JPS63192141A - Log-out system for fault information of multi-processor system - Google Patents

Abstract

PURPOSE:To improve the performance of the titled log-out system by performing a log-out action for the hardware information only as necessary in case an exceptional interruption occurs during the execution of a program. CONSTITUTION:A writable memory means 50 is set into a system control part 40 provided between a main memory 30 and a system bus 10. When the control information is registered into the means 50, the means 50 works to form a mask to decide whether an exceptional log-out action should be carried out or not in response to each bit. As a result, the control parts 2101-210n of processing units 2001-200n transfer the contents of the log-out information parts 2201-220n to the log-out areas 3001-300n of the main memory 30 respectively if the corresponding bit of said mask is set when an exceptional interruption is produced. While the contents of the parts 2201-220n are not transferred if the corresponding bit of the mask is reset. Thus the system maintenance properties can be improved.

Description

[Detailed Description of the Invention] [Objective of the Invention] (Field of Industrial Application) The present invention relates to a fault information logout method for a multiprocessor system that logs out fault information regarding hardware when an exception occurs due to software factors. .

(Prior Art) When a hardware failure occurs in a computer system and an abnormality such as malfunction occurs, this must be detected as a hardware abnormality.

However, it is generally difficult to detect hardware malfunctions, and therefore, in computer system configurations, malfunctions are often detected as software abnormalities such as program exceptions.

Moreover, when such a software abnormality is detected, no fault information related to the hardware remains even if a subsequent analysis reveals that it is actually a hardware abnormality.

Therefore, it was first proposed to log out hardware failure information when an exception occurs, which may be caused by a hardware abnormality, and to save this logged out hardware failure information.

However, the conventionally proposed logout method has a problem in that the logout determination operation is fixed, and the user ends up logging out even when it is not necessary.

(Problem to be Solved by the Invention) In other words, in a computer system, an exception interrupt that occurs during program execution may be caused by a program error, an arithmetic error, an address conversion, etc. Exceptions are classified into several types.

It is not desirable in terms of system performance to log out all of these as hardware failures. Particularly in a system that performs virtual memory control, logging out at the time of an address translation exception will significantly degrade system performance.

As described above, even in cases where system performance is significantly degraded, the conventional logout method has a problem in that the logout determination operation is fixed, resulting in the user being logged out.

The present invention has been made in view of the above circumstances, and an object of the present invention is to log out hardware information only when necessary when an exception interrupt as described above occurs in a system having a multiprocessor configuration.

[Structure of the Invention] (Means for Solving the Problems) In order to achieve the above object, the present invention provides a multiprocessor system in which a plurality of CPUs are connected to a system bus to which a main storage device is connected. A writable storage means provided in a system control unit provided between a main storage device and the system bus, and a writable storage means for this storage means,
registration control means for registering control information indicating the type of exception caused by software factors in response to a predetermined input operation; The logout determining means is provided with a logout determination means when a logout determination occurs, and each time the registration control means changes the control information registered in the storage means, the determination criteria of the logout determination means is changed accordingly. shall be.

(Function) As described above, in the present invention, a writable storage means is provided in the system III control unit provided between the main storage device and the system bus, and the control information registered in this storage means is registered and controlled. By changing the logout determination means, the determination criteria of the logout determination means are changed, so that hardware-related failure information can be logged out in a manner that corresponds to the usage status and purpose of the system.

(Embodiment) FIG. 1 is a block diagram showing an embodiment of a multiprocessor system to which the failure information logout method of the present invention is applied.

In this embodiment, there are a plurality of processors 20 on the system bus 10, and a main memory 30 common to each of these processors 20. A system control module 40, a storage means 50 within this module, and a service processor 60 including a terminal device 70 are connected to form a multiprocessor.

In this multiprocessor configuration, logout areas 3001 to 300n are secured in the main storage device 30 to correspond to the respective processing units 2001 to 200n of the plurality of processors 20.

Exception logout control information is registered in the storage means 50 within the system control module 40 by the service processor 60. This storage means 50 is a non-volatile memory capable of storing N data.

On the other hand, each of the processing units 2001 to 20On of the plurality of processors 20 includes control units 2101 to 20On.
10n is provided, and these control units 2101 to 210n are
Control information is read from the storage means 50 in the system control module 40, a logout determination is made using this control information as a determination criterion, and according to this logout determination, the logout area 1ff3001 of the main storage device 30 is transferred from the logout information section 2201 to 22On. This is to transfer lockout information to 300n.

Further, the storage means 50 in the system control module 40 is a register having a content structure as shown in FIG. 2 when controlling according to the type of exception. Further, if control is to be performed according to the cause of each exception, a plurality of mask registers will be prepared for each exception as shown in FIG.

Next, the operation will be explained. Since the storage means 50 in the system control module 40 is programmable and readable/writable by the service processor 60, each 1-bit control information indicating each type of exception caused by software factors is stored as appropriate. Registration is possible.

Therefore, in cases where system performance is significantly degraded, such as during address translation, for example, it is possible to register so as not to log out. Furthermore, if it is desirable to log out even if the system performance deteriorates during system development or maintenance, it is also possible to register to log out.

In other words, since the type of exception that requires logout and the degree of performance degradation due to logout vary depending on the usage status and purpose of the system, the service processor 60 stores the DJ information in the storage means 50 in a corresponding manner.
can be registered.

In this way, when the control information is registered in the storage means 50, this storage means 50 constitutes a mask for determining whether or not exception logout is performed for each bit. Therefore,
In each of the control units 2101 to 210n of each processing unit 2001 to 200n, if the corresponding bit of this mask is set when an exception interrupt occurs, the contents of the logout information section 2201 to 220n are transferred to the logout areas 3001 to 300n of the main storage device 30. Transfer to 300n. Further, if the corresponding bit of the mask is reset, such transfer operation is not performed.

Furthermore, since the present invention is applied to the configuration of a multiprocessor system, at the initial stage of system startup,
There is a high possibility that an exception will occur that will cause you to log out.

Considering this case, the storage means 50 is non-volatile.
For example, E2 PROM is used. Since this E2 PROM is nonvolatile and self-installable, there is no need to perform initial settings each time the system is started up, and it can respond sufficiently and flexibly to needs.

[Effects of the Invention] As explained above, in a multiprocessor system to which the failure logout method of the present invention is applied, 0) When an exception occurs in the lower control unit, failure information regarding the hardware is set appropriately. You can get logged out on the street.

(2) The system performance degradation due to logout can be kept as small as possible.

(3) Furthermore, since it is possible for the hardware to obtain more timely failure information, the maintainability of the system is greatly improved.

Benefits such as:

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an embodiment of a multiprocessor system to which the fault information logout method of the present invention is applied;
FIGS. 2 and 3 are diagrams each showing an example of the content of control information in the storage means. 10... System bus 20... Processor 30... Main storage device 40... System i control module 50... Storage means 60... Service processor

Claims (1)

[Claims] (1) Multiple CPUs on the system bus connected to the main memory
In a multiprocessor system to which a PU is connected, a writable storage means provided in a system control unit provided between the main storage device and the system bus; registration control means for registering control information indicating the type of exception caused by software factors; and for the plurality of CPUs, determining corresponding control information registered in the storage means to determine logout when an exception interrupt occurs. A multiprocessor comprising: a logout determination means for performing the following: each time the registration control means changes the control information registered in the storage means, the determination criteria of the logout determination means is changed accordingly. System fault information logout method.