Debian Bug report logs - #579575
fat does not sync after cp

Toggle useless messages

Message #5 received at submit@bugs.debian.org (full text, mbox, reply):

From: jamesb <jagginess@videotron.ca>

To: Debian Bug Tracking System <submit@bugs.debian.org>

Subject: fat does not sync after cp

Date: Wed, 28 Apr 2010 10:28:09 -0400

Package: mount
Version: 2.13.1.1-1
Severity: serious

i did a mount with a usb stick (kingston 1gig), 

mount -t auto /dev/sda1 /mnt/sda1
cp /path/foobar.zip .
ls (while under /mnt/sda1) shows the newly copied /mnt/sda1/foobar.zip
cd /
umount /mnt/sda1

i remount the stick on another system and noticed it wasnt there, so i 
came back and did the following

mounted,
cp'd
"sync"
etc..

the sync did the trick.. 

something is missing somewhere, i shouldn't have to sync in order
to keep my fat32 usb drive intact..

the file was only 21mb.. 

i marked this serious, as i dont know what may have happened if i used
the move command.. but i'll try that next :)


-- System Information:
Debian Release: 5.0.3
  APT prefers stable
  APT policy: (500, 'stable')
Architecture: i386 (i686)

Kernel: Linux 2.6.26-2-686 (SMP w/1 CPU core)
Locale: LANG=en_US, LC_CTYPE=en_US (charmap=ISO-8859-1)
Shell: /bin/sh linked to /bin/bash

Versions of packages mount depends on:
ii  libblkid1                     1.41.3-1   block device id library
ii  libc6                         2.7-18     GNU C Library: Shared libraries
ii  libselinux1                   2.0.65-5   SELinux shared libraries
ii  libuuid1                      1.41.3-1   universally unique id library

mount recommends no packages.

Versions of packages mount suggests:
ii  nfs-common               1:1.1.2-6lenny1 NFS support files common to client

-- no debconf information

Message #10 received at 579575@bugs.debian.org (full text, mbox, reply):

From: John Pearson <johnp@gtagalenco.com.au>

To: 579575@bugs.debian.org

Subject: also seen here, with VFAT fs corruption

Date: Thu, 29 Apr 2010 11:27:46 +0930

Hi,

I saw the same behaviour here, accessing a Lexar 4Gb firefly drive under
Gnome.

Initially put down to user error, after seeing this bug I tried again
and confirmed the bug: I inserted the USB stick which Gnome mounted for
me, copied a folder 'data' containing 24 files totalling a little under
4Mb, unmounted and re-mounted the drive as root; it contained an empty
"data" folder.

Copied the folder again (as root), ran "sync", unmounted and remounted;
the volume was mounted read-only, with the following kernel messages:
[1808059.588790] FAT: utf8 is not a recommended IO charset for FAT
filesystems, filesystem will be case sensitive!
[1808059.831298] FAT: Filesystem panic (dev sdb1)
[1808059.831298]     fat_get_cluster: invalid cluster chain (i_pos 0)
[1808059.831298]     File system has been set read-only
[1808059.928105] FAT: Filesystem panic (dev sdb1)
[1808059.928105]     fat_get_cluster: invalid cluster chain (i_pos 0)
[1808059.928680] FAT: Filesystem panic (dev sdb1)
[1808059.928680]     fat_get_cluster: invalid cluster chain (i_pos 0)
[1808066.051727] FAT: Filesystem panic (dev sdb1)
[1808066.051733]     fat_get_cluster: invalid cluster chain (i_pos 0)
[1808066.338154] FAT: Filesystem panic (dev sdb1)
[1808066.338160]     fat_get_cluster: invalid cluster chain (i_pos 0)
[1808067.365064] FAT: Filesystem panic (dev sdb1)
[1808067.365064]     fat_get_cluster: invalid cluster chain (i_pos 0)
[1808085.401633] FAT: Filesystem panic (dev sdb1)
[1808085.401638]     fat_get_cluster: invalid cluster chain (i_pos 0)
[1808104.819785] FAT: Filesystem panic (dev sdb1)
[1808104.819785]     fat_get_cluster: invalid cluster chain (i_pos 0)
[1808156.203844] FAT: Filesystem panic (dev sdb1)
[1808156.203844]     fat_get_cluster: invalid cluster chain (i_pos 0)
[1808156.203844]     File system has been set read-only

The 'data' directory showed up with '?' for all permissions, and wasn't
readable.

