fix(multimodal): harden playback and client lifecycle handling

- retry the server connection prompt after transient playback failures
- serialize multimodal client and session start-stop operations
- tag downstream responses to ignore stale packets and playback events
- complete bounded teardown and clean callbacks after partial startup
This commit is contained in:
chenjintao
2026-08-14 15:24:29 +08:00
parent c56c65462e
commit 0cfb10252c
12 changed files with 436 additions and 188 deletions
@@ -154,8 +154,9 @@ esp_err_t control_plane_start_multimodal_client(void);
* Semantics:
* - Idempotent and safe to call repeatedly.
* - Requests cooperative stop first, then applies bounded waits for worker teardown.
* - Returns an explicit error if teardown does not complete.
*/
void control_plane_stop_multimodal_client(void);
esp_err_t control_plane_stop_multimodal_client(void);
esp_err_t control_plane_register_model_command_callback(control_plane_model_command_callback_t cb,
void *user_data);
@@ -96,7 +96,7 @@ typedef enum {
} control_plane_transport_loss_reason_t;
esp_err_t control_plane_multimodal_start(void);
void control_plane_multimodal_stop(void);
esp_err_t control_plane_multimodal_stop(void);
bool control_plane_multimodal_is_connected(void);
esp_err_t control_plane_multimodal_send_start(void);
esp_err_t control_plane_multimodal_send_stop(const char *reason);
@@ -11,6 +11,7 @@
#include <string.h>
#include "freertos/FreeRTOS.h"
#include "freertos/semphr.h"
#include "freertos/task.h"
#include "esp_log.h"
#include "sdkconfig.h"
@@ -40,6 +41,9 @@
#define CONTROL_PLANE_IDLE_TRANSPORT_SHUTDOWN_MS 30000
#define CONTROL_PLANE_WATCHDOG_TASK_STACK_SIZE 6144
#define CONTROL_PLANE_SERVER_SUCCESS_PROMPT_FILE "/littlefs/audio_prompts/server_connect_success.opus"
#define CONTROL_PLANE_CLIENT_LIFECYCLE_WAIT_MS 30000U
#define CONTROL_PLANE_SERVER_SUCCESS_PROMPT_RETRY_MS 1000U
#define CONTROL_PLANE_SERVER_SUCCESS_PROMPT_MAX_ATTEMPTS 3U
#ifndef CONFIG_TQ_NETWORK_TASK_CORE
#define CONFIG_TQ_NETWORK_TASK_CORE 0
@@ -57,6 +61,10 @@ static bool s_multimodal_watchdog_stop_requested;
static bool s_multimodal_watchdog_task_starting;
static bool s_model_command_dispatched;
static bool s_server_success_prompt_played;
static bool s_server_success_prompt_pending;
static bool s_server_success_prompt_in_progress;
static uint8_t s_server_success_prompt_attempts;
static uint32_t s_server_success_prompt_next_attempt_ms;
static bool s_cloud_chouqian_pending;
static bool s_downstream_stall_recovery_pending;
static bool s_downstream_stall_stop_sent;
@@ -64,6 +72,45 @@ static uint32_t s_downstream_stall_recovery_started_ms;
static control_plane_model_command_callback_t s_model_command_cb;
static void *s_model_command_cb_user_data;
static portMUX_TYPE s_multimodal_runtime_lock = portMUX_INITIALIZER_UNLOCKED;
static SemaphoreHandle_t s_client_lifecycle_mutex;
#if ( configSUPPORT_STATIC_ALLOCATION == 1 )
static StaticSemaphore_t s_client_lifecycle_mutex_buf;
#endif
static portMUX_TYPE s_client_lifecycle_init_lock = portMUX_INITIALIZER_UNLOCKED;
static bool lock_client_lifecycle(void)
{
portENTER_CRITICAL(&s_client_lifecycle_init_lock);
if (s_client_lifecycle_mutex == NULL) {
#if ( configSUPPORT_STATIC_ALLOCATION == 1 )
s_client_lifecycle_mutex =
xSemaphoreCreateMutexStatic(&s_client_lifecycle_mutex_buf);
#else
s_client_lifecycle_mutex = xSemaphoreCreateMutex();
#endif
}
SemaphoreHandle_t mutex = s_client_lifecycle_mutex;
portEXIT_CRITICAL(&s_client_lifecycle_init_lock);
if (mutex == NULL) {
return false;
}
TickType_t timeout_ticks = pdMS_TO_TICKS(CONTROL_PLANE_CLIENT_LIFECYCLE_WAIT_MS);
if (timeout_ticks == 0) {
timeout_ticks = 1;
}
if (xSemaphoreTake(mutex, timeout_ticks) != pdTRUE) {
ESP_LOGW(TAG, "Timeout acquiring multimodal client lifecycle lock");
return false;
}
return true;
}
static void unlock_client_lifecycle(void)
{
if (s_client_lifecycle_mutex != NULL) {
(void)xSemaphoreGive(s_client_lifecycle_mutex);
}
}
static TaskHandle_t get_multimodal_task_handle(TaskHandle_t *slot)
{
@@ -171,6 +218,74 @@ static void control_plane_clear_downstream_stall_recovery(void)
portEXIT_CRITICAL(&s_multimodal_runtime_lock);
}
static void control_plane_schedule_server_success_prompt(void)
{
uint32_t now_ms = (uint32_t)(xTaskGetTickCount() * portTICK_PERIOD_MS);
portENTER_CRITICAL(&s_multimodal_runtime_lock);
if (!s_server_success_prompt_played &&
s_server_success_prompt_attempts <
CONTROL_PLANE_SERVER_SUCCESS_PROMPT_MAX_ATTEMPTS) {
s_server_success_prompt_pending = true;
s_server_success_prompt_next_attempt_ms = now_ms;
}
portEXIT_CRITICAL(&s_multimodal_runtime_lock);
}
static void control_plane_try_server_success_prompt(
const control_plane_multimodal_runtime_state_t *state,
uint32_t now_ms)
{
if (state == NULL || !state->connected || state->has_active_task) {
return;
}
bool should_attempt = false;
uint8_t attempt = 0U;
