On Wed, Sep 13, 2023, Binbin Wu wrote:
> Introduce a new interface get_untagged_addr() to kvm_x86_ops to untag
> the metadata from linear address. Call the interface in linearization
> of instruction emulator for 64-bit mode.
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index ea48ba87dacf..e03313287816 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -8308,6 +8308,15 @@ static void emulator_vm_bugged(struct x86_emulate_ctxt *ctxt)
> kvm_vm_bugged(kvm);
> }
>
> +static gva_t emulator_get_untagged_addr(struct x86_emulate_ctxt *ctxt,
> + gva_t addr, unsigned int flags)
> +{
> + if (!kvm_x86_ops.get_untagged_addr)
> + return addr;
> +
> + return static_call(kvm_x86_get_untagged_addr)(emul_to_vcpu(ctxt), addr, flags);
> +}
> +
> static const struct x86_emulate_ops emulate_ops = {
> .vm_bugged = emulator_vm_bugged,
> .read_gpr = emulator_read_gpr,
> @@ -8352,6 +8361,7 @@ static const struct x86_emulate_ops emulate_ops = {
> .leave_smm = emulator_leave_smm,
> .triple_fault = emulator_triple_fault,
> .set_xcr = emulator_set_xcr,
> + .get_untagged_addr = emulator_get_untagged_addr,
Oh, and for posterity because I've now thought about this on two separate occasions:
We _could_ nullify .get_untagged_addr() if LAM isn't supported, e.g. to save a
RETPOLINE when applicable, but that would require making emulate_ops non-const,
and so I think it's worth eating the indirect call. I suppose we could gate it
on CONFIG_ADDRESS_MASKING, but then we'd have to carry a bunch of #ifdefs in the
VMX code, so I don't think that's worth doing either.
