Re: [PATCH v2 4/6] iio: amplifiers: ad8366: add device tree support

[Andy Shevchenko]

On Wed, Jan 28, 2026 at 09:55:16AM +0000, Rodrigo Alencar wrote:
> On 26/01/27 11:21PM, Andy Shevchenko wrote:
> > On Mon, Jan 26, 2026 at 01:51:05PM +0000, Rodrigo Alencar via B4 Relay wrote:

...

> > > +static size_t ad8366_pack_code(struct ad8366_state *st)
> > > +{
> > > +	u8 ch_a = bitrev8(st->ch[0] & 0x3F);
> > > +	u8 ch_b = bitrev8(st->ch[1] & 0x3F);
> > 
> > GENMASK() in both cases? But I don't see why ch_a needs this at all,
> > isn't the 2 LSBs are not used anyway?
> 
> Yes, I can adjust with:
> 
> u8 ch_a = bitrev8(st->ch[0]) >> 2;
> u8 ch_b = bitrev8(st->ch[1]) >> 2;
> 
> st->data[0] = ch_b >> 2;
> st->data[1] = (ch_b << 6) | ch_a;
> 
> so no need for masking both.

This is better, but let's think a bit more. The data we put seems to be
__be12 (yeah, we don't have the exact type for that) and can be put slightly
differently.

So, something like

	put_unaligned_be16((ch_b << 6) | ch_a, &st->deta[0]);

should be better, no? (Note, you would need linux/unaligned.h).

> > Also missed header inclusion for this? And also perhaps sorting headers first
> > to see what's there and what needs to be updated (ideally another patch to move
> > to IWYU principle).
> 
> linux/bitrev.h is there, but indeed header includes are not sorted.
> I will create a separate patch for that.

Ah, good!

> > > +	st->data[0] = ch_b >> 4;
> > > +	st->data[1] = (ch_b << 4) | (ch_a >> 2);
> > > +	return 2;
> > > +}

-- 
With Best Regards,
Andy Shevchenko