blob: ae0457f8c3f391fa3293db3240e56ed6fd62f667 [file] [log] [blame]
/*
* Copyright (C) 2019 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_TAG "AudioPort"
#include <algorithm>
#include <utility>
#include <android-base/stringprintf.h>
#include <media/AudioPort.h>
#include <utils/Log.h>
namespace android {
void AudioPort::setFlags(uint32_t flags)
{
// force direct flag if offload flag is set: offloading implies a direct output stream
// and all common behaviors are driven by checking only the direct flag
// this should normally be set appropriately in the policy configuration file
if (mRole == AUDIO_PORT_ROLE_SOURCE &&
(flags & AUDIO_OUTPUT_FLAG_COMPRESS_OFFLOAD) != 0) {
flags |= AUDIO_OUTPUT_FLAG_DIRECT;
}
if (useInputChannelMask()) {
mFlags.input = static_cast<audio_input_flags_t>(flags);
} else {
mFlags.output = static_cast<audio_output_flags_t>(flags);
}
}
void AudioPort::importAudioPort(const sp<AudioPort>& port, bool force __unused)
{
for (const auto& profileToImport : port->mProfiles) {
// Import only valid port, i.e. valid format, non empty rates and channels masks
if (!profileToImport->isValid()) {
continue;
}
if (std::find_if(mProfiles.begin(), mProfiles.end(),
[profileToImport](const auto &profile) {
return *profile == *profileToImport; }) == mProfiles.end()) {
addAudioProfile(profileToImport);
}
}
}
void AudioPort::importAudioPort(const audio_port_v7 &port) {
for (size_t i = 0; i < port.num_audio_profiles; ++i) {
if (port.audio_profiles[i].format == AUDIO_FORMAT_DEFAULT) {
// The dynamic format from AudioPort should not be AUDIO_FORMAT_DEFAULT.
continue;
}
sp<AudioProfile> profile = new AudioProfile(port.audio_profiles[i].format,
ChannelMaskSet(port.audio_profiles[i].channel_masks,
port.audio_profiles[i].channel_masks +
port.audio_profiles[i].num_channel_masks),
SampleRateSet(port.audio_profiles[i].sample_rates,
port.audio_profiles[i].sample_rates +
port.audio_profiles[i].num_sample_rates),
port.audio_profiles[i].encapsulation_type);
profile->setDynamicFormat(true);
profile->setDynamicChannels(true);
profile->setDynamicRate(true);
if (!mProfiles.contains(profile)) {
addAudioProfile(profile);
}
}
for (size_t i = 0; i < port.num_extra_audio_descriptors; ++i) {
auto convertedResult = legacy2aidl_audio_extra_audio_descriptor_ExtraAudioDescriptor(
port.extra_audio_descriptors[i]);
if (!convertedResult.ok()) {
ALOGE("%s, failed to convert extra audio descriptor", __func__);
continue;
}
if (std::find(mExtraAudioDescriptors.begin(),
mExtraAudioDescriptors.end(),
convertedResult.value()) == mExtraAudioDescriptors.end()) {
mExtraAudioDescriptors.push_back(std::move(convertedResult.value()));
}
}
}
void AudioPort::toAudioPort(struct audio_port *port) const {
// TODO: update this function once audio_port structure reflects the new profile definition.
// For compatibility reason: flatening the AudioProfile into audio_port structure.
