blob: 39c0fb5c16fb757bb0e8a51a0f2f7cd46c78188e [file] [log] [blame]
/*
* Copyright (C) 2012 The Android Open Source Project
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* 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 "modules.usbaudio.audio_hal"
/* #define LOG_NDEBUG 0 */
#include <errno.h>
#include <inttypes.h>
#include <pthread.h>
#include <stdint.h>
#include <stdlib.h>
#include <string.h>
#include <sys/time.h>
#include <unistd.h>
#include <log/log.h>
#include <cutils/list.h>
#include <cutils/str_parms.h>
#include <cutils/properties.h>
#include <hardware/audio.h>
#include <hardware/audio_alsaops.h>
#include <hardware/hardware.h>
#include <system/audio.h>
#include <tinyalsa/asoundlib.h>
#include <audio_utils/channels.h>
#include "alsa_device_profile.h"
#include "alsa_device_proxy.h"
#include "alsa_logging.h"
/* Lock play & record samples rates at or above this threshold */
#define RATELOCK_THRESHOLD 96000
#define max(a, b) ((a) > (b) ? (a) : (b))
#define min(a, b) ((a) < (b) ? (a) : (b))
struct audio_device {
struct audio_hw_device hw_device;
pthread_mutex_t lock; /* see note below on mutex acquisition order */
/* output */
struct listnode output_stream_list;
/* input */
struct listnode input_stream_list;
/* lock input & output sample rates */
/*FIXME - How do we address multiple output streams? */
uint32_t device_sample_rate; // this should be a rate that is common to both input & output
bool mic_muted;
int32_t inputs_open; /* number of input streams currently open. */
audio_patch_handle_t next_patch_handle; // Increase 1 when create audio patch
};
struct stream_lock {
pthread_mutex_t lock; /* see note below on mutex acquisition order */
pthread_mutex_t pre_lock; /* acquire before lock to avoid DOS by playback thread */
};
struct alsa_device_info {
alsa_device_profile profile; /* The profile of the ALSA device */
alsa_device_proxy proxy; /* The state */
struct listnode list_node;
};
struct stream_out {
struct audio_stream_out stream;
struct stream_lock lock;
bool standby;
struct audio_device *adev; /* hardware information - only using this for the lock */
struct listnode alsa_devices; /* The ALSA devices connected to the stream. */
unsigned hal_channel_count; /* channel count exposed to AudioFlinger.
* This may differ from the device channel count when
* the device is not compatible with AudioFlinger
* capabilities, e.g. exposes too many channels or
* too few channels. */
audio_channel_mask_t hal_channel_mask; /* USB devices deal in channel counts, not masks
* so the proxy doesn't have a channel_mask, but
* audio HALs need to talk about channel masks
* so expose the one calculated by
* adev_open_output_stream */
struct listnode list_node;
void * conversion_buffer; /* any conversions are put into here
* they could come from here too if
* there was a previous conversion */
size_t conversion_buffer_size; /* in bytes */
struct pcm_config config;
audio_io_handle_t handle; // Unique constant for a stream
audio_patch_handle_t patch_handle; // Patch handle for this stream
};
struct stream_in {
struct audio_stream_in stream;
struct stream_lock lock;
bool standby;
struct audio_device *adev; /* hardware information - only using this for the lock */
struct listnode alsa_devices; /* The ALSA devices connected to the stream. */
unsigned hal_channel_count; /* channel count exposed to AudioFlinger.
* This may differ from the device channel count when
* the device is not compatible with AudioFlinger
* capabilities, e.g. exposes too many channels or
* too few channels. */
audio_channel_mask_t hal_channel_mask; /* USB devices deal in channel counts, not masks
* so the proxy doesn't have a channel_mask, but
* audio HALs need to talk about channel masks
* so expose the one calculated by
* adev_open_input_stream */
struct listnode list_node;
/* We may need to read more data from the device in order to data reduce to 16bit, 4chan */
void * conversion_buffer; /* any conversions are put into here
* they could come from here too if
* there was a previous conversion */
size_t conversion_buffer_size; /* in bytes */
struct pcm_config config;
audio_io_handle_t handle; // Unique identifier for a stream
audio_patch_handle_t patch_handle; // Patch handle for this stream
};
// Map channel count to output channel mask
static const audio_channel_mask_t OUT_CHANNEL_MASKS_MAP[FCC_24 + 1] = {
[0] = AUDIO_CHANNEL_NONE, // == 0 (so this line is optional and could be omitted)
// != AUDIO_CHANNEL_INVALID == 0xC0000000u
[1] = AUDIO_CHANNEL_OUT_MONO,
[2] = AUDIO_CHANNEL_OUT_STEREO,
[3] = AUDIO_CHANNEL_OUT_2POINT1,
[4] = AUDIO_CHANNEL_OUT_QUAD,
[5] = AUDIO_CHANNEL_OUT_PENTA,
[6] = AUDIO_CHANNEL_OUT_5POINT1,
[7] = AUDIO_CHANNEL_OUT_6POINT1,
[8] = AUDIO_CHANNEL_OUT_7POINT1,
[9 ... 11] = AUDIO_CHANNEL_NONE, // == 0 (so this line is optional and could be omitted).
[12] = AUDIO_CHANNEL_OUT_7POINT1POINT4,
[13 ... 23] = AUDIO_CHANNEL_NONE, // == 0 (so this line is optional and could be omitted).
[24] = AUDIO_CHANNEL_OUT_22POINT2,
};
static const int OUT_CHANNEL_MASKS_SIZE = AUDIO_ARRAY_SIZE(OUT_CHANNEL_MASKS_MAP);
// Map channel count to input channel mask
static const audio_channel_mask_t IN_CHANNEL_MASKS_MAP[] = {
AUDIO_CHANNEL_NONE, /* 0 */
AUDIO_CHANNEL_IN_MONO, /* 1 */
AUDIO_CHANNEL_IN_STEREO, /* 2 */
/* channel counts greater than this are not considered */
};
static const int IN_CHANNEL_MASKS_SIZE = AUDIO_ARRAY_SIZE(IN_CHANNEL_MASKS_MAP);
// Map channel count to index mask
static const audio_channel_mask_t CHANNEL_INDEX_MASKS_MAP[FCC_24 + 1] = {
[0] = AUDIO_CHANNEL_NONE, // == 0 (so this line is optional and could be omitted).
[1] = AUDIO_CHANNEL_INDEX_MASK_1,
[2] = AUDIO_CHANNEL_INDEX_MASK_2,
[3] = AUDIO_CHANNEL_INDEX_MASK_3,
[4] = AUDIO_CHANNEL_INDEX_MASK_4,
[5] = AUDIO_CHANNEL_INDEX_MASK_5,
[6] = AUDIO_CHANNEL_INDEX_MASK_6,
[7] = AUDIO_CHANNEL_INDEX_MASK_7,
[8] = AUDIO_CHANNEL_INDEX_MASK_8,
[9] = AUDIO_CHANNEL_INDEX_MASK_9,
[10] = AUDIO_CHANNEL_INDEX_MASK_10,
[11] = AUDIO_CHANNEL_INDEX_MASK_11,
[12] = AUDIO_CHANNEL_INDEX_MASK_12,
[13] = AUDIO_CHANNEL_INDEX_MASK_13,
[14] = AUDIO_CHANNEL_INDEX_MASK_14,
[15] = AUDIO_CHANNEL_INDEX_MASK_15,
[16] = AUDIO_CHANNEL_INDEX_MASK_16,
[17] = AUDIO_CHANNEL_INDEX_MASK_17,
[18] = AUDIO_CHANNEL_INDEX_MASK_18,
[19] = AUDIO_CHANNEL_INDEX_MASK_19,
[20] = AUDIO_CHANNEL_INDEX_MASK_20,
[21] = AUDIO_CHANNEL_INDEX_MASK_21,
[22] = AUDIO_CHANNEL_INDEX_MASK_22,
[23] = AUDIO_CHANNEL_INDEX_MASK_23,
[24] = AUDIO_CHANNEL_INDEX_MASK_24,
};
static const int CHANNEL_INDEX_MASKS_SIZE = AUDIO_ARRAY_SIZE(CHANNEL_INDEX_MASKS_MAP);
/*
* Locking Helpers
*/
/*
* NOTE: when multiple mutexes have to be acquired, always take the
* stream_in or stream_out mutex first, followed by the audio_device mutex.
