IBM System/370 145 Manual page 212

statistical data recorder (SDR), and the recording functions of MCAR,
CCH, tape error by volume (TEBV), and error volume analysis (EVA).
Machine Check Analysis and Recording
After IPL of a contr'Ol program c'Ontaining M'Odel 145 RMS routines,
mask bits are set to 'Ones and control register values are set t'O permit
machine check interruptions and l'Ogouts t'O occur as discussed for
os MCR.
When a System Rec'Overy machine check occurs to indicate a successful
microinstruction retry. an environment rec'Ord (recovery report), is
constructed by MCAR. The environment rec'Ord, which c'Ontains pertinent
status information from the fixed area, recovery action, pr'Ogram
identification, date, and time of day, is then passed by MCAR t'O RMSR t'O
be written in the recorder file, wh'Ose symb'Olic unit name is SYSREC
(corresponding to the SYS1.LOGREC recording data set of OS). The
operator is informed that a repressible machine check has occurred.
If a repressible err'Or condition occurs during execution of
nonprivileged code (a processing
program~
for example), err'Or recording
takes place when the interruption Occurs. If the repressible error
condition occurs during execution of privileged code (supervisor c'Ode,
for example), the CPU is disabled f'Or machine checks due to repressible
conditions, an indicati'On that MCAR processing is required is st'Ored,
and control is returned t'O the interrupted privileged code. When
execution of the privileged code terminates, contr'Ol is given t'O MCAR to
record the repressible error condition. Machine check statistics are
lost during the interval between occurrence of the repressible error and
its recording if an'Other machine check c'Ondition arises. Specifically,
a machine check interrupti'On for an exigent c'Onditi'On during this
interval will overlay the machine check logout f'Or the repressible error
condition, and any additi'Onal repressible machine check conditions will
result in the loss of all but 'One of the repressible errors, since the
Model
145
can keep only one machine check pending at a time for each
machine check type.
Prior to relinquishing ·CPU control, MCAR determines whether or not an
automatic mode switch from recording mode to quiet (nonrecording) m'Ode
should
be
made for microinstruction retry recoveries. Quiet mode is
established by HCAR (the System Recovery mask bit is turned off) if the
number of microinstruction retry corrections that occur during a given
time interval exceeds the established error c'Ount threshold value for
these corrections. (The time of day cl'Ock is used by MCAR to determine
whether or not the time interval threshold has been exceeded.) The IBM-
supplied threshold values can be altered during system generati'On or by
the operator MODE c'Ommand during system 'Operation. The 'Operat'Or is
informed of any mode switch made by MCAR and can switch back t'O
recording mode any time thereafter. Quiet mode can be used t'O prevent
SYSREC from being filled with CPU recovery reports when a large number
of transient errors are 'Occurring.
As
described for the OS MCH r'Outine, MCAR also supports an 'Operat'Or
MODE command to permit the operator t'O enable the CPU for interruptions
after single-bit intermittent process'Or and contr'Ol storage correcti'Ons
so that these errors can be logged.
If an error damages the time of day clock, MCAR places the system in
quiet mode for microinstruction retry and ECC c'Orrecti'Ons, a message is
issued to the 'Operator, and processing c'Ontinues. If the interval timer
is damaged, a message is issued and pr'Ocessing continues.
When an Instruction Processing Damage machine check occurs
(uncorrectable or unretryable CPU error, double-bit storage err'Or, or
storage protect key err'Or) during the execution of supervisor (or any
202 A Guide to the IBM System/370 M'Odel 145