• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1
2Miscellaneous Device control operations for the autofs kernel module
3====================================================================
4
5The problem
6===========
7
8There is a problem with active restarts in autofs (that is to say
9restarting autofs when there are busy mounts).
10
11During normal operation autofs uses a file descriptor opened on the
12directory that is being managed in order to be able to issue control
13operations. Using a file descriptor gives ioctl operations access to
14autofs specific information stored in the super block. The operations
15are things such as setting an autofs mount catatonic, setting the
16expire timeout and requesting expire checks. As is explained below,
17certain types of autofs triggered mounts can end up covering an autofs
18mount itself which prevents us being able to use open(2) to obtain a
19file descriptor for these operations if we don't already have one open.
20
21Currently autofs uses "umount -l" (lazy umount) to clear active mounts
22at restart. While using lazy umount works for most cases, anything that
23needs to walk back up the mount tree to construct a path, such as
24getcwd(2) and the proc file system /proc/<pid>/cwd, no longer works
25because the point from which the path is constructed has been detached
26from the mount tree.
27
28The actual problem with autofs is that it can't reconnect to existing
29mounts. Immediately one thinks of just adding the ability to remount
30autofs file systems would solve it, but alas, that can't work. This is
31because autofs direct mounts and the implementation of "on demand mount
32and expire" of nested mount trees have the file system mounted directly
33on top of the mount trigger directory dentry.
34
35For example, there are two types of automount maps, direct (in the kernel
36module source you will see a third type called an offset, which is just
37a direct mount in disguise) and indirect.
38
39Here is a master map with direct and indirect map entries:
40
41/-      /etc/auto.direct
42/test   /etc/auto.indirect
43
44and the corresponding map files:
45
46/etc/auto.direct:
47
48/automount/dparse/g6  budgie:/autofs/export1
49/automount/dparse/g1  shark:/autofs/export1
50and so on.
51
52/etc/auto.indirect:
53
54g1    shark:/autofs/export1
55g6    budgie:/autofs/export1
56and so on.
57
58For the above indirect map an autofs file system is mounted on /test and
59mounts are triggered for each sub-directory key by the inode lookup
60operation. So we see a mount of shark:/autofs/export1 on /test/g1, for
61example.
62
63The way that direct mounts are handled is by making an autofs mount on
64each full path, such as /automount/dparse/g1, and using it as a mount
65trigger. So when we walk on the path we mount shark:/autofs/export1 "on
66top of this mount point". Since these are always directories we can
67use the follow_link inode operation to trigger the mount.
68
69But, each entry in direct and indirect maps can have offsets (making
70them multi-mount map entries).
71
72For example, an indirect mount map entry could also be:
73
74g1  \
75   /        shark:/autofs/export5/testing/test \
76   /s1      shark:/autofs/export/testing/test/s1 \
77   /s2      shark:/autofs/export5/testing/test/s2 \
78   /s1/ss1  shark:/autofs/export1 \
79   /s2/ss2  shark:/autofs/export2
80
81and a similarly a direct mount map entry could also be:
82
83/automount/dparse/g1 \
84    /       shark:/autofs/export5/testing/test \
85    /s1     shark:/autofs/export/testing/test/s1 \
86    /s2     shark:/autofs/export5/testing/test/s2 \
87    /s1/ss1 shark:/autofs/export2 \
88    /s2/ss2 shark:/autofs/export2
89
90One of the issues with version 4 of autofs was that, when mounting an
91entry with a large number of offsets, possibly with nesting, we needed
92to mount and umount all of the offsets as a single unit. Not really a
93problem, except for people with a large number of offsets in map entries.
94This mechanism is used for the well known "hosts" map and we have seen
95cases (in 2.4) where the available number of mounts are exhausted or
96where the number of privileged ports available is exhausted.
97
98In version 5 we mount only as we go down the tree of offsets and
99similarly for expiring them which resolves the above problem. There is
100somewhat more detail to the implementation but it isn't needed for the
101sake of the problem explanation. The one important detail is that these
102offsets are implemented using the same mechanism as the direct mounts
103above and so the mount points can be covered by a mount.
