(RHSA-2012:1541) Moderate: kernel security and bug fix update

These packages contain the Linux kernel.

Security fixes:

• A malicious NFSv4 server could return a crafted reply to a GETACL
  request, causing a denial of service on the client. (CVE-2011-4131,
  Moderate)

• A flaw in the dl2k driver could allow a local, unprivileged user to issue
  potentially harmful IOCTLs, possibly causing Ethernet adapters using the
  driver to malfunction (such as losing network connectivity).
  (CVE-2012-2313, Low)

Red Hat would like to thank Andy Adamson for reporting CVE-2011-4131, and
Stephan Mueller for reporting CVE-2012-2313.

Bug fixes:

• A kernel oops occurred in the nf_nat code when a bogus pointer was
  dereferenced in the nf_conn_nat structure. Consequently, if Source Network
  Address Translation (SNAT) was performed, incorrect information could be
  received by other CTS (Clear to Send) signals. A conntrack entry is now
  placed in the source hash after SNAT has been completed, which prevents the
  described problems. (BZ#865715)

• Previously, the ixgbe_setup_tc() function was called recursively when the
  set_state() CEE (Convergence Enhanced Ethernet) API routine was called in
  IEEE DCBX (Data Center Bridging eXchange) mode. This is considered unsafe
  according to the IEEE standards. With this update, the ixgbe driver has
  been modified to no longer call the set_state() routine in IEEE DCBX mode.
  The driver now calls routines of the PFC (Priority-based Flow Control) and
  ETS (Enhanced Transmission Selection) extensions instead of the CEE
  extension routines in IEEE DCBX mode. (BZ#867859)

• A Symmetric Multi Processing (SMP) race condition between the munmap()
  and exit() function could lead to false-positive triggering of the BUG_ON()
  macro if Transparent Huge Pages (THP) were enabled. This update fixes the
  race condition, which avoids false-positive triggering of the BUG_ON()
  macro in this scenario. (BZ#875121)

• The kernel allows high priority real time tasks, such as tasks scheduled
  with the SCHED_FIFO policy, to be throttled. Previously, the CPU stop tasks
  were scheduled as high priority real time tasks and could be thus throttled
  accordingly. However, the replenishment timer, which is responsible for
  clearing a throttle flag on tasks, could be pending on the just disabled
  CPU. This could lead to the situation that the throttled tasks were never
  scheduled to run. Consequently, if any of such tasks was needed to complete
  the CPU disabling, the system became unresponsive. This update introduces a
  new scheduler class, which gives a task the highest possible system
  priority and such a task cannot be throttled. The stop-task scheduling
  class is now used for the CPU stop tasks, and the system shutdown completes
  as expected in the scenario described. (BZ#876078)

• Previously, XFS log buffers were handled incorrectly so that XFS could,
  in certain circumstances, incorrectly read metadata from the journal during
  XFS log recovery. As a consequence, XFS log recovery terminated with an
  error message and prevented the file system from being mounted. This
  problem could result in a loss of data if the user forcibly emptied the log
  to allow the file system to be mounted. This update ensures that metadata
  is read correctly from the log and journal recovery thus completes
  successfully, and the file system mounts as expected. (BZ#876498)

• Previously, kernel was allowed to reduce the number of unnecessary commit
  calls by skipping the commit when there was a large number of outstanding
  pages being written. However, a test on the number of commits (ncommit) did
  not properly handle the edge case when ncommit was zero. Consequently,
  inodes sometimes remained on the sb->s_dirty list and could not be freed by
  the inode cache shrinker. As a result, the nfs_inode_cache structure grew
  very large over time. With this update, the call to the nfs_write_inode()
  function is immediately returned when commit == 0, thus fixing this bug.
  (BZ#877394)