| Summary: | X fails to start in VM when using VBoxVideo X driver (auto-selected by service_harddrake) | ||
|---|---|---|---|
| Product: | Mageia | Reporter: | Martin Whitaker <mageia> |
| Component: | RPM Packages | Assignee: | Mageia tools maintainers <mageiatools> |
| Status: | RESOLVED FIXED | QA Contact: | |
| Severity: | normal | ||
| Priority: | Normal | CC: | LpSolit, ghibomgx, marja11, tmb |
| Version: | Cauldron | ||
| Target Milestone: | --- | ||
| Hardware: | All | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Source RPM: | CVE: | ||
| Status comment: | |||
| Attachments: |
X server log showing error
journal from boot where X server failed to start |
||
|
Marja Van Waes
2018-10-27 21:10:53 CEST
CC:
(none) =>
marja11 @tmb, what is the correct X driver to use in VirtualBox clients these days? CC:
(none) =>
tmb vboxvideo should be the correct one, if it fails we need to figure out why do you have anything more in the logs, dmesg or full journal ? Created attachment 10448 [details]
journal from boot where X server failed to start
X server log already attached (not very helpful). Here is the accompanying journal.
We can't expect QA to test ISOs with this bug. Unless someone has time to work on this, I am going to revert the change in ldetect.lst. We can revert it for now while I debug it so we can get beta1 built vbox switched back to modesetting in ldetect-lst-0.5.3-2.mga7 If I use Hyper-V for paravirtualisation as suggested in bug 51553 comment 55, then I can see what happens when the vboxvideo driver is used. The following error is thrown again and again: EXT4-fs error (device sda1): ext4_lookup:1577: inode #10301: comm Xorg: deleted inode referenced: 2523 If I use modesetting instead of vboxvideo with Hyper-V, then it boots correctly and no error is thrown. I don't know if that helps. (In reply to Frédéric Buclin from comment #9) > If I use Hyper-V for paravirtualisation as suggested in bug 51553 comment > 55 I meant bug 21553 comment 55. Apparently vboxvideo is not supported well for xorg 1.19 and 1.20. There are a few reports about it in the virtualbox tracker. Probably the fix is in the making. Theoretically vboxvideo should provide also 3d hardware acceleration, though it's not known at which API OpenGL compatibility level; in the past this level was pretty low to be compatible with the latest OpenGL software requirement (like for WebGL in the browsers, GEarth, etc.); beyond this in a few guest VM where there is still xorg older than 1.18, current vboxvideo starts, but I get the screen messed up, e.g. not scaled correctly, and with the mouse pointer in a messed position. Apparently virtualbox has also the possibility to use the "vmware" video driver, using the command: VBoxManage modifyvm mageiaguest --graphicscontroller vmsvga but indeed it's incomplete/experimental and doesn't work properly. CC:
(none) =>
ghibomgx Booting with xdriver=vboxvideo is still broken. Booting with xdriver=modesetting still works. The Live ISOs now default to allowing X to auto-detect the driver, which appears to use a combination of vboxvideo + modesetting. So setting this to fixed. Resolution:
(none) =>
FIXED |
Created attachment 10432 [details] X server log showing error Testing a Live ISO build from cauldron. On boot, service_harddrake reports: service_harddrake[556]: switch X.org driver from 'modesetting' to 'vboxvideo' (The graphics card 'InnoTek Systemberatung GmbH|VirtualBox Graphics Adapter' is no longer supported by driver 'modesetting') But then X fails to start with the error message: [ 24.714] (EE) No devices detected. [ 24.714] (EE) Fatal server error: [ 24.714] (EE) no screens found(EE) If I boot with xdriver=modesetting, X starts OK. In 6.1, the modesetting driver was automatically selected.