agora inbox for [email protected]  

help / color / mirror / Atom feed

[PATCH v11 1/7] Row pattern recognition patch for raw parser.
966+ messages / 2 participants
[nested] [flat]

* [PATCH v11 1/7] Row pattern recognition patch for raw parser.
@ 2023-11-08 06:57  Tatsuo Ishii <[email protected]>
  0 siblings, 0 replies; 966+ messages in thread

From: Tatsuo Ishii @ 2023-11-08 06:57 UTC (permalink / raw)

---
 src/backend/parser/gram.y       | 222 ++++++++++++++++++++++++++++++--
 src/include/nodes/parsenodes.h  |  56 ++++++++
 src/include/parser/kwlist.h     |   8 ++
 src/include/parser/parse_node.h |   1 +
 4 files changed, 273 insertions(+), 14 deletions(-)

diff --git a/src/backend/parser/gram.y b/src/backend/parser/gram.y
index c224df4ecc..e09eb061f8 100644
--- a/src/backend/parser/gram.y
+++ b/src/backend/parser/gram.y
@@ -251,6 +251,8 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);
 	DefElem	   *defelt;
 	SortBy	   *sortby;
 	WindowDef  *windef;
+	RPCommonSyntax	*rpcom;
+	RPSubsetItem	*rpsubset;
 	JoinExpr   *jexpr;
 	IndexElem  *ielem;
 	StatsElem  *selem;
@@ -278,6 +280,7 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);
 	MergeWhenClause *mergewhen;
 	struct KeyActions *keyactions;
 	struct KeyAction *keyaction;
+	RPSkipTo	skipto;
 }
 
 %type <node>	stmt toplevel_stmt schema_stmt routine_body_stmt
@@ -453,8 +456,12 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);
 				TriggerTransitions TriggerReferencing
 				vacuum_relation_list opt_vacuum_relation_list
 				drop_option_list pub_obj_list
-
-%type <node>	opt_routine_body
+				row_pattern_measure_list row_pattern_definition_list
+				opt_row_pattern_subset_clause
+				row_pattern_subset_list row_pattern_subset_rhs
+				row_pattern
+%type <rpsubset>	 row_pattern_subset_item
+%type <node>	opt_routine_body row_pattern_term
 %type <groupclause> group_clause
 %type <list>	group_by_list
 %type <node>	group_by_item empty_grouping_set rollup_clause cube_clause
@@ -551,6 +558,8 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);
 %type <range>	relation_expr_opt_alias
 %type <node>	tablesample_clause opt_repeatable_clause
 %type <target>	target_el set_target insert_column_item
+				row_pattern_measure_item row_pattern_definition
+%type <skipto>	first_or_last
 
 %type <str>		generic_option_name
 %type <node>	generic_option_arg
@@ -633,6 +642,9 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);
 %type <list>	window_clause window_definition_list opt_partition_clause
 %type <windef>	window_definition over_clause window_specification
 				opt_frame_clause frame_extent frame_bound
+%type <rpcom>	opt_row_pattern_common_syntax opt_row_pattern_skip_to
+%type <boolean>	opt_row_pattern_initial_or_seek
+%type <list>	opt_row_pattern_measures
 %type <ival>	opt_window_exclusion_clause
 %type <str>		opt_existing_window_name
 %type <boolean> opt_if_not_exists
@@ -659,7 +671,6 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);
 				json_object_constructor_null_clause_opt
 				json_array_constructor_null_clause_opt
 
-
 /*
  * Non-keyword token types.  These are hard-wired into the "flex" lexer.
  * They must be listed first so that their numeric codes do not depend on
@@ -702,7 +713,7 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);
 	CURRENT_TIME CURRENT_TIMESTAMP CURRENT_USER CURSOR CYCLE
 
 	DATA_P DATABASE DAY_P DEALLOCATE DEC DECIMAL_P DECLARE DEFAULT DEFAULTS
-	DEFERRABLE DEFERRED DEFINER DELETE_P DELIMITER DELIMITERS DEPENDS DEPTH DESC
+	DEFERRABLE DEFERRED DEFINE DEFINER DELETE_P DELIMITER DELIMITERS DEPENDS DEPTH DESC
 	DETACH DICTIONARY DISABLE_P DISCARD DISTINCT DO DOCUMENT_P DOMAIN_P
 	DOUBLE_P DROP
 
@@ -718,7 +729,7 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);
 	HANDLER HAVING HEADER_P HOLD HOUR_P
 
 	IDENTITY_P IF_P ILIKE IMMEDIATE IMMUTABLE IMPLICIT_P IMPORT_P IN_P INCLUDE
-	INCLUDING INCREMENT INDENT INDEX INDEXES INHERIT INHERITS INITIALLY INLINE_P
+	INCLUDING INCREMENT INDENT INDEX INDEXES INHERIT INHERITS INITIAL INITIALLY INLINE_P
 	INNER_P INOUT INPUT_P INSENSITIVE INSERT INSTEAD INT_P INTEGER
 	INTERSECT INTERVAL INTO INVOKER IS ISNULL ISOLATION
 
@@ -731,7 +742,7 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);
 	LEADING LEAKPROOF LEAST LEFT LEVEL LIKE LIMIT LISTEN LOAD LOCAL
 	LOCALTIME LOCALTIMESTAMP LOCATION LOCK_P LOCKED LOGGED
 
-	MAPPING MATCH MATCHED MATERIALIZED MAXVALUE MERGE METHOD
+	MAPPING MATCH MATCHED MATERIALIZED MAXVALUE MEASURES MERGE METHOD
 	MINUTE_P MINVALUE MODE MONTH_P MOVE
 
 	NAME_P NAMES NATIONAL NATURAL NCHAR NEW NEXT NFC NFD NFKC NFKD NO NONE
@@ -743,8 +754,8 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);
 	ORDER ORDINALITY OTHERS OUT_P OUTER_P
 	OVER OVERLAPS OVERLAY OVERRIDING OWNED OWNER
 
-	PARALLEL PARAMETER PARSER PARTIAL PARTITION PASSING PASSWORD
-	PLACING PLANS POLICY
+	PARALLEL PARAMETER PARSER PARTIAL PARTITION PASSING PASSWORD PAST
+	PATTERN_P PERMUTE PLACING PLANS POLICY
 	POSITION PRECEDING PRECISION PRESERVE PREPARE PREPARED PRIMARY
 	PRIOR PRIVILEGES PROCEDURAL PROCEDURE PROCEDURES PROGRAM PUBLICATION
 
@@ -755,12 +766,13 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);
 	RESET RESTART RESTRICT RETURN RETURNING RETURNS REVOKE RIGHT ROLE ROLLBACK ROLLUP
 	ROUTINE ROUTINES ROW ROWS RULE
 
-	SAVEPOINT SCALAR SCHEMA SCHEMAS SCROLL SEARCH SECOND_P SECURITY SELECT
+	SAVEPOINT SCALAR SCHEMA SCHEMAS SCROLL SEARCH SECOND_P SECURITY SEEK SELECT
 	SEQUENCE SEQUENCES
+
 	SERIALIZABLE SERVER SESSION SESSION_USER SET SETS SETOF SHARE SHOW
 	SIMILAR SIMPLE SKIP SMALLINT SNAPSHOT SOME SQL_P STABLE STANDALONE_P
 	START STATEMENT STATISTICS STDIN STDOUT STORAGE STORED STRICT_P STRIP_P
-	SUBSCRIPTION SUBSTRING SUPPORT SYMMETRIC SYSID SYSTEM_P SYSTEM_USER
+	SUBSCRIPTION SUBSET SUBSTRING SUPPORT SYMMETRIC SYSID SYSTEM_P SYSTEM_USER
 
