1.14.  autofs

when connecting to an LDAP server while using SASL authentication, autofs occasionally failed with a segmentation fault, forcing users to restart the autofs service. This failure was caused by a double-free error in the cyrus-sasl module, which has been fixed in this updated package. Connecting to an LDAP server while using SASL authentication now works as expected. (BZ#504566)

Previously, automount did not return its status to its parent while it waited for the autofs daemon to complete its startup. As a result, the init script did not always report success when the service started sucessfully. Automount now returns its status and accurately reports when the service has started. (BZ#244177)

Autofs uses "umount -l" to clear active mounts at restart. This method results in getcwd() failing because the point from which the path is constructed has been detached from the mount tree. To resolve this a miscellaneous device node for routing ioctl commands to these mount points has been implemented in the autofs4 kernel module and a library added to autofs. This provides the ability to re-construct a mount tree from existing mounts and then re-connect them. (BZ#452122)

Previously, the version of autofs shipped with Red Hat Enterprise Linux 5 used the "-hosts" method as its default way to handle /net mounts. Using this method, it was necessary to reboot the client to release processes if if the connection to the server was lost. Now, autofs uses the "intr" option as its default, which allows the mount to be unmounted forcibly if necessary. (BZ#466673)

By default, autofs waits 60 seconds for a server to respond while performing a YP lookup. Previously, repeated attempts to perform lookups for non-existent directories could result in all available ports becoming congested. Autofs now maintains a cache of failed lookups and avoids repeated failures occupying the available ports. (BZ#469387)

The %{dist?} tag that is used by rpm spec files is defined in ~/.rpmmacros for the user building the package. However, this is not a reliable method of providing the "Release:" tag in a package, because the {%dist?} tag might not be defined for the user building the package. Previously, autofs relied on the {%dist?} tag to define "Release:" in its spec file, which meant that building it correctly depended on the user's ~/.rpmmacros file being set up appropriately. "Release:" is now defined directly in the autofs file system, which makes it more likely to build correctly on a greater number of systems. (BZ#471385)

Previously, the LDAP module lacked the ability to lock the server list. When used in SASL authenticated environments, this could cause autofs to fail if the credential for the connection became stale. The LDAP module can now lock a server list, and autofs refreshes and retries failed SASL connections. Autofs therefore performs more reliably when used in authenticated environments. (BZ#481139)

Submounts are detached threads that do not belong to the master map entry list. Previously, autofs did not release mount resources when a mount thread for a submount was terminated. With these resources not released, a segmentation fault during a shutdown or reboot of the system could result. Resources allocated to submounts are now explictly released in the code and the segmentation fault is therefore avoided. (BZ#482988)

Previously, autofs contained an an incorrect %token declaration in the master map parser. In some rare cases this could cause the timeout sent from the tokenizer to the parser to always be zero, which is interpreted as "never". As a result, indirect mounts would never expire, no matter how long they had been inactive. The %token declaration is now corrected, meaning that mounts expire as they should. (BZ#487151)

Previously, autofs used the select() function to process direct-mount maps and was therefore limited by the file descriptor limit (by default, 1024). As a consequence, autofs was not able to use direct-mount maps with numbers of entries larger than the limit, and would stop responding when it used up all available file descriptors. Now, autofs uses poll() instead of select() and is therefore no longer limited by the available file descriptors. Freed of this limitation, autofs can use large direct-mount maps. (BZ#487653)

Previously, autofs reported an incorrect buffer size internally when passing the startup status from the autofs daemon to the parent process. Although no specific consequences of this inaccuracy are known, the buffer size is now reported correctly to avoid any consequences arising in the future. (BZ#487656)

Previously, the additive hashing algorithm used by autofs to generate hash values would result in a clustering of values that favoured a small range of hash indexes and led to reduced performance in large maps. Autofs now uses a "one-at-a-time" hash function which gives a better distribution of hash values in large hash tables. Use of the "one-at-a-time" hash function safeguards lookup performance as maps increase to 8,000 entries and beyond. (BZ#487985)

Previously, autofs would not always read file maps. If a map had been loaded into cache, autofs would rely on checks to determine whether the map was up to date before reading the map. Because file maps require a linear search through the file, large maps consume significant resources to process. Now, autofs automatically loads file-based maps when it starts, and uses the map file mtime parameter to detemine whether the cache needs to be refresed. This avoids the processing overhead of checking a map before deciding whether to load it. (BZ#487986)

Previously, the autofs code contained a logic error that resulted in a crash under conditions of heavy load. When autofs was not able to create a new pthread, it would double free a value. Now, with the error corrected, when heavily loaded, autofs will fail to create a new pthread safely. It reports the failure, but does not crash. (BZ#489658)

Previously, autofs could use the LDAP server on a network only if the location of the LDAP server were specified manually. Now, if no LDAP server is specified, autofs can look up domain SRV server records to make LDAP connections. This functionality simplifies the use of autofs on networks where an LDAP server is available. (BZ#490476)

Previously, if a name lookup failed while creating a TCP or UDP client, automount would destroy the client, but would not set the rpc client to NULL. Therefore, subsequent lookup attempts would attempt to use the invalid rpc client, which would lead to a segmentation fault. Now, when a name lookup fails, autofs sets the rpc client to NULL, and therefore avoids the segmentation fault on subsequent lookup attempts. (BZ#491351)

Previously, in LDAP environments were both Red Hat Enterprise Linux and Solaris were in use, autofs would not correctly interpret master map keys added by Solaris. The auto_master file would therefore contain duplicate entries, where '%' symbols were interspersed between the characters of the map key names. Autofs now correctly parses the Solaris key names and does not create duplicate entries. (BZ#493074)

Previously, a stack variable was not initialized on entry to the create_udp_client() or create_tcp_client() functions. During an error exit, the stack variable was checked, and the corresponding file descriptor was closed if the variable had a value other than -1. This could result in incorrectly closing a file descriptor still in use. The stack variable is now initialized and descriptors currently in use should not be closed. (BZ#493223)

Due to a number of logic errors in the code, autofs could not remount a direct-mount NFS if the mount had expired following a map reload. The mount request would never complete, and "can't find map entry" would appear in the log. The logic errors are now fixed, and autofs can successfully remount an expired direct-mount NFS after a map reload. (BZ#493791)

Previously, thread locking was missing from the st_remove_tasks() function, which meant in turn that its calling function could not get the locks that it required. This could result in a segmentation fault and a crash of autofs. Now, with the thread locking properly in place, the segmentation fault is avoided. (BZ#494319)

Previously, when autofs looked up a host name where when one NFS server name was associated with multiple IP adresses, autofs would repeat the query many times. As a consequence of these multiple queries, the mount would take a long time. Now, redundant queries have been removed, so that autofs performs the mount more quickly. (BZ#495895)

When connecting to an LDAP server while using SASL authentication, autofs occasionally failed with a segmentation fault, forcing users to restart the autofs service. This failure was caused by a double-free error in the cyrus-sasl module, which has been fixed in this updated package. Connecting to an LDAP server while using SASL authentication now works as expected. (BZ#501612)

Previously, the method used by autofs to clean up pthreads was not reliable and could result in a memory leak. If the memory leak occurred, autofs would gradually consume all available memory and then crash. A small semantic change in the code prevents this memory leak from occurring now. (BZ#510530)