I ran mkdosfs on the thumbdrive and re-tried mount; cp, sync; umount;
mount.  This time it vworked, so I'm guessing that the corruption likely
initially occured during the initial "failed" cp operation as user, and
was compunded when the directory was overwritten during the subsequent
cp as root.

The initial mount and cp was performed as an unprivileged user from the
Gnome desktop; all subsequent operations were done as root.

More or less stock Lenny install, with linux-image-2.6.26-2-686 version
2.6.26-21lenny4 and mount 2.13.1.1-1


John Pearson

-- 
johnp@gtagalenco.com.au  |   The greatest problem in communication is
Network Administrator    | the illusion that it has been accomplished
+61 4 0739 1169          |   - George Bernard Shaw: The Sanity of Art

Message #15 received at 579575@bugs.debian.org (full text, mbox, reply):

From: Moritz Muehlenhoff <jmm@inutil.org>

To: jamesb <jagginess@videotron.ca>

Cc: 579575@bugs.debian.org

Subject: Re: fat does not sync after cp

Date: Thu, 29 Jul 2010 22:50:09 -0400

On Wed, Apr 28, 2010 at 10:28:09AM -0400, jamesb wrote:
> Package: mount
> Version: 2.13.1.1-1
> Severity: serious
> 
> i did a mount with a usb stick (kingston 1gig), 
> 
> mount -t auto /dev/sda1 /mnt/sda1
> cp /path/foobar.zip .
> ls (while under /mnt/sda1) shows the newly copied /mnt/sda1/foobar.zip
> cd /
> umount /mnt/sda1
> 
> i remount the stick on another system and noticed it wasnt there, so i 
> came back and did the following
> 
> mounted,
> cp'd
> "sync"
> etc..
> 
> the sync did the trick.. 
> 
> something is missing somewhere, i shouldn't have to sync in order
> to keep my fat32 usb drive intact..
> 
> the file was only 21mb.. 
> 
> i marked this serious, as i dont know what may have happened if i used
> the move command.. but i'll try that next :)

You need to mount your USB stick with the "sync" option if you want your
data written out rightway. See "man mount".

Cheers,
        Moritz

Message #22 received at 579575@bugs.debian.org (full text, mbox, reply):

From: John Pearson <johnp@gtagalenco.com.au>

To: 579575@bugs.debian.org

Cc: Moritz Muehlenhoff <jmm@debian.org>

Subject: fat does not sync on umount

Date: Mon, 02 Aug 2010 15:08:27 +0930

Hi,

The comment to the effect that '-o sync' is required to ensure data is
written out immediately concerns me.  Are you saying that outstanding
data is not necessarily flushed to disk before umount returns if '-o
sync' is not specified? 

If that's the expected behaviour then this remains a serious bug that
can cause data loss - the original reporter mounted his drive from the
command line and so at least had the option of specifying 'sync', but in
my case the drive was mounted automatically during a GNOME session, and
there was no opportunity to specify mount options.


John Pearson

-- 
johnp@gtagalenco.com.au  |   The greatest problem in communication is
Network Administrator    | the illusion that it has been accomplished
+61 4 0739 1169          |   - George Bernard Shaw: The Sanity of Art

Message #31 received at 579575@bugs.debian.org (full text, mbox, reply):

From: Simon Paillard <spaillard@debian.org>

To: Helge Kreutzmann <debian@helgefjell.de>, 603956-done@bugs.debian.org, 579575@bugs.debian.org

Cc: control@bugs.debian.org

Subject: Re: Needs to be mentioned in release notes

Date: Thu, 9 Dec 2010 23:43:28 +0100

Hi Helge,

On Thu, Nov 18, 2010 at 09:42:33PM +0100, Helge Kreutzmann wrote:
> clone 579575 -1
> reassign -1 release-notes
> thanks
> 
> Given that traditionally a returned unmount would indicate that all
> outstanding write operations are completed and it was hence safe to
> remove the media and the ubiquity of fat formatted usb drives and the
> fact that non command line users are affacted too this behaviour
> clearly should be mentioned in the release notes.

