blob: 82678c9e92027e2db00f35034737032d27bb01fe [file] [log] [blame]
** Copyright 2012, 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
** Unless required by applicable law or agreed to in writing, software
** distributed under the License is distributed on an "AS IS" BASIS,
** See the License for the specific language governing permissions and
** limitations under the License.
#define LOG_TAG "AudioHAL_AudioStreamOut"
#include <utils/Log.h>
#include "AudioHardwareOutput.h"
#include "AudioStreamOut.h"
// Set to 1 to print timestamp data in CSV format.
#define ALOGVV(a...) do { } while(0)
namespace android {
AudioStreamOut::AudioStreamOut(AudioHardwareOutput& owner, bool mcOut)
: mFramesRendered(0)
, mOwnerHAL(owner)
, mFramesWritten(0)
, mTgtDevices(0)
, mAudioFlingerTgtDevices(0)
, mIsMCOutput(mcOut)
, mInStandby(false)
, mReportedAvailFail(false)
// Set some reasonable defaults for these. All of this should be eventually
// be overwritten by a specific audio flinger configuration, but it does not
// hurt to have something here by default.
mInputSampleRate = 48000;
mInputFormat = AUDIO_FORMAT_PCM_16_BIT;
mInputNominalChunksInFlight = 4;
mThrottleValid = false;
memset(&mUSecToLocalTime, 0, sizeof(mUSecToLocalTime));
mUSecToLocalTime.a_to_b_numer = mLocalClock.getLocalFreq();
mUSecToLocalTime.a_to_b_denom = 1000000;
status_t AudioStreamOut::set(
audio_format_t *pFormat,
uint32_t *pChannels,
uint32_t *pRate)
Mutex::Autolock _l(mLock);
audio_format_t lFormat = pFormat ? *pFormat : AUDIO_FORMAT_DEFAULT;
uint32_t lChannels = pChannels ? *pChannels : 0;
uint32_t lRate = pRate ? *pRate : 0;
// fix up defaults
if (lFormat == AUDIO_FORMAT_DEFAULT) lFormat = format();
if (lChannels == 0) lChannels = chanMask();
if (lRate == 0) lRate = sampleRate();
if (pFormat) *pFormat = lFormat;
if (pChannels) *pChannels = lChannels;
if (pRate) *pRate = lRate;
if (!audio_is_linear_pcm(lFormat)) {
ALOGW("set: format 0x%08X needs to be wrapped in SPDIF data burst", lFormat);
return BAD_VALUE;
if (!mIsMCOutput) {
// If this is the primary stream out, then demand our defaults.
if ((lFormat != AUDIO_FORMAT_PCM_16_BIT && lFormat != AUDIO_FORMAT_PCM_8_24_BIT) ||
(lChannels != chanMask()) ||
(lRate != sampleRate())) {
ALOGW("set: parameters incompatible with defaults");
return BAD_VALUE;
} else {
// Else check to see if our HDMI sink supports this format before proceeding.
