Re: [PATCH v2] mtd: nand: support for Toshiba BENAND (Built-in ECC NAND)

From: KOBAYASHI Yoshitake
Date: Fri Jun 19 2015 - 10:47:47 EST

Please see my patches how to deal with that. Maybe you can find a better solution.

Thank you for your suggestion. I considered to use your submitted patch on BENAND
driver, but I believe reading twice the same page approach will affect read performance.
Additionally, BENAND does not support Disable ECC. So, I cannot use same approach.

I consider other solutions.

BENAND Read Status CMD (70h) can report Rewrite Recommend. BENAND also has extended
ECC Status CMD (7Ah). This can report number of bit error / sector data.
If I can use an extended ECC Status CMD (7Ah) in BENAND driver (nand_benand.c),
I suggest to implement it as the follows:

+++ b/drivers/mtd/nand/nand_benand.c

+/* ECC Status Read Command */
+ chip->ecc.strength = 8;
+ /* correctable */
+ else if (status & NAND_STATUS_RECOM_REWRT) {
+ if (chip->cmd_ctrl &&
+ /* Check Read ECC Status */
+ chip->cmd_ctrl(mtd, NAND_CMD_ECC_STATUS,
+ ecc_status = chip->read_byte(mtd);
+ bitflips = ecc_status & 0xff;
+ mtd->ecc_stats.corrected += bitflips;

If above implementation cannot be acceptable, I would like to use "mtd->bitflip_threshold"
to inform number of bitflip, as the follows:

+ } else {
+ /*
+ * If can't use chip->cmd_ctrl,
+ * we can't get real number of bitflips.
+ * So, we set bitflips mtd->bitflip_threshold.
+ */
+ bitflips = mtd->bitflip_threshold;
+ mtd->ecc_stats.corrected += bitflips;
+ }
+ }
+ return bitflips;

-- Yoshi

On 2015/06/16 22:53, Richard Weinberger wrote:
Am 16.06.2015 um 15:33 schrieb KOBAYASHI Yoshitake:
+ /* correctable */
+ else if (status & NAND_STATUS_RECOM_REWRT) {
+ pr_info("BENAND : Recommended to rewrite!\n");
+ bitflips = chip->ecc.strength;

In your case this might be okay, as you set strength to 1.
Otherweise you'd have to report the real number of bitflips.

I also thought it is okay in this case.
BENAND return corrected data to Host NAND Controller till uncorrectable status.
The current patch uses this Read Status command 70h to abstract BENAND Multi
bit ECC and Need to Rewrite judgement so BENAND would look like 1bit ECC device.

The layers above MTD need to know how many bits got repaired.
It seems like BENAND suffers from the same shortcomings than On-Die-ECC. ;-\
Please see my patches how to deal with that. Maybe you can find a better solution.


To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
Please read the FAQ at