Re: [PATCH] net-icmp: make icmp{,v6} (ping) sockets available to all by default

["Maciej Żenczykowski" <zenczykowski@gmail.com>]

> Ido's response gave introductory commands which can also be found here:
>     https://www.kernel.org/doc/Documentation/networking/vrf.txt
>
> This should answer most questions about more advanced topics:
>     http://schd.ws/hosted_files/ossna2017/fe/vrf-tutorial-oss.pdf
>
> Lately, I am putting blogs on https://people.kernel.org/dsahern for
> recurring questions.

Thanks for that - I'll give it a look.

> Rumors are ugly. If in doubt, ask. LLA with VRF is a primary requirement
> from the beginning.

LLA? Link Level Aggregation?
Not sure what that has to do with VRF though... that's just bonding/teaming??
and seems to work fine without VRF at a (at least to me conceptually)
even lower layer.

> With 5.3 and up, you can have IPv4 routes with IPv6 LLA gateways with
> and without VRFs.

Ah, see... the latest phone hardware that I still don't have access to
(mostly because of covid),
is only 4.19 based (as such doing dev work on 4.14 or VMs, but getting
wifi/cell emulation
and ipv6 working right in VMs is very very hard, though we're making
slow progress).

5.4 hardware is probably 10 months out (assuming things return to normal).
We're always ~2 years behind, and playing catchup... :-(

And I recently switched some of our servers (those were stuck on 4.3
until very recently)
to use IPv4 egress routing via IPv6 ND, that's a great improvement
(and came at just the right time),
but again I guess I don't see how VRF fits in to the picture - this
seems to be just a use NDv6 for foo
instead of ARP for bar to figure out dst mac type of thing...

Obviously I haven't read your links, so perhaps all my questions will
be answered.
(I'm rambling, mostly writing this email to thank you for the links)