Yesterday I needed to turn on gtf tracing to catch a problem involving about 80 
address spaces. I am tracing user records issued at a really rapid rate by 
those 80 asids.

Unfortunately GTF terminates itself with a dump of title DUMP OF GTF MODULE 
AHLTUSR. Again unfortunately, that dump does not contain any system trace 
table. All I have for debugging is the (2) logrec entries for this:

PSW: 04041001 80000000 00000000 05D02EB8                             
   INSTRUCTION LENGTH: 06   INTERRUPT CODE: 0038                        
   FAILING INSTRUCTION TEXT: 0090B904 0085D9F8 66B4E000                 
   TRANSLATION EXCEPTION ADDRESS: 1E909E74_00023000                     
                                                                        
   BREAKING EVENT ADDRESS: 00000000_05D02E90                            
   AR/GR 0-1    008FE040/00000000_01FF1EA8  00000000/0000001C_1C000000  
   AR/GR 2-3    00000000/20202020_021AB028  00000000/0C0C0C0C_02184400  
   AR/GR 4-5    00000000/7F3EF44E_02184988  00000000/00000000_0000008D  
   AR/GR 6-7    00000000/00000000_02184400  00000000/7F3EF5A4_00000000  
   AR/GR 8-9    00000000/00000000_0000008D  00000000/24080028_00F96650  
   AR/GR 10-11  00000000/00000000_8296F76E  00000000/00000000_02184988  
   AR/GR 12-13  00000000/00000000_05D027A8  00000000/00000000_0244A048  
   AR/GR 14-15  00000000/1E909E74_00023858  00000000/00000000_00000100  
                                                                        
   HOME ASID: 00A1    PRIMARY ASID: 00A1    SECONDARY ASID: 00A1        
   PKM: 00C0          AX: 0000              EAX: 0000                   
                                                                        
   THIS TASK'S ASID/TCB: 00A1/008E33D0                                  

The error occurs on an MVCK instruction. The problem is R14, which contains 
'1E909E74_00023858', obviously a 64bit address that isn't getmained. The code 
that does the gtrace macros reliably fails on a 1.13 system, but runs on 1.12 
and below. There is no maintenance that I could find that describes such an 
abend. UA60894 (the only one that might fit despite running only one instance 
of gtf) is installed.

Has anbody else seen this? Or has an idea how RE could get corrupted by invalid 
gtrace invocations? Or is this a real, new bug, and I have to find someone who 
would open a PMR (we are a business partner running on an ADCD system and don't 
have a customer number to open PMRs with).

Thanks in advance, Barbara

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to