blob: 2e954d673404df6016920fb2b3d4dd9b8bba904f [file] [log] [blame]
/*
* Copyright (C) 2013-2014 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 "audio_hw_primary"
/*#define LOG_NDEBUG 0*/
/*#define VERY_VERY_VERBOSE_LOGGING*/
#ifdef VERY_VERY_VERBOSE_LOGGING
#define ALOGVV ALOGV
#else
#define ALOGVV(a...) do { } while(0)
#endif
#include <errno.h>
#include <pthread.h>
#include <stdint.h>
#include <sys/time.h>
#include <stdlib.h>
#include <math.h>
#include <dlfcn.h>
#include <sys/resource.h>
#include <sys/prctl.h>
#include <cutils/log.h>
#include <cutils/str_parms.h>
#include <cutils/properties.h>
#include <cutils/atomic.h>
#include <cutils/sched_policy.h>
#include <hardware/audio_effect.h>
#include <hardware/audio_alsaops.h>
#include <system/thread_defs.h>
#include <audio_effects/effect_aec.h>
#include <audio_effects/effect_ns.h>
#include "audio_hw.h"
#include "audio_extn.h"
#include "platform_api.h"
#include <platform.h>
#include "voice_extn.h"
#include "sound/compress_params.h"
#define COMPRESS_OFFLOAD_FRAGMENT_SIZE (256 * 1024)
// 2 buffers causes problems with high bitrate files
#define COMPRESS_OFFLOAD_NUM_FRAGMENTS 3
/* ToDo: Check and update a proper value in msec */
#define COMPRESS_OFFLOAD_PLAYBACK_LATENCY 96
#define COMPRESS_PLAYBACK_VOLUME_MAX 0x2000
#define PROXY_OPEN_RETRY_COUNT 100
#define PROXY_OPEN_WAIT_TIME 20
#define MIN_CHANNEL_COUNT 1
#define DEFAULT_CHANNEL_COUNT 2
#ifndef MAX_TARGET_SPECIFIC_CHANNEL_CNT
#define MAX_CHANNEL_COUNT 1
#else
#define MAX_CHANNEL_COUNT atoi(XSTR(MAX_TARGET_SPECIFIC_CHANNEL_CNT))
#define XSTR(x) STR(x)
#define STR(x) #x
#endif
static unsigned int configured_low_latency_capture_period_size =
LOW_LATENCY_CAPTURE_PERIOD_SIZE;
/* This constant enables extended precision handling.
* TODO The flag is off until more testing is done.
*/
static const bool k_enable_extended_precision = false;
struct pcm_config pcm_config_deep_buffer = {
.channels = DEFAULT_CHANNEL_COUNT,
.rate = DEFAULT_OUTPUT_SAMPLING_RATE,
.period_size = DEEP_BUFFER_OUTPUT_PERIOD_SIZE,
.period_count = DEEP_BUFFER_OUTPUT_PERIOD_COUNT,
.format = PCM_FORMAT_S16_LE,
.start_threshold = DEEP_BUFFER_OUTPUT_PERIOD_SIZE / 4,
.stop_threshold = INT_MAX,
.avail_min = DEEP_BUFFER_OUTPUT_PERIOD_SIZE / 4,
};
struct pcm_config pcm_config_low_latency = {
.channels = DEFAULT_CHANNEL_COUNT,
.rate = DEFAULT_OUTPUT_SAMPLING_RATE,
.period_size = LOW_LATENCY_OUTPUT_PERIOD_SIZE,
.period_count = LOW_LATENCY_OUTPUT_PERIOD_COUNT,
.format = PCM_FORMAT_S16_LE,
.start_threshold = LOW_LATENCY_OUTPUT_PERIOD_SIZE / 4,
.stop_threshold = INT_MAX,
.avail_min = LOW_LATENCY_OUTPUT_PERIOD_SIZE / 4,
};
struct pcm_config pcm_config_hdmi_multi = {
.channels = HDMI_MULTI_DEFAULT_CHANNEL_COUNT, /* changed when the stream is opened */
.rate = DEFAULT_OUTPUT_SAMPLING_RATE, /* changed when the stream is opened */
.period_size = HDMI_MULTI_PERIOD_SIZE,
.period_count = HDMI_MULTI_PERIOD_COUNT,
.format = PCM_FORMAT_S16_LE,
.start_threshold = 0,
.stop_threshold = INT_MAX,
.avail_min = 0,
};
struct pcm_config pcm_config_audio_capture = {
.channels = DEFAULT_CHANNEL_COUNT,
.period_count = AUDIO_CAPTURE_PERIOD_COUNT,
.format = PCM_FORMAT_S16_LE,
.stop_threshold = INT_MAX,
.avail_min = 0,
};
#define AFE_PROXY_CHANNEL_COUNT 2
#define AFE_PROXY_SAMPLING_RATE 48000
#define AFE_PROXY_PLAYBACK_PERIOD_SIZE 768
#define AFE_PROXY_PLAYBACK_PERIOD_COUNT 4
struct pcm_config pcm_config_afe_proxy_playback = {
.channels = AFE_PROXY_CHANNEL_COUNT,
.rate = AFE_PROXY_SAMPLING_RATE,
.period_size = AFE_PROXY_PLAYBACK_PERIOD_SIZE,
.period_count = AFE_PROXY_PLAYBACK_PERIOD_COUNT,
.format = PCM_FORMAT_S16_LE,
.start_threshold = AFE_PROXY_PLAYBACK_PERIOD_SIZE,
.stop_threshold = INT_MAX,
.avail_min = AFE_PROXY_PLAYBACK_PERIOD_SIZE,
};
#define AFE_PROXY_RECORD_PERIOD_SIZE 768
#define AFE_PROXY_RECORD_PERIOD_COUNT 4
struct pcm_config pcm_config_afe_proxy_record = {
.channels = AFE_PROXY_CHANNEL_COUNT,
.rate = AFE_PROXY_SAMPLING_RATE,
.period_size = AFE_PROXY_RECORD_PERIOD_SIZE,
.period_count = AFE_PROXY_RECORD_PERIOD_COUNT,
.format = PCM_FORMAT_S16_LE,
.start_threshold = AFE_PROXY_RECORD_PERIOD_SIZE,
.stop_threshold = INT_MAX,
.avail_min = AFE_PROXY_RECORD_PERIOD_SIZE,
};
const char * const use_case_table[AUDIO_USECASE_MAX] = {
[USECASE_AUDIO_PLAYBACK_DEEP_BUFFER] = "deep-buffer-playback",
[USECASE_AUDIO_PLAYBACK_LOW_LATENCY] = "low-latency-playback",
[USECASE_AUDIO_PLAYBACK_MULTI_CH] = "multi-channel-playback",
[USECASE_AUDIO_PLAYBACK_OFFLOAD] = "compress-offload-playback",
[USECASE_AUDIO_PLAYBACK_TTS] = "audio-tts-playback",
[USECASE_AUDIO_PLAYBACK_ULL] = "audio-ull-playback",
[USECASE_AUDIO_RECORD] = "audio-record",
[USECASE_AUDIO_RECORD_LOW_LATENCY] = "low-latency-record",
[USECASE_AUDIO_HFP_SCO] = "hfp-sco",
[USECASE_AUDIO_HFP_SCO_WB] = "hfp-sco-wb",
[USECASE_VOICE_CALL] = "voice-call",
[USECASE_VOICE2_CALL] = "voice2-call",
[USECASE_VOLTE_CALL] = "volte-call",
[USECASE_QCHAT_CALL] = "qchat-call",
[USECASE_VOWLAN_CALL] = "vowlan-call",
[USECASE_AUDIO_SPKR_CALIB_RX] = "spkr-rx-calib",
[USECASE_AUDIO_SPKR_CALIB_TX] = "spkr-vi-record",
[USECASE_AUDIO_PLAYBACK_AFE_PROXY] = "afe-proxy-playback",
[USECASE_AUDIO_RECORD_AFE_PROXY] = "afe-proxy-record",
};
#define STRING_TO_ENUM(string) { #string, string }
struct string_to_enum {
const char *name;
uint32_t value;
};
static const struct string_to_enum out_channels_name_to_enum_table[] = {
STRING_TO_ENUM(AUDIO_CHANNEL_OUT_STEREO),
STRING_TO_ENUM(AUDIO_CHANNEL_OUT_5POINT1),
STRING_TO_ENUM(AUDIO_CHANNEL_OUT_7POINT1),
};
static int set_voice_volume_l(struct audio_device *adev, float volume);
static struct audio_device *adev = NULL;
static pthread_mutex_t adev_init_lock;
static unsigned int audio_device_ref_count;
__attribute__ ((visibility ("default")))
bool audio_hw_send_gain_dep_calibration(int level) {
bool ret_val = false;
ALOGV("%s: enter ... ", __func__);
pthread_mutex_lock(&adev_init_lock);
if (adev != NULL && adev->platform != NULL) {
pthread_mutex_lock(&adev->lock);
ret_val = platform_send_gain_dep_cal(adev->platform, level);
pthread_mutex_unlock(&adev->lock);
} else {
ALOGE("%s: %s is NULL", __func__, adev == NULL ? "adev" : "adev->platform");
}
pthread_mutex_unlock(&adev_init_lock);
ALOGV("%s: exit with ret_val %d ", __func__, ret_val);
return ret_val;
}
static bool is_supported_format(audio_format_t format)
{
switch (format) {
case AUDIO_FORMAT_MP3:
case AUDIO_FORMAT_AAC_LC:
case AUDIO_FORMAT_AAC_HE_V1:
case AUDIO_FORMAT_AAC_HE_V2:
return true;
default:
break;
}
return false;
}
static int get_snd_codec_id(audio_format_t format)
{
int id = 0;
switch (format & AUDIO_FORMAT_MAIN_MASK) {
case AUDIO_FORMAT_MP3:
id = SND_AUDIOCODEC_MP3;
break;
case AUDIO_FORMAT_AAC:
id = SND_AUDIOCODEC_AAC;
break;
default:
ALOGE("%s: Unsupported audio format", __func__);
}
return id;
}
int enable_audio_route(struct audio_device *adev,
struct audio_usecase *usecase)
{
snd_device_t snd_device;
char mixer_path[50];
if (usecase == NULL)
return -EINVAL;
ALOGV("%s: enter: usecase(%d)", __func__, usecase->id);
if (usecase->type == PCM_CAPTURE)
snd_device = usecase->in_snd_device;
else
snd_device = usecase->out_snd_device;
strcpy(mixer_path, use_case_table[usecase->id]);
platform_add_backend_name(adev->platform, mixer_path, snd_device);
ALOGD("%s: apply and update mixer path: %s", __func__, mixer_path);
audio_route_apply_and_update_path(adev->audio_route, mixer_path);
ALOGV("%s: exit", __func__);
return 0;
}
int disable_audio_route(struct audio_device *adev,
struct audio_usecase *usecase)
{
snd_device_t snd_device;
char mixer_path[50];
if (usecase == NULL)
return -EINVAL;
ALOGV("%s: enter: usecase(%d)", __func__, usecase->id);
if (usecase->type == PCM_CAPTURE)
snd_device = usecase->in_snd_device;
else
snd_device = usecase->out_snd_device;
strcpy(mixer_path, use_case_table[usecase->id]);
platform_add_backend_name(adev->platform, mixer_path, snd_device);
ALOGD("%s: reset and update mixer path: %s", __func__, mixer_path);
audio_route_reset_and_update_path(adev->audio_route, mixer_path);
ALOGV("%s: exit", __func__);
return 0;
}
int enable_snd_device(struct audio_device *adev,
snd_device_t snd_device)
{
int i, num_devices = 0;
snd_device_t new_snd_devices[2];
if (snd_device < SND_DEVICE_MIN ||
snd_device >= SND_DEVICE_MAX) {
ALOGE("%s: Invalid sound device %d", __func__, snd_device);
return -EINVAL;
}
platform_send_audio_calibration(adev->platform, snd_device);
adev->snd_dev_ref_cnt[snd_device]++;
if (adev->snd_dev_ref_cnt[snd_device] > 1) {
ALOGV("%s: snd_device(%d: %s) is already active",
__func__, snd_device, platform_get_snd_device_name(snd_device));
return 0;
}
/* due to the possibility of calibration overwrite between listen
and audio, notify sound trigger hal before audio calibration is sent */
audio_extn_sound_trigger_update_device_status(snd_device,
ST_EVENT_SND_DEVICE_BUSY);
if (audio_extn_spkr_prot_is_enabled())
audio_extn_spkr_prot_calib_cancel(adev);
audio_extn_dsm_feedback_enable(adev, snd_device, true);
if ((snd_device == SND_DEVICE_OUT_SPEAKER ||
