JP2009169747A - Information processor and data restoration method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To restore the state of data stored in a nonvolatile main storage part when power supply is restored, and perform backup with a reduced backup area. <P>SOLUTION: A backup part 110 reads, upon detection of a writing address, original data from a nonvolatile main storage part 120 to generate backup data, stores an address contained in the backup data to a history table 118, writes the backup data to a nonvolatile backup memory 130 if an address matched to a writing destination address is not stored in the history table 118, writes the writing destination address to the history table 118 when the backup data is written to the backup storage part, and writes data of the writing object to a storage area designated by the writing destination address in the nonvolatile main storage part 120 after the writing of the history table 118 is completed. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to an information processing apparatus including a nonvolatile main storage unit and a data recovery method.

  In recent years, the development of nonvolatile memory technologies such as flash memory, magnetic random access memory (MRAM), ferro-electric random access memory (FeRAM), and phase change random access memory (PRAM) that will not lose recorded content even when power is turned off. Progressing. In the future, it is expected that these non-volatile memories will replace volatile memories such as DRAMs that are currently widely used as the main memory of computers.

  When such a non-volatile memory is used for the main storage unit of a computer, even if power supply to the main storage unit is interrupted during processing of a program, the data stored in the main storage unit and the state of the system are stored in the main memory. It remains in the department and is not lost. As a result, when the power supply is restored, it becomes easy to restart the operation of the computer from the moment when the power is turned off, the startup time of the computer is shortened, and the operability is improved. In addition, unlike a volatile memory, a non-volatile memory does not require a power supply device for holding stored contents, and therefore, reduction of power consumption, battery size, and battery life are expected as a whole computer. .

  When a computer using such a non-volatile memory is suddenly powered off and then restored, the method of maintaining and restoring the state at the moment when the power is turned off is as follows: A method is disclosed in which an interrupt is notified to a CPU by a power supply monitoring mechanism, and the state of the CPU is saved in a main memory by control software such as an operating system (see Patent Document 1).

JP-A-2005-107573

  However, in the technique described in the above-mentioned Patent Document 1, in order to be able to restore the volatile system state such as the state of the processor when the power after the unexpected power failure is restored, after the unexpected power failure, It is necessary to secure a power source while storing the system state in a nonvolatile memory (hereinafter referred to as a nonvolatile main storage unit). Therefore, a battery such as a capacitor is required separately. That is, a battery for holding the system state stored in the nonvolatile main storage unit is not required, but a battery for storing the system state when the power is suddenly turned off is required. In general, a battery provided for a computer to complete the writing process is often a secondary battery, and there is a possibility that sufficient power cannot be supplied when frequent power interruption occurs.

  Further, in the technique described in Patent Document 1, a process log is taken in order to return the state of the system including the contents of the nonvolatile main storage unit to a state at a specific time before the power is turned off. However, a device that can use only limited resources such as an embedded device may be able to use only a small amount of memory as an area for writing a log due to physical restrictions. In such a case, if log data overflows from a small amount of log area, it is impossible to restore any previous state, and the system as a whole cannot continue processing.

  The present invention has been made in view of the above, and in an information processing apparatus including a nonvolatile main storage unit, even when an unexpected power failure occurs without using an additional power source such as a battery, It is an object of the present invention to provide an information processing apparatus and a data recovery method capable of restoring the state of data stored in a nonvolatile main storage unit at the time of subsequent power recovery and performing backup with a small backup area.

  In order to solve the above-described problems and achieve the object, the present invention provides a nonvolatile main storage unit, a nonvolatile backup storage unit that stores a copy of data stored in the main storage unit, and a processor Storage area specified by the address when a write access including the data to be written to the main storage unit and the write destination address specifying the write destination of the data to be written is detected. Reading means for reading data already stored in the memory, generation means for generating backup data including the data read by the reading means, and the address included in the write access, and writing to the backup storage unit History information storage means for storing the address included in the backup data to be stored, and storage in the history information storage means History information search means for searching for an address that matches the address of the write destination from among the written addresses, and if the address that matches the address of the write destination is not stored in the history information storage means, the backup A first writing unit for writing data to the backup storage unit; and a second writing unit for writing the write destination address to the history information storage unit when the backup data is written to the backup storage unit. And a third writing means for writing the data to be written into the storage area specified by the address of the main memory after the writing by the second writing means is completed. It is characterized by that.

  The present invention is also a data recovery method executed by the information processing apparatus, and includes a nonvolatile main storage unit, and a nonvolatile backup storage unit that stores a copy of data stored in the main storage unit. A data recovery method for an information processing apparatus comprising: a write including data to be written to the main storage unit and a write destination address designating a write destination of the data to be written, sent from a processor unit A read step for reading data already stored in the storage area designated by the address when access is detected; the data read in the read step; and the address included in the write access Included in the generation step of generating backup data and the backup data to be written to the backup storage unit A history information search step for searching for an address that matches the address of the write destination from among the addresses stored in the history information storage means for storing the recorded address; and a match with the address of the write destination in the history information storage means If the address is not stored, a first writing step of writing the backup data to the backup storage unit, and the history information storage means when the backup data is written to the backup storage unit A second writing step for writing the write destination address; and after the writing by the second writing step is completed, the data to be written is written to the storage area specified by the address of the main storage unit. And a third writing step.

