I'm testing database permissions on PostgreSQL and i'm trying to prevent a common user (readuser) from executing an 'ALTER DEFAULT PRIVILEGES' statement. However i cannot find a way to revoke this specific permission, and couldn't find anything about it on documentation.
I started a local PostgreSQL 11.2 instance, removed connect permisssions, created a database testdb and revoked table creation on the public schema.
revoke connect on database postgres from public;
create database testdb with template template0 --lc_collate "pt_BR.utf8" lc_ctype "pt_BR.utf8";
revoke connect on database testdb from public;
\c :database
revoke all on schema public from public;
grant all on schema public to postgres;
create schema private;
After that, I created a user with read permissions only:
create user readuser
with nosuperuser
nocreatedb
nocreaterole
noreplication
login
encrypted password 'testpassword';
grant connect
on database testdb
to readuser;
Then create a schema testschema and granted read permissions on it's tables:
grant usage
on schema testschema
to readuser;
grant select
on all tables
in schema testschema
to readuser;
Even though i only set read permissions on all schemas and tables, the 'readuser' user can still perform 'alter default privileges' query without a permission error:
alter default privileges in schema testschema grant select on tables to readuser;
ALTER DEFAULT PRIVILEGES
I would like some help on preventing a user from altering it's default privileges, so that it cannot mess up permissions for tables created in the future.
Try this by revoking the EXECUTE from the role postgres that granted the default privilege of execute to readuser
ALTER DEFAULT PRIVILEGES FOR ROLE postgres IN SCHEMA testschema REVOKE EXECUTE ON FUNCTIONS FROM readuser;
Related
My Postgres DB has a user - "myuser" who has access to everycommand in schema my_schema.
I want to restrict the user to use only basic sequelize commands like Create, Update, Delete, Select. But the user should not be able to Alter the table (in sequelize, he shouldn't be able to do .sync{} operation).
I have already tried doing:
REVOKE ALTER ALL TABLES IN SCHEMA my_schema from myuser
But the above command is giving error!
Commands I have used to create role is:
CREATE ROLE app_readwrite1;
GRANT CONNECT ON DATABASE dvdrental TO app_readwrite1;
GRANT USAGE, CREATE ON SCHEMA app_schema TO app_readwrite1;
GRANT SELECT, INSERT, UPDATE, DELETE ON ALL TABLES IN SCHEMA app_schema TO app_readwrite1;
ALTER DEFAULT PRIVILEGES IN SCHEMA app_schema GRANT SELECT, INSERT, UPDATE, DELETE ON TABLES TO app_readwrite1;
GRANT USAGE ON ALL SEQUENCES IN SCHEMA app_schema TO app_readwrite1;
ALTER DEFAULT PRIVILEGES IN SCHEMA app_schema GRANT USAGE ON SEQUENCES TO app_readwrite1;
After creating role above, I am creating user and assigning him the role.
Still, I can use Model.sync()- basically "ALTER" command along with CRUD operations.
How can I restrict the user from using alter command?
I was trying to delete PostgreSQL user:
DROP USER ryan;
I received this error:
Error in query:
ERROR: role "ryan" cannot be dropped because some objects depend on it
DETAIL: privileges for database mydatabase
I looked for a solution from these threads:
PostgreSQL - how to quickly drop a user with existing privileges
How to drop user in postgres if it has depending objects
Still have the same error.
This happens after I grant all permission to user "ryan" with:
GRANT ALL PRIVILEGES ON DATABASE mydatabase ON SCHEMA public TO ryan;
DROP USER (or DROP ROLE, same thing) cannot proceed while the role still owns anything or has any granted privileges on other objects.
Get rid of all privileges with DROP OWNED (which isn't too obvious from the wording). The manual:
[...] Any privileges granted to the given roles on objects in the current
database and on shared objects (databases, tablespaces) will also be revoked.
So the reliable sequence of commands to drop a role is:
REASSIGN OWNED BY ryan TO postgres; -- or some other trusted role
DROP OWNED BY ryan;
Run both commands in every database of the same cluster where the role owns anything or has any privileges!
And finally:
DROP USER ryan;
REASSIGN OWNED changes ownership for all objects currently owned by the role.
DROP OWNED then only revokes privileges (ownerships out of the way).
