JPH01273153A - Virtual storage control system - Google Patents

Abstract

PURPOSE:To facilitate the shift from a real storage system and to decrease the overhead by providing plural virtual spaces and mapping them so that virtual addresses of each user program are not overlapped even between different virtual spaces. CONSTITUTION:In a virtual space 1, an OS 5 and all user programs, namely, user programs 6-10 are mapped so that address areas are not overlapped. In a virtual space 2, the OS 5 and the user programs 6-8 are mapped. In a virtual space 3, the OS 5 and a user program 9 and a user program 10 are mapped. In this case, the user programs 6-8 and the user programs 9-10 are constituted so that the address areas are not overlapped mutually not only in the same space but also between the space 2 and 3. In such a way, it can be prevented that the area of other user program is brought to access by mis take when the user program is being executed, and even if the number of user programs increases, an overhead for changing a protective key of a main stor age is not generated.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a virtual memory control method in a computer system having a virtual memory mechanism, and particularly to a virtual memory control method that can be easily migrated from a real memory system and has little overhead.

[Conventional technology]

Generally, the virtual memory system in a computer system is
It has the following advantages compared to real memory systems.

■High efficiency in the use of real memory.

■A single program larger than the actual storage capacity can be executed.

■High main memory protection function.

Therefore, by migrating from a real storage system to a virtual storage system, it is possible to have the above advantages.

Conventionally, regarding virtual memory control methods for computer systems,
As stated in "Function and Structure of rMVs" written by Senda, published by Kindai Kagakusha, pages 5 to 9, only,
A single virtual space method using one virtual space and a multiple virtual space method using multiple virtual spaces are known.

In the single virtual space method, multiple programs to be executed simultaneously are loaded into one space, whereas in the multiple virtual space method, one space is allocated for each program, so a wide space can be used.

On the other hand, when migrating from a real storage system to a virtual storage system, the single virtual space method is easier. The reason for this is as follows.

■The single virtual space method, like the real storage system, has only one space, so the method can be used to allocate memory in the real storage system.

■To access the user program area from O8,
Unlike the multiple virtual space method, there is no need to specify the space in addition to the address. Just like in a real memory system, the address alone determines the memory you want to access.

(Problem to be solved by the invention) In the multiple virtual space system, although O8 manages control tables and buffers in the user area (in each virtual space),
In addition to the virtual space address, it is also necessary to remember a space number for identifying the virtual space, and when referring to addresses in other spaces, space switching overhead occurs. On the other hand, in the single virtual space method, just like the real storage system, only addresses need to be managed.

However, the single virtual space method has the following drawbacks compared to the multiple virtual space method. In other words, in the single virtual space method, in order to protect the main memory between different user programs, the main memory is divided using a protection key provided in the CPU, and during execution, an address conversion table (virtual address provided for each program → real address The main memory is accessible only when the lock value (JOB value) on the conversion table (conversion table) matches the protection key value of the main memory. For this reason, if the number of user programs (JOB number) exceeds the maximum number of keys held by the computer, for example 100 to 200, the protection keys must be dynamically replaced when switching user programs to be executed. The overhead increases accordingly. In addition, if the upper limit number of keys is set to the number of user programs that can be executed simultaneously, the comparison time between keys and locks will become longer, and the address translation table in main memory will also become thicker. As a result, the conversion from a virtual address to a physical address (real address) becomes slow, so the number of keys is generally limited to about 16.On the other hand, in the multiple virtual space method, each user program has an independent Since the main memory between user programs is protected by allocating this space, no such overhead exists.

As mentioned above, in the conventional virtual memory control method, in the case of a single virtual space method, it is necessary to replace keys, and
In the case of the multiple virtual space method, it is necessary to manage each space using a space number, and as a result, there are problems in migrating from a real storage system, such as overhead occurring when switching keys or switching spaces. .

Therefore, it is an object of the present invention to provide a virtual storage control method that can be easily migrated from a real storage system and has little overhead, solving the problems of the prior art described above.

