Black screens after wake from sleep on Fedora 43 KDE

Ask Fedora
,
1

A consistent problem I am having is that after waking my pc from sleep the monitors remain black. The lights and fans on my pc all turn on, the power indicator lights on my monitors turn on, but the screens remain black and I have to reboot.

Kernel version: 6.18.6-200
Graphics Platform: Wayland
GPU: Nvidia RTX 3090
CPU: Intel Core I7-13700K
Motherboard: ASUS Z690-F Strix (recently updated the bios to latest version)
Nvidia Driver Version: 580.119.02

I have tried blacklisting nouveau in the grub config. Also enabling the nvidia sleep service and resume service. I also tried NVreg_PreserveVideoMemoryAllocation=1 in a modprobe nvidia.conf file. No luck.

This is the only issue I have been noticing. Once I am in the desktop it works great, just canā€™t seem to get display to comeback after sleep.

2

We do not know if you have installed the nvidia driver or not. If you have, and did not install it from the rpmfusion repo then problems are usually more prevalent.

Installing those drivers from rpmfusion usually takes care of everything needed to work properly, including blacklisting the nouveau drivers for you.

3

I did install the Nvidia driver from the rpmfuson repo. I am running version 580.119.02. The command ā€œnvidia-smiā€ does show that the driver is loaded and the correct version.

4

Sorry, I missed seeing that in your first post; My error.

What kernel version are you running? The latest is 6.18.7 for fedora.
Sometimes there may be problems with booting and a black screen if a reboot is performed to quickly after an update completes.

What is the output of cat /proc/cmdline and cat /etc/kernel/cmdline

5

I recently updated to 6.18.7 after I made the original post. I restarted a few minutes after the update and was able use it just fine. When I reboot the pc, everything works fine including the displays. The issue only seams to occur when waking from sleep. A strange behavior that Iā€™ve been able to recreate twice now is the displays will comeback after the first time I wake it up after a reboot if I power cycle one of my monitors, but only for the first sleep/wake event cycle. From the 2nd sleep/wake forward, nothing will bring the displays back, only a reboot will bring the displays back up.

6

I seem to recall this has been noted before.
Waking from suspend does not appear to query the displays for their EDID data so they cannot be activated. Without receiving that data the system is not aware that the monitors exist.

I do not use sleep nor suspend since my daily driver is also a 24/7 media server and I only allow the screen to blank and lock but the system never sleeps.

7

Iā€™ve run into something similar. Iā€™m also using an Nvidia card with the latest ā€œstableā€ proprietary drivers, and I think Iā€™ve mostly solved it.

Problem:
My system would fail to wake from sleep about half the time, usually when nothing was open except the Plasma desktop.

What helped:

  1. In System Settings ā†’ Lock Screen, make sure ā€œLock after waking from sleepā€ is enabled.
    Turning this on significantly reduced the issue for me.

  2. I still had occasional wakeā€‘fromā€‘sleep problems, so I also changed another setting:
    Go to System Settings ā†’ Power Management ā†’ Display and Brightness, and set ā€œDim Automaticallyā€ to Never.

Since making those two changes, sleep and wake have been much more reliable. Hope this helps you too!

8

i am facing similar issue in my desktop i have amd cpu and gpu , so i dont think its related to nvidia/amd/intel , itā€™s kernel related or something other , pc waking up after suspend but monitor stays blank , if only i remove power cord of monitor and replug it , then the display comes back.

9

Iā€™m running into this issue as well; I have the akmod-nvidia-580 driver installed from the rpmfusion repo. This problem just started after I switched from using displayport to hdmi, which I did due to a completely different issue. Using the displayport, occasionally my screen would just go black for a brief period of time (~0.5 to 1s duration) before coming back. This seemed to happen more when I was watching videos in full screen mode, which as you can imagine, is incredibly frustrating.

Switching over to HDMI seems to have fixed that issue, but now when my computer wakes from sleep, it I guess forgets that it was using HDMI to talk to the monitor? When the computer wakes, I get a little pulse on my monitor showing that the computer talked to it for a second, but the screen remains black, and after about 5 seconds, the screen goes into powersave mode. I found that if I go into the monitorā€™s settings and switch the input from hdmi1 (which Iā€™m using) to hdmi2 or displayport and then switch it back, the issue resolves itself.

Switching it away from hdmi1 causes the computer to make the hardware disconnect noise, and then switching it back causes the hardware re-connect noise, and all of a sudden it works just fine.

This, also, is incredibly frustrating. Iā€™ve looked in the system settings, and can find no way to force the computer to communicate with the monitor only through the hdmi port. Similarly, there are no options to do so in the nvidia x server settings app. Essentially, the only way I have to ā€œfixā€ the issue is to make the computer never go to sleep, which is no solution at all. That, or switch back to displayport, which will then cause the other incredibly frustrating issue.

10

Ok, I may have fixed my issue. After doing a bunch of poking around online, I found an article talking about a problem similar to mine. In this personā€™s solution, they used xrandr to fix the issue.

I did the following:

sudo xrandr --listactivemonitors

Monitors: 1
 0: +*HDMI-A-1 2560/798x1080/334+0+0  HDMI-A-1

sudo xrandr --output HDMI-A-1 --primary

This seems to have fixed my issue. Iā€™ve slept the computer about a half dozen times now, and every time it wakes up without issue. It could, of course, just be pure luck that itā€™s worked every time, but Iā€™m getting the impression that xrandr fixed my issue.

11

tried this. didnā€™t work for me.
amd cpu and gpu.

12

I guess there could be several root causes for this problem.

Iā€™ve had this issue as well, but in my case itā€™s related to kdwin_wayland (the KDE compositor) not being able to re-acquire DRM master after resume. Root cause unsure yet, but it could have something to do with a slight delay until the correct /dev/cardx is detected (in which case it would be a kdwin_wayland race bug).

Iā€™ve reported a bug; read the details: https://bugs.kde.org/show_bug.cgi?id=516038 Check your journalctl if you see something similar. If that is the case, I have a workaround that seems to be working (for now): set KWIN_DRM_DEVICES=/dev/dri/cardx in /etc/environment (restricting kwin to your main GPU). Donā€™t forget to change x to the number that corresponds with your gpu: lspci | grep -i ā€˜vgaā€™

14

Iā€™ve reported a bug; read the details: https://bugs.kde.org/show_bug.cgi?id=516038 Check your journalctl if you see something similar. If that is the case, I have a workaround that seems to be working (for now): set KWIN_DRM_DEVICES=/dev/dri/cardx in /etc/environment (restricting kwin to your main GPU). Donā€™t forget to change x to the number that corresponds with your gpu: lspci | grep -i ā€˜vgaā€™

Thanks for this! Though, I had to run ls -l /dev/dri/by-path/ after lspci to get the symlink to determine my card.

Side note - this was occurring with my built in AMD Granite Ridge card and I had to dedicate it to my 7900 XTX.