diff mbox

[v2] gpmi-nand: Handle ECC Errors in erased pages

Message ID 1461587712-16403-1-git-send-email-mpa@pengutronix.de (mailing list archive)
State New, archived
Headers show

Commit Message

Markus Pargmann April 25, 2016, 12:35 p.m. UTC
ECC is only calculated for written pages. As erased pages are not
actively written the ECC is always invalid. For this purpose the
Hardware BCH unit is able to check for erased pages and does not raise
an ECC error in this case. This behaviour can be influenced using the
BCH_MODE register which sets the number of allowed bitflips in an erased
page. Unfortunately the unit is not capable of fixing the bitflips in
memory.

To avoid complete software checks for erased pages, we can simply check
buffers with uncorrectable ECC errors because we know that any erased
page with errors is uncorrectable by the BCH unit.

This patch adds the generic nand_check_erased_ecc_chunk() to gpmi-nand
to correct erased pages. To have the valid data in the buffer before
using them, this patch moves the read_page_swap_end() call before the
ECC status checking for-loop.

Signed-off-by: Markus Pargmann <mpa@pengutronix.de>
[Squashed patches by Stefan and Boris to check ECC area]
Cc: Stefan Christ <s.christ@phytec.de>
Cc: Boris Brezillon <boris.brezillon@free-electrons.com>

Signed-off-by: Markus Pargmann <mpa@pengutronix.de>
---
Hi,

thanks for the patches Stefan and Boris, I squashed it into this patch.

Best Regards,

Markus


 drivers/mtd/nand/gpmi-nand/gpmi-nand.c | 78 +++++++++++++++++++++++++++++++---
 1 file changed, 73 insertions(+), 5 deletions(-)

Comments

Boris BREZILLON April 25, 2016, 12:53 p.m. UTC | #1
Hi Markus,

On Mon, 25 Apr 2016 14:35:12 +0200
Markus Pargmann <mpa@pengutronix.de> wrote:

> ECC is only calculated for written pages. As erased pages are not
> actively written the ECC is always invalid. For this purpose the
> Hardware BCH unit is able to check for erased pages and does not raise
> an ECC error in this case. This behaviour can be influenced using the
> BCH_MODE register which sets the number of allowed bitflips in an erased
> page. Unfortunately the unit is not capable of fixing the bitflips in
> memory.
> 
> To avoid complete software checks for erased pages, we can simply check
> buffers with uncorrectable ECC errors because we know that any erased
> page with errors is uncorrectable by the BCH unit.
> 
> This patch adds the generic nand_check_erased_ecc_chunk() to gpmi-nand
> to correct erased pages. To have the valid data in the buffer before
> using them, this patch moves the read_page_swap_end() call before the
> ECC status checking for-loop.
> 
> Signed-off-by: Markus Pargmann <mpa@pengutronix.de>
> [Squashed patches by Stefan and Boris to check ECC area]
> Cc: Stefan Christ <s.christ@phytec.de>
> Cc: Boris Brezillon <boris.brezillon@free-electrons.com>
> 
> Signed-off-by: Markus Pargmann <mpa@pengutronix.de>

Duplicate SoB (no need to resend, I'll drop it while applying the
patch).

Otherwise, it looks good to me. Stefan, could you add your Tested-by,
and I'd also like to have an Ack from Han.

