JPH04145554A - Real memory assigning system - Google Patents

Abstract

PURPOSE:To supply a substitute real memory even when a fault occurs on a real memory page by making addresses of continuous areas in an arbitrary real memory space assignable to continuous page addresses in a virtual memory space. CONSTITUTION:When a virtual memory client 101 requests areas of continuous addresses in a real memory space corresponding to continuous areas 111 in a virtual memory space L pages, a program for making a real memory securing process 104 makes a request 105 for a real memory which is continuously unused for L pages. A real memory management program 106 retrieves real memory addresses which are unused continuously for L pages and returns the addresses to a virtual memory management program 103. The program 103 makes a process 109 for coupling the obtained real memory addresses R with the addresses V of virtual memory areas which are unused continuously for L pages. Therefore, areas 112 which are continuous for L page can be assigned in a real memory space and a substitute memory can be supplied when a real memory becomes defective.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a memory management system for controlling virtual space/address conversion of a supervisor program constituting a control program.

[Conventional technology]

Conventionally, regarding the effectiveness of areas in which memory addresses are consecutive in the real memory space, Japanese Patent Laid-Open No. 61-5358 discloses a method for increasing the speed of memory access.

JP-A-62-260242 or JP-A-63-
Improving the efficiency of data transfer during input/output in JP-A No. 141144, JP-A-59-172188 and JP-A-1-
Publication No. 116849 discussed improving the efficiency of converting virtual memory addresses to real memory addresses.

Furthermore, regarding a general continuous area management method that does not specify whether it is a real memory address or a virtual memory address, Japanese Patent Laid-Open No. 63
This was discussed in JP-A-89950, JP-A-1-274256, and JP-A-2-47744.

In addition, Japanese Patent Laid-Open No. 1-17137 discloses a real memory allocation method that allocates arbitrary real memory addresses to virtual memory in units of fixed-length pages, and provides a larger continuous real memory area. There were two types of page sizes, and some allowed you to select the page size for each area. They do not guarantee that memory addresses in the real memory space are continuous for areas in the virtual memory space whose size exceeds the maximum selectable page size.

Another real memory allocation method is to allocate an area in real memory space with the same real memory address as the virtual memory address to an area in virtual memory space, so that an area that is continuous in virtual memory space is also contiguous in real memory. There was something to ensure that it was a well-established area. (Hereafter referred to as virtual memory address.

(The area where the address in the real memory space allocated to the corresponding area is the same is called the V=R area) [Problem to be solved by the invention] Memory management for providing an area with continuous memory addresses also in the prior art However, the real memory allocation method, which allocates real memory to virtual memory on a page-by-page basis, mainly makes effective use of groups of pages that are discontinuous in the real memory space. When securing an area with contiguous addresses in memory space, any real memory address can be assigned only to an area in virtual space that is less than or equal to the page size, and cannot be allocated to an area that includes more than one page boundary in virtual memory space. In this case, only real memory addresses that matched virtual addresses were allocated.

That is, a means of securing an area in the virtual memory space where addresses in the real memory space are continuous and whose size is larger than a page is to select the V=R area, but there is a problem described below.

(1) Area size limitations; Conventionally, although the maximum size of the virtual memory space is, for example, 2 GB, the area reserved for the V=R area is limited to, for example, an area from address O to less than 16M address. Therefore, the size of the V=R area could not exceed this storage capacity.

In general, the memory addresses that can be used as the V=R area can be limited depending on the real memory layout of the minimum configuration in which the system can operate, and the size of the area can also be limited.

(2) Restrictions on alternative real memory; For areas other than the V=R area, even if a page in the real memory space that was allocated 5 becomes unusable due to a hardware failure, the page at the alternative real memory address can be reallocated, and failure recovery can be performed without changing the area address in the virtual memory space.

However, in the case of the V=R area, alternative memory could not be allocated without changing the layout of the virtual space area. In other words, recovery from a memory failure in the V=R area is as follows:
The correspondence between real memory and virtual memory could not be localized within the memory management program, and it had to affect clients in the virtual memory space.

