SDK的RTC驱动对闹钟的处理不完善导致频繁进中断并看门狗复位问题

[oneiotworld]

在SDK的RTC驱动存在一个BUG，公司生产项目有遇到，最近发现新版本SDK还没修复。
主要问题是RTC_DRV_IRQHandler中没有对alarm触发且alarmConfig为NULL的情况做处理，然后会一直触发RTC中断，最终导致MCU看门狗复位(复位也不会恢复，会一直持续复位)。

复现流程：

• 正常上电
• 设置RTC闹钟
• 在RTC闹钟触发前软复位MCU（会导致alarmConfig变为NULL）
• 到RTC闹钟时间时会一直进RTC中断
• 看门狗复位

代码位置：

void RTC_DRV_IRQHandler(uint32_t instance)
{
    DEV_ASSERT(instance < RTC_INSTANCE_COUNT);

    uint32_t tempSeconds;
    /* Get the alarm configuration */
    rtc_alarm_config_t *alarmConfig = g_rtcRuntimeConfig[instance].alarmConfig;
    /* Get the fault interrupt configuration */
    const rtc_overflow_config_t *const overflowConfig = g_rtcRuntimeConfig[instance].overflowConfig;

    /* Check if an alarm has occurred */
    if (RTC_GetTimeAlarmFlag(g_rtcBase[instance]) == true)
    {
        /* If the alarm interrupt configuration has been defined process the
         * alarm IRQ
         */
        if ((alarmConfig != NULL))
        {
            /* If recurrence is enabled modify the alarm register to the next
             * alarm.
             */
            if ((alarmConfig->numberOfRepeats > 0UL) || (alarmConfig->repeatForever == true))
            {
                /* Get current time */
                tempSeconds = RTC_GetTimeSecondsRegister(g_rtcBase[instance]);
                /* Current time is incremented with the repetition value */
                tempSeconds += alarmConfig->repetitionInterval - 1UL;
                /* Set new value to trigger the alarm */
                RTC_SetTimeAlarmRegister(g_rtcBase[instance], tempSeconds);

                g_rtcRuntimeConfig[instance].isAlarmTimeNew = true;
                /* If the alarm repeats forever, set number of repeats to 0
                 * to avoid an accidental trigger of the core overflow flag
                 */
                alarmConfig->numberOfRepeats = (alarmConfig->repeatForever == false) ? (alarmConfig->numberOfRepeats -
                                                                                        1UL) : 0UL;
            } else
            {
                /* If the alarm does not repeat, write 0 to ALM to clear the
                 * alarm flag.
                 */
                RTC_SetTimeAlarmRegister(g_rtcBase[instance], 0UL);
                /* Set the internal variable which indicates that a new alarm is enabled to false */
                g_rtcRuntimeConfig[instance].isAlarmTimeNew = false;
            }
            /* If the user has defined a callback, call it */
            if (alarmConfig->rtcAlarmCallback != NULL)
            {
                alarmConfig->rtcAlarmCallback(alarmConfig->callbackParams);
            }
        }
        // 这里else 分支没有处理
    }
        /* If the IRQ is not caused by the alarm then call the user callback if
         * defined.
         */
    else if (overflowConfig != NULL)
    {
        if (overflowConfig->rtcOverflowCallback != NULL)
        {
            /* Call the RTC interrupt callback function with callback parameter */
            overflowConfig->rtcOverflowCallback(overflowConfig->callbackParams);
        }
    } else
    {
        /* Do nothing*/
    }
}

修复方案：
补充else分支，清除RTC闹钟

diff --git a/platform/drivers/src/rtc/rtc_driver.c b/platform/drivers/src/rtc/rtc_driver.c
index 439a400..512f49a 100644
--- a/platform/drivers/src/rtc/rtc_driver.c
+++ b/platform/drivers/src/rtc/rtc_driver.c
@@ -736,6 +736,14 @@ void RTC_DRV_IRQHandler(uint32_t instance)
             {
                 alarmConfig->rtcAlarmCallback(alarmConfig->callbackParams);
             }
+        } else
+        {
+            /* If the alarm interrupt configuration is NULL (it may happen if the
+             * alarm is not arrived but the chip has been reset by any reason),
+             * So we should write 0 to ALM to clear the alarm flag */
+            RTC_SetTimeAlarmRegister(g_rtcBase[instance], 0UL);
+            /* Set the internal variable which indicates that a new alarm is enabled to false */
+            g_rtcRuntimeConfig[instance].isAlarmTimeNew = false;
         }
     }
         /* If the IRQ is not caused by the alarm then call the user callback if

[qinzhao]

RTC的寄存器在不断电复位是会保持的，也就是说会继续之前的配置工作，demo的使用方法是先deinit，再init

[oneiotworld]

qinzhao 您说的deinit再init是在MCU启动时初始化吧，但是如果中断是在那之前触发(应该有概率出现?)，是否也一样会一直进中断，之前也和FAE讨论过这个问题。所以我们的最终方案就是加了else分支，在alarmConfig 为NULL时直接清除寄存器的alarm配置

[major]

这个问题出现的原因是RTC不同中断共用相同向量号导致函数未注册时候被异常触发，可以通过如下方式进行修复：

改问题在新版本中修复（新版本尚在发布流程中）