Thanks,

Boris

> ---
> Hi,
> 
> thanks for the patches Stefan and Boris, I squashed it into this patch.
> 
> Best Regards,
> 
> Markus
> 
> 
>  drivers/mtd/nand/gpmi-nand/gpmi-nand.c | 78 +++++++++++++++++++++++++++++++---
>  1 file changed, 73 insertions(+), 5 deletions(-)
> 
> diff --git a/drivers/mtd/nand/gpmi-nand/gpmi-nand.c b/drivers/mtd/nand/gpmi-nand/gpmi-nand.c
> index 235ddcb58f39..71ee1696bcd1 100644
> --- a/drivers/mtd/nand/gpmi-nand/gpmi-nand.c
> +++ b/drivers/mtd/nand/gpmi-nand/gpmi-nand.c
> @@ -1035,14 +1035,87 @@ static int gpmi_ecc_read_page(struct mtd_info *mtd, struct nand_chip *chip,
>  	/* Loop over status bytes, accumulating ECC status. */
>  	status = auxiliary_virt + nfc_geo->auxiliary_status_offset;
>  
> +	read_page_swap_end(this, buf, nfc_geo->payload_size,
> +			   this->payload_virt, this->payload_phys,
> +			   nfc_geo->payload_size,
> +			   payload_virt, payload_phys);
> +
>  	for (i = 0; i < nfc_geo->ecc_chunk_count; i++, status++) {
>  		if ((*status == STATUS_GOOD) || (*status == STATUS_ERASED))
>  			continue;
>  
>  		if (*status == STATUS_UNCORRECTABLE) {
> +			int eccbits = nfc_geo->ecc_strength * nfc_geo->gf_len;
> +			u8 *eccbuf = this->raw_buffer;
> +			int offset, bitoffset;
> +			int eccbytes;
> +			int flips;
> +
> +			/* Read ECC bytes into our internal raw_buffer */
> +			offset = nfc_geo->metadata_size * 8;
> +			offset += ((8 * nfc_geo->ecc_chunk_size) + eccbits) * (i + 1);
> +			offset -= eccbits;
> +			bitoffset = offset % 8;
> +			eccbytes = DIV_ROUND_UP(offset + eccbits, 8);
> +			offset /= 8;
> +			eccbytes -= offset;
> +			chip->cmdfunc(mtd, NAND_CMD_RNDOUT, offset, -1);
> +			chip->read_buf(mtd, eccbuf, eccbytes);
> +
> +			/*
> +			 * ECC data are not byte aligned and we may have
> +			 * in-band data in the first and last byte of
> +			 * eccbuf. Set non-eccbits to one so that
> +			 * nand_check_erased_ecc_chunk() does not count them
> +			 * as bitflips.
> +			 */
> +			if (bitoffset)
> +				eccbuf[0] |= GENMASK(bitoffset - 1, 0);
> +
> +			bitoffset = (bitoffset + eccbits) % 8;
> +			if (bitoffset)
> +				eccbuf[eccbytes - 1] |= GENMASK(7, bitoffset);
> +
> +			/*
> +			 * The ECC hardware has an uncorrectable ECC status
> +			 * code in case we have bitflips in an erased page. As
> +			 * nothing was written into this subpage the ECC is
> +			 * obviously wrong and we can not trust it. We assume
> +			 * at this point that we are reading an erased page and
> +			 * try to correct the bitflips in buffer up to
> +			 * ecc_strength bitflips. If this is a page with random
> +			 * data, we exceed this number of bitflips and have a
> +			 * ECC failure. Otherwise we use the corrected buffer.
> +			 */
> +			if (i == 0) {
> +				/* The first block includes metadata */
> +				flips = nand_check_erased_ecc_chunk(
> +						buf + i * nfc_geo->ecc_chunk_size,
> +						nfc_geo->ecc_chunk_size,
> +						eccbuf, eccbytes,
> +						auxiliary_virt,
> +						nfc_geo->metadata_size,
> +						nfc_geo->ecc_strength);
> +			} else {
> +				flips = nand_check_erased_ecc_chunk(
> +						buf + i * nfc_geo->ecc_chunk_size,
> +						nfc_geo->ecc_chunk_size,
> +						eccbuf, eccbytes,
> +						NULL, 0,
> +						nfc_geo->ecc_strength);
> +			}
> +
> +			if (flips > 0) {
> +				max_bitflips = max_t(unsigned int, max_bitflips,
> +						     flips);
> +				mtd->ecc_stats.corrected += flips;
> +				continue;
> +			}
> +
>  			mtd->ecc_stats.failed++;
>  			continue;
>  		}
> +
>  		mtd->ecc_stats.corrected += *status;
>  		max_bitflips = max_t(unsigned int, max_bitflips, *status);
>  	}
> @@ -1062,11 +1135,6 @@ static int gpmi_ecc_read_page(struct mtd_info *mtd, struct nand_chip *chip,
>  		chip->oob_poi[0] = ((uint8_t *) auxiliary_virt)[0];
>  	}
>  
> -	read_page_swap_end(this, buf, nfc_geo->payload_size,
> -			this->payload_virt, this->payload_phys,
> -			nfc_geo->payload_size,
> -			payload_virt, payload_phys);
> -
>  	return max_bitflips;
>  }
>
Stefan Christ April 26, 2016, 7:51 a.m. UTC | #2
Hi,

