Hello, On Thu, 3 Jan 2013 14:57:56 -0600 Michael David <[email protected]> wrote:
> Happy new year to all. :-) Had a chance to revisit this and have an > update. Hi, happy new year to you too; sorry for not responding to the bug report sooner, didn't have a chance to dig into the issue. > I consider this a bug as the behavior is changed and it silently > breaks existing configs, however the workaround is simple enough. > Ideally the behavior from squeeze could be brought forward, but > perhaps another option is to warn if a manual iface has no up / down > statements? Such warnings could also be surpressed by a flag > in /etc/defaults/networking. I think it is actually a bug; not sure where exactly. Try to debug it by changing the shebang of /lib/bridge-utils/ifupdown.sh to #!/bin/sh -x and see what happens. It was supposed to bring the interface up, as I understand. > In squeeze, br0.XXX devices are treated as their own entities. It > seems like in wheezy they are treated as aliases of the root device. Well, yes, every interface with a period “.” in its name considered a vlan subinterface. Probably, I can teach ifupdown to know about bridge-utils existance, so it won't configure the link if it sees non-empty bridge_ports parameter. How do you think, should it be enough? If I do that, this part: > if test -d /sys/class/net/br0 -a ! -d /sys/class/net/br0.4094 ; > then ip link set up dev br0; > ip link add link br0 name br0.4094 type vlan id 4094; > fi will be executed if and only if bridge_ports parameter isn't specified. Actually, I could solve the first issue on my side as well, but I don't think it's a nice idea. > However, I again feel that the squeeze behavior was appropriate. > We've built our configs such that generally speaking eth$x > corresponds to br$x. So, if you want VLAN 100 on the primary NIC, > br0.100, secondary NIC, br1.100. With this workaround, you'd have to > do tricks like VLAN 100 is br10 for first NIC and br100 for second > NIC. Makes sense, yes. > I did notice there are some issues with the down process. For example, > when br1 goes down, it's also removing VLAN eth0.4094 though > technically eth0.4094 should be responsible for its own VLAN config. Don't really understand what exactly it is, let's try to fix (1) and (2) and then look into this thing — I assume you can try to get some debug logs. > run-parts: executing /etc/network/if-pre-up.d/bridge > Set name-type for VLAN subsystem. Should be visible > in /proc/net/vlan/config ERROR: trying to add VLAN #4094 to IF > -:eth0:- error: File exists run-parts: I guess there may be a bug in bridge-utils. -- WBR, Andrew
signature.asc
Description: PGP signature
