1 ======= 1 =======
2 Locking 2 Locking
3 ======= 3 =======
4 4
5 The text below describes the locking rules for 5 The text below describes the locking rules for VFS-related methods.
6 It is (believed to be) up-to-date. *Please*, i 6 It is (believed to be) up-to-date. *Please*, if you change anything in
7 prototypes or locking protocols - update this 7 prototypes or locking protocols - update this file. And update the relevant
8 instances in the tree, don't leave that to mai 8 instances in the tree, don't leave that to maintainers of filesystems/devices/
9 etc. At the very least, put the list of dubiou 9 etc. At the very least, put the list of dubious cases in the end of this file.
10 Don't turn it into log - maintainers of out-of 10 Don't turn it into log - maintainers of out-of-the-tree code are supposed to
11 be able to use diff(1). 11 be able to use diff(1).
12 12
13 Thing currently missing here: socket operation 13 Thing currently missing here: socket operations. Alexey?
14 14
15 dentry_operations 15 dentry_operations
16 ================= 16 =================
17 17
18 prototypes:: 18 prototypes::
19 19
20 int (*d_revalidate)(struct dentry *, u 20 int (*d_revalidate)(struct dentry *, unsigned int);
21 int (*d_weak_revalidate)(struct dentry 21 int (*d_weak_revalidate)(struct dentry *, unsigned int);
22 int (*d_hash)(const struct dentry *, s 22 int (*d_hash)(const struct dentry *, struct qstr *);
23 int (*d_compare)(const struct dentry * 23 int (*d_compare)(const struct dentry *,
24 unsigned int, const ch 24 unsigned int, const char *, const struct qstr *);
25 int (*d_delete)(struct dentry *); 25 int (*d_delete)(struct dentry *);
26 int (*d_init)(struct dentry *); 26 int (*d_init)(struct dentry *);
27 void (*d_release)(struct dentry *); 27 void (*d_release)(struct dentry *);
28 void (*d_iput)(struct dentry *, struct 28 void (*d_iput)(struct dentry *, struct inode *);
29 char *(*d_dname)((struct dentry *dentr 29 char *(*d_dname)((struct dentry *dentry, char *buffer, int buflen);
30 struct vfsmount *(*d_automount)(struct 30 struct vfsmount *(*d_automount)(struct path *path);
31 int (*d_manage)(const struct path *, b 31 int (*d_manage)(const struct path *, bool);
32 struct dentry *(*d_real)(struct dentry !! 32 struct dentry *(*d_real)(struct dentry *, const struct inode *);
33 33
34 locking rules: 34 locking rules:
35 35
36 ================== =========== ======== 36 ================== =========== ======== ============== ========
37 ops rename_lock ->d_lock 37 ops rename_lock ->d_lock may block rcu-walk
38 ================== =========== ======== 38 ================== =========== ======== ============== ========
39 d_revalidate: no no 39 d_revalidate: no no yes (ref-walk) maybe
40 d_weak_revalidate: no no 40 d_weak_revalidate: no no yes no
41 d_hash no no 41 d_hash no no no maybe
42 d_compare: yes no 42 d_compare: yes no no maybe
43 d_delete: no yes 43 d_delete: no yes no no
44 d_init: no no 44 d_init: no no yes no
45 d_release: no no 45 d_release: no no yes no
46 d_prune: no yes 46 d_prune: no yes no no
47 d_iput: no no 47 d_iput: no no yes no
48 d_dname: no no 48 d_dname: no no no no
49 d_automount: no no 49 d_automount: no no yes no
50 d_manage: no no 50 d_manage: no no yes (ref-walk) maybe
51 d_real no no 51 d_real no no yes no
52 ================== =========== ======== 52 ================== =========== ======== ============== ========
53 53
54 inode_operations 54 inode_operations
55 ================ 55 ================
56 56
57 prototypes:: 57 prototypes::
58 58
59 int (*create) (struct mnt_idmap *, str !! 59 int (*create) (struct inode *,struct dentry *,umode_t, bool);
60 struct dentry * (*lookup) (struct inod 60 struct dentry * (*lookup) (struct inode *,struct dentry *, unsigned int);
61 int (*link) (struct dentry *,struct in 61 int (*link) (struct dentry *,struct inode *,struct dentry *);
62 int (*unlink) (struct inode *,struct d 62 int (*unlink) (struct inode *,struct dentry *);
63 int (*symlink) (struct mnt_idmap *, st !! 63 int (*symlink) (struct inode *,struct dentry *,const char *);
64 int (*mkdir) (struct mnt_idmap *, stru !! 64 int (*mkdir) (struct inode *,struct dentry *,umode_t);
65 int (*rmdir) (struct inode *,struct de 65 int (*rmdir) (struct inode *,struct dentry *);
66 int (*mknod) (struct mnt_idmap *, stru !! 66 int (*mknod) (struct inode *,struct dentry *,umode_t,dev_t);
67 int (*rename) (struct mnt_idmap *, str !! 67 int (*rename) (struct inode *, struct dentry *,
68 struct inode *, struct 68 struct inode *, struct dentry *, unsigned int);
69 int (*readlink) (struct dentry *, char 69 int (*readlink) (struct dentry *, char __user *,int);
70 const char *(*get_link) (struct dentry 70 const char *(*get_link) (struct dentry *, struct inode *, struct delayed_call *);
71 void (*truncate) (struct inode *); 71 void (*truncate) (struct inode *);
72 int (*permission) (struct mnt_idmap *, !! 72 int (*permission) (struct inode *, int, unsigned int);
73 struct posix_acl * (*get_inode_acl)(st !! 73 struct posix_acl * (*get_acl)(struct inode *, int, bool);
74 int (*setattr) (struct mnt_idmap *, st !! 74 int (*setattr) (struct dentry *, struct iattr *);
75 int (*getattr) (struct mnt_idmap *, co !! 75 int (*getattr) (const struct path *, struct kstat *, u32, unsigned int);
76 ssize_t (*listxattr) (struct dentry *, 76 ssize_t (*listxattr) (struct dentry *, char *, size_t);
77 int (*fiemap)(struct inode *, struct f 77 int (*fiemap)(struct inode *, struct fiemap_extent_info *, u64 start, u64 len);
78 void (*update_time)(struct inode *, st 78 void (*update_time)(struct inode *, struct timespec *, int);
79 int (*atomic_open)(struct inode *, str 79 int (*atomic_open)(struct inode *, struct dentry *,
80 struct file *, 80 struct file *, unsigned open_flag,
81 umode_t create 81 umode_t create_mode);
82 int (*tmpfile) (struct mnt_idmap *, st !! 82 int (*tmpfile) (struct inode *, struct dentry *, umode_t);
83 struct file *, umode_t !! 83 int (*fileattr_set)(struct user_namespace *mnt_userns,
84 int (*fileattr_set)(struct mnt_idmap * <<
85 struct dentry *den 84 struct dentry *dentry, struct fileattr *fa);
86 int (*fileattr_get)(struct dentry *den 85 int (*fileattr_get)(struct dentry *dentry, struct fileattr *fa);
87 struct posix_acl * (*get_acl)(struct m <<