* stream pre_lock is always acquired before stream lock to prevent starvation of control thread by
* higher priority playback or capture thread.
*/
static void stream_lock_init(struct stream_lock *lock) {
pthread_mutex_init(&lock->lock, (const pthread_mutexattr_t *) NULL);
pthread_mutex_init(&lock->pre_lock, (const pthread_mutexattr_t *) NULL);
}
static void stream_lock(struct stream_lock *lock) {
if (lock == NULL) {
return;
}
pthread_mutex_lock(&lock->pre_lock);
pthread_mutex_lock(&lock->lock);
pthread_mutex_unlock(&lock->pre_lock);
}
static void stream_unlock(struct stream_lock *lock) {
pthread_mutex_unlock(&lock->lock);
}
static void device_lock(struct audio_device *adev) {
pthread_mutex_lock(&adev->lock);
}
static int device_try_lock(struct audio_device *adev) {
return pthread_mutex_trylock(&adev->lock);
}
static void device_unlock(struct audio_device *adev) {
pthread_mutex_unlock(&adev->lock);
}
/*
* streams list management
*/
static void adev_add_stream_to_list(
struct audio_device* adev, struct listnode* list, struct listnode* stream_node) {
device_lock(adev);
list_add_tail(list, stream_node);
device_unlock(adev);
}
static struct stream_out* adev_get_stream_out_by_io_handle_l(
struct audio_device* adev, audio_io_handle_t handle) {
struct listnode *node;
list_for_each (node, &adev->output_stream_list) {
struct stream_out *out = node_to_item(node, struct stream_out, list_node);
if (out->handle == handle) {
return out;
}
}
return NULL;
}
static struct stream_in* adev_get_stream_in_by_io_handle_l(
struct audio_device* adev, audio_io_handle_t handle) {
struct listnode *node;
list_for_each (node, &adev->input_stream_list) {
struct stream_in *in = node_to_item(node, struct stream_in, list_node);
if (in->handle == handle) {
return in;
}
}
return NULL;
}
static struct stream_out* adev_get_stream_out_by_patch_handle_l(
struct audio_device* adev, audio_patch_handle_t patch_handle) {
struct listnode *node;
list_for_each (node, &adev->output_stream_list) {
struct stream_out *out = node_to_item(node, struct stream_out, list_node);
if (out->patch_handle == patch_handle) {
return out;
}
}
return NULL;
}
static struct stream_in* adev_get_stream_in_by_patch_handle_l(
struct audio_device* adev, audio_patch_handle_t patch_handle) {
struct listnode *node;
list_for_each (node, &adev->input_stream_list) {
struct stream_in *in = node_to_item(node, struct stream_in, list_node);
if (in->patch_handle == patch_handle) {
return in;
}
}
return NULL;
}
/*
* Extract the card and device numbers from the supplied key/value pairs.
* kvpairs A null-terminated string containing the key/value pairs or card and device.
* i.e. "card=1;device=42"
* card A pointer to a variable to receive the parsed-out card number.
* device A pointer to a variable to receive the parsed-out device number.
* NOTE: The variables pointed to by card and device return -1 (undefined) if the
* associated key/value pair is not found in the provided string.
* Return true if the kvpairs string contain a card/device spec, false otherwise.
*/
static bool parse_card_device_params(const char *kvpairs, int *card, int *device)
{
struct str_parms * parms = str_parms_create_str(kvpairs);
char value[32];
int param_val;
// initialize to "undefined" state.
*card = -1;
*device = -1;
param_val = str_parms_get_str(parms, "card", value, sizeof(value));
if (param_val >= 0) {
*card = atoi(value);
}
param_val = str_parms_get_str(parms, "device", value, sizeof(value));
if (param_val >= 0) {
*device = atoi(value);
}
str_parms_destroy(parms);
return *card >= 0 && *device >= 0;
}
static char *device_get_parameters(const alsa_device_profile *profile, const char * keys)
{
if (profile->card < 0 || profile->device < 0) {
return strdup("");
}
struct str_parms *query = str_parms_create_str(keys);
struct str_parms *result = str_parms_create();
/* These keys are from hardware/libhardware/include/audio.h */
/* supported sample rates */
if (str_parms_has_key(query, AUDIO_PARAMETER_STREAM_SUP_SAMPLING_RATES)) {
char* rates_list = profile_get_sample_rate_strs(profile);
str_parms_add_str(result, AUDIO_PARAMETER_STREAM_SUP_SAMPLING_RATES,
rates_list);
free(rates_list);
}
/* supported channel counts */
if (str_parms_has_key(query, AUDIO_PARAMETER_STREAM_SUP_CHANNELS)) {
char* channels_list = profile_get_channel_count_strs(profile);
str_parms_add_str(result, AUDIO_PARAMETER_STREAM_SUP_CHANNELS,
channels_list);
free(channels_list);
}
/* supported sample formats */
if (str_parms_has_key(query, AUDIO_PARAMETER_STREAM_SUP_FORMATS)) {
char * format_params = profile_get_format_strs(profile);
str_parms_add_str(result, AUDIO_PARAMETER_STREAM_SUP_FORMATS,
format_params);
free(format_params);
}
str_parms_destroy(query);
char* result_str = str_parms_to_str(result);
str_parms_destroy(result);
ALOGV("device_get_parameters = %s", result_str);
return result_str;
}
static audio_format_t audio_format_from(enum pcm_format format)
{
switch (format) {
case PCM_FORMAT_S16_LE:
return AUDIO_FORMAT_PCM_16_BIT;
case PCM_FORMAT_S32_LE:
return AUDIO_FORMAT_PCM_32_BIT;
case PCM_FORMAT_S8:
return AUDIO_FORMAT_PCM_8_BIT;
case PCM_FORMAT_S24_LE:
return AUDIO_FORMAT_PCM_8_24_BIT;
case PCM_FORMAT_S24_3LE:
return AUDIO_FORMAT_PCM_24_BIT_PACKED;
default:
return AUDIO_FORMAT_INVALID;
}
}
static unsigned int populate_channel_mask_from_profile(const alsa_device_profile* profile,
bool is_output,
audio_channel_mask_t channel_masks[])
{
unsigned int num_channel_masks = 0;
const audio_channel_mask_t* channel_masks_map =
is_output ? OUT_CHANNEL_MASKS_MAP : IN_CHANNEL_MASKS_MAP;
int channel_masks_size = is_output ? OUT_CHANNEL_MASKS_SIZE : IN_CHANNEL_MASKS_SIZE;
if (channel_masks_size > FCC_LIMIT + 1) {
channel_masks_size = FCC_LIMIT + 1;
}
unsigned int channel_count = 0;
for (size_t i = 0; i < min(channel_masks_size, AUDIO_PORT_MAX_CHANNEL_MASKS) &&
(channel_count = profile->channel_counts[i]) != 0 &&
num_channel_masks < AUDIO_PORT_MAX_CHANNEL_MASKS; ++i) {
if (channel_count < channel_masks_size &&
channel_masks_map[channel_count] != AUDIO_CHANNEL_NONE) {
channel_masks[num_channel_masks++] = channel_masks_map[channel_count];
if (num_channel_masks >= AUDIO_PORT_MAX_CHANNEL_MASKS) {
break;
}
}
if (channel_count < CHANNEL_INDEX_MASKS_SIZE &&
CHANNEL_INDEX_MASKS_MAP[channel_count] != AUDIO_CHANNEL_NONE) {
channel_masks[num_channel_masks++] = CHANNEL_INDEX_MASKS_MAP[channel_count];
}
}
return num_channel_masks;
}
static unsigned int populate_sample_rates_from_profile(const alsa_device_profile* profile,
unsigned int sample_rates[])
{
unsigned int num_sample_rates = 0;
for (;num_sample_rates < min(MAX_PROFILE_SAMPLE_RATES, AUDIO_PORT_MAX_SAMPLING_RATES) &&
profile->sample_rates[num_sample_rates] != 0; num_sample_rates++) {
sample_rates[num_sample_rates] = profile->sample_rates[num_sample_rates];
}
return num_sample_rates;
}
/*
* HAl Functions
*/
/**
* NOTE: when multiple mutexes have to be acquired, always respect the
* following order: hw device > out stream
*/
static struct alsa_device_info* stream_get_first_alsa_device(const struct listnode *alsa_devices) {
if (list_empty(alsa_devices)) {
return NULL;
}
return node_to_item(list_head(alsa_devices), struct alsa_device_info, list_node);
}
/**
* Must be called with holding the stream's lock.
