postgresql/src/test/regress/sql/rowsecurity.sql

1864 lines
62 KiB
MySQL
Raw Normal View History

Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
--
-- Test of Row-level security feature
--
-- Clean up in case a prior regression run failed
-- Suppress NOTICE messages when users/groups don't exist
SET client_min_messages TO 'warning';
DROP USER IF EXISTS regress_rls_alice;
DROP USER IF EXISTS regress_rls_bob;
DROP USER IF EXISTS regress_rls_carol;
DROP USER IF EXISTS regress_rls_dave;
DROP USER IF EXISTS regress_rls_exempt_user;
DROP ROLE IF EXISTS regress_rls_group1;
DROP ROLE IF EXISTS regress_rls_group2;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
DROP SCHEMA IF EXISTS regress_rls_schema CASCADE;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
RESET client_min_messages;
-- initial setup
CREATE USER regress_rls_alice NOLOGIN;
CREATE USER regress_rls_bob NOLOGIN;
CREATE USER regress_rls_carol NOLOGIN;
CREATE USER regress_rls_dave NOLOGIN;
CREATE USER regress_rls_exempt_user BYPASSRLS NOLOGIN;
CREATE ROLE regress_rls_group1 NOLOGIN;
CREATE ROLE regress_rls_group2 NOLOGIN;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
GRANT regress_rls_group1 TO regress_rls_bob;
GRANT regress_rls_group2 TO regress_rls_carol;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
CREATE SCHEMA regress_rls_schema;
GRANT ALL ON SCHEMA regress_rls_schema to public;
SET search_path = regress_rls_schema;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
-- setup of malicious function
CREATE OR REPLACE FUNCTION f_leak(text) RETURNS bool
COST 0.0000001 LANGUAGE plpgsql
AS 'BEGIN RAISE NOTICE ''f_leak => %'', $1; RETURN true; END';
GRANT EXECUTE ON FUNCTION f_leak(text) TO public;
-- BASIC Row-Level Security Scenario
SET SESSION AUTHORIZATION regress_rls_alice;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
CREATE TABLE uaccount (
pguser name primary key,
seclv int
);
GRANT SELECT ON uaccount TO public;
INSERT INTO uaccount VALUES
('regress_rls_alice', 99),
('regress_rls_bob', 1),
('regress_rls_carol', 2),
('regress_rls_dave', 3);
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
CREATE TABLE category (
cid int primary key,
cname text
);
GRANT ALL ON category TO public;
INSERT INTO category VALUES
(11, 'novel'),
(22, 'science fiction'),
(33, 'technology'),
(44, 'manga');
CREATE TABLE document (
did int primary key,
cid int references category(cid),
dlevel int not null,
dauthor name,
dtitle text
);
GRANT ALL ON document TO public;
INSERT INTO document VALUES
( 1, 11, 1, 'regress_rls_bob', 'my first novel'),
( 2, 11, 2, 'regress_rls_bob', 'my second novel'),
( 3, 22, 2, 'regress_rls_bob', 'my science fiction'),
( 4, 44, 1, 'regress_rls_bob', 'my first manga'),
( 5, 44, 2, 'regress_rls_bob', 'my second manga'),
( 6, 22, 1, 'regress_rls_carol', 'great science fiction'),
( 7, 33, 2, 'regress_rls_carol', 'great technology book'),
( 8, 44, 1, 'regress_rls_carol', 'great manga'),
( 9, 22, 1, 'regress_rls_dave', 'awesome science fiction'),
(10, 33, 2, 'regress_rls_dave', 'awesome technology book');
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
ALTER TABLE document ENABLE ROW LEVEL SECURITY;
-- user's security level must be higher than or equal to document's
CREATE POLICY p1 ON document AS PERMISSIVE
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
USING (dlevel <= (SELECT seclv FROM uaccount WHERE pguser = current_user));
-- try to create a policy of bogus type
CREATE POLICY p1 ON document AS UGLY
USING (dlevel <= (SELECT seclv FROM uaccount WHERE pguser = current_user));
-- but Dave isn't allowed to anything at cid 50 or above
-- this is to make sure that we sort the policies by name first
-- when applying WITH CHECK, a later INSERT by Dave should fail due
-- to p1r first
CREATE POLICY p2r ON document AS RESTRICTIVE TO regress_rls_dave
USING (cid <> 44 AND cid < 50);
-- and Dave isn't allowed to see manga documents
CREATE POLICY p1r ON document AS RESTRICTIVE TO regress_rls_dave
USING (cid <> 44);
\dp
\d document
SELECT * FROM pg_policies WHERE schemaname = 'regress_rls_schema' AND tablename = 'document' ORDER BY policyname;
-- viewpoint from regress_rls_bob
SET SESSION AUTHORIZATION regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SET row_security TO ON;
SELECT * FROM document WHERE f_leak(dtitle) ORDER BY did;
SELECT * FROM document NATURAL JOIN category WHERE f_leak(dtitle) ORDER BY did;
Redesign tablesample method API, and do extensive code review. The original implementation of TABLESAMPLE modeled the tablesample method API on index access methods, which wasn't a good choice because, without specialized DDL commands, there's no way to build an extension that can implement a TSM. (Raw inserts into system catalogs are not an acceptable thing to do, because we can't undo them during DROP EXTENSION, nor will pg_upgrade behave sanely.) Instead adopt an API more like procedural language handlers or foreign data wrappers, wherein the only SQL-level support object needed is a single handler function identified by having a special return type. This lets us get rid of the supporting catalog altogether, so that no custom DDL support is needed for the feature. Adjust the API so that it can support non-constant tablesample arguments (the original coding assumed we could evaluate the argument expressions at ExecInitSampleScan time, which is undesirable even if it weren't outright unsafe), and discourage sampling methods from looking at invisible tuples. Make sure that the BERNOULLI and SYSTEM methods are genuinely repeatable within and across queries, as required by the SQL standard, and deal more honestly with methods that can't support that requirement. Make a full code-review pass over the tablesample additions, and fix assorted bugs, omissions, infelicities, and cosmetic issues (such as failure to put the added code stanzas in a consistent ordering). Improve EXPLAIN's output of tablesample plans, too. Back-patch to 9.5 so that we don't have to support the original API in production.
2015-07-25 20:39:00 +02:00
-- try a sampled version
SELECT * FROM document TABLESAMPLE BERNOULLI(50) REPEATABLE(0)
WHERE f_leak(dtitle) ORDER BY did;
-- viewpoint from regress_rls_carol
SET SESSION AUTHORIZATION regress_rls_carol;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SELECT * FROM document WHERE f_leak(dtitle) ORDER BY did;
SELECT * FROM document NATURAL JOIN category WHERE f_leak(dtitle) ORDER BY did;
Redesign tablesample method API, and do extensive code review. The original implementation of TABLESAMPLE modeled the tablesample method API on index access methods, which wasn't a good choice because, without specialized DDL commands, there's no way to build an extension that can implement a TSM. (Raw inserts into system catalogs are not an acceptable thing to do, because we can't undo them during DROP EXTENSION, nor will pg_upgrade behave sanely.) Instead adopt an API more like procedural language handlers or foreign data wrappers, wherein the only SQL-level support object needed is a single handler function identified by having a special return type. This lets us get rid of the supporting catalog altogether, so that no custom DDL support is needed for the feature. Adjust the API so that it can support non-constant tablesample arguments (the original coding assumed we could evaluate the argument expressions at ExecInitSampleScan time, which is undesirable even if it weren't outright unsafe), and discourage sampling methods from looking at invisible tuples. Make sure that the BERNOULLI and SYSTEM methods are genuinely repeatable within and across queries, as required by the SQL standard, and deal more honestly with methods that can't support that requirement. Make a full code-review pass over the tablesample additions, and fix assorted bugs, omissions, infelicities, and cosmetic issues (such as failure to put the added code stanzas in a consistent ordering). Improve EXPLAIN's output of tablesample plans, too. Back-patch to 9.5 so that we don't have to support the original API in production.
2015-07-25 20:39:00 +02:00
-- try a sampled version
SELECT * FROM document TABLESAMPLE BERNOULLI(50) REPEATABLE(0)
WHERE f_leak(dtitle) ORDER BY did;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
EXPLAIN (COSTS OFF) SELECT * FROM document WHERE f_leak(dtitle);
EXPLAIN (COSTS OFF) SELECT * FROM document NATURAL JOIN category WHERE f_leak(dtitle);
-- viewpoint from regress_rls_dave
SET SESSION AUTHORIZATION regress_rls_dave;
SELECT * FROM document WHERE f_leak(dtitle) ORDER BY did;
SELECT * FROM document NATURAL JOIN category WHERE f_leak(dtitle) ORDER BY did;
EXPLAIN (COSTS OFF) SELECT * FROM document WHERE f_leak(dtitle);
EXPLAIN (COSTS OFF) SELECT * FROM document NATURAL JOIN category WHERE f_leak(dtitle);
-- 44 would technically fail for both p2r and p1r, but we should get an error
-- back from p1r for this because it sorts first
INSERT INTO document VALUES (100, 44, 1, 'regress_rls_dave', 'testing sorting of policies'); -- fail
-- Just to see a p2r error
INSERT INTO document VALUES (100, 55, 1, 'regress_rls_dave', 'testing sorting of policies'); -- fail
-- only owner can change policies
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
ALTER POLICY p1 ON document USING (true); --fail
DROP POLICY p1 ON document; --fail
SET SESSION AUTHORIZATION regress_rls_alice;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
ALTER POLICY p1 ON document USING (dauthor = current_user);
-- viewpoint from regress_rls_bob again
SET SESSION AUTHORIZATION regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SELECT * FROM document WHERE f_leak(dtitle) ORDER BY did;
SELECT * FROM document NATURAL JOIN category WHERE f_leak(dtitle) ORDER by did;
-- viewpoint from rls_regres_carol again
SET SESSION AUTHORIZATION regress_rls_carol;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SELECT * FROM document WHERE f_leak(dtitle) ORDER BY did;
SELECT * FROM document NATURAL JOIN category WHERE f_leak(dtitle) ORDER by did;
EXPLAIN (COSTS OFF) SELECT * FROM document WHERE f_leak(dtitle);
EXPLAIN (COSTS OFF) SELECT * FROM document NATURAL JOIN category WHERE f_leak(dtitle);
-- interaction of FK/PK constraints
SET SESSION AUTHORIZATION regress_rls_alice;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
CREATE POLICY p2 ON category
USING (CASE WHEN current_user = 'regress_rls_bob' THEN cid IN (11, 33)
WHEN current_user = 'regress_rls_carol' THEN cid IN (22, 44)
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
ELSE false END);
ALTER TABLE category ENABLE ROW LEVEL SECURITY;
-- cannot delete PK referenced by invisible FK
SET SESSION AUTHORIZATION regress_rls_bob;
SELECT * FROM document d FULL OUTER JOIN category c on d.cid = c.cid ORDER BY d.did, c.cid;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
DELETE FROM category WHERE cid = 33; -- fails with FK violation
-- can insert FK referencing invisible PK
SET SESSION AUTHORIZATION regress_rls_carol;
SELECT * FROM document d FULL OUTER JOIN category c on d.cid = c.cid ORDER BY d.did, c.cid;
INSERT INTO document VALUES (11, 33, 1, current_user, 'hoge');
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
-- UNIQUE or PRIMARY KEY constraint violation DOES reveal presence of row
SET SESSION AUTHORIZATION regress_rls_bob;
INSERT INTO document VALUES (8, 44, 1, 'regress_rls_bob', 'my third manga'); -- Must fail with unique violation, revealing presence of did we can't see
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SELECT * FROM document WHERE did = 8; -- and confirm we can't see it
-- RLS policies are checked before constraints
INSERT INTO document VALUES (8, 44, 1, 'regress_rls_carol', 'my third manga'); -- Should fail with RLS check violation, not duplicate key violation
UPDATE document SET did = 8, dauthor = 'regress_rls_carol' WHERE did = 5; -- Should fail with RLS check violation, not duplicate key violation
-- database superuser does bypass RLS policy when enabled
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
RESET SESSION AUTHORIZATION;
SET row_security TO ON;
SELECT * FROM document;
SELECT * FROM category;
-- database superuser does bypass RLS policy when disabled
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
RESET SESSION AUTHORIZATION;
SET row_security TO OFF;
SELECT * FROM document;
SELECT * FROM category;
-- database non-superuser with bypass privilege can bypass RLS policy when disabled
SET SESSION AUTHORIZATION regress_rls_exempt_user;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SET row_security TO OFF;
SELECT * FROM document;
SELECT * FROM category;
-- RLS policy does not apply to table owner when RLS enabled.
SET SESSION AUTHORIZATION regress_rls_alice;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SET row_security TO ON;
SELECT * FROM document;
SELECT * FROM category;
-- RLS policy does not apply to table owner when RLS disabled.
SET SESSION AUTHORIZATION regress_rls_alice;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SET row_security TO OFF;
SELECT * FROM document;
SELECT * FROM category;
--
-- Table inheritance and RLS policy
--
SET SESSION AUTHORIZATION regress_rls_alice;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SET row_security TO ON;
Remove WITH OIDS support, change oid catalog column visibility. Previously tables declared WITH OIDS, including a significant fraction of the catalog tables, stored the oid column not as a normal column, but as part of the tuple header. This special column was not shown by default, which was somewhat odd, as it's often (consider e.g. pg_class.oid) one of the more important parts of a row. Neither pg_dump nor COPY included the contents of the oid column by default. The fact that the oid column was not an ordinary column necessitated a significant amount of special case code to support oid columns. That already was painful for the existing, but upcoming work aiming to make table storage pluggable, would have required expanding and duplicating that "specialness" significantly. WITH OIDS has been deprecated since 2005 (commit ff02d0a05280e0). Remove it. Removing includes: - CREATE TABLE and ALTER TABLE syntax for declaring the table to be WITH OIDS has been removed (WITH (oids[ = true]) will error out) - pg_dump does not support dumping tables declared WITH OIDS and will issue a warning when dumping one (and ignore the oid column). - restoring an pg_dump archive with pg_restore will warn when restoring a table with oid contents (and ignore the oid column) - COPY will refuse to load binary dump that includes oids. - pg_upgrade will error out when encountering tables declared WITH OIDS, they have to be altered to remove the oid column first. - Functionality to access the oid of the last inserted row (like plpgsql's RESULT_OID, spi's SPI_lastoid, ...) has been removed. The syntax for declaring a table WITHOUT OIDS (or WITH (oids = false) for CREATE TABLE) is still supported. While that requires a bit of support code, it seems unnecessary to break applications / dumps that do not use oids, and are explicit about not using them. The biggest user of WITH OID columns was postgres' catalog. This commit changes all 'magic' oid columns to be columns that are normally declared and stored. To reduce unnecessary query breakage all the newly added columns are still named 'oid', even if a table's column naming scheme would indicate 'reloid' or such. This obviously requires adapting a lot code, mostly replacing oid access via HeapTupleGetOid() with access to the underlying Form_pg_*->oid column. The bootstrap process now assigns oids for all oid columns in genbki.pl that do not have an explicit value (starting at the largest oid previously used), only oids assigned later by oids will be above FirstBootstrapObjectId. As the oid column now is a normal column the special bootstrap syntax for oids has been removed. Oids are not automatically assigned during insertion anymore, all backend code explicitly assigns oids with GetNewOidWithIndex(). For the rare case that insertions into the catalog via SQL are called for the new pg_nextoid() function can be used (which only works on catalog tables). The fact that oid columns on system tables are now normal columns means that they will be included in the set of columns expanded by * (i.e. SELECT * FROM pg_class will now include the table's oid, previously it did not). It'd not technically be hard to hide oid column by default, but that'd mean confusing behavior would either have to be carried forward forever, or it'd cause breakage down the line. While it's not unlikely that further adjustments are needed, the scope/invasiveness of the patch makes it worthwhile to get merge this now. It's painful to maintain externally, too complicated to commit after the code code freeze, and a dependency of a number of other patches. Catversion bump, for obvious reasons. Author: Andres Freund, with contributions by John Naylor Discussion: https://postgr.es/m/20180930034810.ywp2c7awz7opzcfr@alap3.anarazel.de
2018-11-21 00:36:57 +01:00
CREATE TABLE t1 (id int not null primary key, a int, junk1 text, b text);
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
ALTER TABLE t1 DROP COLUMN junk1; -- just a disturbing factor
GRANT ALL ON t1 TO public;
Remove WITH OIDS support, change oid catalog column visibility. Previously tables declared WITH OIDS, including a significant fraction of the catalog tables, stored the oid column not as a normal column, but as part of the tuple header. This special column was not shown by default, which was somewhat odd, as it's often (consider e.g. pg_class.oid) one of the more important parts of a row. Neither pg_dump nor COPY included the contents of the oid column by default. The fact that the oid column was not an ordinary column necessitated a significant amount of special case code to support oid columns. That already was painful for the existing, but upcoming work aiming to make table storage pluggable, would have required expanding and duplicating that "specialness" significantly. WITH OIDS has been deprecated since 2005 (commit ff02d0a05280e0). Remove it. Removing includes: - CREATE TABLE and ALTER TABLE syntax for declaring the table to be WITH OIDS has been removed (WITH (oids[ = true]) will error out) - pg_dump does not support dumping tables declared WITH OIDS and will issue a warning when dumping one (and ignore the oid column). - restoring an pg_dump archive with pg_restore will warn when restoring a table with oid contents (and ignore the oid column) - COPY will refuse to load binary dump that includes oids. - pg_upgrade will error out when encountering tables declared WITH OIDS, they have to be altered to remove the oid column first. - Functionality to access the oid of the last inserted row (like plpgsql's RESULT_OID, spi's SPI_lastoid, ...) has been removed. The syntax for declaring a table WITHOUT OIDS (or WITH (oids = false) for CREATE TABLE) is still supported. While that requires a bit of support code, it seems unnecessary to break applications / dumps that do not use oids, and are explicit about not using them. The biggest user of WITH OID columns was postgres' catalog. This commit changes all 'magic' oid columns to be columns that are normally declared and stored. To reduce unnecessary query breakage all the newly added columns are still named 'oid', even if a table's column naming scheme would indicate 'reloid' or such. This obviously requires adapting a lot code, mostly replacing oid access via HeapTupleGetOid() with access to the underlying Form_pg_*->oid column. The bootstrap process now assigns oids for all oid columns in genbki.pl that do not have an explicit value (starting at the largest oid previously used), only oids assigned later by oids will be above FirstBootstrapObjectId. As the oid column now is a normal column the special bootstrap syntax for oids has been removed. Oids are not automatically assigned during insertion anymore, all backend code explicitly assigns oids with GetNewOidWithIndex(). For the rare case that insertions into the catalog via SQL are called for the new pg_nextoid() function can be used (which only works on catalog tables). The fact that oid columns on system tables are now normal columns means that they will be included in the set of columns expanded by * (i.e. SELECT * FROM pg_class will now include the table's oid, previously it did not). It'd not technically be hard to hide oid column by default, but that'd mean confusing behavior would either have to be carried forward forever, or it'd cause breakage down the line. While it's not unlikely that further adjustments are needed, the scope/invasiveness of the patch makes it worthwhile to get merge this now. It's painful to maintain externally, too complicated to commit after the code code freeze, and a dependency of a number of other patches. Catversion bump, for obvious reasons. Author: Andres Freund, with contributions by John Naylor Discussion: https://postgr.es/m/20180930034810.ywp2c7awz7opzcfr@alap3.anarazel.de
2018-11-21 00:36:57 +01:00
COPY t1 FROM stdin WITH ;
101 1 aba
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
102 2 bbb
103 3 ccc
104 4 dad
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
\.
