JP3903678B2 - Computer system dump processing method - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to acquisition of failure information of a computer system in which a failure has occurred, and restart of the computer system in which a failure has occurred.
[0002]
[Prior art]
Conventionally, the operating system checks data consistency during system operation. When data inconsistency occurs due to some kind of failure, panic processing is executed to urgently stop the computer system.
[0003]
In the panic process, the memory contents at the time of the failure are stored in an auxiliary storage device such as a hard disk so that the system administrator can find out the cause of the failure later. The process of saving the memory contents in the auxiliary storage device is called a dump process, and after completion of the dump process, the computer system is restarted.
[0004]
In the dump processing, there are cases where all the memory of the computer system is dumped (all dumps) and a part of the memory is dumped. When dumping all of the memory, the time required for dump processing is longer than when dumping a portion of memory. Japanese Patent Application Laid-Open No. 64-25255 discloses a technique for selecting a part of memory and dumping only the selected memory in order to shorten the dump processing time.
[0005]
In the system restart process, the memory content stored in the auxiliary storage device such as a hard disk is converted into a file format that allows the system administrator to analyze the cause of the failure. In order to save the capacity of the auxiliary storage device, a file compression process may be executed at the time of file format conversion.
[0006]
A series of processing until a computer system is restarted after a failure has occurred in the computer system is described in “Panic! UNIX System Crash Tracking and Countermeasures” (1997), page 27 of ISBN 4-7561-1912-3. See page 33.
[0007]
[Problems to be solved by the invention]
In recent years, the amount of memory installed in a computer system has increased, and the time required to restart after a failure has occurred in the computer system has increased.
[0008]
As described above, by selecting and dumping a part of memory, the amount of memory that needs to be dumped is smaller than in the case of full dumping. However, it may be difficult to identify the memory range associated with the failure before analyzing the memory contents at the time of the failure. Furthermore, even if a part of the memory is selected and dumped, the restart process cannot be executed until all the selected memories are dumped.
[0009]
Further, as described above, in order to save the use capacity of the auxiliary storage device, when file compression processing is executed at the time of restart, the time required for restart further increases. The computer system cannot provide the service to the user after the failure occurs in the computer system until the restart is finished. For this reason, there has been a problem that the operating time of the computer system during a certain period of time decreases and the utilization efficiency decreases.
[0010]
The object of the present invention is to shorten the time from the occurrence of a failure in the computer system to the end of the restart, improve the operating rate of the computer system, and determine the memory contents necessary for analyzing the cause of the failure. It is to provide to etc.
[0011]
[Means for Solving the Problems]
In order to achieve the above object, the present invention provides a memory range specification information indicating an address or memory amount for specifying an area of a memory to be initialized, and a memory indicating a numerical value for dividing the memory into a plurality of areas according to the memory address or the memory amount. It has configuration information, the dump acquisition range designation means designates the memory range to be saved in the first dump processing, and the dump acquisition means stores the memory contents in the range designated by the dump acquisition range designation means in the dump storage area Further, the dump file generation means converts the memory contents stored in the dump storage area into a file. When a failure occurs in the computer system, when the dump acquisition unit saves the contents of the memory range specified by the dump acquisition range specification unit in the dump storage area and restarts the computer system, Only the memory range saved at the time of occurrence is initialized, the dump file generation means converts the contents of the dump storage area into a file, and after the restart of the computer system is completed, the memory contents that were not saved at the time of failure are In addition to saving in the dump storage area, the contents of the dump storage area are converted into a file.
[0012]
DETAILED DESCRIPTION OF THE INVENTION
Examples of the present invention will be described.
[0013]
FIG. 1 is a block diagram of a computer system showing an embodiment of the present invention. In this figure, a computer system 124 includes an auxiliary storage device 111 such as a magnetic disk for permanently storing program instruction sequences and data, and an EEPROM (Electrically Erasable Programmable Read) for permanently storing program instruction sequences and data. A non-volatile memory 101 such as an only memory), a memory 107 for storing a program instruction sequence and data, a processor 100 for executing the program instruction sequence stored in the non-volatile memory 101 and the memory 107, and a transmission path between each block of the computer It is comprised from the bus | bath 110 which is.
