JP3346670B2 - Central processing unit switching method and switching system - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a switching method and a switching system for a central processing unit in a system having a duplex configuration of an active system and a standby system, such as an electronic exchange and an information processing system, which require high reliability. In particular, the present invention relates to a high-speed and reliable switching method and a switching system for a central processing unit necessary for specifying a failed portion when a device failure occurs.

[0002]

2. Description of the Related Art Conventionally, a rotation method has been well known as a method of specifying a failed part when a failure is detected in a system having the above-mentioned duplex configuration. The rotation method is a method in which various device configuration patterns are sequentially reconfigured for failure detection, and a device that cannot be finally incorporated into a normal system configuration is specified as a failure site.

FIG. 1 is a diagram for explaining a method of specifying a failed part by a rotation method. In a normal system configuration during operation, the central processing unit # 0, the input / output device # 0, and the input / output device # 1 are active devices, and the central processing unit # 1 is in a thermal reserve state. Here, when a failure is detected, first, for example, the central processing unit # 0 is switched to the thermal standby state, and the central processing unit # 1 that was in the thermal standby state is switched to the active unit, and the abnormal time system configuration is changed. # 1 is configured. In the abnormal-time system configuration # 1, if a failure occurs within a certain period of time, a transition is made to another abnormal-time system configuration # 2. For example, the input / output device # 1 in the working state is switched to the non-working state. Even in the abnormal time system configuration # 2, if a failure occurs within a predetermined time, for example, the input / output device #
0 is switched from the working state to the non-working state, and the input / output device #
1 is switched from the non-working state to the working state, and an abnormal system configuration # 3 is formed. In the abnormal system configuration # 3, if no failure occurs after a certain period of time, the input / output device # 0, which is a non-active device, is regarded as a failed device in the system configuration at this time, and the system configuration Is called a quasi-normal system configuration.

In order to specify a faulty device more accurately and quickly in the rotation method, it is necessary to switch the central processing unit at a high speed without being disturbed by the faulty device. Conventionally, two methods are known as a method of switching a central processing unit in an electronic exchange or an information processing system that requires high reliability. The first method is to autonomously decrement the timer in the circuit, determine that the software is abnormal when the timer underflows, and use a circuit for switching between the active system and the standby system of the central processing unit. This is a fixed point restart method for restarting from a fixed point. Switching of the system by the above-mentioned circuit means that an abnormality has occurred in the running of the software. Therefore, in the new working system after the switching, the program on the memory is reinitialized.

A second conventional method for switching the central processing unit is a method of resuming an interruption point without going through software initialization. According to the interruption point resuming method, the register information and the like in the central processing unit are passed from the active central processing unit to the standby central processing unit, and the standby central processing unit is made active,
Switching processing such as making the current central processing unit a standby system is performed by software. In such an interruption point resumption method,
The above-described switching process controlled by software drives a suspected failure device (a device in which a failure may have occurred) such as a bus crossing device that connects an active bus and a standby bus. Since it needs to be performed, there is a problem from the viewpoint of the certainty of the processing. Therefore, in order to solve such a problem of certainty, a dedicated crossover circuit for switching the attributes of the working / standby system necessary for switching the interruption point and taking over the interruption point address is provided between the central processing units. Have been.

[0006]

The above fixed point resuming method has a disadvantage that the service is interrupted for a long time at the time of switching since all initial settings of software are required. On the other hand, in the above interruption point resuming method, since the number of hardware components of the central processing unit increases by providing the dedicated crossover circuit, the probability of occurrence of a failure of the central processing unit itself increases, and the reliability as a system is impaired.

The present invention has been made in view of the above-mentioned problems of the conventional method, and has as its object to provide a reliable switching method of a central processing unit that requires a short service interruption time and does not require a dedicated cross circuit. Another object of the present invention is to provide a more reliable central processing unit switching method.

It is another object of the present invention to provide a central processing unit switching system that can reliably switch central processing units in a short time without requiring a dedicated cross circuit.

