Possible Wayland bug introduced with latest nvidia-dkms-575.57.08

seth · 2025-07-19 22:34:22

On *each* of the edids???? Even the expectably empty ones? (The loop will call it ~8 times)
Even if you remove all but one monitor? (Notably remove the daisy-chained one)

Edit, sidebar: please avoid bloating the thread w/ full quotes of previous posts.

Last edited by seth (2025-07-19 22:38:53)

phpTANKER · 2025-07-19 23:03:08

seth wrote:

On *each* of the edids???? Even the expectably empty ones? (The loop will call it ~8 times)
Even if you remove all but one monitor? (Notably remove the daisy-chained one)

Running it with all 3 monitors displayed just seems to hang, yes. I can try to remove the daisy-chained one as well and run it. Will post an update later.

phpTANKER · 2025-07-20 03:03:56

seth wrote:

On *each* of the edids???? Even the expectably empty ones? (The loop will call it ~8 times)
Even if you remove all but one monitor? (Notably remove the daisy-chained one)

Same with only one confirmed working monitor connected. Weird.

seth · 2025-07-20 07:00:27

strace -o /tmp/ed.strace edid-decode  /sys/class/drm/card?????????????/edid

You'll have to fill in the "?????????????" since I don't think I've seen the topography on your system, the strace gets logged to /tmp/ed.strace

This isn't the case w/ the 570xx drivers?
(ftr, not an issue on @priultimus' system)

Also

seth wrote:

please avoid bloating the thread w/ full quotes of previous posts.

And please don't bump, edit your previous post to mend it if nobody has yet replied.

phpTANKER · 2025-07-20 12:53:52

seth wrote:

strace -o /tmp/ed.strace edid-decode  /sys/class/drm/card?????????????/edid
You'll have to fill in the "?????????????" since I don't think I've seen the topography on your system, the strace gets logged to /tmp/ed.strace
This isn't the case w/ the 570xx drivers?
(ftr, not an issue on @priultimus' system)
Also
seth wrote:
please avoid bloating the thread w/ full quotes of previous posts.
And please don't bump, edit your previous post to mend it if nobody has yet replied.

I went through DP-2 to DP-4 and I could see the hex dump of the EDID on each output, here is the ed.strace output:
https://0x0.st/8ddA.txt

seth · 2025-07-20 13:24:19

Where did that call "hang"?
edid-decode prints the edid for /sys/class/drm/card1-DP-4/edid and exits w/o any error.

What is the actual, complete and unaltered output of

for OUT in /sys/class/drm/card*; do echo $OUT; edid-decode $OUT/edid; echo "================="; done

phpTANKER · 2025-07-20 18:58:00

seth wrote:

Where did that call "hang"?
edid-decode prints the edid for /sys/class/drm/card1-DP-4/edid and exits w/o any error.
What is the actual, complete and unaltered output of
for OUT in /sys/class/drm/card*; do echo $OUT; edid-decode $OUT/edid; echo "================="; done

Maybe I'm being a bit slow here but just running the command without editing it gives me this:

Or did you mean I should edit the card* part to match my card/dp combo?

seth · 2025-07-20 19:14:38

ls -lr /sys/class/drm/card*

But it seems there's a bogus output and that blocks the edid scan and that will throw off everything else.
By your screenshot, it's also gonna be the very first ls entry…

phpTANKER · 2025-07-20 19:53:44

seth wrote:

ls -lr /sys/class/drm/card*
But it seems there's a bogus output and that blocks the edid scan and that will throw off everything else.
By your screenshot, it's also gonna be the very first ls entry…

https://0x0.st/8dhW.txt

Testing the loop with card1:

seth · 2025-07-20 19:59:22

That should not work at all since there is of course no /sys/class/drm/card1/edid … judging from the syntax highlighting: is this fish or a real shell?

phpTANKER · 2025-07-20 20:02:20

seth wrote:

That should not work at all since there is of course no /sys/class/drm/card1/edid … judging from the syntax highlighting: is this fish or a real shell?

Fish, and like you said it's not really "real" since it doesn't run simple syntax correctly apparently, you learn something every day!

Bash gives me a better output:
https://0x0.st/8dhD.txt

seth · 2025-07-20 20:10:06

https://archlinux.org/packages/extra/x86_64/zsh/
https://archlinux.org/packages/extra/an … sh-config/
https://archlinux.org/packages/extra/an … hlighting/
https://archlinux.org/packages/extra/an … ggestions/

But unfortunately there goes the easy explanation w/ some bogus EDID getting in the way
All three EDIDs are perfectly sane and properly parsed, all are 8bpc, none VRR

For recap, have we tested a wayland compositor that's not KDE?

phpTANKER · 2025-07-20 20:13:41

seth wrote:

https://archlinux.org/packages/extra/x86_64/zsh/
https://archlinux.org/packages/extra/an … sh-config/
https://archlinux.org/packages/extra/an … hlighting/
https://archlinux.org/packages/extra/an … ggestions/
But unfortunately there goes the easy explanation w/ some bogus EDID getting in the way
All three EDIDs are perfectly sane and properly parsed, all are 8bpc, none VRR
For recap, have we tested a wayland compositor that's not KDE?

