[U-Boot] jffs2: ECC corrected nand data is ignored

Message ID ff0bd6f4bed7419382eb2e4b4342e482@ise-exch13.ise.de
State Accepted
Delegated to: Tom Rini
Headers show
Series
  • [U-Boot] jffs2: ECC corrected nand data is ignored
Related show

Commit Message

Engling, Uwe Oct. 10, 2017, 2:20 p.m.
Hello,

I ran into a problem with the JFFS2 filesystem driver implemented in U-Boot.

I've got a NAND device that has correctable ECC errors (corrected somewhere in mtd/nand/nand_base.c).
The NAND driver tells the filesystem layer (jffs2_1pass.c) above that there occurred correctable ECC errors and returns with a "value > 0".
The JFFS2 driver recognizes the corrected ECC errors as real error and skips this block because the only accepts a "return value == 0" as correct.

This problem exists for over 8 years (I checked version 2010.09) so I'm a little bit worried that I interpreted something wrong or didn't get the whole context.

Can someone confirm this bug (and the bugfix) in the u-boot jffs2 driver?

There was a mail in 2012 that mentioned the same problem, but there was no patch:
http://u-boot.10912.n7.nabble.com/JFFS2-seems-to-drop-nand-data-with-ECC-corrections-td142008.html
Sometime after this discussion the return value of nand_read() changed from -EUCLEAN as correctable ECC error to a positive value with the count of ECC corrected errors.




With kind reguards,
Uwe Engling

Comments

Tom Rini Oct. 17, 2017, 12:47 a.m. | #1
On Tue, Oct 10, 2017 at 02:20:55PM +0000, Engling, Uwe wrote:

> Hello,
> 
> I ran into a problem with the JFFS2 filesystem driver implemented in U-Boot.
> 
> I've got a NAND device that has correctable ECC errors (corrected somewhere in mtd/nand/nand_base.c).
> The NAND driver tells the filesystem layer (jffs2_1pass.c) above that there occurred correctable ECC errors and returns with a "value > 0".
> The JFFS2 driver recognizes the corrected ECC errors as real error and skips this block because the only accepts a "return value == 0" as correct.
> 
> This problem exists for over 8 years (I checked version 2010.09) so I'm a little bit worried that I interpreted something wrong or didn't get the whole context.
> 
> Can someone confirm this bug (and the bugfix) in the u-boot jffs2 driver?
> 
> There was a mail in 2012 that mentioned the same problem, but there was no patch:
> http://u-boot.10912.n7.nabble.com/JFFS2-seems-to-drop-nand-data-with-ECC-corrections-td142008.html
> Sometime after this discussion the return value of nand_read() changed from -EUCLEAN as correctable ECC error to a positive value with the count of ECC corrected errors.
> 
> 
> 
> 
> With kind reguards,
> Uwe Engling
> 
> diff --git a/fs/jffs2/jffs2_1pass.c b/fs/jffs2/jffs2_1pass.c
> index 4c6dfbf..6bf1943 100644

Applied to u-boot/master, thanks!

Patch

diff --git a/fs/jffs2/jffs2_1pass.c b/fs/jffs2/jffs2_1pass.c
index 4c6dfbf..6bf1943 100644
--- a/fs/jffs2/jffs2_1pass.c
+++ b/fs/jffs2/jffs2_1pass.c
@@ -201,7 +201,7 @@  static int read_nand_cached(u32 off, u32 size, u_char *buf)
 
 			retlen = NAND_CACHE_SIZE;
 			if (nand_read(mtd, nand_cache_off,
-				      &retlen, nand_cache) != 0 ||
+				      &retlen, nand_cache) < 0 ||
 					retlen != NAND_CACHE_SIZE) {
 				printf("read_nand_cached: error reading nand off %#x size %d bytes\n",
 						nand_cache_off, NAND_CACHE_SIZE);
@@ -300,7 +300,7 @@  static int read_onenand_cached(u32 off, u32 size, u_char *buf)
 
 			retlen = ONENAND_CACHE_SIZE;
 			if (onenand_read(&onenand_mtd, onenand_cache_off, retlen,
-						&retlen, onenand_cache) != 0 ||
+						&retlen, onenand_cache) < 0 ||
 					retlen != ONENAND_CACHE_SIZE) {
 				printf("read_onenand_cached: error reading nand off %#x size %d bytes\n",
 					onenand_cache_off, ONENAND_CACHE_SIZE);