1459 lines
42 KiB
C++
Raw Normal View History

/* //device/extlibs/pv/android/AudioTrack.cpp
**
** Copyright 2007, 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.
*/
//#define LOG_NDEBUG 0
#define LOG_TAG "AudioTrack"
#include <stdint.h>
#include <sys/types.h>
#include <limits.h>
#include <sched.h>
#include <sys/resource.h>
#include <private/media/AudioTrackShared.h>
#include <media/AudioSystem.h>
#include <media/AudioTrack.h>
#include <utils/Log.h>
#include <binder/Parcel.h>
#include <binder/IPCThreadState.h>
#include <utils/Timers.h>
#include <utils/Atomic.h>
#include <cutils/bitops.h>
#include <system/audio.h>
#include <system/audio_policy.h>
#define LIKELY( exp ) (__builtin_expect( (exp) != 0, true ))
#define UNLIKELY( exp ) (__builtin_expect( (exp) != 0, false ))
namespace android {
// ---------------------------------------------------------------------------
// static
status_t AudioTrack::getMinFrameCount(
int* frameCount,
int streamType,
uint32_t sampleRate)
{
int afSampleRate;
if (AudioSystem::getOutputSamplingRate(&afSampleRate, streamType) != NO_ERROR) {
return NO_INIT;
}
int afFrameCount;
if (AudioSystem::getOutputFrameCount(&afFrameCount, streamType) != NO_ERROR) {
return NO_INIT;
}
uint32_t afLatency;
if (AudioSystem::getOutputLatency(&afLatency, streamType) != NO_ERROR) {
return NO_INIT;
}
// Ensure that buffer depth covers at least audio hardware latency
uint32_t minBufCount = afLatency / ((1000 * afFrameCount) / afSampleRate);
if (minBufCount < 2) minBufCount = 2;
*frameCount = (sampleRate == 0) ? afFrameCount * minBufCount :
afFrameCount * minBufCount * sampleRate / afSampleRate;
return NO_ERROR;
}
// ---------------------------------------------------------------------------
AudioTrack::AudioTrack()
: mStatus(NO_INIT)
{
}
AudioTrack::AudioTrack(
int streamType,
uint32_t sampleRate,
int format,
int channelMask,
int frameCount,
uint32_t flags,
callback_t cbf,
void* user,
Issue 2667801: [Audio Effect Framework] AudioFlinger, AudioMixer AudioTrack modifications. First drop of audio framework modifications for audio effects support. - AudioTrack/AudioRecord: Added support for auxiliary effects in AudioTrack Added support for audio sessions Fixed left right channel inversion in setVolume() - IAudioFlinger: Added interface methods for effect enumeraiton and instantiation Added support for audio sessions. - IAudioTrack: Added method to attach auxiliary effect. - AudioFlinger Created new classes to control effect engines in effect library and manage effect connections to tracks or output mix: EffectModule: wrapper object controlling the effect engine implementation in the effect library. There is one EffectModule per instance of an effect in a given audio session EffectChain: group of effects associated to one audio session. There is one EffectChain per audio session. EffectChain for session 0 is for output mix effects, other chains are attached to audio tracks with same session ID. Each chain contains a variable number of EffectModules EffectHandle: implements the IEffect interface. There is one EffectHandle object for each application controlling (or using) an effect module. THe EffectModule maintians a list of EffectHandles. Added support for effect modules and effect chains creation in PlaybackThread. modified mixer thread loop to allow track volume control by effect modules and call effect processing. -AudioMixer Each track now specifies its output buffer used by mixer for accumulation Modified mixer process functions to process tracks by groups of tracks with same buffer Modified track process functions to support accumulation to auxiliary channel Change-Id: I26d5f7c9e070a89bdd383e1a659f8b7ca150379c
2010-06-01 23:49:17 -07:00
int notificationFrames,
int sessionId)
: mStatus(NO_INIT)
{
mStatus = set(streamType, sampleRate, format, channelMask,
frameCount, flags, cbf, user, notificationFrames,
0, false, sessionId);
}
AudioTrack::AudioTrack(
int streamType,
uint32_t sampleRate,
int format,
int channelMask,
const sp<IMemory>& sharedBuffer,
uint32_t flags,
callback_t cbf,
void* user,
Issue 2667801: [Audio Effect Framework] AudioFlinger, AudioMixer AudioTrack modifications. First drop of audio framework modifications for audio effects support. - AudioTrack/AudioRecord: Added support for auxiliary effects in AudioTrack Added support for audio sessions Fixed left right channel inversion in setVolume() - IAudioFlinger: Added interface methods for effect enumeraiton and instantiation Added support for audio sessions. - IAudioTrack: Added method to attach auxiliary effect. - AudioFlinger Created new classes to control effect engines in effect library and manage effect connections to tracks or output mix: EffectModule: wrapper object controlling the effect engine implementation in the effect library. There is one EffectModule per instance of an effect in a given audio session EffectChain: group of effects associated to one audio session. There is one EffectChain per audio session. EffectChain for session 0 is for output mix effects, other chains are attached to audio tracks with same session ID. Each chain contains a variable number of EffectModules EffectHandle: implements the IEffect interface. There is one EffectHandle object for each application controlling (or using) an effect module. THe EffectModule maintians a list of EffectHandles. Added support for effect modules and effect chains creation in PlaybackThread. modified mixer thread loop to allow track volume control by effect modules and call effect processing. -AudioMixer Each track now specifies its output buffer used by mixer for accumulation Modified mixer process functions to process tracks by groups of tracks with same buffer Modified track process functions to support accumulation to auxiliary channel Change-Id: I26d5f7c9e070a89bdd383e1a659f8b7ca150379c
2010-06-01 23:49:17 -07:00
int notificationFrames,
int sessionId)
: mStatus(NO_INIT)
{
mStatus = set(streamType, sampleRate, format, channelMask,
0, flags, cbf, user, notificationFrames,
sharedBuffer, false, sessionId);
}
AudioTrack::~AudioTrack()
{
LOGV_IF(mSharedBuffer != 0, "Destructor sharedBuffer: %p", mSharedBuffer->pointer());
if (mStatus == NO_ERROR) {
// Make sure that callback function exits in the case where
// it is looping on buffer full condition in obtainBuffer().
// Otherwise the callback thread will never exit.
stop();
if (mAudioTrackThread != 0) {
mAudioTrackThread->requestExitAndWait();
mAudioTrackThread.clear();
}
mAudioTrack.clear();
IPCThreadState::self()->flushCommands();
AudioSystem::releaseAudioSessionId(mSessionId);
}
}
status_t AudioTrack::set(
int streamType,
uint32_t sampleRate,
int format,
int channelMask,
int frameCount,
uint32_t flags,
callback_t cbf,
void* user,
int notificationFrames,
const sp<IMemory>& sharedBuffer,
Issue 2667801: [Audio Effect Framework] AudioFlinger, AudioMixer AudioTrack modifications. First drop of audio framework modifications for audio effects support. - AudioTrack/AudioRecord: Added support for auxiliary effects in AudioTrack Added support for audio sessions Fixed left right channel inversion in setVolume() - IAudioFlinger: Added interface methods for effect enumeraiton and instantiation Added support for audio sessions. - IAudioTrack: Added method to attach auxiliary effect. - AudioFlinger Created new classes to control effect engines in effect library and manage effect connections to tracks or output mix: EffectModule: wrapper object controlling the effect engine implementation in the effect library. There is one EffectModule per instance of an effect in a given audio session EffectChain: group of effects associated to one audio session. There is one EffectChain per audio session. EffectChain for session 0 is for output mix effects, other chains are attached to audio tracks with same session ID. Each chain contains a variable number of EffectModules EffectHandle: implements the IEffect interface. There is one EffectHandle object for each application controlling (or using) an effect module. THe EffectModule maintians a list of EffectHandles. Added support for effect modules and effect chains creation in PlaybackThread. modified mixer thread loop to allow track volume control by effect modules and call effect processing. -AudioMixer Each track now specifies its output buffer used by mixer for accumulation Modified mixer process functions to process tracks by groups of tracks with same buffer Modified track process functions to support accumulation to auxiliary channel Change-Id: I26d5f7c9e070a89bdd383e1a659f8b7ca150379c
2010-06-01 23:49:17 -07:00
bool threadCanCallJava,
int sessionId)
{
LOGV_IF(sharedBuffer != 0, "sharedBuffer: %p, size: %d", sharedBuffer->pointer(), sharedBuffer->size());
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
AutoMutex lock(mLock);
if (mAudioTrack != 0) {
LOGE("Track already in use");
return INVALID_OPERATION;
}
int afSampleRate;
if (AudioSystem::getOutputSamplingRate(&afSampleRate, streamType) != NO_ERROR) {
return NO_INIT;
}
uint32_t afLatency;
if (AudioSystem::getOutputLatency(&afLatency, streamType) != NO_ERROR) {
return NO_INIT;
}
// handle default values first.
if (streamType == AUDIO_STREAM_DEFAULT) {
streamType = AUDIO_STREAM_MUSIC;
}
if (sampleRate == 0) {
sampleRate = afSampleRate;
}
// these below should probably come from the audioFlinger too...
if (format == 0) {
format = AUDIO_FORMAT_PCM_16_BIT;
}
if (channelMask == 0) {
channelMask = AUDIO_CHANNEL_OUT_STEREO;
}
// validate parameters
if (!audio_is_valid_format(format)) {
LOGE("Invalid format");
return BAD_VALUE;
}
// force direct flag if format is not linear PCM
if (!audio_is_linear_pcm(format)) {
flags |= AUDIO_POLICY_OUTPUT_FLAG_DIRECT;
}
if (!audio_is_output_channel(channelMask)) {
LOGE("Invalid channel mask");
return BAD_VALUE;
}
uint32_t channelCount = popcount(channelMask);
audio_io_handle_t output = AudioSystem::getOutput(
(audio_stream_type_t)streamType,
sampleRate,format, channelMask,
(audio_policy_output_flags_t)flags);
if (output == 0) {
LOGE("Could not get audio output for stream type %d", streamType);
return BAD_VALUE;
}
mVolume[LEFT] = 1.0f;
mVolume[RIGHT] = 1.0f;
Issue 2667801: [Audio Effect Framework] AudioFlinger, AudioMixer AudioTrack modifications. First drop of audio framework modifications for audio effects support. - AudioTrack/AudioRecord: Added support for auxiliary effects in AudioTrack Added support for audio sessions Fixed left right channel inversion in setVolume() - IAudioFlinger: Added interface methods for effect enumeraiton and instantiation Added support for audio sessions. - IAudioTrack: Added method to attach auxiliary effect. - AudioFlinger Created new classes to control effect engines in effect library and manage effect connections to tracks or output mix: EffectModule: wrapper object controlling the effect engine implementation in the effect library. There is one EffectModule per instance of an effect in a given audio session EffectChain: group of effects associated to one audio session. There is one EffectChain per audio session. EffectChain for session 0 is for output mix effects, other chains are attached to audio tracks with same session ID. Each chain contains a variable number of EffectModules EffectHandle: implements the IEffect interface. There is one EffectHandle object for each application controlling (or using) an effect module. THe EffectModule maintians a list of EffectHandles. Added support for effect modules and effect chains creation in PlaybackThread. modified mixer thread loop to allow track volume control by effect modules and call effect processing. -AudioMixer Each track now specifies its output buffer used by mixer for accumulation Modified mixer process functions to process tracks by groups of tracks with same buffer Modified track process functions to support accumulation to auxiliary channel Change-Id: I26d5f7c9e070a89bdd383e1a659f8b7ca150379c
2010-06-01 23:49:17 -07:00
mSendLevel = 0;
Fix issue 2553359: Pandora does not work well with Passion deskdock / Cardock. The problem is due to a too big difference between the buffer size used at the hardware interface and at the A2DP interface. When no resampling occurs we don't notice problems but the timing is very tight. As soon as resampling is activated, the AudioTrack underruns. This is because the AudioTrack buffers are not resized when moving the AudioTrack from hardware to A2DP output. The AudioTrack buffers are calculated based on a hardware output buffer size of 3072 bytes. Which is much less than the A2DP output buffer size (10240). The solution consists in creating new tracks with new buffers in AudioFlinger when the A2DP output is opened instead of just transfering active tracks from hardware output mixer thread to the new A2DP output mixer thread. To avoid synchronization issues between mixer threads and client processes, this is done by invalidating tracks by setting a flag in their control block and having AudioTrack release the handle on this track (IAudioTrack) and create a new IAudioTrack when this flag is detected next time obtainBuffer() or start() is executed. AudioFlinger modifications: - invalidate the tracks when setStreamOutput() is called - make sure that notifications of output opening/closing and change of stream type to output mapping are sent synchronously to client process. This is necessary so that AudioSystem has the new stream to output mapping when the AudioTrack detects the invalidate flag in the client process. Previously their were sent when the corresponding thread loop was executed. AudioTrack modifications: - move frame count calculation and verification from set() to createTrack() so that is is updated every time a new IAudioTrack is created. - detect track invalidate flag in obtainBuffer() and start() and create a new IAudioTrack. AudioTrackShared modifications - group all flags (out, flowControlFlag, forceReady...) into a single bit filed to save space. Change-Id: I9ac26b6192230627d35084e1449640caaf7d56ee
