[vbox-dev] Any known problems with changing CPU core count on a Windows guest.

[Klaus Espenlaub]

Hi Rūdolfs,

On 25.04.2016 14:09, Rūdolfs Bundulis wrote:
> Hi,
>
> some time ago I was struggling with weird longjmp caused crashes when
> running a Windows guest in my framebuffer implementation that I could
> not explain (since the jumps were made from vbox code) and couldn't
> solve it here either. These crashes happened when I chaned the amount of
> vCPU cores on the guest (I had installed it with one). I blamed my code
> and for somewhat reason this happened only on Windows.

What exactly did crash? The VM manager UI? The VM UI? VBoxSVC? Or is it 
not a crash but an error message about needing VT-x or AMD-V for SMP?

All Windows versions which Microsoft still supports (Vista+) have a 
"unified HAL", which means that worst case the boot will take somewhat 
longer as Windows needs to detect some additional hardware and install 
some new drivers.

> This weekend one of the students assigned to this project also
> complained that the VirtualBox GUI crashes when he changes the amount of
> cores at Windows boot up (which made me suspicious). I wonder if there
> is any knowledge of related issues with this scenario? Maybe there is
> some code that tries to handle this in the VirtualBox GUI and that is
> why I did not see any crashes there but only in my frontend? If there is
> no such knowledge I'll just continue to investigate the issue.

Not supposed to happen. Again, what exactly crashes? Is it literally the 
VirtualBox VM manager UI (where one would change the VM settings) or is 
it later the VM UI when starting the VM?

Log files are unfortunately only partially useful when it comes to 
crashes, because the only hint about crashes generally is that the log 
file ends abruptly where it normally should continue.

Klaus