Basic hostapd testing with BE200 and MT7925 m.2 cards

[fhteagle]

Quick and dirty testing of a be200 and mt7925, each in m.2 card format. In each case, I took a working hostapd.conf file optimized for the mt7921 in 2GHz, ax, WPA3 SAE only security and changed the bare minimum to get it to not error out at all. BE200 needed [GF] removed from the HT_Capab definitions, for example. Resulting hostapd.conf here: https://pastebin.com/unpqG6uk

No attempt was made to add any 802.11be or MLO features to the .conf . Host is x86_64 running Arch Linux 6.12.6-hardened kernel and hostapd compiled from git within the last week or so. All measurements of throughput used iperf3, and STA and AP were directly connected over wifi, no intermediate hops or routers.

BE200:

• Worked in 2GHz, ax mode, wpa3 security. Throughput to a Wifi5 android phone on the same lap desk was pretty horrible, ~30Mbps AP->STA, ~50 Mbps STA->AP . Forgot to check if it accepted the 40MHz wide channel or not...
• Absolutely refused to make an AP on the 5GHz band, despite a router with correct 802.11d/h country codes being on air in this band, due to Intel's absolutely laughable Location (Un)Aware Registry trainwreck of an anti-feature.
• All band 4 (6GHz) frequencies are marked no-IR due to regulation.

MT7925:

• Worked in 2GHz, ax mode, wpa3 security, 40MHz channel IIRC. Throughput to same phone in same location was about 90Mbps each way
• Worked in 5GHz, ax mode, wpa3 security. Throughput to same phone in same location was about 140Mbps up and similar down. Forgot to note the channel width in use, pretty sure it was 40MHz though.
• All band 4 (6GHz) frequencies are marked no-IR due to regulation.

If anybody cares I could probably be more thorough, especially around the ax throughput testing. But I think that is good enough to say that AP is possible, and more tuning is a future project.

Also, weird quirk I discovered: if the adapter is assigned to a bridge, neither adapter sent packets correctly. As soon as I commented out the bridge=br0 line in the hostapd.conf and restarted hostapd, pings and iperfs started work as expected. I did not spend much time troubleshooting if creating the bridge before hostapd was brought up, or having a second interface assigned to the bridge would make any difference.

[morrownr]

hI @fhteagle

No attempt was made to add any 802.11be or MLO features to the .conf

I more or less keep up with traffic on linux-wireless. My take on things is that there is still a lot of WiFi 7 work being done on the wireless stack API's. A lot of it is MLO related. I'd say that we are looking at another year before WiFi 7 is fully capable and stabilized.

But I think that is good enough to say that AP is possible, and more tuning is a future project.

I need to do a little AP mode testing with my M.2 mt7925 card soon. Thanks for the report.