Those following the evolution of the Reiser4 filesystem will be interested in learning that it has become “fairly stable for average users”, so much so that Namesys is soon planning to push patches to 2.6 kernel maintainer Andrew Morton. Once the two remaining known bugs are fixed, the warnings against using reiser4 on a production system will likely be removed, KernelTrap reports.
Reiser4 FileSystem Stabilizing
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
Reiser4 is going to rock !!!!
Looks to be truly cutting edge stuff, Linux is quite grown up now. Wow.
“We need a lot more real user testers, because we have run out of scripts that can crash it, and there are distros that would like to ship it soon.”
I’m game. I’ll backup whatever periodically. But how do you get a crash dump or whatever off the disk if the filesystem got messed up? Will it be logged to another partition with a more mature file system on it?
Is there extended ACL support in reiser4 over the stock linux/unix user/group ownership?
Is the rumored repacker a part of this release? I had read somewhere a while back that it might have to wait until v4.1, and that it might make a big performance difference in the long term. Does Reiser4 suffer from fragmentation more than other filesystems? Is that the problem that the repacker will try to address?
This is Linux’s first FS comparable to UFS2+su (IMHO). It’s great to see this kind of intellectual activity. For a while, I though file-system innovation was dead.
If this becomes proven stability wise, Linux users may not need the current plethora of file systems (ext2/3, reiserfs, JFS, XFS. etc). Reiser4 may suit everyone’s needs.
ReiserFS is really in no way comparable to UFS2, except maybe that they’re filesystems on a Unix-like OS. UFS2 is yet another update to the standard Berkeley FFS code that’s been used for ages now. That said, FFS/UFS is rock-solid stable, albeit a sometimes lousy performer due to its design. I use mainly BSD-derived OS’s these days (OSX, NetBSD, FreeBSD), so I’ve gotten quite familiar with the various versions (OSX…UFS1 with log filesystem layout, no soft updates, NetBSD either, although ufs2 isn’t well supported, FreeBSD…UFS2 only in 5.x, which aren’t stable yet)….A faq on the differences between UFS1 and UFS2
http://sixshooter.v6.thrupoint.net/jeroen/faq.html
As for Linux filesystems, Reiser4 has some neat features that JFS and XFS lack (or aren’t fully implemented on Linux). Performance-wise, the benchmarks I’ve seen don’t show a clear performance winner among the three. Reiserfs does well with small files, JFS tends to be the best all-around performer, and XFS does well with big files. What’s going to hurt Reiser4, however, is that it’s quite cpu-intensive. On a slightly underpowered desktop, or a loaded fileserver, you can’t have that overhead. I tend to go with XFS on linux boxes, because I’ve found it to be quite reliable, fast, and the userland tools are great. I’ve never had a problem with JFS, either — it’s just harder to find distro’s that support it well. As for Reiser, twice I’ve had total partition loss with Reiser3 filesystems (both drives are still running, so no, not a hardware problem….I also have had an ext2 corruption before, but it was on a filesystem I was using with GNU/Hurd…). I will give 4 a shot when it’s integrated into the kernel…..but I admit I’m a bit standoffish, and certainly won’t put it on a production machine.
This is Linux’s first FS comparable to UFS2+su (IMHO)
What makes you say that? Linux’s ext3 is very scalable, stable, it has hashed directories, can do ordered and data journalling. And the underlying data structures are apparently pretty similar.
Softupdates are very complex. It is basically acknowledged that nobody except the code’s creator understands the code. Matt Dillon, for example has said he’d rather have a journalling fs than softupdates.
Also keep in mind FreeBSD hasn’t been at the top of the performance pack for a while. At least not in public benchmarks.
What’s going to hurt Reiser4, however, is that it’s quite cpu-intensive.
Although the the code tested was alpha quality, so I’m sure CPU usage will come down.
AFAIK Linux (2.4.x) supports POSIX ACLs. You just have to compile the kernel module. I think Linux’s implementation is FS independent.
It has been widely known for quite some time that that ext3 has numerous FS/size and SMP related scalability issues. See here:
http://www.uwsg.iu.edu/hypermail/linux/kernel/0210.3/2267.html
“ext3 has some SMP scalability problems. The BKL is used to protect many journal operations, and we see huge amounts of CPU spent spinning on it on 4/8/16 proc machines.”
http://bulma.net/body.phtml?nIdNoticia=1154&nIdPage=3
“…that have inherent scalability problem (typical complexity for search is O(n)) and are not adequate for new, vary large capacity disks.”
http://olstrans.sourceforge.net/release/OLS2000-ext3/OLS2000-ext3.h…
“…it does journal data and therefore it has poor write performance for write-intensive operations.”
