audio_manager_win.cc revision 4e180b6a0b4720a9b8e9e959a882386f690f08ff 
1// Copyright (c) 2012 The Chromium Authors. All rights reserved.
2// Use of this source code is governed by a BSD-style license that can be
3// found in the LICENSE file.
4
5#include "media/audio/audio_io.h"
6
7#include <windows.h>
8#include <objbase.h>  // This has to be before initguid.h
9#include <initguid.h>
10#include <mmsystem.h>
11#include <setupapi.h>
12
13#include "base/bind.h"
14#include "base/bind_helpers.h"
15#include "base/command_line.h"
16#include "base/files/file_path.h"
17#include "base/memory/scoped_ptr.h"
18#include "base/message_loop/message_loop.h"
19#include "base/path_service.h"
20#include "base/process/launch.h"
21#include "base/strings/string_number_conversions.h"
22#include "base/strings/string_util.h"
23#include "base/win/windows_version.h"
24#include "media/audio/audio_parameters.h"
25#include "media/audio/win/audio_device_listener_win.h"
26#include "media/audio/win/audio_low_latency_input_win.h"
27#include "media/audio/win/audio_low_latency_output_win.h"
28#include "media/audio/win/audio_manager_win.h"
29#include "media/audio/win/audio_unified_win.h"
30#include "media/audio/win/core_audio_util_win.h"
31#include "media/audio/win/device_enumeration_win.h"
32#include "media/audio/win/wavein_input_win.h"
33#include "media/audio/win/waveout_output_win.h"
34#include "media/base/bind_to_loop.h"
35#include "media/base/channel_layout.h"
36#include "media/base/limits.h"
37#include "media/base/media_switches.h"
38
39// Libraries required for the SetupAPI and Wbem APIs used here.
40#pragma comment(lib, "setupapi.lib")
41
42// The following are defined in various DDK headers, and we (re)define them here
43// to avoid adding the DDK as a chrome dependency.
44#define DRV_QUERYDEVICEINTERFACE 0x80c
45#define DRVM_MAPPER_PREFERRED_GET 0x2015
46#define DRV_QUERYDEVICEINTERFACESIZE 0x80d
47DEFINE_GUID(AM_KSCATEGORY_AUDIO, 0x6994ad04, 0x93ef, 0x11d0,
48            0xa3, 0xcc, 0x00, 0xa0, 0xc9, 0x22, 0x31, 0x96);
49
50namespace media {
51
52// Maximum number of output streams that can be open simultaneously.
53static const int kMaxOutputStreams = 50;
54
55// Up to 8 channels can be passed to the driver.  This should work, given the
56// right drivers, but graceful error handling is needed.
57static const int kWinMaxChannels = 8;
58
59// We use 3 buffers for recording audio so that if a recording callback takes
60// some time to return we won't lose audio. More buffers while recording are
61// ok because they don't introduce any delay in recording, unlike in playback
62// where you first need to fill in that number of buffers before starting to
63// play.
64static const int kNumInputBuffers = 3;
65
66// Buffer size to use for input and output stream when a proper size can't be
67// determined from the system
68static const int kFallbackBufferSize = 2048;
69
70static int GetVersionPartAsInt(DWORDLONG num) {
71  return static_cast<int>(num & 0xffff);
72}
73
74// Returns a string containing the given device's description and installed
75// driver version.
76static string16 GetDeviceAndDriverInfo(HDEVINFO device_info,
77                                       SP_DEVINFO_DATA* device_data) {
78  // Save the old install params setting and set a flag for the
79  // SetupDiBuildDriverInfoList below to return only the installed drivers.
