On 29.06.2023 10:43, Konrad Dybcio wrote:
> On 28.06.2023 21:28, Mark Brown wrote:
>> On Wed, Jun 28, 2023 at 06:29:46PM +0200, Konrad Dybcio wrote:
>>
>>> +// SPDX-License-Identifier: GPL-2.0-only
>>> +/*
>>> + * Copyright (c) 2017, 2019-2020, The Linux Foundation. All rights reserved.
>>> + * Copyright (c) 2023, Linaro Limited
>>> + */
>>
>> Please use a C++ comment for the whole thing for consistency.
> Oh that's new!
>
>>
>>> +static int qcom_sdm845_refgen_enable(struct regulator_dev *rdev)
>>> +{
>>> + struct qcom_refgen *vreg = rdev_get_drvdata(rdev);
>>> +
>>> + regmap_update_bits(vreg->base, REFGEN_REG_BG_CTRL,
>>> + REFGEN_BG_CTRL_MASK, REFGEN_BG_CTRL_ENABLE);
>>> + regmap_write(vreg->base, REFGEN_REG_BIAS_EN, REFGEN_BIAS_EN_ENABLE);
>>
>> For the enable and disable operations we use a mix of _update_bits() and
>> absolute writes with no FIELD_PREP()...
> This absolute write was accidentally fine as the mask began at bit0...
>
>>
>>> +static int qcom_sdm845_refgen_is_enabled(struct regulator_dev *rdev)
>>> +{
>>> + struct qcom_refgen *vreg = rdev_get_drvdata(rdev);
>>> + u32 val;
>>> +
>>> + regmap_read(vreg->base, REFGEN_REG_BG_CTRL, &val);
>>> + if (FIELD_GET(REFGEN_BG_CTRL_MASK, val) != REFGEN_BG_CTRL_ENABLE)
>>> + return 0;
>>> +
>>> + regmap_read(vreg->base, REFGEN_REG_BIAS_EN, &val);
>>> + if (FIELD_GET(REFGEN_BIAS_EN_MASK, val) != REFGEN_BIAS_EN_ENABLE)
>>> + return 0;
>>
>> ...but when we read back the status we use FIELD_GET(). This looks like
>> a bug, and given that one of the fields starts at bit 1 it presumably is
>> one - FIELD_GET() will do shifting.
> ...but a 2-bit-wide field will never equal 6.
> Looks like I put unshifted values in the defines for REFGEN_BG_CTRL..
>
> Thanks for spotting that!
Even worse, I noticed I've been feeding a raw address into regmap
functions.. :)
Konrad
>
>>
>>> +static int qcom_sm8250_refgen_enable(struct regulator_dev *rdev)
>>> +{
>>> + struct qcom_refgen *vreg = rdev_get_drvdata(rdev);
>>> +
>>> + regmap_update_bits(vreg->base, REFGEN_REG_PWRDWN_CTRL5,
>>> + REFGEN_PWRDWN_CTRL5_MASK, REFGEN_PWRDWN_CTRL5_ENABLE);
>>
>> This is a single bit in a single register so could use the standard
>> helpers rather than open coding, the sdm845 does need custom operations
>> due to having two fields to manage.
> Forgot that's a thing!
>
> Konrad
