Re: Missing BEACON_LOSS event

[Alexander Wilhelm <alexander.wilhelm@westermo.com>]

Am Wed, Jul 30, 2025 at 09:52:16AM +0200 schrieb Nicolas Escande:
> On Tue Jul 29, 2025 at 3:03 PM CEST, Alexander Wilhelm wrote:
> > Hello devs,
> >
> > can someone help with the following issue?
> >
> > I'm using a QCN9074-based device in STA mode, connected to an access point. When
> > the AP is powered off (without sending a deauthentication frame), the client
> > remains indefinitely associated. There is no disconnect, no beacon loss event,
> > and the RSSI remains stuck at the last known value.
> [...]
> > From what I can tell, the function `ath11k_mac_handle_beacon_miss()` exists and
> > is wired up via `ath11k_roam_event()`, but the firmware never seems to send
> > `WMI_ROAM_EVENTID`, so the handler is never triggered.
> >
> > Is this expected behavior? Does the firmware need to be configured differently
> > to enable beacon miss detection in STA mode? Or is this a known limitation? Any
> > help or clarification would be appreciated.
> 
> Hello,
> 
> I've brought this up already on the ath11k/ath12k list:
> https://urldefense.com/v3/__https://lore.kernel.org/ath12k/CZA2NS7J83D4.18SU6W9R96KPY@gmail.com/__;!!I9LPvj3b!A_l1xrZCr5Oo1_ZCUkIiXoWPpFN0WLWL0jzZvD_tjkUbOJTC-SHA0CC0HrtcXfPeOOb3729mKsEecEcXslQ0X9e4o9WHOw$ 
> To my knowlege, nothing upstream has been posted so far by QCA.
> 
> In non mainline sources there is a patch to support ath12k:
> https://urldefense.com/v3/__https://git.codelinaro.org/clo/qsdk/oss/system/feeds/wlan-open/-/blob/win.wlan_host_opensource.3.0/patches/ath12k/897-wifi-ath12k-Add-support-for-STA-mode-to-trigger.patch__;!!I9LPvj3b!A_l1xrZCr5Oo1_ZCUkIiXoWPpFN0WLWL0jzZvD_tjkUbOJTC-SHA0CC0HrtcXfPeOOb3729mKsEecEcXslQ0X9dH40t43A$ 
> My guess is that ath11k works the same way.

Thank you for your response, Nicolas. It helps me a lot. Most of the patch code
is already part of `ath11k` driver. I will port the remaining part that
implements the `ath11k_peer_sta_kickout_event` for STA for me until another
firmware/driver solution is proposed.


Best regards
Alexander Wilhelm