Re: [RFC 5/5] [powerpc] Implement a p1010rdb clock source.

[Robin Holt <holt-sJ/iWh9BUns@public.gmane.org>]

On Mon, Aug 08, 2011 at 05:33:53PM +0200, Wolfgang Grandegger wrote:
> On 08/08/2011 05:14 PM, Marc Kleine-Budde wrote:
...

> > On 08/08/2011 04:59 PM, Wolfgang Grandegger wrote:
> >> On 08/08/2011 04:44 PM, Robin Holt wrote:
> >>> On Mon, Aug 08, 2011 at 04:37:44PM +0200, Wolfgang Grandegger wrote:
> >>>> On 08/08/2011 04:21 PM, Robin Holt wrote:
> >>>>> On Mon, Aug 08, 2011 at 04:16:27PM +0200, Wolfgang Grandegger wrote:
> >>>>>> On 08/08/2011 03:56 PM, Robin Holt wrote:
> >>>>>>>> commit 65bb8b060a873fa4f5188f2951081f6011259614
> >>>>>>>> Author: Bhaskar Upadhaya <Bhaskar.Upadhaya-KZfg59tc24xl57MIdRCFDg@public.gmane.org>
> >>>>>>>> Date:   Fri Mar 4 20:27:58 2011 +0530
> >>>>>>>
> >>>>>>> On a side note, that commit fixes up "fsl,flexcan-v1.0"
> >>>>>>> ...
> >>>>>>> +       do_fixup_by_compat_u32(blob, "fsl,flexcan-v1.0",
> >>>>>>> +                       "clock_freq", gd->bus_clk, 1);
> >>>>>>>
> >>>>>>> Should I go back to flexcan-v1.0 in my patches?
> >>>>>>
> >>>>>> Well, no. Let's wait. I don't think we need it. Also, it sets
> >>>>>> "clock_freq" while
> >>>>>>
> >>>>>>  http://lxr.linux.no/#linux+v3.0.1/Documentation/devicetree/bindings/net/can/fsl-flexcan.txt
> >>>>>>
> >>>>>> documents "clock-frequencies"... :-(.
> >>>>>
> >>>>> You answered a different question that I was asking.  I was asking if
> >>>>> I should change fsl,flexcan back to fsl,flexcan-v1.0 as documented on
> >>>>> line 5.  The clock_freq looks like a uboot change will need to be made
> >>>>> as well.
> >>>>
> >>>> Well, I wrote above: "Well, no. Let's wait. I don't think we need it."
> >>>>
> >>>> For the P1010 we can sinmply derive the clock frequency from
> >>>> "fsl_get_sys_freq()", which is fine for the time being. No extra
> >>>> properties, etc. The clk implemetation might go into
> >>>>
> >>>>  http://lxr.linux.no/#linux+v3.0.1/arch/powerpc/platforms/85xx/clock.c
> >>>>
> >>>> or
> >>>>
> >>>>  http://lxr.linux.no/#linux+v3.0.1/arch/powerpc/sysdev/fsl_soc.c
> >>>>
> >>>> And may depend on HAVE_CAN_FLEXCAN
> >>>>
> >>>> BTW, I have not found HAVE_CAN_FLEXCAN in your patch. What kernel are
> >>>> you using?
> >>>
> >>> I am starting with the v3.0 kernel, apply one patch from the freescale BSP
> >>> we receive under NDA which introduces the P1010RDB board into the QorIQ
> >>> platform, and then work from there for the flexcan stuff.  That patch
> >>> introduces the HAVE_CAN_FLEXCAN.  I do not like how freescale structured
> >>> that Kconfig bit, so I have tweaked it to be selected automatically
> >>> when P1010RDB, NET, and CAN are selected.  That allows the CAN_FLEXCAN
> >>> selection to determine is we are going to build the flexcan.c file.
> >>
> >> ARM boards select HAVE_CAN_FLEXCAN and I do not see a good reason why
> >> we should do it differently for PowerPC. 
> >>
> >> For mainline inclusion, you should provide your patches against the
> >> David Millers "net-next-2.6" tree, which already seems to have support
> >> for the P1010RDB:
> >>
> >>   config P1010_RDB
> >>         bool "Freescale P1010RDB"
> >>         select DEFAULT_UIMAGE
> >>         help
> >>           This option enables support for the MPC85xx RDB (P1010 RDB) board
> >>
> >>           P1010RDB contains P1010Si, which provides CPU performance up to 800
> >>           MHz and 1600 DMIPS, additional functionality and faster interfaces
> >>           (DDR3/3L, SATA II, and PCI  Express).
> >>
> >>
> >>> Our contact with Freescale would prefer that I not post that patch until
> >>> we get the OK from freescale to do so since we received it under NDA.
> >>
> >> I don't think we currently need it. I prefer dropping and cleaning up
> >> the device tree stuff as it is not needed for the P1010 anyway. If a
> >> new processor shows up with enhanced capabilities requiring
> >> configuration via device tree, we or somebody else can provide a patch.
> >> Marc, what do you think?
> > 
> > ACK - The device tree bindings as in mainline's Documentation is a mess.
> > If the powerpc guys are happy with a clock interfaces based approach
> > somewhere in arch/ppc, I'm more than happy to remove:
> > - fsl,flexcan-clock-source (not implemented, even in the fsl driver)
> > 
> > - fsl,flexcan-clock-divider \__ replace with code in arch/ppc, or
> > - clock-frequency           /   a single clock-frequency attribute
> 
> In the "net-next-2.6" tree there is also:
> 
>  $ grep flexcan arch/powerpc/boots/dts/*.dts
>   p1010rdb.dts:			fsl,flexcan-clock-source = "platform";
>   p1010rdb.dts:			fsl,flexcan-clock-source = "platform";
>   p1010si.dtsi:			compatible = "fsl,flexcan-v1.0";
>   p1010si.dtsi:			fsl,flexcan-clock-divider = <2>;
>   p1010si.dtsi:			compatible = "fsl,flexcan-v1.0";
>   p1010si.dtsi:			fsl,flexcan-clock-divider = <2>;
> 
> Especially the fsl,flexcan-clock-divider = <2>; might make people think,
> that they could set something else.

I am currently lost on the direction.  I think I need something like:

1) Patch 1/5 removing the "#include <mach/clock.h>" stays.
2) Patch 2/5 abstracting readl/writel stays.
3) Patch 3/5 of_match for ppc and the match string is "fsl,flexcan" stays.
4) Patch 4/5 I have not been given clear direction to not do it but have
   not gotten a favorable response.
5) Patch 5/5 goes from being a powerpc patch back to being a flexcan.c
   patch which determines the clock source not using the device tree
   information, but rather from some system register.  I need more detail
   on how this would work for both arm and powerpc.  How would I absract
   that or am I providing a flexcan_clk_* set of functions like I have
   in earlier generations of the patch set?

Thanks,
Robin