[plug] re: kernel -smp conflict

Jon Miller jlmiller at mmtnetworks.com.au
Wed Jul 23 16:24:15 WST 2003


I would have thought that the newer module/driver would overwrite the older, but for some reason that is not the case.  So either I copy it out of the kernel-smp area (which is probable the best thing and do a rpm -Uvh --force {package} to force the upgrade or delete it and rebuild the database with rpm --rebuilddb.  However, I'm not sure if the database does a check on the installed packages to see what's still on the system.  I guess I could alway do a rpm -e {package} to remove it also ... probable best bet.
Any thoughts on this?

Thanks

Jon L. Miller, MCNE, CNS
Director/Sr Systems Consultant
MMT Networks Pty Ltd
http://www.mmtnetworks.com.au

"I don't know the key to success, but the key to failure
 is trying to please everybody." -Bill Cosby



>>> dsbrown at cyllene.uwa.edu.au 10:32:53 PM 22/07/2003 >>>
Hello Jon.

On 22 Jul 2003, Jon Miller wrote:

> I probably could if I wanted to spend the time.  I guess what I'm asking
> is it should be possible to recompile both kernels with the correct
> module for the broadcom driver needed.  Is this correct?

Sorry, brain fade... what you're saying is that you have pre-compiled
kernels for smp and non-smp?   And you just want to create (rpm's??) the
modules to suit of the BCM5700?   In that case, don't know - only done one
RH install, nothing else yet with RH.

<Off thread>
I'm a bit confused here... if rpm's are like deb's, they manage
dependencies, right?   For "normal" applications and applets I can
understand the benefits of creating rpm's (or deb's) so they can be
applied to differing configurations.   But since a kernel is so intimately
tied with the hardware on which it will run - unless there are gross
inefficiencies trying to match all possible hardware - what is the benefit
of having a kernel done up in an rpm?

I ask this to advance my knowledge here, sorry Jon!
</Off thread>

> Running the server with dual cpu you would want to run the kernel-smp,
> since the driver is available for this, should I run the rpm with either
> --force or whatever parameter to get the installation to install since I
> doubt very seriously if I'll run the server in single cpu mode.

I suppose you may want to have an emergency stand-by kernel of single CPU
flavour in case one of the Xeon's decides on a vacation?  Even without RH
experience, trying to "--force" anything could result in instability,
methinks.

Even having done only a few kernel + module compiles, I
would definitely be tempted to spend the time compiling a custom kernel.
But if this *is* the x335, I suppose the other argument in favour
of an off-the-shelf kernel is in case Mr Big Blue wants to get nasty over
warranties; "we have certified this hardware on X, Y and Z configurations
of <insert your software here>"

HTH,
Denis

 > > On Tue, 2003-07-22 at 19:49, Andrew
Barbara wrote: > > Hu Jon,
> > the smp kernel will run the system in multiprocessor mode and the non-smp 
> > kernel will run the system in a single CPU mode.
> > As for your questions about kernels and roms.... is there a reason why you 







More information about the plug mailing list