[PATCH v4 10/13] KVM: arm64: nVHE: Support CONFIG_CFI_CLANG at EL2

[Pierre-Clément Tosi]

On Mon, Jun 03, 2024 at 03:45:30PM +0100, Will Deacon wrote:
> On Wed, May 29, 2024 at 01:12:16PM +0100, Pierre-Clément Tosi wrote:
> > diff --git a/arch/arm64/kvm/hyp/nvhe/hyp-init.S b/arch/arm64/kvm/hyp/nvhe/hyp-init.S
> > index d859c4de06b6..b1c8977e2812 100644
> > --- a/arch/arm64/kvm/hyp/nvhe/hyp-init.S
> > +++ b/arch/arm64/kvm/hyp/nvhe/hyp-init.S
> > @@ -5,6 +5,7 @@
> >   */
> >  
> >  #include <linux/arm-smccc.h>
> > +#include <linux/cfi_types.h>
> >  #include <linux/linkage.h>
> >  
> >  #include <asm/alternative.h>
> > @@ -267,8 +268,11 @@ SYM_CODE_END(__kvm_handle_stub_hvc)
> >  
> >  /*
> >   * void __pkvm_init_switch_pgd(phys_addr_t pgd, void *sp, void (*fn)(void));
> > + *
> > + * SYM_TYPED_FUNC_START() allows C to call this ID-mapped function indirectly
> > + * using a physical pointer without triggering a kCFI failure.
> >   */
> > -SYM_FUNC_START(__pkvm_init_switch_pgd)
> > +SYM_TYPED_FUNC_START(__pkvm_init_switch_pgd)
> >  	/* Turn the MMU off */
> >  	pre_disable_mmu_workaround
> >  	mrs	x9, sctlr_el2
> 
> I still think this last hunk should be merged with the earlier patch
> fixing up the prototype of __pkvm_init_switch_pgd().

Unfortunately, this is not possible because

  SYM_TYPED_FUNC_START(__pkvm_init_switch_pgd)

makes the assembler generate an unresolved symbol for the function type, which
the compiler only generates (from the C declaration) if it compiles with kCFI so
that moving this hunk to an earlier patch results in a linker error:

  ld.lld: error: undefined symbol: __kvm_nvhe___kcfi_typeid___pkvm_init_switch_pgd

OTOH, moving it to a later patch triggers a kCFI (runtime) panic.

As a result, this hunk *must* be part of this patch.

> 
> With that:
> 
> Acked-by: Will Deacon <will at kernel.org>

As I haven't followed your suggestion, I'll ignore this.