Re: [uml-user] UML 2.6.11 SMP issues

[Blaisorblade <bla...@ya...>]

On Wednesday 25 May 2005 20:18, Patrick Allison wrote:
> Blaisorblade wrote:
> > Well, this particular error (the build one) should be fixed in -bs... for
> > the SMP mode we'll work on it, however SMP support is not very maintained
> > because it's supported for now only in TT mode, which is slower... i.e.
> > with SKAS on the host you get better performances than with SMP.

> OK, I'll try that branch. TT mode is all that's possible for me, as I
> don't control the server (and don't have root). Is there any previous
> 2.6 UML patch that actually works with SMP (and would there be a strong
> reason for me not to use it)?
Hmm, I don't really know... actually SMP worked until very recently IIRC... 
try 2.6.9 vanilla and/or 2.6.9-bs7. I don't have time to test myself 
currently.
> Incidentally, the fix that was posted here for slirp also worked for me
> as well. Slirp now works like a charm. So thanks much for that. :)

> I mainly use UML to have a consistent platform 
> for running simulations on department-maintained machines - basically no
> disk accesses while it's running. I haven't noticed a significant
> slowdown over the native version, but maybe I wasn't paying close enough
> attention.
No, simply in this case you won't get almost any slowdown, because as you note 
this program does not use a lot the kernel services... (memory allocation 
could still be a problem, but that can only make a little difference).

Obviously you must point TMPDIR to a tmpfs / ramfs mount point for UML (as 
otherwise it'll need to sync its virtual memory to disk).
-- 
Paolo Giarrusso, aka Blaisorblade
Skype user "PaoloGiarrusso"
Linux registered user n. 292729
http://www.user-mode-linux.org/~blaisorblade