[radvd-devel-l] radvd wrongly sends all subnets from all interfaces to the first interface

[Reuben Hawkins]

On Wed, Jun 27, 2012 at 7:27 AM, Roman Mamedov <rm at romanrm.ru> wrote:
> On Wed, 27 Jun 2012 07:13:57 -0700
> Reuben Hawkins <reubenhwk at gmail.com> wrote:
>
>> Is this the bug where a prefix for interface Y is send out on interface X?
>>
>> I'm not sure what's happening here.  I added a debug message which
>> verifies the interface before sending a packet and it appears to be
>> working correctly, but isn't.  I haven't been able to figure out if this
>> is some kind of bug with sendmsg or if radvd is just using it incorrectly.
>
> As of now I have determined that in this particular case all the
> advertisements which are supposed to go over VLANs (eth0.2, eth0.3, eth0.9
> etc), ALSO seem to be sent over eth0 (i.e. without the VLAN tag).
>
> I don't believe this is the same bug that has been posted some time ago, where
> adverts intended for one interface were seen on a completely different and
> unrelated one (that one was related to addition and removal of devices).
>
> But regarding the VLAN issue, I have experienced a similar problem with ISC
> DHCPD: it sees all DHCP requests sent by hosts on VLAN 2 (and to be received
> on eth0.2) as though they come untagged to eth0. It turned out that it is
> related to the use of SOCK_RAW, which doesn't seem to play well with VLANs:
> http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=643564
> There's some kind of a patch available there.
>
> In the radvd source code I noticed that it also uses SOCK_RAW. Maybe this is
> where the problem lies? To reproduce, I suggest that you set up a test box
> configuring VLANs on a radvd server (e.g. setting up a VLAN interface eth0.2
> and configuring both eth0 and eth0.2 separately in radvd.conf), and see if your
> clients on the LAN which have no VLAN set up also receive RAs that were
> intended for eth0.2.
>
> --
> With respect,
> Roman
>
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~
> "Stallman had a printer,
> with code he could not see.
> So he began to tinker,
> and set the software free."

Interesting...It'll take me a little while, but I think that'll be
enough info for me to come up with a fix.

Thanks!