[v2] KVM: SVM: hyper-v: placate modpost section mismatch error

  modpost reports section mismatch errors/warnings:
WARNING: modpost: vmlinux.o: section mismatch in reference: svm_hv_hardware_setup (section: .text) -> (unknown) (section: .init.data)
WARNING: modpost: vmlinux.o: section mismatch in reference: svm_hv_hardware_setup (section: .text) -> (unknown) (section: .init.data)
WARNING: modpost: vmlinux.o: section mismatch in reference: svm_hv_hardware_setup (section: .text) -> (unknown) (section: .init.data)

Marking svm_hv_hardware_setup() as __init fixes the warnings.

I don't know why this should be needed -- it seems like a compiler
problem to me since the calling function is marked as __init.

This "(unknown) (section: .init.data)" all refer to svm_x86_ops.

Fixes: 1e0c7d40758b ("KVM: SVM: hyper-v: Remote TLB flush for SVM")
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Vineeth Pillai <viremana@linux.microsoft.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Vitaly Kuznetsov <vkuznets@redhat.com>
Cc: Sean Christopherson <seanjc@google.com>
Cc: kvm@vger.kernel.org
---
v2: also make the empty stub function be __init (Vitaly)

 arch/x86/kvm/svm/svm_onhyperv.h |    4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
  

On Wed, Feb 22, 2023, Randy Dunlap wrote:
> modpost reports section mismatch errors/warnings:
> WARNING: modpost: vmlinux.o: section mismatch in reference: svm_hv_hardware_setup (section: .text) -> (unknown) (section: .init.data)
> WARNING: modpost: vmlinux.o: section mismatch in reference: svm_hv_hardware_setup (section: .text) -> (unknown) (section: .init.data)
> WARNING: modpost: vmlinux.o: section mismatch in reference: svm_hv_hardware_setup (section: .text) -> (unknown) (section: .init.data)
> 
> Marking svm_hv_hardware_setup() as __init fixes the warnings.
> 
> I don't know why this should be needed -- it seems like a compiler
> problem to me since the calling function is marked as __init.

It's not a compiler issue.  __initdata is freed after init and so must not be
accessed by __init-less functions.

This as a changelog?

  Tag svm_hv_hardware_setup() with __init to fix a modpost warning as the
  non-stub implementation accesses __initdata (svm_x86_ops), i.e. would
  generate a use-after-free if svm_hv_hardware_setup() were actually invoked
  post-init.  The helper is only called from svm_hardware_setup(), which is
  also __init, i.e. other than the modpost warning, lack of __init is benign.

With that (in case Paolo grabs this directly):

Reviewed-by: Sean Christopherson <seanjc@google.com>
  
> This "(unknown) (section: .init.data)" all refer to svm_x86_ops.
> 
> Fixes: 1e0c7d40758b ("KVM: SVM: hyper-v: Remote TLB flush for SVM")
> Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
> Cc: Vineeth Pillai <viremana@linux.microsoft.com>
> Cc: Paolo Bonzini <pbonzini@redhat.com>
> Cc: Vitaly Kuznetsov <vkuznets@redhat.com>
> Cc: Sean Christopherson <seanjc@google.com>
> Cc: kvm@vger.kernel.org
> ---
> v2: also make the empty stub function be __init (Vitaly)
> 
>  arch/x86/kvm/svm/svm_onhyperv.h |    4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff -- a/arch/x86/kvm/svm/svm_onhyperv.h b/arch/x86/kvm/svm/svm_onhyperv.h
> --- a/arch/x86/kvm/svm/svm_onhyperv.h
> +++ b/arch/x86/kvm/svm/svm_onhyperv.h
> @@ -30,7 +30,7 @@ static inline void svm_hv_init_vmcb(stru
>  		hve->hv_enlightenments_control.msr_bitmap = 1;
>  }
>  
> -static inline void svm_hv_hardware_setup(void)
> +static inline __init void svm_hv_hardware_setup(void)
>  {
>  	if (npt_enabled &&
>  	    ms_hyperv.nested_features & HV_X64_NESTED_ENLIGHTENED_TLB) {
> @@ -84,7 +84,7 @@ static inline void svm_hv_init_vmcb(stru
>  {
>  }
>  
> -static inline void svm_hv_hardware_setup(void)
> +static inline __init void svm_hv_hardware_setup(void)
>  {
>  }
>  
>
  

Randy Dunlap <rdunlap@infradead.org> writes:

