feat: improve EC800K transparent TCP recovery

applications/VK_Client.c

@@ -35,10 +35,10 @@ no_carrier_match_t _no_carrier_t = {0};
 
 /*关闭TCP连接*/
 void client_tcp_close(Cclient_t *client) {
+  client->connect_sta = false;
   rt_sprintf(cmd_str, AT_QICLOSE, client->connect_ID);
-  BSP_LTE_SendCmd(cmd_str, DEFAULT_EXPECT, DEFAULT_ERROR, LTE_CMD_WAIT_NORMAL);
-  // client->connect_sta = false;
-  // client->error_cnt = 0;
+  BSP_LTE_SendCmd(cmd_str, DEFAULT_EXPECT, DEFAULT_ERROR, LTE_CMD_WAIT_QICLOSE);
+  client->error_cnt = 0;
 }
 
 static int no_carrier_char_parser(uint8_t ch){
@@ -139,11 +139,71 @@ static void no_carrier_match(uint8_t * at_buf,rt_size_t len){
   }
 }
 
+static const char *disconnect_reason_name(lte_disconnect_reason_t reason) {
+  switch (reason) {
+    case LTE_DISC_CLOSED: return "CLOSED";
+    case LTE_DISC_NO_CARRIER: return "NO_CARRIER";
+    case LTE_DISC_NO_CARRIER_SPLIT: return "NO_CARRIER_SPLIT";
+    case LTE_DISC_AT_TIMEOUT: return "AT_TIMEOUT";
+    case LTE_DISC_QIOPEN_FAILED: return "QIOPEN_FAILED";
+    case LTE_DISC_MODULE_RESET: return "MODULE_RESET";
+    default: return "UNKNOWN";
+  }
+}
+
+static rt_tick_t diag_age(rt_tick_t now, rt_tick_t tick) {
+  return tick == 0 ? 0 : now - tick;
+}
+
+void lte_diag_disconnect(lte_disconnect_reason_t reason) {
+  rt_tick_t now = rt_tick_get();
+  rt_size_t heap_total = 0;
+  rt_size_t heap_used = 0;
+  rt_size_t heap_max = 0;
+
+  rt_memory_info(&heap_total, &heap_used, &heap_max);
+  lte_diag.disconnect_count++;
+  lte_diag.last_disconnect_reason = reason;
+  lte_diag.last_disconnect_tick = now;
+
+  LOG_I("[DISC] reason=%s count=%u connected_ticks=%u",
+        disconnect_reason_name(reason), lte_diag.disconnect_count,
+        lte_diag.connect_tick == 0 ? 0 : now - lte_diag.connect_tick);
+  LOG_I("[DISC] uart2_rx=%u uart2_tx=%u events=%u write_fail=%u errors=%u/%u last_err=0x%X",
+        lte_diag.uart2_rx_bytes, lte_diag.uart2_tx_bytes,
+        lte_diag.uart2_rx_events, lte_diag.uart2_write_failures,
+        lte_diag.uart2_error_count, lte_diag.uart3_error_count,
+        lte_diag.last_uart_error_code);
+  LOG_I("[DISC] fmu_rx=%u fmu_tx=%u server_rx=%u read_events=%u silence_timeout=%u",
+        lte_diag.fmu_rx_bytes, lte_diag.fmu_tx_bytes,
+        lte_diag.server_rx_bytes, lte_diag.server_rx_read_events,
+        lte_diag.server_silence_timeout_count);
+  LOG_I("[DISC] age_uart2_rx=%u age_uart2_tx=%u age_fmu_rx=%u age_server_rx=%u",
+        diag_age(now, lte_diag.last_uart2_rx_tick),
+        diag_age(now, lte_diag.last_uart2_tx_tick),
+        diag_age(now, lte_diag.last_fmu_rx_tick),
+        diag_age(now, lte_diag.last_server_rx_tick));
+  LOG_I("[DISC] at_cmd=%s result=%d elapsed_ticks=%u timeouts=%u errors=%u",
+        lte_diag.last_at_cmd, lte_diag.last_at_result,
+        lte_diag.last_at_elapsed, lte_diag.at_timeouts, lte_diag.at_errors);
+  LOG_I("[DISC] attempts=%u reconnects=%u qioopen_fail=%u resets=%u",
+        lte_diag.connect_attempts, lte_diag.reconnect_count,
+        lte_diag.qioopen_failures, lte_diag.reset_count);
+  LOG_I("[DISC] csq=%d cereg=%u pdp=%u heap=%u/%u min_free=%u",
+        lte_dev.csq, lte_dev.eps_sta, lte_dev.reset_flag == DTU_CF_NO_NEED_RESET,
+        heap_used, heap_total, heap_total - heap_max);
+
+  lte_dev.client1->connect_sta = false;
+}
+
 
 /*将EC800K软件复位*/
 int16_t EC800K_Reset(void) {
   int16_t rst = -1;
 
+  lte_diag_module_reset();
+
+  BSP_LTE_SetDataTxPaused(true);
   lte_dev.config_sta = false;
   lte_dev.reset_flag = DTU_CF_NEED_RESET;
   lte_dev.client1->connect_sta = false;
@@ -207,15 +267,17 @@ int16_t client_tcp_connect(Cclient_t *client) {
   if (client->connect_sta == true)
     return rtn;
 