[0014]
The auxiliary storage device 111 includes an operating system (OS) that manages resources such as the dump acquisition range specification file 112 that specifies the range of the memory 107 to be saved when a failure occurs in the computer system, the auxiliary storage device 111 of the computer system 124, the memory 107, and the like. ) Program instruction sequence and data storage area 113 for storing data, and a dump storage area 114 for storing memory contents when a failure occurs. The dump storage area 114 has a primary dump storage area 116 used in the first dump process and a secondary dump storage area 115 used in the second dump process. In the present embodiment, a computer system having two storage areas 115 and 116 in the dump storage area 114 is described, but two or more storage areas may be provided as necessary.
[0015]
The non-volatile memory 101 manages configuration information 102 for holding the size of the memory 107 installed in the computer system 124 and the location of the primary dump storage area 116 and the secondary dump storage area 115 in the auxiliary storage device 111. Dump storage area management information 105, and a system activation procedure 106 executed to activate the computer system 124 when a power switch is turned on or when a reset signal is detected. In the configuration information 102, an actual memory amount storage area 103 that holds the amount of memory used when the computer system 124 is started, and a reserved memory amount storage area that holds a memory amount that is additionally used after the computer system 124 is started. 104.
[0016]
The memory 107 (1) corresponds to the memory range stored in the first dump process. The primary dump acquisition range 108 (1) includes a program instruction sequence used by the user and a user holding data. It has an area 109 (1) and an OS area 119 for holding program instruction sequences and data used by the OS. The OS area 119 includes memory management information 120 used by the OS to allocate memory to the user, a failure processing procedure 121 executed when a failure occurs in the computer system 124, and a dump storage area 114 when the computer system 124 is restarted. A dump file generation procedure 122 for converting the contents of the file into a file, and a memory expansion procedure 123 for adding and using the memory specified by the reserved memory amount 104 after the computer system 124 is restarted. Embodiments of the failure processing procedure 121 and the dump file generation procedure 122 are described in “Panic! UNIX System Crash Tracking and Countermeasures” (1997), ISBN 4-7561-1912-3, pages 29 and 39, respectively. It is familiar with the panic () routine and the savecore program.
[0017]
In the computer system 124, configuration information 102 that holds information necessary for reserving a management area necessary for memory expansion without reinitializing the OS after the OS initialization is completed, and the OS reserves for memory expansion. An embodiment of the memory expansion procedure 123 for updating the management area is described in Japanese Patent Application No. 10-27996.
[0018]
FIG. 2 shows an embodiment of the memory management information 120 in the present invention. The memory management information 120 has a use start address 201 and a use end address 202 for the range of the memory 107 used by the processor 100, and a reservation start address 203 and a reservation end address for another range of the memory 107 used by the processor 100. It has 204. For the range from the reservation start address 203 to the reservation end address 204, the OS cannot be used after the computer system 124 is activated. The OS can use the range from the reservation start address 203 to the reservation end address 204 by executing the memory expansion procedure 123.
[0019]
FIG. 3 shows an embodiment of the dump storage area management information 105 in the present invention. The dump area management information 105 holds an offset, which is a value indicating the location of the primary dump storage area 116 in the auxiliary storage device 111, in the primary dump storage area offset 301, and stores the offset of the secondary dump storage area 116 in the auxiliary storage apparatus 111. The offset indicating the location is held in the secondary dump storage area offset 302, and the maximum offset indicating the end of the dump storage area 114 in the auxiliary storage device 111 is held in the dump storage area maximum offset 303. An offset indicating a location of the memory management information 120 held in the next dump storage area 116 is held in the memory management information offset 304.
[0020]
FIG. 4 is a flowchart showing an embodiment of a schematic processing flow from the occurrence of a failure to the completion of restart in the present invention.
