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          * 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         TTU_BATCH_FLUSH = (1 << 11),    /* Batch TLB flushes where possible
  93                                          * and caller guarantees they will
  94                                          * do a final flush if necessary */
  95 };
  96
  97 #ifdef CONFIG_MMU
  98 static inline void get_anon_vma(struct anon_vma *anon_vma)
  99 {
 100         atomic_inc(&anon_vma->refcount);
 101 }
 102
 103 void __put_anon_vma(struct anon_vma *anon_vma);
 104
 105 static inline void put_anon_vma(struct anon_vma *anon_vma)
 106 {
 107         if (atomic_dec_and_test(&anon_vma->refcount))
 108                 __put_anon_vma(anon_vma);
 109 }
 110
 111 static inline void vma_lock_anon_vma(struct vm_area_struct *vma)
 112 {
 113         struct anon_vma *anon_vma = vma->anon_vma;
 114         if (anon_vma)
 115                 down_write(&anon_vma->root->rwsem);
 116 }
 117
 118 static inline void vma_unlock_anon_vma(struct vm_area_struct *vma)
 119 {
 120         struct anon_vma *anon_vma = vma->anon_vma;
 121         if (anon_vma)
 122                 up_write(&anon_vma->root->rwsem);
 123 }
 124
 125 static inline void anon_vma_lock_write(struct anon_vma *anon_vma)
 126 {
 127         down_write(&anon_vma->root->rwsem);
 128 }
 129
 130 static inline void anon_vma_unlock_write(struct anon_vma *anon_vma)
 131 {
 132         up_write(&anon_vma->root->rwsem);
 133 }
 134
 135 static inline void anon_vma_lock_read(struct anon_vma *anon_vma)
 136 {
 137         down_read(&anon_vma->root->rwsem);
 138 }
 139
 140 static inline void anon_vma_unlock_read(struct anon_vma *anon_vma)
 141 {
 142         up_read(&anon_vma->root->rwsem);
 143 }
 144
 145
 146 /*
 147  * anon_vma helper functions.
 148  */
 149 void anon_vma_init(void);       /* create anon_vma_cachep */
 150 int  anon_vma_prepare(struct vm_area_struct *);
 151 void unlink_anon_vmas(struct vm_area_struct *);
 152 int anon_vma_clone(struct vm_area_struct *, struct vm_area_struct *);
 153 int anon_vma_fork(struct vm_area_struct *, struct vm_area_struct *);
 154
 155 static inline void anon_vma_merge(struct vm_area_struct *vma,
 156                                   struct vm_area_struct *next)
 157 {
 158         VM_BUG_ON_VMA(vma->anon_vma != next->anon_vma, vma);
 159         unlink_anon_vmas(next);
 160 }
 161
 162 struct anon_vma *page_get_anon_vma(struct page *page);
 163
 164 /*
 165  * rmap interfaces called when adding or removing pte of page
 166  */
 167 void page_move_anon_rmap(struct page *, struct vm_area_struct *, unsigned long);
 168 void page_add_anon_rmap(struct page *, struct vm_area_struct *, unsigned long);
 169 void do_page_add_anon_rmap(struct page *, struct vm_area_struct *,
 170                            unsigned long, int);
 171 void page_add_new_anon_rmap(struct page *, struct vm_area_struct *, unsigned long);
 172 void page_add_file_rmap(struct page *);
 173 void page_remove_rmap(struct page *);
 174
 175 void hugepage_add_anon_rmap(struct page *, struct vm_area_struct *,
 176                             unsigned long);
 177 void hugepage_add_new_anon_rmap(struct page *, struct vm_area_struct *,
 178                                 unsigned long);
 179
 180 static inline void page_dup_rmap(struct page *page)
 181 {
 182         atomic_inc(&page->_mapcount);
 183 }
 184
 185 /*
 186  * Called from mm/vmscan.c to handle paging out
 187  */
 188 int page_referenced(struct page *, int is_locked,
 189                         struct mem_cgroup *memcg, unsigned long *vm_flags);
 190
 191 #define TTU_ACTION(x) ((x) & TTU_ACTION_MASK)
 192
 193 int try_to_unmap(struct page *, enum ttu_flags flags);
 194
 195 /*
 196  * Used by uprobes to replace a userspace page safely
 197  */
 198 pte_t *__page_check_address(struct page *, struct mm_struct *,
 199                                 unsigned long, spinlock_t **, int);
 200
 201 static inline pte_t *page_check_address(struct page *page, struct mm_struct *mm,
 202                                         unsigned long address,
 203                                         spinlock_t **ptlp, int sync)
 204 {
 205         pte_t *ptep;
 206
 207         __cond_lock(*ptlp, ptep = __page_check_address(page, mm, address,
 208                                                        ptlp, sync));
 209         return ptep;
 210 }
 211
 212 /*
 213  * Used by swapoff to help locate where page is expected in vma.
 214  */
 215 unsigned long page_address_in_vma(struct page *, struct vm_area_struct *);
 216
 217 /*
 218  * Cleans the PTEs of shared mappings.
 219  * (and since clean PTEs should also be readonly, write protects them too)
 220  *
 221  * returns the number of cleaned PTEs.
 222  */
 223 int page_mkclean(struct page *);
 224
 225 /*
 226  * called in munlock()/munmap() path to check for other vmas holding
 227  * the page mlocked.
 228  */
 229 int try_to_munlock(struct page *);
 230
 231 /*
 232  * Called by memory-failure.c to kill processes.
 233  */
 234 struct anon_vma *page_lock_anon_vma_read(struct page *page);
 235 void page_unlock_anon_vma_read(struct anon_vma *anon_vma);
 236 int page_mapped_in_vma(struct page *page, struct vm_area_struct *vma);
 237
 238 /*
 239  * rmap_walk_control: To control rmap traversing for specific needs
 240  *
 241  * arg: passed to rmap_one() and invalid_vma()
 242  * rmap_one: executed on each vma where page is mapped
 243  * done: for checking traversing termination condition
 244  * anon_lock: for getting anon_lock by optimized way rather than default
 245  * invalid_vma: for skipping uninterested vma
 246  */
 247 struct rmap_walk_control {
 248         void *arg;
 249         int (*rmap_one)(struct page *page, struct vm_area_struct *vma,
 250                                         unsigned long addr, void *arg);
 251         int (*done)(struct page *page);
 252         struct anon_vma *(*anon_lock)(struct page *page);
 253         bool (*invalid_vma)(struct vm_area_struct *vma, void *arg);
 254 };
 255
 256 int rmap_walk(struct page *page, struct rmap_walk_control *rwc);
 257
 258 #else   /* !CONFIG_MMU */
 259
 260 #define anon_vma_init()         do {} while (0)
 261 #define anon_vma_prepare(vma)   (0)
 262 #define anon_vma_link(vma)      do {} while (0)
 263
 264 static inline int page_referenced(struct page *page, int is_locked,
 265                                   struct mem_cgroup *memcg,
 266                                   unsigned long *vm_flags)
 267 {
 268         *vm_flags = 0;
 269         return 0;
 270 }
 271
 272 #define try_to_unmap(page, refs) SWAP_FAIL
 273
 274 static inline int page_mkclean(struct page *page)
 275 {
 276         return 0;
 277 }
 278
 279
 280 #endif  /* CONFIG_MMU */
 281
 282 /*
 283  * Return values of try_to_unmap
 284  */
 285 #define SWAP_SUCCESS    0
 286 #define SWAP_AGAIN      1
 287 #define SWAP_FAIL       2
 288 #define SWAP_MLOCK      3
 289
 290 #endif  /* _LINUX_RMAP_H */