Service manual

5-26 AlphaServer 4000/4100 Service Manual
5.3.4 MCHK 660 IOD-Detected Failure (System Bus Error)
The error log in Example 5-4 shows the following:
➊ CPU0 logged the error in a system with two CPUs.
➋ The External Interface Status Register does not record an error.
➌ Both IOD CAP Error Registers logged an error.
➍ The MC Error Info Registers 0 and 1 captured the error information.
➎ The commander at the time of the error was CPU3 (known from
MC_ERR1).
➏ The command on the bus at the time was a write-back memory command.
Since this is an MCHK 660, the IOD detected the error on the bus, and CPU0 is
logging the error. CPU0 registers are not important in this case since it is servicing
the IOD interrupt. There are three devices that can put data on the system bus:
CPUs, memory, or an IOD. From MC_ERR Register 1 we know that at the time of
the error CPU3 put bad data on the bus while writing to memory. See Section 5.4
for a procedure designed to help with IOD-detected errors.
NOTE: The error log example has been edited to decrease its size; registers of
interest are in bold type. The “Horse” module referred to in the error log is the
system bus to PCI bus bridge module, the B3040 module. The “Saddle” module is
the PCI motherboard, the B3050 module. The “MC” bus is the system bus.
Refer to Table 5-9 for information on decoding commands, and refer to Table 5-10
for information on node IDs.