Testing request: possible ext4 corruption issue

Adam Williamson awilliam at redhat.com
Thu Oct 29 18:42:13 UTC 2009

Hi, guys. Jóhann raised a concern about the potential ext4 corruption
issue that's currently under discussion upstream:


the discussion between Eric and Linus tends towards the theory that this
is not affecting 2.6.31, but we'd like to have somewhat more certainty
of this.

For those who are testing F12 on disposable systems/partitions, it would
be very much appreciated if you could test this. Testing it is
relatively easy - all we need you to do is dirty power-off active
systems with ext4 partitions. If they're physical systems, yank the
plug. If they're virtual machines, do a forced power-off from the VM
interface (not a clean shutdown request). Ideally while they're writing
data. See if you can get some data corruption this way.

There's some suggested operations to reproduce this in the bug report.
e.g. Eric's reproduction method is detailed in
http://bugzilla.kernel.org/show_bug.cgi?id=14354#c141 . But basically
just suddenly powering off the system while something's writing data to
an ext4 partition is the key - ideally you know _what_ data is being
written, so you can check if it's been corrupted on the next boot.

Again, let me emphasize, since this is a potential data corruption bug,
I _hope_ it's obvious you shouldn't test it on any system or data you
actually care about :). Of course, as Dan says, any data you don't have
backed up is data you don't care about, but please bear that in mind.

For my peace of mind it'd be great if a few of you can try this and
confirm whether or not you're able to dependably cause corruption this

It's also worth noting Jóhann and I both came to this report via the
somewhat dubious Ubuntu report at
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/453579 , which
claims to see corruption of large files on ext4 partitions. This
upstream bug was linked from that one, though they don't look related to
me. I don't place a lot of credence in the Ubuntu report, but obviously
if anyone's observed large files being corrupted on F12 ext4 partitions,
PLEASE let us know about this. Check those sha256sums.

Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Fedora Talk: adamwill AT fedoraproject DOT org

More information about the fedora-test-list mailing list