if (!mOwnerHAL.getHDMIAudioCaps().supportsFormat(
lFormat, lRate, audio_channel_count_from_out_mask(lChannels))) {
ALOGW("set: parameters incompatible with hdmi capabilities");
return BAD_VALUE;
mInputFormat = lFormat;
mInputChanMask = lChannels;
mInputSampleRate = lRate;
ALOGI("AudioStreamOut::set: rate = %u, format = 0x%08X\n", lRate, lFormat);
return NO_ERROR;
void AudioStreamOut::setTgtDevices(uint32_t tgtDevices)
Mutex::Autolock _l(mRoutingLock);
if (mTgtDevices != tgtDevices) {
mTgtDevices = tgtDevices;
status_t AudioStreamOut::standbyHardware()
mOwnerHAL.standbyStatusUpdate(true, mIsMCOutput);
mInStandby = true;
return NO_ERROR;
status_t AudioStreamOut::standby()
mFramesRendered = 0;
return standbyHardware();
void AudioStreamOut::releaseAllOutputs() {
Mutex::Autolock _l(mRoutingLock);
ALOGI("releaseAllOutputs: releasing %d mPhysOutputs", mPhysOutputs.size());
AudioOutputList::iterator I;
for (I = mPhysOutputs.begin(); I != mPhysOutputs.end(); ++I)
mOwnerHAL.releaseOutput(*this, *I);
status_t AudioStreamOut::pause()
return standbyHardware();
status_t AudioStreamOut::resume()
return NO_ERROR;
status_t AudioStreamOut::flush()
mFramesRendered = 0;
return NO_ERROR;
void AudioStreamOut::updateInputNums()
// mInputBufSize determines how many audio frames AudioFlinger is going to
// mix at a time. We also use the mInputBufSize to determine the ALSA
// period_size, the number of of samples which need to play out (at most)
// before low level ALSA driver code is required to wake up upper levels of
// SW to fill a new buffer. As it turns out, ALSA is going to apply some
// rules and modify the period_size which we pass to it. One of the things
// ALSA seems to do is attempt to round the period_size up to a value which
// will make the period an integral number of 0.5 mSec. This round-up
// behavior can cause the low levels of ALSA to consume more data per period
// than the AudioFlinger mixer has been told to produce. If there are only
// two buffers in flight at any given point in time, this can lead to a
// situation where the pipeline ends up slipping an extra buffer and
// underflowing. There are two approaches to mitigate this, both of which
// are implemented in this HAL...
// 1) Try as hard as possible to make certain that the buffer size we choose
// results in a period_size which is not going to get rounded up by ALSA.
// This means that we want a buffer size which at the chosen sample rate
// and frame size will be an integral multiple of 1/2 mSec.
// 2) Increate the number of chunks we keep in flight. If the system slips
// a single period, its only really a problem if there is no data left in
// the pipeline waiting to be played out. The mixer should going to mix
// as fast as possible until the buffer has been topped off. By
// decreasing the buffer size and increasing the number of buffers in
// flight, we increase the number of interrups and mix events per second,
// but buy ourselves some insurance against the negative side effects of
// slipping one buffer in the schedule. We end up using 4 buffers at
// 10mSec, making the total audio latency somewhere between 40 and 50
// mSec, depending on when a sample begins playback relative to
// AudioFlinger's mixing schedule.
mInputChanCount = audio_channel_count_from_out_mask(mInputChanMask);
// Picking a chunk duration 10mSec should satisfy #1 for both major families
// of audio sample rates (the 44.1K and 48K families). In the case of 44.1
// (or higher) we will end up with a multiple of 441 frames of audio per
// chunk, while for 48K, we will have a multiple of 480 frames of audio per
// chunk. This will not work well for lower sample rates in the 44.1 family
// (22.05K and 11.025K); it is unlikely that we will ever be configured to
// deliver those rates, and if we ever do, we will need to rely on having
// extra chunks in flight to deal with the jitter problem described above.
mInputChunkFrames = outputSampleRate() / 100;
// FIXME: Currently, audio flinger demands an input buffer size which is a
// multiple of 16 audio frames. Right now, there is no good way to
// reconcile this with ALSA round-up behavior described above when the
// desired sample rate is a member of the 44.1 family. For now, we just
// round up to the nearest multiple of 16 frames and roll the dice, but
// someday it would be good to fix one or the other halves of the problem
// (either ALSA or AudioFlinger)
mInputChunkFrames = (mInputChunkFrames + 0xF) & ~0xF;
ALOGD("AudioStreamOut::updateInputNums: chunk size %u from output rate %u\n",
mInputChunkFrames, outputSampleRate());
mInputFrameSize = mInputChanCount * audio_bytes_per_sample(mInputFormat);
// Buffer size is just the frame size multiplied by the number of
// frames per chunk.
mInputBufSize = mInputChunkFrames * mInputFrameSize;
// The nominal latency is just the duration of a chunk * the number of
// chunks we nominally keep in flight at any given point in time.
