Control unit threshold timeout controls for software missing interrupt handlers in operating systems

• US 5,758,190 A
• Filed: 04/09/1996
• Issued: 05/26/1998
• Est. Priority Date: 04/09/1996
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Ripberger, Richard Anthony, Yudenfriend, Harry Morris, Johnson, Gavin Stuart, Urbanejo, Luis Ricardo
• Primary Examiner(s): AUVE, GLENN ALLEN
• Application Number:
  US08/629,702
• Time in Patent Office: 777 Days
• Field of Search: 395/868, 395/856, 395/838, 395/839, 395/726, 395/728, 395/733, 395/739, 395/735, 395/869
• US Class Current: 710/48
• CPC Class Codes: G06F 11/0757 by exceeding a time limit, i.e. time-out, e.g. watchdogs

Abstract

I/O control unit (CU) features for supporting multiple host operating systems (OSs) which use missing interrupt handler (MIH) timeout functions for detecting potential failures of requested I/O device operations. These CU features support multiple host OSs by preventing them from falsely indicating I/O device failures, when in fact the device has not failed but is merely processing one or more other requests for other host Oss.

Claims

1. Support controls within an I/O control unit (CU) for operating with a plurality of host system adapter interfaces located in the CU for connecting the CU to host systems that send I/O requests to the CU and use timeout detection for determining when a requested I/O device attached to the CU has failed to perform a requested I/O operation, the CU comprising:

a plurality of device lock bits respectively associated with I/O devices connected to the CU to control a connection of any of the I/O devices to any of the adapter interfaces for an I/O operation,

a busy bit, a threshold timer associated with each of the adapter interfaces, and a stored threshold timeout value for operating with the timer for determining occurrence of a threshold timeout,

bit testing controls for each of the adapter interfaces for attempting to set to a lock state an available one of the lock bits associated with an I/O device required by a current I/O request received by the adapter interface which may contend for the lock state with a current I/O request received by any other of the adapter interfaces simultaneously attempting to obtain use of the associated I/O device,

busy bit controls associated with the adapter interface for setting the associated busy bit to a busy state when a waiting I/O request is received by the adapter interface while the associated lock bit is set to the locked state for a current I/O request using the associated I/O device,

timer starting controls for resetting and starting the timer associated with the adapter interface when the associated lock bit is set to a lock state,

long busy controls in the CU associated with each adapter interface for generating a long busy signal to the adapter interface if the associated timer exceeds the threshold timeout while a locked state exists for the lock bit, and the long busy controls transferring the long busy signal from the adapter interface for transmission external to the CU, and

operation completion controls associated with each lock bit for generating a no-longer long-busy signal when the device lock is set to an unlocked state after the threshold timeout is exceeded, and the operation completion controls transferring the no-longer long busy signal to the adapter interface for transmission external to the CU.

2. An I/O control unit (CU) having multiple adapter interfaces as defined in claim 1, further comprising:

regular busy controls in the CU associated with each adapter interface for generating a regular busy signal to the adapter interface when the associated timer does not exceed the threshold timeout while the locked state exists for the lock bit when another I/O request is received by the adapter interface for the associated I/O device, and

operation completion controls associated with each lock bit for generating a no-longer regular-busy signal when the device lock is reset to an unlocked state before the threshold timeout is exceeded upon completion of the current I/O request, and the operation completion controls transferring a no-longer busy signal from the adapter interface for transmission external to the CU to indicate that the I/O device is available for another I/O request.

3. An I/O control unit (CU) having multiple adapter interfaces as defined in claim 2, further comprising:

the I/O device successfully completing operation for the current I/O request after the threshold timeout and then signalling the CU to reset the associated lock bit to an available state, and

the CU generating a device end interrupt signal and status information for enabling the adapter interface to communicate the availability of the I/O device to external software advising that any next I/O request may be presented to the CU for contending for the lock state of the lock bit associated with the I/O device.

4. An I/O control unit (CU) having multiple adapter interfaces as defined in claim 2, further comprising:

the I/O device unsuccessfully completing operation for the current I/O request after the threshold timeout and then failing to signal any completion interrupt to software of a host system making the current I/O request wherein a missing interrupt handler timer controlled by the software exceeds the extended timeout to indicate a failure of operation by the I/O device for the current I/O request.

