Method of executing perform locked operation instructions for supporting recovery of data consistency if lost due to processor failure, and a method of recovering the data consistency after processor failure

• US 5,895,494 A
• Filed: 09/05/1997
• Issued: 04/20/1999
• Est. Priority Date: 09/05/1997
• Status: Expired due to Term

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Scalzi, Casper Anthony, Plambeck, Kenneth Ernest
• Primary Examiner(s): Lim, Krisna
• Application Number:
  US08/924,890
• Time in Patent Office: 592 Days
• Field of Search: 711/144, 711/149, 711/150, 711/152, 395/376, 395/392
• US Class Current: 711/150
• CPC Class Codes: G06F 11/2289 by configuration test

Abstract

Provides a processor method of executing instances of a Perform Locked Operation (PLO) instruction for enabling a recovery of the consistency of a resource unit being changed by a PLO instance when processor failure occurs anywhere during execution of the PLO instance. The method uses a PLO save area for each processor in a computer system capable of executing PLO instructions. Each PLO save area has a resource-inconsistency (RI) indicator having an RI state and a non-RI state, and stores the function code (FC) of the PLO instance. The RI state indicates that the resource is in a non-usable potentially inconsistent state, and the non-RI state indicates the resource is in the consistent state and may be used. A processor executing a PLO instance writes into its PLO save area all resource addresses where a change is to be made in the resource unit, and also writes in its PLO save area all operand values which will be used to change the resource at the associated addresses. After the processor has successfully written all of these addresses and associated operand data values into its PLO save area, the processor sets its RI indicator to the RI state, and then stores into the resource unit the data values stored in its PLO save area at the resource addresses also stored in its PLO save area, according to the FC also stored therein. After completing all changes in the resource, the processor sets its RI indicator to the non-RI (resource consistent) state. If the processor should ever fail while executing a PLO instance, it signals its failing condition to the system. Then, a RCR method is executed by any operational processor (acting as a RCR processor) to recover the consistency of the resource unit. The RCR method uses the state of the RI indicator with the failed processor'"'"'s PLO save area, the FC stored therewith, and the content of the failed processor'"'"'s PLO save area to restore the consistency of the resource unit.

Claims

1. A method of maintaining the integrity of data in a computer system against processor failure while executing perform locked operation (PLO) instructions, the data being stored in data resources accessible through storage of the computer system, each instance of execution of a PLO instruction (PLO instance) comprising the steps of:

structuring a save area in the storage for each processor in the computer system capable of executing PLO instructions, each PLO instance of execution having one or more operands specified by the PLO instruction of the PLO instance, the one or more operands including specifications for generating store operand values of data items to be stored in a resource unit specified by a blocking symbol in the PLO instance,

associating a resource-inconsistency (RI) indicator with each processor'"'"'s save area, the RI indicator being setable to at least a first state (RI state) and a second state (non-RI state), the non-RI state indicating resource data consistency in the resource unit, and the RI state indicating potential resource inconsistency in the resource unit specified by the PLO instance, and store operand values having been generated by using store operands specified in the PLO instance,

initiating execution of the PLO instance by a processor while the RI indicator of the processor'"'"'s save area is set to the non-RI state, and

writing in the processor'"'"'s save area each store operand value targeted for changing a data item in the resource unit while the RI indicator of the processor'"'"'s save area is set to the non-RI state, and also writing in the processor'"'"'s save area in association with each store operand value a resource address at which the store operand value is to be written in the resource unit.

2. A method of maintaining the integrity of data in a computer system against processor failure while executing perform locked operation (PLO) instructions as defined in claim 1, comprising the steps of:

setting the RI indicator to the RI state after the writing step and the also writing step have stored all of the store operand values and associated resource addresses in the processor'"'"'s save area,

copying each store operand value from the processor'"'"'s save area to the associated resource address in the resource unit while the RI indicator is set to the RI state for changing the resource unit, and

setting the RI indicator of the processor'"'"'s save area to the non-RI state after the copying step has completed changing the resource unit by copying all of the store operand values from the processor'"'"'s save area into the resource unit.

