Still wrong reported file size after --rebuild-tree

Still wrong reported file size after --rebuild-tree

[Philippe Gramoullé]

Hi,

I just wanted to let you know that after rebuilding the tree for a 500Go
volume, i still have wrong file size reported :

/dev/sdb1            572418604 174714996 397703608  31% /storage

# ls -l
total 4928006
-r-x-wx-wx    1 daemon   324231      32768 Jan  1  1970 25s.jpg
-r-x--x---    1 19415372 19415373 83388383504122186 Aug 13  1970 25s_lb.jpg
-r--rw-rw-    1 19415354 19415355 83388306194710840 Aug 13  1970 25slight.jpg
-r---wxrwx    1 19415331 19415332 83388207410463009 Aug 13  1970 25slight_ln.jpg
-rw-rw-r--    1 500      500         19616 Dec 13  2000 AnLake.class
-rw-rw-r--    1 500      500         12576 Dec 13  2000 AnLake.jar
-rw-rw-r--    1 500      500          2251 Dec 13  2000 Lware.class
-r----x-w-    1 19415310 19415311 83388117216149772 Aug 13  1970 anfy.class
-rw-rw-r--    1 500      500          5699 Dec 13  2000 anlake.txt
-rw-rw-r--    1 500      500          1963 Dec 13  2000 chesterfield.htm
-rw-rw-r--    1 500      500         21143 Dec 13  2000 filter.jpg

I'll make metadata available as soon as debugreiserfs is finised.
(reiserfsck used was 3.6.4 ( 3.6.4 pre3)

Thanks,

Philippe.

Re: Still wrong reported file size after --rebuild-tree

[Todd Lyons]

[-- Attachment #1: Type: text/plain, Size: 1179 bytes --]

Philippe Gramoullé wanted us to know:

>I just wanted to let you know that after rebuilding the tree for a 500Go
>volume, i still have wrong file size reported :
>/dev/sdb1            572418604 174714996 397703608  31% /storage

Hmmm, ours showed 96% or something like that.  I'll check to see if it's
accurate.

>-r-x--x---    1 19415372 19415373 83388383504122186 Aug 13  1970 25s_lb.jpg
>-r--rw-rw-    1 19415354 19415355 83388306194710840 Aug 13  1970 25slight.jpg
>-r---wxrwx    1 19415331 19415332 83388207410463009 Aug 13  1970 25slight_ln.jpg

We had these too.  Humongous uid and gid numbers.

>I'll make metadata available as soon as debugreiserfs is finised.
>(reiserfsck used was 3.6.4 ( 3.6.4 pre3)

I really don't think it's a requirement to see our metadata as our
problems were caused by faulty hardware and not any bugs in the fs
itself.
-- 
Blue skies...		Todd
| Get a bigger hammer!   | Security is like an onion.  It's made   |
| http://www.mrball.net  | up of several layers and makes you cry. |
| http://faq.mrball.net  |                         --Howard Chu    |
   Linux kernel 2.5.44   3 users,  load average: 0.00, 0.00, 0.00

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: Still wrong reported file size after --rebuild-tree

[Philippe Gramoullé]

On Wed, 30 Oct 2002 19:57:36 -0800
Todd Lyons <todd@mrball.net> wrote:

  |  Philippe Gramoullé wanted us to know:
  |  
  |  >I just wanted to let you know that after rebuilding the tree for a 500Go
  |  >volume, i still have wrong file size reported :
  |  >/dev/sdb1            572418604 174714996 397703608  31% /storage
  |  
  |  Hmmm, ours showed 96% or something like that.  I'll check to see if it's
  |  accurate.

oh, the above figures are correct. I only wanted to report about the bogus files sizes
below.

  |  
  |  >-r-x--x---    1 19415372 19415373 83388383504122186 Aug 13  1970 25s_lb.jpg
  |  >-r--rw-rw-    1 19415354 19415355 83388306194710840 Aug 13  1970 25slight.jpg
  |  >-r---wxrwx    1 19415331 19415332 83388207410463009 Aug 13  1970 25slight_ln.jpg
  |  
  |  We had these too.  Humongous uid and gid numbers.

Though we do use 32 bits UIDs & GIDs ( up to 2 Millions or so right now ), we don't
have ones up to 19415331 :o) Not yet :o)
  |  
  |  >I'll make metadata available as soon as debugreiserfs is finised.
  |  >(reiserfsck used was 3.6.4 ( 3.6.4 pre3)
  |  
  |  I really don't think it's a requirement to see our metadata as our
  |  problems were caused by faulty hardware and not any bugs in the fs
  |  itself.

In my case, i don't think the hardware is broken but there have been some poweroffs that seem
to have damaged the FS somewhat. Raid volumes have been checked, disks are ok , logical volume
consistency as well, so it's up to ReiserFS team to say if something's wrong with reporting those
file sizes even after a full --rebuild-tree.

