netdev - Re: [EXT] Re: [PATCH net-next] net/hanic: Add the hanic network interface for high availability links

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Date:   Tue, 21 Feb 2023 12:03:08 +0100
From:   Ferenc Fejes <primalgamer@...il.com>
To:     Andrew Lunn <andrew@...n.ch>, Jiri Pirko <jiri@...nulli.us>
Cc:     Vladimir Oltean <vladimir.oltean@....com>,
        Steve Williams <steve.williams@...cruise.com>,
        Jakub Kicinski <kuba@...nel.org>, netdev@...r.kernel.org,
        vinicius.gomes@...el.com, xiaoliang.yang_1@....com
Subject: Re: [EXT] Re: [PATCH net-next] net/hanic: Add the hanic network
 interface for high availability links

Hi!

On Wed, 2022-11-23 at 16:12 +0100, Andrew Lunn wrote:
> > I guess for the same reason other soft netdevice driver are in the
> > kernel. You can do bridge, bond, etc in a silimilar way you
> > described
> > here...
> 
> Performance of tun/tap is also not great.

Yes, but mostly because of the throughput. It also that severe the
latency issue too?

FRER tipically used in TSN use-cases, where 10 and 100 Mbps covers like
90% what you need. 1Gbps at max. Yes I agree that the concept can be
useful everywhere when redundancy over multiple hops necessary but I
cant see other than industrial use-cases (like datacenter or so)

> 
> Doing this in the kernel does seem correct. But we need one
> implementation which can be expanded over time to cover everything in
> the standard. From what has been said so far, it seems like this
> implementation focuses on leaf nodes, because that is what the author
> of the code is interested in. But is the design generic enough it can
> be expanded to cover everything else? I'm not saying it actually
> needs
> to implement it now, we just need to have a vision of how it can be
> extended to implement the rest. What we don't want is one way for
> leaf
> nodes, and a completely different code base for other nodes.
> 
>        Andrew

I'm afraid not generic enough.
There are MPLS encapsulated DetNet flows as well (PREOF) which is
different than .1CB but replication/elimination concept is the same.
Also it would be nice to have to store the R-tag seq as a metadata and
replicate it in hybrid L2/L3 fashion (TSN and DetNet member streams)

Best,
Ferenc