[Xenomai-core] [CAN] SIOC* clash

[Xenomai-core] [CAN] SIOC* clash

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 583 bytes --]

Hi Wolfgang,

just realised that there is an ABI mine laid in rtcan.h:

#ifndef SIOCGIFINDEX
#define SIOCGIFINDEX \
            _IOWR(RTIOC_TYPE_CAN, 0x00, struct ifreq)
#endif

So, if you happen to have included sys/ioctl.h earlier in your
application, an incompatible IOCTL code will be used in user space.

I recommend to switch to existing IOCTL codes where available (that's
what we do in RTnet as well). This may also mean betting on the current
IOCTL assignments in Socket-CAN, that they will survive unchanged the
eventual merge into Linux mainline.

Jan


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 250 bytes --]

[Xenomai-core] Re: [CAN] SIOC* clash

[Wolfgang Grandegger]

Jan Kiszka wrote:
> Hi Wolfgang,
> 
> just realised that there is an ABI mine laid in rtcan.h:
> 
> #ifndef SIOCGIFINDEX
> #define SIOCGIFINDEX \
>             _IOWR(RTIOC_TYPE_CAN, 0x00, struct ifreq)
> #endif
> 
> So, if you happen to have included sys/ioctl.h earlier in your
> application, an incompatible IOCTL code will be used in user space.
> 
> I recommend to switch to existing IOCTL codes where available (that's
> what we do in RTnet as well). This may also mean betting on the current
> IOCTL assignments in Socket-CAN, that they will survive unchanged the
> eventual merge into Linux mainline.

D'accord, I will change this with the next update.

Wolfgang.