JPH0531774B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION Field of Industrial Application The present invention relates to interrupt processing in a computer system.

Conventional configuration and its problems Interrupts are a very frequently used and effective function in computer systems, and when some event occurs, the currently executing process is interrupted and a more urgent processing program is started. Used when priority needs to be given to execution.

Since interrupts are a function shared by all processes executed on a computer system, they are handled more uniformly by the operating system. When an interrupt occurs, the interrupt handling program in the operating system is started, saves the process context such as registers and instruction counters used by the interrupted process to a save area such as the interrupt stack, and then starts the next process. After making preparations for restart, execute processing corresponding to the cause of the interrupt. When the interrupt processing is finished, the previously saved registers are restored,
Resume execution of the interrupted process by executing the return-from-interrupt instruction.

Since the interrupt processing executed by the operating system is common to all processes, it is uniform, and it is difficult to perform detailed processing according to the differences between each process. Therefore, in the case of an interrupt caused by a fatal failure in a computer system, if there is no general treatment method, the process that caused the interrupt is normally forcibly terminated. For example, an interrupt that occurs when an attempt is made to reference a non-existent main memory corresponds to this. As mentioned above, if the treatment for interrupts is uniform,
This often causes inconvenience.

In recent years, instructions with advanced functions have often been implemented in computers. These are instructions that are realized in such a way that functions that were conventionally achieved by executing multiple instructions in a series of steps can be executed with a single instruction. There are instructions that have the function of transferring data from one address to another. This type of instruction has complicated contents to be executed, and may include repeated functions, so the execution may take a long time. For example, if 100,000 pieces of data are to be transferred using the above transfer command, it will take about 100 milliseconds. This is an extremely long time for computers, which typically execute an instruction in nanoseconds to microseconds.

An interrupt usually ends the execution of a single instruction,
It is accepted before starting execution of the next instruction. Therefore, the high-performance instructions that are executed while occupying the computer for a long time as described above are not interruptible during their execution, which is inconvenient for the interrupt function. Therefore, many high-performance instructions are implemented so that they can be interrupted even during instruction execution. Of course, even if an interrupt occurs and instruction execution is interrupted, it is easy to resume instruction execution by saving the context at that point. If the address where the currently executed instruction is located is indicated, there will be no inconsistency.

Incidentally, the operating system attempts to uniformly handle interrupts that occur during the execution of high-performance instructions as described above, as well as other interrupts. Therefore, if a fatal failure occurs, the process will be forcibly terminated.
The execution results up to that point will also be lost, resulting in a loss. Even if a process is prevented from being forcibly terminated by interrupt processing, this will occur uniformly, resulting in another inconvenience. In the end, it is necessary to implement an interrupt handling program so that appropriate measures can be taken for each process, but it is difficult for the operating system to add different processing for each process to the interrupt handling program. This results in a lack of versatility, and it is usually very difficult to significantly modify the interrupt processing program of an existing operating system, which is not a practical method.

Therefore, the best method is to handle interrupt processing within the operating system using a uniform procedure, and leave processing for process-specific interrupts to the processes.

Purpose of the Invention It is an object of the present invention to provide a method that eliminates the drawbacks of the conventional interrupt handling described above and allows a process to handle process-specific interrupt handling for interrupts that occur during the execution of one instruction. do.

Structure of the Invention The present invention provides a flag area associated with a process and an error information area in main memory, so that a process and an interrupt processing program can communicate information to each other via the above areas. Accordingly, the interrupt processing program is configured to entrust process-specific interrupt processing to processes in response to requests from processes.

Description of Example If a process wants to perform the corresponding processing itself when an interrupt occurs during the execution of one instruction,
For example, the value “1” is stored in advance in the flag area, and if not, the value “0” is stored in advance, and
A value "0" meaning no error information is stored in advance in the error information area, and then execution of the instruction for which an interrupt in the middle of execution is to be detected is started. If no interrupt occurs,
Upon completion of instruction execution, execution of the instruction at the next address begins. On the other hand, if an interrupt occurs during instruction execution that would conventionally cause the process to be forcibly terminated, the interrupt handling program refers to the flag area and if the value is "0", the process is terminated as usual. If the process is forcibly terminated and the value is "1", a predetermined value other than "0", which means the occurrence of an interrupt, is stored in the error information area. As the value to be stored at this time, it is also possible to predetermine a plurality of different values depending on the type of interrupt. Next, the interrupt handling program adds a certain value determined by the computer to the value of the instruction counter in the saved process context, and the value of the instruction counter is added to the value of the instruction counter in the saved process context. Change it to show the street address. Thereafter, when the procedure for returning from the interrupt is executed as usual, execution of the process is resumed from the instruction following the interrupted instruction.

As you can see from the above explanation, by storing the value "1" in the flag area, even if an interrupt occurs during instruction execution, the process will continue executing the next instruction without being forced to terminate. It becomes possible to do so. Therefore, after the process finishes executing the instruction for which it wants to detect the presence or absence of an interrupt, it checks the error information area, and if the value remains "0", it means that there was no interrupt, so it continues the rest of the process. It is possible to
If it is a value other than "0", it means that an interrupt has occurred, so it is possible to determine that some kind of abnormality has occurred and take measures accordingly. Therefore, even if some abnormality occurs, it is possible for the process to take measures such as saving the execution results up to that point. Note that the present invention is applicable to a case where the location address of the next instruction after the interrupted instruction can be calculated using a fixed procedure. Furthermore, the present invention does not limit the values stored in the flag area and error information area to only the values used in the above description.

Effects of the Invention As described above, according to the method according to the present invention, only a small amount of certain procedures are added to the interrupt handling of the operating system, which previously had no choice but to forcibly terminate processes across the board. It enables a process to deal with missed interrupts in a method specific to that process, and greatly improves the degree of freedom in computer usage, which has great effects.

Claims (1)

[Claims] 1. A flag area and an error information area associated with a process are provided in main memory, and if the process stores a predetermined value in the flag area, the process is activated by the occurrence of an interrupt. The interrupt processing program stores predetermined error information in the error information area, changes the value of the instruction counter in the context saved at the time of the interrupt to indicate the location address of the next instruction, and returns from the interrupt. Therefore, by restarting execution of the above process from the instruction at the next address of the interrupted instruction, and checking the contents of the error information area on the above process side, the interrupt during execution of the previous instruction can be detected. 1. An interrupt processing method characterized by determining the presence or absence of a process and, if an interrupt occurs, executing interrupt processing specific to the process.