Errors resulting in black screen

[soldierofhell]

Hi, trying to run latest docker image on remote server (from ssh console) with 2 NVIDIA GPUs.

When I execute ./run.sh I see the following output with errors (seems like mostly permission errors)

Stopping "xubuntu" container...
Removing "xubuntu" container...
Creating "xubuntu" container...
Done!

sha1 Fingerprint=9C:D9:6E:04:C2:DA:39:E6:2E:73:7A:76:69:B1:BC:10:25:9D:3D:9E
sha256 Fingerprint=60:89:B2:24:64:E2:CA:34:BE:45:E5:13:76:91:37:ED:49:0E:B7:3B:F8:91:C6:C1:2A:47:8B:93:11:C0:AE:08
[20230613-19:45:44] [INFO ] starting xrdp with pid 69
[20230613-19:45:44] [INFO ] starting xrdp-sesman with pid 68
Failed to set receive buffer size for device monitor, ignoring: Operation not permitted
[20230613-19:45:44] [INFO ] address [0.0.0.0] port [3389] mode 1
[20230613-19:45:44] [INFO ] listening to port 3389 on 0.0.0.0
[20230613-19:45:44] [INFO ] xrdp_listen_pp done
kmalloc-8(122704:72a0868a96eb001d1991def099629933bcb4f9af3b382e8981d4a8672c1e010c): Failed to send device, ignoring: Operation not permitted
kmalloc-rcl-64(122704:72a0868a96eb001d1991def099629933bcb4f9af3b382e8981d4a8672c1e010c): Failed to send device, ignoring: Operation not permitted
kmalloc-rcl-128(122704:72a0868a96eb001d1991def099629933bcb4f9af3b382e8981d4a8672c1e010c): Failed to send device, ignoring: Operation not permitted
kmalloc-rcl-96(122704:72a0868a96eb001d1991def099629933bcb4f9af3b382e8981d4a8672c1e010c): Worker [78] did not accept message, killing the worker: Operation not permitted
kmalloc-rcl-96(122704:72a0868a96eb001d1991def099629933bcb4f9af3b382e8981d4a8672c1e010c): Worker [77] did not accept message, killing the worker: Operation not permitted
kmalloc-rcl-96(122704:72a0868a96eb001d1991def099629933bcb4f9af3b382e8981d4a8672c1e010c): Worker [76] did not accept message, killing the worker: Operation not permitted
Worker [76] terminated by signal 9 (KILL)
Worker [78] terminated by signal 9 (KILL)
Worker [77] terminated by signal 9 (KILL)
kmalloc-rcl-96(122704:72a0868a96eb001d1991def099629933bcb4f9af3b382e8981d4a8672c1e010c): Failed to send device, ignoring: Operation not permitted
LNXSYSTM:00: Failed to write 'change' to '/sys/devices/LNXSYSTM:00/uevent': Read-only file system

Then when I connect with remmina client I see log screen, but after logon I see blank screen.

Could you please suggest where to look? Thanks,

[soldierofhell]

I've tried oldest available image (v93) (with run.sh from v93 tag) and

1. some errors after container startup disappeared, but not all

SHA1 Fingerprint=84:18:A8:5D:10:5F:A8:F2:91:17:09:87:08:21:B4:68:9E:BC:F5:37
SHA256 Fingerprint=6B:D8:3D:16:A7:0C:6E:B9:89:ED:E8:41:14:23:8D:24:50:D1:15:A8:AE:D3:B4:A2:7E:E5:98:0C:18:1F:AC:87
[20230613-20:23:27] [INFO ] starting xrdp-sesman with pid 73
[20230613-20:23:27] [INFO ] starting xrdp with pid 74
[20230613-20:23:27] [INFO ] address [0.0.0.0] port [3389] mode 1
[20230613-20:23:27] [INFO ] listening to port 3389 on 0.0.0.0
[20230613-20:23:27] [INFO ] xrdp_listen_pp done
kmalloc-8(122801:b3c08c1f0f01fcb54c35fc900eef232a1cb52697d003cef8b934544eb56facb1): Failed to send device, ignoring: Operation not permitted
kmalloc-rcl-128(122801:b3c08c1f0f01fcb54c35fc900eef232a1cb52697d003cef8b934544eb56facb1): Failed to send device, ignoring: Operation not permitted
kmalloc-rcl-96(122801:b3c08c1f0f01fcb54c35fc900eef232a1cb52697d003cef8b934544eb56facb1): Worker [81] did not accept message, killing the worker: Operation not permitted
kmalloc-rcl-96(122801:b3c08c1f0f01fcb54c35fc900eef232a1cb52697d003cef8b934544eb56facb1): Worker [82] did not accept message, killing the worker: Operation not permitted
Worker [81] terminated by signal 9 (KILL)
Worker [82] terminated by signal 9 (KILL)
kmalloc-rcl-96(122801:b3c08c1f0f01fcb54c35fc900eef232a1cb52697d003cef8b934544eb56facb1): Failed to send device, ignoring: Operation not permitted

2. I can connect to desktop, but VirtualGL doesn't recognize GPU. From what I briefly see, your approach to GPU is more general than NVIDIA Container Toolkit, e.g. you manually install drivers and connect devices. I wonder if run.sh support NVIDIA out-of-the-box or I should modify it (e.g. add --gpus all, etc.)

[soldierofhell]

BTW I hope I will succeed in make it work, because I've read previous "success story" for NVIDIA GPU (vide #7 (comment)). Unfortunately that was before v93 tag, but I guess I can build the image by myself. Still, hope for some valuable insight

[soldierofhell]

I went back to v69, added parameters required by Nvidia Container Toolkit to run.sh and it works! Now I'll try to narrow down what works and what causes troubles

[hectorm]

Thanks for your testing, unfortunately I don't have an NVIDIA card anymore and every time I need to do some testing I have to rent a server.

If you find the cause I would appreciate it if you could update the thread, otherwise I will try to find time later to do my own tests.