[PATCH 3/4] soc: rockchip: power-domain: Add regulator support
Robin Murphy
robin.murphy at arm.com
Wed Dec 22 04:54:06 PST 2021
On 2021-12-22 10:40, Sascha Hauer wrote:
> On Mon, Dec 20, 2021 at 12:53:58PM +0000, Mark Brown wrote:
>> On Mon, Dec 20, 2021 at 10:44:35AM +0100, Sascha Hauer wrote:
>>
>>> Well, all true and on one specific board the regulator is indeed not
>>> optional. However, on all other power domains that don't need a
>>> regulator and all other boards and all other SoCs this driver is used we
>>> now get:
>>
>> This seems unlikely to be board specific, if the chip requires power the
>> chip requires power. If there are power domains that don't take
>> external supplies then they shouldn't be requesting any regulators and
>> should be fixed.
>>
>>> [ 0.185588] rk-power-domain rk-power-domain.8: supply power not found, using dummy regulator
>>
>> It seems vanishingly unlikely that the SoC takes a single supply called
>> "power" shared by everything in the SoC but that is what the code
>> appears to be requesting - the power domains should be requesting the
>> supplies they actually use, and as ever the supplies should be named
>> such that someone looking at the schematic can hook them up. The
>> general recommendation is to use the names used in the datasheet.
>
> Ok. I'll change the patch in a way that only for the GPU power domain on
> rk3568 a supply is requested. That's the one power domain I know that a
> regulator is needed. I'm sure there are more, if not on rk3568 then
> probably on other SoCs the driver handles. Once we notice that other
> domains need a supply we'll have to add the supply name to driver data
> for that domain.
Certainly RK3399 (and I guess RK3288 too) suffers from the same
priority-inversion issue of being unaware that VD_GPU needs power before
PD_GPU can be successfully turned on to probe the GPU to claim and
enable the regulator (via "mali-supply" for DVFS purposes) that needed
to be on in the first place. Currently all the boards are bodging around
this with "regulator-always-on" (e.g. commit 06b2818678d9).
Thanks,
Robin.
More information about the Linux-rockchip
mailing list