[RFC PATCH 09/27] KVM: arm64: Allow using kvm_nvhe_sym() in hyp code

[David Brazdil]

On Mon, Nov 23, 2020 at 02:02:50PM +0000, 'Quentin Perret' via kernel-team wrote:
> On Monday 23 Nov 2020 at 12:57:23 (+0000), David Brazdil wrote:
> > > diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c
> > > index 882eb383bd75..391cf6753a13 100644
> > > --- a/arch/arm64/kvm/arm.c
> > > +++ b/arch/arm64/kvm/arm.c
> > > @@ -1369,7 +1369,7 @@ static void cpu_prepare_hyp_mode(int cpu)
> > >  
> > >  	params->vector_hyp_va = kern_hyp_va((unsigned long)kvm_ksym_ref(__kvm_hyp_host_vector));
> > >  	params->stack_hyp_va = kern_hyp_va(per_cpu(kvm_arm_hyp_stack_page, cpu) + PAGE_SIZE);
> > > -	params->entry_hyp_va = kern_hyp_va((unsigned long)kvm_ksym_ref(__kvm_hyp_psci_cpu_entry));
> > > +	params->entry_hyp_va = kern_hyp_va((unsigned long)kvm_ksym_ref_nvhe(__kvm_hyp_psci_cpu_entry));
> > 
> > Why is this change needed?
> 
> You mean this line specifically or the whole __kvm_hyp_psci_cpu_entry
> thing?
> 
> For the latter, it is to avoid having the compiler complain about
> __kvm_hyp_psci_cpu_entry being re-defined as a different symbol. If
> there is a better way to solve this problem I'm happy to change it -- I
> must admit I got a little confused with the namespacing along the way.

Yeah, we do need a more robust approach. It's getting out of control.

> 
> Thanks,
> Quentin
> 
> -- 
> To unsubscribe from this group and stop receiving emails from it, send an email to kernel-team+unsubscribe at android.com.
>