[OpenWrt-Devel] Regarding bridge multicast-to-unicast patch

Linus Lüssing linus.luessing at c0d3.blue
Wed Jun 3 12:20:00 EDT 2015


On Wed, Jun 03, 2015 at 09:56:41AM +0200, Hans Dedecker wrote:
> On Wed, Jun 3, 2015 at 8:50 AM, Steven Barth <cyrus at openwrt.org> wrote:
> > Thanks for the quick reply.
> >
> >
> > On 03.06.2015 08:26, Linus Lüssing wrote:
> >> For IPv6 and a non-zero IPv4 source address it'll use the normal
> >> IGMP/MLD querier election mechanism, meaning the one with the lowest
> >> IP address will become the selected querier and the only one querying
> >> on the link.
> > So especially for v6 - since the kernel doesn't do IGMPv3 / MLDv2 - this
> > effectively downgrades all the bridge links to ASM then (at least with a
> > 50% chance, i.e. when the kernel wins due to lower MAC-address), which
> > is a bit meh.

Right, that's currently a trade-off, if you don't have
source-specific-multicast or another MLD querier on the network,
it is nice to have it enabled by default. Otherwise there's a
disadvantage. Not sure how often source-specific-multicast is actually
used these days and what the philosophy and typical scenario for the
default values in OpenWRT is. If the defaults were optimized for the
average guys home router, who probably doesn't have an SSM multicast
router, but might have an app to do some multicast music/video/conference
streaming it should probably be enabled by default to avoid his
wifi becoming unusable/congested. Don't feel strongly about the default
for the querier though.

> >
> > Now remaining question is, does the kernel detect if there is a
> > userspace-querier running on the same machine and silence itself then?
> > If not we probably need to address this as well somehow.
> I can confirm the bridge querier silence itself when an
> userspace-querier is running at least for IGMP.
> Did not check the MLD behavior yet.

Yes, that's probably due to the bridge zero-source querier becoming
quiet when it hears another IGMP query. For MLD that should be different
yes.

Steven, could you elaborate a little more on the scenario and
explain why it is a must to shut up automatically when having a
local user-space querier? If people are able to manually install and
configure a complex multicast router, aren't they capable of
manually disabling the bridge querier, too (there's even a netifd
option for that)?

Cheers, Linus
_______________________________________________
openwrt-devel mailing list
openwrt-devel at lists.openwrt.org
https://lists.openwrt.org/cgi-bin/mailman/listinfo/openwrt-devel


More information about the openwrt-devel mailing list