Setuid functions

Started by Mark Volpealmost 25 years ago42 messageshackers
Jump to latest
#1Mark Volpe
volpe.mark@epamail.epa.gov

I know this topic was discussed a few months ago, but I'm wondering if any
decisions have been reached on if, how, and when setuid functions and triggers
might be implemented. If not, I have an idea to throw at it.

Thanks,
Mark

#2Mark Volpe
volpe.mark@epa.gov
In reply to: Mark Volpe (#1)
Re: Setuid functions

Okay, what I'm thinking is to have a pair of commands added to PL/pgSQL. For
the sake of example we'll call them:

ENABLE PRIVLEDGE --> Sets user ID to that of the function's owner
DISABLE PRIVLEDGE --> Restores the original UID

I saw something like this being used in the referential integrity code, and
thought this would be a nice (albeit incomplete) way to generalize this
ability.

Mark

Mark Volpe wrote:

Show quoted text

I know this topic was discussed a few months ago, but I'm wondering if any
decisions have been reached on if, how, and when setuid functions and triggers
might be implemented. If not, I have an idea to throw at it.

Thanks,
Mark

---------------------------(end of broadcast)---------------------------
TIP 1: subscribe and unsubscribe commands go to majordomo@postgresql.org

#3Christopher Kings-Lynne
chriskl@familyhealth.com.au
In reply to: Mark Volpe (#2)
RE: Setuid functions

Okay, what I'm thinking is to have a pair of commands added to
PL/pgSQL. For
the sake of example we'll call them:

ENABLE PRIVLEDGE --> Sets user ID to that of the function's owner
DISABLE PRIVLEDGE --> Restores the original UID

I hope you mean:

ENABLE PRIVILEGE
DISABLE PRIVILEGE

Chris

#4Ross J. Reedstrom
reedstrm@rice.edu
In reply to: Christopher Kings-Lynne (#3)
Re: Setuid functions

On Thu, Jun 21, 2001 at 09:44:43AM +0800, Christopher Kings-Lynne wrote:

Okay, what I'm thinking is to have a pair of commands added to
PL/pgSQL. For
the sake of example we'll call them:

ENABLE PRIVLEDGE --> Sets user ID to that of the function's owner
DISABLE PRIVLEDGE --> Restores the original UID

I hope you mean:

ENABLE PRIVILEGE
DISABLE PRIVILEGE

Come on, Chris, you've never heard about SQL standard LEDGE? That's
the nomenclature they chose to describe a collection of permissions:
a SHELF or LEDGE. PUBLEDGE, USERLEDGE, PRIVLEDGE. So, the last is the
PRIVATE LEDGE, reserved for the owner of the object whose access is
being determined (or was that PRIVITHEDGE? now I'm confused)

... or something. ;-) Actually, not too far from how some of the SQL92
standards docs actually seem to read, especially after falling asleep
face down on the keyboard will trying to understand them, and having
vivid dreams.

Ross (who's in the office much too late, working on budget justifications
for grants that are due tomorrow!)

#5Christopher Kings-Lynne
chriskl@familyhealth.com.au
In reply to: Ross J. Reedstrom (#4)
RE: Setuid functions

Come on, Chris, you've never heard about SQL standard LEDGE? That's
the nomenclature they chose to describe a collection of permissions:
a SHELF or LEDGE. PUBLEDGE, USERLEDGE, PRIVLEDGE. So, the last is the
PRIVATE LEDGE, reserved for the owner of the object whose access is
being determined (or was that PRIVITHEDGE? now I'm confused)

It's alright - Americans don't seem to be able to spell the same as the rest
of the English speaking world anyway ;)

Chris

#6Mark Volpe
volpe.mark@epa.gov
In reply to: Christopher Kings-Lynne (#3)
[PATCH] Re: Setuid functions

Sorry, I have decided not to follow the SQL standard ;-) PRIVILEGE is spelled
correctly in my patch.

This patch will implement the "ENABLE PRIVILEGE" and "DISABLE PRIVILEGE"
commands in PL/pgSQL, which, respectively, change the effective uid to that
of the function owner and back. It doesn't break security (I hope). The
commands can be abbreviated as "ENABLE" and "DISABLE" for the poor saps that
have trouble with "PRIVILEGE" :) Easier than adding a setuid bit to the
catalog, no?

Apologies if the patch is not in the correct format. Apply with

patch -p1 < enable_disable.patch

in the tippety-top of the 7.1.2 tree.

Regression example:

CREATE USER sample_user;
CREATE TABLE test_log(stamp datetime);
GRANT SELECT ON test_log TO PUBLIC;

DROP FUNCTION test_enable();
CREATE FUNCTION test_enable() RETURNS boolean AS
'
DECLARE
user name;
BEGIN
user:=current_user;
RAISE NOTICE ''Username: %'', user;
ENABLE PRIVILEGE;
user:=current_user;
RAISE NOTICE ''Username: %'', user;
INSERT INTO test_log VALUES(''now''::text);
DISABLE PRIVILEGE; -- Actually unnecessary at the end of the function
RETURN TRUE;
END;
' LANGUAGE 'plpgsql';

\c - sample_user
SELECT test_enable();
SELECT * FROM test_log;

stamp
------------------------
2001-06-21 11:17:29-04

(Note current time logged into a table where sample_user could not normally
write)

Hope you will find this useful
- Mark

"Ross J. Reedstrom" wrote:

Show quoted text

Come on, Chris, you've never heard about SQL standard LEDGE? That's
the nomenclature they chose to describe a collection of permissions:
a SHELF or LEDGE. PUBLEDGE, USERLEDGE, PRIVLEDGE. So, the last is the
PRIVATE LEDGE, reserved for the owner of the object whose access is
being determined (or was that PRIVITHEDGE? now I'm confused)

... or something. ;-) Actually, not too far from how some of the SQL92
standards docs actually seem to read, especially after falling asleep
face down on the keyboard will trying to understand them, and having
vivid dreams.

