Re: [Jfs-discussion] dirty FS running as clean - missing directories

[Dave K. <sh...@au...>]

On Sun, 2003-10-05 at 16:45, mt...@am... wrote:
> Hello all,
> 
> I'm missing /usr/ in "ls" output for the FS-root, nontheless
> I can change into the directory and the system is still running...
> ... for the moment!
> Maybe the appended errorlog is helpful to developers - especially
> because fsck replayed journal after crash (took some time), then it
> told me the FS is clean.
> 
> I'll go and obtain a new HD to "dd" the partition before trying
> to fsck-fix it. Hopefully that error is caused by the already fixed
> "too much optimized" bug and fsck + updating the system components
> will help.

I doubt that bug has anything to do with this.  The only inconsistancy
that causes is that the free space is sometimes under-reported.  Did you
build JFS 1.1.3 against this kernel?  If not you won't see the "too much
optimized" bug on this kernel.

> I'd like to know if somebody can already predict the behaviour of
> fsck when told to fix the problems listed below.

For a mounted FS, the output appears very sane.  There are no problems
at all until checking the inode maps and block maps, and some of the
important map data is not kept in-sync on disk until unmount time since
fsck rebuilds it when the journal is replayed.  In fact "fsck -f" will
completely rebuild these maps from scratch.

Nothing here explains the /usr behavior, but I don't expect that fixing
the FS will result in any data loss.

> Best regards
> 
> PS: some funny behaviour related to the FS corruption:
>     # ls -la | grep something
>     ls: reading directory .: Input/output error
> listing without pipe and grep seems to work although...

Strange.

> ------------------------
> Linux l1 2.4.20-4GB-athlon #1 Wed Jun 18 07:44:55 UTC 2003 i686 unknown unknown GNU/Linux
> ------------------------

-- 
David Kleikamp
IBM Linux Technology Center