[0009]

FIG. 2 is a schematic block diagram of a switching system for explaining the principle of the present invention. The switching method of the central processing unit according to the present invention includes a first central processing unit 10 of an active system and a second central processing unit 10 ′ of a standby system.
The first central processing unit 10 and the second central processing unit 1
0 'and an information transmission unit 3 for transmitting information between the first central processing unit 10 and the second central processing unit 10'.
0: a switching method in a duplex system comprising a switching circuit 20 that switches the first central processing unit to the standby system under software control and switches the second central processing unit to the active system.

FIG. 3 is a diagram illustrating the principle of the present invention. In the switching method of the central processing unit of the present invention, the step of interrupting the running software in the first central processing unit 10 and starting predetermined switching software (step 10).
And a register information saving step of saving register information of the first central processing unit 10 to the information transmitting unit 30 (step 2).
0), the second central processing unit is switched to the active system, and the first
Driving the switching circuit 20 so as to switch the central processing unit to the standby system (step 30); and transmitting the saved register information from the information transmitting unit 30 to the second central processing unit so as to continuously start the interrupted software. Device 10 '
(Step 40).

In another aspect of the switching method of the central processing unit according to the present invention, the register information saving step is for transferring the first identification information of the first central processing unit 10 to the second central processing unit 10 '. The first identification information is transmitted to the information transmitting unit 30
Further comprising the step of saving the register information from the information transmitting unit 30 to the second central processing unit 1.
Before loading the data to 0 ', the first identification information inherited by the information transmission unit 30 is compared with the second identification information of the second central processing unit 10', and the first identification information and the second identification information are compared. If the information does not match, the register information taken over by the information transmitting unit 30 is loaded from the information transmitting unit 30 to the second central processing unit 10 'so as to continuously start the software, and the first identification information is loaded. When the ID and the second identification information match, error occurrence information indicating that a software error has occurred at the time of switching is set, and the register information is returned from the information transmission unit 30.

Further, in another aspect of the switching method of the central processing unit according to the present invention, the register information saving step includes:
The second central processing unit 10 'further includes a step of saving the switching information to the information transmitting unit 30 so that the switching information indicating whether or not the central processing unit is being switched is transferred to the second central processing unit 10'. Information transmission unit 3
Before loading from 0 to the second central processing unit 10 ′, it is determined whether or not the central processing unit is being switched based on the switching information passed to the information transmitting unit 30, and if the switching is being performed, the software is continued. The register information passed to the information transmitting unit 30 is loaded from the information transmitting unit 30 to the second central processing unit 10 ′ so as to be activated,
When the switching is not being performed, the initial setting of the entire software including the above software is performed.

[0013]

According to the switching method of the central processing unit of the present invention, the running software is interrupted and the switching software of the predetermined central processing unit is activated (step 1).
0), the register information of the active central processing unit 10 indicating the state of the active central processing unit 10 including the interrupted running software state is saved in the information transmitting unit 30 by the switching software (step 2).
0) Since the standby central processing unit 10 'can load the register information from the information transmitting unit 30 (step 40), the interruption point address can be taken over without providing a dedicated cross circuit. Further, since the switching software drives the switching circuit 20 (step 30), there is no need to use a dedicated crossing circuit for switching the attributes of the working / standby system.

In another aspect, the switching method of the central processing unit according to the present invention includes first identification information representing the old active central processing unit 10 and second identification information representing the new active central processing unit 10 '. Compare with information. As a result of the comparison, when the first and second identification information match, it is detected that the switching circuit 20 has been operated again due to another factor during the switching processing by the switching software, and the interruption is performed. The software that has been interrupted is notified of the fact and appropriate processing can be performed on the software that has been interrupted, thereby improving the reliability of the switching processing.

Further, in the switching method of the central processing unit of the present invention, the switching information indicating whether or not the switching is being performed by the switching software in the old active central processing unit 10 is saved in the information transmitting unit 30, and the new active processing is performed. It is determined whether the switching of the central processing unit is being performed based on the switching information taken over by the central processing unit 10 of the system, the case where the switching circuit 20 is driven by a factor other than the switching software is detected, and the processing is interrupted. Since the entire software including the software can be initialized, the reliability of the switching process is improved.

