Systemd Unit doesn't restart after resume from suspend

[terencode]

I'm on ArchLinux, I installed it from the AUR (https://aur.archlinux.org/packages/amdgpu-clocks-git/) and I enabled the systemd unit which correctly outputs:

Created symlink /etc/systemd/system/multi-user.target.wants/amdgpu-clocks.service → /usr/lib/systemd/system/amdgpu-clocks.service.
Created symlink /etc/systemd/system/sleep.target.wants/amdgpu-clocks.service → /usr/lib/systemd/system/amdgpu-clocks.service.
Created symlink /etc/systemd/system/suspend.target.wants/amdgpu-clocks.service → /usr/lib/systemd/system/amdgpu-clocks.service.
Created symlink /etc/systemd/system/hibernate.target.wants/amdgpu-clocks.service → /usr/lib/systemd/system/amdgpu-clocks.service.

However, after resuming from suspend, the unit is not restarted and thus the different settings not applied.

[sibradzic]

@terencode thanks for report. Can you please try this in your amdgpu-clocks.service file (followed by systemctl daemon-reload) and report back:

[Unit]
Description=Set custom amdgpu clocks & voltages
After=rc-local.service plymouth-start.service systemd-user-sessions.service suspend.target hibernate.target hybrid-sleep.target

[Service]
Type=oneshot
ExecStart=/usr/bin/amdgpu-clocks
RemainAfterExit=yes
StandardOutput=syslog

[Install]
WantedBy=multi-user.target suspend.target hibernate.target hybrid-sleep.target

?

[sibradzic]

https://askubuntu.com/questions/661715/make-a-script-start-after-suspend-in-ubuntu-15-04-systemd

[terencode]

It doesn't restart when resuming because of the RemainAfterExit=true line. You have 3 options: https://github.com/rncbc/rtirq/pull/4#issuecomment-452429118

[sibradzic]

As you can see I've opted for option 3, an additional scriptlet @ /usr/lib/systemd/system-sleep. I wanted to keep RemainAfterExit=true so that the provided service can be stopped, which will now restore the OverDrive states to the initial values (this is something I planned adding long time ago...). Please check if this approach works for you.

[terencode]

Thanks, it works but you should mention in the README that you need to make the script executable (chmod a+x).

Why did you change the path to /usr/local/bin ?

[terencode]

Oh and you need to change/usr/bin/systemctl start amdgpu-clocks to /usr/bin/systemctl restart amdgpu-clocks otherwise it doesn't work.

Also, little typo in your repo description ;)

[sibradzic]

Thanks, it works but you should mention in the README that you need to make the script executable (chmod a+x).

It is already set as executable in the git, no need to do chmod.

Why did you change the path to /usr/local/bin ?

It seems more appropriate that way, but it the end of the day it is just a simple script, feel free to place it anywhere you prefer...

[sibradzic]

Oh and you need to change/usr/bin/systemctl start amdgpu-clocks to /usr/bin/systemctl restart amdgpu-clocks otherwise it doesn't work.

Hmm, are you sure, how does it behave for you? It works just fine for me (Ubuntu 20.04) after suspend with just start.

Also, little typo in your repo description ;)

Fixed, thanks for pointing it out!

[terencode]

It is already set as executable in the git, no need to do chmod.

Ah yeah I needed to chmod it because I mistakenly used install -Dm 644

It seems more appropriate that way, but it the end of the day it is just a simple script, feel free to place it anywhere you prefer... Yeah, I guess, I was wondering from a packaging perspective.

Hmm, are you sure, how does it behave for you? It works just fine for me (Ubuntu 20.04) after suspend with just start.

Yes , the unit is not restarted. You can check this by doing systemctl status amdgpu-clocks.service, make sure it's started and take note of the time when it did, then systemctl start amdgpu-clocks.service and see the time is the same with systemctl status amdgpu-clocks.service again.

[terencode]

To quote the manual (https://www.freedesktop.org/software/systemd/man/systemd.service.html):

Since the unit is considered to be running after the start action has exited, invoking systemctl start on that unit again will cause no action to be taken.

[sibradzic]

Ok, changed to restart. Check it out once more, and feel free to close the issue.

[terencode]

Thank you :)