http://www.bobbydurrettdba.com/2012/07/17/two-ways-to-see-predicates-added-by-vpd-or-fgac/

Two ways to see predicates added by VPD or FGAC

Posted on July 17, 2012 by Bobby

We use a feature called “Virtual Private Database” or VPD on our 11g database.  This looks a lot like what used to be called “Fine Grained Access Control” or FGAC on our 10g database.  The idea behind both of these features is that a particular user in a particular situation would see a tailored view of the data rather than have all users see all of the data all the time.  VPD or FGAC accomplishes this feat by secretly adding predicates to a user’s query’s where clause predicates so they only see the rows allowed by that predicate.

The problem is that when you need to tune a poorly performing query that accesses tables protected by VPD you can’t see the real query through any of the normal methods.  Even a 10046 trace just gives you the unmodified query as the user ran it not the one with the new VPD additions.  I found two ways to see what the real where clause conditions are after the query is modified by VPD – dbms_xplan.display_cursor and 10053 trace.

Here is how to use dbms_xplan.display_cursor to show the VPD predicates:

SQL> select count(*) from test.table_list;

  COUNT(*)
----------
      1858 SQL> select * from table(
dbms_xplan.display_cursor(null,null,'ALLSTATS')); PLAN_TABLE_OUTPUT
-------------------------------------------------------
SQL_ID  2fuam6x1dyt5v, child number 0
-------------------------------------
select count(*) from test.table_list Plan hash value: 1374414456 --------------------------------------------------
| Id  | Operation          | Name       | E-Rows |
--------------------------------------------------
|   0 | SELECT STATEMENT   |            |        |
|   1 |  SORT AGGREGATE    |            |      1 |
|*  2 |   TABLE ACCESS FULL| TABLE_LIST |   2028 |
-------------------------------------------------- Predicate Information (identified by operation id):
---------------------------------------------------    2 - filter("OWNER"<>'SYS')

Note that the predicate owner<>’SYS’ isn’t in the query but was added by the VPD.  The idea here is that the table TEST.TABLE_LIST contains a list of table names but the user doing the query doesn’t have permission to see the names of the tables owned by SYS.

Here is how to use a 10053 trace to see the VPD predicates:

ALTER SESSION SET EVENTS
'10053 trace name context forever, level 1'; select /* comment to force parse */ count(*) from test.table_list; ALTER SESSION SET EVENTS '10053 trace name context OFF'; trace output: Final query after transformations:******* UNPARSED QUERY IS *******
SELECT COUNT(*) "COUNT(*)" FROM "TEST"."TABLE_LIST" "TABLE_LIST"
WHERE "TABLE_LIST"."OWNER"<>'SYS'

I had to add the comment to make sure the query got reparsed.  The 10053 trace only produces a trace when a query is parsed.  Note that the trace file has the description: “Final query after transformations”.  I’m not sure what all transformations are possible but it stands to reason that using a 10053 trace will give you a clearer picture of the real query being parsed.  It shows you the text the parser itself starts with before it starts to break it down into an execution plan that can be run.

alter session set tracefile_identifier='test_lv123';
ALTER SESSION SET EVENTS
'10053 trace name context forever, level 1';

SELECT /* comment to force parse */ * FROM oe_order_headers;

ALTER SESSION SET EVENTS '10053 trace name context OFF';

SELECT U_DUMP.VALUE || '/' || DB_NAME.VALUE || '_ora_' || V$PROCESS.SPID ||
NVL2(V$PROCESS.TRACEID, '_' || V$PROCESS.TRACEID, NULL) || '.trc' "Trace File"
FROM V$PARAMETER U_DUMP
CROSS JOIN V$PARAMETER DB_NAME
CROSS JOIN V$PROCESS
JOIN V$SESSION
ON V$PROCESS.ADDR = V$SESSION.PADDR
WHERE U_DUMP.NAME = 'user_dump_dest'
AND DB_NAME.NAME = 'db_name'
AND V$SESSION.AUDSID = SYS_CONTEXT('userenv', 'sessionid');

Two ways to see predicates added by VPD or FGAC的更多相关文章

  1. C++ Knowledge series 3

    Programming language evolves always along with Compiler's evolvement The Semantics of Data The size ...