 	TABLE TABLES TABLESAMPLE TABLESPACE TEMP TEMPLATE TEMPORARY TEXT_P THEN
 	TIES TIME TIMESTAMP TO TRAILING TRANSACTION TRANSFORM
@@ -853,6 +865,7 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);
  */
 %nonassoc	UNBOUNDED		/* ideally would have same precedence as IDENT */
 %nonassoc	IDENT PARTITION RANGE ROWS GROUPS PRECEDING FOLLOWING CUBE ROLLUP
+%nonassoc	MEASURES AFTER INITIAL SEEK PATTERN_P
 %left		Op OPERATOR		/* multi-character ops and user-defined operators */
 %left		'+' '-'
 %left		'*' '/' '%'
@@ -15901,7 +15914,8 @@ over_clause: OVER window_specification
 		;
 
 window_specification: '(' opt_existing_window_name opt_partition_clause
-						opt_sort_clause opt_frame_clause ')'
+						opt_sort_clause opt_row_pattern_measures opt_frame_clause
+						opt_row_pattern_common_syntax ')'
 				{
 					WindowDef  *n = makeNode(WindowDef);
 
@@ -15909,10 +15923,12 @@ window_specification: '(' opt_existing_window_name opt_partition_clause
 					n->refname = $2;
 					n->partitionClause = $3;
 					n->orderClause = $4;
+					n->rowPatternMeasures = $5;
 					/* copy relevant fields of opt_frame_clause */
-					n->frameOptions = $5->frameOptions;
-					n->startOffset = $5->startOffset;
-					n->endOffset = $5->endOffset;
+					n->frameOptions = $6->frameOptions;
+					n->startOffset = $6->startOffset;
+					n->endOffset = $6->endOffset;
+					n->rpCommonSyntax = $7;
 					n->location = @1;
 					$$ = n;
 				}
@@ -15936,6 +15952,31 @@ opt_partition_clause: PARTITION BY expr_list		{ $$ = $3; }
 			| /*EMPTY*/								{ $$ = NIL; }
 		;
 
+/*
+ * ROW PATTERN_P MEASURES
+ */
+opt_row_pattern_measures: MEASURES row_pattern_measure_list	{ $$ = $2; }
+			| /*EMPTY*/								{ $$ = NIL; }
+		;
+
+row_pattern_measure_list:
+			row_pattern_measure_item
+					{ $$ = list_make1($1); }
+			| row_pattern_measure_list ',' row_pattern_measure_item
+					{ $$ = lappend($1, $3); }
+		;
+
+row_pattern_measure_item:
+			a_expr AS ColLabel
+				{
+					$$ = makeNode(ResTarget);
+					$$->name = $3;
+					$$->indirection = NIL;
+					$$->val = (Node *) $1;
+					$$->location = @1;
+				}
+		;
+
 /*
  * For frame clauses, we return a WindowDef, but only some fields are used:
  * frameOptions, startOffset, and endOffset.
@@ -16095,6 +16136,143 @@ opt_window_exclusion_clause:
 			| /*EMPTY*/				{ $$ = 0; }
 		;
 
+opt_row_pattern_common_syntax:
+opt_row_pattern_skip_to opt_row_pattern_initial_or_seek
+				PATTERN_P '(' row_pattern ')'
+				opt_row_pattern_subset_clause
+				DEFINE row_pattern_definition_list
+			{
+				RPCommonSyntax *n = makeNode(RPCommonSyntax);
+				n->rpSkipTo = $1->rpSkipTo;
+				n->rpSkipVariable = $1->rpSkipVariable;
+				n->initial = $2;
+				n->rpPatterns = $5;
+				n->rpSubsetClause = $7;
+				n->rpDefs = $9;
+				$$ = n;
+			}
+			| /*EMPTY*/		{ $$ = NULL; }
+	;
+
+opt_row_pattern_skip_to:
+			AFTER MATCH SKIP TO NEXT ROW
+				{
+					RPCommonSyntax *n = makeNode(RPCommonSyntax);
+					n->rpSkipTo = ST_NEXT_ROW;
+					n->rpSkipVariable = NULL;
+					$$ = n;
+			}
+			| AFTER MATCH SKIP PAST LAST_P ROW
+				{
+					RPCommonSyntax *n = makeNode(RPCommonSyntax);
+					n->rpSkipTo = ST_PAST_LAST_ROW;
+					n->rpSkipVariable = NULL;
+					$$ = n;
+				}
+			| AFTER MATCH SKIP TO first_or_last ColId
+				{
+					RPCommonSyntax *n = makeNode(RPCommonSyntax);
+					n->rpSkipTo = $5;
+					n->rpSkipVariable = $6;
+					$$ = n;
+				}
+/*
+			| AFTER MATCH SKIP TO LAST_P ColId		%prec LAST_P
+				{
+					RPCommonSyntax *n = makeNode(RPCommonSyntax);
+					n->rpSkipTo = ST_LAST_VARIABLE;
+					n->rpSkipVariable = $6;
+					$$ = n;
+				}
+			| AFTER MATCH SKIP TO ColId
+				{
+					RPCommonSyntax *n = makeNode(RPCommonSyntax);
+					n->rpSkipTo = ST_VARIABLE;
+					n->rpSkipVariable = $5;
+					$$ = n;
+				}
+*/
+			| /*EMPTY*/
+				{
+					RPCommonSyntax *n = makeNode(RPCommonSyntax);
+					/* temporary set default to ST_NEXT_ROW */
+					n->rpSkipTo = ST_PAST_LAST_ROW;
+					n->rpSkipVariable = NULL;
+					$$ = n;
+				}
+	;
+
+first_or_last:
+			FIRST_P		{ $$ = ST_FIRST_VARIABLE; }
+			| LAST_P	{ $$ = ST_LAST_VARIABLE; }
+	;
+
+opt_row_pattern_initial_or_seek:
+			INITIAL			{ $$ = true; }
+			| SEEK
+				{
+					ereport(ERROR,
+							(errcode(ERRCODE_SYNTAX_ERROR),
+							 errmsg("SEEK is not supported"),
+							 errhint("Use INITIAL."),
+							 parser_errposition(@1)));
+				}
+			| /*EMPTY*/		{ $$ = true; }
+		;
+
+row_pattern:
+			row_pattern_term							{ $$ = list_make1($1); }
+			| row_pattern row_pattern_term				{ $$ = lappend($1, $2); }
+		;
+
+row_pattern_term:
+			ColId	{ $$ = (Node *) makeSimpleA_Expr(AEXPR_OP, "", (Node *)makeString($1), NULL, @1); }
+			| ColId '*'	{ $$ = (Node *) makeSimpleA_Expr(AEXPR_OP, "*", (Node *)makeString($1), NULL, @1); }
+			| ColId '+'	{ $$ = (Node *) makeSimpleA_Expr(AEXPR_OP, "+", (Node *)makeString($1), NULL, @1); }
+			| ColId '?'	{ $$ = (Node *) makeSimpleA_Expr(AEXPR_OP, "?", (Node *)makeString($1), NULL, @1); }
+		;
+
+opt_row_pattern_subset_clause:
+			SUBSET row_pattern_subset_list	{ $$ = $2; }
+			| /*EMPTY*/												{ $$ = NIL; }
+		;
+
+row_pattern_subset_list:
+			row_pattern_subset_item									{ $$ = list_make1($1); }
+			| row_pattern_subset_list ',' row_pattern_subset_item	{ $$ = lappend($1, $3); }
+			| /*EMPTY*/												{ $$ = NIL; }
+		;
+
+row_pattern_subset_item: ColId '=' '(' row_pattern_subset_rhs ')'
+			{
+				RPSubsetItem *n = makeNode(RPSubsetItem);
+				n->name = $1;
+				n->rhsVariable = $4;
+				$$ = n;
+			}
+		;
+
+row_pattern_subset_rhs:
+			ColId								{ $$ = list_make1(makeStringConst($1, @1)); }
+			| row_pattern_subset_rhs ',' ColId	{ $$ = lappend($1, makeStringConst($3, @1)); }
+			| /*EMPTY*/							{ $$ = NIL; }
+		;
+
+row_pattern_definition_list:
+			row_pattern_definition										{ $$ = list_make1($1); }
+			| row_pattern_definition_list ',' row_pattern_definition	{ $$ = lappend($1, $3); }
+		;
+
+row_pattern_definition:
+			ColId AS a_expr
+				{
+					$$ = makeNode(ResTarget);
+					$$->name = $1;
+					$$->indirection = NIL;
+					$$->val = (Node *) $3;
+					$$->location = @1;
+				}
+		;
 