> modpost reports section mismatch errors/warnings:
> WARNING: modpost: vmlinux.o: section mismatch in reference: svm_hv_hardware_setup (section: .text) -> (unknown) (section: .init.data)
> WARNING: modpost: vmlinux.o: section mismatch in reference: svm_hv_hardware_setup (section: .text) -> (unknown) (section: .init.data)
> WARNING: modpost: vmlinux.o: section mismatch in reference: svm_hv_hardware_setup (section: .text) -> (unknown) (section: .init.data)
>
> Marking svm_hv_hardware_setup() as __init fixes the warnings.
>
> I don't know why this should be needed -- it seems like a compiler
> problem to me since the calling function is marked as __init.
>
> This "(unknown) (section: .init.data)" all refer to svm_x86_ops.
>
> Fixes: 1e0c7d40758b ("KVM: SVM: hyper-v: Remote TLB flush for SVM")
> Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
> Cc: Vineeth Pillai <viremana@linux.microsoft.com>
> Cc: Paolo Bonzini <pbonzini@redhat.com>
> Cc: Vitaly Kuznetsov <vkuznets@redhat.com>
> Cc: Sean Christopherson <seanjc@google.com>
> Cc: kvm@vger.kernel.org
> ---
> v2: also make the empty stub function be __init (Vitaly)

Thanks!

Reviewed-by: Vitaly Kuznetsov <vkuznets@redhat.com>

>
>  arch/x86/kvm/svm/svm_onhyperv.h |    4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff -- a/arch/x86/kvm/svm/svm_onhyperv.h b/arch/x86/kvm/svm/svm_onhyperv.h
> --- a/arch/x86/kvm/svm/svm_onhyperv.h
> +++ b/arch/x86/kvm/svm/svm_onhyperv.h
> @@ -30,7 +30,7 @@ static inline void svm_hv_init_vmcb(stru
>  		hve->hv_enlightenments_control.msr_bitmap = 1;
>  }
>  
> -static inline void svm_hv_hardware_setup(void)
> +static inline __init void svm_hv_hardware_setup(void)
>  {
>  	if (npt_enabled &&
>  	    ms_hyperv.nested_features & HV_X64_NESTED_ENLIGHTENED_TLB) {
> @@ -84,7 +84,7 @@ static inline void svm_hv_init_vmcb(stru
>  {
>  }
>  
> -static inline void svm_hv_hardware_setup(void)
> +static inline __init void svm_hv_hardware_setup(void)
>  {
>  }
>  
>
  

On 2/22/23 17:46, Sean Christopherson wrote:
>    Tag svm_hv_hardware_setup() with __init to fix a modpost warning as the
>    non-stub implementation accesses __initdata (svm_x86_ops), i.e. would
>    generate a use-after-free if svm_hv_hardware_setup() were actually invoked
>    post-init.  The helper is only called from svm_hardware_setup(), which is
>    also __init, i.e. other than the modpost warning, lack of __init is benign.

Done.  It's caused by the compiler deciding not to inline the function, 
probably.

Also Cc'ed stable.

Paolo
  

On Wed, 22 Feb 2023 19:32:53 +0100
Paolo Bonzini <pbonzini@redhat.com> wrote:

Maybe we can use __always_inline? I just noticed this thread today by chance.

https://lore.kernel.org/all/20210624095147.880513802@infradead.org/

> On 2/22/23 17:46, Sean Christopherson wrote:
> >    Tag svm_hv_hardware_setup() with __init to fix a modpost warning as the
> >    non-stub implementation accesses __initdata (svm_x86_ops), i.e. would
> >    generate a use-after-free if svm_hv_hardware_setup() were actually invoked
> >    post-init.  The helper is only called from svm_hardware_setup(), which is
> >    also __init, i.e. other than the modpost warning, lack of __init is benign.
> 
> Done.  It's caused by the compiler deciding not to inline the function, 
> probably.
> 
> Also Cc'ed stable.
> 
> Paolo
>
  

On Thu, Feb 23, 2023, Zhi Wang wrote:
> On Wed, 22 Feb 2023 19:32:53 +0100
> Paolo Bonzini <pbonzini@redhat.com> wrote:
> 
> Maybe we can use __always_inline? I just noticed this thread today by chance.

Using __always_inline will "fix" the problem, but it's not necessary in this case,
and in some ways it's less correct.  The noinstr case you linked is different
because the helpers in question can (and are) be used in noinstr and regular
sections, i.e. shouldn't be tagged noinstr.  In this case, svm_hv_hardware_setup()
must be called from __init functions, i.e. doesn't need to be unopinionated.

And FWIW, svm_hv_hardware_setup() really doesn't need to be inlined.

> https://lore.kernel.org/all/20210624095147.880513802@infradead.org/
> 
> > On 2/22/23 17:46, Sean Christopherson wrote:
> > >    Tag svm_hv_hardware_setup() with __init to fix a modpost warning as the
> > >    non-stub implementation accesses __initdata (svm_x86_ops), i.e. would
> > >    generate a use-after-free if svm_hv_hardware_setup() were actually invoked
> > >    post-init.  The helper is only called from svm_hardware_setup(), which is
> > >    also __init, i.e. other than the modpost warning, lack of __init is benign.
> > 
> > Done.  It's caused by the compiler deciding not to inline the function, 
> > probably.
> > 
> > Also Cc'ed stable.
> > 
> > Paolo
> > 
>