Ross (who's in the office much too late, working on budget justifications
for grants that are due tomorrow!)

---------------------------(end of broadcast)---------------------------
TIP 2: you can get off all lists at once with the unregister command
(send "unregister YourEmailAddressHere" to majordomo@postgresql.org)

Attachments:

enable_disable.patchtext/plain; charset=us-ascii; name=enable_disable.patchDownload+122-2
#7Bruce Momjian
bruce@momjian.us
In reply to: Mark Volpe (#6)
Re: [PATCH] Re: Setuid functions

Your patch has been added to the PostgreSQL unapplied patches list at:

http://candle.pha.pa.us/cgi-bin/pgpatches

I will try to apply it within the next 48 hours.

Sorry, I have decided not to follow the SQL standard ;-) PRIVILEGE is spelled
correctly in my patch.

This patch will implement the "ENABLE PRIVILEGE" and "DISABLE PRIVILEGE"
commands in PL/pgSQL, which, respectively, change the effective uid to that
of the function owner and back. It doesn't break security (I hope). The
commands can be abbreviated as "ENABLE" and "DISABLE" for the poor saps that
have trouble with "PRIVILEGE" :) Easier than adding a setuid bit to the
catalog, no?

Apologies if the patch is not in the correct format. Apply with

patch -p1 < enable_disable.patch

in the tippety-top of the 7.1.2 tree.

Regression example:

CREATE USER sample_user;
CREATE TABLE test_log(stamp datetime);
GRANT SELECT ON test_log TO PUBLIC;

DROP FUNCTION test_enable();
CREATE FUNCTION test_enable() RETURNS boolean AS
'
DECLARE
user name;
BEGIN
user:=current_user;
RAISE NOTICE ''Username: %'', user;
ENABLE PRIVILEGE;
user:=current_user;
RAISE NOTICE ''Username: %'', user;
INSERT INTO test_log VALUES(''now''::text);
DISABLE PRIVILEGE; -- Actually unnecessary at the end of the function
RETURN TRUE;
END;
' LANGUAGE 'plpgsql';

\c - sample_user
SELECT test_enable();
SELECT * FROM test_log;

stamp
------------------------
2001-06-21 11:17:29-04

(Note current time logged into a table where sample_user could not normally
write)

Hope you will find this useful
- Mark

"Ross J. Reedstrom" wrote:

Come on, Chris, you've never heard about SQL standard LEDGE? That's
the nomenclature they chose to describe a collection of permissions:
a SHELF or LEDGE. PUBLEDGE, USERLEDGE, PRIVLEDGE. So, the last is the
PRIVATE LEDGE, reserved for the owner of the object whose access is
being determined (or was that PRIVITHEDGE? now I'm confused)

... or something. ;-) Actually, not too far from how some of the SQL92
standards docs actually seem to read, especially after falling asleep
face down on the keyboard will trying to understand them, and having
vivid dreams.

Ross (who's in the office much too late, working on budget justifications
for grants that are due tomorrow!)

---------------------------(end of broadcast)---------------------------
TIP 2: you can get off all lists at once with the unregister command
(send "unregister YourEmailAddressHere" to majordomo@postgresql.org)

diff -ur postgresql-7.1.2/src/pl/plpgsql/src/gram.y postgresql-7.1.2-patch/src/pl/plpgsql/src/gram.y
--- postgresql-7.1.2/src/pl/plpgsql/src/gram.y	Wed Jun 20 20:07:45 2001
+++ postgresql-7.1.2-patch/src/pl/plpgsql/src/gram.y	Wed Jun 20 19:48:18 2001
@@ -121,7 +121,7 @@
%type <stmts>	proc_sect, proc_stmts, stmt_else, loop_body
%type <stmt>	proc_stmt, pl_block
%type <stmt>	stmt_assign, stmt_if, stmt_loop, stmt_while, stmt_exit
-%type <stmt>	stmt_return, stmt_raise, stmt_execsql, stmt_fori
+%type <stmt>	stmt_return, stmt_raise, stmt_execsql, stmt_fori, stmt_enable, stmt_disable
%type <stmt>	stmt_fors, stmt_select, stmt_perform
%type <stmt>	stmt_dynexecute, stmt_dynfors, stmt_getdiag
@@ -164,6 +164,9 @@
%token	K_PERFORM
%token	K_ROW_COUNT
%token	K_RAISE
+%token  K_ENABLE
+%token  K_DISABLE
+%token  K_PRIVILEGE
%token	K_RECORD
%token	K_RENAME
%token	K_RESULT_OID
@@ -569,6 +572,10 @@
{ $$ = $1; }
| stmt_raise
{ $$ = $1; }
+				| stmt_enable
+						{ $$ = $1; }
+				| stmt_disable
+						{ $$ = $1; }
| stmt_execsql
{ $$ = $1; }
| stmt_dynexecute
@@ -1033,6 +1040,34 @@
$$ = (PLpgSQL_stmt *)new;
}
;
+
+stmt_enable		: K_ENABLE opt_privilege lno ';'
+				{
+					PLpgSQL_stmt_privilege *new;
+
+					new=malloc(sizeof(PLpgSQL_stmt_privilege));
+
+					new->cmd_type = PLPGSQL_STMT_ENABLE;
+					new->lineno = $3;
+
+					$$ = (PLpgSQL_stmt *)new;
+				}
+
+stmt_disable		: K_DISABLE opt_privilege lno ';'
+				{
+					PLpgSQL_stmt_privilege *new;
+
+					new=malloc(sizeof(PLpgSQL_stmt_privilege));
+
+					new->cmd_type = PLPGSQL_STMT_DISABLE;
+					new->lineno = $3;
+
+					$$ = (PLpgSQL_stmt *)new;
+				}
+
+opt_privilege :
+	| K_PRIVILEGE
+;
stmt_raise		: K_RAISE lno raise_level raise_msg raise_params ';'
{
diff -ur postgresql-7.1.2/src/pl/plpgsql/src/pl_exec.c postgresql-7.1.2-patch/src/pl/plpgsql/src/pl_exec.c
--- postgresql-7.1.2/src/pl/plpgsql/src/pl_exec.c	Wed Jun 20 20:07:45 2001
+++ postgresql-7.1.2-patch/src/pl/plpgsql/src/pl_exec.c	Wed Jun 20 19:48:18 2001
@@ -99,6 +99,8 @@
PLpgSQL_stmt_exit * stmt);
static int exec_stmt_return(PLpgSQL_execstate * estate,
PLpgSQL_stmt_return * stmt);
+static int exec_stmt_privilege(PLpgSQL_execstate * estate,
+				PLpgSQL_stmt_privilege * stmt);
static int exec_stmt_raise(PLpgSQL_execstate * estate,
PLpgSQL_stmt_raise * stmt);
static int exec_stmt_execsql(PLpgSQL_execstate * estate,
@@ -220,6 +222,12 @@
case PLPGSQL_STMT_RETURN:
stmttype = "return";
break;
+					case PLPGSQL_STMT_ENABLE:
+						stmttype = "enable";
+						break;
+					case PLPGSQL_STMT_DISABLE:
+						stmttype = "disable";
+						break;
case PLPGSQL_STMT_RAISE:
stmttype = "raise";
break;
@@ -265,6 +273,8 @@
estate.retistuple = func->fn_retistuple;
estate.retisset = func->fn_retset;
estate.exitlabel = NULL;
+	estate.save_uid = InvalidOid;
+	estate.fn_oid = func->fn_oid;

