audio/device_port_source.cpp - device/generic/goldfish - Gitiles

 /*
  * Copyright (C) 2020 The Android Open Source Project
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
  * You may obtain a copy of the License at
  *
  *      http://www.apache.org/licenses/LICENSE-2.0
  *
  * Unless required by applicable law or agreed to in writing, software
  * distributed under the License is distributed on an "AS IS" BASIS,
  * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  * See the License for the specific language governing permissions and
  * limitations under the License.
  */

 #include <android_audio_policy_configuration_V7_0-enums.h>
 #include <cmath>
 #include <chrono>
 #include <thread>
 #include <unistd.h>
 #include <audio_utils/channels.h>
 #include <audio_utils/format.h>
 #include <log/log.h>
 #include <utils/ThreadDefs.h>
 #include <utils/Timers.h>
 #include "device_port_source.h"
 #include "talsa.h"
 #include "ring_buffer.h"
 #include "audio_ops.h"
 #include "util.h"
 #include "debug.h"

 namespace xsd {
 using namespace ::android::audio::policy::configuration::V7_0;
 }

 namespace android {
 namespace hardware {
 namespace audio {
 namespace V7_0 {
 namespace implementation {

 namespace {

 constexpr int kMaxJitterUs = 3000;  // Enforced by CTS, should be <= 6ms

 struct TinyalsaSource : public DevicePortSource {
     TinyalsaSource(unsigned pcmCard, unsigned pcmDevice,
                    const AudioConfig &cfg, uint64_t &frames)
             : mStartNs(systemTime(SYSTEM_TIME_MONOTONIC))
             , mSampleRateHz(cfg.base.sampleRateHz)
             , mFrameSize(util::countChannels(cfg.base.channelMask) * sizeof(int16_t))
             , mReadSizeFrames(cfg.frameCount)
             , mFrames(frames)
             , mRingBuffer(mFrameSize * cfg.frameCount * 3)
             , mMixer(pcmCard)
             , mPcm(talsa::pcmOpen(pcmCard, pcmDevice,
                                   util::countChannels(cfg.base.channelMask),
                                   cfg.base.sampleRateHz,
                                   cfg.frameCount,
                                   false /* isOut */)) {
         mProduceThread = std::thread(&TinyalsaSource::producerThread, this);
     }

     ~TinyalsaSource() {
         mProduceThreadRunning = false;
         mProduceThread.join();
     }

     Result getCapturePosition(uint64_t &frames, uint64_t &time) override {
         const nsecs_t nowNs = systemTime(SYSTEM_TIME_MONOTONIC);
         const uint64_t nowFrames = getCaptureFrames(nowNs);
         mFrames += (nowFrames - mPreviousFrames);
         mPreviousFrames = nowFrames;

         frames = mFrames;
         time = nowNs;
         return Result::OK;
     }

     uint64_t getCaptureFrames(const nsecs_t nowNs) const {
         return uint64_t(mSampleRateHz) * ns2us(nowNs - mStartNs) / 1000000;
     }

     uint64_t getAvailableFrames(const nsecs_t nowNs) const {
         return getCaptureFrames(nowNs) - mSentFrames;
     }

     uint64_t getAvailableFramesNow() const {
         return getAvailableFrames(systemTime(SYSTEM_TIME_MONOTONIC));
     }

     size_t getWaitFramesNow(const size_t requestedFrames) const {
         const size_t availableFrames = getAvailableFramesNow();
         return (requestedFrames > availableFrames)
             ? (requestedFrames - availableFrames) : 0;
     }

     size_t read(float volume, size_t bytesToRead, IWriter &writer) override {
         const size_t waitFrames = getWaitFramesNow(bytesToRead / mFrameSize);
         const auto blockUntil =
             std::chrono::high_resolution_clock::now() +
                 + std::chrono::microseconds(waitFrames * 1000000 / mSampleRateHz);

         while (bytesToRead > 0) {
             if (mRingBuffer.waitForConsumeAvailable(blockUntil
                     + std::chrono::microseconds(kMaxJitterUs))) {
                 if (mRingBuffer.availableToConsume() >= bytesToRead) {
                     // Since the ring buffer has all bytes we need, make sure we
                     // are not too early here: tinyalsa is jittery, we don't
                     // want to go faster than SYSTEM_TIME_MONOTONIC
                     std::this_thread::sleep_until(blockUntil);
                 }

