On Wed, May 06, 2020 at 06:09:23PM +0200, Paolo Bonzini wrote: > On 06/05/20 18:06, Peter Xu wrote: > > On Wed, May 06, 2020 at 07:10:32AM -0400, Paolo Bonzini wrote: > >> kvm_x86_ops.set_dr6 is only ever called with vcpu->arch.dr6 as the > >> second argument, and for both SVM and VMX the VMCB value is kept > >> synchronized with vcpu->arch.dr6 on #DB; we can therefore remove the > >> read accessor. > >> > >> For the write accessor we can avoid the retpoline penalty on Intel > >> by accepting a NULL value and just skipping the call in that case. > >> > >> Suggested-by: Sean Christopherson <sean.j.christopherson@xxxxxxxxx> > >> Signed-off-by: Paolo Bonzini <pbonzini@xxxxxxxxxx> > > (I think this patch and the previous one seem to be the same as the previous > > version. Anyway...) > > Yes, I placed them here because they are needed to solve the SVM bugs in > patch 8. Sorry for not adding your Reviewed-by. That's not a problem to me. :) Instead I'm more afraid of not noticing some trivial difference in the patch comparing to the last one. Thanks, -- Peter Xu
