page of GC20-1807-7 As Updated April 1, 1981 by TNL GN25-0829 ZAP makes use of control records to control processing. These
records can be submitted either from the terminal or from a disk file. Using the VER and REP control records, you can verify and replace
records or instructions in a control section (CSECT). Using the DUMP control record, you can dump all or part of a CSECT, an entire member of
a LOADLIB or TXTLIB file, or an entire MODULE file.
Debugging CP on a Virtual Machine Many CP problems can be isolated without standalone machine testing. It
is possible to debug CP by running it in a virtual machine. In most
instances, the virtual machine system is an exact replica of the system
running on the real machine. To set up a CP system on a virtual
machine, use the same procedure that is used to generate a CP system on
a real machine. However, remember that the entire procedure of running
service programs is now done on a virtual machine. Also, the virtual
machine must be described in the VM/370 directory. See VML11Q in g Virtual for directions on how to set up
the virtual machine.
CP Internal Trace Table CP has an internal trace table that records events that occur in the
real machine. The events that are traced are: External interruptions SVC interruptions Program interruptions Machine check interruptions I/O interruptions Free storage requests Release of free storage Entry into scheduler Qlleue drop Run user requests Start IIO Unstack I/O interruptions Storing a virtual CSW Test I/O Halt Device Unstack IOBLOK or TRQBLOK NCP BTU (Network Control Program Basic Transmission Unit) Spinning on a lock (attached processor environment) SIGP (X'13') Clear Channel instruction
An installation may optionally specify the size of
table. To do so, use the SYSCOR macro instruction in
Information on using this macro instruction is in the ang QgngratiQ!! the CP trace
module DMKSYS. !UJIQ-Planning· If an installation does not specify the trace table size or the size
specified is smaller than the default size, CP assigns the default size.
For each 256K bytes (or part thereof) of real storage available at IPL time, one page (4096 bytes) is allocated to the CP trace table.
Each entry in the CP trace table is 16 bytes long. There are trace
table entries for each type of event recorded. The first byte of each
42 IBM VM/370 System programmer's Guide
trace table entry, the identification code, identifies the type of event
being recorded. Figure 8 describes the format of each type of trace
table entry. Some trace table entries are generated by ECPS:VM. The first bit of these entries is set to 1 to indicate the entry was generated by the
hardware assist. For example, a trace table entry of type X'86' (FREE)
is the same as an entry of type X'06'. The only difference is that the
first entry was generated by the hardware assist.
The trace table is allocated by the main initialization routine, D8KCPI. The first event traced is placed in the lowest trace table
address. Each subsequent event is recorded 1n the next available trace
table entry. Once the trace table is full, events are recorded at the
lowest address (overlaying the data previously recorded there). Tracing
continues with each new entry replacing an entry from a previous cycle. Use the trace table to determine the events that preceded a CP systea failure. An abend dump contains the CP internal trace table and the
Fointers to it. The address of the start of the. trace table, TRACSTRT, is at location X'OC'. The address of the byte the end of the
trace table, TRACEND, is at location X'10'. And the address of the next
available trace table entry, TRACCURR, is at lecation X'14'. Subtract
16 bytes (X'10') from the address stored at X'14' (TRACCUBR) to obtain
the trace table entry for the last event completed.
The CP internal trace table is initialized during IPL. If you do not
wish to record events in the trace table, issue the MONITOR STOP command to suppress recording. The pages allocated to the trace table are not
released and recording can be restarted at any time by issuing the MONITOR START command. If the V8/370 system should abnormally terminate
and automatically restart, the tracing of events on the real machine
will be active. After a V8/370 IPL (manual or automatic), CP internal
tracing is always active. Part 1. Debugging vith 15/370 43
Previous Page Next Page