There’s no ACL module in either Linux 2.4 or 2.6. The filesystem have to support ACLs. You can enable this option on a per-filesystem basis. Currently, ext3, XFS and JFS do but not Reiserfs. I don’t know for Reiser4.
Er, re-read my post and note that I did not heap praise on ext3. While I would choose it over Reiser3, I focused more on XFS and JFS — real hardcore Unix filesystems. Ext3 is not. SGI has linux running 64 proc itanic machines? I doubt, seriously, that they’re using Ext3. 🙂
I have waited for that for years!
I apologize. Afterall, you didn’t even mention ext3! I meant to direct my post at Anonymous (IP: —.dyn.iinet.net.au) but inadvertantly directed it at you (it’s been a looong day). Indeed XFS and JFS are very proven, reliable and scalable on their mother OSs – IRIX and AIX respectively. On Linux however, they aren’t yet proven as they’ve only recently been ported. This is particularly the case for JFS as a large amount of the interface to the I/O subsystem was re-written in the porting process.
ReiserFS 4 is feature complete, but I would say with a bit of working and input from users, the CPU usage will go down with time. As for the license, the only downside I see is the fact that it is GPL. I would have been nicer had it been licensed under something similar to BSD which would allow its adoption by other operating systems that use different opensource licenses, like FreeBSD or MacOS X.
Lemme set the record a little bit straight here… I have NEVER had a bad experience with ReiserFS and I’ve used it with both SuSE 8.3 and Slack 9.1…. The performance has been a hell of a lot better then Ext3 and this is on under powered DESKTOP systems… The first system was a p-200 with 284 megs of ram an a 20 gig hd and the second one was a p-200 with 64 megs of ram and a 30 gig hd…. The sheer speed of Rieser with both large and small files as far as I can tell with practical experience is ALOT faster then both Vfat32(cause I dual boot because diablo 2 isn’t linux friendly 🙁 ) and Ext3… RieserFS says they need a desktop tester.. Well here’s one right here… 1 year and going strong… No crash/hick-up or anything else… Just smooth/speedy/and responsive performance!
Aha. Didn’t know that (because i don’t use ACLs).
I searches the document: http://www.namesys.com/v4/v4.html for ACL and found it on several places. But because i’m to sick and tired at the moment, I didn’t read it complytly. But what i read was: ACLs COULD be implementet as a FS plugin, and maybe the reiserfs4 developers will implement such a plugin.
>>On Linux however, they aren’t yet proven as they’ve only recently been ported
XFS has been available for Linux for nearly 4years alright it was only included in 2.5/2.6 in last 2years but it is fairly stable, i still prefer IRIX though 😉
As regards ext3 well from what i remember of the 2.5 series (i read linus bitkeeper log everyday) alot if not most of the BKL incidences were removed in ext3, so under the hood it’s quite a different beast to what you find in 2.4
I’ve no experience with Reiser but i will say version 4 looks interesting, hopefully it will get stablised quickly enough.
Hi all…
I’m not sure whether Reiser4 has an ACL plugin yet (I suspect it does, due to where it is funded from!). Even if it doesn’t, ACLs can be constructed using file meta-data, which is what Reiser4 is designed for. Permissions are stored in meta-data for the filesystem object as a sub-object. So, to find out what user permissions are for /home/test/mydoc.txt , you do:
/home/test/mydoc.txt/user
Obviously this requires a plug-in to interpret this sort of thing to the Kernel and classic Unix permission systems in userspace, but that is very, very easy to write. Similarly, ACLs can be produced in exactly the same way, per object. It would take roughly the same amount of time to write plug-ins for both (and both are quick), so I imagine it was done a long time ago
EXT3 SMP scalabilty problems where solved in 2.6. It’s slow because it does “ordered writes”, like ntfs. reiser3, xfs and jfs only care about metadata, not about data, and they won’t care if the data associated with a metadata is written to disk before the metadata or not. Ext3 does – unless you mount it with data=writeback-. Even more, ext3, by default, does a sync every 5 seconds. That’s because the ext3 developers are really paranoid about people’s data. But this default _hurts_ benchmarks (you can tune it with commit=nrofseconds, but lots of people benchmarks ext3 without doing this). Reiser4 has a great design wich allows them to do the “ordered writes” thing that other FSs can’t do without performance loss.
