Commit | Line | Data |
---|---|---|
099c5c7a TH |
1 | .. SPDX-License-Identifier: GPL-2.0 |
2 | ||
90ac11a8 TH |
3 | ========================================= |
4 | Overview of the Linux Virtual File System | |
5 | ========================================= | |
1da177e4 | 6 | |
e66b0457 | 7 | Original author: Richard Gooch <rgooch@atnf.csiro.au> |
1da177e4 | 8 | |
e66b0457 TH |
9 | - Copyright (C) 1999 Richard Gooch |
10 | - Copyright (C) 2005 Pekka Enberg | |
1da177e4 | 11 | |
1da177e4 | 12 | |
cc7d1f8f PE |
13 | Introduction |
14 | ============ | |
1da177e4 | 15 | |
90caa781 TH |
16 | The Virtual File System (also known as the Virtual Filesystem Switch) is |
17 | the software layer in the kernel that provides the filesystem interface | |
18 | to userspace programs. It also provides an abstraction within the | |
19 | kernel which allows different filesystem implementations to coexist. | |
1da177e4 | 20 | |
90caa781 TH |
21 | VFS system calls open(2), stat(2), read(2), write(2), chmod(2) and so on |
22 | are called from a process context. Filesystem locking is described in | |
ec23eb54 | 23 | the document Documentation/filesystems/locking.rst. |
1da177e4 | 24 | |
1da177e4 | 25 | |
cc7d1f8f PE |
26 | Directory Entry Cache (dcache) |
27 | ------------------------------ | |
1da177e4 | 28 | |
cc7d1f8f | 29 | The VFS implements the open(2), stat(2), chmod(2), and similar system |
4ee33ea4 | 30 | calls. The pathname argument that is passed to them is used by the VFS |
cc7d1f8f | 31 | to search through the directory entry cache (also known as the dentry |
4ee33ea4 TH |
32 | cache or dcache). This provides a very fast look-up mechanism to |
33 | translate a pathname (filename) into a specific dentry. Dentries live | |
cc7d1f8f PE |
34 | in RAM and are never saved to disc: they exist only for performance. |
35 | ||
4ee33ea4 | 36 | The dentry cache is meant to be a view into your entire filespace. As |
90caa781 TH |
37 | most computers cannot fit all dentries in the RAM at the same time, some |
38 | bits of the cache are missing. In order to resolve your pathname into a | |
39 | dentry, the VFS may have to resort to creating dentries along the way, | |
40 | and then loading the inode. This is done by looking up the inode. | |
cc7d1f8f PE |
41 | |
42 | ||
43 | The Inode Object | |
44 | ---------------- | |
45 | ||
4ee33ea4 | 46 | An individual dentry usually has a pointer to an inode. Inodes are |
cc7d1f8f | 47 | filesystem objects such as regular files, directories, FIFOs and other |
90caa781 TH |
48 | beasts. They live either on the disc (for block device filesystems) or |
49 | in the memory (for pseudo filesystems). Inodes that live on the disc | |
50 | are copied into the memory when required and changes to the inode are | |
51 | written back to disc. A single inode can be pointed to by multiple | |
cc7d1f8f PE |
52 | dentries (hard links, for example, do this). |
53 | ||
54 | To look up an inode requires that the VFS calls the lookup() method of | |
4ee33ea4 | 55 | the parent directory inode. This method is installed by the specific |
90caa781 TH |
56 | filesystem implementation that the inode lives in. Once the VFS has the |
57 | required dentry (and hence the inode), we can do all those boring things | |
58 | like open(2) the file, or stat(2) it to peek at the inode data. The | |
59 | stat(2) operation is fairly simple: once the VFS has the dentry, it | |
60 | peeks at the inode data and passes some of it back to userspace. | |
cc7d1f8f PE |
61 | |
62 | ||
63 | The File Object | |
64 | --------------- | |
1da177e4 LT |
65 | |
66 | Opening a file requires another operation: allocation of a file | |
90caa781 TH |
67 | structure (this is the kernel-side implementation of file descriptors). |
68 | The freshly allocated file structure is initialized with a pointer to | |
69 | the dentry and a set of file operation member functions. These are | |
70 | taken from the inode data. The open() file method is then called so the | |
71 | specific filesystem implementation can do its work. You can see that | |
72 | this is another switch performed by the VFS. The file structure is | |
73 | placed into the file descriptor table for the process. | |
1da177e4 LT |
74 | |
75 | Reading, writing and closing files (and other assorted VFS operations) | |
76 | is done by using the userspace file descriptor to grab the appropriate | |
cc7d1f8f | 77 | file structure, and then calling the required file structure method to |
4ee33ea4 | 78 | do whatever is required. For as long as the file is open, it keeps the |
cc7d1f8f | 79 | dentry in use, which in turn means that the VFS inode is still in use. |
1da177e4 | 80 | |
5ea626aa PE |
81 | |
82 | Registering and Mounting a Filesystem | |
cc7d1f8f | 83 | ===================================== |
1da177e4 | 84 | |
cc7d1f8f PE |
85 | To register and unregister a filesystem, use the following API |
86 | functions: | |
1da177e4 | 87 | |
af96c1e3 | 88 | .. code-block:: c |
1da177e4 | 89 | |
af96c1e3 TH |
90 | #include <linux/fs.h> |
91 | ||
92 | extern int register_filesystem(struct file_system_type *); | |
93 | extern int unregister_filesystem(struct file_system_type *); | |
1da177e4 | 94 | |
4ee33ea4 | 95 | The passed struct file_system_type describes your filesystem. When a |
90caa781 TH |
96 | request is made to mount a filesystem onto a directory in your |
97 | namespace, the VFS will call the appropriate mount() method for the | |
98 | specific filesystem. New vfsmount referring to the tree returned by | |
99 | ->mount() will be attached to the mountpoint, so that when pathname | |
100 | resolution reaches the mountpoint it will jump into the root of that | |
101 | vfsmount. | |
1da177e4 | 102 | |
cc7d1f8f PE |
103 | You can see all filesystems that are registered to the kernel in the |
104 | file /proc/filesystems. | |
1da177e4 LT |
105 | |
106 | ||
5ea626aa | 107 | struct file_system_type |
cc7d1f8f | 108 | ----------------------- |
1da177e4 | 109 | |
4ee33ea4 | 110 | This describes the filesystem. As of kernel 2.6.39, the following |
1da177e4 LT |
111 | members are defined: |
112 | ||
af96c1e3 TH |
113 | .. code-block:: c |
114 | ||
115 | struct file_system_operations { | |
116 | const char *name; | |
117 | int fs_flags; | |
118 | struct dentry *(*mount) (struct file_system_type *, int, | |
119 | const char *, void *); | |
120 | void (*kill_sb) (struct super_block *); | |
121 | struct module *owner; | |
122 | struct file_system_type * next; | |
123 | struct list_head fs_supers; | |
124 | struct lock_class_key s_lock_key; | |
125 | struct lock_class_key s_umount_key; | |
126 | }; | |
127 | ||
ee5dc049 TH |
128 | ``name`` |
129 | the name of the filesystem type, such as "ext2", "iso9660", | |
1da177e4 LT |
130 | "msdos" and so on |
131 | ||
ee5dc049 TH |
132 | ``fs_flags`` |
133 | various flags (i.e. FS_REQUIRES_DEV, FS_NO_DCACHE, etc.) | |
1da177e4 | 134 | |
ee5dc049 TH |
135 | ``mount`` |
136 | the method to call when a new instance of this filesystem should | |
137 | be mounted | |
1da177e4 | 138 | |
ee5dc049 TH |
139 | ``kill_sb`` |
140 | the method to call when an instance of this filesystem should be | |
141 | shut down | |
5ea626aa | 142 | |
1da177e4 | 143 | |
ee5dc049 TH |
144 | ``owner`` |
145 | for internal VFS use: you should initialize this to THIS_MODULE | |
146 | in most cases. | |
147 | ||
148 | ``next`` | |
149 | for internal VFS use: you should initialize this to NULL | |
5ea626aa | 150 | |
0746aec3 BP |
151 | s_lock_key, s_umount_key: lockdep-specific |
152 | ||
1a102ff9 | 153 | The mount() method has the following arguments: |
1da177e4 | 154 | |
ee5dc049 TH |
155 | ``struct file_system_type *fs_type`` |
156 | describes the filesystem, partly initialized by the specific | |
157 | filesystem code | |
5ea626aa | 158 | |
ee5dc049 TH |
159 | ``int flags`` |
160 | mount flags | |
5ea626aa | 161 | |
ee5dc049 TH |
162 | ``const char *dev_name`` |
163 | the device name we are mounting. | |
1da177e4 | 164 | |
ee5dc049 TH |
165 | ``void *data`` |
166 | arbitrary mount options, usually comes as an ASCII string (see | |
167 | "Mount Options" section) | |
1da177e4 | 168 | |
1a102ff9 AV |
169 | The mount() method must return the root dentry of the tree requested by |
170 | caller. An active reference to its superblock must be grabbed and the | |
171 | superblock must be locked. On failure it should return ERR_PTR(error). | |
1da177e4 | 172 | |
90caa781 TH |
173 | The arguments match those of mount(2) and their interpretation depends |
174 | on filesystem type. E.g. for block filesystems, dev_name is interpreted | |
175 | as block device name, that device is opened and if it contains a | |
176 | suitable filesystem image the method creates and initializes struct | |
177 | super_block accordingly, returning its root dentry to caller. | |
1a102ff9 AV |
178 | |
179 | ->mount() may choose to return a subtree of existing filesystem - it | |
180 | doesn't have to create a new one. The main result from the caller's | |
90caa781 TH |
181 | point of view is a reference to dentry at the root of (sub)tree to be |
182 | attached; creation of new superblock is a common side effect. | |
1da177e4 | 183 | |
90caa781 TH |
184 | The most interesting member of the superblock structure that the mount() |
185 | method fills in is the "s_op" field. This is a pointer to a "struct | |
186 | super_operations" which describes the next level of the filesystem | |
