it's frustrating, I've downloaded the torrent and it restarts in a matter of hours. I had been perfect for 8 days. amazing.
@dtaht Just wanted to give you two more links that I think you'll find helpful:
Configuring NSS QDISC with TC:
Looks like the Kong build may actually have this properly integrated with SQM in Luci?
My my non-ct firmware version is: 10.4-3.9.0.2-00131 (picked up from the dmesg).
About the speed tests related to SUPPORTS_REORDERING_BUFFER, even if 10.4, at least on my R7800. It still has an effect, I mean.
Below iperf3 is ran without (and then with) -R argument. I use the non-ct firmware as at least the 5GHz radio comes online when I activate 160MHz on channel 100. But is not always staying at 1733/1733 tx/rx rate, I even saw 866/1733 or even lower towards 1100s..
Without the SUPPORTS_REORDERING_BUFFER patch the upload speed is constant at around 730 MBits/sec and the download speed (-R) sits at around 386 MBits/sec. Negligable difference between different runs (up to 10% let's say).
iperf3.exe -c 192.168.0.106 -P 8 -w 131072
[SUM] 0.00-10.01 sec 870 MBytes 729 Mbits/sec sender
[SUM] 0.00-10.01 sec 869 MBytes 729 Mbits/sec receiver
iperf3.exe -c 192.168.0.106 -R -P 8 -w 131072
[SUM] 0.00-10.00 sec 460 MBytes 386 Mbits/sec 0 sender
[SUM] 0.00-10.00 sec 459 MBytes 385 Mbits/sec receiver
However, with reordering patch, the upload speed fluctuates a lot, but on average is very slow (from 174 MBits/sec, maybe up to 290MBits/s). The download speed is similar to the case without the patch (again, 10% differences).
iperf3.exe -c 192.168.0.106 -P 8 -w 131072
[SUM] 0.00-10.00 sec 207 MBytes 174 Mbits/sec sender
[SUM] 0.00-10.00 sec 206 MBytes 173 Mbits/sec receiver
iperf3.exe -c 192.168.0.106 -R -P 8 -w 131072
[SUM] 0.00-10.00 sec 458 MBytes 384 Mbits/sec 4 sender
[SUM] 0.00-10.00 sec 457 MBytes 383 Mbits/sec receiver
Hope it helps 
That's what I meant.
SUPPORTS_REORDERING_BUFFER tells mac80211 that firmware or driver will handle the rx reorder.
Upload being so slow just proves that firmware and driver don't do reordering.
1 Like
I'm trying the Master + NSS Hardware Offloading Download on my R7500v2 and it seems very nice. Thanks for your work, @ACwifidude.
The only issue so far is I can't install minidlna or ksmbd-server packages from OpenWRT software repo because of this: The installed version of package kernel is not compatible, require 5.4.110-1-8c187f92… while 5.4.110-1-bf450115… is installed.
Any help will be appreciated.
Awesome. The downside of custom images is that it adds on a hash to the end and opkg will complain when installing kmod packages.
The cleanest method to add those packages is for you to custom build it from scratch - adding the packages you desire. Otherwise you might be successful trying to opkg force install it with your current setup.
Thanks for your quick replay, @ACwifidude. I moved to the 21.02 (Stable) + NSS Hardware Offloading Download build and I was able to install samba4 and other packages without issues. Right now, the only package I need is minidlna, which can't be installed from the software repo.
In order to force install minidlna do I have to use a terminal connection and include the --force-depends option, like this: opkg install minidlna --force-depends? That's my understanding from this reference.
Building from scratch is the cleanest but that should work too.
1 Like
For at least the last 3 versions I have used 3/15, 4/03 and 4/11, I have noticed some strangeness with the 2.5Ghz radio. I can connect from my phone, but it initially says 'no internet' and, though connected, I cannot reach anything on my LAN. I have other devices on the LAN (RPis) that connect and appear to have connectivity. I've tried rebooting the phone to no avail. This morning (running 4/14) the RPi went down (no internet) and I tried to reach it. The router showed it still had an IP, but I could not connect to it. I 'disconnected' the device and rebooted the radio and when it reconnected, It started working again.
looking through the log, I saw this:
Tue Apr 20 06:06:52 2021 daemon.err nlbwmon[4419]: Netlink receive failure: Out of memory
Tue Apr 20 06:06:52 2021 daemon.err nlbwmon[4419]: Unable to dump conntrack: No buffer space available
this happened about the time it lost connectivity. Does this point to anything?
