Re: [PATCH v2] ARM: EXYNOS: Refactor the pm code to use DT based lookup

[Tomasz Figa <t.figa@samsung.com>]

On 06.08.2014 14:58, Vikas Sajjan wrote:
> On Wed, Aug 6, 2014 at 1:35 AM, Tomasz Figa <tomasz.figa@gmail.com> wrote:
>> On 25.07.2014 13:49, Vikas Sajjan wrote:

[snip]

> 
>>> +
>>> +     /* Disable USE_RETENTION of JPEG_MEM_OPTION */
>>> +     tmp = pmu_raw_readl(EXYNOS5_JPEG_MEM_OPTION);
>>> +     tmp &= ~EXYNOS5_OPTION_USE_RETENTION;
>>> +     pmu_raw_writel(tmp, EXYNOS5_JPEG_MEM_OPTION);
>>
>> This looks completely like a copy paste from a vendor tree needed to
>> workaround some issues in early revisions of the SoC. Are you sure this
>> is still needed in production versions of the silicon used on boards
>> supported in mainline?
> 
> This piece of code is NOT copy paste from my side, it is an already
> existing code. I just moved it into the function
> exynos5250_pm_prepare().

Sure, I'm aware of this. It might be good to know though if this is
really necessary, as I don't think we want useless code.

> However I removed this piece of code and
> made a common function for exynos4 and exynos5250, S2R works on 5250
> well even without this piece of code.

Thanks for checking this. I wonder if we could get some clarification
about this from hardware guys. Kukjin, Jingoo, any ideas or people that
might know what this is about?

For now, I believe it could be handled the same way as Exynos4, in PMU
register array as I suggested in my previous reply (quoted below).

> 
>>
>> Even if yes, Exynos4 handles such registers in PMU register array in
>> pmu.c, so I wonder whether it shouldn't be moved there and allow
>> handling of all Exynos SoCs uniformly in this file.
>>
>>> +
>>> +     exynos_pm_enter_sleep_mode();
>>> +}
>>> +
>>> +static void exynos4_pm_prepare(void)
>>> +{
>>> +     /* Set wake-up mask registers */
>>> +     exynos_pm_set_wakeup_mask();
>>> +
>>> +     s3c_pm_do_save(exynos_core_save, ARRAY_SIZE(exynos_core_save));
>>> +
>>> +     exynos_pm_enter_sleep_mode();
>>> +}
>>
>> In fact, exynos4_pm_prepare() is a direct subset of
>> exynos5250_pm_prepare(). This just confirms that it might be a good idea
>> to just merge both functions into a single generic one.
> 
> Right, can be merged into a common function which can be used for
> exynos4 and exynos5250.
> But I am afraid we still need specific functions in case of 5420.

Could you provide some details about Exynos5420 specific things?

> 
>>
>>> +
>>>  static void exynos_pm_central_suspend(void)
>>>  {
>>>       unsigned long tmp;

[snip]

>>> +
>>>       /* Platform-specific GIC callback */
>>>       gic_arch_extn.irq_set_wake = exynos_irq_set_wake;
>>>
>>>       /* All wakeup disable */
>>>       tmp = pmu_raw_readl(S5P_WAKEUP_MASK);
>>> -     tmp |= ((0xFF << 8) | (0x1F << 1));
>>> +     tmp |= pm_data->wake_disable_mask;
>>
>> Does this vary between SoCs?
> 
> Yes, It is different in case of 5420.

Could you provide more information about the difference?

Best regards,
Tomasz