In particular, the larger the secured area, the more likely a memory failure will occur within that area, so consideration regarding failure recovery is important.

An object of the present invention is to provide a virtual space area with continuous addresses in a larger real memory space while taking into account real memory failure recovery.

[Means to solve the problem]

The above purpose is to secure a contiguous free area in the real memory space of the same size as the page size of the area in the virtual memory space when real memory is allocated to an area in the virtual memory space, and to This is achieved by selecting the real memory allocation method of the present invention using means for selecting one of a plurality of types of real memory allocation methods when requesting to secure an area in the virtual memory space. (Hereinafter, the area where the addresses in the real memory space are consecutive for the areas that are continuous in the virtual memory space will be referred to as the v=C area.) Note that the above V=
In the C area allocation process, any real memory address may be allocated to the address of the virtual memory area as long as the real memory address allocated to the address of the corresponding virtual memory area is continuous and unused.

Also, since there is a lot of unused memory when the system starts,
It is easy to secure a real memory area with consecutive addresses as described above.

[Effect]

■=C area can be selected as one of a plurality of real memory allocation methods when requesting to secure an area in the virtual memory space.

Whether or not each page in the real memory space is unused can be determined by a conventional real memory management method. By repeatedly applying this means, it is possible to continuously search for an unused area in the real memory space of the size requested when requesting the v=C area.

After the page address in the real memory space to be allocated to the v=C area request is determined in this way, a virtual memory address can be allocated to the corresponding real memory address using conventional techniques.

Also, since the ■ = C area only guarantees that it is contiguous in the real memory space, even if the real memory used as the - degree v = C area becomes unusable, another same-sized Alternative memory can be replenished by continuously reallocating unused real memory areas.

In addition, in the case of V=R area, even if there is a continuous unused memory area in the real memory space, the area in the virtual space at the same address as the corresponding real memory address must be usable, but v= The independence of the virtual memory address and the real memory address of the C area is such that as long as there is sufficient continuous unused memory in each of the real memory page and virtual memory space, there is no restriction on the layout within each space.

〔Example〕

An embodiment of the present invention will be described below with reference to the drawings.6 Figure 2 shows a program configuration that is the premise of the program configuration of this embodiment shown in Figure 1. The program configuration during real memory allocation processing operation in a method that allocates an arbitrary real memory page is shown below.

First, when a virtual memory client (201) simply requests L pages of virtual memory area (202), a virtual memory management program (203) is called, and a program that performs real memory allocation processing 204 creates one page of unused space. Memory page request (205) is executed L times.

In response to this request, the real memory management program (206) searches for an unused real memory page and returns the obtained real memory page address R to the virtual memory management program (208). The virtual memory management program 203 performs a process (2
09) and set address V to virtual memory client 2.
01 (210).

Through the above-described processing, a group of L pages (212) in the real memory space can be allocated to an area (211) of L consecutive pages in the virtual memory space.

Next, referring to FIG. 1, the virtual memory client lO1
When requesting (102) the v=C area of L/<-ji, the virtual memory management program (103) is called,
A program that performs real memory securing processing (104) requests unused real memory once (105) for L consecutive pages. In response to this request, the real memory management program (106
) returns the real memory address R obtained by searching for unused real memory addresses for L consecutive pages to the virtual memory management program 103 (108).

The virtual memory management program 103 performs a process (109) to link the obtained real memory address R with the address ■ of an unused virtual memory area of L consecutive pages (109), and returns the address V to the virtual memory client 101 (110).
).

This results in an L page continuous area in the virtual memory space (
111), a continuous area (112) of L pages can be allocated in the real memory space as well.

Next, details of the real memory address search process performed by the real memory management program 106 will be explained using FIG. 3.

(302) while changing the index S among the page numbers for all the mounted real memories.

In the loop of (303), the final page number A of the continuous unused real memory area is determined from the page indicated by S, and (304
), it is tested whether A is greater than or equal to (S+L), that is, whether L pages are consecutively unused from the real memory page indicated by S, and the test is passed in the loop of (305) and (306). Make index S to L pages in use. Thereafter, the real memory address R (307) of the corresponding page is obtained from S indicating the page number of the real memory, and R is returned to the calling source (308).

