[Server-devel] Getting closer: XS-0.5-dev5 "preview"

[Jerry Vonau]

Martin Langhoff wrote:
> Fantastic testing - thanks! More notes below...
> 
> On Wed, Oct 15, 2008 at 12:35 PM, Jerry Vonau <jvonau at shaw.ca> wrote:
>> The automated kickstart install pauses at the method screen for input. This
>> could be avoided by adding method=cdrom:/dev/sr0 to the boot prompt  for the
>> kickstart install. Should the cdrom device differ from sr0, that would be a
>> quick "hit tab and edit" before you boot the install. The other alternative
>> is to add the "cdrom" to the kickstart file.
> 
> Looking into that...
> 
>> All the ifcfg-mshbond[012] files use the same CHANNEL=1, that should be
>> 1,6,11 no?
> 
> Doh! Fixed.
> 
>> If you use AAs, don't plug them in until after the first boot, or udev might
>> be a bit confused and may not write the correct udev rules on a fresh
>> install.
> 
> Interesting. I wonder how we can deal with AAs during first-boot ( /
> install? ) and udev picking them up too realy. Hmmm. By the time we're
> in firstboot, our custom udev rules should have been installed
> already, so there is no room for udev to mess up I think.
>
The udev rules aren't written until udev finds the AA plugged into the 
system. It's renaming what was an eth/msh combo to wlan/msh. Think this 
is more of an issue with the firmware not finding the second AA, than 
anything else. I'll dig around more tomorrow.

Upon "firstboot" the server is still somewhat un-configured, you have to 
run domain_config, so waiting to plug the AA until after domain_config 
is run doesn't seem to be a showstopper to me. However "domain_config" 
is leaving its xs_domain_name file in olpc-scripts instead of sysconfig 
causing "make dhcpd.conf" and thus dhcpd to fail.


> Perhaps udev from the _anaconda stage2_ is writing its (uninformed) rules?
> 
Just for eth0, and that should be fine.

>> What is failing, for me anyway, is if 2 AA are plugged in during a reboot,
>> the initscripts fail to load the firmware for the second one:
> 
> Good catch - but that's weird. I just tried to repro, but I have AAs
> from different generations and I locked up the machine badly.
> Mismatched AAs are no go.
> 
> I'll try again with matching AAs tomorrow, but this is low-priority
> for this release. One AA setups are something we want to support (XS
> on the XO is a "one AA setup") -- but multi-AA setups are currently
> not in vogue...
One works perfect, two becomes an issue.
More later,

Jerry