[Cbe-oss-dev] [PATCH 2/2] spufs: Remove spurious WARN_ON for spu_deactivate for NOSCHED contexts

Christoph Hellwig hch at lst.de
Tue Jul 17 23:03:51 EST 2007


On Tue, Jul 17, 2007 at 10:08:51AM +1000, Michael Ellerman wrote:
> It's a little different, but a little the same. My point is from a
> user's perspective the scheduling policy shouldn't effect the fact that
> I can core dump it etc.

NOSCHED contexts are no scheduling policy.  It was a sort of
lite-isolated context setup (which of course never got tested at all
until now..).

> This is broken. I agree that accessing it with the debugger will screw
> with the timing behaviour, there's nothing we can do about that. But if
> I want to, I should still be able to debug it. And when I'm finished it
> should be rescheduled on the _same spu_.

Which means a shitload of special cases in the core spu scheduler for
a feature that has no real life purpose.  I'm dead set against that
special case.  Currently NOSCHED is just an isolated context without
the special loading procedure from the kernel perspective, and the
special treatment is very limited especially in the fastpath and
complicated code.




More information about the cbe-oss-dev mailing list