在上一篇的基础上，本篇实现修改undo segment header中的相关事务槽信息,从而屏蔽数据库在重启或者进程异常的时候,smon的回滚操作,从而比较完美的实现了手工提交数据库事务
实验过程如下
Session 1中
JP@ORCL>select last_name from bbed_test;
LAST_NAME
-------------------------
OConnell
Grant
Whalen
Hartstein
Fay
Mavris
Baer
Higgins
Gietz
King
10 rows selected.
JP@ORCL>update bbed_test set last_name='BADLY9';
10 rows updated.
事务未提交
Session 2中：
SYS@ORCL>select rowid, dbms_rowid.rowid_relative_fno(rowid)rel_fno,
dbms_rowid.rowid_block_number(rowid)blockno,
dbms_rowid.rowid_row_number(rowid) rowno
from 2 3 4
5 JP.BBED_TEST;
ROWID REL_FNO BLOCKNO ROWNO
------------------ ---------- ---------- ----------
AAAM8CAAEAAAAG8AAA 4 444 0
AAAM8CAAEAAAAG8AAB 4 444 1
AAAM8CAAEAAAAG8AAC 4 444 2
AAAM8CAAEAAAAG8AAD 4 444 3
AAAM8CAAEAAAAG8AAE 4 444 4
AAAM8CAAEAAAAG8AAF 4 444 5
AAAM8CAAEAAAAG8AAG 4 444 6
AAAM8CAAEAAAAG8AAH 4 444 7
AAAM8CAAEAAAAG8AAI 4 444 8
AAAM8CAAEAAAAG8AAJ 4 444 9
10 rows selected.
SYS@ORCL>SELECT XIDUSN,XIDSLOT,XIDSQN,UBAFIL,UBABLK,UBASQN,UBAREC FROM v$transaction;
XIDUSN XIDSLOT XIDSQN UBAFIL UBABLK UBASQN UBAREC
---------- ---------- ---------- ---------- ---------- ---------- ----------
7 38 369 2 786 302 16
SYS@ORCL>select segment_name,header_file,header_block
2 from dba_segments where segment_name=(
3 select name from v$rollname where usn=(select xidusn from v$transaction));
SEGMENT_NAME HEADER_FILE HEADER_BLOCK
------------------ ----------- ------------
_SYSSMU7$ 2 105
SYS@ORCL>alter system dump undo header "_SYSSMU7$";
System altered.
SYS@ORCL>oradebug setmypid
Statement processed.
SYS@ORCL>oradebug tracefile_name
/u01/app/oracle/admin/ORCL/udump/orcl_ora_8512.trc
查看/u01/app/oracle/admin/ORCL/udump/orcl_ora_8512.trc文件，在其中可以看到
0x25 9 0x00 0x0171 0x001d 0x0000.000ee92e 0x00000000 0x0000.000.00000000 0x00000000 0x00000000 1402237152
0x26 10 0x80 0x0171 0x0001 0x0000.000f3981 0x00800312 0x0000.000.00000000 0x00000001 0x00000000 0
0x27 9 0x00 0x0171 0x0022 0x0000.000f38cc 0x00800311 0x0000.000.00000000 0x00000001 0x00000000 1402238227
通过结合dump undo header 中的TRN TBL的state为10的为active事务,然后结合scn/dba等信息来确定是哪条记录是需要我们修改.然后通过find命令快速定位到0x0b这条记录,然后进行修改
通过结合bbed的dump命令得出16进制数据分析得出如下结论
--index 0x26
7101 wrap# 0000 12038000 dba 81390f00 00000000 scn 0a s tate 80 cflags 0100 uel 00000000 00000000 parent-xid 00000000 stmt_num 01000000 nub 00000000 cmt
BBED> f /x 0a80
File: /u01/app/oracle/oradata/ORCL/undotbs01.dbf (2)
Block: 105 Offsets: 7736 to 8191 Dba:0x00800069
------------------------------------------------------------------------
0a800100 00000000 00000000 00000000 01000000 00000000 71010000 11038000
<32 bytes per line>
BBED> m /x 0900
Warning: contents of previous BIFILE will be lost. Proceed (Y/N) Y
File: /u01/app/oracle/oradata/ORCL/undotbs01.dbf (2)
Block: 105 Offsets: 7736 to 8191 Dba:0x00800069
------------------------------------------------------------------------
09000100 00000000 00000000 00000000 01000000 00000000 71010000 11038000
<32 bytes per line>
BBED> sum apply
Check value for File 2, Block 105:
current = 0x8e65, required = 0x8e65
接下来修改itl
struct ktbbhitl[1], 24 bytes @68
struct ktbitxid, 8 bytes @68
ub2 kxidusn @68 0x0007
ub2 kxidslt @70 0x0026
ub4 kxidsqn @72 0x00000171
struct ktbituba, 8 bytes @76
ub4 kubadba @76 0x00800312
ub2 kubaseq @80 0x012e
ub1 kubarec @82 0x10
ub2 ktbitflg @84 0x000a (NONE)
union _kt