 /*
  * Supporting nonterminals for expressions.
@@ -17190,6 +17368,7 @@ unreserved_keyword:
 			| INDEXES
 			| INHERIT
 			| INHERITS
+			| INITIAL
 			| INLINE_P
 			| INPUT_P
 			| INSENSITIVE
@@ -17217,6 +17396,7 @@ unreserved_keyword:
 			| MATCHED
 			| MATERIALIZED
 			| MAXVALUE
+			| MEASURES
 			| MERGE
 			| METHOD
 			| MINUTE_P
@@ -17259,6 +17439,9 @@ unreserved_keyword:
 			| PARTITION
 			| PASSING
 			| PASSWORD
+			| PAST
+			| PATTERN_P
+			| PERMUTE
 			| PLANS
 			| POLICY
 			| PRECEDING
@@ -17309,6 +17492,7 @@ unreserved_keyword:
 			| SEARCH
 			| SECOND_P
 			| SECURITY
+			| SEEK
 			| SEQUENCE
 			| SEQUENCES
 			| SERIALIZABLE
@@ -17334,6 +17518,7 @@ unreserved_keyword:
 			| STRICT_P
 			| STRIP_P
 			| SUBSCRIPTION
+			| SUBSET
 			| SUPPORT
 			| SYSID
 			| SYSTEM_P
@@ -17521,6 +17706,7 @@ reserved_keyword:
 			| CURRENT_USER
 			| DEFAULT
 			| DEFERRABLE
+			| DEFINE
 			| DESC
 			| DISTINCT
 			| DO
@@ -17683,6 +17869,7 @@ bare_label_keyword:
 			| DEFAULTS
 			| DEFERRABLE
 			| DEFERRED
+			| DEFINE
 			| DEFINER
 			| DELETE_P
 			| DELIMITER
@@ -17758,6 +17945,7 @@ bare_label_keyword:
 			| INDEXES
 			| INHERIT
 			| INHERITS
+			| INITIAL
 			| INITIALLY
 			| INLINE_P
 			| INNER_P
@@ -17807,6 +17995,7 @@ bare_label_keyword:
 			| MATCHED
 			| MATERIALIZED
 			| MAXVALUE
+			| MEASURES
 			| MERGE
 			| METHOD
 			| MINVALUE
@@ -17860,6 +18049,9 @@ bare_label_keyword:
 			| PARTITION
 			| PASSING
 			| PASSWORD
+			| PAST
+			| PATTERN_P
+			| PERMUTE
 			| PLACING
 			| PLANS
 			| POLICY
@@ -17916,6 +18108,7 @@ bare_label_keyword:
 			| SCROLL
 			| SEARCH
 			| SECURITY
+			| SEEK
 			| SELECT
 			| SEQUENCE
 			| SEQUENCES
@@ -17947,6 +18140,7 @@ bare_label_keyword:
 			| STRICT_P
 			| STRIP_P
 			| SUBSCRIPTION
+			| SUBSET
 			| SUBSTRING
 			| SUPPORT
 			| SYMMETRIC
diff --git a/src/include/nodes/parsenodes.h b/src/include/nodes/parsenodes.h
index e494309da8..094c603887 100644
--- a/src/include/nodes/parsenodes.h
+++ b/src/include/nodes/parsenodes.h
@@ -540,6 +540,44 @@ typedef struct SortBy
 	int			location;		/* operator location, or -1 if none/unknown */
 } SortBy;
 
+/*
+ * AFTER MATCH row pattern skip to types in row pattern common syntax
+ */
+typedef enum RPSkipTo
+{
+	ST_NONE,					/* AFTER MATCH omitted */
+	ST_NEXT_ROW,				/* SKIP TO NEXT ROW */
+	ST_PAST_LAST_ROW,			/* SKIP TO PAST LAST ROW */
+	ST_FIRST_VARIABLE,			/* SKIP TO FIRST variable name */
+	ST_LAST_VARIABLE,			/* SKIP TO LAST variable name */
+	ST_VARIABLE					/* SKIP TO variable name */
+} RPSkipTo;
+
+/*
+ * Row Pattern SUBSET clause item
+ */
+typedef struct RPSubsetItem
+{
+	NodeTag		type;
+	char	   *name;			/* Row Pattern SUBSET clause variable name */
+	List	   *rhsVariable;	/* Row Pattern SUBSET rhs variables (list of char *string) */
+} RPSubsetItem;
+
+/*
+ * RowPatternCommonSyntax - raw representation of row pattern common syntax
+ *
+ */
+typedef struct RPCommonSyntax
+{
+	NodeTag		type;
+	RPSkipTo	rpSkipTo;		/* Row Pattern AFTER MATCH SKIP type */
+	char	   *rpSkipVariable;	/* Row Pattern Skip To variable name, if any */
+	bool		initial;		/* true if <row pattern initial or seek> is initial */
+	List	   *rpPatterns;		/* PATTERN variables (list of A_Expr) */
+	List	   *rpSubsetClause;	/* row pattern subset clause (list of RPSubsetItem), if any */
+	List	   *rpDefs;			/* row pattern definitions clause (list of ResTarget) */
+} RPCommonSyntax;
+
 /*
  * WindowDef - raw representation of WINDOW and OVER clauses
  *
@@ -555,6 +593,8 @@ typedef struct WindowDef
 	char	   *refname;		/* referenced window name, if any */
 	List	   *partitionClause;	/* PARTITION BY expression list */
 	List	   *orderClause;	/* ORDER BY (list of SortBy) */
+	List	   *rowPatternMeasures;	/* row pattern measures (list of ResTarget) */
+	RPCommonSyntax *rpCommonSyntax;	/* row pattern common syntax */
 	int			frameOptions;	/* frame_clause options, see below */
 	Node	   *startOffset;	/* expression for starting bound, if any */
 	Node	   *endOffset;		/* expression for ending bound, if any */
@@ -1476,6 +1516,11 @@ typedef struct GroupingSet
  * the orderClause might or might not be copied (see copiedOrder); the framing
  * options are never copied, per spec.
  *
+ * "defineClause" is Row Pattern Recognition DEFINE clause (list of
+ * TargetEntry). TargetEntry.resname represents row pattern definition
+ * variable name. "patternVariable" and "patternRegexp" represents PATTERN
+ * clause.
+ *
  * The information relevant for the query jumbling is the partition clause
  * type and its bounds.
  */
@@ -1507,6 +1552,17 @@ typedef struct WindowClause
 	Index		winref;			/* ID referenced by window functions */
 	/* did we copy orderClause from refname? */
 	bool		copiedOrder pg_node_attr(query_jumble_ignore);
+	/* Row Pattern AFTER MACH SKIP clause */
+	RPSkipTo	rpSkipTo;		/* Row Pattern Skip To type */
+	bool		initial;		/* true if <row pattern initial or seek> is initial */
+	/* Row Pattern DEFINE clause (list of TargetEntry) */
+	List		*defineClause;
+	/* Row Pattern DEFINE variable initial names (list of String) */
+	List		*defineInitial;
+	/* Row Pattern PATTERN variable name (list of String) */
+	List		*patternVariable;
+	/* Row Pattern PATTERN regular expression quantifier ('+' or ''. list of String) */
+	List		*patternRegexp;
 } WindowClause;
 
