That is great explanation from Linus. Thanks for linking it. He uses AFS (Andrew File System) as an example in that post which I think would be very analogous to ZFS. Linus says
> Historically, there's been things like the original Andrew filesystem
> module: a standard filesystem that really wasn't written for Linux in
> the first place, and just implements a UNIX filesystem. Is that
> derived just because it got ported to Linux that had a reasonably
> similar VFS interface to what other UNIXes did? Personally, I didn't
> feel that I could make that judgment call. Maybe it was, maybe it
> wasn't, but it clearly is a gray area.
> Personally, I think that case wasn't a derived work, and I was willing
> to tell the AFS guys so.
> To me, there seems like their is a good chance than ZFS and the Linux Kernel can be distributed as a combined work (But I Am Not A Lawyer!).
This is a situation the GPL tries to prevent by explicitly using the language "combined work" - however copyright law doesn't typically have a concept of a combined work, only a derived work. It will take a court to settle this.
Also, if you feel that kernel modules should be GPL without exception then you should lobby the Linux kernel maintainers to export all symbols with the EXPORT_SYMBOL_GPL macro. As it now stands, the division of exported symbols into two classes ("normal" and "GPL only") is a strong indication that the copyright holders do believe that kernel modules are not necessarily derivative works.