Re: [PATCH net-next 2/2] net: dsa: microchip: add KSZ8 change_tag_protocol support

[Vladimir Oltean <olteanv@gmail.com>]

Hi Pieter,

On Mon, Aug 19, 2024 at 12:12:35PM +0200, vtpieter@gmail.com wrote:
> From: Pieter Van Trappen <pieter.van.trappen@cern.ch>
> 
> Add support for changing the KSZ8 switches tag protocol. In fact
> these devices can only enable or disable the tail tag, so there's
> really only three supported protocols:
> - DSA_TAG_PROTO_KSZ8795 for KSZ87xx
> - DSA_TAG_PROTO_KSZ9893 for KSZ88x3
> - DSA_TAG_PROTO_NONE
> 
> When disabled, this can be used as a workaround for the 'Common
> pitfalls using DSA setups' [1] to use the conduit network interface as
> a regular one, admittedly forgoing most DSA functionality and using
> the device as an unmanaged switch whilst allowing control
> operations (ethtool, PHY management, WoL).

Concretely, what is it that you wish to accomplish? I see you chose to
ignore my previous NACK due to the lack of a strong justification for
disabling the tagging protocol.
https://lore.kernel.org/netdev/20240801134401.h24ikzuoiakwg4i4@skbuf/

> Implementing the new software-defined DSA tagging protocol tag_8021q
> [2] for these devices seems overkill for this use case at the time
> being.

I think there's a misunderstanding about tag_8021q. It does not disable
the tagging protocol. But rather, it helps you implement a tagging
protocol when the hardware does not want to cooperate. So I don't see
how it would have helped you in your goal (whatever that is), and why
mention it.

tag_8021q exists because it is my goal for DSA_TAG_PROTO_NONE to
eventually disappear. The trend is for drivers to be converted from
DSA_TAG_PROTO_NONE to something else (like DSA_TAG_PROTO_VSC73XX_8021Q),
not the other way around. It's a strong usability concern to not be able
to ping through the port net devices.

At the very least we need consensus among the current DSA maintainers
that accepting 'none' as an alternative tagging protocol is acceptable.