Re: [PATCH wireless-next v2 00/16] wifi: Add Seamless Mobility Domain (SMD) AP support

[Pooventhiran G <pooventhiran.g@oss.qualcomm.com>]

On 10/04/2026 04:19 pm, Johannes Berg wrote:
> On Thu, 2026-09-24 at 08:10 +0530, Pooventhiran G wrote:
>> IEEE P802.11bn introduces the Seamless Mobility Domain (SMD) - a mobility
>> domain uniquely identified by the SMD identifier MAC address - where
>> a non-AP MLD that is associated to the SMD Management Entity (SMD-ME) can
>> transition from its current AP MLD to a non-colocated target AP MLD
>> without requiring reassociation via SMD BSS Transition (ST) procedure.
>>
>> This series implements the nl80211, cfg80211, and mac80211 support for
>> SMD BSS Transition in AP mode. Along with a summary of the ST procedure,
>> the design of this series and patch structure are presented below.
> 
> OK I'm wondering ... more things:
> 
>    1. How is this supposed to update the DS correctly? mac80211 normally
>       does this via cfg80211_send_layer2_update() and I don't think
>       you're changing that, but then it's the wrong place?
>       
>    2. In the kernel here the state is the per-AP MLD state of the STA,
>       not the SMD-ME state, so 4a isn't _really_ "AUTHORIZED", is it?
>       Ties in with 1. though, I guess.
>       

(combining #1 and #2)

Yes, Johannes. 4a is not part of state-4 per se in this implementation.
Patch 6/16
[https://lore.kernel.org/all/20260924-smd-v2-6-bb40094da1d4@oss.qualcomm.com/]
sanitizes only substate-4b and 4c transitions in state-4.

Because, we are not introducing a new entity to track SMD-ME states but rely
on MLD states (subclause 11.3.1, Table 11-6a). During Preparation, on the
target AP MLD, STA will move till just state-3 and then to substate-4a.

On the current AP MLD:
STA moves to: 1 -> 2-> 3 -> 4 -> 4b -> 4c (optionally) -> 1

On the target AP MLD:
STA moves to: 1 -> 2 -> 3 -> 4a -> 4

When STA moves to state-4, L2 update is sent via the existing state transition
handler.
>    3. I commented a few minutes ago about handling of 4a in rx.c, but
>       now I'm beginning to think we also need to handle 4b/4c
>       explicitly.
>       

Yes, I will add frame limitations for all 3 SMD states.

>    4. Adding a station in 4a should have _something_ about powersave, at
>       the very least some documentation that it must be done by drivers
>       (and then perhaps also hwsim changes)?
> 

I will add documentation that drivers should handle this powersave since this
is specific to SMD.
> 
> johannes