mInputNominalLatencyUSec = static_cast<uint32_t>(((
* 1000000 * mInputNominalChunksInFlight)
/ mInputSampleRate));
memset(&mLocalTimeToFrames, 0, sizeof(mLocalTimeToFrames));
mLocalTimeToFrames.a_to_b_numer = mInputSampleRate;
mLocalTimeToFrames.a_to_b_denom = mLocalClock.getLocalFreq();
void AudioStreamOut::finishedWriteOp(size_t framesWritten,
bool needThrottle)
int64_t now = mLocalClock.getLocalTime();
if (!mThrottleValid || !needThrottle) {
mThrottleValid = true;
mWriteStartLT = now;
mFramesWritten = 0;
mFramesWritten += framesWritten;
mFramesRendered += framesWritten;
if (needThrottle) {
int64_t deltaLT;
mLocalTimeToFrames.doReverseTransform(mFramesWritten, &deltaLT);
deltaLT += mWriteStartLT;
deltaLT -= now;
int64_t deltaUSec;
mUSecToLocalTime.doReverseTransform(deltaLT, &deltaUSec);
if (deltaUSec > 0) {
useconds_t sleep_time;
// We should never be a full second ahead of schedule; sanity check
// our throttle time and cap the max sleep time at 1 second.
if (deltaUSec > 1000000) {
ALOGW("throttle time clipped! deltaLT = %lld deltaUSec = %lld",
deltaLT, deltaUSec);
sleep_time = 1000000;
} else {
sleep_time = static_cast<useconds_t>(deltaUSec);
static const String8 keyRouting(AudioParameter::keyRouting);
static const String8 keySupSampleRates("sup_sampling_rates");
static const String8 keySupFormats("sup_formats");
static const String8 keySupChannels("sup_channels");
status_t AudioStreamOut::setParameters(__unused struct audio_stream *stream, const char *kvpairs)
AudioParameter param = AudioParameter(String8(kvpairs));
String8 key = String8(AudioParameter::keyRouting);
int tmpInt;
if (param.getInt(key, tmpInt) == NO_ERROR) {
// The audio HAL handles routing to physical devices entirely
// internally and mostly ignores what audio flinger tells it to do. JiC
// there is something (now or in the future) in audio flinger which
// cares about the routing value in a call to getParameters, we hang on
// to the last routing value set by audio flinger so we can at least be
// consistent when we lie to the upper levels about doing what they told
// us to do.
mAudioFlingerTgtDevices = static_cast<uint32_t>(tmpInt);
return NO_ERROR;
char* AudioStreamOut::getParameters(const char* k)
AudioParameter param = AudioParameter(String8(k));
String8 value;
if (param.get(keyRouting, value) == NO_ERROR) {
param.addInt(keyRouting, (int)mAudioFlingerTgtDevices);
HDMIAudioCaps& hdmiCaps = mOwnerHAL.getHDMIAudioCaps();
if (param.get(keySupSampleRates, value) == NO_ERROR) {
if (mIsMCOutput) {
param.add(keySupSampleRates, value);
} else {
param.add(keySupSampleRates, String8("48000"));
if (param.get(keySupFormats, value) == NO_ERROR) {
if (mIsMCOutput) {
param.add(keySupFormats, value);
} else {
param.add(keySupFormats, String8("AUDIO_FORMAT_PCM_16_BIT|AUDIO_FORMAT_PCM_8_24_BIT"));
if (param.get(keySupChannels, value) == NO_ERROR) {
if (mIsMCOutput) {
param.add(keySupChannels, value);
} else {
param.add(keySupChannels, String8("AUDIO_CHANNEL_OUT_STEREO"));
return strdup(param.toString().string());
uint32_t AudioStreamOut::outputSampleRate() const
return mInputSampleRate;
uint32_t AudioStreamOut::latency() const {
uint32_t uSecLatency = mInputNominalLatencyUSec;
uint32_t vcompDelay = mOwnerHAL.getVideoDelayCompUsec();
if (uSecLatency < vcompDelay)
return 0;
return ((uSecLatency - vcompDelay) / 1000);
// Used to implement get_presentation_position() for Audio HAL.
// According to the prototype in audio.h, the frame count should not get
// reset on standby().
status_t AudioStreamOut::getPresentationPosition(uint64_t *frames,
struct timespec *timestamp)
Mutex::Autolock _l(mRoutingLock);
status_t result = -ENODEV;
// The presentation timestamp should be the same for all devices.
// Also Molly only has one output device at the moment.
