Re: [PATCH net-next 2/3] net: stmmac: Allow supporting coarse adjustment mode

[Kory Maincent <kory.maincent@bootlin.com>]

On Thu, 23 Oct 2025 10:29:26 +0200
Maxime Chevallier <maxime.chevallier@bootlin.com> wrote:

> On 22/10/2025 01:02, Jakub Kicinski wrote:
> > On Tue, 21 Oct 2025 10:02:01 +0200 Maxime Chevallier wrote:  
> >> Let me know if you need more clarifications on this  
> > 
> > The explanation was excellent, thank you. I wonder why it's designed
> > in such an odd way, instead of just having current_time with some
> > extra/fractional bits not visible in the timestamp. Sigh.
> > 
> > In any case, I don't feel strongly but it definitely seems to me like
> > the crucial distinction here is not the precision of the timestamp but
> > whether the user intends to dial the frequency.  
> 
> Yes indeed. I don't have a clear view on wether this is something unique
> to stmmac or if this is common enough to justify using the tsconfig API.
> 
> As we discuss this, I would tend to think devlink is the way, as this
> all boils down to how this particular HW works. Moreover, if we use a
> dedicated hwprov qualifier, where do we make it sit in the current
> hierarchy (precise > approx) that's used for the TS source selection ?

That's ok to me. I was not strongly against devlink in either way, and I didn't
have real arguments. Let's go for devlink, we still can move it to tsconfig API
later if it's needed. 

Regards,
-- 
Köry Maincent, Bootlin
Embedded Linux and kernel engineering
https://bootlin.com