Hi Scott,
Please see my reply inline.
> -----Original Message-----
> From: Scott Wood <oss@xxxxxxxxxxxx>
> Sent: 2018年10月21日 7:54
> To: Andy Tang <andy.tang@xxxxxxx>; mturquette@xxxxxxxxxxxx
> Cc: sboyd@xxxxxxxxxx; robh+dt@xxxxxxxxxx; mark.rutland@xxxxxxx;
> linux-clk@xxxxxxxxxxxxxxx; devicetree@xxxxxxxxxxxxxxx
> Subject: Re: [PATCH 2/3] powerpc: t102x: upgrade the legacy clock node
>
> On Wed, 2018-09-19 at 15:39 +0800, andy.tang@xxxxxxx wrote:
> > +clockgen: global-utilities@e1000 {
> > + compatible = "fsl,qoriq-clockgen";
>
> Where does this compatible string come from?
>
> > + compatible = "fsl,t1023-clockgen";
> > };
>
> And here you overwrite it with only the chip-specific compatible?
>
> Is t1023 incompatible with both fsl,qoriq-clockgen-1.0 and
> fsl,qoriq-clockgen- 2.0? The existing dts says 2.0; is that wrong?
>
> BTW, assuming it is 2.0 compatible and thus the use of
> qoriq-clockgen2.dtsi is correct, the best course of action is probably to to
> remove the legacy stuff from all fsl chips, rather than introduce a new dtsi.
> In fact it'd be nice to see it all removed in any case. :-)
qoriq-clockgen*.dtsi are used by legacy bindings. The contents are all of legacy bindings.
To use new framework, I introduce a new dtsi which contains new bindings and used for all PPC soc.
A chip-specific compatible is needed because driver will use it to get chip-specific clock tree information.
The clock information was defined in driver not in dts in new framework, remember?
This patch set is the first step to convert it to using new framework. After all the chips have been updated,
the legacy stuff will be removed. Including qoriq-clockgen*.dtsi.
>
> Also, please post any patches that you want me to apply to the
> linuxppc-dev mailing list.
Sure, no problem.
BR,
Andy
>
> -Scott
