So with those exact steps I managed to crash a real dual-core machine.
So we can rule out any virtualization. This merely is related to
whatever that command does and likely having more than one cpu. (It also
worked / crashed on a local xen guest, so I can reproduce things
locally). Next step will
Bugger, I made the severity of the messages too low... :/ If at all they
might be in /var/log/syslog on instance reboot... I better make a better
version. But its interesting that this appears on a completely different
virtualization platform. I wonder whether it could even appear on real
hardware.
And another one from the debug kernel on EC2, with a slightly different
call stack:
[ 4389.480352] [ cut here ]
[ 4389.480884] kernel BUG at
/home/smb/precise-amd64/ubuntu-2.6/kernel/sched_fair.c:1239!
[ 4389.480894] invalid opcode: [#1] SMP
[ 4389.480902] CPU 0
[ 4
We have had a crash with Stefan's debugging kernel running. Here is the
output (doesn't look like it contains any more information).
[248587.286290] [ cut here ]
[248587.286765] kernel BUG at
/home/smb/precise-amd64/ubuntu-2.6/kernel/sched_fair.c:1239!
[248587.286775] inv
And we've just reproduced on EC2 with the debug kernel:
[248587.286290] [ cut here ]
[248587.286765] kernel BUG at
/home/smb/precise-amd64/ubuntu-2.6/kernel/sched_fair.c:1239!
[248587.286775] invalid opcode: [#1] SMP
[248587.286783] CPU 0
[248587.286786] Modules lin
We've got small EC2 instances (single processor) that haven't exhibited
this behaviour, but we get it with large EC2 instances (2 CPUs); the
VirtualBox machine I just reproduced it with was specifically set to 2
CPUS.
It seems to me that this bug might only occur on multi-cpu boxes?
** Summary ch