Re: [PATCHv1] rtc: bcm-iproc: Add support for Broadcom iproc rtc

[Arnd Bergmann <arnd@arndb.de>]

On Thursday 12 February 2015 14:17:41 Arun Ramamurthy wrote:
> Hi Arnd
> 
> My apologies for the late reply, I was moved to other work items. I 
> wanted to get more clarification on the syscon issue so that I can 
> submit the next patch set. If I understand correctly, you would like
> me to move the CRMU logic to a new driver under mfd/ and use the syscon
> api calls in my rtc driver? Thanks

It depends a lot on what's in there, I can best advise you if you
have some form of register list.

A common approach would be to not have a driver for the crmu at all,
but just mark it as syscon, and have the other drivers either reference
the syscon node through a phandle, or create them as childrem of
the syscon node. The latter case makes most sense if all uses of
the crmu have no other MMIO registers.


> On 14-12-17 06:31 AM, Arnd Bergmann wrote:
> > On Tuesday 16 December 2014 13:54:04 Arun Ramamurthy wrote:
> >> On 14-12-16 12:27 PM, Ray Jui wrote:
> >>> On 12/16/2014 12:19 PM, Arnd Bergmann wrote:
> >>>>
> >>>> It sounds like CRMU is some other unit aside from the RTC. Could this
> >>>> be something like a generic system controller? I think it should
> >>>> either have its own driver or use the syscon logic if that is what
> >>>> this is.
> >>>>
> >>> Giving that CRMU has scattered, miscellaneous control logic for multiple
> >>> different peripherals, it probably makes more sense to use the syscon
> >>> logic here.
> >>>
> >> Arnd, thanks for the feedback. If I was to write a separate driver for
> >> the CRMU, I would have to export certain functions and create an api
> >> that only this RTC driver would use. I am not sure that is efficient or
> >> required. What is your opinion?
> >> Would it be better if I use the syson api in my current driver and move
> >> the CRMU registers to separate syscon device tree entry?
> >>
> >
> > This is something that's normally up to the platform maintainers, depending
> > on what works best for a given SoC. If you have a control block that
> > wants to export the same high-level API for multiple drivers, that's
> > fine, but if literally every register does something different, a syscon
> > driver works best.
> >
> > It's also possible that some of the functions of the CRMU already have
> > abstractions, like system-reset, device-reset, regulator or clock support.
> > In that case, you can still use syscon but have the more other drivers
> > use that for accessing the registers.
> >
> > 	Arnd
> >
> --
> To unsubscribe from this list: send the line "unsubscribe devicetree" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html