Re: [PATCH] ARM: dts: imx6sx: document SION necessity of ENET1_REF_CLK1

[Christoph Fritz <chf.fritz@googlemail.com>]

On Wed, 2016-08-17 at 11:26 -0300, Fabio Estevam wrote:
> Hi Christoph,
> 
> On Wed, Aug 17, 2016 at 6:25 AM, Christoph Fritz
> <chf.fritz@googlemail.com> wrote:
> 
> > +/*
> > + * SION bit is necessary for ENET1_REF_CLK1 (ENET2_REF_CLK2 untested) if it is
> > + * used as clock output of IMX6SX_CLK_ENET_REF (ENET1_TX_CLK) to e.g. supply a
> > + * PHY in RMII mode. This configuration is valid if:
> > + *  - bit 1 in field IMX6SX_GPR1_FEC_CLOCK_PAD_DIR_MASK is set
> > + *  - bit 1 in field IMX6SX_GPR1_FEC_CLOCK_MUX_SEL_MASK unset
> 
> So in your case the imx6sx_enet_clk_sel() does not do what you need:
> 
> static void __init imx6sx_enet_clk_sel(void)
> {
>     struct regmap *gpr;
> 
>     gpr = syscon_regmap_lookup_by_compatible("fsl,imx6sx-iomuxc-gpr");
>     if (!IS_ERR(gpr)) {
>         regmap_update_bits(gpr, IOMUXC_GPR1,
>                    IMX6SX_GPR1_FEC_CLOCK_MUX_SEL_MASK, 0);
>         regmap_update_bits(gpr, IOMUXC_GPR1,
>                    IMX6SX_GPR1_FEC_CLOCK_PAD_DIR_MASK, 0);
>     } else {
>         pr_err("failed to find fsl,imx6sx-iomux-gpr regmap\n");
>     }
> }
> 
> It seems that it is not a good idea to have imx6sx_enet_clk_sel() in
> common code as the GPR1 setting can change from board to board.

Yes, currently I'm adapting/quirking the fec config there and in
imx6sx_clocks_init() to set the Phy-Clock-Frequency to 50Mhz.

> I think we need the fec driver to be in charge of configuring the GPR1 register.

What about making the Phy-Clock-Frequency IMX6SX_CLK_ENET_REF
adjustable?

> This is unrelated to your patch though :-)

Thanks
 -- Christoph