+  lte_diag_connect_attempt();
+
   if (lte_dev.reset_flag == DTU_CF_NEED_RESET) {
     /*查询PDP场景*/
     if (0 != BSP_LTE_SendCmd(AT_QIACT, "+QIACT: 1,1,1,", DEFAULT_ERROR,
-                             LTE_CMD_WAIT_LLONG)) {
+                             LTE_CMD_WAIT_QIACT)) {
       /*如果查询PDP场景未设置成功，则激活PDP场景模式*/
       if (0 != BSP_LTE_SendCmd(AT_QIACT1, DEFAULT_EXPECT, DEFAULT_ERROR,
-                               LTE_CMD_WAIT_LLONG)) {
+                               LTE_CMD_WAIT_QIACT)) {
         BSP_LTE_SendCmd(AT_QIDEACT, NETCLOSE_SUCCESS, NETCLOSE_FAILURE,
-                        LTE_CMD_WAIT_LLONG);
+                        LTE_CMD_WAIT_QIDEACT);
         // client->error_cnt++;
         return -1;
       }
@@ -223,9 +285,13 @@ int16_t client_tcp_connect(Cclient_t *client) {
 
     //
     BSP_LTE_SendCmd(AT_CFG_TRANS_SIZE, DEFAULT_EXPECT, DEFAULT_ERROR,
-                    LTE_CMD_WAIT_LLONG);
+                    LTE_CMD_WAIT_NORMAL);
     BSP_LTE_SendCmd(AT_CFG_TRANS_WAITTM, DEFAULT_EXPECT, DEFAULT_ERROR,
-                    LTE_CMD_WAIT_LLONG);
+                    LTE_CMD_WAIT_NORMAL);
+    BSP_LTE_SendCmd(AT_CFG_PASSIVE_CLOSED, DEFAULT_EXPECT, DEFAULT_ERROR,
+                    LTE_CMD_WAIT_NORMAL);
+    BSP_LTE_SendCmd(AT_CFG_TCP_KEEPALIVE, DEFAULT_EXPECT, DEFAULT_ERROR,
+                    LTE_CMD_WAIT_NORMAL);
     //
   }
   // 1. 检查网络信号
@@ -236,13 +302,10 @@ int16_t client_tcp_connect(Cclient_t *client) {
   memset(cmd_str, 0, sizeof(cmd_str));
   rt_sprintf(cmd_cmd, AT_QIOPEN_TRANS /*AT_QIOPEN_FORW*/, client->connect_ID,
           client->mode, client->ip, client->port);
-  rt_sprintf(cmd_str, "+QIOPEN: %d,0", client->connect_ID);
-
-  /*打开socket, 这个模块他不返回+QIOPEN:字符串*/
+  /* 透传模式成功返回 CONNECT；一次尝试只发送一次 QIOPEN。 */
   while (0 != BSP_LTE_SendCmd(cmd_cmd, CONNECTE_OK, DEFAULT_ERROR,
-                              LTE_CMD_WAIT_LLONG) &&
-         0 != BSP_LTE_SendCmd(cmd_cmd, cmd_str, DEFAULT_ERROR,
-                              LTE_CMD_WAIT_LLONG)) {
+                              LTE_CMD_WAIT_QIOPEN)) {
+    lte_diag.qioopen_failures++;
     client_tcp_close(client);
 #ifdef debug_mode
     EC800K_Reset();
@@ -251,13 +314,17 @@ int16_t client_tcp_connect(Cclient_t *client) {
 #else
     client->error_cnt++;
     if (client->error_cnt > 5) {
+      lte_diag_disconnect(LTE_DISC_QIOPEN_FAILED);
       EC800K_Reset();
       client->error_cnt = 0;
       return 0;
     }
+    rt_thread_mdelay((client->error_cnt < 5 ? client->error_cnt : 5) * 1000);
 #endif
   }
   client->connect_sta = true;
+  BSP_LTE_SetDataTxPaused(false);
+  lte_diag_connect_success();
   client->error_cnt = 0;
   lte_dev.reset_flag = DTU_CF_NO_NEED_RESET;
 
@@ -278,7 +345,7 @@ int16_t client_tcp_connect(Cclient_t *client) {
   mav_len = mavlink_msg_to_send_buffer(mav_buf, &msg);
 
   for (int i = 0; i < 3; i++) {
-    rt_device_write(lte_dev.com_dev, 0, mav_buf, mav_len);
+    BSP_LTE_SendData(mav_buf, mav_len);
   }
   rt_device_t fmu_com = fmu_com_get();
   if (fmu_com != RT_NULL) {
@@ -289,11 +356,11 @@ int16_t client_tcp_connect(Cclient_t *client) {
   rt_sprintf(buf, "{\"sn\":\"%s\",\"client_id\":\"%s\",\"csq\":\"%d\"}\r\n",
              lte_dev.iccid, lte_dev.imei, lte_dev.csq);
 
-  rt_device_write(lte_dev.com_dev, 0, buf, sizeof(buf));
+  BSP_LTE_SendData((uint8_t *)buf, sizeof(buf));
 
-  rt_device_write(lte_dev.com_dev, 0, buf, sizeof(buf));
+  BSP_LTE_SendData((uint8_t *)buf, sizeof(buf));
 
-  rt_device_write(lte_dev.com_dev, 0, buf, sizeof(buf));
+  BSP_LTE_SendData((uint8_t *)buf, sizeof(buf));
 
  //hw_printf(buf);
 #endif
@@ -301,17 +368,58 @@ int16_t client_tcp_connect(Cclient_t *client) {
   return rtn;
 }
 
+#if LTE_SERVER_SILENCE_TIMEOUT_SEC > 0
 static rt_tick_t last_rx_tick = 0;
-int16_t client_yield(CLte_dev_t *pLte_dev) {
-  int16_t rtn = 0;
+#endif
+
+/* Leave transparent mode only after the UART has been quiet for the
+ * required guard time.  The caller keeps data_tx_paused asserted until the
+ * next successful QIOPEN. */
+static bool client_exit_transparent_mode(CLte_dev_t *pLte_dev) {
+  rt_tick_t deadline;
+  rt_ssize_t len;
 
