On 28/11/19, Marc Gonzalez wrote:
> On 28/11/2019 10:55, Oliver Graute wrote:
>
> > On 28/11/19, Marc Gonzalez wrote:
> >
> >> On 29/10/2019 13:23, Oliver Graute wrote:
> >>
> >>> +&fec1 {
> >>> + pinctrl-names = "default";
> >>> + pinctrl-0 = <&pinctrl_fec1>;
> >>> + phy-mode = "rgmii";
> >>> + phy-handle = <ðphy0>;
> >>> + fsl,magic-packet;
> >>> + fsl,rgmii_txc_dly;
> >>> + fsl,rgmii_rxc_dly;
> >>> + status = "okay";
> >>
> >> The two fsl,rgmii* properties do not exist in mainline.
> >> I suppose there were copied from downstream?
> >
> > you are right, I'll remove them.
>
> You should first check what the downstream driver does for them.
> And check if there is an equivalent action in mainline.
> These delays tend to be required for the PHY to work at all.
ok as far as I see there is currently no equilant action in mainline.
Downstream linux-imx use rgmii_txc_dly and rgmii_rxc_dly in fec_probe()
and fec_restart() for some imx8qm-fec and imx8mq-fec PHY delay quirks.
Perhaps this missing quirks are related to the random crashes in that
driver I have with imx8qm.
[ 129.211959] fec 5b040000.ethernet eth0: rcv is not +last
[ 129.217300] fec 5b040000.ethernet eth0: rcv is not +last
[ 129.222647] fec 5b040000.ethernet eth0: rcv is not +last
[ 129.227966] fec 5b040000.ethernet eth0: rcv is not +last
[ 129.233282] fec 5b040000.ethernet eth0: rcv is not +last
Best regards,
Oliver