187 | implementation. | |
1da177e4 | 188 | |
1a102ff9 | 189 | Usually, a filesystem uses one of the generic mount() implementations |
4ee33ea4 | 190 | and provides a fill_super() callback instead. The generic variants are: |
5ea626aa | 191 | |
ee5dc049 TH |
192 | ``mount_bdev`` |
193 | mount a filesystem residing on a block device | |
1da177e4 | 194 | |
ee5dc049 TH |
195 | ``mount_nodev`` |
196 | mount a filesystem that is not backed by a device | |
5ea626aa | 197 | |
ee5dc049 TH |
198 | ``mount_single`` |
199 | mount a filesystem which shares the instance between all mounts | |
5ea626aa | 200 | |
1a102ff9 | 201 | A fill_super() callback implementation has the following arguments: |
5ea626aa | 202 | |
ee5dc049 TH |
203 | ``struct super_block *sb`` |
204 | the superblock structure. The callback must initialize this | |
205 | properly. | |
5ea626aa | 206 | |
ee5dc049 TH |
207 | ``void *data`` |
208 | arbitrary mount options, usually comes as an ASCII string (see | |
209 | "Mount Options" section) | |
5ea626aa | 210 | |
ee5dc049 TH |
211 | ``int silent`` |
212 | whether or not to be silent on error | |
5ea626aa PE |
213 | |
214 | ||
cc7d1f8f PE |
215 | The Superblock Object |
216 | ===================== | |
217 | ||
218 | A superblock object represents a mounted filesystem. | |
219 | ||
220 | ||
5ea626aa | 221 | struct super_operations |
cc7d1f8f | 222 | ----------------------- |
1da177e4 LT |
223 | |
224 | This describes how the VFS can manipulate the superblock of your | |
4ee33ea4 | 225 | filesystem. As of kernel 2.6.22, the following members are defined: |
1da177e4 | 226 | |
af96c1e3 TH |
227 | .. code-block:: c |
228 | ||
229 | struct super_operations { | |
230 | struct inode *(*alloc_inode)(struct super_block *sb); | |
231 | void (*destroy_inode)(struct inode *); | |
232 | ||
233 | void (*dirty_inode) (struct inode *, int flags); | |
234 | int (*write_inode) (struct inode *, int); | |
235 | void (*drop_inode) (struct inode *); | |
236 | void (*delete_inode) (struct inode *); | |
237 | void (*put_super) (struct super_block *); | |
238 | int (*sync_fs)(struct super_block *sb, int wait); | |
239 | int (*freeze_fs) (struct super_block *); | |
240 | int (*unfreeze_fs) (struct super_block *); | |
241 | int (*statfs) (struct dentry *, struct kstatfs *); | |
242 | int (*remount_fs) (struct super_block *, int *, char *); | |
243 | void (*clear_inode) (struct inode *); | |
244 | void (*umount_begin) (struct super_block *); | |
245 | ||
246 | int (*show_options)(struct seq_file *, struct dentry *); | |
247 | ||
248 | ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t); | |
249 | ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t); | |
250 | int (*nr_cached_objects)(struct super_block *); | |
251 | void (*free_cached_objects)(struct super_block *, int); | |
252 | }; | |
1da177e4 LT |
253 | |
254 | All methods are called without any locks being held, unless otherwise | |
4ee33ea4 | 255 | noted. This means that most methods can block safely. All methods are |
1da177e4 LT |
256 | only called from a process context (i.e. not from an interrupt handler |
257 | or bottom half). | |
258 | ||
ee5dc049 TH |
259 | ``alloc_inode`` |
260 | this method is called by alloc_inode() to allocate memory for | |
261 | struct inode and initialize it. If this function is not | |
50c1f43a TH |
262 | defined, a simple 'struct inode' is allocated. Normally |
263 | alloc_inode will be used to allocate a larger structure which | |
264 | contains a 'struct inode' embedded within it. | |
5ea626aa | 265 | |
ee5dc049 TH |
266 | ``destroy_inode`` |
267 | this method is called by destroy_inode() to release resources | |
268 | allocated for struct inode. It is only required if | |
50c1f43a | 269 | ->alloc_inode was defined and simply undoes anything done by |
341546f5 | 270 | ->alloc_inode. |
5ea626aa | 271 | |
ee5dc049 TH |
272 | ``dirty_inode`` |
273 | this method is called by the VFS to mark an inode dirty. | |
1da177e4 | 274 | |
ee5dc049 TH |
275 | ``write_inode`` |
276 | this method is called when the VFS needs to write an inode to | |
277 | disc. The second parameter indicates whether the write should | |
278 | be synchronous or not, not all filesystems check this flag. | |
1da177e4 | 279 | |
ee5dc049 TH |
280 | ``drop_inode`` |
281 | called when the last access to the inode is dropped, with the | |
282 | inode->i_lock spinlock held. | |
1da177e4 | 283 | |
5ea626aa | 284 | This method should be either NULL (normal UNIX filesystem |
ee5dc049 TH |
285 | semantics) or "generic_delete_inode" (for filesystems that do |
286 | not want to cache inodes - causing "delete_inode" to always be | |
1da177e4 LT |
287 | called regardless of the value of i_nlink) |
288 | ||
ee5dc049 TH |
289 | The "generic_delete_inode()" behavior is equivalent to the old |
290 | practice of using "force_delete" in the put_inode() case, but | |
291 | does not have the races that the "force_delete()" approach had. | |
1da177e4 | 292 | |
ee5dc049 TH |
293 | ``delete_inode`` |
294 | called when the VFS wants to delete an inode | |
1da177e4 | 295 | |
ee5dc049 TH |
296 | ``put_super`` |
297 | called when the VFS wishes to free the superblock | |
4ee33ea4 | 298 | (i.e. unmount). This is called with the superblock lock held |
1da177e4 | 299 | |
ee5dc049 TH |
300 | ``sync_fs`` |
301 | called when VFS is writing out all dirty data associated with a | |
302 | superblock. The second parameter indicates whether the method | |
4ee33ea4 | 303 | should wait until the write out has been completed. Optional. |
5ea626aa | 304 | |
ee5dc049 TH |
305 | ``freeze_fs`` |
306 | called when VFS is locking a filesystem and forcing it into a | |
307 | consistent state. This method is currently used by the Logical | |
308 | Volume Manager (LVM). | |
5ea626aa | 309 | |
ee5dc049 TH |
310 | ``unfreeze_fs`` |
311 | called when VFS is unlocking a filesystem and making it writable | |
50c1f43a | 312 | again. |
5ea626aa | 313 | |
ee5dc049 TH |
314 | ``statfs`` |
315 | called when the VFS needs to get filesystem statistics. | |
1da177e4 | 316 | |
ee5dc049 TH |
317 | ``remount_fs`` |
318 | called when the filesystem is remounted. This is called with | |
319 | the kernel lock held | |
1da177e4 | 320 | |
ee5dc049 TH |
321 | ``clear_inode`` |
322 | called then the VFS clears the inode. Optional | |
1da177e4 | 323 | |
ee5dc049 TH |
324 | ``umount_begin`` |
325 | called when the VFS is unmounting a filesystem. | |
5ea626aa | 326 | |
ee5dc049 TH |
327 | ``show_options`` |
328 | called by the VFS to show mount options for /proc/<pid>/mounts. | |
329 | (see "Mount Options" section) | |
5ea626aa | 330 | |
ee5dc049 TH |
331 | ``quota_read`` |
332 | called by the VFS to read from filesystem quota file. | |
5ea626aa | 333 | |
ee5dc049 TH |
334 | ``quota_write`` |
335 | called by the VFS to write to filesystem quota file. | |
5ea626aa | 336 | |
ee5dc049 TH |
337 | ``nr_cached_objects`` |
338 | called by the sb cache shrinking function for the filesystem to | |
339 | return the number of freeable cached objects it contains. | |
0e1fdafd DC |
340 | Optional. |
341 | ||
ee5dc049 TH |
342 | ``free_cache_objects`` |
343 | called by the sb cache shrinking function for the filesystem to | |
344 | scan the number of objects indicated to try to free them. | |
345 | Optional, but any filesystem implementing this method needs to | |
346 | also implement ->nr_cached_objects for it to be called | |
347 | correctly. | |
0e1fdafd DC |
348 | |
349 | We can't do anything with any errors that the filesystem might | |
ee5dc049 TH |
350 | encountered, hence the void return type. This will never be |
351 | called if the VM is trying to reclaim under GFP_NOFS conditions, | |
352 | hence this method does not need to handle that situation itself. | |
0e1fdafd | 353 | |
ee5dc049 TH |
354 | Implementations must include conditional reschedule calls inside |
355 | any scanning loop that is done. This allows the VFS to | |
356 | determine appropriate scan batch sizes without having to worry | |
357 | about whether implementations will cause holdoff problems due to | |
358 | large scan batch sizes. | |
8ab47664 | 359 | |
90caa781 TH |
360 | Whoever sets up the inode is responsible for filling in the "i_op" |
361 | field. This is a pointer to a "struct inode_operations" which describes | |
362 | the methods that can be performed on individual inodes. | |
1da177e4 | 363 | |
e04c83cd | 364 | |
6c6ef9f2 AG |
365 | struct xattr_handlers |
366 | --------------------- | |
367 | ||
368 | On filesystems that support extended attributes (xattrs), the s_xattr | |
90caa781 TH |
369 | superblock field points to a NULL-terminated array of xattr handlers. |
370 | Extended attributes are name:value pairs. | |
6c6ef9f2 | 371 | |
ee5dc049 TH |
372 | ``name`` |
373 | Indicates that the handler matches attributes with the specified | |
374 | name (such as "system.posix_acl_access"); the prefix field must | |
375 | be NULL. | |
6c6ef9f2 | 376 | |
ee5dc049 TH |
377 | ``prefix`` |
378 | Indicates that the handler matches all attributes with the | |
379 | specified name prefix (such as "user."); the name field must be | |
380 | NULL. | |
6c6ef9f2 | 381 | |
ee5dc049 TH |
382 | ``list`` |
383 | Determine if attributes matching this xattr handler should be | |
384 | listed for a particular dentry. Used by some listxattr | |
385 | implementations like generic_listxattr. | |
6c6ef9f2 | 386 | |
ee5dc049 TH |
387 | ``get`` |
