possible bug - fsck shows perfect results, linux refuses to mount

possible bug - fsck shows perfect results, linux refuses to mount

[Robin H. Johnson]

[-- Attachment #1: Type: text/plain, Size: 3387 bytes --]

Greetings,

A server of mine suddenly started acting very wonky and spontaneously
rebooting. I've been pulling my hair out trying to narrow down what the
problem was.

Today after one of the reboots, it showed a lot of damage to several of
the data partitions (not the system ones luckily), so I ran fsck on them
to see about cleaning up. For two of them, /dev/hde[56], I only had to
get down to --rebuild-tree to fix them up. However for the last one,
/dev/hdb1, looks to be far worse. 

The system refused to mount it originally, so I ran just plain
--fix-fixable. It showed nothing wrong at all. By a fluke of terminals,
I have a copy of this first output
[http://www.orbis-terrarum.net/~robbat2/reiserfs/hdb1.first].

However the system still refused to mount the drive, showing this in
syslog:
Mar 30 22:14:22 [kernel] read_super_block: can't find a reiserfs
filesystem on (dev 03:41, block 64, size 1024)
Mar 30 22:14:22 [kernel] read_super_block: can't find a reiserfs
filesystem on (dev 03:41, block 8, size 1024)

Mount showed this:
server1 kernel # mount /dev/hdb1
mount: wrong fs type, bad option, bad superblock on /dev/hdb1,
       or too many mounted file systems

This baffled me, so I decided it wouldn't hurt to run --rebuild-sb and
--rebuild-tree. For rebuild-sb, I don't have the full output, but I did
get: "Super block seems to be correct"

I ran rebuild-tree, and saw no errors. 

The drive still refused to mount.

I dug in fsck.reiserfs --help, and saw '--scan-whole-partition'. Tried
--rebuild-tree with that on. It showed a LOT of stuff about StatDatas,
and completed successfully.

Still no joy in mounting it.

I repeated this, and still got a massive piles of errors
[http://www.orbis-terrarum.net/~robbat2/reiserfs/hdb1.later].

Kernel: 2.4.20-gentoo-r1 (Amongst the patches are: 3.5G-address-space,
rmap-14a, futuxes, imon, missing.list, preempt-ac, supermount, freeswan,
and lots of netfilter stuff)
however I don't see much at a glance that should affect things I think)
Hardware: Pentium III (Katmai) 500mhz CPU (replaced with identical known
good one in hardware testing). 512Mb PC133 SDRAM (replaced with identical known
good part).
Motherboard: Abit AH6 Rev 1.01 (was Rev 1.00 previously) [Intel 440bx].
reiserfsprogs: 3.6.4

I'm not exactly sure what had the partition open at the time, but I am
reasonably confidident that it was NFS.

The partition (~10GiB in size) hosts several dedicated game server
binaries and related files (NFS, exported readonly+async) to other
diskless machines that actually run the games.  (Logs are written to a
different NFS server).

I'm going to wipe the seemingly corrupt contents of the drive later this
week if I don't manage to fix it. But as with all data it would make
life as a SysAdmin so much easier if data just came back magically 8-).
(I think I've spent too much of the weekend playing AD&D).

I just find that there is something definetly wrong if fsck says the
partition is fine, but Linux refuses to mount it. Either this is a bug
in Linux, or the reiserfsprogs. Either way, somebody has a bug :-)

-- 
Robin Hugh Johnson
E-Mail     : robbat2@orbis-terrarum.net
Home Page  : http://www.orbis-terrarum.net/?l=people.robbat2
ICQ#       : 30269588 or 41961639
GnuPG FP   : 11AC BA4F 4778 E3F6 E4ED  F38E B27B 944E 3488 4E85

[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]

Re: possible bug - fsck shows perfect results, linux refuses to mount

[Oleg Drokin]

Hello!

On Sun, Mar 30, 2003 at 11:22:50PM -0800, Robin H. Johnson wrote:
> The system refused to mount it originally, so I ran just plain
> --fix-fixable. It showed nothing wrong at all. By a fluke of terminals,
> I have a copy of this first output
> [http://www.orbis-terrarum.net/~robbat2/reiserfs/hdb1.first].
> However the system still refused to mount the drive, showing this in
> syslog:
> Mar 30 22:14:22 [kernel] read_super_block: can't find a reiserfs
> filesystem on (dev 03:41, block 64, size 1024)
> Mar 30 22:14:22 [kernel] read_super_block: can't find a reiserfs
> filesystem on (dev 03:41, block 8, size 1024)

This is indeed strange.

