[PATCH 1/9] KVM: arm64: Handle huge mappings for np-guest CMOs

[Vincent Donnefort]

On Fri, Feb 28, 2025 at 06:54:40PM +0000, Quentin Perret wrote:
> On Friday 28 Feb 2025 at 10:25:17 (+0000), Vincent Donnefort wrote:
> > diff --git a/arch/arm64/kvm/hyp/nvhe/mem_protect.c b/arch/arm64/kvm/hyp/nvhe/mem_protect.c
> > index 19c3c631708c..a796e257c41f 100644
> > --- a/arch/arm64/kvm/hyp/nvhe/mem_protect.c
> > +++ b/arch/arm64/kvm/hyp/nvhe/mem_protect.c
> > @@ -219,14 +219,24 @@ static void guest_s2_put_page(void *addr)
> >  
> >  static void clean_dcache_guest_page(void *va, size_t size)
> >  {
> > -	__clean_dcache_guest_page(hyp_fixmap_map(__hyp_pa(va)), size);
> > -	hyp_fixmap_unmap();
> > +	while (size) {
> 
> Nit: not a problem at the moment, but this makes me mildly worried if
> size ever became non-page-aligned, could we make the code robust to
> that?

The fixmap doesn't handle !ALIGNED adresses. (I have a patch in the tracing
series to cover that though). So wonder if that really makes sense to handle
unaligned size while it wouldn't work with unaligned va anyway?

Perhaps just a WARN_ON() then?

> 
> > +		__clean_dcache_guest_page(hyp_fixmap_map(__hyp_pa(va)),
> > +					  PAGE_SIZE);
> > +		hyp_fixmap_unmap();
> > +		va += PAGE_SIZE;
> > +		size -= PAGE_SIZE;
> > +	}
> >  }
> >  
> >  static void invalidate_icache_guest_page(void *va, size_t size)
> >  {
> > -	__invalidate_icache_guest_page(hyp_fixmap_map(__hyp_pa(va)), size);
> > -	hyp_fixmap_unmap();
> > +	while (size) {
> > +		__invalidate_icache_guest_page(hyp_fixmap_map(__hyp_pa(va)),
> > +					       PAGE_SIZE);
> > +		hyp_fixmap_unmap();
> > +		va += PAGE_SIZE;
> > +		size -= PAGE_SIZE;
> > +	}
> >  }
> >  
> >  int kvm_guest_prepare_stage2(struct pkvm_hyp_vm *vm, void *pgd)
> > -- 
> > 2.48.1.711.g2feabab25a-goog
> >