388 | Called by the VFS to get the value of a particular extended | |
389 | attribute. This method is called by the getxattr(2) system | |
390 | call. | |
6c6ef9f2 | 391 | |
ee5dc049 TH |
392 | ``set`` |
393 | Called by the VFS to set the value of a particular extended | |
394 | attribute. When the new value is NULL, called to remove a | |
8286de7c | 395 | particular extended attribute. This method is called by the |
ee5dc049 | 396 | setxattr(2) and removexattr(2) system calls. |
6c6ef9f2 | 397 | |
90caa781 TH |
398 | When none of the xattr handlers of a filesystem match the specified |
399 | attribute name or when a filesystem doesn't support extended attributes, | |
af96c1e3 | 400 | the various ``*xattr(2)`` system calls return -EOPNOTSUPP. |
6c6ef9f2 | 401 | |
1da177e4 | 402 | |
cc7d1f8f PE |
403 | The Inode Object |
404 | ================ | |
405 | ||
406 | An inode object represents an object within the filesystem. | |
407 | ||
408 | ||
5ea626aa | 409 | struct inode_operations |
cc7d1f8f | 410 | ----------------------- |
1da177e4 | 411 | |
90caa781 TH |
412 | This describes how the VFS can manipulate an inode in your filesystem. |
413 | As of kernel 2.6.22, the following members are defined: | |
1da177e4 | 414 | |
af96c1e3 TH |
415 | .. code-block:: c |
416 | ||
417 | struct inode_operations { | |
549c7297 | 418 | int (*create) (struct user_namespace *, struct inode *,struct dentry *, umode_t, bool); |
af96c1e3 TH |
419 | struct dentry * (*lookup) (struct inode *,struct dentry *, unsigned int); |
420 | int (*link) (struct dentry *,struct inode *,struct dentry *); | |
421 | int (*unlink) (struct inode *,struct dentry *); | |
549c7297 CB |
422 | int (*symlink) (struct user_namespace *, struct inode *,struct dentry *,const char *); |
423 | int (*mkdir) (struct user_namespace *, struct inode *,struct dentry *,umode_t); | |
af96c1e3 | 424 | int (*rmdir) (struct inode *,struct dentry *); |
549c7297 CB |
425 | int (*mknod) (struct user_namespace *, struct inode *,struct dentry *,umode_t,dev_t); |
426 | int (*rename) (struct user_namespace *, struct inode *, struct dentry *, | |
af96c1e3 TH |
427 | struct inode *, struct dentry *, unsigned int); |
428 | int (*readlink) (struct dentry *, char __user *,int); | |
429 | const char *(*get_link) (struct dentry *, struct inode *, | |
430 | struct delayed_call *); | |
549c7297 | 431 | int (*permission) (struct user_namespace *, struct inode *, int); |
af96c1e3 | 432 | int (*get_acl)(struct inode *, int); |
549c7297 CB |
433 | int (*setattr) (struct user_namespace *, struct dentry *, struct iattr *); |
434 | int (*getattr) (struct user_namespace *, const struct path *, struct kstat *, u32, unsigned int); | |
af96c1e3 TH |
435 | ssize_t (*listxattr) (struct dentry *, char *, size_t); |
436 | void (*update_time)(struct inode *, struct timespec *, int); | |
437 | int (*atomic_open)(struct inode *, struct dentry *, struct file *, | |
438 | unsigned open_flag, umode_t create_mode); | |
549c7297 CB |
439 | int (*tmpfile) (struct user_namespace *, struct inode *, struct dentry *, umode_t); |
440 | int (*set_acl)(struct user_namespace *, struct inode *, struct posix_acl *, int); | |
af96c1e3 | 441 | }; |
1da177e4 LT |
442 | |
443 | Again, all methods are called without any locks being held, unless | |
444 | otherwise noted. | |
445 | ||
ee5dc049 TH |
446 | ``create`` |
447 | called by the open(2) and creat(2) system calls. Only required | |
448 | if you want to support regular files. The dentry you get should | |
449 | not have an inode (i.e. it should be a negative dentry). Here | |
450 | you will probably call d_instantiate() with the dentry and the | |
451 | newly created inode | |
1da177e4 | 452 | |
ee5dc049 TH |
453 | ``lookup`` |
454 | called when the VFS needs to look up an inode in a parent | |
4ee33ea4 | 455 | directory. The name to look for is found in the dentry. This |
1da177e4 | 456 | method must call d_add() to insert the found inode into the |
4ee33ea4 TH |
457 | dentry. The "i_count" field in the inode structure should be |
458 | incremented. If the named inode does not exist a NULL inode | |
1da177e4 | 459 | should be inserted into the dentry (this is called a negative |
ee5dc049 TH |
460 | dentry). Returning an error code from this routine must only be |
461 | done on a real error, otherwise creating inodes with system | |
1da177e4 LT |
462 | calls like create(2), mknod(2), mkdir(2) and so on will fail. |
463 | If you wish to overload the dentry methods then you should | |
ee5dc049 TH |
464 | initialise the "d_dop" field in the dentry; this is a pointer to |
465 | a struct "dentry_operations". This method is called with the | |
466 | directory inode semaphore held | |
1da177e4 | 467 | |
ee5dc049 TH |
468 | ``link`` |
469 | called by the link(2) system call. Only required if you want to | |
470 | support hard links. You will probably need to call | |
1da177e4 LT |
471 | d_instantiate() just as you would in the create() method |
472 | ||
ee5dc049 TH |
473 | ``unlink`` |
474 | called by the unlink(2) system call. Only required if you want | |
475 | to support deleting inodes | |
1da177e4 | 476 | |
ee5dc049 TH |
477 | ``symlink`` |
478 | called by the symlink(2) system call. Only required if you want | |
479 | to support symlinks. You will probably need to call | |
1da177e4 LT |
480 | d_instantiate() just as you would in the create() method |
481 | ||
ee5dc049 TH |
482 | ``mkdir`` |
483 | called by the mkdir(2) system call. Only required if you want | |
4ee33ea4 | 484 | to support creating subdirectories. You will probably need to |
1da177e4 LT |
485 | call d_instantiate() just as you would in the create() method |
486 | ||
ee5dc049 TH |
487 | ``rmdir`` |
488 | called by the rmdir(2) system call. Only required if you want | |
1da177e4 LT |
489 | to support deleting subdirectories |
490 | ||
ee5dc049 TH |
491 | ``mknod`` |
492 | called by the mknod(2) system call to create a device (char, | |
493 | block) inode or a named pipe (FIFO) or socket. Only required if | |
494 | you want to support creating these types of inodes. You will | |
495 | probably need to call d_instantiate() just as you would in the | |
496 | create() method | |
1da177e4 | 497 | |
ee5dc049 TH |
498 | ``rename`` |
499 | called by the rename(2) system call to rename the object to have | |
500 | the parent and name given by the second inode and dentry. | |
cc7d1f8f | 501 | |
18fc84da | 502 | The filesystem must return -EINVAL for any unsupported or |
ee5dc049 TH |
503 | unknown flags. Currently the following flags are implemented: |
504 | (1) RENAME_NOREPLACE: this flag indicates that if the target of | |
505 | the rename exists the rename should fail with -EEXIST instead of | |
506 | replacing the target. The VFS already checks for existence, so | |
507 | for local filesystems the RENAME_NOREPLACE implementation is | |
508 | equivalent to plain rename. | |
520c8b16 | 509 | (2) RENAME_EXCHANGE: exchange source and target. Both must |
ee5dc049 TH |
510 | exist; this is checked by the VFS. Unlike plain rename, source |
511 | and target may be of different type. | |
512 | ||
513 | ``get_link`` | |
514 | called by the VFS to follow a symbolic link to the inode it | |
515 | points to. Only required if you want to support symbolic links. | |
516 | This method returns the symlink body to traverse (and possibly | |
517 | resets the current position with nd_jump_link()). If the body | |
518 | won't go away until the inode is gone, nothing else is needed; | |
519 | if it needs to be otherwise pinned, arrange for its release by | |
520 | having get_link(..., ..., done) do set_delayed_call(done, | |
521 | destructor, argument). In that case destructor(argument) will | |
522 | be called once VFS is done with the body you've returned. May | |
523 | be called in RCU mode; that is indicated by NULL dentry | |
fceef393 AV |
524 | argument. If request can't be handled without leaving RCU mode, |
525 | have it return ERR_PTR(-ECHILD). | |
cc7d1f8f | 526 | |
dcb2cb1f EB |
527 | If the filesystem stores the symlink target in ->i_link, the |
528 | VFS may use it directly without calling ->get_link(); however, | |
529 | ->get_link() must still be provided. ->i_link must not be | |
530 | freed until after an RCU grace period. Writing to ->i_link | |
531 | post-iget() time requires a 'release' memory barrier. | |
532 | ||
ee5dc049 TH |
533 | ``readlink`` |
534 | this is now just an override for use by readlink(2) for the | |
76fca90e MS |
535 | cases when ->get_link uses nd_jump_link() or object is not in |
536 | fact a symlink. Normally filesystems should only implement | |
537 | ->get_link for symlinks and readlink(2) will automatically use | |
538 | that. | |
539 | ||
ee5dc049 TH |
540 | ``permission`` |
541 | called by the VFS to check for access rights on a POSIX-like | |
50c1f43a | 542 | filesystem. |
5ea626aa | 543 | |
ee5dc049 TH |
544 | May be called in rcu-walk mode (mask & MAY_NOT_BLOCK). If in |
545 | rcu-walk mode, the filesystem must check the permission without | |
546 | blocking or storing to the inode. | |
b74c79e9 | 547 | |
ee5dc049 TH |
548 | If a situation is encountered that rcu-walk cannot handle, |
549 | return | |
b74c79e9 NP |
550 | -ECHILD and it will be called again in ref-walk mode. |
551 | ||
ee5dc049 TH |
552 | ``setattr`` |
553 | called by the VFS to set attributes for a file. This method is | |
554 | called by chmod(2) and related system calls. | |
