LWN: Comments on "Linus on GPLv3 and ZFS"

Netapp and patents

anton — Fri, 22 Jun 2007 07:10:19 +0000

I suppose NetApp has patents too, but perhaps Linus wishes to imply that they would be more tractable to deal with than Sun

Yes, Netapp has patents, and they caused Daniel Phillips to stop working on the tux2 filesystem; I have not followed the story enough to know if Netapp did anything other than file the patents to achieve this result.

Netapp's WAFL is not very interesting for Linux anyway, because it requires special NVRAM hardware to buffer writes during some of the more time-consuming operations (e.g., snapshot creation). I don't think that this hardware dependence can be eliminated without major changes to the WAFL code.

Concerning not breaking Sun patents, you can look for older sources where similar ideas have been described, e.g., various papers on log-structured file systems, e.g., our Freenix 2000 paper, or (maybe too young) my file system ideas.

COW for Flash?

joern — Tue, 19 Jun 2007 08:17:11 +0000

> It has been said that COW is ideal for Flash. Can you explain why ZFS isn't relevant here?

Raw flash behaves sufficiently different to hard disks that some ZFS design assumptions become untrue. Flash has large erase blocks. Within erase blocks, data must be written from front to back. Writing the block again requires erasing all of it. So the filesystem block size either has to be equal to the erase block size, or you need garbage collection. And with garbage collection comes a nasty deadlock problem most people don't even realize exists. :)

Next comes wear out and protection against it. Afaics, ZFS has several hot zones that receive significantly more writes than others.

I guess those two are the big reasons.

COW for Flash?

qu1j0t3 — Mon, 18 Jun 2007 17:26:30 +0000

It has been said that COW is ideal for Flash. Can you explain why ZFS isn't relevant here?

There is no fsck; ZFS is "always consistent on disk" (through COW+atomic transactions). It seems to me this is a necessary invariant to achieve its other features (such as snapshots). Debate flares up (occasionally) as to whether a scavenger will be necessary. If so, it won't much resemble 'fsck' - and certainly won't be run in normal operation or after reset/powerfail/etc (ZFS behaviour under impromptu reset is extremely well tested).

I suspect, but correct me if I'm wrong, once you "know" you've lost data in ZFS (through exhausting redundancy or ditto blocks), it's actually gone by definition, and unrecoverable by re-creating links. No doubt Bonwick et all have explained it better somewhere...

good to hear

joern — Sun, 17 Jun 2007 21:43:18 +0000

> From comments I've heard, and the buzz around Linus' and Jon's posts, there seems to be considerable community interest around ZFS (I don't want to use anything else, or wait, so I switched to Solaris 10 some time ago).

> Look forward to more news on this front. Did you follow-up on LKML? (I have not checked :)

My personal interest is in flash, not hard disks. Therefore ZFS is impressive technology, but solving someone else's problem. It is not the last word in filesystems either, as the fsck will run for hours or days if it ever becomes necessary. So there remain valid reasons to work on different filesystems.

Impressive technology none the less.

good to hear

qu1j0t3 — Sun, 17 Jun 2007 19:15:50 +0000

The only question is when, not if. :)

From comments I've heard, and the buzz around Linus' and Jon's posts, there seems to be considerable community interest around ZFS (I don't want to use anything else, or wait, so I switched to Solaris 10 some time ago).

Look forward to more news on this front. Did you follow-up on LKML? (I have not checked :)

btrfs?

joern — Sun, 17 Jun 2007 19:05:00 +0000

> Maybe so, but there's quite a lot of catch-up to do. Once you have COW, transactions, and checksums, then you want self-healing; then snapshots; pools; quotas; compression; and so on, until you eventually have something like ZFS. :)

Sure, ZFS has an impressive set of features. If nothing else, it has showed how things can be done. And I have little doubt that btrfs, which you quoted, will end up having most of those features relatively soon. And even if Chris dies tomorrow, I'll keep working on LogFS.

The only question is when, not if. :)

btrfs?

qu1j0t3 — Sun, 17 Jun 2007 18:06:14 +0000

Maybe so, but there's quite a lot of catch-up to do. Once you have COW, transactions, and checksums, then you want self-healing; then snapshots; pools; quotas; compression; and so on, until you eventually have something like ZFS. :)

Linus' grandstanding aside, it's possible there is quiet work going on to improve the situation, as David Magda commented on zfs-discuss:

Somewhat off topic, but it seems that someone released a COW file system for Linux (currently in 'alpha'):

	* Extent based file storage (2^64 max file size)
	* Space efficient packing of small files
	* Space efficient indexed directories
	* Dynamic inode allocation
	* Writable snapshots
	* Subvolumes (separate internal filesystem roots)
	- Object level mirroring and striping
	* Checksums on data and metadata (multiple algorithms available)
	- Strong integration with device mapper for multiple device support
	- Online filesystem check
	* Very fast offline filesystem check
	- Efficient incremental backup and FS mirroring

	http://lkml.org/lkml/2007/6/12/242
	http://oss.oracle.com/~mason/btrfs/

Via Storage Mojo

ZFS and WAFL

joern — Sun, 17 Jun 2007 17:29:35 +0000

Checksums are easy to add once you have a COW format. Either you add them to the block pointers, as ZFS did, or you add them to the objects themselves, as JFFS2 and LogFS did.

