[Spice-devel] [PATCH spice-gtk v3] Fix overlay for vaapisink

Snir Sheriber ssheribe at redhat.com
Mon Jan 14 12:54:06 UTC 2019


Hi,

On 1/14/19 11:48 AM, Frediano Ziglio wrote:
>> Hi,
>>
>>
>> On 1/11/19 6:36 PM, Frediano Ziglio wrote:
>>> The vaapisink plugin to support overlay requires the application
>>> to provide the proper context. If you don't do so the plugin will
>>> cause a crash of the application.
>>> To avoid possible thread errors from X11 create a new Display
>>> connection to pass to vaapisink.
>>>
>>> Signed-off-by: Frediano Ziglio <fziglio at redhat.com>
>>> ---
>>> To test it probably you'll have to remove the gstreamer registry,
>>> usually in ~/.cache/gstreamer-1.0/registry.x86_64.bin.
>>>
>>> Changes since v2:
>>> - remove hard dependency to libva-x11.
>>>
>>> Changes since v1:
>>> - updated to master;
>>> - use a different X11 connection as discussed in a previous thread;
>>> - remove some comments, now obsoleted;
>>> - fixed direct X11 link, now code is in spice-widget (as it should be);
>>> - better libva linking, using now build systems.
>>> ---
>>>    configure.ac              |  5 +++++
>>>    meson.build               |  5 +++++
>>>    src/Makefile.am           |  2 ++
>>>    src/channel-display-gst.c | 14 -------------
>>>    src/spice-widget.c        | 44 +++++++++++++++++++++++++++++++++++++++
>>>    5 files changed, 56 insertions(+), 14 deletions(-)
>>>
>>> diff --git a/configure.ac b/configure.ac
>>> index 8c5c4d38..ead6c1bb 100644
>>> --- a/configure.ac
>>> +++ b/configure.ac
>>> @@ -175,6 +175,11 @@ PKG_CHECK_MODULES(GTHREAD, gthread-2.0)
>>>    
>>>    PKG_CHECK_MODULES(JSON, json-glib-1.0)
>>>    
>>> +PKG_CHECK_EXISTS([libva-x11], [
>>> +    PKG_CHECK_MODULES(LIBVA, libva-x11)
>>> +    AC_DEFINE([HAVE_LIBVA], [1], [Define if libva is available])
>>> +])
>>> +
>>>    AC_ARG_ENABLE([webdav],
>>>      AS_HELP_STRING([--enable-webdav=@<:@auto/yes/no@:>@],
>>>                     [Enable webdav support @<:@default=auto@:>@]),
>>> diff --git a/meson.build b/meson.build
>>> index 37377680..6e96cca8 100644
>>> --- a/meson.build
>>> +++ b/meson.build
>>> @@ -134,6 +134,11 @@ if d.found()
>>>      if host_machine.system() != 'windows'
>>>        spice_gtk_deps += dependency('epoxy')
>>>        spice_gtk_deps += dependency('x11')
>>> +    d = dependency('libva-x11', required: false)
>>> +    if d.found()
>>> +      spice_gtk_deps += d
>>> +      spice_gtk_config_data.set('HAVE_LIBVA', '1')
>>> +    endif
>>>      endif
>>>      spice_gtk_has_gtk = true
>>>    endif
>>> diff --git a/src/Makefile.am b/src/Makefile.am
>>> index b876530c..784b18d0 100644
>>> --- a/src/Makefile.am
>>> +++ b/src/Makefile.am
>>> @@ -89,6 +89,7 @@ SPICE_COMMON_CPPFLAGS =						\
>>>    	$(GUDEV_CFLAGS)						\
>>>    	$(SOUP_CFLAGS)						\
>>>    	$(PHODAV_CFLAGS)					\
>>> +	$(LIBVA_CFLAGS)						\
>>>    	$(X11_CFLAGS)					\
>>>    	$(LZ4_CFLAGS)					\
>>>    	$(NULL)
>>> @@ -112,6 +113,7 @@ SPICE_GTK_LIBADD_COMMON =		\
>>>    	$(CAIRO_LIBS)			\
>>>    	$(X11_LIBS)			\
>>>    	$(LIBM)				\
>>> +	$(LIBVA_LIBS)			\
>>>    	$(NULL)
>>>    
>>>    SPICE_GTK_SOURCES_COMMON =		\
>>> diff --git a/src/channel-display-gst.c b/src/channel-display-gst.c
>>> index 2f556fe4..01ee83cb 100644
>>> --- a/src/channel-display-gst.c
>>> +++ b/src/channel-display-gst.c
>>> @@ -406,22 +406,8 @@ static gboolean create_pipeline(SpiceGstDecoder
>>> *decoder)
>>>        } else {
>>>            /* handle has received, it means playbin will render directly
>>>            into
>>>             * widget using the gstvideooverlay interface instead of
>>>             app-sink.
>>> -         * Also avoid using vaapisink if exist since vaapisink could be
>>> -         * buggy when it is combined with playbin. changing its rank to
>>> -         * none will make playbin to avoid of using it.
>>>             */
>>> -        GstRegistry *registry = NULL;
>>> -        GstPluginFeature *vaapisink = NULL;
>>> -
>>>            SPICE_DEBUG("Video is presented using gstreamer's GstVideoOverlay
>>>            interface");
>>> -        registry = gst_registry_get();
>>> -        if (registry) {
>>> -            vaapisink = gst_registry_lookup_feature(registry,
>>> "vaapisink");
>>> -        }
>>> -        if (vaapisink) {
>>> -            gst_plugin_feature_set_rank(vaapisink, GST_RANK_NONE);