portENTER_CRITICAL(&s_multimodal_runtime_lock);
if (s_server_success_prompt_pending &&
!s_server_success_prompt_played &&
!s_server_success_prompt_in_progress &&
s_server_success_prompt_attempts <
CONTROL_PLANE_SERVER_SUCCESS_PROMPT_MAX_ATTEMPTS &&
(int32_t)(now_ms - s_server_success_prompt_next_attempt_ms) >= 0) {
s_server_success_prompt_in_progress = true;
s_server_success_prompt_attempts++;
attempt = s_server_success_prompt_attempts;
should_attempt = true;
}
portEXIT_CRITICAL(&s_multimodal_runtime_lock);
if (!should_attempt) {
return;
}
esp_err_t err = domain_runtime_behavior_play_prompt_file(
CONTROL_PLANE_SERVER_SUCCESS_PROMPT_FILE);
portENTER_CRITICAL(&s_multimodal_runtime_lock);
s_server_success_prompt_in_progress = false;
if (err == ESP_OK) {
s_server_success_prompt_played = true;
s_server_success_prompt_pending = false;
} else if (s_server_success_prompt_attempts >=
CONTROL_PLANE_SERVER_SUCCESS_PROMPT_MAX_ATTEMPTS) {
s_server_success_prompt_pending = false;
} else {
s_server_success_prompt_next_attempt_ms =
now_ms + CONTROL_PLANE_SERVER_SUCCESS_PROMPT_RETRY_MS;
}
portEXIT_CRITICAL(&s_multimodal_runtime_lock);
if (err == ESP_OK) {
ESP_LOGI(TAG, "Server success prompt played on attempt %u", (unsigned)attempt);
} else {
ESP_LOGW(TAG,
"Server success prompt attempt %u/%u failed: %s",
(unsigned)attempt,
(unsigned)CONTROL_PLANE_SERVER_SUCCESS_PROMPT_MAX_ATTEMPTS,
esp_err_to_name(err));
}
}
static bool try_begin_multimodal_task_start(TaskHandle_t *slot, bool *start_flag)
{
bool can_start = false;
@@ -507,6 +622,7 @@ static void multimodal_watchdog_task(void *arg)
}
uint32_t now_ms = control_plane_now_ms();
control_plane_try_server_success_prompt(&state, now_ms);
bool stall_stop_sent = false;
uint32_t stall_recovery_started_ms = 0U;
if (control_plane_get_downstream_stall_recovery(&stall_stop_sent,
@@ -559,7 +675,7 @@ static void multimodal_watchdog_task(void *arg)
"Idle multimodal transport timeout, stopping client after %u ms",
(unsigned)CONTROL_PLANE_IDLE_TRANSPORT_SHUTDOWN_MS);
control_plane_multimodal_stop_uplink();
control_plane_multimodal_stop();
(void)control_plane_multimodal_stop();
idle_transport_started_ms = 0U;
}
continue;
@@ -627,16 +743,7 @@ static void control_plane_multimodal_event_bridge(control_plane_multimodal_event
switch (event) {
case CONTROL_PLANE_MULTIMODAL_EVENT_TRANSPORT_CONNECTED: {
bool should_play_prompt = false;
portENTER_CRITICAL(&s_multimodal_runtime_lock);
if (!s_server_success_prompt_played) {
s_server_success_prompt_played = true;
should_play_prompt = true;
}
portEXIT_CRITICAL(&s_multimodal_runtime_lock);
if (should_play_prompt) {
(void)domain_runtime_behavior_play_prompt_file(CONTROL_PLANE_SERVER_SUCCESS_PROMPT_FILE);
}
control_plane_schedule_server_success_prompt();
break;
}
@@ -718,7 +825,7 @@ static void control_plane_multimodal_event_bridge(control_plane_multimodal_event
}
}
esp_err_t control_plane_start_multimodal_client(void)
static esp_err_t control_plane_start_multimodal_client_locked(void)
{
control_plane_clear_downstream_stall_recovery();
@@ -768,18 +875,35 @@ esp_err_t control_plane_start_multimodal_client(void)
esp_err_t err = control_plane_multimodal_start();
if (err != ESP_OK) {
stop_watchdog_task();
(void)control_plane_multimodal_register_downstream_audio_callback(NULL, NULL);
(void)control_plane_multimodal_register_event_callback(NULL, NULL);
return err;
}
return ESP_OK;
}
void control_plane_stop_multimodal_client(void)
esp_err_t control_plane_start_multimodal_client(void)
{
if (!lock_client_lifecycle()) {
return ESP_ERR_TIMEOUT;
}
esp_err_t err = control_plane_start_multimodal_client_locked();
unlock_client_lifecycle();
return err;
}
esp_err_t control_plane_stop_multimodal_client(void)
{
if (!lock_client_lifecycle()) {
return ESP_ERR_TIMEOUT;
}
control_plane_clear_downstream_stall_recovery();
control_plane_clear_cloud_chouqian_pending("multimodal client stopped");
stop_watchdog_task();
control_plane_multimodal_stop();
esp_err_t err = control_plane_multimodal_stop();
unlock_client_lifecycle();
return err;
}
esp_err_t control_plane_register_model_command_callback(
@@ -88,6 +88,7 @@
#define CONTROL_PLANE_UPLINK_READY_WAIT_MS 6000U
#define CONTROL_PLANE_AUDIO_SEND_FAILURE_WINDOW_MS 100U
#define CONTROL_PLANE_AUDIO_SEND_FAILURE_THRESHOLD 2U
#define CONTROL_PLANE_MULTIMODAL_LIFECYCLE_WAIT_MS 30000U
#ifndef CONFIG_TQ_LOCAL_TASK_CORE
#define CONFIG_TQ_LOCAL_TASK_CORE 1
@@ -154,6 +155,7 @@ static bool s_downstream_stall_stop_pending;
static bool s_drop_downstream_audio;
static bool s_received_downstream_audio;
static bool s_current_response_received_audio;
static uint32_t s_current_playback_generation;
static uint32_t s_audio_send_failure_count;
static TickType_t s_audio_send_failure_window_start_tick;
static size_t s_last_asr_text_len;
@@ -171,9 +173,12 @@ static control_plane_multimodal_downstream_audio_callback_t s_downstream_audio_c
