Re: [PATCH v3] wifi: cfg80211: check radio iface combination for multi radio per wiphy

[Felix Fietkau <nbd@nbd.name>]

On 11.09.24 17:25, Johannes Berg wrote:
> On Wed, 2024-09-11 at 20:51 +0530, Karthikeyan Periyasamy wrote:
>> 
>> ieee80211_link_reserve_chanctx() calls 
>> ieee80211_can_create_new_chanctx() with radio_idx (-1) to calculate the 
>> max channel (ieee80211_max_num_channels) after the iface combination 
>> check (cfg80211_iter_combinations) passed for the global iface 
>> combination. Here the expectation is number of channel context is less 
>> than the number of different channel. So in multi-radio advertisement, 
>> each radio support atleast one channel, so totally multiple different 
>> channels advertised in the global iface combination to pass this 
>> ieee80211_max_num_channels().
> 
> So maybe that's broken then, I dunno. You should figure it out with
> Felix I guess.
> 
> The intent was, and clearly it has to be, that the global combinations
> are something that can be handled regardless of radio information, to be
> backward compatible with existing uses. Therefore, it cannot be
> something where you say two channels and radar detection on both because
> that would imply being able to use channels 36 and 40 with 20 MHz at the
> same time with radar detection, which isn't actually possible.
> In this case, the "two channels" is only possible with also two radios,
> which has to rely on the per-radio advertisement, and the global one has
> to be just one channel for radar detection.
> 
> We still need the part of the patch that calls the validation on each
> radio, but it shouldn't be different from the global one. If you could
> make that patch I'd appreciate it.

It's implemented the way we discussed it back when this topic came up. 
Legacy user space gets the ifcomb data from the first radio for 
compatibility. The actual global ifcomb data is reported in a new attribute.

- Felix