Bug 819234 – x11-drivers/nvidia-drivers: conflicts with CONFIG_DRM_SIMPLEDRM (was: X fails to start (blank screen) with kernel 5.14.14 (with gentoo-sources-5.14.14 and nvidia-drivers-495.29.05-r1))

• Status: RESOLVED FIXED
• Product: Gentoo Linux

[Description Abulhair Saparov 2021-10-21 16:44:40 UTC]

I recently updated my system, and X fails to start with kernel version 5.14.14 and nvidia-drivers-495.29.05-r1.
Falling back to kernel 5.14.9 works correctly.

Xorg.0.log contains:
[   176.453] (II) NVIDIA(0): Setting mode "DP-4: nvidia-auto-select @2560x1440 +0+0 {ViewPortIn=2560x1440, ViewPortOut=2560x1440+0+0}, HDMI-0: 640x480 @640x480 +0+0 {ViewPortIn=640x480, ViewPortOut=640x480+0+0}"
[   176.521] (II) NVIDIA(0): ACPI: failed to connect to the ACPI event daemon; the daemon
[   176.521] (II) NVIDIA(0):     may not be running or the "AcpidSocketPath" X
[   176.521] (II) NVIDIA(0):     configuration option may not be set correctly.  When the
[   176.521] (II) NVIDIA(0):     ACPI event daemon is available, the NVIDIA X driver will
[   176.521] (II) NVIDIA(0):     try to use it to receive ACPI event notifications.  For
[   176.521] (II) NVIDIA(0):     details, please see the "ConnectToAcpid" and
[   176.521] (II) NVIDIA(0):     "AcpidSocketPath" X configuration options in Appendix B: X
[   176.521] (II) NVIDIA(0):     Config Options in the README.


Reproducible: Always

Steps to Reproduce:
1. Install kernel version 5.14.14 and nvidia-drivers-495.29.05-r1
2. Restart
Actual Results:  
X fails to start and the screen remains blank.

Expected Results:  
X should start normally and proceed to start the desktop environment.

[Comment 1 Abulhair Saparov 2021-10-21 16:45:32 UTC]

Created attachment 746058 [details]
emerge --info

[Comment 2 Abulhair Saparov 2021-10-21 16:46:24 UTC]

Created attachment 746061 [details]
Xorg.0.log

[Comment 3 Ionen Wolkens 2021-10-21 23:22:54 UTC]

From a quick look I don't see any error in that Xorg.0.log and it /seems/ like it should be working as expected. The acpi thing is fine and you were likely already getting it before as well, it wouldn't lead to a blank screen.

I've tested 5.14.14 already with 495 (both my own system with a self-built kernel, and tried with mostly-out-of-the-box plasma/gnome and gentoo-kernel-bin-5.14.14) and can't see anything different -- or at least not with my own hardware/setup. May want to try the same gentoo-kernel-bin + rebuild drivers if want to rule out config-related issues.

That aside, this is better suited to continue on support channels[1], difficult to help on bugzilla and right now hard to say to whom even assign this to.

[1] https://www.gentoo.org/support/

[Comment 4 Frederik Pfautsch 2021-10-23 11:35:28 UTC]

I have the same problem, downgrading to nvidia-drivers-470.74 with kernel 5.14.14 also works for me.

I attached my kernel .config, dmesg, syslog and Xorg.log for comparison. GPU is a GTX1650, Gnome with XOrg.

[Comment 5 Frederik Pfautsch 2021-10-23 11:35:52 UTC]

Created attachment 746277 [details]
syslog

[Comment 6 Frederik Pfautsch 2021-10-23 11:36:05 UTC]

Created attachment 746280 [details]
dmesg

[Comment 7 Frederik Pfautsch 2021-10-23 11:36:35 UTC]

Created attachment 746283 [details]
XOrg.log

[Comment 8 Frederik Pfautsch 2021-10-23 11:38:40 UTC]

Created attachment 746286 [details]
emerge --info

[Comment 9 Frederik Pfautsch 2021-10-23 11:39:57 UTC]

Created attachment 746289 [details]
5.14.14 .config

[Comment 10 Abulhair Saparov 2021-10-23 18:23:51 UTC]

I was able to get 5.14.14 working by disabling DRM_SIMPLEDRM in the kernel config.

I tried this after I did some more digging: X and Plasma seemed to be running fine in the background, even though my VT7 screen was blank (which I found by `pidof X` and `ps aux | grep -i plasma` etc). So it seemed like something was going wrong with modeswitching.

My dmesg output with kernel version 5.14.9 contained the following:
[    8.462610] [drm] Initialized simpledrm 1.0.0 20200625 for simple-framebuffer.0 on minor 0
[    8.462631] simple-framebuffer simple-framebuffer.0: [drm] *ERROR* fbdev: Failed to setup generic emulation (ret=-22)

But with 5.14.14, this issue seems to have been fixed:
[    8.457485] [drm] Initialized simpledrm 1.0.0 20200625 for simple-framebuffer.0 on minor 0
[    8.466111] Console: switching to colour frame buffer device 128x48
[    8.469664] simple-framebuffer simple-framebuffer.0: [drm] fb0: simpledrm frame buffer device

[Comment 11 Ionen Wolkens 2021-10-23 20:51:28 UTC]

(In reply to Abulhair Saparov from comment #10)
> I was able to get 5.14.14 working by disabling DRM_SIMPLEDRM in the kernel
> config.
Interesting, thanks for experimenting.

