JPS63284652A - Patrol diagnosis system for shared memory - Google Patents

Abstract

PURPOSE:To improve patrol efficiency of a shared memory by performing the patrol to a specific area of the shared memory via a master processor only. CONSTITUTION:A #0-side host computer 1 and a #1-side host computer 2 share a shared memory 5 via a #0-side processor 3 and a #1-side processor 4 respectively. The processor 3 patrols periodically the corresponding shared memory area via a patrol diagnosing means 36 is response to the interruption signal supplied from a patrol diagnosis executing timer 31 via an interruption generating means 35 and based on the shared memory control information registered into a shared memory control table 32 of the means 36 and the shared memory control information read via an interface 6 and registered into a shared memory control table 42 of the processor 4. The processor 4 confirms the working of the #0-side processor via the interface 6 and then patrols the shared memory area only when the abnormality is detected.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a patrol diagnosis method for a shared memory shared by a plurality of processors of an input/output device constituting an information processing system.

[Prior Art] Conventionally, this type of patrol diagnosis method uses multiple processors to check the entire area of a shared memory after a certain period of time, regardless of the operation of other processors, without having an interface with each other. They were conducting patrol diagnosis.

[Problems to be Solved by the Invention] In the conventional patrol diagnosis method described above, each of the plurality of processors has a patrol diagnosis execution timer 1 to 1, and when the patrol diagnosis execution timer reaches a predetermined time, Since patrol diagnosis is performed on all areas of one shared memory regardless of the relationship, patrol diagnosis is performed more than necessary, which has the drawback of being very inefficient.

SUMMARY OF THE INVENTION Therefore, an object of the present invention is to provide a shared memory patrol diagnosis method that can efficiently perform patrol diagnosis of a shared memory.

[Means for Solving the Problems] A shared memory patrol diagnosis method according to the present invention is a patrol diagnosis method for a shared memory shared by a plurality of processors, and each of the plurality of processors has the shared memory Divide memory into input/output operation units,
a shared memory management table that manages the shared memory while registering a shared memory area in which there has been no read operation after a write operation as shared memory management information; and a patrol diagnosis execution timer that is set when the processor starts up.
means for generating an interrupt when the patrol diagnosis execution timer reaches a predetermined time; and an interface for directly connecting the plurality of processors; When the interrupt occurs, reads the shared memory management information registered in its own shared memory management table and the shared memory management information registered in the shared memory management tables of all other processors via the interface, A means for patrolling and diagnosing the shared memory area corresponding to the shared memory management information registered in the shared memory management table, and a means for notifying all other processors to that effect via the interface when the patrol diagnosis is completed, and means for resetting the patrol diagnosis execution timer of the processor; and means for clearing the shared memory management information registered in its own shared memory management table when the patrol diagnosis is completed; , means for investigating whether or not the predetermined one processor is operating normally via the interface when the interrupt occurs, and patrolling the shared memory only when normal operation cannot be confirmed by the investigation. means for performing a diagnosis; means for resetting its own patrol diagnosis execution timer in response to the notification; and means for clearing shared memory management information registered in its own shared memory management table in response to the notification; It is characterized by having the following.

[Example] Hereinafter, an example of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a block diagram showing the overall configuration of an embodiment of the present invention, in which a 0-side host computer 1.1-side host computer 2.0-side processor 3.1-side processor 4,
It consists of a shared memory 5 and an interface 6 between the 0th processor 3 and the 1st processor 4. The 0th processor 3 and the 1st processor 4 are control devices for input/output devices, such as magnetic disk control devices, and the shared memory 5 stores lock information and the like.

Next, the patrol diagnosis method based on this practical rejection example will be explained in detail using FIG. 0 side processor 3 and 1
Each side processor 4 has patrol diagnosis execution timers 31 and 41. When the processor is started up, the clearing means 33 and 43 clear the shared memory management information registered in the respective shared memory management tables 32 and 42 and divided into input/output operation units, and then the patrol diagnosis execution timer 31 and 41 is set by the notification/resetting means 34 and the resetting means 44, and when the set time comes, the interrupt generating means 35 and 45 generate an interrupt. That is, an interrupt is generated every time a certain period of time elapses.

Each processor 3 and 4 has a host computer 1.
After performing a write operation to the shared memory 5 at the request from 2,
Shared memory management table 3 uses shared memory areas with no read operations as shared memory management information for each input/output operation.
2 and 42.

Furthermore, in each processor, an interrupt occurs after a certain period of time has passed, as described above, regardless of the request from the post computer 1.2.