CREATE TABLE t2 (c float) INHERITS (t1);
RLS fixes, new hooks, and new test module In prepend_row_security_policies(), defaultDeny was always true, so if there were any hook policies, the RLS policies on the table would just get discarded. Fixed to start off with defaultDeny as false and then properly set later if we detect that only the default deny policy exists for the internal policies. The infinite recursion detection in fireRIRrules() didn't properly manage the activeRIRs list in the case of WCOs, so it would incorrectly report infinite recusion if the same relation with RLS appeared more than once in the rtable, for example "UPDATE t ... FROM t ...". Further, the RLS expansion code in fireRIRrules() was handling RLS in the main loop through the rtable, which lead to RTEs being visited twice if they contained sublink subqueries, which prepend_row_security_policies() attempted to handle by exiting early if the RTE already had securityQuals. That doesn't work, however, since if the query involved a security barrier view on top of a table with RLS, the RTE would already have securityQuals (from the view) by the time fireRIRrules() was invoked, and so the table's RLS policies would be ignored. This is fixed in fireRIRrules() by handling RLS in a separate loop at the end, after dealing with any other sublink subqueries, thus ensuring that each RTE is only visited once for RLS expansion. The inheritance planner code didn't correctly handle non-target relations with RLS, which would get turned into subqueries during planning. Thus an update of the form "UPDATE t1 ... FROM t2 ..." where t1 has inheritance and t2 has RLS quals would fail. Fix by making sure to copy in and update the securityQuals when they exist for non-target relations. process_policies() was adding WCOs to non-target relations, which is unnecessary, and could lead to a lot of wasted time in the rewriter and the planner. Fix by only adding WCO policies when working on the result relation. Also in process_policies, we should be copying the USING policies to the WITH CHECK policies on a per-policy basis, fix by moving the copying up into the per-policy loop. Lastly, as noted by Dean, we were simply adding policies returned by the hook provided to the list of quals being AND'd, meaning that they would actually restrict records returned and there was no option to have internal policies and hook-based policies work together permissively (as all internal policies currently work). Instead, explicitly add support for both permissive and restrictive policies by having a hook for each and combining the results appropriately. To ensure this is all done correctly, add a new test module (test_rls_hooks) to test the various combinations of internal, permissive, and restrictive hook policies. Largely from Dean Rasheed (thanks!): CAEZATCVmFUfUOwwhnBTcgi6AquyjQ0-1fyKd0T3xBWJvn+xsFA@mail.gmail.com Author: Dean Rasheed, though I added the new hooks and test module.
2015-04-22 18:01:06 +02:00
GRANT ALL ON t2 TO public;
Remove WITH OIDS support, change oid catalog column visibility. Previously tables declared WITH OIDS, including a significant fraction of the catalog tables, stored the oid column not as a normal column, but as part of the tuple header. This special column was not shown by default, which was somewhat odd, as it's often (consider e.g. pg_class.oid) one of the more important parts of a row. Neither pg_dump nor COPY included the contents of the oid column by default. The fact that the oid column was not an ordinary column necessitated a significant amount of special case code to support oid columns. That already was painful for the existing, but upcoming work aiming to make table storage pluggable, would have required expanding and duplicating that "specialness" significantly. WITH OIDS has been deprecated since 2005 (commit ff02d0a05280e0). Remove it. Removing includes: - CREATE TABLE and ALTER TABLE syntax for declaring the table to be WITH OIDS has been removed (WITH (oids[ = true]) will error out) - pg_dump does not support dumping tables declared WITH OIDS and will issue a warning when dumping one (and ignore the oid column). - restoring an pg_dump archive with pg_restore will warn when restoring a table with oid contents (and ignore the oid column) - COPY will refuse to load binary dump that includes oids. - pg_upgrade will error out when encountering tables declared WITH OIDS, they have to be altered to remove the oid column first. - Functionality to access the oid of the last inserted row (like plpgsql's RESULT_OID, spi's SPI_lastoid, ...) has been removed. The syntax for declaring a table WITHOUT OIDS (or WITH (oids = false) for CREATE TABLE) is still supported. While that requires a bit of support code, it seems unnecessary to break applications / dumps that do not use oids, and are explicit about not using them. The biggest user of WITH OID columns was postgres' catalog. This commit changes all 'magic' oid columns to be columns that are normally declared and stored. To reduce unnecessary query breakage all the newly added columns are still named 'oid', even if a table's column naming scheme would indicate 'reloid' or such. This obviously requires adapting a lot code, mostly replacing oid access via HeapTupleGetOid() with access to the underlying Form_pg_*->oid column. The bootstrap process now assigns oids for all oid columns in genbki.pl that do not have an explicit value (starting at the largest oid previously used), only oids assigned later by oids will be above FirstBootstrapObjectId. As the oid column now is a normal column the special bootstrap syntax for oids has been removed. Oids are not automatically assigned during insertion anymore, all backend code explicitly assigns oids with GetNewOidWithIndex(). For the rare case that insertions into the catalog via SQL are called for the new pg_nextoid() function can be used (which only works on catalog tables). The fact that oid columns on system tables are now normal columns means that they will be included in the set of columns expanded by * (i.e. SELECT * FROM pg_class will now include the table's oid, previously it did not). It'd not technically be hard to hide oid column by default, but that'd mean confusing behavior would either have to be carried forward forever, or it'd cause breakage down the line. While it's not unlikely that further adjustments are needed, the scope/invasiveness of the patch makes it worthwhile to get merge this now. It's painful to maintain externally, too complicated to commit after the code code freeze, and a dependency of a number of other patches. Catversion bump, for obvious reasons. Author: Andres Freund, with contributions by John Naylor Discussion: https://postgr.es/m/20180930034810.ywp2c7awz7opzcfr@alap3.anarazel.de
2018-11-21 00:36:57 +01:00
COPY t2 FROM stdin;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
201 1 abc 1.1
202 2 bcd 2.2
203 3 cde 3.3
204 4 def 4.4
\.
Remove WITH OIDS support, change oid catalog column visibility. Previously tables declared WITH OIDS, including a significant fraction of the catalog tables, stored the oid column not as a normal column, but as part of the tuple header. This special column was not shown by default, which was somewhat odd, as it's often (consider e.g. pg_class.oid) one of the more important parts of a row. Neither pg_dump nor COPY included the contents of the oid column by default. The fact that the oid column was not an ordinary column necessitated a significant amount of special case code to support oid columns. That already was painful for the existing, but upcoming work aiming to make table storage pluggable, would have required expanding and duplicating that "specialness" significantly. WITH OIDS has been deprecated since 2005 (commit ff02d0a05280e0). Remove it. Removing includes: - CREATE TABLE and ALTER TABLE syntax for declaring the table to be WITH OIDS has been removed (WITH (oids[ = true]) will error out) - pg_dump does not support dumping tables declared WITH OIDS and will issue a warning when dumping one (and ignore the oid column). - restoring an pg_dump archive with pg_restore will warn when restoring a table with oid contents (and ignore the oid column) - COPY will refuse to load binary dump that includes oids. - pg_upgrade will error out when encountering tables declared WITH OIDS, they have to be altered to remove the oid column first. - Functionality to access the oid of the last inserted row (like plpgsql's RESULT_OID, spi's SPI_lastoid, ...) has been removed. The syntax for declaring a table WITHOUT OIDS (or WITH (oids = false) for CREATE TABLE) is still supported. While that requires a bit of support code, it seems unnecessary to break applications / dumps that do not use oids, and are explicit about not using them. The biggest user of WITH OID columns was postgres' catalog. This commit changes all 'magic' oid columns to be columns that are normally declared and stored. To reduce unnecessary query breakage all the newly added columns are still named 'oid', even if a table's column naming scheme would indicate 'reloid' or such. This obviously requires adapting a lot code, mostly replacing oid access via HeapTupleGetOid() with access to the underlying Form_pg_*->oid column. The bootstrap process now assigns oids for all oid columns in genbki.pl that do not have an explicit value (starting at the largest oid previously used), only oids assigned later by oids will be above FirstBootstrapObjectId. As the oid column now is a normal column the special bootstrap syntax for oids has been removed. Oids are not automatically assigned during insertion anymore, all backend code explicitly assigns oids with GetNewOidWithIndex(). For the rare case that insertions into the catalog via SQL are called for the new pg_nextoid() function can be used (which only works on catalog tables). The fact that oid columns on system tables are now normal columns means that they will be included in the set of columns expanded by * (i.e. SELECT * FROM pg_class will now include the table's oid, previously it did not). It'd not technically be hard to hide oid column by default, but that'd mean confusing behavior would either have to be carried forward forever, or it'd cause breakage down the line. While it's not unlikely that further adjustments are needed, the scope/invasiveness of the patch makes it worthwhile to get merge this now. It's painful to maintain externally, too complicated to commit after the code code freeze, and a dependency of a number of other patches. Catversion bump, for obvious reasons. Author: Andres Freund, with contributions by John Naylor Discussion: https://postgr.es/m/20180930034810.ywp2c7awz7opzcfr@alap3.anarazel.de
2018-11-21 00:36:57 +01:00
CREATE TABLE t3 (id int not null primary key, c text, b text, a int);
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
ALTER TABLE t3 INHERIT t1;
RLS fixes, new hooks, and new test module In prepend_row_security_policies(), defaultDeny was always true, so if there were any hook policies, the RLS policies on the table would just get discarded. Fixed to start off with defaultDeny as false and then properly set later if we detect that only the default deny policy exists for the internal policies. The infinite recursion detection in fireRIRrules() didn't properly manage the activeRIRs list in the case of WCOs, so it would incorrectly report infinite recusion if the same relation with RLS appeared more than once in the rtable, for example "UPDATE t ... FROM t ...". Further, the RLS expansion code in fireRIRrules() was handling RLS in the main loop through the rtable, which lead to RTEs being visited twice if they contained sublink subqueries, which prepend_row_security_policies() attempted to handle by exiting early if the RTE already had securityQuals. That doesn't work, however, since if the query involved a security barrier view on top of a table with RLS, the RTE would already have securityQuals (from the view) by the time fireRIRrules() was invoked, and so the table's RLS policies would be ignored. This is fixed in fireRIRrules() by handling RLS in a separate loop at the end, after dealing with any other sublink subqueries, thus ensuring that each RTE is only visited once for RLS expansion. The inheritance planner code didn't correctly handle non-target relations with RLS, which would get turned into subqueries during planning. Thus an update of the form "UPDATE t1 ... FROM t2 ..." where t1 has inheritance and t2 has RLS quals would fail. Fix by making sure to copy in and update the securityQuals when they exist for non-target relations. process_policies() was adding WCOs to non-target relations, which is unnecessary, and could lead to a lot of wasted time in the rewriter and the planner. Fix by only adding WCO policies when working on the result relation. Also in process_policies, we should be copying the USING policies to the WITH CHECK policies on a per-policy basis, fix by moving the copying up into the per-policy loop. Lastly, as noted by Dean, we were simply adding policies returned by the hook provided to the list of quals being AND'd, meaning that they would actually restrict records returned and there was no option to have internal policies and hook-based policies work together permissively (as all internal policies currently work). Instead, explicitly add support for both permissive and restrictive policies by having a hook for each and combining the results appropriately. To ensure this is all done correctly, add a new test module (test_rls_hooks) to test the various combinations of internal, permissive, and restrictive hook policies. Largely from Dean Rasheed (thanks!): CAEZATCVmFUfUOwwhnBTcgi6AquyjQ0-1fyKd0T3xBWJvn+xsFA@mail.gmail.com Author: Dean Rasheed, though I added the new hooks and test module.
2015-04-22 18:01:06 +02:00
GRANT ALL ON t3 TO public;
Remove WITH OIDS support, change oid catalog column visibility. Previously tables declared WITH OIDS, including a significant fraction of the catalog tables, stored the oid column not as a normal column, but as part of the tuple header. This special column was not shown by default, which was somewhat odd, as it's often (consider e.g. pg_class.oid) one of the more important parts of a row. Neither pg_dump nor COPY included the contents of the oid column by default. The fact that the oid column was not an ordinary column necessitated a significant amount of special case code to support oid columns. That already was painful for the existing, but upcoming work aiming to make table storage pluggable, would have required expanding and duplicating that "specialness" significantly. WITH OIDS has been deprecated since 2005 (commit ff02d0a05280e0). Remove it. Removing includes: - CREATE TABLE and ALTER TABLE syntax for declaring the table to be WITH OIDS has been removed (WITH (oids[ = true]) will error out) - pg_dump does not support dumping tables declared WITH OIDS and will issue a warning when dumping one (and ignore the oid column). - restoring an pg_dump archive with pg_restore will warn when restoring a table with oid contents (and ignore the oid column) - COPY will refuse to load binary dump that includes oids. - pg_upgrade will error out when encountering tables declared WITH OIDS, they have to be altered to remove the oid column first. - Functionality to access the oid of the last inserted row (like plpgsql's RESULT_OID, spi's SPI_lastoid, ...) has been removed. The syntax for declaring a table WITHOUT OIDS (or WITH (oids = false) for CREATE TABLE) is still supported. While that requires a bit of support code, it seems unnecessary to break applications / dumps that do not use oids, and are explicit about not using them. The biggest user of WITH OID columns was postgres' catalog. This commit changes all 'magic' oid columns to be columns that are normally declared and stored. To reduce unnecessary query breakage all the newly added columns are still named 'oid', even if a table's column naming scheme would indicate 'reloid' or such. This obviously requires adapting a lot code, mostly replacing oid access via HeapTupleGetOid() with access to the underlying Form_pg_*->oid column. The bootstrap process now assigns oids for all oid columns in genbki.pl that do not have an explicit value (starting at the largest oid previously used), only oids assigned later by oids will be above FirstBootstrapObjectId. As the oid column now is a normal column the special bootstrap syntax for oids has been removed. Oids are not automatically assigned during insertion anymore, all backend code explicitly assigns oids with GetNewOidWithIndex(). For the rare case that insertions into the catalog via SQL are called for the new pg_nextoid() function can be used (which only works on catalog tables). The fact that oid columns on system tables are now normal columns means that they will be included in the set of columns expanded by * (i.e. SELECT * FROM pg_class will now include the table's oid, previously it did not). It'd not technically be hard to hide oid column by default, but that'd mean confusing behavior would either have to be carried forward forever, or it'd cause breakage down the line. While it's not unlikely that further adjustments are needed, the scope/invasiveness of the patch makes it worthwhile to get merge this now. It's painful to maintain externally, too complicated to commit after the code code freeze, and a dependency of a number of other patches. Catversion bump, for obvious reasons. Author: Andres Freund, with contributions by John Naylor Discussion: https://postgr.es/m/20180930034810.ywp2c7awz7opzcfr@alap3.anarazel.de
2018-11-21 00:36:57 +01:00
COPY t3(id, a,b,c) FROM stdin;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
301 1 xxx X
302 2 yyy Y
303 3 zzz Z
\.
