Re: [PATCH] KVM: arm64: Inject UNDEF when host is executing an smc with imm16 != 0

[Sebastian Ene <sebastianene@google.com>]

On Tue, Mar 24, 2026 at 02:18:04PM +0000, Vincent Donnefort wrote:
> On Tue, Mar 24, 2026 at 02:09:50PM +0000, Vincent Donnefort wrote:
> > On Tue, Mar 24, 2026 at 02:06:40PM +0000, Vincent Donnefort wrote:
> > > On Tue, Mar 24, 2026 at 01:57:28PM +0000, Sebastian Ene wrote:
> > > > The ARM Service Calling Convention (SMCCC) specifies that the function
> > > > identifier and parameters should be passed in registers, leaving the
> > > > 16-bit immediate field of the SMC instruction un-handled.
> > > > Currently, our pKVM handler ignores the immediate value, which could lead
> > > > to non-compliant software relying on implementation-defined behavior.
> > > > Enforce the host kernel running under pKVM to use an immediate value
> > > > of 0 by decoding the ISS from the ESR_EL2 and inject an undefined
> > > > instruction exception back to the caller.
> > > > 
> > > > Signed-off-by: Sebastian Ene <sebastianene@google.com>
> > > > ---
> > > >  arch/arm64/kvm/hyp/nvhe/hyp-main.c | 7 ++++++-
> > > >  1 file changed, 6 insertions(+), 1 deletion(-)
> > > > 
> > > > diff --git a/arch/arm64/kvm/hyp/nvhe/hyp-main.c b/arch/arm64/kvm/hyp/nvhe/hyp-main.c
> > > > index e7790097db93..ff6a90a4a4c7 100644
> > > > --- a/arch/arm64/kvm/hyp/nvhe/hyp-main.c
> > > > +++ b/arch/arm64/kvm/hyp/nvhe/hyp-main.c
> > > > @@ -756,13 +756,18 @@ static bool handle_host_mte(u64 esr)
> > > >  void handle_trap(struct kvm_cpu_context *host_ctxt)
> > > >  {
> > > >  	u64 esr = read_sysreg_el2(SYS_ESR);
> > > > +	u16 imm16;
> > > >  
> > > >  	switch (ESR_ELx_EC(esr)) {
> > > >  	case ESR_ELx_EC_HVC64:
> > > >  		handle_host_hcall(host_ctxt);
> > > >  		break;
> > > >  	case ESR_ELx_EC_SMC64:
> > > > -		handle_host_smc(host_ctxt);
> > > > +		imm16 = esr & U16_MAX;
> > > > +		if (!imm16)
> > > 
> > > if (ESR_ELx_xVC_IMM_MASK(esr)) ?
> > > 
> > > Also, I can't find void inject_undef64(void); I think you need a vcpu for that?
> > 
> > Ah my bad, handle_host_mte() introduced it!
> > 
> > > 
> > > > +			handle_host_smc(host_ctxt);
> > > > +		else
> > > > +			inject_undef64();
> 
> Could it be better to return SMCCC_RET_NOT_SUPPORTED?

Right, I will do that since it seems this is the correct approach.

> 
> > > >  		break;
> > > >  	case ESR_ELx_EC_IABT_LOW:
> > > >  	case ESR_ELx_EC_DABT_LOW:
> > > > -- 
> > > > 2.53.0.983.g0bb29b3bc5-goog
> > > > 

Thanks,
Sebastian