From OpenVZ Virtuozzo Containers Wiki
Jump to: navigation, search


Since 042stab061.2:

  • [processes] when CT's init is dying, check if a process is a zombie with the parent outside of this CT and is not collected, kill it. This unblocks CT stop in case there was a vzctl enter + Ctrl-z. (PCLIN-31083)
  • [mm] balance between file/anon pressure has been fixed. This prevents the situation when file lru gets much more pressure than anon memory and reclaimer never touches anon pages. (PSBM-14619)
  • [sysrq] Alt+SysRQ+m has been enhanced to show reclaimer statistics.
  • [NFS/lockd/debug] generate a kernel BUG() in case we've got 2 or more identical lock IDs (pids). This is a debug patch for PCLIN-31177.
  • [shmem] a function to install a page into a specific location in a shmem area has been added. This is required for VMs suspend/restore into pram. (PSBM-14664)
  • [ploop] a race in ploop locking has been fixed. (PSBM-14751)
  • [gcc] kernel should now be compilable by gcc 4.7 (#2339)
  • [mm] add a separate bio pool that guarantees swapout progress during reclaiming memory. This is an optimization (not a fix) related to PSBM-14619
  • [block] optimized ->writepages() method for block devices has been implemented.
  • [cpt] missed "\n", other cleanups (PCLIN-31096)
  • [e1000e] VLAN set up over veth inside a CT (veths are bridged on HN) configuration works now (PCLIN-31132)
Warning: if a VLAN configured on the HN, other VLANs inside CTs may still not work, if e1000e driver is used, either all VLANs used inside CTs must be created on HN, or no VLANs on HN at all.
  • [net] netdevice leaks have been fixed (PSBM-14858)
(messages like "kernel:[17597.898188] unregister_netdevice: waiting for lo=ffff880310e52820 to become free. Usage count = 3")
  • several kernel panics on modules load/unload have been fixed in configfs, configfs example, vga16fb and AB3100 regulator
  • [net] kernel panic triggered by NFS server timeouts has been fixed (PCLIN-31212)
  • [venet] kernel panic when using OpenVPN (tun dev) + iptables inside a CT has been fixed (#2322)
  • [net] CPU soft lockup caused by calling inet_twsk_deschedule() with bh enabled has been fixed (#2346)