JPH04172537A - Failure analysis method for information processor - Google Patents

Abstract

PURPOSE:To enhance indicated resolution and precision of a defective part when the defect occurs by referring to a failure dictionary for parts including components of the processor, and by indicating a processor component part group included in a limited cover area as defective parts. CONSTITUTION:A failure dictionary 20 is prepared on the basis of a result of extracting in advance a cover area for total failure detecting circuit, and consists of a failure detecting circuit, processor components, and correspondence information for selector signals, cover for failure defecting circuit, information of connecting between logical elements included in the area, and so on. Failure analysis section 100 recognizes a leading light failure detecting circuit (FD) from a detected state of failure information 10, obtains bus selecting information from information 10 and the corresponding information of selector signals stored in dictionary 20, analyses the bus selecting information, and specifies a data transfer bus in the case of the occurrence of a failure. Failure analysis section 200 extracts a cover area from the content of the register of information 10 and inter-register connecting information. Failure analysis section 300 extracts a cover area for the case where a plurality of FDs exist. With this, a cover area is limited, thereby enhancing the indicated resolution and its accuracy.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention relates to a fault analysis method that analyzes faults that occur during the operation of an information processing device and points out faulty parts, and particularly relates to a fault analysis method that identifies faulty parts by analyzing faults that occur during the operation of an information processing device. In other words, the present invention relates to a failure analysis method suitable for improving the pointing resolution i and the pointing accuracy of faulty parts.

[Conventional technology]

Generally, a concurrent error check diagnosis method is used in information processing devices so that tests can be performed in parallel during operation and even intermittent failures that are not reproducible can be diagnosed. In this current error check diagnostic method, a fault detection circuit is built into the device to detect faults that occur during operation.
When a fault occurs, the detection states of these fault detection circuits are analyzed to find the fault detection circuit that lights up first, and by referring to the fault dictionary that has been activated in advance, the fault maintenance/replacement unit is pointed out. Fault repair processing is performed by replacing this faulty replacement unit.

By the way, the above-mentioned fault dictionary contains a correspondence table for each fault detection circuit that points out the maintenance/replacement unit that includes the fault that caused the fault detection circuit to turn on, and its accuracy is This affects the performance of the concurrent error check diagnosis method, namely resolution (ability to pinpoint fault location) and success rate (accuracy in pinpointing faulty parts). Therefore, it is important to create a highly accurate fault dictionary in order to facilitate investigation of the cause and enable prompt repair when a fault occurs.

However, creating this fault dictionary requires special design knowledge and is highly dependent on manual labor by skilled experts, and the logical scale of computers will expand dramatically in the future as parts become more highly integrated. However, there were concerns about a rapid increase in the number of man-hours required to create dictionaries and a decline in the quality of dictionaries.

It is against this background that the fault diagnosis dictionary creation method described in Japanese Patent Application Laid-open Publication No. 8836/1983 was developed.

The purpose is to automate the creation of a fault dictionary by extracting replacement parts that are expected to cause a fault corresponding to the fault detection circuit from logic design information.

This not only reduces the number of man-hours required to create a dictionary, but also prevents omissions or incorrect indications due to manual errors, and improves the quality of the dictionary.

Furthermore, JP-A-63-10244 describes a fault dictionary creation method for improving the resolution of a fault dictionary by removing areas where fault propagation is not possible using logic simulation. Publication No. 2-10440 describes a fault dictionary creation method for improving the pointing accuracy of a fault dictionary by adding a priority order to each pointing exchange unit. All of these are aimed at improving the accuracy of failure dictionaries, making it possible to improve maintenance efficiency.

(Problems to be Solved by the Invention) Since the above-mentioned conventional technology creates a fault dictionary using logical design information in advance before the computer starts operating, it is difficult to limit the fault location to the logical part that actually operated at the time of the fault. The problem was that the accuracy of identifying faulty parts was insufficient.

Furthermore, in recent years, due to advances in hardware technology such as higher integration of elements and higher density packaging, the maintenance/replacement units that constitute devices have become larger and their number has been decreasing. The failure dictionary created by the above-mentioned conventional technology does not take into consideration the environment surrounding the information processing device, and is premised on being used in the maintenance stage.
Failure parts (L) installed on large-scale maintenance and replacement units
There was a problem that the pointing resolution of SI) was insufficient.

Additionally, from maintenance replacement units replaced at the maintenance site using the concurrent error check diagnostic method.

When narrowing down the faulty parts that are the cause of a failure, the cause is investigated based on the results of a failure reproduction test using a hardware tester or diagnostic program. However, in this reproduction test, it often takes a long time for the failure to be reproduced, and if the failure cannot be reproduced, the expensive failed replacement unit is often discarded. This means that they are powerless against intermittent failures, and there is a strong need for an analysis method that does not depend on the type of failure, even in factories.