snd_device == SND_DEVICE_OUT_VOICE_SPEAKER) &&
audio_extn_spkr_prot_is_enabled()) {
if (audio_extn_spkr_prot_get_acdb_id(snd_device) < 0) {
adev->snd_dev_ref_cnt[snd_device]--;
return -EINVAL;
}
if (audio_extn_spkr_prot_start_processing(snd_device)) {
ALOGE("%s: spkr_start_processing failed", __func__);
return -EINVAL;
}
} else if (platform_can_split_snd_device(snd_device, &num_devices, new_snd_devices)) {
for (i = 0; i < num_devices; i++) {
enable_snd_device(adev, new_snd_devices[i]);
}
platform_set_speaker_gain_in_combo(adev, snd_device, true);
} else {
const char * dev_path = platform_get_snd_device_name(snd_device);
ALOGD("%s: snd_device(%d: %s)", __func__, snd_device, dev_path);
audio_route_apply_and_update_path(adev->audio_route, dev_path);
}
return 0;
}
int disable_snd_device(struct audio_device *adev,
snd_device_t snd_device)
{
int i, num_devices = 0;
snd_device_t new_snd_devices[2];
if (snd_device < SND_DEVICE_MIN ||
snd_device >= SND_DEVICE_MAX) {
ALOGE("%s: Invalid sound device %d", __func__, snd_device);
return -EINVAL;
}
if (adev->snd_dev_ref_cnt[snd_device] <= 0) {
ALOGE("%s: device ref cnt is already 0", __func__);
return -EINVAL;
}
adev->snd_dev_ref_cnt[snd_device]--;
if (adev->snd_dev_ref_cnt[snd_device] == 0) {
const char * dev_path = platform_get_snd_device_name(snd_device);
ALOGD("%s: snd_device(%d: %s)", __func__, snd_device, dev_path);
audio_extn_dsm_feedback_enable(adev, snd_device, false);
if ((snd_device == SND_DEVICE_OUT_SPEAKER ||
snd_device == SND_DEVICE_OUT_VOICE_SPEAKER) &&
audio_extn_spkr_prot_is_enabled()) {
audio_extn_spkr_prot_stop_processing(snd_device);
} else if (platform_can_split_snd_device(snd_device, &num_devices, new_snd_devices)) {
for (i = 0; i < num_devices; i++) {
disable_snd_device(adev, new_snd_devices[i]);
}
platform_set_speaker_gain_in_combo(adev, snd_device, false);
} else {
audio_route_reset_and_update_path(adev->audio_route, dev_path);
}
audio_extn_sound_trigger_update_device_status(snd_device,
ST_EVENT_SND_DEVICE_FREE);
}
return 0;
}
static void check_and_route_playback_usecases(struct audio_device *adev,
struct audio_usecase *uc_info,
snd_device_t snd_device)
{
struct listnode *node;
struct audio_usecase *usecase;
bool switch_device[AUDIO_USECASE_MAX];
int i, num_uc_to_switch = 0;
/*
* This function is to make sure that all the usecases that are active on
* the hardware codec backend are always routed to any one device that is
* handled by the hardware codec.
* For example, if low-latency and deep-buffer usecases are currently active
* on speaker and out_set_parameters(headset) is received on low-latency
* output, then we have to make sure deep-buffer is also switched to headset,
* because of the limitation that both the devices cannot be enabled
* at the same time as they share the same backend.
*/
/* Disable all the usecases on the shared backend other than the
specified usecase */
for (i = 0; i < AUDIO_USECASE_MAX; i++)
switch_device[i] = false;
list_for_each(node, &adev->usecase_list) {
usecase = node_to_item(node, struct audio_usecase, list);
if (usecase->type != PCM_CAPTURE &&
usecase != uc_info &&
usecase->out_snd_device != snd_device &&
usecase->devices & AUDIO_DEVICE_OUT_ALL_CODEC_BACKEND &&
platform_check_backends_match(snd_device, usecase->out_snd_device)) {
ALOGV("%s: Usecase (%s) is active on (%s) - disabling ..",
__func__, use_case_table[usecase->id],
platform_get_snd_device_name(usecase->out_snd_device));
disable_audio_route(adev, usecase);
switch_device[usecase->id] = true;
num_uc_to_switch++;
}
}
if (num_uc_to_switch) {
list_for_each(node, &adev->usecase_list) {
usecase = node_to_item(node, struct audio_usecase, list);
if (switch_device[usecase->id]) {
disable_snd_device(adev, usecase->out_snd_device);
}
}
list_for_each(node, &adev->usecase_list) {
usecase = node_to_item(node, struct audio_usecase, list);
if (switch_device[usecase->id]) {
enable_snd_device(adev, snd_device);
}
}
/* Re-route all the usecases on the shared backend other than the
specified usecase to new snd devices */
list_for_each(node, &adev->usecase_list) {
usecase = node_to_item(node, struct audio_usecase, list);
/* Update the out_snd_device only before enabling the audio route */
if (switch_device[usecase->id] ) {
usecase->out_snd_device = snd_device;
enable_audio_route(adev, usecase);
}
}
}
}
static void check_and_route_capture_usecases(struct audio_device *adev,
struct audio_usecase *uc_info,
snd_device_t snd_device)
{
struct listnode *node;
struct audio_usecase *usecase;
bool switch_device[AUDIO_USECASE_MAX];
int i, num_uc_to_switch = 0;
/*
* This function is to make sure that all the active capture usecases
* are always routed to the same input sound device.
* For example, if audio-record and voice-call usecases are currently
* active on speaker(rx) and speaker-mic (tx) and out_set_parameters(earpiece)
* is received for voice call then we have to make sure that audio-record
* usecase is also switched to earpiece i.e. voice-dmic-ef,
* because of the limitation that two devices cannot be enabled
* at the same time if they share the same backend.
*/
for (i = 0; i < AUDIO_USECASE_MAX; i++)
switch_device[i] = false;
list_for_each(node, &adev->usecase_list) {
usecase = node_to_item(node, struct audio_usecase, list);
if (usecase->type != PCM_PLAYBACK &&
usecase != uc_info &&
usecase->in_snd_device != snd_device &&
(usecase->id != USECASE_AUDIO_SPKR_CALIB_TX)) {
ALOGV("%s: Usecase (%s) is active on (%s) - disabling ..",
__func__, use_case_table[usecase->id],
platform_get_snd_device_name(usecase->in_snd_device));
disable_audio_route(adev, usecase);
switch_device[usecase->id] = true;
num_uc_to_switch++;
}
}
if (num_uc_to_switch) {
list_for_each(node, &adev->usecase_list) {
usecase = node_to_item(node, struct audio_usecase, list);
if (switch_device[usecase->id]) {
disable_snd_device(adev, usecase->in_snd_device);
}
}
list_for_each(node, &adev->usecase_list) {
usecase = node_to_item(node, struct audio_usecase, list);
if (switch_device[usecase->id]) {
enable_snd_device(adev, snd_device);
}
}
/* Re-route all the usecases on the shared backend other than the
specified usecase to new snd devices */
list_for_each(node, &adev->usecase_list) {
usecase = node_to_item(node, struct audio_usecase, list);
/* Update the in_snd_device only before enabling the audio route */
if (switch_device[usecase->id] ) {
usecase->in_snd_device = snd_device;
enable_audio_route(adev, usecase);
}
}
}
}
/* must be called with hw device mutex locked */
static int read_hdmi_channel_masks(struct stream_out *out)
{
int ret = 0;
int channels = platform_edid_get_max_channels(out->dev->platform);
switch (channels) {
/*
* Do not handle stereo output in Multi-channel cases
* Stereo case is handled in normal playback path
*/
case 6:
ALOGV("%s: HDMI supports 5.1", __func__);
out->supported_channel_masks[0] = AUDIO_CHANNEL_OUT_5POINT1;
break;
case 8:
ALOGV("%s: HDMI supports 5.1 and 7.1 channels", __func__);
out->supported_channel_masks[0] = AUDIO_CHANNEL_OUT_5POINT1;
out->supported_channel_masks[1] = AUDIO_CHANNEL_OUT_7POINT1;
break;
default:
ALOGE("HDMI does not support multi channel playback");
ret = -ENOSYS;
break;
}
return ret;
}
static audio_usecase_t get_voice_usecase_id_from_list(struct audio_device *adev)
{
struct audio_usecase *usecase;
struct listnode *node;
list_for_each(node, &adev->usecase_list) {
usecase = node_to_item(node, struct audio_usecase, list);
if (usecase->type == VOICE_CALL) {
ALOGV("%s: usecase id %d", __func__, usecase->id);
return usecase->id;
}
}
return USECASE_INVALID;
}
struct audio_usecase *get_usecase_from_list(struct audio_device *adev,
audio_usecase_t uc_id)
{
struct audio_usecase *usecase;
struct listnode *node;
list_for_each(node, &adev->usecase_list) {
usecase = node_to_item(node, struct audio_usecase, list);
if (usecase->id == uc_id)
return usecase;
}
return NULL;
}
int select_devices(struct audio_device *adev,
audio_usecase_t uc_id)
{
snd_device_t out_snd_device = SND_DEVICE_NONE;
snd_device_t in_snd_device = SND_DEVICE_NONE;
struct audio_usecase *usecase = NULL;
struct audio_usecase *vc_usecase = NULL;
struct audio_usecase *hfp_usecase = NULL;
audio_usecase_t hfp_ucid;
struct listnode *node;
int status = 0;
usecase = get_usecase_from_list(adev, uc_id);
if (usecase == NULL) {
ALOGE("%s: Could not find the usecase(%d)", __func__, uc_id);
return -EINVAL;
}
if ((usecase->type == VOICE_CALL) ||
(usecase->type == PCM_HFP_CALL)) {
out_snd_device = platform_get_output_snd_device(adev->platform,
usecase->stream.out->devices);
in_snd_device = platform_get_input_snd_device(adev->platform, usecase->stream.out->devices);
usecase->devices = usecase->stream.out->devices;
} else {
/*
* If the voice call is active, use the sound devices of voice call usecase
* so that it would not result any device switch. All the usecases will
* be switched to new device when select_devices() is called for voice call
* usecase. This is to avoid switching devices for voice call when
* check_and_route_playback_usecases() is called below.
