IPQ806x NSS Drivers

jack338c · 2020-06-26T10:52:54.895Z

I didn't try qsdk10 or 6 yet
Do you have any idea? I really want to try openwrt master with nss driver

thanks

Ansuel · 2020-06-26T11:23:56.504Z

My repo... we are testing stability and you need to do some workaround to make it stable... just read some post up... also i need to sync qith quarky changes so the package to select are different

jack338c · 2020-06-26T11:40:06.167Z

ok,I got it
I guess it's hard for me to test nss driver at the monent
maybe I need wait for the final stable version later
thanks for your great job

rog · 2020-06-26T13:56:39.585Z

Hi all. I had to revert back to non-nss firmware due to the wifi ping problem but router did not have any reboots since yesterday.

@quarky: echo 1 > /sys/class/net/wlan[x]/brport/hairpin_mode did not work for me. In addition, I did not have time to figure out how to change the ap_isolate option to "0" to see if that would solve the problem.

rog · 2020-06-26T14:00:25.472Z

neonman63:

try to disable ap_isolation, i have the same problem when devices connected to the same wifi dev

Could you give me more details on how to do that? I did not find any easy fix for that.

Gram · 2020-06-26T14:19:25.817Z

Hi guys,

Managed to merge Ansuel "kernel5.4-nss-qsdk10.0" branch and building image is ok.
But what is the way to force NSS cores to 800MHz (for expected stability) for testing?

I cannot find a direct answer for that in above messages.

quarky · 2020-06-26T14:30:02.835Z

Gram:

But what is the way to force NSS cores to 800MHz (for expected stability) for testing?

echo 800000000 > /proc/sys/dev/nss/clock/current_freq

neonman63 · 2020-06-26T15:36:24.396Z

I just fixed it in /lib/netifd/hostapd.sh: set_default isolate 0

fantom-x · 2020-06-26T15:43:09.794Z

The current behaviour is intentional:

IPQ806x NSS Drivers For Developers

This may clarify things a bit:

Gram · 2020-06-26T15:50:25.407Z

Very interesting results after loading the image.

Finally got internet to work after manual reboot but the IP and default GW for my desktop in wired LAN is from my ISP public range.
Like the router is in bridged mode and serving DHCP IPs directly from my ISP.

Based on wireshark spam taken from my desktop, even my NAS box seem to have received IP from my ISP.

Got no clue what's the router IP if any.
Any pointers before returning to standard master image?

Beside default config only thing I can think of I've changed is:

dsnmasq-full & stubby for DoT

rog · 2020-06-26T16:54:34.311Z

neonman63:

I just fixed it in /lib/netifd/hostapd.sh: set_default isolate 0

I had seen that, but it is already set at '0' in my router. Actually, it is set at '0' in both the normal firmware as well as in the nss version. In both cases, "ap_isolate" in /var/run/hostapd-phy0.conf is always '1'.

Thanks anyway. I'll dig deeper into that over the weekend.

quarky · 2020-06-26T21:13:48.044Z

Gram:

Based on wireshark spam taken from my desktop, even my NAS box seem to have received IP from my ISP.

This can only happen when you connected your ISP’s modem/equipment to your router’s LAN port instead of WAN and that your ISP modem/equipment serves IPs using DHCP.

Otherwise using default OpenWrt config should not have resulted in what you have described.

Gram · 2020-06-27T03:15:37.379Z

Thanks and most likely reason.

Too many beers down already before started to "troubleshoot" why the box did not come up.
I'll give it another try later.

perceival · 2020-06-27T04:05:38.766Z

I've done some tests on R7800 in access point mode and here is short summary:

@Ansuel repo:

Without frequency lock reboots every few minutes.
iperf3 test with AP as a server and WiFi client shows pretty high CPU utilization.
For me compilation worked only with single thread.

@sqter repo:

Quite stable out of the box although noticed some hick up that looked like device reboot but with no impact (?) on uptime nor any dmesg/syslog trace. After almost 24h router rebooted itself.
iperf3 test with AP as a server and WiFi client with significantly better throughput and lower CPU utilization.
iperf3 test with wired client goes high with CPU util but I haven't compared with Ansuel's.
Might be broken currently.

sqter · 2020-06-27T08:44:43.294Z

I explain.
My repository is a copy of @Ansuel sources.
All I have is authored by @Ansuel . On my repository I introduce selectively changes that help me achieve stability, but I am not the author of these changes.

