Initially Common Clock Framework isn't aware of the clock-enable status, this results in enabling of clocks that were enabled by bootloader. This is not a big deal for a regular clock-gates, but for PLL's it may have some unpleasant consequences. Thus re-enabling PLLX (the main CPU parent clock) may result in extra long period of PLL re-locking. Signed-off-by: Dmitry Osipenko <digetx@gmail.com> --- drivers/clk/tegra/clk-pll.c | 50 +++++++++++++++++++++++++++---------- 1 file changed, 37 insertions(+), 13 deletions(-) diff --git a/drivers/clk/tegra/clk-pll.c b/drivers/clk/tegra/clk-pll.c index b50b7460014b..ebc8481a2122 100644 --- a/drivers/clk/tegra/clk-pll.c +++ b/drivers/clk/tegra/clk-pll.c @@ -444,6 +444,9 @@ static int clk_pll_enable(struct clk_hw *hw) unsigned long flags = 0; int ret; + if (clk_pll_is_enabled(hw)) + return 0; + if (pll->lock) spin_lock_irqsave(pll->lock, flags); @@ -940,11 +943,16 @@ static int clk_plle_training(struct tegra_clk_pll *pll) static int clk_plle_enable(struct clk_hw *hw) { struct tegra_clk_pll *pll = to_clk_pll(hw); - unsigned long input_rate = clk_hw_get_rate(clk_hw_get_parent(hw)); struct tegra_clk_pll_freq_table sel; + unsigned long input_rate; u32 val; int err; + if (clk_pll_is_enabled(hw)) + return 0; + + input_rate = clk_hw_get_rate(clk_hw_get_parent(hw)); + if (_get_table_rate(hw, &sel, pll->params->fixed_rate, input_rate)) return -EINVAL; @@ -1355,6 +1363,9 @@ static int clk_pllc_enable(struct clk_hw *hw) int ret; unsigned long flags = 0; + if (clk_pll_is_enabled(hw)) + return 0; + if (pll->lock) spin_lock_irqsave(pll->lock, flags); @@ -1567,7 +1578,12 @@ static int clk_plle_tegra114_enable(struct clk_hw *hw) u32 val; int ret; unsigned long flags = 0; - unsigned long input_rate = clk_hw_get_rate(clk_hw_get_parent(hw)); + unsigned long input_rate; + + if (clk_pll_is_enabled(hw)) + return 0; + + input_rate = clk_hw_get_rate(clk_hw_get_parent(hw)); if (_get_table_rate(hw, &sel, pll->params->fixed_rate, input_rate)) return -EINVAL; @@ -1704,6 +1720,9 @@ static int clk_pllu_tegra114_enable(struct clk_hw *hw) return -EINVAL; } + if (clk_pll_is_enabled(hw)) + return 0; + input_rate = clk_hw_get_rate(__clk_get_hw(osc)); if (pll->lock) @@ -2379,6 +2398,16 @@ struct clk *tegra_clk_register_pllre_tegra210(const char *name, return clk; } +static int clk_plle_tegra210_is_enabled(struct clk_hw *hw) +{ + struct tegra_clk_pll *pll = to_clk_pll(hw); + u32 val; + + val = pll_readl_base(pll); + + return val & PLLE_BASE_ENABLE ? 1 : 0; +} + static int clk_plle_tegra210_enable(struct clk_hw *hw) { struct tegra_clk_pll *pll = to_clk_pll(hw); @@ -2386,7 +2415,12 @@ static int clk_plle_tegra210_enable(struct clk_hw *hw) u32 val; int ret = 0; unsigned long flags = 0; - unsigned long input_rate = clk_hw_get_rate(clk_hw_get_parent(hw)); + unsigned long input_rate; + + if (clk_plle_tegra210_is_enabled(hw)) + return 0; + + input_rate = clk_hw_get_rate(clk_hw_get_parent(hw)); if (_get_table_rate(hw, &sel, pll->params->fixed_rate, input_rate)) return -EINVAL; @@ -2497,16 +2531,6 @@ static void clk_plle_tegra210_disable(struct clk_hw *hw) spin_unlock_irqrestore(pll->lock, flags); } -static int clk_plle_tegra210_is_enabled(struct clk_hw *hw) -{ - struct tegra_clk_pll *pll = to_clk_pll(hw); - u32 val; - - val = pll_readl_base(pll); - - return val & PLLE_BASE_ENABLE ? 1 : 0; -} - static const struct clk_ops tegra_clk_plle_tegra210_ops = { .is_enabled = clk_plle_tegra210_is_enabled, .enable = clk_plle_tegra210_enable, -- 2.20.1
24.02.2019 18:32, Dmitry Osipenko пишет:
> Initially Common Clock Framework isn't aware of the clock-enable status,
> this results in enabling of clocks that were enabled by bootloader. This
> is not a big deal for a regular clock-gates, but for PLL's it may have
> some unpleasant consequences. Thus re-enabling PLLX (the main CPU parent
> clock) may result in extra long period of PLL re-locking.
