Linux Kernels 5.13.10 and older worked stably, maybe 25% slower than with direct hardware Virtualization. However, after upgrading my VBox Guest VM to linux 5.14.9 it's having kernel Oops in do_sync_core quite regularly. Either during boot or shortly after. Sometimes the stack is printed on screen, but most often it just freezes with even the caret not blinking. In this case only one line is printed to VBox.log:
Code: Select all
00:00:58.129967 VMMDev: vmmDevHeartbeatFlatlinedTimer: Guest seems to be unresponsive. Last heartbeat received 4 seconds ago
With the help of "downgrade" utility I figured out that 5.13.10 was the last good package and 5.13.12 and later were bad.
I even created a minimal VM for my colleagues to compare the setups: https://drive.google.com/file/d/1KD7ixa ... sp=sharing
After boot it automatically starts to build a package from AUR (the one I was building the first time I encountered the bug). This is how I was able to track it to "Windows Hypervisor Platform". Uninstalling it makes the guest work stably, but that's not an option.
Reducing the number of Guest CPUs from 4 to 1 makes the panic go away.
Can you try to reproduce the issue with the .ova above and back me up when I report this to kernel.org? Windows Hypervisor Platform has to be installed. Not sure what's the needed hardware.
Currently installed package on guest: linux 5.14.9
Host VirtualBox 6.1.26
Host Windows 10 Professional 64-bit (Build 19043)
video https://drive.google.com/file/d/1I6y0XK ... sp=sharing
VBoxLogs.zip, cpuz-report.zip, and console.zip with Oops details are attached