 /*
diff --git a/src/include/parser/kwlist.h b/src/include/parser/kwlist.h
index 5984dcfa4b..2804333b53 100644
--- a/src/include/parser/kwlist.h
+++ b/src/include/parser/kwlist.h
@@ -128,6 +128,7 @@ PG_KEYWORD("default", DEFAULT, RESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("defaults", DEFAULTS, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("deferrable", DEFERRABLE, RESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("deferred", DEFERRED, UNRESERVED_KEYWORD, BARE_LABEL)
+PG_KEYWORD("define", DEFINE, RESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("definer", DEFINER, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("delete", DELETE_P, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("delimiter", DELIMITER, UNRESERVED_KEYWORD, BARE_LABEL)
@@ -212,6 +213,7 @@ PG_KEYWORD("index", INDEX, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("indexes", INDEXES, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("inherit", INHERIT, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("inherits", INHERITS, UNRESERVED_KEYWORD, BARE_LABEL)
+PG_KEYWORD("initial", INITIAL, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("initially", INITIALLY, RESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("inline", INLINE_P, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("inner", INNER_P, TYPE_FUNC_NAME_KEYWORD, BARE_LABEL)
@@ -265,6 +267,7 @@ PG_KEYWORD("match", MATCH, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("matched", MATCHED, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("materialized", MATERIALIZED, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("maxvalue", MAXVALUE, UNRESERVED_KEYWORD, BARE_LABEL)
+PG_KEYWORD("measures", MEASURES, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("merge", MERGE, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("method", METHOD, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("minute", MINUTE_P, UNRESERVED_KEYWORD, AS_LABEL)
@@ -326,6 +329,9 @@ PG_KEYWORD("partial", PARTIAL, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("partition", PARTITION, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("passing", PASSING, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("password", PASSWORD, UNRESERVED_KEYWORD, BARE_LABEL)
+PG_KEYWORD("past", PAST, UNRESERVED_KEYWORD, BARE_LABEL)
+PG_KEYWORD("pattern", PATTERN_P, UNRESERVED_KEYWORD, BARE_LABEL)
+PG_KEYWORD("permute", PERMUTE, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("placing", PLACING, RESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("plans", PLANS, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("policy", POLICY, UNRESERVED_KEYWORD, BARE_LABEL)
@@ -385,6 +391,7 @@ PG_KEYWORD("scroll", SCROLL, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("search", SEARCH, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("second", SECOND_P, UNRESERVED_KEYWORD, AS_LABEL)
 PG_KEYWORD("security", SECURITY, UNRESERVED_KEYWORD, BARE_LABEL)
+PG_KEYWORD("seek", SEEK, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("select", SELECT, RESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("sequence", SEQUENCE, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("sequences", SEQUENCES, UNRESERVED_KEYWORD, BARE_LABEL)
@@ -416,6 +423,7 @@ PG_KEYWORD("stored", STORED, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("strict", STRICT_P, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("strip", STRIP_P, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("subscription", SUBSCRIPTION, UNRESERVED_KEYWORD, BARE_LABEL)
+PG_KEYWORD("subset", SUBSET, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("substring", SUBSTRING, COL_NAME_KEYWORD, BARE_LABEL)
 PG_KEYWORD("support", SUPPORT, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("symmetric", SYMMETRIC, RESERVED_KEYWORD, BARE_LABEL)
diff --git a/src/include/parser/parse_node.h b/src/include/parser/parse_node.h
index f589112d5e..6640090910 100644
--- a/src/include/parser/parse_node.h
+++ b/src/include/parser/parse_node.h
@@ -51,6 +51,7 @@ typedef enum ParseExprKind
 	EXPR_KIND_WINDOW_FRAME_RANGE,	/* window frame clause with RANGE */
 	EXPR_KIND_WINDOW_FRAME_ROWS,	/* window frame clause with ROWS */
 	EXPR_KIND_WINDOW_FRAME_GROUPS,	/* window frame clause with GROUPS */
+	EXPR_KIND_RPR_DEFINE,		/* DEFINE */
 	EXPR_KIND_SELECT_TARGET,	/* SELECT target list item */
 	EXPR_KIND_INSERT_TARGET,	/* INSERT target list item */
 	EXPR_KIND_UPDATE_SOURCE,	/* UPDATE assignment source item */
-- 
2.25.1


----Next_Part(Wed_Nov__8_16_37_05_2023_872)--
Content-Type: Text/X-Patch; charset=us-ascii
Content-Transfer-Encoding: 7bit
Content-Disposition: inline;
 filename="v11-0002-Row-pattern-recognition-patch-parse-analysis.patch"



^ permalink  raw  reply  [nested|flat] 966+ messages in thread

* [PATCH] Lock upgrade without deadlocks.
@ 2026-04-17 13:13  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 966+ messages in thread

From: Antonin Houska @ 2026-04-17 13:13 UTC (permalink / raw)

For REPACK (CONCURRENTLY), it's essential to upgrade ShareUpdateExclusive lock
to AccessExclusiveLock for the final stage of table processing. However, we
cannot release the earlier before requesting the latter, because if another
process ran DDL on the table in between, REPACK would have to cancel its
transaction, and waste possibly a lot of work. Without releasing
ShareUpdateExclusive lock, the upgrade can cause deadlock, and it's not really
deterministic whether REPACK or the other process will get canceled.

This patch adds a new field to the LOCK structure which tells which process is
trying to upgrade the lock - this field is set right before the upgrade. All
processes waiting in the lock queue are then woken up to check the flag.

When another process tries to get the lock after that, it checks this field,
and if it's set, it checks for deadlocks even if deadlock timeout hasn't
expired yet. If the process is already in the lock's queue and sleeping, the
lock upgrading process wakes it up so it checks the flag immediately.

At the time the upgrading process starts to sleep, all the other process
should be aware that they need to check for deadlock during their lock
acquisitions, so the upgrading process does not have to check for deadlocks
when it gets woken up. Thus it should never receive deadlock error report.
---
 src/backend/commands/repack.c   |   4 +-
 src/backend/storage/lmgr/lmgr.c |  35 ++++++++---
 src/backend/storage/lmgr/lock.c | 107 +++++++++++++++++++++++++++++++-
 src/backend/storage/lmgr/proc.c |  32 +++++++---
 src/include/storage/lmgr.h      |   1 +
 src/include/storage/lock.h      |   5 +-
 src/include/storage/proc.h      |   4 +-
 7 files changed, 166 insertions(+), 22 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index 58e3867246f..310b2a65099 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -3055,8 +3055,10 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Acquire AccessExclusiveLock on the table, its TOAST relation (if there
 	 * is one), all its indexes, so that we can swap the files.
+	 *
+	 * TODO The same for indexes and TOAST?
 	 */
-	LockRelationOid(old_table_oid, AccessExclusiveLock);
+	LockRelationOidUpgrade(old_table_oid, AccessExclusiveLock);
 
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
diff --git a/src/backend/storage/lmgr/lmgr.c b/src/backend/storage/lmgr/lmgr.c
index 2ccf7237fee..0b103485d54 100644
--- a/src/backend/storage/lmgr/lmgr.c
+++ b/src/backend/storage/lmgr/lmgr.c
@@ -58,6 +58,8 @@ typedef struct XactLockTableWaitInfo
 	const ItemPointerData *ctid;
 } XactLockTableWaitInfo;
 
+static void LockRelationOidCommon(Oid relid, LOCKMODE lockmode,
+								  bool isUpgrade);
 static void XactLockTableWaitErrorCb(void *arg);
 
 /*
@@ -105,6 +107,21 @@ SetLocktagRelationOid(LOCKTAG *tag, Oid relid)
  */
 void
 LockRelationOid(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, false);
+}
+
+/*
+ * Like above, but upgrade an existing lock.
+ */
+void
+LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, true);
+}
+
+static void
+LockRelationOidCommon(Oid relid, LOCKMODE lockmode, bool isUpgrade)
 {
 	LOCKTAG		tag;
 	LOCALLOCK  *locallock;
@@ -113,7 +130,7 @@ LockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, isUpgrade);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages, so that we
@@ -157,7 +174,7 @@ ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -191,7 +208,7 @@ LockRelationId(LockRelId *relid, LOCKMODE lockmode)
 	SET_LOCKTAG_RELATION(tag, relid->dbId, relid->relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -254,7 +271,7 @@ LockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -286,7 +303,7 @@ ConditionalLockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -591,7 +608,7 @@ ConditionalLockTuple(Relation relation, const ItemPointerData *tid, LOCKMODE loc
 					  ItemPointerGetOffsetNumber(tid));
 
 	return (LockAcquireExtended(&tag, lockmode, false, true, true, NULL,
-								logLockFailure) != LOCKACQUIRE_NOT_AVAIL);
+								logLockFailure, false) != LOCKACQUIRE_NOT_AVAIL);
 }
 
