Re: [PATCH v2 3/7] clk: thead: Mark essential bus clocks as CLK_IGNORE_UNUSED

[Drew Fustini <drew@pdp7.com>]

On Tue, Jun 10, 2025 at 02:52:51PM +0200, Michal Wilczynski wrote:
> Probing peripherals in the AON and PERI domains, such as the PVT thermal
> sensor and the PWM controller, can lead to boot hangs or unresponsive
> devices on the LPi4A board. The root cause is that their parent bus
> clocks ('CLK_CPU2AON_X2H' and the 'CLK_PERISYS_APB' clocks) are
> automatically gated by the kernel's power-saving mechanisms when the bus
> is perceived as idle.
> 
> Alternative solutions were investigated, including modeling the parent
> bus in the Device Tree with 'simple-pm-bus' or refactoring the clock
> driver's parentage. The 'simple-pm-bus' approach is not viable due to
> the lack of defined bus address ranges in the hardware manual and its
> creation of improper dependencies on the 'pm_runtime' API for consumer
> drivers.
> 
> Therefore, applying the'`CLK_IGNORE_UNUSED' flag directly to the
> essential bus clocks is the most direct and targeted fix. This prevents
> the kernel from auto-gating these buses and ensures peripherals remain
> accessible.
> 
> This change fixes the boot hang associated with the PVT sensor and
> resolves the functional issues with the PWM controller.
> 
> [1] - https://lore.kernel.org/all/9e8a12db-236d-474c-b110-b3be96edf057@samsung.com/
> 
> Signed-off-by: Michal Wilczynski <m.wilczynski@samsung.com>
> ---
>  drivers/clk/thead/clk-th1520-ap.c | 5 +++--
>  1 file changed, 3 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/clk/thead/clk-th1520-ap.c b/drivers/clk/thead/clk-th1520-ap.c
> index ebfb1d59401d05443716eb0029403b01775e8f73..cf7f6bd428a0faa4611b3fc61edbbc6690e565d9 100644
> --- a/drivers/clk/thead/clk-th1520-ap.c
> +++ b/drivers/clk/thead/clk-th1520-ap.c
> @@ -792,11 +792,12 @@ static CCU_GATE(CLK_AON2CPU_A2X, aon2cpu_a2x_clk, "aon2cpu-a2x", axi4_cpusys2_ac
>  		0x134, BIT(8), 0);
>  static CCU_GATE(CLK_X2X_CPUSYS, x2x_cpusys_clk, "x2x-cpusys", axi4_cpusys2_aclk_pd,
>  		0x134, BIT(7), 0);
> -static CCU_GATE(CLK_CPU2AON_X2H, cpu2aon_x2h_clk, "cpu2aon-x2h", axi_aclk_pd, 0x138, BIT(8), 0);
> +static CCU_GATE(CLK_CPU2AON_X2H, cpu2aon_x2h_clk, "cpu2aon-x2h", axi_aclk_pd,
> +		0x138, BIT(8), CLK_IGNORE_UNUSED);
>  static CCU_GATE(CLK_CPU2PERI_X2H, cpu2peri_x2h_clk, "cpu2peri-x2h", axi4_cpusys2_aclk_pd,
>  		0x140, BIT(9), CLK_IGNORE_UNUSED);
>  static CCU_GATE(CLK_PERISYS_APB1_HCLK, perisys_apb1_hclk, "perisys-apb1-hclk", perisys_ahb_hclk_pd,
> -		0x150, BIT(9), 0);
> +		0x150, BIT(9), CLK_IGNORE_UNUSED);
>  static CCU_GATE(CLK_PERISYS_APB2_HCLK, perisys_apb2_hclk, "perisys-apb2-hclk", perisys_ahb_hclk_pd,
>  		0x150, BIT(10), CLK_IGNORE_UNUSED);
>  static CCU_GATE(CLK_PERISYS_APB3_HCLK, perisys_apb3_hclk, "perisys-apb3-hclk", perisys_ahb_hclk_pd,
> 
> -- 
> 2.34.1
> 

I'm okay with fixing it this way for now and revisiting the parent
relationships later.

Reviewed-by: Drew Fustini <drew@pdp7.com>