/*
* 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>
*
* 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.
*/
/**
* SECTION:element-wasapisink
* @title: wasapisink
*
* Provides audio playback using the Windows Audio Session API available with
* Vista and newer.
*
* ## Example pipelines
* |[
* gst-launch-1.0 -v audiotestsrc samplesperbuffer=160 ! wasapisink
* ]| Generate 20 ms buffers and render to the default audio device.
*
* |[
* gst-launch-1.0 -v audiotestsrc samplesperbuffer=160 ! wasapisink low-latency=true
* ]| Same as above, but with the minimum possible latency
*
*/
#ifdef HAVE_CONFIG_H
# include <config.h>
#endif
#include "gstwasapisink.h"
#include <avrt.h>
GST_DEBUG_CATEGORY_STATIC (gst_wasapi_sink_debug);
#define GST_CAT_DEFAULT gst_wasapi_sink_debug
static GstStaticPadTemplate sink_template = GST_STATIC_PAD_TEMPLATE ("sink",
GST_PAD_SINK,
GST_PAD_ALWAYS,
GST_STATIC_CAPS (GST_WASAPI_STATIC_CAPS));
#define DEFAULT_ROLE GST_WASAPI_DEVICE_ROLE_CONSOLE
#define DEFAULT_MUTE FALSE
#define DEFAULT_EXCLUSIVE FALSE
#define DEFAULT_LOW_LATENCY FALSE
#define DEFAULT_AUDIOCLIENT3 TRUE
enum
{
PROP_0,
PROP_ROLE,
PROP_MUTE,
PROP_DEVICE,
PROP_EXCLUSIVE,
PROP_LOW_LATENCY,
PROP_AUDIOCLIENT3
};
static void gst_wasapi_sink_dispose (GObject * object);
static void gst_wasapi_sink_finalize (GObject * object);
static void gst_wasapi_sink_set_property (GObject * object, guint prop_id,
const GValue * value, GParamSpec * pspec);
static void gst_wasapi_sink_get_property (GObject * object, guint prop_id,
GValue * value, GParamSpec * pspec);
static GstCaps *gst_wasapi_sink_get_caps (GstBaseSink * bsink,
GstCaps * filter);
static gboolean gst_wasapi_sink_prepare (GstAudioSink * asink,
GstAudioRingBufferSpec * spec);
static gboolean gst_wasapi_sink_unprepare (GstAudioSink * asink);
static gboolean gst_wasapi_sink_open (GstAudioSink * asink);
static gboolean gst_wasapi_sink_close (GstAudioSink * asink);
static gint gst_wasapi_sink_write (GstAudioSink * asink,
gpointer data, guint length);
static guint gst_wasapi_sink_delay (GstAudioSink * asink);
static void gst_wasapi_sink_reset (GstAudioSink * asink);
#define gst_wasapi_sink_parent_class parent_class
G_DEFINE_TYPE (GstWasapiSink, gst_wasapi_sink, GST_TYPE_AUDIO_SINK);
static void
gst_wasapi_sink_class_init (GstWasapiSinkClass * klass)
{
GObjectClass *gobject_class = G_OBJECT_CLASS (klass);
GstElementClass *gstelement_class = GST_ELEMENT_CLASS (klass);
GstBaseSinkClass *gstbasesink_class = GST_BASE_SINK_CLASS (klass);
GstAudioSinkClass *gstaudiosink_class = GST_AUDIO_SINK_CLASS (klass);
gobject_class->dispose = gst_wasapi_sink_dispose;
gobject_class->finalize = gst_wasapi_sink_finalize;
gobject_class->set_property = gst_wasapi_sink_set_property;
gobject_class->get_property = gst_wasapi_sink_get_property;
g_object_class_install_property (gobject_class,
PROP_ROLE,
g_param_spec_enum ("role", "Role",
"Role of the device: communications, multimedia, etc",
GST_WASAPI_DEVICE_TYPE_ROLE, DEFAULT_ROLE, G_PARAM_READWRITE |
G_PARAM_STATIC_STRINGS | GST_PARAM_MUTABLE_READY));
g_object_class_install_property (gobject_class,
PROP_MUTE,
g_param_spec_boolean ("mute", "Mute", "Mute state of this stream",