3. A method of maintaining the integrity of data in a computer system against processor failure while executing perform locked operation (PLO) instructions as defined in claim 1, comprising the steps of:

performing the writing and also writing steps in the processor'"'"'s save area during the execution of the PLO instance before the processor attempts to obtain a lock on the resource unit specified by the PLO instance, and while the RI indicator of the processor'"'"'s save area is set to the non-RI state.

4. A method of maintaining the integrity of data in a computer system against processor failure while executing perform locked operation (PLO) instructions as defined in claim 1, comprising the steps of:

performing the writing and also writing steps in the processor'"'"'s save area during the execution of the PLO instance after the processor obtains a lock on the resource unit specified by the PLO instance, and while the RI indicator of the processor'"'"'s save area is set to the non-RI state.

5. A method of maintaining the integrity of data in a computer system against processor failure while executing PLO instructions as defined in claim 1, comprising the step of:

storing with the processor'"'"'s save area a function code (FC) obtained from the PLO instance of execution, and the storing step making the FC available in the processor'"'"'s save area to a future recovery processor if the processor should fail while executing the PLO instance.

6. A method of maintaining the integrity of data in a computer system against processor failure while executing a perform locked operation (PLO) instruction, as defined in claim 2, further comprising the steps of:

performing the steps of writing in, and also writing in, the processor'"'"'s save area without the processor having obtained a lock on the resource specified by the PLO instance,

attempting by the processor to obtain a lock on the resource unit specified by the PLO instance while the RI indicator of the processor'"'"'s save area is set to the non-RI state, and

performing the setting and copying steps only if the lock on the resource is obtained by the processor.

7. A method of maintaining the integrity of data in a computer system against processor failure while executing a perform locked operation (PLO) instruction, as defined in claim 2, further comprising the steps of:

obtaining a lock on the resource unit specified by the PLO instance while the RI indicator of the processor'"'"'s save area is set to the non-RI state, and

then executing the steps of setting the RI indicator to RI state and copying each store operand value to the associated resources address in the resource unit.

8. A method of maintaining the integrity of data in a computer system against processor failure while executing PLO instructions as defined in claim 1, further comprising the steps of:

providing one or more comparand operands for the PLO instance, each comparand operand specifying a comparand location associated with the resource unit which contains a comparand value that is changed when the resource unit is changed,

executing instructions, other than the PLO instruction of the PLO instance, prior to initiating execution of the PLO instance, for preparing operand values for the PLO instance including a prepared comparand value,

during execution of the PLO instance, comparing the prepared comparand value with a current comparand value currently fetched from the comparand location of the resource unit, and

continuing execution of the PLO instance only if the prepared comparand value is the same as the current comparand value.

9. A method of maintaining the integrity of data in a computer system against processor failure while executing PLO instructions as defined in claim 8, further comprising the steps of:

terminating execution of the PLO instance if the prepared comparand value is different from the current comparand value.

10. A method of maintaining the integrity of data in a computer system against processor failure while executing PLO instructions as defined in claim 8, further comprising:

performing the steps of structuring, associating, initiating, writing and also writing in response to the step of continuing execution,

then obtaining a lock on the resource unit in response to the continuing execution step,

setting the RI indicator to the RI state after the writing step and the also writing step have stored all of the store operand value and associated resource addresses in the processor'"'"'s save area,

copying each store operand value from the processor'"'"'s save to the associated resource address in the resource unit while the RI indicator is set to the RI state for changing the resource unit,

setting the RI indicator of the processor'"'"'s save area to the non-RI state after the copying step has completed changing the resource unit by copying all of the store operand values from the processor'"'"'s save area into the resource unit, and

releasing the lock on the resource unit.

11. A method of maintaining the integrity of data in a computer system against processor failure while executing PLO instructions as defined in claim 1, further comprising the steps of:

providing one or more comparand operands for the PLO instance, each comparand operand specifying a comparand location associated with the resource unit which contains a comparand value that is changed when the resource unit is changed,

executing instructions, other than the PLO instruction of the PLO instance, prior to initiating execution of the PLO instance, for preparing operand values for the PLO instance including a prepared comparand value,

obtaining a lock on the resource unit upon initiating execution of the PLO instance,

then comparing the prepared comparand value with a current comparand value currently fetched from the comparand location of the resource unit, and

