JPH1078884A - Method for restoring data using virtual storage - Google Patents

Abstract

PROBLEM TO BE SOLVED: To preserver and restore data without describing any data to be restored or any timing of preservation or the like in a program by a person who prepares a program. SOLUTION: The writable bits of a page descripter in a page table 122 are turned off. A program processing part 111 advances the processing of a program, and generates interruption when a writing request is made to a page in which the writable bits of the page descripter in the page table 122 are turned off. A data preservation processing part 112 reserves the pertinent page content of a memory 120 in a disk 130 according to this interruption, and turns on the writable bits. Afterwards, a program processing part 111 operates writing to the pertinent page. When abnormality is found at a check point, a data restoration processing part 113 returns a page preserved in the disk 130 at that point to the memory 120.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a data recovery method in a computer having a virtual storage system.

[0002]

2. Description of the Related Art Conventionally, in information processing by a computer, an unexpected situation for a program creator may occur during execution due to a program error or a hardware failure. The above data is stored in a non-volatile storage device such as a disk, and after the storage is completed, the intended processing is advanced. If it is found that an abnormal situation has occurred, the processing is interrupted. A series of recovery processing of returning the stored value to the memory, resuming the state at the time of storage (the context that is the state of the processor), and restarting the processing has been performed. FIG. 4 shows a flowchart of such a conventional restoration process.

On the other hand, a computer of the virtual storage system has an address conversion mechanism for obtaining an address (real address) of a memory actually provided in the computer from an address (logical address) described in a program for processing by a processor. I have. FIG. 2 shows an address conversion mechanism of the segmentation paging system.
The actual page address in the page descriptor is obtained by sequentially indexing 0-N, and the relative address in the page is combined with the actual page address to obtain the address (address conversion destination address) of the corresponding page 230 in the memory. Is shown. FIG. 3 shows an address conversion mechanism of a paging system, and a logical address 300
Indicates that the actual page address in the page descriptor is obtained by directly indexing the page table 310 using the page field Pn, and the relative address D in the page is combined with the actual page address to obtain the address conversion destination address. In addition,
The high-speed conversion buffers 240 and 330 shown in FIGS. 2 and 3 store the correspondence between the logical address and the real page address obtained by recent address conversion, and store the conversion buffer before the index of the segment table or the page table. If the real address corresponding to the logical address is stored by reference, the real address is directly obtained.

Here, the page descriptor of the page table includes reference bits, update bit access right information, and the like.
Used for swapping. Each entry in the high-speed conversion buffer also stores a copy of this page descriptor. The address translation mechanism turns on the reference bit of the page descriptor when the processor refers to the memory address of the page associated with the page descriptor including the reference bit. The address translation mechanism turns on the update bit of the page descriptor when the processor performs a write operation on the memory address of the page associated with the page descriptor including the update bit. The operating system of the computer refers to the values of both bits to prevent page out of pages where the contents have been referenced or written, or to update the contents of pages that have been paged out once. If not, the same content is not paged out again. The access right information of the page descriptor is composed of a readable bit, a writable bit, and the like.
Here, the writable bit is OFF (write prohibited)
When the processor attempts to write to the address of the page corresponding to the page descriptor of the page in the memory, a memory protection interrupt occurs in the processor, and the processor executes the interrupt processing routine.

[0005]

In saving and restoring data, "what data" and "when to save" are important. However, in the conventional method as shown in FIG. 4, the program creator grasps in detail and exactly which data value is to be updated in what order, and determines the data to be restored and the timing of storage. Because of the need to decide, there was a problem that it was complicated and error-prone.

SUMMARY OF THE INVENTION An object of the present invention is to use a page descriptor in a page table so that when a processor attempts to write data to a memory, a nonvolatile memory such as a disk is used in units of a page including the address to be written. It is an object of the present invention to provide a data recovery method using virtual storage, in which a program creator does not need to consider “what data” and “when to save” by storing the data in a storage device.

[0007]

The data that needs to be restored is limited to data whose value has been changed by the processor. Therefore, if the state immediately before the processor attempts to change the value is stored, a recovery method can be realized in which the state is reproduced later and the program is executed again. For this reason, in the present invention, the write enable bit included in the page descriptor is set to OFF, and when the processor attempts to change the data of a certain page, an apparent memory protection interrupt is generated, and the interrupt processing routine executes At the same time that the contents of the page to be changed are saved on a disk or the like, the context (consisting of the program counter and register values) of the processor is saved. After the storage is completed, the processor proceeds with the original processing of changing the data. By proceeding with this procedure, it is possible to save the memory and context before updating each time data is updated. It is not necessary to provide the writable bit of the page descriptor separately from the original page by determining whether the page is the original write-protected area by the interrupt handler or the like. In the restoration process, the content of the page saved from the disk is returned to the original memory, and the context at that time is set, contrary to the order of saving. By implementing this series of procedures in an operating system, a creator of a program other than the operating system does not need to consider what data is stored and when.

[0008]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing a configuration example of a computer system for implementing the present invention. In FIG. 1, a processor 110, a memory 120, and a disk (non-volatile storage device) are connected by a bus 140. The processor 110 includes a program processing unit 111 that executes a process of a program, a data storage processing unit 112 that saves a page and a context before writing in the memory 120 to the disk 130 in an interrupt process, and a page and a context saved from the disk 130 in the memory 120. And a high-speed conversion buffer 114. Note that the processor 110 includes an operating system (OS), but is omitted in FIG. In the memory 120, there are a segment table 121 and a page table 122 for address translation, and the page descriptor of each entry of the page table 122 includes a reference bit, an update bit, and a reference bit shown in FIGS. The access right information of a readable bit and a writable bit is stored. The high-speed conversion buffer 114 is a memory 12
It may be provided within 0.

