Re: [PATCH 1/2] clk: samsung: exynos4x12: Enable ARMCLK down feature

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

Hi Krzysztof,

On 18.07.2014 11:53, Krzysztof Kozlowski wrote:
> Enable ARMCLK down and up features on Exynos4212 and 4412 SoCs. The
> frequency of ARMCLK will be reduced upon entering idle mode (WFI or
> WFE).  Additionally upon exiting from idle mode the divider for ARMCLK
> will be brought back to 1.
> 
> These are exactly the same settings as for Exynos5250
> (clk-exynos5250.c), except of Exynos4412 where ARMCLK down is enabled
> for all 4 cores.

Could you add a sentence or two about the purpose of this change? E.g.
what it improves, in what conditions, etc.

> 
> Tested on Trats2 board (Exynos4412) and Samsung Gear 1 (Exynos4212).
> 
> Signed-off-by: Krzysztof Kozlowski <k.kozlowski@samsung.com>
> ---
>  drivers/clk/samsung/clk-exynos4.c | 53 +++++++++++++++++++++++++++++++++++++++
>  1 file changed, 53 insertions(+)
> 
> diff --git a/drivers/clk/samsung/clk-exynos4.c b/drivers/clk/samsung/clk-exynos4.c
> index 7f4a473a7ad7..b8ea37b23984 100644
> --- a/drivers/clk/samsung/clk-exynos4.c
> +++ b/drivers/clk/samsung/clk-exynos4.c
> @@ -114,11 +114,34 @@
>  #define DIV_CPU1		0x14504
>  #define GATE_SCLK_CPU		0x14800
>  #define GATE_IP_CPU		0x14900
> +#define PWR_CTRL1		0x15020
> +#define PWR_CTRL2		0x15024

I guess these registers should be also added to save/restore list below
in this driver.

>  #define E4X12_DIV_ISP0		0x18300
>  #define E4X12_DIV_ISP1		0x18304
>  #define E4X12_GATE_ISP0		0x18800
>  #define E4X12_GATE_ISP1		0x18804
>  
> +/* Below definitions are used for PWR_CTRL settings */
> +#define PWR_CTRL1_CORE2_DOWN_RATIO		(7 << 28)
> +#define PWR_CTRL1_CORE1_DOWN_RATIO		(7 << 16)

I think these macros could be defined to take the ratio as its argument,
e.g.

#define PWR_CTRL1_CORE2_DOWN_RATIO(x)		((x) << 28)

> +#define PWR_CTRL1_DIV2_DOWN_EN			(1 << 9)
> +#define PWR_CTRL1_DIV1_DOWN_EN			(1 << 8)
> +#define PWR_CTRL1_USE_CORE3_WFE			(1 << 7)
> +#define PWR_CTRL1_USE_CORE2_WFE			(1 << 6)
> +#define PWR_CTRL1_USE_CORE1_WFE			(1 << 5)
> +#define PWR_CTRL1_USE_CORE0_WFE			(1 << 4)
> +#define PWR_CTRL1_USE_CORE3_WFI			(1 << 3)
> +#define PWR_CTRL1_USE_CORE2_WFI			(1 << 2)
> +#define PWR_CTRL1_USE_CORE1_WFI			(1 << 1)
> +#define PWR_CTRL1_USE_CORE0_WFI			(1 << 0)
> +
> +#define PWR_CTRL2_DIV2_UP_EN			(1 << 25)
> +#define PWR_CTRL2_DIV1_UP_EN			(1 << 24)
> +#define PWR_CTRL2_DUR_STANDBY2_VAL		(1 << 16)
> +#define PWR_CTRL2_DUR_STANDBY1_VAL		(1 << 8)

These two too.

> +#define PWR_CTRL2_CORE2_UP_RATIO		(1 << 4)
> +#define PWR_CTRL2_CORE1_UP_RATIO		(1 << 0)

These two as well.

> +
>  /* the exynos4 soc type */
>  enum exynos4_soc {
>  	EXYNOS4210,
> @@ -1164,6 +1187,34 @@ static struct samsung_pll_clock exynos4x12_plls[nr_plls] __initdata = {
>  			VPLL_LOCK, VPLL_CON0, NULL),
>  };
>  
> +static void __init exynos4_core_down_clock(void)
> +{
> +	unsigned int tmp;
> +
> +	/*
> +	 * Enable arm clock down (in idle) and set arm divider
> +	 * ratios in WFI/WFE state.
> +	 */
> +	tmp = (PWR_CTRL1_CORE2_DOWN_RATIO | PWR_CTRL1_CORE1_DOWN_RATIO |
> +		PWR_CTRL1_DIV2_DOWN_EN | PWR_CTRL1_DIV1_DOWN_EN |
> +		PWR_CTRL1_USE_CORE1_WFE | PWR_CTRL1_USE_CORE0_WFE |
> +		PWR_CTRL1_USE_CORE1_WFI | PWR_CTRL1_USE_CORE0_WFI);
> +	if (of_machine_is_compatible("samsung,exynos4412"))

Maybe you could use num_possible_cpus() here instead?

> +		tmp |= PWR_CTRL1_USE_CORE3_WFE | PWR_CTRL1_USE_CORE2_WFE |
> +		       PWR_CTRL1_USE_CORE3_WFI | PWR_CTRL1_USE_CORE2_WFI;
> +	__raw_writel(tmp, reg_base + PWR_CTRL1);
> +
> +	/*
> +	 * Enable arm clock up (on exiting idle). Set arm divider
> +	 * ratios when not in idle along with the standby duration
> +	 * ratios.
> +	 */
> +	tmp = (PWR_CTRL2_DIV2_UP_EN | PWR_CTRL2_DIV1_UP_EN |
> +		PWR_CTRL2_DUR_STANDBY2_VAL | PWR_CTRL2_DUR_STANDBY1_VAL |
> +		PWR_CTRL2_CORE2_UP_RATIO | PWR_CTRL2_CORE1_UP_RATIO);
> +	__raw_writel(tmp, reg_base + PWR_CTRL2);

Do we need the clock up feature at all? The values you set seem to
configure both dividers to 2 (resulting in arm_clk = apll / 4) for a
very short period of time (16 ticks of some, unfortunately unspecified,
clock) between wake-up and setting the frequency back to normal.
Moreover, for certain settings (div_core * div_core2 set to > 4 by
cpufreq) this might cause issues with activating higher frequency than
current voltage allows.

I believe the same comments apply for patch 2/2 as well.

Best regards,
Tomasz