Re: [ath11k] Roaming latency with Compex WLE300HX (Q CN9074) — firmware/driver issue/Latency

Aliliche Larbi <[email protected]>
Newsgroups org.infradead.lists.ath11k
Message-ID <CAPEqGQo=mQ23uP-+y+eMYraTpjx2iHxBpgkXpTC8P4kgfeWUTA@mail.gmail.com>
Hello,

I noticed an error in my previous message regarding roaming latency
measurements sent : Tue Dec 9 10:25:01 PST 2025. Please see the
corrected information below.

The radio card is : Compex WLE3000HX

larbi.

Le mar. 9 déc. 2025 à 19:25, Aliliche Larbi <[email protected]> a écrit :
>
> Hello,
>
> I am experiencing persistent roaming latency with a Compex WLE300HX
> (QCN9074-based) Wi-Fi 6 PCIe card using the ath11k driver.
>
> I have a roaming test platform with multiple APs and controllable
> attenuators.
>
> The roaming test is performed between two 11a access points:
> - AP1 on channel 36
> - AP2 on channel 149
>
> Using Wireshark monitors, I measured the time between:
> 1) the last data packet on channel 36, and
> 2) the first data packet on channel 149.
>
> I tested four devices:
>
> 1. ACKSYS Railbox (WLE300HX), firmware QCN9074:
>    https://git.codelinaro.org/clo/ath-firmware/ath11k-firmware.git
>
> 2. 8devices Mango kit, CPU IPQ6018 + WLE300HX (QCN9074 firmware)
>
> 3. ACKSYS Airwave, CPU IPQ6018 + QCA firmware:
>    https://github.com/andreybondar59/firmware_qca-wireless
>
> 4. Moxa AWK-1161C, CPU IPQ6010, firmware unknown
>
> Roaming time (last data packet on ch36 → first data packet on ch149):
> - Moxa: ~104 ms
> - Railbox: ~180 ms
> - AirWave: ~104 ms
> - Mango: ~140 ms
>
> Wireshark measurement (last data packet ch36 → authentication response ch149):
> - Moxa: ~44 ms
> - Railbox: ~101 ms
> - AirWave: ~51 ms
> - Mango: ~100 ms
>
> I traced the Railbox using ftrace and extracted several WMI commands and
> their response times, from the last data packet on channel 36 until
> authentication success on channel 149:
>
> subif_start_xmit -> wmi_install_key_cmdid: 19.9 ms
> wmi_install_key_cmdid -> wmi_install_key_cmdid_response: 1.2 ms
> wmi_install_key_cmdid_response -> wmi_perr_delete_cmdid: 0.26 ms
> wmi_perr_delete_cmdid -> wmi_perr_delete_cmdid_response: 5 ms
> wmi_perr_delete_cmdid_response -> ath11k_mac_vdev_stop(): 0.5 ms
> ath11k_mac_vdev_stop() -> wmi_vdev_stop_cmdid: 19.3 ms
> wmi_vdev_stop_cmdid -> wmi_vdev_stop_cmdid_response: 0.27 ms
> wmi_vdev_stop_cmdid_response -> ath11k_wmi_vdev_start(): 0.03 ms
> ath11k_wmi_vdev_start() -> wmi_vdev_start_cmdid: 6.3 ms
> wmi_vdev_start_cmdid -> wmi_vdev_start_cmdid_response: 38.7 ms
> wmi_vdev_start_cmdid_response -> ieee80211_mgd_auth: 18.7 ms
> ieee80211_mgd_auth -> wmi_send_auth_req_cmdid: 5 ms
> wmi_send_auth_req_cmdid -> wmi_send_auth_req_cmdid_completed: 2.7 ms
> wmi_send_auth_req_cmdid_completed -> wmi_rx_auth_response: 1.39 ms
> wmi_rx_auth_response -> ieee80211_mark_sta_authenticated(): 3.57 ms
>
> Total: ~123 ms between the last packet on ch36 and authentication
> success on ch149
>
> Note: These timings are specific to the Railbox (QCN9074).
> The Airwave (IPQ6018 + QCA firmware) behaves differently; for instance,
> the vdev start command responds in ~10 ms on the Airwave firmware.
>
> Given this, it seems the QCN9074 firmware responds significantly slower.
>
> What are the differences between the WLE300HX QCN9074 firmware and the
> QCA firmware used on the Airwave device? Is there anything that can be
> tuned or configured to reduce roaming latency?
>
>
> Thank you,
> Larbi
> ------------------------------------------------------wireshark
> measures in ms--------------------------------------------------
> Event                         | MOXA         | RailboxV2 + WLE3000HX
>   |AirWave (New FW)     | Mango + WLE3000HX
> --------------------------+-------------------+--------------------------------------+-----------------------------+-------------------
> LAST DATA PACKET    | T0                | T0
>                 | T0                       | T0
>
> Disassociate                  | T0+{16}       | XX
>                       | Deauth XX          | XX
> Auth request                  | T0+{42}       | T0+{94;97}
>                   | T0+{48}               | T0+{96}
> Auth response               | T0+{44}       | T0+{98;101}
>                | T0+{51}               | T0+{99}
> Assoc request                | T0+{46}      | T0+{113;116}
>               | T0+{57}               | T0+{104}
> Assoc response             | T0+{51}      | T0+{115;118}
>             | T0+{58}               | T0+{106}
> EAPOL 1/4                    | T0+{61}       | T0+{123;126}
>                | T0+{67}               | T0+{113}
> EAPOL 2/4                    | T0+{64}      | T0+{142;144}
>                | T0+{73}               | T0+{124}
> EAPOL 3/4                    | T0+{67}      | T0+{145;147}
>                | T0+{76}               | T0+{125}
> EAPOL 4/4                    | T0+{68}      | T0+{148;151}
>                | T0+{78}               | T0+{129}
>
> LAST DATA PACKET    | T0+{104}    | T0+{177;197}
>      | T0+{104}             | T0+{140}
lmpx.com only provides a reader for public news (NNTP) servers. It is not affiliated with the servers or forums shown here and is not responsible for the content of articles, which is written by their respective authors.