2010-05-14 03:26:45 -07:00
mFrameCount = frameCount;
mNotificationFramesReq = notificationFrames;
Issue 2667801: [Audio Effect Framework] AudioFlinger, AudioMixer AudioTrack modifications. First drop of audio framework modifications for audio effects support. - AudioTrack/AudioRecord: Added support for auxiliary effects in AudioTrack Added support for audio sessions Fixed left right channel inversion in setVolume() - IAudioFlinger: Added interface methods for effect enumeraiton and instantiation Added support for audio sessions. - IAudioTrack: Added method to attach auxiliary effect. - AudioFlinger Created new classes to control effect engines in effect library and manage effect connections to tracks or output mix: EffectModule: wrapper object controlling the effect engine implementation in the effect library. There is one EffectModule per instance of an effect in a given audio session EffectChain: group of effects associated to one audio session. There is one EffectChain per audio session. EffectChain for session 0 is for output mix effects, other chains are attached to audio tracks with same session ID. Each chain contains a variable number of EffectModules EffectHandle: implements the IEffect interface. There is one EffectHandle object for each application controlling (or using) an effect module. THe EffectModule maintians a list of EffectHandles. Added support for effect modules and effect chains creation in PlaybackThread. modified mixer thread loop to allow track volume control by effect modules and call effect processing. -AudioMixer Each track now specifies its output buffer used by mixer for accumulation Modified mixer process functions to process tracks by groups of tracks with same buffer Modified track process functions to support accumulation to auxiliary channel Change-Id: I26d5f7c9e070a89bdd383e1a659f8b7ca150379c
2010-06-01 23:49:17 -07:00
mSessionId = sessionId;
mAuxEffectId = 0;
Issue 2667801: [Audio Effect Framework] AudioFlinger, AudioMixer AudioTrack modifications. First drop of audio framework modifications for audio effects support. - AudioTrack/AudioRecord: Added support for auxiliary effects in AudioTrack Added support for audio sessions Fixed left right channel inversion in setVolume() - IAudioFlinger: Added interface methods for effect enumeraiton and instantiation Added support for audio sessions. - IAudioTrack: Added method to attach auxiliary effect. - AudioFlinger Created new classes to control effect engines in effect library and manage effect connections to tracks or output mix: EffectModule: wrapper object controlling the effect engine implementation in the effect library. There is one EffectModule per instance of an effect in a given audio session EffectChain: group of effects associated to one audio session. There is one EffectChain per audio session. EffectChain for session 0 is for output mix effects, other chains are attached to audio tracks with same session ID. Each chain contains a variable number of EffectModules EffectHandle: implements the IEffect interface. There is one EffectHandle object for each application controlling (or using) an effect module. THe EffectModule maintians a list of EffectHandles. Added support for effect modules and effect chains creation in PlaybackThread. modified mixer thread loop to allow track volume control by effect modules and call effect processing. -AudioMixer Each track now specifies its output buffer used by mixer for accumulation Modified mixer process functions to process tracks by groups of tracks with same buffer Modified track process functions to support accumulation to auxiliary channel Change-Id: I26d5f7c9e070a89bdd383e1a659f8b7ca150379c
2010-06-01 23:49:17 -07:00
// create the IAudioTrack
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
status_t status = createTrack_l(streamType,
sampleRate,
(uint32_t)format,
(uint32_t)channelMask,
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
frameCount,
flags,
sharedBuffer,
output,
true);
if (status != NO_ERROR) {
return status;
}
if (cbf != 0) {
mAudioTrackThread = new AudioTrackThread(*this, threadCanCallJava);
if (mAudioTrackThread == 0) {
LOGE("Could not create callback thread");
return NO_INIT;
}
}
mStatus = NO_ERROR;
mStreamType = streamType;
mFormat = (uint32_t)format;
mChannelMask = (uint32_t)channelMask;
mChannelCount = channelCount;
mSharedBuffer = sharedBuffer;
mMuted = false;
mActive = 0;
mCbf = cbf;
mUserData = user;
mLoopCount = 0;
mMarkerPosition = 0;
mMarkerReached = false;
mNewPosition = 0;
mUpdatePeriod = 0;
mFlushed = false;
mFlags = flags;
AudioSystem::acquireAudioSessionId(mSessionId);
return NO_ERROR;
}
status_t AudioTrack::initCheck() const
{
return mStatus;
}
// -------------------------------------------------------------------------
uint32_t AudioTrack::latency() const
{
return mLatency;
}
int AudioTrack::streamType() const
{
return mStreamType;
}
int AudioTrack::format() const
{
return mFormat;
}
int AudioTrack::channelCount() const
{
return mChannelCount;
}
uint32_t AudioTrack::frameCount() const
{
Fix issue 2553359: Pandora does not work well with Passion deskdock / Cardock. The problem is due to a too big difference between the buffer size used at the hardware interface and at the A2DP interface. When no resampling occurs we don't notice problems but the timing is very tight. As soon as resampling is activated, the AudioTrack underruns. This is because the AudioTrack buffers are not resized when moving the AudioTrack from hardware to A2DP output. The AudioTrack buffers are calculated based on a hardware output buffer size of 3072 bytes. Which is much less than the A2DP output buffer size (10240). The solution consists in creating new tracks with new buffers in AudioFlinger when the A2DP output is opened instead of just transfering active tracks from hardware output mixer thread to the new A2DP output mixer thread. To avoid synchronization issues between mixer threads and client processes, this is done by invalidating tracks by setting a flag in their control block and having AudioTrack release the handle on this track (IAudioTrack) and create a new IAudioTrack when this flag is detected next time obtainBuffer() or start() is executed. AudioFlinger modifications: - invalidate the tracks when setStreamOutput() is called - make sure that notifications of output opening/closing and change of stream type to output mapping are sent synchronously to client process. This is necessary so that AudioSystem has the new stream to output mapping when the AudioTrack detects the invalidate flag in the client process. Previously their were sent when the corresponding thread loop was executed. AudioTrack modifications: - move frame count calculation and verification from set() to createTrack() so that is is updated every time a new IAudioTrack is created. - detect track invalidate flag in obtainBuffer() and start() and create a new IAudioTrack. AudioTrackShared modifications - group all flags (out, flowControlFlag, forceReady...) into a single bit filed to save space. Change-Id: I9ac26b6192230627d35084e1449640caaf7d56ee
2010-05-14 03:26:45 -07:00
return mCblk->frameCount;
}
int AudioTrack::frameSize() const
{
if (audio_is_linear_pcm(mFormat)) {
return channelCount()*audio_bytes_per_sample(mFormat);
} else {
return sizeof(uint8_t);
}
}
sp<IMemory>& AudioTrack::sharedBuffer()
{
return mSharedBuffer;
}
// -------------------------------------------------------------------------
void AudioTrack::start()
{
sp<AudioTrackThread> t = mAudioTrackThread;
status_t status = NO_ERROR;
LOGV("start %p", this);
if (t != 0) {
if (t->exitPending()) {
if (t->requestExitAndWait() == WOULD_BLOCK) {
LOGE("AudioTrack::start called from thread");
return;
}
}
t->mLock.lock();
}
AutoMutex lock(mLock);
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
// acquire a strong reference on the IMemory and IAudioTrack so that they cannot be destroyed
// while we are accessing the cblk
sp <IAudioTrack> audioTrack = mAudioTrack;
sp <IMemory> iMem = mCblkMemory;
audio_track_cblk_t* cblk = mCblk;
if (mActive == 0) {
mFlushed = false;
mActive = 1;
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
mNewPosition = cblk->server + mUpdatePeriod;
cblk->lock.lock();
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
cblk->bufferTimeoutMs = MAX_STARTUP_TIMEOUT_MS;
cblk->waitTimeMs = 0;
android_atomic_and(~CBLK_DISABLED_ON, &cblk->flags);
if (t != 0) {
t->run("AudioTrackThread", ANDROID_PRIORITY_AUDIO);
} else {
setpriority(PRIO_PROCESS, 0, ANDROID_PRIORITY_AUDIO);
}
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
LOGV("start %p before lock cblk %p", this, mCblk);
if (!(cblk->flags & CBLK_INVALID_MSK)) {
cblk->lock.unlock();
Fix issue 2553359: Pandora does not work well with Passion deskdock / Cardock. The problem is due to a too big difference between the buffer size used at the hardware interface and at the A2DP interface. When no resampling occurs we don't notice problems but the timing is very tight. As soon as resampling is activated, the AudioTrack underruns. This is because the AudioTrack buffers are not resized when moving the AudioTrack from hardware to A2DP output. The AudioTrack buffers are calculated based on a hardware output buffer size of 3072 bytes. Which is much less than the A2DP output buffer size (10240). The solution consists in creating new tracks with new buffers in AudioFlinger when the A2DP output is opened instead of just transfering active tracks from hardware output mixer thread to the new A2DP output mixer thread. To avoid synchronization issues between mixer threads and client processes, this is done by invalidating tracks by setting a flag in their control block and having AudioTrack release the handle on this track (IAudioTrack) and create a new IAudioTrack when this flag is detected next time obtainBuffer() or start() is executed. AudioFlinger modifications: - invalidate the tracks when setStreamOutput() is called - make sure that notifications of output opening/closing and change of stream type to output mapping are sent synchronously to client process. This is necessary so that AudioSystem has the new stream to output mapping when the AudioTrack detects the invalidate flag in the client process. Previously their were sent when the corresponding thread loop was executed. AudioTrack modifications: - move frame count calculation and verification from set() to createTrack() so that is is updated every time a new IAudioTrack is created. - detect track invalidate flag in obtainBuffer() and start() and create a new IAudioTrack. AudioTrackShared modifications - group all flags (out, flowControlFlag, forceReady...) into a single bit filed to save space. Change-Id: I9ac26b6192230627d35084e1449640caaf7d56ee
2010-05-14 03:26:45 -07:00
status = mAudioTrack->start();
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
cblk->lock.lock();
if (status == DEAD_OBJECT) {
android_atomic_or(CBLK_INVALID_ON, &cblk->flags);
}
}
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
if (cblk->flags & CBLK_INVALID_MSK) {
status = restoreTrack_l(cblk, true);
}
cblk->lock.unlock();
if (status != NO_ERROR) {
LOGV("start() failed");
mActive = 0;
if (t != 0) {
t->requestExit();
} else {
setpriority(PRIO_PROCESS, 0, ANDROID_PRIORITY_NORMAL);
}
}
}
if (t != 0) {
t->mLock.unlock();
}
}
void AudioTrack::stop()
{
sp<AudioTrackThread> t = mAudioTrackThread;
LOGV("stop %p", this);
if (t != 0) {
t->mLock.lock();
}
AutoMutex lock(mLock);
if (mActive == 1) {
mActive = 0;
mCblk->cv.signal();
mAudioTrack->stop();
// Cancel loops (If we are in the middle of a loop, playback
// would not stop until loopCount reaches 0).
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
setLoop_l(0, 0, 0);
// the playback head position will reset to 0, so if a marker is set, we need
// to activate it again
mMarkerReached = false;
// Force flush if a shared buffer is used otherwise audioflinger
// will not stop before end of buffer is reached.
