Hi.

My machine runs
1 x 500 GB SATA EXT3
2 x 500 GB IDE EXT3
2 x 200 GB IDE EXT3
1 x 80 GB IDE EXT3

on a 2.6.25-2-686 / 2.6.26-1-686 kernel in a Debian distro.

Lately, I've been having trouble with data integrity. The following usually happens:

I want to copy a file to a harddrive, and I'm told that the drive is R/O. I unmount the drive and check for errors. FSCK asks me if I want to correct bad inodes, I say yes. When I inspect the harddrive, whole folders have been neatly renamed and restored to the `lost+found` folder, effectively becoming garbage. I lost another 300 G of data.

The drives are all fairly new, and diagnostic software cannot find anything wrong with the hardware itself.

I've lost about 1 TB of data so far, and I'm getting a bit irritated.

So far, only the 2 500 GB IDE EXT3 drives have been affected, but I'm worried for the rest.

I've been doing some digging, and the only thing I've been able to come up with that explains this is a thread on a Ubuntu forum (which I can't seem to locate right now; I'll provide a link later) in which it was stated that this behavior was due to a kernel bug, from kernel 2.6.15 upward. The bug was expected to be fixed in 2.6.27, but debian doesn't offer that upgrade yet, and I'm unsure whether that will provide a solution anyway.

Do any of you know anything about this? Possible solutions perhaps?

Thanks in advance!

So what's stopping you from downloading the most recent kernel and compiling it yourself just to test?

I haven't seen any issues with EXT3 filesystems going corrupt and I'm currently running kernel 2.6.23.

Never rule out hardware problems either, might not be the drives themselves, could be the controller they're attached to. Make sure there are no firmware updates for your hardware, I've seen these issues in the past with Read-Only drives due to such things and or other hardware malfunctions. So before you just blame the kernel, do more digging.

Fear. Unmoderated, pure ad simple. I think it'd be easier to revise the entire system, replace the IDE drives with SATA and install Ubuntu than that...

Ouch. Another one of those things. Okay, I guess I'll have to learn how to do stuff like that, without wiping the entire system.

I hadn't thought of firmware updates, true. But the post I was referring to was the only result matching the troubles I had that I could find after 3 hours of googling.

Well, thanks. I'll be at Compiling Kernels 101, and Upgrading Firmware Without Totaling Your System 101...

No need to be worried about compiling your own kernel. The worst that can happen is that the new kernel won't boot, in which case you just reboot the machine and use the previous kernel, figure out what you did wrong and try again.

Said like a true gentoo user ...

I run EXT3 almost exclusively on a truck load of different (recent) kernels and have seen no issues.
Not to say there aren't potential errors in code out there, but I've not heard of any.

Did some more digging. This time I went digging physically: into the bowels of my ctr. Found out that the two suspect discs were both connected to the same PCI IDE controller.

04:01.0 Mass storage controller: Integrated Technology Express, Inc. IT/ITE8212 Dual channel ATA RAID controller (rev 13)

It's set in passthrough mode, and has a bios revision some 3 years before the latest update.

I've spent some time and a couple of CD's trying to find out how to flash its ROM, but fearing after a couple of unsuccessful attempts that I might end up disabling my mainboard, I've decided to take the n00bs way out and just buy some 1 TB SATA drives and replace the IDE controller with a better one so that I can at least copy the remaining data, and perhaps keep some backup storage space.

Last question: any preferences as to what brand and model of controller I should buy? Is Promise any good?

Thanks!

Anything with LSI chipset is what I sort of prefer nowadays. Promise is okay but from my experience, it's rather on the lower end of the list of quality. 3ware is okay but it's kind of overpriced in my opinion.