88 struct offset_ctx *(*get_offset_ctx)(s <<
89 86
90 locking rules: 87 locking rules:
91 all may block 88 all may block
92 89
93 ============== ============================== !! 90 ============= =============================================
94 ops i_rwsem(inode) 91 ops i_rwsem(inode)
95 ============== ============================== !! 92 ============= =============================================
96 lookup: shared 93 lookup: shared
97 create: exclusive 94 create: exclusive
98 link: exclusive (both) 95 link: exclusive (both)
99 mknod: exclusive 96 mknod: exclusive
100 symlink: exclusive 97 symlink: exclusive
101 mkdir: exclusive 98 mkdir: exclusive
102 unlink: exclusive (both) 99 unlink: exclusive (both)
103 rmdir: exclusive (both)(see below) 100 rmdir: exclusive (both)(see below)
104 rename: exclusive (both parents, some 101 rename: exclusive (both parents, some children) (see below)
105 readlink: no 102 readlink: no
106 get_link: no 103 get_link: no
107 setattr: exclusive 104 setattr: exclusive
108 permission: no (may not block if called in 105 permission: no (may not block if called in rcu-walk mode)
109 get_inode_acl: no <<
110 get_acl: no 106 get_acl: no
111 getattr: no 107 getattr: no
112 listxattr: no 108 listxattr: no
113 fiemap: no 109 fiemap: no
114 update_time: no 110 update_time: no
115 atomic_open: shared (exclusive if O_CREAT i 111 atomic_open: shared (exclusive if O_CREAT is set in open flags)
116 tmpfile: no 112 tmpfile: no
117 fileattr_get: no or exclusive 113 fileattr_get: no or exclusive
118 fileattr_set: exclusive 114 fileattr_set: exclusive
119 get_offset_ctx no !! 115 ============= =============================================
120 ============== ============================== <<
121 116
122 117
123 Additionally, ->rmdir(), ->unlink() an 118 Additionally, ->rmdir(), ->unlink() and ->rename() have ->i_rwsem
124 exclusive on victim. 119 exclusive on victim.
125 cross-directory ->rename() has (per-su 120 cross-directory ->rename() has (per-superblock) ->s_vfs_rename_sem.
126 ->unlink() and ->rename() have ->i_rws 121 ->unlink() and ->rename() have ->i_rwsem exclusive on all non-directories
127 involved. 122 involved.
128 ->rename() has ->i_rwsem exclusive on 123 ->rename() has ->i_rwsem exclusive on any subdirectory that changes parent.
129 124
130 See Documentation/filesystems/directory-lockin 125 See Documentation/filesystems/directory-locking.rst for more detailed discussion
131 of the locking scheme for directory operations 126 of the locking scheme for directory operations.
132 127
133 xattr_handler operations 128 xattr_handler operations
134 ======================== 129 ========================
135 130
136 prototypes:: 131 prototypes::
137 132
138 bool (*list)(struct dentry *dentry); 133 bool (*list)(struct dentry *dentry);
139 int (*get)(const struct xattr_handler 134 int (*get)(const struct xattr_handler *handler, struct dentry *dentry,
140 struct inode *inode, const 135 struct inode *inode, const char *name, void *buffer,
141 size_t size); 136 size_t size);
142 int (*set)(const struct xattr_handler 137 int (*set)(const struct xattr_handler *handler,
143 struct mnt_idmap *idmap, !! 138 struct user_namespace *mnt_userns,
144 struct dentry *dentry, stru 139 struct dentry *dentry, struct inode *inode, const char *name,
145 const void *buffer, size_t 140 const void *buffer, size_t size, int flags);
146 141
147 locking rules: 142 locking rules:
148 all may block 143 all may block
149 144
150 ===== ============== 145 ===== ==============
151 ops i_rwsem(inode) 146 ops i_rwsem(inode)
152 ===== ============== 147 ===== ==============
153 list: no 148 list: no
154 get: no 149 get: no
155 set: exclusive 150 set: exclusive
156 ===== ============== 151 ===== ==============
157 152
158 super_operations 153 super_operations
159 ================ 154 ================
160 155
161 prototypes:: 156 prototypes::
162 157
163 struct inode *(*alloc_inode)(struct su 158 struct inode *(*alloc_inode)(struct super_block *sb);
164 void (*free_inode)(struct inode *); 159 void (*free_inode)(struct inode *);
165 void (*destroy_inode)(struct inode *); 160 void (*destroy_inode)(struct inode *);
166 void (*dirty_inode) (struct inode *, i 161 void (*dirty_inode) (struct inode *, int flags);
167 int (*write_inode) (struct inode *, st 162 int (*write_inode) (struct inode *, struct writeback_control *wbc);
168 int (*drop_inode) (struct inode *); 163 int (*drop_inode) (struct inode *);
169 void (*evict_inode) (struct inode *); 164 void (*evict_inode) (struct inode *);
170 void (*put_super) (struct super_block 165 void (*put_super) (struct super_block *);
171 int (*sync_fs)(struct super_block *sb, 166 int (*sync_fs)(struct super_block *sb, int wait);
172 int (*freeze_fs) (struct super_block * 167 int (*freeze_fs) (struct super_block *);
173 int (*unfreeze_fs) (struct super_block 168 int (*unfreeze_fs) (struct super_block *);
174 int (*statfs) (struct dentry *, struct 169 int (*statfs) (struct dentry *, struct kstatfs *);
175 int (*remount_fs) (struct super_block 170 int (*remount_fs) (struct super_block *, int *, char *);
176 void (*umount_begin) (struct super_blo 171 void (*umount_begin) (struct super_block *);
177 int (*show_options)(struct seq_file *, 172 int (*show_options)(struct seq_file *, struct dentry *);
178 ssize_t (*quota_read)(struct super_blo 173 ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t);
179 ssize_t (*quota_write)(struct super_bl 174 ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t);
>> 175 int (*bdev_try_to_free_page)(struct super_block*, struct page*, gfp_t);
180 176
181 locking rules: 177 locking rules:
182 All may block [not true, see below] 178 All may block [not true, see below]
183 179
184 ====================== ============ ====== 180 ====================== ============ ========================
185 ops s_umount note 181 ops s_umount note
186 ====================== ============ ====== 182 ====================== ============ ========================
187 alloc_inode: 183 alloc_inode:
188 free_inode: called 184 free_inode: called from RCU callback
189 destroy_inode: 185 destroy_inode:
190 dirty_inode: 186 dirty_inode:
191 write_inode: 187 write_inode:
192 drop_inode: !!!ino 188 drop_inode: !!!inode->i_lock!!!