if (mSharedBuffer != 0) {
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
flush_l();
}
if (t != 0) {
t->requestExit();
} else {
setpriority(PRIO_PROCESS, 0, ANDROID_PRIORITY_NORMAL);
}
}
if (t != 0) {
t->mLock.unlock();
}
}
bool AudioTrack::stopped() const
{
return !mActive;
}
void AudioTrack::flush()
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
{
AutoMutex lock(mLock);
flush_l();
}
// must be called with mLock held
void AudioTrack::flush_l()
{
LOGV("flush");
// clear playback marker and periodic update counter
mMarkerPosition = 0;
mMarkerReached = false;
mUpdatePeriod = 0;
if (!mActive) {
mFlushed = true;
mAudioTrack->flush();
// Release AudioTrack callback thread in case it was waiting for new buffers
// in AudioTrack::obtainBuffer()
mCblk->cv.signal();
}
}
void AudioTrack::pause()
{
LOGV("pause");
AutoMutex lock(mLock);
if (mActive == 1) {
mActive = 0;
mAudioTrack->pause();
}
}
void AudioTrack::mute(bool e)
{
mAudioTrack->mute(e);
mMuted = e;
}
bool AudioTrack::muted() const
{
return mMuted;
}
Issue 2667801: [Audio Effect Framework] AudioFlinger, AudioMixer AudioTrack modifications. First drop of audio framework modifications for audio effects support. - AudioTrack/AudioRecord: Added support for auxiliary effects in AudioTrack Added support for audio sessions Fixed left right channel inversion in setVolume() - IAudioFlinger: Added interface methods for effect enumeraiton and instantiation Added support for audio sessions. - IAudioTrack: Added method to attach auxiliary effect. - AudioFlinger Created new classes to control effect engines in effect library and manage effect connections to tracks or output mix: EffectModule: wrapper object controlling the effect engine implementation in the effect library. There is one EffectModule per instance of an effect in a given audio session EffectChain: group of effects associated to one audio session. There is one EffectChain per audio session. EffectChain for session 0 is for output mix effects, other chains are attached to audio tracks with same session ID. Each chain contains a variable number of EffectModules EffectHandle: implements the IEffect interface. There is one EffectHandle object for each application controlling (or using) an effect module. THe EffectModule maintians a list of EffectHandles. Added support for effect modules and effect chains creation in PlaybackThread. modified mixer thread loop to allow track volume control by effect modules and call effect processing. -AudioMixer Each track now specifies its output buffer used by mixer for accumulation Modified mixer process functions to process tracks by groups of tracks with same buffer Modified track process functions to support accumulation to auxiliary channel Change-Id: I26d5f7c9e070a89bdd383e1a659f8b7ca150379c
2010-06-01 23:49:17 -07:00
status_t AudioTrack::setVolume(float left, float right)
{
Issue 2667801: [Audio Effect Framework] AudioFlinger, AudioMixer AudioTrack modifications. First drop of audio framework modifications for audio effects support. - AudioTrack/AudioRecord: Added support for auxiliary effects in AudioTrack Added support for audio sessions Fixed left right channel inversion in setVolume() - IAudioFlinger: Added interface methods for effect enumeraiton and instantiation Added support for audio sessions. - IAudioTrack: Added method to attach auxiliary effect. - AudioFlinger Created new classes to control effect engines in effect library and manage effect connections to tracks or output mix: EffectModule: wrapper object controlling the effect engine implementation in the effect library. There is one EffectModule per instance of an effect in a given audio session EffectChain: group of effects associated to one audio session. There is one EffectChain per audio session. EffectChain for session 0 is for output mix effects, other chains are attached to audio tracks with same session ID. Each chain contains a variable number of EffectModules EffectHandle: implements the IEffect interface. There is one EffectHandle object for each application controlling (or using) an effect module. THe EffectModule maintians a list of EffectHandles. Added support for effect modules and effect chains creation in PlaybackThread. modified mixer thread loop to allow track volume control by effect modules and call effect processing. -AudioMixer Each track now specifies its output buffer used by mixer for accumulation Modified mixer process functions to process tracks by groups of tracks with same buffer Modified track process functions to support accumulation to auxiliary channel Change-Id: I26d5f7c9e070a89bdd383e1a659f8b7ca150379c
2010-06-01 23:49:17 -07:00
if (left > 1.0f || right > 1.0f) {
return BAD_VALUE;
}
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
AutoMutex lock(mLock);
mVolume[LEFT] = left;
mVolume[RIGHT] = right;
// write must be atomic
Issue 2667801: [Audio Effect Framework] AudioFlinger, AudioMixer AudioTrack modifications. First drop of audio framework modifications for audio effects support. - AudioTrack/AudioRecord: Added support for auxiliary effects in AudioTrack Added support for audio sessions Fixed left right channel inversion in setVolume() - IAudioFlinger: Added interface methods for effect enumeraiton and instantiation Added support for audio sessions. - IAudioTrack: Added method to attach auxiliary effect. - AudioFlinger Created new classes to control effect engines in effect library and manage effect connections to tracks or output mix: EffectModule: wrapper object controlling the effect engine implementation in the effect library. There is one EffectModule per instance of an effect in a given audio session EffectChain: group of effects associated to one audio session. There is one EffectChain per audio session. EffectChain for session 0 is for output mix effects, other chains are attached to audio tracks with same session ID. Each chain contains a variable number of EffectModules EffectHandle: implements the IEffect interface. There is one EffectHandle object for each application controlling (or using) an effect module. THe EffectModule maintians a list of EffectHandles. Added support for effect modules and effect chains creation in PlaybackThread. modified mixer thread loop to allow track volume control by effect modules and call effect processing. -AudioMixer Each track now specifies its output buffer used by mixer for accumulation Modified mixer process functions to process tracks by groups of tracks with same buffer Modified track process functions to support accumulation to auxiliary channel Change-Id: I26d5f7c9e070a89bdd383e1a659f8b7ca150379c
2010-06-01 23:49:17 -07:00
mCblk->volumeLR = (uint32_t(uint16_t(right * 0x1000)) << 16) | uint16_t(left * 0x1000);
return NO_ERROR;
}
void AudioTrack::getVolume(float* left, float* right)
{
Issue 2667801: [Audio Effect Framework] AudioFlinger, AudioMixer AudioTrack modifications. First drop of audio framework modifications for audio effects support. - AudioTrack/AudioRecord: Added support for auxiliary effects in AudioTrack Added support for audio sessions Fixed left right channel inversion in setVolume() - IAudioFlinger: Added interface methods for effect enumeraiton and instantiation Added support for audio sessions. - IAudioTrack: Added method to attach auxiliary effect. - AudioFlinger Created new classes to control effect engines in effect library and manage effect connections to tracks or output mix: EffectModule: wrapper object controlling the effect engine implementation in the effect library. There is one EffectModule per instance of an effect in a given audio session EffectChain: group of effects associated to one audio session. There is one EffectChain per audio session. EffectChain for session 0 is for output mix effects, other chains are attached to audio tracks with same session ID. Each chain contains a variable number of EffectModules EffectHandle: implements the IEffect interface. There is one EffectHandle object for each application controlling (or using) an effect module. THe EffectModule maintians a list of EffectHandles. Added support for effect modules and effect chains creation in PlaybackThread. modified mixer thread loop to allow track volume control by effect modules and call effect processing. -AudioMixer Each track now specifies its output buffer used by mixer for accumulation Modified mixer process functions to process tracks by groups of tracks with same buffer Modified track process functions to support accumulation to auxiliary channel Change-Id: I26d5f7c9e070a89bdd383e1a659f8b7ca150379c
2010-06-01 23:49:17 -07:00
if (left != NULL) {
*left = mVolume[LEFT];
}
if (right != NULL) {
*right = mVolume[RIGHT];
}
}
status_t AudioTrack::setAuxEffectSendLevel(float level)
Issue 2667801: [Audio Effect Framework] AudioFlinger, AudioMixer AudioTrack modifications. First drop of audio framework modifications for audio effects support. - AudioTrack/AudioRecord: Added support for auxiliary effects in AudioTrack Added support for audio sessions Fixed left right channel inversion in setVolume() - IAudioFlinger: Added interface methods for effect enumeraiton and instantiation Added support for audio sessions. - IAudioTrack: Added method to attach auxiliary effect. - AudioFlinger Created new classes to control effect engines in effect library and manage effect connections to tracks or output mix: EffectModule: wrapper object controlling the effect engine implementation in the effect library. There is one EffectModule per instance of an effect in a given audio session EffectChain: group of effects associated to one audio session. There is one EffectChain per audio session. EffectChain for session 0 is for output mix effects, other chains are attached to audio tracks with same session ID. Each chain contains a variable number of EffectModules EffectHandle: implements the IEffect interface. There is one EffectHandle object for each application controlling (or using) an effect module. THe EffectModule maintians a list of EffectHandles. Added support for effect modules and effect chains creation in PlaybackThread. modified mixer thread loop to allow track volume control by effect modules and call effect processing. -AudioMixer Each track now specifies its output buffer used by mixer for accumulation Modified mixer process functions to process tracks by groups of tracks with same buffer Modified track process functions to support accumulation to auxiliary channel Change-Id: I26d5f7c9e070a89bdd383e1a659f8b7ca150379c
2010-06-01 23:49:17 -07:00
{
LOGV("setAuxEffectSendLevel(%f)", level);
Issue 2667801: [Audio Effect Framework] AudioFlinger, AudioMixer AudioTrack modifications. First drop of audio framework modifications for audio effects support. - AudioTrack/AudioRecord: Added support for auxiliary effects in AudioTrack Added support for audio sessions Fixed left right channel inversion in setVolume() - IAudioFlinger: Added interface methods for effect enumeraiton and instantiation Added support for audio sessions. - IAudioTrack: Added method to attach auxiliary effect. - AudioFlinger Created new classes to control effect engines in effect library and manage effect connections to tracks or output mix: EffectModule: wrapper object controlling the effect engine implementation in the effect library. There is one EffectModule per instance of an effect in a given audio session EffectChain: group of effects associated to one audio session. There is one EffectChain per audio session. EffectChain for session 0 is for output mix effects, other chains are attached to audio tracks with same session ID. Each chain contains a variable number of EffectModules EffectHandle: implements the IEffect interface. There is one EffectHandle object for each application controlling (or using) an effect module. THe EffectModule maintians a list of EffectHandles. Added support for effect modules and effect chains creation in PlaybackThread. modified mixer thread loop to allow track volume control by effect modules and call effect processing. -AudioMixer Each track now specifies its output buffer used by mixer for accumulation Modified mixer process functions to process tracks by groups of tracks with same buffer Modified track process functions to support accumulation to auxiliary channel Change-Id: I26d5f7c9e070a89bdd383e1a659f8b7ca150379c
2010-06-01 23:49:17 -07:00
if (level > 1.0f) {
return BAD_VALUE;
}
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
AutoMutex lock(mLock);
Issue 2667801: [Audio Effect Framework] AudioFlinger, AudioMixer AudioTrack modifications. First drop of audio framework modifications for audio effects support. - AudioTrack/AudioRecord: Added support for auxiliary effects in AudioTrack Added support for audio sessions Fixed left right channel inversion in setVolume() - IAudioFlinger: Added interface methods for effect enumeraiton and instantiation Added support for audio sessions. - IAudioTrack: Added method to attach auxiliary effect. - AudioFlinger Created new classes to control effect engines in effect library and manage effect connections to tracks or output mix: EffectModule: wrapper object controlling the effect engine implementation in the effect library. There is one EffectModule per instance of an effect in a given audio session EffectChain: group of effects associated to one audio session. There is one EffectChain per audio session. EffectChain for session 0 is for output mix effects, other chains are attached to audio tracks with same session ID. Each chain contains a variable number of EffectModules EffectHandle: implements the IEffect interface. There is one EffectHandle object for each application controlling (or using) an effect module. THe EffectModule maintians a list of EffectHandles. Added support for effect modules and effect chains creation in PlaybackThread. modified mixer thread loop to allow track volume control by effect modules and call effect processing. -AudioMixer Each track now specifies its output buffer used by mixer for accumulation Modified mixer process functions to process tracks by groups of tracks with same buffer Modified track process functions to support accumulation to auxiliary channel Change-Id: I26d5f7c9e070a89bdd383e1a659f8b7ca150379c
2010-06-01 23:49:17 -07:00
mSendLevel = level;
mCblk->sendLevel = uint16_t(level * 0x1000);
return NO_ERROR;
}
void AudioTrack::getAuxEffectSendLevel(float* level)
Issue 2667801: [Audio Effect Framework] AudioFlinger, AudioMixer AudioTrack modifications. First drop of audio framework modifications for audio effects support. - AudioTrack/AudioRecord: Added support for auxiliary effects in AudioTrack Added support for audio sessions Fixed left right channel inversion in setVolume() - IAudioFlinger: Added interface methods for effect enumeraiton and instantiation Added support for audio sessions. - IAudioTrack: Added method to attach auxiliary effect. - AudioFlinger Created new classes to control effect engines in effect library and manage effect connections to tracks or output mix: EffectModule: wrapper object controlling the effect engine implementation in the effect library. There is one EffectModule per instance of an effect in a given audio session EffectChain: group of effects associated to one audio session. There is one EffectChain per audio session. EffectChain for session 0 is for output mix effects, other chains are attached to audio tracks with same session ID. Each chain contains a variable number of EffectModules EffectHandle: implements the IEffect interface. There is one EffectHandle object for each application controlling (or using) an effect module. THe EffectModule maintians a list of EffectHandles. Added support for effect modules and effect chains creation in PlaybackThread. modified mixer thread loop to allow track volume control by effect modules and call effect processing. -AudioMixer Each track now specifies its output buffer used by mixer for accumulation Modified mixer process functions to process tracks by groups of tracks with same buffer Modified track process functions to support accumulation to auxiliary channel Change-Id: I26d5f7c9e070a89bdd383e1a659f8b7ca150379c
2010-06-01 23:49:17 -07:00
{
if (level != NULL) {
*level = mSendLevel;
}
}
status_t AudioTrack::setSampleRate(int rate)
{
int afSamplingRate;
if (AudioSystem::getOutputSamplingRate(&afSamplingRate, mStreamType) != NO_ERROR) {
return NO_INIT;
}
// Resampler implementation limits input sampling rate to 2 x output sampling rate.