> The drive still refused to mount.
> I dug in fsck.reiserfs --help, and saw '--scan-whole-partition'. Tried
> --rebuild-tree with that on. It showed a LOT of stuff about StatDatas,
> and completed successfully.

Well, this is likely to destroy data, but still it should be mountable at the point of
completion.

Can you please make a metadata dump for us?
"debugreiserfs -p /dev/hdb1 | bzip2 -9c >metadata.bz2" and make this file
available for us to download.

> I just find that there is something definetly wrong if fsck says the
> partition is fine, but Linux refuses to mount it. Either this is a bug
> in Linux, or the reiserfsprogs. Either way, somebody has a bug :-)

Sure, and we are interested in resolving the problem.

Thank you.

Bye,
    Oleg

Re: possible bug - fsck shows perfect results, linux refuses to mount

[Robin H. Johnson]

[-- Attachment #1: Type: text/plain, Size: 1876 bytes --]

On Mon, Mar 31, 2003 at 11:34:46AM +0400, Oleg Drokin wrote:
> Hello!
> 
> On Sun, Mar 30, 2003 at 11:22:50PM -0800, Robin H. Johnson wrote:
> > The drive still refused to mount.
> > I dug in fsck.reiserfs --help, and saw '--scan-whole-partition'. Tried
> > --rebuild-tree with that on. It showed a LOT of stuff about StatDatas,
> > and completed successfully.
> Well, this is likely to destroy data, but still it should be mountable at the point of
> completion.
Now I get to prove to my boss why I wanted to spend so much on backups
:-).

> Can you please make a metadata dump for us?
> "debugreiserfs -p /dev/hdb1 | bzip2 -9c >metadata.bz2" and make this file
> available for us to download.
Stderr output of debugreiserfs here:
http://www.orbis-terrarum.net/~robbat2/reiserfs/debugreiserfs.run
metadata file:
http://www.orbis-terrarum.net/~robbat2/reiserfs/metadata.bz2

> > I just find that there is something definitely wrong if fsck says the
> > partition is fine, but Linux refuses to mount it. Either this is a bug
> > in Linux, or the reiserfsprogs. Either way, somebody has a bug :-)
> Sure, and we are interested in resolving the problem.
Same here.

> Thank you.
No, most definitely thank you. Now I have more ammunition against my boss
as to why Open Source is better. You guys really rock. I'd never get a
response from any other vendor this quickly (even though we pay some of
them a fair sum).

A question for you, so I can support things. Is there a list somewhere
of features requested, and how much you money you want to be paid to add
them in? Especially the smaller features (hint hint).

-- 
Robin Hugh Johnson
E-Mail     : robbat2@orbis-terrarum.net
Home Page  : http://www.orbis-terrarum.net/?l=people.robbat2
ICQ#       : 30269588 or 41961639
GnuPG FP   : 11AC BA4F 4778 E3F6 E4ED  F38E B27B 944E 3488 4E85

[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]

Re: possible bug - fsck shows perfect results, linux refuses to mount

[Oleg Drokin]

Hello!

On Sun, Mar 30, 2003 at 11:49:57PM -0800, Robin H. Johnson wrote:
> > Can you please make a metadata dump for us?
> > "debugreiserfs -p /dev/hdb1 | bzip2 -9c >metadata.bz2" and make this file
> > available for us to download.
> Stderr output of debugreiserfs here:
> http://www.orbis-terrarum.net/~robbat2/reiserfs/debugreiserfs.run

"bread: Cannot read the block (18)."
Well, now this is something.
Do you have I/O errors messages in kernel logs, or do you know another reason for
it to fail reading 18th block?
Perhaps try to run "badblocks -b 4096" to find out if there are more blocks
that cannot be read?

> > Thank you.
> No, most definitely thank you. Now I have more ammunition against my boss
> as to why Open Source is better. You guys really rock. I'd never get a
> response from any other vendor this quickly (even though we pay some of
> them a fair sum).

