On Fri, Oct 20, 2023 at 03:48:36PM +0200, Peter Zijlstra wrote:
> On Fri, Oct 20, 2023 at 02:31:13PM +0200, Frederic Weisbecker wrote:
>
> > Yeah I can't say I really like the old coverage of PF_IDLE either. The new one
> > (after Liam's patch) is only halfway better defined though: it makes the boot
> > CPU's idle behave quite well: PF_IDLE is set on idle entry. And secondary
> > CPU's idle behave quite well also except when they go offline and then online
> > again. And then the secondary boot code becomes PF_IDLE.
>
> Bah offline, yeah, we should just not do that :-)
>
> > We probably need something like this:
> >
> > diff --git a/kernel/cpu.c b/kernel/cpu.c
> > index 3b9d5c7eb4a2..b24d7937b989 100644
> > --- a/kernel/cpu.c
> > +++ b/kernel/cpu.c
> > @@ -1394,7 +1394,9 @@ void cpuhp_report_idle_dead(void)
> > {
> > struct cpuhp_cpu_state *st = this_cpu_ptr(&cpuhp_state);
> >
> > + current->flags &= ~PF_IDLE;
> > BUG_ON(st->state != CPUHP_AP_OFFLINE);
> > +
> > rcutree_report_cpu_dead();
> > st->state = CPUHP_AP_IDLE_DEAD;
> > /*
> > @@ -1642,6 +1644,8 @@ void cpuhp_online_idle(enum cpuhp_state state)
> > {
> > struct cpuhp_cpu_state *st = this_cpu_ptr(&cpuhp_state);
> >
> > + current->flags |= PF_IDLE;
> > +
> > /* Happens for the boot cpu */
> > if (state != CPUHP_AP_ONLINE_IDLE)
> > return;
>
> Yeah that works I suppose.
Booting up kernels being what it is, there might not be a completely
pretty solution. Not that I would say "no" to such a solution should
it appear, mind you! ;-)
Thanx, Paul