On Mon, Apr 25, 2016 at 02:35:12PM +0200, Markus Pargmann wrote:
> ECC is only calculated for written pages. As erased pages are not
> actively written the ECC is always invalid. For this purpose the
> Hardware BCH unit is able to check for erased pages and does not raise
> an ECC error in this case. This behaviour can be influenced using the
> BCH_MODE register which sets the number of allowed bitflips in an erased
> page. Unfortunately the unit is not capable of fixing the bitflips in
> memory.
> 
> To avoid complete software checks for erased pages, we can simply check
> buffers with uncorrectable ECC errors because we know that any erased
> page with errors is uncorrectable by the BCH unit.
> 
> This patch adds the generic nand_check_erased_ecc_chunk() to gpmi-nand
> to correct erased pages. To have the valid data in the buffer before
> using them, this patch moves the read_page_swap_end() call before the
> ECC status checking for-loop.
> 
> Signed-off-by: Markus Pargmann <mpa@pengutronix.de>
> [Squashed patches by Stefan and Boris to check ECC area]
> Cc: Stefan Christ <s.christ@phytec.de>
> Cc: Boris Brezillon <boris.brezillon@free-electrons.com>
> 
> Signed-off-by: Markus Pargmann <mpa@pengutronix.de>

Tested-by: Stefan Christ <s.christ@phytec.de>

I verified the fix again on our board.

Mit freundlichen Grüßen / Kind regards,
	Stefan Christ
Han Xu April 26, 2016, 3:14 p.m. UTC | #3

Boris BREZILLON April 26, 2016, 5:30 p.m. UTC | #4
On Mon, 25 Apr 2016 14:35:12 +0200
Markus Pargmann <mpa@pengutronix.de> wrote:

> ECC is only calculated for written pages. As erased pages are not
> actively written the ECC is always invalid. For this purpose the
> Hardware BCH unit is able to check for erased pages and does not raise
> an ECC error in this case. This behaviour can be influenced using the
> BCH_MODE register which sets the number of allowed bitflips in an erased
> page. Unfortunately the unit is not capable of fixing the bitflips in
> memory.
> 
> To avoid complete software checks for erased pages, we can simply check
> buffers with uncorrectable ECC errors because we know that any erased
> page with errors is uncorrectable by the BCH unit.
> 
> This patch adds the generic nand_check_erased_ecc_chunk() to gpmi-nand
> to correct erased pages. To have the valid data in the buffer before
> using them, this patch moves the read_page_swap_end() call before the
> ECC status checking for-loop.
> 
> Signed-off-by: Markus Pargmann <mpa@pengutronix.de>
> [Squashed patches by Stefan and Boris to check ECC area]
> Cc: Stefan Christ <s.christ@phytec.de>
> Cc: Boris Brezillon <boris.brezillon@free-electrons.com>
> 
> Signed-off-by: Markus Pargmann <mpa@pengutronix.de>
> ---

Applied, thanks.

Boris
Markus Pargmann April 28, 2016, 8:17 a.m. UTC | #5
On Tuesday 26 April 2016 19:30:02 Boris Brezillon wrote:
> On Mon, 25 Apr 2016 14:35:12 +0200
> Markus Pargmann <mpa@pengutronix.de> wrote:
> 
> > ECC is only calculated for written pages. As erased pages are not
> > actively written the ECC is always invalid. For this purpose the
> > Hardware BCH unit is able to check for erased pages and does not raise
> > an ECC error in this case. This behaviour can be influenced using the
> > BCH_MODE register which sets the number of allowed bitflips in an erased
> > page. Unfortunately the unit is not capable of fixing the bitflips in
> > memory.
> > 
> > To avoid complete software checks for erased pages, we can simply check
> > buffers with uncorrectable ECC errors because we know that any erased
> > page with errors is uncorrectable by the BCH unit.
> > 
> > This patch adds the generic nand_check_erased_ecc_chunk() to gpmi-nand
> > to correct erased pages. To have the valid data in the buffer before
> > using them, this patch moves the read_page_swap_end() call before the
> > ECC status checking for-loop.
> > 
> > Signed-off-by: Markus Pargmann <mpa@pengutronix.de>
> > [Squashed patches by Stefan and Boris to check ECC area]
> > Cc: Stefan Christ <s.christ@phytec.de>
> > Cc: Boris Brezillon <boris.brezillon@free-electrons.com>
> > 
> > Signed-off-by: Markus Pargmann <mpa@pengutronix.de>
> > ---
> 
> Applied, thanks.

Great, thanks all of you.

