gstreamer/subprojects/gst-plugins-bad/sys/wasapi2/gstwasapi2client.cpp

984 lines
28 KiB
C++
Raw Normal View History

/*
* Copyright (C) 2008 Ole André Vadla Ravnås <ole.andre.ravnas@tandberg.com>
* Copyright (C) 2013 Collabora Ltd.
* Author: Sebastian Dröge <sebastian.droege@collabora.co.uk>
* Copyright (C) 2018 Centricular Ltd.
* Author: Nirbheek Chauhan <nirbheek@centricular.com>
* Copyright (C) 2020 Seungha Yang <seungha@centricular.com>
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Library General Public
* License as published by the Free Software Foundation; either
* version 2 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Library General Public License for more details.
*
* You should have received a copy of the GNU Library General Public
* License along with this library; if not, write to the
* Free Software Foundation, Inc., 51 Franklin St, Fifth Floor,
* Boston, MA 02110-1301, USA.
*/
#ifdef HAVE_CONFIG_H
#include "config.h"
#endif
#include "AsyncOperations.h"
#include "gstwasapi2client.h"
#include "gstwasapi2util.h"
#include <initguid.h>
#include <windows.foundation.h>
#include <windows.ui.core.h>
#include <wrl.h>
#include <wrl/wrappers/corewrappers.h>
#include <audioclient.h>
#include <mmdeviceapi.h>
#include <string.h>
#include <string>
#include <locale>
#include <codecvt>
/* *INDENT-OFF* */
using namespace ABI::Windows::ApplicationModel::Core;
using namespace ABI::Windows::Foundation;
using namespace ABI::Windows::Foundation::Collections;
using namespace ABI::Windows::UI::Core;
using namespace ABI::Windows::Media::Devices;
using namespace ABI::Windows::Devices::Enumeration;
using namespace Microsoft::WRL;
using namespace Microsoft::WRL::Wrappers;
G_BEGIN_DECLS
GST_DEBUG_CATEGORY_EXTERN (gst_wasapi2_client_debug);
#define GST_CAT_DEFAULT gst_wasapi2_client_debug
G_END_DECLS
/* *INDENT-ON* */
static void
gst_wasapi2_client_on_device_activated (GstWasapi2Client * client,
wasapi2: Rewrite plugin and implement audioringbuffer subclass ... based on MediaFoundation work queue API. By this commit, wasapi2 plugin will make use of pull mode scheduling with audioringbuffer subclass. There are several drawbacks of audiosrc/audiosink subclassing (not audiobasesrc/audiobasesink) for WASAPI API, which are: * audiosrc/audiosink classes try to set high priority to read/write thread via MMCSS (Multimedia Class Scheduler Service) but it's not allowed in case of UWP application. In order to use MMCSS in UWP, application should use MediaFoundation work queue indirectly. Since audiosrc/audiosink scheduling model is not compatible with MediaFoundation's work queue model, audioringbuffer subclassing is required. * WASAPI capture device might report larger packet size than expected (i.e., larger frames we can read than expected frame size per period). Meanwhile, in any case, application should drain all packets at that moment. In order to handle the case, wasapi/wasapi2 plugins were making use of GstAdapter which is obviously sub-optimal because it requires additional memory allocation and copy. By implementing audioringbuffer subclassing, we can avoid such inefficiency. In this commit, all the device read/write operations will be moved to newly implemented wasapi2ringbuffer class and existing wasapi2client class will take care of device enumeration and activation parts only. Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2306>
2021-05-10 11:45:28 +00:00
IAudioClient * audio_client);
/* *INDENT-OFF* */
class GstWasapiDeviceActivator
: public RuntimeClass<RuntimeClassFlags<ClassicCom>, FtmBase,
IActivateAudioInterfaceCompletionHandler>
{
public:
GstWasapiDeviceActivator ()
{
g_weak_ref_init (&listener_, nullptr);
}
~GstWasapiDeviceActivator ()
{
g_weak_ref_set (&listener_, nullptr);
}
HRESULT
RuntimeClassInitialize (GstWasapi2Client * listener, gpointer dispatcher)
{
if (!listener)
return E_INVALIDARG;
g_weak_ref_set (&listener_, listener);
if (dispatcher) {
ComPtr<IInspectable> inspectable =
reinterpret_cast<IInspectable*> (dispatcher);
HRESULT hr;
hr = inspectable.As (&dispatcher_);
if (gst_wasapi2_result (hr))
GST_INFO("Main UI dispatcher is available");
}
return S_OK;
}
STDMETHOD(ActivateCompleted)
(IActivateAudioInterfaceAsyncOperation *async_op)
{
wasapi2: Rewrite plugin and implement audioringbuffer subclass ... based on MediaFoundation work queue API. By this commit, wasapi2 plugin will make use of pull mode scheduling with audioringbuffer subclass. There are several drawbacks of audiosrc/audiosink subclassing (not audiobasesrc/audiobasesink) for WASAPI API, which are: * audiosrc/audiosink classes try to set high priority to read/write thread via MMCSS (Multimedia Class Scheduler Service) but it's not allowed in case of UWP application. In order to use MMCSS in UWP, application should use MediaFoundation work queue indirectly. Since audiosrc/audiosink scheduling model is not compatible with MediaFoundation's work queue model, audioringbuffer subclassing is required. * WASAPI capture device might report larger packet size than expected (i.e., larger frames we can read than expected frame size per period). Meanwhile, in any case, application should drain all packets at that moment. In order to handle the case, wasapi/wasapi2 plugins were making use of GstAdapter which is obviously sub-optimal because it requires additional memory allocation and copy. By implementing audioringbuffer subclassing, we can avoid such inefficiency. In this commit, all the device read/write operations will be moved to newly implemented wasapi2ringbuffer class and existing wasapi2client class will take care of device enumeration and activation parts only. Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2306>
2021-05-10 11:45:28 +00:00
ComPtr<IAudioClient> audio_client;
HRESULT hr = S_OK;
HRESULT hr_async_op = S_OK;
ComPtr<IUnknown> audio_interface;
GstWasapi2Client *client;
client = (GstWasapi2Client *) g_weak_ref_get (&listener_);
if (!client) {
this->Release ();
GST_WARNING ("No listener was configured");
return S_OK;
}
GST_INFO_OBJECT (client, "AsyncOperation done");
hr = async_op->GetActivateResult(&hr_async_op, &audio_interface);
if (!gst_wasapi2_result (hr)) {
GST_WARNING_OBJECT (client, "Failed to get activate result, hr: 0x%x", hr);
goto done;
}
if (!gst_wasapi2_result (hr_async_op)) {
GST_WARNING_OBJECT (client, "Failed to activate device");
goto done;
}
hr = audio_interface.As (&audio_client);
if (!gst_wasapi2_result (hr)) {
GST_ERROR_OBJECT (client, "Failed to get IAudioClient3 interface");
goto done;
}
done:
/* Should call this method anyway, listener will wait this event */
gst_wasapi2_client_on_device_activated (client, audio_client.Get());
gst_object_unref (client);
/* return S_OK anyway, but listener can know it's succeeded or not
* by passed IAudioClient handle via gst_wasapi2_client_on_device_activated
*/
this->Release ();
return S_OK;
}
HRESULT
ActivateDeviceAsync(const std::wstring &device_id)
{
ComPtr<IAsyncAction> async_action;
bool run_async = false;
HRESULT hr;
auto work_item = Callback<Implements<RuntimeClassFlags<ClassicCom>,
IDispatchedHandler, FtmBase>>([this, device_id]{
ComPtr<IActivateAudioInterfaceAsyncOperation> async_op;
HRESULT async_hr = S_OK;
async_hr = ActivateAudioInterfaceAsync (device_id.c_str (),
__uuidof(IAudioClient3), nullptr, this, &async_op);
/* for debugging */
gst_wasapi2_result (async_hr);
return async_hr;
});
if (dispatcher_) {
boolean can_now;
hr = dispatcher_->get_HasThreadAccess (&can_now);
if (!gst_wasapi2_result (hr))
return hr;
if (!can_now)
run_async = true;
}
if (run_async && dispatcher_) {
hr = dispatcher_->RunAsync (CoreDispatcherPriority_Normal,
work_item.Get (), &async_action);
} else {
hr = work_item->Invoke ();
}
/* We should hold activator object until activation callback has executed,
* because OS doesn't hold reference of this callback COM object.
