[PATCH v10 3/5] cfg80211/mac80211: move interface counting for combination check to mac80211

[Wen Gong]

On 2/14/2023 10:41 PM, Johannes Berg wrote:
> On Tue, 2023-01-10 at 15:23 +0800, Wen Gong wrote:
>> On 1/9/2023 6:05 PM, Johannes Berg wrote:
>>> On Mon, 2023-01-09 at 17:39 +0800, Wen Gong wrote:
>>>> On 3/11/2014 10:16 PM, Luciano Coelho wrote:
>>>>> ...
>>>>> +int ieee80211_check_combinations(struct ieee80211_sub_if_data *sdata,
>>>>> +				 const struct cfg80211_chan_def *chandef,
>>>>> +				 enum ieee80211_chanctx_mode chanmode,
>>>>> +				 u8 radar_detect)
>> ...
>>>>> +
>>>>> +	if (total == 1 && !radar_detect)
>>>>> +		return 0;
>>>>> +
>>>> should also check with cfg80211_check_combinations() when total == 1 and
>>>> num_different_channels > 1 ?
>>>>
>>>> When MLO is enabled, it could have 2 channels for one ieee80211_sub_if_data.
>>>>
>>> Heh. You're commenting on a patch from 2014, well before MLO :-)
>>>
>>> Not sure what happens in the code now?
>>>
>>> johannes
>> Yes, it is 2014. Each interface only has one channel at 2014.
>> I did not hit issue for the code.
>>
>> the story is like this:
>> When station interface and p2p device interface both running.
>> the station connect to MLO AP which has 2 links.
>> The ieee80211_link_use_channel()/ieee80211_check_combinations() call
>> cfg80211_check_combinations()
>> for the channel1 and channel2 because total==2.
>> When only station interface is running, not called for channel1/channel2
>> because total==1.
>> That is the little thing I hit.
>>
> So ... I guess you found a bug? Not sure what I'm supposed to say here.
> Send a fix?
I consider it NOT a bug. It is only the check is not strict here.
>
> johannes