Hi, Like earlier, Paul's rcutorture is fixed by #1, but now it actually makes sense and I can explain how it happens. Therefore this patch is _much_ better than the last one :-)
I was going to push 1-3 into sched/urgent, #3 removes the horrible BUG_ON crap in kernel/smp.c and trades them for slightly less horrible unions. I was planning on keeping #4,#5 back for the next round, but if Linus wants them now, that can certainly be arranged too. I have some further work on top that creates and uses irq_work_queue_remote() to replace many (hopefully all) smp_call_function_single_async() users so we can get rid of that head-ache, but those need to cook a little more. (#5 has 'trivial' conflicts with patches from hch that are targeted at the next round)
