JP6914430B2 - Memory diagnostic device and memory diagnostic method - Google Patents

Description

The present invention relates to a memory diagnostic apparatus and a memory diagnostic method, and more specifically to a diagnosis of the presence or absence of soft errors and hard errors in a rewritable storage medium such as RAM (Random Access Memory).

Rewritable storage media such as RAM are used in information processing devices such as PCs (Personal Computers) and PLCs (Programmable Logic Controllers) as storage areas for various settings that are appropriately changed during the development of control programs and the use of the devices. Be done. When the storage medium is used in this way, damage to the data storage area or damage to the data itself due to a failure of the storage medium may cause a malfunction of the information processing apparatus.

The functional safety standardized in IEC 61508 by the International Electrotechnical Commission (IEC) is based on the premise that the machine is broken, and the occurrence of dangerous failures is detected by the function of the device, and the safety state is achieved. The idea is to ensure safety by shifting to. In order to prevent malfunction of the information processing device due to the failure of the storage medium, it is important to have a mechanism to detect a dangerous failure as soon as possible and shift the device to a safe state in accordance with the above concept of functional safety. be.

It is known that soft errors occur in memory as well as hard errors caused by physical factors. A soft error is a temporary error caused by α rays, neutron rays, EMI noise, internal crosstalk, and the like. That is, unlike a hard error, a soft error is not a permanent error and can be returned to normal by rewriting data or restarting the device. On the other hand, the occurrence of a soft error in the operating memory may cause a malfunction of the information processing device, so that reliable detection is required.

According to Japanese Patent Application Laid-Open No. 2012-104064 (Patent Document 1), a storage area for storing safety-related data is subjected to RAM diagnosis by a double RAM method when the safety-related data is accessed (hereinafter, also referred to as “double RAM diagnosis”). ), A failure diagnosis device for reliably and early detecting a RAM failure by periodically executing RAM diagnosis using test pattern data (hereinafter, also referred to as "test pattern diagnosis") is described. ..

JP 2012-104064

In the test pattern diagnosis, the presence or absence of a RAM failure is diagnosed by writing and reading the test pattern data after copying the data originally stored in the RAM to another area. Therefore, while the failure of the memory cell itself (that is, a hard error) can be detected, the certainty of the data itself stored in the RAM is not confirmed. Therefore, it is difficult to detect an abnormality in the data content due to the above-mentioned soft error.

Further, in the test pattern diagnosis, since it is necessary to write and read the test pattern data, it becomes impossible to read the data for the diagnosis target area. Therefore, if data output from the diagnostic area is requested during the test pattern diagnosis, the test pattern diagnosis needs to be interrupted. As a result, if the time required for the test pattern diagnosis becomes long, the diagnosis cycle of each region becomes long, so that the time from the occurrence of the RAM failure to the detection by the diagnosis may become long.

On the other hand, in the double RAM diagnosis, the presence or absence of a RAM failure is diagnosed by inverting and comparing the data corresponding to the two memories when writing the data to the RAM and the inverted data when writing and reading the data. Will be done. Since the double RAM diagnosis can be executed every time the data is read from the RAM area, the abnormality can be detected immediately when the data that has become abnormal due to a hard error or a soft error is used for the calculation. However, a RAM area with low access frequency, in which both writing and reading are rarely performed, is less likely to be subject to double RAM diagnosis. Therefore, in a usage environment where there is a difference in access frequency between regions, there is a risk that an abnormality in the region cannot be detected for a long period of time. Further, since the data written in the RAM differs depending on the environment in which the information processing device is used, for example, "1" is continuously written to the memory cell in which the memory value is stuck to "1". In such a usage environment, it is difficult to detect an abnormality in the memory cell by double RAM diagnosis.

If RAM failures are not detected for a long period of time due to these factors, failures may accumulate in the same area of the two memories, leading to malfunction of the information processing device. Therefore, it is desirable to detect a RAM failure as soon as possible after it occurs.

In Patent Document 1, the self-diagnosis rate is improved by using both the double RAM diagnosis and the test pattern diagnosis, but it is necessary to suspend the test pattern diagnosis due to an interrupt when executing the safety-related application process. Therefore, since the total time until the test pattern diagnosis is completed becomes long, the failure detection by the test pattern diagnosis may be delayed in the system in which the safety function execution process needs to be performed frequently. In addition, the address indicating the storage area of safety-related data is held, and when access to the corresponding address is detected, double RAM diagnosis is performed, but in the memory cell with low access frequency as described above, after an abnormality occurs. Since it takes time to detect the above, the accumulation of failures may cause a failure.

The present invention has been made to solve such a problem, and an object of the present invention is to promptly and surely detect an abnormality occurring in data stored using a data area in which a test pattern diagnosis is executed. To detect.

In one aspect of the present invention, the memory diagnostic apparatus includes a memory and a calculation unit for diagnosing a failure of the memory. The memory has three or more data areas for commonly storing the same predetermined data (RAM data). The calculation unit includes a memory management unit, a comparison processing unit, a diagnosis unit, and a data copy unit. The calculation unit specifies two first data areas in which the same data is currently stored from the memory management unit and a plurality of data areas, and test pattern data so as not to overlap with the two data areas. Specify the second data area to be the target of the test pattern diagnosis accompanied by the writing of. When the comparison processing unit requests the output of the same data to the memory, the comparison and collation processing of the stored data are executed between the two first data areas, and the abnormality detection is notified when the collation result is abnormal. The diagnosis unit executes the test pattern diagnosis on the second data area, and notifies the abnormality detection when the abnormality is detected by the test pattern diagnosis. The memory management unit updates the second data area in response to the normal completion notification of the test pattern diagnosis by the diagnosis unit, and the updated second data area is included in the two first data areas. If so, the updated second data area and the third data area that is neither of the current two first data areas are determined among the plurality of data areas. The data copy unit copies the data stored in the updated second data area to the third data area.

