[PATCH v2 12/19] ARM: at91/rtc-at91sam9: each SoC can select the RTT device to use

[Jean-Christophe PLAGNIOL-VILLARD]

On 16:47 Wed 22 Feb     , Arnd Bergmann wrote:
> On Wednesday 22 February 2012, Jean-Christophe PLAGNIOL-VILLARD wrote:
> > On 14:50 Wed 22 Feb     , Arnd Bergmann wrote:
> > > On Wednesday 22 February 2012, Nicolas Ferre wrote:
> > > > From: Jean-Christophe PLAGNIOL-VILLARD <plagnioj at jcrosoft.com>
> > > > 
> > > > For the RTT as RTC driver rtc-at91sam9, the platform_device structure
> > > > is filled during SoC initialization. This will allow to convert this
> > > > RTC driver as a standard platform driver.
> > > 
> > > Can you make this more elaborate? I don't see from this or the code why
> > > you don't just always register the RTT as "rtc-at91sam9". There seems to
> > > be no driver for the "at91_rtt" in tree, so I don't know if there is
> > > an out of tree driver binding to it.
> > > 
> > > Would it be possible to remove the compile time #if and the resetting
> > > of the device name if both drivers bind to the rtc name and the other
> > > rtt driver binds to both names?
> > because I see a quite some people use the rtt for something else
> > 
> > so I don't want to change on existing kernel
> 
> But if you change the rtc driver to always bind to the existing
> "at91_rtt" platform_device and fail the probe() function for the
> case that you are on at91sam9263 probing the non-RTC device,
> you should get the exact same result without any extra code
> in the per-soc files.
agreed but as on at91sam9263 you have 2 RTTs so you can have 2 drivers at the
same time that use the RTT. So we can not do this except if we set the second
RTT to an other device name.
> 
> I guess it also makes sense to specify a device tree property
> that lets you detect whether the RTT is used as RTC or something
> else.
>
agreed for DT I think of something like this with a phandle maybe

but today not all the soc are DT and will not be switch to DT at first

Best Regards,
J.