*/
if (voice_is_in_call(adev)) {
vc_usecase = get_usecase_from_list(adev,
get_voice_usecase_id_from_list(adev));
if ((vc_usecase != NULL) &&
((vc_usecase->devices & AUDIO_DEVICE_OUT_ALL_CODEC_BACKEND) ||
(usecase->devices == AUDIO_DEVICE_IN_VOICE_CALL))) {
in_snd_device = vc_usecase->in_snd_device;
out_snd_device = vc_usecase->out_snd_device;
}
} else if (audio_extn_hfp_is_active(adev)) {
hfp_ucid = audio_extn_hfp_get_usecase();
hfp_usecase = get_usecase_from_list(adev, hfp_ucid);
if (hfp_usecase->devices & AUDIO_DEVICE_OUT_ALL_CODEC_BACKEND) {
in_snd_device = hfp_usecase->in_snd_device;
out_snd_device = hfp_usecase->out_snd_device;
}
}
if (usecase->type == PCM_PLAYBACK) {
usecase->devices = usecase->stream.out->devices;
in_snd_device = SND_DEVICE_NONE;
if (out_snd_device == SND_DEVICE_NONE) {
out_snd_device = platform_get_output_snd_device(adev->platform,
usecase->stream.out->devices);
if (usecase->stream.out == adev->primary_output &&
adev->active_input &&
(adev->active_input->source == AUDIO_SOURCE_VOICE_COMMUNICATION ||
adev->mode == AUDIO_MODE_IN_COMMUNICATION) &&
out_snd_device != usecase->out_snd_device) {
select_devices(adev, adev->active_input->usecase);
}
}
} else if (usecase->type == PCM_CAPTURE) {
usecase->devices = usecase->stream.in->device;
out_snd_device = SND_DEVICE_NONE;
if (in_snd_device == SND_DEVICE_NONE) {
audio_devices_t out_device = AUDIO_DEVICE_NONE;
if (adev->active_input &&
(adev->active_input->source == AUDIO_SOURCE_VOICE_COMMUNICATION ||
adev->mode == AUDIO_MODE_IN_COMMUNICATION)) {
platform_set_echo_reference(adev, false, AUDIO_DEVICE_NONE);
if (usecase->id == USECASE_AUDIO_RECORD_AFE_PROXY) {
out_device = AUDIO_DEVICE_OUT_TELEPHONY_TX;
} else if (adev->primary_output) {
out_device = adev->primary_output->devices;
}
}
in_snd_device = platform_get_input_snd_device(adev->platform, out_device);
}
}
}
if (out_snd_device == usecase->out_snd_device &&
in_snd_device == usecase->in_snd_device) {
return 0;
}
ALOGD("%s: out_snd_device(%d: %s) in_snd_device(%d: %s)", __func__,
out_snd_device, platform_get_snd_device_name(out_snd_device),
in_snd_device, platform_get_snd_device_name(in_snd_device));
/*
* Limitation: While in call, to do a device switch we need to disable
* and enable both RX and TX devices though one of them is same as current
* device.
*/
if ((usecase->type == VOICE_CALL) &&
(usecase->in_snd_device != SND_DEVICE_NONE) &&
(usecase->out_snd_device != SND_DEVICE_NONE)) {
status = platform_switch_voice_call_device_pre(adev->platform);
/* Disable sidetone only if voice call already exists */
if (voice_is_call_state_active(adev))
voice_set_sidetone(adev, usecase->out_snd_device, false);
}
/* Disable current sound devices */
if (usecase->out_snd_device != SND_DEVICE_NONE) {
disable_audio_route(adev, usecase);
disable_snd_device(adev, usecase->out_snd_device);
}
if (usecase->in_snd_device != SND_DEVICE_NONE) {
disable_audio_route(adev, usecase);
disable_snd_device(adev, usecase->in_snd_device);
}
/* Applicable only on the targets that has external modem.
* New device information should be sent to modem before enabling
* the devices to reduce in-call device switch time.
*/
if ((usecase->type == VOICE_CALL) &&
(usecase->in_snd_device != SND_DEVICE_NONE) &&
(usecase->out_snd_device != SND_DEVICE_NONE)) {
status = platform_switch_voice_call_enable_device_config(adev->platform,
out_snd_device,
in_snd_device);
}
/* Enable new sound devices */
if (out_snd_device != SND_DEVICE_NONE) {
if (usecase->devices & AUDIO_DEVICE_OUT_ALL_CODEC_BACKEND)
check_and_route_playback_usecases(adev, usecase, out_snd_device);
enable_snd_device(adev, out_snd_device);
}
if (in_snd_device != SND_DEVICE_NONE) {
check_and_route_capture_usecases(adev, usecase, in_snd_device);
enable_snd_device(adev, in_snd_device);
}
if (usecase->type == VOICE_CALL)
status = platform_switch_voice_call_device_post(adev->platform,
out_snd_device,
in_snd_device);
usecase->in_snd_device = in_snd_device;
usecase->out_snd_device = out_snd_device;
enable_audio_route(adev, usecase);
/* Applicable only on the targets that has external modem.
* Enable device command should be sent to modem only after
* enabling voice call mixer controls
*/
if (usecase->type == VOICE_CALL) {
status = platform_switch_voice_call_usecase_route_post(adev->platform,
out_snd_device,
in_snd_device);
/* Enable sidetone only if voice call already exists */
if (voice_is_call_state_active(adev))
voice_set_sidetone(adev, out_snd_device, true);
}
return status;
}
static int stop_input_stream(struct stream_in *in)
{
int i, ret = 0;
struct audio_usecase *uc_info;
struct audio_device *adev = in->dev;
adev->active_input = NULL;
ALOGV("%s: enter: usecase(%d: %s)", __func__,
in->usecase, use_case_table[in->usecase]);
uc_info = get_usecase_from_list(adev, in->usecase);
if (uc_info == NULL) {
ALOGE("%s: Could not find the usecase (%d) in the list",
__func__, in->usecase);
return -EINVAL;
}
/* 1. Disable stream specific mixer controls */
disable_audio_route(adev, uc_info);
/* 2. Disable the tx device */
disable_snd_device(adev, uc_info->in_snd_device);
list_remove(&uc_info->list);
free(uc_info);
ALOGV("%s: exit: status(%d)", __func__, ret);
return ret;
}
int start_input_stream(struct stream_in *in)
{
/* 1. Enable output device and stream routing controls */
int ret = 0;
struct audio_usecase *uc_info;
struct audio_device *adev = in->dev;
ALOGV("%s: enter: usecase(%d)", __func__, in->usecase);
in->pcm_device_id = platform_get_pcm_device_id(in->usecase, PCM_CAPTURE);
if (in->pcm_device_id < 0) {
ALOGE("%s: Could not find PCM device id for the usecase(%d)",
__func__, in->usecase);
ret = -EINVAL;
goto error_config;
}
adev->active_input = in;
uc_info = (struct audio_usecase *)calloc(1, sizeof(struct audio_usecase));
uc_info->id = in->usecase;
uc_info->type = PCM_CAPTURE;
uc_info->stream.in = in;
uc_info->devices = in->device;
uc_info->in_snd_device = SND_DEVICE_NONE;
uc_info->out_snd_device = SND_DEVICE_NONE;
list_add_tail(&adev->usecase_list, &uc_info->list);
audio_extn_perf_lock_acquire();
select_devices(adev, in->usecase);
ALOGV("%s: Opening PCM device card_id(%d) device_id(%d), channels %d",
__func__, adev->snd_card, in->pcm_device_id, in->config.channels);
unsigned int flags = PCM_IN;
unsigned int pcm_open_retry_count = 0;
if (in->usecase == USECASE_AUDIO_RECORD_AFE_PROXY) {
flags |= PCM_MMAP | PCM_NOIRQ;
pcm_open_retry_count = PROXY_OPEN_RETRY_COUNT;
}
while (1) {
in->pcm = pcm_open(adev->snd_card, in->pcm_device_id,
flags, &in->config);
if (in->pcm == NULL || !pcm_is_ready(in->pcm)) {
ALOGE("%s: %s", __func__, pcm_get_error(in->pcm));
if (in->pcm != NULL) {
pcm_close(in->pcm);
in->pcm = NULL;
}
if (pcm_open_retry_count-- == 0) {
ret = -EIO;
goto error_open;
}
usleep(PROXY_OPEN_WAIT_TIME * 1000);
continue;
}
break;
}
ALOGV("%s: pcm_prepare start", __func__);
pcm_prepare(in->pcm);
audio_extn_perf_lock_release();
ALOGV("%s: exit", __func__);
return ret;
error_open:
stop_input_stream(in);
audio_extn_perf_lock_release();
error_config:
adev->active_input = NULL;
ALOGD("%s: exit: status(%d)", __func__, ret);
return ret;
}
void lock_input_stream(struct stream_in *in)
{
pthread_mutex_lock(&in->pre_lock);
pthread_mutex_lock(&in->lock);
pthread_mutex_unlock(&in->pre_lock);
}
void lock_output_stream(struct stream_out *out)
{
pthread_mutex_lock(&out->pre_lock);
pthread_mutex_lock(&out->lock);
pthread_mutex_unlock(&out->pre_lock);
}
/* must be called with out->lock locked */
static int send_offload_cmd_l(struct stream_out* out, int command)
{
struct offload_cmd *cmd = (struct offload_cmd *)calloc(1, sizeof(struct offload_cmd));
ALOGVV("%s %d", __func__, command);
cmd->cmd = command;
list_add_tail(&out->offload_cmd_list, &cmd->node);
pthread_cond_signal(&out->offload_cond);
return 0;
}
/* must be called iwth out->lock locked */
static void stop_compressed_output_l(struct stream_out *out)
{
out->offload_state = OFFLOAD_STATE_IDLE;
out->playback_started = 0;
out->send_new_metadata = 1;
if (out->compr != NULL) {
compress_stop(out->compr);
while (out->offload_thread_blocked) {
pthread_cond_wait(&out->cond, &out->lock);
}
}
}
static void *offload_thread_loop(void *context)
{
struct stream_out *out = (struct stream_out *) context;
struct listnode *item;
out->offload_state = OFFLOAD_STATE_IDLE;
out->playback_started = 0;
setpriority(PRIO_PROCESS, 0, ANDROID_PRIORITY_AUDIO);
set_sched_policy(0, SP_FOREGROUND);
prctl(PR_SET_NAME, (unsigned long)"Offload Callback", 0, 0, 0);
ALOGV("%s", __func__);
lock_output_stream(out);
for (;;) {
struct offload_cmd *cmd = NULL;
stream_callback_event_t event;
bool send_callback = false;
ALOGVV("%s offload_cmd_list %d out->offload_state %d",
__func__, list_empty(&out->offload_cmd_list),
out->offload_state);
if (list_empty(&out->offload_cmd_list)) {
ALOGV("%s SLEEPING", __func__);
pthread_cond_wait(&out->offload_cond, &out->lock);
ALOGV("%s RUNNING", __func__);
continue;
}
item = list_head(&out->offload_cmd_list);
cmd = node_to_item(item, struct offload_cmd, node);
