Re: Kernel dump/segmentation failt

[Green <green@namesys.com>]

Hello!

   I am looking at the kernel code that does this trap riight now.
   Can you tell me which "mv" commandlines you tried?
   Have you tried "mv /mnt/lost+found/somename /mnt/temp" for example?
   (not there is no ".." path element used at all).
   Can you create directory in the root of that partition?
   You sade if you change stuff in /lost+fond, it crashes.
   What other operations besides rename have you tried?

   I think it may be possible to hand-correct directory structure
   and add missing ".", ".." and fix wrong offset in lost+found
   directory entry. (I have your 2 days old metadata snapshot and looking at
   it).

Bye,
    Oleg
On Thu, Apr 25, 2002 at 02:48:28PM +0200, Anders Widman wrote:
> True, it takes 12 hours minimum to run --rebuild-tree. Also. I did
> actually run --rebuild-tree (without -S) after I ran with -S. The
> reason for this is that --check reported an error that could only be
> fixed with "--rebuild-tree". The error is now gone, but I can't change
> the data on the volume (in the lost+found folder).
> 
> Sure.. Backup would be best thing, but it seems a bit difficult to
> find 800GB backup media.
> 
> 
> 
> //Anders
> 
> 
> 
> > Hello!
> 
> >    Given the size of your volume, I guess you do not want to run --rebuild-tree
> >    (without -S) again.
> >    And the only thing that you can do yourself is to backup all the stuff,
> >    reformat and then put all the stuff back. (you said yo can access it in
> >    read-only mode)
> 
> > Bye,
> >     Oleg
> > On Thu, Apr 25, 2002 at 02:27:02PM +0200, Anders Widman wrote:
> >>   Yes, it seems so. I just had a look, and the "." and ".." are
> >>   missing. Is there anything that I can do at this point to fix this
> >>   problem myself?
> >> 
> >> 
> >> //Anders
> >> 
> >> 
> >> 
> >> > Hello!
> >> 
> >> >    Thank you for the report.
> >> >    This is a very similar crashdump like you've sent some time ago
> >> >    (BUG at namei.c:1160). The problem itself seems to be related to
> >> >    missing "." and ".." entries in lost+found dir (though it is not very
> >> >    clear at this point why these entries were not created by reiserfsck).
> >> 
> >> >    We do not have any real solution right now, unfortunatelly.
> >> >    But we are working on it.
> >> 
> >> > Bye,
> >> >     Oleg
> >> > On Thu, Apr 25, 2002 at 01:57:17PM +0200, Anders Widman wrote:
> >> >> 
> >> >>   After running reiserfsck -S --ebuild-tree to restore a filesystem I
> >> >>   get lots of folders and files in lost+found named with numbers like
> >> >>   2_1328. This is itself isn't the biggest problems. All subfolders
> >> >>   are named correktly, as the files in those subfolders.
> >> >> 
> >> >>   However, when I try to change anything in the lost+found folder I
> >> >>   get a segmentation fault and a kernel dump. I am using latest
> >> >>   reiserfs tools with kernel 2.4.18-6 and kernel 2.4.18-7:
> >> >> 
> >> >>   ~~~~~~~~~~~~~~~~~~~~
> >> >> Apr 25 12:24:22 server kernel: kernel BUG at namei.c:1160!
> >> >> Apr 25 12:24:22 server kernel: invalid operand: 0000
> >> >> Apr 25 12:24:22 server kernel: CPU:    0
> >> >> Apr 25 12:24:22 server kernel: EIP:    0010:[<c017ad24>]    Not tainted  
> >> >> Apr 25 12:24:22 server kernel: EFLAGS: 00010297
> >> >> Apr 25 12:24:22 server kernel: eax: 00000000   ebx: dac57d1c   ecx: 062bb000   edx: dac57e0c
> >> >> Apr 25 12:24:22 server kernel: esi: ffffffff   edi: dac57c7c   ebp: dab22c80   esp: dac57c40
> >> >> Apr 25 12:24:22 server kernel: ds: 0018   es: 0018   ss: 0018 
> >> >> Apr 25 12:24:22 server kernel: Process mv (pid: 1536, stackpage=dac57000)
> >> >> Apr 25 12:24:22 server kernel: Stack: 00000002 00000000 dac57ca8 dac57d48 00000000 00000000 00000000 c0298f48
> >> >> Apr 25 12:24:22 server kernel:        00000001 00000039 000f65d1 df638000 00001000 c0137088 00003a00 db084ba0
> >> >> Apr 25 12:24:22 server kernel:        00000003 daf9a060 00000001 daf9a418 00000008 00000002 daf9aeb0 00000000
> >> >> Apr 25 12:24:22 server kernel: Call Trace: [<c0137088>] [<c018af9b>] [<c018abff>] [<c0137088>] [<c0179773>]
> >> >> Apr 25 12:24:22 server kernel:    [<c01663e1>] [<c0140590>] [<c012657e>] [<c0126e7a>] [<c0140b79>] [<c013e120>]
> >> >> Apr 25 12:24:22 server kernel:    [<c013ee99>] [<c0140d7a>] [<c013dedd>] [<c010899b>]
> >> >> Apr 25 12:24:22 server kernel:
> >> >> Apr 25 12:24:22 server kernel: Code: 0f 0b 88 04 8e 49 29 c0 8b 84 24 cc 01 00 00 8b 94 c4 d0 01
> >> >>   ~~~~~~~~~~~~~~~~~~~~
> >> >> If I am mounting as read only I can read all files without a kernel
> >> >> dump or segmentation fault.
> >> >> 
> >> >> //Anders
> >> >> 
> >> 
>