ath11k: unable to connect to 6ghz networks after cycling interface (IWD restart)
James Prestwood
jprestwood at locusrobotics.com
Wed Jul 22 07:17:52 PDT 2026
Hi,
Recently I've noticed an issue scanning 6GHz-only SSIDs. I am running:
Card: Qualcomm Atheros QCNFA765 (rev 01)
Driver: ath11k
Kernel: 6.8
Firmware: fw_version 0x1106196e fw_build_timestamp 2024-01-12 11:30
fw_build_id WLAN.HSP.1.1-03125-QCAHSPSWPL_V1_V2_SILICONZ_LITE-3.6510.37
Running IWD
I can't reproduce this 100% of the time, but it happens quite frequently
when I perform the following steps:
- Device is connected to a 5GHz SSID
- Switch the IWD profile to a 6GHz-only SSID
- Restart IWD
In some cases this works, but frequently IWD starts up and scans the
entire spectrum (2.4/5/6ghz) and sees no 6GHz SSIDs. IWD continues to
scan indefinitely and in some cases never sees a 6GHz SSID until its
restarted. Restarting IWD does remove and create a wireless interface in
the process. AFAICT the regulatory domain is set correctly and very
early on when IWD scans and the kernel detects a US-FCC SSID.
This leaves the device unconnected until our external watchdog kicks in
and restarts IWD which generally gets things working again. I've
captured the netlink comms below and nothing looks wrong or incorrect
about how IWD scans, the regdom, or enabled/disabled frequencies in the
wiphy dump:
The regulatory domain is set:
> Result: Get Reg (0x1f) len 508 Jul 22 13:32:08
Wiphy: 0 (0x00000000)
Regulatory Alpha2: US
DFS Region: 1 (0x01)
...
Rule 7: len 56
Flags: len 4
No Outdoor
Auto BW
Frequency Start: 5925 MHz
Frequency End: 7125 MHz
Frequency Range Max BW: 160 MHz
Max Antenna Gain: 0 (0x00000000)
Max EIRP: 2400 (0x00000960)
DFS CAC Time: 0 (0x00000000)
During the wiphy dump when IWD starts up I see the frequency of my SSID
and it is not disabled/no-IR:
> Result: New Wiphy (0x03) len 72 [multi] Jul 22 13:32:08
Wiphy: 0 (0x00000000)
Wiphy Name: phy0
Generation: 1 (0x00000001)
Wiphy Bands: len 40
Band 3: len 36
Frequencies: len 32
Frequency 40: len 28
* Frequency: 6755 (0x00001a63)*
Unknown: 20 len 4
00 00 00 00 ....
Indoor only: true
Max TX Power: 2400 (0x00000960)
IWD will then issues a trigger scan, and requests to scan on frequency
6755 (among many others):
< Request: Trigger Scan (0x21) len 848 [ack] Jul 22 13:32:08
Wireless Device: 9 (0x0000000000000009)
Information Elements: len 12
Extended Capabilities: len 10
Capability: bit 2: Extended channel switching
Capability: bit 19: BSS transition
Capability: bit 22: Multiple BSSID
Capability: bit 32: QoS Map
Capability: bit 62: Opmode Notification
04 00 48 00 01 00 00 40 00 21 ..H.... at .!
Scan Frequencies: len 816
<removed for readability>
* 69: 6755 (0x00001a63)*
<removed for readability>
The kernel responds and includes 6755 in the response:
> Event: Trigger Scan (0x21) len 844 Jul 22 13:32:08
Wiphy: 0 (0x00000000)
Interface Index: 15 (0x0000000f)
Wireless Device: 9 (0x0000000000000009)
Scan SSIDs: len 0
Scan Frequencies: len 792
<removed for readability>
* 69: 6755 (0x00001a63)*
<removed for readability>
IWD receives a new scan result event, which also includes 6755:
> Event: New Scan Results (0x22) len 844 Jul 22 13:32:14
Wiphy: 0 (0x00000000)
Interface Index: 15 (0x0000000f)
Wireless Device: 9 (0x0000000000000009)
Scan SSIDs: len 0
Scan Frequencies: len 792
<removed for readability>
* 69: 6755 (0x00001a63)*
<removed for readability>
But in the scan results there are no 6GHz BSS's found. Plenty of 2.4 and
5GHz though so the kernel is scanning. IWD also pulls a survey after the
scan, which includes 6755:
> Result: New Survey Results (0x33) len 52 [multi] Jul 22
13:32:14
Interface Index: 15 (0x0000000f)
Survey Info: len 40
* Frequency: 6755 (0x00001a63)*
Noise dBm: -96
Survey Time: 30 (0x000000000000001e)
Busy Time: 0 (0x0000000000000000)
I'm open to upgrading the kernel to debug, or trying patches. But I'm
curious if there are any reports of this, or patches that have gone in
since 6.8 to address scanning behavior similar to this. We have other
devices connected to this SSID during this process so the APs are
clearly sending beacons to maintain other connections. It feels to me
that the kernel is not actually scanning 6GHz despite the netlink
messaging saying it is.
Thanks,
James
More information about the ath11k
mailing list