if (rate <= 0 || rate > afSamplingRate*2 ) return BAD_VALUE;
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
AutoMutex lock(mLock);
mCblk->sampleRate = rate;
return NO_ERROR;
}
uint32_t AudioTrack::getSampleRate()
{
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
AutoMutex lock(mLock);
return mCblk->sampleRate;
}
status_t AudioTrack::setLoop(uint32_t loopStart, uint32_t loopEnd, int loopCount)
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
{
AutoMutex lock(mLock);
return setLoop_l(loopStart, loopEnd, loopCount);
}
// must be called with mLock held
status_t AudioTrack::setLoop_l(uint32_t loopStart, uint32_t loopEnd, int loopCount)
{
audio_track_cblk_t* cblk = mCblk;
Mutex::Autolock _l(cblk->lock);
if (loopCount == 0) {
cblk->loopStart = UINT_MAX;
cblk->loopEnd = UINT_MAX;
cblk->loopCount = 0;
mLoopCount = 0;
return NO_ERROR;
}
if (loopStart >= loopEnd ||
loopEnd - loopStart > cblk->frameCount ||
cblk->server > loopStart) {
Fix issue 2553359: Pandora does not work well with Passion deskdock / Cardock. The problem is due to a too big difference between the buffer size used at the hardware interface and at the A2DP interface. When no resampling occurs we don't notice problems but the timing is very tight. As soon as resampling is activated, the AudioTrack underruns. This is because the AudioTrack buffers are not resized when moving the AudioTrack from hardware to A2DP output. The AudioTrack buffers are calculated based on a hardware output buffer size of 3072 bytes. Which is much less than the A2DP output buffer size (10240). The solution consists in creating new tracks with new buffers in AudioFlinger when the A2DP output is opened instead of just transfering active tracks from hardware output mixer thread to the new A2DP output mixer thread. To avoid synchronization issues between mixer threads and client processes, this is done by invalidating tracks by setting a flag in their control block and having AudioTrack release the handle on this track (IAudioTrack) and create a new IAudioTrack when this flag is detected next time obtainBuffer() or start() is executed. AudioFlinger modifications: - invalidate the tracks when setStreamOutput() is called - make sure that notifications of output opening/closing and change of stream type to output mapping are sent synchronously to client process. This is necessary so that AudioSystem has the new stream to output mapping when the AudioTrack detects the invalidate flag in the client process. Previously their were sent when the corresponding thread loop was executed. AudioTrack modifications: - move frame count calculation and verification from set() to createTrack() so that is is updated every time a new IAudioTrack is created. - detect track invalidate flag in obtainBuffer() and start() and create a new IAudioTrack. AudioTrackShared modifications - group all flags (out, flowControlFlag, forceReady...) into a single bit filed to save space. Change-Id: I9ac26b6192230627d35084e1449640caaf7d56ee
2010-05-14 03:26:45 -07:00
LOGE("setLoop invalid value: loopStart %d, loopEnd %d, loopCount %d, framecount %d, user %d", loopStart, loopEnd, loopCount, cblk->frameCount, cblk->user);
return BAD_VALUE;
}
if ((mSharedBuffer != 0) && (loopEnd > cblk->frameCount)) {
LOGE("setLoop invalid value: loop markers beyond data: loopStart %d, loopEnd %d, framecount %d",
Fix issue 2553359: Pandora does not work well with Passion deskdock / Cardock. The problem is due to a too big difference between the buffer size used at the hardware interface and at the A2DP interface. When no resampling occurs we don't notice problems but the timing is very tight. As soon as resampling is activated, the AudioTrack underruns. This is because the AudioTrack buffers are not resized when moving the AudioTrack from hardware to A2DP output. The AudioTrack buffers are calculated based on a hardware output buffer size of 3072 bytes. Which is much less than the A2DP output buffer size (10240). The solution consists in creating new tracks with new buffers in AudioFlinger when the A2DP output is opened instead of just transfering active tracks from hardware output mixer thread to the new A2DP output mixer thread. To avoid synchronization issues between mixer threads and client processes, this is done by invalidating tracks by setting a flag in their control block and having AudioTrack release the handle on this track (IAudioTrack) and create a new IAudioTrack when this flag is detected next time obtainBuffer() or start() is executed. AudioFlinger modifications: - invalidate the tracks when setStreamOutput() is called - make sure that notifications of output opening/closing and change of stream type to output mapping are sent synchronously to client process. This is necessary so that AudioSystem has the new stream to output mapping when the AudioTrack detects the invalidate flag in the client process. Previously their were sent when the corresponding thread loop was executed. AudioTrack modifications: - move frame count calculation and verification from set() to createTrack() so that is is updated every time a new IAudioTrack is created. - detect track invalidate flag in obtainBuffer() and start() and create a new IAudioTrack. AudioTrackShared modifications - group all flags (out, flowControlFlag, forceReady...) into a single bit filed to save space. Change-Id: I9ac26b6192230627d35084e1449640caaf7d56ee
2010-05-14 03:26:45 -07:00
loopStart, loopEnd, cblk->frameCount);
return BAD_VALUE;
}
cblk->loopStart = loopStart;
cblk->loopEnd = loopEnd;
cblk->loopCount = loopCount;
mLoopCount = loopCount;
return NO_ERROR;
}
status_t AudioTrack::getLoop(uint32_t *loopStart, uint32_t *loopEnd, int *loopCount)
{
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
AutoMutex lock(mLock);
if (loopStart != 0) {
*loopStart = mCblk->loopStart;
}
if (loopEnd != 0) {
*loopEnd = mCblk->loopEnd;
}
if (loopCount != 0) {
if (mCblk->loopCount < 0) {
*loopCount = -1;
} else {
*loopCount = mCblk->loopCount;
}
}
return NO_ERROR;
}
status_t AudioTrack::setMarkerPosition(uint32_t marker)
{
if (mCbf == 0) return INVALID_OPERATION;
mMarkerPosition = marker;
mMarkerReached = false;
return NO_ERROR;
}
status_t AudioTrack::getMarkerPosition(uint32_t *marker)
{
if (marker == 0) return BAD_VALUE;
*marker = mMarkerPosition;
return NO_ERROR;
}
status_t AudioTrack::setPositionUpdatePeriod(uint32_t updatePeriod)
{
if (mCbf == 0) return INVALID_OPERATION;
uint32_t curPosition;
getPosition(&curPosition);
mNewPosition = curPosition + updatePeriod;
mUpdatePeriod = updatePeriod;
return NO_ERROR;
}
status_t AudioTrack::getPositionUpdatePeriod(uint32_t *updatePeriod)
{
if (updatePeriod == 0) return BAD_VALUE;
*updatePeriod = mUpdatePeriod;
return NO_ERROR;
}
status_t AudioTrack::setPosition(uint32_t position)
{
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
AutoMutex lock(mLock);
Mutex::Autolock _l(mCblk->lock);
if (!stopped()) return INVALID_OPERATION;
if (position > mCblk->user) return BAD_VALUE;
mCblk->server = position;
android_atomic_or(CBLK_FORCEREADY_ON, &mCblk->flags);
return NO_ERROR;
}
status_t AudioTrack::getPosition(uint32_t *position)
{
if (position == 0) return BAD_VALUE;
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
AutoMutex lock(mLock);
*position = mFlushed ? 0 : mCblk->server;
return NO_ERROR;
}
status_t AudioTrack::reload()
{
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
AutoMutex lock(mLock);
if (!stopped()) return INVALID_OPERATION;
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
flush_l();
Fix issue 2553359: Pandora does not work well with Passion deskdock / Cardock. The problem is due to a too big difference between the buffer size used at the hardware interface and at the A2DP interface. When no resampling occurs we don't notice problems but the timing is very tight. As soon as resampling is activated, the AudioTrack underruns. This is because the AudioTrack buffers are not resized when moving the AudioTrack from hardware to A2DP output. The AudioTrack buffers are calculated based on a hardware output buffer size of 3072 bytes. Which is much less than the A2DP output buffer size (10240). The solution consists in creating new tracks with new buffers in AudioFlinger when the A2DP output is opened instead of just transfering active tracks from hardware output mixer thread to the new A2DP output mixer thread. To avoid synchronization issues between mixer threads and client processes, this is done by invalidating tracks by setting a flag in their control block and having AudioTrack release the handle on this track (IAudioTrack) and create a new IAudioTrack when this flag is detected next time obtainBuffer() or start() is executed. AudioFlinger modifications: - invalidate the tracks when setStreamOutput() is called - make sure that notifications of output opening/closing and change of stream type to output mapping are sent synchronously to client process. This is necessary so that AudioSystem has the new stream to output mapping when the AudioTrack detects the invalidate flag in the client process. Previously their were sent when the corresponding thread loop was executed. AudioTrack modifications: - move frame count calculation and verification from set() to createTrack() so that is is updated every time a new IAudioTrack is created. - detect track invalidate flag in obtainBuffer() and start() and create a new IAudioTrack. AudioTrackShared modifications - group all flags (out, flowControlFlag, forceReady...) into a single bit filed to save space. Change-Id: I9ac26b6192230627d35084e1449640caaf7d56ee
2010-05-14 03:26:45 -07:00
mCblk->stepUser(mCblk->frameCount);
return NO_ERROR;
}
audio_io_handle_t AudioTrack::getOutput()
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
{
AutoMutex lock(mLock);
return getOutput_l();
}
// must be called with mLock held
audio_io_handle_t AudioTrack::getOutput_l()
{
return AudioSystem::getOutput((audio_stream_type_t)mStreamType,
mCblk->sampleRate, mFormat, mChannelMask, (audio_policy_output_flags_t)mFlags);
}
Issue 2667801: [Audio Effect Framework] AudioFlinger, AudioMixer AudioTrack modifications. First drop of audio framework modifications for audio effects support. - AudioTrack/AudioRecord: Added support for auxiliary effects in AudioTrack Added support for audio sessions Fixed left right channel inversion in setVolume() - IAudioFlinger: Added interface methods for effect enumeraiton and instantiation Added support for audio sessions. - IAudioTrack: Added method to attach auxiliary effect. - AudioFlinger Created new classes to control effect engines in effect library and manage effect connections to tracks or output mix: EffectModule: wrapper object controlling the effect engine implementation in the effect library. There is one EffectModule per instance of an effect in a given audio session EffectChain: group of effects associated to one audio session. There is one EffectChain per audio session. EffectChain for session 0 is for output mix effects, other chains are attached to audio tracks with same session ID. Each chain contains a variable number of EffectModules EffectHandle: implements the IEffect interface. There is one EffectHandle object for each application controlling (or using) an effect module. THe EffectModule maintians a list of EffectHandles. Added support for effect modules and effect chains creation in PlaybackThread. modified mixer thread loop to allow track volume control by effect modules and call effect processing. -AudioMixer Each track now specifies its output buffer used by mixer for accumulation Modified mixer process functions to process tracks by groups of tracks with same buffer Modified track process functions to support accumulation to auxiliary channel Change-Id: I26d5f7c9e070a89bdd383e1a659f8b7ca150379c
2010-06-01 23:49:17 -07:00
int AudioTrack::getSessionId()
{
return mSessionId;
}
status_t AudioTrack::attachAuxEffect(int effectId)
{
LOGV("attachAuxEffect(%d)", effectId);
status_t status = mAudioTrack->attachAuxEffect(effectId);
if (status == NO_ERROR) {
mAuxEffectId = effectId;
}
return status;
Issue 2667801: [Audio Effect Framework] AudioFlinger, AudioMixer AudioTrack modifications. First drop of audio framework modifications for audio effects support. - AudioTrack/AudioRecord: Added support for auxiliary effects in AudioTrack Added support for audio sessions Fixed left right channel inversion in setVolume() - IAudioFlinger: Added interface methods for effect enumeraiton and instantiation Added support for audio sessions. - IAudioTrack: Added method to attach auxiliary effect. - AudioFlinger Created new classes to control effect engines in effect library and manage effect connections to tracks or output mix: EffectModule: wrapper object controlling the effect engine implementation in the effect library. There is one EffectModule per instance of an effect in a given audio session EffectChain: group of effects associated to one audio session. There is one EffectChain per audio session. EffectChain for session 0 is for output mix effects, other chains are attached to audio tracks with same session ID. Each chain contains a variable number of EffectModules EffectHandle: implements the IEffect interface. There is one EffectHandle object for each application controlling (or using) an effect module. THe EffectModule maintians a list of EffectHandles. Added support for effect modules and effect chains creation in PlaybackThread. modified mixer thread loop to allow track volume control by effect modules and call effect processing. -AudioMixer Each track now specifies its output buffer used by mixer for accumulation Modified mixer process functions to process tracks by groups of tracks with same buffer Modified track process functions to support accumulation to auxiliary channel Change-Id: I26d5f7c9e070a89bdd383e1a659f8b7ca150379c
2010-06-01 23:49:17 -07:00
}
// -------------------------------------------------------------------------
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
// must be called with mLock held
status_t AudioTrack::createTrack_l(
int streamType,
uint32_t sampleRate,
uint32_t format,
uint32_t channelMask,
int frameCount,
uint32_t flags,
const sp<IMemory>& sharedBuffer,
Fix issue 2553359: Pandora does not work well with Passion deskdock / Cardock. The problem is due to a too big difference between the buffer size used at the hardware interface and at the A2DP interface. When no resampling occurs we don't notice problems but the timing is very tight. As soon as resampling is activated, the AudioTrack underruns. This is because the AudioTrack buffers are not resized when moving the AudioTrack from hardware to A2DP output. The AudioTrack buffers are calculated based on a hardware output buffer size of 3072 bytes. Which is much less than the A2DP output buffer size (10240). The solution consists in creating new tracks with new buffers in AudioFlinger when the A2DP output is opened instead of just transfering active tracks from hardware output mixer thread to the new A2DP output mixer thread. To avoid synchronization issues between mixer threads and client processes, this is done by invalidating tracks by setting a flag in their control block and having AudioTrack release the handle on this track (IAudioTrack) and create a new IAudioTrack when this flag is detected next time obtainBuffer() or start() is executed. AudioFlinger modifications: - invalidate the tracks when setStreamOutput() is called - make sure that notifications of output opening/closing and change of stream type to output mapping are sent synchronously to client process. This is necessary so that AudioSystem has the new stream to output mapping when the AudioTrack detects the invalidate flag in the client process. Previously their were sent when the corresponding thread loop was executed. AudioTrack modifications: - move frame count calculation and verification from set() to createTrack() so that is is updated every time a new IAudioTrack is created. - detect track invalidate flag in obtainBuffer() and start() and create a new IAudioTrack. AudioTrackShared modifications - group all flags (out, flowControlFlag, forceReady...) into a single bit filed to save space. Change-Id: I9ac26b6192230627d35084e1449640caaf7d56ee