Thanks,

Philippe

Re: Still wrong reported file size after --rebuild-tree

[Nikita Danilov]

Philippe Gramoullé writes:
 > On Wed, 30 Oct 2002 19:57:36 -0800
 > Todd Lyons <todd@mrball.net> wrote:
 > 
 >   |  Philippe Gramoullé wanted us to know:
 >   |  
 >   |  >I just wanted to let you know that after rebuilding the tree for a 500Go
 >   |  >volume, i still have wrong file size reported :
 >   |  >/dev/sdb1            572418604 174714996 397703608  31% /storage
 >   |  
 >   |  Hmmm, ours showed 96% or something like that.  I'll check to see if it's
 >   |  accurate.
 > 
 > oh, the above figures are correct. I only wanted to report about the bogus files sizes
 > below.
 > 
 >   |  
 >   |  >-r-x--x---    1 19415372 19415373 83388383504122186 Aug 13  1970 25s_lb.jpg
 >   |  >-r--rw-rw-    1 19415354 19415355 83388306194710840 Aug 13  1970 25slight.jpg
 >   |  >-r---wxrwx    1 19415331 19415332 83388207410463009 Aug 13  1970 25slight_ln.jpg
 >   |  
 >   |  We had these too.  Humongous uid and gid numbers.

They are also -consequent-, and Oleg found that they look conspicuous
when printed in hex (with file sizes).

 > 
 > Though we do use 32 bits UIDs & GIDs ( up to 2 Millions or so right now ), we don't
 > have ones up to 19415331 :o) Not yet :o)
 >   |  
 >   |  >I'll make metadata available as soon as debugreiserfs is finised.
 >   |  >(reiserfsck used was 3.6.4 ( 3.6.4 pre3)
 >   |  
 >   |  I really don't think it's a requirement to see our metadata as our
 >   |  problems were caused by faulty hardware and not any bugs in the fs
 >   |  itself.
 > 
 > In my case, i don't think the hardware is broken but there have been some poweroffs that seem
 > to have damaged the FS somewhat. Raid volumes have been checked, disks are ok , logical volume
 > consistency as well, so it's up to ReiserFS team to say if something's wrong with reporting those
 > file sizes even after a full --rebuild-tree.
 > 
 > Thanks,
 > 
 > Philippe

Nikita.

Re: Still wrong reported file size after --rebuild-tree

[Todd Lyons]

[-- Attachment #1: Type: text/plain, Size: 904 bytes --]

Nikita Danilov wanted us to know:

> >   |  >-r-x--x---    1 19415372 19415373 83388383504122186 Aug 13  1970 25s_lb.jpg
> >   |  >-r--rw-rw-    1 19415354 19415355 83388306194710840 Aug 13  1970 25slight.jpg
> >   |  >-r---wxrwx    1 19415331 19415332 83388207410463009 Aug 13  1970 25slight_ln.jpg
> >   |  We had these too.  Humongous uid and gid numbers.
>They are also -consequent-, and Oleg found that they look conspicuous
>when printed in hex (with file sizes).

Now that you mention it, yes.  Interesting that they all have the same
date and how the perms have a funny pattern too.  
-- 
Blue skies...		Todd
| Get a bigger hammer!   |   Never mind me.  I'm speaking out of  |
| http://www.mrball.net  |   my /dev/ass anyway.                  |
| http://faq.mrball.net  |                        --Gash Teshome  |
   Linux kernel 2.5.44   2 users,  load average: 0.00, 0.00, 0.00

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: Still wrong reported file size after --rebuild-tree

[Philippe Gramoullé]

Hi,

Please, find the metata here :

http://philippemm.multimania.com/metadata-2.4.19-pre6.bz2

ahem , it's a 100Mo file :o)

Thanks,

Philippe


<-------------debugreiserfs, 2002------------->
reiserfsprogs 3.6.4-pre3

Loading on-disk bitmap .. 43683118 bits set - done
super block..ok
bitmaps..(4368).. ok
journal (from 18 to 8210)..ok
Super block, bitmaps, journal - 12562 blocks - done, 43670556 blocks left
0%....20%....40%....60%....80%....                                                            left 0, 7989 /sec
Packed 647143 blocks:
        compessed 637664
        full blocks 9479
                leaves with broken block head 15
                corrupted leaves 0
                internals 4509
                descriptors 0
data packed with ratio 0.11

Re: Still wrong reported file size after --rebuild-tree

[Vitaly Fertman]

Hi, 

On Wednesday 30 October 2002 21:30, Philippe Gramoullé wrote:
> Hi,
>
> I just wanted to let you know that after rebuilding the tree for a 500Go
> volume, i still have wrong file size reported :
>
> /dev/sdb1            572418604 174714996 397703608  31% /storage
>
> # ls -l
> total 4928006
> -r-x-wx-wx 1 daemon   324231      32768 Jan  1  1970 25s.jpg
> -r-x--x--- 1 19415372 19415373 83388383504122186 Aug 13 1970 25s_lb.jpg
> -r--rw-rw- 1 19415354 19415355 83388306194710840 Aug 13 1970 25slight.jpg 
> -r---wxrwx 1 19415331 19415332 83388207410463009 Aug 13 1970 25slight_ln.jpg 
>

ReiserFS can handle 64 bit file sizes, so from the reiserfsck point of view  
these are correct sizes. 

