Re: [RESEND PATCH V2 2/2] mtd: core: NAND filling block

[Boris Brezillon <bbrezillon@xxxxxxxxxx>]

On Fri, 18 Jan 2019 22:12:04 +0000
"Bean Huo (beanhuo)" <beanhuo@xxxxxxxxxx> wrote:

> +static int check_page_if_emtpy(struct nand_chip *chip, char *data)
> +{
> +	struct mtd_info *mtd = nand_to_mtd(chip);
> +	int ret, i;
> +	void *databuf, *eccbuf;
> +	int max_bitflips;
> +	struct mtd_oob_region oobregion;
> +
> +	mtd_ooblayout_ecc(mtd, 0, &oobregion);
> +	eccbuf = chip->oob_poi + oobregion.offset;
> +	databuf = data;
> +	max_bitflips = 0;
> +
> +	for (i = 0; i < chip->ecc.steps; i++) {
> +		ret = nand_check_erased_ecc_chunk(data,
> +						  chip->ecc.size,
> +						  eccbuf,
> +						  chip->ecc.bytes,
> +						  NULL, 0,
> +						  chip->ecc.strength);
> +		if (ret >= 0)
> +			max_bitflips = max(ret, max_bitflips);
> +		else
> +			return false;
> +
> +		databuf += chip->ecc.size;
> +		eccbuf += chip->ecc.bytes;
> +
> +	}

You should check the whole page and not only in-band-data+ECC bytes.
Plus, the "overwrite page" trigger is not dependent on the ECC
strength, but more something related to the NAND chip itself so using
ecc->strength as a threshold sounds like a bad idea to me.

______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/