[0016]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A switching method and a switching system of a central processing unit according to the present invention will be described below in detail with reference to the drawings. FIG. 4 is a configuration diagram of the switching system of the central processing unit of the present invention. The switching system includes an active first central processing unit 10 and a standby second central processing unit 10 ′.
The first central processing unit 10 is switched to the standby system by detecting an abnormality in the running state of the software, and the second central processing unit 1
A switching circuit 20 for switching 0 'to the active system.

The switching system is connected to a first central processing unit 10 and a second central processing unit 10 ',
The storage unit 30 further stores information passed from the central processing unit 10 to the second central processing unit 10 ′. The first central processing unit 10 saves the register information, for example, the register information of the LSI in the central processing unit 10 into the storage unit 30, switches the first central processing unit 10 to the standby system, and It has an evacuation / circuit driving unit 12 that drives a switching circuit 20 that switches 10 ′ to the active system. The second central processing unit 10 ′ switched from the standby system to the active system by the switching circuit 20 has a load unit 14 for loading the saved register information into the storage unit 30.

Hereinafter, the storage unit 30 will be described in more detail. As shown in FIG. 4, the storage unit 30 includes a first memory 32 accessed by the first central processing unit 10, a second memory 34 accessed by the second central processing unit 10 ′,
A writing circuit 36 for writing data written to the first memory 32 by the first central processing unit 10 to the second memory 34; The writing circuit 36 is preferably a circuit having a function of always keeping the first memory 32 and the second memory the same. For example, the contents of the first memory 32, which is the active system, are in a thermal reserve state. A circuit having a function of copying to the second memory 36 and a function of writing data from the processor to the active memory 32 to the thermal standby state memory 36 (hereinafter, referred to as a memory crossing circuit).

The switching circuit 20 is controlled by software to switch the first central control unit 10 from the active system to the standby system and to switch the second central control unit 10 'from the standby system to the active system. For example, a circuit that switches when there is no access from software within a predetermined time is preferable. The switching circuit 20 has, in particular, a timer for autonomously decrementing, detects an abnormality in the running state of software when the timer underflows, switches the active device to the standby system, and switches the standby device to the active system. It is preferable to use a circuit (hereinafter, referred to as an EMA circuit) that starts the central processing unit that can be switched to the active system from a fixed address of the memory.

In the above description, the first central processing unit 10
Is described as comprising the save / circuit drive unit 12 and the second central processing unit 10 'is composed of the load unit 14, but the first central processing unit 10 and the second central processing unit 1
0 'may have both the save / circuit drive unit 12 and the load unit 14. In the configuration in this case, first,
The second central processing unit 10 'is an active system, and the first central processing unit 10 is a standby system.

FIG. 5 is a hardware configuration diagram of a central processing unit switching system according to an embodiment of the present invention. The active system side device includes a central processing unit 10 and an input / output device 50.
And a bus 52 to which the central processing unit 10 and the input / output device 50 are connected. The standby system side device includes a central processing unit 10 ′, an input / output device 50 ′, and a central processing unit 10 ′. And a bus 52 'to which the device 50' is connected. The central processing unit 10 includes a microprocessor (M
PU) 100, a memory (MM) 102, and a memory access circuit (MA) connected between the microprocessor 100 and the memory 102 for accessing the memory 102.
C) 104, a memory crossover circuit (MXC) 106 for connecting the memory 102 and the standby system side device, and a switching circuit connected between the microprocessor 100 and the standby side device for switching between the active system and the standby system ( EMA) 108;
And a bus interface control circuit (BIC) 110 that performs an interface with the bus 52. Here, the memory crossing circuit is, as described above, a function of copying the contents of the active system memory to the thermal standby state memory so that the memory contents of both the active system and the standby system are always the same. This is a circuit having a function of writing to the thermal standby state memory in response to a memory write from the memory. Similarly,
The central processing unit 10 'is a microprocessor (MPU)
100 ′, a memory (MM) 102 ′, a memory access circuit (MAC) 104 ′, and a memory cross circuit (MX)
C) 106 ′, a switching circuit (EMA) 108 ′,
And a bus interface control circuit (BIC) 110 '.

