When Load command from cursor fails with SQL0437W and user is
prompted with c/d/t messages, and upon selecting any option,
db2bp might try to double free the memory that has already been
freed vis clpPostLoadFreeDynMem().
Error while running load:
Do you want to continue(c), terminate this device only(d), abort
the utility(t) ? (c/d/t) c
SQL0437W Performance of this complex query may be sub-optimal.
Reason code: "1". SQLSTATE=01602
Do you want to continue(c), terminate this device only(d), abort
the utility(t) ? (c/d/t)
In db2diag.log we can see entries as below:
012-02-22-17.49.36.688978+060 I1083415A1469 LEVEL: Severe
PID : 53477450 TID : 1 PROC : db2bp
INSTANCE: db2inst1 NODE : 000
APPID : *N0.db2inst1.120222164343
EDUID : 1
FUNCTION: DB2 UDB, SQO Memory Management,
sqloDiagnoseFreeBlockFailure, probe:10
MESSAGE : Possible memory corruption detected.
DATA #1 : ZRC, PD_TYPE_ZRC, 4 bytes
0x820F0002
DATA #2 : Corrupt block address, PD_TYPE_CORRUPT_BLK_PTR, 8
bytes
0x000000011011aee0
DATA #3 : Block header, PD_TYPE_BLK_HEADER, 24 bytes
0x000000011011AEC8 : FAB0 0000 0011 0114 0000 0001 1011 8148
...............H
0x000000011011AED8 : 0000 0000 0000 0000
........
DATA #4 : Data header, PD_TYPE_BLK_DATA_HEAD, 48 bytes
0x000000011011AEE0 : 0000 0000 0000 9120 0001 0000 0000 0000
....... ........
0x000000011011AEF0 : 0000 0000 0002 0000 0000 0000 0000 0000
................
0x000000011011AF00 : 0003 0000 0000 0000 0000 0000 0004 0000
................
CALLSTCK:
[0] 0x0900000002FF6CD4 pdLog + 0xA8
[1] 0x0900000003175E2C
@28@sqloDiagnoseFreeBlockFailure__FP8SMemFBlk + 0x15C
[2] 0x0900000003838ED4 sqlofmblkEx + 0x88
[3] 0x0900000002F5E46C clpPostLoadFreeDynMem + 0xB0
[4] 0x00000001000422B0 clpbp_exe_sqluload__Fv + 0x130
[5] 0x0000000100001794 clp_bp_con__Fv + 0x634
[6] 0x0000000100000B24 main + 0x224
[7] 0x0000000100000320 __start + 0x98
[8] 0x0000000000000000 ?unknown + 0x0
[9] 0x0000000000000000 ?unknown + 0x0
2012-02-22-17.49.36.689490+060 I1084885A337 LEVEL: Severe
PID : 53477450 TID : 1 PROC : db2bp
INSTANCE: db2inst1 NODE : 000
APPID : *N0.db2inst1.120222164343
EDUID : 1
FUNCTION: DB2 UDB, SQO Memory Management,
sqloDiagnoseFreeBlockFailure, probe:30
DATA #1 : String, 12 bytes
Double free.
The issue happened because once the memory was allocated
successfully and also de-allocated by OSS, then clp tried
freeing the same memory again and again after supplying the
input i.e (c/d/t). clp_bp was reading the input queue again and
again which was duplicating the memory pointer, because of which
there was attempt to free the memory
pointer again and again, hence the double free.
In the traces we will have the following entries:
1779 | | clp_bp_con entry
849439 | | | clp_bp_exe entry
849442 | | | | clpbp_exe_sqluload entry
849523 | | | | | clppostloadfreedynmem entry
849523 entry DB2 UDB command line process clppostloadfreedynmem
fnc (1.3.41.788.0)
pid 3068074 tid 1 cpid 1221 node 0 sec 4 nsec 419553800
bytes 64
Data1 (PD_TYPE_PTR,8) Pointer:
0x0000000110046ff0
Data2 (PD_TYPE_HEXDUMP,40) Hexdump:
0000 0000 0000 0000 0000 0000 0000 0000 ................
0000 0000 0000 0000 0000 0001 1011 BC40 ...............@
0000 07D2 0000 0000 ........ |