static void *s_downstream_audio_cb_user_data;
static SemaphoreHandle_t s_state_mutex;
static SemaphoreHandle_t s_lifecycle_mutex;
#if ( configSUPPORT_STATIC_ALLOCATION == 1 )
static StaticSemaphore_t s_state_mutex_buf;
static StaticSemaphore_t s_lifecycle_mutex_buf;
#endif
static portMUX_TYPE s_mutex_init_lock = portMUX_INITIALIZER_UNLOCKED;
typedef enum {
LOCAL_RESPONDING_EVENT_STARTED = 0,
@@ -299,23 +304,60 @@ static void log_downstream_binary_frame(size_t len, bool drop_audio, bool has_ca
}
}
static bool ensure_session_mutexes(void)
{
portENTER_CRITICAL(&s_mutex_init_lock);
if (s_state_mutex == NULL) {
#if ( configSUPPORT_STATIC_ALLOCATION == 1 )
s_state_mutex = xSemaphoreCreateMutexStatic(&s_state_mutex_buf);
#else
s_state_mutex = xSemaphoreCreateMutex();
#endif
}
if (s_lifecycle_mutex == NULL) {
#if ( configSUPPORT_STATIC_ALLOCATION == 1 )
s_lifecycle_mutex = xSemaphoreCreateMutexStatic(&s_lifecycle_mutex_buf);
#else
s_lifecycle_mutex = xSemaphoreCreateMutex();
#endif
}
bool ready = s_state_mutex != NULL && s_lifecycle_mutex != NULL;
portEXIT_CRITICAL(&s_mutex_init_lock);
return ready;
}
static bool ensure_state_mutex(void)
{
if (s_state_mutex != NULL) {
return true;
}
#if ( configSUPPORT_STATIC_ALLOCATION == 1 )
s_state_mutex = xSemaphoreCreateMutexStatic(&s_state_mutex_buf);
#else
s_state_mutex = xSemaphoreCreateMutex();
#endif
if (s_state_mutex == NULL) {
ESP_LOGE(TAG, "Failed to create multimodal state mutex");
if (!ensure_session_mutexes()) {
ESP_LOGE(TAG, "Failed to create multimodal lifecycle mutexes");
return false;
}
return true;
}
static bool lock_lifecycle(void)
{
if (!ensure_session_mutexes()) {
return false;
}
TickType_t timeout_ticks = pdMS_TO_TICKS(CONTROL_PLANE_MULTIMODAL_LIFECYCLE_WAIT_MS);
if (timeout_ticks == 0) {
timeout_ticks = 1;
}
if (xSemaphoreTake(s_lifecycle_mutex, timeout_ticks) != pdTRUE) {
ESP_LOGW(TAG, "Timeout acquiring multimodal lifecycle lock");
return false;
}
return true;
}
static void unlock_lifecycle(void)
{
if (s_lifecycle_mutex != NULL) {
(void)xSemaphoreGive(s_lifecycle_mutex);
}
}
static bool lock_state(void)
{
if (!ensure_state_mutex()) {
@@ -402,6 +444,7 @@ static void reset_session_state_locked(void)
s_drop_downstream_audio = false;
s_received_downstream_audio = false;
s_current_response_received_audio = false;
s_current_playback_generation = 0U;
s_audio_send_failure_count = 0;
s_audio_send_failure_window_start_tick = 0;
s_local_responding_started_send_pending = false;
@@ -1403,7 +1446,13 @@ static void process_text_json(const char *json_text)
unlock_state();
}
ESP_LOGI(TAG, "TTS responding started");
esp_err_t begin_err = domain_runtime_multimodal_audio_begin_downstream_response();
uint32_t playback_generation = 0U;
esp_err_t begin_err = domain_runtime_multimodal_audio_begin_downstream_response(
&playback_generation);
if (begin_err == ESP_OK && lock_state()) {
s_current_playback_generation = playback_generation;
unlock_state();
}
if (begin_err != ESP_OK && begin_err != ESP_ERR_INVALID_STATE) {
ESP_LOGW(TAG,
"Begin downstream response failed: %s",
@@ -1636,6 +1685,21 @@ static void downstream_playback_event_bridge(
return;
}
bool generation_matches = false;
if (lock_state()) {
generation_matches = s_current_playback_generation != 0U &&
s_current_playback_generation ==
event->playback_generation;
unlock_state();
}
if (!generation_matches) {
ESP_LOGD(TAG,
"Ignore stale or local playback event type=%d generation=%u",
(int)event->type,
(unsigned)event->playback_generation);
return;
}
switch (event->type) {
case DOMAIN_MULTIMODAL_DOWNSTREAM_PLAYBACK_STARTED: {
bool should_send_started = false;
@@ -1656,12 +1720,17 @@ static void downstream_playback_event_bridge(
case DOMAIN_MULTIMODAL_DOWNSTREAM_PLAYBACK_COMPLETED:
if (lock_state()) {
s_downstream_playback_completed = true;
s_current_playback_generation = 0U;
unlock_state();
}
maybe_send_local_responding_ended_if_ready();
break;
case DOMAIN_MULTIMODAL_DOWNSTREAM_PLAYBACK_STOPPED:
if (lock_state()) {
s_current_playback_generation = 0U;
unlock_state();
}
send_local_responding_ended_after_playback_stop();
break;
@@ -1671,6 +1740,7 @@ static void downstream_playback_event_bridge(
cloud_response_active = s_task_id[0] != '\0' &&
(s_remote_responding ||
s_current_response_received_audio);
s_current_playback_generation = 0U;
unlock_state();
}
@@ -1697,7 +1767,20 @@ static esp_err_t control_plane_audio_uplink_send_cb(const uint8_t *data, size_t
return control_plane_multimodal_send_audio_realtime(data, len);
}
static esp_err_t control_plane_multimodal_start_locked(void);
static esp_err_t control_plane_multimodal_stop_locked(void);
esp_err_t control_plane_multimodal_start(void)
{
if (!lock_lifecycle()) {
return ESP_ERR_TIMEOUT;
}
esp_err_t err = control_plane_multimodal_start_locked();
unlock_lifecycle();
return err;
}
static esp_err_t control_plane_multimodal_start_locked(void)
