Re: st_pressure: lps22: unused default data

[Jonathan Cameron <jic23@jic23.retrosnub.co.uk>]

On Thu, 08 Mar 2018 15:19:39 +0100
Michael Nosthoff <posted@heine.so> wrote:

> On 2018-03-08 14:54, Michael Nosthoff wrote:
> > Hi,  
> 
> Sorry for answering my own mail, but I got a bit further. See below
> 
> > I have an LPS22 sensor which i used sucessfully with Kernel 4.9. Upon
> > switching to 4.14 it was not loading properly anymore. This failure
> > occured:
> > 
> > [    3.827094] iio iio:device1: DRDY on pdata not valid.
> > [    3.832578] st-press-i2c: probe of 0-005c failed with error -22
> > 
> > So I started digging. It started working again when I add this entry
> > to my devicetree:
> > 
> > st,drdy-int-pin = <1>;
> > 
> > So it seems the default value seems to be not set in 4.14 but was in 
> > 4.9
> > 
> > I came to the conclusion that the problem happens in
> > drivers/iio/pressure/st_pressure_core.c in
> > 
> > st_press_common_probe:
> > 
> > 
> > struct st_sensors_platform_data *pdata =
> > 	(struct st_sensors_platform_data *)press_data->dev->platform_data;
> > ...
> > if (!pdata && press_data->sensor_settings->drdy_irq.addr)
> > 	pdata =	(struct st_sensors_platform_data *)&default_press_pdata;
> > 
> > err = st_sensors_init_sensor(indio_dev, 
> > press_data->dev->platform_data);
> > 
> > 
> > If pdata is not set (which is correct in my case) it gets assigned a
> > default value. But in the call to st_sensors_init_sensor it's still
> > using the data stored in press_data.
> > 
> > Okay, simple solution would be to pass pdata here as well. But I'm a
> > bit confused about the commit [1] which introduced this. pdata is not
> > used anymore in this function, so it seems completely useless to set it
> > there. And so I don't understand what the commit tries to achieve at 
> > all
> > besides never using the default data.
> >   
> 
> Somebody seems also to have noticed in the accel driver. See this commit 
> [2]
> and decided to remove the default initialization completely. Which makes 
> sense
> as this code was not doing anything.
> 
> I guess the what the intial commit was trying to achieve was to not 
> overwrite
> the pointer in dev->platform_data which i can imagine might be a problem 
> when
> using it as a module. But maybe forgot to alter the following usage of 
> the data.
> 
> So I would suggest to revert commit [2] and pass pdata to 
> st_sensors_init_sensor
> in both cases (st_pressure and st_accel).
> 
> Does this make sense?
> 
Yes,  I was clearly dozing when Colin's patch came through and failed
to notice that it should have been doing something with that pointer.
Should always have been passed the pdata copy.

Thanks,


Jonathan

> 
> 
> > 
> > [1]
> > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/drivers/iio/pressure/st_pressure_core.c?id=7383d44b84c94aaca4bf695a6bd8a69f2295ef1a  
> [2] 
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/drivers/iio/accel/st_accel_core.c?id=585ed27d06151f98e39238298f43ee261314ae74