I/O control apparatus having check recovery function

• US 5,931,954 A
• Filed: 01/13/1997
• Issued: 08/03/1999
• Est. Priority Date: 01/31/1996
• Status: Expired due to Fees

• Current Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Original Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Inventors: Sakai, Hiroshi, Sakuma, Takeshi, Hoshina, Satoshi
• Primary Examiner(s): Beausoliel, Jr., Robert W.
• Assistant Examiner(s): Elmore, Stephen C.
• Application Number:
  US08/782,077
• Time in Patent Office: 932 Days
• Field of Search: 395/182.13, 395/181, 395/180, 364/285.2, 364/230.2, 364/267, 364/267.8
• US Class Current: 714/15
• CPC Class Codes: G06F 11/1407 Checkpointing the instruction stream

Abstract

In a computer system, when a CPU performs a state setting operation, such as setting an appropriate operation mode for I/O devices, the log data of the state setting operation is stored in a log storage area. Upon occurrence of a fault in the computer system, the I/O devices are cleared, and the state setting operation of the I/O devices is performed on the basis of the log data of the prior state setting operation stored in the log storage area. Therefore, the states of the I/O devices can be recovered to a state at a checkpoint when the process is restarted.

Claims

1. An I/O control apparatus in a computer system which has one or more CPUs, a main memory, and one or more I/O devices and in which said CPUs periodically save the internal state of said CPUs and the contents of said main memory as a checkpoint, and the internal status of said CPUs and the contents of said main memory of a most recent checkpoint are restored when a fault occurs in said computer system to restart data processing comprising:

I/O device state storing means for logging parameters of a state setup sequence of said I/O devices performed by said CPUs, said I/O device state storing means including means for erasing part of previously logged parameters of an I/O device which is made unnecessary when a new state of said I/O device is set; and

I/O device state restoring means for restoring the state of said I/O devices to that of the most recent checkpoint by first initializing said I/O devices and by second reexecuting the state setup sequence according to said logged parameters stored by said I/O device state storing means.

2. An apparatus according to claim 1, further comprising:

request block creating means for creating, when an application process in said computer system makes an I/O request, a request block in said main memory which contains information necessary to perform said I/O request;

I/O execution processes for performing an I/O operation by executing I/O device driver routines according to a request block; and

I/O execution process initializing means for initializing, upon restart from the most recent checkpoint after a fault occurrence, said I/O execution processes other than in initial state and causing I/O operations being performed by said I/O execution processes to be performed again from the beginning.

3. An apparatus according to claim 2, wherein, of request blocks held in said main memory, said I/O execution processes begin to perform an I/O operation according to a request block created before the most recent checkpoint, while said I/O execution processes postpone an I/O operation according to a request block created after the most recent checkpoint until a new checkpoint acquisition.

4. An apparatus according to claim 2, wherein the CPU which executes an application process which made an I/O request also executes an I/O execution process which is responsible for the request block created based on said I/O request.

5. An apparatus according to claim 2, wherein the number of CPUs which are assigned to said I/O execution processes is properly determined depending on the number of request blocks to be processed.

6. An apparatus according to claim 2, further comprising means for making, when a fault occurs, an error reply to the application process without reexecuting the requested I/O operation, in case that said I/O device state restoring means does not manage to restore the state of the I/O device which relates to said I/O request.

7. An apparatus according to claim 2, further comprising means for reporting, when a fault occurs, a successful I/O completion to the application process without re-executing the requested I/O operation, in a case that the I/O request is an output request and the I/O request has been completed before the fault occurrence.

8. An apparatus according to claim 3, wherein said CPUs are assigned to said I/O execution processes when a new checkpoint acquisition has been completed.

9. An apparatus according to claim 3, further comprising means for making, when a fault occurs, an error reply to the application process without reexecuting the requested I/O operation, in case that said I/O device state restoring means does not manage to restore the state of the I/O device which relates to said I/O request.

10. An apparatus according to claim 3, further comprising means for reporting, when a fault of said computer system occurs, a successful I/O completion to the I/O requesting process without re-executing the requested I/O operation, in a case that the I/O request was an output request and the I/O request has been completed before the fault occurrence.

