Re: in_interrupt race

[Rusty Russell <rusty@rustcorp.com.au>]

In message <20020423093151.A17302@flint.arm.linux.org.uk> you write:
> On Tue, Apr 23, 2002 at 01:25:24PM +1000, Rusty Russell wrote:
> > Yes: the old CPU happens to be processing an interrupt now.
> > The neat solution is to follow Linus' original instinct and make
> > PREEMPT an option only for UP: I only like preempt because it brings
> > UP into line with SMP, effectively enlarging the SMP userbase to reasonable
> > size.
> 
> > -bool 'Preemptible kernel' CONFIG_PREEMPT
> > +dep_bool 'Preemptible kernel' CONFIG_PREEMPT $CONFIG_SMP
> > -bool 'Preemptible Kernel' CONFIG_PREEMPT
> > +dep_bool 'Preemptible Kernel' CONFIG_PREEMPT $CONFIG_SMP
> 
> Do you really mean that CONFIG_PREEMPT is only available if CONFIG_SMP is
> 'y' or undefined?

<sigh>... Of course that should be reversed.
if [ "$CONFIG_SMP" != y ]; then
   bool 'Preemptible Kernel' CONFIG_PREEMPT
fi

Thanks,
Rusty.
--
  Anyone who quotes me in their sig is an idiot. -- Rusty Russell.