ZFS native on Ubuntu 16.04 LTS

[TuxDude]

So when I initially read this thread, I thought that there might be some GPL violations going on with this. And now after coming across this article today it sounds like this is almost certainly a GPL violation (as would be Proxmox with ZFS included)

Interpreting, enforcing and changing the GNU GPL, as applied to combining Linux and ZFS — Free Software Foundation — working together for free software

 

[canta]

So when I initially read this thread, I thought that there might be some GPL violations going on with this. And now after coming across this article today it sounds like this is almost certainly a GPL violation (as would be Proxmox with ZFS included)

Interpreting, enforcing and changing the GNU GPL, as applied to combining Linux and ZFS — Free Software Foundation — working together for free software

I believe this is not vialotion on ZFS on linux

Proxmox and ubuntu, provide ZoL as a package that will wrapped in DKMS, not directly to kernel.

this is a hassle when upgrading ZoL, the systme must run DKMS wrapper again to make sure picks up correctly

I had an issue before, when upgrading ZoL on centos 7.X will rendered dead ZoL, Manual running ZoL script and editing was the only way to fix.

 

[TuxDude]

Well I'm not a lawyer, but these snips from that article would seem to imply that it is a violation:

This means that when you distribute a work that contains material from a GPL-covered program, you must distribute the entire work, as a whole, under the conditions of appropriate versions of the GPL.

if you distribute modules meant to be linked together by the user, you have made them into a combined work, and you must release the entire combined work under the GNU GPL.

And this bit at the bottom to me implies that if Richard Stallman or the FSF did have a significant copywrite interest in the linux kernel, that they would be claiming a violation:

The FSF is not a copyright holder of Linux, except for small pieces copied from programs which we released under the Lesser GPL or GPL plus exceptions, and perhaps an obscure processor port. We do not presume on the strength of these pieces of code to tell the developers of Linux proper how to approach GPL enforcement for Linux, and we won't advocate a choice as if we had a say.

I suspect the entire situation is one of those legal grey-water areas that won't actually be decided until/unless someone who is a copywrite holder takes it to court and sees it through to the end to set a legal precedent.

 

[ttabbal]

It "feels" the same as the Nvidia binary drivers, or even the Tivo case. If those aren't a violation, I don't see how ZFS is. It's a kernel module, the interface code between the module and the kernel complies, seems alright to me. More importantly, the kernel developers that hold the copyrights have stated that they won't fight binary modules. So it's a moot point, as the only people that could bring a suit, have said they won't.

Of all the possible targets for GPL ire, ZFS seems a poor one. It's an open-source project, not the same license, but a reasonable one. It's a far better situation than Nvidia with the binaries wrapped with a thin source layer to interface to the kernel. I get where FSF/RMS are coming from, and I think we need people like them to keep everyone's eyes open. I just think there are better targets to worry over.

 

[TuxDude]

In the case of binary nVidia drivers - they aren't distributed together. And I agree that this is a stupid fight for a stupid reason and wish all the open-source projects could just get along. But it's also kinda a large grey cloud hanging over the situation - I wouldn't want to use the combo in production at enterprise scale if there is any risk of random internet people deciding they do want to fight over license issues because they were in a bad mood that day.

I'm not trying to say ZFS is bad, or that linux and ZFS shouldn't be distributed together. I just wish there was a clear answer to whether it is allowed to be distributed together like that or not.

 

[canta]

In the case of binary nVidia drivers - they aren't distributed together. And I agree that this is a stupid fight for a stupid reason and wish all the open-source projects could just get along. But it's also kinda a large grey cloud hanging over the situation - I wouldn't want to use the combo in production at enterprise scale if there is any risk of random internet people deciding they do want to fight over license issues because they were in a bad mood that day.

I'm not trying to say ZFS is bad, or that linux and ZFS shouldn't be distributed together. I just wish there was a clear answer to whether it is allowed to be distributed together like that or not.

these always happen since LGPL and GPL that linux kernel is used

not to explain in detail of my work, I had been told after moving to other department that involve linux and GPL development. I can not use GPL to link my works since the company that I am working, will never releases the source code due on competition and security .
other twisted, Yes I can lin LGPL and never release source code .

I would say, we are using Red Hat Model ...

ZFS model is != ZoL
ZoL born due on license compatibility and work well on DKMS space so far..
ideally, ZoL must be treated as supported filesystem and kernel module directly. this never happens due in != licensing..

how about GPL3? I am reading in detail...

 

[zyntax]

I'm having this issue still on my esxi server with flashed M1015 passthrough (newest firmware) to Ubuntu 17.04.
No matter what i do i see no disks in esxi newer than 6.2.
In any 6.5 version i cannot find disks, tried both mpt3sas.msix_disable=1 and mpt2sas.msix_disable=1 (and both). It will not work with esxi 6.5.x.
Even tried out Ubuntu 17.10 daily, that has a newer (less old) mpt2/3sas driver.

Still getting: mpt2sas_cm0: failure at /build/linux-qOfREr/linux-4.12.0/drivers/scsi/mpt3sas/mpt3sas_scsih.c:8937/_scsih_probe()!

Am i the only one? Works fine with older esxi. I'm thinking of either scrapping esxi for proxmox, or getting other hba card.
Will try out esxi 6.5 with passthrough on BSD and see if that picks the disks up.

 

[FuriousPy]

I'm having this issue still on my esxi server with flashed M1015 passthrough (newest firmware) to Ubuntu 17.04.
No matter what i do i see no disks in esxi newer than 6.2.
In any 6.5 version i cannot find disks, tried both mpt3sas.msix_disable=1 and mpt2sas.msix_disable=1 (and both). It will not work with esxi 6.5.x.
Even tried out Ubuntu 17.10 daily, that has a newer (less old) mpt2/3sas driver.

Still getting: mpt2sas_cm0: failure at /build/linux-qOfREr/linux-4.12.0/drivers/scsi/mpt3sas/mpt3sas_scsih.c:8937/_scsih_probe()!

Am i the only one? Works fine with older esxi. I'm thinking of either scrapping esxi for proxmox, or getting other hba card.
Will try out esxi 6.5 with passthrough on BSD and see if that picks the disks up.

Same problem on HP H220 LSI SAS2308 (9207-8i) with ubuntu, proxmox and redhat...

not problems with Windows 10 or Windows 2016 Server