Product Fault Management,
Continued
NOTE
Although the VMS error handler has built-in features
to aid services in memory repair, good judgment is
needed by the Service Engineer. It is essential to
understand that in many, if not most cases, correctable
ECC errors are transient in nature. No amount of repair
will fix them, as generally there is nothing to be fixed.
Uncorrectable
ECC Errors
For uncorrectable ECC errors, a memory subpacket is logged as
indicated by memory subpacket listed in the third column of the
FLAGS software register (
!
). Also, the hardware register MESR
<11> (
"
) of the processor register subpacket is set equal to 1, and
MEAR latches the error address (
&
).
Examine the MEMCON software register (
#
) under the memory
subpacket. The MEMCON register provides memory configuration
information and a MEMORY ERROR STATUS buffer (
$
) that
points to the memory module(s) that is the most likely FRU.
Replace the indicated memory module.
The VMS error handler marks each page bad and attempts page
replacement, indicated in SYSTAT (
%
). The DCL command
SHOW MEMORY also indicates the result of VMS page
replacement.
Uncorrectable memory errors increment the VMS global counter,
which can be viewed using the DCL command SHOW ERROR.
NOTE
If register MESR <11> was set equal to 1, but MESR
<19:12> syndrome equals 07, no memory subpacket
will be logged as a result of incorrect check bits
written to memory because of an NDAL bus parity error
detected by the NMC. In short, this indicates a problem
with the CPU module, not memory. There should be a
previous entry with MESR <22>, NDAL data parity error
set equal to 1.
NOTE
An uncorrectable ECC error due to a ‘‘disown write’’
results in a CRD entry like those for correctable ECC
Continued on next page
5–73
