Re: Linux 4.9.180 build fails with gcc 9 and 'cleanup_module' specifies less restrictive attribute than its target â

[Miguel Ojeda]

On Fri, Aug 2, 2019 at 12:33 PM Greg KH <greg@xxxxxxxxx> wrote:
>
> On Fri, Aug 02, 2019 at 12:19:33PM +0200, Miguel Ojeda wrote:
> > On Fri, Aug 2, 2019 at 10:17 AM Rolf Eike Beer <eb@xxxxxxxxx> wrote:
> > >
> > > Am Samstag, 8. Juni 2019, 14:00:34 CEST schrieb Miguel Ojeda:
> > > > On Thu, Jun 6, 2019 at 8:59 PM Greg KH <greg@xxxxxxxxx> wrote:
> > > > > "manually fixing it up" means "hacked it to pieces" to me, I have no
> > > > > idea what the end result really was :)
> > > > >
> > > > > If someone wants to send me some patches I can actually apply, that
> > > > > would be best...
> > > >
> > > > I will give it a go whenever I get some free time :)
> > >
> > > I fear this has never happened, did it?
> >
> > No. Between summer, holidays and a conference I didn't get to do it.
> >
> > Done the minimal approach here:
> >
> >   https://github.com/ojeda/linux/commits/compiler-attributes-backport
> >
> > Tested building a handful of drivers with gcc 4.6.4, 8.3.0 and 9.1.1.
> >
> > Greg, I could backport the entire compiler_attributes.h, but given
> > this is stable, we are supposed to minimize changes, right?
> >
> > I tried to imitate what you do in other stable patches, please check
> > the Cc:, Link: lines and the "commit ... upstream" just in case.
>
> If only those 2 patches are all that is needed, nice!  I'll gladly take
> them, can you send them to me (and cc: the stable list) in email so I
> can queue them up for the next round of releases after this one?

At least for that particular problem, yeah -- I haven't done a full allmod.

By the way, I just checked 4.14.y and I noticed you had already
backported it, although going for another solution:

+#if GCC_VERSION >= 90100
+#define __copy(symbol)                 __attribute__((__copy__(symbol)))
+#endif

and then:

+#ifndef __copy
+# define __copy(symbol)
+#endif

Cheers,
Miguel