Re: [Lse-tech] NUMA scheduler issue

[Rick L. <ric...@us...>]

    However, I still assert that the occasional task is being ativated
    on the same cpu that's running numatest, and that's why this is
    happening.

Looking at the code I can see where that *could* happen; we just
don't have any evidence that it's happening here.  I think it's more
likely the node-balancing code than the wakeup code (and if so, that's
intentional.) I think we're in agreement, though, that the proposed
patch probably addresses a symptom rather than a cause.

    A modified try_to_wakeup will fix this.

I agree, that would be a good thing.

Rick