XFS is a GPL'd high-performance journaling file system developed by SGI since 1994. Version 1.0 was released for Linux on May 1, 2001, with the current release being version 1.1. Christoph Hellwig has provided patches aimed towards getting XFS included into the 2.5 development kernel [earlier story].
A very length thread on the lkml started when 'khromy' asked, "What's up with XFS in linux-2.5? I've seen some patches sent to the list but I havn't seen any replies from linus.. What needs to be done to finally merge it?". Buried amongst typical tangents flaming Linus Torvalds for dropping patches, it was pointed out that there's little left to be merged for XFS support. Alan Cox [earlier interview] explains, "The problem has always been that XFS was very invasive code so it might break stuff for people who dont choose to use experimental xfs stuff. Thats slowly improving". Eric Sandeen, part of the SGI XFS development team, points out that the only "invasive" items left to merge are: 1 new process flag, 1 new CTL_VM name, 1 new CTL_FS name and 1 exported symbol. That said, support for XFS in the mainline 2.5 kernel looks to be quite likely.
From: khromy To: linux-kernel mailing list Subject: XFS? Date: Mon, 9 Sep 2002 15:38:20 -0400 What's up with XFS in linux-2.5? I've seen some patches sent to the list but I havn't seen any replies from linus.. What needs to be done to finally merge it? -- L1: khromy
From: Thunder from the hill Subject: Re: XFS? Date: Mon, 9 Sep 2002 15:00:18 -0600 (MDT) Hi, It has been stated quite regularly that XFS a) doesn't always work like it should yet b) involves some changes which Linus doesn't like in particular, for pretty good reasons. Go read the archives on that if you want more. Thunder
From: Joe Kellner Subject: Re: XFS? Date: Mon, 9 Sep 2002 18:29:24 -0400 I use XFS on a number of production servers, and it has always served me well. The only problem is of course getting a kernel that supports XFS. I'm glad many of the more mainstream linux distributions are starting to have support for XFS support "out of the box", but I feel until it's in the mainline kernel many people will never even have a chance to try it.
From: Wade Subject: Re: XFS? Date: Tue, 10 Sep 2002 06:08:06 +1000 I've noticed this too. And I must say, ITS QUITE FUCKING RUDE OF LINUS. [Editorial comment: Who's being rude?]
From: Arador Subject: Re: XFS? Date: Mon, 9 Sep 2002 22:43:43 +0200 Well, you have the patches, so you can't say that you can't use it. If you don't like you can make a branch from the 2.5 kernel including xfs. But personally i don't mind if xfs is not included, I assume that there's some reason for not merging it, and i'd be glad to hear the reasons for not merging it, insteand of blaming to Linus "The evil non-merger" Torvalds
From: Shawn Subject: Re: XFS? Date: Mon, 9 Sep 2002 16:18:46 -0500 Linus has proudly declared himself to be a non-patch accepting bastard many times in the past. This has the interesting side effect of spurring debate and flame wars about the joys and horrors of ac-rmap-xfs-O1-devfs-lvm-xfs-lowlat-preempt-blah blah patchsets. This is a Good Thing(tm). You should thank Linus for being Linus, and also thank those who actively maintain alternate branches of the kernel. Shawn Leas
From: Rik van Riel Subject: Re: XFS? Date: Mon, 9 Sep 2002 19:31:44 -0300 (BRT) If you think you can do things better you should fork Wadix ;) Linus might not be the easiest user interface and I've had some problems too in the past, but everybody will have to admit that the current system just works in the long run. cheers, Rik
From: Wade Subject: Re: XFS? Date: Tue, 10 Sep 2002 16:23:48 +1000 In the long run, for sure. I'm not saying Linus needs to be replaced, but his _manners_ could do with some work. Those XFS guys have worked quite hard to make the merge fairly painless, and Linus wont even comment (that I've seen, see http://marc.theaimsgroup.com/?l=linux-kernel&m=103118187002156&w=2). BTW: I'm only a lurker, I don't even use XFS :-)
From: Gerhard Mack Subject: Re: XFS? Date: Tue, 10 Sep 2002 09:24:06 -0400 (EDT) Keep in mind Linus gets flooded by patches on a constant basis and also needs to focus on getting core functions working again so I wouldn't be supprised if he hasn't had time to deal with XFS yet. Hes also known for just deleting his mail que when it gets overloaded so he may not have even read the messages in question. Gerhard
From: Andi Kleen Subject: Re: XFS? Date: 09 Sep 2002 23:12:10 +0200 Thunder from the hill writes: > It has been stated quite regularly that XFS > a) doesn't always work like it should yet That's quite bogus. While not being perfect XFS just works fine for lots of people in production and performs very well for a lot of tasks. > b) involves some changes which Linus doesn't like in particular, for > pretty good reasons. I think that's FUD too. That last patch had 6 lines or so of changes to generic code, everything else was already merged. I guess it just ended up in Linus' spam filters, like some other things... -Andi
From: Shawn Subject: Re: XFS? Date: Mon, 9 Sep 2002 16:20:50 -0500 XFS needs a sponser. Who amung Linus's circle of trust cares to comment or re-evaluate? If no one, I guess it's a moot point. -- Shawn Leas
From: Robert Love Subject: Re: XFS? Date: 09 Sep 2002 17:27:40 -0400 On Mon, 2002-09-09 at 17:20, Shawn wrote: Christoph Hellwig (hch) is working on the patches... I cannot speak for Linus, but I think most of us trust him. I do. Robert Love
From: Mike Galbraith Subject: Re: XFS? Date: Tue, 10 Sep 2002 19:15:08 +0200 (see below) [...snip...] > I think that's FUD too. That last patch had 6 lines or so of changes > to generic code, everything else was already merged.
