Re: [PATCH v2 2/7] iio: adc: ti-adc128s052: Simplify using be16_to_cpu()

[Matti Vaittinen]

On 03/04/2025 00:04, David Lechner wrote:
> On 4/2/25 1:09 AM, Matti Vaittinen wrote:
> > The register data is 12-bit big-endian data. Use be16_to_cpu() to do
> > the conversion, and simple bitwise AND for masking to make it more
> > obvious.
> >
> > Signed-off-by: Matti Vaittinen <mazziesaccount@xxxxxxxxx>
> > ---
> > Revision history:
> > v1 => v2:
> >   - Fix commit msg to reflect the fact there was no bug
> >   - Drop Fixes tag
> >   - Use union for rx / tx buffer to avoid casting
> >   - Keep the shared message protected by the mutex
> > ---
> >   drivers/iio/adc/ti-adc128s052.c | 18 ++++++++++--------
> >   1 file changed, 10 insertions(+), 8 deletions(-)
> >
> > diff --git a/drivers/iio/adc/ti-adc128s052.c b/drivers/iio/adc/ti-adc128s052.c
> > index a456ea78462f..3e69a5fce010 100644
> > --- a/drivers/iio/adc/ti-adc128s052.c
> > +++ b/drivers/iio/adc/ti-adc128s052.c
> > @@ -28,32 +28,34 @@ struct adc128 {
> >   	struct regulator *reg;
> >   	struct mutex lock;
> >   
> > -	u8 buffer[2] __aligned(IIO_DMA_MINALIGN);
> > +	union {
> > +		__be16 rx_buffer;
> > +		u8 tx_buffer[2];
> > +	} __aligned(IIO_DMA_MINALIGN);
> >   };
> >   
> >   static int adc128_adc_conversion(struct adc128 *adc, u8 channel)
> >   {
> >   	int ret;
> > +	char *msg = &adc->tx_buffer[0];
> >   
> >   	mutex_lock(&adc->lock);
> >   
> > -	adc->buffer[0] = channel << 3;
> > -	adc->buffer[1] = 0;
> > +	msg[0] = channel << 3;
> > +	msg[1] = 0;
> >   
> > -	ret = spi_write(adc->spi, &adc->buffer, 2);
> > +	ret = spi_write(adc->spi, msg, sizeof(adc->tx_buffer));
> >   	if (ret < 0) {
> >   		mutex_unlock(&adc->lock);
> >   		return ret;
> >   	}
> >   
> > -	ret = spi_read(adc->spi, &adc->buffer, 2);
> > -
> > +	ret = spi_read(adc->spi, &adc->rx_buffer, 2);
> >   	mutex_unlock(&adc->lock);
> > -
> >   	if (ret < 0)
> >   		return ret;
> >   
> > -	return ((adc->buffer[0] << 8 | adc->buffer[1]) & 0xFFF);
> > +	return be16_to_cpu(adc->rx_buffer) & 0xFFF;
>
>
> The cast isn't exactly beautiful, but this would save a lot of
> lines of diff and a few lines of code by avoiding the need for
> the union and the local msg variable.
>
> 	return be16_to_cpup((__be16 *)adc->buffer) & 0xFFF;

Thanks again for the review David :)

I am unsure which way to go. I kind of like having the __be16 in the 
struct, as it immediately yells "data from device is big-endian". OTOH, 
I've never loved unions (and, it silences the above "yelling" quite a 
bit). I still think this might be the first time I really see a valid 
use-case for an union :) And, you're right this adds more lines, 
besides, the cast doesn't look that ugly to me. Yet, I originally had a 
cast probably as simple as this (and I also had the __be16 in the 
struct), and Jonathan suggested using union to avoid it...

At the end of the day, I suppose I am Okay with any of these 3 
approaches. Original cast, union or this cast you suggest. Jonathan, any 
preferences on your side?

> >   }
> >   
> >   static int adc128_read_raw(struct iio_dev *indio_dev,

Yours,
	-- Matti