07-Jul-2005

More on the 8.2 crash

Apparently the crash that I saw in 8.2 is caused by a very small timing window in $GETJPI, where the system service goes to get information about the process and the process is no longer there.

Engineering tells me that the fix will be in the first UPDATE patch for 8.2 (or was it the first SYS patch? As I'm not running 8.2 I didn't write it down).

Footprint:


Failing Instruction:
OTS$MOVEM_C+00034:  	LDQ_U		R28,(R18)

Instruction Stream (last 20 instructions):
SYS$BASE_IMAGE+06A84:   BIS             R28,R27,R28
SYS$BASE_IMAGE+06A88:   SUBQ            R18,R20,R27
SYS$BASE_IMAGE+06A8C:   STL             R28,(R17)
SYS$BASE_IMAGE+06A90:   BR              R31,#XFFFF96
SYS$BASE_IMAGE+06A94:   HALT
SYS$BASE_IMAGE+06A98:   HALT
SYS$BASE_IMAGE+06A9C:   HALT
OTS$MOVEM_C:            SUBQ            R17,R19,R20
OTS$MOVEM_C+00004:      LDQ_U           R31,(SP)
OTS$MOVEM_C+00008:      CMOVLT          R20,R17,R19
OTS$MOVEM_C+0000C:      AMASK           R31,#X01,R21
OTS$MOVEM_C+00010:      BLE             R19,#X00027F
OTS$MOVEM_C+00014:      SUBQ            R17,R19,R17
OTS$MOVEM_C+00018:      SUBQ            R19,#X04,R20
OTS$MOVEM_C+0001C:      LDQ_U           R31,(SP)
OTS$MOVEM_C+00020:      BNE             R21,#X000253
OTS$MOVEM_C+00024:      LDQ_U           R31,(SP)
OTS$MOVEM_C+00028:      LDQ_U           R31,(SP)
OTS$MOVEM_C+0002C:      LDQ_U           R31,(SP)
OTS$MOVEM_C+00030:      BLT             R20,#X000063
OTS$MOVEM_C+00034:      LDQ_U           R28,(R18)
OTS$MOVEM_C+00038:      ADDQ            R19,R18,R27
OTS$MOVEM_C+0003C:      SUBQ            R20,#X04,R19
OTS$MOVEM_C+00040:      BIS             R31,R16,R21
OTS$MOVEM_C+00044:      BLE             R19,#X00004A
Posted at July 7, 2005 11:55 AM
Tag Set:

Comments are closed