vfat corruption in 2.6.0?

vfat corruption in 2.6.0?

[Ludovico Gardenghi]

Hi.

A friend of mine had the same problem with 2 different test releases
(test3 and test6), and I don't know if this problem is still here as
of test7.

He has a quite big vfat partition (60 GB) created with mkfs.vfat; he
ran a program that had to write ~5000 files summing up to 18 GB but
some hour after that program started (it's a simulation tool that
runs for ~20 hours on an athlon XP 2500+) his /var started to fill with
log errors of "attempt to access beyond the end of the device".
The files are very fragmented because they are written line by line
more or less in parallel.

Moreover, the partition resulted unmountable and fsck.vfat could not
manage to repair it --- the only solution being running MS win's
scandisk tool. After the repair some of the smaller files on the disk
got lost and some part of the bigger files got corrupted.

This happened twice (with test3 and test6) and the partition was
completely erased and re-created between the 2 crashes.

I can't tell much more than this because my friend had to erase his logs
because they filled up /var.

Ludovico
-- 
<dunadan@libero.it>          #acheronte (irc.freenode.net) ICQ: 64483080
GPG ID: 07F89BB8              Jabber: garden@jabber.students.cs.unibo.it
-- This is signature nr. 1249

Re: vfat corruption in 2.6.0?

[OGAWA Hirofumi]

Ludovico Gardenghi <garden@despammed.com> writes:

> He has a quite big vfat partition (60 GB) created with mkfs.vfat; he
> ran a program that had to write ~5000 files summing up to 18 GB but
> some hour after that program started (it's a simulation tool that
> runs for ~20 hours on an athlon XP 2500+) his /var started to fill with
> log errors of "attempt to access beyond the end of the device".
> The files are very fragmented because they are written line by line
> more or less in parallel.
> 
> Moreover, the partition resulted unmountable and fsck.vfat could not
> manage to repair it --- the only solution being running MS win's
> scandisk tool. After the repair some of the smaller files on the disk
> got lost and some part of the bigger files got corrupted.
> 
> This happened twice (with test3 and test6) and the partition was
> completely erased and re-created between the 2 crashes.
> 
> I can't tell much more than this because my friend had to erase his logs
> because they filled up /var.
> 
> Ludovico
> -- 
> <dunadan@libero.it>          #acheronte (irc.freenode.net) ICQ: 64483080
> GPG ID: 07F89BB8              Jabber: garden@jabber.students.cs.unibo.it
> -- This is signature nr. 1249
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
> 

-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

Re: vfat corruption in 2.6.0?

[OGAWA Hirofumi]

Ludovico Gardenghi <garden@despammed.com> writes:

> He has a quite big vfat partition (60 GB) created with mkfs.vfat; he
> ran a program that had to write ~5000 files summing up to 18 GB but
> some hour after that program started (it's a simulation tool that
> runs for ~20 hours on an athlon XP 2500+) his /var started to fill with
> log errors of "attempt to access beyond the end of the device".
> The files are very fragmented because they are written line by line
> more or less in parallel.

Can I get your program or how can I make it reproduce? Any hint?

[...]

> I can't tell much more than this because my friend had to erase his logs
> because they filled up /var.

If it happen next time, please send the dmesg and .config.
-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

Re: [despammed] Re: vfat corruption in 2.6.0?

[Ludovico Gardenghi]

On Mon, Oct 13, 2003 at 02:18:02AM +0900, OGAWA Hirofumi wrote:

> > He has a quite big vfat partition (60 GB) created with mkfs.vfat; he
> > ran a program that had to write ~5000 files summing up to 18 GB but
> Can I get your program or how can I make it reproduce? Any hint?

It's quite simple: given a program than echoes a list of lines like

EVENT:xxxxx:something:somethingelse:somethingelsemore

with xxxxx a number (you can think this as a random number between 00000
and 05000)

I wrote a bash shell script that contains something like

./myprogram | grep ^EVENT | while read; do
	echo "$REPLY" >> "clientdata-${REPLY:6:5}"
done

That's all.

It continues up to ~6GB and 5000 different files then starts reading
the files back and so on. The problem *should* happen in the writing
phase.

> If it happen next time, please send the dmesg and .config.

Ok. I'll also ask him if he has a copy somewhere else, I think he sends
syslog messages to a remote machine too.

Ludovico
-- 
<dunadan@despammed.com>          #acheronte (irc.freenode.net) ICQ: 64483080
GPG ID: 07F89BB8              Jabber: garden@jabber.students.cs.unibo.it
-- This is signature nr. 1252