On Wed, Dec 31, 2014 at 06:55:25PM +0100, Antonio Quartulli wrote:
On 27/12/14 18:03, Markus Pargmann wrote:
On Sat, Dec 27, 2014 at 03:30:53PM +0100, Antonio Quartulli wrote:
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 email@example.com
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.
these are probably pieces of code that have not yet been cleaned up. Still, this does not allow us to introduce more code like this: David (the networking tree maintainer) would refuse such additions.
Okay, thanks. I removed the bitops.h include for the moment as I don't have an idea how to keep the header usable for the userspace application.