estate.found_varno = func->found_varno;
estate.ndatums = func->ndatums;
@@ -385,6 +395,9 @@
elog(ERROR, "control reaches end of function without RETURN");
}

+	if (estate.save_uid!=InvalidOid)
+		SetUserId(estate.save_uid);
+
/*
* We got a return value - process it
*/
@@ -565,6 +578,8 @@
estate.retistuple = func->fn_retistuple;
estate.retisset = func->fn_retset;
estate.exitlabel = NULL;
+	estate.save_uid = InvalidOid;
+	estate.fn_oid = func->fn_oid;

estate.found_varno = func->found_varno;
estate.ndatums = func->ndatums;
@@ -742,6 +757,9 @@
elog(ERROR, "control reaches end of trigger procedure without RETURN");
}

+ if (estate.save_uid!=InvalidOid)
+ SetUserId(estate.save_uid);
+
/*
* Check that the returned tuple structure has the same attributes,
* the relation that fired the trigger has.
@@ -986,6 +1004,11 @@
rc = exec_stmt_return(estate, (PLpgSQL_stmt_return *) stmt);
break;

+		case PLPGSQL_STMT_ENABLE:
+		case PLPGSQL_STMT_DISABLE:
+			rc = exec_stmt_privilege(estate, (PLpgSQL_stmt_privilege *) stmt);
+			break;
+
case PLPGSQL_STMT_RAISE:
rc = exec_stmt_raise(estate, (PLpgSQL_stmt_raise *) stmt);
break;
@@ -1536,6 +1559,41 @@
&(estate->rettype));
return PLPGSQL_RC_RETURN;
+}
+
+/* ----------
+ * exec_stmt_privilege          Changes user ID to/from
+ *                              that of the function owner's
+ * ----------
+ */
+
+static int
+exec_stmt_privilege(PLpgSQL_execstate * estate, PLpgSQL_stmt_privilege * stmt)
+{
+	HeapTuple procedureTuple;
+	Form_pg_proc procedureStruct;
+
+	if (stmt->cmd_type==PLPGSQL_STMT_ENABLE)
+	{
+		if (estate->save_uid!=InvalidOid) return PLPGSQL_RC_OK; /* Already enabled */
+		procedureTuple = SearchSysCache(PROCOID, ObjectIdGetDatum(estate->fn_oid), 0, 0, 0);
+
+		if (!HeapTupleIsValid(procedureTuple))
+			elog(ERROR, "exec_stmt_privilege: cache lookup failed");
+
+		procedureStruct = (Form_pg_proc) GETSTRUCT(procedureTuple);
+
+		estate->save_uid = GetUserId();
+		SetUserId(procedureStruct->proowner);
+		ReleaseSysCache(procedureTuple);
+	} else
+	{
+		if (estate->save_uid==InvalidOid) return PLPGSQL_RC_OK; /* Already disabled */
+		SetUserId(estate->save_uid);
+		estate->save_uid = InvalidOid;
+	}
+
+	return PLPGSQL_RC_OK;
}
diff -ur postgresql-7.1.2/src/pl/plpgsql/src/pl_funcs.c postgresql-7.1.2-patch/src/pl/plpgsql/src/pl_funcs.c
--- postgresql-7.1.2/src/pl/plpgsql/src/pl_funcs.c	Wed Jun 20 20:07:45 2001
+++ postgresql-7.1.2-patch/src/pl/plpgsql/src/pl_funcs.c	Wed Jun 20 20:05:14 2001
@@ -382,6 +382,7 @@
static void dump_select(PLpgSQL_stmt_select * stmt);
static void dump_exit(PLpgSQL_stmt_exit * stmt);
static void dump_return(PLpgSQL_stmt_return * stmt);
+static void dump_privilege(PLpgSQL_stmt_privilege * stmt);
static void dump_raise(PLpgSQL_stmt_raise * stmt);
static void dump_execsql(PLpgSQL_stmt_execsql * stmt);
static void dump_dynexecute(PLpgSQL_stmt_dynexecute * stmt);
@@ -435,6 +436,10 @@
case PLPGSQL_STMT_RETURN:
dump_return((PLpgSQL_stmt_return *) stmt);
break;
+		case PLPGSQL_STMT_ENABLE:
+		case PLPGSQL_STMT_DISABLE:
+			dump_privilege((PLpgSQL_stmt_privilege *) stmt);
+			break;
case PLPGSQL_STMT_RAISE:
dump_raise((PLpgSQL_stmt_raise *) stmt);
break;
@@ -647,6 +652,16 @@
dump_expr(stmt->expr);
}
printf("\n");
+}
+
+static void
+dump_privilege(PLpgSQL_stmt_privilege * stmt)
+{
+	dump_ind();
+	if (stmt->cmd_type==PLPGSQL_STMT_ENABLE)
+		printf("ENABLE PRIVILEGE\n");
+	else
+		printf("DISABLE PRIVILEGE\n");
}
static void
diff -ur postgresql-7.1.2/src/pl/plpgsql/src/plpgsql.h postgresql-7.1.2-patch/src/pl/plpgsql/src/plpgsql.h
--- postgresql-7.1.2/src/pl/plpgsql/src/plpgsql.h	Wed Jun 20 20:07:45 2001
+++ postgresql-7.1.2-patch/src/pl/plpgsql/src/plpgsql.h	Wed Jun 20 20:08:57 2001
@@ -94,7 +94,9 @@
PLPGSQL_STMT_EXECSQL,
PLPGSQL_STMT_DYNEXECUTE,
PLPGSQL_STMT_DYNFORS,
-	PLPGSQL_STMT_GETDIAG
+	PLPGSQL_STMT_GETDIAG,
+	PLPGSQL_STMT_ENABLE,
+	PLPGSQL_STMT_DISABLE
};