CREATE POLICY p1 ON t1 FOR ALL TO PUBLIC USING (a % 2 = 0); -- be even number
CREATE POLICY p2 ON t2 FOR ALL TO PUBLIC USING (a % 2 = 1); -- be odd number
ALTER TABLE t1 ENABLE ROW LEVEL SECURITY;
ALTER TABLE t2 ENABLE ROW LEVEL SECURITY;
SET SESSION AUTHORIZATION regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SELECT * FROM t1;
EXPLAIN (COSTS OFF) SELECT * FROM t1;
SELECT * FROM t1 WHERE f_leak(b);
EXPLAIN (COSTS OFF) SELECT * FROM t1 WHERE f_leak(b);
-- reference to system column
Remove WITH OIDS support, change oid catalog column visibility. Previously tables declared WITH OIDS, including a significant fraction of the catalog tables, stored the oid column not as a normal column, but as part of the tuple header. This special column was not shown by default, which was somewhat odd, as it's often (consider e.g. pg_class.oid) one of the more important parts of a row. Neither pg_dump nor COPY included the contents of the oid column by default. The fact that the oid column was not an ordinary column necessitated a significant amount of special case code to support oid columns. That already was painful for the existing, but upcoming work aiming to make table storage pluggable, would have required expanding and duplicating that "specialness" significantly. WITH OIDS has been deprecated since 2005 (commit ff02d0a05280e0). Remove it. Removing includes: - CREATE TABLE and ALTER TABLE syntax for declaring the table to be WITH OIDS has been removed (WITH (oids[ = true]) will error out) - pg_dump does not support dumping tables declared WITH OIDS and will issue a warning when dumping one (and ignore the oid column). - restoring an pg_dump archive with pg_restore will warn when restoring a table with oid contents (and ignore the oid column) - COPY will refuse to load binary dump that includes oids. - pg_upgrade will error out when encountering tables declared WITH OIDS, they have to be altered to remove the oid column first. - Functionality to access the oid of the last inserted row (like plpgsql's RESULT_OID, spi's SPI_lastoid, ...) has been removed. The syntax for declaring a table WITHOUT OIDS (or WITH (oids = false) for CREATE TABLE) is still supported. While that requires a bit of support code, it seems unnecessary to break applications / dumps that do not use oids, and are explicit about not using them. The biggest user of WITH OID columns was postgres' catalog. This commit changes all 'magic' oid columns to be columns that are normally declared and stored. To reduce unnecessary query breakage all the newly added columns are still named 'oid', even if a table's column naming scheme would indicate 'reloid' or such. This obviously requires adapting a lot code, mostly replacing oid access via HeapTupleGetOid() with access to the underlying Form_pg_*->oid column. The bootstrap process now assigns oids for all oid columns in genbki.pl that do not have an explicit value (starting at the largest oid previously used), only oids assigned later by oids will be above FirstBootstrapObjectId. As the oid column now is a normal column the special bootstrap syntax for oids has been removed. Oids are not automatically assigned during insertion anymore, all backend code explicitly assigns oids with GetNewOidWithIndex(). For the rare case that insertions into the catalog via SQL are called for the new pg_nextoid() function can be used (which only works on catalog tables). The fact that oid columns on system tables are now normal columns means that they will be included in the set of columns expanded by * (i.e. SELECT * FROM pg_class will now include the table's oid, previously it did not). It'd not technically be hard to hide oid column by default, but that'd mean confusing behavior would either have to be carried forward forever, or it'd cause breakage down the line. While it's not unlikely that further adjustments are needed, the scope/invasiveness of the patch makes it worthwhile to get merge this now. It's painful to maintain externally, too complicated to commit after the code code freeze, and a dependency of a number of other patches. Catversion bump, for obvious reasons. Author: Andres Freund, with contributions by John Naylor Discussion: https://postgr.es/m/20180930034810.ywp2c7awz7opzcfr@alap3.anarazel.de
2018-11-21 00:36:57 +01:00
SELECT tableoid::regclass, * FROM t1;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
EXPLAIN (COSTS OFF) SELECT *, t1 FROM t1;
-- reference to whole-row reference
SELECT *, t1 FROM t1;
EXPLAIN (COSTS OFF) SELECT *, t1 FROM t1;
-- for share/update lock
SELECT * FROM t1 FOR SHARE;
EXPLAIN (COSTS OFF) SELECT * FROM t1 FOR SHARE;
SELECT * FROM t1 WHERE f_leak(b) FOR SHARE;
EXPLAIN (COSTS OFF) SELECT * FROM t1 WHERE f_leak(b) FOR SHARE;
-- union all query
Remove WITH OIDS support, change oid catalog column visibility. Previously tables declared WITH OIDS, including a significant fraction of the catalog tables, stored the oid column not as a normal column, but as part of the tuple header. This special column was not shown by default, which was somewhat odd, as it's often (consider e.g. pg_class.oid) one of the more important parts of a row. Neither pg_dump nor COPY included the contents of the oid column by default. The fact that the oid column was not an ordinary column necessitated a significant amount of special case code to support oid columns. That already was painful for the existing, but upcoming work aiming to make table storage pluggable, would have required expanding and duplicating that "specialness" significantly. WITH OIDS has been deprecated since 2005 (commit ff02d0a05280e0). Remove it. Removing includes: - CREATE TABLE and ALTER TABLE syntax for declaring the table to be WITH OIDS has been removed (WITH (oids[ = true]) will error out) - pg_dump does not support dumping tables declared WITH OIDS and will issue a warning when dumping one (and ignore the oid column). - restoring an pg_dump archive with pg_restore will warn when restoring a table with oid contents (and ignore the oid column) - COPY will refuse to load binary dump that includes oids. - pg_upgrade will error out when encountering tables declared WITH OIDS, they have to be altered to remove the oid column first. - Functionality to access the oid of the last inserted row (like plpgsql's RESULT_OID, spi's SPI_lastoid, ...) has been removed. The syntax for declaring a table WITHOUT OIDS (or WITH (oids = false) for CREATE TABLE) is still supported. While that requires a bit of support code, it seems unnecessary to break applications / dumps that do not use oids, and are explicit about not using them. The biggest user of WITH OID columns was postgres' catalog. This commit changes all 'magic' oid columns to be columns that are normally declared and stored. To reduce unnecessary query breakage all the newly added columns are still named 'oid', even if a table's column naming scheme would indicate 'reloid' or such. This obviously requires adapting a lot code, mostly replacing oid access via HeapTupleGetOid() with access to the underlying Form_pg_*->oid column. The bootstrap process now assigns oids for all oid columns in genbki.pl that do not have an explicit value (starting at the largest oid previously used), only oids assigned later by oids will be above FirstBootstrapObjectId. As the oid column now is a normal column the special bootstrap syntax for oids has been removed. Oids are not automatically assigned during insertion anymore, all backend code explicitly assigns oids with GetNewOidWithIndex(). For the rare case that insertions into the catalog via SQL are called for the new pg_nextoid() function can be used (which only works on catalog tables). The fact that oid columns on system tables are now normal columns means that they will be included in the set of columns expanded by * (i.e. SELECT * FROM pg_class will now include the table's oid, previously it did not). It'd not technically be hard to hide oid column by default, but that'd mean confusing behavior would either have to be carried forward forever, or it'd cause breakage down the line. While it's not unlikely that further adjustments are needed, the scope/invasiveness of the patch makes it worthwhile to get merge this now. It's painful to maintain externally, too complicated to commit after the code code freeze, and a dependency of a number of other patches. Catversion bump, for obvious reasons. Author: Andres Freund, with contributions by John Naylor Discussion: https://postgr.es/m/20180930034810.ywp2c7awz7opzcfr@alap3.anarazel.de
2018-11-21 00:36:57 +01:00
SELECT a, b, tableoid::regclass FROM t2 UNION ALL SELECT a, b, tableoid::regclass FROM t3;
EXPLAIN (COSTS OFF) SELECT a, b, tableoid::regclass FROM t2 UNION ALL SELECT a, b, tableoid::regclass FROM t3;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
-- superuser is allowed to bypass RLS checks
RESET SESSION AUTHORIZATION;
SET row_security TO OFF;
SELECT * FROM t1 WHERE f_leak(b);
EXPLAIN (COSTS OFF) SELECT * FROM t1 WHERE f_leak(b);
-- non-superuser with bypass privilege can bypass RLS policy when disabled
SET SESSION AUTHORIZATION regress_rls_exempt_user;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SET row_security TO OFF;
SELECT * FROM t1 WHERE f_leak(b);
EXPLAIN (COSTS OFF) SELECT * FROM t1 WHERE f_leak(b);
--
-- Partitioned Tables
--
SET SESSION AUTHORIZATION regress_rls_alice;
CREATE TABLE part_document (
did int,
cid int,
dlevel int not null,
dauthor name,
dtitle text
) PARTITION BY RANGE (cid);
GRANT ALL ON part_document TO public;
-- Create partitions for document categories
CREATE TABLE part_document_fiction PARTITION OF part_document FOR VALUES FROM (11) to (12);
CREATE TABLE part_document_satire PARTITION OF part_document FOR VALUES FROM (55) to (56);
CREATE TABLE part_document_nonfiction PARTITION OF part_document FOR VALUES FROM (99) to (100);
GRANT ALL ON part_document_fiction TO public;
GRANT ALL ON part_document_satire TO public;
GRANT ALL ON part_document_nonfiction TO public;
INSERT INTO part_document VALUES
( 1, 11, 1, 'regress_rls_bob', 'my first novel'),
( 2, 11, 2, 'regress_rls_bob', 'my second novel'),
( 3, 99, 2, 'regress_rls_bob', 'my science textbook'),
( 4, 55, 1, 'regress_rls_bob', 'my first satire'),
( 5, 99, 2, 'regress_rls_bob', 'my history book'),
( 6, 11, 1, 'regress_rls_carol', 'great science fiction'),
( 7, 99, 2, 'regress_rls_carol', 'great technology book'),
( 8, 55, 2, 'regress_rls_carol', 'great satire'),
( 9, 11, 1, 'regress_rls_dave', 'awesome science fiction'),
(10, 99, 2, 'regress_rls_dave', 'awesome technology book');
ALTER TABLE part_document ENABLE ROW LEVEL SECURITY;
-- Create policy on parent
-- user's security level must be higher than or equal to document's
CREATE POLICY pp1 ON part_document AS PERMISSIVE
USING (dlevel <= (SELECT seclv FROM uaccount WHERE pguser = current_user));
-- Dave is only allowed to see cid < 55
CREATE POLICY pp1r ON part_document AS RESTRICTIVE TO regress_rls_dave
USING (cid < 55);
\d+ part_document
SELECT * FROM pg_policies WHERE schemaname = 'regress_rls_schema' AND tablename like '%part_document%' ORDER BY policyname;
-- viewpoint from regress_rls_bob
SET SESSION AUTHORIZATION regress_rls_bob;
SET row_security TO ON;
SELECT * FROM part_document WHERE f_leak(dtitle) ORDER BY did;
EXPLAIN (COSTS OFF) SELECT * FROM part_document WHERE f_leak(dtitle);
-- viewpoint from regress_rls_carol
SET SESSION AUTHORIZATION regress_rls_carol;
SELECT * FROM part_document WHERE f_leak(dtitle) ORDER BY did;
EXPLAIN (COSTS OFF) SELECT * FROM part_document WHERE f_leak(dtitle);
-- viewpoint from regress_rls_dave
SET SESSION AUTHORIZATION regress_rls_dave;
SELECT * FROM part_document WHERE f_leak(dtitle) ORDER BY did;
EXPLAIN (COSTS OFF) SELECT * FROM part_document WHERE f_leak(dtitle);
-- pp1 ERROR
INSERT INTO part_document VALUES (100, 11, 5, 'regress_rls_dave', 'testing pp1'); -- fail
-- pp1r ERROR
INSERT INTO part_document VALUES (100, 99, 1, 'regress_rls_dave', 'testing pp1r'); -- fail
-- Show that RLS policy does not apply for direct inserts to children
-- This should fail with RLS POLICY pp1r violation.
INSERT INTO part_document VALUES (100, 55, 1, 'regress_rls_dave', 'testing RLS with partitions'); -- fail
-- But this should succeed.
INSERT INTO part_document_satire VALUES (100, 55, 1, 'regress_rls_dave', 'testing RLS with partitions'); -- success
-- We still cannot see the row using the parent
SELECT * FROM part_document WHERE f_leak(dtitle) ORDER BY did;
-- But we can if we look directly
SELECT * FROM part_document_satire WHERE f_leak(dtitle) ORDER BY did;
-- Turn on RLS and create policy on child to show RLS is checked before constraints
SET SESSION AUTHORIZATION regress_rls_alice;
ALTER TABLE part_document_satire ENABLE ROW LEVEL SECURITY;
CREATE POLICY pp3 ON part_document_satire AS RESTRICTIVE
USING (cid < 55);
-- This should fail with RLS violation now.
SET SESSION AUTHORIZATION regress_rls_dave;
INSERT INTO part_document_satire VALUES (101, 55, 1, 'regress_rls_dave', 'testing RLS with partitions'); -- fail
-- And now we cannot see directly into the partition either, due to RLS
SELECT * FROM part_document_satire WHERE f_leak(dtitle) ORDER BY did;
-- The parent looks same as before
-- viewpoint from regress_rls_dave
SELECT * FROM part_document WHERE f_leak(dtitle) ORDER BY did;
EXPLAIN (COSTS OFF) SELECT * FROM part_document WHERE f_leak(dtitle);
-- viewpoint from regress_rls_carol
SET SESSION AUTHORIZATION regress_rls_carol;
SELECT * FROM part_document WHERE f_leak(dtitle) ORDER BY did;
EXPLAIN (COSTS OFF) SELECT * FROM part_document WHERE f_leak(dtitle);
-- only owner can change policies
ALTER POLICY pp1 ON part_document USING (true); --fail
DROP POLICY pp1 ON part_document; --fail
SET SESSION AUTHORIZATION regress_rls_alice;
ALTER POLICY pp1 ON part_document USING (dauthor = current_user);
-- viewpoint from regress_rls_bob again
SET SESSION AUTHORIZATION regress_rls_bob;
SELECT * FROM part_document WHERE f_leak(dtitle) ORDER BY did;
-- viewpoint from rls_regres_carol again
SET SESSION AUTHORIZATION regress_rls_carol;
SELECT * FROM part_document WHERE f_leak(dtitle) ORDER BY did;
EXPLAIN (COSTS OFF) SELECT * FROM part_document WHERE f_leak(dtitle);
-- database superuser does bypass RLS policy when enabled
RESET SESSION AUTHORIZATION;
SET row_security TO ON;
SELECT * FROM part_document ORDER BY did;
SELECT * FROM part_document_satire ORDER by did;
-- database non-superuser with bypass privilege can bypass RLS policy when disabled
SET SESSION AUTHORIZATION regress_rls_exempt_user;
SET row_security TO OFF;
SELECT * FROM part_document ORDER BY did;
SELECT * FROM part_document_satire ORDER by did;
-- RLS policy does not apply to table owner when RLS enabled.
SET SESSION AUTHORIZATION regress_rls_alice;
SET row_security TO ON;
SELECT * FROM part_document ORDER by did;
SELECT * FROM part_document_satire ORDER by did;
-- When RLS disabled, other users get ERROR.
SET SESSION AUTHORIZATION regress_rls_dave;
SET row_security TO OFF;
SELECT * FROM part_document ORDER by did;
SELECT * FROM part_document_satire ORDER by did;
-- Check behavior with a policy that uses a SubPlan not an InitPlan.
SET SESSION AUTHORIZATION regress_rls_alice;
SET row_security TO ON;
CREATE POLICY pp3 ON part_document AS RESTRICTIVE
USING ((SELECT dlevel <= seclv FROM uaccount WHERE pguser = current_user));
SET SESSION AUTHORIZATION regress_rls_carol;
INSERT INTO part_document VALUES (100, 11, 5, 'regress_rls_carol', 'testing pp3'); -- fail
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
----- Dependencies -----
SET SESSION AUTHORIZATION regress_rls_alice;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SET row_security TO ON;
CREATE TABLE dependee (x integer, y integer);
CREATE TABLE dependent (x integer, y integer);
CREATE POLICY d1 ON dependent FOR ALL
TO PUBLIC
USING (x = (SELECT d.x FROM dependee d WHERE d.y = y));
DROP TABLE dependee; -- Should fail without CASCADE due to dependency on row security qual?
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
DROP TABLE dependee CASCADE;
EXPLAIN (COSTS OFF) SELECT * FROM dependent; -- After drop, should be unqualified
----- RECURSION ----
--
-- Simple recursion
--
SET SESSION AUTHORIZATION regress_rls_alice;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
CREATE TABLE rec1 (x integer, y integer);
CREATE POLICY r1 ON rec1 USING (x = (SELECT r.x FROM rec1 r WHERE y = r.y));
ALTER TABLE rec1 ENABLE ROW LEVEL SECURITY;
SET SESSION AUTHORIZATION regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SELECT * FROM rec1; -- fail, direct recursion
--
-- Mutual recursion
--
SET SESSION AUTHORIZATION regress_rls_alice;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
CREATE TABLE rec2 (a integer, b integer);
ALTER POLICY r1 ON rec1 USING (x = (SELECT a FROM rec2 WHERE b = y));
CREATE POLICY r2 ON rec2 USING (a = (SELECT x FROM rec1 WHERE y = b));
ALTER TABLE rec2 ENABLE ROW LEVEL SECURITY;
SET SESSION AUTHORIZATION regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SELECT * FROM rec1; -- fail, mutual recursion
--
-- Mutual recursion via views
--
SET SESSION AUTHORIZATION regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
CREATE VIEW rec1v AS SELECT * FROM rec1;
CREATE VIEW rec2v AS SELECT * FROM rec2;
SET SESSION AUTHORIZATION regress_rls_alice;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
ALTER POLICY r1 ON rec1 USING (x = (SELECT a FROM rec2v WHERE b = y));
ALTER POLICY r2 ON rec2 USING (a = (SELECT x FROM rec1v WHERE y = b));
SET SESSION AUTHORIZATION regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SELECT * FROM rec1; -- fail, mutual recursion via views
--
-- Mutual recursion via .s.b views
--
SET SESSION AUTHORIZATION regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
DROP VIEW rec1v, rec2v CASCADE;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
CREATE VIEW rec1v WITH (security_barrier) AS SELECT * FROM rec1;
CREATE VIEW rec2v WITH (security_barrier) AS SELECT * FROM rec2;
SET SESSION AUTHORIZATION regress_rls_alice;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
CREATE POLICY r1 ON rec1 USING (x = (SELECT a FROM rec2v WHERE b = y));
CREATE POLICY r2 ON rec2 USING (a = (SELECT x FROM rec1v WHERE y = b));
SET SESSION AUTHORIZATION regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SELECT * FROM rec1; -- fail, mutual recursion via s.b. views
--
-- recursive RLS and VIEWs in policy
--
SET SESSION AUTHORIZATION regress_rls_alice;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
CREATE TABLE s1 (a int, b text);
INSERT INTO s1 (SELECT x, md5(x::text) FROM generate_series(-10,10) x);
CREATE TABLE s2 (x int, y text);
INSERT INTO s2 (SELECT x, md5(x::text) FROM generate_series(-6,6) x);
GRANT SELECT ON s1, s2 TO regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
CREATE POLICY p1 ON s1 USING (a in (select x from s2 where y like '%2f%'));
CREATE POLICY p2 ON s2 USING (x in (select a from s1 where b like '%22%'));
CREATE POLICY p3 ON s1 FOR INSERT WITH CHECK (a = (SELECT a FROM s1));
ALTER TABLE s1 ENABLE ROW LEVEL SECURITY;
ALTER TABLE s2 ENABLE ROW LEVEL SECURITY;
SET SESSION AUTHORIZATION regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
CREATE VIEW v2 AS SELECT * FROM s2 WHERE y like '%af%';
SELECT * FROM s1 WHERE f_leak(b); -- fail (infinite recursion)
INSERT INTO s1 VALUES (1, 'foo'); -- fail (infinite recursion)
SET SESSION AUTHORIZATION regress_rls_alice;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
DROP POLICY p3 on s1;
ALTER POLICY p2 ON s2 USING (x % 2 = 0);
SET SESSION AUTHORIZATION regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SELECT * FROM s1 WHERE f_leak(b); -- OK
EXPLAIN (COSTS OFF) SELECT * FROM only s1 WHERE f_leak(b);
SET SESSION AUTHORIZATION regress_rls_alice;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
ALTER POLICY p1 ON s1 USING (a in (select x from v2)); -- using VIEW in RLS policy
SET SESSION AUTHORIZATION regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SELECT * FROM s1 WHERE f_leak(b); -- OK
EXPLAIN (COSTS OFF) SELECT * FROM s1 WHERE f_leak(b);
SELECT (SELECT x FROM s1 LIMIT 1) xx, * FROM s2 WHERE y like '%28%';
EXPLAIN (COSTS OFF) SELECT (SELECT x FROM s1 LIMIT 1) xx, * FROM s2 WHERE y like '%28%';
SET SESSION AUTHORIZATION regress_rls_alice;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
ALTER POLICY p2 ON s2 USING (x in (select a from s1 where b like '%d2%'));
SET SESSION AUTHORIZATION regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SELECT * FROM s1 WHERE f_leak(b); -- fail (infinite recursion via view)
-- prepared statement with regress_rls_alice privilege
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
PREPARE p1(int) AS SELECT * FROM t1 WHERE a <= $1;
EXECUTE p1(2);
EXPLAIN (COSTS OFF) EXECUTE p1(2);
-- superuser is allowed to bypass RLS checks
RESET SESSION AUTHORIZATION;
SET row_security TO OFF;
SELECT * FROM t1 WHERE f_leak(b);
EXPLAIN (COSTS OFF) SELECT * FROM t1 WHERE f_leak(b);
-- plan cache should be invalidated
EXECUTE p1(2);
EXPLAIN (COSTS OFF) EXECUTE p1(2);
PREPARE p2(int) AS SELECT * FROM t1 WHERE a = $1;
EXECUTE p2(2);
EXPLAIN (COSTS OFF) EXECUTE p2(2);
-- also, case when privilege switch from superuser
SET SESSION AUTHORIZATION regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SET row_security TO ON;
EXECUTE p2(2);
EXPLAIN (COSTS OFF) EXECUTE p2(2);
--
-- UPDATE / DELETE and Row-level security
--
SET SESSION AUTHORIZATION regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
EXPLAIN (COSTS OFF) UPDATE t1 SET b = b || b WHERE f_leak(b);
UPDATE t1 SET b = b || b WHERE f_leak(b);
EXPLAIN (COSTS OFF) UPDATE only t1 SET b = b || '_updt' WHERE f_leak(b);
UPDATE only t1 SET b = b || '_updt' WHERE f_leak(b);
-- returning clause with system column
