ext3/reiserfs: ext3fine, reiser got OOPS!

ext3/reiserfs: ext3fine, reiser got OOPS!

[Nico Schottelius]

Hello guys!

I 've been using ext3 and reiserfs for somedays with
2.4.7. Using mkreiserfs I recieved some null pointer
problems and recieved a kernel oops.
While ext3 is mounted as fast as ext2, reiserfs
seems is slower.

ext3, 10 GB: ~ 0.5 seconds
reisferfs 10 GB: ~ 3-5 seconds

Generating the journal with tune2fs was
again much faster than mkreiserfs. But I think
this is because reiser creates a new fs, wherefore
mkfs.ext2 did that before.

While running there occured some problems
with reiserfs.

I am wondering that reiserfs first got into the kernel
before ext3!


Nico

Re: ext3/reiserfs: ext3fine, reiser got OOPS!

[Chris Wedgwood]

On Sun, Aug 05, 2001 at 04:24:14AM +0200, Nico Schottelius wrote:

    I 've been using ext3 and reiserfs for somedays with 2.4.7. Using
    mkreiserfs I recieved some null pointer problems and recieved a
    kernel oops.

Odd --- nobody else has reported this.  Can you plese supply more
details (ksymoops) such that these bugs may be fixed.

    While ext3 is mounted as fast as ext2, reiserfs seems is slower.

Slower to mount? Or slower to use?

    ext3, 10 GB: ~ 0.5 seconds reisferfs 10 GB: ~ 3-5 seconds

Probably journal replay, still, you might have slow disks.  A journal
reply for me of 60+ events takes about 1 second on a single spindle
(SCSI, U160).

    Generating the journal with tune2fs was again much faster than
    mkreiserfs. But I think this is because reiser creates a new fs,
    wherefore mkfs.ext2 did that before.

Yes.


Do it really matter (within reason) which fs mounts and is made
faster?  It's not something you do every other minute.



    While running there occured some problems with reiserfs.

Such as?

    I am wondering that reiserfs first got into the kernel before
    ext3!

It was ready first (mostly).





   --cw

Re: ext3/reiserfs: ext3fine, reiser got OOPS!

[Nico Schottelius]

>     I 've been using ext3 and reiserfs for somedays with 2.4.7. Using
>     mkreiserfs I recieved some null pointer problems and recieved a
>     kernel oops.
>
> Odd --- nobody else has reported this.  Can you plese supply more
> details (ksymoops) such that these bugs may be fixed.

Retrying it...
Seems like it is _not_ reproduceable currently!
The mkreiserfs ran through it without any problems!
If I can reproduce it again, I will send all
informations.


>     While ext3 is mounted as fast as ext2, reiserfs seems is slower.
>
> Slower to mount? Or slower to use?

Mount and delete.

reiserfs

I created 200000 empty files, and then deleted them.

ext2: creating took a very long time, I aborted (was about 7 minutes)
reiserfs: creating was done in about 3 minutes

ext2: deleting started with find | xargs rm, no problem.

reiserfs: started deleting, recieved NULL pointer
(kernel oops)



>     ext3, 10 GB: ~ 0.5 seconds reisferfs 10 GB: ~ 3-5 seconds
>
> Probably journal replay, still, you might have slow disks.

It's been the same disk.
and i don't think it was a reply, because I created the
filesystem some seconds before.

> A journal
> reply for me of 60+ events takes about 1 second on a single spindle
> (SCSI, U160).

Hmm.... I used udma33 :(

> Do it really matter (within reason) which fs mounts and is made
> faster?  It's not something you do every other minute.

You are right. I was only surprised that the so called
fast reiserfs takes soo long at mounting.

>     While running there occured some problems with reiserfs.
>
> Such as?

It was impossible to delete the files on the one partition.
(see above) The problem is, after I reformated the partition
it was not reproduceable.

It seems there are some strange things in reiserfs, which
happen at special situations.


Nico

Re: ext3/reiserfs: ext3fine, reiser got OOPS!

[Chris Wedgwood]

On Sun, Aug 05, 2001 at 03:16:40PM +0200, Nico Schottelius wrote:

    Mount and delete.

    reiserfs

    I created 200000 empty files, and then deleted them.

In many cases (ie. lots of files in a small number of directories),
reiserfs is _much_ faster then ext2 and ext3.

This is fairly well established.

For example, on a rather humble machine (dual PIII 866) the time to
extract a while bunch of files, including compression time and delete
them (whilst the machine was doing other things):


cw:tty2@tapu(cw)$ /usr/bin/time tx archive/freedb/freedb-complete-20010617.tar.bz2
128.87user 152.89system 5:33.36elapsed 84%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (2141major+2086minor)pagefaults 0swaps

cw:tty2@tapu(cw)$ du -s freedb-complete-20010617/
1380821 freedb-complete-20010617

cw:tty2@tapu(cw)$ find freedb-complete-20010617/ -type f -print|wc -l
 350065

cw:tty2@tapu(cw)$ ls -F freedb-complete-20010617/
COPYING  blues/      country/  folk/  misc/    reggae/  soundtrack/
README   classical/  data/     jazz/  newage/  rock/

<stress memory and cause swapping to purge much of page cache>

cw:tty2@tapu(cw)$ /usr/bin/time rm -rf freedb-complete-20010617/
0.68user 66.48system 3:22.71elapsed 33%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (111major+17minor)pagefaults 0swaps

Hot-cache, the delete time is about 15seconds from memory (I posted
results to the reiserfs list a few weeks back).




  --cw