Re: nfs: __setup_str_nfs_root_setup causes a section type conflict

[Helge Deller]

On 08/04/2009 12:11 AM, Trond Myklebust wrote:
> On Mon, 2009-08-03 at 23:52 +0200, Sam Ravnborg wrote:
> > On Mon, Aug 03, 2009 at 05:07:23PM -0400, Trond Myklebust wrote:
> > > On Mon, 2009-08-03 at 22:57 +0200, Frans Pop wrote:
> > > > $ make init/do_mounts.o
> > > >    CHK     include/linux/version.h
> > > >    CHK     include/linux/utsrelease.h
> > > >    SYMLINK include/asm ->  include/asm-parisc
> > > >    CALL    scripts/checksyscalls.sh
> > > >    CC      init/do_mounts.o
> > > >
> > > > while:
> > > >
> > > > $ make fs/nfs/nfsroot.o
> > > >    CHK     include/linux/version.h
> > > >    CHK     include/linux/utsrelease.h
> > > >    SYMLINK include/asm ->  include/asm-parisc
> > > >    CALL    scripts/checksyscalls.sh
> > > >    CC      fs/nfs/nfsroot.o
> > > > fs/nfs/nfsroot.c:403: error: __setup_str_nfs_root_setup causes a section type
> > > > conflict
> > > To me that looks like some kind of compiler bug.
> > unspecified behaviour is a better name for it.
> >
> > We have two variables we have forced a section on.
> > One variable is marked RO by the compiler while the other is not.
> > This results in a section type conflict because all
> > symbols needs to have the same falgs.
> >
> > We usually triggers this with powerpc builds (64 bit IIRC),
> > and now with parisc too.
> >
> > The only way to fix it is to move one of the offending
> > variables to __initdata.
> > There is two variales to consider - the compiler only
> > mention one of them.

Yes.
It seems this was already once described on the parisc mailing list:
http://article.gmane.org/gmane.linux.ports.parisc/1816

Helge


> > Trying to declare the variables const etc usually has
> > no good effect.
>
> So why would it be happening in the NFSroot case, but not in
> do_mounts.c? They're both using the standard __setup() macro with a
> constant string, and an __init function argument.
>
> Futhermore, when grepping for '__initconst', it seems that the
> combination with 'static const' is the norm rather than the exception.

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/