Always detecting ssh.service as service to restart

[micw]

Hi,

I use needrestart on jessie from jessie-backports. I have the problem that ssh.service is always detected, e.g.:
[main] #9761 exe => /usr/bin/lxc-start
[main] #9761 is ssh.service

[main] #4166 exe => /usr/lib/jvm/java-8-oracle/bin/java
[Core] #4166 is a NeedRestart::Interp::Java
[Core] #4166 source is UNKNOWN
[main] #4166 is ssh.service

[liske]

Hi,

does the system runs with systemd? Could you please provide the content of /proc/9761/cgroup? Could you give needrestart 2.10 from stretch a try?

[Korni22]

Hi,

here is the output from my machine, since I have the same problem, this might help you 😊

[main] #2491 exe => /usr/bin/screen-4.4.0
[main] #2491 is sshd.service
[main] #2492 exe => /usr/bin/zsh
[main] #2492 is sshd.service
[main] #3776 exe => /usr/bin/screen-4.4.0
[main] #3776 is sshd.service
[main] #3777 exe => /usr/bin/zsh
[main] #3777 is sshd.service
[main] #10556 exe => /usr/bin/screen-4.4.0
[main] #10556 is sshd.service
[main] #10562 exe => /usr/bin/bash
[main] #10562 is sshd.service
[main] #29337 exe => /usr/bin/screen-4.4.0
[main] #29337 is sshd.service
n
user@host ~ $ cat /proc/2491/cgroup
9:net_cls:/
8:cpuset:/
7:blkio:/
6:devices:/system.slice/sshd.service
5:cpu,cpuacct:/
4:memory:/
3:freezer:/
2:pids:/system.slice/sshd.service
1:name=systemd:/system.slice/sshd.service
user@host ~ $ cat /proc/2492/cgroup
9:net_cls:/
8:cpuset:/
7:blkio:/
6:devices:/system.slice/sshd.service
5:cpu,cpuacct:/
4:memory:/
3:freezer:/
2:pids:/system.slice/sshd.service
1:name=systemd:/system.slice/sshd.service
user@host ~ $ cat /proc/3776/cgroup
9:net_cls:/
8:cpuset:/
7:blkio:/
6:devices:/system.slice/sshd.service
5:cpu,cpuacct:/
4:memory:/
3:freezer:/
2:pids:/system.slice/sshd.service
1:name=systemd:/system.slice/sshd.service
user@host ~ $ cat /proc/3777/cgroup
9:net_cls:/
8:cpuset:/
7:blkio:/
6:devices:/system.slice/sshd.service
5:cpu,cpuacct:/
4:memory:/
3:freezer:/
2:pids:/system.slice/sshd.service
1:name=systemd:/system.slice/sshd.service
user@host ~ $ cat /proc/10556/cgroup
9:net_cls:/
8:cpuset:/
7:blkio:/
6:devices:/system.slice/sshd.service
5:cpu,cpuacct:/
4:memory:/
3:freezer:/
2:pids:/system.slice/sshd.service
1:name=systemd:/system.slice/sshd.service
user@host ~ $ cat /proc/10562/cgroup
9:net_cls:/
8:cpuset:/
7:blkio:/
6:devices:/system.slice/sshd.service
5:cpu,cpuacct:/
4:memory:/
3:freezer:/
2:pids:/system.slice/sshd.service
1:name=systemd:/system.slice/sshd.service
user@host ~ $ cat /proc/29337/cgroup
9:net_cls:/
8:cpuset:/
7:blkio:/
6:devices:/system.slice/sshd.service
5:cpu,cpuacct:/
4:memory:/
3:freezer:/
2:pids:/system.slice/sshd.service
1:name=systemd:/system.slice/sshd.service

I installed needrestart from the AUR, which seems to be version 2.10-1, but needrestart --version tells me its needrestart 2.9 - Restart daemons after library updates. 🤔

[liske]

@Korni22: do you use systemd's PAM module (libpam-systemd on Debian)? The content of the cgroup files shows that your screen instances are part of the sshd.service. If the systemd PAM module is used the user's processes should be part of a user session:

$ ssh localhost
$ cat /proc/$$/cgroup
10:freezer:/
9:cpuset:/
8:devices:/user.slice/user-1000.slice
7:perf_event:/
6:net_cls,net_prio:/
5:pids:/user.slice/user-1000.slice/session-309219.scope
4:memory:/user.slice/user-1000.slice
3:blkio:/user.slice/user-1000.slice
2:cpu,cpuacct:/user.slice/user-1000.slice
1:name=systemd:/user.slice/user-1000.slice/session-309219.scope

Without the systemd PAM module needrestart is unable to detect user sessions.

PS: There was a bug showing the wrong version number in the needrestart 2.10 release.

[Korni22]

@liske it seems that Arch Linux does not use libpam-systemd, so I'm out of luck, but thanks for your help!