Hi Andrew,
On Tue, Apr 16, 2024 at 3:05 PM Andrew Lunn <andrew@xxxxxxx> wrote:
> > > > + if (!of_property_read_u32(np, "rx-internal-delay-ps", &delay))
> > > > + if (delay)
> > > > + val |= GPOUT_RDM;
> > > > +
> > > > + /* Valid values are 0 and 2000, according to DT bindings */
> > > > + if (!of_property_read_u32(np, "tx-internal-delay-ps", &delay))
> > > > + if (delay)
> > > > + val |= GPOUT_TDM;
> > > > +
> > > > + rtsn_write(priv, GPOUT, val);
> > >
> > > So you seem to be using it as bool?
> >
> > Yes.
> >
> > > That is wrong. It is a number of pico seconds!
> >
> > The issue is that the hardware only supports no delay or a fixed delay
> > that can depend on electric properties of the board.
>
> The general convention is that the MAC does not add delays, it leaves
> it to the PHY. Probably 95% of boards are like this, and many MAC
> drivers don't even add support for configuring their hardware delays,
> it is not needed. Those that do, it is generally for fine tuning the
> delays, being able to add/remove 100s of pico seconds, not the full
> 2us. This hardware cannot do that.
>
> So i suggest you drop all this code, and just hard code the delay to
> 0ps.
IIRC (from users of RAVB, which have similar delay bits), the issue
is that some boards require a larger delay than the maximum delay
supported by the Micrel KSZ9031 PHY (960 ps). Hence these
boards need to enable the MAC delay.
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx
In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
