Re: [PATCH] kvm/x86 : set meaningful return value

[Paolo Bonzini]

On 28/09/2018 18:41, Peng Hao wrote:
> From: Peng Hao <peng.h...@zte.com.cn>
> 
> kvm_irq_delivery_to_apic_fast()-->
>     kvm_apic_map_get_dest_lapic()-->
>         kvm_apic_disabled_lapic_found()
> kvm_apic_map_get_dest_lapic return with bitmap==0 and dst[i]==NULL,
> then (*r == -1) will be returned to qemu and "KVM: injection failed,
> MSI lost(Operation not permitted)" will be recorded in qemu log. The
> output is puzzling.

This makes sense, but I am not sure that ENXIO is any better...

In fact, in this case I think it's okay to just return zero.  What about
this fix:

diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c
index b756f1237ce3..d02515b8018f 100644
--- a/arch/x86/kvm/lapic.c
+++ b/arch/x86/kvm/lapic.c
@@ -957,14 +957,14 @@ bool kvm_irq_delivery_to_apic_fast(struct kvm
*kvm, struct kvm_lapic *src,
        map = rcu_dereference(kvm->arch.apic_map);

        ret = kvm_apic_map_get_dest_lapic(kvm, &src, irq, map, &dst, &bitmap);
-       if (ret)
+       if (ret) {
+               *r = 0;
                for_each_set_bit(i, &bitmap, 16) {
                        if (!dst[i])
                                continue;
-                       if (*r < 0)
-                               *r = 0;
                        *r += kvm_apic_set_irq(dst[i]->vcpu, irq, dest_map);
                }
+       }

        rcu_read_unlock();
        return ret;

Paolo

> Signed-off-by: Peng Hao <peng.h...@zte.com.cn>
> ---
>  arch/x86/kvm/lapic.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c
> index fbb0e6d..a8896b3 100644
> --- a/arch/x86/kvm/lapic.c
> +++ b/arch/x86/kvm/lapic.c
> @@ -944,7 +944,7 @@ bool kvm_irq_delivery_to_apic_fast(struct kvm *kvm, 
> struct kvm_lapic *src,
>       int i;
>       bool ret;
>  
> -     *r = -1;
> +     *r = -ENXIO;
>  
>       if (irq->shorthand == APIC_DEST_SELF) {
>               *r = kvm_apic_set_irq(src->vcpu, irq, dest_map);
>