Xserver and Linux containers

Christian Stussak stussak at mfo.de
Sat May 16 09:33:11 PDT 2015


There is currently a huge hype around Linux containers, Docker and such and there are quite a few attempts to run GUI apps inside Linux containers. This is usually done by mounting the host’s X socket and sometimes a few other files/directories into the container (such as /dev/shm, /dev/dri/...). In order to get 3D hardware acceleration working (over DRI, I think), they install the very same version of the graphics driver into the container and the host although the Linux distro inside the container might be completely different from the hosts distro (different library versions, distro specific patches etc.)

Here are a few links to get an idea of these approaches:
https://github.com/thewtex/docker-opengl-nvidia/blob/master/run.sh
https://github.com/yantis/docker-virtualgl

I tried it my self and it seems to work well.

My question is: Is this approach of implementing the communication between host and container X server safe in the sense that it is supposed to work like that or is it likely to break at any time because it’s just an ugly hack that works by chance? If so, what would be the correct way to tackle that problem, i.e. utilizing an X server running on the host from within the container while still having hardware acceleration for 3d.

I saw demos where this „solution“ has been used to run 3D accelerated apps remotely on Amazon EC2 web services and I am considering to try something similar with containers if this is not a dead end.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 5091 bytes
Desc: not available
URL: <http://lists.x.org/archives/xorg/attachments/20150516/47b69b59/attachment.bin>


More information about the xorg mailing list