netdev - Re: [net][hsr] Question regarding HSR RedBox functionality implementation (preferably on KSZ9477)

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: Windows password security audit tool. GUI, reports in PDF.

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

Message-ID: <20240109125205.u6yc3z4neter24ae@skbuf>
Date: Tue, 9 Jan 2024 14:52:05 +0200
From: Vladimir Oltean <olteanv@...il.com>
To: Lukasz Majewski <lukma@...x.de>
Cc: Andrew Lunn <andrew@...n.ch>, Tristram.Ha@...rochip.com,
	Oleksij Rempel <o.rempel@...gutronix.de>,
	UNGLinuxDriver@...rochip.com, netdev@...r.kernel.org,
	Sebastian Andrzej Siewior <bigeasy@...utronix.de>,
	George McCollister <george.mccollister@...il.com>
Subject: Re: [net][hsr] Question regarding HSR RedBox functionality
 implementation (preferably on KSZ9477)

Hi Lukasz,

On Tue, Jan 09, 2024 at 01:32:34PM +0100, Lukasz Majewski wrote:
> However, I'm wondering how the mainline Linux kernel could handle HSR
> RedBox functionality (on document [1], Figure 2. we do have "bridge" -
> OSI L2).
> 
> To be more interesting - br0 can be created between hsr0 and e.g. lan3.
> But as expected communication breaks on both directions (to SAN and to
> HSR ring).

Yes, I suppose this is how a RedBox should be modeled. In principle it's
identical to how bridging with LAG ports (bond, team) works - either in
software or offloaded. The trouble is that the HSR driver seems to only
work with the DANH/DANP roles (as also mentioned in
Documentation/networking/dsa/dsa.rst). I don't remember what doesn't
work (or if I ever knew at all). It might be the address substitution
from hsr_xmit() that masks the MAC address of the SAN side device?

> Is there a similar functionality already present in the Linux kernel
> (so this approach could be reused)?
> 
> My (very rough idea) would be to extend KSZ9477 bridge join functions
> to check if HSR capable interface is "bridged" and then handle frames
> in a special way.
> 
> However, I would like to first ask for as much input as possible - to
> avoid any unnecessary work.

First I'd figure out why the software data path isn't working, and if it
can be fixed. Then, fix that if possible, and add a new selftest to
tools/testing/selftests/net/forwarding/, that should pass using veth
interfaces as lower ports.

Then, offloading something that has a clear model in software should be
relatively easy, though you might need to add some logic to DSA. This is
one place that needs to be edited, there may be others.

	/* dsa_port_pre_hsr_leave is not yet necessary since hsr devices cannot
	 * meaningfully placed under a bridge yet
	 */

> 
> Thanks in advance for help :-)
> 
> Link:
> 
> [1] -
> https://ww1.microchip.com/downloads/en/Appnotes/AN3474-KSZ9477-High-Availability-Seamless-Redundancy-Application-Note-00003474A.pdf
> 
> 
> Best regards,
> 
> Lukasz Majewski