list_remove(item);
ALOGVV("%s STATE %d CMD %d out->compr %p",
__func__, out->offload_state, cmd->cmd, out->compr);
if (cmd->cmd == OFFLOAD_CMD_EXIT) {
free(cmd);
break;
}
if (out->compr == NULL) {
ALOGE("%s: Compress handle is NULL", __func__);
pthread_cond_signal(&out->cond);
continue;
}
out->offload_thread_blocked = true;
pthread_mutex_unlock(&out->lock);
send_callback = false;
switch(cmd->cmd) {
case OFFLOAD_CMD_WAIT_FOR_BUFFER:
compress_wait(out->compr, -1);
send_callback = true;
event = STREAM_CBK_EVENT_WRITE_READY;
break;
case OFFLOAD_CMD_PARTIAL_DRAIN:
compress_next_track(out->compr);
compress_partial_drain(out->compr);
send_callback = true;
event = STREAM_CBK_EVENT_DRAIN_READY;
/* Resend the metadata for next iteration */
out->send_new_metadata = 1;
break;
case OFFLOAD_CMD_DRAIN:
compress_drain(out->compr);
send_callback = true;
event = STREAM_CBK_EVENT_DRAIN_READY;
break;
default:
ALOGE("%s unknown command received: %d", __func__, cmd->cmd);
break;
}
lock_output_stream(out);
out->offload_thread_blocked = false;
pthread_cond_signal(&out->cond);
if (send_callback) {
ALOGVV("%s: sending offload_callback event %d", __func__, event);
out->offload_callback(event, NULL, out->offload_cookie);
}
free(cmd);
}
pthread_cond_signal(&out->cond);
while (!list_empty(&out->offload_cmd_list)) {
item = list_head(&out->offload_cmd_list);
list_remove(item);
free(node_to_item(item, struct offload_cmd, node));
}
pthread_mutex_unlock(&out->lock);
return NULL;
}
static int create_offload_callback_thread(struct stream_out *out)
{
pthread_cond_init(&out->offload_cond, (const pthread_condattr_t *) NULL);
list_init(&out->offload_cmd_list);
pthread_create(&out->offload_thread, (const pthread_attr_t *) NULL,
offload_thread_loop, out);
return 0;
}
static int destroy_offload_callback_thread(struct stream_out *out)
{
lock_output_stream(out);
stop_compressed_output_l(out);
send_offload_cmd_l(out, OFFLOAD_CMD_EXIT);
pthread_mutex_unlock(&out->lock);
pthread_join(out->offload_thread, (void **) NULL);
pthread_cond_destroy(&out->offload_cond);
return 0;
}
static bool allow_hdmi_channel_config(struct audio_device *adev)
{
struct listnode *node;
struct audio_usecase *usecase;
bool ret = true;
list_for_each(node, &adev->usecase_list) {
usecase = node_to_item(node, struct audio_usecase, list);
if (usecase->devices & AUDIO_DEVICE_OUT_AUX_DIGITAL) {
/*
* If voice call is already existing, do not proceed further to avoid
* disabling/enabling both RX and TX devices, CSD calls, etc.
* Once the voice call done, the HDMI channels can be configured to
* max channels of remaining use cases.
*/
if (usecase->id == USECASE_VOICE_CALL) {
ALOGD("%s: voice call is active, no change in HDMI channels",
__func__);
ret = false;
break;
} else if (usecase->id == USECASE_AUDIO_PLAYBACK_MULTI_CH) {
ALOGD("%s: multi channel playback is active, "
"no change in HDMI channels", __func__);
ret = false;
break;
}
}
}
return ret;
}
static int check_and_set_hdmi_channels(struct audio_device *adev,
unsigned int channels)
{
struct listnode *node;
struct audio_usecase *usecase;
/* Check if change in HDMI channel config is allowed */
if (!allow_hdmi_channel_config(adev))
return 0;
if (channels == adev->cur_hdmi_channels) {
ALOGD("%s: Requested channels are same as current", __func__);
return 0;
}
platform_set_hdmi_channels(adev->platform, channels);
adev->cur_hdmi_channels = channels;
/*
* Deroute all the playback streams routed to HDMI so that
* the back end is deactivated. Note that backend will not
* be deactivated if any one stream is connected to it.
*/
list_for_each(node, &adev->usecase_list) {
usecase = node_to_item(node, struct audio_usecase, list);
if (usecase->type == PCM_PLAYBACK &&
usecase->devices & AUDIO_DEVICE_OUT_AUX_DIGITAL) {
disable_audio_route(adev, usecase);
}
}
/*
* Enable all the streams disabled above. Now the HDMI backend
* will be activated with new channel configuration
*/
list_for_each(node, &adev->usecase_list) {
usecase = node_to_item(node, struct audio_usecase, list);
if (usecase->type == PCM_PLAYBACK &&
usecase->devices & AUDIO_DEVICE_OUT_AUX_DIGITAL) {
enable_audio_route(adev, usecase);
}
}
return 0;
}
static int stop_output_stream(struct stream_out *out)
{
int i, ret = 0;
struct audio_usecase *uc_info;
struct audio_device *adev = out->dev;
ALOGV("%s: enter: usecase(%d: %s)", __func__,
out->usecase, use_case_table[out->usecase]);
uc_info = get_usecase_from_list(adev, out->usecase);
if (uc_info == NULL) {
ALOGE("%s: Could not find the usecase (%d) in the list",
__func__, out->usecase);
return -EINVAL;
}
if (out->usecase == USECASE_AUDIO_PLAYBACK_OFFLOAD) {
if (adev->visualizer_stop_output != NULL)
adev->visualizer_stop_output(out->handle, out->pcm_device_id);
if (adev->offload_effects_stop_output != NULL)
adev->offload_effects_stop_output(out->handle, out->pcm_device_id);
}
/* 1. Get and set stream specific mixer controls */
disable_audio_route(adev, uc_info);
/* 2. Disable the rx device */
disable_snd_device(adev, uc_info->out_snd_device);
list_remove(&uc_info->list);
free(uc_info);
audio_extn_extspk_update(adev->extspk);
/* Must be called after removing the usecase from list */
if (out->devices & AUDIO_DEVICE_OUT_AUX_DIGITAL)
check_and_set_hdmi_channels(adev, DEFAULT_HDMI_OUT_CHANNELS);
ALOGV("%s: exit: status(%d)", __func__, ret);
return ret;
}
int start_output_stream(struct stream_out *out)
{
int ret = 0;
struct audio_usecase *uc_info;
struct audio_device *adev = out->dev;
ALOGV("%s: enter: usecase(%d: %s) devices(%#x)",
__func__, out->usecase, use_case_table[out->usecase], out->devices);
out->pcm_device_id = platform_get_pcm_device_id(out->usecase, PCM_PLAYBACK);
if (out->pcm_device_id < 0) {
ALOGE("%s: Invalid PCM device id(%d) for the usecase(%d)",
__func__, out->pcm_device_id, out->usecase);
ret = -EINVAL;
goto error_config;
}
uc_info = (struct audio_usecase *)calloc(1, sizeof(struct audio_usecase));
uc_info->id = out->usecase;
uc_info->type = PCM_PLAYBACK;
uc_info->stream.out = out;
uc_info->devices = out->devices;
uc_info->in_snd_device = SND_DEVICE_NONE;
uc_info->out_snd_device = SND_DEVICE_NONE;
/* This must be called before adding this usecase to the list */
if (out->devices & AUDIO_DEVICE_OUT_AUX_DIGITAL)
check_and_set_hdmi_channels(adev, out->config.channels);
list_add_tail(&adev->usecase_list, &uc_info->list);
audio_extn_perf_lock_acquire();
select_devices(adev, out->usecase);
audio_extn_extspk_update(adev->extspk);
ALOGV("%s: Opening PCM device card_id(%d) device_id(%d) format(%#x)",
__func__, adev->snd_card, out->pcm_device_id, out->config.format);
if (out->usecase != USECASE_AUDIO_PLAYBACK_OFFLOAD) {
unsigned int flags = PCM_OUT;
unsigned int pcm_open_retry_count = 0;
if (out->usecase == USECASE_AUDIO_PLAYBACK_AFE_PROXY) {
flags |= PCM_MMAP | PCM_NOIRQ;
pcm_open_retry_count = PROXY_OPEN_RETRY_COUNT;
} else
flags |= PCM_MONOTONIC;
while (1) {
out->pcm = pcm_open(adev->snd_card, out->pcm_device_id,
flags, &out->config);
if (out->pcm == NULL || !pcm_is_ready(out->pcm)) {
ALOGE("%s: %s", __func__, pcm_get_error(out->pcm));
if (out->pcm != NULL) {
pcm_close(out->pcm);
out->pcm = NULL;
}
if (pcm_open_retry_count-- == 0) {
ret = -EIO;
goto error_open;
}
usleep(PROXY_OPEN_WAIT_TIME * 1000);
continue;
}
break;
}
ALOGV("%s: pcm_prepare start", __func__);
if (pcm_is_ready(out->pcm))
pcm_prepare(out->pcm);
} else {
out->pcm = NULL;
out->compr = compress_open(adev->snd_card, out->pcm_device_id,
COMPRESS_IN, &out->compr_config);
if (out->compr && !is_compress_ready(out->compr)) {
ALOGE("%s: %s", __func__, compress_get_error(out->compr));
compress_close(out->compr);
out->compr = NULL;
ret = -EIO;
goto error_open;
}
if (out->offload_callback)
compress_nonblock(out->compr, out->non_blocking);
if (adev->visualizer_start_output != NULL)
adev->visualizer_start_output(out->handle, out->pcm_device_id);
if (adev->offload_effects_start_output != NULL)
adev->offload_effects_start_output(out->handle, out->pcm_device_id);
}
audio_extn_perf_lock_release();
ALOGV("%s: exit", __func__);
return 0;
error_open:
audio_extn_perf_lock_release();
stop_output_stream(out);
error_config:
return ret;
}
static int check_input_parameters(uint32_t sample_rate,
audio_format_t format,
int channel_count)
{
if (format != AUDIO_FORMAT_PCM_16_BIT) {
ALOGE("%s: unsupported AUDIO FORMAT (%d) ", __func__, format);
return -EINVAL;
}
if ((channel_count < MIN_CHANNEL_COUNT) || (channel_count > MAX_CHANNEL_COUNT)) {
ALOGE("%s: unsupported channel count (%d) passed Min / Max (%d / %d)", __func__,
channel_count, MIN_CHANNEL_COUNT, MAX_CHANNEL_COUNT);
return -EINVAL;
}
switch (sample_rate) {
case 8000:
case 11025:
case 12000:
case 16000:
case 22050:
case 24000:
case 32000:
case 44100:
case 48000:
break;
default:
ALOGE("%s: unsupported (%d) samplerate passed ", __func__, sample_rate);
return -EINVAL;
}
return 0;
}
static size_t get_input_buffer_size(uint32_t sample_rate,
audio_format_t format,
int channel_count,
bool is_low_latency)
{
size_t size = 0;