80  SP_DEVINSTALL_PARAMS old_device_install_params;
81  old_device_install_params.cbSize = sizeof(old_device_install_params);
82  SetupDiGetDeviceInstallParams(device_info, device_data,
83                                &old_device_install_params);
84  SP_DEVINSTALL_PARAMS device_install_params = old_device_install_params;
85  device_install_params.FlagsEx |= DI_FLAGSEX_INSTALLEDDRIVER;
86  SetupDiSetDeviceInstallParams(device_info, device_data,
87                                &device_install_params);
88
89  SP_DRVINFO_DATA driver_data;
90  driver_data.cbSize = sizeof(driver_data);
91  string16 device_and_driver_info;
92  if (SetupDiBuildDriverInfoList(device_info, device_data,
93                                 SPDIT_COMPATDRIVER)) {
94    if (SetupDiEnumDriverInfo(device_info, device_data, SPDIT_COMPATDRIVER, 0,
95                              &driver_data)) {
96      DWORDLONG version = driver_data.DriverVersion;
97      device_and_driver_info = string16(driver_data.Description) + L" v" +
98          base::IntToString16(GetVersionPartAsInt((version >> 48))) + L"." +
99          base::IntToString16(GetVersionPartAsInt((version >> 32))) + L"." +
100          base::IntToString16(GetVersionPartAsInt((version >> 16))) + L"." +
101          base::IntToString16(GetVersionPartAsInt(version));
102    }
103    SetupDiDestroyDriverInfoList(device_info, device_data, SPDIT_COMPATDRIVER);
104  }
105
106  SetupDiSetDeviceInstallParams(device_info, device_data,
107                                &old_device_install_params);
108
109  return device_and_driver_info;
110}
111
112static int NumberOfWaveOutBuffers() {
113  // Use the user provided buffer count if provided.
114  int buffers = 0;
115  std::string buffers_str(CommandLine::ForCurrentProcess()->GetSwitchValueASCII(
116      switches::kWaveOutBuffers));
117  if (base::StringToInt(buffers_str, &buffers) && buffers > 0) {
118    return buffers;
119  }
120
121  // Use 4 buffers for Vista, 3 for everyone else:
122  //  - The entire Windows audio stack was rewritten for Windows Vista and wave
123  //    out performance was degraded compared to XP.
124  //  - The regression was fixed in Windows 7 and most configurations will work
125  //    with 2, but some (e.g., some Sound Blasters) still need 3.
126  //  - Some XP configurations (even multi-processor ones) also need 3.
127  return (base::win::GetVersion() == base::win::VERSION_VISTA) ? 4 : 3;
128}
129
130AudioManagerWin::AudioManagerWin() {
131  if (!CoreAudioUtil::IsSupported()) {
132    // Use the Wave API for device enumeration if XP or lower.
133    enumeration_type_ = kWaveEnumeration;
134  } else {
135    // Use the MMDevice API for device enumeration if Vista or higher.
136    enumeration_type_ = kMMDeviceEnumeration;
137  }
138
139  SetMaxOutputStreamsAllowed(kMaxOutputStreams);
140
141  // Task must be posted last to avoid races from handing out "this" to the
142  // audio thread.
143  GetMessageLoop()->PostTask(FROM_HERE, base::Bind(
144      &AudioManagerWin::CreateDeviceListener, base::Unretained(this)));
145}
146
147AudioManagerWin::~AudioManagerWin() {
148  // It's safe to post a task here since Shutdown() will wait for all tasks to
149  // complete before returning.
150  GetMessageLoop()->PostTask(FROM_HERE, base::Bind(
151      &AudioManagerWin::DestroyDeviceListener, base::Unretained(this)));
152  Shutdown();
153}
154
155bool AudioManagerWin::HasAudioOutputDevices() {
156  return (::waveOutGetNumDevs() != 0);
157}
158
159bool AudioManagerWin::HasAudioInputDevices() {
160  return (::waveInGetNumDevs() != 0);
161}
162
163void AudioManagerWin::CreateDeviceListener() {
164  // AudioDeviceListenerWin must be initialized on a COM thread and should only
165  // be used if WASAPI / Core Audio is supported.