193 evict_inode: 189 evict_inode:
194 put_super: write 190 put_super: write
195 sync_fs: read 191 sync_fs: read
196 freeze_fs: write 192 freeze_fs: write
197 unfreeze_fs: write 193 unfreeze_fs: write
198 statfs: maybe(read) (see b 194 statfs: maybe(read) (see below)
199 remount_fs: write 195 remount_fs: write
200 umount_begin: no 196 umount_begin: no
201 show_options: no (names 197 show_options: no (namespace_sem)
202 quota_read: no (see b 198 quota_read: no (see below)
203 quota_write: no (see b 199 quota_write: no (see below)
>> 200 bdev_try_to_free_page: no (see below)
204 ====================== ============ ====== 201 ====================== ============ ========================
205 202
206 ->statfs() has s_umount (shared) when called b 203 ->statfs() has s_umount (shared) when called by ustat(2) (native or
207 compat), but that's an accident of bad API; s_ 204 compat), but that's an accident of bad API; s_umount is used to pin
208 the superblock down when we only have dev_t gi 205 the superblock down when we only have dev_t given us by userland to
209 identify the superblock. Everything else (sta 206 identify the superblock. Everything else (statfs(), fstatfs(), etc.)
210 doesn't hold it when calling ->statfs() - supe 207 doesn't hold it when calling ->statfs() - superblock is pinned down
211 by resolving the pathname passed to syscall. 208 by resolving the pathname passed to syscall.
212 209
213 ->quota_read() and ->quota_write() functions a 210 ->quota_read() and ->quota_write() functions are both guaranteed to
214 be the only ones operating on the quota file b 211 be the only ones operating on the quota file by the quota code (via
215 dqio_sem) (unless an admin really wants to scr 212 dqio_sem) (unless an admin really wants to screw up something and
216 writes to quota files with quotas on). For oth 213 writes to quota files with quotas on). For other details about locking
217 see also dquot_operations section. 214 see also dquot_operations section.
218 215
>> 216 ->bdev_try_to_free_page is called from the ->releasepage handler of
>> 217 the block device inode. See there for more details.
>> 218
219 file_system_type 219 file_system_type
220 ================ 220 ================
221 221
222 prototypes:: 222 prototypes::
223 223
224 struct dentry *(*mount) (struct file_s 224 struct dentry *(*mount) (struct file_system_type *, int,
225 const char *, void *); 225 const char *, void *);
226 void (*kill_sb) (struct super_block *) 226 void (*kill_sb) (struct super_block *);
227 227
228 locking rules: 228 locking rules:
229 229
230 ======= ========= 230 ======= =========
231 ops may block 231 ops may block
232 ======= ========= 232 ======= =========
233 mount yes 233 mount yes
234 kill_sb yes 234 kill_sb yes
235 ======= ========= 235 ======= =========
236 236
237 ->mount() returns ERR_PTR or the root dentry; 237 ->mount() returns ERR_PTR or the root dentry; its superblock should be locked
238 on return. 238 on return.
239 239
240 ->kill_sb() takes a write-locked superblock, d 240 ->kill_sb() takes a write-locked superblock, does all shutdown work on it,
241 unlocks and drops the reference. 241 unlocks and drops the reference.
242 242
243 address_space_operations 243 address_space_operations
244 ======================== 244 ========================
245 prototypes:: 245 prototypes::
246 246
247 int (*writepage)(struct page *page, st 247 int (*writepage)(struct page *page, struct writeback_control *wbc);
248 int (*read_folio)(struct file *, struc !! 248 int (*readpage)(struct file *, struct page *);
249 int (*writepages)(struct address_space 249 int (*writepages)(struct address_space *, struct writeback_control *);
250 bool (*dirty_folio)(struct address_spa !! 250 int (*set_page_dirty)(struct page *page);
251 void (*readahead)(struct readahead_con 251 void (*readahead)(struct readahead_control *);
>> 252 int (*readpages)(struct file *filp, struct address_space *mapping,
>> 253 struct list_head *pages, unsigned nr_pages);
252 int (*write_begin)(struct file *, stru 254 int (*write_begin)(struct file *, struct address_space *mapping,
253 loff_t pos, un !! 255 loff_t pos, unsigned len, unsigned flags,
254 struct folio * !! 256 struct page **pagep, void **fsdata);
255 int (*write_end)(struct file *, struct 257 int (*write_end)(struct file *, struct address_space *mapping,
256 loff_t pos, un 258 loff_t pos, unsigned len, unsigned copied,
257 struct folio * !! 259 struct page *page, void *fsdata);
258 sector_t (*bmap)(struct address_space 260 sector_t (*bmap)(struct address_space *, sector_t);
259 void (*invalidate_folio) (struct folio !! 261 void (*invalidatepage) (struct page *, unsigned int, unsigned int);
260 bool (*release_folio)(struct folio *, !! 262 int (*releasepage) (struct page *, int);
261 void (*free_folio)(struct folio *); !! 263 void (*freepage)(struct page *);
262 int (*direct_IO)(struct kiocb *, struc 264 int (*direct_IO)(struct kiocb *, struct iov_iter *iter);
263 int (*migrate_folio)(struct address_sp !! 265 bool (*isolate_page) (struct page *, isolate_mode_t);
264 struct folio *src, enu !! 266 int (*migratepage)(struct address_space *, struct page *, struct page *);
265 int (*launder_folio)(struct folio *); !! 267 void (*putback_page) (struct page *);
266 bool (*is_partially_uptodate)(struct f !! 268 int (*launder_page)(struct page *);
267 int (*error_remove_folio)(struct addre !! 269 int (*is_partially_uptodate)(struct page *, unsigned long, unsigned long);
268 int (*swap_activate)(struct swap_info_ !! 270 int (*error_remove_page)(struct address_space *, struct page *);
>> 271 int (*swap_activate)(struct file *);
269 int (*swap_deactivate)(struct file *); 272 int (*swap_deactivate)(struct file *);
270 int (*swap_rw)(struct kiocb *iocb, str <<
271 273
272 locking rules: 274 locking rules:
273 All except dirty_folio and free_folio !! 275 All except set_page_dirty and freepage may block
274 276
275 ====================== ====================== 277 ====================== ======================== ========= ===============
276 ops folio locked !! 278 ops PageLocked(page) i_rwsem invalidate_lock
277 ====================== ====================== 279 ====================== ======================== ========= ===============
278 writepage: yes, unlocks (see belo 280 writepage: yes, unlocks (see below)
279 read_folio: yes, unlocks !! 281 readpage: yes, unlocks shared
280 writepages: 282 writepages:
281 dirty_folio: maybe !! 283 set_page_dirty no
282 readahead: yes, unlocks 284 readahead: yes, unlocks shared
283 write_begin: locks the folio !! 285 readpages: no shared
>> 286 write_begin: locks the page exclusive
284 write_end: yes, unlocks 287 write_end: yes, unlocks exclusive
285 bmap: 288 bmap:
286 invalidate_folio: yes !! 289 invalidatepage: yes exclusive
287 release_folio: yes !! 290 releasepage: yes
288 free_folio: yes !! 291 freepage: yes
289 direct_IO: 292 direct_IO:
290 migrate_folio: yes (both) !! 293 isolate_page: yes
291 launder_folio: yes !! 294 migratepage: yes (both)
>> 295 putback_page: yes
>> 296 launder_page: yes
292 is_partially_uptodate: yes 297 is_partially_uptodate: yes
293 error_remove_folio: yes !! 298 error_remove_page: yes
294 swap_activate: no 299 swap_activate: no
295 swap_deactivate: no 300 swap_deactivate: no
296 swap_rw: yes, unlocks <<
297 ====================== ====================== 301 ====================== ======================== ========= ===============
298 302
299 ->write_begin(), ->write_end() and ->read_foli !! 303 ->write_begin(), ->write_end() and ->readpage() may be called from
300 the request handler (/dev/loop). 304 the request handler (/dev/loop).