DEFAULT_MUTE, G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS |
GST_PARAM_MUTABLE_PLAYING));
g_object_class_install_property (gobject_class,
PROP_DEVICE,
g_param_spec_string ("device", "Device",
"WASAPI playback device as a GUID string",
NULL, G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
g_object_class_install_property (gobject_class,
PROP_EXCLUSIVE,
g_param_spec_boolean ("exclusive", "Exclusive mode",
"Open the device in exclusive mode",
DEFAULT_EXCLUSIVE, G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
g_object_class_install_property (gobject_class,
PROP_LOW_LATENCY,
g_param_spec_boolean ("low-latency", "Low latency",
"Optimize all settings for lowest latency. Always safe to enable.",
DEFAULT_LOW_LATENCY, G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
g_object_class_install_property (gobject_class,
PROP_AUDIOCLIENT3,
g_param_spec_boolean ("use-audioclient3", "Use the AudioClient3 API",
"Use the Windows 10 AudioClient3 API when available and if the "
"low-latency property is set to TRUE",
DEFAULT_AUDIOCLIENT3, G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
gst_element_class_add_static_pad_template (gstelement_class, &sink_template);
gst_element_class_set_static_metadata (gstelement_class, "WasapiSrc",
"Sink/Audio",
"Stream audio to an audio capture device through WASAPI",
"Nirbheek Chauhan <nirbheek@centricular.com>, "
"Ole André Vadla Ravnås <ole.andre.ravnas@tandberg.com>");
gstbasesink_class->get_caps = GST_DEBUG_FUNCPTR (gst_wasapi_sink_get_caps);
gstaudiosink_class->prepare = GST_DEBUG_FUNCPTR (gst_wasapi_sink_prepare);
gstaudiosink_class->unprepare = GST_DEBUG_FUNCPTR (gst_wasapi_sink_unprepare);
gstaudiosink_class->open = GST_DEBUG_FUNCPTR (gst_wasapi_sink_open);
gstaudiosink_class->close = GST_DEBUG_FUNCPTR (gst_wasapi_sink_close);
gstaudiosink_class->write = GST_DEBUG_FUNCPTR (gst_wasapi_sink_write);
gstaudiosink_class->delay = GST_DEBUG_FUNCPTR (gst_wasapi_sink_delay);
gstaudiosink_class->reset = GST_DEBUG_FUNCPTR (gst_wasapi_sink_reset);
GST_DEBUG_CATEGORY_INIT (gst_wasapi_sink_debug, "wasapisink",
0, "Windows audio session API sink");
}
static void
gst_wasapi_sink_init (GstWasapiSink * self)
{
self->role = DEFAULT_ROLE;
self->mute = DEFAULT_MUTE;
self->sharemode = AUDCLNT_SHAREMODE_SHARED;
self->low_latency = DEFAULT_LOW_LATENCY;
self->try_audioclient3 = DEFAULT_AUDIOCLIENT3;
self->event_handle = CreateEvent (NULL, FALSE, FALSE, NULL);
self->client_needs_restart = FALSE;
CoInitialize (NULL);
}
static void
gst_wasapi_sink_dispose (GObject * object)
{
GstWasapiSink *self = GST_WASAPI_SINK (object);
if (self->event_handle != NULL) {
CloseHandle (self->event_handle);
self->event_handle = NULL;
}
if (self->client != NULL) {
IUnknown_Release (self->client);
self->client = NULL;
}
if (self->render_client != NULL) {
IUnknown_Release (self->render_client);
self->render_client = NULL;
}
G_OBJECT_CLASS (gst_wasapi_sink_parent_class)->dispose (object);
}
static void
gst_wasapi_sink_finalize (GObject * object)
{
GstWasapiSink *self = GST_WASAPI_SINK (object);
g_clear_pointer (&self->mix_format, CoTaskMemFree);
if (self->cached_caps != NULL) {
