Re: [PATCH 7/7] cpuidle/poll_state: replace cpu_relax with smp_cond_load_relaxed

[Mihai Carabas <mihai.carabas@oracle.com>]

>>>> cpu_relax on ARM64 does a simple "yield". Thus we replace it with
>>>> smp_cond_load_relaxed which basically does a "wfe".
>>>>
>>>> Suggested-by: Peter Zijlstra <peterz@infradead.org>
>>>> Signed-off-by: Mihai Carabas <mihai.carabas@oracle.com>
>>>> ---
>>>>    drivers/cpuidle/poll_state.c | 14 +++++++++-----
>>>>    1 file changed, 9 insertions(+), 5 deletions(-)
>>>>
>>>> diff --git a/drivers/cpuidle/poll_state.c b/drivers/cpuidle/poll_state.c
>>>> index 9b6d90a72601..440cd713e39a 100644
>>>> --- a/drivers/cpuidle/poll_state.c
>>>> +++ b/drivers/cpuidle/poll_state.c
>>>> @@ -26,12 +26,16 @@ static int __cpuidle poll_idle(struct cpuidle_device *dev,
>>>>    		limit = cpuidle_poll_time(drv, dev);
>>>> -		while (!need_resched()) {
>>>> -			cpu_relax();
>>>> -			if (loop_count++ < POLL_IDLE_RELAX_COUNT)
>>>> -				continue;
>>>> -
>>>> +		for (;;) {
>>>>    			loop_count = 0;
>>>> +
>>>> +			smp_cond_load_relaxed(&current_thread_info()->flags,
>>>> +					      (VAL & _TIF_NEED_RESCHED) ||
>>>> +					      (loop_count++ >= POLL_IDLE_RELAX_COUNT));
>>>> +
>>>> +			if (loop_count < POLL_IDLE_RELAX_COUNT)
>>>> +				break;
>>>> +
>>>>    			if (local_clock_noinstr() - time_start > limit) {
>>>>    				dev->poll_time_limit = true;
>>>>    				break;
>>> Doesn't this make ARCH_HAS_CPU_RELAX a complete misnomer?
>> This controls the build of poll_state.c and the generic definition of
>> smp_cond_load_relaxed (used by x86) is using cpu_relax(). Do you propose
>> other approach here?
> Give it a better name? Having ARCH_HAS_CPU_RELAX control a piece of code
> that doesn't use cpu_relax() doesn't make sense to me.

The generic code for smp_cond_load_relaxed is using cpu_relax and this 
one is used on x86 - so ARCH_HAS_CPU_RELAX is a prerequisite on x86 when 
using haltpoll. Only on ARM64 this is overwritten. Moreover 
ARCH_HAS_CPU_RELAX is controlling the function definition for 
cpuidle_poll_state_init (this is how it was originally designed).

Thanks,
Mihai