"wrong bytes" with files >=4GiB

"wrong bytes" with files >=4GiB

[Michael Weissenbacher]

I'm using kernel 2.6.9-rc1-mm1 (patched for the page-size bug) and 
reiser4progs1.0.0. I am able to reproduce a "wrong bytes" problem by 
doing the following:
# dd if=/dev/urandom bs=4096 count=1048576 of=/mnt/tmp/testfile
# umount /mnt/tmp
# fsck.reiser4 /dev/sda1
...
FSCK: Node (45223670), item (3), [2a:7465737466696c:1000f] (stat40): 
wrong bytes(4294967296), Should be (0).
...
1 fixable corruptions were detected in the FileSystem. Run with --fix 
option to fix them.

When doing a fsck.reiser4 with --fix parameter it says
FSCK: Node (45223670), item (3), [2a:7465737466696c:1000f] (stat40): 
wrong bytes
(4294967296), Fixed to (0).

and after that it says that the fs is consistent. I cannot find any 
difference in the file on the filesystem though.
Is this actually harmful or is it just "statistic" data?
If i'm doing the same test with count=1048575 no corruption is reported 
and if i'm doing it for files >4GiB i always get the problem (with 
different numbers of course).
I can sent the full fsck output if needed.
BTW this problem also occurs with kernel 2.6.8.1-mm2.

regards
Michael

Re: "wrong bytes" with files >=4GiB

[Vladimir Saveliev]

Hello

On Mon, 2004-09-06 at 16:16, Michael Weissenbacher wrote:
> I'm using kernel 2.6.9-rc1-mm1 (patched for the page-size bug) and 
> reiser4progs1.0.0. I am able to reproduce a "wrong bytes" problem by 
> doing the following:
> # dd if=/dev/urandom bs=4096 count=1048576 of=/mnt/tmp/testfile
> # umount /mnt/tmp
> # fsck.reiser4 /dev/sda1
> ...
> FSCK: Node (45223670), item (3), [2a:7465737466696c:1000f] (stat40): 
> wrong bytes(4294967296), Should be (0).
> ...
> 1 fixable corruptions were detected in the FileSystem. Run with --fix 
> option to fix them.
> 
> When doing a fsck.reiser4 with --fix parameter it says
> FSCK: Node (45223670), item (3), [2a:7465737466696c:1000f] (stat40): 
> wrong bytes
> (4294967296), Fixed to (0).
> 
> and after that it says that the fs is consistent. I cannot find any 
> difference in the file on the filesystem though.
> Is this actually harmful or is it just "statistic" data?
> If i'm doing the same test with count=1048575 no corruption is reported 
> and if i'm doing it for files >4GiB i always get the problem (with 
> different numbers of course).
> I can sent the full fsck output if needed.
> BTW this problem also occurs with kernel 2.6.8.1-mm2.
> 

This is reiser4progs problem. It will be fixed soon

> regards
> Michael
> 

Re: "wrong bytes" with files >=4GiB

[Vitaly Fertman]

On Monday 06 September 2004 16:16, Michael Weissenbacher wrote:
> I'm using kernel 2.6.9-rc1-mm1 (patched for the page-size bug) and
> reiser4progs1.0.0. I am able to reproduce a "wrong bytes" problem by
> doing the following:
> # dd if=/dev/urandom bs=4096 count=1048576 of=/mnt/tmp/testfile
> # umount /mnt/tmp
> # fsck.reiser4 /dev/sda1
> ...
> FSCK: Node (45223670), item (3), [2a:7465737466696c:1000f] (stat40):
> wrong bytes(4294967296), Should be (0).
> ...
> 1 fixable corruptions were detected in the FileSystem. Run with --fix
> option to fix them.
>
> When doing a fsck.reiser4 with --fix parameter it says
> FSCK: Node (45223670), item (3), [2a:7465737466696c:1000f] (stat40):
> wrong bytes
> (4294967296), Fixed to (0).
>
> and after that it says that the fs is consistent. I cannot find any
> difference in the file on the filesystem though.
> Is this actually harmful or is it just "statistic" data?

the only problem here is that bytes field becomes wrong. 
this does not corrupt data.

> If i'm doing the same test with count=1048575 no corruption is reported
> and if i'm doing it for files >4GiB i always get the problem (with
> different numbers of course).
> I can sent the full fsck output if needed.
> BTW this problem also occurs with kernel 2.6.8.1-mm2.

I have found it, thank you for hitting the bug, fixed.

-- 
Thanks,
Vitaly Fertman