CoreAudio: add audio device module

This patch adds support for CoreAudio driven devices under Mac OS X. It
is typically instanciated by the CoreAudio device detection module and
handles all available streams on a specific device.

Sinks are created according to the reported stream configuration.
Float32 is used as default audio sample format at it is the only format
CoreAudio speaks natively.

Hardware volume control is not implemented yet.

2 files changed, 828 insertions, 1 deletions

diff --git a/src/Makefile.am b/src/Makefile.am
index 70ab5b07..fa5d1702 100644
--- a/src/Makefile.am
+++ b/src/Makefile.am
@@ -1068,7 +1068,8 @@ endif
 
 if HAVE_COREAUDIO
 modlibexec_LTLIBRARIES += \
-		module-coreaudio-detect.la
+		module-coreaudio-detect.la \
+		module-coreaudio-device.la
 endif
 
 pulselibexec_PROGRAMS =
@@ -1244,6 +1245,7 @@ SYMDEF_FILES = \
 		modules/alsa/module-alsa-source-symdef.h \
 		modules/alsa/module-alsa-card-symdef.h \
 		modules/coreaudio/module-coreaudio-detect-symdef.h \
+		modules/coreaudio/module-coreaudio-device-symdef.h \
 		modules/module-solaris-symdef.h \
 		modules/module-waveout-symdef.h \
 		modules/module-detect-symdef.h \
@@ -1483,6 +1485,11 @@ module_coreaudio_detect_la_LDFLAGS = $(MODULE_LDFLAGS) \
 			-Wl,-framework -Wl,AudioUnit -framework AudioUnit
 module_coreaudio_detect_la_LIBADD = $(AM_LIBADD) libpulsecore-@PA_MAJORMINORMICRO@.la libpulsecommon-@PA_MAJORMINORMICRO@.la libpulse.la
 
+module_coreaudio_device_la_SOURCES = modules/coreaudio/module-coreaudio-device.c
+module_coreaudio_device_la_LDFLAGS = $(MODULE_LDFLAGS) \
+			-Wl,-framework -Wl,Cocoa -framework CoreAudio \
+			-Wl,-framework -Wl,AudioUnit -framework AudioUnit
+module_coreaudio_device_la_LIBADD = $(AM_LIBADD) libpulsecore-@PA_MAJORMINORMICRO@.la libpulsecommon-@PA_MAJORMINORMICRO@.la libpulse.la
 # ALSA
 
 libalsa_util_la_SOURCES = modules/alsa/alsa-util.c modules/alsa/alsa-util.h modules/alsa/alsa-mixer.c modules/alsa/alsa-mixer.h modules/alsa/alsa-sink.c modules/alsa/alsa-sink.h modules/alsa/alsa-source.c modules/alsa/alsa-source.h modules/reserve-wrap.c modules/reserve-wrap.h
diff --git a/src/modules/coreaudio/module-coreaudio-device.c b/src/modules/coreaudio/module-coreaudio-device.c
new file mode 100644
index 00000000..a3699e68
--- /dev/null
+++ b/src/modules/coreaudio/module-coreaudio-device.c
@@ -0,0 +1,820 @@
+/***
+  This file is part of PulseAudio.
+
+  Copyright 2009 Daniel Mack <daniel@caiaq.de>
+
+  PulseAudio is free software; you can redistribute it and/or modify
+  it under the terms of the GNU Lesser General Public License as published
+  by the Free Software Foundation; either version 2.1 of the License,
+  or (at your option) any later version.
+
+  PulseAudio 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
+  General Public License for more details.
+
+  You should have received a copy of the GNU Lesser General Public License
+  along with PulseAudio; if not, write to the Free Software
+  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307
+  USA.
