File corruption with 2.4.20pre6

File corruption with 2.4.20pre6

[Joachim Bremer]

Hi all,

I've got an easy to produce file corruption with
linux-2.4.20pre6 and then tested with the file-write
patch on pre4. I verified the problem on 2 different machines -
1. Uniprozessor with an Athlon XP1600+
2. Dualprozessor with 2 Athlon MP2000+

On both machines exactly the same files are damaged.

To reproduce:

- unpack libtool-1.4.2.tar.gz
- switch into the resulting directory
- enter: ./configure --prefix=/usr

The generated file "libtool" is damaged. Just do an
./libtool and you will se weird errors. The same procedure
works OK on a ext2 filesystem with the same kernel.

Bye

Joachim

Re: File corruption with 2.4.20pre6

[Oleg Drokin]

Hello!

On Wed, Sep 11, 2002 at 12:41:25PM +0200, Joachim Bremer wrote:

> I've got an easy to produce file corruption with
> linux-2.4.20pre6 and then tested with the file-write
> patch on pre4. I verified the problem on 2 different machines -
> 1. Uniprozessor with an Athlon XP1600+
> 2. Dualprozessor with 2 Athlon MP2000+

Thanks a lot for your report!

This is my stupid bug and I am starting to wonder how did it passed through
my tests...

This trivial patch below should take care of it. Can you try it? 
Thank you.


===== fs/reiserfs/file.c 1.10 vs edited =====
--- 1.10/fs/reiserfs/file.c	Tue Sep 10 16:06:23 2002
+++ edited/fs/reiserfs/file.c	Wed Sep 11 17:43:48 2002
@@ -1189,7 +1189,7 @@
 
 	already_written += write_bytes;
 	buf += write_bytes;
-	pos = *ppos += write_bytes;
+	*ppos = pos += write_bytes;
 	count -= write_bytes;
     }
 

Bye,
    Oleg

Re: File corruption with 2.4.20pre6

[Joachim Bremer]

Hi all,

> Hello!
> 
> On Wed, Sep 11, 2002 at 12:41:25PM +0200, Joachim Bremer wrote:
> 
> > I've got an easy to produce file corruption with
> > linux-2.4.20pre6 and then tested with the file-write
> > patch on pre4. I verified the problem on 2 different machines -
> > 1. Uniprozessor with an Athlon XP1600+
> > 2. Dualprozessor with 2 Athlon MP2000+
> 
> Thanks a lot for your report!
> 
> This is my stupid bug and I am starting to wonder how did it passed through
> my tests...
> 
> This trivial patch below should take care of it. Can you try it? 
> Thank you.

pretty short test on the single CPU system showed OK. If there is a problem
on the Dual CPU (I dont't think so) I will file a report.

Thanks for the quick fix

Joachim

> 
> 
> ===== fs/reiserfs/file.c 1.10 vs edited =====
> --- 1.10/fs/reiserfs/file.c	Tue Sep 10 16:06:23 2002
> +++ edited/fs/reiserfs/file.c	Wed Sep 11 17:43:48 2002
> @@ -1189,7 +1189,7 @@
>  
>  	already_written += write_bytes;
>  	buf += write_bytes;
> -	pos = *ppos += write_bytes;
> +	*ppos = pos += write_bytes;
>  	count -= write_bytes;
>      }
>  
> 
> Bye,
>     Oleg
>