now that #188 / #486 are done, we could actually use mutex critical sections in the x86 kernel. this would prevent of the tracing-in-ISRs deadlock rakes i've stupidly gone out of my way to step on.

disabling all IRQs in a critical section is not ideal. we could probably pick and choose a bit. but this is probably better than not doing it. if spin_lock_irqsave is good enough for linux, it's good enough for us.