Cannot start/switch DEs more than once with Nvidia GPU

TheDcoder · Yesterday 13:01:03

I noticed that I cannot start any DE more than once after booting, i.e. if I start a DE session and then logout, I cannot start another session as the DE compositor will just crash during initialization, it happens will all of the DEs that I tested: KDE, Hyperland (both Wayland) and Xfce (X11).

They all start normally if they're the first to start after a fresh boot. The issue persists until I reboot the system. Here's a log of a boot where I reproduced the issue by starting Hyprland twice. I suspect the issue might be related to my Nvidia GPU, I'm currently using the open drivers and the modules are loaded early in the initramfs.

seth · Yesterday 14:21:20

Feb 28 17:39:04 arch systemd[1]: Started Lemurs.
Feb 28 17:39:33 arch lemurs[840]: pam_loginuid(lemurs:session): Error writing /proc/self/loginuid: Operation not permitted
Feb 28 17:39:33 arch lemurs[840]: pam_loginuid(lemurs:session): set_loginuid failed

what is you don't use lemur but start hyprland or whatever from a getty login?

The log btw. doesn't reflect your narration.

Feb 28 17:39:33 arch systemd-logind[761]: New session '1' of user 'TheDcoder' with class 'manager' and type 'unspecified'.
…
Feb 28 17:39:33 arch systemd-journald[456]: Time jumped backwards, rotating.
Feb 28 17:39:33 arch systemd-timesyncd[490]: Initial clock synchronization to Sat 2026-02-28 17:39:33.306366 IST.
Feb 28 17:39:42 arch systemd-resolved[726]: Using degraded feature set UDP instead of UDP+EDNS0 for DNS server 194.242.2.2#dns.mullvad.net.
Feb 28 17:40:20 arch kernel: Hyprland[1244]: segfault at 10 ip 00007fcf7f0153e1 sp 00007ffdbb151450 error 4 in libaquamarine.so.0.10.0[863e1,7fcf7efb5000+c4000] likely on CPU 8 (core 16, socket 0)
Feb 28 17:40:20 arch kernel: Code: c0 48 c7 47 18 00 00 00 00 c5 fa 7f 47 20 c3 66 2e 0f 1f 84 00 00 00 00 00 f3 0f 1e fa 41 55 41 54 55 53 48 89 fb 48 83 ec 58 <48> 83 7f 10 00 64 48 8b 2c 25 28 00 00 00 48 89 6c 24 48 89 f5 0f
Feb 28 17:40:20 arch systemd-coredump[1447]: Process 1244 (Hyprland) of user 1000 terminated abnormally with signal 11/SEGV, processing...
Feb 28 17:40:20 arch systemd[1]: Created slice Slice /system/systemd-coredump.
Feb 28 17:40:20 arch systemd[1]: Started Process Core Dump (PID 1447/UID 0).
Feb 28 17:40:20 arch systemd-coredump[1448]: Process 1244 (Hyprland) of user 1000 dumped core.
…
Feb 28 17:40:20 arch lemurs[840]: pam_unix(lemurs:session): session closed for user TheDcoder
Feb 28 17:40:33 arch lemurs[840]: pam_loginuid(lemurs:session): Error writing /proc/self/loginuid: Operation not permitted
Feb 28 17:40:33 arch lemurs[840]: pam_loginuid(lemurs:session): set_loginuid failed
Feb 28 17:40:33 arch lemurs[840]: pam_unix(lemurs:session): session opened for user TheDcoder(uid=1000) by (uid=0)
Feb 28 17:40:34 arch systemd-coredump[1505]: Process 1491 (Hyprland) of user 1000 terminated abnormally with signal 6/ABRT, processing...
Feb 28 17:40:34 arch systemd[1]: Started Process Core Dump (PID 1505/UID 0).
Feb 28 17:40:34 arch systemd-coredump[1506]: Process 1491 (Hyprland) of user 1000 dumped core.
                                             
                                             Stack trace of thread 1491:
                                             #0  0x00007f1681225663 abort (libc.so.6 + 0x25663)
                                             #1  0x0000561fc673f4c0 n/a (/usr/bin/Hyprland + 0x2714c0)
                                             #2  0x00007f168123e2d0 n/a (libc.so.6 + 0x3e2d0)
                                             #3  0x00007f1681298a2c n/a (libc.so.6 + 0x98a2c)
                                             #4  0x00007f168123e1a0 raise (libc.so.6 + 0x3e1a0)
                                             #5  0x00007f16812255fe abort (libc.so.6 + 0x255fe)
                                             #6  0x00007f168169502b n/a (libstdc++.so.6 + 0x9502b)

You login, get network access the clock jumps backwards* and then hyprland crashes then lemurs tries to start hyprland again (there's no logout as far as logind is concerned most likely because of the lemurs bugs) and then hyprland crashes again.

*I assume there's a parallel windows installation
=> https://wiki.archlinux.org/title/Dual_b … e_standard
And also see the 3rd link below. Mandatory.
Disable it (it's NOT the BIOS setting!) and reboot windows and linux twice for voodo reasons.

TheDcoder · Yesterday 18:31:39

seth wrote:

what is you don't use lemur but start hyprland or whatever from a getty login?

Not sure I understand, I did use lemurs as the login manager, in fact I used it for the current boot as I write, I did notice those errors but chalked them up to being nothing major since everything else worked fine (aside from this thread's issue). I can also use getty to login manually via the shell but the issue remains, so I don't think lemurs is related.

seth wrote:

You login, get network access the clock jumps backwards* and then hyprland crashes then lemurs tries to start hyprland again (there's no logout as far as logind is concerned most likely because of the lemurs bugs) and then hyprland crashes again.

Perhaps the bug with lemurs didn't initialize the "login" properly, I am quite certain that I logged in twice, or at-least I tried, because hyprland crashed on the 2nd attempt. It never fails on the first attempt after a fresh boot.

seth wrote:

*I assume there's a parallel windows installation

There is but I haven't booted into it for months, also the time jump was just ~1 second, it's caused due to natural inaccuracy of my system's RTC, not due to Windows setting the wrong non-local timezone, which I have disabled in the registry as well.

seth wrote:

And also see the 3rd link below. Mandatory.

Ha, that one's a classic, I've had fast boot disabled ever since Windows 8.1 came out

Back then it only caused issues with mounting NTFS partitions, does it cause more trouble these days?

seth · Yesterday 20:56:31

I can also use getty to login manually via the shell but the issue remains, so I don't think lemurs is related.

Do you have a journal for that?

because hyprland crashed on the 2nd attempt. It never fails on the first attempt after a fresh boot.

It *does* crash on exit (apparently) from the first session in what looks like
https://github.com/hyprwm/Hyprland/discussions/12467
https://github.com/hyprwm/Hyprland/discussions/12750
https://gitlab.archlinux.org/archlinux/ … -/issues/5 - from the conversation there the patch might not have been downstreamed.

Leaving aside concerns of that resulting in a polluted state causing subsequent aborts (the hyprland log might even have messages about what it doesn't like since the crash of the second run is not a segfault but a "controlled" abort) and whether those are related to the lemurs/logind situation
=> you could try to apply the aquamarine patch locally.

Arch Linux

#1 Yesterday 13:01:03

Cannot start/switch DEs more than once with Nvidia GPU

#2 Yesterday 14:21:20

Re: Cannot start/switch DEs more than once with Nvidia GPU

#3 Yesterday 18:31:39

Re: Cannot start/switch DEs more than once with Nvidia GPU

#4 Yesterday 20:56:31

Re: Cannot start/switch DEs more than once with Nvidia GPU

Board footer