Re: [PATCH v1] spi: spi-mtk-nor: Add recovery mechanism for dma read timeout

[AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>]

Il 05/12/22 09:24, Bayi Cheng ha scritto:
> From: bayi cheng <bayi.cheng@mediatek.com>
> 
> The state machine of MTK spi nor controller may be disturbed by some
> glitch signals from the relevant BUS during dma read, Although the
> possibility of causing the dma read to fail is next to nothing,
> However, if error-handling is not implemented, which makes the feature
> somewhat risky.
> 
> Add an error-handling mechanism here, reset the state machine and
> re-read the data when an error occurs.
> 
> Signed-off-by: bayi cheng <bayi.cheng@mediatek.com>
> ---
> Change in v1:
>    -Reset the state machine when dma read fails and read again.
> ---
> ---
>   drivers/spi/spi-mtk-nor.c | 16 +++++++++++++++-
>   1 file changed, 15 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/spi/spi-mtk-nor.c b/drivers/spi/spi-mtk-nor.c
> index d167699a1a96..c77d79da9a4c 100644
> --- a/drivers/spi/spi-mtk-nor.c
> +++ b/drivers/spi/spi-mtk-nor.c
> @@ -80,6 +80,9 @@
>   #define MTK_NOR_REG_DMA_FADR		0x71c
>   #define MTK_NOR_REG_DMA_DADR		0x720
>   #define MTK_NOR_REG_DMA_END_DADR	0x724
> +#define MTK_NOR_REG_CG_DIS		0x728
> +#define MTK_NOR_SFC_SW_RST		BIT(2)
> +
>   #define MTK_NOR_REG_DMA_DADR_HB		0x738
>   #define MTK_NOR_REG_DMA_END_DADR_HB	0x73c
>   
> @@ -616,7 +619,18 @@ static int mtk_nor_exec_op(struct spi_mem *mem, const struct spi_mem_op *op)
>   			mtk_nor_set_addr(sp, op);
>   			return mtk_nor_read_pio(sp, op);
>   		} else {
> -			return mtk_nor_read_dma(sp, op);
> +			ret = mtk_nor_read_dma(sp, op);
> +			if (ret) {
> +				dev_err(sp->dev, "try to read again\n");
> +				mtk_nor_rmw(sp, MTK_NOR_REG_CG_DIS, 0, MTK_NOR_SFC_SW_RST);
> +				mb(); /* flush previous writes */
> +				mtk_nor_rmw(sp, MTK_NOR_REG_CG_DIS, MTK_NOR_SFC_SW_RST, 0);
> +				mb(); /* flush previous writes */
> +				writel(MTK_NOR_ENABLE_SF_CMD, sp->base + MTK_NOR_REG_WP);

 From what I understand, you're introducing a way to perform a flush+reset on
the controller.

At this point, I'd put that in a separate function like `mtk_nor_reset()`, as
to both increase readability and to possibly reuse it somewhere else in the
future, if needed.

So this would become...

		} else {
			ret = mtk_nor_read_dma(sp, op);
			if (unlikely(ret)) {
				/* Handle rare bus glitch */
				mtk_nor_reset(sp);
				mtk_nor_setup_bus(sp, op);
				return mtk_nor_read_dma(sp, op);
			}
			return ret;
		}

...or something alike :-)

Regards,
Angelo