Another aspect of the present invention is a memory diagnostic method for a memory having three or more data areas for commonly storing the same predetermined data, wherein the same data can be obtained from the plurality of data areas. A step of designating two first data areas currently stored, and a test pattern diagnosis target that involves writing test pattern data from a plurality of data areas so as not to overlap with the two data areas. When the step of designating the second data area to be In response to the step of notifying the detection, the step of executing the test pattern diagnosis for the second data area, the step of notifying the abnormality detection when the abnormality is detected by the test pattern diagnosis, and the notification of the normal completion of the test pattern diagnosis, the first step. When the step of updating the second data area and the second data area after the update are included in the two first data areas, the second data area after the update among the plurality of data areas. The step of determining the data area and the third data area that is neither of the current two first data areas, and the data stored in the updated second data area are combined with the third data area. It has a step to copy to.

According to the present invention, a test pattern diagnosis for one data area among a plurality of data areas for storing the same predetermined data in common, and comparison and collation between the other two data areas are performed. Since the processing can be executed in parallel, the same data can be output from the memory together with the collation processing without interrupting the test pattern diagnosis, and an abnormality in which the same data is changed due to a hard error or a soft error can be detected. Can be detected. As a result, it is possible to detect an abnormality that has occurred in the data stored using the plurality of data areas early and reliably.

It is a functional block diagram explaining the structure of the memory diagnostic apparatus according to Embodiment 1 of this invention. It is a flowchart which shows the process by the RAM management part shown in FIG. It is a flowchart which shows the processing by the comparison processing part shown in FIG. It is a flowchart which shows the process by the diagnosis part shown in FIG. It is a flowchart which shows the process by the data copy part shown in FIG. FIG. 5 is a conceptual diagram illustrating an example of a diagnostic procedure for three data areas of the RAM shown in FIG. 1 by the memory diagnostic apparatus according to the first embodiment. It is a functional block diagram explaining the structure of the memory diagnostic apparatus according to Embodiment 2 of this invention. It is a functional block diagram explaining the structure of the memory diagnostic apparatus according to Embodiment 3 of this invention. It is a conceptual diagram explaining the division example of the data confirmation area and the data confirmation unnecessary area about the area inside each of a plurality of data areas. It is a flowchart which shows the process by the data copy part shown in FIG.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. In the following, the same or corresponding parts in the drawings are designated by the same reference numerals, and the explanations will not be repeated in principle.

Embodiment 1.
FIG. 1 is a functional block diagram illustrating a configuration of a memory diagnostic device according to a first embodiment of the present invention.

With reference to FIG. 1, the memory diagnostic apparatus 10 includes a memory 20 and a calculation unit 30. In the present embodiment, RAM is assumed as a typical example of a rewritable storage medium as the memory to be diagnosed. Therefore, in the following, the memory 20 is also referred to as a RAM 20, and the memory diagnostic device 10 is also referred to as a RAM diagnostic device 10.

The RAM diagnostic device 10 corresponds to a functional portion that executes self-diagnosis of RAM by the controller main body of an information processing device (such as the PLC described above) equipped with the RAM diagnostic device 10 including the RAM 20. The arithmetic unit 30 is composed of a CPU (Central Processing Unit) as hardware. The arithmetic unit 30 has a RAM management unit 40 and a comparison processing unit 50, which are functional parts for memory diagnostic processing, in addition to a function of executing processing of a main function of an information processing device in which the RAM diagnostic device 10 is incorporated. , A diagnostic unit 60, and a data copy unit 70. The comparison processing unit 50 includes a comparison processing execution unit 51 and a data abnormality detection unit 52. The diagnosis unit 60 includes a test pattern diagnosis execution unit 61 and a RAM failure detection unit 62. The data copy unit 70 includes an inter-data area copy unit 71 and a data copy confirmation unit 72.

Each functional block included in the arithmetic unit 30 shown in FIG. 1 is transmitted by the CPU to another memory such as a RAM 20 and a ROM (Read Only Memory) (not shown) via an address line and a data line (not shown). Corresponds to each function realized by reading and executing a program or reading and writing data with the access of. Alternatively, some or all of the functions of each functional block can be realized by hardware such as a dedicated electronic circuit.

The RAM 20 is a storage medium on which stored data can be rewritten. The RAM 20 has three or more storage areas (data areas) for storing the same specific data (hereinafter, also referred to as “RAM data”), which are predetermined. At each time point, the RAM data is commonly written in at least two of the plurality of storage areas. As an example in FIG. 1, three storage areas of a data area 21 (data area 1), a data area 22 (data area 2), and a data area 23 (data area 3) are provided in a RAM 20 of the same hardware. The configuration is described. The data areas 21 to 23 may be configured by using a plurality of RAMs of different hardware.

The RAM data assumes programs and parameters used in an information processing device equipped with a RAM diagnostic device 10 (RAM 20). For example, the RAM data includes data used for keeping the operation of the information processing apparatus safe, as typified by "safety-related data" in Patent Document 1.

When the RAM data output request is generated from the inside or the outside of the calculation unit 30, it is possible to output the RAM data from the RAM 20 after the comparison processing unit 50 compares and collates the data. .. Hereinafter, the RAM diagnosis by the comparison processing unit 50 is also referred to as “reading diagnosis”.

The memory diagnostic process described below is automatically started after the information processing device equipped with the RAM diagnostic device 10 is activated. The memory diagnosis process includes a read-time diagnosis performed when a RAM data output request is made in a RAM operation in which the information processing apparatus operates using the data of the RAM 20, and a RAM diagnosis using a test pattern (hereinafter, also referred to as "test pattern diagnosis"). ) And.

The RAM management unit 40 has a function of designating two areas to be compared and processed for data in the read-time diagnosis among a plurality of three or more data areas of the RAM 20 and notifying the comparison processing unit 50, and a test pattern diagnosis. It has a function of designating one area for performing the above and notifying the diagnosis unit 60, and a function of determining when copying is required between the data areas and notifying the data copy unit 70.

FIG. 2 is a flowchart showing processing by the RAM management unit 40.
With reference to FIG. 2, after the RAM diagnostic device 10 is started, the RAM management unit 40 uses steps 101 (hereinafter, simply referred to as “S”) 101 to read the initial values of the data areas 21 to 23. The RAM data is written to at least two of the above. Further, during the operation of the RAM diagnostic device 10, the loop processing from S102 (loop start) to S112 (loop end) is repeatedly executed.