  According to the present invention, in an information processing apparatus including a nonvolatile main storage unit, even when an unexpected power failure occurs without using an additional power source, when the power is restored, the stable nonvolatile main memory closest to the power failure is restored. Can be restored, and backup processing and necessary backup areas can be reduced as much as possible, reducing the amount of backup memory used and reducing the performance of existing information processing devices. .

  Exemplary embodiments of an information processing apparatus and a data recovery method according to the present invention are explained in detail below with reference to the accompanying drawings. Although the example which applied the information processing apparatus concerning this invention to a personal computer (henceforth PC) is shown, it is not limited to this, If it is an apparatus provided with a main memory part, a server apparatus, a game machine, etc. It can be applied to various devices.

(First embodiment)
A first embodiment will be described with reference to the accompanying drawings. First, a configuration example of a PC to which the present invention is applied will be described. FIG. 1 is a block diagram showing the configuration of the PC 100 according to the first embodiment.

  The PC 100 according to the present embodiment includes a backup unit 110, a nonvolatile main storage unit 120, a nonvolatile backup memory 130, a commit unit 140, a rollback unit 150, a processor unit 160, control software 170, And a bus 180. Further, the PC 100 is supplied with power from the power source 200 via the power line 10.

  The nonvolatile main storage unit 120 is a nonvolatile memory such as MRAM or FeRAM. The nonvolatile main storage unit 120 stores a software program operating on the processor unit 160 and data used by the software. The processor unit 160 sends a write access to the main memory access control unit 111 when writing to the nonvolatile main memory unit 120. The main memory access control unit 111 performs writing to the nonvolatile main storage unit 120 in accordance with the write access received from the processor unit 160. Here, the write access refers to the data that the processor unit 160 writes to the nonvolatile main storage unit 120 and the address of the nonvolatile main storage unit 120 that specifies the write destination of the data (hereinafter referred to as the write destination address). It is information to include.

  The non-volatile backup memory 130 is a non-volatile memory like the non-volatile main storage unit 120. The non-volatile backup memory 130 records backup data. Here, the backup data refers to data already stored in the storage area of the nonvolatile storage unit 120 indicated by the write destination address included in the write access when the write access to the nonvolatile main storage unit 120 is detected. A write destination address that is replicated data, specifically, an address of data that is updated at least in the nonvolatile main storage unit 120 to enable recovery of the data to the nonvolatile storage unit 120, and the address Data before update (hereinafter referred to as original data) stored in the.

  The backup unit 110 creates backup data at the time of writing to the nonvolatile main storage unit 120 and writes the backup data to the nonvolatile backup memory 130. Specifically, when the access to the nonvolatile main storage unit 120 is monitored and the write access to the nonvolatile main storage unit 120 is detected, the storage area specified by the write destination address included in the write access The original data already stored in the non-volatile main storage unit 120 is read out. Next, the backup unit 110 generates backup data from the write destination address and the original data included in the write access, and writes the backup data in the nonvolatile backup memory 130. Thereafter, the backup unit 110 actually writes the data instructed to write to the nonvolatile main storage unit 120 by the write access to the storage area designated by the write destination address of the nonvolatile main storage unit 120.

  The backup unit 110 further includes a main memory access control unit 111, a backup control unit 112, a backup memory access control unit 113, a backup data generation unit 114, a backup memory start address holding unit 115, and a write pointer holding unit 116. A history table management unit 117, and a history table 118.

  The main memory access control unit 111 reads data stored in the non-volatile main storage unit 120 in response to read / write access to the non-volatile main storage unit 120 received via the bus 180, and stores data in the non-volatile main storage unit 120. Write. However, when a write access to the nonvolatile main storage unit 120 via the bus 180 is detected, the main memory access control unit 111 does not immediately write data to the nonvolatile main storage unit 120, but temporarily performs backup control. The write destination address is sent to the unit 112.

  Further, the main memory access control unit 111 reads / writes data stored in the nonvolatile main storage unit 120 in response to access from the backup control unit 112 to the nonvolatile main storage unit 120. The main memory access control unit 111 sends access to the non-volatile backup memory 130 via the bus 180 to the backup control unit 112. Further, the main memory access control unit 111 sends access to the write pointer holding unit 116 to the backup control unit 112. The main memory access control unit 111 corresponds to the reading unit and the third writing unit described in the claims.

  When the backup control unit 112 receives an access from the main memory access control unit 111 to the non-volatile main storage unit 120, the backup control unit 112 passes the main memory access control unit 111 to back up the data. To read data stored in the storage area indicated by the write destination address. In addition, the backup control unit 112 sends the data read from the write destination address and the nonvolatile main storage unit 120 to the backup data generation unit 114.