301 305
302 ->read_folio() unlocks the folio, either synch !! 306 ->readpage() unlocks the page, either synchronously or via I/O
303 completion. 307 completion.
304 308
305 ->readahead() unlocks the folios that I/O is a !! 309 ->readahead() unlocks the pages that I/O is attempted on like ->readpage().
>> 310
>> 311 ->readpages() populates the pagecache with the passed pages and starts
>> 312 I/O against them. They come unlocked upon I/O completion.
306 313
307 ->writepage() is used for two purposes: for "m 314 ->writepage() is used for two purposes: for "memory cleansing" and for
308 "sync". These are quite different operations 315 "sync". These are quite different operations and the behaviour may differ
309 depending upon the mode. 316 depending upon the mode.
310 317
311 If writepage is called for sync (wbc->sync_mod 318 If writepage is called for sync (wbc->sync_mode != WBC_SYNC_NONE) then
312 it *must* start I/O against the page, even if 319 it *must* start I/O against the page, even if that would involve
313 blocking on in-progress I/O. 320 blocking on in-progress I/O.
314 321
315 If writepage is called for memory cleansing (s 322 If writepage is called for memory cleansing (sync_mode ==
316 WBC_SYNC_NONE) then its role is to get as much 323 WBC_SYNC_NONE) then its role is to get as much writeout underway as
317 possible. So writepage should try to avoid bl 324 possible. So writepage should try to avoid blocking against
318 currently-in-progress I/O. 325 currently-in-progress I/O.
319 326
320 If the filesystem is not called for "sync" and 327 If the filesystem is not called for "sync" and it determines that it
321 would need to block against in-progress I/O to 328 would need to block against in-progress I/O to be able to start new I/O
322 against the page the filesystem should redirty 329 against the page the filesystem should redirty the page with
323 redirty_page_for_writepage(), then unlock the 330 redirty_page_for_writepage(), then unlock the page and return zero.
324 This may also be done to avoid internal deadlo 331 This may also be done to avoid internal deadlocks, but rarely.
325 332
326 If the filesystem is called for sync then it m 333 If the filesystem is called for sync then it must wait on any
327 in-progress I/O and then start new I/O. 334 in-progress I/O and then start new I/O.
328 335
329 The filesystem should unlock the page synchron 336 The filesystem should unlock the page synchronously, before returning to the
330 caller, unless ->writepage() returns special W 337 caller, unless ->writepage() returns special WRITEPAGE_ACTIVATE
331 value. WRITEPAGE_ACTIVATE means that page cann 338 value. WRITEPAGE_ACTIVATE means that page cannot really be written out
332 currently, and VM should stop calling ->writep 339 currently, and VM should stop calling ->writepage() on this page for some
333 time. VM does this by moving page to the head 340 time. VM does this by moving page to the head of the active list, hence the
334 name. 341 name.
335 342
336 Unless the filesystem is going to redirty_page 343 Unless the filesystem is going to redirty_page_for_writepage(), unlock the page
337 and return zero, writepage *must* run set_page 344 and return zero, writepage *must* run set_page_writeback() against the page,
338 followed by unlocking it. Once set_page_write 345 followed by unlocking it. Once set_page_writeback() has been run against the
339 page, write I/O can be submitted and the write 346 page, write I/O can be submitted and the write I/O completion handler must run
340 end_page_writeback() once the I/O is complete. 347 end_page_writeback() once the I/O is complete. If no I/O is submitted, the
341 filesystem must run end_page_writeback() again 348 filesystem must run end_page_writeback() against the page before returning from
342 writepage. 349 writepage.
343 350
344 That is: after 2.5.12, pages which are under w 351 That is: after 2.5.12, pages which are under writeout are *not* locked. Note,
345 if the filesystem needs the page to be locked 352 if the filesystem needs the page to be locked during writeout, that is ok, too,
346 the page is allowed to be unlocked at any poin 353 the page is allowed to be unlocked at any point in time between the calls to
347 set_page_writeback() and end_page_writeback(). 354 set_page_writeback() and end_page_writeback().
348 355
349 Note, failure to run either redirty_page_for_w 356 Note, failure to run either redirty_page_for_writepage() or the combination of
350 set_page_writeback()/end_page_writeback() on a 357 set_page_writeback()/end_page_writeback() on a page submitted to writepage
351 will leave the page itself marked clean but it 358 will leave the page itself marked clean but it will be tagged as dirty in the
352 radix tree. This incoherency can lead to all 359 radix tree. This incoherency can lead to all sorts of hard-to-debug problems
353 in the filesystem like having dirty inodes at 360 in the filesystem like having dirty inodes at umount and losing written data.
