svn commit: r344487 - in head/sys: conf gnu/gcov

[Warner Losh]

On Tue, Feb 26, 2019 at 8:46 AM Shawn Webb <shawn.webb at hardenedbsd.org>
wrote:

> On Mon, Feb 25, 2019 at 06:18:42PM -0800, Rodney W. Grimes wrote:
> > > > The modest increase in activation energy for that task seems worth it
> > > > for the short-term gains of reduced integration cost (this code will
> > > > greatly improve our ZFS-on-Linux test coverage.)
> > > >
> > > > Rod rightly points out that we haven't accepted SPDX tags alone as
> > > > license statements.  The standard GPL v2.0 boiler plate should be
> added
> > > > to this file along side the tag.
> > >
> > > I've copied the full copyright attribution that is in the
> > > corresponding files on Linux. Is there some reason why FreeBSD
> > > requires the files to be inflated with the full license text where the
> > > original lacks it?
> >
> > I think for a few reasons, I doubt you copied the whole distribution
> > that this file came from, as I am sure that distribution included
> > a LICENSE file.  Second if you actually read the GPL v2 documentation
> > and follow what it says it says you must do this, just because some
> > one else does not follow the rules of what the GPL v2 says does not
> > give us to knowingling not do it.  Third this is a particular dangerious
> > area for BSD to be mixing a GPL code with its kernel, to my knowlege
> > we have never had any gpl code in the kernel, no have we ever
> > allowed it, but thats a seperate argument, that should be made.
>
> Would the arm64 DTS/DTB files count as "GPL code in the kernel?"
>

No. dts gets compiled into dtb. dtb is a separate work loaded by the boot
loader. While one can compile it into the kernel, we don't ship like that.

There's also a question as to whether or not these files are text
representation of the hardware, and there being only one way to represent
it (making it not copyrightable under at least US case law since it's a
database). That question hasn't been litigated. Many hardware companies
also dual license the dts. Since we're not incorporating it into the
kernel, but merely using it as a standardized table (there's a separate
group that controls the dts/dtb spec), I think we're safe from that angle
as well.

There's benefit from having it in-tree because the version of the spec
evolves over time, and having the right version makes it harder to push
this off into a port. Also, having them in-tree makes the project's
compliance with GPL a no-op because it's all there in the open in a tagged
VCS.

tl;dr: I don't think this is an issue.


> I, too, would like less GPL in project, both in userland in kernel.
> But, I can understand the desire for gcov. Note that I'm not
> advocating either way that FreeBSD perform an action. ;)
>

Given this is for TEST kernels, there's no issue here.  While we'd like to
be GPL free, let's not cut off our nose to spite our face. Given the
interactions between different bits, the FreeBSD selling point of "well
integrated" I think trumps the purity arguments because it's not code
anybody would ever ship (and if they did, they'd get the proper warnings).

Warner