>>> On 23.02.17 at 10:41, <haozhong.zh...@intel.com> wrote: > The current implementation of nested VMX cannot work without HAP.
Of course the better route would be to fix the actual problem, the more that - according to other feedback you've got elsewhere - this apparently is a regression. Nevertheless I can see you perhaps not having the time to do so, so as a band aid it's likely fine. However ... > --- a/xen/arch/x86/hvm/hvm.c > +++ b/xen/arch/x86/hvm/hvm.c > @@ -4123,7 +4123,7 @@ static int hvmop_set_param( > * Remove the check below once we have > * shadow-on-shadow. > */ > - if ( cpu_has_svm && !paging_mode_hap(d) && a.value ) > + if ( (cpu_has_svm || cpu_has_vmx) && !paging_mode_hap(d) && a.value ) ... you want to simply drop the cpu_has_svm check here instead of adding to it, as with neither SVM nor VMX available execution will never come here. Jan _______________________________________________ Xen-devel mailing list Xen-devel@lists.xen.org https://lists.xen.org/xen-devel