I have searched why I always get "[drm] could not load kernel module "vboxvideo". While that I found out, that the OSE-Version lacks this feature. Reproducible: Always Steps to Reproduce: 1. merge xf86-video-virtualbox 2. enable dri 3. read the messages, no DRI Actual Results: no acceleration Expected Results: DRI should work by loading the kernel module "vboxvideo"
Please provide info: - which version of virtualbox - emerge --info
I am using Virtualbox 2.2.0 on a Mac. (The newest version available). I have already tried upgrading all VirtualBox related things (the additions and the input and video driver). emerge --info follows as attachment.
Created attachment 189998 [details] emerge --info from guest emerge --info from a VBox guest running in VBox 2.2.0 on a MacBook Pro (Late 2008)
So if I understand this correctly virtualbox-ose is released without the DRI functionality. That's not something we can fix, either use virtualbox-bin, don't use DRI or complain at virtualbox upstream until they release a "fixed" version. Closing bug as we can't fix this.
No, virtualbox-bin didn't work either. I think I found the problem. "export_modules" doesn't export the "drm" directory, and a whole lot more. I'll try to put together an ebuild once this works for me...
I was able to "fix" this problem by... * unpacking the OSE-Tarball * go to $unpackdir/src/VBox/Additions/linux/drm * run kmk * go to $unpackdir/out/linux.x86/release/bin/additions/src/vboxvideo_drm * make * make install So it IS included in the OSE. But it doesn't get build by the ebuild it seems... I hope this helps somehow...
Created attachment 191265 [details] Ebuild
Created attachment 191269 [details, diff] Patch
I was able to create a patch for export_modules that causes it to export the drm directory, and modified the ebuild to make use of it and install. My only problem with this approach is that the drm module is being installed as part of the guest-additions, and it feels like it should be part of xf86-video-virtualbox when the drm use-flag is enabled.(In reply to comment #8)
I'd like to point out that this bug is not to be resolved upstream, but gets fixed with the attached ebuild and patch by Josh Williams. In contrast to comment #9, though, I'd keep the patch for this ebuild (and not in the xf86 video package) by applying the patch only if the X use flag is enabled.