11. An I/O control apparatus in a computer system which has one or more CPUs, a main memory, and one or more I/O devices and in which said CPUs periodically save the internal state of said CPUs and the contents of said main memory as a checkpoint, and the internal state of said CPUs and the contents of said main memory of a most recent checkpoint are restored when a fault occurs in said computer system to restart data processing, comprising:

I/O device state storing means for logging parameters of a state setup sequence of said I/O devices performed by said CPUs; and

I/O device state restoring means for restoring the state of said I/O devices to that of the most recent checkpoint by first initializing said I/O devices and by second reexecuting the state setup sequence according to said logged parameters stored by said I/O device state storing means, said I/O device state restoring means including means for skipping initializing and reexecuting the state setup sequence of an I/O device in which a new state setup sequence has not been performed since the most recent checkpoint.

12. An apparatus according to claim 11, further comprising:

request block creating means for creating, when an application process in said computer system makes an I/O request, a request block in said main memory which contains information necessary to perform said I/O request;

I/O execution processes for performing an I/O operation by executing I/O device driver routines according to a request block; and

I/O execution processes initializing means for initializing, upon restart from the most recent checkpoint after a fault occurrence, said I/O execution processes other than in initial state and causing I/O operations being performed by said I/O execution processes to be performed again from the beginning.

13. An apparatus according to claim 12, wherein, of request blocks held in said main memory, said I/O execution processes begin to perform an I/O operation according to a request block created before the most recent checkpoint, while said I/O execution processes postpone an I/O operation according to a request block created after the most recent checkpoint until a new checkpoint acquisition.

14. An apparatus according to claim 12, wherein the CPU which executes an application process which made an I/O request also executes an I/O execution process which is responsible for the request block created based on said I/O request.

15. An apparatus according to claim 12, wherein the number of CPUs which are assigned to said I/O execution processes is properly determined depending on the number of request blocks to be processed.

16. An apparatus according to claim 12, further comprising means for making, when a fault occurs, an error reply to the application process without reexecuting the requested I/O operation, in case that said I/O device state restoring means does not manage to restore the state of the I/O device which relates to said I/O request.

17. An apparatus according to claim 12, further comprising means for reporting, when a fault occurs, a successful I/O completion to the application process without re-executing the requested I/O operation, in a case that the I/O request is an output request and the I/O request has been completed before the fault occurrence.

18. An apparatus according to claim 13, wherein said CPUs assigned to said I/O execution processes when a new checkpoint acquisition has been completed.

19. An apparatus according to claim 13, further comprising means for making, when a fault occurs, an error reply to the application process without reexecuting the requested I/O operation, in case that said I/O device state restoring means does not manage to restore the state of the I/O device which relates to said I/O request.

20. An apparatus according to claim 13, further comprising means for reporting, when a fault occurs, a successful I/O completion to the application process without re-executing the requested I/O operation, in a case that the I/O request is an output request and the I/O request has been completed before the fault occurrence.

21. An I/O control apparatus in a computer system which has one or more CPUs, a main memory, and one or more I/O devices and in which said CPUs periodically save the internal state of said CPUs and the contents of said main memory as a checkpoint, and the internal state of said CPUs and the contents of said main memory of a most recent checkpoint are restored when a fault occurs in said computer system to restart data processing, comprising:

I/O device state storing means for logging parameters of a state setup sequence of said I/O devices performed by said CPUs;

I/O device state restoring means for restoring the state of said I/O devices to that of the most recent checkpoint by first initializing said 10 devices and by second reexecuting the state setup sequence according to said logged parameters stored by said I/O device state storing means;

request block creating means for creating, when an application process in said computer system makes an I/O request, a request block in said main memory which contains information necessary to perform said I/O request;

I/O execution processes for performing an I/O operation by executing I/O device driver routines according to a request block; and

I/O execution processes initializing means for initializing, upon restart from the most recent checkpoint after a fault occurrence, said I/O execution processes other than in initial state and causing I/O operations being performed by said I/O execution processes to be reexecuted from the beginning.