continuing execution of the PLO instance only if the prepared comparand value is the same as the current comparand value,

performing steps of structuring, associating, initiating, writing and also writing to store the processor'"'"'s save area in response to the step of continuing execution,

next setting the RI indicator to the RI state after the writing step and the also writing step have stored all of the store operand values and associated resource addresses in the processor'"'"'s save area,

copying each store operand value from the processor'"'"'s save area to the associated resource address in the resource unit while the RI indicator is set to the RI state for changing the resource unit,

setting the RI indicator of the processor'"'"'s save area to the non-RI state after the copying step has completed changing the resource unit by copying all the store operand values form the processor'"'"'s save area into the resource unit to change the resource, and

releasing the lock on the resource unit.

12. A method of maintaining integrity of data in a computer system by a processor in the computer system executing a program containing a perform locked operation (PLO) instruction specifying a function for changing data in a resource unit stored in a computer system, the PLO instruction having operands for specifying where and what in the resource unit is to be changed by execution of the PLO instruction, the processor performing the steps of:

executing by a processor preparatory instructions for determining values of operands of a PLO instance to be executed representing a PLO instruction specifying where and what in a resource unit is to be changed by execution of the PLO instance, including preparing operand addresses for locating in the resource insertions and/or deletions of data and for preparing operand values of data to be inserted and/or deleted in the resource unit at prepared operand addresses, operands of the PLO instance including one or more comparands, each comparand value prepared by fetching a predetermined location associated with the resource unit, each comparand value being changeable by another processor executing another PLO instance for changing the resource unit and,

storing each prepared operand value, including each comparand value, for the operands of a PLO instance to be executed,

initiating execution of the PLO instance by the processor,

comparing each prepared comparand value with a current value of the comparand operand fetched from a predetermined location associated with the resource unit to test if the resource unit has been changed since the prepared comparand value was prepared,

if the comparing step finds an equal condition, generating store operand values of the PLO instance for changing the resource unit and generating the addresses in the resource unit at which the store operand values are to be stored, and storing the store operand values and the addresses in a save area of the processor executing the PLO instance,

setting to a locked state a hardware lock associated with a blocking symbol specified by the PLO instance for locking the resource unit from being changed by any other processor in the computer system,

again comparing each prepared comparand value with a current value of the comparand operand fetched from a predetermined location associated with the resource unit to test if the resource unit has been changed since the prepared comparand value was prepared,

if the again comparing step finds an equal condition, setting a resource-inconsistency (RI) indicator of the processor'"'"'s PLO save area to a RI state in preparation for changing the resource unit,

accessing the resource unit at locations specified by the addresses stored in the processor'"'"'s save area to change the resource unit with the operand values stored in the processor'"'"'s save area in accordance with a function code (FC) specified in the PLO instance being executed,

resetting the RI indicator to a non-RI state after completion of changes of the resource unit, and

resetting to an unlocked state the hardware lock associated with the blocking symbol specified in the PLO instance to enable the resource unit to be accessed by any processor in the computer system.

13. A method of maintaining the integrity of data in a computer system by executing a perform locked operation (PLO) instruction as defined in claim 12, further including the steps of:

terminating the execution of the PLO instance if either the comparing step, or the again comparing step, finds an unequal condition, and

signaling an issuing program a termination of the execution of the PLO instance to enable the issuing program to again execute preparatory instructions for a reissue of the PLO instruction of the PLO instance.

14. A method of maintaining integrity of data in a computer system by a processor in the computer system executing a program containing a perform locked operation (PLO) instruction specifying a function for changing data in a resource unit stored in a computer system, the PLO instruction having operands for specifying where and what in the resource unit is to be changed by execution of the PLO instruction, the processor performing the steps of:

executing by a processor preparatory instructions for determining values of operands of a PLO instance to be executed representing a PLO instruction specifying where and what in a resource unit is to be changed by execution of the PLO instance, including preparing operand addresses for locating in the resource insertions and/or deletions of data and for preparing operand values of data to be inserted and/or deleted in the resource unit at prepared operand addresses, operands of the PLO instance including one or more comparands, each comparand value prepared by fetching a predetermined location associated with the resource unit, each comparand value being changeable by another processor executing another PLO instance for changing the resource unit and, storing each prepared operand value, including each comparand value, for the operands of a PLO instance to be executed,