Remove WITH OIDS support, change oid catalog column visibility. Previously tables declared WITH OIDS, including a significant fraction of the catalog tables, stored the oid column not as a normal column, but as part of the tuple header. This special column was not shown by default, which was somewhat odd, as it's often (consider e.g. pg_class.oid) one of the more important parts of a row. Neither pg_dump nor COPY included the contents of the oid column by default. The fact that the oid column was not an ordinary column necessitated a significant amount of special case code to support oid columns. That already was painful for the existing, but upcoming work aiming to make table storage pluggable, would have required expanding and duplicating that "specialness" significantly. WITH OIDS has been deprecated since 2005 (commit ff02d0a05280e0). Remove it. Removing includes: - CREATE TABLE and ALTER TABLE syntax for declaring the table to be WITH OIDS has been removed (WITH (oids[ = true]) will error out) - pg_dump does not support dumping tables declared WITH OIDS and will issue a warning when dumping one (and ignore the oid column). - restoring an pg_dump archive with pg_restore will warn when restoring a table with oid contents (and ignore the oid column) - COPY will refuse to load binary dump that includes oids. - pg_upgrade will error out when encountering tables declared WITH OIDS, they have to be altered to remove the oid column first. - Functionality to access the oid of the last inserted row (like plpgsql's RESULT_OID, spi's SPI_lastoid, ...) has been removed. The syntax for declaring a table WITHOUT OIDS (or WITH (oids = false) for CREATE TABLE) is still supported. While that requires a bit of support code, it seems unnecessary to break applications / dumps that do not use oids, and are explicit about not using them. The biggest user of WITH OID columns was postgres' catalog. This commit changes all 'magic' oid columns to be columns that are normally declared and stored. To reduce unnecessary query breakage all the newly added columns are still named 'oid', even if a table's column naming scheme would indicate 'reloid' or such. This obviously requires adapting a lot code, mostly replacing oid access via HeapTupleGetOid() with access to the underlying Form_pg_*->oid column. The bootstrap process now assigns oids for all oid columns in genbki.pl that do not have an explicit value (starting at the largest oid previously used), only oids assigned later by oids will be above FirstBootstrapObjectId. As the oid column now is a normal column the special bootstrap syntax for oids has been removed. Oids are not automatically assigned during insertion anymore, all backend code explicitly assigns oids with GetNewOidWithIndex(). For the rare case that insertions into the catalog via SQL are called for the new pg_nextoid() function can be used (which only works on catalog tables). The fact that oid columns on system tables are now normal columns means that they will be included in the set of columns expanded by * (i.e. SELECT * FROM pg_class will now include the table's oid, previously it did not). It'd not technically be hard to hide oid column by default, but that'd mean confusing behavior would either have to be carried forward forever, or it'd cause breakage down the line. While it's not unlikely that further adjustments are needed, the scope/invasiveness of the patch makes it worthwhile to get merge this now. It's painful to maintain externally, too complicated to commit after the code code freeze, and a dependency of a number of other patches. Catversion bump, for obvious reasons. Author: Andres Freund, with contributions by John Naylor Discussion: https://postgr.es/m/20180930034810.ywp2c7awz7opzcfr@alap3.anarazel.de
2018-11-21 00:36:57 +01:00
UPDATE only t1 SET b = b WHERE f_leak(b) RETURNING tableoid::regclass, *, t1;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
UPDATE t1 SET b = b WHERE f_leak(b) RETURNING *;
Remove WITH OIDS support, change oid catalog column visibility. Previously tables declared WITH OIDS, including a significant fraction of the catalog tables, stored the oid column not as a normal column, but as part of the tuple header. This special column was not shown by default, which was somewhat odd, as it's often (consider e.g. pg_class.oid) one of the more important parts of a row. Neither pg_dump nor COPY included the contents of the oid column by default. The fact that the oid column was not an ordinary column necessitated a significant amount of special case code to support oid columns. That already was painful for the existing, but upcoming work aiming to make table storage pluggable, would have required expanding and duplicating that "specialness" significantly. WITH OIDS has been deprecated since 2005 (commit ff02d0a05280e0). Remove it. Removing includes: - CREATE TABLE and ALTER TABLE syntax for declaring the table to be WITH OIDS has been removed (WITH (oids[ = true]) will error out) - pg_dump does not support dumping tables declared WITH OIDS and will issue a warning when dumping one (and ignore the oid column). - restoring an pg_dump archive with pg_restore will warn when restoring a table with oid contents (and ignore the oid column) - COPY will refuse to load binary dump that includes oids. - pg_upgrade will error out when encountering tables declared WITH OIDS, they have to be altered to remove the oid column first. - Functionality to access the oid of the last inserted row (like plpgsql's RESULT_OID, spi's SPI_lastoid, ...) has been removed. The syntax for declaring a table WITHOUT OIDS (or WITH (oids = false) for CREATE TABLE) is still supported. While that requires a bit of support code, it seems unnecessary to break applications / dumps that do not use oids, and are explicit about not using them. The biggest user of WITH OID columns was postgres' catalog. This commit changes all 'magic' oid columns to be columns that are normally declared and stored. To reduce unnecessary query breakage all the newly added columns are still named 'oid', even if a table's column naming scheme would indicate 'reloid' or such. This obviously requires adapting a lot code, mostly replacing oid access via HeapTupleGetOid() with access to the underlying Form_pg_*->oid column. The bootstrap process now assigns oids for all oid columns in genbki.pl that do not have an explicit value (starting at the largest oid previously used), only oids assigned later by oids will be above FirstBootstrapObjectId. As the oid column now is a normal column the special bootstrap syntax for oids has been removed. Oids are not automatically assigned during insertion anymore, all backend code explicitly assigns oids with GetNewOidWithIndex(). For the rare case that insertions into the catalog via SQL are called for the new pg_nextoid() function can be used (which only works on catalog tables). The fact that oid columns on system tables are now normal columns means that they will be included in the set of columns expanded by * (i.e. SELECT * FROM pg_class will now include the table's oid, previously it did not). It'd not technically be hard to hide oid column by default, but that'd mean confusing behavior would either have to be carried forward forever, or it'd cause breakage down the line. While it's not unlikely that further adjustments are needed, the scope/invasiveness of the patch makes it worthwhile to get merge this now. It's painful to maintain externally, too complicated to commit after the code code freeze, and a dependency of a number of other patches. Catversion bump, for obvious reasons. Author: Andres Freund, with contributions by John Naylor Discussion: https://postgr.es/m/20180930034810.ywp2c7awz7opzcfr@alap3.anarazel.de
2018-11-21 00:36:57 +01:00
UPDATE t1 SET b = b WHERE f_leak(b) RETURNING tableoid::regclass, *, t1;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
RLS fixes, new hooks, and new test module In prepend_row_security_policies(), defaultDeny was always true, so if there were any hook policies, the RLS policies on the table would just get discarded. Fixed to start off with defaultDeny as false and then properly set later if we detect that only the default deny policy exists for the internal policies. The infinite recursion detection in fireRIRrules() didn't properly manage the activeRIRs list in the case of WCOs, so it would incorrectly report infinite recusion if the same relation with RLS appeared more than once in the rtable, for example "UPDATE t ... FROM t ...". Further, the RLS expansion code in fireRIRrules() was handling RLS in the main loop through the rtable, which lead to RTEs being visited twice if they contained sublink subqueries, which prepend_row_security_policies() attempted to handle by exiting early if the RTE already had securityQuals. That doesn't work, however, since if the query involved a security barrier view on top of a table with RLS, the RTE would already have securityQuals (from the view) by the time fireRIRrules() was invoked, and so the table's RLS policies would be ignored. This is fixed in fireRIRrules() by handling RLS in a separate loop at the end, after dealing with any other sublink subqueries, thus ensuring that each RTE is only visited once for RLS expansion. The inheritance planner code didn't correctly handle non-target relations with RLS, which would get turned into subqueries during planning. Thus an update of the form "UPDATE t1 ... FROM t2 ..." where t1 has inheritance and t2 has RLS quals would fail. Fix by making sure to copy in and update the securityQuals when they exist for non-target relations. process_policies() was adding WCOs to non-target relations, which is unnecessary, and could lead to a lot of wasted time in the rewriter and the planner. Fix by only adding WCO policies when working on the result relation. Also in process_policies, we should be copying the USING policies to the WITH CHECK policies on a per-policy basis, fix by moving the copying up into the per-policy loop. Lastly, as noted by Dean, we were simply adding policies returned by the hook provided to the list of quals being AND'd, meaning that they would actually restrict records returned and there was no option to have internal policies and hook-based policies work together permissively (as all internal policies currently work). Instead, explicitly add support for both permissive and restrictive policies by having a hook for each and combining the results appropriately. To ensure this is all done correctly, add a new test module (test_rls_hooks) to test the various combinations of internal, permissive, and restrictive hook policies. Largely from Dean Rasheed (thanks!): CAEZATCVmFUfUOwwhnBTcgi6AquyjQ0-1fyKd0T3xBWJvn+xsFA@mail.gmail.com Author: Dean Rasheed, though I added the new hooks and test module.
2015-04-22 18:01:06 +02:00
-- updates with from clause
EXPLAIN (COSTS OFF) UPDATE t2 SET b=t2.b FROM t3
WHERE t2.a = 3 and t3.a = 2 AND f_leak(t2.b) AND f_leak(t3.b);
UPDATE t2 SET b=t2.b FROM t3
WHERE t2.a = 3 and t3.a = 2 AND f_leak(t2.b) AND f_leak(t3.b);
EXPLAIN (COSTS OFF) UPDATE t1 SET b=t1.b FROM t2
WHERE t1.a = 3 and t2.a = 3 AND f_leak(t1.b) AND f_leak(t2.b);
UPDATE t1 SET b=t1.b FROM t2
WHERE t1.a = 3 and t2.a = 3 AND f_leak(t1.b) AND f_leak(t2.b);
EXPLAIN (COSTS OFF) UPDATE t2 SET b=t2.b FROM t1
WHERE t1.a = 3 and t2.a = 3 AND f_leak(t1.b) AND f_leak(t2.b);
UPDATE t2 SET b=t2.b FROM t1
WHERE t1.a = 3 and t2.a = 3 AND f_leak(t1.b) AND f_leak(t2.b);
-- updates with from clause self join
EXPLAIN (COSTS OFF) UPDATE t2 t2_1 SET b = t2_2.b FROM t2 t2_2
WHERE t2_1.a = 3 AND t2_2.a = t2_1.a AND t2_2.b = t2_1.b
AND f_leak(t2_1.b) AND f_leak(t2_2.b) RETURNING *, t2_1, t2_2;
UPDATE t2 t2_1 SET b = t2_2.b FROM t2 t2_2
WHERE t2_1.a = 3 AND t2_2.a = t2_1.a AND t2_2.b = t2_1.b
AND f_leak(t2_1.b) AND f_leak(t2_2.b) RETURNING *, t2_1, t2_2;
EXPLAIN (COSTS OFF) UPDATE t1 t1_1 SET b = t1_2.b FROM t1 t1_2
WHERE t1_1.a = 4 AND t1_2.a = t1_1.a AND t1_2.b = t1_1.b
AND f_leak(t1_1.b) AND f_leak(t1_2.b) RETURNING *, t1_1, t1_2;
UPDATE t1 t1_1 SET b = t1_2.b FROM t1 t1_2
WHERE t1_1.a = 4 AND t1_2.a = t1_1.a AND t1_2.b = t1_1.b
AND f_leak(t1_1.b) AND f_leak(t1_2.b) RETURNING *, t1_1, t1_2;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
RESET SESSION AUTHORIZATION;
SET row_security TO OFF;
RLS fixes, new hooks, and new test module In prepend_row_security_policies(), defaultDeny was always true, so if there were any hook policies, the RLS policies on the table would just get discarded. Fixed to start off with defaultDeny as false and then properly set later if we detect that only the default deny policy exists for the internal policies. The infinite recursion detection in fireRIRrules() didn't properly manage the activeRIRs list in the case of WCOs, so it would incorrectly report infinite recusion if the same relation with RLS appeared more than once in the rtable, for example "UPDATE t ... FROM t ...". Further, the RLS expansion code in fireRIRrules() was handling RLS in the main loop through the rtable, which lead to RTEs being visited twice if they contained sublink subqueries, which prepend_row_security_policies() attempted to handle by exiting early if the RTE already had securityQuals. That doesn't work, however, since if the query involved a security barrier view on top of a table with RLS, the RTE would already have securityQuals (from the view) by the time fireRIRrules() was invoked, and so the table's RLS policies would be ignored. This is fixed in fireRIRrules() by handling RLS in a separate loop at the end, after dealing with any other sublink subqueries, thus ensuring that each RTE is only visited once for RLS expansion. The inheritance planner code didn't correctly handle non-target relations with RLS, which would get turned into subqueries during planning. Thus an update of the form "UPDATE t1 ... FROM t2 ..." where t1 has inheritance and t2 has RLS quals would fail. Fix by making sure to copy in and update the securityQuals when they exist for non-target relations. process_policies() was adding WCOs to non-target relations, which is unnecessary, and could lead to a lot of wasted time in the rewriter and the planner. Fix by only adding WCO policies when working on the result relation. Also in process_policies, we should be copying the USING policies to the WITH CHECK policies on a per-policy basis, fix by moving the copying up into the per-policy loop. Lastly, as noted by Dean, we were simply adding policies returned by the hook provided to the list of quals being AND'd, meaning that they would actually restrict records returned and there was no option to have internal policies and hook-based policies work together permissively (as all internal policies currently work). Instead, explicitly add support for both permissive and restrictive policies by having a hook for each and combining the results appropriately. To ensure this is all done correctly, add a new test module (test_rls_hooks) to test the various combinations of internal, permissive, and restrictive hook policies. Largely from Dean Rasheed (thanks!): CAEZATCVmFUfUOwwhnBTcgi6AquyjQ0-1fyKd0T3xBWJvn+xsFA@mail.gmail.com Author: Dean Rasheed, though I added the new hooks and test module.
2015-04-22 18:01:06 +02:00
SELECT * FROM t1 ORDER BY a,b;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SET SESSION AUTHORIZATION regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SET row_security TO ON;
EXPLAIN (COSTS OFF) DELETE FROM only t1 WHERE f_leak(b);
EXPLAIN (COSTS OFF) DELETE FROM t1 WHERE f_leak(b);
Remove WITH OIDS support, change oid catalog column visibility. Previously tables declared WITH OIDS, including a significant fraction of the catalog tables, stored the oid column not as a normal column, but as part of the tuple header. This special column was not shown by default, which was somewhat odd, as it's often (consider e.g. pg_class.oid) one of the more important parts of a row. Neither pg_dump nor COPY included the contents of the oid column by default. The fact that the oid column was not an ordinary column necessitated a significant amount of special case code to support oid columns. That already was painful for the existing, but upcoming work aiming to make table storage pluggable, would have required expanding and duplicating that "specialness" significantly. WITH OIDS has been deprecated since 2005 (commit ff02d0a05280e0). Remove it. Removing includes: - CREATE TABLE and ALTER TABLE syntax for declaring the table to be WITH OIDS has been removed (WITH (oids[ = true]) will error out) - pg_dump does not support dumping tables declared WITH OIDS and will issue a warning when dumping one (and ignore the oid column). - restoring an pg_dump archive with pg_restore will warn when restoring a table with oid contents (and ignore the oid column) - COPY will refuse to load binary dump that includes oids. - pg_upgrade will error out when encountering tables declared WITH OIDS, they have to be altered to remove the oid column first. - Functionality to access the oid of the last inserted row (like plpgsql's RESULT_OID, spi's SPI_lastoid, ...) has been removed. The syntax for declaring a table WITHOUT OIDS (or WITH (oids = false) for CREATE TABLE) is still supported. While that requires a bit of support code, it seems unnecessary to break applications / dumps that do not use oids, and are explicit about not using them. The biggest user of WITH OID columns was postgres' catalog. This commit changes all 'magic' oid columns to be columns that are normally declared and stored. To reduce unnecessary query breakage all the newly added columns are still named 'oid', even if a table's column naming scheme would indicate 'reloid' or such. This obviously requires adapting a lot code, mostly replacing oid access via HeapTupleGetOid() with access to the underlying Form_pg_*->oid column. The bootstrap process now assigns oids for all oid columns in genbki.pl that do not have an explicit value (starting at the largest oid previously used), only oids assigned later by oids will be above FirstBootstrapObjectId. As the oid column now is a normal column the special bootstrap syntax for oids has been removed. Oids are not automatically assigned during insertion anymore, all backend code explicitly assigns oids with GetNewOidWithIndex(). For the rare case that insertions into the catalog via SQL are called for the new pg_nextoid() function can be used (which only works on catalog tables). The fact that oid columns on system tables are now normal columns means that they will be included in the set of columns expanded by * (i.e. SELECT * FROM pg_class will now include the table's oid, previously it did not). It'd not technically be hard to hide oid column by default, but that'd mean confusing behavior would either have to be carried forward forever, or it'd cause breakage down the line. While it's not unlikely that further adjustments are needed, the scope/invasiveness of the patch makes it worthwhile to get merge this now. It's painful to maintain externally, too complicated to commit after the code code freeze, and a dependency of a number of other patches. Catversion bump, for obvious reasons. Author: Andres Freund, with contributions by John Naylor Discussion: https://postgr.es/m/20180930034810.ywp2c7awz7opzcfr@alap3.anarazel.de
2018-11-21 00:36:57 +01:00
DELETE FROM only t1 WHERE f_leak(b) RETURNING tableoid::regclass, *, t1;
DELETE FROM t1 WHERE f_leak(b) RETURNING tableoid::regclass, *, t1;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
RLS fixes, new hooks, and new test module In prepend_row_security_policies(), defaultDeny was always true, so if there were any hook policies, the RLS policies on the table would just get discarded. Fixed to start off with defaultDeny as false and then properly set later if we detect that only the default deny policy exists for the internal policies. The infinite recursion detection in fireRIRrules() didn't properly manage the activeRIRs list in the case of WCOs, so it would incorrectly report infinite recusion if the same relation with RLS appeared more than once in the rtable, for example "UPDATE t ... FROM t ...". Further, the RLS expansion code in fireRIRrules() was handling RLS in the main loop through the rtable, which lead to RTEs being visited twice if they contained sublink subqueries, which prepend_row_security_policies() attempted to handle by exiting early if the RTE already had securityQuals. That doesn't work, however, since if the query involved a security barrier view on top of a table with RLS, the RTE would already have securityQuals (from the view) by the time fireRIRrules() was invoked, and so the table's RLS policies would be ignored. This is fixed in fireRIRrules() by handling RLS in a separate loop at the end, after dealing with any other sublink subqueries, thus ensuring that each RTE is only visited once for RLS expansion. The inheritance planner code didn't correctly handle non-target relations with RLS, which would get turned into subqueries during planning. Thus an update of the form "UPDATE t1 ... FROM t2 ..." where t1 has inheritance and t2 has RLS quals would fail. Fix by making sure to copy in and update the securityQuals when they exist for non-target relations. process_policies() was adding WCOs to non-target relations, which is unnecessary, and could lead to a lot of wasted time in the rewriter and the planner. Fix by only adding WCO policies when working on the result relation. Also in process_policies, we should be copying the USING policies to the WITH CHECK policies on a per-policy basis, fix by moving the copying up into the per-policy loop. Lastly, as noted by Dean, we were simply adding policies returned by the hook provided to the list of quals being AND'd, meaning that they would actually restrict records returned and there was no option to have internal policies and hook-based policies work together permissively (as all internal policies currently work). Instead, explicitly add support for both permissive and restrictive policies by having a hook for each and combining the results appropriately. To ensure this is all done correctly, add a new test module (test_rls_hooks) to test the various combinations of internal, permissive, and restrictive hook policies. Largely from Dean Rasheed (thanks!): CAEZATCVmFUfUOwwhnBTcgi6AquyjQ0-1fyKd0T3xBWJvn+xsFA@mail.gmail.com Author: Dean Rasheed, though I added the new hooks and test module.
