[PATCH v4 1/3] wifi: nl80211: add link id of transmitted profile for MLO MBSSID

[Aloka Dixit]

On 3/22/2025 2:04 AM, Johannes Berg wrote:
> On Thu, 2025-03-13 at 15:59 -0700, Aloka Dixit wrote:
>> On 3/12/2025 1:55 AM, Johannes Berg wrote:
>>> Hi Aloka, all,
>>>
>>>> + * @NL80211_MBSSID_CONFIG_ATTR_TX_LINK_ID: Link ID of the transmitted profile.
>>>> + *	This parameter is mandatory if the transmitted profile is part of an MLD
>>>> + *	and the interface getting configured is a non-transmitted profile. For all
>>>> + *	other cases it will be ignored.
>>>
>>> So I guess it's a question of what "the interface getting configured"
>>> means, but I guess you could set up the transmitting interface?
>>>
>> Right, if the interface getting configured is the transmitted profile
>> the link_id is ignored.
> 
> But I guess I don't understand yet why it's not needed? If you want to
> configure MBSSID stuff for the transmitting profile it seems you'd need
> the right link anyway?
> 
> But if it's not needed, should we at least validate it's correct or so?
> 
>> The 'else' case in the highlighted snippet above is the same scenario as
>> the 'else' case you referred to, which is one layer out: "else if
>> (!config->index) {".
>>
>> Both are executed when the interface getting configured is the
>> transmitting interface. The difference between these two 'else's is that
>> in one case userspace explicitly provides
>> "NL80211_MBSSID_CONFIG_ATTR_TX_IFINDEX" and in other it hasn't.
> 
> Right.
> 
>> Similarly, for MLO, mac80211 does not need the link_id of the tx
>> interface because it already has the link_conf for it as part of basic
>> AP configuration parameters, hence link_id is ignored in both 'else's.
> 
> Oh so you're saying that you have an outer link_id already in
> nl80211_start_ap() (or other callers). So then I guess we're back to
> validating it, in line with how we say:
> 
>                  if ((!config->index && tx_ifindex != dev->ifindex) ||
>                      (config->index && tx_ifindex == dev->ifindex))
>                          return -EINVAL;
> 
> now?
> 

Sure, let me know if any more comments regarding this patch ant 2/2, I 
will send a new version together for all.

Thanks.