vmread | A library to read/write memory to Windows on KVM | Monitoring library

trying to build ffmpeg on ubuntu 18.10

https://trac.ffmpeg.org/wiki/CompilationGuide/Ubuntu

first install dependencies

I have worked out a toy hypervisor with VMX in Linux, and it can work well with my testing VM. Now I am trying to test suspend and resume my testing VM with my toy hypervisor. I did followings to suspend the VM.

The VM starts in real-mode, and then goes to protected mode. It call a wrmsr, which will cause VMEXIT. The hypervisor gets the VMEXIT, then saves the CPU contexts (gprs, segment regs, descriptor tables, etc) and the VM's memory into a disk file, and quit. (VMXON/VMCS regions are not cleared).

Now, restart the VM loader, it will read the disk file, save the CPU contexts into the VMCS fields, and populate the VM's memory. Then tell the hypervisor to start the VM by calling 'vmlaunch'. (Before that, the hypervisor will call vmxon, vmclear, vmptrld, and vmwrite the VMCS fields and guest CPU contexts). My expectation is the VM will resume from its RIP saved before.
But it failed with 'invalid guest state'

Below is the VMCS dump just at the WRMSR VMEXIT.

I'm currently working on automating a Win32 UI application that cannot be altered. So far my approach is using the standard message queue of the target application to inject my inputs. I've gotten quite far with that:

• "Clicking" buttons using WM_COMMAND works
• Reading the names of tabs via TCM_GETITEMA and activating them via virtual mouse clicks WM_LBUTTONDOWN/WM_LBUTTONUP works
• Reading enabled/disabled states of controls works

Where I'm stuck, however, is modifying the text of an editable ComboBox and its Edit control. I try using the WM_SETTEXT message like so:

Logically thinking, writes to any CPU stored variable should be faster than the respective memory operation, because there are no chances of a cache miss. These states are cached in the CPU, and they do not change any CPU state before the next VMLAUNCH/VMRESUME operation. Therefore, they should be faster than an equivalent operation on a memory address

This question arises when looking at different virtualization solutions provided by AMD and Intel. Intel has mandated that all changes to the VMCS data structure should always go through a VMREAD/VMWRITE interface and not through regular memory R/W operations. However, AMD does not pose any such restriction, and its VMCB region is modified by regular memory operations.

The gains in Intel approach should be faster VMExit/VMResume times vs AMD. However, Intel would loose out on the flexibility front with adding new instructions.

However, in reality, VMREAD/VMWRITE operations are slower than regular memory operations. This does not make any sense to me.