{
#if !CONFIG_ENABLE_MULTIMODAL_INTERACTION
ESP_LOGW(TAG, "Multimodal interaction is disabled by config");
@@ -1707,25 +1790,26 @@ esp_err_t control_plane_multimodal_start(void)
#if CONFIG_TQ_MULTIMODAL_DIAGNOSTICS_ENABLE
uint32_t session_generation = 0U;
#endif
if (lock_state()) {
bool already_started = s_started || s_auth.token[0] != '\0';
bool starting = s_starting;
bool stopping = s_stopping;
if (already_started) {
unlock_state();
return ESP_OK;
}
if (starting || stopping) {
unlock_state();
return ESP_ERR_INVALID_STATE;
}
s_starting = true;
#if CONFIG_TQ_MULTIMODAL_DIAGNOSTICS_ENABLE
s_session_generation++;
session_generation = s_session_generation;
#endif
unlock_state();
if (!lock_state()) {
return ESP_ERR_TIMEOUT;
}
bool already_started = s_started || s_auth.token[0] != '\0';
bool starting = s_starting;
bool stopping = s_stopping;
if (already_started) {
unlock_state();
return ESP_OK;
}
if (starting || stopping) {
unlock_state();
return ESP_ERR_INVALID_STATE;
}
s_starting = true;
#if CONFIG_TQ_MULTIMODAL_DIAGNOSTICS_ENABLE
s_session_generation++;
session_generation = s_session_generation;
#endif
unlock_state();
#if CONFIG_TQ_MULTIMODAL_DIAGNOSTICS_ENABLE
ESP_LOGI(TAG,
@@ -1804,14 +1888,28 @@ esp_err_t control_plane_multimodal_start(void)
return err;
}
if (lock_state()) {
s_auth = auth;
unlock_state();
if (!lock_state()) {
(void)domain_runtime_multimodal_transport_stop(2000);
(void)domain_runtime_multimodal_audio_register_downstream_playback_callback(NULL, NULL);
stop_local_responding_event_worker();
return ESP_ERR_TIMEOUT;
}
s_auth = auth;
unlock_state();
return ESP_OK;
}
void control_plane_multimodal_stop(void)
esp_err_t control_plane_multimodal_stop(void)
{
if (!lock_lifecycle()) {
return ESP_ERR_TIMEOUT;
}
esp_err_t err = control_plane_multimodal_stop_locked();
unlock_lifecycle();
return err;
}
static esp_err_t control_plane_multimodal_stop_locked(void)
{
bool already_stopping = false;
if (lock_state()) {
@@ -1823,8 +1921,7 @@ void control_plane_multimodal_stop(void)
unlock_state();
}
if (already_stopping) {
ESP_LOGI(TAG, "Multimodal stop already in progress");
return;
ESP_LOGI(TAG, "Completing previously started multimodal teardown");
}
control_plane_multimodal_stop_uplink();
@@ -1846,7 +1943,10 @@ void control_plane_multimodal_stop(void)
memset(&s_auth, 0, sizeof(s_auth));
reset_session_state_locked();
unlock_state();
} else if (transport_stop_err == ESP_OK) {
transport_stop_err = ESP_ERR_TIMEOUT;
}
return transport_stop_err;
}
bool control_plane_multimodal_is_connected(void)
@@ -1777,7 +1777,7 @@ static void control_plane_wifi_state_task(void *arg)
if (!disconnected_expression_active) {
ESP_LOGI(TAG, "Wi-Fi STA disconnect persisted; stopping multimodal client");
control_plane_stop_multimodal_client();
(void)control_plane_stop_multimodal_client();
control_plane_stop_user_config_server();
}
@@ -2455,7 +2455,7 @@ void control_plane_stop_runtime_workers(void)
portEXIT_CRITICAL(&s_runtime_lock);
control_plane_delete_model_cmd_queue(queue, queue_with_caps);
control_plane_stop_multimodal_client();
(void)control_plane_stop_multimodal_client();
control_plane_set_runtime_stop_requested(false);
}
+3 -1
View File
@@ -103,6 +103,7 @@ typedef enum {
typedef struct {
domain_multimodal_downstream_playback_event_type_t type;
uint32_t playback_generation;
} domain_multimodal_downstream_playback_event_t;
typedef void (*domain_multimodal_downstream_playback_callback_t)(
@@ -228,7 +229,8 @@ esp_err_t domain_runtime_local_command_stop(int timeout_ms);
esp_err_t domain_runtime_local_command_get_snapshot(
domain_local_command_snapshot_t *out_snapshot);
esp_err_t domain_runtime_multimodal_audio_play_downstream_opus(const uint8_t *data, size_t len);
esp_err_t domain_runtime_multimodal_audio_begin_downstream_response(void);
esp_err_t domain_runtime_multimodal_audio_begin_downstream_response(
uint32_t *out_playback_generation);
esp_err_t domain_runtime_multimodal_audio_finish_downstream_response(void);
esp_err_t domain_runtime_multimodal_audio_stop_downstream_playback(void);
esp_err_t domain_runtime_multimodal_audio_register_downstream_playback_callback(
@@ -435,7 +435,9 @@ static const char *domain_wifi_prompt_file_path(domain_wifi_prompt_id_t prompt)
}
}
static void domain_downstream_playback_dispatch(domain_multimodal_downstream_playback_event_type_t type)
static void domain_downstream_playback_dispatch(
domain_multimodal_downstream_playback_event_type_t type,
uint32_t playback_generation)
{
domain_multimodal_downstream_playback_callback_t cb = NULL;
void *cb_user_data = NULL;
@@ -446,32 +448,41 @@ static void domain_downstream_playback_dispatch(domain_multimodal_downstream_pla
const domain_multimodal_downstream_playback_event_t event = {
.type = type,
.playback_generation = playback_generation,
};
cb(&event, cb_user_data);
}
