JPH03127241A - Memory control method for paging virtual storage system - Google Patents

Abstract

PURPOSE:To carry on the working of a computer system despite the presence of a faulty memory by reloading the data on a faulty page into an idle page for application when this data is not updated yet, that is, when the same data remain in a page file. CONSTITUTION:A faulty page list 7 is provided and the numbers of faulty pages are registered into the list 7. The use of the registered pages are inhibited. When the data on a faulty page is not updated yet, this data is reloaded into an idle page from a page file of a disk 9. At the same time, an address conversion table 12 and an idle page list 6 are rewritten in terms of the idle page. Thus the working of a computer system can be carried on despite a memory trouble.

Description

[Detailed description of the invention] [Industrial application field]

The present invention relates to a memory management system in a paging virtual memory system that allows a processor device to continue operating normally using the remaining memory even if a part of the main memory fails.

[Conventional technology]

In the paging virtual memory system, a virtual address space made up of virtual addresses set by the CPU (Central Processing Unit) and a real address space made up of real main memory addresses are divided into uniform parts called pages. Memory is managed by dividing it into different sizes. Therefore, one page is a collection of several addresses. In a computer system employing such a paging virtual storage method, if an abnormality occurs in a page memory storing data necessary for processing performed by a processor device, processing cannot be continued. Conventionally, countermeasures have been taken to address this problem, such as adding a memory correction circuit or running a fault diagnosis program when starting up the computer system. The memory correction circuit is a circuit that corrects any errors in the contents of the memory. The fault diagnosis program writes predetermined data into the main memory when the computer system is started up, and then reads the data to diagnose the presence or absence of a fault. If the data that was supposed to be written is not read out, the memory is diagnosed as faulty.

[Question H that the invention attempts to solve]

(Problem) However, if the memory is faulty, the memory correction circuit can no longer deal with the problem, and the fault diagnosis program only discovers the fault, so the computer system cannot continue operating. was there. (Explanation of the problem) First, there is a memory correction circuit, which corrects erroneous data and writes correct data, so if the memory to be written to is malfunctioning, writing cannot be done. The memory correction circuit cannot deal with faulty memories. Next is the fault diagnosis program.If you run this program when starting up the computer system, it will naturally detect faulty memory if it exists.If it finds faulty memory, it will display a message indicating that there is a fault and stop the computer system from operating. This does not mean that measures will be taken to overcome the failure and continue operation. Operation can only begin after repairs have been made and the failure diagnosis program has been run again to confirm that there are no abnormalities. Therefore, it was not possible to continue operation with the faulty memory. An object of the present invention is to solve the above-mentioned problems.

[Means to solve the problem]

In order to solve the above problem, in the memory management method in the paging virtual memory system of the present invention, a fault page list is provided in which the page number of the faulty page is registered, and the fault page is prevented from being used from now on, and the data of the fault page is If the data has not been updated, the data is reloaded from the page file into the empty page, and the address conversion table and empty page list are rewritten with respect to the empty page.

[For production]

With the memory management method of the paging virtual memory method as described above, if the data on the failed page has not been updated, that is, if the same data remains in the page file, the data is reloaded into an empty page. It is now possible to use the Therefore, processes using the data can continue even though the data was initially stored in the failed page. On the other hand, since the number of the faulty page is registered in the faulty page list and its use is prohibited, the operation of the computer system will no longer be hindered by the faulty page.