555 | ||
556 | ``getattr`` | |
557 | called by the VFS to get attributes of a file. This method is | |
558 | called by stat(2) and related system calls. | |
559 | ||
560 | ``listxattr`` | |
561 | called by the VFS to list all extended attributes for a given | |
562 | file. This method is called by the listxattr(2) system call. | |
563 | ||
564 | ``update_time`` | |
565 | called by the VFS to update a specific time or the i_version of | |
566 | an inode. If this is not defined the VFS will update the inode | |
567 | itself and call mark_inode_dirty_sync. | |
568 | ||
569 | ``atomic_open`` | |
570 | called on the last component of an open. Using this optional | |
571 | method the filesystem can look up, possibly create and open the | |
572 | file in one atomic operation. If it wants to leave actual | |
573 | opening to the caller (e.g. if the file turned out to be a | |
574 | symlink, device, or just something filesystem won't do atomic | |
575 | open for), it may signal this by returning finish_no_open(file, | |
576 | dentry). This method is only called if the last component is | |
577 | negative or needs lookup. Cached positive dentries are still | |
578 | handled by f_op->open(). If the file was created, FMODE_CREATED | |
579 | flag should be set in file->f_mode. In case of O_EXCL the | |
580 | method must only succeed if the file didn't exist and hence | |
581 | FMODE_CREATED shall always be set on success. | |
582 | ||
583 | ``tmpfile`` | |
584 | called in the end of O_TMPFILE open(). Optional, equivalent to | |
585 | atomically creating, opening and unlinking a file in given | |
586 | directory. | |
48bde8d3 | 587 | |
e04c83cd | 588 | |
cc7d1f8f PE |
589 | The Address Space Object |
590 | ======================== | |
591 | ||
341546f5 | 592 | The address space object is used to group and manage pages in the page |
90caa781 TH |
593 | cache. It can be used to keep track of the pages in a file (or anything |
594 | else) and also track the mapping of sections of the file into process | |
595 | address spaces. | |
341546f5 N |
596 | |
597 | There are a number of distinct yet related services that an | |
90caa781 TH |
598 | address-space can provide. These include communicating memory pressure, |
599 | page lookup by address, and keeping track of pages tagged as Dirty or | |
600 | Writeback. | |
341546f5 | 601 | |
a9e102b6 | 602 | The first can be used independently to the others. The VM can try to |
90caa781 TH |
603 | either write dirty pages in order to clean them, or release clean pages |
604 | in order to reuse them. To do this it can call the ->writepage method | |
605 | on dirty pages, and ->releasepage on clean pages with PagePrivate set. | |
606 | Clean pages without PagePrivate and with no external references will be | |
607 | released without notice being given to the address_space. | |
341546f5 | 608 | |
a9e102b6 | 609 | To achieve this functionality, pages need to be placed on an LRU with |
90caa781 TH |
610 | lru_cache_add and mark_page_active needs to be called whenever the page |
611 | is used. | |
341546f5 | 612 | |
4ee33ea4 | 613 | Pages are normally kept in a radix tree index by ->index. This tree |
90caa781 TH |
614 | maintains information about the PG_Dirty and PG_Writeback status of each |
615 | page, so that pages with either of these flags can be found quickly. | |
341546f5 N |
616 | |
617 | The Dirty tag is primarily used by mpage_writepages - the default | |
618 | ->writepages method. It uses the tag to find dirty pages to call | |
619 | ->writepage on. If mpage_writepages is not used (i.e. the address | |
90caa781 TH |
620 | provides its own ->writepages) , the PAGECACHE_TAG_DIRTY tag is almost |
621 | unused. write_inode_now and sync_inode do use it (through | |
341546f5 N |
622 | __sync_single_inode) to check if ->writepages has been successful in |
623 | writing out the whole address_space. | |
624 | ||
90caa781 TH |
625 | The Writeback tag is used by filemap*wait* and sync_page* functions, via |
626 | filemap_fdatawait_range, to wait for all writeback to complete. | |
341546f5 N |
627 | |
628 | An address_space handler may attach extra information to a page, | |
629 | typically using the 'private' field in the 'struct page'. If such | |
630 | information is attached, the PG_Private flag should be set. This will | |
a9e102b6 | 631 | cause various VM routines to make extra calls into the address_space |
341546f5 N |
632 | handler to deal with that data. |
633 | ||
634 | An address space acts as an intermediate between storage and | |
635 | application. Data is read into the address space a whole page at a | |
90caa781 TH |
636 | time, and provided to the application either by copying of the page, or |
637 | by memory-mapping the page. Data is written into the address space by | |
638 | the application, and then written-back to storage typically in whole | |
639 | pages, however the address_space has finer control of write sizes. | |
341546f5 N |
640 | |
641 | The read process essentially only requires 'readpage'. The write | |
4e02ed4b | 642 | process is more complicated and uses write_begin/write_end or |
90caa781 TH |
643 | set_page_dirty to write data into the address_space, and writepage and |
644 | writepages to writeback data to storage. | |
341546f5 N |
645 | |
646 | Adding and removing pages to/from an address_space is protected by the | |
647 | inode's i_mutex. | |
648 | ||
649 | When data is written to a page, the PG_Dirty flag should be set. It | |
650 | typically remains set until writepage asks for it to be written. This | |
90caa781 TH |
651 | should clear PG_Dirty and set PG_Writeback. It can be actually written |
652 | at any point after PG_Dirty is clear. Once it is known to be safe, | |
653 | PG_Writeback is cleared. | |
341546f5 | 654 | |
acbf3c34 | 655 | Writeback makes use of a writeback_control structure to direct the |
8286de7c | 656 | operations. This gives the writepage and writepages operations some |
acbf3c34 JL |
657 | information about the nature of and reason for the writeback request, |
658 | and the constraints under which it is being done. It is also used to | |
659 | return information back to the caller about the result of a writepage or | |
660 | writepages request. | |
661 | ||
e04c83cd | 662 | |
acbf3c34 JL |
663 | Handling errors during writeback |
664 | -------------------------------- | |
e04c83cd | 665 | |
acbf3c34 JL |
666 | Most applications that do buffered I/O will periodically call a file |
667 | synchronization call (fsync, fdatasync, msync or sync_file_range) to | |
668 | ensure that data written has made it to the backing store. When there | |
669 | is an error during writeback, they expect that error to be reported when | |
670 | a file sync request is made. After an error has been reported on one | |
671 | request, subsequent requests on the same file descriptor should return | |
672 | 0, unless further writeback errors have occurred since the previous file | |
673 | syncronization. | |
674 | ||
675 | Ideally, the kernel would report errors only on file descriptions on | |
676 | which writes were done that subsequently failed to be written back. The | |
677 | generic pagecache infrastructure does not track the file descriptions | |
678 | that have dirtied each individual page however, so determining which | |
679 | file descriptors should get back an error is not possible. | |
680 | ||
681 | Instead, the generic writeback error tracking infrastructure in the | |
682 | kernel settles for reporting errors to fsync on all file descriptions | |
683 | that were open at the time that the error occurred. In a situation with | |
90caa781 TH |
684 | multiple writers, all of them will get back an error on a subsequent |
685 | fsync, even if all of the writes done through that particular file | |
686 | descriptor succeeded (or even if there were no writes on that file | |
687 | descriptor at all). | |
acbf3c34 JL |
688 | |
689 | Filesystems that wish to use this infrastructure should call | |
690 | mapping_set_error to record the error in the address_space when it | |
691 | occurs. Then, after writing back data from the pagecache in their | |
692 | file->fsync operation, they should call file_check_and_advance_wb_err to | |
693 | ensure that the struct file's error cursor has advanced to the correct | |
694 | point in the stream of errors emitted by the backing device(s). | |
5ea626aa | 695 | |
e04c83cd | 696 | |
5ea626aa | 697 | struct address_space_operations |
cc7d1f8f | 698 | ------------------------------- |
5ea626aa | 699 | |
90caa781 TH |
700 | This describes how the VFS can manipulate mapping of a file to page |
701 | cache in your filesystem. The following members are defined: | |
5ea626aa | 702 | |
af96c1e3 TH |
703 | .. code-block:: c |
704 | ||
705 | struct address_space_operations { | |
706 | int (*writepage)(struct page *page, struct writeback_control *wbc); | |
707 | int (*readpage)(struct file *, struct page *); | |
708 | int (*writepages)(struct address_space *, struct writeback_control *); | |
709 | int (*set_page_dirty)(struct page *page); | |
8151b4c8 | 710 | void (*readahead)(struct readahead_control *); |
af96c1e3 TH |
711 | int (*readpages)(struct file *filp, struct address_space *mapping, |
712 | struct list_head *pages, unsigned nr_pages); | |
713 | int (*write_begin)(struct file *, struct address_space *mapping, | |
