Re: [PATCH] qemu-kvm: Eliminate _kvm_arch_init_vcpu

[Marcelo Tosatti <mtosatti@redhat.com>]

On Mon, Apr 23, 2012 at 10:42:25AM +0200, Jan Kiszka wrote:
> On 2012-03-09 23:17, Jan Kiszka wrote:
> > From: Jan Kiszka <jan.kiszka@siemens.com>
> > 
> > There is no need for an explicit kvm_arch_reset_vcpu on arch cpu init
> > anymore, kvm_init_vcpu does it already. Call the remaining
> > kvm_update_ioport_access directly from kvm_arch_init_vcpu.
> > 
> > Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
> > ---
> > 
> > This depends on the TPR switch-over patch.
> > 
> >  qemu-kvm-x86.c    |    7 -------
> >  target-i386/kvm.c |    4 +---
> >  2 files changed, 1 insertions(+), 10 deletions(-)
> > 
> > diff --git a/qemu-kvm-x86.c b/qemu-kvm-x86.c
> > index 159779c..7de51dc 100644
> > --- a/qemu-kvm-x86.c
> > +++ b/qemu-kvm-x86.c
> > @@ -77,13 +77,6 @@ int kvm_set_pit2(KVMState *s, struct kvm_pit_state2 *ps2)
> >      return kvm_vm_ioctl(s, KVM_SET_PIT2, ps2);
> >  }
> >  
> > -static int _kvm_arch_init_vcpu(CPUState *env)
> > -{
> > -    kvm_arch_reset_vcpu(env);
> > -
> > -    return kvm_update_ioport_access(env);
> > -}
> > -
> >  #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
> >  int kvm_arch_set_ioport_access(unsigned long start, unsigned long size,
> >                                 bool enable)
> > diff --git a/target-i386/kvm.c b/target-i386/kvm.c
> > index 461734d..91110c1 100644
> > --- a/target-i386/kvm.c
> > +++ b/target-i386/kvm.c
> > @@ -345,8 +345,6 @@ static void cpu_update_state(void *opaque, int running, RunState state)
> >      }
> >  }
> >  
> > -static int _kvm_arch_init_vcpu(CPUState *env);
> > -
> >  int kvm_arch_init_vcpu(CPUState *env)
> >  {
> >      struct {
> > @@ -360,7 +358,7 @@ int kvm_arch_init_vcpu(CPUState *env)
> >      uint32_t signature[3];
> >      int r;
> >  
> > -    r = _kvm_arch_init_vcpu(env);
> > +    r = kvm_update_ioport_access(env);
> >      if (r < 0) {
> >          return r;
> >      }
> 
> Marcelo, what is the state of understanding why this change can cause
> troubles in autotest? As I wrote offlist, I can neither reproduce nor
> understand how it can cause any regression (I suspect it rather triggers
> a so far hidden issue).
> 
> Jan

Still no understanding of previous failures, but results are
consistently OK now. Maybe the gods are happy with us now :)

Applied, thanks.