【Example】

Embodiments of the present invention will be described in detail below with reference to the drawings. FIG. 1 shows a computer system to which the memory management method in the paging virtual memory method of the present invention is applied.
In FIG. 1, 1 is a processor device, 2 is a CPU (
3 is MMU (memory management unit), 4 is system bus, 5 is main memory, 6 is
is an empty page list, 7 is a failed page list, 8 is a disk controller, and 9 is a disk. In the Combiator system that employs the paging virtual memory method, the address output from the CPU 2 is a virtual address. The MMU 3 converts this virtual address into a real address, which is an address in the main memory 5, and stores it in the main memory 5.
access. The main memory 5 is composed of a memory IC and the like, and part of the data stored on the disk 9 is copied thereto as necessary. Conversely, the contents of main memory 5 are returned to disk 9 as needed. A predetermined number of memory areas on the disk 9 are also grouped together and named a page. Data is stored in the form of a file (page file) in units of pages. Page numbers are added to distinguish between pages. Data exchange between the disk 9 and the main memory 5 is as follows:
This is not done in addresses 1-lower, but in page units. The disk controller 8 controls data input/output to/from the disk 9 . The main memory 5 is provided with an empty page list 6 and a failed page restore. Although the empty page list 6 has existed in the past, failure page restoration is provided for the first time in the present invention. The roles of the empty page list 6 and failed page restoration will be explained with reference to FIG. 2 below. FIG. 2 is a diagram showing the status of access from the virtual address space to the real address space. In Figure 2, 9 is a disk, 10 is a virtual address space, 11 is a page number of the virtual address space, 12 is an address conversion table, 13 is an update display bit, 14 is a real address space, and 15 is a page number of the real address space. It is. The virtual address space 10 is an address space configured by virtual addresses set by the CPU 2, and the page number 11 is a number assigned to each page of the virtual address space IO. The address issued by the CPU 2 is a virtual address belonging to the virtual address space 10. The real address space 14 is an address space made up of real addresses in the main memory 5, and is also divided into pages, each of which is assigned a page number 15. Among the pages in the real address space 14, pages 0 and 2.3, which are filled with dots, represent empty pages. An empty page is a page in which no data is stored. Furthermore, page 5 marked with a rose mark is a failure page. A failed page is a page that contains failed memory. The address conversion table 12 is a table for converting a virtual address into a real address, and its management and conversion processing is performed by the MMU3.
done by. Note that an update display bit 13 provided in association with the address conversion table 12 indicates whether or not the data of each page currently in the real address space 14 has been updated. If the value of this bit is "0", it means that it has not been updated (that is, the value remains the same as when it was copied from disk 9), and if it is "l", it means that it has been updated (that is, the value that was copied from disk 9 remains the same). 9). FIG. 3 is a diagram showing an empty page list and a failed page list in the case of FIG. 2. Since the empty page in the real address space 14 in Figure 2 is page 0.2.3, Figure 3 (A)
"rO, 2, 3, . . ." is written in the empty page list. Furthermore, since the failed page in the real address space 14 in FIG. 2 is page 5, "5. . . ." is written in the failed page restore in FIG. 3 (b). When CPU2 performs processing, real address space 1
4 (that is, main memory 5) at the required address, which is done as follows. Now, suppose that the virtual address issued by CPU 2 belongs to page number 1 of virtual address space 10,
It is converted by the address conversion table 12, and the real address belonging to the real address space 14 (main memory 5) of page number 4 is accessed through the path shown by the solid line. If there is no page in the real address space 14 corresponding to the virtual address issued by the CPU 2, the corresponding page is loaded from the disk 9 into an empty page in the real address space 14. Accordingly, the address conversion table 12 is also modified. Then, access the real address in the same way as above. In the present invention, by newly providing a faulty page restore and using it to manage memory as follows, it is possible to continue operating the computer system even if there is a faulty memory if predetermined conditions are met. I can do it. FIG. 4 shows a flowchart illustrating the operation when a faulty memory is discovered in the present invention. A program that operates as shown in this flowchart is prepared in the memory management section of O3 (operating system). Item numbers ■ to ■ in the following explanation refer to steps ■ in Figure 4.
It corresponds to ~■. ■ A faulty memory is detected. Detection is performed by a fault detection circuit provided in hardware when accessing the main memory 5 (dotted line path in FIG. 2), or when a fault diagnosis program is run. ■ Check whether the page to which the faulty memory belongs (hereinafter referred to as the "failure page") is in use. In other words, it is checked whether the page has data copied from the disk 9 or whether it is an empty page. ■ If the page is in use, check whether the data that was lost in the failed memory section can be provided again, that is, whether it is possible to recover to a state where all the required data can be provided. This is done by checking update indication bit 13 in FIG. In the example in Figure 2, the failure page is page 5, but when you look at the corresponding update display bit 13, it is "0", which means that the data on page 5 has not been updated. Therefore, it is the same as the data corresponding to 5 pages remaining in the page file on disk 9. Therefore, by loading it into an empty page (for example, page 2) of the real address space 14, all the same data as in page 5 can be arranged. That is, in such a case, recovery is possible. However, if the update indicator bit 13 is "l", it means that any of the five pages of data has been updated after being loaded into the real address space 14, and the page on disk 9 has been updated. 5 remaining in the file
The page-wise data is no longer the same; in such a case, it is impossible to recover the data that was on page 5. (2) If it is recoverable, select an empty page from the empty page list 6 as appropriate, and load the same data from the page file on the disk 9 into that page as was loaded into the failed page. ■ The page number “5” listed in address conversion table 12
Rewrite ノ (faulty page number) with the empty page number used in step ① (“2” on the top). Also, "2" is deleted from the empty page list 6. ■ When recovery is not possible, the process using the data of the failed page cannot proceed, so it is terminated because it is assumed that there is an abnormality in the computer system. (2) In step (2), if the failed page is not in use, that is, if the failed page is an empty page, the page number is deleted from the empty page list 6. This is because if data is loaded from the page file of the disk 9 into such an empty page, data will obviously be lost, and the empty page will not be usable. ■ As shown in FIG. 3 (b), register the failed page in the failed page restore. Pages registered in the failed page restore are prohibited from being used thereafter. As a result of memory management as described above, as long as the faulty page has not been updated, the data is recovered and the process that the processor device 1 was attempting to perform continues. In addition, even if there is a faulty page, if the data can be recovered, the computer system can continue operating, so the contents of the faulty page restoration will be automatically notified through the communication means and output means connected to the computer system. I can do it. This allows for quick repairs and improves maintainability.

【Effect of the invention】

As described above, the memory management method in the paging virtual storage method of the present invention provides the following effects. ■ Even if there is a faulty page in the real address space, as long as the data on the faulty page has not been updated, it is now possible to reload the data from the page file to an empty page and continue operating the computer system. Ta. (2) A faulty page list was created, the page number of the faulty page was registered, and the use of the registered page was prohibited, so that the faulty page no longer interfered with the operation of the computer system.

[Brief explanation of the drawing]

Figure 1: A diagram showing a computer system to which the memory management method in the paging virtual memory system of the present invention is applied. Figure 2: A diagram showing the status of access from the virtual address space to the real address space. . . . A diagram showing the empty page list and the faulty page list in FIG. 2. FIG. 4 . . . A flowchart explaining the operation when a faulty memory is discovered in the present invention. In the flowchart diagram, 1 is a processor device, and 2 is a CPU. 3 is MMU, 4 is system bus, 5 is main memory, 6 is empty page list, 7 is failed page list, 8 is disk
10 is a virtual address space, 11 is a page number, 12 is an address conversion table, I3 is an update display bit, 14 is a real address space, and 15 is a page number.

Claims (1)

[Claims] A failure page list is provided to register the page number of the failure page, so that the failure page cannot be used from now on, and if the data of the failure page has not been updated, the data is reloaded from the page file to an empty page, and A memory management system in a paging virtual memory system, characterized in that an address translation table and an empty page list are rewritten with respect to the empty page.