  2. 【leetcode】Decode Ways(medium)

    A message containing letters from A-Z is being encoded to numbers using the following mapping: 'A' - ...

  3. ASP.NET MVC3 Dynamically added form fields model binding

    Adding  new Item to a list of items, inline is a very nice feature you can provide to your user. Thi ...

  4. QMP ( qemu monitor protocol ) and Different ways of accessing it

    The QEMU Monitor Protocol (QMP) is a JSON-based protocol which allows applications to communicate wi ...

  5. dapper extensions (predicates)

    https://github.com/tmsmith/Dapper-Extensions/wiki/Predicates The predicate system in Dapper Extensio ...

  6. [AngularJS] 5 simple ways to speed up your AngularJS application

    Nowdays, Single page apps are becoming increasingly popular among the fornt-end developers. It is th ...

  7. ASP.NET MVC:4 Ways To Prevent Duplicate Form Submission(转载)

    原文地址:http://technoesis.net/prevent-double-form-submission/. Double form submission in a multi-user w ...

  8. Four Ways to Create a Thread

    Blaise Pascal Magazine Rerun #5: Four Ways to Create a Thread   This article was originally written ...

  9. Recommend ways to overwrite hashCode() in java

    Perface In the former chapter, I talk about topics about hashCode, And I will continue to finish the ...

随机推荐

  1. linux kernel 的配置及编译

    1. 执行make menuconfig 配置内核 2. 执行make zImage 编译内核 3. 执行make modules 编译模块 4. 内核源代码的配置及编译系统 Makefile Kco ...

  2. sqldatareader无法得到output参数的解决

    只需要在所有的sqldatareader结束后,加上一句就可以得到输出参数了. sdr.Close(); Object ObjCount = cmd.Parameters["@Count_P ...

  3. oracle,PL/SQL新建表

    创建用户 -- Create the user create user U_HQ_JAVA default tablespace USERS temporary tablespace TEMP pro ...

  4. Python3 chr() 函数

    Python3 chr() 函数 Python3 内置函数 描述 chr() 用一个整数作参数,返回一个对应的字符. 语法 以下是 chr() 方法的语法: chr(i) 参数 i -- 可以是 10 ...

  5. goim源码分析与二次开发-comet分析一

    因为要完成一个聊天的项目,所以借鉴了goim,第一篇分析打算半原版,先摘抄http://www.jianshu.com/p/8bd96a9a473d他的一些理解,写这些还是为了让自己更好的理解这个项目 ...

  6. leetcdoe 175. Combine Two Tables

    给定两个表,一个是人,一个是地址,要求查询所有人,可以没有地址. select a.FirstName, a.LastName, b.City, b.State from Person as a le ...

  7. Django的models操作

    一.先看单表操作 增 方式1: models.book.objects.create( Book_name = "aaa", Book_info = "bbb" ...

  8. 240. Search a 2D Matrix II&&74. Search a 2D Matrix 都用不严格递增的方法

    一句话思路:从左下角开始找.复杂度是o(m+n). 一刷报错: 应该是一个while循环中有几个条件判断语句,而不是每个条件判断语句里去加while,很麻烦 数组越界了.从0开始,y的最大条件应该是& ...

  9. Spring的3.0提供了一种:SpEL注入方式(了解)

    1. SpEL:Spring Expression Language是Spring的表达式语言,有一些自己的语法 2. 语法 * #{SpEL} 3. 例如如下的代码 <!-- SpEL的方式 ...

  10. swagger ui js 错误:Failed to execute 'serializeToString' on 'XMLSerializer': parameter 1 is not of type 'Node'.

    经过排查,引发此错误的原因是,表中有一个字段名称为“NodeName”,应该是在前台xml解析时引发冲突所致.我的解决办法是: 修改列名,修改映射. 如下: [Column("NodeNam ...