[uml-devel] Re: SMP, locking and UML

[Jeff D. <jd...@ka...>]

gr...@na... said:
> schedule_tail() called from new_thread_handler()
>     when you start idle threads in turn tries to unlock rq lock for
>     its CPU, but this lock is not locked. 

Yes, that's definitely a problem.

> I suspect that the problem
>     is you should put these run queue locks to locked state prior to
>     each idle task start. Nikita have different opinion that I do not
>     understand yet.

That sounds reasonable, what does Nikita think about this?

				Jeff