Re: [PATCH rtw-next v13 2/3] wifi: rtw88: support channel switch in AP mode
From: Mehmet Fide
Date: Fri Nov 20 2026 - 12:03:31 EST
Hi Luka,
> I think this can finish before the last count gets a TBTT if the work
> runs a full interval late.
Right, at this commit. With a test build that delays the step at count
2 by 150 ms, count 1 never went on the air with 2/3 alone (0 of 5 runs)
and always did with 3/3 applied (5 of 5). I kept the TBTT pacing in its
own patch to keep 2/3 easier to review. If Ping-Ke prefers, v14 can
carry it in 2/3.
> What happens if the page can never fit? On RTL8821C the driver has
> eight 128 byte pages.
The switch would stay pending, but hostapd refuses such a request
first. On an RTL8821CU at 2.4 GHz a beacon still fit with 778 bytes of
vendor elements, while hostapd could not build the channel switch
request above 586 bytes ("nl80211: Could not build channel switch
request"). That message carries the beacon and the probe response
twice, before and after the switch, in the default 4 KB of
nlmsg_alloc(). No chip has less than eight 128 byte pages, and an AP
next to a station interface is only possible on the RTL8822C, which has
16. So I would rather not add a reject path that cannot be reached.
> Could we retry the first beacon download after the switch too?
[ ... ]
> Could we queue a retry for the station if this refresh fails?
Neither download is retried, like every other beacon update in rtw88
today. In about 2360 kernel logs from our test runs, a reserved page
download failed only in unbind and firmware crash tests, and with one
experimental change that is not in this series. I would keep both
paths as they are.
Mehmet