Re: [uml-devel] Some Problems

[David C. <da...@da...>]

po...@ja... wrote:
> * the new 2.4.19-45um seems to produce some compilation-errors
>   in context with "skas" (for example in ptrace.c , FRAME_SIZE
>   isn't always well defined, as it seems)

Just disable skas during the compilation - Since you're using the SuSE 
kernel, unless you rebuild it with the ptrace patch, you can't use skas.

> * if "skas"-versions get compiled, kernel panics about
>   New_mm_context (wich doesn't seem to have to do with /proc/mm).

See above.

> * jail-mode really is unexpected slow on my system
>   (a factor of ~10) with 2.4.19-5

Yep. If anything, it's probably more slow than that. Hence why we use skas.

> * [the already fixed] sighandling-prob makes 2.4.10 
>   depressing to use

2.4.10? That's the SuSE kernel - Upgrade it.

> * I'm having serious mem-managment-problems.
>   For example, 2.4.19-5um, nearly exclusivly running on
>   2.4.10 host with 256 MB RAM (+ same swap), UML
>   without limits + 400 MB swapfile produced a  SEGFAULT
>   when using YAST1.

What 'mem=' parameter did you pass to the UML?

>   I planned to use ~5 umls on that host (Cel. 1100 Mhz), but a 
>   single one seems to produce more overhead I expected
>   5 to do.

Could you elaborate on the overhead you are seeing?

>   I guess it would help, if I would turn off disk-buffering
>   on host, but the strange it sounds, I don't find the
>   files to do this in my procfs on host (although kerneldocu
>   tell where to find, they simply don't exist).

Sounds like SuSE poke with their kernel then.

> * I didn't manage to install the package "makedev"
>   ("permission denied"). Am I not allowed to create devices
>   on uml?

Sure. Maybe you should investigate why 'makedev' failed. Can you write 
to other locations on the root partition?

> I'd expect this switch to refer to physical mem of the host
> (wich may be the physical machine, or host's host's phys. mem 
> otherwise).
> Can sombody in short explain, wich swap will be used
> (inside[wich I expected] or outside one?)

Host swap is used to swap out physical RAM. UML swap is used just as the 
host.

> |Something to note if you have a small /tmp is that UML creates a 
> |file in /tmp which is the same size as the memory you specified. 
> 
> And why the hell this? I mean, I didn't proove that, but
> at least this sounds strange to me.

The file in /tmp is your 'physical' RAM for UML.

David

-- 
David Coulson                                    email: d...@vi...
Linux Developer /                          web: http://davidcoulson.net/
Network Engineer                                   phone: (216) 533-6967