@@ -387,6 +389,11 @@
int retrecno;
} PLpgSQL_stmt_return;

+typedef struct
+{                               /* ENABLE/DISABLE statement */
+    int         cmd_type;
+    int         lineno;
+}           PLpgSQL_stmt_privilege;

typedef struct
{ /* RAISE statement */
@@ -464,6 +471,8 @@
int found_varno;
int ndatums;
PLpgSQL_datum **datums;
+ Oid save_uid;
+ Oid fn_oid;
} PLpgSQL_execstate;

Only in postgresql-7.1.2/src/pl/plpgsql/src: plpgsql.h.orig
Only in postgresql-7.1.2/src/pl/plpgsql/src: plpgsql.h.rej
diff -ur postgresql-7.1.2/src/pl/plpgsql/src/scan.l postgresql-7.1.2-patch/src/pl/plpgsql/src/scan.l
--- postgresql-7.1.2/src/pl/plpgsql/src/scan.l	Wed Jun 20 20:07:45 2001
+++ postgresql-7.1.2-patch/src/pl/plpgsql/src/scan.l	Wed Jun 20 19:48:18 2001
@@ -115,6 +115,9 @@
null			{ return K_NULL;			}
perform			{ return K_PERFORM;			}
raise			{ return K_RAISE;			}
+enable			{ return K_ENABLE;			}
+disable		{ return K_DISABLE;			}
+privilege		{ return K_PRIVILEGE;			}
record			{ return K_RECORD;			}
rename			{ return K_RENAME;			}
result_oid		{ return K_RESULT_OID;		}

---------------------------(end of broadcast)---------------------------
TIP 3: if posting/reading through Usenet, please send an appropriate
subscribe-nomail command to majordomo@postgresql.org so that your
message can get through to the mailing list cleanly

-- 
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 853-3000
  +  If your life is a hard drive,     |  830 Blythe Avenue
  +  Christ can be your backup.        |  Drexel Hill, Pennsylvania 19026
#8Peter Eisentraut
peter_e@gmx.net
In reply to: Mark Volpe (#6)
Re: [PATCH] Re: Setuid functions

Mark Volpe writes:

This patch will implement the "ENABLE PRIVILEGE" and "DISABLE PRIVILEGE"
commands in PL/pgSQL, which, respectively, change the effective uid to that
of the function owner and back.

May I suggest better names? When I do DISABLE PRIVILEGE, do I no longer
have any privilege? Also, in SQL, the term "privilege" refers to
select/insert/update/etc. right on some table, so "enable privilege" would
be "grant". The term for user identity is "authorization", so I would
call these commands

SET AUTHORIZATION { INVOKER | DEFINER }

("invoker" and "definer" are part of the SQL CREATE FUNCTION syntax) and
the default would be invoker.

--
Peter Eisentraut peter_e@gmx.net http://funkturm.homeip.net/~peter

#9Peter Eisentraut
peter_e@gmx.net
In reply to: Mark Volpe (#1)
Re: Setuid functions

Mark Volpe writes:

I know this topic was discussed a few months ago, but I'm wondering if any
decisions have been reached on if, how, and when setuid functions and triggers
might be implemented. If not, I have an idea to throw at it.

Add a boolean column to pg_proc, when that column is true you select a
special function class handler in fmgr_info(), that function handler will
save and restore the user id and then call the real function. See
backend/utils/fmgr/fmgr.c for the action.

--
Peter Eisentraut peter_e@gmx.net http://funkturm.homeip.net/~peter

#10Tom Lane
tgl@sss.pgh.pa.us
In reply to: Peter Eisentraut (#8)
Re: [PATCH] Re: Setuid functions

Peter Eisentraut <peter_e@gmx.net> writes:

The term for user identity is "authorization", so I would
call these commands

SET AUTHORIZATION { INVOKER | DEFINER }

I like that better, too.

Overall, the only objection I can see to doing things this way is that
we have to do it over again for each function language (eg, adding such
a thing to SQL functions is doable, but much more tedious than for
plpgsql). But it seems more flexible than the pg_proc-attribute
approach.

regards, tom lane

#11Philip Warner
pjw@rhyme.com.au
In reply to: Tom Lane (#10)
Re: [PATCH] Re: Setuid functions

At 20:47 23/06/01 -0400, Tom Lane wrote:

Peter Eisentraut <peter_e@gmx.net> writes:

The term for user identity is "authorization", so I would
call these commands

SET AUTHORIZATION { INVOKER | DEFINER }

I like that better, too.