                 auto chunk = mRingBuffer.getConsumeChunk();
                 const size_t writeBufSzBytes = std::min(chunk.size, bytesToRead);

                 aops::multiplyByVolume(volume,
                                        static_cast<int16_t *>(chunk.data),
                                        writeBufSzBytes / sizeof(int16_t));

                 writer(chunk.data, writeBufSzBytes);
                 LOG_ALWAYS_FATAL_IF(mRingBuffer.consume(chunk, writeBufSzBytes) < writeBufSzBytes);

                 bytesToRead -= writeBufSzBytes;
                 mSentFrames += writeBufSzBytes / mFrameSize;
             } else {
                 ALOGW("TinyalsaSource::%s:%d pcm_read was late delivering "
                       "frames, inserting %zu us of silence",
                       __func__, __LINE__,
                       size_t(1000000 * bytesToRead / mFrameSize / mSampleRateHz));

                 static const uint8_t zeroes[256] = {0};

                 while (bytesToRead > 0) {
                     const size_t nZeroFrames =
                         std::min(bytesToRead, sizeof(zeroes)) / mFrameSize;
                     const size_t nZeroBytes = nZeroFrames * mFrameSize;

                     writer(zeroes, nZeroBytes);
                     mSentFrames += nZeroFrames;
                     bytesToRead -= nZeroBytes;
                 }
                 break;
             }
         }

         return mFramesLost.exchange(0);
     }

     void producerThread() {
         util::setThreadPriority(PRIORITY_URGENT_AUDIO);
         std::vector<uint8_t> readBuf(mReadSizeFrames * mFrameSize);

         while (mProduceThreadRunning) {
             const size_t bytesLost = mRingBuffer.makeRoomForProduce(readBuf.size());
             mFramesLost += bytesLost / mFrameSize;

             auto produceChunk = mRingBuffer.getProduceChunk();
             if (produceChunk.size < readBuf.size()) {
                 const size_t sz = doRead(readBuf.data(), readBuf.size());
                 if (sz > 0) {
                     LOG_ALWAYS_FATAL_IF(mRingBuffer.produce(readBuf.data(), sz) < sz);
                 }
             } else {
                 const size_t sz = doRead(produceChunk.data, readBuf.size());
                 if (sz > 0) {
                     LOG_ALWAYS_FATAL_IF(mRingBuffer.produce(readBuf.size()) < sz);
                 }
             }
         }
     }

     size_t doRead(void *dst, size_t sz) {
         const int res = ::pcm_read(mPcm.get(), dst, sz);
         if (res < 0) {
             ALOGW("TinyalsaSource::%s:%d pcm_read failed with res=%d",
                   __func__, __LINE__, res);
             return 0;
         }

         return sz;
     }

     static std::unique_ptr<TinyalsaSource> create(unsigned pcmCard,
                                                   unsigned pcmDevice,
                                                   const AudioConfig &cfg,
                                                   size_t writerBufferSizeHint,
                                                   uint64_t &frames) {
         (void)writerBufferSizeHint;

         auto src = std::make_unique<TinyalsaSource>(pcmCard, pcmDevice,
                                                     cfg, frames);
         if (src->mMixer && src->mPcm) {
             return src;
         } else {
             return FAILURE(nullptr);
         }
     }

 private:
     const nsecs_t mStartNs;
     const unsigned mSampleRateHz;
     const unsigned mFrameSize;
     const unsigned mReadSizeFrames;
     uint64_t &mFrames;
     uint64_t mPreviousFrames = 0;
     uint64_t mSentFrames = 0;
     std::atomic<uint32_t> mFramesLost = 0;
     RingBuffer mRingBuffer;
     talsa::Mixer mMixer;
     talsa::PcmPtr mPcm;
     std::thread mProduceThread;
     std::atomic<bool> mProduceThreadRunning = true;
 };

 template <class G> struct GeneratedSource : public DevicePortSource {
     GeneratedSource(const AudioConfig &cfg,
                     size_t writerBufferSizeHint,
                     uint64_t &frames,
                     G generator)
             : mWriteBuffer(writerBufferSizeHint / sizeof(int16_t))
             , mFrames(frames)
             , mStartNs(systemTime(SYSTEM_TIME_MONOTONIC))
             , mSampleRateHz(cfg.base.sampleRateHz)
             , mNChannels(util::countChannels(cfg.base.channelMask))
             , mGenerator(std::move(generator)) {}