// So just use the first one in the list.
if (!mPhysOutputs.isEmpty()) {
const unsigned int kInsaneAvail = 10 * 48000;
unsigned int avail = 0;
sp<AudioOutput> audioOutput = mPhysOutputs.itemAt(0);
if (audioOutput->getHardwareTimestamp(&avail, timestamp) == 0) {
if (avail < kInsaneAvail) {
// FIXME av sync fudge factor
// Use a fudge factor to account for hidden buffering in the
// HDMI output path. This is a hack until we can determine the
// actual buffer sizes.
// Increasing kFudgeMSec will move the audio earlier in
// relation to the video.
const int kFudgeMSec = 50;
int fudgeFrames = kFudgeMSec * sampleRate() / 1000;
int64_t framesInDriverBuffer =
(int64_t)audioOutput->getKernelBufferSize() - (int64_t)avail;
int64_t pendingFrames = framesInDriverBuffer + fudgeFrames;
int64_t signedFrames = mFramesRendered - pendingFrames;
if (pendingFrames < 0) {
ALOGE("getPresentationPosition: negative pendingFrames = %lld",
} else if (signedFrames < 0) {
ALOGI("getPresentationPosition: playing silent preroll"
", mFramesRendered = %llu, pendingFrames = %lld",
mFramesRendered, pendingFrames);
} else {
// Print comma separated values for spreadsheet analysis.
uint64_t nanos = (((uint64_t)timestamp->tv_sec) * 1000000000L)
+ timestamp->tv_nsec;
ALOGI("getPresentationPosition, %lld, %4u, %lld, %llu",
mFramesRendered, avail, signedFrames, nanos);
*frames = (uint64_t) signedFrames;
result = NO_ERROR;
} else {
ALOGE("getPresentationPosition: avail too large = %u", avail);
mReportedAvailFail = false;
} else {
"getPresentationPosition: getHardwareTimestamp returned non-zero");
mReportedAvailFail = true;
} else {
ALOGVV("getPresentationPosition: no physical outputs! This HAL is inactive!");
return result;
status_t AudioStreamOut::getRenderPosition(__unused uint32_t *dspFrames)
if (dspFrames == NULL) {
return -EINVAL;
*dspFrames = (uint32_t) mFramesRendered;
return NO_ERROR;
void AudioStreamOut::updateTargetOutputs()
Mutex::Autolock _l(mRoutingLock);
AudioOutputList::iterator I;
uint32_t cur_outputs = 0;
for (I = mPhysOutputs.begin(); I != mPhysOutputs.end(); ++I)
cur_outputs |= (*I)->devMask();
if (cur_outputs == mTgtDevices)
uint32_t outputsToObtain = mTgtDevices & ~cur_outputs;
uint32_t outputsToRelease = cur_outputs & ~mTgtDevices;
// Start by releasing any outputs we should no longer have back to the HAL.
if (outputsToRelease) {
I = mPhysOutputs.begin();
while (I != mPhysOutputs.end()) {
if (!(outputsToRelease & (*I)->devMask())) {
outputsToRelease &= ~((*I)->devMask());
mOwnerHAL.releaseOutput(*this, *I);
I = mPhysOutputs.erase(I);
if (outputsToRelease) {
ALOGW("Bookkeeping error! Still have outputs to release (%08x), but"
" none of them appear to be in the mPhysOutputs list!",
// Now attempt to obtain any outputs we should be using, but are not
// currently.
if (outputsToObtain) {
uint32_t mask;
// Buffer configuration may need updating now that we have decoded
// the start of a stream. For example, EAC3, needs 4X sampleRate.
for (mask = 0x1; outputsToObtain; mask <<= 1) {
if (!(mask & outputsToObtain))
sp<AudioOutput> newOutput;
status_t res;
res = mOwnerHAL.obtainOutput(*this, mask, &newOutput);
outputsToObtain &= ~mask;
if (OK != res) {
// If we get an error back from obtain output, it means that
// something went really wrong at a lower level (probably failed
// to open the driver). We should not try to obtain this output
// again, at least until the next routing change.
ALOGW("Failed to obtain output %08x for %s audio stream out."