+  BSP_LTE_SetDataTxPaused(true);
+  rt_thread_mdelay(1000);
+  BSP_LTE_SendData((uint8_t *)"+++", 3);
+  rt_thread_mdelay(1000);
+
+  deadline = rt_tick_get() + 2000;
+  while ((rt_int32_t)(rt_tick_get() - deadline) < 0) {
+    if (pLte_dev->com_dev_rx_sem != RT_NULL) {
+      rt_sem_take(pLte_dev->com_dev_rx_sem, 100);
+    }
+    rt_memset(pLte_dev->at_buf, 0, sizeof(pLte_dev->at_buf));
+    len = rt_device_read(pLte_dev->com_dev, 0, pLte_dev->at_buf,
+                         sizeof(pLte_dev->at_buf) - 1);
+    if (len > 0) {
+      lte_diag_uart2_rx((rt_size_t)len);
+      pLte_dev->at_buf[len] = '\0';
+      if (rt_strstr((const char *)pLte_dev->at_buf, DEFAULT_EXPECT) != RT_NULL) {
+        return true;
+      }
+    }
+  }
+  return false;
+}
+
+void client_yield(CLte_dev_t *pLte_dev) {
   while (1) {
     if (pLte_dev->client1->connect_sta == true) {
       rt_err_t ret = rt_sem_take(pLte_dev->com_dev_rx_sem, 1000);
-#ifdef EDU_TRAINING     
-      if (rt_tick_get() - last_rx_tick > RT_TICK_PER_SECOND * 6) {
-        LOG_I("timeout!!! reconnect.\n");
-        pLte_dev->client1->connect_sta = false;
+#if LTE_SERVER_SILENCE_TIMEOUT_SEC > 0
+      if (ret != RT_EOK && rt_tick_get() - last_rx_tick >
+          RT_TICK_PER_SECOND * LTE_SERVER_SILENCE_TIMEOUT_SEC) {
+        lte_diag.server_silence_timeout_count++;
+        LOG_W("server silence timeout, reconnect\n");
+        if (client_exit_transparent_mode(pLte_dev)) {
+          pLte_dev->client1->connect_sta = false;
+          client_tcp_close(pLte_dev->client1);
+        } else {
+          lte_diag_disconnect(LTE_DISC_AT_TIMEOUT);
+          EC800K_Reset();
+        }
+        continue;
       }
 #endif
       if (RT_EOK == ret) {
@@ -321,18 +429,18 @@ int16_t client_yield(CLte_dev_t *pLte_dev) {
           len = rt_device_read(pLte_dev->com_dev, 0, pLte_dev->at_buf,
                                sizeof(pLte_dev->at_buf));
           if (len > 0) {
+            lte_diag_uart2_rx((rt_size_t)len);
             /*收到TCP关闭的消息*/
             if (strstr((const char *)pLte_dev->at_buf, CLOSED)) {
-              memset(cmd_str, 0, sizeof(cmd_str));
-              rt_sprintf(cmd_str, "\"closed\",%d", pLte_dev->client1->connect_ID);
-              if (NULL != strstr((char *)lte_dev.at_buf, cmd_str)) {
-                EC800K_Reset();
-                pLte_dev->client1->connect_sta = false;
-                break;
-              }
+              LOG_W("[URC] closed received");
+              lte_diag_disconnect(LTE_DISC_CLOSED);
+              BSP_LTE_SetDataTxPaused(true);
+              client_tcp_close(pLte_dev->client1);
               break;
             } else if (strstr((const char *)lte_dev.at_buf, "NO CARRIER")) {
-              pLte_dev->client1->connect_sta = false;
+              lte_diag_disconnect(LTE_DISC_NO_CARRIER);
+              BSP_LTE_SetDataTxPaused(true);
+              client_tcp_close(pLte_dev->client1);
               _no_carrier_t.match_flag = NO_MATCH;
               break;
             } else if(strstr((const char *)lte_dev.at_buf, "\r\n")) {
@@ -342,21 +450,31 @@ int16_t client_yield(CLte_dev_t *pLte_dev) {
             if (_no_carrier_t.match_flag == HALF_MATCH) {
               no_carrier_match(pLte_dev->at_buf, len);
               if (_no_carrier_t.match_flag == COMPLETE_MATCH) {
-                LOG_I("disconnect!!!\n");
-                pLte_dev->client1->connect_sta = false;
+                lte_diag_disconnect(LTE_DISC_NO_CARRIER_SPLIT);
+                BSP_LTE_SetDataTxPaused(true);
+                client_tcp_close(pLte_dev->client1);
                 _no_carrier_t.match_flag = NO_MATCH;
                 break;
               }
             }
 
             rt_device_t fmu_com = fmu_com_get();
+            lte_diag.server_rx_bytes += len;
+            lte_diag.server_rx_read_events++;
+            lte_diag.last_server_rx_tick = rt_tick_get();
             if (fmu_com) {
-              rt_device_write(fmu_com, 0, pLte_dev->at_buf, len);
+              rt_ssize_t written = rt_device_write(fmu_com, 0,
+                                                   pLte_dev->at_buf, len);
+              if (written > 0) {
+                lte_diag_fmu_tx((rt_size_t)written);
+              }
             }
           }else{
             break;
           }
+#if LTE_SERVER_SILENCE_TIMEOUT_SEC > 0
           last_rx_tick = rt_tick_get();
+#endif
         }
       }
 