+***/
+
+/* TODO:
+    - implement hardware volume controls
+    - handle audio device stream format changes (will require changes to the core)
+*/
+
+#ifdef HAVE_CONFIG_H
+#include <config.h>
+#endif
+
+#include <pulse/xmalloc.h>
+#include <pulse/util.h>
+
+#include <pulsecore/core-error.h>
+#include <pulsecore/sink.h>
+#include <pulsecore/source.h>
+#include <pulsecore/module.h>
+#include <pulsecore/sample-util.h>
+#include <pulsecore/core-util.h>
+#include <pulsecore/modargs.h>
+#include <pulsecore/log.h>
+#include <pulsecore/macro.h>
+#include <pulsecore/llist.h>
+#include <pulsecore/card.h>
+#include <pulsecore/strbuf.h>
+#include <pulsecore/thread.h>
+#include <pulsecore/thread-mq.h>
+
+#include <CoreAudio/CoreAudio.h>
+#include <CoreAudio/CoreAudioTypes.h>
+#include <CoreAudio/AudioHardware.h>
+
+#include "module-coreaudio-device-symdef.h"
+
+#define DEFAULT_FRAMES_PER_IOPROC 512
+
+PA_MODULE_AUTHOR("Daniel Mack");
+PA_MODULE_DESCRIPTION("CoreAudio device");
+PA_MODULE_VERSION(PACKAGE_VERSION);
+PA_MODULE_LOAD_ONCE(FALSE);
+PA_MODULE_USAGE("device_id=<the CoreAudio device id> "
+                "ioproc_frames=<audio frames per IOProc call> ");
+
+static const char* const valid_modargs[] = {
+    "device_id",
+    "ioproc_frames",
+    NULL
+};
+
+enum {
+    CA_MESSAGE_RENDER = PA_SINK_MESSAGE_MAX,
+};
+
+typedef struct coreaudio_sink coreaudio_sink;
+typedef struct coreaudio_source coreaudio_source;
+
+struct userdata {
+    AudioDeviceID device_id;
+    AudioDeviceIOProcID proc_id;
+
+    pa_thread_mq thread_mq;
+    pa_asyncmsgq *async_msgq;
+
+    pa_rtpoll *rtpoll;
+    pa_thread *thread;
+
+    pa_module *module;
+    pa_card *card;
+    pa_bool_t running;
+
+    char *device_name, *vendor_name;
+
+    const AudioBufferList *render_input_data;
+    AudioBufferList       *render_output_data;
+
+    AudioStreamBasicDescription stream_description;
+
+    PA_LLIST_HEAD(coreaudio_sink, sinks);
+    PA_LLIST_HEAD(coreaudio_source, sources);
+};
+
+struct coreaudio_sink {
+    pa_sink *pa_sink;
+    struct userdata *userdata;
+
+    char *name;
+    unsigned int channel_idx;
+    pa_bool_t active;
+
+    pa_channel_map map;
+    pa_sample_spec ss;
+
+    PA_LLIST_FIELDS(coreaudio_sink);
+};
+
+struct coreaudio_source {
+    pa_source *pa_source;
+    struct userdata *userdata;
+
+    char *name;
+    unsigned int channel_idx;
+    pa_bool_t active;
+
+    pa_channel_map map;
+    pa_sample_spec ss;
+
+    PA_LLIST_FIELDS(coreaudio_source);
+};
+
+static OSStatus io_render_proc (AudioDeviceID          device,
+                                const AudioTimeStamp  *now,
+                                const AudioBufferList *inputData,
+                                const AudioTimeStamp  *inputTime,
+                                AudioBufferList       *outputData,
+                                const AudioTimeStamp  *outputTime,
+                                void                  *clientData)
+{
+    struct userdata *u = clientData;
+
+    pa_assert(u);
+    pa_assert(device == u->device_id);
+
+    u->render_input_data = inputData;
+    u->render_output_data = outputData;
+
+    if (u->sinks)
+        pa_assert_se(pa_asyncmsgq_send(u->async_msgq, PA_MSGOBJECT(u->sinks->pa_sink),
+                                        CA_MESSAGE_RENDER, NULL, 0, NULL) == 0);
+
+    if (u->sources)
+        pa_assert_se(pa_asyncmsgq_send(u->async_msgq, PA_MSGOBJECT(u->sources->pa_source),
+                                        CA_MESSAGE_RENDER, NULL, 0, NULL) == 0);
+
+    return 0;
+}
+
+static OSStatus ca_stream_format_changed(AudioDeviceID inDevice,
+                                         UInt32 inChannel,
+                                         Boolean isInput,
+                                         AudioDevicePropertyID inPropertyID,
+                                         void *inClientData)
+{
+    struct userdata *u = inClientData;
+
+    pa_assert(u);
+
+    /* REVISIT: PA can't currently handle external format change requests.
+     * Hence, we set the original format back in this callback to avoid horrible audio artefacts.
+     * The device settings will appear to be 'locked' for any application as long as the PA daemon is running.
