John Stoffel wrote: >>>>>> "marty" == marty <marty@xxxxxxxxxxxxxxxx> writes: > > marty> Greg KH wrote: >>>> On Tue, Aug 18, 2009 at 04:27:04PM -0400, marty wrote: >>>>>> I got trouble... >>>>>> (duplicate MAC addresses) >>>> That's a bug in your hardware, have you asked your manufacturer to >>>> resolve this for you? That violates the ethernet spec... > > marty> I have resolved that problem as of today. I found this was > marty> caused by the software I had been using. If a hardware issue > marty> remains, it is moot. > > marty> The bonding driver/utilities normally sets the bond address to > marty> the MAC of the first NIC. But it also set the MAC of the slave > marty> (eth3) to the MAC of the first NIC. This persists through > marty> reboots so that is how my MACs got duplicated. > > marty> Resetting the MAC corrected those problems and everything works > marty> fine now. > > Doesn't this point to a udev rules problem? What should happen if > there are conflicting devices which both satisfy a condition, but > where only one device is allowed to match? > > Now I realize that with MAC addresses you're actually allowed to have > multiple NICs on a host all with the SAME Mac addr, but only if > they're on different segments. Older Sun boxes all used to have a > single MAC address across all ports. This usually isn't a problem > since the ethernet spec says that MAC addresses are local to the > segment, and with switches and bridges, the segment is is limited. > > Fails when you have bonding drivers and other HA tricks which I'm not > up on though. > > John > > > OOPS... Duplicate MACS won't work on a single box. On a network, yes. Duplicate MACS mess everything up, because the lower networking layers do not use IP addresses. They depend on the MAC to route the traffic. I thought this was a udev problem. Greg KH suggested a hardware problem, but I fixed it by removing the bonding driver from my config. Took a lot of debug. I am using shorewall to configure iptables, which has another means to handle multiple ISP's using packet marking. Works and as far as I can see no issues. I was able to make the bonding driver work, but only if I manually corrected the borked MAC beforehand. My changes didn't survive reboot. Something is broken in that driver. I haven't looked as yet but I'm sure someone will discover it. There was a issue with udev, however not a rule; the LFS bootscripts I use were guilty. --retry-failed is in invalid option on udev-1.46. Caused a big delay for some reason. I commented it out and it boots fast now. BTW, this is a handy thing we can do on linux. ip link set eth0 address 01:02:03:04:05:06 That will set a MAC address and survives reboot (on my system anyway). Marty B. -- An artist who is forced to work a specific schedule, is no longer an artist; he is just hired help. Inspiration cannot be purchased.
Attachment:
signature.asc
Description: OpenPGP digital signature
