blob: be574506e6a93074983f73b9ba62c07a50b4deeb [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001#ifndef _LINUX_RMAP_H
2#define _LINUX_RMAP_H
3/*
4 * Declarations for Reverse Mapping functions in mm/rmap.c
5 */
6
Linus Torvalds1da177e2005-04-16 15:20:36 -07007#include <linux/list.h>
8#include <linux/slab.h>
9#include <linux/mm.h>
Ingo Molnar5a505082012-12-02 19:56:46 +000010#include <linux/rwsem.h>
Balbir Singhbed71612008-02-07 00:14:01 -080011#include <linux/memcontrol.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070012
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 {
Ingo Molnar5a505082012-12-02 19:56:46 +000028 struct anon_vma *root; /* Root of this anon_vma tree */
29 struct rw_semaphore rwsem; /* W: modification, R: walking the list */
Mel Gorman7f60c212010-05-24 14:32:18 -070030 /*
Peter Zijlstra83813262011-03-22 16:32:48 -070031 * The refcount is taken on an anon_vma when there is no
Mel Gorman7f60c212010-05-24 14:32:18 -070032 * 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 */
Peter Zijlstra83813262011-03-22 16:32:48 -070037 atomic_t refcount;
38
Andrea Arcangeli7906d002008-07-28 15:46:26 -070039 /*
Michel Lespinassebf181b92012-10-08 16:31:39 -070040 * NOTE: the LSB of the rb_root.rb_node is set by
Andrea Arcangeli7906d002008-07-28 15:46:26 -070041 * mm_take_all_locks() _after_ taking the above lock. So the
Michel Lespinassebf181b92012-10-08 16:31:39 -070042 * rb_root must only be read/written after taking the above lock
Andrea Arcangeli7906d002008-07-28 15:46:26 -070043 * 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 */
Michel Lespinassebf181b92012-10-08 16:31:39 -070047 struct rb_root rb_root; /* Interval tree of private "related" vmas */
Rik van Riel5beb4932010-03-05 13:42:07 -080048};
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.
Michel Lespinassebf181b92012-10-08 16:31:39 -070060 * The "rb" field indexes on an interval tree the anon_vma_chains
Rik van Riel5beb4932010-03-05 13:42:07 -080061 * which link all the VMAs associated with this anon_vma.
62 */
63struct 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 */
Ingo Molnar5a505082012-12-02 19:56:46 +000067 struct rb_node rb; /* locked by anon_vma->rwsem */
Michel Lespinassebf181b92012-10-08 16:31:39 -070068 unsigned long rb_subtree_last;
Michel Lespinasseed8ea812012-10-08 16:31:45 -070069#ifdef CONFIG_DEBUG_VM_RB
70 unsigned long cached_vma_start, cached_vma_last;
71#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -070072};
73
Minchan Kim02c6de82012-10-08 16:31:55 -070074enum ttu_flags {
Konstantin Khlebnikovdaa5ba72014-06-04 16:10:52 -070075 TTU_UNMAP = 1, /* unmap mode */
76 TTU_MIGRATION = 2, /* migration mode */
77 TTU_MUNLOCK = 4, /* munlock mode */
Minchan Kim02c6de82012-10-08 16:31:55 -070078
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
Linus Torvalds1da177e2005-04-16 15:20:36 -070084#ifdef CONFIG_MMU
Rik van Riel76545062010-08-09 17:18:41 -070085static inline void get_anon_vma(struct anon_vma *anon_vma)
86{
Peter Zijlstra83813262011-03-22 16:32:48 -070087 atomic_inc(&anon_vma->refcount);
Rik van Riel76545062010-08-09 17:18:41 -070088}
89
Peter Zijlstra01d8b202011-03-22 16:32:49 -070090void __put_anon_vma(struct anon_vma *anon_vma);
91
92static 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}
Linus Torvalds1da177e2005-04-16 15:20:36 -070097
Hugh Dickins3ca7b3c2009-12-14 17:58:57 -080098static 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
Rik van Rielbb4a3402010-08-09 17:18:37 -0700106static inline void vma_lock_anon_vma(struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107{
108 struct anon_vma *anon_vma = vma->anon_vma;
109 if (anon_vma)
Ingo Molnar5a505082012-12-02 19:56:46 +0000110 down_write(&anon_vma->root->rwsem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111}
112
Rik van Rielbb4a3402010-08-09 17:18:37 -0700113static inline void vma_unlock_anon_vma(struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114{
115 struct anon_vma *anon_vma = vma->anon_vma;
116 if (anon_vma)
Ingo Molnar5a505082012-12-02 19:56:46 +0000117 up_write(&anon_vma->root->rwsem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700118}
