[vbox-dev] VirtualBox 4.1.8 segfault

[Klaus Espenlaub]

On 06.07.2012 16:17, isdtor wrote:
> I'm observing the following behaviour when running a custom setup
> script to create a new instance:
>
> + VBoxManage createvm --name vbox-image --ostype WindowsXP --register
> --basefolder /var/tmp/virtualbox
> Virtual machine 'vbox-image' is created and registered.
> UUID: 28ae47e2-37c5-40f3-bc46-d98b40ee158b
> Settings file: '/var/tmp/virtualbox/vbox-image/vbox-image.vbox'
> + VBoxManage storagectl vbox-image --name 'IDE Controller' --add ide
> --controller PIIX4
> + VBoxManage modifyvm vbox-image --memory 1024 --vram 32 --audio alsa
> --accelerate2dvideo on  --accelerate3d on --nic1 bridged
> --bridgeadapter1 eth0 --hda
> /var/tmp/virtualbox/vbox-image/vbox-image.vdi --usb on --usbehci on
> + pgrep VirtualBox
> + VirtualBox --startvm vbox-image
> vbox: line 204: 25651 Segmentation fault      VirtualBox --startvm ${vmname}
> + exit 0
>
> Environment: CentOS6.2 (fully patched),
> VirtualBox-4.1-4.1.18_78361_rhel6-1.x86_64,  NVidia Quadro 600, elrepo
> nvidia driver 295.59-1.
>
> When I remove the "--accelerate3d on" option, the software starts normally.

Could be a bug in the 3D code in VirtualBox, or somewhere in the host's 
3D driver. Hard to tell without knowing what 3D application you're 
running, and without having VBox.log.

> I have not tested older versions, but I have used VirtualBox on
> CentOS5 x86_64 on very similar hardware since version 3.1, and this
> has never happened before.

CentOS5 has a much older X server than CentOS 6.2, so there can be 
significant differences.

> Curiously, I do not get a core dump, despite unlimiting core size
> explicitly in environment and script.

You need to enable cores for setuid executables...
https://www.virtualbox.org/wiki/Core_dump

Klaus