[SOLVED] nvidia produces black screen at sddm login screen

haigioli · 2022-10-12 20:40:31

On my new install of Arch on a brand new system, I'm able to install everything in UEFI mode (confirmed with all the vars showing up in /sys/firmware/efi/efivars/) and boot up in UEFI mode. However, I get a black screen whenever I start sddm or startx. Journald produces the following error regarding nvidia:

Oct 12 12:57:09 kernel: nvidia: loading out-of-tree module taints kernel.
Oct 12 12:57:09 kernel: nvidia: module license 'NVIDIA' taints kernel.
Oct 12 12:57:09 kernel: nvidia: module verification failed: signature and/or required key missing - tainting kernel
Oct 12 12:57:10 kernel: nvidia-nvlink: Nvlink Core is being initialized, major device number 236
Oct 12 12:57:10 kernel: nvidia 0000:09:00.0: vgaarb: changed VGA decodes: olddecodes=io+mem,decodes=none:owns=none
Oct 12 12:57:10 systemd-udevd[476]: nvidia: Process '/usr/bin/bash -c '/usr/bin/mknod -Z -m 666 /dev/nvidiactl c $(grep nvidia-frontend /proc/devices | cut -d \  -f 1) 255'' failed with exit code 1.
Oct 12 12:57:10 systemd-udevd[457]: nvidia: Process '/usr/bin/bash -c 'for i in $(cat /proc/driver/nvidia/gpus/*/information | grep Minor | cut -d \  -f 4); do /usr/bin/mknod -Z -m 666 /dev/nvidia${i} c $(grep nvidia-frontend /proc/devices | cut -d \  -f 1) ${i}; done'' failed with exit code 1.
Oct 12 12:57:10 kernel: nvidia_uvm: module uses symbols nvUvmInterfaceDisableAccessCntr from proprietary module nvidia, inheriting taint.
Oct 12 12:57:10 kernel: nvidia-modeset: Loading NVIDIA Kernel Mode Setting Driver for UNIX platforms  515.76  Mon Sep 12 19:11:54 UTC 2022
Oct 12 12:57:10 systemd-modules-load[410]: Inserted module 'nvidia_uvm'
Oct 12 12:57:10 kernel: [drm] [nvidia-drm] [GPU ID 0x00000900] Loading driver
Oct 12 12:57:10 kernel: [drm] Initialized nvidia-drm 0.0.0 20160202 for 0000:09:00.0 on minor 0
Oct 12 12:57:10 kernel: nvidia-uvm: Loaded the UVM driver, major device number 234.
Oct 12 12:57:28 kernel: nvidia-modeset: WARNING: GPU:0: Lost display notification (0:0x00000000); continuing.

The last entry of nvidia-modeset: WARNING: GPU:0: Lost display notification appears only with CSM disabled in BIOS. The kernel taint messages appear regardless of whether CSM is enabled or disabled. Here's the journald output on startx/sddm with CSM disabled:

Oct 12 12:58:25 kernel: nvidia: loading out-of-tree module taints kernel.
Oct 12 12:58:25 kernel: nvidia: module license 'NVIDIA' taints kernel.
Oct 12 12:58:25 kernel: nvidia: module verification failed: signature and/or required key missing - tainting kernel
Oct 12 12:58:26 kernel: nvidia-nvlink: Nvlink Core is being initialized, major device number 236
Oct 12 12:58:26 kernel: nvidia 0000:09:00.0: vgaarb: changed VGA decodes: olddecodes=io+mem,decodes=none:owns=io+mem
Oct 12 12:58:26 systemd-udevd[492]: nvidia: Process '/usr/bin/bash -c '/usr/bin/mknod -Z -m 666 /dev/nvidiactl c $(grep nvidia-frontend /proc/devices | cut -d \  -f 1) 255'' failed with exit code 1.
Oct 12 12:58:26 systemd-udevd[492]: nvidia: Process '/usr/bin/bash -c 'for i in $(cat /proc/driver/nvidia/gpus/*/information | grep Minor | cut -d \  -f 4); do /usr/bin/mknod -Z -m 666 /dev/nvidia${i} c $(grep nvidia-frontend /proc/devices | cut -d \  -f 1) ${i}; done'' failed with exit code 1.
Oct 12 12:58:26 kernel: nvidia_uvm: module uses symbols nvUvmInterfaceDisableAccessCntr from proprietary module nvidia, inheriting taint.
Oct 12 12:58:26 kernel: nvidia-modeset: Loading NVIDIA Kernel Mode Setting Driver for UNIX platforms  515.76  Mon Sep 12 19:11:54 UTC 2022
Oct 12 12:58:26 systemd-modules-load[412]: Inserted module 'nvidia_uvm'
Oct 12 12:58:26 kernel: [drm] [nvidia-drm] [GPU ID 0x00000900] Loading driver
Oct 12 12:58:26 kernel: [drm] Initialized nvidia-drm 0.0.0 20160202 for 0000:09:00.0 on minor 0
Oct 12 12:58:26 kernel: nvidia-uvm: Loaded the UVM driver, major device number 234.
Oct 12 12:59:39 dbus-daemon[834]: [session uid=1000 pid=834] Activating via systemd: service name='org.a11y.Bus' unit='at-spi-dbus-bus.service' requested by ':1.42' (uid=1000 pid=1316 comm="/usr/bin/nvidia-settings")
Oct 12 12:59:39 nvidia-settings[1316]: g_object_unref: assertion 'G_IS_OBJECT (object)' failed
Oct 12 12:59:39 at-spi-bus-launcher[1324]: dbus-daemon[1324]: Activating service name='org.a11y.atspi.Registry' requested by ':1.0' (uid=1000 pid=1316 comm="/usr/bin/nvidia-settings")
Oct 12 13:12:04 nvidia-settings[3136]: g_object_unref: assertion 'G_IS_OBJECT (object)' failed

