Hi Antonio,

On Sat, Dec 27, 2014 at 03:30:53PM +0100, Antonio Quartulli wrote:
> Hi Markus,
> 
> first of all you should know that I really appreciate your hard work.
> Thank you very much for this!
> 
> On 26/12/14 12:41, Markus Pargmann wrote:
> > Signed-off-by: Markus Pargmann <mpa@pengutronix.de>
> > ---
> >  packet.h | 5 +++++
> >  1 file changed, 5 insertions(+)
> > 
> > diff --git a/packet.h b/packet.h
> > index facd1feddd4e..5fd0d814b6de 100644
> > --- a/packet.h
> > +++ b/packet.h
> > @@ -18,6 +18,11 @@
> >  #ifndef _NET_BATMAN_ADV_PACKET_H_
> >  #define _NET_BATMAN_ADV_PACKET_H_
> >  
> > +#ifdef __KERNEL__
> > +#include <linux/bitops.h>
> > +#include <uapi/linux/if_ether.h>
> > +#endif /* __KERNEL__ */
> 
> Unfortunately we can't do this :-(
> 
> Patches sent to the kernel cannot contain conditional code on being in
> the kernel or not (same is true for checks against a particular kernel
> version - such code can't be in the kernel).
> 
> Patches sent to the kernel must assume that they are only for the kernel
> (and for that particular version we are sending the patches against).

Oh, I checked by grepping through the kernel before changing the patch
like this.
	git grep "^#if.*def.*__KERNEL__" | grep -v include | wc -l
	145

So there are 145 uses of an ifdef __KERNEL__ outside of include
directories (I excluded include directories as they may be exported to
userspace). So I thought it would be ok.

> Of course this makes everything a bit trickier for us since we share
> files between kernel and userspace .....
> 
> This is one of the reasons why we have compat.c/h in the batman-adv
> package: you will see that in those files we had to implement all sort
> of hackish/dirty tricks in order to keep the code in the other files
> clean (compat.h/c are not sent to the kernel but are part of the
> batman-adv out-of-the-tree package).

Yes I saw that. I still don't really understand why this whole out of
tree package is necessary. But I am wondering if it wouldn't be easier
to develope and build directly in the kernel tree.

> 
> However, as far as I know, you should be able to simply do:
> 
> #include <linux/if_ether.h> (without the uapi subfolder)
> 
> and this should happily work for both the kernel and the userspace. Have
> you tried that? I might be wrong - it's quite some time I haven't used
> uapi headers.

Yep, that works for the kernel, thanks. However, bitops.h remains as we
don't have bitops.h for the userspace.

> Otherwise we might need to find a different solution - or live we what
> we have now :)

Yeah it works normally as the includes leak through other include files.
But I prefer explicit includes. I am thinking about some solution for
the bitops.h

Best Regards,

Markus

-- 
Pengutronix e.K.                           |                             |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |