Windows 8.1 Guest crashes and aborts after install Guest Additions

[fabiohsouza]

After installing Windows 8.1 when I install Guest Additions the system stops responding and aborts.

Host: Kubuntu 14.10 64bit - Kernel 3.17.0-031700-lowlatency

VirtualBox 4.3.18

Log Error: Guest Log: VBoxMP::DxgkDdiQueryCurrentFenceNew: WARNING! :uncompleted fences, u32FenceSubmitted(1728), u32FenceCompleted(1725) u32FenceProcessed(1725)

[fabiohsouza]

Windows 10 Preview Guest with same error after install Guest Additions.

[fabiohsouza]

Obs.: I have a Windows 7 Ultimate (64bit) and Windows XP Pro (32bit) running OK on this host.

[slinde]

I have the same issue, VB 4.3.20r96996 on a debian linux host with a 3.2.0-4-amd64 kernel. Windows 8.1 guest is allocated 6GB RAM and full access to the CPU.

[beikaigun]

This appears to be related to my problem: ​https://forums.virtualbox.org/posting.php?mode=reply&f=2&t=66152

It appears to be specific to the Windows 8 WDDM display driver in the Guest Additions. You should be able to restart the VM with 3D acceleration disabled, then, through settings, rollback to the default driver. This will allow you to boot as normal with 3D enabled - however, you will loose functionality such as full screen capability.

If you go to the discussion linked above, I have a few other references to background surrounding the problem. This may be specific to interactions between the VB GA driver and at least some the linux 3.17 kernel series versions as well (I'm seeing it on Fedora 21). If I read correctly, the error is stating that there is graphical data (contained in data structures called "fences") that are not returning from being sent to the GPU. In the case of the original error listed, there would be 3 (1728-1725). This data has somehow been lost.

These also doesn't seem to be specific to a particular GPU. I'm using the Intel HD GPU packaged with a gen 3 Core i7. However, this ticket (https://www.virtualbox.org/ticket/13599) seems to refer to a similar issue with an nVidia GPU.

This ticket also suggests more fully that this is an interaction between VB and the kernel, since the author appears to have been able to replicate with a non-windows VM. Hence, even though removing the Windows 8 driver from the VM should bring back some functionality, the driver itself may not be to blame but may be hitting on something either in VB or the kernel code.

[aeichner]

Please reopen if still relevant with a recent VirtualBox release.