714 | loff_t pos, unsigned len, unsigned flags, | |
afddba49 | 715 | struct page **pagep, void **fsdata); |
af96c1e3 TH |
716 | int (*write_end)(struct file *, struct address_space *mapping, |
717 | loff_t pos, unsigned len, unsigned copied, | |
718 | struct page *page, void *fsdata); | |
719 | sector_t (*bmap)(struct address_space *, sector_t); | |
720 | void (*invalidatepage) (struct page *, unsigned int, unsigned int); | |
721 | int (*releasepage) (struct page *, int); | |
722 | void (*freepage)(struct page *); | |
723 | ssize_t (*direct_IO)(struct kiocb *, struct iov_iter *iter); | |
724 | /* isolate a page for migration */ | |
725 | bool (*isolate_page) (struct page *, isolate_mode_t); | |
726 | /* migrate the contents of a page to the specified target */ | |
727 | int (*migratepage) (struct page *, struct page *); | |
728 | /* put migration-failed page back to right list */ | |
729 | void (*putback_page) (struct page *); | |
730 | int (*launder_page) (struct page *); | |
731 | ||
732 | int (*is_partially_uptodate) (struct page *, unsigned long, | |
733 | unsigned long); | |
734 | void (*is_dirty_writeback) (struct page *, bool *, bool *); | |
735 | int (*error_remove_page) (struct mapping *mapping, struct page *page); | |
736 | int (*swap_activate)(struct file *); | |
737 | int (*swap_deactivate)(struct file *); | |
738 | }; | |
739 | ||
ee5dc049 TH |
740 | ``writepage`` |
741 | called by the VM to write a dirty page to backing store. This | |
742 | may happen for data integrity reasons (i.e. 'sync'), or to free | |
743 | up memory (flush). The difference can be seen in | |
744 | wbc->sync_mode. The PG_Dirty flag has been cleared and | |
745 | PageLocked is true. writepage should start writeout, should set | |
746 | PG_Writeback, and should make sure the page is unlocked, either | |
747 | synchronously or asynchronously when the write operation | |
748 | completes. | |
749 | ||
750 | If wbc->sync_mode is WB_SYNC_NONE, ->writepage doesn't have to | |
751 | try too hard if there are problems, and may choose to write out | |
752 | other pages from the mapping if that is easier (e.g. due to | |
753 | internal dependencies). If it chooses not to start writeout, it | |
754 | should return AOP_WRITEPAGE_ACTIVATE so that the VM will not | |
755 | keep calling ->writepage on that page. | |
756 | ||
757 | See the file "Locking" for more details. | |
758 | ||
759 | ``readpage`` | |
760 | called by the VM to read a page from backing store. The page | |
761 | will be Locked when readpage is called, and should be unlocked | |
762 | and marked uptodate once the read completes. If ->readpage | |
763 | discovers that it needs to unlock the page for some reason, it | |
764 | can do so, and then return AOP_TRUNCATED_PAGE. In this case, | |
765 | the page will be relocated, relocked and if that all succeeds, | |
766 | ->readpage will be called again. | |
767 | ||
768 | ``writepages`` | |
769 | called by the VM to write out pages associated with the | |
e9b2f15b | 770 | address_space object. If wbc->sync_mode is WB_SYNC_ALL, then |
50c1f43a | 771 | the writeback_control will specify a range of pages that must be |
e9b2f15b | 772 | written out. If it is WB_SYNC_NONE, then a nr_to_write is |
ee5dc049 TH |
773 | given and that many pages should be written if possible. If no |
774 | ->writepages is given, then mpage_writepages is used instead. | |
775 | This will choose pages from the address space that are tagged as | |
776 | DIRTY and will pass them to ->writepage. | |
777 | ||
778 | ``set_page_dirty`` | |
779 | called by the VM to set a page dirty. This is particularly | |
780 | needed if an address space attaches private data to a page, and | |
781 | that data needs to be updated when a page is dirtied. This is | |
782 | called, for example, when a memory mapped page gets modified. | |
341546f5 | 783 | If defined, it should set the PageDirty flag, and the |
1b44ae63 | 784 | PAGECACHE_TAG_DIRTY tag in the radix tree. |
5ea626aa | 785 | |
8151b4c8 MWO |
786 | ``readahead`` |
787 | Called by the VM to read pages associated with the address_space | |
788 | object. The pages are consecutive in the page cache and are | |
789 | locked. The implementation should decrement the page refcount | |
790 | after starting I/O on each page. Usually the page will be | |
791 | unlocked by the I/O completion handler. If the filesystem decides | |
792 | to stop attempting I/O before reaching the end of the readahead | |
793 | window, it can simply return. The caller will decrement the page | |
794 | refcount and unlock the remaining pages for you. Set PageUptodate | |
795 | if the I/O completes successfully. Setting PageError on any page | |
796 | will be ignored; simply unlock the page if an I/O error occurs. | |
797 | ||
ee5dc049 TH |
798 | ``readpages`` |
799 | called by the VM to read pages associated with the address_space | |
800 | object. This is essentially just a vector version of readpage. | |
801 | Instead of just one page, several pages are requested. | |
a9e102b6 | 802 | readpages is only used for read-ahead, so read errors are |
50c1f43a | 803 | ignored. If anything goes wrong, feel free to give up. |
8151b4c8 MWO |
804 | This interface is deprecated and will be removed by the end of |
805 | 2020; implement readahead instead. | |
1da177e4 | 806 | |
ee5dc049 TH |
807 | ``write_begin`` |
808 | Called by the generic buffered write code to ask the filesystem | |
809 | to prepare to write len bytes at the given offset in the file. | |
810 | The address_space should check that the write will be able to | |
811 | complete, by allocating space if necessary and doing any other | |
812 | internal housekeeping. If the write will update parts of any | |
813 | basic-blocks on storage, then those blocks should be pre-read | |
814 | (if they haven't been read already) so that the updated blocks | |
815 | can be written out properly. | |
afddba49 | 816 | |
ee5dc049 TH |
817 | The filesystem must return the locked pagecache page for the |
818 | specified offset, in ``*pagep``, for the caller to write into. | |
afddba49 | 819 | |
ee5dc049 TH |
820 | It must be able to cope with short writes (where the length |
821 | passed to write_begin is greater than the number of bytes copied | |
822 | into the page). | |
4e02ed4b | 823 | |
afddba49 NP |
824 | flags is a field for AOP_FLAG_xxx flags, described in |
825 | include/linux/fs.h. | |
826 | ||
1b44ae63 TH |
827 | A void * may be returned in fsdata, which then gets passed into |
828 | write_end. | |
afddba49 | 829 | |
ee5dc049 TH |
830 | Returns 0 on success; < 0 on failure (which is the error code), |
831 | in which case write_end is not called. | |
832 | ||
833 | ``write_end`` | |
834 | After a successful write_begin, and data copy, write_end must be | |
835 | called. len is the original len passed to write_begin, and | |
836 | copied is the amount that was able to be copied. | |
837 | ||
838 | The filesystem must take care of unlocking the page and | |
839 | releasing it refcount, and updating i_size. | |
840 | ||
841 | Returns < 0 on failure, otherwise the number of bytes (<= | |
842 | 'copied') that were able to be copied into pagecache. | |
843 | ||
844 | ``bmap`` | |
845 | called by the VFS to map a logical block offset within object to | |
846 | physical block number. This method is used by the FIBMAP ioctl | |
847 | and for working with swap-files. To be able to swap to a file, | |
848 | the file must have a stable mapping to a block device. The swap | |
849 | system does not go through the filesystem but instead uses bmap | |
850 | to find out where the blocks in the file are and uses those | |
851 | addresses directly. | |
852 | ||
853 | ``invalidatepage`` | |
854 | If a page has PagePrivate set, then invalidatepage will be | |
855 | called when part or all of the page is to be removed from the | |
856 | address space. This generally corresponds to either a | |
857 | truncation, punch hole or a complete invalidation of the address | |
d47992f8 | 858 | space (in the latter case 'offset' will always be 0 and 'length' |
4ee33ea4 | 859 | will be PAGE_SIZE). Any private data associated with the page |
ee5dc049 TH |
860 | should be updated to reflect this truncation. If offset is 0 |
861 | and length is PAGE_SIZE, then the private data should be | |
862 | released, because the page must be able to be completely | |
863 | discarded. This may be done by calling the ->releasepage | |
864 | function, but in this case the release MUST succeed. | |
865 | ||
866 | ``releasepage`` | |
867 | releasepage is called on PagePrivate pages to indicate that the | |
868 | page should be freed if possible. ->releasepage should remove | |
869 | any private data from the page and clear the PagePrivate flag. | |
870 | If releasepage() fails for some reason, it must indicate failure | |
871 | with a 0 return value. releasepage() is used in two distinct | |
872 | though related cases. The first is when the VM finds a clean | |
873 | page with no active users and wants to make it a free page. If | |
874 | ->releasepage succeeds, the page will be removed from the | |
875 | address_space and become free. | |
341546f5 | 876 | |
bc5b1d55 | 877 | The second case is when a request has been made to invalidate |
ee5dc049 TH |
878 | some or all pages in an address_space. This can happen through |
879 | the fadvise(POSIX_FADV_DONTNEED) system call or by the | |
