RE: [PATCH] kvm: x86: fix RSM when PCID is non-zero

[Christopherson, Sean J]

On Thu, Dec 21, 2017 at 07:30:35AM -0800, Andy Lutomirski wrote:
> On Thu, Dec 21, 2017 at 3:56 AM, Paolo Bonzini <pbonzini@xxxxxxxxxx> wrote:
> > rsm_load_state_64() and rsm_enter_protected_mode() load CR3, then
> > CR4 & ~PCIDE, then CR0, then CR4.
> >
> > However, setting CR4.PCIDE fails if CR3[11:0] != 0.  It's probably easier
> > in the long run to replace rsm_enter_protected_mode() with an emulator
> > callback that sets all the special registers (like KVM_SET_SREGS would
> > do).  For now, set the PCID field of CR3 only after CR4.PCIDE is 1.
>
> Out of curiosity, has anyone ever tested RSM returning to PAE mode?  I
> *think* it's supposed to restore the PDPTR registers directly rather
> than reloading them from the memory pointed to by CR3, and it doesn't
> look like the new or old code does it.

Yes, PDPTRs are saved/loaded to/from SMRAM, at least on Intel hardware.