 /*
@@ -749,7 +766,7 @@ ConditionalXactLockTableWait(TransactionId xid, bool logLockFailure)
 		SET_LOCKTAG_TRANSACTION(tag, xid);
 
 		if (LockAcquireExtended(&tag, ShareLock, false, true, true, NULL,
-								logLockFailure)
+								logLockFailure, false)
 			== LOCKACQUIRE_NOT_AVAIL)
 			return false;
 
@@ -1043,7 +1060,7 @@ ConditionalLockDatabaseObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -1123,7 +1140,7 @@ ConditionalLockSharedObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
diff --git a/src/backend/storage/lmgr/lock.c b/src/backend/storage/lmgr/lock.c
index c221fe96889..bcbbf36c091 100644
--- a/src/backend/storage/lmgr/lock.c
+++ b/src/backend/storage/lmgr/lock.c
@@ -810,7 +810,7 @@ LockAcquire(const LOCKTAG *locktag,
 			bool dontWait)
 {
 	return LockAcquireExtended(locktag, lockmode, sessionLock, dontWait,
-							   true, NULL, false);
+							   true, NULL, false, false);
 }
 
 /*
@@ -829,6 +829,9 @@ LockAcquire(const LOCKTAG *locktag,
  *
  * logLockFailure indicates whether to log details when a lock acquisition
  * fails with dontWait = true.
+ *
+ * isUpgrade should be true if the backend already holds this lock in a lower
+ * mode.
  */
 LockAcquireResult
 LockAcquireExtended(const LOCKTAG *locktag,
@@ -837,7 +840,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 					bool dontWait,
 					bool reportMemoryError,
 					LOCALLOCK **locallockp,
-					bool logLockFailure)
+					bool logLockFailure,
+					bool isUpgrade)
 {
 	LOCKMETHODID lockmethodid = locktag->locktag_lockmethodid;
 	LockMethod	lockMethodTable;
@@ -1119,7 +1123,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		 * case, because JoinWaitQueue() may discover that we can acquire the
 		 * lock immediately after all.
 		 */
-		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait);
+		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait,
+								   isUpgrade);
 	}
 
 	if (waitResult == PROC_WAIT_STATUS_ERROR)
@@ -1225,8 +1230,85 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		Assert(!dontWait);
 		PROCLOCK_PRINT("LockAcquire: sleeping on lock", proclock);
 		LOCK_PRINT("LockAcquire: sleeping on lock", lock, lockmode);
+
+		/*
+		 * Lock upgrade can introduce deadlock, therefore enforce special
+		 * behavior of other processes that deal with this lock. In
+		 * particular, any waiter that sees upgradedBy set is expected to
+		 * perform deadlock check as soon as it's woken up. (Since we are
+		 * already in the queue, the deadlock detector has all the information
+		 * it needs.)
+		 */
+		if (isUpgrade)
+		{
+			dlist_iter	iter;
+
+			/*
+			 * There should not be multiple upgrades at the same time. XXX
+			 * ERROR ?
+			 */
+			Assert(lock->upgradedBy == NULL);
+
+			lock->upgradedBy = MyProc;
+
+			/*
+			 * Now, before I sleep myself, wake up all the existing waiters
+			 * (except for me), so they check for deadlock.
+			 */
+			dclist_foreach(iter, &lock->waitProcs)
+			{
+				PGPROC	   *proc = dlist_container(PGPROC, waitLink,
+												   iter.cur);
+
+				if (proc != MyProc)
+					SetLatch(&proc->procLatch);
+			}
+		}
 		LWLockRelease(partitionLock);
 
+		if (!isUpgrade)
+		{
+			HASH_SEQ_STATUS status;
+			LOCALLOCK	*llock;
+			bool	check_deadlock = false;
+
+			/*
+			 * Likewise, new waiters should check the isUpgrade flag and
+			 * engage deadlock detector if needed. The lock being acquired
+			 * should already be in the hash.
+			 *
+			 * Note that, to make deadlock detection happen as soon as
+			 * possible (i.e. regardless deadlock timeout), we check the other
+			 * locks too. XXX There might be ways to get into the deadlock
+			 * indirectly, but that needs more analysis. As long as the
+			 * upgrading process skips deadlock detection altogether, the
+			 * worst consequence of missing some lock wait cycle here is that
+			 * the other process will be kicked-off no sooner than after its
+			 * deadlock timeout has elapsed. (Which in turn means that the
+			 * lock upgrade will take longer than expected.)
+			 */
+			hash_seq_init(&status, LockMethodLocalHash);
+			while ((llock = (LOCALLOCK *) hash_seq_search(&status)) != NULL)
+			{
+				LOCK	*mylock = llock->lock;
+
+				if (mylock && mylock->upgradedBy)
+				{
+					check_deadlock = true;
+					break;
+				}
+			}
+			/* Check for deadlock if needed. */
+			if (check_deadlock)
+			{
+				DeadLockState	deadlock_state;
+
+				deadlock_state = CheckDeadLock();
+				if (deadlock_state == DS_HARD_DEADLOCK)
+					DeadLockReport();
+			}
+		}
+
 		waitResult = WaitOnLock(locallock, owner);
 
 		/*
@@ -1245,6 +1327,18 @@ LockAcquireExtended(const LOCKTAG *locktag,
 			DeadLockReport();
 			/* DeadLockReport() will not return */
 		}
+
+		/*
+		 * If finishing the lock upgrade, we should not get into a deadlock
+		 * anymore, so let others know that they do not have to care either.
+		 */
+		if (isUpgrade)
+		{
+			LWLockAcquire(partitionLock, LW_EXCLUSIVE);
+			Assert(lock->upgradedBy != NULL);
+			lock->upgradedBy = NULL;
+			LWLockRelease(partitionLock);
+		}
 	}
 	else
 		LWLockRelease(partitionLock);
@@ -1322,6 +1416,7 @@ SetupLockInTable(LockMethod lockMethodTable, PGPROC *proc,
 		lock->nGranted = 0;
 		MemSet(lock->requested, 0, sizeof(int) * MAX_LOCKMODES);
 		MemSet(lock->granted, 0, sizeof(int) * MAX_LOCKMODES);
+		lock->upgradedBy = NULL;
 		LOCK_PRINT("LockAcquire: new", lock, lockmode);
 	}
 	else
@@ -1768,6 +1863,12 @@ CleanUpLock(LOCK *lock, PROCLOCK *proclock,
 			elog(PANIC, "proclock table corrupted");
 	}
 
+	/*
+	 * Was this backend upgrading the lock?
+	 */
+	if (lock->upgradedBy == MyProc)
+		lock->upgradedBy = NULL;
+
 	if (lock->nRequested == 0)
 	{
 		/*
diff --git a/src/backend/storage/lmgr/proc.c b/src/backend/storage/lmgr/proc.c
index 1ac25068d62..7b31370db70 100644
--- a/src/backend/storage/lmgr/proc.c
+++ b/src/backend/storage/lmgr/proc.c
@@ -95,7 +95,6 @@ static volatile sig_atomic_t got_deadlock_timeout;
 static void RemoveProcFromArray(int code, Datum arg);
 static void ProcKill(int code, Datum arg);
 static void AuxiliaryProcKill(int code, Datum arg);
-static DeadLockState CheckDeadLock(void);
 
