Any chance to include EBTABLES for both IPV4 and IPV6 for this build ? Now when I try to install it it conflicts with kernel..
Intermediate update:
root@OpenWrt:~# uptime
17:52:30 up 4 days, 19:27, load average: 0.16, 0.04, 0.02
root@ap-R7800:~# uptime
17:52:45 up 4 days, 17:50, load average: 0.01, 0.00, 0.00
root@ap-c2600:~# uptime
17:53:08 up 4 days, 21:40, load average: 0.09, 0.10, 0.08
So far no complaints yet from house members (knock on wood 
)
4 Likes
If anyone is tempted, there are some cpu frequency patches that Ansuel made a PR for recently. Hnyman has tried to compile those patches and his R7800 boots. Iāll look into creating a newer build with those patches from Ansuel and the latest from @tishipp . Grabbing patches from multiple sources isnāt something Iām very familiar with so it might take a few days to get it ready to build.
Just wanted to add my 2 cents on how this thread is going, just wanting to be sure not to be misinterpreted.
I appreciate a lot all the effort you all guys put in trying this and that, but i also think that since we need the more testers the possible, we'd need to have ONE shared repository where to put fixes to try.
I can build my images myself, but i do not know how to insert specific patches, why don't we evaluate the idea of having one repository (that could be @ACwifidude 's one, for example, maybe with a "official" branch and a "development" branch, also on master) where you guys can put things so that we can test the latest development?
I see you, @vochong and other guys reporting tests on new patches, but it's always one environment, and we know it changes a lot case by case..
Sorry, just my two cents 
1 Like
Well said. I've mentioned before I'm not a developer, I know my way around a Linux prompt enough to get systems installed, up and running and maintaining them. If they're broken, mostly I can fix them too. But when it comes to developing code to fix them; learning to swim basically. I can clone a GitHub repository and then add in some other patches via git, but when I want to push that back to my cloned GitHub repository, a lot more gets committed that I don't want committed. I agree with you, but I'm glad I can at least put pieces together in a build. Cloning and branching is a bit vague to me, to be honest. Doing this stuff isn't also my daily job so I need to learn in my own time.
1 Like
i fully understand your point, but i don't think this is helping in sorting out the code problems we have. Maybe @vochong is a little bit confident in managing code and it and can mantain a "side" repository with code to test, while @ACwifidude keeps the "official" nss repository where to put developments that we tested as good? Really, i don't want to give homeworks to people, i'm just trying to think to a way to turn more effective all the developments you guys are getting into the code..
Sadly, this thread has been highjacked by 'virologists' and other self taught """"devs"""" with... OP blessing. Most probably personal problems / not much time to allocate to this project is allowing this.
Anyone can make a build. Very few understand what they are doing and blindly add / remove various commits hoping a DEV badge.
Original stable build was properly running for months without any issues. 'Plagued' WiFi clients can bring any router to a halt.
i don't fully agree with the first part of your post, but even less with this part.
With nss we have ALWAYS had self reset problems, starting from the first builds from @quarky and @Ansuel and with 5.10 much more.
We've never had a rock solid build with NSS (i can remember some builds last for maybe one month, but "properly rulling without any issue" seems to much to me
That's why i think we should all focus on trying to sort out the problems. Than for sure not all of us (almost noone lol) will be able to fix it, but if we can find where the problem is, we can be helpful for who knows where to put his hands.
That's why i do not like going in seventeen different ways.. @ACwifidude did a wonderful job putting everything together, allowing almost anyone to have a working nss build, we should not waste this committment.
Now, topic closed to me, no more cents to add Thanks
2 Likes
imho times are mature, there is a good dev here (not referring to me) that wants to refresh this and finally progress with new kernel and new firmware. With NSS stability is improved but my theory is that with NSS offload the cpu is just scaled less and the silent problem is less present.
3 Likes
I add in the commits once they are fully functional / add stability / fix issues. So your idea of my repo being the main with lots of other devs / custom builds is exactly my thought. That is why I donāt mind the additional builds / additional ideas - it helps everyone test and improve ipq806x NSS.
For anyone having stability issues - lock your cpu to one freq (max or the next step down available for your cpu) or use the performance gov. Some devices have more cpu issues than others depending on your config and cpu usage. Looks like Ansuel recently found a cpu issue and is pushing a patch for 5.15.
Iām using the build defaults with irqbalance on (irqbalance prob not needed). Have dedicated x2 dumb AP r7800 that only do wifi. Below are the main router (wifi turned off) stats from today. Pretty much all accessory programs turned off (Adblock, vpn, etc). Note the very low loads and infrequent need to change frequencies after 2 weeks of uptime.
