Client Shutdown

[echolimazulu]

Hello Adrien,

Today I encountered glorytun client crashing twice. Periodically, the glorytun client process exits itself. Is it possible to track the reason for the closure of the process and what could be the reason for what is happening? The client is launched as a regular user with setcap cap_net_admin = ep.

Startup command: glorytun bind 0.0.0.0 to $SERVER_IP keyfile gt.key &

Similar issue: https://github.com/angt/glorytun/issues/61 But I am not using systemd yet. OS Version: Debian 10 (kernel-4.19.126)

[angt]

Hi, Do you use glorytun in OverTheBox/openMPTCProuter ? Are you using the last version 0.3.4 ?

[echolimazulu]

Hi, Do you use glorytun in OverTheBox/openMPTCProuter ? Are you using the last version 0.3.4 ?

No, i use clean Debian 10 installation with implemented kernel 4.19.126-mptcp and net.mptcp.mptcp_enabled=0. Version of Glorytun is 0.3.4 binary build from releases on github.

[echolimazulu]

The strange thing is that it works for several hours (some time), and then it closes itself for no reason.

[angt]

Do you have a coredump or any information ? The only way to stop the client without printing in stderr is to send it a signal. Now we need to know what signal it was..

[echolimazulu]

At the moment, there is no file with information. But if you tell me how to assemble it, I will definitely do it as soon as possible.

[angt]

With the exception of signals that cannot be ignored, we stop the client on: SIGHUP, SIGTERM, SIGQUIT, SIGINT.

[angt]

do you have coredumpctl or something similar ?

[angt]

On debian, I think you can run the systemd-coredump service. Next time, just run coredumpctl list If you see nothing about glorytun, then you need track the process that sends signal to it.

[echolimazulu]

For now, I have installed the systemd-coredump package. Is it enough to just install the package, or is there something else to do while the command prints:

sudo coredumpctl list No coredumps found.

[angt]

This is too late, you need to start/enable the service systemd-coredump. Next time you see glorytun going down you can coredumpctl. If there is an entry for glorytun then you can send it to me by email so that I can find the problem. But, there is no memory allocation while running and the 0.3.4 binary from github is compiled with -fstack-protector-strong, so in the case of a stack overflow there is a (very) good chance that a message will follow the bug.

[echolimazulu]

Maybe we have a preference for kernel.core_pattern? Or startup options for systemd-coredump?

[echolimazulu]

sudo service systemd-coredump@.service status ● systemd-coredump@.service.service - Process Core Dump Loaded: loaded (/lib/systemd/system/systemd-coredump@.service; static; vendor preset: enabled) Active: inactive (dead) Docs: man:systemd-coredump(8)

Oct 29 10:18:23 superrouter systemd[1]: Started Process Core Dump. Oct 29 10:18:23 superrouter systemd-coredump[4361]: Not enough arguments passed by the kernel (0, expected 7).

[angt]

I have no idea about debian, I've never used this OS. But I would expect that if you install and start something it works out of the box.. But yes, maybe you need to tweak debian's systcl about cordeump to make it works ?

[echolimazulu]

Yes, after installing a separate package. The service apparently requires it to be configured in a certain way before starting to be monitored. Would you recommend the startup options for the service systemd-coredump? What OS do you prefer to use for glorytun in which it is most stable?

[angt]

I use archlinux personally but I don't necessarily recommend it because every OS has its advantages and disadvantages. However I can recommend OpenSUSE simply because it is the only distribution that had the will and took the time to distribute glorytun by itself :)

[angt]

If you don't see anything in your logs (dmesg/journald/etc) and no coredump-like files, most likely another process has sent a signal to glorytun.

[angt]

I'm closing, but don't hesitate to reopen with a coredump if the problem comes back.