@@ -365,11 +483,12 @@ int16_t client_yield(CLte_dev_t *pLte_dev) {
         BSP_LTE_Config();
       }
       client_tcp_connect(pLte_dev->client1);
+#if LTE_SERVER_SILENCE_TIMEOUT_SEC > 0
       last_rx_tick = rt_tick_get();
+#endif
     }
   }
 
-  return rtn;
 }
 
 void client_task(void) {

applications/VK_Client.h

@@ -17,7 +17,7 @@ typedef struct {
 
 void client_task(void);
 
-int16_t client_yield( CLte_dev_t *lte_dev );
+void client_yield( CLte_dev_t *lte_dev );
 
 #ifdef __cplusplus
 }

applications/VK_Com.c

@@ -4,9 +4,53 @@
 #include "rtdef.h"
 #include "rtthread.h"
 
+#define ENABLE_MEMORY_USAGE_LOG 0
+
+#if ENABLE_MEMORY_USAGE_LOG
+#define DBG_TAG "MEM_LOG"
+#define DBG_LVL DBG_INFO
+#include <rtdbg.h>
+#endif
+
 rt_device_t fmu_com = RT_NULL;
 rt_sem_t fmu_com_rx_sem = RT_NULL;
 
+#if ENABLE_MEMORY_USAGE_LOG
+static rt_size_t thread_stack_used(rt_thread_t thread) {
+  rt_uint8_t *stack;
+  rt_size_t unused = 0;
+
+  if (thread == RT_NULL || thread->stack_addr == RT_NULL) {
+    return 0;
+  }
+
+  stack = (rt_uint8_t *)thread->stack_addr;
+  while (unused < thread->stack_size && stack[unused] == '#') {
+    unused++;
+  }
+
+  return thread->stack_size - unused;
+}
+
+static void log_memory_usage(void) {
+  rt_size_t total = 0;
+  rt_size_t used = 0;
+  rt_size_t max_used = 0;
+  rt_thread_t lte_thread;
+
+  rt_memory_info(&total, &used, &max_used);
+  LOG_I("heap used=%u total=%u free=%u max=%u", used, total, total - used,
+        max_used);
+
+  lte_thread = rt_thread_find("lte");
+  if (lte_thread != RT_NULL) {
+    rt_size_t stack_used = thread_stack_used(lte_thread);
+    LOG_I("lte stack used=%u total=%u free=%u", stack_used,
+          lte_thread->stack_size, lte_thread->stack_size - stack_used);
+  }
+}
+#endif
+
 static rt_err_t fmu_com_rx_cb(rt_device_t dev, rt_size_t size) {
   if (fmu_com_rx_sem != RT_NULL) {
     rt_sem_release(fmu_com_rx_sem);
@@ -33,22 +77,32 @@ void com_task(void) {
   static uint8_t buff[128];
 
   uint32_t len = 0;
+#if ENABLE_MEMORY_USAGE_LOG
+  rt_tick_t memory_log_start = rt_tick_get();
+#endif
 
   while (1) {
     if (rt_sem_take(fmu_com_rx_sem, 1000) == RT_EOK) {
       while (1) {
         len = rt_device_read(fmu_com, 0, buff, sizeof(buff));
         if (len > 0) {
-          if (lte_dev.client1->connect_sta == true) {
+          lte_diag_fmu_rx(len);
+          if (lte_dev.client1->connect_sta == true &&
+              lte_dev.data_tx_paused == false) {
             /* 仅在客户端2连接时转发 */
-            rt_device_write(lte_dev.com_dev, 0, buff, len);
+            BSP_LTE_SendData(buff, len);
           }
         } else {
           break;
         }
       }
     }
+
+#if ENABLE_MEMORY_USAGE_LOG
+    if (rt_tick_get() - memory_log_start >= RT_TICK_PER_SECOND * 5) {
+      memory_log_start = rt_tick_get();
+      log_memory_usage();
+    }
+#endif
   }
 }
-
-

applications/drv_ec800k.c

@@ -46,11 +46,104 @@ CLte_dev_t lte_dev = {
     .config_sta = false,
     .reset_flag = DTU_CF_NEED_RESET,
     .dev_init_flag = true,
+    .data_tx_paused = false,
 };
 