     Result getCapturePosition(uint64_t &frames, uint64_t &time) override {
         const nsecs_t nowNs = systemTime(SYSTEM_TIME_MONOTONIC);
         const uint64_t nowFrames = getCaptureFrames(nowNs);
         mFrames += (nowFrames - mPreviousFrames);
         mPreviousFrames = nowFrames;
         frames = mFrames;
         time = nowNs;
         return Result::OK;
     }

     uint64_t getCaptureFrames(const nsecs_t nowNs) const {
         return uint64_t(mSampleRateHz) * ns2us(nowNs - mStartNs) / 1000000;
     }

     uint64_t getAvailableFrames(const nsecs_t nowNs) const {
         return getCaptureFrames(nowNs) - mSentFrames;
     }

     size_t read(float volume, size_t bytesToRead, IWriter &writer) override {
         mWriteBuffer.resize(bytesToRead / sizeof(int16_t));

         int16_t *samples = mWriteBuffer.data();
         const unsigned nChannels = mNChannels;
         const unsigned requestedFrames = bytesToRead / nChannels / sizeof(*samples);

         unsigned availableFrames;
         while (true) {
             const nsecs_t nowNs = systemTime(SYSTEM_TIME_MONOTONIC);
             availableFrames = getAvailableFrames(nowNs);
             if (availableFrames < requestedFrames / 2) {
                 const unsigned neededMoreFrames = requestedFrames / 2 - availableFrames;

                 using namespace std::chrono_literals;
                 std::this_thread::sleep_for(1s * neededMoreFrames / mSampleRateHz);
             } else {
                 break;
             }
         }

         const unsigned nFrames = std::min(requestedFrames, availableFrames);
         mGenerator(samples, nFrames);
         const size_t sizeBytes = nFrames * nChannels * sizeof(*samples);
         if (nChannels > 1) {
             adjust_channels(samples, 1, samples, nChannels,
                             sizeof(*samples), sizeBytes);
         }
         mSentFrames += nFrames;

         aops::multiplyByVolume(volume,
                                mWriteBuffer.data(),
                                sizeBytes / sizeof(int16_t));

         writer(mWriteBuffer.data(), sizeBytes);
         return 0;
     }

 private:
     std::vector<int16_t> mWriteBuffer;
     uint64_t &mFrames;
     const nsecs_t mStartNs;
     const unsigned mSampleRateHz;
     const unsigned mNChannels;
     uint64_t mPreviousFrames = 0;
     uint64_t mSentFrames = 0;
     G mGenerator;
 };

 std::vector<int16_t> convertFloatsToInt16(const std::vector<float> &pcmFloat) {
     std::vector<int16_t> pcmI16(pcmFloat.size());

     memcpy_by_audio_format(pcmI16.data(),   AUDIO_FORMAT_PCM_16_BIT,
                            pcmFloat.data(), AUDIO_FORMAT_PCM_FLOAT,
                            pcmFloat.size());

     return pcmI16;
 }

 // https://en.wikipedia.org/wiki/Busy_signal
 struct BusySignalGenerator {
     explicit BusySignalGenerator(const uint32_t sampleRateHz) : mSampleRateHz(sampleRateHz) {
         // 24/480 = 31/620, mValues must contain 50ms of audio samples
         const size_t sz = sampleRateHz / 20;
         std::vector<float> pcm(sz);
         for (unsigned i = 0; i < sz; ++i) {
             const double a = double(i) * M_PI * 2 / sampleRateHz;
             pcm[i] = .5 * (sin(480 * a) + sin(620 * a));
         }
         mValues = convertFloatsToInt16(pcm);
     }

     void operator()(int16_t* s, size_t n) {
         const unsigned rate = mSampleRateHz;
         const unsigned rateHalf = rate / 2;
         const int16_t *const vals = mValues.data();
         const size_t valsSz = mValues.size();
         size_t i = mI;

         while (n > 0) {
             size_t len;
             if (i < rateHalf) {
                 const size_t valsOff = i % valsSz;
                 len = std::min(n, std::min(rateHalf - i, valsSz - valsOff));
                 memcpy(s, vals + valsOff, len * sizeof(*s));
             } else {
                 len = std::min(n, rate - i);
                 memset(s, 0, len * sizeof(*s));
             }
             s += len;
             i = (i + len) % rate;
             n -= len;
         }

         mI = i;
     }

 private:
     const unsigned mSampleRateHz;
     std::vector<int16_t> mValues;
     size_t mI = 0;
 };