" (res %d)", mask, getName(), res);
mTgtDevices &= ~mask;
if (newOutput != NULL) {
// If we actually got an output, go ahead and add it to our list
// of physical outputs. The rest of the system will handle
// starting it up. If we didn't get an output, but also go no
// error code, it just means that the output is currently busy
// and should become available soon.
ALOGI("updateTargetOutputs: adding output back to mPhysOutputs");
void AudioStreamOut::adjustOutputs(int64_t maxTime)
int64_t a_zero_original = mLocalTimeToFrames.a_zero;
int64_t b_zero_original = mLocalTimeToFrames.b_zero;
AudioOutputList::iterator I;
// Check to see if any outputs are active and see what their buffer levels
// are.
for (I = mPhysOutputs.begin(); I != mPhysOutputs.end(); ++I) {
if ((*I)->getState() == AudioOutput::DMA_START) {
int64_t lastWriteTS = (*I)->getLastNextWriteTS();
int64_t padAmt;
mLocalTimeToFrames.a_zero = lastWriteTS;
mLocalTimeToFrames.b_zero = 0;
if (mLocalTimeToFrames.doForwardTransform(maxTime,
&padAmt)) {
// Restore original offset so that the sleep time calculation for
// throttling is not broken in finishedWriteOp().
mLocalTimeToFrames.a_zero = a_zero_original;
mLocalTimeToFrames.b_zero = b_zero_original;
ssize_t AudioStreamOut::write(const void* buffer, size_t bytes)
uint8_t *data = (uint8_t *)buffer;
ALOGVV("AudioStreamOut::write_l(%u) 0x%02X, 0x%02X, 0x%02X, 0x%02X,"
" 0x%02X, 0x%02X, 0x%02X, 0x%02X,"
" 0x%02X, 0x%02X, 0x%02X, 0x%02X,"
" 0x%02X, 0x%02X, 0x%02X, 0x%02X",
bytes, data[0], data[1], data[2], data[3],
data[4], data[5], data[6], data[7],
data[8], data[9], data[10], data[11],
data[12], data[13], data[14], data[15]
// Note: no lock is obtained here. Calls to write and getNextWriteTimestamp
// happen only on the AudioFlinger mixer thread which owns this particular
// output stream, so there is no need to worry that there will be two
// threads in this instance method concurrently.
// In addition, only calls to write change the contents of the mPhysOutputs
// collection (during the call to updateTargetOutputs). updateTargetOutputs
// will hold the routing lock during the operation, as should any reader of
// mPhysOutputs, unless the reader is a call to write or
// getNextWriteTimestamp (we know that it is safe for write and gnwt to read
// the collection because the only collection mutator is the same thread
// which calls write and gnwt).
// If the stream is in standby, then the first write should bring it out
// of standby
if (mInStandby) {
mOwnerHAL.standbyStatusUpdate(false, mIsMCOutput);
mInStandby = false;
// If any of our outputs is in the PRIMED state when ::write is called, it
// means one of two things. First, it could be that the DMA output really
// has not started yet. This is odd, but certainly not impossible. The
// other possibility is that AudioFlinger is in its silence-pushing mode and
// is not calling getNextWriteTimestamp. After an output is primed, its in
// GNWTS where the amount of padding to compensate for different DMA start
// times is taken into account. Go ahead and force a call to GNWTS, just to
// be certain that we have checked recently and are not stuck in silence
// fill mode. Failure to do this will cause the AudioOutput state machine
// to eventually give up on DMA starting and reset the output over and over
// again (spamming the log and producing general confusion).
// While we are in the process of checking our various output states, check
// to see if any outputs have made it to the ACTIVE state. Pass this
// information along to the call to processOneChunk. If any of our outputs
// are waiting to be primed while other outputs have made it to steady
// state, we need to change our priming behavior slightly. Instead of
// filling an output's buffer completely, we want to fill it to slightly
// less than full and let the adjustDelay mechanism take care of the rest.
// Failure to do this during steady state operation will almost certainly
// lead to the new output being over-filled relative to the other outputs
// causing it to be slightly out of sync.
