Patchwork [RFC,20/20] target-i386: move default init of cpuid_kvm_features bitmap into CPU initializer from cpudef

login
register
mail settings
Submitter Igor Mammedov
Date Aug. 10, 2012, 11:22 a.m.
Message ID <1344597756-2890-21-git-send-email-imammedo@redhat.com>
Download mbox | patch
Permalink /patch/176449/
State New
Headers show

Comments

Igor Mammedov - Aug. 10, 2012, 11:22 a.m.
Moving it inside CPU initializer from cpudef will help to split
cpu_x86_find_by_name() into default init and user settable properties.

PS:
  Is kvm_features field necessary in cpudef, what the point
  if it's almost imediately overwritten to ~0? Could it be removed
  from cpudef?

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
---
 target-i386/cpu.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
Eduardo Habkost - Aug. 10, 2012, 3:24 p.m.
On Fri, Aug 10, 2012 at 01:22:36PM +0200, Igor Mammedov wrote:
> Moving it inside CPU initializer from cpudef will help to split
> cpu_x86_find_by_name() into default init and user settable properties.
> 
> PS:
>   Is kvm_features field necessary in cpudef, what the point
>   if it's almost imediately overwritten to ~0? Could it be removed
>   from cpudef?

We could probably drop it, but: maybe better than dropping it is to
initialize it properly with all the currently-supported KVM feature
flags, instead of initializing it to ~0 and then filter the bits later.
Otherwise "-cpu enforce" would not be able to check the KVM feature bits
properly.


> 
> Signed-off-by: Igor Mammedov <imammedo@redhat.com>
> ---
>  target-i386/cpu.c | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/target-i386/cpu.c b/target-i386/cpu.c
> index 43601a3..e266792 100644
> --- a/target-i386/cpu.c
> +++ b/target-i386/cpu.c
> @@ -1093,6 +1093,9 @@ static void cpudef_2_x86_cpu(X86CPU *cpu, x86_def_t *def, Error **errp)
>      env->cpuid_7_0_ebx = def->cpuid_7_0_ebx_features;
>      env->cpuid_xlevel2 = def->xlevel2;
>  
> +    /* not supported bits will be filtered out later */
> +    env->cpuid_kvm_features = ~0;
> +
>      object_property_set_bool(OBJECT(cpu), true, "hypervisor", errp);
>  }
>  
> @@ -1175,9 +1178,6 @@ static int cpu_x86_find_by_name(X86CPU *cpu, x86_def_t *x86_cpu_def,
>  
>      cpudef_2_x86_cpu(cpu, def, errp);
>  
> -    /* not supported bits will be filtered out later */
> -    env->cpuid_kvm_features = ~0;
> -
>      for (ent = qdict_first(features); ent; ent = qdict_next(features, ent)) {
>          const QString *qval = qobject_to_qstring(qdict_entry_value(ent));
>          object_property_parse(OBJECT(cpu), qstring_get_str(qval),
> -- 
> 1.7.11.2
>
Igor Mammedov - Aug. 15, 2012, 12:23 p.m.
On Fri, 10 Aug 2012 12:24:48 -0300
Eduardo Habkost <ehabkost@redhat.com> wrote:

> On Fri, Aug 10, 2012 at 01:22:36PM +0200, Igor Mammedov wrote:
> > Moving it inside CPU initializer from cpudef will help to split
> > cpu_x86_find_by_name() into default init and user settable properties.
> > 
> > PS:
> >   Is kvm_features field necessary in cpudef, what the point
> >   if it's almost imediately overwritten to ~0? Could it be removed
> >   from cpudef?
> 
> We could probably drop it, but: maybe better than dropping it is to
> initialize it properly with all the currently-supported KVM feature
> flags, instead of initializing it to ~0 and then filter the bits later.
> Otherwise "-cpu enforce" would not be able to check the KVM feature bits
> properly.
Looking at current core, It should be doable, lets postpone it to a separate
series.
For now lets keep it initialized to ~0 and plan to improve it later.

> 
> > 
> > Signed-off-by: Igor Mammedov <imammedo@redhat.com>
> > ---
> >  target-i386/cpu.c | 6 +++---
> >  1 file changed, 3 insertions(+), 3 deletions(-)
> > 
> > diff --git a/target-i386/cpu.c b/target-i386/cpu.c
> > index 43601a3..e266792 100644
> > --- a/target-i386/cpu.c
> > +++ b/target-i386/cpu.c
> > @@ -1093,6 +1093,9 @@ static void cpudef_2_x86_cpu(X86CPU *cpu, x86_def_t *def, Error **errp)
> >      env->cpuid_7_0_ebx = def->cpuid_7_0_ebx_features;
> >      env->cpuid_xlevel2 = def->xlevel2;
> >  
> > +    /* not supported bits will be filtered out later */
> > +    env->cpuid_kvm_features = ~0;
> > +
> >      object_property_set_bool(OBJECT(cpu), true, "hypervisor", errp);
> >  }
> >  
> > @@ -1175,9 +1178,6 @@ static int cpu_x86_find_by_name(X86CPU *cpu, x86_def_t *x86_cpu_def,
> >  
> >      cpudef_2_x86_cpu(cpu, def, errp);
> >  
> > -    /* not supported bits will be filtered out later */
> > -    env->cpuid_kvm_features = ~0;
> > -
> >      for (ent = qdict_first(features); ent; ent = qdict_next(features, ent)) {
> >          const QString *qval = qobject_to_qstring(qdict_entry_value(ent));
> >          object_property_parse(OBJECT(cpu), qstring_get_str(qval),
> > -- 
> > 1.7.11.2
> > 
> 
> -- 
> Eduardo
>
Eduardo Habkost - Aug. 15, 2012, 12:32 p.m.
On Wed, Aug 15, 2012 at 02:23:46PM +0200, Igor Mammedov wrote:
> On Fri, 10 Aug 2012 12:24:48 -0300
> Eduardo Habkost <ehabkost@redhat.com> wrote:
> 
> > On Fri, Aug 10, 2012 at 01:22:36PM +0200, Igor Mammedov wrote:
> > > Moving it inside CPU initializer from cpudef will help to split
> > > cpu_x86_find_by_name() into default init and user settable properties.
> > > 
> > > PS:
> > >   Is kvm_features field necessary in cpudef, what the point
> > >   if it's almost imediately overwritten to ~0? Could it be removed
> > >   from cpudef?
> > 
> > We could probably drop it, but: maybe better than dropping it is to
> > initialize it properly with all the currently-supported KVM feature
> > flags, instead of initializing it to ~0 and then filter the bits later.
> > Otherwise "-cpu enforce" would not be able to check the KVM feature bits
> > properly.
> Looking at current core, It should be doable, lets postpone it to a separate
> series.
> For now lets keep it initialized to ~0 and plan to improve it later.

Agreed. Let's move one step at a time. The good thing is that the
cleanup is making those weird exceptions in the code easier to spot.

> 
> > 
> > > 
> > > Signed-off-by: Igor Mammedov <imammedo@redhat.com>
> > > ---
> > >  target-i386/cpu.c | 6 +++---
> > >  1 file changed, 3 insertions(+), 3 deletions(-)
> > > 
> > > diff --git a/target-i386/cpu.c b/target-i386/cpu.c
> > > index 43601a3..e266792 100644
> > > --- a/target-i386/cpu.c
> > > +++ b/target-i386/cpu.c
> > > @@ -1093,6 +1093,9 @@ static void cpudef_2_x86_cpu(X86CPU *cpu, x86_def_t *def, Error **errp)
> > >      env->cpuid_7_0_ebx = def->cpuid_7_0_ebx_features;
> > >      env->cpuid_xlevel2 = def->xlevel2;
> > >  
> > > +    /* not supported bits will be filtered out later */
> > > +    env->cpuid_kvm_features = ~0;
> > > +
> > >      object_property_set_bool(OBJECT(cpu), true, "hypervisor", errp);
> > >  }
> > >  
> > > @@ -1175,9 +1178,6 @@ static int cpu_x86_find_by_name(X86CPU *cpu, x86_def_t *x86_cpu_def,
> > >  
> > >      cpudef_2_x86_cpu(cpu, def, errp);
> > >  
> > > -    /* not supported bits will be filtered out later */
> > > -    env->cpuid_kvm_features = ~0;
> > > -
> > >      for (ent = qdict_first(features); ent; ent = qdict_next(features, ent)) {
> > >          const QString *qval = qobject_to_qstring(qdict_entry_value(ent));
> > >          object_property_parse(OBJECT(cpu), qstring_get_str(qval),
> > > -- 
> > > 1.7.11.2
> > > 
> > 
> > -- 
> > Eduardo
> > 
> 
> 
> -- 
> Regards,
>   Igor

Patch

diff --git a/target-i386/cpu.c b/target-i386/cpu.c
index 43601a3..e266792 100644
--- a/target-i386/cpu.c
+++ b/target-i386/cpu.c
@@ -1093,6 +1093,9 @@  static void cpudef_2_x86_cpu(X86CPU *cpu, x86_def_t *def, Error **errp)
     env->cpuid_7_0_ebx = def->cpuid_7_0_ebx_features;
     env->cpuid_xlevel2 = def->xlevel2;
 
+    /* not supported bits will be filtered out later */
+    env->cpuid_kvm_features = ~0;
+
     object_property_set_bool(OBJECT(cpu), true, "hypervisor", errp);
 }
 
@@ -1175,9 +1178,6 @@  static int cpu_x86_find_by_name(X86CPU *cpu, x86_def_t *x86_cpu_def,
 
     cpudef_2_x86_cpu(cpu, def, errp);
 
-    /* not supported bits will be filtered out later */
-    env->cpuid_kvm_features = ~0;
-
     for (ent = qdict_first(features); ent; ent = qdict_next(features, ent)) {
         const QString *qval = qobject_to_qstring(qdict_entry_value(ent));
         object_property_parse(OBJECT(cpu), qstring_get_str(qval),