  The backup control unit 112 reads and writes data stored in the nonvolatile backup memory 130 via the backup memory access control unit 113. Further, the backup control unit 112 reads and writes the write pointer stored in the write pointer holding unit 116 and the backup memory start address stored in the backup memory start address holding unit 115. The backup control unit 112 corresponds to the first writing unit and the second writing unit described in the claims.

  The backup memory access control unit 113 reads / writes data stored in the non-volatile backup memory 130 in response to an access from the backup control unit 112.

  The backup data generation unit 114 generates backup data from the write destination address of the nonvolatile main storage unit 120 and the original data read from the storage area indicated by the write destination address of the nonvolatile main storage unit 120 transmitted from the backup control unit 112 To do.

  The backup memory head address holding unit 115 holds the head address of the nonvolatile backup memory 130. The backup memory head address holding unit 115 updates the content in response to a request from the backup control unit 112 or sends the content to the backup control unit 112.

  The write pointer holding unit 116 holds a write pointer. Here, the write pointer is an address on the non-volatile backup memory 130 into which backup data is written. The write pointer holding unit 116 updates the content in response to a request from the backup control unit 112 or sends the content to the backup control unit 112.

  The history table management unit 117 operates the history table according to an instruction from the backup control unit 112. There are three instructions from the backup control unit 112 to the history table management unit 117. The first instruction is to register the address of the non-volatile main storage unit 120 that has already been backed up in the history table 118. Second, it instructs to search the history table 118 for the address of the nonvolatile main storage unit 120 to be backed up. The third is an instruction to clear the history table 118.

  The history table 118 is a storage area for storing the address of the nonvolatile main storage unit 120 backed up in the nonvolatile backup memory 130. The history table 118 is accessed and managed by the history table management unit 117.

  Here, the history table 118 will be further described. FIG. 2 is an explanatory diagram showing an example of the data configuration of the history table 118. The history table 118 is configured as a simple table as shown in FIG. In the storage area of the history table 118 (hereinafter referred to as history entry 21), the addresses in the nonvolatile main storage unit 120 that have already been backed up are stored. When the history table 118 has a simple table format as shown in FIG. 2, the registration process to the history table 118 can be realized by writing an address in the empty history entry 21. The search for the backed-up address in the history table 118 can be realized by comparing whether or not the search target address is registered in the history entry 21 in the history table 118, that is, by comparing the addresses. Further, clearing of the history table 118 can be realized, for example, by storing an illegal address as the address of the nonvolatile main storage unit 120 in all the history entries 21 so that they do not coincide at the time of retrieval.

  In addition, the backup unit 110 includes a switch for determining whether or not to perform the backup process itself, and also has a function of enabling or disabling the entire backup process for writing to the nonvolatile main storage unit 120. It should be noted that this switch indicating whether or not to perform backup processing when a write access to the nonvolatile main storage unit 120 is detected may be a physical switch or information held in a memory.

  The commit unit 140 performs a commit process. The commit process is a case where the power is restored after a sudden power interruption, and when the state of the nonvolatile main storage unit 120 needs to be restored, the state of the nonvolatile main storage unit 120 which is a restoration target It is a series of processes for creating (hereinafter referred to as a checkpoint). Specifically, when the commit unit 140 rewrites the backup data newly stored in the non-volatile backup memory 130 by the backup unit 110 from the time when the commit process is instructed, the storage content of the non-volatile main storage unit 120 is rewritten. This is realized by setting the read permission as the backup data used for. Note that the details of the commit process depend on the method of using data in the nonvolatile backup memory 130 (stack or ring buffer, etc.).

  The rollback unit 150 has read and read the backup data set as read permission necessary for restoring the contents of the nonvolatile main storage unit 120 from the backup data stored in the nonvolatile backup memory 130. The original data extracted from the backup data is written back to the storage area indicated by the write destination address of the nonvolatile main storage unit 120 extracted from the backup data. By this processing, the state of the checkpoint that was most recently created in the nonvolatile main storage unit 120 is restored. The backup control unit 112 corresponds to the extraction unit and the fourth writing unit described in the claims.

  In order for the rollback unit 150 to perform correct restoration processing, the state of the nonvolatile main storage unit 120 is changed to the latest checkpoint state from one or a plurality of backup data stored in the nonvolatile backup memory 130. It is important that the backup data necessary for restoration can be read out. That is, the nonvolatile backup memory 130 needs to manage backup data necessary for restoring the state of the nonvolatile main storage unit 120 to the latest checkpoint state. Therefore, in the commit process, as one or a plurality of backup data necessary for the restoration process that is managed by the nonvolatile backup memory 130, the backup data managed before the commit process is newly transferred to the nonvolatile backup memory 130. A process for setting read permission to the backup data to be stored is performed.

