[PATCH 21/33] clk: ux500: Add Device Tree support for the PRCC Kernel clock

Linus Walleij linus.walleij at linaro.org
Thu Sep 12 08:50:02 EDT 2013


On Tue, Aug 27, 2013 at 10:23 AM, Lee Jones <lee.jones at linaro.org> wrote:
> On Fri, 23 Aug 2013, Linus Walleij wrote:
>> On Thu, Aug 22, 2013 at 11:21 AM, Lee Jones <lee.jones at linaro.org> wrote:

>> > 1. Duplicate each of the; clk_reg_prcmu_*(), clk_reg_prcc_pclk(),
>> > clk_reg_prcc_kclk() calls into your proposed u8500_clk_init_dt(),
>> > which, while keeping everything separate would be unrealistic.
>>
>> I think this is perfectly realistic.
>>
>> You're not going to duplicate each clk_register_clkdev(),
>> which makes it way smaller than the original function,
>> and since one of the function will be inside a
>>
>> #ifdef CONFIG_OF
>> #endif
>>
>> After we switch the entire platform to DT-only it will be pretty
>> obvious which big chunk of code that needs to go away, it's
>> a clean cut.
>>
>> (Note: I know the #ifdef CONFIG_OF is not necessary anymore
>> since we switched to multiplatform, but I intend that marker for
>> humans, not machines.)
>
> This sounds gross. To duplicate; u8500_clk_init(), u8540_clk_init()
> and u9540_clk_init() just for the sake of loading a few pointers into
> an array for a small part of the development cycle sounds obscene.
>
> I genuinely think keeping the current patch in this series and then
> removing the clk_register_clkdev() in the remove ATAG support series
> is the best way to go.

So what I am worrying about is not only the looks of the code
and what is beautiful or not may be something of an opinion
anway.

What I worry about is leaving all the calls to clk_register_clkdev()
in the DT boot path. Because that has the potential to hide a lot
of bugs, as clk_get() from drivers that should've got named and
probed randomly now will still find their clocks from their old
device names, instead of using the <&ampersand> clocks from
the device tree.

But if you still don't like this, let me cook a counter-patch so
I can realized on my own how terribly wrong I am...

Yours,
Linus Walleij



More information about the linux-arm-kernel mailing list