Either way you have an incompatible format change. But the amount of code affected if rather small. Took about 1-2% of the effort to design a new filesystem in the LogFS case.

ZFS and WAFL

qu1j0t3 — Sun, 17 Jun 2007 15:30:37 +0000

As you see I haven't studied XFS in depth, but I was under the impression it was COW like ZFS. AFAIK, WAFL also lacks the more interesting features of ZFS (foremost being end-to-end checksumming).

ZFS and WAFL

joern — Fri, 15 Jun 2007 00:02:54 +0000

> XFS is already integrated, and that has about as much in common with ZFS as WAFL does.

That is plain wrong. XFS is in the huge class of traditional filesystems with a static mapping between file offsets and device offsets. ZFS is in the somewhat smaller (ignoring reseach projects) class of COW filesystems, just like WAFL. Anyone unable to see the similarities is well advised to read more and write less. ;)

what epoch are you posting from?

drag — Thu, 14 Jun 2007 07:36:34 +0000

Those things seem to have triggered some sort of bug.

Rest assured people do have real-world Linux boxes that have over 512 cpus in a single system image. SGI has boxes, at least, have Linux boxes with 4096 cpus in a single system image.

As far as clustering goes.. there are Linux systems with tens of thousands of cpus going.

Linux kernel itself does scale past 8 cpus. Of course nothing is perfect.

Linus on GPLv3 and ZFS

error27 — Thu, 14 Jun 2007 01:59:23 +0000

Instead of "ridiculous and unfortunate" I would say "justified by current events."

Linus on GPLv3 and ZFS

notamisfit — Wed, 13 Jun 2007 21:56:58 +0000

It creates the possibility that code created in a downstream work may not be usable upstream. Linus has put his cards on the table in the past; he wants code back.

okay but,

dlang — Wed, 13 Jun 2007 19:51:38 +0000

note that the zfs code released for grub is not enough to actually be able to write to the filesystem, just enough for grub to be able to find the files that it needs.

Linus on GPLv3 and ZFS

man_ls — Wed, 13 Jun 2007 13:36:54 +0000

It is not so unreasonable: Linus said:

How can you _ever_ sign anything sight unseen? That's just stupid, and that's totally regardless of any worries about the FSF.

Said that way, it looks like the correct thing to do. However, given that (as you say) "v2 or later" licensing gives the choice to the user, I'm not particularly worried about misuse.

Linus on GPLv3 and ZFS

job — Wed, 13 Jun 2007 13:09:12 +0000

Maaaybe there was a good reason why the FSF recommended the use of "v2 or later" licensing. Then you basically leave the choice to the user. I never understood what Linus didn't like about that, except some unspecified fear of the FSF, which would be not only ridicolous but also unfortunate.

competition

qu1j0t3 — Wed, 13 Jun 2007 12:47:40 +0000

There difference may be that it's augmentative competition rather than destructive - In business, it's optimal to completely eliminate rivals. In open source, you don't have to do that; you can just do better. It's a more pure meritocracy. I hope. :)

That said, there's still some dirty pool played from time to time, but since it's played in the open, it hardly festers.

Linus on GPLv3 and ZFS

paulj — Wed, 13 Jun 2007 12:36:14 +0000

Of course CDDL is hopeless is this regard as hackers must transfer copyright to Sun, who want to do that?

This is false.

You need to sign a contributors agreement with Sun, granting Sun joint ownership, if you wish to have Sun incorporate any contribution to various open-source projects which Sun founded and maintain (such as OpenSolaris, amongst other projects).

However the CDDL does not require any copyright transfership, and you're quite free to take and hack away on CDDLed code, like OpenSolaris, without giving copyright ownership to Sun or anyone else.

Linus on GPLv3 and ZFS

marduk — Wed, 13 Jun 2007 12:25:12 +0000

Reads like typical salesmanship. Sales talk always "sounds" good...

I actually disagreed with his implication that communities don't compete (only corporations). There exists competition in the OSS community.

Linus on GPLv3 and ZFS

venkatesh045 — Wed, 13 Jun 2007 10:03:21 +0000

You could read Jonathan Schwartz's reply to this post. I think he clarifies a lot of issues as to what Sun is looking at. This post is actually worth a good read.

okay but,

mjthayer — Wed, 13 Jun 2007 09:38:06 +0000

That said "GPL v2 or later" if I read correctly. I didn't take the time to read the code, but presumably that is only code for reading and would not affect potential patents on writing parts.