+lte_diag_t lte_diag = {0};
+
+void lte_diag_at_begin(const char *cmd) {
+  rt_size_t i;
+
+  lte_diag.at_commands++;
+  lte_diag.last_at_elapsed = rt_tick_get();
+  lte_diag.last_at_result = -2;
+  for (i = 0; i < sizeof(lte_diag.last_at_cmd) - 1 && cmd[i] != '\0'; i++) {
+    lte_diag.last_at_cmd[i] = (cmd[i] == '\r' || cmd[i] == '\n') ? '_' : cmd[i];
+  }
+  lte_diag.last_at_cmd[i] = '\0';
+}
+
+void lte_diag_at_end(rt_int32_t result) {
+  rt_tick_t now = rt_tick_get();
+  lte_diag.last_at_elapsed = now - lte_diag.last_at_elapsed;
+  lte_diag.last_at_result = result;
+  if (result == -1) {
+    lte_diag.at_timeouts++;
+  } else if (result != 0) {
+    lte_diag.at_errors++;
+  }
+}
+
+void lte_diag_uart2_rx(rt_size_t size) {
+  lte_diag.uart2_rx_bytes += size;
+  lte_diag.last_uart2_rx_tick = rt_tick_get();
+}
+
+void lte_diag_uart2_tx(rt_size_t size) {
+  lte_diag.uart2_tx_bytes += size;
+  lte_diag.last_uart2_tx_tick = rt_tick_get();
+}
+
+void lte_diag_uart_error(const char *name, rt_uint32_t error_code) {
+  lte_diag.last_uart_error_code = error_code;
+  if (name != RT_NULL && rt_strcmp(name, "uart2") == 0) {
+    lte_diag.uart2_error_count++;
+  } else if (name != RT_NULL && rt_strcmp(name, "uart3") == 0) {
+    lte_diag.uart3_error_count++;
+  }
+}
+
+void lte_diag_fmu_rx(rt_size_t size) {
+  lte_diag.fmu_rx_bytes += size;
+  lte_diag.last_fmu_rx_tick = rt_tick_get();
+}
+
+void lte_diag_fmu_tx(rt_size_t size) {
+  lte_diag.fmu_tx_bytes += size;
+  lte_diag.last_fmu_tx_tick = rt_tick_get();
+}
+
+void lte_diag_connect_attempt(void) {
+  lte_diag.connect_attempts++;
+  if (lte_diag.connect_attempts > 1) {
+    lte_diag.reconnect_count++;
+  }
+}
+
+void lte_diag_connect_success(void) {
+  lte_diag.connect_tick = rt_tick_get();
+  lte_diag.last_server_rx_tick = lte_diag.connect_tick;
+}
+
+void lte_diag_module_reset(void) {
+  lte_diag.reset_count++;
+}
+
 void BSP_LTE_SendData(uint8_t *pdata, uint32_t length) {
   if (lte_dev.com_dev != RT_NULL) {
-    rt_device_write(lte_dev.com_dev, 0, pdata, length);
+    if (lte_dev.data_tx_mutex != RT_NULL) {
+      rt_mutex_take(lte_dev.data_tx_mutex, RT_WAITING_FOREVER);
+    }
+    rt_ssize_t written = rt_device_write(lte_dev.com_dev, 0, pdata, length);
+    if (written > 0) {
+      lte_diag_uart2_tx((rt_size_t)written);
+    }
+    if (written != (rt_ssize_t)length) {
+      lte_diag.uart2_write_failures++;
+    }
+    if (lte_dev.data_tx_mutex != RT_NULL) {
+      rt_mutex_release(lte_dev.data_tx_mutex);
+    }
+  }
+}
+
+void BSP_LTE_SetDataTxPaused(bool paused) {
+  if (lte_dev.data_tx_mutex != RT_NULL) {
+    rt_mutex_take(lte_dev.data_tx_mutex, RT_WAITING_FOREVER);
+  }
+  lte_dev.data_tx_paused = paused;
+  if (lte_dev.data_tx_mutex != RT_NULL) {
+    rt_mutex_release(lte_dev.data_tx_mutex);
   }
 }
 
@@ -63,7 +156,11 @@ int16_t BSP_LTE_CipSend(uint8_t *pdata, uint32_t length, uint32_t waittime) {
 
     rt_memset(lte_dev.at_buf, 0, sizeof(lte_dev.at_buf));
 
-    rt_device_read(lte_dev.com_dev, 0, lte_dev.at_buf, sizeof(lte_dev.at_buf));
+    rt_ssize_t read_len = rt_device_read(lte_dev.com_dev, 0, lte_dev.at_buf,
+                                         sizeof(lte_dev.at_buf));
+    if (read_len > 0) {
+      lte_diag_uart2_rx((rt_size_t)read_len);
+    }
 
     /* 清空缓存 */
     uint8_t c;
@@ -81,6 +178,7 @@ int16_t BSP_LTE_CipSend(uint8_t *pdata, uint32_t length, uint32_t waittime) {
 
 int16_t BSP_LTE_SendCmd(const char *cmd, const char *expect, const char *error,
                         uint32_t waittime) {
+  lte_diag_at_begin(cmd);
   LOG_I("%s\n", cmd);
   BSP_LTE_SendData((uint8_t *)cmd, rt_strlen(cmd));
 
@@ -90,7 +188,11 @@ int16_t BSP_LTE_SendCmd(const char *cmd, const char *expect, const char *error,
 
     rt_memset(lte_dev.at_buf, 0, sizeof(lte_dev.at_buf));
 
-    rt_device_read(lte_dev.com_dev, 0, lte_dev.at_buf, sizeof(lte_dev.at_buf));
+    rt_ssize_t read_len = rt_device_read(lte_dev.com_dev, 0, lte_dev.at_buf,
+                                         sizeof(lte_dev.at_buf));
+    if (read_len > 0) {
+      lte_diag_uart2_rx((rt_size_t)read_len);
+    }
 
     LOG_I("%s\n", lte_dev.at_buf);
     /* 清空缓存 */
@@ -99,14 +201,17 @@ int16_t BSP_LTE_SendCmd(const char *cmd, const char *expect, const char *error,
     }
 
     if (expect != NULL && NULL != rt_strstr((char *)lte_dev.at_buf, expect)) {
+      lte_diag_at_end(0);
       return 0;
-    } else if (NULL != rt_strstr((char *)lte_dev.at_buf, error) &&
-               error != NULL) {
+    } else if (error != NULL && NULL != rt_strstr((char *)lte_dev.at_buf, error)) {
+      lte_diag_at_end(1);
       return 1;
     } else {
+      lte_diag_at_end(2);
       return 2;
     }
   } else {
+    lte_diag_at_end(-1);
     return -1;
   }
 }
@@ -307,6 +412,7 @@ int16_t BSP_LTE_GetCEREG(int16_t retry) {
 }
 
 static rt_err_t lte_com_rx_cb(rt_device_t dev, rt_size_t size) {
+  lte_diag.uart2_rx_events++;
   if (lte_dev.com_dev_rx_sem != RT_NULL) {
     rt_sem_release(lte_dev.com_dev_rx_sem);
   }
@@ -319,6 +425,10 @@ void BSP_LET_Register_Com(rt_device_t com_dev) {
   }
 
   lte_dev.com_dev = com_dev;
+  if (lte_dev.data_tx_mutex == RT_NULL) {
+    lte_dev.data_tx_mutex = rt_mutex_create("lteTx", RT_IPC_FLAG_PRIO);
+  }
+  BSP_LTE_SetDataTxPaused(false);
 }
 
