Even if we ignore the ZFS license itself, the SF Conservancy appears to be arguing that _all_ proprietary kernel modules are GPL violations.
Most previous arguments about zfs.ko were about the derived work clause, which is much more grey. I belive that nvidia.ko would be more likely to be ruled a derived work than ZFS.ko, but that's hard to tell.
It does seem silly to argue about ZFS in this context. ZFS is open source and much less offensive to kernel developers than nvidia.ko.[1]
If nvidia.ko is fine, then so is zfs.ko. If nvidia.ko isn't fine, zfs.ko may or may not be fine. Ubuntu has been shipping nvidia.ko for years.
This is the argument the Illumos developers make about including the GPL'ed KVM code, KVM does not depend on Illumos to exist, as such does loading the KVM module into an Illumos kernel constitute a derived work?
Conveniently, this is also precedent that could be set during the VMWare case the SFC is also perusing - we could either see ZFS relegated to being available anywhere but Linux and binary kernel modules totally banned (bad for nVidia, good for AMD), or a total change in the landscape of open source licensing (for better or worse).
This is entirely unlike VMWare, who really did make and ship their own butchered linux distro with big binary blobs planted right in the middle.
Is that true? I thought Ubuntu's support for nvidia.ko involved downloading source code to the local machine and compiling it there, to avoid distributing a GPL violation.
Pure distribution of source with no binaries is undeniably different. When distributing source code and no binaries, requirements in those sections of GPLv2 and CDDLv1 that cover modification and/or binary (or “Executable”, as CDDLv1 calls it) distribution do not activate. Therefore, the analysis is simpler, and we find no specific clause in either license that prohibits source-only redistribution of Linux and ZFS, even on the same distribution media.
(Conservancy mentions the possibility of contributory/indirect infringement, though. Also, proprietary kernel modules written specifically for Linux, or using Linux-specific kernel features, may be GPL violations by virtue of their source code being a derivative work. But in the case of ZFS and several other cases such as OpenAFS and many hardware drivers, the source code is not a derivative work of Linux.)
I believe the key here is the distribution of the resulting combination. Were you to apply these patches after the distribution of Ubuntu, then would this still apply? Asked another way, if Ubuntu sets up a separate package channel that installs ZFS through a downstream patch via apt, would it be in violation of the license?