[PATCH v9 6/6] clk: samsung: remove unused clock aliases and update clock flags

[Thomas Abraham]

On Fri, Aug 1, 2014 at 12:05 AM, Tomasz Figa <tomasz.figa at gmail.com> wrote:
> On 31.07.2014 20:24, Thomas Abraham wrote:
>> Hi Tomasz,
>>
>> On Thu, Jul 31, 2014 at 7:43 PM, Tomasz Figa <tomasz.figa at gmail.com> wrote:
>>> On 30.07.2014 10:07, Thomas Abraham wrote:
>>>> With some of the Exynos SoCs switched over to use the generic CPUfreq drivers,
>>>> the unused clock aliases can be removed. In addition to this, the individual
>>>> clock blocks which are now encapsulated with the consolidate CPU clock type
>>>> can now be marked with read-only flags.
>>>
>>> [snip]
>>>
>>>> @@ -1500,6 +1499,7 @@ static void __init exynos4_clk_init(struct device_node *np,
>>>>               exynos4_soc == EXYNOS4210 ? "Exynos4210" : "Exynos4x12",
>>>>               _get_rate("sclk_apll"), _get_rate("sclk_mpll"),
>>>>               _get_rate("sclk_epll"), _get_rate("sclk_vpll"),
>>>> +             exynos4_soc == EXYNOS4210 ? _get_rate("armclk") :
>>>>               _get_rate("div_core2"));
>>>
>>> I believe "div_core2" should work fine here for all SoCs without the
>>> need for this if.
>>
>> The following patch is a pre-requisite for this patch.
>> http://www.spinics.net/lists/arm-kernel/msg351540.html
>>
>> The rate can be obtained from div_core2 as well but with the cpu clock
>> now registered, the rate can be obtained from the cpu clock instance
>> instead of the div_core2 divider. And when Exynos4412 also add cpu
>> clock instance, the 'if' above will be removed.
>>
>>>
>>>>  }
>>>>
>>>> diff --git a/drivers/clk/samsung/clk-exynos5250.c b/drivers/clk/samsung/clk-exynos5250.c
>>>> index e19e365..1d958f1 100644
>>>> --- a/drivers/clk/samsung/clk-exynos5250.c
>>>> +++ b/drivers/clk/samsung/clk-exynos5250.c
>>>
>>> [snip]
>>>
>>>> @@ -848,6 +851,6 @@ static void __init exynos5250_clk_init(struct device_node *np)
>>>>       samsung_clk_of_add_provider(np, ctx);
>>>>
>>>>       pr_info("Exynos5250: clock setup completed, armclk=%ld\n",
>>>> -                     _get_rate("div_arm2"));
>>>> +                     _get_rate("armclk"));
>>>
>>> Similarly here, no need for this change.
>>
>> Same here. Instead of getting the rate from div_core2 divider block,
>> the cpu clock instance is used to find the rate. I would prefer to use
>> cpu clock here. Is there any reason to prefer div_core2 over the cpu
>> clock instance?
>
> Well, the reason is simple: if you don't need to change something (i.e.
> the change doesn't have any advantages), don't change it.

The advantage with using cpu clock would be that get_rate can obtain
the cached rate whereas when reading div_core2 rate, the clock tree
will have to be traversed to determine the rate.

>
> There is no difference between obtaining the rate from div_{arm,core}2
> and armclk, so I don't see the point of changing this.
>
> In fact now when thinking of it, this has revealed one hole that will be
> unhandled by your code - if cpufreq is disabled and the bootloader
> configures div_{arm,core}{,2} with non-zero values, armclk will return
> incorrect rate. However since I haven't observed such case on existing
> platforms, fixing this might be done on top of this series, in a
> separate patch.

Right. I will fix this later.

Thanks,
Thomas.

>
> Best regards,
> Tomasz