354 361
355 ->writepages() is used for periodic writeback 362 ->writepages() is used for periodic writeback and for syscall-initiated
356 sync operations. The address_space should sta 363 sync operations. The address_space should start I/O against at least
357 ``*nr_to_write`` pages. ``*nr_to_write`` must 364 ``*nr_to_write`` pages. ``*nr_to_write`` must be decremented for each page
358 which is written. The address_space implement 365 which is written. The address_space implementation may write more (or less)
359 pages than ``*nr_to_write`` asks for, but it s 366 pages than ``*nr_to_write`` asks for, but it should try to be reasonably close.
360 If nr_to_write is NULL, all dirty pages must b 367 If nr_to_write is NULL, all dirty pages must be written.
361 368
362 writepages should _only_ write pages which are 369 writepages should _only_ write pages which are present on
363 mapping->io_pages. 370 mapping->io_pages.
364 371
365 ->dirty_folio() is called from various places !! 372 ->set_page_dirty() is called from various places in the kernel
366 the target folio is marked as needing writebac !! 373 when the target page is marked as needing writeback. It may be called
367 truncated because either the caller holds the !! 374 under spinlock (it cannot block) and is sometimes called with the page
368 has found the folio while holding the page tab !! 375 not locked.
369 truncation. <<
370 376
371 ->bmap() is currently used by legacy ioctl() ( 377 ->bmap() is currently used by legacy ioctl() (FIBMAP) provided by some
372 filesystems and by the swapper. The latter wil 378 filesystems and by the swapper. The latter will eventually go away. Please,
373 keep it that way and don't breed new callers. 379 keep it that way and don't breed new callers.
374 380
375 ->invalidate_folio() is called when the filesy !! 381 ->invalidatepage() is called when the filesystem must attempt to drop
376 some or all of the buffers from the page when 382 some or all of the buffers from the page when it is being truncated. It
377 returns zero on success. The filesystem must !! 383 returns zero on success. If ->invalidatepage is zero, the kernel uses
378 invalidate_lock before invalidating page cache !! 384 block_invalidatepage() instead. The filesystem must exclusively acquire
379 path (and thus calling into ->invalidate_folio !! 385 invalidate_lock before invalidating page cache in truncate / hole punch path
380 cache invalidation and page cache filling func !! 386 (and thus calling into ->invalidatepage) to block races between page cache
381 !! 387 invalidation and page cache filling functions (fault, read, ...).
382 ->release_folio() is called when the MM wants !! 388
383 folio that would invalidate the filesystem's p !! 389 ->releasepage() is called when the kernel is about to try to drop the
384 it may be about to be removed from the address !! 390 buffers from the page in preparation for freeing it. It returns zero to
385 is locked and not under writeback. It may be !! 391 indicate that the buffers are (or may be) freeable. If ->releasepage is zero,
386 is not usually used for allocation, but rather !! 392 the kernel assumes that the fs has no private interest in the buffers.
387 filesystem may do to attempt to free the priva <<
388 return false to indicate that the folio's priv <<
389 If it returns true, it should have already rem <<
390 the folio. If a filesystem does not provide a <<
391 the pagecache will assume that private data is <<
392 try_to_free_buffers(). <<
393 393
394 ->free_folio() is called when the kernel has d !! 394 ->freepage() is called when the kernel is done dropping the page
395 from the page cache. 395 from the page cache.
396 396
397 ->launder_folio() may be called prior to relea !! 397 ->launder_page() may be called prior to releasing a page if
398 it is still found to be dirty. It returns zero !! 398 it is still found to be dirty. It returns zero if the page was successfully
399 cleaned, or an error value if not. Note that i !! 399 cleaned, or an error value if not. Note that in order to prevent the page
400 getting mapped back in and redirtied, it needs 400 getting mapped back in and redirtied, it needs to be kept locked
401 across the entire operation. 401 across the entire operation.
402 402
403 ->swap_activate() will be called to prepare th !! 403 ->swap_activate will be called with a non-zero argument on
404 should perform any validation and preparation !! 404 files backing (non block device backed) swapfiles. A return value
405 writes can be performed with minimal memory al !! 405 of zero indicates success, in which case this file can be used for
406 add_swap_extent(), or the helper iomap_swapfil !! 406 backing swapspace. The swapspace operations will be proxied to the
407 the number of extents added. If IO should be !! 407 address space operations.
408 ->swap_rw(), it should set SWP_FS_OPS, otherwi <<
409 directly to the block device ``sis->bdev``. <<
410 408
411 ->swap_deactivate() will be called in the sys_ 409 ->swap_deactivate() will be called in the sys_swapoff()
412 path after ->swap_activate() returned success. 410 path after ->swap_activate() returned success.
413 411
414 ->swap_rw will be called for swap IO if SWP_FS <<
415 <<
416 file_lock_operations 412 file_lock_operations
417 ==================== 413 ====================
418 414
419 prototypes:: 415 prototypes::
420 416
421 void (*fl_copy_lock)(struct file_lock 417 void (*fl_copy_lock)(struct file_lock *, struct file_lock *);
422 void (*fl_release_private)(struct file 418 void (*fl_release_private)(struct file_lock *);
423 419
424 420
425 locking rules: 421 locking rules:
426 422
427 =================== ============= ====== 423 =================== ============= =========
428 ops inode->i_lock may bl 424 ops inode->i_lock may block
429 =================== ============= ====== 425 =================== ============= =========
430 fl_copy_lock: yes no 426 fl_copy_lock: yes no
431 fl_release_private: maybe maybe[ 427 fl_release_private: maybe maybe[1]_
432 =================== ============= ====== 428 =================== ============= =========
433 429
434 .. [1]: 430 .. [1]:
435 ->fl_release_private for flock or POSIX loc 431 ->fl_release_private for flock or POSIX locks is currently allowed
436 to block. Leases however can still be freed 432 to block. Leases however can still be freed while the i_lock is held and
437 so fl_release_private called on a lease sho 433 so fl_release_private called on a lease should not block.
