can't ls, then rebuild-tree fails, no hardware error (kow)

can't ls, then rebuild-tree fails, no hardware error (kow)

[Casey Callendrello]

Hello all,

    I'd been using Reiserfs on a 180 GiB partition.  One day my computer 
hung while I was out.  I came home and had to hard-reboot it.  Fine.  
(Unfortunately, this is not the / partition but /home!  egad!)  
Unfortunately, when I rebooted I could not ls in the /home directory, 
which is mount --bind 'ed to /mnt/big/home (big? get it!)  sensing a 
major problem, I had to hard-reboot to kill the ls which was hung on 
waiting for IO (with no errors in /var/log/messages or the error console). 

I ran fsck and it found some errors, so I ran fsck --fix-fixable.  The 
whole directory structure was there, I saw it flash by during the fsck 
(well, as far as I could tell.  There certainly wasn't anything obvious 
missing).  So, I remount it and try to cat a file I know is in there, 
and it shows up fine.  I feel better.

But, I still can't ls.  I can't log in with a non-root user since it 
hangs waiting for IO.  I figure it's a problem with the tree (I 
shouldn't jump to conclusions).  So I run rebuild-tree and it fails.  
Now i'm up *** creek. 
I don't have a hard drive large enough to copy this to.  I also don't 
believe it is a hardware issue.  I started copying with dd_rescue, 
telling it to halt on all errors, and managed to copy 2.5 gigs before I 
was convinced there weren't any critical hardware issues.  i have about 
50 gigs I can scrap together for recovery on multiple disks... but the 
partition was 180(ish) GiB.  Only 38 GiB was used though.  Here's what 
happens:


####### Pass 0 #######
Could not find a hash in use. Using "r5"
####### Pass 1 #######
####### Pass 2 #######
Will rebuild the filesystem (/dev/hdb1) tree
Will put log info to 'stdout'
Replaying journal..
Reiserfs journal '/dev/hdb1' in blocks [18..8211]: 0 transactions replayed
###########
reiserfsck --rebuild-tree started at Mon Jun  7 01:15:31 2004
###########
Pass 0:
Loading on-disk bitmap .. ok, 27189 blocks marked used
Skipping 9701 blocks (super block, journal, bitmaps) 17488 blocks will 
be read

**does the reading here**
"r5" hash is selected
Flushing..finished
    Read blocks (but not data blocks) 17488
        Leaves among those 0
        Objectids found 2


Pass 1 (will try to insert 0 leaves):
Looking for allocable blocks .. finished

Flushing..finished
    0 leaves read
        0 inserted

Flushing..finished

No reiserfs metadata found.  If you are sure that you had the reiserfs
on this partition...



Will running with -S get it back?  Like I said, I'm almost certain 
there's no hardware issue, and I'd really like to get this data back.


Thanks a lot,
    Casey Callendrello

Re: can't ls, then rebuild-tree fails, no hardware error (kow)

[Vitaly Fertman]

On Monday 07 June 2004 09:25, Casey Callendrello wrote:
> Hello all,
>
>     I'd been using Reiserfs on a 180 GiB partition.  One day my computer
> hung while I was out.  I came home and had to hard-reboot it.  Fine.
> (Unfortunately, this is not the / partition but /home!  egad!)
> Unfortunately, when I rebooted I could not ls in the /home directory,
> which is mount --bind 'ed to /mnt/big/home (big? get it!)  sensing a
> major problem, I had to hard-reboot to kill the ls which was hung on
> waiting for IO (with no errors in /var/log/messages or the error console).
>
> I ran fsck and it found some errors, so I ran fsck --fix-fixable.  The
> whole directory structure was there, I saw it flash by during the fsck
> (well, as far as I could tell.  There certainly wasn't anything obvious
> missing).  So, I remount it and try to cat a file I know is in there,
> and it shows up fine.  I feel better.
>
> But, I still can't ls.  I can't log in with a non-root user since it
> hangs waiting for IO.  I figure it's a problem with the tree (I
> shouldn't jump to conclusions).  So I run rebuild-tree and it fails.
> Now i'm up *** creek.
> I don't have a hard drive large enough to copy this to.  I also don't
> believe it is a hardware issue.  I started copying with dd_rescue,
> telling it to halt on all errors, and managed to copy 2.5 gigs before I
> was convinced there weren't any critical hardware issues.  i have about
> 50 gigs I can scrap together for recovery on multiple disks... but the
> partition was 180(ish) GiB.  Only 38 GiB was used though.  Here's what
> happens:
>
>
> Will rebuild the filesystem (/dev/hdb1) tree
>
> No reiserfs metadata found.  If you are sure that you had the reiserfs
> on this partition...

Is this /dev/hdb1 the original /home ?
So you had it mountable and you could cat some files there but could
not ls the root of the /home, getting no error into the syslog (?) and
then you run reiserfsck --rebuild-tree and it found no metadata?

It looks like before rebuilding the partition table was altered. Did 
you use some partition editor or another program that could change 
the partition table? To tell for sure I need to look at the partition itself.

-- 
Thanks,
Vitaly Fertman

Re: can't ls, then rebuild-tree fails, no hardware error (kow)

[Casey Callendrello]

Vitaly Fertman wrote:

>Is this /dev/hdb1 the original /home ?
>So you had it mountable and you could cat some files there but could
>not ls the root of the /home, getting no error into the syslog (?) and
>then you run reiserfsck --rebuild-tree and it found no metadata?
>
>It looks like before rebuilding the partition table was altered. Did 
>you use some partition editor or another program that could change 
>the partition table? To tell for sure I need to look at the partition itself.
>
>  
>
I figured out what is probably the error.
The drive is 180 gigs, but the mobo only recognizes 137.  However, linux 
recognized 180 gigs through and through...
and now is only recognizing 137.

However, Soyo says it should be readable.  I guess not.  I'm going to 
try an -S and give up otherwise, creating a new 137 gig partition.

Thanks, Casey.