Re: [Qemu-devel] [PATCH v5 4/4] s390x/virtio-ccw: Wire up irq routing and irqfds.

[Cornelia Huck]

On Mon, 12 May 2014 10:58:46 +0200
Christian Borntraeger <address@hidden> wrote:

> On 08/05/14 15:03, Cornelia Huck wrote:
> > Make use of the new s390 adapter irq routing support to enable real
> > in-kernel irqfds for virtio-ccw with adapter interrupts.
> > 
> > Note that s390 doesn't provide the common KVM_CAP_IRQCHIP capability, but
> > rather needs KVM_CAP_S390_IRQCHIP to be enabled. This is to ensure backward
> > compatibility.
> > 
> > Reviewed-by: Thomas Huth <address@hidden>
> > Signed-off-by: Cornelia Huck <address@hidden>
> 
> Small things, otherwise
> 
> Reviewed-by: Christian Borntraeger <address@hidden>
> 
> 
> [...]
> > --- /dev/null
> > +++ b/include/hw/s390x/adapter.h
> > @@ -0,0 +1,23 @@
> > +/*
> > + * s390 adapter definitions
> > + *
> > + * Copyright 2013 IBM Corp.
> 
> 2014 as well

ok

> 
> [...]
> > diff --git a/kvm-all.c b/kvm-all.c
> > index 5cb7f26..51983bf 100644
> > --- a/kvm-all.c
> > +++ b/kvm-all.c

> > @@ -1267,6 +1297,11 @@ int kvm_irqchip_add_msi_route(KVMState *s, 
> > MSIMessage msg)
> >      return -ENOSYS;
> >  }
> > 
> > +int kvm_irqchip_add_adapter_route(KVMState *s, AdapterInfo *adapter)
> > +{
> > +    return -ENOSYS;
> > +}
> > +
> >  static int kvm_irqchip_assign_irqfd(KVMState *s, int fd, int virq, bool 
> > assign)
> >  {
> >      abort();
> > @@ -1296,7 +1331,8 @@ static int kvm_irqchip_create(KVMState *s)
> >      int ret;
> > 
> >      if (!qemu_opt_get_bool(qemu_get_machine_opts(), "kernel_irqchip", 
> > true) ||
> > -        !kvm_check_extension(s, KVM_CAP_IRQCHIP)) {
> > +        (!kvm_check_extension(s, KVM_CAP_IRQCHIP) &&
> > +         (kvm_vm_enable_cap(s, KVM_CAP_S390_IRQCHIP, 0) < 0))) {
> >          return 0;
> >      }
> 
> Do we need a comment here and/or in target-s390x/kvm.c  why 
> kvm_halt_in_kernel_allowed does not work without breaking backward 
> compatibility  on s390?

Probably not here but...

> 
> 
> > 

> > diff --git a/target-s390x/kvm.c b/target-s390x/kvm.c
> > index b7b0edc..43859e9 100644
> > --- a/target-s390x/kvm.c
> > +++ b/target-s390x/kvm.c
> > @@ -938,6 +938,11 @@ void kvm_s390_enable_css_support(S390CPU *cpu)
> > 
> >  void kvm_arch_init_irq_routing(KVMState *s)
> >  {
> > +    if (kvm_check_extension(s, KVM_CAP_IRQ_ROUTING)) {
> > +        kvm_irqfds_allowed = true;
> > +        kvm_gsi_routing_allowed = true;
> > +        kvm_halt_in_kernel_allowed = false;
> > +    }
> 
> Do we need a comment here and/or in kvm-all why kvm_halt_in_kernel_allowed 
> does not work without breaking backward compatibility on s390?

...I'll add the following here:

/*
 * Note that while irqchip capabilities generally imply that cpustates
 * are handled in-kernel, it is not true for s390 (yet); therefore, we
 * have to override the common code kvm_halt_in_kernel_allowed setting.
 */

> 
> [...]
> 
> 
> Should Paolo ack the common kvm changes? (CCed)
> 
> Christian