UFS2 is just a hack in top of UFS, very much like ext3 WRT ext2. reiser, xfs, jfs are modern filesystems, where ext3/ufs are old unix designs ej: they don’t use b-trees, like reiser, xfs, jfs, ntfs…ufs2 and ext3 (htree)have hacks to workaround this, but thats what they are: hacks. Someone is trying to make XFS work under freebsd, IIRC.
That’s because the ext3 developers are really paranoid about people’s data.
Funny that ext3 is the only Linux filesystem I’ve ever tried that has actually corrupted my data. I’m running Reiser3 on Fedora Core 1 right here and it has been nothing but rock solid. Performance is much better than ext3 and stability has been great.
Anyway, to get back on topic…Reiser4 looks awesome, can’t wait for it to become stable enough for production use. It really is the most promising filesystem available for Linux.
Does anybody know whether there’s a tool for converting ReiserFS 3 partitions to ReiserFS 4? Or do I have to reformat everything?
Hi
“ReiserFS 4 is feature complete, but I would say with a bit of working and input from users, the CPU usage will go down with time. As for the license, the only downside I see is the fact that it is GPL. I would have been nicer had it been licensed under something similar to BSD which would allow its adoption by other operating systems that use different opensource licenses, like FreeBSD or MacOS X.”
bring out better reasons on the table. if its bsd how will namesys make money out of it. there is no reason why bsd os’es cant include gpl’ed stuff in it. freebsd has an optional gpl components for example.
regards
Jess
Hi
“http://packages.debian.org/unstable/admin/convertfs“
converts from one fs to another. Reiserfs4 is a pretty complete rewrite and is very different from reiserfs3 so i guess a clean partition is better
regards
Jess
How can it be that the GPL is recently being attacked as “restrictive”, and a “downside” to use in Free Software projects?
I mean, its free to the maximum extent possible, and its only purpose is to ensure that nobody restricts oters the rights he himself got granted by the same GPL.
How by gods sake can it happen that this noble goad is suddendly being more and more opposed as free software gains momentum and begins to interest the pro-proprietary folks? How can it be that the first thought one can have when he discovers some innovative Free Software is to moan how he cant unlock it away from the wide public, which this code is wrote for?
The first damt thought is about how you cant unlock it and make a quick buck out of it!
Its just unbelievable. I respect the reiser team and other innovative developers who still release code under the classic GPL (the for not selling out THAT easilly!
Peace.
It’s not actually GPL (ignore the claims, read the license).
For instance
http://www.namesys.com/legalese.htm
This in particular is an additional restriction that
the GPL does not cover:
Finally, nothing in this license shall be interpreted to allow you to
fail to fairly credit me, or to remove my credits such as by creating
a front end that hides my credits from the user or renaming mkreiser4
to mkyourcompanyfs or even just make_filesystem, without my
permission, unless you are an end user not redistributing to others.
If you have doubts about how to properly do that, or about what is
fair, ask. (Last I spoke with him Richard was contemplating how best
to address the fair crediting issue in the next GPL version.)
Well you can get a commercial licence for Reiser as well if you want to incorporate it into closed source systems, this provides a form of income for them. If it was under BSD like licesnce they wouldn’t have this option, so in that way the GPL is better for them.
thats odd, I am been able to compress files with windows 2000’s NTFS since 1998. Seems you get what you pay for….
Hi
its gpl. read the actual link
http://www.namesys.com/legalese.html
http://lxr.linux.no/source/fs/reiserfs/README
”
Source code files that contain the phrase “licensing governed by 7 reiserfs/README” are “governed files” throughout this file. Governed 8 files are licensed under the GPL. ”
regards
Jess
Hi
“thats odd, I am been able to compress files with windows 2000’s NTFS since 1998. Seems you get what you pay for….”
ha ha. have you bothered to compare the compression ratio,speed etc. for you info ext2 has compression support for a very long time. user space utilities exist
regards
Jess
“Jess” again. No surprise there. I was addressing the
possible reader who would actually read the license.
