[PATCH v3] clk: tegra: pll: Improve PLLM enable-state detection

[PATCH v3] clk: tegra: pll: Improve PLLM enable-state detection

[Dmitry Osipenko]

Power Management Controller (PMC) can override the PLLM clock settings,
including the enable-state. Although PMC could only act as a second level
gate, meaning that PLLM needs to be enabled by the Clock and Reset
Controller (CaR) anyways if we want it to be enabled. Hence, when PLLM is
overridden by PMC, it needs to be enabled by CaR and ungated by PMC in
order to be functional. Please note that this patch doesn't fix any known
problem, and thus, it's merely a minor improvement.

Link: https://lore.kernel.org/linux-arm-kernel/20191210120909.GA2703785@ulmo/T/
Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
---

Changelog:

v3: - Dropped unintended code change that was accidentally added to v2.

v2: - Added clarifying comment to the code.

    - Prettified the code.

 drivers/clk/tegra/clk-pll.c | 20 +++++++++++++++-----
 1 file changed, 15 insertions(+), 5 deletions(-)

diff --git a/drivers/clk/tegra/clk-pll.c b/drivers/clk/tegra/clk-pll.c
index b2d39a66f0fa..37cfcd6836c9 100644
--- a/drivers/clk/tegra/clk-pll.c
+++ b/drivers/clk/tegra/clk-pll.c
@@ -327,16 +327,26 @@ int tegra_pll_wait_for_lock(struct tegra_clk_pll *pll)
 	return clk_pll_wait_for_lock(pll);
 }
 
+static bool pllm_pmc_clk_enabled(struct tegra_clk_pll *pll)
+{
+	u32 val = readl_relaxed(pll->pmc + PMC_PLLP_WB0_OVERRIDE);
+
+	return !(val & PMC_PLLP_WB0_OVERRIDE_PLLM_OVERRIDE) ||
+		(val & PMC_PLLP_WB0_OVERRIDE_PLLM_ENABLE);
+}
+
 static int clk_pll_is_enabled(struct clk_hw *hw)
 {
 	struct tegra_clk_pll *pll = to_clk_pll(hw);
 	u32 val;
 
-	if (pll->params->flags & TEGRA_PLLM) {
-		val = readl_relaxed(pll->pmc + PMC_PLLP_WB0_OVERRIDE);
-		if (val & PMC_PLLP_WB0_OVERRIDE_PLLM_OVERRIDE)
-			return val & PMC_PLLP_WB0_OVERRIDE_PLLM_ENABLE ? 1 : 0;
-	}
+	/*
+	 * Power Management Controller (PMC) can override the PLLM clock
+	 * settings, including the enable-state. The PLLM is enabled when
+	 * PLLM's CaR state is ON and when PLLM isn't disabled by PMC.
+	 */
+	if ((pll->params->flags & TEGRA_PLLM) && !pllm_pmc_clk_enabled(pll))
+		return 0;
 
 	val = pll_readl_base(pll);
 
-- 
2.26.0

Re: [PATCH v3] clk: tegra: pll: Improve PLLM enable-state detection

[Jon Hunter]

On 08/07/2020 08:54, Dmitry Osipenko wrote:
> Power Management Controller (PMC) can override the PLLM clock settings,
> including the enable-state. Although PMC could only act as a second level
> gate, meaning that PLLM needs to be enabled by the Clock and Reset
> Controller (CaR) anyways if we want it to be enabled. Hence, when PLLM is
> overridden by PMC, it needs to be enabled by CaR and ungated by PMC in
> order to be functional. Please note that this patch doesn't fix any known
> problem, and thus, it's merely a minor improvement.
> 
> Link: https://lore.kernel.org/linux-arm-kernel/20191210120909.GA2703785@ulmo/T/
> Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
> ---
> 
> Changelog:
> 
> v3: - Dropped unintended code change that was accidentally added to v2.
> 
> v2: - Added clarifying comment to the code.
> 
>     - Prettified the code.
> 
>  drivers/clk/tegra/clk-pll.c | 20 +++++++++++++++-----
>  1 file changed, 15 insertions(+), 5 deletions(-)
> 
> diff --git a/drivers/clk/tegra/clk-pll.c b/drivers/clk/tegra/clk-pll.c
> index b2d39a66f0fa..37cfcd6836c9 100644
> --- a/drivers/clk/tegra/clk-pll.c
> +++ b/drivers/clk/tegra/clk-pll.c
> @@ -327,16 +327,26 @@ int tegra_pll_wait_for_lock(struct tegra_clk_pll *pll)
>  	return clk_pll_wait_for_lock(pll);
>  }
>  
> +static bool pllm_pmc_clk_enabled(struct tegra_clk_pll *pll)
> +{
> +	u32 val = readl_relaxed(pll->pmc + PMC_PLLP_WB0_OVERRIDE);
> +
> +	return !(val & PMC_PLLP_WB0_OVERRIDE_PLLM_OVERRIDE) ||
> +		(val & PMC_PLLP_WB0_OVERRIDE_PLLM_ENABLE);
> +}
> +