Hereinafter, a method of resuming an interruption point of an application program according to an embodiment of the present invention will be described with reference to FIG. An operating system (OS) that provides hardware control means to an application program is incorporated in the central processing units 10 and 10 ′. The application program requests the switching of the central processing unit by activating the central processing unit switching system call. Such a system call, which will be described in detail below, switches the central processing unit and restarts the application program based on the interruption point restart method. That is, the system call changes the state of the central processing unit before the occurrence of the system call from the old active central processing unit 10 to the new active central processing unit 1.
By taking over to 0 'and returning to the instruction immediately after the start of the system call, the interruption point of the application program is resumed.

Next, the operation of the switching system call in one embodiment of the present invention will be described with reference to the flowchart of the system call shown in FIG. First, the register information is copied onto the memory to take over the information in the central processing unit (step 100). By using the above memory crossing circuits 106 and 106 ', the standby memory 102' always holds the same contents as the active memory 102. That is, a write access to the active memory 102 'causes a write to the thermal standby memory 102'. Before switching the system, the active central processing unit 10 saves the register information in the central processing unit 10 in the active memory 102, so that the register information is written into the thermal standby memory 102 ′, The same content as 102 is reflected. Here, saving of the register information of the LSI in the central processing unit 10 is performed by the MPU.
, Except for the program counter (PC), the result of reading the register is saved in a specific area on the memory. On the other hand, for the program counter,
Save the start address of the program that normally ends the switching system call.

Next, the EMA circuit 108 is started (step 110). The activation of the EMA circuit 108 is realized by performing a loop process (spin wait) without resetting the timer in the EMA circuit 108 for a fixed time.
If the EMA circuit is not activated after a certain period of time,
That is, when the program immediately after the loop processing is executed, it is preferable that the system call program be terminated with an error.

The activation of the EMA circuits 108, 108 'causes the central processing unit to be switched by hardware and firmware (step 120). At this point, the dual writing of the memory, that is, the writing to the standby memory by the write access to the active memory is stopped (step 130). With the switching of the central processing unit, the new active central processing unit 10 'switched from the standby system to the active system is started from the fixed address of the memory 102' and starts executing (step 140).

Finally, in the new active central processing unit 10 ', the register information of the old active central processing unit 10 saved in step 100 is restored (step 15).
0). That is, the register information saved in the memory 102 'is loaded into the corresponding register of the LSI in the new active central processing unit 10'. After the system is switched, the contents of the old spare system memory are loaded into the register in the central processing unit, whereby the information in the central processing unit is taken over. The double writing of the memory suspended in the above step 130 is restarted by reloading each register of the LSI of the memory crossing circuit 106 '. With respect to the program counter, the normal end program of the switching system call whose head address was saved in step 130 is started, and the program returns to the application program which issued the switching system call.

As described above, according to one embodiment of the present invention, a running application program is interrupted and a predetermined central processing unit switching system call is activated without using a dedicated crossover circuit. By the switching system call, the register information of the active central processing unit 10 indicating the state of the active central processing unit 10 including the state of the interrupted application program is taken over by the new active central processing unit 10 'after the switching. Then, the suspended application program is resumed.

FIG. 8 is a flowchart showing an operation of a switching system call according to another embodiment of the present invention. The switching system call according to another embodiment of the present invention determines whether the central processing unit is switched based on the activation of the switching system call, and further determines whether the switching of the central processing unit by the switching system call is successful. Determine whether In the figure, the steps denoted by the same reference numbers as the steps in the operation flowchart shown in FIG. 7 perform the same functions as the steps in FIG.

First, the register information is copied onto the memory to take over the information in the central processing unit (step 100). Next, the identification information of the active system central processing unit 10, that is, information indicating which device is the active system, for example, the device number, is saved at a specific address on the memory 102 (step 102). The same contents as the information saved in the memory 102 are stored in the memory 1 as described above.
02 '.

The system switching data indicating that the switching system call has been activated is set to a specific address on the memory 102 (step 104). The data during system switching is reflected in the memory 102 'as described above.
Next, the EMA circuit 108 is started by performing a loop process (spin wait) without resetting the timer in the EMA circuit 108 for a predetermined time (step 110).

