Re: [PATCH v3 1/6] soc: mediatek: mtk-devapc: refine devapc interrupt handler

[CK Hu (胡俊光)]

On Thu, 2026-04-16 at 11:12 +0800, Xiaoshun Xu wrote:
> Because the violation IRQ uses a while loop, it might cause the
> system to remain in the interrupt handler indefinitely. We are
> currently optimizing this part of the process to handle only 20
> violations for debug violation issues, and then exit the loop
> 
> Signed-off-by: Xiaoshun Xu <xiaoshun.xu@xxxxxxxxxxxx>
> ---
>  drivers/soc/mediatek/mtk-devapc.c | 8 +++++++-
>  1 file changed, 7 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/soc/mediatek/mtk-devapc.c b/drivers/soc/mediatek/mtk-devapc.c
> index f54c966138b5..c9e1401315ad 100644
> --- a/drivers/soc/mediatek/mtk-devapc.c
> +++ b/drivers/soc/mediatek/mtk-devapc.c
> @@ -12,6 +12,7 @@
>  #include <linux/of_irq.h>
>  #include <linux/of_address.h>
>  
> +#define MAX_VIO_NUM 20
>  #define VIO_MOD_TO_REG_IND(m)	((m) / 32)
>  #define VIO_MOD_TO_REG_OFF(m)	((m) % 32)
>  
> @@ -188,13 +189,18 @@ static void devapc_extract_vio_dbg(struct mtk_devapc_context *ctx)
>   */
>  static irqreturn_t devapc_violation_irq(int irq_number, void *data)
>  {
> +	u32 vio_num = 0;
>  	struct mtk_devapc_context *ctx = data;
>  
> -	while (devapc_sync_vio_dbg(ctx))
> +	mask_module_irq(ctx, true);

mask irq is not related to this patch. This patch care about the infinite loop.
So separate mask irq part to an independent patch and describe why do this.

Regards,
CK

> +
> +	for (vio_num = 0; (vio_num < MAX_VIO_NUM) && (devapc_sync_vio_dbg(ctx)); ++vio_num)
>  		devapc_extract_vio_dbg(ctx);
>  
>  	clear_vio_status(ctx);
>  
> +	mask_module_irq(ctx, false);
> +
>  	return IRQ_HANDLED;
>  }
>