services/audiopolicy/managerdefault/AudioPolicyManager.cpp

/*
 * Copyright (C) 2009 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 "APM_AudioPolicyManager"

// Need to keep the log statements even in production builds
// to enable VERBOSE logging dynamically.
// You can enable VERBOSE logging as follows:
// adb shell setprop log.tag.APM_AudioPolicyManager V
#define LOG_NDEBUG 0

//#define VERY_VERBOSE_LOGGING
#ifdef VERY_VERBOSE_LOGGING
#define ALOGVV ALOGV
#else
#define ALOGVV(a...) do { } while(0)
#endif

#define AUDIO_POLICY_XML_CONFIG_FILE_PATH_MAX_LENGTH 128
#define AUDIO_POLICY_XML_CONFIG_FILE_NAME "audio_policy_configuration.xml"
#define AUDIO_POLICY_A2DP_OFFLOAD_DISABLED_XML_CONFIG_FILE_NAME \
        "audio_policy_configuration_a2dp_offload_disabled.xml"
#define AUDIO_POLICY_BLUETOOTH_LEGACY_HAL_XML_CONFIG_FILE_NAME \
        "audio_policy_configuration_bluetooth_legacy_hal.xml"

#include <algorithm>
#include <inttypes.h>
#include <math.h>
#include <set>
#include <unordered_set>
#include <vector>
#include <cutils/properties.h>
#include <utils/Log.h>
#include <media/AudioParameter.h>
#include <private/android_filesystem_config.h>
#include <soundtrigger/SoundTrigger.h>
#include <system/audio.h>
#include "AudioPolicyManager.h"
#include <Serializer.h>
#include "TypeConverter.h"
#include <policy.h>

namespace android {

//FIXME: workaround for truncated touch sounds
// to be removed when the problem is handled by system UI
#define TOUCH_SOUND_FIXED_DELAY_MS 100

// Largest difference in dB on earpiece in call between the voice volume and another
// media / notification / system volume.
constexpr float IN_CALL_EARPIECE_HEADROOM_DB = 3.f;

// Compressed formats for MSD module, ordered from most preferred to least preferred.
static const std::vector<audio_format_t> compressedFormatsOrder = {{
        AUDIO_FORMAT_MAT_2_1, AUDIO_FORMAT_MAT_2_0, AUDIO_FORMAT_E_AC3,
        AUDIO_FORMAT_AC3, AUDIO_FORMAT_PCM_16_BIT }};
// Channel masks for MSD module, 3D > 2D > 1D ordering (most preferred to least preferred).
static const std::vector<audio_channel_mask_t> surroundChannelMasksOrder = {{
        AUDIO_CHANNEL_OUT_3POINT1POINT2, AUDIO_CHANNEL_OUT_3POINT0POINT2,
        AUDIO_CHANNEL_OUT_2POINT1POINT2, AUDIO_CHANNEL_OUT_2POINT0POINT2,
        AUDIO_CHANNEL_OUT_5POINT1, AUDIO_CHANNEL_OUT_STEREO }};

// ----------------------------------------------------------------------------
// AudioPolicyInterface implementation
// ----------------------------------------------------------------------------

status_t AudioPolicyManager::setDeviceConnectionState(audio_devices_t device,
                                                      audio_policy_dev_state_t state,
                                                      const char *device_address,
                                                      const char *device_name,
                                                      audio_format_t encodedFormat)
{
    status_t status = setDeviceConnectionStateInt(device, state, device_address,
                                                  device_name, encodedFormat);
    nextAudioPortGeneration();
    return status;
}

void AudioPolicyManager::broadcastDeviceConnectionState(const sp<DeviceDescriptor> &device,
                                                        audio_policy_dev_state_t state)
{
    AudioParameter param(device->address());
    const String8 key(state == AUDIO_POLICY_DEVICE_STATE_AVAILABLE ?
                AudioParameter::keyStreamConnect : AudioParameter::keyStreamDisconnect);
    param.addInt(key, device->type());
    mpClientInterface->setParameters(AUDIO_IO_HANDLE_NONE, param.toString());
}

status_t AudioPolicyManager::setDeviceConnectionStateInt(audio_devices_t deviceType,
                                                         audio_policy_dev_state_t state,
                                                         const char *device_address,
                                                         const char *device_name,
                                                         audio_format_t encodedFormat)
{
    ALOGV("setDeviceConnectionStateInt() device: 0x%X, state %d, address %s name %s format 0x%X",
            deviceType, state, device_address, device_name, encodedFormat);

    // connect/disconnect only 1 device at a time
    if (!audio_is_output_device(deviceType) && !audio_is_input_device(deviceType)) return BAD_VALUE;

    sp<DeviceDescriptor> device =
            mHwModules.getDeviceDescriptor(deviceType, device_address, device_name, encodedFormat,
                                           state == AUDIO_POLICY_DEVICE_STATE_AVAILABLE);
    if (device == 0) {
        return INVALID_OPERATION;
    }

    // handle output devices
    if (audio_is_output_device(deviceType)) {
        SortedVector <audio_io_handle_t> outputs;

        ssize_t index = mAvailableOutputDevices.indexOf(device);

        // save a copy of the opened output descriptors before any output is opened or closed
        // by checkOutputsForDevice(). This will be needed by checkOutputForAllStrategies()
        mPreviousOutputs = mOutputs;
        switch (state)
        {
        // handle output device connection
        case AUDIO_POLICY_DEVICE_STATE_AVAILABLE: {
            if (index >= 0) {
                ALOGW("%s() device already connected: %s", __func__, device->toString().c_str());
                return INVALID_OPERATION;
            }
            ALOGV("%s() connecting device %s format %x",
                    __func__, device->toString().c_str(), encodedFormat);

            // register new device as available
            if (mAvailableOutputDevices.add(device) < 0) {
                return NO_MEMORY;
            }

            // Before checking outputs, broadcast connect event to allow HAL to retrieve dynamic
            // parameters on newly connected devices (instead of opening the outputs...)
            broadcastDeviceConnectionState(device, state);

            if (checkOutputsForDevice(device, state, outputs) != NO_ERROR) {
                mAvailableOutputDevices.remove(device);

                mHwModules.cleanUpForDevice(device);

                broadcastDeviceConnectionState(device, AUDIO_POLICY_DEVICE_STATE_UNAVAILABLE);
                return INVALID_OPERATION;
            }

            // outputs should never be empty here
            ALOG_ASSERT(outputs.size() != 0, "setDeviceConnectionState():"
                    "checkOutputsForDevice() returned no outputs but status OK");
            ALOGV("%s() checkOutputsForDevice() returned %zu outputs", __func__, outputs.size());

            } break;
        // handle output device disconnection
        case AUDIO_POLICY_DEVICE_STATE_UNAVAILABLE: {
            if (index < 0) {
                ALOGW("%s() device not connected: %s", __func__, device->toString().c_str());
                return INVALID_OPERATION;
            }

            ALOGV("%s() disconnecting output device %s", __func__, device->toString().c_str());

            // Send Disconnect to HALs
            broadcastDeviceConnectionState(device, state);

            // remove device from available output devices
            mAvailableOutputDevices.remove(device);

            mOutputs.clearSessionRoutesForDevice(device);

            checkOutputsForDevice(device, state, outputs);

            // Reset active device codec
            device->setEncodedFormat(AUDIO_FORMAT_DEFAULT);

            } break;

        default:
            ALOGE("%s() invalid state: %x", __func__, state);
            return BAD_VALUE;
        }

        // Propagate device availability to Engine
        setEngineDeviceConnectionState(device, state);

        // No need to evaluate playback routing when connecting a remote submix
        // output device used by a dynamic policy of type recorder as no
        // playback use case is affected.
        bool doCheckForDeviceAndOutputChanges = true;
        if (device->type() == AUDIO_DEVICE_OUT_REMOTE_SUBMIX
                && strncmp(device_address, "0", AUDIO_DEVICE_MAX_ADDRESS_LEN) != 0) {
            for (audio_io_handle_t output : outputs) {
                sp<SwAudioOutputDescriptor> desc = mOutputs.valueFor(output);
                sp<AudioPolicyMix> policyMix = desc->mPolicyMix.promote();
                if (policyMix != nullptr
                        && policyMix->mMixType == MIX_TYPE_RECORDERS
                        && strncmp(device_address,
                                   policyMix->mDeviceAddress.string(),
                                   AUDIO_DEVICE_MAX_ADDRESS_LEN) == 0) {
                    doCheckForDeviceAndOutputChanges = false;
                    break;
                }
            }
        }

        auto checkCloseOutputs = [&]() {
            // outputs must be closed after checkOutputForAllStrategies() is executed
            if (!outputs.isEmpty()) {
                for (audio_io_handle_t output : outputs) {
                    sp<SwAudioOutputDescriptor> desc = mOutputs.valueFor(output);
                    // close unused outputs after device disconnection or direct outputs that have
                    // been opened by checkOutputsForDevice() to query dynamic parameters
                    if ((state == AUDIO_POLICY_DEVICE_STATE_UNAVAILABLE) ||
                            (((desc->mFlags & AUDIO_OUTPUT_FLAG_DIRECT) != 0) &&
                                (desc->mDirectOpenCount == 0))) {
                        closeOutput(output);
                    }
                }
                // check A2DP again after closing A2DP output to reset mA2dpSuspended if needed
                return true;
            }
            return false;
        };

        if (doCheckForDeviceAndOutputChanges) {
            checkForDeviceAndOutputChanges(checkCloseOutputs);
        } else {
            checkCloseOutputs();
        }

        if (mEngine->getPhoneState() == AUDIO_MODE_IN_CALL && hasPrimaryOutput()) {
            DeviceVector newDevices = getNewOutputDevices(mPrimaryOutput, false /*fromCache*/);
            updateCallRouting(newDevices);
        }
        const DeviceVector msdOutDevices = getMsdAudioOutDevices();
        for (size_t i = 0; i < mOutputs.size(); i++) {
            sp<SwAudioOutputDescriptor> desc = mOutputs.valueAt(i);
            if ((mEngine->getPhoneState() != AUDIO_MODE_IN_CALL) || (desc != mPrimaryOutput)) {
                DeviceVector newDevices = getNewOutputDevices(desc, true /*fromCache*/);
                // do not force device change on duplicated output because if device is 0, it will
                // also force a device 0 for the two outputs it is duplicated to which may override
                // a valid device selection on those outputs.
                bool force = (msdOutDevices.isEmpty() || msdOutDevices != desc->devices())
                        && !desc->isDuplicated()
                        && (!device_distinguishes_on_address(deviceType)
                                // always force when disconnecting (a non-duplicated device)
                                || (state == AUDIO_POLICY_DEVICE_STATE_UNAVAILABLE));
                setOutputDevices(desc, newDevices, force, 0);
            }
        }

        if (state == AUDIO_POLICY_DEVICE_STATE_UNAVAILABLE) {
            cleanUpForDevice(device);
        }

        mpClientInterface->onAudioPortListUpdate();
        return NO_ERROR;
    }  // end if is output device

    // handle input devices
    if (audio_is_input_device(deviceType)) {
        ssize_t index = mAvailableInputDevices.indexOf(device);
        switch (state)
        {
        // handle input device connection
        case AUDIO_POLICY_DEVICE_STATE_AVAILABLE: {
            if (index >= 0) {
                ALOGW("%s() device already connected: %s", __func__, device->toString().c_str());
                return INVALID_OPERATION;
            }

            if (mAvailableInputDevices.add(device) < 0) {
                return NO_MEMORY;
            }

            // Before checking intputs, broadcast connect event to allow HAL to retrieve dynamic
            // parameters on newly connected devices (instead of opening the inputs...)
            broadcastDeviceConnectionState(device, state);

            if (checkInputsForDevice(device, state) != NO_ERROR) {
                mAvailableInputDevices.remove(device);

                broadcastDeviceConnectionState(device, AUDIO_POLICY_DEVICE_STATE_UNAVAILABLE);

                mHwModules.cleanUpForDevice(device);

                return INVALID_OPERATION;
            }

        } break;

        // handle input device disconnection
        case AUDIO_POLICY_DEVICE_STATE_UNAVAILABLE: {
            if (index < 0) {
                ALOGW("%s() device not connected: %s", __func__, device->toString().c_str());
                return INVALID_OPERATION;
            }

            ALOGV("%s() disconnecting input device %s", __func__, device->toString().c_str());

            // Set Disconnect to HALs
            broadcastDeviceConnectionState(device, state);

            mAvailableInputDevices.remove(device);

            checkInputsForDevice(device, state);
        } break;

        default:
            ALOGE("%s() invalid state: %x", __func__, state);
            return BAD_VALUE;
        }

        // Propagate device availability to Engine
        setEngineDeviceConnectionState(device, state);

        checkCloseInputs();
        // As the input device list can impact the output device selection, update
        // getDeviceForStrategy() cache
        updateDevicesAndOutputs();

        if (mEngine->getPhoneState() == AUDIO_MODE_IN_CALL && hasPrimaryOutput()) {
            DeviceVector newDevices = getNewOutputDevices(mPrimaryOutput, false /*fromCache*/);
            updateCallRouting(newDevices);
        }

        if (state == AUDIO_POLICY_DEVICE_STATE_UNAVAILABLE) {
            cleanUpForDevice(device);
        }

        mpClientInterface->onAudioPortListUpdate();
        return NO_ERROR;
    } // end if is input device

    ALOGW("%s() invalid device: %s", __func__, device->toString().c_str());
    return BAD_VALUE;
}

void AudioPolicyManager::setEngineDeviceConnectionState(const sp<DeviceDescriptor> device,
                                      audio_policy_dev_state_t state) {

    // the Engine does not have to know about remote submix devices used by dynamic audio policies
    if (audio_is_remote_submix_device(device->type()) && device->address() != "0") {
        return;
    }
    mEngine->setDeviceConnectionState(device, state);
}


audio_policy_dev_state_t AudioPolicyManager::getDeviceConnectionState(audio_devices_t device,
                                                                      const char *device_address)
{
    sp<DeviceDescriptor> devDesc =
            mHwModules.getDeviceDescriptor(device, device_address, "", AUDIO_FORMAT_DEFAULT,
                                           false /* allowToCreate */,
                                           (strlen(device_address) != 0)/*matchAddress*/);

    if (devDesc == 0) {
        ALOGV("getDeviceConnectionState() undeclared device, type %08x, address: %s",
              device, device_address);
        return AUDIO_POLICY_DEVICE_STATE_UNAVAILABLE;
    }

    DeviceVector *deviceVector;

    if (audio_is_output_device(device)) {
        deviceVector = &mAvailableOutputDevices;
    } else if (audio_is_input_device(device)) {
        deviceVector = &mAvailableInputDevices;
    } else {
        ALOGW("%s() invalid device type %08x", __func__, device);
        return AUDIO_POLICY_DEVICE_STATE_UNAVAILABLE;
    }

    return (deviceVector->getDevice(
                device, String8(device_address), AUDIO_FORMAT_DEFAULT) != 0) ?
            AUDIO_POLICY_DEVICE_STATE_AVAILABLE : AUDIO_POLICY_DEVICE_STATE_UNAVAILABLE;
}

status_t AudioPolicyManager::handleDeviceConfigChange(audio_devices_t device,
                                                      const char *device_address,
                                                      const char *device_name,
                                                      audio_format_t encodedFormat)
{
    status_t status;
    String8 reply;
    AudioParameter param;
    int isReconfigA2dpSupported = 0;

    ALOGV("handleDeviceConfigChange(() device: 0x%X, address %s name %s encodedFormat: 0x%X",
          device, device_address, device_name, encodedFormat);

    // connect/disconnect only 1 device at a time
    if (!audio_is_output_device(device) && !audio_is_input_device(device)) return BAD_VALUE;

    // Check if the device is currently connected
    DeviceVector deviceList = mAvailableOutputDevices.getDevicesFromTypeMask(device);
    if (deviceList.empty()) {
        // Nothing to do: device is not connected
        return NO_ERROR;
    }
    sp<DeviceDescriptor> devDesc = deviceList.itemAt(0);

    // For offloaded A2DP, Hw modules may have the capability to
    // configure codecs.
    // Handle two specific cases by sending a set parameter to
    // configure A2DP codecs. No need to toggle device state.
    // Case 1: A2DP active device switches from primary to primary
    // module
    // Case 2: A2DP device config changes on primary module.
    if (device & AUDIO_DEVICE_OUT_ALL_A2DP) {
        sp<HwModule> module = mHwModules.getModuleForDeviceTypes(device, encodedFormat);
        audio_module_handle_t primaryHandle = mPrimaryOutput->getModuleHandle();
        if (availablePrimaryOutputDevices().contains(devDesc) &&
           (module != 0 && module->getHandle() == primaryHandle)) {
            reply = mpClientInterface->getParameters(
                        AUDIO_IO_HANDLE_NONE,
                        String8(AudioParameter::keyReconfigA2dpSupported));
            AudioParameter repliedParameters(reply);
            repliedParameters.getInt(
                    String8(AudioParameter::keyReconfigA2dpSupported), isReconfigA2dpSupported);
            if (isReconfigA2dpSupported) {
                const String8 key(AudioParameter::keyReconfigA2dp);
                param.add(key, String8("true"));
                mpClientInterface->setParameters(AUDIO_IO_HANDLE_NONE, param.toString());
                devDesc->setEncodedFormat(encodedFormat);
                return NO_ERROR;
            }
        }
    }

    // Toggle the device state: UNAVAILABLE -> AVAILABLE
    // This will force reading again the device configuration
    status = setDeviceConnectionState(device,
                                      AUDIO_POLICY_DEVICE_STATE_UNAVAILABLE,
                                      device_address, device_name,
                                      devDesc->getEncodedFormat());
    if (status != NO_ERROR) {
        ALOGW("handleDeviceConfigChange() error disabling connection state: %d",
              status);
        return status;
    }

    status = setDeviceConnectionState(device,
                                      AUDIO_POLICY_DEVICE_STATE_AVAILABLE,
                                      device_address, device_name, encodedFormat);
    if (status != NO_ERROR) {
        ALOGW("handleDeviceConfigChange() error enabling connection state: %d",
              status);
        return status;
    }

    return NO_ERROR;
}

status_t AudioPolicyManager::getHwOffloadEncodingFormatsSupportedForA2DP(
                                    std::vector<audio_format_t> *formats)
{
    ALOGV("getHwOffloadEncodingFormatsSupportedForA2DP()");
    status_t status = NO_ERROR;
    std::unordered_set<audio_format_t> formatSet;
    sp<HwModule> primaryModule =
            mHwModules.getModuleFromName(AUDIO_HARDWARE_MODULE_ID_PRIMARY);
    DeviceVector declaredDevices = primaryModule->getDeclaredDevices().getDevicesFromTypeMask(
            AUDIO_DEVICE_OUT_ALL_A2DP);
    for (const auto& device : declaredDevices) {
        formatSet.insert(device->encodedFormats().begin(), device->encodedFormats().end());
    }
    formats->assign(formatSet.begin(), formatSet.end());
    return status;
}

uint32_t AudioPolicyManager::updateCallRouting(const DeviceVector &rxDevices, uint32_t delayMs)
{
    bool createTxPatch = false;
    bool createRxPatch = false;
    uint32_t muteWaitMs = 0;

    if(!hasPrimaryOutput() || mPrimaryOutput->devices().types() == AUDIO_DEVICE_OUT_STUB) {
        return muteWaitMs;
    }
    ALOG_ASSERT(!rxDevices.isEmpty(), "updateCallRouting() no selected output device");

    audio_attributes_t attr = { .source = AUDIO_SOURCE_VOICE_COMMUNICATION };
    auto txSourceDevice = mEngine->getInputDeviceForAttributes(attr);
    ALOG_ASSERT(txSourceDevice != 0, "updateCallRouting() input selected device not available");

    ALOGV("updateCallRouting device rxDevice %s txDevice %s",
          rxDevices.itemAt(0)->toString().c_str(), txSourceDevice->toString().c_str());

    // release existing RX patch if any
    if (mCallRxPatch != 0) {
        mpClientInterface->releaseAudioPatch(mCallRxPatch->mAfPatchHandle, 0);
        mCallRxPatch.clear();
    }
    // release TX patch if any
    if (mCallTxPatch != 0) {
        mpClientInterface->releaseAudioPatch(mCallTxPatch->mAfPatchHandle, 0);
        mCallTxPatch.clear();
    }

    auto telephonyRxModule =
        mHwModules.getModuleForDeviceTypes(AUDIO_DEVICE_IN_TELEPHONY_RX, AUDIO_FORMAT_DEFAULT);
    auto telephonyTxModule =
        mHwModules.getModuleForDeviceTypes(AUDIO_DEVICE_OUT_TELEPHONY_TX, AUDIO_FORMAT_DEFAULT);
    // retrieve Rx Source and Tx Sink device descriptors
    sp<DeviceDescriptor> rxSourceDevice =
        mAvailableInputDevices.getDevice(AUDIO_DEVICE_IN_TELEPHONY_RX,
                                         String8(),
                                         AUDIO_FORMAT_DEFAULT);
    sp<DeviceDescriptor> txSinkDevice =
        mAvailableOutputDevices.getDevice(AUDIO_DEVICE_OUT_TELEPHONY_TX,
                                          String8(),
                                          AUDIO_FORMAT_DEFAULT);

    // RX and TX Telephony device are declared by Primary Audio HAL
    if (isPrimaryModule(telephonyRxModule) && isPrimaryModule(telephonyTxModule) &&
            (telephonyRxModule->getHalVersionMajor() >= 3)) {
        if (rxSourceDevice == 0 || txSinkDevice == 0) {
            // RX / TX Telephony device(s) is(are) not currently available
            ALOGE("updateCallRouting() no telephony Tx and/or RX device");
            return muteWaitMs;
        }
        // do not create a patch (aka Sw Bridging) if Primary HW module has declared supporting a
        // route between telephony RX to Sink device and Source device to telephony TX
        const auto &primaryModule = telephonyRxModule;
        createRxPatch = !primaryModule->supportsPatch(rxSourceDevice, rxDevices.itemAt(0));
        createTxPatch = !primaryModule->supportsPatch(txSourceDevice, txSinkDevice);
    } else {
        // If the RX device is on the primary HW module, then use legacy routing method for
        // voice calls via setOutputDevice() on primary output.
        // Otherwise, create two audio patches for TX and RX path.
        createRxPatch = !(availablePrimaryOutputDevices().contains(rxDevices.itemAt(0))) &&
                (rxSourceDevice != 0);
        // If the TX device is also on the primary HW module, setOutputDevice() will take care
        // of it due to legacy implementation. If not, create a patch.
        createTxPatch = !(availablePrimaryModuleInputDevices().contains(txSourceDevice)) &&
                (txSinkDevice != 0);
    }
    // Use legacy routing method for voice calls via setOutputDevice() on primary output.
    // Otherwise, create two audio patches for TX and RX path.
    if (!createRxPatch) {
        muteWaitMs = setOutputDevices(mPrimaryOutput, rxDevices, true, delayMs);
    } else { // create RX path audio patch
        mCallRxPatch = createTelephonyPatch(true /*isRx*/, rxDevices.itemAt(0), delayMs);

        // If the TX device is on the primary HW module but RX device is
        // on other HW module, SinkMetaData of telephony input should handle it
        // assuming the device uses audio HAL V5.0 and above
    }
    if (createTxPatch) { // create TX path audio patch
        mCallTxPatch = createTelephonyPatch(false /*isRx*/, txSourceDevice, delayMs);
    }

    return muteWaitMs;
}

sp<AudioPatch> AudioPolicyManager::createTelephonyPatch(
        bool isRx, const sp<DeviceDescriptor> &device, uint32_t delayMs) {
    PatchBuilder patchBuilder;

    if (device == nullptr) {
        return nullptr;
    }
    if (isRx) {
        patchBuilder.addSink(device).
                addSource(mAvailableInputDevices.getDevice(
                    AUDIO_DEVICE_IN_TELEPHONY_RX, String8(), AUDIO_FORMAT_DEFAULT));
    } else {
        patchBuilder.addSource(device).
                addSink(mAvailableOutputDevices.getDevice(
                    AUDIO_DEVICE_OUT_TELEPHONY_TX, String8(), AUDIO_FORMAT_DEFAULT));
    }

    // @TODO: still ignoring the address, or not dealing platform with mutliple telephonydevices
    const sp<DeviceDescriptor> outputDevice = isRx ?
                device : mAvailableOutputDevices.getDevice(
                    AUDIO_DEVICE_OUT_TELEPHONY_TX, String8(), AUDIO_FORMAT_DEFAULT);
    SortedVector<audio_io_handle_t> outputs =
            getOutputsForDevices(DeviceVector(outputDevice), mOutputs);
    const audio_io_handle_t output = selectOutput(outputs);
    // request to reuse existing output stream if one is already opened to reach the target device
    if (output != AUDIO_IO_HANDLE_NONE) {
        sp<AudioOutputDescriptor> outputDesc = mOutputs.valueFor(output);
        ALOG_ASSERT(!outputDesc->isDuplicated(), "%s() %s device output %d is duplicated", __func__,
                    outputDevice->toString().c_str(), output);
        patchBuilder.addSource(outputDesc, { .stream = AUDIO_STREAM_PATCH });
    }

    if (!isRx) {
        // terminate active capture if on the same HW module as the call TX source device
        // FIXME: would be better to refine to only inputs whose profile connects to the
        // call TX device but this information is not in the audio patch and logic here must be
        // symmetric to the one in startInput()
        for (const auto& activeDesc : mInputs.getActiveInputs()) {
            if (activeDesc->hasSameHwModuleAs(device)) {
                closeActiveClients(activeDesc);
            }
        }
    }

    audio_patch_handle_t afPatchHandle = AUDIO_PATCH_HANDLE_NONE;
    status_t status = mpClientInterface->createAudioPatch(
            patchBuilder.patch(), &afPatchHandle, delayMs);
    ALOGW_IF(status != NO_ERROR,
            "%s() error %d creating %s audio patch", __func__, status, isRx ? "RX" : "TX");
    sp<AudioPatch> audioPatch;
    if (status == NO_ERROR) {
        audioPatch = new AudioPatch(patchBuilder.patch(), mUidCached);
        audioPatch->mAfPatchHandle = afPatchHandle;
        audioPatch->mUid = mUidCached;
    }
    return audioPatch;
}

sp<DeviceDescriptor> AudioPolicyManager::findDevice(
        const DeviceVector& devices, audio_devices_t device) const {
    DeviceVector deviceList = devices.getDevicesFromTypeMask(device);
    ALOG_ASSERT(!deviceList.isEmpty(),
            "%s() selected device type %#x is not in devices list", __func__, device);
    return deviceList.itemAt(0);
}

audio_devices_t AudioPolicyManager::getModuleDeviceTypes(
        const DeviceVector& devices, const char *moduleId) const {
    sp<HwModule> mod = mHwModules.getModuleFromName(moduleId);
    return mod != 0 ? devices.getDeviceTypesFromHwModule(mod->getHandle()) : AUDIO_DEVICE_NONE;
}

bool AudioPolicyManager::isDeviceOfModule(
        const sp<DeviceDescriptor>& devDesc, const char *moduleId) const {
    sp<HwModule> module = mHwModules.getModuleFromName(moduleId);
    if (module != 0) {
        return mAvailableOutputDevices.getDevicesFromHwModule(module->getHandle())
                .indexOf(devDesc) != NAME_NOT_FOUND
                || mAvailableInputDevices.getDevicesFromHwModule(module->getHandle())
                .indexOf(devDesc) != NAME_NOT_FOUND;
    }
    return false;
}

void AudioPolicyManager::setPhoneState(audio_mode_t state)
{
    ALOGV("setPhoneState() state %d", state);
    // store previous phone state for management of sonification strategy below
    int oldState = mEngine->getPhoneState();

    if (mEngine->setPhoneState(state) != NO_ERROR) {
        ALOGW("setPhoneState() invalid or same state %d", state);
        return;
    }
    /// Opens: can these line be executed after the switch of volume curves???
    if (isStateInCall(oldState)) {
        ALOGV("setPhoneState() in call state management: new state is %d", state);
        // force reevaluating accessibility routing when call stops
        mpClientInterface->invalidateStream(AUDIO_STREAM_ACCESSIBILITY);
    }

    /**
     * Switching to or from incall state or switching between telephony and VoIP lead to force
     * routing command.
     */
    bool force = ((is_state_in_call(oldState) != is_state_in_call(state))
                  || (is_state_in_call(state) && (state != oldState)));

    // check for device and output changes triggered by new phone state
    checkForDeviceAndOutputChanges();

    int delayMs = 0;
    if (isStateInCall(state)) {
        nsecs_t sysTime = systemTime();
        auto musicStrategy = streamToStrategy(AUDIO_STREAM_MUSIC);
        auto sonificationStrategy = streamToStrategy(AUDIO_STREAM_ALARM);
        for (size_t i = 0; i < mOutputs.size(); i++) {
            sp<SwAudioOutputDescriptor> desc = mOutputs.valueAt(i);
            // mute media and sonification strategies and delay device switch by the largest
            // latency of any output where either strategy is active.
            // This avoid sending the ring tone or music tail into the earpiece or headset.
            if ((desc->isStrategyActive(musicStrategy, SONIFICATION_HEADSET_MUSIC_DELAY, sysTime) ||
                 desc->isStrategyActive(sonificationStrategy, SONIFICATION_HEADSET_MUSIC_DELAY,
                                        sysTime)) &&
                    (delayMs < (int)desc->latency()*2)) {
                delayMs = desc->latency()*2;
            }
            setStrategyMute(musicStrategy, true, desc);
            setStrategyMute(musicStrategy, false, desc, MUTE_TIME_MS,
                mEngine->getOutputDevicesForAttributes(attributes_initializer(AUDIO_USAGE_MEDIA),
                                                       nullptr, true /*fromCache*/).types());
            setStrategyMute(sonificationStrategy, true, desc);
            setStrategyMute(sonificationStrategy, false, desc, MUTE_TIME_MS,
                mEngine->getOutputDevicesForAttributes(attributes_initializer(AUDIO_USAGE_ALARM),
                                                       nullptr, true /*fromCache*/).types());
        }
    }

    if (hasPrimaryOutput()) {
        // Note that despite the fact that getNewOutputDevices() is called on the primary output,
        // the device returned is not necessarily reachable via this output
        DeviceVector rxDevices = getNewOutputDevices(mPrimaryOutput, false /*fromCache*/);
        // force routing command to audio hardware when ending call
        // even if no device change is needed
        if (isStateInCall(oldState) && rxDevices.isEmpty()) {
            rxDevices = mPrimaryOutput->devices();
        }

        if (state == AUDIO_MODE_IN_CALL) {
            updateCallRouting(rxDevices, delayMs);
        } else if (oldState == AUDIO_MODE_IN_CALL) {
            if (mCallRxPatch != 0) {
                mpClientInterface->releaseAudioPatch(mCallRxPatch->mAfPatchHandle, 0);
                mCallRxPatch.clear();
            }
            if (mCallTxPatch != 0) {
                mpClientInterface->releaseAudioPatch(mCallTxPatch->mAfPatchHandle, 0);
                mCallTxPatch.clear();
            }
            setOutputDevices(mPrimaryOutput, rxDevices, force, 0);
        } else {
            setOutputDevices(mPrimaryOutput, rxDevices, force, 0);
        }
    }

    // reevaluate routing on all outputs in case tracks have been started during the call
    for (size_t i = 0; i < mOutputs.size(); i++) {
        sp<SwAudioOutputDescriptor> desc = mOutputs.valueAt(i);
        DeviceVector newDevices = getNewOutputDevices(desc, true /*fromCache*/);
        if (state != AUDIO_MODE_IN_CALL || desc != mPrimaryOutput) {
            setOutputDevices(desc, newDevices, !newDevices.isEmpty(), 0 /*delayMs*/);
        }
    }

    if (isStateInCall(state)) {
        ALOGV("setPhoneState() in call state management: new state is %d", state);
        // force reevaluating accessibility routing when call starts
        mpClientInterface->invalidateStream(AUDIO_STREAM_ACCESSIBILITY);
    }

    // Flag that ringtone volume must be limited to music volume until we exit MODE_RINGTONE
    mLimitRingtoneVolume = (state == AUDIO_MODE_RINGTONE &&
                            isStreamActive(AUDIO_STREAM_MUSIC, SONIFICATION_HEADSET_MUSIC_DELAY));
}

audio_mode_t AudioPolicyManager::getPhoneState() {
    return mEngine->getPhoneState();
}

void AudioPolicyManager::setForceUse(audio_policy_force_use_t usage,
                                     audio_policy_forced_cfg_t config)
{
    ALOGV("setForceUse() usage %d, config %d, mPhoneState %d", usage, config, mEngine->getPhoneState());
    if (config == mEngine->getForceUse(usage)) {
        return;
    }

    if (mEngine->setForceUse(usage, config) != NO_ERROR) {
        ALOGW("setForceUse() could not set force cfg %d for usage %d", config, usage);
        return;
    }
    bool forceVolumeReeval = (usage == AUDIO_POLICY_FORCE_FOR_COMMUNICATION) ||
            (usage == AUDIO_POLICY_FORCE_FOR_DOCK) ||
            (usage == AUDIO_POLICY_FORCE_FOR_SYSTEM);

    // check for device and output changes triggered by new force usage
    checkForDeviceAndOutputChanges();

    //FIXME: workaround for truncated touch sounds
    // to be removed when the problem is handled by system UI
    uint32_t delayMs = 0;
    uint32_t waitMs = 0;
    if (usage == AUDIO_POLICY_FORCE_FOR_COMMUNICATION) {
        delayMs = TOUCH_SOUND_FIXED_DELAY_MS;
    }
    if (mEngine->getPhoneState() == AUDIO_MODE_IN_CALL && hasPrimaryOutput()) {
        DeviceVector newDevices = getNewOutputDevices(mPrimaryOutput, true /*fromCache*/);
        waitMs = updateCallRouting(newDevices, delayMs);
    }
    for (size_t i = 0; i < mOutputs.size(); i++) {
        sp<SwAudioOutputDescriptor> outputDesc = mOutputs.valueAt(i);
        DeviceVector newDevices = getNewOutputDevices(outputDesc, true /*fromCache*/);
        if ((mEngine->getPhoneState() != AUDIO_MODE_IN_CALL) || (outputDesc != mPrimaryOutput)) {
            // As done in setDeviceConnectionState, we could also fix default device issue by
            // preventing the force re-routing in case of default dev that distinguishes on address.
            // Let's give back to engine full device choice decision however.
            waitMs = setOutputDevices(outputDesc, newDevices, !newDevices.isEmpty(), delayMs);
        }
        if (forceVolumeReeval && !newDevices.isEmpty()) {
            applyStreamVolumes(outputDesc, newDevices.types(), waitMs, true);
        }
    }

    for (const auto& activeDesc : mInputs.getActiveInputs()) {
        auto newDevice = getNewInputDevice(activeDesc);
        // Force new input selection if the new device can not be reached via current input
        if (activeDesc->mProfile->getSupportedDevices().contains(newDevice)) {
            setInputDevice(activeDesc->mIoHandle, newDevice);
        } else {
            closeInput(activeDesc->mIoHandle);
        }
    }
}

void AudioPolicyManager::setSystemProperty(const char* property, const char* value)
{
    ALOGV("setSystemProperty() property %s, value %s", property, value);
}

// Find an output profile compatible with the parameters passed. When "directOnly" is set, restrict
// search to profiles for direct outputs.
sp<IOProfile> AudioPolicyManager::getProfileForOutput(
                                                   const DeviceVector& devices,
                                                   uint32_t samplingRate,
                                                   audio_format_t format,
                                                   audio_channel_mask_t channelMask,
                                                   audio_output_flags_t flags,
                                                   bool directOnly)
{
    if (directOnly) {
        // only retain flags that will drive the direct output profile selection
        // if explicitly requested
        static const uint32_t kRelevantFlags =
                (AUDIO_OUTPUT_FLAG_HW_AV_SYNC | AUDIO_OUTPUT_FLAG_COMPRESS_OFFLOAD |
                 AUDIO_OUTPUT_FLAG_VOIP_RX);
        flags =
            (audio_output_flags_t)((flags & kRelevantFlags) | AUDIO_OUTPUT_FLAG_DIRECT);
    }

    sp<IOProfile> profile;

    for (const auto& hwModule : mHwModules) {
        for (const auto& curProfile : hwModule->getOutputProfiles()) {
            if (!curProfile->isCompatibleProfile(devices,
                    samplingRate, NULL /*updatedSamplingRate*/,
                    format, NULL /*updatedFormat*/,
                    channelMask, NULL /*updatedChannelMask*/,
                    flags)) {
                continue;
            }
            // reject profiles not corresponding to a device currently available
            if (!mAvailableOutputDevices.containsAtLeastOne(curProfile->getSupportedDevices())) {
                continue;
            }
            // reject profiles if connected device does not support codec
            if (!curProfile->deviceSupportsEncodedFormats(devices.types())) {
                continue;
            }
            if (!directOnly) return curProfile;
            // when searching for direct outputs, if several profiles are compatible, give priority
            // to one with offload capability
            if (profile != 0 && ((curProfile->getFlags() & AUDIO_OUTPUT_FLAG_COMPRESS_OFFLOAD) == 0)) {
                continue;
            }
            profile = curProfile;
            if ((profile->getFlags() & AUDIO_OUTPUT_FLAG_COMPRESS_OFFLOAD) != 0) {
                break;
            }
        }
    }
    return profile;
}

audio_io_handle_t AudioPolicyManager::getOutput(audio_stream_type_t stream)
{
    DeviceVector devices = mEngine->getOutputDevicesForStream(stream, false /*fromCache*/);

    // Note that related method getOutputForAttr() uses getOutputForDevice() not selectOutput().
    // We use selectOutput() here since we don't have the desired AudioTrack sample rate,
    // format, flags, etc. This may result in some discrepancy for functions that utilize
    // getOutput() solely on audio_stream_type such as AudioSystem::getOutputFrameCount()
    // and AudioSystem::getOutputSamplingRate().