If there is no S that passes the above test and the index S has been changed among the page numbers for all the implemented real memories, that is, if there is no contiguous area in the real memory space of the requested size, ( 309) is reached, and the same process as in the prior art when the real memory search fails is performed.

Here, the real memory address search process is described in Japanese Patent Laid-Open No. 63-899.
It is also possible to improve the efficiency by using improvements in memory release as shown in Japanese Patent No. 50 and Japanese Patent Application Laid-Open No. 63-103344, but when the system is started without any memory release requests, this example A simple real memory management method is sufficient.

Furthermore, although FIG. 1 shows the process of allocating a real memory area immediately after requesting a virtual memory area,
= The R area does not necessarily reside in real memory,
In the same way that real memory is released when swapping out, and real memory is allocated and the area contents are restored when swapping in, real memory is reallocated for the C area as well by the process shown in Figure 1 when swapping in. Then, the recovery of the area contents is V
= Swapping can be done in the same way as in the R area.

In addition, in the conventional technology, recovery from a memory failure in an area where real memory is allocated to virtual memory on a page-by-page basis is possible by using the page if the data in the page where the memory failure was detected has been paged out to the auxiliary storage device. It was possible to allocate alternative memory using the real memory reallocation process as part of the paging process.

Furthermore, to recover from a memory failure in the v=C area, if the data in the area where the memory failure has occurred has been swapped out to auxiliary storage, it can be swapped out by reallocating a contiguous area of real memory as described above. This can be achieved by replenishing the alternative real memory as part of the in-processing.

Figure 4 shows an example in which one table spans discontinuous real memory pages (real pages). In such a case, the table structure can be simplified by using the v=C area function of the present invention. . 401 is a first real page, 405 is a second real page, 402, 403, and 406 are table elements such as CCV (4 channel control word), and 404 is a command for passing control from page to page. In the v=C area, there is no need to insert a branch command for the next page at the end of the page.

〔Effect of the invention〕

According to the present invention, it is possible to secure a table of consecutive addresses on the real memory that is larger than the V=R area.

Particularly when there is a large amount of unused memory, such as when the system is started, the memory can be efficiently secured even with the simple process shown in the embodiment.

Also, compared to creating a table in a discontinuous area in page units in the real memory space, creating a table in a contiguous area in a larger real memory space can simplify the table structure and reduce the associated The performance of the process of creating the corresponding table can be improved.

Further, according to the present invention, by moving the table that was conventionally created in the V=R area to the v=C area, alternative memory can be supplied in the event of a failure, thereby improving the failure recovery function.

In addition, if the present invention is applied to a system that originally had a large virtual memory space and adopted a page-based memory allocation method, the above-mentioned effects can be obtained. The effect of adding functions to satisfy the configuration of the present invention by mainly applying page-based memory allocation to a system that was only able to allocate contiguous areas is that the memory allocation method of the parent system A group of fragmented pages in the real memory space, which could only be allocated to a virtual area with a size smaller than this size, can be allocated to a larger virtual memory area. In this case, the fragmented real memory can be effectively used without sacrificing the memory failure recovery function in the host system or the advantages of a contiguous area in the real memory space.

[Brief explanation of drawings]

FIG. 1 is a configuration diagram showing one embodiment of the present invention, and FIG.
FIG. 3 is a flowchart showing a real memory address search process, and FIG. 4 is a diagram showing a conventional example in which a table spans discontinuous real pages. 106...Real memory management program. Mihidai da

Claims (1)

[Claims] 1. In a virtual memory management method with a fixed-length page size, in a real memory allocation method that allocates page addresses in virtual memory space and page addresses in real memory space, real memory is assigned to consecutive page addresses in virtual memory. The main method is to allocate arbitrary page addresses in memory, but it is also possible to allocate addresses in any continuous area in real memory space to continuous page addresses in virtual memory space, and failures occur in real memory pages. A real memory allocation method that is characterized in that it is possible to replenish alternative real memory even when the memory is used.