2015-04-22 18:01:06 +02:00
--
-- S.b. view on top of Row-level security
--
SET SESSION AUTHORIZATION regress_rls_alice;
RLS fixes, new hooks, and new test module In prepend_row_security_policies(), defaultDeny was always true, so if there were any hook policies, the RLS policies on the table would just get discarded. Fixed to start off with defaultDeny as false and then properly set later if we detect that only the default deny policy exists for the internal policies. The infinite recursion detection in fireRIRrules() didn't properly manage the activeRIRs list in the case of WCOs, so it would incorrectly report infinite recusion if the same relation with RLS appeared more than once in the rtable, for example "UPDATE t ... FROM t ...". Further, the RLS expansion code in fireRIRrules() was handling RLS in the main loop through the rtable, which lead to RTEs being visited twice if they contained sublink subqueries, which prepend_row_security_policies() attempted to handle by exiting early if the RTE already had securityQuals. That doesn't work, however, since if the query involved a security barrier view on top of a table with RLS, the RTE would already have securityQuals (from the view) by the time fireRIRrules() was invoked, and so the table's RLS policies would be ignored. This is fixed in fireRIRrules() by handling RLS in a separate loop at the end, after dealing with any other sublink subqueries, thus ensuring that each RTE is only visited once for RLS expansion. The inheritance planner code didn't correctly handle non-target relations with RLS, which would get turned into subqueries during planning. Thus an update of the form "UPDATE t1 ... FROM t2 ..." where t1 has inheritance and t2 has RLS quals would fail. Fix by making sure to copy in and update the securityQuals when they exist for non-target relations. process_policies() was adding WCOs to non-target relations, which is unnecessary, and could lead to a lot of wasted time in the rewriter and the planner. Fix by only adding WCO policies when working on the result relation. Also in process_policies, we should be copying the USING policies to the WITH CHECK policies on a per-policy basis, fix by moving the copying up into the per-policy loop. Lastly, as noted by Dean, we were simply adding policies returned by the hook provided to the list of quals being AND'd, meaning that they would actually restrict records returned and there was no option to have internal policies and hook-based policies work together permissively (as all internal policies currently work). Instead, explicitly add support for both permissive and restrictive policies by having a hook for each and combining the results appropriately. To ensure this is all done correctly, add a new test module (test_rls_hooks) to test the various combinations of internal, permissive, and restrictive hook policies. Largely from Dean Rasheed (thanks!): CAEZATCVmFUfUOwwhnBTcgi6AquyjQ0-1fyKd0T3xBWJvn+xsFA@mail.gmail.com Author: Dean Rasheed, though I added the new hooks and test module.
2015-04-22 18:01:06 +02:00
CREATE TABLE b1 (a int, b text);
INSERT INTO b1 (SELECT x, md5(x::text) FROM generate_series(-10,10) x);
CREATE POLICY p1 ON b1 USING (a % 2 = 0);
ALTER TABLE b1 ENABLE ROW LEVEL SECURITY;
GRANT ALL ON b1 TO regress_rls_bob;
RLS fixes, new hooks, and new test module In prepend_row_security_policies(), defaultDeny was always true, so if there were any hook policies, the RLS policies on the table would just get discarded. Fixed to start off with defaultDeny as false and then properly set later if we detect that only the default deny policy exists for the internal policies. The infinite recursion detection in fireRIRrules() didn't properly manage the activeRIRs list in the case of WCOs, so it would incorrectly report infinite recusion if the same relation with RLS appeared more than once in the rtable, for example "UPDATE t ... FROM t ...". Further, the RLS expansion code in fireRIRrules() was handling RLS in the main loop through the rtable, which lead to RTEs being visited twice if they contained sublink subqueries, which prepend_row_security_policies() attempted to handle by exiting early if the RTE already had securityQuals. That doesn't work, however, since if the query involved a security barrier view on top of a table with RLS, the RTE would already have securityQuals (from the view) by the time fireRIRrules() was invoked, and so the table's RLS policies would be ignored. This is fixed in fireRIRrules() by handling RLS in a separate loop at the end, after dealing with any other sublink subqueries, thus ensuring that each RTE is only visited once for RLS expansion. The inheritance planner code didn't correctly handle non-target relations with RLS, which would get turned into subqueries during planning. Thus an update of the form "UPDATE t1 ... FROM t2 ..." where t1 has inheritance and t2 has RLS quals would fail. Fix by making sure to copy in and update the securityQuals when they exist for non-target relations. process_policies() was adding WCOs to non-target relations, which is unnecessary, and could lead to a lot of wasted time in the rewriter and the planner. Fix by only adding WCO policies when working on the result relation. Also in process_policies, we should be copying the USING policies to the WITH CHECK policies on a per-policy basis, fix by moving the copying up into the per-policy loop. Lastly, as noted by Dean, we were simply adding policies returned by the hook provided to the list of quals being AND'd, meaning that they would actually restrict records returned and there was no option to have internal policies and hook-based policies work together permissively (as all internal policies currently work). Instead, explicitly add support for both permissive and restrictive policies by having a hook for each and combining the results appropriately. To ensure this is all done correctly, add a new test module (test_rls_hooks) to test the various combinations of internal, permissive, and restrictive hook policies. Largely from Dean Rasheed (thanks!): CAEZATCVmFUfUOwwhnBTcgi6AquyjQ0-1fyKd0T3xBWJvn+xsFA@mail.gmail.com Author: Dean Rasheed, though I added the new hooks and test module.
2015-04-22 18:01:06 +02:00
SET SESSION AUTHORIZATION regress_rls_bob;
RLS fixes, new hooks, and new test module In prepend_row_security_policies(), defaultDeny was always true, so if there were any hook policies, the RLS policies on the table would just get discarded. Fixed to start off with defaultDeny as false and then properly set later if we detect that only the default deny policy exists for the internal policies. The infinite recursion detection in fireRIRrules() didn't properly manage the activeRIRs list in the case of WCOs, so it would incorrectly report infinite recusion if the same relation with RLS appeared more than once in the rtable, for example "UPDATE t ... FROM t ...". Further, the RLS expansion code in fireRIRrules() was handling RLS in the main loop through the rtable, which lead to RTEs being visited twice if they contained sublink subqueries, which prepend_row_security_policies() attempted to handle by exiting early if the RTE already had securityQuals. That doesn't work, however, since if the query involved a security barrier view on top of a table with RLS, the RTE would already have securityQuals (from the view) by the time fireRIRrules() was invoked, and so the table's RLS policies would be ignored. This is fixed in fireRIRrules() by handling RLS in a separate loop at the end, after dealing with any other sublink subqueries, thus ensuring that each RTE is only visited once for RLS expansion. The inheritance planner code didn't correctly handle non-target relations with RLS, which would get turned into subqueries during planning. Thus an update of the form "UPDATE t1 ... FROM t2 ..." where t1 has inheritance and t2 has RLS quals would fail. Fix by making sure to copy in and update the securityQuals when they exist for non-target relations. process_policies() was adding WCOs to non-target relations, which is unnecessary, and could lead to a lot of wasted time in the rewriter and the planner. Fix by only adding WCO policies when working on the result relation. Also in process_policies, we should be copying the USING policies to the WITH CHECK policies on a per-policy basis, fix by moving the copying up into the per-policy loop. Lastly, as noted by Dean, we were simply adding policies returned by the hook provided to the list of quals being AND'd, meaning that they would actually restrict records returned and there was no option to have internal policies and hook-based policies work together permissively (as all internal policies currently work). Instead, explicitly add support for both permissive and restrictive policies by having a hook for each and combining the results appropriately. To ensure this is all done correctly, add a new test module (test_rls_hooks) to test the various combinations of internal, permissive, and restrictive hook policies. Largely from Dean Rasheed (thanks!): CAEZATCVmFUfUOwwhnBTcgi6AquyjQ0-1fyKd0T3xBWJvn+xsFA@mail.gmail.com Author: Dean Rasheed, though I added the new hooks and test module.
2015-04-22 18:01:06 +02:00
CREATE VIEW bv1 WITH (security_barrier) AS SELECT * FROM b1 WHERE a > 0 WITH CHECK OPTION;
GRANT ALL ON bv1 TO regress_rls_carol;
RLS fixes, new hooks, and new test module In prepend_row_security_policies(), defaultDeny was always true, so if there were any hook policies, the RLS policies on the table would just get discarded. Fixed to start off with defaultDeny as false and then properly set later if we detect that only the default deny policy exists for the internal policies. The infinite recursion detection in fireRIRrules() didn't properly manage the activeRIRs list in the case of WCOs, so it would incorrectly report infinite recusion if the same relation with RLS appeared more than once in the rtable, for example "UPDATE t ... FROM t ...". Further, the RLS expansion code in fireRIRrules() was handling RLS in the main loop through the rtable, which lead to RTEs being visited twice if they contained sublink subqueries, which prepend_row_security_policies() attempted to handle by exiting early if the RTE already had securityQuals. That doesn't work, however, since if the query involved a security barrier view on top of a table with RLS, the RTE would already have securityQuals (from the view) by the time fireRIRrules() was invoked, and so the table's RLS policies would be ignored. This is fixed in fireRIRrules() by handling RLS in a separate loop at the end, after dealing with any other sublink subqueries, thus ensuring that each RTE is only visited once for RLS expansion. The inheritance planner code didn't correctly handle non-target relations with RLS, which would get turned into subqueries during planning. Thus an update of the form "UPDATE t1 ... FROM t2 ..." where t1 has inheritance and t2 has RLS quals would fail. Fix by making sure to copy in and update the securityQuals when they exist for non-target relations. process_policies() was adding WCOs to non-target relations, which is unnecessary, and could lead to a lot of wasted time in the rewriter and the planner. Fix by only adding WCO policies when working on the result relation. Also in process_policies, we should be copying the USING policies to the WITH CHECK policies on a per-policy basis, fix by moving the copying up into the per-policy loop. Lastly, as noted by Dean, we were simply adding policies returned by the hook provided to the list of quals being AND'd, meaning that they would actually restrict records returned and there was no option to have internal policies and hook-based policies work together permissively (as all internal policies currently work). Instead, explicitly add support for both permissive and restrictive policies by having a hook for each and combining the results appropriately. To ensure this is all done correctly, add a new test module (test_rls_hooks) to test the various combinations of internal, permissive, and restrictive hook policies. Largely from Dean Rasheed (thanks!): CAEZATCVmFUfUOwwhnBTcgi6AquyjQ0-1fyKd0T3xBWJvn+xsFA@mail.gmail.com Author: Dean Rasheed, though I added the new hooks and test module.
2015-04-22 18:01:06 +02:00
SET SESSION AUTHORIZATION regress_rls_carol;
RLS fixes, new hooks, and new test module In prepend_row_security_policies(), defaultDeny was always true, so if there were any hook policies, the RLS policies on the table would just get discarded. Fixed to start off with defaultDeny as false and then properly set later if we detect that only the default deny policy exists for the internal policies. The infinite recursion detection in fireRIRrules() didn't properly manage the activeRIRs list in the case of WCOs, so it would incorrectly report infinite recusion if the same relation with RLS appeared more than once in the rtable, for example "UPDATE t ... FROM t ...". Further, the RLS expansion code in fireRIRrules() was handling RLS in the main loop through the rtable, which lead to RTEs being visited twice if they contained sublink subqueries, which prepend_row_security_policies() attempted to handle by exiting early if the RTE already had securityQuals. That doesn't work, however, since if the query involved a security barrier view on top of a table with RLS, the RTE would already have securityQuals (from the view) by the time fireRIRrules() was invoked, and so the table's RLS policies would be ignored. This is fixed in fireRIRrules() by handling RLS in a separate loop at the end, after dealing with any other sublink subqueries, thus ensuring that each RTE is only visited once for RLS expansion. The inheritance planner code didn't correctly handle non-target relations with RLS, which would get turned into subqueries during planning. Thus an update of the form "UPDATE t1 ... FROM t2 ..." where t1 has inheritance and t2 has RLS quals would fail. Fix by making sure to copy in and update the securityQuals when they exist for non-target relations. process_policies() was adding WCOs to non-target relations, which is unnecessary, and could lead to a lot of wasted time in the rewriter and the planner. Fix by only adding WCO policies when working on the result relation. Also in process_policies, we should be copying the USING policies to the WITH CHECK policies on a per-policy basis, fix by moving the copying up into the per-policy loop. Lastly, as noted by Dean, we were simply adding policies returned by the hook provided to the list of quals being AND'd, meaning that they would actually restrict records returned and there was no option to have internal policies and hook-based policies work together permissively (as all internal policies currently work). Instead, explicitly add support for both permissive and restrictive policies by having a hook for each and combining the results appropriately. To ensure this is all done correctly, add a new test module (test_rls_hooks) to test the various combinations of internal, permissive, and restrictive hook policies. Largely from Dean Rasheed (thanks!): CAEZATCVmFUfUOwwhnBTcgi6AquyjQ0-1fyKd0T3xBWJvn+xsFA@mail.gmail.com Author: Dean Rasheed, though I added the new hooks and test module.
2015-04-22 18:01:06 +02:00
EXPLAIN (COSTS OFF) SELECT * FROM bv1 WHERE f_leak(b);
SELECT * FROM bv1 WHERE f_leak(b);
INSERT INTO bv1 VALUES (-1, 'xxx'); -- should fail view WCO
INSERT INTO bv1 VALUES (11, 'xxx'); -- should fail RLS check
INSERT INTO bv1 VALUES (12, 'xxx'); -- ok
EXPLAIN (COSTS OFF) UPDATE bv1 SET b = 'yyy' WHERE a = 4 AND f_leak(b);
UPDATE bv1 SET b = 'yyy' WHERE a = 4 AND f_leak(b);
EXPLAIN (COSTS OFF) DELETE FROM bv1 WHERE a = 6 AND f_leak(b);
DELETE FROM bv1 WHERE a = 6 AND f_leak(b);
SET SESSION AUTHORIZATION regress_rls_alice;
RLS fixes, new hooks, and new test module In prepend_row_security_policies(), defaultDeny was always true, so if there were any hook policies, the RLS policies on the table would just get discarded. Fixed to start off with defaultDeny as false and then properly set later if we detect that only the default deny policy exists for the internal policies. The infinite recursion detection in fireRIRrules() didn't properly manage the activeRIRs list in the case of WCOs, so it would incorrectly report infinite recusion if the same relation with RLS appeared more than once in the rtable, for example "UPDATE t ... FROM t ...". Further, the RLS expansion code in fireRIRrules() was handling RLS in the main loop through the rtable, which lead to RTEs being visited twice if they contained sublink subqueries, which prepend_row_security_policies() attempted to handle by exiting early if the RTE already had securityQuals. That doesn't work, however, since if the query involved a security barrier view on top of a table with RLS, the RTE would already have securityQuals (from the view) by the time fireRIRrules() was invoked, and so the table's RLS policies would be ignored. This is fixed in fireRIRrules() by handling RLS in a separate loop at the end, after dealing with any other sublink subqueries, thus ensuring that each RTE is only visited once for RLS expansion. The inheritance planner code didn't correctly handle non-target relations with RLS, which would get turned into subqueries during planning. Thus an update of the form "UPDATE t1 ... FROM t2 ..." where t1 has inheritance and t2 has RLS quals would fail. Fix by making sure to copy in and update the securityQuals when they exist for non-target relations. process_policies() was adding WCOs to non-target relations, which is unnecessary, and could lead to a lot of wasted time in the rewriter and the planner. Fix by only adding WCO policies when working on the result relation. Also in process_policies, we should be copying the USING policies to the WITH CHECK policies on a per-policy basis, fix by moving the copying up into the per-policy loop. Lastly, as noted by Dean, we were simply adding policies returned by the hook provided to the list of quals being AND'd, meaning that they would actually restrict records returned and there was no option to have internal policies and hook-based policies work together permissively (as all internal policies currently work). Instead, explicitly add support for both permissive and restrictive policies by having a hook for each and combining the results appropriately. To ensure this is all done correctly, add a new test module (test_rls_hooks) to test the various combinations of internal, permissive, and restrictive hook policies. Largely from Dean Rasheed (thanks!): CAEZATCVmFUfUOwwhnBTcgi6AquyjQ0-1fyKd0T3xBWJvn+xsFA@mail.gmail.com Author: Dean Rasheed, though I added the new hooks and test module.
2015-04-22 18:01:06 +02:00
SELECT * FROM b1;
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
--
-- INSERT ... ON CONFLICT DO UPDATE and Row-level security
--
SET SESSION AUTHORIZATION regress_rls_alice;
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
DROP POLICY p1 ON document;
DROP POLICY p1r ON document;
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
CREATE POLICY p1 ON document FOR SELECT USING (true);
CREATE POLICY p2 ON document FOR INSERT WITH CHECK (dauthor = current_user);
CREATE POLICY p3 ON document FOR UPDATE
USING (cid = (SELECT cid from category WHERE cname = 'novel'))
WITH CHECK (dauthor = current_user);
SET SESSION AUTHORIZATION regress_rls_bob;
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
-- Exists...
SELECT * FROM document WHERE did = 2;
-- ...so violates actual WITH CHECK OPTION within UPDATE (not INSERT, since
-- alternative UPDATE path happens to be taken):
INSERT INTO document VALUES (2, (SELECT cid from category WHERE cname = 'novel'), 1, 'regress_rls_carol', 'my first novel')
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
ON CONFLICT (did) DO UPDATE SET dtitle = EXCLUDED.dtitle, dauthor = EXCLUDED.dauthor;
-- Violates USING qual for UPDATE policy p3.
--
-- UPDATE path is taken, but UPDATE fails purely because *existing* row to be
-- updated is not a "novel"/cid 11 (row is not leaked, even though we have
-- SELECT privileges sufficient to see the row in this instance):
INSERT INTO document VALUES (33, 22, 1, 'regress_rls_bob', 'okay science fiction'); -- preparation for next statement
INSERT INTO document VALUES (33, (SELECT cid from category WHERE cname = 'novel'), 1, 'regress_rls_bob', 'Some novel, replaces sci-fi') -- takes UPDATE path
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
ON CONFLICT (did) DO UPDATE SET dtitle = EXCLUDED.dtitle;
-- Fine (we UPDATE, since INSERT WCOs and UPDATE security barrier quals + WCOs
-- not violated):
INSERT INTO document VALUES (2, (SELECT cid from category WHERE cname = 'novel'), 1, 'regress_rls_bob', 'my first novel')
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
ON CONFLICT (did) DO UPDATE SET dtitle = EXCLUDED.dtitle RETURNING *;
-- Fine (we INSERT, so "cid = 33" ("technology") isn't evaluated):
INSERT INTO document VALUES (78, (SELECT cid from category WHERE cname = 'novel'), 1, 'regress_rls_bob', 'some technology novel')
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
ON CONFLICT (did) DO UPDATE SET dtitle = EXCLUDED.dtitle, cid = 33 RETURNING *;
-- Fine (same query, but we UPDATE, so "cid = 33", ("technology") is not the
-- case in respect of *existing* tuple):
INSERT INTO document VALUES (78, (SELECT cid from category WHERE cname = 'novel'), 1, 'regress_rls_bob', 'some technology novel')
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
ON CONFLICT (did) DO UPDATE SET dtitle = EXCLUDED.dtitle, cid = 33 RETURNING *;
-- Same query a third time, but now fails due to existing tuple finally not
-- passing quals:
INSERT INTO document VALUES (78, (SELECT cid from category WHERE cname = 'novel'), 1, 'regress_rls_bob', 'some technology novel')
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
ON CONFLICT (did) DO UPDATE SET dtitle = EXCLUDED.dtitle, cid = 33 RETURNING *;
-- Don't fail just because INSERT doesn't satisfy WITH CHECK option that
-- originated as a barrier/USING() qual from the UPDATE. Note that the UPDATE
-- path *isn't* taken, and so UPDATE-related policy does not apply:
INSERT INTO document VALUES (79, (SELECT cid from category WHERE cname = 'technology'), 1, 'regress_rls_bob', 'technology book, can only insert')
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
ON CONFLICT (did) DO UPDATE SET dtitle = EXCLUDED.dtitle RETURNING *;
-- But this time, the same statement fails, because the UPDATE path is taken,
-- and updating the row just inserted falls afoul of security barrier qual
-- (enforced as WCO) -- what we might have updated target tuple to is
-- irrelevant, in fact.