+     * Once we're able to propagate such events up in the core, this needs to be changed. */
+
+    return AudioDeviceSetProperty(inDevice, NULL, inChannel, isInput,
+                                  kAudioDevicePropertyStreamFormat, sizeof(u->stream_description), &u->stream_description);
+}
+
+static pa_usec_t get_latency_us(pa_object *o) {
+    struct userdata *u;
+    pa_sample_spec *ss;
+    bool is_source;
+    UInt32 v, total = 0;
+    UInt32 err, size = sizeof(v);
+
+    if (pa_sink_isinstance(o)) {
+        coreaudio_sink *sink = PA_SINK(o)->userdata;
+
+        u = sink->userdata;
+        ss = &sink->ss;
+        is_source = FALSE;
+    } else if (pa_source_isinstance(o)) {
+        coreaudio_source *source = PA_SOURCE(o)->userdata;
+
+        u = source->userdata;
+        ss = &source->ss;
+        is_source = TRUE;
+    } else
+        pa_assert_not_reached();
+
+    pa_assert(u);
+
+    /* get the device latency */
+    size = sizeof(total);
+    AudioDeviceGetProperty(u->device_id, 0, is_source, kAudioDevicePropertyLatency, &size, &v);
+    total += v;
+
+    /* the the IOProc buffer size */
+    size = sizeof(v);
+    AudioDeviceGetProperty(u->device_id, 0, is_source, kAudioDevicePropertyBufferFrameSize, &size, &v);
+    total += v;
+
+    /* IOProc safety offset - this value is the same for both directions, hence we divide it by 2 */
+    size = sizeof(v);
+    AudioDeviceGetProperty(u->device_id, 0, is_source, kAudioDevicePropertySafetyOffset, &size, &v);
+    total += v / 2;
+
+    /* get the stream latency.
+     * FIXME: this assumes the stream latency is the same for all streams */
+    err = AudioStreamGetProperty(0, is_source, kAudioStreamPropertyLatency, &size, &v);
+    if (!err)
+        total += v;
+
+    return pa_bytes_to_usec(total * pa_frame_size(ss), ss);
+}
+
+static void ca_device_check_device_state(struct userdata *u) {
+    coreaudio_sink *ca_sink;
+    coreaudio_source *ca_source;
+    pa_bool_t active = FALSE;
+
+    pa_assert(u);
+
+    for (ca_sink = u->sinks; ca_sink; ca_sink = ca_sink->next)
+        if (ca_sink->active)
+            active = TRUE;
+
+    for (ca_source = u->sources; ca_source; ca_source = ca_source->next)
+        if (ca_source->active)
+            active = TRUE;
+
+    if (active && !u->running)
+        AudioDeviceStart(u->device_id, u->proc_id);
+    else if (!active && u->running)
+        AudioDeviceStop(u->device_id, u->proc_id);
+
+    u->running = active;
+}
+
+static int sink_process_msg(pa_msgobject *o, int code, void *data, int64_t offset, pa_memchunk *chunk) {
+    coreaudio_sink *sink = PA_SINK(o)->userdata;
+    struct userdata *u = sink->userdata;
+    unsigned int i;
+    pa_memchunk audio_chunk;
+
+    switch (code) {
+        case CA_MESSAGE_RENDER: {
+            /* audio out */
+            for (i = 0; i < u->render_output_data->mNumberBuffers; i++) {
+                AudioBuffer *buf = u->render_output_data->mBuffers + i;
+
+                pa_assert(sink);
+
+                if (PA_SINK_IS_OPENED(sink->pa_sink->thread_info.state)) {
+                    if (sink->pa_sink->thread_info.rewind_requested)
+                        pa_sink_process_rewind(sink->pa_sink, 0);
+
+                    audio_chunk.memblock = pa_memblock_new_fixed(u->module->core->mempool, buf->mData, buf->mDataByteSize, FALSE);
+                    audio_chunk.length = buf->mDataByteSize;
+                    audio_chunk.index = 0;
+
+                    pa_sink_render_into_full(sink->pa_sink, &audio_chunk);
+                    pa_memblock_unref_fixed(audio_chunk.memblock);
+                }
+
+                sink = sink->next;