Will take a look, I have a simple oh my zsh setup for running stuff in zsh at least.

Yes I did try Weston on the newer drivers.

seth · 2025-07-20 21:28:15

I have a simple oh my zsh setup

That's actually worse than fish … (feel free to search the forum)

So wayland™ struggles w/ your output config…
The multi-user.target (linux console) isn't affected?

phpTANKER · 2025-07-20 22:38:16

seth wrote:

I have a simple oh my zsh setup
That's actually worse than fish … (feel free to search the forum)
So wayland™ struggles w/ your output config…
The multi-user.target (linux console) isn't affected?

A topic for another thread perhaps..

No, the linux console shows correctly on all 3 displays, as does SDDM. It's only once I login to the Plasma/Weston session that the daisy-chained display stops working. I'm currently building a new machine that is fully AMD so this won't be a problem for me in the future. However might be worth finding a solution since a lot of users might still be on the same gpu.

seth · 2025-07-21 07:18:22

If this works on the framebuffer console and X11, it's not gonna be a "GPU" thing.
Can you explicitly activate the output,

hyprctl monitors

?

phpTANKER · 2025-07-21 11:29:22

seth wrote:

If this works on the framebuffer console and X11, it's not gonna be a "GPU" thing.
Can you explicitly activate the output,
hyprctl monitors
?

On the new GPU now so any tests will be kind of redundant. I am putting together the old parts into a spare machine which I will test this on later.

Interesting fact: The new GPU doesn't always detect the problematic screen, which leads me to think it might be a physical issue with the cable. But I can't puzzle together why that would break between driver installs persistently and work 100% on the old one..

seth · 2025-07-21 13:51:48

which leads me to think it might be a physical issue with the cable

seth wrote:

If this works on the framebuffer console and X11, it's not gonna be a "GPU" thing.

Does this also affect the console or X11 on the new GPU now?

phpTANKER · 2025-07-21 14:06:23

seth wrote:

which leads me to think it might be a physical issue with the cable
seth wrote:
If this works on the framebuffer console and X11, it's not gonna be a "GPU" thing.
Does this also affect the console or X11 on the new GPU now?

Only intermittently, I just put in the new gpu today so I haven't rebooted many times outside of cleaning out old Nvidia stuff/replacing with AMD.

It has happened twice so far, and I switched the displays around between a reboot to see if it was the port or the cable that's messed up and to my surprise the other display went black until i replugged it on the next reboot. I have rebooted once since then and it is fine so fingers crossed. The console appeared on all 3 displays on the current boot. When you say affect X11, do you mean X11 apps? Everything I normally run seems fine.

This is starting to move into the occult territory for me tbh, since I just replaced the entire system and still seeing these weird issues it feels more like a display/cable issue (even though it seemed to happen to 2 different displays when i swapped the cables). I will reboot a lot today to see if it still keeps happening.

seth · 2025-07-21 20:34:15

When you say affect X11, do you mean X11 apps?

No, X11 display server, KDE on Xorg etc.
The clients have no relation to this (on wayland they cannot even know whether or what output exists)

move into the occult territory … cable issue

It's usually the cable, except for your previous findings suggesting *only* wayland display servers (but then all of those tested) were affected.

phpTANKER · 2025-07-21 21:07:30

seth wrote:

When you say affect X11, do you mean X11 apps?
No, X11 display server, KDE on Xorg etc.
The clients have no relation to this (on wayland they cannot even know whether or what output exists)
move into the occult territory … cable issue
It's usually the cable, except for your previous findings suggesting *only* wayland display servers (but then all of those tested) were affected.

The reason I mention occult is how a faulty cable can break Wayland between Nvidia driver versions..

priultimus · 2025-08-03 17:11:07

seth wrote:

Try to add "nvidia_drm.modeset=1" here to block the simpledrm device and see what happens.
Also only boot the multi-user.target (2nd link below) and start Hyprland (or better: some X11 session) manually and avoid all the other OOT modules:
· evdi(OE)
· vmnet(OE)
· vmmon(OE)
· vmw_vmci vboxnetflt(OE)
· vboxnetadp(OE)
· vboxdrv(OE)
· legion_laptop(OE)
(I get that you're probably using those for a reason, but we need to find some stable base to operate from)

Sorry to post back so late, I've been pretty busy. But I've attempted booting without any of those out of tree modules and into multi-user.target:
http://0x0.st/87y8.txt
This is also a boot without the daisy-chained monitors plugged in, notably doesn't have the "Failed to add connector for NvKmsKapiDisplay":
http://0x0.st/87y-.txt
as for the symptoms, the behaviour is identical. I didn't bother trying an X11 session because the console was not outputting to the other monitor either.

seth wrote:

Last resort: downgrade to 570xx nvidia-dkms to see whether this is actually related to the 575xx drivers.

