Belkin RT3200/Linksys E8450 WiFi AX discussion

[MeIsReallyBa]

wireless-tools is needed to use iwpriv.

[Mushoz]

I used to watch IPTV just fine on my home network with a rt3200 via igmpproxy on an older snapshot. However, since updating to the 22.03.0 version, it's totally broken:

1. If I disable igmp snooping, WiFi gets flooded with multicast traffic as expected when someone is watching TV, and slows down to an absolute crawl.
2. If I enable igmp snooping, multicast traffic is not working at all and I am unable to watch TV.

On a previous snapshot, igmp snooping was working just fine to prevent multicast flooding on WiFi, but multicast worked just fine and I was able to watch IPTV. Is someone experiencing the same issue?

[Anteus]

You could try latest 22.03-SNAPSHOT. This commit mention igmp snooping and multicast

hostapd: rename hostapd multicast_to_unicast option to multicast_to_unicast_all

[Mushoz]

As the description already tells you, and the git diff shows, it's merely a rename. It doesn't actually change anything functionality wise.

[MeIsReallyBa]

There are many mt76 patches in MTK SDK. For example, mtk turn off airtime fairness.

https://git01.mediatek.com/plugins/gitiles/openwrt/feeds/mtk-openwrt-feeds/+/a833e5a20bd93393ff3342c4fc903174cecd786f/autobuild_mac80211_release/package/kernel/mt76/patches/1115-mt76-airtime-fairness-feature-off-in-mac80211.patch

Did anybody test if these patches make influence on speed?

[namidairo]

@MeIsReallyBa

Did anybody test if these patches make influence on speed?

I'm not sure if anyone's posted results with it off, but there has been a fairly lengthy discussion in the AQL thread trying to get a good balance between speed and latency/trying to track down throughput bugs. (Although the user is testing with a mt7621+7615, not mt7622+7915.)

AQL and the ath10k is *lovely*

I built openwrt head last night when the final patches for AQL and the ath10k-ct landed. Oh! what a difference in bufferbloat-related performance! Where, on this purposely terrible link (12mbits or so), delays would spike past 4 seconds, and things like the babel routing protocol fail with any background load, now it holds things steady well below 60ms with totally solid throughput. Before: [before] After: [after] I tested this on the uap mesh and mesh pro using the smallbuffers version of…

The thread is AQL in general, despite ath10k in the title.

One should probably take into account that it looks like mediatek were trying to pass cert, thus they may have just disabled it in order for a test suite to pass/complete. It's nice to see them openly developing on top of mt76 though.

[amteza]

@namidairo

user is testing with a mt7621+7615, not mt7622+7915.)

Yes, that's what I've at hand! Not buying new devices to test and patch, but happy for you to send one my way.

Please, note that ATF can be manually disabled if you want to via sysfs, if anyone want to test.

[anon11662642]

@LameDuck

I'm stopping with the proprietary wifi drivers - they are good if you want a simple setup and the performance is better, but the missing features are a deal breaker for me. Please let me know if any of these are working for you, maybe I just didn't figure them out.

Proprietary drivers:

• Get better speeds
• Have marginally better range
• Don't support multiple APs on the same frequency (e.g. guest network)
• Only support wireless Mesh (FT over DS isn't supported)
• Don't support 802.11r fast roaming for hand-off between APs

The open source drivers have their own quirks, but hopefully will improve.

• Stable at AC speeds on 5ghz
• Proper wireless handoff using 3 APs with FT over DS
• Allows setting up a guest network on its own VLAN
• Can leverage usteer to instruct client roaming
• AX was a little flaky for me when doing hand-off between APs
• Slower speeds than proprietary driver (still getting 400 mbit in most places around the house)

I'm still playing to get the best settings, but since I don't need gbit speeds over wifi - the open source driver is more suited to my needs.

@geminis3

Had to revert to stock mt76 on 802.11ac mode, bufferbloat was terrible over 5GHz WiFi (probably due to no AQL nor improper integration with the rest of the system).

So far this is a superb 4x4 WiFi 5 router for its price till WiFi 6 issues get ironed out, as always thank you to all the devs that make OpenWrt and OSS Linux drivers possible.

[Anteus]

Also went back to mt76. A quick bufferbloat test showed +17ms on mtk and +4ms on mt76.

[Lynx]

Well so much for the proprietary drivers then! The cake-autorate project has shown me that a trade-off exists between bandwidth and latency. There's no point in offering very high bandwidth if latency is all over the shop. The latter has a huge impact on everything.

[16098]

@Lynx

proprietary

I didn't use a dedicated testing tool to test for latency. But from my own experience of playing games, the stability of proprietary driver on iphone is much higher than mt76.

[ishiryoku]

If this is important, once you have exceeded 100 Mbps of bandwidth or at most 200, I think that for most people it is not necessary anymore, but the ping is super important.

I have noticed that now ISPs offer high bandwidths, but now the latencies have increased to an average of 40 to 60 milliseconds.

I really don't know what the cause is, but I assume that having many users affects performance.

[16098]

The network delay displayed in the iphone game is the same for mtk proprietary driver and mt76 most of the time. But mt76 will often suddenly have high network latency, while proprietary driver does not have this problem. I have no need for bandwidth, but the unstable latency of mt76 makes me temporarily choose the proprietary driver.

[anon4457646]

all that is not true, is just a legend, configure mtk the right way
mtk work better on buffer and on speed too.
also on range
also stable
also ax

[Lynx]

Significance of the following?

Sun Sep 11 09:43:35 2022 kern.err kernel: [177156.570058] cpu0: failed to scale up voltage!
Sun Sep 11 09:43:35 2022 kern.err kernel: [177156.574535] cpufreq: __target_index: Failed to change cpu frequency: -110

[jebradl]

I have been looking at the install procedure for putting OpenWrt on my Linksys E8450 router. I have been using openwrt for several years, beginning with a WRT54Gv2. When do you use the .bin files vs. the .itb files? I've always used the sysupgrade.bin to upgrade, but I see that there are both present.

[mike]

It's been a while since I used the installer, but I don't think you'll need the .bin file at all in normal usage, just .itb.

[geminis3]

Has anyone tried building mt76 without airtime fairness? It causes more harm than good on most environments.

[Mushoz]

No need to recompile. You can simply disable it by setting the debug kernel flags.

[geminis3]

Can you enlighten me? thanks