2010-05-14 03:26:45 -07:00
audio_io_handle_t output,
bool enforceFrameCount)
{
status_t status;
const sp<IAudioFlinger>& audioFlinger = AudioSystem::get_audio_flinger();
if (audioFlinger == 0) {
LOGE("Could not get audioflinger");
return NO_INIT;
}
Fix issue 2553359: Pandora does not work well with Passion deskdock / Cardock. The problem is due to a too big difference between the buffer size used at the hardware interface and at the A2DP interface. When no resampling occurs we don't notice problems but the timing is very tight. As soon as resampling is activated, the AudioTrack underruns. This is because the AudioTrack buffers are not resized when moving the AudioTrack from hardware to A2DP output. The AudioTrack buffers are calculated based on a hardware output buffer size of 3072 bytes. Which is much less than the A2DP output buffer size (10240). The solution consists in creating new tracks with new buffers in AudioFlinger when the A2DP output is opened instead of just transfering active tracks from hardware output mixer thread to the new A2DP output mixer thread. To avoid synchronization issues between mixer threads and client processes, this is done by invalidating tracks by setting a flag in their control block and having AudioTrack release the handle on this track (IAudioTrack) and create a new IAudioTrack when this flag is detected next time obtainBuffer() or start() is executed. AudioFlinger modifications: - invalidate the tracks when setStreamOutput() is called - make sure that notifications of output opening/closing and change of stream type to output mapping are sent synchronously to client process. This is necessary so that AudioSystem has the new stream to output mapping when the AudioTrack detects the invalidate flag in the client process. Previously their were sent when the corresponding thread loop was executed. AudioTrack modifications: - move frame count calculation and verification from set() to createTrack() so that is is updated every time a new IAudioTrack is created. - detect track invalidate flag in obtainBuffer() and start() and create a new IAudioTrack. AudioTrackShared modifications - group all flags (out, flowControlFlag, forceReady...) into a single bit filed to save space. Change-Id: I9ac26b6192230627d35084e1449640caaf7d56ee
2010-05-14 03:26:45 -07:00
int afSampleRate;
if (AudioSystem::getOutputSamplingRate(&afSampleRate, streamType) != NO_ERROR) {
return NO_INIT;
}
int afFrameCount;
if (AudioSystem::getOutputFrameCount(&afFrameCount, streamType) != NO_ERROR) {
return NO_INIT;
}
uint32_t afLatency;
if (AudioSystem::getOutputLatency(&afLatency, streamType) != NO_ERROR) {
return NO_INIT;
}
mNotificationFramesAct = mNotificationFramesReq;
if (!audio_is_linear_pcm(format)) {
Fix issue 2553359: Pandora does not work well with Passion deskdock / Cardock. The problem is due to a too big difference between the buffer size used at the hardware interface and at the A2DP interface. When no resampling occurs we don't notice problems but the timing is very tight. As soon as resampling is activated, the AudioTrack underruns. This is because the AudioTrack buffers are not resized when moving the AudioTrack from hardware to A2DP output. The AudioTrack buffers are calculated based on a hardware output buffer size of 3072 bytes. Which is much less than the A2DP output buffer size (10240). The solution consists in creating new tracks with new buffers in AudioFlinger when the A2DP output is opened instead of just transfering active tracks from hardware output mixer thread to the new A2DP output mixer thread. To avoid synchronization issues between mixer threads and client processes, this is done by invalidating tracks by setting a flag in their control block and having AudioTrack release the handle on this track (IAudioTrack) and create a new IAudioTrack when this flag is detected next time obtainBuffer() or start() is executed. AudioFlinger modifications: - invalidate the tracks when setStreamOutput() is called - make sure that notifications of output opening/closing and change of stream type to output mapping are sent synchronously to client process. This is necessary so that AudioSystem has the new stream to output mapping when the AudioTrack detects the invalidate flag in the client process. Previously their were sent when the corresponding thread loop was executed. AudioTrack modifications: - move frame count calculation and verification from set() to createTrack() so that is is updated every time a new IAudioTrack is created. - detect track invalidate flag in obtainBuffer() and start() and create a new IAudioTrack. AudioTrackShared modifications - group all flags (out, flowControlFlag, forceReady...) into a single bit filed to save space. Change-Id: I9ac26b6192230627d35084e1449640caaf7d56ee
2010-05-14 03:26:45 -07:00
if (sharedBuffer != 0) {
frameCount = sharedBuffer->size();
}
} else {
// Ensure that buffer depth covers at least audio hardware latency
uint32_t minBufCount = afLatency / ((1000 * afFrameCount)/afSampleRate);
if (minBufCount < 2) minBufCount = 2;
int minFrameCount = (afFrameCount*sampleRate*minBufCount)/afSampleRate;
if (sharedBuffer == 0) {
if (frameCount == 0) {
frameCount = minFrameCount;
}
if (mNotificationFramesAct == 0) {
mNotificationFramesAct = frameCount/2;
}
// Make sure that application is notified with sufficient margin
// before underrun
if (mNotificationFramesAct > (uint32_t)frameCount/2) {
mNotificationFramesAct = frameCount/2;
}
if (frameCount < minFrameCount) {
if (enforceFrameCount) {
LOGE("Invalid buffer size: minFrameCount %d, frameCount %d", minFrameCount, frameCount);
return BAD_VALUE;
} else {
frameCount = minFrameCount;
}
}
} else {
// Ensure that buffer alignment matches channelcount
int channelCount = popcount(channelMask);
Fix issue 2553359: Pandora does not work well with Passion deskdock / Cardock. The problem is due to a too big difference between the buffer size used at the hardware interface and at the A2DP interface. When no resampling occurs we don't notice problems but the timing is very tight. As soon as resampling is activated, the AudioTrack underruns. This is because the AudioTrack buffers are not resized when moving the AudioTrack from hardware to A2DP output. The AudioTrack buffers are calculated based on a hardware output buffer size of 3072 bytes. Which is much less than the A2DP output buffer size (10240). The solution consists in creating new tracks with new buffers in AudioFlinger when the A2DP output is opened instead of just transfering active tracks from hardware output mixer thread to the new A2DP output mixer thread. To avoid synchronization issues between mixer threads and client processes, this is done by invalidating tracks by setting a flag in their control block and having AudioTrack release the handle on this track (IAudioTrack) and create a new IAudioTrack when this flag is detected next time obtainBuffer() or start() is executed. AudioFlinger modifications: - invalidate the tracks when setStreamOutput() is called - make sure that notifications of output opening/closing and change of stream type to output mapping are sent synchronously to client process. This is necessary so that AudioSystem has the new stream to output mapping when the AudioTrack detects the invalidate flag in the client process. Previously their were sent when the corresponding thread loop was executed. AudioTrack modifications: - move frame count calculation and verification from set() to createTrack() so that is is updated every time a new IAudioTrack is created. - detect track invalidate flag in obtainBuffer() and start() and create a new IAudioTrack. AudioTrackShared modifications - group all flags (out, flowControlFlag, forceReady...) into a single bit filed to save space. Change-Id: I9ac26b6192230627d35084e1449640caaf7d56ee
2010-05-14 03:26:45 -07:00
if (((uint32_t)sharedBuffer->pointer() & (channelCount | 1)) != 0) {
LOGE("Invalid buffer alignement: address %p, channelCount %d", sharedBuffer->pointer(), channelCount);
return BAD_VALUE;
}
frameCount = sharedBuffer->size()/channelCount/sizeof(int16_t);
}
}
sp<IAudioTrack> track = audioFlinger->createTrack(getpid(),
streamType,
sampleRate,
format,
channelMask,
frameCount,
((uint16_t)flags) << 16,
sharedBuffer,
output,
Issue 2667801: [Audio Effect Framework] AudioFlinger, AudioMixer AudioTrack modifications. First drop of audio framework modifications for audio effects support. - AudioTrack/AudioRecord: Added support for auxiliary effects in AudioTrack Added support for audio sessions Fixed left right channel inversion in setVolume() - IAudioFlinger: Added interface methods for effect enumeraiton and instantiation Added support for audio sessions. - IAudioTrack: Added method to attach auxiliary effect. - AudioFlinger Created new classes to control effect engines in effect library and manage effect connections to tracks or output mix: EffectModule: wrapper object controlling the effect engine implementation in the effect library. There is one EffectModule per instance of an effect in a given audio session EffectChain: group of effects associated to one audio session. There is one EffectChain per audio session. EffectChain for session 0 is for output mix effects, other chains are attached to audio tracks with same session ID. Each chain contains a variable number of EffectModules EffectHandle: implements the IEffect interface. There is one EffectHandle object for each application controlling (or using) an effect module. THe EffectModule maintians a list of EffectHandles. Added support for effect modules and effect chains creation in PlaybackThread. modified mixer thread loop to allow track volume control by effect modules and call effect processing. -AudioMixer Each track now specifies its output buffer used by mixer for accumulation Modified mixer process functions to process tracks by groups of tracks with same buffer Modified track process functions to support accumulation to auxiliary channel Change-Id: I26d5f7c9e070a89bdd383e1a659f8b7ca150379c
2010-06-01 23:49:17 -07:00
&mSessionId,
&status);
if (track == 0) {
LOGE("AudioFlinger could not create track, status: %d", status);
return status;
}
sp<IMemory> cblk = track->getCblk();
if (cblk == 0) {
LOGE("Could not get control block");
return NO_INIT;
}
mAudioTrack.clear();
mAudioTrack = track;
mCblkMemory.clear();
mCblkMemory = cblk;
mCblk = static_cast<audio_track_cblk_t*>(cblk->pointer());
android_atomic_or(CBLK_DIRECTION_OUT, &mCblk->flags);
if (sharedBuffer == 0) {
mCblk->buffers = (char*)mCblk + sizeof(audio_track_cblk_t);
} else {
mCblk->buffers = sharedBuffer->pointer();
// Force buffer full condition as data is already present in shared memory
Fix issue 2553359: Pandora does not work well with Passion deskdock / Cardock. The problem is due to a too big difference between the buffer size used at the hardware interface and at the A2DP interface. When no resampling occurs we don't notice problems but the timing is very tight. As soon as resampling is activated, the AudioTrack underruns. This is because the AudioTrack buffers are not resized when moving the AudioTrack from hardware to A2DP output. The AudioTrack buffers are calculated based on a hardware output buffer size of 3072 bytes. Which is much less than the A2DP output buffer size (10240). The solution consists in creating new tracks with new buffers in AudioFlinger when the A2DP output is opened instead of just transfering active tracks from hardware output mixer thread to the new A2DP output mixer thread. To avoid synchronization issues between mixer threads and client processes, this is done by invalidating tracks by setting a flag in their control block and having AudioTrack release the handle on this track (IAudioTrack) and create a new IAudioTrack when this flag is detected next time obtainBuffer() or start() is executed. AudioFlinger modifications: - invalidate the tracks when setStreamOutput() is called - make sure that notifications of output opening/closing and change of stream type to output mapping are sent synchronously to client process. This is necessary so that AudioSystem has the new stream to output mapping when the AudioTrack detects the invalidate flag in the client process. Previously their were sent when the corresponding thread loop was executed. AudioTrack modifications: - move frame count calculation and verification from set() to createTrack() so that is is updated every time a new IAudioTrack is created. - detect track invalidate flag in obtainBuffer() and start() and create a new IAudioTrack. AudioTrackShared modifications - group all flags (out, flowControlFlag, forceReady...) into a single bit filed to save space. Change-Id: I9ac26b6192230627d35084e1449640caaf7d56ee
2010-05-14 03:26:45 -07:00
mCblk->stepUser(mCblk->frameCount);
}
Issue 2667801: [Audio Effect Framework] AudioFlinger, AudioMixer AudioTrack modifications. First drop of audio framework modifications for audio effects support. - AudioTrack/AudioRecord: Added support for auxiliary effects in AudioTrack Added support for audio sessions Fixed left right channel inversion in setVolume() - IAudioFlinger: Added interface methods for effect enumeraiton and instantiation Added support for audio sessions. - IAudioTrack: Added method to attach auxiliary effect. - AudioFlinger Created new classes to control effect engines in effect library and manage effect connections to tracks or output mix: EffectModule: wrapper object controlling the effect engine implementation in the effect library. There is one EffectModule per instance of an effect in a given audio session EffectChain: group of effects associated to one audio session. There is one EffectChain per audio session. EffectChain for session 0 is for output mix effects, other chains are attached to audio tracks with same session ID. Each chain contains a variable number of EffectModules EffectHandle: implements the IEffect interface. There is one EffectHandle object for each application controlling (or using) an effect module. THe EffectModule maintians a list of EffectHandles. Added support for effect modules and effect chains creation in PlaybackThread. modified mixer thread loop to allow track volume control by effect modules and call effect processing. -AudioMixer Each track now specifies its output buffer used by mixer for accumulation Modified mixer process functions to process tracks by groups of tracks with same buffer Modified track process functions to support accumulation to auxiliary channel Change-Id: I26d5f7c9e070a89bdd383e1a659f8b7ca150379c
2010-06-01 23:49:17 -07:00
mCblk->volumeLR = (uint32_t(uint16_t(mVolume[RIGHT] * 0x1000)) << 16) | uint16_t(mVolume[LEFT] * 0x1000);
mCblk->sendLevel = uint16_t(mSendLevel * 0x1000);
mAudioTrack->attachAuxEffect(mAuxEffectId);
mCblk->bufferTimeoutMs = MAX_STARTUP_TIMEOUT_MS;
mCblk->waitTimeMs = 0;