880 | filesystem explicitly requesting it as nfs and 9fs do (when they | |
881 | believe the cache may be out of date with storage) by calling | |
882 | invalidate_inode_pages2(). If the filesystem makes such a call, | |
883 | and needs to be certain that all pages are invalidated, then its | |
884 | releasepage will need to ensure this. Possibly it can clear the | |
885 | PageUptodate bit if it cannot free private data yet. | |
886 | ||
887 | ``freepage`` | |
888 | freepage is called once the page is no longer visible in the | |
889 | page cache in order to allow the cleanup of any private data. | |
890 | Since it may be called by the memory reclaimer, it should not | |
891 | assume that the original address_space mapping still exists, and | |
892 | it should not block. | |
893 | ||
894 | ``direct_IO`` | |
895 | called by the generic read/write routines to perform direct_IO - | |
896 | that is IO requests which bypass the page cache and transfer | |
897 | data directly between the storage and the application's address | |
898 | space. | |
899 | ||
900 | ``isolate_page`` | |
901 | Called by the VM when isolating a movable non-lru page. If page | |
902 | is successfully isolated, VM marks the page as PG_isolated via | |
903 | __SetPageIsolated. | |
904 | ||
905 | ``migrate_page`` | |
906 | This is used to compact the physical memory usage. If the VM | |
907 | wants to relocate a page (maybe off a memory card that is | |
908 | signalling imminent failure) it will pass a new page and an old | |
909 | page to this function. migrate_page should transfer any private | |
910 | data across and update any references that it has to the page. | |
911 | ||
912 | ``putback_page`` | |
913 | Called by the VM when isolated page's migration fails. | |
914 | ||
915 | ``launder_page`` | |
916 | Called before freeing a page - it writes back the dirty page. | |
917 | To prevent redirtying the page, it is kept locked during the | |
918 | whole operation. | |
919 | ||
920 | ``is_partially_uptodate`` | |
921 | Called by the VM when reading a file through the pagecache when | |
922 | the underlying blocksize != pagesize. If the required block is | |
923 | up to date then the read can complete without needing the IO to | |
924 | bring the whole page up to date. | |
925 | ||
926 | ``is_dirty_writeback`` | |
927 | Called by the VM when attempting to reclaim a page. The VM uses | |
928 | dirty and writeback information to determine if it needs to | |
929 | stall to allow flushers a chance to complete some IO. | |
930 | Ordinarily it can use PageDirty and PageWriteback but some | |
931 | filesystems have more complex state (unstable pages in NFS | |
932 | prevent reclaim) or do not set those flags due to locking | |
933 | problems. This callback allows a filesystem to indicate to the | |
934 | VM if a page should be treated as dirty or writeback for the | |
935 | purposes of stalling. | |
936 | ||
937 | ``error_remove_page`` | |
938 | normally set to generic_error_remove_page if truncation is ok | |
939 | for this address space. Used for memory failure handling. | |
25718736 AK |
940 | Setting this implies you deal with pages going away under you, |
941 | unless you have them locked or reference counts increased. | |
942 | ||
ee5dc049 TH |
943 | ``swap_activate`` |
944 | Called when swapon is used on a file to allocate space if | |
945 | necessary and pin the block lookup information in memory. A | |
946 | return value of zero indicates success, in which case this file | |
947 | can be used to back swapspace. | |
62c230bc | 948 | |
ee5dc049 TH |
949 | ``swap_deactivate`` |
950 | Called during swapoff on files where swap_activate was | |
951 | successful. | |
62c230bc | 952 | |
25718736 | 953 | |
cc7d1f8f PE |
954 | The File Object |
955 | =============== | |
956 | ||
4ee33ea4 | 957 | A file object represents a file opened by a process. This is also known |
acbf3c34 | 958 | as an "open file description" in POSIX parlance. |
cc7d1f8f PE |
959 | |
960 | ||
5ea626aa | 961 | struct file_operations |
cc7d1f8f | 962 | ---------------------- |
1da177e4 | 963 | |
4ee33ea4 | 964 | This describes how the VFS can manipulate an open file. As of kernel |
17ef445f | 965 | 4.18, the following members are defined: |
1da177e4 | 966 | |
af96c1e3 TH |
967 | .. code-block:: c |
968 | ||
969 | struct file_operations { | |
970 | struct module *owner; | |
971 | loff_t (*llseek) (struct file *, loff_t, int); | |
972 | ssize_t (*read) (struct file *, char __user *, size_t, loff_t *); | |
973 | ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *); | |
974 | ssize_t (*read_iter) (struct kiocb *, struct iov_iter *); | |
975 | ssize_t (*write_iter) (struct kiocb *, struct iov_iter *); | |
976 | int (*iopoll)(struct kiocb *kiocb, bool spin); | |
977 | int (*iterate) (struct file *, struct dir_context *); | |
978 | int (*iterate_shared) (struct file *, struct dir_context *); | |
979 | __poll_t (*poll) (struct file *, struct poll_table_struct *); | |
980 | long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long); | |
981 | long (*compat_ioctl) (struct file *, unsigned int, unsigned long); | |
982 | int (*mmap) (struct file *, struct vm_area_struct *); | |
983 | int (*open) (struct inode *, struct file *); | |
984 | int (*flush) (struct file *, fl_owner_t id); | |
985 | int (*release) (struct inode *, struct file *); | |
986 | int (*fsync) (struct file *, loff_t, loff_t, int datasync); | |
987 | int (*fasync) (int, struct file *, int); | |
988 | int (*lock) (struct file *, int, struct file_lock *); | |
989 | ssize_t (*sendpage) (struct file *, struct page *, int, size_t, loff_t *, int); | |
990 | unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); | |
991 | int (*check_flags)(int); | |
992 | int (*flock) (struct file *, int, struct file_lock *); | |
993 | ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *, size_t, unsigned int); | |
994 | ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *, size_t, unsigned int); | |
995 | int (*setlease)(struct file *, long, struct file_lock **, void **); | |
996 | long (*fallocate)(struct file *file, int mode, loff_t offset, | |
997 | loff_t len); | |
998 | void (*show_fdinfo)(struct seq_file *m, struct file *f); | |
999 | #ifndef CONFIG_MMU | |
1000 | unsigned (*mmap_capabilities)(struct file *); | |
1001 | #endif | |
1002 | ssize_t (*copy_file_range)(struct file *, loff_t, struct file *, loff_t, size_t, unsigned int); | |
1003 | loff_t (*remap_file_range)(struct file *file_in, loff_t pos_in, | |
1004 | struct file *file_out, loff_t pos_out, | |
1005 | loff_t len, unsigned int remap_flags); | |
1006 | int (*fadvise)(struct file *, loff_t, loff_t, int); | |
1007 | }; | |
1da177e4 LT |
1008 | |
1009 | Again, all methods are called without any locks being held, unless | |
1010 | otherwise noted. | |
1011 | ||
ee5dc049 TH |
1012 | ``llseek`` |
1013 | called when the VFS needs to move the file position index | |
1da177e4 | 1014 | |
ee5dc049 TH |
1015 | ``read`` |
1016 | called by read(2) and related system calls | |
1da177e4 | 1017 | |
ee5dc049 TH |
1018 | ``read_iter`` |
1019 | possibly asynchronous read with iov_iter as destination | |
5ea626aa | 1020 | |
ee5dc049 TH |
1021 | ``write`` |
1022 | called by write(2) and related system calls | |
1da177e4 | 1023 | |
ee5dc049 TH |
1024 | ``write_iter`` |
1025 | possibly asynchronous write with iov_iter as source | |
5ea626aa | 1026 | |
ee5dc049 TH |
1027 | ``iopoll`` |
1028 | called when aio wants to poll for completions on HIPRI iocbs | |
fb7e1600 | 1029 | |
ee5dc049 TH |
1030 | ``iterate`` |
1031 | called when the VFS needs to read the directory contents | |
1da177e4 | 1032 | |
ee5dc049 TH |
1033 | ``iterate_shared`` |
1034 | called when the VFS needs to read the directory contents when | |
1035 | filesystem supports concurrent dir iterators | |
17ef445f | 1036 | |
ee5dc049 TH |
1037 | ``poll`` |
1038 | called by the VFS when a process wants to check if there is | |
1da177e4 | 1039 | activity on this file and (optionally) go to sleep until there |
4ee33ea4 | 1040 | is activity. Called by the select(2) and poll(2) system calls |
1da177e4 | 1041 | |
ee5dc049 TH |
1042 | ``unlocked_ioctl`` |
1043 | called by the ioctl(2) system call. | |
5ea626aa | 1044 | |
ee5dc049 TH |
1045 | ``compat_ioctl`` |
1046 | called by the ioctl(2) system call when 32 bit system calls are | |
1047 | used on 64 bit kernels. | |
5ea626aa | 1048 | |
ee5dc049 TH |
1049 | ``mmap`` |
1050 | called by the mmap(2) system call | |
1da177e4 | 1051 | |
ee5dc049 TH |
1052 | ``open`` |
1053 | called by the VFS when an inode should be opened. When the VFS | |
4ee33ea4 TH |
1054 | opens a file, it creates a new "struct file". It then calls the |
1055 | open method for the newly allocated file structure. You might | |
ee5dc049 TH |
1056 | think that the open method really belongs in "struct |
1057 | inode_operations", and you may be right. I think it's done the | |
1058 | way it is because it makes filesystems simpler to implement. | |
1059 | The open() method is a good place to initialize the | |
5ea626aa PE |
1060 | "private_data" member in the file structure if you want to point |
1061 | to a device structure | |
1062 | ||
ee5dc049 TH |
1063 | ``flush`` |
1064 | called by the close(2) system call to flush a file | |
1da177e4 | 1065 | |
ee5dc049 TH |
1066 | ``release`` |
1067 | called when the last reference to an open file is closed | |
1da177e4 | 1068 | |
ee5dc049 TH |
1069 | ``fsync`` |
