So, if I understand it correctly, when LWKT-scheduler schedules any thread with TDPRI_USER_* priority to run, then usched of the process, associated with chosen thread, decide what user process to schedule and according to usched's scheduling discipline this chosen process may differ from that process, which thread is initially scheduled by LWKT-scheduler. And fact that LWKT-scheduler will schedule user threads according to round-robin discipline give us possibility to implement multiply userland schedulers because every user thread will have it chance to run, thus every usched will be active at least once. Right? But if so, then we have a problem: those uscheds which have more threads will get a higher priority over those which have less threads.
-- Sergey Glushchenko