We accept money too ;)

> A question for you, so I can support things. Is there a list somewhere
> of features requested, and how much you money you want to be paid to add
> them in? Especially the smaller features (hint hint).

Well, there is http://namesys.com/support.html , as of features, usually
we expect customers to come up with features they need. If you lack something
in reiserfs and you are willing to pay, we can implement it.
If you are happy with the current features-set, why to pay for features that
would be never used?

Bye,
    Oleg

Re: possible bug - fsck shows perfect results, linux refuses to mount

[Soeren Sonnenburg]

On Mon, 31 Mar 2003, Oleg Drokin wrote:

> Hello!
> 
> On Sun, Mar 30, 2003 at 11:49:57PM -0800, Robin H. Johnson wrote:
> > > Can you please make a metadata dump for us?
> > > "debugreiserfs -p /dev/hdb1 | bzip2 -9c >metadata.bz2" and make this file
> > > available for us to download.
> > Stderr output of debugreiserfs here:
> > http://www.orbis-terrarum.net/~robbat2/reiserfs/debugreiserfs.run
> 
> "bread: Cannot read the block (18)."
> Well, now this is something.
> Do you have I/O errors messages in kernel logs, or do you know another reason for
> it to fail reading 18th block?
> Perhaps try to run "badblocks -b 4096" to find out if there are more blocks
> that cannot be read?
> 
> > > Thank you.
> > No, most definitely thank you. Now I have more ammunition against my boss
> > as to why Open Source is better. You guys really rock. I'd never get a
> > response from any other vendor this quickly (even though we pay some of
> > them a fair sum).
> 
> We accept money too ;)
> 
> > A question for you, so I can support things. Is there a list somewhere
> > of features requested, and how much you money you want to be paid to add
> > them in? Especially the smaller features (hint hint).
> 
> Well, there is http://namesys.com/support.html , as of features, usually
> we expect customers to come up with features they need. If you lack something
> in reiserfs and you are willing to pay, we can implement it.
> If you are happy with the current features-set, why to pay for features that
> would be never used?
> 
> Bye,
>     Oleg
> 

Today a raid0 (stripe set) crashed here (there is a bug in western
digital disks - they sometimes need a power cycle to accept commands
again) and I ran the latest reiserfsck on /dev/md0 with --rebuild-sb
--rebuild-tree.

It successfully finished rebuilding the tree but now I observe the same
thing robin observes: I cannot mount the device... so I ran reiserfsck
/dev/md0 again and it complained that there is no superblock on the
drive. So I tried to create one with reiserfstune --make-standard-journal
/dev/md0 but that one just segfaults (unfortunately I have gcc on that
machine as /usr /var are on the stripeset :-( ).
Also I tried to to reiserfsck --rebuild-tree -rebuild-sb again, but that
only segfault too ...

This is really strange since reiserfsck was really enumerating over all
the files and everything seemed to be fine after the first rebuild-tree.

So robin, you are not alone :-)

Hmm, I am running kernel 2.4.21pre6

Soeren.

Re: possible bug - fsck shows perfect results, linux refuses to mount

[Robin H. Johnson]

[-- Attachment #1: Type: text/plain, Size: 2168 bytes --]

On Mon, Mar 31, 2003 at 12:00:14PM +0400, Oleg Drokin wrote:
> > http://www.orbis-terrarum.net/~robbat2/reiserfs/debugreiserfs.run
> "bread: Cannot read the block (18)."
> Well, now this is something.
> Do you have I/O errors messages in kernel logs, or do you know another reason for
> it to fail reading 18th block?
Nope, no I/O messages in the logs at all.

> Perhaps try to run "badblocks -b 4096" to find out if there are more blocks
> that cannot be read?
I had previously scanned the partition only, but I've scanned the entire
drive now, just to be entirely certain, and there are no bad blocks at
all. I've tested with both the standard readonly test, and the
non-destructive read/write.

The drive is a FUJITSU MPC3102AT-E, one of the ones without the firmware
issues (these drives pre-Feb 2001 had a firmware bug).

> We accept money too ;)
I'll keep that in mind then.

