On Fri, Jun 12, 2020 at 11:27:21AM +0200, Thomas Gleixner wrote:
> "Paul E. McKenney" <paulmck@xxxxxxxxxx> writes:
> > +static __always_inline bool rcu_needs_irq_enter(void)
> > +{
> > + return !IS_ENABLED(CONFIG_TINY_RCU) &&
> > + (context_tracking_enabled_cpu(smp_processor_id()) || is_idle_task(current));
>
> This reintroduces the #PF problem which started the whole conditional
> RCU entry discussion:
>
> https://lore.kernel.org/lkml/20200515235125.628629605@xxxxxxxxxxxxx/
>
> and which made us all come to the conclusion that we can do it always
> conditional. No biscuit for you. :)
We can only be thankful that source-code control systems mean that my
coding session yesterday afternoon will have no permanent effect.
Let this be a lesson to all of you: Hacking RCU while sleep-deprived
is a really bad idea. ;-)
Thanx, Paul