I have not read the whole thread, but I am used to module level definitions
(which easily translate to funtion level):

Create Module ZZZ [Authorization <auth-name>]

Where <auth-name> is any valid UID or Role (we don't have roles yet). We
definitely should not limit ourselves to DEFINER or INVOKER (and obviously,
when undefined, it is CURRENT_USER).

----------------------------------------------------------------
Philip Warner | __---_____
Albatross Consulting Pty. Ltd. |----/ - \
(A.B.N. 75 008 659 498) | /(@) ______---_
Tel: (+61) 0500 83 82 81 | _________ \
Fax: (+61) 0500 83 82 82 | ___________ |
Http://www.rhyme.com.au | / \|
| --________--
PGP key available upon request, | /
and from pgp5.ai.mit.edu:11371 |/

#12Bruce Momjian
bruce@momjian.us
In reply to: Mark Volpe (#6)
Re: [PATCH] Re: Setuid functions

Removed from queue pending review.

Sorry, I have decided not to follow the SQL standard ;-) PRIVILEGE is spelled
correctly in my patch.

This patch will implement the "ENABLE PRIVILEGE" and "DISABLE PRIVILEGE"
commands in PL/pgSQL, which, respectively, change the effective uid to that
of the function owner and back. It doesn't break security (I hope). The
commands can be abbreviated as "ENABLE" and "DISABLE" for the poor saps that
have trouble with "PRIVILEGE" :) Easier than adding a setuid bit to the
catalog, no?

Apologies if the patch is not in the correct format. Apply with

patch -p1 < enable_disable.patch

in the tippety-top of the 7.1.2 tree.

Regression example:

CREATE USER sample_user;
CREATE TABLE test_log(stamp datetime);
GRANT SELECT ON test_log TO PUBLIC;

DROP FUNCTION test_enable();
CREATE FUNCTION test_enable() RETURNS boolean AS
'
DECLARE
user name;
BEGIN
user:=current_user;
RAISE NOTICE ''Username: %'', user;
ENABLE PRIVILEGE;
user:=current_user;
RAISE NOTICE ''Username: %'', user;
INSERT INTO test_log VALUES(''now''::text);
DISABLE PRIVILEGE; -- Actually unnecessary at the end of the function
RETURN TRUE;
END;
' LANGUAGE 'plpgsql';

\c - sample_user
SELECT test_enable();
SELECT * FROM test_log;

stamp
------------------------
2001-06-21 11:17:29-04

(Note current time logged into a table where sample_user could not normally
write)

Hope you will find this useful
- Mark

"Ross J. Reedstrom" wrote:

Come on, Chris, you've never heard about SQL standard LEDGE? That's
the nomenclature they chose to describe a collection of permissions:
a SHELF or LEDGE. PUBLEDGE, USERLEDGE, PRIVLEDGE. So, the last is the
PRIVATE LEDGE, reserved for the owner of the object whose access is
being determined (or was that PRIVITHEDGE? now I'm confused)

... or something. ;-) Actually, not too far from how some of the SQL92
standards docs actually seem to read, especially after falling asleep
face down on the keyboard will trying to understand them, and having
vivid dreams.

Ross (who's in the office much too late, working on budget justifications
for grants that are due tomorrow!)

---------------------------(end of broadcast)---------------------------
TIP 2: you can get off all lists at once with the unregister command
(send "unregister YourEmailAddressHere" to majordomo@postgresql.org)

diff -ur postgresql-7.1.2/src/pl/plpgsql/src/gram.y postgresql-7.1.2-patch/src/pl/plpgsql/src/gram.y
--- postgresql-7.1.2/src/pl/plpgsql/src/gram.y	Wed Jun 20 20:07:45 2001
+++ postgresql-7.1.2-patch/src/pl/plpgsql/src/gram.y	Wed Jun 20 19:48:18 2001
@@ -121,7 +121,7 @@
%type <stmts>	proc_sect, proc_stmts, stmt_else, loop_body
%type <stmt>	proc_stmt, pl_block
%type <stmt>	stmt_assign, stmt_if, stmt_loop, stmt_while, stmt_exit
-%type <stmt>	stmt_return, stmt_raise, stmt_execsql, stmt_fori
+%type <stmt>	stmt_return, stmt_raise, stmt_execsql, stmt_fori, stmt_enable, stmt_disable
%type <stmt>	stmt_fors, stmt_select, stmt_perform
%type <stmt>	stmt_dynexecute, stmt_dynfors, stmt_getdiag
@@ -164,6 +164,9 @@
%token	K_PERFORM
%token	K_ROW_COUNT
%token	K_RAISE
+%token  K_ENABLE
+%token  K_DISABLE
+%token  K_PRIVILEGE
%token	K_RECORD
%token	K_RENAME
%token	K_RESULT_OID
@@ -569,6 +572,10 @@
{ $$ = $1; }
| stmt_raise
{ $$ = $1; }
+				| stmt_enable
+						{ $$ = $1; }
+				| stmt_disable
+						{ $$ = $1; }
| stmt_execsql
{ $$ = $1; }
| stmt_dynexecute
@@ -1033,6 +1040,34 @@
$$ = (PLpgSQL_stmt *)new;
}
;
+
+stmt_enable		: K_ENABLE opt_privilege lno ';'
+				{
+					PLpgSQL_stmt_privilege *new;
+
+					new=malloc(sizeof(PLpgSQL_stmt_privilege));
+
+					new->cmd_type = PLPGSQL_STMT_ENABLE;
+					new->lineno = $3;
+
+					$$ = (PLpgSQL_stmt *)new;
+				}
+
+stmt_disable		: K_DISABLE opt_privilege lno ';'
+				{
+					PLpgSQL_stmt_privilege *new;
+
+					new=malloc(sizeof(PLpgSQL_stmt_privilege));
+
+					new->cmd_type = PLPGSQL_STMT_DISABLE;
+					new->lineno = $3;
+
+					$$ = (PLpgSQL_stmt *)new;
+				}
+
+opt_privilege :
+	| K_PRIVILEGE
+;
stmt_raise		: K_RAISE lno raise_level raise_msg raise_params ';'
{
diff -ur postgresql-7.1.2/src/pl/plpgsql/src/pl_exec.c postgresql-7.1.2-patch/src/pl/plpgsql/src/pl_exec.c
--- postgresql-7.1.2/src/pl/plpgsql/src/pl_exec.c	Wed Jun 20 20:07:45 2001
+++ postgresql-7.1.2-patch/src/pl/plpgsql/src/pl_exec.c	Wed Jun 20 19:48:18 2001
@@ -99,6 +99,8 @@
PLpgSQL_stmt_exit * stmt);
static int exec_stmt_return(PLpgSQL_execstate * estate,
PLpgSQL_stmt_return * stmt);
+static int exec_stmt_privilege(PLpgSQL_execstate * estate,
+				PLpgSQL_stmt_privilege * stmt);
static int exec_stmt_raise(PLpgSQL_execstate * estate,
PLpgSQL_stmt_raise * stmt);
static int exec_stmt_execsql(PLpgSQL_execstate * estate,
@@ -220,6 +222,12 @@
case PLPGSQL_STMT_RETURN:
stmttype = "return";
break;
+					case PLPGSQL_STMT_ENABLE:
+						stmttype = "enable";
+						break;
+					case PLPGSQL_STMT_DISABLE:
+						stmttype = "disable";
+						break;
case PLPGSQL_STMT_RAISE:
stmttype = "raise";
break;
@@ -265,6 +273,8 @@
estate.retistuple = func->fn_retistuple;
estate.retisset = func->fn_retset;
estate.exitlabel = NULL;
+	estate.save_uid = InvalidOid;
+	estate.fn_oid = func->fn_oid;

