On Thu, 30 Aug 2007, Brian J. Murrell wrote:

> On Thu, 2007-08-30 at 16:16 -0400, Bret Hammond wrote:
> > 
> > A kernel-source rpm is fine, the problem comes into play when a kernel
> > patch is required.
> 
> Meaning you need to apply a patch to the CFS supplied kernel and
> rebuild?

Yes that, or go through the vanilla kernel build process, selecting the
proper patch set via quilt, etc.
> 
> > The instructions contained in the lustre operations
> > manual (for 1.4.x at least), for building custom lustre/kernel rpms,
> > leave you short a lustre-source rpm in the end.
> 
> Perhaps.  I have not reviewed those instructions.
> 
> > They also leave you 
> > short an initrd, if I recall correctly, which leads me to the conclusion 
> > that some other build magic is happening behind the scenes at CFS.
> 
> The CFS build does not supply an initrd.  That is job of the vendor's
> kernel installation process.  You should get an initrd when you install
> the CFS kernel RPM.

Exactly, which is why the kernel srpm would come in handy ;-)
> 
> > It would just make life a tad easier from this side to have access to
> > the srpm or the spec file used to generate the kernel rpms.
> 
> I see.  Yes, I don't know what provision we make available for getting
> from our kernel-source rpm to another binary kernel rpm.  Maybe somebody
> more knowledgeable in our current build processes might have an answer.
> Barring that, you could of course file a bug or request an update on
> status of a bug (i.e. in the bug itself).

I'll file a bug, thanks for your help.

> > We need to build Myrinet drivers from time to time, which works using
> > the CFS supplied kernel-source rpm.
> 
> Well that is good to know at least.  :-)
> 
> b.

Thanks,
Bret

Attachment: pgpMkJpQ1J8wK.pgp
Description: PGP signature

_______________________________________________
Lustre-discuss mailing list
[email protected]
https://mail.clusterfs.com/mailman/listinfo/lustre-discuss

Reply via email to