On 20.07.2011 00:02, Marc Kleine-Budde wrote: > > The funny thing is you have to disable self reception, it's enabled by > default. > > > However - it turned out in a discussion here (or on -developers) some > time ago that the IFF_ECHO via self reception has a drawback. The > information that the rx'ed packet is a IFF_ECHO'ed one gets lost.
This should be fixed in general. We had two attempts of the networking guys to kill a sk-reference in the skb %-) But all the stable kernels do not have this problem in the latest versions, e.g. for the 2.6.35 the fix 'can_raw: Fix skb_orphan_try handling' became part of the 2.6.35.4. For that reason i do some checks on each upcoming net-next-2.6 if the CAN frame flow is still working like expected. The tests have been committed to the Linux test project (LTP) and also to the SocketCAN SVN in trunk/test, see http://svn.berlios.de/wsvn/socketcan?op=comp&compare[]=%2F@1264&compare[]=%2F@1265 The other thing could be the new message flags introduced in the can_raw sockets in 2.6.37 by Kurt: http://lxr.linux.no/#linux+v2.6.39/Documentation/networking/can.txt#L474 But i did not hear anything bad from this functionality so far. Anything new i should know? Regards, Oliver _______________________________________________ Socketcan-users mailing list [email protected] https://lists.berlios.de/mailman/listinfo/socketcan-users
