agora inbox for [email protected]  

help / color / mirror / Atom feed

[PATCH v9 2/2] wip: report nanoseconds in pg_test_timing
173+ messages / 2 participants
[nested] [flat]

* [PATCH v9 2/2] wip: report nanoseconds in pg_test_timing
@ 2023-01-16 19:19 Andres Freund <[email protected]>
  0 siblings, 0 replies; 173+ messages in thread

From: Andres Freund @ 2023-01-16 19:19 UTC (permalink / raw)

This commit also updates pg_test_timing's documentation:

- compare EXPLAIN (ANALYZE, TIMING ON/OFF) instead of comparing performance of
  of statement with/without EXPLAIN ANALYZE
- explain the 2x overhead (due to two timestamp acquisitions per row)
- remove old section about old versions of linux - I couldn't update the
  numbers, and it's old enough nobody would care
---
 src/bin/pg_test_timing/pg_test_timing.c |  74 +++++++++------
 doc/src/sgml/ref/pgtesttiming.sgml      | 117 ++++++++++--------------
 2 files changed, 95 insertions(+), 96 deletions(-)

diff --git a/src/bin/pg_test_timing/pg_test_timing.c b/src/bin/pg_test_timing/pg_test_timing.c
index c29d6f87629..e20718669a5 100644
--- a/src/bin/pg_test_timing/pg_test_timing.c
+++ b/src/bin/pg_test_timing/pg_test_timing.c
@@ -19,8 +19,8 @@ static void handle_args(int argc, char *argv[]);
 static uint64 test_timing(unsigned int duration);
 static void output(uint64 loop_count);
 
-/* record duration in powers of 2 microseconds */
-long long int histogram[32];
+/* record duration in powers of 2 nanoseconds */
+uint64		histogram[64];
 
 int
 main(int argc, char *argv[])
@@ -121,35 +121,48 @@ handle_args(int argc, char *argv[])
 static uint64
 test_timing(unsigned int duration)
 {
-	uint64		total_time;
-	int64		time_elapsed = 0;
 	uint64		loop_count = 0;
-	uint64		prev,
-				cur;
+	instr_time	until_time,
+				total_time;
 	instr_time	start_time,
-				end_time,
-				temp;
-
-	total_time = duration > 0 ? duration * INT64CONST(1000000) : 0;
+				end_time;
+	instr_time	cur;
 
 	INSTR_TIME_SET_CURRENT(start_time);
-	cur = INSTR_TIME_GET_MICROSEC(start_time);
 
-	while (time_elapsed < total_time)
+	/*
+	 * To reduce loop overhead, check loop condition in instr_time domain.
+	 */
+	INSTR_TIME_SET_SECONDS(total_time, duration);
+	until_time = start_time;
+	INSTR_TIME_ADD(until_time, total_time);
+
+	cur = start_time;
+
+	while (INSTR_TIME_IS_LT(cur, until_time))
 	{
-		int32		diff,
-					bits = 0;
+		instr_time	temp;
+		instr_time	prev;
+		int64		diff;
+		int32		bits = 0;
 
 		prev = cur;
-		INSTR_TIME_SET_CURRENT(temp);
-		cur = INSTR_TIME_GET_MICROSEC(temp);
-		diff = cur - prev;
+		INSTR_TIME_SET_CURRENT(cur);
+		temp = cur;
+		INSTR_TIME_SUBTRACT(temp, prev);
+		diff = INSTR_TIME_GET_NANOSEC(temp);
 
 		/* Did time go backwards? */
-		if (diff < 0)
+		if (unlikely(diff <= 0))
 		{
+			/* can't do anything with that measurement */
+			if (diff == 0)
+			{
+				loop_count++;
+				continue;
+			}
 			fprintf(stderr, _("Detected clock going backwards in time.\n"));
-			fprintf(stderr, _("Time warp: %d ms\n"), diff);
+			fprintf(stderr, _("Time warp: %lld ns\n"), (long long) diff);
 			exit(1);
 		}
 
@@ -164,8 +177,6 @@ test_timing(unsigned int duration)
 		histogram[bits]++;
 
 		loop_count++;
-		INSTR_TIME_SUBTRACT(temp, start_time);
-		time_elapsed = INSTR_TIME_GET_MICROSEC(temp);
 	}
 
 	INSTR_TIME_SET_CURRENT(end_time);
@@ -173,7 +184,7 @@ test_timing(unsigned int duration)
 	INSTR_TIME_SUBTRACT(end_time, start_time);
 
 	printf(_("Per loop time including overhead: %0.2f ns\n"),
-		   INSTR_TIME_GET_DOUBLE(end_time) * 1e9 / loop_count);
+		   (INSTR_TIME_GET_DOUBLE(end_time) * NS_PER_S) / loop_count);
 
 	return loop_count;
 }
@@ -181,9 +192,10 @@ test_timing(unsigned int duration)
 static void
 output(uint64 loop_count)
 {
-	int64		max_bit = 31,
+	int64		low_bit = 0,
+				max_bit = 63,
 				i;
-	char	   *header1 = _("< us");
+	char	   *header1 = _("< ns");
 	char	   *header2 = /* xgettext:no-c-format */ _("% of total");
 	char	   *header3 = _("count");
 	int			len1 = strlen(header1);
@@ -194,15 +206,19 @@ output(uint64 loop_count)
 	while (max_bit > 0 && histogram[max_bit] == 0)
 		max_bit--;
 
+	/* find lowest bit value */
+	while (low_bit < max_bit && histogram[low_bit] == 0)
+		low_bit++;
+
 	printf(_("Histogram of timing durations:\n"));
 	printf("%*s   %*s %*s\n",
-		   Max(6, len1), header1,
+		   Max(9, len1), header1,
 		   Max(10, len2), header2,
 		   Max(10, len3), header3);
 
-	for (i = 0; i <= max_bit; i++)
-		printf("%*ld    %*.5f %*lld\n",
-			   Max(6, len1), 1l << i,
+	for (i = low_bit; i <= max_bit; i++)
+		printf("%*ld    %*.5f %*llu\n",
+			   Max(9, len1), 1l << i,
 			   Max(10, len2) - 1, (double) histogram[i] * 100 / loop_count,
-			   Max(10, len3), histogram[i]);
+			   Max(10, len3), (long long unsigned) histogram[i]);
 }
diff --git a/doc/src/sgml/ref/pgtesttiming.sgml b/doc/src/sgml/ref/pgtesttiming.sgml
index a5eb3aa25e0..7e0266cf58b 100644
--- a/doc/src/sgml/ref/pgtesttiming.sgml
+++ b/doc/src/sgml/ref/pgtesttiming.sgml
@@ -93,28 +93,34 @@ PostgreSQL documentation
 
   <para>
    Good results will show most (>90%) individual timing calls take less than
-   one microsecond. Average per loop overhead will be even lower, below 100
-   nanoseconds. This example from an Intel i7-860 system using a TSC clock
-   source shows excellent performance:
+   one microsecond (1000 nanoseconds). Average per loop overhead will be even
+   lower, below 100 nanoseconds. This example from an Intel i9-9880H system
+   using a TSC clock source shows excellent performance:
 
 <screen><![CDATA[
 Testing timing overhead for 3 seconds.
-Per loop time including overhead: 35.96 ns
+Per loop time including overhead: 13.74 ns
 Histogram of timing durations:
-  < us   % of total      count
-     1     96.40465   80435604
-     2      3.59518    2999652
-     4      0.00015        126
-     8      0.00002         13
-    16      0.00000          2
+     < ns   % of total      count
+       16     97.00221  211857215
+       32      2.99555    6542412
+       64      0.00115       2505
+      128      0.00035        759
+      256      0.00004         78
+      512      0.00000          3
+     1024      0.00000          4
+     2048      0.00034        732
+     4096      0.00000          6
+     8192      0.00000          8
+    16384      0.00019        409
+    32768      0.00018        403
+    65536      0.00000          1
 ]]></screen>
   </para>
 
   <para>
-   Note that different units are used for the per loop time than the
-   histogram. The loop can have resolution within a few nanoseconds (ns),
-   while the individual timing calls can only resolve down to one microsecond
-   (us).
+   Note that the accuracy of the histogram entries may be lower than the
+   per loop time.
   </para>
 
  </refsect2>
@@ -125,24 +131,25 @@ Histogram of timing durations:
    When the query executor is running a statement using
    <command>EXPLAIN ANALYZE</command>, individual operations are timed as well
    as showing a summary.  The overhead of your system can be checked by
-   counting rows with the <application>psql</application> program:
+   disabling the per-row timing, using the <literal>TIMING OFF</literal>
+   option:
 
 <screen>
-CREATE TABLE t AS SELECT * FROM generate_series(1,100000);
-\timing
-SELECT COUNT(*) FROM t;
-EXPLAIN ANALYZE SELECT COUNT(*) FROM t;
+CREATE TABLE t AS SELECT * FROM generate_series(1, 100000);
+EXPLAIN (ANALYZE, TIMING OFF) SELECT COUNT(*) FROM t;
+EXPLAIN (ANALYZE, TIMING ON) SELECT COUNT(*) FROM t;
 </screen>
   </para>
 
   <para>
-   The i7-860 system measured runs the count query in 9.8 ms while
-   the <command>EXPLAIN ANALYZE</command> version takes 16.6 ms, each
-   processing just over 100,000 rows.  That 6.8 ms difference means the timing
-   overhead per row is 68 ns, about twice what pg_test_timing estimated it
-   would be.  Even that relatively small amount of overhead is making the fully
-   timed count statement take almost 70% longer.  On more substantial queries,
-   the timing overhead would be less problematic.
+   The i9-9880H system measured shows an execution time of 4.116 ms for the
+   <literal>TIMING OFF</literal> query, and 6.965 ms for the
+   <literal>TIMING ON</literal>, each processing 100,000 rows.
+
+   That 2.849 ms difference means the timing overhead per row is 28 ns.  As
+   <literal>TIMING ON</literal> measures timestamps twice per row returned by
+   an executor node, the overhead is very close to what pg_test_timing
+   estimated it would be.
   </para>
 
  </refsect2>
@@ -157,28 +164,31 @@ EXPLAIN ANALYZE SELECT COUNT(*) FROM t;
 
 <screen><![CDATA[
 # cat /sys/devices/system/clocksource/clocksource0/available_clocksource
-tsc hpet acpi_pm
+tsc acpi_pm
 # echo acpi_pm > /sys/devices/system/clocksource/clocksource0/current_clocksource
 # pg_test_timing
-Per loop time including overhead: 722.92 ns
+Testing timing overhead for 3 seconds.
+Per loop time including overhead: 708.58 ns
 Histogram of timing durations:
-  < us   % of total      count
-     1     27.84870    1155682
-     2     72.05956    2990371
-     4      0.07810       3241
-     8      0.01357        563
-    16      0.00007          3
+     < ns   % of total      count
+     1024     99.79796    4225270
+     2048      0.15560       6588
+     4096      0.00035         15
+     8192      0.01738        736
+    16384      0.01679        711
+    32768      0.01190        504
 ]]></screen>
   </para>
 
   <para>
    In this configuration, the sample <command>EXPLAIN ANALYZE</command> above
-   takes 115.9 ms.  That's 1061 ns of timing overhead, again a small multiple
-   of what's measured directly by this utility.  That much timing overhead
-   means the actual query itself is only taking a tiny fraction of the
-   accounted for time, most of it is being consumed in overhead instead.  In
-   this configuration, any <command>EXPLAIN ANALYZE</command> totals involving
-   many timed operations would be inflated significantly by timing overhead.
+   shows an execution time of 148.7 ms.  That's 1392 ns of per-row timing
+   overhead.  Taking the two timestamps per row into account, that's again
+   close to what pg_test_timing estimated.  That much timing overhead means
+   the actual query itself is only taking a tiny fraction of the accounted for
+   time, most of it is being consumed in overhead instead.  In this
+   configuration, any <command>EXPLAIN ANALYZE</command> totals involving many
+   timed operations would be inflated significantly by timing overhead.
   </para>
 
   <para>
@@ -196,33 +206,6 @@ kern.timecounter.hardware: ACPI-fast -> TSC
 </screen>
   </para>
 
-  <para>
-   Other systems may only allow setting the time source on boot.  On older
-   Linux systems the "clock" kernel setting is the only way to make this sort
-   of change.  And even on some more recent ones, the only option you'll see
-   for a clock source is "jiffies".  Jiffies are the older Linux software clock
-   implementation, which can have good resolution when it's backed by fast
-   enough timing hardware, as in this example:
-
-<screen><![CDATA[
-$ cat /sys/devices/system/clocksource/clocksource0/available_clocksource
-jiffies
-$ dmesg | grep time.c
-time.c: Using 3.579545 MHz WALL PM GTOD PIT/TSC timer.
-time.c: Detected 2400.153 MHz processor.
-$ pg_test_timing
-Testing timing overhead for 3 seconds.
-Per timing duration including loop overhead: 97.75 ns
-Histogram of timing durations:
-  < us   % of total      count
-     1     90.23734   27694571
-     2      9.75277    2993204
-     4      0.00981       3010
-     8      0.00007         22
-    16      0.00000          1
-    32      0.00000          1
-]]></screen></para>
-
  </refsect2>
 
  <refsect2>
-- 
2.38.0


--2e47imhqa2hvpvyc--





^ permalink  raw  reply  [nested|flat] 173+ messages in thread

* [PATCH v1] Add RoleNameGetOid() with invalidation-based retry loop for DropRole()/GrantRole()
@ 2026-07-04 04:47 Bertrand Drouvot <[email protected]>
  0 siblings, 0 replies; 173+ messages in thread

From: Bertrand Drouvot @ 2026-07-04 04:47 UTC (permalink / raw)

DropRole() and GrantRole() resolve the role name to an OID before acquiring
LockSharedObject() on the role. A concurrent session that commits a DROP ROLE
between the read and the lock acquisition leaves the first session acting on a
stale OID.

This commit fixes the races by using the same approach as RangeVarGetRelidExtended():
It encapsulates name resolution, permission checking (via a caller-supplied callback),
and lock acquisition inside a retry loop driven by SharedInvalidMessageCounter.
If invalidation messages arrive between name resolution and locking, indicating
concurrent DDL, the function retries.

The lock is kept across retries and only released if the name resolves to a
different OID on the next iteration.

Two callbacks are provided:
- RoleNameCallbackForDropRole(): checks current/session user, superuser attribute,
and ADMIN OPTION privilege before locking. This is similar to what DropRole() is
currently doing before LockSharedObject().
- RoleNameCallbackForGrantRole(): calls check_role_membership_authorization() to
verify the current user can grant/revoke membership. This is similar to what GrantRole()
is currently doing before calling AddRoleMems()/DelRoleMems().

DropRole() and GrantRole() now call RoleNameGetOid() with appropriate lock
levels.

AlterRole() does not need the fix because it calls CatalogTupleUpdate() on the
pg_authid tuple before AddRoleMems(), which blocks a concurrent DROP ROLE.

Author: Bertrand Drouvot <[email protected]>
Reviewed-by:
Discussion:
---
 src/backend/commands/user.c | 246 ++++++++++++++++++++++++++----------
 src/include/commands/user.h |   9 ++
 2 files changed, 191 insertions(+), 64 deletions(-)
  95.5% src/backend/commands/
   4.4% src/include/commands/

diff --git a/src/backend/commands/user.c b/src/backend/commands/user.c
index be11c49f919..6d8e2fa8813 100644
--- a/src/backend/commands/user.c
+++ b/src/backend/commands/user.c
@@ -34,6 +34,7 @@
 #include "miscadmin.h"
 #include "port/pg_bitutils.h"
 #include "storage/lmgr.h"
+#include "storage/sinval.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
 #include "utils/catcache.h"
@@ -116,6 +117,10 @@ static void plan_recursive_revoke(CatCList *memlist,
 								  bool revoke_admin_option_only,
 								  DropBehavior behavior);
 static void InitGrantRoleOptions(GrantRoleOptions *popt);
+static void RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+static void RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+										 Oid oldroleid, void *callback_arg);
 
 
 /* Check if current user has createrole privileges */
@@ -126,6 +131,92 @@ have_createrole_privilege(void)
 }
 
 
+/*
+ * RoleNameGetOid
+ *		Given a role name, look up its OID, lock it, and return the OID.
+ *
+ * This follows the same pattern as RangeVarGetRelidExtended():
+ * name resolution, permission check (via callback), and lock acquisition are
+ * performed inside a retry loop. If invalidation messages arrive during the
+ * process (indicating concurrent DDL), we retry to ensure the name still
+ * resolves to the same OID.
+ *
+ * The callback is invoked before locking, giving callers a chance to check
+ * permissions. It receives the current rolename, the resolved OID, the
+ * previous OID (InvalidOid on first iteration), and a caller-supplied arg.
+ * If the callback raises an error, the function aborts without locking.
+ *
+ * If missing_ok is true and the role does not exist, returns InvalidOid.
+ * Otherwise, raises an error.
+ */
+Oid
+RoleNameGetOid(const char *rolename, LOCKMODE lockmode, bool missing_ok,
+			   RoleNameGetOidCallback callback, void *callback_arg)
+{
+	uint64		inval_count;
+	Oid			roleid;
+	Oid			oldroleid = InvalidOid;
+	bool		retry = false;
+
+	for (;;)
+	{
+		/*
+		 * Remember the current invalidation count so we can detect concurrent
+		 * DDL after locking.
+		 */
+		inval_count = SharedInvalidMessageCounter;
+
+		/* Look up the role name */
+		roleid = get_role_oid(rolename, true);
+
+		if (!OidIsValid(roleid))
+		{
+			if (!missing_ok)
+				ereport(ERROR,
+						(errcode(ERRCODE_UNDEFINED_OBJECT),
+						 errmsg("role \"%s\" does not exist", rolename)));
+			return InvalidOid;
+		}
+
+		/*
+		 * Invoke caller-supplied callback before locking. This is a good
+		 * place to check permissions: we haven't taken the lock yet, but we
+		 * know the OID we intend to lock. If concurrent DDL changes things,
+		 * the callback will be invoked again on the next iteration.
+		 */
+		if (callback)
+			callback(rolename, roleid, oldroleid, callback_arg);
+
+		/*
+		 * If upon retry we get back the same OID, the invalidation messages
+		 * did not change the final answer. So we're done.
+		 *
+		 * If we got a different OID, we've locked the role that used to have
+		 * this name rather than the one that does now. Release the old lock.
+		 */
+		if (retry)
+		{
+			if (roleid == oldroleid)
+				break;
+			UnlockSharedObject(AuthIdRelationId, oldroleid, 0, lockmode);
+		}
+
+		/* Lock the role */
+		LockSharedObject(AuthIdRelationId, roleid, 0, lockmode);
+
+		/* If no invalidation messages were processed, we're done */
+		if (inval_count == SharedInvalidMessageCounter)
+			break;
+
+		/* Something may have changed, retry */
+		retry = true;
+		oldroleid = roleid;
+	}
+
+	return roleid;
+}
+
+
 /*
  * CREATE ROLE
  */
@@ -1090,6 +1181,59 @@ AlterRoleSet(AlterRoleSetStmt *stmt)
 }
 
 
+/*
+ * Before acquiring a role lock for DROP ROLE, check that the role is not the
+ * current/session user and that the caller has sufficient privileges to drop it.
+ */
+static void
+RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+							Oid oldroleid, void *callback_arg)
+{
+	HeapTuple	tuple;
+	Form_pg_authid roleform;
+
+	if (roleid == GetUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetOuterUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetSessionUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("session user cannot be dropped")));
+
+	tuple = SearchSysCache1(AUTHOID, ObjectIdGetDatum(roleid));
+	if (!HeapTupleIsValid(tuple))
+		ereport(ERROR,
+				(errcode(ERRCODE_UNDEFINED_OBJECT),
+				 errmsg("role \"%s\" does not exist", rolename)));
+
+	roleform = (Form_pg_authid) GETSTRUCT(tuple);
+
+	/*
+	 * For safety's sake, we allow createrole holders to drop ordinary roles
+	 * but not superuser roles, and only if they also have ADMIN OPTION.
+	 */
+	if (roleform->rolsuper && !superuser())
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
+						   "SUPERUSER", "SUPERUSER")));
+	if (!is_admin_of_role(GetUserId(), roleid))
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
+						   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
+
+	ReleaseSysCache(tuple);
+}
+
+
 /*
  * DROP ROLE
  */
@@ -1119,9 +1263,7 @@ DropRole(DropRoleStmt *stmt)
 	{
 		RoleSpec   *rolspec = lfirst(item);
 		char	   *role;
-		HeapTuple	tuple,
-					tmp_tuple;
-		Form_pg_authid roleform;
+		HeapTuple	tmp_tuple;
 		ScanKeyData scankey;
 		SysScanDesc sscan;
 		Oid			roleid;
@@ -1132,71 +1274,27 @@ DropRole(DropRoleStmt *stmt)
 					 errmsg("cannot use special role specifier in DROP ROLE")));
 		role = rolspec->rolename;
 
-		tuple = SearchSysCache1(AUTHNAME, PointerGetDatum(role));
-		if (!HeapTupleIsValid(tuple))
-		{
-			if (!stmt->missing_ok)
-			{
-				ereport(ERROR,
-						(errcode(ERRCODE_UNDEFINED_OBJECT),
-						 errmsg("role \"%s\" does not exist", role)));
-			}
-			else
-			{
-				ereport(NOTICE,
-						(errmsg("role \"%s\" does not exist, skipping",
-								role)));
-			}
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP or ALTER commits between name
+		 * resolution and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(role, AccessExclusiveLock, stmt->missing_ok,
+								RoleNameCallbackForDropRole, NULL);
 
+		if (!OidIsValid(roleid))
+		{
+			/* missing_ok case: role doesn't exist */
+			ereport(NOTICE,
+					(errmsg("role \"%s\" does not exist, skipping",
+							role)));
 			continue;
 		}
 
-		roleform = (Form_pg_authid) GETSTRUCT(tuple);
-		roleid = roleform->oid;
-
-		if (roleid == GetUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetOuterUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetSessionUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("session user cannot be dropped")));
-
-		/*
-		 * For safety's sake, we allow createrole holders to drop ordinary
-		 * roles but not superuser roles, and only if they also have ADMIN
-		 * OPTION.
-		 */
-		if (roleform->rolsuper && !superuser())
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
-							   "SUPERUSER", "SUPERUSER")));
-		if (!is_admin_of_role(GetUserId(), roleid))
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
-							   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
-
 		/* DROP hook for the role being removed */
 		InvokeObjectDropHook(AuthIdRelationId, roleid, 0);
 
-		/* Don't leak the syscache tuple */
-		ReleaseSysCache(tuple);
-
-		/*
-		 * Lock the role, so nobody can add dependencies to her while we drop
-		 * her.  We keep the lock until the end of transaction.
-		 */
-		LockSharedObject(AuthIdRelationId, roleid, 0, AccessExclusiveLock);
-
 		/*
 		 * If there is a pg_auth_members entry that has one of the roles to be
 		 * dropped as the roleid or member, it should be silently removed, but
@@ -1484,6 +1582,21 @@ RenameRole(const char *oldname, const char *newname)
 	return address;
 }
 
+/*
+ * Before acquiring a role lock for GRANT/REVOKE, check that the current user
+ * has authorization to grant/revoke membership in the specified role.
+ */
+static void
+RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+							 Oid oldroleid, void *callback_arg)
+{
+	bool		is_grant = *((bool *) callback_arg);
+	Oid			currentUserId = GetUserId();
+
+	check_role_membership_authorization(currentUserId, roleid, is_grant);
+}
+
+
 /*
  * GrantRoleStmt
  *
@@ -1568,9 +1681,14 @@ GrantRole(ParseState *pstate, GrantRoleStmt *stmt)
 					(errcode(ERRCODE_INVALID_GRANT_OPERATION),
 					 errmsg("column names cannot be included in GRANT/REVOKE ROLE")));
 
-		roleid = get_role_oid(rolename, false);
-		check_role_membership_authorization(currentUserId,
-											roleid, stmt->is_grant);
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP commits between name resolution
+		 * and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(rolename, ShareUpdateExclusiveLock, false,
+								RoleNameCallbackForGrantRole, &stmt->is_grant);
 		if (stmt->is_grant)
 			AddRoleMems(currentUserId, rolename, roleid,
 						stmt->grantee_roles, grantee_ids,
diff --git a/src/include/commands/user.h b/src/include/commands/user.h
index 97dcb93791b..17263452250 100644
--- a/src/include/commands/user.h
+++ b/src/include/commands/user.h
@@ -21,6 +21,15 @@
 extern PGDLLIMPORT int Password_encryption; /* values from enum PasswordType */
 extern PGDLLIMPORT char *createrole_self_grant;
 
+/* Callback for RoleNameGetOid, invoked after name resolution but before locking */
+typedef void (*RoleNameGetOidCallback) (const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+
+extern Oid	RoleNameGetOid(const char *rolename, LOCKMODE lockmode,
+						   bool missing_ok,
+						   RoleNameGetOidCallback callback,
+						   void *callback_arg);
+
 /* Hook to check passwords in CreateRole() and AlterRole() */
 typedef void (*check_password_hook_type) (const char *username, const char *shadow_pass, PasswordType password_type, Datum validuntil_time, bool validuntil_null);
 
-- 
2.34.1


--xlWYagR37YwJXkF8--





^ permalink  raw  reply  [nested|flat] 173+ messages in thread

* [PATCH v1] Add RoleNameGetOid() with invalidation-based retry loop for DropRole()/GrantRole()
@ 2026-07-04 04:47 Bertrand Drouvot <[email protected]>
  0 siblings, 0 replies; 173+ messages in thread

From: Bertrand Drouvot @ 2026-07-04 04:47 UTC (permalink / raw)

DropRole() and GrantRole() resolve the role name to an OID before acquiring
LockSharedObject() on the role. A concurrent session that commits a DROP ROLE
between the read and the lock acquisition leaves the first session acting on a
stale OID.

This commit fixes the races by using the same approach as RangeVarGetRelidExtended():
It encapsulates name resolution, permission checking (via a caller-supplied callback),
and lock acquisition inside a retry loop driven by SharedInvalidMessageCounter.
If invalidation messages arrive between name resolution and locking, indicating
concurrent DDL, the function retries.

The lock is kept across retries and only released if the name resolves to a
different OID on the next iteration.

Two callbacks are provided:
- RoleNameCallbackForDropRole(): checks current/session user, superuser attribute,
and ADMIN OPTION privilege before locking. This is similar to what DropRole() is
currently doing before LockSharedObject().
- RoleNameCallbackForGrantRole(): calls check_role_membership_authorization() to
verify the current user can grant/revoke membership. This is similar to what GrantRole()
is currently doing before calling AddRoleMems()/DelRoleMems().

DropRole() and GrantRole() now call RoleNameGetOid() with appropriate lock
levels.

AlterRole() does not need the fix because it calls CatalogTupleUpdate() on the
pg_authid tuple before AddRoleMems(), which blocks a concurrent DROP ROLE.

Author: Bertrand Drouvot <[email protected]>
Reviewed-by:
Discussion:
---
 src/backend/commands/user.c | 246 ++++++++++++++++++++++++++----------
 src/include/commands/user.h |   9 ++
 2 files changed, 191 insertions(+), 64 deletions(-)
  95.5% src/backend/commands/
   4.4% src/include/commands/

diff --git a/src/backend/commands/user.c b/src/backend/commands/user.c
index be11c49f919..6d8e2fa8813 100644
--- a/src/backend/commands/user.c
+++ b/src/backend/commands/user.c
@@ -34,6 +34,7 @@
 #include "miscadmin.h"
 #include "port/pg_bitutils.h"
 #include "storage/lmgr.h"
+#include "storage/sinval.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
 #include "utils/catcache.h"
@@ -116,6 +117,10 @@ static void plan_recursive_revoke(CatCList *memlist,
 								  bool revoke_admin_option_only,
 								  DropBehavior behavior);
 static void InitGrantRoleOptions(GrantRoleOptions *popt);
+static void RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+static void RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+										 Oid oldroleid, void *callback_arg);
 
 
 /* Check if current user has createrole privileges */
@@ -126,6 +131,92 @@ have_createrole_privilege(void)
 }
 
 
+/*
+ * RoleNameGetOid
+ *		Given a role name, look up its OID, lock it, and return the OID.
+ *
+ * This follows the same pattern as RangeVarGetRelidExtended():
+ * name resolution, permission check (via callback), and lock acquisition are
+ * performed inside a retry loop. If invalidation messages arrive during the
+ * process (indicating concurrent DDL), we retry to ensure the name still
+ * resolves to the same OID.
+ *
+ * The callback is invoked before locking, giving callers a chance to check
+ * permissions. It receives the current rolename, the resolved OID, the
+ * previous OID (InvalidOid on first iteration), and a caller-supplied arg.
+ * If the callback raises an error, the function aborts without locking.
+ *
+ * If missing_ok is true and the role does not exist, returns InvalidOid.
+ * Otherwise, raises an error.
+ */
+Oid
+RoleNameGetOid(const char *rolename, LOCKMODE lockmode, bool missing_ok,
+			   RoleNameGetOidCallback callback, void *callback_arg)
+{
+	uint64		inval_count;
+	Oid			roleid;
+	Oid			oldroleid = InvalidOid;
+	bool		retry = false;
+
+	for (;;)
+	{
+		/*
+		 * Remember the current invalidation count so we can detect concurrent
+		 * DDL after locking.
+		 */
+		inval_count = SharedInvalidMessageCounter;
+
+		/* Look up the role name */
+		roleid = get_role_oid(rolename, true);
+
+		if (!OidIsValid(roleid))
+		{
+			if (!missing_ok)
+				ereport(ERROR,
+						(errcode(ERRCODE_UNDEFINED_OBJECT),
+						 errmsg("role \"%s\" does not exist", rolename)));
+			return InvalidOid;
+		}
+
+		/*
+		 * Invoke caller-supplied callback before locking. This is a good
+		 * place to check permissions: we haven't taken the lock yet, but we
+		 * know the OID we intend to lock. If concurrent DDL changes things,
+		 * the callback will be invoked again on the next iteration.
+		 */
+		if (callback)
+			callback(rolename, roleid, oldroleid, callback_arg);
+
+		/*
+		 * If upon retry we get back the same OID, the invalidation messages
+		 * did not change the final answer. So we're done.
+		 *
+		 * If we got a different OID, we've locked the role that used to have
+		 * this name rather than the one that does now. Release the old lock.
+		 */
+		if (retry)
+		{
+			if (roleid == oldroleid)
+				break;
+			UnlockSharedObject(AuthIdRelationId, oldroleid, 0, lockmode);
+		}
+
+		/* Lock the role */
+		LockSharedObject(AuthIdRelationId, roleid, 0, lockmode);
+
+		/* If no invalidation messages were processed, we're done */
+		if (inval_count == SharedInvalidMessageCounter)
+			break;
+
+		/* Something may have changed, retry */
+		retry = true;
+		oldroleid = roleid;
+	}
+
+	return roleid;
+}
+
+
 /*
  * CREATE ROLE
  */
@@ -1090,6 +1181,59 @@ AlterRoleSet(AlterRoleSetStmt *stmt)
 }
 
 
+/*
+ * Before acquiring a role lock for DROP ROLE, check that the role is not the
+ * current/session user and that the caller has sufficient privileges to drop it.
+ */
+static void
+RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+							Oid oldroleid, void *callback_arg)
+{
+	HeapTuple	tuple;
+	Form_pg_authid roleform;
+
+	if (roleid == GetUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetOuterUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetSessionUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("session user cannot be dropped")));
+
+	tuple = SearchSysCache1(AUTHOID, ObjectIdGetDatum(roleid));
+	if (!HeapTupleIsValid(tuple))
+		ereport(ERROR,
+				(errcode(ERRCODE_UNDEFINED_OBJECT),
+				 errmsg("role \"%s\" does not exist", rolename)));
+
+	roleform = (Form_pg_authid) GETSTRUCT(tuple);
+
+	/*
+	 * For safety's sake, we allow createrole holders to drop ordinary roles
+	 * but not superuser roles, and only if they also have ADMIN OPTION.
+	 */
+	if (roleform->rolsuper && !superuser())
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
+						   "SUPERUSER", "SUPERUSER")));
+	if (!is_admin_of_role(GetUserId(), roleid))
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
+						   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
+
+	ReleaseSysCache(tuple);
+}
+
+
 /*
  * DROP ROLE
  */
@@ -1119,9 +1263,7 @@ DropRole(DropRoleStmt *stmt)
 	{
 		RoleSpec   *rolspec = lfirst(item);
 		char	   *role;
-		HeapTuple	tuple,
-					tmp_tuple;
-		Form_pg_authid roleform;
+		HeapTuple	tmp_tuple;
 		ScanKeyData scankey;
 		SysScanDesc sscan;
 		Oid			roleid;
@@ -1132,71 +1274,27 @@ DropRole(DropRoleStmt *stmt)
 					 errmsg("cannot use special role specifier in DROP ROLE")));
 		role = rolspec->rolename;
 
-		tuple = SearchSysCache1(AUTHNAME, PointerGetDatum(role));
-		if (!HeapTupleIsValid(tuple))
-		{
-			if (!stmt->missing_ok)
-			{
-				ereport(ERROR,
-						(errcode(ERRCODE_UNDEFINED_OBJECT),
-						 errmsg("role \"%s\" does not exist", role)));
-			}
-			else
-			{
-				ereport(NOTICE,
-						(errmsg("role \"%s\" does not exist, skipping",
-								role)));
-			}
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP or ALTER commits between name
+		 * resolution and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(role, AccessExclusiveLock, stmt->missing_ok,
+								RoleNameCallbackForDropRole, NULL);
 
+		if (!OidIsValid(roleid))
+		{
+			/* missing_ok case: role doesn't exist */
+			ereport(NOTICE,
+					(errmsg("role \"%s\" does not exist, skipping",
+							role)));
 			continue;
 		}
 
-		roleform = (Form_pg_authid) GETSTRUCT(tuple);
-		roleid = roleform->oid;
-
-		if (roleid == GetUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetOuterUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetSessionUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("session user cannot be dropped")));
-
-		/*
-		 * For safety's sake, we allow createrole holders to drop ordinary
-		 * roles but not superuser roles, and only if they also have ADMIN
-		 * OPTION.
-		 */
-		if (roleform->rolsuper && !superuser())
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
-							   "SUPERUSER", "SUPERUSER")));
-		if (!is_admin_of_role(GetUserId(), roleid))
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
-							   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
-
 		/* DROP hook for the role being removed */
 		InvokeObjectDropHook(AuthIdRelationId, roleid, 0);
 
-		/* Don't leak the syscache tuple */
-		ReleaseSysCache(tuple);
-
-		/*
-		 * Lock the role, so nobody can add dependencies to her while we drop
-		 * her.  We keep the lock until the end of transaction.
-		 */
-		LockSharedObject(AuthIdRelationId, roleid, 0, AccessExclusiveLock);
-
 		/*
 		 * If there is a pg_auth_members entry that has one of the roles to be
 		 * dropped as the roleid or member, it should be silently removed, but
@@ -1484,6 +1582,21 @@ RenameRole(const char *oldname, const char *newname)
 	return address;
 }
 
+/*
+ * Before acquiring a role lock for GRANT/REVOKE, check that the current user
+ * has authorization to grant/revoke membership in the specified role.
+ */
+static void
+RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+							 Oid oldroleid, void *callback_arg)
+{
+	bool		is_grant = *((bool *) callback_arg);
+	Oid			currentUserId = GetUserId();
+
+	check_role_membership_authorization(currentUserId, roleid, is_grant);
+}
+
+
 /*
  * GrantRoleStmt
  *
@@ -1568,9 +1681,14 @@ GrantRole(ParseState *pstate, GrantRoleStmt *stmt)
 					(errcode(ERRCODE_INVALID_GRANT_OPERATION),
 					 errmsg("column names cannot be included in GRANT/REVOKE ROLE")));
 
-		roleid = get_role_oid(rolename, false);
-		check_role_membership_authorization(currentUserId,
-											roleid, stmt->is_grant);
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP commits between name resolution
+		 * and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(rolename, ShareUpdateExclusiveLock, false,
+								RoleNameCallbackForGrantRole, &stmt->is_grant);
 		if (stmt->is_grant)
 			AddRoleMems(currentUserId, rolename, roleid,
 						stmt->grantee_roles, grantee_ids,
diff --git a/src/include/commands/user.h b/src/include/commands/user.h
index 97dcb93791b..17263452250 100644
--- a/src/include/commands/user.h
+++ b/src/include/commands/user.h
@@ -21,6 +21,15 @@
 extern PGDLLIMPORT int Password_encryption; /* values from enum PasswordType */
 extern PGDLLIMPORT char *createrole_self_grant;
 
+/* Callback for RoleNameGetOid, invoked after name resolution but before locking */
+typedef void (*RoleNameGetOidCallback) (const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+
+extern Oid	RoleNameGetOid(const char *rolename, LOCKMODE lockmode,
+						   bool missing_ok,
+						   RoleNameGetOidCallback callback,
+						   void *callback_arg);
+
 /* Hook to check passwords in CreateRole() and AlterRole() */
 typedef void (*check_password_hook_type) (const char *username, const char *shadow_pass, PasswordType password_type, Datum validuntil_time, bool validuntil_null);
 
-- 
2.34.1


--xlWYagR37YwJXkF8--





^ permalink  raw  reply  [nested|flat] 173+ messages in thread

* [PATCH v1] Add RoleNameGetOid() with invalidation-based retry loop for DropRole()/GrantRole()
@ 2026-07-04 04:47 Bertrand Drouvot <[email protected]>
  0 siblings, 0 replies; 173+ messages in thread

From: Bertrand Drouvot @ 2026-07-04 04:47 UTC (permalink / raw)

DropRole() and GrantRole() resolve the role name to an OID before acquiring
LockSharedObject() on the role. A concurrent session that commits a DROP ROLE
between the read and the lock acquisition leaves the first session acting on a
stale OID.

This commit fixes the races by using the same approach as RangeVarGetRelidExtended():
It encapsulates name resolution, permission checking (via a caller-supplied callback),
and lock acquisition inside a retry loop driven by SharedInvalidMessageCounter.
If invalidation messages arrive between name resolution and locking, indicating
concurrent DDL, the function retries.

The lock is kept across retries and only released if the name resolves to a
different OID on the next iteration.

Two callbacks are provided:
- RoleNameCallbackForDropRole(): checks current/session user, superuser attribute,
and ADMIN OPTION privilege before locking. This is similar to what DropRole() is
currently doing before LockSharedObject().
- RoleNameCallbackForGrantRole(): calls check_role_membership_authorization() to
verify the current user can grant/revoke membership. This is similar to what GrantRole()
is currently doing before calling AddRoleMems()/DelRoleMems().

DropRole() and GrantRole() now call RoleNameGetOid() with appropriate lock
levels.

AlterRole() does not need the fix because it calls CatalogTupleUpdate() on the
pg_authid tuple before AddRoleMems(), which blocks a concurrent DROP ROLE.