AudioOutputList::iterator I;
bool checkDMAStart = false;
bool hasActiveOutputs = false;
for (I = mPhysOutputs.begin(); I != mPhysOutputs.end(); ++I) {
if (AudioOutput::PRIMED == (*I)->getState())
checkDMAStart = true;
if ((*I)->getState() == AudioOutput::ACTIVE)
hasActiveOutputs = true;
if (checkDMAStart) {
int64_t junk;
// We always call processOneChunk on the outputs, as it is the
// tick for their state machines.
for (I = mPhysOutputs.begin(); I != mPhysOutputs.end(); ++I) {
(*I)->processOneChunk((uint8_t *)buffer, bytes, hasActiveOutputs, mInputFormat);
// If we don't actually have any physical outputs to write to, just sleep
// for the proper amt of time in order to simulate the throttle that writing
// to the hardware would impose.
finishedWriteOp(bytes / mInputFrameSize, (0 == mPhysOutputs.size()));
return static_cast<ssize_t>(bytes);
status_t AudioStreamOut::getNextWriteTimestamp(int64_t *timestamp)
return getNextWriteTimestamp_internal(timestamp);
status_t AudioStreamOut::getNextWriteTimestamp_internal(
int64_t *timestamp)
int64_t max_time = LLONG_MIN;
bool max_time_valid = false;
bool need_adjust = false;
// Across all of our physical outputs, figure out the max time when
// a write operation will hit the speakers. Assume that if an
// output cannot answer the question, its because it has never
// started or because it has recently underflowed and needs to be
// restarted. If this is the case, we will need to prime the
// pipeline with a chunk's worth of data before proceeding.
// If any of the outputs indicate a discontinuity (meaning that the
// DMA start time was valid and is now invalid, or was and is valid
// but was different from before; almost certainly caused by a low
// level underfow), then just stop now. We will need to reset and
// re-prime all of the outputs in order to make certain that the
// lead-times on all of the outputs match.
AudioOutputList::iterator I;
bool discon = false;
// Find the largest next write timestamp. The goal is to make EVERY
// output have the same value, but we also need this to pass back
// up the layers.
for (I = mPhysOutputs.begin(); I != mPhysOutputs.end(); ++I) {
int64_t tmp;
if (OK == (*I)->getNextWriteTimestamp(&tmp, &discon)) {
if (!max_time_valid || (max_time < tmp)) {
max_time = tmp;
max_time_valid = true;
// Check the state of each output and determine if we need to align them.
// Make sure to do this after we have called each outputs'
// getNextWriteTimestamp as the transition from PRIMED to DMA_START happens
// there.
for (I = mPhysOutputs.begin(); I != mPhysOutputs.end(); ++I) {
if ((*I)->getState() == AudioOutput::DMA_START) {
need_adjust = true;
// At this point, if we still have not found at least one output
// who knows when their data is going to hit the speakers, then we
// just can't answer the getNextWriteTimestamp question and we
// should give up.
if (!max_time_valid) {
// Stuff silence into the non-aligned outputs so that the effective
// timestamp is the same for all the outputs.
if (need_adjust)
// We are done. The time at which the next written audio should
// hit the speakers is just max_time plus the maximum amt of delay
// compensation in the system.
*timestamp = max_time;
return OK;
#define DUMP(a...) \
snprintf(buffer, SIZE, a); \
buffer[SIZE - 1] = 0; \
#define B2STR(b) b ? "true" : "false"
status_t AudioStreamOut::dump(int fd)
const size_t SIZE = 256;
char buffer[SIZE];
String8 result;
DUMP("\n%s AudioStreamOut::dump\n", getName());
DUMP("\tsample rate : %d\n", sampleRate());
DUMP("\tbuffer size : %d\n", bufferSize());
DUMP("\tchannel mask : 0x%04x\n", chanMask());
DUMP("\tformat : %d\n", format());
DUMP("\tdevice mask : 0x%04x\n", mTgtDevices);
DUMP("\tIn standby : %s\n", mInStandby? "yes" : "no");
AudioOutputList outSnapshot(mPhysOutputs);
AudioOutputList::iterator I;
for (I = outSnapshot.begin(); I != outSnapshot.end(); ++I)
::write(fd, result.string(), result.size());
return NO_ERROR;
#undef B2STR
#undef DUMP
} // android