FormatSet flatenedFormats;
SampleRateSet flatenedRates;
ChannelMaskSet flatenedChannels;
for (const auto& profile : mProfiles) {
if (profile->isValid()) {
audio_format_t formatToExport = profile->getFormat();
const SampleRateSet &ratesToExport = profile->getSampleRates();
const ChannelMaskSet &channelsToExport = profile->getChannels();
flatenedFormats.insert(formatToExport);
flatenedRates.insert(ratesToExport.begin(), ratesToExport.end());
flatenedChannels.insert(channelsToExport.begin(), channelsToExport.end());
if (flatenedRates.size() > AUDIO_PORT_MAX_SAMPLING_RATES ||
flatenedChannels.size() > AUDIO_PORT_MAX_CHANNEL_MASKS ||
flatenedFormats.size() > AUDIO_PORT_MAX_FORMATS) {
ALOGE("%s: bailing out: cannot export profiles to port config", __func__);
return;
}
}
}
toAudioPortBase(port);
port->num_sample_rates = flatenedRates.size();
port->num_channel_masks = flatenedChannels.size();
port->num_formats = flatenedFormats.size();
std::copy(flatenedRates.begin(), flatenedRates.end(), port->sample_rates);
std::copy(flatenedChannels.begin(), flatenedChannels.end(), port->channel_masks);
std::copy(flatenedFormats.begin(), flatenedFormats.end(), port->formats);
}
void AudioPort::toAudioPort(struct audio_port_v7 *port) const {
toAudioPortBase(port);
port->num_audio_profiles = 0;
for (const auto& profile : mProfiles) {
if (profile->isValid()) {
const SampleRateSet &sampleRates = profile->getSampleRates();
const ChannelMaskSet &channelMasks = profile->getChannels();
if (sampleRates.size() > AUDIO_PORT_MAX_SAMPLING_RATES ||
channelMasks.size() > AUDIO_PORT_MAX_CHANNEL_MASKS ||
port->num_audio_profiles >= AUDIO_PORT_MAX_AUDIO_PROFILES) {
ALOGE("%s: bailing out: cannot export profiles to port config", __func__);
break;
}
auto& dstProfile = port->audio_profiles[port->num_audio_profiles++];
dstProfile.format = profile->getFormat();
dstProfile.num_sample_rates = sampleRates.size();
std::copy(sampleRates.begin(), sampleRates.end(),
std::begin(dstProfile.sample_rates));
dstProfile.num_channel_masks = channelMasks.size();
std::copy(channelMasks.begin(), channelMasks.end(),
std::begin(dstProfile.channel_masks));
dstProfile.encapsulation_type = profile->getEncapsulationType();
}
}
port->num_extra_audio_descriptors = 0;
for (const auto& desc : mExtraAudioDescriptors) {
if (port->num_extra_audio_descriptors >= AUDIO_PORT_MAX_EXTRA_AUDIO_DESCRIPTORS) {
ALOGE("%s: bailing out: cannot export extra audio descriptor to port config", __func__);
return;
}
auto convertedResult = aidl2legacy_ExtraAudioDescriptor_audio_extra_audio_descriptor(desc);
if (!convertedResult.ok()) {
ALOGE("%s: failed to convert extra audio descriptor", __func__);
continue;
}
port->extra_audio_descriptors[port->num_extra_audio_descriptors++] =
std::move(convertedResult.value());
}
}
void AudioPort::dump(std::string *dst, int spaces, const char* extraInfo, bool verbose) const {
if (!mName.empty()) {
dst->append(base::StringPrintf("\"%s\"%s", mName.c_str(),
extraInfo != nullptr ? "; " : ""));
}
if (extraInfo != nullptr) {
dst->append(base::StringPrintf("%s", extraInfo));
}
if (!mName.empty() || extraInfo != nullptr) {
dst->append("\n");
}
if (verbose) {
std::string profilesStr;
mProfiles.dump(&profilesStr, spaces);
dst->append(profilesStr);
if (!mExtraAudioDescriptors.empty()) {
dst->append(base::StringPrintf("%*s- extra audio descriptors: \n", spaces, ""));
const int eadSpaces = spaces + 4;
const int descSpaces = eadSpaces + 4;
for (size_t i = 0; i < mExtraAudioDescriptors.size(); i++) {
dst->append(
base::StringPrintf("%*s extra audio descriptor %zu:\n", eadSpaces, "", i));
dst->append(base::StringPrintf(
"%*s- standard: %u\n", descSpaces, "", mExtraAudioDescriptors[i].standard));
dst->append(base::StringPrintf("%*s- descriptor:", descSpaces, ""));
for (auto v : mExtraAudioDescriptors[i].audioDescriptor) {
dst->append(base::StringPrintf(" %02x", v));
}
dst->append("\n");
}
}
if (mGains.size() != 0) {
dst->append(base::StringPrintf("%*s- gains:\n", spaces, ""));
for (size_t i = 0; i < mGains.size(); i++) {
