KS2009: Realtime preemption

Then again, due to all the subtleness in the BKL (e.g. http://lwn.net/Articles/283066/ ), it wouldn't surprise me if many of these drivers would then be broken anyway, since they could depend on some other piece of code acquiring the BKL at some suitable point.

Anyway, I suppose that's the price of progress. One can at least hope that in the longer term, the elimination of the BKL will not only improve latencies and corner-case scalability, but also quality as the replacement locking will be better designed and understood.