| blob | 39d02e19fb6288f79769f356d7759a997bf9461f |
1 <head>
2 <style> p { max-width:50em} ol, ul {max-width: 40em}</style>
3 </head>
5 autofs - how it works
6 =====================
8 Purpose
9 -------
11 The goal of autofs is to provide on-demand mounting and race free
12 automatic unmounting of various other filesystems. This provides two
13 key advantages:
15 1. There is no need to delay boot until all filesystems that
16 might be needed are mounted. Processes that try to access those
17 slow filesystems might be delayed but other processes can
18 continue freely. This is particularly important for
19 network filesystems (e.g. NFS) or filesystems stored on
20 media with a media-changing robot.
22 2. The names and locations of filesystems can be stored in
23 a remote database and can change at any time. The content
24 in that data base at the time of access will be used to provide
25 a target for the access. The interpretation of names in the
26 filesystem can even be programmatic rather than database-backed,
27 allowing wildcards for example, and can vary based on the user who
28 first accessed a name.
30 Context
31 -------
33 The "autofs4" filesystem module is only one part of an autofs system.
34 There also needs to be a user-space program which looks up names
35 and mounts filesystems. This will often be the "automount" program,
36 though other tools including "systemd" can make use of "autofs4".
37 This document describes only the kernel module and the interactions
38 required with any user-space program. Subsequent text refers to this
39 as the "automount daemon" or simply "the daemon".
41 "autofs4" is a Linux kernel module with provides the "autofs"
42 filesystem type. Several "autofs" filesystems can be mounted and they
43 can each be managed separately, or all managed by the same daemon.
45 Content
46 -------
48 An autofs filesystem can contain 3 sorts of objects: directories,
49 symbolic links and mount traps. Mount traps are directories with
50 extra properties as described in the next section.
52 Objects can only be created by the automount daemon: symlinks are
53 created with a regular `symlink` system call, while directories and
54 mount traps are created with `mkdir`. The determination of whether a
55 directory should be a mount trap or not is quite _ad hoc_, largely for
56 historical reasons, and is determined in part by the
57 *direct*/*indirect*/*offset* mount options, and the *maxproto* mount option.
59 If neither the *direct* or *offset* mount options are given (so the
60 mount is considered to be *indirect*), then the root directory is
61 always a regular directory, otherwise it is a mount trap when it is
62 empty and a regular directory when not empty. Note that *direct* and
63 *offset* are treated identically so a concise summary is that the root
64 directory is a mount trap only if the filesystem is mounted *direct*
65 and the root is empty.
67 Directories created in the root directory are mount traps only if the
68 filesystem is mounted *indirect* and they are empty.
70 Directories further down the tree depend on the *maxproto* mount
71 option and particularly whether it is less than five or not.
72 When *maxproto* is five, no directories further down the
73 tree are ever mount traps, they are always regular directories. When
74 the *maxproto* is four (or three), these directories are mount traps
75 precisely when they are empty.
77 So: non-empty (i.e. non-leaf) directories are never mount traps. Empty
78 directories are sometimes mount traps, and sometimes not depending on
79 where in the tree they are (root, top level, or lower), the *maxproto*,
80 and whether the mount was *indirect* or not.
82 Mount Traps
83 ---------------
85 A core element of the implementation of autofs is the Mount Traps
86 which are provided by the Linux VFS. Any directory provided by a
87 filesystem can be designated as a trap. This involves two separate
88 features that work together to allow autofs to do its job.
90 **DCACHE_NEED_AUTOMOUNT**
92 If a dentry has the DCACHE_NEED_AUTOMOUNT flag set (which gets set if
93 the inode has S_AUTOMOUNT set, or can be set directly) then it is
94 (potentially) a mount trap. Any access to this directory beyond a
95 "`stat`" will (normally) cause the `d_op->d_automount()` dentry operation
96 to be called. The task of this method is to find the filesystem that
97 should be mounted on the directory and to return it. The VFS is
98 responsible for actually mounting the root of this filesystem on the
99 directory.
101 autofs doesn't find the filesystem itself but sends a message to the
102 automount daemon asking it to find and mount the filesystem. The
103 autofs `d_automount` method then waits for the daemon to report that
104 everything is ready. It will then return "`NULL`" indicating that the
105 mount has already happened. The VFS doesn't try to mount anything but
106 follows down the mount that is already there.
108 This functionality is sufficient for some users of mount traps such
109 as NFS which creates traps so that mountpoints on the server can be
110 reflected on the client. However it is not sufficient for autofs. As
111 mounting onto a directory is considered to be "beyond a `stat`", the
112 automount daemon would not be able to mount a filesystem on the 'trap'
113 directory without some way to avoid getting caught in the trap. For
114 that purpose there is another flag.
