Tree issues, fsck output

Tree issues, fsck output

[mjt]

Hi!

***** Tree Traverse Pass: scanning the reiser4 internal tree.
FSCK: Node (2013064): The last key [377f5f:4(FB):2e616339375f63:3a45d8:99] in 2 the node is greater then the right delimiting key [2fa6e:4(FB):6c69626d70342e:1e88ad:1f7c].
FSCK: Node (469959), item (28), unit (0): Points to the node [2013064] with wrong delimiting keys. The whole subtree is skipped.
Error: Can't open node 2014146.
FSCK: Node (2014146): failed to open the node pointed by the node (469959), item(29), unit (0) on the level (2). The whole subtree is skipped.
Error: Can't open node 2014147.
FSCK: Node (2014147): failed to open the node pointed by the node (469959), item(30), unit (0) on the level (2). The whole subtree is skipped.
FSCK: Node (469959), item (31), unit (0): Points to the block (2014148) which isin the tree already. The whole subtree is skipped.

How deep is the shit I'm in? :P

What brought this on?

I'll just send this post now, I'll send a follow-up when I'm re-doing
fsck with some switches. This was just a basic fsck /dev/hda1, so maybe
no need to panic yet...

-- 
mjt

Re: Tree issues, fsck output

[mjt]

On Tue, Jun 29, 2004 at 11:19:56PM +0300, Markus Törnqvist wrote:
>
>I'll just send this post now, I'll send a follow-up when I'm re-doing
>fsck with some switches. This was just a basic fsck /dev/hda1, so maybe
>no need to panic yet...

So I proceeded to --build-fs

FSCK: Node (481195), item (5), [2fa6e:6c69626d70342e:1e88ae] (stat40): wrong
bytes (939), Fixed to (0).
FSCK: Directory [2fa68:496e7075742e64:2fa6e]: can't find the object
[2fa6e:6c69626d70342e:1e88ad] pointed by the entry [libmp4.so]. Entry is
removed.

Anyway, that finished with a consistent fs..

Now that I'm back in my real environment:
mjt@shrike:~$ xmms &
[1] 3752
mjt@shrike:~$ /usr/lib/xmms/Input/libtonegen.so: invalid ELF header
/usr/lib/xmms/Input/libwav.so: invalid ELF header
/usr/lib/xmms/Effect/libecho.so: invalid ELF header
/usr/lib/xmms/Effect/libstereo.so: invalid ELF header
/usr/lib/xmms/Effect/libvoice.so: invalid ELF header

So I try to reinstall xmms:
mjt@shrike:~$ sudo apt-get install xmms
Reading Package Lists... 93%

And it sticks there AGAIN!

So now I'll just compile the new auto-snapshot and see what happens...

-- 
mjt

Re: Tree issues, fsck output

[mjt]

On Wed, Jun 30, 2004 at 09:45:47AM +0300, Markus Törnqvist wrote:
>So now I'll just compile the new auto-snapshot and see what happens...

It did not work.

http://mjt.nysv.org/reiser/Reiser4-2004.06.29/
Pictures once again courtesy of my trusty phone, so the quality truly
does suck.

So, err, I'm doing a plain fsck now again, to see if that is the result
of some corruption, but it went to the extremely non-verbose semantic
tree pass...

I'm at a loss here, if the new kernel doesn't bother to boot and the
old one doesn't really work.

Vitaly, do you have some ideas for me? Do you want my metadata somewhere?

I doubt anyone checked yet if the mainline 2.6.7 port of the snapshot
I used was to blame, but at least now I'm trying with a standard one..

If the semantic tree pass goes cleanly and the fs is consistent, I will
try older auto-snapshots, despite the fact that the earlier mm's are
not highly recommended.

Thanks!

-- 
mjt

Re: Tree issues, fsck output

[Vitaly Fertman]

On Wednesday 30 June 2004 13:03, Markus Törnqvist wrote:
> On Wed, Jun 30, 2004 at 09:45:47AM +0300, Markus Törnqvist wrote:
> >So now I'll just compile the new auto-snapshot and see what happens...
>
> It did not work.
>
> http://mjt.nysv.org/reiser/Reiser4-2004.06.29/
> Pictures once again courtesy of my trusty phone, so the quality truly
> does suck.
>
> So, err, I'm doing a plain fsck now again, to see if that is the result
> of some corruption, but it went to the extremely non-verbose semantic
> tree pass...

it prints all found corruptions. could you send them ?

btw, I hvae already added gauge to the semantic pass. Hopefully will
release another progs snapshot or internal-snapshot on days.

> I'm at a loss here, if the new kernel doesn't bother to boot and the
> old one doesn't really work.

if I recall correctly the only corruption was wrong directory entry, right?
or it fould some more while you were sleeping?

> Vitaly, do you have some ideas for me? Do you want my metadata somewhere?

could you run fsck.reiser4 in check mode again after rebuilding to be sure 
that evth has been fixed. If you have matadata packed before rebuilding, 
not after, and you can provide them for downloading, it would be useful.

> I doubt anyone checked yet if the mainline 2.6.7 port of the snapshot
> I used was to blame, but at least now I'm trying with a standard one..
>
> If the semantic tree pass goes cleanly and the fs is consistent, I will
> try older auto-snapshots, despite the fact that the earlier mm's are
> not highly recommended.

to reproduce this we need to know which steps lead to such corruptions.

-- 
Thanks,
Vitaly Fertman

Re: Tree issues, fsck output

[mjt]

On Wed, Jun 30, 2004 at 01:24:34PM +0400, Vitaly Fertman wrote:
>it prints all found corruptions. could you send them ?

The thing is, now that 2.6.7-mm4, latest auto-snapshot, failed,
I didn't get any errors with fsck... So it seems my filesystem
is all alright, at least for the time being.

>btw, I hvae already added gauge to the semantic pass. Hopefully will
>release another progs snapshot or internal-snapshot on days.

That would be great!
Any hints on what else to expect beyond the gauge?-)

>if I recall correctly the only corruption was wrong directory entry, right?
>or it fould some more while you were sleeping?

Nope, nada, the last I pasted to you, libmp4, was also the last failure.
Then I think I posted here that xmms wouldn't reinstall and there were
bad ELF headers.

>could you run fsck.reiser4 in check mode again after rebuilding to be sure 
>that evth has been fixed. If you have matadata packed before rebuilding, 
>not after, and you can provide them for downloading, it would be useful.

I'm taking a debugfs.reiser4 -P /dev/hda1 now.
Then I will run additional stuff, but it still seems like the FS is ok....

>> If the semantic tree pass goes cleanly and the fs is consistent, I will
>> try older auto-snapshots, despite the fact that the earlier mm's are
>> not highly recommended.
>to reproduce this we need to know which steps lead to such corruptions.

Yeah, but it ain't easy.
I'd like to hear what changes were made when porting the snapshot
to mainline. If that could possibly induce corruption.

Anyway, I think this part of the discussion is redundant, as the big
problem now remaining is that the 2004.06.29 auto-snapshot doesn't
boot. If it presents these problems, with sibling check and all that
enabled, I will have more details for you.

-- 
mjt

Re: Tree issues, fsck output

[mjt]

On Wed, Jun 30, 2004 at 12:42:25PM +0300, Markus Törnqvist wrote:
>
>I'm taking a debugfs.reiser4 -P /dev/hda1 now.
>Then I will run additional stuff, but it still seems like the FS is ok....

Yeah, I have the metadata dump now, and I will send it to you if you
feel you need it, but the additional fsck reported clean results.

I will proceed to compiling a new kernel, with an older auto-snap.

Thanks!

-- 
mjt