438 434
439 lock_manager_operations 435 lock_manager_operations
440 ======================= 436 =======================
441 437
442 prototypes:: 438 prototypes::
443 439
444 void (*lm_notify)(struct file_lock *); 440 void (*lm_notify)(struct file_lock *); /* unblock callback */
445 int (*lm_grant)(struct file_lock *, st 441 int (*lm_grant)(struct file_lock *, struct file_lock *, int);
446 void (*lm_break)(struct file_lock *); 442 void (*lm_break)(struct file_lock *); /* break_lease callback */
447 int (*lm_change)(struct file_lock **, 443 int (*lm_change)(struct file_lock **, int);
448 bool (*lm_breaker_owns_lease)(struct f 444 bool (*lm_breaker_owns_lease)(struct file_lock *);
449 bool (*lm_lock_expirable)(struct file_ 445 bool (*lm_lock_expirable)(struct file_lock *);
450 void (*lm_expire_lock)(void); 446 void (*lm_expire_lock)(void);
451 447
452 locking rules: 448 locking rules:
453 449
454 ====================== ============= ====== 450 ====================== ============= ================= =========
455 ops flc_lock blocke 451 ops flc_lock blocked_lock_lock may block
456 ====================== ============= ====== 452 ====================== ============= ================= =========
457 lm_notify: no yes 453 lm_notify: no yes no
458 lm_grant: no no 454 lm_grant: no no no
459 lm_break: yes no 455 lm_break: yes no no
460 lm_change yes no 456 lm_change yes no no
461 lm_breaker_owns_lease: yes no 457 lm_breaker_owns_lease: yes no no
462 lm_lock_expirable yes no 458 lm_lock_expirable yes no no
463 lm_expire_lock no no 459 lm_expire_lock no no yes
464 ====================== ============= ====== 460 ====================== ============= ================= =========
465 461
466 buffer_head 462 buffer_head
467 =========== 463 ===========
468 464
469 prototypes:: 465 prototypes::
470 466
471 void (*b_end_io)(struct buffer_head *b 467 void (*b_end_io)(struct buffer_head *bh, int uptodate);
472 468
473 locking rules: 469 locking rules:
474 470
475 called from interrupts. In other words, extrem 471 called from interrupts. In other words, extreme care is needed here.
476 bh is locked, but that's all warranties we hav 472 bh is locked, but that's all warranties we have here. Currently only RAID1,
477 highmem, fs/buffer.c, and fs/ntfs/aops.c are p 473 highmem, fs/buffer.c, and fs/ntfs/aops.c are providing these. Block devices
478 call this method upon the IO completion. 474 call this method upon the IO completion.
479 475
480 block_device_operations 476 block_device_operations
481 ======================= 477 =======================
482 prototypes:: 478 prototypes::
483 479
484 int (*open) (struct block_device *, fm 480 int (*open) (struct block_device *, fmode_t);
485 int (*release) (struct gendisk *, fmod 481 int (*release) (struct gendisk *, fmode_t);
486 int (*ioctl) (struct block_device *, f 482 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
487 int (*compat_ioctl) (struct block_devi 483 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
488 int (*direct_access) (struct block_dev 484 int (*direct_access) (struct block_device *, sector_t, void **,
489 unsigned long 485 unsigned long *);
490 void (*unlock_native_capacity) (struct 486 void (*unlock_native_capacity) (struct gendisk *);
491 int (*getgeo)(struct block_device *, s 487 int (*getgeo)(struct block_device *, struct hd_geometry *);
492 void (*swap_slot_free_notify) (struct 488 void (*swap_slot_free_notify) (struct block_device *, unsigned long);
493 489
494 locking rules: 490 locking rules:
495 491
496 ======================= =================== 492 ======================= ===================
497 ops open_mutex 493 ops open_mutex
498 ======================= =================== 494 ======================= ===================
499 open: yes 495 open: yes
500 release: yes 496 release: yes
501 ioctl: no 497 ioctl: no
502 compat_ioctl: no 498 compat_ioctl: no
503 direct_access: no 499 direct_access: no
504 unlock_native_capacity: no 500 unlock_native_capacity: no
505 getgeo: no 501 getgeo: no
506 swap_slot_free_notify: no (see below) 502 swap_slot_free_notify: no (see below)
507 ======================= =================== 503 ======================= ===================
508 504
509 swap_slot_free_notify is called with swap_lock 505 swap_slot_free_notify is called with swap_lock and sometimes the page lock
510 held. 506 held.
511 507
512 508
513 file_operations 509 file_operations
514 =============== 510 ===============
515 511
516 prototypes:: 512 prototypes::
517 513
518 loff_t (*llseek) (struct file *, loff_ 514 loff_t (*llseek) (struct file *, loff_t, int);
519 ssize_t (*read) (struct file *, char _ 515 ssize_t (*read) (struct file *, char __user *, size_t, loff_t *);
520 ssize_t (*write) (struct file *, const 516 ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *);
521 ssize_t (*read_iter) (struct kiocb *, 517 ssize_t (*read_iter) (struct kiocb *, struct iov_iter *);
522 ssize_t (*write_iter) (struct kiocb *, 518 ssize_t (*write_iter) (struct kiocb *, struct iov_iter *);
523 int (*iopoll) (struct kiocb *kiocb, bo 519 int (*iopoll) (struct kiocb *kiocb, bool spin);
>> 520 int (*iterate) (struct file *, struct dir_context *);
524 int (*iterate_shared) (struct file *, 521 int (*iterate_shared) (struct file *, struct dir_context *);
525 __poll_t (*poll) (struct file *, struc 522 __poll_t (*poll) (struct file *, struct poll_table_struct *);
526 long (*unlocked_ioctl) (struct file *, 523 long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long);
527 long (*compat_ioctl) (struct file *, u 524 long (*compat_ioctl) (struct file *, unsigned int, unsigned long);
528 int (*mmap) (struct file *, struct vm_ 525 int (*mmap) (struct file *, struct vm_area_struct *);
529 int (*open) (struct inode *, struct fi 526 int (*open) (struct inode *, struct file *);
530 int (*flush) (struct file *); 527 int (*flush) (struct file *);
531 int (*release) (struct inode *, struct 528 int (*release) (struct inode *, struct file *);
532 int (*fsync) (struct file *, loff_t st 529 int (*fsync) (struct file *, loff_t start, loff_t end, int datasync);
533 int (*fasync) (int, struct file *, int 530 int (*fasync) (int, struct file *, int);
534 int (*lock) (struct file *, int, struc 531 int (*lock) (struct file *, int, struct file_lock *);
>> 532 ssize_t (*sendpage) (struct file *, struct page *, int, size_t,
>> 533 loff_t *, int);
535 unsigned long (*get_unmapped_area)(str 534 unsigned long (*get_unmapped_area)(struct file *, unsigned long,
536 unsigned long, unsigne 535 unsigned long, unsigned long, unsigned long);
537 int (*check_flags)(int); 536 int (*check_flags)(int);
538 int (*flock) (struct file *, int, stru 537 int (*flock) (struct file *, int, struct file_lock *);
539 ssize_t (*splice_write)(struct pipe_in 538 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *,
540 size_t, unsigned int); 539 size_t, unsigned int);
541 ssize_t (*splice_read)(struct file *, 540 ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *,
542 size_t, unsigned int); 541 size_t, unsigned int);
543 int (*setlease)(struct file *, long, s 542 int (*setlease)(struct file *, long, struct file_lock **, void **);
544 long (*fallocate)(struct file *, int, 543 long (*fallocate)(struct file *, int, loff_t, loff_t);
545 void (*show_fdinfo)(struct seq_file *m 544 void (*show_fdinfo)(struct seq_file *m, struct file *f);
546 unsigned (*mmap_capabilities)(struct f 545 unsigned (*mmap_capabilities)(struct file *);
547 ssize_t (*copy_file_range)(struct file 546 ssize_t (*copy_file_range)(struct file *, loff_t, struct file *,
548 loff_t, size_t, unsign 547 loff_t, size_t, unsigned int);
549 loff_t (*remap_file_range)(struct file 548 loff_t (*remap_file_range)(struct file *file_in, loff_t pos_in,
550 struct file *file_out, 549 struct file *file_out, loff_t pos_out,
551 loff_t len, unsigned i 550 loff_t len, unsigned int remap_flags);
552 int (*fadvise)(struct file *, loff_t, 551 int (*fadvise)(struct file *, loff_t, loff_t, int);
553 552
554 locking rules: 553 locking rules:
