Re: [PATCH v4 25/27] KVM: s390: arm64: Implement vCPU IOCTLs

[Marc Zyngier]

On Mon, 06 Jul 2026 09:52:25 +0100,
Steffen Eiden <seiden@xxxxxxxxxxxxx> wrote:
> 
> +static void adjust_pc(struct kvm_vcpu *vcpu)
> +{
> +	if (vcpu_get_flag(vcpu, INCREMENT_PC)) {
> +		kvm_skip_instr(vcpu);
> +		vcpu_clear_flag(vcpu, INCREMENT_PC);
> +	}
> +}

Can you clarify the semantics of adjust_pc() here? arm64 also deals
with exceptions in the same code, and I wonder how you deal with this.

> +
> +static void arm_vcpu_run(struct kvm_vcpu *vcpu)
> +{
> +	struct kvm_sae_block *sae_block = &vcpu->arch.sae_block;
> +
> +	adjust_pc(vcpu);
> +
> +	local_irq_disable();
> +	guest_enter_irqoff();
> +	local_irq_enable();
> +
> +	sae_block->icptr = 0;
> +
> +	sae64a(sae_block);
> +
> +	local_irq_disable();
> +	guest_exit_irqoff();
> +	local_irq_enable();
> +}
> +
> +/** kvm_arch_vcpu_ioctl_run() - run arm64 vCPU
> + *
> + * Execute arm64 guest instructions using SAE.
> + *
> + * Returns:
> + * 1 enter the guest (should not be observed by userspace)
> + * 0 exit to userspace
> + * < 0 exit to userspace, where the return value indicates n error
> + *
> + *
> + */
> +int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu)
> +{
> +	DECLARE_KERNEL_FPU_ONSTACK32(fpu_save);
> +	struct kvm_run *kvm_run = vcpu->run;
> +	int ret;
> +
> +	if (kvm_run->exit_reason == KVM_EXIT_MMIO) {
> +		ret = kvm_handle_mmio_return(vcpu);
> +		if (ret <= 0)
> +			return ret;
> +	}
> +
> +	vcpu_load(vcpu);
> +
> +	kernel_fpu_begin(&fpu_save, KERNEL_FPC | KERNEL_VXR);
> +	load_vx_regs((vcpu->arch.ctxt.vregs));
> +
> +	if (!vcpu->wants_to_run) {
> +		ret = -EINTR;
> +		goto out;
> +	}
> +
> +	kvm_sigset_activate(vcpu);
> +
> +	might_fault();
> +
> +	ret = 1;
> +	do {
> +		if (signal_pending(current)) {
> +			kvm_run->exit_reason = KVM_EXIT_INTR;
> +			ret = -EINTR;
> +			continue;
> +		}
> +
> +		if (need_resched())
> +			schedule();
> +
> +		if (ret > 0)
> +			ret = check_vcpu_requests(vcpu);
> +
> +		vcpu->arch.sae_block.icptr = 0;
> +
> +		arm_vcpu_run(vcpu);
> +
> +		ret = handle_exit(vcpu);
> +
> +	} while (ret > 0);
> +
> +	kvm_sigset_deactivate(vcpu);
> +out:
> +	if (unlikely(vcpu_get_flag(vcpu, INCREMENT_PC)))
> +		adjust_pc(vcpu);

arm64 has the following statements:

	if (unlikely(vcpu_get_flag(vcpu, PENDING_EXCEPTION) ||
		     vcpu_get_flag(vcpu, INCREMENT_PC)))
		kvm_call_hyp(__kvm_adjust_pc, vcpu);

It isn't clear to me why you can afford not to deal with pending
exceptions when returning to userspace.

Thanks,

	M.

-- 
Jazz isn't dead. It just smells funny.