*/
static void stream_standby_l(struct listnode *alsa_devices, bool *standby)
{
if (!*standby) {
struct listnode *node;
list_for_each (node, alsa_devices) {
struct alsa_device_info *device_info =
node_to_item(node, struct alsa_device_info, list_node);
proxy_close(&device_info->proxy);
}
*standby = true;
}
}
static void stream_clear_devices(struct listnode *alsa_devices)
{
struct listnode *node, *temp;
struct alsa_device_info *device_info = NULL;
list_for_each_safe (node, temp, alsa_devices) {
device_info = node_to_item(node, struct alsa_device_info, list_node);
if (device_info != NULL) {
list_remove(&device_info->list_node);
free(device_info);
}
}
}
static int stream_set_new_devices(struct pcm_config *config,
struct listnode *alsa_devices,
unsigned int num_devices,
const int cards[],
const int devices[],
int direction)
{
int status = 0;
stream_clear_devices(alsa_devices);
for (unsigned int i = 0; i < num_devices; ++i) {
struct alsa_device_info *device_info =
(struct alsa_device_info *) calloc(1, sizeof(struct alsa_device_info));
profile_init(&device_info->profile, direction);
device_info->profile.card = cards[i];
device_info->profile.device = devices[i];
status = profile_read_device_info(&device_info->profile) ? 0 : -EINVAL;
if (status != 0) {
ALOGE("%s failed to read device info card=%d;device=%d",
__func__, cards[i], devices[i]);
goto exit;
}
status = proxy_prepare(&device_info->proxy, &device_info->profile, config);
if (status != 0) {
ALOGE("%s failed to prepare device card=%d;device=%d",
__func__, cards[i], devices[i]);
goto exit;
}
list_add_tail(alsa_devices, &device_info->list_node);
}
exit:
if (status != 0) {
stream_clear_devices(alsa_devices);
}
return status;
}
static void stream_dump_alsa_devices(const struct listnode *alsa_devices, int fd) {
struct listnode *node;
size_t i = 0;
list_for_each(node, alsa_devices) {
struct alsa_device_info *device_info =
node_to_item(node, struct alsa_device_info, list_node);
dprintf(fd, "Output Profile %zu:\n", i);
profile_dump(&device_info->profile, fd);
dprintf(fd, "Output Proxy %zu:\n", i);
proxy_dump(&device_info->proxy, fd);
}
}
/*
* OUT functions
*/
static uint32_t out_get_sample_rate(const struct audio_stream *stream)
{
struct alsa_device_info *device_info = stream_get_first_alsa_device(
&((struct stream_out*)stream)->alsa_devices);
if (device_info == NULL) {
ALOGW("%s device info is null", __func__);
return 0;
}
uint32_t rate = proxy_get_sample_rate(&device_info->proxy);
ALOGV("out_get_sample_rate() = %d", rate);
return rate;
}
static int out_set_sample_rate(struct audio_stream *stream, uint32_t rate)
{
return 0;
}
static size_t out_get_buffer_size(const struct audio_stream *stream)
{
const struct stream_out* out = (const struct stream_out*)stream;
const struct alsa_device_info* device_info = stream_get_first_alsa_device(&out->alsa_devices);
if (device_info == NULL) {
ALOGW("%s device info is null", __func__);
return 0;
}
return proxy_get_period_size(&device_info->proxy) * audio_stream_out_frame_size(&(out->stream));
}
static uint32_t out_get_channels(const struct audio_stream *stream)
{
const struct stream_out *out = (const struct stream_out*)stream;
return out->hal_channel_mask;
}
static audio_format_t out_get_format(const struct audio_stream *stream)
{
/* Note: The HAL doesn't do any FORMAT conversion at this time. It
* Relies on the framework to provide data in the specified format.
* This could change in the future.
*/
struct alsa_device_info *device_info = stream_get_first_alsa_device(
&((struct stream_out*)stream)->alsa_devices);
if (device_info == NULL) {
ALOGW("%s device info is null", __func__);
return AUDIO_FORMAT_DEFAULT;
}
audio_format_t format = audio_format_from_pcm_format(proxy_get_format(&device_info->proxy));
return format;
}
static int out_set_format(struct audio_stream *stream, audio_format_t format)
{
return 0;
}
static int out_standby(struct audio_stream *stream)
{
struct stream_out *out = (struct stream_out *)stream;
stream_lock(&out->lock);
device_lock(out->adev);
stream_standby_l(&out->alsa_devices, &out->standby);
device_unlock(out->adev);
stream_unlock(&out->lock);
return 0;
}
static int out_dump(const struct audio_stream *stream, int fd) {
const struct stream_out* out_stream = (const struct stream_out*) stream;
if (out_stream != NULL) {
stream_dump_alsa_devices(&out_stream->alsa_devices, fd);
}
return 0;
}
static int out_set_parameters(struct audio_stream *stream __unused, const char *kvpairs)
{
ALOGV("out_set_parameters() keys:%s", kvpairs);
// The set parameters here only matters when the routing devices are changed.
// When the device version is not less than 3.0, the framework will use create
// audio patch API instead of set parameters to chanage audio routing.