 /*

applications/drv_ec800k.h

@@ -197,6 +197,17 @@ extern "C" {
 //AT CFG
 #define AT_CFG_TRANS_SIZE "AT+QICFG=\"transpktsize\",512\r\n"
 #define AT_CFG_TRANS_WAITTM "AT+QICFG=\"transwaittm\",0\r\n"
+#define AT_CFG_PASSIVE_CLOSED "AT+QICFG=\"passiveclosed\",1\r\n"
+#define AT_CFG_TCP_KEEPALIVE "AT+QICFG=\"tcp/keepalive\",1,30,25,3\r\n"
+
+#define LTE_CMD_WAIT_QIOPEN (150000)
+#define LTE_CMD_WAIT_QICLOSE (10000)
+#define LTE_CMD_WAIT_QIACT (150000)
+#define LTE_CMD_WAIT_QIDEACT (40000)
+
+/* Set after confirming the server MAVLink heartbeat period. 0 disables the
+ * application-level silence recovery and leaves NO CARRIER/keepalive active. */
+#define LTE_SERVER_SILENCE_TIMEOUT_SEC (15)
 
 // #define GSM_BT_LIVE		5*60*1000			// 5min
 typedef struct _connect_state {
@@ -223,6 +234,63 @@ typedef struct {
   unsigned char error_cnt;
 } Cclient_t;
 
+typedef enum {
+  LTE_DISC_CLOSED = 0,
+  LTE_DISC_NO_CARRIER,
+  LTE_DISC_NO_CARRIER_SPLIT,
+  LTE_DISC_AT_TIMEOUT,
+  LTE_DISC_QIOPEN_FAILED,
+  LTE_DISC_MODULE_RESET,
+} lte_disconnect_reason_t;
+
+typedef struct {
+  rt_uint32_t uart2_rx_bytes;
+  rt_uint32_t uart2_tx_bytes;
+  rt_uint32_t uart2_rx_events;
+  rt_uint32_t uart2_write_failures;
+  rt_uint32_t uart2_error_count;
+  rt_uint32_t uart3_error_count;
+  rt_uint32_t last_uart_error_code;
+  rt_uint32_t fmu_rx_bytes;
+  rt_uint32_t fmu_tx_bytes;
+  rt_uint32_t at_commands;
+  rt_uint32_t at_timeouts;
+  rt_uint32_t at_errors;
+  rt_uint32_t connect_attempts;
+  rt_uint32_t reconnect_count;
+  rt_uint32_t qioopen_failures;
+  rt_uint32_t reset_count;
+  rt_uint32_t disconnect_count;
+  rt_uint32_t server_rx_bytes;
+  rt_uint32_t server_rx_read_events;
+  rt_uint32_t server_silence_timeout_count;
+  rt_tick_t last_uart2_rx_tick;
+  rt_tick_t last_uart2_tx_tick;
+  rt_tick_t last_fmu_rx_tick;
+  rt_tick_t last_fmu_tx_tick;
+  rt_tick_t last_server_rx_tick;
+  rt_tick_t connect_tick;
+  rt_tick_t last_disconnect_tick;
+  rt_tick_t last_at_elapsed;
+  rt_int32_t last_at_result;
+  char last_at_cmd[20];
+  lte_disconnect_reason_t last_disconnect_reason;
+} lte_diag_t;
+
+extern lte_diag_t lte_diag;
+
+void lte_diag_at_begin(const char *cmd);
+void lte_diag_at_end(rt_int32_t result);
+void lte_diag_uart2_rx(rt_size_t size);
+void lte_diag_uart2_tx(rt_size_t size);
+void lte_diag_uart_error(const char *name, rt_uint32_t error_code);
+void lte_diag_fmu_rx(rt_size_t size);
+void lte_diag_fmu_tx(rt_size_t size);
+void lte_diag_connect_attempt(void);
+void lte_diag_connect_success(void);
+void lte_diag_disconnect(lte_disconnect_reason_t reason);
+void lte_diag_module_reset(void);
+
 typedef struct {
   char hard_ver[32];
   char imei[32];
@@ -241,6 +309,7 @@ typedef struct {
 
   rt_device_t com_dev;
   rt_sem_t com_dev_rx_sem;
+  rt_mutex_t data_tx_mutex;
   rt_uint8_t at_buf[256];
 
   rt_thread_t tid1;
@@ -248,6 +317,7 @@ typedef struct {
 
   bool reset_flag;
   bool dev_init_flag;
+  volatile bool data_tx_paused;
 } CLte_dev_t;
 extern CLte_dev_t lte_dev;
 
