Re: [PATCH wireless-next v4 03/21] wifi: nl80211: add SMD BSS Transition sub-state STA flags

From: Johannes Berg

Date: Mon Nov 16 2026 - 05:50:47 EST


On Mon, 2026-11-16 at 15:51 +0530, Pooventhiran G wrote:
>
> > I'm also not sure why you need to distinguish between the latter two at
> > all, hostapd is going to send the ST execution response and immediately
> > transition to DL drain (4c) or back to 4 anyway?
> >
>
> The 4b/4c differentiation is needed to resolve frame filtering (especially UL
> data) since the requirements differ between these two states. There will be
> some time here between the reception of ST Exec Request (set 4b, communicate
> with the target AP MLD, and get the confirmation back) and sending the ST Exec
> Response to the non-AP STA MLD.

Well, technically yes, but there shouldn't _be_ any UL data (from a
compliant non-AP MLD). If that's the only reason, do we really want to
make the implementation more complex for effectively non-compliant
stations?

I understand that you need to get the ST context at some time between
those points, or already have it attached to the ST execution request
frame, and then receiving and accepting UL data from a non-compliant
peer will ... break things for that peer after it transitions. Actually
it's broken anyway, regardless of the AP dropping the frames, since the
peer continued to transmit. A (buggy or malicious) peer can already
duplicate the frame on the DS by just sending it multiple times. In your
(planned) implementation, the AP can't even protect against it fully
because the ST context snapshot is at the RX time, while moving the
state takes a hostapd reaction, unless you say that the FW also has to
move the state when it takes the ST context snapshot. In the case you
just discussed with Allen, you could move the state before taking the
context.

So ... that's a fairly long-winded way of saying "is it really worth
it"? It doesn't seem to buy either side anything if the peer isn't
compliant, and has no effect if it is compliant?

johannes