I am not sure that the name of the above function really reflects what
it is doing. If it was enabled, isn't it the AND of these bits?

Futhermore, what we really want to know is if the override is enabled,
but the PMC PLLM enable is not set. In other words, the PMC is gating
the clock. So maybe we should have a function that is called something
like pllm_clk_is_gated_by_pmc().

Jon

-- 
nvpublic

Re: [PATCH v3] clk: tegra: pll: Improve PLLM enable-state detection

[Dmitry Osipenko]

09.07.2020 12:49, Jon Hunter пишет:
> 
> On 08/07/2020 08:54, Dmitry Osipenko wrote:
>> Power Management Controller (PMC) can override the PLLM clock settings,
>> including the enable-state. Although PMC could only act as a second level
>> gate, meaning that PLLM needs to be enabled by the Clock and Reset
>> Controller (CaR) anyways if we want it to be enabled. Hence, when PLLM is
>> overridden by PMC, it needs to be enabled by CaR and ungated by PMC in
>> order to be functional. Please note that this patch doesn't fix any known
>> problem, and thus, it's merely a minor improvement.
>>
>> Link: https://lore.kernel.org/linux-arm-kernel/20191210120909.GA2703785@ulmo/T/
>> Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
>> ---
>>
>> Changelog:
>>
>> v3: - Dropped unintended code change that was accidentally added to v2.
>>
>> v2: - Added clarifying comment to the code.
>>
>>     - Prettified the code.
>>
>>  drivers/clk/tegra/clk-pll.c | 20 +++++++++++++++-----
>>  1 file changed, 15 insertions(+), 5 deletions(-)
>>
>> diff --git a/drivers/clk/tegra/clk-pll.c b/drivers/clk/tegra/clk-pll.c
>> index b2d39a66f0fa..37cfcd6836c9 100644
>> --- a/drivers/clk/tegra/clk-pll.c
>> +++ b/drivers/clk/tegra/clk-pll.c
>> @@ -327,16 +327,26 @@ int tegra_pll_wait_for_lock(struct tegra_clk_pll *pll)
>>  	return clk_pll_wait_for_lock(pll);
>>  }
>>  
>> +static bool pllm_pmc_clk_enabled(struct tegra_clk_pll *pll)
>> +{
>> +	u32 val = readl_relaxed(pll->pmc + PMC_PLLP_WB0_OVERRIDE);
>> +
>> +	return !(val & PMC_PLLP_WB0_OVERRIDE_PLLM_OVERRIDE) ||
>> +		(val & PMC_PLLP_WB0_OVERRIDE_PLLM_ENABLE);
>> +}
>> +
> 
> 
> I am not sure that the name of the above function really reflects what
> it is doing. If it was enabled, isn't it the AND of these bits?
> 
> Futhermore, what we really want to know is if the override is enabled,
> but the PMC PLLM enable is not set. In other words, the PMC is gating
> the clock. So maybe we should have a function that is called something
> like pllm_clk_is_gated_by_pmc().

Yeah, the name indeed could be improved + the logic could be inverted in
order to make it all more clear. Thank you for the suggestion! I'll
prepare the v4.