Re: [PATCH v4 2/2] iio: vadc: Qualcomm SPMI PMIC voltage ADC driver

From: "Ivan T. Ivanov" <iivanov@mm-sol.com>
To: Hartmut Knaack <knaack.h@gmx.de>
Cc: Jonathan Cameron <jic23@kernel.org>,
	Lars-Peter Clausen <lars@metafoo.de>,
	Peter Meerwald <pmeerw@pmeerw.net>,
	Stanimir Varbanov <svarbanov@mm-sol.com>,
	Angelo Compagnucci <angelo.compagnucci@gmail.com>,
	Grant Likely <grant.likely@linaro.org>,
	linux-kernel@vger.kernel.org, linux-iio@vger.kernel.org,
	devicetree@vger.kernel.org, linux-arm-msm@vger.kernel.org
Subject: Re: [PATCH v4 2/2] iio: vadc: Qualcomm SPMI PMIC voltage ADC driver
Date: Tue, 11 Nov 2014 10:21:49 +0200	[thread overview]
Message-ID: <1415694109.22935.13.camel@mm-sol.com> (raw)
In-Reply-To: <54612A0D.7020308@gmx.de>

Hi Hartmut,

On Mon, 2014-11-10 at 22:11 +0100, Hartmut Knaack wrote:
> Ivan T. Ivanov schrieb am 03.11.2014 16:24:
> > From: Stanimir Varbanov <svarbanov@mm-sol.com>
> > 
> > The voltage ADC is peripheral of Qualcomm SPMI PMIC chips. It has
> > 15 bits resolution and register space inside PMIC accessible across
> > SPMI bus.
> > 
> > The vadc driver registers itself through IIO interface.
> Reviewing again, I got the feeling that due to the complexity of adc reads (writing to register 
> to start conversion, waiting a decent time for the conversion to complete, reading the result), 
> it would be beneficial to use a mutex in vadc_read_raw or its depending functions.

Hm, yes, but there is such a nice info_exist_lock :-) in core functions,
which in practice serve the same purpose.

> Also, I would want to double-check that the CPU in this type of chips is little endian.

Yes, it is.

> Besides that, a few minor issues inline.

> > +static int vadc_probe(struct platform_device *pdev)
> > +{
> > +       struct device_node *node = pdev->dev.of_node;
> > +       struct device *dev = &pdev->dev;
> > +       struct iio_dev *indio_dev;
> > +       struct vadc_priv *vadc;
> > +       struct regmap *regmap;
> > +       int ret, irq_eoc;
> > +       u32 res;
> Since vadc->base is u16, shouldn't res be u16 as well? Also think about naming it reg, as this is 
> the property name in DT.

Yes, but #address-cells for the node is 1, so I would like to keep it 32 bits.

> > +
> > +       regmap = dev_get_regmap(dev->parent, NULL);
> > +       if (!regmap)
> > +       return -ENODEV;
> > +
> > +       ret = of_property_read_u32(node, "reg", &res);
> For u16, there would be of_property_read_u16().
> > +       if (ret < 0)
> > +       return -ENODEV;
> Just return ret here?

I am usually trying to follow these recommendations[1]. In practice driver
core cares only for EPROBE_DEFER, ENODEV and ENXIO, while of_property_read_u32()
can return ENODATA and EOVERFLOW, which did't not make sense for the core.

Thank you,
Ivan

[1] http://www.gossamer-threads.com/lists/linux/kernel/1010603