Patchwork oobavail size calculation error in nand_base.c

login
register
mail settings
Submitter Jinyoung Park
Date May 14, 2009, 8:28 a.m.
Message ID <280b69ac0905140128u135f8b2aua50114faebb593e0@mail.gmail.com>
Download mbox | patch
Permalink /patch/27202/
State New, archived
Headers show

Comments

Jinyoung Park - May 14, 2009, 8:28 a.m.
Hello.

I found error in nand_base.c.
In nand_scan_tail() function, oobavail size calculation result is
wrong as follow circumstances.
    - Define nand_ecclayout using full oobfree array like below.
	.oobfree = { {2, 6}, {11, 2}, {16, 8}, {27, 2}, {32, 8}, {43, 2},{48,
8}, {59, 2} }
    - Using ARM EABI cross compiler when kernel compile also enable
"Use the ARM EABI to compile the kernl" feature in kernel menuconfig.
      I using ARM EABI compiler that codesourcery release arm-2008q3-51.

In this case, right oobavail size is 38 but result is too much
value(e.g. 2703572308...). It's random.
I think sometimes happen that after beyond oobfree array(oobfree[8])
is not 0 when ARM EABI compilation.
Below codes are my modified code. Please check...

Thanks.

Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/
Artem Bityutskiy - May 15, 2009, 2:11 p.m.
On Thu, 2009-05-14 at 17:28 +0900, Jinyoung Park wrote:
> Hello.
> 
> I found error in nand_base.c.
> In nand_scan_tail() function, oobavail size calculation result is
> wrong as follow circumstances.
>     - Define nand_ecclayout using full oobfree array like below.
> 	.oobfree = { {2, 6}, {11, 2}, {16, 8}, {27, 2}, {32, 8}, {43, 2},{48,
> 8}, {59, 2} }
>     - Using ARM EABI cross compiler when kernel compile also enable
> "Use the ARM EABI to compile the kernl" feature in kernel menuconfig.
>       I using ARM EABI compiler that codesourcery release arm-2008q3-51.
> 
> In this case, right oobavail size is 38 but result is too much
> value(e.g. 2703572308...). It's random.
> I think sometimes happen that after beyond oobfree array(oobfree[8])
> is not 0 when ARM EABI compilation.
> Below codes are my modified code. Please check...

It seems the following patch which sits in my l2-mtd-2.6.git fixes
this as well: http://git.infradead.org/users/dedekind/l2-mtd-2.6.git

Author: David Brownell <dbrownell@users.sourceforge.net>
Date:   Tue Apr 21 19:51:20 2009 -0700

    MTD: NAND: don't walk past end of oobfree[]

    Resolve issue noted by Sneha:  when computing oobavail from
    the list of free areas in the OOB, don't assume there will
    always be an unused slot at the end.  With ECC_HW_SYNDROME
    and 4KB page chips, it's fairly likely there *won't* be one.

    Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
    Cc: "Narnakaje, Snehaprabha" <nsnehaprabha@ti.com>"
    Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>

Patch

diff --git a/drivers/mtd/nand/nand_base.c b/drivers/mtd/nand/nand_base.c
index 3d7ed43..abb5998 100644
--- a/drivers/mtd/nand/nand_base.c
+++ b/drivers/mtd/nand/nand_base.c
@@ -2756,9 +2756,12 @@  int nand_scan_tail(struct mtd_info *mtd)
	* the out of band area
	*/
	chip->ecc.layout->oobavail = 0;
-	for (i = 0; chip->ecc.layout->oobfree[i].length; i++)
+	for (i = 0; i < MTD_MAX_OOBFREE_ENTRIES; i++) {
+	if (chip->ecc.layout->oobfree[i].length == 0)
+		break;
	chip->ecc.layout->oobavail +=
		chip->ecc.layout->oobfree[i].length;
+	}
	mtd->oobavail = chip->ecc.layout->oobavail;

______________________________________________________