    SortedVector<audio_io_handle_t> outputs = getOutputsForDevices(devices, mOutputs);
    const audio_io_handle_t output = selectOutput(outputs);

    ALOGV("getOutput() stream %d selected devices %s, output %d", stream,
          devices.toString().c_str(), output);
    return output;
}

status_t AudioPolicyManager::getAudioAttributes(audio_attributes_t *dstAttr,
                                                const audio_attributes_t *srcAttr,
                                                audio_stream_type_t srcStream)
{
    if (srcAttr != NULL) {
        if (!isValidAttributes(srcAttr)) {
            ALOGE("%s invalid attributes: usage=%d content=%d flags=0x%x tags=[%s]",
                    __func__,
                    srcAttr->usage, srcAttr->content_type, srcAttr->flags,
                    srcAttr->tags);
            return BAD_VALUE;
        }
        *dstAttr = *srcAttr;
    } else {
        if (srcStream < AUDIO_STREAM_MIN || srcStream >= AUDIO_STREAM_PUBLIC_CNT) {
            ALOGE("%s:  invalid stream type", __func__);
            return BAD_VALUE;
        }
        *dstAttr = mEngine->getAttributesForStreamType(srcStream);
    }
    return NO_ERROR;
}

status_t AudioPolicyManager::getOutputForAttrInt(
        audio_attributes_t *resultAttr,
        audio_io_handle_t *output,
        audio_session_t session,
        const audio_attributes_t *attr,
        audio_stream_type_t *stream,
        uid_t uid,
        const audio_config_t *config,
        audio_output_flags_t *flags,
        audio_port_handle_t *selectedDeviceId,
        bool *isRequestedDeviceForExclusiveUse,
        std::vector<sp<SwAudioOutputDescriptor>> *secondaryDescs)
{
    DeviceVector outputDevices;
    const audio_port_handle_t requestedPortId = *selectedDeviceId;
    DeviceVector msdDevices = getMsdAudioOutDevices();
    const sp<DeviceDescriptor> requestedDevice =
        mAvailableOutputDevices.getDeviceFromId(requestedPortId);

    status_t status = getAudioAttributes(resultAttr, attr, *stream);
    if (status != NO_ERROR) {
        return status;
    }
    if (auto it = mAllowedCapturePolicies.find(uid); it != end(mAllowedCapturePolicies)) {
        resultAttr->flags |= it->second;
    }
    *stream = mEngine->getStreamTypeForAttributes(*resultAttr);

    ALOGV("%s() attributes=%s stream=%s session %d selectedDeviceId %d", __func__,
          toString(*resultAttr).c_str(), toString(*stream).c_str(), session, requestedPortId);

    // The primary output is the explicit routing (eg. setPreferredDevice) if specified,
    //       otherwise, fallback to the dynamic policies, if none match, query the engine.
    // Secondary outputs are always found by dynamic policies as the engine do not support them
    sp<SwAudioOutputDescriptor> policyDesc;
    status = mPolicyMixes.getOutputForAttr(*resultAttr, uid, *flags, policyDesc, secondaryDescs);
    if (status != OK) {
        return status;
    }

    // Explicit routing is higher priority then any dynamic policy primary output
    bool usePrimaryOutputFromPolicyMixes = requestedDevice == nullptr && policyDesc != nullptr;

    // FIXME: in case of RENDER policy, the output capabilities should be checked
    if ((usePrimaryOutputFromPolicyMixes || !secondaryDescs->empty())
        && !audio_is_linear_pcm(config->format)) {
        ALOGD("%s: rejecting request as dynamic audio policy only support pcm", __func__);
        return BAD_VALUE;
    }
    if (usePrimaryOutputFromPolicyMixes) {
        *output = policyDesc->mIoHandle;
        sp<AudioPolicyMix> mix = policyDesc->mPolicyMix.promote();
        sp<DeviceDescriptor> deviceDesc =
                mAvailableOutputDevices.getDevice(mix->mDeviceType,
                                                  mix->mDeviceAddress,
                                                  AUDIO_FORMAT_DEFAULT);
        *selectedDeviceId = deviceDesc != 0 ? deviceDesc->getId() : AUDIO_PORT_HANDLE_NONE;
        ALOGV("getOutputForAttr() returns output %d", *output);
        return NO_ERROR;
    }
    // Virtual sources must always be dynamicaly or explicitly routed
    if (resultAttr->usage == AUDIO_USAGE_VIRTUAL_SOURCE) {
        ALOGW("getOutputForAttr() no policy mix found for usage AUDIO_USAGE_VIRTUAL_SOURCE");
        return BAD_VALUE;
    }
    // explicit routing managed by getDeviceForStrategy in APM is now handled by engine
    // in order to let the choice of the order to future vendor engine
    outputDevices = mEngine->getOutputDevicesForAttributes(*resultAttr, requestedDevice, false);

    if ((resultAttr->flags & AUDIO_FLAG_HW_AV_SYNC) != 0) {
        *flags = (audio_output_flags_t)(*flags | AUDIO_OUTPUT_FLAG_HW_AV_SYNC);
    }

    // Set incall music only if device was explicitly set, and fallback to the device which is
    // chosen by the engine if not.
    // FIXME: provide a more generic approach which is not device specific and move this back
    // to getOutputForDevice.
    // TODO: Remove check of AUDIO_STREAM_MUSIC once migration is completed on the app side.
    if (outputDevices.types() == AUDIO_DEVICE_OUT_TELEPHONY_TX &&
        (*stream == AUDIO_STREAM_MUSIC  || resultAttr->usage == AUDIO_USAGE_VOICE_COMMUNICATION) &&
        audio_is_linear_pcm(config->format) &&
        isInCall()) {
        if (requestedPortId != AUDIO_PORT_HANDLE_NONE) {
            *flags = (audio_output_flags_t)AUDIO_OUTPUT_FLAG_INCALL_MUSIC;
            *isRequestedDeviceForExclusiveUse = true;
        }
    }

    ALOGV("%s() device %s, sampling rate %d, format %#x, channel mask %#x, flags %#x stream %s",
          __func__, outputDevices.toString().c_str(), config->sample_rate, config->format,
          config->channel_mask, *flags, toString(*stream).c_str());

    *output = AUDIO_IO_HANDLE_NONE;
    if (!msdDevices.isEmpty()) {
        *output = getOutputForDevices(msdDevices, session, *stream, config, flags);
        sp<DeviceDescriptor> device = outputDevices.isEmpty() ? nullptr : outputDevices.itemAt(0);
        if (*output != AUDIO_IO_HANDLE_NONE && setMsdPatch(device) == NO_ERROR) {
            ALOGV("%s() Using MSD devices %s instead of devices %s",
                  __func__, msdDevices.toString().c_str(), outputDevices.toString().c_str());
            outputDevices = msdDevices;
        } else {
            *output = AUDIO_IO_HANDLE_NONE;
        }
    }
    if (*output == AUDIO_IO_HANDLE_NONE) {
        *output = getOutputForDevices(outputDevices, session, *stream, config,
                flags, resultAttr->flags & AUDIO_FLAG_MUTE_HAPTIC);
    }
    if (*output == AUDIO_IO_HANDLE_NONE) {
        return INVALID_OPERATION;
    }

    *selectedDeviceId = getFirstDeviceId(outputDevices);

    ALOGV("%s returns output %d selectedDeviceId %d", __func__, *output, *selectedDeviceId);

    return NO_ERROR;
}

status_t AudioPolicyManager::getOutputForAttr(const audio_attributes_t *attr,
                                              audio_io_handle_t *output,
                                              audio_session_t session,
                                              audio_stream_type_t *stream,
                                              uid_t uid,
                                              const audio_config_t *config,
                                              audio_output_flags_t *flags,
                                              audio_port_handle_t *selectedDeviceId,
                                              audio_port_handle_t *portId,
                                              std::vector<audio_io_handle_t> *secondaryOutputs)
{
    // The supplied portId must be AUDIO_PORT_HANDLE_NONE
    if (*portId != AUDIO_PORT_HANDLE_NONE) {
        return INVALID_OPERATION;
    }
    const audio_port_handle_t requestedPortId = *selectedDeviceId;
    audio_attributes_t resultAttr;
    bool isRequestedDeviceForExclusiveUse = false;
    std::vector<sp<SwAudioOutputDescriptor>> secondaryOutputDescs;
    const sp<DeviceDescriptor> requestedDevice =
      mAvailableOutputDevices.getDeviceFromId(requestedPortId);

    // Prevent from storing invalid requested device id in clients
    const audio_port_handle_t sanitizedRequestedPortId =
      requestedDevice != nullptr ? requestedPortId : AUDIO_PORT_HANDLE_NONE;
    *selectedDeviceId = sanitizedRequestedPortId;

    status_t status = getOutputForAttrInt(&resultAttr, output, session, attr, stream, uid,
            config, flags, selectedDeviceId, &isRequestedDeviceForExclusiveUse,
            &secondaryOutputDescs);
    if (status != NO_ERROR) {
        return status;
    }
    std::vector<wp<SwAudioOutputDescriptor>> weakSecondaryOutputDescs;
    for (auto& secondaryDesc : secondaryOutputDescs) {
        secondaryOutputs->push_back(secondaryDesc->mIoHandle);
        weakSecondaryOutputDescs.push_back(secondaryDesc);
    }

    audio_config_base_t clientConfig = {.sample_rate = config->sample_rate,
        .format = config->format,
        .channel_mask = config->channel_mask };
    *portId = AudioPort::getNextUniqueId();

    sp<TrackClientDescriptor> clientDesc =
        new TrackClientDescriptor(*portId, uid, session, resultAttr, clientConfig,
                                  sanitizedRequestedPortId, *stream,
                                  mEngine->getProductStrategyForAttributes(resultAttr),
                                  toVolumeSource(resultAttr),
                                  *flags, isRequestedDeviceForExclusiveUse,
                                  std::move(weakSecondaryOutputDescs));
    sp<SwAudioOutputDescriptor> outputDesc = mOutputs.valueFor(*output);
    outputDesc->addClient(clientDesc);

    ALOGV("%s() returns output %d requestedPortId %d selectedDeviceId %d for port ID %d", __func__,
          *output, requestedPortId, *selectedDeviceId, *portId);

    return NO_ERROR;
}

audio_io_handle_t AudioPolicyManager::getOutputForDevices(
        const DeviceVector &devices,
        audio_session_t session,
        audio_stream_type_t stream,
        const audio_config_t *config,
        audio_output_flags_t *flags,
        bool forceMutingHaptic)
{
    audio_io_handle_t output = AUDIO_IO_HANDLE_NONE;
    status_t status;

    // Discard haptic channel mask when forcing muting haptic channels.
    audio_channel_mask_t channelMask = forceMutingHaptic
            ? (config->channel_mask & ~AUDIO_CHANNEL_HAPTIC_ALL) : config->channel_mask;

    // open a direct output if required by specified parameters
    //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 ((*flags & AUDIO_OUTPUT_FLAG_COMPRESS_OFFLOAD) != 0) {
        *flags = (audio_output_flags_t)(*flags | AUDIO_OUTPUT_FLAG_DIRECT);
    }
    if ((*flags & AUDIO_OUTPUT_FLAG_HW_AV_SYNC) != 0) {
        *flags = (audio_output_flags_t)(*flags | AUDIO_OUTPUT_FLAG_DIRECT);
    }
    // only allow deep buffering for music stream type
    if (stream != AUDIO_STREAM_MUSIC) {
        *flags = (audio_output_flags_t)(*flags &~AUDIO_OUTPUT_FLAG_DEEP_BUFFER);
    } else if (/* stream == AUDIO_STREAM_MUSIC && */
            *flags == AUDIO_OUTPUT_FLAG_NONE &&
            property_get_bool("audio.deep_buffer.media", false /* default_value */)) {
        // use DEEP_BUFFER as default output for music stream type
        *flags = (audio_output_flags_t)AUDIO_OUTPUT_FLAG_DEEP_BUFFER;
    }
    if (stream == AUDIO_STREAM_TTS) {
        *flags = AUDIO_OUTPUT_FLAG_TTS;
    } else if (stream == AUDIO_STREAM_VOICE_CALL &&
               audio_is_linear_pcm(config->format) &&
               (*flags & AUDIO_OUTPUT_FLAG_INCALL_MUSIC) == 0) {
        *flags = (audio_output_flags_t)(AUDIO_OUTPUT_FLAG_VOIP_RX |
                                       AUDIO_OUTPUT_FLAG_DIRECT);
        ALOGV("Set VoIP and Direct output flags for PCM format");
    }


    sp<IOProfile> profile;

    // skip direct output selection if the request can obviously be attached to a mixed output
    // and not explicitly requested
    if (((*flags & AUDIO_OUTPUT_FLAG_DIRECT) == 0) &&
            audio_is_linear_pcm(config->format) && config->sample_rate <= SAMPLE_RATE_HZ_MAX &&
            audio_channel_count_from_out_mask(channelMask) <= 2) {
        goto non_direct_output;
    }

    // Do not allow offloading if one non offloadable effect is enabled or MasterMono is enabled.
    // This prevents creating an offloaded track and tearing it down immediately after start
    // when audioflinger detects there is an active non offloadable effect.
    // FIXME: We should check the audio session here but we do not have it in this context.
    // This may prevent offloading in rare situations where effects are left active by apps
    // in the background.

    if (((*flags & AUDIO_OUTPUT_FLAG_COMPRESS_OFFLOAD) == 0) ||
            !(mEffects.isNonOffloadableEffectEnabled() || mMasterMono)) {
        profile = getProfileForOutput(devices,
                                   config->sample_rate,
                                   config->format,
                                   channelMask,
                                   (audio_output_flags_t)*flags,
                                   true /* directOnly */);
    }

    if (profile != 0) {
        // exclusive outputs for MMAP and Offload are enforced by different session ids.
        for (size_t i = 0; i < mOutputs.size(); i++) {
            sp<SwAudioOutputDescriptor> desc = mOutputs.valueAt(i);
            if (!desc->isDuplicated() && (profile == desc->mProfile)) {
                // reuse direct output if currently open by the same client
                // and configured with same parameters
                if ((config->sample_rate == desc->mSamplingRate) &&
                    (config->format == desc->mFormat) &&
                    (channelMask == desc->mChannelMask) &&
                    (session == desc->mDirectClientSession)) {
                    desc->mDirectOpenCount++;
                    ALOGI("%s reusing direct output %d for session %d", __func__, 
                        mOutputs.keyAt(i), session);
                    return mOutputs.keyAt(i);
                }
            }
        }

        if (!profile->canOpenNewIo()) {
            goto non_direct_output;
        }

        sp<SwAudioOutputDescriptor> outputDesc =
                new SwAudioOutputDescriptor(profile, mpClientInterface);

        String8 address = getFirstDeviceAddress(devices);

        // MSD patch may be using the only output stream that can service this request. Release
        // MSD patch to prioritize this request over any active output on MSD.
        AudioPatchCollection msdPatches = getMsdPatches();
        for (size_t i = 0; i < msdPatches.size(); i++) {
            const auto& patch = msdPatches[i];
            for (size_t j = 0; j < patch->mPatch.num_sinks; ++j) {
                const struct audio_port_config *sink = &patch->mPatch.sinks[j];
                if (sink->type == AUDIO_PORT_TYPE_DEVICE &&
                        (sink->ext.device.type & devices.types()) != AUDIO_DEVICE_NONE &&
                        (address.isEmpty() || strncmp(sink->ext.device.address, address.string(),
                                AUDIO_DEVICE_MAX_ADDRESS_LEN) == 0)) {
                    releaseAudioPatch(patch->mHandle, mUidCached);
                    break;
                }
            }
        }

        status = outputDesc->open(config, devices, stream, *flags, &output);

        // only accept an output with the requested parameters
        if (status != NO_ERROR ||
            (config->sample_rate != 0 && config->sample_rate != outputDesc->mSamplingRate) ||
            (config->format != AUDIO_FORMAT_DEFAULT && config->format != outputDesc->mFormat) ||
            (channelMask != 0 && channelMask != outputDesc->mChannelMask)) {
            ALOGV("%s failed opening direct output: output %d sample rate %d %d," 
                    "format %d %d, channel mask %04x %04x", __func__, output, config->sample_rate,
                    outputDesc->mSamplingRate, config->format, outputDesc->mFormat,
                    channelMask, outputDesc->mChannelMask);
            if (output != AUDIO_IO_HANDLE_NONE) {
                outputDesc->close();
            }
            // fall back to mixer output if possible when the direct output could not be open
            if (audio_is_linear_pcm(config->format) &&
                    config->sample_rate  <= SAMPLE_RATE_HZ_MAX) {
                goto non_direct_output;
            }
            return AUDIO_IO_HANDLE_NONE;
        }
        outputDesc->mDirectOpenCount = 1;
        outputDesc->mDirectClientSession = session;

        addOutput(output, outputDesc);
        mPreviousOutputs = mOutputs;
        ALOGV("%s returns new direct output %d", __func__, output);
        mpClientInterface->onAudioPortListUpdate();
        return output;
    }

non_direct_output:

    // A request for HW A/V sync cannot fallback to a mixed output because time
    // stamps are embedded in audio data
    if ((*flags & (AUDIO_OUTPUT_FLAG_HW_AV_SYNC | AUDIO_OUTPUT_FLAG_MMAP_NOIRQ)) != 0) {
        return AUDIO_IO_HANDLE_NONE;
    }

    // ignoring channel mask due to downmix capability in mixer

    // open a non direct output

    // for non direct outputs, only PCM is supported
    if (audio_is_linear_pcm(config->format)) {
        // get which output is suitable for the specified stream. The actual
        // routing change will happen when startOutput() will be called
        SortedVector<audio_io_handle_t> outputs = getOutputsForDevices(devices, mOutputs);

        // at this stage we should ignore the DIRECT flag as no direct output could be found earlier
        *flags = (audio_output_flags_t)(*flags & ~AUDIO_OUTPUT_FLAG_DIRECT);
        output = selectOutput(outputs, *flags, config->format, channelMask, config->sample_rate);
    }
    ALOGW_IF((output == 0), "getOutputForDevices() could not find output for stream %d, "
            "sampling rate %d, format %#x, channels %#x, flags %#x",
            stream, config->sample_rate, config->format, channelMask, *flags);

    return output;
}

sp<DeviceDescriptor> AudioPolicyManager::getMsdAudioInDevice() const {
    auto msdInDevices = mHwModules.getAvailableDevicesFromModuleName(AUDIO_HARDWARE_MODULE_ID_MSD,
                                                                     mAvailableInputDevices);
    return msdInDevices.isEmpty()? nullptr : msdInDevices.itemAt(0);
}

DeviceVector AudioPolicyManager::getMsdAudioOutDevices() const {
    return mHwModules.getAvailableDevicesFromModuleName(AUDIO_HARDWARE_MODULE_ID_MSD,
                                                        mAvailableOutputDevices);
}

const AudioPatchCollection AudioPolicyManager::getMsdPatches() const {
    AudioPatchCollection msdPatches;
    sp<HwModule> msdModule = mHwModules.getModuleFromName(AUDIO_HARDWARE_MODULE_ID_MSD);
    if (msdModule != 0) {
        for (size_t i = 0; i < mAudioPatches.size(); ++i) {
            sp<AudioPatch> patch = mAudioPatches.valueAt(i);
            for (size_t j = 0; j < patch->mPatch.num_sources; ++j) {
                const struct audio_port_config *source = &patch->mPatch.sources[j];
                if (source->type == AUDIO_PORT_TYPE_DEVICE &&
                        source->ext.device.hw_module == msdModule->getHandle()) {
                    msdPatches.addAudioPatch(patch->mHandle, patch);
                }
            }
        }
    }
    return msdPatches;
}

status_t AudioPolicyManager::getBestMsdAudioProfileFor(const sp<DeviceDescriptor> &outputDevice,
        bool hwAvSync, audio_port_config *sourceConfig, audio_port_config *sinkConfig) const
{
    sp<HwModule> msdModule = mHwModules.getModuleFromName(AUDIO_HARDWARE_MODULE_ID_MSD);
    if (msdModule == nullptr) {
        ALOGE("%s() unable to get MSD module", __func__);
        return NO_INIT;
    }
    sp<HwModule> deviceModule = mHwModules.getModuleForDevice(outputDevice, AUDIO_FORMAT_DEFAULT);
    if (deviceModule == nullptr) {
        ALOGE("%s() unable to get module for %s", __func__, outputDevice->toString().c_str());
        return NO_INIT;
    }
    const InputProfileCollection &inputProfiles = msdModule->getInputProfiles();
    if (inputProfiles.isEmpty()) {
        ALOGE("%s() no input profiles for MSD module", __func__);
        return NO_INIT;
    }
    const OutputProfileCollection &outputProfiles = deviceModule->getOutputProfiles();
    if (outputProfiles.isEmpty()) {
        ALOGE("%s() no output profiles for device %s", __func__, outputDevice->toString().c_str());
        return NO_INIT;
    }
    AudioProfileVector msdProfiles;
    // Each IOProfile represents a MixPort from audio_policy_configuration.xml
    for (const auto &inProfile : inputProfiles) {
        if (hwAvSync == ((inProfile->getFlags() & AUDIO_INPUT_FLAG_HW_AV_SYNC) != 0)) {
            msdProfiles.appendVector(inProfile->getAudioProfiles());
        }
    }
    AudioProfileVector deviceProfiles;
    for (const auto &outProfile : outputProfiles) {
        if (hwAvSync == ((outProfile->getFlags() & AUDIO_OUTPUT_FLAG_HW_AV_SYNC) != 0)) {
            deviceProfiles.appendVector(outProfile->getAudioProfiles());
        }
    }
    struct audio_config_base bestSinkConfig;
    status_t result = msdProfiles.findBestMatchingOutputConfig(deviceProfiles,
            compressedFormatsOrder, surroundChannelMasksOrder, true /*preferHigherSamplingRates*/,
            &bestSinkConfig);
    if (result != NO_ERROR) {
        ALOGD("%s() no matching profiles found for device: %s, hwAvSync: %d",
                __func__, outputDevice->toString().c_str(), hwAvSync);
        return result;
    }
    sinkConfig->sample_rate = bestSinkConfig.sample_rate;
    sinkConfig->channel_mask = bestSinkConfig.channel_mask;
    sinkConfig->format = bestSinkConfig.format;
    // For encoded streams force direct flag to prevent downstream mixing.
    sinkConfig->flags.output = static_cast<audio_output_flags_t>(
            sinkConfig->flags.output | AUDIO_OUTPUT_FLAG_DIRECT);
    sourceConfig->sample_rate = bestSinkConfig.sample_rate;
    // Specify exact channel mask to prevent guessing by bit count in PatchPanel.
    sourceConfig->channel_mask = audio_channel_mask_out_to_in(bestSinkConfig.channel_mask);
    sourceConfig->format = bestSinkConfig.format;
    // Copy input stream directly without any processing (e.g. resampling).
    sourceConfig->flags.input = static_cast<audio_input_flags_t>(
            sourceConfig->flags.input | AUDIO_INPUT_FLAG_DIRECT);
    if (hwAvSync) {
        sinkConfig->flags.output = static_cast<audio_output_flags_t>(
                sinkConfig->flags.output | AUDIO_OUTPUT_FLAG_HW_AV_SYNC);
        sourceConfig->flags.input = static_cast<audio_input_flags_t>(
                sourceConfig->flags.input | AUDIO_INPUT_FLAG_HW_AV_SYNC);
    }
    const unsigned int config_mask = AUDIO_PORT_CONFIG_SAMPLE_RATE |
            AUDIO_PORT_CONFIG_CHANNEL_MASK | AUDIO_PORT_CONFIG_FORMAT | AUDIO_PORT_CONFIG_FLAGS;
    sinkConfig->config_mask |= config_mask;
    sourceConfig->config_mask |= config_mask;
    return NO_ERROR;
}

PatchBuilder AudioPolicyManager::buildMsdPatch(const sp<DeviceDescriptor> &outputDevice) const
{
    PatchBuilder patchBuilder;
    patchBuilder.addSource(getMsdAudioInDevice()).addSink(outputDevice);
    audio_port_config sourceConfig = patchBuilder.patch()->sources[0];
    audio_port_config sinkConfig = patchBuilder.patch()->sinks[0];
    // TODO: Figure out whether MSD module has HW_AV_SYNC flag set in the AP config file.
    // For now, we just forcefully try with HwAvSync first.
    status_t res = getBestMsdAudioProfileFor(outputDevice, true /*hwAvSync*/,
            &sourceConfig, &sinkConfig) == NO_ERROR ? NO_ERROR :
            getBestMsdAudioProfileFor(
                    outputDevice, false /*hwAvSync*/, &sourceConfig, &sinkConfig);
    if (res == NO_ERROR) {
        // Found a matching profile for encoded audio. Re-create PatchBuilder with this config.
        return (PatchBuilder()).addSource(sourceConfig).addSink(sinkConfig);
    }
    ALOGV("%s() no matching profile found. Fall through to default PCM patch"
            " supporting PCM format conversion.", __func__);
    return patchBuilder;
}

status_t AudioPolicyManager::setMsdPatch(const sp<DeviceDescriptor> &outputDevice) {
    sp<DeviceDescriptor> device = outputDevice;
    if (device == nullptr) {
        // Use media strategy for unspecified output device. This should only
        // occur on checkForDeviceAndOutputChanges(). Device connection events may
        // therefore invalidate explicit routing requests.
        DeviceVector devices = mEngine->getOutputDevicesForAttributes(
                    attributes_initializer(AUDIO_USAGE_MEDIA), nullptr, false /*fromCache*/);
        LOG_ALWAYS_FATAL_IF(devices.isEmpty(), "no outpudevice to set Msd Patch");
        device = devices.itemAt(0);
    }
    ALOGV("%s() for device %s", __func__, device->toString().c_str());
    PatchBuilder patchBuilder = buildMsdPatch(device);
    const struct audio_patch* patch = patchBuilder.patch();
    const AudioPatchCollection msdPatches = getMsdPatches();
    if (!msdPatches.isEmpty()) {
        LOG_ALWAYS_FATAL_IF(msdPatches.size() > 1,
                "The current MSD prototype only supports one output patch");
        sp<AudioPatch> currentPatch = msdPatches.valueAt(0);
        if (audio_patches_are_equal(&currentPatch->mPatch, patch)) {
            return NO_ERROR;
        }
        releaseAudioPatch(currentPatch->mHandle, mUidCached);
    }
    status_t status = installPatch(__func__, -1 /*index*/, nullptr /*patchHandle*/,
            patch, 0 /*delayMs*/, mUidCached, nullptr /*patchDescPtr*/);
    ALOGE_IF(status != NO_ERROR, "%s() error %d creating MSD audio patch", __func__, status);
    ALOGI_IF(status == NO_ERROR, "%s() Patch created from MSD_IN to "
           "device:%s (format:%#x channels:%#x samplerate:%d)", __func__,
             device->toString().c_str(), patch->sources[0].format,
             patch->sources[0].channel_mask, patch->sources[0].sample_rate);
    return status;
}

audio_io_handle_t AudioPolicyManager::selectOutput(const SortedVector<audio_io_handle_t>& outputs,
                                                       audio_output_flags_t flags,
                                                       audio_format_t format,
                                                       audio_channel_mask_t channelMask,
                                                       uint32_t samplingRate)
{
    LOG_ALWAYS_FATAL_IF(!(format == AUDIO_FORMAT_INVALID || audio_is_linear_pcm(format)),
        "%s called with format %#x", __func__, format);

    // Flags disqualifying an output: the match must happen before calling selectOutput()
    static const audio_output_flags_t kExcludedFlags = (audio_output_flags_t)
        (AUDIO_OUTPUT_FLAG_HW_AV_SYNC | AUDIO_OUTPUT_FLAG_MMAP_NOIRQ | AUDIO_OUTPUT_FLAG_DIRECT);

    // Flags expressing a functional request: must be honored in priority over
    // other criteria
    static const audio_output_flags_t kFunctionalFlags = (audio_output_flags_t)
        (AUDIO_OUTPUT_FLAG_VOIP_RX | AUDIO_OUTPUT_FLAG_INCALL_MUSIC |
            AUDIO_OUTPUT_FLAG_TTS | AUDIO_OUTPUT_FLAG_DIRECT_PCM);
    // Flags expressing a performance request: have lower priority than serving
    // requested sampling rate or channel mask
    static const audio_output_flags_t kPerformanceFlags = (audio_output_flags_t)
        (AUDIO_OUTPUT_FLAG_FAST | AUDIO_OUTPUT_FLAG_DEEP_BUFFER |
            AUDIO_OUTPUT_FLAG_RAW | AUDIO_OUTPUT_FLAG_SYNC);

    const audio_output_flags_t functionalFlags =
        (audio_output_flags_t)(flags & kFunctionalFlags);
    const audio_output_flags_t performanceFlags =
        (audio_output_flags_t)(flags & kPerformanceFlags);

    audio_io_handle_t bestOutput = (outputs.size() == 0) ? AUDIO_IO_HANDLE_NONE : outputs[0];

