Hi Lorenzo, On Tue, 22 Sep 2026 19:07:53 +0100, Oliver Upton <[email protected]> wrote: [...] > I'd favor doing something like below and sticking the comment inline in > the ioctl handler. Unless I'm missing something blatantly obvious, I > don't see why the x86 or s390 pre-conditions can't be tested early too. [...] > diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c [...] > +int kvm_arch_pre_fault_allowed(struct kvm_vcpu *vcpu) > +{ > + if (!kvm_vcpu_initialized(vcpu)) > + return -ENOEXEC; > + > + return 0; > +}
The same goes for arm64's pKVM check: kvm_arch_vcpu_pre_fault_memory() returns -EOPNOTSUPP for is_protected_kvm_enabled() only after vcpu_load(), so on a pKVM host the ioctl loads a vCPU the hypervisor may not have created yet before rejecting it. Could that check move into kvm_arch_pre_fault_allowed() as well? Cheers, /fuad