Author: Bertrand Drouvot <[email protected]>
Reviewed-by:
Discussion:
---
 src/backend/commands/user.c | 246 ++++++++++++++++++++++++++----------
 src/include/commands/user.h |   9 ++
 2 files changed, 191 insertions(+), 64 deletions(-)
  95.5% src/backend/commands/
   4.4% src/include/commands/

diff --git a/src/backend/commands/user.c b/src/backend/commands/user.c
index be11c49f919..6d8e2fa8813 100644
--- a/src/backend/commands/user.c
+++ b/src/backend/commands/user.c
@@ -34,6 +34,7 @@
 #include "miscadmin.h"
 #include "port/pg_bitutils.h"
 #include "storage/lmgr.h"
+#include "storage/sinval.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
 #include "utils/catcache.h"
@@ -116,6 +117,10 @@ static void plan_recursive_revoke(CatCList *memlist,
 								  bool revoke_admin_option_only,
 								  DropBehavior behavior);
 static void InitGrantRoleOptions(GrantRoleOptions *popt);
+static void RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+static void RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+										 Oid oldroleid, void *callback_arg);
 
 
 /* Check if current user has createrole privileges */
@@ -126,6 +131,92 @@ have_createrole_privilege(void)
 }
 
 
+/*
+ * RoleNameGetOid
+ *		Given a role name, look up its OID, lock it, and return the OID.
+ *
+ * This follows the same pattern as RangeVarGetRelidExtended():
+ * name resolution, permission check (via callback), and lock acquisition are
+ * performed inside a retry loop. If invalidation messages arrive during the
+ * process (indicating concurrent DDL), we retry to ensure the name still
+ * resolves to the same OID.
+ *
+ * The callback is invoked before locking, giving callers a chance to check
+ * permissions. It receives the current rolename, the resolved OID, the
+ * previous OID (InvalidOid on first iteration), and a caller-supplied arg.
+ * If the callback raises an error, the function aborts without locking.
+ *
+ * If missing_ok is true and the role does not exist, returns InvalidOid.
+ * Otherwise, raises an error.
+ */
+Oid
+RoleNameGetOid(const char *rolename, LOCKMODE lockmode, bool missing_ok,
+			   RoleNameGetOidCallback callback, void *callback_arg)
+{
+	uint64		inval_count;
+	Oid			roleid;
+	Oid			oldroleid = InvalidOid;
+	bool		retry = false;
+
+	for (;;)
+	{
+		/*
+		 * Remember the current invalidation count so we can detect concurrent
+		 * DDL after locking.
+		 */
+		inval_count = SharedInvalidMessageCounter;
+
+		/* Look up the role name */
+		roleid = get_role_oid(rolename, true);
+
+		if (!OidIsValid(roleid))
+		{
+			if (!missing_ok)
+				ereport(ERROR,
+						(errcode(ERRCODE_UNDEFINED_OBJECT),
+						 errmsg("role \"%s\" does not exist", rolename)));
+			return InvalidOid;
+		}
+
+		/*
+		 * Invoke caller-supplied callback before locking. This is a good
+		 * place to check permissions: we haven't taken the lock yet, but we
+		 * know the OID we intend to lock. If concurrent DDL changes things,
+		 * the callback will be invoked again on the next iteration.
+		 */
+		if (callback)
+			callback(rolename, roleid, oldroleid, callback_arg);
+
+		/*
+		 * If upon retry we get back the same OID, the invalidation messages
+		 * did not change the final answer. So we're done.
+		 *
+		 * If we got a different OID, we've locked the role that used to have
+		 * this name rather than the one that does now. Release the old lock.
+		 */
+		if (retry)
+		{
+			if (roleid == oldroleid)
+				break;
+			UnlockSharedObject(AuthIdRelationId, oldroleid, 0, lockmode);
+		}
+
+		/* Lock the role */
+		LockSharedObject(AuthIdRelationId, roleid, 0, lockmode);
+
+		/* If no invalidation messages were processed, we're done */
+		if (inval_count == SharedInvalidMessageCounter)
+			break;
+
+		/* Something may have changed, retry */
+		retry = true;
+		oldroleid = roleid;
+	}
+
+	return roleid;
+}
+
+
 /*
  * CREATE ROLE
  */
@@ -1090,6 +1181,59 @@ AlterRoleSet(AlterRoleSetStmt *stmt)
 }
 
 
+/*
+ * Before acquiring a role lock for DROP ROLE, check that the role is not the
+ * current/session user and that the caller has sufficient privileges to drop it.
+ */
+static void
+RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+							Oid oldroleid, void *callback_arg)
+{
+	HeapTuple	tuple;
+	Form_pg_authid roleform;
+
+	if (roleid == GetUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetOuterUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetSessionUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("session user cannot be dropped")));
+
+	tuple = SearchSysCache1(AUTHOID, ObjectIdGetDatum(roleid));
+	if (!HeapTupleIsValid(tuple))
+		ereport(ERROR,
+				(errcode(ERRCODE_UNDEFINED_OBJECT),
+				 errmsg("role \"%s\" does not exist", rolename)));
+
+	roleform = (Form_pg_authid) GETSTRUCT(tuple);
+
+	/*
+	 * For safety's sake, we allow createrole holders to drop ordinary roles
+	 * but not superuser roles, and only if they also have ADMIN OPTION.
+	 */
+	if (roleform->rolsuper && !superuser())
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
+						   "SUPERUSER", "SUPERUSER")));
+	if (!is_admin_of_role(GetUserId(), roleid))
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
+						   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
+
+	ReleaseSysCache(tuple);
+}
+
+
 /*
  * DROP ROLE
  */
@@ -1119,9 +1263,7 @@ DropRole(DropRoleStmt *stmt)
 	{
 		RoleSpec   *rolspec = lfirst(item);
 		char	   *role;
-		HeapTuple	tuple,
-					tmp_tuple;
-		Form_pg_authid roleform;
+		HeapTuple	tmp_tuple;
 		ScanKeyData scankey;
 		SysScanDesc sscan;
 		Oid			roleid;
@@ -1132,71 +1274,27 @@ DropRole(DropRoleStmt *stmt)
 					 errmsg("cannot use special role specifier in DROP ROLE")));
 		role = rolspec->rolename;
 
-		tuple = SearchSysCache1(AUTHNAME, PointerGetDatum(role));
-		if (!HeapTupleIsValid(tuple))
-		{
-			if (!stmt->missing_ok)
-			{
-				ereport(ERROR,
-						(errcode(ERRCODE_UNDEFINED_OBJECT),
-						 errmsg("role \"%s\" does not exist", role)));
-			}
-			else
-			{
-				ereport(NOTICE,
-						(errmsg("role \"%s\" does not exist, skipping",
-								role)));
-			}
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP or ALTER commits between name
+		 * resolution and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(role, AccessExclusiveLock, stmt->missing_ok,
+								RoleNameCallbackForDropRole, NULL);
 
+		if (!OidIsValid(roleid))
+		{
+			/* missing_ok case: role doesn't exist */
+			ereport(NOTICE,
+					(errmsg("role \"%s\" does not exist, skipping",
+							role)));
 			continue;
 		}
 
-		roleform = (Form_pg_authid) GETSTRUCT(tuple);
-		roleid = roleform->oid;
-
-		if (roleid == GetUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetOuterUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetSessionUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("session user cannot be dropped")));
-
-		/*
-		 * For safety's sake, we allow createrole holders to drop ordinary
-		 * roles but not superuser roles, and only if they also have ADMIN
-		 * OPTION.
-		 */
-		if (roleform->rolsuper && !superuser())
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
-							   "SUPERUSER", "SUPERUSER")));
-		if (!is_admin_of_role(GetUserId(), roleid))
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
-							   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
-
 		/* DROP hook for the role being removed */
 		InvokeObjectDropHook(AuthIdRelationId, roleid, 0);
 
-		/* Don't leak the syscache tuple */
-		ReleaseSysCache(tuple);
-
-		/*
-		 * Lock the role, so nobody can add dependencies to her while we drop
-		 * her.  We keep the lock until the end of transaction.
-		 */
-		LockSharedObject(AuthIdRelationId, roleid, 0, AccessExclusiveLock);
-
 		/*
 		 * If there is a pg_auth_members entry that has one of the roles to be
 		 * dropped as the roleid or member, it should be silently removed, but
@@ -1484,6 +1582,21 @@ RenameRole(const char *oldname, const char *newname)
 	return address;
 }
 
+/*
+ * Before acquiring a role lock for GRANT/REVOKE, check that the current user
+ * has authorization to grant/revoke membership in the specified role.
+ */
+static void
+RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+							 Oid oldroleid, void *callback_arg)
+{
+	bool		is_grant = *((bool *) callback_arg);
+	Oid			currentUserId = GetUserId();
+
+	check_role_membership_authorization(currentUserId, roleid, is_grant);
+}
+
+
 /*
  * GrantRoleStmt
  *
@@ -1568,9 +1681,14 @@ GrantRole(ParseState *pstate, GrantRoleStmt *stmt)
 					(errcode(ERRCODE_INVALID_GRANT_OPERATION),
 					 errmsg("column names cannot be included in GRANT/REVOKE ROLE")));
 
-		roleid = get_role_oid(rolename, false);
-		check_role_membership_authorization(currentUserId,
-											roleid, stmt->is_grant);
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP commits between name resolution
+		 * and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(rolename, ShareUpdateExclusiveLock, false,
+								RoleNameCallbackForGrantRole, &stmt->is_grant);
 		if (stmt->is_grant)
 			AddRoleMems(currentUserId, rolename, roleid,
 						stmt->grantee_roles, grantee_ids,
diff --git a/src/include/commands/user.h b/src/include/commands/user.h
index 97dcb93791b..17263452250 100644
--- a/src/include/commands/user.h
+++ b/src/include/commands/user.h
@@ -21,6 +21,15 @@
 extern PGDLLIMPORT int Password_encryption; /* values from enum PasswordType */
 extern PGDLLIMPORT char *createrole_self_grant;
 
+/* Callback for RoleNameGetOid, invoked after name resolution but before locking */
+typedef void (*RoleNameGetOidCallback) (const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+
+extern Oid	RoleNameGetOid(const char *rolename, LOCKMODE lockmode,
+						   bool missing_ok,
+						   RoleNameGetOidCallback callback,
+						   void *callback_arg);
+
 /* Hook to check passwords in CreateRole() and AlterRole() */
 typedef void (*check_password_hook_type) (const char *username, const char *shadow_pass, PasswordType password_type, Datum validuntil_time, bool validuntil_null);
 
-- 
2.34.1


--xlWYagR37YwJXkF8--





^ permalink  raw  reply  [nested|flat] 173+ messages in thread

* [PATCH v1] Add RoleNameGetOid() with invalidation-based retry loop for DropRole()/GrantRole()
@ 2026-07-04 04:47 Bertrand Drouvot <[email protected]>
  0 siblings, 0 replies; 173+ messages in thread

From: Bertrand Drouvot @ 2026-07-04 04:47 UTC (permalink / raw)

DropRole() and GrantRole() resolve the role name to an OID before acquiring
LockSharedObject() on the role. A concurrent session that commits a DROP ROLE
between the read and the lock acquisition leaves the first session acting on a
stale OID.

This commit fixes the races by using the same approach as RangeVarGetRelidExtended():
It encapsulates name resolution, permission checking (via a caller-supplied callback),
and lock acquisition inside a retry loop driven by SharedInvalidMessageCounter.
If invalidation messages arrive between name resolution and locking, indicating
concurrent DDL, the function retries.

The lock is kept across retries and only released if the name resolves to a
different OID on the next iteration.

Two callbacks are provided:
- RoleNameCallbackForDropRole(): checks current/session user, superuser attribute,
and ADMIN OPTION privilege before locking. This is similar to what DropRole() is
currently doing before LockSharedObject().
- RoleNameCallbackForGrantRole(): calls check_role_membership_authorization() to
verify the current user can grant/revoke membership. This is similar to what GrantRole()
is currently doing before calling AddRoleMems()/DelRoleMems().

DropRole() and GrantRole() now call RoleNameGetOid() with appropriate lock
levels.

AlterRole() does not need the fix because it calls CatalogTupleUpdate() on the
pg_authid tuple before AddRoleMems(), which blocks a concurrent DROP ROLE.

Author: Bertrand Drouvot <[email protected]>
Reviewed-by:
Discussion:
---
 src/backend/commands/user.c | 246 ++++++++++++++++++++++++++----------
 src/include/commands/user.h |   9 ++
 2 files changed, 191 insertions(+), 64 deletions(-)
  95.5% src/backend/commands/
   4.4% src/include/commands/

diff --git a/src/backend/commands/user.c b/src/backend/commands/user.c
index be11c49f919..6d8e2fa8813 100644
--- a/src/backend/commands/user.c
+++ b/src/backend/commands/user.c
@@ -34,6 +34,7 @@
 #include "miscadmin.h"
 #include "port/pg_bitutils.h"
 #include "storage/lmgr.h"
+#include "storage/sinval.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
 #include "utils/catcache.h"
@@ -116,6 +117,10 @@ static void plan_recursive_revoke(CatCList *memlist,
 								  bool revoke_admin_option_only,
 								  DropBehavior behavior);
 static void InitGrantRoleOptions(GrantRoleOptions *popt);
+static void RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+static void RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+										 Oid oldroleid, void *callback_arg);
 
 
 /* Check if current user has createrole privileges */
@@ -126,6 +131,92 @@ have_createrole_privilege(void)
 }
 
 
+/*
+ * RoleNameGetOid
+ *		Given a role name, look up its OID, lock it, and return the OID.
+ *
+ * This follows the same pattern as RangeVarGetRelidExtended():
+ * name resolution, permission check (via callback), and lock acquisition are
+ * performed inside a retry loop. If invalidation messages arrive during the
+ * process (indicating concurrent DDL), we retry to ensure the name still
+ * resolves to the same OID.
+ *
+ * The callback is invoked before locking, giving callers a chance to check
+ * permissions. It receives the current rolename, the resolved OID, the
+ * previous OID (InvalidOid on first iteration), and a caller-supplied arg.
+ * If the callback raises an error, the function aborts without locking.
+ *
+ * If missing_ok is true and the role does not exist, returns InvalidOid.
+ * Otherwise, raises an error.
+ */
+Oid
+RoleNameGetOid(const char *rolename, LOCKMODE lockmode, bool missing_ok,
+			   RoleNameGetOidCallback callback, void *callback_arg)
+{
+	uint64		inval_count;
+	Oid			roleid;
+	Oid			oldroleid = InvalidOid;
+	bool		retry = false;
+
+	for (;;)
+	{
+		/*
+		 * Remember the current invalidation count so we can detect concurrent
+		 * DDL after locking.
+		 */
+		inval_count = SharedInvalidMessageCounter;
+
+		/* Look up the role name */
+		roleid = get_role_oid(rolename, true);
+
+		if (!OidIsValid(roleid))
+		{
+			if (!missing_ok)
+				ereport(ERROR,
+						(errcode(ERRCODE_UNDEFINED_OBJECT),
+						 errmsg("role \"%s\" does not exist", rolename)));
+			return InvalidOid;
+		}
+
+		/*
+		 * Invoke caller-supplied callback before locking. This is a good
+		 * place to check permissions: we haven't taken the lock yet, but we
+		 * know the OID we intend to lock. If concurrent DDL changes things,
+		 * the callback will be invoked again on the next iteration.
+		 */
+		if (callback)
+			callback(rolename, roleid, oldroleid, callback_arg);
+
+		/*
+		 * If upon retry we get back the same OID, the invalidation messages
+		 * did not change the final answer. So we're done.
+		 *
+		 * If we got a different OID, we've locked the role that used to have
+		 * this name rather than the one that does now. Release the old lock.
+		 */
+		if (retry)
+		{
+			if (roleid == oldroleid)
+				break;
+			UnlockSharedObject(AuthIdRelationId, oldroleid, 0, lockmode);
+		}
+
+		/* Lock the role */
+		LockSharedObject(AuthIdRelationId, roleid, 0, lockmode);
+
+		/* If no invalidation messages were processed, we're done */
+		if (inval_count == SharedInvalidMessageCounter)
+			break;
+
+		/* Something may have changed, retry */
+		retry = true;
+		oldroleid = roleid;
+	}
+
+	return roleid;
+}
+
+
 /*
  * CREATE ROLE
  */
@@ -1090,6 +1181,59 @@ AlterRoleSet(AlterRoleSetStmt *stmt)
 }
 
 
+/*
+ * Before acquiring a role lock for DROP ROLE, check that the role is not the
+ * current/session user and that the caller has sufficient privileges to drop it.
+ */
+static void
+RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+							Oid oldroleid, void *callback_arg)
+{
+	HeapTuple	tuple;
+	Form_pg_authid roleform;
+
+	if (roleid == GetUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetOuterUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetSessionUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("session user cannot be dropped")));
+
+	tuple = SearchSysCache1(AUTHOID, ObjectIdGetDatum(roleid));
+	if (!HeapTupleIsValid(tuple))
+		ereport(ERROR,
+				(errcode(ERRCODE_UNDEFINED_OBJECT),
+				 errmsg("role \"%s\" does not exist", rolename)));
+
+	roleform = (Form_pg_authid) GETSTRUCT(tuple);
+
+	/*
+	 * For safety's sake, we allow createrole holders to drop ordinary roles
+	 * but not superuser roles, and only if they also have ADMIN OPTION.
+	 */
+	if (roleform->rolsuper && !superuser())
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
+						   "SUPERUSER", "SUPERUSER")));
+	if (!is_admin_of_role(GetUserId(), roleid))
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
+						   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
+
+	ReleaseSysCache(tuple);
+}
+
+
 /*
  * DROP ROLE
  */
@@ -1119,9 +1263,7 @@ DropRole(DropRoleStmt *stmt)
 	{
 		RoleSpec   *rolspec = lfirst(item);
 		char	   *role;
-		HeapTuple	tuple,
-					tmp_tuple;
-		Form_pg_authid roleform;
+		HeapTuple	tmp_tuple;
 		ScanKeyData scankey;
 		SysScanDesc sscan;
 		Oid			roleid;
@@ -1132,71 +1274,27 @@ DropRole(DropRoleStmt *stmt)
 					 errmsg("cannot use special role specifier in DROP ROLE")));
 		role = rolspec->rolename;
 
-		tuple = SearchSysCache1(AUTHNAME, PointerGetDatum(role));
-		if (!HeapTupleIsValid(tuple))
-		{
-			if (!stmt->missing_ok)
-			{
-				ereport(ERROR,
-						(errcode(ERRCODE_UNDEFINED_OBJECT),
-						 errmsg("role \"%s\" does not exist", role)));
-			}
-			else
-			{
-				ereport(NOTICE,
-						(errmsg("role \"%s\" does not exist, skipping",
-								role)));
-			}
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP or ALTER commits between name
+		 * resolution and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(role, AccessExclusiveLock, stmt->missing_ok,
+								RoleNameCallbackForDropRole, NULL);
 
+		if (!OidIsValid(roleid))
+		{
+			/* missing_ok case: role doesn't exist */
+			ereport(NOTICE,
+					(errmsg("role \"%s\" does not exist, skipping",
+							role)));
 			continue;
 		}
 
-		roleform = (Form_pg_authid) GETSTRUCT(tuple);
-		roleid = roleform->oid;
-
-		if (roleid == GetUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetOuterUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetSessionUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("session user cannot be dropped")));
-
-		/*
-		 * For safety's sake, we allow createrole holders to drop ordinary
-		 * roles but not superuser roles, and only if they also have ADMIN
-		 * OPTION.
-		 */
-		if (roleform->rolsuper && !superuser())
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
-							   "SUPERUSER", "SUPERUSER")));
-		if (!is_admin_of_role(GetUserId(), roleid))
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
-							   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
-
 		/* DROP hook for the role being removed */
 		InvokeObjectDropHook(AuthIdRelationId, roleid, 0);
 
-		/* Don't leak the syscache tuple */
-		ReleaseSysCache(tuple);
-
-		/*
-		 * Lock the role, so nobody can add dependencies to her while we drop
-		 * her.  We keep the lock until the end of transaction.
-		 */
-		LockSharedObject(AuthIdRelationId, roleid, 0, AccessExclusiveLock);
-
 		/*
 		 * If there is a pg_auth_members entry that has one of the roles to be
 		 * dropped as the roleid or member, it should be silently removed, but
@@ -1484,6 +1582,21 @@ RenameRole(const char *oldname, const char *newname)
 	return address;
 }
 
+/*
+ * Before acquiring a role lock for GRANT/REVOKE, check that the current user
+ * has authorization to grant/revoke membership in the specified role.
+ */
+static void
+RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+							 Oid oldroleid, void *callback_arg)
+{
+	bool		is_grant = *((bool *) callback_arg);
+	Oid			currentUserId = GetUserId();
+
+	check_role_membership_authorization(currentUserId, roleid, is_grant);
+}
+
+
 /*
  * GrantRoleStmt
  *
@@ -1568,9 +1681,14 @@ GrantRole(ParseState *pstate, GrantRoleStmt *stmt)
 					(errcode(ERRCODE_INVALID_GRANT_OPERATION),
 					 errmsg("column names cannot be included in GRANT/REVOKE ROLE")));
 
-		roleid = get_role_oid(rolename, false);
-		check_role_membership_authorization(currentUserId,
-											roleid, stmt->is_grant);
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP commits between name resolution
+		 * and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(rolename, ShareUpdateExclusiveLock, false,
+								RoleNameCallbackForGrantRole, &stmt->is_grant);
 		if (stmt->is_grant)
 			AddRoleMems(currentUserId, rolename, roleid,
 						stmt->grantee_roles, grantee_ids,
diff --git a/src/include/commands/user.h b/src/include/commands/user.h
index 97dcb93791b..17263452250 100644
--- a/src/include/commands/user.h
+++ b/src/include/commands/user.h
@@ -21,6 +21,15 @@
 extern PGDLLIMPORT int Password_encryption; /* values from enum PasswordType */
 extern PGDLLIMPORT char *createrole_self_grant;
 
+/* Callback for RoleNameGetOid, invoked after name resolution but before locking */
+typedef void (*RoleNameGetOidCallback) (const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+
+extern Oid	RoleNameGetOid(const char *rolename, LOCKMODE lockmode,
+						   bool missing_ok,
+						   RoleNameGetOidCallback callback,
+						   void *callback_arg);
+
 /* Hook to check passwords in CreateRole() and AlterRole() */
 typedef void (*check_password_hook_type) (const char *username, const char *shadow_pass, PasswordType password_type, Datum validuntil_time, bool validuntil_null);
 
-- 
2.34.1


--xlWYagR37YwJXkF8--





^ permalink  raw  reply  [nested|flat] 173+ messages in thread

* [PATCH v1] Add RoleNameGetOid() with invalidation-based retry loop for DropRole()/GrantRole()
@ 2026-07-04 04:47 Bertrand Drouvot <[email protected]>
  0 siblings, 0 replies; 173+ messages in thread

From: Bertrand Drouvot @ 2026-07-04 04:47 UTC (permalink / raw)

DropRole() and GrantRole() resolve the role name to an OID before acquiring
LockSharedObject() on the role. A concurrent session that commits a DROP ROLE
between the read and the lock acquisition leaves the first session acting on a
stale OID.

This commit fixes the races by using the same approach as RangeVarGetRelidExtended():
It encapsulates name resolution, permission checking (via a caller-supplied callback),
and lock acquisition inside a retry loop driven by SharedInvalidMessageCounter.
If invalidation messages arrive between name resolution and locking, indicating
concurrent DDL, the function retries.

The lock is kept across retries and only released if the name resolves to a
different OID on the next iteration.

Two callbacks are provided:
- RoleNameCallbackForDropRole(): checks current/session user, superuser attribute,
and ADMIN OPTION privilege before locking. This is similar to what DropRole() is
currently doing before LockSharedObject().
- RoleNameCallbackForGrantRole(): calls check_role_membership_authorization() to
verify the current user can grant/revoke membership. This is similar to what GrantRole()
is currently doing before calling AddRoleMems()/DelRoleMems().

DropRole() and GrantRole() now call RoleNameGetOid() with appropriate lock
levels.

AlterRole() does not need the fix because it calls CatalogTupleUpdate() on the
pg_authid tuple before AddRoleMems(), which blocks a concurrent DROP ROLE.

Author: Bertrand Drouvot <[email protected]>
Reviewed-by:
Discussion:
---
 src/backend/commands/user.c | 246 ++++++++++++++++++++++++++----------
 src/include/commands/user.h |   9 ++
 2 files changed, 191 insertions(+), 64 deletions(-)
  95.5% src/backend/commands/
   4.4% src/include/commands/

diff --git a/src/backend/commands/user.c b/src/backend/commands/user.c
index be11c49f919..6d8e2fa8813 100644
--- a/src/backend/commands/user.c
+++ b/src/backend/commands/user.c
@@ -34,6 +34,7 @@
 #include "miscadmin.h"
 #include "port/pg_bitutils.h"
 #include "storage/lmgr.h"
+#include "storage/sinval.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
 #include "utils/catcache.h"
@@ -116,6 +117,10 @@ static void plan_recursive_revoke(CatCList *memlist,
 								  bool revoke_admin_option_only,
 								  DropBehavior behavior);
 static void InitGrantRoleOptions(GrantRoleOptions *popt);
+static void RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+static void RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+										 Oid oldroleid, void *callback_arg);
 
 
 /* Check if current user has createrole privileges */
@@ -126,6 +131,92 @@ have_createrole_privilege(void)
 }
 
 
+/*
+ * RoleNameGetOid
+ *		Given a role name, look up its OID, lock it, and return the OID.
+ *
+ * This follows the same pattern as RangeVarGetRelidExtended():
+ * name resolution, permission check (via callback), and lock acquisition are
+ * performed inside a retry loop. If invalidation messages arrive during the
+ * process (indicating concurrent DDL), we retry to ensure the name still
+ * resolves to the same OID.
+ *
+ * The callback is invoked before locking, giving callers a chance to check
+ * permissions. It receives the current rolename, the resolved OID, the
+ * previous OID (InvalidOid on first iteration), and a caller-supplied arg.
+ * If the callback raises an error, the function aborts without locking.
+ *
+ * If missing_ok is true and the role does not exist, returns InvalidOid.
+ * Otherwise, raises an error.
+ */
+Oid
+RoleNameGetOid(const char *rolename, LOCKMODE lockmode, bool missing_ok,
+			   RoleNameGetOidCallback callback, void *callback_arg)
+{
+	uint64		inval_count;
+	Oid			roleid;
+	Oid			oldroleid = InvalidOid;
+	bool		retry = false;
+
+	for (;;)
+	{
+		/*
+		 * Remember the current invalidation count so we can detect concurrent
+		 * DDL after locking.
+		 */
+		inval_count = SharedInvalidMessageCounter;
+
+		/* Look up the role name */
+		roleid = get_role_oid(rolename, true);
+
+		if (!OidIsValid(roleid))
+		{
+			if (!missing_ok)
+				ereport(ERROR,
+						(errcode(ERRCODE_UNDEFINED_OBJECT),
+						 errmsg("role \"%s\" does not exist", rolename)));
+			return InvalidOid;
+		}
+
+		/*
+		 * Invoke caller-supplied callback before locking. This is a good
+		 * place to check permissions: we haven't taken the lock yet, but we
+		 * know the OID we intend to lock. If concurrent DDL changes things,
+		 * the callback will be invoked again on the next iteration.
+		 */
+		if (callback)
+			callback(rolename, roleid, oldroleid, callback_arg);
+
+		/*
+		 * If upon retry we get back the same OID, the invalidation messages
+		 * did not change the final answer. So we're done.
+		 *
+		 * If we got a different OID, we've locked the role that used to have
+		 * this name rather than the one that does now. Release the old lock.
+		 */
+		if (retry)
+		{
+			if (roleid == oldroleid)
+				break;
+			UnlockSharedObject(AuthIdRelationId, oldroleid, 0, lockmode);
+		}
+
+		/* Lock the role */
+		LockSharedObject(AuthIdRelationId, roleid, 0, lockmode);
+
+		/* If no invalidation messages were processed, we're done */
+		if (inval_count == SharedInvalidMessageCounter)
+			break;
+
+		/* Something may have changed, retry */
+		retry = true;
+		oldroleid = roleid;
+	}
+
+	return roleid;
+}
+
+
 /*
  * CREATE ROLE
  */
@@ -1090,6 +1181,59 @@ AlterRoleSet(AlterRoleSetStmt *stmt)
 }
 
 
+/*
+ * Before acquiring a role lock for DROP ROLE, check that the role is not the
+ * current/session user and that the caller has sufficient privileges to drop it.
+ */
+static void
+RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+							Oid oldroleid, void *callback_arg)
+{
+	HeapTuple	tuple;
+	Form_pg_authid roleform;
+
+	if (roleid == GetUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetOuterUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetSessionUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("session user cannot be dropped")));
+
+	tuple = SearchSysCache1(AUTHOID, ObjectIdGetDatum(roleid));
+	if (!HeapTupleIsValid(tuple))
+		ereport(ERROR,
+				(errcode(ERRCODE_UNDEFINED_OBJECT),
+				 errmsg("role \"%s\" does not exist", rolename)));
+
+	roleform = (Form_pg_authid) GETSTRUCT(tuple);
+
+	/*
+	 * For safety's sake, we allow createrole holders to drop ordinary roles
+	 * but not superuser roles, and only if they also have ADMIN OPTION.
+	 */
+	if (roleform->rolsuper && !superuser())
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
+						   "SUPERUSER", "SUPERUSER")));
+	if (!is_admin_of_role(GetUserId(), roleid))
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
+						   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
+
+	ReleaseSysCache(tuple);
+}
+
+
 /*
  * DROP ROLE
  */
@@ -1119,9 +1263,7 @@ DropRole(DropRoleStmt *stmt)
 	{
 		RoleSpec   *rolspec = lfirst(item);
 		char	   *role;
-		HeapTuple	tuple,
-					tmp_tuple;
-		Form_pg_authid roleform;
+		HeapTuple	tmp_tuple;
 		ScanKeyData scankey;
 		SysScanDesc sscan;
 		Oid			roleid;
@@ -1132,71 +1274,27 @@ DropRole(DropRoleStmt *stmt)
 					 errmsg("cannot use special role specifier in DROP ROLE")));
 		role = rolspec->rolename;
 
-		tuple = SearchSysCache1(AUTHNAME, PointerGetDatum(role));
-		if (!HeapTupleIsValid(tuple))
-		{
-			if (!stmt->missing_ok)
-			{
-				ereport(ERROR,
-						(errcode(ERRCODE_UNDEFINED_OBJECT),
-						 errmsg("role \"%s\" does not exist", role)));
-			}
-			else
-			{
-				ereport(NOTICE,
-						(errmsg("role \"%s\" does not exist, skipping",
-								role)));
-			}
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP or ALTER commits between name
+		 * resolution and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(role, AccessExclusiveLock, stmt->missing_ok,
+								RoleNameCallbackForDropRole, NULL);
 
+		if (!OidIsValid(roleid))
+		{
+			/* missing_ok case: role doesn't exist */
+			ereport(NOTICE,
+					(errmsg("role \"%s\" does not exist, skipping",
+							role)));
 			continue;
 		}
 
-		roleform = (Form_pg_authid) GETSTRUCT(tuple);
-		roleid = roleform->oid;
-
-		if (roleid == GetUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetOuterUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetSessionUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("session user cannot be dropped")));
-
-		/*
-		 * For safety's sake, we allow createrole holders to drop ordinary
-		 * roles but not superuser roles, and only if they also have ADMIN
-		 * OPTION.
-		 */
-		if (roleform->rolsuper && !superuser())
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
-							   "SUPERUSER", "SUPERUSER")));
-		if (!is_admin_of_role(GetUserId(), roleid))
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
-							   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
-
 		/* DROP hook for the role being removed */
 		InvokeObjectDropHook(AuthIdRelationId, roleid, 0);
 
-		/* Don't leak the syscache tuple */
-		ReleaseSysCache(tuple);
-
-		/*
-		 * Lock the role, so nobody can add dependencies to her while we drop
-		 * her.  We keep the lock until the end of transaction.
-		 */
-		LockSharedObject(AuthIdRelationId, roleid, 0, AccessExclusiveLock);
-
 		/*
 		 * If there is a pg_auth_members entry that has one of the roles to be
 		 * dropped as the roleid or member, it should be silently removed, but
@@ -1484,6 +1582,21 @@ RenameRole(const char *oldname, const char *newname)
 	return address;
 }
 
+/*
+ * Before acquiring a role lock for GRANT/REVOKE, check that the current user
+ * has authorization to grant/revoke membership in the specified role.
+ */
+static void
+RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+							 Oid oldroleid, void *callback_arg)
+{
+	bool		is_grant = *((bool *) callback_arg);
+	Oid			currentUserId = GetUserId();
+
+	check_role_membership_authorization(currentUserId, roleid, is_grant);
+}
+
+
 /*
  * GrantRoleStmt
  *
@@ -1568,9 +1681,14 @@ GrantRole(ParseState *pstate, GrantRoleStmt *stmt)
 					(errcode(ERRCODE_INVALID_GRANT_OPERATION),
 					 errmsg("column names cannot be included in GRANT/REVOKE ROLE")));
 
-		roleid = get_role_oid(rolename, false);
-		check_role_membership_authorization(currentUserId,
-											roleid, stmt->is_grant);
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP commits between name resolution
+		 * and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(rolename, ShareUpdateExclusiveLock, false,
+								RoleNameCallbackForGrantRole, &stmt->is_grant);
 		if (stmt->is_grant)
 			AddRoleMems(currentUserId, rolename, roleid,
 						stmt->grantee_roles, grantee_ids,
diff --git a/src/include/commands/user.h b/src/include/commands/user.h
index 97dcb93791b..17263452250 100644
--- a/src/include/commands/user.h
+++ b/src/include/commands/user.h
@@ -21,6 +21,15 @@
 extern PGDLLIMPORT int Password_encryption; /* values from enum PasswordType */
 extern PGDLLIMPORT char *createrole_self_grant;
 
+/* Callback for RoleNameGetOid, invoked after name resolution but before locking */
+typedef void (*RoleNameGetOidCallback) (const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+
+extern Oid	RoleNameGetOid(const char *rolename, LOCKMODE lockmode,
+						   bool missing_ok,
+						   RoleNameGetOidCallback callback,
+						   void *callback_arg);
+
 /* Hook to check passwords in CreateRole() and AlterRole() */
 typedef void (*check_password_hook_type) (const char *username, const char *shadow_pass, PasswordType password_type, Datum validuntil_time, bool validuntil_null);
 
-- 
2.34.1


--xlWYagR37YwJXkF8--





^ permalink  raw  reply  [nested|flat] 173+ messages in thread

* [PATCH v1] Add RoleNameGetOid() with invalidation-based retry loop for DropRole()/GrantRole()
@ 2026-07-04 04:47 Bertrand Drouvot <[email protected]>
  0 siblings, 0 replies; 173+ messages in thread

From: Bertrand Drouvot @ 2026-07-04 04:47 UTC (permalink / raw)

DropRole() and GrantRole() resolve the role name to an OID before acquiring
LockSharedObject() on the role. A concurrent session that commits a DROP ROLE
between the read and the lock acquisition leaves the first session acting on a
stale OID.

This commit fixes the races by using the same approach as RangeVarGetRelidExtended():
It encapsulates name resolution, permission checking (via a caller-supplied callback),
and lock acquisition inside a retry loop driven by SharedInvalidMessageCounter.
If invalidation messages arrive between name resolution and locking, indicating
concurrent DDL, the function retries.

The lock is kept across retries and only released if the name resolves to a
different OID on the next iteration.

Two callbacks are provided:
- RoleNameCallbackForDropRole(): checks current/session user, superuser attribute,
and ADMIN OPTION privilege before locking. This is similar to what DropRole() is
currently doing before LockSharedObject().
- RoleNameCallbackForGrantRole(): calls check_role_membership_authorization() to
verify the current user can grant/revoke membership. This is similar to what GrantRole()
is currently doing before calling AddRoleMems()/DelRoleMems().

DropRole() and GrantRole() now call RoleNameGetOid() with appropriate lock
levels.

AlterRole() does not need the fix because it calls CatalogTupleUpdate() on the
pg_authid tuple before AddRoleMems(), which blocks a concurrent DROP ROLE.

Author: Bertrand Drouvot <[email protected]>
Reviewed-by:
Discussion:
---
 src/backend/commands/user.c | 246 ++++++++++++++++++++++++++----------
 src/include/commands/user.h |   9 ++
 2 files changed, 191 insertions(+), 64 deletions(-)
  95.5% src/backend/commands/
   4.4% src/include/commands/

diff --git a/src/backend/commands/user.c b/src/backend/commands/user.c
index be11c49f919..6d8e2fa8813 100644
--- a/src/backend/commands/user.c
+++ b/src/backend/commands/user.c
@@ -34,6 +34,7 @@
 #include "miscadmin.h"
 #include "port/pg_bitutils.h"
 #include "storage/lmgr.h"
+#include "storage/sinval.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
 #include "utils/catcache.h"
@@ -116,6 +117,10 @@ static void plan_recursive_revoke(CatCList *memlist,
 								  bool revoke_admin_option_only,
 								  DropBehavior behavior);
 static void InitGrantRoleOptions(GrantRoleOptions *popt);
+static void RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+static void RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+										 Oid oldroleid, void *callback_arg);
 
 
 /* Check if current user has createrole privileges */
@@ -126,6 +131,92 @@ have_createrole_privilege(void)
 }
 
 
+/*
+ * RoleNameGetOid
+ *		Given a role name, look up its OID, lock it, and return the OID.
+ *
+ * This follows the same pattern as RangeVarGetRelidExtended():
+ * name resolution, permission check (via callback), and lock acquisition are
+ * performed inside a retry loop. If invalidation messages arrive during the
+ * process (indicating concurrent DDL), we retry to ensure the name still
+ * resolves to the same OID.
+ *
+ * The callback is invoked before locking, giving callers a chance to check
+ * permissions. It receives the current rolename, the resolved OID, the
+ * previous OID (InvalidOid on first iteration), and a caller-supplied arg.
+ * If the callback raises an error, the function aborts without locking.
+ *
+ * If missing_ok is true and the role does not exist, returns InvalidOid.
+ * Otherwise, raises an error.
+ */
+Oid
+RoleNameGetOid(const char *rolename, LOCKMODE lockmode, bool missing_ok,
+			   RoleNameGetOidCallback callback, void *callback_arg)
+{
+	uint64		inval_count;
+	Oid			roleid;
+	Oid			oldroleid = InvalidOid;
+	bool		retry = false;
+
+	for (;;)
+	{
+		/*
+		 * Remember the current invalidation count so we can detect concurrent
+		 * DDL after locking.
+		 */
+		inval_count = SharedInvalidMessageCounter;
+
+		/* Look up the role name */
+		roleid = get_role_oid(rolename, true);
+
+		if (!OidIsValid(roleid))
+		{
+			if (!missing_ok)
+				ereport(ERROR,
+						(errcode(ERRCODE_UNDEFINED_OBJECT),
+						 errmsg("role \"%s\" does not exist", rolename)));
+			return InvalidOid;
+		}
+
+		/*
+		 * Invoke caller-supplied callback before locking. This is a good
+		 * place to check permissions: we haven't taken the lock yet, but we
+		 * know the OID we intend to lock. If concurrent DDL changes things,
+		 * the callback will be invoked again on the next iteration.
+		 */
+		if (callback)
+			callback(rolename, roleid, oldroleid, callback_arg);
+
+		/*
+		 * If upon retry we get back the same OID, the invalidation messages
+		 * did not change the final answer. So we're done.
+		 *
+		 * If we got a different OID, we've locked the role that used to have
+		 * this name rather than the one that does now. Release the old lock.
+		 */
+		if (retry)
+		{
+			if (roleid == oldroleid)
+				break;
+			UnlockSharedObject(AuthIdRelationId, oldroleid, 0, lockmode);
+		}
+
+		/* Lock the role */
+		LockSharedObject(AuthIdRelationId, roleid, 0, lockmode);
+
+		/* If no invalidation messages were processed, we're done */
+		if (inval_count == SharedInvalidMessageCounter)
+			break;
+
+		/* Something may have changed, retry */
+		retry = true;
+		oldroleid = roleid;
+	}
+
+	return roleid;
+}
+
+
 /*
  * CREATE ROLE
  */
@@ -1090,6 +1181,59 @@ AlterRoleSet(AlterRoleSetStmt *stmt)
 }
 
 
+/*
+ * Before acquiring a role lock for DROP ROLE, check that the role is not the
+ * current/session user and that the caller has sufficient privileges to drop it.
+ */
+static void
+RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+							Oid oldroleid, void *callback_arg)
+{
+	HeapTuple	tuple;
+	Form_pg_authid roleform;
+
+	if (roleid == GetUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetOuterUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetSessionUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("session user cannot be dropped")));
+
+	tuple = SearchSysCache1(AUTHOID, ObjectIdGetDatum(roleid));
+	if (!HeapTupleIsValid(tuple))
+		ereport(ERROR,
+				(errcode(ERRCODE_UNDEFINED_OBJECT),
+				 errmsg("role \"%s\" does not exist", rolename)));
+
+	roleform = (Form_pg_authid) GETSTRUCT(tuple);
+
+	/*
+	 * For safety's sake, we allow createrole holders to drop ordinary roles
+	 * but not superuser roles, and only if they also have ADMIN OPTION.
+	 */
+	if (roleform->rolsuper && !superuser())
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
+						   "SUPERUSER", "SUPERUSER")));
+	if (!is_admin_of_role(GetUserId(), roleid))
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
+						   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
+
+	ReleaseSysCache(tuple);
+}
+
+
 /*
  * DROP ROLE
  */
@@ -1119,9 +1263,7 @@ DropRole(DropRoleStmt *stmt)
 	{
 		RoleSpec   *rolspec = lfirst(item);
 		char	   *role;
-		HeapTuple	tuple,
-					tmp_tuple;
-		Form_pg_authid roleform;
+		HeapTuple	tmp_tuple;
 		ScanKeyData scankey;
 		SysScanDesc sscan;
 		Oid			roleid;
@@ -1132,71 +1274,27 @@ DropRole(DropRoleStmt *stmt)
 					 errmsg("cannot use special role specifier in DROP ROLE")));
 		role = rolspec->rolename;
 
-		tuple = SearchSysCache1(AUTHNAME, PointerGetDatum(role));
-		if (!HeapTupleIsValid(tuple))
-		{
-			if (!stmt->missing_ok)
-			{
-				ereport(ERROR,
-						(errcode(ERRCODE_UNDEFINED_OBJECT),
-						 errmsg("role \"%s\" does not exist", role)));
-			}
-			else
-			{
-				ereport(NOTICE,
-						(errmsg("role \"%s\" does not exist, skipping",
-								role)));
-			}
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP or ALTER commits between name
+		 * resolution and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(role, AccessExclusiveLock, stmt->missing_ok,
+								RoleNameCallbackForDropRole, NULL);
 
+		if (!OidIsValid(roleid))
+		{
+			/* missing_ok case: role doesn't exist */
+			ereport(NOTICE,
+					(errmsg("role \"%s\" does not exist, skipping",
+							role)));
 			continue;
 		}
 
-		roleform = (Form_pg_authid) GETSTRUCT(tuple);
-		roleid = roleform->oid;
-
-		if (roleid == GetUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetOuterUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetSessionUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("session user cannot be dropped")));
-
-		/*
-		 * For safety's sake, we allow createrole holders to drop ordinary
-		 * roles but not superuser roles, and only if they also have ADMIN
-		 * OPTION.
-		 */
-		if (roleform->rolsuper && !superuser())
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
-							   "SUPERUSER", "SUPERUSER")));
-		if (!is_admin_of_role(GetUserId(), roleid))
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
-							   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
-
 		/* DROP hook for the role being removed */
 		InvokeObjectDropHook(AuthIdRelationId, roleid, 0);
 
-		/* Don't leak the syscache tuple */
-		ReleaseSysCache(tuple);
-
-		/*
-		 * Lock the role, so nobody can add dependencies to her while we drop
-		 * her.  We keep the lock until the end of transaction.
-		 */
-		LockSharedObject(AuthIdRelationId, roleid, 0, AccessExclusiveLock);
-
 		/*
 		 * If there is a pg_auth_members entry that has one of the roles to be
 		 * dropped as the roleid or member, it should be silently removed, but
@@ -1484,6 +1582,21 @@ RenameRole(const char *oldname, const char *newname)
 	return address;
 }
 
+/*
+ * Before acquiring a role lock for GRANT/REVOKE, check that the current user
+ * has authorization to grant/revoke membership in the specified role.
+ */
+static void
+RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+							 Oid oldroleid, void *callback_arg)
+{
+	bool		is_grant = *((bool *) callback_arg);
+	Oid			currentUserId = GetUserId();
+
+	check_role_membership_authorization(currentUserId, roleid, is_grant);
+}
+
+
 /*
  * GrantRoleStmt
  *
@@ -1568,9 +1681,14 @@ GrantRole(ParseState *pstate, GrantRoleStmt *stmt)
 					(errcode(ERRCODE_INVALID_GRANT_OPERATION),
 					 errmsg("column names cannot be included in GRANT/REVOKE ROLE")));
 
-		roleid = get_role_oid(rolename, false);
-		check_role_membership_authorization(currentUserId,
-											roleid, stmt->is_grant);
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP commits between name resolution
+		 * and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(rolename, ShareUpdateExclusiveLock, false,
+								RoleNameCallbackForGrantRole, &stmt->is_grant);
 		if (stmt->is_grant)
 			AddRoleMems(currentUserId, rolename, roleid,
 						stmt->grantee_roles, grantee_ids,
diff --git a/src/include/commands/user.h b/src/include/commands/user.h
index 97dcb93791b..17263452250 100644
--- a/src/include/commands/user.h
+++ b/src/include/commands/user.h
@@ -21,6 +21,15 @@
 extern PGDLLIMPORT int Password_encryption; /* values from enum PasswordType */
 extern PGDLLIMPORT char *createrole_self_grant;
 
+/* Callback for RoleNameGetOid, invoked after name resolution but before locking */
+typedef void (*RoleNameGetOidCallback) (const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+
+extern Oid	RoleNameGetOid(const char *rolename, LOCKMODE lockmode,
+						   bool missing_ok,
+						   RoleNameGetOidCallback callback,
+						   void *callback_arg);
+
 /* Hook to check passwords in CreateRole() and AlterRole() */
 typedef void (*check_password_hook_type) (const char *username, const char *shadow_pass, PasswordType password_type, Datum validuntil_time, bool validuntil_null);
 
-- 
2.34.1


--xlWYagR37YwJXkF8--





^ permalink  raw  reply  [nested|flat] 173+ messages in thread

* [PATCH v1] Add RoleNameGetOid() with invalidation-based retry loop for DropRole()/GrantRole()
@ 2026-07-04 04:47 Bertrand Drouvot <[email protected]>
  0 siblings, 0 replies; 173+ messages in thread

From: Bertrand Drouvot @ 2026-07-04 04:47 UTC (permalink / raw)

DropRole() and GrantRole() resolve the role name to an OID before acquiring
LockSharedObject() on the role. A concurrent session that commits a DROP ROLE
between the read and the lock acquisition leaves the first session acting on a
stale OID.

