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         * 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         * Count of child anon_vmas and VMAs which points to this anon_vma.
  51         *
  52         * This counter is used for making decision about reusing anon_vma
  53         * instead of forking new one. See comments in function anon_vma_clone.
  54         */
  55        RH_KABI_EXTEND(unsigned degree)
  56
  57        RH_KABI_EXTEND(struct anon_vma *parent) /* Parent of this anon_vma */
  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        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
  99static inline void get_anon_vma(struct anon_vma *anon_vma)
 100{
 101        atomic_inc(&anon_vma->refcount);
 102}
 103
 104void __put_anon_vma(struct anon_vma *anon_vma);
 105
 106static 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
 112static inline struct anon_vma *page_anon_vma(struct page *page)
 113{
 114        if (((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) !=
 115                                            PAGE_MAPPING_ANON)
 116                return NULL;
 117        return page_rmapping(page);
 118}
 119
 120static inline void vma_lock_anon_vma(struct vm_area_struct *vma)
 121{
 122        struct anon_vma *anon_vma = vma->anon_vma;
 123        if (anon_vma)
 124                down_write(&anon_vma->root->rwsem);
 125}
 126
 127static inline void vma_unlock_anon_vma(struct vm_area_struct *vma)
 128{
 129        struct anon_vma *anon_vma = vma->anon_vma;
 130        if (anon_vma)
 131                up_write(&anon_vma->root->rwsem);
 132}
 133
 134static inline void anon_vma_lock_write(struct anon_vma *anon_vma)
 135{
 136        down_write(&anon_vma->root->rwsem);
 137}
 138
 139static inline void anon_vma_unlock_write(struct anon_vma *anon_vma)
 140{
 141        up_write(&anon_vma->root->rwsem);
 142}
 143
 144static inline void anon_vma_lock_read(struct anon_vma *anon_vma)
 145{
 146        down_read(&anon_vma->root->rwsem);
 147}
 148
 149static inline void anon_vma_unlock_read(struct anon_vma *anon_vma)
 150{
 151        up_read(&anon_vma->root->rwsem);
 152}
 153
 154
 155/*
 156 * anon_vma helper functions.
 157 */
 158void anon_vma_init(void);       /* create anon_vma_cachep */
 159int  anon_vma_prepare(struct vm_area_struct *);
 160void unlink_anon_vmas(struct vm_area_struct *);
 161int anon_vma_clone(struct vm_area_struct *, struct vm_area_struct *);
 162int anon_vma_fork(struct vm_area_struct *, struct vm_area_struct *);
 163
 164static inline void anon_vma_merge(struct vm_area_struct *vma,
 165                                  struct vm_area_struct *next)
 166{
 167        VM_BUG_ON(vma->anon_vma != next->anon_vma);
 168        unlink_anon_vmas(next);
 169}
 170
 171struct anon_vma *page_get_anon_vma(struct page *page);
 172
 173/*
 174 * rmap interfaces called when adding or removing pte of page
 175 */
 176void page_move_anon_rmap(struct page *, struct vm_area_struct *, unsigned long);
 177void page_add_anon_rmap(struct page *, struct vm_area_struct *, unsigned long);
 178void do_page_add_anon_rmap(struct page *, struct vm_area_struct *,
 179                           unsigned long, int);
 180void page_add_new_anon_rmap(struct page *, struct vm_area_struct *, unsigned long);
 181void page_add_file_rmap(struct page *);
 182void page_remove_rmap(struct page *);
 183
 184void hugepage_add_anon_rmap(struct page *, struct vm_area_struct *,
 185                            unsigned long);
 186void hugepage_add_new_anon_rmap(struct page *, struct vm_area_struct *,
 187                                unsigned long);
 188
 189static inline void page_dup_rmap(struct page *page)
 190{
 191        atomic_inc(&page->_mapcount);
 192}
 193
 194/*
 195 * Called from mm/vmscan.c to handle paging out
 196 */
 197int page_referenced(struct page *, int is_locked,
 198                        struct mem_cgroup *memcg, unsigned long *vm_flags);
 199int page_referenced_one(struct page *, struct vm_area_struct *,
 200        unsigned long address, void *arg);
 201
 202int try_to_unmap(struct page *, enum ttu_flags flags);
 203int try_to_unmap_one(struct page *, struct vm_area_struct *,
 204                        unsigned long address, void *arg);
 205
 206/*
 207 * Used by uprobes to replace a userspace page safely
 208 */
 209pte_t *__page_check_address(struct page *, struct mm_struct *,
 210                                unsigned long, spinlock_t **, int);
 211
 212static inline pte_t *page_check_address(struct page *page, struct mm_struct *mm,
 213                                        unsigned long address,
 214                                        spinlock_t **ptlp, int sync)
 215{
 216        pte_t *ptep;
 217
 218        __cond_lock(*ptlp, ptep = __page_check_address(page, mm, address,
 219                                                       ptlp, sync));
 220        return ptep;
 221}
 222
 223/*
 224 * Used by swapoff to help locate where page is expected in vma.
