[Solved] 802.11ax worse than 802.11ac with mt76 driver?

Does that remove vht160mhz support also works on 7986?

This patch also works for the Xiaomi ax3600, tested with an iPhone 13 Pro and IPad Air 5 :v:

@jkdf2 could you prepare a pull request on Github? Would be good to move some part of that discussion there.

Hi, I'm unclear on what the PR would actually contain. I don't think we want to remove 160 Mhz 802.11ac for all devices, do we?

Secondly it seems that PRs on GitHub are not looked at, since @Anteus PR for fixing the mt7915 never got attention on GitHub until they submitted some other way.

1 Like

At this point, I suspect we are just waiting for nbd to back port the MT7915 fix in snapshot to 23.05 so that it is in available in the next 23.05 release candidate.

Perhaps we do not want to remove 802.11ac 160 (VHT160) for all mediatek devices, but we certainly want to do that for MT7915 devices for the reasons summarized in odrt's post here.

Your report that the alternative script modification approach that removes VHT160 fixes the same ax upload problem for MT7986 / Filogic 830 suggests the fix should be expanded to more than just MT7915.

Well, I would certainly like to stop having this problem on my non-MT7915 too, like you follow up with in the end.

Yup exactly.

@odrt do you have any advice on the matter of a larger-scale solution beyond just fixing mt7915?

although I have not experienced any issues on my mt7981 based Cudy WR3000 . I think it should be an tick box in wireless configuration to disable VHT160 announcement from hostapd . So this PR should go to luci repo apart openwrt of course

1 Like

I think for Filogic 830 MT7986 the situation is slightly different compared to MT7915. Since MT7986 has vendor official 160 MHz channel support it should be verified if the issue is present in stock firmware or not. If yes then it's a case for customer support and a Mediatek developer fix. Removing 160 MHz channel support like Mediatek did for MT7915 would not be an option for MT7986 in this case.

If the issue is also present in MT7986 OEM firmware then Mediatek could in theory fix this together with Apple, if Mediatek finds out that the root cause of the bug can only be fixed on Apple side.

If AX capable Apple devices perform as expected with their AX mode uplink on OEM firmware then it's probably a MT7986 driver or OpenWrt issue. Since MT7986 has official 160 MHz support it could be useful to debug the issue for this SoC until it is understood and then implement a real fix. Or when it's a hardware or firmware blob related bug: know that it needs to be accepted and if dropping VHT160 capability would be a reasonable workaround, too.

6 Likes

Hello,

Could the patch be backported to openwrt-23.05 so it can be tested in 23.05-snapshot and be ready for a potential RC3?

Yes, the patch could be backported to 23.05 branch, if a developer with rights to commit would decide to do so. The patch worked with kernel 5.15 before.

2 Likes

The patch has now landed in 23.05: https://git.openwrt.org/?p=openwrt/openwrt.git;a=commit;h=dc370ad19adaedfaca2f14c3bb3aea0af28f625c

9 Likes

Thank you for highlighting this. The upcoming 23.05-snapshot images https://downloads.openwrt.org/releases/23.05-SNAPSHOT/targets/mediatek/ should include the patch.

Fingers crossed for an RC3 some time in the near future!

Next iperf3 performance measurement update: testing the first 23.05 images with the patch included and stock mac80211.sh.

Test on 23.05-SNAPSHOT r23308-5ded0a3975, image build time: Thu Jul 27 00:45:16 2023
AX 80 MHz with stock mac80211.sh

432.3 Mbit/s, 80 MHz, HE-MCS 4, HE-NSS 2
680.6 Mbit/s, 80 MHz, HE-MCS 7, HE-NSS 2, HE-GI 1

iperf3 upload: 223 Mbit/s

On past tests with 23.05-rc2 and the mac80211.sh modification I got 292 Mbit/s and a more stable uplink rate. But the drop of the uplink connection to HE-MCS0 and 14 Mbit/s upload rate is fixed.

I couldn't wait. :smile: @Anteus' patch needs to be tested ASAP.

1 Like

23.05-snapshot-r23308-5ded0a3975 with WAX206 + iPhone 12

First few tests with online speedtest show great results. Unfortunately I can't test anymore with iperf as the iperf client I was using on the phone just vanished from the store.

This app provides an iperf3 function:

3 Likes

I am very cheered to see this.

1 Like

btw setting:

config wifi-device 'radio1'
        ...
        option short_gi_160 '0'
        option vht160 '0'

with radio1 as your 5Ghz device in /etc/config/wirless has the same effect like:

sed -i '/SHORT-GI-160/d;/vht160_hw/d;/$vht160/d;/# supported/d' /lib/netifd/wireless/mac80211.sh

for me.
(using HE80)

3 Likes

that's supported by 2.4ghz ax too, no?