linux - df shows negative values for used?

I have a CentOS 5.2 server and running df -h I get this:

Filesystem            Size  Used Avail Use% Mounted on
/dev/mapper/VolGroup00-LogVol00
                      672G -551M  638G   0% /
/dev/hda1              99M   12M   82M  13% /boot
tmpfs                 2.0G     0  2.0G   0% /dev/shm

that space wasn't even near 10% usage the last time it showed a correct value. I'm at a loss with whats going on.

EDIT #1

Ok so I had to reboot the server because SSHD hanged up, I'm guessing it was related to this.

Some new info, after rebooting, df -h showed 12Gb used (2%), but if I run du -hcs / it shows 46Gb total, theres a big difference here.

EDIT #2

After about 15mins of uptime df -h is showing negative values again:

Filesystem            Size  Used Avail Use% Mounted on  
/dev/mapper/VolGroup00-LogVol00
                      672G  -24G  660G   -  /

EDIT #3

More info, ran a fsck and this is the output:

Checking all file systems.
  [/sbin/fsck.ext3 (1) -- /] fsck.ext3 -f -y /dev/VolGroup00/LogVol00 
Pass 1: Checking inodes, blocks, and sizes 
Pass 2: Checking directory structure 
Pass 3: Checking directory connectivity 
Pass 4: Checking reference counts 
Pass 5: Checking group summary information
 /dev/VolGroup00/LogVol00: 204158/181633024 files (1.3% non-contiguous), 9224806/181633024 blocks 

[/sbin/fsck.ext3 (1) -- /boot] fsck.ext3 -f -y /dev/hda1 
Pass 1: Checking inodes, blocks, and sizes 
Pass 2: Checking directory structure 
Pass 3: Checking directory connectivity 
Pass 4: Checking reference counts 
Pass 5: Checking group summary information 
/boot: 34/26104 files (5.9% non-contiguous), 15339/104388 blocks

Answer

I think it is a file system corruption. You should unmount the partition and run a fsck.
Check also the logs and the console for any file system errors.