Merge branch 'wifi_fail_too_quickly_fix' into wifi_timeout

esphome/components/wifi/wifi_component.cpp

@@ -205,6 +205,21 @@ static constexpr uint32_t WIFI_COOLDOWN_DURATION_MS = 500;
 /// While connecting, WiFi can't beacon the AP properly, so needs longer cooldown
 static constexpr uint32_t WIFI_COOLDOWN_WITH_AP_ACTIVE_MS = 30000;
 
+/// Timeout for WiFi scan operations
+/// This is a fallback in case we don't receive a scan done callback from the WiFi driver.
+/// Normal scans complete via callback; this only triggers if something goes wrong.
+static constexpr uint32_t WIFI_SCAN_TIMEOUT_MS = 31000;
+
+/// Timeout for WiFi connection attempts
+/// This is a fallback in case we don't receive connection success/failure callbacks.
+/// Some platforms (especially LibreTiny/Beken) can take 30-60 seconds to connect,
+/// particularly with fast_connect enabled where no prior scan provides channel info.
+/// Do not lower this value - connection failures are detected via callbacks, not timeout.
+/// If this timeout fires prematurely while a connection is still in progress, it causes
+/// cascading failures: the subsequent scan will also fail because the WiFi driver is
+/// still busy with the previous connection attempt.
+static constexpr uint32_t WIFI_CONNECT_TIMEOUT_MS = 46000;
+
 static constexpr uint8_t get_max_retries_for_phase(WiFiRetryPhase phase) {
   switch (phase) {
     case WiFiRetryPhase::INITIAL_CONNECT:
@@ -1035,7 +1050,7 @@ __attribute__((noinline)) static void log_scan_result(const WiFiScanResult &res)
 
 void WiFiComponent::check_scanning_finished() {
   if (!this->scan_done_) {
-    if (millis() - this->action_started_ > 30000) {
+    if (millis() - this->action_started_ > WIFI_SCAN_TIMEOUT_MS) {
       ESP_LOGE(TAG, "Scan timeout");
       this->retry_connect();
     }
@@ -1184,8 +1199,9 @@ void WiFiComponent::check_connecting_finished() {
   }
 
   uint32_t now = millis();
-  if (now - this->action_started_ > 30000) {
-    ESP_LOGW(TAG, "Connection timeout");
+  if (now - this->action_started_ > WIFI_CONNECT_TIMEOUT_MS) {
+    ESP_LOGW(TAG, "Connection timeout, aborting connection attempt");
+    this->wifi_disconnect_();
     this->retry_connect();
     return;
   }
@@ -1405,6 +1421,10 @@ bool WiFiComponent::transition_to_phase_(WiFiRetryPhase new_phase) {
       // without disrupting the captive portal/improv connection
       if (!this->is_captive_portal_active_() && !this->is_esp32_improv_active_()) {
         this->restart_adapter();
+      } else {
+        // Even when skipping full restart, disconnect to clear driver state
+        // Without this, platforms like LibreTiny may think we're still connecting
+        this->wifi_disconnect_();
       }
       // Clear scan flag - we're starting a new retry cycle
       this->did_scan_this_cycle_ = false;

esphome/components/wifi/wifi_component_esp_idf.cpp

@@ -720,6 +720,7 @@ void WiFiComponent::wifi_process_event_(IDFWiFiEvent *data) {
   } else if (data->event_base == WIFI_EVENT && data->event_id == WIFI_EVENT_STA_STOP) {
     ESP_LOGV(TAG, "STA stop");
     s_sta_started = false;
+    s_sta_connecting = false;
 
   } else if (data->event_base == WIFI_EVENT && data->event_id == WIFI_EVENT_STA_AUTHMODE_CHANGE) {
     const auto &it = data->data.sta_authmode_change;

esphome/components/wifi/wifi_component_libretiny.cpp

@@ -291,6 +291,7 @@ void WiFiComponent::wifi_event_callback_(esphome_wifi_event_id_t event, esphome_
     }
     case ESPHOME_EVENT_ID_WIFI_STA_STOP: {
       ESP_LOGV(TAG, "STA stop");
+      s_sta_connecting = false;
       break;
     }
     case ESPHOME_EVENT_ID_WIFI_STA_CONNECTED: {
@@ -322,7 +323,7 @@ void WiFiComponent::wifi_event_callback_(esphome_wifi_event_id_t event, esphome_
       // wifi_sta_connect_status_() to return IDLE. The main loop then sees
       // "Unknown connection status 0" (wifi_component.cpp check_connecting_finished)
       // and calls retry_connect(), aborting a connection that may succeed moments later.
-      // Real connection failures will have ssid/bssid populated, or we'll hit the 30s timeout.
+      // Real connection failures will have ssid/bssid populated, or we'll hit the connection timeout.
       if (it.ssid_len == 0 && s_sta_connecting) {
         ESP_LOGV(TAG, "Ignoring disconnect event with empty ssid while connecting (reason=%s)",
                  get_disconnect_reason_str(it.reason));
@@ -527,7 +528,12 @@ bool WiFiComponent::wifi_start_ap_(const WiFiAP &ap) {
 network::IPAddress WiFiComponent::wifi_soft_ap_ip() { return {WiFi.softAPIP()}; }
 #endif  // USE_WIFI_AP
 
-bool WiFiComponent::wifi_disconnect_() { return WiFi.disconnect(); }
+bool WiFiComponent::wifi_disconnect_() {
+  // Clear connecting flag first so disconnect events aren't ignored
+  // and wifi_sta_connect_status_() returns IDLE instead of CONNECTING
+  s_sta_connecting = false;
+  return WiFi.disconnect();
+}
 
 bssid_t WiFiComponent::wifi_bssid() {
   bssid_t bssid{};

esphome/espota2.py

@@ -322,8 +322,8 @@ def perform_ota(
         hash_func, nonce_size, hash_name = _AUTH_METHODS[auth]
         perform_auth(sock, password, hash_func, nonce_size, hash_name)
 
-    # Set higher timeout during upload
-    sock.settimeout(30.0)
+    # Timeout must match device-side OTA_SOCKET_TIMEOUT_DATA to prevent premature failures
+    sock.settimeout(90.0)
 
     upload_size = len(upload_contents)
     upload_size_encoded = [