Ok, I think we must take another approach here. It is damn hard to write a working locking mechanism that has low latency for the periodic work and is still free of races. It is so hard, because we call into the driver from many different contexts. (And even if we consider dscape, and we really want to, I don't even know all the contexts for sure). This would always result in _lots_ of magic scattered all around the code and I am not really a friend of magic. ;)
The other approach would be to insert voluntarily preemtion points into the heavy LO recalc code. I will come up with a new experimental patch, as soon as I finished it. Thanks again for your help in testing the stuff. -- Greetings Michael. [2007: Knorkator (Germany) - 12 points] _______________________________________________ Bcm43xx-dev mailing list [email protected] http://lists.berlios.de/mailman/listinfo/bcm43xx-dev