if (check_input_parameters(sample_rate, format, channel_count) != 0)
return 0;
size = (sample_rate * AUDIO_CAPTURE_PERIOD_DURATION_MSEC) / 1000;
if (is_low_latency)
size = configured_low_latency_capture_period_size;
/* ToDo: should use frame_size computed based on the format and
channel_count here. */
size *= sizeof(short) * channel_count;
/* make sure the size is multiple of 32 bytes
* At 48 kHz mono 16-bit PCM:
* 5.000 ms = 240 frames = 15*16*1*2 = 480, a whole multiple of 32 (15)
* 3.333 ms = 160 frames = 10*16*1*2 = 320, a whole multiple of 32 (10)
*/
size += 0x1f;
size &= ~0x1f;
return size;
}
static uint32_t out_get_sample_rate(const struct audio_stream *stream)
{
struct stream_out *out = (struct stream_out *)stream;
return out->sample_rate;
}
static int out_set_sample_rate(struct audio_stream *stream __unused, uint32_t rate __unused)
{
return -ENOSYS;
}
static size_t out_get_buffer_size(const struct audio_stream *stream)
{
struct stream_out *out = (struct stream_out *)stream;
if (out->usecase == USECASE_AUDIO_PLAYBACK_OFFLOAD) {
return out->compr_config.fragment_size;
}
return out->config.period_size *
audio_stream_out_frame_size((const struct audio_stream_out *)stream);
}
static uint32_t out_get_channels(const struct audio_stream *stream)
{
struct stream_out *out = (struct stream_out *)stream;
return out->channel_mask;
}
static audio_format_t out_get_format(const struct audio_stream *stream)
{
struct stream_out *out = (struct stream_out *)stream;
return out->format;
}
static int out_set_format(struct audio_stream *stream __unused, audio_format_t format __unused)
{
return -ENOSYS;
}
static int out_standby(struct audio_stream *stream)
{
struct stream_out *out = (struct stream_out *)stream;
struct audio_device *adev = out->dev;
ALOGV("%s: enter: usecase(%d: %s)", __func__,
out->usecase, use_case_table[out->usecase]);
lock_output_stream(out);
if (!out->standby) {
if (adev->adm_deregister_stream)
adev->adm_deregister_stream(adev->adm_data, out->handle);
pthread_mutex_lock(&adev->lock);
out->standby = true;
if (out->usecase != USECASE_AUDIO_PLAYBACK_OFFLOAD) {
if (out->pcm) {
pcm_close(out->pcm);
out->pcm = NULL;
}
} else {
stop_compressed_output_l(out);
out->gapless_mdata.encoder_delay = 0;
out->gapless_mdata.encoder_padding = 0;
if (out->compr != NULL) {
compress_close(out->compr);
out->compr = NULL;
}
}
stop_output_stream(out);
pthread_mutex_unlock(&adev->lock);
}
pthread_mutex_unlock(&out->lock);
ALOGV("%s: exit", __func__);
return 0;
}
static int out_dump(const struct audio_stream *stream __unused, int fd __unused)
{
return 0;
}
static int parse_compress_metadata(struct stream_out *out, struct str_parms *parms)
{
int ret = 0;
char value[32];
struct compr_gapless_mdata tmp_mdata;
if (!out || !parms) {
return -EINVAL;
}
ret = str_parms_get_str(parms, AUDIO_OFFLOAD_CODEC_DELAY_SAMPLES, value, sizeof(value));
if (ret >= 0) {
tmp_mdata.encoder_delay = atoi(value); //whats a good limit check?
} else {
return -EINVAL;
}
ret = str_parms_get_str(parms, AUDIO_OFFLOAD_CODEC_PADDING_SAMPLES, value, sizeof(value));
if (ret >= 0) {
tmp_mdata.encoder_padding = atoi(value);
} else {
return -EINVAL;
}
out->gapless_mdata = tmp_mdata;
out->send_new_metadata = 1;
ALOGV("%s new encoder delay %u and padding %u", __func__,
out->gapless_mdata.encoder_delay, out->gapless_mdata.encoder_padding);
return 0;
}
static bool output_drives_call(struct audio_device *adev, struct stream_out *out)
{
return out == adev->primary_output || out == adev->voice_tx_output;
}
static int out_set_parameters(struct audio_stream *stream, const char *kvpairs)
{
struct stream_out *out = (struct stream_out *)stream;
struct audio_device *adev = out->dev;
struct audio_usecase *usecase;
struct listnode *node;
struct str_parms *parms;
char value[32];
int ret, val = 0;
bool select_new_device = false;
int status = 0;
ALOGD("%s: enter: usecase(%d: %s) kvpairs: %s",
__func__, out->usecase, use_case_table[out->usecase], kvpairs);
parms = str_parms_create_str(kvpairs);
ret = str_parms_get_str(parms, AUDIO_PARAMETER_STREAM_ROUTING, value, sizeof(value));
if (ret >= 0) {
val = atoi(value);
lock_output_stream(out);
pthread_mutex_lock(&adev->lock);
/*
* When HDMI cable is unplugged the music playback is paused and
* the policy manager sends routing=0. But the audioflinger
* continues to write data until standby time (3sec).
* As the HDMI core is turned off, the write gets blocked.
* Avoid this by routing audio to speaker until standby.
*/
if (out->devices == AUDIO_DEVICE_OUT_AUX_DIGITAL &&
val == AUDIO_DEVICE_NONE) {
val = AUDIO_DEVICE_OUT_SPEAKER;
}
/*
* select_devices() call below switches all the usecases on the same
* backend to the new device. Refer to check_and_route_playback_usecases() in
* the select_devices(). But how do we undo this?
*
* For example, music playback is active on headset (deep-buffer usecase)
* and if we go to ringtones and select a ringtone, low-latency usecase
* will be started on headset+speaker. As we can't enable headset+speaker
* and headset devices at the same time, select_devices() switches the music
* playback to headset+speaker while starting low-lateny usecase for ringtone.
* So when the ringtone playback is completed, how do we undo the same?
*
* We are relying on the out_set_parameters() call on deep-buffer output,
* once the ringtone playback is ended.
* NOTE: We should not check if the current devices are same as new devices.
* Because select_devices() must be called to switch back the music
* playback to headset.
*/
if (val != 0) {
out->devices = val;
if (!out->standby)
select_devices(adev, out->usecase);
if (output_drives_call(adev, out)) {
if (!voice_is_in_call(adev)) {
if (adev->mode == AUDIO_MODE_IN_CALL) {
adev->current_call_output = out;
ret = voice_start_call(adev);
}
} else {
adev->current_call_output = out;
voice_update_devices_for_all_voice_usecases(adev);
}
}
}
pthread_mutex_unlock(&adev->lock);
pthread_mutex_unlock(&out->lock);
/*handles device and call state changes*/
audio_extn_extspk_update(adev->extspk);
}
if (out->usecase == USECASE_AUDIO_PLAYBACK_OFFLOAD) {
parse_compress_metadata(out, parms);
}
str_parms_destroy(parms);
ALOGV("%s: exit: code(%d)", __func__, status);
return status;
}
static char* out_get_parameters(const struct audio_stream *stream, const char *keys)
{
struct stream_out *out = (struct stream_out *)stream;
struct str_parms *query = str_parms_create_str(keys);
char *str;
char value[256];
struct str_parms *reply = str_parms_create();
size_t i, j;
int ret;
bool first = true;
ALOGV("%s: enter: keys - %s", __func__, keys);
ret = str_parms_get_str(query, AUDIO_PARAMETER_STREAM_SUP_CHANNELS, value, sizeof(value));
if (ret >= 0) {
value[0] = '\0';
i = 0;
while (out->supported_channel_masks[i] != 0) {
for (j = 0; j < ARRAY_SIZE(out_channels_name_to_enum_table); j++) {
if (out_channels_name_to_enum_table[j].value == out->supported_channel_masks[i]) {
if (!first) {
strcat(value, "|");
}
strcat(value, out_channels_name_to_enum_table[j].name);
first = false;
break;
}
}
i++;
}
str_parms_add_str(reply, AUDIO_PARAMETER_STREAM_SUP_CHANNELS, value);
str = str_parms_to_str(reply);
} else {
str = strdup(keys);
}
str_parms_destroy(query);
str_parms_destroy(reply);
ALOGV("%s: exit: returns - %s", __func__, str);
return str;
}
static uint32_t out_get_latency(const struct audio_stream_out *stream)
{
struct stream_out *out = (struct stream_out *)stream;
if (out->usecase == USECASE_AUDIO_PLAYBACK_OFFLOAD)
return COMPRESS_OFFLOAD_PLAYBACK_LATENCY;
return (out->config.period_count * out->config.period_size * 1000) /
(out->config.rate);
}
static int out_set_volume(struct audio_stream_out *stream, float left,
float right)
{
struct stream_out *out = (struct stream_out *)stream;
int volume[2];
if (out->usecase == USECASE_AUDIO_PLAYBACK_MULTI_CH) {
/* only take left channel into account: the API is for stereo anyway */
out->muted = (left == 0.0f);
return 0;
} else if (out->usecase == USECASE_AUDIO_PLAYBACK_OFFLOAD) {
const char *mixer_ctl_name = "Compress Playback Volume";
struct audio_device *adev = out->dev;
struct mixer_ctl *ctl;
ctl = mixer_get_ctl_by_name(adev->mixer, mixer_ctl_name);
if (!ctl) {
/* try with the control based on device id */
int pcm_device_id = platform_get_pcm_device_id(out->usecase,
PCM_PLAYBACK);
char ctl_name[128] = {0};
snprintf(ctl_name, sizeof(ctl_name),
"Compress Playback %d Volume", pcm_device_id);
ctl = mixer_get_ctl_by_name(adev->mixer, ctl_name);
if (!ctl) {
ALOGE("%s: Could not get volume ctl mixer cmd", __func__);
return -EINVAL;
}
}
volume[0] = (int)(left * COMPRESS_PLAYBACK_VOLUME_MAX);
volume[1] = (int)(right * COMPRESS_PLAYBACK_VOLUME_MAX);
mixer_ctl_set_array(ctl, volume, sizeof(volume)/sizeof(volume[0]));
return 0;
}
return -ENOSYS;
}
#ifdef NO_AUDIO_OUT
static ssize_t out_write_for_no_output(struct audio_stream_out *stream,
const void *buffer, size_t bytes)
{
struct stream_out *out = (struct stream_out *)stream;
/* No Output device supported other than BT for playback.