 struct RepeatGenerator {
     explicit RepeatGenerator(const std::vector<float> &pcm)
             : mValues(convertFloatsToInt16(pcm)) {}

     void operator()(int16_t* s, size_t n) {
         const int16_t *const vals = mValues.data();
         const size_t valsSz = mValues.size();
         size_t i = mI;

         while (n > 0) {
             const size_t len = std::min(n, valsSz - i);
             memcpy(s, vals + i, len * sizeof(*s));
             s += len;
             i = (i + len) % valsSz;
             n -= len;
         }

         mI = i;
     }

 private:
     const std::vector<int16_t> mValues;
     size_t mI = 0;
 };

 std::vector<float> generateSinePattern(uint32_t sampleRateHz,
                                        double freq,
                                        double amp) {
     std::vector<float> result(3 * sampleRateHz / freq + .5);

     for (size_t i = 0; i < result.size(); ++i) {
         const double a = double(i) * M_PI * 2 / sampleRateHz;
         result[i] = amp * sin(a * freq);
     }

     return result;
 }

 template <class G> std::unique_ptr<GeneratedSource<G>>
 createGeneratedSource(const AudioConfig &cfg,
                       size_t writerBufferSizeHint,
                       uint64_t &frames,
                       G generator) {
     return std::make_unique<GeneratedSource<G>>(cfg,
                                                 writerBufferSizeHint,
                                                 frames,
                                                 std::move(generator));
 }

 }  // namespace

 std::unique_ptr<DevicePortSource>
 DevicePortSource::create(size_t writerBufferSizeHint,
                          const DeviceAddress &address,
                          const AudioConfig &cfg,
                          const hidl_vec<AudioInOutFlag> &flags,
                          uint64_t &frames) {
     (void)flags;

     if (xsd::stringToAudioFormat(cfg.base.format) != xsd::AudioFormat::AUDIO_FORMAT_PCM_16_BIT) {
         ALOGE("%s:%d, unexpected format: '%s'", __func__, __LINE__, cfg.base.format.c_str());
         return FAILURE(nullptr);
     }

     switch (xsd::stringToAudioDevice(address.deviceType)) {
     case xsd::AudioDevice::AUDIO_DEVICE_IN_DEFAULT:
     case xsd::AudioDevice::AUDIO_DEVICE_IN_BUILTIN_MIC:
         {
             auto sourceptr = TinyalsaSource::create(talsa::kPcmCard, talsa::kPcmDevice,
                                                     cfg, writerBufferSizeHint, frames);
             if (sourceptr != nullptr) {
                 return sourceptr;
             } else {
                 ALOGW("%s:%d failed to create alsa source for '%s'; creating a tone source instead.",
                       __func__, __LINE__, address.deviceType.c_str());
             }
         }
         break;

     case xsd::AudioDevice::AUDIO_DEVICE_IN_TELEPHONY_RX:
         return createGeneratedSource(cfg, writerBufferSizeHint, frames,
                                      BusySignalGenerator(cfg.base.sampleRateHz));

     case xsd::AudioDevice::AUDIO_DEVICE_IN_FM_TUNER:
         return createGeneratedSource(
             cfg, writerBufferSizeHint, frames,
             RepeatGenerator(generateSinePattern(cfg.base.sampleRateHz, 440.0, 1.0)));

     default:
         ALOGW("%s:%d unsupported device: '%s', creating a tone source",
               __func__, __LINE__, address.deviceType.c_str());
         break;
     }

     return createGeneratedSource(
         cfg, writerBufferSizeHint, frames,
         RepeatGenerator(generateSinePattern(cfg.base.sampleRateHz, 220.0, 1.0)));
 }

 bool DevicePortSource::validateDeviceAddress(const DeviceAddress& address) {
     switch (xsd::stringToAudioDevice(address.deviceType)) {
     default:
         ALOGW("%s:%d unsupported device: '%s'", __func__, __LINE__, address.deviceType.c_str());
         return FAILURE(false);

     case xsd::AudioDevice::AUDIO_DEVICE_IN_DEFAULT:
     case xsd::AudioDevice::AUDIO_DEVICE_IN_BUILTIN_MIC:
     case xsd::AudioDevice::AUDIO_DEVICE_IN_TELEPHONY_RX:
     case xsd::AudioDevice::AUDIO_DEVICE_IN_FM_TUNER:
     case xsd::AudioDevice::AUDIO_DEVICE_IN_BUS:
         break;
     }

     return true;
 }

 }  // namespace implementation
 }  // namespace V7_0
 }  // namespace audio
 }  // namespace hardware
 }  // namespace android
	/*
	* Copyright (C) 2020 The Android Open Source Project
	*
	* Licensed under the Apache License, Version 2.0 (the "License");
	* you may not use this file except in compliance with the License.
	* You may obtain a copy of the License at
	*
	* http://www.apache.org/licenses/LICENSE-2.0
	*
	* Unless required by applicable law or agreed to in writing, software
	* distributed under the License is distributed on an "AS IS" BASIS,
	* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
	* See the License for the specific language governing permissions and
	* limitations under the License.
	*/