initiating execution of the PLO instance by the processor setting to a locked state a hardware lock associated with a blocking symbol specified by the PLO instance for locking the resource unit from being changed by any other processor in the computer system,

comparing each prepared comparand value with a current value of the comparand operand fetched from a predetermined location associated with the resource unit to test if the resource unit has been changed since the prepared comparand value was prepared,

if the comparing step finds an equal condition, generating store operand values of the PLO instance for changing the resource unit and generating the addresses in the resource unit at which the store operand values are to be stored, and storing the store operand values and the addresses in a save area of the processor executing the PLO instance,

setting a resource-inconsistency (RI) indicator of the processor'"'"'s PLO save area to a RI state in preparation for changing the resource unit,

accessing the resource unit at locations specified by the addresses stored in the processor'"'"'s save area to change the resource unit with the operand values stored in the processor'"'"'s save area in accordance with a function code (FC) specified in the PLO instance being executed,

resetting the RI indicator to a non-RI state after completion of changes of the resource unit, and

resetting to an unlocked state the hardware lock associated with the blocking symbol specified in the PLO instance to enable the resource unit to be accessed by any processor in the computer system.

15. A method of maintaining the integrity of data in a computer system against processor failure while executing a PLO instance of a PLO instruction, each PLO instance specifying a blocking symbol associated with a resource unit, each processor in the computer system capable of executing PLO instances having a PLO save area, and during execution of each PLO instance the PLO save area storing PLO operand values, and addresses of the operand values, which change the resource unit associated with the blocking symbol in an executing PLO instance, the method comprising the steps of:

detecting by an operational processor a processor identifier (PID) with a fail signal communicated from a failed processor in the computer system executing a PLO instance of a PLO instruction,

accepting by the operational processor of the communicated PID to temporarily become a recovery processor for preserving the consistency of the resource unit,

accessing by the recovery processor a resource-inconsistency (RI) indicator associated with the failed processor'"'"'s PLO save area, and detecting a setting of a RI indicator which may be set to an RI state to indicate the resource unit may be in an inconsistent state or to a non-RI state to indicate the resource unit is in a consistent state, the RI state existing only if the lock on the resource unit was in locked state when the processor failed,

if the RI indicator is found by the recovery processor to be in the RI state, restoring the resource unit to the consistent state by the recovery processor utilizing the contents of the failed processor'"'"'s PLO save area to change the resource unit in the same way as would have been done by the failed processor if it had not failed,

locating by the recovery processor a hardware lock entry for the resource unit, and resetting the lock entry to the unlocked state,

terminating the recovery operation, and the recovery processor returning to other operations.

16. A method of maintaining the integrity of data in a computer system against processor failure while executing instances of PLO instructions for changing resource units associated with blocking symbols specified in the PLO instructions, as defined in claim 15 in which the RI indicator of the failed processor'"'"'s PLO save area indicates the RI state, further comprising:

the restoring step copying into the resource unit operand values (found stored in the failed processor'"'"'s PLO save area) to associated resource addresses (also found stored in the PLO save area) as specified by a function code (FC) stored with the failed processor'"'"'s PLO save area.

17. A method of maintaining the integrity of data in a computer system against processor failure while executing instances of PLO instructions for changing resource units associated with blocking symbols specified in the PLO instructions, as defined in claim 15 in which the RI indicator of the failed processor'"'"'s PLO save area indicates the non-RI state, further comprising the steps of:

the restoring step not being performed, and the locking step resetting the hardware lock on the resource unit to unlocked state to allow the resource unit to be used in the consistent state by any processor in the computer system.

18. A method of maintaining the integrity of data in a computer system against processor failure while executing instances of PLO instructions for changing resource units associated with blocking symbols specified in the PLO instructions, as defined in claim 15, in which the lock must be found by the recovery processor before the lock can be set to the unlocked state, the lock finding process being dependent on whether the hardware locks in the computer system are in:

blocking-symbol-hash-located entries (hash entries), or in processor-associated entries (CLE entries), the method for locating hash entries further comprising the steps of;

scanning the PLO lock entries in a predetermined order to locate any lock entry containing the PID of the failed processor, and

if any lock entry is found to contain the PID of the failed lock entry, performing the steps of;

clearing the PID from the lock entry,

setting a lock field in the lock entry to an unlocked state, and

ending operations of the recovery processor for the failed processor'"'"'s PLO instance, since the resource unit is now enabled for use by another PLO instance.

