Re: Strange behaviour with boottime loading of ipw2200

["darrell pfeifer" <darrellpf gmail com>]

On 11/3/06, Bill Nottingham <notting redhat com> wrote:
> Adrian Reber (adrian lisas de) said:
> > I am seeing the same thing on a server using bonding with two ethernet
> > interfaces. I have two different cards (e1000, acenic) and on every boot
> > one of the card fails. The bond0 interface still works with only one
> > card by I also have to do an rmmod, modprobe to get both cards working.
> > So I have the same error in a different setup.
>
> Not to sound like a broken record - but do you have ifcfg files with
> HWADDR in them?
>
> Background:
>
> PCI modules are loaded (essentially) in parallel. If one driver takes
> longer to initialize on a particular boot, the device ordering will change.
>
> We have code run by udev (/lib/udev/rename_device) to handle this, but
> it requires that ifcfg files have the HWADDR in them for the appropriate
> devices so it knows what should remain what.
>
> Bill
>
> --
> fedora-devel-list mailing list
> fedora-devel-list redhat com
> https://www.redhat.com/mailman/listinfo/fedora-devel-list

In my case the BCM4400 device (eth0) has HWADDR but the IPW2200 device
(eth1) doesn't. I'm assuming that the

/etc/sysconfig/networking/profiles/default

path is the correct one, not either of

/etc/sysconfig/networking/devices/
/etc/sysconfig/network-scripts/

> From the failure I see in /var/log/messages it looks like the Broadcom
driver picked up eth1 first (ignoring the HWADDR?), then IPW2200 was
left to find a random name.

I'll try putting HWADDR in the IPW eth1 ifcfg file(s).

darrell