In S103, the RAM management unit 40 designates two data areas to be compared in the read-time diagnosis by selecting two of the data areas 21 to 23 according to a preset order. As a result, two "first data areas" are designated. In the first S103 after the RAM diagnostic device 10 is started, the above order needs to be set so that the data area in which the specific data is written in the S101 is set as the comparison processing target area.

The RAM management unit 40 designates a test pattern diagnosis area from the data area not set in the comparison processing target area in S103 by S104. When three data areas 21 to 23 are provided as shown in FIG. 1, the remaining one data area that is not set as the comparison processing target area in S103 is automatically set as a test pattern diagnosis area. Is specified in. As a result, the "second data area" to be the target of the test pattern diagnosis is designated so as not to overlap with the "first data area (two)" specified in S103.

The RAM management unit 40 outputs a test pattern diagnosis start request for the data area set in the test pattern diagnosis area in S104 to the diagnosis unit 60 by S105. After outputting the test pattern diagnosis start request, the RAM management unit 40 waits until the diagnosis unit 60 notifies the normal completion of the test pattern diagnosis by S106.

When the diagnosis unit 60 notifies the RAM management unit 40 of the normal completion of the test pattern diagnosis (when the YES determination in S106 is made), the RAM management unit 40 next performs the test pattern diagnosis in the data area (hereinafter, hereinafter, according to the preset order) according to S107. Determine the "next test pattern diagnostic area"). As a result, the "second data area after update" is determined. Further, in S108, it is determined whether or not the data area (S103) currently set in the comparison target area and the next test pattern diagnosis area by S107 are the same. When three data areas 21 to 23 are provided as shown in FIG. 1, S108 is always determined to be YES.

The next test pattern diagnosis area cannot be used for the comparison process because the test pattern is overwritten during the test pattern diagnosis. Therefore, when the data area (S103) currently set in the comparison target area and the next test pattern diagnosis area overlap (when YES is determined in S108), the data in the next test pattern diagnosis area is determined by S109. The data area to copy to is selected. Specifically, a data area that is neither the following test pattern diagnosis area nor the data area currently set as the comparison target area is selected as the copy destination. When three data areas 21 to 23 are provided as shown in FIG. 1, in S109, the data area for which the test pattern diagnosis (S105) has been completed is inevitably selected as the copy destination. As a result, the "third data area" is determined.

Then, the RAM management unit 40 outputs a data copy request to the data copy unit 70 by S110 to copy the data of the next test pattern diagnosis area set in S107 to the data area selected in S109. ..

After sending the data copy request, the RAM management unit 40 waits until the data copy unit 70 notifies the normal completion of the data copy by S111. When the data copy unit 70 notifies the data copy that the data copy has been normally completed (YES in S111), the RAM management unit 40 proceeds to S112. As a result, the processing is returned to S102, and a new loop processing is started.

In a configuration in which the RAM 20 is provided with four or more data areas, S108 is determined to be NO because the data area (S103) currently set as the comparison target area and the next test pattern diagnosis area do not overlap. There will be cases where it is done. In this case, since the above data copy is not necessary, S109 to S111 are skipped and the process proceeds to S112, so that a new loop process is started.

In the new loop processing, two RAM data are stored from the remaining data areas (two or more) excluding the data area determined in the next test pattern diagnosis area in S107 of the previous loop processing by S103. Data area is set as the comparison processing target area. Further, in S104, the data area determined as the next test pattern diagnosis area in S107 of the previous loop processing is designated as the test pattern diagnosis area.

In this way, the RAM management unit 40 can target one of the three or more data areas for the test pattern diagnosis for each loop process. The information processing device equipped with the RAM diagnostic device 10 (RAM 20) can operate using the data in the data area in which the test pattern diagnosis is not executed. Further, the data of two data areas out of the data areas for which the test pattern diagnosis is not executed is set as the comparison processing target area.

As shown in FIG. 1, the comparison processing unit 50 has a data abnormality with the comparison processing execution unit 51 that performs comparison processing between two of the plurality of data areas of the RAM 20 when a discrepancy is detected as a result of the comparison processing. It has a data abnormality detection unit 52 that outputs a detection signal.

FIG. 3 is a flowchart showing processing by the comparison processing unit 50.
With reference to FIG. 3, the comparison processing unit 50 repeatedly executes the loop processing from S201 (loop start) to S208 (loop end) during the operation of the RAM diagnostic device 10 to perform the read-time diagnosis. To realize the function of.

The comparison processing unit 50 confirms with S202 whether or not there is a request for output of RAM data from the inside or the outside of the calculation unit 30. The output request is generated when it is necessary to read and use the RAM data in a device other than the RAM 20 for a general-purpose function (the RAM operation) other than the memory diagnosis process in the information processing device. The target of the output request may be a part of the entire RAM data stored in the data areas 21 to 23.

When the output of RAM data is requested (when YES is determined in S202), the two RAM management units 40 are currently set in the comparison target area by S103 or S109 (FIG. 2) by S203. In addition to confirming the data area, S204 performs comparison processing (comparison and collation confirmation) between the RAM data stored in each of the two data areas. At this time, the comparison processing target may be only a part of the RAM data for which the output request has been received, out of the entire RAM data stored in the entire data area. The process by S204 is executed by the comparison process execution unit 51.

The comparison processing unit 50 determines in S205 whether or not there is an abnormality in the RAM data based on the comparison processing result in S204. Specifically, when the data contents match between the RAM data of the two data areas set in the comparison target area, no abnormality is detected. If no abnormality is detected (NO determination in S205), S206 executes the output processing of the RAM data for which the output request has been received. Further, when the processing is advanced to S208, a new loop processing is started, so that a standby state for the next RAM data output request is formed.

On the other hand, when the data contents do not match between the RAM data of the two data areas set in the comparison target area, S205 is determined to be YES, and an abnormality in the RAM data is detected. In this case, S206 is not executed, and S207 outputs a data abnormality signal from the data abnormality detection unit 52.

As shown in FIG. 1, the diagnosis unit 60 includes a test pattern diagnosis execution unit 61 for executing a test pattern diagnosis for one data area of a plurality of data areas of the RAM 20, and a test pattern diagnosis. It has a RAM failure detection unit 62 that outputs a RAM failure signal when an abnormality is detected.