> > A question for you, so I can support things. Is there a list somewhere
> > of features requested, and how much you money you want to be paid to add
> > them in? Especially the smaller features (hint hint).
> Well, there is http://namesys.com/support.html , as of features, usually
> we expect customers to come up with features they need. If you lack something
> in reiserfs and you are willing to pay, we can implement it.
> If you are happy with the current features-set, why to pay for features that
> would be never used?
I've been trying and using reiserfs since before it was officially in
2.4, and this is actually the first real issue I've hit with it. I had
just wondered if you had some feature list of minor things you would
like, but never had time for. Eg nice to have, but not crucial.
Random thing I saw while reading ReiserFS docs, like maybe the repacker?
But I suspect that may be a much larger task (and I certianly don't want
my drives and RAID arrays ever getting that full in the first place).

-- 
Robin Hugh Johnson
E-Mail     : robbat2@orbis-terrarum.net
Home Page  : http://www.orbis-terrarum.net/?l=people.robbat2
ICQ#       : 30269588 or 41961639
GnuPG FP   : 11AC BA4F 4778 E3F6 E4ED  F38E B27B 944E 3488 4E85

[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]

Re: possible bug - fsck shows perfect results, linux refuses to mount

[Oleg Drokin]

Hello!

On Mon, Mar 31, 2003 at 10:34:30AM +0200, Soeren Sonnenburg wrote:

> Today a raid0 (stripe set) crashed here (there is a bug in western
> digital disks - they sometimes need a power cycle to accept commands
> again) and I ran the latest reiserfsck on /dev/md0 with --rebuild-sb
> --rebuild-tree.
> It successfully finished rebuilding the tree but now I observe the same
> thing robin observes: I cannot mount the device... so I ran reiserfsck
> /dev/md0 again and it complained that there is no superblock on the
> drive. So I tried to create one with reiserfstune --make-standard-journal

Well, if there is no superblock, it will certainly refuse to mount.

> /dev/md0 but that one just segfaults (unfortunately I have gcc on that
> machine as /usr /var are on the stripeset :-( ).

Is it just segfaults without any messages?
Anything interesting in the logs?
What reiserfsprogs version do you use?

> This is really strange since reiserfsck was really enumerating over all
> the files and everything seemed to be fine after the first rebuild-tree.

Hm, might it be that your harddrive eaten the superblock that it was supposed to write
to disk? ;)

> Hmm, I am running kernel 2.4.21pre6

Hm, new IDE changes? What if you try same stuff on 2.4.20?

Bye,
    Oleg

Re: possible bug - fsck shows perfect results, linux refuses to mount

[Soeren Sonnenburg]

On Mon, 31 Mar 2003, Oleg Drokin wrote:

> Hello!
> 
> On Mon, Mar 31, 2003 at 10:34:30AM +0200, Soeren Sonnenburg wrote:
> 
> > Today a raid0 (stripe set) crashed here (there is a bug in western
> > digital disks - they sometimes need a power cycle to accept commands
> > again) and I ran the latest reiserfsck on /dev/md0 with --rebuild-sb
> > --rebuild-tree.
> > It successfully finished rebuilding the tree but now I observe the same
> > thing robin observes: I cannot mount the device... so I ran reiserfsck
> > /dev/md0 again and it complained that there is no superblock on the
> > drive. So I tried to create one with reiserfstune --make-standard-journal
> 
> Well, if there is no superblock, it will certainly refuse to mount.
> 
> > /dev/md0 but that one just segfaults (unfortunately I have gcc on that
> > machine as /usr /var are on the stripeset :-( ).
> 
> Is it just segfaults without any messages?

right. it says no standard journal somewhen before crashing but thats
it...

> Anything interesting in the logs?

nothing. the disks are fine (only two partitions work as raid0 there, some
others are accessible and mountable and working nicely :)

> What reiserfsprogs version do you use?

reiserfsprogs-3.6.5-pre2

> > This is really strange since reiserfsck was really enumerating over all
> > the files and everything seemed to be fine after the first rebuild-tree.
> 
> Hm, might it be that your harddrive eaten the superblock that it was supposed to write
> to disk? ;)

don't think so... while this drive _HAS_ a certain firmware bug this
definitely is something else (at least this time).
 
> > Hmm, I am running kernel 2.4.21pre6
> 
> Hm, new IDE changes? What if you try same stuff on 2.4.20?

