automountdaemon. The daemon automatically mounts file systems when in use and unmounts them when they are not busy.
- Under certain circumstances, the autofs utility did not respect all configured settings and used the UDP protocol to probe availability of network file systems. This can lead to some servers refusing the connection with the message:
Client x.x.x.x is violating the NFSv4 specification by sending a UDP/IP datagram to the NFSv4 server.With this update, autofs has been modified to respect explicitly defined NFSv4 requests, thus fixing this bug.
- Due to changes made to the autofs utility, when probing server availability at mount time, mounts using the RDMA protocol are no longer recognized. With this update, autofs has been modified not to probe availability for mounts that use the RDMA protocol.
- Previously, the autofs utility ignored the
--random-multimount-selectionoption. Consequently, this setting was not used when mounting local file systems even when it was given. This bug has been fixed and
--random-multimount-selectionnow works as expected.
- Previously, when two nearly simultaneous mount requests appeared, NFS mounts mounted by the autofs utility sometimes terminated. This was caused by using invalid protoent structures to identify the protocol. With this update, autofs has been modified to use numeric protocol IDs, instead of protoent structures. As a result, attempts to mount NFS no longer fail in the described scenario.
- Prior to this update, the autofs master map parser did not recognize the
SELinux context=option and returned a syntax error when the option was used. The master map parser has been updated to recognize
SELinux context=that can now be used without complications.
- Previously, the autofs utility did not recognize the allowed limit of maximum opened files after it was increased by the system administrator. Consequently, the default limit was used regardless of the new configuration. With this update, autofs has been modified to check for changes of this limit and to apply them correctly.
- Previously, the
libldaplibrary was not initialized in a thread-safe manner. Consequently, when running automount, the
ber_memalloc_x()function could have terminated unexpectedly with a segmentation fault. With this update, the initializaliton of
libldaphas been modified to be thread-safe and
ber_memalloc_x()no longer crashes in the aforementioned scenario. (BZ#996749)
- When the
automountdaemon was checking host availability and one of the network interfaces was marked "DOWN",
automountterminated with a segmentation fault. With this update, a check for this case has been added and the segmentation fault no longer occurs.
- When the
automountdaemon received a shutdown signal, executing the
autofs reloadcommand caused
automountto stop running when multiple maps were being removed from the auto.master map. A patch has been added to fix this bug and
automountno longer terminates in the described case.
- A change that removed a code for adding the current map entry caused wildcard indirect multi-mount map entries to fail to mount. A patch to fix wildcard multi-map regression has been added and map entries now mount successfully.
- Due to an execution order race that occurred when creating an expire thread, the
automountdaemon became unresponsive. The code that handled the expire thread creation has been modified to prevent the aforementioned problem.
- Previously, no locking was performed around LDAP initialization calls. However, these functions are not thread-safe and race conditions could have occurred. With this update, the locking has been added and the risk of race condition is now reduced.
- The description of the
TIMEOUTconfiguration option has been enhanced in the autofs man page. The description now explains the internal default configuration more clearly.
- The autofs utility has been updated to provide the ability to dump its mount maps in a simple
<key, value>format in addition to the existing informational format.