Re: [PATCH 1/7] iio: backend: add API for interface get

["Nuno Sá" <noname.nuno@gmail.com>]

On Sat, 2024-09-28 at 18:23 +0100, Jonathan Cameron wrote:
> On Thu, 26 Sep 2024 12:52:39 +0200
> Nuno Sá <noname.nuno@gmail.com> wrote:
> 
> > On Thu, 2024-09-26 at 10:40 +0200, David Lechner wrote:
> > > On Mon, Sep 23, 2024 at 12:15 PM Antoniu Miclaus
> > > <antoniu.miclaus@analog.com> wrote:  
> > > > 
> > > > Add backend support for obtaining the interface type used.
> > > > 
> > > > Signed-off-by: Antoniu Miclaus <antoniu.miclaus@analog.com>
> > > > ---
> > > >  drivers/iio/industrialio-backend.c | 24 ++++++++++++++++++++++++
> > > >  include/linux/iio/backend.h        | 10 ++++++++++
> > > >  2 files changed, 34 insertions(+)
> > > > 
> > > > diff --git a/drivers/iio/industrialio-backend.c
> > > > b/drivers/iio/industrialio-
> > > > backend.c
> > > > index efe05be284b6..53ab6bc86a50 100644
> > > > --- a/drivers/iio/industrialio-backend.c
> > > > +++ b/drivers/iio/industrialio-backend.c
> > > > @@ -449,6 +449,30 @@ ssize_t iio_backend_ext_info_set(struct iio_dev
> > > > *indio_dev,
> > > > uintptr_t private,
> > > >  }
> > > >  EXPORT_SYMBOL_NS_GPL(iio_backend_ext_info_set, IIO_BACKEND);
> > > > 
> > > > +/**
> > > > + * iio_backend_interface_type_get - get the interace type used.
> > > > + * @back: Backend device
> > > > + * @type: Interface type
> > > > + *
> > > > + * RETURNS:
> > > > + * 0 on success, negative error number on failure.
> > > > + */
> > > > +int iio_backend_interface_type_get(struct iio_backend *back,
> > > > +                                  enum iio_backend_interface_type
> > > > *type)
> > > > +{
> > > > +       int ret;
> > > > +
> > > > +       ret = iio_backend_op_call(back, interface_type_get, type);
> > > > +       if (ret)
> > > > +               return ret;
> > > > +
> > > > +       if (*type > IIO_BACKEND_INTERFACE_CMOS)
> Put a COUNT entry or similar on the end of the enum so this doesn't need
> updating for more types.
> 
> > > > +               return -EINVAL;
> > > > +
> > > > +       return 0;
> > > > +}
> > > > +EXPORT_SYMBOL_NS_GPL(iio_backend_interface_type_get, IIO_BACKEND);
> > > > +
> > > >  /**
> > > >   * iio_backend_extend_chan_spec - Extend an IIO channel
> > > >   * @indio_dev: IIO device
> > > > diff --git a/include/linux/iio/backend.h b/include/linux/iio/backend.h
> > > > index 8099759d7242..ba8ad30ac9ba 100644
> > > > --- a/include/linux/iio/backend.h
> > > > +++ b/include/linux/iio/backend.h
> > > > @@ -63,6 +63,11 @@ enum iio_backend_sample_trigger {
> > > >         IIO_BACKEND_SAMPLE_TRIGGER_MAX
> > > >  };
> > > > 
> > > > +enum iio_backend_interface_type {
> > > > +       IIO_BACKEND_INTERFACE_LVDS,
> > > > +       IIO_BACKEND_INTERFACE_CMOS
> 
> trailing comma.
> 
> This is going to get bigger!
> 
> > > > +};
> > > > +
> > > >  /**
> > > >   * struct iio_backend_ops - operations structure for an iio_backend
> > > >   * @enable: Enable backend.
> > > > @@ -81,6 +86,7 @@ enum iio_backend_sample_trigger {
> > > >   * @extend_chan_spec: Extend an IIO channel.
> > > >   * @ext_info_set: Extended info setter.
> > > >   * @ext_info_get: Extended info getter.
> > > > + * @interface_type_get: Interface type.
> > > >   **/
> > > >  struct iio_backend_ops {
> > > >         int (*enable)(struct iio_backend *back);
> > > > @@ -113,6 +119,8 @@ struct iio_backend_ops {
> > > >                             const char *buf, size_t len);
> > > >         int (*ext_info_get)(struct iio_backend *back, uintptr_t private,
> > > >                             const struct iio_chan_spec *chan, char
> > > > *buf);
> > > > +       int (*interface_type_get)(struct iio_backend *back,
> > > > +                                 enum iio_backend_interface_type
> > > > *type);
> > > >  };
> > > > 
> > > >  int iio_backend_chan_enable(struct iio_backend *back, unsigned int
> > > > chan);
> > > > @@ -142,6 +150,8 @@ ssize_t iio_backend_ext_info_set(struct iio_dev
> > > > *indio_dev,
> > > > uintptr_t private,
> > > >  ssize_t iio_backend_ext_info_get(struct iio_dev *indio_dev, uintptr_t
> > > > private,
> > > >                                  const struct iio_chan_spec *chan, char
> > > > *buf);
> > > > 
> > > > +int iio_backend_interface_type_get(struct iio_backend *back,
> > > > +                                  enum iio_backend_interface_type
> > > > *type);
> > > >  int iio_backend_extend_chan_spec(struct iio_dev *indio_dev,
> > > >                                  struct iio_backend *back,
> > > >                                  struct iio_chan_spec *chan);
> > > > --
> > > > 2.46.0
> > > >   
> > > 
> > > This seems very specific to the AD485x chips and the AXI ADC backend.
> > > Since it is describing how the chip is wired to the AXI DAC IP block,
> > > I would be tempted to use the devicetree for this info instead of
> > > adding a new backend function.  
> > 
> > Not sure If I'm following your point but I think this is the typical case
> > where the
> > chip (being it a DAC or ADC) supports both CMOS and LVDS interfaces.
> > Naturally you
> > only use one on your system and this is a synthesis parameter on the FPGA IP
> > core.
> > Therefore, it makes sense for the frontend to have way to ask for this
> > information to
> > the backend.
> > 
> > That said, we could also have a DT parameter but, ideally, we would then
> > need a way
> > to match the parameter with the backend otherwise we could have DT stating
> > LVDS and
> > the backend built with CMOS.
> 
> That would be a DTS bug that you should fix :)  For this to make sense you are
> relying on an FPGA that also has pins flexible enough to support LVDS and CMOS
> so it's only a firmware thing. Been a while since I last messed with FPGAs,
> but that seems unlikely to be true in general.
> 

Sure, but if this is something the FPGA can give us as part of it's register
map, it makes sense to me to have an interface like this...

> So far I'm with David on this, feels like something we shouldn't be
> discovering
> at runtime though maybe that's a convenience that we do want to enable.
> 

To be clear, I'm not against a DT parameter as it indeed describes how the HW is
being used (even though we could get it done solely with the interface_get())
and while I agree with you that having a mismatch in interface types would be a
DT bug, it's always better to be able to detect and catch it early on (and fail
early) then going against the wall until we realize the issue. So, I do see
value in an interface like this even if only to match and validate against a DT
parameter.

- Nuno Sá
>