[PATCH v1 1/9] rockchip: RK3328: Read the reset cause from clock reset unit for RK3328 SoC

[Quentin Schulz]

Hi Anand,

On 5/16/24 10:59 AM, Anand Moon wrote:
> Read the reset cause from clock reset unit for RK3328 SoC.
> 
> Cc: Jagan Teki <jagan at edgeble.ai>
> Signed-off-by: Anand Moon <anand at edgeble.ai>
> ---
>   arch/arm/mach-rockchip/cpu-info.c | 7 ++++++-
>   1 file changed, 6 insertions(+), 1 deletion(-)
> 
> diff --git a/arch/arm/mach-rockchip/cpu-info.c b/arch/arm/mach-rockchip/cpu-info.c
> index 14c7331e1a..fce4bd7541 100644
> --- a/arch/arm/mach-rockchip/cpu-info.c
> +++ b/arch/arm/mach-rockchip/cpu-info.c
> @@ -8,12 +8,17 @@
>   #include <init.h>
>   #include <asm/arch-rockchip/clock.h>
>   #include <asm/arch-rockchip/cru.h>
> +#if IS_ENABLED(CONFIG_ROCKCHIP_RK3328)
> +#include <asm/arch-rockchip/cru_rk3328.h>
> +#endif
>   #include <asm/arch-rockchip/hardware.h>
>   #include <linux/err.h>
>   
>   char *get_reset_cause(void)
>   {
> -	struct rockchip_cru *cru = rockchip_get_cru();
> +#if IS_ENABLED(CONFIG_ROCKCHIP_RK3328)
> +	struct rk3328_cru *cru = rockchip_get_cru();
> +#endif

NACK.

If I only apply this patch, it breaks support for rk3288 and rk3399 (the 
ones which have a rockchip_cru struct).

Do:

"""

#if IS_ENABLED(CONFIG_ROCKCHIP_RK3328)
     struct rk3328_cru *cru = rockchip_get_cru();
#else
     struct rockchip_cru *cru = rockchip_get_cru();
#endif
"""

instead to keep the current support and add stuff.

However, I very much don't like this. There's no need for cpuinfo.c to 
become SoC-specific.

I would recommend to move get_reset_cause() in some file that is SoC 
specific, and out of the DISPLAY_CPUINFO ifdef since it may be useful 
outside of this very specific print_cpuinfo function (c.f. the rk3399 
firefly board).

Yes there's a bit of code duplication then, BUT, nothing guarantees us 
the reset reason will always be stored in that glb_rst_st register and 
moreover, RK3588 already has more reasons that currently supported, and 
this prevents us from expanding it.

You can have a

"""
__weak char *get_reset_cause(void)
{
     return "could not get reset cause";
}
"""
in replacement and let SoC-specific files expand it by overriding it.

And.... to think even further than that, maybe we should actually expand 
drivers/sysreset/sysreset_rockchip.c to support more SoCs.

Then we could get the status via get_status callback from sysreset_ops in

https://elixir.bootlin.com/u-boot/latest/source/common/board_f.c#L904

and remove this cpu-info.c file from the arch/arm/mach-rockchip.

Reporting the SoC could be done with a CPU UCLASS (c.f. what's done for 
IMX8 already), from 
https://elixir.bootlin.com/u-boot/latest/source/common/board_f.c#L176.

This honestly makes more sense to me.

Cheers,
Quentin