This commit fixes the races by using the same approach as RangeVarGetRelidExtended():
It encapsulates name resolution, permission checking (via a caller-supplied callback),
and lock acquisition inside a retry loop driven by SharedInvalidMessageCounter.
If invalidation messages arrive between name resolution and locking, indicating
concurrent DDL, the function retries.

The lock is kept across retries and only released if the name resolves to a
different OID on the next iteration.

Two callbacks are provided:
- RoleNameCallbackForDropRole(): checks current/session user, superuser attribute,
and ADMIN OPTION privilege before locking. This is similar to what DropRole() is
currently doing before LockSharedObject().
- RoleNameCallbackForGrantRole(): calls check_role_membership_authorization() to
verify the current user can grant/revoke membership. This is similar to what GrantRole()
is currently doing before calling AddRoleMems()/DelRoleMems().

DropRole() and GrantRole() now call RoleNameGetOid() with appropriate lock
levels.

AlterRole() does not need the fix because it calls CatalogTupleUpdate() on the
pg_authid tuple before AddRoleMems(), which blocks a concurrent DROP ROLE.

Author: Bertrand Drouvot <[email protected]>
Reviewed-by:
Discussion:
---
 src/backend/commands/user.c | 246 ++++++++++++++++++++++++++----------
 src/include/commands/user.h |   9 ++
 2 files changed, 191 insertions(+), 64 deletions(-)
  95.5% src/backend/commands/
   4.4% src/include/commands/

diff --git a/src/backend/commands/user.c b/src/backend/commands/user.c
index be11c49f919..6d8e2fa8813 100644
--- a/src/backend/commands/user.c
+++ b/src/backend/commands/user.c
@@ -34,6 +34,7 @@
 #include "miscadmin.h"
 #include "port/pg_bitutils.h"
 #include "storage/lmgr.h"
+#include "storage/sinval.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
 #include "utils/catcache.h"
@@ -116,6 +117,10 @@ static void plan_recursive_revoke(CatCList *memlist,
 								  bool revoke_admin_option_only,
 								  DropBehavior behavior);
 static void InitGrantRoleOptions(GrantRoleOptions *popt);
+static void RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+static void RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+										 Oid oldroleid, void *callback_arg);
 
 
 /* Check if current user has createrole privileges */
@@ -126,6 +131,92 @@ have_createrole_privilege(void)
 }
 
 
+/*
+ * RoleNameGetOid
+ *		Given a role name, look up its OID, lock it, and return the OID.
+ *
+ * This follows the same pattern as RangeVarGetRelidExtended():
+ * name resolution, permission check (via callback), and lock acquisition are
+ * performed inside a retry loop. If invalidation messages arrive during the
+ * process (indicating concurrent DDL), we retry to ensure the name still
+ * resolves to the same OID.
+ *
+ * The callback is invoked before locking, giving callers a chance to check
+ * permissions. It receives the current rolename, the resolved OID, the
+ * previous OID (InvalidOid on first iteration), and a caller-supplied arg.
+ * If the callback raises an error, the function aborts without locking.
+ *
+ * If missing_ok is true and the role does not exist, returns InvalidOid.
+ * Otherwise, raises an error.
+ */
+Oid
+RoleNameGetOid(const char *rolename, LOCKMODE lockmode, bool missing_ok,
+			   RoleNameGetOidCallback callback, void *callback_arg)
+{
+	uint64		inval_count;
+	Oid			roleid;
+	Oid			oldroleid = InvalidOid;
+	bool		retry = false;
+
+	for (;;)
+	{
+		/*
+		 * Remember the current invalidation count so we can detect concurrent
+		 * DDL after locking.
+		 */
+		inval_count = SharedInvalidMessageCounter;
+
+		/* Look up the role name */
+		roleid = get_role_oid(rolename, true);
+
+		if (!OidIsValid(roleid))
+		{
+			if (!missing_ok)
+				ereport(ERROR,
+						(errcode(ERRCODE_UNDEFINED_OBJECT),
+						 errmsg("role \"%s\" does not exist", rolename)));
+			return InvalidOid;
+		}
+
+		/*
+		 * Invoke caller-supplied callback before locking. This is a good
+		 * place to check permissions: we haven't taken the lock yet, but we
+		 * know the OID we intend to lock. If concurrent DDL changes things,
+		 * the callback will be invoked again on the next iteration.
+		 */
+		if (callback)
+			callback(rolename, roleid, oldroleid, callback_arg);
+
+		/*
+		 * If upon retry we get back the same OID, the invalidation messages
+		 * did not change the final answer. So we're done.
+		 *
+		 * If we got a different OID, we've locked the role that used to have
+		 * this name rather than the one that does now. Release the old lock.
+		 */
+		if (retry)
+		{
+			if (roleid == oldroleid)
+				break;
+			UnlockSharedObject(AuthIdRelationId, oldroleid, 0, lockmode);
+		}
+
+		/* Lock the role */
+		LockSharedObject(AuthIdRelationId, roleid, 0, lockmode);
+
+		/* If no invalidation messages were processed, we're done */
+		if (inval_count == SharedInvalidMessageCounter)
+			break;
+
+		/* Something may have changed, retry */
+		retry = true;
+		oldroleid = roleid;
+	}
+
+	return roleid;
+}
+
+
 /*
  * CREATE ROLE
  */
@@ -1090,6 +1181,59 @@ AlterRoleSet(AlterRoleSetStmt *stmt)
 }
 
 
+/*
+ * Before acquiring a role lock for DROP ROLE, check that the role is not the
+ * current/session user and that the caller has sufficient privileges to drop it.
+ */
+static void
+RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+							Oid oldroleid, void *callback_arg)
+{
+	HeapTuple	tuple;
+	Form_pg_authid roleform;
+
+	if (roleid == GetUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetOuterUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetSessionUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("session user cannot be dropped")));
+
+	tuple = SearchSysCache1(AUTHOID, ObjectIdGetDatum(roleid));
+	if (!HeapTupleIsValid(tuple))
+		ereport(ERROR,
+				(errcode(ERRCODE_UNDEFINED_OBJECT),
+				 errmsg("role \"%s\" does not exist", rolename)));
+
+	roleform = (Form_pg_authid) GETSTRUCT(tuple);
+
+	/*
+	 * For safety's sake, we allow createrole holders to drop ordinary roles
+	 * but not superuser roles, and only if they also have ADMIN OPTION.
+	 */
+	if (roleform->rolsuper && !superuser())
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
+						   "SUPERUSER", "SUPERUSER")));
+	if (!is_admin_of_role(GetUserId(), roleid))
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
+						   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
+
+	ReleaseSysCache(tuple);
+}
+
+
 /*
  * DROP ROLE
  */
@@ -1119,9 +1263,7 @@ DropRole(DropRoleStmt *stmt)
 	{
 		RoleSpec   *rolspec = lfirst(item);
 		char	   *role;
-		HeapTuple	tuple,
-					tmp_tuple;
-		Form_pg_authid roleform;
+		HeapTuple	tmp_tuple;
 		ScanKeyData scankey;
 		SysScanDesc sscan;
 		Oid			roleid;
@@ -1132,71 +1274,27 @@ DropRole(DropRoleStmt *stmt)
 					 errmsg("cannot use special role specifier in DROP ROLE")));
 		role = rolspec->rolename;
 
-		tuple = SearchSysCache1(AUTHNAME, PointerGetDatum(role));
-		if (!HeapTupleIsValid(tuple))
-		{
-			if (!stmt->missing_ok)
-			{
-				ereport(ERROR,
-						(errcode(ERRCODE_UNDEFINED_OBJECT),
-						 errmsg("role \"%s\" does not exist", role)));
-			}
-			else
-			{
-				ereport(NOTICE,
-						(errmsg("role \"%s\" does not exist, skipping",
-								role)));
-			}
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP or ALTER commits between name
+		 * resolution and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(role, AccessExclusiveLock, stmt->missing_ok,
+								RoleNameCallbackForDropRole, NULL);
 
+		if (!OidIsValid(roleid))
+		{
+			/* missing_ok case: role doesn't exist */
+			ereport(NOTICE,
+					(errmsg("role \"%s\" does not exist, skipping",
+							role)));
 			continue;
 		}
 
-		roleform = (Form_pg_authid) GETSTRUCT(tuple);
-		roleid = roleform->oid;
-
-		if (roleid == GetUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetOuterUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetSessionUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("session user cannot be dropped")));
-
-		/*
-		 * For safety's sake, we allow createrole holders to drop ordinary
-		 * roles but not superuser roles, and only if they also have ADMIN
-		 * OPTION.
-		 */
-		if (roleform->rolsuper && !superuser())
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
-							   "SUPERUSER", "SUPERUSER")));
-		if (!is_admin_of_role(GetUserId(), roleid))
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
-							   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
-
 		/* DROP hook for the role being removed */
 		InvokeObjectDropHook(AuthIdRelationId, roleid, 0);
 
-		/* Don't leak the syscache tuple */
-		ReleaseSysCache(tuple);
-
-		/*
-		 * Lock the role, so nobody can add dependencies to her while we drop
-		 * her.  We keep the lock until the end of transaction.
-		 */
-		LockSharedObject(AuthIdRelationId, roleid, 0, AccessExclusiveLock);
-
 		/*
 		 * If there is a pg_auth_members entry that has one of the roles to be
 		 * dropped as the roleid or member, it should be silently removed, but
@@ -1484,6 +1582,21 @@ RenameRole(const char *oldname, const char *newname)
 	return address;
 }
 
+/*
+ * Before acquiring a role lock for GRANT/REVOKE, check that the current user
+ * has authorization to grant/revoke membership in the specified role.
+ */
+static void
+RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+							 Oid oldroleid, void *callback_arg)
+{
+	bool		is_grant = *((bool *) callback_arg);
+	Oid			currentUserId = GetUserId();
+
+	check_role_membership_authorization(currentUserId, roleid, is_grant);
+}
+
+
 /*
  * GrantRoleStmt
  *
@@ -1568,9 +1681,14 @@ GrantRole(ParseState *pstate, GrantRoleStmt *stmt)
 					(errcode(ERRCODE_INVALID_GRANT_OPERATION),
 					 errmsg("column names cannot be included in GRANT/REVOKE ROLE")));
 
-		roleid = get_role_oid(rolename, false);
-		check_role_membership_authorization(currentUserId,
-											roleid, stmt->is_grant);
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP commits between name resolution
+		 * and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(rolename, ShareUpdateExclusiveLock, false,
+								RoleNameCallbackForGrantRole, &stmt->is_grant);
 		if (stmt->is_grant)
 			AddRoleMems(currentUserId, rolename, roleid,
 						stmt->grantee_roles, grantee_ids,
diff --git a/src/include/commands/user.h b/src/include/commands/user.h
index 97dcb93791b..17263452250 100644
--- a/src/include/commands/user.h
+++ b/src/include/commands/user.h
@@ -21,6 +21,15 @@
 extern PGDLLIMPORT int Password_encryption; /* values from enum PasswordType */
 extern PGDLLIMPORT char *createrole_self_grant;
 
+/* Callback for RoleNameGetOid, invoked after name resolution but before locking */
+typedef void (*RoleNameGetOidCallback) (const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+
+extern Oid	RoleNameGetOid(const char *rolename, LOCKMODE lockmode,
+						   bool missing_ok,
+						   RoleNameGetOidCallback callback,
+						   void *callback_arg);
+
 /* Hook to check passwords in CreateRole() and AlterRole() */
 typedef void (*check_password_hook_type) (const char *username, const char *shadow_pass, PasswordType password_type, Datum validuntil_time, bool validuntil_null);
 
-- 
2.34.1


--xlWYagR37YwJXkF8--





^ permalink  raw  reply  [nested|flat] 173+ messages in thread

* [PATCH v1] Add RoleNameGetOid() with invalidation-based retry loop for DropRole()/GrantRole()
@ 2026-07-04 04:47 Bertrand Drouvot <[email protected]>
  0 siblings, 0 replies; 173+ messages in thread

From: Bertrand Drouvot @ 2026-07-04 04:47 UTC (permalink / raw)

DropRole() and GrantRole() resolve the role name to an OID before acquiring
LockSharedObject() on the role. A concurrent session that commits a DROP ROLE
between the read and the lock acquisition leaves the first session acting on a
stale OID.

This commit fixes the races by using the same approach as RangeVarGetRelidExtended():
It encapsulates name resolution, permission checking (via a caller-supplied callback),
and lock acquisition inside a retry loop driven by SharedInvalidMessageCounter.
If invalidation messages arrive between name resolution and locking, indicating
concurrent DDL, the function retries.

The lock is kept across retries and only released if the name resolves to a
different OID on the next iteration.

Two callbacks are provided:
- RoleNameCallbackForDropRole(): checks current/session user, superuser attribute,
and ADMIN OPTION privilege before locking. This is similar to what DropRole() is
currently doing before LockSharedObject().
- RoleNameCallbackForGrantRole(): calls check_role_membership_authorization() to
verify the current user can grant/revoke membership. This is similar to what GrantRole()
is currently doing before calling AddRoleMems()/DelRoleMems().

DropRole() and GrantRole() now call RoleNameGetOid() with appropriate lock
levels.

AlterRole() does not need the fix because it calls CatalogTupleUpdate() on the
pg_authid tuple before AddRoleMems(), which blocks a concurrent DROP ROLE.

Author: Bertrand Drouvot <[email protected]>
Reviewed-by:
Discussion:
---
 src/backend/commands/user.c | 246 ++++++++++++++++++++++++++----------
 src/include/commands/user.h |   9 ++
 2 files changed, 191 insertions(+), 64 deletions(-)
  95.5% src/backend/commands/
   4.4% src/include/commands/

diff --git a/src/backend/commands/user.c b/src/backend/commands/user.c
index be11c49f919..6d8e2fa8813 100644
--- a/src/backend/commands/user.c
+++ b/src/backend/commands/user.c
@@ -34,6 +34,7 @@
 #include "miscadmin.h"
 #include "port/pg_bitutils.h"
 #include "storage/lmgr.h"
+#include "storage/sinval.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
 #include "utils/catcache.h"
@@ -116,6 +117,10 @@ static void plan_recursive_revoke(CatCList *memlist,
 								  bool revoke_admin_option_only,
 								  DropBehavior behavior);
 static void InitGrantRoleOptions(GrantRoleOptions *popt);
+static void RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+static void RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+										 Oid oldroleid, void *callback_arg);
 
 
 /* Check if current user has createrole privileges */
@@ -126,6 +131,92 @@ have_createrole_privilege(void)
 }
 
 
+/*
+ * RoleNameGetOid
+ *		Given a role name, look up its OID, lock it, and return the OID.
+ *
+ * This follows the same pattern as RangeVarGetRelidExtended():
+ * name resolution, permission check (via callback), and lock acquisition are
+ * performed inside a retry loop. If invalidation messages arrive during the
+ * process (indicating concurrent DDL), we retry to ensure the name still
+ * resolves to the same OID.
+ *
+ * The callback is invoked before locking, giving callers a chance to check
+ * permissions. It receives the current rolename, the resolved OID, the
+ * previous OID (InvalidOid on first iteration), and a caller-supplied arg.
+ * If the callback raises an error, the function aborts without locking.
+ *
+ * If missing_ok is true and the role does not exist, returns InvalidOid.
+ * Otherwise, raises an error.
+ */
+Oid
+RoleNameGetOid(const char *rolename, LOCKMODE lockmode, bool missing_ok,
+			   RoleNameGetOidCallback callback, void *callback_arg)
+{
+	uint64		inval_count;
+	Oid			roleid;
+	Oid			oldroleid = InvalidOid;
+	bool		retry = false;
+
+	for (;;)
+	{
+		/*
+		 * Remember the current invalidation count so we can detect concurrent
+		 * DDL after locking.
+		 */
+		inval_count = SharedInvalidMessageCounter;
+
+		/* Look up the role name */
+		roleid = get_role_oid(rolename, true);
+
+		if (!OidIsValid(roleid))
+		{
+			if (!missing_ok)
+				ereport(ERROR,
+						(errcode(ERRCODE_UNDEFINED_OBJECT),
+						 errmsg("role \"%s\" does not exist", rolename)));
+			return InvalidOid;
+		}
+
+		/*
+		 * Invoke caller-supplied callback before locking. This is a good
+		 * place to check permissions: we haven't taken the lock yet, but we
+		 * know the OID we intend to lock. If concurrent DDL changes things,
+		 * the callback will be invoked again on the next iteration.
+		 */
+		if (callback)
+			callback(rolename, roleid, oldroleid, callback_arg);
+
+		/*
+		 * If upon retry we get back the same OID, the invalidation messages
+		 * did not change the final answer. So we're done.
+		 *
+		 * If we got a different OID, we've locked the role that used to have
+		 * this name rather than the one that does now. Release the old lock.
+		 */
+		if (retry)
+		{
+			if (roleid == oldroleid)
+				break;
+			UnlockSharedObject(AuthIdRelationId, oldroleid, 0, lockmode);
+		}
+
+		/* Lock the role */
+		LockSharedObject(AuthIdRelationId, roleid, 0, lockmode);
+
+		/* If no invalidation messages were processed, we're done */
+		if (inval_count == SharedInvalidMessageCounter)
+			break;
+
+		/* Something may have changed, retry */
+		retry = true;
+		oldroleid = roleid;
+	}
+
+	return roleid;
+}
+
+
 /*
  * CREATE ROLE
  */
@@ -1090,6 +1181,59 @@ AlterRoleSet(AlterRoleSetStmt *stmt)
 }
 
 
+/*
+ * Before acquiring a role lock for DROP ROLE, check that the role is not the
+ * current/session user and that the caller has sufficient privileges to drop it.
+ */
+static void
+RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+							Oid oldroleid, void *callback_arg)
+{
+	HeapTuple	tuple;
+	Form_pg_authid roleform;
+
+	if (roleid == GetUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetOuterUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetSessionUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("session user cannot be dropped")));
+
+	tuple = SearchSysCache1(AUTHOID, ObjectIdGetDatum(roleid));
+	if (!HeapTupleIsValid(tuple))
+		ereport(ERROR,
+				(errcode(ERRCODE_UNDEFINED_OBJECT),
+				 errmsg("role \"%s\" does not exist", rolename)));
+
+	roleform = (Form_pg_authid) GETSTRUCT(tuple);
+
+	/*
+	 * For safety's sake, we allow createrole holders to drop ordinary roles
+	 * but not superuser roles, and only if they also have ADMIN OPTION.
+	 */
+	if (roleform->rolsuper && !superuser())
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
+						   "SUPERUSER", "SUPERUSER")));
+	if (!is_admin_of_role(GetUserId(), roleid))
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
+						   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
+
+	ReleaseSysCache(tuple);
+}
+
+
 /*
  * DROP ROLE
  */
@@ -1119,9 +1263,7 @@ DropRole(DropRoleStmt *stmt)
 	{
 		RoleSpec   *rolspec = lfirst(item);
 		char	   *role;
-		HeapTuple	tuple,
-					tmp_tuple;
-		Form_pg_authid roleform;
+		HeapTuple	tmp_tuple;
 		ScanKeyData scankey;
 		SysScanDesc sscan;
 		Oid			roleid;
@@ -1132,71 +1274,27 @@ DropRole(DropRoleStmt *stmt)
 					 errmsg("cannot use special role specifier in DROP ROLE")));
 		role = rolspec->rolename;
 
-		tuple = SearchSysCache1(AUTHNAME, PointerGetDatum(role));
-		if (!HeapTupleIsValid(tuple))
-		{
-			if (!stmt->missing_ok)
-			{
-				ereport(ERROR,
-						(errcode(ERRCODE_UNDEFINED_OBJECT),
-						 errmsg("role \"%s\" does not exist", role)));
-			}
-			else
-			{
-				ereport(NOTICE,
-						(errmsg("role \"%s\" does not exist, skipping",
-								role)));
-			}
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP or ALTER commits between name
+		 * resolution and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(role, AccessExclusiveLock, stmt->missing_ok,
+								RoleNameCallbackForDropRole, NULL);
 
+		if (!OidIsValid(roleid))
+		{
+			/* missing_ok case: role doesn't exist */
+			ereport(NOTICE,
+					(errmsg("role \"%s\" does not exist, skipping",
+							role)));
 			continue;
 		}
 
-		roleform = (Form_pg_authid) GETSTRUCT(tuple);
-		roleid = roleform->oid;
-
-		if (roleid == GetUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetOuterUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetSessionUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("session user cannot be dropped")));
-
-		/*
-		 * For safety's sake, we allow createrole holders to drop ordinary
-		 * roles but not superuser roles, and only if they also have ADMIN
-		 * OPTION.
-		 */
-		if (roleform->rolsuper && !superuser())
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
-							   "SUPERUSER", "SUPERUSER")));
-		if (!is_admin_of_role(GetUserId(), roleid))
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
-							   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
-
 		/* DROP hook for the role being removed */
 		InvokeObjectDropHook(AuthIdRelationId, roleid, 0);
 
-		/* Don't leak the syscache tuple */
-		ReleaseSysCache(tuple);
-
-		/*
-		 * Lock the role, so nobody can add dependencies to her while we drop
-		 * her.  We keep the lock until the end of transaction.
-		 */
-		LockSharedObject(AuthIdRelationId, roleid, 0, AccessExclusiveLock);
-
 		/*
 		 * If there is a pg_auth_members entry that has one of the roles to be
 		 * dropped as the roleid or member, it should be silently removed, but
@@ -1484,6 +1582,21 @@ RenameRole(const char *oldname, const char *newname)
 	return address;
 }
 
+/*
+ * Before acquiring a role lock for GRANT/REVOKE, check that the current user
+ * has authorization to grant/revoke membership in the specified role.
+ */
+static void
+RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+							 Oid oldroleid, void *callback_arg)
+{
+	bool		is_grant = *((bool *) callback_arg);
+	Oid			currentUserId = GetUserId();
+
+	check_role_membership_authorization(currentUserId, roleid, is_grant);
+}
+
+
 /*
  * GrantRoleStmt
  *
@@ -1568,9 +1681,14 @@ GrantRole(ParseState *pstate, GrantRoleStmt *stmt)
 					(errcode(ERRCODE_INVALID_GRANT_OPERATION),
 					 errmsg("column names cannot be included in GRANT/REVOKE ROLE")));
 
-		roleid = get_role_oid(rolename, false);
-		check_role_membership_authorization(currentUserId,
-											roleid, stmt->is_grant);
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP commits between name resolution
+		 * and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(rolename, ShareUpdateExclusiveLock, false,
+								RoleNameCallbackForGrantRole, &stmt->is_grant);
 		if (stmt->is_grant)
 			AddRoleMems(currentUserId, rolename, roleid,
 						stmt->grantee_roles, grantee_ids,
diff --git a/src/include/commands/user.h b/src/include/commands/user.h
index 97dcb93791b..17263452250 100644
--- a/src/include/commands/user.h
+++ b/src/include/commands/user.h
@@ -21,6 +21,15 @@
 extern PGDLLIMPORT int Password_encryption; /* values from enum PasswordType */
 extern PGDLLIMPORT char *createrole_self_grant;
 
+/* Callback for RoleNameGetOid, invoked after name resolution but before locking */
+typedef void (*RoleNameGetOidCallback) (const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+
+extern Oid	RoleNameGetOid(const char *rolename, LOCKMODE lockmode,
+						   bool missing_ok,
+						   RoleNameGetOidCallback callback,
+						   void *callback_arg);
+
 /* Hook to check passwords in CreateRole() and AlterRole() */
 typedef void (*check_password_hook_type) (const char *username, const char *shadow_pass, PasswordType password_type, Datum validuntil_time, bool validuntil_null);
 
-- 
2.34.1


--xlWYagR37YwJXkF8--





^ permalink  raw  reply  [nested|flat] 173+ messages in thread

* [PATCH v1] Add RoleNameGetOid() with invalidation-based retry loop for DropRole()/GrantRole()
@ 2026-07-04 04:47 Bertrand Drouvot <[email protected]>
  0 siblings, 0 replies; 173+ messages in thread

From: Bertrand Drouvot @ 2026-07-04 04:47 UTC (permalink / raw)

DropRole() and GrantRole() resolve the role name to an OID before acquiring
LockSharedObject() on the role. A concurrent session that commits a DROP ROLE
between the read and the lock acquisition leaves the first session acting on a
stale OID.

This commit fixes the races by using the same approach as RangeVarGetRelidExtended():
It encapsulates name resolution, permission checking (via a caller-supplied callback),
and lock acquisition inside a retry loop driven by SharedInvalidMessageCounter.
If invalidation messages arrive between name resolution and locking, indicating
concurrent DDL, the function retries.

The lock is kept across retries and only released if the name resolves to a
different OID on the next iteration.

Two callbacks are provided:
- RoleNameCallbackForDropRole(): checks current/session user, superuser attribute,
and ADMIN OPTION privilege before locking. This is similar to what DropRole() is
currently doing before LockSharedObject().
- RoleNameCallbackForGrantRole(): calls check_role_membership_authorization() to
verify the current user can grant/revoke membership. This is similar to what GrantRole()
is currently doing before calling AddRoleMems()/DelRoleMems().

DropRole() and GrantRole() now call RoleNameGetOid() with appropriate lock
levels.

AlterRole() does not need the fix because it calls CatalogTupleUpdate() on the
pg_authid tuple before AddRoleMems(), which blocks a concurrent DROP ROLE.

Author: Bertrand Drouvot <[email protected]>
Reviewed-by:
Discussion:
---
 src/backend/commands/user.c | 246 ++++++++++++++++++++++++++----------
 src/include/commands/user.h |   9 ++
 2 files changed, 191 insertions(+), 64 deletions(-)
  95.5% src/backend/commands/
   4.4% src/include/commands/

diff --git a/src/backend/commands/user.c b/src/backend/commands/user.c
index be11c49f919..6d8e2fa8813 100644
--- a/src/backend/commands/user.c
+++ b/src/backend/commands/user.c
@@ -34,6 +34,7 @@
 #include "miscadmin.h"
 #include "port/pg_bitutils.h"
 #include "storage/lmgr.h"
+#include "storage/sinval.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
 #include "utils/catcache.h"
@@ -116,6 +117,10 @@ static void plan_recursive_revoke(CatCList *memlist,
 								  bool revoke_admin_option_only,
 								  DropBehavior behavior);
 static void InitGrantRoleOptions(GrantRoleOptions *popt);
+static void RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+static void RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+										 Oid oldroleid, void *callback_arg);
 
 
 /* Check if current user has createrole privileges */
@@ -126,6 +131,92 @@ have_createrole_privilege(void)
 }
 
 
+/*
+ * RoleNameGetOid
+ *		Given a role name, look up its OID, lock it, and return the OID.
+ *
+ * This follows the same pattern as RangeVarGetRelidExtended():
+ * name resolution, permission check (via callback), and lock acquisition are
+ * performed inside a retry loop. If invalidation messages arrive during the
+ * process (indicating concurrent DDL), we retry to ensure the name still
+ * resolves to the same OID.
+ *
+ * The callback is invoked before locking, giving callers a chance to check
+ * permissions. It receives the current rolename, the resolved OID, the
+ * previous OID (InvalidOid on first iteration), and a caller-supplied arg.
+ * If the callback raises an error, the function aborts without locking.
+ *
+ * If missing_ok is true and the role does not exist, returns InvalidOid.
+ * Otherwise, raises an error.
+ */
+Oid
+RoleNameGetOid(const char *rolename, LOCKMODE lockmode, bool missing_ok,
+			   RoleNameGetOidCallback callback, void *callback_arg)
+{
+	uint64		inval_count;
+	Oid			roleid;
+	Oid			oldroleid = InvalidOid;
+	bool		retry = false;
+
+	for (;;)
+	{
+		/*
+		 * Remember the current invalidation count so we can detect concurrent
+		 * DDL after locking.
+		 */
+		inval_count = SharedInvalidMessageCounter;
+
+		/* Look up the role name */
+		roleid = get_role_oid(rolename, true);
+
+		if (!OidIsValid(roleid))
+		{
+			if (!missing_ok)
+				ereport(ERROR,
+						(errcode(ERRCODE_UNDEFINED_OBJECT),
+						 errmsg("role \"%s\" does not exist", rolename)));
+			return InvalidOid;
+		}
+
+		/*
+		 * Invoke caller-supplied callback before locking. This is a good
+		 * place to check permissions: we haven't taken the lock yet, but we
+		 * know the OID we intend to lock. If concurrent DDL changes things,
+		 * the callback will be invoked again on the next iteration.
+		 */
+		if (callback)
+			callback(rolename, roleid, oldroleid, callback_arg);
+
+		/*
+		 * If upon retry we get back the same OID, the invalidation messages
+		 * did not change the final answer. So we're done.
+		 *
+		 * If we got a different OID, we've locked the role that used to have
+		 * this name rather than the one that does now. Release the old lock.
+		 */
+		if (retry)
+		{
+			if (roleid == oldroleid)
+				break;
+			UnlockSharedObject(AuthIdRelationId, oldroleid, 0, lockmode);
+		}
+
+		/* Lock the role */
+		LockSharedObject(AuthIdRelationId, roleid, 0, lockmode);
+
+		/* If no invalidation messages were processed, we're done */
+		if (inval_count == SharedInvalidMessageCounter)
+			break;
+
+		/* Something may have changed, retry */
+		retry = true;
+		oldroleid = roleid;
+	}
+
+	return roleid;
+}
+
+
 /*
  * CREATE ROLE
  */
@@ -1090,6 +1181,59 @@ AlterRoleSet(AlterRoleSetStmt *stmt)
 }
 
 
+/*
+ * Before acquiring a role lock for DROP ROLE, check that the role is not the
+ * current/session user and that the caller has sufficient privileges to drop it.
+ */
+static void
+RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+							Oid oldroleid, void *callback_arg)
+{
+	HeapTuple	tuple;
+	Form_pg_authid roleform;
+
+	if (roleid == GetUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetOuterUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetSessionUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("session user cannot be dropped")));
+
+	tuple = SearchSysCache1(AUTHOID, ObjectIdGetDatum(roleid));
+	if (!HeapTupleIsValid(tuple))
+		ereport(ERROR,
+				(errcode(ERRCODE_UNDEFINED_OBJECT),
+				 errmsg("role \"%s\" does not exist", rolename)));
+
+	roleform = (Form_pg_authid) GETSTRUCT(tuple);
+
+	/*
+	 * For safety's sake, we allow createrole holders to drop ordinary roles
+	 * but not superuser roles, and only if they also have ADMIN OPTION.
+	 */
+	if (roleform->rolsuper && !superuser())
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
+						   "SUPERUSER", "SUPERUSER")));
+	if (!is_admin_of_role(GetUserId(), roleid))
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
+						   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
+
+	ReleaseSysCache(tuple);
+}
+
+
 /*
  * DROP ROLE
  */
@@ -1119,9 +1263,7 @@ DropRole(DropRoleStmt *stmt)
 	{
 		RoleSpec   *rolspec = lfirst(item);
 		char	   *role;
-		HeapTuple	tuple,
-					tmp_tuple;
-		Form_pg_authid roleform;
+		HeapTuple	tmp_tuple;
 		ScanKeyData scankey;
 		SysScanDesc sscan;
 		Oid			roleid;
@@ -1132,71 +1274,27 @@ DropRole(DropRoleStmt *stmt)
 					 errmsg("cannot use special role specifier in DROP ROLE")));
 		role = rolspec->rolename;
 
-		tuple = SearchSysCache1(AUTHNAME, PointerGetDatum(role));
-		if (!HeapTupleIsValid(tuple))
-		{
-			if (!stmt->missing_ok)
-			{
-				ereport(ERROR,
-						(errcode(ERRCODE_UNDEFINED_OBJECT),
-						 errmsg("role \"%s\" does not exist", role)));
-			}
-			else
-			{
-				ereport(NOTICE,
-						(errmsg("role \"%s\" does not exist, skipping",
-								role)));
-			}
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP or ALTER commits between name
+		 * resolution and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(role, AccessExclusiveLock, stmt->missing_ok,
+								RoleNameCallbackForDropRole, NULL);
 
+		if (!OidIsValid(roleid))
+		{
+			/* missing_ok case: role doesn't exist */
+			ereport(NOTICE,
+					(errmsg("role \"%s\" does not exist, skipping",
+							role)));
 			continue;
 		}
 
-		roleform = (Form_pg_authid) GETSTRUCT(tuple);
-		roleid = roleform->oid;
-
-		if (roleid == GetUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetOuterUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetSessionUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("session user cannot be dropped")));
-
-		/*
-		 * For safety's sake, we allow createrole holders to drop ordinary
-		 * roles but not superuser roles, and only if they also have ADMIN
-		 * OPTION.
-		 */
-		if (roleform->rolsuper && !superuser())
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
-							   "SUPERUSER", "SUPERUSER")));
-		if (!is_admin_of_role(GetUserId(), roleid))
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
-							   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
-
 		/* DROP hook for the role being removed */
 		InvokeObjectDropHook(AuthIdRelationId, roleid, 0);
 
-		/* Don't leak the syscache tuple */
-		ReleaseSysCache(tuple);
-
-		/*
-		 * Lock the role, so nobody can add dependencies to her while we drop
-		 * her.  We keep the lock until the end of transaction.
-		 */
-		LockSharedObject(AuthIdRelationId, roleid, 0, AccessExclusiveLock);
-
 		/*
 		 * If there is a pg_auth_members entry that has one of the roles to be
 		 * dropped as the roleid or member, it should be silently removed, but
@@ -1484,6 +1582,21 @@ RenameRole(const char *oldname, const char *newname)
 	return address;
 }
 
+/*
+ * Before acquiring a role lock for GRANT/REVOKE, check that the current user
+ * has authorization to grant/revoke membership in the specified role.
+ */
+static void
+RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+							 Oid oldroleid, void *callback_arg)
+{
+	bool		is_grant = *((bool *) callback_arg);
+	Oid			currentUserId = GetUserId();
+
+	check_role_membership_authorization(currentUserId, roleid, is_grant);
+}
+
+
 /*
  * GrantRoleStmt
  *
@@ -1568,9 +1681,14 @@ GrantRole(ParseState *pstate, GrantRoleStmt *stmt)
 					(errcode(ERRCODE_INVALID_GRANT_OPERATION),
 					 errmsg("column names cannot be included in GRANT/REVOKE ROLE")));
 
-		roleid = get_role_oid(rolename, false);
-		check_role_membership_authorization(currentUserId,
-											roleid, stmt->is_grant);
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP commits between name resolution
+		 * and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(rolename, ShareUpdateExclusiveLock, false,
+								RoleNameCallbackForGrantRole, &stmt->is_grant);
 		if (stmt->is_grant)
 			AddRoleMems(currentUserId, rolename, roleid,
 						stmt->grantee_roles, grantee_ids,
diff --git a/src/include/commands/user.h b/src/include/commands/user.h
index 97dcb93791b..17263452250 100644
--- a/src/include/commands/user.h
+++ b/src/include/commands/user.h
@@ -21,6 +21,15 @@
 extern PGDLLIMPORT int Password_encryption; /* values from enum PasswordType */
 extern PGDLLIMPORT char *createrole_self_grant;
 
+/* Callback for RoleNameGetOid, invoked after name resolution but before locking */
+typedef void (*RoleNameGetOidCallback) (const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+
+extern Oid	RoleNameGetOid(const char *rolename, LOCKMODE lockmode,
+						   bool missing_ok,
+						   RoleNameGetOidCallback callback,
+						   void *callback_arg);
+
 /* Hook to check passwords in CreateRole() and AlterRole() */
 typedef void (*check_password_hook_type) (const char *username, const char *shadow_pass, PasswordType password_type, Datum validuntil_time, bool validuntil_null);
 
-- 
2.34.1


--xlWYagR37YwJXkF8--





^ permalink  raw  reply  [nested|flat] 173+ messages in thread

* [PATCH v1] Add RoleNameGetOid() with invalidation-based retry loop for DropRole()/GrantRole()
@ 2026-07-04 04:47 Bertrand Drouvot <[email protected]>
  0 siblings, 0 replies; 173+ messages in thread

From: Bertrand Drouvot @ 2026-07-04 04:47 UTC (permalink / raw)

DropRole() and GrantRole() resolve the role name to an OID before acquiring
LockSharedObject() on the role. A concurrent session that commits a DROP ROLE
between the read and the lock acquisition leaves the first session acting on a
stale OID.

This commit fixes the races by using the same approach as RangeVarGetRelidExtended():
It encapsulates name resolution, permission checking (via a caller-supplied callback),
and lock acquisition inside a retry loop driven by SharedInvalidMessageCounter.
If invalidation messages arrive between name resolution and locking, indicating
concurrent DDL, the function retries.

The lock is kept across retries and only released if the name resolves to a
different OID on the next iteration.

Two callbacks are provided:
- RoleNameCallbackForDropRole(): checks current/session user, superuser attribute,
and ADMIN OPTION privilege before locking. This is similar to what DropRole() is
currently doing before LockSharedObject().
- RoleNameCallbackForGrantRole(): calls check_role_membership_authorization() to
verify the current user can grant/revoke membership. This is similar to what GrantRole()
is currently doing before calling AddRoleMems()/DelRoleMems().

DropRole() and GrantRole() now call RoleNameGetOid() with appropriate lock
levels.

AlterRole() does not need the fix because it calls CatalogTupleUpdate() on the
pg_authid tuple before AddRoleMems(), which blocks a concurrent DROP ROLE.