FIG. 4 is a flowchart showing processing by the diagnosis unit 60.
With reference to FIG. 4, the diagnosis unit 60 determines by S301 whether or not there is a test pattern diagnosis start request (S105) from the RAM management unit 40. When there is a test pattern diagnosis start request (when YES is determined in S301), the diagnosis unit 60 confirms the data area currently determined as the test pattern diagnosis area by the RAM management unit 40 by S302. Then, according to S303, RAM diagnosis using the test pattern is executed for the test pattern diagnosis area. The processing by S301 to S303 can be executed by the test pattern diagnosis execution unit 61.

The diagnosis unit 60 is convinced of the result of the test pattern diagnosis by S304, and when a RAM abnormality is not detected (at the time of NO determination in S304), the diagnosis unit 60 notifies the RAM management unit 40 of the normal completion of the test pattern diagnosis by S305. Output. The test pattern diagnosis completion notification is used for the determination in S106 of FIG.

On the other hand, when a RAM abnormality is detected by the test pattern diagnosis (when YES is determined in S304), the RAM failure detection unit 62 outputs a RAM failure signal by S306 without executing S305.

In this way, every time the RAM management unit 40 generates a test pattern diagnosis start request by S105 of FIG. 2, the processing of FIG. 4 is executed by the diagnosis unit 60. As the test pattern data used in the test pattern diagnosis, known ones can be arbitrarily applied, but it is preferable that the test pattern data are two types of pattern data in which the constituent bits are inverted with each other. By writing each of the two types of pattern data to the diagnostic area, reading it again, and collating the written data with the read data, it is possible to detect a hard error in the data area targeted for test pattern diagnosis. It is possible. For example, in IEC 61508, as a RAM test method, galpat (or transparent galpat), Abraham, and the like are recommended as techniques for obtaining a high diagnostic rate. By using these techniques for the test pattern diagnosis in S303, the detection of hard errors can be made more reliable.

As shown in FIG. 1, the data copy unit 70 includes a data area copy unit 71 that copies data between data areas and a data copy confirmation unit 72 that confirms that the data has been copied normally after the data copy. Have.

FIG. 5 is a flowchart showing processing by the data copy unit 70.
With reference to FIG. 5, the data copy unit 70 determines whether or not there is a data copy request (S110) from the RAM management unit 40 by S401. When there is a data copy request (when YES is determined in S401), the data copy unit 70 sets the data area as the data copy source and the data area as the copy destination according to the contents of the data copy request according to S402. decide. As described above, in S110 of FIG. 2, the next test pattern diagnosis area set in S107 is designated as the copy source data area, while the data area selected in S109 is the copy destination data area. It is specified in.

The data copy unit 70 copies the stored data of the copy source data area determined in S402 by S403 to the copy destination data area. The processing by S401 to S403 is executed by the inter-data area copy unit 71.

When the data copy is completed, S404 outputs a data copy confirmation request from the inter-data area copy unit 71 to the data copy confirmation unit 72. In S405, the data in the copy destination data area is compared and collated not with the copy source data area but with another data area in which the same data as the copy source data area is stored. .. This not only confirms that the data has been copied normally, but also detects the occurrence of the soft error when a soft error occurs in either the copy source data area or the other data area. It becomes possible.

The data copy confirmation unit 72 executes a match confirmation between the data in the copy destination data area and the data in the other data area according to S405 in accordance with the data copy confirmation request (S404). Further, the data copy confirmation unit 72 determines by S406 whether or not there is an abnormality in the RAM data based on the match confirmation result in S405. Specifically, when the data in the copy destination data area and the data in the other data area match, it is determined that there is no abnormality. When it is determined that there is no abnormality (NO determination in S406), a copy confirmation notification is output from the data copy confirmation unit 72 to the inter-data area copy unit 71.

When the inter-data area copy unit 71 confirms the copy confirmation notification (when YES is determined in S407), the data copy normal completion notification is output to the RAM management unit 40 by S408. Correspondingly, S111 in FIG. 2 is determined to be YES.

On the other hand, when the data in the copy destination data area and the data in the other data area do not match, it is determined that there is an abnormality. When it is determined that there is an abnormality (YES determination in S406), S407 and S408 are not executed, and the data copy confirmation unit 72 outputs a data abnormality signal by S409. That is, the occurrence of a soft error is detected by outputting a data abnormality signal.

The RAM management unit 40 repeatedly executes the process shown in the flowchart of FIG. 2, and the comparison processing unit 50, the diagnosis unit 60, and the data copy unit 70 are shown according to various selections of the data area by the RAM management unit 40. By repeating the processes shown in the flowcharts of FIGS. 3 to 5, the RAM diagnostic apparatus 10 can realize the read-time diagnosis and the periodic test pattern diagnosis for the data areas 21 to 23 of the RAM 20. ..

FIG. 6 shows an example of a diagnostic procedure for the data areas 21 to 23 (FIG. 1) of the RAM 20 by the RAM diagnostic apparatus according to the first embodiment.

With reference to FIG. 6, in the state 501, the RAM data is written to the data areas 21 and 22 by reading the initial value by S101 of FIG. Further, according to S103 of FIG. 2, the data areas 21 and 22 are set as the data area to be compared, and according to S104 of FIG. 2, the data area 23 is set as the test pattern diagnosis area.

In process 1, the diagnosis unit 60 executes the test pattern diagnosis for the data area 23 in response to the output of the test pattern diagnosis start request in S105 of FIG. When a RAM data output request is generated during process 1, that is, during test pattern diagnosis for the data area 23, the comparison processing unit 50 performs a comparison process between the data areas 21 and 22 (FIG. 3). After it is confirmed by the execution of S204) that there is no data abnormality, the RAM data can be output from the RAM 20 by S206 of FIG.