 
 /*
@@ -1143,7 +1142,8 @@ AuxiliaryPidGetProc(int pid)
  * NOTES: The process queue is now a priority queue for locking.
  */
 ProcWaitStatus
-JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
+JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait,
+			  bool isUpgrade)
 {
 	LOCKMODE	lockmode = locallock->tag.mode;
 	LOCK	   *lock = locallock->lock;
@@ -1230,8 +1230,12 @@ JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
 			/* Must he wait for me? */
 			if (lockMethodTable->conflictTab[proc->waitLockMode] & myHeldLocks)
 			{
-				/* Must I wait for him ? */
-				if (lockMethodTable->conflictTab[lockmode] & proc->heldLocks)
+				/*
+				 * Must I wait for him? I don't want a deadlock during lock
+				 * upgrade - other processes should fail on it.
+				 */
+				if ((lockMethodTable->conflictTab[lockmode] & proc->heldLocks) &&
+					!isUpgrade)
 				{
 					/*
 					 * Yes, so we have a deadlock.  Easiest way to clean up
@@ -1461,8 +1465,22 @@ ProcSleep(LOCALLOCK *locallock)
 			(void) WaitLatch(MyLatch, WL_LATCH_SET | WL_EXIT_ON_PM_DEATH, 0,
 							 PG_WAIT_LOCK | locallock->tag.lock.locktag_type);
 			ResetLatch(MyLatch);
-			/* check for deadlocks first, as that's probably log-worthy */
-			if (got_deadlock_timeout)
+			/*
+			 * Check for deadlocks first, as that's probably log-worthy.
+			 *
+			 * Do not wait for the timeout if the lock is being upgraded since
+			 * the risk of deadlock is higher now. However, do not check for
+			 * deadlock if the lock is being upgraded by this process - other
+			 * processes should take care.
+			 *
+			 * TODO Possible optimization: if this is the only lock of the
+			 * backend and if it did not have any weaker lock on the table so
+			 * far, it should be safe to skip the deadlock check. However, to
+			 * evaluate the situation, we need to take fast-path locks into
+			 * account.
+			 */
+			if ((got_deadlock_timeout || lock->upgradedBy) &&
+				lock->upgradedBy != MyProc)
 			{
 				deadlock_state = CheckDeadLock();
 				got_deadlock_timeout = false;
@@ -1819,7 +1837,7 @@ ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock)
  * not, just return.  If we have a real deadlock, remove ourselves from the
  * lock's wait queue.
  */
-static DeadLockState
+DeadLockState
 CheckDeadLock(void)
 {
 	int			i;
diff --git a/src/include/storage/lmgr.h b/src/include/storage/lmgr.h
index 2a985ce5e15..eddbe1e8622 100644
--- a/src/include/storage/lmgr.h
+++ b/src/include/storage/lmgr.h
@@ -38,6 +38,7 @@ extern void RelationInitLockInfo(Relation relation);
 
 /* Lock a relation */
 extern void LockRelationOid(Oid relid, LOCKMODE lockmode);
+extern void LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode);
 extern void LockRelationId(LockRelId *relid, LOCKMODE lockmode);
 extern bool ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode);
 extern void UnlockRelationId(LockRelId *relid, LOCKMODE lockmode);
diff --git a/src/include/storage/lock.h b/src/include/storage/lock.h
index ee3cb1dc203..080b69eddfd 100644
--- a/src/include/storage/lock.h
+++ b/src/include/storage/lock.h
@@ -131,6 +131,7 @@ typedef const LockMethodData *LockMethod;
  * nRequested -- total requested locks of all types.
  * granted -- count of each lock type currently granted on the lock.
  * nGranted -- total granted locks of all types.
+ * upgrader -- process that intends to upgrade the lock
  *
  * Note: these counts count 1 for each backend.  Internally to a backend,
  * there may be multiple grabs on a particular lock, but this is not reflected
@@ -150,6 +151,7 @@ typedef struct LOCK
 	int			nRequested;		/* total of requested[] array */
 	int			granted[MAX_LOCKMODES]; /* counts of granted locks */
 	int			nGranted;		/* total of granted[] array */
+	PGPROC		*upgradedBy;	/* is lock being upgraded by this process? */
 } LOCK;
 
 #define LOCK_LOCKMETHOD(lock) ((LOCKMETHODID) (lock).tag.locktag_lockmethodid)
@@ -390,7 +392,8 @@ extern LockAcquireResult LockAcquireExtended(const LOCKTAG *locktag,
 											 bool dontWait,
 											 bool reportMemoryError,
 											 LOCALLOCK **locallockp,
-											 bool logLockFailure);
+											 bool logLockFailure,
+											 bool isUpgrade);
 extern void AbortStrongLockAcquire(void);
 extern void MarkLockClear(LOCALLOCK *locallock);
 extern bool LockRelease(const LOCKTAG *locktag,
diff --git a/src/include/storage/proc.h b/src/include/storage/proc.h
index 3e1d1fad5f9..c4bc3a24044 100644
--- a/src/include/storage/proc.h
+++ b/src/include/storage/proc.h
@@ -563,10 +563,12 @@ extern bool HaveNFreeProcs(int n, int *nfree);
 extern void ProcReleaseLocks(bool isCommit);
 
 extern ProcWaitStatus JoinWaitQueue(LOCALLOCK *locallock,
-									LockMethod lockMethodTable, bool dontWait);
+									LockMethod lockMethodTable, bool dontWait,
+									bool isUpgrade);
 extern ProcWaitStatus ProcSleep(LOCALLOCK *locallock);
 extern void ProcWakeup(PGPROC *proc, ProcWaitStatus waitStatus);
 extern void ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock);
+extern DeadLockState CheckDeadLock(void);
 extern void CheckDeadLockAlert(void);
 extern void LockErrorCleanup(void);
 extern void GetLockHoldersAndWaiters(LOCALLOCK *locallock,
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 966+ messages in thread

* [PATCH] Lock upgrade without deadlocks.
@ 2026-04-17 13:13  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 966+ messages in thread

From: Antonin Houska @ 2026-04-17 13:13 UTC (permalink / raw)

For REPACK (CONCURRENTLY), it's essential to upgrade ShareUpdateExclusive lock
to AccessExclusiveLock for the final stage of table processing. However, we
cannot release the earlier before requesting the latter, because if another
process ran DDL on the table in between, REPACK would have to cancel its
transaction, and waste possibly a lot of work. Without releasing
ShareUpdateExclusive lock, the upgrade can cause deadlock, and it's not really
deterministic whether REPACK or the other process will get canceled.

This patch adds a new field to the LOCK structure which tells which process is
trying to upgrade the lock - this field is set right before the upgrade. All
processes waiting in the lock queue are then woken up to check the flag.

When another process tries to get the lock after that, it checks this field,
and if it's set, it checks for deadlocks even if deadlock timeout hasn't
expired yet. If the process is already in the lock's queue and sleeping, the
lock upgrading process wakes it up so it checks the flag immediately.

At the time the upgrading process starts to sleep, all the other process
should be aware that they need to check for deadlock during their lock
acquisitions, so the upgrading process does not have to check for deadlocks
when it gets woken up. Thus it should never receive deadlock error report.
---
 src/backend/commands/repack.c   |   4 +-
 src/backend/storage/lmgr/lmgr.c |  35 ++++++++---
 src/backend/storage/lmgr/lock.c | 107 +++++++++++++++++++++++++++++++-
 src/backend/storage/lmgr/proc.c |  32 +++++++---
 src/include/storage/lmgr.h      |   1 +
 src/include/storage/lock.h      |   5 +-
 src/include/storage/proc.h      |   4 +-
 7 files changed, 166 insertions(+), 22 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index 58e3867246f..310b2a65099 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -3055,8 +3055,10 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Acquire AccessExclusiveLock on the table, its TOAST relation (if there
 	 * is one), all its indexes, so that we can swap the files.
+	 *
+	 * TODO The same for indexes and TOAST?
 	 */
-	LockRelationOid(old_table_oid, AccessExclusiveLock);
+	LockRelationOidUpgrade(old_table_oid, AccessExclusiveLock);
 