return 0;
}
static char * out_get_parameters(const struct audio_stream *stream, const char *keys)
{
struct stream_out *out = (struct stream_out *)stream;
stream_lock(&out->lock);
struct alsa_device_info *device_info = stream_get_first_alsa_device(&out->alsa_devices);
char *params_str = NULL;
if (device_info != NULL) {
params_str = device_get_parameters(&device_info->profile, keys);
}
stream_unlock(&out->lock);
return params_str;
}
static uint32_t out_get_latency(const struct audio_stream_out *stream)
{
struct alsa_device_info *device_info = stream_get_first_alsa_device(
&((struct stream_out*)stream)->alsa_devices);
if (device_info == NULL) {
ALOGW("%s device info is null", __func__);
return 0;
}
return proxy_get_latency(&device_info->proxy);
}
static int out_set_volume(struct audio_stream_out *stream, float left, float right)
{
return -ENOSYS;
}
/* must be called with hw device and output stream mutexes locked */
static int start_output_stream(struct stream_out *out)
{
int status = 0;
struct listnode *node;
list_for_each(node, &out->alsa_devices) {
struct alsa_device_info *device_info =
node_to_item(node, struct alsa_device_info, list_node);
ALOGV("start_output_stream(card:%d device:%d)",
device_info->profile.card, device_info->profile.device);
status = proxy_open(&device_info->proxy);
if (status != 0) {
ALOGE("%s failed to open device(card: %d device: %d)",
__func__, device_info->profile.card, device_info->profile.device);
goto exit;
}
}
exit:
if (status != 0) {
list_for_each(node, &out->alsa_devices) {
struct alsa_device_info *device_info =
node_to_item(node, struct alsa_device_info, list_node);
proxy_close(&device_info->proxy);
}
}
return status;
}
static ssize_t out_write(struct audio_stream_out *stream, const void* buffer, size_t bytes)
{
int ret;
struct stream_out *out = (struct stream_out *)stream;
stream_lock(&out->lock);
if (out->standby) {
ret = start_output_stream(out);
if (ret != 0) {
goto err;
}
out->standby = false;
}
struct listnode* node;
list_for_each(node, &out->alsa_devices) {
struct alsa_device_info* device_info =
node_to_item(node, struct alsa_device_info, list_node);
alsa_device_proxy* proxy = &device_info->proxy;
const void * write_buff = buffer;
int num_write_buff_bytes = bytes;
const int num_device_channels = proxy_get_channel_count(proxy); /* what we told alsa */
const int num_req_channels = out->hal_channel_count; /* what we told AudioFlinger */
if (num_device_channels != num_req_channels) {
/* allocate buffer */
const size_t required_conversion_buffer_size =
bytes * num_device_channels / num_req_channels;
if (required_conversion_buffer_size > out->conversion_buffer_size) {
out->conversion_buffer_size = required_conversion_buffer_size;
out->conversion_buffer = realloc(out->conversion_buffer,
out->conversion_buffer_size);
}
/* convert data */
const audio_format_t audio_format = out_get_format(&(out->stream.common));
const unsigned sample_size_in_bytes = audio_bytes_per_sample(audio_format);
num_write_buff_bytes =
adjust_channels(write_buff, num_req_channels,
out->conversion_buffer, num_device_channels,
sample_size_in_bytes, num_write_buff_bytes);
write_buff = out->conversion_buffer;
}
if (write_buff != NULL && num_write_buff_bytes != 0) {
proxy_write(proxy, write_buff, num_write_buff_bytes);
}
}
stream_unlock(&out->lock);
return bytes;
err:
stream_unlock(&out->lock);
if (ret != 0) {
usleep(bytes * 1000000 / audio_stream_out_frame_size(stream) /
out_get_sample_rate(&stream->common));
}
return bytes;
}
static int out_get_render_position(const struct audio_stream_out *stream, uint32_t *dsp_frames)
{
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; // discard const qualifier
stream_lock(&out->lock);
const struct alsa_device_info* device_info = stream_get_first_alsa_device(&out->alsa_devices);
const int ret = device_info == NULL ? -ENODEV :
proxy_get_presentation_position(&device_info->proxy, frames, timestamp);
stream_unlock(&out->lock);
return ret;
}
static int out_add_audio_effect(const struct audio_stream *stream, effect_handle_t effect)
{
return 0;
}
static int out_remove_audio_effect(const struct audio_stream *stream, effect_handle_t effect)
{
return 0;
}
static int out_get_next_write_timestamp(const struct audio_stream_out *stream, int64_t *timestamp)
{
return -EINVAL;
}
static int adev_open_output_stream(struct audio_hw_device *hw_dev,
audio_io_handle_t handle,
audio_devices_t devicesSpec __unused,
audio_output_flags_t flags,
struct audio_config *config,
struct audio_stream_out **stream_out,
const char *address /*__unused*/)
{
ALOGV("adev_open_output_stream() handle:0x%X, devicesSpec:0x%X, flags:0x%X, addr:%s",
handle, devicesSpec, flags, address);
struct stream_out *out;
out = (struct stream_out *)calloc(1, sizeof(struct stream_out));
if (out == NULL) {
return -ENOMEM;
}
/* setup function pointers */
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;
out->stream.write = out_write;
out->stream.get_render_position = out_get_render_position;
out->stream.get_presentation_position = out_get_presentation_position;
out->stream.get_next_write_timestamp = out_get_next_write_timestamp;
out->handle = handle;
stream_lock_init(&out->lock);
out->adev = (struct audio_device *)hw_dev;
list_init(&out->alsa_devices);
struct alsa_device_info *device_info =
(struct alsa_device_info *)calloc(1, sizeof(struct alsa_device_info));
profile_init(&device_info->profile, PCM_OUT);
// build this to hand to the alsa_device_proxy
struct pcm_config proxy_config = {};
/* Pull out the card/device pair */
parse_card_device_params(address, &device_info->profile.card, &device_info->profile.device);
profile_read_device_info(&device_info->profile);
int ret = 0;
/* Rate */
if (config->sample_rate == 0) {
proxy_config.rate = profile_get_default_sample_rate(&device_info->profile);
} else if (profile_is_sample_rate_valid(&device_info->profile, config->sample_rate)) {
proxy_config.rate = config->sample_rate;
} else {
proxy_config.rate = config->sample_rate =
profile_get_default_sample_rate(&device_info->profile);
ret = -EINVAL;
}
/* TODO: This is a problem if the input does not support this rate */
device_lock(out->adev);
out->adev->device_sample_rate = config->sample_rate;
device_unlock(out->adev);
/* Format */
if (config->format == AUDIO_FORMAT_DEFAULT) {
proxy_config.format = profile_get_default_format(&device_info->profile);
config->format = audio_format_from_pcm_format(proxy_config.format);
} else {
enum pcm_format fmt = pcm_format_from_audio_format(config->format);
if (profile_is_format_valid(&device_info->profile, fmt)) {
proxy_config.format = fmt;
} else {
proxy_config.format = profile_get_default_format(&device_info->profile);
config->format = audio_format_from_pcm_format(proxy_config.format);
ret = -EINVAL;
}
}
/* Channels */
bool calc_mask = false;
if (config->channel_mask == AUDIO_CHANNEL_NONE) {
/* query case */
out->hal_channel_count = profile_get_default_channel_count(&device_info->profile);
calc_mask = true;
} else {
/* explicit case */
out->hal_channel_count = audio_channel_count_from_out_mask(config->channel_mask);
}
/* The Framework is currently limited to no more than this number of channels */
if (out->hal_channel_count > FCC_LIMIT) {
out->hal_channel_count = FCC_LIMIT;
calc_mask = true;
}
if (calc_mask) {
/* need to calculate the mask from channel count either because this is the query case
* or the specified mask isn't valid for this device, or is more than the FW can handle */
config->channel_mask = out->hal_channel_count <= FCC_2
/* position mask for mono and stereo*/
? audio_channel_out_mask_from_count(out->hal_channel_count)
/* otherwise indexed */
: audio_channel_mask_for_index_assignment_from_count(out->hal_channel_count);
}
out->hal_channel_mask = config->channel_mask;
// Validate the "logical" channel count against support in the "actual" profile.
// if they differ, choose the "actual" number of channels *closest* to the "logical".
// and store THAT in proxy_config.channels
proxy_config.channels =
profile_get_closest_channel_count(&device_info->profile, out->hal_channel_count);
proxy_prepare(&device_info->proxy, &device_info->profile, &proxy_config);
out->config = proxy_config;
list_add_tail(&out->alsa_devices, &device_info->list_node);
/* TODO The retry mechanism isn't implemented in AudioPolicyManager/AudioFlinger
* So clear any errors that may have occurred above.
