[gPXE] Issue 55

[Joshua Oreman]

On Thu, Nov 19, 2009 at 4:52 PM, Patrick Cervicek
<patrick at etherboot.spamtrap.hs-esslingen.de> wrote:
> On Thu, 19 Nov 2009, Thomas Miletich wrote:
>
>> Hi
>> Even if the UNDI driver works some times, in your setup it isn't
>> really supposed to work, and I'm surprised that it does.
>
> What should cause a problem in my setup? Booting from a usb stick?
>
> btw: Why does "ifstat" show two different nic?
> PCI00:19.0
> UNDI-PCI00:19.0
>
> Does this mean, using DEBUG changes the behavior of gPXE?

The only thing I can think of is that DEBUG could change the order in
which some files are linked, which changes the order in which buses
are probed, which might put the UNDI driver before the raw PCI one. If
this is the case, it's an unexpected side effect, and probably
something the developers should fix.

>> Please try a gPXE build with DEBUG=e1000. It's definitely helpful if
>> you can use the serial console because there will be a lot of output.
>> Please try 'dhcp net0' with the debugging options enabled and capture
>> all the output on the serial port. Wait until the output stops, or at
>> least slows down(that means dhcp is finished). Then please put your
>> capture at http://etherboot.pastebin.com/ or attach it as a file to a
>> mail
>>
>
> I tried DEBUG=e1000 and get much output but still does not work.
> Unfortunately this notebook has no serial console. I will try to get a USB2SERIAL-Adapter and try to get
> the Debugging output tomorrow, before the owner takes back his notebook
> next week.

A USB-to-serial adapter won't work if the gPXE machine has the USB
end, because it looks like a USB device and gPXE doesn't know how to
talk to USB devices. If your laptop and development machine both have
a FireWire port, I have an experimental FireWire debugging capability
in my personal repository, and will be happy to help you set it up.

-- Josh