In the state 502 in which the process 1 is completed, the test pattern diagnosis for the data area 23 in the process 1 is completed, so that the test pattern data is overwritten in the data area 23. Upon completion of the test pattern diagnosis, S106 in FIG. 2 is determined to be YES, and the data area 21 is determined to be the next test pattern diagnosis area (S107 in FIG. 2). As a result, since the comparison target area (data areas 21 and 22) at the present stage is the same as the next test pattern diagnosis area (data area 21), S108 in FIG. 2 is determined to be YES, and the test pattern data. S108 and S109 of FIG. 2 are executed to copy the data in the data area 21 that is to be overwritten. Specifically, according to S109, the data area 23, which is neither the next test pattern diagnosis area (data area 21) nor the data area currently set as the comparison target area (data areas 21 and 22), is copied. Selected first.

In process 2, in response to the output of the data copy request in S110 of FIG. 2, the data copy unit 70 transfers data from the data area 21 which is the next test pattern diagnosis area to the data area 23 selected as the copy destination. The copy is executed. After copying the data, if the data collation is normally completed between the copy destination data area 23 and the other data area 22 in which the same data as the copy source data area 21 is stored, S111 in FIG. 2 is YES. As a determination, the following loop processing is executed in FIG.

In the state 503 in which the process 2 is completed, the data areas 22 and 23 are set as the data areas to be compared and processed for diagnosis at the time of reading by S103 of FIG. 2, and the data area 21 is tested by S104 of FIG. It is set in the pattern diagnosis area.

In the process 3, the diagnosis unit 60 executes the test pattern diagnosis for the data area 21 in response to the output of the test pattern diagnosis start request in S105 of FIG. When a RAM data output request is generated during the process 3, that is, during the test pattern diagnosis for the data area 21, the comparison process 50 performs a comparison process between the data areas 22 and 23 (FIG. 3). RAM data can be output from the RAM 20 with S204) (S206 in FIG. 3).

In the state 504 in which the process 3 is completed, the test pattern diagnosis for the data area 21 in the process 3 is completed, so that the test pattern data is overwritten in the data area 21. When the data area 22 is determined to be the next test pattern diagnosis area according to S107 of FIG. 2, the comparison target area (data areas 22 and 23) at the present stage becomes the next test pattern diagnosis area (data area 22). Since they are the same, S108 in FIG. 2 is determined to be YES. Then, the test pattern data is overwritten in the data area 21 which is neither the next test pattern diagnosis area (data area 22) nor the data area (data areas 22 and 23) currently set as the comparison target area. It is selected as the copy destination of the data from the data area 22 to be scheduled (S109 in FIG. 2).

In process 4, in response to the output of the data copy request in S110 of FIG. 2, the data copy unit 70 transfers data from the data area 22 which is the next test pattern diagnosis area to the data area 21 selected as the copy destination. The copy is executed. After copying the data, if the data collation is normally completed between the copy destination data area 21 and the other data area 23 in which the same data as the copy source data area 22 is stored, S111 in FIG. 2 is YES. As a determination, the next loop processing is further executed in FIG.

In the state 505 in which the process 4 is completed, the data areas 21 and 23 are set as the data areas to be compared by S103 in FIG. 2, and the data areas 22 are set as the test pattern diagnosis area by S104 in FIG. NS.

In process 5, the diagnosis unit 60 executes a test pattern diagnosis for the data area 22. When a RAM data output request is generated during process 5, that is, during test pattern diagnosis for the data area 22, comparison processing between the data areas 21 and 23 by the comparison processing unit 50 (S204 in FIG. 3). RAM data can be output from the RAM 20 (S206 in FIG. 3).

In the state 506 in which the process 5 is completed, the test pattern diagnosis for the data area 22 in the process 5 is completed, so that the test pattern data is overwritten in the data area 22. When the data area 23 is determined to be the next test pattern diagnosis area according to S107 of FIG. 2, the comparison target area (data areas 21 and 23) at the present stage becomes the next test pattern diagnosis area (data area 23). Since they are the same, S108 in FIG. 2 is determined to be YES. Then, the test pattern data is overwritten in the data area 22, which is neither the next test pattern diagnosis area (data area 23) nor the data area (data areas 21 and 22) currently set as the comparison target area. It is selected as the copy destination of the data in the data area 23 (S109 in FIG. 2).

In process 6, the data copy unit 70 executes data copy from the data area 23 (next test pattern diagnosis area) to the data area 22 selected as the copy destination. After copying the data, if the data collation is normally completed between the copy destination data area 22 and the other data area 21 in which the same data as the copy source data area 23 is stored, S111 in FIG. 2 is YES. As a determination, the next loop processing is further executed in FIG. After the end of the process 6, RAM data is written in the data areas 21 and 22, and the data area 23 is scheduled for the next test pattern diagnosis area, and the RAM 20 is in the same state as the state 501. Become.

In the memory diagnostic apparatus according to the first embodiment, the test pattern diagnosis can be executed for each of the data areas 21 to 23 in order by repeating the processes 1 to 6 thereafter. At the same time, in response to the RAM data output request, normal RAM data can be output after confirming the presence or absence of data abnormality by comparison processing (diagnosis at the time of reading). In particular, as can be understood from FIG. 6, the output of RAM data accompanied by the read-time diagnosis by comparison and collation between the two data areas is output except when the data is copied in the process 2, the process 4, and the process 6. It is possible to do it. That is, even if a RAM data output request is generated during the test pattern diagnosis in the process 1, the process 3, and the process 5, the RAM data can be output from the RAM 20 without interrupting the test pattern diagnosis.

As a result, the time required for the test pattern diagnosis for each of the plurality of data areas for storing the RAM data can be shortened, and the cycle for the test pattern diagnosis in each data area can be shortened. It is possible to detect the occurrence of a hard error in the data area of the above at an early stage. Further, by using the test pattern diagnosis and the read-time diagnosis together, it is possible to reliably detect an abnormality in which the RAM data stored in the plurality of data areas is changed due to a hard error or a soft error.

In the comparison process, the test pattern diagnosis process, and the copy process between the processes 1 to 6, whether there is an abnormality in the RAM data according to S205 in FIG. 3, S304 in FIG. 4, and S406 in FIG. Can be determined. Therefore, when a RAM failure is detected, the data abnormality signal in S207 of FIG. 3, the RAM failure signal of S306 of FIG. 4, and the data abnormality signal of S409 of FIG. 5 can be output. By stopping the use of RAM data when signals indicating these abnormalities are output, it is possible to prevent the information processing apparatus from malfunctioning due to the use of abnormal RAM data. For example, in response to the signal indicating the above abnormality, the operation of the information processing device can be stopped, an alert can be sent to the user, or the output of RAM data can be stopped by processing in the comparison processing unit 50 of the RAM diagnostic device 10. ..

