Re: [PATCH v4 02/11] iio: adc: Add basic support for AD4170

[Jonathan Cameron]

On Tue, 3 Jun 2025 16:43:42 +0300
Andy Shevchenko <andy@xxxxxxxxxx> wrote:

> On Tue, Jun 03, 2025 at 09:02:56AM -0300, Marcelo Schmitt wrote:
> > On 06/03, Andy Shevchenko wrote:  
> > > On Mon, Jun 02, 2025 at 01:54:25PM -0300, Marcelo Schmitt wrote:  
> 
> ...
> 
> > > > > > +static bool ad4170_setup_eq(struct ad4170_setup *a, struct ad4170_setup *b)
> > > > > > +{
> > > > > > +	/*
> > > > > > +	 * The use of static_assert() here is to make sure that the comparison
> > > > > > +	 * is adapted whenever struct ad4170_setup is changed.
> > > > > > +	 */  
> > Does the reason given in the comment justify the use of static_assert?  
> 
> Should I repeat myself? It makes a little sense when no memcmp() is involved.

The intent I think is to reduce the chance of a field being added without
this match function being updated.   Not a very strong test but maybe
better than nothing...

Not sure how memcmp() is relevant.

> 
> > > > > > +	static_assert(sizeof(*a) ==
> > > > > > +		      sizeof(struct {
> > > > > > +				     u16 misc;
> > > > > > +				     u16 afe;
> > > > > > +				     u16 filter;
> > > > > > +				     u16 filter_fs;
> > > > > > +				     u32 offset;
> > > > > > +				     u32 gain;
> > > > > > +			     }));  
> > > > > 
> > > > > I think it doesn't make much sense unless one uses memcpy().  
> > > > 
> > > > memcpy() is used to update the setups after reg write succeeds.
> > > > Also, previously, memcmp() was used to compare setups.
> > > > Since struct ad4170_setup has only unsigned integers (no floating point fields
> > > > like ad7124 had [1]), ad4170 works properly when comparing setups with memcmp().
> > > > Though, it was asked to do explicit field matching on previous reviews [2] so
> > > > that's how it had been since then. Well, both ways work for ad4170. We can
> > > > compare setup with memcmp(), or do the comparison field by field. I don't mind
> > > > changing it again if requested. I guess we only need to reach an agreement about
> > > > what to go with.  
> > > 
> > > The question was "why do you need the static_assert() now?"  
> > 
> > To ensure that the comparison function gets updated if struct ad4170_setup is
> > ever modified? This intends to be similar to what was implemented in ad7124
> > driver as the chips have similar channel configuration mechanisms. We also
> > have ad7173 and ad4130 using static_assert for analogous purpose. There was
> > also a comment about static_assert above.  
> 
> Does this won;t work if you changes field types? (Assuming only integers to
> integers) I believe it doesn't affect the field-by-field comparison.
> 
> The other drivers may have different approach, have you studied them? Do they
> use memcmp()
> 
> > > > [1]: https://lore.kernel.org/all/20250303114659.1672695-13-u.kleine-koenig@xxxxxxxxxxxx/
> > > > [2]: https://lore.kernel.org/linux-iio/20250504192117.5e19f44b@jic23-huawei/
> > > >   
> > > > > > +	if (a->misc != b->misc ||
> > > > > > +	    a->afe != b->afe ||
> > > > > > +	    a->filter != b->filter ||
> > > > > > +	    a->filter_fs != b->filter_fs ||
> > > > > > +	    a->offset != b->offset ||
> > > > > > +	    a->gain != b->gain)
> > > > > > +		return false;
> > > > > > +
> > > > > > +	return true;
> > > > > > +}  
>