Re: [PATCH v7 3/3] net: stmmac: dwmac-sophgo: Add phy interface filter

On Mon, Nov 10, 2025 at 08:09:10PM +0000, Simon Horman wrote:
> On Fri, Nov 07, 2025 at 07:17:15PM +0800, Inochi Amaoto wrote:
> > As the SG2042 has an internal rx delay, the delay should be removed
> > when initializing the mac, otherwise the phy will be misconfigurated.
> > 
> > Fixes: 543009e2d4cd ("net: stmmac: dwmac-sophgo: Add support for Sophgo SG2042 SoC")
> > Signed-off-by: Inochi Amaoto <[email protected]>
> > Tested-by: Han Gao <[email protected]>
> > Reviewed-by: Andrew Lunn <[email protected]>
> > ---
> >  .../ethernet/stmicro/stmmac/dwmac-sophgo.c    | 20 ++++++++++++++++++-
> >  1 file changed, 19 insertions(+), 1 deletion(-)
> > 
> > diff --git a/drivers/net/ethernet/stmicro/stmmac/dwmac-sophgo.c b/drivers/net/ethernet/stmicro/stmmac/dwmac-sophgo.c
> 
> ...
> 
> > @@ -50,11 +56,23 @@ static int sophgo_dwmac_probe(struct platform_device *pdev)
> >  	if (ret)
> >  		return ret;
> >  
> > +	data = device_get_match_data(&pdev->dev);
> > +	if (data && data->has_internal_rx_delay) {
> > +		plat_dat->phy_interface = phy_fix_phy_mode_for_mac_delays(plat_dat->phy_interface,
> > +									  false, true);
> > +		if (plat_dat->phy_interface == PHY_INTERFACE_MODE_NA)
> > +			return -EINVAL;
> 
> I'm sorry if this is a false positive. Because, more so than Russell [1], I
> confused about how about the treatment of phy_interface. But it seems that
> there is a miss match between the use of phy_fix_phy_mode_for_mac_delays()
> above and the binding.
> 
> The call to phy_fix_phy_mode_for_mac_delays() above will return
> PHY_INTERFACE_MODE_NA unless phy_interface is PHY_INTERFACE_MODE_RGMII_ID
> or PHY_INTERFACE_MODE_RGMII_RXID.
> 
>   phy_interface_t phy_fix_phy_mode_for_mac_delays(phy_interface_t interface,
> 						bool mac_txid, bool mac_rxid)
>   ...
> 	if (mac_rxid) {
> 		if (interface == PHY_INTERFACE_MODE_RGMII_ID)
> 			return PHY_INTERFACE_MODE_RGMII_TXID;
> 		if (interface == PHY_INTERFACE_MODE_RGMII_RXID)
> 			return PHY_INTERFACE_MODE_RGMII;
> 		return PHY_INTERFACE_MODE_NA;
> 	}
>   ...
> 
> Looking at phy_modes(), unsurprisingly, the following mappings occur:
> * "rgmii" -> PHY_INTERFACE_MODE_RGMII
> * "rgmii-id" -> PHY_INTERFACE_MODE_RGMII_ID
> * "rgmii-rxid" -> PHY_INTERFACE_MODE_RGMII_RXID
> * "rgmii-txid" -> PHY_INTERFACE_MODE_RGMII_TXID
> 
> And in the binding, patch 1/3 of this series, only phy-mode rgmii-txid or
> rgmii-id is allowed.
> 

rgmii-txid is a mistake and should be rgmii-rxid. This is because
the mac of SG2042 add rx delay, and the phy can only add tx delay or
no delay. So the phy-mode can only be rgmii-id or rgmii-rxid. I will
fix it in the next version.

> But if rgmii-txid is used, PHY_INTERFACE_MODE_RGMII_TXID will be passed to
> phy_fix_phy_mode_for_mac_delays(), which will return PHY_INTERFACE_MODE_NA.
> 
> Again, I'm confused about the mapping in phy_fix_phy_mode_for_mac_delays().
> But there does seem to be some inconsistency between the binding and
> the driver implementation here.
> 

I think this inconsistency begin with the change for ethernet-controller
binding.
https://lore.kernel.org/all/[email protected]/

And this function serves as an helper so the driver can only add
the delay once.

> Flagged by Claude Code with https://github.com/masoncl/review-prompts/ 
> 
> [1] https://lore.kernel.org/all/[email protected]/
> 
> ...

Regards,
Inochi