On another note, if Sun make Solaris GPL3 and accept external contributions, it might get tricky to keep parts (i.e. ZFS) under another licence.

Linus on GPLv3 and ZFS

man_ls — Wed, 13 Jun 2007 09:37:07 +0000

That kind of compatibility is not much help, unless all of the kernel is licensed as "v2 or later". As long as there is a single file licensed under "v2 only", it becomes impossible to link with a single "v3 only" file.

Meanwhile, relicensing all files under a "v2 or later" license might seem to be a necessary first step to a GPLv3 kernel. But given Linus' reluctance to blanket license, I would rather expect a "dual v2-v3" license, if the migration is to be done at all.

Linus on GPLv3 and ZFS

forthy — Wed, 13 Jun 2007 08:20:17 +0000

The FSF took great efforts that GPL versions can be made compatible. The paragraph that deals with it is section 9 of the GPL. Read it, especially the last part - many files in the Linux kernel are not explicitely restricted to a specific GPL versions, which means "any version". And section 6 makes sure that everybody receives a license from the original licensor, not from a compilation editor like Linus Torvalds.

The compilation editor (Linus Torvalds) can set terms under which he redistributes the work, i.e. conditions he has to follow. But since everybody receives the license from the original licensors, this "restriction" is null and void, you still can make a compilation yourself which does not restrict the license version, and then, most parts of Linux are compatible with GPLv3 (because you can either choose any GPL or explicitely v2 or later).

okay but,

TRS-80 — Wed, 13 Jun 2007 07:57:14 +0000

One could always start from Sun's GPLv2 ZFS code in GRUB. And Jonathon Schwartz has just posted saying Linux ZFS would have full patent indemnity.

what epoch are you posting from?

Nick — Wed, 13 Jun 2007 03:03:55 +0000

Poster is probably talking about this blog entry

http://jeffr-tech.livejournal.com/5705.html

So he is right, and Linux did have a problem on this
workload. Basically it was a combination of a glibc
inefficiency and the fact that nobody seems to have
reported such a workload before. The fix was basically
a small change to the way malloc/free works, and a
little patch to the kernel to optimise the new path
used by glibc.

http://www.thisishull.net/showpost.php?s=5d2bfa8b5a070728...

That post found the fixes to have eliminated the big
dropoff. Note it still doesn't scale past 8-way, but
this is likely to be a MySQL issue -- BSD doesn't do
any better.

what epoch are you posting from?

xoddam — Wed, 13 Jun 2007 02:37:58 +0000

> linux kernel having problem scaling beyond 8-way compared to bsd.

You're several years behind. A long time_t in LKML-land.

Linus on GPLv3 and ZFS

wolfrider — Wed, 13 Jun 2007 01:48:36 +0000

> JFS [is] rarely used

--Depends on who you ask. I use JFS now almost exclusively for Vmware and "bkps" (read: large) filesystems, where before I would use ReiserfsV3 with notail.

--After seeing how fast (and reliable) JFS is, I switched almost all my Reiser filesystems over to it - and have been much happier. Reiser is great for root and squid (tail-packing) but not ideal when you're trying to run a VM from a USB2 IDE drive. JFS makes it usable.

okay but,

qu1j0t3 — Wed, 13 Jun 2007 00:21:53 +0000

This still does not fully explain to me why, to date, kernel devs aren't looking dispassionately as the affordances of ZFS and how they might have them without stepping on anyone's patent*. Max V. Yudin recently asked on zfs-discuss,

... is it legal to write ZFS clone from scratch while maintaining binary compatibility with original?
Jeff mentioned in his blog that Sun filled 56 patents on ZFS related technologies. Can anybody from the company provide me with more information about this?
If porting ZFS to Linux kernel is not an option and I were to implement different file system with ZFS ideas in mind how can I be safe and not break any Sun patents?

There has been no meaningful resolution of his questions. At least it may prove that, thanks to software patents, interesting development is now impossible. So much for stimulating innovation...

* - I suppose NetApp has patents too, but perhaps Linus wishes to imply that they would be more tractable to deal with than Sun (maybe he actually knows somebody @ NetApp). Let's dream for a moment, and imagine that Linus and Jonathan, over a piña colada one Sunday, work out a magical way to free ZFS for kernel inclusion. That would be a P/R coup for Sun an order of magnitude greater than even the Apple buzz. Since Solaris 10 famously runs on all varieties of hardware (IBM, HP, Dell, even Macs), I don't seriously think Jonathan believes this would damage hardware sales. Then again, I only have the ponytail, not an MBA, and my bonuses are a few zeroes short of his. ;-)

linux core better?