555 All may block. 554 All may block.
556 555
557 ->llseek() locking has moved from llseek to th 556 ->llseek() locking has moved from llseek to the individual llseek
558 implementations. If your fs is not using gene 557 implementations. If your fs is not using generic_file_llseek, you
559 need to acquire and release the appropriate lo 558 need to acquire and release the appropriate locks in your ->llseek().
560 For many filesystems, it is probably safe to a 559 For many filesystems, it is probably safe to acquire the inode
561 mutex or just to use i_size_read() instead. 560 mutex or just to use i_size_read() instead.
562 Note: this does not protect the file->f_pos ag 561 Note: this does not protect the file->f_pos against concurrent modifications
563 since this is something the userspace has to t 562 since this is something the userspace has to take care about.
564 563
565 ->iterate_shared() is called with i_rwsem held !! 564 ->iterate() is called with i_rwsem exclusive.
566 file f_pos_lock held exclusively !! 565
>> 566 ->iterate_shared() is called with i_rwsem at least shared.
567 567
568 ->fasync() is responsible for maintaining the 568 ->fasync() is responsible for maintaining the FASYNC bit in filp->f_flags.
569 Most instances call fasync_helper(), which doe 569 Most instances call fasync_helper(), which does that maintenance, so it's
570 not normally something one needs to worry abou 570 not normally something one needs to worry about. Return values > 0 will be
571 mapped to zero in the VFS layer. 571 mapped to zero in the VFS layer.
572 572
573 ->readdir() and ->ioctl() on directories must 573 ->readdir() and ->ioctl() on directories must be changed. Ideally we would
574 move ->readdir() to inode_operations and use a 574 move ->readdir() to inode_operations and use a separate method for directory
575 ->ioctl() or kill the latter completely. One o 575 ->ioctl() or kill the latter completely. One of the problems is that for
576 anything that resembles union-mount we won't h 576 anything that resembles union-mount we won't have a struct file for all
577 components. And there are other reasons why th 577 components. And there are other reasons why the current interface is a mess...
578 578
579 ->read on directories probably must go away - 579 ->read on directories probably must go away - we should just enforce -EISDIR
580 in sys_read() and friends. 580 in sys_read() and friends.
581 581
582 ->setlease operations should call generic_setl 582 ->setlease operations should call generic_setlease() before or after setting
583 the lease within the individual filesystem to 583 the lease within the individual filesystem to record the result of the
584 operation 584 operation
585 585
586 ->fallocate implementation must be really care 586 ->fallocate implementation must be really careful to maintain page cache
587 consistency when punching holes or performing 587 consistency when punching holes or performing other operations that invalidate
588 page cache contents. Usually the filesystem ne 588 page cache contents. Usually the filesystem needs to call
589 truncate_inode_pages_range() to invalidate rel 589 truncate_inode_pages_range() to invalidate relevant range of the page cache.
590 However the filesystem usually also needs to u 590 However the filesystem usually also needs to update its internal (and on disk)
591 view of file offset -> disk block mapping. Unt 591 view of file offset -> disk block mapping. Until this update is finished, the
592 filesystem needs to block page faults and read 592 filesystem needs to block page faults and reads from reloading now-stale page
593 cache contents from the disk. Since VFS acquir 593 cache contents from the disk. Since VFS acquires mapping->invalidate_lock in
594 shared mode when loading pages from disk (file 594 shared mode when loading pages from disk (filemap_fault(), filemap_read(),
595 readahead paths), the fallocate implementation 595 readahead paths), the fallocate implementation must take the invalidate_lock to
596 prevent reloading. 596 prevent reloading.
597 597
598 ->copy_file_range and ->remap_file_range imple 598 ->copy_file_range and ->remap_file_range implementations need to serialize
599 against modifications of file data while the o 599 against modifications of file data while the operation is running. For
600 blocking changes through write(2) and similar 600 blocking changes through write(2) and similar operations inode->i_rwsem can be
601 used. To block changes to file contents via a 601 used. To block changes to file contents via a memory mapping during the
602 operation, the filesystem must take mapping->i 602 operation, the filesystem must take mapping->invalidate_lock to coordinate
603 with ->page_mkwrite. 603 with ->page_mkwrite.
604 604
605 dquot_operations 605 dquot_operations
606 ================ 606 ================
607 607
608 prototypes:: 608 prototypes::
609 609
610 int (*write_dquot) (struct dquot *); 610 int (*write_dquot) (struct dquot *);
611 int (*acquire_dquot) (struct dquot *); 611 int (*acquire_dquot) (struct dquot *);
612 int (*release_dquot) (struct dquot *); 612 int (*release_dquot) (struct dquot *);
613 int (*mark_dirty) (struct dquot *); 613 int (*mark_dirty) (struct dquot *);
614 int (*write_info) (struct super_block 614 int (*write_info) (struct super_block *, int);
615 615
616 These operations are intended to be more or le 616 These operations are intended to be more or less wrapping functions that ensure
617 a proper locking wrt the filesystem and call t 617 a proper locking wrt the filesystem and call the generic quota operations.
618 618
619 What filesystem should expect from the generic 619 What filesystem should expect from the generic quota functions:
620 620
621 ============== ============ ============== 621 ============== ============ =========================
622 ops FS recursion Held locks whe 622 ops FS recursion Held locks when called
623 ============== ============ ============== 623 ============== ============ =========================
624 write_dquot: yes dqonoff_sem or 624 write_dquot: yes dqonoff_sem or dqptr_sem
625 acquire_dquot: yes dqonoff_sem or 625 acquire_dquot: yes dqonoff_sem or dqptr_sem
626 release_dquot: yes dqonoff_sem or 626 release_dquot: yes dqonoff_sem or dqptr_sem
627 mark_dirty: no - 627 mark_dirty: no -
628 write_info: yes dqonoff_sem 628 write_info: yes dqonoff_sem
629 ============== ============ ============== 629 ============== ============ =========================
630 630
631 FS recursion means calling ->quota_read() and 631 FS recursion means calling ->quota_read() and ->quota_write() from superblock
632 operations. 632 operations.
