last 541 kernel damages fileystem .. starting fresh FC3T2 candidate.

[Stephen C. Tweedie]

Hi,

On Mon, 2004-09-06 at 17:31, Andy Green wrote:

> > statfs("/boot", {f_type="EXT2_SUPER_MAGIC", f_bsize=1024, f_blocks=248895,
> > f_bfree=389288, f_bavail=376438, f_files=64256, f_ffree=64210, f_fsid={0,
> > 0}, f_namelen=255, f_frsize=1024}) = 0 write(1, "/dev/sda1              
> > 248895 -"..., 74) = 74
> > /dev/sda1        248895 -18446744073709411223    376438 101% /boot
> >
> > At the moment they look like separate problems.
> 
> Hum, an fsck makes the df -h problem go away.  It's hard to see how if the 
> data returned from the kernel is really correct in both cases?

Tim pointed out that f_bfree was > f_blocks, so it's hardly surprising
that "df" was getting upset.  Definitely a kernel problem.

> BTW what kind of unhappiness can we expect from this "inode problem"?  Just 
> things that e2fsck will make go away?

That's all I was able to reproduce myself: bad block counts in the
various block groups.  Other people have reported file corruption,
though.

--Stephen