Author: Bertrand Drouvot <[email protected]>
Reviewed-by:
Discussion:
---
 src/backend/commands/user.c | 246 ++++++++++++++++++++++++++----------
 src/include/commands/user.h |   9 ++
 2 files changed, 191 insertions(+), 64 deletions(-)
  95.5% src/backend/commands/
   4.4% src/include/commands/

diff --git a/src/backend/commands/user.c b/src/backend/commands/user.c
index be11c49f919..6d8e2fa8813 100644
--- a/src/backend/commands/user.c
+++ b/src/backend/commands/user.c
@@ -34,6 +34,7 @@
 #include "miscadmin.h"
 #include "port/pg_bitutils.h"
 #include "storage/lmgr.h"
+#include "storage/sinval.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
 #include "utils/catcache.h"
@@ -116,6 +117,10 @@ static void plan_recursive_revoke(CatCList *memlist,
 								  bool revoke_admin_option_only,
 								  DropBehavior behavior);
 static void InitGrantRoleOptions(GrantRoleOptions *popt);
+static void RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+static void RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+										 Oid oldroleid, void *callback_arg);
 
 
 /* Check if current user has createrole privileges */
@@ -126,6 +131,92 @@ have_createrole_privilege(void)
 }
 
 
+/*
+ * RoleNameGetOid
+ *		Given a role name, look up its OID, lock it, and return the OID.
+ *
+ * This follows the same pattern as RangeVarGetRelidExtended():
+ * name resolution, permission check (via callback), and lock acquisition are
+ * performed inside a retry loop. If invalidation messages arrive during the
+ * process (indicating concurrent DDL), we retry to ensure the name still
+ * resolves to the same OID.
+ *
+ * The callback is invoked before locking, giving callers a chance to check
+ * permissions. It receives the current rolename, the resolved OID, the
+ * previous OID (InvalidOid on first iteration), and a caller-supplied arg.
+ * If the callback raises an error, the function aborts without locking.
+ *
+ * If missing_ok is true and the role does not exist, returns InvalidOid.
+ * Otherwise, raises an error.
+ */
+Oid
+RoleNameGetOid(const char *rolename, LOCKMODE lockmode, bool missing_ok,
+			   RoleNameGetOidCallback callback, void *callback_arg)
+{
+	uint64		inval_count;
+	Oid			roleid;
+	Oid			oldroleid = InvalidOid;
+	bool		retry = false;
+
+	for (;;)
+	{
+		/*
+		 * Remember the current invalidation count so we can detect concurrent
+		 * DDL after locking.
+		 */
+		inval_count = SharedInvalidMessageCounter;
+
+		/* Look up the role name */
+		roleid = get_role_oid(rolename, true);
+
+		if (!OidIsValid(roleid))
+		{
+			if (!missing_ok)
+				ereport(ERROR,
+						(errcode(ERRCODE_UNDEFINED_OBJECT),
+						 errmsg("role \"%s\" does not exist", rolename)));
+			return InvalidOid;
+		}
+
+		/*
+		 * Invoke caller-supplied callback before locking. This is a good
+		 * place to check permissions: we haven't taken the lock yet, but we
+		 * know the OID we intend to lock. If concurrent DDL changes things,
+		 * the callback will be invoked again on the next iteration.
+		 */
+		if (callback)
+			callback(rolename, roleid, oldroleid, callback_arg);
+
+		/*
+		 * If upon retry we get back the same OID, the invalidation messages
+		 * did not change the final answer. So we're done.
+		 *
+		 * If we got a different OID, we've locked the role that used to have
+		 * this name rather than the one that does now. Release the old lock.
+		 */
+		if (retry)
+		{
+			if (roleid == oldroleid)
+				break;
+			UnlockSharedObject(AuthIdRelationId, oldroleid, 0, lockmode);
+		}
+
+		/* Lock the role */
+		LockSharedObject(AuthIdRelationId, roleid, 0, lockmode);
+
+		/* If no invalidation messages were processed, we're done */
+		if (inval_count == SharedInvalidMessageCounter)
+			break;
+
+		/* Something may have changed, retry */
+		retry = true;
+		oldroleid = roleid;
+	}
+
+	return roleid;
+}
+
+
 /*
  * CREATE ROLE
  */
@@ -1090,6 +1181,59 @@ AlterRoleSet(AlterRoleSetStmt *stmt)
 }
 
 
+/*
+ * Before acquiring a role lock for DROP ROLE, check that the role is not the
+ * current/session user and that the caller has sufficient privileges to drop it.
+ */
+static void
+RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+							Oid oldroleid, void *callback_arg)
+{
+	HeapTuple	tuple;
+	Form_pg_authid roleform;
+
+	if (roleid == GetUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetOuterUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetSessionUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("session user cannot be dropped")));
+
+	tuple = SearchSysCache1(AUTHOID, ObjectIdGetDatum(roleid));
+	if (!HeapTupleIsValid(tuple))
+		ereport(ERROR,
+				(errcode(ERRCODE_UNDEFINED_OBJECT),
+				 errmsg("role \"%s\" does not exist", rolename)));
+
+	roleform = (Form_pg_authid) GETSTRUCT(tuple);
+
+	/*
+	 * For safety's sake, we allow createrole holders to drop ordinary roles
+	 * but not superuser roles, and only if they also have ADMIN OPTION.
+	 */
+	if (roleform->rolsuper && !superuser())
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
+						   "SUPERUSER", "SUPERUSER")));
+	if (!is_admin_of_role(GetUserId(), roleid))
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
+						   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
+
+	ReleaseSysCache(tuple);
+}
+
+
 /*
  * DROP ROLE
  */
@@ -1119,9 +1263,7 @@ DropRole(DropRoleStmt *stmt)
 	{
 		RoleSpec   *rolspec = lfirst(item);
 		char	   *role;
-		HeapTuple	tuple,
-					tmp_tuple;
-		Form_pg_authid roleform;
+		HeapTuple	tmp_tuple;
 		ScanKeyData scankey;
 		SysScanDesc sscan;
 		Oid			roleid;
@@ -1132,71 +1274,27 @@ DropRole(DropRoleStmt *stmt)
 					 errmsg("cannot use special role specifier in DROP ROLE")));
 		role = rolspec->rolename;
 
-		tuple = SearchSysCache1(AUTHNAME, PointerGetDatum(role));
-		if (!HeapTupleIsValid(tuple))
-		{
-			if (!stmt->missing_ok)
-			{
-				ereport(ERROR,
-						(errcode(ERRCODE_UNDEFINED_OBJECT),
-						 errmsg("role \"%s\" does not exist", role)));
-			}
-			else
-			{
-				ereport(NOTICE,
-						(errmsg("role \"%s\" does not exist, skipping",
-								role)));
-			}
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP or ALTER commits between name
+		 * resolution and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(role, AccessExclusiveLock, stmt->missing_ok,
+								RoleNameCallbackForDropRole, NULL);
 
+		if (!OidIsValid(roleid))
+		{
+			/* missing_ok case: role doesn't exist */
+			ereport(NOTICE,
+					(errmsg("role \"%s\" does not exist, skipping",
+							role)));
 			continue;
 		}
 
-		roleform = (Form_pg_authid) GETSTRUCT(tuple);
-		roleid = roleform->oid;
-
-		if (roleid == GetUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetOuterUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetSessionUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("session user cannot be dropped")));
-
-		/*
-		 * For safety's sake, we allow createrole holders to drop ordinary
-		 * roles but not superuser roles, and only if they also have ADMIN
-		 * OPTION.
-		 */
-		if (roleform->rolsuper && !superuser())
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
-							   "SUPERUSER", "SUPERUSER")));
-		if (!is_admin_of_role(GetUserId(), roleid))
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
-							   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
-
 		/* DROP hook for the role being removed */
 		InvokeObjectDropHook(AuthIdRelationId, roleid, 0);
 
-		/* Don't leak the syscache tuple */
-		ReleaseSysCache(tuple);
-
-		/*
-		 * Lock the role, so nobody can add dependencies to her while we drop
-		 * her.  We keep the lock until the end of transaction.
-		 */
-		LockSharedObject(AuthIdRelationId, roleid, 0, AccessExclusiveLock);
-
 		/*
 		 * If there is a pg_auth_members entry that has one of the roles to be
 		 * dropped as the roleid or member, it should be silently removed, but
@@ -1484,6 +1582,21 @@ RenameRole(const char *oldname, const char *newname)
 	return address;
 }
 
+/*
+ * Before acquiring a role lock for GRANT/REVOKE, check that the current user
+ * has authorization to grant/revoke membership in the specified role.
+ */
+static void
+RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+							 Oid oldroleid, void *callback_arg)
+{
+	bool		is_grant = *((bool *) callback_arg);
+	Oid			currentUserId = GetUserId();
+
+	check_role_membership_authorization(currentUserId, roleid, is_grant);
+}
+
+
 /*
  * GrantRoleStmt
  *
@@ -1568,9 +1681,14 @@ GrantRole(ParseState *pstate, GrantRoleStmt *stmt)
 					(errcode(ERRCODE_INVALID_GRANT_OPERATION),
 					 errmsg("column names cannot be included in GRANT/REVOKE ROLE")));
 
-		roleid = get_role_oid(rolename, false);
-		check_role_membership_authorization(currentUserId,
-											roleid, stmt->is_grant);
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP commits between name resolution
+		 * and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(rolename, ShareUpdateExclusiveLock, false,
+								RoleNameCallbackForGrantRole, &stmt->is_grant);
 		if (stmt->is_grant)
 			AddRoleMems(currentUserId, rolename, roleid,
 						stmt->grantee_roles, grantee_ids,
diff --git a/src/include/commands/user.h b/src/include/commands/user.h
index 97dcb93791b..17263452250 100644
--- a/src/include/commands/user.h
+++ b/src/include/commands/user.h
@@ -21,6 +21,15 @@
 extern PGDLLIMPORT int Password_encryption; /* values from enum PasswordType */
 extern PGDLLIMPORT char *createrole_self_grant;
 
+/* Callback for RoleNameGetOid, invoked after name resolution but before locking */
+typedef void (*RoleNameGetOidCallback) (const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+
+extern Oid	RoleNameGetOid(const char *rolename, LOCKMODE lockmode,
+						   bool missing_ok,
+						   RoleNameGetOidCallback callback,
+						   void *callback_arg);
+
 /* Hook to check passwords in CreateRole() and AlterRole() */
 typedef void (*check_password_hook_type) (const char *username, const char *shadow_pass, PasswordType password_type, Datum validuntil_time, bool validuntil_null);
 
-- 
2.34.1


--xlWYagR37YwJXkF8--





^ permalink  raw  reply  [nested|flat] 173+ messages in thread

* [PATCH v1] Add RoleNameGetOid() with invalidation-based retry loop for DropRole()/GrantRole()
@ 2026-07-04 04:47 Bertrand Drouvot <[email protected]>
  0 siblings, 0 replies; 173+ messages in thread

From: Bertrand Drouvot @ 2026-07-04 04:47 UTC (permalink / raw)

DropRole() and GrantRole() resolve the role name to an OID before acquiring
LockSharedObject() on the role. A concurrent session that commits a DROP ROLE
between the read and the lock acquisition leaves the first session acting on a
stale OID.

This commit fixes the races by using the same approach as RangeVarGetRelidExtended():
It encapsulates name resolution, permission checking (via a caller-supplied callback),
and lock acquisition inside a retry loop driven by SharedInvalidMessageCounter.
If invalidation messages arrive between name resolution and locking, indicating
concurrent DDL, the function retries.

The lock is kept across retries and only released if the name resolves to a
different OID on the next iteration.

Two callbacks are provided:
- RoleNameCallbackForDropRole(): checks current/session user, superuser attribute,
and ADMIN OPTION privilege before locking. This is similar to what DropRole() is
currently doing before LockSharedObject().
- RoleNameCallbackForGrantRole(): calls check_role_membership_authorization() to
verify the current user can grant/revoke membership. This is similar to what GrantRole()
is currently doing before calling AddRoleMems()/DelRoleMems().

DropRole() and GrantRole() now call RoleNameGetOid() with appropriate lock
levels.

AlterRole() does not need the fix because it calls CatalogTupleUpdate() on the
pg_authid tuple before AddRoleMems(), which blocks a concurrent DROP ROLE.

Author: Bertrand Drouvot <[email protected]>
Reviewed-by:
Discussion:
---
 src/backend/commands/user.c | 246 ++++++++++++++++++++++++++----------
 src/include/commands/user.h |   9 ++
 2 files changed, 191 insertions(+), 64 deletions(-)
  95.5% src/backend/commands/
   4.4% src/include/commands/

diff --git a/src/backend/commands/user.c b/src/backend/commands/user.c
index be11c49f919..6d8e2fa8813 100644
--- a/src/backend/commands/user.c
+++ b/src/backend/commands/user.c
@@ -34,6 +34,7 @@
 #include "miscadmin.h"
 #include "port/pg_bitutils.h"
 #include "storage/lmgr.h"
+#include "storage/sinval.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
 #include "utils/catcache.h"
@@ -116,6 +117,10 @@ static void plan_recursive_revoke(CatCList *memlist,
 								  bool revoke_admin_option_only,
 								  DropBehavior behavior);
 static void InitGrantRoleOptions(GrantRoleOptions *popt);
+static void RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+static void RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+										 Oid oldroleid, void *callback_arg);
 
 
 /* Check if current user has createrole privileges */
@@ -126,6 +131,92 @@ have_createrole_privilege(void)
 }
 
 
+/*
+ * RoleNameGetOid
+ *		Given a role name, look up its OID, lock it, and return the OID.
+ *
+ * This follows the same pattern as RangeVarGetRelidExtended():
+ * name resolution, permission check (via callback), and lock acquisition are
+ * performed inside a retry loop. If invalidation messages arrive during the
+ * process (indicating concurrent DDL), we retry to ensure the name still
+ * resolves to the same OID.
+ *
+ * The callback is invoked before locking, giving callers a chance to check
+ * permissions. It receives the current rolename, the resolved OID, the
+ * previous OID (InvalidOid on first iteration), and a caller-supplied arg.
+ * If the callback raises an error, the function aborts without locking.
+ *
+ * If missing_ok is true and the role does not exist, returns InvalidOid.
+ * Otherwise, raises an error.
+ */
+Oid
+RoleNameGetOid(const char *rolename, LOCKMODE lockmode, bool missing_ok,
+			   RoleNameGetOidCallback callback, void *callback_arg)
+{
+	uint64		inval_count;
+	Oid			roleid;
+	Oid			oldroleid = InvalidOid;
+	bool		retry = false;
+
+	for (;;)
+	{
+		/*
+		 * Remember the current invalidation count so we can detect concurrent
+		 * DDL after locking.
+		 */
+		inval_count = SharedInvalidMessageCounter;
+
+		/* Look up the role name */
+		roleid = get_role_oid(rolename, true);
+
+		if (!OidIsValid(roleid))
+		{
+			if (!missing_ok)
+				ereport(ERROR,
+						(errcode(ERRCODE_UNDEFINED_OBJECT),
+						 errmsg("role \"%s\" does not exist", rolename)));
+			return InvalidOid;
+		}
+
+		/*
+		 * Invoke caller-supplied callback before locking. This is a good
+		 * place to check permissions: we haven't taken the lock yet, but we
+		 * know the OID we intend to lock. If concurrent DDL changes things,
+		 * the callback will be invoked again on the next iteration.
+		 */
+		if (callback)
+			callback(rolename, roleid, oldroleid, callback_arg);
+
+		/*
+		 * If upon retry we get back the same OID, the invalidation messages
+		 * did not change the final answer. So we're done.
+		 *
+		 * If we got a different OID, we've locked the role that used to have
+		 * this name rather than the one that does now. Release the old lock.
+		 */
+		if (retry)
+		{
+			if (roleid == oldroleid)
+				break;
+			UnlockSharedObject(AuthIdRelationId, oldroleid, 0, lockmode);
+		}
+
+		/* Lock the role */
+		LockSharedObject(AuthIdRelationId, roleid, 0, lockmode);
+
+		/* If no invalidation messages were processed, we're done */
+		if (inval_count == SharedInvalidMessageCounter)
+			break;
+
+		/* Something may have changed, retry */
+		retry = true;
+		oldroleid = roleid;
+	}
+
+	return roleid;
+}
+
+
 /*
  * CREATE ROLE
  */
@@ -1090,6 +1181,59 @@ AlterRoleSet(AlterRoleSetStmt *stmt)
 }
 
 
+/*
+ * Before acquiring a role lock for DROP ROLE, check that the role is not the
+ * current/session user and that the caller has sufficient privileges to drop it.
+ */
+static void
+RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+							Oid oldroleid, void *callback_arg)
+{
+	HeapTuple	tuple;
+	Form_pg_authid roleform;
+
+	if (roleid == GetUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetOuterUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetSessionUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("session user cannot be dropped")));
+
+	tuple = SearchSysCache1(AUTHOID, ObjectIdGetDatum(roleid));
+	if (!HeapTupleIsValid(tuple))
+		ereport(ERROR,
+				(errcode(ERRCODE_UNDEFINED_OBJECT),
+				 errmsg("role \"%s\" does not exist", rolename)));
+
+	roleform = (Form_pg_authid) GETSTRUCT(tuple);
+
+	/*
+	 * For safety's sake, we allow createrole holders to drop ordinary roles
+	 * but not superuser roles, and only if they also have ADMIN OPTION.
+	 */
+	if (roleform->rolsuper && !superuser())
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
+						   "SUPERUSER", "SUPERUSER")));
+	if (!is_admin_of_role(GetUserId(), roleid))
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
+						   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
+
+	ReleaseSysCache(tuple);
+}
+
+
 /*
  * DROP ROLE
  */
@@ -1119,9 +1263,7 @@ DropRole(DropRoleStmt *stmt)
 	{
 		RoleSpec   *rolspec = lfirst(item);
 		char	   *role;
-		HeapTuple	tuple,
-					tmp_tuple;
-		Form_pg_authid roleform;
+		HeapTuple	tmp_tuple;
 		ScanKeyData scankey;
 		SysScanDesc sscan;
 		Oid			roleid;
@@ -1132,71 +1274,27 @@ DropRole(DropRoleStmt *stmt)
 					 errmsg("cannot use special role specifier in DROP ROLE")));
 		role = rolspec->rolename;
 
-		tuple = SearchSysCache1(AUTHNAME, PointerGetDatum(role));
-		if (!HeapTupleIsValid(tuple))
-		{
-			if (!stmt->missing_ok)
-			{
-				ereport(ERROR,
-						(errcode(ERRCODE_UNDEFINED_OBJECT),
-						 errmsg("role \"%s\" does not exist", role)));
-			}
-			else
-			{
-				ereport(NOTICE,
-						(errmsg("role \"%s\" does not exist, skipping",
-								role)));
-			}
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP or ALTER commits between name
+		 * resolution and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(role, AccessExclusiveLock, stmt->missing_ok,
+								RoleNameCallbackForDropRole, NULL);
 
+		if (!OidIsValid(roleid))
+		{
+			/* missing_ok case: role doesn't exist */
+			ereport(NOTICE,
+					(errmsg("role \"%s\" does not exist, skipping",
+							role)));
 			continue;
 		}
 
-		roleform = (Form_pg_authid) GETSTRUCT(tuple);
-		roleid = roleform->oid;
-
-		if (roleid == GetUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetOuterUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetSessionUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("session user cannot be dropped")));
-
-		/*
-		 * For safety's sake, we allow createrole holders to drop ordinary
-		 * roles but not superuser roles, and only if they also have ADMIN
-		 * OPTION.
-		 */
-		if (roleform->rolsuper && !superuser())
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
-							   "SUPERUSER", "SUPERUSER")));
-		if (!is_admin_of_role(GetUserId(), roleid))
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
-							   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
-
 		/* DROP hook for the role being removed */
 		InvokeObjectDropHook(AuthIdRelationId, roleid, 0);
 
-		/* Don't leak the syscache tuple */
-		ReleaseSysCache(tuple);
-
-		/*
-		 * Lock the role, so nobody can add dependencies to her while we drop
-		 * her.  We keep the lock until the end of transaction.
-		 */
-		LockSharedObject(AuthIdRelationId, roleid, 0, AccessExclusiveLock);
-
 		/*
 		 * If there is a pg_auth_members entry that has one of the roles to be
 		 * dropped as the roleid or member, it should be silently removed, but
@@ -1484,6 +1582,21 @@ RenameRole(const char *oldname, const char *newname)
 	return address;
 }
 
+/*
+ * Before acquiring a role lock for GRANT/REVOKE, check that the current user
+ * has authorization to grant/revoke membership in the specified role.
+ */
+static void
+RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+							 Oid oldroleid, void *callback_arg)
+{
+	bool		is_grant = *((bool *) callback_arg);
+	Oid			currentUserId = GetUserId();
+
+	check_role_membership_authorization(currentUserId, roleid, is_grant);
+}
+
+
 /*
  * GrantRoleStmt
  *
@@ -1568,9 +1681,14 @@ GrantRole(ParseState *pstate, GrantRoleStmt *stmt)
 					(errcode(ERRCODE_INVALID_GRANT_OPERATION),
 					 errmsg("column names cannot be included in GRANT/REVOKE ROLE")));
 
-		roleid = get_role_oid(rolename, false);
-		check_role_membership_authorization(currentUserId,
-											roleid, stmt->is_grant);
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP commits between name resolution
+		 * and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(rolename, ShareUpdateExclusiveLock, false,
+								RoleNameCallbackForGrantRole, &stmt->is_grant);
 		if (stmt->is_grant)
 			AddRoleMems(currentUserId, rolename, roleid,
 						stmt->grantee_roles, grantee_ids,
diff --git a/src/include/commands/user.h b/src/include/commands/user.h
index 97dcb93791b..17263452250 100644
--- a/src/include/commands/user.h
+++ b/src/include/commands/user.h
@@ -21,6 +21,15 @@
 extern PGDLLIMPORT int Password_encryption; /* values from enum PasswordType */
 extern PGDLLIMPORT char *createrole_self_grant;
 
+/* Callback for RoleNameGetOid, invoked after name resolution but before locking */
+typedef void (*RoleNameGetOidCallback) (const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+
+extern Oid	RoleNameGetOid(const char *rolename, LOCKMODE lockmode,
+						   bool missing_ok,
+						   RoleNameGetOidCallback callback,
+						   void *callback_arg);
+
 /* Hook to check passwords in CreateRole() and AlterRole() */
 typedef void (*check_password_hook_type) (const char *username, const char *shadow_pass, PasswordType password_type, Datum validuntil_time, bool validuntil_null);
 
-- 
2.34.1


--xlWYagR37YwJXkF8--





^ permalink  raw  reply  [nested|flat] 173+ messages in thread

* [PATCH v1] Add RoleNameGetOid() with invalidation-based retry loop for DropRole()/GrantRole()
@ 2026-07-04 04:47 Bertrand Drouvot <[email protected]>
  0 siblings, 0 replies; 173+ messages in thread

From: Bertrand Drouvot @ 2026-07-04 04:47 UTC (permalink / raw)

DropRole() and GrantRole() resolve the role name to an OID before acquiring
LockSharedObject() on the role. A concurrent session that commits a DROP ROLE
between the read and the lock acquisition leaves the first session acting on a
stale OID.

This commit fixes the races by using the same approach as RangeVarGetRelidExtended():
It encapsulates name resolution, permission checking (via a caller-supplied callback),
and lock acquisition inside a retry loop driven by SharedInvalidMessageCounter.
If invalidation messages arrive between name resolution and locking, indicating
concurrent DDL, the function retries.

The lock is kept across retries and only released if the name resolves to a
different OID on the next iteration.

Two callbacks are provided:
- RoleNameCallbackForDropRole(): checks current/session user, superuser attribute,
and ADMIN OPTION privilege before locking. This is similar to what DropRole() is
currently doing before LockSharedObject().
- RoleNameCallbackForGrantRole(): calls check_role_membership_authorization() to
verify the current user can grant/revoke membership. This is similar to what GrantRole()
is currently doing before calling AddRoleMems()/DelRoleMems().

DropRole() and GrantRole() now call RoleNameGetOid() with appropriate lock
levels.

AlterRole() does not need the fix because it calls CatalogTupleUpdate() on the
pg_authid tuple before AddRoleMems(), which blocks a concurrent DROP ROLE.

Author: Bertrand Drouvot <[email protected]>
Reviewed-by:
Discussion:
---
 src/backend/commands/user.c | 246 ++++++++++++++++++++++++++----------
 src/include/commands/user.h |   9 ++
 2 files changed, 191 insertions(+), 64 deletions(-)
  95.5% src/backend/commands/
   4.4% src/include/commands/

diff --git a/src/backend/commands/user.c b/src/backend/commands/user.c
index be11c49f919..6d8e2fa8813 100644
--- a/src/backend/commands/user.c
+++ b/src/backend/commands/user.c
@@ -34,6 +34,7 @@
 #include "miscadmin.h"
 #include "port/pg_bitutils.h"
 #include "storage/lmgr.h"
+#include "storage/sinval.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
 #include "utils/catcache.h"
@@ -116,6 +117,10 @@ static void plan_recursive_revoke(CatCList *memlist,
 								  bool revoke_admin_option_only,
 								  DropBehavior behavior);
 static void InitGrantRoleOptions(GrantRoleOptions *popt);
+static void RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+static void RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+										 Oid oldroleid, void *callback_arg);
 
 
 /* Check if current user has createrole privileges */
@@ -126,6 +131,92 @@ have_createrole_privilege(void)
 }
 
 
+/*
+ * RoleNameGetOid
+ *		Given a role name, look up its OID, lock it, and return the OID.
+ *
+ * This follows the same pattern as RangeVarGetRelidExtended():
+ * name resolution, permission check (via callback), and lock acquisition are
+ * performed inside a retry loop. If invalidation messages arrive during the
+ * process (indicating concurrent DDL), we retry to ensure the name still
+ * resolves to the same OID.
+ *
+ * The callback is invoked before locking, giving callers a chance to check
+ * permissions. It receives the current rolename, the resolved OID, the
+ * previous OID (InvalidOid on first iteration), and a caller-supplied arg.
+ * If the callback raises an error, the function aborts without locking.
+ *
+ * If missing_ok is true and the role does not exist, returns InvalidOid.
+ * Otherwise, raises an error.
+ */
+Oid
+RoleNameGetOid(const char *rolename, LOCKMODE lockmode, bool missing_ok,
+			   RoleNameGetOidCallback callback, void *callback_arg)
+{
+	uint64		inval_count;
+	Oid			roleid;
+	Oid			oldroleid = InvalidOid;
+	bool		retry = false;
+
+	for (;;)
+	{
+		/*
+		 * Remember the current invalidation count so we can detect concurrent
+		 * DDL after locking.
+		 */
+		inval_count = SharedInvalidMessageCounter;
+
+		/* Look up the role name */
+		roleid = get_role_oid(rolename, true);
+
+		if (!OidIsValid(roleid))
+		{
+			if (!missing_ok)
+				ereport(ERROR,
+						(errcode(ERRCODE_UNDEFINED_OBJECT),
+						 errmsg("role \"%s\" does not exist", rolename)));
+			return InvalidOid;
+		}
+
+		/*
+		 * Invoke caller-supplied callback before locking. This is a good
+		 * place to check permissions: we haven't taken the lock yet, but we
+		 * know the OID we intend to lock. If concurrent DDL changes things,
+		 * the callback will be invoked again on the next iteration.
+		 */
+		if (callback)
+			callback(rolename, roleid, oldroleid, callback_arg);
+
+		/*
+		 * If upon retry we get back the same OID, the invalidation messages
+		 * did not change the final answer. So we're done.
+		 *
+		 * If we got a different OID, we've locked the role that used to have
+		 * this name rather than the one that does now. Release the old lock.
+		 */
+		if (retry)
+		{
+			if (roleid == oldroleid)
+				break;
+			UnlockSharedObject(AuthIdRelationId, oldroleid, 0, lockmode);
+		}
+
+		/* Lock the role */
+		LockSharedObject(AuthIdRelationId, roleid, 0, lockmode);
+
+		/* If no invalidation messages were processed, we're done */
+		if (inval_count == SharedInvalidMessageCounter)
+			break;
+
+		/* Something may have changed, retry */
+		retry = true;
+		oldroleid = roleid;
+	}
+
+	return roleid;
+}
+
+
 /*
  * CREATE ROLE
  */
@@ -1090,6 +1181,59 @@ AlterRoleSet(AlterRoleSetStmt *stmt)
 }
 
 
+/*
+ * Before acquiring a role lock for DROP ROLE, check that the role is not the
+ * current/session user and that the caller has sufficient privileges to drop it.
+ */
+static void
+RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+							Oid oldroleid, void *callback_arg)
+{
+	HeapTuple	tuple;
+	Form_pg_authid roleform;
+
+	if (roleid == GetUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetOuterUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetSessionUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("session user cannot be dropped")));
+
+	tuple = SearchSysCache1(AUTHOID, ObjectIdGetDatum(roleid));
+	if (!HeapTupleIsValid(tuple))
+		ereport(ERROR,
+				(errcode(ERRCODE_UNDEFINED_OBJECT),
+				 errmsg("role \"%s\" does not exist", rolename)));
+
+	roleform = (Form_pg_authid) GETSTRUCT(tuple);
+
+	/*
+	 * For safety's sake, we allow createrole holders to drop ordinary roles
+	 * but not superuser roles, and only if they also have ADMIN OPTION.
+	 */
+	if (roleform->rolsuper && !superuser())
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
+						   "SUPERUSER", "SUPERUSER")));
+	if (!is_admin_of_role(GetUserId(), roleid))
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
+						   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
+
+	ReleaseSysCache(tuple);
+}
+
+
 /*
  * DROP ROLE
  */
@@ -1119,9 +1263,7 @@ DropRole(DropRoleStmt *stmt)
 	{
 		RoleSpec   *rolspec = lfirst(item);
 		char	   *role;
-		HeapTuple	tuple,
-					tmp_tuple;
-		Form_pg_authid roleform;
+		HeapTuple	tmp_tuple;
 		ScanKeyData scankey;
 		SysScanDesc sscan;
 		Oid			roleid;
@@ -1132,71 +1274,27 @@ DropRole(DropRoleStmt *stmt)
 					 errmsg("cannot use special role specifier in DROP ROLE")));
 		role = rolspec->rolename;
 
-		tuple = SearchSysCache1(AUTHNAME, PointerGetDatum(role));
-		if (!HeapTupleIsValid(tuple))
-		{
-			if (!stmt->missing_ok)
-			{
-				ereport(ERROR,
-						(errcode(ERRCODE_UNDEFINED_OBJECT),
-						 errmsg("role \"%s\" does not exist", role)));
-			}
-			else
-			{
-				ereport(NOTICE,
-						(errmsg("role \"%s\" does not exist, skipping",
-								role)));
-			}
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP or ALTER commits between name
+		 * resolution and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(role, AccessExclusiveLock, stmt->missing_ok,
+								RoleNameCallbackForDropRole, NULL);
 
+		if (!OidIsValid(roleid))
+		{
+			/* missing_ok case: role doesn't exist */
+			ereport(NOTICE,
+					(errmsg("role \"%s\" does not exist, skipping",
+							role)));
 			continue;
 		}
 
-		roleform = (Form_pg_authid) GETSTRUCT(tuple);
-		roleid = roleform->oid;
-
-		if (roleid == GetUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetOuterUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetSessionUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("session user cannot be dropped")));
-
-		/*
-		 * For safety's sake, we allow createrole holders to drop ordinary
-		 * roles but not superuser roles, and only if they also have ADMIN
-		 * OPTION.
-		 */
-		if (roleform->rolsuper && !superuser())
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
-							   "SUPERUSER", "SUPERUSER")));
-		if (!is_admin_of_role(GetUserId(), roleid))
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
-							   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
-
 		/* DROP hook for the role being removed */
 		InvokeObjectDropHook(AuthIdRelationId, roleid, 0);
 
-		/* Don't leak the syscache tuple */
-		ReleaseSysCache(tuple);
-
-		/*
-		 * Lock the role, so nobody can add dependencies to her while we drop
-		 * her.  We keep the lock until the end of transaction.
-		 */
-		LockSharedObject(AuthIdRelationId, roleid, 0, AccessExclusiveLock);
-
 		/*
 		 * If there is a pg_auth_members entry that has one of the roles to be
 		 * dropped as the roleid or member, it should be silently removed, but
@@ -1484,6 +1582,21 @@ RenameRole(const char *oldname, const char *newname)
 	return address;
 }
 
+/*
+ * Before acquiring a role lock for GRANT/REVOKE, check that the current user
+ * has authorization to grant/revoke membership in the specified role.
+ */
+static void
+RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+							 Oid oldroleid, void *callback_arg)
+{
+	bool		is_grant = *((bool *) callback_arg);
+	Oid			currentUserId = GetUserId();
+
+	check_role_membership_authorization(currentUserId, roleid, is_grant);
+}
+
+
 /*
  * GrantRoleStmt
  *
@@ -1568,9 +1681,14 @@ GrantRole(ParseState *pstate, GrantRoleStmt *stmt)
 					(errcode(ERRCODE_INVALID_GRANT_OPERATION),
 					 errmsg("column names cannot be included in GRANT/REVOKE ROLE")));
 
-		roleid = get_role_oid(rolename, false);
-		check_role_membership_authorization(currentUserId,
-											roleid, stmt->is_grant);
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP commits between name resolution
+		 * and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(rolename, ShareUpdateExclusiveLock, false,
+								RoleNameCallbackForGrantRole, &stmt->is_grant);
 		if (stmt->is_grant)
 			AddRoleMems(currentUserId, rolename, roleid,
 						stmt->grantee_roles, grantee_ids,
diff --git a/src/include/commands/user.h b/src/include/commands/user.h
index 97dcb93791b..17263452250 100644
--- a/src/include/commands/user.h
+++ b/src/include/commands/user.h
@@ -21,6 +21,15 @@
 extern PGDLLIMPORT int Password_encryption; /* values from enum PasswordType */
 extern PGDLLIMPORT char *createrole_self_grant;
 
+/* Callback for RoleNameGetOid, invoked after name resolution but before locking */
+typedef void (*RoleNameGetOidCallback) (const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+
+extern Oid	RoleNameGetOid(const char *rolename, LOCKMODE lockmode,
+						   bool missing_ok,
+						   RoleNameGetOidCallback callback,
+						   void *callback_arg);
+
 /* Hook to check passwords in CreateRole() and AlterRole() */
 typedef void (*check_password_hook_type) (const char *username, const char *shadow_pass, PasswordType password_type, Datum validuntil_time, bool validuntil_null);
 
-- 
2.34.1


--xlWYagR37YwJXkF8--





^ permalink  raw  reply  [nested|flat] 173+ messages in thread

* [PATCH v1] Add RoleNameGetOid() with invalidation-based retry loop for DropRole()/GrantRole()
@ 2026-07-04 04:47 Bertrand Drouvot <[email protected]>
  0 siblings, 0 replies; 173+ messages in thread

From: Bertrand Drouvot @ 2026-07-04 04:47 UTC (permalink / raw)

DropRole() and GrantRole() resolve the role name to an OID before acquiring
LockSharedObject() on the role. A concurrent session that commits a DROP ROLE
between the read and the lock acquisition leaves the first session acting on a
stale OID.

This commit fixes the races by using the same approach as RangeVarGetRelidExtended():
It encapsulates name resolution, permission checking (via a caller-supplied callback),
and lock acquisition inside a retry loop driven by SharedInvalidMessageCounter.
If invalidation messages arrive between name resolution and locking, indicating
concurrent DDL, the function retries.

The lock is kept across retries and only released if the name resolves to a
different OID on the next iteration.

Two callbacks are provided:
- RoleNameCallbackForDropRole(): checks current/session user, superuser attribute,
and ADMIN OPTION privilege before locking. This is similar to what DropRole() is
currently doing before LockSharedObject().
- RoleNameCallbackForGrantRole(): calls check_role_membership_authorization() to
verify the current user can grant/revoke membership. This is similar to what GrantRole()
is currently doing before calling AddRoleMems()/DelRoleMems().

DropRole() and GrantRole() now call RoleNameGetOid() with appropriate lock
levels.

AlterRole() does not need the fix because it calls CatalogTupleUpdate() on the
pg_authid tuple before AddRoleMems(), which blocks a concurrent DROP ROLE.

Author: Bertrand Drouvot <[email protected]>
Reviewed-by:
Discussion:
---
 src/backend/commands/user.c | 246 ++++++++++++++++++++++++++----------
 src/include/commands/user.h |   9 ++
 2 files changed, 191 insertions(+), 64 deletions(-)
  95.5% src/backend/commands/
   4.4% src/include/commands/

diff --git a/src/backend/commands/user.c b/src/backend/commands/user.c
index be11c49f919..6d8e2fa8813 100644
--- a/src/backend/commands/user.c
+++ b/src/backend/commands/user.c
@@ -34,6 +34,7 @@
 #include "miscadmin.h"
 #include "port/pg_bitutils.h"
 #include "storage/lmgr.h"
+#include "storage/sinval.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
 #include "utils/catcache.h"
@@ -116,6 +117,10 @@ static void plan_recursive_revoke(CatCList *memlist,
 								  bool revoke_admin_option_only,
 								  DropBehavior behavior);
 static void InitGrantRoleOptions(GrantRoleOptions *popt);
+static void RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+static void RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+										 Oid oldroleid, void *callback_arg);
 
 
 /* Check if current user has createrole privileges */
@@ -126,6 +131,92 @@ have_createrole_privilege(void)
 }
 
 
+/*
+ * RoleNameGetOid
+ *		Given a role name, look up its OID, lock it, and return the OID.
+ *
+ * This follows the same pattern as RangeVarGetRelidExtended():
+ * name resolution, permission check (via callback), and lock acquisition are
+ * performed inside a retry loop. If invalidation messages arrive during the
+ * process (indicating concurrent DDL), we retry to ensure the name still
+ * resolves to the same OID.
+ *
+ * The callback is invoked before locking, giving callers a chance to check
+ * permissions. It receives the current rolename, the resolved OID, the
+ * previous OID (InvalidOid on first iteration), and a caller-supplied arg.
+ * If the callback raises an error, the function aborts without locking.
+ *
+ * If missing_ok is true and the role does not exist, returns InvalidOid.
+ * Otherwise, raises an error.
+ */
+Oid
+RoleNameGetOid(const char *rolename, LOCKMODE lockmode, bool missing_ok,
+			   RoleNameGetOidCallback callback, void *callback_arg)
+{
+	uint64		inval_count;
+	Oid			roleid;
+	Oid			oldroleid = InvalidOid;
+	bool		retry = false;
+
+	for (;;)
+	{
+		/*
+		 * Remember the current invalidation count so we can detect concurrent
+		 * DDL after locking.
+		 */
+		inval_count = SharedInvalidMessageCounter;
+
+		/* Look up the role name */
+		roleid = get_role_oid(rolename, true);
+
+		if (!OidIsValid(roleid))
+		{
+			if (!missing_ok)
+				ereport(ERROR,
+						(errcode(ERRCODE_UNDEFINED_OBJECT),
+						 errmsg("role \"%s\" does not exist", rolename)));
+			return InvalidOid;
+		}
+
+		/*
+		 * Invoke caller-supplied callback before locking. This is a good
+		 * place to check permissions: we haven't taken the lock yet, but we
+		 * know the OID we intend to lock. If concurrent DDL changes things,
+		 * the callback will be invoked again on the next iteration.
+		 */
+		if (callback)
+			callback(rolename, roleid, oldroleid, callback_arg);
+
+		/*
+		 * If upon retry we get back the same OID, the invalidation messages
+		 * did not change the final answer. So we're done.
+		 *
+		 * If we got a different OID, we've locked the role that used to have
+		 * this name rather than the one that does now. Release the old lock.
+		 */
+		if (retry)
+		{
+			if (roleid == oldroleid)
+				break;
+			UnlockSharedObject(AuthIdRelationId, oldroleid, 0, lockmode);
+		}
+
+		/* Lock the role */
+		LockSharedObject(AuthIdRelationId, roleid, 0, lockmode);
+
+		/* If no invalidation messages were processed, we're done */
+		if (inval_count == SharedInvalidMessageCounter)
+			break;
+
+		/* Something may have changed, retry */
+		retry = true;
+		oldroleid = roleid;
+	}
+
+	return roleid;
+}
+
+
 /*
  * CREATE ROLE
  */
@@ -1090,6 +1181,59 @@ AlterRoleSet(AlterRoleSetStmt *stmt)
 }
 
 
+/*
+ * Before acquiring a role lock for DROP ROLE, check that the role is not the
+ * current/session user and that the caller has sufficient privileges to drop it.
+ */
+static void
+RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+							Oid oldroleid, void *callback_arg)
+{
+	HeapTuple	tuple;
+	Form_pg_authid roleform;
+
+	if (roleid == GetUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetOuterUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetSessionUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("session user cannot be dropped")));
+
+	tuple = SearchSysCache1(AUTHOID, ObjectIdGetDatum(roleid));
+	if (!HeapTupleIsValid(tuple))
+		ereport(ERROR,
+				(errcode(ERRCODE_UNDEFINED_OBJECT),
+				 errmsg("role \"%s\" does not exist", rolename)));
+
+	roleform = (Form_pg_authid) GETSTRUCT(tuple);
+
+	/*
+	 * For safety's sake, we allow createrole holders to drop ordinary roles
+	 * but not superuser roles, and only if they also have ADMIN OPTION.
+	 */
+	if (roleform->rolsuper && !superuser())
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
+						   "SUPERUSER", "SUPERUSER")));
+	if (!is_admin_of_role(GetUserId(), roleid))
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
+						   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
+
+	ReleaseSysCache(tuple);
+}
+
+
 /*
  * DROP ROLE
  */
@@ -1119,9 +1263,7 @@ DropRole(DropRoleStmt *stmt)
 	{
 		RoleSpec   *rolspec = lfirst(item);
 		char	   *role;
-		HeapTuple	tuple,
-					tmp_tuple;
-		Form_pg_authid roleform;
+		HeapTuple	tmp_tuple;
 		ScanKeyData scankey;
 		SysScanDesc sscan;
 		Oid			roleid;
@@ -1132,71 +1274,27 @@ DropRole(DropRoleStmt *stmt)
 					 errmsg("cannot use special role specifier in DROP ROLE")));
 		role = rolspec->rolename;
 
-		tuple = SearchSysCache1(AUTHNAME, PointerGetDatum(role));
-		if (!HeapTupleIsValid(tuple))
-		{
-			if (!stmt->missing_ok)
-			{
-				ereport(ERROR,
-						(errcode(ERRCODE_UNDEFINED_OBJECT),
-						 errmsg("role \"%s\" does not exist", role)));
-			}
-			else
-			{
-				ereport(NOTICE,
-						(errmsg("role \"%s\" does not exist, skipping",
-								role)));
-			}
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP or ALTER commits between name
+		 * resolution and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(role, AccessExclusiveLock, stmt->missing_ok,
+								RoleNameCallbackForDropRole, NULL);
 
+		if (!OidIsValid(roleid))
+		{
+			/* missing_ok case: role doesn't exist */
+			ereport(NOTICE,
+					(errmsg("role \"%s\" does not exist, skipping",
+							role)));
 			continue;
 		}
 
-		roleform = (Form_pg_authid) GETSTRUCT(tuple);
-		roleid = roleform->oid;
-
-		if (roleid == GetUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetOuterUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetSessionUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("session user cannot be dropped")));
-
-		/*
-		 * For safety's sake, we allow createrole holders to drop ordinary
-		 * roles but not superuser roles, and only if they also have ADMIN
-		 * OPTION.
-		 */
-		if (roleform->rolsuper && !superuser())
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
-							   "SUPERUSER", "SUPERUSER")));
-		if (!is_admin_of_role(GetUserId(), roleid))
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
-							   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
-
 		/* DROP hook for the role being removed */
 		InvokeObjectDropHook(AuthIdRelationId, roleid, 0);
 
-		/* Don't leak the syscache tuple */
-		ReleaseSysCache(tuple);
-
-		/*
-		 * Lock the role, so nobody can add dependencies to her while we drop
-		 * her.  We keep the lock until the end of transaction.
-		 */
-		LockSharedObject(AuthIdRelationId, roleid, 0, AccessExclusiveLock);
-
 		/*
 		 * If there is a pg_auth_members entry that has one of the roles to be
 		 * dropped as the roleid or member, it should be silently removed, but
@@ -1484,6 +1582,21 @@ RenameRole(const char *oldname, const char *newname)
 	return address;
 }
 
+/*
+ * Before acquiring a role lock for GRANT/REVOKE, check that the current user
+ * has authorization to grant/revoke membership in the specified role.
+ */
+static void
+RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+							 Oid oldroleid, void *callback_arg)
+{
+	bool		is_grant = *((bool *) callback_arg);
+	Oid			currentUserId = GetUserId();
+
+	check_role_membership_authorization(currentUserId, roleid, is_grant);
+}
+
+
 /*
  * GrantRoleStmt
  *
@@ -1568,9 +1681,14 @@ GrantRole(ParseState *pstate, GrantRoleStmt *stmt)
 					(errcode(ERRCODE_INVALID_GRANT_OPERATION),
 					 errmsg("column names cannot be included in GRANT/REVOKE ROLE")));
 
-		roleid = get_role_oid(rolename, false);
-		check_role_membership_authorization(currentUserId,
-											roleid, stmt->is_grant);
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP commits between name resolution
+		 * and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(rolename, ShareUpdateExclusiveLock, false,
+								RoleNameCallbackForGrantRole, &stmt->is_grant);
 		if (stmt->is_grant)
 			AddRoleMems(currentUserId, rolename, roleid,
 						stmt->grantee_roles, grantee_ids,
diff --git a/src/include/commands/user.h b/src/include/commands/user.h
index 97dcb93791b..17263452250 100644
--- a/src/include/commands/user.h
+++ b/src/include/commands/user.h
@@ -21,6 +21,15 @@
 extern PGDLLIMPORT int Password_encryption; /* values from enum PasswordType */
 extern PGDLLIMPORT char *createrole_self_grant;
 
+/* Callback for RoleNameGetOid, invoked after name resolution but before locking */
+typedef void (*RoleNameGetOidCallback) (const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+
+extern Oid	RoleNameGetOid(const char *rolename, LOCKMODE lockmode,
+						   bool missing_ok,
+						   RoleNameGetOidCallback callback,
+						   void *callback_arg);
+
 /* Hook to check passwords in CreateRole() and AlterRole() */
 typedef void (*check_password_hook_type) (const char *username, const char *shadow_pass, PasswordType password_type, Datum validuntil_time, bool validuntil_null);
 
-- 
2.34.1


--xlWYagR37YwJXkF8--





^ permalink  raw  reply  [nested|flat] 173+ messages in thread

* [PATCH v1] Add RoleNameGetOid() with invalidation-based retry loop for DropRole()/GrantRole()
@ 2026-07-04 04:47 Bertrand Drouvot <[email protected]>
  0 siblings, 0 replies; 173+ messages in thread

From: Bertrand Drouvot @ 2026-07-04 04:47 UTC (permalink / raw)

DropRole() and GrantRole() resolve the role name to an OID before acquiring
LockSharedObject() on the role. A concurrent session that commits a DROP ROLE
between the read and the lock acquisition leaves the first session acting on a
stale OID.

