[PATCH 1/3] ARM: OMAP2+: 32k-counter: Use hwmod lookup to check presence of 32k timer

[Ming Lei]

On Sun, Apr 1, 2012 at 3:10 AM, Shilimkar, Santosh
<santosh.shilimkar at ti.com> wrote:
> On Sat, Mar 31, 2012 at 2:09 PM, Ming Lei <tom.leiming at gmail.com> wrote:
>> On Sat, Mar 31, 2012 at 2:30 PM, Shilimkar, Santosh
>>> Since you need to recompile the kernel, you can very much tweak the
>>> clocksource to use GPTIMER with sysl clock. Support for that is still
>>> in place.
>>
>> With current kernel, running 'make menuconfig' can do it, but after applying
>> Hiremath's patch[1], I have to edit the source code manually to get it, so looks
>> this kind of tweaking is not friendly enough, :-(
>>
> It's not friendly but doable. But above can be supported I guess.
>
> Since you are talking about doing menuconfig
> and rebuilding kernel so what can be done is when one disable CPUIDLE
> and PM, one can disabled 32K source as well. And then 32K clocksource
> init should fail and fallback on gptimer clock source.

OK, it should work, but looks OMAP_32K_TIMER option has to be kept
for the usage, which may be a bit divergent to the purpose of the patch set.

So how about introducing a kernel parameter to decide if bypassing
32K source and using gptimer source directly, and let which depend
on PM?

>
> Vaibhav,
> Can you update your patch to support above. The patch which I
> did was doing exactly that but was not using hwmod. The problem with your
> patch is synctimer hwmod lookup on OMAP will never fail if the hwmod
> entry is supported.
>
> So you might better use failure case of omap_init_clocksource_32k() for
> fall back on gptimer source as I tried in my patch. That should support

You patch still doesn't work for the usage because you removed 32K option.


Thanks,
-- 
Ming Lei