Fix issue 2553359: Pandora does not work well with Passion deskdock / Cardock. The problem is due to a too big difference between the buffer size used at the hardware interface and at the A2DP interface. When no resampling occurs we don't notice problems but the timing is very tight. As soon as resampling is activated, the AudioTrack underruns. This is because the AudioTrack buffers are not resized when moving the AudioTrack from hardware to A2DP output. The AudioTrack buffers are calculated based on a hardware output buffer size of 3072 bytes. Which is much less than the A2DP output buffer size (10240). The solution consists in creating new tracks with new buffers in AudioFlinger when the A2DP output is opened instead of just transfering active tracks from hardware output mixer thread to the new A2DP output mixer thread. To avoid synchronization issues between mixer threads and client processes, this is done by invalidating tracks by setting a flag in their control block and having AudioTrack release the handle on this track (IAudioTrack) and create a new IAudioTrack when this flag is detected next time obtainBuffer() or start() is executed. AudioFlinger modifications: - invalidate the tracks when setStreamOutput() is called - make sure that notifications of output opening/closing and change of stream type to output mapping are sent synchronously to client process. This is necessary so that AudioSystem has the new stream to output mapping when the AudioTrack detects the invalidate flag in the client process. Previously their were sent when the corresponding thread loop was executed. AudioTrack modifications: - move frame count calculation and verification from set() to createTrack() so that is is updated every time a new IAudioTrack is created. - detect track invalidate flag in obtainBuffer() and start() and create a new IAudioTrack. AudioTrackShared modifications - group all flags (out, flowControlFlag, forceReady...) into a single bit filed to save space. Change-Id: I9ac26b6192230627d35084e1449640caaf7d56ee
2010-05-14 03:26:45 -07:00
mRemainingFrames = mNotificationFramesAct;
mLatency = afLatency + (1000*mCblk->frameCount) / sampleRate;
return NO_ERROR;
}
status_t AudioTrack::obtainBuffer(Buffer* audioBuffer, int32_t waitCount)
{
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
AutoMutex lock(mLock);
int active;
status_t result = NO_ERROR;
audio_track_cblk_t* cblk = mCblk;
uint32_t framesReq = audioBuffer->frameCount;
uint32_t waitTimeMs = (waitCount < 0) ? cblk->bufferTimeoutMs : WAIT_PERIOD_MS;
audioBuffer->frameCount = 0;
audioBuffer->size = 0;
uint32_t framesAvail = cblk->framesAvailable();
cblk->lock.lock();
if (cblk->flags & CBLK_INVALID_MSK) {
goto create_new_track;
}
cblk->lock.unlock();
if (framesAvail == 0) {
cblk->lock.lock();
goto start_loop_here;
while (framesAvail == 0) {
active = mActive;
if (UNLIKELY(!active)) {
LOGV("Not active and NO_MORE_BUFFERS");
cblk->lock.unlock();
return NO_MORE_BUFFERS;
}
if (UNLIKELY(!waitCount)) {
cblk->lock.unlock();
return WOULD_BLOCK;
}
Fix issue 2553359: Pandora does not work well with Passion deskdock / Cardock. The problem is due to a too big difference between the buffer size used at the hardware interface and at the A2DP interface. When no resampling occurs we don't notice problems but the timing is very tight. As soon as resampling is activated, the AudioTrack underruns. This is because the AudioTrack buffers are not resized when moving the AudioTrack from hardware to A2DP output. The AudioTrack buffers are calculated based on a hardware output buffer size of 3072 bytes. Which is much less than the A2DP output buffer size (10240). The solution consists in creating new tracks with new buffers in AudioFlinger when the A2DP output is opened instead of just transfering active tracks from hardware output mixer thread to the new A2DP output mixer thread. To avoid synchronization issues between mixer threads and client processes, this is done by invalidating tracks by setting a flag in their control block and having AudioTrack release the handle on this track (IAudioTrack) and create a new IAudioTrack when this flag is detected next time obtainBuffer() or start() is executed. AudioFlinger modifications: - invalidate the tracks when setStreamOutput() is called - make sure that notifications of output opening/closing and change of stream type to output mapping are sent synchronously to client process. This is necessary so that AudioSystem has the new stream to output mapping when the AudioTrack detects the invalidate flag in the client process. Previously their were sent when the corresponding thread loop was executed. AudioTrack modifications: - move frame count calculation and verification from set() to createTrack() so that is is updated every time a new IAudioTrack is created. - detect track invalidate flag in obtainBuffer() and start() and create a new IAudioTrack. AudioTrackShared modifications - group all flags (out, flowControlFlag, forceReady...) into a single bit filed to save space. Change-Id: I9ac26b6192230627d35084e1449640caaf7d56ee
2010-05-14 03:26:45 -07:00
if (!(cblk->flags & CBLK_INVALID_MSK)) {
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
mLock.unlock();
Fix issue 2553359: Pandora does not work well with Passion deskdock / Cardock. The problem is due to a too big difference between the buffer size used at the hardware interface and at the A2DP interface. When no resampling occurs we don't notice problems but the timing is very tight. As soon as resampling is activated, the AudioTrack underruns. This is because the AudioTrack buffers are not resized when moving the AudioTrack from hardware to A2DP output. The AudioTrack buffers are calculated based on a hardware output buffer size of 3072 bytes. Which is much less than the A2DP output buffer size (10240). The solution consists in creating new tracks with new buffers in AudioFlinger when the A2DP output is opened instead of just transfering active tracks from hardware output mixer thread to the new A2DP output mixer thread. To avoid synchronization issues between mixer threads and client processes, this is done by invalidating tracks by setting a flag in their control block and having AudioTrack release the handle on this track (IAudioTrack) and create a new IAudioTrack when this flag is detected next time obtainBuffer() or start() is executed. AudioFlinger modifications: - invalidate the tracks when setStreamOutput() is called - make sure that notifications of output opening/closing and change of stream type to output mapping are sent synchronously to client process. This is necessary so that AudioSystem has the new stream to output mapping when the AudioTrack detects the invalidate flag in the client process. Previously their were sent when the corresponding thread loop was executed. AudioTrack modifications: - move frame count calculation and verification from set() to createTrack() so that is is updated every time a new IAudioTrack is created. - detect track invalidate flag in obtainBuffer() and start() and create a new IAudioTrack. AudioTrackShared modifications - group all flags (out, flowControlFlag, forceReady...) into a single bit filed to save space. Change-Id: I9ac26b6192230627d35084e1449640caaf7d56ee
2010-05-14 03:26:45 -07:00
result = cblk->cv.waitRelative(cblk->lock, milliseconds(waitTimeMs));
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
cblk->lock.unlock();
mLock.lock();
if (mActive == 0) {
return status_t(STOPPED);
}
cblk->lock.lock();
Fix issue 2553359: Pandora does not work well with Passion deskdock / Cardock. The problem is due to a too big difference between the buffer size used at the hardware interface and at the A2DP interface. When no resampling occurs we don't notice problems but the timing is very tight. As soon as resampling is activated, the AudioTrack underruns. This is because the AudioTrack buffers are not resized when moving the AudioTrack from hardware to A2DP output. The AudioTrack buffers are calculated based on a hardware output buffer size of 3072 bytes. Which is much less than the A2DP output buffer size (10240). The solution consists in creating new tracks with new buffers in AudioFlinger when the A2DP output is opened instead of just transfering active tracks from hardware output mixer thread to the new A2DP output mixer thread. To avoid synchronization issues between mixer threads and client processes, this is done by invalidating tracks by setting a flag in their control block and having AudioTrack release the handle on this track (IAudioTrack) and create a new IAudioTrack when this flag is detected next time obtainBuffer() or start() is executed. AudioFlinger modifications: - invalidate the tracks when setStreamOutput() is called - make sure that notifications of output opening/closing and change of stream type to output mapping are sent synchronously to client process. This is necessary so that AudioSystem has the new stream to output mapping when the AudioTrack detects the invalidate flag in the client process. Previously their were sent when the corresponding thread loop was executed. AudioTrack modifications: - move frame count calculation and verification from set() to createTrack() so that is is updated every time a new IAudioTrack is created. - detect track invalidate flag in obtainBuffer() and start() and create a new IAudioTrack. AudioTrackShared modifications - group all flags (out, flowControlFlag, forceReady...) into a single bit filed to save space. Change-Id: I9ac26b6192230627d35084e1449640caaf7d56ee
2010-05-14 03:26:45 -07:00
}
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
Fix issue 2553359: Pandora does not work well with Passion deskdock / Cardock. The problem is due to a too big difference between the buffer size used at the hardware interface and at the A2DP interface. When no resampling occurs we don't notice problems but the timing is very tight. As soon as resampling is activated, the AudioTrack underruns. This is because the AudioTrack buffers are not resized when moving the AudioTrack from hardware to A2DP output. The AudioTrack buffers are calculated based on a hardware output buffer size of 3072 bytes. Which is much less than the A2DP output buffer size (10240). The solution consists in creating new tracks with new buffers in AudioFlinger when the A2DP output is opened instead of just transfering active tracks from hardware output mixer thread to the new A2DP output mixer thread. To avoid synchronization issues between mixer threads and client processes, this is done by invalidating tracks by setting a flag in their control block and having AudioTrack release the handle on this track (IAudioTrack) and create a new IAudioTrack when this flag is detected next time obtainBuffer() or start() is executed. AudioFlinger modifications: - invalidate the tracks when setStreamOutput() is called - make sure that notifications of output opening/closing and change of stream type to output mapping are sent synchronously to client process. This is necessary so that AudioSystem has the new stream to output mapping when the AudioTrack detects the invalidate flag in the client process. Previously their were sent when the corresponding thread loop was executed. AudioTrack modifications: - move frame count calculation and verification from set() to createTrack() so that is is updated every time a new IAudioTrack is created. - detect track invalidate flag in obtainBuffer() and start() and create a new IAudioTrack. AudioTrackShared modifications - group all flags (out, flowControlFlag, forceReady...) into a single bit filed to save space. Change-Id: I9ac26b6192230627d35084e1449640caaf7d56ee
2010-05-14 03:26:45 -07:00
if (cblk->flags & CBLK_INVALID_MSK) {
goto create_new_track;
}
if (__builtin_expect(result!=NO_ERROR, false)) {
cblk->waitTimeMs += waitTimeMs;
if (cblk->waitTimeMs >= cblk->bufferTimeoutMs) {
// timing out when a loop has been set and we have already written upto loop end
// is a normal condition: no need to wake AudioFlinger up.
if (cblk->user < cblk->loopEnd) {
LOGW( "obtainBuffer timed out (is the CPU pegged?) %p "
"user=%08x, server=%08x", this, cblk->user, cblk->server);
//unlock cblk mutex before calling mAudioTrack->start() (see issue #1617140)
cblk->lock.unlock();
result = mAudioTrack->start();
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
cblk->lock.lock();
if (result == DEAD_OBJECT) {
android_atomic_or(CBLK_INVALID_ON, &cblk->flags);
Fix issue 2553359: Pandora does not work well with Passion deskdock / Cardock. The problem is due to a too big difference between the buffer size used at the hardware interface and at the A2DP interface. When no resampling occurs we don't notice problems but the timing is very tight. As soon as resampling is activated, the AudioTrack underruns. This is because the AudioTrack buffers are not resized when moving the AudioTrack from hardware to A2DP output. The AudioTrack buffers are calculated based on a hardware output buffer size of 3072 bytes. Which is much less than the A2DP output buffer size (10240). The solution consists in creating new tracks with new buffers in AudioFlinger when the A2DP output is opened instead of just transfering active tracks from hardware output mixer thread to the new A2DP output mixer thread. To avoid synchronization issues between mixer threads and client processes, this is done by invalidating tracks by setting a flag in their control block and having AudioTrack release the handle on this track (IAudioTrack) and create a new IAudioTrack when this flag is detected next time obtainBuffer() or start() is executed. AudioFlinger modifications: - invalidate the tracks when setStreamOutput() is called - make sure that notifications of output opening/closing and change of stream type to output mapping are sent synchronously to client process. This is necessary so that AudioSystem has the new stream to output mapping when the AudioTrack detects the invalidate flag in the client process. Previously their were sent when the corresponding thread loop was executed. AudioTrack modifications: - move frame count calculation and verification from set() to createTrack() so that is is updated every time a new IAudioTrack is created. - detect track invalidate flag in obtainBuffer() and start() and create a new IAudioTrack. AudioTrackShared modifications - group all flags (out, flowControlFlag, forceReady...) into a single bit filed to save space. Change-Id: I9ac26b6192230627d35084e1449640caaf7d56ee
2010-05-14 03:26:45 -07:00
create_new_track:
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
result = restoreTrack_l(cblk, false);
}
if (result != NO_ERROR) {
LOGW("obtainBuffer create Track error %d", result);
cblk->lock.unlock();
return result;
}
}
cblk->waitTimeMs = 0;
}
if (--waitCount == 0) {
cblk->lock.unlock();
return TIMED_OUT;
}
}
// read the server count again
start_loop_here:
framesAvail = cblk->framesAvailable_l();
}
cblk->lock.unlock();
}
// restart track if it was disabled by audioflinger due to previous underrun
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
if (mActive && (cblk->flags & CBLK_DISABLED_MSK)) {
android_atomic_and(~CBLK_DISABLED_ON, &cblk->flags);
LOGW("obtainBuffer() track %p disabled, restarting", this);
mAudioTrack->start();
}
cblk->waitTimeMs = 0;
if (framesReq > framesAvail) {
framesReq = framesAvail;
}
uint32_t u = cblk->user;
uint32_t bufferEnd = cblk->userBase + cblk->frameCount;
if (u + framesReq > bufferEnd) {
framesReq = bufferEnd - u;
}
audioBuffer->flags = mMuted ? Buffer::MUTE : 0;
audioBuffer->channelCount = mChannelCount;
audioBuffer->frameCount = framesReq;
audioBuffer->size = framesReq * cblk->frameSize;
if (audio_is_linear_pcm(mFormat)) {
audioBuffer->format = AUDIO_FORMAT_PCM_16_BIT;
} else {
audioBuffer->format = mFormat;
}
audioBuffer->raw = (int8_t *)cblk->buffer(u);
active = mActive;
return active ? status_t(NO_ERROR) : status_t(STOPPED);
}
void AudioTrack::releaseBuffer(Buffer* audioBuffer)
{
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
AutoMutex lock(mLock);
mCblk->stepUser(audioBuffer->frameCount);
}
// -------------------------------------------------------------------------
ssize_t AudioTrack::write(const void* buffer, size_t userSize)
{
if (mSharedBuffer != 0) return INVALID_OPERATION;
if (ssize_t(userSize) < 0) {
// sanity-check. user is most-likely passing an error code.
