Thanks so much for your advice!
My host kernel in the 64 bit SUSE Linux Enterprise Server is
18.104.22.168-0.30-smp. I also tried Ubuntu 8.04 64 bit, which uses 2.6.24
My guest kernel is linux-um-22.214.171.124-bb2-xt-4m in both cases, provided
by VNUML project (compiled for 32 bit). In case I didn't express it
correctly I summarise. (I hope I'm not confusing between the concept of
host (physical kernel) with guest (UML kernel) kernels, is it not?)
Case 1) Host: 32 bits (2.6.24). Guest, UML Kernel: 32 bits.
Case 2) Host: 64 bits (2.6.24, 2.6.16). Guest: 32 bits. (the same
In both cases the host kernel is not too old (at least in the case of
Ubuntu). VNUML provides a tutorial to compile your own VNUML kernel, so
I'll try downloading a recent image from kernel.org and compiling it
myself and see what happens.
I will also try to see if there is something strange about libraries or
Anyway, thanks again for your reply.
Gordon Russell wrote:
> Having made the swap from 32 to 64 bit hosts myself this month, I have
> found them to be very reliable and without the problems you seem to be
> You are using a really old kernel. I suggest upgrading both the host
> and guest kernels to something more modern. Previously I had been
> using 2.6.7, and this definitely caused problems in the move to 64
> bits. 2.6.18 for me ran but had the odd little quirk which made it
> problematic. You dont say what version your guest kernel is, but try
> to get both updated.
> The other thing which caught me out was the uml tools, which I
> compiled on another machine. When I upgraded the readline library had
> undergone major revisions, and so the tools needed to be recompiled.
> Do "ldd" on the tools (like uml_mconsole) to make sure they all link
> to current dlls.
> You should really have /dev/shm as a tmpfs mount. Never seen a
> standard install without this as default (but I use fedora).
> I also originally had con2 to 7 in the inittab of the virtual machine.
> Why not give all the con2..7 devices a pty? But thats a longshot.
> Best of luck.
> 2008/10/29 Gerson Rodríguez de los Santos López <gerson@...>:
>> 1) In case 2, the "linux" process becomes "defunct" according to "ps -e"
>> output. As a result, the VM does not boot. The key point seems to be the
>> command line invoked by VNUML presented here (full trace attached in
>> /daemon: /usr/local/share/vnuml/kernels/linux
>> hostfs=/root/.vnuml/simulations/dynavnuml/vms/H1/hostfs mem=32M
>> uml_dir=/root/.vnuml/simulations/dynavnuml/vms/H1/ umid=run con=null
>> con1=pts mconsole=notify:/tmp/vnuml_notify.ctl.X16646
>> 2) I have tried executing the UML kernel myself with that same command line.
>> After printing "Checking syscall emulation patch for ptrace...missing", the
>> proccess finishes with no further error message displayed in the console. I
>> don't even receive an status notification on how the process ends (Stopped,
>> Aborted or Segmentation Fault).
>> 3) I have also tried executing the following line at
>> /usr/local/share/vnuml/kernels: "./linux
>> ubda=../filesystems/root_fs_tutorial mem=32M". The VM starts succesfully and
>> I can log as root as usual. (May it have something to do with the COW
>> feature?) No one in the VNUML forum seems to know anything about it and the
>> forum administrator advised to expose my problem here.
>> I have seen the archives to see if there is a similar problem, but I haven't
>> found someone with the same problem as me I am specially interested in using
>> 64 bit host OS as it is the only way we can fully use the system resources
>> our servers have. Is there anything I am doing wrong?