[PATCH v4 1/4] ethernet: arc: Add support for Rockchip SoC layer device tree bindings

[Heiko Stübner]

Am Mittwoch, 3. September 2014, 13:14:54 schrieb Arnd Bergmann:
> On Wednesday 03 September 2014 10:27:13 Romain Perier wrote:
> > +static const struct emac_rockchip_soc_data emac_rockchip_dt_data[] = {
> > +       { .grf_offset = 0x154 }, /* rk3066 */
> > +       { .grf_offset = 0x0a4 }, /* rk3188 */
> > +};
> > +
> > +static const struct of_device_id emac_rockchip_dt_ids[] = {
> > +       { .compatible = "rockchip,rk3066-emac", .data =
> > &emac_rockchip_dt_data[0] }, +       { .compatible =
> > "rockchip,rk3188-emac", .data = &emac_rockchip_dt_data[1] }, +       { /*
> > Sentinel */ }
> > +};
> > +
> 
> One last question: is this the location given as .grf_offset the
> only thing in grf that is potentially of concern to this driver?
> 
> If it is, you can change the binding to include the register number
> in the syscon reference, like
> 
> 	rockchip,grf = <&grf 0x154>;
> 
> and then read it from there, to simplify the code needed to get the
> number from the device id.

I would disagree here :-)

Specific to the emac, there also exists a second register in the grf (0xa8 for 
the rk3188) that contains a field 

----
emac_newrcv_en - the selection of RMII receive selection
0: don't support the data package without header
1: support the data package without header
----

which we don't handle currently but somebody might want to in the future.
[There also is no documentation of this at all]


The dt maintainers also generally suggest to define compatibles for the 
individual socs anyway, even if only one is matched, so I don't see the 
necessity to encode this 2 times
	compatible = "rockchip,rk3188-emac", "rockchip,rk3066-emac";
	rockchip,grf = <&grf 0xa4>;

instead of the more flexible
	compatible = "rockchip,rk3188-emac";
	rockchip,grf = <&grf>;

which other rockchip drivers already use in this form.


Heiko