I was running 2.4.21pre5-ac5 for around 14days without any trouble. No I
guess not.. and all the other 5 ide disks in the machine did not happen to
have any problems...

 > Bye,
>     Oleg

Thanks,
Soeren 

Re: possible bug - fsck shows perfect results, linux refuses to mount

[Oleg Drokin]

Hello!

On Mon, Mar 31, 2003 at 12:37:58AM -0800, Robin H. Johnson wrote:
> On Mon, Mar 31, 2003 at 12:00:14PM +0400, Oleg Drokin wrote:
> > > http://www.orbis-terrarum.net/~robbat2/reiserfs/debugreiserfs.run
> > "bread: Cannot read the block (18)."
> > Well, now this is something.
> > Do you have I/O errors messages in kernel logs, or do you know another reason for
> > it to fail reading 18th block?
> Nope, no I/O messages in the logs at all.

Hm...
And you can do "dd if=/dev/hdb1 of=/tmp/somefile bs=4096 count=2 skip=18" and it does
not produce any errors?
What if you run that debugreiserfs command again and see if it complains about being unable
to read 18th block?

> > Perhaps try to run "badblocks -b 4096" to find out if there are more blocks
> > that cannot be read?
> I had previously scanned the partition only, but I've scanned the entire
> drive now, just to be entirely certain, and there are no bad blocks at
> all. I've tested with both the standard readonly test, and the
> non-destructive read/write.

Hmm, but debugreiserfs cannot read it for some reason, this is weird.

> > > A question for you, so I can support things. Is there a list somewhere
> > > of features requested, and how much you money you want to be paid to add
> > > them in? Especially the smaller features (hint hint).
> > Well, there is http://namesys.com/support.html , as of features, usually
> > we expect customers to come up with features they need. If you lack something
> > in reiserfs and you are willing to pay, we can implement it.
> > If you are happy with the current features-set, why to pay for features that
> > would be never used?
> I've been trying and using reiserfs since before it was officially in
> 2.4, and this is actually the first real issue I've hit with it. I had
> just wondered if you had some feature list of minor things you would
> like, but never had time for. Eg nice to have, but not crucial.
> Random thing I saw while reading ReiserFS docs, like maybe the repacker?

Well, repacker, delayed block allocations, fine-grained locking is coming to my
head right now, but all of this is going to be in reiser4, and Hans intends reiser4 to
replace reiserfs eventually, so he feels that spending money/time on such things is
just a waste of resources. Probably you can look at the list of all of new features of
reiser4 and think of those as of features missed in reiserfs.
On the other hand if somebody actually needs these features in reiserfs right now,
just vote with the money ;)

> But I suspect that may be a much larger task (and I certianly don't want
> my drives and RAID arrays ever getting that full in the first place).

Repacker is not that trivialindeed, if you read this list for some time already, you might have
noticed discussion in which Hans and Chris have agreed that while repacker is cool and can be
implemented, nobody knows what layout for the data should be chosen.

Bye,
    Oleg

Re: possible bug - fsck shows perfect results, linux refuses to mount

[Oleg Drokin]

Hello!

On Mon, Mar 31, 2003 at 10:48:51AM +0200, Soeren Sonnenburg wrote:
> > > /dev/md0 but that one just segfaults (unfortunately I have gcc on that
> > > machine as /usr /var are on the stripeset :-( ).
> > Is it just segfaults without any messages?
> right. it says no standard journal somewhen before crashing but thats
> it...

Can you quote the exact message?

> > What reiserfsprogs version do you use?
> reiserfsprogs-3.6.5-pre2

And can you try reiserfsprogs-3.6.5? hmm, your gcc is not accessible. I can supply you
with the binary.

You can supply -l /path/somefile to reiserfsck and it will put its log to
/path/somefile instead of stdout. Also if you then capture 
stderr with mouse (cut'n'paste) or by using script(1), or in some other way,
and then in case of crash you will supply us with those stdout/stderr outputs
(and core file, if you'd run our binary)

Bye,
    Oleg

Re: possible bug - fsck shows perfect results, linux refuses to mount

[Robin H. Johnson]

[-- Attachment #1: Type: text/plain, Size: 2810 bytes --]

On Mon, Mar 31, 2003 at 12:56:52PM +0400, Oleg Drokin wrote:
> > Nope, no I/O messages in the logs at all.
> Hm...
> And you can do "dd if=/dev/hdb1 of=/tmp/somefile bs=4096 count=2
> skip=18" and it does not produce any errors?
Works fine, no errors. I copied the first 256mb of the drive to a
ramdisk a few times to be certain as well.