* Sleep for the amount of buffer duration
*/
lock_output_stream(out);
usleep(bytes * 1000000 / audio_stream_frame_size(&out->stream.common) /
out_get_sample_rate(&out->stream.common));
pthread_mutex_unlock(&out->lock);
return bytes;
}
#endif
static ssize_t out_write(struct audio_stream_out *stream, const void *buffer,
size_t bytes)
{
struct stream_out *out = (struct stream_out *)stream;
struct audio_device *adev = out->dev;
ssize_t ret = 0;
lock_output_stream(out);
if (out->standby) {
out->standby = false;
pthread_mutex_lock(&adev->lock);
ret = start_output_stream(out);
pthread_mutex_unlock(&adev->lock);
/* ToDo: If use case is compress offload should return 0 */
if (ret != 0) {
out->standby = true;
goto exit;
}
if (out->usecase != USECASE_AUDIO_PLAYBACK_OFFLOAD && adev->adm_register_output_stream)
adev->adm_register_output_stream(adev->adm_data, out->handle, out->flags);
}
if (out->usecase == USECASE_AUDIO_PLAYBACK_OFFLOAD) {
ALOGVV("%s: writing buffer (%d bytes) to compress device", __func__, bytes);
if (out->send_new_metadata) {
ALOGVV("send new gapless metadata");
compress_set_gapless_metadata(out->compr, &out->gapless_mdata);
out->send_new_metadata = 0;
}
ret = compress_write(out->compr, buffer, bytes);
ALOGVV("%s: writing buffer (%d bytes) to compress device returned %d", __func__, bytes, ret);
if (ret >= 0 && ret < (ssize_t)bytes) {
send_offload_cmd_l(out, OFFLOAD_CMD_WAIT_FOR_BUFFER);
}
if (!out->playback_started) {
compress_start(out->compr);
out->playback_started = 1;
out->offload_state = OFFLOAD_STATE_PLAYING;
}
pthread_mutex_unlock(&out->lock);
return ret;
} else {
if (out->pcm) {
if (out->muted)
memset((void *)buffer, 0, bytes);
ALOGVV("%s: writing buffer (%d bytes) to pcm device", __func__, bytes);
if (adev->adm_request_focus)
adev->adm_request_focus(adev->adm_data, out->handle);
if (out->usecase == USECASE_AUDIO_PLAYBACK_AFE_PROXY) {
ret = pcm_mmap_write(out->pcm, (void *)buffer, bytes);
}
else
ret = pcm_write(out->pcm, (void *)buffer, bytes);
if (ret == 0)
out->written += bytes / (out->config.channels * sizeof(short));
if (adev->adm_abandon_focus)
adev->adm_abandon_focus(adev->adm_data, out->handle);
}
}
exit:
pthread_mutex_unlock(&out->lock);
if (ret != 0) {
if (out->pcm)
ALOGE("%s: error %zu - %s", __func__, ret, pcm_get_error(out->pcm));
out_standby(&out->stream.common);
usleep(bytes * 1000000 / audio_stream_out_frame_size(stream) /
out_get_sample_rate(&out->stream.common));
}
return bytes;
}
static int out_get_render_position(const struct audio_stream_out *stream,
uint32_t *dsp_frames)
{
struct stream_out *out = (struct stream_out *)stream;
*dsp_frames = 0;
if ((out->usecase == USECASE_AUDIO_PLAYBACK_OFFLOAD) && (dsp_frames != NULL)) {
lock_output_stream(out);
if (out->compr != NULL) {
compress_get_tstamp(out->compr, (unsigned long *)dsp_frames,
&out->sample_rate);
ALOGVV("%s rendered frames %d sample_rate %d",
__func__, *dsp_frames, out->sample_rate);
}
pthread_mutex_unlock(&out->lock);
return 0;
} else
return -EINVAL;
}
static int out_add_audio_effect(const struct audio_stream *stream __unused,
effect_handle_t effect __unused)
{
return 0;
}
static int out_remove_audio_effect(const struct audio_stream *stream __unused,
effect_handle_t effect __unused)
{
return 0;
}
static int out_get_next_write_timestamp(const struct audio_stream_out *stream __unused,
int64_t *timestamp __unused)
{
return -EINVAL;
}
static int out_get_presentation_position(const struct audio_stream_out *stream,
uint64_t *frames, struct timespec *timestamp)
{
struct stream_out *out = (struct stream_out *)stream;
int ret = -1;
unsigned long dsp_frames;
lock_output_stream(out);
if (out->usecase == USECASE_AUDIO_PLAYBACK_OFFLOAD) {
if (out->compr != NULL) {
compress_get_tstamp(out->compr, &dsp_frames,
&out->sample_rate);
ALOGVV("%s rendered frames %ld sample_rate %d",
__func__, dsp_frames, out->sample_rate);
*frames = dsp_frames;
ret = 0;
/* this is the best we can do */
clock_gettime(CLOCK_MONOTONIC, timestamp);
}
} else {
if (out->pcm) {
unsigned int avail;
if (pcm_get_htimestamp(out->pcm, &avail, timestamp) == 0) {
size_t kernel_buffer_size = out->config.period_size * out->config.period_count;
int64_t signed_frames = out->written - kernel_buffer_size + avail;
// This adjustment accounts for buffering after app processor.
// It is based on estimated DSP latency per use case, rather than exact.
signed_frames -=
(platform_render_latency(out->usecase) * out->sample_rate / 1000000LL);
// It would be unusual for this value to be negative, but check just in case ...
if (signed_frames >= 0) {
*frames = signed_frames;
ret = 0;
}
}
}
}
pthread_mutex_unlock(&out->lock);
return ret;
}
static int out_set_callback(struct audio_stream_out *stream,
stream_callback_t callback, void *cookie)
{
struct stream_out *out = (struct stream_out *)stream;
ALOGV("%s", __func__);
lock_output_stream(out);
out->offload_callback = callback;
out->offload_cookie = cookie;
pthread_mutex_unlock(&out->lock);
return 0;
}
static int out_pause(struct audio_stream_out* stream)
{
struct stream_out *out = (struct stream_out *)stream;
int status = -ENOSYS;
ALOGV("%s", __func__);
if (out->usecase == USECASE_AUDIO_PLAYBACK_OFFLOAD) {
lock_output_stream(out);
if (out->compr != NULL && out->offload_state == OFFLOAD_STATE_PLAYING) {
status = compress_pause(out->compr);
out->offload_state = OFFLOAD_STATE_PAUSED;
}
pthread_mutex_unlock(&out->lock);
}
return status;
}
static int out_resume(struct audio_stream_out* stream)
{
struct stream_out *out = (struct stream_out *)stream;
int status = -ENOSYS;
ALOGV("%s", __func__);
if (out->usecase == USECASE_AUDIO_PLAYBACK_OFFLOAD) {
status = 0;
lock_output_stream(out);
if (out->compr != NULL && out->offload_state == OFFLOAD_STATE_PAUSED) {
status = compress_resume(out->compr);
out->offload_state = OFFLOAD_STATE_PLAYING;
}
pthread_mutex_unlock(&out->lock);
}
return status;
}
static int out_drain(struct audio_stream_out* stream, audio_drain_type_t type )
{
struct stream_out *out = (struct stream_out *)stream;
int status = -ENOSYS;
ALOGV("%s", __func__);
if (out->usecase == USECASE_AUDIO_PLAYBACK_OFFLOAD) {
lock_output_stream(out);
if (type == AUDIO_DRAIN_EARLY_NOTIFY)
status = send_offload_cmd_l(out, OFFLOAD_CMD_PARTIAL_DRAIN);
else
status = send_offload_cmd_l(out, OFFLOAD_CMD_DRAIN);
pthread_mutex_unlock(&out->lock);
}
return status;
}
static int out_flush(struct audio_stream_out* stream)
{
struct stream_out *out = (struct stream_out *)stream;
ALOGV("%s", __func__);
if (out->usecase == USECASE_AUDIO_PLAYBACK_OFFLOAD) {
lock_output_stream(out);
stop_compressed_output_l(out);
pthread_mutex_unlock(&out->lock);
return 0;
}
return -ENOSYS;
}
/** audio_stream_in implementation **/
static uint32_t in_get_sample_rate(const struct audio_stream *stream)
{
struct stream_in *in = (struct stream_in *)stream;
return in->config.rate;
}
static int in_set_sample_rate(struct audio_stream *stream __unused, uint32_t rate __unused)
{
return -ENOSYS;
}
static size_t in_get_buffer_size(const struct audio_stream *stream)
{
struct stream_in *in = (struct stream_in *)stream;
return in->config.period_size *
audio_stream_in_frame_size((const struct audio_stream_in *)stream);
}
static uint32_t in_get_channels(const struct audio_stream *stream)
{
struct stream_in *in = (struct stream_in *)stream;
return in->channel_mask;
}
static audio_format_t in_get_format(const struct audio_stream *stream __unused)
{
return AUDIO_FORMAT_PCM_16_BIT;
}
static int in_set_format(struct audio_stream *stream __unused, audio_format_t format __unused)
{
return -ENOSYS;
}
static int in_standby(struct audio_stream *stream)
{
struct stream_in *in = (struct stream_in *)stream;
struct audio_device *adev = in->dev;
int status = 0;
ALOGV("%s: enter", __func__);
lock_input_stream(in);
if (!in->standby && in->is_st_session) {
ALOGD("%s: sound trigger pcm stop lab", __func__);
audio_extn_sound_trigger_stop_lab(in);
in->standby = true;
}
if (!in->standby) {
if (adev->adm_deregister_stream)
adev->adm_deregister_stream(adev->adm_data, in->capture_handle);
pthread_mutex_lock(&adev->lock);
in->standby = true;
if (in->pcm) {
pcm_close(in->pcm);
in->pcm = NULL;
}
adev->enable_voicerx = false;
platform_set_echo_reference(adev, false, AUDIO_DEVICE_NONE );
status = stop_input_stream(in);
pthread_mutex_unlock(&adev->lock);
}
pthread_mutex_unlock(&in->lock);
ALOGV("%s: exit: status(%d)", __func__, status);
return status;
}
static int in_dump(const struct audio_stream *stream __unused, int fd __unused)
{
return 0;
}
static int in_set_parameters(struct audio_stream *stream, const char *kvpairs)
{
struct stream_in *in = (struct stream_in *)stream;
struct audio_device *adev = in->dev;
struct str_parms *parms;
char *str;
char value[32];
int ret, val = 0;
int status = 0;
ALOGV("%s: enter: kvpairs=%s", __func__, kvpairs);
parms = str_parms_create_str(kvpairs);
ret = str_parms_get_str(parms, AUDIO_PARAMETER_STREAM_INPUT_SOURCE, value, sizeof(value));
lock_input_stream(in);
pthread_mutex_lock(&adev->lock);
if (ret >= 0) {
val = atoi(value);
/* no audio source uses val == 0 */
if ((in->source != val) && (val != 0)) {
in->source = val;
}
}
ret = str_parms_get_str(parms, AUDIO_PARAMETER_STREAM_ROUTING, value, sizeof(value));
if (ret >= 0) {
val = atoi(value);
if (((int)in->device != val) && (val != 0)) {
in->device = val;
/* If recording is in progress, change the tx device to new device */
if (!in->standby)
status = select_devices(adev, in->usecase);
}
}
pthread_mutex_unlock(&adev->lock);
pthread_mutex_unlock(&in->lock);
str_parms_destroy(parms);
ALOGV("%s: exit: status(%d)", __func__, status);
return status;
}
static char* in_get_parameters(const struct audio_stream *stream __unused,
const char *keys __unused)
{
return strdup("");
}
static int in_set_gain(struct audio_stream_in *stream __unused, float gain __unused)
{
return 0;
}
static ssize_t in_read(struct audio_stream_in *stream, void *buffer,
size_t bytes)
{
struct stream_in *in = (struct stream_in *)stream;
struct audio_device *adev = in->dev;
int i, ret = -1;
lock_input_stream(in);
if (in->is_st_session) {
ALOGVV(" %s: reading on st session bytes=%d", __func__, bytes);
/* Read from sound trigger HAL */
audio_extn_sound_trigger_read(in, buffer, bytes);
pthread_mutex_unlock(&in->lock);
return bytes;
}
if (in->standby) {
pthread_mutex_lock(&adev->lock);
ret = start_input_stream(in);
pthread_mutex_unlock(&adev->lock);
if (ret != 0) {
goto exit;
}
in->standby = 0;
if (adev->adm_register_input_stream)
adev->adm_register_input_stream(adev->adm_data, in->capture_handle, in->flags);
}
if (adev->adm_request_focus)
adev->adm_request_focus(adev->adm_data, in->capture_handle);
if (in->pcm) {
if (in->usecase == USECASE_AUDIO_RECORD_AFE_PROXY) {
ret = pcm_mmap_read(in->pcm, buffer, bytes);
} else
ret = pcm_read(in->pcm, buffer, bytes);
}
if (adev->adm_abandon_focus)
adev->adm_abandon_focus(adev->adm_data, in->capture_handle);
/*
* Instead of writing zeroes here, we could trust the hardware
* to always provide zeroes when muted.