	#include <android_audio_policy_configuration_V7_0-enums.h>
	#include <cmath>
	#include <chrono>
	#include <thread>
	#include <unistd.h>
	#include <audio_utils/channels.h>
	#include <audio_utils/format.h>
	#include <log/log.h>
	#include <utils/ThreadDefs.h>
	#include <utils/Timers.h>
	#include "device_port_source.h"
	#include "talsa.h"
	#include "ring_buffer.h"
	#include "audio_ops.h"
	#include "util.h"
	#include "debug.h"

	namespace xsd {
	using namespace ::android::audio::policy::configuration::V7_0;
	}

	namespace android {
	namespace hardware {
	namespace audio {
	namespace V7_0 {
	namespace implementation {

	namespace {

	constexpr int kMaxJitterUs = 3000; // Enforced by CTS, should be <= 6ms

	struct TinyalsaSource : public DevicePortSource {
	TinyalsaSource(unsigned pcmCard, unsigned pcmDevice,
	const AudioConfig &cfg, uint64_t &frames)
	: mStartNs(systemTime(SYSTEM_TIME_MONOTONIC))
	, mSampleRateHz(cfg.base.sampleRateHz)
	, mFrameSize(util::countChannels(cfg.base.channelMask) * sizeof(int16_t))
	, mReadSizeFrames(cfg.frameCount)
	, mFrames(frames)
	, mRingBuffer(mFrameSize * cfg.frameCount * 3)
	, mMixer(pcmCard)
	, mPcm(talsa::pcmOpen(pcmCard, pcmDevice,
	util::countChannels(cfg.base.channelMask),
	cfg.base.sampleRateHz,
	cfg.frameCount,
	false /* isOut */)) {
	mProduceThread = std::thread(&TinyalsaSource::producerThread, this);
	}

	~TinyalsaSource() {
	mProduceThreadRunning = false;
	mProduceThread.join();
	}

	Result getCapturePosition(uint64_t &frames, uint64_t &time) override {
	const nsecs_t nowNs = systemTime(SYSTEM_TIME_MONOTONIC);
	const uint64_t nowFrames = getCaptureFrames(nowNs);
	mFrames += (nowFrames - mPreviousFrames);
	mPreviousFrames = nowFrames;

	frames = mFrames;
	time = nowNs;
	return Result::OK;
	}

	uint64_t getCaptureFrames(const nsecs_t nowNs) const {
	return uint64_t(mSampleRateHz) * ns2us(nowNs - mStartNs) / 1000000;
	}

	uint64_t getAvailableFrames(const nsecs_t nowNs) const {
	return getCaptureFrames(nowNs) - mSentFrames;
	}

	uint64_t getAvailableFramesNow() const {
	return getAvailableFrames(systemTime(SYSTEM_TIME_MONOTONIC));
	}

	size_t getWaitFramesNow(const size_t requestedFrames) const {
	const size_t availableFrames = getAvailableFramesNow();
	return (requestedFrames > availableFrames)
	? (requestedFrames - availableFrames) : 0;
	}

	size_t read(float volume, size_t bytesToRead, IWriter &writer) override {
	const size_t waitFrames = getWaitFramesNow(bytesToRead / mFrameSize);
	const auto blockUntil =
	std::chrono::high_resolution_clock::now() +
	+ std::chrono::microseconds(waitFrames * 1000000 / mSampleRateHz);

	while (bytesToRead > 0) {
	if (mRingBuffer.waitForConsumeAvailable(blockUntil
	+ std::chrono::microseconds(kMaxJitterUs))) {
	if (mRingBuffer.availableToConsume() >= bytesToRead) {
	// Since the ring buffer has all bytes we need, make sure we
	// are not too early here: tinyalsa is jittery, we don't
	// want to go faster than SYSTEM_TIME_MONOTONIC
	std::this_thread::sleep_until(blockUntil);
	}