At that time, the patrol diagnostic means 36 of the processor 3 on the 0th side
unconditionally reads the shared memory management information registered in its own shared memory management table 32 and the shared memory management information registered in the shared memory management table 42 in the first processor 4 via the interface 6. , patrol diagnoses the shared memory areas corresponding to the shared memory management information registered in these shared memory management tables, and confirms the normality of the shared memory 5.

On the other hand, when an interrupt occurs, the checking means 47 of the first processor 4 uses the interface 6 with the zero processor 3 to check whether or not the zero processor 3 is operating normally. If it is operating normally, it is assumed that the patrol diagnosis for the shared memory 5 is also being performed normally, and the diagnostic means 47 does nothing. However, if it is not operating normally, the diagnostic means 47 notifies the patrol diagnostic means 46 of this fact, and the patrol diagnostic means 4
6 performs patrol diagnosis on the shared memory 5 and confirms the normality of the shared memory 5.

Incidentally, when the patrol diagnosis means 36 of the 0th side processor 3 completes the patrol diagnosis of the shared memory 5, the patrol diagnosis means 36 sends an end signal to the notification/resetting means 34 and the clearing means 33. Upon receiving the end signal, the notification/resetting means 34 notifies the first processor 4 of this via the interface 6, and then resets the patrol diagnosis execution timer 31. Upon receiving the end signal, the clearing means 33 clears the shared memory management information registered in its own shared memory management table 32. Also, 1
In the second processor 4, upon receiving the above notification, the reset means 44 resets the patrol diagnosis execution timer 41. Then, upon receiving the above notification, the clearing means 43 clears the shared memory management information registered in its own shared memory management table 42.

In this way, instead of performing patrol diagnosis on the shared memory 5 from both processors, the 0-side processor 3 mainly performs a patrol diagnosis on a portion of the shared memory 5, and the 0-side processor 3 Only when it is not operating, the shared memory 5 is partially patrolled and diagnosed by the first processor 4.

In the above embodiment, the 0th processor 3 primarily performs the patrol diagnosis of the shared memory 5, but the 1st processor 4 may primarily perform the patrol diagnosis of the shared memory 5. Furthermore, in the above embodiment, the number of processors may be two or any number greater than or equal to three. In this case, an interface is provided between each processor, a predetermined processor is operated as a mask like the 0th processor 3 in the above embodiment, and all other processors are operated as the 1st processor 4 in the above embodiment. You can operate it as a slave like this.

[Effects of the Invention] As explained above, the present invention does not perform shared memory patrol diagnosis on the entire area of the shared memory from each processor, but utilizes the interface between the processors and performs shared memory patrol diagnosis on the master processor 10. = By performing the patrol diagnosis on a specific area of the shared memory only from the server, there is an effect that the shared memory can be efficiently patrolled and diagnosed.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the configuration of one practical example of the present invention. 1... No. 0 side host computer, 2... No. 1 side host computer, 3... No. 0 side processor, 4...
- 1st processor, 5... shared memory, 6... interface.

Claims (1)

[Scope of Claims] 1. In a patrol diagnosis method for a shared memory shared by a plurality of processors, each of the plurality of processors is provided with a system that divides the shared memory into input/output operation units and performs a read operation after a write operation. A shared memory management table that manages the shared memory while registering inactive shared memory areas as shared memory management information, a patrol diagnosis execution timer that is set when the processor starts up, and a patrol diagnosis execution timer that is set in a predetermined manner. means for generating an interrupt when the time comes, and an interface for directly connecting the plurality of processors, and a predetermined one of the plurality of processors is configured to generate an interrupt when the interrupt occurs. Reads the shared memory management information registered in its own shared memory management table and the shared memory management information registered in the shared memory management tables of all other processors via the interface, and processes the shared memory management information registered in these shared memory management tables. means for performing a patrol diagnosis on a shared memory area corresponding to the shared memory management information, and when the patrol diagnosis is completed, notifying all other processors to that effect via the interface and resetting their own patrol diagnosis execution timer. and means for clearing shared memory management information registered in its own shared memory management table when the patrol diagnosis is completed, and each of the other processors is provided with a means for clearing the shared memory management information registered in its own shared memory management table when the patrol diagnosis is completed, and each of the other processors is configured to clear the interface when the interrupt occurs. means for investigating whether or not the predetermined one processor is operating normally through the notification; means for performing patrol diagnosis on the shared memory only when normal operation cannot be confirmed by the investigation; A sharing device characterized by comprising: means for resetting its own patrol diagnosis execution timer in response; and means for clearing shared memory management information registered in its own shared memory management table in response to the notification. Memory patrol diagnostic method.