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
-Mike
From: Shawn
Subject: Re: XFS?
Date: Tue, 10 Sep 2002 14:23:47 -0500
I'm not sure what this is intended to communicate.
The question was specifically regarding filesystem support, so I'll
assume you meant to point out that XFS does not always work like it
should.
Then, am I incorrect that since almost all of XFS that's left to merge
is XFS code and not changes to the kernel at large?
If this is correct, could I then make the assumption that merging XFS
would be minimally impactful for those kernel user who do not enable it?
Linus incorporated reiserfs long before it "always functioned as it is
supposed to", so I find myself wondering where your point was.
(see below) and "^^^^^^^^^^^^^" don't fully cover your thoughts I'm
afraid.
From: Robert Love
Subject: Re: XFS?
Date: 10 Sep 2002 15:29:17 -0400
Yes, and I think it will go in. It has just not made it passed Linus's
filters yet, but it will - hch knows what he is doing.
Robert Love
From: Thunder from the hill
Subject: Re: XFS?
Date: Tue, 10 Sep 2002 13:33:41 -0600 (MDT)
Hi,
It was my point, actually. I was referring to some crashes caused by XFS,
however, they seem resolved. I also want to be commonly known as not
hating XFS. It seems pretty cool.
Thunder
From: John Alvord
Subject: Re: XFS?
Date: Tue, 10 Sep 2002 13:06:27 -0700
If memory serves, Linus incorporated reiserfs after several major
distributors started including it. Linus seems to pay a lot of
attention to distributions in areas where he isn't so much interested.
So does Redhat/Suse/??? ship XFS yet?
johm
From: Hans Reiser
Subject: Re: XFS?
Date: Wed, 11 Sep 2002 00:17:14 +0400
Mandrake does if I remember right.
XFS is cool, and their allocation at flush innovation has influenced
reiser4 deeply. I wish them well.
Hans
From: Joe Kellner
Subject: Re: XFS?
Date: Tue, 10 Sep 2002 16:17:52 -0400
Mandrake has had XFS support in the default boot kernel since 8.0. AFAIK, Suse
and Slackware also have XFS capable kernels now too.
From: Tomas Szepe
Subject: Re: XFS?
Date: Tue, 10 Sep 2002 22:31:22 +0200
Slackware lets you use a special install kernel that has XFS [1] compiled in
and that the installer can take advantage of. However, you have to provide
your own XFS vmlinuz for actually booting into the new system, as Slackware
rightly continues to use bare Marcelo(tm) kernels.
[1] and JFS, too.
T.
From: Bernd Eckenfels
Subject: Re: XFS?
Date: Wed, 11 Sep 2002 04:34:33 +0200
so does debian do support XFS:
http://www.physik.tu-cottbus.de/~george/woody_xfs/
Greetings
Bernd
From: David Lang
Subject: Re: XFS?
Date: Tue, 10 Sep 2002 13:26:05 -0700 (PDT)
and you know something is mainstream when slackware includes it.
said as a longtime slackware user :-)
I know slackware 8.1 included XFS, I don't think it was in 8.0.
David Lang
From: Nick LeRoy
Subject: Re: XFS?
Date: Tue, 10 Sep 2002 15:18:31 -0700
Don't know about RedHat & others, but SuSE _does_ ship XFS.
-Nick
From: Steve Lord
Subject: Re: XFS?
Date: 10 Sep 2002 16:01:25 -0500
I should probably keep out of the discussion and I am not presenting
this as an argument for inclusion, but for an incomplete list of XFS
users and distribution including it look here:
http://oss.sgi.com/projects/xfs/xfs_users.html
Steve
From: Mike Galbraith
Subject: Re: XFS?
Date: Wed, 11 Sep 2002 06:56:22 +0200
(sigh)
If "everything else" the XFS team has asked for has gone in, it seems
unlikely that
sponsorship is needed.
-Mike
From: Shawn
Subject: Re: XFS?
Date: Wed, 11 Sep 2002 09:55:29 -0500
Which is why I pointed out that the issue at hand was not regarding the
everything else, but in fact the actual filesystem support.
As far as why the rest is still pending, I was just offering ideas.
A lot of this thread is advocacy as opposed to substantive conversation
about the how and/or why/why not of inclusion of XFS into mainline.
Fankly, there is no /real/ answer except "Linus has not weighed in on
the current question".
I lost my ability to invest emotions in either side of huge kernel
debates when the devfs and lvm wars happened.