  Here, a series of backup data recorded in the nonvolatile backup memory 130 from one commit process to the next commit process and to be managed is called a sequence. A series of backup data recorded in the nonvolatile backup memory 130 after the last commit process is referred to as the latest sequence. That is, the sequence managed by the nonvolatile backup memory 130 at a certain point in time is the latest sequence, and the process of setting the backup data newly stored in the nonvolatile backup memory 130 as the latest sequence as read permission is specifically In other words, it is the latest sequence switching process.

  The processor unit 160 causes the control software 170 to operate. In normal processing, the control software 170 operates the data in the nonvolatile main storage unit 120 to provide a system-specific function. In addition, the control software 170 issues a request to the commit unit 140 at an appropriate timing (for example, at regular intervals), and creates a checkpoint for the nonvolatile main storage unit 120. The control software 170 issues a request to the rollback unit 150 when power is restored, and performs processing to return the state of the nonvolatile main storage unit 120 to the latest checkpoint.

  The bus 180 connects the commit unit 140, the rollback unit 150, the processor unit 160, and the backup unit 110. The nonvolatile main storage unit 120 is not directly connected to the bus 180 but is connected to the bus 180 via the backup unit 110. Thereby, since all accesses to the nonvolatile main storage unit 120 are via the backup unit 110, the backup unit 110 can perform backup processing according to the access to the nonvolatile main storage unit 120. Note that the arrangement position of the backup unit 110 is not limited to the position between the bus 180 and the nonvolatile main storage unit 120, and can be received before the nonvolatile main storage unit 120 receives access to the nonvolatile main storage unit 120. As long as it is, it may be arranged at any position. In addition to the nonvolatile main storage unit 120, a nonvolatile backup memory 130 is connected to the backup unit 110.

  Next, backup processing by the backup unit 110 of the PC 100 configured as described above, commit processing by the commit unit 140, and rollback processing by the rollback unit 150 will be described.

  FIGS. 3A and 3B are flowcharts illustrating the backup processing procedure performed by the backup unit 110. The backup unit 110 operates by receiving an access from the bus 180.

  First, the main memory access control unit 111 detects an access from the bus 180 (step S301). The main memory access control unit 111 determines whether or not the detected access is a read access to the nonvolatile main storage unit 120 (step S302). Specifically, it is determined whether or not it is a read access to the nonvolatile main storage unit 120 from an instruction included in the access. In addition, not only for read access but also for other accesses such as write access, the type of access and the storage medium to be accessed are determined from the instructions included in the access.

  When it is determined that the detected access is a read access to the nonvolatile main storage unit 120 (step S302: Yes), the main memory access control unit 111 reads data from the nonvolatile main storage unit 120 (step S303), The read data is returned to the bus 180.

  When it is determined that the detected access is not a read access to the nonvolatile main storage unit 120 (step S302: No), the main storage access control unit 111 sends the access to the backup control unit 112, and the backup control unit 112 accesses Is received (step S304). The backup control unit 112 determines whether the received access is a write access to the nonvolatile main storage unit 120 (step S305).

  When it is determined that the received access is a write access to the nonvolatile main storage unit 120 (step S305: Yes), the backup control unit 112 determines whether the backup process is valid (step S306). If it is determined that the backup process is not valid (step S306: No), the backup process is not performed and the process proceeds to step S317. When it is determined that the backup process is valid (step S306: Yes), the backup control unit 112 sends the write destination address to the history table management unit 117 (step S307). The history table management unit 117 searches whether the received write destination address is recorded in the history table 118 (step S308). The history table management unit 117 determines whether or not the received write destination address is recorded in the history table 118 (step S309). When it is determined that the received write destination address is recorded in the history table 118 (step S309: Yes), the process proceeds to step S317.

  As a result, in the latest sequence, when a write access is made to a certain address and a write access is made to the same address, if there is backup data for the first write access, it can be rolled back to the checkpoint state. Therefore, backup data is not stored in the nonvolatile backup memory 130.

  That is, by performing such processing, a plurality of backup data for the same address is not written in the same sequence, and the same function can be realized with a backup memory having a small capacity. Furthermore, since the number of backup data writing processes themselves can be reduced, it is possible to suppress the performance degradation of the entire PC 100 due to the backup processes.

  When it is determined that the received write destination address is not recorded in the history table 118 (step S309: No), that is, when it is the first write access to the address after the latest commit process, the backup control unit 112 A read request is sent to the storage access control unit 111, and the main storage access control unit 111 reads the original data stored in the storage area indicated by the write destination address from the nonvolatile main storage unit 120 (step S310). Next, the backup unit 110 sends the write destination address and the original data to the backup data generation unit 140, and the backup data generation unit 140 generates backup data from the received write destination address and the original data (step S311).

  The backup control unit 112 reads the address on the nonvolatile backup memory 130 into which the generated backup data is written from the write pointer holding unit 116 (step S312). The backup memory access control unit 113 writes the backup data in the storage area indicated by the read address on the nonvolatile backup memory 130 (step S313). The backup control unit 112 updates the value of the pointer held in the write pointer holding unit 116 (step S314). The backup control unit 112 sends the write destination address to the history table management unit 117 and instructs recording (step S315). The history table management unit 117 determines a storage area (history entry) in the history table 118 according to the structure of the managed history table and records the write destination address (step S316). The main memory access control unit 111 writes data in the storage area indicated by the write destination address of the nonvolatile main storage unit 120 (step S317).

