IBM RS/6000 SP, Problem Determination Manual

Page: 141 / 316

This soft copy for use by IBM employees only.
Figure 55. The flt File
The flt file has some nomenclature that differentiates device types, but it also
lists the device number in a field that is generically labeled “id.” With this in
mind, it was decided to differentiate switch IDs from nodes IDs by adding 100,000
to them. In this way, you cannot confuse a node with a switch, unless you have
more that 10,000 nodes in your system. Why 10,000 nodes? The last digit of the
switch ID tells you the chip on that switch. So, you are actually adding 10,000 to
the switch board number. This leaves 10,000 node IDs before you hit the first
switch ID at 100,000.
Note
• Therefore if the ID is < 100,000 you know that it is an adapter.
•
If the ID > 100,000 you know it is a switch.
You can translate the flt notation to out.top notation so that you can look in the
out.top file to determine the physical location of the connection.
The important information in the flt file is:
•
The device ID, which tells you which device reported seeing the fault.
•
For switch chips IDs, the WHY registers information. This is really important
to determine master/slave miscompares and which port is reporting the
problem. The actual error type in the WHY is of lesser importance. If the
chip can narrow down the information to a port, it will report WHY xx TO xx
TE xx DO xx RF xx FC xx DE xx, where TO, TE, DO, RF, FC, and DE are
registers for each different error category underneath that particular WHY
code. The register value, xx, indicates which port saw that error. Each bit
stands for a port. Port 0 is the least significant bit, and port 7 is the most
significant bit. If a particular error, such as TO, was not seen, then its
register value will be 00. If a TE was found on port 2, the value will be 04,
Chapter 4. The Switch
121