[0021]
First, when a failure occurs in the computer system, that is, when the OS detects a state such as data inconsistency during execution of the program instruction sequence, step 401 is executed. In step 401, the failure processing procedure 121 is executed, and the contents of the memory 107 designated by the dump acquisition range designation file 112 are stored in the primary dump storage area 116. Here, the primary dump acquisition range 108 (1) is a range determined by the dump acquisition range specification file 12. Here, an example in which the size of the primary dump acquisition range 108 (1) is 1 GB will be described. In the dump acquisition range specification file, “1 GB” is described. Since 1 GB corresponds to address 0x40000000 (“0x” means a hexadecimal number), the dump acquisition range is the memory contents from address 0 to address 0x40000000. In another example, “FROM = 0, TO = 0x40000000” is described in the file. In this case as well, the dump acquisition range is the memory contents from address 0 to address 0x40000000.
[0022]
Next, step 402 is executed to restart the computer system. In step 402, the system is restarted using only the memory in the primary dump acquisition range 108 (1). Further, a memory other than the primary dump acquisition range 108 (1), that is, a memory in the secondary dump acquisition range 108 (2) is set so that it cannot be used from the OS. The reason for setting the secondary dump acquisition range so that the OS cannot be used is to not destroy the contents of the memory when a failure occurs. If the OS uses the memory 108 (2) in the secondary dump acquisition range, the memory contents at the time of the failure are lost and the cause of the failure cannot be analyzed.
[0023]
Next, when the restart of the computer system is completed, step 403 is executed. When the restart of the computer system is completed, the user's program instruction sequence can be executed. In step 403, a dump file generation procedure 122 is executed. The dump file generation procedure 122 converts the memory contents saved in the primary dump storage area 116 into a file format that can be analyzed by a system administrator or the like with a dump analysis tool. Further, the memory contents of the secondary dump acquisition range 108 (2) are stored in the secondary dump storage area 115 and converted into a file format that can be analyzed by the dump analysis tool. In the present embodiment, an example will be described in which the memory contents of the secondary dump acquisition range 108 (2) are stored in the secondary dump storage area 115 and then converted into a file format. Here, the memory contents may be directly converted into a file format without being stored in the secondary dump storage area 115.
[0024]
Next, when the execution of the dump file generation procedure 122 is completed, step 404 is executed. In step 404, the memory expansion procedure 123 is executed so that the OS can use the memory in the secondary dump acquisition range 108 (2) that was not used when the system was restarted.
[0025]
FIG. 5 is a flowchart showing an embodiment of the failure processing procedure in the present invention.
[0026]
First, in step 501, the offset 301 of the primary dump storage area of the dump area management information 105 specifies the memory contents from the use start address 201 of the memory management information 120 to the address specified by the dump acquisition range specification file 112. Saved in the primary dump storage area 116.
[0027]
In step 502, the last position (offset) of the primary dump storage area 116 saved in step 501 is set as the secondary dump storage area offset 302.
[0028]
In step 503, the position (offset) of the memory management information 120 stored in the primary dump storage area 116 is set in the memory management information offset 304.
[0029]
In step 504, the reserved memory amount 104 is set to a value obtained by subtracting the memory amount corresponding to the address specified in the dump acquisition range specifying file 112 from the actual memory amount 103.
[0030]
In step 505, the memory amount corresponding to the address specified in the dump acquisition range specification file 112 is set as the real memory amount 103. For example, when “1 GB” is described in the dump acquisition range designation file 112, 1 GB is set as the actual memory amount. As another example, when “FROM = 0, TO = 0x40000000” is described in the file, the memory amount 1 GB corresponding to the address 0x40000000 is set as the actual memory amount.
[0031]
FIG. 6 is a flowchart showing an embodiment of the system startup procedure in the present invention.
[0032]
First, in step 601, the processor 100, the auxiliary storage device 111, and the amount of memory 107 held by the real memory amount 103 are initialized so that the processor 100 can read and write to the auxiliary storage device 111 and the memory 107.
[0033]
In step 602, the processor 100 reads the OS program instruction sequence and data from the OS storage area 113, writes the OS program instruction sequence and data in the memory 107 initialized in step 601, and transfers control to the OS program instruction sequence.
[0034]
In step 603, the configuration information 102 is read and the memory management information 120 is created. When creating the memory management information 120, with respect to the use start address 201 and the use end address 202, for example, the use start address 201 is set to 0, and the real memory amount 103 of the configuration information 102 is read and converted into a corresponding address. To use final address 202. When 1 GB is set in the real memory amount 103, the use end address 202 is set to 0x40000000.