    // select one output among several that provide a path to a particular device or set of
    // devices (the list was previously build by getOutputsForDevices()).
    // The priority is as follows:
    // 1: the output supporting haptic playback when requesting haptic playback
    // 2: the output with the highest number of requested functional flags
    // 3: the output supporting the exact channel mask
    // 4: the output with a higher channel count than requested
    // 5: the output with a higher sampling rate than requested
    // 6: the output with the highest number of requested performance flags
    // 7: the output with the bit depth the closest to the requested one
    // 8: the primary output
    // 9: the first output in the list

    // matching criteria values in priority order for best matching output so far
    std::vector<uint32_t> bestMatchCriteria(8, 0);

    const uint32_t channelCount = audio_channel_count_from_out_mask(channelMask);
    const uint32_t hapticChannelCount = audio_channel_count_from_out_mask(
        channelMask & AUDIO_CHANNEL_HAPTIC_ALL);

    for (audio_io_handle_t output : outputs) {
        sp<SwAudioOutputDescriptor> outputDesc = mOutputs.valueFor(output);
        // matching criteria values in priority order for current output
        std::vector<uint32_t> currentMatchCriteria(8, 0);

        if (outputDesc->isDuplicated()) {
            continue;
        }
        if ((kExcludedFlags & outputDesc->mFlags) != 0) {
            continue;
        }

        // If haptic channel is specified, use the haptic output if present.
        // When using haptic output, same audio format and sample rate are required.
        const uint32_t outputHapticChannelCount = audio_channel_count_from_out_mask(
            outputDesc->mChannelMask & AUDIO_CHANNEL_HAPTIC_ALL);
        if ((hapticChannelCount == 0) != (outputHapticChannelCount == 0)) {
            continue;
        }
        if (outputHapticChannelCount >= hapticChannelCount
            && format == outputDesc->mFormat
            && samplingRate == outputDesc->mSamplingRate) {
                currentMatchCriteria[0] = outputHapticChannelCount;
        }

        // functional flags match
        currentMatchCriteria[1] = popcount(outputDesc->mFlags & functionalFlags);

        // channel mask and channel count match
        uint32_t outputChannelCount = audio_channel_count_from_out_mask(outputDesc->mChannelMask);
        if (channelMask != AUDIO_CHANNEL_NONE && channelCount > 2 &&
            channelCount <= outputChannelCount) {
            if ((audio_channel_mask_get_representation(channelMask) ==
                    audio_channel_mask_get_representation(outputDesc->mChannelMask)) &&
                    ((channelMask & outputDesc->mChannelMask) == channelMask)) {
                currentMatchCriteria[2] = outputChannelCount;
            }
            currentMatchCriteria[3] = outputChannelCount;
        }

        // sampling rate match
        if (samplingRate > SAMPLE_RATE_HZ_DEFAULT &&
                samplingRate <= outputDesc->mSamplingRate) {
            currentMatchCriteria[4] = outputDesc->mSamplingRate;
        }

        // performance flags match
        currentMatchCriteria[5] = popcount(outputDesc->mFlags & performanceFlags);

        // format match
        if (format != AUDIO_FORMAT_INVALID) {
            currentMatchCriteria[6] =
                AudioPort::kFormatDistanceMax -
                AudioPort::formatDistance(format, outputDesc->mFormat);
        }

        // primary output match
        currentMatchCriteria[7] = outputDesc->mFlags & AUDIO_OUTPUT_FLAG_PRIMARY;

        // compare match criteria by priority then value
        if (std::lexicographical_compare(bestMatchCriteria.begin(), bestMatchCriteria.end(),
                currentMatchCriteria.begin(), currentMatchCriteria.end())) {
            bestMatchCriteria = currentMatchCriteria;
            bestOutput = output;

            std::stringstream result;
            std::copy(bestMatchCriteria.begin(), bestMatchCriteria.end(),
                std::ostream_iterator<int>(result, " "));
            ALOGV("%s new bestOutput %d criteria %s",
                __func__, bestOutput, result.str().c_str());
        }
    }

    return bestOutput;
}

status_t AudioPolicyManager::startOutput(audio_port_handle_t portId)
{
    ALOGV("%s portId %d", __FUNCTION__, portId);

    sp<SwAudioOutputDescriptor> outputDesc = mOutputs.getOutputForClient(portId);
    if (outputDesc == 0) {
        ALOGW("startOutput() no output for client %d", portId);
        return BAD_VALUE;
    }
    sp<TrackClientDescriptor> client = outputDesc->getClient(portId);

    ALOGV("startOutput() output %d, stream %d, session %d",
          outputDesc->mIoHandle, client->stream(), client->session());

    status_t status = outputDesc->start();
    if (status != NO_ERROR) {
        return status;
    }

    uint32_t delayMs;
    status = startSource(outputDesc, client, &delayMs);

    if (status != NO_ERROR) {
        outputDesc->stop();
        return status;
    }
    if (delayMs != 0) {
        usleep(delayMs * 1000);
    }

    return status;
}

status_t AudioPolicyManager::startSource(const sp<SwAudioOutputDescriptor>& outputDesc,
                                         const sp<TrackClientDescriptor>& client,
                                         uint32_t *delayMs)
{
    // cannot start playback of STREAM_TTS if any other output is being used
    uint32_t beaconMuteLatency = 0;

    *delayMs = 0;
    audio_stream_type_t stream = client->stream();
    auto clientVolSrc = client->volumeSource();
    auto clientStrategy = client->strategy();
    auto clientAttr = client->attributes();
    if (stream == AUDIO_STREAM_TTS) {
        ALOGV("\t found BEACON stream");
        if (!mTtsOutputAvailable && mOutputs.isAnyOutputActive(
                                    toVolumeSource(AUDIO_STREAM_TTS) /*sourceToIgnore*/)) {
            return INVALID_OPERATION;
        } else {
            beaconMuteLatency = handleEventForBeacon(STARTING_BEACON);
        }
    } else {
        // some playback other than beacon starts
        beaconMuteLatency = handleEventForBeacon(STARTING_OUTPUT);
    }

    // force device change if the output is inactive and no audio patch is already present.
    // check active before incrementing usage count
    bool force = !outputDesc->isActive() &&
            (outputDesc->getPatchHandle() == AUDIO_PATCH_HANDLE_NONE);

    DeviceVector devices;
    sp<AudioPolicyMix> policyMix = outputDesc->mPolicyMix.promote();
    const char *address = NULL;
    if (policyMix != NULL) {
        audio_devices_t newDeviceType;
        address = policyMix->mDeviceAddress.string();
        if ((policyMix->mRouteFlags & MIX_ROUTE_FLAG_LOOP_BACK) == MIX_ROUTE_FLAG_LOOP_BACK) {
            newDeviceType = AUDIO_DEVICE_OUT_REMOTE_SUBMIX;
        } else {
            newDeviceType = policyMix->mDeviceType;
        }
        sp device = mAvailableOutputDevices.getDevice(newDeviceType, String8(address),
                                                        AUDIO_FORMAT_DEFAULT);
        ALOG_ASSERT(device, "%s: no device found t=%u, a=%s", __func__, newDeviceType, address);
        devices.add(device);
    }

    // requiresMuteCheck is false when we can bypass mute strategy.
    // It covers a common case when there is no materially active audio
    // and muting would result in unnecessary delay and dropped audio.
    const uint32_t outputLatencyMs = outputDesc->latency();
    bool requiresMuteCheck = outputDesc->isActive(outputLatencyMs * 2);  // account for drain

    // increment usage count for this stream on the requested output:
    // NOTE that the usage count is the same for duplicated output and hardware output which is
    // necessary for a correct control of hardware output routing by startOutput() and stopOutput()
    outputDesc->setClientActive(client, true);

    if (client->hasPreferredDevice(true)) {
        if (outputDesc->clientsList(true /*activeOnly*/).size() == 1 &&
                client->isPreferredDeviceForExclusiveUse()) {
            // Preferred device may be exclusive, use only if no other active clients on this output
            devices = DeviceVector(
                        mAvailableOutputDevices.getDeviceFromId(client->preferredDeviceId()));
        } else {
            devices = getNewOutputDevices(outputDesc, false /*fromCache*/);
        }
        if (devices != outputDesc->devices()) {
            checkStrategyRoute(clientStrategy, outputDesc->mIoHandle);
        }
    }

    if (followsSameRouting(clientAttr, attributes_initializer(AUDIO_USAGE_MEDIA))) {
        selectOutputForMusicEffects();
    }

    if (outputDesc->getActivityCount(clientVolSrc) == 1 || !devices.isEmpty()) {
        // starting an output being rerouted?
        if (devices.isEmpty()) {
            devices = getNewOutputDevices(outputDesc, false /*fromCache*/);
        }
        bool shouldWait =
            (followsSameRouting(clientAttr, attributes_initializer(AUDIO_USAGE_ALARM)) ||
             followsSameRouting(clientAttr, attributes_initializer(AUDIO_USAGE_NOTIFICATION)) ||
             (beaconMuteLatency > 0));
        uint32_t waitMs = beaconMuteLatency;
        for (size_t i = 0; i < mOutputs.size(); i++) {
            sp<SwAudioOutputDescriptor> desc = mOutputs.valueAt(i);
            if (desc != outputDesc) {
                // An output has a shared device if
                // - managed by the same hw module
                // - supports the currently selected device
                const bool sharedDevice = outputDesc->sharesHwModuleWith(desc)
                        && (!desc->filterSupportedDevices(devices).isEmpty());

                // force a device change if any other output is:
                // - managed by the same hw module
                // - supports currently selected device
                // - has a current device selection that differs from selected device.
                // - has an active audio patch
                // In this case, the audio HAL must receive the new device selection so that it can
                // change the device currently selected by the other output.
                if (sharedDevice &&
                        desc->devices() != devices &&
                        desc->getPatchHandle() != AUDIO_PATCH_HANDLE_NONE) {
                    force = true;
                }
                // wait for audio on other active outputs to be presented when starting
                // a notification so that audio focus effect can propagate, or that a mute/unmute
                // event occurred for beacon
                const uint32_t latencyMs = desc->latency();
                const bool isActive = desc->isActive(latencyMs * 2);  // account for drain

                if (shouldWait && isActive && (waitMs < latencyMs)) {
                    waitMs = latencyMs;
                }

                // Require mute check if another output is on a shared device
                // and currently active to have proper drain and avoid pops.
                // Note restoring AudioTracks onto this output needs to invoke
                // a volume ramp if there is no mute.
                requiresMuteCheck |= sharedDevice && isActive;
            }
        }

        const uint32_t muteWaitMs =
                setOutputDevices(outputDesc, devices, force, 0, NULL, requiresMuteCheck);

        // apply volume rules for current stream and device if necessary
        auto &curves = getVolumeCurves(client->attributes());
        checkAndSetVolume(curves, client->volumeSource(),
                          curves.getVolumeIndex(outputDesc->devices().types()),
                          outputDesc,
                          outputDesc->devices().types());

        // update the outputs if starting an output with a stream that can affect notification
        // routing
        handleNotificationRoutingForStream(stream);

        // force reevaluating accessibility routing when ringtone or alarm starts
        if (followsSameRouting(clientAttr, attributes_initializer(AUDIO_USAGE_ALARM))) {
            mpClientInterface->invalidateStream(AUDIO_STREAM_ACCESSIBILITY);
        }

        if (waitMs > muteWaitMs) {
            *delayMs = waitMs - muteWaitMs;
        }

        // FIXME: A device change (muteWaitMs > 0) likely introduces a volume change.
        // A volume change enacted by APM with 0 delay is not synchronous, as it goes
        // via AudioCommandThread to AudioFlinger.  Hence it is possible that the volume
        // change occurs after the MixerThread starts and causes a stream volume
        // glitch.
        //
        // We do not introduce additional delay here.
    }

    if (stream == AUDIO_STREAM_ENFORCED_AUDIBLE &&
            mEngine->getForceUse(AUDIO_POLICY_FORCE_FOR_SYSTEM) == AUDIO_POLICY_FORCE_SYSTEM_ENFORCED) {
        setStrategyMute(streamToStrategy(AUDIO_STREAM_ALARM), true, outputDesc);
    }

    // Automatically enable the remote submix input when output is started on a re routing mix
    // of type MIX_TYPE_RECORDERS
    if (audio_is_remote_submix_device(devices.types()) && policyMix != NULL &&
        policyMix->mMixType == MIX_TYPE_RECORDERS) {
        setDeviceConnectionStateInt(AUDIO_DEVICE_IN_REMOTE_SUBMIX,
                                    AUDIO_POLICY_DEVICE_STATE_AVAILABLE,
                                    address,
                                    "remote-submix",
                                    AUDIO_FORMAT_DEFAULT);
    }

    return NO_ERROR;
}

status_t AudioPolicyManager::stopOutput(audio_port_handle_t portId)
{
    ALOGV("%s portId %d", __FUNCTION__, portId);

    sp<SwAudioOutputDescriptor> outputDesc = mOutputs.getOutputForClient(portId);
    if (outputDesc == 0) {
        ALOGW("stopOutput() no output for client %d", portId);
        return BAD_VALUE;
    }
    sp<TrackClientDescriptor> client = outputDesc->getClient(portId);

    ALOGV("stopOutput() output %d, stream %d, session %d",
          outputDesc->mIoHandle, client->stream(), client->session());

    status_t status = stopSource(outputDesc, client);

    if (status == NO_ERROR ) {
        outputDesc->stop();
    }
    return status;
}

status_t AudioPolicyManager::stopSource(const sp<SwAudioOutputDescriptor>& outputDesc,
                                        const sp<TrackClientDescriptor>& client)
{
    // always handle stream stop, check which stream type is stopping
    audio_stream_type_t stream = client->stream();
    auto clientVolSrc = client->volumeSource();

    handleEventForBeacon(stream == AUDIO_STREAM_TTS ? STOPPING_BEACON : STOPPING_OUTPUT);

    if (outputDesc->getActivityCount(clientVolSrc) > 0) {
        if (outputDesc->getActivityCount(clientVolSrc) == 1) {
            // Automatically disable the remote submix input when output is stopped on a
            // re routing mix of type MIX_TYPE_RECORDERS
            sp<AudioPolicyMix> policyMix = outputDesc->mPolicyMix.promote();
            if (audio_is_remote_submix_device(outputDesc->devices().types()) &&
                policyMix != NULL &&
                policyMix->mMixType == MIX_TYPE_RECORDERS) {
                setDeviceConnectionStateInt(AUDIO_DEVICE_IN_REMOTE_SUBMIX,
                                            AUDIO_POLICY_DEVICE_STATE_UNAVAILABLE,
                                            policyMix->mDeviceAddress,
                                            "remote-submix", AUDIO_FORMAT_DEFAULT);
            }
        }
        bool forceDeviceUpdate = false;
        if (client->hasPreferredDevice(true)) {
            checkStrategyRoute(client->strategy(), AUDIO_IO_HANDLE_NONE);
            forceDeviceUpdate = true;
        }

        // decrement usage count of this stream on the output
        outputDesc->setClientActive(client, false);

        // store time at which the stream was stopped - see isStreamActive()
        if (outputDesc->getActivityCount(clientVolSrc) == 0 || forceDeviceUpdate) {
            outputDesc->setStopTime(client, systemTime());
            DeviceVector newDevices = getNewOutputDevices(outputDesc, false /*fromCache*/);
            // delay the device switch by twice the latency because stopOutput() is executed when
            // the track stop() command is received and at that time the audio track buffer can
            // still contain data that needs to be drained. The latency only covers the audio HAL
            // and kernel buffers. Also the latency does not always include additional delay in the
            // audio path (audio DSP, CODEC ...)
            setOutputDevices(outputDesc, newDevices, false, outputDesc->latency()*2);

            // force restoring the device selection on other active outputs if it differs from the
            // one being selected for this output
            uint32_t delayMs = outputDesc->latency()*2;
            for (size_t i = 0; i < mOutputs.size(); i++) {
                sp<SwAudioOutputDescriptor> desc = mOutputs.valueAt(i);
                if (desc != outputDesc &&
                        desc->isActive() &&
                        outputDesc->sharesHwModuleWith(desc) &&
                        (newDevices != desc->devices())) {
                    DeviceVector newDevices2 = getNewOutputDevices(desc, false /*fromCache*/);
                    bool force = desc->devices() != newDevices2;

                    setOutputDevices(desc, newDevices2, force, delayMs);

                    // re-apply device specific volume if not done by setOutputDevice()
                    if (!force) {
                        applyStreamVolumes(desc, newDevices2.types(), delayMs);
                    }
                }
            }
            // update the outputs if stopping one with a stream that can affect notification routing
            handleNotificationRoutingForStream(stream);
        }

        if (stream == AUDIO_STREAM_ENFORCED_AUDIBLE &&
                mEngine->getForceUse(AUDIO_POLICY_FORCE_FOR_SYSTEM) == AUDIO_POLICY_FORCE_SYSTEM_ENFORCED) {
            setStrategyMute(streamToStrategy(AUDIO_STREAM_ALARM), false, outputDesc);
        }

        if (followsSameRouting(client->attributes(), attributes_initializer(AUDIO_USAGE_MEDIA))) {
            selectOutputForMusicEffects();
        }
        return NO_ERROR;
    } else {
        ALOGW("stopOutput() refcount is already 0");
        return INVALID_OPERATION;
    }
}

void AudioPolicyManager::releaseOutput(audio_port_handle_t portId)
{
    ALOGV("%s portId %d", __FUNCTION__, portId);

    sp<SwAudioOutputDescriptor> outputDesc = mOutputs.getOutputForClient(portId);
    if (outputDesc == 0) {
        // If an output descriptor is closed due to a device routing change,
        // then there are race conditions with releaseOutput from tracks
        // that may be destroyed (with no PlaybackThread) or a PlaybackThread
        // destroyed shortly thereafter.
        //
        // Here we just log a warning, instead of a fatal error.
        ALOGW("releaseOutput() no output for client %d", portId);
        return;
    }

    ALOGV("releaseOutput() %d", outputDesc->mIoHandle);

    if (outputDesc->mFlags & AUDIO_OUTPUT_FLAG_DIRECT) {
        if (outputDesc->mDirectOpenCount <= 0) {
            ALOGW("releaseOutput() invalid open count %d for output %d",
                  outputDesc->mDirectOpenCount, outputDesc->mIoHandle);
            return;
        }
        if (--outputDesc->mDirectOpenCount == 0) {
            closeOutput(outputDesc->mIoHandle);
            mpClientInterface->onAudioPortListUpdate();
        }
    }
    // stopOutput() needs to be successfully called before releaseOutput()
    // otherwise there may be inaccurate stream reference counts.
    // This is checked in outputDesc->removeClient below.
    outputDesc->removeClient(portId);
}

status_t AudioPolicyManager::getInputForAttr(const audio_attributes_t *attr,
                                             audio_io_handle_t *input,
                                             audio_unique_id_t riid,
                                             audio_session_t session,
                                             uid_t uid,
                                             const audio_config_base_t *config,
                                             audio_input_flags_t flags,
                                             audio_port_handle_t *selectedDeviceId,
                                             input_type_t *inputType,
                                             audio_port_handle_t *portId)
{
    ALOGV("%s() source %d, sampling rate %d, format %#x, channel mask %#x, session %d, "
          "flags %#x attributes=%s", __func__, attr->source, config->sample_rate,
          config->format, config->channel_mask, session, flags, toString(*attr).c_str());

    status_t status = NO_ERROR;
    audio_source_t halInputSource;
    audio_attributes_t attributes = *attr;
    sp<AudioPolicyMix> policyMix;
    sp<DeviceDescriptor> device;
    sp<AudioInputDescriptor> inputDesc;
    sp<RecordClientDescriptor> clientDesc;
    audio_port_handle_t requestedDeviceId = *selectedDeviceId;
    bool isSoundTrigger;

    // The supplied portId must be AUDIO_PORT_HANDLE_NONE
    if (*portId != AUDIO_PORT_HANDLE_NONE) {
        return INVALID_OPERATION;
    }

    if (attr->source == AUDIO_SOURCE_DEFAULT) {
        attributes.source = AUDIO_SOURCE_MIC;
    }

    // Explicit routing?
    sp<DeviceDescriptor> explicitRoutingDevice = 
            mAvailableInputDevices.getDeviceFromId(*selectedDeviceId);

    // special case for mmap capture: if an input IO handle is specified, we reuse this input if
    // possible
    if ((flags & AUDIO_INPUT_FLAG_MMAP_NOIRQ) == AUDIO_INPUT_FLAG_MMAP_NOIRQ &&
            *input != AUDIO_IO_HANDLE_NONE) {
        ssize_t index = mInputs.indexOfKey(*input);
        if (index < 0) {
            ALOGW("getInputForAttr() unknown MMAP input %d", *input);
            status = BAD_VALUE;
            goto error;
        }
        sp<AudioInputDescriptor> inputDesc = mInputs.valueAt(index);
        RecordClientVector clients = inputDesc->getClientsForSession(session);
        if (clients.size() == 0) {
            ALOGW("getInputForAttr() unknown session %d on input %d", session, *input);
            status = BAD_VALUE;
            goto error;
        }
        // For MMAP mode, the first call to getInputForAttr() is made on behalf of audioflinger.
        // The second call is for the first active client and sets the UID. Any further call
        // corresponds to a new client and is only permitted from the same UID.
        // If the first UID is silenced, allow a new UID connection and replace with new UID
        if (clients.size() > 1) {
            for (const auto& client : clients) {
                // The client map is ordered by key values (portId) and portIds are allocated
                // incrementaly. So the first client in this list is the one opened by audio flinger
                // when the mmap stream is created and should be ignored as it does not correspond
                // to an actual client
                if (client == *clients.cbegin()) {
                    continue;
                }
                if (uid != client->uid() && !client->isSilenced()) {
                    ALOGW("getInputForAttr() bad uid %d for client %d uid %d",
                          uid, client->portId(), client->uid());
                    status = INVALID_OPERATION;
                    goto error;
                }
            }
        }
        *inputType = API_INPUT_LEGACY;
        device = inputDesc->getDevice();

        ALOGI("%s reusing MMAP input %d for session %d", __FUNCTION__, *input, session);
        goto exit;
    }

    *input = AUDIO_IO_HANDLE_NONE;
    *inputType = API_INPUT_INVALID;

    halInputSource = attributes.source;

    if (attributes.source == AUDIO_SOURCE_REMOTE_SUBMIX &&
            strncmp(attributes.tags, "addr=", strlen("addr=")) == 0) {
        status = mPolicyMixes.getInputMixForAttr(attributes, &policyMix);
        if (status != NO_ERROR) {
            ALOGW("%s could not find input mix for attr %s",
                    __func__, toString(attributes).c_str());
            goto error;
        }
        device = mAvailableInputDevices.getDevice(AUDIO_DEVICE_IN_REMOTE_SUBMIX,
                                                  String8(attr->tags + strlen("addr=")),
                                                  AUDIO_FORMAT_DEFAULT);
        if (device == nullptr) {
            ALOGW("%s could not find in Remote Submix device for source %d, tags %s",
                    __func__, attributes.source, attributes.tags);
            status = BAD_VALUE;
            goto error;
        }

        if (is_mix_loopback_render(policyMix->mRouteFlags)) {
            *inputType = API_INPUT_MIX_PUBLIC_CAPTURE_PLAYBACK;
        } else {
            *inputType = API_INPUT_MIX_EXT_POLICY_REROUTE;
        }
    } else {
        if (explicitRoutingDevice != nullptr) {
            device = explicitRoutingDevice;
        } else {
            // Prevent from storing invalid requested device id in clients
            requestedDeviceId = AUDIO_PORT_HANDLE_NONE;
            device = mEngine->getInputDeviceForAttributes(attributes, &policyMix);
        }
        if (device == nullptr) {
            ALOGW("getInputForAttr() could not find device for source %d", attributes.source);
            status = BAD_VALUE;
            goto error;
        }
        if (policyMix) {
            ALOG_ASSERT(policyMix->mMixType == MIX_TYPE_RECORDERS, "Invalid Mix Type");
            // there is an external policy, but this input is attached to a mix of recorders,
            // meaning it receives audio injected into the framework, so the recorder doesn't
            // know about it and is therefore considered "legacy"
            *inputType = API_INPUT_LEGACY;
        } else if (audio_is_remote_submix_device(device->type())) {
            *inputType = API_INPUT_MIX_CAPTURE;
        } else if (device->type() == AUDIO_DEVICE_IN_TELEPHONY_RX) {
            *inputType = API_INPUT_TELEPHONY_RX;
        } else {
            *inputType = API_INPUT_LEGACY;
        }

    }

    *input = getInputForDevice(device, session, attributes, config, flags, policyMix);
    if (*input == AUDIO_IO_HANDLE_NONE) {
        status = INVALID_OPERATION;
        goto error;
    }

exit:

    *selectedDeviceId = mAvailableInputDevices.contains(device) ?
                device->getId() : AUDIO_PORT_HANDLE_NONE;

    isSoundTrigger = attributes.source == AUDIO_SOURCE_HOTWORD &&
        mSoundTriggerSessions.indexOfKey(session) >= 0;
    *portId = AudioPort::getNextUniqueId();

    clientDesc = new RecordClientDescriptor(*portId, riid, uid, session, attributes, *config,
                                            requestedDeviceId, attributes.source, flags,
                                            isSoundTrigger);
    inputDesc = mInputs.valueFor(*input);
    inputDesc->addClient(clientDesc);

    ALOGV("getInputForAttr() returns input %d type %d selectedDeviceId %d for port ID %d",
            *input, *inputType, *selectedDeviceId, *portId);

    return NO_ERROR;

error:
    return status;
}


audio_io_handle_t AudioPolicyManager::getInputForDevice(const sp<DeviceDescriptor> &device,
                                                        audio_session_t session,
                                                        const audio_attributes_t &attributes,
                                                        const audio_config_base_t *config,
                                                        audio_input_flags_t flags,
                                                        const sp<AudioPolicyMix> &policyMix)
{
    audio_io_handle_t input = AUDIO_IO_HANDLE_NONE;
    audio_source_t halInputSource = attributes.source;
    bool isSoundTrigger = false;

    if (attributes.source == AUDIO_SOURCE_HOTWORD) {
        ssize_t index = mSoundTriggerSessions.indexOfKey(session);
        if (index >= 0) {
            input = mSoundTriggerSessions.valueFor(session);
            isSoundTrigger = true;
            flags = (audio_input_flags_t)(flags | AUDIO_INPUT_FLAG_HW_HOTWORD);
            ALOGV("SoundTrigger capture on session %d input %d", session, input);
        } else {
            halInputSource = AUDIO_SOURCE_VOICE_RECOGNITION;
        }
    } else if (attributes.source == AUDIO_SOURCE_VOICE_COMMUNICATION &&
               audio_is_linear_pcm(config->format)) {
        flags = (audio_input_flags_t)(flags | AUDIO_INPUT_FLAG_VOIP_TX);
    }

    // find a compatible input profile (not necessarily identical in parameters)
    sp<IOProfile> profile;
    // sampling rate and flags may be updated by getInputProfile
    uint32_t profileSamplingRate = (config->sample_rate == 0) ?
            SAMPLE_RATE_HZ_DEFAULT : config->sample_rate;
    audio_format_t profileFormat;
    audio_channel_mask_t profileChannelMask = config->channel_mask;
    audio_input_flags_t profileFlags = flags;
    for (;;) {
        profileFormat = config->format; // reset each time through loop, in case it is updated
        profile = getInputProfile(device, profileSamplingRate, profileFormat, profileChannelMask,
                                  profileFlags);
        if (profile != 0) {
            break; // success
        } else if (profileFlags & AUDIO_INPUT_FLAG_RAW) {
            profileFlags = (audio_input_flags_t) (profileFlags & ~AUDIO_INPUT_FLAG_RAW); // retry
        } else if (profileFlags != AUDIO_INPUT_FLAG_NONE) {
            profileFlags = AUDIO_INPUT_FLAG_NONE; // retry
        } else { // fail
            ALOGW("%s could not find profile for device %s, sampling rate %u, format %#x, "
                  "channel mask 0x%X, flags %#x", __func__, device->toString().c_str(), 
                  config->sample_rate, config->format, config->channel_mask, flags);
            return input;
        }
    }
    // Pick input sampling rate if not specified by client
    uint32_t samplingRate = config->sample_rate;
    if (samplingRate == 0) {
        samplingRate = profileSamplingRate;
    }

    if (profile->getModuleHandle() == 0) {
        ALOGE("getInputForAttr(): HW module %s not opened", profile->getModuleName());
        return input;
    }

    if (!profile->canOpenNewIo()) {
        for (size_t i = 0; i < mInputs.size(); ) {
            sp <AudioInputDescriptor> desc = mInputs.valueAt(i);
            if (desc->mProfile != profile) {
                i++;
                continue;
            }
            // if sound trigger, reuse input if used by other sound trigger on same session
            // else
            //    reuse input if active client app is not in IDLE state
            //
            RecordClientVector clients = desc->clientsList();
            bool doClose = false;
            for (const auto& client : clients) {
                if (isSoundTrigger != client->isSoundTrigger()) {
                    continue;
                }
                if (client->isSoundTrigger()) {
                    if (session == client->session()) {
                        return desc->mIoHandle;
                    }
                    continue;
                }
                if (client->active() && client->appState() != APP_STATE_IDLE) {
                    return desc->mIoHandle;
                }
                doClose = true;
            }
            if (doClose) {
                closeInput(desc->mIoHandle);
            } else {
                i++;
            }
        }
    }

    sp<AudioInputDescriptor> inputDesc = new AudioInputDescriptor(profile, mpClientInterface);

    audio_config_t lConfig = AUDIO_CONFIG_INITIALIZER;
    lConfig.sample_rate = profileSamplingRate;
    lConfig.channel_mask = profileChannelMask;
    lConfig.format = profileFormat;

    status_t status = inputDesc->open(&lConfig, device, halInputSource, profileFlags, &input);

    // only accept input with the exact requested set of parameters
    if (status != NO_ERROR || input == AUDIO_IO_HANDLE_NONE ||
        (profileSamplingRate != lConfig.sample_rate) ||
        !audio_formats_match(profileFormat, lConfig.format) ||
        (profileChannelMask != lConfig.channel_mask)) {
        ALOGW("getInputForAttr() failed opening input: sampling rate %d"
              ", format %#x, channel mask %#x",
              profileSamplingRate, profileFormat, profileChannelMask);
        if (input != AUDIO_IO_HANDLE_NONE) {
            inputDesc->close();
        }
        return AUDIO_IO_HANDLE_NONE;
    }

    inputDesc->mPolicyMix = policyMix;

    addInput(input, inputDesc);
    mpClientInterface->onAudioPortListUpdate();

    return input;
}

status_t AudioPolicyManager::startInput(audio_port_handle_t portId)
{
    ALOGV("%s portId %d", __FUNCTION__, portId);

    sp<AudioInputDescriptor> inputDesc = mInputs.getInputForClient(portId);
    if (inputDesc == 0) {
        ALOGW("%s no input for client %d", __FUNCTION__, portId);
        return BAD_VALUE;
    }
    audio_io_handle_t input = inputDesc->mIoHandle;
    sp<RecordClientDescriptor> client = inputDesc->getClient(portId);
    if (client->active()) {
        ALOGW("%s input %d client %d already started", __FUNCTION__, input, client->portId());
        return INVALID_OPERATION;
    }

    audio_session_t session = client->session();

    ALOGV("%s input:%d, session:%d)", __FUNCTION__, input, session);

    Vector<sp<AudioInputDescriptor>> activeInputs = mInputs.getActiveInputs();

    status_t status = inputDesc->start();
    if (status != NO_ERROR) {
        return status;
    }

  // increment activity count before calling getNewInputDevice() below as only active sessions
    // are considered for device selection
    inputDesc->setClientActive(client, true);

    // indicate active capture to sound trigger service if starting capture from a mic on
    // primary HW module
    sp<DeviceDescriptor> device = getNewInputDevice(inputDesc);
    setInputDevice(input, device, true /* force */);

    if (inputDesc->activeCount()  == 1) {
        sp<AudioPolicyMix> policyMix = inputDesc->mPolicyMix.promote();
        // if input maps to a dynamic policy with an activity listener, notify of state change
        if ((policyMix != NULL)
                && ((policyMix->mCbFlags & AudioMix::kCbFlagNotifyActivity) != 0)) {
            mpClientInterface->onDynamicPolicyMixStateUpdate(policyMix->mDeviceAddress,
                    MIX_STATE_MIXING);
        }

        DeviceVector primaryInputDevices = availablePrimaryModuleInputDevices();
        if (primaryInputDevices.contains(device) &&
                mInputs.activeInputsCountOnDevices(primaryInputDevices) == 1) {
            SoundTrigger::setCaptureState(true);
        }

        // automatically enable the remote submix output when input is started if not
        // used by a policy mix of type MIX_TYPE_RECORDERS
        // For remote submix (a virtual device), we open only one input per capture request.
        if (audio_is_remote_submix_device(inputDesc->getDeviceType())) {
            String8 address = String8("");
            if (policyMix == NULL) {
                address = String8("0");
            } else if (policyMix->mMixType == MIX_TYPE_PLAYERS) {
                address = policyMix->mDeviceAddress;
            }
            if (address != "") {
                setDeviceConnectionStateInt(AUDIO_DEVICE_OUT_REMOTE_SUBMIX,
                        AUDIO_POLICY_DEVICE_STATE_AVAILABLE,
                        address, "remote-submix", AUDIO_FORMAT_DEFAULT);
            }
        }
    }

