Re: build breakage due to br_multicast.c referencing ipv6_dev_get_saddr()

[Brian Haley <brian.haley@hp.com>]

On 03/17/2011 03:57 AM, Jan Beulich wrote:
>>>> On 16.03.11 at 18:34, Brian Haley <brian.haley@hp.com> wrote:
>> On 03/16/2011 08:34 AM, Jan Beulich wrote:
>>> With BRIDGE=y and IPV6=m commit
>>> fe29ec41aaa51902aebd63658dfb04fe6fea8be5 ("bridge: Use IPv6
>>> link-local address for multicast listener queries") causes the build to
>>> break.
>>>
>>> Similary, even if both are =m, but ipv6.ko got blacklisted (as is
>>> happening in various SuSE distros when disabling IPv6), there's
>>> a runtime problem since bridge.ko then won't load anymore due
>>> to the missing symbol.
>>
>> Load the ipv6 module with disable=1, which is why I added it :)
> 
> Indeed, I realized there is such an option only after I sent
> that mail. Nevertheless, I think it is overkill to load a huge
> module like this just to satisfy never actually used symbol
> references.

I could also argue that the bridge code could change into IPv4-only
and IPv6-only pieces to avoid this too (without actually looking
at the code of course).  But in this case you actually built the kernel
with IPV6=m right?

> In fact, just like it seems bogus to load ipv6.ko in a pure IPv4
> environment, I think the opposite is also true: IPv4 support
> should be in a module, and it should be possible to not load
> it in a pure IPv6 environment.

That's a much bigger nut to crack...

-Brian