std::string gainStr;
mGains[i]->dump(&gainStr, spaces + 2, i);
dst->append(gainStr);
}
}
}
}
void AudioPort::log(const char* indent) const
{
ALOGI("%s Port[nm:%s, type:%d, role:%d]", indent, mName.c_str(), mType, mRole);
}
bool AudioPort::equals(const sp<AudioPort> &other) const
{
return other != nullptr &&
mGains.equals(other->getGains()) &&
mName.compare(other->getName()) == 0 &&
mType == other->getType() &&
mRole == other->getRole() &&
mProfiles.equals(other->getAudioProfiles()) &&
getFlags() == other->getFlags() &&
mExtraAudioDescriptors == other->getExtraAudioDescriptors();
}
status_t AudioPort::writeToParcelable(media::AudioPortFw* parcelable) const {
parcelable->hal.name = mName;
parcelable->sys.type = VALUE_OR_RETURN_STATUS(
legacy2aidl_audio_port_type_t_AudioPortType(mType));
parcelable->sys.role = VALUE_OR_RETURN_STATUS(
legacy2aidl_audio_port_role_t_AudioPortRole(mRole));
auto aidlProfiles = VALUE_OR_RETURN_STATUS(
legacy2aidl_AudioProfileVector(mProfiles, useInputChannelMask()));
parcelable->hal.profiles = aidlProfiles.first;
parcelable->sys.profiles = aidlProfiles.second;
parcelable->hal.flags = VALUE_OR_RETURN_STATUS(
legacy2aidl_audio_io_flags_AudioIoFlags(mFlags, useInputChannelMask()));
parcelable->hal.extraAudioDescriptors = mExtraAudioDescriptors;
auto aidlGains = VALUE_OR_RETURN_STATUS(legacy2aidl_AudioGains(mGains));
parcelable->hal.gains = aidlGains.first;
parcelable->sys.gains = aidlGains.second;
if (mType == AUDIO_PORT_TYPE_MIX) {
media::audio::common::AudioPortMixExt mixExt{};
mixExt.maxOpenStreamCount = maxOpenCount;
mixExt.maxActiveStreamCount = maxActiveCount;
mixExt.recommendedMuteDurationMs = recommendedMuteDurationMs;
parcelable->hal.ext = media::audio::common::AudioPortExt::make<
media::audio::common::AudioPortExt::mix>(mixExt);
}
return OK;
}
status_t AudioPort::readFromParcelable(const media::AudioPortFw& parcelable) {
mName = parcelable.hal.name;
mType = VALUE_OR_RETURN_STATUS(
aidl2legacy_AudioPortType_audio_port_type_t(parcelable.sys.type));
mRole = VALUE_OR_RETURN_STATUS(
aidl2legacy_AudioPortRole_audio_port_role_t(parcelable.sys.role));
mProfiles = VALUE_OR_RETURN_STATUS(
aidl2legacy_AudioProfileVector(
std::make_pair(parcelable.hal.profiles, parcelable.sys.profiles),
useInputChannelMask()));
mFlags = VALUE_OR_RETURN_STATUS(
aidl2legacy_AudioIoFlags_audio_io_flags(parcelable.hal.flags, useInputChannelMask()));
mExtraAudioDescriptors = parcelable.hal.extraAudioDescriptors;
mGains = VALUE_OR_RETURN_STATUS(
aidl2legacy_AudioGains(std::make_pair(parcelable.hal.gains, parcelable.sys.gains)));
if (mType == AUDIO_PORT_TYPE_MIX) {
const media::audio::common::AudioPortMixExt& mixExt =
parcelable.hal.ext.get<media::audio::common::AudioPortExt::mix>();
maxOpenCount = mixExt.maxOpenStreamCount;
maxActiveCount = mixExt.maxActiveStreamCount;
recommendedMuteDurationMs = mixExt.recommendedMuteDurationMs;
}
return OK;
}
// --- AudioPortConfig class implementation
status_t AudioPortConfig::applyAudioPortConfig(
const struct audio_port_config *config,
struct audio_port_config *backupConfig __unused)
{
if (config->config_mask & AUDIO_PORT_CONFIG_SAMPLE_RATE) {
mSamplingRate = config->sample_rate;
}
if (config->config_mask & AUDIO_PORT_CONFIG_CHANNEL_MASK) {
mChannelMask = config->channel_mask;
}
if (config->config_mask & AUDIO_PORT_CONFIG_FORMAT) {
mFormat = config->format;
}
if (config->config_mask & AUDIO_PORT_CONFIG_GAIN) {
mGain = config->gain;
}
if (config->config_mask & AUDIO_PORT_CONFIG_FLAGS) {
mFlags = config->flags;
}
return NO_ERROR;
}
namespace {
template<typename T>
void updateField(
const T& portConfigField, T audio_port_config::*port_config_field,
struct audio_port_config *dstConfig, const struct audio_port_config *srcConfig,
unsigned int configMask, T defaultValue)
{
if (dstConfig->config_mask & configMask) {
if ((srcConfig != nullptr) && (srcConfig->config_mask & configMask)) {
dstConfig->*port_config_field = srcConfig->*port_config_field;
} else {
dstConfig->*port_config_field = portConfigField;
}
} else {
dstConfig->*port_config_field = defaultValue;
