OSDev.org

Looking for testers for Bochs VMX support.
I have working version of Bochs with VMX support including most of the required functionality.
I hope anybody have something meaningful to play with on this version !

Thanks,
Stanislav

Will VMX become the default for x86?

Love4Boobies wrote:Will VMX become the default for x86?

If you like it could

Stanislav

Well I for one think it's a good idea I'd like to test... Do you have precompiled binaries or do I have to build it?

Is it like something taking advantage AMD-V or Intel VT-x like the way VirtualBox or Hyper-V does? Or is it some kind of set of specialized drivers for a guest OS to take advantage of special features?

If it works on x86_64 I'll test it, otherwise I can't

Troy Martin wrote:Is it like something taking advantage AMD-V or Intel VT-x like the way VirtualBox or Hyper-V does? Or is it some kind of set of specialized drivers for a guest OS to take advantage of special features?

No, it will help you ONLY if you writing your own Hyper-V engine and have no why it doesn't work.
This is complete emulation of VMX hardware, not getting advantage of host VMX hardware (take Xen if you need it).
If you still need it - I'll post sources + precompiled Windows binaries here in the middle of next week.

Stanislav

Oh, I thought Bochs was taking advantage of VMX not emulating it That's still nice and I am up for testing; still, why not take advantage of this technology on compatible processors?

I see two ways of testing this:
- by running your own VMX code (which in my case, has yet to be written, as well as a buch of other dependencies)
- by installing a VM inside bochs (which everybody can do).

what would be fun is installing virtualPC in bochs in virtualPC in bochs

Combuster wrote:what would be fun is installing virtualPC in bochs in virtualPC in bochs

I don't know about fun - slow would perhaps be a better choice of description

Cheers,
Adam

Love4Boobies wrote:Oh, I thought Bochs was taking advantage of VMX not emulating it That's still nice and I am up for testing; still, why not take advantage of this technology on compatible processors?

If this happened you would lose your all-important Bochs processor state dump. when running native code, it's much harder for an emulator to keep a precise state.

If somebody still interested:

Attached VMX emulation patch.
Apply to CVS from 23 Jan 2009.
You are welcome to try to run Xen or kvm inside Bochs, this time I am pretty optimistic and believe it should run !
Waiting for any feedback.

Thanks,
Stanislav

----------------------------------------------------------------------
Patch name: VMX emulation
Date: 23 Jan 2009
Status:
VMX emulation support for Bochs.
Almost complete Core 2 Duo VMX support, not implemented yet:
- dual-monitor treatment of SMIs not implemented
- APIC virtualization not implemented yet
- NMI virtualization not implemented yet
- VMENTER to not-active state not implemented yet
Also I made my own VMCS layout, not compatible with any existing
hardware + VMCS revision id (0xE) not necessary related to
existing Intel hardware.

Instructions:
To patch, apply patch to main bochs directory:
> patch -p1 < vmx-patch
Automatically generated files skipped from this patch to
reduce its absolute size.
To regenerate configure file:
> autoconf configure.in > configure

Will patch right away. Won't this be included in the CVS?

Love4Boobies wrote:Will patch right away. Won't this be included in the CVS?

Of course, depends on your inputs here. Hope it works fine and I will merge it into the CVS.

Thanks,
Stanislav

One major bug fixed - MSR save/restore on VMENTER/VMEXIT was completely broken.
Plus few minor fixes.
New patch: