Re: [Qemu-devel] [PATCH v2] qobject: Replace property list with GHashTable

[Daniel P. Berrange]

Nit-pick on subject line - s/qobject/qom/ and also copy
Andreas Färber on the patch, since he's the primary
maintainer of the QOM subsystem.

On Tue, Oct 06, 2015 at 04:48:53PM +0300, Pavel Fedin wrote:
> ARM GICv3 systems with large number of CPUs create lots of IRQ pins. Since
> every pin is represented as a property, number of these properties becomes
> very large. Every property add first makes sure there's no duplicates.
> Traversing the list becomes very slow, therefore qemu initialization takes
> significant time (several seconds for e. g. 16 CPUs).
> 
> This patch replaces list with GHashTable, making lookup very fast. The only
> drawback is that object_child_foreach() and object_child_foreach_recursive()
> cannot modify their objects during traversal, since GHashTableIter does not
> have modify-safe version. However, the code seems not to modify objects via
> these functions.
> 
> Signed-off-by: Pavel Fedin <address@hidden>
> ---
> v1 => v2:
> - Fixed stupid bug in object_unparent(), use correct object
> ---
>  include/qom/object.h |  4 +--
>  qmp.c                |  8 +++--
>  qom/object.c         | 98 
> +++++++++++++++++++++++-----------------------------
>  vl.c                 |  4 ++-
>  4 files changed, 54 insertions(+), 60 deletions(-)
> 
> diff --git a/include/qom/object.h b/include/qom/object.h
> index be7280c..b100923 100644
> --- a/include/qom/object.h
> +++ b/include/qom/object.h
> @@ -345,7 +345,7 @@ typedef struct ObjectProperty
>      ObjectPropertyRelease *release;
>      void *opaque;
>  
> -    QTAILQ_ENTRY(ObjectProperty) node;
> +    Object *obj;

I have a patch which adds property registration against the
class, so requiring ObjectProperty to have a back-poointer
to an object instance is not desirable.

> diff --git a/qom/object.c b/qom/object.c
> index 4805328..1c7c42a 100644
> --- a/qom/object.c
> +++ b/qom/object.c
> @@ -326,6 +326,20 @@ static void object_post_init_with_type(Object *obj, 
> TypeImpl *ti)
>      }
>  }
>  
> +static void property_destroy(gpointer data)
> +{
> +    ObjectProperty *prop = data;
> +
> +    if (prop->release) {
> +        prop->release(prop->obj, prop->name, prop->opaque);
> +    }

Instead of calling the release() callback here, use a hash iterator
to call it from the object_finalize() method, before unref'ing the
hash table, and also call it from objecT_property_del directly. That
way you don't need to store the Object * back pointer in the
ObjectProperty struct.

> +
> +    g_free(prop->name);
> +    g_free(prop->type);
> +    g_free(prop->description);
> +    g_free(prop);
> +}

> @@ -410,7 +404,7 @@ static void object_finalize(void *data)
>      Object *obj = data;
>      TypeImpl *ti = obj->class->type;
>  
> -    object_property_del_all(obj);
> +    g_hash_table_unref(obj->properties);
>      object_deinit(obj, ti);
>  
>      g_assert(obj->ref == 0);

>  void object_property_del(Object *obj, const char *name, Error **errp)
>  {
> -    ObjectProperty *prop = object_property_find(obj, name, errp);
> -    if (prop == NULL) {
> +    if (g_hash_table_remove(obj->properties, name)) {
>          return;
>      }
>  
> -    if (prop->release) {
> -        prop->release(obj, name, prop->opaque);
> -    }
> -
> -    QTAILQ_REMOVE(&obj->properties, prop, node);
> -
> -    g_free(prop->name);
> -    g_free(prop->type);
> -    g_free(prop->description);
> -    g_free(prop);
> +    error_setg(errp, "Property '.%s' not found", name);
>  }

Overall I think this is a good improvement to make

Regards,
Daniel
-- 
|: http://berrange.com      -o-    http://www.flickr.com/photos/dberrange/ :|
|: http://libvirt.org              -o-             http://virt-manager.org :|
|: http://autobuild.org       -o-         http://search.cpan.org/~danberr/ :|
|: http://entangle-photo.org       -o-       http://live.gnome.org/gtk-vnc :|