Actually this bug has not been confirmed by util-linux maintainers, and is not
reproducible (linux-image-2.6.32-5-amd64 2.6.32-28 with util-linux 2.17.2-3.3).

It seems misunderstanding happened between reporters and Moritz (on whether
they actually umounted it or not).

-> Closing the bug against release-notes.
 
> The better alternative would be to remedy this behaviour, i.e. cause
> umount to actually write the data, but given the bug log in 579575
> this does not seem to be the case for squeeze.

-- 
Simon Paillard

Message #36 received at 579575@bugs.debian.org (full text, mbox, reply):

From: Helge Kreutzmann <debian@helgefjell.de>

To: Simon Paillard <spaillard@debian.org>

Cc: 603956@bugs.debian.org, 579575@bugs.debian.org

Subject: Re: Needs to be mentioned in release notes

Date: Fri, 10 Dec 2010 19:10:09 +0100

[Message part 1 (text/plain, inline)]

Hello Simon,
On Thu, Dec 09, 2010 at 11:43:28PM +0100, Simon Paillard wrote:
> On Thu, Nov 18, 2010 at 09:42:33PM +0100, Helge Kreutzmann wrote:
> > clone 579575 -1
> > reassign -1 release-notes
> > thanks
> > 
> > Given that traditionally a returned unmount would indicate that all
> > outstanding write operations are completed and it was hence safe to
> > remove the media and the ubiquity of fat formatted usb drives and the
> > fact that non command line users are affacted too this behaviour
> > clearly should be mentioned in the release notes.
> 
> Actually this bug has not been confirmed by util-linux maintainers, and is not
> reproducible (linux-image-2.6.32-5-amd64 2.6.32-28 with util-linux 2.17.2-3.3).

Glad to hear that!

I assumed that the "no confirmation" was a "I have not idea how to fix
this", this assumption has now been corrected, thanks.

> It seems misunderstanding happened between reporters and Moritz (on whether
> they actually umounted it or not).

Rereading the bug trail I see your point. Unfortunately Moritz did not
reply to #22 and so I understood it was a silent agreement.

Greetings

          Helge
-- 
      Dr. Helge Kreutzmann                     debian@helgefjell.de
           Dipl.-Phys.                   http://www.helgefjell.de/debian.php
        64bit GNU powered                     gpg signed mail preferred
           Help keep free software "libre": http://www.ffii.de/

[signature.asc (application/pgp-signature, inline)]

Message #41 received at 579575-done@bugs.debian.org (full text, mbox, reply):

From: Phillip Susi <psusi@ubuntu.com>

To: 579575-done@bugs.debian.org

Subject: fat does not sync after cp

Date: Fri, 01 Nov 2013 09:42:23 -0400

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

If you want the fat driver to sync automatically when the cp ends so
you can yank out the drive without unmounting, use -o flush.

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.17 (MingW32)
Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/

iQEcBAEBAgAGBQJSc6+/AAoJEJrBOlT6nu7530cIAJCHPJXwoPx577fWuL8b05Kz
unXSxfx6xk7BF6gZ5u2Je2fzQDqKyyhi3+MeWTEIjafJjGiHphqd44L/Ebaazyb5
gAfs+uJCiqKqzW5HHumkLS+a8y++6lqapVXXH0ig25ltNhbKYTCcYwFhlqxrYwyO
U0J59q6XJW5sRbMyN3/RUywg8W18LF11zbppwygLrZWnjH5HCujhE3I8tLuKE50p
kKdDE1JlxpaHdu39Kk09Zy3mTLV8lQx5rYulB7LwLxfTeY+NEmQAmQHb6hFIh/+4
13MpO+x6RTNsgECiTM/eNUZ85ncuRafjBRUZTUDCos3fDguCkKfm3L2JWQdpaWo=
=CE4Z
-----END PGP SIGNATURE-----

Send a report that this bug log contains spam.