INSERT INTO document VALUES (79, (SELECT cid from category WHERE cname = 'technology'), 1, 'regress_rls_bob', 'technology book, can only insert')
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
ON CONFLICT (did) DO UPDATE SET dtitle = EXCLUDED.dtitle RETURNING *;
-- Test default USING qual enforced as WCO
SET SESSION AUTHORIZATION regress_rls_alice;
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
DROP POLICY p1 ON document;
DROP POLICY p2 ON document;
DROP POLICY p3 ON document;
CREATE POLICY p3_with_default ON document FOR UPDATE
USING (cid = (SELECT cid from category WHERE cname = 'novel'));
SET SESSION AUTHORIZATION regress_rls_bob;
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
-- Just because WCO-style enforcement of USING quals occurs with
-- existing/target tuple does not mean that the implementation can be allowed
-- to fail to also enforce this qual against the final tuple appended to
-- relation (since in the absence of an explicit WCO, this is also interpreted
-- as an UPDATE/ALL WCO in general).
--
-- UPDATE path is taken here (fails due to existing tuple). Note that this is
-- not reported as a "USING expression", because it's an RLS UPDATE check that originated as
-- a USING qual for the purposes of RLS in general, as opposed to an explicit
-- USING qual that is ordinarily a security barrier. We leave it up to the
-- UPDATE to make this fail:
INSERT INTO document VALUES (79, (SELECT cid from category WHERE cname = 'technology'), 1, 'regress_rls_bob', 'technology book, can only insert')
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
ON CONFLICT (did) DO UPDATE SET dtitle = EXCLUDED.dtitle RETURNING *;
-- UPDATE path is taken here. Existing tuple passes, since its cid
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
-- corresponds to "novel", but default USING qual is enforced against
-- post-UPDATE tuple too (as always when updating with a policy that lacks an
-- explicit WCO), and so this fails:
INSERT INTO document VALUES (2, (SELECT cid from category WHERE cname = 'technology'), 1, 'regress_rls_bob', 'my first novel')
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
ON CONFLICT (did) DO UPDATE SET cid = EXCLUDED.cid, dtitle = EXCLUDED.dtitle RETURNING *;
SET SESSION AUTHORIZATION regress_rls_alice;
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
DROP POLICY p3_with_default ON document;
--
-- Test ALL policies with ON CONFLICT DO UPDATE (much the same as existing UPDATE
-- tests)
--
CREATE POLICY p3_with_all ON document FOR ALL
USING (cid = (SELECT cid from category WHERE cname = 'novel'))
WITH CHECK (dauthor = current_user);
SET SESSION AUTHORIZATION regress_rls_bob;
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
-- Fails, since ALL WCO is enforced in insert path:
INSERT INTO document VALUES (80, (SELECT cid from category WHERE cname = 'novel'), 1, 'regress_rls_carol', 'my first novel')
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
ON CONFLICT (did) DO UPDATE SET dtitle = EXCLUDED.dtitle, cid = 33;
-- Fails, since ALL policy USING qual is enforced (existing, target tuple is in
-- violation, since it has the "manga" cid):
INSERT INTO document VALUES (4, (SELECT cid from category WHERE cname = 'novel'), 1, 'regress_rls_bob', 'my first novel')
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
ON CONFLICT (did) DO UPDATE SET dtitle = EXCLUDED.dtitle;
-- Fails, since ALL WCO are enforced:
INSERT INTO document VALUES (1, (SELECT cid from category WHERE cname = 'novel'), 1, 'regress_rls_bob', 'my first novel')
ON CONFLICT (did) DO UPDATE SET dauthor = 'regress_rls_carol';
RLS fixes, new hooks, and new test module In prepend_row_security_policies(), defaultDeny was always true, so if there were any hook policies, the RLS policies on the table would just get discarded. Fixed to start off with defaultDeny as false and then properly set later if we detect that only the default deny policy exists for the internal policies. The infinite recursion detection in fireRIRrules() didn't properly manage the activeRIRs list in the case of WCOs, so it would incorrectly report infinite recusion if the same relation with RLS appeared more than once in the rtable, for example "UPDATE t ... FROM t ...". Further, the RLS expansion code in fireRIRrules() was handling RLS in the main loop through the rtable, which lead to RTEs being visited twice if they contained sublink subqueries, which prepend_row_security_policies() attempted to handle by exiting early if the RTE already had securityQuals. That doesn't work, however, since if the query involved a security barrier view on top of a table with RLS, the RTE would already have securityQuals (from the view) by the time fireRIRrules() was invoked, and so the table's RLS policies would be ignored. This is fixed in fireRIRrules() by handling RLS in a separate loop at the end, after dealing with any other sublink subqueries, thus ensuring that each RTE is only visited once for RLS expansion. The inheritance planner code didn't correctly handle non-target relations with RLS, which would get turned into subqueries during planning. Thus an update of the form "UPDATE t1 ... FROM t2 ..." where t1 has inheritance and t2 has RLS quals would fail. Fix by making sure to copy in and update the securityQuals when they exist for non-target relations. process_policies() was adding WCOs to non-target relations, which is unnecessary, and could lead to a lot of wasted time in the rewriter and the planner. Fix by only adding WCO policies when working on the result relation. Also in process_policies, we should be copying the USING policies to the WITH CHECK policies on a per-policy basis, fix by moving the copying up into the per-policy loop. Lastly, as noted by Dean, we were simply adding policies returned by the hook provided to the list of quals being AND'd, meaning that they would actually restrict records returned and there was no option to have internal policies and hook-based policies work together permissively (as all internal policies currently work). Instead, explicitly add support for both permissive and restrictive policies by having a hook for each and combining the results appropriately. To ensure this is all done correctly, add a new test module (test_rls_hooks) to test the various combinations of internal, permissive, and restrictive hook policies. Largely from Dean Rasheed (thanks!): CAEZATCVmFUfUOwwhnBTcgi6AquyjQ0-1fyKd0T3xBWJvn+xsFA@mail.gmail.com Author: Dean Rasheed, though I added the new hooks and test module.
2015-04-22 18:01:06 +02:00
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
--
-- ROLE/GROUP
--
SET SESSION AUTHORIZATION regress_rls_alice;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
CREATE TABLE z1 (a int, b text);
CREATE TABLE z2 (a int, b text);
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
GRANT SELECT ON z1,z2 TO regress_rls_group1, regress_rls_group2,
regress_rls_bob, regress_rls_carol;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
INSERT INTO z1 VALUES
(1, 'aba'),
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
(2, 'bbb'),
(3, 'ccc'),
(4, 'dad');
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
CREATE POLICY p1 ON z1 TO regress_rls_group1 USING (a % 2 = 0);
CREATE POLICY p2 ON z1 TO regress_rls_group2 USING (a % 2 = 1);
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
ALTER TABLE z1 ENABLE ROW LEVEL SECURITY;
SET SESSION AUTHORIZATION regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SELECT * FROM z1 WHERE f_leak(b);
EXPLAIN (COSTS OFF) SELECT * FROM z1 WHERE f_leak(b);
PREPARE plancache_test AS SELECT * FROM z1 WHERE f_leak(b);
EXPLAIN (COSTS OFF) EXECUTE plancache_test;
Allow user control of CTE materialization, and change the default behavior. Historically we've always materialized the full output of a CTE query, treating WITH as an optimization fence (so that, for example, restrictions from the outer query cannot be pushed into it). This is appropriate when the CTE query is INSERT/UPDATE/DELETE, or is recursive; but when the CTE query is non-recursive and side-effect-free, there's no hazard of changing the query results by pushing restrictions down. Another argument for materialization is that it can avoid duplicate computation of an expensive WITH query --- but that only applies if the WITH query is called more than once in the outer query. Even then it could still be a net loss, if each call has restrictions that would allow just a small part of the WITH query to be computed. Hence, let's change the behavior for WITH queries that are non-recursive and side-effect-free. By default, we will inline them into the outer query (removing the optimization fence) if they are called just once. If they are called more than once, we will keep the old behavior by default, but the user can override this and force inlining by specifying NOT MATERIALIZED. Lastly, the user can force the old behavior by specifying MATERIALIZED; this would mainly be useful when the query had deliberately been employing WITH as an optimization fence to prevent a poor choice of plan. Andreas Karlsson, Andrew Gierth, David Fetter Discussion: https://postgr.es/m/87sh48ffhb.fsf@news-spur.riddles.org.uk
2019-02-16 22:11:12 +01:00
PREPARE plancache_test2 AS WITH q AS MATERIALIZED (SELECT * FROM z1 WHERE f_leak(b)) SELECT * FROM q,z2;
EXPLAIN (COSTS OFF) EXECUTE plancache_test2;
Allow user control of CTE materialization, and change the default behavior. Historically we've always materialized the full output of a CTE query, treating WITH as an optimization fence (so that, for example, restrictions from the outer query cannot be pushed into it). This is appropriate when the CTE query is INSERT/UPDATE/DELETE, or is recursive; but when the CTE query is non-recursive and side-effect-free, there's no hazard of changing the query results by pushing restrictions down. Another argument for materialization is that it can avoid duplicate computation of an expensive WITH query --- but that only applies if the WITH query is called more than once in the outer query. Even then it could still be a net loss, if each call has restrictions that would allow just a small part of the WITH query to be computed. Hence, let's change the behavior for WITH queries that are non-recursive and side-effect-free. By default, we will inline them into the outer query (removing the optimization fence) if they are called just once. If they are called more than once, we will keep the old behavior by default, but the user can override this and force inlining by specifying NOT MATERIALIZED. Lastly, the user can force the old behavior by specifying MATERIALIZED; this would mainly be useful when the query had deliberately been employing WITH as an optimization fence to prevent a poor choice of plan. Andreas Karlsson, Andrew Gierth, David Fetter Discussion: https://postgr.es/m/87sh48ffhb.fsf@news-spur.riddles.org.uk
2019-02-16 22:11:12 +01:00
PREPARE plancache_test3 AS WITH q AS MATERIALIZED (SELECT * FROM z2) SELECT * FROM q,z1 WHERE f_leak(z1.b);
EXPLAIN (COSTS OFF) EXECUTE plancache_test3;
SET ROLE regress_rls_group1;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SELECT * FROM z1 WHERE f_leak(b);
EXPLAIN (COSTS OFF) SELECT * FROM z1 WHERE f_leak(b);
EXPLAIN (COSTS OFF) EXECUTE plancache_test;
EXPLAIN (COSTS OFF) EXECUTE plancache_test2;
EXPLAIN (COSTS OFF) EXECUTE plancache_test3;
SET SESSION AUTHORIZATION regress_rls_carol;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SELECT * FROM z1 WHERE f_leak(b);
EXPLAIN (COSTS OFF) SELECT * FROM z1 WHERE f_leak(b);
EXPLAIN (COSTS OFF) EXECUTE plancache_test;
EXPLAIN (COSTS OFF) EXECUTE plancache_test2;
EXPLAIN (COSTS OFF) EXECUTE plancache_test3;
SET ROLE regress_rls_group2;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SELECT * FROM z1 WHERE f_leak(b);
EXPLAIN (COSTS OFF) SELECT * FROM z1 WHERE f_leak(b);
EXPLAIN (COSTS OFF) EXECUTE plancache_test;
EXPLAIN (COSTS OFF) EXECUTE plancache_test2;
EXPLAIN (COSTS OFF) EXECUTE plancache_test3;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
--
-- Views should follow policy for view owner.
--
-- View and Table owner are the same.
SET SESSION AUTHORIZATION regress_rls_alice;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
CREATE VIEW rls_view AS SELECT * FROM z1 WHERE f_leak(b);
GRANT SELECT ON rls_view TO regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
-- Query as role that is not owner of view or table. Should return all records.
SET SESSION AUTHORIZATION regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SELECT * FROM rls_view;
EXPLAIN (COSTS OFF) SELECT * FROM rls_view;
-- Query as view/table owner. Should return all records.
SET SESSION AUTHORIZATION regress_rls_alice;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SELECT * FROM rls_view;
EXPLAIN (COSTS OFF) SELECT * FROM rls_view;
DROP VIEW rls_view;
-- View and Table owners are different.
SET SESSION AUTHORIZATION regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
CREATE VIEW rls_view AS SELECT * FROM z1 WHERE f_leak(b);
GRANT SELECT ON rls_view TO regress_rls_alice;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
-- Query as role that is not owner of view but is owner of table.
-- Should return records based on view owner policies.
SET SESSION AUTHORIZATION regress_rls_alice;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SELECT * FROM rls_view;
EXPLAIN (COSTS OFF) SELECT * FROM rls_view;
-- Query as role that is not owner of table but is owner of view.
-- Should return records based on view owner policies.
SET SESSION AUTHORIZATION regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SELECT * FROM rls_view;
EXPLAIN (COSTS OFF) SELECT * FROM rls_view;
-- Query as role that is not the owner of the table or view without permissions.
SET SESSION AUTHORIZATION regress_rls_carol;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SELECT * FROM rls_view; --fail - permission denied.
EXPLAIN (COSTS OFF) SELECT * FROM rls_view; --fail - permission denied.
-- Query as role that is not the owner of the table or view with permissions.
SET SESSION AUTHORIZATION regress_rls_bob;
GRANT SELECT ON rls_view TO regress_rls_carol;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SELECT * FROM rls_view;
EXPLAIN (COSTS OFF) SELECT * FROM rls_view;
SET SESSION AUTHORIZATION regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
DROP VIEW rls_view;
--
-- Command specific
--
SET SESSION AUTHORIZATION regress_rls_alice;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
CREATE TABLE x1 (a int, b text, c text);
GRANT ALL ON x1 TO PUBLIC;
INSERT INTO x1 VALUES
(1, 'abc', 'regress_rls_bob'),
(2, 'bcd', 'regress_rls_bob'),
(3, 'cde', 'regress_rls_carol'),
(4, 'def', 'regress_rls_carol'),
(5, 'efg', 'regress_rls_bob'),
(6, 'fgh', 'regress_rls_bob'),
(7, 'fgh', 'regress_rls_carol'),
(8, 'fgh', 'regress_rls_carol');
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
CREATE POLICY p0 ON x1 FOR ALL USING (c = current_user);
CREATE POLICY p1 ON x1 FOR SELECT USING (a % 2 = 0);
CREATE POLICY p2 ON x1 FOR INSERT WITH CHECK (a % 2 = 1);
CREATE POLICY p3 ON x1 FOR UPDATE USING (a % 2 = 0);
CREATE POLICY p4 ON x1 FOR DELETE USING (a < 8);
ALTER TABLE x1 ENABLE ROW LEVEL SECURITY;
SET SESSION AUTHORIZATION regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SELECT * FROM x1 WHERE f_leak(b) ORDER BY a ASC;
UPDATE x1 SET b = b || '_updt' WHERE f_leak(b) RETURNING *;
SET SESSION AUTHORIZATION regress_rls_carol;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SELECT * FROM x1 WHERE f_leak(b) ORDER BY a ASC;
UPDATE x1 SET b = b || '_updt' WHERE f_leak(b) RETURNING *;
DELETE FROM x1 WHERE f_leak(b) RETURNING *;
--
-- Duplicate Policy Names
--
SET SESSION AUTHORIZATION regress_rls_alice;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
CREATE TABLE y1 (a int, b text);
CREATE TABLE y2 (a int, b text);
GRANT ALL ON y1, y2 TO regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
CREATE POLICY p1 ON y1 FOR ALL USING (a % 2 = 0);
CREATE POLICY p2 ON y1 FOR SELECT USING (a > 2);
CREATE POLICY p1 ON y1 FOR SELECT USING (a % 2 = 1); --fail
CREATE POLICY p1 ON y2 FOR ALL USING (a % 2 = 0); --OK
ALTER TABLE y1 ENABLE ROW LEVEL SECURITY;
ALTER TABLE y2 ENABLE ROW LEVEL SECURITY;
--
-- Expression structure with SBV
--
-- Create view as table owner. RLS should NOT be applied.
SET SESSION AUTHORIZATION regress_rls_alice;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
CREATE VIEW rls_sbv WITH (security_barrier) AS
SELECT * FROM y1 WHERE f_leak(b);
EXPLAIN (COSTS OFF) SELECT * FROM rls_sbv WHERE (a = 1);
DROP VIEW rls_sbv;
-- Create view as role that does not own table. RLS should be applied.
SET SESSION AUTHORIZATION regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
CREATE VIEW rls_sbv WITH (security_barrier) AS
SELECT * FROM y1 WHERE f_leak(b);
EXPLAIN (COSTS OFF) SELECT * FROM rls_sbv WHERE (a = 1);
DROP VIEW rls_sbv;
--
-- Expression structure
--
SET SESSION AUTHORIZATION regress_rls_alice;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
INSERT INTO y2 (SELECT x, md5(x::text) FROM generate_series(0,20) x);
CREATE POLICY p2 ON y2 USING (a % 3 = 0);
CREATE POLICY p3 ON y2 USING (a % 4 = 0);
SET SESSION AUTHORIZATION regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SELECT * FROM y2 WHERE f_leak(b);
EXPLAIN (COSTS OFF) SELECT * FROM y2 WHERE f_leak(b);
--
-- Qual push-down of leaky functions, when not referring to table
--
SELECT * FROM y2 WHERE f_leak('abc');
EXPLAIN (COSTS OFF) SELECT * FROM y2 WHERE f_leak('abc');
CREATE TABLE test_qual_pushdown (
abc text
);
INSERT INTO test_qual_pushdown VALUES ('abc'),('def');
SELECT * FROM y2 JOIN test_qual_pushdown ON (b = abc) WHERE f_leak(abc);
EXPLAIN (COSTS OFF) SELECT * FROM y2 JOIN test_qual_pushdown ON (b = abc) WHERE f_leak(abc);
SELECT * FROM y2 JOIN test_qual_pushdown ON (b = abc) WHERE f_leak(b);
EXPLAIN (COSTS OFF) SELECT * FROM y2 JOIN test_qual_pushdown ON (b = abc) WHERE f_leak(b);
DROP TABLE test_qual_pushdown;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
--
-- Plancache invalidate on user change.
