-
-
Notifications
You must be signed in to change notification settings - Fork 106
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
VirtualGL on ubuntu hangs at the entrypoint and has no errors #160
Comments
This sentence raised a couple of red flags. What do you mean by "first tty"? Do you have multiple display managers or multiple X servers running? What do you mean by "didn't allow vglrun to do its thing?" Do you mean that VirtualGL doesn't work properly when the 3D X server is at the login prompt? If so, then that's the root of the issue. How specifically does it fail? I assume that you ran |
To make sure it's clear, I'm doing By "first tty", I mean the screen I reach when I press ctrl-alt-F1 on the physical keyboard. When I run
I made the change to the greeter autostart, but the file I Also, I have no attachment to gdm/kdm/lightdm/etc; I'm happy to make any changes you advise to debug & get this working. I have only cursory familiarity with the Wayland vs X differences, and if Ubuntu's wayland defaults are making this more painful, I would like your suggestion as to what to remove and what to use instead. My only constraint is that the CAD program is proprietary, and only supports Ubuntu 20.04 :( |
I'm sorry to double-post, but I did identify something: the 2nd Xorg running as me that was bound to vt3 and later vt5 is due to having checked the "automatically log me in" option. I unselected that option, restarted GDM again, and now there's only GDM running on vt1 with no other xorg sessions. I reverted the greeter autostart to
Thank you again for your help. |
I'm sorry again to triple post, but I've made significant progress in debugging this, but I do not understand what's happening.
|
VirtualGL works perfectly well with headless configurations, including nVidia adapters with no monitor outputs. However, the 3D X server VT must always be the active VT in order to use hardware-accelerated OpenGL. That is a system limitation, and it has existed for as long as VirtualGL has existed (although, in the pre-Wayland days, it tended to manifest as a Bear in mind that VirtualGL is primarily designed around the needs of large-scale multi-user systems, and on such systems, the best practices are to use a headless 3D X server and never log into it. VirtualGL works on single-user workstations as well, but it isn't compatible with every single X server configuration that is possible on a single-user workstation. If you need to log into the 3D X server locally, then you'll need to redirect the 3D X server connection to :1 for as long as you are logged in. If you need to switch VTs, then you'll need to switch back to the 3D X server VT before running VirtualGL. Those are not limitations of VirtualGL. They're limitations of the system architecture. You can try the EGL back end in the evolving VirtualGL 3.0 builds as a possible workaround, since it doesn't require a 3D X server (but it doesn't yet support as many corner cases as the GLX back end does, so there may be some rare applications that don't like it.) |
Thank you for your help. It's very helpful to know that the "active VT" is a concept that's relevant for VirtualGL. I think this software is really great, and I'm excited to be able to use it. Thank you again for all your help! |
Yes, it's relevant for hardware-accelerated OpenGL in general. I don't have a deep understanding as to why, but I'm guessing that it has something to do with DRM, i.e. with how hardware-accelerated OpenGL bypasses the X server. |
Hello, I am trying to set up VirtualGL & TurboVNC to create a headless linux desktop for CAD that I can access remotely. I installed VirtualGL according to the directions, and I chose to not restrict access to the vgluser group, and I did not disable the XTEST extension. I'm installing this on Ubuntu 20.04. I'm able to run my CAD software locally on the machine.
xdpyinfo -display :0
returns 3599 lines and exit code 0. The first lines of its output are at [1]./opt/VirtualGL/bin/glxinfo -display :0 -c
just hangs after the following output (it generates a newline after thename of display: :0
):GLXgears hangs in the same way:
I included an
ltrace
ofglxinfo
at [2] and a the tail of anstrace
at [3]:I've searched for information on debugging this further, but it seems like most users get errors rather than hanging situations. The only idea I had (beyond collecting debugging info) is that since I'm using gnome desktop and gdm, I disabled auto-login, so that my first tty (which is running gdm) is just sitting at the login prompt; however, this didn't allow
vglrun
to do its thing.I would appreciate any advice you have to collect more debugging information, or if I'm misinterpreting the nature of the issue. Thank you for your time.
[1]
[2]
[3]
The text was updated successfully, but these errors were encountered: