[arch-general] journalctl and I/O errors

[Anatol Pomozov]

Hi

I agree on both points.

On Wed, Feb 5, 2014 at 10:03 PM, Sébastien Leblanc
<leblancsebas at gmail.com> wrote:
> Conclusion (as I understand it):
>
> 1. There is definitely a bug in Journalctl: it crashes (segfaults) on I/O
> errors.

A few months ago I had a problem with btrfs. I set +C attribute
(disable copy-on-write) on existing journal files. Btrfs recommends to
put the attribute on empty files and seems was confused that I applied
it to non-empty files. Btrfs started returning IO error when I was
trying to read the file with 'less' and journalctl started crashing
with segfault. It is very similar to what being discussed here. And I
agree that journalctl should play nicer.

If anyone still sees this problem please run 'strace journalctl ...'.
If it shows that a filesystem operation returns IO error right before
SEGFAULT then it proves current thesis.

> 2. You have a drive that is failing, or your BIOS might not be set
> correctly. This is causing the I/O errors. How large is the drive? You
> might have to turn off settings such as "SATA legacy compatibility" or the
> like -- I had a 3TB drive that would cause ATA command errors on a ~2006
> computer; I found this option in the BIOS and as soon as I turned it off
> everything worked perfectly.
>
> Although journalctl should not crash on I/O errors, I think it is not
> unreasonable to assume that many other apps do not tolerate I/O errors
> either. So I would say: you should still report this bug upstream.
>
> --
> Sébastien Leblanc