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
diff --git a/src/backend/storage/lmgr/lmgr.c b/src/backend/storage/lmgr/lmgr.c
index 2ccf7237fee..0b103485d54 100644
--- a/src/backend/storage/lmgr/lmgr.c
+++ b/src/backend/storage/lmgr/lmgr.c
@@ -58,6 +58,8 @@ typedef struct XactLockTableWaitInfo
 	const ItemPointerData *ctid;
 } XactLockTableWaitInfo;
 
+static void LockRelationOidCommon(Oid relid, LOCKMODE lockmode,
+								  bool isUpgrade);
 static void XactLockTableWaitErrorCb(void *arg);
 
 /*
@@ -105,6 +107,21 @@ SetLocktagRelationOid(LOCKTAG *tag, Oid relid)
  */
 void
 LockRelationOid(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, false);
+}
+
+/*
+ * Like above, but upgrade an existing lock.
+ */
+void
+LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, true);
+}
+
+static void
+LockRelationOidCommon(Oid relid, LOCKMODE lockmode, bool isUpgrade)
 {
 	LOCKTAG		tag;
 	LOCALLOCK  *locallock;
@@ -113,7 +130,7 @@ LockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, isUpgrade);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages, so that we
@@ -157,7 +174,7 @@ ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -191,7 +208,7 @@ LockRelationId(LockRelId *relid, LOCKMODE lockmode)
 	SET_LOCKTAG_RELATION(tag, relid->dbId, relid->relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -254,7 +271,7 @@ LockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -286,7 +303,7 @@ ConditionalLockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -591,7 +608,7 @@ ConditionalLockTuple(Relation relation, const ItemPointerData *tid, LOCKMODE loc
 					  ItemPointerGetOffsetNumber(tid));
 
 	return (LockAcquireExtended(&tag, lockmode, false, true, true, NULL,
-								logLockFailure) != LOCKACQUIRE_NOT_AVAIL);
+								logLockFailure, false) != LOCKACQUIRE_NOT_AVAIL);
 }
 
 /*
@@ -749,7 +766,7 @@ ConditionalXactLockTableWait(TransactionId xid, bool logLockFailure)
 		SET_LOCKTAG_TRANSACTION(tag, xid);
 
 		if (LockAcquireExtended(&tag, ShareLock, false, true, true, NULL,
-								logLockFailure)
+								logLockFailure, false)
 			== LOCKACQUIRE_NOT_AVAIL)
 			return false;
 
@@ -1043,7 +1060,7 @@ ConditionalLockDatabaseObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -1123,7 +1140,7 @@ ConditionalLockSharedObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
diff --git a/src/backend/storage/lmgr/lock.c b/src/backend/storage/lmgr/lock.c
index c221fe96889..bcbbf36c091 100644
--- a/src/backend/storage/lmgr/lock.c
+++ b/src/backend/storage/lmgr/lock.c
@@ -810,7 +810,7 @@ LockAcquire(const LOCKTAG *locktag,
 			bool dontWait)
 {
 	return LockAcquireExtended(locktag, lockmode, sessionLock, dontWait,
-							   true, NULL, false);
+							   true, NULL, false, false);
 }
 
 /*
@@ -829,6 +829,9 @@ LockAcquire(const LOCKTAG *locktag,
  *
  * logLockFailure indicates whether to log details when a lock acquisition
  * fails with dontWait = true.
+ *
+ * isUpgrade should be true if the backend already holds this lock in a lower
+ * mode.
  */
 LockAcquireResult
 LockAcquireExtended(const LOCKTAG *locktag,
@@ -837,7 +840,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 					bool dontWait,
 					bool reportMemoryError,
 					LOCALLOCK **locallockp,
-					bool logLockFailure)
+					bool logLockFailure,
+					bool isUpgrade)
 {
 	LOCKMETHODID lockmethodid = locktag->locktag_lockmethodid;
 	LockMethod	lockMethodTable;
@@ -1119,7 +1123,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		 * case, because JoinWaitQueue() may discover that we can acquire the
 		 * lock immediately after all.
 		 */
-		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait);
+		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait,
+								   isUpgrade);
 	}
 
 	if (waitResult == PROC_WAIT_STATUS_ERROR)
@@ -1225,8 +1230,85 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		Assert(!dontWait);
 		PROCLOCK_PRINT("LockAcquire: sleeping on lock", proclock);
 		LOCK_PRINT("LockAcquire: sleeping on lock", lock, lockmode);
+
+		/*
+		 * Lock upgrade can introduce deadlock, therefore enforce special
+		 * behavior of other processes that deal with this lock. In
+		 * particular, any waiter that sees upgradedBy set is expected to
+		 * perform deadlock check as soon as it's woken up. (Since we are
+		 * already in the queue, the deadlock detector has all the information
+		 * it needs.)
+		 */
+		if (isUpgrade)
+		{
+			dlist_iter	iter;
+
+			/*
+			 * There should not be multiple upgrades at the same time. XXX
+			 * ERROR ?
+			 */
+			Assert(lock->upgradedBy == NULL);
+
+			lock->upgradedBy = MyProc;
+
+			/*
+			 * Now, before I sleep myself, wake up all the existing waiters
+			 * (except for me), so they check for deadlock.
+			 */
+			dclist_foreach(iter, &lock->waitProcs)
+			{
+				PGPROC	   *proc = dlist_container(PGPROC, waitLink,
+												   iter.cur);
+
+				if (proc != MyProc)
+					SetLatch(&proc->procLatch);
+			}
+		}
 		LWLockRelease(partitionLock);
 
+		if (!isUpgrade)
+		{
+			HASH_SEQ_STATUS status;
+			LOCALLOCK	*llock;
+			bool	check_deadlock = false;
+
+			/*
+			 * Likewise, new waiters should check the isUpgrade flag and
+			 * engage deadlock detector if needed. The lock being acquired
+			 * should already be in the hash.
+			 *
+			 * Note that, to make deadlock detection happen as soon as
+			 * possible (i.e. regardless deadlock timeout), we check the other
+			 * locks too. XXX There might be ways to get into the deadlock
+			 * indirectly, but that needs more analysis. As long as the
+			 * upgrading process skips deadlock detection altogether, the
+			 * worst consequence of missing some lock wait cycle here is that
+			 * the other process will be kicked-off no sooner than after its
+			 * deadlock timeout has elapsed. (Which in turn means that the
+			 * lock upgrade will take longer than expected.)
+			 */
+			hash_seq_init(&status, LockMethodLocalHash);
+			while ((llock = (LOCALLOCK *) hash_seq_search(&status)) != NULL)
+			{
+				LOCK	*mylock = llock->lock;
+
+				if (mylock && mylock->upgradedBy)
+				{
+					check_deadlock = true;
+					break;
+				}
+			}
+			/* Check for deadlock if needed. */
+			if (check_deadlock)
+			{
+				DeadLockState	deadlock_state;
+
+				deadlock_state = CheckDeadLock();
+				if (deadlock_state == DS_HARD_DEADLOCK)
+					DeadLockReport();
+			}
+		}
+
 		waitResult = WaitOnLock(locallock, owner);
 
 		/*
@@ -1245,6 +1327,18 @@ LockAcquireExtended(const LOCKTAG *locktag,
 			DeadLockReport();
 			/* DeadLockReport() will not return */
 		}
+
+		/*
+		 * If finishing the lock upgrade, we should not get into a deadlock
+		 * anymore, so let others know that they do not have to care either.
+		 */
+		if (isUpgrade)
+		{
+			LWLockAcquire(partitionLock, LW_EXCLUSIVE);
+			Assert(lock->upgradedBy != NULL);
+			lock->upgradedBy = NULL;
+			LWLockRelease(partitionLock);
+		}
 	}
 	else
 		LWLockRelease(partitionLock);
@@ -1322,6 +1416,7 @@ SetupLockInTable(LockMethod lockMethodTable, PGPROC *proc,
 		lock->nGranted = 0;
 		MemSet(lock->requested, 0, sizeof(int) * MAX_LOCKMODES);
 		MemSet(lock->granted, 0, sizeof(int) * MAX_LOCKMODES);
+		lock->upgradedBy = NULL;
 		LOCK_PRINT("LockAcquire: new", lock, lockmode);
 	}
 	else
@@ -1768,6 +1863,12 @@ CleanUpLock(LOCK *lock, PROCLOCK *proclock,
 			elog(PANIC, "proclock table corrupted");
 	}
 
