[PATCH 1/3] mtd: rawnand: qcom: Update read_loc size to 512

Miquel Raynal miquel.raynal at bootlin.com
Fri Aug 18 12:59:08 PDT 2023


Hi Md,

quic_mdalam at quicinc.com wrote on Fri, 18 Aug 2023 20:20:59 +0530:

> For parameter page read upper layer is passing len
> as 256 bytes and if we try to configure 256 bytes
> size in read loaction register then subsequent bam
> transaction is getting timed out for 4K nand devices.
> So update this length as one step size if its
> less than NANDC_STEP_SIZE.
> 
> Signed-off-by: Sricharan Ramabadhran <quic_srichara at quicinc.com>
> Signed-off-by: Md Sadre Alam <quic_mdalam at quicinc.com>

I'm fine with patches 2 and 3 and will take them. But this one does not
seem legitimate. I don't like it. Are you sure the ECC engine was not
enabled when it timed out? Default should be having the ECC disabled
and it should just get enabled when you need it. There is no reason
why, specifically on NAND devices, it would not be possible to read 256
bytes.

> ---
>  drivers/mtd/nand/raw/qcom_nandc.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/drivers/mtd/nand/raw/qcom_nandc.c b/drivers/mtd/nand/raw/qcom_nandc.c
> index d4ba0d04c970..413e214c8e87 100644
> --- a/drivers/mtd/nand/raw/qcom_nandc.c
> +++ b/drivers/mtd/nand/raw/qcom_nandc.c
> @@ -2885,6 +2885,9 @@ static int qcom_param_page_type_exec(struct nand_chip *chip,  const struct nand_
>  	op_id = q_op.data_instr_idx;
>  	len = nand_subop_get_data_len(subop, op_id);
>  
> +	if (len < NANDC_STEP_SIZE)
> +		len = NANDC_STEP_SIZE;
> +
>  	nandc_set_read_loc(chip, 0, 0, 0, len, 1);
>  
>  	if (!nandc->props->qpic_v2) {


Thanks,
Miquèl



More information about the linux-mtd mailing list