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

SiXX · 2022-02-24T02:52:06.477Z

Kudos to all of the effort you're putting into debugging your issue - I hope you'll get to the bottom of it soon! I do think it's important to point out that everyone's use case is different, so as a result everyone's config will be different. For example - I've used packet steering from Day-1 on this build with the R7800's I support and have never experienced the types of crashes you're having. Then again - I don't use LAN side promisc mode.

Until it's certain - I think it's safer to say "there may be an issue with Packet Steering if LAN promisc mode is enabled"... just to not confuse anyone that is interested in this build -or- may not fully understand the technical ins & outs of what's being discussed. By default - LAN promisc mode is disabled.

shelterx · 2022-02-24T09:29:45.457Z

Yes, that’s basically what I meant, try it, you might have success with it.
In this debug case it was packet steering that caused the crash, I did not try without promisc mode so I don't know about that.

To be fair, packet steering isn't something that's enabled by default either and I don't think any benchmark's been done with it using NSS. Does it help at all? How's the CPU usage with it enabled, and so on...

quarky · 2022-02-24T10:02:16.757Z

shelterx:

To be fair, packet steering isn't something that's enabled by default either and I don't think any benchmark's been done with it using NSS. Does it help at all? How's the CPU usage with it enabled, and so on...

From what I understand, packet steering is basically an optimisation method to 'steer' processing of network packets to the same CPU that is handling the interrupt that receives the packet. This has the benefit of making it more efficient as the CPU do not have to reload it's caches since it already have the network packet data in it.

For your case, I don't understand how two methods that are supposed to be executed in two different threads are executing one after another, causing the deadlock. My 'kung fu' is not strong enough to figure out why ... haha.

nicoh88 · 2022-02-24T13:01:26.656Z

Hello, with the current versions I have lost connection to the internet after a few hours / days.

The router Netgear R7800 is connected to a Vodafone Cable (Germany) modem in bridge mode. The Vodafone Cable Modem establishes the connection to the Internet and the Netgear router with OpenWRT receives the public IP addresses via DHCP.

It starts out quite easy, with a few drop-outs in the ping, after a few hours there are more and more until in the end hardly a packet arrives.

The problems have been known to me since version R7800-20211202-Stable2012NSS-sysupgrade, the current one also causes problems R7800-20220127-Stable2012NSS-sysupgrade.
The last stable running version for me is R7800-20211010-Stable2012NSS-sysupgrade without any issues.

logread doesn't show me any errors. Also in the top no problem is recognizable.

Do you have an idea? Are there any other log files we could look at?

sppmaster · 2022-02-24T22:40:05.253Z

It's interesting but after using the router for several months with different master versions (even for over a month without reboot) I've never experienced a loss of Internet connection until today. After 5 days of work since last restart tonight I've lost Internet connection. Restarting the WAN interface didn't help. I had to reboot the router to recover the Internet. Nothing in the syslog.

quarky · 2022-02-25T00:10:56.535Z

@nicoh88, @sppmaster Was it that your wireless devices all lost connectivity? Does it also affect your LAN clients? I suspect it is just your wireless clients affected? You may want to try restarting your wireless interfaces.

You may also want to try out @KONG suggestion here.

boldow · 2022-02-25T03:45:52.654Z

I also had such problem with Kong build r18645-900a7ed2a9. the router lost internet after several hours or one or two days. cannot get public IP from my modem.

sppmaster · 2022-02-25T07:26:28.599Z

@quarky Initially I thought it was only for the WLAN part of the network but when I get to the desktop PC connected to the LAN via cable to check what was going on I saw there was no connection to the Internet. I have two set-top boxes for TV connected with cable to the router and they had no connection to the service either.
Till now with ath10k drivers I haven't had any stalls nor other significant issues. I had issues with -ct drivers including disconnection of WLAN clients and really low throughput. Usually I have 1-3 to 7-8 WLAN clients. My setup includes a guest network with WDS and VLAN connection over 2.4GHz band for 20 clients separated from my LAN.

nicoh88 · 2022-02-25T07:29:50.324Z

Hi, my problem affects all devices - not just wireless devices. My devices connected via LAN such as Linux servers, Raspberry Pi, TV and iMac also have this problem.

Even from my root server, in a Hetzner data center, the public IP address of my Internet connection can no longer ping or the ping has timed out.

