1. The engineers modify the J1 byte to check whether the optical cable connection is faulty. It is found that the optical cable connection is correct.
2. The engineers collect K bytes for analysis:
1 5644 SD_CLEARS 0x0002
1 5645 T2_START 0x0001
1 5646 K_SENDS 0x5ba2
1 5647 K_DIR 0x0002
1 5648 K_SENDS 0x5baa
1 5649 K_DIR 0x0000
1 5650 STATE_TRANS 0x2415
1 5651 K_RECEIVED 0x5aba
1 5652 K_DIR 0x0000
1 5653 K_RECEIVED 0xb328
It can be seen that the transmission of K bytes is normal and the T2 timer is started normally. However, the T2 timer cannot time out. As a result, the switching cannot be restored.
3. The engineers perform the following analysis: After the WTR timer (T2_START) is started, because the timeout event cannot be received, the WTR timer maintains in the WTR state and is aborted by the algorithm module abnormally. Because the crossover algorithm module does not set the timeout timer for writing the reserved memory to an invalid value after the timer times out, the algorithm module regards the timer as a valid timer. The platform times out the non-periodic timer and then releases the resource and adds the ID of the timer to the end of the unused timer linked list.
The MS starts a timer every two minutes for active and standby state synchronization. The timer IDs are applied and then released repeatedly. Other tasks also need to apply for timer IDs. If the MS applies for the ID successfully at the moment and service is delivered, the crossover algorithm module stores the ID and the cross algorithm actively stops the timer. Because timer started by the MS has been deleted, the timeout event can no longer be received and the MS enters the WTR state. However, the crossover algorithm still thinks that the MS has a timer and does not start a new timer for the MS. Instead, the crossover algorithm module directly adds the ID to ID list that the MS has applied for. Because the crossover algorithm module cannot receive the timeout event, the MS cannot switch to the idle state from the WTR state. When the T2 is started, this problem will occur again.
It is very difficult to reproduce this problem for the following reason: This problem involves a timer ID that has been used to deliver a service and the ID must be applied for the algorithm module for writing the reserved memory. This ID must be also applied for by the MS module at the same time and a service operation must be delivered to stop the timer ID for writing the reserved memory That is started in the last service delivery.4. Solution: When the non-periodic timer times out, the upper layer software should set the timer ID to an invalid value to prevent incorrect timer stopping. This problem can be solved by upgrading the version to the R