Hi Jonathan,
On 3/1/2025 8:55 AM, Jonathan Cameron wrote:
> On Wed, 26 Feb 2025 14:22:05 +0530
> Jishnu Prakash <jishnu.prakash@xxxxxxxxxxxxxxxx> wrote:
>
...
>>>> +void adc5_take_mutex_lock(struct device *dev, bool lock)
>>>> +{
>>>> + struct iio_dev *indio_dev = dev_get_drvdata(dev->parent);
>>>> + struct adc5_chip *adc = iio_priv(indio_dev);
>>>> +
>>>> + if (lock)
>>>> + mutex_lock(&adc->lock);
>>>> + else
>>>> + mutex_unlock(&adc->lock);
>>>> +}
>>>> +EXPORT_SYMBOL_NS_GPL(adc5_take_mutex_lock, "QCOM_SPMI_ADC5_GEN3");
>>>
>>> This is potentially going to make a mess for sparse. Might be better to split
>>> it in two so you can had __acquires and __releases markings.
>>>
>>> If you don't get any warnings with sparse then I guess we are fine.
>>>
>>
>> I had tried building with sparse in my local workspace and I did not get any errors in this file. Do you think I can keep this unchanged?
>> Also, would any kernel bots run sparse later on this patch, if it's not already done?
>
> Problems around this tend to turn up a bit late in build tests as requires
> particular combinations of features. Here you may not see problems because
> sparse can't see far enough to understand the locking.
>
> I would still split this into lock / unlock as that matches better
> with common syntax for locks. We can then add markings
> as necessary later.
>
OK, I can split this into separate lock and unlock functions.
And for markings, you mean I should add these:
__acquires(&adc->lock)
__releases(&adc->lock)
under the lock and unlock functions respectively?
Thanks,
Jishnu
>>>> +/*
>>>
>>> Looks like valid kernel doc, so /** and check it builds fine
>>> with the kernel-doc script.
>>>