+            }
+
+            return 0;
+        }
+
+        case PA_SINK_MESSAGE_GET_LATENCY: {
+            *((pa_usec_t *) data) = get_latency_us(PA_OBJECT(o));
+            return 0;
+        }
+
+        case PA_SINK_MESSAGE_SET_STATE:
+            switch ((pa_sink_state_t) PA_PTR_TO_UINT(data)) {
+                case PA_SINK_SUSPENDED:
+                case PA_SINK_IDLE:
+                    sink->active = FALSE;
+                    break;
+
+                case PA_SINK_RUNNING:
+                    sink->active = TRUE;
+                    break;
+
+                case PA_SINK_UNLINKED:
+                case PA_SINK_INIT:
+                case PA_SINK_INVALID_STATE:
+                    ;
+            }
+
+            ca_device_check_device_state(sink->userdata);
+            break;
+    }
+
+    return pa_sink_process_msg(o, code, data, offset, chunk);
+}
+
+static int source_process_msg(pa_msgobject *o, int code, void *data, int64_t offset, pa_memchunk *chunk) {
+    coreaudio_source *source = PA_SOURCE(o)->userdata;
+    struct userdata *u = source->userdata;
+    unsigned int i;
+    pa_memchunk audio_chunk;
+
+    switch (code) {
+        case CA_MESSAGE_RENDER: {
+            /* audio in */
+            for (i = 0; i < u->render_input_data->mNumberBuffers; i++) {
+                const AudioBuffer *buf = u->render_input_data->mBuffers + i;
+
+                pa_assert(source);
+
+                if (PA_SOURCE_IS_OPENED(source->pa_source->thread_info.state)) {
+                    audio_chunk.memblock = pa_memblock_new_fixed(u->module->core->mempool, buf->mData, buf->mDataByteSize, TRUE);
+                    audio_chunk.length = buf->mDataByteSize;
+                    audio_chunk.index = 0;
+
+                    pa_source_post(source->pa_source, &audio_chunk);
+                    pa_memblock_unref_fixed(audio_chunk.memblock);
+                }
+
+                source = source->next;
+            }
+
+            return 0;
+        }
+
+        case PA_SOURCE_MESSAGE_GET_LATENCY: {
+            *((pa_usec_t *) data) = get_latency_us(PA_OBJECT(o));
+            return 0;
+        }
+
+        case PA_SOURCE_MESSAGE_SET_STATE:
+            switch ((pa_source_state_t) PA_PTR_TO_UINT(data)) {
+                case PA_SOURCE_SUSPENDED:
+                case PA_SOURCE_IDLE:
+                    source->active = FALSE;
+                    break;
+
+                case PA_SOURCE_RUNNING:
+                    source->active = TRUE;
+                    break;
+
+                case PA_SOURCE_UNLINKED:
+                case PA_SOURCE_INIT:
+                case PA_SOURCE_INVALID_STATE:
+                    ;
+            }
+
+            ca_device_check_device_state(source->userdata);
+            break;
+    }
+
+    return pa_source_process_msg(o, code, data, offset, chunk);
+}
+
+static int ca_device_create_sink(pa_module *m, AudioBuffer *buf, int channel_idx) {
+    OSStatus err;
+    UInt32 size;
+    struct userdata *u = m->userdata;
+    pa_sink_new_data new_data;
+    pa_sink_flags_t flags = PA_SINK_LATENCY | PA_SINK_HARDWARE;
+    coreaudio_sink *ca_sink;
+    pa_sink *sink;
+    unsigned int i;
+    char tmp[255];
+    pa_strbuf *strbuf;
+
+    ca_sink = pa_xnew0(coreaudio_sink, 1);
+    ca_sink->map.channels = buf->mNumberChannels;
+    ca_sink->ss.channels = buf->mNumberChannels;
+    ca_sink->channel_idx = channel_idx;
+
+    /* build a name for this stream */
+    strbuf = pa_strbuf_new();
+
+    for (i = 0; i < buf->mNumberChannels; i++) {
+        size = sizeof(tmp);
+        err = AudioDeviceGetProperty(u->device_id, channel_idx + i + 1, 0, kAudioObjectPropertyElementName, &size, tmp);