In addition, the lack of skilled professionals has led to a decline in repair efficiency.

The purpose of the present invention is to improve the resolution and accuracy of pointing out faulty parts when a fault occurs, without necessarily assuming advanced technical knowledge, by implementing systematic standard processing based on the knowledge of experts. The objective is to provide an analysis method.

[Means to solve the problem]

In order to achieve the above object, the present invention defines the first lighting fault detection circuit as the first lighting fault detection circuit based on fault information indicating the detection state of the fault detection circuit collected by a maintenance support device when a fault occurs. However, if there is path selection information at the time of failure regarding the leading lighting failure detection circuit recognized based on the failure information, a data transfer path at the time of failure is identified according to the path selection information and circuit tracing is performed. Accordingly, there is provided a means for increasing the cover area (fault detectable area) of a leading lighting fault detection circuit that eliminates logic unrelated to actual operation, and a leading lighting fault detecting circuit recognized based on the fault information. A parity check is performed on all registers, and if there is an error register in which the fault data that caused the first lighting fault detection circuit to turn on, that is, parity error information remains, the inter-register connection information is also checked. means for narrowing down and extracting the cover area of the leading lighting fault detection circuit by first identifying the error register in which the fault data is set and performing circuit tracing using the error register as a starting point; If there are multiple leading lighting fault detection circuits recognized based on the fault detection circuits at the same time, a common area of the cover area of each leading lighting fault detection circuit is extracted, and the common area is used for each leading lighting fault detection circuit. and a means for pointing out as the most suspicious part where the fault that caused the detection circuit to turn on is present, and by referring to a fault dictionary containing device component information, the device configuration included in the limited coverage area is detected. This system is designed to point out a group of parts as a failed part.

[Effect]

In the present invention, since the coverage area of the fault detection circuit is narrowed down based on the fault information frozen at the time of a fault, it is possible to limit the fault location to the logic part that actually operated, and the fault detection resolution and fault detection accuracy are improved. In addition, since it is based on fault information, it is possible to perform analysis regardless of the type of fault, and it is possible to identify faults at the maintenance site and at the factory where information processing equipment is produced. In this case, there is no need for a reproduction test, and the faulty part (LSI) on the fault maintenance/replacement unit is pointed out. Additionally, by standardizing the process, maintenance and repairs can be performed without necessarily requiring advanced technical knowledge.

〔Example〕

Hereinafter, the operating principle and one embodiment of the present invention will be explained using the drawings.

First, the operating principle of the present invention will be explained.

The area including the fault that causes the fault detection circuit to turn on when 1 is lit is the coverage area of the fault detection circuit (fault detectable area).
The basic information of the fault dictionary used in the concurrent error check diagnostic method called ``Fault Detection Circuit'' is the correspondence information between such a fault detection circuit and the components included within its coverage area.

When creating a fault dictionary, conventional coverage area recognition methods automatically recognize the coverage area of fault detection circuits from logic design information, and start from each fault detection circuit in the opposite direction to the logic flow. Trace (backtrace) to
By setting the parity generation target register corresponding to the fault detection circuit and the parity check target register corresponding to the fault detection circuit other than the fault detection circuit as stopping points, the coverage area of the fault detection circuit is recognized. (See Figure 5(a)).

In the present invention, the starting point and stopping point of a trace are clearly determined based on the path selection information regarding the fault detection circuit that is lit and the register information that holds the fault data, which is included in the fault information collected at the time of a fault. After setting, by performing a cover area extraction trace, it becomes possible to narrow down as much as possible the areas most likely to include the failure that caused the failure.

FIG. 1 is a schematic diagram showing the configuration of an information processing apparatus failure analysis method 1, which is an embodiment of the present invention. The input information is fault information 10 and fault dictionary 20. The fault information 10 includes information indicating the detection state of a fault detection circuit built into the device, information indicating the set/reset state of the selector signal, information indicating the set/reset state of the tracer signal, and the contents of the register at the time of freezing. It consists of information such as the number of faulty machine cycles, etc. The fault dictionary 20 is created based on the results of previously extracting the coverage areas of all fault detection circuits.
It consists of correspondence information between the failure detection circuit and device components, correspondence information between the failure detection circuit and selector signals, and connection relationship information between logical elements such as gates and registers included within the coverage area of the failure detection circuit. The output information is a failed parts indication list 3°. The faulty parts indication list 30 includes the name of the leading lighting fault detection circuit, the name of the suspected faulty part, and its mounting position information, and may be displayed on a display device or printed out by a printer.

