[RFC v2 3/7] wifi: cfg80211: extend interface combination check for multi-radio

[Felix Fietkau]

On 06.06.24 11:52, Karthikeyan Periyasamy wrote:
> 
> 
> On 6/6/2024 2:28 PM, Felix Fietkau wrote:
>> On 06.06.24 10:56, Karthikeyan Periyasamy wrote:
>>>
>>>
>>> On 6/6/2024 1:25 PM, Felix Fietkau wrote:
>>>> On 06.06.24 09:20, Karthikeyan Periyasamy wrote:
>>>>>
>>>>>
>>>>> On 6/6/2024 12:01 AM, Felix Fietkau wrote:
>>>>>>       /*
>>>>>>        * This is a bit strange, since the iteration used to rely 
>>>>>> only on
>>>>>> @@ -2384,8 +2383,10 @@ int cfg80211_iter_combinations(struct wiphy 
>>>>>> *wiphy,
>>>>>>        * cfg80211 already - the only thing not would appear to be 
>>>>>> any new
>>>>>>        * interfaces (while being brought up) and channel/radar data.
>>>>>>        */
>>>>>> -    cfg80211_calculate_bi_data(wiphy, params->new_beacon_int,
>>>>>> -                   &beacon_int_gcd, &beacon_int_different);
>>>>>> +    if (!radio)
>>>>>> +            cfg80211_calculate_bi_data(wiphy, params->new_beacon_int,
>>>>>> +                           &beacon_int_gcd,
>>>>>> +                           &beacon_int_different);
>>>>>
>>>>> Why its avoid for radio specific iface combination ?
>>>>
>>>> Because it iterates over all wdevs within cfg80211. I didn't think 
>>>> this was necessary, given that it already excludes MLO wdevs.
>>>>
>>>
>>> Dont tie the radio specific iface advertisement with MLO.
>>>
>>> Usually the existing code consider "params->new_beacon_int" the MLO
>>> scenario also.
>> 
>> For your hardware, do beacon intervals need to be matched/aligned per 
>> radio or globally?
>> 
> 
> Our hardware supports radio aligned beacon interval.
> 
> Currently, ath12k use use same beacon interval configuration all radio
> iface combination.
> 
> Even in radio specific iface combination, we should check the beacon
> interval for the non MLO VAPs.
> 
> so dont avoid the beacon interval check.

Okay, I'll look into making this work.

- Felix