[Means to solve the problem]

In order to achieve the above object, the present invention first provides a basic invention (first invention) in which a plurality of virtual spaces are provided in a virtual memory control method, and each virtual space has an operating system and mutual memory protection. Map only a few (within a few) user programs in order. (for example,
In a computer with eight types of keys, seven user programs are mapped to one virtual space, and the remaining one key is
Used to protect SW4 area. Furthermore, the computer of the ring protection mechanism maps only one user program. ) At this time, each user program area is allocated so that the user program areas do not overlap even in different virtual spaces (so that the virtual addresses of each user program do not overlap). Each user program is configured to be executed in the virtual space to which the user program is mapped.

Further, as a second invention, the present invention provides another virtual space in addition to the plurality of virtual spaces in the first invention. The operating system and all user programs are mapped to this added virtual space, and the operating system is configured to be executed in this virtual space.

[Effect]

The operation based on the above configuration will be explained.

In the first invention, only the number of user programs that can be mutually protected (the number of keys) is mapped in one virtual space, and the user programs can mutually protect their memories even if they are placed in the same space. In addition, other user programs that exceed the number that can be protected are mapped to a separate virtual space, so it is possible to prevent the areas of other user programs from being accessed by mistake while the user program is running. . Therefore, unlike the single virtual space system, even if the number of user programs increases, there is no overhead of replacing the protection key in the main memory. In addition, the virtual address of each user program does not overlap in the same virtual space or in different virtual spaces, so even if the virtual spaces are different, the user program area can be identified only by the virtual address. , O8 eliminates the need to manage space identifiers together, unlike the multiple virtual space system. That is, the virtual address (for example, its upper bit) of each user program also functions as a space identifier.

Further, in the second invention, in addition to the plurality of virtual spaces, another virtual space is provided in which all the user programs are mapped, and the OS is executed in the other virtual space. Since O5 can directly access all user program areas in the virtual space, there is no need to find the virtual space in which the user program area is mapped from the virtual address.

〔Example〕

First, before describing specific embodiments, an overview of the present invention will be explained in comparison with the prior art.

It is clear that the present invention differs from the conventional single virtual space system in that a plurality of virtual spaces are provided.

Further, the difference from the conventional multiple virtual space system will be explained using FIGS. 6 and 7. As shown in FIG. 6, in the conventional multiple virtual space system, the virtual addresses of each user program area are allowed to overlap, but in the present invention, the virtual addresses are allocated so as not to overlap. Therefore, when O3 specifies the address of the user program area, as shown in FIG.
As shown in FIG. 7(b), in the conventional multiple virtual space system, a space identifier must be specified in addition to the virtual address, whereas in the present invention, as shown in FIG. 7(b), the upper field of the virtual address is a space identifier. Since it also serves as an identifier, there is no need to manage space identifiers in addition to virtual addresses. Note that the present invention does not limit the method of memory allocation for the user program area (the length of the space identifier field of the virtual address is handled as a variable length).

The present invention further includes adding another virtual space in addition to the plurality of virtual spaces, and mapping O5 and all user programs to the other virtual space so as to overlap with the plurality of virtual spaces, By executing O3 in the other virtual space, all user program areas can be accessed from O3 in the other virtual space. The user program is not executed in the other virtual space.

Next, embodiments of the present invention will be described with reference to the drawings.

FIG. 1 is a diagram showing the arrangement of t'1-grams in the virtual space in this embodiment. O85 and all the user programs shown in FIG. 2, ie, user programs 6 to 10, are mapped in the virtual space 1 so that their address areas do not overlap. In virtual space 2, O35 and user programs 6 to 8 are mapped. O3 in virtual space 3
5 and user program 9 and user program 1
Map 0. In this case, the features of this embodiment are that user programs 6 to 8 and user programs 9 to 10 are
Address areas are made to not overlap not only within the same space but also between different spaces 2 and 3. Second
The figure shows the configuration of the program. User programs 6 to 10 are running under 035. FIG. 3 is a diagram showing an address conversion procedure from a virtual address to a physical address (real address) for realizing the above virtual space using a conversion table. Although FIG. 3 shows only the necessary conversion procedure for one program, in reality, a different conversion table is used for each program. Virtual address 20 is 32 bits long. Segment table pointer 5TP21 indicates segment table pointer 5TP21.
Starting physical address of table 5T22 (table 5T2
This is a register that indicates the starting real address of the position where 2 exists.