Best Regards,

Markus
diff mbox

Patch

diff --git a/drivers/mtd/nand/gpmi-nand/gpmi-nand.c b/drivers/mtd/nand/gpmi-nand/gpmi-nand.c
index 235ddcb58f39..71ee1696bcd1 100644
--- a/drivers/mtd/nand/gpmi-nand/gpmi-nand.c
+++ b/drivers/mtd/nand/gpmi-nand/gpmi-nand.c
@@ -1035,14 +1035,87 @@  static int gpmi_ecc_read_page(struct mtd_info *mtd, struct nand_chip *chip,
 	/* Loop over status bytes, accumulating ECC status. */
 	status = auxiliary_virt + nfc_geo->auxiliary_status_offset;
 
+	read_page_swap_end(this, buf, nfc_geo->payload_size,
+			   this->payload_virt, this->payload_phys,
+			   nfc_geo->payload_size,
+			   payload_virt, payload_phys);
+
 	for (i = 0; i < nfc_geo->ecc_chunk_count; i++, status++) {
 		if ((*status == STATUS_GOOD) || (*status == STATUS_ERASED))
 			continue;
 
 		if (*status == STATUS_UNCORRECTABLE) {
+			int eccbits = nfc_geo->ecc_strength * nfc_geo->gf_len;
+			u8 *eccbuf = this->raw_buffer;
+			int offset, bitoffset;
+			int eccbytes;
+			int flips;
+
+			/* Read ECC bytes into our internal raw_buffer */
+			offset = nfc_geo->metadata_size * 8;
+			offset += ((8 * nfc_geo->ecc_chunk_size) + eccbits) * (i + 1);
+			offset -= eccbits;
+			bitoffset = offset % 8;
+			eccbytes = DIV_ROUND_UP(offset + eccbits, 8);
+			offset /= 8;
+			eccbytes -= offset;
+			chip->cmdfunc(mtd, NAND_CMD_RNDOUT, offset, -1);
+			chip->read_buf(mtd, eccbuf, eccbytes);
+
+			/*
+			 * ECC data are not byte aligned and we may have
+			 * in-band data in the first and last byte of
+			 * eccbuf. Set non-eccbits to one so that
+			 * nand_check_erased_ecc_chunk() does not count them
+			 * as bitflips.
+			 */
+			if (bitoffset)
+				eccbuf[0] |= GENMASK(bitoffset - 1, 0);
+
+			bitoffset = (bitoffset + eccbits) % 8;
+			if (bitoffset)
+				eccbuf[eccbytes - 1] |= GENMASK(7, bitoffset);
+
+			/*
+			 * The ECC hardware has an uncorrectable ECC status
+			 * code in case we have bitflips in an erased page. As
+			 * nothing was written into this subpage the ECC is
+			 * obviously wrong and we can not trust it. We assume
+			 * at this point that we are reading an erased page and
+			 * try to correct the bitflips in buffer up to
+			 * ecc_strength bitflips. If this is a page with random
+			 * data, we exceed this number of bitflips and have a
+			 * ECC failure. Otherwise we use the corrected buffer.
+			 */
+			if (i == 0) {
+				/* The first block includes metadata */
+				flips = nand_check_erased_ecc_chunk(
+						buf + i * nfc_geo->ecc_chunk_size,
+						nfc_geo->ecc_chunk_size,
+						eccbuf, eccbytes,
+						auxiliary_virt,
+						nfc_geo->metadata_size,
+						nfc_geo->ecc_strength);
+			} else {
+				flips = nand_check_erased_ecc_chunk(
+						buf + i * nfc_geo->ecc_chunk_size,
+						nfc_geo->ecc_chunk_size,
+						eccbuf, eccbytes,
+						NULL, 0,
+						nfc_geo->ecc_strength);
+			}
+
+			if (flips > 0) {
+				max_bitflips = max_t(unsigned int, max_bitflips,
+						     flips);
+				mtd->ecc_stats.corrected += flips;
+				continue;
+			}
+
 			mtd->ecc_stats.failed++;
 			continue;
 		}
+
 		mtd->ecc_stats.corrected += *status;
 		max_bitflips = max_t(unsigned int, max_bitflips, *status);
 	}
@@ -1062,11 +1135,6 @@  static int gpmi_ecc_read_page(struct mtd_info *mtd, struct nand_chip *chip,
 		chip->oob_poi[0] = ((uint8_t *) auxiliary_virt)[0];
 	}
 
-	read_page_swap_end(this, buf, nfc_geo->payload_size,
-			this->payload_virt, this->payload_phys,
-			nfc_geo->payload_size,
-			payload_virt, payload_phys);
-
 	return max_bitflips;
 }