This is what I build so far. It needs testing, but it seems I got all things in place. I extended the rx-offload so it can be used easier, ie. without mailbox magic around. I needed this because the IRQ of c_can takes an initial sample of filled mailboxes, and until those have been emptied, no others are done. This requires a bit more context that is not in rx-offload so far. Adding something there smells like over-engineering the problem. The solution I took is to make advantage of rx-offload as much as I need, and no more. The changes to the c_can driver remain minimal now, so easier to test. I created the patches on top of Jeroen Hofstee's work. I will test this only tomorrow. Kind regards, Kurt
