Re: [PATCH net-next] net: phy: vitesse: add inband caps and configuration

[Charles Perry <charles.perry@microchip.com>]

On Wed, Mar 11, 2026 at 02:26:11PM +0000, Russell King (Oracle) wrote:
> On Wed, Mar 11, 2026 at 03:01:35PM +0100, Geert Uytterhoeven wrote:
> > 	Hi Russell,
> > 
> > On Wed, 11 Mar 2026 at 01:07, Russell King <rmk@armlinux.org.uk> wrote:
> > > Add support for VSC8662 reporting its inband capabilities, and also
> > > hook to configure the PHY's inband mode.
> > >
> > > This fixes a regression in the macb driver caused by commit
> > > 1338cfef1ff1 ("net: macb: fix SGMII with inband aneg disabled")
> > 
> > Fixes: 1338cfef1ff1b958 ("net: macb: fix SGMII with inband aneg disabled")
> > 
> > > Reported-by: Conor Dooley <conor@kernel.org>
> > > Link: https://lore.kernel.org/r/20260304-nebulizer-rounding-40fbc81a2ba1@spud
> > 
> > s/Link/Closes/
> 
> I avoided it because any other PHY used with macb that is also connected
> via SGMII will run into this same issue. It just happens that adding
> this support fixes the above commit.
> 
> Conversely, we have stmmac, which unconditionally enables SGMII inband
> at the MAC end no matter what phylink says to do, and should this
> PHY be used with this patch in a stmmac system, it will cause that to
> break, because we end up with either end misconfigured.
> 

macb also unconditionally activated inband aneg before I added the
->pcs_config() callback.

One of the way I think we could fix macb is by adding:

        if (bp->phy_interface == PHY_INTERFACE_MODE_SGMII) {
                bp->phylink_config.poll_fixed_state = true;
                bp->phylink_config.get_fixed_state = macb_get_pcs_fixed_state;
+               bp->phylink_config.default_an_inband = true;
        }

in macb_mii_probe(). This would force phylink into using inband aneg
every time, unless there's a fixed link or the PHY explicitly doesn't
support in-band. This would make it compatible with legacy system that were
setup with in-band aneg activated. It would fix MPFS without the need for
this patch or device-tree modifications ('managed = "in-band-status"').

Thanks,
Charles

> Also, if this commit is applied without the macb change, you'll
> probably also find that macb breaks (please test that.) 
> 
> So, while this is a solution and a step forward, I'd rather not suggest
> that it is an official fix to the macb issue - it's merely filling in
> a missing piece of the jigsaw.
> 
> Sadly, this is a fundamental issue with SGMII implementations today:
> whether SGMII inband is used/required is completely random both at the
> MAC/PCS end but also at the PHY end. If one ends up with a mismatch,
> then things stop working.
> 
> As I say, the problem with fixing the PHY end of the link is that
> it could cause a different network driver to regress.
> 
> So yes, this is a solution to the macb problem, but I regard it as
> high risk.
> 
> That said, given the SGMII mess, there is no easy way forward.
> 
> -- 
> RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
> FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!