Re: [PATCH v3 1/3] net: dsa: microchip: implement KSZ87xx Module 3 low-loss cable errata

[Fidelio LAWSON <lawson.fidelio@gmail.com>]

On 4/16/26 14:25, Andrew Lunn wrote:
>> Yes, I think a reasonable compromise could be to expose three tunables:
>>
>> - a boolean "short-cable" tunable, which applies the known good settings
>>    (LPF 62 MHz BW, DSP EQ initial value 0).
>>
>> - an integer LPF bandwidth tunable, for advanced use cases where further
>>    tuning is needed;
>>
>> - an integer DSP EQ initial value tunable, for the same advanced cases.
>>
>> The boolean tunable would follow the KISS principle and cover the common
>> scenario, while the more granular controls would remain optional.
> 
> How do the three interact? Do you need to first enable short-cable
> before you set LPG bandwidth or DSP EQ? If it is not enabled, do you
> get -EINVAL?
> 
> It seems like having extack would be useful to return informative
> error messages to user space, however, that requires netlink
> ethtool. And ETHTOOL_PHY_STUNABLE has not been added to netlink
> ethtool yet :-(
> 
> 	Andrew

My intention would be to keep the interactions as simple and
non-surprising as possible, and avoid requiring any particular ordering
or state machine between the tunables.

The boolean short-cable tunable would simply apply the preset
in one step. The LPF bandwidth and DSP EQ initial value tunables would 
be orthogonal knobs which can be set independently at any time, 
regardless of whether short-cable is enabled or not.

With this model, we don’t need to return -EINVAL for combinations or
ordering, and userspace does not need detailed error reporting. The
tunables behave more like simple setters than a mode switch, which
keeps the API predictable and avoids the need for extack or netlink
ethtool support at this point.

Fidelio