LOGE("AudioTrack::write(buffer=%p, size=%u (%d)",
buffer, userSize, userSize);
return BAD_VALUE;
}
LOGV("write %p: %d bytes, mActive=%d", this, userSize, mActive);
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
// acquire a strong reference on the IMemory and IAudioTrack so that they cannot be destroyed
// while we are accessing the cblk
mLock.lock();
sp <IAudioTrack> audioTrack = mAudioTrack;
sp <IMemory> iMem = mCblkMemory;
mLock.unlock();
ssize_t written = 0;
const int8_t *src = (const int8_t *)buffer;
Buffer audioBuffer;
size_t frameSz = (size_t)frameSize();
do {
audioBuffer.frameCount = userSize/frameSz;
// Calling obtainBuffer() with a negative wait count causes
// an (almost) infinite wait time.
status_t err = obtainBuffer(&audioBuffer, -1);
if (err < 0) {
// out of buffers, return #bytes written
if (err == status_t(NO_MORE_BUFFERS))
break;
return ssize_t(err);
}
size_t toWrite;
if (mFormat == AUDIO_FORMAT_PCM_8_BIT && !(mFlags & AUDIO_POLICY_OUTPUT_FLAG_DIRECT)) {
// Divide capacity by 2 to take expansion into account
toWrite = audioBuffer.size>>1;
// 8 to 16 bit conversion
int count = toWrite;
int16_t *dst = (int16_t *)(audioBuffer.i8);
while(count--) {
*dst++ = (int16_t)(*src++^0x80) << 8;
}
} else {
toWrite = audioBuffer.size;
memcpy(audioBuffer.i8, src, toWrite);
src += toWrite;
}
userSize -= toWrite;
written += toWrite;
releaseBuffer(&audioBuffer);
} while (userSize >= frameSz);
return written;
}
// -------------------------------------------------------------------------
bool AudioTrack::processAudioBuffer(const sp<AudioTrackThread>& thread)
{
Buffer audioBuffer;
uint32_t frames;
size_t writtenSize;
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
mLock.lock();
// acquire a strong reference on the IMemory and IAudioTrack so that they cannot be destroyed
// while we are accessing the cblk
sp <IAudioTrack> audioTrack = mAudioTrack;
sp <IMemory> iMem = mCblkMemory;
audio_track_cblk_t* cblk = mCblk;
mLock.unlock();
// Manage underrun callback
if (mActive && (cblk->framesAvailable() == cblk->frameCount)) {
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
LOGV("Underrun user: %x, server: %x, flags %04x", cblk->user, cblk->server, cblk->flags);
if (!(android_atomic_or(CBLK_UNDERRUN_ON, &cblk->flags) & CBLK_UNDERRUN_MSK)) {
mCbf(EVENT_UNDERRUN, mUserData, 0);
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
if (cblk->server == cblk->frameCount) {
mCbf(EVENT_BUFFER_END, mUserData, 0);
}
if (mSharedBuffer != 0) return false;
}
}
// Manage loop end callback
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
while (mLoopCount > cblk->loopCount) {
int loopCount = -1;
mLoopCount--;
if (mLoopCount >= 0) loopCount = mLoopCount;
mCbf(EVENT_LOOP_END, mUserData, (void *)&loopCount);
}
// Manage marker callback
if (!mMarkerReached && (mMarkerPosition > 0)) {
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
if (cblk->server >= mMarkerPosition) {
mCbf(EVENT_MARKER, mUserData, (void *)&mMarkerPosition);
mMarkerReached = true;
}
}
// Manage new position callback
if (mUpdatePeriod > 0) {
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
while (cblk->server >= mNewPosition) {
mCbf(EVENT_NEW_POS, mUserData, (void *)&mNewPosition);
mNewPosition += mUpdatePeriod;
}
}
// If Shared buffer is used, no data is requested from client.
if (mSharedBuffer != 0) {
frames = 0;
} else {
frames = mRemainingFrames;
}
do {
audioBuffer.frameCount = frames;
// Calling obtainBuffer() with a wait count of 1
// limits wait time to WAIT_PERIOD_MS. This prevents from being
// stuck here not being able to handle timed events (position, markers, loops).
status_t err = obtainBuffer(&audioBuffer, 1);
if (err < NO_ERROR) {
if (err != TIMED_OUT) {
LOGE_IF(err != status_t(NO_MORE_BUFFERS), "Error obtaining an audio buffer, giving up.");
return false;
}
break;
}
if (err == status_t(STOPPED)) return false;
// Divide buffer size by 2 to take into account the expansion
// due to 8 to 16 bit conversion: the callback must fill only half
// of the destination buffer
if (mFormat == AUDIO_FORMAT_PCM_8_BIT && !(mFlags & AUDIO_POLICY_OUTPUT_FLAG_DIRECT)) {
audioBuffer.size >>= 1;
}
size_t reqSize = audioBuffer.size;
mCbf(EVENT_MORE_DATA, mUserData, &audioBuffer);
writtenSize = audioBuffer.size;
// Sanity check on returned size
if (ssize_t(writtenSize) <= 0) {
// The callback is done filling buffers
// Keep this thread going to handle timed events and
// still try to get more data in intervals of WAIT_PERIOD_MS
// but don't just loop and block the CPU, so wait
usleep(WAIT_PERIOD_MS*1000);
break;
}
if (writtenSize > reqSize) writtenSize = reqSize;
if (mFormat == AUDIO_FORMAT_PCM_8_BIT && !(mFlags & AUDIO_POLICY_OUTPUT_FLAG_DIRECT)) {
// 8 to 16 bit conversion
const int8_t *src = audioBuffer.i8 + writtenSize-1;
int count = writtenSize;
int16_t *dst = audioBuffer.i16 + writtenSize-1;
while(count--) {
*dst-- = (int16_t)(*src--^0x80) << 8;
}
writtenSize <<= 1;
}
audioBuffer.size = writtenSize;
// NOTE: mCblk->frameSize is not equal to AudioTrack::frameSize() for
// 8 bit PCM data: in this case, mCblk->frameSize is based on a sampel size of
// 16 bit.
audioBuffer.frameCount = writtenSize/mCblk->frameSize;
frames -= audioBuffer.frameCount;
releaseBuffer(&audioBuffer);
}
while (frames);
if (frames == 0) {
Fix issue 2553359: Pandora does not work well with Passion deskdock / Cardock. The problem is due to a too big difference between the buffer size used at the hardware interface and at the A2DP interface. When no resampling occurs we don't notice problems but the timing is very tight. As soon as resampling is activated, the AudioTrack underruns. This is because the AudioTrack buffers are not resized when moving the AudioTrack from hardware to A2DP output. The AudioTrack buffers are calculated based on a hardware output buffer size of 3072 bytes. Which is much less than the A2DP output buffer size (10240). The solution consists in creating new tracks with new buffers in AudioFlinger when the A2DP output is opened instead of just transfering active tracks from hardware output mixer thread to the new A2DP output mixer thread. To avoid synchronization issues between mixer threads and client processes, this is done by invalidating tracks by setting a flag in their control block and having AudioTrack release the handle on this track (IAudioTrack) and create a new IAudioTrack when this flag is detected next time obtainBuffer() or start() is executed. AudioFlinger modifications: - invalidate the tracks when setStreamOutput() is called - make sure that notifications of output opening/closing and change of stream type to output mapping are sent synchronously to client process. This is necessary so that AudioSystem has the new stream to output mapping when the AudioTrack detects the invalidate flag in the client process. Previously their were sent when the corresponding thread loop was executed. AudioTrack modifications: - move frame count calculation and verification from set() to createTrack() so that is is updated every time a new IAudioTrack is created. - detect track invalidate flag in obtainBuffer() and start() and create a new IAudioTrack. AudioTrackShared modifications - group all flags (out, flowControlFlag, forceReady...) into a single bit filed to save space. Change-Id: I9ac26b6192230627d35084e1449640caaf7d56ee
2010-05-14 03:26:45 -07:00
mRemainingFrames = mNotificationFramesAct;
} else {
mRemainingFrames = frames;
}
return true;
}
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
// must be called with mLock and cblk.lock held. Callers must also hold strong references on
// the IAudioTrack and IMemory in case they are recreated here.
// If the IAudioTrack is successfully restored, the cblk pointer is updated
status_t AudioTrack::restoreTrack_l(audio_track_cblk_t*& cblk, bool fromStart)
{
status_t result;
if (!(android_atomic_or(CBLK_RESTORING_ON, &cblk->flags) & CBLK_RESTORING_MSK)) {
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
LOGW("dead IAudioTrack, creating a new one from %s",
fromStart ? "start()" : "obtainBuffer()");
// signal old cblk condition so that other threads waiting for available buffers stop
// waiting now
cblk->cv.broadcast();
cblk->lock.unlock();
// refresh the audio configuration cache in this process to make sure we get new
// output parameters in getOutput_l() and createTrack_l()
AudioSystem::clearAudioConfigCache();
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
// if the new IAudioTrack is created, createTrack_l() will modify the
// following member variables: mAudioTrack, mCblkMemory and mCblk.
// It will also delete the strong references on previous IAudioTrack and IMemory
result = createTrack_l(mStreamType,
cblk->sampleRate,
mFormat,
mChannelMask,
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
mFrameCount,
mFlags,
mSharedBuffer,
getOutput_l(),
false);
if (result == NO_ERROR) {
uint32_t user = cblk->user;
uint32_t server = cblk->server;
// restore write index and set other indexes to reflect empty buffer status
mCblk->user = user;
mCblk->server = user;
mCblk->userBase = user;
mCblk->serverBase = user;
// restore loop: this is not guaranteed to succeed if new frame count is not
// compatible with loop length
setLoop_l(cblk->loopStart, cblk->loopEnd, cblk->loopCount);
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
if (!fromStart) {
mCblk->bufferTimeoutMs = MAX_RUN_TIMEOUT_MS;
// Make sure that a client relying on callback events indicating underrun or
// the actual amount of audio frames played (e.g SoundPool) receives them.
if (mSharedBuffer == 0) {
uint32_t frames = 0;
if (user > server) {
frames = ((user - server) > mCblk->frameCount) ?
mCblk->frameCount : (user - server);
memset(mCblk->buffers, 0, frames * mCblk->frameSize);
}
// restart playback even if buffer is not completely filled.
android_atomic_or(CBLK_FORCEREADY_ON, &mCblk->flags);
// stepUser() clears CBLK_UNDERRUN_ON flag enabling underrun callbacks to
// the client
mCblk->stepUser(frames);
}
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
}
if (mActive) {
result = mAudioTrack->start();
}
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
if (fromStart && result == NO_ERROR) {
mNewPosition = mCblk->server + mUpdatePeriod;
}
}
if (result != NO_ERROR) {
mActive = false;
}
// signal old cblk condition for other threads waiting for restore completion
android_atomic_or(CBLK_RESTORED_ON, &cblk->flags);
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
cblk->cv.broadcast();
} else {
if (!(cblk->flags & CBLK_RESTORED_MSK)) {
LOGW("dead IAudioTrack, waiting for a new one");
mLock.unlock();
result = cblk->cv.waitRelative(cblk->lock, milliseconds(RESTORE_TIMEOUT_MS));
cblk->lock.unlock();
mLock.lock();
} else {
LOGW("dead IAudioTrack, already restored");
result = NO_ERROR;
cblk->lock.unlock();
}
if (result != NO_ERROR || mActive == 0) {
result = status_t(STOPPED);
}
}
LOGV("restoreTrack_l() status %d mActive %d cblk %p, old cblk %p flags %08x old flags %08x",
result, mActive, mCblk, cblk, mCblk->flags, cblk->flags);
if (result == NO_ERROR) {
// from now on we switch to the newly created cblk
cblk = mCblk;
}
cblk->lock.lock();
LOGW_IF(result != NO_ERROR, "restoreTrack_l() error %d", result);
return result;
}
status_t AudioTrack::dump(int fd, const Vector<String16>& args) const
{
const size_t SIZE = 256;
char buffer[SIZE];
String8 result;
result.append(" AudioTrack::dump\n");
snprintf(buffer, 255, " stream type(%d), left - right volume(%f, %f)\n", mStreamType, mVolume[0], mVolume[1]);
result.append(buffer);
Fix issue 2553359: Pandora does not work well with Passion deskdock / Cardock. The problem is due to a too big difference between the buffer size used at the hardware interface and at the A2DP interface. When no resampling occurs we don't notice problems but the timing is very tight. As soon as resampling is activated, the AudioTrack underruns. This is because the AudioTrack buffers are not resized when moving the AudioTrack from hardware to A2DP output. The AudioTrack buffers are calculated based on a hardware output buffer size of 3072 bytes. Which is much less than the A2DP output buffer size (10240). The solution consists in creating new tracks with new buffers in AudioFlinger when the A2DP output is opened instead of just transfering active tracks from hardware output mixer thread to the new A2DP output mixer thread. To avoid synchronization issues between mixer threads and client processes, this is done by invalidating tracks by setting a flag in their control block and having AudioTrack release the handle on this track (IAudioTrack) and create a new IAudioTrack when this flag is detected next time obtainBuffer() or start() is executed. AudioFlinger modifications: - invalidate the tracks when setStreamOutput() is called - make sure that notifications of output opening/closing and change of stream type to output mapping are sent synchronously to client process. This is necessary so that AudioSystem has the new stream to output mapping when the AudioTrack detects the invalidate flag in the client process. Previously their were sent when the corresponding thread loop was executed. AudioTrack modifications: - move frame count calculation and verification from set() to createTrack() so that is is updated every time a new IAudioTrack is created. - detect track invalidate flag in obtainBuffer() and start() and create a new IAudioTrack. AudioTrackShared modifications - group all flags (out, flowControlFlag, forceReady...) into a single bit filed to save space. Change-Id: I9ac26b6192230627d35084e1449640caaf7d56ee
2010-05-14 03:26:45 -07:00
snprintf(buffer, 255, " format(%d), channel count(%d), frame count(%d)\n", mFormat, mChannelCount, mCblk->frameCount);
result.append(buffer);
snprintf(buffer, 255, " sample rate(%d), status(%d), muted(%d)\n", (mCblk == 0) ? 0 : mCblk->sampleRate, mStatus, mMuted);
result.append(buffer);
snprintf(buffer, 255, " active(%d), latency (%d)\n", mActive, mLatency);
result.append(buffer);
::write(fd, result.string(), result.size());
return NO_ERROR;
}
// =========================================================================
AudioTrack::AudioTrackThread::AudioTrackThread(AudioTrack& receiver, bool bCanCallJava)
: Thread(bCanCallJava), mReceiver(receiver)
{
}
bool AudioTrack::AudioTrackThread::threadLoop()
{
return mReceiver.processAudioBuffer(this);
}
status_t AudioTrack::AudioTrackThread::readyToRun()
{
return NO_ERROR;
}
void AudioTrack::AudioTrackThread::onFirstRef()
{
}
// =========================================================================
audio_track_cblk_t::audio_track_cblk_t()
: lock(Mutex::SHARED), cv(Condition::SHARED), user(0), server(0),
userBase(0), serverBase(0), buffers(0), frameCount(0),
loopStart(UINT_MAX), loopEnd(UINT_MAX), loopCount(0), volumeLR(0),
sendLevel(0), flags(0)
{
}
uint32_t audio_track_cblk_t::stepUser(uint32_t frameCount)
{
uint32_t u = this->user;
u += frameCount;
// Ensure that user is never ahead of server for AudioRecord
Fix issue 2553359: Pandora does not work well with Passion deskdock / Cardock. The problem is due to a too big difference between the buffer size used at the hardware interface and at the A2DP interface. When no resampling occurs we don't notice problems but the timing is very tight. As soon as resampling is activated, the AudioTrack underruns. This is because the AudioTrack buffers are not resized when moving the AudioTrack from hardware to A2DP output. The AudioTrack buffers are calculated based on a hardware output buffer size of 3072 bytes. Which is much less than the A2DP output buffer size (10240). The solution consists in creating new tracks with new buffers in AudioFlinger when the A2DP output is opened instead of just transfering active tracks from hardware output mixer thread to the new A2DP output mixer thread. To avoid synchronization issues between mixer threads and client processes, this is done by invalidating tracks by setting a flag in their control block and having AudioTrack release the handle on this track (IAudioTrack) and create a new IAudioTrack when this flag is detected next time obtainBuffer() or start() is executed. AudioFlinger modifications: - invalidate the tracks when setStreamOutput() is called - make sure that notifications of output opening/closing and change of stream type to output mapping are sent synchronously to client process. This is necessary so that AudioSystem has the new stream to output mapping when the AudioTrack detects the invalidate flag in the client process. Previously their were sent when the corresponding thread loop was executed. AudioTrack modifications: - move frame count calculation and verification from set() to createTrack() so that is is updated every time a new IAudioTrack is created. - detect track invalidate flag in obtainBuffer() and start() and create a new IAudioTrack. AudioTrackShared modifications - group all flags (out, flowControlFlag, forceReady...) into a single bit filed to save space. Change-Id: I9ac26b6192230627d35084e1449640caaf7d56ee
2010-05-14 03:26:45 -07:00
if (flags & CBLK_DIRECTION_MSK) {
// If stepServer() has been called once, switch to normal obtainBuffer() timeout period
if (bufferTimeoutMs == MAX_STARTUP_TIMEOUT_MS-1) {
bufferTimeoutMs = MAX_RUN_TIMEOUT_MS;
}
} else if (u > this->server) {
LOGW("stepServer occured after track reset");
u = this->server;
}
if (u >= userBase + this->frameCount) {
userBase += this->frameCount;
}
this->user = u;
// Clear flow control error condition as new data has been written/read to/from buffer.