* otherwise access violation would happen
* See https://docs.microsoft.com/en-us/windows/win32/api/mmdeviceapi/nf-mmdeviceapi-activateaudiointerfaceasync
*
* This reference count will be decreased by self later on callback,
* which will be called from device worker thread.
*/
if (gst_wasapi2_result (hr))
this->AddRef ();
return hr;
}
private:
GWeakRef listener_;
ComPtr<ICoreDispatcher> dispatcher_;
};
/* *INDENT-ON* */
wasapi2: Rewrite plugin and implement audioringbuffer subclass ... based on MediaFoundation work queue API. By this commit, wasapi2 plugin will make use of pull mode scheduling with audioringbuffer subclass. There are several drawbacks of audiosrc/audiosink subclassing (not audiobasesrc/audiobasesink) for WASAPI API, which are: * audiosrc/audiosink classes try to set high priority to read/write thread via MMCSS (Multimedia Class Scheduler Service) but it's not allowed in case of UWP application. In order to use MMCSS in UWP, application should use MediaFoundation work queue indirectly. Since audiosrc/audiosink scheduling model is not compatible with MediaFoundation's work queue model, audioringbuffer subclassing is required. * WASAPI capture device might report larger packet size than expected (i.e., larger frames we can read than expected frame size per period). Meanwhile, in any case, application should drain all packets at that moment. In order to handle the case, wasapi/wasapi2 plugins were making use of GstAdapter which is obviously sub-optimal because it requires additional memory allocation and copy. By implementing audioringbuffer subclassing, we can avoid such inefficiency. In this commit, all the device read/write operations will be moved to newly implemented wasapi2ringbuffer class and existing wasapi2client class will take care of device enumeration and activation parts only. Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2306>
2021-05-10 11:45:28 +00:00
typedef enum
{
GST_WASAPI2_CLIENT_ACTIVATE_FAILED = -1,
GST_WASAPI2_CLIENT_ACTIVATE_INIT = 0,
GST_WASAPI2_CLIENT_ACTIVATE_WAIT,
GST_WASAPI2_CLIENT_ACTIVATE_DONE,
} GstWasapi2ClientActivateState;
enum
{
PROP_0,
PROP_DEVICE,
PROP_DEVICE_NAME,
PROP_DEVICE_INDEX,
PROP_DEVICE_CLASS,
PROP_DISPATCHER,
PROP_CAN_AUTO_ROUTING,
};
#define DEFAULT_DEVICE_INDEX -1
#define DEFAULT_DEVICE_CLASS GST_WASAPI2_CLIENT_DEVICE_CLASS_CAPTURE
struct _GstWasapi2Client
{
GstObject parent;
GstWasapi2ClientDeviceClass device_class;
gchar *device_id;
gchar *device_name;
gint device_index;
gpointer dispatcher;
gboolean can_auto_routing;
wasapi2: Rewrite plugin and implement audioringbuffer subclass ... based on MediaFoundation work queue API. By this commit, wasapi2 plugin will make use of pull mode scheduling with audioringbuffer subclass. There are several drawbacks of audiosrc/audiosink subclassing (not audiobasesrc/audiobasesink) for WASAPI API, which are: * audiosrc/audiosink classes try to set high priority to read/write thread via MMCSS (Multimedia Class Scheduler Service) but it's not allowed in case of UWP application. In order to use MMCSS in UWP, application should use MediaFoundation work queue indirectly. Since audiosrc/audiosink scheduling model is not compatible with MediaFoundation's work queue model, audioringbuffer subclassing is required. * WASAPI capture device might report larger packet size than expected (i.e., larger frames we can read than expected frame size per period). Meanwhile, in any case, application should drain all packets at that moment. In order to handle the case, wasapi/wasapi2 plugins were making use of GstAdapter which is obviously sub-optimal because it requires additional memory allocation and copy. By implementing audioringbuffer subclassing, we can avoid such inefficiency. In this commit, all the device read/write operations will be moved to newly implemented wasapi2ringbuffer class and existing wasapi2client class will take care of device enumeration and activation parts only. Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2306>
2021-05-10 11:45:28 +00:00
IAudioClient *audio_client;
GstWasapiDeviceActivator *activator;
GstCaps *supported_caps;
GThread *thread;
GMutex lock;
GCond cond;
GMainContext *context;
GMainLoop *loop;
/* To wait ActivateCompleted event */
GMutex init_lock;
GCond init_cond;
GstWasapi2ClientActivateState activate_state;
};
GType
gst_wasapi2_client_device_class_get_type (void)
{
static GType class_type = 0;
static const GEnumValue types[] = {
{GST_WASAPI2_CLIENT_DEVICE_CLASS_CAPTURE, "Capture", "capture"},
{GST_WASAPI2_CLIENT_DEVICE_CLASS_RENDER, "Render", "render"},
{GST_WASAPI2_CLIENT_DEVICE_CLASS_LOOPBACK_CAPTURE, "Loopback-Capture",
"loopback-capture"},
wasapi2: Rewrite plugin and implement audioringbuffer subclass ... based on MediaFoundation work queue API. By this commit, wasapi2 plugin will make use of pull mode scheduling with audioringbuffer subclass. There are several drawbacks of audiosrc/audiosink subclassing (not audiobasesrc/audiobasesink) for WASAPI API, which are: * audiosrc/audiosink classes try to set high priority to read/write thread via MMCSS (Multimedia Class Scheduler Service) but it's not allowed in case of UWP application. In order to use MMCSS in UWP, application should use MediaFoundation work queue indirectly. Since audiosrc/audiosink scheduling model is not compatible with MediaFoundation's work queue model, audioringbuffer subclassing is required. * WASAPI capture device might report larger packet size than expected (i.e., larger frames we can read than expected frame size per period). Meanwhile, in any case, application should drain all packets at that moment. In order to handle the case, wasapi/wasapi2 plugins were making use of GstAdapter which is obviously sub-optimal because it requires additional memory allocation and copy. By implementing audioringbuffer subclassing, we can avoid such inefficiency. In this commit, all the device read/write operations will be moved to newly implemented wasapi2ringbuffer class and existing wasapi2client class will take care of device enumeration and activation parts only. Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2306>
2021-05-10 11:45:28 +00:00
{0, nullptr, nullptr}
};
if (g_once_init_enter (&class_type)) {
GType gtype = g_enum_register_static ("GstWasapi2ClientDeviceClass", types);
g_once_init_leave (&class_type, gtype);
}
return class_type;
}
static void gst_wasapi2_client_constructed (GObject * object);
static void gst_wasapi2_client_finalize (GObject * object);
static void gst_wasapi2_client_get_property (GObject * object, guint prop_id,
GValue * value, GParamSpec * pspec);
static void gst_wasapi2_client_set_property (GObject * object, guint prop_id,
const GValue * value, GParamSpec * pspec);
static gpointer gst_wasapi2_client_thread_func (GstWasapi2Client * self);
static gboolean
gst_wasapi2_client_main_loop_running_cb (GstWasapi2Client * self);
#define gst_wasapi2_client_parent_class parent_class
G_DEFINE_TYPE (GstWasapi2Client, gst_wasapi2_client, GST_TYPE_OBJECT);
static void
gst_wasapi2_client_class_init (GstWasapi2ClientClass * klass)
{
GObjectClass *gobject_class = G_OBJECT_CLASS (klass);
GParamFlags param_flags =
(GParamFlags) (G_PARAM_READWRITE | G_PARAM_CONSTRUCT_ONLY |
G_PARAM_STATIC_STRINGS);
gobject_class->constructed = gst_wasapi2_client_constructed;
gobject_class->finalize = gst_wasapi2_client_finalize;
gobject_class->get_property = gst_wasapi2_client_get_property;
gobject_class->set_property = gst_wasapi2_client_set_property;
g_object_class_install_property (gobject_class, PROP_DEVICE,
g_param_spec_string ("device", "Device",
