> I've sent mine to Niels, so he can look how it (should?) works. > For instance, it don't bufferized packets on driver-side: each packet > received raise the read_hook semaphore, > A good design, the shorter the better. Another note though: the Rx uses a completely unrelated register (the CBR, not that it rings a bell), I tried to use that one, but it didn't work. (When you come to think of it, it is quite logical). What I will do is replicate the driver behaviour at init time, perhaps my grave bug is there. Niels