(RHSA-2010:0291) Moderate: gfs-kmod security, bug fix and enhancement update

The gfs-kmod packages contain modules that provide the ability to mount and
use GFS file systems.

A flaw was found in the gfs_lock() implementation. The GFS locking code
could skip the lock operation for files that have the S_ISGID bit
(set-group-ID on execution) in their mode set. A local, unprivileged user
on a system that has a GFS file system mounted could use this flaw to cause
a kernel panic. (CVE-2010-0727)

These updated gfs-kmod packages are in sync with the latest kernel
(2.6.18-194.el5). The modules in earlier gfs-kmod packages failed to load
because they did not match the running kernel. It was possible to
force-load the modules. With this update, however, users no longer need to.

These updated gfs-kmod packages also fix the following bugs:

• when SELinux was in permissive mode, a race condition during file
  creation could have caused one or more cluster nodes to be fenced and lock
  the remaining nodes out of the GFS file system. This race condition no
  longer occurs with this update. (BZ#471258)

• when ACLs (Access Control Lists) are enabled on a GFS file system, if a
  transaction that has started to do a write request does not have enough
  spare blocks for the operation it causes a kernel panic. This update
  ensures that there are enough blocks for the write request before starting
  the operation. (BZ#513885)

• requesting a “flock” on a file in GFS in either read-only or read-write
  mode would sometimes cause a “Resource temporarily unavailable” state error
  (error 11 for EWOULDBLOCK) to occur. In these cases, a flock could not be
  obtained on the file in question. This has been fixed with this update so
  that flocks can successfully be obtained on GFS files without this error
  occurring. (BZ#515717)

• the GFS withdraw function is a data integrity feature of GFS file systems
  in a cluster. If the GFS kernel module detects an inconsistency in a GFS
  file system following an I/O operation, the file system becomes unavailable
  to the cluster. The GFS withdraw function is less severe than a kernel
  panic, which would cause another node to fence the node. With this update,
  you can override the GFS withdraw function by mounting the file system with
  the “-o errors=panic” option specified. When this option is specified, any
  errors that would normally cause the system to withdraw cause the system to
  panic instead. This stops the node’s cluster communications, which causes
  the node to be fenced. (BZ#517145)

Finally, these updated gfs-kmod packages provide the following enhancement:

• the GFS kernel modules have been updated to use the new generic freeze
  and unfreeze ioctl interface that is also supported by the following file
  systems: ext3, ext4, GFS2, JFS and ReiserFS. With this update, GFS supports
  freeze/unfreeze through the VFS-level FIFREEZE/FITHAW ioctl interface.
  (BZ#487610)

Users are advised to upgrade to these latest gfs-kmod packages, updated for
use with the 2.6.18-194.el5 kernel, which contain backported patches to
correct these issues, fix these bugs, and add this enhancement.