  If it is determined in step S305 that the received access is not a write access to the nonvolatile main storage unit 120 (step S305: No), the backup control unit 112 determines whether the access is to the nonvolatile backup memory 130 or not. Judgment is made (step S318). If it is determined that the access is to the nonvolatile backup memory 130 (step S318: Yes), the backup memory access control unit 113 reads data from the nonvolatile backup memory 130 or writes data to the nonvolatile backup memory 130. (Step S319).

  If it is determined that the access is not to the nonvolatile backup memory 130 (step S318: No), the backup control unit 112 determines whether the access is to the write pointer holding unit 116 (step S320). If it is determined that the access is to the write pointer holding unit 116 (step S320: Yes), the backup control unit 112 reads or updates the value of the write pointer holding unit 116 (step S321). If it is determined that the access is not to the write pointer holding unit 116 (step S320: No), the process ends.

  Next, commit processing will be described. FIG. 4 is a flowchart showing a commit processing procedure performed by the commit unit 140. The commit unit 140 is activated in response to an instruction for commit processing for the backup unit 110 from the control software 170 or the rollback unit 150 operating on the processor unit 160.

  First, the backup control unit 112 instructs the history table management unit 117 to clear the history table 118 (step S401). When the history table management unit 117 receives an instruction to clear the history table 118 from the backup control unit 112, the history table management unit 117 clears the data in the history table 118 according to the structure of the managed history table 118 (step S402). Next, the commit unit 140 switches the latest sequence on the nonvolatile backup memory 130 (step S403). By this commit process, the state of the nonvolatile main storage unit 120 at this point becomes a new check point of the nonvolatile main storage unit 120. The details of the commit process depend on the configuration of the backup unit 110 and the format of the backup data. For example, when the data structure of the non-volatile backup memory 130 is a stack, switching of the latest sequence is realized by storing the value held in the backup memory head address holding unit 115 in the write pointer holding unit 116.

  Next, the rollback process will be described. FIG. 5 is a flowchart showing a rollback processing procedure performed by the rollback unit 150. The rollback unit 150 is activated by an instruction from the control software 170 operating on the processor unit 160.

  First, the rollback unit 150 sends a request to invalidate the backup process to the backup unit 110 (step S501). The backup unit 110 invalidates the backup process (step S502). This is so that when the backup data is written back to the non-volatile main storage unit 120, the data of the non-volatile main storage unit 120 that is indefinite due to an unexpected power interruption is not stored in the non-volatile backup memory 130 as backup data. It is to do.

  Next, the rollback unit 150 identifies the latest sequence, which is a series of backup data necessary for the rollback process (step S503). Note that the latest sequence identification method depends on the configuration of the backup unit 110 and the format of the backup data. The rollback unit 150 stores the latest backup data in the read latest sequence in E (step S504). The rollback unit 150 determines whether E is empty (step S505).

  If it is determined that E is not empty (step S505: No), the rollback unit 150 sends an E read request to the backup unit 110 (step S506). The backup unit 110 reads the backup data E from the nonvolatile backup memory 130 (step S507). The rollback unit 150 sends to the backup unit 110 a request to write back the original data also included in E to the storage area indicated by the write destination address of the nonvolatile main memory 120 included in E (step S508). The backup unit 110 writes the original data to the designated address in the nonvolatile main storage unit 120 (step S509). The rollback unit 150 stores the next backup data in the latest sequence in E (step S510). As a result, the backup data can be read out in the reverse order of the times recorded in the latest sequence, that is, in the newly stored order. Returning to step S505, this process is repeated for all backup data in the latest sequence.

  If it is determined in step S505 that E is empty (step S505: Yes), the rollback unit 150 sends a request to enable backup processing to the backup unit 110 (step S511). The backup unit 110 validates the backup process (step S512). The rollback unit 150 sends a commit request to the commit unit 140 (step S513). The commit unit 140 performs a commit process (step S514). Thereby, a checkpoint at the time when the rollback process is completed is generated.

  Considering the case where the history table 118 does not exist in the backup process, the data stored in the nonvolatile main storage unit 120 must be read once for every writing to the nonvolatile main storage unit 120. Even if it is assumed that a non-volatile memory element having the same performance is used for the non-volatile backup memory 130 and the non-volatile main storage unit 120, it is considered that the writing process to the non-volatile main storage unit 120 takes twice as long. A decrease in the processing performance of the PC 100 is expected.

  In addition, the control software 170 operating on the processor unit 160 has a feature of accessing the same or close storage areas if they are close in time (called temporal locality and spatial locality), and is short. It is known that access including writing is often made to the same or peripheral addresses of the main memory unit in time.

