[gentoo-dev] [Fwd: [reiserfs-list] glibc check problem on reiserfs]

[gentoo-dev] [Fwd: [reiserfs-list] glibc check problem on reiserfs]

[Achim Gottinger]

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



[-- Attachment #2: Type: message/rfc822, Size: 3110 bytes --]

From: Russell Coker <russell@coker.com.au>
To: achim@gentoo.org, 320095285153-0001@t-online.de (Achim Gottinger), reiserfs-list@namesys.com
Subject: Re: [reiserfs-list] glibc check problem on reiserfs
Date: Mon, 19 Feb 2001 10:09:12 +0100
Message-ID: <01021910091202.00754@lyta>

On Saturday 17 February 2001 23:16, Achim Gottinger wrote:
> If I compile the latest glibc-2.2.2 on a system with a 2.4.1-ac17 or
> 2.4.0-ac13 kernel on a reiserfs partition,
> "make check" failed during io/test-lfs.
> I took a quick look into the code and increased a timeout value from 20
> to 40 seconds. After that it worked.
> The test creates a tmp file using mkstemp64 and hangs if it tries to
> write "hello" to that test file.
> I tested test-lfs with an ext2 partition mounted on /tmp and it runs in
> under a second (28 sec's on reiserfs).
> I attach the code for test-lfs.c to that mail but it requires the
> glibc-2.2.2. sources to compile.

ReiserFS has a bug in that it will take large amounts of CPU time to create 
holes in files.  On my machine (p3-650) I can do that operation in about 10 
seconds.  If you get a faster CPU then ReiserFS will pass this test.

But really this is a known bug in ReiserFS that the developers will fix after 
they fix the more important bugs.

As long as fsck is eating file systems I hope that the developers don't 
bother about this issue too much...

-- 
http://www.coker.com.au/bonnie++/     Bonnie++ hard drive benchmark
http://www.coker.com.au/postal/       Postal SMTP/POP benchmark
http://www.coker.com.au/projects.html Projects I am working on
http://www.coker.com.au/~russell/     My home page

[gentoo-dev] [Fwd: [reiserfs-list] glibc check problem on reiserfs]

[Achim Gottinger]

[-- Attachment #1: Type: text/plain, Size: 34 bytes --]

Just for your information.

achim~

[-- Attachment #2: Type: message/rfc822, Size: 3323 bytes --]

From: Hans Reiser <reiser@namesys.com>
To: Xuan Baldauf <xuan--reiserfs@baldauf.org>
Cc: Chris Wedgwood <cw@f00f.org>, achim@gentoo.org, reiserfs-list@namesys.com
Subject: Re: [reiserfs-list] glibc check problem on reiserfs
Date: Wed, 21 Feb 2001 17:22:22 +0300
Message-ID: <3A93CF1E.49C022C7@namesys.com>

Xuan Baldauf wrote:
> 
> Chris Wedgwood wrote:
> 
> > On Tue, Feb 20, 2001 at 08:32:15AM +0100, Achim Gottinger wrote:
> >
> >     On an ext2 partition this test runs in under a second and it does
> >     not hurt people using ext2 to increase the timeout as long as no
> >     other issues fail the test.
> >
> > The fact it takes over 10 times longer for reiserfs to be indicates
> > it's behavior is not only undesirable but a bug.
> >
> > Sparse files are real -- the are used often in some places. This
> > really does need to be 'fixed' some way or another.
> >
> >   --cw
> 
> How is reiserfs creating files with only ones '1' (e.g. 0xFF-bytes)
> compared to ext2? :-) Both cases (0x00-bytes and 0xFF-bytes) are special
> cases. I don't think that it's a bug to be not optimized for some special
> cases, it's just undesirable. How does FAT or NTFS perform on creating
> files with 0x00-bytes and 0xFF bytes? :-)
> 
> This does not mean that reiserfs should not be optimized for sparse
> files, but as long as it works (albeit suboptimal), it works.
> 
> Wether bug or not, it should be solved. :-)
> 
> XuБn.
It will be solved in v4.  Extents missed code freeze.

Mason's proposed solution is an interesting one though.  Not to be done until we
leave code freeze though.

Hans