Re: FWD: Re: [PATCH net-next v1 1/1] net: dsa: microchip: ksz8: Make flow control, speed, and duplex on CPU port configurable

[Vladimir Oltean <olteanv@gmail.com>]

On Tue, Apr 11, 2023 at 12:50:28PM +0100, Russell King (Oracle) wrote:
> On Tue, Apr 11, 2023 at 02:16:09PM +0300, Vladimir Oltean wrote:
> > On Tue, Apr 11, 2023 at 10:17:47AM +0100, Russell King (Oracle) wrote:
> > > Since we can't manually control the tx and rx pause enables, I think
> > > the only sensible way forward with this would be to either globally
> > > disable pause on the device, and not report support for any pause
> > > modes,
> > 
> > This implies restarting autoneg on all the other switch ports when one
> > port's flow control mode is changed?
> 
> From my reading of these global register descriptions, no it doesn't,
> and even if we did restart aneg, it would have no overall system effect
> because the advertisements for each port haven't been changed. It's
> mad hardware.
> 
> What I was meaning above is that we configure the entire switch to
> either do autonegotiated flow control at setup time, or we configure
> the switch to never do flow control.

I was thinking you were suggesting to also modify the advertisement in
software from those other ports to 00 when the flow control was forced
off on one. Otherwise (it seems you weren't), I think it's a bit
counter-productive to configure the switch to never do flow control,
when the only problem seems to be with the forced modes but autoneg is fine.

> > I don't object to documenting that manually forcing flow control off is
> > broken and leaving it at that (and the way to force it off would be to
> > not advertise any of the 2 bits).
> > 
> > But why advertise only 11 (Asym_Pause | Pause) when the PHYs integrated
> > here have the advertisement configurable (presumably also through the
> > micrel.c PHY driver)? They would advertise in accordance with ethtool, no?
> > 
> > I may have missed something.
> 
> I think you have. I'm only talking about the ability to control flow
> control manually via ethtool -A. Changing it via the advertisement
> (ethtool -s) would still work.

That part was understood.