>
> Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
> ---
Peter (and everyone else), yours r-b and ACK will be appreciated since you kinda already took a look at this patch in the past and were okay with it (IIRC, we had some brief discussion on the #tegra IRC a few months ago), thanks.
On Wed, Feb 27, 2019 at 06:15:04PM +0300, Dmitry Osipenko wrote:
> 24.02.2019 18:32, Dmitry Osipenko пишет:
> > Initially Common Clock Framework isn't aware of the clock-enable status,
> > this results in enabling of clocks that were enabled by bootloader. This
> > is not a big deal for a regular clock-gates, but for PLL's it may have
> > some unpleasant consequences. Thus re-enabling PLLX (the main CPU parent
> > clock) may result in extra long period of PLL re-locking.
> >
> > Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
> > ---
>
> Peter (and everyone else), yours r-b and ACK will be appreciated since you kinda already took a look at this patch in the past and were okay with it (IIRC, we had some brief discussion on the #tegra IRC a few months ago), thanks.
>
Acked-By: Peter De Schrijver <pdeschrijver@nvidia.com>
04.03.2019 11:18, Peter De Schrijver пишет:
> On Wed, Feb 27, 2019 at 06:15:04PM +0300, Dmitry Osipenko wrote:
>> 24.02.2019 18:32, Dmitry Osipenko пишет:
>>> Initially Common Clock Framework isn't aware of the clock-enable status,
>>> this results in enabling of clocks that were enabled by bootloader. This
>>> is not a big deal for a regular clock-gates, but for PLL's it may have
>>> some unpleasant consequences. Thus re-enabling PLLX (the main CPU parent
>>> clock) may result in extra long period of PLL re-locking.
>>>
>>> Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
>>> ---
>>
>> Peter (and everyone else), yours r-b and ACK will be appreciated since you kinda already took a look at this patch in the past and were okay with it (IIRC, we had some brief discussion on the #tegra IRC a few months ago), thanks.
>>
>
> Acked-By: Peter De Schrijver <pdeschrijver@nvidia.com>
>
Awesome, thanks again!
04.03.2019 16:42, Dmitry Osipenko пишет:
> 04.03.2019 11:18, Peter De Schrijver пишет:
>> On Wed, Feb 27, 2019 at 06:15:04PM +0300, Dmitry Osipenko wrote:
>>> 24.02.2019 18:32, Dmitry Osipenko пишет:
>>>> Initially Common Clock Framework isn't aware of the clock-enable status,
>>>> this results in enabling of clocks that were enabled by bootloader. This
>>>> is not a big deal for a regular clock-gates, but for PLL's it may have
>>>> some unpleasant consequences. Thus re-enabling PLLX (the main CPU parent
>>>> clock) may result in extra long period of PLL re-locking.
>>>>
>>>> Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
>>>> ---
>>>
>>> Peter (and everyone else), yours r-b and ACK will be appreciated since you kinda already took a look at this patch in the past and were okay with it (IIRC, we had some brief discussion on the #tegra IRC a few months ago), thanks.
>>>
>>
>> Acked-By: Peter De Schrijver <pdeschrijver@nvidia.com>
>>
>
> Awesome, thanks again!
>
Hello Stephen and Michael,
Could you please pick up this patch to get it into linux-next to get some more testing exposure before it will hit upstream? Thanks in advance.