This commit fixes the races by using the same approach as RangeVarGetRelidExtended():
It encapsulates name resolution, permission checking (via a caller-supplied callback),
and lock acquisition inside a retry loop driven by SharedInvalidMessageCounter.
If invalidation messages arrive between name resolution and locking, indicating
concurrent DDL, the function retries.

The lock is kept across retries and only released if the name resolves to a
different OID on the next iteration.

Two callbacks are provided:
- RoleNameCallbackForDropRole(): checks current/session user, superuser attribute,
and ADMIN OPTION privilege before locking. This is similar to what DropRole() is
currently doing before LockSharedObject().
- RoleNameCallbackForGrantRole(): calls check_role_membership_authorization() to
verify the current user can grant/revoke membership. This is similar to what GrantRole()
is currently doing before calling AddRoleMems()/DelRoleMems().

DropRole() and GrantRole() now call RoleNameGetOid() with appropriate lock
levels.

AlterRole() does not need the fix because it calls CatalogTupleUpdate() on the
pg_authid tuple before AddRoleMems(), which blocks a concurrent DROP ROLE.

Author: Bertrand Drouvot <[email protected]>
Reviewed-by:
Discussion:
---
 src/backend/commands/user.c | 246 ++++++++++++++++++++++++++----------
 src/include/commands/user.h |   9 ++
 2 files changed, 191 insertions(+), 64 deletions(-)
  95.5% src/backend/commands/
   4.4% src/include/commands/

diff --git a/src/backend/commands/user.c b/src/backend/commands/user.c
index be11c49f919..6d8e2fa8813 100644
--- a/src/backend/commands/user.c
+++ b/src/backend/commands/user.c
@@ -34,6 +34,7 @@
 #include "miscadmin.h"
 #include "port/pg_bitutils.h"
 #include "storage/lmgr.h"
+#include "storage/sinval.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
 #include "utils/catcache.h"
@@ -116,6 +117,10 @@ static void plan_recursive_revoke(CatCList *memlist,
 								  bool revoke_admin_option_only,
 								  DropBehavior behavior);
 static void InitGrantRoleOptions(GrantRoleOptions *popt);
+static void RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+static void RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+										 Oid oldroleid, void *callback_arg);
 
 
 /* Check if current user has createrole privileges */
@@ -126,6 +131,92 @@ have_createrole_privilege(void)
 }
 
 
+/*
+ * RoleNameGetOid
+ *		Given a role name, look up its OID, lock it, and return the OID.
+ *
+ * This follows the same pattern as RangeVarGetRelidExtended():
+ * name resolution, permission check (via callback), and lock acquisition are
+ * performed inside a retry loop. If invalidation messages arrive during the
+ * process (indicating concurrent DDL), we retry to ensure the name still
+ * resolves to the same OID.
+ *
+ * The callback is invoked before locking, giving callers a chance to check
+ * permissions. It receives the current rolename, the resolved OID, the
+ * previous OID (InvalidOid on first iteration), and a caller-supplied arg.
+ * If the callback raises an error, the function aborts without locking.
+ *
+ * If missing_ok is true and the role does not exist, returns InvalidOid.
+ * Otherwise, raises an error.
+ */
+Oid
+RoleNameGetOid(const char *rolename, LOCKMODE lockmode, bool missing_ok,
+			   RoleNameGetOidCallback callback, void *callback_arg)
+{
+	uint64		inval_count;
+	Oid			roleid;
+	Oid			oldroleid = InvalidOid;
+	bool		retry = false;
+
+	for (;;)
+	{
+		/*
+		 * Remember the current invalidation count so we can detect concurrent
+		 * DDL after locking.
+		 */
+		inval_count = SharedInvalidMessageCounter;
+
+		/* Look up the role name */
+		roleid = get_role_oid(rolename, true);
+
+		if (!OidIsValid(roleid))
+		{
+			if (!missing_ok)
+				ereport(ERROR,
+						(errcode(ERRCODE_UNDEFINED_OBJECT),
+						 errmsg("role \"%s\" does not exist", rolename)));
+			return InvalidOid;
+		}
+
+		/*
+		 * Invoke caller-supplied callback before locking. This is a good
+		 * place to check permissions: we haven't taken the lock yet, but we
+		 * know the OID we intend to lock. If concurrent DDL changes things,
+		 * the callback will be invoked again on the next iteration.
+		 */
+		if (callback)
+			callback(rolename, roleid, oldroleid, callback_arg);
+
+		/*
+		 * If upon retry we get back the same OID, the invalidation messages
+		 * did not change the final answer. So we're done.
+		 *
+		 * If we got a different OID, we've locked the role that used to have
+		 * this name rather than the one that does now. Release the old lock.
+		 */
+		if (retry)
+		{
+			if (roleid == oldroleid)
+				break;
+			UnlockSharedObject(AuthIdRelationId, oldroleid, 0, lockmode);
+		}
+
+		/* Lock the role */
+		LockSharedObject(AuthIdRelationId, roleid, 0, lockmode);
+
+		/* If no invalidation messages were processed, we're done */
+		if (inval_count == SharedInvalidMessageCounter)
+			break;
+
+		/* Something may have changed, retry */
+		retry = true;
+		oldroleid = roleid;
+	}
+
+	return roleid;
+}
+
+
 /*
  * CREATE ROLE
  */
@@ -1090,6 +1181,59 @@ AlterRoleSet(AlterRoleSetStmt *stmt)
 }
 
 
+/*
+ * Before acquiring a role lock for DROP ROLE, check that the role is not the
+ * current/session user and that the caller has sufficient privileges to drop it.
+ */
+static void
+RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+							Oid oldroleid, void *callback_arg)
+{
+	HeapTuple	tuple;
+	Form_pg_authid roleform;
+
+	if (roleid == GetUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetOuterUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetSessionUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("session user cannot be dropped")));
+
+	tuple = SearchSysCache1(AUTHOID, ObjectIdGetDatum(roleid));
+	if (!HeapTupleIsValid(tuple))
+		ereport(ERROR,
+				(errcode(ERRCODE_UNDEFINED_OBJECT),
+				 errmsg("role \"%s\" does not exist", rolename)));
+
+	roleform = (Form_pg_authid) GETSTRUCT(tuple);
+
+	/*
+	 * For safety's sake, we allow createrole holders to drop ordinary roles
+	 * but not superuser roles, and only if they also have ADMIN OPTION.
+	 */
+	if (roleform->rolsuper && !superuser())
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
+						   "SUPERUSER", "SUPERUSER")));
+	if (!is_admin_of_role(GetUserId(), roleid))
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
+						   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
+
+	ReleaseSysCache(tuple);
+}
+
+
 /*
  * DROP ROLE
  */
@@ -1119,9 +1263,7 @@ DropRole(DropRoleStmt *stmt)
 	{
 		RoleSpec   *rolspec = lfirst(item);
 		char	   *role;
-		HeapTuple	tuple,
-					tmp_tuple;
-		Form_pg_authid roleform;
+		HeapTuple	tmp_tuple;
 		ScanKeyData scankey;
 		SysScanDesc sscan;
 		Oid			roleid;
@@ -1132,71 +1274,27 @@ DropRole(DropRoleStmt *stmt)
 					 errmsg("cannot use special role specifier in DROP ROLE")));
 		role = rolspec->rolename;
 
-		tuple = SearchSysCache1(AUTHNAME, PointerGetDatum(role));
-		if (!HeapTupleIsValid(tuple))
-		{
-			if (!stmt->missing_ok)
-			{
-				ereport(ERROR,
-						(errcode(ERRCODE_UNDEFINED_OBJECT),
-						 errmsg("role \"%s\" does not exist", role)));
-			}
-			else
-			{
-				ereport(NOTICE,
-						(errmsg("role \"%s\" does not exist, skipping",
-								role)));
-			}
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP or ALTER commits between name
+		 * resolution and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(role, AccessExclusiveLock, stmt->missing_ok,
+								RoleNameCallbackForDropRole, NULL);
 
+		if (!OidIsValid(roleid))
+		{
+			/* missing_ok case: role doesn't exist */
+			ereport(NOTICE,
+					(errmsg("role \"%s\" does not exist, skipping",
+							role)));
 			continue;
 		}
 
-		roleform = (Form_pg_authid) GETSTRUCT(tuple);
-		roleid = roleform->oid;
-
-		if (roleid == GetUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetOuterUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetSessionUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("session user cannot be dropped")));
-
-		/*
-		 * For safety's sake, we allow createrole holders to drop ordinary
-		 * roles but not superuser roles, and only if they also have ADMIN
-		 * OPTION.
-		 */
-		if (roleform->rolsuper && !superuser())
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
-							   "SUPERUSER", "SUPERUSER")));
-		if (!is_admin_of_role(GetUserId(), roleid))
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
-							   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
-
 		/* DROP hook for the role being removed */
 		InvokeObjectDropHook(AuthIdRelationId, roleid, 0);
 
-		/* Don't leak the syscache tuple */
-		ReleaseSysCache(tuple);
-
-		/*
-		 * Lock the role, so nobody can add dependencies to her while we drop
-		 * her.  We keep the lock until the end of transaction.
-		 */
-		LockSharedObject(AuthIdRelationId, roleid, 0, AccessExclusiveLock);
-
 		/*
 		 * If there is a pg_auth_members entry that has one of the roles to be
 		 * dropped as the roleid or member, it should be silently removed, but
@@ -1484,6 +1582,21 @@ RenameRole(const char *oldname, const char *newname)
 	return address;
 }
 
+/*
+ * Before acquiring a role lock for GRANT/REVOKE, check that the current user
+ * has authorization to grant/revoke membership in the specified role.
+ */
+static void
+RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+							 Oid oldroleid, void *callback_arg)
+{
+	bool		is_grant = *((bool *) callback_arg);
+	Oid			currentUserId = GetUserId();
+
+	check_role_membership_authorization(currentUserId, roleid, is_grant);
+}
+
+
 /*
  * GrantRoleStmt
  *
@@ -1568,9 +1681,14 @@ GrantRole(ParseState *pstate, GrantRoleStmt *stmt)
 					(errcode(ERRCODE_INVALID_GRANT_OPERATION),
 					 errmsg("column names cannot be included in GRANT/REVOKE ROLE")));
 
-		roleid = get_role_oid(rolename, false);
-		check_role_membership_authorization(currentUserId,
-											roleid, stmt->is_grant);
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP commits between name resolution
+		 * and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(rolename, ShareUpdateExclusiveLock, false,
+								RoleNameCallbackForGrantRole, &stmt->is_grant);
 		if (stmt->is_grant)
 			AddRoleMems(currentUserId, rolename, roleid,
 						stmt->grantee_roles, grantee_ids,
diff --git a/src/include/commands/user.h b/src/include/commands/user.h
index 97dcb93791b..17263452250 100644
--- a/src/include/commands/user.h
+++ b/src/include/commands/user.h
@@ -21,6 +21,15 @@
 extern PGDLLIMPORT int Password_encryption; /* values from enum PasswordType */
 extern PGDLLIMPORT char *createrole_self_grant;
 
+/* Callback for RoleNameGetOid, invoked after name resolution but before locking */
+typedef void (*RoleNameGetOidCallback) (const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+
+extern Oid	RoleNameGetOid(const char *rolename, LOCKMODE lockmode,
+						   bool missing_ok,
+						   RoleNameGetOidCallback callback,
+						   void *callback_arg);
+
 /* Hook to check passwords in CreateRole() and AlterRole() */
 typedef void (*check_password_hook_type) (const char *username, const char *shadow_pass, PasswordType password_type, Datum validuntil_time, bool validuntil_null);
 
-- 
2.34.1


--xlWYagR37YwJXkF8--





^ permalink  raw  reply  [nested|flat] 173+ messages in thread

* [PATCH v1] Add RoleNameGetOid() with invalidation-based retry loop for DropRole()/GrantRole()
@ 2026-07-04 04:47 Bertrand Drouvot <[email protected]>
  0 siblings, 0 replies; 173+ messages in thread

From: Bertrand Drouvot @ 2026-07-04 04:47 UTC (permalink / raw)

DropRole() and GrantRole() resolve the role name to an OID before acquiring
LockSharedObject() on the role. A concurrent session that commits a DROP ROLE
between the read and the lock acquisition leaves the first session acting on a
stale OID.

This commit fixes the races by using the same approach as RangeVarGetRelidExtended():
It encapsulates name resolution, permission checking (via a caller-supplied callback),
and lock acquisition inside a retry loop driven by SharedInvalidMessageCounter.
If invalidation messages arrive between name resolution and locking, indicating
concurrent DDL, the function retries.

The lock is kept across retries and only released if the name resolves to a
different OID on the next iteration.

Two callbacks are provided:
- RoleNameCallbackForDropRole(): checks current/session user, superuser attribute,
and ADMIN OPTION privilege before locking. This is similar to what DropRole() is
currently doing before LockSharedObject().
- RoleNameCallbackForGrantRole(): calls check_role_membership_authorization() to
verify the current user can grant/revoke membership. This is similar to what GrantRole()
is currently doing before calling AddRoleMems()/DelRoleMems().

DropRole() and GrantRole() now call RoleNameGetOid() with appropriate lock
levels.

AlterRole() does not need the fix because it calls CatalogTupleUpdate() on the
pg_authid tuple before AddRoleMems(), which blocks a concurrent DROP ROLE.

Author: Bertrand Drouvot <[email protected]>
Reviewed-by:
Discussion:
---
 src/backend/commands/user.c | 246 ++++++++++++++++++++++++++----------
 src/include/commands/user.h |   9 ++
 2 files changed, 191 insertions(+), 64 deletions(-)
  95.5% src/backend/commands/
   4.4% src/include/commands/

diff --git a/src/backend/commands/user.c b/src/backend/commands/user.c
index be11c49f919..6d8e2fa8813 100644
--- a/src/backend/commands/user.c
+++ b/src/backend/commands/user.c
@@ -34,6 +34,7 @@
 #include "miscadmin.h"
 #include "port/pg_bitutils.h"
 #include "storage/lmgr.h"
+#include "storage/sinval.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
 #include "utils/catcache.h"
@@ -116,6 +117,10 @@ static void plan_recursive_revoke(CatCList *memlist,
 								  bool revoke_admin_option_only,
 								  DropBehavior behavior);
 static void InitGrantRoleOptions(GrantRoleOptions *popt);
+static void RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+static void RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+										 Oid oldroleid, void *callback_arg);
 
 
 /* Check if current user has createrole privileges */
@@ -126,6 +131,92 @@ have_createrole_privilege(void)
 }
 
 
+/*
+ * RoleNameGetOid
+ *		Given a role name, look up its OID, lock it, and return the OID.
+ *
+ * This follows the same pattern as RangeVarGetRelidExtended():
+ * name resolution, permission check (via callback), and lock acquisition are
+ * performed inside a retry loop. If invalidation messages arrive during the
+ * process (indicating concurrent DDL), we retry to ensure the name still
+ * resolves to the same OID.
+ *
+ * The callback is invoked before locking, giving callers a chance to check
+ * permissions. It receives the current rolename, the resolved OID, the
+ * previous OID (InvalidOid on first iteration), and a caller-supplied arg.
+ * If the callback raises an error, the function aborts without locking.
+ *
+ * If missing_ok is true and the role does not exist, returns InvalidOid.
+ * Otherwise, raises an error.
+ */
+Oid
+RoleNameGetOid(const char *rolename, LOCKMODE lockmode, bool missing_ok,
+			   RoleNameGetOidCallback callback, void *callback_arg)
+{
+	uint64		inval_count;
+	Oid			roleid;
+	Oid			oldroleid = InvalidOid;
+	bool		retry = false;
+
+	for (;;)
+	{
+		/*
+		 * Remember the current invalidation count so we can detect concurrent
+		 * DDL after locking.
+		 */
+		inval_count = SharedInvalidMessageCounter;
+
+		/* Look up the role name */
+		roleid = get_role_oid(rolename, true);
+
+		if (!OidIsValid(roleid))
+		{
+			if (!missing_ok)
+				ereport(ERROR,
+						(errcode(ERRCODE_UNDEFINED_OBJECT),
+						 errmsg("role \"%s\" does not exist", rolename)));
+			return InvalidOid;
+		}
+
+		/*
+		 * Invoke caller-supplied callback before locking. This is a good
+		 * place to check permissions: we haven't taken the lock yet, but we
+		 * know the OID we intend to lock. If concurrent DDL changes things,
+		 * the callback will be invoked again on the next iteration.
+		 */
+		if (callback)
+			callback(rolename, roleid, oldroleid, callback_arg);
+
+		/*
+		 * If upon retry we get back the same OID, the invalidation messages
+		 * did not change the final answer. So we're done.
+		 *
+		 * If we got a different OID, we've locked the role that used to have
+		 * this name rather than the one that does now. Release the old lock.
+		 */
+		if (retry)
+		{
+			if (roleid == oldroleid)
+				break;
+			UnlockSharedObject(AuthIdRelationId, oldroleid, 0, lockmode);
+		}
+
+		/* Lock the role */
+		LockSharedObject(AuthIdRelationId, roleid, 0, lockmode);
+
+		/* If no invalidation messages were processed, we're done */
+		if (inval_count == SharedInvalidMessageCounter)
+			break;
+
+		/* Something may have changed, retry */
+		retry = true;
+		oldroleid = roleid;
+	}
+
+	return roleid;
+}
+
+
 /*
  * CREATE ROLE
  */
@@ -1090,6 +1181,59 @@ AlterRoleSet(AlterRoleSetStmt *stmt)
 }
 
 
+/*
+ * Before acquiring a role lock for DROP ROLE, check that the role is not the
+ * current/session user and that the caller has sufficient privileges to drop it.
+ */
+static void
+RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+							Oid oldroleid, void *callback_arg)
+{
+	HeapTuple	tuple;
+	Form_pg_authid roleform;
+
+	if (roleid == GetUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetOuterUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetSessionUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("session user cannot be dropped")));
+
+	tuple = SearchSysCache1(AUTHOID, ObjectIdGetDatum(roleid));
+	if (!HeapTupleIsValid(tuple))
+		ereport(ERROR,
+				(errcode(ERRCODE_UNDEFINED_OBJECT),
+				 errmsg("role \"%s\" does not exist", rolename)));
+
+	roleform = (Form_pg_authid) GETSTRUCT(tuple);
+
+	/*
+	 * For safety's sake, we allow createrole holders to drop ordinary roles
+	 * but not superuser roles, and only if they also have ADMIN OPTION.
+	 */
+	if (roleform->rolsuper && !superuser())
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
+						   "SUPERUSER", "SUPERUSER")));
+	if (!is_admin_of_role(GetUserId(), roleid))
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
+						   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
+
+	ReleaseSysCache(tuple);
+}
+
+
 /*
  * DROP ROLE
  */
@@ -1119,9 +1263,7 @@ DropRole(DropRoleStmt *stmt)
 	{
 		RoleSpec   *rolspec = lfirst(item);
 		char	   *role;
-		HeapTuple	tuple,
-					tmp_tuple;
-		Form_pg_authid roleform;
+		HeapTuple	tmp_tuple;
 		ScanKeyData scankey;
 		SysScanDesc sscan;
 		Oid			roleid;
@@ -1132,71 +1274,27 @@ DropRole(DropRoleStmt *stmt)
 					 errmsg("cannot use special role specifier in DROP ROLE")));
 		role = rolspec->rolename;
 
-		tuple = SearchSysCache1(AUTHNAME, PointerGetDatum(role));
-		if (!HeapTupleIsValid(tuple))
-		{
-			if (!stmt->missing_ok)
-			{
-				ereport(ERROR,
-						(errcode(ERRCODE_UNDEFINED_OBJECT),
-						 errmsg("role \"%s\" does not exist", role)));
-			}
-			else
-			{
-				ereport(NOTICE,
-						(errmsg("role \"%s\" does not exist, skipping",
-								role)));
-			}
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP or ALTER commits between name
+		 * resolution and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(role, AccessExclusiveLock, stmt->missing_ok,
+								RoleNameCallbackForDropRole, NULL);
 
+		if (!OidIsValid(roleid))
+		{
+			/* missing_ok case: role doesn't exist */
+			ereport(NOTICE,
+					(errmsg("role \"%s\" does not exist, skipping",
+							role)));
 			continue;
 		}
 
-		roleform = (Form_pg_authid) GETSTRUCT(tuple);
-		roleid = roleform->oid;
-
-		if (roleid == GetUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetOuterUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetSessionUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("session user cannot be dropped")));
-
-		/*
-		 * For safety's sake, we allow createrole holders to drop ordinary
-		 * roles but not superuser roles, and only if they also have ADMIN
-		 * OPTION.
-		 */
-		if (roleform->rolsuper && !superuser())
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
-							   "SUPERUSER", "SUPERUSER")));
-		if (!is_admin_of_role(GetUserId(), roleid))
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
-							   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
-
 		/* DROP hook for the role being removed */
 		InvokeObjectDropHook(AuthIdRelationId, roleid, 0);
 
-		/* Don't leak the syscache tuple */
-		ReleaseSysCache(tuple);
-
-		/*
-		 * Lock the role, so nobody can add dependencies to her while we drop
-		 * her.  We keep the lock until the end of transaction.
-		 */
-		LockSharedObject(AuthIdRelationId, roleid, 0, AccessExclusiveLock);
-
 		/*
 		 * If there is a pg_auth_members entry that has one of the roles to be
 		 * dropped as the roleid or member, it should be silently removed, but
@@ -1484,6 +1582,21 @@ RenameRole(const char *oldname, const char *newname)
 	return address;
 }
 
+/*
+ * Before acquiring a role lock for GRANT/REVOKE, check that the current user
+ * has authorization to grant/revoke membership in the specified role.
+ */
+static void
+RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+							 Oid oldroleid, void *callback_arg)
+{
+	bool		is_grant = *((bool *) callback_arg);
+	Oid			currentUserId = GetUserId();
+
+	check_role_membership_authorization(currentUserId, roleid, is_grant);
+}
+
+
 /*
  * GrantRoleStmt
  *
@@ -1568,9 +1681,14 @@ GrantRole(ParseState *pstate, GrantRoleStmt *stmt)
 					(errcode(ERRCODE_INVALID_GRANT_OPERATION),
 					 errmsg("column names cannot be included in GRANT/REVOKE ROLE")));
 
-		roleid = get_role_oid(rolename, false);
-		check_role_membership_authorization(currentUserId,
-											roleid, stmt->is_grant);
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP commits between name resolution
+		 * and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(rolename, ShareUpdateExclusiveLock, false,
+								RoleNameCallbackForGrantRole, &stmt->is_grant);
 		if (stmt->is_grant)
 			AddRoleMems(currentUserId, rolename, roleid,
 						stmt->grantee_roles, grantee_ids,
diff --git a/src/include/commands/user.h b/src/include/commands/user.h
index 97dcb93791b..17263452250 100644
--- a/src/include/commands/user.h
+++ b/src/include/commands/user.h
@@ -21,6 +21,15 @@
 extern PGDLLIMPORT int Password_encryption; /* values from enum PasswordType */
 extern PGDLLIMPORT char *createrole_self_grant;
 
+/* Callback for RoleNameGetOid, invoked after name resolution but before locking */
+typedef void (*RoleNameGetOidCallback) (const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+
+extern Oid	RoleNameGetOid(const char *rolename, LOCKMODE lockmode,
+						   bool missing_ok,
+						   RoleNameGetOidCallback callback,
+						   void *callback_arg);
+
 /* Hook to check passwords in CreateRole() and AlterRole() */
 typedef void (*check_password_hook_type) (const char *username, const char *shadow_pass, PasswordType password_type, Datum validuntil_time, bool validuntil_null);
 
-- 
2.34.1


--xlWYagR37YwJXkF8--





^ permalink  raw  reply  [nested|flat] 173+ messages in thread

* [PATCH v1] Add RoleNameGetOid() with invalidation-based retry loop for DropRole()/GrantRole()
@ 2026-07-04 04:47 Bertrand Drouvot <[email protected]>
  0 siblings, 0 replies; 173+ messages in thread

From: Bertrand Drouvot @ 2026-07-04 04:47 UTC (permalink / raw)

DropRole() and GrantRole() resolve the role name to an OID before acquiring
LockSharedObject() on the role. A concurrent session that commits a DROP ROLE
between the read and the lock acquisition leaves the first session acting on a
stale OID.

This commit fixes the races by using the same approach as RangeVarGetRelidExtended():
It encapsulates name resolution, permission checking (via a caller-supplied callback),
and lock acquisition inside a retry loop driven by SharedInvalidMessageCounter.
If invalidation messages arrive between name resolution and locking, indicating
concurrent DDL, the function retries.

The lock is kept across retries and only released if the name resolves to a
different OID on the next iteration.

Two callbacks are provided:
- RoleNameCallbackForDropRole(): checks current/session user, superuser attribute,
and ADMIN OPTION privilege before locking. This is similar to what DropRole() is
currently doing before LockSharedObject().
- RoleNameCallbackForGrantRole(): calls check_role_membership_authorization() to
verify the current user can grant/revoke membership. This is similar to what GrantRole()
is currently doing before calling AddRoleMems()/DelRoleMems().

DropRole() and GrantRole() now call RoleNameGetOid() with appropriate lock
levels.

AlterRole() does not need the fix because it calls CatalogTupleUpdate() on the
pg_authid tuple before AddRoleMems(), which blocks a concurrent DROP ROLE.

Author: Bertrand Drouvot <[email protected]>
Reviewed-by:
Discussion:
---
 src/backend/commands/user.c | 246 ++++++++++++++++++++++++++----------
 src/include/commands/user.h |   9 ++
 2 files changed, 191 insertions(+), 64 deletions(-)
  95.5% src/backend/commands/
   4.4% src/include/commands/

diff --git a/src/backend/commands/user.c b/src/backend/commands/user.c
index be11c49f919..6d8e2fa8813 100644
--- a/src/backend/commands/user.c
+++ b/src/backend/commands/user.c
@@ -34,6 +34,7 @@
 #include "miscadmin.h"
 #include "port/pg_bitutils.h"
 #include "storage/lmgr.h"
+#include "storage/sinval.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
 #include "utils/catcache.h"
@@ -116,6 +117,10 @@ static void plan_recursive_revoke(CatCList *memlist,
 								  bool revoke_admin_option_only,
 								  DropBehavior behavior);
 static void InitGrantRoleOptions(GrantRoleOptions *popt);
+static void RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+static void RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+										 Oid oldroleid, void *callback_arg);
 
 
 /* Check if current user has createrole privileges */
@@ -126,6 +131,92 @@ have_createrole_privilege(void)
 }
 
 
+/*
+ * RoleNameGetOid
+ *		Given a role name, look up its OID, lock it, and return the OID.
+ *
+ * This follows the same pattern as RangeVarGetRelidExtended():
+ * name resolution, permission check (via callback), and lock acquisition are
+ * performed inside a retry loop. If invalidation messages arrive during the
+ * process (indicating concurrent DDL), we retry to ensure the name still
+ * resolves to the same OID.
+ *
+ * The callback is invoked before locking, giving callers a chance to check
+ * permissions. It receives the current rolename, the resolved OID, the
+ * previous OID (InvalidOid on first iteration), and a caller-supplied arg.
+ * If the callback raises an error, the function aborts without locking.
+ *
+ * If missing_ok is true and the role does not exist, returns InvalidOid.
+ * Otherwise, raises an error.
+ */
+Oid
+RoleNameGetOid(const char *rolename, LOCKMODE lockmode, bool missing_ok,
+			   RoleNameGetOidCallback callback, void *callback_arg)
+{
+	uint64		inval_count;
+	Oid			roleid;
+	Oid			oldroleid = InvalidOid;
+	bool		retry = false;
+
+	for (;;)
+	{
+		/*
+		 * Remember the current invalidation count so we can detect concurrent
+		 * DDL after locking.
+		 */
+		inval_count = SharedInvalidMessageCounter;
+
+		/* Look up the role name */
+		roleid = get_role_oid(rolename, true);
+
+		if (!OidIsValid(roleid))
+		{
+			if (!missing_ok)
+				ereport(ERROR,
+						(errcode(ERRCODE_UNDEFINED_OBJECT),
+						 errmsg("role \"%s\" does not exist", rolename)));
+			return InvalidOid;
+		}
+
+		/*
+		 * Invoke caller-supplied callback before locking. This is a good
+		 * place to check permissions: we haven't taken the lock yet, but we
+		 * know the OID we intend to lock. If concurrent DDL changes things,
+		 * the callback will be invoked again on the next iteration.
+		 */
+		if (callback)
+			callback(rolename, roleid, oldroleid, callback_arg);
+
+		/*
+		 * If upon retry we get back the same OID, the invalidation messages
+		 * did not change the final answer. So we're done.
+		 *
+		 * If we got a different OID, we've locked the role that used to have
+		 * this name rather than the one that does now. Release the old lock.
+		 */
+		if (retry)
+		{
+			if (roleid == oldroleid)
+				break;
+			UnlockSharedObject(AuthIdRelationId, oldroleid, 0, lockmode);
+		}
+
+		/* Lock the role */
+		LockSharedObject(AuthIdRelationId, roleid, 0, lockmode);
+
+		/* If no invalidation messages were processed, we're done */
+		if (inval_count == SharedInvalidMessageCounter)
+			break;
+
+		/* Something may have changed, retry */
+		retry = true;
+		oldroleid = roleid;
+	}
+
+	return roleid;
+}
+
+
 /*
  * CREATE ROLE
  */
@@ -1090,6 +1181,59 @@ AlterRoleSet(AlterRoleSetStmt *stmt)
 }
 
 
+/*
+ * Before acquiring a role lock for DROP ROLE, check that the role is not the
+ * current/session user and that the caller has sufficient privileges to drop it.
+ */
+static void
+RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+							Oid oldroleid, void *callback_arg)
+{
+	HeapTuple	tuple;
+	Form_pg_authid roleform;
+
+	if (roleid == GetUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetOuterUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetSessionUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("session user cannot be dropped")));
+
+	tuple = SearchSysCache1(AUTHOID, ObjectIdGetDatum(roleid));
+	if (!HeapTupleIsValid(tuple))
+		ereport(ERROR,
+				(errcode(ERRCODE_UNDEFINED_OBJECT),
+				 errmsg("role \"%s\" does not exist", rolename)));
+
+	roleform = (Form_pg_authid) GETSTRUCT(tuple);
+
+	/*
+	 * For safety's sake, we allow createrole holders to drop ordinary roles
+	 * but not superuser roles, and only if they also have ADMIN OPTION.
+	 */
+	if (roleform->rolsuper && !superuser())
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
+						   "SUPERUSER", "SUPERUSER")));
+	if (!is_admin_of_role(GetUserId(), roleid))
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
+						   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
+
+	ReleaseSysCache(tuple);
+}
+
+
 /*
  * DROP ROLE
  */
@@ -1119,9 +1263,7 @@ DropRole(DropRoleStmt *stmt)
 	{
 		RoleSpec   *rolspec = lfirst(item);
 		char	   *role;
-		HeapTuple	tuple,
-					tmp_tuple;
-		Form_pg_authid roleform;
+		HeapTuple	tmp_tuple;
 		ScanKeyData scankey;
 		SysScanDesc sscan;
 		Oid			roleid;
@@ -1132,71 +1274,27 @@ DropRole(DropRoleStmt *stmt)
 					 errmsg("cannot use special role specifier in DROP ROLE")));
 		role = rolspec->rolename;
 
-		tuple = SearchSysCache1(AUTHNAME, PointerGetDatum(role));
-		if (!HeapTupleIsValid(tuple))
-		{
-			if (!stmt->missing_ok)
-			{
-				ereport(ERROR,
-						(errcode(ERRCODE_UNDEFINED_OBJECT),
-						 errmsg("role \"%s\" does not exist", role)));
-			}
-			else
-			{
-				ereport(NOTICE,
-						(errmsg("role \"%s\" does not exist, skipping",
-								role)));
-			}
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP or ALTER commits between name
+		 * resolution and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(role, AccessExclusiveLock, stmt->missing_ok,
+								RoleNameCallbackForDropRole, NULL);
 
+		if (!OidIsValid(roleid))
+		{
+			/* missing_ok case: role doesn't exist */
+			ereport(NOTICE,
+					(errmsg("role \"%s\" does not exist, skipping",
+							role)));
 			continue;
 		}
 
-		roleform = (Form_pg_authid) GETSTRUCT(tuple);
-		roleid = roleform->oid;
-
-		if (roleid == GetUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetOuterUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetSessionUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("session user cannot be dropped")));
-
-		/*
-		 * For safety's sake, we allow createrole holders to drop ordinary
-		 * roles but not superuser roles, and only if they also have ADMIN
-		 * OPTION.
-		 */
-		if (roleform->rolsuper && !superuser())
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
-							   "SUPERUSER", "SUPERUSER")));
-		if (!is_admin_of_role(GetUserId(), roleid))
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
-							   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
-
 		/* DROP hook for the role being removed */
 		InvokeObjectDropHook(AuthIdRelationId, roleid, 0);
 
-		/* Don't leak the syscache tuple */
-		ReleaseSysCache(tuple);
-
-		/*
-		 * Lock the role, so nobody can add dependencies to her while we drop
-		 * her.  We keep the lock until the end of transaction.
-		 */
-		LockSharedObject(AuthIdRelationId, roleid, 0, AccessExclusiveLock);
-
 		/*
 		 * If there is a pg_auth_members entry that has one of the roles to be
 		 * dropped as the roleid or member, it should be silently removed, but
@@ -1484,6 +1582,21 @@ RenameRole(const char *oldname, const char *newname)
 	return address;
 }
 
+/*
+ * Before acquiring a role lock for GRANT/REVOKE, check that the current user
+ * has authorization to grant/revoke membership in the specified role.
+ */
+static void
+RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+							 Oid oldroleid, void *callback_arg)
+{
+	bool		is_grant = *((bool *) callback_arg);
+	Oid			currentUserId = GetUserId();
+
+	check_role_membership_authorization(currentUserId, roleid, is_grant);
+}
+
+
 /*
  * GrantRoleStmt
  *
@@ -1568,9 +1681,14 @@ GrantRole(ParseState *pstate, GrantRoleStmt *stmt)
 					(errcode(ERRCODE_INVALID_GRANT_OPERATION),
 					 errmsg("column names cannot be included in GRANT/REVOKE ROLE")));
 
-		roleid = get_role_oid(rolename, false);
-		check_role_membership_authorization(currentUserId,
-											roleid, stmt->is_grant);
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP commits between name resolution
+		 * and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(rolename, ShareUpdateExclusiveLock, false,
+								RoleNameCallbackForGrantRole, &stmt->is_grant);
 		if (stmt->is_grant)
 			AddRoleMems(currentUserId, rolename, roleid,
 						stmt->grantee_roles, grantee_ids,
diff --git a/src/include/commands/user.h b/src/include/commands/user.h
index 97dcb93791b..17263452250 100644
--- a/src/include/commands/user.h
+++ b/src/include/commands/user.h
@@ -21,6 +21,15 @@
 extern PGDLLIMPORT int Password_encryption; /* values from enum PasswordType */
 extern PGDLLIMPORT char *createrole_self_grant;
 
+/* Callback for RoleNameGetOid, invoked after name resolution but before locking */
+typedef void (*RoleNameGetOidCallback) (const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+
+extern Oid	RoleNameGetOid(const char *rolename, LOCKMODE lockmode,
+						   bool missing_ok,
+						   RoleNameGetOidCallback callback,
+						   void *callback_arg);
+
 /* Hook to check passwords in CreateRole() and AlterRole() */
 typedef void (*check_password_hook_type) (const char *username, const char *shadow_pass, PasswordType password_type, Datum validuntil_time, bool validuntil_null);
 
-- 
2.34.1


--xlWYagR37YwJXkF8--





^ permalink  raw  reply  [nested|flat] 173+ messages in thread

* [PATCH v1] Add RoleNameGetOid() with invalidation-based retry loop for DropRole()/GrantRole()
@ 2026-07-04 04:47 Bertrand Drouvot <[email protected]>
  0 siblings, 0 replies; 173+ messages in thread

From: Bertrand Drouvot @ 2026-07-04 04:47 UTC (permalink / raw)

DropRole() and GrantRole() resolve the role name to an OID before acquiring
LockSharedObject() on the role. A concurrent session that commits a DROP ROLE
between the read and the lock acquisition leaves the first session acting on a
stale OID.

This commit fixes the races by using the same approach as RangeVarGetRelidExtended():
It encapsulates name resolution, permission checking (via a caller-supplied callback),
and lock acquisition inside a retry loop driven by SharedInvalidMessageCounter.
If invalidation messages arrive between name resolution and locking, indicating
concurrent DDL, the function retries.

The lock is kept across retries and only released if the name resolves to a
different OID on the next iteration.

Two callbacks are provided:
- RoleNameCallbackForDropRole(): checks current/session user, superuser attribute,
and ADMIN OPTION privilege before locking. This is similar to what DropRole() is
currently doing before LockSharedObject().
- RoleNameCallbackForGrantRole(): calls check_role_membership_authorization() to
verify the current user can grant/revoke membership. This is similar to what GrantRole()
is currently doing before calling AddRoleMems()/DelRoleMems().

DropRole() and GrantRole() now call RoleNameGetOid() with appropriate lock
levels.

AlterRole() does not need the fix because it calls CatalogTupleUpdate() on the
pg_authid tuple before AddRoleMems(), which blocks a concurrent DROP ROLE.