116 **DCACHE_MANAGE_TRANSIT**
118 If a dentry has DCACHE_MANAGE_TRANSIT set then two very different but
119 related behaviors are invoked, both using the `d_op->d_manage()`
120 dentry operation.
122 Firstly, before checking to see if any filesystem is mounted on the
123 directory, d_manage() will be called with the `rcu_walk` parameter set
124 to `false`. It may return one of three things:
126 - A return value of zero indicates that there is nothing special
127 about this dentry and normal checks for mounts and automounts
128 should proceed.
130 autofs normally returns zero, but first waits for any
131 expiry (automatic unmounting of the mounted filesystem) to
132 complete. This avoids races.
134 - A return value of `-EISDIR` tells the VFS to ignore any mounts
135 on the directory and to not consider calling `->d_automount()`.
136 This effectively disables the **DCACHE_NEED_AUTOMOUNT** flag
137 causing the directory not be a mount trap after all.
139 autofs returns this if it detects that the process performing the
140 lookup is the automount daemon and that the mount has been
141 requested but has not yet completed. How it determines this is
142 discussed later. This allows the automount daemon not to get
143 caught in the mount trap.
145 There is a subtlety here. It is possible that a second autofs
146 filesystem can be mounted below the first and for both of them to
147 be managed by the same daemon. For the daemon to be able to mount
148 something on the second it must be able to "walk" down past the
149 first. This means that d_manage cannot *always* return -EISDIR for
150 the automount daemon. It must only return it when a mount has
151 been requested, but has not yet completed.
153 `d_manage` also returns `-EISDIR` if the dentry shouldn't be a
154 mount trap, either because it is a symbolic link or because it is
155 not empty.
157 - Any other negative value is treated as an error and returned
158 to the caller.
160 autofs can return
162 - -ENOENT if the automount daemon failed to mount anything,
163 - -ENOMEM if it ran out of memory,
164 - -EINTR if a signal arrived while waiting for expiry to
165 complete
166 - or any other error sent down by the automount daemon.
169 The second use case only occurs during an "RCU-walk" and so `rcu_walk`
170 will be set.
172 An RCU-walk is a fast and lightweight process for walking down a
173 filename path (i.e. it is like running on tip-toes). RCU-walk cannot
174 cope with all situations so when it finds a difficulty it falls back
175 to "REF-walk", which is slower but more robust.
177 RCU-walk will never call `->d_automount`; the filesystems must already
178 be mounted or RCU-walk cannot handle the path.
179 To determine if a mount-trap is safe for RCU-walk mode it calls
180 `->d_manage()` with `rcu_walk` set to `true`.
182 In this case `d_manage()` must avoid blocking and should avoid taking
183 spinlocks if at all possible. Its sole purpose is to determine if it
184 would be safe to follow down into any mounted directory and the only
185 reason that it might not be is if an expiry of the mount is
186 underway.
188 In the `rcu_walk` case, `d_manage()` cannot return -EISDIR to tell the
189 VFS that this is a directory that doesn't require d_automount. If
190 `rcu_walk` sees a dentry with DCACHE_NEED_AUTOMOUNT set but nothing
191 mounted, it *will* fall back to REF-walk. `d_manage()` cannot make the
192 VFS remain in RCU-walk mode, but can only tell it to get out of
193 RCU-walk mode by returning `-ECHILD`.
195 So `d_manage()`, when called with `rcu_walk` set, should either return
196 -ECHILD if there is any reason to believe it is unsafe to end the
197 mounted filesystem, and otherwise should return 0.
199 autofs will return `-ECHILD` if an expiry of the filesystem has been
200 initiated or is being considered, otherwise it returns 0.
203 Mountpoint expiry
204 -----------------
206 The VFS has a mechansim for automatically expiring unused mounts,
207 much as it can expire any unused dentry information from the dcache.
208 This is guided by the MNT_SHRINKABLE flag. This only applies to
209 mounts that were created by `d_automount()` returning a filesystem to be
210 mounted. As autofs doesn't return such a filesystem but leaves the
211 mounting to the automount daemon, it must involve the automount daemon
212 in unmounting as well. This also means that autofs has more control
213 of expiry.
215 The VFS also supports "expiry" of mounts using the MNT_EXPIRE flag to
216 the `umount` system call. Unmounting with MNT_EXPIRE will fail unless
217 a previous attempt had been made, and the filesystem has been inactive
218 and untouched since that previous attempt. autofs4 does not depend on
219 this but has its own internal tracking of whether filesystems were
220 recently used. This allows individual names in the autofs directory
221 to expire separately.