CSM enabled on the motherboard produces no black screen. With CSM enabled I also have 'UEFI Only' turned on for all options underneath, ensuring that everything starts in UEFI mode. For the secure boot section of my BIOS, I have it set to 'Other OS', which according to the latest info out there and from past experience with ASUS motherboards, is the same as turning Secure Boot off. In either case, I've deleted all the keys, just in case. The resulting black screen is the same whether keys are generated or not.

Relevant hardware is as follows:

Motherboard: ASUS TUF X570-Pro
CPU: AMD Ryzen 9 5900X
RAM: DDR4 64GB
GPU: ASUS ROG Strix Nvidia RTX3090

I've tried adding nvidia-drm.modeset=1 to my kernel parameters with no luck. On the off chance the this is related to another bug, I added ibt=off, which still didn't help. The hint appears to be in that it work with CSM enabled, but not when disabled, even though my Arch install is clearly UEFI from install media to local install.

Edit: I should also mention that I've tried launching sddm/startx with the Nouveau driver with CSM disabled and it works fine. So, clearly, this an nvidia/kernel issue, I gather.

Last edited by haigioli (2022-10-13 22:53:44)

seth · 2022-10-12 21:01:44

https://bbs.archlinux.org/viewtopic.php?id=280109
https://bbs.archlinux.org/viewtopic.php … 0#p2060090

haigioli · 2022-10-12 21:13:25

Thanks for the reply @seth.

I read this post, as well as others, at length before posting this one. What's curious about my situation is that is works with CSM enabled, but produced a black screen (on all tty's, too) with CSM disabled. Is it indeed the kernel?

I'll gladly post any other logs that anyone might find relevant to help solve this issue.

seth · 2022-10-12 21:24:49

Hard to tell w/ofull comparative journals, but probably efifb?
Did you try to downgrade the nvidia driver?

haigioli · 2022-10-12 21:47:08

This is a fresh install of Arch. Afraid I don't have a cached nvidia driver to downgrade to. Is there a safe way, otherwise?

Also, I don't know anything about efifb (I'm assuming the 'fb' stands for 'frame buffer'?). Is there something I could do there?

I updated my initial post to include the fact that Nouveau with CSM disabled doesn't produce this issue. Perhaps the Nvidia driver is the problem then?

Last edited by haigioli (2022-10-12 21:50:47)

seth · 2022-10-12 21:52:39

https://wiki.archlinux.org/title/Downgrading_packages
For nvidia you want to switch to nvidia-dkms, otherwise you'll have to downgrade the kernel as well.

Is there something I could do there?

https://bugs.archlinux.org/task/75995#comment211646 seems a workaround for that specific bug.

Edit:

Perhaps the Nvidia driver is the problem then?

Yes. That's why you'd need to downgrade it.

Last edited by seth (2022-10-12 21:53:37)

haigioli · 2022-10-12 22:21:03

@seth I downgraded the nvidia-dkms driver (I always use that one), as well as nvidia-utils the the previous working version and it now works with CSM disabled. Thanks for the lead and quick responses!

The question is whether I should add my voice to the bug report link you sent me for Arch, or to nvidia? This needs to get resolved. It's pretty serious.

Also, should I mark this as solved? I don't know if a driver freeze is a solution. What do you think?

haigioli · 2022-10-13 01:46:28

Looks like it *might* be fixed in the latest testing driver: https://www.nvidia.com/Download/driverR … 764/en-us/

We'll have to see. Once I can tell for sure, I'll mark this post as solved.

haigioli · 2022-10-13 22:51:37

[SOLUTION] Yes, this problem is indeed resolved with the current linux kernel (6.0.10 and nvidia driver (520.56.06). Marking this post as solved now.

Last edited by haigioli (2022-10-13 22:52:22)

Arch Linux

#1 2022-10-12 20:40:31

[SOLVED] nvidia produces black screen at sddm login screen

#2 2022-10-12 21:01:44

Re: [SOLVED] nvidia produces black screen at sddm login screen

#3 2022-10-12 21:13:25

Re: [SOLVED] nvidia produces black screen at sddm login screen

#4 2022-10-12 21:24:49

Re: [SOLVED] nvidia produces black screen at sddm login screen

#5 2022-10-12 21:47:08

Re: [SOLVED] nvidia produces black screen at sddm login screen

#6 2022-10-12 21:52:39

Re: [SOLVED] nvidia produces black screen at sddm login screen

#7 2022-10-12 22:21:03

Re: [SOLVED] nvidia produces black screen at sddm login screen

#8 2022-10-13 01:46:28

Re: [SOLVED] nvidia produces black screen at sddm login screen

#9 2022-10-13 22:51:37

Re: [SOLVED] nvidia produces black screen at sddm login screen

Board footer