--
RESET SESSION AUTHORIZATION;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
DROP TABLE t1 CASCADE;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
CREATE TABLE t1 (a integer);
GRANT SELECT ON t1 TO regress_rls_bob, regress_rls_carol;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
CREATE POLICY p1 ON t1 TO regress_rls_bob USING ((a % 2) = 0);
CREATE POLICY p2 ON t1 TO regress_rls_carol USING ((a % 4) = 0);
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
ALTER TABLE t1 ENABLE ROW LEVEL SECURITY;
-- Prepare as regress_rls_bob
SET ROLE regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
PREPARE role_inval AS SELECT * FROM t1;
Reset plan->row_security_env and planUserId In the plancache, we check if the environment we planned the query under has changed in a way which requires us to re-plan, such as when the user for whom the plan was prepared changes and RLS is being used (and, therefore, there may be different policies to apply). Unfortunately, while those values were set and checked, they were not being reset when the query was re-planned and therefore, in cases where we change role, re-plan, and then change role again, we weren't re-planning again. This leads to potentially incorrect policies being applied in cases where role-specific policies are used and a given query is planned under one role and then executed under other roles, which could happen under security definer functions or when a common user and query is planned initially and then re-used across multiple SET ROLEs. Further, extensions which made use of CopyCachedPlan() may suffer from similar issues as the RLS-related fields were not properly copied as part of the plan and therefore RevalidateCachedQuery() would copy in the current settings without invalidating the query. Fix by using the same approach used for 'search_path', where we set the correct values in CompleteCachedPlan(), check them early on in RevalidateCachedQuery() and then properly reset them if re-planning. Also, copy through the values during CopyCachedPlan(). Pointed out by Ashutosh Bapat. Reviewed by Michael Paquier. Back-patch to 9.5 where RLS was introduced. Security: CVE-2016-2193
2016-03-28 15:03:20 +02:00
-- Check plan
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
EXPLAIN (COSTS OFF) EXECUTE role_inval;
-- Change to regress_rls_carol
SET ROLE regress_rls_carol;
Reset plan->row_security_env and planUserId In the plancache, we check if the environment we planned the query under has changed in a way which requires us to re-plan, such as when the user for whom the plan was prepared changes and RLS is being used (and, therefore, there may be different policies to apply). Unfortunately, while those values were set and checked, they were not being reset when the query was re-planned and therefore, in cases where we change role, re-plan, and then change role again, we weren't re-planning again. This leads to potentially incorrect policies being applied in cases where role-specific policies are used and a given query is planned under one role and then executed under other roles, which could happen under security definer functions or when a common user and query is planned initially and then re-used across multiple SET ROLEs. Further, extensions which made use of CopyCachedPlan() may suffer from similar issues as the RLS-related fields were not properly copied as part of the plan and therefore RevalidateCachedQuery() would copy in the current settings without invalidating the query. Fix by using the same approach used for 'search_path', where we set the correct values in CompleteCachedPlan(), check them early on in RevalidateCachedQuery() and then properly reset them if re-planning. Also, copy through the values during CopyCachedPlan(). Pointed out by Ashutosh Bapat. Reviewed by Michael Paquier. Back-patch to 9.5 where RLS was introduced. Security: CVE-2016-2193
2016-03-28 15:03:20 +02:00
-- Check plan- should be different
EXPLAIN (COSTS OFF) EXECUTE role_inval;
-- Change back to regress_rls_bob
SET ROLE regress_rls_bob;
Reset plan->row_security_env and planUserId In the plancache, we check if the environment we planned the query under has changed in a way which requires us to re-plan, such as when the user for whom the plan was prepared changes and RLS is being used (and, therefore, there may be different policies to apply). Unfortunately, while those values were set and checked, they were not being reset when the query was re-planned and therefore, in cases where we change role, re-plan, and then change role again, we weren't re-planning again. This leads to potentially incorrect policies being applied in cases where role-specific policies are used and a given query is planned under one role and then executed under other roles, which could happen under security definer functions or when a common user and query is planned initially and then re-used across multiple SET ROLEs. Further, extensions which made use of CopyCachedPlan() may suffer from similar issues as the RLS-related fields were not properly copied as part of the plan and therefore RevalidateCachedQuery() would copy in the current settings without invalidating the query. Fix by using the same approach used for 'search_path', where we set the correct values in CompleteCachedPlan(), check them early on in RevalidateCachedQuery() and then properly reset them if re-planning. Also, copy through the values during CopyCachedPlan(). Pointed out by Ashutosh Bapat. Reviewed by Michael Paquier. Back-patch to 9.5 where RLS was introduced. Security: CVE-2016-2193
2016-03-28 15:03:20 +02:00
-- Check plan- should be back to original
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
EXPLAIN (COSTS OFF) EXECUTE role_inval;
--
-- CTE and RLS
--
RESET SESSION AUTHORIZATION;
DROP TABLE t1 CASCADE;
CREATE TABLE t1 (a integer, b text);
CREATE POLICY p1 ON t1 USING (a % 2 = 0);
ALTER TABLE t1 ENABLE ROW LEVEL SECURITY;
GRANT ALL ON t1 TO regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
INSERT INTO t1 (SELECT x, md5(x::text) FROM generate_series(0,20) x);
SET SESSION AUTHORIZATION regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
Allow user control of CTE materialization, and change the default behavior. Historically we've always materialized the full output of a CTE query, treating WITH as an optimization fence (so that, for example, restrictions from the outer query cannot be pushed into it). This is appropriate when the CTE query is INSERT/UPDATE/DELETE, or is recursive; but when the CTE query is non-recursive and side-effect-free, there's no hazard of changing the query results by pushing restrictions down. Another argument for materialization is that it can avoid duplicate computation of an expensive WITH query --- but that only applies if the WITH query is called more than once in the outer query. Even then it could still be a net loss, if each call has restrictions that would allow just a small part of the WITH query to be computed. Hence, let's change the behavior for WITH queries that are non-recursive and side-effect-free. By default, we will inline them into the outer query (removing the optimization fence) if they are called just once. If they are called more than once, we will keep the old behavior by default, but the user can override this and force inlining by specifying NOT MATERIALIZED. Lastly, the user can force the old behavior by specifying MATERIALIZED; this would mainly be useful when the query had deliberately been employing WITH as an optimization fence to prevent a poor choice of plan. Andreas Karlsson, Andrew Gierth, David Fetter Discussion: https://postgr.es/m/87sh48ffhb.fsf@news-spur.riddles.org.uk
2019-02-16 22:11:12 +01:00
WITH cte1 AS MATERIALIZED (SELECT * FROM t1 WHERE f_leak(b)) SELECT * FROM cte1;
EXPLAIN (COSTS OFF)
WITH cte1 AS MATERIALIZED (SELECT * FROM t1 WHERE f_leak(b)) SELECT * FROM cte1;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
WITH cte1 AS (UPDATE t1 SET a = a + 1 RETURNING *) SELECT * FROM cte1; --fail
WITH cte1 AS (UPDATE t1 SET a = a RETURNING *) SELECT * FROM cte1; --ok
WITH cte1 AS (INSERT INTO t1 VALUES (21, 'Fail') RETURNING *) SELECT * FROM cte1; --fail
WITH cte1 AS (INSERT INTO t1 VALUES (20, 'Success') RETURNING *) SELECT * FROM cte1; --ok
--
-- Rename Policy
--
RESET SESSION AUTHORIZATION;
ALTER POLICY p1 ON t1 RENAME TO p1; --fail
SELECT polname, relname
FROM pg_policy pol
JOIN pg_class pc ON (pc.oid = pol.polrelid)
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
WHERE relname = 't1';
ALTER POLICY p1 ON t1 RENAME TO p2; --ok
SELECT polname, relname
FROM pg_policy pol
JOIN pg_class pc ON (pc.oid = pol.polrelid)
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
WHERE relname = 't1';
--
-- Check INSERT SELECT
--
SET SESSION AUTHORIZATION regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
CREATE TABLE t2 (a integer, b text);
INSERT INTO t2 (SELECT * FROM t1);
EXPLAIN (COSTS OFF) INSERT INTO t2 (SELECT * FROM t1);
SELECT * FROM t2;
EXPLAIN (COSTS OFF) SELECT * FROM t2;
CREATE TABLE t3 AS SELECT * FROM t1;
SELECT * FROM t3;
SELECT * INTO t4 FROM t1;
SELECT * FROM t4;
--
-- RLS with JOIN
--
SET SESSION AUTHORIZATION regress_rls_alice;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
CREATE TABLE blog (id integer, author text, post text);
CREATE TABLE comment (blog_id integer, message text);
GRANT ALL ON blog, comment TO regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
CREATE POLICY blog_1 ON blog USING (id % 2 = 0);
ALTER TABLE blog ENABLE ROW LEVEL SECURITY;
INSERT INTO blog VALUES
(1, 'alice', 'blog #1'),
(2, 'bob', 'blog #1'),
(3, 'alice', 'blog #2'),
(4, 'alice', 'blog #3'),
(5, 'john', 'blog #1');
INSERT INTO comment VALUES
(1, 'cool blog'),
(1, 'fun blog'),
(3, 'crazy blog'),
(5, 'what?'),
(4, 'insane!'),
(2, 'who did it?');
SET SESSION AUTHORIZATION regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
-- Check RLS JOIN with Non-RLS.
SELECT id, author, message FROM blog JOIN comment ON id = blog_id;
-- Check Non-RLS JOIN with RLS.
SELECT id, author, message FROM comment JOIN blog ON id = blog_id;
SET SESSION AUTHORIZATION regress_rls_alice;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
CREATE POLICY comment_1 ON comment USING (blog_id < 4);
ALTER TABLE comment ENABLE ROW LEVEL SECURITY;
SET SESSION AUTHORIZATION regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
-- Check RLS JOIN RLS
SELECT id, author, message FROM blog JOIN comment ON id = blog_id;
SELECT id, author, message FROM comment JOIN blog ON id = blog_id;
SET SESSION AUTHORIZATION regress_rls_alice;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
DROP TABLE blog, comment;
--
-- Default Deny Policy
--
RESET SESSION AUTHORIZATION;
DROP POLICY p2 ON t1;
ALTER TABLE t1 OWNER TO regress_rls_alice;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
-- Check that default deny does not apply to superuser.
RESET SESSION AUTHORIZATION;
SELECT * FROM t1;
EXPLAIN (COSTS OFF) SELECT * FROM t1;
-- Check that default deny does not apply to table owner.
SET SESSION AUTHORIZATION regress_rls_alice;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SELECT * FROM t1;
EXPLAIN (COSTS OFF) SELECT * FROM t1;
-- Check that default deny applies to non-owner/non-superuser when RLS on.
SET SESSION AUTHORIZATION regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SET row_security TO ON;
SELECT * FROM t1;
EXPLAIN (COSTS OFF) SELECT * FROM t1;
SET SESSION AUTHORIZATION regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SELECT * FROM t1;
EXPLAIN (COSTS OFF) SELECT * FROM t1;
--
-- COPY TO/FROM
--
RESET SESSION AUTHORIZATION;
DROP TABLE copy_t CASCADE;
CREATE TABLE copy_t (a integer, b text);
CREATE POLICY p1 ON copy_t USING (a % 2 = 0);
ALTER TABLE copy_t ENABLE ROW LEVEL SECURITY;
GRANT ALL ON copy_t TO regress_rls_bob, regress_rls_exempt_user;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
INSERT INTO copy_t (SELECT x, md5(x::text) FROM generate_series(0,10) x);
-- Check COPY TO as Superuser/owner.
RESET SESSION AUTHORIZATION;
SET row_security TO OFF;
COPY (SELECT * FROM copy_t ORDER BY a ASC) TO STDOUT WITH DELIMITER ',';
SET row_security TO ON;
COPY (SELECT * FROM copy_t ORDER BY a ASC) TO STDOUT WITH DELIMITER ',';
-- Check COPY TO as user with permissions.
SET SESSION AUTHORIZATION regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SET row_security TO OFF;
COPY (SELECT * FROM copy_t ORDER BY a ASC) TO STDOUT WITH DELIMITER ','; --fail - would be affected by RLS
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SET row_security TO ON;
COPY (SELECT * FROM copy_t ORDER BY a ASC) TO STDOUT WITH DELIMITER ','; --ok
-- Check COPY TO as user with permissions and BYPASSRLS
SET SESSION AUTHORIZATION regress_rls_exempt_user;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SET row_security TO OFF;
COPY (SELECT * FROM copy_t ORDER BY a ASC) TO STDOUT WITH DELIMITER ','; --ok
SET row_security TO ON;
COPY (SELECT * FROM copy_t ORDER BY a ASC) TO STDOUT WITH DELIMITER ','; --ok
-- Check COPY TO as user without permissions. SET row_security TO OFF;
SET SESSION AUTHORIZATION regress_rls_carol;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SET row_security TO OFF;
COPY (SELECT * FROM copy_t ORDER BY a ASC) TO STDOUT WITH DELIMITER ','; --fail - would be affected by RLS
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SET row_security TO ON;
COPY (SELECT * FROM copy_t ORDER BY a ASC) TO STDOUT WITH DELIMITER ','; --fail - permission denied
-- Check COPY relation TO; keep it just one row to avoid reordering issues
RESET SESSION AUTHORIZATION;
SET row_security TO ON;
CREATE TABLE copy_rel_to (a integer, b text);
CREATE POLICY p1 ON copy_rel_to USING (a % 2 = 0);
ALTER TABLE copy_rel_to ENABLE ROW LEVEL SECURITY;
GRANT ALL ON copy_rel_to TO regress_rls_bob, regress_rls_exempt_user;
INSERT INTO copy_rel_to VALUES (1, md5('1'));
-- Check COPY TO as Superuser/owner.
RESET SESSION AUTHORIZATION;
SET row_security TO OFF;
COPY copy_rel_to TO STDOUT WITH DELIMITER ',';
SET row_security TO ON;
COPY copy_rel_to TO STDOUT WITH DELIMITER ',';
-- Check COPY TO as user with permissions.
SET SESSION AUTHORIZATION regress_rls_bob;
SET row_security TO OFF;
COPY copy_rel_to TO STDOUT WITH DELIMITER ','; --fail - would be affected by RLS
SET row_security TO ON;
COPY copy_rel_to TO STDOUT WITH DELIMITER ','; --ok
-- Check COPY TO as user with permissions and BYPASSRLS
SET SESSION AUTHORIZATION regress_rls_exempt_user;
SET row_security TO OFF;
COPY copy_rel_to TO STDOUT WITH DELIMITER ','; --ok
SET row_security TO ON;
COPY copy_rel_to TO STDOUT WITH DELIMITER ','; --ok
-- Check COPY TO as user without permissions. SET row_security TO OFF;
SET SESSION AUTHORIZATION regress_rls_carol;
SET row_security TO OFF;
COPY copy_rel_to TO STDOUT WITH DELIMITER ','; --fail - permission denied
SET row_security TO ON;
COPY copy_rel_to TO STDOUT WITH DELIMITER ','; --fail - permission denied
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
-- Check COPY FROM as Superuser/owner.
RESET SESSION AUTHORIZATION;
SET row_security TO OFF;
COPY copy_t FROM STDIN; --ok
1 abc
2 bcd
3 cde
4 def
\.
SET row_security TO ON;
COPY copy_t FROM STDIN; --ok
1 abc
2 bcd
3 cde
4 def
\.
-- Check COPY FROM as user with permissions.
SET SESSION AUTHORIZATION regress_rls_bob;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SET row_security TO OFF;
COPY copy_t FROM STDIN; --fail - would be affected by RLS.
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SET row_security TO ON;
COPY copy_t FROM STDIN; --fail - COPY FROM not supported by RLS.
-- Check COPY FROM as user with permissions and BYPASSRLS
SET SESSION AUTHORIZATION regress_rls_exempt_user;
SET row_security TO ON;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
COPY copy_t FROM STDIN; --ok
1 abc
2 bcd
3 cde
4 def
\.
-- Check COPY FROM as user without permissions.
SET SESSION AUTHORIZATION regress_rls_carol;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
SET row_security TO OFF;
COPY copy_t FROM STDIN; --fail - permission denied.
SET row_security TO ON;
COPY copy_t FROM STDIN; --fail - permission denied.