wasapi2: Rewrite plugin and implement audioringbuffer subclass ... based on MediaFoundation work queue API. By this commit, wasapi2 plugin will make use of pull mode scheduling with audioringbuffer subclass. There are several drawbacks of audiosrc/audiosink subclassing (not audiobasesrc/audiobasesink) for WASAPI API, which are: * audiosrc/audiosink classes try to set high priority to read/write thread via MMCSS (Multimedia Class Scheduler Service) but it's not allowed in case of UWP application. In order to use MMCSS in UWP, application should use MediaFoundation work queue indirectly. Since audiosrc/audiosink scheduling model is not compatible with MediaFoundation's work queue model, audioringbuffer subclassing is required. * WASAPI capture device might report larger packet size than expected (i.e., larger frames we can read than expected frame size per period). Meanwhile, in any case, application should drain all packets at that moment. In order to handle the case, wasapi/wasapi2 plugins were making use of GstAdapter which is obviously sub-optimal because it requires additional memory allocation and copy. By implementing audioringbuffer subclassing, we can avoid such inefficiency. In this commit, all the device read/write operations will be moved to newly implemented wasapi2ringbuffer class and existing wasapi2client class will take care of device enumeration and activation parts only. Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2306>
2021-05-10 11:45:28 +00:00
"WASAPI playback device as a GUID string", nullptr, param_flags));
g_object_class_install_property (gobject_class, PROP_DEVICE_NAME,
g_param_spec_string ("device-name", "Device Name",
wasapi2: Rewrite plugin and implement audioringbuffer subclass ... based on MediaFoundation work queue API. By this commit, wasapi2 plugin will make use of pull mode scheduling with audioringbuffer subclass. There are several drawbacks of audiosrc/audiosink subclassing (not audiobasesrc/audiobasesink) for WASAPI API, which are: * audiosrc/audiosink classes try to set high priority to read/write thread via MMCSS (Multimedia Class Scheduler Service) but it's not allowed in case of UWP application. In order to use MMCSS in UWP, application should use MediaFoundation work queue indirectly. Since audiosrc/audiosink scheduling model is not compatible with MediaFoundation's work queue model, audioringbuffer subclassing is required. * WASAPI capture device might report larger packet size than expected (i.e., larger frames we can read than expected frame size per period). Meanwhile, in any case, application should drain all packets at that moment. In order to handle the case, wasapi/wasapi2 plugins were making use of GstAdapter which is obviously sub-optimal because it requires additional memory allocation and copy. By implementing audioringbuffer subclassing, we can avoid such inefficiency. In this commit, all the device read/write operations will be moved to newly implemented wasapi2ringbuffer class and existing wasapi2client class will take care of device enumeration and activation parts only. Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2306>
2021-05-10 11:45:28 +00:00
"The human-readable device name", nullptr, param_flags));
g_object_class_install_property (gobject_class, PROP_DEVICE_INDEX,
g_param_spec_int ("device-index", "Device Index",
"The zero-based device index", -1, G_MAXINT, DEFAULT_DEVICE_INDEX,
param_flags));
g_object_class_install_property (gobject_class, PROP_DEVICE_CLASS,
g_param_spec_enum ("device-class", "Device Class",
"Device class", GST_TYPE_WASAPI2_CLIENT_DEVICE_CLASS,
DEFAULT_DEVICE_CLASS, param_flags));
g_object_class_install_property (gobject_class, PROP_DISPATCHER,
g_param_spec_pointer ("dispatcher", "Dispatcher",
"ICoreDispatcher COM object to use", param_flags));
g_object_class_install_property (gobject_class, PROP_CAN_AUTO_ROUTING,
g_param_spec_boolean ("auto-routing", "Auto Routing",
"Whether client can support automatic stream routing", FALSE,
(GParamFlags) (G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS)));
}
static void
gst_wasapi2_client_init (GstWasapi2Client * self)
{
self->device_index = DEFAULT_DEVICE_INDEX;
self->device_class = DEFAULT_DEVICE_CLASS;
self->can_auto_routing = FALSE;
g_mutex_init (&self->lock);
g_cond_init (&self->cond);
g_mutex_init (&self->init_lock);
g_cond_init (&self->init_cond);
self->activate_state = GST_WASAPI2_CLIENT_ACTIVATE_INIT;
self->context = g_main_context_new ();
self->loop = g_main_loop_new (self->context, FALSE);
}
static void
gst_wasapi2_client_constructed (GObject * object)
{
GstWasapi2Client *self = GST_WASAPI2_CLIENT (object);
/* *INDENT-OFF* */
ComPtr<GstWasapiDeviceActivator> activator;
/* *INDENT-ON* */
/* Create a new thread to ensure that COM thread can be MTA thread.
* We cannot ensure whether CoInitializeEx() was called outside of here for
* this thread or not. If it was called with non-COINIT_MULTITHREADED option,
* we cannot update it */
g_mutex_lock (&self->lock);
self->thread = g_thread_new ("GstWasapi2ClientWinRT",
(GThreadFunc) gst_wasapi2_client_thread_func, self);
while (!self->loop || !g_main_loop_is_running (self->loop))
g_cond_wait (&self->cond, &self->lock);
g_mutex_unlock (&self->lock);
G_OBJECT_CLASS (parent_class)->constructed (object);
}
static void
wasapi2: Rewrite plugin and implement audioringbuffer subclass ... based on MediaFoundation work queue API. By this commit, wasapi2 plugin will make use of pull mode scheduling with audioringbuffer subclass. There are several drawbacks of audiosrc/audiosink subclassing (not audiobasesrc/audiobasesink) for WASAPI API, which are: * audiosrc/audiosink classes try to set high priority to read/write thread via MMCSS (Multimedia Class Scheduler Service) but it's not allowed in case of UWP application. In order to use MMCSS in UWP, application should use MediaFoundation work queue indirectly. Since audiosrc/audiosink scheduling model is not compatible with MediaFoundation's work queue model, audioringbuffer subclassing is required. * WASAPI capture device might report larger packet size than expected (i.e., larger frames we can read than expected frame size per period). Meanwhile, in any case, application should drain all packets at that moment. In order to handle the case, wasapi/wasapi2 plugins were making use of GstAdapter which is obviously sub-optimal because it requires additional memory allocation and copy. By implementing audioringbuffer subclassing, we can avoid such inefficiency. In this commit, all the device read/write operations will be moved to newly implemented wasapi2ringbuffer class and existing wasapi2client class will take care of device enumeration and activation parts only. Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2306>
2021-05-10 11:45:28 +00:00
gst_wasapi2_client_finalize (GObject * object)
{
GstWasapi2Client *self = GST_WASAPI2_CLIENT (object);
if (self->loop) {
g_main_loop_quit (self->loop);
g_thread_join (self->thread);
g_main_context_unref (self->context);
g_main_loop_unref (self->loop);
wasapi2: Rewrite plugin and implement audioringbuffer subclass ... based on MediaFoundation work queue API. By this commit, wasapi2 plugin will make use of pull mode scheduling with audioringbuffer subclass. There are several drawbacks of audiosrc/audiosink subclassing (not audiobasesrc/audiobasesink) for WASAPI API, which are: * audiosrc/audiosink classes try to set high priority to read/write thread via MMCSS (Multimedia Class Scheduler Service) but it's not allowed in case of UWP application. In order to use MMCSS in UWP, application should use MediaFoundation work queue indirectly. Since audiosrc/audiosink scheduling model is not compatible with MediaFoundation's work queue model, audioringbuffer subclassing is required. * WASAPI capture device might report larger packet size than expected (i.e., larger frames we can read than expected frame size per period). Meanwhile, in any case, application should drain all packets at that moment. In order to handle the case, wasapi/wasapi2 plugins were making use of GstAdapter which is obviously sub-optimal because it requires additional memory allocation and copy. By implementing audioringbuffer subclassing, we can avoid such inefficiency. In this commit, all the device read/write operations will be moved to newly implemented wasapi2ringbuffer class and existing wasapi2client class will take care of device enumeration and activation parts only. Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2306>
