systemd-logind crash when suspend with nvidia-suspend.service masked, bringing session down with it

[Ratchanan Srirattanamet]

Steps to reproduce:
1. Run `sudo apt install nvidia-driver-465`. Then, without a need for a restart, run `sudo apt autoremove nvidia-driver-465`.
2. Leaving the terminal open, suspend the machine.

Expected behavior: the machine suspend successfully and can be woken up successfully.
Actual behavior: the machine doesn't suspend. It either:
- hang with no respond other than SysRq+REISUB (or maybe network, but I didn't test), or
- return you back to the login screen. Upon logging in, you'll notice that the terminal you opened is gone.

Upon further inspection (on a session that doesn't hang), it's been found that X server died with:

Fatal server error:
[ 66.422] (EE) systemd-logind disappeared (stopped/restarted?)

And checking journal for systemd-logind log, it said:

Error during inhibitor-delayed operation (already returned success to client): Unit nvidia-suspend.service is masked.

before the new process takes it place.

The system is Ubuntu 20.04, X.org session.

ProblemType: Bug
DistroRelease: Ubuntu 20.04
Package: systemd 245.4-4ubuntu3.7
ProcVersionSignature: Ubuntu 5.11.0-22.23~20.04.1-generic 5.11.21
Uname: Linux 5.11.0-22-generic x86_64
ApportVersion: 2.20.11-0ubuntu27.18
Architecture: amd64
CasperMD5CheckResult: skip
CurrentDesktop: ubuntu:GNOME
Date: Tue Jun 29 03:34:00 2021
InstallationDate: Installed on 2021-03-15 (105 days ago)
InstallationMedia: Ubuntu 20.04.2.0 LTS "Focal Fossa" - Release amd64 (20210209.1)
MachineType: LENOVO 82B5
ProcEnviron:
 TERM=xterm-256color
 PATH=(custom, no user)
 XDG_RUNTIME_DIR=<set>
 LANG=th_TH.UTF-8
 SHELL=/bin/bash
ProcKernelCmdLine: BOOT_IMAGE=/boot/vmlinuz-5.11.0-22-generic root=UUID=06f2a676-a62c-443a-8bc8-4e0eda4600f4 ro log_buf_len=2M quiet splash vt.handoff=7
SourcePackage: systemd
UpgradeStatus: No upgrade log present (probably fresh install)
dmi.bios.date: 01/01/2021
dmi.bios.release: 1.31
dmi.bios.vendor: LENOVO
dmi.bios.version: EUCN31WW
dmi.board.asset.tag: NO Asset Tag
dmi.board.name: LNVNB161216
dmi.board.vendor: LENOVO
dmi.board.version: SDK0Q55756 WIN
dmi.chassis.asset.tag: NO Asset Tag
dmi.chassis.type: 10
dmi.chassis.vendor: LENOVO
dmi.chassis.version: Lenovo Legion 5 15ARH05
dmi.ec.firmware.release: 1.31
dmi.modalias: dmi:bvnLENOVO:bvrEUCN31WW:bd01/01/2021:br1.31:efr1.31:svnLENOVO:pn82B5:pvrLenovoLegion515ARH05:rvnLENOVO:rnLNVNB161216:rvrSDK0Q55756WIN:cvnLENOVO:ct10:cvrLenovoLegion515ARH05:
dmi.product.family: Legion 5 15ARH05
dmi.product.name: 82B5
dmi.product.sku: LENOVO_MT_82B5_BU_idea_FM_Legion 5 15ARH05
dmi.product.version: Lenovo Legion 5 15ARH05
dmi.sys.vendor: LENOVO

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[Alex Boisvert (alex-boisvert)]

Not sure if it's going to help anybody else but I fixed my problem by correcting broken symlinks with nvidia-suspend.service. It was pointing to a missing file/location (/lib/systemd/system/nvidia-suspend.service). I simply created empty files to match the broken links.

[apienk (andrzej-pienkowski)]

Alex, thanks a lot. Your workaround fixed my suspend problems.

[soccer193 (soccer193)]

A workaround of my own was to correct the nvidia driver's behaviour.

Running Ubuntu 21.04 and nvidia-drivers-460, I upgraded to nvidia-drivers-470 (which I had installed once before), resulting in no masked services, and files paths to real .service files.

Before:
❯ systemctl status nvidia-suspend nvidia-hibernate nvidia-resume
● nvidia-suspend.service
     Loaded: masked (Reason: Unit nvidia-suspend.service is masked.)
     Active: inactive (dead)

● nvidia-hibernate.service
     Loaded: masked (Reason: Unit nvidia-hibernate.service is masked.)
     Active: inactive (dead)

● nvidia-resume.service
     Loaded: masked (Reason: Unit nvidia-resume.service is masked.)
     Active: inactive (dead)

After:
systemctl status nvidia-suspend nvidia-hibernate nvidia-resume
● nvidia-suspend.service - NVIDIA system suspend actions
     Loaded: loaded (/lib/systemd/system/nvidia-suspend.service; enabled; vendor preset: enabled)
     Active: inactive (dead)

● nvidia-hibernate.service - NVIDIA system hibernate actions
     Loaded: loaded (/lib/systemd/system/nvidia-hibernate.service; enabled; vendor preset: enabled)
     Active: inactive (dead)

● nvidia-resume.service - NVIDIA system resume actions
     Loaded: loaded (/lib/systemd/system/nvidia-resume.service; enabled; vendor preset: enabled)
     Active: inactive (dead)

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[Markus (markonetri)]

Confirming this problem.
The work-around does not work for me.

[apienk (andrzej-pienkowski)]

For me, with nvidia-driver-495, the simple solution was to remove the damaged symlinks from systemd. You most likely have them if you upgraded from nvidia-driver-470 or nvidia-driver-465, because 470 still included the .service files in /lib/systemd/system/. The files are no longer included in 495 but the postinst script does not remove the symlinks. So, remove them with:

sudo rm /etc/systemd/system/systemd-hibernate.service.requires/nvidia-resume.service
sudo rm /etc/systemd/system/systemd-hibernate.service.requires/nvidia-hibernate.service
sudo rm /etc/systemd/system/systemd-suspend.service.requires/nvidia-resume.service
sudo rm /etc/systemd/system/systemd-suspend.service.requires/nvidia-suspend.service

Works instantly, no need of rebooting or logging off.

According to user punyidea at https://gist.github.com/bmcbm/375f14eaa17f88756b4bdbbebbcfd029, the services are not needed on 470 either and may be safely removed.

Hope that will fix your problems as well.

[Shuhao (shuhao)]

Can confirm that removing the .service files in /etc/systemd/system/systemd-suspend.service.requires works.

The reason this occurred to me is because i switched GPU from nvidia to AMD, and some programs thought I have nvenc (because the library is installed) when i don't, which necessitates the removal of the nvidia-driver packages.

This seem like an issue with one of the nvidia package's postrm debian package scripts as opposed to systemd, tho.

[Luigi Calligaris (luigicalligaris)]

I confirm that removing the .service files described above solves the issue for me as well.

[Rocio Platini (ubuntu-lover)]

I can confirm that solution of apienk worked for me. My pc doesn't freeze any more after suspend.

[Nick Rosbrook (enr0n)]

Based on the workaround, it seems that the problem is with nvidia systemd units, and not systemd itself.