5. An I/O control unit (CU) having multiple adapter interfaces as defined in claim 1, further comprising:

controls of the adapter interface generating a signal and status information representing the long-busy signal to the adapter interface when the threshold timeout is exceeded, the long-busy signal enabling the adapter interface to communicate non-availability of the I/O device to external software in order to advise the external software to extend a missing-interrupt-handler timeout for preventing a false timeout detection which would indicate operation failure by the associated I/O device in handling the current I/O request when the I/O device has not failed, and

the lock bit remaining in locked state and the I/O device continuing operation for the current I/O request after the threshold timeout has been exceeded.

6. An I/O control unit (CU) having multiple adapter interfaces as defined in claim 1, further comprising:

controls of the adapter interface generating an interrupt signal and status information representing the long-busy signal to the adapter interface when the threshold timeout is exceeded for enabling the adapter interface to communicate a long-busy interrupt signal to host system software sending the waiting I/O request to the adapter interface in order to advise the host system software that the CU no longer takes responsibility for handling the waiting I/O request and to enable the host system software to stop any missing interrupt handler timeout operation for the waiting I/O request which can again be requested by the host system software to the CU for a future operation by the required I/O device, and

the lock bit and associated I/O device continuing operations for the current I/O request at the adapter interface.

7. An I/O control unit (CU) having multiple adapter interfaces as defined in claim 6, further comprising:

the associated timer terminating an operation after the interrupt signal and status information have been communicated from controls of the adapter interface.

8. An I/O control unit (CU) having multiple adapter interfaces as defined in claim 1, further comprising:

the controls of each adapter interface continuing a timing operation in the CU by the associated timer after the threshold timeout is exceeded by replacing the threshold timeout value with a second threshold timeout value, the second threshold timeout to support communication to external software of a further extension to the long-busy timeout for use by a missing interrupt handler timeout in the external software for avoiding a false detection of I/O device failure for the current I/O request by increasing granularity of the timeout extensions, and

the controls of each adapter interface generating an interrupt signal and status information to communicate to the external software the need for extending the long-busy signal when the second threshold timeout is exceeded while the associated lock bit continues its locked state and the associated I/O device continues its operation for the current I/O request.

9. A method within an I/O control unit (CU) for supporting I/O device failure detection by any of a plurality of host systems which may be connected to I/O interface adapters in the CU that receives I/O requests from the host systems, the CU method communicating information from the CU to the host systems to support missing interrupt timeout detection in the host systems in a manner that prevents any host system from falsely indicating operational failure by a requested I/O device attached to the CU, the CU performing the steps of:

associating a plurality of lock bits with respective I/O devices connected to the CU to control a connection of any of the I/O devices to any of the adapter interfaces for an I/O operation,

structuring with each interface adapter a busy bit, a threshold timer, and a stored threshold timeout value for operating the timer to determine an occurrence of a threshold timeout,

conditionally testing by any interface adapter the state of the lock bit associated with an I/O device required by a current I/O request received by the interface adapter while simultaneous I/O requests are being made for plural interface adapters trying to obtain use of the same I/O device,

setting the associated busy bit to a busy state when a waiting I/O request is received by the interface adapter while the lock bit is set to the locked state for a current I/O request using the associated I/O device,

resetting and starting the timer for the interface adapter when a tested lock bit is set to a lock state,

generating a long busy signal for the interface adapter if the timer exceeds the threshold timeout while a locked state exists for the lock bit, and transferring the long busy signal from the interface adapter to a host system requesting the I/O device to communicate that the I/O device remains busy with a current I/O request obtaining the lock state, and

outputting a no-longer long-busy signal from the interface adapter to indicate the I/O device has completed a requested I/O operation when the device lock is set to an unlocked state after the threshold timeout is exceeded.

10. A method within an I/O control unit (CU) having multiple interface adapters for supporting I/O device failure detection by any of a plurality of host systems as defined in claim 9, the method further comprising the steps of:

providing a regular busy signal to the interface adapter when the associated timer does not exceed the threshold timeout while the locked state exists for the lock bit when another I/O request is received by the interface adapter for use of the associated I/O device, and

generating a no-longer regular-busy signal on completion of the current I/O request before the threshold timeout is exceeded, and then resetting the lock bit to an unlocked state, and transmitting the no-longer busy signal from the interface adapter to an external system to indicate that the I/O device is available for another I/O request.