gst_caps_unref (self->cached_caps);
self->cached_caps = NULL;
}
g_clear_pointer (&self->positions, g_free);
g_clear_pointer (&self->device_strid, g_free);
self->mute = FALSE;
CoUninitialize ();
G_OBJECT_CLASS (gst_wasapi_sink_parent_class)->finalize (object);
}
static void
gst_wasapi_sink_set_property (GObject * object, guint prop_id,
const GValue * value, GParamSpec * pspec)
{
GstWasapiSink *self = GST_WASAPI_SINK (object);
switch (prop_id) {
case PROP_ROLE:
self->role = gst_wasapi_device_role_to_erole (g_value_get_enum (value));
break;
case PROP_MUTE:
self->mute = g_value_get_boolean (value);
break;
case PROP_DEVICE:
{
const gchar *device = g_value_get_string (value);
g_free (self->device_strid);
self->device_strid =
device ? g_utf8_to_utf16 (device, -1, NULL, NULL, NULL) : NULL;
break;
}
case PROP_EXCLUSIVE:
self->sharemode = g_value_get_boolean (value)
? AUDCLNT_SHAREMODE_EXCLUSIVE : AUDCLNT_SHAREMODE_SHARED;
break;
case PROP_LOW_LATENCY:
self->low_latency = g_value_get_boolean (value);
break;
case PROP_AUDIOCLIENT3:
self->try_audioclient3 = g_value_get_boolean (value);
break;
default:
G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
break;
}
}
static void
gst_wasapi_sink_get_property (GObject * object, guint prop_id,
GValue * value, GParamSpec * pspec)
{
GstWasapiSink *self = GST_WASAPI_SINK (object);
switch (prop_id) {
case PROP_ROLE:
g_value_set_enum (value, gst_wasapi_erole_to_device_role (self->role));
break;
case PROP_MUTE:
g_value_set_boolean (value, self->mute);
break;
case PROP_DEVICE:
g_value_take_string (value, self->device_strid ?
g_utf16_to_utf8 (self->device_strid, -1, NULL, NULL, NULL) : NULL);
break;
case PROP_EXCLUSIVE:
g_value_set_boolean (value,
self->sharemode == AUDCLNT_SHAREMODE_EXCLUSIVE);
break;
case PROP_LOW_LATENCY:
g_value_set_boolean (value, self->low_latency);
break;
case PROP_AUDIOCLIENT3:
g_value_set_boolean (value, self->try_audioclient3);
break;
default:
G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
break;
}
}
static gboolean
gst_wasapi_sink_can_audioclient3 (GstWasapiSink * self)
{
/* AudioClient3 API only makes sense in shared mode */
if (self->sharemode != AUDCLNT_SHAREMODE_SHARED)
return FALSE;
if (!self->try_audioclient3) {
GST_INFO_OBJECT (self, "AudioClient3 disabled by user");
return FALSE;
}
if (!gst_wasapi_util_have_audioclient3 ()) {
GST_INFO_OBJECT (self, "AudioClient3 not available on this OS");
return FALSE;
}
/* Only use audioclient3 when low-latency is requested because otherwise
* very slow machines and VMs with 1 CPU allocated will get glitches:
* https://bugzilla.gnome.org/show_bug.cgi?id=794497 */
if (!self->low_latency) {
GST_INFO_OBJECT (self, "AudioClient3 disabled because low-latency mode "
"was not requested");
return FALSE;
}
return TRUE;
}
static GstCaps *
gst_wasapi_sink_get_caps (GstBaseSink * bsink, GstCaps * filter)
{
GstWasapiSink *self = GST_WASAPI_SINK (bsink);
WAVEFORMATEX *format = NULL;
GstCaps *caps = NULL;
GST_DEBUG_OBJECT (self, "entering get caps");
if (self->cached_caps) {
caps = gst_caps_ref (self->cached_caps);
} else {
GstCaps *template_caps;
gboolean ret;
template_caps = gst_pad_get_pad_template_caps (bsink->sinkpad);
