On Fri, Feb 21, 2025 at 04:53:26PM +0000, Peter Maydell wrote:
> On Fri, 21 Feb 2025 at 14:02, Wei Liu <li...@linux.microsoft.com> wrote:
> >
> > Hi,
> >
> > Microsoft's Linux Systems Group developed a Linux driver for the Microsoft
> > Hypervisor (MSHV for short). The driver is being upstreamed. The first
> > supported VMM is Cloud Hypervisor. QEMU will be the second supported
> > VMM.
> >
> > The plan is to write an mshv accelerator in QEMU. The accelerator is still 
> > in
> > the works.
> >
> > MSHV doesn't emulate instructions. VMMs are supposed to bring their own
> > instruction emulator. The path we've chosen is to reuse what's already in 
> > QEMU.
> > The instruction emulator in HVF looks good for what we need.
> >
> > This patch series attempts to make the instruction emulator in HVF a common
> > component for the i386 target. It removes HVF specific code by either using 
> > a
> > set of hooks or moving it to better locations. The new incoming MSHV
> > accelerator will implement the hooks, and where necessary, enhance the 
> > emulator
> > and / or add new hooks.
> 
> If you want to make the hvf decoder more widely used you might want
> to look at this old patch to it that was never applied (issues in
> code review not addressed by the submitter):
> 
> https://lore.kernel.org/qemu-devel/CAFEAcA8yaBOD3KXc-DY94oqzC5wkCENPkePgVCybqR=9nmd...@mail.gmail.com/
> 
> which is trying to fix a problem where an overlong string of
> prefix bytes causes the decoder to misbehave.
> 

Thanks for the information.

> (PS: if in the future you should ever find yourself wanting to do an
> equivalent "decode loads/stores the hypervisor doesn't handle"
> for Arm, use decodetree, not a hand-rolled decoder...)
> 

Noted. Yep, we have plans to add ARM64 support in the future.

Thanks,
Wei.

> thanks
> -- PMM

Reply via email to