Receive OMI support in mac80211?

Receive OMI support in mac80211?

[Johannes Berg]

Hey all,

We're contemplating adding some support for using (sending) OMI to save
power or perhaps other reasons.

The implementation logic would seem to end up being somewhat similar to
some other cases of changing parameters (CSA or simply AP doing
HT/VHT/HE/EHT operation changes, VHT operating mode notification action
frame), but 

 - for widening bandwidth/increasing NSS:
    - update chanctx
    - send OMI / wait / response
    - update TX configuration for the STA it was sent to

 - for reducing bandwidth/NSS:
    - update TX configuration for the STA it was sent to
    - send OMI / wait / response
    - update chanctx

FWIW, one thing I'm still unsure about is how OMI interacts with VHT
operating mode notification, but we don't transmit the latter right now.


Initially I thought it might be best to have some kind of request_omi()
function the driver can call, but since sending the OM field is done
with any data frame, I'm not sure how to really get that done - a new
driver op seems a bit overkill?


So perhaps it'd be better to have at least two new functions that always
have to be called in pairs, like prepare_omi() and finalize_omi() and
it'd be up to the driver to make sure the field was sent inbetween?

I'm tending towards the latter, any thoughts?

johannes

Re: Receive OMI support in mac80211?

[Johannes Berg]

On Fri, 2024-06-21 at 14:31 +0200, Johannes Berg wrote:
> Hey all,
> 
> We're contemplating adding some support for using (sending) OMI to save
> power or perhaps other reasons.
> 
> The implementation logic would seem to end up being somewhat similar to
> some other cases of changing parameters (CSA or simply AP doing
> HT/VHT/HE/EHT operation changes, VHT operating mode notification action
> frame), but 
> 
>  - for widening bandwidth/increasing NSS:
>     - update chanctx
>     - send OMI / wait / response
>     - update TX configuration for the STA it was sent to
> 
>  - for reducing bandwidth/NSS:
>     - update TX configuration for the STA it was sent to
>     - send OMI / wait / response
>     - update chanctx

Hmm. Maybe I sent this too early ;)

Maybe mac80211 shouldn't even do the TX configuration for the STA? It
would have to track this as a separate data point because it's due to
the *local* configuration, so it doesn't actually match the other things
that it's doing today ...

OTOH, if mac80211 does rate control? But maybe we don't really care
about that case? I don't really for iwlwifi ...

And then if we don't care about that part, all we need is to be able to
notify mac80211 about the BW change on that given connection.

johannes