104
105The current autofs implementation uses an ioctl file descriptor opened
106on the mount point for control operations. The references held by the
107descriptor are accounted for in checks made to determine if a mount is
108in use and is also used to access autofs file system information held
109in the mount super block. So the use of a file handle needs to be
110retained.
111
112
113The Solution
114============
115
116To be able to restart autofs leaving existing direct, indirect and
117offset mounts in place we need to be able to obtain a file handle
118for these potentially covered autofs mount points. Rather than just
119implement an isolated operation it was decided to re-implement the
120existing ioctl interface and add new operations to provide this
121functionality.
122
123In addition, to be able to reconstruct a mount tree that has busy mounts,
124the uid and gid of the last user that triggered the mount needs to be
125available because these can be used as macro substitution variables in
126autofs maps. They are recorded at mount request time and an operation
127has been added to retrieve them.
128
129Since we're re-implementing the control interface, a couple of other
130problems with the existing interface have been addressed. First, when
131a mount or expire operation completes a status is returned to the
132kernel by either a "send ready" or a "send fail" operation. The
133"send fail" operation of the ioctl interface could only ever send
134ENOENT so the re-implementation allows user space to send an actual
135status. Another expensive operation in user space, for those using
136very large maps, is discovering if a mount is present. Usually this
137involves scanning /proc/mounts and since it needs to be done quite
138often it can introduce significant overhead when there are many entries
139in the mount table. An operation to lookup the mount status of a mount
140point dentry (covered or not) has also been added.
141
142Current kernel development policy recommends avoiding the use of the
143ioctl mechanism in favor of systems such as Netlink. An implementation
144using this system was attempted to evaluate its suitability and it was
145found to be inadequate, in this case. The Generic Netlink system was
146used for this as raw Netlink would lead to a significant increase in
147complexity. There's no question that the Generic Netlink system is an
148elegant solution for common case ioctl functions but it's not a complete
149replacement probably because its primary purpose in life is to be a
150message bus implementation rather than specifically an ioctl replacement.
151While it would be possible to work around this there is one concern
152that lead to the decision to not use it. This is that the autofs
153expire in the daemon has become far to complex because umount
154candidates are enumerated, almost for no other reason than to "count"
155the number of times to call the expire ioctl. This involves scanning
156the mount table which has proved to be a big overhead for users with
157large maps. The best way to improve this is try and get back to the
158way the expire was done long ago. That is, when an expire request is
159issued for a mount (file handle) we should continually call back to
160the daemon until we can't umount any more mounts, then return the
161appropriate status to the daemon. At the moment we just expire one
162mount at a time. A Generic Netlink implementation would exclude this
163possibility for future development due to the requirements of the
164message bus architecture.
165
166
167autofs Miscellaneous Device mount control interface
168====================================================
169
170The control interface is opening a device node, typically /dev/autofs.
171
172All the ioctls use a common structure to pass the needed parameter
173information and return operation results:
174
175struct autofs_dev_ioctl {
176	__u32 ver_major;
177	__u32 ver_minor;
178	__u32 size;             /* total size of data passed in
179				 * including this struct */
180	__s32 ioctlfd;          /* automount command fd */
181
182	/* Command parameters */
183	union {
184		struct args_protover		protover;
185		struct args_protosubver		protosubver;
186		struct args_openmount		openmount;
187		struct args_ready		ready;
188		struct args_fail		fail;
189		struct args_setpipefd		setpipefd;
190		struct args_timeout		timeout;
191		struct args_requester		requester;
192		struct args_expire		expire;
193		struct args_askumount		askumount;
194		struct args_ismountpoint	ismountpoint;
195	};
196
197	char path[0];
198};
199
200The ioctlfd field is a mount point file descriptor of an autofs mount
201point. It is returned by the open call and is used by all calls except
202the check for whether a given path is a mount point, where it may
203optionally be used to check a specific mount corresponding to a given
204mount point file descriptor, and when requesting the uid and gid of the
205last successful mount on a directory within the autofs file system.
206
207The union is used to communicate parameters and results of calls made
208as described below.
209
210The path field is used to pass a path where it is needed and the size field
211is used account for the increased structure length when translating the
212structure sent from user space.
213
214This structure can be initialized before setting specific fields by using
215the void function call init_autofs_dev_ioctl(struct autofs_dev_ioctl *).