OpenWrt SNAPSHOT, r20439+24-a96382c1bb
-----------------------------------------------------
root@OpenWrt:~# uname -a
Linux OpenWrt 5.10.138 #0 SMP Wed Aug 31 19:04:56 2022 armv7l GNU/Linux
root@OpenWrt:~# uptime
10:36:29 up 14 days, 17:29, load average: 0.06, 0.01, 0.00
root@OpenWrt:~# cat /sys/devices/system/cpu/cpufreq/policy1/stats/trans_table
From : To
: 384000 600000 800000 1000000 1400000 1725000
384000: 0 19 11 9 10 20
600000: 17 0 2024 3 9 29771
800000: 13 1925 0 3 14 275
1000000: 14 5 5 0 8 11
1400000: 8 7 12 17 0 9
1725000: 16 29869 178 11 12 0
root@OpenWrt:~# cat /sys/devices/system/cpu/cpufreq/policy*/stats/time_in_state
384000 2145
600000 126010480
800000 12027
1000000 275
1400000 257
1725000 1236105
384000 1470
600000 126554385
800000 6806
1000000 148
1400000 160
1725000 698320
Here is one of the APs (dunno why it reset once? Has been steady otherwise, having two APs makes the home users happy- no complaints)
OpenWrt SNAPSHOT, r20439+24-a96382c1bb
-----------------------------------------------------
root@OpenWrt:~# uname -a
Linux OpenWrt 5.10.138 #0 SMP Wed Aug 31 19:04:56 2022 armv7l GNU/Linux
root@OpenWrt:~# uptime
04:33:50 up 13 days, 13:28, load average: 0.00, 0.00, 0.00
root@OpenWrt:~# cat /sys/devices/system/cpu/cpufreq/policy1/stats/trans_table
From : To
: 384000 600000 800000 1000000 1400000 1725000
384000: 0 13 15 14 12 16
600000: 19 0 7732 2 7 43360
800000: 10 6750 0 4 9 1443
1000000: 12 5 1 0 15 12
1400000: 13 6 6 19 0 15
1725000: 15 44347 462 6 16 0
root@OpenWrt:~# cat /sys/devices/system/cpu/cpufreq/policy*/stats/time_in_state
384000 1091
600000 116089765
800000 29895
1000000 223
1400000 269
1725000 1055086
384000 2565
600000 116179406
800000 21193
1000000 157
1400000 217
1725000 972791
Separately - anyone working on 5.15? If you have a working / in progress repo, we can all start to collaborate on preparing for the inevitable bump coming soon to master.
5 Likes
that made me laugh 
Please don't waste time responding to people who are only capable of spitting out virulent language. They're like TV anchors going on air exhorting people to riot and be anti-vaxxers while at the same time getting themselves, their family fully vaccinated to be eligible to attend the Correspondentsā Associationās annual dinner, and hiding in the comfort and safety of their home watching people getting killed, being sent to prison, or dead from viruses (aka their very own mouth).
This is exactly what a """"virus"""" is for this good build... until it got """"infected""""...
For those who are not familiar with NSS, here's a nice explanation.
4 Likes
My EA7500V1 keeps running out of RAM and rebooting once every few hours, which is a shame since stock owrt has bad network speeds over pppoe. I tried putting the gov to performance. Any tips?
If you are running out of RAM I would turn off less essential programs. Unfortunately the ea7500v1 only has 256mb so you have to be cautious.
You can add something like this to your startup script in luci to stop all the extra programs:
for i in adblock nlbwmon openvpn sqm; do
if /etc/init.d/"$i" enabled; then
/etc/init.d/"$i" disable
/etc/init.d/"$i" stop
fi
done
Restart your router and see how your memory is doing.
kmod-ath10k-ct-smallbuffers might be worth looking at.
Just to give some context to this answer - there's 2 paths to the startup script (/etc/rc.local)
Luci login: system -> startup, select right tab of 2, use the edit window.
Ssh to root@your-router, manually edit /etc/rc.local adding the lines @ACwifidude gave.
Kernel 5.15 (QSDK 11) -> https://github.com/qosmio/openwrt-ip806x
NSS packages -> https://github.com/qosmio/nss-packages
I decided to break out the nss-packages to follow @robimarko 's methodology. I was going to fork it, but most of the work is on QSDK 12 and 11.4 seems to be the last version for ipq806x. With @Ansuel recent upload of QSDK 11.4 FW for ipq806x, I was hoping get that working as well, but it's been a huge PITA. I haven't been able to get past kernel panics. The firmware just refuses to load.
In other news I've worked on tweaking a lot of the existing packages, and backported changes from QCA that were pretty platform agnostic. Still needs a lot of testing, especially the PPPoE/PPP, since I don't use that with my ISP.
3 Likes
I tried turning off those services and a few more.
About kmod-ath10k-ct-smallbuffers, nss repos don't seem to have it and openwrt repos don't install it with incompatible version
Now here's the weird thing, the memory usage seems to plateau at about two thirds of the available RAM, that must have something to do with it
LE: There seems to be a memory leak too, when fresh booted it starts at 74MB and slowly creeps up until 132MB where it plateaus.