19. A method of maintaining the integrity of data in a computer system against processor failure while executing instances of PLO instructions for changing resource units associated with blocking symbols specified in the PLO instructions, as defined in claim 15, in which the lock must be found by the recovery processor before the lock can be set to the unlocked state, the lock finding process being dependent on whether the hardware locks in the computer system are in:

blocking-symbol-hash-located entries (hash entries), or in processor-associated entries (CLE entries), the method for locating hash entries further comprising the steps of;

scanning the PLO lock entries in a predetermined order to locate any lock entry containing the PID of the failed processor, and

if no lock entry is found to contain the PID of the failed lock entry, performing the step of;

ending operations of the recovery processor for the failed processor'"'"'s PLO instance, since no lock entry was found for the failed processor.

20. A method of maintaining the integrity of data in a computer system against processor failure while executing instances of PLO instructions for changing resource units associated with blocking symbols specified in the PLO instructions, as defined in claim 15, in which the lock must be found by the recovery processor before the lock can be set to the unlocked state, the lock finding process being dependent on whether the hardware locks in the computer system are in:

blocking-symbol-hash-located entries (hash entries), or in processor-associated entries (CLE entries), the method for locating CLE entries further comprising the steps of;

obtaining a scan lock on all CLE PLO lock entries to allow only one processor at a time to access the lock entries,

accessing lock entry for the failed processor by using the PID of the failed processor,

testing a lock in lock entry to find if the lock is set to the locked state,

clearing a blocking symbol from the lock entry if found in the locked state by the test step,

determining if a waiting field in the lock entry contains any waiting CLE address,

setting the lock field to the unlocked state in the lock entry,

releasing the scan lock when the wait field does not contain any waiting CLE address, and

ending operations of the recovery processor for the failed processor'"'"'s PLO instance to enable the processor to perform other tasks.

21. A method of maintaining the integrity of data in a computer system against processor failure while executing instances of PLO instructions for changing resource units associated with blocking symbols specified in the PLO instructions, as defined in claim 15, in which the lock must be found by the recovery processor before the lock can be set to the unlocked state, the lock finding process being dependent on whether the hardware locks in the computer system are in:

blocking-symbol-hash-located entries (hash entries), or in processor-associated entries (CLE entries), the method for locating CLE entries further comprising the steps of;

obtaining a scan lock on all CLE PLO lock entries to allow only one processor at a time to access the lock entries,

accessing lock entry for the failed processor by using the PID of the failed processor,

testing a lock in lock entry to find if the lock is set to the locked state,

accessing a blocking symbol field in the lock entry if the lock is set to unlocked state,

releasing the scan lock if no blocking symbol is found in the lock entry, and

ending operations of the recovery processor for the failed processor'"'"'s PLO instance to enable the processor to perform other tasks.

22. A method of maintaining the integrity of data in a computer system against processor failure while executing instances of PLO instructions for changing resource units associated with blocking symbols specified in the PLO instructions, as defined in claim 15, in which the lock must be found by the recovery processor before the lock can be set to the unlocked state, the lock finding process being dependent on whether the hardware locks in the computer system are in:

blocking-symbol-hash-located entries (hash entries), or in processor-associated entries (CLE entries), the method for locating CLE entries further comprising the steps of;

obtaining a scan lock on all CLE PLO lock entries to allow only one processor at a time to access the lock entries,

accessing lock entry for the failed processor by using the PID of the failed processor,

testing a lock in lock entry to find if the lock is set to the locked state,

accessing a blocking symbol field in the lock entry if the lock is set to unlocked state,

finding that the failed processor was in a wait chain of CLEs waiting for the resource unit if a blocking symbol is found in the lock entry in an unlocked state,

removing the CLE of the failed process from the waiting chain for the resource, and

ending operations of the recovery processor for the failed processor'"'"'s PLO instance to enable the processor to perform other tasks.

