I've been trying to eliminate sources of unpredictability in the latency of interrupt response in my real-time application, and I'm struggling a bit with the CFS scheduler. If a SCHED_OTHER task was preempted the rt scheduler eventually calls put_prev_task() which results in a rb tree insert in the CFS scheduler. So the latency of my real-time task depends on the scheduling class of the task I'm interrupting and how many SCHED_OTHER tasks are in the runqueue, which seems like a priority inversion. Are there any ways to mitigate this other than trying to minimize the number of SCHED_OTHER tasks in the system? Ideally what I'd really like is to defer the work in put_prev_task_fair() until there's no real-time work to do. -Allen -- To unsubscribe from this list: send the line "unsubscribe linux-rt-users" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
