If your application accesses an IBM Informix Server engine
running on Windows and the 'set explain on' command is used in
the application, it can cause a corruption in shared memory
alocated by the engine - a content of a memory block can
overflow to the subsequent block, corrupting its header.
Depending on how the subsequent block is used by the engine one
can see various types of AF events, for example:
02:32:59 Explain file for session 1274 :
d:informixsqexplninformix.out
02:32:59 Found during mt_shm_malloc_segid 2, error seen 1 times
since last print to the log
02:32:59 Pool 'mt' (0x1477ca040)
02:32:59 Bad block header 0x16e2a4420
02:32:59 Found during mt_shm_malloc_segid 2, error seen 1 times
since last print to the log
02:32:59 Pool 'mt' (0x1477ca040)
02:32:59 Bad block header 0x16e2a4420
02:32:59 Assert Failed: Memory block header corruption detected
in mt_shm_malloc_segid 2
02:32:59 IBM Informix Dynamic Server Version 12.10.FC5WE
02:32:59 Who: Session(1274, informix@hostname, 5632,
0000000000000000)
Thread(5188, sqlexec, 0, 10)
File: mtshpool.c Line: 1937
02:32:59 Results: Unable to repair pool
02:32:59 Action: Please notify IBM Informix Techical Support.
02:32:59 stack trace for pid 1200 written to
d:informix mpaf.182ceb4a
02:33:02 See Also: d:informix mpaf.182ceb4a
or
02:21:31 Explain file for session 856 :
d:informixsqexplninformix.out
...
02:27:29 Found during mt_shm_malloc_segid 8, error seen 1 times
since last print to the log
02:27:29 Pool 'mt' (0x1445d4040)
02:27:29 Bad free block 0x148334d38
02:27:29 Assert Failed: Unhandled NT Exception!
02:27:29 See Also: d:informix mpaf.10ace110
02:27:29 Assert Failed: No Exception Handler
02:27:29 IBM Informix Dynamic Server Version 12.10.FC5WE
02:27:29 Who: Session(856, informix@hostname, 1652,
0000000000000000)
Thread(3268, sqlexec, 0, 12)
File: mtex.c Line: 1157
02:27:29 Results: Exception Caught. Type: MT_EX_OS, Context:
mem
02:27:29 Action: Please notify IBM Informix Techical Support.
02:27:29 stack trace for pid 3496 written to
d:informix mpaf.10ace110
02:27:30 See Also: d:informix mpaf.10ace110
02:27:46 No Exception Handler
The stack trace of the failing thread can vary, but at the top
of it the functions for allocating memory can always be seen,
for example:
afhandler
affail_interface
mt_shm_malloc_segid
mt_malloc
mt_create_mutex
ha_init_hash_table
init_hash_structs
group_open
....
or
recover_pool_bad_free
mt_shm_malloc_segid
mt_malloc
mt_create_mutex
rsla_open
fmla_write
insertla_next
doselintotmp
...
Another symptom is that the text of the SQL statement run by the
failing session is incomplete in the explain file:
QUERY: (OPTIMIZATION TIMESTAMP: 11-03-2015 16:18:25)
------
select
sysrstcb.indx,
sysrstcb.address,
sysrstcb.txp,
... (skipped some lines)
sysrstcb.tmpsize,
sy |