*/
ret = 0;
out->conversion_buffer = NULL;
out->conversion_buffer_size = 0;
out->standby = true;
/* Save the stream for adev_dump() */
adev_add_stream_to_list(out->adev, &out->adev->output_stream_list, &out->list_node);
*stream_out = &out->stream;
return ret;
}
static void adev_close_output_stream(struct audio_hw_device *hw_dev,
struct audio_stream_out *stream)
{
struct stream_out *out = (struct stream_out *)stream;
stream_lock(&out->lock);
/* Close the pcm device */
stream_standby_l(&out->alsa_devices, &out->standby);
stream_clear_devices(&out->alsa_devices);
free(out->conversion_buffer);
out->conversion_buffer = NULL;
out->conversion_buffer_size = 0;
device_lock(out->adev);
list_remove(&out->list_node);
out->adev->device_sample_rate = 0;
device_unlock(out->adev);
stream_unlock(&out->lock);
free(stream);
}
static size_t adev_get_input_buffer_size(const struct audio_hw_device *hw_dev,
const struct audio_config *config)
{
/* TODO This needs to be calculated based on format/channels/rate */
return 320;
}
/*
* IN functions
*/
static uint32_t in_get_sample_rate(const struct audio_stream *stream)
{
struct alsa_device_info *device_info = stream_get_first_alsa_device(
&((const struct stream_in *)stream)->alsa_devices);
if (device_info == NULL) {
ALOGW("%s device info is null", __func__);
return 0;
}
uint32_t rate = proxy_get_sample_rate(&device_info->proxy);
ALOGV("in_get_sample_rate() = %d", rate);
return rate;
}
static int in_set_sample_rate(struct audio_stream *stream, uint32_t rate)
{
ALOGV("in_set_sample_rate(%d) - NOPE", rate);
return -ENOSYS;
}
static size_t in_get_buffer_size(const struct audio_stream *stream)
{
const struct stream_in * in = ((const struct stream_in*)stream);
struct alsa_device_info *device_info = stream_get_first_alsa_device(&in->alsa_devices);
if (device_info == NULL) {
ALOGW("%s device info is null", __func__);
return 0;
}
return proxy_get_period_size(&device_info->proxy) * audio_stream_in_frame_size(&(in->stream));
}
static uint32_t in_get_channels(const struct audio_stream *stream)
{
const struct stream_in *in = (const struct stream_in*)stream;
return in->hal_channel_mask;
}
static audio_format_t in_get_format(const struct audio_stream *stream)
{
struct alsa_device_info *device_info = stream_get_first_alsa_device(
&((const struct stream_in *)stream)->alsa_devices);
if (device_info == NULL) {
ALOGW("%s device info is null", __func__);
return AUDIO_FORMAT_DEFAULT;
}
alsa_device_proxy *proxy = &device_info->proxy;
audio_format_t format = audio_format_from_pcm_format(proxy_get_format(proxy));
return format;
}
static int in_set_format(struct audio_stream *stream, audio_format_t format)
{
ALOGV("in_set_format(%d) - NOPE", format);
return -ENOSYS;
}
static int in_standby(struct audio_stream *stream)
{
struct stream_in *in = (struct stream_in *)stream;
stream_lock(&in->lock);
device_lock(in->adev);
stream_standby_l(&in->alsa_devices, &in->standby);
device_unlock(in->adev);
stream_unlock(&in->lock);
return 0;
}
static int in_dump(const struct audio_stream *stream, int fd)
{
const struct stream_in* in_stream = (const struct stream_in*)stream;
if (in_stream != NULL) {
stream_dump_alsa_devices(&in_stream->alsa_devices, fd);
}
return 0;
}
static int in_set_parameters(struct audio_stream *stream, const char *kvpairs)
{
ALOGV("in_set_parameters() keys:%s", kvpairs);
// The set parameters here only matters when the routing devices are changed.
// When the device version higher than 3.0, the framework will use create_audio_patch
// API instead of set_parameters to change audio routing.
return 0;
}
static char * in_get_parameters(const struct audio_stream *stream, const char *keys)
{
struct stream_in *in = (struct stream_in *)stream;
stream_lock(&in->lock);
struct alsa_device_info *device_info = stream_get_first_alsa_device(&in->alsa_devices);
char *params_str = NULL;
if (device_info != NULL) {
params_str = device_get_parameters(&device_info->profile, keys);
}
stream_unlock(&in->lock);
return params_str;
}
static int in_add_audio_effect(const struct audio_stream *stream, effect_handle_t effect)
{
return 0;
}
static int in_remove_audio_effect(const struct audio_stream *stream, effect_handle_t effect)
{
return 0;
}
static int in_set_gain(struct audio_stream_in *stream, float gain)
{
return 0;
}
/* must be called with hw device and output stream mutexes locked */
static int start_input_stream(struct stream_in *in)
{
// Only care about the first device as only one input device is allowed.
struct alsa_device_info *device_info = stream_get_first_alsa_device(&in->alsa_devices);
if (device_info == NULL) {
return -ENODEV;
}
ALOGV("start_input_stream(card:%d device:%d)",
device_info->profile.card, device_info->profile.device);
return proxy_open(&device_info->proxy);
}
/* TODO mutex stuff here (see out_write) */
static ssize_t in_read(struct audio_stream_in *stream, void* buffer, size_t bytes)
{
size_t num_read_buff_bytes = 0;
void * read_buff = buffer;
void * out_buff = buffer;
int ret = 0;
struct stream_in * in = (struct stream_in *)stream;
stream_lock(&in->lock);
if (in->standby) {
ret = start_input_stream(in);
if (ret != 0) {
goto err;
}
in->standby = false;
}
// Only care about the first device as only one input device is allowed.
struct alsa_device_info *device_info = stream_get_first_alsa_device(&in->alsa_devices);
if (device_info == NULL) {
return 0;
}
/*
* OK, we need to figure out how much data to read to be able to output the requested
* number of bytes in the HAL format (16-bit, stereo).
*/
num_read_buff_bytes = bytes;
int num_device_channels = proxy_get_channel_count(&device_info->proxy); /* what we told Alsa */
int num_req_channels = in->hal_channel_count; /* what we told AudioFlinger */
if (num_device_channels != num_req_channels) {
num_read_buff_bytes = (num_device_channels * num_read_buff_bytes) / num_req_channels;
}
/* Setup/Realloc the conversion buffer (if necessary). */
if (num_read_buff_bytes != bytes) {
if (num_read_buff_bytes > in->conversion_buffer_size) {
/*TODO Remove this when AudioPolicyManger/AudioFlinger support arbitrary formats
(and do these conversions themselves) */
in->conversion_buffer_size = num_read_buff_bytes;
in->conversion_buffer = realloc(in->conversion_buffer, in->conversion_buffer_size);
}
read_buff = in->conversion_buffer;
}
ret = proxy_read(&device_info->proxy, read_buff, num_read_buff_bytes);
if (ret == 0) {
if (num_device_channels != num_req_channels) {
// ALOGV("chans dev:%d req:%d", num_device_channels, num_req_channels);
out_buff = buffer;
/* Num Channels conversion */
if (num_device_channels != num_req_channels) {
audio_format_t audio_format = in_get_format(&(in->stream.common));
unsigned sample_size_in_bytes = audio_bytes_per_sample(audio_format);
num_read_buff_bytes =
adjust_channels(read_buff, num_device_channels,
out_buff, num_req_channels,
sample_size_in_bytes, num_read_buff_bytes);
}
}
/* no need to acquire in->adev->lock to read mic_muted here as we don't change its state */
if (num_read_buff_bytes > 0 && in->adev->mic_muted)
memset(buffer, 0, num_read_buff_bytes);
} else {
num_read_buff_bytes = 0; // reset the value after USB headset is unplugged
}
err:
stream_unlock(&in->lock);
return num_read_buff_bytes;
}
