[PATCH v8 01/11] cpuidle/poll_state: poll via smp_cond_load_relaxed()

[Catalin Marinas]

On Fri, Oct 18, 2024 at 12:00:34PM -0700, Ankur Arora wrote:
> Catalin Marinas <catalin.marinas at arm.com> writes:
> > On Thu, Oct 17, 2024 at 03:47:31PM -0700, Ankur Arora wrote:
> >> So maybe the right thing to do would be to keep smp_cond_load_timeout()
> >> but only allow polling if WFxT or event-stream is enabled. And enhance
> >> cpuidle_poll_state_init() to fail if the above condition is not met.
> >
> > We could do this as well. Maybe hide this behind another function like
> > arch_has_efficient_smp_cond_load_timeout() (well, some shorter name),
> > checked somewhere in or on the path to cpuidle_poll_state_init(). Well,
> > it might be simpler to do this in haltpoll_want(), backed by an
> > arch_haltpoll_want() function.
> 
> Yeah, checking in arch_haltpoll_want() would mean that we can leave all
> the cpuidle_poll_state_init() call sites unchanged.
> 
> However, I suspect that even acpi-idle on arm64 might end up using
> poll_idle() (as this patch tries to do:
> https://lore.kernel.org/lkml/f8a1f85b-c4bf-4c38-81bf-728f72a4f2fe@huawei.com/).
> 
> So, let me try doing it both ways to see which one is simpler.
> Given that the event-stream can be assumed to be always-on it might just
> be more straight-forward to fallback to cpu_relax() in that edge case.

I agree, let's go with the simplest. If one has some strong case for
running with the event stream disabled and idle polling becomes too
energy inefficient, we can revisit and add some run-time checks.

-- 
Catalin