    ALOGV("%s input %d source = %d exit", __FUNCTION__, input, client->source());

    return NO_ERROR;
}

status_t AudioPolicyManager::stopInput(audio_port_handle_t portId)
{
    ALOGV("%s portId %d", __FUNCTION__, portId);

    sp<AudioInputDescriptor> inputDesc = mInputs.getInputForClient(portId);
    if (inputDesc == 0) {
        ALOGW("%s no input for client %d", __FUNCTION__, portId);
        return BAD_VALUE;
    }
    audio_io_handle_t input = inputDesc->mIoHandle;
    sp<RecordClientDescriptor> client = inputDesc->getClient(portId);
    if (!client->active()) {
        ALOGW("%s input %d client %d already stopped", __FUNCTION__, input, client->portId());
        return INVALID_OPERATION;
    }

    inputDesc->setClientActive(client, false);

    inputDesc->stop();
    if (inputDesc->isActive()) {
        setInputDevice(input, getNewInputDevice(inputDesc), false /* force */);
    } else {
        sp<AudioPolicyMix> policyMix = inputDesc->mPolicyMix.promote();
        // if input maps to a dynamic policy with an activity listener, notify of state change
        if ((policyMix != NULL)
                && ((policyMix->mCbFlags & AudioMix::kCbFlagNotifyActivity) != 0)) {
            mpClientInterface->onDynamicPolicyMixStateUpdate(policyMix->mDeviceAddress,
                    MIX_STATE_IDLE);
        }

        // automatically disable the remote submix output when input is stopped if not
        // used by a policy mix of type MIX_TYPE_RECORDERS
        if (audio_is_remote_submix_device(inputDesc->getDeviceType())) {
            String8 address = String8("");
            if (policyMix == NULL) {
                address = String8("0");
            } else if (policyMix->mMixType == MIX_TYPE_PLAYERS) {
                address = policyMix->mDeviceAddress;
            }
            if (address != "") {
                setDeviceConnectionStateInt(AUDIO_DEVICE_OUT_REMOTE_SUBMIX,
                                         AUDIO_POLICY_DEVICE_STATE_UNAVAILABLE,
                                         address, "remote-submix", AUDIO_FORMAT_DEFAULT);
            }
        }
        resetInputDevice(input);

        // indicate inactive capture to sound trigger service if stopping capture from a mic on
        // primary HW module
        DeviceVector primaryInputDevices = availablePrimaryModuleInputDevices();
        if (primaryInputDevices.contains(inputDesc->getDevice()) &&
                mInputs.activeInputsCountOnDevices(primaryInputDevices) == 0) {
            SoundTrigger::setCaptureState(false);
        }
        inputDesc->clearPreemptedSessions();
    }
    return NO_ERROR;
}

void AudioPolicyManager::releaseInput(audio_port_handle_t portId)
{
    ALOGV("%s portId %d", __FUNCTION__, portId);

    sp<AudioInputDescriptor> inputDesc = mInputs.getInputForClient(portId);
    if (inputDesc == 0) {
        ALOGW("%s no input for client %d", __FUNCTION__, portId);
        return;
    }
    sp<RecordClientDescriptor> client = inputDesc->getClient(portId);
    audio_io_handle_t input = inputDesc->mIoHandle;

    ALOGV("%s %d", __FUNCTION__, input);

    inputDesc->removeClient(portId);

    if (inputDesc->getClientCount() > 0) {
        ALOGV("%s(%d) %zu clients remaining", __func__, portId, inputDesc->getClientCount());
        return;
    }

    closeInput(input);
    mpClientInterface->onAudioPortListUpdate();
    ALOGV("%s exit", __FUNCTION__);
}

void AudioPolicyManager::closeActiveClients(const sp<AudioInputDescriptor>& input)
{
    RecordClientVector clients = input->clientsList(true);

    for (const auto& client : clients) {
        closeClient(client->portId());
    }
}

void AudioPolicyManager::closeClient(audio_port_handle_t portId)
{
    stopInput(portId);
    releaseInput(portId);
}

void AudioPolicyManager::checkCloseInputs() {
    // After connecting or disconnecting an input device, close input if:
    // - it has no client (was just opened to check profile)  OR
    // - none of its supported devices are connected anymore OR
    // - one of its clients cannot be routed to one of its supported
    // devices anymore. Otherwise update device selection
    std::vector<audio_io_handle_t> inputsToClose;
    for (size_t i = 0; i < mInputs.size(); i++) {
        const sp<AudioInputDescriptor> input = mInputs.valueAt(i);
        if (input->clientsList().size() == 0
                || !mAvailableInputDevices.containsAtLeastOne(input->supportedDevices())) {
            inputsToClose.push_back(mInputs.keyAt(i));
        } else {
            bool close = false;
            for (const auto& client : input->clientsList()) {
                sp<DeviceDescriptor> device =
                    mEngine->getInputDeviceForAttributes(client->attributes());
                if (!input->supportedDevices().contains(device)) {
                    close = true;
                    break;
                }
            }
            if (close) {
                inputsToClose.push_back(mInputs.keyAt(i));
            } else {
                setInputDevice(input->mIoHandle, getNewInputDevice(input));
            }
        }
    }

    for (const audio_io_handle_t handle : inputsToClose) {
        ALOGV("%s closing input %d", __func__, handle);
        closeInput(handle);
    }
}

void AudioPolicyManager::initStreamVolume(audio_stream_type_t stream, int indexMin, int indexMax)
{
    ALOGV("initStreamVolume() stream %d, min %d, max %d", stream , indexMin, indexMax);
    if (indexMin < 0 || indexMax < 0) {
        ALOGE("%s for stream %d: invalid min %d or max %d", __func__, stream , indexMin, indexMax);
        return;
    }
    getVolumeCurves(stream).initVolume(indexMin, indexMax);

    // initialize other private stream volumes which follow this one
    for (int curStream = 0; curStream < AUDIO_STREAM_FOR_POLICY_CNT; curStream++) {
        if (!streamsMatchForvolume(stream, (audio_stream_type_t)curStream)) {
            continue;
        }
        getVolumeCurves((audio_stream_type_t)curStream).initVolume(indexMin, indexMax);
    }
}

status_t AudioPolicyManager::setStreamVolumeIndex(audio_stream_type_t stream,
                                                  int index,
                                                  audio_devices_t device)
{
    auto attributes = mEngine->getAttributesForStreamType(stream);
    ALOGV("%s: stream %s attributes=%s", __func__,
          toString(stream).c_str(), toString(attributes).c_str());
    return setVolumeIndexForAttributes(attributes, index, device);
}

status_t AudioPolicyManager::getStreamVolumeIndex(audio_stream_type_t stream,
                                                  int *index,
                                                  audio_devices_t device)
{
    // if device is AUDIO_DEVICE_OUT_DEFAULT_FOR_VOLUME, return volume for device selected for this
    // stream by the engine.
    if (device == AUDIO_DEVICE_OUT_DEFAULT_FOR_VOLUME) {
        device = mEngine->getOutputDevicesForStream(stream, true /*fromCache*/).types();
    }
    return getVolumeIndex(getVolumeCurves(stream), *index, device);
}

status_t AudioPolicyManager::setVolumeIndexForAttributes(const audio_attributes_t &attributes,
                                                         int index,
                                                         audio_devices_t device)
{
    // Get Volume group matching the Audio Attributes
    auto group = mEngine->getVolumeGroupForAttributes(attributes);
    if (group == VOLUME_GROUP_NONE) {
        ALOGD("%s: no group matching with %s", __FUNCTION__, toString(attributes).c_str());
        return BAD_VALUE;
    }
    ALOGV("%s: group %d matching with %s", __FUNCTION__, group, toString(attributes).c_str());
    status_t status = NO_ERROR;
    IVolumeCurves &curves = getVolumeCurves(attributes);
    VolumeSource vs = toVolumeSource(group);
    product_strategy_t strategy = mEngine->getProductStrategyForAttributes(attributes);

    status = setVolumeCurveIndex(index, device, curves);
    if (status != NO_ERROR) {
        ALOGE("%s failed to set curve index for group %d device 0x%X", __func__, group, device);
        return status;
    }

    audio_devices_t curSrcDevice;
    auto curCurvAttrs = curves.getAttributes();
    if (!curCurvAttrs.empty() && curCurvAttrs.front() != defaultAttr) {
        auto attr = curCurvAttrs.front();
        curSrcDevice = mEngine->getOutputDevicesForAttributes(attr, nullptr, false).types();
    } else if (!curves.getStreamTypes().empty()) {
        auto stream = curves.getStreamTypes().front();
        curSrcDevice = mEngine->getOutputDevicesForStream(stream, false).types();
    } else {
        ALOGE("%s: Invalid src %d: no valid attributes nor stream",__func__, vs);
        return BAD_VALUE;
    }
    curSrcDevice = Volume::getDeviceForVolume(curSrcDevice);

    // update volume on all outputs and streams matching the following:
    // - The requested stream (or a stream matching for volume control) is active on the output
    // - The device (or devices) selected by the engine for this stream includes
    // the requested device
    // - For non default requested device, currently selected device on the output is either the
    // requested device or one of the devices selected by the engine for this stream
    // - For default requested device (AUDIO_DEVICE_OUT_DEFAULT_FOR_VOLUME), apply volume only if
    // no specific device volume value exists for currently selected device.
    for (size_t i = 0; i < mOutputs.size(); i++) {
        sp<SwAudioOutputDescriptor> desc = mOutputs.valueAt(i);
        audio_devices_t curDevice = desc->devices().types();

        if (curDevice & AUDIO_DEVICE_OUT_SPEAKER_SAFE) {
            curDevice |= AUDIO_DEVICE_OUT_SPEAKER;
            curDevice &= ~AUDIO_DEVICE_OUT_SPEAKER_SAFE;
        }

        // Inter / intra volume group priority management: Loop on strategies arranged by priority
        // If a higher priority strategy is active, and the output is routed to a device with a
        // HW Gain management, do not change the volume
        bool applyVolume = false;
        if (desc->useHwGain()) {
            if (!(desc->isActive(toVolumeSource(group)) || isInCall())) {
                continue;
            }
            for (const auto &productStrategy : mEngine->getOrderedProductStrategies()) {
                auto activeClients = desc->clientsList(true /*activeOnly*/, productStrategy,
                                                       false /*preferredDevice*/);
                if (activeClients.empty()) {
                    continue;
                }
                bool isPreempted = false;
                bool isHigherPriority = productStrategy < strategy;
                for (const auto &client : activeClients) {
                    if (isHigherPriority && (client->volumeSource() != vs)) {
                        ALOGV("%s: Strategy=%d (\nrequester:\n"
                              " group %d, volumeGroup=%d attributes=%s)\n"
                              " higher priority source active:\n"
                              " volumeGroup=%d attributes=%s) \n"
                              " on output %zu, bailing out", __func__, productStrategy,
                              group, group, toString(attributes).c_str(),
                              client->volumeSource(), toString(client->attributes()).c_str(), i);
                        applyVolume = false;
                        isPreempted = true;
                        break;
                    }
                    // However, continue for loop to ensure no higher prio clients running on output
                    if (client->volumeSource() == vs) {
                        applyVolume = true;
                    }
                }
                if (isPreempted || applyVolume) {
                    break;
                }
            }
            if (!applyVolume) {
                continue; // next output
            }
            status_t volStatus = checkAndSetVolume(curves, vs, index, desc, curDevice,
                                                   (vs == toVolumeSource(AUDIO_STREAM_SYSTEM)?
                                                        TOUCH_SOUND_FIXED_DELAY_MS : 0));
            if (volStatus != NO_ERROR) {
                status = volStatus;
            }
            continue;
        }
        if (!(desc->isActive(vs) || isInCall())) {
            continue;
        }
        if ((device != AUDIO_DEVICE_OUT_DEFAULT_FOR_VOLUME) && ((curDevice & device) == 0)) {
            continue;
        }
        if (device != AUDIO_DEVICE_OUT_DEFAULT_FOR_VOLUME) {
            curSrcDevice |= device;
            applyVolume = (Volume::getDeviceForVolume(curDevice) & curSrcDevice) != 0;
        } else {
            applyVolume = !curves.hasVolumeIndexForDevice(curSrcDevice);
        }
        if (applyVolume) {
            //FIXME: workaround for truncated touch sounds
            // delayed volume change for system stream to be removed when the problem is
            // handled by system UI
            status_t volStatus = checkAndSetVolume(
                        curves, vs, index, desc, curDevice,
                        ((vs == toVolumeSource(AUDIO_STREAM_SYSTEM))?
                             TOUCH_SOUND_FIXED_DELAY_MS : 0));
            if (volStatus != NO_ERROR) {
                status = volStatus;
            }
        }
    }
    mpClientInterface->onAudioVolumeGroupChanged(group, 0 /*flags*/);
    return status;
}

status_t AudioPolicyManager::setVolumeCurveIndex(int index,
                                                 audio_devices_t device,
                                                 IVolumeCurves &volumeCurves)
{
    // VOICE_CALL stream has minVolumeIndex > 0  but can be muted directly by an
    // app that has MODIFY_PHONE_STATE permission.
    bool hasVoice = hasVoiceStream(volumeCurves.getStreamTypes());
    if (((index < volumeCurves.getVolumeIndexMin()) && !(hasVoice && index == 0)) ||
            (index > volumeCurves.getVolumeIndexMax())) {
        ALOGD("%s: wrong index %d min=%d max=%d", __FUNCTION__, index,
              volumeCurves.getVolumeIndexMin(), volumeCurves.getVolumeIndexMax());
        return BAD_VALUE;
    }
    if (!audio_is_output_device(device)) {
        return BAD_VALUE;
    }

    // Force max volume if stream cannot be muted
    if (!volumeCurves.canBeMuted()) index = volumeCurves.getVolumeIndexMax();

    ALOGV("%s device %08x, index %d", __FUNCTION__ , device, index);
    volumeCurves.addCurrentVolumeIndex(device, index);
    return NO_ERROR;
}

status_t AudioPolicyManager::getVolumeIndexForAttributes(const audio_attributes_t &attr,
                                                         int &index,
                                                         audio_devices_t device)
{
    // if device is AUDIO_DEVICE_OUT_DEFAULT_FOR_VOLUME, return volume for device selected for this
    // stream by the engine.
    if (device == AUDIO_DEVICE_OUT_DEFAULT_FOR_VOLUME) {
        device = mEngine->getOutputDevicesForAttributes(attr, nullptr, true /*fromCache*/).types();
    }
    return getVolumeIndex(getVolumeCurves(attr), index, device);
}

status_t AudioPolicyManager::getVolumeIndex(const IVolumeCurves &curves,
                                            int &index,
                                            audio_devices_t device) const
{
    if (!audio_is_output_device(device)) {
        return BAD_VALUE;
    }
    device = Volume::getDeviceForVolume(device);
    index = curves.getVolumeIndex(device);
    ALOGV("%s: device %08x index %d", __FUNCTION__, device, index);
    return NO_ERROR;
}

status_t AudioPolicyManager::getMinVolumeIndexForAttributes(const audio_attributes_t &attr,
                                                            int &index)
{
    index = getVolumeCurves(attr).getVolumeIndexMin();
    return NO_ERROR;
}

status_t AudioPolicyManager::getMaxVolumeIndexForAttributes(const audio_attributes_t &attr,
                                                            int &index)
{
    index = getVolumeCurves(attr).getVolumeIndexMax();
    return NO_ERROR;
}

audio_io_handle_t AudioPolicyManager::selectOutputForMusicEffects()
{
    // select one output among several suitable for global effects.
    // The priority is as follows:
    // 1: An offloaded output. If the effect ends up not being offloadable,
    //    AudioFlinger will invalidate the track and the offloaded output
    //    will be closed causing the effect to be moved to a PCM output.
    // 2: A deep buffer output
    // 3: The primary output
    // 4: the first output in the list

    DeviceVector devices = mEngine->getOutputDevicesForAttributes(
                attributes_initializer(AUDIO_USAGE_MEDIA), nullptr, false /*fromCache*/);
    SortedVector<audio_io_handle_t> outputs = getOutputsForDevices(devices, mOutputs);

    if (outputs.size() == 0) {
        return AUDIO_IO_HANDLE_NONE;
    }

    audio_io_handle_t output = AUDIO_IO_HANDLE_NONE;
    bool activeOnly = true;

    while (output == AUDIO_IO_HANDLE_NONE) {
        audio_io_handle_t outputOffloaded = AUDIO_IO_HANDLE_NONE;
        audio_io_handle_t outputDeepBuffer = AUDIO_IO_HANDLE_NONE;
        audio_io_handle_t outputPrimary = AUDIO_IO_HANDLE_NONE;

        for (audio_io_handle_t output : outputs) {
            sp<SwAudioOutputDescriptor> desc = mOutputs.valueFor(output);
            if (activeOnly && !desc->isActive(toVolumeSource(AUDIO_STREAM_MUSIC))) {
                continue;
            }
            ALOGV("selectOutputForMusicEffects activeOnly %d output %d flags 0x%08x",
                  activeOnly, output, desc->mFlags);
            if ((desc->mFlags & AUDIO_OUTPUT_FLAG_COMPRESS_OFFLOAD) != 0) {
                outputOffloaded = output;
            }
            if ((desc->mFlags & AUDIO_OUTPUT_FLAG_DEEP_BUFFER) != 0) {
                outputDeepBuffer = output;
            }
            if ((desc->mFlags & AUDIO_OUTPUT_FLAG_PRIMARY) != 0) {
                outputPrimary = output;
            }
        }
        if (outputOffloaded != AUDIO_IO_HANDLE_NONE) {
            output = outputOffloaded;
        } else if (outputDeepBuffer != AUDIO_IO_HANDLE_NONE) {
            output = outputDeepBuffer;
        } else if (outputPrimary != AUDIO_IO_HANDLE_NONE) {
            output = outputPrimary;
        } else {
            output = outputs[0];
        }
        activeOnly = false;
    }

    if (output != mMusicEffectOutput) {
        mEffects.moveEffects(AUDIO_SESSION_OUTPUT_MIX, mMusicEffectOutput, output);
        mpClientInterface->moveEffects(AUDIO_SESSION_OUTPUT_MIX, mMusicEffectOutput, output);
        mMusicEffectOutput = output;
    }

    ALOGV("selectOutputForMusicEffects selected output %d", output);
    return output;
}

audio_io_handle_t AudioPolicyManager::getOutputForEffect(const effect_descriptor_t *desc __unused)
{
    return selectOutputForMusicEffects();
}

status_t AudioPolicyManager::registerEffect(const effect_descriptor_t *desc,
                                audio_io_handle_t io,
                                uint32_t strategy,
                                int session,
                                int id)
{
    ssize_t index = mOutputs.indexOfKey(io);
    if (index < 0) {
        index = mInputs.indexOfKey(io);
        if (index < 0) {
            ALOGW("registerEffect() unknown io %d", io);
            return INVALID_OPERATION;
        }
    }
    return mEffects.registerEffect(desc, io, session, id,
                                   (strategy == streamToStrategy(AUDIO_STREAM_MUSIC) ||
                                   strategy == PRODUCT_STRATEGY_NONE));
}

status_t AudioPolicyManager::unregisterEffect(int id)
{
    if (mEffects.getEffect(id) == nullptr) {
        return INVALID_OPERATION;
    }
    if (mEffects.isEffectEnabled(id)) {
        ALOGW("%s effect %d enabled", __FUNCTION__, id);
        setEffectEnabled(id, false);
    }
    return mEffects.unregisterEffect(id);
}

void AudioPolicyManager::cleanUpEffectsForIo(audio_io_handle_t io)
{
    EffectDescriptorCollection effects = mEffects.getEffectsForIo(io);
    for (size_t i = 0; i < effects.size(); i++) {
        ALOGW("%s removing stale effect %s, id %d on closed IO %d",
              __func__, effects.valueAt(i)->mDesc.name, effects.keyAt(i), io);
        unregisterEffect(effects.keyAt(i));
    }
}

status_t AudioPolicyManager::setEffectEnabled(int id, bool enabled)
{
    sp<EffectDescriptor> effect = mEffects.getEffect(id);
    if (effect == nullptr) {
        return INVALID_OPERATION;
    }

    status_t status = mEffects.setEffectEnabled(id, enabled);
    if (status == NO_ERROR) {
        mInputs.trackEffectEnabled(effect, enabled);
    }
    return status;
}


status_t AudioPolicyManager::moveEffectsToIo(const std::vector<int>& ids, audio_io_handle_t io)
{
   mEffects.moveEffects(ids, io);
   return NO_ERROR;
}

bool AudioPolicyManager::isStreamActive(audio_stream_type_t stream, uint32_t inPastMs) const
{
    return mOutputs.isActive(toVolumeSource(stream), inPastMs);
}

bool AudioPolicyManager::isStreamActiveRemotely(audio_stream_type_t stream, uint32_t inPastMs) const
{
    return mOutputs.isActiveRemotely(toVolumeSource(stream), inPastMs);
}

bool AudioPolicyManager::isSourceActive(audio_source_t source) const
{
    for (size_t i = 0; i < mInputs.size(); i++) {
        const sp<AudioInputDescriptor>  inputDescriptor = mInputs.valueAt(i);
        if (inputDescriptor->isSourceActive(source)) {
            return true;
        }
    }
    return false;
}

// Register a list of custom mixes with their attributes and format.
// When a mix is registered, corresponding input and output profiles are
// added to the remote submix hw module. The profile contains only the
// parameters (sampling rate, format...) specified by the mix.
// The corresponding input remote submix device is also connected.
//
// When a remote submix device is connected, the address is checked to select the
// appropriate profile and the corresponding input or output stream is opened.
//
// When capture starts, getInputForAttr() will:
//  - 1 look for a mix matching the address passed in attribtutes tags if any
//  - 2 if none found, getDeviceForInputSource() will:
//     - 2.1 look for a mix matching the attributes source
//     - 2.2 if none found, default to device selection by policy rules
// At this time, the corresponding output remote submix device is also connected
// and active playback use cases can be transferred to this mix if needed when reconnecting
// after AudioTracks are invalidated
//
// When playback starts, getOutputForAttr() will:
//  - 1 look for a mix matching the address passed in attribtutes tags if any
//  - 2 if none found, look for a mix matching the attributes usage
//  - 3 if none found, default to device and output selection by policy rules.

status_t AudioPolicyManager::registerPolicyMixes(const Vector<AudioMix>& mixes)
{
    ALOGV("registerPolicyMixes() %zu mix(es)", mixes.size());
    status_t res = NO_ERROR;

    sp<HwModule> rSubmixModule;
    // examine each mix's route type
    for (size_t i = 0; i < mixes.size(); i++) {
        AudioMix mix = mixes[i];
        // Only capture of playback is allowed in LOOP_BACK & RENDER mode
        if (is_mix_loopback_render(mix.mRouteFlags) && mix.mMixType != MIX_TYPE_PLAYERS) {
            ALOGE("Unsupported Policy Mix %zu of %zu: "
                  "Only capture of playback is allowed in LOOP_BACK & RENDER mode",
                   i, mixes.size());
            res = INVALID_OPERATION;
            break;
        }
        // LOOP_BACK and LOOP_BACK | RENDER have the same remote submix backend and are handled
        // in the same way.
        if ((mix.mRouteFlags & MIX_ROUTE_FLAG_LOOP_BACK) == MIX_ROUTE_FLAG_LOOP_BACK) {
            ALOGV("registerPolicyMixes() mix %zu of %zu is LOOP_BACK %d", i, mixes.size(),
                  mix.mRouteFlags);
            if (rSubmixModule == 0) {
                rSubmixModule = mHwModules.getModuleFromName(
                        AUDIO_HARDWARE_MODULE_ID_REMOTE_SUBMIX);
                if (rSubmixModule == 0) {
                    ALOGE("Unable to find audio module for submix, aborting mix %zu registration",
                            i);
                    res = INVALID_OPERATION;
                    break;
                }
            }

            String8 address = mix.mDeviceAddress;
            if (mix.mMixType == MIX_TYPE_PLAYERS) {
                mix.mDeviceType = AUDIO_DEVICE_IN_REMOTE_SUBMIX;
            } else {
                mix.mDeviceType = AUDIO_DEVICE_OUT_REMOTE_SUBMIX;
            }

            if (mPolicyMixes.registerMix(address, mix, 0 /*output desc*/) != NO_ERROR) {
                ALOGE("Error registering mix %zu for address %s", i, address.string());
                res = INVALID_OPERATION;
                break;
            }
            audio_config_t outputConfig = mix.mFormat;
            audio_config_t inputConfig = mix.mFormat;
            // NOTE: audio flinger mixer does not support mono output: configure remote submix HAL in
            // stereo and let audio flinger do the channel conversion if needed.
            outputConfig.channel_mask = AUDIO_CHANNEL_OUT_STEREO;
            inputConfig.channel_mask = AUDIO_CHANNEL_IN_STEREO;
            rSubmixModule->addOutputProfile(address, &outputConfig,
                    AUDIO_DEVICE_OUT_REMOTE_SUBMIX, address);
            rSubmixModule->addInputProfile(address, &inputConfig,
                    AUDIO_DEVICE_IN_REMOTE_SUBMIX, address);

            if ((res = setDeviceConnectionStateInt(mix.mDeviceType,
                    AUDIO_POLICY_DEVICE_STATE_AVAILABLE,
                    address.string(), "remote-submix", AUDIO_FORMAT_DEFAULT)) != NO_ERROR) {
                ALOGE("Failed to set remote submix device available, type %u, address %s",
                        mix.mDeviceType, address.string());
                break;
            }
        } else if ((mix.mRouteFlags & MIX_ROUTE_FLAG_RENDER) == MIX_ROUTE_FLAG_RENDER) {
            String8 address = mix.mDeviceAddress;
            audio_devices_t type = mix.mDeviceType;
            ALOGV(" registerPolicyMixes() mix %zu of %zu is RENDER, dev=0x%X addr=%s",
                    i, mixes.size(), type, address.string());

            sp<DeviceDescriptor> device = mHwModules.getDeviceDescriptor(
                    mix.mDeviceType, mix.mDeviceAddress,
                    String8(), AUDIO_FORMAT_DEFAULT);
            if (device == nullptr) {
                res = INVALID_OPERATION;
                break;
            }

            bool foundOutput = false;
            for (size_t j = 0 ; j < mOutputs.size() ; j++) {
                sp<SwAudioOutputDescriptor> desc = mOutputs.valueAt(j);

                if (desc->supportedDevices().contains(device)) {
                    if (mPolicyMixes.registerMix(address, mix, desc) != NO_ERROR) {
                        ALOGE("Could not register mix RENDER,  dev=0x%X addr=%s", type,
                              address.string());
                        res = INVALID_OPERATION;
                    } else {
                        foundOutput = true;
                    }
                    break;
                }
            }

            if (res != NO_ERROR) {
                ALOGE(" Error registering mix %zu for device 0x%X addr %s",
                        i, type, address.string());
                res = INVALID_OPERATION;
                break;
            } else if (!foundOutput) {
                ALOGE(" Output not found for mix %zu for device 0x%X addr %s",
                        i, type, address.string());
                res = INVALID_OPERATION;
                break;
            }
        }
    }
    if (res != NO_ERROR) {
        unregisterPolicyMixes(mixes);
    }
    return res;
}

status_t AudioPolicyManager::unregisterPolicyMixes(Vector<AudioMix> mixes)
{
    ALOGV("unregisterPolicyMixes() num mixes %zu", mixes.size());
    status_t res = NO_ERROR;
    sp<HwModule> rSubmixModule;
    // examine each mix's route type
    for (const auto& mix : mixes) {
        if ((mix.mRouteFlags & MIX_ROUTE_FLAG_LOOP_BACK) == MIX_ROUTE_FLAG_LOOP_BACK) {

            if (rSubmixModule == 0) {
                rSubmixModule = mHwModules.getModuleFromName(
                        AUDIO_HARDWARE_MODULE_ID_REMOTE_SUBMIX);
                if (rSubmixModule == 0) {
                    res = INVALID_OPERATION;
                    continue;
                }
            }

            String8 address = mix.mDeviceAddress;

            if (mPolicyMixes.unregisterMix(address) != NO_ERROR) {
                res = INVALID_OPERATION;
                continue;
            }

            for (auto device : {AUDIO_DEVICE_IN_REMOTE_SUBMIX, AUDIO_DEVICE_OUT_REMOTE_SUBMIX}) {
                if (getDeviceConnectionState(device, address.string()) ==
                        AUDIO_POLICY_DEVICE_STATE_AVAILABLE)  {
                    res = setDeviceConnectionStateInt(device, AUDIO_POLICY_DEVICE_STATE_UNAVAILABLE,
                                                      address.string(), "remote-submix",
                                                      AUDIO_FORMAT_DEFAULT);
                    if (res != OK) {
                        ALOGE("Error making RemoteSubmix device unavailable for mix "
                              "with type %d, address %s", device, address.string());
                    }
                }
            }
            rSubmixModule->removeOutputProfile(address);
            rSubmixModule->removeInputProfile(address);

        } else if ((mix.mRouteFlags & MIX_ROUTE_FLAG_RENDER) == MIX_ROUTE_FLAG_RENDER) {
            if (mPolicyMixes.unregisterMix(mix.mDeviceAddress) != NO_ERROR) {
                res = INVALID_OPERATION;
                continue;
            }
        }
    }
    return res;
}

void AudioPolicyManager::dumpManualSurroundFormats(String8 *dst) const
{
    size_t i = 0;
    constexpr size_t audioFormatPrefixLen = sizeof("AUDIO_FORMAT_");
    for (const auto& fmt : mManualSurroundFormats) {
        if (i++ != 0) dst->append(", ");
        std::string sfmt;
        FormatConverter::toString(fmt, sfmt);
        dst->append(sfmt.size() >= audioFormatPrefixLen ?
                sfmt.c_str() + audioFormatPrefixLen - 1 : sfmt.c_str());
    }
}

status_t AudioPolicyManager::setUidDeviceAffinities(uid_t uid,
        const Vector<AudioDeviceTypeAddr>& devices) {
    ALOGV("%s() uid=%d num devices %zu", __FUNCTION__, uid, devices.size());
    // uid/device affinity is only for output devices
    for (size_t i = 0; i < devices.size(); i++) {
        if (!audio_is_output_device(devices[i].mType)) {
            ALOGE("setUidDeviceAffinities() device=%08x is NOT an output device",
                    devices[i].mType);
            return BAD_VALUE;
        }
    }
    status_t res =  mPolicyMixes.setUidDeviceAffinities(uid, devices);
    if (res == NO_ERROR) {
        // reevaluate outputs for all given devices
        for (size_t i = 0; i < devices.size(); i++) {
            sp<DeviceDescriptor> devDesc = mHwModules.getDeviceDescriptor(
                            devices[i].mType, devices[i].mAddress, String8(),
                            AUDIO_FORMAT_DEFAULT);
            SortedVector<audio_io_handle_t> outputs;
            if (checkOutputsForDevice(devDesc, AUDIO_POLICY_DEVICE_STATE_AVAILABLE,
                    outputs) != NO_ERROR) {
                ALOGE("setUidDeviceAffinities() error in checkOutputsForDevice for device=%08x"
                        " addr=%s", devices[i].mType, devices[i].mAddress.string());
                return INVALID_OPERATION;
            }
        }
    }
    return res;
}

status_t AudioPolicyManager::removeUidDeviceAffinities(uid_t uid) {
    ALOGV("%s() uid=%d", __FUNCTION__, uid);
    status_t res = mPolicyMixes.removeUidDeviceAffinities(uid);
    if (res != NO_ERROR) {
        ALOGE("%s() Could not remove all device affinities fo uid = %d",
            __FUNCTION__, uid);
        return INVALID_OPERATION;
    }

    return res;
}

void AudioPolicyManager::dump(String8 *dst) const
{
    dst->appendFormat("\nAudioPolicyManager Dump: %p\n", this);
    dst->appendFormat(" Primary Output: %d\n",
             hasPrimaryOutput() ? mPrimaryOutput->mIoHandle : AUDIO_IO_HANDLE_NONE);
    std::string stateLiteral;
    AudioModeConverter::toString(mEngine->getPhoneState(), stateLiteral);
    dst->appendFormat(" Phone state: %s\n", stateLiteral.c_str());
    const char* forceUses[AUDIO_POLICY_FORCE_USE_CNT] = {
        "communications", "media", "record", "dock", "system",
        "HDMI system audio", "encoded surround output", "vibrate ringing" };
    for (audio_policy_force_use_t i = AUDIO_POLICY_FORCE_FOR_COMMUNICATION;
         i < AUDIO_POLICY_FORCE_USE_CNT; i = (audio_policy_force_use_t)((int)i + 1)) {
        audio_policy_forced_cfg_t forceUseValue = mEngine->getForceUse(i);
        dst->appendFormat(" Force use for %s: %d", forceUses[i], forceUseValue);
        if (i == AUDIO_POLICY_FORCE_FOR_ENCODED_SURROUND &&
                forceUseValue == AUDIO_POLICY_FORCE_ENCODED_SURROUND_MANUAL) {
            dst->append(" (MANUAL: ");
            dumpManualSurroundFormats(dst);
            dst->append(")");
        }
        dst->append("\n");
    }
    dst->appendFormat(" TTS output %savailable\n", mTtsOutputAvailable ? "" : "not ");
    dst->appendFormat(" Master mono: %s\n", mMasterMono ? "on" : "off");
    dst->appendFormat(" Config source: %s\n", mConfig.getSource().c_str()); // getConfig not const
    mAvailableOutputDevices.dump(dst, String8("Available output"));
    mAvailableInputDevices.dump(dst, String8("Available input"));
    mHwModulesAll.dump(dst);
    mOutputs.dump(dst);
    mInputs.dump(dst);
    mEffects.dump(dst);
    mAudioPatches.dump(dst);
    mPolicyMixes.dump(dst);
    mAudioSources.dump(dst);

    dst->appendFormat(" AllowedCapturePolicies:\n");
    for (auto& policy : mAllowedCapturePolicies) {
        dst->appendFormat("   - uid=%d flag_mask=%#x\n", policy.first, policy.second);
    }

    dst->appendFormat("\nPolicy Engine dump:\n");
    mEngine->dump(dst);
}

status_t AudioPolicyManager::dump(int fd)
{
    String8 result;
    dump(&result);
    write(fd, result.string(), result.size());
    return NO_ERROR;
}

status_t AudioPolicyManager::setAllowedCapturePolicy(uid_t uid, audio_flags_mask_t capturePolicy)
{
    mAllowedCapturePolicies[uid] = capturePolicy;
    return NO_ERROR;
}