Hi
You need to point to correct links and quote stuff thats actually in that link then. Try doing that
Jess
used google to find ext2 / ext3 transparant file compression. i would love it for my slackware + postfix + dovecot mailserver 😉
and i found:
http://www.tldp.org/HOWTO/Filesystems-HOWTO-6.html
uh-oh:
Software consists of a patch to the linux kernel (…)
e2fsprogs i.e. e2fsck and friends). Although some people have been relying on it for years, THIS SOFTWARE IS STILL IN DEVELOPMENT, AND IS NOT ,END-USER`-READY.
ok, the link to http://opensource.captech.com/e2compr/ isn’t working…
too bad! no easy and reliable transparant filesystem compression, like for my windows xp laptop (i like filesystem compression for laptops because their harddrives are so slow, compression speeds up r/w times and cpu power is no problem)
so, i say again:
– choose windows XP for your laptop / desktop
– choose linux for your specialized server
Hi
“so, i say again:
– choose windows XP for your laptop / desktop ”
do you have any idea how many desktop users use compression?. encyrption and compression is actually more used for servers than for typical desktops
regards
Jess
Depending on you needs, you might want to check out cloop. Generally, though, you’re right, Linux doesn’t support filesystem compression to any degree for desktops. The majority of the tools to support that are targetted for embedded use.
That said, your conclusion of “Linux doesn’t support filesystem compression, use WinXP” is inane. Who still uses filesystem compression? I don’t know about other people’s computers, but on mine, anything worth compressing is already compressed. The majority of the really space-hogging things are compressed formats like video and audio files. At least on my machine, the second biggest space hog are binaries, and I’ve only got 700MB of those (pretty full configuration, too). So what’s the point?
Oh, for encryption, there is crypto-loop, which can be used to create encrypted filesystems or directories:
http://linux.co.uk/Members/oddjob/howtos/Cryptoloop-HOWTO.html
Does anyone know if Reiser4 will have any Windowz support? It would be nice if you could access your Linux stuff from windowz.
Rayiner: thanks for your comments, and the Cryptoloop link. It seems Cryptoloop can be very useful
I agree about binaries only taking a very small size. But it’s a very RELEVANT size. I do not use it to save space, I use it to SPEED UP my laptop. Laptop drives are slow, while cpu / ram power is no bottleneck.
How often do you USE that 700 MB of binaries? Very often, they are OS and program files. Smalles sizes makes faster read times. Especially during boot-up and when starting big programs very useful to have them compressed.
That being said, I really believe many more arguments exists to recommend windows xp for desktop / laptop use, like more simple GUI design and wider support / compatability. But that may change soon, if Gnome Storage succeeds. Or M$ might win the battle using it’s very promising Longhorn.
offtopic: just installed maildrop + bogofilter on my mailserver. trained it, now about 99% of spam messages are detected. Great!
I don’t disagree with you there are many reasons to choose Windows XP over Linux on the desktop. There are certain groups of users for whome I would recommend a Linux desktop over an XP one, and many groups for whome I would not. I just think that availability of compressed filesystems is really a sticking point.
This in particular is an additional restriction that
the GPL does not cover:
Finally, nothing in this license shall be interpreted to allow you to
fail to fairly credit me, or to remove my credits such as by creating
a front end that hides my credits from the user or renaming mkreiser4
to mkyourcompanyfs or even just make_filesystem, without my
permission, unless you are an end user not redistributing to others.
If you have doubts about how to properly do that, or about what is
fair, ask. (Last I spoke with him Richard was contemplating how best
to address the fair crediting issue in the next GPL version.)
If I understand you/that correctly (the link you provided is 404), it sounds like a perfectly normal moral right and is covered by copyright and so the licence doesn’t need to mention it…
Umm…doesn’t that make the license GPL-incompatible? Isn’t that just an advertising clause like the X-Free one that everyone is complaining about?
On a side note, I’m not at all surprised that Hans Reiser put that there. He is an absolute egomaniac, much like that X screesaver guy who insists on having that ugly X logo on the unlock screen “to credit him”.
How about a graphical boot sreen? That hides his “credits” from the user. Is that not allowed? That clause looks to be a seriously bad thing.
HI
I dont find any such restrictions on the license or the documentation that accompanies it. nobody has questioned it and i dont find any such claims on the legalese page
what are you talking about
Jess
“Does anyone know if Reiser4 will have any Windowz support?”
You got it the wrong way around. You should be asking if
“Windowz” is finally going to include drivers for filesystems
other that fat/ntfs.
Well… I heard some rumors about some guy that had WinXP up and running on a Reiser4 partition. But those are just rumors. I haven’t seen any actually *proof* of it.