Also, and I think this may be a phone problem and not a router/firmware problem, but thought I'd mention it in case anyone here has any insight. My child has a Motorola e5. It used to connect to the 5GHz radio, but for a few months, it has not. I have been using @ACwifidude builds since September 2020, or so, and it used to connect but now it only connects to teh 2.5GHz and with the flakiness of the 2.5GHz radio, I get a lot of static from the child...
Should I try the other driver (non-ct)? If so, can anyone give me a step by step on how to try it?
the 2.5GHz radio is seriously hampered. Woke up this AM and noticed that all 2.5GHz devices were off line. The log shows this again:
Thu Apr 22 19:03:59 2021 daemon.err nlbwmon[4419]: Netlink receive failure: Out of memory
Thu Apr 22 19:03:59 2021 daemon.err nlbwmon[4419]: Unable to dump conntrack: No buffer space available
...
Thu Apr 22 21:38:20 2021 daemon.err nlbwmon[4419]: Netlink receive failure: Out of memory
Thu Apr 22 21:38:20 2021 daemon.err nlbwmon[4419]: Unable to dump conntrack: No buffer space available
...
Fri Apr 23 01:49:21 2021 daemon.err nlbwmon[4419]: Unable to dump conntrack: I/O error
Fri Apr 23 01:49:21 2021 daemon.err nlbwmon[4419]: Unable to dump conntrack: I/O error
...
Fri Apr 23 01:58:22 2021 daemon.err nlbwmon[4419]: Netlink receive failure: Out of memory
Fri Apr 23 01:58:22 2021 daemon.err nlbwmon[4419]: Unable to dump conntrack: No buffer space available
Also, I have a MAC address that I want the router to ignore and not provide an address to. I list its IP as 'ignore' yet I see this in the log:
Thu Apr 22 17:59:55 2021 daemon.err dnsmasq[5177]: bad address at /tmp/hosts/dhcp.cfg01411c line 16
That line of the dhcp.cfg01411c file is:
ignore Unknown.lan
From the hints in the Luci UI, this appears to be correct. I have no quotes around it as Luci will not let me do this. Is this correct?
Also, I still would like to know if I can/should try the non-ct ath10k driver and how to do that.
Don't think @ACwifidude build the non-ct driver
I stay true to the defaults in master. You can swap to non-ct with the command in the second post or you could build with non-ct. See if changing helps for your clients.
Here is your answer but no NSS:
All builds are with the default ath10k-ct wifi driver. The mainline ath10k wifi driver is being offered as a downloadable .ipk in the download directory of each build.
That only replaces the firmware.
Maybe you can build the non-ct driver as a module so people can just install the ipk if they want to use non-ct driver.
1 Like
After 10 days uptime, NSS cores set at 800MHz and onDemand governor set at 800Mhz minimum frequency, the router rebooted during a large rsync operation (from laptop / wifi -> server / lan).
I didn't get the kernel panic, but the last lines console-ramoops-0 are:
[957667.702309] ath10k_pci 0000:01:00.0: wmi command 36967 timeout, restarting h
ardware
[957668.314909] ath10k_pci 0000:01:00.0: failed to send pdev bss chan info reque
st
[957668.315287] ath10k_pci 0000:01:00.0: failed to set beacon mode for vdev 0: -
108
[957668.321032] ath10k_pci 0000:01:00.0: failed to set dtim period for vdev 0: -
108
[957668.328836] ath10k_pci 0000:01:00.0: failed to recalculate rts/cts prot for
vdev 0: -108
[957668.336132] ath10k_pci 0000:01:00.0: failed to set cts protection for vdev 0
: -108
[957668.344203] ath10k_pci 0000:01:00.0: failed to set preamble for vdev 0: -108
[957668.351669] ath10k_pci 0000:01:00.0: failed to set mgmt tx rate -108
[957671.381293] ath10k_pci 0001:01:00.0: bss channel survey timed out
Not sure if this has triggered the reset (I'm running the non-ct firmware to have access to the 160Mhz channels).
I doubt that the problem is the wifi driver, at least not only that would be the problem. my routers stable with nss as long as i dont use wired torrent for example. if I turn on torrent it restarts in hours. multiple connections overflow memory? I don't know .... I don't understand anything about this topic, I'm sorry, I'm sorry I can't help.
Have you been monitoring active connections on the status page, how many connections do you see when you torrent? Do you also use samba?
Current builds are not optimized for torrent usage, it is possible, that the unit has not enough reserved memory configured. I can add a few changes to the next build, if you want to test?
connections do not usually exceed 1000, in total. the torrent client is very limited as it is a synology nas. I am willing to try any modification, no problem.
problems are only with nss.
I noticed, that we have not added the min freq setting in the nss build, this is very likely the issue for the sporadic problems.
Who can solve this issue?
Please do something on this