2021-05-10 11:45:28 +00:00
self->thread = nullptr;
self->context = nullptr;
self->loop = nullptr;
}
wasapi2: Rewrite plugin and implement audioringbuffer subclass ... based on MediaFoundation work queue API. By this commit, wasapi2 plugin will make use of pull mode scheduling with audioringbuffer subclass. There are several drawbacks of audiosrc/audiosink subclassing (not audiobasesrc/audiobasesink) for WASAPI API, which are: * audiosrc/audiosink classes try to set high priority to read/write thread via MMCSS (Multimedia Class Scheduler Service) but it's not allowed in case of UWP application. In order to use MMCSS in UWP, application should use MediaFoundation work queue indirectly. Since audiosrc/audiosink scheduling model is not compatible with MediaFoundation's work queue model, audioringbuffer subclassing is required. * WASAPI capture device might report larger packet size than expected (i.e., larger frames we can read than expected frame size per period). Meanwhile, in any case, application should drain all packets at that moment. In order to handle the case, wasapi/wasapi2 plugins were making use of GstAdapter which is obviously sub-optimal because it requires additional memory allocation and copy. By implementing audioringbuffer subclassing, we can avoid such inefficiency. In this commit, all the device read/write operations will be moved to newly implemented wasapi2ringbuffer class and existing wasapi2client class will take care of device enumeration and activation parts only. Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2306>
2021-05-10 11:45:28 +00:00
gst_clear_caps (&self->supported_caps);
g_free (self->device_id);
g_free (self->device_name);
g_mutex_clear (&self->lock);
g_cond_clear (&self->cond);
g_mutex_clear (&self->init_lock);
g_cond_clear (&self->init_cond);
G_OBJECT_CLASS (parent_class)->finalize (object);
}
static void
gst_wasapi2_client_get_property (GObject * object, guint prop_id,
GValue * value, GParamSpec * pspec)
{
GstWasapi2Client *self = GST_WASAPI2_CLIENT (object);
switch (prop_id) {
case PROP_DEVICE:
g_value_set_string (value, self->device_id);
break;
case PROP_DEVICE_NAME:
g_value_set_string (value, self->device_name);
break;
case PROP_DEVICE_INDEX:
g_value_set_int (value, self->device_index);
break;
case PROP_DEVICE_CLASS:
g_value_set_enum (value, self->device_class);
break;
case PROP_DISPATCHER:
g_value_set_pointer (value, self->dispatcher);
break;
case PROP_CAN_AUTO_ROUTING:
g_value_set_boolean (value, self->can_auto_routing);
break;
default:
G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
break;
}
}
static void
gst_wasapi2_client_set_property (GObject * object, guint prop_id,
const GValue * value, GParamSpec * pspec)
{
GstWasapi2Client *self = GST_WASAPI2_CLIENT (object);
switch (prop_id) {
case PROP_DEVICE:
g_free (self->device_id);
self->device_id = g_value_dup_string (value);
break;
case PROP_DEVICE_NAME:
g_free (self->device_name);
self->device_name = g_value_dup_string (value);
break;
case PROP_DEVICE_INDEX:
self->device_index = g_value_get_int (value);
break;
case PROP_DEVICE_CLASS:
self->device_class =
(GstWasapi2ClientDeviceClass) g_value_get_enum (value);
break;
case PROP_DISPATCHER:
self->dispatcher = g_value_get_pointer (value);
break;
default:
G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
break;
}
}
static gboolean
gst_wasapi2_client_main_loop_running_cb (GstWasapi2Client * self)
{
GST_DEBUG_OBJECT (self, "Main loop running now");
g_mutex_lock (&self->lock);
g_cond_signal (&self->cond);
g_mutex_unlock (&self->lock);
return G_SOURCE_REMOVE;
}
static void
gst_wasapi2_client_on_device_activated (GstWasapi2Client * self,
wasapi2: Rewrite plugin and implement audioringbuffer subclass ... based on MediaFoundation work queue API. By this commit, wasapi2 plugin will make use of pull mode scheduling with audioringbuffer subclass. There are several drawbacks of audiosrc/audiosink subclassing (not audiobasesrc/audiobasesink) for WASAPI API, which are: * audiosrc/audiosink classes try to set high priority to read/write thread via MMCSS (Multimedia Class Scheduler Service) but it's not allowed in case of UWP application. In order to use MMCSS in UWP, application should use MediaFoundation work queue indirectly. Since audiosrc/audiosink scheduling model is not compatible with MediaFoundation's work queue model, audioringbuffer subclassing is required. * WASAPI capture device might report larger packet size than expected (i.e., larger frames we can read than expected frame size per period). Meanwhile, in any case, application should drain all packets at that moment. In order to handle the case, wasapi/wasapi2 plugins were making use of GstAdapter which is obviously sub-optimal because it requires additional memory allocation and copy. By implementing audioringbuffer subclassing, we can avoid such inefficiency. In this commit, all the device read/write operations will be moved to newly implemented wasapi2ringbuffer class and existing wasapi2client class will take care of device enumeration and activation parts only. Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2306>
2021-05-10 11:45:28 +00:00
IAudioClient * audio_client)
{
GST_INFO_OBJECT (self, "Device activated");
g_mutex_lock (&self->init_lock);
if (audio_client) {
audio_client->AddRef ();
self->audio_client = audio_client;
self->activate_state = GST_WASAPI2_CLIENT_ACTIVATE_DONE;
} else {
GST_WARNING_OBJECT (self, "IAudioClient is unavailable");
self->activate_state = GST_WASAPI2_CLIENT_ACTIVATE_FAILED;
}
g_cond_broadcast (&self->init_cond);
g_mutex_unlock (&self->init_lock);
}
/* *INDENT-OFF* */
static std::string
convert_wstring_to_string (const std::wstring &wstr)
{
std::wstring_convert<std::codecvt_utf8<wchar_t>, wchar_t> converter;
return converter.to_bytes (wstr.c_str());
}
static std::string
convert_hstring_to_string (HString * hstr)
{
const wchar_t *raw_hstr;
if (!hstr)
return std::string();
raw_hstr = hstr->GetRawBuffer (nullptr);
if (!raw_hstr)
return std::string();
return convert_wstring_to_string (std::wstring (raw_hstr));
}
static std::wstring
gst_wasapi2_client_get_default_device_id (GstWasapi2Client * self)
{
HRESULT hr;
PWSTR default_device_id_wstr = nullptr;
if (self->device_class == GST_WASAPI2_CLIENT_DEVICE_CLASS_CAPTURE)
hr = StringFromIID (DEVINTERFACE_AUDIO_CAPTURE, &default_device_id_wstr);
else
hr = StringFromIID (DEVINTERFACE_AUDIO_RENDER, &default_device_id_wstr);
if (!gst_wasapi2_result (hr))
return std::wstring();
std::wstring ret = std::wstring (default_device_id_wstr);
CoTaskMemFree (default_device_id_wstr);
return ret;
}
/* *INDENT-ON* */
static gboolean
gst_wasapi2_client_activate_async (GstWasapi2Client * self,
GstWasapiDeviceActivator * activator)
{
/* *INDENT-OFF* */
ComPtr<IDeviceInformationStatics> device_info_static;
ComPtr<IAsyncOperation<DeviceInformationCollection*>> async_op;
ComPtr<IVectorView<DeviceInformation*>> device_list;
HStringReference hstr_device_info =
HStringReference(RuntimeClass_Windows_Devices_Enumeration_DeviceInformation);
/* *INDENT-ON* */
HRESULT hr;
DeviceClass device_class;
unsigned int count = 0;
gint device_index = 0;
std::wstring default_device_id_wstring;
std::string default_device_id;
std::wstring target_device_id_wstring;
std::string target_device_id;
std::string target_device_name;
gboolean use_default_device = FALSE;
GST_INFO_OBJECT (self,
"requested device info, device-class: %s, device: %s, device-index: %d",
self->device_class ==
GST_WASAPI2_CLIENT_DEVICE_CLASS_CAPTURE ? "capture" : "render",
GST_STR_NULL (self->device_id), self->device_index);
if (self->device_class == GST_WASAPI2_CLIENT_DEVICE_CLASS_CAPTURE) {
device_class = DeviceClass::DeviceClass_AudioCapture;
} else {
device_class = DeviceClass::DeviceClass_AudioRender;
}