> What if you run that debugreiserfs command again and see if it
> complains about being unable to read 18th block?
No change in the output of debugreiserfs.

> > I had previously scanned the partition only, but I've scanned the entire
> > drive now, just to be entirely certain, and there are no bad blocks at
> > all. I've tested with both the standard readonly test, and the
> > non-destructive read/write.
> Hmm, but debugreiserfs cannot read it for some reason, this is weird.
Yes, very.

> Well, repacker, delayed block allocations, fine-grained locking is
> coming to my head right now, but all of this is going to be in
> reiser4, and Hans intends reiser4 to replace reiserfs eventually, so
> he feels that spending money/time on such things is just a waste of
> resources.
I'd agree with Hans on that. Duplication of effort is most often not
worthwhile. 

> Probably you can look at the list of all of new features of reiser4
> and think of those as of features missed in reiserfs.  On the other
> hand if somebody actually needs these features in reiserfs right now,
> just vote with the money ;)
I don't need them myself. ReiserFS works out great on a large number of
places that I use it, including a 1.2TiB IDE RAID3 array. I'm eagerly
anticapting Reiser4. I am however waiting until it's offical declared more
stable before I try it out.

> > But I suspect that may be a much larger task (and I certianly don't want
> > my drives and RAID arrays ever getting that full in the first place).
> Repacker is not that trivialindeed, if you read this list for some
> time already, you might have noticed discussion in which Hans and
> Chris have agreed that while repacker is cool and can be implemented,
> nobody knows what layout for the data should be chosen.
Nope, I just subscribed to the list today, when I had this problem.

I've taken the system home for now, I so I can hack on it some more.
Mostly to find what was causing the reboots previously, but also so this
can get worked on.

However, I realize some of my assignments aren't waiting for me. Argh,
nasty evil signals and systems AM/FM mathematics. I'm a full-time
SysAdmin, and an almost full-time student as well.

-- 
Robin Hugh Johnson
E-Mail     : robbat2@orbis-terrarum.net
Home Page  : http://www.orbis-terrarum.net/?l=people.robbat2
ICQ#       : 30269588 or 41961639
GnuPG FP   : 11AC BA4F 4778 E3F6 E4ED  F38E B27B 944E 3488 4E85

[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]

Re: possible bug - fsck shows perfect results, linux refuses to mount

[Oleg Drokin]

Hello!

On Mon, Mar 31, 2003 at 01:16:21AM -0800, Robin H. Johnson wrote:

> > What if you run that debugreiserfs command again and see if it
> > complains about being unable to read 18th block?
> No change in the output of debugreiserfs.

Ok, plase run it under strace, kind of like this:
"strace -o /tmp/out debugreiserfs -p /dev/hdb1 >/dev/null" and
send me this /tmp/out file.

> > > But I suspect that may be a much larger task (and I certianly don't want
> > > my drives and RAID arrays ever getting that full in the first place).
> > Repacker is not that trivialindeed, if you read this list for some
> > time already, you might have noticed discussion in which Hans and
> > Chris have agreed that while repacker is cool and can be implemented,
> > nobody knows what layout for the data should be chosen.
> Nope, I just subscribed to the list today, when I had this problem.
> I've taken the system home for now, I so I can hack on it some more.
> Mostly to find what was causing the reboots previously, but also so this
> can get worked on.

Hmm, I also wonder what happens if you dd the whole partition (/dev/hdb1) to
some file and then run debugreiserfs on that file will it succeed?
If not, then probably you can compress the file and somehow make it available for
download?
If it will succeed, then there is certainly some interaction between harware and
debugreiserfs is going on.