	auto chunk = mRingBuffer.getConsumeChunk();
	const size_t writeBufSzBytes = std::min(chunk.size, bytesToRead);

	aops::multiplyByVolume(volume,
	static_cast<int16_t *>(chunk.data),
	writeBufSzBytes / sizeof(int16_t));

	writer(chunk.data, writeBufSzBytes);
	LOG_ALWAYS_FATAL_IF(mRingBuffer.consume(chunk, writeBufSzBytes) < writeBufSzBytes);

	bytesToRead -= writeBufSzBytes;
	mSentFrames += writeBufSzBytes / mFrameSize;
	} else {
	ALOGW("TinyalsaSource::%s:%d pcm_read was late delivering "
	"frames, inserting %zu us of silence",
	__func__, __LINE__,
	size_t(1000000 * bytesToRead / mFrameSize / mSampleRateHz));

	static const uint8_t zeroes[256] = {0};

	while (bytesToRead > 0) {
	const size_t nZeroFrames =
	std::min(bytesToRead, sizeof(zeroes)) / mFrameSize;
	const size_t nZeroBytes = nZeroFrames * mFrameSize;

	writer(zeroes, nZeroBytes);
	mSentFrames += nZeroFrames;
	bytesToRead -= nZeroBytes;
	}
	break;
	}
	}

	return mFramesLost.exchange(0);
	}

	void producerThread() {
	util::setThreadPriority(PRIORITY_URGENT_AUDIO);
	std::vector<uint8_t> readBuf(mReadSizeFrames * mFrameSize);

	while (mProduceThreadRunning) {
	const size_t bytesLost = mRingBuffer.makeRoomForProduce(readBuf.size());
	mFramesLost += bytesLost / mFrameSize;

	auto produceChunk = mRingBuffer.getProduceChunk();
	if (produceChunk.size < readBuf.size()) {
	const size_t sz = doRead(readBuf.data(), readBuf.size());
	if (sz > 0) {
	LOG_ALWAYS_FATAL_IF(mRingBuffer.produce(readBuf.data(), sz) < sz);
	}
	} else {
	const size_t sz = doRead(produceChunk.data, readBuf.size());
	if (sz > 0) {
	LOG_ALWAYS_FATAL_IF(mRingBuffer.produce(readBuf.size()) < sz);
	}
	}
	}
	}

	size_t doRead(void *dst, size_t sz) {
	const int res = ::pcm_read(mPcm.get(), dst, sz);
	if (res < 0) {
	ALOGW("TinyalsaSource::%s:%d pcm_read failed with res=%d",
	__func__, __LINE__, res);
	return 0;
	}

	return sz;
	}

	static std::unique_ptr<TinyalsaSource> create(unsigned pcmCard,
	unsigned pcmDevice,
	const AudioConfig &cfg,
	size_t writerBufferSizeHint,
	uint64_t &frames) {
	(void)writerBufferSizeHint;

	auto src = std::make_unique<TinyalsaSource>(pcmCard, pcmDevice,
	cfg, frames);
	if (src->mMixer && src->mPcm) {
	return src;
	} else {
	return FAILURE(nullptr);
	}
	}

	private:
	const nsecs_t mStartNs;
	const unsigned mSampleRateHz;
	const unsigned mFrameSize;
	const unsigned mReadSizeFrames;
	uint64_t &mFrames;
	uint64_t mPreviousFrames = 0;
	uint64_t mSentFrames = 0;
	std::atomic<uint32_t> mFramesLost = 0;
	RingBuffer mRingBuffer;
	talsa::Mixer mMixer;
	talsa::PcmPtr mPcm;
	std::thread mProduceThread;
	std::atomic<bool> mProduceThreadRunning = true;
	};

	template <class G> struct GeneratedSource : public DevicePortSource {
	GeneratedSource(const AudioConfig &cfg,
	size_t writerBufferSizeHint,
	uint64_t &frames,
	G generator)
	: mWriteBuffer(writerBufferSizeHint / sizeof(int16_t))
	, mFrames(frames)
	, mStartNs(systemTime(SYSTEM_TIME_MONOTONIC))
	, mSampleRateHz(cfg.base.sampleRateHz)
	, mNChannels(util::countChannels(cfg.base.channelMask))
	, mGenerator(std::move(generator)) {}

