[linux-lvm] reiserfs oops at init

[linux-lvm] reiserfs oops at init

[Chris Worley]

I installed SuSE 7.3 on a new disk at the end of December, using LVM 
and a reiserfs as a root partition.

I've been having problems with the ide controller and this disk, 
causing many lock-ups.

I put the new disk on the secondary controller (which, lvm shouldn't 
care that the disk has changed from hdb to hdd), and the DMA/interrupt 
problems seem to end... and a new problem begins: I can't boot this 
partition as root.  The kernel still boots off /dev/hda1, and the 
initrd runs, and finds the moved volume without any errors, but "init" 
immediately oopses (after reporting it's version number) with "invalid 
operand: 0000"...  I'm guessing this is coming from the reiserfs 
(sorry for asking this question to the lvm group, I don't know a good 
reiserfs list server).

I can boot with "init=/bin/sh".  I've run "reiserfsck" with the 
"--check" and "--fix*" options... and there are a lot of errors, and 
it ends saying there are ten errors that can be fixed with 
"--rebuild-tree" (which I can't run with the disk mounted). the 
"--rebuild-sb" option trys to run, but immediately aborts.

Booted with "init=/bin/sh", I can remount the file system read/write, 
and everything looks fine.  Given that "init" won't run from the 
command line (it will run, but only to communicate with the 
controlling "init") it's hard to figure out why we're oopsing.

Booted with "init=/bin/sh", I can run vgscan.  It gives a "modprobe" 
error about device major number 35(??? that's a midi device), but 
still says it's writing the /etc/lv* files.  But, if I then run 
lvscan, vgdisplay, etc..., it says the kernel is out of sync with the 
file systems and I need to rerun vgscan (which doesn't seem to help).

Any ideas on how to proceed?

Thanks,

Chris

Re: [linux-lvm] reiserfs oops at init

[Chris Mason]

On Monday, January 14, 2002 08:36:36 AM -0700 Chris Worley
<cworley@liberate.com> wrote:

> 
> I installed SuSE 7.3 on a new disk at the end of December, using LVM and a
> reiserfs as a root partition.
> 
> I've been having problems with the ide controller and this disk, causing
> many lock-ups.
> 
> I put the new disk on the secondary controller (which, lvm shouldn't care
> that the disk has changed from hdb to hdd), and the DMA/interrupt problems
> seem to end... and a new problem begins: I can't boot this partition as
> root.  The kernel still boots off /dev/hda1, and the initrd runs, and finds
> the moved volume without any errors, but "init" immediately oopses (after
> reporting it's version number) with "invalid operand: 0000"...  I'm
> guessing this is coming from the reiserfs (sorry for asking this question
> to the lvm group, I don't know a good reiserfs list server).

reiserfs-list@namesys.com ;-)

> 
> I can boot with "init=/bin/sh".  I've run "reiserfsck" with the "--check"
> and "--fix*" options... and there are a lot of errors, and it ends saying
> there are ten errors that can be fixed with "--rebuild-tree" (which I can't
> run with the disk mounted). the "--rebuild-sb" option trys to run, but
> immediately aborts.

Ok, first grab the latest reiserfsck 3.x.1a from
ftp.namesys.com/pub/reiserfsprogs.  What you probably need to do is boot off
the suse CD, and then you'll be able to run reiserfsck --rebuild-tree on your
unmounted root drive.

I'd put the new fsck on a floppy, and run it from there.

If you want to send the errors from reiserfsck --check, I can confirm things
a little better.

-chris

RE: [linux-lvm] reiserfs oops at init

[Chris Worley]

Chris Mason wrote:

> Ok, first grab the latest reiserfsck 3.x.1a from


The SuSE 7.3 version was 3.x.0k-pre9-28, which, knowing SuSE, was 
probably a heavily patched version of the reiser distribution.

I've upgraded to the version you suggest.


> If you want to send the errors from reiserfsck --check, I can confirm things
> a little better.


It's going to be a big file... do you just want the tail?

Thanks,

Chris

RE: [linux-lvm] reiserfs oops at init

[Chris Mason]

On Monday, January 14, 2002 10:48:49 AM -0700 Chris Worley
<cworley@liberate.com> wrote:

> Chris Mason wrote:
> 
>> Ok, first grab the latest reiserfsck 3.x.1a from
> 
> 
> The SuSE 7.3 version was 3.x.0k-pre9-28, which, knowing SuSE, was probably
> a heavily patched version of the reiser distribution.

Not too heavily, we try to keep the utils patches to a minimum ;-)

> 
> I've upgraded to the version you suggest.
> 
> 
>> If you want to send the errors from reiserfsck --check, I can confirm
>> things a little better.
> 
> 
> It's going to be a big file... do you just want the tail?
> 

Lets start with the tail, my internet connection isn't that hot right now.

-chris

Re: [linux-lvm] reiserfs oops at init

[Chris Worley]

Chris Mason wrote:

>>>If you want to send the errors from reiserfsck --check, I can confirm
>>>things a little better.
>>>
>>
>>It's going to be a big file... do you just want the tail?
>>
>>
> 
> Lets start with the tail, my internet connection isn't that hot right now.


Every error I could see was of the form:

bad_directory_item: block 141874: item 698385 744805 0x1 DIR (3), len 
350, location 2576 entry count 14, fsck need 0, format old has entry 
".." with wrong deh_state 140744 - expected 4 - fixed

This goes on for ~11MB.

The tail is:

free block count 3306944 mismatches with a correct one 4783644.
on-disk bitmap does not match to the correct one. 1476700 bits differ

With the new reiserfs version, it actually says "fixed", the older 
version didn't say that (except at the beginning), and I don't think 
they really fixed anything.

Although those "free block count mismatch/correct" numbers look like 
they did for all previous runs, this run with the new software said 
nothing of (something like) "10 ... must be fixed with 
--rebuild-tree".  It still doesn't boot.

On to the "--rebuild-tree" pass.  Wish me luck.

Chris