166  if (CoreAudioUtil::IsSupported()) {
167    output_device_listener_.reset(new AudioDeviceListenerWin(BindToLoop(
168        GetMessageLoop(), base::Bind(
169            &AudioManagerWin::NotifyAllOutputDeviceChangeListeners,
170            base::Unretained(this)))));
171  }
172}
173
174void AudioManagerWin::DestroyDeviceListener() {
175  output_device_listener_.reset();
176}
177
178string16 AudioManagerWin::GetAudioInputDeviceModel() {
179  // Get the default audio capture device and its device interface name.
180  DWORD device_id = 0;
181  waveInMessage(reinterpret_cast<HWAVEIN>(WAVE_MAPPER),
182                DRVM_MAPPER_PREFERRED_GET,
183                reinterpret_cast<DWORD_PTR>(&device_id), NULL);
184  ULONG device_interface_name_size = 0;
185  waveInMessage(reinterpret_cast<HWAVEIN>(device_id),
186                DRV_QUERYDEVICEINTERFACESIZE,
187                reinterpret_cast<DWORD_PTR>(&device_interface_name_size), 0);
188  size_t bytes_in_char16 = sizeof(string16::value_type);
189  DCHECK_EQ(0u, device_interface_name_size % bytes_in_char16);
190  if (device_interface_name_size <= bytes_in_char16)
191    return string16();  // No audio capture device.
192
193  string16 device_interface_name;
194  string16::value_type* name_ptr = WriteInto(&device_interface_name,
195      device_interface_name_size / bytes_in_char16);
196  waveInMessage(reinterpret_cast<HWAVEIN>(device_id),
197                DRV_QUERYDEVICEINTERFACE,
198                reinterpret_cast<DWORD_PTR>(name_ptr),
199                static_cast<DWORD_PTR>(device_interface_name_size));
200
201  // Enumerate all audio devices and find the one matching the above device
202  // interface name.
203  HDEVINFO device_info = SetupDiGetClassDevs(
204      &AM_KSCATEGORY_AUDIO, 0, 0, DIGCF_DEVICEINTERFACE | DIGCF_PRESENT);
205  if (device_info == INVALID_HANDLE_VALUE)
206    return string16();
207
208  DWORD interface_index = 0;
209  SP_DEVICE_INTERFACE_DATA interface_data;
210  interface_data.cbSize = sizeof(interface_data);
211  while (SetupDiEnumDeviceInterfaces(device_info, 0, &AM_KSCATEGORY_AUDIO,
212                                     interface_index++, &interface_data)) {
213    // Query the size of the struct, allocate it and then query the data.
214    SP_DEVINFO_DATA device_data;
215    device_data.cbSize = sizeof(device_data);
216    DWORD interface_detail_size = 0;
217    SetupDiGetDeviceInterfaceDetail(device_info, &interface_data, 0, 0,
218                                    &interface_detail_size, &device_data);
219    if (!interface_detail_size)
220      continue;
221
222    scoped_ptr<char[]> interface_detail_buffer(new char[interface_detail_size]);
223    SP_DEVICE_INTERFACE_DETAIL_DATA* interface_detail =
224        reinterpret_cast<SP_DEVICE_INTERFACE_DETAIL_DATA*>(
225            interface_detail_buffer.get());
226    interface_detail->cbSize = interface_detail_size;
227    if (!SetupDiGetDeviceInterfaceDetail(device_info, &interface_data,
228                                         interface_detail,
229                                         interface_detail_size, NULL,
230                                         &device_data))
231      return string16();
232
233    bool device_found = (device_interface_name == interface_detail->DevicePath);
234
235    if (device_found)
236      return GetDeviceAndDriverInfo(device_info, &device_data);
237  }
238
239  return string16();
240}
241
242void AudioManagerWin::ShowAudioInputSettings() {
243  std::wstring program;
244  std::string argument;
245  if (!CoreAudioUtil::IsSupported()) {
246    program = L"sndvol32.exe";
247    argument = "-R";
248  } else {
249    program = L"control.exe";
250    argument = "mmsys.cpl,,1";
251  }
252
253  base::FilePath path;
254  PathService::Get(base::DIR_SYSTEM, &path);
255  path = path.Append(program);
256  CommandLine command_line(path);
257  command_line.AppendArg(argument);
258  base::LaunchProcess(command_line, base::LaunchOptions(), NULL);
259}
260
261void AudioManagerWin::GetAudioDeviceNamesImpl(
262    bool input,
263    AudioDeviceNames* device_names) {
264  DCHECK(device_names->empty());
265  DCHECK(enumeration_type() !=  kUninitializedEnumeration);
266  // Enumerate all active audio-endpoint capture devices.