estate.found_varno = func->found_varno;
estate.ndatums = func->ndatums;
@@ -385,6 +395,9 @@
elog(ERROR, "control reaches end of function without RETURN");
}

+	if (estate.save_uid!=InvalidOid)
+		SetUserId(estate.save_uid);
+
/*
* We got a return value - process it
*/
@@ -565,6 +578,8 @@
estate.retistuple = func->fn_retistuple;
estate.retisset = func->fn_retset;
estate.exitlabel = NULL;
+	estate.save_uid = InvalidOid;
+	estate.fn_oid = func->fn_oid;

estate.found_varno = func->found_varno;
estate.ndatums = func->ndatums;
@@ -742,6 +757,9 @@
elog(ERROR, "control reaches end of trigger procedure without RETURN");
}

+ if (estate.save_uid!=InvalidOid)
+ SetUserId(estate.save_uid);
+
/*
* Check that the returned tuple structure has the same attributes,
* the relation that fired the trigger has.
@@ -986,6 +1004,11 @@
rc = exec_stmt_return(estate, (PLpgSQL_stmt_return *) stmt);
break;

+		case PLPGSQL_STMT_ENABLE:
+		case PLPGSQL_STMT_DISABLE:
+			rc = exec_stmt_privilege(estate, (PLpgSQL_stmt_privilege *) stmt);
+			break;
+
case PLPGSQL_STMT_RAISE:
rc = exec_stmt_raise(estate, (PLpgSQL_stmt_raise *) stmt);
break;
@@ -1536,6 +1559,41 @@
&(estate->rettype));
return PLPGSQL_RC_RETURN;
+}
+
+/* ----------
+ * exec_stmt_privilege          Changes user ID to/from
+ *                              that of the function owner's
+ * ----------
+ */
+
+static int
+exec_stmt_privilege(PLpgSQL_execstate * estate, PLpgSQL_stmt_privilege * stmt)
+{
+	HeapTuple procedureTuple;
+	Form_pg_proc procedureStruct;
+
+	if (stmt->cmd_type==PLPGSQL_STMT_ENABLE)
+	{
+		if (estate->save_uid!=InvalidOid) return PLPGSQL_RC_OK; /* Already enabled */
+		procedureTuple = SearchSysCache(PROCOID, ObjectIdGetDatum(estate->fn_oid), 0, 0, 0);
+
+		if (!HeapTupleIsValid(procedureTuple))
+			elog(ERROR, "exec_stmt_privilege: cache lookup failed");
+
+		procedureStruct = (Form_pg_proc) GETSTRUCT(procedureTuple);
+
+		estate->save_uid = GetUserId();
+		SetUserId(procedureStruct->proowner);
+		ReleaseSysCache(procedureTuple);
+	} else
+	{
+		if (estate->save_uid==InvalidOid) return PLPGSQL_RC_OK; /* Already disabled */
+		SetUserId(estate->save_uid);
+		estate->save_uid = InvalidOid;
+	}
+
+	return PLPGSQL_RC_OK;
}
diff -ur postgresql-7.1.2/src/pl/plpgsql/src/pl_funcs.c postgresql-7.1.2-patch/src/pl/plpgsql/src/pl_funcs.c
--- postgresql-7.1.2/src/pl/plpgsql/src/pl_funcs.c	Wed Jun 20 20:07:45 2001
+++ postgresql-7.1.2-patch/src/pl/plpgsql/src/pl_funcs.c	Wed Jun 20 20:05:14 2001
@@ -382,6 +382,7 @@
static void dump_select(PLpgSQL_stmt_select * stmt);
static void dump_exit(PLpgSQL_stmt_exit * stmt);
static void dump_return(PLpgSQL_stmt_return * stmt);
+static void dump_privilege(PLpgSQL_stmt_privilege * stmt);
static void dump_raise(PLpgSQL_stmt_raise * stmt);
static void dump_execsql(PLpgSQL_stmt_execsql * stmt);
static void dump_dynexecute(PLpgSQL_stmt_dynexecute * stmt);
@@ -435,6 +436,10 @@
case PLPGSQL_STMT_RETURN:
dump_return((PLpgSQL_stmt_return *) stmt);
break;
+		case PLPGSQL_STMT_ENABLE:
+		case PLPGSQL_STMT_DISABLE:
+			dump_privilege((PLpgSQL_stmt_privilege *) stmt);
+			break;
case PLPGSQL_STMT_RAISE:
dump_raise((PLpgSQL_stmt_raise *) stmt);
break;
@@ -647,6 +652,16 @@
dump_expr(stmt->expr);
}
printf("\n");
+}
+
+static void
+dump_privilege(PLpgSQL_stmt_privilege * stmt)
+{
+	dump_ind();
+	if (stmt->cmd_type==PLPGSQL_STMT_ENABLE)
+		printf("ENABLE PRIVILEGE\n");
+	else
+		printf("DISABLE PRIVILEGE\n");
}
static void
diff -ur postgresql-7.1.2/src/pl/plpgsql/src/plpgsql.h postgresql-7.1.2-patch/src/pl/plpgsql/src/plpgsql.h
--- postgresql-7.1.2/src/pl/plpgsql/src/plpgsql.h	Wed Jun 20 20:07:45 2001
+++ postgresql-7.1.2-patch/src/pl/plpgsql/src/plpgsql.h	Wed Jun 20 20:08:57 2001
@@ -94,7 +94,9 @@
PLPGSQL_STMT_EXECSQL,
PLPGSQL_STMT_DYNEXECUTE,
PLPGSQL_STMT_DYNFORS,
-	PLPGSQL_STMT_GETDIAG
+	PLPGSQL_STMT_GETDIAG,
+	PLPGSQL_STMT_ENABLE,
+	PLPGSQL_STMT_DISABLE
};