5T22 has 1024 (21°) entries 5TE2
3, each entry 23 has a page table P
It shows the starting physical address of T24 (the starting real address of the position where table PT24 exists). The upper 10 bits of virtual address 20 represent the entry number of 5TE23, and the upper 10 bits of virtual address 20 represent the entry number of 5TE23.
One of the 5TE23 of the entry number is selected by the bit. The FT24 has 1024 (21°) entries PTE25, and each PTH25 has a PN field 26 indicating the upper 20 bits of the physical address 28 corresponding to the virtual address 20 and a 2-bit length indicating memory protection data. field (key field)
Contains 27. The 10 bits in the range of the 12th to 21st bits of the virtual address 20 represent the entry number of the PTH 25. During address conversion, the 10 bits from the 12th bit to the 21st bit of the virtual address 20 represent the entry number of the PTH 25.
One of the PTH25 of T24 (the entry corresponding to the current entry number) is selected. The physical address 2 is 32 bits long, and as a result of address conversion, the lower 12 bits of the virtual address 20 are reflected as they are in the lower 12 bits of the physical address 28, and the PTH in the selected PTH 25 is
The upper 20 bits of the physical address indicated by the N field 26 are reflected in the upper 20 bits of the physical address 28, resulting in a 32-bit physical address (address on real main memory) 28.
is generated. The control register 29 is a register containing a 2-bit length access Jti-30 to the memory during program execution (the right to access the main memory only from the program and not from other programs, that is, lock). Access rights 30 in control register 29
The selected PTH 25 is compared with the storage protection data in the field 27, and only when the two values match, access to the main memory indicated by the virtual address 20 becomes possible. However, the access right value "0" is special, and the memory indicated by any virtual address 20 can be accessed regardless of the value of the K field. This is to enable access to all user program areas during O3 execution. Further, when control is transferred from the user program to O5, the value of the access right 30 is automatically set to "0".

Figure 4 shows the PTHs corresponding to each program area in Figure 1.
25 is a table showing the values of field 27.

In this embodiment, the K field is 2 bits as described above, and the operating system O8 and up to three user programs, Durham, can be specified in one virtual space as programs that can be executed by the user. It shows. However, the values of the fields in user program 6 and user program 9 are both "1",
In the virtual space 1, the values of both fields overlap, but since only O35 is executed in the virtual space 1, as will be described later, this does not pose a main memory protection problem when the user program is executed. The same thing can be said about the user program 7 and the user program 10. FIG. 5 shows the virtual space 1 shown in FIG. 2. 3 is a diagram showing the relationship between a segment table ST and a page table PT for mapping 3. FIG.

Here, one segment table corresponds to one space, and one page table corresponds to one user program or O8.

5T40 is virtual space 1, ST41 is virtual space 2, S
T42 is a segment table for mapping each virtual space 3. Segment table 5T40
．． Each of 41 and 42 corresponds to the segment table 5T22 in FIG.

PT45 has 035, PT46 has user program 6, PT41 has user program 7, PT48 has user program 8, PT49 has user program 9, P
T50 is a page table for mapping each user program 10. Page table PT4
Each of 5 to 50 is the page table PT in Figure 3.
It corresponds to 24. Since O35 is commonly mapped to each virtual space, Sr10 to 5T42 are mapped to the same PT45.
is the point. Since user program 6 to user program 8 are mapped to virtual space 1 and virtual space 2, 5T40 and Sr11 both point to PT46 to PT48. Similarly, since user program 9 to user program 10 are mapped to virtual space 1 and virtual space 3, Sr10 and Sr1
2 both point to PT49 or PT50.

As mentioned above, all user programs 6 to 10
Since the address areas do not overlap even between different virtual spaces, by looking at the upper bits of the virtual address 20 (different for each user program), you can tell which space the virtual address belongs to. In other words, this upper bit can be said to also serve as a space identifier.

