Re: [PATCH] iio: dac: adi-axi-dac: drop io_mode check

[Nuno Sá]

On Mon, 2025-02-10 at 19:13 +0000, Jonathan Cameron wrote:
> On Mon, 10 Feb 2025 10:05:47 +0000
> Nuno Sá <noname.nuno@xxxxxxxxx> wrote:
> 
> > On Sat, 2025-02-08 at 15:45 +0000, Jonathan Cameron wrote:
> > > On Thu, 06 Feb 2025 09:36:14 +0100
> > > Angelo Dureghello <adureghello@xxxxxxxxxxxx> wrote:
> > >   
> > > > From: Angelo Dureghello <adureghello@xxxxxxxxxxxx>
> > > > 
> > > > Drop mode check, producing the following robot test warning:
> > > > 
> > > > smatch warnings:
> > > > drivers/iio/dac/adi-axi-dac.c:731 axi_dac_bus_set_io_mode()
> > > >   warn: always true condition '(mode >= 0) => (0-u32max >= 0)'
> > > > 
> > > > The range check results not useful since these are the only
> > > > plausible modes for enum ad3552r_io_mode.
> > > > 
> > > > Fixes: 493122c53af1 ("iio: dac: adi-axi-dac: add bus mode setup")
> > > > Signed-off-by: Angelo Dureghello <adureghello@xxxxxxxxxxxx>  
> > > Ah. I missed this.  Anyhow made the same change directly so all is well
> > > than ends well!
> > >   
> > 
> > Hi Angelo, Jonathan,
> > 
> > I wanted to reply to this one when I saw it but I haven't done right away
> > and
> > then totally forgot. Sorry about that!
> > 
> > I don't really agree with the "fix" in this patch. AFAIU, smatch is
> > complaining
> > since the enum is apparently defaulting to an unsigned type which means
> > doing
> > the >= 0 check is useless. But we should keep the upper bound...
> 
> Why? It's an enum so unless we are messing around with deliberate casts the
> compiler should always be able to spot this. The check may be needed on a
> future

I do not think the compiler will catch this:

diff --git a/drivers/iio/dac/ad3552r-hs.c b/drivers/iio/dac/ad3552r-hs.c
index c1dae58c1975..5234dd5e227d 100644
--- a/drivers/iio/dac/ad3552r-hs.c
+++ b/drivers/iio/dac/ad3552r-hs.c
@@ -293,7 +293,7 @@ static int ad3552r_hs_buffer_postenable(struct iio_dev
*indio_dev)
         * Back bus to simple SPI, this must be executed together with above
         * target mode unwind, and can be done only after it.
         */
-       st->data->bus_set_io_mode(st->back, AD3552R_IO_MODE_SPI);
+       st->data->bus_set_io_mode(st->back, -1);

A W=1 build (clang) did not complained at all... Maybe tools like smatch will.

> date if we add more types to that enum.
> 
> So I agree the check wasn't terrible and perhaps acted as hardening but it
> isn't strictly speaking doing anything today.
> 

It's not a very super important check, I agree... and being an enum will be
easier to spot a raw value being passed during a review but since we already had
the check, I don't see why we should remove it completely and not keep the upper
bound.

- Nuno Sá