On Wed, 9 Sep 2020 at 16:36, Vitaly Kuznetsov <vkuznets@xxxxxxxxxx> wrote:
>
> Wanpeng Li <kernellwp@xxxxxxxxx> writes:
>
> > From: Wanpeng Li <wanpengli@xxxxxxxxxxx>
> >
> > Moving svm_complete_interrupts() into svm_vcpu_run() which can align VMX
> > and SVM with respect to completing interrupts.
> >
> > Suggested-by: Sean Christopherson <sean.j.christopherson@xxxxxxxxx>
> > Cc: Paul K. <kronenpj@xxxxxxxxxxxxxxxxxxx>
> > Signed-off-by: Wanpeng Li <wanpengli@xxxxxxxxxxx>
> > ---
> > arch/x86/kvm/svm/svm.c | 4 ++--
> > 1 file changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/arch/x86/kvm/svm/svm.c b/arch/x86/kvm/svm/svm.c
> > index c61bc3b..74bcf0a 100644
> > --- a/arch/x86/kvm/svm/svm.c
> > +++ b/arch/x86/kvm/svm/svm.c
> > @@ -2938,8 +2938,6 @@ static int handle_exit(struct kvm_vcpu *vcpu, fastpath_t exit_fastpath)
> > if (npt_enabled)
> > vcpu->arch.cr3 = svm->vmcb->save.cr3;
> >
> > - svm_complete_interrupts(svm);
> > -
> > if (is_guest_mode(vcpu)) {
> > int vmexit;
> >
> > @@ -3530,6 +3528,8 @@ static __no_kcsan fastpath_t svm_vcpu_run(struct kvm_vcpu *vcpu)
> > SVM_EXIT_EXCP_BASE + MC_VECTOR))
> > svm_handle_mce(svm);
> >
> > + svm_complete_interrupts(svm);
> > +
> > vmcb_mark_all_clean(svm->vmcb);
> > return exit_fastpath;
> > }
>
> This seems to be the right thing to do, however, the amount of code
> between kvm_x86_ops.run() and kvm_x86_ops.handle_exit() is non-trivial,
> hope it won't blow up in testing...
>
> Reviewed-by: Vitaly Kuznetsov <vkuznets@xxxxxxxxxx>
>
> One more thing:
>
> VMX version does
>
> vmx_complete_interrupts(vmx);
> if (is_guest_mode(vcpu))
> return EXIT_FASTPATH_NONE;
>
> and on SVM we analyze is_guest_mode() inside
> svm_exit_handlers_fastpath() - should we also change that for
> conformity?
Agreed, will do in v2.
Wanpeng