Bye,
    Oleg

Re: possible bug - fsck shows perfect results, linux refuses to mount

[Robin H. Johnson]

[-- Attachment #1: Type: text/plain, Size: 1811 bytes --]

On Mon, Mar 31, 2003 at 01:24:54PM +0400, Oleg Drokin wrote:
> > No change in the output of debugreiserfs.
> Ok, plase run it under strace, kind of like this:
> "strace -o /tmp/out debugreiserfs -p /dev/hdb1 >/dev/null" and
> send me this /tmp/out file.
http://www.orbis-terrarum.net/~robbat2/reiserfs/debugreiserfs.strace.bz2

> Hmm, I also wonder what happens if you dd the whole partition (/dev/hdb1) to
> some file and then run debugreiserfs on that file will it succeed?
Same output from debugreiserfs. 
http://www.orbis-terrarum.net/~robbat2/reiserfs/debugreiserfs.run.fromfile
I did an strace as well on this (it IS different somewhere, didn't look
at the output myself tho, just the md5sum):
http://www.orbis-terrarum.net/~robbat2/reiserfs/debugreiserfs-fromfile.strace.bz2

MD5SUMS for everything:
http://www.orbis-terrarum.net/~robbat2/reiserfs/MD5SUMS

You can just browse that directory to grab things again.

> If not, then probably you can compress the file and somehow make it available for
> download?
> If it will succeed, then there is certainly some interaction between harware and
> debugreiserfs is going on.
Ok, I think we can rule out hardware since it still happens anyway.

Next step is getting that data to you. I'm running bzip2 -9vvkz on the
dd'd contents of the drive now, I expect that to take a few hours.

Depending on the size of the output, I'll either see about getting the
data up on a relatively high speed site somewhere, or splitting it into
two parts and sending them to you on DVD-R if you can accept those?

-- 
Robin Hugh Johnson
E-Mail     : robbat2@orbis-terrarum.net
Home Page  : http://www.orbis-terrarum.net/?l=people.robbat2
ICQ#       : 30269588 or 41961639
GnuPG FP   : 11AC BA4F 4778 E3F6 E4ED  F38E B27B 944E 3488 4E85

[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]

Re: possible bug - fsck shows perfect results, linux refuses to mount

[Oleg Drokin]

Hello!

On Mon, Mar 31, 2003 at 02:17:14AM -0800, Robin H. Johnson wrote:
> On Mon, Mar 31, 2003 at 01:24:54PM +0400, Oleg Drokin wrote:
> > > No change in the output of debugreiserfs.
> > Ok, plase run it under strace, kind of like this:
> > "strace -o /tmp/out debugreiserfs -p /dev/hdb1 >/dev/null" and
> > send me this /tmp/out file.
> http://www.orbis-terrarum.net/~robbat2/reiserfs/debugreiserfs.strace.bz2

_llseek(0, 73728, 0xbffff450, SEEK_SET) = -1 ESPIPE (Illegal seek)

Hm, seeking on stdin instead of file? That's strange.

> > If not, then probably you can compress the file and somehow make it available for
> > download?
> > If it will succeed, then there is certainly some interaction between harware and
> > debugreiserfs is going on.
> Ok, I think we can rule out hardware since it still happens anyway.

Sure.

> Next step is getting that data to you. I'm running bzip2 -9vvkz on the
> dd'd contents of the drive now, I expect that to take a few hours.
> Depending on the size of the output, I'll either see about getting the
> data up on a relatively high speed site somewhere, or splitting it into
> two parts and sending them to you on DVD-R if you can accept those?

Also you can just give me a shell on your box, so that I can tweak and
recompile reiserfsprogs (btw, you are not using debugreiserfs from
reiserfsprogs v3.6.5), also I need r/w access to the file containing
copy of partition data.
In order for me to be able to verify mountability of partition I need
a way to mount/umount it).

Bye,
    Oleg

Re: possible bug - fsck shows perfect results, linux refuses to mount

[Robin H. Johnson]

[-- Attachment #1: Type: text/plain, Size: 1692 bytes --]

On Mon, Mar 31, 2003 at 02:36:23PM +0400, Oleg Drokin wrote:
> > http://www.orbis-terrarum.net/~robbat2/reiserfs/debugreiserfs.strace.bz2
> _llseek(0, 73728, 0xbffff450, SEEK_SET) = -1 ESPIPE (Illegal seek)
> Hm, seeking on stdin instead of file? That's strange.
Very.