	Result getCapturePosition(uint64_t &frames, uint64_t &time) override {
	const nsecs_t nowNs = systemTime(SYSTEM_TIME_MONOTONIC);
	const uint64_t nowFrames = getCaptureFrames(nowNs);
	mFrames += (nowFrames - mPreviousFrames);
	mPreviousFrames = nowFrames;
	frames = mFrames;
	time = nowNs;
	return Result::OK;
	}

	uint64_t getCaptureFrames(const nsecs_t nowNs) const {
	return uint64_t(mSampleRateHz) * ns2us(nowNs - mStartNs) / 1000000;
	}

	uint64_t getAvailableFrames(const nsecs_t nowNs) const {
	return getCaptureFrames(nowNs) - mSentFrames;
	}

	size_t read(float volume, size_t bytesToRead, IWriter &writer) override {
	mWriteBuffer.resize(bytesToRead / sizeof(int16_t));

	int16_t *samples = mWriteBuffer.data();
	const unsigned nChannels = mNChannels;
	const unsigned requestedFrames = bytesToRead / nChannels / sizeof(*samples);

	unsigned availableFrames;
	while (true) {
	const nsecs_t nowNs = systemTime(SYSTEM_TIME_MONOTONIC);
	availableFrames = getAvailableFrames(nowNs);
	if (availableFrames < requestedFrames / 2) {
	const unsigned neededMoreFrames = requestedFrames / 2 - availableFrames;

	using namespace std::chrono_literals;
	std::this_thread::sleep_for(1s * neededMoreFrames / mSampleRateHz);
	} else {
	break;
	}
	}

	const unsigned nFrames = std::min(requestedFrames, availableFrames);
	mGenerator(samples, nFrames);
	const size_t sizeBytes = nFrames * nChannels * sizeof(*samples);
	if (nChannels > 1) {
	adjust_channels(samples, 1, samples, nChannels,
	sizeof(*samples), sizeBytes);
	}
	mSentFrames += nFrames;

	aops::multiplyByVolume(volume,
	mWriteBuffer.data(),
	sizeBytes / sizeof(int16_t));

	writer(mWriteBuffer.data(), sizeBytes);
	return 0;
	}

	private:
	std::vector<int16_t> mWriteBuffer;
	uint64_t &mFrames;
	const nsecs_t mStartNs;
	const unsigned mSampleRateHz;
	const unsigned mNChannels;
	uint64_t mPreviousFrames = 0;
	uint64_t mSentFrames = 0;
	G mGenerator;
	};

	std::vector<int16_t> convertFloatsToInt16(const std::vector<float> &pcmFloat) {
	std::vector<int16_t> pcmI16(pcmFloat.size());

	memcpy_by_audio_format(pcmI16.data(), AUDIO_FORMAT_PCM_16_BIT,
	pcmFloat.data(), AUDIO_FORMAT_PCM_FLOAT,
	pcmFloat.size());

	return pcmI16;
	}

	// https://en.wikipedia.org/wiki/Busy_signal
	struct BusySignalGenerator {
	explicit BusySignalGenerator(const uint32_t sampleRateHz) : mSampleRateHz(sampleRateHz) {
	// 24/480 = 31/620, mValues must contain 50ms of audio samples
	const size_t sz = sampleRateHz / 20;
	std::vector<float> pcm(sz);
	for (unsigned i = 0; i < sz; ++i) {
	const double a = double(i) * M_PI * 2 / sampleRateHz;
	pcm[i] = .5 * (sin(480 * a) + sin(620 * a));
	}
	mValues = convertFloatsToInt16(pcm);
	}

	void operator()(int16_t* s, size_t n) {
	const unsigned rate = mSampleRateHz;
	const unsigned rateHalf = rate / 2;
	const int16_t *const vals = mValues.data();
	const size_t valsSz = mValues.size();
	size_t i = mI;

	while (n > 0) {
	size_t len;
	if (i < rateHalf) {
	const size_t valsOff = i % valsSz;
	len = std::min(n, std::min(rateHalf - i, valsSz - valsOff));
	memcpy(s, vals + valsOff, len * sizeof(*s));
	} else {
	len = std::min(n, rate - i);
	memset(s, 0, len * sizeof(*s));
	}
	s += len;
	i = (i + len) % rate;
	n -= len;
	}

	mI = i;
	}

	private:
	const unsigned mSampleRateHz;
	std::vector<int16_t> mValues;
	size_t mI = 0;
	};

