Changes
Since 042stab053.5:
- Rebased on 2.6.32-220.13.1.el6 RHEL6.2 kernel (security, bug fixes, and enhancements, see below)
- [procfs] make reading /proc/mounts consistent. Now we should get much less timeouts on CT stops (PSBM-12119)
- [ext4] online resize has been significantly enhanced (PSBM-12197)
- [timer] kmemsize leaks for 5457 has been fixed (PSBM-11885)
- [oom] now OOM can effectively struggle against fork bombs (PSBM-11920)
- 5 possible memory leaks were fixed in various places, found internally
- [bridge] kernel panic in bridge code in case NAT rules are configured improperly (PSBM-11782)
- [cpt] kernel panic in case a dead task gets into cpt image (PCLIN-30811)
- [cpt] we do align the size of cpt image. Online migration could fail due to tar which exits before reading the align tail. (PCLIN-30804)
- [fuse] FUSE improvements (write-back cache, AIO support, all off by default)
- [ploop] possible deadlock detected by lockdep fixed (PSBM-12210)
- [block] support for in-kernel direct aio added. (patch-set backports a few Zach Brown's patches from Chris Mason) (PCLIN-30658)
- [idle] add boot param 'intel_auto_demotion=force' to forcibly enable intel auto demotion. For debug (PSBM-11477)
- [proc] kernel panic in /proc dentry put fixed (PSBM-12306)
- [ploop] add discard support
- [bridge] tcpdump shows incoming traffic in bridged veth CT (PCLIN-30814)
- [block] order of nested locking fixed in partition resize ioctl
- [ext4/ploop] preserve balloon_ino option on remount. Otherwise this option is dropped when some distrib scripts (unaware of ploop/balloon) remount root inside a CT. Use "balloon_ino=0" mount option if you need to really drop balloon. (#2225, PSBM-12337)
- [cpt] bypass limits on inotify files creation during CT checkpoint (PSBM-12475)
- [ploop] possible deadlock fixed
- [ploop] fixes for in-kernel grow of expanded images (#2226)
- [ploop] out-of-space error message has ben made more descriptive
- [raid] avoid deadlock on RAID1/RAID10 during resync recover
- [ubc] livelock in bdev cache allocation fixed (PSBM-12523)
- [cpt] a CT with tun device inside could not be migrated twice (PSBM-12502)
- [ploop] ENOSPC threshold defaults set: 2Gb for root, 4Gb for users. These thresholds can be changed via ploop.ko parameters
- [ploop] debug messages on error paths added
- [ubc] UB ram limit check skipped for __GFP_NOFAIL allocations. This prevents possible deadlocks (PSBM-12523)
- [ploop] max disk-space preallocation limited to 128MB independently on ploop cluster-block size. This reduces initial size of ploop image (PSBM-12546)
- [megaraid_sas] mask off flags in the ioctl path to prevent memory scribble with older MegaCLI versions (#2240)
- [atl1] kernel panic fixed in case of DMA errors
- [scheduler] warning added if huge IPI latency is detected (debug for PSBM-11683)
- [timer] kernel panic in timer interrupt in kexec on nodes with buggy BIOS fixed (PSBM-12579)
- [memory reclaimer] possibility to use sync writeback, make nodes much more responsive on high memory pressure (PSBM-11683, #2188)
- [oom] OOM priorities configuration has been reworked completely (PSBM-12263)
- [cgroups] kernel panic in devcgroup_seq_read() has been fixed (PSBM-12254)
- [cpt] task credentials fix
- [sheduler stat] /proc/uptime should now show correct idle time inside CT (PCLIN-30774)
- [fs] fix occasional NULL pointer dereference in show_vfsmnt() -> prepare_mnt_root_mangle() (PCLIN-30843, #2188#c62)
- [fs] fdatasync() optimization and statistics (/sys/fs/ext4/$dev{inflight_requests,optimized_flushes}) (PSBM-12094)
- [cpt] fix locking for netfilter conntrack expectations (PSBM-12691)
- [cpt] do not try to restore CT which can't be undumped (PSBM-12693, possibly #2244)
- [ploop] initial (incomplete) implementation of io_kaio - ploop io module based on in-kernel direct aio (PCLIN-30658)
- [ploop] optimization to more effectively merge requests by converting bio reqs to ploop reqs earlier
- fix sysinfo in containers: don't clean uptime, procs and loads (#2051)
- [ms] [cpufreq] fix memory leak in cpufreq_add_dev
- [ms] [ext4] check for zero length extent to prevent BUG_ON (bz.k.org #42859, PSBM-12241)
- [nfs] lockd: change ve in locks reclaimer thread (PSBM-12799)
- [ext4] Fix race leading to endless wait in
ext4_ioend_wait()(PSBM-12831, PCLIN-30867, #2258) - [ext4] Add fsync optimization statistics (PCLIN-30870)
- [nfs] kernel panic on a CT stop with NFS client inside fixed (PCLIN-30874)
- [cpt] online migration with tmpfs inside from kernels older than 54.1 failures fixed (PCLIN-30881)
- [fs] sync has been made per-container (was global) (PCLIN-30842)
- [vzquota] soft lockups on inode_lock have been fixed (PCLIN-30858, #2255)