Stardust XR•6mo ago

StardustXR Server causes Monado to segfault immediately

Not sure where the problem lies, but xrgears runs fine. I'm running the latest garuda linux on basically a fresh install, and I've tested this on xfce, and kde/wayland flavors. Both behave the same. This is with the newest monado v24.0.0 and latest stardust. This is inside a vfio VM, but I've tested steamvr inside it, so I suspect that's not relevant. Should I be using an earlier version of monado? I tried getting v21, but it was far too old to compile easily on garuda.

Monado.txt

stardustxr_precrash....

26 Replies

Nova•6mo ago

what stardust server version are you using? is it main or dev branch?

LostOP•6mo ago

Ah, the main branch. Let me try dev

Nova•6mo ago

dev is... very broken :p anyway, this is so weirdd can you run monado using a debugger? envision has one

LostOP•6mo ago

Haha, okay, I won't try dev. I can grab envision. How would I open the debugger? Or could you point me to some docs?

Nova•6mo ago

oh if you don't use envision then just run gdb on it

LostOP•6mo ago

[New Thread 0x7fffc29f36c0 (LWP 13293)]
 INFO [client_loop] Client 1 connected
 INFO [ipc_handle_instance_describe_client] Client info:
    id: 1
    application_name: 'Stardust XR'
    pid: 13280
    extensions:
        ext_hand_tracking_enabled: true
        ext_eye_gaze_interaction_enabled: true
        ext_hand_interaction_enabled: false
[New Thread 0x7fffc21f26c0 (LWP 13294)]

Thread 16 "monado-service" received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7fffc21f26c0 (LWP 13294)]
0x00007fffe6a07a87 in ?? () from /usr/lib/libnvidia-glcore.so.550.90.07
(gdb) backtrace
#0  0x00007fffe6a07a87 in ?? () from /usr/lib/libnvidia-glcore.so.550.90.07
#1  0x00007fffe6e22d99 in ?? () from /usr/lib/libnvidia-glcore.so.550.90.07
#2  0x00007fffe6e07790 in ?? () from /usr/lib/libnvidia-glcore.so.550.90.07
#3  0x00007fffe6d237ee in ?? () from /usr/lib/libnvidia-glcore.so.550.90.07
#4  0x00005555556b73d3 in fence_wait (xcf=<optimized out>, timeout=<optimized out>) at /usr/src/debug/monado/monado-v24.0.0/src/xrt/compositor/util/comp_sync.c:64
#5  0x00005555556c841f in xrt_compositor_fence_wait (xcf=0x7fffb4029d80, timeout=100000000) at /usr/src/debug/monado/monado-v24.0.0/src/xrt/include/xrt/xrt_compositor.h:753
#6  wait_fence (xcf_ptr=<optimized out>) at /usr/src/debug/monado/monado-v24.0.0/src/xrt/compositor/multi/comp_multi_compositor.c:132
#7  run_func (ptr=0x7fffb4000bf0) at /usr/src/debug/monado/monado-v24.0.0/src/xrt/compositor/multi/comp_multi_compositor.c:291
#8  0x00007ffff38a6ded in start_thread (arg=<optimized out>) at pthread_create.c:447
#9  0x00007ffff392a0dc in clone3 () at ../sysdeps/unix/sysv/linux/x86_64/clone3.S:78
(gdb)

[New Thread 0x7fffc29f36c0 (LWP 13293)]
 INFO [client_loop] Client 1 connected
 INFO [ipc_handle_instance_describe_client] Client info:
    id: 1
    application_name: 'Stardust XR'
    pid: 13280
    extensions:
        ext_hand_tracking_enabled: true
        ext_eye_gaze_interaction_enabled: true
        ext_hand_interaction_enabled: false
[New Thread 0x7fffc21f26c0 (LWP 13294)]