if (flags & CBLK_UNDERRUN_MSK) {
android_atomic_and(~CBLK_UNDERRUN_MSK, &flags);
}
return u;
}
bool audio_track_cblk_t::stepServer(uint32_t frameCount)
{
if (!tryLock()) {
LOGW("stepServer() could not lock cblk");
return false;
}
uint32_t s = this->server;
s += frameCount;
Fix issue 2553359: Pandora does not work well with Passion deskdock / Cardock. The problem is due to a too big difference between the buffer size used at the hardware interface and at the A2DP interface. When no resampling occurs we don't notice problems but the timing is very tight. As soon as resampling is activated, the AudioTrack underruns. This is because the AudioTrack buffers are not resized when moving the AudioTrack from hardware to A2DP output. The AudioTrack buffers are calculated based on a hardware output buffer size of 3072 bytes. Which is much less than the A2DP output buffer size (10240). The solution consists in creating new tracks with new buffers in AudioFlinger when the A2DP output is opened instead of just transfering active tracks from hardware output mixer thread to the new A2DP output mixer thread. To avoid synchronization issues between mixer threads and client processes, this is done by invalidating tracks by setting a flag in their control block and having AudioTrack release the handle on this track (IAudioTrack) and create a new IAudioTrack when this flag is detected next time obtainBuffer() or start() is executed. AudioFlinger modifications: - invalidate the tracks when setStreamOutput() is called - make sure that notifications of output opening/closing and change of stream type to output mapping are sent synchronously to client process. This is necessary so that AudioSystem has the new stream to output mapping when the AudioTrack detects the invalidate flag in the client process. Previously their were sent when the corresponding thread loop was executed. AudioTrack modifications: - move frame count calculation and verification from set() to createTrack() so that is is updated every time a new IAudioTrack is created. - detect track invalidate flag in obtainBuffer() and start() and create a new IAudioTrack. AudioTrackShared modifications - group all flags (out, flowControlFlag, forceReady...) into a single bit filed to save space. Change-Id: I9ac26b6192230627d35084e1449640caaf7d56ee
2010-05-14 03:26:45 -07:00
if (flags & CBLK_DIRECTION_MSK) {
// Mark that we have read the first buffer so that next time stepUser() is called
// we switch to normal obtainBuffer() timeout period
if (bufferTimeoutMs == MAX_STARTUP_TIMEOUT_MS) {
bufferTimeoutMs = MAX_STARTUP_TIMEOUT_MS - 1;
}
// It is possible that we receive a flush()
// while the mixer is processing a block: in this case,
// stepServer() is called After the flush() has reset u & s and
// we have s > u
if (s > this->user) {
LOGW("stepServer occured after track reset");
s = this->user;
}
}
if (s >= loopEnd) {
LOGW_IF(s > loopEnd, "stepServer: s %u > loopEnd %u", s, loopEnd);
s = loopStart;
if (--loopCount == 0) {
loopEnd = UINT_MAX;
loopStart = UINT_MAX;
}
}
if (s >= serverBase + this->frameCount) {
serverBase += this->frameCount;
}
this->server = s;
Fix issue 3439872: video chat and bluetooth SCO This change fixes the stability problems experienced when using a bluetooth headset supporting both A2DP and SCO. Problems occur when starting the video chat at which time the A2DP output is being stopped to start SCO. At that time, active AudioTracks are invalidated by AudioFlinger so that a new AudioTrack binder interface can be recreated by the client process on the new mixer thread with correct parameters. The problem was that the process to restore the binder interface was not protected against concurrent requests which caused 2 binder interfaces to be created sometimes. This could lead to permanent client deadlock if one of the client threads was waiting for a condition of the first created binder interface while the second one was created (as the AudioFlinger would only signal conditions on the last one created). This concurrent request situation is more likely to happen when a client uses the JAVA AudioTrack as the JNI implementation uses simultaneously the native AudioTrack callback and write push mechanisms. By doing so, the code that checks if the binder interface should be restored (in obtainBuffer()) is much more likely to be called concurrently from two different threads. The fix consists in protecting the critical binder interface restore phase with a flag in the AudioTrack control block. The first thread acting upon the binder interface restore request will raise the flag and the second thread will just wait for a condition to be signaled when the restore process is complete. Also protected all accesses to the AudioTrack control block by a mutex to prevent access while the track is being destroyed and restored. If a mutex cannot be held (e.g because we call a callback function), acquire a strong reference on the IAudioTrack to prevent its destruction while the cblk is being accessed. Modified AudioTrack JNI to use GetByteArrayElements() instead of GetPrimitiveArrayCritical() when writing audio buffers. Entering a critical section would cause the JNI to abort if a mediaserver crash occurs during a write due to the AudioSystem callback being called during the critical section when media server process restarts. Anyway with current JNI implementation, either versions do not copy data most of the times and the criticial version does not guaranty no data copy. The same modifications have been made to AudioRecord. Change-Id: Idc5aa711a04c3eee180cdd03f44fe17f3c4dcb52
2011-03-07 14:52:59 -08:00
if (!(flags & CBLK_INVALID_MSK)) {
cv.signal();
}
lock.unlock();
return true;
}
void* audio_track_cblk_t::buffer(uint32_t offset) const
{
return (int8_t *)this->buffers + (offset - userBase) * this->frameSize;
}
uint32_t audio_track_cblk_t::framesAvailable()
{
Mutex::Autolock _l(lock);
return framesAvailable_l();
}
uint32_t audio_track_cblk_t::framesAvailable_l()
{
uint32_t u = this->user;
uint32_t s = this->server;
Fix issue 2553359: Pandora does not work well with Passion deskdock / Cardock. The problem is due to a too big difference between the buffer size used at the hardware interface and at the A2DP interface. When no resampling occurs we don't notice problems but the timing is very tight. As soon as resampling is activated, the AudioTrack underruns. This is because the AudioTrack buffers are not resized when moving the AudioTrack from hardware to A2DP output. The AudioTrack buffers are calculated based on a hardware output buffer size of 3072 bytes. Which is much less than the A2DP output buffer size (10240). The solution consists in creating new tracks with new buffers in AudioFlinger when the A2DP output is opened instead of just transfering active tracks from hardware output mixer thread to the new A2DP output mixer thread. To avoid synchronization issues between mixer threads and client processes, this is done by invalidating tracks by setting a flag in their control block and having AudioTrack release the handle on this track (IAudioTrack) and create a new IAudioTrack when this flag is detected next time obtainBuffer() or start() is executed. AudioFlinger modifications: - invalidate the tracks when setStreamOutput() is called - make sure that notifications of output opening/closing and change of stream type to output mapping are sent synchronously to client process. This is necessary so that AudioSystem has the new stream to output mapping when the AudioTrack detects the invalidate flag in the client process. Previously their were sent when the corresponding thread loop was executed. AudioTrack modifications: - move frame count calculation and verification from set() to createTrack() so that is is updated every time a new IAudioTrack is created. - detect track invalidate flag in obtainBuffer() and start() and create a new IAudioTrack. AudioTrackShared modifications - group all flags (out, flowControlFlag, forceReady...) into a single bit filed to save space. Change-Id: I9ac26b6192230627d35084e1449640caaf7d56ee
2010-05-14 03:26:45 -07:00
if (flags & CBLK_DIRECTION_MSK) {
uint32_t limit = (s < loopStart) ? s : loopStart;
return limit + frameCount - u;
} else {
return frameCount + u - s;
}
}
uint32_t audio_track_cblk_t::framesReady()
{
uint32_t u = this->user;
uint32_t s = this->server;
Fix issue 2553359: Pandora does not work well with Passion deskdock / Cardock. The problem is due to a too big difference between the buffer size used at the hardware interface and at the A2DP interface. When no resampling occurs we don't notice problems but the timing is very tight. As soon as resampling is activated, the AudioTrack underruns. This is because the AudioTrack buffers are not resized when moving the AudioTrack from hardware to A2DP output. The AudioTrack buffers are calculated based on a hardware output buffer size of 3072 bytes. Which is much less than the A2DP output buffer size (10240). The solution consists in creating new tracks with new buffers in AudioFlinger when the A2DP output is opened instead of just transfering active tracks from hardware output mixer thread to the new A2DP output mixer thread. To avoid synchronization issues between mixer threads and client processes, this is done by invalidating tracks by setting a flag in their control block and having AudioTrack release the handle on this track (IAudioTrack) and create a new IAudioTrack when this flag is detected next time obtainBuffer() or start() is executed. AudioFlinger modifications: - invalidate the tracks when setStreamOutput() is called - make sure that notifications of output opening/closing and change of stream type to output mapping are sent synchronously to client process. This is necessary so that AudioSystem has the new stream to output mapping when the AudioTrack detects the invalidate flag in the client process. Previously their were sent when the corresponding thread loop was executed. AudioTrack modifications: - move frame count calculation and verification from set() to createTrack() so that is is updated every time a new IAudioTrack is created. - detect track invalidate flag in obtainBuffer() and start() and create a new IAudioTrack. AudioTrackShared modifications - group all flags (out, flowControlFlag, forceReady...) into a single bit filed to save space. Change-Id: I9ac26b6192230627d35084e1449640caaf7d56ee
2010-05-14 03:26:45 -07:00
if (flags & CBLK_DIRECTION_MSK) {
if (u < loopEnd) {
return u - s;
} else {
// do not block on mutex shared with client on AudioFlinger side
if (!tryLock()) {
LOGW("framesReady() could not lock cblk");
return 0;
}
uint32_t frames = UINT_MAX;
if (loopCount >= 0) {
frames = (loopEnd - loopStart)*loopCount + u - s;
}
lock.unlock();
return frames;
}
} else {
return s - u;
}
}
bool audio_track_cblk_t::tryLock()
{
// the code below simulates lock-with-timeout
// we MUST do this to protect the AudioFlinger server
// as this lock is shared with the client.
status_t err;
err = lock.tryLock();
if (err == -EBUSY) { // just wait a bit
usleep(1000);
err = lock.tryLock();
}
if (err != NO_ERROR) {
// probably, the client just died.
return false;
}
return true;
}
// -------------------------------------------------------------------------
}; // namespace android