[Bridge] device eth0 is already a member of a bridge; can't enslave it to bridge Net6

[Bridge] device eth0 is already a member of a bridge; can't enslave it to bridge Net6

[Tim Keupen]

[-- Attachment #1: Type: text/plain, Size: 1323 bytes --]

Hello,
I am working on Network-Simulation (VNUML). Our simulator uses linux bridging to connect the UMLs.
So there is one problem:
The example is the following: There are two hosts simulating one big net.
The two hosts have connection over the external nets Net3 and Net6 (see http://www.uni-koblenz.de/~timbub/verteilteSim3.GIF), but in fact there is only one physical connection between the host: connected through a switch an each host has one interface card eth0. (so normaly you should bind the ip adresses related Net3 and to Net6 to the eth0 and it already is part of the two nets - even if there is one phsical connection)

So my virtual nets are implemented by linux bridges. But if I now connect the physical interface eth0 to the bridge Net3 and then to Net6 the following error occurs: "device eth0 is already a member of a bridge; can't enslave it to bridge Net6" 
Is this a bug or a feature? Why is it not allowed to bind an Interface to two different bridges? 
Is there a workaround for this issue?


Problem: 
UML1 ---- bridge1  \
                             eth0  
UML2 ---- bridge2  /


Workaround?:
UML1 ---- bridge1  \
                             bridge3 ---- eth0
UML2 ---- bridge2  /


UML1   \
             bridge1 ---- eth0
UML2   /



Thanks for your help
Tim Keupen

[-- Attachment #2: Type: text/html, Size: 3498 bytes --]

Re: [Bridge] device eth0 is already a member of a bridge; can't enslave it to bridge Net6

[Stephen Hemminger]

On Mon, 31 Oct 2005 11:04:32 +0100
"Tim Keupen" <timbub@uni-koblenz.de> wrote:

> Hello,
> I am working on Network-Simulation (VNUML). Our simulator uses linux bridging to connect the UMLs.
> So there is one problem:
> The example is the following: There are two hosts simulating one big net.
> The two hosts have connection over the external nets Net3 and Net6 (see http://www.uni-koblenz.de/~timbub/verteilteSim3.GIF), but in fact there is only one physical connection between the host: connected through a switch an each host has one interface card eth0. (so normaly you should bind the ip adresses related Net3 and to Net6 to the eth0 and it already is part of the two nets - even if there is one phsical connection)

Have you looked at umlsim

> So my virtual nets are implemented by linux bridges. But if I now connect the physical interface eth0 to the bridge Net3 and then to Net6 the following error occurs: "device eth0 is already a member of a bridge; can't enslave it to bridge Net6" 
> Is this a bug or a feature? Why is it not allowed to bind an Interface to two different bridges? 
> Is there a workaround for this issue?
> 
> 
> Problem: 
> UML1 ---- bridge1  \
>                              eth0  
> UML2 ---- bridge2  /
> 
> 
> Workaround?:
> UML1 ---- bridge1  \
>                              bridge3 ---- eth0
> UML2 ---- bridge2  /
> 
> 
> UML1   \
>              bridge1 ---- eth0
> UML2   /
> 

There was no good reason to allow it, I guess. Also, I would worry about races on deleting
a bridge and the RCU issues.


-- 
Stephen Hemminger <shemminger@osdl.org>
OSDL http://developer.osdl.org/~shemminger

Re: [Bridge] device eth0 is already a member of a bridge; can'tenslave it to bridge Net6

[Tim Keupen]

> Have you looked at umlsim
no, problem has to be solved with vnuml

   >> Is this a bug or a feature? Why is it not allowed to bind an Interface 
to two different bridges?
   >> Is there a workaround for this issue?
> There was no good reason to allow it, I guess. Also, I would worry about 
> races on deleting
> a bridge and the RCU issues.
But what should be a reasen not to allow it?
What du you mean with races and RCU issues?


thanks
Tim

Re: [Bridge] device eth0 is already a member of a bridge; can'tenslave it to bridge Net6

[Stephen Hemminger]

On Mon, 31 Oct 2005 19:43:01 +0100
"Tim Keupen" <timbub@uni-koblenz.de> wrote:

> > Have you looked at umlsim
> no, problem has to be solved with vnuml
> 
>    >> Is this a bug or a feature? Why is it not allowed to bind an Interface 
> to two different bridges?
>    >> Is there a workaround for this issue?
> > There was no good reason to allow it, I guess. Also, I would worry about 
> > races on deleting
> > a bridge and the RCU issues.
> But what should be a reasen not to allow it?

Because it the input path hook how bridge's get data would
cause increased stack depth (if it even worked).

> What du you mean with races and RCU issues?
Look at how interfaces are removed from bridges, and the
code that notifies a bridge about interface changes. It would
probably break as well.


-- 
Stephen Hemminger <shemminger@osdl.org>
OSDL http://developer.osdl.org/~shemminger