FIG. 5 is a processing flowchart of an embodiment of the data recovery method according to the present invention, which will be described below with reference to FIG.

Since the program is divided into several processing stages or the passage of time (checkpoint), they are named as processing (1), processing (2),..., Processing (n) in order, and hereinafter, processing (n) k). k represents 1, 2,..., n. The method of dividing the program into several stages can be realized by the operating system notifying the program of an exception typified by a timer exception. In this case, the program creator does not need to consider the steps.

Step 1: Initial setting The update bits of the page descriptors of all entries in the page table are turned off. Similarly, the writable bits of the page descriptors of all entries in the page table are turned off. Also, the number of times the restoration process is repeated or the value is set. Thereafter, the process proceeds to step 2.

Step 2: Program processing and data storage Step 2 is Step 2-1 of the processing of the program processing unit 111 and Step 2- of the processing of the data storage processing unit 112
Divided into two.

Step 2-1: Processing (K) The processing of the program is advanced. Here, if there is a write to a page whose write enable bit of the page descriptor of the page table is OFF, an interrupt occurs and the process proceeds to step 2-2. Then, when the processing of step 2-2 is completed, writing to the corresponding page is executed. At this time,
The update bit of the page descriptor is turned ON. When the process (k) is completed, that is, when the next check point is reached, the process proceeds to step 3.

Step 2-2: Data Storage The context of the memory, the page table, and the page contents before writing are stored in the disk by interrupt processing. At this time, it is made clear which page descriptor of the stored page table corresponds to the stored page. Set the writable bit of the corresponding page descriptor of the page table to O.
N and return to step 2-1. When writing is attempted on the original write-protected page (in this case also, the writable bit of the corresponding page descriptor is OFF.
This step 2-2 is not executed, and an original access exception interrupt is generated instead.

Step 3: Judgment of end of process (k) It is judged whether or not the process (k) of the program has been completed without any program error or hardware failure. This determination is realized by using a failure interrupt function of the processor or by providing a check processing procedure in a program. If the operation has been completed normally, the process proceeds to step S4. If it is not normal, go to step 5. Step 4: Set page table Turn off the writable bit of each page descriptor of the page table. Similarly, the update bit of each page descriptor of the table is turned off. Thereafter, the process proceeds to step 2 in order to execute the process (k + 1) at the next stage of the process (k). Alternatively, when the processing ends in the process (k), the processing of the program ends.

Step 5: Restoring Memory and Context The page contents after the start of the processing (k) are described in step 2-
Since the data is stored on the disk by 2, it is written back to the memory. The current page table is changed to the page table saved first after the process (k) starts. The current context is changed to the first saved context after the processing (k) starts. The process moves to step 2 when the context is restored. This allows
The re-execution of the process (k) is started. In step 5, the number of restoration processes is counted. Then, in step 3, this count value is set in step 1
If the number of repetitions of the recovery processing set in the step is exceeded, the program is terminated as "unrecoverable".

[0018]

According to the present invention, in a computer of a virtual storage system, which data in a memory is updated in what order is stored in a page descriptor of a page table in a unit called a page. Creators do not need to create programs that represent complex and error-prone procedures that accurately and precisely reflect the order in which data to be recovered and the values between data are updated in order to enable recovery processing. Can be. Also, the saving process can be realized by the same procedure as the conventional paging, so that the performance degradation due to the saving process can be almost the same as the degradation due to the realization of the virtual storage.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration example of a computer system that implements the present invention.

FIG. 2 is a diagram illustrating an address conversion of a segmentation paging method.

FIG. 3 is a diagram illustrating address conversion in a paging system.

FIG. 4 is a processing flowchart of a conventional data recovery method.

FIG. 5 is a processing flowchart of an embodiment of the data recovery method of the present invention.

[Explanation of symbols]

 110 processor 111 program processing unit 112 data storage processing unit 113 data recovery processing unit 114 high-speed conversion buffer 120 memory 121 segment table 122 page table 130 disk (non-volatile storage device)

Claims (2)

[Claims] 1. A computer provided with a virtual storage system,
If an abnormal condition occurs during the execution of the program processing,
This is a data recovery method for returning to the state at the time when processing was normally executed before that, and in the initial state, the update bit and the writable bit included in the page descriptor in the page table are set to OFF. When the processor attempts to write data to a page in which the writable bit included in the page descriptor is OFF, the processing shifts to interrupt processing, and the page contents before writing, the page table, and the context are stored in a nonvolatile manner. And save the writable bit to O
After returning from the interrupt processing as N and performing the data write processing, the update bit included in the page descriptor is turned ON, and whether or not the program processing has been normally completed at a predetermined processing stage or at every elapse of time. If the processing is completed normally, the program processing is continued after turning off the writable bit and the update bit included in the page descriptor in the page table. If the processing is not completed normally, the nonvolatile memory is used. By writing back the page content stored in the non-volatile storage device to the memory and changing the current page table and context to the page table and context stored in the non-volatile storage device, the state before writing the data is restored. A data recovery method using virtual storage. 2. In the initial state, the number of repetitions of the data recovery process when the process is not normally completed at the time of the determination is set, and when the number of the data recovery processes exceeds the number of repetitions, the program process is executed. 2. The data recovery method using virtual storage according to claim 1, wherein the process is terminated.