1070 | called by the fsync(2) system call. Also see the section above | |
1071 | entitled "Handling errors during writeback". | |
1da177e4 | 1072 | |
ee5dc049 TH |
1073 | ``fasync`` |
1074 | called by the fcntl(2) system call when asynchronous | |
1da177e4 LT |
1075 | (non-blocking) mode is enabled for a file |
1076 | ||
ee5dc049 TH |
1077 | ``lock`` |
1078 | called by the fcntl(2) system call for F_GETLK, F_SETLK, and | |
1079 | F_SETLKW commands | |
5ea626aa | 1080 | |
ee5dc049 TH |
1081 | ``get_unmapped_area`` |
1082 | called by the mmap(2) system call | |
5ea626aa | 1083 | |
ee5dc049 TH |
1084 | ``check_flags`` |
1085 | called by the fcntl(2) system call for F_SETFL command | |
5ea626aa | 1086 | |
ee5dc049 TH |
1087 | ``flock`` |
1088 | called by the flock(2) system call | |
5ea626aa | 1089 | |
ee5dc049 TH |
1090 | ``splice_write`` |
1091 | called by the VFS to splice data from a pipe to a file. This | |
1092 | method is used by the splice(2) system call | |
d1195c51 | 1093 | |
ee5dc049 TH |
1094 | ``splice_read`` |
1095 | called by the VFS to splice data from file to a pipe. This | |
1096 | method is used by the splice(2) system call | |
d1195c51 | 1097 | |
ee5dc049 TH |
1098 | ``setlease`` |
1099 | called by the VFS to set or release a file lock lease. setlease | |
1100 | implementations should call generic_setlease to record or remove | |
1101 | the lease in the inode after setting it. | |
17cf28af | 1102 | |
ee5dc049 TH |
1103 | ``fallocate`` |
1104 | called by the VFS to preallocate blocks or punch a hole. | |
17cf28af | 1105 | |
ee5dc049 TH |
1106 | ``copy_file_range`` |
1107 | called by the copy_file_range(2) system call. | |
17ef445f | 1108 | |
ee5dc049 TH |
1109 | ``remap_file_range`` |
1110 | called by the ioctl(2) system call for FICLONERANGE and FICLONE | |
1111 | and FIDEDUPERANGE commands to remap file ranges. An | |
1112 | implementation should remap len bytes at pos_in of the source | |
1113 | file into the dest file at pos_out. Implementations must handle | |
1114 | callers passing in len == 0; this means "remap to the end of the | |
1115 | source file". The return value should the number of bytes | |
1116 | remapped, or the usual negative error code if errors occurred | |
1117 | before any bytes were remapped. The remap_flags parameter | |
1118 | accepts REMAP_FILE_* flags. If REMAP_FILE_DEDUP is set then the | |
1119 | implementation must only remap if the requested file ranges have | |
cb56ecae | 1120 | identical contents. If REMAP_FILE_CAN_SHORTEN is set, the caller is |
ee5dc049 TH |
1121 | ok with the implementation shortening the request length to |
1122 | satisfy alignment or EOF requirements (or any other reason). | |
17ef445f | 1123 | |
ee5dc049 TH |
1124 | ``fadvise`` |
1125 | possibly called by the fadvise64() system call. | |
45cd0faa | 1126 | |
1da177e4 | 1127 | Note that the file operations are implemented by the specific |
4ee33ea4 | 1128 | filesystem in which the inode resides. When opening a device node |
1da177e4 LT |
1129 | (character or block special) most filesystems will call special |
1130 | support routines in the VFS which will locate the required device | |
4ee33ea4 | 1131 | driver information. These support routines replace the filesystem file |
1da177e4 | 1132 | operations with those for the device driver, and then proceed to call |
4ee33ea4 | 1133 | the new open() method for the file. This is how opening a device file |
1da177e4 | 1134 | in the filesystem eventually ends up calling the device driver open() |
5ea626aa | 1135 | method. |
1da177e4 LT |
1136 | |
1137 | ||
5ea626aa PE |
1138 | Directory Entry Cache (dcache) |
1139 | ============================== | |
1140 | ||
1da177e4 LT |
1141 | |
1142 | struct dentry_operations | |
5ea626aa | 1143 | ------------------------ |
1da177e4 LT |
1144 | |
1145 | This describes how a filesystem can overload the standard dentry | |
4ee33ea4 TH |
1146 | operations. Dentries and the dcache are the domain of the VFS and the |
1147 | individual filesystem implementations. Device drivers have no business | |
1148 | here. These methods may be set to NULL, as they are either optional or | |
1149 | the VFS uses a default. As of kernel 2.6.22, the following members are | |
1da177e4 LT |
1150 | defined: |
1151 | ||
af96c1e3 TH |
1152 | .. code-block:: c |
1153 | ||
1154 | struct dentry_operations { | |
1155 | int (*d_revalidate)(struct dentry *, unsigned int); | |
1156 | int (*d_weak_revalidate)(struct dentry *, unsigned int); | |
1157 | int (*d_hash)(const struct dentry *, struct qstr *); | |
1158 | int (*d_compare)(const struct dentry *, | |
1159 | unsigned int, const char *, const struct qstr *); | |
1160 | int (*d_delete)(const struct dentry *); | |
1161 | int (*d_init)(struct dentry *); | |
1162 | void (*d_release)(struct dentry *); | |
1163 | void (*d_iput)(struct dentry *, struct inode *); | |
1164 | char *(*d_dname)(struct dentry *, char *, int); | |
1165 | struct vfsmount *(*d_automount)(struct path *); | |
1166 | int (*d_manage)(const struct path *, bool); | |
1167 | struct dentry *(*d_real)(struct dentry *, const struct inode *); | |
1168 | }; | |
1169 | ||
ee5dc049 TH |
1170 | ``d_revalidate`` |
1171 | called when the VFS needs to revalidate a dentry. This is | |
1172 | called whenever a name look-up finds a dentry in the dcache. | |
1173 | Most local filesystems leave this as NULL, because all their | |
1174 | dentries in the dcache are valid. Network filesystems are | |
1175 | different since things can change on the server without the | |
1176 | client necessarily being aware of it. | |
1177 | ||
1178 | This function should return a positive value if the dentry is | |
1179 | still valid, and zero or a negative error code if it isn't. | |
1180 | ||
1181 | d_revalidate may be called in rcu-walk mode (flags & | |
1182 | LOOKUP_RCU). If in rcu-walk mode, the filesystem must | |
1183 | revalidate the dentry without blocking or storing to the dentry, | |
1184 | d_parent and d_inode should not be used without care (because | |
1185 | they can change and, in d_inode case, even become NULL under | |
1186 | us). | |
1187 | ||
1188 | If a situation is encountered that rcu-walk cannot handle, | |
1189 | return | |
34286d66 NP |
1190 | -ECHILD and it will be called again in ref-walk mode. |
1191 | ||
ee5dc049 TH |
1192 | ``_weak_revalidate`` |
1193 | called when the VFS needs to revalidate a "jumped" dentry. This | |
1194 | is called when a path-walk ends at dentry that was not acquired | |
1195 | by doing a lookup in the parent directory. This includes "/", | |
1196 | "." and "..", as well as procfs-style symlinks and mountpoint | |
1197 | traversal. | |
ecf3d1f1 | 1198 | |
ee5dc049 TH |
1199 | In this case, we are less concerned with whether the dentry is |
1200 | still fully correct, but rather that the inode is still valid. | |
1201 | As with d_revalidate, most local filesystems will set this to | |
1202 | NULL since their dcache entries are always valid. | |
ecf3d1f1 | 1203 | |
ee5dc049 TH |
1204 | This function has the same return code semantics as |
1205 | d_revalidate. | |
ecf3d1f1 JL |
1206 | |
1207 | d_weak_revalidate is only called after leaving rcu-walk mode. | |
1208 | ||
ee5dc049 TH |
1209 | ``d_hash`` |
1210 | called when the VFS adds a dentry to the hash table. The first | |
621e155a | 1211 | dentry passed to d_hash is the parent directory that the name is |
da53be12 | 1212 | to be hashed into. |
b1e6a015 NP |
1213 | |
1214 | Same locking and synchronisation rules as d_compare regarding | |
1215 | what is safe to dereference etc. | |
1da177e4 | 1216 | |
ee5dc049 TH |
1217 | ``d_compare`` |
1218 | called to compare a dentry name with a given name. The first | |
621e155a | 1219 | dentry is the parent of the dentry to be compared, the second is |
ee5dc049 TH |
1220 | the child dentry. len and name string are properties of the |
1221 | dentry to be compared. qstr is the name to compare it with. | |
621e155a NP |
1222 | |
1223 | Must be constant and idempotent, and should not take locks if | |
ee5dc049 TH |
1224 | possible, and should not or store into the dentry. Should not |
1225 | dereference pointers outside the dentry without lots of care | |
1226 | (eg. d_parent, d_inode, d_name should not be used). | |
1227 | ||
1228 | However, our vfsmount is pinned, and RCU held, so the dentries | |
1229 | and inodes won't disappear, neither will our sb or filesystem | |
1230 | module. ->d_sb may be used. | |
1231 | ||
1232 | It is a tricky calling convention because it needs to be called | |
1233 | under "rcu-walk", ie. without any locks or references on things. | |
1234 | ||
1235 | ``d_delete`` | |
1236 | called when the last reference to a dentry is dropped and the | |
1237 | dcache is deciding whether or not to cache it. Return 1 to | |
1238 | delete immediately, or 0 to cache the dentry. Default is NULL | |
1239 | which means to always cache a reachable dentry. d_delete must | |
1240 | be constant and idempotent. | |
1241 | ||
1242 | ``d_init`` | |
1243 | called when a dentry is allocated | |
1244 | ||
1245 | ``d_release`` | |
1246 | called when a dentry is really deallocated | |
1247 | ||
1248 | ``d_iput`` | |
1249 | called when a dentry loses its inode (just prior to its being | |
1250 | deallocated). The default when this is NULL is that the VFS | |
1251 | calls iput(). If you define this method, you must call iput() | |
1252 | yourself | |
1253 | ||