>>> -            gst_object_unref(vaapisink);
>>> -        }
>>>        }
>>>    
>>>        g_signal_connect(playbin, "source-setup",
>>>        G_CALLBACK(app_source_setup), decoder);
>>> diff --git a/src/spice-widget.c b/src/spice-widget.c
>>> index af0301c1..85707704 100644
>>> --- a/src/spice-widget.c
>>> +++ b/src/spice-widget.c
>>> @@ -28,6 +28,9 @@
>>>    #ifdef GDK_WINDOWING_X11
>>>    #include <X11/Xlib.h>
>>>    #include <gdk/gdkx.h>
>>> +#ifdef HAVE_LIBVA
>>> +#include <va/va_x11.h>
>>> +#endif
>>>    #endif
>>>    #ifdef G_OS_WIN32
>>>    #include <windows.h>
>>> @@ -2592,6 +2595,47 @@ static void gst_sync_bus_call(GstBus *bus,
>>> GstMessage *msg, SpiceDisplay *displa
>>>            }
>>>            break;
>>>        }
>>> +    case GST_MESSAGE_NEED_CONTEXT:
>>> +    {
>>> +        const gchar *context_type;
>>> +
>>> +        gst_message_parse_context_type(msg, &context_type);
>>> +        g_debug("got need context %s from %s", context_type,
>>> GST_MESSAGE_SRC_NAME(msg));
>>
>> Maybe worth mentioning it's gstreamer related (Gstreamer: ?)
>>
> Added.
>
>>> +
>>> +#if defined(GDK_WINDOWING_X11) && defined(HAVE_LIBVA)
>>> +        if (g_strcmp0(context_type, "gst.vaapi.app.Display") == 0) {
>>> +            static Display *x11_display = NULL;
>>> +            static VADisplay va_display = NULL;
>>> +
>>> +            // note that if VAAPI do not get the context for the
>>> +            // overlay it crashes the entire program!
>>> +            GdkDisplay *display = gdk_display_get_default();
>>> +            g_assert_nonnull(display);
>>> +
>>> +            // Compute display pointers
>>> +            if (!x11_display && GDK_IS_X11_DISPLAY(display)) {
>>> +                x11_display = gdk_x11_display_get_xdisplay(display);
>>> +                // for thread problems we need a different Display,
>>> +                // VAAPI access the Display from another thread
>>> +                x11_display = XOpenDisplay(XDisplayString(x11_display));
>>> +                g_assert_nonnull(x11_display);
>>> +                va_display = vaGetDisplay(x11_display);
>>> +                g_assert_nonnull(va_display);
>>> +            }
>>> +
>>> +            GstContext *context = gst_context_new("gst.vaapi.app.Display",
>>> FALSE);
>>> +            GstStructure *structure =
>>> gst_context_writable_structure(context);
>>> +            if (x11_display) {
>>> +                gst_structure_set(structure, "x11-display",
>>> G_TYPE_POINTER, x11_display, NULL);
>>> +            }
>>> +            gst_structure_set(structure, "va-display", G_TYPE_POINTER,
>>> va_display, NULL);
>>> +
>>> +            gst_element_set_context(GST_ELEMENT(msg->src), context);
>>> +            gst_context_unref(context);
>>> +        }
>>> +#endif
>>> +        break;
>>> +    }
>>>        default:
>>>            /* not being handled */
>>>            break;
>>
>> It seems there is some kind of a race, every few runs it gets frozen and
>> then crashes, i looked a bit,
>>
>> could not found a clear cause.
>>
> I had similar issues in the past but patch was quite different.
> Also I had Fedora 28, now I have Fedora 29, which OS are you running?


f29

(I guess it should not matter much but i'm using cinnamon not gnome)


> How often does it happen?


Very often, every 3-4 consecutive runs


> Does it happen doing something specific like screen resize?


No, when it happens it does not event get to the point of a clear image, 
immediately freezes.

fwiw i disabled the resize related callbacks and it still happened.


>
>> One time i got these errors but may not be always related since it
>> usually crashes without any hints
>>
>>
>> ** (remote-viewer:19435): CRITICAL **: 11:44:18.845:
>> gst_vaapi_window_reconfigure: assertion 'window != NULL' failed
>>
>> ** (remote-viewer:19435): CRITICAL **: 11:44:18.845:
>> gst_vaapi_window_get_size: assertion 'window != NULL' failed
>>
>>
>> Snir.
>>
> None of these.
> Are you sure your machine is using VAAPI, did you dump the pipeline?
> window != NULL seems to suggest there's no overlay involved.


This specific warning i got only once, usually i get it without it, i do 
have vaapi, and seems

it sets the vaapicontext correctly in replay to the context msg


Snir.

>
> Frediano


More information about the Spice-devel mailing list