[GH-ISSUE #640] Grafics acceleration with --x11=xpra #443

Open
opened 2026-05-05 05:52:28 -06:00 by gitea-mirror · 6 comments
Owner

Originally created by @fkarl423 on GitHub (Jul 19, 2016).
Original GitHub issue: https://github.com/netblue30/firejail/issues/640

When running firefox with x11 sandboxing using xpra, the hardware acceleration seems broken.

It seems like the application is using VMware grafics drivers instead of the native one.

about:support shows that with xpra it says
Adapter Description: VMware, Inc. -- Gallium 0.4 on llvmpipe (LLVM 3.8, 256 bits)
without it:
Adapter Description: nouveau -- Gallium 0.4 on NVE4

Like this it is barely useable.
Is there a way to change the used grafics driver in xpra?
I would suggest to atomatically set this bz firejail to make it useable on low end hardware.

Originally created by @fkarl423 on GitHub (Jul 19, 2016). Original GitHub issue: https://github.com/netblue30/firejail/issues/640 When running firefox with x11 sandboxing using xpra, the hardware acceleration seems broken. It seems like the application is using VMware grafics drivers instead of the native one. about:support shows that with xpra it says Adapter Description: VMware, Inc. -- Gallium 0.4 on llvmpipe (LLVM 3.8, 256 bits) without it: Adapter Description: nouveau -- Gallium 0.4 on NVE4 Like this it is barely useable. Is there a way to change the used grafics driver in xpra? I would suggest to atomatically set this bz firejail to make it useable on low end hardware.
gitea-mirror added the
enhancement
label 2026-05-05 05:52:28 -06:00
Author
Owner

@netblue30 commented on GitHub (Jul 20, 2016):

Unfortunately there is no 3D acceleration support in xpra. They have a setting in /etc/xpra/xpra.conf (opengl = auto) but this one picks up the software implementation (llvmpipe). It will work fine for browsing and text editing on powerful computers, but that's about it - definitely not good for games.

<!-- gh-comment-id:233988803 --> @netblue30 commented on GitHub (Jul 20, 2016): Unfortunately there is no 3D acceleration support in xpra. They have a setting in /etc/xpra/xpra.conf (opengl = auto) but this one picks up the software implementation (llvmpipe). It will work fine for browsing and text editing on powerful computers, but that's about it - definitely not good for games.
Author
Owner

@msva commented on GitHub (Jul 28, 2016):

@netblue30 actually, not only 3d but also 2d acceleration.
And, yes I've same with chrome-based apps like slack. When they tries to use acceleration there are rendering artefacts appearing.

<!-- gh-comment-id:235966116 --> @msva commented on GitHub (Jul 28, 2016): @netblue30 actually, not only 3d but also 2d acceleration. And, yes I've same with chrome-based apps like slack. When they tries to use acceleration there are rendering artefacts appearing.
Author
Owner

@l29ah commented on GitHub (Aug 8, 2016):

I've tried to use virtualgl with firejail with --x11=xpra: seems working for glxgears, but firefox renders a gray rectangle that can be closed with Enter (an "already running" message box?) and that's it.

<!-- gh-comment-id:238166282 --> @l29ah commented on GitHub (Aug 8, 2016): I've tried to use virtualgl with firejail with --x11=xpra: seems working for glxgears, but firefox renders a gray rectangle that can be closed with Enter (an "already running" message box?) and that's it.
Author
Owner

@netblue30 commented on GitHub (Aug 9, 2016):

I'll try virtualgl, maybe it can solve the problem.

<!-- gh-comment-id:238531070 --> @netblue30 commented on GitHub (Aug 9, 2016): I'll try virtualgl, maybe it can solve the problem.
Author
Owner

@chiraag-nataraj commented on GitHub (Oct 3, 2018):

@netblue30, did you ever try virtualgl?

<!-- gh-comment-id:426656309 --> @chiraag-nataraj commented on GitHub (Oct 3, 2018): @netblue30, did you ever try virtualgl?
Author
Owner

@totaam commented on GitHub (Jul 20, 2019):

There is some confusion here.
The opengl switch is used mostly by the xpra client to enable opengl accelerated rendering.
On the server, it is only used for querying the adapter (with v3, use "noprobe" to skip this).
If you want to enable accelerated rendering on the server, start your client applications using virtualgl.
More info here: Xpra OpenGL
This will mean opening up an X11 channel from the container back to the host for VirtualGL, but this is still much safer than allowing client applications to talk to X11 directly. At the end of the day, real acceleration requires access to the hardware...

<!-- gh-comment-id:513442732 --> @totaam commented on GitHub (Jul 20, 2019): There is some confusion here. The _opengl_ switch is used mostly by the xpra client to enable opengl accelerated rendering. On the server, it is only used for querying the adapter (with v3, use "noprobe" to skip this). If you want to enable accelerated rendering on the server, start your client applications using virtualgl. More info here: [Xpra OpenGL](https://xpra.org/trac/wiki/Usage/OpenGL) This will mean opening up an X11 channel from the container back to the host for VirtualGL, but this is still much safer than allowing client applications to talk to X11 directly. At the end of the day, real acceleration requires access to the hardware...
Sign in to join this conversation.
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference: github-starred/firejail#443
No description provided.