223 With version 4 of the protocol, the automount daemon can try to
224 unmount any filesystems mounted on the autofs filesystem or remove any
225 symbolic links or empty directories any time it likes. If the unmount
226 or removal is successful the filesystem will be returned to the state
227 it was before the mount or creation, so that any access of the name
228 will trigger normal auto-mount processing. In particlar, `rmdir` and
229 `unlink` do not leave negative entries in the dcache as a normal
230 filesystem would, so an attempt to access a recently-removed object is
231 passed to autofs for handling.
233 With version 5, this is not safe except for unmounting from top-level
234 directories. As lower-level directories are never mount traps, other
235 processes will see an empty directory as soon as the filesystem is
236 unmounted. So it is generally safest to use the autofs expiry
237 protocol described below.
239 Normally the daemon only wants to remove entries which haven't been
240 used for a while. For this purpose autofs maintains a "`last_used`"
241 time stamp on each directory or symlink. For symlinks it genuinely
242 does record the last time the symlink was "used" or followed to find
243 out where it points to. For directories the field is a slight
244 misnomer. It actually records the last time that autofs checked if
245 the directory or one of its descendents was busy and found that it
246 was. This is just as useful and doesn't require updating the field so
247 often.
249 The daemon is able to ask autofs if anything is due to be expired,
250 using an `ioctl` as discussed later. For a *direct* mount, autofs
251 considers if the entire mount-tree can be unmounted or not. For an
252 *indirect* mount, autofs considers each of the names in the top level
253 directory to determine if any of those can be unmounted and cleaned
254 up.
256 There is an option with indirect mounts to consider each of the leaves
257 that has been mounted on instead of considering the top-level names.
258 This is intended for compatability with version 4 of autofs and should
259 be considered as deprecated.
261 When autofs considers a directory it checks the `last_used` time and
262 compares it with the "timeout" value set when the filesystem was
263 mounted, though this check is ignored in some cases. It also checks if
264 the directory or anything below it is in use. For symbolic links,
265 only the `last_used` time is ever considered.
267 If both appear to support expiring the directory or symlink, an action
268 is taken.
270 There are two ways to ask autofs to consider expiry. The first is to
271 use the **AUTOFS_IOC_EXPIRE** ioctl. This only works for indirect
272 mounts. If it finds something in the root directory to expire it will
273 return the name of that thing. Once a name has been returned the
274 automount daemon needs to unmount any filesystems mounted below the
275 name normally. As described above, this is unsafe for non-toplevel
276 mounts in a version-5 autofs. For this reason the current `automountd`
277 does not use this ioctl.
279 The second mechanism uses either the **AUTOFS_DEV_IOCTL_EXPIRE_CMD** or
280 the **AUTOFS_IOC_EXPIRE_MULTI** ioctl. This will work for both direct and
281 indirect mounts. If it selects an object to expire, it will notify
282 the daemon using the notification mechanism described below. This
283 will block until the daemon acknowledges the expiry notification.
284 This implies that the "`EXPIRE`" ioctl must be sent from a different
285 thread than the one which handles notification.
287 While the ioctl is blocking, the entry is marked as "expiring" and
288 `d_manage` will block until the daemon affirms that the unmount has
289 completed (together with removing any directories that might have been
290 necessary), or has been aborted.
292 Communicating with autofs: detecting the daemon
293 -----------------------------------------------
295 There are several forms of communication between the automount daemon
296 and the filesystem. As we have already seen, the daemon can create and
297 remove directories and symlinks using normal filesystem operations.
298 autofs knows whether a process requesting some operation is the daemon
299 or not based on its process-group id number (see getpgid(1)).
301 When an autofs filesystem it mounted the pgid of the mounting
302 processes is recorded unless the "pgrp=" option is given, in which
303 case that number is recorded instead. Any request arriving from a
304 process in that process group is considered to come from the daemon.
305 If the daemon ever has to be stopped and restarted a new pgid can be
306 provided through an ioctl as will be described below.
308 Communicating with autofs: the event pipe
309 -----------------------------------------
311 When an autofs filesystem is mounted, the 'write' end of a pipe must
312 be passed using the 'fd=' mount option. autofs will write
313 notification messages to this pipe for the daemon to respond to.