@@ -387,6 +389,11 @@
int retrecno;
} PLpgSQL_stmt_return;

+typedef struct
+{                               /* ENABLE/DISABLE statement */
+    int         cmd_type;
+    int         lineno;
+}           PLpgSQL_stmt_privilege;

typedef struct
{ /* RAISE statement */
@@ -464,6 +471,8 @@
int found_varno;
int ndatums;
PLpgSQL_datum **datums;
+ Oid save_uid;
+ Oid fn_oid;
} PLpgSQL_execstate;

Only in postgresql-7.1.2/src/pl/plpgsql/src: plpgsql.h.orig
Only in postgresql-7.1.2/src/pl/plpgsql/src: plpgsql.h.rej
diff -ur postgresql-7.1.2/src/pl/plpgsql/src/scan.l postgresql-7.1.2-patch/src/pl/plpgsql/src/scan.l
--- postgresql-7.1.2/src/pl/plpgsql/src/scan.l	Wed Jun 20 20:07:45 2001
+++ postgresql-7.1.2-patch/src/pl/plpgsql/src/scan.l	Wed Jun 20 19:48:18 2001
@@ -115,6 +115,9 @@
null			{ return K_NULL;			}
perform			{ return K_PERFORM;			}
raise			{ return K_RAISE;			}
+enable			{ return K_ENABLE;			}
+disable		{ return K_DISABLE;			}
+privilege		{ return K_PRIVILEGE;			}
record			{ return K_RECORD;			}
rename			{ return K_RENAME;			}
result_oid		{ return K_RESULT_OID;		}

---------------------------(end of broadcast)---------------------------
TIP 3: if posting/reading through Usenet, please send an appropriate
subscribe-nomail command to majordomo@postgresql.org so that your
message can get through to the mailing list cleanly

-- 
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 853-3000
  +  If your life is a hard drive,     |  830 Blythe Avenue
  +  Christ can be your backup.        |  Drexel Hill, Pennsylvania 19026
#13Philip Warner
pjw@rhyme.com.au
In reply to: Peter Eisentraut (#9)
Re: Setuid functions

As a matter of interest, whatever happened to Perter's & Jan's plans for
setuid functions?

From memory, Jan's proposal was shelved in favour of something Peter had
previously proposed, is that right?

----------------------------------------------------------------
Philip Warner | __---_____
Albatross Consulting Pty. Ltd. |----/ - \
(A.B.N. 75 008 659 498) | /(@) ______---_
Tel: (+61) 0500 83 82 81 | _________ \
Fax: (+61) 0500 83 82 82 | ___________ |
Http://www.rhyme.com.au | / \|
| --________--
PGP key available upon request, | /
and from pgp5.ai.mit.edu:11371 |/

#14Peter Eisentraut
peter_e@gmx.net
In reply to: Philip Warner (#13)
Re: Setuid functions

Philip Warner writes:

As a matter of interest, whatever happened to Perter's & Jan's plans for
setuid functions?

If someone wants to implement them, be my guest. I originally needed them
for fixing the RI permission problems, but they couldn't be used for that
after all.

--
Peter Eisentraut peter_e@gmx.net http://funkturm.homeip.net/~peter

#15Philip Warner
pjw@rhyme.com.au
In reply to: Peter Eisentraut (#14)
Re: Setuid functions

At 13:28 24/06/01 +0200, Peter Eisentraut wrote:

If someone wants to implement them, be my guest. I originally needed them
for fixing the RI permission problems, but they couldn't be used for that
after all.

They were part of a larger permissions overhaul that Jan proposed - IIRC,
at the time you objected, citing your prior proposal. Are you now saying
you are happy with Jan's original proposal? Or just the setuid functions?

----------------------------------------------------------------
Philip Warner | __---_____
Albatross Consulting Pty. Ltd. |----/ - \
(A.B.N. 75 008 659 498) | /(@) ______---_
Tel: (+61) 0500 83 82 81 | _________ \
Fax: (+61) 0500 83 82 82 | ___________ |
Http://www.rhyme.com.au | / \|
| --________--
PGP key available upon request, | /
and from pgp5.ai.mit.edu:11371 |/

#16Peter Eisentraut
peter_e@gmx.net
In reply to: Philip Warner (#15)
Re: Setuid functions

Philip Warner writes:

If someone wants to implement them, be my guest. I originally needed them
for fixing the RI permission problems, but they couldn't be used for that
after all.

They were part of a larger permissions overhaul that Jan proposed - IIRC,
at the time you objected, citing your prior proposal. Are you now saying
you are happy with Jan's original proposal? Or just the setuid functions?