+        if (err || !strlen(tmp))
+            snprintf(tmp, sizeof(tmp), "Channel %d", channel_idx + i + 1);
+
+        if (i > 0)
+            pa_strbuf_puts(strbuf, ", ");
+
+        pa_strbuf_puts(strbuf, tmp);
+    }
+
+    ca_sink->name = pa_strbuf_tostring_free(strbuf);
+
+    pa_log_debug("Stream name is >%s<", ca_sink->name);
+
+    /* default to mono streams */
+    for (i = 0; i < ca_sink->map.channels; i++)
+        ca_sink->map.map[i] = PA_CHANNEL_POSITION_MONO;
+
+    if (buf->mNumberChannels == 2) {
+        ca_sink->map.map[0] = PA_CHANNEL_POSITION_LEFT;
+        ca_sink->map.map[1] = PA_CHANNEL_POSITION_RIGHT;
+    }
+
+    ca_sink->ss.rate = u->stream_description.mSampleRate;
+    ca_sink->ss.format = PA_SAMPLE_FLOAT32LE;
+
+    pa_sink_new_data_init(&new_data);
+    new_data.card = u->card;
+    new_data.driver = __FILE__;
+    new_data.module = u->module;
+    new_data.namereg_fail = FALSE;
+    pa_sink_new_data_set_name(&new_data, ca_sink->name);
+    pa_sink_new_data_set_channel_map(&new_data, &ca_sink->map);
+    pa_sink_new_data_set_sample_spec(&new_data, &ca_sink->ss);
+    pa_proplist_sets(new_data.proplist, PA_PROP_DEVICE_STRING, u->device_name);
+    pa_proplist_sets(new_data.proplist, PA_PROP_DEVICE_PRODUCT_NAME, u->device_name);
+    pa_proplist_sets(new_data.proplist, PA_PROP_DEVICE_DESCRIPTION, u->device_name);
+    pa_proplist_sets(new_data.proplist, PA_PROP_DEVICE_ACCESS_MODE, "mmap");
+    pa_proplist_sets(new_data.proplist, PA_PROP_DEVICE_CLASS, "sound");
+    pa_proplist_sets(new_data.proplist, PA_PROP_DEVICE_API, "CoreAudio");
+    pa_proplist_setf(new_data.proplist, PA_PROP_DEVICE_BUFFERING_BUFFER_SIZE, "%lu", (unsigned long) buf->mDataByteSize);
+
+    if (u->vendor_name)
+        pa_proplist_sets(new_data.proplist, PA_PROP_DEVICE_VENDOR_NAME, u->vendor_name);
+
+    sink = pa_sink_new(m->core, &new_data, flags);
+    pa_sink_new_data_done(&new_data);
+
+    if (!sink) {
+        pa_log("unable to create sink.");
+        return -1;
+    }
+
+    sink->parent.process_msg = sink_process_msg;
+    sink->userdata = ca_sink;
+
+    pa_sink_set_asyncmsgq(sink, u->thread_mq.inq);
+    pa_sink_set_rtpoll(sink, u->rtpoll);
+
+    ca_sink->pa_sink = sink;
+    ca_sink->userdata = u;
+
+    PA_LLIST_PREPEND(coreaudio_sink, u->sinks, ca_sink);
+
+    return 0;
+}
+
+static int ca_device_create_source(pa_module *m, AudioBuffer *buf, int channel_idx) {
+    OSStatus err;
+    UInt32 size;
+    struct userdata *u = m->userdata;
+    pa_source_new_data new_data;
+    pa_source_flags_t flags = PA_SOURCE_LATENCY | PA_SOURCE_HARDWARE;
+    coreaudio_source *ca_source;
+    pa_source *source;
+    unsigned int i;
+    char tmp[255];
+    pa_strbuf *strbuf;
+
+    ca_source = pa_xnew0(coreaudio_source, 1);
+    ca_source->map.channels = buf->mNumberChannels;
+    ca_source->ss.channels = buf->mNumberChannels;
+    ca_source->channel_idx = channel_idx;
+
+    /* build a name for this stream */
+    strbuf = pa_strbuf_new();
+
+    for (i = 0; i < buf->mNumberChannels; i++) {
+        size = sizeof(tmp);
+        err = AudioDeviceGetProperty(u->device_id, channel_idx + i + 1, 0, kAudioObjectPropertyElementName, &size, tmp);
+        if (err || !strlen(tmp))
+            snprintf(tmp, sizeof(tmp), "Channel %d", channel_idx + i + 1);
+
+        if (i > 0)
+            pa_strbuf_puts(strbuf, ", ");
+
+        pa_strbuf_puts(strbuf, tmp);
+    }
+
+    ca_source->name = pa_strbuf_tostring_free(strbuf);
+
+    pa_log_debug("Stream name is >%s<", ca_source->name);
+
+    /* default to mono streams */