As said, R7800-20211010-Stable2012NSS-sysupgrade it's been working fine for over a week now! All newer ones, including the current ones, still have these problems for a few hours/days.

quarky · 2022-02-25T07:32:05.359Z

So it looks like it is something introduced in master recently then. I'm running 21.02 for my R7800 and Askey RT4230W, and it's been stable so far for me, except for the Wi-Fi issue I'm facing.

xeonpj · 2022-02-25T08:35:03.741Z

this seems silly, but a chinese wifi camera gave me a lot of problems, it left the network without internet, even to wired equipment. Have you introduced any new devices to the network in the last month? I know it sounds silly, but it is so. regards

sppmaster · 2022-02-25T09:32:07.416Z

In my case nothing new. I was using my regular setup and devices.

ACwifidude · 2022-02-27T02:41:53.737Z

Updated the master build. Master has some updates for the ea7500v1 and ea8500 so I started building for them again. Let me know how they work.

Edit: Added kmod-ramoops to the build. Latest master has .dts changes that support ramoops for the R7800. Will look at the other dts files and hopefully should be easy to add functional ramoops across the board. I haven’t tried it yet and have never used ramoops before. Will have to learn more about it and make sure it is turn key when you load the build (I haven’t done any further configuration beyond master defaults so I’m not sure if ramoops works currently or not).

SiXX · 2022-02-28T00:50:25.481Z

EDIT - disregard... operator error

PerkelSimon · 2022-02-28T02:05:43.148Z

Hi,

I have a problem with my nss build, I've cloned NSS repo with master, usually it is working but tonight I got an error, don't know if you have an idea what to do, here the log

Applying ./patches/subsys/999-mac80211-NSS-support.patch using plaintext: 
patching file local-symbols
patching file net/mac80211/Kconfig
patching file net/mac80211/ieee80211_i.h
patching file net/mac80211/iface.c
Hunk #2 succeeded at 1166 (offset 64 lines).
Hunk #3 succeeded at 1480 (offset 64 lines).
patching file net/mac80211/rx.c
Hunk #4 FAILED at 4538.
1 out of 4 hunks FAILED -- saving rejects to file net/mac80211/rx.c.rej
patching file net/mac80211/tx.c
Hunk #3 succeeded at 4390 (offset -5 lines).
Patch failed!  Please fix ./patches/subsys/999-mac80211-NSS-support.patch!
make[3]: *** [Makefile:573: /run/media/james/D/NSS/build_dir/target-arm_cortex-a15+neon-vfpv4_musl_eabi/linux-ipq806x_generic/backports-5.15.8-1/.prepared_7b0a321e2b51032ec9e121293eedfcae_6664517399ebbbc92a37c5bb081b5c53] Error 1
make[3]: Leaving directory '/run/media/james/D/NSS/package/kernel/mac80211'
time: package/kernel/mac80211/regular/compile#0.81#0.27#0.91
    ERROR: package/kernel/mac80211 failed to build (build variant: regular).
make[2]: *** [package/Makefile:116: package/kernel/mac80211/compile] Error 1

ACwifidude · 2022-02-28T02:47:11.002Z

If you have an older clone you need this fix to the mac80211 patch, fixes the hunk that changed in master:

github.com/ACwifidude/openwrt

ipq806x: NSS Hardware Offloading update mac80211 patch

committed 04:36AM - 16 Feb 22 UTC

ACwifidude

+4 -4

In other news - The 21.02 build has been updated. Changed the ea8500 and ea7500v1 back to what was working with the 5.4 kernel. Let me know how it works.

quarky · 2022-02-28T03:02:47.729Z

The patch for the ieee80211_rx_8023() (that sends the skb to NSS) can be removed, or you will be sending the same skb to NSS twice, once with ieee80211_deliver_skb_to_local_stack() and next just after calling that method. But since we're not receiving de-capped skb from firmware (i.e. de-cap offload) for ath10k, this path will not be called. Will be a problem in future if hardware de-cap offload is implemented.

palibrzuch · 2022-02-28T05:33:15.662Z

Hi. I recently asked for a g10 image. Forgot

castiel652 · 2022-02-28T06:43:08.208Z

quarky:

Will be a problem in future if hardware de-cap offload is implemented.

Currently ath10k doesn't do buffer reordering and de-duplication in driver or firmware both are need for fast-rx and fast-rx is the prerequisite for decap offload.

There were some ancient patches to implement but it's quite complicated to adapt to latest ath10k.

argor · 2022-02-28T07:38:20.171Z

I think the internet connectivity issue is DNS related. None of my devices would have internet access[wired/wireless] but they successfully get ip#'s. The only way to get internet access was for me to manually set DNS on each devices. Dont have much time to mess with things so I switched to Kongs build for now.