Ipq806x NSS build (Netgear R7800 / TP-Link C2600 / Linksys EA8500)

How is your IPv6 configured? R7800 with NSS should accelerate IPv6 if it is configured using standard DHCPv6.

Well - static IPv6-addresses for inner Vlans (private subnets) and DHCPv6 for WAN-interface - I have testet IPv6-Performance only between different inner Vlans (around 300 Mbit/s throughput, while IPv4 between same hosts is reaching around 900 Mbit/s)

That’s odd. If you are only testing within LAN, it should just bridge the traffic. You should get full line speed.

No, it´s not a test within a Vlan - the Test is about routing between different Vlan-interfaces.

Just switching within the same Vlan has normal performance around 940 Mbit/s... - but I am using different Vlans and routing between them

let´s make it more clear:

  • Interface "LAN" = Vlan 1: 'fd88::1' + public IPv6 + private IPv4
  • Interface "Secure" = Vlan 2: 'fd88:0:0:17::1' + public IPv6 + private IPv4
  • Interface "wan" = vlan 999: DCHP-IPv4
  • Interface "wan6"= alias of "wan": DHCPv6 with route-delegation

When I do an iperf3 between hosts in subnets "fd88::" and "fd88:0:0:17::" then I get something around 300Mbit/s, but between same interfaces with IPv4 it is around 900Mbit/s

It's not really spamming the logs, i've seen it happen right after boot a few times, one theory I read was that it could occur when a device (br-lan) is set to promisc mode, which I have when using NSS.
For me the MAC-address seems different each time too and isn't related to a mac address on the router itself, it's related to devices I have on wifi.

I saw your thread about the very same issue, in your case it was ipv6?

But everything seems to work and I can't troubleshoot it as I don't get the error very often.

I don't think i know.

I found that removing some duplicate ipv6 link local address (lla) from interfaces (in particular bridge interfaces) eliminated the messages. This post describes one attempt at a hotplug script to do this.

I still have a variant of the hotplug script that is adapted for dsa but i just checked my switch0.x bridges and they have duplicate ipv6 lla's and i see no error messages atm.

1 Like

Did some cleanup and consolidation of the commits on the 5.10 kernel build. Mostly the build has a kernel bump and it seems some of the firewall4 issues are potentially fixed in master. Anyone with a ea8500 or ea7500v1 let me know how the .dts file likes it best for stability on the 5.10 kernel. I've excluded these two linksys devices on 5.10 builds (they seem to work on 5.4 builds per report).

2 Likes

Hi @ACwifidude , the new driver setup is tested OK on my Hong Kong version ea8500.

For all those running recent ipq806x builds, whether from @ACwifidude or your own, do you encounter Wi-Fi slowdown/non-responsive issues after a few days of uptime? Trying to gauge if the issues are due to recent commits to the 21.02 (since Nov 2021) and master (since Jul 2021) branch.

2 Likes

I noticed slowdowns back in april 2021... but I can't say I have them now, but it could be because the router reboots by itself now from time to time.

The reply to that was that it could be related to the ath10k-ct drivers.

Are you using upstream ath10k with encap offload?

Personally, the Not-CT controllers work much better for me, they don't fail over time, the speed is quite high, reaching WAN limits.

The problem of random reboots, for some unknown reason, in my case they are directly related to the "ifconfig br-lan promisc" command, if I use it, I have a reboot after 2 or 3 days.
A colleague pointed to this command and firewall, I don't think it's my case, since my servers work from WAN, my firewall rules are simple, simple port forwarding, and allow openVPN in firewall rule.

@ACwifidude (thanks!!!)
OpenWrt 21.02-SNAPSHOT r16461+16-5beaa75d94 / LuCI openwrt-21.02 branch git-22.007.66044-dd9390c
Kernel Version 5.4.168

uptime 4 days and 19H and testing, but I estimate that you will not see restart problems, and the operation is perfect, very fast, very stable.

For some reason I discovered that a webcam I was using was causing instability in the router (I don't know), I removed it from my house, and the problems were solved.

currently waiting to see how much uptime i am able to get, but i think there will be no time limit.

I use factory router, with pppoe connection, OPENVPN with 2 active clients 24/7 and adblock (210,000 entries)

thanks to all, and sorry if in the past I was something annoying, I needed a solution.

Yes, I'm using the ath10k driver with encap offload patch. Is this the issue?

I suspect my Linksys E8450 with mt76 driver is also affected with the same issue, but I'm not sure. I'm preparing another build with the old mac80211 behaviour to test it out.

My R7800 running the latest openwrt 21.02 (kernel 5.4.171) with ath10k and encap offload, with the old mac80211 airtime fairness seems good so far. 4 days uptime without any noticeable degradation to the Wi-Fi traffic. Previously by the 3rd or 4th day, the Wi-Fi would be almost unusable.

Do you have a lot of Wi-Fi client connected to your ipq806x router? I notice that when you have many clients connected to it, it will get a lot worst fast.

Hmm... ok. Interesting. however wlan0/1 are always in promisc mode for some reason, at least for me. So why would setting br-lan to promisc mode cause issues?

Not that many, I have an wifi bridge tho' so each client that's behind that bridge show up as a wifi connection to the router, with those counted I have around 12 clients connected.
I can't say I've noticed any issus with slow wifi for a while, my wife complains tho'.

yes, wlan if it works as if the mode was activated, I really do not know to check if it is

I also experienced the issue but I don't think it's because of it.
The patch doesn't touch any airtime related stuff.

We should probably ask Felix about it on IRC.

In this case, it would seems that the Wi-Fi bridge technically only have 1 Wi-Fi client, and the air-time fairness don't come into play. This is at least consistent with what I see with my R7800. If only 1 or 2 clients gets connected, everything seems fine.

I just deployed my E8450 with the old airtime fairness algo (i.e. round robin instead of the virtual time based algo.) Will monitor it maybe a week before I conclude. So far my R7800 is running good, so that's one data point.

I 6 real clients connected right now, usually a few more, so maybe around 8 during normal "use".