BTW. Refresh the source, build now and check stability. I added two changes that work for me.

perceival · 2020-06-27T16:25:29.633Z

Sure. If I understand correctly that would mean in your repo WiFi offloading was enabled while in Ansuel's must have been disabled.

jack338c · 2020-06-28T04:27:45.705Z

@quarky
I tried your QCA shortcut-fe fast-path driver for openwrt master,It's awesome
if possiable add something for kernel4.14 and kernel4.19,that's will be real good
thank you

quarky · 2020-06-28T04:46:59.206Z

jack338c:

if possiable add something for kernel4.14

It’s already running for 4.14. It needs the event notifier patch I’ve done this past week tho, or other modules relying on conntrack notification will fail. The code can be found here:

github.com

GitHub - quarkysg/openwrt at openwrt-19.07-ipq806x-nss

openwrt-19.07-ipq806x-nss

This repository is a mirror of https://git.openwrt.org/openwrt/openwrt.git It is for reference only and is not active for check-ins or for reporting issues. We will continue to accept Pull Request...

Patch should be simple to apply.

jack338c · 2020-06-28T04:56:03.762Z

I did try this before
but it's failed on netfilter conntrack,just like on kernel5.4 while selecting dnsmasq-full

jack338c · 2020-06-28T09:41:25.683Z

I tried change something on 999-01-shortcut-fe-support.patch,then get error while compiling.
Here are my steps

jack@jackr:/media/jack/sam128/Gmobile/opmaster$ patch -p1 < d5bea4bd4889f6b8b2a3b03f9907c12e6a547b75.patch
patching file package/kernel/fast-classifier/Makefile
patching file package/kernel/fast-classifier/src/Makefile
patching file package/kernel/fast-classifier/src/fast-classifier.c
patching file package/kernel/fast-classifier/src/fast-classifier.h
patching file package/kernel/fast-classifier/src/nl_classifier_test.c
patching file package/kernel/fast-classifier/src/userspace_example.c
patching file package/kernel/shortcut-fe/Makefile
patching file package/kernel/shortcut-fe/files/etc/init.d/shortcut-fe
patching file package/kernel/shortcut-fe/files/usr/bin/sfe_dump
patching file package/kernel/shortcut-fe/src/Kconfig
patching file package/kernel/shortcut-fe/src/Makefile
patching file package/kernel/shortcut-fe/src/sfe.h
patching file package/kernel/shortcut-fe/src/sfe_backport.h
patching file package/kernel/shortcut-fe/src/sfe_cm.c
patching file package/kernel/shortcut-fe/src/sfe_cm.h
patching file package/kernel/shortcut-fe/src/sfe_ipv4.c
patching file package/kernel/shortcut-fe/src/sfe_ipv6.c
patching file target/linux/generic/hack-4.14/999-01-shortcut-fe-support.patch
jack@jackr:/media/jack/sam128/Gmobile/opmaster$ wget https://github.com/quarkysg/openwrt/commit/5846d564980ff2f3eb70f0430b9a6c0a86a29da8.patch
--2020-06-28 17:37:25--  https://github.com/quarkysg/openwrt/commit/5846d564980ff2f3eb70f0430b9a6c0a86a29da8.patch
Resolving github.com (github.com)... 140.82.112.4
Connecting to github.com (github.com)|140.82.112.4|:443... connected.
HTTP request sent, awaiting response... 200 OK
Length: 10864 (11K) [text/plain]
Saving to: ‘5846d564980ff2f3eb70f0430b9a6c0a86a29da8.patch’

5846d564980ff2f3eb70f0430b9a6c0a86a29da8.patch      100%[=================================================================================================================>]  10.61K  --.-KB/s    in 0.02s   

2020-06-28 17:37:26 (468 KB/s) - ‘5846d564980ff2f3eb70f0430b9a6c0a86a29da8.patch’ saved [10864/10864]

jack@jackr:/media/jack/sam128/Gmobile/opmaster$ patch -p1 < 5846d564980ff2f3eb70f0430b9a6c0a86a29da8.patch
patching file package/qca/shortcut-fe/fast-classifier/patches/201-rework-nfct-notification.patch
patching file package/qca/shortcut-fe/shortcut-fe/patches/201-rework-nfct-notification.patch
patching file target/linux/generic/hack-4.14/999-01-shortcut-fe-support.patch
Hunk #1 FAILED at 23.
Hunk #2 succeeded at 269 (offset 2 lines).
Hunk #3 succeeded at 317 (offset 2 lines).
1 out of 3 hunks FAILED -- saving rejects to file target/linux/generic/hack-4.14/999-01-shortcut-fe-support.patch.rej