// This function checks for the parameters which can be offloaded.
// This can be enhanced depending on the capability of the DSP and policy
// of the system.
bool AudioPolicyManager::isOffloadSupported(const audio_offload_info_t& offloadInfo)
{
    ALOGV("isOffloadSupported: SR=%u, CM=0x%x, Format=0x%x, StreamType=%d,"
     " BitRate=%u, duration=%" PRId64 " us, has_video=%d",
     offloadInfo.sample_rate, offloadInfo.channel_mask,
     offloadInfo.format,
     offloadInfo.stream_type, offloadInfo.bit_rate, offloadInfo.duration_us,
     offloadInfo.has_video);

    if (mMasterMono) {
        return false; // no offloading if mono is set.
    }

    // Check if offload has been disabled
    if (property_get_bool("audio.offload.disable", false /* default_value */)) {
        ALOGV("offload disabled by audio.offload.disable");
        return false;
    }

    // Check if stream type is music, then only allow offload as of now.
    if (offloadInfo.stream_type != AUDIO_STREAM_MUSIC)
    {
        ALOGV("isOffloadSupported: stream_type != MUSIC, returning false");
        return false;
    }

    //TODO: enable audio offloading with video when ready
    const bool allowOffloadWithVideo =
            property_get_bool("audio.offload.video", false /* default_value */);
    if (offloadInfo.has_video && !allowOffloadWithVideo) {
        ALOGV("isOffloadSupported: has_video == true, returning false");
        return false;
    }

    //If duration is less than minimum value defined in property, return false
    const int min_duration_secs = property_get_int32(
            "audio.offload.min.duration.secs", -1 /* default_value */);
    if (min_duration_secs >= 0) {
        if (offloadInfo.duration_us < min_duration_secs * 1000000LL) {
            ALOGV("Offload denied by duration < audio.offload.min.duration.secs(=%d)",
                    min_duration_secs);
            return false;
        }
    } else if (offloadInfo.duration_us < OFFLOAD_DEFAULT_MIN_DURATION_SECS * 1000000) {
        ALOGV("Offload denied by duration < default min(=%u)", OFFLOAD_DEFAULT_MIN_DURATION_SECS);
        return false;
    }

    // Do not allow offloading if one non offloadable effect is enabled. This prevents from
    // creating an offloaded track and tearing it down immediately after start when audioflinger
    // detects there is an active non offloadable effect.
    // FIXME: We should check the audio session here but we do not have it in this context.
    // This may prevent offloading in rare situations where effects are left active by apps
    // in the background.
    if (mEffects.isNonOffloadableEffectEnabled()) {
        return false;
    }

    // See if there is a profile to support this.
    // AUDIO_DEVICE_NONE
    sp<IOProfile> profile = getProfileForOutput(DeviceVector() /*ignore device */,
                                            offloadInfo.sample_rate,
                                            offloadInfo.format,
                                            offloadInfo.channel_mask,
                                            AUDIO_OUTPUT_FLAG_COMPRESS_OFFLOAD,
                                            true /* directOnly */);
    ALOGV("isOffloadSupported() profile %sfound", profile != 0 ? "" : "NOT ");
    return (profile != 0);
}

bool AudioPolicyManager::isDirectOutputSupported(const audio_config_base_t& config,
                                                 const audio_attributes_t& attributes) {
    audio_output_flags_t output_flags = AUDIO_OUTPUT_FLAG_NONE;
    audio_flags_to_audio_output_flags(attributes.flags, &output_flags);
    sp<IOProfile> profile = getProfileForOutput(DeviceVector() /*ignore device */,
                                            config.sample_rate,
                                            config.format,
                                            config.channel_mask,
                                            output_flags,
                                            true /* directOnly */);
    ALOGV("%s() profile %sfound with name: %s, "
        "sample rate: %u, format: 0x%x, channel_mask: 0x%x, output flags: 0x%x",
        __FUNCTION__, profile != 0 ? "" : "NOT ",
        (profile != 0 ? profile->getTagName().string() : "null"),
        config.sample_rate, config.format, config.channel_mask, output_flags);
    return (profile != 0);
}

status_t AudioPolicyManager::listAudioPorts(audio_port_role_t role,
                                            audio_port_type_t type,
                                            unsigned int *num_ports,
                                            struct audio_port *ports,
                                            unsigned int *generation)
{
    if (num_ports == NULL || (*num_ports != 0 && ports == NULL) ||
            generation == NULL) {
        return BAD_VALUE;
    }
    ALOGV("listAudioPorts() role %d type %d num_ports %d ports %p", role, type, *num_ports, ports);
    if (ports == NULL) {
        *num_ports = 0;
    }

    size_t portsWritten = 0;
    size_t portsMax = *num_ports;
    *num_ports = 0;
    if (type == AUDIO_PORT_TYPE_NONE || type == AUDIO_PORT_TYPE_DEVICE) {
        // do not report devices with type AUDIO_DEVICE_IN_STUB or AUDIO_DEVICE_OUT_STUB
        // as they are used by stub HALs by convention
        if (role == AUDIO_PORT_ROLE_SINK || role == AUDIO_PORT_ROLE_NONE) {
            for (const auto& dev : mAvailableOutputDevices) {
                if (dev->type() == AUDIO_DEVICE_OUT_STUB) {
                    continue;
                }
                if (portsWritten < portsMax) {
                    dev->toAudioPort(&ports[portsWritten++]);
                }
                (*num_ports)++;
            }
        }
        if (role == AUDIO_PORT_ROLE_SOURCE || role == AUDIO_PORT_ROLE_NONE) {
            for (const auto& dev : mAvailableInputDevices) {
                if (dev->type() == AUDIO_DEVICE_IN_STUB) {
                    continue;
                }
                if (portsWritten < portsMax) {
                    dev->toAudioPort(&ports[portsWritten++]);
                }
                (*num_ports)++;
            }
        }
    }
    if (type == AUDIO_PORT_TYPE_NONE || type == AUDIO_PORT_TYPE_MIX) {
        if (role == AUDIO_PORT_ROLE_SINK || role == AUDIO_PORT_ROLE_NONE) {
            for (size_t i = 0; i < mInputs.size() && portsWritten < portsMax; i++) {
                mInputs[i]->toAudioPort(&ports[portsWritten++]);
            }
            *num_ports += mInputs.size();
        }
        if (role == AUDIO_PORT_ROLE_SOURCE || role == AUDIO_PORT_ROLE_NONE) {
            size_t numOutputs = 0;
            for (size_t i = 0; i < mOutputs.size(); i++) {
                if (!mOutputs[i]->isDuplicated()) {
                    numOutputs++;
                    if (portsWritten < portsMax) {
                        mOutputs[i]->toAudioPort(&ports[portsWritten++]);
                    }
                }
            }
            *num_ports += numOutputs;
        }
    }
    *generation = curAudioPortGeneration();
    ALOGV("listAudioPorts() got %zu ports needed %d", portsWritten, *num_ports);
    return NO_ERROR;
}

status_t AudioPolicyManager::getAudioPort(struct audio_port *port)
{
    if (port == nullptr || port->id == AUDIO_PORT_HANDLE_NONE) {
        return BAD_VALUE;
    }
    sp<DeviceDescriptor> dev = mAvailableOutputDevices.getDeviceFromId(port->id);
    if (dev != 0) {
        dev->toAudioPort(port);
        return NO_ERROR;
    }
    dev = mAvailableInputDevices.getDeviceFromId(port->id);
    if (dev != 0) {
        dev->toAudioPort(port);
        return NO_ERROR;
    }
    sp<SwAudioOutputDescriptor> out = mOutputs.getOutputFromId(port->id);
    if (out != 0) {
        out->toAudioPort(port);
        return NO_ERROR;
    }
    sp<AudioInputDescriptor> in = mInputs.getInputFromId(port->id);
    if (in != 0) {
        in->toAudioPort(port);
        return NO_ERROR;
    }
    return BAD_VALUE;
}

status_t AudioPolicyManager::createAudioPatch(const struct audio_patch *patch,
                                               audio_patch_handle_t *handle,
                                               uid_t uid)
{
    ALOGV("createAudioPatch()");

    if (handle == NULL || patch == NULL) {
        return BAD_VALUE;
    }
    ALOGV("createAudioPatch() num sources %d num sinks %d", patch->num_sources, patch->num_sinks);

    if (!audio_patch_is_valid(patch)) {
        return BAD_VALUE;
    }
    // only one source per audio patch supported for now
    if (patch->num_sources > 1) {
        return INVALID_OPERATION;
    }

    if (patch->sources[0].role != AUDIO_PORT_ROLE_SOURCE) {
        return INVALID_OPERATION;
    }
    for (size_t i = 0; i < patch->num_sinks; i++) {
        if (patch->sinks[i].role != AUDIO_PORT_ROLE_SINK) {
            return INVALID_OPERATION;
        }
    }

    sp<AudioPatch> patchDesc;
    ssize_t index = mAudioPatches.indexOfKey(*handle);

    ALOGV("createAudioPatch source id %d role %d type %d", patch->sources[0].id,
                                                           patch->sources[0].role,
                                                           patch->sources[0].type);
#if LOG_NDEBUG == 0
    for (size_t i = 0; i < patch->num_sinks; i++) {
        ALOGV("createAudioPatch sink %zu: id %d role %d type %d", i, patch->sinks[i].id,
                                                             patch->sinks[i].role,
                                                             patch->sinks[i].type);
    }
#endif

    if (index >= 0) {
        patchDesc = mAudioPatches.valueAt(index);
        ALOGV("createAudioPatch() mUidCached %d patchDesc->mUid %d uid %d",
                                                                  mUidCached, patchDesc->mUid, uid);
        if (patchDesc->mUid != mUidCached && uid != patchDesc->mUid) {
            return INVALID_OPERATION;
        }
    } else {
        *handle = AUDIO_PATCH_HANDLE_NONE;
    }

    if (patch->sources[0].type == AUDIO_PORT_TYPE_MIX) {
        sp<SwAudioOutputDescriptor> outputDesc = mOutputs.getOutputFromId(patch->sources[0].id);
        if (outputDesc == NULL) {
            ALOGV("createAudioPatch() output not found for id %d", patch->sources[0].id);
            return BAD_VALUE;
        }
        ALOG_ASSERT(!outputDesc->isDuplicated(),"duplicated output %d in source in ports",
                                                outputDesc->mIoHandle);
        if (patchDesc != 0) {
            if (patchDesc->mPatch.sources[0].id != patch->sources[0].id) {
                ALOGV("createAudioPatch() source id differs for patch current id %d new id %d",
                                          patchDesc->mPatch.sources[0].id, patch->sources[0].id);
                return BAD_VALUE;
            }
        }
        DeviceVector devices;
        for (size_t i = 0; i < patch->num_sinks; i++) {
            // Only support mix to devices connection
            // TODO add support for mix to mix connection
            if (patch->sinks[i].type != AUDIO_PORT_TYPE_DEVICE) {
                ALOGV("createAudioPatch() source mix but sink is not a device");
                return INVALID_OPERATION;
            }
            sp<DeviceDescriptor> devDesc =
                    mAvailableOutputDevices.getDeviceFromId(patch->sinks[i].id);
            if (devDesc == 0) {
                ALOGV("createAudioPatch() out device not found for id %d", patch->sinks[i].id);
                return BAD_VALUE;
            }

            if (!outputDesc->mProfile->isCompatibleProfile(DeviceVector(devDesc),
                                                           patch->sources[0].sample_rate,
                                                           NULL,  // updatedSamplingRate
                                                           patch->sources[0].format,
                                                           NULL,  // updatedFormat
                                                           patch->sources[0].channel_mask,
                                                           NULL,  // updatedChannelMask
                                                           AUDIO_OUTPUT_FLAG_NONE /*FIXME*/)) {
                ALOGV("createAudioPatch() profile not supported for device %08x",
                        devDesc->type());
                return INVALID_OPERATION;
            }
            devices.add(devDesc);
        }
        if (devices.size() == 0) {
            return INVALID_OPERATION;
        }

        // TODO: reconfigure output format and channels here
        ALOGV("createAudioPatch() setting device %08x on output %d",
              devices.types(), outputDesc->mIoHandle);
        setOutputDevices(outputDesc, devices, true, 0, handle);
        index = mAudioPatches.indexOfKey(*handle);
        if (index >= 0) {
            if (patchDesc != 0 && patchDesc != mAudioPatches.valueAt(index)) {
                ALOGW("createAudioPatch() setOutputDevice() did not reuse the patch provided");
            }
            patchDesc = mAudioPatches.valueAt(index);
            patchDesc->mUid = uid;
            ALOGV("createAudioPatch() success");
        } else {
            ALOGW("createAudioPatch() setOutputDevice() failed to create a patch");
            return INVALID_OPERATION;
        }
    } else if (patch->sources[0].type == AUDIO_PORT_TYPE_DEVICE) {
        if (patch->sinks[0].type == AUDIO_PORT_TYPE_MIX) {
            // input device to input mix connection
            // only one sink supported when connecting an input device to a mix
            if (patch->num_sinks > 1) {
                return INVALID_OPERATION;
            }
            sp<AudioInputDescriptor> inputDesc = mInputs.getInputFromId(patch->sinks[0].id);
            if (inputDesc == NULL) {
                return BAD_VALUE;
            }
            if (patchDesc != 0) {
                if (patchDesc->mPatch.sinks[0].id != patch->sinks[0].id) {
                    return BAD_VALUE;
                }
            }
            sp<DeviceDescriptor> device =
                    mAvailableInputDevices.getDeviceFromId(patch->sources[0].id);
            if (device == 0) {
                return BAD_VALUE;
            }

            if (!inputDesc->mProfile->isCompatibleProfile(DeviceVector(device),
                                                          patch->sinks[0].sample_rate,
                                                          NULL, /*updatedSampleRate*/
                                                          patch->sinks[0].format,
                                                          NULL, /*updatedFormat*/
                                                          patch->sinks[0].channel_mask,
                                                          NULL, /*updatedChannelMask*/
                                                          // FIXME for the parameter type,
                                                          // and the NONE
                                                          (audio_output_flags_t)
                                                            AUDIO_INPUT_FLAG_NONE)) {
                return INVALID_OPERATION;
            }
            // TODO: reconfigure output format and channels here
            ALOGV("%s() setting device %s on output %d", __func__,
                  device->toString().c_str(), inputDesc->mIoHandle);
            setInputDevice(inputDesc->mIoHandle, device, true, handle);
            index = mAudioPatches.indexOfKey(*handle);
            if (index >= 0) {
                if (patchDesc != 0 && patchDesc != mAudioPatches.valueAt(index)) {
                    ALOGW("createAudioPatch() setInputDevice() did not reuse the patch provided");
                }
                patchDesc = mAudioPatches.valueAt(index);
                patchDesc->mUid = uid;
                ALOGV("createAudioPatch() success");
            } else {
                ALOGW("createAudioPatch() setInputDevice() failed to create a patch");
                return INVALID_OPERATION;
            }
        } else if (patch->sinks[0].type == AUDIO_PORT_TYPE_DEVICE) {
            // device to device connection
            if (patchDesc != 0) {
                if (patchDesc->mPatch.sources[0].id != patch->sources[0].id) {
                    return BAD_VALUE;
                }
            }
            sp<DeviceDescriptor> srcDevice =
                    mAvailableInputDevices.getDeviceFromId(patch->sources[0].id);
            if (srcDevice == 0) {
                return BAD_VALUE;
            }

            //update source and sink with our own data as the data passed in the patch may
            // be incomplete.
            struct audio_patch newPatch = *patch;
            srcDevice->toAudioPortConfig(&newPatch.sources[0], &patch->sources[0]);

            for (size_t i = 0; i < patch->num_sinks; i++) {
                if (patch->sinks[i].type != AUDIO_PORT_TYPE_DEVICE) {
                    ALOGV("createAudioPatch() source device but one sink is not a device");
                    return INVALID_OPERATION;
                }

                sp<DeviceDescriptor> sinkDevice =
                        mAvailableOutputDevices.getDeviceFromId(patch->sinks[i].id);
                if (sinkDevice == 0) {
                    return BAD_VALUE;
                }
                sinkDevice->toAudioPortConfig(&newPatch.sinks[i], &patch->sinks[i]);

                // create a software bridge in PatchPanel if:
                // - source and sink devices are on different HW modules OR
                // - audio HAL version is < 3.0
                // - audio HAL version is >= 3.0 but no route has been declared between devices
                if (!srcDevice->hasSameHwModuleAs(sinkDevice) ||
                        (srcDevice->getModuleVersionMajor() < 3) ||
                        !srcDevice->getModule()->supportsPatch(srcDevice, sinkDevice)) {
                    // support only one sink device for now to simplify output selection logic
                    if (patch->num_sinks > 1) {
                        return INVALID_OPERATION;
                    }
                    SortedVector<audio_io_handle_t> outputs =
                            getOutputsForDevices(DeviceVector(sinkDevice), mOutputs);
                    // if the sink device is reachable via an opened output stream, request to go via
                    // this output stream by adding a second source to the patch description
                    const audio_io_handle_t output = selectOutput(outputs);
                    if (output != AUDIO_IO_HANDLE_NONE) {
                        sp<AudioOutputDescriptor> outputDesc = mOutputs.valueFor(output);
                        if (outputDesc->isDuplicated()) {
                            return INVALID_OPERATION;
                        }
                        outputDesc->toAudioPortConfig(&newPatch.sources[1], &patch->sources[0]);
                        newPatch.sources[1].ext.mix.usecase.stream = AUDIO_STREAM_PATCH;
                        newPatch.num_sources = 2;
                    }
                }
            }
            // TODO: check from routing capabilities in config file and other conflicting patches

            status_t status = installPatch(__func__, index, handle, &newPatch, 0, uid, &patchDesc);
            if (status != NO_ERROR) {
                ALOGW("createAudioPatch() patch panel could not connect device patch, error %d",
                status);
                return INVALID_OPERATION;
            }
        } else {
            return BAD_VALUE;
        }
    } else {
        return BAD_VALUE;
    }
    return NO_ERROR;
}

status_t AudioPolicyManager::releaseAudioPatch(audio_patch_handle_t handle,
                                                  uid_t uid)
{
    ALOGV("releaseAudioPatch() patch %d", handle);

    ssize_t index = mAudioPatches.indexOfKey(handle);

    if (index < 0) {
        return BAD_VALUE;
    }
    sp<AudioPatch> patchDesc = mAudioPatches.valueAt(index);
    ALOGV("releaseAudioPatch() mUidCached %d patchDesc->mUid %d uid %d",
          mUidCached, patchDesc->mUid, uid);
    if (patchDesc->mUid != mUidCached && uid != patchDesc->mUid) {
        return INVALID_OPERATION;
    }

    struct audio_patch *patch = &patchDesc->mPatch;
    patchDesc->mUid = mUidCached;
    if (patch->sources[0].type == AUDIO_PORT_TYPE_MIX) {
        sp<SwAudioOutputDescriptor> outputDesc = mOutputs.getOutputFromId(patch->sources[0].id);
        if (outputDesc == NULL) {
            ALOGV("releaseAudioPatch() output not found for id %d", patch->sources[0].id);
            return BAD_VALUE;
        }

        setOutputDevices(outputDesc,
                         getNewOutputDevices(outputDesc, true /*fromCache*/),
                         true,
                         0,
                         NULL);
    } else if (patch->sources[0].type == AUDIO_PORT_TYPE_DEVICE) {
        if (patch->sinks[0].type == AUDIO_PORT_TYPE_MIX) {
            sp<AudioInputDescriptor> inputDesc = mInputs.getInputFromId(patch->sinks[0].id);
            if (inputDesc == NULL) {
                ALOGV("releaseAudioPatch() input not found for id %d", patch->sinks[0].id);
                return BAD_VALUE;
            }
            setInputDevice(inputDesc->mIoHandle,
                           getNewInputDevice(inputDesc),
                           true,
                           NULL);
        } else if (patch->sinks[0].type == AUDIO_PORT_TYPE_DEVICE) {
            status_t status = mpClientInterface->releaseAudioPatch(patchDesc->mAfPatchHandle, 0);
            ALOGV("releaseAudioPatch() patch panel returned %d patchHandle %d",
                                                              status, patchDesc->mAfPatchHandle);
            removeAudioPatch(patchDesc->mHandle);
            nextAudioPortGeneration();
            mpClientInterface->onAudioPatchListUpdate();
        } else {
            return BAD_VALUE;
        }
    } else {
        return BAD_VALUE;
    }
    return NO_ERROR;
}

status_t AudioPolicyManager::listAudioPatches(unsigned int *num_patches,
                                              struct audio_patch *patches,
                                              unsigned int *generation)
{
    if (generation == NULL) {
        return BAD_VALUE;
    }
    *generation = curAudioPortGeneration();
    return mAudioPatches.listAudioPatches(num_patches, patches);
}

status_t AudioPolicyManager::setAudioPortConfig(const struct audio_port_config *config)
{
    ALOGV("setAudioPortConfig()");

    if (config == NULL) {
        return BAD_VALUE;
    }
    ALOGV("setAudioPortConfig() on port handle %d", config->id);
    // Only support gain configuration for now
    if (config->config_mask != AUDIO_PORT_CONFIG_GAIN) {
        return INVALID_OPERATION;
    }

    sp<AudioPortConfig> audioPortConfig;
    if (config->type == AUDIO_PORT_TYPE_MIX) {
        if (config->role == AUDIO_PORT_ROLE_SOURCE) {
            sp<SwAudioOutputDescriptor> outputDesc = mOutputs.getOutputFromId(config->id);
            if (outputDesc == NULL) {
                return BAD_VALUE;
            }
            ALOG_ASSERT(!outputDesc->isDuplicated(),
                        "setAudioPortConfig() called on duplicated output %d",
                        outputDesc->mIoHandle);
            audioPortConfig = outputDesc;
        } else if (config->role == AUDIO_PORT_ROLE_SINK) {
            sp<AudioInputDescriptor> inputDesc = mInputs.getInputFromId(config->id);
            if (inputDesc == NULL) {
                return BAD_VALUE;
            }
            audioPortConfig = inputDesc;
        } else {
            return BAD_VALUE;
        }
    } else if (config->type == AUDIO_PORT_TYPE_DEVICE) {
        sp<DeviceDescriptor> deviceDesc;
        if (config->role == AUDIO_PORT_ROLE_SOURCE) {
            deviceDesc = mAvailableInputDevices.getDeviceFromId(config->id);
        } else if (config->role == AUDIO_PORT_ROLE_SINK) {
            deviceDesc = mAvailableOutputDevices.getDeviceFromId(config->id);
        } else {
            return BAD_VALUE;
        }
        if (deviceDesc == NULL) {
            return BAD_VALUE;
        }
        audioPortConfig = deviceDesc;
    } else {
        return BAD_VALUE;
    }

    struct audio_port_config backupConfig = {};
    status_t status = audioPortConfig->applyAudioPortConfig(config, &backupConfig);
    if (status == NO_ERROR) {
        struct audio_port_config newConfig = {};
        audioPortConfig->toAudioPortConfig(&newConfig, config);
        status = mpClientInterface->setAudioPortConfig(&newConfig, 0);
    }
    if (status != NO_ERROR) {
        audioPortConfig->applyAudioPortConfig(&backupConfig);
    }

    return status;
}

void AudioPolicyManager::releaseResourcesForUid(uid_t uid)
{
    clearAudioSources(uid);
    clearAudioPatches(uid);
    clearSessionRoutes(uid);
}

void AudioPolicyManager::clearAudioPatches(uid_t uid)
{
    for (ssize_t i = (ssize_t)mAudioPatches.size() - 1; i >= 0; i--)  {
        sp<AudioPatch> patchDesc = mAudioPatches.valueAt(i);
        if (patchDesc->mUid == uid) {
            releaseAudioPatch(mAudioPatches.keyAt(i), uid);
        }
    }
}

void AudioPolicyManager::checkStrategyRoute(product_strategy_t ps, audio_io_handle_t ouptutToSkip)
{
    // Take the first attributes following the product strategy as it is used to retrieve the routed
    // device. All attributes wihin a strategy follows the same "routing strategy"
    auto attributes = mEngine->getAllAttributesForProductStrategy(ps).front();
    DeviceVector devices = mEngine->getOutputDevicesForAttributes(attributes, nullptr, false);
    SortedVector<audio_io_handle_t> outputs = getOutputsForDevices(devices, mOutputs);
    for (size_t j = 0; j < mOutputs.size(); j++) {
        if (mOutputs.keyAt(j) == ouptutToSkip) {
            continue;
        }
        sp<SwAudioOutputDescriptor> outputDesc = mOutputs.valueAt(j);
        if (!outputDesc->isStrategyActive(ps)) {
            continue;
        }
        // If the default device for this strategy is on another output mix,
        // invalidate all tracks in this strategy to force re connection.
        // Otherwise select new device on the output mix.
        if (outputs.indexOf(mOutputs.keyAt(j)) < 0) {
            for (auto stream : mEngine->getStreamTypesForProductStrategy(ps)) {
                mpClientInterface->invalidateStream(stream);
            }
        } else {
            setOutputDevices(
                        outputDesc, getNewOutputDevices(outputDesc, false /*fromCache*/), false);
        }
    }
}

void AudioPolicyManager::clearSessionRoutes(uid_t uid)
{
    // remove output routes associated with this uid
    std::vector<product_strategy_t> affectedStrategies;
    for (size_t i = 0; i < mOutputs.size(); i++) {
        sp<AudioOutputDescriptor> outputDesc = mOutputs.valueAt(i);
        for (const auto& client : outputDesc->getClientIterable()) {
            if (client->hasPreferredDevice() && client->uid() == uid) {
                client->setPreferredDeviceId(AUDIO_PORT_HANDLE_NONE);
                auto clientStrategy = client->strategy();
                if (std::find(begin(affectedStrategies), end(affectedStrategies), clientStrategy) !=
                        end(affectedStrategies)) {
                    continue;
                }
                affectedStrategies.push_back(client->strategy());
            }
        }
    }
    // reroute outputs if necessary
    for (const auto& strategy : affectedStrategies) {
        checkStrategyRoute(strategy, AUDIO_IO_HANDLE_NONE);
    }

    // remove input routes associated with this uid
    SortedVector<audio_source_t> affectedSources;
    for (size_t i = 0; i < mInputs.size(); i++) {
        sp<AudioInputDescriptor> inputDesc = mInputs.valueAt(i);
        for (const auto& client : inputDesc->getClientIterable()) {
            if (client->hasPreferredDevice() && client->uid() == uid) {
                client->setPreferredDeviceId(AUDIO_PORT_HANDLE_NONE);
                affectedSources.add(client->source());
            }
        }
    }
    // reroute inputs if necessary
    SortedVector<audio_io_handle_t> inputsToClose;
    for (size_t i = 0; i < mInputs.size(); i++) {
        sp<AudioInputDescriptor> inputDesc = mInputs.valueAt(i);
        if (affectedSources.indexOf(inputDesc->source()) >= 0) {
            inputsToClose.add(inputDesc->mIoHandle);
        }
    }
    for (const auto& input : inputsToClose) {
        closeInput(input);
    }
}

void AudioPolicyManager::clearAudioSources(uid_t uid)
{
    for (ssize_t i = (ssize_t)mAudioSources.size() - 1; i >= 0; i--)  {
        sp<SourceClientDescriptor> sourceDesc = mAudioSources.valueAt(i);
        if (sourceDesc->uid() == uid) {
            stopAudioSource(mAudioSources.keyAt(i));
        }
    }
}

status_t AudioPolicyManager::acquireSoundTriggerSession(audio_session_t *session,
                                       audio_io_handle_t *ioHandle,
                                       audio_devices_t *device)
{
    *session = (audio_session_t)mpClientInterface->newAudioUniqueId(AUDIO_UNIQUE_ID_USE_SESSION);
    *ioHandle = (audio_io_handle_t)mpClientInterface->newAudioUniqueId(AUDIO_UNIQUE_ID_USE_INPUT);
    audio_attributes_t attr = { .source = AUDIO_SOURCE_HOTWORD };
    *device = mEngine->getInputDeviceForAttributes(attr)->type();

    return mSoundTriggerSessions.acquireSession(*session, *ioHandle);
}

status_t AudioPolicyManager::startAudioSource(const struct audio_port_config *source,
                                              const audio_attributes_t *attributes,
                                              audio_port_handle_t *portId,
                                              uid_t uid)
{
    ALOGV("%s", __FUNCTION__);
    *portId = AUDIO_PORT_HANDLE_NONE;

    if (source == NULL || attributes == NULL || portId == NULL) {
        ALOGW("%s invalid argument: source %p attributes %p handle %p",
              __FUNCTION__, source, attributes, portId);
        return BAD_VALUE;
    }

    if (source->role != AUDIO_PORT_ROLE_SOURCE ||
            source->type != AUDIO_PORT_TYPE_DEVICE) {
        ALOGW("%s INVALID_OPERATION source->role %d source->type %d",
              __FUNCTION__, source->role, source->type);
        return INVALID_OPERATION;
    }

    sp<DeviceDescriptor> srcDevice =
            mAvailableInputDevices.getDevice(source->ext.device.type,
                                             String8(source->ext.device.address),
                                             AUDIO_FORMAT_DEFAULT);
    if (srcDevice == 0) {
        ALOGW("%s source->ext.device.type %08x not found", __FUNCTION__, source->ext.device.type);
        return BAD_VALUE;
    }

    *portId = AudioPort::getNextUniqueId();

    struct audio_patch dummyPatch = {};
    sp<AudioPatch> patchDesc = new AudioPatch(&dummyPatch, uid);

    sp<SourceClientDescriptor> sourceDesc =
        new SourceClientDescriptor(*portId, uid, *attributes, patchDesc, srcDevice,
                                   mEngine->getStreamTypeForAttributes(*attributes),
                                   mEngine->getProductStrategyForAttributes(*attributes),
                                   toVolumeSource(*attributes));

    status_t status = connectAudioSource(sourceDesc);
    if (status == NO_ERROR) {
        mAudioSources.add(*portId, sourceDesc);
    }
    return status;
}

status_t AudioPolicyManager::connectAudioSource(const sp<SourceClientDescriptor>& sourceDesc)
{
    ALOGV("%s handle %d", __FUNCTION__, sourceDesc->portId());

    // make sure we only have one patch per source.
    disconnectAudioSource(sourceDesc);

    audio_attributes_t attributes = sourceDesc->attributes();
    audio_stream_type_t stream = sourceDesc->stream();
    sp<DeviceDescriptor> srcDevice = sourceDesc->srcDevice();

    DeviceVector sinkDevices =
            mEngine->getOutputDevicesForAttributes(attributes, nullptr, true);
    ALOG_ASSERT(!sinkDevices.isEmpty(), "connectAudioSource(): no device found for attributes");
    sp<DeviceDescriptor> sinkDevice = sinkDevices.itemAt(0);
    ALOG_ASSERT(mAvailableOutputDevices.contains(sinkDevice), "%s: Device %s not available",
                __FUNCTION__, sinkDevice->toString().c_str());

    audio_patch_handle_t afPatchHandle = AUDIO_PATCH_HANDLE_NONE;

    if (srcDevice->hasSameHwModuleAs(sinkDevice) &&
            srcDevice->getModuleVersionMajor() >= 3 &&
            sinkDevice->getModule()->supportsPatch(srcDevice, sinkDevice) &&
            srcDevice->getAudioPort()->mGains.size() > 0) {
        ALOGV("%s Device to Device route supported by >=3.0 HAL", __FUNCTION__);
        // TODO: may explicitly specify whether we should use HW or SW patch
        //   create patch between src device and output device
        //   create Hwoutput and add to mHwOutputs
    } else {
        audio_attributes_t resultAttr;
        audio_io_handle_t output = AUDIO_IO_HANDLE_NONE;
        audio_config_t config = AUDIO_CONFIG_INITIALIZER;
        config.sample_rate = sourceDesc->config().sample_rate;
        config.channel_mask = sourceDesc->config().channel_mask;
        config.format = sourceDesc->config().format;
        audio_output_flags_t flags = AUDIO_OUTPUT_FLAG_NONE;
        audio_port_handle_t selectedDeviceId = AUDIO_PORT_HANDLE_NONE;
        bool isRequestedDeviceForExclusiveUse = false;
        std::vector<sp<SwAudioOutputDescriptor>> secondaryOutputs;
        getOutputForAttrInt(&resultAttr, &output, AUDIO_SESSION_NONE,
                &attributes, &stream, sourceDesc->uid(), &config, &flags,
                &selectedDeviceId, &isRequestedDeviceForExclusiveUse,
                &secondaryOutputs);
        if (output == AUDIO_IO_HANDLE_NONE) {
            ALOGV("%s no output for device %08x", __FUNCTION__, sinkDevices.types());
            return INVALID_OPERATION;
        }
        sp<SwAudioOutputDescriptor> outputDesc = mOutputs.valueFor(output);
        if (outputDesc->isDuplicated()) {
            ALOGV("%s output for device %08x is duplicated", __FUNCTION__, sinkDevices.types());
            return INVALID_OPERATION;
        }
        status_t status = outputDesc->start();
        if (status != NO_ERROR) {
            return status;
        }