  Therefore, as in the present embodiment, the data of the same address in the nonvolatile main storage unit 120 is backed up by recording the address of the backed up nonvolatile main storage unit 120 in the history table 118 and checking at the time of backup. Therefore, it is possible to prevent the performance of the PC 100 from being deteriorated and to reduce the amount of use of the nonvolatile backup memory 130. Further, the PC 100 can be configured with a smaller amount of the non-volatile backup memory 130.

  Next, an example of a history table configuration different from the above-described configuration will be described. FIG. 6 is an explanatory diagram showing an example of the data configuration of the history table 118.

  The history table 118 shown in FIG. 6 stores the history entry 21 and the valid bit 22 in association with each other. When the history table 118 shown in FIG. 6 is used, when the write destination address included in the detected write access to the nonvolatile main storage unit 120 is already backed up, the value of the valid bit 22 is used. It can be determined by whether or not the corresponding history entry 21 is used. Also, the process of clearing the history table 118 can be realized by clearing only the valid bit 22.

  As another configuration example, a configuration example of a history table using a hash will be described. FIG. 7 is an explanatory diagram showing an example of the data configuration of the history table 118.

  The history table 118 illustrated in FIG. 7 includes a storage area (history entry 21) whose index is expressed by 0 to N-1. The history table management unit 117 includes a hash value calculation unit 31 and a selector 32. When registering the backup destination address in the history table 118, the hash value calculation unit 31 calculates the index 23 using the write destination address as an input, and the history entry 21 indicated by the calculated index 23 in the history table 118. The write destination address is stored in.

  Further, when searching for the write destination address from the history table 118, the hash value calculation unit 31 calculates the index 23 with the write destination address to be searched as an input, and the history indicated by the calculated index 23 in the history table 118. It is checked whether or not the write destination address stored in the entry 21 matches the address being searched. If they match, the search is successful, and if they do not match, the search fails. Finally, the process of clearing the history table 118 is the same as in the case of a simple table format. If a valid bit is provided, the history table 118 can be cleared by clearing the valid bit.

  For example, when the number of history entries 21 included in the history table 118 is N as illustrated in FIG. 7, the hash value calculation unit 31 sets the remainder obtained by dividing the write destination address by N as the output hash value. And so on. However, it only needs to satisfy the properties as a hash value and does not depend on a specific calculation method.

(Second Embodiment)
The second embodiment will be described with reference to the attached drawings, centering on the point that data read from the nonvolatile main storage unit is recorded in a history table, which is a different part from the first embodiment. The other parts are the same as those in the first embodiment, so the description is omitted with reference to the above description. In the present embodiment, the configuration of the history table 118 and backup processing will be described in detail.

  FIG. 8 is an explanatory diagram of an example of a data configuration of the history table 118 according to the second embodiment. The history table 118 according to the present embodiment stores the address of the nonvolatile main storage unit 120, the read data, the valid bit, and the backed up flag in association with each other.

  The history table management unit 117 operates the history table 118 according to an instruction from the backup control unit 112 in the same manner as described above. Specifically, the history table management unit 117 registers the address and data of the nonvolatile main storage unit 120 that has read the data in the history table 118. The history table management unit 117 searches the history table 118 for the address of the nonvolatile main storage unit 120 to be backed up. The history table management unit 117 clears the history table 118. In the present embodiment, unlike the first embodiment described above, in addition to the address of the backed-up nonvolatile main storage unit 120, the read data of the nonvolatile main storage unit 120 is also recorded.

  9A and 9B are flowcharts illustrating the backup processing procedure performed by the backup unit 110. The backup unit 110 operates by receiving an access from the bus 180. The processing for registering data in the history table 118 and the processing for searching for data from the history table 118 performed in the present embodiment are different from the first embodiment in processing content and processing position.

  First, the main memory access control unit 111 detects an access from the bus 180 (step S901). The main memory access control unit 111 determines whether or not the detected access is a read access to the nonvolatile main storage unit 120 (step S902).

  If it is determined that the detected access is a read access to the nonvolatile main storage unit 120 (step S902: Yes), the main memory access control unit 111 reads data from the nonvolatile main storage unit 120 (step S903), The read data is returned to the bus 180. The backup control unit 112 determines whether the backup process is valid (step S904). If it is determined that the backup process is not valid (step S904: No), the process ends. If it is determined that the backup processing is valid (step S904: Yes), the backup control unit 112 sends the read destination address and the read data to the history table management unit 117 and instructs to record in the history table 118. (Step S905). The history table management unit 117 determines a storage area in the history table 118 and records a read destination address and data (step S906). Since the backup process is not performed in the read process, the history table management unit 117 sets the backed-up flag corresponding to the recorded history entry to OFF, which is a value that is not backed up (step S907).

  If it is determined that the detected access is not a read access to the nonvolatile main storage unit 120 (step S902: No), the main storage access control unit 111 sends the access to the backup control unit 112, and the backup control unit 112 accesses Is received (step S908). The backup control unit 112 determines whether the received access is a write access to the nonvolatile main storage unit 120 (step S909).