if (!self->client) {
caps = template_caps;
goto out;
}
ret = gst_wasapi_util_get_device_format (GST_ELEMENT (self),
self->sharemode, self->device, self->client, &format);
if (!ret) {
GST_ELEMENT_ERROR (self, STREAM, FORMAT, (NULL),
("failed to detect format"));
gst_caps_unref (template_caps);
return NULL;
}
gst_wasapi_util_parse_waveformatex ((WAVEFORMATEXTENSIBLE *) format,
template_caps, &caps, &self->positions);
if (caps == NULL) {
GST_ELEMENT_ERROR (self, STREAM, FORMAT, (NULL), ("unknown format"));
gst_caps_unref (template_caps);
return NULL;
}
{
gchar *pos_str = gst_audio_channel_positions_to_string (self->positions,
format->nChannels);
GST_INFO_OBJECT (self, "positions are: %s", pos_str);
g_free (pos_str);
}
self->mix_format = format;
gst_caps_replace (&self->cached_caps, caps);
gst_caps_unref (template_caps);
}
if (filter) {
GstCaps *filtered =
gst_caps_intersect_full (filter, caps, GST_CAPS_INTERSECT_FIRST);
gst_caps_unref (caps);
caps = filtered;
}
out:
GST_DEBUG_OBJECT (self, "returning caps %" GST_PTR_FORMAT, caps);
return caps;
}
static gboolean
gst_wasapi_sink_open (GstAudioSink * asink)
{
GstWasapiSink *self = GST_WASAPI_SINK (asink);
gboolean res = FALSE;
IMMDevice *device = NULL;
IAudioClient *client = NULL;
GST_DEBUG_OBJECT (self, "opening device");
if (self->client)
return TRUE;
/* FIXME: Switching the default device does not switch the stream to it,
* even if the old device was unplugged. We need to handle this somehow.
* For example, perhaps we should automatically switch to the new device if
* the default device is changed and a device isn't explicitly selected. */
if (!gst_wasapi_util_get_device_client (GST_ELEMENT (self), FALSE,
self->role, self->device_strid, &device, &client)) {
if (!self->device_strid)
GST_ELEMENT_ERROR (self, RESOURCE, OPEN_WRITE, (NULL),
("Failed to get default device"));
else
GST_ELEMENT_ERROR (self, RESOURCE, OPEN_WRITE, (NULL),
("Failed to open device %S", self->device_strid));
goto beach;
}
self->client = client;
self->device = device;
res = TRUE;
beach:
return res;
}
static gboolean
gst_wasapi_sink_close (GstAudioSink * asink)
{
GstWasapiSink *self = GST_WASAPI_SINK (asink);
if (self->device != NULL) {
IUnknown_Release (self->device);
self->device = NULL;
}
if (self->client != NULL) {
IUnknown_Release (self->client);
self->client = NULL;
}
return TRUE;
}
/* Get the empty space in the buffer that we have to write to */
static gint
gst_wasapi_sink_get_can_frames (GstWasapiSink * self)
{
HRESULT hr;
guint n_frames_padding;
/* There is no padding in exclusive mode since there is no ringbuffer */
if (self->sharemode == AUDCLNT_SHAREMODE_EXCLUSIVE) {
GST_DEBUG_OBJECT (self, "exclusive mode, can write: %i",
self->buffer_frame_count);
return self->buffer_frame_count;
}
/* Frames the card hasn't rendered yet */
hr = IAudioClient_GetCurrentPadding (self->client, &n_frames_padding);
HR_FAILED_RET (hr, IAudioClient::GetCurrentPadding, -1);
GST_DEBUG_OBJECT (self, "%i unread frames (padding)", n_frames_padding);
/* We can write out these many frames */
return self->buffer_frame_count - n_frames_padding;
}
static gboolean
