[PATCH 2/2] ARM: dts: socfpga: Add a 3.3V fixed regulator node

[Doug Anderson]

Dinh,

On Fri, Oct 17, 2014 at 1:04 PM, Dinh Nguyen
<dinguyen at opensource.altera.com> wrote:
> On 10/17/2014 02:43 PM, Doug Anderson wrote:
>> Hi,
>>
>> On Fri, Oct 17, 2014 at 12:19 PM, Dinh Nguyen
>> <dinguyen at opensource.altera.com> wrote:
>>> On 10/17/2014 11:57 AM, Doug Anderson wrote:
>>>> Dinh,
>>>>
>>>> On Thu, Oct 16, 2014 at 2:03 PM,  <dinguyen at opensource.altera.com> wrote:
>>>>> From: Dinh Nguyen <dinguyen at opensource.altera.com>
>>>>>
>>>>> Without the 3.3V regulator node, the SDMMC driver will give these warnings:
>>>>>
>>>>> dw_mmc ff704000.dwmmc0: No vmmc regulator found
>>>>> dw_mmc ff704000.dwmmc0: No vqmmc regulator found
>>>>>
>>>>> This patch adds the regulator node, and points the SD/MMC to the regulator.
>>>>>
>>> [...]
>>>>>
>>>>> -               dwmmc0 at ff704000 {
>>>>> +               mmc0: dwmmc0 at ff704000 {
>>>>>                         num-slots = <1>;
>>>>>                         broken-cd;
>>>>>                         bus-width = <4>;
>>>>> @@ -41,4 +41,13 @@
>>>>>                         cpu1-start-addr = <0xffd080c4>;
>>>>>                 };
>>>>>         };
>>>>> +
>>>>> +       regulator_3_3v_hps: fixed_3_3v_hps_regulator at 0 {
>>>>
>>>> nit: no @0 since there is no "reg" (register) under this node.
>>>>
>>>> nit: usually people don't like "_" in node names.  ...I would probably
>>>> do this but I'm not an expert:
>>>>   regulator_3_3v_hps: hps-regulator {
>>>
>>> Ok..will fix up.
>>>
>>>>
>>>> This regulator also looks pretty bogus to me.  Is this really a
>>>> regulator that software has no control over?  It means you can't fully
>>>> reset a card but I guess that's OK.
>>>
>>> Yes, this is a generic 3.3V regulator that is used for the 3.3V power
>>> rail for the SOC, and any IP the needs 3.3V.
>>>
>>> Schematics are here:
>>> http://www.rocketboards.org/pub/Documentation/AlteraSoCDevelopmentBoard/c5_soc_devkit_c.pdf?t=1365712679
>>>
>>> http://www.rocketboards.org/pub/Documentation/ArrowSoCKitEvaluationBoard/SoCKit_RevC.pdf
>>>
>>> And right, I don't see anyway to be able to reset the card.
>>>
>>>>
>>>> I'd also expect this regulator to be defined in the same dts / dtsi
>>>> file that it's used in.  Your current patch says "there's a generic
>>>> 3.3V regulator on all boards of the socfpga_arria5 class even if they
>>>> don't use MMC, but the "socfpga_arria5_socdk" uses it for MMC.  Is
>>>> that really true?
>>>>
>>>
>>> yeah...My original patch had the regulator placed in it's appropriate
>>> dts board file, but I saw that as a bit of duplication. I figured that
>>> most people are just copying Altera's devkit schematic, that if a new
>>> board file shows up that has a different controllable regulator, then
>>> the board file can have the new regulator node. Meanwhile, this patch
>>> saves adding an additional regulator node for the existing board files.
>>>
>>> The 3.3V is not just used by MMC, but the for all 3.3V peripherals.
>>
>> The advantage of putting it in each board file is you could name it
>> based on the schematics.  I see SoCKit_RevC it is named "VCC3P3"
>>
>
> That's fine. I can edit the patch so that each regulator node goes into
> it's respective board dts file.
>
>> It also allows you to specify the 3.3 supply as the output of a real
>> regulator driver.  In s5_soc_devkit it looks like this comes from
>> LTC2978.  That's got i2c connections.  What happens when you add a
>> driver for that?
>>
>
> Do you mean the c5_soc_devkit? If so, then the 3.3V is coming from the
> LTC3885 regulator. The LTC2978 is a regulator for the the 3.3V_HPS or
> 3.3V_REG_HPS which is used by others.

Oops, right.  c5_soc_devkit_c.

I guess I got confused since your regulator name in the dts was "HPS
3.3V".  When I saw "3.3V_PM_HPS" I assumed it was the same.

On <http://www.rocketboards.org/pub/Documentation/AlteraSoCDevelopmentBoard/c5_soc_devkit_c.pdf?t=1365712679>
I don't see a 3885.  Oh, 3855.  I see.


> So I should change the regulator name from 3.3V_HPS to just 3.3V for the
> c5_soc_devkit.

Yes, that's my opinion.  Matching the schematics can be a huge benefit
and can avoid confusion!  :)

-Doug