Re: iio: iio_channel_get: ntc_thermistor isn't working since patch for of_iio_channel_get_by_name()

[Guenter Roeck <linux@roeck-us.net>]

On Fri, Aug 22, 2014 at 01:50:23PM +0200, Johannes Pointner wrote:
> Hello Guenter,
> 
> thank you for your quick answer.
> 
> 2014-08-21 15:25 GMT+02:00 Guenter Roeck <linux@roeck-us.net>:
> > On Thu, Aug 21, 2014 at 09:29:56AM +0200, Johannes Pointner wrote:
> >> Hello,
> >>
> >> I found out that since this patch https://lkml.org/lkml/2014/6/11/201,
> >> which fixes the return values in this function, the ntc_thermistor
> >> driver isn't working anymore for me.
> >> The problem is that before the driver asked for probe deferral because
> >> the adc driver isn't loaded at this point and now the probe fails in
> >> the function iio_channel_get_sys and returns -ENODEV.
> >> My question is: Who is in this case responsible to request a deferral?
> >> The driver or the subsystem?
> >>
> > Unless I am missing something, iio_get_channel() should request probe deferral.
> > There is no other way for the driver to know that the adc driver is not loaded
> > yet.
> 
> Yes, I missed that. In the function __of_iio_channel_get the subsystem
> returns a -EPROBE_DEFER.
> 
> >
> > The problem here may be that the devicetree entry for the ntc_thermistor driver
> > may not be correct. The above referenced patch fixes a bug in that area, and
> > devicetree data is now better validated.
> >
> But the problem is, that this statement:
> if (np && !of_get_property(np, "io-channel-ranges", NULL))
>             return NULL;
> overrides the -EPROBE_DEFER. Before the fix there was a break
> statement and the -EPROBE_DEFER was returned.
> 
I think the code further up should be

	if (!IS_ERR(chan) || PTR_ERR(chan) == -EPROBE_DEFER)
		break;

> Do I have made a mistake in the devicetree if I put the ntc
> declaration directly under the root node?
> 
No, I don't think so. I would suggest to submit a patch to fix the above.

Guenter