static void domain_downstream_playback_started_adapter(void *user_data)
static void domain_downstream_playback_started_adapter(uint32_t playback_generation,
void *user_data)
{
(void)user_data;
domain_downstream_playback_dispatch(DOMAIN_MULTIMODAL_DOWNSTREAM_PLAYBACK_STARTED);
domain_downstream_playback_dispatch(DOMAIN_MULTIMODAL_DOWNSTREAM_PLAYBACK_STARTED,
playback_generation);
}
static void domain_downstream_playback_completed_adapter(void *user_data)
static void domain_downstream_playback_completed_adapter(uint32_t playback_generation,
void *user_data)
{
(void)user_data;
domain_downstream_playback_dispatch(DOMAIN_MULTIMODAL_DOWNSTREAM_PLAYBACK_COMPLETED);
domain_downstream_playback_dispatch(DOMAIN_MULTIMODAL_DOWNSTREAM_PLAYBACK_COMPLETED,
playback_generation);
}
static void domain_downstream_playback_stopped_adapter(void *user_data)
static void domain_downstream_playback_stopped_adapter(uint32_t playback_generation,
void *user_data)
{
(void)user_data;
domain_downstream_playback_dispatch(DOMAIN_MULTIMODAL_DOWNSTREAM_PLAYBACK_STOPPED);
domain_downstream_playback_dispatch(DOMAIN_MULTIMODAL_DOWNSTREAM_PLAYBACK_STOPPED,
playback_generation);
}
static void domain_downstream_playback_stalled_adapter(void *user_data)
static void domain_downstream_playback_stalled_adapter(uint32_t playback_generation,
void *user_data)
{
(void)user_data;
domain_downstream_playback_dispatch(DOMAIN_MULTIMODAL_DOWNSTREAM_PLAYBACK_STALLED);
domain_downstream_playback_dispatch(DOMAIN_MULTIMODAL_DOWNSTREAM_PLAYBACK_STALLED,
playback_generation);
}
static void domain_wifi_set_keep_awake_adapter(bool keep_awake, void *user_data)
@@ -687,9 +698,10 @@ esp_err_t domain_runtime_multimodal_audio_play_downstream_opus(const uint8_t *da
return platform_multimodal_audio_play_downstream_opus(data, len);
}
esp_err_t domain_runtime_multimodal_audio_begin_downstream_response(void)
esp_err_t domain_runtime_multimodal_audio_begin_downstream_response(
uint32_t *out_playback_generation)
{
return platform_multimodal_audio_begin_downstream_response();
return platform_multimodal_audio_begin_downstream_response(out_playback_generation);
}
esp_err_t domain_runtime_multimodal_audio_finish_downstream_response(void)
+5 -5
View File
@@ -43,10 +43,10 @@ typedef struct {
} platform_multimodal_transport_callbacks_t;
typedef struct {
void (*on_started)(void *user_data);
void (*on_completed)(void *user_data);
void (*on_stopped)(void *user_data);
void (*on_stalled)(void *user_data);
void (*on_started)(uint32_t playback_generation, void *user_data);
void (*on_completed)(uint32_t playback_generation, void *user_data);
void (*on_stopped)(uint32_t playback_generation, void *user_data);
void (*on_stalled)(uint32_t playback_generation, void *user_data);
void *user_data;
} platform_multimodal_downstream_playback_callbacks_t;
@@ -399,7 +399,7 @@ esp_err_t platform_multimodal_audio_unregister_pcm_observer(
void *user_data,
int timeout_ms);
esp_err_t platform_multimodal_audio_play_downstream_opus(const uint8_t *data, size_t len);
esp_err_t platform_multimodal_audio_begin_downstream_response(void);
esp_err_t platform_multimodal_audio_begin_downstream_response(uint32_t *out_playback_generation);
esp_err_t platform_multimodal_audio_finish_downstream_response(void);
esp_err_t platform_multimodal_audio_stop_downstream_playback(void);
esp_err_t platform_multimodal_audio_register_downstream_playback_callbacks(
@@ -25,6 +25,7 @@ typedef enum {
typedef struct {
multimodal_audio_downstream_playback_event_type_t type;
uint32_t playback_generation;
} multimodal_audio_downstream_playback_event_t;
typedef void (*multimodal_audio_downstream_playback_callback_t)(
@@ -47,7 +48,7 @@ esp_err_t multimodal_audio_unregister_pcm_observer(
int timeout_ms);
esp_err_t multimodal_audio_handle_downstream_opus(const uint8_t *data, size_t len);
esp_err_t multimodal_audio_begin_downstream_response(void);
esp_err_t multimodal_audio_begin_downstream_response(uint32_t *out_playback_generation);
esp_err_t multimodal_audio_finish_downstream_response(void);
esp_err_t multimodal_audio_stop_downstream_playback(void);
esp_err_t multimodal_audio_register_downstream_playback_callback(
@@ -461,16 +461,6 @@ static void drain_downstream_playback_queue(multimodal_audio_ctx_t *ctx)
unlock_downstream_playback_queue();
}
static void drain_downstream_playback_queue_now(multimodal_audio_ctx_t *ctx)
{
if (ctx == NULL || !lock_downstream_playback_queue(0)) {
return;
}
QueueHandle_t queue = ctx->downstream_playback_queue;
drain_downstream_playback_queue_handle(queue);
unlock_downstream_playback_queue();
}
static void track_uplink_packet_queued(size_t len)
{
#if CONFIG_MULTIMODAL_AUDIO_LOG_UPLINK_LEVELS
@@ -581,10 +571,11 @@ static esp_err_t set_downstream_output_mute_if_needed(multimodal_audio_ctx_t *ct
static esp_err_t ensure_downstream_idle_mute_task(multimodal_audio_ctx_t *ctx);