[0035]
As for the reservation start address 203 and the reservation end address 204, for example, the reservation start address 203 is set to the same address as the use end address 202, and the reserved memory amount 104 of the configuration information 102 is read and converted into an address for reservation. An address added to the head address 203 is set as a reserved final address 204. For example, if 0x40000000 is set as the reservation start address and 1 GB is set as the reservation memory amount 104, the address 0x80000000, which is the final address when 1 GB of memory is added from the address 0x40000000, is set as the reservation final address 204. .
[0036]
FIG. 7 is a flowchart showing an embodiment of the dump file generation procedure in the present invention.
[0037]
First, in step 701, the memory management information offset 304 is read from the dump area management information 105, and the use final address 202 saved in step 501 is read from the auxiliary storage device 111 using the offset.
[0038]
In step 702, the secondary dump storage area offset 302 is read to determine the secondary dump storage area 115.
[0039]
In step 703, the contents of the memory 107 from the address corresponding to the actual memory amount 103 set in step 505 to the use final address 202 read in step 701 are stored in the secondary dump storage area 115 determined in step 702. .
[0040]
In step 704, a savecore procedure is executed. An embodiment of the savecore procedure is described in “Panic! UNIX System Crash Tracking and Countermeasures” (1997), page 39 of ISBN 4-7561-1912-3.
[0041]
In step 705, the memory expansion procedure 123 is executed.
[0042]
In this embodiment, after the memory contents at the time of failure are stored in the primary dump storage area 116 and the secondary dump storage area 115 in step 704, they are converted into a file format (dump file) that can be analyzed by the dump analysis tool. It explains the computer system to do. As another embodiment, a savecore procedure call is executed before step 701 to convert only the primary dump storage area 116 into a dump file, and at step 704, only the secondary dump storage area 115 is converted into a dump file. It may be a computer system.
[0043]
FIG. 8 is a flowchart showing an embodiment of the memory expansion procedure according to the present invention.
[0044]
First, in step 801, the memory 107 is initialized from an address corresponding to the real memory amount 103 to an address obtained by adding an address corresponding to the reserved memory amount 104 to an address corresponding to the real memory amount 103. For example, the address corresponding to the real memory amount 103 is the address 0x40000000 corresponding to 1 GB when 1 GB is set in the real memory amount 103. When 1 GB is set in the reserved memory amount 104, the memory up to the address 0x80000000 is initialized by adding 0x40000000 corresponding to the reserved memory amount 1GB to the address 0x40000000 corresponding to the actual memory amount.
[0045]
In step 802, the memory amount obtained by adding the real memory amount 103 and the reserved memory amount 104 is set as the real memory amount 103.
[0046]
In step 803, the reserved memory amount is set to zero.
[0047]
In step 804, the OS memory management information 120 is changed. In order to use the memory that has been initialized in step 801, the use final address 202 is changed. For example, when the use final address 202 and the reservation head address 203 are the same address, the value held in the reservation final address 204 is set as the use final address. Thereafter, in order to indicate that the reserved memory amount is 0, the value of the use final address 202 is set to the reservation start address 203 and the reservation final address 204.
[0048]
Next, an embodiment in which the present invention is applied to a computer system that performs virtual storage control will be described.
[0049]
Virtual storage control is a system that manages the address space divided into fixed sizes, and associates logical addresses, which are composed of management unit numbers (called pages) and offsets in pages, with physical memory addresses. is there. In the user program, a logical address is used when the processor accesses an instruction sequence or data of the user program. By using the logical address in the user program, the programmer of the user program does not need to be aware of the address of the physical memory where the program operates. Further, in the OS, it is possible to effectively use the physical memory by evicting the contents of pages that are not frequently used from the physical memory to the auxiliary storage device. Details on virtual memory control can be found on pages 80-85 of Barney Goodhart, James Cox, and translated by Takashi Sakuragawa, Prentice Hall, "Unix Kernel Magic System V Release 4 Architecture" (1997). Has been.
[0050]
FIG. 9 shows an embodiment of the memory management information 120 of the computer system for controlling virtual storage in the present invention.
