From 4fb7e945f86c143e69bb95e77213d45b2e278950 Mon Sep 17 00:00:00 2001
From: "J. Nick Koston" <nick@koston.org>
Date: Thu, 21 Dec 2023 16:59:24 -1000
Subject: [PATCH] Fix unexpected disconnects when outgoing buffer is full
 during keepalive (#5988)

---
 esphome/components/api/api_connection.cpp | 24 +++++++++++++++++++----
 esphome/components/api/api_connection.h   |  3 +++
 2 files changed, 23 insertions(+), 4 deletions(-)

diff --git a/esphome/components/api/api_connection.cpp b/esphome/components/api/api_connection.cpp
index d5ab00a822..4ebfb7582e 100644
--- a/esphome/components/api/api_connection.cpp
+++ b/esphome/components/api/api_connection.cpp
@@ -118,7 +118,9 @@ void APIConnection::loop() {
   this->list_entities_iterator_.advance();
   this->initial_state_iterator_.advance();
 
-  const uint32_t keepalive = 60000;
+  static uint32_t keepalive = 60000;
+  static uint8_t max_ping_retries = 60;
+  static uint16_t ping_retry_interval = 1000;
   const uint32_t now = millis();
   if (this->sent_ping_) {
     // Disconnect if not responded within 2.5*keepalive
@@ -126,10 +128,24 @@ void APIConnection::loop() {
       on_fatal_error();
       ESP_LOGW(TAG, "%s didn't respond to ping request in time. Disconnecting...", this->client_combined_info_.c_str());
     }
-  } else if (now - this->last_traffic_ > keepalive) {
+  } else if (now - this->last_traffic_ > keepalive && now > this->next_ping_retry_) {
     ESP_LOGVV(TAG, "Sending keepalive PING...");
-    this->sent_ping_ = true;
-    this->send_ping_request(PingRequest());
+    this->sent_ping_ = this->send_ping_request(PingRequest());
+    if (!this->sent_ping_) {
+      this->next_ping_retry_ = now + ping_retry_interval;
+      this->ping_retries_++;
+      if (this->ping_retries_ >= max_ping_retries) {
+        on_fatal_error();
+        ESP_LOGE(TAG, "%s: Sending keepalive failed %d time(s). Disconnecting...", this->client_combined_info_.c_str(),
+                 this->ping_retries_);
+      } else if (this->ping_retries_ >= 10) {
+        ESP_LOGW(TAG, "%s: Sending keepalive failed %d time(s), will retry in %d ms",
+                 this->client_combined_info_.c_str(), this->ping_retries_, ping_retry_interval);
+      } else {
+        ESP_LOGD(TAG, "%s: Sending keepalive failed %d time(s), will retry in %d ms",
+                 this->client_combined_info_.c_str(), this->ping_retries_, ping_retry_interval);
+      }
+    }
   }
 
 #ifdef USE_ESP32_CAMERA
diff --git a/esphome/components/api/api_connection.h b/esphome/components/api/api_connection.h
index 09b595bb71..9d01468807 100644
--- a/esphome/components/api/api_connection.h
+++ b/esphome/components/api/api_connection.h
@@ -140,6 +140,7 @@ class APIConnection : public APIServerConnection {
   void on_disconnect_response(const DisconnectResponse &value) override;
   void on_ping_response(const PingResponse &value) override {
     // we initiated ping
+    this->ping_retries_ = 0;
     this->sent_ping_ = false;
   }
   void on_home_assistant_state_response(const HomeAssistantStateResponse &msg) override;
@@ -217,6 +218,8 @@ class APIConnection : public APIServerConnection {
   bool state_subscription_{false};
   int log_subscription_{ESPHOME_LOG_LEVEL_NONE};
   uint32_t last_traffic_;
+  uint32_t next_ping_retry_{0};
+  uint8_t ping_retries_{0};
   bool sent_ping_{false};
   bool service_call_subscription_{false};
   bool next_close_ = false;