[Xen-bugs] Dom0 crash with xen 4

[Emile den Tex <emile@xxxxxxxxxxxx>]

Hello,

We are running Linux 2.6.32.13 (x86_64) as dom0 with a Xen 4.0.0
hypervisor and experienced a dom0 kernel crash. This was in the night,
with no one logged in, no xen commands issued.

This is what to the kernel said before it died (copied from a screen
shot):

[..] xen_restore_fl_direct_end+0x0/0x1
[..] _spin_unlock_irqrestore+0x2b/0x50
[..] down_trylock+0x30/0x50
[..] oops_end+0xe1/0xf0
[..] do_divide_error+0x7e/0xa
[..] find_busiest_group+0x5f6/0xb50
[..] br_forward_finish+0x0/0x60 [bridge]
[..] nf_hook_slow+0xa6/0x100
[..] br_forward_finish+0x0/0x60 [bridge]
[..] divide_error+0x1b/0x20
[..] find_busiest_group+0x5f6/0xb50
[..] run_rebalance_domains+0x190/0x5c0
[..] __do_softirq+0xae/0x140
[..] call_softirq+0x65/0xa0
[..] xen_evtchn_do_upcall+0x3a/0x60
[..] xen_do_hypervisor_callback+0x1e/0x30
<EOI> [..] hypercall_page+0x3aa/0x1010
[..] xen_safe_halt+0xc/0x20
[..] xen_idle+0x3d/0x70
[..] cpu_idle+0xae/0xf0
[..] xen_irq_enable_direct_end+0x0/0x7
[..] cpu_bringup+0x68/0x6d

Unfortunatly this is all we have.

Only two things unusual te report in the setup:
- During dom0 boot we do got 'Unable to read CPU state' messages, but xm
info and the xenpm tool does give me an accurate report on the cpus.
- when doing 'ps aux' the xend daemon process has an incorrect age,
claims to run for over something like 21170349:10 hours (which is
impossible because unix wasn't developed back then) The system clock is
operating normally with tsc clocksource.

The hardware is an IBM X3550, 64GB ram, 2x quad-core Xeon CPU (E5506)

Any help would be greatly appreciated. We would love to bring xen 4 into
production, but are obviously a little intimidated by this crash.
Please mail if you need any more information.

Thanks alot,

Emile


_______________________________________________
Xen-bugs mailing list
Xen-bugs@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-bugs