Further, in the memory diagnostic apparatus according to the first embodiment, by performing the test pattern diagnosis accompanied by the above-mentioned data copy, even if the RAM data is changed due to a hard error or a soft error in a data area that is not accessed for a long period of time. , Relatively early detection is possible.

In general test pattern diagnosis, the data in the data area to be diagnosed is copied to another storage medium (copy destination) such as RAM, and the diagnosis is completed after failure diagnosis involving writing and reading of test pattern data. It is necessary to rewrite the data from the copy destination to the data area. On the other hand, in the memory diagnostic apparatus according to the first embodiment, after the data is copied in the process 2, the process 4, and the process 6 of FIG. 6, the data saved in the copy destination data area is the RAM data. Will be used for output processing of. As a result, it is not necessary to write back the data to the data area where the test pattern diagnosis is completed. Therefore, by reducing the number of times of copying the data in the test pattern diagnosis, the time required for the diagnosis for each data area can be shortened. As a result, it is possible to shorten the execution cycle of the test pattern diagnosis of each data area, so that even if a hard error occurs in a memory cell that is not accessed for a long period of time, the failure can be detected at a relatively high speed. You can expect it.

Further, when copying data, the data in the copy destination data area is not with the copy source data area, but with another data area in which the same data as the copy source data area is stored. , Therefore, even if a soft error occurs in a memory cell that is not accessed for a long period of time, it can be expected to detect the failure at a relatively high speed.

It is also possible to obtain a higher diagnostic rate by applying the double RAM technique described in Patent Document 1 to the memory diagnostic apparatus according to the first embodiment. In this case, in the initial value reading in S101 of FIG. 2, RAM data and its inverted data (hereinafter, inverted RAM data) are written, and the comparison processing unit 50 (comparison processing execution unit 51) is shown in FIG. In S204, after inverting the inverted RAM data stored in one data area of the two data areas to be compared, data comparison processing (comparison and collation confirmation) is performed with the RAM data stored in the other data area. ) Is executed. Similarly, in S405 of FIG. 5, the data copy unit 70 (data copy confirmation unit 72) inverts the data in one data area and then two data areas (the copy destination data area and the copy source data area). Data comparison processing is executed between the data area and other data areas where the same data is stored.

Embodiment 2.
FIG. 7 is a functional block diagram illustrating the configuration of the memory diagnostic apparatus according to the second embodiment of the present invention.

With reference to FIG. 7, the RAM diagnostic apparatus 610 according to the second embodiment includes a RAM 20, a main calculation unit 630, and a sub calculation unit 640. The main calculation unit 630 includes a comparison processing unit 50 having a comparison processing execution unit 51 and a data abnormality detection unit 52, as in FIG. 1. The sub-calculation unit 640 includes a RAM management unit 40, a diagnosis unit 60 having a test pattern diagnosis execution unit 61 and a RAM failure detection unit 62, a data area copy unit 71, and a data copy confirmation unit 72, as in FIG. It is provided with a data copy unit 70 having the data copy unit 70. Since the configuration of the RAM 20 and the functions and processes of the RAM management unit 40, the comparison processing unit 50, the diagnosis unit 60, and the data copy unit 70 are the same as those in the first embodiment, the description will not be repeated. That is, the RAM management unit 40, the comparison processing unit 50, the diagnosis unit 60, and the data copy unit 70 execute the processes shown in FIGS. Diagnosis can be performed.

In the RAM diagnostic device 610 according to the second embodiment, a sub-calculation unit for memory diagnosis is performed separately from the main calculation unit 630 that executes the main function of the information processing device on which the RAM diagnostic device 610 including the RAM 20 is mounted. 640 is provided. For example, the sub-calculation unit 640 can be configured by using a core or the like different from the CPU constituting the main calculation unit 630 in a microcomputer or a multi-core CPU. The main arithmetic unit 630 corresponds to an embodiment of the "first arithmetic unit", and the sub arithmetic unit 640 corresponds to an embodiment of the "second arithmetic unit".

According to the configuration of the second embodiment, the main calculation unit 630 is provided with the function of the comparison processing unit 50 in response to the RAM data output request, while the main processing for executing the test pattern diagnosis and data copying is performed. It can be executed in parallel with the arithmetic unit 630. As described above, when the RAM data output process and the test pattern diagnosis process are simultaneously executed except for the data copy period in the process 2, process 4, and process 6 of FIG. 6, the main calculation unit 630 and the sub Parallel processing by the arithmetic unit 640 becomes possible. As a result, the test pattern diagnosis can be further speeded up as compared with the memory diagnostic apparatus of the first embodiment, so that the execution cycle of the test pattern diagnosis of each data area can be shortened.

In the first and second embodiments, three data areas for storing RAM data are provided, but a plurality of four or more data areas can be provided. In this case as well, the test pattern diagnosis can be executed for each data area in order, and in response to the RAM data output request, between the two sequentially selected data areas. Normal RAM data can be output with comparison processing. As described above, when four or more data areas are provided, the next time the data area to be the target of the test pattern diagnosis is determined (S107 in FIG. 2), S108 in FIG. 2 is determined as NO. As a result, there will be cases where data copying becomes unnecessary.

Embodiment 3.
FIG. 8 is a functional block diagram illustrating the configuration of the memory diagnostic apparatus according to the third embodiment of the present invention.

With reference to FIG. 8, the RAM diagnostic apparatus 710 according to the third embodiment includes a RAM 20 and a calculation unit 730. The calculation unit 730 includes a RAM management unit 40, a comparison processing unit 50, a diagnosis unit 60, and a data copy unit 770, which are the same as those in FIG. Since the configuration of the RAM 20 and the functions and processes of the RAM management unit 40, the comparison processing unit 50, and the diagnosis unit 60 are the same as those in the first embodiment, the description will not be repeated.

The data copy unit 770 further has data copy confirmation unnecessary area information 73 in addition to the inter-data area copy unit 71 and the data copy confirmation unit 72 similar to the data copy unit 70 in FIG.