default_device_id_wstring = gst_wasapi2_client_get_default_device_id (self);
if (default_device_id_wstring.empty ()) {
GST_WARNING_OBJECT (self, "Couldn't get default device id");
goto failed;
}
default_device_id = convert_wstring_to_string (default_device_id_wstring);
GST_DEBUG_OBJECT (self, "Default device id: %s", default_device_id.c_str ());
/* When
* 1) default device was requested or
* 2) no explicitly requested device or
* 3) requested device string id is null but device index is zero
* will use default device
*
* Note that default device is much preferred
* See https://docs.microsoft.com/en-us/windows/win32/coreaudio/automatic-stream-routing
*/
if (self->device_id &&
g_ascii_strcasecmp (self->device_id, default_device_id.c_str ()) == 0) {
GST_DEBUG_OBJECT (self, "Default device was requested");
use_default_device = TRUE;
} else if (self->device_index < 0 && !self->device_id) {
GST_DEBUG_OBJECT (self,
"No device was explicitly requested, use default device");
use_default_device = TRUE;
} else if (!self->device_id && self->device_index == 0) {
GST_DEBUG_OBJECT (self, "device-index == zero means default device");
use_default_device = TRUE;
}
if (use_default_device) {
target_device_id_wstring = default_device_id_wstring;
target_device_id = default_device_id;
if (self->device_class == GST_WASAPI2_CLIENT_DEVICE_CLASS_CAPTURE)
target_device_name = "Default Audio Capture Device";
else
target_device_name = "Default Audio Render Device";
goto activate;
}
hr = GetActivationFactory (hstr_device_info.Get (), &device_info_static);
if (!gst_wasapi2_result (hr))
goto failed;
hr = device_info_static->FindAllAsyncDeviceClass (device_class, &async_op);
device_info_static.Reset ();
if (!gst_wasapi2_result (hr))
goto failed;
/* *INDENT-OFF* */
hr = SyncWait<DeviceInformationCollection*>(async_op.Get ());
/* *INDENT-ON* */
if (!gst_wasapi2_result (hr))
goto failed;
hr = async_op->GetResults (&device_list);
async_op.Reset ();
if (!gst_wasapi2_result (hr))
goto failed;
hr = device_list->get_Size (&count);
if (!gst_wasapi2_result (hr))
goto failed;
if (count == 0) {
GST_WARNING_OBJECT (self, "No available device");
goto failed;
}
/* device_index 0 will be assigned for default device
* so the number of available device is count + 1 (for default device) */
if (self->device_index >= 0 && self->device_index > (gint) count) {
GST_WARNING_OBJECT (self, "Device index %d is unavailable",
self->device_index);
goto failed;
}
GST_DEBUG_OBJECT (self, "Available device count: %d", count);
/* zero is for default device */
device_index = 1;
for (unsigned int i = 0; i < count; i++) {
/* *INDENT-OFF* */
ComPtr<IDeviceInformation> device_info;
/* *INDENT-ON* */
HString id;
HString name;
boolean b_value;
std::string cur_device_id;
std::string cur_device_name;
hr = device_list->GetAt (i, &device_info);
if (!gst_wasapi2_result (hr))
continue;
hr = device_info->get_IsEnabled (&b_value);
if (!gst_wasapi2_result (hr))
continue;
/* select only enabled device */
if (!b_value) {
GST_DEBUG_OBJECT (self, "Device index %d is disabled", i);
continue;
}
/* To ensure device id and device name are available,
* will query this later again once target device is determined */
hr = device_info->get_Id (id.GetAddressOf ());
if (!gst_wasapi2_result (hr))
continue;
if (!id.IsValid ()) {
GST_WARNING_OBJECT (self, "Device index %d has invalid id", i);
continue;
}
hr = device_info->get_Name (name.GetAddressOf ());
if (!gst_wasapi2_result (hr))
continue;
if (!name.IsValid ()) {
GST_WARNING_OBJECT (self, "Device index %d has invalid name", i);
continue;
}
cur_device_id = convert_hstring_to_string (&id);
if (cur_device_id.empty ()) {
GST_WARNING_OBJECT (self, "Device index %d has empty id", i);
continue;
}
cur_device_name = convert_hstring_to_string (&name);
if (cur_device_name.empty ()) {
GST_WARNING_OBJECT (self, "Device index %d has empty device name", i);
continue;
}
GST_DEBUG_OBJECT (self, "device [%d] id: %s, name: %s",
device_index, cur_device_id.c_str (), cur_device_name.c_str ());
if (self->device_id &&
g_ascii_strcasecmp (self->device_id, cur_device_id.c_str ()) == 0) {
GST_INFO_OBJECT (self,
"Device index %d has matching device id %s", device_index,
cur_device_id.c_str ());
target_device_id_wstring = id.GetRawBuffer (nullptr);
target_device_id = cur_device_id;
target_device_name = cur_device_name;
break;
}
if (self->device_index >= 0 && self->device_index == device_index) {
GST_INFO_OBJECT (self, "Select device index %d, device id %s",
device_index, cur_device_id.c_str ());
target_device_id_wstring = id.GetRawBuffer (nullptr);
target_device_id = cur_device_id;
target_device_name = cur_device_name;
break;
}
/* count only available devices */
device_index++;
}
if (target_device_id_wstring.empty ()) {
GST_WARNING_OBJECT (self, "Couldn't find target device");
goto failed;
}
activate:
/* fill device id and name */
g_free (self->device_id);
self->device_id = g_strdup (target_device_id.c_str ());
g_free (self->device_name);
self->device_name = g_strdup (target_device_name.c_str ());
self->device_index = device_index;
/* default device supports automatic stream routing */
self->can_auto_routing = use_default_device;
hr = activator->ActivateDeviceAsync (target_device_id_wstring);
if (!gst_wasapi2_result (hr)) {
GST_WARNING_OBJECT (self, "Failed to activate device");
goto failed;
}
g_mutex_lock (&self->lock);
if (self->activate_state == GST_WASAPI2_CLIENT_ACTIVATE_INIT)
self->activate_state = GST_WASAPI2_CLIENT_ACTIVATE_WAIT;
g_mutex_unlock (&self->lock);
return TRUE;
failed:
self->activate_state = GST_WASAPI2_CLIENT_ACTIVATE_FAILED;
return FALSE;
}
static const gchar *
activate_state_to_string (GstWasapi2ClientActivateState state)
{
switch (state) {
case GST_WASAPI2_CLIENT_ACTIVATE_FAILED:
return "FAILED";
case GST_WASAPI2_CLIENT_ACTIVATE_INIT:
return "INIT";
case GST_WASAPI2_CLIENT_ACTIVATE_WAIT:
return "WAIT";
case GST_WASAPI2_CLIENT_ACTIVATE_DONE:
return "DONE";
}
g_assert_not_reached ();
return "Undefined";
}
static gpointer
gst_wasapi2_client_thread_func (GstWasapi2Client * self)
{
RoInitializeWrapper initialize (RO_INIT_MULTITHREADED);
GSource *source;
HRESULT hr;
/* *INDENT-OFF* */
ComPtr<GstWasapiDeviceActivator> activator;
hr = MakeAndInitialize<GstWasapiDeviceActivator> (&activator,
self, self->dispatcher);
/* *INDENT-ON* */
if (!gst_wasapi2_result (hr)) {
GST_ERROR_OBJECT (self, "Could not create activator object");
self->activate_state = GST_WASAPI2_CLIENT_ACTIVATE_FAILED;
goto run_loop;
}
gst_wasapi2_client_activate_async (self, activator.Get ());
if (!self->dispatcher) {
/* In case that dispatcher is unavailable, wait activation synchroniously */
GST_DEBUG_OBJECT (self, "Wait device activation");
gst_wasapi2_client_ensure_activation (self);
GST_DEBUG_OBJECT (self, "Device activation result %s",
activate_state_to_string (self->activate_state));
}
run_loop:
g_main_context_push_thread_default (self->context);
source = g_idle_source_new ();
g_source_set_callback (source,
wasapi2: Rewrite plugin and implement audioringbuffer subclass ... based on MediaFoundation work queue API. By this commit, wasapi2 plugin will make use of pull mode scheduling with audioringbuffer subclass. There are several drawbacks of audiosrc/audiosink subclassing (not audiobasesrc/audiobasesink) for WASAPI API, which are: * audiosrc/audiosink classes try to set high priority to read/write thread via MMCSS (Multimedia Class Scheduler Service) but it's not allowed in case of UWP application. In order to use MMCSS in UWP, application should use MediaFoundation work queue indirectly. Since audiosrc/audiosink scheduling model is not compatible with MediaFoundation's work queue model, audioringbuffer subclassing is required. * WASAPI capture device might report larger packet size than expected (i.e., larger frames we can read than expected frame size per period). Meanwhile, in any case, application should drain all packets at that moment. In order to handle the case, wasapi/wasapi2 plugins were making use of GstAdapter which is obviously sub-optimal because it requires additional memory allocation and copy. By implementing audioringbuffer subclassing, we can avoid such inefficiency. In this commit, all the device read/write operations will be moved to newly implemented wasapi2ringbuffer class and existing wasapi2client class will take care of device enumeration and activation parts only. Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2306>