[0051]
In a computer that performs virtual storage control, the memory management information includes the use start address 201, use end address 202, reservation start address 203, and reservation end address 204 described with reference to FIG. It consists of an address 902, a logical address / physical address conversion table 903, and a free list 904.
[0052]
An address conversion unnecessary first address 901 and an address conversion unnecessary final address 902 indicate areas that do not need to be converted from a logical address to a physical address.
[0053]
The logical address / physical address conversion table 903 is a correspondence table between logical addresses and physical addresses.
[0054]
The free list 904 manages information for managing memory resources used in the computer system 124.
[0055]
In virtual storage control, when a processor accesses a certain logical address, the accessed page may exist in an auxiliary storage device instead of physical memory. In this case, the selected page content is transferred to the auxiliary storage device after selecting the page allocated to the physical memory based on information such as the frequency of use, and the accessed page content is transferred from the auxiliary storage device to the physical memory instead. Forward to.
[0056]
Here, if a page including a program instruction sequence for transferring the page contents from the auxiliary storage device to the physical memory exists in the auxiliary storage device, the page accessed by the processor cannot be transferred to the physical memory. In order to prevent such a situation from occurring, a page including a program instruction sequence for transferring the page contents from the auxiliary storage device to the physical memory must always exist in the physical memory.
[0057]
Such a page is included in an area where no address conversion is required by an address conversion unnecessary start address 901 and an address conversion unnecessary final address 902. Since the data in the area that does not require address translation is usually managed in a fixed size, if the memory used by the processor changes after the computer system restarts, reserve it in advance. Is required.
[0058]
Data that needs to be reserved includes a logical address / physical address conversion table, a free list, and the like, which differ depending on the OS implementation. In this embodiment, an OS that requires reservation of a logical address / physical address conversion table and a free list will be described.
[0059]
FIG. 10 shows an embodiment of the logical address / physical address conversion table 903 in the present invention.
[0060]
The logical address / physical address conversion table 903 includes a physical page number 1001 that holds the page number of the physical address, a logical page number 1002 that holds the page number of the logical address, and a conversion pair of the physical page number and the logical page number in the same row. Management information 1003 including a flag indicating whether or not is valid.
[0061]
This figure shows that in a computer system with a page size of 4 KB (0x1000) in which about 2 GB of memory is mounted, the memory management information 120 is the physical page number 0-0x1000, the OS area is the physical page number 0-0x8000, and the user area 109 (1 ) Is the physical page number 0x8001 to 0x40000, the user area 109 (2) is the physical page number 0x40001 to 0x80000, the primary dump acquisition range 108 (1) is the physical page number 0-0x40000, and the secondary dump acquisition range 108 (2) is Each corresponds to physical page numbers 0x40001 to 0x80000.
[0062]
In this figure, a logical address 0x10000123 composed of a logical page number 0x10000 and an offset 0x123 in the page is converted into a physical address 0x4000123 using a physical address number 0x40000 and an offset 0x123 corresponding to the logical page number 0x10000.
[0063]
For details of the logical address / physical address conversion table 903, pages 3-9 to 3 of Gary Kane, published by Prentice Hall "PA-RISC 2.0 ARCHITECTURE" ISBN 0-13-182734-0 (1996). It is described on page 16.
[0064]
A system startup procedure 106 when applied to a computer system that performs virtual storage control will be described. The basic processing content is the same as the flowchart (601 to 603) described in FIG. In a computer system that performs virtual storage control, an address conversion unnecessary start address 901, an address conversion unnecessary final address 902, a logical address / physical address conversion table 903, and a free list 904 are added to the memory management information 120. Therefore, in step 603, the start address and the final address of the OS area 119 are set to the address conversion unnecessary start address 901 and the address conversion unnecessary final address 902, respectively, and the total memory amount 103 and the reserved memory amount 104 are managed. An area for storing the logical address / physical address conversion table 903 and the free list 903 is secured as possible. Of the pages registered in the free list 903, the page corresponding to the reserved memory amount 104 is set as reserved using the page lock of the PAGE structure so that the OS cannot be used. Embodiments of page locks are described on pages 80-85 of Prentice Hall's "Unix Kernel Magic System V Release 4 Architecture" (1997), written by Bernie Goodhart, James Cox, and translated by Takashi Sakuragawa. ing.