@@ -282,6 +352,7 @@ void BSP_LTE_PwrOff(void);
 void BSP_LTE_ResetByKey(void);
 
 void BSP_LTE_SendData(uint8_t *pdata, uint32_t length);
+void BSP_LTE_SetDataTxPaused(bool paused);
 
 int16_t BSP_LTE_SendDataAT(uint8_t *pdata, uint32_t length);
 

applications/main.c

@@ -56,7 +56,7 @@ int main(void) {
   
   /* 创建任务2和任务3，分别从U2和U3收发数据 */
   lte_dev.tid1 =
-      rt_thread_create("lte", task_let_entry, RT_NULL, 2048, 10, 5);
+      rt_thread_create("lte", task_let_entry, RT_NULL, 1024, 10, 5);
   if (lte_dev.tid1) {
     rt_thread_startup(lte_dev.tid1);
   }

docs/EC800K_TCP透传模式优化方案.md

@@ -0,0 +1,250 @@
+# EC800K TCP 透传模式优化方案
+
+## 1. 适用范围
+
+当前工程使用 Quectel EC800K，通过 UART2 连接模块，通过 `AT+QIOPEN` 的透传模式承载 TCP 数据：
+
+```text
+AT+QIOPEN=1,0,"TCP",<server>,<port>,0,2
+```
+
+飞控数据保持原有透传路径，不增加新的 TCP 连接，也不要求第一阶段解析完整 MAVLink 数据。
+
+## 2. 目标
+
+- 保持飞控数据的透明传输。
+- 使用服务器现有的 MAVLink 心跳检测下行链路。
+- 使用 EC800K 的 TCP keepalive 处理网络黑洞。
+- 正确处理 `NO CARRIER` 和分包 URC。
+- TCP 断开时优先关闭 Socket 并重连，只有 AT 状态机无响应时才复位模块。
+- 避免重复发送 `QIOPEN` 和连续快速重连。
+
+## 3. 连接初始化
+
+在 `QIOPEN` 前完成 PDP 激活和透传参数配置：
+
+```text
+AT+QIACT?
+AT+QIACT=1                         // PDP 未激活时执行
+AT+QICFG="transpktsize",512
+AT+QICFG="transwaittm",0
+AT+QICFG="passiveclosed",1
+AT+QICFG="tcp/keepalive",1,30,25,3
+```
+
+实际配置前先执行：
+
+```text
+AT+QICFG=?
+```
+
+确认当前 EC800K 固件支持的参数范围。`tcp/keepalive` 中的 `interval_time` 必须遵守当前固件文档范围，不能直接使用其他模块的参数。
+
+配置后查询并记录：
+
+```text
+AT+QICFG="passiveclosed"
+AT+QICFG="tcp/keepalive"
+AT+QICFG="transpktsize"
+AT+QICFG="transwaittm"
+```
+
+## 4. 透传状态机
+
+建议将连接状态划分为：
+
+```text
+DISCONNECTED
+CONNECTING
+TRANSPARENT_CONNECTED
+SUSPECTED
+EXITING_TRANSPARENT
+CLOSING
+RECONNECT_WAIT
+```
+
+正常连接流程：
+
+```text
+DISCONNECTED
+    -> QIACT
+    -> CONNECTING
+    -> QIOPEN
+    -> 收到 CONNECT
+    -> TRANSPARENT_CONNECTED
+```
+
+一次连接尝试只能发送一次 `QIOPEN`。等待 `CONNECT` 或 `+QIOPEN:<connectID>,<err>` 后再决定成功或失败，不能在一个循环条件中重复发送同一条 `QIOPEN`。
+
+重连间隔建议退避：
+
+```text
+1 秒 -> 2 秒 -> 5 秒 -> 10 秒 -> 30 秒
+```
+
+## 5. 服务器心跳检测
+
+服务器连接建立后会定期发送 MAVLink 心跳包。第一阶段先不解析 MAVLink，只要透传接收路径收到服务器数据，就更新：
+
+```text
+last_server_rx_tick
+server_rx_bytes
+server_rx_read_events
+```
+
+设备超时应设置为服务器心跳周期的 2~3 倍：
+
+| 服务器心跳周期 | 建议设备超时 |
+|---|---:|
+| 3 秒 | 10~12 秒 |
+| 5 秒 | 15 秒左右 |
+| 10 秒 | 25~30 秒 |
+
+当前已确认服务器每 3 秒发送一次 MAVLink 心跳，工程宏 `LTE_SERVER_SILENCE_TIMEOUT_SEC` 设置为 `15`，允许连续丢失多个心跳后再进入恢复流程。若服务器心跳周期发生变化，应同步调整该值。一般应设置为心跳周期的 2~3 倍，并保留网络抖动余量。
+
+不要只允许丢失一个心跳包就重连。超时后先标记 `SUSPECTED`，停止普通发送，再执行恢复流程。
+
+第一阶段的“任意下行数据”只用于快速验证链路，不能证明收到的必然是 MAVLink HEARTBEAT，也不能证明服务器一定收到了设备上行数据。读取次数也不等于 MAVLink 包数量，因此应记录为 `server_rx_read_events`。后续如果需要严格判断心跳，再使用现有 MAVLink 库只识别 `MAVLINK_MSG_ID_HEARTBEAT`，不解析其他消息。服务器如果没有固定下行心跳，不能使用该方法作为连接判断。
+
+## 6. 断连事件处理
+
+### 6.1 `NO CARRIER`
+
+在透传模式下，EC800K 因网络错误或 Socket 断开通常会上报：
+
+```text
+NO CARRIER
+```
+
+模块会退出透传模式。程序应支持完整和分包匹配：
+
+```text