Thereafter, O35 selects and executes the user program 6,
The operation of this embodiment will be explained by taking as an example a case where the user program 6 requests the O35 for data input service from an external storage device (not shown), and then the O85 selects and executes the user program 10. do. When O35 schedules a program and selects user program 6, it sets in 5TP21 the address of 5T41 that maps the virtual space 2 where user program 6 exists (as described above, the leading physical address). Thereafter, address translation is performed based on Sr11, and virtual space 2 is selected. Next, the access right 30 of the control register 29 is set to the value "
1", control is transferred to the user program 6. At the time the control is transferred to the user program 6, the value of the access right 30 is u 1 n, so the PTE
User program 6 (where the value of the field is 1n)
(see Figure 4), but the user program 7 or user program 8 area mapped to the same virtual space 2 cannot be accessed because the field values are different as shown in Figure 4. I can't.

User program 6 performs processing, specifies a buffer address, requests data input service from external storage, and returns control to O35. When control returns to 035, the value of access right 30 of control register 29 automatically changes to “
°0''.O35 is 5TP21 to 5T40
and switch to virtual space 1.

After storing the specified buffer address in the internal table, the O55 starts up the external storage device, schedules the program again, and executes the user program 10.
Select. The address of Sr12 that maps the virtual space 3 where the user program 10 exists is set in 5TP21. Next, the value "2" is set in the access right 30 of the control register 29, and control is transferred to the user program 10 at the same time. The user program 10 performs processing, but when a completion interrupt from the previously started external storage device occurs during the processing, control immediately returns to O35.

In this case as well, the access right 3 of the control register 29 is automatically
A value of 0 switches to "0". O35 sets the address of 5T40 in 5TP21, switches to virtual space 1,
The buffer address stored in the internal table is retrieved and the read data is written to the designated buffer in the user program 6 area. In this way, O85 is a virtual space 1 where all user programs are mapped.
Since the user program area is accessed in , there is no need to remember the virtual space identifier at runtime to manage buffer addresses. Then, schedule the program again and select user program 6.
The address of 5T41 that maps the virtual space 2 where the user program 6 exists is set in TP21. Next, the value "1" is set in the access right 30 of the control register 29, and control is transferred to the user program 6 at the same time.

According to this embodiment, since a plurality of user programs can be placed in the same virtual space, the number of virtual spaces can be reduced and the memory capacity required for the address translation table can be reduced.

〔Effect of the invention〕

As described in detail above, according to the virtual memory control method of the present invention, a plurality of virtual spaces are provided and the virtual addresses of each user program are mapped so that they do not overlap even between different virtual spaces. In order to manage control tables and buffers in the user program area within the system, there is no need to manage space identifiers in addition to addresses as in the conventional multiple virtual space system, so it is easy to migrate from a real storage system. In addition, unlike the conventional single virtual space method, there is no need to increase or replace main memory protection keys as the number of user programs increases, making it possible to realize a virtual memory control method with low overhead. .

Additionally, in addition to the multiple virtual spaces mentioned above, another virtual space is created in which all user programs are mapped, and O3 can access all user program areas. This has the effect that it is not necessary to obtain each one from the virtual address.

[Brief explanation of the drawing]

Fig. 1 is a diagram showing the configuration of a virtual space in an embodiment of the present invention, Fig. 2 is a diagram showing a program and duram configuration, Fig. 3 is a diagram showing an address conversion procedure, and Fig. 4 is a diagram showing a program protection key. FIG. 5 is a diagram showing the relationship between address translation tables, and FIGS. 6 and 7 are diagrams showing the difference between the present invention and the multiple virtual storage system. 1~3-・-・-・Virtual space, 5・・・・・・・・・0
316-10---user program, 20-...-
・Virtual address, 21...Segment table pointer STP, 22°40~42------
- Segment table ST, 23...- Segment table entry STE, 24°45~50-
・−1111 page table PT, 25・−・−・・
-Page table entry PTE, 28...
- Physical address, 29-- Control register. Figure 1 Figure 2 Figure 4 Figure 5 Figure 6 Figure 7 (a)

Claims (1)

[Scope of Claims] 1. A plurality of virtual spaces, and for each of the plurality of virtual spaces, an operating system and a number of user programs that can be protected mutually, and the virtual addresses of these user program areas are different. A virtual storage control method comprising: means for sequentially mapping virtual spaces so that there is no overlap; and means for executing each user program in the space to which the user program is mapped. 2. In addition to the plurality of virtual spaces, another virtual space is provided to which an operating system and all user programs are mapped, and the operating system is configured to be executed in the other virtual space. 1. The virtual memory control method according to 1.