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

[Vincent Touquet]

Hello fellow UML users :)

I had the privilege of meeting some of you already.

Thanks for your support, write-ups and all the stuff
you do that makes UML nicer/easier to use.

I have a rackserver in colocation, on which I am
running instances of uml.

One of them is doing fine and hasn't needed a reboot
for some time now. It doesn't have any significant
load either, it's just running tinydns as a nameserver.

The other host is running Apache and mysql and
is a usermachine where sites are hosted.
It needs a reboot every two hours :(
It just locks up, its not able to get to
it through the network, nor through the
uml_mconsole...

I need to remove the tap device and respawn
the uml to get it back in to place.
If I 'forget' to remove and reinstantiate
the tap device, networking simply fails to
work.

I'm running 2.4.20 vanilla on the host,
compiled with gcc 3.2 and -march athlon-mp.
Its a dual athlon 1600, with 1Gb Ram and
highmem enabled.

The uml kernels are 2.4.19-17um, the nameserver
with 64M of RAM and the server that likes
to lockup with 384MB of RAM because 512M
RAM makes the uml segfault. Both the host
and the user mode linux instances are running
Debian testing and they use devfs.

If you need any more information,
I will be happy to provide it.

Sorry if I run a version of uml that
is out of date or I missed some obvious
patch. I would like to hear it if this
is the case.

best regards
Vincent Touquet