267  if (enumeration_type() == kWaveEnumeration) {
268    // Utilize the Wave API for Windows XP.
269    if (input)
270      GetInputDeviceNamesWinXP(device_names);
271    else
272      GetOutputDeviceNamesWinXP(device_names);
273  } else {
274    // Utilize the MMDevice API (part of Core Audio) for Vista and higher.
275    if (input)
276      GetInputDeviceNamesWin(device_names);
277    else
278      GetOutputDeviceNamesWin(device_names);
279  }
280
281  // Always add default device parameters as first element.
282  if (!device_names->empty()) {
283    AudioDeviceName name;
284    name.device_name = AudioManagerBase::kDefaultDeviceName;
285    name.unique_id = AudioManagerBase::kDefaultDeviceId;
286    device_names->push_front(name);
287  }
288}
289
290void AudioManagerWin::GetAudioInputDeviceNames(AudioDeviceNames* device_names) {
291  GetAudioDeviceNamesImpl(true, device_names);
292}
293
294void AudioManagerWin::GetAudioOutputDeviceNames(
295    AudioDeviceNames* device_names) {
296  GetAudioDeviceNamesImpl(false, device_names);
297}
298
299AudioParameters AudioManagerWin::GetInputStreamParameters(
300    const std::string& device_id) {
301  int sample_rate = 48000;
302  ChannelLayout channel_layout = CHANNEL_LAYOUT_STEREO;
303  if (CoreAudioUtil::IsSupported()) {
304    int hw_sample_rate = WASAPIAudioInputStream::HardwareSampleRate(device_id);
305    if (hw_sample_rate)
306      sample_rate = hw_sample_rate;
307    channel_layout =
308        WASAPIAudioInputStream::HardwareChannelCount(device_id) == 1 ?
309            CHANNEL_LAYOUT_MONO : CHANNEL_LAYOUT_STEREO;
310  }
311
312  // TODO(Henrika): improve the default buffer size value for input stream.
313  return AudioParameters(
314      AudioParameters::AUDIO_PCM_LOW_LATENCY, channel_layout,
315      sample_rate, 16, kFallbackBufferSize);
316}
317
318std::string AudioManagerWin::GetAssociatedOutputDeviceID(
319    const std::string& input_device_id) {
320  if (!CoreAudioUtil::IsSupported()) {
321    NOTIMPLEMENTED()
322        << "GetAssociatedOutputDeviceID is not supported on this OS";
323    return std::string();
324  }
325  return CoreAudioUtil::GetMatchingOutputDeviceID(input_device_id);
326}
327
328// Factory for the implementations of AudioOutputStream for AUDIO_PCM_LINEAR
329// mode.
330// - PCMWaveOutAudioOutputStream: Based on the waveOut API.
331AudioOutputStream* AudioManagerWin::MakeLinearOutputStream(
332    const AudioParameters& params) {
333  DCHECK_EQ(AudioParameters::AUDIO_PCM_LINEAR, params.format());
334  if (params.channels() > kWinMaxChannels)
335    return NULL;
336
337  return new PCMWaveOutAudioOutputStream(this,
338                                         params,
339                                         NumberOfWaveOutBuffers(),
340                                         WAVE_MAPPER);
341}
342
343// Factory for the implementations of AudioOutputStream for
344// AUDIO_PCM_LOW_LATENCY mode. Two implementations should suffice most
345// windows user's needs.
346// - PCMWaveOutAudioOutputStream: Based on the waveOut API.