+    for (i = 0; i < ca_source->map.channels; i++)
+        ca_source->map.map[i] = PA_CHANNEL_POSITION_MONO;
+
+    if (buf->mNumberChannels == 2) {
+        ca_source->map.map[0] = PA_CHANNEL_POSITION_LEFT;
+        ca_source->map.map[1] = PA_CHANNEL_POSITION_RIGHT;
+    }
+
+    ca_source->ss.rate = u->stream_description.mSampleRate;
+    ca_source->ss.format = PA_SAMPLE_FLOAT32LE;
+
+    pa_source_new_data_init(&new_data);
+    new_data.card = u->card;
+    new_data.driver = __FILE__;
+    new_data.module = u->module;
+    new_data.namereg_fail = FALSE;
+    pa_source_new_data_set_name(&new_data, ca_source->name);
+    pa_source_new_data_set_channel_map(&new_data, &ca_source->map);
+    pa_source_new_data_set_sample_spec(&new_data, &ca_source->ss);
+    pa_proplist_sets(new_data.proplist, PA_PROP_DEVICE_STRING, u->device_name);
+    pa_proplist_sets(new_data.proplist, PA_PROP_DEVICE_PRODUCT_NAME, u->device_name);
+    pa_proplist_sets(new_data.proplist, PA_PROP_DEVICE_DESCRIPTION, u->device_name);
+    pa_proplist_sets(new_data.proplist, PA_PROP_DEVICE_ACCESS_MODE, "mmap");
+    pa_proplist_sets(new_data.proplist, PA_PROP_DEVICE_CLASS, "sound");
+    pa_proplist_sets(new_data.proplist, PA_PROP_DEVICE_API, "CoreAudio");
+    pa_proplist_setf(new_data.proplist, PA_PROP_DEVICE_BUFFERING_BUFFER_SIZE, "%lu", (unsigned long) buf->mDataByteSize);
+
+    if (u->vendor_name)
+        pa_proplist_sets(new_data.proplist, PA_PROP_DEVICE_VENDOR_NAME, u->vendor_name);
+
+    source = pa_source_new(m->core, &new_data, flags);
+    pa_source_new_data_done(&new_data);
+
+    if (!source) {
+        pa_log("unable to create source.");
+        return -1;
+    }
+
+    source->parent.process_msg = source_process_msg;
+    source->userdata = ca_source;
+
+    pa_source_set_asyncmsgq(source, u->thread_mq.inq);
+    pa_source_set_rtpoll(source, u->rtpoll);
+
+    ca_source->pa_source = source;
+    ca_source->userdata = u;
+
+    PA_LLIST_PREPEND(coreaudio_source, u->sources, ca_source);
+
+    return 0;
+}
+
+static int ca_device_create_streams(pa_module *m, bool direction_in) {
+    OSStatus err;
+    UInt32 size, i, channel_idx;
+    struct userdata *u = m->userdata;
+    int section = direction_in ? 1 : 0;
+    AudioBufferList *buffer_list;
+
+    /* get current stream format */
+    size = sizeof(AudioStreamBasicDescription);
+    err = AudioDeviceGetProperty(u->device_id, 0, section, kAudioDevicePropertyStreamFormat, &size, &u->stream_description);
+    if (err) {
+        /* no appropriate streams found - silently bail. */
+        return -1;
+    }
+
+    if (u->stream_description.mFormatID != kAudioFormatLinearPCM) {
+        pa_log("Unsupported audio format '%c%c%c%c'",
+            (char) (u->stream_description.mFormatID >> 24),
+            (char) (u->stream_description.mFormatID >> 16) & 0xff,
+            (char) (u->stream_description.mFormatID >> 8) & 0xff,
+            (char) (u->stream_description.mFormatID & 0xff));
+        return -1;
+    }
+
+    /* get stream configuration */
+    size = 0;
+    err = AudioDeviceGetPropertyInfo(u->device_id, 0, section, kAudioDevicePropertyStreamConfiguration, &size, NULL);
+    if (err) {
+        pa_log("Failed to get kAudioDevicePropertyStreamConfiguration (%s).", direction_in ? "input" : "output");
+        return -1;
+    }
+
+    if (!size)
+        return 0;
+
+    buffer_list = (AudioBufferList *) pa_xmalloc(size);
+    err = AudioDeviceGetProperty(u->device_id, 0, section, kAudioDevicePropertyStreamConfiguration, &size, buffer_list);
+
+    if (!err) {
+        pa_log_debug("Sample rate: %f", u->stream_description.mSampleRate);
