> > diff --git a/target/i386/cpu.c b/target/i386/cpu.c > > index 0d35e95430fe..bf290262cbfe 100644 > > --- a/target/i386/cpu.c > > +++ b/target/i386/cpu.c > > @@ -9044,6 +9044,12 @@ static void x86_cpu_post_initfn(Object *obj) > > X86_CONFIDENTIAL_GUEST(current_machine->cgs), (CPU(obj))); > > } > > #endif > > + > > + /* > > + * Re-apply the "feature[=foo]" from '-cpu' option since they might > > + * be overwritten by above > > + */ > > + qdev_prop_set_globals(DEVICE(obj)); > > } > > This patch LGTM.
This solution will call qdev_prop_set_globals() twice. My concern is that this masks the problem: previous x86 CPU assumptions about the order of global property initialization break down... Per the commit message of Paolo's commit: "This is incorrect because the leaf class cannot observe property values applied by the superclasses; for example, a compat property will be set on a device *after* the class's post_init callback has run." X86 CPUs have the issue (e.g., "vendor" doesn't work) now because they - as leaf class, don't care about the property values of superclass - the DeviceState. If a property is just for initialization, like "vendor", it should be placed in the instance_init() instead of instance_post_init(). In addition, if other places handle it similarly, the device's post_init seems pointless. :-( Thanks, Zhao
