EXT4 Data Corruption with kernel 3.6.x

[Fintan]

I just came across this:
http://www.phoronix.com/scan.php?pag...tem&px=MTIxNDQ

[dmeyer]

Don't worry about this. Look at the Kernel version you are running with uname -r. You are on the 3.2 (no fears) series if on 12.04 and if you are on 12.10 then its 3.5(a little concern.) By the way its all the kernels back to 3.4 as that change was backported.

Honestly though, the use case to get that type or corruption is quite tricky. Unless you restart your computer every few seconds after making big changes like a kernel developer you won't run into any issues. A patch has been released and will arrive shortly (I hope.)

[dmeyer]

My bad. I should have looked at your profile. Obviously you weren't stressed about it because you are a long time user and was just making people aware.

I spoke to somebody on IRC who was freaking out over this and saw another person out on the openSUSE forum so when I see something like this I just give a generic response.

No offense meant. I just don't want people to get the wrong impressions about this.

[claydoh]

I dunno --- having had 2 drives recently "die" with data corruption, it gives one pause. One of my drives really did die, but the other one after reformatting seems fine. Both exhibited the same symptoms - many if not all user config files lost their write permissions. Seeing a few reports of similar symptoms with other users might make one think this bug may be the cause.

It does seem to be easily overblown, though, from reading the mailing list thread on it. The backporting of the patch that might be the cause is what is freaking people out. I was concerned and have looked into it, and the only thing that concerns me is that it had been hard to discover, so how do we as non-technical users really know how bad it is , or not.

[dibl]

The nature and extent of the issue is still evolving. So it's WAY too early to either panic or change anything in your system. BTW shutting off your running Linux system with the power switch was NEVER a good idea ...

[dmeyer]

I dunno --- having had 2 drives recently "die" with data corruption, it gives one pause. One of my drives really did die, but the other one after reformatting seems fine.

3.6.2 is only like a week old. I really don't think it got into all that many systems. Even with a backport, most distro's still don't really update their kernels all that often except for important things (which the original patch doesn't really seem to be.) So I doubt that might be the cause of your problem but it does raise a little concern over the security of EXT4 - possibly another bug somewhere else. I've personally never had data corruption or a drive die which is amazing luck but I have had to fix a few drives before. EXT4 partitions are usually an easy job to fix. Heck even NTFS is pretty easy thanks to fairly nice tools available.

Good thing though that I am fanatical about my back-ups though! It peeves me though when people back up their laptop and then carry that HDD around with them. Then when the bag gets stolen they blame everybody but themselves. Back-ups are worthless unless they are safe, secure and comprehensive. Storage is so cheap nowadays that its really not a burden on the pocket to fork out a little cash for a drive dedicated to just back-ups. Also back-ups are so easy to do nowadays! There are tons of amazing linux utilities but I stick to the old fashioned "dd"!

[dibl]

Actually 3.6-3 has been in Debian sid for a few days now:

don@imerabox:~$ uname -a && uptime 
Linux imerabox 3.6-3.towo-siduction-amd64 #1 SMP PREEMPT Sun Oct 21 17:59:02 UTC 2012 x86_64 GNU/Linux
 10:17:30 up 3 days,  1:03,  2 users,  load average: 0.08, 0.14, 0.15

No sign of any issues here -- my OS and several VMs are on ext4 filesystems, but the rest of my data is on a 2-disk BTRFS filesystem (which was supposedly the experimental one .....).

[dibl]

It turns out only a very unusual (and very far from default) configuration can trigger this bug. It is nothing for typical Kubuntu users to be concerned about.

[Fintan]

It does seem difficult to reproduce and I not about to try Just wanted to point it out.