Re: [PATCH V3 4/4] thermal: qcom: add support for PMIC5 Gen2 ADCTM

[Jishnu Prakash <quic_jprakash@quicinc.com>]

Hi Dmitry,

On 11/29/2021 8:02 AM, Dmitry Baryshkov wrote:
> On 23/11/2021 08:57, Jishnu Prakash wrote:
>> Add support for PMIC5 Gen2 ADC_TM, used on PMIC7 chips. It is a
>> close counterpart of PMIC7 ADC and has the same functionality as
>> PMIC5 ADC_TM, for threshold monitoring and interrupt generation.
>> It is present on PMK8350 alone, like PMIC7 ADC and can be used
>> to monitor up to 8 ADC channels, from any of the PMIC7 PMICs
>> having ADC on a target, through PBS(Programmable Boot Sequence).
>>
>> Signed-off-by: Jishnu Prakash <quic_jprakash@quicinc.com>
>> ---
>>   drivers/thermal/qcom/qcom-spmi-adc-tm5.c | 375 
>> ++++++++++++++++++++++++++++++-
>>   1 file changed, 372 insertions(+), 3 deletions(-)
>>
>> diff --git a/drivers/thermal/qcom/qcom-spmi-adc-tm5.c 
>> b/drivers/thermal/qcom/qcom-spmi-adc-tm5.c
>> index fc8cd45..a7b33a8 100644

>>   static int adc_tm5_set_trips(void *data, int low, int high)
>>   {
>>       struct adc_tm5_channel *channel = data;
>> @@ -422,6 +739,11 @@ static int adc_tm5_init(struct adc_tm5_chip *chip)
>>           }
>>       }
>>   +    mutex_init(&chip->adc_mutex_lock);
>
> Minor issue. This way, the mutex is left uninitialized for ADC_TM5_HC 
> devices. I'd move the mutex_init() call to the _probe function itself.


The mutex is needed only for Gen2 ADC_TM devices, I have mentioned this 
in the adc_tm5_chip struct description. I'll keep it in the Gen2 ADC_TM 
init function.


>
>> +
>> +    if (chip->data->gen == ADC_TM5_GEN2)
>> +        return ret;
>> +
>
> Please do not do this. Create a separate adc_tm5_gen2_init function.
> Add init() callback to adc_tm5_data structure.


Will make this change in next post.


>
>>       buf[0] = chip->decimation;
>>       buf[1] = chip->avg_samples | ADC_TM5_FAST_AVG_EN;
>>       buf[2] = ADC_TM5_TIMER1;
>> @@ -442,7 +764,7 @@ static int adc_tm5_get_dt_channel_data(struct 
>> adc_tm5_chip *adc_tm,
>>                          struct device_node *node)
>>   {
>>       const char *name = node->name;
>> -    u32 chan, value, varr[2];
>> +    u32 chan, value, adc_channel, varr[2];
>>       int ret;
>>       struct device *dev = adc_tm->dev;
>>       struct of_phandle_args args;
>> @@ -472,7 +794,11 @@ static int adc_tm5_get_dt_channel_data(struct 
>> adc_tm5_chip *adc_tm,
>>       }
>>       of_node_put(args.np);
>>   -    if (args.args_count != 1 || args.args[0] >= ADC5_MAX_CHANNEL) {
>> +    adc_channel = args.args[0];
>> +    if (adc_tm->data->gen == ADC_TM5_GEN2)
>> +        adc_channel &= 0xff;
>> +
>> +    if (args.args_count != 1 || adc_channel >= ADC5_MAX_CHANNEL) {
>
> Here you read the data (args.args[0]) before checking that it is 
> actually available (args.args_count is not zero). Please correct the 
> sequence.


Will correct this in next post.


>
>>           dev_err(dev, "%s: invalid ADC channel number %d\n", name, 
>> chan);
>>           return -EINVAL;
>>       }
>> @@ -518,6 +844,32 @@ static int adc_tm5_get_dt_channel_data(struct 
>> adc_tm5_chip *adc_tm,
>>       else
>>           channel->cal_method = ADC_TM5_ABSOLUTE_CAL;
>>   +    if (adc_tm->data->gen == ADC_TM5_GEN2) {
>> +        ret = of_property_read_u32(node, "qcom,decimation", &value);
>> +        if (!ret) {
>> +            ret = qcom_adc5_decimation_from_dt(value, 
>> adc_tm->data->decimation);
>> +            if (ret < 0) {
>> +                dev_err(dev, "invalid decimation %d\n", value);
>> +                return ret;

Thanks,

Jishnu