        // create a special patch with no sink and two sources:
        // - the second source indicates to PatchPanel through which output mix this patch should
        // be connected as well as the stream type for volume control
        // - the sink is defined by whatever output device is currently selected for the output
        // though which this patch is routed.
        PatchBuilder patchBuilder;
        patchBuilder.addSource(srcDevice).addSource(outputDesc, { .stream = stream });
        status = mpClientInterface->createAudioPatch(patchBuilder.patch(),
                                                              &afPatchHandle,
                                                              0);
        ALOGV("%s patch panel returned %d patchHandle %d", __FUNCTION__,
                                                              status, afPatchHandle);
        sourceDesc->patchDesc()->mPatch = *patchBuilder.patch();
        if (status != NO_ERROR) {
            ALOGW("%s patch panel could not connect device patch, error %d",
                  __FUNCTION__, status);
            return INVALID_OPERATION;
        }

        if (outputDesc->getClient(sourceDesc->portId()) != nullptr) {
            ALOGW("%s source portId has already been attached to outputDesc", __func__);
            return INVALID_OPERATION;
        }
        outputDesc->addClient(sourceDesc);

        uint32_t delayMs = 0;
        status = startSource(outputDesc, sourceDesc, &delayMs);

        if (status != NO_ERROR) {
            mpClientInterface->releaseAudioPatch(sourceDesc->patchDesc()->mAfPatchHandle, 0);
            return status;
        }
        sourceDesc->setSwOutput(outputDesc);
        if (delayMs != 0) {
            usleep(delayMs * 1000);
        }
    }

    sourceDesc->patchDesc()->mAfPatchHandle = afPatchHandle;
    addAudioPatch(sourceDesc->patchDesc()->mHandle, sourceDesc->patchDesc());

    return NO_ERROR;
}

status_t AudioPolicyManager::stopAudioSource(audio_port_handle_t portId)
{
    sp<SourceClientDescriptor> sourceDesc = mAudioSources.valueFor(portId);
    ALOGV("%s port ID %d", __FUNCTION__, portId);
    if (sourceDesc == 0) {
        ALOGW("%s unknown source for port ID %d", __FUNCTION__, portId);
        return BAD_VALUE;
    }
    status_t status = disconnectAudioSource(sourceDesc);

    mAudioSources.removeItem(portId);
    return status;
}

status_t AudioPolicyManager::setMasterMono(bool mono)
{
    if (mMasterMono == mono) {
        return NO_ERROR;
    }
    mMasterMono = mono;
    // if enabling mono we close all offloaded devices, which will invalidate the
    // corresponding AudioTrack. The AudioTrack client/MediaPlayer is responsible
    // for recreating the new AudioTrack as non-offloaded PCM.
    //
    // If disabling mono, we leave all tracks as is: we don't know which clients
    // and tracks are able to be recreated as offloaded. The next "song" should
    // play back offloaded.
    if (mMasterMono) {
        Vector<audio_io_handle_t> offloaded;
        for (size_t i = 0; i < mOutputs.size(); ++i) {
            sp<SwAudioOutputDescriptor> desc = mOutputs.valueAt(i);
            if (desc->mFlags & AUDIO_OUTPUT_FLAG_COMPRESS_OFFLOAD) {
                offloaded.push(desc->mIoHandle);
            }
        }
        for (const auto& handle : offloaded) {
            closeOutput(handle);
        }
    }
    // update master mono for all remaining outputs
    for (size_t i = 0; i < mOutputs.size(); ++i) {
        updateMono(mOutputs.keyAt(i));
    }
    return NO_ERROR;
}

status_t AudioPolicyManager::getMasterMono(bool *mono)
{
    *mono = mMasterMono;
    return NO_ERROR;
}

float AudioPolicyManager::getStreamVolumeDB(
        audio_stream_type_t stream, int index, audio_devices_t device)
{
    return computeVolume(getVolumeCurves(stream), toVolumeSource(stream), index, device);
}

status_t AudioPolicyManager::getSurroundFormats(unsigned int *numSurroundFormats,
                                                audio_format_t *surroundFormats,
                                                bool *surroundFormatsEnabled,
                                                bool reported)
{
    if (numSurroundFormats == NULL || (*numSurroundFormats != 0 &&
            (surroundFormats == NULL || surroundFormatsEnabled == NULL))) {
        return BAD_VALUE;
    }
    ALOGV("%s() numSurroundFormats %d surroundFormats %p surroundFormatsEnabled %p reported %d",
            __func__, *numSurroundFormats, surroundFormats, surroundFormatsEnabled, reported);

    size_t formatsWritten = 0;
    size_t formatsMax = *numSurroundFormats;
    std::unordered_set<audio_format_t> formats; // Uses primary surround formats only
    if (reported) {
        // Return formats from all device profiles that have already been resolved by
        // checkOutputsForDevice().
        for (size_t i = 0; i < mAvailableOutputDevices.size(); i++) {
            sp<DeviceDescriptor> device = mAvailableOutputDevices[i];
            FormatVector supportedFormats =
                    device->getAudioPort()->getAudioProfiles().getSupportedFormats();
            for (size_t j = 0; j < supportedFormats.size(); j++) {
                if (mConfig.getSurroundFormats().count(supportedFormats[j]) != 0) {
                    formats.insert(supportedFormats[j]);
                } else {
                    for (const auto& pair : mConfig.getSurroundFormats()) {
                        if (pair.second.count(supportedFormats[j]) != 0) {
                            formats.insert(pair.first);
                            break;
                        }
                    }
                }
            }
        }
    } else {
        for (const auto& pair : mConfig.getSurroundFormats()) {
            formats.insert(pair.first);
        }
    }
    *numSurroundFormats = formats.size();
    audio_policy_forced_cfg_t forceUse = mEngine->getForceUse(
            AUDIO_POLICY_FORCE_FOR_ENCODED_SURROUND);
    for (const auto& format: formats) {
        if (formatsWritten < formatsMax) {
            surroundFormats[formatsWritten] = format;
            bool formatEnabled = true;
            switch (forceUse) {
                case AUDIO_POLICY_FORCE_ENCODED_SURROUND_MANUAL:
                    formatEnabled = mManualSurroundFormats.count(format) != 0;
                    break;
                case AUDIO_POLICY_FORCE_ENCODED_SURROUND_NEVER:
                    formatEnabled = false;
                    break;
                default: // AUTO or ALWAYS => true
                    break;
            }
            surroundFormatsEnabled[formatsWritten++] = formatEnabled;
        }
    }
    return NO_ERROR;
}

status_t AudioPolicyManager::setSurroundFormatEnabled(audio_format_t audioFormat, bool enabled)
{
    ALOGV("%s() format 0x%X enabled %d", __func__, audioFormat, enabled);
    const auto& formatIter = mConfig.getSurroundFormats().find(audioFormat);
    if (formatIter == mConfig.getSurroundFormats().end()) {
        ALOGW("%s() format 0x%X is not a known surround format", __func__, audioFormat);
        return BAD_VALUE;
    }

    if (mEngine->getForceUse(AUDIO_POLICY_FORCE_FOR_ENCODED_SURROUND) !=
            AUDIO_POLICY_FORCE_ENCODED_SURROUND_MANUAL) {
        ALOGW("%s() not in manual mode for surround sound format selection", __func__);
        return INVALID_OPERATION;
    }

    if ((mManualSurroundFormats.count(audioFormat) != 0) == enabled) {
        return NO_ERROR;
    }

    std::unordered_set<audio_format_t> surroundFormatsBackup(mManualSurroundFormats);
    if (enabled) {
        mManualSurroundFormats.insert(audioFormat);
        for (const auto& subFormat : formatIter->second) {
            mManualSurroundFormats.insert(subFormat);
        }
    } else {
        mManualSurroundFormats.erase(audioFormat);
        for (const auto& subFormat : formatIter->second) {
            mManualSurroundFormats.erase(subFormat);
        }
    }

    sp<SwAudioOutputDescriptor> outputDesc;
    bool profileUpdated = false;
    DeviceVector hdmiOutputDevices = mAvailableOutputDevices.getDevicesFromTypeMask(
            AUDIO_DEVICE_OUT_HDMI);
    for (size_t i = 0; i < hdmiOutputDevices.size(); i++) {
        // Simulate reconnection to update enabled surround sound formats.
        String8 address = hdmiOutputDevices[i]->address();
        String8 name = hdmiOutputDevices[i]->getName();
        status_t status = setDeviceConnectionStateInt(AUDIO_DEVICE_OUT_HDMI,
                                                      AUDIO_POLICY_DEVICE_STATE_UNAVAILABLE,
                                                      address.c_str(),
                                                      name.c_str(),
                                                      AUDIO_FORMAT_DEFAULT);
        if (status != NO_ERROR) {
            continue;
        }
        status = setDeviceConnectionStateInt(AUDIO_DEVICE_OUT_HDMI,
                                             AUDIO_POLICY_DEVICE_STATE_AVAILABLE,
                                             address.c_str(),
                                             name.c_str(),
                                             AUDIO_FORMAT_DEFAULT);
        profileUpdated |= (status == NO_ERROR);
    }
    // FIXME: Why doing this for input HDMI devices if we don't augment their reported formats?
    DeviceVector hdmiInputDevices = mAvailableInputDevices.getDevicesFromTypeMask(
                AUDIO_DEVICE_IN_HDMI);
    for (size_t i = 0; i < hdmiInputDevices.size(); i++) {
        // Simulate reconnection to update enabled surround sound formats.
        String8 address = hdmiInputDevices[i]->address();
        String8 name = hdmiInputDevices[i]->getName();
        status_t status = setDeviceConnectionStateInt(AUDIO_DEVICE_IN_HDMI,
                                                      AUDIO_POLICY_DEVICE_STATE_UNAVAILABLE,
                                                      address.c_str(),
                                                      name.c_str(),
                                                      AUDIO_FORMAT_DEFAULT);
        if (status != NO_ERROR) {
            continue;
        }
        status = setDeviceConnectionStateInt(AUDIO_DEVICE_IN_HDMI,
                                             AUDIO_POLICY_DEVICE_STATE_AVAILABLE,
                                             address.c_str(),
                                             name.c_str(),
                                             AUDIO_FORMAT_DEFAULT);
        profileUpdated |= (status == NO_ERROR);
    }

    if (!profileUpdated) {
        ALOGW("%s() no audio profiles updated, undoing surround formats change", __func__);
        mManualSurroundFormats = std::move(surroundFormatsBackup);
    }

    return profileUpdated ? NO_ERROR : INVALID_OPERATION;
}

void AudioPolicyManager::setAppState(uid_t uid, app_state_t state)
{
    ALOGV("%s(uid:%d, state:%d)", __func__, uid, state);
    for (size_t i = 0; i < mInputs.size(); i++) {
        mInputs.valueAt(i)->setAppState(uid, state);
    }
}

bool AudioPolicyManager::isHapticPlaybackSupported()
{
    for (const auto& hwModule : mHwModules) {
        const OutputProfileCollection &outputProfiles = hwModule->getOutputProfiles();
        for (const auto &outProfile : outputProfiles) {
            struct audio_port audioPort;
            outProfile->toAudioPort(&audioPort);
            for (size_t i = 0; i < audioPort.num_channel_masks; i++) {
                if (audioPort.channel_masks[i] & AUDIO_CHANNEL_HAPTIC_ALL) {
                    return true;
                }
            }
        }
    }
    return false;
}

status_t AudioPolicyManager::disconnectAudioSource(const sp<SourceClientDescriptor>& sourceDesc)
{
    ALOGV("%s port Id %d", __FUNCTION__, sourceDesc->portId());

    sp<AudioPatch> patchDesc = mAudioPatches.valueFor(sourceDesc->patchDesc()->mHandle);
    if (patchDesc == 0) {
        ALOGW("%s source has no patch with handle %d", __FUNCTION__,
              sourceDesc->patchDesc()->mHandle);
        return BAD_VALUE;
    }
    removeAudioPatch(sourceDesc->patchDesc()->mHandle);

    sp<SwAudioOutputDescriptor> swOutputDesc = sourceDesc->swOutput().promote();
    if (swOutputDesc != 0) {
        status_t status = stopSource(swOutputDesc, sourceDesc);
        if (status == NO_ERROR) {
            swOutputDesc->stop();
        }
        mpClientInterface->releaseAudioPatch(patchDesc->mAfPatchHandle, 0);
    } else {
        sp<HwAudioOutputDescriptor> hwOutputDesc = sourceDesc->hwOutput().promote();
        if (hwOutputDesc != 0) {
          //   release patch between src device and output device
          //   close Hwoutput and remove from mHwOutputs
        } else {
            ALOGW("%s source has neither SW nor HW output", __FUNCTION__);
        }
    }
    return NO_ERROR;
}

sp<SourceClientDescriptor> AudioPolicyManager::getSourceForAttributesOnOutput(
        audio_io_handle_t output, const audio_attributes_t &attr)
{
    sp<SourceClientDescriptor> source;
    for (size_t i = 0; i < mAudioSources.size(); i++)  {
        sp<SourceClientDescriptor> sourceDesc = mAudioSources.valueAt(i);
        sp<SwAudioOutputDescriptor> outputDesc = sourceDesc->swOutput().promote();
        if (followsSameRouting(attr, sourceDesc->attributes()) &&
                               outputDesc != 0 && outputDesc->mIoHandle == output) {
            source = sourceDesc;
            break;
        }
    }
    return source;
}

// ----------------------------------------------------------------------------
// AudioPolicyManager
// ----------------------------------------------------------------------------
uint32_t AudioPolicyManager::nextAudioPortGeneration()
{
    return mAudioPortGeneration++;
}

// Treblized audio policy xml config will be located in /odm/etc or /vendor/etc.
static const char *kConfigLocationList[] =
        {"/odm/etc", "/vendor/etc", "/system/etc"};
static const int kConfigLocationListSize =
        (sizeof(kConfigLocationList) / sizeof(kConfigLocationList[0]));

static status_t deserializeAudioPolicyXmlConfig(AudioPolicyConfig &config) {
    char audioPolicyXmlConfigFile[AUDIO_POLICY_XML_CONFIG_FILE_PATH_MAX_LENGTH];
    std::vector<const char*> fileNames;
    status_t ret;

    if (property_get_bool("ro.bluetooth.a2dp_offload.supported", false)) {
        if (property_get_bool("persist.bluetooth.bluetooth_audio_hal.disabled", false) &&
            property_get_bool("persist.bluetooth.a2dp_offload.disabled", false)) {
            // Both BluetoothAudio@2.0 and BluetoothA2dp@1.0 (Offlaod) are disabled, and uses
            // the legacy hardware module for A2DP and hearing aid.
            fileNames.push_back(AUDIO_POLICY_BLUETOOTH_LEGACY_HAL_XML_CONFIG_FILE_NAME);
        } else if (property_get_bool("persist.bluetooth.a2dp_offload.disabled", false)) {
            // A2DP offload supported but disabled: try to use special XML file
            fileNames.push_back(AUDIO_POLICY_A2DP_OFFLOAD_DISABLED_XML_CONFIG_FILE_NAME);
        }
    } else if (property_get_bool("persist.bluetooth.bluetooth_audio_hal.disabled", false)) {
        fileNames.push_back(AUDIO_POLICY_BLUETOOTH_LEGACY_HAL_XML_CONFIG_FILE_NAME);
    }
    fileNames.push_back(AUDIO_POLICY_XML_CONFIG_FILE_NAME);

    for (const char* fileName : fileNames) {
        for (int i = 0; i < kConfigLocationListSize; i++) {
            snprintf(audioPolicyXmlConfigFile, sizeof(audioPolicyXmlConfigFile),
                     "%s/%s", kConfigLocationList[i], fileName);
            ret = deserializeAudioPolicyFile(audioPolicyXmlConfigFile, &config);
            if (ret == NO_ERROR) {
                config.setSource(audioPolicyXmlConfigFile);
                return ret;
            }
        }
    }
    return ret;
}

AudioPolicyManager::AudioPolicyManager(AudioPolicyClientInterface *clientInterface,
                                       bool /*forTesting*/)
    :
    mUidCached(AID_AUDIOSERVER), // no need to call getuid(), there's only one of us running.
    mpClientInterface(clientInterface),
    mLimitRingtoneVolume(false), mLastVoiceVolume(-1.0f),
    mA2dpSuspended(false),
    mConfig(mHwModulesAll, mAvailableOutputDevices, mAvailableInputDevices, mDefaultOutputDevice),
    mAudioPortGeneration(1),
    mBeaconMuteRefCount(0),
    mBeaconPlayingRefCount(0),
    mBeaconMuted(false),
    mTtsOutputAvailable(false),
    mMasterMono(false),
    mMusicEffectOutput(AUDIO_IO_HANDLE_NONE)
{
}

AudioPolicyManager::AudioPolicyManager(AudioPolicyClientInterface *clientInterface)
        : AudioPolicyManager(clientInterface, false /*forTesting*/)
{
    loadConfig();
    initialize();
}

void AudioPolicyManager::loadConfig() {
    if (deserializeAudioPolicyXmlConfig(getConfig()) != NO_ERROR) {
        ALOGE("could not load audio policy configuration file, setting defaults");
        getConfig().setDefault();
    }
}

status_t AudioPolicyManager::initialize() {
    {
        auto engLib = EngineLibrary::load(
                        "libaudiopolicyengine" + getConfig().getEngineLibraryNameSuffix() + ".so");
        if (!engLib) {
            ALOGE("%s: Failed to load the engine library", __FUNCTION__);
            return NO_INIT;
        }
        mEngine = engLib->createEngine();
        if (mEngine == nullptr) {
            ALOGE("%s: Failed to instantiate the APM engine", __FUNCTION__);
            return NO_INIT;
        }
    }
    mEngine->setObserver(this);
    status_t status = mEngine->initCheck();
    if (status != NO_ERROR) {
        LOG_FATAL("Policy engine not initialized(err=%d)", status);
        return status;
    }

    // mAvailableOutputDevices and mAvailableInputDevices now contain all attached devices
    // open all output streams needed to access attached devices
    for (const auto& hwModule : mHwModulesAll) {
        hwModule->setHandle(mpClientInterface->loadHwModule(hwModule->getName()));
        if (hwModule->getHandle() == AUDIO_MODULE_HANDLE_NONE) {
            ALOGW("could not open HW module %s", hwModule->getName());
            continue;
        }
        mHwModules.push_back(hwModule);
        // open all output streams needed to access attached devices
        // except for direct output streams that are only opened when they are actually
        // required by an app.
        // This also validates mAvailableOutputDevices list
        for (const auto& outProfile : hwModule->getOutputProfiles()) {
            if (!outProfile->canOpenNewIo()) {
                ALOGE("Invalid Output profile max open count %u for profile %s",
                      outProfile->maxOpenCount, outProfile->getTagName().c_str());
                continue;
            }
            if (!outProfile->hasSupportedDevices()) {
                ALOGW("Output profile contains no device on module %s", hwModule->getName());
                continue;
            }
            if ((outProfile->getFlags() & AUDIO_OUTPUT_FLAG_TTS) != 0) {
                mTtsOutputAvailable = true;
            }

            if ((outProfile->getFlags() & AUDIO_OUTPUT_FLAG_DIRECT) != 0) {
                continue;
            }
            const DeviceVector &supportedDevices = outProfile->getSupportedDevices();
            DeviceVector availProfileDevices = supportedDevices.filter(mAvailableOutputDevices);
            sp<DeviceDescriptor> supportedDevice = 0;
            if (supportedDevices.contains(mDefaultOutputDevice)) {
                supportedDevice = mDefaultOutputDevice;
            } else {
                // choose first device present in profile's SupportedDevices also part of
                // mAvailableOutputDevices.
                if (availProfileDevices.isEmpty()) {
                    continue;
                }
                supportedDevice = availProfileDevices.itemAt(0);
            }
            if (!mAvailableOutputDevices.contains(supportedDevice)) {
                continue;
            }
            sp<SwAudioOutputDescriptor> outputDesc = new SwAudioOutputDescriptor(outProfile,
                                                                                 mpClientInterface);
            audio_io_handle_t output = AUDIO_IO_HANDLE_NONE;
            status_t status = outputDesc->open(nullptr, DeviceVector(supportedDevice),
                                               AUDIO_STREAM_DEFAULT,
                                               AUDIO_OUTPUT_FLAG_NONE, &output);
            if (status != NO_ERROR) {
                ALOGW("Cannot open output stream for devices %s on hw module %s",
                      supportedDevice->toString().c_str(), hwModule->getName());
                continue;
            }
            for (const auto &device : availProfileDevices) {
                // give a valid ID to an attached device once confirmed it is reachable
                if (!device->isAttached()) {
                    device->attach(hwModule);
                }
            }
            if (mPrimaryOutput == 0 &&
                    outProfile->getFlags() & AUDIO_OUTPUT_FLAG_PRIMARY) {
                mPrimaryOutput = outputDesc;
            }
            addOutput(output, outputDesc);
            setOutputDevices(outputDesc,
                             DeviceVector(supportedDevice),
                             true,
                             0,
                             NULL);
        }
        // open input streams needed to access attached devices to validate
        // mAvailableInputDevices list
        for (const auto& inProfile : hwModule->getInputProfiles()) {
            if (!inProfile->canOpenNewIo()) {
                ALOGE("Invalid Input profile max open count %u for profile %s",
                      inProfile->maxOpenCount, inProfile->getTagName().c_str());
                continue;
            }
            if (!inProfile->hasSupportedDevices()) {
                ALOGW("Input profile contains no device on module %s", hwModule->getName());
                continue;
            }
            // chose first device present in profile's SupportedDevices also part of
            // available input devices
            const DeviceVector &supportedDevices = inProfile->getSupportedDevices();
            DeviceVector availProfileDevices = supportedDevices.filter(mAvailableInputDevices);
            if (availProfileDevices.isEmpty()) {
                ALOGE("%s: Input device list is empty!", __FUNCTION__);
                continue;
            }
            sp<AudioInputDescriptor> inputDesc =
                    new AudioInputDescriptor(inProfile, mpClientInterface);

            audio_io_handle_t input = AUDIO_IO_HANDLE_NONE;
            status_t status = inputDesc->open(nullptr,
                                              availProfileDevices.itemAt(0),
                                              AUDIO_SOURCE_MIC,
                                              AUDIO_INPUT_FLAG_NONE,
                                              &input);
            if (status != NO_ERROR) {
                ALOGW("Cannot open input stream for device %s on hw module %s",
                      availProfileDevices.toString().c_str(),
                      hwModule->getName());
                continue;
            }
            for (const auto &device : availProfileDevices) {
                // give a valid ID to an attached device once confirmed it is reachable
                if (!device->isAttached()) {
                    device->attach(hwModule);
                    device->importAudioPort(inProfile, true);
                }
            }
            inputDesc->close();
        }
    }
    // make sure all attached devices have been allocated a unique ID
    auto checkAndSetAvailable = [this](auto& devices) {
        for (size_t i = 0; i < devices.size();) {
            const auto &device = devices[i];
            if (!device->isAttached()) {
                ALOGW("device %s is unreachable", device->toString().c_str());
                devices.remove(device);
                continue;
            }
            // Device is now validated and can be appended to the available devices of the engine
            setEngineDeviceConnectionState(device, AUDIO_POLICY_DEVICE_STATE_AVAILABLE);
            i++;
        }
    };
    checkAndSetAvailable(mAvailableOutputDevices);
    checkAndSetAvailable(mAvailableInputDevices);

    // make sure default device is reachable
    if (mDefaultOutputDevice == 0 || !mAvailableOutputDevices.contains(mDefaultOutputDevice)) {
        ALOGE_IF(mDefaultOutputDevice != 0, "Default device %s is unreachable",
                 mDefaultOutputDevice->toString().c_str());
        status = NO_INIT;
    }
    // If microphones address is empty, set it according to device type
    for (size_t i = 0; i < mAvailableInputDevices.size(); i++) {
        if (mAvailableInputDevices[i]->address().isEmpty()) {
            if (mAvailableInputDevices[i]->type() == AUDIO_DEVICE_IN_BUILTIN_MIC) {
                mAvailableInputDevices[i]->setAddress(String8(AUDIO_BOTTOM_MICROPHONE_ADDRESS));
            } else if (mAvailableInputDevices[i]->type() == AUDIO_DEVICE_IN_BACK_MIC) {
                mAvailableInputDevices[i]->setAddress(String8(AUDIO_BACK_MICROPHONE_ADDRESS));
            }
        }
    }

    if (mPrimaryOutput == 0) {
        ALOGE("Failed to open primary output");
        status = NO_INIT;
    }

    // Silence ALOGV statements
    property_set("log.tag." LOG_TAG, "D");

    updateDevicesAndOutputs();
    return status;
}

AudioPolicyManager::~AudioPolicyManager()
{
   for (size_t i = 0; i < mOutputs.size(); i++) {
        mOutputs.valueAt(i)->close();
   }
   for (size_t i = 0; i < mInputs.size(); i++) {
        mInputs.valueAt(i)->close();
   }
   mAvailableOutputDevices.clear();
   mAvailableInputDevices.clear();
   mOutputs.clear();
   mInputs.clear();
   mHwModules.clear();
   mHwModulesAll.clear();
   mManualSurroundFormats.clear();
}

status_t AudioPolicyManager::initCheck()
{
    return hasPrimaryOutput() ? NO_ERROR : NO_INIT;
}

// ---

void AudioPolicyManager::addOutput(audio_io_handle_t output,
                                   const sp<SwAudioOutputDescriptor>& outputDesc)
{
    mOutputs.add(output, outputDesc);
    applyStreamVolumes(outputDesc, AUDIO_DEVICE_NONE, 0 /* delayMs */, true /* force */);
    updateMono(output); // update mono status when adding to output list
    selectOutputForMusicEffects();
    nextAudioPortGeneration();
}

void AudioPolicyManager::removeOutput(audio_io_handle_t output)
{
    mOutputs.removeItem(output);
    selectOutputForMusicEffects();
}

void AudioPolicyManager::addInput(audio_io_handle_t input,
                                  const sp<AudioInputDescriptor>& inputDesc)
{
    mInputs.add(input, inputDesc);
    nextAudioPortGeneration();
}

status_t AudioPolicyManager::checkOutputsForDevice(const sp<DeviceDescriptor>& device,
                                                   audio_policy_dev_state_t state,
                                                   SortedVector<audio_io_handle_t>& outputs)
{
    audio_devices_t deviceType = device->type();
    const String8 &address = device->address();
    sp<SwAudioOutputDescriptor> desc;

    if (audio_device_is_digital(deviceType)) {
        // erase all current sample rates, formats and channel masks
        device->clearAudioProfiles();
    }

    if (state == AUDIO_POLICY_DEVICE_STATE_AVAILABLE) {
        // first list already open outputs that can be routed to this device
        for (size_t i = 0; i < mOutputs.size(); i++) {
            desc = mOutputs.valueAt(i);
            if (!desc->isDuplicated() && desc->supportsDevice(device)
                    && desc->deviceSupportsEncodedFormats(deviceType)) {
                ALOGV("checkOutputsForDevice(): adding opened output %d on device %s",
                      mOutputs.keyAt(i), device->toString().c_str());
                outputs.add(mOutputs.keyAt(i));
            }
        }
        // then look for output profiles that can be routed to this device
        SortedVector< sp<IOProfile> > profiles;
        for (const auto& hwModule : mHwModules) {
            for (size_t j = 0; j < hwModule->getOutputProfiles().size(); j++) {
                sp<IOProfile> profile = hwModule->getOutputProfiles()[j];
                if (profile->supportsDevice(device)) {
                    profiles.add(profile);
                    ALOGV("checkOutputsForDevice(): adding profile %zu from module %s",
                          j, hwModule->getName());
                }
            }
        }

        ALOGV("  found %zu profiles, %zu outputs", profiles.size(), outputs.size());

        if (profiles.isEmpty() && outputs.isEmpty()) {
            ALOGW("checkOutputsForDevice(): No output available for device %04x", deviceType);
            return BAD_VALUE;
        }

        // open outputs for matching profiles if needed. Direct outputs are also opened to
        // query for dynamic parameters and will be closed later by setDeviceConnectionState()
        for (ssize_t profile_index = 0; profile_index < (ssize_t)profiles.size(); profile_index++) {
            sp<IOProfile> profile = profiles[profile_index];

            // nothing to do if one output is already opened for this profile
            size_t j;
            for (j = 0; j < outputs.size(); j++) {
                desc = mOutputs.valueFor(outputs.itemAt(j));
                if (!desc->isDuplicated() && desc->mProfile == profile) {
                    // matching profile: save the sample rates, format and channel masks supported
                    // by the profile in our device descriptor
                    if (audio_device_is_digital(deviceType)) {
                        device->importAudioPort(profile);
                    }
                    break;
                }
            }
            if (j != outputs.size()) {
                continue;
            }

            if (!profile->canOpenNewIo()) {
                ALOGW("Max Output number %u already opened for this profile %s",
                      profile->maxOpenCount, profile->getTagName().c_str());
                continue;
            }

            ALOGV("opening output for device %08x with params %s profile %p name %s",
                  deviceType, address.string(), profile.get(), profile->getName().string());
            desc = new SwAudioOutputDescriptor(profile, mpClientInterface);
            audio_io_handle_t output = AUDIO_IO_HANDLE_NONE;
            status_t status = desc->open(nullptr, DeviceVector(device),
                                         AUDIO_STREAM_DEFAULT, AUDIO_OUTPUT_FLAG_NONE, &output);

            if (status == NO_ERROR) {
                // Here is where the out_set_parameters() for card & device gets called
                if (!address.isEmpty()) {
                    char *param = audio_device_address_to_parameter(deviceType, address);
                    mpClientInterface->setParameters(output, String8(param));
                    free(param);
                }
                updateAudioProfiles(device, output, profile->getAudioProfiles());
                if (!profile->hasValidAudioProfile()) {
                    ALOGW("checkOutputsForDevice() missing param");
                    desc->close();
                    output = AUDIO_IO_HANDLE_NONE;
                } else if (profile->hasDynamicAudioProfile()) {
                    desc->close();
                    output = AUDIO_IO_HANDLE_NONE;
                    audio_config_t config = AUDIO_CONFIG_INITIALIZER;
                    profile->pickAudioProfile(
                            config.sample_rate, config.channel_mask, config.format);
                    config.offload_info.sample_rate = config.sample_rate;
                    config.offload_info.channel_mask = config.channel_mask;
                    config.offload_info.format = config.format;

                    status_t status = desc->open(&config, DeviceVector(device),
                                                 AUDIO_STREAM_DEFAULT,
                                                 AUDIO_OUTPUT_FLAG_NONE, &output);
                    if (status != NO_ERROR) {
                        output = AUDIO_IO_HANDLE_NONE;
                    }
                }

                if (output != AUDIO_IO_HANDLE_NONE) {
                    addOutput(output, desc);
                    if (device_distinguishes_on_address(deviceType) && address != "0") {
                        sp<AudioPolicyMix> policyMix;
                        if (mPolicyMixes.getAudioPolicyMix(address, policyMix) == NO_ERROR) {
                            policyMix->setOutput(desc);
                            desc->mPolicyMix = policyMix;
                        } else {
                            ALOGW("checkOutputsForDevice() cannot find policy for address %s",
                                  address.string());
                        }

                    } else if (((desc->mFlags & AUDIO_OUTPUT_FLAG_DIRECT) == 0) &&
                                    hasPrimaryOutput()) {
                        // no duplicated output for direct outputs and
                        // outputs used by dynamic policy mixes
                        audio_io_handle_t duplicatedOutput = AUDIO_IO_HANDLE_NONE;