> > Next step is getting that data to you. I'm running bzip2 -9vvkz on the
> > dd'd contents of the drive now, I expect that to take a few hours.
> > Depending on the size of the output, I'll either see about getting the
> > data up on a relatively high speed site somewhere, or splitting it into
> > two parts and sending them to you on DVD-R if you can accept those?
> Also you can just give me a shell on your box, so that I can tweak and
> recompile reiserfsprogs (btw, you are not using debugreiserfs from
> reiserfsprogs v3.6.5), also I need r/w access to the file containing
> copy of partition data.
> In order for me to be able to verify mountability of partition I need
> a way to mount/umount it).
I'm afraid I can't do that Dave. Shells are out of the question. It
opens up too many possible security holes. Paranoid I am.

At the present rate, it looks like it will be at least 50% compression,
I'm thinking more since the drive was about 60% full.

~5gb would be an acceptable download I think, split into 10x512Mb parts,
using MD5SUMS. 

I was going to upgrade to it, but it seems Gentoo currently has some
issues with it, so I'm holding off on it.

-- 
Robin Hugh Johnson
E-Mail     : robbat2@orbis-terrarum.net
Home Page  : http://www.orbis-terrarum.net/?l=people.robbat2
ICQ#       : 30269588 or 41961639
GnuPG FP   : 11AC BA4F 4778 E3F6 E4ED  F38E B27B 944E 3488 4E85

[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]

Re: possible bug - fsck shows perfect results, linux refuses to mount

[Oleg Drokin]

Hello!

On Mon, Mar 31, 2003 at 03:13:32AM -0800, Robin H. Johnson wrote:
> > > Next step is getting that data to you. I'm running bzip2 -9vvkz on the
> > > dd'd contents of the drive now, I expect that to take a few hours.
> > > Depending on the size of the output, I'll either see about getting the
> > > data up on a relatively high speed site somewhere, or splitting it into
> > > two parts and sending them to you on DVD-R if you can accept those?
> > Also you can just give me a shell on your box, so that I can tweak and
> > recompile reiserfsprogs (btw, you are not using debugreiserfs from
> > reiserfsprogs v3.6.5), also I need r/w access to the file containing
> > copy of partition data.
> > In order for me to be able to verify mountability of partition I need
> > a way to mount/umount it).
> I'm afraid I can't do that Dave. Shells are out of the question. It

Ok, I fully understand that.

> opens up too many possible security holes. Paranoid I am.
> At the present rate, it looks like it will be at least 50% compression,
> I'm thinking more since the drive was about 60% full.

Ok. We will look at code meanwhile to see why it might have done lseek on stdin.

> ~5gb would be an acceptable download I think, split into 10x512Mb parts,
> using MD5SUMS. 
> I was going to upgrade to it, but it seems Gentoo currently has some
> issues with it, so I'm holding off on it.

You can uild it yourself if you want, or I can supply you with binary.

Bye,
    Oleg

Re: possible bug - fsck shows perfect results, linux refuses to mount

[Oleg Drokin]

Hello!

On Mon, Mar 31, 2003 at 03:13:32AM -0800, Robin H. Johnson wrote:

> At the present rate, it looks like it will be at least 50% compression,
> I'm thinking more since the drive was about 60% full.

Ok, so we have found that your filesystem have relocated journal
enabled for some reason.
"reiserfstune --make-journal-standard /dev/hdb1" should make the partition mountable.
Another question is how did you got it to this state. We are failing to see it yet.
Do you have any idea?
BTW we are still interested in copy of your filesystem.

Bye,
    Oleg

Re: possible bug - fsck shows perfect results, linux refuses to mount

[Hans Reiser]

Oleg Drokin wrote:

>  
>
>>A question for you, so I can support things. Is there a list somewhere
>>of features requested, and how much you money you want to be paid to add
>>them in? Especially the smaller features (hint hint).
>>    
>>
>
>Well, there is http://namesys.com/support.html , as of features, usually
>we expect customers to come up with features they need. If you lack something
>in reiserfs and you are willing to pay, we can implement it.
>If you are happy with the current features-set, why to pay for features that
>would be never used?
>
>Bye,
>    Oleg
>
>
>  
>
This was the wrong answer.;-)

There are lots of features needed but not implemented.  You could 
sponsor our integrating support for badblock handling to the current 
reiser3 and reiser4 for $1000.

-- 
Hans