119
Ingo Molnar4fc3f1d2012-12-02 19:56:50 +0000120static inline void anon_vma_lock_write(struct anon_vma *anon_vma)
Rik van Rielcba48b92010-08-09 17:18:38 -0700121{
Ingo Molnar5a505082012-12-02 19:56:46 +0000122 down_write(&anon_vma->root->rwsem);
Rik van Rielcba48b92010-08-09 17:18:38 -0700123}
124
Konstantin Khlebnikov08b52702013-02-22 16:34:40 -0800125static inline void anon_vma_unlock_write(struct anon_vma *anon_vma)
Rik van Rielcba48b92010-08-09 17:18:38 -0700126{
Ingo Molnar5a505082012-12-02 19:56:46 +0000127 up_write(&anon_vma->root->rwsem);
Rik van Rielcba48b92010-08-09 17:18:38 -0700128}
129
Ingo Molnar4fc3f1d2012-12-02 19:56:50 +0000130static inline void anon_vma_lock_read(struct anon_vma *anon_vma)
131{
132 down_read(&anon_vma->root->rwsem);
133}
134
135static inline void anon_vma_unlock_read(struct anon_vma *anon_vma)
136{
137 up_read(&anon_vma->root->rwsem);
138}
139
140
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141/*
142 * anon_vma helper functions.
143 */
144void anon_vma_init(void); /* create anon_vma_cachep */
145int anon_vma_prepare(struct vm_area_struct *);
Rik van Riel5beb4932010-03-05 13:42:07 -0800146void unlink_anon_vmas(struct vm_area_struct *);
147int anon_vma_clone(struct vm_area_struct *, struct vm_area_struct *);
148int anon_vma_fork(struct vm_area_struct *, struct vm_area_struct *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149
Rik van Riel5beb4932010-03-05 13:42:07 -0800150static 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
Peter Zijlstra01d8b202011-03-22 16:32:49 -0700157struct anon_vma *page_get_anon_vma(struct page *page);
158
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159/*
160 * rmap interfaces called when adding or removing pte of page
161 */
Rik van Rielc44b6742010-03-05 13:42:09 -0800162void page_move_anon_rmap(struct page *, struct vm_area_struct *, unsigned long);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700163void page_add_anon_rmap(struct page *, struct vm_area_struct *, unsigned long);
Rik van Rielad8c2ee2010-08-09 17:19:48 -0700164void do_page_add_anon_rmap(struct page *, struct vm_area_struct *,
165 unsigned long, int);
Nick Piggin9617d952006-01-06 00:11:12 -0800166void page_add_new_anon_rmap(struct page *, struct vm_area_struct *, unsigned long);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167void page_add_file_rmap(struct page *);
Hugh Dickinsedc315f2009-01-06 14:40:11 -0800168void page_remove_rmap(struct page *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169
Naoya Horiguchi0fe6e202010-05-28 09:29:16 +0900170void hugepage_add_anon_rmap(struct page *, struct vm_area_struct *,
171 unsigned long);
172void hugepage_add_new_anon_rmap(struct page *, struct vm_area_struct *,
173 unsigned long);
174
Hugh Dickins21333b22009-09-21 17:01:59 -0700175static inline void page_dup_rmap(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700176{
177 atomic_inc(&page->_mapcount);
178}
179
180/*
181 * Called from mm/vmscan.c to handle paging out
182 */
Wu Fengguang6fe6b7e2009-06-16 15:33:05 -0700183int page_referenced(struct page *, int is_locked,
Johannes Weiner72835c82012-01-12 17:18:32 -0800184 struct mem_cgroup *memcg, unsigned long *vm_flags);
Hugh Dickins5ad64682009-12-14 17:59:24 -0800185
Andi Kleen14fa31b2009-09-16 11:50:10 +0200186#define TTU_ACTION(x) ((x) & TTU_ACTION_MASK)
187
188int try_to_unmap(struct page *, enum ttu_flags flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189
190/*
Carsten Otteceffc072005-06-23 22:05:25 -0700191 * Called from mm/filemap_xip.c to unmap empty zero page
192 */
Namhyung Kime9a81a82010-10-26 14:22:01 -0700193pte_t *__page_check_address(struct page *, struct mm_struct *,
Nick Piggin479db0b2008-08-20 14:09:18 -0700194 unsigned long, spinlock_t **, int);
Carsten Otteceffc072005-06-23 22:05:25 -0700195
Namhyung Kime9a81a82010-10-26 14:22:01 -0700196static 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
Carsten Otteceffc072005-06-23 22:05:25 -0700207/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208 * Used by swapoff to help locate where page is expected in vma.
