Re: [Qemu-devel] [PATCH v4 02/10] accel: introduce AccelClass.global_props

[Peter Xu]

On Wed, Jun 21, 2017 at 09:23:11AM -0300, Eduardo Habkost wrote:
> On Wed, Jun 21, 2017 at 03:52:00PM +0800, Peter Xu wrote:
> > Introduce this new field for the accelerator classes so that each
> > specific accelerator in the future can register its own global
> > properties to be used further by the system. It works just like how the
> > old machine compatible properties do, but only tailored for
> > accelerators.
> > 
> > Use the newly exported register_compat_prop() to pass the accelerator
> > global properties to the global_props list.
> > 
> > Suggested-by: Eduardo Habkost <address@hidden>
> > Signed-off-by: Peter Xu <address@hidden>
> > ---
> >  accel/accel.c          | 10 ++++++++++
> >  include/sysemu/accel.h | 10 ++++++++++
> >  vl.c                   |  1 +
> >  3 files changed, 21 insertions(+)
> > 
> > diff --git a/accel/accel.c b/accel/accel.c
> > index 7c079a5..212581c 100644
> > --- a/accel/accel.c
> > +++ b/accel/accel.c
> > @@ -120,6 +120,16 @@ void configure_accelerator(MachineState *ms)
> >      }
> >  }
> >  
> > +void accel_register_compat_props(AccelState *accel)
> > +{
> > +    AccelClass *class = ACCEL_GET_CLASS(accel);
> > +    GlobalProperty *prop;
> > +
> > +    for (prop = class->global_props; prop && prop->driver; prop++) {
> > +        register_compat_prop(prop->driver, prop->property, prop->value);
> > +    }
> > +}
> 
> I suggest adding a generic register_compat_props_array(GlobalProperty *props)
> helper, so we can reuse it at machine_register_compat_props()
> later (once we fix global property ordering in
> qdev_prop_set_globals() and make the object_class_foreach() hack
> in machine_register_compat_props() unnecessary).

Sure.

> 
> > +
> >  static void register_accel_types(void)
> >  {
> >      type_register_static(&accel_type);
> > diff --git a/include/sysemu/accel.h b/include/sysemu/accel.h
> > index 15944c1..8a01e51 100644
> > --- a/include/sysemu/accel.h
> > +++ b/include/sysemu/accel.h
> > @@ -24,6 +24,7 @@
> >  #define HW_ACCEL_H
> >  
> >  #include "qom/object.h"
> > +#include "hw/qdev-properties.h"
> >  
> >  typedef struct AccelState {
> >      /*< private >*/
> > @@ -40,6 +41,13 @@ typedef struct AccelClass {
> >      int (*available)(void);
> >      int (*init_machine)(MachineState *ms);
> >      bool *allowed;
> > +    /*
> > +     * Array of gobal properties that would be applied when specific
> 
> "global"

Will fix.

> 
> > +     * accelerator is chosen. It works just like
> > +     * MachineClass.compat_props but it's for accelerators not
> > +     * machines.
> > +     */
> > +    GlobalProperty *global_props;
> >  } AccelClass;
> >  
> >  #define TYPE_ACCEL "accel"
> > @@ -57,5 +65,7 @@ typedef struct AccelClass {
> >  extern int tcg_tb_size;
> >  
> >  void configure_accelerator(MachineState *ms);
> > +/* Register accelerator specific global properties */
> > +void accel_register_compat_props(AccelState *accel);
> >  
> >  #endif
> > diff --git a/vl.c b/vl.c
> > index 59fea15..4452d7a 100644
> > --- a/vl.c
> > +++ b/vl.c
> > @@ -4571,6 +4571,7 @@ int main(int argc, char **argv, char **envp)
> >              exit (i == 1 ? 1 : 0);
> >      }
> >  
> 
> I suggest a comment here warning about global property
> registration ordering.  e.g.:
> 
>  /*
>   * Ordering of global property registration matters:
>   * - machine compat_props should override accelerator-specific
>   *   globals.
>   * - user-provided globals (-global, and global properties
>   *   derived from other command-line options like -cpu) should
>   *   override machine compat_props and accelerator-specific
>   *   globals.
>   */

I have a standalone patch (next one) to unify these three places and
added some comment. Would that suite?

Thanks,

-- 
Peter Xu