}
}
} // namespace
void AudioPortConfig::toAudioPortConfig(
struct audio_port_config *dstConfig,
const struct audio_port_config *srcConfig) const
{
updateField(mSamplingRate, &audio_port_config::sample_rate,
dstConfig, srcConfig, AUDIO_PORT_CONFIG_SAMPLE_RATE, 0u);
updateField(mChannelMask, &audio_port_config::channel_mask,
dstConfig, srcConfig, AUDIO_PORT_CONFIG_CHANNEL_MASK,
(audio_channel_mask_t)AUDIO_CHANNEL_NONE);
updateField(mFormat, &audio_port_config::format,
dstConfig, srcConfig, AUDIO_PORT_CONFIG_FORMAT, AUDIO_FORMAT_INVALID);
dstConfig->id = mId;
sp<AudioPort> audioport = getAudioPort();
if ((dstConfig->config_mask & AUDIO_PORT_CONFIG_GAIN) && audioport != NULL) {
dstConfig->gain = mGain;
if ((srcConfig != NULL) && (srcConfig->config_mask & AUDIO_PORT_CONFIG_GAIN)
&& audioport->checkGain(&srcConfig->gain, srcConfig->gain.index) == OK) {
dstConfig->gain = srcConfig->gain;
}
} else {
dstConfig->gain.index = -1;
}
if (dstConfig->gain.index != -1) {
dstConfig->config_mask |= AUDIO_PORT_CONFIG_GAIN;
} else {
dstConfig->config_mask &= ~AUDIO_PORT_CONFIG_GAIN;
}
updateField(mFlags, &audio_port_config::flags,
dstConfig, srcConfig, AUDIO_PORT_CONFIG_FLAGS, { AUDIO_INPUT_FLAG_NONE });
}
bool AudioPortConfig::hasGainController(bool canUseForVolume) const
{
sp<AudioPort> audioport = getAudioPort();
if (!audioport) {
return false;
}
return canUseForVolume ? audioport->getGains().canUseForVolume()
: audioport->getGains().size() > 0;
}
bool AudioPortConfig::equals(const sp<AudioPortConfig> &other, bool isInput) const
{
return other != nullptr &&
mSamplingRate == other->getSamplingRate() &&
mFormat == other->getFormat() &&
mChannelMask == other->getChannelMask() &&
(isInput ? mFlags.input == other->getFlags().input :
mFlags.output == other->getFlags().output )&&
// Compare audio gain config
mGain.index == other->mGain.index &&
mGain.mode == other->mGain.mode &&
mGain.channel_mask == other->mGain.channel_mask &&
std::equal(std::begin(mGain.values), std::end(mGain.values),
std::begin(other->mGain.values)) &&
mGain.ramp_duration_ms == other->mGain.ramp_duration_ms;
}
status_t AudioPortConfig::writeToParcelable(
media::audio::common::AudioPortConfig* parcelable, bool isInput) const {
media::audio::common::Int aidl_sampleRate;
aidl_sampleRate.value = VALUE_OR_RETURN_STATUS(convertIntegral<int32_t>(mSamplingRate));
parcelable->sampleRate = aidl_sampleRate;
parcelable->format = VALUE_OR_RETURN_STATUS(
legacy2aidl_audio_format_t_AudioFormatDescription(mFormat));
parcelable->channelMask = VALUE_OR_RETURN_STATUS(
legacy2aidl_audio_channel_mask_t_AudioChannelLayout(mChannelMask, isInput));
parcelable->id = VALUE_OR_RETURN_STATUS(legacy2aidl_audio_port_handle_t_int32_t(mId));
media::audio::common::AudioGainConfig aidl_gain = VALUE_OR_RETURN_STATUS(
legacy2aidl_audio_gain_config_AudioGainConfig(mGain, isInput));
parcelable->gain = aidl_gain;
parcelable->flags = VALUE_OR_RETURN_STATUS(
legacy2aidl_audio_io_flags_AudioIoFlags(mFlags, isInput));
return OK;
}
status_t AudioPortConfig::readFromParcelable(
const media::audio::common::AudioPortConfig& parcelable, bool isInput) {
if (parcelable.sampleRate.has_value()) {
mSamplingRate = VALUE_OR_RETURN_STATUS(
convertIntegral<unsigned int>(parcelable.sampleRate.value().value));
}
if (parcelable.format.has_value()) {
mFormat = VALUE_OR_RETURN_STATUS(
aidl2legacy_AudioFormatDescription_audio_format_t(parcelable.format.value()));
}
if (parcelable.channelMask.has_value()) {
mChannelMask = VALUE_OR_RETURN_STATUS(
aidl2legacy_AudioChannelLayout_audio_channel_mask_t(
parcelable.channelMask.value(), isInput));
}
mId = VALUE_OR_RETURN_STATUS(aidl2legacy_int32_t_audio_port_handle_t(parcelable.id));
if (parcelable.gain.has_value()) {
mGain = VALUE_OR_RETURN_STATUS(
aidl2legacy_AudioGainConfig_audio_gain_config(parcelable.gain.value(), isInput));
}
if (parcelable.flags.has_value()) {
mFlags = VALUE_OR_RETURN_STATUS(
aidl2legacy_AudioIoFlags_audio_io_flags(parcelable.flags.value(), isInput));
}
return OK;
}
} // namespace android