serious regression: mt7916 firmware: 5ghz AP not working

[lts-rad]

• 2.4ghz interface (wlan0) okay
• 5ghz interface (wlan1) does not appear to be broadcasting beacons to nearby devices

firmware d6611d0, broken

checked out 65bbd4c, confirmed working

[maurerle]

Which device is here affected?

Can you trace the bug down to one of the commits in the mentioned code delta, by bisecting further?

65bbd4c...d6611d0

[lts-rad]

To clarify -- the bug happens with the new firmware (not kernel code)

[stratixx]

Hi,

I have some observations from my side.
I have Netgear EAX12 with MT7915E working as WiFi Access Point, connected to router over wire. Latest release Openwrt 23.05.4 (r24012-d8dd03c46f) (kernel 5.15.162) (2024-07-17 11:16:49) has working 2.4GHz and not starting 5GHz wifi. Snapshot version SNAPSHOT (r27426-232cc239b8) (2024-09-16 12:27:11) (kernel 6.6.51) has working both wifi's.
I don't see older images to test for my device on firmware-selector.openwrt.org page.

By not working wifi in my case I mean that SSID is not broadcasted, no wifi access. But radio is enabled, it is able to perform scan via luci network/wireless scan option.

Kernel logs from 23.05.4 while restarting radio looks like that:
[79324.336625] device phy1-ap0 left promiscuous mode
[79324.341560] br-lan: port 3(phy1-ap0) entered disabled state
[79325.801774] br-lan: port 3(phy1-ap0) entered blocking state
[79325.807601] br-lan: port 3(phy1-ap0) entered disabled state
[79325.813937] device phy1-ap0 entered promiscuous mode
[79325.844345] device phy1-ap0 left promiscuous mode
[79325.849625] br-lan: port 3(phy1-ap0) entered disabled state
[79325.921278] br-lan: port 3(phy1-ap0) entered blocking state
[79325.927061] br-lan: port 3(phy1-ap0) entered disabled state
[79325.933162] device phy1-ap0 entered promiscuous mode

Kernel logs from snapshot:
[ 1904.404544] mt7915e 0000:02:00.0 phy1-ap0: left allmulticast mode
[ 1904.410850] mt7915e 0000:02:00.0 phy1-ap0: left promiscuous mode
[ 1904.417124] br-lan: port 3(phy1-ap0) entered disabled state
[ 1905.036022] br-lan: port 3(phy1-ap0) entered blocking state
[ 1905.041777] br-lan: port 3(phy1-ap0) entered disabled state
[ 1905.047487] mt7915e 0000:02:00.0 phy1-ap0: entered allmulticast mode
[ 1905.054692] mt7915e 0000:02:00.0 phy1-ap0: entered promiscuous mode
[ 1905.061889] br-lan: port 3(phy1-ap0) entered blocking state
[ 1905.067583] br-lan: port 3(phy1-ap0) entered forwarding state
[ 1906.122242] br-lan: left allmulticast mode
[ 1906.308905] br-lan: entered allmulticast mode
[ 2615.024473] mt7915e 0000:02:00.0 phy1-ap0: left allmulticast mode
[ 2615.030747] mt7915e 0000:02:00.0 phy1-ap0: left promiscuous mode
[ 2615.036968] br-lan: port 3(phy1-ap0) entered disabled state
[ 2616.877933] br-lan: port 3(phy1-ap0) entered blocking state
[ 2616.883710] br-lan: port 3(phy1-ap0) entered disabled state
[ 2616.889584] mt7915e 0000:02:00.0 phy1-ap0: entered allmulticast mode
[ 2616.896500] mt7915e 0000:02:00.0 phy1-ap0: entered promiscuous mode
[ 2616.903550] br-lan: port 3(phy1-ap0) entered blocking state
[ 2616.909272] br-lan: port 3(phy1-ap0) entered forwarding state
[ 2616.916154] br-lan: port 3(phy1-ap0) entered disabled state
[ 2617.363271] br-lan: port 3(phy1-ap0) entered blocking state
[ 2617.369028] br-lan: port 3(phy1-ap0) entered forwarding state

There are some differences in logs. Not sure how and why.

[lts-rad]

@stratixx so is Firmware Version 20240823 working for you then ? you should see it in dmesg when the driver loads. thats great news if the firmware works on the newer kernel. in my case the new firmware failed with kernel 6.8.12 (noble).

[stratixx]

I see firmware version 6.6.51.2024.09.05~65cc3daf-r1 in snapshot image. It seems working. Later I can reflash and check what firmware version is used in latest release. I'm rather not going to compile custom firmware versions.

[stratixx]

version 23.05.4 has mt7915 firmware version 5.15.162+2024-04-03-1e336a85-1.

weird thing. After flashing back latest release image, which was already tested previously and was failing, now 5GHz wifi is working.

[lts-rad]

maybe unrelated? the firmware that was committed in d6611d0 is even more recent, 20240823 (vs 04 03)

[cristian-ciobanu]

I can confirm the most recent firmware for MT7916 (version 20240823) is broken and 5GHz interface does not send any beacons. I'm using a very recent snapshot for OpenWRT
Once I have downgrade the firmware files to an older version 20230202 it started to work fine.

[leow149]

I can also confirm those issues for the MT7916 (AsiaRF AW7916-NPD on Banana Pi R4). Firmware version 20240823 breaks 5ghz. I downgraded to version 20230202 and now its working as expected.

[ZerionSeven]

On the latest commit (6a11994), 5Ghz works now on the new firmware for me.

[Headcrabed]

On the latest commit (6a11994), 5Ghz works now on the new firmware for me.

@ZerionSeven Hello, which bandwidth and channel configuration are you using? Have you tried 160mhz?

[ZerionSeven]

On the latest commit (6a11994), 5Ghz works now on the new firmware for me.

@ZerionSeven Hello, which bandwidth and channel configuration are you using? Have you tried 160mhz?

I use AC mode on channel 100 at 160 MHz. I get about 1.2 Gbps of bandwith on 2x2.

AX mode doesn't work reliably with my phone, but other devices work fine, so It's probably a problem with the phone.