SCIENTIFIC-LINUX-USERS Archives

January 2006

SCIENTIFIC-LINUX-USERS@LISTSERV.FNAL.GOV

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
Stephan Wiesand <[log in to unmask]>
Reply To:
Stephan Wiesand <[log in to unmask]>
Date:
Thu, 12 Jan 2006 19:30:23 +0100
Content-Type:
TEXT/PLAIN
Parts/Attachments:
TEXT/PLAIN (94 lines)
Hi Connie, Andreas, All,

On Thu, 12 Jan 2006, Connie Sieh wrote:

> On Thu, 12 Jan 2006, Andreas Heiss wrote:
>
>> Hi Folks,
>> I have a problem with SL4:
>> We got some new Opteron-Nodes and want to install them with SL4 with=
>> =20
>> Quattor.
>> Unfortunately, the new boxes have a bcm network chip which needs the
>> bcm5700 kernel module to work. This module is not available in the=
>> =20
>> installation initrd.img but we have the source-code for it.
>> How can I put this module into the initrd.img ?
>> Which version is the installation kernel?
>
> SL 4.0 kernel-2.6.9-5.0.5.EL.x86_64.rpm
> SL 4.1 kernel-2.6.9-11.EL.x86_64.rpm
> SL 4.2 kernel-2.6.9-22.0.1.EL.x86_64.rpm

This seems to confirm my guess that SL4 uses the plain non-SMP kernel
for the installation system, and building additional modules to be used
during installation against the content of the kernel-devel rpm with the 
right version should genereally be ok. Is this correct?

>> Where can I get the sources and config file for the installation kern=
>> el?
>
> ftp://ftp.scientificlinux.org/linux/scientific/4x/SRPMS/vendor/errata/
>
>> How can I build an installation kernel and initrd which finds the=
>> =20
>> correct root partition of the installation system?
>
> I think you are better off with making a "driver disk".  A tool kit to
> help with this process is available from
>
> http://people.redhat.com/mwaite/ddiskit-0.9.1.tgz

Great. I repeatedly googled for a definition of what exactly it takes to 
make a driver disk, but to no avail. Thanks.

Alas, I think a dd won't help Andreas much since it's the network driver 
he has to add, and I figure he'd like to avoid having to insert a floppy 
or CD into all his new nodes before a kickstart installation over the 
network is possible. (Or will anaconda retrieve the driver disk over the 
network with the settings inherited from the PXE stack?)

So here are my notes from replacing the e1000 driver on the SL4.1rc1 
initrd with the latest version:

--8<--
initrd.40r.e1000 - modified initrd for pxeboot for SL4.1rc1/i386

Modifying the initrd:
  gzip -cd initrd.img > initrd.uz
  mount -oloop initrd.uz mp
  vi mp/modules/pci.ids
  vi mp/modules/modules.pcimap
  gzip -cd mp/modules/modules.cgz | cpio -id
  cp e1000.ko.2.6.9-11.EL 2.6.9-11.EL/i686/e1000.ko
  rm -f mp/modules/modules.cgz
  find ./2.6.9-11.EL -type f | cpio -oc | gzip -9 > mp/modules/modules.cgz
  umount mp
  gzip -9 <initrd.uz >initrd.40r.e1000

The changes to pci.ids and modules.pcimap are completely straightforward
-->8--

Since adding a new driver is a slightly different case, it probably takes 
modifying a few more text files on the initrd. If I had this problem, and 
were desperate, I'd possibly consider overwriting the tg3 driver 
instead...

Hope this helps,
 	Stephan

PS The release notes for the beate of the upstream vendor's Update 3 talk
    about support for additional Broadcom chips, hence the whole problem
    may vanish within a couple of weeks...

-- 

  ----------------------------------------------------
| Stephan Wiesand  |                                |
|                  |                                |
| DESY     - DV -  | phone  +49 33762 7 7370        |
| Platanenallee 6  | fax    +49 33762 7 7216        |
| 15738 Zeuthen    |                                |
| Germany          |                                |
  ----------------------------------------------------

ATOM RSS1 RSS2