Re: another hard disk broken or xfs problems?

[Nathan Scott <nathans@sgi.com>]

On Thu, Feb 26, 2004 at 09:25:51AM +0100, Nico Schottelius wrote:
> Nathan Scott [Thu, Feb 26, 2004 at 02:27:41PM +1100]:
> > On Thu, Feb 26, 2004 at 12:49:44AM +0100, Nico Schottelius wrote:
> > > There are still some questions open for me:
> > > 
> > > 1. why is it an internal xfs error?
> > 
> > Your loopback file seems to have got corrupted, XFS reports
> > this as an internal error (generic error message).
> 
> I am really wondering about the error message, as "internal errors" 
> indicate for me an error in the kernel.

It is a misleading error message, I'll look into correcting that.

> > > 2. why does it print a call trace?
> > 
> > XFS detected corruption, and tried to dump out some state info
> > at the point where it noticed the problem.
> 
> I am wondering how my dmesg will look like if I've to recover some
> Gigabytes of date.

The filesystem typically shuts down on detection of corruption,
so you should get just the one error report.

> And btw, do all filesystem drivers behave in this way, printing internal
> errors and displaying call traces when they find errors in the
> filesystem?

No, not all filesystem behave this way.  And it is configurable
in XFS; if you don't want this to happen, you can switch it off
via the sysctl/procfs interface - see the "error_level" section
in Documentation/filesystems/xfs.txt.

> For me this is really confusing.

Hope this helps.

cheers.

-- 
Nathan