FS#62261 - [linux] Kernel panics when attempting to wake from suspend for linux 4.20 and above

Attached to Project: Arch Linux
Opened by modula t. (defaultxr) - Sunday, 07 April 2019, 01:23 GMT
Last edited by Andreas Radke (AndyRTR) - Tuesday, 01 March 2022, 21:32 GMT
Task Type Bug Report
Category Kernel
Status Closed
Assigned To Jan Alexander Steffens (heftig)
Architecture x86_64
Severity High
Priority Normal
Reported Version
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Votes 3
Private No

Details

Description:

Kernel appears to panic when resuming from suspend on any version of linux above 4.19.12. linux-4.19.12 suspends and resumes fine, but linux-4.20 or above seemed to panic, as the caps lock light blinked and nothing appears on the screen. Since panic happens before the screen turns on, I have no way to record the kernel panic message as far as I am aware because it doesn't appear on the screen. I've tried using netconsole as well and no messages are sent to it either (messages are sent fine prior to suspending, but none are sent when resuming is attempted). I've tried using the latest kernel (5.0.6) in hopes that the problem would be fixed but it still occurs with that version. The only difference in the latest kernel is that instead of the caps lock blinking, it now seems to turn on and stay on when resume is attempted.


Additional info:
* package version(s):

linux-4.19.12: suspend/resume works normally.
linux-4.20.0 and above: suspend seems to work; resuming results in kernel panic and screen stays black.

* config and/or log files etc.
no special config files.

hardware (in case it's relevant):

Laptop; MSI GT72S-6QF
graphics card: Nvidia GTX980
using Nvidia proprietary drivers

* link to upstream bug report, if any:
none

Steps to reproduce:

suspend system:
# systemctl suspend
resume by opening laptop lid.

Here are the last lines from "journalctl -b -1" when suspending and resuming using latest (5.0.6) kernel:

---

Apr 06 19:51:55 hypermulti polkitd[2003]: Registered Authentication Agent for unix-process:1999:2092 (system bus name :1.46 [/usr/bin/pkttyagent --notify-fd 5 --fallback], object path /org/freedesktop/PolicyKit1/AuthenticationAgent, locale en_US.UTF-8)
Apr 06 19:51:55 hypermulti systemd[1]: Created slice system-resume.slice.
Apr 06 19:51:55 hypermulti polkitd[2003]: Unregistered Authentication Agent for unix-process:1999:2092 (system bus name :1.46, object path /org/freedesktop/PolicyKit1/AuthenticationAgent, locale en_US.UTF-8) (disconnected from bus)
Apr 06 19:51:55 hypermulti systemd[1]: Starting TLP suspend/resume...
Apr 06 19:51:55 hypermulti systemd[1]: Started TLP suspend/resume.
Apr 06 19:51:55 hypermulti audit[1]: SERVICE_START pid=1 uid=0 auid=4294967295 ses=4294967295 msg='unit=tlp-sleep comm="systemd" exe="/usr/lib/systemd/systemd" hostname=? addr=? terminal=? res=success'
Apr 06 19:51:55 hypermulti systemd[1]: Reached target Sleep.
Apr 06 19:51:55 hypermulti systemd[1]: Starting Suspend...
Apr 06 19:51:55 hypermulti kernel: audit: type=1130 audit(1554598315.533:40): pid=1 uid=0 auid=4294967295 ses=4294967295 msg='unit=tlp-sleep comm="systemd" exe="/usr/lib/systemd/systemd" hostname=? addr=? terminal=? res=success'
Apr 06 19:51:55 hypermulti systemd-sleep[2152]: Suspending system...
Apr 06 19:51:55 hypermulti kernel: PM: suspend entry (deep)

---

As you can see, nothing seems to be written after the suspend entries. No resume events are logged. The system simply stops responding other than the caps lock lighting up. Can't SSH to the machine or anything, I have to hold the power switch to power it down and reboot.

Since suspending is critical for me, I've been using the 4.19.12 kernel, but some programs such as steam won't start unless they're used with the most recent drivers and kernel, so I'm hoping there is some way to get the suspend/resume behavior reverted back to how it worked pre linux-4.20. If there is anything I can do to provide more information, please let me know.
This task depends upon

Closed by  Andreas Radke (AndyRTR)
Tuesday, 01 March 2022, 21:32 GMT
Reason for closing:  None
Additional comments about closing:  No response in years. Assuming fixed upstream.
Comment by t-ask (tAsk) - Thursday, 02 May 2019, 11:46 GMT
I have the same for weeks now. Did you try downgrading systemd to eg.

local/systemd 241.7-1
local/lib32-systemd 241.7-1
local/systemd-libs 241.7-1
local/systemd-sysvcompat 241.7-1

or just any other lower version?
Comment by modula t. (defaultxr) - Friday, 21 June 2019, 23:56 GMT
Unfortunately this problem persists even with the latest kernel (5.1.12.arch1-1). It also seems to occur with linux-lts (4.19.53-1).

@t-ask Thanks for the suggestion. I tried those versions of those libraries with both linux 5.1.12.arch1-1 and linux-lts 4.19.53-1 with the same issue. I also tried systemd and libs 240.34-3 with the latest main and lts kernels and still had the same issue. It seems like anything above linux 4.19.12-arch1-1 and systemd 240.34-3 will cause this issue for me.
Comment by loqs (loqs) - Saturday, 22 June 2019, 16:47 GMT
Can you please try to find the first linux-lts release affected from https://archive.archlinux.org/packages/l/linux-lts/
Edit:
With linux 4.19.12 does updating systemd to 241 also trigger the issue?
Comment by mattia (nTia89) - Tuesday, 08 December 2020, 20:34 GMT
HI @modula t.
I experienced this kind of issue since years;
I do not remember exact versions but starting from a certain release, suspend & resume stopped working;
After years, now (8th-Dec-2020, kernel version 5.9.9), it has been fixed;
Suspend and resume is a minefield e.g. this [https://bugs.archlinux.org/task/68762] is only the latest issue related;

Anyway, current kernel versions are fixed. Is the same for you?

Loading...