df is wrong for mounted device

[Myrmornis]

I am using an external hard drive (an apple ipod) and I am confident that there is 41G space left on it. That is also what du tells me:

~> du --max-depth=0 /mnt/ipod/
42319056        /mnt/ipod/

However, df reports that it is almost full:

~> df /mnt/ipod/
Filesystem           1K-blocks      Used Available Use% Mounted on
/dev/sdb2             77973488  77797872    175616 100% /mnt/ipod

And indeed I am told that I'm out of space when I try to copy large files.

I believe that df is wrong. But what is causing the problem and how do I correct it?

Thanks a lot,

Dan

p.s. I have previously seen this sort of thing being caused by files stored below the mountpoint, that are "hidden" by the mounted device. However, when I unmount the device, there is nothing stored at /mnt/ipod

~> umount /mnt/ipod
~> du /mnt/ipod/
4       /mnt/ipod/

[artemn]

Please paste here output of this command:
mount -l
(this will show options for all mounded file systems).

I also suggest to use:
du -sh /mnt/ipod/
instead of
du --max-depth=0 /mnt/ipod/

and

df -h /mnt/ipod
instead of
df /mnt/ipod

[Myrmornis]

Thanks artemn, here goes:

First this seems to show a clear inconsistency between du and df:

~> du -sh /mnt/ipod/
41G     /mnt/ipod/

~> df -h /mnt/ipod
Filesystem            Size  Used Avail Use% Mounted on
/dev/sdb2              75G   75G   14M 100% /mnt/ipod

And here's the output of mount -l:

~> mount -l
/dev/sda3 on / type ext3 (rw) [/]
proc on /proc type proc (rw)
sysfs on /sys type sysfs (rw)
devpts on /dev/pts type devpts (rw,gid=5,mode=620)
tmpfs on /dev/shm type tmpfs (rw)
/dev/mapper/vg00-lvhome on /home type ext3 (rw)
/dev/mapper/vg00-lvtmp on /tmp type ext3 (rw)
/dev/mapper/vg00-lvusr on /usr type ext3 (rw)
/dev/mapper/vg00-lvvar on /var type ext3 (rw)
/dev/sda2 on /mnt/shared type vfat (rw,uid=502,gid=502)
none on /proc/sys/fs/binfmt_misc type binfmt_misc (rw)
sunrpc on /var/lib/nfs/rpc_pipefs type rpc_pipefs (rw)
automount(pid2521) on /net type autofs (rw,fd=4,pgrp=2521,minproto=2,maxproto=4)
/dev/sdb2 on /mnt/ipod type vfat (rw,noexec,nosuid,nodev,user=dan) [IPOD]
/dev/sdc1 on /mnt/xd type vfat (rw) [DSK2_VOL1]

[artemn]

Hmm...

Maybe it makes sense to back up data and perform:
fsck -t vfat /mnt/ipod

By the way, here is discussion of similar problem for XFS filesystem:
du vs. df inconsistency

[Myrmornis]

artemn suggested

Maybe it makes sense to back up data and perform:
fsck -t vfat /mnt/ipod

I did that (output below), but the problem seems to persist, even though I selected the option to correct something that fsck thought was wrong.

(EDIT: I just checked under Windows, which reported the correct situation: 41G used, 34G available.)

/home/dan# fsck -V -t vfat /mnt/ipod ; echo exit code is $?     
fsck 1.38 (30-Jun-2005)
[/sbin/fsck.vfat (1) -- /mnt/ipod] fsck.vfat /dev/sdb2 
dosfsck 2.11, 12 Mar 2005, FAT32, LFN
Free cluster summary wrong (846 vs. really 2218272)
1) Correct
2) Don't correct
? 1
Leaving file system unchanged.
/dev/sdb2: 8126 files, 2655071/4873343 clusters
exit code is 1

/home/dan# umount /mnt/ipod/
/home/dan# mount /mnt/ipod/
/home/dan# df -h /mnt/ipod/
Filesystem            Size  Used Avail Use% Mounted on
/dev/sdb2              75G   75G   14M 100% /mnt/ipod
/home/dan# du -sh /mnt/ipod/
41G     /mnt/ipod/

[artemn]

Look at these lines:

Free cluster summary wrong (846 vs. really 2218272)

and

Leaving file system unchanged.

It seems it's necessary to start fsck with autorepair option:

fsck -a -V -t vfat /mnt/ipod

[Myrmornis]

Thanks for that. I noticed that fsck had refused my request to repair, but I didn't know hot to use the autorepair option.

However, in the end I gave up and reformatted the disk (following the instructions at IpodConversionToFAT32 < Main < TWiki), and I think it's OK now.

Thanks again for the help.