 225 */
 226unsigned long page_address_in_vma(struct page *, struct vm_area_struct *);
 227
 228/*
 229 * Cleans the PTEs of shared mappings.
 230 * (and since clean PTEs should also be readonly, write protects them too)
 231 *
 232 * returns the number of cleaned PTEs.
 233 */
 234int page_mkclean(struct page *);
 235
 236/*
 237 * called in munlock()/munmap() path to check for other vmas holding
 238 * the page mlocked.
 239 */
 240int try_to_munlock(struct page *);
 241
 242/*
 243 * Called by memory-failure.c to kill processes.
 244 */
 245struct anon_vma *page_lock_anon_vma_read(struct page *page);
 246void page_unlock_anon_vma_read(struct anon_vma *anon_vma);
 247int page_mapped_in_vma(struct page *page, struct vm_area_struct *vma);
 248
 249/*
 250 * rmap_walk_control: To control rmap traversing for specific needs
 251 *
 252 * arg: passed to rmap_one() and invalid_vma()
 253 * rmap_one: executed on each vma where page is mapped
 254 * done: for checking traversing termination condition
 255 * file_nonlinear: for handling file nonlinear mapping
 256 * anon_lock: for getting anon_lock by optimized way rather than default
 257 * invalid_vma: for skipping uninterested vma
 258 */
 259struct 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        int (*file_nonlinear)(struct page *, struct address_space *,
 265                                        struct vm_area_struct *vma);
 266        struct anon_vma *(*anon_lock)(struct page *page);
 267        bool (*invalid_vma)(struct vm_area_struct *vma, void *arg);
 268};
 269
 270int rmap_walk(struct page *page, struct rmap_walk_control *rwc);
 271
 272#else   /* !CONFIG_MMU */
 273
 274#define anon_vma_init()         do {} while (0)
 275#define anon_vma_prepare(vma)   (0)
 276#define anon_vma_link(vma)      do {} while (0)
 277
 278static inline int page_referenced(struct page *page, int is_locked,
 279                                  struct mem_cgroup *memcg,
 280                                  unsigned long *vm_flags)
 281{
 282        *vm_flags = 0;
 283        return 0;
 284}
 285
 286#define try_to_unmap(page, refs) SWAP_FAIL
 287
 288static inline int page_mkclean(struct page *page)
 289{
 290        return 0;
 291}
 292
 293
 294#endif  /* CONFIG_MMU */
 295
 296/*
 297 * Return values of try_to_unmap
 298 */
 299#define SWAP_SUCCESS    0
 300#define SWAP_AGAIN      1
 301#define SWAP_FAIL       2
 302#define SWAP_MLOCK      3
 303#define SWAP_LZFREE     4
 304
 305#endif  /* _LINUX_RMAP_H */
 306