It is already enabled.

Just to exclude the ath10k-ct versus ath10k differences; which version are you running? I suspect the ath10k-ct driver. It might be worthwhile to flash the ath10k variant of the image. Easy enough to do, just upload the sysupgrade image and keep your current settings while flashing. If that driver is the culprit, you'll notice a difference after a short period of usage. I keep having issues with the ath10k-ct driver, but not all have the same experience.

I'm using the r7500v2 so there is only a anth10 version. I was using that when I first posted. Now I am using the other build just the regular upgrade image.

Sorry I missed that, my bad

No problem thanks for trying.

I have done some more testing. I connected my PC directly to the modem to rule out my internet connection. Everything works beautifully.

Next, I downloaded every v21 build I could find. I flashed them all one at a time. I did no change anything except offloading. All of them broke my connection. This rules out wifi and drivers. I was hard wired with wireless disabled.

Finally, I retested every v21 build again. This time I made sure that all services that could interfere with the loading of webpages were turned off or removed completely. My connection was still broken.

Now, I am on your kernel 5.4 nss build ath10k and everything is working as intended. I do not understand why this is happening. However, something to do with the new v21 builds destroys the internet for r7500v2 users.

Thank you very much @ACwifidude !!! :bowing_man: :bowing_man:
I'll try the new 21.01 build today

1 Like

That is interesting. Wonder what it is about 21.02. Settings are the same in the .dts files. Let me know if you stumble upon something unique for 21.02 + r7500v2 so that I can find a fix.

1 Like

My conclusion thus far: the best connection quality is achieved by flashing 0618-master-ath10k NSS build and setting it to reboot every night.

It rivals the connection quality of recent NSS builds and of hnyman's builds (even with ath10k installed). I can talk on my phone via 2.4mhz from places in the house which used to lose audio, and 5ghz connection is also faster and more stable.

Maybe there's just a bug in that version of firmware which maxes out wifi signal strength, or maybe it's actually got something right. It will start to flake out after 1 day though, i.e. 5ghz will drop out for a minute, so I set it to reboot nightly.

I did find out there was a performance issue with my ISP. Xfinity has been working on the internet in my town for the last two days. I do not think this was the problem as everything seamed to be fine n 5.4 also with stock and ddwrt on the r7500v2. Things should be fixed today though so I am going to retest just to make 100% sure. I am wondering what other users of the r7500v2 experience is with v21 builds. I would like to hear what they say if they would chime in.

Hello friends, an annoying thing that is happening to me since the latest versions, I could not say from which, is that from my windows pc, I start it, and it takes a minute or more to surf the internet, I have it with DHCP, for that with fixed IP sometimes does not allow to navigate. Internal network if you have, it seems something related to DNS. Does the problem sound familiar to you? Thank you.

Hi, @ACwifidude today, I got this? How can I fix it?

Applying ./patches/subsys/999-mac80211-add-option-for-NSS-support.patch using plaintext: 
patching file net/mac80211/Kconfig
patching file local-symbols
patching file net/mac80211/ieee80211_i.h
patching file net/mac80211/iface.c
patch: **** malformed patch at line 137: <<<<<<< HEAD

Patch failed!  Please fix ./patches/subsys/999-mac80211-add-option-for-NSS-support.patch!
make[3]: *** [Makefile:573: /home/jerry/openwrt-21.02-nss/build_dir/target-arm_cortex-a15+neon-vfpv4_musl_eabi/linux-ipq806x_generic/backports-5.10.68-1/.prepared_b510272e7889bd07a66223ab22912932_18f1e190c5d53547fed41a3eaa76e9e9] Error 1
make[3]: Leaving directory '/home/jerry/openwrt-21.02-nss/package/kernel/mac80211'
time: package/kernel/mac80211/compile#3.55#2.38#4.91
    ERROR: package/kernel/mac80211 failed to build.
make[2]: *** [package/Makefile:114: package/kernel/mac80211/compile] Error 1
make[2]: Leaving directory '/home/jerry/openwrt-21.02-nss'
make[1]: *** [package/Makefile:108: /home/jerry/openwrt-21.02-nss/staging_dir/target-arm_cortex-a15+neon-vfpv4_musl_eabi/stamp/.package_compile] Error 2
make[1]: Leaving directory '/home/jerry/openwrt-21.02-nss'
make: *** [/home/jerry/openwrt-21.02-nss/include/toplevel.mk:230:world] Error 2

Interesting. I’ll try to rebuild today and see what’s up with that patch.

Are you building with my newer modifications to that patch?

yes, before I built it, I rebase upstream/openwrt-21.02 and merge some conflicts manually.

hello,
to abandon the master branch and go back to your 21 branch, is it enough just to edit my link to your repository? can i then use same config files and packages?
Thanks

2 Likes

thank you and everyone involved for spending time in getting these drivers working. 21.02 build is solid, didnt try the kernel5.4 build yet.

Throughput is close to what i use to get on OEM firmware. latency in some cases, is way better than the OEM version. Guest wlan didnt work though, latency went up 15x and speed maxed out at 11-18mbps on a 200/200 line. Except for removing some luci, vpn packages, didnt deviate much from your build config.
thanks again.

1 Like

Updated the 21.02 build to the latest 21.02 commit. Latest has the new ath10k-ct commit that fixes 160mhz devices. I don’t have any 160mhz devices - if anybody wants to test with speedtest/iperf I would love to see results. I’d be interested how much it improves performance with encap offloading + NSS gives more performance vs regular OpenWrt.

2 Likes

Thank you @ACwifidude

I have been putting this into my startup since Jan 2021, is it still necessary with your latest builds?

echo performance > /sys/devices/system/cpu/cpufreq/policy0/scaling_governor; echo performance > /sys/devices/system/cpu/cpufreq/policy1/scaling_governor

uci set irqbalance.irqbalance.enabled=1; uci set network.globals.packet_steering=1; uci commit
1 Like

My build has more aggressive on demand settings built in. If you keep settings between builds, no need to re-input irq and global packets (those settings are kept). The cpu can be switched to the performance governor like you have above. I’d expect the aggressive ondemand to be similar in performance (the ondemand allows your CPU to be cooler at the lower frequency when the router is under low use).

1 Like

Thanks for the reply!

Since I don't "keep settings" between builds, I don't need to put either of those into my startup?