22. An apparatus according to claim 21, wherein., of request blocks held in said main memory, said I/O execution processes begin to perform an I/O operation according to a request block created before the most recent checkpoint, while said I/O execution processes postpone an I/O operation according to a request block created after the most recent checkpoint until a new checkpoint acquisition.

23. An apparatus according to claim 21, wherein the CPU which executes an application process which made an I/O request also executes an I/O execution process which is responsible for the request block created based on said I/O request.

24. An apparatus according to claim 21, wherein the number of CPUs which are assigned to said I/O execution processes is properly determined depending on the number of request blocks to be processed.

25. An apparatus according to claim 21, further comprising means for making, when a fault occurs, an error reply to the application process without reexecuting the requested I/O operation, in a case that said I/O device state restoring means fails to restore the state of the I/O device which relates to said I/O request.

26. An apparatus according to claim 21, further comprising means for reporting, when a fault occurs, a successful I/O completion to the application process without re-executing the requested I/O operation, in a case that the I/O request is an output request and the I/O request has been completed before the fault occurrence.

27. An apparatus according to claim 22, wherein said CPUs are assigned to said I/O execution processes when a new checkpoint acquisition has been completed.

28. An apparatus according to claim 22, further comprising means for making, when a fault occurs, an error reply to the application process without reexecuting the requested I/O operation, in case that said I/O device state restoring means does not manage to restore the state of the I/O device which relates to said I/O request.

29. An apparatus according to claim 22, further comprising means for reporting, when a fault occurs, a successful I/O completion to the application process without re-executing the requested I/O operation, in a case that the I/O request is an output request and the I/O request has been completed before the fault occurrence.

30. An I/O control method in a computer system which has one or more CPUs, a main memory, and one or more I/O devices and in which said CPUs periodically save the internal state of said CPUs and the contents of said main memory as a checkpoint, and the internal state of said CPUs and the contents of said main memory of the most recent checkpoint are restored when a fault occurs in said computer system to restart data processing, comprising the steps of:

logging parameters of a state setup sequence of said I/O devices performed by said CPUs, said logging step including erasing part of previously logged parameters of an I/O device which is made unnecessary when a new state of said I/O device is set; and

restoring the state of said I/O devices to that of the most recent checkpoint by first initializing said I/O devices and by second reexecuting the state setup sequence according to said logged parameters.

31. An I/O control method in a computer system which has one or more CPUs, a main memory, and one or more I/O devices and in which said CPUs periodically save the internal state of said CPUs and the contents of said main memory as a checkpoint, and the internal state of said CPUs and the contents of said main memory of a most recent checkpoint are restored when a fault occurs in said computer system to restart data processing, comprising the steps of:

logging parameters of a state setup sequence of said I/O devices performed by said CPUs; and

restoring the state of said I/O devices to that of the most recent checkpoint by first initializing said I/O devices and by second reexecuting the state setup sequence according to said logged parameters, said restoring step including initializing and reexecuting the state setup sequence of an I/O device in which a new state setup sequence has been performed since the most recent checkpoint.

32. An I/O control method in a computer system which has one or more CPUs, a main memory, and one or more I/O devices and in which said CPUs periodically save the internal state of said CPUs and the contents of said main memory as a checkpoint, and the internal state of said CPUs and the contents of said main memory of a most recent checkpoint are restored when a fault occurs in said computer system to restart data processing, comprising the steps of:

logging parameters of a state setup sequence of said I/O devices performed by said CPUs;

restoring the state of said I/O devices to that of the most recent checkpoint by first initializing said I/O devices and by second reexecuting the state setup sequence according to said stored logged parameters;

creating, when an application process in said computer system makes an I/O request, a request block in said main memory which contains information necessary to perform said I/O request;

performing I/O operations, using I/O execution processes, by executing I/O device driver routines according to a request block; and

initializing, upon restart from the most recent checkpoint after a fault occurrence, said I/O execution processes other than in initial state and causing I/O operations being performed by said I/O execution processes to be reexecuted from the beginning.