gst_wasapi_sink_prepare (GstAudioSink * asink, GstAudioRingBufferSpec * spec)
{
GstWasapiSink *self = GST_WASAPI_SINK (asink);
gboolean res = FALSE;
REFERENCE_TIME latency_rt;
guint bpf, rate, devicep_frames;
HRESULT hr;
CoInitialize (NULL);
if (gst_wasapi_sink_can_audioclient3 (self)) {
if (!gst_wasapi_util_initialize_audioclient3 (GST_ELEMENT (self), spec,
(IAudioClient3 *) self->client, self->mix_format, self->low_latency,
&devicep_frames))
goto beach;
} else {
if (!gst_wasapi_util_initialize_audioclient (GST_ELEMENT (self), spec,
self->client, self->mix_format, self->sharemode, self->low_latency,
&devicep_frames))
goto beach;
}
bpf = GST_AUDIO_INFO_BPF (&spec->info);
rate = GST_AUDIO_INFO_RATE (&spec->info);
/* Total size of the allocated buffer that we will write to */
hr = IAudioClient_GetBufferSize (self->client, &self->buffer_frame_count);
HR_FAILED_GOTO (hr, IAudioClient::GetBufferSize, beach);
GST_INFO_OBJECT (self, "buffer size is %i frames, device period is %i "
"frames, bpf is %i bytes, rate is %i Hz", self->buffer_frame_count,
devicep_frames, bpf, rate);
/* Actual latency-time/buffer-time will be different now */
spec->segsize = devicep_frames * bpf;
/* We need a minimum of 2 segments to ensure glitch-free playback */
spec->segtotal = MAX (self->buffer_frame_count * bpf / spec->segsize, 2);
GST_INFO_OBJECT (self, "segsize is %i, segtotal is %i", spec->segsize,
spec->segtotal);
/* Get latency for logging */
hr = IAudioClient_GetStreamLatency (self->client, &latency_rt);
HR_FAILED_GOTO (hr, IAudioClient::GetStreamLatency, beach);
GST_INFO_OBJECT (self, "wasapi stream latency: %" G_GINT64_FORMAT " (%"
G_GINT64_FORMAT "ms)", latency_rt, latency_rt / 10000);
/* Set the event handler which will trigger writes */
hr = IAudioClient_SetEventHandle (self->client, self->event_handle);
HR_FAILED_GOTO (hr, IAudioClient::SetEventHandle, beach);
/* Get render sink client and start it up */
if (!gst_wasapi_util_get_render_client (GST_ELEMENT (self), self->client,
&self->render_client)) {
goto beach;
}
GST_INFO_OBJECT (self, "got render client");
/* To avoid start-up glitches, before starting the streaming, we fill the
* buffer with silence as recommended by the documentation:
* https://msdn.microsoft.com/en-us/library/windows/desktop/dd370879%28v=vs.85%29.aspx */
{
gint n_frames, len;
gint16 *dst = NULL;
n_frames = gst_wasapi_sink_get_can_frames (self);
if (n_frames < 1) {
GST_ELEMENT_ERROR (self, RESOURCE, WRITE, (NULL),
("should have more than %i frames to write", n_frames));
goto beach;
}
len = n_frames * self->mix_format->nBlockAlign;
hr = IAudioRenderClient_GetBuffer (self->render_client, n_frames,
(BYTE **) & dst);
HR_FAILED_GOTO (hr, IAudioRenderClient::GetBuffer, beach);
GST_DEBUG_OBJECT (self, "pre-wrote %i bytes of silence", len);
hr = IAudioRenderClient_ReleaseBuffer (self->render_client, n_frames,
AUDCLNT_BUFFERFLAGS_SILENT);
HR_FAILED_GOTO (hr, IAudioRenderClient::ReleaseBuffer, beach);
}
hr = IAudioClient_Start (self->client);
HR_FAILED_GOTO (hr, IAudioClient::Start, beach);
gst_audio_ring_buffer_set_channel_positions (GST_AUDIO_BASE_SINK
(self)->ringbuffer, self->positions);
/* Increase the thread priority to reduce glitches */