The idea of setuid functions has surely existed much longer than that
proposal, and the implementation is more or less "obvious" for someone
knowledgeable.

The proposal as a whole was rather vague and went amiss of the goal to
become SQL compliant, IIRC.

--
Peter Eisentraut peter_e@gmx.net http://funkturm.homeip.net/~peter

#17Peter Eisentraut
peter_e@gmx.net
In reply to: Tom Lane (#10)
Re: [PATCH] Re: Setuid functions

Tom Lane writes:

Overall, the only objection I can see to doing things this way is that
we have to do it over again for each function language (eg, adding such
a thing to SQL functions is doable, but much more tedious than for
plpgsql). But it seems more flexible than the pg_proc-attribute
approach.

Both approaches could be complementary, like they are in Unix as well.

--
Peter Eisentraut peter_e@gmx.net http://funkturm.homeip.net/~peter

#18Zeugswetter Andreas SB
ZeugswetterA@wien.spardat.at
In reply to: Peter Eisentraut (#17)
AW: [PATCH] Re: Setuid functions

This patch will implement the "ENABLE PRIVILEGE" and "DISABLE PRIVILEGE"
commands in PL/pgSQL, which, respectively, change the effective uid to that
of the function owner and back. It doesn't break security (I hope). The
commands can be abbreviated as "ENABLE" and "DISABLE" for the poor saps that

Anybody else want to object to this abbreviation idea ? Seems
reading ENABLE; or DISABLE; is very hard to interpret in source code
(enable what ?) and should thus not be allowed (or allow "ENABLE PRIV").

Andreas

#19Jan Wieck
JanWieck@Yahoo.com
In reply to: Philip Warner (#15)
Re: Setuid functions

Philip Warner wrote:

At 13:28 24/06/01 +0200, Peter Eisentraut wrote:

If someone wants to implement them, be my guest. I originally needed them
for fixing the RI permission problems, but they couldn't be used for that
after all.

They were part of a larger permissions overhaul that Jan proposed - IIRC,
at the time you objected, citing your prior proposal. Are you now saying
you are happy with Jan's original proposal? Or just the setuid functions?

More a bunch of ideas instead of a fully thought through
proposal. But IIRC we wanted to attack the overhaul of the
access control system while doing schemas - no?

Jan

--

#======================================================================#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me. #
#================================================== JanWieck@Yahoo.com #

_________________________________________________________
Do You Yahoo!?
Get your free @yahoo.com address at http://mail.yahoo.com

#20Mark Volpe
volpe.mark@epa.gov
In reply to: Zeugswetter Andreas SB (#18)
Re: AW: [PATCH] Re: Setuid functions

Actually, I liked the SET AUTHORIZATION { DEFINER | INVOKER } terminology
mentioned earlier.

Mark

Zeugswetter Andreas SB wrote:

Show quoted text

This patch will implement the "ENABLE PRIVILEGE" and "DISABLE PRIVILEGE"
commands in PL/pgSQL, which, respectively, change the effective uid to that
of the function owner and back. It doesn't break security (I hope). The
commands can be abbreviated as "ENABLE" and "DISABLE" for the poor saps that

Anybody else want to object to this abbreviation idea ? Seems
reading ENABLE; or DISABLE; is very hard to interpret in source code
(enable what ?) and should thus not be allowed (or allow "ENABLE PRIV").

Andreas

---------------------------(end of broadcast)---------------------------
TIP 3: if posting/reading through Usenet, please send an appropriate
subscribe-nomail command to majordomo@postgresql.org so that your
message can get through to the mailing list cleanly

#21Tom Lane
tgl@sss.pgh.pa.us
In reply to: Zeugswetter Andreas SB (#18)
#22Zeugswetter Andreas SB
ZeugswetterA@wien.spardat.at
In reply to: Tom Lane (#21)
#23Tom Lane
tgl@sss.pgh.pa.us
In reply to: Zeugswetter Andreas SB (#22)
#24Zeugswetter Andreas SB
ZeugswetterA@wien.spardat.at
In reply to: Tom Lane (#23)
#25Tom Lane
tgl@sss.pgh.pa.us
In reply to: Zeugswetter Andreas SB (#24)
#26Peter Eisentraut
peter_e@gmx.net
In reply to: Tom Lane (#25)
#27Peter Eisentraut
peter_e@gmx.net
In reply to: Zeugswetter Andreas SB (#24)
#28Mark Volpe
volpe.mark@epa.gov
In reply to: Peter Eisentraut (#26)
#29Bruce Momjian
bruce@momjian.us
In reply to: Mark Volpe (#28)
#30Peter Eisentraut
peter_e@gmx.net
In reply to: Bruce Momjian (#29)
#31Bruce Momjian
bruce@momjian.us
In reply to: Peter Eisentraut (#30)
#32Bruce Momjian
bruce@momjian.us
In reply to: Peter Eisentraut (#30)
#33Mark Volpe
volpe.mark@epa.gov
In reply to: Bruce Momjian (#32)
#34Bruce Momjian
bruce@momjian.us
In reply to: Mark Volpe (#33)
#35Mark Volpe
volpe.mark@epa.gov
In reply to: Bruce Momjian (#34)
#36Peter Eisentraut
peter_e@gmx.net
In reply to: Bruce Momjian (#34)
#37Mark Volpe
volpe.mark@epa.gov
In reply to: Peter Eisentraut (#36)
#38Peter Eisentraut
peter_e@gmx.net
In reply to: Mark Volpe (#37)
#39Bruce Momjian
bruce@momjian.us
In reply to: Peter Eisentraut (#38)
#40Mark Volpe
volpe.mark@epa.gov
In reply to: Peter Eisentraut (#38)
#41Tom Lane
tgl@sss.pgh.pa.us
In reply to: Mark Volpe (#40)
#42Mark Volpe
volpe.mark@epa.gov
In reply to: Peter Eisentraut (#38)