                        //TODO: configure audio effect output stage here

                        // open a duplicating output thread for the new output and the primary output
                        sp<SwAudioOutputDescriptor> dupOutputDesc =
                                new SwAudioOutputDescriptor(NULL, mpClientInterface);
                        status_t status = dupOutputDesc->openDuplicating(mPrimaryOutput, desc,
                                                                         &duplicatedOutput);
                        if (status == NO_ERROR) {
                            // add duplicated output descriptor
                            addOutput(duplicatedOutput, dupOutputDesc);
                        } else {
                            ALOGW("checkOutputsForDevice() could not open dup output for %d and %d",
                                    mPrimaryOutput->mIoHandle, output);
                            desc->close();
                            removeOutput(output);
                            nextAudioPortGeneration();
                            output = AUDIO_IO_HANDLE_NONE;
                        }
                    }
                }
            } else {
                output = AUDIO_IO_HANDLE_NONE;
            }
            if (output == AUDIO_IO_HANDLE_NONE) {
                ALOGW("checkOutputsForDevice() could not open output for device %x", deviceType);
                profiles.removeAt(profile_index);
                profile_index--;
            } else {
                outputs.add(output);
                // Load digital format info only for digital devices
                if (audio_device_is_digital(deviceType)) {
                    device->importAudioPort(profile);
                }

                if (device_distinguishes_on_address(deviceType)) {
                    ALOGV("checkOutputsForDevice(): setOutputDevices %s",
                            device->toString().c_str());
                    setOutputDevices(desc, DeviceVector(device), true/*force*/, 0/*delay*/,
                                     NULL/*patch handle*/);
                }
                ALOGV("checkOutputsForDevice(): adding output %d", output);
            }
        }

        if (profiles.isEmpty()) {
            ALOGW("checkOutputsForDevice(): No output available for device %04x", deviceType);
            return BAD_VALUE;
        }
    } else { // Disconnect
        // check if one opened output is not needed any more after disconnecting one device
        for (size_t i = 0; i < mOutputs.size(); i++) {
            desc = mOutputs.valueAt(i);
            if (!desc->isDuplicated()) {
                // exact match on device
                if (device_distinguishes_on_address(deviceType) && desc->supportsDevice(device)
                        && desc->deviceSupportsEncodedFormats(deviceType)) {
                    outputs.add(mOutputs.keyAt(i));
                } else if (!mAvailableOutputDevices.containsAtLeastOne(desc->supportedDevices())) {
                    ALOGV("checkOutputsForDevice(): disconnecting adding output %d",
                            mOutputs.keyAt(i));
                    outputs.add(mOutputs.keyAt(i));
                }
            }
        }
        // Clear any profiles associated with the disconnected device.
        for (const auto& hwModule : mHwModules) {
            for (size_t j = 0; j < hwModule->getOutputProfiles().size(); j++) {
                sp<IOProfile> profile = hwModule->getOutputProfiles()[j];
                if (profile->supportsDevice(device)) {
                    ALOGV("checkOutputsForDevice(): "
                            "clearing direct output profile %zu on module %s",
                            j, hwModule->getName());
                    profile->clearAudioProfiles();
                }
            }
        }
    }
    return NO_ERROR;
}

status_t AudioPolicyManager::checkInputsForDevice(const sp<DeviceDescriptor>& device,
                                                  audio_policy_dev_state_t state)
{
    sp<AudioInputDescriptor> desc;

    if (audio_device_is_digital(device->type())) {
        // erase all current sample rates, formats and channel masks
        device->clearAudioProfiles();
    }

    if (state == AUDIO_POLICY_DEVICE_STATE_AVAILABLE) {
        // look for input profiles that can be routed to this device
        SortedVector< sp<IOProfile> > profiles;
        for (const auto& hwModule : mHwModules) {
            for (size_t profile_index = 0;
                 profile_index < hwModule->getInputProfiles().size();
                 profile_index++) {
                sp<IOProfile> profile = hwModule->getInputProfiles()[profile_index];

                if (profile->supportsDevice(device)) {
                    profiles.add(profile);
                    ALOGV("checkInputsForDevice(): adding profile %zu from module %s",
                          profile_index, hwModule->getName());
                }
            }
        }

        if (profiles.isEmpty()) {
            ALOGW("%s: No input profile available for device %s",
                __func__, device->toString().c_str());
            return BAD_VALUE;
        }

        // open inputs for matching profiles if needed. Direct inputs are also opened to
        // query for dynamic parameters and will be closed later by setDeviceConnectionState()
        for (ssize_t profile_index = 0; profile_index < (ssize_t)profiles.size(); profile_index++) {

            sp<IOProfile> profile = profiles[profile_index];

            // nothing to do if one input is already opened for this profile
            size_t input_index;
            for (input_index = 0; input_index < mInputs.size(); input_index++) {
                desc = mInputs.valueAt(input_index);
                if (desc->mProfile == profile) {
                    if (audio_device_is_digital(device->type())) {
                        device->importAudioPort(profile);
                    }
                    break;
                }
            }
            if (input_index != mInputs.size()) {
                continue;
            }

            if (!profile->canOpenNewIo()) {
                ALOGW("Max Input number %u already opened for this profile %s",
                      profile->maxOpenCount, profile->getTagName().c_str());
                continue;
            }

            desc = new AudioInputDescriptor(profile, mpClientInterface);
            audio_io_handle_t input = AUDIO_IO_HANDLE_NONE;
            status_t status = desc->open(nullptr,
                                         device,
                                         AUDIO_SOURCE_MIC,
                                         AUDIO_INPUT_FLAG_NONE,
                                         &input);

            if (status == NO_ERROR) {
                const String8& address = device->address();
                if (!address.isEmpty()) {
                    char *param = audio_device_address_to_parameter(device->type(), address);
                    mpClientInterface->setParameters(input, String8(param));
                    free(param);
                }
                updateAudioProfiles(device, input, profile->getAudioProfiles());
                if (!profile->hasValidAudioProfile()) {
                    ALOGW("checkInputsForDevice() direct input missing param");
                    desc->close();
                    input = AUDIO_IO_HANDLE_NONE;
                }

                if (input != AUDIO_IO_HANDLE_NONE) {
                    addInput(input, desc);
                }
            } // endif input != 0

            if (input == AUDIO_IO_HANDLE_NONE) {
                ALOGW("%s could not open input for device %s", __func__,  
                       device->toString().c_str());
                profiles.removeAt(profile_index);
                profile_index--;
            } else {
                if (audio_device_is_digital(device->type())) {
                    device->importAudioPort(profile);
                }
                ALOGV("checkInputsForDevice(): adding input %d", input);
            }
        } // end scan profiles

        if (profiles.isEmpty()) {
            ALOGW("%s: No input available for device %s", __func__,  device->toString().c_str());
            return BAD_VALUE;
        }
    } else {
        // Disconnect
        // Clear any profiles associated with the disconnected device.
        for (const auto& hwModule : mHwModules) {
            for (size_t profile_index = 0;
                 profile_index < hwModule->getInputProfiles().size();
                 profile_index++) {
                sp<IOProfile> profile = hwModule->getInputProfiles()[profile_index];
                if (profile->supportsDevice(device)) {
                    ALOGV("checkInputsForDevice(): clearing direct input profile %zu on module %s",
                            profile_index, hwModule->getName());
                    profile->clearAudioProfiles();
                }
            }
        }
    } // end disconnect

    return NO_ERROR;
}


void AudioPolicyManager::closeOutput(audio_io_handle_t output)
{
    ALOGV("closeOutput(%d)", output);

    sp<SwAudioOutputDescriptor> closingOutput = mOutputs.valueFor(output);
    if (closingOutput == NULL) {
        ALOGW("closeOutput() unknown output %d", output);
        return;
    }
    const bool closingOutputWasActive = closingOutput->isActive();
    mPolicyMixes.closeOutput(closingOutput);

    // look for duplicated outputs connected to the output being removed.
    for (size_t i = 0; i < mOutputs.size(); i++) {
        sp<SwAudioOutputDescriptor> dupOutput = mOutputs.valueAt(i);
        if (dupOutput->isDuplicated() &&
                (dupOutput->mOutput1 == closingOutput || dupOutput->mOutput2 == closingOutput)) {
            sp<SwAudioOutputDescriptor> remainingOutput =
                dupOutput->mOutput1 == closingOutput ? dupOutput->mOutput2 : dupOutput->mOutput1;
            // As all active tracks on duplicated output will be deleted,
            // and as they were also referenced on the other output, the reference
            // count for their stream type must be adjusted accordingly on
            // the other output.
            const bool wasActive = remainingOutput->isActive();
            // Note: no-op on the closing output where all clients has already been set inactive
            dupOutput->setAllClientsInactive();
            // stop() will be a no op if the output is still active but is needed in case all
            // active streams refcounts where cleared above
            if (wasActive) {
                remainingOutput->stop();
            }
            audio_io_handle_t duplicatedOutput = mOutputs.keyAt(i);
            ALOGV("closeOutput() closing also duplicated output %d", duplicatedOutput);

            mpClientInterface->closeOutput(duplicatedOutput);
            removeOutput(duplicatedOutput);
        }
    }

    nextAudioPortGeneration();

    ssize_t index = mAudioPatches.indexOfKey(closingOutput->getPatchHandle());
    if (index >= 0) {
        sp<AudioPatch> patchDesc = mAudioPatches.valueAt(index);
        (void) /*status_t status*/ mpClientInterface->releaseAudioPatch(patchDesc->mAfPatchHandle, 0);
        mAudioPatches.removeItemsAt(index);
        mpClientInterface->onAudioPatchListUpdate();
    }

    if (closingOutputWasActive) {
        closingOutput->stop();
    }
    closingOutput->close();

    removeOutput(output);
    mPreviousOutputs = mOutputs;

    // MSD patches may have been released to support a non-MSD direct output. Reset MSD patch if
    // no direct outputs are open.
    if (!getMsdAudioOutDevices().isEmpty()) {
        bool directOutputOpen = false;
        for (size_t i = 0; i < mOutputs.size(); i++) {
            if (mOutputs[i]->mFlags & AUDIO_OUTPUT_FLAG_DIRECT) {
                directOutputOpen = true;
                break;
            }
        }
        if (!directOutputOpen) {
            ALOGV("no direct outputs open, reset MSD patch");
            setMsdPatch();
        }
    }

    cleanUpEffectsForIo(output);
}

void AudioPolicyManager::closeInput(audio_io_handle_t input)
{
    ALOGV("closeInput(%d)", input);

    sp<AudioInputDescriptor> inputDesc = mInputs.valueFor(input);
    if (inputDesc == NULL) {
        ALOGW("closeInput() unknown input %d", input);
        return;
    }

    nextAudioPortGeneration();

    sp<DeviceDescriptor> device = inputDesc->getDevice();
    ssize_t index = mAudioPatches.indexOfKey(inputDesc->getPatchHandle());
    if (index >= 0) {
        sp<AudioPatch> patchDesc = mAudioPatches.valueAt(index);
        (void) /*status_t status*/ mpClientInterface->releaseAudioPatch(patchDesc->mAfPatchHandle, 0);
        mAudioPatches.removeItemsAt(index);
        mpClientInterface->onAudioPatchListUpdate();
    }

    inputDesc->close();
    mInputs.removeItem(input);

    DeviceVector primaryInputDevices = availablePrimaryModuleInputDevices();
    if (primaryInputDevices.contains(device) &&
            mInputs.activeInputsCountOnDevices(primaryInputDevices) == 0) {
        SoundTrigger::setCaptureState(false);
    }

    cleanUpEffectsForIo(input);
}

SortedVector<audio_io_handle_t> AudioPolicyManager::getOutputsForDevices(
            const DeviceVector &devices,
            const SwAudioOutputCollection& openOutputs)
{
    SortedVector<audio_io_handle_t> outputs;

    ALOGVV("%s() devices %s", __func__, devices.toString().c_str());
    for (size_t i = 0; i < openOutputs.size(); i++) {
        ALOGVV("output %zu isDuplicated=%d device=%s",
                i, openOutputs.valueAt(i)->isDuplicated(),
                openOutputs.valueAt(i)->supportedDevices().toString().c_str());
        if (openOutputs.valueAt(i)->supportsAllDevices(devices)
                && openOutputs.valueAt(i)->deviceSupportsEncodedFormats(devices.types())) {
            ALOGVV("%s() found output %d", __func__, openOutputs.keyAt(i));
            outputs.add(openOutputs.keyAt(i));
        }
    }
    return outputs;
}

void AudioPolicyManager::checkForDeviceAndOutputChanges(std::function<bool()> onOutputsChecked)
{
    // checkA2dpSuspend must run before checkOutputForAllStrategies so that A2DP
    // output is suspended before any tracks are moved to it
    checkA2dpSuspend();
    checkOutputForAllStrategies();
    checkSecondaryOutputs();
    if (onOutputsChecked != nullptr && onOutputsChecked()) checkA2dpSuspend();
    updateDevicesAndOutputs();
    if (mHwModules.getModuleFromName(AUDIO_HARDWARE_MODULE_ID_MSD) != 0) {
        setMsdPatch();
    }
}

bool AudioPolicyManager::followsSameRouting(const audio_attributes_t &lAttr,
                                            const audio_attributes_t &rAttr) const
{
    return mEngine->getProductStrategyForAttributes(lAttr) ==
            mEngine->getProductStrategyForAttributes(rAttr);
}

void AudioPolicyManager::checkOutputForAttributes(const audio_attributes_t &attr)
{
    auto psId = mEngine->getProductStrategyForAttributes(attr);

    DeviceVector oldDevices = mEngine->getOutputDevicesForAttributes(attr, 0, true /*fromCache*/);
    DeviceVector newDevices = mEngine->getOutputDevicesForAttributes(attr, 0, false /*fromCache*/);
    SortedVector<audio_io_handle_t> srcOutputs = getOutputsForDevices(oldDevices, mPreviousOutputs);
    SortedVector<audio_io_handle_t> dstOutputs = getOutputsForDevices(newDevices, mOutputs);

    // also take into account external policy-related changes: add all outputs which are
    // associated with policies in the "before" and "after" output vectors
    ALOGVV("%s(): policy related outputs", __func__);
    for (size_t i = 0 ; i < mPreviousOutputs.size() ; i++) {
        const sp<SwAudioOutputDescriptor> desc = mPreviousOutputs.valueAt(i);
        if (desc != 0 && desc->mPolicyMix != NULL) {
            srcOutputs.add(desc->mIoHandle);
            ALOGVV(" previous outputs: adding %d", desc->mIoHandle);
        }
    }
    for (size_t i = 0 ; i < mOutputs.size() ; i++) {
        const sp<SwAudioOutputDescriptor> desc = mOutputs.valueAt(i);
        if (desc != 0 && desc->mPolicyMix != NULL) {
            dstOutputs.add(desc->mIoHandle);
            ALOGVV(" new outputs: adding %d", desc->mIoHandle);
        }
    }

    if (srcOutputs != dstOutputs) {
        // get maximum latency of all source outputs to determine the minimum mute time guaranteeing
        // audio from invalidated tracks will be rendered when unmuting
        uint32_t maxLatency = 0;
        for (audio_io_handle_t srcOut : srcOutputs) {
            sp<SwAudioOutputDescriptor> desc = mPreviousOutputs.valueFor(srcOut);
            if (desc != 0 && maxLatency < desc->latency()) {
                maxLatency = desc->latency();
            }
        }
        ALOGV_IF(!(srcOutputs.isEmpty() || dstOutputs.isEmpty()),
              "%s: strategy %d, moving from output %s to output %s", __func__, psId,
              std::to_string(srcOutputs[0]).c_str(),
              std::to_string(dstOutputs[0]).c_str());
        // mute strategy while moving tracks from one output to another
        for (audio_io_handle_t srcOut : srcOutputs) {
            sp<SwAudioOutputDescriptor> desc = mPreviousOutputs.valueFor(srcOut);
            if (desc != 0 && desc->isStrategyActive(psId)) {
                setStrategyMute(psId, true, desc);
                setStrategyMute(psId, false, desc, maxLatency * LATENCY_MUTE_FACTOR,
                                newDevices.types());
            }
            sp<SourceClientDescriptor> source = getSourceForAttributesOnOutput(srcOut, attr);
            if (source != 0){
                connectAudioSource(source);
            }
        }

        // Move effects associated to this stream from previous output to new output
        if (followsSameRouting(attr, attributes_initializer(AUDIO_USAGE_MEDIA))) {
            selectOutputForMusicEffects();
        }
        // Move tracks associated to this stream (and linked) from previous output to new output
        for (auto stream :  mEngine->getStreamTypesForProductStrategy(psId)) {
            mpClientInterface->invalidateStream(stream);
        }
    }
}

void AudioPolicyManager::checkOutputForAllStrategies()
{
    for (const auto &strategy : mEngine->getOrderedProductStrategies()) {
        auto attributes = mEngine->getAllAttributesForProductStrategy(strategy).front();
        checkOutputForAttributes(attributes);
    }
}

void AudioPolicyManager::checkSecondaryOutputs() {
    std::set<audio_stream_type_t> streamsToInvalidate;
    for (size_t i = 0; i < mOutputs.size(); i++) {
        const sp<SwAudioOutputDescriptor>& outputDescriptor = mOutputs[i];
        for (const sp<TrackClientDescriptor>& client : outputDescriptor->getClientIterable()) {
            sp<SwAudioOutputDescriptor> desc;
            std::vector<sp<SwAudioOutputDescriptor>> secondaryDescs;
            status_t status = mPolicyMixes.getOutputForAttr(client->attributes(), client->uid(),
                                                            client->flags(), desc, &secondaryDescs);
            if (status != OK ||
                !std::equal(client->getSecondaryOutputs().begin(),
                            client->getSecondaryOutputs().end(),
                            secondaryDescs.begin(), secondaryDescs.end())) {
                streamsToInvalidate.insert(client->stream());
            }
        }
    }
    for (audio_stream_type_t stream : streamsToInvalidate) {
        ALOGD("%s Invalidate stream %d due to secondary output change", __func__, stream);
        mpClientInterface->invalidateStream(stream);
    }
}

void AudioPolicyManager::checkA2dpSuspend()
{
    audio_io_handle_t a2dpOutput = mOutputs.getA2dpOutput();
    if (a2dpOutput == 0 || mOutputs.isA2dpOffloadedOnPrimary()) {
        mA2dpSuspended = false;
        return;
    }

    bool isScoConnected =
            ((mAvailableInputDevices.types() & AUDIO_DEVICE_IN_BLUETOOTH_SCO_HEADSET &
                    ~AUDIO_DEVICE_BIT_IN) != 0) ||
            ((mAvailableOutputDevices.types() & AUDIO_DEVICE_OUT_ALL_SCO) != 0);

    // if suspended, restore A2DP output if:
    //      ((SCO device is NOT connected) ||
    //       ((forced usage communication is NOT SCO) && (forced usage for record is NOT SCO) &&
    //        (phone state is NOT in call) && (phone state is NOT ringing)))
    //
    // if not suspended, suspend A2DP output if:
    //      (SCO device is connected) &&
    //       ((forced usage for communication is SCO) || (forced usage for record is SCO) ||
    //       ((phone state is in call) || (phone state is ringing)))
    //
    if (mA2dpSuspended) {
        if (!isScoConnected ||
             ((mEngine->getForceUse(AUDIO_POLICY_FORCE_FOR_COMMUNICATION) !=
                     AUDIO_POLICY_FORCE_BT_SCO) &&
              (mEngine->getForceUse(AUDIO_POLICY_FORCE_FOR_RECORD) !=
                      AUDIO_POLICY_FORCE_BT_SCO) &&
              (mEngine->getPhoneState() != AUDIO_MODE_IN_CALL) &&
              (mEngine->getPhoneState() != AUDIO_MODE_RINGTONE))) {

            mpClientInterface->restoreOutput(a2dpOutput);
            mA2dpSuspended = false;
        }
    } else {
        if (isScoConnected &&
             ((mEngine->getForceUse(AUDIO_POLICY_FORCE_FOR_COMMUNICATION) ==
                     AUDIO_POLICY_FORCE_BT_SCO) ||
              (mEngine->getForceUse(AUDIO_POLICY_FORCE_FOR_RECORD) ==
                      AUDIO_POLICY_FORCE_BT_SCO) ||
              (mEngine->getPhoneState() == AUDIO_MODE_IN_CALL) ||
              (mEngine->getPhoneState() == AUDIO_MODE_RINGTONE))) {

            mpClientInterface->suspendOutput(a2dpOutput);
            mA2dpSuspended = true;
        }
    }
}

DeviceVector AudioPolicyManager::getNewOutputDevices(const sp<SwAudioOutputDescriptor>& outputDesc,
                                                     bool fromCache)
{
    DeviceVector devices;

    ssize_t index = mAudioPatches.indexOfKey(outputDesc->getPatchHandle());
    if (index >= 0) {
        sp<AudioPatch> patchDesc = mAudioPatches.valueAt(index);
        if (patchDesc->mUid != mUidCached) {
            ALOGV("%s device %s forced by patch %d", __func__,
                  outputDesc->devices().toString().c_str(), outputDesc->getPatchHandle());
            return  outputDesc->devices();
        }
    }

    // Honor explicit routing requests only if no client using default routing is active on this
    // input: a specific app can not force routing for other apps by setting a preferred device.
    bool active; // unused
    sp<DeviceDescriptor> device =
        findPreferredDevice(outputDesc, PRODUCT_STRATEGY_NONE, active, mAvailableOutputDevices);
    if (device != nullptr) {
        return DeviceVector(device);
    }

    // Legacy Engine cannot take care of bus devices and mix, so we need to handle the conflict
    // of setForceUse / Default Bus device here
    device = mPolicyMixes.getDeviceAndMixForOutput(outputDesc, mAvailableOutputDevices);
    if (device != nullptr) {
        return DeviceVector(device);
    }

    for (const auto &productStrategy : mEngine->getOrderedProductStrategies()) {
        StreamTypeVector streams = mEngine->getStreamTypesForProductStrategy(productStrategy);
        auto attr = mEngine->getAllAttributesForProductStrategy(productStrategy).front();

        if ((hasVoiceStream(streams) &&
             (isInCall() || mOutputs.isStrategyActiveOnSameModule(productStrategy, outputDesc))) ||
             ((hasStream(streams, AUDIO_STREAM_ALARM) || hasStream(streams, AUDIO_STREAM_ENFORCED_AUDIBLE)) &&
                mOutputs.isStrategyActiveOnSameModule(productStrategy, outputDesc)) ||
                outputDesc->isStrategyActive(productStrategy)) {
            // Retrieval of devices for voice DL is done on primary output profile, cannot
            // check the route (would force modifying configuration file for this profile)
            devices = mEngine->getOutputDevicesForAttributes(attr, nullptr, fromCache);
            break;
        }
    }
    ALOGV("%s selected devices %s", __func__, devices.toString().c_str());
    return devices;
}

sp<DeviceDescriptor> AudioPolicyManager::getNewInputDevice(
        const sp<AudioInputDescriptor>& inputDesc)
{
    sp<DeviceDescriptor> device;

    ssize_t index = mAudioPatches.indexOfKey(inputDesc->getPatchHandle());
    if (index >= 0) {
        sp<AudioPatch> patchDesc = mAudioPatches.valueAt(index);
        if (patchDesc->mUid != mUidCached) {
            ALOGV("getNewInputDevice() device %s forced by patch %d",
                  inputDesc->getDevice()->toString().c_str(), inputDesc->getPatchHandle());
            return inputDesc->getDevice();
        }
    }

    // Honor explicit routing requests only if no client using default routing is active on this
    // input: a specific app can not force routing for other apps by setting a preferred device.
    bool active;
    device = findPreferredDevice(inputDesc, AUDIO_SOURCE_DEFAULT, active, mAvailableInputDevices);
    if (device != nullptr) {
        return device;
    }

    // If we are not in call and no client is active on this input, this methods returns
    // a null sp<>, causing the patch on the input stream to be released.
    audio_attributes_t attributes = inputDesc->getHighestPriorityAttributes();
    if (attributes.source == AUDIO_SOURCE_DEFAULT && isInCall()) {
        attributes.source = AUDIO_SOURCE_VOICE_COMMUNICATION;
    }
    if (attributes.source != AUDIO_SOURCE_DEFAULT) {
        device = mEngine->getInputDeviceForAttributes(attributes);
    }

    return device;
}

bool AudioPolicyManager::streamsMatchForvolume(audio_stream_type_t stream1,
                                               audio_stream_type_t stream2) {
    return (stream1 == stream2);
}

audio_devices_t AudioPolicyManager::getDevicesForStream(audio_stream_type_t stream) {
    // By checking the range of stream before calling getStrategy, we avoid
    // getOutputDevicesForStream's behavior for invalid streams.
    // engine's getOutputDevicesForStream would fallback on its default behavior (most probably
    // device for music stream), but we want to return the empty set.
    if (stream < AUDIO_STREAM_MIN || stream >= AUDIO_STREAM_PUBLIC_CNT) {
        return AUDIO_DEVICE_NONE;
    }
    DeviceVector activeDevices;
    DeviceVector devices;
    for (audio_stream_type_t curStream = AUDIO_STREAM_MIN; curStream < AUDIO_STREAM_PUBLIC_CNT;
         curStream = (audio_stream_type_t) (curStream + 1)) {
        if (!streamsMatchForvolume(stream, curStream)) {
            continue;
        }
        DeviceVector curDevices = mEngine->getOutputDevicesForStream(curStream, false/*fromCache*/);
        devices.merge(curDevices);
        for (audio_io_handle_t output : getOutputsForDevices(curDevices, mOutputs)) {
            sp<AudioOutputDescriptor> outputDesc = mOutputs.valueFor(output);
            if (outputDesc->isActive(toVolumeSource(curStream))) {
                activeDevices.merge(outputDesc->devices());
            }
        }
    }

    // Favor devices selected on active streams if any to report correct device in case of
    // explicit device selection
    if (!activeDevices.isEmpty()) {
        devices = activeDevices;
    }
    /*Filter SPEAKER_SAFE out of results, as AudioService doesn't know about it
      and doesn't really need to.*/
    DeviceVector speakerSafeDevices = devices.getDevicesFromTypeMask(AUDIO_DEVICE_OUT_SPEAKER_SAFE);
    if (!speakerSafeDevices.isEmpty()) {
        devices.merge(mAvailableOutputDevices.getDevicesFromTypeMask(AUDIO_DEVICE_OUT_SPEAKER));
        devices.remove(speakerSafeDevices);
    }
    return devices.types();
}

void AudioPolicyManager::handleNotificationRoutingForStream(audio_stream_type_t stream) {
    switch(stream) {
    case AUDIO_STREAM_MUSIC:
        checkOutputForAttributes(attributes_initializer(AUDIO_USAGE_NOTIFICATION));
        updateDevicesAndOutputs();
        break;
    default:
        break;
    }
}

uint32_t AudioPolicyManager::handleEventForBeacon(int event) {

    // skip beacon mute management if a dedicated TTS output is available
    if (mTtsOutputAvailable) {
        return 0;
    }

    switch(event) {
    case STARTING_OUTPUT:
        mBeaconMuteRefCount++;
        break;
    case STOPPING_OUTPUT:
        if (mBeaconMuteRefCount > 0) {
            mBeaconMuteRefCount--;
        }
        break;
    case STARTING_BEACON:
        mBeaconPlayingRefCount++;
        break;
    case STOPPING_BEACON:
        if (mBeaconPlayingRefCount > 0) {
            mBeaconPlayingRefCount--;
        }
        break;
    }

    if (mBeaconMuteRefCount > 0) {
        // any playback causes beacon to be muted
        return setBeaconMute(true);
    } else {
        // no other playback: unmute when beacon starts playing, mute when it stops
        return setBeaconMute(mBeaconPlayingRefCount == 0);
    }
}

uint32_t AudioPolicyManager::setBeaconMute(bool mute) {
    ALOGV("setBeaconMute(%d) mBeaconMuteRefCount=%d mBeaconPlayingRefCount=%d",
            mute, mBeaconMuteRefCount, mBeaconPlayingRefCount);
    // keep track of muted state to avoid repeating mute/unmute operations
    if (mBeaconMuted != mute) {
        // mute/unmute AUDIO_STREAM_TTS on all outputs
        ALOGV("\t muting %d", mute);
        uint32_t maxLatency = 0;
        auto ttsVolumeSource = toVolumeSource(AUDIO_STREAM_TTS);
        for (size_t i = 0; i < mOutputs.size(); i++) {
            sp<SwAudioOutputDescriptor> desc = mOutputs.valueAt(i);
            setVolumeSourceMute(ttsVolumeSource, mute/*on*/, desc, 0 /*delay*/, AUDIO_DEVICE_NONE);
            const uint32_t latency = desc->latency() * 2;
            if (latency > maxLatency) {
                maxLatency = latency;
            }
        }
        mBeaconMuted = mute;
        return maxLatency;
    }
    return 0;
}

void AudioPolicyManager::updateDevicesAndOutputs()
{
    mEngine->updateDeviceSelectionCache();
    mPreviousOutputs = mOutputs;
}

uint32_t AudioPolicyManager::checkDeviceMuteStrategies(const sp<AudioOutputDescriptor>& outputDesc,
                                                       const DeviceVector &prevDevices,
                                                       uint32_t delayMs)
{
    // mute/unmute strategies using an incompatible device combination
    // if muting, wait for the audio in pcm buffer to be drained before proceeding
    // if unmuting, unmute only after the specified delay
    if (outputDesc->isDuplicated()) {
        return 0;
    }

    uint32_t muteWaitMs = 0;
    DeviceVector devices = outputDesc->devices();
    bool shouldMute = outputDesc->isActive() && (devices.size() >= 2);

    auto productStrategies = mEngine->getOrderedProductStrategies();
    for (const auto &productStrategy : productStrategies) {
        auto attributes = mEngine->getAllAttributesForProductStrategy(productStrategy).front();
        DeviceVector curDevices =
                mEngine->getOutputDevicesForAttributes(attributes, nullptr, false/*fromCache*/);
        curDevices = curDevices.filter(outputDesc->supportedDevices());
        bool mute = shouldMute && curDevices.containsAtLeastOne(devices) && curDevices != devices;
        bool doMute = false;

        if (mute && !outputDesc->isStrategyMutedByDevice(productStrategy)) {
            doMute = true;
            outputDesc->setStrategyMutedByDevice(productStrategy, true);
        } else if (!mute && outputDesc->isStrategyMutedByDevice(productStrategy)) {
            doMute = true;
            outputDesc->setStrategyMutedByDevice(productStrategy, false);
        }
        if (doMute) {
            for (size_t j = 0; j < mOutputs.size(); j++) {
                sp<AudioOutputDescriptor> desc = mOutputs.valueAt(j);
                // skip output if it does not share any device with current output
                if (!desc->supportedDevices().containsAtLeastOne(outputDesc->supportedDevices())) {
                    continue;
                }
                ALOGVV("%s() %s (curDevice %s)", __func__,
                      mute ? "muting" : "unmuting", curDevices.toString().c_str());
                setStrategyMute(productStrategy, mute, desc, mute ? 0 : delayMs);
                if (desc->isStrategyActive(productStrategy)) {
                    if (mute) {
                        // FIXME: should not need to double latency if volume could be applied
                        // immediately by the audioflinger mixer. We must account for the delay
                        // between now and the next time the audioflinger thread for this output
                        // will process a buffer (which corresponds to one buffer size,
                        // usually 1/2 or 1/4 of the latency).
                        if (muteWaitMs < desc->latency() * 2) {
                            muteWaitMs = desc->latency() * 2;
                        }
                    }
                }
            }
        }
    }

    // temporary mute output if device selection changes to avoid volume bursts due to
    // different per device volumes
    if (outputDesc->isActive() && (devices != prevDevices)) {
        uint32_t tempMuteWaitMs = outputDesc->latency() * 2;
        // temporary mute duration is conservatively set to 4 times the reported latency
        uint32_t tempMuteDurationMs = outputDesc->latency() * 4;
        if (muteWaitMs < tempMuteWaitMs) {
            muteWaitMs = tempMuteWaitMs;
        }
        for (const auto &activeVs : outputDesc->getActiveVolumeSources()) {
            // make sure that we do not start the temporary mute period too early in case of
            // delayed device change
            setVolumeSourceMute(activeVs, true, outputDesc, delayMs);
            setVolumeSourceMute(activeVs, false, outputDesc, delayMs + tempMuteDurationMs,
                                devices.types());
        }
    }

    // wait for the PCM output buffers to empty before proceeding with the rest of the command
    if (muteWaitMs > delayMs) {
        muteWaitMs -= delayMs;
        usleep(muteWaitMs * 1000);
        return muteWaitMs;
    }
    return 0;
}

uint32_t AudioPolicyManager::setOutputDevices(const sp<SwAudioOutputDescriptor>& outputDesc,
                                              const DeviceVector &devices,
                                              bool force,
                                              int delayMs,
                                              audio_patch_handle_t *patchHandle,
                                              bool requiresMuteCheck)
{
    ALOGV("%s device %s delayMs %d", __func__, devices.toString().c_str(), delayMs);
    uint32_t muteWaitMs;

    if (outputDesc->isDuplicated()) {
        muteWaitMs = setOutputDevices(outputDesc->subOutput1(), devices, force, delayMs,
                nullptr /* patchHandle */, requiresMuteCheck);
        muteWaitMs += setOutputDevices(outputDesc->subOutput2(), devices, force, delayMs,
                nullptr /* patchHandle */, requiresMuteCheck);
        return muteWaitMs;
    }

    // filter devices according to output selected
    DeviceVector filteredDevices = outputDesc->filterSupportedDevices(devices);
    DeviceVector prevDevices = outputDesc->devices();

    // no need to proceed if new device is not AUDIO_DEVICE_NONE and not supported by current
    // output profile or if new device is not supported AND previous device(s) is(are) still
    // available (otherwise reset device must be done on the output)
    if (!devices.isEmpty() && filteredDevices.isEmpty() &&
            !mAvailableOutputDevices.filter(prevDevices).empty()) {
        ALOGV("%s: unsupported device %s for output", __func__, devices.toString().c_str());
        return 0;
    }

    ALOGV("setOutputDevices() prevDevice %s", prevDevices.toString().c_str());

    if (!filteredDevices.isEmpty()) {
        outputDesc->setDevices(filteredDevices);
    }

    // if the outputs are not materially active, there is no need to mute.
    if (requiresMuteCheck) {
        muteWaitMs = checkDeviceMuteStrategies(outputDesc, prevDevices, delayMs);
    } else {
        ALOGV("%s: suppressing checkDeviceMuteStrategies", __func__);
        muteWaitMs = 0;
    }

    // Do not change the routing if:
    //      the requested device is AUDIO_DEVICE_NONE
    //      OR the requested device is the same as current device
    //  AND force is not specified
    //  AND the output is connected by a valid audio patch.
    // Doing this check here allows the caller to call setOutputDevices() without conditions
    if ((filteredDevices.isEmpty() || filteredDevices == prevDevices) &&
            !force && outputDesc->getPatchHandle() != 0) {
        ALOGV("%s setting same device %s or null device, force=%d, patch handle=%d", __func__,
              filteredDevices.toString().c_str(), force, outputDesc->getPatchHandle());
        return muteWaitMs;
    }

    ALOGV("%s changing device to %s", __func__, filteredDevices.toString().c_str());

