Re: [EXT] Re: [RFC, net-next] net: qos: introduce a redundancy flow action

[Vladimir Oltean <vladimir.oltean@nxp.com>]

On Fri, Nov 20, 2020 at 09:39:12AM +0000, Xiaoliang Yang wrote:
> On 2020-11-18 3:01 Joergen Andreasen wrote:
> > I like your idea about using filter actions for FRER configuration.
> >
> > I think this is a good starting point but I think that this approach
> > will only allow us to configure end systems and not relay systems in
> > bridges/switches.
> >
> > In the following I refer to sections and figures in 802.1CB-2017.
> >
> > I am missing the following possibilities:
> > Configure split without adding an r-tag (Figure C-4 Relay system C).
> > Configure recovery without popping the r-tag (Figure C4 Relay system F).
> > Disable flooding and learning per VLAN (Section C.7).
> > Select between vector and match recovery algorithm (Section 7.4.3.4 and 7.4.3.5).
> > Configure history length if vector algorithm is used (Section 10.4.1.6).
> > Configure reset timeout (Section 10.4.1.7).
> > Adding an individual recovery function (Section 7.5).
> > Counters to be used for latent error detection (Section 7.4.4).
> >
> > I would prefer to use the term 'frer' instead of 'red' or 'redundancy'
> > in all definitions and functions except for 'redundancy-tag'.
> 
> Thanks for your suggestion, it's very useful to me. I ignored frer on
> relay system. I will study sections and features you mentioned on
> Spec. If using a new tc-frer action is ok, I will perfect and update
> it.

Is replicated IP multicast (with IGMP/MLD snooping) something that is
going to work using your current abstraction? I think this is one area
that is required to work at a higher level than the level of a physical
port.