static uint32_t in_get_input_frames_lost(struct audio_stream_in *stream)
{
return 0;
}
static int in_get_capture_position(const struct audio_stream_in *stream,
int64_t *frames, int64_t *time)
{
struct stream_in *in = (struct stream_in *)stream; // discard const qualifier
stream_lock(&in->lock);
struct alsa_device_info *device_info = stream_get_first_alsa_device(&in->alsa_devices);
const int ret = device_info == NULL ? -ENODEV
: proxy_get_capture_position(&device_info->proxy, frames, time);
stream_unlock(&in->lock);
return ret;
}
static int in_get_active_microphones(const struct audio_stream_in *stream,
struct audio_microphone_characteristic_t *mic_array,
size_t *mic_count) {
(void)stream;
(void)mic_array;
(void)mic_count;
return -ENOSYS;
}
static int in_set_microphone_direction(const struct audio_stream_in *stream,
audio_microphone_direction_t dir) {
(void)stream;
(void)dir;
ALOGV("---- in_set_microphone_direction()");
return -ENOSYS;
}
static int in_set_microphone_field_dimension(const struct audio_stream_in *stream, float zoom) {
(void)zoom;
ALOGV("---- in_set_microphone_field_dimension()");
return -ENOSYS;
}
static int adev_open_input_stream(struct audio_hw_device *hw_dev,
audio_io_handle_t handle,
audio_devices_t devicesSpec __unused,
struct audio_config *config,
struct audio_stream_in **stream_in,
audio_input_flags_t flags __unused,
const char *address,
audio_source_t source __unused)
{
ALOGV("adev_open_input_stream() rate:%" PRIu32 ", chanMask:0x%" PRIX32 ", fmt:%" PRIu8,
config->sample_rate, config->channel_mask, config->format);
/* Pull out the card/device pair */
int32_t card, device;
if (!parse_card_device_params(address, &card, &device)) {
ALOGW("%s fail - invalid address %s", __func__, address);
*stream_in = NULL;
return -EINVAL;
}
struct stream_in * const in = (struct stream_in *)calloc(1, sizeof(struct stream_in));
if (in == NULL) {
*stream_in = NULL;
return -ENOMEM;
}
/* setup function pointers */
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->stream.get_capture_position = in_get_capture_position;
in->stream.get_active_microphones = in_get_active_microphones;
in->stream.set_microphone_direction = in_set_microphone_direction;
in->stream.set_microphone_field_dimension = in_set_microphone_field_dimension;
in->handle = handle;
stream_lock_init(&in->lock);
in->adev = (struct audio_device *)hw_dev;
list_init(&in->alsa_devices);
struct alsa_device_info *device_info =
(struct alsa_device_info *)calloc(1, sizeof(struct alsa_device_info));
profile_init(&device_info->profile, PCM_IN);
memset(&in->config, 0, sizeof(in->config));
int ret = 0;
device_lock(in->adev);
int num_open_inputs = in->adev->inputs_open;
device_unlock(in->adev);
/* Check if an input stream is already open */
if (num_open_inputs > 0) {
if (!profile_is_cached_for(&device_info->profile, card, device)) {
ALOGW("%s fail - address card:%d device:%d doesn't match existing profile",
__func__, card, device);
ret = -EINVAL;
}
} else {
/* Read input profile only if necessary */
device_info->profile.card = card;
device_info->profile.device = device;
if (!profile_read_device_info(&device_info->profile)) {
ALOGW("%s fail - cannot read profile", __func__);
ret = -EINVAL;
}
}
if (ret != 0) {
free(in);
*stream_in = NULL;
return ret;
}
/* Rate */
int request_config_rate = config->sample_rate;
if (config->sample_rate == 0) {
config->sample_rate = profile_get_default_sample_rate(&device_info->profile);
}
if (in->adev->device_sample_rate != 0 && /* we are playing, so lock the rate if possible */
in->adev->device_sample_rate >= RATELOCK_THRESHOLD) {/* but only for high sample rates */
if (config->sample_rate != in->adev->device_sample_rate) {
unsigned highest_rate = profile_get_highest_sample_rate(&device_info->profile);
if (highest_rate == 0) {
ret = -EINVAL; /* error with device */
} else {
in->config.rate = config->sample_rate =
min(highest_rate, in->adev->device_sample_rate);
if (request_config_rate != 0 && in->config.rate != config->sample_rate) {
/* Changing the requested rate */
ret = -EINVAL;
} else {
/* Everything AOK! */
ret = 0;
}
}
}
} else if (profile_is_sample_rate_valid(&device_info->profile, config->sample_rate)) {
in->config.rate = config->sample_rate;
} else {
in->config.rate = config->sample_rate =
profile_get_default_sample_rate(&device_info->profile);
ret = -EINVAL;
}
/* Format */
if (config->format == AUDIO_FORMAT_DEFAULT) {
in->config.format = profile_get_default_format(&device_info->profile);
config->format = audio_format_from_pcm_format(in->config.format);
} else {
enum pcm_format fmt = pcm_format_from_audio_format(config->format);
if (profile_is_format_valid(&device_info->profile, fmt)) {
in->config.format = fmt;
} else {
in->config.format = profile_get_default_format(&device_info->profile);
config->format = audio_format_from_pcm_format(in->config.format);
ret = -EINVAL;
}
}
/* Channels */
bool calc_mask = false;
if (config->channel_mask == AUDIO_CHANNEL_NONE) {
/* query case */
in->hal_channel_count = profile_get_default_channel_count(&device_info->profile);
calc_mask = true;
} else {
/* explicit case */
in->hal_channel_count = audio_channel_count_from_in_mask(config->channel_mask);
}
/* The Framework is currently limited to no more than this number of channels */
if (in->hal_channel_count > FCC_LIMIT) {
in->hal_channel_count = FCC_LIMIT;
calc_mask = true;
}
if (calc_mask) {
/* need to calculate the mask from channel count either because this is the query case
* or the specified mask isn't valid for this device, or is more than the FW can handle */
in->hal_channel_mask = in->hal_channel_count <= FCC_2
/* position mask for mono & stereo */
? audio_channel_in_mask_from_count(in->hal_channel_count)
/* otherwise indexed */
: audio_channel_mask_for_index_assignment_from_count(in->hal_channel_count);
// if we change the mask...
if (in->hal_channel_mask != config->channel_mask &&
config->channel_mask != AUDIO_CHANNEL_NONE) {
config->channel_mask = in->hal_channel_mask;
ret = -EINVAL;
}
} else {
in->hal_channel_mask = config->channel_mask;
}
if (ret == 0) {
// Validate the "logical" channel count against support in the "actual" profile.
// if they differ, choose the "actual" number of channels *closest* to the "logical".
// and store THAT in proxy_config.channels
in->config.channels =
profile_get_closest_channel_count(&device_info->profile, in->hal_channel_count);
ret = proxy_prepare(&device_info->proxy, &device_info->profile, &in->config);
if (ret == 0) {
in->standby = true;
in->conversion_buffer = NULL;
in->conversion_buffer_size = 0;
*stream_in = &in->stream;
/* Save this for adev_dump() */
adev_add_stream_to_list(in->adev, &in->adev->input_stream_list, &in->list_node);
} else {
ALOGW("proxy_prepare error %d", ret);
unsigned channel_count = proxy_get_channel_count(&device_info->proxy);
config->channel_mask = channel_count <= FCC_2
? audio_channel_in_mask_from_count(channel_count)
: audio_channel_mask_for_index_assignment_from_count(channel_count);
config->format = audio_format_from_pcm_format(proxy_get_format(&device_info->proxy));
config->sample_rate = proxy_get_sample_rate(&device_info->proxy);
}
}
if (ret != 0) {
// Deallocate this stream on error, because AudioFlinger won't call
// adev_close_input_stream() in this case.