+        pa_log_debug("%d bytes per packet",   (unsigned int) u->stream_description.mBytesPerPacket);
+        pa_log_debug("%d frames per packet",  (unsigned int) u->stream_description.mFramesPerPacket);
+        pa_log_debug("%d bytes per frame",    (unsigned int) u->stream_description.mBytesPerFrame);
+        pa_log_debug("%d channels per frame", (unsigned int) u->stream_description.mChannelsPerFrame);
+        pa_log_debug("%d bits per channel",   (unsigned int) u->stream_description.mBitsPerChannel);
+
+        for (channel_idx = 0, i = 0; i < buffer_list->mNumberBuffers; i++) {
+            AudioBuffer *buf = buffer_list->mBuffers + i;
+
+            if (direction_in)
+                ca_device_create_source(m, buf, channel_idx);
+            else
+                ca_device_create_sink(m, buf, channel_idx);
+
+            channel_idx += buf->mNumberChannels;
+        }
+    }
+
+    pa_xfree(buffer_list);
+    return 0;
+}
+
+static void thread_func(void *userdata) {
+    struct userdata *u = userdata;
+
+    pa_assert(u);
+    pa_assert(u->module);
+    pa_assert(u->module->core);
+
+    pa_log_debug("Thread starting up");
+
+    if (u->module->core->realtime_scheduling)
+        pa_make_realtime(u->module->core->realtime_priority);
+
+    pa_thread_mq_install(&u->thread_mq);
+
+    for (;;) {
+        int ret;
+
+        ret = pa_rtpoll_run(u->rtpoll, TRUE);
+
+        if (ret < 0)
+            goto fail;
+
+        if (ret == 0)
+            goto finish;
+    }
+
+fail:
+    /* If this was no regular exit from the loop we have to continue
+     * processing messages until we received PA_MESSAGE_SHUTDOWN */
+    pa_asyncmsgq_post(u->thread_mq.outq, PA_MSGOBJECT(u->module->core), PA_CORE_MESSAGE_UNLOAD_MODULE, u->module, 0, NULL, NULL);
+    pa_asyncmsgq_wait_for(u->thread_mq.inq, PA_MESSAGE_SHUTDOWN);
+
+finish:
+    pa_log_debug("Thread shutting down");
+}
+
+int pa__init(pa_module *m) {
+    OSStatus err;
+    UInt32 size, frames;
+    struct userdata *u = NULL;
+    pa_modargs *ma = NULL;
+    char tmp[64];
+    pa_card_new_data card_new_data;
+    coreaudio_sink *ca_sink;
+    coreaudio_source *ca_source;
+
+    pa_assert(m);
+
+    if (!(ma = pa_modargs_new(m->argument, valid_modargs))) {
+        pa_log("Failed to parse module arguments.");
+        goto fail;
+    }
+
+    u = pa_xnew0(struct userdata, 1);
+    u->module = m;
+    m->userdata = u;
+
+    if (pa_modargs_get_value_u32(ma, "device_id", (unsigned int *) &u->device_id) != 0) {
+        pa_log("Failed to parse device_id argument.");
+        goto fail;
+    }
+
+    /* get device product name */
+    size = sizeof(tmp);
+    err = AudioDeviceGetProperty(u->device_id, 0, 0, kAudioDevicePropertyDeviceName, &size, &tmp);
+    if (err) {
+        pa_log("Failed to get kAudioDevicePropertyDeviceName (err = %08x).", (int) err);
+        goto fail;
+    }
+
+    u->device_name = pa_xstrdup(tmp);
+
+    pa_card_new_data_init(&card_new_data);
+    pa_proplist_sets(card_new_data.proplist, PA_PROP_DEVICE_STRING, tmp);
+    card_new_data.driver = __FILE__;
+    pa_card_new_data_set_name(&card_new_data, tmp);
+    pa_log_info("Initializing module for CoreAudio device '%s' (id %d)", tmp, (unsigned int) u->device_id);
+
+    /* get device vendor name (may fail) */
+    size = sizeof(tmp);
+    err = AudioDeviceGetProperty(u->device_id, 0, 0, kAudioDevicePropertyDeviceManufacturer, &size, &tmp);
+    if (!err)
+        u->vendor_name = pa_xstrdup(tmp);
+
+    /* create the card object */
+    u->card = pa_card_new(m->core, &card_new_data);
+    if (!u->card) {
+        pa_log("Unable to create card.\n");
+        goto fail;
+    }
+
