Re: [PATCH v4 4/6] KVM: X86: Implement PV IPIs send hypercall

[Paolo Bonzini <pbonzini@redhat.com>]

On 20/07/2018 18:28, Wanpeng Li wrote:
> +a0: ipi_bitmap low 64 bits
> +a1: ipi_bitmap high 64 bits
> +a2: the lowest APIC ID in bitmap
> +a3: APIC ICR
> +
> +The hypercall lets a guest send multicast IPIs at most can handle
> +128 vCPUs per hypercall on 64-bit machines and 64 vCPUs per hypercall
> +on 32-bit machines.
> +
> +Returns 0 if successfully delivery the IPIs and 1 if discarded.

This description does not mention what happens in 32-bit mode.

> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index 2b812b3..016c7e2 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -6691,6 +6691,41 @@ static void kvm_pv_kick_cpu_op(struct kvm *kvm, unsigned long flags, int apicid)
>  	kvm_irq_delivery_to_apic(kvm, NULL, &lapic_irq, NULL);
>  }
>  
> +/*
> + * Return 0 if successfully added and 1 if discarded.
> + */
> +static int kvm_pv_send_ipi(struct kvm *kvm, unsigned long ipi_bitmap_low,
> +		unsigned long ipi_bitmap_high, int min, int vector, int op_64_bit)
> +{
> +	int i;
> +	struct kvm_apic_map *map;
> +	struct kvm_vcpu *vcpu;
> +	struct kvm_lapic_irq irq = {
> +		.delivery_mode = APIC_DM_FIXED,
> +		.vector = vector,
> +	};
> +
> +	rcu_read_lock();
> +	map = rcu_dereference(kvm->arch.apic_map);
> +
> +	for_each_set_bit(i, &ipi_bitmap_low, BITS_PER_LONG) {
> +		vcpu = map->phys_map[min + i]->vcpu;
> +		if (!kvm_apic_set_irq(vcpu, &irq, NULL))
> +			return 1;
> +	}
> +
> +	if (op_64_bit) {
> +		for_each_set_bit(i, &ipi_bitmap_high, BITS_PER_LONG) {
> +			vcpu = map->phys_map[min + i + BITS_PER_LONG]->vcpu;
> +			if (!kvm_apic_set_irq(vcpu, &irq, NULL))
> +				return 1;
> +		}
> +	}

The second loop processes the second argument, and it should always run,
even in 32-bit mode.  However, the phys_map index should be min + i + 32
in 32-bit mode and min + i + 64 in 64-bit mode.  (Using BITS_PER_LONG in
the for_each_set_bit length is not a bug instead; you could write it
explicitly as 32 in 32-bit mode, and 64 in 64-bit mode, but I think it's
a little bit more efficient if it's constant).

Paolo
> +