On Sat, Dec 09, 2017 at 07:32:38PM +0200, Thomas Backlund wrote:
> Den 09.12.2017 kl. 19:13, skrev Greg Kroah-Hartman:
> > On Sat, Dec 09, 2017 at 07:56:40AM +0000, Ivan Kozik wrote:
> > > On Sat, Dec 9, 2017 at 7:45 AM, Greg Kroah-Hartman
> > > <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
> > > > On Sat, Dec 09, 2017 at 03:34:24AM +0000, Ivan Kozik wrote:
> > > > > I saw no problems on 8 of 9 machines, but the last one had a problem
> > > > > because it used NVIDIA drivers (387); DKMS reported:
> > > > >
> > > > > FATAL: modpost: GPL-incompatible module nvidia-drm.ko uses GPL-only
> > > > > symbol 'ex_handler_refcount'
> > > > > //usr/src/linux-headers-4.14.0-11-common/scripts/Makefile.modpost:92:
> > > > > recipe for target '__modpost' failed
> > > > > make[3]: *** [__modpost] Error 1
> > > >
> > > > Is this a new issue? Does 4.14.4 have this issue?
> > >
> > > I believe it is a new issue, because I have a 4.14.4 build and an
> > > NVIDIA DKMS log for that 4.14.4 showing build success.
> > >
> > > > Odd, is 564c9cc84e2a ("locking/refcounts, x86/asm: Use unique .text
> > > > section for refcount exceptions") causing this?
> > >
> > > That was my guess too, but I did not verify.
> >
> > That feels really wrong here, I'd like to get some confirmation before I
> > add this patch...
> >
>
> It's needed.
>
> The reason you hit in 4.14.5 queue is because of:
>
> [PATCH 4.14 64/75] locking/refcounts, x86/asm: Enable
> CONFIG_ARCH_HAS_REFCOUNT
>
> From foo@baz Wed Dec 6 18:04:41 CET 2017
> From: Kees Cook <keescook@xxxxxxxxxxxx>
> Date: Sat, 2 Sep 2017 13:09:46 -0700
> Subject: locking/refcounts, x86/asm: Enable CONFIG_ARCH_HAS_REFCOUNT
>
>
> that does this:
>
> - select ARCH_HAS_REFCOUNT if BROKEN
> + select ARCH_HAS_REFCOUNT
>
>
>
> So it exposes previously hidden code....
Ah yes, thanks for the explaination. I've queued up that patch now.
greg k-h