Alternatively, you can skip REASSIGN OWNED. Then DROP OWNED will (also) drop all objects owned by the user. (Are you sure?!)
Related:
Drop a role with privileges (with a function to generate commands for all relevant DBs)
Find objects linked to a PostgreSQL role
What worked for me was to follow these steps:
Connecting to the database
\c mydatabase
Reassigning ownership
REASSIGN OWNED BY ryan TO <newuser>;
Or/and just deleting the object
DROP OWNED BY ryan;
Executing REVOKE PRIVILEGES
REVOKE ALL PRIVILEGES ON ALL TABLES IN SCHEMA public FROM ryan;
REVOKE ALL PRIVILEGES ON ALL SEQUENCES IN SCHEMA public FROM ryan;
REVOKE ALL PRIVILEGES ON ALL FUNCTIONS IN SCHEMA public FROM ryan;
Dropping the user
DROP USER ryan;
PS: You might not need to execute both Step 2 and 3, just one of the two steps might be usually enough.
What worked for me on RDS Postgres 13:
REVOKE ALL PRIVILEGES ON DATABASE <my_db> FROM <my_user>;
I also had a similar error where the role was owner for tables so it couldn't be dropped, had to re-assign table owner with:
ALTER TABLE <my_table> OWNER TO <trusted_role>;
Doing a REASSIGN like this didn't work for me on RDS because AWS doesn't give you full superuser to your master user:
REASSIGN OWNED BY <olduser> TO <newuser>;
REVOKE ALL ON SCHEMA "public" FROM "<user>";
Worked for me.
What worked for me was to recreate template1 database and then drop some role:
$ psql -U postgres postgres
postgres=# update pg_database set datistemplate = false where datname='template1';
UPDATE 1
postgres=# drop database template1;
DROP DATABASE
postgres=# create database template1 template=template0;
CREATE DATABASE
postgres=# update pg_database set datistemplate = true where datname='template1';
UPDATE 1
postgres=# DROP ROLE test;
DROP ROLE
For people who use AWS Postgresql RDS, you may try following
login to postgres user, then grant owner
postgres=> GRANT target_user to old_user;
GRANT ROLE
Login to target db using user that would like to remove(old_user), then reassign
target_db=> REASSIGN OWNED BY old_user TO target_user;
REASSIGN OWNED
Login back to postgres user, revoke all privileges then drop user
postgres=> REVOKE ALL PRIVILEGES ON DATABASE target_db FROM old_user;
REVOKE
postgres=> DROP USER old_user;
DROP ROLE
Ref. https://aws.amazon.com/premiumsupport/knowledge-center/rds-postgresql-drop-user-role/
I am running PostgreSQL 9.3.1. I have test database and backup user which is used to backup the database. I have no problems with granting privileges to all current tables, but I have to grant privileges each time the new table is added to schema.
createdb test
psql test
test=# create table foo();
CREATE TABLE
test=# grant all on all tables in schema public to backup;
GRANT
test=# create table bar();
CREATE TABLE
psql -U backup test
test=> select * from foo;
test=> select * from bar;
ERROR: permission denied for relation bar
Is it possible to grant access to tables which will be created in future without making user owner of the table?
It looks like the solution is to alter default privileges for backup user:
alter default privileges in schema public grant all on tables to backup;
alter default privileges in schema public grant all on sequences to backup;
From the comment by Matt Schaffer:
As caveat, the default only applies to the user that executed the
alter statement. This confused me since I was driving most of my
permissions statements from the postgres user but creating tables from
an app user. In short, you might need something like this depending on
your setup:
ALTER DEFAULT PRIVILEGES FOR USER webapp IN SCHEMA public GRANT SELECT ON SEQUENCES TO backup;
ALTER DEFAULT PRIVILEGES FOR USER webapp IN SCHEMA public GRANT SELECT ON TABLES TO backup;
Where webapp is the user that will be creating new tables in the futrue and backup is the user that will be able to read from new tables created by webapp.
If you want the backup user to have access to the future tables of userN,
you must run the code below under each userN who creates new tables,
because ALTER DEFAULT PRIVILEGES...
works only for objects by that user under whom you run ALTER DEFAULT PRIVILEGES...
