JPH0340048A - Memory error processing system - Google Patents

Abstract

PURPOSE:To improve the reliability of a system by calculating the check sum of a program at the time point of actuating this program, and actuating this program after confirming a fact that the check sum coincides with a check sum value stored in a check sum table. CONSTITUTION:Programs P of the same contents are stored in plural parts on a memory 1, and also, a check sum table 2 for calculating and storing a check sum CS at every program at the time of storing each program on the memory 1 is provided. At the time of actuating some program P, the check sum of this program is calculated at its time point, compared with the check sum value of this program stored in the check sum table 2, and after confirming a fact that they coincide with each other, this program is actuated. When they do not coincide with each other, the check sum is calculated and compared in the same way with regard to the same program stored in other part. In such a manner, the reliability of a system can be improved.

Description

[Detailed Description of the Invention] [Summary] The purpose of this invention is to provide a means for checking the validity of a program in a data processor system that does not have a core ECC circuit, and to store multiple programs with the same content in memory. In addition, when each program is stored in memory, the checksum is calculated and stored for each program.
A checksum table is provided, and when a program is started, the checksum of the program is calculated at that time, and the checksum value of the program is compared with the checksum value of the program stored in the checksum table. After confirming that the program is running, start the program.

[Industrial application field]

The present invention relates to reliability design for a type of processor application system in which memory errors cannot be recognized by hardware.

In recent years, space microprocessor application systems such as Zosnam require low power consumption and lightweight systems. For this reason, a system without an ECC circuit is conceivable, but it is necessary to perform a reliability design equivalent to an ECC circuit.

[Conventional technology]

In conventional main memory circuits, E is used to automatically repair 1-bit inversion errors and recognize 2-bit errors.
It incorporates a CC circuit. On the other hand, space microbe a
In fields where low power consumption and weight reduction are desired, such as processor application systems, methods without adding an ECC circuit have been studied in order to reduce power consumption and weight.

However, in the example of a space microprocessor application system, bit inversion occurs due to the influence of cosmic rays, so memory errors such as bit inversion cannot be ignored, and a method of adding an ECC circuit to the main memory circuit is selected. .

[Problem to be solved by the invention]

The present invention aims to improve the reliability of the system by recognizing memory errors by software even if the IC1- is not added with an ECC circuit.

[Means for resolving liability]

In the present invention, as shown in FIG. 1, programs P with the same content are stored in multiple locations on the memory 1, and when each program is stored on the memory, its checksum C8f, A checksum table 2 is provided to calculate and store the checksum table 2, and when a program Pi is started, the checksum of the program is calculated at that time, and the checksum value of the program stored in the checksum table 21C is calculated. After confirming that they match, the program is started, and if they do not match, the checksum of the same program stored elsewhere is calculated and compared in the same way. What he did was a credit to Kim Ji Chong.

[Effect]

By doing the above, it is possible to check the validity of the program to be started before starting it, without applying ECC1 to each data and instruction on the memory and checking it with the ECC circuit. If possible, although it takes a long processing time, it is possible to guarantee sufficient reliability for practical use without an ECC circuit.

〔Example〕

This method uses a checksum to check whether the currently running program has a memory error in the program to which it is about to transfer control, and then transfers control to the program with no memory errors. be.

The present invention will be described using a virtual system with a main controller and a main memory.

The virtual software system has a module structure as shown in FIG. 2, and the same module is stored redundantly due to a retry when an important module 1 block memory error occurs. In the case of this virtual system, three main programs, three subprograms, and three B programs are independently stored in separate areas.

(See FIG. 3) Also, when determining the memory layout, a checksum table 2 is created for each program.

(See FIG. 4) Here, the checksum refers to a pattern resulting from addition or exclusive OR operation of all 1.0 patterns of data (including instructions) in a certain memory area. Alternatively, a specific value (for example, the name of the program) may be added to the result. Therefore, the checksum value (pattern) should be different before and after the error occurs due to cosmic rays.

A checksum table 2 as shown in Figure 4 is provided for each valley program, and each table has an ID that is the program name.
21, a program length (number of words or number of bytes) 22, a start address 23 of the program, and a checksum 24 which is the exclusive OR result of all the words/bytes. The program ID has the same value for the same program, for example, program A1. , A2 A
All rcs have the same ID.

Let's consider a case where subprogram Ai is called from the king program (see the flowchart in Figure 5). Then, search for subprogram k from the checksum table. Most likely, the checksum table for subprogram A1 will be found.

At this time, if the checksum matches the calculated value, the main program may transfer control to the subprogram A1.

If a subprogram AIK memory error occurs, the checksum will not match the calculated value. In this case, the checksum table is searched again. Then, subprogram A2 (i7 will be found. In this way, repeat this process until the calculated value and the checksum match. If the target module is not found on the checksum table, give up and Perform error handling.

In the flowchart in Section 5, the target module is in the form of a subroutine, but in some cases it may not return to the caller.

In the present invention, as described above, memory errors are checked before a module is executed, and if a memory error occurs, the module is not executed.

Therefore, even if a memory error occurs, it can be recognized on the program, making it possible to prevent serious problems such as runaway.

〔Effect of the invention〕

As explained above, according to the present invention, even in a system without an ECC circuit, a system that is resistant to memory errors can be created, the continuous operation time can be extended, and memory errors can be localized. This greatly contributes to the reliability of the system.

[Brief explanation of drawings]

Figure 1 is a diagram of the principle of the present invention, Figure 2 is a diagram of the program module structure of the virtual system, Figure 3 is a memory error in the virtual system, and Figure 4 is a diagram of the specifications on the checksum table of the virtual system. 70 is a chart of an embodiment of the present invention applied to a virtual system. In the figure, 1 is a memory and 2 is a checksum table.

Claims (1)

[Claims] A check that stores programs with the same content in multiple locations on a memory, and calculates and stores a checksum for each program when storing each program on the memory. A sum table is provided, and when a program is started, the checksum of the program is calculated at that time, and the checksum value of the program is compared with the checksum value of the program stored in the checksum table. The program is started after checking the program, and if they do not match, the checksums of the same program stored in other locations are similarly calculated and compared. Processing method.