Re: [PATCH 2/3] mfd: db8500-prcmu: Fetch the PRCMU TCDM base addressfrom Device Tree

From: Linus Walleij
Date: Tue Mar 26 2013 - 05:11:31 EST


On Tue, Mar 26, 2013 at 9:17 AM, Lee Jones <lee.jones@xxxxxxxxxx> wrote:
> On Mon, 25 Mar 2013, Linus Walleij wrote:
>> On Mon, Mar 25, 2013 at 5:46 PM, Lee Jones <lee.jones@xxxxxxxxxx> wrote:
>>
>> >> > res = platform_get_resource_byname(pdev, IORESOURCE_MEM,
>> >> > "prcmu-tcpm");
>> >> > if (!res) {
>> >> > - dev_err(&pdev->dev,
>> >> > - "Error: no prcmu tcpm memory region provided\n");
>> >> > - return;
>> >> > - }
>> >> > - tcpm_base = ioremap(res->start, resource_size(res));
>> >> > + if (np) {
>> >> > + tcpm_np = of_find_node_by_name(np->parent,
>> >> > + "prcmu-tcpm-per4");
>> >> > + if (!tcpm_np) {
>> >> > + dev_err(&pdev->dev,
>> >> > + "no prcmu tcpm mem region provided\n");
>> >> > + return;
>> >> > + }
>> >> > + tcpm_base = of_iomap(tcpm_np, 0);
>> >> > + }
>> >> > + } else
>> >> > + tcpm_base = ioremap(res->start, resource_size(res));
>> >> > +
>> >>
>> >> Why don't you just add these extra registers to the prcmu node itself like
>> >>
>> >> diff --git a/arch/arm/boot/dts/dbx5x0.dtsi b/arch/arm/boot/dts/dbx5x0.dtsi
>> >> index 9de9309..6ee6c31 100644
>> >> --- a/arch/arm/boot/dts/dbx5x0.dtsi
>> >> +++ b/arch/arm/boot/dts/dbx5x0.dtsi
>> >> @@ -191,8 +191,8 @@
>> >>
>> >> prcmu: prcmu@80157000 {
>> >> compatible = "stericsson,db8500-prcmu";
>> >> - reg = <0x80157000 0x1000>;
>> >> - reg-names = "prcmu";
>> >> + reg = <0x80157000 0x1000>, <0x801b0000 0x1000>, <0x801b8000 0x1000>;
>> >> + reg-names = "prcmu", "prcmu-tcpm", "prcmu-tcdm-per4";
>> >> interrupts = <0 47 0x4>;
>> >> #address-cells = <1>;
>> >> #size-cells = <1>;
>> >>
>> >>
>> >> and document those in the prcmu binding?
>> >>
>> >> That would keep the code simpler and the same for both cases.
>> >
>> > I'm not sure you can. These addresses differ from platform to
>> > platform, so I'd need to provide the base PRCMU address in
>> > dbx5x0.dtsi, then the other two in the dts files. Can you do that?
>>
>> Better ask on devicetree-discuss.
>>
>> What it reflects are differences between SoC:s, so I think the
>> right solution is to split it in a three-layer cake:
>>
>> dbx5x0.dtsi -> included by
>> db8500.dtsi -> included by
>> snowball/hrefv60plus.dts
>>
>> dbx5x0.dtsi -> included by
>> dbx540.dtsi -> included by
>> ccu9540/ccu8540.dts
>>
>> Here you may insert the reg property into the node at the
>> db8500.dtsi or dbx540.dtsi layer.
>
> It's easy done, if that's the right way.
>
>> If you prefer I can cook a patch...
>
> ... and I'm fine to do it.
>
> [5 mins later]
>
> Hmm... it appears that in STE's kernel, the same address is used for
> all SoCs:
>
> #define U8500_PRCMU_TCDM_BASE (U8500_PER4_BASE + 0x68000)

Yes that's correct...

> However, in Mainline there is another definition:
>
> #define U9540_PRCMU_TCDM_BASE (U8500_PER4_BASE + 0x6A000)
>
> ... which doesn't seem to be in use anywhere - even for the u9540.
>
> What's the deal big-man?

Probably there was another address at some point in some FPGA or
early silicon and it now needs fixing.

So you should be able to use the same thing on all ASICs, put one
single value in the dbx5x0.dtsi *and* fix this bug at the same time,
good!

Yours,
Linus Walleij
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/