ALTER DEFAULT PRIVILEGES IN SCHEMA public GRANT ALL ON TABLES TO backup;
I am trying to create a role, grant connect access to the role and then alter default privileges to keep access for future objects. However, it seems that the below command doesn't work at role level.
alter default privileges in schema public grant all on tables to backup;
I followed the below documentation but seems that there are two command do not work for roles.
DOC: https://aws.amazon.com/blogs/database/managing-postgresql-users-and-roles/
First command:
GRANT CONNECT ON DATABASE mydatabase TO readonly;
Second command:
GRANT USAGE ON SCHEMA myschema TO readonly;
(For ROLES usually it needs TO ROLE, I also tried TO ROLE but still doesn't work.
I would like to give a user all the permissions on a database without making it an admin.
The reason why I want to do that is that at the moment DEV and PROD are different DBs on the same cluster so I don't want a user to be able to change production objects but it must be able to change objects on DEV.
I tried:
grant ALL on database MY_DB to group MY_GROUP;
but it doesn't seem to give any permission.
Then I tried:
grant all privileges on schema MY_SCHEMA to group MY_GROUP;
and it seems to give me permission to create objects but not to query\delete objects on that schema that belong to other users
I could go on by giving USAGE permission to the user on MY_SCHEMA but then it would complain about not having permissions on the table ...
So I guess my question is: is there any easy way of giving all the permissions to a user on a DB?
I'm working on PostgreSQL 8.1.23.
All commands must be executed while connected to the right database cluster. Make sure of it.
Roles are objects of the database cluster. All databases of the same cluster share the set of defined roles. Privileges are granted / revoked per database / schema / table etc.
A role needs access to the database, obviously. That's granted to PUBLIC by default. Else:
GRANT CONNECT ON DATABASE my_db TO my_user;
Basic privileges for Postgres 14 or later
Postgres 14 adds the predefined, non-login roles pg_read_all_data / pg_write_all_data.
They have SELECT / INSERT, UPDATE, DELETE privileges for all tables, views, and sequences. Plus USAGE on schemas. We can GRANT membership in these roles:
GRANT pg_read_all_data TO my_user;
GRANT pg_write_all_data TO my_user;
This covers all basic DML commands (but not DDL, and not some special commands like TRUNCATE or the EXECUTE privilege for functions!). The manual:
pg_read_all_data
Read all data (tables, views, sequences), as if having SELECT rights
on those objects, and USAGE rights on all schemas, even without
having it explicitly. This role does not have the role attribute
BYPASSRLS set. If RLS is being used, an administrator may wish to
set BYPASSRLS on roles which this role is GRANTed to.
pg_write_all_data
Write all data (tables, views, sequences), as if having INSERT,
UPDATE, and DELETE rights on those objects, and USAGE rights on
all schemas, even without having it explicitly. This role does not
have the role attribute BYPASSRLS set. If RLS is being used, an
administrator may wish to set BYPASSRLS on roles which this role is
GRANTed to.
All privileges without using predefined roles (any Postgres version)
Commands must be executed while connected to the right database. Make sure of it.
The role needs (at least) the USAGE privilege on the schema. Again, if that's granted to PUBLIC, you are covered. Else:
GRANT USAGE ON SCHEMA public TO my_user;
Or grant USAGE on all custom schemas:
DO
$$
BEGIN
-- RAISE NOTICE '%', ( -- use instead of EXECUTE to see generated commands
EXECUTE (
SELECT string_agg(format('GRANT USAGE ON SCHEMA %I TO my_user', nspname), '; ')
FROM pg_namespace
WHERE nspname <> 'information_schema' -- exclude information schema and ...
AND nspname NOT LIKE 'pg\_%' -- ... system schemas
);
END
$$;
Then, all permissions for all tables (requires Postgres 9.0 or later).
And don't forget sequences (if any):
GRANT ALL PRIVILEGES ON ALL TABLES IN SCHEMA public TO my_user;
GRANT ALL PRIVILEGES ON ALL SEQUENCES IN SCHEMA public TO my_user;
Alternatively, you could use the "Grant Wizard" of pgAdmin 4 to work with a GUI.
This covers privileges for existing objects. To also cover future objects, set DEFAULT PRIVILEGES. See:
Grant privileges for a particular database in PostgreSQL
How to manage DEFAULT PRIVILEGES for USERs on a DATABASE vs SCHEMA?
