Re: [PATCH 3/4] arm64: barrier: Add smp_cond_load_relaxed_timewait()

[Ankur Arora]

Catalin Marinas <catalin.marinas@xxxxxxx> writes:

> On Mon, Feb 03, 2025 at 01:49:10PM -0800, Ankur Arora wrote:
>> diff --git a/arch/arm64/include/asm/barrier.h b/arch/arm64/include/asm/barrier.h
>> index 1ca947d5c939..25721275a5a2 100644
>> --- a/arch/arm64/include/asm/barrier.h
>> +++ b/arch/arm64/include/asm/barrier.h
>> @@ -216,6 +216,44 @@ do {									\
>>  	(typeof(*ptr))VAL;						\
>>  })
>>
>> +#define __smp_cond_load_relaxed_timewait(ptr, cond_expr,		\
>> +					 time_expr_ns, time_limit_ns)	\
>> +({									\
>> +	typeof(ptr) __PTR = (ptr);					\
>> +	__unqual_scalar_typeof(*ptr) VAL;				\
>> +	for (;;) {							\
>> +		VAL = READ_ONCE(*__PTR);				\
>> +		if (cond_expr)						\
>> +			break;						\
>> +		__cmpwait_relaxed(__PTR, VAL);				\
>> +		if ((time_expr_ns) >= (time_limit_ns))			\
>> +			break;						\
>> +	}								\
>> +	(typeof(*ptr))VAL;						\
>> +})
>
> Rename this to something like *_evstrm as this doesn't really work
> unless we have the event stream.

Ack.

> Another one would be *_wfet.

Hadn't sent out the WFET version yet.

Did you mean that this should be *_evtstrm or *_wfet?

>> +
>> +/*
>> + * For the unlikely case that the event-stream is unavailable,
>> + * ward off the possibility of waiting forever by falling back
>> + * to the generic spin-wait.
>> + */
>> +#define smp_cond_load_relaxed_timewait(ptr, cond_expr,			\
>> +				       time_expr_ns, time_limit_ns)	\
>> +({									\
>> +	__unqual_scalar_typeof(*ptr) _val;				\
>> +	int __wfe = arch_timer_evtstrm_available();			\
>
> This should be a bool.

Yeah. Will fix.

>> +									\
>> +	if (likely(__wfe))						\
>> +		_val = __smp_cond_load_relaxed_timewait(ptr, cond_expr,	\
>> +							time_expr_ns,	\
>> +							time_limit_ns);	\
>> +	else								\
>> +		_val = __smp_cond_load_relaxed_spinwait(ptr, cond_expr,	\
>> +							time_expr_ns,	\
>> +							time_limit_ns);	\
>> +	(typeof(*ptr))_val;						\
>> +})
>
> Not sure there's much to say here, this depends on the actual interface
> introduced by patch 1. If we make some statements about granularity of
> some time_cond_expr check, we'll have to take that into account.

Agreed.

Thanks for the review!

--
ankur