Error handling

The DFHSR1 module gets control from the kernel or from other AP domain modules. It determines whether it is dealing with an internal error, or an external error such as a program check. Internal errors are dealt with by exiting from the recovery environment and issuing the appropriate kernel call. If the DFHXFP or DFHEMS module has caused a program check, the DFHSR1 module exits from the recovery environment and passes control to DFHXFP or DFHEMS, respectively. All other external errors are passed on to the DFHSRP module. If control returns from the DFHSRP module, DFHSR1 issues a transaction abend. If control returns from the abend call, it is because the XPCTA exit has requested retry; in which case, DFHSR1 restores the registers and other information, and branches to the resume address.

The DFHSRP module makes an exception trace entry, ensures that it is running on the QR TCB and then deals with one of the following:
Note: The kernel recovery environment is terminated very soon after DFHSRP receives control. This ensures that DFHSRP gets driven again if a subsequent error occurs in DFHSRP itself (rather than the kernel percolating the error to the next kernel stack entry). DFHSRP is therefore in a position to detect such recursive errors, and can take the appropriate action.

If DFHSRP can abend the transaction, it builds a Transaction Abend Control Block (TACB) to describe the abend. The TACB is a task-lifetime control block that records details of a transaction abend. This TACB may be used by the rest of AP domain that needs information about the abend. DFHSRP builds the TACB, rather than letting Program Control build it as part of DFHPC TYPE=ABEND processing. This enables DFHSRP to include extra information in the TACB that would otherwise be lost, such as GP registers, PSW, and FP registers at the time of the error.



dfhs386.html | Timestamp icon Last updated: Monday, 5 June 2017