diff mbox series

Enable the read ECC before program the page

Message ID 59b97a4e.cf02b00a.d8b82.a310@mx.google.com
State Superseded
Headers show
Series Enable the read ECC before program the page | expand

Commit Message

Arun Nagendran Sept. 13, 2017, 6:34 p.m. UTC 
Current program_page function did following operation:

1. read page (with ECC OFF)
2. modify the page
3. write the page (with ECC ON)

For some case(buggy flash Chip), while read the page without ECC ON,
we may read the page with bit flip error and modify that bad page without
knowing the bit flip error on that page.
also we re-calculate the hash for bad page and write it.
This could bring potential in-consistency problem with Flash data.

Verify this logic with GIGA DEVICE Part(GD5F2GQ4RCFIG):
we see this in-conststency problem wit Giga Device and fix on
this patch resovle that issue.
---
 drivers/staging/mt29f_spinand/mt29f_spinand.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

Comments

gregkh@linuxfoundation.org Sept. 13, 2017, 7:04 p.m. UTC | #1 
On Wed, Sep 13, 2017 at 02:34:53PM -0400, Arun Nagendran wrote:
> Current program_page function did following operation:
> 
> 1. read page (with ECC OFF)
> 2. modify the page
> 3. write the page (with ECC ON)
> 
> For some case(buggy flash Chip), while read the page without ECC ON,
> we may read the page with bit flip error and modify that bad page without
> knowing the bit flip error on that page.
> also we re-calculate the hash for bad page and write it.
> This could bring potential in-consistency problem with Flash data.
> 
> Verify this logic with GIGA DEVICE Part(GD5F2GQ4RCFIG):
> we see this in-conststency problem wit Giga Device and fix on
> this patch resovle that issue.
> ---
>  drivers/staging/mt29f_spinand/mt29f_spinand.c | 9 +++++++--
>  1 file changed, 7 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/staging/mt29f_spinand/mt29f_spinand.c b/drivers/staging/mt29f_spinand/mt29f_spinand.c
> index a4e3ae8..202d222 100644
> --- a/drivers/staging/mt29f_spinand/mt29f_spinand.c
> +++ b/drivers/staging/mt29f_spinand/mt29f_spinand.c
> @@ -496,8 +496,13 @@ static int spinand_program_page(struct spi_device *spi_nand,
>  	if (!wbuf)
>  		return -ENOMEM;
>  
> -	enable_read_hw_ecc = 0;
> -	spinand_read_page(spi_nand, page_id, 0, CACHE_BUF, wbuf);
> +	enable_read_hw_ecc = 1;
> +	retval = spinand_read_page(spi_nand, page_id, 0, CACHE_BUF, wbuf);
> +	if( retval < 0 )
> +	{
> +		dev_err(&spi_nand->dev, "ecc error on read page!!!\n");
> +		return retval;
> +	}

Always use scripts/checkpatch.pl so you don't get grumpy maintainers
telling you to use scripts/checkpatch.pl :(
diff mbox series

Patch

diff --git a/drivers/staging/mt29f_spinand/mt29f_spinand.c b/drivers/staging/mt29f_spinand/mt29f_spinand.c
index a4e3ae8..202d222 100644
--- a/drivers/staging/mt29f_spinand/mt29f_spinand.c
+++ b/drivers/staging/mt29f_spinand/mt29f_spinand.c
@@ -496,8 +496,13 @@  static int spinand_program_page(struct spi_device *spi_nand,
 	if (!wbuf)
 		return -ENOMEM;
 
-	enable_read_hw_ecc = 0;
-	spinand_read_page(spi_nand, page_id, 0, CACHE_BUF, wbuf);
+	enable_read_hw_ecc = 1;
+	retval = spinand_read_page(spi_nand, page_id, 0, CACHE_BUF, wbuf);
+	if( retval < 0 )
+	{
+		dev_err(&spi_nand->dev, "ecc error on read page!!!\n");
+		return retval;
+	}
 
 	for (i = offset, j = 0; i < len; i++, j++)
 		wbuf[i] &= buf[j];