wpa_supplicant bouncing between AP nodes

[Mark Pearson]

On Fri, Jul 10, 2026, at 5:27 PM, Dennis Bland wrote:
>> Hi all,
>>
>> I have a customer issue where they are running Ubuntu 24.04 and at their office site finding that their system is bouncing between AP nodes in the office mesh.
>>
>> I have full logs if useful, but it looks to me like wpa_supplicant is being perhaps over aggressive in switching between nodes when looking for better throughput.
>> e.g. in below switches from ba:ee to b7:ee to ba:ee to b7:ee in less than a minute.
>>
>> Jul 01 09:57:19 ccalifan-l-pw0la38t wpa_supplicant[3444710]: wlp9s0f0: Considering within-ESS reassociation
>> Jul 01 09:57:19 ccalifan-l-pw0la38t wpa_supplicant[3444710]: wlp9s0f0: Current BSS: a4:00:4e:05:ba:ee freq=5680 level=-66 snr=26 est_throughput=324302
>> Jul 01 09:57:19 ccalifan-l-pw0la38t wpa_supplicant[3444710]: wlp9s0f0: Selected BSS: a4:00:4e:05:b7:ee freq=5280 level=-65 snr=27 est_throughput=331502
>> Jul 01 09:57:19 ccalifan-l-pw0la38t wpa_supplicant[3444710]: wlp9s0f0: Using signal poll values for the current BSS: level=-66 snr=26 est_throughput=324302
>> Jul 01 09:57:19 ccalifan-l-pw0la38t wpa_supplicant[3444710]: wlp9s0f0: Allow reassociation - selected BSS has better estimated throughput
>>
>> Jul 01 09:57:53 ccalifan-l-pw0la38t wpa_supplicant[3444710]: wlp9s0f0: Considering within-ESS reassociation
>> Jul 01 09:57:53 ccalifan-l-pw0la38t wpa_supplicant[3444710]: wlp9s0f0: Current BSS: a4:00:4e:05:b7:ee freq=5280 level=-66 snr=26 est_throughput=324302
>> Jul 01 09:57:53 ccalifan-l-pw0la38t wpa_supplicant[3444710]: wlp9s0f0: Selected BSS: a4:00:4e:05:ba:ee freq=5680 level=-65 snr=27 est_throughput=331502
>> Jul 01 09:57:53 ccalifan-l-pw0la38t wpa_supplicant[3444710]: wlp9s0f0: Using signal poll values for the current BSS: level=-66 snr=26 est_throughput=324302
>> Jul 01 09:57:53 ccalifan-l-pw0la38t wpa_supplicant[3444710]: wlp9s0f0: Allow reassociation - selected BSS has better estimated throughput
>>
>> Jul 01 09:58:26 ccalifan-l-pw0la38t wpa_supplicant[3444710]: wlp9s0f0: Considering within-ESS reassociation
>> Jul 01 09:58:26 ccalifan-l-pw0la38t wpa_supplicant[3444710]: wlp9s0f0: Current BSS: a4:00:4e:05:ba:ee freq=5680 level=-66 snr=26 est_throughput=324302
>> Jul 01 09:58:26 ccalifan-l-pw0la38t wpa_supplicant[3444710]: wlp9s0f0: Selected BSS: a4:00:4e:05:b7:ee freq=5280 level=-62 snr=30 est_throughput=353102
>> Jul 01 09:58:26 ccalifan-l-pw0la38t wpa_supplicant[3444710]: wlp9s0f0: Using signal poll values for the current BSS: level=-67 snr=25 est_throughput=306252
>> Jul 01 09:58:26 ccalifan-l-pw0la38t wpa_supplicant[3444710]: wlp9s0f0: Allow reassociation - selected BSS has better estimated throughput
>>
>> Is this 'normal' and are there any recommendations on how to avoid this?
>>
>> Thanks
>> Mark
>>
>
> When wpa_supplicant estimates the throughput of a channel, it takes
> into account the channel width advertised by the AP in 802.11 beacon
> frames.
>
> However, if the Ubuntu 24.04 WLAN STA driver actually connects with a
> smaller channel width, then a ping-pong affect can occur where the
> actual throughput becomes lower, and then the original channel looks
> like a better candidate.  According to your wpa_supplicant logs, the
> calculated throughput went down after switching to the new channel,
> and the previous channel then had a higher estimated throughput.  The
> lower channel width chosen by the STA can be due to spectrum
> interference, or regulatory channel width limitations in the STA
> driver (5280 MHz and 5680 MHz are DFS channels in Canada).
>
Ah - interesting. Thank you for the detail

> Ubuntu 24.04 does not have a way to specify channel width in the GUI.
> The WLAN driver automatically selects the highest possible channel
> width supported by both the STA and AP.  I'm assuming you've already
> updated Ubuntu to the latest patch release and firmware files. If
> possible, try reducing the AP channel width to see if the problem
> still occurs.  Another option, if possible, is to change the APs to
> non-DFS channels, or other DFS channels that may have less RF
> interference due to overlap.
>
Not tried latest patch release - it's a bit tricky as it is on a customers system which he's using for work, and it only happens in the office :) But will definitely look to try that next.

Thanks for the other tips and all the detail - very much appreciated.

Mark