There are some other objects, the manual for GRANT has the complete list. As of Postgres 14:
privileges on a database object (table, column, view, foreign table, sequence, database, foreign-data wrapper, foreign server, function, procedure, procedural language, schema, or tablespace)
But the rest is rarely needed. More details:
Grant privileges for a particular database in PostgreSQL
How to grant all privileges on views to arbitrary user
Consider upgrading to a current version.
GRANT ALL PRIVILEGES ON DATABASE "my_db" to my_user;
In PostgreSQL 9.0+ you would do the following:
GRANT ALL PRIVILEGES ON ALL TABLES IN SCHEMA MY_SCHEMA TO MY_GROUP;
GRANT ALL PRIVILEGES ON ALL SEQUENCES IN SCHEMA MY_SCHEMA TO MY_GROUP;
If you want to enable this for newly created relations too, then set the default permissions:
ALTER DEFAULT PRIVILEGES IN SCHEMA MY_SCHEMA
GRANT ALL PRIVILEGES ON TABLES TO MY_GROUP;
ALTER DEFAULT PRIVILEGES IN SCHEMA MY_SCHEMA
GRANT ALL PRIVILEGES ON SEQUENCES TO MY_GROUP;
However, seeing that you use 8.1 you have to code it yourself:
CREATE FUNCTION grant_all_in_schema (schname name, grant_to name) RETURNS integer AS $$
DECLARE
rel RECORD;
BEGIN
FOR rel IN
SELECT c.relname
FROM pg_class c
JOIN pg_namespace s ON c.namespace = s.oid
WHERE s.nspname = schname
LOOP
EXECUTE 'GRANT ALL PRIVILEGES ON ' || quote_ident(schname) || '.' || rel.relname || ' TO ' || quote_ident(grant_to);
END LOOP;
RETURN 1;
END; $$ LANGUAGE plpgsql STRICT;
REVOKE ALL ON FUNCTION grant_all_in_schema(name, name) FROM PUBLIC;
This will set the privileges on all relations: tables, views, indexes, sequences, etc. If you want to restrict that, filter on pg_class.relkind. See the pg_class docs for details.
You should run this function as superuser and as regular as your application requires. An option would be to package this in a cron job that executes every day or every hour.
I did the following to add a role 'eSumit' on PostgreSQL 9.4.15 database and provide all permission to this role :
CREATE ROLE eSumit;
GRANT ALL PRIVILEGES ON ALL TABLES IN SCHEMA public TO eSumit;
GRANT ALL PRIVILEGES ON DATABASE "postgres" to eSumit;
ALTER USER eSumit WITH SUPERUSER;
Also checked the pg_table enteries via :
select * from pg_roles;
Database queries snapshot :
In PostgreSQL 12 and later, it is possible to grant all privileges of a table in a database to a role/user/account.
The syntax is:
GRANT ALL ON table_name TO role_name;
If you want to grant it to all tables in the database then the syntax will be:
GRANT ALL ON ALL TABLES TO role_name;
If you want to grant it to all tables of a schema in the database then the syntax will be:
GRANT ALL ON ALL TABLES IN SCHEMA schema_name TO role_name;
Note: Remember you will need to select the database before you can grant its privileges to a user.
Resources: PostgreSQL GRANT
That's all
I hope this helps
GRANT USAGE ON SCHEMA schema_name TO user;
GRANT ALL ON SCHEMA schema_name TO user_name;
Give all permissions to a user on a PostgreSQL database:
Command:
grant all privileges on database [database_name] to [database_user_name];
Example:
grant all privileges on database studentdb to shaifullah;
OR
GRANT ALL PRIVILEGES ON DATABASE studentdb TO shaifullah;
My goal is to only enable a specific user to execute functions in a specific schema, list the functions available by name but not see the source code of the function or list other schema.