23. A method of maintaining the integrity of data in a computer system against processor failure while executing instances of PLO instructions for changing resource units associated with blocking symbols specified in the PLO instructions, as defined in claim 22, in which the removing step further comprises the steps of:

scanning the CLEs in the locked state to find a CLE containing the blocking symbol specified in the CLE of the failed processor (failed CLE) in order to find the beginning CLE currently at the beginning of a waiting chain which must contain the failed CLE,

tracing the CLEs in the waiting chain starting with the beginning CLE to find the failed CLE and to identify a predecessor CLE and any successor CLE of the failed CLE,

removing the failed CLE from the waiting chain by changing the waiting field in the predecessor CLE to a CLE address of the successor CLE if a successor CLE exists in the waiting chain, or setting the waiting field in the predecessor CLE to indicate it is the last CLE in the waiting chain if no successor CLE is found by the tracing step,

clearing the blocking symbol and clearing the waiting field from the failed CLE,

releasing the scan lock, and

ending operations of the recovery processor for the failed processor'"'"'s PLO instance with the PLO lock entry set to a condition for enabling another PLO instance to request changing the resource.

24. A method of maintaining the integrity of data in a computer system against processor failure while executing instances of PLO instructions for changing resource units associated with blocking symbols specified in the PLO instructions, as defined in claim 23, in which the tracing step further comprises the steps of:

examining a waiting field in the beginning CLE to find a pointer to a next CLE, and then accessing any pointer in a waiting field in each next CLE in the waiting chain until the failed CLE is located, and identifying a predecessor CLE of the failed CLE and any successor CLE of the failed CLE while tracing the CLEs in the waiting chain, and no successor CLE existing if the failed CLE does not have any pointer in its waiting field.

25. A method of maintaining the integrity of data in a computer system against processor failure while executing instances of PLO instructions for changing resource units associated with blocking symbols specified in the PLO instructions, as defined in claim 24, in which the tracing step further comprises the steps of:

writing in the waiting field of the predecessor CLE a pointer to the successor CLE to remove the failed CLE from the waiting chain when the failed CLE is not the last CLE in the waiting chain.

26. A method of maintaining the integrity of data in a computer system against processor failure while executing instances of PLO instructions for changing resource units associated with blocking symbols specified in the PLO instructions, as defined in claim 24, in which the tracing step further comprises the steps of:

writing in the waiting field of the predecessor CLE a last CLE indication to remove the failed CLE from the queue if no successor CLE exists in the waiting chain, due to the failed CLE having been the last CLE in the waiting chain.

27. A method of maintaining integrity of data in a computer system when a processor in the computer system is executing a program containing a perform locked operation (PLO) instruction specifying a function for changing data in a resource unit stored in a computer system, the PLO instruction having operands for specifying where and what in the resource unit is to be changed by execution of the PLO instruction, the processor performing the steps of:

detecting an occurance of a failure by the processor while executing a PLO instance of a PLO instruction,

signaling to a predesignated entity in the computer system a processor failing indication with associated information including the PID of the failing processor and an indication that the failing processor is executing a PLO instance,

receiving by the predesignated entity the processor failing indication with the associated information,

sending by the predesignated entity a recovery signal with the associated information to one or more operational processors in the computer system, and one of the operational processors accepting the recovery signal to temporarily become a resource-consistency-recovery (RCR) processor for recovering the consistency of a resource subjected to the PLO instance in execution by the failing processor,

performing by the RCR processor the steps of setting the RI indicator to the RI state after the writing step and the also writing step have stored all of the store operand values and associated resource addresses in the processor'"'"'s save area,

copying each store operand value from the processor'"'"'s save area to the associated resource address in the resource unit while the RI indicator is set to the RI state for changing the resource unit,

setting the RI indicator of the processor'"'"'s save area to the non-RI state after the copying step has completed changing the resource unit by copying all of the store operand values from the processor'"'"'s save area into the resource unit for recovering the consistency of the resource and leaving the resource unit in an unlocked condition for the resource unit to be reliably used in an execution of a later PLO instance by any operational processor in the computer system,

terminating operations as a RCR processor to enable the processor to return to other processing tasks which may be unrelated to RCR operations.