Failure analysis method 1 is a failure analysis unit 10 based on path selection information.
Failure analysis unit based on 0 and error register information 2. It is composed of a failure analysis unit 300 based on OQ and common coverage area information. The above three analysis units are implemented depending on the type of target logic, and may be selectively implemented individually or in combination.

The details of each analysis section will be explained below.

FIG. 2 is a flowchart showing the processing procedure of the failure analysis unit 100 based on path selection information.

In step 110, from the information indicating the detection state of the fault detection circuit included in the fault information 10, the fault detection circuit that is turned on due to the first detection of a fault is recognized and defined as the first lighting fault detection circuit. Recognition of the leading lighting failure detection circuit has conventionally been easily realized using information on the mutual dependency of the failure detection circuits or an identification counter circuit built into the device as a hardware function. At this time, based on the information indicating the set/reset state of the selector signal included in the fault information 10 and the correspondence information between the fault detection circuit and the selector signal included in the fault dictionary 2, Check whether a selector exists or not, and if so, obtain the path selection information frozen at the time of failure detection (120), then step 13
0, the path selection information is analyzed to identify the data transfer path at the time of failure. If there is no guarantee that the path selection information is frozen, it is also possible to correct the selection information by analyzing the operating state at the time of the failure based on the tracer signal. Once the data transfer path has been identified, similar to the conventional coverage area recognition method, backtrace is performed starting from the first lighting fault detection circuit, and covers up to the stopping point of the parity check target register of other fault detection circuits. Extract as an area (140). At this time, if a selector is reached during tracing, excluding the specified data transfer path,
Since tracing is stopped, the covered area is narrowed down compared to conventional methods. When the extraction of the cover area is completed,
Point out equipment components within the area as faulty parts (15
0).

FIG. 3 shows a failure analysis unit 200 based on error register information.
2 is a flowchart showing a processing procedure.

In step 210, the leading lighting fault detection circuit is recognized based on the fault information 10, and information indicating the contents of the register at the time of freezing included in the fault information 10 and within the coverage area of the fault detection circuit included in the fault dictionary 20 are detected. A parity check is performed on all registers related to the leading lighting failure detection circuit based on the inter-register connection information (220). If there is an error register in which parity error information (failure data) that caused the leading lighting failure detection circuit to turn on remains (230), the error register is selected based on the connection information between registers. First, identify the error register in which the fault data is set and use it as the starting point for extracting the cover area (240). To extract the cover area of the leading lighting fault detection circuit, a backtrace is performed from this starting point (240).
250).

At this time, the area between this first error register and the leading lighting failure detection circuit can be excluded as outside the coverage area. When the extraction of the cover area is completed, the device components within the area are pointed out as failed parts (260).

FIG. 4 shows a failure analysis unit 300 based on common coverage area information.
2 is a flowchart showing a processing procedure.

In step 310, the leading lighting failure detection circuit is recognized based on the failure information 10. Normally, there is only one leading lighting fault detection circuit, but in rare cases, multiple leading lighting fault detection circuits may exist because multiple fault detection circuits detect the fault at the same time. Therefore, it is necessary to check whether there are a plurality of leading lighting failure detection circuits (320).

If there are more than one, extract the cover area of each leading lighting failure detection circuit and recognize the common area (330)
. In step 340, a coverage area is extracted for the only leading lighting failure detection circuit. Step 33
The extraction of the cover area in 0 to 340 may be performed according to the processing procedure of the failure analysis unit 100 based on the path selection information and the failure analysis unit 200 based on the error register information.

A conventional method may be used. Device components included within the common area of the obtained coverage areas are pointed out as faulty parts (350).

The effects of this embodiment will be explained using FIGS. 5 and 6. In FIG. 5, 510 to 513 are failure detection circuits;
20 to 528 are registers.

530 is a selector signal (XXSELA), and when the logical value is 11011, it selects the data transfer path 540 and “
1”, select 541. 531 (XXSELB
) is also similar. Assume that under such a circuit configuration, the failure detection circuit 510 (ACK) lights up first. Fifth
Figure (cL) shows a leading lighting failure detection circuit 5 using a conventional method.
10 coverage areas 560 are shown, and the identified failed parts are:
LSII (550), LSI2 (551), LSI3
(552), LSI4 (553). Figure 5 (
b) indicates that when the selector signal 530 is set to the logical value "1", the cover area 561 has been extracted as a result of the failure analysis 100 based on the path selection information.

The indicated failed part is LSII (550).

They are LSI3 (552) and LSI4 (553).