[0065]
Next, the memory expansion procedure 123 when applied to a computer system that performs virtual storage control will be described. The basic processing contents are the same as those in the flowchart (801 to 804) described in FIG. In step 804, by releasing the page lock of the PAGE structure corresponding to the reserved memory amount 104, the memory of the reserved memory amount 104 can be used without reinitializing the OS.
[0066]
【The invention's effect】
According to the present invention, it is possible to reduce the time required to complete the restart after the computer system is stopped and to provide the service to the user, and as a result, the operating rate of the computer system is improved during a certain period. To do.
[0067]
Further, after rebooting, the memory range that was not saved in the first first process can be converted into a file, and the memory contents necessary for failure cause analysis can be provided to the system administrator. Even in a computer system having a dump storage area smaller than the size of the memory installed in the computer system, a file including all memory contents can be obtained, and the cause of the failure can be sufficiently analyzed.
[Brief description of the drawings]
FIG. 1 is a block diagram of a computer system 124 showing an embodiment of the present invention.
FIG. 2 is an example of memory management information 120 in the present invention.
FIG. 3 is an example of dump storage area management information 105 according to the present invention.
FIG. 4 is a flowchart showing an embodiment of a schematic processing flow from the occurrence of a failure to the completion of restart in the present invention.
FIG. 5 is a flowchart showing an embodiment of a failure processing procedure 121 in the present invention.
FIG. 6 is a flowchart showing an embodiment of a system activation procedure 106 according to the present invention.
FIG. 7 is a flowchart showing an embodiment of a dump file generation procedure 122 according to the present invention.
FIG. 8 is a flowchart showing an embodiment of a memory expansion procedure 123 in the present invention.
FIG. 9 shows an example of memory management information 120 of a computer system that controls virtual storage according to the present invention.
FIG. 10 shows an example of a logical address / physical address conversion table 903 according to the present invention.
[Explanation of symbols]
100: Processor, 101: Non-volatile memory, 102: Configuration information, 105 Dump storage area management information, 106: System startup procedure, 107: Memory, 108: Dump acquisition range, 109: User area, 110: Bus, 111: Auxiliary Storage device 112: Dump acquisition range specification file 113: OS storage area 114: Dump storage area 119: OS area 120: Memory management information 121: Failure processing procedure 122: Dump file generation procedure 123: Memory Expansion procedure

Claims (3)

A computer system dump processing method comprising a memory, an auxiliary storage device, and a control means for controlling the memory and the auxiliary storage device ,
The computer system includes: a dump storage unit that stores, in an auxiliary storage device, memory contents of a plurality of areas divided based on memory management information that the control unit allocates a memory to a user; and memory contents stored in the auxiliary storage device. A dump file generating means for converting into a file, a memory initializing means for initializing the memory based on memory configuration information indicating a numerical value for dividing the memory into a plurality of areas according to a memory address or a memory amount , A dump area management means indicating a storage location for storing the memory contents when storing the memory contents in the auxiliary storage device;
When said control means detects a failure of the computer system, the dump storage means, said auxiliary storage memory contents of the first areas divided on the basis of the memory management information and dump area selected stored in the storage location indicated by the dump area management means, wherein the first by the memory initialization means with the control means sets as the actual amount of memory the amount of memory region in which the dump storage means has stored in the memory configuration information of the memory region and restart the computer system is initialized, after the control unit has completed the rebooting of said computer system, read the first th region stored at the dump storage means from the auxiliary storage device, wherein converting the dump file generation means memory areas read in the file, No. 2 which is divided based on the memory management information Wherein stored in the dump area management storage location means illustrated, dump processing method of a computer system and converting the second and subsequent area from the auxiliary storage device to a file and later region an auxiliary storage device. 2. The dump processing method for a computer system according to claim 1, wherein the dump area management means holds at least two entries having values indicating the location of the dump storage area. 3. The computer system dump processing method according to claim 1, wherein the control means has a virtual storage control function.

Images