    // do the routing
    if (filteredDevices.isEmpty()) {
        resetOutputDevice(outputDesc, delayMs, NULL);
    } else {
        PatchBuilder patchBuilder;
        patchBuilder.addSource(outputDesc);
        ALOG_ASSERT(filteredDevices.size() <= AUDIO_PATCH_PORTS_MAX, "Too many sink ports");
        for (const auto &filteredDevice : filteredDevices) {
            patchBuilder.addSink(filteredDevice);
        }

        installPatch(__func__, patchHandle, outputDesc.get(), patchBuilder.patch(), delayMs);
    }

    // update stream volumes according to new device
    applyStreamVolumes(outputDesc, filteredDevices.types(), delayMs);

    return muteWaitMs;
}

status_t AudioPolicyManager::resetOutputDevice(const sp<AudioOutputDescriptor>& outputDesc,
                                               int delayMs,
                                               audio_patch_handle_t *patchHandle)
{
    ssize_t index;
    if (patchHandle) {
        index = mAudioPatches.indexOfKey(*patchHandle);
    } else {
        index = mAudioPatches.indexOfKey(outputDesc->getPatchHandle());
    }
    if (index < 0) {
        return INVALID_OPERATION;
    }
    sp< AudioPatch> patchDesc = mAudioPatches.valueAt(index);
    status_t status = mpClientInterface->releaseAudioPatch(patchDesc->mAfPatchHandle, delayMs);
    ALOGV("resetOutputDevice() releaseAudioPatch returned %d", status);
    outputDesc->setPatchHandle(AUDIO_PATCH_HANDLE_NONE);
    removeAudioPatch(patchDesc->mHandle);
    nextAudioPortGeneration();
    mpClientInterface->onAudioPatchListUpdate();
    return status;
}

status_t AudioPolicyManager::setInputDevice(audio_io_handle_t input,
                                            const sp<DeviceDescriptor> &device,
                                            bool force,
                                            audio_patch_handle_t *patchHandle)
{
    status_t status = NO_ERROR;

    sp<AudioInputDescriptor> inputDesc = mInputs.valueFor(input);
    if ((device != nullptr) && ((device != inputDesc->getDevice()) || force)) {
        inputDesc->setDevice(device);

        if (mAvailableInputDevices.contains(device)) {
            PatchBuilder patchBuilder;
            patchBuilder.addSink(inputDesc,
            // AUDIO_SOURCE_HOTWORD is for internal use only:
            // handled as AUDIO_SOURCE_VOICE_RECOGNITION by the audio HAL
                    [inputDesc](const PatchBuilder::mix_usecase_t& usecase) {
                        auto result = usecase;
                        if (result.source == AUDIO_SOURCE_HOTWORD && !inputDesc->isSoundTrigger()) {
                            result.source = AUDIO_SOURCE_VOICE_RECOGNITION;
                        }
                        return result; }).
            //only one input device for now
                    addSource(device);
            status = installPatch(__func__, patchHandle, inputDesc.get(), patchBuilder.patch(), 0);
        }
    }
    return status;
}

status_t AudioPolicyManager::resetInputDevice(audio_io_handle_t input,
                                              audio_patch_handle_t *patchHandle)
{
    sp<AudioInputDescriptor> inputDesc = mInputs.valueFor(input);
    ssize_t index;
    if (patchHandle) {
        index = mAudioPatches.indexOfKey(*patchHandle);
    } else {
        index = mAudioPatches.indexOfKey(inputDesc->getPatchHandle());
    }
    if (index < 0) {
        return INVALID_OPERATION;
    }
    sp< AudioPatch> patchDesc = mAudioPatches.valueAt(index);
    status_t status = mpClientInterface->releaseAudioPatch(patchDesc->mAfPatchHandle, 0);
    ALOGV("resetInputDevice() releaseAudioPatch returned %d", status);
    inputDesc->setPatchHandle(AUDIO_PATCH_HANDLE_NONE);
    removeAudioPatch(patchDesc->mHandle);
    nextAudioPortGeneration();
    mpClientInterface->onAudioPatchListUpdate();
    return status;
}

sp<IOProfile> AudioPolicyManager::getInputProfile(const sp<DeviceDescriptor> &device,
                                                  uint32_t& samplingRate,
                                                  audio_format_t& format,
                                                  audio_channel_mask_t& channelMask,
                                                  audio_input_flags_t flags)
{
    // Choose an input profile based on the requested capture parameters: select the first available
    // profile supporting all requested parameters.
    //
    // TODO: perhaps isCompatibleProfile should return a "matching" score so we can return
    // the best matching profile, not the first one.

    sp<IOProfile> firstInexact;
    uint32_t updatedSamplingRate = 0;
    audio_format_t updatedFormat = AUDIO_FORMAT_INVALID;
    audio_channel_mask_t updatedChannelMask = AUDIO_CHANNEL_INVALID;
    for (const auto& hwModule : mHwModules) {
        for (const auto& profile : hwModule->getInputProfiles()) {
            // profile->log();
            //updatedFormat = format;
            if (profile->isCompatibleProfile(DeviceVector(device), samplingRate,
                                             &samplingRate  /*updatedSamplingRate*/,
                                             format,
                                             &format,       /*updatedFormat*/
                                             channelMask,
                                             &channelMask   /*updatedChannelMask*/,
                                             // FIXME ugly cast
                                             (audio_output_flags_t) flags,
                                             true /*exactMatchRequiredForInputFlags*/)) {
                return profile;
            }
            if (firstInexact == nullptr && profile->isCompatibleProfile(DeviceVector(device),
                                             samplingRate,
                                             &updatedSamplingRate,
                                             format,
                                             &updatedFormat,
                                             channelMask,
                                             &updatedChannelMask,
                                             // FIXME ugly cast
                                             (audio_output_flags_t) flags,
                                             false /*exactMatchRequiredForInputFlags*/)) {
                firstInexact = profile;
            }

        }
    }
    if (firstInexact != nullptr) {
        samplingRate = updatedSamplingRate;
        format = updatedFormat;
        channelMask = updatedChannelMask;
        return firstInexact;
    }
    return NULL;
}

float AudioPolicyManager::computeVolume(IVolumeCurves &curves,
                                        VolumeSource volumeSource,
                                        int index,
                                        audio_devices_t device)
{
    float volumeDb = curves.volIndexToDb(Volume::getDeviceCategory(device), index);

    // handle the case of accessibility active while a ringtone is playing: if the ringtone is much
    // louder than the accessibility prompt, the prompt cannot be heard, thus masking the touch
    // exploration of the dialer UI. In this situation, bring the accessibility volume closer to
    // the ringtone volume
    const auto callVolumeSrc = toVolumeSource(AUDIO_STREAM_VOICE_CALL);
    const auto ringVolumeSrc = toVolumeSource(AUDIO_STREAM_RING);
    const auto musicVolumeSrc = toVolumeSource(AUDIO_STREAM_MUSIC);
    const auto alarmVolumeSrc = toVolumeSource(AUDIO_STREAM_ALARM);

    if (volumeSource == toVolumeSource(AUDIO_STREAM_ACCESSIBILITY)
            && (AUDIO_MODE_RINGTONE == mEngine->getPhoneState()) &&
            mOutputs.isActive(ringVolumeSrc, 0)) {
        auto &ringCurves = getVolumeCurves(AUDIO_STREAM_RING);
        const float ringVolumeDb = computeVolume(ringCurves, ringVolumeSrc, index, device);
        return ringVolumeDb - 4 > volumeDb ? ringVolumeDb - 4 : volumeDb;
    }

    // in-call: always cap volume by voice volume + some low headroom
    if ((volumeSource != callVolumeSrc && (isInCall() ||
                                           mOutputs.isActiveLocally(callVolumeSrc))) &&
            (volumeSource == toVolumeSource(AUDIO_STREAM_SYSTEM) ||
             volumeSource == ringVolumeSrc || volumeSource == musicVolumeSrc ||
             volumeSource == alarmVolumeSrc ||
             volumeSource == toVolumeSource(AUDIO_STREAM_NOTIFICATION) ||
             volumeSource == toVolumeSource(AUDIO_STREAM_ENFORCED_AUDIBLE) ||
             volumeSource == toVolumeSource(AUDIO_STREAM_DTMF) ||
             volumeSource == toVolumeSource(AUDIO_STREAM_ACCESSIBILITY))) {
        auto &voiceCurves = getVolumeCurves(callVolumeSrc);
        int voiceVolumeIndex = voiceCurves.getVolumeIndex(device);
        const float maxVoiceVolDb =
                computeVolume(voiceCurves, callVolumeSrc, voiceVolumeIndex, device)
                + IN_CALL_EARPIECE_HEADROOM_DB;
        // FIXME: Workaround for call screening applications until a proper audio mode is defined
        // to support this scenario : Exempt the RING stream from the audio cap if the audio was
        // programmatically muted.
        // VOICE_CALL stream has minVolumeIndex > 0 : Users cannot set the volume of voice calls to
        // 0. We don't want to cap volume when the system has programmatically muted the voice call
        // stream. See setVolumeCurveIndex() for more information.
        bool exemptFromCapping = (volumeSource == ringVolumeSrc) && (voiceVolumeIndex == 0);
        ALOGV_IF(exemptFromCapping, "%s volume source %d at vol=%f not capped", __func__,
                 volumeSource, volumeDb);
        if ((volumeDb > maxVoiceVolDb) && !exemptFromCapping) {
            ALOGV("%s volume source %d at vol=%f overriden by volume group %d at vol=%f", __func__,
                  volumeSource, volumeDb, callVolumeSrc, maxVoiceVolDb);
            volumeDb = maxVoiceVolDb;
        }
    }
    // if a headset is connected, apply the following rules to ring tones and notifications
    // to avoid sound level bursts in user's ears:
    // - always attenuate notifications volume by 6dB
    // - attenuate ring tones volume by 6dB unless music is not playing and
    // speaker is part of the select devices
    // - if music is playing, always limit the volume to current music volume,
    // with a minimum threshold at -36dB so that notification is always perceived.
    if ((device & (AUDIO_DEVICE_OUT_BLUETOOTH_A2DP | AUDIO_DEVICE_OUT_BLUETOOTH_A2DP_HEADPHONES |
                   AUDIO_DEVICE_OUT_WIRED_HEADSET | AUDIO_DEVICE_OUT_WIRED_HEADPHONE |
                   AUDIO_DEVICE_OUT_USB_HEADSET | AUDIO_DEVICE_OUT_HEARING_AID)) &&
            ((volumeSource == alarmVolumeSrc ||
              volumeSource == ringVolumeSrc) ||
             (volumeSource == toVolumeSource(AUDIO_STREAM_NOTIFICATION)) ||
             (volumeSource == toVolumeSource(AUDIO_STREAM_SYSTEM)) ||
             ((volumeSource == toVolumeSource(AUDIO_STREAM_ENFORCED_AUDIBLE)) &&
              (mEngine->getForceUse(AUDIO_POLICY_FORCE_FOR_SYSTEM) == AUDIO_POLICY_FORCE_NONE))) &&
            curves.canBeMuted()) {

        // when the phone is ringing we must consider that music could have been paused just before
        // by the music application and behave as if music was active if the last music track was
        // just stopped
        if (isStreamActive(AUDIO_STREAM_MUSIC, SONIFICATION_HEADSET_MUSIC_DELAY) ||
                mLimitRingtoneVolume) {
            volumeDb += SONIFICATION_HEADSET_VOLUME_FACTOR_DB;
            audio_devices_t musicDevice =
                    mEngine->getOutputDevicesForAttributes(attributes_initializer(AUDIO_USAGE_MEDIA),
                                                           nullptr, true /*fromCache*/).types();
            auto &musicCurves = getVolumeCurves(AUDIO_STREAM_MUSIC);
            float musicVolDb = computeVolume(musicCurves, musicVolumeSrc,
                                             musicCurves.getVolumeIndex(musicDevice), musicDevice);
            float minVolDb = (musicVolDb > SONIFICATION_HEADSET_VOLUME_MIN_DB) ?
                        musicVolDb : SONIFICATION_HEADSET_VOLUME_MIN_DB;
            if (volumeDb > minVolDb) {
                volumeDb = minVolDb;
                ALOGV("computeVolume limiting volume to %f musicVol %f", minVolDb, musicVolDb);
            }
            if (device & (AUDIO_DEVICE_OUT_BLUETOOTH_A2DP |
                          AUDIO_DEVICE_OUT_BLUETOOTH_A2DP_HEADPHONES)) {
                // on A2DP, also ensure notification volume is not too low compared to media when
                // intended to be played
                if ((volumeDb > -96.0f) &&
                        (musicVolDb - SONIFICATION_A2DP_MAX_MEDIA_DIFF_DB > volumeDb)) {
                    ALOGV("%s increasing volume for volume source=%d device=0x%X from %f to %f",
                          __func__, volumeSource, device, volumeDb,
                          musicVolDb - SONIFICATION_A2DP_MAX_MEDIA_DIFF_DB);
                    volumeDb = musicVolDb - SONIFICATION_A2DP_MAX_MEDIA_DIFF_DB;
                }
            }
        } else if ((Volume::getDeviceForVolume(device) != AUDIO_DEVICE_OUT_SPEAKER) ||
                   (!(volumeSource == alarmVolumeSrc || volumeSource == ringVolumeSrc))) {
            volumeDb += SONIFICATION_HEADSET_VOLUME_FACTOR_DB;
        }
    }

    return volumeDb;
}

int AudioPolicyManager::rescaleVolumeIndex(int srcIndex,
                                           VolumeSource fromVolumeSource,
                                           VolumeSource toVolumeSource)
{
    if (fromVolumeSource == toVolumeSource) {
        return srcIndex;
    }
    auto &srcCurves = getVolumeCurves(fromVolumeSource);
    auto &dstCurves = getVolumeCurves(toVolumeSource);
    float minSrc = (float)srcCurves.getVolumeIndexMin();
    float maxSrc = (float)srcCurves.getVolumeIndexMax();
    float minDst = (float)dstCurves.getVolumeIndexMin();
    float maxDst = (float)dstCurves.getVolumeIndexMax();

    // preserve mute request or correct range
    if (srcIndex < minSrc) {
        if (srcIndex == 0) {
            return 0;
        }
        srcIndex = minSrc;
    } else if (srcIndex > maxSrc) {
        srcIndex = maxSrc;
    }
    return (int)(minDst + ((srcIndex - minSrc) * (maxDst - minDst)) / (maxSrc - minSrc));
}

status_t AudioPolicyManager::checkAndSetVolume(IVolumeCurves &curves,
                                               VolumeSource volumeSource,
                                               int index,
                                               const sp<AudioOutputDescriptor>& outputDesc,
                                               audio_devices_t device,
                                               int delayMs,
                                               bool force)
{
    // do not change actual attributes volume if the attributes is muted
    if (outputDesc->isMuted(volumeSource)) {
        ALOGVV("%s: volume source %d muted count %d active=%d", __func__, volumeSource,
               outputDesc->getMuteCount(volumeSource), outputDesc->isActive(volumeSource));
        return NO_ERROR;
    }
    VolumeSource callVolSrc = toVolumeSource(AUDIO_STREAM_VOICE_CALL);
    VolumeSource btScoVolSrc = toVolumeSource(AUDIO_STREAM_BLUETOOTH_SCO);
    bool isVoiceVolSrc = callVolSrc == volumeSource;
    bool isBtScoVolSrc = btScoVolSrc == volumeSource;

    audio_policy_forced_cfg_t forceUseForComm =
            mEngine->getForceUse(AUDIO_POLICY_FORCE_FOR_COMMUNICATION);
    // do not change in call volume if bluetooth is connected and vice versa
    // if sco and call follow same curves, bypass forceUseForComm
    if ((callVolSrc != btScoVolSrc) &&
            ((isVoiceVolSrc && forceUseForComm == AUDIO_POLICY_FORCE_BT_SCO) ||
             (isBtScoVolSrc && forceUseForComm != AUDIO_POLICY_FORCE_BT_SCO))) {
        ALOGV("%s cannot set volume group %d volume with force use = %d for comm", __func__,
             volumeSource, forceUseForComm);
        return INVALID_OPERATION;
    }
    if (device == AUDIO_DEVICE_NONE) {
        device = outputDesc->devices().types();
    }

    float volumeDb = computeVolume(curves, volumeSource, index, device);
    if (outputDesc->isFixedVolume(device) ||
            // Force VoIP volume to max for bluetooth SCO
            ((isVoiceVolSrc || isBtScoVolSrc) && (device & AUDIO_DEVICE_OUT_ALL_SCO) != 0)) {
        volumeDb = 0.0f;
    }
    outputDesc->setVolume(volumeDb, volumeSource, curves.getStreamTypes(), device, delayMs, force);

    if (isVoiceVolSrc || isBtScoVolSrc) {
        float voiceVolume;
        // Force voice volume to max for bluetooth SCO as volume is managed by the headset
        if (isVoiceVolSrc) {
            voiceVolume = (float)index/(float)curves.getVolumeIndexMax();
        } else {
            voiceVolume = 1.0;
        }
        if (voiceVolume != mLastVoiceVolume) {
            mpClientInterface->setVoiceVolume(voiceVolume, delayMs);
            mLastVoiceVolume = voiceVolume;
        }
    }
    return NO_ERROR;
}

void AudioPolicyManager::applyStreamVolumes(const sp<AudioOutputDescriptor>& outputDesc,
                                                audio_devices_t device,
                                                int delayMs,
                                                bool force)
{
    ALOGVV("applyStreamVolumes() for device %08x", device);
    for (const auto &volumeGroup : mEngine->getVolumeGroups()) {
        auto &curves = getVolumeCurves(toVolumeSource(volumeGroup));
        checkAndSetVolume(curves, toVolumeSource(volumeGroup),
                          curves.getVolumeIndex(device), outputDesc, device, delayMs, force);
    }
}

void AudioPolicyManager::setStrategyMute(product_strategy_t strategy,
                                         bool on,
                                         const sp<AudioOutputDescriptor>& outputDesc,
                                         int delayMs,
                                         audio_devices_t device)
{
    std::vector<VolumeSource> sourcesToMute;
    for (auto attributes: mEngine->getAllAttributesForProductStrategy(strategy)) {
        ALOGVV("%s() attributes %s, mute %d, output ID %d", __func__,
               toString(attributes).c_str(), on, outputDesc->getId());
        VolumeSource source = toVolumeSource(attributes);
        if (std::find(begin(sourcesToMute), end(sourcesToMute), source) == end(sourcesToMute)) {
            sourcesToMute.push_back(source);
        }
    }
    for (auto source : sourcesToMute) {
        setVolumeSourceMute(source, on, outputDesc, delayMs, device);
    }

}

void AudioPolicyManager::setVolumeSourceMute(VolumeSource volumeSource,
                                             bool on,
                                             const sp<AudioOutputDescriptor>& outputDesc,
                                             int delayMs,
                                             audio_devices_t device)
{
    if (device == AUDIO_DEVICE_NONE) {
        device = outputDesc->devices().types();
    }
    auto &curves = getVolumeCurves(volumeSource);
    if (on) {
        if (!outputDesc->isMuted(volumeSource)) {
            if (curves.canBeMuted() &&
                    (volumeSource != toVolumeSource(AUDIO_STREAM_ENFORCED_AUDIBLE) ||
                     (mEngine->getForceUse(AUDIO_POLICY_FORCE_FOR_SYSTEM) ==
                      AUDIO_POLICY_FORCE_NONE))) {
                checkAndSetVolume(curves, volumeSource, 0, outputDesc, device, delayMs);
            }
        }
        // increment mMuteCount after calling checkAndSetVolume() so that volume change is not
        // ignored
        outputDesc->incMuteCount(volumeSource);
    } else {
        if (!outputDesc->isMuted(volumeSource)) {
            ALOGV("%s unmuting non muted attributes!", __func__);
            return;
        }
        if (outputDesc->decMuteCount(volumeSource) == 0) {
            checkAndSetVolume(curves, volumeSource,
                              curves.getVolumeIndex(device),
                              outputDesc,
                              device,
                              delayMs);
        }
    }
}

bool AudioPolicyManager::isValidAttributes(const audio_attributes_t *paa)
{
    // has flags that map to a stream type?
    if ((paa->flags & (AUDIO_FLAG_AUDIBILITY_ENFORCED | AUDIO_FLAG_SCO | AUDIO_FLAG_BEACON)) != 0) {
        return true;
    }

    // has known usage?
    switch (paa->usage) {
    case AUDIO_USAGE_UNKNOWN:
    case AUDIO_USAGE_MEDIA:
    case AUDIO_USAGE_VOICE_COMMUNICATION:
    case AUDIO_USAGE_VOICE_COMMUNICATION_SIGNALLING:
    case AUDIO_USAGE_ALARM:
    case AUDIO_USAGE_NOTIFICATION:
    case AUDIO_USAGE_NOTIFICATION_TELEPHONY_RINGTONE:
    case AUDIO_USAGE_NOTIFICATION_COMMUNICATION_REQUEST:
    case AUDIO_USAGE_NOTIFICATION_COMMUNICATION_INSTANT:
    case AUDIO_USAGE_NOTIFICATION_COMMUNICATION_DELAYED:
    case AUDIO_USAGE_NOTIFICATION_EVENT:
    case AUDIO_USAGE_ASSISTANCE_ACCESSIBILITY:
    case AUDIO_USAGE_ASSISTANCE_NAVIGATION_GUIDANCE:
    case AUDIO_USAGE_ASSISTANCE_SONIFICATION:
    case AUDIO_USAGE_GAME:
    case AUDIO_USAGE_VIRTUAL_SOURCE:
    case AUDIO_USAGE_ASSISTANT:
        break;
    default:
        return false;
    }
    return true;
}

audio_policy_forced_cfg_t AudioPolicyManager::getForceUse(audio_policy_force_use_t usage)
{
    return mEngine->getForceUse(usage);
}

bool AudioPolicyManager::isInCall()
{
    return isStateInCall(mEngine->getPhoneState());
}

bool AudioPolicyManager::isStateInCall(int state)
{
    return is_state_in_call(state);
}

void AudioPolicyManager::cleanUpForDevice(const sp<DeviceDescriptor>& deviceDesc)
{
    for (ssize_t i = (ssize_t)mAudioSources.size() - 1; i >= 0; i--)  {
        sp<SourceClientDescriptor> sourceDesc = mAudioSources.valueAt(i);
        if (sourceDesc->srcDevice()->equals(deviceDesc)) {
            ALOGV("%s releasing audio source %d", __FUNCTION__, sourceDesc->portId());
            stopAudioSource(sourceDesc->portId());
        }
    }

    for (ssize_t i = (ssize_t)mAudioPatches.size() - 1; i >= 0; i--)  {
        sp<AudioPatch> patchDesc = mAudioPatches.valueAt(i);
        bool release = false;
        for (size_t j = 0; j < patchDesc->mPatch.num_sources && !release; j++)  {
            const struct audio_port_config *source = &patchDesc->mPatch.sources[j];
            if (source->type == AUDIO_PORT_TYPE_DEVICE &&
                    source->ext.device.type == deviceDesc->type()) {
                release = true;
            }
        }
        for (size_t j = 0; j < patchDesc->mPatch.num_sinks && !release; j++)  {
            const struct audio_port_config *sink = &patchDesc->mPatch.sinks[j];
            if (sink->type == AUDIO_PORT_TYPE_DEVICE &&
                    sink->ext.device.type == deviceDesc->type()) {
                release = true;
            }
        }
        if (release) {
            ALOGV("%s releasing patch %u", __FUNCTION__, patchDesc->mHandle);
            releaseAudioPatch(patchDesc->mHandle, patchDesc->mUid);
        }
    }

    mInputs.clearSessionRoutesForDevice(deviceDesc);

    mHwModules.cleanUpForDevice(deviceDesc);
}

void AudioPolicyManager::modifySurroundFormats(
        const sp<DeviceDescriptor>& devDesc, FormatVector *formatsPtr) {
    std::unordered_set<audio_format_t> enforcedSurround(
            devDesc->encodedFormats().begin(), devDesc->encodedFormats().end());
    std::unordered_set<audio_format_t> allSurround;  // A flat set of all known surround formats
    for (const auto& pair : mConfig.getSurroundFormats()) {
        allSurround.insert(pair.first);
        for (const auto& subformat : pair.second) allSurround.insert(subformat);
    }

    audio_policy_forced_cfg_t forceUse = mEngine->getForceUse(
            AUDIO_POLICY_FORCE_FOR_ENCODED_SURROUND);
    ALOGD("%s: forced use = %d", __FUNCTION__, forceUse);
    // This is the resulting set of formats depending on the surround mode:
    //   'all surround' = allSurround
    //   'enforced surround' = enforcedSurround [may include IEC69137 which isn't raw surround fmt]
    //   'non-surround' = not in 'all surround' and not in 'enforced surround'
    //   'manual surround' = mManualSurroundFormats
    // AUTO:   formats v 'enforced surround'
    // ALWAYS: formats v 'all surround' v 'enforced surround'
    // NEVER:  formats ^ 'non-surround'
    // MANUAL: formats ^ ('non-surround' v 'manual surround' v (IEC69137 ^ 'enforced surround'))

    std::unordered_set<audio_format_t> formatSet;
    if (forceUse == AUDIO_POLICY_FORCE_ENCODED_SURROUND_MANUAL
            || forceUse == AUDIO_POLICY_FORCE_ENCODED_SURROUND_NEVER) {
        // formatSet is (formats ^ 'non-surround')
        for (auto formatIter = formatsPtr->begin(); formatIter != formatsPtr->end(); ++formatIter) {
            if (allSurround.count(*formatIter) == 0 && enforcedSurround.count(*formatIter) == 0) {
                formatSet.insert(*formatIter);
            }
        }
    } else {
        formatSet.insert(formatsPtr->begin(), formatsPtr->end());
    }
    formatsPtr->clear();  // Re-filled from the formatSet at the end.

    if (forceUse == AUDIO_POLICY_FORCE_ENCODED_SURROUND_MANUAL) {
        formatSet.insert(mManualSurroundFormats.begin(), mManualSurroundFormats.end());
        // Enable IEC61937 when in MANUAL mode if it's enforced for this device.
        if (enforcedSurround.count(AUDIO_FORMAT_IEC61937) != 0) {
            formatSet.insert(AUDIO_FORMAT_IEC61937);
        }
    } else if (forceUse != AUDIO_POLICY_FORCE_ENCODED_SURROUND_NEVER) { // AUTO or ALWAYS
        if (forceUse == AUDIO_POLICY_FORCE_ENCODED_SURROUND_ALWAYS) {
            formatSet.insert(allSurround.begin(), allSurround.end());
        }
        formatSet.insert(enforcedSurround.begin(), enforcedSurround.end());
    }
    for (const auto& format : formatSet) {
        formatsPtr->push(format);
    }
}

void AudioPolicyManager::modifySurroundChannelMasks(ChannelsVector *channelMasksPtr) {
    ChannelsVector &channelMasks = *channelMasksPtr;
    audio_policy_forced_cfg_t forceUse = mEngine->getForceUse(
            AUDIO_POLICY_FORCE_FOR_ENCODED_SURROUND);

    // If NEVER, then remove support for channelMasks > stereo.
    if (forceUse == AUDIO_POLICY_FORCE_ENCODED_SURROUND_NEVER) {
        for (size_t maskIndex = 0; maskIndex < channelMasks.size(); ) {
            audio_channel_mask_t channelMask = channelMasks[maskIndex];
            if (channelMask & ~AUDIO_CHANNEL_OUT_STEREO) {
                ALOGI("%s: force NEVER, so remove channelMask 0x%08x", __FUNCTION__, channelMask);
                channelMasks.removeAt(maskIndex);
            } else {
                maskIndex++;
            }
        }
    // If ALWAYS or MANUAL, then make sure we at least support 5.1
    } else if (forceUse == AUDIO_POLICY_FORCE_ENCODED_SURROUND_ALWAYS
            || forceUse == AUDIO_POLICY_FORCE_ENCODED_SURROUND_MANUAL) {
        bool supports5dot1 = false;
        // Are there any channel masks that can be considered "surround"?
        for (audio_channel_mask_t channelMask : channelMasks) {
            if ((channelMask & AUDIO_CHANNEL_OUT_5POINT1) == AUDIO_CHANNEL_OUT_5POINT1) {
                supports5dot1 = true;
                break;
            }
        }
        // If not then add 5.1 support.
        if (!supports5dot1) {
            channelMasks.add(AUDIO_CHANNEL_OUT_5POINT1);
            ALOGI("%s: force MANUAL or ALWAYS, so adding channelMask for 5.1 surround", __func__);
        }
    }
}

void AudioPolicyManager::updateAudioProfiles(const sp<DeviceDescriptor>& devDesc,
                                             audio_io_handle_t ioHandle,
                                             AudioProfileVector &profiles)
{
    String8 reply;
    audio_devices_t device = devDesc->type();

    // Format MUST be checked first to update the list of AudioProfile
    if (profiles.hasDynamicFormat()) {
        reply = mpClientInterface->getParameters(
                ioHandle, String8(AudioParameter::keyStreamSupportedFormats));
        ALOGV("%s: supported formats %d, %s", __FUNCTION__, ioHandle, reply.string());
        AudioParameter repliedParameters(reply);
        if (repliedParameters.get(
                String8(AudioParameter::keyStreamSupportedFormats), reply) != NO_ERROR) {
            ALOGE("%s: failed to retrieve format, bailing out", __FUNCTION__);
            return;
        }
        FormatVector formats = formatsFromString(reply.string());
        if (device == AUDIO_DEVICE_OUT_HDMI
                || isDeviceOfModule(devDesc, AUDIO_HARDWARE_MODULE_ID_MSD)) {
            modifySurroundFormats(devDesc, &formats);
        }
        profiles.setFormats(formats);
    }

    for (audio_format_t format : profiles.getSupportedFormats()) {
        ChannelsVector channelMasks;
        SampleRateVector samplingRates;
        AudioParameter requestedParameters;
        requestedParameters.addInt(String8(AudioParameter::keyFormat), format);

        if (profiles.hasDynamicRateFor(format)) {
            reply = mpClientInterface->getParameters(
                    ioHandle,
                    requestedParameters.toString() + ";" +
                    AudioParameter::keyStreamSupportedSamplingRates);
            ALOGV("%s: supported sampling rates %s", __FUNCTION__, reply.string());
            AudioParameter repliedParameters(reply);
            if (repliedParameters.get(
                    String8(AudioParameter::keyStreamSupportedSamplingRates), reply) == NO_ERROR) {
                samplingRates = samplingRatesFromString(reply.string());
            }
        }
        if (profiles.hasDynamicChannelsFor(format)) {
            reply = mpClientInterface->getParameters(ioHandle,
                                                     requestedParameters.toString() + ";" +
                                                     AudioParameter::keyStreamSupportedChannels);
            ALOGV("%s: supported channel masks %s", __FUNCTION__, reply.string());
            AudioParameter repliedParameters(reply);
            if (repliedParameters.get(
                    String8(AudioParameter::keyStreamSupportedChannels), reply) == NO_ERROR) {
                channelMasks = channelMasksFromString(reply.string());
                if (device == AUDIO_DEVICE_OUT_HDMI
                        || isDeviceOfModule(devDesc, AUDIO_HARDWARE_MODULE_ID_MSD)) {
                    modifySurroundChannelMasks(&channelMasks);
                }
            }
        }
        profiles.addProfileFromHal(new AudioProfile(format, channelMasks, samplingRates));
    }
}

status_t AudioPolicyManager::installPatch(const char *caller,
                                          audio_patch_handle_t *patchHandle,
                                          AudioIODescriptorInterface *ioDescriptor,
                                          const struct audio_patch *patch,
                                          int delayMs)
{
    ssize_t index = mAudioPatches.indexOfKey(
            patchHandle && *patchHandle != AUDIO_PATCH_HANDLE_NONE ?
            *patchHandle : ioDescriptor->getPatchHandle());
    sp<AudioPatch> patchDesc;
    status_t status = installPatch(
            caller, index, patchHandle, patch, delayMs, mUidCached, &patchDesc);
    if (status == NO_ERROR) {
        ioDescriptor->setPatchHandle(patchDesc->mHandle);
    }
    return status;
}

status_t AudioPolicyManager::installPatch(const char *caller,
                                          ssize_t index,
                                          audio_patch_handle_t *patchHandle,
                                          const struct audio_patch *patch,
                                          int delayMs,
                                          uid_t uid,
                                          sp<AudioPatch> *patchDescPtr)
{
    sp<AudioPatch> patchDesc;
    audio_patch_handle_t afPatchHandle = AUDIO_PATCH_HANDLE_NONE;
    if (index >= 0) {
        patchDesc = mAudioPatches.valueAt(index);
        afPatchHandle = patchDesc->mAfPatchHandle;
    }

    status_t status = mpClientInterface->createAudioPatch(patch, &afPatchHandle, delayMs);
    ALOGV("%s() AF::createAudioPatch returned %d patchHandle %d num_sources %d num_sinks %d",
            caller, status, afPatchHandle, patch->num_sources, patch->num_sinks);
    if (status == NO_ERROR) {
        if (index < 0) {
            patchDesc = new AudioPatch(patch, uid);
            addAudioPatch(patchDesc->mHandle, patchDesc);
        } else {
            patchDesc->mPatch = *patch;
        }
        patchDesc->mAfPatchHandle = afPatchHandle;
        if (patchHandle) {
            *patchHandle = patchDesc->mHandle;
        }
        nextAudioPortGeneration();
        mpClientInterface->onAudioPatchListUpdate();
    }
    if (patchDescPtr) *patchDescPtr = patchDesc;
    return status;
}

} // namespace android