The comparison processing unit 50 executes the same read-time diagnosis as in the first embodiment according to the flowchart of FIG. The RAM management unit 40 manages the test pattern diagnosis and the data copy associated with the test pattern diagnosis according to the flowchart of FIG. 2, as in the first embodiment. Similarly, the diagnosis unit 60 executes the test pattern diagnosis according to the flowchart of FIG. Similar to the data copy unit 70 of FIG. 1, the data copy unit 770 executes data copy in response to a data copy request (S110 of FIG. 1) from the RAM management unit 40 according to the flowchart of FIG. 9 described later. Therefore, in the third embodiment, only the data copy is different from the first embodiment, but the other processing of the RAM diagnosis is the same as that of the first embodiment.

In the RAM diagnosis described in the first embodiment, when the RAM management unit 40 outputs a data copy request (S110 in FIG. 2), the RAM 20 is subjected to until the data copy completion notification is received (NO determination in S111). Reading and writing of RAM data is prohibited. This is because the RAM data cannot be changed during the data copy execution, and the reliability of the RAM data before the completion of the data copy confirmation is low because it is unconfirmed that no soft error has occurred.

Therefore, as the time required for the data copy processing (S402 to S408) according to the flowchart shown in FIG. 5 becomes longer, the execution frequency of the main original function of the information processing device equipped with the RAM diagnostic device increases accordingly. There is concern that it will decline. Therefore, in the third embodiment, the time required for the data copy processing is shortened by selectively executing the data copy confirmation (S405) process after the data copy.

The data copy confirmation unnecessary area information 73 included in the data copy unit 770 of FIG. 8 is information for specifying an area for which data copy confirmation is unnecessary for each of the plurality of data areas 21 to 23 for storing RAM data. For example, as described above, when a part of the RAM data is subject to an output request as needed, the internal areas of the plurality of data areas 21 to 23 are weighted according to the frequency of access. , Only the area with low access frequency is targeted for data copy confirmation, while the data copy confirmation can be omitted for the data area with high access frequency. That is, the data copy confirmation unnecessary area information 73 stores a part of the RAM data that is frequently targeted for output request, and specifies an area that does not require data copy confirmation.

FIG. 9 is a conceptual diagram illustrating a division example of a data confirmation area and a data confirmation unnecessary area for an area inside each data area.

Since the same RAM data is written in the data areas 1 to 3 with reference to FIG. 9, the division of the data confirmation area and the data confirmation unnecessary area is the same in each data area. In the example of FIG. 10, each data area is divided into a data confirmation area 901 and 903 and a data confirmation unnecessary area 902 and 904.

In the data area where the access frequency is sufficiently high compared to the frequency of data copy confirmation, the comparison processing associated with the data access for outputting RAM data is performed frequently, so the occurrence of soft errors is compared. It can be found by the reading diagnosis by the processing unit 50. Therefore, even if the data copy confirmation is omitted, it is possible to find a soft error in the area. Therefore, it can be determined that the data copy confirmation is unnecessary in the data confirmation unnecessary areas 902 and 904 because the RAM data stored in the area is frequently accessed and the like. The data copy confirmation unnecessary area information 73 identifies which of the data confirmation areas 901 and 903 and the data confirmation unnecessary areas 902 and 904 corresponds to each of the internal areas of the plurality of data areas 21 to 23. can do.

FIG. 10 is a flowchart showing processing by the data copy unit 770 shown in FIG.

With reference to FIG. 10, the data copy unit 770 executes the processes of S450 and S455 in addition to the processes of S401 to S404 and S405 to S409 similar to those of FIG. Similar to FIG. 5, S401 to S404 and S407 to S409 are executed by the inter-data area copy unit 71. On the other hand, S450, S455, and S406 are executed by the data copy confirmation unit 72.

In S450, after the data copy confirmation request by S404, the data copy confirmation unit 72 has a high access frequency and data copy confirmation is performed for the internal area of the data copy source data area, which is indicated by the data copy confirmation unnecessary area information 73. The necessity of data copy confirmation is determined according to whether or not the area corresponds to an unnecessary area.

In S455, the data copy confirmation unit 72 determines whether or not data copy confirmation is necessary based on the determination result in S450, and when it is an area where data copy confirmation is necessary (when YES determination in S455), the data copy confirmation unit 72 determines whether or not data copy confirmation is necessary. The process proceeds to S405 similar to FIG. 5, and the data between the copy destination data area and the copy source data area is between other data areas in which the same data is stored in the internal area. Data copy confirmation is performed by comparison and collation.

On the other hand, when the data copy confirmation unit 72 is an area where data copy confirmation is unnecessary (when NO determination in S455), S405 and S406 are skipped, and when NO determination in S406 (that is, there is no RAM data abnormality). ), The process proceeds to S407. The data copy confirmation unit 72 waits for the output of the copy confirmation notification with S407 as a NO determination until the processes S450 to S406 are executed for all the areas inside the data area of the data copy source. As a result, when the data copy confirmation (S405) is executed only in the data confirmation areas 901 and 903 of the data copy source data areas and the RAM data in the data confirmation area is determined to be "no abnormality", S407 is executed. A YES determination is made, and a copy confirmation notification is output from the data copy confirmation unit 72 to the inter-data area copy unit 71.

As described above, in the RAM diagnostic apparatus according to the third embodiment, in the data copy processing during the test pattern diagnosis, the data copy confirmation (S405) process is performed only for a part of the internal area of each data area, in other words, a part. The time required for the data copy process can be shortened by selectively executing only the RAM data of the above. As a result, in the information processing device equipped with the RAM diagnosis device, it is possible to avoid that the execution time of the original function is limited by the RAM diagnosis.

The content of the data copy confirmation unnecessary area information 73 may be fixed or may be changed according to the operation of the information processing device equipped with the RAM diagnostic device 710. For example, when the data copy confirmation unnecessary area information 73 is fixed, the developer of the information processing apparatus using the RAM data can determine the access frequency and set the data copy confirmation unnecessary area information 73 in advance.