2021-05-10 11:45:28 +00:00
(GSourceFunc) gst_wasapi2_client_main_loop_running_cb, self, nullptr);
g_source_attach (source, self->context);
g_source_unref (source);
GST_DEBUG_OBJECT (self, "Starting main loop");
g_main_loop_run (self->loop);
GST_DEBUG_OBJECT (self, "Stopped main loop");
g_main_context_pop_thread_default (self->context);
wasapi2: Rewrite plugin and implement audioringbuffer subclass ... based on MediaFoundation work queue API. By this commit, wasapi2 plugin will make use of pull mode scheduling with audioringbuffer subclass. There are several drawbacks of audiosrc/audiosink subclassing (not audiobasesrc/audiobasesink) for WASAPI API, which are: * audiosrc/audiosink classes try to set high priority to read/write thread via MMCSS (Multimedia Class Scheduler Service) but it's not allowed in case of UWP application. In order to use MMCSS in UWP, application should use MediaFoundation work queue indirectly. Since audiosrc/audiosink scheduling model is not compatible with MediaFoundation's work queue model, audioringbuffer subclassing is required. * WASAPI capture device might report larger packet size than expected (i.e., larger frames we can read than expected frame size per period). Meanwhile, in any case, application should drain all packets at that moment. In order to handle the case, wasapi/wasapi2 plugins were making use of GstAdapter which is obviously sub-optimal because it requires additional memory allocation and copy. By implementing audioringbuffer subclassing, we can avoid such inefficiency. In this commit, all the device read/write operations will be moved to newly implemented wasapi2ringbuffer class and existing wasapi2client class will take care of device enumeration and activation parts only. Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2306>
2021-05-10 11:45:28 +00:00
GST_WASAPI2_CLEAR_COM (self->audio_client);
/* Reset explicitly to ensure that it happens before
* RoInitializeWrapper dtor is called */
activator.Reset ();
GST_DEBUG_OBJECT (self, "Exit thread function");
wasapi2: Rewrite plugin and implement audioringbuffer subclass ... based on MediaFoundation work queue API. By this commit, wasapi2 plugin will make use of pull mode scheduling with audioringbuffer subclass. There are several drawbacks of audiosrc/audiosink subclassing (not audiobasesrc/audiobasesink) for WASAPI API, which are: * audiosrc/audiosink classes try to set high priority to read/write thread via MMCSS (Multimedia Class Scheduler Service) but it's not allowed in case of UWP application. In order to use MMCSS in UWP, application should use MediaFoundation work queue indirectly. Since audiosrc/audiosink scheduling model is not compatible with MediaFoundation's work queue model, audioringbuffer subclassing is required. * WASAPI capture device might report larger packet size than expected (i.e., larger frames we can read than expected frame size per period). Meanwhile, in any case, application should drain all packets at that moment. In order to handle the case, wasapi/wasapi2 plugins were making use of GstAdapter which is obviously sub-optimal because it requires additional memory allocation and copy. By implementing audioringbuffer subclassing, we can avoid such inefficiency. In this commit, all the device read/write operations will be moved to newly implemented wasapi2ringbuffer class and existing wasapi2client class will take care of device enumeration and activation parts only. Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2306>
2021-05-10 11:45:28 +00:00
return nullptr;
}
GstCaps *
gst_wasapi2_client_get_caps (GstWasapi2Client * client)
{
wasapi2: Rewrite plugin and implement audioringbuffer subclass ... based on MediaFoundation work queue API. By this commit, wasapi2 plugin will make use of pull mode scheduling with audioringbuffer subclass. There are several drawbacks of audiosrc/audiosink subclassing (not audiobasesrc/audiobasesink) for WASAPI API, which are: * audiosrc/audiosink classes try to set high priority to read/write thread via MMCSS (Multimedia Class Scheduler Service) but it's not allowed in case of UWP application. In order to use MMCSS in UWP, application should use MediaFoundation work queue indirectly. Since audiosrc/audiosink scheduling model is not compatible with MediaFoundation's work queue model, audioringbuffer subclassing is required. * WASAPI capture device might report larger packet size than expected (i.e., larger frames we can read than expected frame size per period). Meanwhile, in any case, application should drain all packets at that moment. In order to handle the case, wasapi/wasapi2 plugins were making use of GstAdapter which is obviously sub-optimal because it requires additional memory allocation and copy. By implementing audioringbuffer subclassing, we can avoid such inefficiency. In this commit, all the device read/write operations will be moved to newly implemented wasapi2ringbuffer class and existing wasapi2client class will take care of device enumeration and activation parts only. Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2306>
2021-05-10 11:45:28 +00:00
WAVEFORMATEX *mix_format = nullptr;
static GstStaticCaps static_caps = GST_STATIC_CAPS (GST_WASAPI2_STATIC_CAPS);
GstCaps *scaps;
HRESULT hr;
wasapi2: Rewrite plugin and implement audioringbuffer subclass ... based on MediaFoundation work queue API. By this commit, wasapi2 plugin will make use of pull mode scheduling with audioringbuffer subclass. There are several drawbacks of audiosrc/audiosink subclassing (not audiobasesrc/audiobasesink) for WASAPI API, which are: * audiosrc/audiosink classes try to set high priority to read/write thread via MMCSS (Multimedia Class Scheduler Service) but it's not allowed in case of UWP application. In order to use MMCSS in UWP, application should use MediaFoundation work queue indirectly. Since audiosrc/audiosink scheduling model is not compatible with MediaFoundation's work queue model, audioringbuffer subclassing is required. * WASAPI capture device might report larger packet size than expected (i.e., larger frames we can read than expected frame size per period). Meanwhile, in any case, application should drain all packets at that moment. In order to handle the case, wasapi/wasapi2 plugins were making use of GstAdapter which is obviously sub-optimal because it requires additional memory allocation and copy. By implementing audioringbuffer subclassing, we can avoid such inefficiency. In this commit, all the device read/write operations will be moved to newly implemented wasapi2ringbuffer class and existing wasapi2client class will take care of device enumeration and activation parts only. Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2306>
2021-05-10 11:45:28 +00:00
g_return_val_if_fail (GST_IS_WASAPI2_CLIENT (client), nullptr);
if (client->supported_caps)
return gst_caps_ref (client->supported_caps);
if (!client->audio_client) {
GST_WARNING_OBJECT (client, "IAudioClient3 wasn't configured");