11. A method within an I/O control unit (CU) having multiple interface adapters for supporting I/O device failure detection by any of a plurality of host systems as defined in claim 10, the method further comprising the steps of:

successfully completing operation by the associated I/O device for the current I/O request after the threshold timeout, then signalling the CU to reset the lock bit associated with the I/O device to an available state, and

providing a device-end interrupt signal and status information for enabling the interface adapter to communicate the availability of the I/O device to a host system to advise that any next I/O request for the I/O device may be again presented to the CU to contend for obtaining the lock state of the lock bit associated with the I/O device.

12. A method within an I/O control unit (CU) having multiple interface adapters for supporting I/O device failure detection by any of a plurality of host systems as defined in claim 10, the method further comprising the steps of:

unsuccessfully completing operation by the I/O device for the current I/O request after the threshold timeout, wherein the host system continues a secondary timeout operation until the secondary timeout operation exceeds a secondary timeout value to indicate in a missing interrupt handler a failure of operation by the I/O device for the current I/O request.

13. A method within an I/O control unit (CU) having multiple interface adapters for supporting I/O device failure detection by any of a plurality of host systems as defined in claim 9, the method further comprising the steps of:

generating by the interface adapter a long-busy signal and status information representing the long-busy signal when the threshold timeout is exceeded while the lock state exists, the long-busy signal enabling the interface adapter to communicate non-availability of the I/O device to a host system in order to advise the host system to extend a timeout for a missing-interrupt-handler in order to prevent a false timeout detection by the host system which would falsely indicate an operational failure by the associated I/O device in handling the current I/O request when the I/O device operation has not failed, and

continuing the locked state and the I/O device operation for the current I/O request after the threshold timeout has been exceeded.

14. A method within an I/O control unit (CU) having multiple interface adapters for supporting I/O device failure detection by any of a plurality of host systems as defined in claim 9, the method further comprising the steps of:

generating a long-busy interrupt signal and status information for the interface adapter when the threshold timeout is exceeded for enabling the interface adapter to communicate the long-busy interrupt signal to the host system sending the waiting I/O request to the interface adapter in order to advise the host system that the CU no longer will handle the waiting I/O request and that the host system should stop any missing interrupt handler timeout operation for the waiting I/O request and to again presented the request to the CU in the future when the required I/O device becomes available, and

continuing the locked state for the lock bit and operations for the associated I/O device for the current I/O request associated with the interface adapter.

15. A method within an I/O control unit (CU) having multiple interface adapters for supporting I/O device failure detection by any of a plurality of host systems as defined in claim 14, the method further comprising the steps of:

terminating operation by the timer associated with the interface adapter after the interrupt signal and status information have been communicated by controls of the interface adapter to the host system making the I/O request.

16. A method within an I/O control unit (CU) having multiple interface adapters for supporting I/O device failure detection by any of a plurality of host systems as defined in claim 9, the method further comprising the steps of:

continuing a timing operation for the interface adapter by the timer associated with the interface adapter after the threshold timeout is exceeded by replacing the threshold timeout value with a second threshold timeout value, the second threshold timeout supporting a communication to the host system by the interface adapter of a further extension to the long-busy timeout for use by a missing interrupt handler timeout in the host system for avoiding a false detection of I/O device failure for the current I/O request by increasing granularity of the timeout, and

generating an interrupt signal and status information to communicate from the interface adapter to a host system the need for an extended-long-busy signal when the second threshold timeout is exceeded while the lock bit associated with the I/O device continues in the locked state and the I/O device continues to operate for the current I/O request.

17. Support controls within an I/O control unit (CU) for operating with one or more host operating systems (OSs) that send I/O requests to the CU for an operation by a requested I/O device attached to the CU, the CU comprising:

an operation code signal for indicating an I/O command being currently performed by the I/O device for an I/O program associated with an I/O request from an OS,

a command detector for testing the operation code signal to determine whether the command is a long command,

an interrupt signal generator for generating a long-busy interrupt signal if the command detector determines the command is a long command,

sending the long-busy interrupt signal to a processor for communicating the long busy interrupt signal to the OS which issued the I/O request to enable the OS to extend a primary missing interrupt handler (MIH) timeout period to a secondary MIH timeout period before the end of the primary MIH timeout period for preventing the primary MIH timeout period from ending with a timeout which would falsely indicate a potential I/O device operation failure.