OpenVZ bug 1284 (http://bugzilla.openvz.org/show_bug.cgi?id=1284) also affects openvz-sources-2.6.27.2.1 and I am attaching a fix for this issue. It has been tested with the above version and it fixes the mentioned problem. Reproducible: Always
Created attachment 202049 [details, diff] printk patch
Thank you, Jernej. Could you reopen upstream bug report so upstream provide us with a fix. As I see attached patch differs from upstream so ... we need another eyes to review it.
Well patches differ because that one is for kernel 2.6.18 and this one is for 2.6.27 and there have been some changes in between. I have checked the code and discovered that in stock 2.6.27 printk_cpu is always set to UINT_MAX just before releasing locks but in OpenVZ, where another branch is added, it is not set there. This patch only adds this one set into that branch and is functionally equivalent to that of the previous version. I have also posted this information to the upstream bugzilla just before opening this bug report.
(In reply to comment #3) > I have also posted this information to the upstream bugzilla just before > opening this bug report. I saw that, but I asked you to _reopen_ upstream bug report :) If bug is closed, developers may miss any changes there. Any way, I just did that. Let's see what upstream tell us or probably I'll get and review code there.
This bug was fixed upstream and fix is included in 2.6.27-kiprensky.1.