Re: [PATCHv1 3/3] clk: meson: g12a: set cpu clock divider flags too CLK_IS_CRITICAL

From: Jerome Brunet
Date: Mon Feb 17 2020 - 03:02:27 EST



On Sun 16 Feb 2020 at 18:34, Anand Moon <linux.amoon@xxxxxxxxx> wrote:

> Odroid N2 would fail to boot using microSD unless we set
> cpu freq clk divider flags to CLK_IS_CRITICAL to avoid stalling of
> cpu when booting, most likely because of PWM module linked to

Where did you see a PWM ?

> the CPU for DVFS is getting disabled in between the late_init call,

between the late_init call and what ?

> so gaiting the clock source shuts down the power to the codes.

what code ?

> Setting clk divider flags to CLK_IS_CRITICAL help resolve the issue.
>
> Cc: Martin Blumenstingl <martin.blumenstingl@xxxxxxxxxxxxxx>
> Cc: Jerome Brunet <jbrunet@xxxxxxxxxxxx>
> Cc: Neil Armstrong <narmstrong@xxxxxxxxxxxx>
> Suggested-by: Neil Armstrong <narmstrong@xxxxxxxxxxxx>
> Signed-off-by: Anand Moon <linux.amoon@xxxxxxxxx>
> ---
>
> Following Neil's suggestion, I have prepared this patch.
> https://patchwork.kernel.org/patch/11177441/#22964889
> ---
> drivers/clk/meson/g12a.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/clk/meson/g12a.c b/drivers/clk/meson/g12a.c
> index d2760a021301..accae3695fe5 100644
> --- a/drivers/clk/meson/g12a.c
> +++ b/drivers/clk/meson/g12a.c
> @@ -283,6 +283,7 @@ static struct clk_fixed_factor g12a_fclk_div2_div = {
> .ops = &clk_fixed_factor_ops,
> .parent_hws = (const struct clk_hw *[]) { &g12a_fixed_pll.hw },
> .num_parents = 1,
> + .flags = CLK_IS_CRITICAL,

This makes no sense for because:
* This clock cannot gate and none of its parents can either. IOW, the
output of this clock is never disabled.
* I cannot guess the relation between fdiv2 and the commit description

> },
> };
>
> @@ -681,7 +682,7 @@ static struct clk_regmap g12b_cpub_clk = {
> &g12a_sys_pll.hw
> },
> .num_parents = 2,
> - .flags = CLK_SET_RATE_PARENT,
> + .flags = CLK_SET_RATE_PARENT | CLK_IS_CRITICAL,

Why not. Neil what do you think of this ?
If nothing is claiming this clock and enabling it then I suppose it
could make sense.


> },
> };