Re: [PATCH v2] rtc: add Abracon ABx80x driver

[Philippe De Muyter <phdm@macqel.be>]

On Tue, Mar 03, 2015 at 09:50:32PM +0100, Alexandre Belloni wrote:
> On 03/03/2015 at 12:20:12 -0800, Andrew Morton wrote :
> > On Tue, 3 Mar 2015 02:11:16 +0100 Alexandre Belloni <alexandre.belloni@free-electrons.com> wrote:
> > 
> > > On 02/03/2015 at 15:53:37 -0800, Andrew Morton wrote :
> > > > On Sun,  1 Mar 2015 11:27:15 +0100 Alexandre Belloni <alexandre.belloni@free-electrons.com> wrote:
> > > > 
> > > > > Add support for the i2c RTC from Abracon.
> > > > 
> > > > What is the relationship between this patch and
> > > > http://ozlabs.org/~akpm/mmots/broken-out/rtc-add-rtc-abx805-a-driver-for-the-abracon-ab-1805-i2c-rtc.patch?
> > > > 
> > > 
> > > I'd say drop mine, I couldn't find the other one before writing it...
> > > 
> > > I'll try to build on Philippe's driver.
> > 
> > Which driver supports the most devices?  Your driver has
> > 
> > +static const struct i2c_device_id abx80x_id[] = {
> > +	{ "abx80x", ABX80X },
> > +	{ "ab0801", AB0801 },
> > +	{ "ab0802", AB0802 },
> > +	{ "ab0803", AB0803 },
> > +	{ "ab0804", AB0804 },
> > +	{ "ab0805", AB0805 },
> > +	{ "ab1801", AB1801 },
> > +	{ "ab1802", AB1802 },
> > +	{ "ab1803", AB1803 },
> > +	{ "ab1804", AB1804 },
> > +	{ "ab1805", AB1805 },
> > +	{ }
> > +};
> > 
> > And Philippe's has
> > 
> > +static struct i2c_device_id abx805_id[] = {
> > +	{ "abx805-rtc", 0 },

The only part my driver is actually tested with, is the 'AB1805', but see
below.

> > +	{ }
> > +};
> > 
> > 
> > And is the naming in Philippe's driver appropriate?  If it supports the
> > AB1801 (for example) then why is it described as an "abx805" driver?
> 
> The real naming is in the form ABx8yz. With:
> 
> x: 0 or 1, indicate the presence of the reset management
> y: 0 or 1: 0 is i2c, 1 is spi
> z: [1-5]: different amount of on chip SRAM. From what I understand, only
> ABx8y5 are actually recommended for new designs.

My driver is based on the datashet called 'AB18X5 Real-Time Clock with
Power Management Family', which calls the parts 'AB18X5 family' with X
being '0' for I2c parts and '1' for SPI parts.

As the part I have and the driver I wrote are I2C, not SPI, I fixed the
'X' as 0 in the name of the driver.

The datasheet lists only one part as being 'software and pin compatible'
with the 'AB1805': the 'AB0805', hence the 'x' in the name I choose : abx805.

> 
> They all share the same register set, apart from the reset management
> that is only available on AB18yz.
> 
> >From my point of view, but I'm obviously biased, I'd take my patch
> because it declares and supports all the available rtc and it also uses
> the i2c_smbus_xxx API that is more robust than the raw i2c_transfer.

I have no opinion on that.  The driver I started from uses 'raw i2c_transfer'.

> 
> I also take care of the 12/24 mode bit and the write RTC bit which is
> necessary to be able to write to the RTC.

Actually, my driver is used in production and works fine, because the
default(reset) value of the 12/24 mode bit is '24 hour mode' and the
default value of the 'write RTC bit' enables writing.  There is
no real need to play with them.

> 
> What I like in Philippe's driver is the info printed at probe time and
> the support for trickle charging. However, I wouldn't enable it
> unconditionally.

My hardware colleagues told me that the only way to enable the 'ultra low-power'
functionality is enabling the trickle charger.  And the 'ultra low-power'
functionality is the reason we choose that chip, so I would at least
keep that as the default behaviour.

Regards

Philippe

-- 
Philippe De Muyter +32 2 6101532 Macq SA rue de l'Aeronef 2 B-1140 Bruxelles