wasapi2: Rewrite plugin and implement audioringbuffer subclass ... based on MediaFoundation work queue API. By this commit, wasapi2 plugin will make use of pull mode scheduling with audioringbuffer subclass. There are several drawbacks of audiosrc/audiosink subclassing (not audiobasesrc/audiobasesink) for WASAPI API, which are: * audiosrc/audiosink classes try to set high priority to read/write thread via MMCSS (Multimedia Class Scheduler Service) but it's not allowed in case of UWP application. In order to use MMCSS in UWP, application should use MediaFoundation work queue indirectly. Since audiosrc/audiosink scheduling model is not compatible with MediaFoundation's work queue model, audioringbuffer subclassing is required. * WASAPI capture device might report larger packet size than expected (i.e., larger frames we can read than expected frame size per period). Meanwhile, in any case, application should drain all packets at that moment. In order to handle the case, wasapi/wasapi2 plugins were making use of GstAdapter which is obviously sub-optimal because it requires additional memory allocation and copy. By implementing audioringbuffer subclassing, we can avoid such inefficiency. In this commit, all the device read/write operations will be moved to newly implemented wasapi2ringbuffer class and existing wasapi2client class will take care of device enumeration and activation parts only. Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2306>
2021-05-10 11:45:28 +00:00
return nullptr;
}
wasapi2: Rewrite plugin and implement audioringbuffer subclass ... based on MediaFoundation work queue API. By this commit, wasapi2 plugin will make use of pull mode scheduling with audioringbuffer subclass. There are several drawbacks of audiosrc/audiosink subclassing (not audiobasesrc/audiobasesink) for WASAPI API, which are: * audiosrc/audiosink classes try to set high priority to read/write thread via MMCSS (Multimedia Class Scheduler Service) but it's not allowed in case of UWP application. In order to use MMCSS in UWP, application should use MediaFoundation work queue indirectly. Since audiosrc/audiosink scheduling model is not compatible with MediaFoundation's work queue model, audioringbuffer subclassing is required. * WASAPI capture device might report larger packet size than expected (i.e., larger frames we can read than expected frame size per period). Meanwhile, in any case, application should drain all packets at that moment. In order to handle the case, wasapi/wasapi2 plugins were making use of GstAdapter which is obviously sub-optimal because it requires additional memory allocation and copy. By implementing audioringbuffer subclassing, we can avoid such inefficiency. In this commit, all the device read/write operations will be moved to newly implemented wasapi2ringbuffer class and existing wasapi2client class will take care of device enumeration and activation parts only. Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2306>
2021-05-10 11:45:28 +00:00
hr = client->audio_client->GetMixFormat (&mix_format);
if (!gst_wasapi2_result (hr)) {
GST_WARNING_OBJECT (client, "Failed to get mix format");
return nullptr;
}
scaps = gst_static_caps_get (&static_caps);
wasapi2: Rewrite plugin and implement audioringbuffer subclass ... based on MediaFoundation work queue API. By this commit, wasapi2 plugin will make use of pull mode scheduling with audioringbuffer subclass. There are several drawbacks of audiosrc/audiosink subclassing (not audiobasesrc/audiobasesink) for WASAPI API, which are: * audiosrc/audiosink classes try to set high priority to read/write thread via MMCSS (Multimedia Class Scheduler Service) but it's not allowed in case of UWP application. In order to use MMCSS in UWP, application should use MediaFoundation work queue indirectly. Since audiosrc/audiosink scheduling model is not compatible with MediaFoundation's work queue model, audioringbuffer subclassing is required. * WASAPI capture device might report larger packet size than expected (i.e., larger frames we can read than expected frame size per period). Meanwhile, in any case, application should drain all packets at that moment. In order to handle the case, wasapi/wasapi2 plugins were making use of GstAdapter which is obviously sub-optimal because it requires additional memory allocation and copy. By implementing audioringbuffer subclassing, we can avoid such inefficiency. In this commit, all the device read/write operations will be moved to newly implemented wasapi2ringbuffer class and existing wasapi2client class will take care of device enumeration and activation parts only. Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2306>
2021-05-10 11:45:28 +00:00
gst_wasapi2_util_parse_waveformatex (mix_format,
scaps, &client->supported_caps, nullptr);
gst_caps_unref (scaps);
wasapi2: Rewrite plugin and implement audioringbuffer subclass ... based on MediaFoundation work queue API. By this commit, wasapi2 plugin will make use of pull mode scheduling with audioringbuffer subclass. There are several drawbacks of audiosrc/audiosink subclassing (not audiobasesrc/audiobasesink) for WASAPI API, which are: * audiosrc/audiosink classes try to set high priority to read/write thread via MMCSS (Multimedia Class Scheduler Service) but it's not allowed in case of UWP application. In order to use MMCSS in UWP, application should use MediaFoundation work queue indirectly. Since audiosrc/audiosink scheduling model is not compatible with MediaFoundation's work queue model, audioringbuffer subclassing is required. * WASAPI capture device might report larger packet size than expected (i.e., larger frames we can read than expected frame size per period). Meanwhile, in any case, application should drain all packets at that moment. In order to handle the case, wasapi/wasapi2 plugins were making use of GstAdapter which is obviously sub-optimal because it requires additional memory allocation and copy. By implementing audioringbuffer subclassing, we can avoid such inefficiency. In this commit, all the device read/write operations will be moved to newly implemented wasapi2ringbuffer class and existing wasapi2client class will take care of device enumeration and activation parts only. Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2306>
2021-05-10 11:45:28 +00:00
CoTaskMemFree (mix_format);
if (!client->supported_caps) {
GST_ERROR_OBJECT (client, "No caps from subclass");
wasapi2: Rewrite plugin and implement audioringbuffer subclass ... based on MediaFoundation work queue API. By this commit, wasapi2 plugin will make use of pull mode scheduling with audioringbuffer subclass. There are several drawbacks of audiosrc/audiosink subclassing (not audiobasesrc/audiobasesink) for WASAPI API, which are: * audiosrc/audiosink classes try to set high priority to read/write thread via MMCSS (Multimedia Class Scheduler Service) but it's not allowed in case of UWP application. In order to use MMCSS in UWP, application should use MediaFoundation work queue indirectly. Since audiosrc/audiosink scheduling model is not compatible with MediaFoundation's work queue model, audioringbuffer subclassing is required. * WASAPI capture device might report larger packet size than expected (i.e., larger frames we can read than expected frame size per period). Meanwhile, in any case, application should drain all packets at that moment. In order to handle the case, wasapi/wasapi2 plugins were making use of GstAdapter which is obviously sub-optimal because it requires additional memory allocation and copy. By implementing audioringbuffer subclassing, we can avoid such inefficiency. In this commit, all the device read/write operations will be moved to newly implemented wasapi2ringbuffer class and existing wasapi2client class will take care of device enumeration and activation parts only. Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2306>
2021-05-10 11:45:28 +00:00
return nullptr;
}
return gst_caps_ref (client->supported_caps);
}
gboolean
gst_wasapi2_client_ensure_activation (GstWasapi2Client * client)
{
g_return_val_if_fail (GST_IS_WASAPI2_CLIENT (client), FALSE);
/* should not happen */
g_assert (client->activate_state != GST_WASAPI2_CLIENT_ACTIVATE_INIT);
g_mutex_lock (&client->init_lock);
while (client->activate_state == GST_WASAPI2_CLIENT_ACTIVATE_WAIT)
g_cond_wait (&client->init_cond, &client->init_lock);
g_mutex_unlock (&client->init_lock);
return client->activate_state == GST_WASAPI2_CLIENT_ACTIVATE_DONE;
}
static HRESULT