314 For version 5, the format of the message is:
316 struct autofs_v5_packet {
317 int proto_version; /* Protocol version */
318 int type; /* Type of packet */
319 autofs_wqt_t wait_queue_token;
320 __u32 dev;
321 __u64 ino;
322 __u32 uid;
323 __u32 gid;
324 __u32 pid;
325 __u32 tgid;
326 __u32 len;
327 char name[NAME_MAX+1];
328 };
330 where the type is one of
332 autofs_ptype_missing_indirect
333 autofs_ptype_expire_indirect
334 autofs_ptype_missing_direct
335 autofs_ptype_expire_direct
337 so messages can indicate that a name is missing (something tried to
338 access it but it isn't there) or that it has been selected for expiry.
340 The pipe will be set to "packet mode" (equivalent to passing
341 `O_DIRECT`) to _pipe2(2)_ so that a read from the pipe will return at
342 most one packet, and any unread portion of a packet will be discarded.
344 The `wait_queue_token` is a unique number which can identify a
345 particular request to be acknowledged. When a message is sent over
346 the pipe the affected dentry is marked as either "active" or
347 "expiring" and other accesses to it block until the message is
348 acknowledged using one of the ioctls below and the relevant
349 `wait_queue_token`.
351 Communicating with autofs: root directory ioctls
352 ------------------------------------------------
354 The root directory of an autofs filesystem will respond to a number of
355 ioctls. The process issuing the ioctl must have the CAP_SYS_ADMIN
356 capability, or must be the automount daemon.
358 The available ioctl commands are:
360 - **AUTOFS_IOC_READY**: a notification has been handled. The argument
361 to the ioctl command is the "wait_queue_token" number
362 corresponding to the notification being acknowledged.
363 - **AUTOFS_IOC_FAIL**: similar to above, but indicates failure with
364 the error code `ENOENT`.
365 - **AUTOFS_IOC_CATATONIC**: Causes the autofs to enter "catatonic"
366 mode meaning that it stops sending notifications to the daemon.
367 This mode is also entered if a write to the pipe fails.
368 - **AUTOFS_IOC_PROTOVER**: This returns the protocol version in use.
369 - **AUTOFS_IOC_PROTOSUBVER**: Returns the protocol sub-version which
370 is really a version number for the implementation. It is
371 currently 2.
372 - **AUTOFS_IOC_SETTIMEOUT**: This passes a pointer to an unsigned
373 long. The value is used to set the timeout for expiry, and
374 the current timeout value is stored back through the pointer.
375 - **AUTOFS_IOC_ASKUMOUNT**: Returns, in the pointed-to `int`, 1 if
376 the filesystem could be unmounted. This is only a hint as
377 the situation could change at any instant. This call can be
378 use to avoid a more expensive full unmount attempt.
379 - **AUTOFS_IOC_EXPIRE**: as described above, this asks if there is
380 anything suitable to expire. A pointer to a packet:
382 struct autofs_packet_expire_multi {
383 int proto_version; /* Protocol version */
384 int type; /* Type of packet */
385 autofs_wqt_t wait_queue_token;
386 int len;
387 char name[NAME_MAX+1];
388 };
390 is required. This is filled in with the name of something
391 that can be unmounted or removed. If nothing can be expired,
392 `errno` is set to `EAGAIN`. Even though a `wait_queue_token`
393 is present in the structure, no "wait queue" is established
394 and no acknowledgment is needed.
395 - **AUTOFS_IOC_EXPIRE_MULTI**: This is similar to
396 **AUTOFS_IOC_EXPIRE** except that it causes notification to be
397 sent to the daemon, and it blocks until the daemon acknowledges.
398 The argument is an integer which can contain two different flags.
400 **AUTOFS_EXP_IMMEDIATE** causes `last_used` time to be ignored
401 and objects are expired if the are not in use.
403 **AUTOFS_EXP_LEAVES** will select a leaf rather than a top-level
404 name to expire. This is only safe when *maxproto* is 4.
406 Communicating with autofs: char-device ioctls
407 ---------------------------------------------
409 It is not always possible to open the root of an autofs filesystem,
410 particularly a *direct* mounted filesystem. If the automount daemon
411 is restarted there is no way for it to regain control of existing
412 mounts using any of the above communication channels. To address this
413 need there is a "miscellaneous" character device (major 10, minor 235)
414 which can be used to communicate directly with the autofs filesystem.
415 It requires CAP_SYS_ADMIN for access.
417 The `ioctl`s that can be used on this device are described in a separate
418 document `autofs4-mount-control.txt`, and are summarized briefly here.