It's a new option in 5.14, and that it wasn't working .9 and does in .14 does explain the situation.

Assigning to myself and I'll give it a try later to confirm and likely have the nvidia drivers ebuild suggest disabling the option if it's set. Not seeing much else I can do about this at the moment.

Thankfully it's not enabled by default, and is also disabled in gentoo-kernel-bin last I checked -- so hopefully not hitting too many people.

(In reply to Frederik Pfautsch from comment #4)
> I have the same problem, downgrading to nvidia-drivers-470.74 with kernel
> 5.14.14 also works for me.
That doesn't seem to be the same issue, getting a blank screen is a broad topic with many possible causes. I'd suggest to try support channels first.

https://www.gentoo.org/support/

Albeit 495 remains a new and experimental branch, I considered leaving it unkeyworded/masked but it does fix major gsync issues introduced in 470.74 which is rather important. If in doubt I'd suggest sticking to stable drivers (470.63.01 at the moment, 495 non-production branch will never be stable).

[Comment 12 Larry the Git Cow 2021-10-24 03:30:10 UTC]

The bug has been closed via the following commit(s):

https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=25e0e3ac7ab9febd7ead4bec0e79e6cd7003f38c

commit 25e0e3ac7ab9febd7ead4bec0e79e6cd7003f38c
Author:     Ionen Wolkens <[e-mail]>
AuthorDate: 2021-10-24 03:08:29 +0000
Commit:     Ionen Wolkens <[e-mail]>
CommitDate: 2021-10-24 03:27:44 +0000

    x11-drivers/nvidia-drivers: suggest disabling CONFIG_DRM_SIMPLEDRM
    
    Mostly as a precaution, being set may not always be an issue.
    
    New in kernel 5.14 and may conflict with nvidia's mode switching
    if used with fbdev emulation. Previously benign but fixes have
    caused to become active between kernel 5.14.9 and 5.14.14.
    
    Will hopefully revisit this in the future given this is meant to
    replace efifb/simplefb at some point. Will notably keep an eye on
    what gentoo-kernel-bin's config upstream does, currently kept
    disabled.
    
    Done in <5.14 ebuilds as well to ensure this doesn't get lost if
    these branches gain 5.14 compat.
    
    Thanks-to: Abulhair Saparov
    Closes: https://bugs.gentoo.org/819234
    Signed-off-by: Ionen Wolkens <[e-mail]>

 x11-drivers/nvidia-drivers/nvidia-drivers-390.144.ebuild      | 1 +
 x11-drivers/nvidia-drivers/nvidia-drivers-460.91.03.ebuild    | 1 +
 x11-drivers/nvidia-drivers/nvidia-drivers-470.62.05.ebuild    | 1 +
 x11-drivers/nvidia-drivers/nvidia-drivers-470.63.01.ebuild    | 1 +
 x11-drivers/nvidia-drivers/nvidia-drivers-470.74.ebuild       | 1 +
 x11-drivers/nvidia-drivers/nvidia-drivers-495.29.05-r1.ebuild | 1 +
 6 files changed, 6 insertions(+)

[Comment 13 Frederik Pfautsch 2021-10-27 11:35:30 UTC]

> (In reply to Frederik Pfautsch from comment #4)
> > I have the same problem, downgrading to nvidia-drivers-470.74 with kernel
> > 5.14.14 also works for me.
> That doesn't seem to be the same issue, getting a blank screen is a broad
> topic with many possible causes. I'd suggest to try support channels first.
> 
> https://www.gentoo.org/support/
> 
> Albeit 495 remains a new and experimental branch, I considered leaving it
> unkeyworded/masked but it does fix major gsync issues introduced in 470.74
> which is rather important. If in doubt I'd suggest sticking to stable
> drivers (470.63.01 at the moment, 495 non-production branch will never be
> stable).

Just an update and in case somebody else has similar problems:
I disabled CONFIG_DRM (leftover from my secondary AMD GPU, currently not in use) and AMD SME as suggested in the wiki, also reset my /etc/X11/xorg.conf.d/nvidia.conf to wiki-defaults. Rebuilding nvidia-drivers (and meta packet xork-drivers). After a couple of seconds after complete boot GDM is launching now and everything is working.

[Comment 14 Frederik Pfautsch 2021-10-27 11:58:58 UTC]

Sorry for the double post, after reading https://www.tomshardware.com/news/amd-memory-encryption-disabled-in-linux I enabled SME again to try it, same problem as before, blank screen with blinking cursor. So it does also cause problems in 5.14.14 with 495, but not with 470.