ChangesEdit
Since 042stab076.8:
- rebase to 2.6.32-358.6.2.el6 Red Hat kernel (security, bug fixes and enhancements, see below)
- implemented posix timer TID preserve support during checkpoint/restore (PSBM-19073)
- fixed kernel panic while resuming a Container due to an unexpected empty task signal handler (PSBM-19177, PCLIN-31850)
- fixed a number of issues found by Trinity test (PSBM-18611)
- eliminated a potential file system corruption by preventing pipe writes and fsyncs on a frozen file system. Those writes-after-freeze were accompanied by kernel warnings: " WARNING: at fs/ext4/super.c:248 ext4_journal_start_sb" (PSBM-19255, PSBM-19460)
- corrected CBQ-based traffic shaping algorithm to properly limit the bandwidth when the limit is greater than 50% of total bandwidth (PSBM-18636)
- enhanced checkpoint functionality to operate on the systems with the /tmp directory mounted with the "noexec" mount option (PSBM-17995)
- fixed NFS vzquota bug that caused random memory corruptions (PSBM-17619, #2506)
- fixed Ubuntu 12.04 LTS syslog-ng fail to log kernel messages due to internal capability CAP_FS_FREEZE which was introduced in the previous kernel update (PSBM-17503, #2567)
- fixed a memory leak of a struct file object (PCLIN-31738)
- fixed triggering a kernel BUG() in try_to_release_page() by memory shrinker on an attempt to strip buffer heads from swap-backed pages (PCLIN-31748)
- protected the quota sync operation by a mutex, preventing possible NFS quota corruptions (PCLIN-31715)
- reworked fsync locking in order to prevent possible node lockups (PSBM-19460)
- enhanced memory management algorithm to give more priority to containers with unlimited memory barriers set
- fixed kernel panic on nfs unmount issued inside a container if volume had been mounted on a Hardware Node (PCLIN-31620)
- corrected /proc/stat:btime values on a container after live migration. This should also fix glitches with process start time reports (#2572)
- fixed issue when container checkpoint/restore cycle could break process capabilities (PSBM-18649)
- fixed issue when under certain conditions direct memory reclaiming could trigger fs-related activity and could cause the Node deadlock in case of high memory pressure (PSBM-20210)
- fixed issue when checkpointing a container with a tty under inotify watch could trigger a kernel panic (PCLIN-31849)
- implemented NFSv4 support inside CT, plus online migration of such CTs is supported now (PSBM-17798, PSBM-17795)
- ploop deltas should be opened with O_DIRECT flag, otherwise it works in buffered mode and might deadlock (PSBM-18642)
- ploop writeback completion should be awaited on fsync (PSBM-18049)
- introduced "cpt.suspend_timeout_sec" sysctl -- time to wait for tasks to be frozen (PSBM-18523)
- performance optimization in sys_fairsched_mvpr() (PCLIN-26766)
- fixed vmscan deadlock near global oom point if there are no anonymous pages (PCLIN-31619)
- support for ploop size > 2Tb (PSBM-15670)
- ploop tweaks to increase the random write performance (PSBM-18017)