linux/include/linux/rmap.h
<<
>>
Prefs
   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 */
  27struct 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 */
  73struct 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
  84enum 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
  95static inline void get_anon_vma(struct anon_vma *anon_vma)
  96{
  97        atomic_inc(&anon_vma->refcount);
  98}
  99
 100void __put_anon_vma(struct anon_vma *anon_vma);
 101
 102static 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
 108static inline void vma_lock_anon_vma(struct vm_area_struct *vma)
 109{
 110        struct anon_vma *anon_vma = vma->anon_vma;
 111        if (anon_vma)
 112                down_write(&anon_vma->root->rwsem);
 113}
 114
 115static inline void vma_unlock_anon_vma(struct vm_area_struct *vma)
 116{
 117        struct anon_vma *anon_vma = vma->anon_vma;
 118        if (anon_vma)
 119                up_write(&anon_vma->root->rwsem);
 120}
 121
 122static inline void anon_vma_lock_write(struct anon_vma *anon_vma)
 123{
 124        down_write(&anon_vma->root->rwsem);
 125}
 126
 127static inline void anon_vma_unlock_write(struct anon_vma *anon_vma)
 128{
 129        up_write(&anon_vma->root->rwsem);
 130}
 131
 132static inline void anon_vma_lock_read(struct anon_vma *anon_vma)
 133{
 134        down_read(&anon_vma->root->rwsem);
 135}
 136
 137static inline void anon_vma_unlock_read(struct anon_vma *anon_vma)
 138{
 139        up_read(&anon_vma->root->rwsem);
 140}
 141
 142
 143/*
 144 * anon_vma helper functions.
 145 */
 146void anon_vma_init(void);       /* create anon_vma_cachep */
 147int  anon_vma_prepare(struct vm_area_struct *);
 148void unlink_anon_vmas(struct vm_area_struct *);
 149int anon_vma_clone(struct vm_area_struct *, struct vm_area_struct *);
 150int anon_vma_fork(struct vm_area_struct *, struct vm_area_struct *);
 151
 152static inline void anon_vma_merge(struct vm_area_struct *vma,
 153                                  struct vm_area_struct *next)
 154{
 155        VM_BUG_ON_VMA(vma->anon_vma != next->anon_vma, vma);
 156        unlink_anon_vmas(next);
 157}
 158
 159struct anon_vma *page_get_anon_vma(struct page *page);
 160
 161/*
 162 * rmap interfaces called when adding or removing pte of page
 163 */
 164void page_move_anon_rmap(struct page *, struct vm_area_struct *, unsigned long);
 165void page_add_anon_rmap(struct page *, struct vm_area_struct *, unsigned long);
 166void do_page_add_anon_rmap(struct page *, struct vm_area_struct *,
 167                           unsigned long, int);
 168void page_add_new_anon_rmap(struct page *, struct vm_area_struct *, unsigned long);
 169void page_add_file_rmap(struct page *);
 170void page_remove_rmap(struct page *);
 171
 172void hugepage_add_anon_rmap(struct page *, struct vm_area_struct *,
 173                            unsigned long);
 174void hugepage_add_new_anon_rmap(struct page *, struct vm_area_struct *,
 175                                unsigned long);
 176
 177static inline void page_dup_rmap(struct page *page)
 178{
 179        atomic_inc(&page->_mapcount);
 180}
 181
 182/*
 183 * Called from mm/vmscan.c to handle paging out
 184 */
 185int page_referenced(struct page *, int is_locked,
 186                        struct mem_cgroup *memcg, unsigned long *vm_flags);
 187
 188#define TTU_ACTION(x) ((x) & TTU_ACTION_MASK)
 189
 190int try_to_unmap(struct page *, enum ttu_flags flags);
 191
 192/*
 193 * Used by uprobes to replace a userspace page safely
 194 */
 195pte_t *__page_check_address(struct page *, struct mm_struct *,
 196                                unsigned long, spinlock_t **, int);
 197
 198static inline pte_t *page_check_address(struct page *page, struct mm_struct *mm,
 199                                        unsigned long address,
 200                                        spinlock_t **ptlp, int sync)
 201{
 202        pte_t *ptep;
 203
 204        __cond_lock(*ptlp, ptep = __page_check_address(page, mm, address,
 205                                                       ptlp, sync));
 206        return ptep;
 207}
 208
 209/*
 210 * Used by swapoff to help locate where page is expected in vma.
 211 */
 212unsigned long page_address_in_vma(struct page *, struct vm_area_struct *);
 213
 214/*
 215 * Cleans the PTEs of shared mappings.
 216 * (and since clean PTEs should also be readonly, write protects them too)
 217 *
 218 * returns the number of cleaned PTEs.
 219 */
 220int page_mkclean(struct page *);
 221
 222/*
 223 * called in munlock()/munmap() path to check for other vmas holding
 224 * the page mlocked.
 225 */
 226int try_to_munlock(struct page *);
 227
 228/*
 229 * Called by memory-failure.c to kill processes.
 230 */
 231struct anon_vma *page_lock_anon_vma_read(struct page *page);
 232void page_unlock_anon_vma_read(struct anon_vma *anon_vma);
 233int page_mapped_in_vma(struct page *page, struct vm_area_struct *vma);
 234
 235/*
 236 * rmap_walk_control: To control rmap traversing for specific needs
 237 *
 238 * arg: passed to rmap_one() and invalid_vma()
 239 * rmap_one: executed on each vma where page is mapped
 240 * done: for checking traversing termination condition
 241 * anon_lock: for getting anon_lock by optimized way rather than default
 242 * invalid_vma: for skipping uninterested vma
 243 */
 244struct rmap_walk_control {
 245        void *arg;
 246        int (*rmap_one)(struct page *page, struct vm_area_struct *vma,
 247                                        unsigned long addr, void *arg);
 248        int (*done)(struct page *page);
 249        struct anon_vma *(*anon_lock)(struct page *page);
 250        bool (*invalid_vma)(struct vm_area_struct *vma, void *arg);
 251};
 252
 253int 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
 261static 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
 271static 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 */
 288