IBM System/370 Manual page 96

EXPANDED MACHINE CHECK INTERRUPT FACILITIES
Implementation of the machine check level of interrupt on the
System/370 Model 165 has been expanded in order to enhance error
recording and error recovery procedures. Programming support of the
extended machine check interrupt is provided by the Model 165 MCH
routine.
The machine check interrupt facilities of the Model 165 differ from
those of Models 65 and 75 as follows.
• Four subclasses of machine check interrupt are defined.
• A machine check interrupt occurs to permit the recording of
errors corrected by the hardware (soft machine check interrupt)
as well as to allow recovery routines to handle errors that
cannot be corrected by hardware (hard machine check interrupt).
• Machine check interrupt masking is expanded to handle selective
disabling and enabling of the interrupt subclasses defined,
the CPU retry facility, and the occurrence of extended logouts.
• The size of the fixed storage area in lower processor storage
is increased to accommodate the storing of additional machine
status and diagnostic information when a machine check interrupt
occurs.
• A hard stop error condition is defined that causes the Model
165 system to stop functioning immediately because the nature
of the machine malfunction prevents valid processing from
continuing.
• A hang detect check is implemented that causes a machine check
interrupt if a CPU operation has not been completed within 16.6 ms.
The Model 165 presents one of four subclasses of machine check
interrupt, depending on the specific machine malfunction. Each
interrupt subclass is maskable and causes either a soft machine check
or a hard machine check interrupt when enabled. A soft machine check
occurs after the hardware has been successful in correcting an error
or when the error does not affect the interrupted program. This is
done so that the failure can be recorded. System operation continues
after the error is logged. For example. if an error occurs during
the execution of an instruction and if the CPU retry hardware corrects
the error by reexecuting the failing instruction, a soft machine check
interrupt occurs at the completion of the successful execution.
A hard machine check occurs when hardware retry fails or is not
possible. For example, if the CPU retry hardware has not corrected
a failing instruction after seven retries. a hard rather than a soft
machine check interrupt occurs after the last unsuccessful retry.
Fiqure 50.10.3 shows the layout of fixed processor storage in the
Model-165. Fixed storage consists of three areas: the fixed locations
in decimal addresses 0-127, the fixed loqout area in locations 176-
511, and the CPU extended logout area of 1000 bytes. which normally
begins at location 512 unless the logout pointer is altered by
programming.
Fixed locations 0-127 are identical in layout and content to these
locations in System/360 models, with the exception of the EBCDIC/ASCII
bit in the current PSW, which must be set to zero.
A logout to the fixed logout area (176-511) occurs when any type
of machine check interrupt is taken. The data stored varies slightly
74