802.11k/v based roaming not making the transition to another AP (IDFGH-12973)

evoon commented 3 months ago

Answers checklist.

[X] I have read the documentation ESP-IDF Programming Guide and the issue is not addressed there.
[X] I have updated my IDF branch (master or release) to the latest version and checked that the issue is present there.
[X] I have searched the issue tracker for a similar issue and not found a similar issue.

IDF version.

v5.3

Espressif SoC revision.

ESP32-C6 (v0.0)

Operating System used.

Windows

How did you build your project?

VS Code IDE

If you are using Windows, please specify command line type.

None

Development Kit.

ESP32-C6-WROOM-1

Power Supply used.

USB

What is the expected behavior?

Expected behavior is seamless roaming from AP1 to AP2 via a neighbor report followed by a BSS transition. Network is a 802.11ax with WPA2-PSK and TP-Link Omada hardware (1 router and 1 AP) that has toggled on the roaming features. APs are not on the same channel. In the TP-Link controller, I have tried to disable Fast Roaming/AI Roaming and I tried with Force Disassociate feature.

What is the actual behavior?

The current behavior is this, with the roaming_11kvr code :

I set a RSSI threshold to -40 dBm -I move the STA away from the AP to which it is connected (bssid=f0: ...), while moving closer to another AP (bssid=3c: ...) -RSSI threshold is reached at some point, a RRM neighbor report is retrieved properly from the current AP : I (66045) wifi_manager: WIFI_EVENT_STA_BSS_RSSI_LOW I (66045) wifi_manager: wifi_manager_event_handler:bss rssi is=-42 I (66055) wifi_manager: RMM neighbor report bssid=3c:52:a1:fc:71:7e info=0x7 op_class=0 chan=1 phy_type=0

Then, a BSS transition query is sent by the ESP32 with :

esp_wnm_send_bss_transition_mgmt_query(REASON_RSSI, neighbor_list, cand_list);

And no other BSS packets are observed after with a wifi adapter in monitor mode (we expect a request from the AP and then optionally a response from STA to AP) : Here is the BSS Transition Query packet : Screenshot BSS Query Packet

Shouldn't the BSS Transition QUery Reason be "Better AP Found" instead of "Load Balancing" (for REASON_RSSI) or "Excessive frame loss/poor conditions" (for REASON_FRAME_LOSS) ? It would make more sense but I don't know if it changes the behavior then. I didn't get a request frame back when using REASON_RSSI or REASON_FRAME_LOSS.

The same scenario repeats itself after, because i set the rssi threshold after each low rssi event (and upon first connexion of course).

As roaming_11kvr is an official roaming example in v5.3 i presume it works with other network setups, but I don't see what would make this fail with mine. Thanks for the help.

Steps to reproduce.

^

Debug Logs.

No response

More Information.

No response

evoon commented 3 months ago

Isn't it possible to initiate a FT request to the connected AP, following this figure from IEEE 802.11-2020 page 2622? The idea "over-the-DS" would be to talk to the neighbor from the neighbor report to initiate the FT, I don't know if that is possible. What did you implement when adding the "ft_enabled" boolean in a STA's config ?

Thank you for your help!

kapilkedawat commented 3 months ago

Hi @evoon, unfortunately how AP behaves is not in our control. Ideally we should get an BTM request with the candidate AP info and station will move to the candidate.

ESP32 only support over the air which is supported by majorly of APs(very less APs support over the DS so we don't plan to add that overhead).

Since your AP network is not behaving in a standard way, I would suggest you to give a try to our new roaming app, which takes care of these situations. It supports legacy roaming along with BTM and has lots of parameter which you can tune as per your convenience. You can read more about the app here https://github.com/espressif/esp-idf/blob/master/components/esp_wifi/wifi_apps/roaming_app/src/README.md .

app is available here https://github.com/espressif/esp-idf/blob/master/components/esp_wifi/wifi_apps/roaming_app/src/README.md

evoon commented 3 months ago

@kapilkedawat Indeed the legacy method in roaming_app works well, but defeats the purpose of fast roaming.

Can you recommend a router/AP that you had good results with in terms of fast/seamless roaming ? My objective is to get the lowest offline time during the roaming (<150ms).

kapilkedawat commented 2 months ago

Hi @evoon, sorry for the delay,. somehow I missed your reply. You can use tplink deco series which supports 11kvr and generally give response to btm query.

However most new implementations of mesh actually uses their own algo internally and that is the reason they don't respond to btm query and tried to move the station as they seems fit.

evoon commented 2 months ago

Thanks for your help.