  If it is determined that the received access is a write access to the nonvolatile main storage unit 120 (step S909: Yes), the backup control unit 112 determines whether the backup process is valid (step S910). When it is determined that the backup process is not valid (step S910: No), the backup process is not performed and the process proceeds to step S922. When it is determined that the backup process is valid (step S910: Yes), the backup control unit 112 sends the write destination address to the history table management unit 117 to search whether it is recorded (step S911). . The history table management unit 117 searches whether the received write destination address is recorded in the history table 118 (step S912). The history table management unit 117 determines whether or not the received write destination address is recorded in the history table 118 (step S913).

  When it is determined that the received write destination address is not recorded in the history table 118 (step S913: No), that is, when it is the first access to the address after the latest commit process, the backup control unit 112 stores the main memory. A read request is sent to the access control unit 111, and the main memory access control unit 111 reads the original data stored in the storage area indicated by the write destination address from the nonvolatile main storage unit 120 (step S914). Next, the backup data generation unit 114 generates backup data from the write destination address and the original data (step S915).

  The backup control unit 112 reads the address on the nonvolatile backup memory 130 into which the generated backup data is written from the write pointer holding unit 116 (step S916). The backup memory access control unit 113 writes the backup data in the storage area indicated by the read address on the nonvolatile backup memory 130 (step S917). The backup control unit 112 updates the value of the pointer held in the write pointer holding unit 116 (step S918). The backup control unit 112 sends the write destination address and the read data to the history table management unit 117, and instructs recording (step S919). The history table management unit 117 determines and records a storage area in the history table 118 according to the structure of the managed history table (step S920). The history table management unit 117 sets the backup completed flag corresponding to the recorded history entry to ON, which is the value of the backup (step S921), because the backup process was performed in this writing process. The main memory access control unit 111 writes data to the write destination address of the nonvolatile main storage unit 120 (step S922).

  In step S913, when it is determined that the received write destination address is recorded in the history table 118 (step S913: Yes), the history table management unit 117 determines whether backup has been completed (step S923). . When the history table management unit 117 determines that the received write destination address has been backed up (step S923: Yes), the history table management unit 117 proceeds to step S922.

  When the history table management unit 117 determines that the backup has not been completed (step S923: No), the history table management unit 117 reads an address and data from the history table 118 (step S924). The history table management unit 117 sets the backed up flag corresponding to the recorded history entry to ON (step S925). Thereafter, the process proceeds to step S915. As a result, the address and data recorded in the history table by the read process can be used, so that the data stored in the storage area indicated by the write destination address can be stored without accessing the nonvolatile main storage unit 120. Can be obtained, and the processing time can be shortened.

  If it is determined in step S909 that the received access is not a write access to the nonvolatile main storage unit 120 (step S909: No), the backup control unit 112 determines whether the access is to the nonvolatile backup memory 130 or not. Judgment is made (step S926). When it is determined that the access is to the non-volatile backup memory 130 (step S926: Yes), the backup control unit 112 sends an access destination address to the backup memory access control unit 113, and the backup memory access control unit 113 performs non-volatile backup. Data in the memory 130 is read or written in the nonvolatile backup memory 130 (step S927).

  When determining that the access is not to the nonvolatile backup memory 130 (step S926: No), the backup control unit 112 determines whether the access is to the write pointer holding unit 116 (step S928). When it is determined that the access is to the write pointer holding unit 116 (step S928: Yes), the backup control unit 112 reads or updates the value of the write pointer holding unit 116 (step S929). If it is determined that the access is not to the write pointer holding unit 116 (step S928: No), the process is terminated.

  As described above, in this embodiment, not only the address of the backed-up non-volatile main storage unit 120 but also the address at which data is read from the non-volatile main storage unit 120 and the read data are recorded, thereby The access to the data at the same address in the nonvolatile main storage unit 120 can be reduced. Thereby, the performance fall of PC100 can be reduced.

  It should be noted that the present invention is not limited to the above-described embodiment as it is, and can be embodied by modifying the constituent elements without departing from the scope of the invention in the implementation stage. In addition, various inventions can be formed by appropriately combining a plurality of constituent elements disclosed in the above embodiments. For example, some components may be deleted from all the components shown in the embodiment. Furthermore, constituent elements over different embodiments may be appropriately combined.

It is a block diagram which shows the structure of PC concerning 1st Embodiment. It is explanatory drawing which shows an example of a data structure of a log | history table. It is a flowchart which shows the backup processing procedure which a backup part performs. It is a flowchart which shows the backup processing procedure which a backup part performs. It is a flowchart which shows the commit processing procedure which a commit part performs. It is a flowchart which shows the rollback process sequence which a rollback part performs. It is explanatory drawing which shows an example of a data structure of a log | history table. It is explanatory drawing which shows an example of a data structure of a log | history table. It is explanatory drawing which shows an example in the data structure of the log | history table concerning 2nd Embodiment. It is a flowchart which shows the backup processing procedure which a backup part performs. It is a flowchart which shows the backup processing procedure which a backup part performs.