419 Each ioctl is passed a pointer to an `autofs_dev_ioctl` structure:
421 struct autofs_dev_ioctl {
422 __u32 ver_major;
423 __u32 ver_minor;
424 __u32 size; /* total size of data passed in
425 * including this struct */
426 __s32 ioctlfd; /* automount command fd */
428 __u32 arg1; /* Command parameters */
429 __u32 arg2;
431 char path[0];
432 };
434 For the **OPEN_MOUNT** and **IS_MOUNTPOINT** commands, the target
435 filesystem is identified by the `path`. All other commands identify
436 the filesystem by the `ioctlfd` which is a file descriptor open on the
437 root, and which can be returned by **OPEN_MOUNT**.
439 The `ver_major` and `ver_minor` are in/out parameters which check that
440 the requested version is supported, and report the maximum version
441 that the kernel module can support.
443 Commands are:
445 - **AUTOFS_DEV_IOCTL_VERSION_CMD**: does nothing, except validate and
446 set version numbers.
447 - **AUTOFS_DEV_IOCTL_OPENMOUNT_CMD**: return an open file descriptor
448 on the root of an autofs filesystem. The filesystem is identified
449 by name and device number, which is stored in `arg1`. Device
450 numbers for existing filesystems can be found in
451 `/proc/self/mountinfo`.
452 - **AUTOFS_DEV_IOCTL_CLOSEMOUNT_CMD**: same as `close(ioctlfd)`.
453 - **AUTOFS_DEV_IOCTL_SETPIPEFD_CMD**: if the filesystem is in
454 catatonic mode, this can provide the write end of a new pipe
455 in `arg1` to re-establish communication with a daemon. The
456 process group of the calling process is used to identify the
457 daemon.
458 - **AUTOFS_DEV_IOCTL_REQUESTER_CMD**: `path` should be a
459 name within the filesystem that has been auto-mounted on.
460 arg1 is the dev number of the underlying autofs. On successful
461 return, `arg1` and `arg2` will be the UID and GID of the process
462 which triggered that mount.
464 - **AUTOFS_DEV_IOCTL_ISMOUNTPOINT_CMD**: Check if path is a
465 mountpoint of a particular type - see separate documentation for
466 details.
468 - **AUTOFS_DEV_IOCTL_PROTOVER_CMD**:
469 - **AUTOFS_DEV_IOCTL_PROTOSUBVER_CMD**:
470 - **AUTOFS_DEV_IOCTL_READY_CMD**:
471 - **AUTOFS_DEV_IOCTL_FAIL_CMD**:
472 - **AUTOFS_DEV_IOCTL_CATATONIC_CMD**:
473 - **AUTOFS_DEV_IOCTL_TIMEOUT_CMD**:
474 - **AUTOFS_DEV_IOCTL_EXPIRE_CMD**:
475 - **AUTOFS_DEV_IOCTL_ASKUMOUNT_CMD**: These all have the same
476 function as the similarly named **AUTOFS_IOC** ioctls, except
477 that **FAIL** can be given an explicit error number in `arg1`
478 instead of assuming `ENOENT`, and this **EXPIRE** command
479 corresponds to **AUTOFS_IOC_EXPIRE_MULTI**.
481 Catatonic mode
482 --------------
484 As mentioned, an autofs mount can enter "catatonic" mode. This
485 happens if a write to the notification pipe fails, or if it is
486 explicitly requested by an `ioctl`.
488 When entering catatonic mode, the pipe is closed and any pending
489 notifications are acknowledged with the error `ENOENT`.
491 Once in catatonic mode attempts to access non-existing names will
492 result in `ENOENT` while attempts to access existing directories will
493 be treated in the same way as if they came from the daemon, so mount
494 traps will not fire.
496 When the filesystem is mounted a _uid_ and _gid_ can be given which
497 set the ownership of directories and symbolic links. When the
498 filesystem is in catatonic mode, any process with a matching UID can
499 create directories or symlinks in the root directory, but not in other
500 directories.
502 Catatonic mode can only be left via the
503 **AUTOFS_DEV_IOCTL_OPENMOUNT_CMD** ioctl on the `/dev/autofs`.
505 autofs, name spaces, and shared mounts
506 --------------------------------------
508 With bind mounts and name spaces it is possible for an autofs
509 filesystem to appear at multiple places in one or more filesystem
510 name spaces. For this to work sensibly, the autofs filesystem should
511 always be mounted "shared". e.g.
513 > `mount --make-shared /autofs/mount/point`
515 The automount daemon is only able to mange a single mount location for
516 an autofs filesystem and if mounts on that are not 'shared', other
517 locations will not behave as expected. In particular access to those
518 other locations will likely result in the `ELOOP` error
520 > Too many levels of symbolic links