Am 07.08.2007 um 16:09 schrieb Ian Darwin:
On Tue, Aug 07, 2007 at 11:47:38AM -0400, Dr. H. Nikolaus Schaller
wrote:
I am analyzing why the AJZaurusUSB driver creates a new interface on
MacOS X each time the OpenMoko is rebooted.
Now I think I have found that OM does initialize the MAC address of
the RNDIS/Ethernet gadget driver randomly on each reboot (and MacOS X
uses that to find devices it already knows). Can anybody confirm
that?
If it is so, I would suggest to derive the MAC address from the
device serial number - or if that is not available - from the
Bluetooth MAC (but not using the same).
Good analysis. I don't have my cable with me so I can't confirm
this. However, if this is the problem, it is OS X that should be
fixed. I think the FIC people have enough to do. Other OSes including
Are you kidding? Have you ever tried to ask Apple to fix something?
They use the Mac addresses as an index into a configuration table
so that you don't have to configure everything (manually) over and
over again.
And each device has its own, nicely separated set of DHCP/TCP
configuration values.
Unfortunately, devices are differentiated on that level by MAC
address (only)...
Therefore I think it is a general design flaw if a device does
present random
Mac addresses...
Well, the AJZaurusUSB driver can be made to offer a faked MAC address to
the networking stack, but that is just a workaround and not a good
solution.
But before we blame anybody or urge someone to solve it for us we
must find out
if this is really the case.
Linux and also the various BSDs do not have this problem; I use it
on OpenBSD and it is always cdce0.
They may simply not have the auto-config feature of MacOS X that
relies on
well-behaving devices...
Of course, the phone is open source, so if you want it done that
way, you could submit a patch to make it work that way. That's the
idea of Phase 1, after all!
Yes, that is the idea of open source. Everybody can do it - but in
practice
you can't until you have found how to do this correctly and managed
to recompile the kernel. Since I am not a Linux kernel specialist,
it will become Phase 3 or 4...
-- hns
