BPCC in per-STA profile in link reconfiguration response

BPCC in per-STA profile in link reconfiguration response

[Johannes Berg]

Hi,

Jouni pointed out to me that mac80211 currently requires the BSS
Parameter Change Count in the (Re)Association response multi-link
element per-STA profile(s), and via sharing the code, also in Link
Reconfiguration response, as you note in hostap commit ce8a121287bf ("AP
MLD: Include BPCC only in (Re)Associtiation Response frames").


Also, as you note there, the spec seems to suggest this should not be
the case, per section 9.4.2.322.2.4, where it is only in (re)association
response frames.


However, in the spec it seems to be explicitly required in section
35.3.6.4:

      If the AP MLD accepts link additions for one or more links, it
      shall include [...] For each Per-STA Profile subelement included
      in the BSS Multi-Link Element, [...] and the STA Profile field
      corresponding to that AP shall be complete and consists of all the
      elements and fields that would be included in the STA Profile
      field for that AP in a Reassociation Response frame that includes
      the corresponding AP as a reported AP [...].

We had some discussion about this internally and think that this is
inconsistent in the spec, and also that the 35.3.6.4 text (that suggests
including the field) makes more sense, so we'd suggest clarifying the
spec in section 9.4.2.322.2.4.


It seems to me that your commit was more meant to fix the ML Probe
response (no disagreement here with that, I believe) than the Link
Reconfiguration response, so two questions:

 1) Did you actually look into the various texts about it and decide
    that despite the seeming spec inconsistency, it should _not_ be
    present in Link Reconfiguration responses? As I said above, we
    think it makes more sense to have it present.

 2) Are you aware of any APs that don't include it, which would mean
    regardless of that, mac80211 should accept both?

Thanks,
johannes

Re: BPCC in per-STA profile in link reconfiguration response

[Mohan Kumar G]

On 7/15/2025 5:47 PM, Johannes Berg wrote:
> Hi,
>
> Jouni pointed out to me that mac80211 currently requires the BSS
> Parameter Change Count in the (Re)Association response multi-link
> element per-STA profile(s), and via sharing the code, also in Link
> Reconfiguration response, as you note in hostap commit ce8a121287bf ("AP
> MLD: Include BPCC only in (Re)Associtiation Response frames").
>
>
> Also, as you note there, the spec seems to suggest this should not be
> the case, per section 9.4.2.322.2.4, where it is only in (re)association
> response frames.
>
>
> However, in the spec it seems to be explicitly required in section
> 35.3.6.4:
>
>        If the AP MLD accepts link additions for one or more links, it
>        shall include [...] For each Per-STA Profile subelement included
>        in the BSS Multi-Link Element, [...] and the STA Profile field
>        corresponding to that AP shall be complete and consists of all the
>        elements and fields that would be included in the STA Profile
>        field for that AP in a Reassociation Response frame that includes
>        the corresponding AP as a reported AP [...].
>
> We had some discussion about this internally and think that this is
> inconsistent in the spec, and also that the 35.3.6.4 text (that suggests
> including the field) makes more sense, so we'd suggest clarifying the
> spec in section 9.4.2.322.2.4.
>
>
> It seems to me that your commit was more meant to fix the ML Probe
> response (no disagreement here with that, I believe) than the Link
> Reconfiguration response, so two questions:
>
>   1) Did you actually look into the various texts about it and decide
>      that despite the seeming spec inconsistency, it should _not_ be
>      present in Link Reconfiguration responses? As I said above, we
>      think it makes more sense to have it present.

Hey Johannes,
As you mentioned, this commit was initially intended to fix ML Probe 
Response,
which is why section 9.4.2.322.2.4 was referenced.
However, after reviewing section 35.3.6.4 — which says that the
Per-STA Profile in the ML Reconfiguration Response should be the
same as in the Reassociation Response — I agree that it makes sense
to include the BPCC in this case as well.

>
>   2) Are you aware of any APs that don't include it, which would mean
>      regardless of that, mac80211 should accept both?

As far as I know, I'm not aware of any APs in the market that don't 
include it.
Maybe people in community can comment on this. If such APs do exist, we 
can consider accepting both.

>
> Thanks,
> johannes
>

Re: BPCC in per-STA profile in link reconfiguration response

[Johannes Berg]

Hi,

> As you mentioned, this commit was initially intended to fix ML Probe 
> Response, which is why section 9.4.2.322.2.4 was referenced.

Right, makes sense.

> However, after reviewing section 35.3.6.4 — which says that the
> Per-STA Profile in the ML Reconfiguration Response should be the
> same as in the Reassociation Response — I agree that it makes sense
> to include the BPCC in this case as well.

OK, so I guess we should have a hostapd patch to remove the comment
there about not including it. Do you want to do that?

> > 
> >   2) Are you aware of any APs that don't include it, which would mean
> >      regardless of that, mac80211 should accept both?
> 
> As far as I know, I'm not aware of any APs in the market that don't 
> include it.

OK, cool, so at least we don't have any knowledge of this being
necessary.

> Maybe people in community can comment on this. If such APs do exist, we 
> can consider accepting both.

I don't know either, and the only reason for now would be to accept it
only because of the spec inconsistency, but I guess we can leave it for
now.

Thanks!

johannes