RESET SESSION AUTHORIZATION;
DROP TABLE copy_t;
DROP TABLE copy_rel_to CASCADE;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
-- Check WHERE CURRENT OF
SET SESSION AUTHORIZATION regress_rls_alice;
CREATE TABLE current_check (currentid int, payload text, rlsuser text);
GRANT ALL ON current_check TO PUBLIC;
INSERT INTO current_check VALUES
(1, 'abc', 'regress_rls_bob'),
(2, 'bcd', 'regress_rls_bob'),
(3, 'cde', 'regress_rls_bob'),
(4, 'def', 'regress_rls_bob');
CREATE POLICY p1 ON current_check FOR SELECT USING (currentid % 2 = 0);
CREATE POLICY p2 ON current_check FOR DELETE USING (currentid = 4 AND rlsuser = current_user);
CREATE POLICY p3 ON current_check FOR UPDATE USING (currentid = 4) WITH CHECK (rlsuser = current_user);
ALTER TABLE current_check ENABLE ROW LEVEL SECURITY;
SET SESSION AUTHORIZATION regress_rls_bob;
-- Can SELECT even rows
SELECT * FROM current_check;
-- Cannot UPDATE row 2
UPDATE current_check SET payload = payload || '_new' WHERE currentid = 2 RETURNING *;
BEGIN;
DECLARE current_check_cursor SCROLL CURSOR FOR SELECT * FROM current_check;
-- Returns rows that can be seen according to SELECT policy, like plain SELECT
-- above (even rows)
FETCH ABSOLUTE 1 FROM current_check_cursor;
-- Still cannot UPDATE row 2 through cursor
UPDATE current_check SET payload = payload || '_new' WHERE CURRENT OF current_check_cursor RETURNING *;
-- Can update row 4 through cursor, which is the next visible row
FETCH RELATIVE 1 FROM current_check_cursor;
UPDATE current_check SET payload = payload || '_new' WHERE CURRENT OF current_check_cursor RETURNING *;
SELECT * FROM current_check;
-- Plan should be a subquery TID scan
EXPLAIN (COSTS OFF) UPDATE current_check SET payload = payload WHERE CURRENT OF current_check_cursor;
-- Similarly can only delete row 4
FETCH ABSOLUTE 1 FROM current_check_cursor;
DELETE FROM current_check WHERE CURRENT OF current_check_cursor RETURNING *;
FETCH RELATIVE 1 FROM current_check_cursor;
DELETE FROM current_check WHERE CURRENT OF current_check_cursor RETURNING *;
SELECT * FROM current_check;
COMMIT;
--
-- check pg_stats view filtering
--
SET row_security TO ON;
SET SESSION AUTHORIZATION regress_rls_alice;
ANALYZE current_check;
-- Stats visible
SELECT row_security_active('current_check');
SELECT attname, most_common_vals FROM pg_stats
WHERE tablename = 'current_check'
ORDER BY 1;
SET SESSION AUTHORIZATION regress_rls_bob;
-- Stats not visible
SELECT row_security_active('current_check');
SELECT attname, most_common_vals FROM pg_stats
WHERE tablename = 'current_check'
ORDER BY 1;
--
-- Collation support
--
BEGIN;
CREATE TABLE coll_t (c) AS VALUES ('bar'::text);
CREATE POLICY coll_p ON coll_t USING (c < ('foo'::text COLLATE "C"));
ALTER TABLE coll_t ENABLE ROW LEVEL SECURITY;
GRANT SELECT ON coll_t TO regress_rls_alice;
SELECT (string_to_array(polqual, ':'))[7] AS inputcollid FROM pg_policy WHERE polrelid = 'coll_t'::regclass;
SET SESSION AUTHORIZATION regress_rls_alice;
SELECT * FROM coll_t;
ROLLBACK;
--
-- Shared Object Dependencies
--
RESET SESSION AUTHORIZATION;
BEGIN;
CREATE ROLE regress_rls_eve;
CREATE ROLE regress_rls_frank;
CREATE TABLE tbl1 (c) AS VALUES ('bar'::text);
GRANT SELECT ON TABLE tbl1 TO regress_rls_eve;
CREATE POLICY P ON tbl1 TO regress_rls_eve, regress_rls_frank USING (true);
SELECT refclassid::regclass, deptype
FROM pg_depend
WHERE classid = 'pg_policy'::regclass
AND refobjid = 'tbl1'::regclass;
SELECT refclassid::regclass, deptype
FROM pg_shdepend
WHERE classid = 'pg_policy'::regclass
AND refobjid IN ('regress_rls_eve'::regrole, 'regress_rls_frank'::regrole);
SAVEPOINT q;
DROP ROLE regress_rls_eve; --fails due to dependency on POLICY p
ROLLBACK TO q;
ALTER POLICY p ON tbl1 TO regress_rls_frank USING (true);
SAVEPOINT q;
DROP ROLE regress_rls_eve; --fails due to dependency on GRANT SELECT
ROLLBACK TO q;
REVOKE ALL ON TABLE tbl1 FROM regress_rls_eve;
SAVEPOINT q;
DROP ROLE regress_rls_eve; --succeeds
ROLLBACK TO q;
SAVEPOINT q;
DROP ROLE regress_rls_frank; --fails due to dependency on POLICY p
ROLLBACK TO q;
DROP POLICY p ON tbl1;
SAVEPOINT q;
DROP ROLE regress_rls_frank; -- succeeds
ROLLBACK TO q;
ROLLBACK; -- cleanup
--
-- Converting table to view
--
BEGIN;
CREATE TABLE t (c int);
CREATE POLICY p ON t USING (c % 2 = 1);
ALTER TABLE t ENABLE ROW LEVEL SECURITY;
SAVEPOINT q;
CREATE RULE "_RETURN" AS ON SELECT TO t DO INSTEAD
SELECT * FROM generate_series(1,5) t0(c); -- fails due to row level security enabled
ROLLBACK TO q;
ALTER TABLE t DISABLE ROW LEVEL SECURITY;
SAVEPOINT q;
CREATE RULE "_RETURN" AS ON SELECT TO t DO INSTEAD
SELECT * FROM generate_series(1,5) t0(c); -- fails due to policy p on t
ROLLBACK TO q;
DROP POLICY p ON t;
CREATE RULE "_RETURN" AS ON SELECT TO t DO INSTEAD
SELECT * FROM generate_series(1,5) t0(c); -- succeeds
ROLLBACK;
--
-- Policy expression handling
--
BEGIN;
CREATE TABLE t (c) AS VALUES ('bar'::text);
CREATE POLICY p ON t USING (max(c)); -- fails: aggregate functions are not allowed in policy expressions
ROLLBACK;
--
-- Non-target relations are only subject to SELECT policies
--
SET SESSION AUTHORIZATION regress_rls_alice;
CREATE TABLE r1 (a int);
CREATE TABLE r2 (a int);
INSERT INTO r1 VALUES (10), (20);
INSERT INTO r2 VALUES (10), (20);
GRANT ALL ON r1, r2 TO regress_rls_bob;
CREATE POLICY p1 ON r1 USING (true);
ALTER TABLE r1 ENABLE ROW LEVEL SECURITY;
CREATE POLICY p1 ON r2 FOR SELECT USING (true);
CREATE POLICY p2 ON r2 FOR INSERT WITH CHECK (false);
CREATE POLICY p3 ON r2 FOR UPDATE USING (false);
CREATE POLICY p4 ON r2 FOR DELETE USING (false);
ALTER TABLE r2 ENABLE ROW LEVEL SECURITY;
SET SESSION AUTHORIZATION regress_rls_bob;
SELECT * FROM r1;
SELECT * FROM r2;
-- r2 is read-only
INSERT INTO r2 VALUES (2); -- Not allowed
UPDATE r2 SET a = 2 RETURNING *; -- Updates nothing
DELETE FROM r2 RETURNING *; -- Deletes nothing
-- r2 can be used as a non-target relation in DML
INSERT INTO r1 SELECT a + 1 FROM r2 RETURNING *; -- OK
UPDATE r1 SET a = r2.a + 2 FROM r2 WHERE r1.a = r2.a RETURNING *; -- OK
DELETE FROM r1 USING r2 WHERE r1.a = r2.a + 2 RETURNING *; -- OK
SELECT * FROM r1;
SELECT * FROM r2;
SET SESSION AUTHORIZATION regress_rls_alice;
DROP TABLE r1;
DROP TABLE r2;
--
-- FORCE ROW LEVEL SECURITY applies RLS to owners too
--
SET SESSION AUTHORIZATION regress_rls_alice;
SET row_security = on;
CREATE TABLE r1 (a int);
INSERT INTO r1 VALUES (10), (20);
CREATE POLICY p1 ON r1 USING (false);
ALTER TABLE r1 ENABLE ROW LEVEL SECURITY;
ALTER TABLE r1 FORCE ROW LEVEL SECURITY;
-- No error, but no rows
TABLE r1;
-- RLS error
INSERT INTO r1 VALUES (1);
-- No error (unable to see any rows to update)
UPDATE r1 SET a = 1;
TABLE r1;
-- No error (unable to see any rows to delete)
DELETE FROM r1;
TABLE r1;
SET row_security = off;
-- these all fail, would be affected by RLS
TABLE r1;
UPDATE r1 SET a = 1;
DELETE FROM r1;
DROP TABLE r1;
--
-- FORCE ROW LEVEL SECURITY does not break RI
--
SET SESSION AUTHORIZATION regress_rls_alice;
SET row_security = on;
CREATE TABLE r1 (a int PRIMARY KEY);
CREATE TABLE r2 (a int REFERENCES r1);
INSERT INTO r1 VALUES (10), (20);
INSERT INTO r2 VALUES (10), (20);
-- Create policies on r2 which prevent the
-- owner from seeing any rows, but RI should
-- still see them.
CREATE POLICY p1 ON r2 USING (false);
ALTER TABLE r2 ENABLE ROW LEVEL SECURITY;
ALTER TABLE r2 FORCE ROW LEVEL SECURITY;
-- Errors due to rows in r2
DELETE FROM r1;
-- Reset r2 to no-RLS
DROP POLICY p1 ON r2;
ALTER TABLE r2 NO FORCE ROW LEVEL SECURITY;
ALTER TABLE r2 DISABLE ROW LEVEL SECURITY;
-- clean out r2 for INSERT test below
DELETE FROM r2;
-- Change r1 to not allow rows to be seen
CREATE POLICY p1 ON r1 USING (false);
ALTER TABLE r1 ENABLE ROW LEVEL SECURITY;
ALTER TABLE r1 FORCE ROW LEVEL SECURITY;
-- No rows seen
TABLE r1;
-- No error, RI still sees that row exists in r1
INSERT INTO r2 VALUES (10);
DROP TABLE r2;
DROP TABLE r1;
-- Ensure cascaded DELETE works
CREATE TABLE r1 (a int PRIMARY KEY);
CREATE TABLE r2 (a int REFERENCES r1 ON DELETE CASCADE);
INSERT INTO r1 VALUES (10), (20);
INSERT INTO r2 VALUES (10), (20);
-- Create policies on r2 which prevent the
-- owner from seeing any rows, but RI should
-- still see them.
CREATE POLICY p1 ON r2 USING (false);
ALTER TABLE r2 ENABLE ROW LEVEL SECURITY;
ALTER TABLE r2 FORCE ROW LEVEL SECURITY;
-- Deletes all records from both
DELETE FROM r1;
-- Remove FORCE from r2
ALTER TABLE r2 NO FORCE ROW LEVEL SECURITY;
-- As owner, we now bypass RLS
-- verify no rows in r2 now
TABLE r2;
DROP TABLE r2;
DROP TABLE r1;
-- Ensure cascaded UPDATE works
CREATE TABLE r1 (a int PRIMARY KEY);
CREATE TABLE r2 (a int REFERENCES r1 ON UPDATE CASCADE);
INSERT INTO r1 VALUES (10), (20);
INSERT INTO r2 VALUES (10), (20);
-- Create policies on r2 which prevent the
-- owner from seeing any rows, but RI should
-- still see them.
CREATE POLICY p1 ON r2 USING (false);
ALTER TABLE r2 ENABLE ROW LEVEL SECURITY;
ALTER TABLE r2 FORCE ROW LEVEL SECURITY;
-- Updates records in both
UPDATE r1 SET a = a+5;
-- Remove FORCE from r2
ALTER TABLE r2 NO FORCE ROW LEVEL SECURITY;
-- As owner, we now bypass RLS
-- verify records in r2 updated
TABLE r2;
DROP TABLE r2;
DROP TABLE r1;
--
-- Test INSERT+RETURNING applies SELECT policies as
-- WithCheckOptions (meaning an error is thrown)
--
SET SESSION AUTHORIZATION regress_rls_alice;
SET row_security = on;
CREATE TABLE r1 (a int);
CREATE POLICY p1 ON r1 FOR SELECT USING (false);
CREATE POLICY p2 ON r1 FOR INSERT WITH CHECK (true);
ALTER TABLE r1 ENABLE ROW LEVEL SECURITY;
ALTER TABLE r1 FORCE ROW LEVEL SECURITY;
-- Works fine
INSERT INTO r1 VALUES (10), (20);
-- No error, but no rows
TABLE r1;
SET row_security = off;
-- fail, would be affected by RLS
TABLE r1;
SET row_security = on;
-- Error
INSERT INTO r1 VALUES (10), (20) RETURNING *;
DROP TABLE r1;
--
-- Test UPDATE+RETURNING applies SELECT policies as
-- WithCheckOptions (meaning an error is thrown)
--
SET SESSION AUTHORIZATION regress_rls_alice;
SET row_security = on;
CREATE TABLE r1 (a int PRIMARY KEY);
CREATE POLICY p1 ON r1 FOR SELECT USING (a < 20);
CREATE POLICY p2 ON r1 FOR UPDATE USING (a < 20) WITH CHECK (true);
CREATE POLICY p3 ON r1 FOR INSERT WITH CHECK (true);
INSERT INTO r1 VALUES (10);
ALTER TABLE r1 ENABLE ROW LEVEL SECURITY;
ALTER TABLE r1 FORCE ROW LEVEL SECURITY;
-- Works fine
UPDATE r1 SET a = 30;
-- Show updated rows
ALTER TABLE r1 NO FORCE ROW LEVEL SECURITY;
TABLE r1;
-- reset value in r1 for test with RETURNING
UPDATE r1 SET a = 10;
-- Verify row reset
TABLE r1;
ALTER TABLE r1 FORCE ROW LEVEL SECURITY;
-- Error
UPDATE r1 SET a = 30 RETURNING *;
-- UPDATE path of INSERT ... ON CONFLICT DO UPDATE should also error out
INSERT INTO r1 VALUES (10)
ON CONFLICT (a) DO UPDATE SET a = 30 RETURNING *;
-- Should still error out without RETURNING (use of arbiter always requires
-- SELECT permissions)
INSERT INTO r1 VALUES (10)
ON CONFLICT (a) DO UPDATE SET a = 30;
INSERT INTO r1 VALUES (10)
ON CONFLICT ON CONSTRAINT r1_pkey DO UPDATE SET a = 30;
DROP TABLE r1;
-- Check dependency handling
RESET SESSION AUTHORIZATION;
CREATE TABLE dep1 (c1 int);
CREATE TABLE dep2 (c1 int);
CREATE POLICY dep_p1 ON dep1 TO regress_rls_bob USING (c1 > (select max(dep2.c1) from dep2));
ALTER POLICY dep_p1 ON dep1 TO regress_rls_bob,regress_rls_carol;
-- Should return one
SELECT count(*) = 1 FROM pg_depend
WHERE objid = (SELECT oid FROM pg_policy WHERE polname = 'dep_p1')
AND refobjid = (SELECT oid FROM pg_class WHERE relname = 'dep2');
ALTER POLICY dep_p1 ON dep1 USING (true);
-- Should return one
SELECT count(*) = 1 FROM pg_shdepend
WHERE objid = (SELECT oid FROM pg_policy WHERE polname = 'dep_p1')
AND refobjid = (SELECT oid FROM pg_authid WHERE rolname = 'regress_rls_bob');
-- Should return one
SELECT count(*) = 1 FROM pg_shdepend
WHERE objid = (SELECT oid FROM pg_policy WHERE polname = 'dep_p1')
AND refobjid = (SELECT oid FROM pg_authid WHERE rolname = 'regress_rls_carol');
-- Should return zero
SELECT count(*) = 0 FROM pg_depend
WHERE objid = (SELECT oid FROM pg_policy WHERE polname = 'dep_p1')
AND refobjid = (SELECT oid FROM pg_class WHERE relname = 'dep2');
-- DROP OWNED BY testing
RESET SESSION AUTHORIZATION;
CREATE ROLE regress_rls_dob_role1;
CREATE ROLE regress_rls_dob_role2;
CREATE TABLE dob_t1 (c1 int);
CREATE TABLE dob_t2 (c1 int) PARTITION BY RANGE (c1);
CREATE POLICY p1 ON dob_t1 TO regress_rls_dob_role1 USING (true);
DROP OWNED BY regress_rls_dob_role1;
DROP POLICY p1 ON dob_t1; -- should fail, already gone
CREATE POLICY p1 ON dob_t1 TO regress_rls_dob_role1,regress_rls_dob_role2 USING (true);
DROP OWNED BY regress_rls_dob_role1;
DROP POLICY p1 ON dob_t1; -- should succeed
CREATE POLICY p1 ON dob_t2 TO regress_rls_dob_role1,regress_rls_dob_role2 USING (true);
DROP OWNED BY regress_rls_dob_role1;
DROP POLICY p1 ON dob_t2; -- should succeed
DROP USER regress_rls_dob_role1;
DROP USER regress_rls_dob_role2;
-- Bug #15708: view + table with RLS should check policies as view owner
CREATE TABLE ref_tbl (a int);
INSERT INTO ref_tbl VALUES (1);
CREATE TABLE rls_tbl (a int);
INSERT INTO rls_tbl VALUES (10);
ALTER TABLE rls_tbl ENABLE ROW LEVEL SECURITY;
CREATE POLICY p1 ON rls_tbl USING (EXISTS (SELECT 1 FROM ref_tbl));
GRANT SELECT ON ref_tbl TO regress_rls_bob;
GRANT SELECT ON rls_tbl TO regress_rls_bob;
CREATE VIEW rls_view AS SELECT * FROM rls_tbl;
ALTER VIEW rls_view OWNER TO regress_rls_bob;
GRANT SELECT ON rls_view TO regress_rls_alice;
SET SESSION AUTHORIZATION regress_rls_alice;
SELECT * FROM ref_tbl; -- Permission denied
SELECT * FROM rls_tbl; -- Permission denied
SELECT * FROM rls_view; -- OK
RESET SESSION AUTHORIZATION;
DROP VIEW rls_view;
DROP TABLE rls_tbl;
DROP TABLE ref_tbl;
-- Leaky operator test
CREATE TABLE rls_tbl (a int);
INSERT INTO rls_tbl SELECT x/10 FROM generate_series(1, 100) x;
ANALYZE rls_tbl;
ALTER TABLE rls_tbl ENABLE ROW LEVEL SECURITY;
GRANT SELECT ON rls_tbl TO regress_rls_alice;
SET SESSION AUTHORIZATION regress_rls_alice;
CREATE FUNCTION op_leak(int, int) RETURNS bool
AS 'BEGIN RAISE NOTICE ''op_leak => %, %'', $1, $2; RETURN $1 < $2; END'
LANGUAGE plpgsql;
CREATE OPERATOR <<< (procedure = op_leak, leftarg = int, rightarg = int,
restrict = scalarltsel);
SELECT * FROM rls_tbl WHERE a <<< 1000;
DROP OPERATOR <<< (int, int);
DROP FUNCTION op_leak(int, int);
RESET SESSION AUTHORIZATION;
DROP TABLE rls_tbl;
-- Bug #16006: whole-row Vars in a policy don't play nice with sub-selects
SET SESSION AUTHORIZATION regress_rls_alice;
CREATE TABLE rls_tbl (a int, b int, c int);
CREATE POLICY p1 ON rls_tbl USING (rls_tbl >= ROW(1,1,1));
ALTER TABLE rls_tbl ENABLE ROW LEVEL SECURITY;
ALTER TABLE rls_tbl FORCE ROW LEVEL SECURITY;
INSERT INTO rls_tbl SELECT 10, 20, 30;
EXPLAIN (VERBOSE, COSTS OFF)
INSERT INTO rls_tbl
SELECT * FROM (SELECT b, c FROM rls_tbl ORDER BY a) ss;
INSERT INTO rls_tbl
SELECT * FROM (SELECT b, c FROM rls_tbl ORDER BY a) ss;
SELECT * FROM rls_tbl;
DROP TABLE rls_tbl;
RESET SESSION AUTHORIZATION;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
--
-- Clean up objects
--
RESET SESSION AUTHORIZATION;
DROP SCHEMA regress_rls_schema CASCADE;
Row-Level Security Policies (RLS) Building on the updatable security-barrier views work, add the ability to define policies on tables to limit the set of rows which are returned from a query and which are allowed to be added to a table. Expressions defined by the policy for filtering are added to the security barrier quals of the query, while expressions defined to check records being added to a table are added to the with-check options of the query. New top-level commands are CREATE/ALTER/DROP POLICY and are controlled by the table owner. Row Security is able to be enabled and disabled by the owner on a per-table basis using ALTER TABLE .. ENABLE/DISABLE ROW SECURITY. Per discussion, ROW SECURITY is disabled on tables by default and must be enabled for policies on the table to be used. If no policies exist on a table with ROW SECURITY enabled, a default-deny policy is used and no records will be visible. By default, row security is applied at all times except for the table owner and the superuser. A new GUC, row_security, is added which can be set to ON, OFF, or FORCE. When set to FORCE, row security will be applied even for the table owner and superusers. When set to OFF, row security will be disabled when allowed and an error will be thrown if the user does not have rights to bypass row security. Per discussion, pg_dump sets row_security = OFF by default to ensure that exports and backups will have all data in the table or will error if there are insufficient privileges to bypass row security. A new option has been added to pg_dump, --enable-row-security, to ask pg_dump to export with row security enabled. A new role capability, BYPASSRLS, which can only be set by the superuser, is added to allow other users to be able to bypass row security using row_security = OFF. Many thanks to the various individuals who have helped with the design, particularly Robert Haas for his feedback. Authors include Craig Ringer, KaiGai Kohei, Adam Brightwell, Dean Rasheed, with additional changes and rework by me. Reviewers have included all of the above, Greg Smith, Jeff McCormick, and Robert Haas.
2014-09-19 17:18:35 +02:00
DROP USER regress_rls_alice;
DROP USER regress_rls_bob;
DROP USER regress_rls_carol;
DROP USER regress_rls_dave;
DROP USER regress_rls_exempt_user;
DROP ROLE regress_rls_group1;
DROP ROLE regress_rls_group2;
-- Arrange to have a few policies left over, for testing
-- pg_dump/pg_restore
CREATE SCHEMA regress_rls_schema;
CREATE TABLE rls_tbl (c1 int);
ALTER TABLE rls_tbl ENABLE ROW LEVEL SECURITY;
CREATE POLICY p1 ON rls_tbl USING (c1 > 5);
CREATE POLICY p2 ON rls_tbl FOR SELECT USING (c1 <= 3);
CREATE POLICY p3 ON rls_tbl FOR UPDATE USING (c1 <= 3) WITH CHECK (c1 > 5);
CREATE POLICY p4 ON rls_tbl FOR DELETE USING (c1 <= 3);
CREATE TABLE rls_tbl_force (c1 int);
ALTER TABLE rls_tbl_force ENABLE ROW LEVEL SECURITY;
ALTER TABLE rls_tbl_force FORCE ROW LEVEL SECURITY;
CREATE POLICY p1 ON rls_tbl_force USING (c1 = 5) WITH CHECK (c1 < 5);
CREATE POLICY p2 ON rls_tbl_force FOR SELECT USING (c1 = 8);
CREATE POLICY p3 ON rls_tbl_force FOR UPDATE USING (c1 = 8) WITH CHECK (c1 >= 5);
CREATE POLICY p4 ON rls_tbl_force FOR DELETE USING (c1 = 8);