[PATCH 5/8] mtd: nand: mxc: Fix failed/corrected values for v1 controllers

Boris Brezillon boris.brezillon at free-electrons.com
Fri Jan 12 07:28:50 PST 2018


On Tue,  9 Jan 2018 11:11:45 +0100
Sascha Hauer <s.hauer at pengutronix.de> wrote:

> The v1 controller code has several flaws:
> - We do not forward the number of corrected bitflips to the upper layers
> - For 2k page NAND chips only the status results from the fourth subpage
>   read are evaluated, so ECC failures in the other subpages remain
>   uncovered
> - When there are uncorrectable errors we have to increase the statistics
>   counter, but still have to return successfully. Currently we return
>   an error
> 
> This patch fixes this by introducing a v1 specific read_page function.
> 
> Signed-off-by: Sascha Hauer <s.hauer at pengutronix.de>
> ---
>  drivers/mtd/nand/mxc_nand.c | 78 +++++++++++++++++++++++++++++++++++----------
>  1 file changed, 62 insertions(+), 16 deletions(-)
> 
> diff --git a/drivers/mtd/nand/mxc_nand.c b/drivers/mtd/nand/mxc_nand.c
> index d698f7c6666c..83bb01ab7079 100644
> --- a/drivers/mtd/nand/mxc_nand.c
> +++ b/drivers/mtd/nand/mxc_nand.c
> @@ -738,22 +738,68 @@ static void mxc_nand_enable_hwecc(struct mtd_info *mtd, int mode)
>  static int mxc_nand_correct_data_v1(struct mtd_info *mtd, u_char *dat,
>  				 u_char *read_ecc, u_char *calc_ecc)
>  {
> +	return 0;
> +}
> +
> +static int mxc_nand_read_page_v1(struct mtd_info *mtd, struct nand_chip *chip,
> +				 void *buf, void *oob, bool ecc, int page)
> +{
>  	struct nand_chip *nand_chip = mtd_to_nand(mtd);
> -	struct mxc_nand_host *host = nand_get_controller_data(nand_chip);
> +	struct mxc_nand_host *host = nand_get_controller_data(chip);
> +	unsigned int max_bitflips = 0;
> +	int no_subpages;
> +	int i;
>  
> -	/*
> -	 * 1-Bit errors are automatically corrected in HW.  No need for
> -	 * additional correction.  2-Bit errors cannot be corrected by
> -	 * HW ECC, so we need to return failure
> -	 */
> -	uint16_t ecc_status = get_ecc_status_v1(host);
> +	host->devtype_data->enable_hwecc(nand_chip, ecc);
> +
> +	host->devtype_data->send_cmd(host, NAND_CMD_READ0, false);
> +	mxc_do_addr_cycle(mtd, 0, page);
> +
> +	if (mtd->writesize > 512)
> +		host->devtype_data->send_cmd(host, NAND_CMD_READSTART, true);
>  
> -	if (((ecc_status & 0x3) == 2) || ((ecc_status >> 2) == 2)) {
> -		dev_dbg(host->dev, "HWECC uncorrectable 2-bit ECC error\n");
> -		return -EBADMSG;
> +	no_subpages = mtd->writesize >> 9;
> +
> +	for (i = 0; i < no_subpages; i++) {
> +		int flips = 0;
> +		uint16_t ecc_stats;
> +
> +		/* NANDFC buffer 0 is used for page read/write */
> +		writew((host->active_cs << 4) | i, NFC_V1_V2_BUF_ADDR);
> +
> +		writew(NFC_OUTPUT, NFC_V1_V2_CONFIG2);
> +
> +		/* Wait for operation to complete */
> +		wait_op_done(host, true);
> +
> +		ecc_stats = get_ecc_status_v1(host);
> +
> +		ecc_stats >>= 2;
> +
> +		if (buf && ecc) {
> +			switch (ecc_stats & 0x3) {
> +			case 0:
> +			default:
> +				break;
> +			case 1:
> +				mtd->ecc_stats.corrected++;
> +				flips++;

Since the engine is anyway able to correct a maximum of 1 bitflip you
can get rid of flips and do:

				max_bitflips = 1;


You could even rename max_bitflips into bitflips_corrected.

> +				break;
> +			case 2:
> +				mtd->ecc_stats.failed++;
> +				break;
> +			}
> +		}
> +
> +		max_bitflips = max_t(unsigned int, max_bitflips, flips);

If you get rid of flips, this is not needed.

>  	}
>  
> -	return 0;
> +	if (buf)
> +		memcpy32_fromio(buf, host->main_area0, mtd->writesize);
> +	if (oob)
> +		copy_spare(mtd, true, oob);
> +
> +	return max_bitflips;
>  }
>  
>  static int mxc_nand_correct_data_v2_v3(struct mtd_info *mtd, u_char *dat,
> @@ -1494,6 +1540,7 @@ static struct nand_bbt_descr bbt_mirror_descr = {
>  /* v1 + irqpending_quirk: i.MX21 */
>  static const struct mxc_nand_devtype_data imx21_nand_devtype_data = {
>  	.preset = preset_v1,
> +	.read_page = mxc_nand_read_page_v1,
>  	.send_cmd = send_cmd_v1_v2,
>  	.send_addr = send_addr_v1_v2,
>  	.send_page = send_page_v1,
> @@ -1518,6 +1565,7 @@ static const struct mxc_nand_devtype_data imx21_nand_devtype_data = {
>  /* v1 + !irqpending_quirk: i.MX27, i.MX31 */
>  static const struct mxc_nand_devtype_data imx27_nand_devtype_data = {
>  	.preset = preset_v1,
> +	.read_page = mxc_nand_read_page_v1,
>  	.send_cmd = send_cmd_v1_v2,
>  	.send_addr = send_addr_v1_v2,
>  	.send_page = send_page_v1,
> @@ -1856,11 +1904,9 @@ static int mxcnd_probe(struct platform_device *pdev)
>  
>  	switch (this->ecc.mode) {
>  	case NAND_ECC_HW:
> -		if (host->devtype_data->read_page) {
> -			this->ecc.read_page = mxc_nand_read_page;
> -			this->ecc.read_page_raw = mxc_nand_read_page_raw;
> -			this->ecc.read_oob = mxc_nand_read_oob;
> -		}
> +		this->ecc.read_page = mxc_nand_read_page;
> +		this->ecc.read_page_raw = mxc_nand_read_page_raw;
> +		this->ecc.read_oob = mxc_nand_read_oob;
>  		this->ecc.calculate = mxc_nand_calculate_ecc;
>  		this->ecc.hwctl = mxc_nand_enable_hwecc;
>  		this->ecc.correct = host->devtype_data->correct_data;




More information about the linux-mtd mailing list