find_dispatcher (ICoreDispatcher ** dispatcher)
{
/* *INDENT-OFF* */
HStringReference hstr_core_app =
HStringReference(RuntimeClass_Windows_ApplicationModel_Core_CoreApplication);
ComPtr<ICoreApplication> core_app;
ComPtr<ICoreApplicationView> core_app_view;
ComPtr<ICoreWindow> core_window;
/* *INDENT-ON* */
HRESULT hr;
hr = GetActivationFactory (hstr_core_app.Get (), &core_app);
if (FAILED (hr))
return hr;
hr = core_app->GetCurrentView (&core_app_view);
if (FAILED (hr))
return hr;
hr = core_app_view->get_CoreWindow (&core_window);
if (FAILED (hr))
return hr;
return core_window->get_Dispatcher (dispatcher);
}
GstWasapi2Client *
gst_wasapi2_client_new (GstWasapi2ClientDeviceClass device_class,
wasapi2: Rewrite plugin and implement audioringbuffer subclass ... based on MediaFoundation work queue API. By this commit, wasapi2 plugin will make use of pull mode scheduling with audioringbuffer subclass. There are several drawbacks of audiosrc/audiosink subclassing (not audiobasesrc/audiobasesink) for WASAPI API, which are: * audiosrc/audiosink classes try to set high priority to read/write thread via MMCSS (Multimedia Class Scheduler Service) but it's not allowed in case of UWP application. In order to use MMCSS in UWP, application should use MediaFoundation work queue indirectly. Since audiosrc/audiosink scheduling model is not compatible with MediaFoundation's work queue model, audioringbuffer subclassing is required. * WASAPI capture device might report larger packet size than expected (i.e., larger frames we can read than expected frame size per period). Meanwhile, in any case, application should drain all packets at that moment. In order to handle the case, wasapi/wasapi2 plugins were making use of GstAdapter which is obviously sub-optimal because it requires additional memory allocation and copy. By implementing audioringbuffer subclassing, we can avoid such inefficiency. In this commit, all the device read/write operations will be moved to newly implemented wasapi2ringbuffer class and existing wasapi2client class will take care of device enumeration and activation parts only. Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2306>
2021-05-10 11:45:28 +00:00
gint device_index, const gchar * device_id, gpointer dispatcher)
{
GstWasapi2Client *self;
/* *INDENT-OFF* */
ComPtr<ICoreDispatcher> core_dispatcher;
/* *INDENT-ON* */
/* Multiple COM init is allowed */
RoInitializeWrapper init_wrapper (RO_INIT_MULTITHREADED);
/* If application didn't pass ICoreDispatcher object,
* try to get dispatcher object for the current thread */
if (!dispatcher) {
HRESULT hr;
hr = find_dispatcher (&core_dispatcher);
if (SUCCEEDED (hr)) {
GST_DEBUG ("UI dispatcher is available");
dispatcher = core_dispatcher.Get ();
} else {
GST_DEBUG ("UI dispatcher is unavailable");
}
} else {
GST_DEBUG ("Use user passed UI dispatcher");
}
self = (GstWasapi2Client *) g_object_new (GST_TYPE_WASAPI2_CLIENT,
wasapi2: Rewrite plugin and implement audioringbuffer subclass ... based on MediaFoundation work queue API. By this commit, wasapi2 plugin will make use of pull mode scheduling with audioringbuffer subclass. There are several drawbacks of audiosrc/audiosink subclassing (not audiobasesrc/audiobasesink) for WASAPI API, which are: * audiosrc/audiosink classes try to set high priority to read/write thread via MMCSS (Multimedia Class Scheduler Service) but it's not allowed in case of UWP application. In order to use MMCSS in UWP, application should use MediaFoundation work queue indirectly. Since audiosrc/audiosink scheduling model is not compatible with MediaFoundation's work queue model, audioringbuffer subclassing is required. * WASAPI capture device might report larger packet size than expected (i.e., larger frames we can read than expected frame size per period). Meanwhile, in any case, application should drain all packets at that moment. In order to handle the case, wasapi/wasapi2 plugins were making use of GstAdapter which is obviously sub-optimal because it requires additional memory allocation and copy. By implementing audioringbuffer subclassing, we can avoid such inefficiency. In this commit, all the device read/write operations will be moved to newly implemented wasapi2ringbuffer class and existing wasapi2client class will take care of device enumeration and activation parts only. Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2306>
2021-05-10 11:45:28 +00:00
"device-class", device_class, "device-index", device_index,
"device", device_id, "dispatcher", dispatcher, nullptr);
/* Reset explicitly to ensure that it happens before
* RoInitializeWrapper dtor is called */
core_dispatcher.Reset ();
if (self->activate_state == GST_WASAPI2_CLIENT_ACTIVATE_FAILED) {
gst_object_unref (self);
wasapi2: Rewrite plugin and implement audioringbuffer subclass ... based on MediaFoundation work queue API. By this commit, wasapi2 plugin will make use of pull mode scheduling with audioringbuffer subclass. There are several drawbacks of audiosrc/audiosink subclassing (not audiobasesrc/audiobasesink) for WASAPI API, which are: * audiosrc/audiosink classes try to set high priority to read/write thread via MMCSS (Multimedia Class Scheduler Service) but it's not allowed in case of UWP application. In order to use MMCSS in UWP, application should use MediaFoundation work queue indirectly. Since audiosrc/audiosink scheduling model is not compatible with MediaFoundation's work queue model, audioringbuffer subclassing is required. * WASAPI capture device might report larger packet size than expected (i.e., larger frames we can read than expected frame size per period). Meanwhile, in any case, application should drain all packets at that moment. In order to handle the case, wasapi/wasapi2 plugins were making use of GstAdapter which is obviously sub-optimal because it requires additional memory allocation and copy. By implementing audioringbuffer subclassing, we can avoid such inefficiency. In this commit, all the device read/write operations will be moved to newly implemented wasapi2ringbuffer class and existing wasapi2client class will take care of device enumeration and activation parts only. Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2306>
2021-05-10 11:45:28 +00:00
return nullptr;
}
gst_object_ref_sink (self);
return self;
}
wasapi2: Rewrite plugin and implement audioringbuffer subclass ... based on MediaFoundation work queue API. By this commit, wasapi2 plugin will make use of pull mode scheduling with audioringbuffer subclass. There are several drawbacks of audiosrc/audiosink subclassing (not audiobasesrc/audiobasesink) for WASAPI API, which are: * audiosrc/audiosink classes try to set high priority to read/write thread via MMCSS (Multimedia Class Scheduler Service) but it's not allowed in case of UWP application. In order to use MMCSS in UWP, application should use MediaFoundation work queue indirectly. Since audiosrc/audiosink scheduling model is not compatible with MediaFoundation's work queue model, audioringbuffer subclassing is required. * WASAPI capture device might report larger packet size than expected (i.e., larger frames we can read than expected frame size per period). Meanwhile, in any case, application should drain all packets at that moment. In order to handle the case, wasapi/wasapi2 plugins were making use of GstAdapter which is obviously sub-optimal because it requires additional memory allocation and copy. By implementing audioringbuffer subclassing, we can avoid such inefficiency. In this commit, all the device read/write operations will be moved to newly implemented wasapi2ringbuffer class and existing wasapi2client class will take care of device enumeration and activation parts only. Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2306>
2021-05-10 11:45:28 +00:00
IAudioClient *
gst_wasapi2_client_get_handle (GstWasapi2Client * client)
{
g_return_val_if_fail (GST_IS_WASAPI2_CLIENT (client), nullptr);
return client->audio_client;
}