Re: Missing mali regulator in Amlogic dts

[Robin Murphy <robin.murphy@arm.com>]

On 2022-03-08 19:29, Martin Blumenstingl wrote:
> Hi Heiner,
> 
> On Tue, Mar 8, 2022 at 7:24 PM Heiner Kallweit <hkallweit1@gmail.com> wrote:
>>
>> I get following error on a  S905W-based system.
>> lima d00c0000.gpu: dev_pm_opp_set_regulators: no regulator (mali) found: -19
> This warning is harmless (see below)
> 
>> I see mali-supply in the DTS of few other vendors, but in no Amlogic DTS.
>> Is it missing, or are Amlogic SoCs different here?
> As far as I understand, on GXBB and GXL SoCs the VDDEE regulator is
> managed by the firmware (which is unfortunately closed-source).
> IIRC on the hardware side VDDEE is just another PWM regulator, just
> like with the 32-bit SoCs or with G12A (and later) SoCs.
> 
>> In the latter case supposedly drivers/gpu/drm/lima/lima_devfreq.c would
>> have to be changed to check for the optional regulator and not call
>> devm_pm_opp_set_regulators() if it's missing.
> I think I just copied the logic from panfrost
> (drivers/gpu/drm/panfrost/panfrost_devfreq.c) to lima.
> So GXM boards (using panfrost) will probably report a similar warning.
> That said, I don't know enough to judge whether your approach makes sense.

As a quality-of-life thing, it might make the most sense to refactor the 
API a bit so that it can offer a dedicated helper for setting optional 
regulators - possibly specifically a single optional regulator, since 
it's not necessarily clear how optionality should scale to a set of 
multiple items.

FWIW I do indeed see the same thing with panfrost, on an Arm Juno board 
where OPPs are completely abstracted behind the arm_scmi DVFS "clock".

Robin.

_______________________________________________
linux-amlogic mailing list
linux-amlogic@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-amlogic