Re: [REPOST PATCH 0/2] Table lookup for mux clock type

[Stephen Warren <swarren@wwwdotorg.org>]

On 03/05/2013 03:48 PM, Stephen Warren wrote:
> On 02/22/2013 08:02 AM, Peter De Schrijver wrote:
>> This patchset adds a table lookup feature to the mux clock type. This will
>> be used by the forthcoming Tegra114 clock implementation. Also instead of
>> a fixed field width, a mask is used. This is because Tegra114 has some muxes
>> where the parent selector is spread over several bitfields. The second patch
>> adapts the tegra periph clock implementation which uses struct clk_mux
>> directly.
>>
>> --
>>
>> Mike,
>>
>> This patch is a dependency for the Tegra114 CCF implementation. Could you
>> review it and merge on a topic branch so Stephen can pull it in to verify
>> the integration?
> 
> Peter, does this repost you resolve the issue I pointed out at:
> http://www.spinics.net/lists/arm-kernel/msg220873.html

Indeed, I just tested applying patch 1/2 and the build is broken without
patch 2/2.

Can you either rework the patches to maintain bisectability, or squash
the two together?

> 
> To quote:
> 
>>>> Just a quick note on patch dependencies here:
>>>>
>>>> Patch 1/2 can presumably be taken through the clk tree whenever Mike is
>>>> OK with it.
>>>>
>>>> Patch 2/2 depends on patches in the Tegra tree for 3.9. Since patch 2/2
>>>> is useful mostly for the Tegra114 clock driver, and I don't imagine that
>>>> will get posted/merged in time for 3.9, it's probably easiest to just
>>>> take patch 2/2 for 3.10 along with the Tegra114 clock driver. Also, I
>>>> imagine there won't be any more clk/Tegra tree dependencies in 3.10, so
>>>> patch 2/2 and the Tegra114 clk driver patches can likely go through the
>>>> clk tree itself for 3.10.
>>>
>>> No. Because 1/2 changes struct clk_mux and the tegra peripheral clock type 
>>> uses struct clk_mux directly, 2/2 needs to be applied together with 1/2, even
>>> if the new functionality is not yet used.
>>
>> Oh, then they can't be two separate patches then, or "git bisect" won't
>> work. I guess it's best to wait for 3.10 for this:-(