Re: [PATCH v3 1/3] x86/vmx: implement VMExit based guest Bus Lock detection

[Roger Pau Monné <roger.pau@xxxxxxxxxx>]

On Mon, Jul 04, 2022 at 11:27:37AM +0200, Jan Beulich wrote:
> On 01.07.2022 15:16, Roger Pau Monne wrote:
> > --- a/xen/arch/x86/hvm/vmx/vmx.c
> > +++ b/xen/arch/x86/hvm/vmx/vmx.c
> > @@ -4065,6 +4065,11 @@ void vmx_vmexit_handler(struct cpu_user_regs *regs)
> >  
> >      if ( unlikely(exit_reason & VMX_EXIT_REASONS_FAILED_VMENTRY) )
> >          return vmx_failed_vmentry(exit_reason, regs);
> > +    if ( unlikely(exit_reason & VMX_EXIT_REASONS_BUS_LOCK) )
> > +    {
> > +        perfc_incr(buslock);
> > +        exit_reason &= ~VMX_EXIT_REASONS_BUS_LOCK;
> > +    }
> 
> To cover for the flag bit, don't you also need to mask it off in
> nvmx_idtv_handling()? Or (didn't go into detail with checking whether
> there aren't any counter indications) pass the exit reason there from
> vmx_vmexit_handler(), instead of re-reading it from the VMCS?

This seem to be an existing issue with nvmx_idtv_handling(), as it
should use just the low 16bits to check against the VM Exit reason
codes.

I can send a pre-patch to fix it, could pass exit reason from
vmx_vmexit_handler(), but I would still need to cast to uint16_t for
comparing against exit reason codes, as there's a jump into the 'out'
label before VMX_EXIT_REASONS_BUS_LOCK is masked out.

I think there's a similar issue with nvmx_n2_vmexit_handler() that
doesn't cast the value to uint16_t and is called before
VMX_EXIT_REASONS_BUS_LOCK is removed from exit reason.

Thanks, Roger.