static void notify_downstream_audio_active(multimodal_audio_ctx_t *ctx);
static void notify_downstream_playback_started(uint32_t expected_generation);
static void notify_downstream_playback_completed(void);
static void notify_downstream_playback_stopped(void);
static void notify_downstream_playback_completed(uint32_t expected_generation);
static esp_err_t prime_downstream_output_path(multimodal_audio_ctx_t *ctx);
static esp_err_t handle_downstream_opus_sync(const uint8_t *data, size_t len);
static esp_err_t handle_downstream_opus_sync(const uint8_t *data,
size_t len,
uint32_t expected_generation);
static esp_err_t ensure_downstream_playback_task(multimodal_audio_ctx_t *ctx);
static esp_err_t ensure_uplink_worker_tasks(multimodal_audio_ctx_t *ctx);
static esp_err_t stop_downstream_idle_mute_task(multimodal_audio_ctx_t *ctx, int timeout_ms);
@@ -1563,20 +1554,44 @@ static void snapshot_downstream_playback_state(uint32_t *out_generation,
portEXIT_CRITICAL(&s_audio_runtime_lock);
}
static uint32_t advance_downstream_playback_generation(void)
static uint32_t advance_downstream_playback_generation_locked(void)
{
uint32_t generation = 0U;
portENTER_CRITICAL(&s_audio_runtime_lock);
s_audio.downstream_playback_generation++;
if (s_audio.downstream_playback_generation == 0U) {
s_audio.downstream_playback_generation = 1U;
}
s_audio.downstream_decoder_reset_pending = true;
generation = s_audio.downstream_playback_generation;
portEXIT_CRITICAL(&s_audio_runtime_lock);
return s_audio.downstream_playback_generation;
}
return generation;
static bool claim_downstream_response_terminal(uint32_t expected_generation,
bool invalidate_generation,
uint32_t *out_terminal_generation)
{
if (out_terminal_generation == NULL) {
return false;
}
portENTER_CRITICAL(&s_audio_runtime_lock);
bool generation_matches = expected_generation == 0U ||
s_audio.downstream_playback_generation ==
expected_generation;
bool should_notify = generation_matches &&
(s_audio.downstream_response_active ||
s_audio.downstream_response_end_received ||
s_audio.downstream_playback_in_progress);
*out_terminal_generation = s_audio.downstream_playback_generation;
if (generation_matches && invalidate_generation &&
(should_notify || expected_generation == 0U)) {
(void)advance_downstream_playback_generation_locked();
}
if (should_notify) {
s_audio.downstream_response_active = false;
s_audio.downstream_response_end_received = false;
s_audio.downstream_playback_in_progress = false;
}
portEXIT_CRITICAL(&s_audio_runtime_lock);
return should_notify;
}
static bool clear_downstream_decoder_reset_pending(uint32_t expected_generation)
@@ -1628,7 +1643,8 @@ static void snapshot_downstream_playback_callback(
}
static void dispatch_downstream_playback_event(
multimodal_audio_downstream_playback_event_type_t type)
multimodal_audio_downstream_playback_event_type_t type,
uint32_t playback_generation)
{
multimodal_audio_downstream_playback_callback_t cb = NULL;
void *cb_ctx = NULL;
@@ -1639,6 +1655,7 @@ static void dispatch_downstream_playback_event(
multimodal_audio_downstream_playback_event_t event = {
.type = type,
.playback_generation = playback_generation,
};
cb(&event, cb_ctx);
}
@@ -1656,51 +1673,32 @@ static void notify_downstream_playback_started(uint32_t expected_generation)
portEXIT_CRITICAL(&s_audio_runtime_lock);
if (should_notify) {
dispatch_downstream_playback_event(MULTIMODAL_AUDIO_DOWNSTREAM_PLAYBACK_STARTED);
dispatch_downstream_playback_event(MULTIMODAL_AUDIO_DOWNSTREAM_PLAYBACK_STARTED,
expected_generation);
}
}
static void notify_downstream_playback_finished(
multimodal_audio_downstream_playback_event_type_t type)
static void notify_downstream_playback_completed(uint32_t expected_generation)
{
bool should_notify = false;
portENTER_CRITICAL(&s_audio_runtime_lock);
if (s_audio.downstream_playback_in_progress) {
s_audio.downstream_playback_in_progress = false;
should_notify = true;
}
portEXIT_CRITICAL(&s_audio_runtime_lock);
if (should_notify) {
dispatch_downstream_playback_event(type);
uint32_t playback_generation = 0U;
if (claim_downstream_response_terminal(expected_generation,
false,
&playback_generation)) {
dispatch_downstream_playback_event(MULTIMODAL_AUDIO_DOWNSTREAM_PLAYBACK_COMPLETED,
playback_generation);
}
}
static void notify_downstream_playback_completed(void)
{
bool should_notify = false;
portENTER_CRITICAL(&s_audio_runtime_lock);
if (s_audio.downstream_playback_in_progress) {
s_audio.downstream_playback_in_progress = false;
should_notify = true;
} else if (s_audio.downstream_response_end_received) {
should_notify = true;
}
portEXIT_CRITICAL(&s_audio_runtime_lock);
if (should_notify) {
dispatch_downstream_playback_event(MULTIMODAL_AUDIO_DOWNSTREAM_PLAYBACK_COMPLETED);
}
}
static bool snapshot_downstream_response_end_received(void)
static bool snapshot_downstream_response_end_received(
uint32_t *out_playback_generation)
{