	struct RepeatGenerator {
	explicit RepeatGenerator(const std::vector<float> &pcm)
	: mValues(convertFloatsToInt16(pcm)) {}

	void operator()(int16_t* s, size_t n) {
	const int16_t *const vals = mValues.data();
	const size_t valsSz = mValues.size();
	size_t i = mI;

	while (n > 0) {
	const size_t len = std::min(n, valsSz - i);
	memcpy(s, vals + i, len * sizeof(*s));
	s += len;
	i = (i + len) % valsSz;
	n -= len;
	}

	mI = i;
	}

	private:
	const std::vector<int16_t> mValues;
	size_t mI = 0;
	};

	std::vector<float> generateSinePattern(uint32_t sampleRateHz,
	double freq,
	double amp) {
	std::vector<float> result(3 * sampleRateHz / freq + .5);

	for (size_t i = 0; i < result.size(); ++i) {
	const double a = double(i) * M_PI * 2 / sampleRateHz;
	result[i] = amp * sin(a * freq);
	}

	return result;
	}

	template <class G> std::unique_ptr<GeneratedSource<G>>
	createGeneratedSource(const AudioConfig &cfg,
	size_t writerBufferSizeHint,
	uint64_t &frames,
	G generator) {
	return std::make_unique<GeneratedSource<G>>(cfg,
	writerBufferSizeHint,
	frames,
	std::move(generator));
	}

	} // namespace

	std::unique_ptr<DevicePortSource>
	DevicePortSource::create(size_t writerBufferSizeHint,
	const DeviceAddress &address,
	const AudioConfig &cfg,
	const hidl_vec<AudioInOutFlag> &flags,
	uint64_t &frames) {
	(void)flags;

	if (xsd::stringToAudioFormat(cfg.base.format) != xsd::AudioFormat::AUDIO_FORMAT_PCM_16_BIT) {
	ALOGE("%s:%d, unexpected format: '%s'", __func__, __LINE__, cfg.base.format.c_str());
	return FAILURE(nullptr);
	}

	switch (xsd::stringToAudioDevice(address.deviceType)) {
	case xsd::AudioDevice::AUDIO_DEVICE_IN_DEFAULT:
	case xsd::AudioDevice::AUDIO_DEVICE_IN_BUILTIN_MIC:
	{
	auto sourceptr = TinyalsaSource::create(talsa::kPcmCard, talsa::kPcmDevice,
	cfg, writerBufferSizeHint, frames);
	if (sourceptr != nullptr) {
	return sourceptr;
	} else {
	ALOGW("%s:%d failed to create alsa source for '%s'; creating a tone source instead.",
	__func__, __LINE__, address.deviceType.c_str());
	}
	}
	break;

	case xsd::AudioDevice::AUDIO_DEVICE_IN_TELEPHONY_RX:
	return createGeneratedSource(cfg, writerBufferSizeHint, frames,
	BusySignalGenerator(cfg.base.sampleRateHz));

	case xsd::AudioDevice::AUDIO_DEVICE_IN_FM_TUNER:
	return createGeneratedSource(
	cfg, writerBufferSizeHint, frames,
	RepeatGenerator(generateSinePattern(cfg.base.sampleRateHz, 440.0, 1.0)));

	default:
	ALOGW("%s:%d unsupported device: '%s', creating a tone source",
	__func__, __LINE__, address.deviceType.c_str());
	break;
	}

	return createGeneratedSource(
	cfg, writerBufferSizeHint, frames,
	RepeatGenerator(generateSinePattern(cfg.base.sampleRateHz, 220.0, 1.0)));
	}

	bool DevicePortSource::validateDeviceAddress(const DeviceAddress& address) {
	switch (xsd::stringToAudioDevice(address.deviceType)) {
	default:
	ALOGW("%s:%d unsupported device: '%s'", __func__, __LINE__, address.deviceType.c_str());
	return FAILURE(false);

	case xsd::AudioDevice::AUDIO_DEVICE_IN_DEFAULT:
	case xsd::AudioDevice::AUDIO_DEVICE_IN_BUILTIN_MIC:
	case xsd::AudioDevice::AUDIO_DEVICE_IN_TELEPHONY_RX:
	case xsd::AudioDevice::AUDIO_DEVICE_IN_FM_TUNER:
	case xsd::AudioDevice::AUDIO_DEVICE_IN_BUS:
	break;
	}

	return true;
	}

	} // namespace implementation
	} // namespace V7_0
	} // namespace audio
	} // namespace hardware
	} // namespace android