216
217All of the ioctls perform a copy of this structure from user space to
218kernel space and return -EINVAL if the size parameter is smaller than
219the structure size itself, -ENOMEM if the kernel memory allocation fails
220or -EFAULT if the copy itself fails. Other checks include a version check
221of the compiled in user space version against the module version and a
222mismatch results in a -EINVAL return. If the size field is greater than
223the structure size then a path is assumed to be present and is checked to
224ensure it begins with a "/" and is NULL terminated, otherwise -EINVAL is
225returned. Following these checks, for all ioctl commands except
226AUTOFS_DEV_IOCTL_VERSION_CMD, AUTOFS_DEV_IOCTL_OPENMOUNT_CMD and
227AUTOFS_DEV_IOCTL_CLOSEMOUNT_CMD the ioctlfd is validated and if it is
228not a valid descriptor or doesn't correspond to an autofs mount point
229an error of -EBADF, -ENOTTY or -EINVAL (not an autofs descriptor) is
230returned.
231
232
233The ioctls
234==========
235
236An example of an implementation which uses this interface can be seen
237in autofs version 5.0.4 and later in file lib/dev-ioctl-lib.c of the
238distribution tar available for download from kernel.org in directory
239/pub/linux/daemons/autofs/v5.
240
241The device node ioctl operations implemented by this interface are:
242
243
244AUTOFS_DEV_IOCTL_VERSION
245------------------------
246
247Get the major and minor version of the autofs device ioctl kernel module
248implementation. It requires an initialized struct autofs_dev_ioctl as an
249input parameter and sets the version information in the passed in structure.
250It returns 0 on success or the error -EINVAL if a version mismatch is
251detected.
252
253
254AUTOFS_DEV_IOCTL_PROTOVER_CMD and AUTOFS_DEV_IOCTL_PROTOSUBVER_CMD
255------------------------------------------------------------------
256
257Get the major and minor version of the autofs protocol version understood
258by loaded module. This call requires an initialized struct autofs_dev_ioctl
259with the ioctlfd field set to a valid autofs mount point descriptor
260and sets the requested version number in version field of struct args_protover
261or sub_version field of struct args_protosubver. These commands return
2620 on success or one of the negative error codes if validation fails.
263
264
265AUTOFS_DEV_IOCTL_OPENMOUNT and AUTOFS_DEV_IOCTL_CLOSEMOUNT
266----------------------------------------------------------
267
268Obtain and release a file descriptor for an autofs managed mount point
269path. The open call requires an initialized struct autofs_dev_ioctl with
270the path field set and the size field adjusted appropriately as well
271as the devid field of struct args_openmount set to the device number of
272the autofs mount. The device number can be obtained from the mount options
273shown in /proc/mounts. The close call requires an initialized struct
274autofs_dev_ioct with the ioctlfd field set to the descriptor obtained
275from the open call. The release of the file descriptor can also be done
276with close(2) so any open descriptors will also be closed at process exit.
277The close call is included in the implemented operations largely for
278completeness and to provide for a consistent user space implementation.
279
280
281AUTOFS_DEV_IOCTL_READY_CMD and AUTOFS_DEV_IOCTL_FAIL_CMD
282--------------------------------------------------------
283
284Return mount and expire result status from user space to the kernel.
285Both of these calls require an initialized struct autofs_dev_ioctl
286with the ioctlfd field set to the descriptor obtained from the open
287call and the token field of struct args_ready or struct args_fail set
288to the wait queue token number, received by user space in the foregoing
289mount or expire request. The status field of struct args_fail is set to
290the errno of the operation. It is set to 0 on success.
291
292
293AUTOFS_DEV_IOCTL_SETPIPEFD_CMD
294------------------------------
295
296Set the pipe file descriptor used for kernel communication to the daemon.
297Normally this is set at mount time using an option but when reconnecting
298to a existing mount we need to use this to tell the autofs mount about
299the new kernel pipe descriptor. In order to protect mounts against
300incorrectly setting the pipe descriptor we also require that the autofs
301mount be catatonic (see next call).
302
303The call requires an initialized struct autofs_dev_ioctl with the
304ioctlfd field set to the descriptor obtained from the open call and
305the pipefd field of struct args_setpipefd set to descriptor of the pipe.