Thread 16 "monado-service" received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7fffc21f26c0 (LWP 13294)]
0x00007fffe6a07a87 in ?? () from /usr/lib/libnvidia-glcore.so.550.90.07
(gdb) backtrace
#0  0x00007fffe6a07a87 in ?? () from /usr/lib/libnvidia-glcore.so.550.90.07
#1  0x00007fffe6e22d99 in ?? () from /usr/lib/libnvidia-glcore.so.550.90.07
#2  0x00007fffe6e07790 in ?? () from /usr/lib/libnvidia-glcore.so.550.90.07
#3  0x00007fffe6d237ee in ?? () from /usr/lib/libnvidia-glcore.so.550.90.07
#4  0x00005555556b73d3 in fence_wait (xcf=<optimized out>, timeout=<optimized out>) at /usr/src/debug/monado/monado-v24.0.0/src/xrt/compositor/util/comp_sync.c:64
#5  0x00005555556c841f in xrt_compositor_fence_wait (xcf=0x7fffb4029d80, timeout=100000000) at /usr/src/debug/monado/monado-v24.0.0/src/xrt/include/xrt/xrt_compositor.h:753
#6  wait_fence (xcf_ptr=<optimized out>) at /usr/src/debug/monado/monado-v24.0.0/src/xrt/compositor/multi/comp_multi_compositor.c:132
#7  run_func (ptr=0x7fffb4000bf0) at /usr/src/debug/monado/monado-v24.0.0/src/xrt/compositor/multi/comp_multi_compositor.c:291
#8  0x00007ffff38a6ded in start_thread (arg=<optimized out>) at pthread_create.c:447
#9  0x00007ffff392a0dc in clone3 () at ../sysdeps/unix/sysv/linux/x86_64/clone3.S:78
(gdb)

Let me know if you need more and thanks for the help!

Nova•6mo ago

ugh nvidia uhhh, so i think this is a monado problem https://discord.gg/dTdPVDVN you should probably ask here since it's on the service side stardust shouldn't be able to cause a crash regardless of how broken it might be

LostOP•6mo ago

Fair enough! I'll bring it up there. Thanks again

Nova•6mo ago

no problem

Schmarni•6mo ago

this looks like the same issue i was having btw, it's something in stereokit that trigger the issue

Nova•6mo ago

hmm well if it's stereokit, maybe the dev branch on the server will help you then i updated it there but don't expect any clients to work yet (except maybe flatland on dev branch), just the hands

Schmarni•6mo ago

i don't think it will sine even building stereokit. has the same issue

Nova•6mo ago

wait building stereokit has an issue?

Schmarni•6mo ago

ye but i meant after I tested stardust, a clean stereokit-rs example, a stereokitC app with precompiled libs and the StereokitC test app completely compiled on my machine, they all crash the runtime the same way and i have had no problems like this with any other vr app

Nova•6mo ago

huh that's very strange

Schmarni•5mo ago

i should probably just bite the bullet and make a github issue on stereokit since i have been trying to fix this issue on and off for months and trying to debug a codebase i don't know in a language i don't know doesn't seem like the best use of my time

Nova•5mo ago

i wish i could help but i'm so out of practice with that codebase you could ask in the stereokit server tho

Schmarni•5mo ago

made an issue here https://github.com/StereoKit/StereoKit/issues/1033 i made a "fix" for this, it's a problem with the nvidia drivers not being able to cope with the use of a fence. i had to patch monado

Nova•5mo ago

oh thank!!!!

Schmarni•5mo ago

in my testing this didn't break anything but i didn't test it that mutch yet

nvidia_stereokit_fix...

Skull•5mo ago

Awesome! Huge thanks Think stereokit will take this upstream?

Schmarni•5mo ago

this is a monado patch?

Skull•5mo ago

Oh! Are you able to add an nvidia vendor check and PR monado? This would solve a giant class of issues for a large group of users currently looking to sample the wares

Nova•5mo ago

ah hi

Skull•5mo ago

It will apparently cause graphical artifacts if you use this patch "Look for insert_fence_func in comp_egl_client.c, set it to NULL." Is what I was told to relay

Schmarni•5mo ago

yeah i was expecting issues with it, it was just an extremely hacky solution that worked on my machine. for a upstream solution i'll look into a better solution

Gaming

Programming

StardustXR Server causes Monado to segfault immediately