Author: Bertrand Drouvot <[email protected]>
Reviewed-by:
Discussion:
---
 src/backend/commands/user.c | 246 ++++++++++++++++++++++++++----------
 src/include/commands/user.h |   9 ++
 2 files changed, 191 insertions(+), 64 deletions(-)
  95.5% src/backend/commands/
   4.4% src/include/commands/

diff --git a/src/backend/commands/user.c b/src/backend/commands/user.c
index be11c49f919..6d8e2fa8813 100644
--- a/src/backend/commands/user.c
+++ b/src/backend/commands/user.c
@@ -34,6 +34,7 @@
 #include "miscadmin.h"
 #include "port/pg_bitutils.h"
 #include "storage/lmgr.h"
+#include "storage/sinval.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
 #include "utils/catcache.h"
@@ -116,6 +117,10 @@ static void plan_recursive_revoke(CatCList *memlist,
 								  bool revoke_admin_option_only,
 								  DropBehavior behavior);
 static void InitGrantRoleOptions(GrantRoleOptions *popt);
+static void RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+static void RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+										 Oid oldroleid, void *callback_arg);
 
 
 /* Check if current user has createrole privileges */
@@ -126,6 +131,92 @@ have_createrole_privilege(void)
 }
 
 
+/*
+ * RoleNameGetOid
+ *		Given a role name, look up its OID, lock it, and return the OID.
+ *
+ * This follows the same pattern as RangeVarGetRelidExtended():
+ * name resolution, permission check (via callback), and lock acquisition are
+ * performed inside a retry loop. If invalidation messages arrive during the
+ * process (indicating concurrent DDL), we retry to ensure the name still
+ * resolves to the same OID.
+ *
+ * The callback is invoked before locking, giving callers a chance to check
+ * permissions. It receives the current rolename, the resolved OID, the
+ * previous OID (InvalidOid on first iteration), and a caller-supplied arg.
+ * If the callback raises an error, the function aborts without locking.
+ *
+ * If missing_ok is true and the role does not exist, returns InvalidOid.
+ * Otherwise, raises an error.
+ */
+Oid
+RoleNameGetOid(const char *rolename, LOCKMODE lockmode, bool missing_ok,
+			   RoleNameGetOidCallback callback, void *callback_arg)
+{
+	uint64		inval_count;
+	Oid			roleid;
+	Oid			oldroleid = InvalidOid;
+	bool		retry = false;
+
+	for (;;)
+	{
+		/*
+		 * Remember the current invalidation count so we can detect concurrent
+		 * DDL after locking.
+		 */
+		inval_count = SharedInvalidMessageCounter;
+
+		/* Look up the role name */
+		roleid = get_role_oid(rolename, true);
+
+		if (!OidIsValid(roleid))
+		{
+			if (!missing_ok)
+				ereport(ERROR,
+						(errcode(ERRCODE_UNDEFINED_OBJECT),
+						 errmsg("role \"%s\" does not exist", rolename)));
+			return InvalidOid;
+		}
+
+		/*
+		 * Invoke caller-supplied callback before locking. This is a good
+		 * place to check permissions: we haven't taken the lock yet, but we
+		 * know the OID we intend to lock. If concurrent DDL changes things,
+		 * the callback will be invoked again on the next iteration.
+		 */
+		if (callback)
+			callback(rolename, roleid, oldroleid, callback_arg);
+
+		/*
+		 * If upon retry we get back the same OID, the invalidation messages
+		 * did not change the final answer. So we're done.
+		 *
+		 * If we got a different OID, we've locked the role that used to have
+		 * this name rather than the one that does now. Release the old lock.
+		 */
+		if (retry)
+		{
+			if (roleid == oldroleid)
+				break;
+			UnlockSharedObject(AuthIdRelationId, oldroleid, 0, lockmode);
+		}
+
+		/* Lock the role */
+		LockSharedObject(AuthIdRelationId, roleid, 0, lockmode);
+
+		/* If no invalidation messages were processed, we're done */
+		if (inval_count == SharedInvalidMessageCounter)
+			break;
+
+		/* Something may have changed, retry */
+		retry = true;
+		oldroleid = roleid;
+	}
+
+	return roleid;
+}
+
+
 /*
  * CREATE ROLE
  */
@@ -1090,6 +1181,59 @@ AlterRoleSet(AlterRoleSetStmt *stmt)
 }
 
 
+/*
+ * Before acquiring a role lock for DROP ROLE, check that the role is not the
+ * current/session user and that the caller has sufficient privileges to drop it.
+ */
+static void
+RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+							Oid oldroleid, void *callback_arg)
+{
+	HeapTuple	tuple;
+	Form_pg_authid roleform;
+
+	if (roleid == GetUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetOuterUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetSessionUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("session user cannot be dropped")));
+
+	tuple = SearchSysCache1(AUTHOID, ObjectIdGetDatum(roleid));
+	if (!HeapTupleIsValid(tuple))
+		ereport(ERROR,
+				(errcode(ERRCODE_UNDEFINED_OBJECT),
+				 errmsg("role \"%s\" does not exist", rolename)));
+
+	roleform = (Form_pg_authid) GETSTRUCT(tuple);
+
+	/*
+	 * For safety's sake, we allow createrole holders to drop ordinary roles
+	 * but not superuser roles, and only if they also have ADMIN OPTION.
+	 */
+	if (roleform->rolsuper && !superuser())
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
+						   "SUPERUSER", "SUPERUSER")));
+	if (!is_admin_of_role(GetUserId(), roleid))
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
+						   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
+
+	ReleaseSysCache(tuple);
+}
+
+
 /*
  * DROP ROLE
  */
@@ -1119,9 +1263,7 @@ DropRole(DropRoleStmt *stmt)
 	{
 		RoleSpec   *rolspec = lfirst(item);
 		char	   *role;
-		HeapTuple	tuple,
-					tmp_tuple;
-		Form_pg_authid roleform;
+		HeapTuple	tmp_tuple;
 		ScanKeyData scankey;
 		SysScanDesc sscan;
 		Oid			roleid;
@@ -1132,71 +1274,27 @@ DropRole(DropRoleStmt *stmt)
 					 errmsg("cannot use special role specifier in DROP ROLE")));
 		role = rolspec->rolename;
 
-		tuple = SearchSysCache1(AUTHNAME, PointerGetDatum(role));
-		if (!HeapTupleIsValid(tuple))
-		{
-			if (!stmt->missing_ok)
-			{
-				ereport(ERROR,
-						(errcode(ERRCODE_UNDEFINED_OBJECT),
-						 errmsg("role \"%s\" does not exist", role)));
-			}
-			else
-			{
-				ereport(NOTICE,
-						(errmsg("role \"%s\" does not exist, skipping",
-								role)));
-			}
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP or ALTER commits between name
+		 * resolution and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(role, AccessExclusiveLock, stmt->missing_ok,
+								RoleNameCallbackForDropRole, NULL);
 
+		if (!OidIsValid(roleid))
+		{
+			/* missing_ok case: role doesn't exist */
+			ereport(NOTICE,
+					(errmsg("role \"%s\" does not exist, skipping",
+							role)));
 			continue;
 		}
 
-		roleform = (Form_pg_authid) GETSTRUCT(tuple);
-		roleid = roleform->oid;
-
-		if (roleid == GetUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetOuterUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetSessionUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("session user cannot be dropped")));
-
-		/*
-		 * For safety's sake, we allow createrole holders to drop ordinary
-		 * roles but not superuser roles, and only if they also have ADMIN
-		 * OPTION.
-		 */
-		if (roleform->rolsuper && !superuser())
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
-							   "SUPERUSER", "SUPERUSER")));
-		if (!is_admin_of_role(GetUserId(), roleid))
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
-							   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
-
 		/* DROP hook for the role being removed */
 		InvokeObjectDropHook(AuthIdRelationId, roleid, 0);
 
-		/* Don't leak the syscache tuple */
-		ReleaseSysCache(tuple);
-
-		/*
-		 * Lock the role, so nobody can add dependencies to her while we drop
-		 * her.  We keep the lock until the end of transaction.
-		 */
-		LockSharedObject(AuthIdRelationId, roleid, 0, AccessExclusiveLock);
-
 		/*
 		 * If there is a pg_auth_members entry that has one of the roles to be
 		 * dropped as the roleid or member, it should be silently removed, but
@@ -1484,6 +1582,21 @@ RenameRole(const char *oldname, const char *newname)
 	return address;
 }
 
+/*
+ * Before acquiring a role lock for GRANT/REVOKE, check that the current user
+ * has authorization to grant/revoke membership in the specified role.
+ */
+static void
+RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+							 Oid oldroleid, void *callback_arg)
+{
+	bool		is_grant = *((bool *) callback_arg);
+	Oid			currentUserId = GetUserId();
+
+	check_role_membership_authorization(currentUserId, roleid, is_grant);
+}
+
+
 /*
  * GrantRoleStmt
  *
@@ -1568,9 +1681,14 @@ GrantRole(ParseState *pstate, GrantRoleStmt *stmt)
 					(errcode(ERRCODE_INVALID_GRANT_OPERATION),
 					 errmsg("column names cannot be included in GRANT/REVOKE ROLE")));
 
-		roleid = get_role_oid(rolename, false);
-		check_role_membership_authorization(currentUserId,
-											roleid, stmt->is_grant);
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP commits between name resolution
+		 * and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(rolename, ShareUpdateExclusiveLock, false,
+								RoleNameCallbackForGrantRole, &stmt->is_grant);
 		if (stmt->is_grant)
 			AddRoleMems(currentUserId, rolename, roleid,
 						stmt->grantee_roles, grantee_ids,
diff --git a/src/include/commands/user.h b/src/include/commands/user.h
index 97dcb93791b..17263452250 100644
--- a/src/include/commands/user.h
+++ b/src/include/commands/user.h
@@ -21,6 +21,15 @@
 extern PGDLLIMPORT int Password_encryption; /* values from enum PasswordType */
 extern PGDLLIMPORT char *createrole_self_grant;
 
+/* Callback for RoleNameGetOid, invoked after name resolution but before locking */
+typedef void (*RoleNameGetOidCallback) (const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+
+extern Oid	RoleNameGetOid(const char *rolename, LOCKMODE lockmode,
+						   bool missing_ok,
+						   RoleNameGetOidCallback callback,
+						   void *callback_arg);
+
 /* Hook to check passwords in CreateRole() and AlterRole() */
 typedef void (*check_password_hook_type) (const char *username, const char *shadow_pass, PasswordType password_type, Datum validuntil_time, bool validuntil_null);
 
-- 
2.34.1


--xlWYagR37YwJXkF8--





^ permalink  raw  reply  [nested|flat] 173+ messages in thread

* [PATCH v1] Add RoleNameGetOid() with invalidation-based retry loop for DropRole()/GrantRole()
@ 2026-07-04 04:47 Bertrand Drouvot <[email protected]>
  0 siblings, 0 replies; 173+ messages in thread

From: Bertrand Drouvot @ 2026-07-04 04:47 UTC (permalink / raw)

DropRole() and GrantRole() resolve the role name to an OID before acquiring
LockSharedObject() on the role. A concurrent session that commits a DROP ROLE
between the read and the lock acquisition leaves the first session acting on a
stale OID.

This commit fixes the races by using the same approach as RangeVarGetRelidExtended():
It encapsulates name resolution, permission checking (via a caller-supplied callback),
and lock acquisition inside a retry loop driven by SharedInvalidMessageCounter.
If invalidation messages arrive between name resolution and locking, indicating
concurrent DDL, the function retries.

The lock is kept across retries and only released if the name resolves to a
different OID on the next iteration.

Two callbacks are provided:
- RoleNameCallbackForDropRole(): checks current/session user, superuser attribute,
and ADMIN OPTION privilege before locking. This is similar to what DropRole() is
currently doing before LockSharedObject().
- RoleNameCallbackForGrantRole(): calls check_role_membership_authorization() to
verify the current user can grant/revoke membership. This is similar to what GrantRole()
is currently doing before calling AddRoleMems()/DelRoleMems().

DropRole() and GrantRole() now call RoleNameGetOid() with appropriate lock
levels.

AlterRole() does not need the fix because it calls CatalogTupleUpdate() on the
pg_authid tuple before AddRoleMems(), which blocks a concurrent DROP ROLE.

Author: Bertrand Drouvot <[email protected]>
Reviewed-by:
Discussion:
---
 src/backend/commands/user.c | 246 ++++++++++++++++++++++++++----------
 src/include/commands/user.h |   9 ++
 2 files changed, 191 insertions(+), 64 deletions(-)
  95.5% src/backend/commands/
   4.4% src/include/commands/

diff --git a/src/backend/commands/user.c b/src/backend/commands/user.c
index be11c49f919..6d8e2fa8813 100644
--- a/src/backend/commands/user.c
+++ b/src/backend/commands/user.c
@@ -34,6 +34,7 @@
 #include "miscadmin.h"
 #include "port/pg_bitutils.h"
 #include "storage/lmgr.h"
+#include "storage/sinval.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
 #include "utils/catcache.h"
@@ -116,6 +117,10 @@ static void plan_recursive_revoke(CatCList *memlist,
 								  bool revoke_admin_option_only,
 								  DropBehavior behavior);
 static void InitGrantRoleOptions(GrantRoleOptions *popt);
+static void RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+static void RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+										 Oid oldroleid, void *callback_arg);
 
 
 /* Check if current user has createrole privileges */
@@ -126,6 +131,92 @@ have_createrole_privilege(void)
 }
 
 
+/*
+ * RoleNameGetOid
+ *		Given a role name, look up its OID, lock it, and return the OID.
+ *
+ * This follows the same pattern as RangeVarGetRelidExtended():
+ * name resolution, permission check (via callback), and lock acquisition are
+ * performed inside a retry loop. If invalidation messages arrive during the
+ * process (indicating concurrent DDL), we retry to ensure the name still
+ * resolves to the same OID.
+ *
+ * The callback is invoked before locking, giving callers a chance to check
+ * permissions. It receives the current rolename, the resolved OID, the
+ * previous OID (InvalidOid on first iteration), and a caller-supplied arg.
+ * If the callback raises an error, the function aborts without locking.
+ *
+ * If missing_ok is true and the role does not exist, returns InvalidOid.
+ * Otherwise, raises an error.
+ */
+Oid
+RoleNameGetOid(const char *rolename, LOCKMODE lockmode, bool missing_ok,
+			   RoleNameGetOidCallback callback, void *callback_arg)
+{
+	uint64		inval_count;
+	Oid			roleid;
+	Oid			oldroleid = InvalidOid;
+	bool		retry = false;
+
+	for (;;)
+	{
+		/*
+		 * Remember the current invalidation count so we can detect concurrent
+		 * DDL after locking.
+		 */
+		inval_count = SharedInvalidMessageCounter;
+
+		/* Look up the role name */
+		roleid = get_role_oid(rolename, true);
+
+		if (!OidIsValid(roleid))
+		{
+			if (!missing_ok)
+				ereport(ERROR,
+						(errcode(ERRCODE_UNDEFINED_OBJECT),
+						 errmsg("role \"%s\" does not exist", rolename)));
+			return InvalidOid;
+		}
+
+		/*
+		 * Invoke caller-supplied callback before locking. This is a good
+		 * place to check permissions: we haven't taken the lock yet, but we
+		 * know the OID we intend to lock. If concurrent DDL changes things,
+		 * the callback will be invoked again on the next iteration.
+		 */
+		if (callback)
+			callback(rolename, roleid, oldroleid, callback_arg);
+
+		/*
+		 * If upon retry we get back the same OID, the invalidation messages
+		 * did not change the final answer. So we're done.
+		 *
+		 * If we got a different OID, we've locked the role that used to have
+		 * this name rather than the one that does now. Release the old lock.
+		 */
+		if (retry)
+		{
+			if (roleid == oldroleid)
+				break;
+			UnlockSharedObject(AuthIdRelationId, oldroleid, 0, lockmode);
+		}
+
+		/* Lock the role */
+		LockSharedObject(AuthIdRelationId, roleid, 0, lockmode);
+
+		/* If no invalidation messages were processed, we're done */
+		if (inval_count == SharedInvalidMessageCounter)
+			break;
+
+		/* Something may have changed, retry */
+		retry = true;
+		oldroleid = roleid;
+	}
+
+	return roleid;
+}
+
+
 /*
  * CREATE ROLE
  */
@@ -1090,6 +1181,59 @@ AlterRoleSet(AlterRoleSetStmt *stmt)
 }
 
 
+/*
+ * Before acquiring a role lock for DROP ROLE, check that the role is not the
+ * current/session user and that the caller has sufficient privileges to drop it.
+ */
+static void
+RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+							Oid oldroleid, void *callback_arg)
+{
+	HeapTuple	tuple;
+	Form_pg_authid roleform;
+
+	if (roleid == GetUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetOuterUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetSessionUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("session user cannot be dropped")));
+
+	tuple = SearchSysCache1(AUTHOID, ObjectIdGetDatum(roleid));
+	if (!HeapTupleIsValid(tuple))
+		ereport(ERROR,
+				(errcode(ERRCODE_UNDEFINED_OBJECT),
+				 errmsg("role \"%s\" does not exist", rolename)));
+
+	roleform = (Form_pg_authid) GETSTRUCT(tuple);
+
+	/*
+	 * For safety's sake, we allow createrole holders to drop ordinary roles
+	 * but not superuser roles, and only if they also have ADMIN OPTION.
+	 */
+	if (roleform->rolsuper && !superuser())
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
+						   "SUPERUSER", "SUPERUSER")));
+	if (!is_admin_of_role(GetUserId(), roleid))
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
+						   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
+
+	ReleaseSysCache(tuple);
+}
+
+
 /*
  * DROP ROLE
  */
@@ -1119,9 +1263,7 @@ DropRole(DropRoleStmt *stmt)
 	{
 		RoleSpec   *rolspec = lfirst(item);
 		char	   *role;
-		HeapTuple	tuple,
-					tmp_tuple;
-		Form_pg_authid roleform;
+		HeapTuple	tmp_tuple;
 		ScanKeyData scankey;
 		SysScanDesc sscan;
 		Oid			roleid;
@@ -1132,71 +1274,27 @@ DropRole(DropRoleStmt *stmt)
 					 errmsg("cannot use special role specifier in DROP ROLE")));
 		role = rolspec->rolename;
 
-		tuple = SearchSysCache1(AUTHNAME, PointerGetDatum(role));
-		if (!HeapTupleIsValid(tuple))
-		{
-			if (!stmt->missing_ok)
-			{
-				ereport(ERROR,
-						(errcode(ERRCODE_UNDEFINED_OBJECT),
-						 errmsg("role \"%s\" does not exist", role)));
-			}
-			else
-			{
-				ereport(NOTICE,
-						(errmsg("role \"%s\" does not exist, skipping",
-								role)));
-			}
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP or ALTER commits between name
+		 * resolution and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(role, AccessExclusiveLock, stmt->missing_ok,
+								RoleNameCallbackForDropRole, NULL);
 
+		if (!OidIsValid(roleid))
+		{
+			/* missing_ok case: role doesn't exist */
+			ereport(NOTICE,
+					(errmsg("role \"%s\" does not exist, skipping",
+							role)));
 			continue;
 		}
 
-		roleform = (Form_pg_authid) GETSTRUCT(tuple);
-		roleid = roleform->oid;
-
-		if (roleid == GetUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetOuterUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetSessionUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("session user cannot be dropped")));
-
-		/*
-		 * For safety's sake, we allow createrole holders to drop ordinary
-		 * roles but not superuser roles, and only if they also have ADMIN
-		 * OPTION.
-		 */
-		if (roleform->rolsuper && !superuser())
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
-							   "SUPERUSER", "SUPERUSER")));
-		if (!is_admin_of_role(GetUserId(), roleid))
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
-							   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
-
 		/* DROP hook for the role being removed */
 		InvokeObjectDropHook(AuthIdRelationId, roleid, 0);
 
-		/* Don't leak the syscache tuple */
-		ReleaseSysCache(tuple);
-
-		/*
-		 * Lock the role, so nobody can add dependencies to her while we drop
-		 * her.  We keep the lock until the end of transaction.
-		 */
-		LockSharedObject(AuthIdRelationId, roleid, 0, AccessExclusiveLock);
-
 		/*
 		 * If there is a pg_auth_members entry that has one of the roles to be
 		 * dropped as the roleid or member, it should be silently removed, but
@@ -1484,6 +1582,21 @@ RenameRole(const char *oldname, const char *newname)
 	return address;
 }
 
+/*
+ * Before acquiring a role lock for GRANT/REVOKE, check that the current user
+ * has authorization to grant/revoke membership in the specified role.
+ */
+static void
+RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+							 Oid oldroleid, void *callback_arg)
+{
+	bool		is_grant = *((bool *) callback_arg);
+	Oid			currentUserId = GetUserId();
+
+	check_role_membership_authorization(currentUserId, roleid, is_grant);
+}
+
+
 /*
  * GrantRoleStmt
  *
@@ -1568,9 +1681,14 @@ GrantRole(ParseState *pstate, GrantRoleStmt *stmt)
 					(errcode(ERRCODE_INVALID_GRANT_OPERATION),
 					 errmsg("column names cannot be included in GRANT/REVOKE ROLE")));
 
-		roleid = get_role_oid(rolename, false);
-		check_role_membership_authorization(currentUserId,
-											roleid, stmt->is_grant);
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP commits between name resolution
+		 * and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(rolename, ShareUpdateExclusiveLock, false,
+								RoleNameCallbackForGrantRole, &stmt->is_grant);
 		if (stmt->is_grant)
 			AddRoleMems(currentUserId, rolename, roleid,
 						stmt->grantee_roles, grantee_ids,
diff --git a/src/include/commands/user.h b/src/include/commands/user.h
index 97dcb93791b..17263452250 100644
--- a/src/include/commands/user.h
+++ b/src/include/commands/user.h
@@ -21,6 +21,15 @@
 extern PGDLLIMPORT int Password_encryption; /* values from enum PasswordType */
 extern PGDLLIMPORT char *createrole_self_grant;
 
+/* Callback for RoleNameGetOid, invoked after name resolution but before locking */
+typedef void (*RoleNameGetOidCallback) (const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+
+extern Oid	RoleNameGetOid(const char *rolename, LOCKMODE lockmode,
+						   bool missing_ok,
+						   RoleNameGetOidCallback callback,
+						   void *callback_arg);
+
 /* Hook to check passwords in CreateRole() and AlterRole() */
 typedef void (*check_password_hook_type) (const char *username, const char *shadow_pass, PasswordType password_type, Datum validuntil_time, bool validuntil_null);
 
-- 
2.34.1


--xlWYagR37YwJXkF8--





^ permalink  raw  reply  [nested|flat] 173+ messages in thread

* [PATCH v1] Add RoleNameGetOid() with invalidation-based retry loop for DropRole()/GrantRole()
@ 2026-07-04 04:47 Bertrand Drouvot <[email protected]>
  0 siblings, 0 replies; 173+ messages in thread

From: Bertrand Drouvot @ 2026-07-04 04:47 UTC (permalink / raw)

DropRole() and GrantRole() resolve the role name to an OID before acquiring
LockSharedObject() on the role. A concurrent session that commits a DROP ROLE
between the read and the lock acquisition leaves the first session acting on a
stale OID.

This commit fixes the races by using the same approach as RangeVarGetRelidExtended():
It encapsulates name resolution, permission checking (via a caller-supplied callback),
and lock acquisition inside a retry loop driven by SharedInvalidMessageCounter.
If invalidation messages arrive between name resolution and locking, indicating
concurrent DDL, the function retries.

The lock is kept across retries and only released if the name resolves to a
different OID on the next iteration.

Two callbacks are provided:
- RoleNameCallbackForDropRole(): checks current/session user, superuser attribute,
and ADMIN OPTION privilege before locking. This is similar to what DropRole() is
currently doing before LockSharedObject().
- RoleNameCallbackForGrantRole(): calls check_role_membership_authorization() to
verify the current user can grant/revoke membership. This is similar to what GrantRole()
is currently doing before calling AddRoleMems()/DelRoleMems().

DropRole() and GrantRole() now call RoleNameGetOid() with appropriate lock
levels.

AlterRole() does not need the fix because it calls CatalogTupleUpdate() on the
pg_authid tuple before AddRoleMems(), which blocks a concurrent DROP ROLE.

Author: Bertrand Drouvot <[email protected]>
Reviewed-by:
Discussion:
---
 src/backend/commands/user.c | 246 ++++++++++++++++++++++++++----------
 src/include/commands/user.h |   9 ++
 2 files changed, 191 insertions(+), 64 deletions(-)
  95.5% src/backend/commands/
   4.4% src/include/commands/

diff --git a/src/backend/commands/user.c b/src/backend/commands/user.c
index be11c49f919..6d8e2fa8813 100644
--- a/src/backend/commands/user.c
+++ b/src/backend/commands/user.c
@@ -34,6 +34,7 @@
 #include "miscadmin.h"
 #include "port/pg_bitutils.h"
 #include "storage/lmgr.h"
+#include "storage/sinval.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
 #include "utils/catcache.h"
@@ -116,6 +117,10 @@ static void plan_recursive_revoke(CatCList *memlist,
 								  bool revoke_admin_option_only,
 								  DropBehavior behavior);
 static void InitGrantRoleOptions(GrantRoleOptions *popt);
+static void RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+static void RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+										 Oid oldroleid, void *callback_arg);
 
 
 /* Check if current user has createrole privileges */
@@ -126,6 +131,92 @@ have_createrole_privilege(void)
 }
 
 
+/*
+ * RoleNameGetOid
+ *		Given a role name, look up its OID, lock it, and return the OID.
+ *
+ * This follows the same pattern as RangeVarGetRelidExtended():
+ * name resolution, permission check (via callback), and lock acquisition are
+ * performed inside a retry loop. If invalidation messages arrive during the
+ * process (indicating concurrent DDL), we retry to ensure the name still
+ * resolves to the same OID.
+ *
+ * The callback is invoked before locking, giving callers a chance to check
+ * permissions. It receives the current rolename, the resolved OID, the
+ * previous OID (InvalidOid on first iteration), and a caller-supplied arg.
+ * If the callback raises an error, the function aborts without locking.
+ *
+ * If missing_ok is true and the role does not exist, returns InvalidOid.
+ * Otherwise, raises an error.
+ */
+Oid
+RoleNameGetOid(const char *rolename, LOCKMODE lockmode, bool missing_ok,
+			   RoleNameGetOidCallback callback, void *callback_arg)
+{
+	uint64		inval_count;
+	Oid			roleid;
+	Oid			oldroleid = InvalidOid;
+	bool		retry = false;
+
+	for (;;)
+	{
+		/*
+		 * Remember the current invalidation count so we can detect concurrent
+		 * DDL after locking.
+		 */
+		inval_count = SharedInvalidMessageCounter;
+
+		/* Look up the role name */
+		roleid = get_role_oid(rolename, true);
+
+		if (!OidIsValid(roleid))
+		{
+			if (!missing_ok)
+				ereport(ERROR,
+						(errcode(ERRCODE_UNDEFINED_OBJECT),
+						 errmsg("role \"%s\" does not exist", rolename)));
+			return InvalidOid;
+		}
+
+		/*
+		 * Invoke caller-supplied callback before locking. This is a good
+		 * place to check permissions: we haven't taken the lock yet, but we
+		 * know the OID we intend to lock. If concurrent DDL changes things,
+		 * the callback will be invoked again on the next iteration.
+		 */
+		if (callback)
+			callback(rolename, roleid, oldroleid, callback_arg);
+
+		/*
+		 * If upon retry we get back the same OID, the invalidation messages
+		 * did not change the final answer. So we're done.
+		 *
+		 * If we got a different OID, we've locked the role that used to have
+		 * this name rather than the one that does now. Release the old lock.
+		 */
+		if (retry)
+		{
+			if (roleid == oldroleid)
+				break;
+			UnlockSharedObject(AuthIdRelationId, oldroleid, 0, lockmode);
+		}
+
+		/* Lock the role */
+		LockSharedObject(AuthIdRelationId, roleid, 0, lockmode);
+
+		/* If no invalidation messages were processed, we're done */
+		if (inval_count == SharedInvalidMessageCounter)
+			break;
+
+		/* Something may have changed, retry */
+		retry = true;
+		oldroleid = roleid;
+	}
+
+	return roleid;
+}
+
+
 /*
  * CREATE ROLE
  */
@@ -1090,6 +1181,59 @@ AlterRoleSet(AlterRoleSetStmt *stmt)
 }
 
 
+/*
+ * Before acquiring a role lock for DROP ROLE, check that the role is not the
+ * current/session user and that the caller has sufficient privileges to drop it.
+ */
+static void
+RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+							Oid oldroleid, void *callback_arg)
+{
+	HeapTuple	tuple;
+	Form_pg_authid roleform;
+
+	if (roleid == GetUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetOuterUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetSessionUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("session user cannot be dropped")));
+
+	tuple = SearchSysCache1(AUTHOID, ObjectIdGetDatum(roleid));
+	if (!HeapTupleIsValid(tuple))
+		ereport(ERROR,
+				(errcode(ERRCODE_UNDEFINED_OBJECT),
+				 errmsg("role \"%s\" does not exist", rolename)));
+
+	roleform = (Form_pg_authid) GETSTRUCT(tuple);
+
+	/*
+	 * For safety's sake, we allow createrole holders to drop ordinary roles
+	 * but not superuser roles, and only if they also have ADMIN OPTION.
+	 */
+	if (roleform->rolsuper && !superuser())
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
+						   "SUPERUSER", "SUPERUSER")));
+	if (!is_admin_of_role(GetUserId(), roleid))
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
+						   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
+
+	ReleaseSysCache(tuple);
+}
+
+
 /*
  * DROP ROLE
  */
@@ -1119,9 +1263,7 @@ DropRole(DropRoleStmt *stmt)
 	{
 		RoleSpec   *rolspec = lfirst(item);
 		char	   *role;
-		HeapTuple	tuple,
-					tmp_tuple;
-		Form_pg_authid roleform;
+		HeapTuple	tmp_tuple;
 		ScanKeyData scankey;
 		SysScanDesc sscan;
 		Oid			roleid;
@@ -1132,71 +1274,27 @@ DropRole(DropRoleStmt *stmt)
 					 errmsg("cannot use special role specifier in DROP ROLE")));
 		role = rolspec->rolename;
 
-		tuple = SearchSysCache1(AUTHNAME, PointerGetDatum(role));
-		if (!HeapTupleIsValid(tuple))
-		{
-			if (!stmt->missing_ok)
-			{
-				ereport(ERROR,
-						(errcode(ERRCODE_UNDEFINED_OBJECT),
-						 errmsg("role \"%s\" does not exist", role)));
-			}
-			else
-			{
-				ereport(NOTICE,
-						(errmsg("role \"%s\" does not exist, skipping",
-								role)));
-			}
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP or ALTER commits between name
+		 * resolution and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(role, AccessExclusiveLock, stmt->missing_ok,
+								RoleNameCallbackForDropRole, NULL);
 
+		if (!OidIsValid(roleid))
+		{
+			/* missing_ok case: role doesn't exist */
+			ereport(NOTICE,
+					(errmsg("role \"%s\" does not exist, skipping",
+							role)));
 			continue;
 		}
 
-		roleform = (Form_pg_authid) GETSTRUCT(tuple);
-		roleid = roleform->oid;
-
-		if (roleid == GetUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetOuterUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetSessionUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("session user cannot be dropped")));
-
-		/*
-		 * For safety's sake, we allow createrole holders to drop ordinary
-		 * roles but not superuser roles, and only if they also have ADMIN
-		 * OPTION.
-		 */
-		if (roleform->rolsuper && !superuser())
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
-							   "SUPERUSER", "SUPERUSER")));
-		if (!is_admin_of_role(GetUserId(), roleid))
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
-							   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
-
 		/* DROP hook for the role being removed */
 		InvokeObjectDropHook(AuthIdRelationId, roleid, 0);
 
-		/* Don't leak the syscache tuple */
-		ReleaseSysCache(tuple);
-
-		/*
-		 * Lock the role, so nobody can add dependencies to her while we drop
-		 * her.  We keep the lock until the end of transaction.
-		 */
-		LockSharedObject(AuthIdRelationId, roleid, 0, AccessExclusiveLock);
-
 		/*
 		 * If there is a pg_auth_members entry that has one of the roles to be
 		 * dropped as the roleid or member, it should be silently removed, but
@@ -1484,6 +1582,21 @@ RenameRole(const char *oldname, const char *newname)
 	return address;
 }
 
+/*
+ * Before acquiring a role lock for GRANT/REVOKE, check that the current user
+ * has authorization to grant/revoke membership in the specified role.
+ */
+static void
+RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+							 Oid oldroleid, void *callback_arg)
+{
+	bool		is_grant = *((bool *) callback_arg);
+	Oid			currentUserId = GetUserId();
+
+	check_role_membership_authorization(currentUserId, roleid, is_grant);
+}
+
+
 /*
  * GrantRoleStmt
  *
@@ -1568,9 +1681,14 @@ GrantRole(ParseState *pstate, GrantRoleStmt *stmt)
 					(errcode(ERRCODE_INVALID_GRANT_OPERATION),
 					 errmsg("column names cannot be included in GRANT/REVOKE ROLE")));
 
-		roleid = get_role_oid(rolename, false);
-		check_role_membership_authorization(currentUserId,
-											roleid, stmt->is_grant);
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP commits between name resolution
+		 * and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(rolename, ShareUpdateExclusiveLock, false,
+								RoleNameCallbackForGrantRole, &stmt->is_grant);
 		if (stmt->is_grant)
 			AddRoleMems(currentUserId, rolename, roleid,
 						stmt->grantee_roles, grantee_ids,
diff --git a/src/include/commands/user.h b/src/include/commands/user.h
index 97dcb93791b..17263452250 100644
--- a/src/include/commands/user.h
+++ b/src/include/commands/user.h
@@ -21,6 +21,15 @@
 extern PGDLLIMPORT int Password_encryption; /* values from enum PasswordType */
 extern PGDLLIMPORT char *createrole_self_grant;
 
+/* Callback for RoleNameGetOid, invoked after name resolution but before locking */
+typedef void (*RoleNameGetOidCallback) (const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+
+extern Oid	RoleNameGetOid(const char *rolename, LOCKMODE lockmode,
+						   bool missing_ok,
+						   RoleNameGetOidCallback callback,
+						   void *callback_arg);
+
 /* Hook to check passwords in CreateRole() and AlterRole() */
 typedef void (*check_password_hook_type) (const char *username, const char *shadow_pass, PasswordType password_type, Datum validuntil_time, bool validuntil_null);
 
-- 
2.34.1


--xlWYagR37YwJXkF8--





^ permalink  raw  reply  [nested|flat] 173+ messages in thread

* [PATCH v1] Add RoleNameGetOid() with invalidation-based retry loop for DropRole()/GrantRole()
@ 2026-07-04 04:47 Bertrand Drouvot <[email protected]>
  0 siblings, 0 replies; 173+ messages in thread

From: Bertrand Drouvot @ 2026-07-04 04:47 UTC (permalink / raw)

DropRole() and GrantRole() resolve the role name to an OID before acquiring
LockSharedObject() on the role. A concurrent session that commits a DROP ROLE
between the read and the lock acquisition leaves the first session acting on a
stale OID.

This commit fixes the races by using the same approach as RangeVarGetRelidExtended():
It encapsulates name resolution, permission checking (via a caller-supplied callback),
and lock acquisition inside a retry loop driven by SharedInvalidMessageCounter.
If invalidation messages arrive between name resolution and locking, indicating
concurrent DDL, the function retries.

The lock is kept across retries and only released if the name resolves to a
different OID on the next iteration.

Two callbacks are provided:
- RoleNameCallbackForDropRole(): checks current/session user, superuser attribute,
and ADMIN OPTION privilege before locking. This is similar to what DropRole() is
currently doing before LockSharedObject().
- RoleNameCallbackForGrantRole(): calls check_role_membership_authorization() to
verify the current user can grant/revoke membership. This is similar to what GrantRole()
is currently doing before calling AddRoleMems()/DelRoleMems().

DropRole() and GrantRole() now call RoleNameGetOid() with appropriate lock
levels.

AlterRole() does not need the fix because it calls CatalogTupleUpdate() on the
pg_authid tuple before AddRoleMems(), which blocks a concurrent DROP ROLE.

Author: Bertrand Drouvot <[email protected]>
Reviewed-by:
Discussion:
---
 src/backend/commands/user.c | 246 ++++++++++++++++++++++++++----------
 src/include/commands/user.h |   9 ++
 2 files changed, 191 insertions(+), 64 deletions(-)
  95.5% src/backend/commands/
   4.4% src/include/commands/

diff --git a/src/backend/commands/user.c b/src/backend/commands/user.c
index be11c49f919..6d8e2fa8813 100644
--- a/src/backend/commands/user.c
+++ b/src/backend/commands/user.c
@@ -34,6 +34,7 @@
 #include "miscadmin.h"
 #include "port/pg_bitutils.h"
 #include "storage/lmgr.h"
+#include "storage/sinval.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
 #include "utils/catcache.h"
@@ -116,6 +117,10 @@ static void plan_recursive_revoke(CatCList *memlist,
 								  bool revoke_admin_option_only,
 								  DropBehavior behavior);
 static void InitGrantRoleOptions(GrantRoleOptions *popt);
+static void RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+static void RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+										 Oid oldroleid, void *callback_arg);
 
 
 /* Check if current user has createrole privileges */
@@ -126,6 +131,92 @@ have_createrole_privilege(void)
 }
 
 
+/*
+ * RoleNameGetOid
+ *		Given a role name, look up its OID, lock it, and return the OID.
+ *
+ * This follows the same pattern as RangeVarGetRelidExtended():
+ * name resolution, permission check (via callback), and lock acquisition are
+ * performed inside a retry loop. If invalidation messages arrive during the
+ * process (indicating concurrent DDL), we retry to ensure the name still
+ * resolves to the same OID.
+ *
+ * The callback is invoked before locking, giving callers a chance to check
+ * permissions. It receives the current rolename, the resolved OID, the
+ * previous OID (InvalidOid on first iteration), and a caller-supplied arg.
+ * If the callback raises an error, the function aborts without locking.
+ *
+ * If missing_ok is true and the role does not exist, returns InvalidOid.
+ * Otherwise, raises an error.
+ */
+Oid
+RoleNameGetOid(const char *rolename, LOCKMODE lockmode, bool missing_ok,
+			   RoleNameGetOidCallback callback, void *callback_arg)
+{
+	uint64		inval_count;
+	Oid			roleid;
+	Oid			oldroleid = InvalidOid;
+	bool		retry = false;
+
+	for (;;)
+	{
+		/*
+		 * Remember the current invalidation count so we can detect concurrent
+		 * DDL after locking.
+		 */
+		inval_count = SharedInvalidMessageCounter;
+
+		/* Look up the role name */
+		roleid = get_role_oid(rolename, true);
+
+		if (!OidIsValid(roleid))
+		{
+			if (!missing_ok)
+				ereport(ERROR,
+						(errcode(ERRCODE_UNDEFINED_OBJECT),
+						 errmsg("role \"%s\" does not exist", rolename)));
+			return InvalidOid;
+		}
+
+		/*
+		 * Invoke caller-supplied callback before locking. This is a good
+		 * place to check permissions: we haven't taken the lock yet, but we
+		 * know the OID we intend to lock. If concurrent DDL changes things,
+		 * the callback will be invoked again on the next iteration.
+		 */
+		if (callback)
+			callback(rolename, roleid, oldroleid, callback_arg);
+
+		/*
+		 * If upon retry we get back the same OID, the invalidation messages
+		 * did not change the final answer. So we're done.
+		 *
+		 * If we got a different OID, we've locked the role that used to have
+		 * this name rather than the one that does now. Release the old lock.
+		 */
+		if (retry)
+		{
+			if (roleid == oldroleid)
+				break;
+			UnlockSharedObject(AuthIdRelationId, oldroleid, 0, lockmode);
+		}
+
+		/* Lock the role */
+		LockSharedObject(AuthIdRelationId, roleid, 0, lockmode);
+
+		/* If no invalidation messages were processed, we're done */
+		if (inval_count == SharedInvalidMessageCounter)
+			break;
+
+		/* Something may have changed, retry */
+		retry = true;
+		oldroleid = roleid;
+	}
+
+	return roleid;
+}
+
+
 /*
  * CREATE ROLE
  */
@@ -1090,6 +1181,59 @@ AlterRoleSet(AlterRoleSetStmt *stmt)
 }
 
 
+/*
+ * Before acquiring a role lock for DROP ROLE, check that the role is not the
+ * current/session user and that the caller has sufficient privileges to drop it.
+ */
+static void
+RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+							Oid oldroleid, void *callback_arg)
+{
+	HeapTuple	tuple;
+	Form_pg_authid roleform;
+
+	if (roleid == GetUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetOuterUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetSessionUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("session user cannot be dropped")));
+
+	tuple = SearchSysCache1(AUTHOID, ObjectIdGetDatum(roleid));
+	if (!HeapTupleIsValid(tuple))
+		ereport(ERROR,
+				(errcode(ERRCODE_UNDEFINED_OBJECT),
+				 errmsg("role \"%s\" does not exist", rolename)));
+
+	roleform = (Form_pg_authid) GETSTRUCT(tuple);
+
+	/*
+	 * For safety's sake, we allow createrole holders to drop ordinary roles
+	 * but not superuser roles, and only if they also have ADMIN OPTION.
+	 */
+	if (roleform->rolsuper && !superuser())
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
+						   "SUPERUSER", "SUPERUSER")));
+	if (!is_admin_of_role(GetUserId(), roleid))
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
+						   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
+
+	ReleaseSysCache(tuple);
+}
+
+
 /*
  * DROP ROLE
  */
@@ -1119,9 +1263,7 @@ DropRole(DropRoleStmt *stmt)
 	{
 		RoleSpec   *rolspec = lfirst(item);
 		char	   *role;
-		HeapTuple	tuple,
-					tmp_tuple;
-		Form_pg_authid roleform;
+		HeapTuple	tmp_tuple;
 		ScanKeyData scankey;
 		SysScanDesc sscan;
 		Oid			roleid;
@@ -1132,71 +1274,27 @@ DropRole(DropRoleStmt *stmt)
 					 errmsg("cannot use special role specifier in DROP ROLE")));
 		role = rolspec->rolename;
 
-		tuple = SearchSysCache1(AUTHNAME, PointerGetDatum(role));
-		if (!HeapTupleIsValid(tuple))
-		{
-			if (!stmt->missing_ok)
-			{
-				ereport(ERROR,
-						(errcode(ERRCODE_UNDEFINED_OBJECT),
-						 errmsg("role \"%s\" does not exist", role)));
-			}
-			else
-			{
-				ereport(NOTICE,
-						(errmsg("role \"%s\" does not exist, skipping",
-								role)));
-			}
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP or ALTER commits between name
+		 * resolution and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(role, AccessExclusiveLock, stmt->missing_ok,
+								RoleNameCallbackForDropRole, NULL);
 
+		if (!OidIsValid(roleid))
+		{
+			/* missing_ok case: role doesn't exist */
+			ereport(NOTICE,
+					(errmsg("role \"%s\" does not exist, skipping",
+							role)));
 			continue;
 		}
 
-		roleform = (Form_pg_authid) GETSTRUCT(tuple);
-		roleid = roleform->oid;
-
-		if (roleid == GetUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetOuterUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetSessionUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("session user cannot be dropped")));
-
-		/*
-		 * For safety's sake, we allow createrole holders to drop ordinary
-		 * roles but not superuser roles, and only if they also have ADMIN
-		 * OPTION.
-		 */
-		if (roleform->rolsuper && !superuser())
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
-							   "SUPERUSER", "SUPERUSER")));
-		if (!is_admin_of_role(GetUserId(), roleid))
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
-							   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
-
 		/* DROP hook for the role being removed */
 		InvokeObjectDropHook(AuthIdRelationId, roleid, 0);
 
-		/* Don't leak the syscache tuple */
-		ReleaseSysCache(tuple);
-
-		/*
-		 * Lock the role, so nobody can add dependencies to her while we drop
-		 * her.  We keep the lock until the end of transaction.
-		 */
-		LockSharedObject(AuthIdRelationId, roleid, 0, AccessExclusiveLock);
-
 		/*
 		 * If there is a pg_auth_members entry that has one of the roles to be
 		 * dropped as the roleid or member, it should be silently removed, but
@@ -1484,6 +1582,21 @@ RenameRole(const char *oldname, const char *newname)
 	return address;
 }
 
+/*
+ * Before acquiring a role lock for GRANT/REVOKE, check that the current user
+ * has authorization to grant/revoke membership in the specified role.
+ */
+static void
+RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+							 Oid oldroleid, void *callback_arg)
+{
+	bool		is_grant = *((bool *) callback_arg);
+	Oid			currentUserId = GetUserId();
+
+	check_role_membership_authorization(currentUserId, roleid, is_grant);
+}
+
+
 /*
  * GrantRoleStmt
  *
@@ -1568,9 +1681,14 @@ GrantRole(ParseState *pstate, GrantRoleStmt *stmt)
 					(errcode(ERRCODE_INVALID_GRANT_OPERATION),
 					 errmsg("column names cannot be included in GRANT/REVOKE ROLE")));
 
-		roleid = get_role_oid(rolename, false);
-		check_role_membership_authorization(currentUserId,
-											roleid, stmt->is_grant);
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP commits between name resolution
+		 * and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(rolename, ShareUpdateExclusiveLock, false,
+								RoleNameCallbackForGrantRole, &stmt->is_grant);
 		if (stmt->is_grant)
 			AddRoleMems(currentUserId, rolename, roleid,
 						stmt->grantee_roles, grantee_ids,
diff --git a/src/include/commands/user.h b/src/include/commands/user.h
index 97dcb93791b..17263452250 100644
--- a/src/include/commands/user.h
+++ b/src/include/commands/user.h
@@ -21,6 +21,15 @@
 extern PGDLLIMPORT int Password_encryption; /* values from enum PasswordType */
 extern PGDLLIMPORT char *createrole_self_grant;
 
+/* Callback for RoleNameGetOid, invoked after name resolution but before locking */
+typedef void (*RoleNameGetOidCallback) (const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+
+extern Oid	RoleNameGetOid(const char *rolename, LOCKMODE lockmode,
+						   bool missing_ok,
+						   RoleNameGetOidCallback callback,
+						   void *callback_arg);
+
 /* Hook to check passwords in CreateRole() and AlterRole() */
 typedef void (*check_password_hook_type) (const char *username, const char *shadow_pass, PasswordType password_type, Datum validuntil_time, bool validuntil_null);
 
-- 
2.34.1


--xlWYagR37YwJXkF8--





^ permalink  raw  reply  [nested|flat] 173+ messages in thread

* [PATCH v1] Add RoleNameGetOid() with invalidation-based retry loop for DropRole()/GrantRole()
@ 2026-07-04 04:47 Bertrand Drouvot <[email protected]>
  0 siblings, 0 replies; 173+ messages in thread

From: Bertrand Drouvot @ 2026-07-04 04:47 UTC (permalink / raw)

DropRole() and GrantRole() resolve the role name to an OID before acquiring
LockSharedObject() on the role. A concurrent session that commits a DROP ROLE
between the read and the lock acquisition leaves the first session acting on a
stale OID.