bool end_received = false;
portENTER_CRITICAL(&s_audio_runtime_lock);
end_received = s_audio.downstream_response_end_received;
if (out_playback_generation != NULL) {
*out_playback_generation = s_audio.downstream_playback_generation;
}
portEXIT_CRITICAL(&s_audio_runtime_lock);
return end_received;
@@ -1717,27 +1715,24 @@ static bool snapshot_downstream_response_active(void)
return active;
}
static TickType_t snapshot_downstream_last_audio_tick(const multimodal_audio_ctx_t *ctx)
static bool snapshot_downstream_response_watchdog(
const multimodal_audio_ctx_t *ctx,
uint32_t *out_playback_generation,
TickType_t *out_last_audio_tick,
TickType_t *out_last_response_activity_tick)
{
TickType_t tick = 0;
portENTER_CRITICAL(&s_audio_runtime_lock);
if (ctx != NULL) {
tick = ctx->downstream_last_audio_tick;
if (ctx == NULL || out_playback_generation == NULL ||
out_last_audio_tick == NULL || out_last_response_activity_tick == NULL) {
return false;
}
portEXIT_CRITICAL(&s_audio_runtime_lock);
return tick;
}
static TickType_t snapshot_downstream_last_response_activity_tick(
const multimodal_audio_ctx_t *ctx)
{
TickType_t tick = 0;
portENTER_CRITICAL(&s_audio_runtime_lock);
if (ctx != NULL) {
tick = ctx->downstream_last_response_activity_tick;
}
bool active = s_audio.downstream_response_active;
*out_playback_generation = s_audio.downstream_playback_generation;
*out_last_audio_tick = ctx->downstream_last_audio_tick;
*out_last_response_activity_tick = ctx->downstream_last_response_activity_tick;
portEXIT_CRITICAL(&s_audio_runtime_lock);
return tick;
return active;
}
static void update_downstream_last_audio_tick(multimodal_audio_ctx_t *ctx,
@@ -1769,37 +1764,28 @@ static void update_downstream_response_state(bool active, bool end_received)
portEXIT_CRITICAL(&s_audio_runtime_lock);
}
static void notify_downstream_playback_stopped(void)
{
notify_downstream_playback_finished(MULTIMODAL_AUDIO_DOWNSTREAM_PLAYBACK_STOPPED);
}
static void notify_downstream_playback_stalled(void)
{
notify_downstream_playback_finished(MULTIMODAL_AUDIO_DOWNSTREAM_PLAYBACK_STALLED);
}
static void abort_downstream_response_after_active_stall(multimodal_audio_ctx_t *ctx,
uint32_t idle_ms)
uint32_t idle_ms,
uint32_t expected_generation)
{
(void)ctx;
uint32_t stopped_generation = 0U;
if (!claim_downstream_response_terminal(expected_generation,
true,
&stopped_generation)) {
return;
}
ESP_LOGW(TAG,
"Downstream protection watchdog: abort stalled response idle_ms=%u",
(unsigned)idle_ms);
(void)advance_downstream_playback_generation();
notify_downstream_playback_stalled();
update_downstream_response_state(false, false);
if (audio_get_downstream_playback_task(ctx) != NULL ||
audio_get_downstream_playback_starting(ctx)) {
audio_set_downstream_playback_stopping(ctx, true);
audio_set_downstream_playback_keep_stopping_on_exit(ctx, false);
audio_set_downstream_playback_stop_requested(ctx, true);
}
drain_downstream_playback_queue_now(ctx);
dispatch_downstream_playback_event(MULTIMODAL_AUDIO_DOWNSTREAM_PLAYBACK_STALLED,
stopped_generation);
}
static void maybe_notify_downstream_playback_completed_after_drain(const char *reason)
{
if (!snapshot_downstream_response_end_received()) {
uint32_t playback_generation = 0U;
if (!snapshot_downstream_response_end_received(&playback_generation)) {
return;
}
@@ -1820,8 +1806,7 @@ static void maybe_notify_downstream_playback_completed_after_drain(const char *r
"Downstream playback completed after response end (%s)",
reason ? reason : "drained");
log_downstream_diagnostics("playback_complete", true);
notify_downstream_playback_completed();
update_downstream_response_state(false, false);
notify_downstream_playback_completed(playback_generation);
}
static esp_err_t set_downstream_output_mute_if_needed(multimodal_audio_ctx_t *ctx,
@@ -1933,11 +1918,17 @@ static void downstream_idle_mute_timer_cb(void *arg)
return;
}
const bool response_active = snapshot_downstream_response_active();
uint32_t playback_generation = 0U;
TickType_t last_audio_tick = 0;
TickType_t last_response_activity_tick = 0;
const bool response_active = snapshot_downstream_response_watchdog(
ctx,
&playback_generation,
&last_audio_tick,
&last_response_activity_tick);
if (response_active) {
log_downstream_active_diagnostics();
TickType_t now = xTaskGetTickCount();
TickType_t last_audio_tick = snapshot_downstream_last_audio_tick(ctx);
uint32_t idle_ms = last_audio_tick != 0U ?
(uint32_t)pdTICKS_TO_MS(now - last_audio_tick) : 0U;
uint32_t idle_timeout_ms = get_downstream_idle_mute_timeout_ms();
@@ -1960,14 +1951,15 @@ static void downstream_idle_mute_timer_cb(void *arg)
return;
}
TickType_t last_activity_tick = snapshot_downstream_last_response_activity_tick(ctx);
uint32_t response_idle_ms = last_activity_tick != 0U ?
(uint32_t)pdTICKS_TO_MS(now - last_activity_tick) : idle_ms;
uint32_t response_idle_ms = last_response_activity_tick != 0U ?