Explanation of symbols

100 Personal computer (PC)
DESCRIPTION OF SYMBOLS 110 Backup part 111 Main memory access control part 112 Backup control part 113 Backup memory access control part 114 Backup data generation part 115 Backup memory head address holding part 116 Write pointer holding part 117 History table management part 118 History table 120 Nonvolatile main memory Unit 130 nonvolatile backup memory 140 commit unit 150 rollback unit 160 processor unit 170 control software 180 bus 200 power supply

Claims (10)

A non-volatile main memory unit;
A non-volatile backup storage unit that stores a copy of the data stored in the main storage unit;
A memory specified by the address when a write access including a data to be written to the main storage unit and a write destination address for designating a write destination of the data to be written sent from the processor unit is detected. Reading means for reading data already stored in the area;
Generating means for generating backup data including the data read by the reading means and the address included in the write access;
History information storage means for storing the address included in the backup data to be written to the backup storage unit;
History information search means for searching for an address that matches the address of the write destination from among the addresses stored in the history information storage means;
If the history information storage means does not store an address that matches the address of the write destination, a first writing means for writing the backup data to the backup storage unit;
Second writing means for writing the write destination address in the history information storage means when the backup data is written in the backup storage unit;
A third writing means for writing the data to be written into the storage area specified by the address of the main storage unit after the writing by the second writing means is completed;
An information processing apparatus comprising: The first writing unit writes the backup data into the backup storage unit when an address that matches the address of the writing destination is stored in the history information storage unit;
After it is determined that the address that matches the address of the write destination is stored in the history information storage unit, the third writing unit stores the address in the storage area specified by the address of the main storage unit. The information processing apparatus according to claim 1, wherein data to be written is written. Commit means for setting the backup data newly written to the backup storage unit from the current time by the first writing means as read permission;
When it is instructed to rewrite the storage contents of the main storage unit, the backup data stored in the backup storage unit is read from the data and the address included in the backup data set as read permission. Extracting means for extracting;
And a fourth writing means for writing the data into the storage area on the main memory indicated by the address for each of the data and the address extracted from the backup data. The information processing apparatus according to claim 1 or 2. The reading means reads data stored in a storage area indicated by the address from the main storage unit when detecting a read access including a read destination address designating a read destination for the main storage unit;
The history information storage means stores the read data, the address, and backed up information indicating whether or not the backup data is stored in the backup storage unit in association with each other,
The history information search means searches for an address that matches the address of the write destination from the history information storage means when detecting the write access,
The first writing means indicates that the history information storage means stores an address that matches the address of the writing destination, and that the backed up information has written the backup data to the backup storage unit 4. The information processing apparatus according to claim 1, wherein the backup data is not written to the backup storage unit.   The first writing means stores an address that matches the address of the writing destination in the history information storage means, and that the backed up information has not written the backup data in the backup storage unit. 5. The information processing apparatus according to claim 4, wherein in the case shown, the backup data is written to the backup storage unit.   The first writing unit writes the backup data into the backup storage unit when the history information storage unit does not store an address that matches the address of the write destination, and the history information storage unit 6. The information processing apparatus according to claim 5, wherein the read data, the write destination address, and the backed up information are written. The history information storage means stores flag information indicating whether the storage area storing the address is valid in association with the address,
4. The history information searching means searches the history information storage means for an address that indicates that the flag information is valid and matches the address of the write destination. The information processing apparatus according to item 1. The history information storage means stores flag information indicating whether or not the storage area storing the data and the address is valid in association with the data and the address;
8. The history information search means searches the history information storage means for an address indicating that the flag information is valid and coincides with the address of the write destination. 8. The information processing apparatus according to item 1.   The first writing means is connected between the processor unit and the main storage unit, receives the write access from the processor unit, and sends the received write access to the main storage unit. The information processing apparatus according to claim 1, wherein the information processing apparatus is characterized. A data recovery method for an information processing apparatus comprising: a nonvolatile main storage unit; and a nonvolatile backup storage unit that stores a copy of data stored in the main storage unit,
A memory specified by the address when a write access including a data to be written to the main storage unit and a write destination address for designating a write destination of the data to be written sent from the processor unit is detected. A reading step of reading data already stored in the area;
Generating the backup data including the data read in the reading step and the address included in the write access;
A history information search step for searching for an address that matches the address of the write destination from among the addresses stored in the history information storage means for storing the address included in the backup data to be written to the backup storage unit;
A first writing step of writing the backup data to the backup storage unit when the history information storage means does not store an address that matches the address of the write destination;
A second writing step of writing the write destination address in the history information storage means when the backup data is written in the backup storage unit;
A third writing step of writing the data to be written into the storage area specified by the address of the main storage unit after the writing by the second writing step is completed;
A data recovery method characterized by.

Images