From: Mike Galbraith
Subject: Re: XFS?
Date: Wed, 11 Sep 2002 19:52:51 +0200
At 09:55 AM 9/11/2002 -0500, Shawn wrote:
>Which is why I pointed out that the issue at hand was not regarding the
>everything else, but in fact the actual filesystem support.
I was just trying to say that everything _appears_ to be on track from my
(remote) perspective. I've noticed no gripes from the XFS team, only
evidence that development continues. If there are 6 lines of generic code
changes left, that means a lot has happened.
>As far as why the rest is still pending, I was just offering ideas.
>
>A lot of this thread is advocacy as opposed to substantive conversation
>about the how and/or why/why not of inclusion of XFS into mainline.
Advocacy without technical meat sucks.
>Fankly, there is no /real/ answer except "Linus has not weighed in on
>the current question".
Hey, maybe he's trying to convince (blackmail;) them to port some bandwidth
guarantee stuff ;))))) (I hope that's enough smilies)
>I lost my ability to invest emotions in either side of huge kernel
>debates when the devfs and lvm wars happened.
I love it when the heavyweights square off (oooo:). Unfortunately, that often
leads to a bunch of dipsticks hollering "food fight!" ;-)
-Mike
From: Bill Davidsen
Subject: Re: XFS?
Date: Wed, 11 Sep 2002 11:12:54 -0400 (EDT)
On 9 Sep 2002, Andi Kleen wrote:
> Thunder from the hill writes:
>
> > It has been stated quite regularly that XFS
> > a) doesn't always work like it should yet
>
> That's quite bogus. While not being perfect XFS just works fine for lots
> of people in production and performs very well for a lot of tasks.
More to the point, a quick scan of LKML will show that there are fixes for
ext3 and reisser on a regular basis, so one must assume that they don't
always work as they should either. XFS is in a number of distributions,
and is stable for users.
> > b) involves some changes which Linus doesn't like in particular, for
> > pretty good reasons.
>
> I think that's FUD too. That last patch had 6 lines or so of changes
> to generic code, everything else was already merged.
Does that mean he should only dislike it a little because it's small? At
this stage I would hope he will at least tell you why it wasn't accepted,
since XFS is a desirable feature for many people (as evidence vendors
providing it).
I'd like XFS, I think it's a good feature politically, hopefully it will
not just drop just as it's becoming stable for non-critical production
use.
--
bill davidsen
From: Alan Cox
Subject: Re: XFS?
Date: 11 Sep 2002 17:03:49 +0100
Thats never been the big concern. The problem has always been that XFS
was very invasive code so it might break stuff for people who dont
choose to use experimental xfs stuff. Thats slowly improving
From: Eric Sandeen
Subject: Re: XFS?
Date: 11 Sep 2002 13:55:07 -0500
Alan -
The last patch Christoph posted against 2.5 is not the least bit
invasive. Excluding documentation and configuration files, these are
the changes:
o 1 new process flag: +#define PF_FSTRANS 0x00100000
o 1 new CTL_VM name: + VM_PAGEBUF=18
o 1 new CTL_FS name: + FS_XFS=17
o 1 exported symbol: +EXPORT_SYMBOL(mark_page_accessed);
and of course an addition to fs/Makefile:
+obj-$(CONFIG_XFS_FS) += xfs/
That's it. The rest is under fs/xfs.
(2.4 is more invasive, but this thread started out talking about XFS in
2.5).
-Eric
--
Eric Sandeen XFS for Linux http://oss.sgi.com/projects/xfs
From: Alan Cox
Subject: Re: XFS?
Date: 11 Sep 2002 22:37:39 +0100
As I said its improving
Related Links:
Alan
>From: Alan Cox
>Subject: Re: XFS?
>Date: 11 Sep 2002 22:37:39 +0100
>
>As I said its improving
Exactly what more is there to improve upon? This is what they need to know, and [so far] Linus has refused to answer.
Linus is sometimes the wrong person to ask
Many of these things could be merged at the distro level.
Lots of kernels included riserfs long before Linus did. In fact, the extra testing from distro users influenced Linus's decision to merge it.
KDB is another thing that makes sense in theory to include at the distro level. As far I know, no distros include it. Causing me to wonder whether the theory is correct.
KDB in distro
No, KDB does not belong in the distros. The distros include
features that their target audience would want/need. This is
why ext3/reiserfs got pushed in fairly early - large disks took
too long to fsck if they crashed, and so a lot of enterprise users
complained and so it got into the distros and tested hard by Red
Hat, SuSe and others. Same thing with supermount - lots of
users wanted it, so Mandrake and others include it, but I don't
think it's in the main kernel.
KDB is not even remotely useful for these types of users. It's
mainly a kernel debugging type of thing which is not something
the typical distro user is going to want or care about.
Pete Flugstad
kdb is useful in "enterprise" situations
at least in theory.
Say you are Red Hat and have a customer in Alaska with a problem. You can just ask them over the phone to reproduce it, press a key-combo and read off all the debug info you need to email them a patch.
You are right that most customers don't need this, and that even the ones who do probably don't know they need it.