[PATCH 3/5] perf/arm-cmn: Refactor event filter data

[Robin Murphy]

On 2026-07-05 4:44 pm, Leo Yan wrote:
> On Tue, Jun 30, 2026 at 04:19:18PM +0100, Robin Murphy wrote:
> 
> [...]
> 
>> +struct arm_cmn_filter_attr {
>> +	enum cmn_filter_select sel;
>> +	u8 val;
>> +};
> 
>> -#define _CMN_EVENT_ATTR(_model, _name, _type, _eventid, _filter, _fsel)\
>> +#define _CMN_EVENT_ATTR(_model, _name, _type, _eventid, _filter, ...)	\
>>   	(&((struct arm_cmn_event_attr[]) {{				\
>>   		.attr = __ATTR(_name, 0444, arm_cmn_event_show, NULL),	\
>>   		.model = _model,					\
>>   		.type = _type,						\
>>   		.eventid = _eventid,					\
>> -		.filter = _filter,					\
>> -		.fsel = _fsel,						\
>> +		.filter = {{_filter}},					\
> 
> This patch might break as the "val" field is never initialized. Should
> it use two parameters "_fa, _fb" instead?

Ah, you're right - this evolved through several different designs, and 
this line is still from when "_filter..." was the whole varargs, but 
then I brought forward more of the changes from patch #5 to try to 
minimise intra-series churn, and I guess the fact that it didn't 
completely fail to compile (as I was accustomed to by that point...) 
meant I missed this subtlety.

> #define _CMN_EVENT_ATTR(_model, _name, _type, _eventid, _fa, _fb, ...) \
>        (&((struct arm_cmn_event_attr[]) {{                              \
>                ...                                                      \
>                .filter = {{_fa, _fb}},                                  \
> 
> Although the issue is fixed by a later patch, I think it is good to
> correct this patch for bisection.

Agreed, I'll fix that for a v2.

Thanks,
Robin.