Re: [lwip-users] driver clarifications for ipv6: loopback, multicast

[Ivan Delamer]

Just got back and saw this.

you should filter out loopback packets at the driver level. It's not 
possible to do it at IP / layer-3. It may be possible at layer-2 but I'm 
not sure it's the best way to do it.

You can enable reception of all multicast packets if you want. Upper 
layers will filter out packets to groups we are not subscribed to. The 
hardware filter just offloads the CPU. But it will work either way.

Cheers
Ivan


> Date: Thu, 30 Jun 2016 14:00:27 +0200
> From: chrysn <address@hidden>
> To: address@hidden
> Subject: [lwip-users] driver clarifications for ipv6: loopback,
>         multicast
> Message-ID: <address@hidden>
> Content-Type: text/plain; charset="us-ascii"
>
> hi,
>
> as of fixing my enc28j60 driver[1] to work well with ipv6 as well, 
> i've
> stumbled over some corners where there might be a better solution i
> haven't found.
>
> * when operating in half duplex mode, ipv6 address autoconfiguration
>   failed because i had missed to configure the network card to 
> suppress
>   loopback of sent packages; as a result, the returning nd6 packages
>   were interpreted as another device's attempt to use the same one.
>
>   is it a general requirement of lwip netif drivers to suppress sent
>   packages? or should the nd6 have checked for whether the package 
> comes
>   from an own netif?
>
> * i've enabled unconditional reception of multicasts for now. i could
>   leave it off by default and fully implement igmp_mac_filter and
>   mld_mac_filter, but i could still not know if i missed any other
>   potential indications of protocol's willingness to receive 
> multicast.
>
>   is there any interface i cen implement to be informed of the general
>   need to receive multicasts?
>
> best regards
> chrysn
>
> [1]
> https://gitlab.com/enc28j60driver/enc28j60driver/commit/f60093dd58eabaaaa92c0c87b172d1e5310e08f7