Re: [PATCH v13 02/10] KVM: x86/vmx: Make vmx_set_intercept_for_msr() non-static and expose it

[Xu, Like]

Hi Sean,

On 2020/9/29 11:13, Sean Christopherson wrote:
> On Sun, Jul 26, 2020 at 11:32:21PM +0800, Like Xu wrote:
> > It's reasonable to call vmx_set_intercept_for_msr() in other vmx-specific
> > files (e.g. pmu_intel.c), so expose it without semantic changes hopefully.
> I suppose it's reasonable, but you still need to state what is actually
> going to use it.

Sure, I will add more here that
one of its usage is to pass through LBR-related msrs later.

> > Signed-off-by: Like Xu <like.xu@xxxxxxxxxxxxxxx>
> > ---
> >   arch/x86/kvm/vmx/vmx.c | 4 ++--
> >   arch/x86/kvm/vmx/vmx.h | 2 ++
> >   2 files changed, 4 insertions(+), 2 deletions(-)
> >
> > diff --git a/arch/x86/kvm/vmx/vmx.c b/arch/x86/kvm/vmx/vmx.c
> > index dcde73a230c6..162c668d58f5 100644
> > --- a/arch/x86/kvm/vmx/vmx.c
> > +++ b/arch/x86/kvm/vmx/vmx.c
> > @@ -3772,8 +3772,8 @@ static __always_inline void vmx_enable_intercept_for_msr(unsigned long *msr_bitm
> >   	}
> >   }
> >   
> > -static __always_inline void vmx_set_intercept_for_msr(unsigned long *msr_bitmap,
> > -			     			      u32 msr, int type, bool value)
> > +__always_inline void vmx_set_intercept_for_msr(unsigned long *msr_bitmap,
> > +					 u32 msr, int type, bool value)
> >   {
> >   	if (value)
> >   		vmx_enable_intercept_for_msr(msr_bitmap, msr, type);
> > diff --git a/arch/x86/kvm/vmx/vmx.h b/arch/x86/kvm/vmx/vmx.h
> > index 0d06951e607c..08c850596cfc 100644
> > --- a/arch/x86/kvm/vmx/vmx.h
> > +++ b/arch/x86/kvm/vmx/vmx.h
> > @@ -356,6 +356,8 @@ void vmx_update_host_rsp(struct vcpu_vmx *vmx, unsigned long host_rsp);
> >   int vmx_find_msr_index(struct vmx_msrs *m, u32 msr);
> >   int vmx_handle_memory_failure(struct kvm_vcpu *vcpu, int r,
> >   			      struct x86_exception *e);
> > +void vmx_set_intercept_for_msr(unsigned long *msr_bitmap,
> > +			      u32 msr, int type, bool value);
> This completely defeats the purpose of __always_inline.
My motivation is to use vmx_set_intercept_for_msr() in pmu_intel.c,
which helps to extract pmu-specific code from vmx.c

I assume modern compilers will still make it inline even in this way,
or do you have a better solution for this?

Please let me if you have more comments on the patch series.

Thanks,
Like Xu
> >   
> >   #define POSTED_INTR_ON  0
> >   #define POSTED_INTR_SN  1
> > --
> > 2.21.3