Re: [PATCH wireless-next v4 08/21] wifi: nl80211/cfg80211: pack SMD dynamic context along with frame
From: Allen Ye (葉芷勳)
Date: Sun Nov 15 2026 - 22:38:49 EST
On Fri, 2026-11-13 at 23:49 +0530, Pooventhiran G wrote:
>
> External email : Please do not click links or open attachments until
> you have verified the sender or the content.
>
>
> On 11/11/2026 02:44 pm, Allen Ye wrote:
> > On Fri, 2026-11-06 at 13:28 +0530, Pooventhiran G wrote:
> >
> > Hi Pooventhiran,
> >
> > A question on how the context is collected here.
> >
> > > While packing the received management frame in
> > > nl80211_send_mgmt() to
> > > send to userspace, nest the SMD dynamic context along with the
> > > frame
> > > message.
> >
> > On MediaTek chipsets (mt76), the dynamic context is kept in FW/HW,
> > so
> > the driver has to fetch it from the firmware with an MCU command.
> > This
> > means the driver has to hold the ST request frame until the
> > firmware
> > replies with an event, and only then pass the frame up. Holding RX
> > frames in the driver is the problematic part: the frame has to be
> > kept
> > outside the normal RX path, kept in order with the frames that
> > follow
> > it, and handled if the firmware reply is delayed or never arrives.
> >
> > 1. With the context attached to the frame, how it gets collected
> > depends on each vendor driver's own implementation (on mt76, this
> > may
> > mean holding the frame in the driver as described above). If
> > userspace
> > instead queried it via GET_SMD_CTX / SMD_CTX_EVENT for ST
> > preparation
> > and ST execution via the current AP MLD, as it already does for ST
> > execution via the target AP MLD, the flow would be the same for
> > every
> > driver, regardless of where the context is kept. Would that be
> > acceptable?
> >
> > 2. If not, could the frame-attached context be optional (e.g.
> > behind a
> > capability flag), with userspace falling back to GET_SMD_CTX?
> >
>
> Hi Allen. Sure, that option could also be added if some drivers
> choose not to
> pack in the rx_mgmt frame itself. Since it will be a fallback, I can
> change
> hostapd to directly query the driver for context using GET_SMD_CTX if
> it does
> not find the context attached to the received frame?
>
Hi Pooventhiran,
That sounds good to put the logic in hostapd.
Thanks,
Allen
> > Thanks,
> > Allen
>