Alternatively, the data copy confirmation unnecessary area information 73 can be set based on the access history to each data area when the information processing apparatus is operating. For example, a method in which AI (Artificial Intelligence) is mounted on the data copy confirmation unit 72 to determine an area where access frequency is high and data copy confirmation is unnecessary is also effective. In this case, the access record of the internal area of each data area is monitored by the above AI, and some areas determined to have high access frequency are periodically set and updated as data copy confirmation unnecessary area information 73. can do.

In the information processing device using the RAM diagnostic device 710, the area with high access frequency may change depending on the usage method and timing of the RAM data area. Even in such a case, the data copy confirmation unnecessary area information 73 By optimizing the settings, it is possible to optimize the ratio between the soft error detection rate and the main processing operation rate of the information processing device.

It is also possible to combine the second embodiment with the RAM diagnostic device 710 according to the third embodiment. That is, in the RAM diagnostic apparatus 710 of FIG. 8, the main calculation unit 630 and the sub calculation unit 640 that can operate in parallel are arranged as in FIG. 7, while the main calculation unit 630 has the function of the comparison processing unit 50. The functions of the RAM management unit 40, the diagnosis unit 60, and the data copy unit 770 can be provided to the sub-calculation unit 640. Also in the third embodiment, any plurality of three or more data areas for storing RAM data can be provided.

It should be considered that the embodiments disclosed this time are exemplary in all respects and not restrictive. The scope of the present invention is shown by the claims rather than the above description, and it is intended to include all modifications within the meaning and scope equivalent to the claims.

10,610 Memory diagnostic device (RAM diagnostic device), 20 memory (RAM), 21-23 data area, 30 arithmetic unit, 40 RAM management unit, 50 comparison processing unit, 51 comparison processing execution unit, 52 data abnormality detection unit, 60 Diagnosis unit, 61 Test pattern diagnosis execution unit, 62 RAM failure detection unit, 70 data copy unit, 71 data area copy unit, 72 data copy confirmation unit, 501-506 status, 630 main calculation unit, 640 sub calculation unit, 901,903 data confirmation area, 902,904 data confirmation unnecessary area.

Claims (10)

A memory having three or more data areas for storing the same predetermined data in common, and a memory
It is provided with an arithmetic unit for diagnosing the memory failure.
The calculation unit
From the plurality of data areas, two first data areas in which the same data is currently stored are designated, and test pattern data is written so as not to overlap with the two first data areas. a memory management unit for designating the second data area as a target of the test pattern diagnosis involving,
At the time of requesting the output of the same data to the memory, the comparison processing unit that executes the comparison and collation processing of the stored data between the two first data areas and notifies the abnormality detection when the collation result is abnormal, and the comparison processing unit.
It includes a diagnostic unit that executes the test pattern diagnosis on the second data area and notifies the abnormality detection when an abnormality is detected by the test pattern diagnosis.
The memory management unit
The second data area is updated in response to the normal completion notification of the test pattern diagnosis by the diagnosis unit, and the updated second data area is included in the two first data areas. If so, a third data area that is neither the updated second data area nor the current two first data areas among the plurality of data areas is determined.
The calculation unit
A memory diagnostic apparatus further including a data copy unit that copies data stored in the updated second data area to the third data area. When the memory diagnostic apparatus is started, the memory management unit writes the same data to two or more data areas among the plurality of data areas, and the two or more data areas are said to be the same data. The memory diagnostic apparatus according to claim 1, wherein the first data area of the above is determined. The memory management unit updates the two first data areas and the second data area so as to shift them in a predetermined order in response to the notification of normal completion of the test pattern diagnosis. Item 3. The memory diagnostic apparatus according to item 1 or 2. After copying the data to the third data area, the data copy unit is different from the third data area and the updated second data area among the plurality of data areas, and The item according to any one of claims 1 to 3, wherein the stored data is compared and collated with the data area storing the same data, and the abnormality detection is notified when the collation result is abnormal. Memory diagnostic device. When the number of the plurality of data areas is 3, the memory management unit responds to the normal completion notification of the test pattern diagnosis by the diagnosis unit, and the memory management unit is one of the two first data areas. The data area of is determined to be the second data area after the update, and the data area for which the test pattern diagnosis has been completed is determined to be the third data area.
After copying the data to the third data area, the data copy unit performs the comparison and comparison between the third data area and the other data area of the two first data areas. The memory diagnostic device according to claim 4, which executes a collation process. After copying the data to the third data area, the data copy unit is a part of the inner area of the second data area in which the stored data copied is stored, which is frequently accessed. The memory diagnostic apparatus according to claim 4 or 5, wherein the comparison and collation processing for the area of the above is not executed. The first calculation unit is composed of a first calculation unit including the comparison processing unit, a memory management unit, a diagnosis unit, and a second calculation unit having the data copy unit. The memory diagnostic apparatus according to any one of 6. A memory diagnostic method for a memory having three or more data areas for storing the same predetermined data in common.
From the plurality of data areas, a step of designating two first data areas in which the same data is currently stored, and a step of designating the two first data areas.
A step of designating a second data area to be a target of test pattern diagnosis accompanied by writing test pattern data so as not to overlap with the two first data areas from the plurality of data areas.
At the time of requesting the output of the same data to the memory, the step of executing the comparison and collation processing of the stored data between the two first data areas and notifying the abnormality detection when the collation result is abnormal, and the step.
A step of executing the test pattern diagnosis on the second data area and notifying the abnormality detection when an abnormality is detected by the test pattern diagnosis, and a step of notifying the abnormality detection.
In response to the notification of normal completion of the test pattern diagnosis, the step of updating the second data area and
When the updated second data area is included in the two first data areas, the second data area after the update and the current data area among the plurality of data areas are present. A step of determining a third data area that is neither of the two first data areas,
A memory diagnostic method comprising a step of copying data stored in the updated second data area to the third data area. After copying the data to the third data area, the third data area is different from the updated second data area of the plurality of data areas, and the same data is stored. The memory diagnostic method according to claim 8, further comprising a step of performing comparison and collation processing of stored data with the data area. After copying the data to the third data area, the stored data for a part of the area inside the second data area in which the stored data copied is stored, which is frequently accessed. The memory diagnostic method according to claim 9, further comprising a step of non-execution of the comparison and collation processing.

Images