This commit fixes the races by using the same approach as RangeVarGetRelidExtended():
It encapsulates name resolution, permission checking (via a caller-supplied callback),
and lock acquisition inside a retry loop driven by SharedInvalidMessageCounter.
If invalidation messages arrive between name resolution and locking, indicating
concurrent DDL, the function retries.

The lock is kept across retries and only released if the name resolves to a
different OID on the next iteration.

Two callbacks are provided:
- RoleNameCallbackForDropRole(): checks current/session user, superuser attribute,
and ADMIN OPTION privilege before locking. This is similar to what DropRole() is
currently doing before LockSharedObject().
- RoleNameCallbackForGrantRole(): calls check_role_membership_authorization() to
verify the current user can grant/revoke membership. This is similar to what GrantRole()
is currently doing before calling AddRoleMems()/DelRoleMems().

DropRole() and GrantRole() now call RoleNameGetOid() with appropriate lock
levels.

AlterRole() does not need the fix because it calls CatalogTupleUpdate() on the
pg_authid tuple before AddRoleMems(), which blocks a concurrent DROP ROLE.

Author: Bertrand Drouvot <[email protected]>
Reviewed-by:
Discussion:
---
 src/backend/commands/user.c | 246 ++++++++++++++++++++++++++----------
 src/include/commands/user.h |   9 ++
 2 files changed, 191 insertions(+), 64 deletions(-)
  95.5% src/backend/commands/
   4.4% src/include/commands/

diff --git a/src/backend/commands/user.c b/src/backend/commands/user.c
index be11c49f919..6d8e2fa8813 100644
--- a/src/backend/commands/user.c
+++ b/src/backend/commands/user.c
@@ -34,6 +34,7 @@
 #include "miscadmin.h"
 #include "port/pg_bitutils.h"
 #include "storage/lmgr.h"
+#include "storage/sinval.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
 #include "utils/catcache.h"
@@ -116,6 +117,10 @@ static void plan_recursive_revoke(CatCList *memlist,
 								  bool revoke_admin_option_only,
 								  DropBehavior behavior);
 static void InitGrantRoleOptions(GrantRoleOptions *popt);
+static void RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+static void RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+										 Oid oldroleid, void *callback_arg);
 
 
 /* Check if current user has createrole privileges */
@@ -126,6 +131,92 @@ have_createrole_privilege(void)
 }
 
 
+/*
+ * RoleNameGetOid
+ *		Given a role name, look up its OID, lock it, and return the OID.
+ *
+ * This follows the same pattern as RangeVarGetRelidExtended():
+ * name resolution, permission check (via callback), and lock acquisition are
+ * performed inside a retry loop. If invalidation messages arrive during the
+ * process (indicating concurrent DDL), we retry to ensure the name still
+ * resolves to the same OID.
+ *
+ * The callback is invoked before locking, giving callers a chance to check
+ * permissions. It receives the current rolename, the resolved OID, the
+ * previous OID (InvalidOid on first iteration), and a caller-supplied arg.
+ * If the callback raises an error, the function aborts without locking.
+ *
+ * If missing_ok is true and the role does not exist, returns InvalidOid.
+ * Otherwise, raises an error.
+ */
+Oid
+RoleNameGetOid(const char *rolename, LOCKMODE lockmode, bool missing_ok,
+			   RoleNameGetOidCallback callback, void *callback_arg)
+{
+	uint64		inval_count;
+	Oid			roleid;
+	Oid			oldroleid = InvalidOid;
+	bool		retry = false;
+
+	for (;;)
+	{
+		/*
+		 * Remember the current invalidation count so we can detect concurrent
+		 * DDL after locking.
+		 */
+		inval_count = SharedInvalidMessageCounter;
+
+		/* Look up the role name */
+		roleid = get_role_oid(rolename, true);
+
+		if (!OidIsValid(roleid))
+		{
+			if (!missing_ok)
+				ereport(ERROR,
+						(errcode(ERRCODE_UNDEFINED_OBJECT),
+						 errmsg("role \"%s\" does not exist", rolename)));
+			return InvalidOid;
+		}
+
+		/*
+		 * Invoke caller-supplied callback before locking. This is a good
+		 * place to check permissions: we haven't taken the lock yet, but we
+		 * know the OID we intend to lock. If concurrent DDL changes things,
+		 * the callback will be invoked again on the next iteration.
+		 */
+		if (callback)
+			callback(rolename, roleid, oldroleid, callback_arg);
+
+		/*
+		 * If upon retry we get back the same OID, the invalidation messages
+		 * did not change the final answer. So we're done.
+		 *
+		 * If we got a different OID, we've locked the role that used to have
+		 * this name rather than the one that does now. Release the old lock.
+		 */
+		if (retry)
+		{
+			if (roleid == oldroleid)
+				break;
+			UnlockSharedObject(AuthIdRelationId, oldroleid, 0, lockmode);
+		}
+
+		/* Lock the role */
+		LockSharedObject(AuthIdRelationId, roleid, 0, lockmode);
+
+		/* If no invalidation messages were processed, we're done */
+		if (inval_count == SharedInvalidMessageCounter)
+			break;
+
+		/* Something may have changed, retry */
+		retry = true;
+		oldroleid = roleid;
+	}
+
+	return roleid;
+}
+
+
 /*
  * CREATE ROLE
  */
@@ -1090,6 +1181,59 @@ AlterRoleSet(AlterRoleSetStmt *stmt)
 }
 
 
+/*
+ * Before acquiring a role lock for DROP ROLE, check that the role is not the
+ * current/session user and that the caller has sufficient privileges to drop it.
+ */
+static void
+RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+							Oid oldroleid, void *callback_arg)
+{
+	HeapTuple	tuple;
+	Form_pg_authid roleform;
+
+	if (roleid == GetUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetOuterUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetSessionUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("session user cannot be dropped")));
+
+	tuple = SearchSysCache1(AUTHOID, ObjectIdGetDatum(roleid));
+	if (!HeapTupleIsValid(tuple))
+		ereport(ERROR,
+				(errcode(ERRCODE_UNDEFINED_OBJECT),
+				 errmsg("role \"%s\" does not exist", rolename)));
+
+	roleform = (Form_pg_authid) GETSTRUCT(tuple);
+
+	/*
+	 * For safety's sake, we allow createrole holders to drop ordinary roles
+	 * but not superuser roles, and only if they also have ADMIN OPTION.
+	 */
+	if (roleform->rolsuper && !superuser())
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
+						   "SUPERUSER", "SUPERUSER")));
+	if (!is_admin_of_role(GetUserId(), roleid))
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
+						   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
+
+	ReleaseSysCache(tuple);
+}
+
+
 /*
  * DROP ROLE
  */
@@ -1119,9 +1263,7 @@ DropRole(DropRoleStmt *stmt)
 	{
 		RoleSpec   *rolspec = lfirst(item);
 		char	   *role;
-		HeapTuple	tuple,
-					tmp_tuple;
-		Form_pg_authid roleform;
+		HeapTuple	tmp_tuple;
 		ScanKeyData scankey;
 		SysScanDesc sscan;
 		Oid			roleid;
@@ -1132,71 +1274,27 @@ DropRole(DropRoleStmt *stmt)
 					 errmsg("cannot use special role specifier in DROP ROLE")));
 		role = rolspec->rolename;
 
-		tuple = SearchSysCache1(AUTHNAME, PointerGetDatum(role));
-		if (!HeapTupleIsValid(tuple))
-		{
-			if (!stmt->missing_ok)
-			{
-				ereport(ERROR,
-						(errcode(ERRCODE_UNDEFINED_OBJECT),
-						 errmsg("role \"%s\" does not exist", role)));
-			}
-			else
-			{
-				ereport(NOTICE,
-						(errmsg("role \"%s\" does not exist, skipping",
-								role)));
-			}
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP or ALTER commits between name
+		 * resolution and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(role, AccessExclusiveLock, stmt->missing_ok,
+								RoleNameCallbackForDropRole, NULL);
 
+		if (!OidIsValid(roleid))
+		{
+			/* missing_ok case: role doesn't exist */
+			ereport(NOTICE,
+					(errmsg("role \"%s\" does not exist, skipping",
+							role)));
 			continue;
 		}
 
-		roleform = (Form_pg_authid) GETSTRUCT(tuple);
-		roleid = roleform->oid;
-
-		if (roleid == GetUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetOuterUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetSessionUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("session user cannot be dropped")));
-
-		/*
-		 * For safety's sake, we allow createrole holders to drop ordinary
-		 * roles but not superuser roles, and only if they also have ADMIN
-		 * OPTION.
-		 */
-		if (roleform->rolsuper && !superuser())
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
-							   "SUPERUSER", "SUPERUSER")));
-		if (!is_admin_of_role(GetUserId(), roleid))
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
-							   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
-
 		/* DROP hook for the role being removed */
 		InvokeObjectDropHook(AuthIdRelationId, roleid, 0);
 
-		/* Don't leak the syscache tuple */
-		ReleaseSysCache(tuple);
-
-		/*
-		 * Lock the role, so nobody can add dependencies to her while we drop
-		 * her.  We keep the lock until the end of transaction.
-		 */
-		LockSharedObject(AuthIdRelationId, roleid, 0, AccessExclusiveLock);
-
 		/*
 		 * If there is a pg_auth_members entry that has one of the roles to be
 		 * dropped as the roleid or member, it should be silently removed, but
@@ -1484,6 +1582,21 @@ RenameRole(const char *oldname, const char *newname)
 	return address;
 }
 
+/*
+ * Before acquiring a role lock for GRANT/REVOKE, check that the current user
+ * has authorization to grant/revoke membership in the specified role.
+ */
+static void
+RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+							 Oid oldroleid, void *callback_arg)
+{
+	bool		is_grant = *((bool *) callback_arg);
+	Oid			currentUserId = GetUserId();
+
+	check_role_membership_authorization(currentUserId, roleid, is_grant);
+}
+
+
 /*
  * GrantRoleStmt
  *
@@ -1568,9 +1681,14 @@ GrantRole(ParseState *pstate, GrantRoleStmt *stmt)
 					(errcode(ERRCODE_INVALID_GRANT_OPERATION),
 					 errmsg("column names cannot be included in GRANT/REVOKE ROLE")));
 
-		roleid = get_role_oid(rolename, false);
-		check_role_membership_authorization(currentUserId,
-											roleid, stmt->is_grant);
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP commits between name resolution
+		 * and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(rolename, ShareUpdateExclusiveLock, false,
+								RoleNameCallbackForGrantRole, &stmt->is_grant);
 		if (stmt->is_grant)
 			AddRoleMems(currentUserId, rolename, roleid,
 						stmt->grantee_roles, grantee_ids,
diff --git a/src/include/commands/user.h b/src/include/commands/user.h
index 97dcb93791b..17263452250 100644
--- a/src/include/commands/user.h
+++ b/src/include/commands/user.h
@@ -21,6 +21,15 @@
 extern PGDLLIMPORT int Password_encryption; /* values from enum PasswordType */
 extern PGDLLIMPORT char *createrole_self_grant;
 
+/* Callback for RoleNameGetOid, invoked after name resolution but before locking */
+typedef void (*RoleNameGetOidCallback) (const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+
+extern Oid	RoleNameGetOid(const char *rolename, LOCKMODE lockmode,
+						   bool missing_ok,
+						   RoleNameGetOidCallback callback,
+						   void *callback_arg);
+
 /* Hook to check passwords in CreateRole() and AlterRole() */
 typedef void (*check_password_hook_type) (const char *username, const char *shadow_pass, PasswordType password_type, Datum validuntil_time, bool validuntil_null);
 
-- 
2.34.1


--xlWYagR37YwJXkF8--





^ permalink  raw  reply  [nested|flat] 173+ messages in thread

* [PATCH v1] Add RoleNameGetOid() with invalidation-based retry loop for DropRole()/GrantRole()
@ 2026-07-04 04:47 Bertrand Drouvot <[email protected]>
  0 siblings, 0 replies; 173+ messages in thread

From: Bertrand Drouvot @ 2026-07-04 04:47 UTC (permalink / raw)

DropRole() and GrantRole() resolve the role name to an OID before acquiring
LockSharedObject() on the role. A concurrent session that commits a DROP ROLE
between the read and the lock acquisition leaves the first session acting on a
stale OID.

This commit fixes the races by using the same approach as RangeVarGetRelidExtended():
It encapsulates name resolution, permission checking (via a caller-supplied callback),
and lock acquisition inside a retry loop driven by SharedInvalidMessageCounter.
If invalidation messages arrive between name resolution and locking, indicating
concurrent DDL, the function retries.

The lock is kept across retries and only released if the name resolves to a
different OID on the next iteration.

Two callbacks are provided:
- RoleNameCallbackForDropRole(): checks current/session user, superuser attribute,
and ADMIN OPTION privilege before locking. This is similar to what DropRole() is
currently doing before LockSharedObject().
- RoleNameCallbackForGrantRole(): calls check_role_membership_authorization() to
verify the current user can grant/revoke membership. This is similar to what GrantRole()
is currently doing before calling AddRoleMems()/DelRoleMems().

DropRole() and GrantRole() now call RoleNameGetOid() with appropriate lock
levels.

AlterRole() does not need the fix because it calls CatalogTupleUpdate() on the
pg_authid tuple before AddRoleMems(), which blocks a concurrent DROP ROLE.

Author: Bertrand Drouvot <[email protected]>
Reviewed-by:
Discussion:
---
 src/backend/commands/user.c | 246 ++++++++++++++++++++++++++----------
 src/include/commands/user.h |   9 ++
 2 files changed, 191 insertions(+), 64 deletions(-)
  95.5% src/backend/commands/
   4.4% src/include/commands/

diff --git a/src/backend/commands/user.c b/src/backend/commands/user.c
index be11c49f919..6d8e2fa8813 100644
--- a/src/backend/commands/user.c
+++ b/src/backend/commands/user.c
@@ -34,6 +34,7 @@
 #include "miscadmin.h"
 #include "port/pg_bitutils.h"
 #include "storage/lmgr.h"
+#include "storage/sinval.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
 #include "utils/catcache.h"
@@ -116,6 +117,10 @@ static void plan_recursive_revoke(CatCList *memlist,
 								  bool revoke_admin_option_only,
 								  DropBehavior behavior);
 static void InitGrantRoleOptions(GrantRoleOptions *popt);
+static void RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+static void RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+										 Oid oldroleid, void *callback_arg);
 
 
 /* Check if current user has createrole privileges */
@@ -126,6 +131,92 @@ have_createrole_privilege(void)
 }
 
 
+/*
+ * RoleNameGetOid
+ *		Given a role name, look up its OID, lock it, and return the OID.
+ *
+ * This follows the same pattern as RangeVarGetRelidExtended():
+ * name resolution, permission check (via callback), and lock acquisition are
+ * performed inside a retry loop. If invalidation messages arrive during the
+ * process (indicating concurrent DDL), we retry to ensure the name still
+ * resolves to the same OID.
+ *
+ * The callback is invoked before locking, giving callers a chance to check
+ * permissions. It receives the current rolename, the resolved OID, the
+ * previous OID (InvalidOid on first iteration), and a caller-supplied arg.
+ * If the callback raises an error, the function aborts without locking.
+ *
+ * If missing_ok is true and the role does not exist, returns InvalidOid.
+ * Otherwise, raises an error.
+ */
+Oid
+RoleNameGetOid(const char *rolename, LOCKMODE lockmode, bool missing_ok,
+			   RoleNameGetOidCallback callback, void *callback_arg)
+{
+	uint64		inval_count;
+	Oid			roleid;
+	Oid			oldroleid = InvalidOid;
+	bool		retry = false;
+
+	for (;;)
+	{
+		/*
+		 * Remember the current invalidation count so we can detect concurrent
+		 * DDL after locking.
+		 */
+		inval_count = SharedInvalidMessageCounter;
+
+		/* Look up the role name */
+		roleid = get_role_oid(rolename, true);
+
+		if (!OidIsValid(roleid))
+		{
+			if (!missing_ok)
+				ereport(ERROR,
+						(errcode(ERRCODE_UNDEFINED_OBJECT),
+						 errmsg("role \"%s\" does not exist", rolename)));
+			return InvalidOid;
+		}
+
+		/*
+		 * Invoke caller-supplied callback before locking. This is a good
+		 * place to check permissions: we haven't taken the lock yet, but we
+		 * know the OID we intend to lock. If concurrent DDL changes things,
+		 * the callback will be invoked again on the next iteration.
+		 */
+		if (callback)
+			callback(rolename, roleid, oldroleid, callback_arg);
+
+		/*
+		 * If upon retry we get back the same OID, the invalidation messages
+		 * did not change the final answer. So we're done.
+		 *
+		 * If we got a different OID, we've locked the role that used to have
+		 * this name rather than the one that does now. Release the old lock.
+		 */
+		if (retry)
+		{
+			if (roleid == oldroleid)
+				break;
+			UnlockSharedObject(AuthIdRelationId, oldroleid, 0, lockmode);
+		}
+
+		/* Lock the role */
+		LockSharedObject(AuthIdRelationId, roleid, 0, lockmode);
+
+		/* If no invalidation messages were processed, we're done */
+		if (inval_count == SharedInvalidMessageCounter)
+			break;
+
+		/* Something may have changed, retry */
+		retry = true;
+		oldroleid = roleid;
+	}
+
+	return roleid;
+}
+
+
 /*
  * CREATE ROLE
  */
@@ -1090,6 +1181,59 @@ AlterRoleSet(AlterRoleSetStmt *stmt)
 }
 
 
+/*
+ * Before acquiring a role lock for DROP ROLE, check that the role is not the
+ * current/session user and that the caller has sufficient privileges to drop it.
+ */
+static void
+RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+							Oid oldroleid, void *callback_arg)
+{
+	HeapTuple	tuple;
+	Form_pg_authid roleform;
+
+	if (roleid == GetUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetOuterUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetSessionUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("session user cannot be dropped")));
+
+	tuple = SearchSysCache1(AUTHOID, ObjectIdGetDatum(roleid));
+	if (!HeapTupleIsValid(tuple))
+		ereport(ERROR,
+				(errcode(ERRCODE_UNDEFINED_OBJECT),
+				 errmsg("role \"%s\" does not exist", rolename)));
+
+	roleform = (Form_pg_authid) GETSTRUCT(tuple);
+
+	/*
+	 * For safety's sake, we allow createrole holders to drop ordinary roles
+	 * but not superuser roles, and only if they also have ADMIN OPTION.
+	 */
+	if (roleform->rolsuper && !superuser())
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
+						   "SUPERUSER", "SUPERUSER")));
+	if (!is_admin_of_role(GetUserId(), roleid))
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
+						   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
+
+	ReleaseSysCache(tuple);
+}
+
+
 /*
  * DROP ROLE
  */
@@ -1119,9 +1263,7 @@ DropRole(DropRoleStmt *stmt)
 	{
 		RoleSpec   *rolspec = lfirst(item);
 		char	   *role;
-		HeapTuple	tuple,
-					tmp_tuple;
-		Form_pg_authid roleform;
+		HeapTuple	tmp_tuple;
 		ScanKeyData scankey;
 		SysScanDesc sscan;
 		Oid			roleid;
@@ -1132,71 +1274,27 @@ DropRole(DropRoleStmt *stmt)
 					 errmsg("cannot use special role specifier in DROP ROLE")));
 		role = rolspec->rolename;
 
-		tuple = SearchSysCache1(AUTHNAME, PointerGetDatum(role));
-		if (!HeapTupleIsValid(tuple))
-		{
-			if (!stmt->missing_ok)
-			{
-				ereport(ERROR,
-						(errcode(ERRCODE_UNDEFINED_OBJECT),
-						 errmsg("role \"%s\" does not exist", role)));
-			}
-			else
-			{
-				ereport(NOTICE,
-						(errmsg("role \"%s\" does not exist, skipping",
-								role)));
-			}
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP or ALTER commits between name
+		 * resolution and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(role, AccessExclusiveLock, stmt->missing_ok,
+								RoleNameCallbackForDropRole, NULL);
 
+		if (!OidIsValid(roleid))
+		{
+			/* missing_ok case: role doesn't exist */
+			ereport(NOTICE,
+					(errmsg("role \"%s\" does not exist, skipping",
+							role)));
 			continue;
 		}
 
-		roleform = (Form_pg_authid) GETSTRUCT(tuple);
-		roleid = roleform->oid;
-
-		if (roleid == GetUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetOuterUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetSessionUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("session user cannot be dropped")));
-
-		/*
-		 * For safety's sake, we allow createrole holders to drop ordinary
-		 * roles but not superuser roles, and only if they also have ADMIN
-		 * OPTION.
-		 */
-		if (roleform->rolsuper && !superuser())
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
-							   "SUPERUSER", "SUPERUSER")));
-		if (!is_admin_of_role(GetUserId(), roleid))
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
-							   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
-
 		/* DROP hook for the role being removed */
 		InvokeObjectDropHook(AuthIdRelationId, roleid, 0);
 
-		/* Don't leak the syscache tuple */
-		ReleaseSysCache(tuple);
-
-		/*
-		 * Lock the role, so nobody can add dependencies to her while we drop
-		 * her.  We keep the lock until the end of transaction.
-		 */
-		LockSharedObject(AuthIdRelationId, roleid, 0, AccessExclusiveLock);
-
 		/*
 		 * If there is a pg_auth_members entry that has one of the roles to be
 		 * dropped as the roleid or member, it should be silently removed, but
@@ -1484,6 +1582,21 @@ RenameRole(const char *oldname, const char *newname)
 	return address;
 }
 
+/*
+ * Before acquiring a role lock for GRANT/REVOKE, check that the current user
+ * has authorization to grant/revoke membership in the specified role.
+ */
+static void
+RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+							 Oid oldroleid, void *callback_arg)
+{
+	bool		is_grant = *((bool *) callback_arg);
+	Oid			currentUserId = GetUserId();
+
+	check_role_membership_authorization(currentUserId, roleid, is_grant);
+}
+
+
 /*
  * GrantRoleStmt
  *
@@ -1568,9 +1681,14 @@ GrantRole(ParseState *pstate, GrantRoleStmt *stmt)
 					(errcode(ERRCODE_INVALID_GRANT_OPERATION),
 					 errmsg("column names cannot be included in GRANT/REVOKE ROLE")));
 
-		roleid = get_role_oid(rolename, false);
-		check_role_membership_authorization(currentUserId,
-											roleid, stmt->is_grant);
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP commits between name resolution
+		 * and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(rolename, ShareUpdateExclusiveLock, false,
+								RoleNameCallbackForGrantRole, &stmt->is_grant);
 		if (stmt->is_grant)
 			AddRoleMems(currentUserId, rolename, roleid,
 						stmt->grantee_roles, grantee_ids,
diff --git a/src/include/commands/user.h b/src/include/commands/user.h
index 97dcb93791b..17263452250 100644
--- a/src/include/commands/user.h
+++ b/src/include/commands/user.h
@@ -21,6 +21,15 @@
 extern PGDLLIMPORT int Password_encryption; /* values from enum PasswordType */
 extern PGDLLIMPORT char *createrole_self_grant;
 
+/* Callback for RoleNameGetOid, invoked after name resolution but before locking */
+typedef void (*RoleNameGetOidCallback) (const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+
+extern Oid	RoleNameGetOid(const char *rolename, LOCKMODE lockmode,
+						   bool missing_ok,
+						   RoleNameGetOidCallback callback,
+						   void *callback_arg);
+
 /* Hook to check passwords in CreateRole() and AlterRole() */
 typedef void (*check_password_hook_type) (const char *username, const char *shadow_pass, PasswordType password_type, Datum validuntil_time, bool validuntil_null);
 
-- 
2.34.1


--xlWYagR37YwJXkF8--





^ permalink  raw  reply  [nested|flat] 173+ messages in thread

* [PATCH v1] Add RoleNameGetOid() with invalidation-based retry loop for DropRole()/GrantRole()
@ 2026-07-04 04:47 Bertrand Drouvot <[email protected]>
  0 siblings, 0 replies; 173+ messages in thread

From: Bertrand Drouvot @ 2026-07-04 04:47 UTC (permalink / raw)

DropRole() and GrantRole() resolve the role name to an OID before acquiring
LockSharedObject() on the role. A concurrent session that commits a DROP ROLE
between the read and the lock acquisition leaves the first session acting on a
stale OID.

This commit fixes the races by using the same approach as RangeVarGetRelidExtended():
It encapsulates name resolution, permission checking (via a caller-supplied callback),
and lock acquisition inside a retry loop driven by SharedInvalidMessageCounter.
If invalidation messages arrive between name resolution and locking, indicating
concurrent DDL, the function retries.

The lock is kept across retries and only released if the name resolves to a
different OID on the next iteration.

Two callbacks are provided:
- RoleNameCallbackForDropRole(): checks current/session user, superuser attribute,
and ADMIN OPTION privilege before locking. This is similar to what DropRole() is
currently doing before LockSharedObject().
- RoleNameCallbackForGrantRole(): calls check_role_membership_authorization() to
verify the current user can grant/revoke membership. This is similar to what GrantRole()
is currently doing before calling AddRoleMems()/DelRoleMems().

DropRole() and GrantRole() now call RoleNameGetOid() with appropriate lock
levels.

AlterRole() does not need the fix because it calls CatalogTupleUpdate() on the
pg_authid tuple before AddRoleMems(), which blocks a concurrent DROP ROLE.

Author: Bertrand Drouvot <[email protected]>
Reviewed-by:
Discussion:
---
 src/backend/commands/user.c | 246 ++++++++++++++++++++++++++----------
 src/include/commands/user.h |   9 ++
 2 files changed, 191 insertions(+), 64 deletions(-)
  95.5% src/backend/commands/
   4.4% src/include/commands/

diff --git a/src/backend/commands/user.c b/src/backend/commands/user.c
index be11c49f919..6d8e2fa8813 100644
--- a/src/backend/commands/user.c
+++ b/src/backend/commands/user.c
@@ -34,6 +34,7 @@
 #include "miscadmin.h"
 #include "port/pg_bitutils.h"
 #include "storage/lmgr.h"
+#include "storage/sinval.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
 #include "utils/catcache.h"
@@ -116,6 +117,10 @@ static void plan_recursive_revoke(CatCList *memlist,
 								  bool revoke_admin_option_only,
 								  DropBehavior behavior);
 static void InitGrantRoleOptions(GrantRoleOptions *popt);
+static void RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+static void RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+										 Oid oldroleid, void *callback_arg);
 
 
 /* Check if current user has createrole privileges */
@@ -126,6 +131,92 @@ have_createrole_privilege(void)
 }
 
 
+/*
+ * RoleNameGetOid
+ *		Given a role name, look up its OID, lock it, and return the OID.
+ *
+ * This follows the same pattern as RangeVarGetRelidExtended():
+ * name resolution, permission check (via callback), and lock acquisition are
+ * performed inside a retry loop. If invalidation messages arrive during the
+ * process (indicating concurrent DDL), we retry to ensure the name still
+ * resolves to the same OID.
+ *
+ * The callback is invoked before locking, giving callers a chance to check
+ * permissions. It receives the current rolename, the resolved OID, the
+ * previous OID (InvalidOid on first iteration), and a caller-supplied arg.
+ * If the callback raises an error, the function aborts without locking.
+ *
+ * If missing_ok is true and the role does not exist, returns InvalidOid.
+ * Otherwise, raises an error.
+ */
+Oid
+RoleNameGetOid(const char *rolename, LOCKMODE lockmode, bool missing_ok,
+			   RoleNameGetOidCallback callback, void *callback_arg)
+{
+	uint64		inval_count;
+	Oid			roleid;
+	Oid			oldroleid = InvalidOid;
+	bool		retry = false;
+
+	for (;;)
+	{
+		/*
+		 * Remember the current invalidation count so we can detect concurrent
+		 * DDL after locking.
+		 */
+		inval_count = SharedInvalidMessageCounter;
+
+		/* Look up the role name */
+		roleid = get_role_oid(rolename, true);
+
+		if (!OidIsValid(roleid))
+		{
+			if (!missing_ok)
+				ereport(ERROR,
+						(errcode(ERRCODE_UNDEFINED_OBJECT),
+						 errmsg("role \"%s\" does not exist", rolename)));
+			return InvalidOid;
+		}
+
+		/*
+		 * Invoke caller-supplied callback before locking. This is a good
+		 * place to check permissions: we haven't taken the lock yet, but we
+		 * know the OID we intend to lock. If concurrent DDL changes things,
+		 * the callback will be invoked again on the next iteration.
+		 */
+		if (callback)
+			callback(rolename, roleid, oldroleid, callback_arg);
+
+		/*
+		 * If upon retry we get back the same OID, the invalidation messages
+		 * did not change the final answer. So we're done.
+		 *
+		 * If we got a different OID, we've locked the role that used to have
+		 * this name rather than the one that does now. Release the old lock.
+		 */
+		if (retry)
+		{
+			if (roleid == oldroleid)
+				break;
+			UnlockSharedObject(AuthIdRelationId, oldroleid, 0, lockmode);
+		}
+
+		/* Lock the role */
+		LockSharedObject(AuthIdRelationId, roleid, 0, lockmode);
+
+		/* If no invalidation messages were processed, we're done */
+		if (inval_count == SharedInvalidMessageCounter)
+			break;
+
+		/* Something may have changed, retry */
+		retry = true;
+		oldroleid = roleid;
+	}
+
+	return roleid;
+}
+
+
 /*
  * CREATE ROLE
  */
@@ -1090,6 +1181,59 @@ AlterRoleSet(AlterRoleSetStmt *stmt)
 }
 
 
+/*
+ * Before acquiring a role lock for DROP ROLE, check that the role is not the
+ * current/session user and that the caller has sufficient privileges to drop it.
+ */
+static void
+RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+							Oid oldroleid, void *callback_arg)
+{
+	HeapTuple	tuple;
+	Form_pg_authid roleform;
+
+	if (roleid == GetUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetOuterUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetSessionUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("session user cannot be dropped")));
+
+	tuple = SearchSysCache1(AUTHOID, ObjectIdGetDatum(roleid));
+	if (!HeapTupleIsValid(tuple))
+		ereport(ERROR,
+				(errcode(ERRCODE_UNDEFINED_OBJECT),
+				 errmsg("role \"%s\" does not exist", rolename)));
+
+	roleform = (Form_pg_authid) GETSTRUCT(tuple);
+
+	/*
+	 * For safety's sake, we allow createrole holders to drop ordinary roles
+	 * but not superuser roles, and only if they also have ADMIN OPTION.
+	 */
+	if (roleform->rolsuper && !superuser())
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
+						   "SUPERUSER", "SUPERUSER")));
+	if (!is_admin_of_role(GetUserId(), roleid))
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
+						   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
+
+	ReleaseSysCache(tuple);
+}
+
+
 /*
  * DROP ROLE
  */
@@ -1119,9 +1263,7 @@ DropRole(DropRoleStmt *stmt)
 	{
 		RoleSpec   *rolspec = lfirst(item);
 		char	   *role;
-		HeapTuple	tuple,
-					tmp_tuple;
-		Form_pg_authid roleform;
+		HeapTuple	tmp_tuple;
 		ScanKeyData scankey;
 		SysScanDesc sscan;
 		Oid			roleid;
@@ -1132,71 +1274,27 @@ DropRole(DropRoleStmt *stmt)
 					 errmsg("cannot use special role specifier in DROP ROLE")));
 		role = rolspec->rolename;
 
-		tuple = SearchSysCache1(AUTHNAME, PointerGetDatum(role));
-		if (!HeapTupleIsValid(tuple))
-		{
-			if (!stmt->missing_ok)
-			{
-				ereport(ERROR,
-						(errcode(ERRCODE_UNDEFINED_OBJECT),
-						 errmsg("role \"%s\" does not exist", role)));
-			}
-			else
-			{
-				ereport(NOTICE,
-						(errmsg("role \"%s\" does not exist, skipping",
-								role)));
-			}
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP or ALTER commits between name
+		 * resolution and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(role, AccessExclusiveLock, stmt->missing_ok,
+								RoleNameCallbackForDropRole, NULL);
 
+		if (!OidIsValid(roleid))
+		{
+			/* missing_ok case: role doesn't exist */
+			ereport(NOTICE,
+					(errmsg("role \"%s\" does not exist, skipping",
+							role)));
 			continue;
 		}
 
-		roleform = (Form_pg_authid) GETSTRUCT(tuple);
-		roleid = roleform->oid;
-
-		if (roleid == GetUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetOuterUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetSessionUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("session user cannot be dropped")));
-
-		/*
-		 * For safety's sake, we allow createrole holders to drop ordinary
-		 * roles but not superuser roles, and only if they also have ADMIN
-		 * OPTION.
-		 */
-		if (roleform->rolsuper && !superuser())
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
-							   "SUPERUSER", "SUPERUSER")));
-		if (!is_admin_of_role(GetUserId(), roleid))
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
-							   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
-
 		/* DROP hook for the role being removed */
 		InvokeObjectDropHook(AuthIdRelationId, roleid, 0);
 
-		/* Don't leak the syscache tuple */
-		ReleaseSysCache(tuple);
-
-		/*
-		 * Lock the role, so nobody can add dependencies to her while we drop
-		 * her.  We keep the lock until the end of transaction.
-		 */
-		LockSharedObject(AuthIdRelationId, roleid, 0, AccessExclusiveLock);
-
 		/*
 		 * If there is a pg_auth_members entry that has one of the roles to be
 		 * dropped as the roleid or member, it should be silently removed, but
@@ -1484,6 +1582,21 @@ RenameRole(const char *oldname, const char *newname)
 	return address;
 }
 
+/*
+ * Before acquiring a role lock for GRANT/REVOKE, check that the current user
+ * has authorization to grant/revoke membership in the specified role.
+ */
+static void
+RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+							 Oid oldroleid, void *callback_arg)
+{
+	bool		is_grant = *((bool *) callback_arg);
+	Oid			currentUserId = GetUserId();
+
+	check_role_membership_authorization(currentUserId, roleid, is_grant);
+}
+
+
 /*
  * GrantRoleStmt
  *
@@ -1568,9 +1681,14 @@ GrantRole(ParseState *pstate, GrantRoleStmt *stmt)
 					(errcode(ERRCODE_INVALID_GRANT_OPERATION),
 					 errmsg("column names cannot be included in GRANT/REVOKE ROLE")));
 
-		roleid = get_role_oid(rolename, false);
-		check_role_membership_authorization(currentUserId,
-											roleid, stmt->is_grant);
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP commits between name resolution
+		 * and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(rolename, ShareUpdateExclusiveLock, false,
+								RoleNameCallbackForGrantRole, &stmt->is_grant);
 		if (stmt->is_grant)
 			AddRoleMems(currentUserId, rolename, roleid,
 						stmt->grantee_roles, grantee_ids,
diff --git a/src/include/commands/user.h b/src/include/commands/user.h
index 97dcb93791b..17263452250 100644
--- a/src/include/commands/user.h
+++ b/src/include/commands/user.h
@@ -21,6 +21,15 @@
 extern PGDLLIMPORT int Password_encryption; /* values from enum PasswordType */
 extern PGDLLIMPORT char *createrole_self_grant;
 
+/* Callback for RoleNameGetOid, invoked after name resolution but before locking */
+typedef void (*RoleNameGetOidCallback) (const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+
+extern Oid	RoleNameGetOid(const char *rolename, LOCKMODE lockmode,
+						   bool missing_ok,
+						   RoleNameGetOidCallback callback,
+						   void *callback_arg);
+
 /* Hook to check passwords in CreateRole() and AlterRole() */
 typedef void (*check_password_hook_type) (const char *username, const char *shadow_pass, PasswordType password_type, Datum validuntil_time, bool validuntil_null);
 
-- 
2.34.1


--xlWYagR37YwJXkF8--





^ permalink  raw  reply  [nested|flat] 173+ messages in thread

* [PATCH v1] Add RoleNameGetOid() with invalidation-based retry loop for DropRole()/GrantRole()
@ 2026-07-04 04:47 Bertrand Drouvot <[email protected]>
  0 siblings, 0 replies; 173+ messages in thread

From: Bertrand Drouvot @ 2026-07-04 04:47 UTC (permalink / raw)

DropRole() and GrantRole() resolve the role name to an OID before acquiring
LockSharedObject() on the role. A concurrent session that commits a DROP ROLE
between the read and the lock acquisition leaves the first session acting on a
stale OID.

This commit fixes the races by using the same approach as RangeVarGetRelidExtended():
It encapsulates name resolution, permission checking (via a caller-supplied callback),
and lock acquisition inside a retry loop driven by SharedInvalidMessageCounter.
If invalidation messages arrive between name resolution and locking, indicating
concurrent DDL, the function retries.

The lock is kept across retries and only released if the name resolves to a
different OID on the next iteration.

Two callbacks are provided:
- RoleNameCallbackForDropRole(): checks current/session user, superuser attribute,
and ADMIN OPTION privilege before locking. This is similar to what DropRole() is
currently doing before LockSharedObject().
- RoleNameCallbackForGrantRole(): calls check_role_membership_authorization() to
verify the current user can grant/revoke membership. This is similar to what GrantRole()
is currently doing before calling AddRoleMems()/DelRoleMems().

DropRole() and GrantRole() now call RoleNameGetOid() with appropriate lock
levels.

AlterRole() does not need the fix because it calls CatalogTupleUpdate() on the
pg_authid tuple before AddRoleMems(), which blocks a concurrent DROP ROLE.