Best filesystem ever! OK, enough with the Simpsons reference. It looks really great. Now all i have to hope is that the Fedora people will include it in the final version of FC2. Hey, maybe we’ll see it in test 2 tomorrow. Is that too much to hope for?
There is a lot of exciting stuff happening in the Linux world. ReiserFS is looking great, mono looks like it is going to be really nice, Gnome has always been great for me, and the SuSE/Novell/Ximian thing looks like it could do nice things for the GNU world (frankly, I find it weird that Novell bought the biggest KDE pusher as well as the home of Gnome – just kinda ironic).
(Note: I am not a lawyer. The following is my understanding following the introduction and discussion of the Australian versions of the Creative Commons licences and my own research made in the past week (since its introduction).)
Umm…doesn’t that make the license GPL-incompatible? Isn’t that just an advertising clause like the X-Free one that everyone is complaining about?
No. The problem with the old BSD and new XFree86 licences is that they require you to attribute you in extra places, for instance the documentation.
Consider the following. Under copyright laws which protect moral rights (those of ‘droit d’auteur states’—right-of-the-author states, which includes all Berne Convention states apart from the US, though by being a signatary to it, I think they have to protect it somehow) if you took the following:
A short poem, by Felix Frogman
There was was a man.
A happy man.
He heard his mother died.
He was no longer happy.
and changed it to:
A short poem, by A.K.H.
There was was a man.
A happy man.
He heard his mother died.
He was no longer happy.
you would be breaking copyright law, even if the reproduction (with the name unchanged) was legal. (Note: My name is not Felix Frogman, so the work is published pseudonymously. However, even if it was published anonymously, you can’t say you wrote it.)
(Indeed, I suspect that even
A short poem, by A.K.H.
There was was a man.
Oh! a happy man.
He heard his mother died.
Wo! he was no longer happy.
would be wrong, or at least suspect. Note that not all copyright laws throughout the world forbid the creation of derivative works as strongly as the United States’.)
This, clearly, is a totally different idea from the old BSD/new XFree86 licence, which states that in reproducing the poem, you have to say somewhere else that you include a poem by Felix Frogman. If, for example, you published an anthology of poems by Australian authors, and included ‘A short poem’, and if I’d declared that reproduction of my poems can only be done under the terms of the old BSD licence, then in any advertising material for the anthology, you’d have to say ‘This anthology contains a work by Felix Frogman’.
Personally, I would be offended if someone took this post verbatim or with minor and/or inconsequential changes and said it was by them instead of by me. It certainly stands without of the spirit of the General Public Licence (the intention of which is to allow consequential changes, and indeed by requiring that these modifications are marked as such, it seems to imply fair and correct attribution ).
(Note: I haven’t read the text of the GPL recently so I’m going by memory. As I said at the start of this, I’m not a lawyer and have only had a week to research this anyway.)
Under copyright laws which protect moral rights… if you took… A short poem, by Felix Frogman… and changed it to: A short poem, by A.K.H…. you would be breaking copyright law, even if the reproduction (with the name unchanged) was legal.
US common law does not recognize plagiarism as a fair use, so for unauthorized distribution the case would be much the same. For authorized distribution, an American author may decide whether to require attribution. You are correct in stating that this is already covered by copyright law. However, the “credits” to which Reiser refers are the paragraphs (over a screenful on some terminals) that reiserfsprogs spew to stdout and, from what you quoted, the name of the mkreiser4 binary. This is well beyond the standard of attribution set by law and the GPL and imposes additional restrictions on distribution.
@By capitalist
> thats odd, I am been able to compress files
> with windows 2000’s NTFS since 1998.
The Amiga had compressed filessystem since 1990.
> Seems you get what you pay for….
You say it.
Cheers
Gunnar
with a lot more distros if they actually made it possible to find their miserable pathetic stinking tools to mkfs a whatever type lousy partition. It’s like pulling teeth to put together a set of tools for any of them. I’m sure most distros just, “screw it I’m sick of hunting, we’ve already got ext2/3 and xfs”, and just flat give up, it’s a real pain in the ass I’m sure few people realize, they aint all in busybox.
I do not understand this post. I’ve never had a problem installing reiserfstools on any distro I’ve encountered. Can someone clarify what he’s trying to say? Thanks.
http://groups.google.com/groups?q=+reiser+group:linux.debian.legal&…
Since Reiser’s license is not GPL, how is it that
it is bundled with Linux? (I don’t use it of course).