Ticket #15860 (closed defect: obsolete)
RHEL/CENTOS 7.2 does not work any longer after installing guest extensions after 5.0.14 -> possibly obsolete
Reported by: | HRZunibi | Owned by: | |
---|---|---|---|
Component: | guest additions/x11/graphics | Version: | VirtualBox 5.1.4 |
Keywords: | RedHat | Cc: | raynebc |
Guest type: | Linux | Host type: | Windows |
Description
Redhat 7.2 (and may other versions) crashes login screen when installing guest extensions above 5.0.14 (until 5.1.14). After rebooting before logging in I see message "Oh no! Smomething went wrong. A problem has occured and the system can't recover. Please log out and try again." - But I didn't log in before - so the messsage does not match the problem... But it is a freshly installed and patched dev. system of RHEL 7.2. A coworker told me, that this also happens under CENTOS 7.2. Clean shutdown via ACPI is possible.
Attachments
Change History
Changed 6 years ago by HRZunibi
-
attachment
VBox.log_install_ext
added
Changed 6 years ago by HRZunibi
-
attachment
VBox.log_with_ext
added
vbox.log from first (unsusccessful) boot after installing guest extensions
comment:2 Changed 6 years ago by frank
And is there any change if you increaste the amount of video memory to 128MB (you have 3D enabled for your guest)?
comment:3 Changed 6 years ago by HRZunibi
3d acceleration is enabled and video memory is currently 32MB. I just increased it to 128MB and retried (from snapshot of machine). Problem still exists.
Changed 6 years ago by HRZunibi
-
attachment
Xorg.0.log_b4_reboot
added
Logfile after installation of extensions before reboot
comment:4 Changed 6 years ago by michael
Does this still happen with the latest Additions test build<1>? If it does, can you try looking for SELinux alerts?
sealert -a /var/log/audit/audit.log > selinux.log
This will generate a log file - selinux.log - that you can inspect and attach to this ticket.
comment:5 Changed 6 years ago by HRZunibi
Hello!
I just installed the fresh released 5.1.6 version of guest extensions. Installation was successful and everything worked until I logged out. Before I was able to log in again the well known "Oh no! ..." screen appeared. That's what sealert dumps out:
found 0 alerts in /var/log/audit/audit.log
Changed 6 years ago by HRZunibi
-
attachment
messages.txt
added
messages from the last reboot before installation until failure
comment:6 Changed 6 years ago by michael
Can you reproduce this with a fresh installation of CentOS 7.2? If so, please give exact reproduction steps. (So far I have failed to reproduce it.)
comment:7 Changed 6 years ago by HRZunibi
Yes I was able to reproduce it with a fresh install. I made some snapshots of the machine when doing this. So I can send you the complete machine packed as an appliance / upload it to our cloud provider. Can I send you the link and credentials via another way (not publically readable) ie. PM?
Peter.
comment:8 Changed 6 years ago by michael
You can e-mail me at michael dot thayer at oracle. How long will the link be usable? I may not get the chance to look this week.
comment:9 Changed 6 years ago by HRZunibi
The link to our cloud has been sent by PM. Link (w. PW) is valid until Oct, 15th.
comment:10 Changed 6 years ago by michael
I tried mailing you back, will try here: I got the link but could not see any password.
comment:11 Changed 6 years ago by michael
Had a quick look, and it looks to me (based on the time stamps on the log files) as if CentOS does not even try to start gdm. Have to work out why not.
comment:12 Changed 6 years ago by michael
It is in runlevel 3, whatever that means when systemd is running. When I switch to runlevel 5 gdm starts, stops (crashes?) once and starts again. At that point I can log in.
comment:13 Changed 6 years ago by michael
Following a systemd FAQ on freedesktop.org<1> I did the following to change the runlevel, and the machine happily starts GDM.
# ln -sf /usr/lib/systemd/system/graphical.target /etc/systemd/system/default.target
I assume I am missing something here? Or perhaps it is due to the fact that I am running a development version of VirtualBox rather than 5.1.6. (I did not update the Additions in the virtual machine though.)
<1> https://www.freedesktop.org/wiki/Software/systemd/FrequentlyAskedQuestions/
comment:14 Changed 6 years ago by HRZunibi
When I log in in runlevel 3 and give "init 5" to enable gdm login I get the well known but not loved "Oops!" and only a black screen after press the logout button. This is the behaviour since VBox 5.0.16. My version of VBox is 5.1.6r110634. Using VBox Extensions 5.0.14 everything works fine.
comment:15 Changed 6 years ago by michael
Does disabling 3D make a difference? If I read your log file correctly you are using the Windows software renderer ("GDI generic renderer") rather than a graphics card driver for 3D.
comment:16 Changed 6 years ago by HRZunibi
After disabling 3D acceleration it starts correctly.
comment:17 Changed 6 years ago by michael
- Out of interest, are you using RDP or something similar? That might explain why the generic renderer is being used.
- The current development snapshot (not the 5.1 build!) of the Guest Additions<1> should automatically detect this and disable 3D in the guest. It would be nice if you could test it and see if it solves your problem without disabling 3D in the machine settings. (In fact the Linux/X11 Additions code in the development snapshot is not very different to the 5.1 branch at the moment.)
comment:18 Changed 6 years ago by HRZunibi
RDP is enabled but not in use. I've never tried to run a VBox machine inside a RDP session. Sounds funny but I think it does not make really sense (for GUI VBOXes)...
I'll give the TestBuild a try...
comment:19 Changed 6 years ago by HRZunibi
I just tried VBox 5.1.7r111271 with the matching extensions and it works. gdm login is possible. Re-enabling 3d acceleration results in a crash. So just before it only runs with disabled 3D-accelleration. Maybe this can be mentioned in the FAQ and/or docs...
comment:20 Changed 6 years ago by michael
A crash of the VirtualBox process on the host? If so, would you be able to provide us with a crash trace and/or a mini-dump<1>? I'm afraid that we currently can't provide anywhere to upload a mini-dump too, but if you can provide somewhere we can download it to, please send me an e-mail at michael dot thayer at oracle.
<1> https://www.virtualbox.org/wiki/Core_dump#WindowsMiniDumps
comment:21 Changed 6 years ago by HRZunibi
HellO!
It's not a crash on the host (sorry for the misunderstanding). It's just the "Oh no! Something has gone wrong."-Window on the guest side.
comment:23 Changed 6 years ago by michael
Could you please check the result of VBoxClient --check3d (i.e. VBoxClient --check3d || echo "No 3D") on your guest machine when 3D is enabled with the latest Additions, from a text console or whatever? It should say "No 3D". If it does, perhaps you could take a look at the /sbin/rcvboxadd service script to see why the VirtualBox 3D libraries are still being set up?
comment:24 Changed 6 years ago by HRZunibi
echo "No 3D"" returns nothing (so 3D is enabled when enabling it in the VirtualBox Manager. |
"/sbin/rcvboxadd status" returna "The VirtualBox Additions are currently running". "lsmod|grep -i vbox" just shows vboxguest and vboxsf. "/sbin/rcvboxadd-x11 status" returns nothing
comment:25 Changed 5 years ago by michael
- Summary changed from RHEL/CENTOS 7.2 does not work any longer after installing guest extensions after 5.0.14 to RHEL/CENTOS 7.2 does not work any longer after installing guest extensions after 5.0.14 -> possibly obsolete
Is this still an issue with the current version?
comment:26 Changed 5 years ago by michael
If it is, please try a current test build.
comment:27 Changed 5 years ago by michael
- Status changed from new to closed
- Resolution set to obsolete
vbox.log when installing the guest extensions