On 05.12.2010 10:54, Michal Sojka wrote: > On Sun, 05 Dec 2010, Marc Kleine-Budde wrote: >> On 12/05/2010 01:00 AM, Michal Sojka wrote:
>>> I'm experiencing strange kernel panics with CAN gateway running on >>> MPC5200. The panic happens with 2.6.36.1 where I manually added gw.c and >>> hw.h from subversion. It also happens with 2.6.33.7 where the whole >>> socketcan was copied from SVN. Interestingly, it does not occur with >>> 2.6.33.7-rt29 (rt_preempt) with the same socketcan from svn. The details >>> are bellow. This is 100% reproducible and the gateway configuration is >>> "cangw -A -s can0 -d can1". Do you have any clue what can be the cause >>> of the panic? >> >> Maybe it's a locking problem. With RT the whole locking stuff is changed >> in the background. Use the 2.6.36 kernel go to the kernel hacking menu >> and switch on all the lock checking stuff, then try to reproduce the >> problem. Hi Michal, can you give some more details about you setup, that leads to the crash? 1. Are you running the setup using this patch: http://rtime.felk.cvut.cz/gitweb/shark/linux.git/commitdiff/92487e4f349cd7518cc3202662f42fea7d42ba73 2. Are you using the '-e' option that creates the local echo of sent CAN frames? We're using the cangw in a 'productive' research environment for over 6 month now for routing frames between real/virtual, virtual/real and real/real CAN interfaces without getting the crash. Our non-RT setup runs a 2.6.28.10 on a mpc5...@400mhz patched with the SocketCAN SVN (removing the original 2.6.28.10 CAN subsystem). I currently worry about your patch (see 1.) as it may lead to some hard/soft-irq mixup ... Regards, Oliver _______________________________________________ Socketcan-users mailing list [email protected] https://lists.berlios.de/mailman/listinfo/socketcan-users