*stream_in = NULL;
free(in);
return ret;
}
list_add_tail(&in->alsa_devices, &device_info->list_node);
device_lock(in->adev);
++in->adev->inputs_open;
device_unlock(in->adev);
return ret;
}
static void adev_close_input_stream(struct audio_hw_device *hw_dev,
struct audio_stream_in *stream)
{
struct stream_in *in = (struct stream_in *)stream;
stream_lock(&in->lock);
device_lock(in->adev);
list_remove(&in->list_node);
--in->adev->inputs_open;
struct alsa_device_info *device_info = stream_get_first_alsa_device(&in->alsa_devices);
if (device_info != NULL) {
ALOGV("adev_close_input_stream(c:%d d:%d)",
device_info->profile.card, device_info->profile.device);
}
LOG_ALWAYS_FATAL_IF(in->adev->inputs_open < 0,
"invalid inputs_open: %d", in->adev->inputs_open);
stream_standby_l(&in->alsa_devices, &in->standby);
device_unlock(in->adev);
stream_clear_devices(&in->alsa_devices);
stream_unlock(&in->lock);
free(in->conversion_buffer);
free(stream);
}
/*
* ADEV Functions
*/
static int adev_set_parameters(struct audio_hw_device *hw_dev, const char *kvpairs)
{
return 0;
}
static char * adev_get_parameters(const struct audio_hw_device *hw_dev, const char *keys)
{
return strdup("");
}
static int adev_init_check(const struct audio_hw_device *hw_dev)
{
return 0;
}
static int adev_set_voice_volume(struct audio_hw_device *hw_dev, float volume)
{
return -ENOSYS;
}
static int adev_set_master_volume(struct audio_hw_device *hw_dev, float volume)
{
return -ENOSYS;
}
static int adev_set_mode(struct audio_hw_device *hw_dev, audio_mode_t mode)
{
return 0;
}
static int adev_set_mic_mute(struct audio_hw_device *hw_dev, bool state)
{
struct audio_device * adev = (struct audio_device *)hw_dev;
device_lock(adev);
adev->mic_muted = state;
device_unlock(adev);
return -ENOSYS;
}
static int adev_get_mic_mute(const struct audio_hw_device *hw_dev, bool *state)
{
return -ENOSYS;
}
static int adev_create_audio_patch(struct audio_hw_device *dev,
unsigned int num_sources,
const struct audio_port_config *sources,
unsigned int num_sinks,
const struct audio_port_config *sinks,
audio_patch_handle_t *handle) {
if (num_sources != 1 || num_sinks == 0 || num_sinks > AUDIO_PATCH_PORTS_MAX) {
// Only accept mix->device and device->mix cases. In that case, the number of sources
// must be 1. The number of sinks must be in the range of (0, AUDIO_PATCH_PORTS_MAX].
return -EINVAL;
}
if (sources[0].type == AUDIO_PORT_TYPE_DEVICE) {
// If source is a device, the number of sinks should be 1.
if (num_sinks != 1 || sinks[0].type != AUDIO_PORT_TYPE_MIX) {
return -EINVAL;
}
} else if (sources[0].type == AUDIO_PORT_TYPE_MIX) {
// If source is a mix, all sinks should be device.
for (unsigned int i = 0; i < num_sinks; i++) {
if (sinks[i].type != AUDIO_PORT_TYPE_DEVICE) {
ALOGE("%s() invalid sink type %#x for mix source", __func__, sinks[i].type);
return -EINVAL;
}
}
} else {
// All other cases are invalid.
return -EINVAL;
}
struct audio_device* adev = (struct audio_device*) dev;
bool generatedPatchHandle = false;
device_lock(adev);
if (*handle == AUDIO_PATCH_HANDLE_NONE) {
*handle = ++adev->next_patch_handle;
generatedPatchHandle = true;
}
int cards[AUDIO_PATCH_PORTS_MAX];
int devices[AUDIO_PATCH_PORTS_MAX];
const struct audio_port_config *port_configs =
sources[0].type == AUDIO_PORT_TYPE_DEVICE ? sources : sinks;
int num_configs = 0;
audio_io_handle_t io_handle = 0;
bool wasStandby = true;
int direction = PCM_OUT;
audio_patch_handle_t *patch_handle = NULL;
struct listnode *alsa_devices = NULL;
struct stream_lock *lock = NULL;
struct pcm_config *config = NULL;
struct stream_in *in = NULL;
struct stream_out *out = NULL;
unsigned int num_saved_devices = 0;
int saved_cards[AUDIO_PATCH_PORTS_MAX];
int saved_devices[AUDIO_PATCH_PORTS_MAX];
struct listnode *node;
// Only handle patches for mix->devices and device->mix case.
if (sources[0].type == AUDIO_PORT_TYPE_DEVICE) {
in = adev_get_stream_in_by_io_handle_l(adev, sinks[0].ext.mix.handle);
if (in == NULL) {
ALOGE("%s()can not find stream with handle(%d)", __func__, sinks[0].ext.mix.handle);
device_unlock(adev);
return -EINVAL;
}
direction = PCM_IN;
wasStandby = in->standby;
io_handle = in->handle;
num_configs = num_sources;
patch_handle = &in->patch_handle;
alsa_devices = &in->alsa_devices;
lock = &in->lock;
config = &in->config;
} else {
out = adev_get_stream_out_by_io_handle_l(adev, sources[0].ext.mix.handle);
if (out == NULL) {
ALOGE("%s()can not find stream with handle(%d)", __func__, sources[0].ext.mix.handle);
device_unlock(adev);
return -EINVAL;
}
direction = PCM_OUT;
wasStandby = out->standby;
io_handle = out->handle;
num_configs = num_sinks;
patch_handle = &out->patch_handle;
alsa_devices = &out->alsa_devices;
lock = &out->lock;
config = &out->config;
}
// Check if the patch handle match the recorded one if a valid patch handle is passed.
if (!generatedPatchHandle && *patch_handle != *handle) {
ALOGE("%s() the patch handle(%d) does not match recorded one(%d) for stream "
"with handle(%d) when creating audio patch",
__func__, *handle, *patch_handle, io_handle);
device_unlock(adev);
return -EINVAL;
}
device_unlock(adev);
for (unsigned int i = 0; i < num_configs; ++i) {
if (!parse_card_device_params(port_configs[i].ext.device.address, &cards[i], &devices[i])) {
ALOGE("%s, failed to parse card and device %s",
__func__, port_configs[i].ext.device.address);
return -EINVAL;
}
}
stream_lock(lock);
list_for_each (node, alsa_devices) {
struct alsa_device_info *device_info =
node_to_item(node, struct alsa_device_info, list_node);
saved_cards[num_saved_devices] = device_info->profile.card;
saved_devices[num_saved_devices++] = device_info->profile.device;
}
device_lock(adev);
stream_standby_l(alsa_devices, out == NULL ? &in->standby : &out->standby);
device_unlock(adev);
// Timestamps:
// Audio timestamps assume continuous PCM frame counts which are maintained
// with the device proxy.transferred variable. Technically it would be better
// associated with in or out stream, not the device; here we save and restore
// using the first alsa device as a simplification.
uint64_t saved_transferred_frames = 0;
struct alsa_device_info *device_info = stream_get_first_alsa_device(alsa_devices);
if (device_info != NULL) saved_transferred_frames = device_info->proxy.transferred;
int ret = stream_set_new_devices(config, alsa_devices, num_configs, cards, devices, direction);
if (ret != 0) {
*handle = generatedPatchHandle ? AUDIO_PATCH_HANDLE_NONE : *handle;
stream_set_new_devices(
config, alsa_devices, num_saved_devices, saved_cards, saved_devices, direction);
} else {
*patch_handle = *handle;
}
// Timestamps: Restore transferred frames.