1254 | ``d_dname`` | |
1255 | called when the pathname of a dentry should be generated. | |
1256 | Useful for some pseudo filesystems (sockfs, pipefs, ...) to | |
1257 | delay pathname generation. (Instead of doing it when dentry is | |
1258 | created, it's done only when the path is needed.). Real | |
1259 | filesystems probably dont want to use it, because their dentries | |
1260 | are present in global dcache hash, so their hash should be an | |
1261 | invariant. As no lock is held, d_dname() should not try to | |
1262 | modify the dentry itself, unless appropriate SMP safety is used. | |
1263 | CAUTION : d_path() logic is quite tricky. The correct way to | |
1264 | return for example "Hello" is to put it at the end of the | |
1265 | buffer, and returns a pointer to the first char. | |
1266 | dynamic_dname() helper function is provided to take care of | |
1267 | this. | |
c23fbb6b | 1268 | |
0cac643c MS |
1269 | Example : |
1270 | ||
af96c1e3 TH |
1271 | .. code-block:: c |
1272 | ||
0cac643c MS |
1273 | static char *pipefs_dname(struct dentry *dent, char *buffer, int buflen) |
1274 | { | |
1275 | return dynamic_dname(dentry, buffer, buflen, "pipe:[%lu]", | |
1276 | dentry->d_inode->i_ino); | |
1277 | } | |
1278 | ||
ee5dc049 TH |
1279 | ``d_automount`` |
1280 | called when an automount dentry is to be traversed (optional). | |
1281 | This should create a new VFS mount record and return the record | |
1282 | to the caller. The caller is supplied with a path parameter | |
1283 | giving the automount directory to describe the automount target | |
1284 | and the parent VFS mount record to provide inheritable mount | |
1285 | parameters. NULL should be returned if someone else managed to | |
1286 | make the automount first. If the vfsmount creation failed, then | |
1287 | an error code should be returned. If -EISDIR is returned, then | |
1288 | the directory will be treated as an ordinary directory and | |
1289 | returned to pathwalk to continue walking. | |
1290 | ||
1291 | If a vfsmount is returned, the caller will attempt to mount it | |
1292 | on the mountpoint and will remove the vfsmount from its | |
1293 | expiration list in the case of failure. The vfsmount should be | |
1294 | returned with 2 refs on it to prevent automatic expiration - the | |
1295 | caller will clean up the additional ref. | |
1296 | ||
1297 | This function is only used if DCACHE_NEED_AUTOMOUNT is set on | |
1298 | the dentry. This is set by __d_instantiate() if S_AUTOMOUNT is | |
1299 | set on the inode being added. | |
1300 | ||
1301 | ``d_manage`` | |
1302 | called to allow the filesystem to manage the transition from a | |
1303 | dentry (optional). This allows autofs, for example, to hold up | |
1304 | clients waiting to explore behind a 'mountpoint' while letting | |
1305 | the daemon go past and construct the subtree there. 0 should be | |
1306 | returned to let the calling process continue. -EISDIR can be | |
1307 | returned to tell pathwalk to use this directory as an ordinary | |
1308 | directory and to ignore anything mounted on it and not to check | |
1309 | the automount flag. Any other error code will abort pathwalk | |
1310 | completely. | |
cc53ce53 | 1311 | |
ab90911f | 1312 | If the 'rcu_walk' parameter is true, then the caller is doing a |
ee5dc049 TH |
1313 | pathwalk in RCU-walk mode. Sleeping is not permitted in this |
1314 | mode, and the caller can be asked to leave it and call again by | |
1315 | returning -ECHILD. -EISDIR may also be returned to tell | |
1316 | pathwalk to ignore d_automount or any mounts. | |
ab90911f | 1317 | |
ee5dc049 TH |
1318 | This function is only used if DCACHE_MANAGE_TRANSIT is set on |
1319 | the dentry being transited from. | |
cc53ce53 | 1320 | |
ee5dc049 TH |
1321 | ``d_real`` |
1322 | overlay/union type filesystems implement this method to return | |
1323 | one of the underlying dentries hidden by the overlay. It is | |
1324 | used in two different modes: | |
c23fbb6b | 1325 | |
ee5dc049 TH |
1326 | Called from file_dentry() it returns the real dentry matching |
1327 | the inode argument. The real dentry may be from a lower layer | |
1328 | already copied up, but still referenced from the file. This | |
1329 | mode is selected with a non-NULL inode argument. | |
e698b8a4 | 1330 | |
fb16043b | 1331 | With NULL inode the topmost real underlying dentry is returned. |
c23fbb6b | 1332 | |
1da177e4 | 1333 | Each dentry has a pointer to its parent dentry, as well as a hash list |
4ee33ea4 | 1334 | of child dentries. Child dentries are basically like files in a |
1da177e4 LT |
1335 | directory. |
1336 | ||
5ea626aa | 1337 | |
cc7d1f8f | 1338 | Directory Entry Cache API |
1da177e4 LT |
1339 | -------------------------- |
1340 | ||
1341 | There are a number of functions defined which permit a filesystem to | |
1342 | manipulate dentries: | |
1343 | ||
ee5dc049 TH |
1344 | ``dget`` |
1345 | open a new handle for an existing dentry (this just increments | |
1da177e4 LT |
1346 | the usage count) |
1347 | ||
ee5dc049 TH |
1348 | ``dput`` |
1349 | close a handle for a dentry (decrements the usage count). If | |
fe15ce44 NP |
1350 | the usage count drops to 0, and the dentry is still in its |
1351 | parent's hash, the "d_delete" method is called to check whether | |
ee5dc049 TH |
1352 | it should be cached. If it should not be cached, or if the |
1353 | dentry is not hashed, it is deleted. Otherwise cached dentries | |
1354 | are put into an LRU list to be reclaimed on memory shortage. | |
1355 | ||
1356 | ``d_drop`` | |
1357 | this unhashes a dentry from its parents hash list. A subsequent | |
1358 | call to dput() will deallocate the dentry if its usage count | |
1359 | drops to 0 | |
1360 | ||
1361 | ``d_delete`` | |
1362 | delete a dentry. If there are no other open references to the | |
1363 | dentry then the dentry is turned into a negative dentry (the | |
1364 | d_iput() method is called). If there are other references, then | |
1365 | d_drop() is called instead | |
1366 | ||
1367 | ``d_add`` | |
1368 | add a dentry to its parents hash list and then calls | |
1da177e4 LT |
1369 | d_instantiate() |
1370 | ||
ee5dc049 TH |
1371 | ``d_instantiate`` |
1372 | add a dentry to the alias hash list for the inode and updates | |
1373 | the "d_inode" member. The "i_count" member in the inode | |
1374 | structure should be set/incremented. If the inode pointer is | |
1375 | NULL, the dentry is called a "negative dentry". This function | |
1376 | is commonly called when an inode is created for an existing | |
1377 | negative dentry | |
1378 | ||
1379 | ``d_lookup`` | |
1380 | look up a dentry given its parent and path name component It | |
1381 | looks up the child of that given name from the dcache hash | |
1382 | table. If it is found, the reference count is incremented and | |
1383 | the dentry is returned. The caller must use dput() to free the | |
1384 | dentry when it finishes using it. | |
1da177e4 | 1385 | |
e04c83cd | 1386 | |
f84e3f52 MS |
1387 | Mount Options |
1388 | ============= | |
1389 | ||
e04c83cd | 1390 | |
f84e3f52 MS |
1391 | Parsing options |
1392 | --------------- | |
1393 | ||
1394 | On mount and remount the filesystem is passed a string containing a | |
1395 | comma separated list of mount options. The options can have either of | |
1396 | these forms: | |
1397 | ||
1398 | option | |
1399 | option=value | |
1400 | ||
1401 | The <linux/parser.h> header defines an API that helps parse these | |
1402 | options. There are plenty of examples on how to use it in existing | |
1403 | filesystems. | |
1404 | ||
e04c83cd | 1405 | |
f84e3f52 MS |
1406 | Showing options |
1407 | --------------- | |
1408 | ||
90caa781 TH |
1409 | If a filesystem accepts mount options, it must define show_options() to |
1410 | show all the currently active options. The rules are: | |
f84e3f52 MS |
1411 | |
1412 | - options MUST be shown which are not default or their values differ | |
1413 | from the default | |
1414 | ||
1415 | - options MAY be shown which are enabled by default or have their | |
1416 | default value | |
1417 | ||
90caa781 TH |
1418 | Options used only internally between a mount helper and the kernel (such |
1419 | as file descriptors), or which only have an effect during the mounting | |
1420 | (such as ones controlling the creation of a journal) are exempt from the | |
1421 | above rules. | |
f84e3f52 | 1422 | |
90caa781 TH |
1423 | The underlying reason for the above rules is to make sure, that a mount |
1424 | can be accurately replicated (e.g. umounting and mounting again) based | |
1425 | on the information found in /proc/mounts. | |
f84e3f52 | 1426 | |
e04c83cd | 1427 | |
cc7d1f8f PE |
1428 | Resources |
1429 | ========= | |
1430 | ||
1431 | (Note some of these resources are not up-to-date with the latest kernel | |
1432 | version.) | |
1433 | ||
1434 | Creating Linux virtual filesystems. 2002 | |
c69f22f2 | 1435 | <https://lwn.net/Articles/13325/> |
cc7d1f8f PE |
1436 | |
1437 | The Linux Virtual File-system Layer by Neil Brown. 1999 | |
1438 | <http://www.cse.unsw.edu.au/~neilb/oss/linux-commentary/vfs.html> | |
1439 | ||
1440 | A tour of the Linux VFS by Michael K. Johnson. 1996 | |
c69f22f2 | 1441 | <https://www.tldp.org/LDP/khg/HyperNews/get/fs/vfstour.html> |
cc7d1f8f PE |
1442 | |
1443 | A small trail through the Linux kernel by Andries Brouwer. 2001 | |
c69f22f2 | 1444 | <https://www.win.tue.nl/~aeb/linux/vfs/trail.html> |