347// - WASAPIAudioOutputStream: Based on Core Audio (WASAPI) API.
348AudioOutputStream* AudioManagerWin::MakeLowLatencyOutputStream(
349    const AudioParameters& params,
350    const std::string& device_id,
351    const std::string& input_device_id) {
352  DCHECK_EQ(AudioParameters::AUDIO_PCM_LOW_LATENCY, params.format());
353  if (params.channels() > kWinMaxChannels)
354    return NULL;
355
356  if (!CoreAudioUtil::IsSupported()) {
357    // Fall back to Windows Wave implementation on Windows XP or lower.
358    DLOG_IF(ERROR, !device_id.empty())
359        << "Opening by device id not supported by PCMWaveOutAudioOutputStream";
360    DVLOG(1) << "Using WaveOut since WASAPI requires at least Vista.";
361    return new PCMWaveOutAudioOutputStream(
362        this, params, NumberOfWaveOutBuffers(), WAVE_MAPPER);
363  }
364
365  // TODO(rtoy): support more than stereo input.
366  if (params.input_channels() > 0) {
367    DVLOG(1) << "WASAPIUnifiedStream is created.";
368    DLOG_IF(ERROR, !device_id.empty())
369        << "Opening by device id not supported by WASAPIUnifiedStream";
370    return new WASAPIUnifiedStream(this, params, input_device_id);
371  }
372
373  return new WASAPIAudioOutputStream(this, device_id, params, eConsole);
374}
375
376// Factory for the implementations of AudioInputStream for AUDIO_PCM_LINEAR
377// mode.
378AudioInputStream* AudioManagerWin::MakeLinearInputStream(
379    const AudioParameters& params, const std::string& device_id) {
380  DCHECK_EQ(AudioParameters::AUDIO_PCM_LINEAR, params.format());
381  return CreatePCMWaveInAudioInputStream(params, device_id);
382}
383
384// Factory for the implementations of AudioInputStream for
385// AUDIO_PCM_LOW_LATENCY mode.
386AudioInputStream* AudioManagerWin::MakeLowLatencyInputStream(
387    const AudioParameters& params, const std::string& device_id) {
388  DCHECK_EQ(AudioParameters::AUDIO_PCM_LOW_LATENCY, params.format());
389  AudioInputStream* stream = NULL;
390  if (!CoreAudioUtil::IsSupported()) {
391    // Fall back to Windows Wave implementation on Windows XP or lower.
392    DVLOG(1) << "Using WaveIn since WASAPI requires at least Vista.";
393    stream = CreatePCMWaveInAudioInputStream(params, device_id);
394  } else {
395    stream = new WASAPIAudioInputStream(this, params, device_id);
396  }
397
398  return stream;
399}
400
401std::string AudioManagerWin::GetDefaultOutputDeviceID() {
402  if (!CoreAudioUtil::IsSupported())
403    return std::string();
404  return CoreAudioUtil::GetDefaultOutputDeviceID();
405}
406
407AudioParameters AudioManagerWin::GetPreferredOutputStreamParameters(
408    const std::string& output_device_id,
409    const AudioParameters& input_params) {
410  const bool core_audio_supported = CoreAudioUtil::IsSupported();
411  DLOG_IF(ERROR, !core_audio_supported && !output_device_id.empty())
412      << "CoreAudio is required to open non-default devices.";
413
414  const CommandLine* cmd_line = CommandLine::ForCurrentProcess();
415  ChannelLayout channel_layout = CHANNEL_LAYOUT_STEREO;
416  int sample_rate = 48000;
417  int buffer_size = kFallbackBufferSize;
418  int bits_per_sample = 16;
419  int input_channels = 0;
420  bool use_input_params = !core_audio_supported;
421  if (core_audio_supported) {
422    if (cmd_line->HasSwitch(switches::kEnableExclusiveAudio)) {
423      // TODO(rtoy): tune these values for best possible WebAudio
424      // performance. WebRTC works well at 48kHz and a buffer size of 480
425      // samples will be used for this case. Note that exclusive mode is
426      // experimental. This sample rate will be combined with a buffer size of
427      // 256 samples, which corresponds to an output delay of ~5.33ms.