306On success the call also sets the process group id used to identify the
307controlling process (eg. the owning automount(8) daemon) to the process
308group of the caller.
309
310
311AUTOFS_DEV_IOCTL_CATATONIC_CMD
312------------------------------
313
314Make the autofs mount point catatonic. The autofs mount will no longer
315issue mount requests, the kernel communication pipe descriptor is released
316and any remaining waits in the queue released.
317
318The call requires an initialized struct autofs_dev_ioctl with the
319ioctlfd field set to the descriptor obtained from the open call.
320
321
322AUTOFS_DEV_IOCTL_TIMEOUT_CMD
323----------------------------
324
325Set the expire timeout for mounts within an autofs mount point.
326
327The call requires an initialized struct autofs_dev_ioctl with the
328ioctlfd field set to the descriptor obtained from the open call.
329
330
331AUTOFS_DEV_IOCTL_REQUESTER_CMD
332------------------------------
333
334Return the uid and gid of the last process to successfully trigger a the
335mount on the given path dentry.
336
337The call requires an initialized struct autofs_dev_ioctl with the path
338field set to the mount point in question and the size field adjusted
339appropriately. Upon return the uid field of struct args_requester contains
340the uid and gid field the gid.
341
342When reconstructing an autofs mount tree with active mounts we need to
343re-connect to mounts that may have used the original process uid and
344gid (or string variations of them) for mount lookups within the map entry.
345This call provides the ability to obtain this uid and gid so they may be
346used by user space for the mount map lookups.
347
348
349AUTOFS_DEV_IOCTL_EXPIRE_CMD
350---------------------------
351
352Issue an expire request to the kernel for an autofs mount. Typically
353this ioctl is called until no further expire candidates are found.
354
355The call requires an initialized struct autofs_dev_ioctl with the
356ioctlfd field set to the descriptor obtained from the open call. In
357addition an immediate expire that's independent of the mount timeout,
358and a forced expire that's independent of whether the mount is busy,
359can be requested by setting the how field of struct args_expire to
360AUTOFS_EXP_IMMEDIATE or AUTOFS_EXP_FORCED, respectively . If no
361expire candidates can be found the ioctl returns -1 with errno set to
362EAGAIN.
363
364This call causes the kernel module to check the mount corresponding
365to the given ioctlfd for mounts that can be expired, issues an expire
366request back to the daemon and waits for completion.
367
368AUTOFS_DEV_IOCTL_ASKUMOUNT_CMD
369------------------------------
370
371Checks if an autofs mount point is in use.
372
373The call requires an initialized struct autofs_dev_ioctl with the
374ioctlfd field set to the descriptor obtained from the open call and
375it returns the result in the may_umount field of struct args_askumount,
3761 for busy and 0 otherwise.
377
378
379AUTOFS_DEV_IOCTL_ISMOUNTPOINT_CMD
380---------------------------------
381
382Check if the given path is a mountpoint.
383
384The call requires an initialized struct autofs_dev_ioctl. There are two
385possible variations. Both use the path field set to the path of the mount
386point to check and the size field adjusted appropriately. One uses the
387ioctlfd field to identify a specific mount point to check while the other
388variation uses the path and optionally in.type field of struct args_ismountpoint
389set to an autofs mount type. The call returns 1 if this is a mount point
390and sets out.devid field to the device number of the mount and out.magic
391field to the relevant super block magic number (described below) or 0 if
392it isn't a mountpoint. In both cases the the device number (as returned
393by new_encode_dev()) is returned in out.devid field.
394
395If supplied with a file descriptor we're looking for a specific mount,
396not necessarily at the top of the mounted stack. In this case the path
397the descriptor corresponds to is considered a mountpoint if it is itself
398a mountpoint or contains a mount, such as a multi-mount without a root
399mount. In this case we return 1 if the descriptor corresponds to a mount
400point and and also returns the super magic of the covering mount if there
401is one or 0 if it isn't a mountpoint.
402
403If a path is supplied (and the ioctlfd field is set to -1) then the path
404is looked up and is checked to see if it is the root of a mount. If a
405type is also given we are looking for a particular autofs mount and if
406a match isn't found a fail is returned. If the the located path is the
407root of a mount 1 is returned along with the super magic of the mount
408or 0 otherwise.
409