(uint32_t)pdTICKS_TO_MS(now - last_response_activity_tick) : idle_ms;
if (response_idle_ms < stall_timeout_ms) {
restart_downstream_idle_mute_timer(ctx, stall_timeout_ms - response_idle_ms);
return;
}
abort_downstream_response_after_active_stall(ctx, idle_ms);
abort_downstream_response_after_active_stall(ctx,
idle_ms,
playback_generation);
return;
}
@@ -2221,7 +2213,7 @@ static void downstream_playback_task(void *arg)
if (generation_started && queued_packets > 0U && first_recovery_packet_tick == 0U) {
first_recovery_packet_tick = now;
}
bool response_ended = snapshot_downstream_response_end_received();
bool response_ended = snapshot_downstream_response_end_received(NULL);
if (queued_packets >= target_packets || (response_ended && queued_packets > 0U)) {
playback_buffer_ready = true;
uint32_t buffered_ms = (uint32_t)queued_packets * ctx->opus_frame_duration_ms;
@@ -2288,7 +2280,7 @@ static void downstream_playback_task(void *arg)
}
if (received != pdTRUE || packet.data == NULL || packet.len == 0U) {
if (snapshot_downstream_response_end_received()) {
if (snapshot_downstream_response_end_received(NULL)) {
maybe_notify_downstream_playback_completed_after_drain("queue_empty");
} else if (playback_buffer_ready && generation_started &&
snapshot_downstream_response_active()) {
@@ -2338,7 +2330,9 @@ static void downstream_playback_task(void *arg)
}
if (!audio_get_downstream_playback_stop_requested(ctx)) {
esp_err_t err = handle_downstream_opus_sync(packet.data, packet.len);
esp_err_t err = handle_downstream_opus_sync(packet.data,
packet.len,
packet.playback_generation);
if (err != ESP_OK && err != ESP_ERR_INVALID_STATE) {
note_downstream_playback_failure();
ESP_LOGW(TAG, "Downstream playback packet failed: %s", esp_err_to_name(err));
@@ -2586,9 +2580,11 @@ esp_err_t multimodal_audio_stop_downstream_playback(void)
{
multimodal_audio_ctx_t *ctx = &s_audio;
update_downstream_response_state(false, false);
(void)advance_downstream_playback_generation();
notify_downstream_playback_stopped();
uint32_t stopped_generation = 0U;
if (claim_downstream_response_terminal(0U, true, &stopped_generation)) {
dispatch_downstream_playback_event(MULTIMODAL_AUDIO_DOWNSTREAM_PLAYBACK_STOPPED,
stopped_generation);
}
if (audio_get_deinit_in_progress(ctx)) {
esp_err_t task_stop_err = stop_downstream_playback_task(ctx, 1000, true);
@@ -2608,7 +2604,9 @@ esp_err_t multimodal_audio_stop_downstream_playback(void)
return set_downstream_output_mute_if_needed(ctx, true, "Mute");
}
static esp_err_t handle_downstream_opus_sync(const uint8_t *data, size_t len)
static esp_err_t handle_downstream_opus_sync(const uint8_t *data,
size_t len,
uint32_t expected_generation)
{
multimodal_audio_ctx_t *ctx = &s_audio;
if (!ctx->initialized || !ctx->codec_dev || !ctx->opus_dec) {
@@ -2635,6 +2633,9 @@ static esp_err_t handle_downstream_opus_sync(const uint8_t *data, size_t len)
uint32_t playback_generation = 0U;
bool decoder_reset_pending = false;
snapshot_downstream_playback_state(&playback_generation, &decoder_reset_pending);
if (playback_generation != expected_generation) {
return ESP_ERR_INVALID_STATE;
}
if (decoder_reset_pending) {
esp_err_t reset_err = reset_downstream_decoder(ctx);
if (reset_err != ESP_OK) {
@@ -2929,17 +2930,21 @@ esp_err_t multimodal_audio_handle_downstream_opus(const uint8_t *data, size_t le
return ESP_OK;
}
esp_err_t multimodal_audio_begin_downstream_response(void)
esp_err_t multimodal_audio_begin_downstream_response(
uint32_t *out_playback_generation)
{
multimodal_audio_ctx_t *ctx = &s_audio;
if (out_playback_generation == NULL) {
return ESP_ERR_INVALID_ARG;
}
*out_playback_generation = 0U;
if (!ctx->initialized || ctx->codec_dev == NULL || ctx->opus_dec == NULL) {
return ESP_ERR_INVALID_STATE;
}
uint32_t playback_generation = advance_downstream_playback_generation();
TickType_t begin_tick = xTaskGetTickCount();
portENTER_CRITICAL(&s_audio_runtime_lock);
uint32_t playback_generation = advance_downstream_playback_generation_locked();
ctx->downstream_response_active = true;
ctx->downstream_response_end_received = false;
ctx->downstream_output_primed = false;
@@ -2947,6 +2952,7 @@ esp_err_t multimodal_audio_begin_downstream_response(void)
ctx->downstream_last_audio_tick = begin_tick;
ctx->downstream_last_response_activity_tick = begin_tick;
portEXIT_CRITICAL(&s_audio_runtime_lock);
*out_playback_generation = playback_generation;
s_downstream_queued_packet_total = 0U;
s_downstream_played_packet_total = 0U;
@@ -235,7 +235,8 @@ esp_err_t voice_prompt_play_file(const char *path)
return ESP_ERR_NOT_FOUND;
}
esp_err_t err = multimodal_audio_begin_downstream_response();
uint32_t playback_generation = 0U;
esp_err_t err = multimodal_audio_begin_downstream_response(&playback_generation);
if (err != ESP_OK) {
ESP_LOGW(VOICE_TAG, "Begin prompt playback failed: %s", esp_err_to_name(err));
fclose(file);
@@ -200,9 +200,10 @@ esp_err_t platform_multimodal_audio_play_downstream_opus(const uint8_t *data, si
return multimodal_audio_handle_downstream_opus(data, len);
}
esp_err_t platform_multimodal_audio_begin_downstream_response(void)
esp_err_t platform_multimodal_audio_begin_downstream_response(
uint32_t *out_playback_generation)
{
return multimodal_audio_begin_downstream_response();
return multimodal_audio_begin_downstream_response(out_playback_generation);
}
esp_err_t platform_multimodal_audio_finish_downstream_response(void)
@@ -230,22 +231,22 @@ static void platform_multimodal_downstream_playback_bridge(
switch (event->type) {
case MULTIMODAL_AUDIO_DOWNSTREAM_PLAYBACK_STARTED:
if (callbacks.on_started != NULL) {
callbacks.on_started(callbacks.user_data);
callbacks.on_started(event->playback_generation, callbacks.user_data);
}
break;
case MULTIMODAL_AUDIO_DOWNSTREAM_PLAYBACK_COMPLETED:
if (callbacks.on_completed != NULL) {
callbacks.on_completed(callbacks.user_data);
callbacks.on_completed(event->playback_generation, callbacks.user_data);
}
break;
case MULTIMODAL_AUDIO_DOWNSTREAM_PLAYBACK_STOPPED:
if (callbacks.on_stopped != NULL) {
callbacks.on_stopped(callbacks.user_data);
callbacks.on_stopped(event->playback_generation, callbacks.user_data);
}
break;
case MULTIMODAL_AUDIO_DOWNSTREAM_PLAYBACK_STALLED:
if (callbacks.on_stalled != NULL) {
callbacks.on_stalled(callbacks.user_data);
callbacks.on_stalled(event->playback_generation, callbacks.user_data);
}
break;
default: