Re: [Qemu-devel] [PATCH 04/10] target-i386: convert 'hv_spinlocks' to static property

[Eduardo Habkost]

On Tue, Feb 26, 2013 at 04:50:22PM +0100, Igor Mammedov wrote:
> On Tue, 26 Feb 2013 11:42:08 -0300
> Eduardo Habkost <address@hidden> wrote:
> 
> > On Mon, Feb 25, 2013 at 02:03:03AM +0100, Igor Mammedov wrote:
> > [...]
> > > +static void x86_set_hv_spinlocks(Object *obj, Visitor *v, void *opaque,
> > > +                                 const char *name, Error **errp)
> > > +{
> > > +    const int64_t min = 0xFFF;
> > > +    const int64_t max = UINT_MAX;
> > > +    X86CPU *cpu = X86_CPU(obj);
> > > +    int64_t value;
> > > +
> > > +    visit_type_int(v, &value, name, errp);
> > > +    if (error_is_set(errp)) {
> > > +        return;
> > > +    }
> > > +
> > > +    if (value < min || value > max) {
> > > +        error_setg(errp, "Property %s.%s doesn't take value %" PRId64
> > > +                  " (minimum: %" PRId64 ", maximum: %" PRId64 ")",
> > > +                  object_get_typename(obj), name ? name : "null",
> > > +                  value, min, max);
> > > +        return;
> > > +    }
> > > +    cpu->env.hyperv_spinlock_attempts = value;
> > > +}
> > [...]
> > > @@ -1421,6 +1466,7 @@ static void cpu_x86_parse_featurestr(X86CPU *cpu,
> > > char *features, Error **errp) } else if (!strcmp(featurestr,
> > > "hv-spinlocks")) { char *err;
> > >                  const int min = 0xFFF;
> > > +                char num[32];
> > >                  numvalue = strtoul(val, &err, 0);
> > >                  if (!*val || *err) {
> > >                      error_setg(errp, "bad numerical value %s", val);
> > > @@ -1432,7 +1478,8 @@ static void cpu_x86_parse_featurestr(X86CPU *cpu,
> > > char *features, Error **errp) min);
> > >                      numvalue = min;
> > >                  }
> > > -                env->hyperv_spinlock_attempts = numvalue;
> > > +                snprintf(num, sizeof(num), "%" PRId32, numvalue);
> > 
> > Isn't the above validation code already inside x86_set_hv_spinlocks()?
> > Why do we need it here in cpu_x86_parse_featurestr() too?
> Purpose of above is not a validation, it's silent fix-up made visible in
> 3/10, like we did for xlevel.

OK.

> And will be accompanied by entry in Changelog, once patch is in.
> 
> > 
> > > +                object_property_parse(OBJECT(cpu), num, featurestr,
> > > errp); } else {

I was going to ask: "in that case, isn't it easiser to use
object_property_set_int() instead of snprintf() +
object_property_parse()?"

But then I remembered that we're going to eventually replace all
object_property_parse() calls here with qdev_prop_set_global() calls.
:-)


> > >                  error_setg(errp, "unrecognized feature %s", featurestr);
> > >                  goto out;
> > > @@ -1597,7 +1644,6 @@ static void cpu_x86_register(X86CPU *cpu, const
> > > char *name, Error **errp) def->kvm_features |= kvm_default_features;
> > >      }
> > >      def->ext_features |= CPUID_EXT_HYPERVISOR;
> > > -    env->hyperv_spinlock_attempts = HYPERV_SPINLOCK_NEVER_RETRY;
> > >  
> > >      object_property_set_str(OBJECT(cpu), def->vendor, "vendor", errp);
> > >      object_property_set_int(OBJECT(cpu), def->level, "level", errp);
> > > -- 
> > > 1.7.1
> > > 
> > 
> 

-- 
Eduardo