-- 

Thanks,
Vitaly Fertman

Re: Still wrong reported file size after --rebuild-tree

[Hans Reiser]

Vitaly Fertman wrote:

>Hi, 
>
>On Wednesday 30 October 2002 21:30, Philippe Gramoullé wrote:
>  
>
>>Hi,
>>
>>I just wanted to let you know that after rebuilding the tree for a 500Go
>>volume, i still have wrong file size reported :
>>
>>/dev/sdb1            572418604 174714996 397703608  31% /storage
>>
>># ls -l
>>total 4928006
>>-r-x-wx-wx 1 daemon   324231      32768 Jan  1  1970 25s.jpg
>>-r-x--x--- 1 19415372 19415373 83388383504122186 Aug 13 1970 25s_lb.jpg
>>-r--rw-rw- 1 19415354 19415355 83388306194710840 Aug 13 1970 25slight.jpg 
>>-r---wxrwx 1 19415331 19415332 83388207410463009 Aug 13 1970 25slight_ln.jpg 
>>
>>    
>>
>
>ReiserFS can handle 64 bit file sizes, so from the reiserfsck point of view  
>these are correct sizes. 
>
>  
>
You mean to say that you haven't written code to check that file sizes 
are correct, yes?

-- 
Hans

Re: Still wrong reported file size after --rebuild-tree

[Vitaly Fertman]

On Saturday 02 November 2002 00:07, Hans Reiser wrote:
> Vitaly Fertman wrote:
> >Hi,
> >
> >On Wednesday 30 October 2002 21:30, Philippe Gramoullé wrote:
> >>Hi,
> >>
> >>I just wanted to let you know that after rebuilding the tree for a 500Go
> >>volume, i still have wrong file size reported :
> >>
> >>/dev/sdb1            572418604 174714996 397703608  31% /storage
> >>
> >># ls -l
> >>total 4928006
> >>-r-x-wx-wx 1 daemon   324231      32768 Jan  1  1970 25s.jpg
> >>-r-x--x--- 1 19415372 19415373 83388383504122186 Aug 13 1970 25s_lb.jpg
> >>-r--rw-rw- 1 19415354 19415355 83388306194710840 Aug 13 1970 25slight.jpg
> >>-r---wxrwx 1 19415331 19415332 83388207410463009 Aug 13 1970
> >> 25slight_ln.jpg
> >
> >ReiserFS can handle 64 bit file sizes, so from the reiserfsck point of
> > view these are correct sizes.
>
> You mean to say that you haven't written code to check that file sizes
> are correct, yes?

No, I meant that it is considered as a legal situation to have a size in 
a file statdata greater then its real body size. Thus you can truncate 
the file to the greater size. 

However, if you are sure that you do not have files with such holes at 
the end you can use --ajust-file-sizes option in rebuild and fix-fixable
modes - fixable here - to correct such statdata's file sizes to file body 
sizes. (Sorry that I have not mentioned about this option in my previous 
email). 

-- 

Thanks,
Vitaly Fertman

Re: Still wrong reported file size after --rebuild-tree

[reiser]

Vitaly Fertman wrote:

>On Saturday 02 November 2002 00:07, Hans Reiser wrote:
>  
>
>>Vitaly Fertman wrote:
>>    
>>
>>>Hi,
>>>
>>>On Wednesday 30 October 2002 21:30, Philippe Gramoullé wrote:
>>>      
>>>
>>>>Hi,
>>>>
>>>>I just wanted to let you know that after rebuilding the tree for a 500Go
>>>>volume, i still have wrong file size reported :
>>>>
>>>>/dev/sdb1            572418604 174714996 397703608  31% /storage
>>>>
>>>># ls -l
>>>>total 4928006
>>>>-r-x-wx-wx 1 daemon   324231      32768 Jan  1  1970 25s.jpg
>>>>-r-x--x--- 1 19415372 19415373 83388383504122186 Aug 13 1970 25s_lb.jpg
>>>>-r--rw-rw- 1 19415354 19415355 83388306194710840 Aug 13 1970 25slight.jpg
>>>>-r---wxrwx 1 19415331 19415332 83388207410463009 Aug 13 1970
>>>>25slight_ln.jpg
>>>>        
>>>>
>>>ReiserFS can handle 64 bit file sizes, so from the reiserfsck point of
>>>view these are correct sizes.
>>>      
>>>
>>You mean to say that you haven't written code to check that file sizes
>>are correct, yes?
>>    
>>
>
>No, I meant that it is considered as a legal situation to have a size in 
>a file statdata greater then its real body size. Thus you can truncate 
>the file to the greater size. 
>
>However, if you are sure that you do not have files with such holes at 
>the end you can use --ajust-file-sizes option in rebuild and fix-fixable
>modes - fixable here - to correct such statdata's file sizes to file body 
>sizes. (Sorry that I have not mentioned about this option in my previous 
>email). 
>
>  
>
Ah, interesting.

Hans