The activation of the EMA circuits 108, 108 'causes the central processing unit to be switched by hardware and firmware (step 120). At this point, the double writing of the memory is stopped (step 13).
0). By switching the central processing unit, the new active central processing unit 1 is switched from the standby system to the active system.
0 'is activated from the fixed address of the memory 102' and starts executing (step 140).

The system switching data saved in step 104 is read, and it is determined whether the switching is due to a central processing unit switching by a switching system call or a switching due to an abnormality during running of the software (step 142). If it is caused by the switching system call, the above data during system switching is reset to the non-system switching state, and the process proceeds to step 146. If the switching is due to a software error, the initialization of the operating system is started (fixed point restart), and the processing of the switching system call is terminated (step 144).

If the switching of the central unit is caused by the switching system call, the unit number of the old active central processing unit saved in step 102 and the unit number of the new active central processing unit after the switching are changed. A comparison is made (step 146). If the old active central processing unit and the new active central processing unit after switching have the same unit number,
Proceed to step 144. If they do not match, the process proceeds to step 150.

When the device numbers of the old active central processing unit and the new active central processing unit match, during the processing from the above-described activation of the EMA circuit (step 110) to the determination of system switching (step 142). Then, the EMA circuit is activated, the central processing unit is switched, and it is determined that switching of the central processing unit by the switching system call has not been performed normally, and error information indicating that the switching system call has ended in error is set. (Step 148),
Restore the register information of the old central processing unit (Step 1)
50).

If the unit numbers of the old active central processing unit and the new active central processing unit do not match, it is determined that the switching of the central processing unit by the switching system call has been normally performed, and the register of the old central processing unit is determined. The information is restored (step 150). According to another embodiment of the present invention, the device number representing the old active central processing unit is compared with the device number representing the new active central processing unit, and during the switching process by the switching system call, , EMA
Since it is possible to determine whether or not the circuit has been operated again due to another factor, and to notify the application program of the determination result, the reliability of the switching process based on the switching system call is improved.

Further, information indicating that switching is being performed by the switching system call is transferred from the old active central processing unit to the new active central processing unit.
Due to factors other than the above switching system call,
Since the case where the MA circuit is driven can be detected and the entire software including the interrupted application program can be initialized, the reliability of the switching process is improved.

In the other embodiment of the present invention, the operation of the switching system call is described for the case where the register information of the active central processing unit, the device number, and the data during the system switching are saved in the memory in this order. However, the order of saving the register information, the device number, and the data during system switching is not limited to the order shown in the example.

[0038]

As described above, according to the present invention, it is not necessary to add a dedicated crossover circuit by switching the central processing unit by using a switching circuit for detecting an abnormality during running of the software. When switching is performed due to a failure of a device other than the device, the possibility of failure in the switching process is reduced, and the reliability of the switching is improved. Therefore, the faulty device can be specified by the high-speed rotation method.

[Brief description of the drawings]

FIG. 1 is an explanatory diagram of a rotation method for specifying a faulty device.

FIG. 2 is a schematic configuration diagram of a switching system for explaining the principle of the present invention.

FIG. 3 is a diagram illustrating the principle of the present invention.

FIG. 4 is a configuration diagram of a switching system of a central processing unit of the present invention.

FIG. 5 is a configuration diagram of a switching system of a central processing unit according to an embodiment of the present invention.

FIG. 6 is a diagram for explaining the concept of the interruption point resuming method according to one embodiment of the present invention.

FIG. 7 is an operation flowchart of a switching system call according to an embodiment of the present invention.

FIG. 8 is an operation flowchart of a switching system call according to another embodiment of the present invention.

[Explanation of symbols]

 10, 10 'central processing unit 12 save / circuit drive unit 14 load unit 20 switching circuit 30 information transmission unit 32, 36, 102, 102' memory 34 write circuit 50, 50 'input / output device 52, 52' bus 100, 100 'Microprocessor 104, 104' Memory access circuit 106, 106 'Memory crossing circuit 108, 108' EMA circuit 110, 110 'Bus interface circuit

Continuation of the front page (58) Field surveyed (Int.Cl. ⁷ , DB name) G06F 11/16-11/20 G06F 15/16-15/177 H04M 3/22

Claims (4)

