An I/O interruption allows the processor to respond to signals from an I/O device that indicate either termination of a phase of I/O operations or external action on the device. A complete explanation of I/O interruptions is contained in the publication z/Architecture Principles of Operation. For descriptions of interruption by specific devices, see the IBM® publications for each device.

If error conditions are associated with an interruption, the system schedules the appropriate device-dependent error routine. The operating system might then start another request that is not related to the channel program in error. If the error recovery procedures (ERPs) fail to correct the error, the system places an error code in the IOB, IEDB, and then in the ECB, after appendages have been called.

A channel program might depend upon the successful completion of a previous channel program (as when one channel program retrieves data to be used in building another). The previous channel program is called a related request and must be identified to the system. For a description of this procedure, see IOBFLAGS in Input/Output Block (IOB) Fields and Purging and Restoring I/O Requests (PURGE and RESTORE macros).

If a permanent error occurs in the channel program of a related request, the system removes the request queue elements for all dependent channel programs and returns them to the caller without executing the request. For all requests dependent on the channel program in error, the system places completion codes in the ECBs. The system also places ones in the first two bit positions of the DCBIFLGS field of the DCB. Any new requests for a DCB with error flags are posted complete without execution. To reissue requests that depend on the channel program in error, reset the first two bits of the DCBIFLGS field of the DCB to zero. Then reissue EXCP for each desired channel program.