It is possible to achieve the above without the ability to list the available function names via carrying out the following:
First create a test user role:
CREATE ROLE test_user WITH LOGIN PASSWORD 'secret';
Now revoke all permissions from the public on all schemas:
REVOKE ALL PRIVILEGES ON DATABASE test_db FROM PUBLIC;
REVOKE ALL PRIVILEGES ON ALL TABLES IN SCHEMA public FROM PUBLIC;
REVOKE ALL PRIVILEGES ON ALL FUNCTIONS IN SCHEMA public FROM PUBLIC;
REVOKE ALL PRIVILEGES ON ALL SEQUENCES IN SCHEMA public FROM PUBLIC;
REVOKE ALL ON SCHEMA public FROM PUBLIC;
REVOKE ALL PRIVILEGES ON ALL TABLES IN SCHEMA function_schema FROM PUBLIC;
REVOKE ALL PRIVILEGES ON ALL FUNCTIONS IN SCHEMA function_schema FROM PUBLIC;
REVOKE ALL PRIVILEGES ON ALL SEQUENCES IN SCHEMA function_schema FROM PUBLIC;
REVOKE ALL ON SCHEMA function_schema FROM PUBLIC;
REVOKE ALL PRIVILEGES ON ALL TABLES IN SCHEMA table_schema FROM PUBLIC;
REVOKE ALL PRIVILEGES ON ALL FUNCTIONS IN SCHEMA table_schema FROM PUBLIC;
REVOKE ALL PRIVILEGES ON ALL SEQUENCES IN SCHEMA table_schema FROM PUBLIC;
REVOKE ALL ON SCHEMA table_schema FROM PUBLIC;
Setup restricted access for test user:
GRANT CONNECT ON DATABASE test_db TO test_user;
GRANT USAGE ON SCHEMA function_schema TO test_user;
REVOKE ALL ON SCHEMA public FROM test_user;
REVOKE ALL ON SCHEMA table_schema FROM test_user;
GRANT EXECUTE ON FUNCTION function_schema.function1() TO test_user;
GRANT EXECUTE ON FUNCTION function_schema.function2(integer) TO test_user;
Now to hid the schema structure and code form the test users and public:
REVOKE SELECT ON TABLE pg_proc FROM public;
REVOKE SELECT ON TABLE pg_proc FROM test_user;
This all works well, the test user can execute the functions but they can't see the code inside the functions, nor can they see the schema and table structure.
--
I'd like to allow the test user to now see the functions by name in the test_functions schema.I've tried the following according to GRANT Postgresql 9.3 (this is granting select on every column in pg_proc):
GRANT SELECT (proname,pronamespace,proowner,prolang,procost,prorows,
provariadic,protransform,proisagg,proiswindow,prosecdef,proleakproof,
proisstrict,proretset,provolatile,pronargs,pronargdefaults,prorettype,
proargtypes,proallargtypes,proargmodes,proargnames,proargdefaults,prosrc,
probin,proconfig,proacl) ON TABLE pg_proc TO test_user;
The result here is that the test user does not get all the same select permissions as if they had access to the whole table. They still can't see the function names.
Another test was to do the reverse, grant select to the table then revoke select on all columns accouding to REVOKE postgresql 9.3:
GRANT SELECT ON TABLE pg_proc TO test_user;
REVOKE SELECT (proname,pronamespace,proowner,prolang,procost,prorows,
provariadic,protransform,proisagg,proiswindow,prosecdef,proleakproof,
proisstrict,proretset,provolatile,pronargs,pronargdefaults,prorettype,
proargtypes,proallargtypes,proargmodes,proargnames,proargdefaults,prosrc,
probin,proconfig,proacl) ON TABLE pg_proc FROM test_user;
Again this doesn't work, they now can see all schemas, code and tables (on allowed schemas).
It appears that the grant/revoke on specific columns doesn't work the way the documentation suggests.
Searching extensively yielded How to restrict access to code in a function wich suggests revoking access to only the pg_proc.prosrc column which clearly doesn't work from the tests above.
I'm using postgresql 9.3
Please feel free to suggest any other solution that comes to mind.
Working with column permissions should work but is probably not the best approach here. Maintaining privileges in this way is rather burdensome in PostgreSQL. Using this approach you have to set and track your security policy for individual users on all affected system catalog relations. Just imagine having to update your security policy when you have more than a few users.
Rather than applying detailed security roles on relations, I suggest that you lock everything in the system catalogs down (as you did) and then create views to selectively expose parts of the system catalogs (this is already the case in the standard setup, but apparently not strict enough for your case). GRANT SELECT on those views to a group role and then grant that group role to login roles (which should have the INHERIT property). This way it is much easier to track what you have done and to update your security policy as all policy is contained in a set of views (content) and group roles (accessibility). If you are using PostgreSQL-9.2+ check out the with security-barrier option on the views as this will prevent optimizer-spoofing by malicious users.