(57) [Claims] A first central processing unit of an active system, a second central processing unit of a standby system, an information transmission unit between the first central processing unit and the second central processing unit, and software And a switching circuit for switching the first central processing unit to the standby system under control of the first central processing unit and the second central processing unit to the active system. Interrupting running software in one central processing unit to start predetermined switching software; register information saving step of saving register information of the first central processing unit to the information transmitting unit; Driving the switching circuit to switch the second central processing unit to the active system and switch the first central processing unit to the standby system; and continuously activating the interrupted software. Loading the saved register information from the information transmission unit to the second central processing unit. The register information saving stage includes a first identification of the first central processing unit. Evacuating information to the information transmitting unit, wherein the continuation start-up step includes the step of taking over the register information taken over by the information transmitting unit before loading the register information from the information transmitting unit to the second central processing unit. The first identification information is compared with the second identification information of the second central processing unit, and if the first identification information does not match the second identification information, the software is continuously started. Loading the register information passed to the information transmitting unit from the information transmitting unit to the second central processing unit so that the first identification information and the second identification information match. When a software error occurs during switching A method of switching a central processing unit, comprising setting error occurrence information to be represented and returning the register information from the information transmitting unit. 2. An active first central processing unit, a standby second central processing unit, an information transmission unit between the first central processing unit and the second central processing unit, and software And a switching circuit for switching the first central processing unit to the standby system under control of the first central processing unit and the second central processing unit to the active system. Interrupting running software in one central processing unit to start predetermined switching software; register information saving step of saving register information of the first central processing unit to the information transmitting unit; Driving the switching circuit to switch the second central processing unit to the active system and switch the first central processing unit to the standby system; and continuously activating the interrupted software. Loading the saved register information from the information transmitting unit to the second central processing unit. The register information saving stage indicates whether the central processing unit is being switched. Evacuating switching information to the information transmitting unit, wherein the continuation start-up step is performed by the information transmitting unit before the register information is loaded from the information transmitting unit to the second central processing unit. It is determined whether or not the central processing unit is being switched based on the switching information, and when the switching is being performed, the register information passed to the information transmitting unit is transmitted from the information transmitting unit to the software so as to continuously start the software. A method for switching a central processing unit, wherein the method is loaded into a second central processing unit, and initializes the entire software including the software when the switching is not being performed. 3. The method according to claim 1, wherein the register information saving step includes the step of:
Further comprising the step of saving the first identification information of the central processing unit to the information transmitting unit, wherein the continuous start-up step is performed before the register information is loaded from the information transmitting unit to the second central processing unit. The first identification information passed to the information transmission unit is compared with the second identification information of the second central processing unit, and the first identification information and the second identification information match. Loading the register information taken over by the information transmitting unit from the information transmitting unit to the second central processing unit so that the software is continuously activated when there is no such information; 3. The central processing unit according to claim 2, wherein when the identification information matches, error occurrence information indicating that a software error has occurred at the time of switching is set, and the register information is returned from the information transmission unit. Method of switching. 4. A central processing unit, an input / output unit, and a system bus, which are duplicated in a working system and a standby system, wherein each central processing unit autonomously decrements a timer and a software error occurs when a timer underflows. And a switching circuit for switching between the active system and the standby system. The system is activated when a failure of the active system central processing unit is detected, and operates the switching circuit to switch between the active system and the standby system. The system further includes a switching unit, wherein the switching unit saves the active system memory, the standby system memory, and the register information of the active system central processing unit to the active system memory, and stores the saved register information in the active system memory. Means for taking over to the memory of the standby system, and causing a timer underflow by not resetting the timer of the switching circuit for a fixed time to cause the switching circuit to operate. And a means for loading the register information taken over by the memory of the standby system into a new active system central processing unit after switching. The switching unit further comprises: a current system before switching. Means for passing the device number of the central processing unit to the new active central processing unit after the switching, wherein the new active central processing unit after the switching includes the device number of the active central processing unit before the switching and the unit number after the switching. Means for notifying the active central processing unit before the switchover of the occurrence of an error when the device numbers of the active central processing unit are compared with each other. A system wherein software running on the device is continuously activated on a new active central processing unit.