On Tue, 19 Sep 2006, Sverre Froyen wrote:

> While I am on the topic of LFS, I have noticed that fsck on a cleanly 
> shutdown LFS file system usually shows some type of error or 
> inconsistency.  Is this to be expected or does it indicate some other 
> problem.

This type of inconsistency is not expected, though I don't think it will 
cause you a future panic either.

I've been working over the summer on getting the checkpoints solidly 
consistent, using a test script that prevents the filesystem from wrapping 
the log, and examines each checkpoint in turn for validity (using fsck_lfs 
-n).  This seems to be working now.  Recently I've added "fsck_lfs -p" to 
this test, rolling forward from every non-checkpoint partial-segment 
write; there are some cases in which "fsck_lfs -p" still gets it wrong, 
the cause of which I have yet to investigate.

I should point out, though, that if the checkpoints are indeed valid, 
fsck_lfs -p will never give a more consistent filesystem, only a slightly 
more recent view of the filesystem.

I haven't been explicitly testing shutdown/reboot, of course, nor even 
umount, since the checkpoint should be the same no matter where it comes 
from.

Take care,
 						Konrad Schroder
 						perseant@hhhh.org