ZFS Targeted change against OpenZFS on Linux?

Can't complain if was developed by Linux for Linux.

Noob question: Could Linux copy something from BSD and re-license it under GPL? :-/
 
rbranco said:
I bet it's mostly exaggeration for clickbait.
Click to expand...
It's not.

It's not the first time there something like this happens and it won't be the last. They removed an exported symbol and the replacement one (writeback_iter) is marked EXPORT_SYMBOL_GPL and thus ZFS cannot use it.
 
If this comes to pass, I wonder if iXsystems will keep going with Truenas Scale, or if they'll pivot back to Core. I was planning on deploying a ZFS-friendly Linux distro for my wife sometime in the next 6 months, just so I could automate ZFS send/receive backups of the workstation to my servers. I guess that's a dead idea...
 
fmc000 said:
It's not.

It's not the first time there something like this happens and it won't be the last. They removed an exported symbol and the replacement one (writeback_iter) is marked EXPORT_SYMBOL_GPL and thus ZFS cannot use it.
Click to expand...
It was way worse when lots of crypto functions were exported GPL only. I'm sure a workaround will be found.
 
Yeah, I suspect ZFS people were already expecting this change once the two last mainline users were gone.

Nice timing BTW, kernel 6.18 which will be 2025LTS unless someting strange happens between now and December.
 
msplsh said:
Seems like the people signing off aren't paying attention to out-of-tree consumers, so I guess the point is to raise a stink visibility before they make a huge mistake (by accident??)
Click to expand...
Oh they do pay attention, believe me, especially HCH who is a prominent GPL evangelist.

Anyways, ZFS friendly distros will probably revert the change, thus allowing ZFS to compile and I guess normal users that build their own kernel will follow suit. A third option is that this change could be carried by OpenZFS directly in the form of an additional patch.
 
cracauer@ said:
Looks pretty serious to me. Certainly not clickbait.
Click to expand...
Yeah, I realized that was my thread title that could appear as clickbait-y. I tried to fix that to the best of my English knowledge (to my justification it was 2AM in the third consecutive sleepless night in a hospital bed but I guess I should have done better nonetheless).
 
A boon for FreeBSD. And while they’re futzing around with meme-cachefs, we’re still winning with ZFS.

Now if we could fix the ZFS ARC and mmap/page cache issue, it'll further instill FreeBSD as the better option.
 
If you need to be a kernel module—and if you're a file system, you really do need to be—then limiting interface access to the kernel means becoming less effective & less efficient, and more, until, at the end of the road, you're no longer operating as a kernel module.

This has all the ingredients of becoming a power play, if it isn't already one; technical reasons will become obfuscated. I imagine the OpenZFS developer community isn't looking forward to that. Linu{s|x} has free rein, until the powers that be say otherwise. Somehow, I don't think mere end users would be able to change that.

Background reading from some 5 years ago: Linus Torvalds says “Don’t use ZFS”—but doesn’t seem to understand it
 
fmc000 said:
Oh they do pay attention, believe me, especially HCH who is a prominent GPL evangelist.

Anyways, ZFS friendly distros will probably revert the change, thus allowing ZFS to compile and I guess normal users that build their own kernel will follow suit. A third option is that this change could be carried by OpenZFS directly in the form of an additional patch.
Click to expand...
I believe stuff marked as EXPORT_SYMBOL_GPL can't be reverted back to workaround a license issue. You either fork before the change or do something else.

It's more serious than I thought but less serious than the past issue, IMHO.
 
  • Like
Reactions: mer
rbranco said:
I believe stuff marked as EXPORT_SYMBOL_GPL can't be reverted back to workaround a license issue. You either fork before the change or do something else.

It's more serious than I thought but less serious than the past issue, IMHO.
Click to expand...
I'm not talking about this. I'm talking about reverting the change that removed the old obsolete interface that ZFS currently uses.

The old interface works perfectly fine as it is, there is absolutely no need to use the new one (which I'm sure has its merits, otherwise it would not have been developed in the first place; I simply refuse to believe that the only motivation behind this change is removing an accessible function just to replace it with a restricted one).
 
fmc000 said:
I'm not talking about this. I'm talking about reverting the change that removed the old obsolete interface that ZFS currently uses.

The old interface works perfectly fine as it is, there is absolutely no need to use the new one (which I'm sure has its merits, otherwise it would not have been developed in the first place; I simply refuse to believe that the only motivation behind this change is removing an accessible function just to replace it with a restricted one).
Click to expand...

 
Is it realistic for OpenZFS to relicense under the GPL? If that ever happened, Linux users would gain enormously since ZFS could finally become a first-class, in-tree filesystem. FreeBSD users wouldn’t lose access to ZFS, but they would face political and licensing complications. Rather than presenting ZFS as a native FreeBSD filesystem, it would have to be described as a GPL-licensed kernel module bundled with the system.
 
CyberCr33p said:
Is it realistic for OpenZFS to relicense under the GPL? If that ever happened, Linux users would gain enormously since ZFS could finally become a first-class, in-tree filesystem. FreeBSD users wouldn’t lose access to ZFS, but they would face political and licensing complications. Rather than presenting ZFS as a native FreeBSD filesystem, it would have to be described as a GPL-licensed kernel module bundled with the system.
Click to expand...
My goodness, NO. Who cares that ZFS becomes a "first-class" Linux filesystem? Quotes intended, of course. It's their problem, not ours.
 
epower53 said:
If this comes to pass, I wonder if iXsystems will keep going with Truenas Scale, or if they'll pivot back to Core. I was planning on deploying a ZFS-friendly Linux distro for my wife sometime in the next 6 months, just so I could automate ZFS send/receive backups of the workstation to my servers. I guess that's a dead idea...
Click to expand...
Lol, they're already well past that point. There is no more TrueNAS SCALE or CORE. There is only TrueNAS CE (Community Edition), which is basically the continuation of SCALE.
 
CyberCr33p said:
Is it realistic for OpenZFS to relicense under the GPL? If that ever happened, Linux users would gain enormously since ZFS could finally become a first-class, in-tree filesystem. FreeBSD users wouldn’t lose access to ZFS, but they would face political and licensing complications. Rather than presenting ZFS as a native FreeBSD filesystem, it would have to be described as a GPL-licensed kernel module bundled with the system.
Click to expand...

ZFS wouldn't have to be become GPL. It would just have to drop its custom license and become BSD or MIT licensed. That would make it compatible with the GPL.

Realistic? No, it's Oracle.
 
You must log in or register to reply here.
Back
Top