Author: Bertrand Drouvot <[email protected]>
Reviewed-by:
Discussion:
---
 src/backend/commands/user.c | 246 ++++++++++++++++++++++++++----------
 src/include/commands/user.h |   9 ++
 2 files changed, 191 insertions(+), 64 deletions(-)
  95.5% src/backend/commands/
   4.4% src/include/commands/

diff --git a/src/backend/commands/user.c b/src/backend/commands/user.c
index be11c49f919..6d8e2fa8813 100644
--- a/src/backend/commands/user.c
+++ b/src/backend/commands/user.c
@@ -34,6 +34,7 @@
 #include "miscadmin.h"
 #include "port/pg_bitutils.h"
 #include "storage/lmgr.h"
+#include "storage/sinval.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
 #include "utils/catcache.h"
@@ -116,6 +117,10 @@ static void plan_recursive_revoke(CatCList *memlist,
 								  bool revoke_admin_option_only,
 								  DropBehavior behavior);
 static void InitGrantRoleOptions(GrantRoleOptions *popt);
+static void RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+static void RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+										 Oid oldroleid, void *callback_arg);
 
 
 /* Check if current user has createrole privileges */
@@ -126,6 +131,92 @@ have_createrole_privilege(void)
 }
 
 
+/*
+ * RoleNameGetOid
+ *		Given a role name, look up its OID, lock it, and return the OID.
+ *
+ * This follows the same pattern as RangeVarGetRelidExtended():
+ * name resolution, permission check (via callback), and lock acquisition are
+ * performed inside a retry loop. If invalidation messages arrive during the
+ * process (indicating concurrent DDL), we retry to ensure the name still
+ * resolves to the same OID.
+ *
+ * The callback is invoked before locking, giving callers a chance to check
+ * permissions. It receives the current rolename, the resolved OID, the
+ * previous OID (InvalidOid on first iteration), and a caller-supplied arg.
+ * If the callback raises an error, the function aborts without locking.
+ *
+ * If missing_ok is true and the role does not exist, returns InvalidOid.
+ * Otherwise, raises an error.
+ */
+Oid
+RoleNameGetOid(const char *rolename, LOCKMODE lockmode, bool missing_ok,
+			   RoleNameGetOidCallback callback, void *callback_arg)
+{
+	uint64		inval_count;
+	Oid			roleid;
+	Oid			oldroleid = InvalidOid;
+	bool		retry = false;
+
+	for (;;)
+	{
+		/*
+		 * Remember the current invalidation count so we can detect concurrent
+		 * DDL after locking.
+		 */
+		inval_count = SharedInvalidMessageCounter;
+
+		/* Look up the role name */
+		roleid = get_role_oid(rolename, true);
+
+		if (!OidIsValid(roleid))
+		{
+			if (!missing_ok)
+				ereport(ERROR,
+						(errcode(ERRCODE_UNDEFINED_OBJECT),
+						 errmsg("role \"%s\" does not exist", rolename)));
+			return InvalidOid;
+		}
+
+		/*
+		 * Invoke caller-supplied callback before locking. This is a good
+		 * place to check permissions: we haven't taken the lock yet, but we
+		 * know the OID we intend to lock. If concurrent DDL changes things,
+		 * the callback will be invoked again on the next iteration.
+		 */
+		if (callback)
+			callback(rolename, roleid, oldroleid, callback_arg);
+
+		/*
+		 * If upon retry we get back the same OID, the invalidation messages
+		 * did not change the final answer. So we're done.
+		 *
+		 * If we got a different OID, we've locked the role that used to have
+		 * this name rather than the one that does now. Release the old lock.
+		 */
+		if (retry)
+		{
+			if (roleid == oldroleid)
+				break;
+			UnlockSharedObject(AuthIdRelationId, oldroleid, 0, lockmode);
+		}
+
+		/* Lock the role */
+		LockSharedObject(AuthIdRelationId, roleid, 0, lockmode);
+
+		/* If no invalidation messages were processed, we're done */
+		if (inval_count == SharedInvalidMessageCounter)
+			break;
+
+		/* Something may have changed, retry */
+		retry = true;
+		oldroleid = roleid;
+	}
+
+	return roleid;
+}
+
+
 /*
  * CREATE ROLE
  */
@@ -1090,6 +1181,59 @@ AlterRoleSet(AlterRoleSetStmt *stmt)
 }
 
 
+/*
+ * Before acquiring a role lock for DROP ROLE, check that the role is not the
+ * current/session user and that the caller has sufficient privileges to drop it.
+ */
+static void
+RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+							Oid oldroleid, void *callback_arg)
+{
+	HeapTuple	tuple;
+	Form_pg_authid roleform;
+
+	if (roleid == GetUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetOuterUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetSessionUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("session user cannot be dropped")));
+
+	tuple = SearchSysCache1(AUTHOID, ObjectIdGetDatum(roleid));
+	if (!HeapTupleIsValid(tuple))
+		ereport(ERROR,
+				(errcode(ERRCODE_UNDEFINED_OBJECT),
+				 errmsg("role \"%s\" does not exist", rolename)));
+
+	roleform = (Form_pg_authid) GETSTRUCT(tuple);
+
+	/*
+	 * For safety's sake, we allow createrole holders to drop ordinary roles
+	 * but not superuser roles, and only if they also have ADMIN OPTION.
+	 */
+	if (roleform->rolsuper && !superuser())
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
+						   "SUPERUSER", "SUPERUSER")));
+	if (!is_admin_of_role(GetUserId(), roleid))
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
+						   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
+
+	ReleaseSysCache(tuple);
+}
+
+
 /*
  * DROP ROLE
  */
@@ -1119,9 +1263,7 @@ DropRole(DropRoleStmt *stmt)
 	{
 		RoleSpec   *rolspec = lfirst(item);
 		char	   *role;
-		HeapTuple	tuple,
-					tmp_tuple;
-		Form_pg_authid roleform;
+		HeapTuple	tmp_tuple;
 		ScanKeyData scankey;
 		SysScanDesc sscan;
 		Oid			roleid;
@@ -1132,71 +1274,27 @@ DropRole(DropRoleStmt *stmt)
 					 errmsg("cannot use special role specifier in DROP ROLE")));
 		role = rolspec->rolename;
 
-		tuple = SearchSysCache1(AUTHNAME, PointerGetDatum(role));
-		if (!HeapTupleIsValid(tuple))
-		{
-			if (!stmt->missing_ok)
-			{
-				ereport(ERROR,
-						(errcode(ERRCODE_UNDEFINED_OBJECT),
-						 errmsg("role \"%s\" does not exist", role)));
-			}
-			else
-			{
-				ereport(NOTICE,
-						(errmsg("role \"%s\" does not exist, skipping",
-								role)));
-			}
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP or ALTER commits between name
+		 * resolution and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(role, AccessExclusiveLock, stmt->missing_ok,
+								RoleNameCallbackForDropRole, NULL);
 
+		if (!OidIsValid(roleid))
+		{
+			/* missing_ok case: role doesn't exist */
+			ereport(NOTICE,
+					(errmsg("role \"%s\" does not exist, skipping",
+							role)));
 			continue;
 		}
 
-		roleform = (Form_pg_authid) GETSTRUCT(tuple);
-		roleid = roleform->oid;
-
-		if (roleid == GetUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetOuterUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetSessionUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("session user cannot be dropped")));
-
-		/*
-		 * For safety's sake, we allow createrole holders to drop ordinary
-		 * roles but not superuser roles, and only if they also have ADMIN
-		 * OPTION.
-		 */
-		if (roleform->rolsuper && !superuser())
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
-							   "SUPERUSER", "SUPERUSER")));
-		if (!is_admin_of_role(GetUserId(), roleid))
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
-							   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
-
 		/* DROP hook for the role being removed */
 		InvokeObjectDropHook(AuthIdRelationId, roleid, 0);
 
-		/* Don't leak the syscache tuple */
-		ReleaseSysCache(tuple);
-
-		/*
-		 * Lock the role, so nobody can add dependencies to her while we drop
-		 * her.  We keep the lock until the end of transaction.
-		 */
-		LockSharedObject(AuthIdRelationId, roleid, 0, AccessExclusiveLock);
-
 		/*
 		 * If there is a pg_auth_members entry that has one of the roles to be
 		 * dropped as the roleid or member, it should be silently removed, but
@@ -1484,6 +1582,21 @@ RenameRole(const char *oldname, const char *newname)
 	return address;
 }
 
+/*
+ * Before acquiring a role lock for GRANT/REVOKE, check that the current user
+ * has authorization to grant/revoke membership in the specified role.
+ */
+static void
+RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+							 Oid oldroleid, void *callback_arg)
+{
+	bool		is_grant = *((bool *) callback_arg);
+	Oid			currentUserId = GetUserId();
+
+	check_role_membership_authorization(currentUserId, roleid, is_grant);
+}
+
+
 /*
  * GrantRoleStmt
  *
@@ -1568,9 +1681,14 @@ GrantRole(ParseState *pstate, GrantRoleStmt *stmt)
 					(errcode(ERRCODE_INVALID_GRANT_OPERATION),
 					 errmsg("column names cannot be included in GRANT/REVOKE ROLE")));
 
-		roleid = get_role_oid(rolename, false);
-		check_role_membership_authorization(currentUserId,
-											roleid, stmt->is_grant);
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP commits between name resolution
+		 * and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(rolename, ShareUpdateExclusiveLock, false,
+								RoleNameCallbackForGrantRole, &stmt->is_grant);
 		if (stmt->is_grant)
 			AddRoleMems(currentUserId, rolename, roleid,
 						stmt->grantee_roles, grantee_ids,
diff --git a/src/include/commands/user.h b/src/include/commands/user.h
index 97dcb93791b..17263452250 100644
--- a/src/include/commands/user.h
+++ b/src/include/commands/user.h
@@ -21,6 +21,15 @@
 extern PGDLLIMPORT int Password_encryption; /* values from enum PasswordType */
 extern PGDLLIMPORT char *createrole_self_grant;
 
+/* Callback for RoleNameGetOid, invoked after name resolution but before locking */
+typedef void (*RoleNameGetOidCallback) (const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+
+extern Oid	RoleNameGetOid(const char *rolename, LOCKMODE lockmode,
+						   bool missing_ok,
+						   RoleNameGetOidCallback callback,
+						   void *callback_arg);
+
 /* Hook to check passwords in CreateRole() and AlterRole() */
 typedef void (*check_password_hook_type) (const char *username, const char *shadow_pass, PasswordType password_type, Datum validuntil_time, bool validuntil_null);
 
-- 
2.34.1


--xlWYagR37YwJXkF8--





^ permalink  raw  reply  [nested|flat] 173+ messages in thread

* [PATCH v1] Add RoleNameGetOid() with invalidation-based retry loop for DropRole()/GrantRole()
@ 2026-07-04 04:47 Bertrand Drouvot <[email protected]>
  0 siblings, 0 replies; 173+ messages in thread

From: Bertrand Drouvot @ 2026-07-04 04:47 UTC (permalink / raw)

DropRole() and GrantRole() resolve the role name to an OID before acquiring
LockSharedObject() on the role. A concurrent session that commits a DROP ROLE
between the read and the lock acquisition leaves the first session acting on a
stale OID.

This commit fixes the races by using the same approach as RangeVarGetRelidExtended():
It encapsulates name resolution, permission checking (via a caller-supplied callback),
and lock acquisition inside a retry loop driven by SharedInvalidMessageCounter.
If invalidation messages arrive between name resolution and locking, indicating
concurrent DDL, the function retries.

The lock is kept across retries and only released if the name resolves to a
different OID on the next iteration.

Two callbacks are provided:
- RoleNameCallbackForDropRole(): checks current/session user, superuser attribute,
and ADMIN OPTION privilege before locking. This is similar to what DropRole() is
currently doing before LockSharedObject().
- RoleNameCallbackForGrantRole(): calls check_role_membership_authorization() to
verify the current user can grant/revoke membership. This is similar to what GrantRole()
is currently doing before calling AddRoleMems()/DelRoleMems().

DropRole() and GrantRole() now call RoleNameGetOid() with appropriate lock
levels.

AlterRole() does not need the fix because it calls CatalogTupleUpdate() on the
pg_authid tuple before AddRoleMems(), which blocks a concurrent DROP ROLE.

Author: Bertrand Drouvot <[email protected]>
Reviewed-by:
Discussion:
---
 src/backend/commands/user.c | 246 ++++++++++++++++++++++++++----------
 src/include/commands/user.h |   9 ++
 2 files changed, 191 insertions(+), 64 deletions(-)
  95.5% src/backend/commands/
   4.4% src/include/commands/

diff --git a/src/backend/commands/user.c b/src/backend/commands/user.c
index be11c49f919..6d8e2fa8813 100644
--- a/src/backend/commands/user.c
+++ b/src/backend/commands/user.c
@@ -34,6 +34,7 @@
 #include "miscadmin.h"
 #include "port/pg_bitutils.h"
 #include "storage/lmgr.h"
+#include "storage/sinval.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
 #include "utils/catcache.h"
@@ -116,6 +117,10 @@ static void plan_recursive_revoke(CatCList *memlist,
 								  bool revoke_admin_option_only,
 								  DropBehavior behavior);
 static void InitGrantRoleOptions(GrantRoleOptions *popt);
+static void RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+static void RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+										 Oid oldroleid, void *callback_arg);
 
 
 /* Check if current user has createrole privileges */
@@ -126,6 +131,92 @@ have_createrole_privilege(void)
 }
 
 
+/*
+ * RoleNameGetOid
+ *		Given a role name, look up its OID, lock it, and return the OID.
+ *
+ * This follows the same pattern as RangeVarGetRelidExtended():
+ * name resolution, permission check (via callback), and lock acquisition are
+ * performed inside a retry loop. If invalidation messages arrive during the
+ * process (indicating concurrent DDL), we retry to ensure the name still
+ * resolves to the same OID.
+ *
+ * The callback is invoked before locking, giving callers a chance to check
+ * permissions. It receives the current rolename, the resolved OID, the
+ * previous OID (InvalidOid on first iteration), and a caller-supplied arg.
+ * If the callback raises an error, the function aborts without locking.
+ *
+ * If missing_ok is true and the role does not exist, returns InvalidOid.
+ * Otherwise, raises an error.
+ */
+Oid
+RoleNameGetOid(const char *rolename, LOCKMODE lockmode, bool missing_ok,
+			   RoleNameGetOidCallback callback, void *callback_arg)
+{
+	uint64		inval_count;
+	Oid			roleid;
+	Oid			oldroleid = InvalidOid;
+	bool		retry = false;
+
+	for (;;)
+	{
+		/*
+		 * Remember the current invalidation count so we can detect concurrent
+		 * DDL after locking.
+		 */
+		inval_count = SharedInvalidMessageCounter;
+
+		/* Look up the role name */
+		roleid = get_role_oid(rolename, true);
+
+		if (!OidIsValid(roleid))
+		{
+			if (!missing_ok)
+				ereport(ERROR,
+						(errcode(ERRCODE_UNDEFINED_OBJECT),
+						 errmsg("role \"%s\" does not exist", rolename)));
+			return InvalidOid;
+		}
+
+		/*
+		 * Invoke caller-supplied callback before locking. This is a good
+		 * place to check permissions: we haven't taken the lock yet, but we
+		 * know the OID we intend to lock. If concurrent DDL changes things,
+		 * the callback will be invoked again on the next iteration.
+		 */
+		if (callback)
+			callback(rolename, roleid, oldroleid, callback_arg);
+
+		/*
+		 * If upon retry we get back the same OID, the invalidation messages
+		 * did not change the final answer. So we're done.
+		 *
+		 * If we got a different OID, we've locked the role that used to have
+		 * this name rather than the one that does now. Release the old lock.
+		 */
+		if (retry)
+		{
+			if (roleid == oldroleid)
+				break;
+			UnlockSharedObject(AuthIdRelationId, oldroleid, 0, lockmode);
+		}
+
+		/* Lock the role */
+		LockSharedObject(AuthIdRelationId, roleid, 0, lockmode);
+
+		/* If no invalidation messages were processed, we're done */
+		if (inval_count == SharedInvalidMessageCounter)
+			break;
+
+		/* Something may have changed, retry */
+		retry = true;
+		oldroleid = roleid;
+	}
+
+	return roleid;
+}
+
+
 /*
  * CREATE ROLE
  */
@@ -1090,6 +1181,59 @@ AlterRoleSet(AlterRoleSetStmt *stmt)
 }
 
 
+/*
+ * Before acquiring a role lock for DROP ROLE, check that the role is not the
+ * current/session user and that the caller has sufficient privileges to drop it.
+ */
+static void
+RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+							Oid oldroleid, void *callback_arg)
+{
+	HeapTuple	tuple;
+	Form_pg_authid roleform;
+
+	if (roleid == GetUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetOuterUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetSessionUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("session user cannot be dropped")));
+
+	tuple = SearchSysCache1(AUTHOID, ObjectIdGetDatum(roleid));
+	if (!HeapTupleIsValid(tuple))
+		ereport(ERROR,
+				(errcode(ERRCODE_UNDEFINED_OBJECT),
+				 errmsg("role \"%s\" does not exist", rolename)));
+
+	roleform = (Form_pg_authid) GETSTRUCT(tuple);
+
+	/*
+	 * For safety's sake, we allow createrole holders to drop ordinary roles
+	 * but not superuser roles, and only if they also have ADMIN OPTION.
+	 */
+	if (roleform->rolsuper && !superuser())
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
+						   "SUPERUSER", "SUPERUSER")));
+	if (!is_admin_of_role(GetUserId(), roleid))
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
+						   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
+
+	ReleaseSysCache(tuple);
+}
+
+
 /*
  * DROP ROLE
  */
@@ -1119,9 +1263,7 @@ DropRole(DropRoleStmt *stmt)
 	{
 		RoleSpec   *rolspec = lfirst(item);
 		char	   *role;
-		HeapTuple	tuple,
-					tmp_tuple;
-		Form_pg_authid roleform;
+		HeapTuple	tmp_tuple;
 		ScanKeyData scankey;
 		SysScanDesc sscan;
 		Oid			roleid;
@@ -1132,71 +1274,27 @@ DropRole(DropRoleStmt *stmt)
 					 errmsg("cannot use special role specifier in DROP ROLE")));
 		role = rolspec->rolename;
 
-		tuple = SearchSysCache1(AUTHNAME, PointerGetDatum(role));
-		if (!HeapTupleIsValid(tuple))
-		{
-			if (!stmt->missing_ok)
-			{
-				ereport(ERROR,
-						(errcode(ERRCODE_UNDEFINED_OBJECT),
-						 errmsg("role \"%s\" does not exist", role)));
-			}
-			else
-			{
-				ereport(NOTICE,
-						(errmsg("role \"%s\" does not exist, skipping",
-								role)));
-			}
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP or ALTER commits between name
+		 * resolution and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(role, AccessExclusiveLock, stmt->missing_ok,
+								RoleNameCallbackForDropRole, NULL);
 
+		if (!OidIsValid(roleid))
+		{
+			/* missing_ok case: role doesn't exist */
+			ereport(NOTICE,
+					(errmsg("role \"%s\" does not exist, skipping",
+							role)));
 			continue;
 		}
 
-		roleform = (Form_pg_authid) GETSTRUCT(tuple);
-		roleid = roleform->oid;
-
-		if (roleid == GetUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetOuterUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetSessionUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("session user cannot be dropped")));
-
-		/*
-		 * For safety's sake, we allow createrole holders to drop ordinary
-		 * roles but not superuser roles, and only if they also have ADMIN
-		 * OPTION.
-		 */
-		if (roleform->rolsuper && !superuser())
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
-							   "SUPERUSER", "SUPERUSER")));
-		if (!is_admin_of_role(GetUserId(), roleid))
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
-							   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
-
 		/* DROP hook for the role being removed */
 		InvokeObjectDropHook(AuthIdRelationId, roleid, 0);
 
-		/* Don't leak the syscache tuple */
-		ReleaseSysCache(tuple);
-
-		/*
-		 * Lock the role, so nobody can add dependencies to her while we drop
-		 * her.  We keep the lock until the end of transaction.
-		 */
-		LockSharedObject(AuthIdRelationId, roleid, 0, AccessExclusiveLock);
-
 		/*
 		 * If there is a pg_auth_members entry that has one of the roles to be
 		 * dropped as the roleid or member, it should be silently removed, but
@@ -1484,6 +1582,21 @@ RenameRole(const char *oldname, const char *newname)
 	return address;
 }
 
+/*
+ * Before acquiring a role lock for GRANT/REVOKE, check that the current user
+ * has authorization to grant/revoke membership in the specified role.
+ */
+static void
+RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+							 Oid oldroleid, void *callback_arg)
+{
+	bool		is_grant = *((bool *) callback_arg);
+	Oid			currentUserId = GetUserId();
+
+	check_role_membership_authorization(currentUserId, roleid, is_grant);
+}
+
+
 /*
  * GrantRoleStmt
  *
@@ -1568,9 +1681,14 @@ GrantRole(ParseState *pstate, GrantRoleStmt *stmt)
 					(errcode(ERRCODE_INVALID_GRANT_OPERATION),
 					 errmsg("column names cannot be included in GRANT/REVOKE ROLE")));
 
-		roleid = get_role_oid(rolename, false);
-		check_role_membership_authorization(currentUserId,
-											roleid, stmt->is_grant);
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP commits between name resolution
+		 * and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(rolename, ShareUpdateExclusiveLock, false,
+								RoleNameCallbackForGrantRole, &stmt->is_grant);
 		if (stmt->is_grant)
 			AddRoleMems(currentUserId, rolename, roleid,
 						stmt->grantee_roles, grantee_ids,
diff --git a/src/include/commands/user.h b/src/include/commands/user.h
index 97dcb93791b..17263452250 100644
--- a/src/include/commands/user.h
+++ b/src/include/commands/user.h
@@ -21,6 +21,15 @@
 extern PGDLLIMPORT int Password_encryption; /* values from enum PasswordType */
 extern PGDLLIMPORT char *createrole_self_grant;
 
+/* Callback for RoleNameGetOid, invoked after name resolution but before locking */
+typedef void (*RoleNameGetOidCallback) (const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+
+extern Oid	RoleNameGetOid(const char *rolename, LOCKMODE lockmode,
+						   bool missing_ok,
+						   RoleNameGetOidCallback callback,
+						   void *callback_arg);
+
 /* Hook to check passwords in CreateRole() and AlterRole() */
 typedef void (*check_password_hook_type) (const char *username, const char *shadow_pass, PasswordType password_type, Datum validuntil_time, bool validuntil_null);
 
-- 
2.34.1


--xlWYagR37YwJXkF8--





^ permalink  raw  reply  [nested|flat] 173+ messages in thread

* [PATCH v1] Add RoleNameGetOid() with invalidation-based retry loop for DropRole()/GrantRole()
@ 2026-07-04 04:47 Bertrand Drouvot <[email protected]>
  0 siblings, 0 replies; 173+ messages in thread

From: Bertrand Drouvot @ 2026-07-04 04:47 UTC (permalink / raw)

DropRole() and GrantRole() resolve the role name to an OID before acquiring
LockSharedObject() on the role. A concurrent session that commits a DROP ROLE
between the read and the lock acquisition leaves the first session acting on a
stale OID.

This commit fixes the races by using the same approach as RangeVarGetRelidExtended():
It encapsulates name resolution, permission checking (via a caller-supplied callback),
and lock acquisition inside a retry loop driven by SharedInvalidMessageCounter.
If invalidation messages arrive between name resolution and locking, indicating
concurrent DDL, the function retries.

The lock is kept across retries and only released if the name resolves to a
different OID on the next iteration.

Two callbacks are provided:
- RoleNameCallbackForDropRole(): checks current/session user, superuser attribute,
and ADMIN OPTION privilege before locking. This is similar to what DropRole() is
currently doing before LockSharedObject().
- RoleNameCallbackForGrantRole(): calls check_role_membership_authorization() to
verify the current user can grant/revoke membership. This is similar to what GrantRole()
is currently doing before calling AddRoleMems()/DelRoleMems().

DropRole() and GrantRole() now call RoleNameGetOid() with appropriate lock
levels.

AlterRole() does not need the fix because it calls CatalogTupleUpdate() on the
pg_authid tuple before AddRoleMems(), which blocks a concurrent DROP ROLE.

Author: Bertrand Drouvot <[email protected]>
Reviewed-by:
Discussion:
---
 src/backend/commands/user.c | 246 ++++++++++++++++++++++++++----------
 src/include/commands/user.h |   9 ++
 2 files changed, 191 insertions(+), 64 deletions(-)
  95.5% src/backend/commands/
   4.4% src/include/commands/

diff --git a/src/backend/commands/user.c b/src/backend/commands/user.c
index be11c49f919..6d8e2fa8813 100644
--- a/src/backend/commands/user.c
+++ b/src/backend/commands/user.c
@@ -34,6 +34,7 @@
 #include "miscadmin.h"
 #include "port/pg_bitutils.h"
 #include "storage/lmgr.h"
+#include "storage/sinval.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
 #include "utils/catcache.h"
@@ -116,6 +117,10 @@ static void plan_recursive_revoke(CatCList *memlist,
 								  bool revoke_admin_option_only,
 								  DropBehavior behavior);
 static void InitGrantRoleOptions(GrantRoleOptions *popt);
+static void RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+static void RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+										 Oid oldroleid, void *callback_arg);
 
 
 /* Check if current user has createrole privileges */
@@ -126,6 +131,92 @@ have_createrole_privilege(void)
 }
 
 
+/*
+ * RoleNameGetOid
+ *		Given a role name, look up its OID, lock it, and return the OID.
+ *
+ * This follows the same pattern as RangeVarGetRelidExtended():
+ * name resolution, permission check (via callback), and lock acquisition are
+ * performed inside a retry loop. If invalidation messages arrive during the
+ * process (indicating concurrent DDL), we retry to ensure the name still
+ * resolves to the same OID.
+ *
+ * The callback is invoked before locking, giving callers a chance to check
+ * permissions. It receives the current rolename, the resolved OID, the
+ * previous OID (InvalidOid on first iteration), and a caller-supplied arg.
+ * If the callback raises an error, the function aborts without locking.
+ *
+ * If missing_ok is true and the role does not exist, returns InvalidOid.
+ * Otherwise, raises an error.
+ */
+Oid
+RoleNameGetOid(const char *rolename, LOCKMODE lockmode, bool missing_ok,
+			   RoleNameGetOidCallback callback, void *callback_arg)
+{
+	uint64		inval_count;
+	Oid			roleid;
+	Oid			oldroleid = InvalidOid;
+	bool		retry = false;
+
+	for (;;)
+	{
+		/*
+		 * Remember the current invalidation count so we can detect concurrent
+		 * DDL after locking.
+		 */
+		inval_count = SharedInvalidMessageCounter;
+
+		/* Look up the role name */
+		roleid = get_role_oid(rolename, true);
+
+		if (!OidIsValid(roleid))
+		{
+			if (!missing_ok)
+				ereport(ERROR,
+						(errcode(ERRCODE_UNDEFINED_OBJECT),
+						 errmsg("role \"%s\" does not exist", rolename)));
+			return InvalidOid;
+		}
+
+		/*
+		 * Invoke caller-supplied callback before locking. This is a good
+		 * place to check permissions: we haven't taken the lock yet, but we
+		 * know the OID we intend to lock. If concurrent DDL changes things,
+		 * the callback will be invoked again on the next iteration.
+		 */
+		if (callback)
+			callback(rolename, roleid, oldroleid, callback_arg);
+
+		/*
+		 * If upon retry we get back the same OID, the invalidation messages
+		 * did not change the final answer. So we're done.
+		 *
+		 * If we got a different OID, we've locked the role that used to have
+		 * this name rather than the one that does now. Release the old lock.
+		 */
+		if (retry)
+		{
+			if (roleid == oldroleid)
+				break;
+			UnlockSharedObject(AuthIdRelationId, oldroleid, 0, lockmode);
+		}
+
+		/* Lock the role */
+		LockSharedObject(AuthIdRelationId, roleid, 0, lockmode);
+
+		/* If no invalidation messages were processed, we're done */
+		if (inval_count == SharedInvalidMessageCounter)
+			break;
+
+		/* Something may have changed, retry */
+		retry = true;
+		oldroleid = roleid;
+	}
+
+	return roleid;
+}
+
+
 /*
  * CREATE ROLE
  */
@@ -1090,6 +1181,59 @@ AlterRoleSet(AlterRoleSetStmt *stmt)
 }
 
 
+/*
+ * Before acquiring a role lock for DROP ROLE, check that the role is not the
+ * current/session user and that the caller has sufficient privileges to drop it.
+ */
+static void
+RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+							Oid oldroleid, void *callback_arg)
+{
+	HeapTuple	tuple;
+	Form_pg_authid roleform;
+
+	if (roleid == GetUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetOuterUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetSessionUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("session user cannot be dropped")));
+
+	tuple = SearchSysCache1(AUTHOID, ObjectIdGetDatum(roleid));
+	if (!HeapTupleIsValid(tuple))
+		ereport(ERROR,
+				(errcode(ERRCODE_UNDEFINED_OBJECT),
+				 errmsg("role \"%s\" does not exist", rolename)));
+
+	roleform = (Form_pg_authid) GETSTRUCT(tuple);
+
+	/*
+	 * For safety's sake, we allow createrole holders to drop ordinary roles
+	 * but not superuser roles, and only if they also have ADMIN OPTION.
+	 */
+	if (roleform->rolsuper && !superuser())
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
+						   "SUPERUSER", "SUPERUSER")));
+	if (!is_admin_of_role(GetUserId(), roleid))
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
+						   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
+
+	ReleaseSysCache(tuple);
+}
+
+
 /*
  * DROP ROLE
  */
@@ -1119,9 +1263,7 @@ DropRole(DropRoleStmt *stmt)
 	{
 		RoleSpec   *rolspec = lfirst(item);
 		char	   *role;
-		HeapTuple	tuple,
-					tmp_tuple;
-		Form_pg_authid roleform;
+		HeapTuple	tmp_tuple;
 		ScanKeyData scankey;
 		SysScanDesc sscan;
 		Oid			roleid;
@@ -1132,71 +1274,27 @@ DropRole(DropRoleStmt *stmt)
 					 errmsg("cannot use special role specifier in DROP ROLE")));
 		role = rolspec->rolename;
 
-		tuple = SearchSysCache1(AUTHNAME, PointerGetDatum(role));
-		if (!HeapTupleIsValid(tuple))
-		{
-			if (!stmt->missing_ok)
-			{
-				ereport(ERROR,
-						(errcode(ERRCODE_UNDEFINED_OBJECT),
-						 errmsg("role \"%s\" does not exist", role)));
-			}
-			else
-			{
-				ereport(NOTICE,
-						(errmsg("role \"%s\" does not exist, skipping",
-								role)));
-			}
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP or ALTER commits between name
+		 * resolution and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(role, AccessExclusiveLock, stmt->missing_ok,
+								RoleNameCallbackForDropRole, NULL);
 
+		if (!OidIsValid(roleid))
+		{
+			/* missing_ok case: role doesn't exist */
+			ereport(NOTICE,
+					(errmsg("role \"%s\" does not exist, skipping",
+							role)));
 			continue;
 		}
 
-		roleform = (Form_pg_authid) GETSTRUCT(tuple);
-		roleid = roleform->oid;
-
-		if (roleid == GetUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetOuterUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetSessionUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("session user cannot be dropped")));
-
-		/*
-		 * For safety's sake, we allow createrole holders to drop ordinary
-		 * roles but not superuser roles, and only if they also have ADMIN
-		 * OPTION.
-		 */
-		if (roleform->rolsuper && !superuser())
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
-							   "SUPERUSER", "SUPERUSER")));
-		if (!is_admin_of_role(GetUserId(), roleid))
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
-							   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
-
 		/* DROP hook for the role being removed */
 		InvokeObjectDropHook(AuthIdRelationId, roleid, 0);
 
-		/* Don't leak the syscache tuple */
-		ReleaseSysCache(tuple);
-
-		/*
-		 * Lock the role, so nobody can add dependencies to her while we drop
-		 * her.  We keep the lock until the end of transaction.
-		 */
-		LockSharedObject(AuthIdRelationId, roleid, 0, AccessExclusiveLock);
-
 		/*
 		 * If there is a pg_auth_members entry that has one of the roles to be
 		 * dropped as the roleid or member, it should be silently removed, but
@@ -1484,6 +1582,21 @@ RenameRole(const char *oldname, const char *newname)
 	return address;
 }
 
+/*
+ * Before acquiring a role lock for GRANT/REVOKE, check that the current user
+ * has authorization to grant/revoke membership in the specified role.
+ */
+static void
+RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+							 Oid oldroleid, void *callback_arg)
+{
+	bool		is_grant = *((bool *) callback_arg);
+	Oid			currentUserId = GetUserId();
+
+	check_role_membership_authorization(currentUserId, roleid, is_grant);
+}
+
+
 /*
  * GrantRoleStmt
  *
@@ -1568,9 +1681,14 @@ GrantRole(ParseState *pstate, GrantRoleStmt *stmt)
 					(errcode(ERRCODE_INVALID_GRANT_OPERATION),
 					 errmsg("column names cannot be included in GRANT/REVOKE ROLE")));
 
-		roleid = get_role_oid(rolename, false);
-		check_role_membership_authorization(currentUserId,
-											roleid, stmt->is_grant);
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP commits between name resolution
+		 * and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(rolename, ShareUpdateExclusiveLock, false,
+								RoleNameCallbackForGrantRole, &stmt->is_grant);
 		if (stmt->is_grant)
 			AddRoleMems(currentUserId, rolename, roleid,
 						stmt->grantee_roles, grantee_ids,
diff --git a/src/include/commands/user.h b/src/include/commands/user.h
index 97dcb93791b..17263452250 100644
--- a/src/include/commands/user.h
+++ b/src/include/commands/user.h
@@ -21,6 +21,15 @@
 extern PGDLLIMPORT int Password_encryption; /* values from enum PasswordType */
 extern PGDLLIMPORT char *createrole_self_grant;
 
+/* Callback for RoleNameGetOid, invoked after name resolution but before locking */
+typedef void (*RoleNameGetOidCallback) (const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+
+extern Oid	RoleNameGetOid(const char *rolename, LOCKMODE lockmode,
+						   bool missing_ok,
+						   RoleNameGetOidCallback callback,
+						   void *callback_arg);
+
 /* Hook to check passwords in CreateRole() and AlterRole() */
 typedef void (*check_password_hook_type) (const char *username, const char *shadow_pass, PasswordType password_type, Datum validuntil_time, bool validuntil_null);
 
-- 
2.34.1


--xlWYagR37YwJXkF8--





^ permalink  raw  reply  [nested|flat] 173+ messages in thread

* [PATCH v1] Add RoleNameGetOid() with invalidation-based retry loop for DropRole()/GrantRole()
@ 2026-07-04 04:47 Bertrand Drouvot <[email protected]>
  0 siblings, 0 replies; 173+ messages in thread

From: Bertrand Drouvot @ 2026-07-04 04:47 UTC (permalink / raw)

DropRole() and GrantRole() resolve the role name to an OID before acquiring
LockSharedObject() on the role. A concurrent session that commits a DROP ROLE
between the read and the lock acquisition leaves the first session acting on a
stale OID.

This commit fixes the races by using the same approach as RangeVarGetRelidExtended():
It encapsulates name resolution, permission checking (via a caller-supplied callback),
and lock acquisition inside a retry loop driven by SharedInvalidMessageCounter.
If invalidation messages arrive between name resolution and locking, indicating
concurrent DDL, the function retries.

The lock is kept across retries and only released if the name resolves to a
different OID on the next iteration.

Two callbacks are provided:
- RoleNameCallbackForDropRole(): checks current/session user, superuser attribute,
and ADMIN OPTION privilege before locking. This is similar to what DropRole() is
currently doing before LockSharedObject().
- RoleNameCallbackForGrantRole(): calls check_role_membership_authorization() to
verify the current user can grant/revoke membership. This is similar to what GrantRole()
is currently doing before calling AddRoleMems()/DelRoleMems().

DropRole() and GrantRole() now call RoleNameGetOid() with appropriate lock
levels.

AlterRole() does not need the fix because it calls CatalogTupleUpdate() on the
pg_authid tuple before AddRoleMems(), which blocks a concurrent DROP ROLE.

Author: Bertrand Drouvot <[email protected]>
Reviewed-by:
Discussion:
---
 src/backend/commands/user.c | 246 ++++++++++++++++++++++++++----------
 src/include/commands/user.h |   9 ++
 2 files changed, 191 insertions(+), 64 deletions(-)
  95.5% src/backend/commands/
   4.4% src/include/commands/

diff --git a/src/backend/commands/user.c b/src/backend/commands/user.c
index be11c49f919..6d8e2fa8813 100644
--- a/src/backend/commands/user.c
+++ b/src/backend/commands/user.c
@@ -34,6 +34,7 @@
 #include "miscadmin.h"
 #include "port/pg_bitutils.h"
 #include "storage/lmgr.h"
+#include "storage/sinval.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
 #include "utils/catcache.h"
@@ -116,6 +117,10 @@ static void plan_recursive_revoke(CatCList *memlist,
 								  bool revoke_admin_option_only,
 								  DropBehavior behavior);
 static void InitGrantRoleOptions(GrantRoleOptions *popt);
+static void RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+static void RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+										 Oid oldroleid, void *callback_arg);
 
 
 /* Check if current user has createrole privileges */
@@ -126,6 +131,92 @@ have_createrole_privilege(void)
 }
 
 
+/*
+ * RoleNameGetOid
+ *		Given a role name, look up its OID, lock it, and return the OID.
+ *
+ * This follows the same pattern as RangeVarGetRelidExtended():
+ * name resolution, permission check (via callback), and lock acquisition are
+ * performed inside a retry loop. If invalidation messages arrive during the
+ * process (indicating concurrent DDL), we retry to ensure the name still
+ * resolves to the same OID.
+ *
+ * The callback is invoked before locking, giving callers a chance to check
+ * permissions. It receives the current rolename, the resolved OID, the
+ * previous OID (InvalidOid on first iteration), and a caller-supplied arg.
+ * If the callback raises an error, the function aborts without locking.
+ *
+ * If missing_ok is true and the role does not exist, returns InvalidOid.
+ * Otherwise, raises an error.
+ */
+Oid
+RoleNameGetOid(const char *rolename, LOCKMODE lockmode, bool missing_ok,
+			   RoleNameGetOidCallback callback, void *callback_arg)
+{
+	uint64		inval_count;
+	Oid			roleid;
+	Oid			oldroleid = InvalidOid;
+	bool		retry = false;
+
+	for (;;)
+	{
+		/*
+		 * Remember the current invalidation count so we can detect concurrent
+		 * DDL after locking.
+		 */
+		inval_count = SharedInvalidMessageCounter;
+
+		/* Look up the role name */
+		roleid = get_role_oid(rolename, true);
+
+		if (!OidIsValid(roleid))
+		{
+			if (!missing_ok)
+				ereport(ERROR,
+						(errcode(ERRCODE_UNDEFINED_OBJECT),
+						 errmsg("role \"%s\" does not exist", rolename)));
+			return InvalidOid;
+		}
+
+		/*
+		 * Invoke caller-supplied callback before locking. This is a good
+		 * place to check permissions: we haven't taken the lock yet, but we
+		 * know the OID we intend to lock. If concurrent DDL changes things,
+		 * the callback will be invoked again on the next iteration.
+		 */
+		if (callback)
+			callback(rolename, roleid, oldroleid, callback_arg);
+
+		/*
+		 * If upon retry we get back the same OID, the invalidation messages
+		 * did not change the final answer. So we're done.
+		 *
+		 * If we got a different OID, we've locked the role that used to have
+		 * this name rather than the one that does now. Release the old lock.
+		 */
+		if (retry)
+		{
+			if (roleid == oldroleid)
+				break;
+			UnlockSharedObject(AuthIdRelationId, oldroleid, 0, lockmode);
+		}
+
+		/* Lock the role */
+		LockSharedObject(AuthIdRelationId, roleid, 0, lockmode);
+
+		/* If no invalidation messages were processed, we're done */
+		if (inval_count == SharedInvalidMessageCounter)
+			break;
+
+		/* Something may have changed, retry */
+		retry = true;
+		oldroleid = roleid;
+	}
+
+	return roleid;
+}
+
+
 /*
  * CREATE ROLE
  */
@@ -1090,6 +1181,59 @@ AlterRoleSet(AlterRoleSetStmt *stmt)
 }
 
 
+/*
+ * Before acquiring a role lock for DROP ROLE, check that the role is not the
+ * current/session user and that the caller has sufficient privileges to drop it.
+ */
+static void
+RoleNameCallbackForDropRole(const char *rolename, Oid roleid,
+							Oid oldroleid, void *callback_arg)
+{
+	HeapTuple	tuple;
+	Form_pg_authid roleform;
+
+	if (roleid == GetUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetOuterUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("current user cannot be dropped")));
+	if (roleid == GetSessionUserId())
+		ereport(ERROR,
+				(errcode(ERRCODE_OBJECT_IN_USE),
+				 errmsg("session user cannot be dropped")));
+
+	tuple = SearchSysCache1(AUTHOID, ObjectIdGetDatum(roleid));
+	if (!HeapTupleIsValid(tuple))
+		ereport(ERROR,
+				(errcode(ERRCODE_UNDEFINED_OBJECT),
+				 errmsg("role \"%s\" does not exist", rolename)));
+
+	roleform = (Form_pg_authid) GETSTRUCT(tuple);
+
+	/*
+	 * For safety's sake, we allow createrole holders to drop ordinary roles
+	 * but not superuser roles, and only if they also have ADMIN OPTION.
+	 */
+	if (roleform->rolsuper && !superuser())
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
+						   "SUPERUSER", "SUPERUSER")));
+	if (!is_admin_of_role(GetUserId(), roleid))
+		ereport(ERROR,
+				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
+				 errmsg("permission denied to drop role"),
+				 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
+						   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
+
+	ReleaseSysCache(tuple);
+}
+
+
 /*
  * DROP ROLE
  */
@@ -1119,9 +1263,7 @@ DropRole(DropRoleStmt *stmt)
 	{
 		RoleSpec   *rolspec = lfirst(item);
 		char	   *role;
-		HeapTuple	tuple,
-					tmp_tuple;
-		Form_pg_authid roleform;
+		HeapTuple	tmp_tuple;
 		ScanKeyData scankey;
 		SysScanDesc sscan;
 		Oid			roleid;
@@ -1132,71 +1274,27 @@ DropRole(DropRoleStmt *stmt)
 					 errmsg("cannot use special role specifier in DROP ROLE")));
 		role = rolspec->rolename;
 
-		tuple = SearchSysCache1(AUTHNAME, PointerGetDatum(role));
-		if (!HeapTupleIsValid(tuple))
-		{
-			if (!stmt->missing_ok)
-			{
-				ereport(ERROR,
-						(errcode(ERRCODE_UNDEFINED_OBJECT),
-						 errmsg("role \"%s\" does not exist", role)));
-			}
-			else
-			{
-				ereport(NOTICE,
-						(errmsg("role \"%s\" does not exist, skipping",
-								role)));
-			}
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP or ALTER commits between name
+		 * resolution and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(role, AccessExclusiveLock, stmt->missing_ok,
+								RoleNameCallbackForDropRole, NULL);
 
+		if (!OidIsValid(roleid))
+		{
+			/* missing_ok case: role doesn't exist */
+			ereport(NOTICE,
+					(errmsg("role \"%s\" does not exist, skipping",
+							role)));
 			continue;
 		}
 
-		roleform = (Form_pg_authid) GETSTRUCT(tuple);
-		roleid = roleform->oid;
-
-		if (roleid == GetUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetOuterUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("current user cannot be dropped")));
-		if (roleid == GetSessionUserId())
-			ereport(ERROR,
-					(errcode(ERRCODE_OBJECT_IN_USE),
-					 errmsg("session user cannot be dropped")));
-
-		/*
-		 * For safety's sake, we allow createrole holders to drop ordinary
-		 * roles but not superuser roles, and only if they also have ADMIN
-		 * OPTION.
-		 */
-		if (roleform->rolsuper && !superuser())
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute may drop roles with the %s attribute.",
-							   "SUPERUSER", "SUPERUSER")));
-		if (!is_admin_of_role(GetUserId(), roleid))
-			ereport(ERROR,
-					(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
-					 errmsg("permission denied to drop role"),
-					 errdetail("Only roles with the %s attribute and the %s option on role \"%s\" may drop this role.",
-							   "CREATEROLE", "ADMIN", NameStr(roleform->rolname))));
-
 		/* DROP hook for the role being removed */
 		InvokeObjectDropHook(AuthIdRelationId, roleid, 0);
 
-		/* Don't leak the syscache tuple */
-		ReleaseSysCache(tuple);
-
-		/*
-		 * Lock the role, so nobody can add dependencies to her while we drop
-		 * her.  We keep the lock until the end of transaction.
-		 */
-		LockSharedObject(AuthIdRelationId, roleid, 0, AccessExclusiveLock);
-
 		/*
 		 * If there is a pg_auth_members entry that has one of the roles to be
 		 * dropped as the roleid or member, it should be silently removed, but
@@ -1484,6 +1582,21 @@ RenameRole(const char *oldname, const char *newname)
 	return address;
 }
 
+/*
+ * Before acquiring a role lock for GRANT/REVOKE, check that the current user
+ * has authorization to grant/revoke membership in the specified role.
+ */
+static void
+RoleNameCallbackForGrantRole(const char *rolename, Oid roleid,
+							 Oid oldroleid, void *callback_arg)
+{
+	bool		is_grant = *((bool *) callback_arg);
+	Oid			currentUserId = GetUserId();
+
+	check_role_membership_authorization(currentUserId, roleid, is_grant);
+}
+
+
 /*
  * GrantRoleStmt
  *
@@ -1568,9 +1681,14 @@ GrantRole(ParseState *pstate, GrantRoleStmt *stmt)
 					(errcode(ERRCODE_INVALID_GRANT_OPERATION),
 					 errmsg("column names cannot be included in GRANT/REVOKE ROLE")));
 
-		roleid = get_role_oid(rolename, false);
-		check_role_membership_authorization(currentUserId,
-											roleid, stmt->is_grant);
+		/*
+		 * Use RoleNameGetOid to resolve the name, check permissions, and lock
+		 * the role atomically with a retry loop. This prevents race
+		 * conditions where a concurrent DROP commits between name resolution
+		 * and lock acquisition.
+		 */
+		roleid = RoleNameGetOid(rolename, ShareUpdateExclusiveLock, false,
+								RoleNameCallbackForGrantRole, &stmt->is_grant);
 		if (stmt->is_grant)
 			AddRoleMems(currentUserId, rolename, roleid,
 						stmt->grantee_roles, grantee_ids,
diff --git a/src/include/commands/user.h b/src/include/commands/user.h
index 97dcb93791b..17263452250 100644
--- a/src/include/commands/user.h
+++ b/src/include/commands/user.h
@@ -21,6 +21,15 @@
 extern PGDLLIMPORT int Password_encryption; /* values from enum PasswordType */
 extern PGDLLIMPORT char *createrole_self_grant;
 
+/* Callback for RoleNameGetOid, invoked after name resolution but before locking */
+typedef void (*RoleNameGetOidCallback) (const char *rolename, Oid roleid,
+										Oid oldroleid, void *callback_arg);
+
+extern Oid	RoleNameGetOid(const char *rolename, LOCKMODE lockmode,
+						   bool missing_ok,
+						   RoleNameGetOidCallback callback,
+						   void *callback_arg);
+
 /* Hook to check passwords in CreateRole() and AlterRole() */
 typedef void (*check_password_hook_type) (const char *username, const char *shadow_pass, PasswordType password_type, Datum validuntil_time, bool validuntil_null);
 
-- 
2.34.1


--xlWYagR37YwJXkF8--





^ permalink  raw  reply  [nested|flat] 173+ messages in thread