209 */
210unsigned long page_address_in_vma(struct page *, struct vm_area_struct *);
211
Peter Zijlstrad08b3852006-09-25 23:30:57 -0700212/*
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 */
218int page_mkclean(struct page *);
219
Nick Pigginb291f002008-10-18 20:26:44 -0700220/*
221 * called in munlock()/munmap() path to check for other vmas holding
222 * the page mlocked.
223 */
224int try_to_munlock(struct page *);
Nick Pigginb291f002008-10-18 20:26:44 -0700225
Andi Kleen10be22d2009-09-16 11:50:04 +0200226/*
227 * Called by memory-failure.c to kill processes.
228 */
Ingo Molnar4fc3f1d2012-12-02 19:56:50 +0000229struct anon_vma *page_lock_anon_vma_read(struct page *page);
230void page_unlock_anon_vma_read(struct anon_vma *anon_vma);
Andi Kleen6a460792009-09-16 11:50:15 +0200231int page_mapped_in_vma(struct page *page, struct vm_area_struct *vma);
Andi Kleen10be22d2009-09-16 11:50:04 +0200232
Joonsoo Kim0dd1c7b2014-01-21 15:49:49 -0800233/*
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 */
Joonsoo Kim051ac832014-01-21 15:49:48 -0800243struct rmap_walk_control {
244 void *arg;
245 int (*rmap_one)(struct page *page, struct vm_area_struct *vma,
246 unsigned long addr, void *arg);
Joonsoo Kim0dd1c7b2014-01-21 15:49:49 -0800247 int (*done)(struct page *page);
Hugh Dickins7e09e732014-03-20 21:52:17 -0700248 int (*file_nonlinear)(struct page *, struct address_space *, void *arg);
Joonsoo Kim0dd1c7b2014-01-21 15:49:49 -0800249 struct anon_vma *(*anon_lock)(struct page *page);
250 bool (*invalid_vma)(struct vm_area_struct *vma, void *arg);
Joonsoo Kim051ac832014-01-21 15:49:48 -0800251};
252
Joonsoo Kim051ac832014-01-21 15:49:48 -0800253int rmap_walk(struct page *page, struct rmap_walk_control *rwc);
Hugh Dickinse9995ef2009-12-14 17:59:31 -0800254
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255#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
Mike Frysinger01ff53f2009-06-23 12:37:01 -0700261static inline int page_referenced(struct page *page, int is_locked,
Johannes Weiner72835c82012-01-12 17:18:32 -0800262 struct mem_cgroup *memcg,
Mike Frysinger01ff53f2009-06-23 12:37:01 -0700263 unsigned long *vm_flags)
264{
265 *vm_flags = 0;
Johannes Weiner645747462010-03-05 13:42:22 -0800266 return 0;
Mike Frysinger01ff53f2009-06-23 12:37:01 -0700267}
268
Christoph Lametera48d07a2006-02-01 03:05:38 -0800269#define try_to_unmap(page, refs) SWAP_FAIL
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270
Peter Zijlstrad08b3852006-09-25 23:30:57 -0700271static inline int page_mkclean(struct page *page)
272{
273 return 0;
274}
275
276
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277#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
Nick Pigginb291f002008-10-18 20:26:44 -0700285#define SWAP_MLOCK 3
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286
287#endif /* _LINUX_RMAP_H */