RT-AC86U cfg_server - what is it?

[mymatenige]

Hi,

I notice that my RT-AC86Us make regular attempts at connecting to other machines on my network on UDP port 7788. This appears to be to locate other ASUS routers running cfg_server. Simple question ... what is cfg_server?

Thanks.

 

[RMerlin]

Used by AiMesh to communicate between nodes.

 

[Cam]

Used by AiMesh to communicate between nodes.

Sorry to hijack this thread, but every now and then i notice I have multiple cfg_server processes running (at least 3 times over past couple of weeks).

~$ pidof cfg_server | wc -w
255

At the moment I simply issue 'killall cfg_server' and then some watchdog restarts it with 3 pid's.

I don't use AiMesh, and can't see any reference to cfg_server in syslog. It's not a big issue obviously, but is there a way I track down what is causing it to constantly spawn in my system (RT-AC68U)?

 

[RMerlin]

It might possibly be running by default, so it will be ready to handle the addition of nodes. I don't recommend messing with it, just leave it alone.

 

[Cam]

Yeah well it's not eating memory so I'm not fussed, but 255 instances seems a lot ! Thanks.

Sent from my Nokia 7 plus using Tapatalk

 

[AndreiGuru]

Ran into the same "cfg_server" processes here on my RT-AC5300 using fw 384.19 - over 200 processes found which were causing wifi stability issues. There's definitely a bug with the watchdog that restarts them. I don't use AI Mesh either, but if I login to the webui, I can see the procs climb after a killall. What's the right place to report the bug to? I'll get some details together in the meantime.

 

[AndreiGuru]

Apparently 'cfg_server' is referenced all over the place in numerous scripts.. I setup a small script that runs on boot now:

admin@RT-AC5300-63B0:/tmp/home/root# cat /jffs/bin/zomg_cfg_server
#!/bin/sh
while true;do
  procs=$(/bin/pidof cfg_server|/usr/bin/wc -w)
  if [ "$procs" -gt "3" ];then /usr/bin/killall -9 cfg_server;/opt/bin/curl -k -m 10 -so/dev/null https://thing.com/?wtfrouter=${procs};else /opt/bin/curl -k -m 10 -so/dev/null https://thing.com/?chkrouter=${procs};fi
  sleep 10
done
admin@RT-AC5300-63B0:/tmp/home/root#

and the log entries show repeated spikes, one even crossing 1100 cfg_server processes!

# grep -P 'router\=[0-9]{2,5}' /var/log/nginx/thing.access_log ..
[05/Oct/2020:11:08:13 -0400] "GET /?wtfrouter=199 HTTP/1.1" 200 1312 "-" "curl/7.69.0"
[05/Oct/2020:15:06:39 -0400] "GET /?wtfrouter=98 HTTP/1.1" 200 1312 "-" "curl/7.69.0"
[05/Oct/2020:16:58:17 -0400] "GET /?wtfrouter=646 HTTP/1.1" 200 1312 "-" "curl/7.69.0"
[05/Oct/2020:16:58:51 -0400] "GET /?wtfrouter=1113 HTTP/1.1" 200 1312 "-" "curl/7.69.0"
[05/Oct/2020:18:06:45 -0400] "GET /?wtfrouter=189 HTTP/1.1" 200 1312 "-" "curl/7.69.0"
[05/Oct/2020:18:07:19 -0400] "GET /?wtfrouter=145 HTTP/1.1" 200 1312 "-" "curl/7.69.0"
[05/Oct/2020:18:08:26 -0400] "GET /?wtfrouter=502 HTTP/1.1" 200 1312 "-" "curl/7.69.0"
[05/Oct/2020:18:08:50 -0400] "GET /?wtfrouter=444 HTTP/1.1" 200 1312 "-" "curl/7.69.0"
[07/Oct/2020:15:58:30 -0400] "GET /?wtfrouter=183 HTTP/1.1" 200 1312 "-" "curl/7.69.0"

Is there something similar to 'stap'/SystemTap available that can be used to debug what's launching these?

 

[AndreiGuru]

bump - these large processes spikes seem to cause intermittent wifi drops on my end. I have no use for AiMesh, how can I disable?

 

[RMerlin]

Don't mess with it, it's part of the system. If you kill it, it will just respawn.

 

[AndreiGuru]

Don't mess with it, it's part of the system. If you kill it, it will just respawn.

You keep saying that, but that makes little to no sense for someone like me who works in devops. AiMesh service has to have a bug if I see it spawning hundreds of cfg_server processes. More than likely, it doesn't check for a lock file to know that it's already running. Pretty clear either way that it's failing to work as intended, yet nothing is said aside from "don't mess with it". It literally makes wifi unstable, how is that not an issue to be considered or reported upstream?

 

[AndreiGuru]

Look at the proclist in the attachment for example

ASUSWRT-Merlin RT-AC5300 384.19_0 Fri Aug 14 19:18:44 UTC 2020

AiMesh isn't even in use/has ever been in use.

 

[Scotovod]

Same issue with cfg_server on AC-5300
mutiple processes lock down local network and dnsmasq on router

im use suggestion from @AndreiGuru `killall -9 cfg_server`
and now router work again as normal

 

[Meshkoff]

Hi there
One time i decide to scan myself using nmap with router WAN ip and discover cfg_server has an open port 7788 (TCP)
Any thoughts why AiMesh functionality needs WAN open port?

 

[RMerlin]

Any thoughts why AiMesh functionality needs WAN open port?

It's not open on the WAN. Only on the LAN.

Scanning the WAN IP from inside the LAN is the same as scanning the LAN. You need to scan from outside to truly scan the WAN interface.

 

[jonnyc55]

Was there ever a command that stopped this issue with the CFG server broadcasting to UDP 7788 ? I see andrei mentions issues even after "`killall -9 cfg_server": "and the log entries show repeated spikes, one even crossing 1100 cfg_server processes!"

 

[ColinTaylor]

Was there ever a command that stopped this issue with the CFG server broadcasting to UDP 7788 ? I see andrei mentions issues even after "`killall -9 cfg_server": "and the log entries show repeated spikes, one even crossing 1100 cfg_server processes!"

The problem of hundreds of instances of this process running at the same time was a bug. AFAIK it was fixed long ago.

 

[jonnyc55]

The problem of hundreds of instances of this process running at the same time was a bug. AFAIK it was fixed long ago.

Ah thanks Colin.

Do you reckon 'killall cfg_server' or 'killall -9 cfg_server' will work now then? Or which one would I use? Thanks.

 

[ColinTaylor]

Do you reckon 'killall cfg_server' or 'killall -9 cfg_server' will work now then? Or which one would I use? Thanks.

Killing those processes was a workaround for the bug. Doing that now on a working system serves no purpose as the router will just restart it.