self->thread_priority_handle = gst_wasapi_util_set_thread_characteristics ();
res = TRUE;
beach:
/* unprepare() is not called if prepare() fails, but we want it to be, so call
* it manually when needed */
if (!res)
gst_wasapi_sink_unprepare (asink);
return res;
}
static gboolean
gst_wasapi_sink_unprepare (GstAudioSink * asink)
{
GstWasapiSink *self = GST_WASAPI_SINK (asink);
CoUninitialize ();
if (self->thread_priority_handle != NULL) {
gst_wasapi_util_revert_thread_characteristics
(self->thread_priority_handle);
self->thread_priority_handle = NULL;
}
if (self->client != NULL) {
IAudioClient_Stop (self->client);
}
if (self->render_client != NULL) {
IUnknown_Release (self->render_client);
self->render_client = NULL;
}
return TRUE;
}
static gint
gst_wasapi_sink_write (GstAudioSink * asink, gpointer data, guint length)
{
GstWasapiSink *self = GST_WASAPI_SINK (asink);
HRESULT hr;
gint16 *dst = NULL;
guint pending = length;
GST_OBJECT_LOCK (self);
if (self->client_needs_restart) {
hr = IAudioClient_Start (self->client);
HR_FAILED_AND (hr, IAudioClient::Start, length = 0; goto beach);
self->client_needs_restart = FALSE;
}
GST_OBJECT_UNLOCK (self);
while (pending > 0) {
DWORD dwWaitResult;
guint can_frames, have_frames, n_frames, write_len;
dwWaitResult = WaitForSingleObject (self->event_handle, INFINITE);
if (dwWaitResult != WAIT_OBJECT_0) {
GST_ERROR_OBJECT (self, "Error waiting for event handle: %x",
(guint) dwWaitResult);
length -= pending;
goto beach;
}
/* We have N frames to be written out */
have_frames = pending / (self->mix_format->nBlockAlign);
/* We have can_frames space in the output buffer */
can_frames = gst_wasapi_sink_get_can_frames (self);
/* We will write out these many frames, and this much length */
n_frames = MIN (can_frames, have_frames);
write_len = n_frames * self->mix_format->nBlockAlign;
GST_DEBUG_OBJECT (self, "total: %i, have_frames: %i (%i bytes), "
"can_frames: %i, will write: %i (%i bytes)", self->buffer_frame_count,
have_frames, pending, can_frames, n_frames, write_len);
hr = IAudioRenderClient_GetBuffer (self->render_client, n_frames,
(BYTE **) & dst);
HR_FAILED_AND (hr, IAudioRenderClient::GetBuffer, length = 0; goto beach);
memcpy (dst, data, write_len);
hr = IAudioRenderClient_ReleaseBuffer (self->render_client, n_frames,
self->mute ? AUDCLNT_BUFFERFLAGS_SILENT : 0);
HR_FAILED_AND (hr, IAudioRenderClient::ReleaseBuffer, length = 0;
goto beach);
pending -= write_len;
}
beach:
return length;
}
static guint
gst_wasapi_sink_delay (GstAudioSink * asink)
{
GstWasapiSink *self = GST_WASAPI_SINK (asink);
guint delay = 0;
HRESULT hr;
hr = IAudioClient_GetCurrentPadding (self->client, &delay);
HR_FAILED_RET (hr, IAudioClient::GetCurrentPadding, 0);
return delay;
}
static void
gst_wasapi_sink_reset (GstAudioSink * asink)
{
GstWasapiSink *self = GST_WASAPI_SINK (asink);
HRESULT hr;
GST_INFO_OBJECT (self, "reset called");
if (!self->client)
return;
GST_OBJECT_LOCK (self);
hr = IAudioClient_Stop (self->client);
HR_FAILED_AND (hr, IAudioClient::Stop,);
hr = IAudioClient_Reset (self->client);
HR_FAILED_AND (hr, IAudioClient::Reset,);
self->client_needs_restart = TRUE;
GST_OBJECT_UNLOCK (self);
}