Benchmark errors

Benchmark errors

[Lorenzo PARISI]

Hi all,

when I benchmark the NAND with Bonnie++, I receive many 
errors:

nand_write_oob: Failed write verify, page 0x0....
     :                  :                :
     :                  :                :

and Bonnie++ end with this messages:

Bonnie: drastic I/O error (rmdir): Directory not empty
Cleaning up test directory after error
Bonnie: drastic I/O error (rmdir): No such file or directory

Initially, I thinked that a problem of Bonnie++, but if I 
run Bonnie++ on a NAND with JFFS2, I don't receive any 
errors.

Why?

Thanks.

Re: Benchmark errors

[Ben Dooks]

On Fri, Oct 08, 2004 at 11:52:23AM +0200, Lorenzo PARISI wrote:
> Hi all,
> 
> when I benchmark the NAND with Bonnie++, I receive many 
> errors:
> 
> nand_write_oob: Failed write verify, page 0x0....
>      :                  :                :
>      :                  :                :
> 
> and Bonnie++ end with this messages:
> 
> Bonnie: drastic I/O error (rmdir): Directory not empty
> Cleaning up test directory after error
> Bonnie: drastic I/O error (rmdir): No such file or directory
> 
> Initially, I thinked that a problem of Bonnie++, but if I 
> run Bonnie++ on a NAND with JFFS2, I don't receive any 
> errors.

erm, what filesystem are you using the gives errors?

NAND is not guaranteed to be defectless, the fs must
be able to deal with bad blocks in the NAND.

-- 
Ben (ben@fluff.org, http://www.fluff.org/)

  'a smiley only costs 4 bytes'

Re: Benchmark errors

[Thomas Gleixner]

On Fri, 2004-10-08 at 11:52, Lorenzo PARISI wrote:
> Hi all,
> nand_write_oob: Failed write verify, page 0x0....

So it works with JFFS2 and it fails on YAFFS. The MTD internal interface
was reworked some time ago and it might be that the write verify does
not work with YAFFS anymore, as YAFFS is using a legacy variant of the
interface. Can you please switch of the write verify and test again ?

tglx

Re: Benchmark errors

[Thomas Gleixner]

[-- Attachment #1: Type: text/plain, Size: 584 bytes --]

On Fri, 2004-10-08 at 12:00, Thomas Gleixner wrote:
> On Fri, 2004-10-08 at 11:52, Lorenzo PARISI wrote:
> > Hi all,
> > nand_write_oob: Failed write verify, page 0x0....
> 
> So it works with JFFS2 and it fails on YAFFS. The MTD internal interface
> was reworked some time ago and it might be that the write verify does
> not work with YAFFS anymore, as YAFFS is using a legacy variant of the
> interface. Can you please switch of the write verify and test again ?
> 
Can you please figure out, where the verify exactly fails ?
Please apply the attached patch to nand_base.c

tglx



[-- Attachment #2: nand_base.diff --]
[-- Type: text/x-patch, Size: 890 bytes --]

diff -u -r1.121 nand_base.c
--- nand_base.c	6 Oct 2004 19:53:11 -0000	1.121
+++ nand_base.c	8 Oct 2004 10:34:39 -0000
@@ -310,11 +310,17 @@
 {
 	int i;
 	struct nand_chip *this = mtd->priv;
+	u_char tmp;
 
-	for (i=0; i<len; i++)
-		if (buf[i] != readb(this->IO_ADDR_R))
+	for (i=0; i<len; i++) {
+		tmp = readb(this->IO_ADDR_R);
+		if (buf[i] != tmp) {
+			DEBUG (MTD_DEBUG_LEVEL0, 
+			       "Verify failed: 0x%02x instead of 0x%02x at offset %d\n",
+			       tmp, buf[i], i);
 			return -EFAULT;
-
+		}
+	}
 	return 0;
 }
 
@@ -955,7 +961,7 @@
 		 */
 		if (oobmode) {
 			if (this->verify_buf(mtd, &oob_buf[oobofs], mtd->oobsize - hweccbytes * eccsteps)) {
-				DEBUG (MTD_DEBUG_LEVEL0, "%s: " "Failed write verify, page 0x%08x ", __FUNCTION__, page);
+				DEBUG (MTD_DEBUG_LEVEL0, "%s: " "Failed write verify oob, page 0x%08x ", __FUNCTION__, page);
 				goto out;
 			}
 		} else {