[uml-user] Re: uml lockups using 2.4.19-17um on a 2.4.20 SMP host

[High M. <hi...@pr...>]

Jeff Dike said: 
> vincent@ul... said: 
> > It just locks up, its not able to get to it through the network, nor
> > through the uml_mconsole... 
> 
> Interesting console output? 
> 
> Stack trace from the hang? 
> 
> > Sorry if I run a version of uml that is out of date or I missed some
> > obvious patch. 
> 
> I want to hear about new bad things for anything that's not grossly
>out 
> of date.  There will be bugs that show up only after a long uptime,
>and 
> those systems will be automatically out of date by virtue of that
>uptime. 
> 
> Jeff 
> 

  We are running a Mandrake Linux 8.2 SMP host with kernel
2.4.19-16mdksmp. We have about 6 UML guests running on UML kernel
2.4.19-46um, each of which has its own tap device. Generally speaking
everything works fine... except for one particular guest, called
Lithium. Lithium will from time to time stop responding to network
requests (ssh, ping, anything). If I shutdown Lithium using the command
"uml_mconsole lithium cad", and then start it up again, it appears to
start normally, but still does not respond to network requests.

  To bring back Lithium's networking, we have to shutdown Lithium, bring
the tap interface down (via command "ifdown tap1"), bring the interface
back up ("ifup tap1"), then startup Lithium.

  What gives here? None of our other UML hosts has this problem. The
only thing unique about the host Lithium is that it does generate a lot
of network traffic as it's an SNMP monitoring host running Cacti. Any
suggestions or revelations from the list? Thanks for your help everyone.
Jeff, thanks for some really cool software!

-- 
High Mobley
Network Systems Administrator
Priority Networks