OS: Ubuntu (same thing happens on a straight Debian system)
Kernel: i686 generic
Hardware: using PCI SATA card (maybe this is the problem?)
Filesystem: ReisferFS(same thing happens with EXT3 though)

As you can see from the output, there is 10 Gigs of free space on this ReisferFS partition.

The problem is that you can also see it reports being 100% used.

I don't care if I loose 10 gigs of space but here is the problem.

I make FTP backups, and sometimes the space gets taken up before I can clear it out daily.

Once this situation happens I can still write data to the drive! The data gets successfully written, but guess what....the data written at the end of the partition gets written over, with the new data.

I verified this using md5sums of the original backup files.

So the question is, what is causing this undesired behaviour of data corruption at the end of the partition, or more like overwriting parts of other files to make them corrupt?

What can be done about this?

And also smart reports everything with this drive is fine.

Any help would be appreciated.

Never seen anything like that - and I run systems hard under test.
Might be that card, you never know - multi-processor(/core) system by any chance ???.

Thanks syg00, it is not hardcore or multi-core though.
I suspect it must be an issue with this Sil3114 based card.

I've read lots of mailing lists with people having constant corruption.

The thing is I don't get any corrupt data until I reach the end of the partition. I guess the only way to know is to find the time to try a different SATA card on this machine.

I can't see this being a bug with different kernels or filesystems that I've tried, so it must be the card I'm guessing.

It would be a huge bug for both filesystems to have this kind of bug where it allows data to be written when 100% full that overwrites other valid data.