Re: [PATCH 6/6] arm64: dts: imx8mm-kontron: Add support for reading SD_VSEL signal

[Frieder Schrempf]

Hi Marco,

On 14.02.23 09:10, Marco Felsch wrote:
> On 23-02-13, Marek Vasut wrote:
>> On 2/13/23 20:56, Marco Felsch wrote:
>>> Hi Marek, Frieder,
>>
>> Hi,
>>
>>> On 23-02-13, Marek Vasut wrote:
>>>> On 2/13/23 17:15, Marco Felsch wrote:
>>>>
>>>> [...]
>>>>
>>>>>> @@ -347,7 +347,7 @@ MX8MM_IOMUXC_SD2_DATA1_USDHC2_DATA1		0x1d6
>>>>>>    			MX8MM_IOMUXC_SD2_DATA2_USDHC2_DATA2		0x1d6
>>>>>>    			MX8MM_IOMUXC_SD2_DATA3_USDHC2_DATA3		0x1d6
>>>>>>    			MX8MM_IOMUXC_SD2_CD_B_GPIO2_IO12		0x019
>>>>>> -			MX8MM_IOMUXC_GPIO1_IO04_USDHC2_VSELECT		0x1d0
>>>>>> +			MX8MM_IOMUXC_GPIO1_IO04_USDHC2_VSELECT		0x400001d0
>>>>>
>>>>> The VSELECT pin should be driven by the (u)sdhc core...
>>>>>
>>>>>>    		>;
>>>>>>    	};
>>>>>>    };
>>>>>> diff --git a/arch/arm64/boot/dts/freescale/imx8mm-kontron-osm-s.dtsi b/arch/arm64/boot/dts/freescale/imx8mm-kontron-osm-s.dtsi
>>>>>> index 5172883717d1..90daaf54e704 100644
>>>>>> --- a/arch/arm64/boot/dts/freescale/imx8mm-kontron-osm-s.dtsi
>>>>>> +++ b/arch/arm64/boot/dts/freescale/imx8mm-kontron-osm-s.dtsi
>>>>>> @@ -196,6 +196,7 @@ reg_nvcc_sd: LDO5 {
>>>>>>    				regulator-name = "NVCC_SD (LDO5)";
>>>>>>    				regulator-min-microvolt = <1800000>;
>>>>>>    				regulator-max-microvolt = <3300000>;
>>>>>> +				sd-vsel-gpios = <&gpio1 4 GPIO_ACTIVE_HIGH>;
>>>>>
>>>>> and by using the sd-vsel-gpios property the IOMUXC_GPIO1_IO04 have to be
>>>>> muxed as GPIO, which is not the case. So I think that u-boot have a bug
>>>>> within the (u)sdhc core.
>>>>
>>>> The trick here is that the VSELECT is operated by the usdhc block as a
>>>> function pin, but the PMIC driver can read the current state of the VSELECT
>>>> pin by reading out the GPIO block SR register. Since the IOMUX SION bit is
>>>> set on the VSELECT pin, the state of the pin is reflected in the GPIO block
>>>> SR register even if the pin is muxed as function pin.
>>>>
>>>
>>> Thanks for this explanation :) Why does the regulator driver need to
>>> know the current state of this pin?
>>
>> Because that regulator has an input pin which selects between two states of
>> that regulator, L and H, and whatever L or H is depends on what is
>> configured into the regulator via I2C. To correctly report the state of the
>> regulator, you have to know the state of that input (selector) pin.
>>
>>> Since the voltage switching requires
>>> some cmd's before the actual voltage level switch. So this must be
>>> handled within the core.
>>>
>>> Also after checking the driver, adding the sd-vsel-gpios will request
>>> the specified gpio as output-high.
>>
>> The GPIO would have to be requested as input, obviously.
> 
> But that isn't the case. According the driver comment they just want to
> make sure that this GPIO is high to ensure that the correct regulator
> config registers are used.

It seems like you look at the wrong code. We previously had the
sd-vsel-gpios used as you describe, as an output set to a fixed high
level to make sure that the SD_VSEL state matches the driver using the H
register. This code is reverted in patch 3 and patch 5 implements the
sd-vsel-gpios as input as described by Marek. See:

https://lore.kernel.org/lkml/20230213155833.1644366-4-frieder@xxxxxxx/
https://lore.kernel.org/lkml/20230213155833.1644366-6-frieder@xxxxxxx/

Sorry, my scripts didn't cc everyone for all patches, which is probably
why you missed these changes.


> 
>>> Out of curiosity, what's the bug you
>>> triggering within U-Boot?
>>
>> AFAICT the readback of the initial state of the regulator (see paragraph
>> above), which affects Linux all the same.
> 
> According the binding the driver should check this and apply the value
> to the corresponding L/H register but unfortunately this isn't the case
> yet. Does U-Boot handle this correctly?

See above.

Thanks
Frieder