FIG. 5(c) shows that when the register 521 first sets fault data, a cover area 562 is extracted as a result of fault analysis 200 based on error register information. What must be noted here is that the registers related to the failure detection circuit 510 are 520 to 525.
However, since the register 520 is the target register for the parity check of the failure detection circuit 510, the register in which the failure data is set first is searched from among the other registers. This can be easily achieved using logical connection relationships between registers. The faulty parts pointed out here are
They are LSI2 (551) and LSI3 (552).

In FIG. 6, a fault detection circuit 610 (CHKl), 6
20 (CHK2) are both leading lighting failure detection circuits. As a result of performing failure analysis 300 using common coverage area information, a common area indicated by diagonal lines is extracted. This common area allows 630 failed replacement units (FRU
I) is not only pointed out, but also the parts mounted on the replacement unit 630 are pointed out.

In any of the above failure analyzes, the coverage area is limited and extracted compared to conventional methods, so it is possible to improve the resolution and accuracy of pointing out failed parts.

Furthermore, by standardizing the processing, efficient maintenance and repair becomes possible.

It is clear that the invention is not limited to the above embodiments.

〔Effect of the invention〕

According to the present invention, since the coverage area of the fault detection circuit is narrowed down based on the fault information frozen at the time of the fault, it is possible to limit the fault location to the logic part that actually operated, and the resolution of faulty parts is improved. This is effective in improving pointing accuracy.

In addition, since it is based on fault information, it is possible to perform analysis regardless of the type of fault, such as fixed faults or intermittent faults, and at maintenance sites, it is possible to point out faults in units of maintenance replacements, and to use information processing equipment. There is no need for reproduction tests at the production factory.

Since it is possible to point out a failed component on a failure maintenance replacement unit, it has the effect of shortening the work time for failure recovery processing. In addition, standardization of processing allows maintenance and repairs that do not require advanced specialized knowledge.

[Brief explanation of the drawing]

FIG. 1 is a schematic diagram of the configuration of an information processing device failure analysis method that is an embodiment of the present invention, FIG. 2 is a diagram showing the processing procedure of the failure analysis unit based on the path selection information included in FIG. 1, and FIG. 2 is a diagram showing a processing procedure of a failure analysis unit based on error register information included in FIG. 1. FIG. Figure 4 is a diagram showing the processing procedure of the failure analysis unit based on the common coverage area information included in Figure 1, and Figures 5 (α), (b), and
) is a conceptual diagram of a fault detection circuit using a model circuit and a cover area, and FIG. 6 is a conceptual diagram of a common cover area. 1... Information processing device failure analysis method % formula % 30... Failure parts indication list 100-150... Failure analysis processing using path selection information 200-260... Failure analysis processing using error register information 300-350 ... Fault analysis processing based on common coverage area information 510-513 ... Fault detection circuits 520-528 ... Registers 530-531 ... Selector signals 540-543 ... Data transfer bus routes 550-5
53...Device components (LSI) 560-562...
・Cover area 610-620...Failure detection circuit 611-621...Cover area 630-630...Maintenance/replacement unit unit collection Figure 5 (b) Figure 5 (C) Figure 5

Claims (1)

[Scope of Claims] 1. Fault information when a fault occurs in the device, including information indicating the detection state of a fault detection circuit built into the device, in order to detect a fault that occurs during the operation of the information processing device. , a fault detection circuit that lights up first because a fault has been detected when a fault occurs is defined as the first lighting fault detection circuit, and the first lighting fault detection circuit is the target of fault detection. A fault analysis method that points out a group of equipment component parts included in the logic (fault detectable area or coverage area) as a faulty part, and in the event of a fault related to the first lighting fault detection circuit recognized based on the fault information. If path selection information exists,
means for extracting a cover area of a leading lighting failure detection circuit that excludes logic unrelated to actual operation by identifying a data transfer path at the time of failure according to all of the path selection information and performing circuit tracing; A parity check is performed on all registers related to the leading lighting fault detection circuit recognized based on the fault information, and the fault data that caused the leading lighting fault detection circuit to turn on, that is, parity error information, is determined. If there are remaining error registers, the error register in which the fault data was first set is identified based on the connection information between the registers, and circuit tracing is performed using the error register as a starting point, so that the first light is turned on. a means for narrowing down and extracting a coverage area of a fault detection circuit; and a means for narrowing down and extracting a cover area of a first lighting fault detection circuit; and if a plurality of leading lighting fault detection circuits recognized based on the fault information detect faults at the same time; An information processing device failure characterized by comprising: means for extracting a common area of the areas and pointing out the common area as the most suspicious part where the failure that caused each head lighting failure detection circuit to turn on is present. Analysis method.