[Gllug] Kernel source on Red Hat

Dave Jones davej at suse.de
Thu Jun 6 14:03:17 UTC 2002


On Thu, Jun 06, 2002 at 01:58:32PM +0100, Vincent AE Scott wrote:

 > it does not perform aswell as a native system.

Ah, thanks for setting me straight on that. Always better to defer
to someone with actual experience 8) 

 > but it's performance is heavily dependant on having host and UML using /tmpfs.

Hmm, another way could be NFS exporting the host and mounting it inside
the UML instance perhaps.

 > and nope, theres currently no direct hardware access.

Something that could be /very/ interesting, would be valgrind+UML.
a. It could find leaks and the like in kernel code.
b. Valgrinds virtualisation of the CPU (whilst slowing things down
   slightly) could be useful for trapping hardware poking, and
   redirecting it to the host kernel.
   Eg, a cpuid instruction gets transformed by uml into an open() &
   read() of /dev/cpu/0/cpuid 

Jeff has been talking with the valgrind author, so this might come
to light one day..

 > as dj mentioned, processor affinity is the answer.  the patches for
 > which are bundled with severaljumbo kernel patch sets, -jp and -wolk
 > spring to mind.

I think they're in Andrea Arcangeli's -aa patches too.
 
 > IIRC, UML is being integrated into the 2.5 tree, dave?

Jeff last pushed it for inclusion circa 2.5.5 iirc. I'm not sure
how up to date it is wrt the current tree, but it'd be nice to see
it merged at some point.

        Dave

-- 
| Dave Jones.        http://www.codemonkey.org.uk
| SuSE Labs

-- 
Gllug mailing list  -  Gllug at linux.co.uk
http://list.ftech.net/mailman/listinfo/gllug




More information about the GLLUG mailing list