Re: [PATCH v6 13/18] media: platform: Fill stats buffer on ISP_START

[Laurent Pinchart <laurent.pinchart@ideasonboard.com>]

Hi Dan,

Thank you for the patch.

On Tue, Jul 09, 2024 at 02:29:01PM +0100, Daniel Scally wrote:
> On ISP_START, fill the stats buffer by reading out the metering space
> in the ISP's memory. This is done for the non-active config just as
> the dma transfer of the registers is. To acheive that, move the

s/acheive/achieve/

> checking of the current config outside of mali_c55_swap_next_config()
> so we can use it for both functions.
> 
> Acked-by: Nayden Kanchev  <nayden.kanchev@arm.com>
> Co-developed-by: Jacopo Mondi <jacopo.mondi@ideasonboard.com>
> Signed-off-by: Jacopo Mondi <jacopo.mondi@ideasonboard.com>
> Signed-off-by: Daniel Scally <dan.scally@ideasonboard.com>
> ---
> Changes in v6:
> 
> 	- None
> 
> Changes in v5:
> 
> 	- New patch
> 
>  .../platform/arm/mali-c55/mali-c55-core.c     | 34 ++++++++++++++-----
>  1 file changed, 26 insertions(+), 8 deletions(-)
> 
> diff --git a/drivers/media/platform/arm/mali-c55/mali-c55-core.c b/drivers/media/platform/arm/mali-c55/mali-c55-core.c
> index eedc8f450184..ed0db34767a4 100644
> --- a/drivers/media/platform/arm/mali-c55/mali-c55-core.c
> +++ b/drivers/media/platform/arm/mali-c55/mali-c55-core.c
> @@ -567,15 +567,9 @@ static int mali_c55_check_hwcfg(struct mali_c55 *mali_c55)
>  	return 0;
>  }
>  
> -static void mali_c55_swap_next_config(struct mali_c55 *mali_c55)
> +static void mali_c55_swap_next_config(struct mali_c55 *mali_c55, u32 next_config)
>  {
>  	struct mali_c55_context *ctx = mali_c55_get_active_context(mali_c55);
> -	u32 curr_config, next_config;
> -
> -	curr_config = mali_c55_read(mali_c55, MALI_C55_REG_PING_PONG_READ);
> -	curr_config = (curr_config & MALI_C55_REG_PING_PONG_READ_MASK)
> -		      >> (ffs(MALI_C55_REG_PING_PONG_READ_MASK) - 1);
> -	next_config = curr_config ^ 1;
>  
>  	mali_c55_update_bits(mali_c55, MALI_C55_REG_MCU_CONFIG,
>  			     MALI_C55_REG_MCU_CONFIG_WRITE_MASK,
> @@ -588,6 +582,7 @@ static irqreturn_t mali_c55_isr(int irq, void *context)
>  {
>  	struct device *dev = context;
>  	struct mali_c55 *mali_c55 = dev_get_drvdata(dev);
> +	u32 curr_config, next_config;
>  	u32 interrupt_status;
>  	unsigned int i, j;
>  
> @@ -612,7 +607,30 @@ static irqreturn_t mali_c55_isr(int irq, void *context)
>  			for (j = i; j < MALI_C55_NUM_CAP_DEVS; j++)
>  				mali_c55_set_next_buffer(&mali_c55->cap_devs[j]);
>  
> -			mali_c55_swap_next_config(mali_c55);
> +			/*
> +			 * When the ISP starts a frame we have some work to do:
> +			 *
> +			 * 1. Copy over the config for the **next** frame
> +			 * 2. Read out the metering stats for the **last** frame
> +			 */
> +
> +			curr_config = mali_c55_read(mali_c55,
> +						    MALI_C55_REG_PING_PONG_READ);
> +			curr_config &= MALI_C55_REG_PING_PONG_READ_MASK;
> +			curr_config >>= ffs(MALI_C55_REG_PING_PONG_READ_MASK) - 1;
> +			next_config = curr_config ^ 1;
> +
> +			/*
> +			 * The ordering of these two is currently important as
> +			 * mali_c55_stats_fill_buffer() is asynchronous whereas
> +			 * mali_c55_swap_next_config() is not.
> +			 *
> +			 * TODO: Should mali_c55_swap_next_config() be async?

Isn't it in this version of the series, at least when using DMA ?

As I wrote in the review of 07/18, I think the reconfiguration probably
needs more careful consideration.

> +			 */
> +			mali_c55_stats_fill_buffer(mali_c55,
> +				next_config ? MALI_C55_CONFIG_PING :
> +				MALI_C55_CONFIG_PONG);
> +			mali_c55_swap_next_config(mali_c55, next_config);
>  
>  			break;
>  		case MALI_C55_IRQ_ISP_DONE:

-- 
Regards,

Laurent Pinchart