Bug 1242

Summary: Virtualbox daemon won't start
Product: Mageia Reporter: Doug Laidlaw <laidlaws>
Component: RPM PackagesAssignee: Mageia Bug Squad <bugsquad>
Status: RESOLVED FIXED QA Contact:
Severity: normal    
Priority: Normal CC: davidwhodgins, dmorganec
Version: Cauldron   
Target Milestone: ---   
Hardware: i586   
OS: Linux   
Whiteboard:
Source RPM: virtualbox-4.0.6-3.mga1.src.rpm CVE:
Status comment:

Description Doug Laidlaw 2011-05-11 12:43:44 CEST
Description of problem:
I can set up a virtual machine entry, but when it comes time to start it, the daemon won't start.  . The usual command returns [FAILED]  dmesg has:

VirtualBox[4836]: segfault at 128 ip b528a695 sp bffedbf0 error 4 in libQtNetwork.so.4.7.3[b51d9000+125000]


Version-Release number of selected component (if applicable):
4.0.6_OSE


How reproducible:
Start a VM.

Steps to Reproduce:
1.
2.
3.
Comment 1 Ahmad Samir 2011-05-11 22:17:49 CEST
Just to be sure, the system is fully up-to-date? as qt has been updated to 4.7.3 recently.

Keywords: (none) => NEEDINFO

Comment 2 Doug Laidlaw 2011-05-12 01:29:55 CEST
Yes I have all updates, which stopped a day or so ago, including all Qt4 packages now being 4.7.3.  The error message mentions a Qt 4.7.3 file as being involved.
Comment 3 Ahmad Samir 2011-05-12 02:40:37 CEST
Yes, I noticed; just hoped it's an update issue.

Keywords: NEEDINFO => (none)

Comment 4 Ahmad Samir 2011-05-12 03:04:02 CEST
I think you should file a bug report upstream, follow the instructions here: http://www.virtualbox.org/wiki/Bugtracker
Comment 5 Doug Laidlaw 2011-05-12 04:45:32 CEST
Will do.  There seems to be no similar bug reported.
Comment 6 Doug Laidlaw 2011-05-12 12:10:03 CEST
Ticket #8879 filed with VirtualBox Bugtracker.

http://www.virtualbox.org/ticket/8879#preview.
Comment 7 Doug Laidlaw 2011-05-15 05:56:21 CEST
As at Sun 15 May 0400 UTC, after installing all updates, the VB daemon has started normally.  I haven't changed anything specific to this bug.
Comment 8 Ahmad Samir 2011-05-15 08:02:22 CEST
Weird... but good outcome, of course....
Comment 9 Doug Laidlaw 2011-05-15 08:24:07 CEST
Weird that nobody else reported it.  I mentioned it on the Mandriva ng, and it seemed to be happening there as well.  I wonder if there was an updated KDE library...
Comment 10 D Morgan 2011-05-17 15:42:43 CEST
closing as it is working ( comment #9 )

Status: NEW => RESOLVED
CC: (none) => dmorganec
Resolution: (none) => FIXED

Comment 11 Doug Laidlaw 2011-05-17 15:58:41 CEST
Yes, this bug seems to be fixed.

After a re-install, the daemon won't start and there is nothing in dmesg. I suspect a leftover config file.
Comment 12 Doug Laidlaw 2011-05-17 16:46:57 CEST
It started after I rebuilt the virtualbox kernel with dkms.  My NVidia setup is in much the same position.  Going to rebuild it next.
Comment 13 Doug Laidlaw 2011-05-17 17:17:48 CEST
NVidia fixed as well. New bug against DKMS: 1311.
Comment 14 Ahmad Samir 2011-05-17 17:59:29 CEST
(In reply to comment #12)
> It started after I rebuilt the virtualbox kernel with dkms.  My NVidia setup is
> in much the same position.  Going to rebuild it next.

Manually you mean? the rebuild should happen automatically when you boot a new kernel for the first time (assuming the matching kernel-devel is installed).
Comment 15 Doug Laidlaw 2011-05-17 18:09:14 CEST
Yes, manually.  Rebuilt all 3 dkms modules (vboxdrv, vboxadditions, nvidia-current) manually.  See Bug # 1311.  Modules are being built automatically, and shown on the tree, but they don't work.  At bug # 1311, I have included history from this bug.
Comment 16 Ahmad Samir 2011-05-17 18:40:58 CEST
I have two different installs and never had a problem like that with DKMS.... a possible scenario is you installed a kernel, then uninstalled it, then reinstalled it, that wouldn't have cleaned the dkms tree (the situation is better starting from kernel-2.6.38.6-1.mga1 as it removes the relevant /lib/modules/ dirs when removing a kernel)....
Comment 17 Doug Laidlaw 2011-05-17 19:01:26 CEST
It didn't happen with my initial install.  I don't recall uninstalling then reinstalling a kernel.  I regularly install an update then uninstall the previous kernel, but so does everybody.

Neither of these packages was part of the initial reinstall. The nvidia packages were installed with XFdrake, the VirtualBox with urpmi.  I think that for a while I ran the system on the DVD before installing updates.  The dkms files are in /var/lib and weren't deleted, while /usr was formatted.  By removing the modules, I may have cleaned out /var/lib/dkms, but I don't know enough for that to be more than a guess.

I suppose that I can wait to see what happens with the next kernel upgrade.
Comment 18 Doug Laidlaw 2011-05-17 19:28:23 CEST
There may be something in what I just wrote.  When I installed Virtualbox, there were messages that the modules already existed and needed to be removed first.  They would have had to be in /var/  Normally, that doesn't cause any problem.  I noticed that the command used included rpm-upgrade or similar, no doubt to get around that message.
Comment 19 Dave Hodgins 2011-05-17 21:03:20 CEST
My understanding is that the rpm package virtualbox-guest-additions should
only be installed if Mageia has been installed as a guest under VirtualBox.

It should not be installed on a host system, that will then be running
VirtualBox.

CC: (none) => davidwhodgins

Comment 20 Doug Laidlaw 2011-05-17 21:51:21 CEST
Thanks for explaining that, Dave.  Under Mandriva, I have always used the package from Oracle, and the guest additions are in fact installed on the guest OS.  Surely the guest additions should match the host engine?

But that is a side issue to the bug.  I think that the answer to the bug was not cleaning out /var/lib/dkms  There used to be a problem with MySQL caused by an existing log file in /var/log.  I usually want to keep /var/lib/mysql and /var/www, so I don't format /var, but it would be good practice to do so.
Comment 21 Doug Laidlaw 2011-05-20 07:35:33 CEST
(In reply to comment #20)
> Thanks for explaining that, Dave.  Under Mandriva, I have always used the
> package from Oracle, and the guest additions are in fact installed on the guest
> OS.  Surely the guest additions should match the host engine?
> 
> But that is a side issue to the bug.  I think that the answer to the bug was
> not cleaning out /var/lib/dkms  There used to be a problem with MySQL caused by
> an existing log file in /var/log.  I usually want to keep /var/lib/mysql and
> /var/www, so I don't format /var, but it would be good practice to do so.

Just installed a kernel upgrade with no problems.  IMO, that identifies the cause.
Comment 22 Ahmad Samir 2011-06-16 18:57:00 CEST
(Thanks for the feedback).