Re: brcmfmac disconnects when iwd scans

On 5/7/2025 3:54 PM, KeithG wrote:
> further information...
> 
> I tried this with 2 different pi4s and have been able to confirm
> wpa_supplicant/hostapd is stable on the Pi4 (will try others when I
> get a chance). I also tried both the 'shipping' brcmfmac43455 firmware
> and the 'latest'. It is only with the 'latest' that I get reliable
> connections. With the 'shipping' version, I can get connections for
> both wlan0 with wpa_suppolicant and ap0 with hostapd, but the ap0
> connection is flaky without any connection on wlan0. And I cannot
> sustain both connections at all. This 'latest' firmware is required
> for a sustainable, reliable ap0 with connection with hostapd.
> 
> this is the firmware required:
> [    3.658201] brcmfmac: brcmf_fw_alloc_request: using
> brcm/brcmfmac43455-sdio for chip BCM4345/6
> [    3.660643] usbcore: registered new interface driver brcmfmac
> [    4.041524] brcmfmac: brcmf_c_process_txcap_blob: no txcap_blob
> available (err=-2)
> [    4.042269] brcmfmac: brcmf_c_preinit_dcmds: Firmware: BCM4345/6
> wl0: Oct 28 2024 23:27:00 version 7.45.286 (be70ab3 CY) FWID
> 01-95efe7fa
> 
> The only annoyance I get is that the log fills up with failed scan messages:
> May 07 08:43:11 deskrune wpa_supplicant[604]: wlan0: Failed to
> initiate sched scan

The scheduled scan is basically a scan offload in which user-space tells 
the driver what to look for and how often (hence scheduled). The 
advantage of the scheduled scan is that the host (RPi) goes to sleep 
while the wifi chip is on the lookout for the ssid. No idea why it is 
failing though.

> the Pi still works and is scanning because when a known ssid appears,
> it immediately connects, so it is scanning, I just get this message.

What I recall from wpa_supplicant behavior is that it does a regular 
scan for a configured ssid and after a number of attempts it does try to 
setup a scheduled scan if the driver supports it. When it fails it 
simply continues with the regular scans. The brcmfmac driver supports 
scheduled scan so it seems like regression although apart from the log 
message the end-user is not really affected. I expect it will not make 
much noticable difference on the energy bill. ;-)

> Now that I have this working well with wpa_supplicant and know how to
> get it all 'up' and stable, I will start investigating what
> differences there are with iwd and why it is flaky. Please guide me as
> to the best way to do this.

Were you successful using nlmon. I suspect differences should show up 
between wpa_supplicant and iwd.

Regards,
Arend