* No need to acquire adev->lock to read mic_muted here as we don't change its state.
*/
if (ret == 0 && adev->mic_muted && in->usecase != USECASE_AUDIO_RECORD_AFE_PROXY)
memset(buffer, 0, bytes);
exit:
pthread_mutex_unlock(&in->lock);
if (ret != 0) {
in_standby(&in->stream.common);
ALOGV("%s: read failed - sleeping for buffer duration", __func__);
usleep(bytes * 1000000 / audio_stream_in_frame_size(stream) /
in_get_sample_rate(&in->stream.common));
}
return bytes;
}
static uint32_t in_get_input_frames_lost(struct audio_stream_in *stream __unused)
{
return 0;
}
static int add_remove_audio_effect(const struct audio_stream *stream,
effect_handle_t effect,
bool enable)
{
struct stream_in *in = (struct stream_in *)stream;
struct audio_device *adev = in->dev;
int status = 0;
effect_descriptor_t desc;
status = (*effect)->get_descriptor(effect, &desc);
if (status != 0)
return status;
lock_input_stream(in);
pthread_mutex_lock(&in->dev->lock);
if ((in->source == AUDIO_SOURCE_VOICE_COMMUNICATION ||
adev->mode == AUDIO_MODE_IN_COMMUNICATION) &&
in->enable_aec != enable &&
(memcmp(&desc.type, FX_IID_AEC, sizeof(effect_uuid_t)) == 0)) {
in->enable_aec = enable;
if (!enable)
platform_set_echo_reference(in->dev, enable, AUDIO_DEVICE_NONE);
adev->enable_voicerx = enable;
struct audio_usecase *usecase;
struct listnode *node;
list_for_each(node, &adev->usecase_list) {
usecase = node_to_item(node, struct audio_usecase, list);
if (usecase->type == PCM_PLAYBACK) {
select_devices(adev, usecase->id);
break;
}
}
if (!in->standby)
select_devices(in->dev, in->usecase);
}
if (in->enable_ns != enable &&
(memcmp(&desc.type, FX_IID_NS, sizeof(effect_uuid_t)) == 0)) {
in->enable_ns = enable;
if (!in->standby)
select_devices(in->dev, in->usecase);
}
pthread_mutex_unlock(&in->dev->lock);
pthread_mutex_unlock(&in->lock);
return 0;
}
static int in_add_audio_effect(const struct audio_stream *stream,
effect_handle_t effect)
{
ALOGV("%s: effect %p", __func__, effect);
return add_remove_audio_effect(stream, effect, true);
}
static int in_remove_audio_effect(const struct audio_stream *stream,
effect_handle_t effect)
{
ALOGV("%s: effect %p", __func__, effect);
return add_remove_audio_effect(stream, effect, false);
}
static int adev_open_output_stream(struct audio_hw_device *dev,
audio_io_handle_t handle,
audio_devices_t devices,
audio_output_flags_t flags,
struct audio_config *config,
struct audio_stream_out **stream_out,
const char *address __unused)
{
struct audio_device *adev = (struct audio_device *)dev;
struct stream_out *out;
int i, ret;
ALOGV("%s: enter: sample_rate(%d) channel_mask(%#x) devices(%#x) flags(%#x)",
__func__, config->sample_rate, config->channel_mask, devices, flags);
*stream_out = NULL;
out = (struct stream_out *)calloc(1, sizeof(struct stream_out));
if (devices == AUDIO_DEVICE_NONE)
devices = AUDIO_DEVICE_OUT_SPEAKER;
out->flags = flags;
out->devices = devices;
out->dev = adev;
out->format = config->format;
out->sample_rate = config->sample_rate;
out->channel_mask = AUDIO_CHANNEL_OUT_STEREO;
out->supported_channel_masks[0] = AUDIO_CHANNEL_OUT_STEREO;
out->handle = handle;
/* Init use case and pcm_config */
if (out->flags & AUDIO_OUTPUT_FLAG_DIRECT &&
!(out->flags & AUDIO_OUTPUT_FLAG_COMPRESS_OFFLOAD) &&
out->devices & AUDIO_DEVICE_OUT_AUX_DIGITAL) {
pthread_mutex_lock(&adev->lock);
ret = read_hdmi_channel_masks(out);
pthread_mutex_unlock(&adev->lock);
if (ret != 0)
goto error_open;
if (config->sample_rate == 0)
config->sample_rate = DEFAULT_OUTPUT_SAMPLING_RATE;
if (config->channel_mask == 0)
config->channel_mask = AUDIO_CHANNEL_OUT_5POINT1;
out->channel_mask = config->channel_mask;
out->sample_rate = config->sample_rate;
out->usecase = USECASE_AUDIO_PLAYBACK_MULTI_CH;
out->config = pcm_config_hdmi_multi;
out->config.rate = config->sample_rate;
out->config.channels = audio_channel_count_from_out_mask(out->channel_mask);
out->config.period_size = HDMI_MULTI_PERIOD_BYTES / (out->config.channels * 2);
} else if (out->flags & AUDIO_OUTPUT_FLAG_COMPRESS_OFFLOAD) {
if (config->offload_info.version != AUDIO_INFO_INITIALIZER.version ||
config->offload_info.size != AUDIO_INFO_INITIALIZER.size) {
ALOGE("%s: Unsupported Offload information", __func__);
ret = -EINVAL;
goto error_open;
}
if (!is_supported_format(config->offload_info.format)) {
ALOGE("%s: Unsupported audio format", __func__);
ret = -EINVAL;
goto error_open;
}
out->compr_config.codec = (struct snd_codec *)
calloc(1, sizeof(struct snd_codec));
out->usecase = USECASE_AUDIO_PLAYBACK_OFFLOAD;
if (config->offload_info.channel_mask)
out->channel_mask = config->offload_info.channel_mask;
else if (config->channel_mask)
out->channel_mask = config->channel_mask;
out->format = config->offload_info.format;
out->sample_rate = config->offload_info.sample_rate;
out->stream.set_callback = out_set_callback;
out->stream.pause = out_pause;
out->stream.resume = out_resume;
out->stream.drain = out_drain;
out->stream.flush = out_flush;
out->compr_config.codec->id =
get_snd_codec_id(config->offload_info.format);
out->compr_config.fragment_size = COMPRESS_OFFLOAD_FRAGMENT_SIZE;
out->compr_config.fragments = COMPRESS_OFFLOAD_NUM_FRAGMENTS;
out->compr_config.codec->sample_rate = config->offload_info.sample_rate;
out->compr_config.codec->bit_rate =
config->offload_info.bit_rate;
out->compr_config.codec->ch_in =
audio_channel_count_from_out_mask(config->channel_mask);
out->compr_config.codec->ch_out = out->compr_config.codec->ch_in;
if (flags & AUDIO_OUTPUT_FLAG_NON_BLOCKING)
out->non_blocking = 1;
out->send_new_metadata = 1;
create_offload_callback_thread(out);
ALOGV("%s: offloaded output offload_info version %04x bit rate %d",
__func__, config->offload_info.version,
config->offload_info.bit_rate);
} else if (out->devices == AUDIO_DEVICE_OUT_TELEPHONY_TX) {
if (config->sample_rate == 0)
config->sample_rate = AFE_PROXY_SAMPLING_RATE;
if (config->sample_rate != 48000 && config->sample_rate != 16000 &&
config->sample_rate != 8000) {
config->sample_rate = AFE_PROXY_SAMPLING_RATE;
ret = -EINVAL;
goto error_open;
}
out->sample_rate = config->sample_rate;
out->config.rate = config->sample_rate;
if (config->format == AUDIO_FORMAT_DEFAULT)
config->format = AUDIO_FORMAT_PCM_16_BIT;
if (config->format != AUDIO_FORMAT_PCM_16_BIT) {
config->format = AUDIO_FORMAT_PCM_16_BIT;
ret = -EINVAL;
goto error_open;
}
out->format = config->format;
out->usecase = USECASE_AUDIO_PLAYBACK_AFE_PROXY;
out->config = pcm_config_afe_proxy_playback;
adev->voice_tx_output = out;
} else {
if (out->flags & AUDIO_OUTPUT_FLAG_DEEP_BUFFER) {
out->usecase = USECASE_AUDIO_PLAYBACK_DEEP_BUFFER;
out->config = pcm_config_deep_buffer;
} else if (out->flags & AUDIO_OUTPUT_FLAG_TTS) {
out->usecase = USECASE_AUDIO_PLAYBACK_TTS;
out->config = pcm_config_deep_buffer;
} else if (out->flags & AUDIO_OUTPUT_FLAG_RAW) {
out->usecase = USECASE_AUDIO_PLAYBACK_ULL;
out->config = pcm_config_low_latency;
} else {
out->usecase = USECASE_AUDIO_PLAYBACK_LOW_LATENCY;
out->config = pcm_config_low_latency;
}
if (config->format != audio_format_from_pcm_format(out->config.format)) {
if (k_enable_extended_precision
&& pcm_params_format_test(adev->use_case_table[out->usecase],
pcm_format_from_audio_format(config->format))) {
out->config.format = pcm_format_from_audio_format(config->format);
/* out->format already set to config->format */
} else {
/* deny the externally proposed config format
* and use the one specified in audio_hw layer configuration.
* Note: out->format is returned by out->stream.common.get_format()
* and is used to set config->format in the code several lines below.