428      sample_rate = 48000;
429      buffer_size = 256;
430      if (input_params.IsValid())
431        channel_layout = input_params.channel_layout();
432    } else {
433      AudioParameters params;
434      HRESULT hr = CoreAudioUtil::GetPreferredAudioParameters(
435          output_device_id.empty() ?
436              GetDefaultOutputDeviceID() : output_device_id,
437          &params);
438      if (SUCCEEDED(hr)) {
439        bits_per_sample = params.bits_per_sample();
440        buffer_size = params.frames_per_buffer();
441        channel_layout = params.channel_layout();
442        sample_rate = params.sample_rate();
443      } else {
444        use_input_params = true;
445      }
446    }
447  }
448
449  if (input_params.IsValid()) {
450    if (core_audio_supported &&
451        cmd_line->HasSwitch(switches::kTrySupportedChannelLayouts)) {
452      // Check if it is possible to open up at the specified input channel
453      // layout but avoid checking if the specified layout is the same as the
454      // hardware (preferred) layout. We do this extra check to avoid the
455      // CoreAudioUtil::IsChannelLayoutSupported() overhead in most cases.
456      if (input_params.channel_layout() != channel_layout) {
457        // TODO(henrika): Use |output_device_id| here.
458        // Internally, IsChannelLayoutSupported does many of the operations
459        // that have already been done such as opening up a client and fetching
460        // the WAVEFORMATPCMEX format.  Ideally we should only do that once and
461        // do it for the requested device.  Then here, we can check the layout
462        // from the data we already hold.
463        if (CoreAudioUtil::IsChannelLayoutSupported(
464                eRender, eConsole, input_params.channel_layout())) {
465          // Open up using the same channel layout as the source if it is
466          // supported by the hardware.
467          channel_layout = input_params.channel_layout();
468          VLOG(1) << "Hardware channel layout is not used; using same layout"
469                  << " as the source instead (" << channel_layout << ")";
470        }
471      }
472    }
473    input_channels = input_params.input_channels();
474    if (use_input_params) {
475      // If WASAPI isn't supported we'll fallback to WaveOut, which will take
476      // care of resampling and bits per sample changes.  By setting these
477      // equal to the input values, AudioOutputResampler will skip resampling
478      // and bit per sample differences (since the input parameters will match
479      // the output parameters).
480      bits_per_sample = input_params.bits_per_sample();
481      buffer_size = input_params.frames_per_buffer();
482      channel_layout = input_params.channel_layout();
483      sample_rate = input_params.sample_rate();
484    }
485  }
486
487  int user_buffer_size = GetUserBufferSize();
488  if (user_buffer_size)
489    buffer_size = user_buffer_size;
490
491  return AudioParameters(
492      AudioParameters::AUDIO_PCM_LOW_LATENCY, channel_layout, input_channels,
493      sample_rate, bits_per_sample, buffer_size);
494}
495
496AudioInputStream* AudioManagerWin::CreatePCMWaveInAudioInputStream(
497    const AudioParameters& params,
498    const std::string& device_id) {
499  std::string xp_device_id = device_id;
500  if (device_id != AudioManagerBase::kDefaultDeviceId &&
501      enumeration_type_ == kMMDeviceEnumeration) {
502    xp_device_id = ConvertToWinXPInputDeviceId(device_id);
503    if (xp_device_id.empty()) {
504      DLOG(ERROR) << "Cannot find a waveIn device which matches the device ID "
505                  << device_id;
506      return NULL;
507    }
508  }
509
510  return new PCMWaveInAudioInputStream(this, params, kNumInputBuffers,
511                                       xp_device_id);
512}
513
514/// static
515AudioManager* CreateAudioManager() {
516  return new AudioManagerWin();
517}
518
519}  // namespace media
520