Re: [RFC PATCH net-next v3 4/5] net: dsa: qca8k: flag QCA8337 internal CPU PHYs for SmartSpeed

[Yongzhao Chen <yongzhao.derek@gmail.com>]

Hi Andrew,

> Also tricky. What we don't want is other developers trying to abuse
> this to turn it into a configuration option, rather than a hardware
> property. So i don't think it should be a generic property. Lets make
> it a qualcomm specific property. I would also put 'workaround' in the
> property name, again making it clear this is not intended to be used
> for configuration.

Thank you, that makes sense. I'll use a Qualcomm-specific property with
"workaround" in its name, and word the commit message around what has
been observed, as you suggested: an on-board PHY-to-PHY connection
without a cable, where SmartSpeed has been seen to downshift
incorrectly. The workaround would then apply only to boards that set
the property, rather than to every QCA8337 internal CPU PHY as in v3.

One correction to the data I sent you: as Ziyang pointed out, the
IPQ5018 short-cable DAC values were not being applied on this board,
because the driver does not shift the field value. My earlier A/B runs
were therefore done without them. I have repeated the A/B with the DAC
values corrected, and the result is the same. The details are in my
reply to Ziyang in this thread.

Before sending the next revision, I would like to check one more
thing. In these boots the QCA8337 CPU PHY is reset about 38 s before
the IPQ5018 PHY has its DAC values configured. If setting the DAC before
the switch-side PHY starts negotiating avoids the downshift, that would
suggest fixing the initialisation order rather than adding a DT
workaround. I'll report back on what I find before sending the next
revision. If the workaround is still needed, I'll propose the property
name and binding with it.

Thanks,
Yongzhao Chen