[PATCH 1/1] ARM: dts: Use more descriptive names for Exynos5420 PDs

[Sylwester Nawrocki]

On 10/02/15 14:14, Krzysztof Kozlowski wrote:
> On wto, 2015-02-10 at 14:00 +0100, Sylwester Nawrocki wrote:
>> > On 10/02/15 13:46, Javier Martinez Canillas wrote:
>>>>>> > >>>> This debugfs code iterates over list of generic_pm_domains (gpd_list). I
>>>>>>>>> > >>>> >> > cannot find function for translating from genpd to its platform device
>>>>>>>>> > >>>> >> > so only genpd->name can be printed.
>>>>>>> > >>> >> 
>>>>>>> > >>> >>     Then why power domains aren't just named with the platform device names?
>>>>> > >> > 
>>>>> > >> > Right, the mach-exynos/pm_domains.c set the name equal to OF node name.
>>>>> > >> > I'll send a patch extending the name.
>>>>> > >> >
>>> > > IIRC the OF core uses the device node unit address and node name to create
>>> > > the platform device names so you will have something like 10044000.power-domain.
>>> > > 
>>> > > Same if using the node full_name since it will /power-domain at 10044000. In both
>>> > > cases the DTS should have to be checked to know which power domain really is
>>> > > unless someone knows by heart the power domains addresses.
> For the kernel developer that would be descriptive enough to find the
> real domain but... as you said each time one would have to grep through
> manual or DTS which is slower. However for end-user that still won't be
> descriptive enough.
> 
>>> > > 
>>> > > But if using generic names for the power domains as suggested by ePAPR is so
>>> > > important then we should change all the other Exynos DTS files which don't do.
>> > 
>> > Perhaps we could assign OF aliases to the power domain device nodes in DT
>> > and then in the power domains driver map those aliases to more descriptive
>> > names when creating the power domains?
>
> That would required additional alias in DT but it could be the most
> descriptive for a user.

Yes, we could fall back to of_node->full_name if alias is not present in DT.

-- 
Regards,
Sylwester