633 633
634 More details about quota locking can be found 634 More details about quota locking can be found in fs/dquot.c.
635 635
636 vm_operations_struct 636 vm_operations_struct
637 ==================== 637 ====================
638 638
639 prototypes:: 639 prototypes::
640 640
641 void (*open)(struct vm_area_struct *); !! 641 void (*open)(struct vm_area_struct*);
642 void (*close)(struct vm_area_struct *) !! 642 void (*close)(struct vm_area_struct*);
643 vm_fault_t (*fault)(struct vm_fault *) !! 643 vm_fault_t (*fault)(struct vm_area_struct*, struct vm_fault *);
644 vm_fault_t (*huge_fault)(struct vm_fau <<
645 vm_fault_t (*map_pages)(struct vm_faul <<
646 vm_fault_t (*page_mkwrite)(struct vm_a 644 vm_fault_t (*page_mkwrite)(struct vm_area_struct *, struct vm_fault *);
647 vm_fault_t (*pfn_mkwrite)(struct vm_ar 645 vm_fault_t (*pfn_mkwrite)(struct vm_area_struct *, struct vm_fault *);
648 int (*access)(struct vm_area_struct *, 646 int (*access)(struct vm_area_struct *, unsigned long, void*, int, int);
649 647
650 locking rules: 648 locking rules:
651 649
652 ============= ========== ============== !! 650 ============= ========= ===========================
653 ops mmap_lock PageLocked(pag 651 ops mmap_lock PageLocked(page)
654 ============= ========== ============== !! 652 ============= ========= ===========================
655 open: write !! 653 open: yes
656 close: read/write !! 654 close: yes
657 fault: read can return wit !! 655 fault: yes can return with page locked
658 huge_fault: maybe-read !! 656 map_pages: yes
659 map_pages: maybe-read !! 657 page_mkwrite: yes can return with page locked
660 page_mkwrite: read can return wit !! 658 pfn_mkwrite: yes
661 pfn_mkwrite: read !! 659 access: yes
662 access: read !! 660 ============= ========= ===========================
663 ============= ========== ============== <<
664 661
665 ->fault() is called when a previously not pres 662 ->fault() is called when a previously not present pte is about to be faulted
666 in. The filesystem must find and return the pa 663 in. The filesystem must find and return the page associated with the passed in
667 "pgoff" in the vm_fault structure. If it is po 664 "pgoff" in the vm_fault structure. If it is possible that the page may be
668 truncated and/or invalidated, then the filesys 665 truncated and/or invalidated, then the filesystem must lock invalidate_lock,
669 then ensure the page is not already truncated 666 then ensure the page is not already truncated (invalidate_lock will block
670 subsequent truncate), and then return with VM_ 667 subsequent truncate), and then return with VM_FAULT_LOCKED, and the page
671 locked. The VM will unlock the page. 668 locked. The VM will unlock the page.
672 669
673 ->huge_fault() is called when there is no PUD <<
674 gives the filesystem the opportunity to instal <<
675 Filesystems can also use the ->fault method to <<
676 so implementing this function may not be neces <<
677 filesystems should not call filemap_fault() fr <<
678 The mmap_lock may not be held when this method <<
679 <<
680 ->map_pages() is called when VM asks to map ea 670 ->map_pages() is called when VM asks to map easy accessible pages.
681 Filesystem should find and map pages associate 671 Filesystem should find and map pages associated with offsets from "start_pgoff"
682 till "end_pgoff". ->map_pages() is called with !! 672 till "end_pgoff". ->map_pages() is called with page table locked and must
683 not block. If it's not possible to reach a pa 673 not block. If it's not possible to reach a page without blocking,
684 filesystem should skip it. Filesystem should u !! 674 filesystem should skip it. Filesystem should use do_set_pte() to setup
685 page table entry. Pointer to entry associated 675 page table entry. Pointer to entry associated with the page is passed in
686 "pte" field in vm_fault structure. Pointers to 676 "pte" field in vm_fault structure. Pointers to entries for other offsets
687 should be calculated relative to "pte". 677 should be calculated relative to "pte".
688 678
689 ->page_mkwrite() is called when a previously r 679 ->page_mkwrite() is called when a previously read-only pte is about to become
690 writeable. The filesystem again must ensure th 680 writeable. The filesystem again must ensure that there are no
691 truncate/invalidate races or races with operat 681 truncate/invalidate races or races with operations such as ->remap_file_range
692 or ->copy_file_range, and then return with the 682 or ->copy_file_range, and then return with the page locked. Usually
693 mapping->invalidate_lock is suitable for prope 683 mapping->invalidate_lock is suitable for proper serialization. If the page has
694 been truncated, the filesystem should not look 684 been truncated, the filesystem should not look up a new page like the ->fault()
695 handler, but simply return with VM_FAULT_NOPAG 685 handler, but simply return with VM_FAULT_NOPAGE, which will cause the VM to
696 retry the fault. 686 retry the fault.
697 687
698 ->pfn_mkwrite() is the same as page_mkwrite bu 688 ->pfn_mkwrite() is the same as page_mkwrite but when the pte is
699 VM_PFNMAP or VM_MIXEDMAP with a page-less entr 689 VM_PFNMAP or VM_MIXEDMAP with a page-less entry. Expected return is
700 VM_FAULT_NOPAGE. Or one of the VM_FAULT_ERROR 690 VM_FAULT_NOPAGE. Or one of the VM_FAULT_ERROR types. The default behavior
701 after this call is to make the pte read-write, 691 after this call is to make the pte read-write, unless pfn_mkwrite returns
702 an error. 692 an error.
703 693
704 ->access() is called when get_user_pages() fai 694 ->access() is called when get_user_pages() fails in
705 access_process_vm(), typically used to debug a 695 access_process_vm(), typically used to debug a process through
706 /proc/pid/mem or ptrace. This function is nee 696 /proc/pid/mem or ptrace. This function is needed only for
707 VM_IO | VM_PFNMAP VMAs. 697 VM_IO | VM_PFNMAP VMAs.
708 698
709 ---------------------------------------------- 699 --------------------------------------------------------------------------------
710 700
711 Dubious stuff 701 Dubious stuff
712 702
713 (if you break something or notice that it is b 703 (if you break something or notice that it is broken and do not fix it yourself
714 - at least put it here) 704 - at least put it here)
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.