+    pa_card_new_data_done(&card_new_data);
+    u->card->userdata = u;
+
+    u->rtpoll = pa_rtpoll_new();
+    pa_thread_mq_init(&u->thread_mq, m->core->mainloop, u->rtpoll);
+    u->async_msgq = pa_asyncmsgq_new(0);
+    pa_rtpoll_item_new_asyncmsgq_read(u->rtpoll, PA_RTPOLL_EARLY-1, u->async_msgq);
+
+    PA_LLIST_HEAD_INIT(coreaudio_sink, u->sinks);
+
+    /* create sinks */
+    ca_device_create_streams(m, FALSE);
+
+    /* create sources */
+    ca_device_create_streams(m, TRUE);
+
+    /* create the message thread */
+    if (!(u->thread = pa_thread_new(thread_func, u))) {
+        pa_log("Failed to create thread.");
+        goto fail;
+    }
+
+    /* register notification callback for stream format changes */
+    AudioDeviceAddPropertyListener(u->device_id, 0, 0, kAudioDevicePropertyStreamFormat, ca_stream_format_changed, u);
+
+    /* set number of frames in IOProc */
+    frames = DEFAULT_FRAMES_PER_IOPROC;
+    pa_modargs_get_value_u32(ma, "ioproc_frames", (unsigned int *) &frames);
+
+    AudioDeviceSetProperty(u->device_id, NULL, 0, 0, kAudioDevicePropertyBufferFrameSize, sizeof(frames), &frames);
+    pa_log_debug("%u frames per IOProc\n", (unsigned int) frames);
+
+    /* create one ioproc for both directions */
+    err = AudioDeviceCreateIOProcID(u->device_id, io_render_proc, u, &u->proc_id);
+    if (err) {
+        pa_log("AudioDeviceCreateIOProcID() failed (err = %08x\n).", (int) err);
+        goto fail;
+    }
+
+    for (ca_sink = u->sinks; ca_sink; ca_sink = ca_sink->next)
+        pa_sink_put(ca_sink->pa_sink);
+
+    for (ca_source = u->sources; ca_source; ca_source = ca_source->next)
+        pa_source_put(ca_source->pa_source);
+
+    pa_modargs_free(ma);
+
+    return 0;
+
+fail:
+    if (u)
+        pa__done(m);
+
+    if (ma)
+        pa_modargs_free(ma);
+
+    return -1;
+}
+
+void pa__done(pa_module *m) {
+    struct userdata *u;
+    coreaudio_sink *ca_sink;
+    coreaudio_source *ca_source;
+
+    pa_assert(m);
+
+    u = m->userdata;
+    pa_assert(u);
+
+    /* unlink sinks */
+    for (ca_sink = u->sinks; ca_sink; ca_sink = ca_sink->next)
+        if (ca_sink->pa_sink)
+            pa_sink_unlink(ca_sink->pa_sink);
+
+    /* unlink sources */
+    for (ca_source = u->sources; ca_source; ca_source = ca_source->next)
+        if (ca_source->pa_source)
+            pa_source_unlink(ca_source->pa_source);
+
+    if (u->thread) {
+        pa_asyncmsgq_send(u->thread_mq.inq, NULL, PA_MESSAGE_SHUTDOWN, NULL, 0, NULL);
+        pa_thread_free(u->thread);
+        pa_thread_mq_done(&u->thread_mq);
+        pa_asyncmsgq_unref(u->async_msgq);
+    }
+
+    /* free sinks */
+    for (ca_sink = u->sinks; ca_sink;) {
+        coreaudio_sink *next = ca_sink->next;
+
+        if (ca_sink->pa_sink)
+            pa_sink_unref(ca_sink->pa_sink);
+
+        pa_xfree(ca_sink->name);
+        pa_xfree(ca_sink);
+        ca_sink = next;
+    }
+
+    /* free sources */
+    for (ca_source = u->sources; ca_source;) {
+        coreaudio_source *next = ca_source->next;
+
+        if (ca_source->pa_source)
+            pa_source_unref(ca_source->pa_source);
+
+        pa_xfree(ca_source->name);
+        pa_xfree(ca_source);
+        ca_source = next;
+    }
+
+    if (u->proc_id) {
+        AudioDeviceStop(u->device_id, u->proc_id);
+        AudioDeviceDestroyIOProcID(u->device_id, u->proc_id);
+    }
+
+    AudioDeviceRemovePropertyListener(u->device_id, 0, 0, kAudioDevicePropertyStreamFormat, ca_stream_format_changed);
+
+    pa_xfree(u->device_name);
+    pa_xfree(u->vendor_name);
+    pa_rtpoll_free(u->rtpoll);
+    pa_card_free(u->card);
+
+    pa_xfree(u);
+}