if (saved_transferred_frames != 0) {
device_info = stream_get_first_alsa_device(alsa_devices);
if (device_info != NULL) device_info->proxy.transferred = saved_transferred_frames;
}
if (!wasStandby) {
device_lock(adev);
if (in != NULL) {
start_input_stream(in);
}
if (out != NULL) {
start_output_stream(out);
}
device_unlock(adev);
}
stream_unlock(lock);
return ret;
}
static int adev_release_audio_patch(struct audio_hw_device *dev,
audio_patch_handle_t patch_handle)
{
struct audio_device* adev = (struct audio_device*) dev;
device_lock(adev);
struct stream_out *out = adev_get_stream_out_by_patch_handle_l(adev, patch_handle);
device_unlock(adev);
if (out != NULL) {
stream_lock(&out->lock);
device_lock(adev);
stream_standby_l(&out->alsa_devices, &out->standby);
device_unlock(adev);
out->patch_handle = AUDIO_PATCH_HANDLE_NONE;
stream_unlock(&out->lock);
return 0;
}
device_lock(adev);
struct stream_in *in = adev_get_stream_in_by_patch_handle_l(adev, patch_handle);
device_unlock(adev);
if (in != NULL) {
stream_lock(&in->lock);
device_lock(adev);
stream_standby_l(&in->alsa_devices, &in->standby);
device_unlock(adev);
in->patch_handle = AUDIO_PATCH_HANDLE_NONE;
stream_unlock(&in->lock);
return 0;
}
ALOGE("%s cannot find stream with patch handle as %d", __func__, patch_handle);
return -EINVAL;
}
static int adev_get_audio_port(struct audio_hw_device *dev, struct audio_port *port)
{
if (port->type != AUDIO_PORT_TYPE_DEVICE) {
return -EINVAL;
}
alsa_device_profile profile;
const bool is_output = audio_is_output_device(port->ext.device.type);
profile_init(&profile, is_output ? PCM_OUT : PCM_IN);
if (!parse_card_device_params(port->ext.device.address, &profile.card, &profile.device)) {
return -EINVAL;
}
if (!profile_read_device_info(&profile)) {
return -ENOENT;
}
port->num_formats = 0;;
for (size_t i = 0; i < min(MAX_PROFILE_FORMATS, AUDIO_PORT_MAX_FORMATS) &&
profile.formats[i] != 0; ++i) {
audio_format_t format = audio_format_from(profile.formats[i]);
if (format != AUDIO_FORMAT_INVALID) {
port->formats[port->num_formats++] = format;
}
}
port->num_sample_rates = populate_sample_rates_from_profile(&profile, port->sample_rates);
port->num_channel_masks = populate_channel_mask_from_profile(
&profile, is_output, port->channel_masks);
return 0;
}
static int adev_get_audio_port_v7(struct audio_hw_device *dev, struct audio_port_v7 *port)
{
if (port->type != AUDIO_PORT_TYPE_DEVICE) {
return -EINVAL;
}
alsa_device_profile profile;
const bool is_output = audio_is_output_device(port->ext.device.type);
profile_init(&profile, is_output ? PCM_OUT : PCM_IN);
if (!parse_card_device_params(port->ext.device.address, &profile.card, &profile.device)) {
return -EINVAL;
}
if (!profile_read_device_info(&profile)) {
return -ENOENT;
}
audio_channel_mask_t channel_masks[AUDIO_PORT_MAX_CHANNEL_MASKS];
unsigned int num_channel_masks = populate_channel_mask_from_profile(
&profile, is_output, channel_masks);
unsigned int sample_rates[AUDIO_PORT_MAX_SAMPLING_RATES];
const unsigned int num_sample_rates =
populate_sample_rates_from_profile(&profile, sample_rates);
port->num_audio_profiles = 0;;
for (size_t i = 0; i < min(MAX_PROFILE_FORMATS, AUDIO_PORT_MAX_AUDIO_PROFILES) &&
profile.formats[i] != 0; ++i) {
audio_format_t format = audio_format_from(profile.formats[i]);
if (format == AUDIO_FORMAT_INVALID) {
continue;
}
const unsigned int j = port->num_audio_profiles++;
port->audio_profiles[j].format = format;
port->audio_profiles[j].num_sample_rates = num_sample_rates;
memcpy(port->audio_profiles[j].sample_rates,
sample_rates,
num_sample_rates * sizeof(unsigned int));
port->audio_profiles[j].num_channel_masks = num_channel_masks;
memcpy(port->audio_profiles[j].channel_masks,
channel_masks,
num_channel_masks* sizeof(audio_channel_mask_t));
}
return 0;
}
static int adev_dump(const struct audio_hw_device *device, int fd)
{
dprintf(fd, "\nUSB audio module:\n");
struct audio_device* adev = (struct audio_device*)device;
const int kNumRetries = 3;
const int kSleepTimeMS = 500;
// use device_try_lock() in case we dumpsys during a deadlock
int retry = kNumRetries;
while (retry > 0 && device_try_lock(adev) != 0) {
sleep(kSleepTimeMS);
retry--;
}
if (retry > 0) {
if (list_empty(&adev->output_stream_list)) {
dprintf(fd, " No output streams.\n");
} else {
struct listnode* node;
list_for_each(node, &adev->output_stream_list) {
struct audio_stream* stream =
(struct audio_stream *)node_to_item(node, struct stream_out, list_node);
out_dump(stream, fd);
}
}
if (list_empty(&adev->input_stream_list)) {
dprintf(fd, "\n No input streams.\n");
} else {
struct listnode* node;
list_for_each(node, &adev->input_stream_list) {
struct audio_stream* stream =
(struct audio_stream *)node_to_item(node, struct stream_in, list_node);
in_dump(stream, fd);
}
}
device_unlock(adev);
} else {
// Couldn't lock
dprintf(fd, " Could not obtain device lock.\n");
}
return 0;
}
static int adev_close(hw_device_t *device)
{
free(device);
return 0;
}
static int adev_open(const hw_module_t* module, const char* name, hw_device_t** device)
{
if (strcmp(name, AUDIO_HARDWARE_INTERFACE) != 0)
return -EINVAL;
struct audio_device *adev = calloc(1, sizeof(struct audio_device));
if (!adev)
return -ENOMEM;
pthread_mutex_init(&adev->lock, (const pthread_mutexattr_t *) NULL);
list_init(&adev->output_stream_list);
list_init(&adev->input_stream_list);
adev->hw_device.common.tag = HARDWARE_DEVICE_TAG;
adev->hw_device.common.version = AUDIO_DEVICE_API_VERSION_3_2;
adev->hw_device.common.module = (struct hw_module_t *)module;
adev->hw_device.common.close = adev_close;
adev->hw_device.init_check = adev_init_check;
adev->hw_device.set_voice_volume = adev_set_voice_volume;
adev->hw_device.set_master_volume = adev_set_master_volume;
adev->hw_device.set_mode = adev_set_mode;
adev->hw_device.set_mic_mute = adev_set_mic_mute;
adev->hw_device.get_mic_mute = adev_get_mic_mute;
adev->hw_device.set_parameters = adev_set_parameters;
adev->hw_device.get_parameters = adev_get_parameters;
adev->hw_device.get_input_buffer_size = adev_get_input_buffer_size;
adev->hw_device.open_output_stream = adev_open_output_stream;
adev->hw_device.close_output_stream = adev_close_output_stream;
adev->hw_device.open_input_stream = adev_open_input_stream;
adev->hw_device.close_input_stream = adev_close_input_stream;
adev->hw_device.create_audio_patch = adev_create_audio_patch;
adev->hw_device.release_audio_patch = adev_release_audio_patch;
adev->hw_device.get_audio_port = adev_get_audio_port;
adev->hw_device.get_audio_port_v7 = adev_get_audio_port_v7;
adev->hw_device.dump = adev_dump;
*device = &adev->hw_device.common;
return 0;
}
static struct hw_module_methods_t hal_module_methods = {
.open = adev_open,
};
struct audio_module HAL_MODULE_INFO_SYM = {
.common = {
.tag = HARDWARE_MODULE_TAG,
.module_api_version = AUDIO_MODULE_API_VERSION_0_1,
.hal_api_version = HARDWARE_HAL_API_VERSION,
.id = AUDIO_HARDWARE_MODULE_ID,
.name = "USB audio HW HAL",
.author = "The Android Open Source Project",
.methods = &hal_module_methods,
},
};