*/
out->format = audio_format_from_pcm_format(out->config.format);
}
}
out->sample_rate = out->config.rate;
}
ALOGV("%s: Usecase(%s) config->format %#x out->config.format %#x\n",
__func__, use_case_table[out->usecase], config->format, out->config.format);
if (flags & AUDIO_OUTPUT_FLAG_PRIMARY) {
if (adev->primary_output == NULL)
adev->primary_output = out;
else {
ALOGE("%s: Primary output is already opened", __func__);
ret = -EEXIST;
goto error_open;
}
}
/* Check if this usecase is already existing */
pthread_mutex_lock(&adev->lock);
if (get_usecase_from_list(adev, out->usecase) != NULL) {
ALOGE("%s: Usecase (%d) is already present", __func__, out->usecase);
pthread_mutex_unlock(&adev->lock);
ret = -EEXIST;
goto error_open;
}
pthread_mutex_unlock(&adev->lock);
out->stream.common.get_sample_rate = out_get_sample_rate;
out->stream.common.set_sample_rate = out_set_sample_rate;
out->stream.common.get_buffer_size = out_get_buffer_size;
out->stream.common.get_channels = out_get_channels;
out->stream.common.get_format = out_get_format;
out->stream.common.set_format = out_set_format;
out->stream.common.standby = out_standby;
out->stream.common.dump = out_dump;
out->stream.common.set_parameters = out_set_parameters;
out->stream.common.get_parameters = out_get_parameters;
out->stream.common.add_audio_effect = out_add_audio_effect;
out->stream.common.remove_audio_effect = out_remove_audio_effect;
out->stream.get_latency = out_get_latency;
out->stream.set_volume = out_set_volume;
#ifdef NO_AUDIO_OUT
out->stream.write = out_write_for_no_output;
#else
out->stream.write = out_write;
#endif
out->stream.get_render_position = out_get_render_position;
out->stream.get_next_write_timestamp = out_get_next_write_timestamp;
out->stream.get_presentation_position = out_get_presentation_position;
out->standby = 1;
/* out->muted = false; by calloc() */
/* out->written = 0; by calloc() */
pthread_mutex_init(&out->lock, (const pthread_mutexattr_t *) NULL);
pthread_mutex_init(&out->pre_lock, (const pthread_mutexattr_t *) NULL);
pthread_cond_init(&out->cond, (const pthread_condattr_t *) NULL);
config->format = out->stream.common.get_format(&out->stream.common);
config->channel_mask = out->stream.common.get_channels(&out->stream.common);
config->sample_rate = out->stream.common.get_sample_rate(&out->stream.common);
*stream_out = &out->stream;
ALOGV("%s: exit", __func__);
return 0;
error_open:
free(out);
*stream_out = NULL;
ALOGD("%s: exit: ret %d", __func__, ret);
return ret;
}
static void adev_close_output_stream(struct audio_hw_device *dev __unused,
struct audio_stream_out *stream)
{
struct stream_out *out = (struct stream_out *)stream;
struct audio_device *adev = out->dev;
ALOGV("%s: enter", __func__);
out_standby(&stream->common);
if (out->usecase == USECASE_AUDIO_PLAYBACK_OFFLOAD) {
destroy_offload_callback_thread(out);
if (out->compr_config.codec != NULL)
free(out->compr_config.codec);
}
if (adev->voice_tx_output == out)
adev->voice_tx_output = NULL;
pthread_cond_destroy(&out->cond);
pthread_mutex_destroy(&out->lock);
free(stream);
ALOGV("%s: exit", __func__);
}
static int adev_set_parameters(struct audio_hw_device *dev, const char *kvpairs)
{
struct audio_device *adev = (struct audio_device *)dev;
struct str_parms *parms;
char *str;
char value[32];
int val;
int ret;
int status = 0;
ALOGD("%s: enter: %s", __func__, kvpairs);
pthread_mutex_lock(&adev->lock);
parms = str_parms_create_str(kvpairs);
status = voice_set_parameters(adev, parms);
if (status != 0) {
goto done;
}
ret = str_parms_get_str(parms, AUDIO_PARAMETER_KEY_BT_NREC, value, sizeof(value));
if (ret >= 0) {
/* When set to false, HAL should disable EC and NS */
if (strcmp(value, AUDIO_PARAMETER_VALUE_ON) == 0)
adev->bluetooth_nrec = true;
else
adev->bluetooth_nrec = false;
}
ret = str_parms_get_str(parms, "screen_state", value, sizeof(value));
if (ret >= 0) {
if (strcmp(value, AUDIO_PARAMETER_VALUE_ON) == 0)
adev->screen_off = false;
else
adev->screen_off = true;
}
ret = str_parms_get_int(parms, "rotation", &val);
if (ret >= 0) {
bool reverse_speakers = false;
switch(val) {
// FIXME: note that the code below assumes that the speakers are in the correct placement
// relative to the user when the device is rotated 90deg from its default rotation. This
// assumption is device-specific, not platform-specific like this code.
case 270:
reverse_speakers = true;
break;
case 0:
case 90:
case 180:
break;
default:
ALOGE("%s: unexpected rotation of %d", __func__, val);
status = -EINVAL;
}
if (status == 0) {
platform_swap_lr_channels(adev, reverse_speakers);
}
}
ret = str_parms_get_str(parms, AUDIO_PARAMETER_KEY_BT_SCO_WB, value, sizeof(value));
if (ret >= 0) {
adev->bt_wb_speech_enabled = !strcmp(value, AUDIO_PARAMETER_VALUE_ON);
}
audio_extn_hfp_set_parameters(adev, parms);
done:
str_parms_destroy(parms);
pthread_mutex_unlock(&adev->lock);
ALOGV("%s: exit with code(%d)", __func__, status);
return status;
}
static char* adev_get_parameters(const struct audio_hw_device *dev,
const char *keys)
{
struct audio_device *adev = (struct audio_device *)dev;
struct str_parms *reply = str_parms_create();
struct str_parms *query = str_parms_create_str(keys);
char *str;
pthread_mutex_lock(&adev->lock);
voice_get_parameters(adev, query, reply);
str = str_parms_to_str(reply);
str_parms_destroy(query);
str_parms_destroy(reply);
pthread_mutex_unlock(&adev->lock);
ALOGV("%s: exit: returns - %s", __func__, str);
return str;
}
static int adev_init_check(const struct audio_hw_device *dev __unused)
{
return 0;
}
static int adev_set_voice_volume(struct audio_hw_device *dev, float volume)
{
int ret;
struct audio_device *adev = (struct audio_device *)dev;
audio_extn_extspk_set_voice_vol(adev->extspk, volume);
pthread_mutex_lock(&adev->lock);
ret = voice_set_volume(adev, volume);
pthread_mutex_unlock(&adev->lock);
return ret;
}
static int adev_set_master_volume(struct audio_hw_device *dev __unused, float volume __unused)
{
return -ENOSYS;
}
static int adev_get_master_volume(struct audio_hw_device *dev __unused,
float *volume __unused)
{
return -ENOSYS;
}
static int adev_set_master_mute(struct audio_hw_device *dev __unused, bool muted __unused)
{
return -ENOSYS;
}
static int adev_get_master_mute(struct audio_hw_device *dev __unused, bool *muted __unused)
{
return -ENOSYS;
}
static int adev_set_mode(struct audio_hw_device *dev, audio_mode_t mode)
{
struct audio_device *adev = (struct audio_device *)dev;
pthread_mutex_lock(&adev->lock);
if (adev->mode != mode) {
ALOGD("%s: mode %d\n", __func__, mode);
adev->mode = mode;
if ((mode == AUDIO_MODE_NORMAL || mode == AUDIO_MODE_IN_COMMUNICATION) &&
voice_is_in_call(adev)) {
voice_stop_call(adev);
adev->current_call_output = NULL;
}
}
pthread_mutex_unlock(&adev->lock);
audio_extn_extspk_set_mode(adev->extspk, mode);
return 0;
}
static int adev_set_mic_mute(struct audio_hw_device *dev, bool state)
{
int ret;
struct audio_device *adev = (struct audio_device *)dev;
ALOGD("%s: state %d\n", __func__, state);
pthread_mutex_lock(&adev->lock);
ret = voice_set_mic_mute(adev, state);
adev->mic_muted = state;
pthread_mutex_unlock(&adev->lock);
return ret;
}
static int adev_get_mic_mute(const struct audio_hw_device *dev, bool *state)
{
*state = voice_get_mic_mute((struct audio_device *)dev);
return 0;
}
static size_t adev_get_input_buffer_size(const struct audio_hw_device *dev __unused,
const struct audio_config *config)
{
int channel_count = audio_channel_count_from_in_mask(config->channel_mask);
return get_input_buffer_size(config->sample_rate, config->format, channel_count,
false /* is_low_latency: since we don't know, be conservative */);
}
static int adev_open_input_stream(struct audio_hw_device *dev,
audio_io_handle_t handle,
audio_devices_t devices,
struct audio_config *config,
struct audio_stream_in **stream_in,
audio_input_flags_t flags,
const char *address __unused,
audio_source_t source )
{
struct audio_device *adev = (struct audio_device *)dev;
struct stream_in *in;
int ret = 0, buffer_size, frame_size;
int channel_count = audio_channel_count_from_in_mask(config->channel_mask);
bool is_low_latency = false;
ALOGV("%s: enter", __func__);
*stream_in = NULL;
if (check_input_parameters(config->sample_rate, config->format, channel_count) != 0)
return -EINVAL;
in = (struct stream_in *)calloc(1, sizeof(struct stream_in));
pthread_mutex_init(&in->lock, (const pthread_mutexattr_t *) NULL);
pthread_mutex_init(&in->pre_lock, (const pthread_mutexattr_t *) NULL);
in->stream.common.get_sample_rate = in_get_sample_rate;
in->stream.common.set_sample_rate = in_set_sample_rate;
in->stream.common.get_buffer_size = in_get_buffer_size;
in->stream.common.get_channels = in_get_channels;
in->stream.common.get_format = in_get_format;
in->stream.common.set_format = in_set_format;
in->stream.common.standby = in_standby;
in->stream.common.dump = in_dump;
in->stream.common.set_parameters = in_set_parameters;
in->stream.common.get_parameters = in_get_parameters;
in->stream.common.add_audio_effect = in_add_audio_effect;
in->stream.common.remove_audio_effect = in_remove_audio_effect;
in->stream.set_gain = in_set_gain;
in->stream.read = in_read;
in->stream.get_input_frames_lost = in_get_input_frames_lost;
in->device = devices;
in->source = source;
in->dev = adev;
in->standby = 1;
in->channel_mask = config->channel_mask;
in->capture_handle = handle;
in->flags = flags;
/* Update config params with the requested sample rate and channels */
if (in->device == AUDIO_DEVICE_IN_TELEPHONY_RX) {
if (config->sample_rate == 0)
config->sample_rate = AFE_PROXY_SAMPLING_RATE;
if (config->sample_rate != 48000 && config->sample_rate != 16000 &&
config->sample_rate != 8000) {
config->sample_rate = AFE_PROXY_SAMPLING_RATE;
ret = -EINVAL;
goto err_open;
}
if (config->format == AUDIO_FORMAT_DEFAULT)
config->format = AUDIO_FORMAT_PCM_16_BIT;
if (config->format != AUDIO_FORMAT_PCM_16_BIT) {
config->format = AUDIO_FORMAT_PCM_16_BIT;
ret = -EINVAL;
goto err_open;
}
in->usecase = USECASE_AUDIO_RECORD_AFE_PROXY;
in->config = pcm_config_afe_proxy_record;
} else {
in->usecase = USECASE_AUDIO_RECORD;
if (config->sample_rate == LOW_LATENCY_CAPTURE_SAMPLE_RATE &&