https://bugzilla.novell.com/show_bug.cgi?id=736268 https://bugzilla.novell.com/show_bug.cgi?id=736268#c2 Mike Galbraith <mgalbraith@suse.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |NEEDINFO InfoProvider| |adrian@suse.com --- Comment #2 from Mike Galbraith <mgalbraith@suse.com> 2012-04-12 08:19:23 UTC --- (In reply to comment #0)
We observe a significant higher number of build failures in OBS using the 12.1 kernel due to hanging guest kernels during system shutdown invoked by "echo o
/proc/sysrq-trigger".
Thanks to Michael and Alex this got traced down to nmi watchdog, which keeps one CPU looping forever in migrate_task() function.
What I see in the crash dump is RCU stalls. I wonder if ipv6 is being used at all, because f2c31e32, which is in 3.1 introduced an rcu_read_lock() imbalance that causes RCU stalls all over the place. The fix for that, 94f826b8 is not in 3.1 kernels. That fix is currently in stable-review for 3.0, 3.3 and 3.3, but 3.1 is not receiving any more stable updates, so I'll add it myself. It'd be good to try the kernel I just pushed. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.