This worked! My both my monitors are on and functioning correctly now.
http://0x0.st/87Gg.txt

phpTANKER · 2025-08-03 17:18:36

priultimus wrote:

seth wrote:
Try to add "nvidia_drm.modeset=1" here to block the simpledrm device and see what happens.
Also only boot the multi-user.target (2nd link below) and start Hyprland (or better: some X11 session) manually and avoid all the other OOT modules:
· evdi(OE)
· vmnet(OE)
· vmmon(OE)
· vmw_vmci vboxnetflt(OE)
· vboxnetadp(OE)
· vboxdrv(OE)
· legion_laptop(OE)
(I get that you're probably using those for a reason, but we need to find some stable base to operate from)
Sorry to post back so late, I've been pretty busy. But I've attempted booting without any of those out of tree modules and into multi-user.target:
http://0x0.st/87y8.txt
This is also a boot without the daisy-chained monitors plugged in, notably doesn't have the "Failed to add connector for NvKmsKapiDisplay":
http://0x0.st/87y-.txt
as for the symptoms, the behaviour is identical. I didn't bother trying an X11 session because the console was not outputting to the other monitor either.
seth wrote:
Last resort: downgrade to 570xx nvidia-dkms to see whether this is actually related to the 575xx drivers.
This worked! My both my monitors are on and functioning correctly now.
http://0x0.st/87Gg.txt

I'm glad downgrading solved it for you. Since I built a new system this is not an issue, and my frankensteined old gaming rig with the 1080 now only uses one display Problem not solved but it's gone for me!

priultimus · 2025-08-03 17:22:13

Glad you've got a resolution as well! Although I do find it strange that you only had this issue on Wayland sessions. I do wonder if it's a bug in the driver and something I could make a report for. At least for me, the behaviour I got was pretty consistent. I'd be willing to do more troubleshooting, I don't think it'd be ideal to be on an older version long term.

Arch Linux

#51 2025-07-19 22:34:22

Re: Possible Wayland bug introduced with latest nvidia-dkms-575.57.08

#52 2025-07-19 23:03:08

Re: Possible Wayland bug introduced with latest nvidia-dkms-575.57.08

#53 2025-07-20 03:03:56

Re: Possible Wayland bug introduced with latest nvidia-dkms-575.57.08

#54 2025-07-20 07:00:27

Re: Possible Wayland bug introduced with latest nvidia-dkms-575.57.08

#55 2025-07-20 12:53:52

Re: Possible Wayland bug introduced with latest nvidia-dkms-575.57.08

#56 2025-07-20 13:24:19

Re: Possible Wayland bug introduced with latest nvidia-dkms-575.57.08

#57 2025-07-20 18:58:00

Re: Possible Wayland bug introduced with latest nvidia-dkms-575.57.08

#58 2025-07-20 19:14:38

Re: Possible Wayland bug introduced with latest nvidia-dkms-575.57.08

#59 2025-07-20 19:53:44

Re: Possible Wayland bug introduced with latest nvidia-dkms-575.57.08

#60 2025-07-20 19:59:22

Re: Possible Wayland bug introduced with latest nvidia-dkms-575.57.08

#61 2025-07-20 20:02:20

Re: Possible Wayland bug introduced with latest nvidia-dkms-575.57.08

#62 2025-07-20 20:10:06

Re: Possible Wayland bug introduced with latest nvidia-dkms-575.57.08

#63 2025-07-20 20:13:41

Re: Possible Wayland bug introduced with latest nvidia-dkms-575.57.08

#64 2025-07-20 21:28:15

Re: Possible Wayland bug introduced with latest nvidia-dkms-575.57.08

#65 2025-07-20 22:38:16

Re: Possible Wayland bug introduced with latest nvidia-dkms-575.57.08

#66 2025-07-21 07:18:22

Re: Possible Wayland bug introduced with latest nvidia-dkms-575.57.08

#67 2025-07-21 11:29:22

Re: Possible Wayland bug introduced with latest nvidia-dkms-575.57.08

#68 2025-07-21 13:51:48

Re: Possible Wayland bug introduced with latest nvidia-dkms-575.57.08

#69 2025-07-21 14:06:23

Re: Possible Wayland bug introduced with latest nvidia-dkms-575.57.08

#70 2025-07-21 20:34:15

Re: Possible Wayland bug introduced with latest nvidia-dkms-575.57.08

#71 2025-07-21 21:07:30

Re: Possible Wayland bug introduced with latest nvidia-dkms-575.57.08

#72 2025-08-03 17:11:07

Re: Possible Wayland bug introduced with latest nvidia-dkms-575.57.08

#73 2025-08-03 17:18:36

Re: Possible Wayland bug introduced with latest nvidia-dkms-575.57.08

#74 2025-08-03 17:22:13

Re: Possible Wayland bug introduced with latest nvidia-dkms-575.57.08

Board footer