+记录断连原因
+停止 UART3 -> UART2 的发送
+connect_sta = false
+执行 AT+QICLOSE=0
+进入 RECONNECT_WAIT
+```
+
+收到 `NO CARRIER` 后通常不需要再次发送 `+++`，因为模块已经退出透传模式。
+
+### 6.2 服务器心跳超时
+
+如果超过配置的心跳超时时间没有收到服务器数据：
+
+```text
+TRANSPARENT_CONNECTED
+    -> SUSPECTED
+    -> 暂停飞控数据发送
+    -> 严格退出透传
+    -> QICLOSE
+    -> QIOPEN 重连
+```
+
+### 6.3 `+++` 退出透传
+
+只有在程序仍处于透传模式、但需要主动恢复时才发送 `+++`。必须满足：
+
+```text
+发送前至少 1 秒没有其他 UART 数据
++++ 输入期间没有其他数据
+发送后至少 1 秒没有其他数据
+等待模块返回 OK
+```
+
+执行期间必须暂停飞控发送线程，否则 `+++` 可能被当作普通 TCP 数据发送。
+
+暂停不能只依赖 `connect_sta=false`。所有 UART2 写路径必须检查统一的发送闸门，必要时使用互斥锁，确保停止当前写操作后再等待 1 秒静默。`+++`、退出后的 AT 命令和恢复透传必须与飞控发送互斥。
+
+## 7. Socket 恢复和模块复位
+
+推荐恢复顺序：
+
+```text
+NO CARRIER 或心跳超时
+    -> 停止业务发送
+    -> 必要时 +++ 并等待 OK
+    -> AT+QICLOSE=0
+    -> 查询/恢复 PDP
+    -> 单次 QIOPEN
+    -> 收到 CONNECT 后恢复透传
+```
+
+以下情况才执行 EC800K 复位：
+
+- `+++` 超时；
+- `AT+QICLOSE` 超时；
+- `AT+QIACT` 长时间无响应；
+- 连续多个 AT 命令无响应；
+- UART2 长时间没有任何模块响应。
+
+TCP 断开本身不等于 EC800K 死机，不能每次 `NO CARRIER` 都直接复位模块。
+
+AT 等待时间必须按官方上限设计：`QIACT` 最长可达 150 秒，`QIOPEN` 建议最长等待 150 秒，`QICLOSE` 默认最长约 10 秒，`QIDEACT` 最长可达 40 秒。当前工程的 7 秒和 500 毫秒等待不能直接用于这些命令，否则会把模块仍在处理的请求误判为超时。一次 `QIOPEN` 超时后必须先完成清理和退避，不能立即重复发送。
+
+## 8. TCP keepalive 与业务心跳的关系
+
+```text
+AT+QICFG="tcp/keepalive",1,30,25,3
+```
+
+用于 TCP 协议栈探测对端是否仍可达。它不能确认服务器业务程序是否处理了飞控心跳，也不能替代服务器的 MAVLink 心跳。
+
+`AT+QICFG="passiveclosed",1` 用于处理服务器 FIN/RST 已经到达模块的被动关闭。网络黑洞导致 FIN 没有到达时，仍需依靠 TCP keepalive 或服务器心跳超时。
+
+## 9. 日志内容
+
+建议在断连日志中记录：
+
+```text
+断连原因
+server_rx_bytes
+server_rx_read_events
+age_server_rx
+心跳超时次数
+当前连接状态
+重连退避时间
+UART2 收发字节和错误计数
+CSQ/CEREG/PDP 状态
+AT 最后一条命令、结果和耗时
+```
+
+这些信息可以初步区分：
+
+- 模块明确报告 TCP 断开；
+- 服务器心跳停止；
+- 网络黑洞导致模块尚未收到 FIN；
+- UART 丢数据或溢出；
+- PDP/注册状态异常；
+- AT 状态机无响应。
+
+## 10. 内存预算
+
+第一阶段不解析 MAVLink，只增加计时器、状态和计数器：
+
+```text
+纯状态字段：约 32~64 字节
+加 UART2 发送互斥对象：通常约 100~300 字节，取决于 RT-Thread 配置
+Flash：约 1~3 KB
+```
+
+不新增线程，不新增大块动态内存。
+
+如果后续只增加一个 MAVLink HEARTBEAT 识别器，而不是解析所有消息，预计额外增加：
+
+```text
+RAM：约 50~100 字节
+Flash：约 1~2 KB
+```
+
+## 11. 实施顺序
+
+1. 确认服务器实际 MAVLink 心跳周期。
+2. 增加 `passiveclosed` 和 `tcp/keepalive` 配置及查询日志。
+3. 增加服务器下行数据计时器和超时状态。
+4. 修正 `NO CARRIER` 分包解析。
+5. 修正 `QIOPEN` 重复发送问题。
+6. 实现 `QICLOSE` 优先、复位兜底的恢复流程。
+7. 实现严格的 `+++` 退出保护。
+8. 增加重连退避和断连诊断日志。
+9. 根据现场日志决定是否增加轻量级 MAVLink HEARTBEAT 识别。

libraries/HAL_Drivers/drivers/drv_usart_v2.c

@@ -11,6 +11,8 @@
 #include "board.h"
 #include "drv_usart_v2.h"
 
+extern void lte_diag_uart_error(const char *name, rt_uint32_t error_code);
+
 #ifdef RT_USING_SERIAL_V2
 
 //#define DRV_DEBUG
@@ -1237,6 +1239,7 @@ void HAL_UART_ErrorCallback(UART_HandleTypeDef *huart)
     RT_ASSERT(huart != NULL);
     struct stm32_uart *uart = (struct stm32_uart *)huart;
     LOG_D("%s: %s %d\n", __FUNCTION__, uart->config->name, huart->ErrorCode);
+    lte_diag_uart_error(uart->config->name, huart->ErrorCode);
     UNUSED(uart);
 }