Patchwork mtd: nand: Fix regression in BBM detection

login
register
mail settings
Submitter Brian Norris
Date Aug. 18, 2010, 6:25 p.m.
Message ID <1282155904-10113-1-git-send-email-norris@broadcom.com>
Download mbox | patch
Permalink /patch/62077/
State New
Headers show

Comments

Brian Norris - Aug. 18, 2010, 6:25 p.m.
Commit c7b28e25cb9beb943aead770ff14551b55fa8c79 caused a regression
in detection of factory-set bad block markers, especially for certain
small-page NAND. This fix removes some unneeded constraints on using
NAND_SMALL_BADBLOCK_POS, making the detection code more correct.

This regression can be seen, for example, in Hynix HY27US081G1M and
similar.

Signed-off-by: Brian Norris <norris@broadcom.com>
---
 drivers/mtd/nand/nand_base.c |   10 +++-------
 1 files changed, 3 insertions(+), 7 deletions(-)
Abdoulaye Walsimou GAYE - Aug. 18, 2010, 7:30 p.m.
On 08/18/2010 08:25 PM, Brian Norris wrote:
> Commit c7b28e25cb9beb943aead770ff14551b55fa8c79 caused a regression
> in detection of factory-set bad block markers, especially for certain
> small-page NAND. This fix removes some unneeded constraints on using
> NAND_SMALL_BADBLOCK_POS, making the detection code more correct.
>
> This regression can be seen, for example, in Hynix HY27US081G1M and
> similar.
>
> Signed-off-by: Brian Norris<norris@broadcom.com>
> ---
>   drivers/mtd/nand/nand_base.c |   10 +++-------
>   1 files changed, 3 insertions(+), 7 deletions(-)
>
> diff --git a/drivers/mtd/nand/nand_base.c b/drivers/mtd/nand/nand_base.c
> index a3c7473..a22ed7b 100644
> --- a/drivers/mtd/nand/nand_base.c
> +++ b/drivers/mtd/nand/nand_base.c
> @@ -2934,14 +2934,10 @@ static struct nand_flash_dev *nand_get_flash_type(struct mtd_info *mtd,
>   		chip->chip_shift = ffs((unsigned)(chip->chipsize>>  32)) + 32 - 1;
>
>   	/* Set the bad block position */
> -	if (!(busw&  NAND_BUSWIDTH_16)&&  (*maf_id == NAND_MFR_STMICRO ||
> -				(*maf_id == NAND_MFR_SAMSUNG&&
> -				 mtd->writesize == 512) ||
> -				*maf_id == NAND_MFR_AMD))
> -		chip->badblockpos = NAND_SMALL_BADBLOCK_POS;
> -	else
> +	if (mtd->writesize>  512 || (busw&  NAND_BUSWIDTH_16))
>   		chip->badblockpos = NAND_LARGE_BADBLOCK_POS;
> -
> +	else
> +		chip->badblockpos = NAND_SMALL_BADBLOCK_POS;
>
>   	/* Get chip options, preserve non chip based options */
>   	chip->options&= ~NAND_CHIPOPTIONS_MSK;
>    

Brian,
Sorry for the long delay!
I tested the above patch unfortunately it does not help in my case!
And when I go further and put a JFFS2 in that partition and boot the
board I have

(S3c2410 nand hardware ECC enable):
mtd->read(0x400 bytes from 0x1274000) returned ECC error
mtd->read(0x3c08 bytes from 0x12743f8) returned ECC error

(without S3c2410 nand hardware ECC enable):
uncorrectable error :
uncorrectable error :
uncorrectable error :
uncorrectable error :
mtd->read(0x400 bytes from 0x1274000) returned ECC error
uncorrectable error :
uncorrectable error :
uncorrectable error :
[...]
uncorrectable error :
uncorrectable error :
mtd->read(0x3c08 bytes from 0x12743f8) returned ECC error

Despite these errors I can actually use the board (no kernel panic)!
The part is Samsung K9F1208U0C - PCB0

Hope that helps!

Thanks,
AWG
Brian Norris - Aug. 19, 2010, 12:04 a.m.
On 08/18/2010 12:30 PM, Abdoulaye Walsimou GAYE wrote:
> Brian,
> Sorry for the long delay!
> I tested the above patch unfortunately it does not help in my case!

Understood. That makes sense. In fact, your problem is most likely *not* 
related to this commit. As I mentioned before, please try narrowing down 
what specifically caused this; if I read correctly, you jumped from 
2.6.33 to 2.6.36-rc1. There have been several important changes between 
those releases. Notably, this commit may be giving Samsung chips problems:
426c457a3216fac74e

This thread is covering a few problems with Samsung:
http://lists.infradead.org/pipermail/linux-mtd/2010-August/031590.html

> And when I go further and put a JFFS2 in that partition and boot the
> board I have
<snip>
> Despite these errors I can actually use the board (no kernel panic)!
> The part is Samsung K9F1208U0C - PCB0

Unless you really know what you're doing, I wouldn't be writing/erasing 
the flash if it's not detecting bad blocks properly.

Let me know if you have trouble with narrowing down to the problem commit.

Brian

Patch

diff --git a/drivers/mtd/nand/nand_base.c b/drivers/mtd/nand/nand_base.c
index a3c7473..a22ed7b 100644
--- a/drivers/mtd/nand/nand_base.c
+++ b/drivers/mtd/nand/nand_base.c
@@ -2934,14 +2934,10 @@  static struct nand_flash_dev *nand_get_flash_type(struct mtd_info *mtd,
 		chip->chip_shift = ffs((unsigned)(chip->chipsize >> 32)) + 32 - 1;
 
 	/* Set the bad block position */
-	if (!(busw & NAND_BUSWIDTH_16) && (*maf_id == NAND_MFR_STMICRO ||
-				(*maf_id == NAND_MFR_SAMSUNG &&
-				 mtd->writesize == 512) ||
-				*maf_id == NAND_MFR_AMD))
-		chip->badblockpos = NAND_SMALL_BADBLOCK_POS;
-	else
+	if (mtd->writesize > 512 || (busw & NAND_BUSWIDTH_16))
 		chip->badblockpos = NAND_LARGE_BADBLOCK_POS;
-
+	else
+		chip->badblockpos = NAND_SMALL_BADBLOCK_POS;
 
 	/* Get chip options, preserve non chip based options */
 	chip->options &= ~NAND_CHIPOPTIONS_MSK;