JoeBuck — Wed, 13 Jun 2007 00:11:42 +0000

You are seriously out of date; folks at SGI have Linux running on 1024 processors.

The Solaris and BSD folks cannot claim to be more scalable than Linux at this point; it appears that the reverse is true.

forthright != Linus

JoeBuck — Wed, 13 Jun 2007 00:03:10 +0000

The issue with Sun is not that they prefer a particular license, but that they are choosing to license patents only to code that uses their particular license, while IBM, Red Hat, Novell, and others are licensing a number of patents (or in Red Hat's case, all their patents) to developers who use a much larger set of open source licenses.

Linus on GPLv3 and ZFS

JoeBuck — Wed, 13 Jun 2007 00:00:19 +0000

To be fair, Sun's launching a project, called "Indiana", to correct that deficiency and produce something that would resemble a GNU/Linux distribution. It will take them some time to do it, but I'm looking forward to it.

linux core better?

genius — Tue, 12 Jun 2007 23:54:32 +0000

i dont think i can agree with it. that was a blog last time about the linux kernel having problem scaling beyond 8-way compared to bsd. not sure whether they have solved it. on the other hand, linux has definitely revived interest in unix.

forthright != Linus

qu1j0t3 — Tue, 12 Jun 2007 23:43:51 +0000

It's not very "forthright" to inject snide Linusisms such as "the only really interesting thing they have is ZFS, and even there, I suspect we'd be better off talking to NetApp"! XFS is already integrated, and that has about as much in common with ZFS as WAFL does.

It comes across as sour grapes about the license, and even some N-I-H ("core Solaris: who are you kidding - Linux code is _better_"). Btw, there is as much spurious rancor of the opposite polarity from the Sun camp, as recent zfs-discuss flamefests can attest.

Why can't we all just get along? - Admit that some people like BSD license, some people like GPL, Sun likes CDDL for now, and ZFS plain rocks... :)

Linux devs ignore it at their peril; Linus, being an engineer of Sun's calibre, could do a much more helpful job of deconstructing the issue.

WAFL != ZFS

allesfresser — Tue, 12 Jun 2007 23:03:30 +0000

I don't think it's anywhere near FUD, personally. It sounded simply like classic Linus--he's being very transparently honest. He hopes and wishes that Schwartz and company are being as open and forthright as they claim to be, but knowing human nature and the temptations that beset us, he is keeping his powder dry and his head down, so to speak.

Linus on GPLv3 and ZFS

cyperpunks — Tue, 12 Jun 2007 22:28:46 +0000

Linus is right, Sun don't want Linux source code, they want Linux' kernel hackers (and
then later Linux' users).

Of course CDDL is hopeless is this regard as hackers must transfer copyright to Sun, who
want to do that?

Sun have to fix the bootstrap problem too: it's now not possible to build a complete
free Solaris "distribution". You must use some non free Sun tools at some point.

Who wants to contribute to project you can't build yourself?

WAFL != ZFS

qu1j0t3 — Tue, 12 Jun 2007 21:56:28 +0000

I have to assume Linus knows that. Sigh. If not, like another poster here, he should Google... I'm tired of posting ZFS linkfests ;-)

He's treading close to the FUD-line with this one. There's also a hidden assumption here that Jonathan Schwartz is being disingenuous with his massively revamped corporate strategy.

Sun's a hardware company. They're happy for you to run Linux on your Sun gear if you prefer - it's a supported option - heck, they even support Windows.

for the clueless

huerlisi — Tue, 12 Jun 2007 21:14:17 +0000

Because it's a nice product of computer engineering. Here's a quote from a nice-to-read geeky background article :

64 bits would have been plenty ... but then you can't talk out of your ass about boiling oceans then, can you?

Simon

Linus on GPLv3 and ZFS

man_ls — Tue, 12 Jun 2007 21:04:07 +0000

True:

When we say that GPLv2 and GPLv3 are incompatible, it means there is no legal way to combine code under GPLv2 with code under GPLv3 in a single program. This is because both GPLv2 and GPLv3 are copyleft licenses: each of them says, If you include code under this license in a larger program, the larger program must be under this license too. There is no way to make them compatible.

Linus on GPLv3 and ZFS

JoeBuck — Tue, 12 Jun 2007 20:43:09 +0000

If all of the kernel code were GPLv2 || GPLv3, it could be combined with a GPLv3 ZFS. The collection as a whole would be GPLv3 only if ZFS were added, but ZFS could be a module, and everything would be legal, while embedded software developers who want to do DRM could still use the rest of Linux (except ZFS).

Google is your friend

rfunk — Tue, 12 Jun 2007 20:21:14 +0000

http://www.opensolaris.org/os/community/zfs/