+	/*
+	 * Was this backend upgrading the lock?
+	 */
+	if (lock->upgradedBy == MyProc)
+		lock->upgradedBy = NULL;
+
 	if (lock->nRequested == 0)
 	{
 		/*
diff --git a/src/backend/storage/lmgr/proc.c b/src/backend/storage/lmgr/proc.c
index 1ac25068d62..7b31370db70 100644
--- a/src/backend/storage/lmgr/proc.c
+++ b/src/backend/storage/lmgr/proc.c
@@ -95,7 +95,6 @@ static volatile sig_atomic_t got_deadlock_timeout;
 static void RemoveProcFromArray(int code, Datum arg);
 static void ProcKill(int code, Datum arg);
 static void AuxiliaryProcKill(int code, Datum arg);
-static DeadLockState CheckDeadLock(void);
 
 
 /*
@@ -1143,7 +1142,8 @@ AuxiliaryPidGetProc(int pid)
  * NOTES: The process queue is now a priority queue for locking.
  */
 ProcWaitStatus
-JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
+JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait,
+			  bool isUpgrade)
 {
 	LOCKMODE	lockmode = locallock->tag.mode;
 	LOCK	   *lock = locallock->lock;
@@ -1230,8 +1230,12 @@ JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
 			/* Must he wait for me? */
 			if (lockMethodTable->conflictTab[proc->waitLockMode] & myHeldLocks)
 			{
-				/* Must I wait for him ? */
-				if (lockMethodTable->conflictTab[lockmode] & proc->heldLocks)
+				/*
+				 * Must I wait for him? I don't want a deadlock during lock
+				 * upgrade - other processes should fail on it.
+				 */
+				if ((lockMethodTable->conflictTab[lockmode] & proc->heldLocks) &&
+					!isUpgrade)
 				{
 					/*
 					 * Yes, so we have a deadlock.  Easiest way to clean up
@@ -1461,8 +1465,22 @@ ProcSleep(LOCALLOCK *locallock)
 			(void) WaitLatch(MyLatch, WL_LATCH_SET | WL_EXIT_ON_PM_DEATH, 0,
 							 PG_WAIT_LOCK | locallock->tag.lock.locktag_type);
 			ResetLatch(MyLatch);
-			/* check for deadlocks first, as that's probably log-worthy */
-			if (got_deadlock_timeout)
+			/*
+			 * Check for deadlocks first, as that's probably log-worthy.
+			 *
+			 * Do not wait for the timeout if the lock is being upgraded since
+			 * the risk of deadlock is higher now. However, do not check for
+			 * deadlock if the lock is being upgraded by this process - other
+			 * processes should take care.
+			 *
+			 * TODO Possible optimization: if this is the only lock of the
+			 * backend and if it did not have any weaker lock on the table so
+			 * far, it should be safe to skip the deadlock check. However, to
+			 * evaluate the situation, we need to take fast-path locks into
+			 * account.
+			 */
+			if ((got_deadlock_timeout || lock->upgradedBy) &&
+				lock->upgradedBy != MyProc)
 			{
 				deadlock_state = CheckDeadLock();
 				got_deadlock_timeout = false;
@@ -1819,7 +1837,7 @@ ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock)
  * not, just return.  If we have a real deadlock, remove ourselves from the
  * lock's wait queue.
  */
-static DeadLockState
+DeadLockState
 CheckDeadLock(void)
 {
 	int			i;
diff --git a/src/include/storage/lmgr.h b/src/include/storage/lmgr.h
index 2a985ce5e15..eddbe1e8622 100644
--- a/src/include/storage/lmgr.h
+++ b/src/include/storage/lmgr.h
@@ -38,6 +38,7 @@ extern void RelationInitLockInfo(Relation relation);
 
 /* Lock a relation */
 extern void LockRelationOid(Oid relid, LOCKMODE lockmode);
+extern void LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode);
 extern void LockRelationId(LockRelId *relid, LOCKMODE lockmode);
 extern bool ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode);
 extern void UnlockRelationId(LockRelId *relid, LOCKMODE lockmode);
diff --git a/src/include/storage/lock.h b/src/include/storage/lock.h
index ee3cb1dc203..080b69eddfd 100644
--- a/src/include/storage/lock.h
+++ b/src/include/storage/lock.h
@@ -131,6 +131,7 @@ typedef const LockMethodData *LockMethod;
  * nRequested -- total requested locks of all types.
  * granted -- count of each lock type currently granted on the lock.
  * nGranted -- total granted locks of all types.
+ * upgrader -- process that intends to upgrade the lock
  *
  * Note: these counts count 1 for each backend.  Internally to a backend,
  * there may be multiple grabs on a particular lock, but this is not reflected
@@ -150,6 +151,7 @@ typedef struct LOCK
 	int			nRequested;		/* total of requested[] array */
 	int			granted[MAX_LOCKMODES]; /* counts of granted locks */
 	int			nGranted;		/* total of granted[] array */
+	PGPROC		*upgradedBy;	/* is lock being upgraded by this process? */
 } LOCK;
 
 #define LOCK_LOCKMETHOD(lock) ((LOCKMETHODID) (lock).tag.locktag_lockmethodid)
@@ -390,7 +392,8 @@ extern LockAcquireResult LockAcquireExtended(const LOCKTAG *locktag,
 											 bool dontWait,
 											 bool reportMemoryError,
 											 LOCALLOCK **locallockp,
-											 bool logLockFailure);
+											 bool logLockFailure,
+											 bool isUpgrade);
 extern void AbortStrongLockAcquire(void);
 extern void MarkLockClear(LOCALLOCK *locallock);
 extern bool LockRelease(const LOCKTAG *locktag,
diff --git a/src/include/storage/proc.h b/src/include/storage/proc.h
index 3e1d1fad5f9..c4bc3a24044 100644
--- a/src/include/storage/proc.h
+++ b/src/include/storage/proc.h
@@ -563,10 +563,12 @@ extern bool HaveNFreeProcs(int n, int *nfree);
 extern void ProcReleaseLocks(bool isCommit);
 
 extern ProcWaitStatus JoinWaitQueue(LOCALLOCK *locallock,
-									LockMethod lockMethodTable, bool dontWait);
+									LockMethod lockMethodTable, bool dontWait,
+									bool isUpgrade);
 extern ProcWaitStatus ProcSleep(LOCALLOCK *locallock);
 extern void ProcWakeup(PGPROC *proc, ProcWaitStatus waitStatus);
 extern void ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock);
+extern DeadLockState CheckDeadLock(void);
 extern void CheckDeadLockAlert(void);
 extern void LockErrorCleanup(void);
 extern void GetLockHoldersAndWaiters(LOCALLOCK *locallock,
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 966+ messages in thread

* [PATCH] Lock upgrade without deadlocks.
@ 2026-04-17 13:13  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 966+ messages in thread

From: Antonin Houska @ 2026-04-17 13:13 UTC (permalink / raw)

For REPACK (CONCURRENTLY), it's essential to upgrade ShareUpdateExclusive lock
to AccessExclusiveLock for the final stage of table processing. However, we
cannot release the earlier before requesting the latter, because if another
process ran DDL on the table in between, REPACK would have to cancel its
transaction, and waste possibly a lot of work. Without releasing
ShareUpdateExclusive lock, the upgrade can cause deadlock, and it's not really
deterministic whether REPACK or the other process will get canceled.

This patch adds a new field to the LOCK structure which tells which process is
trying to upgrade the lock - this field is set right before the upgrade. All
processes waiting in the lock queue are then woken up to check the flag.

When another process tries to get the lock after that, it checks this field,
and if it's set, it checks for deadlocks even if deadlock timeout hasn't
expired yet. If the process is already in the lock's queue and sleeping, the
lock upgrading process wakes it up so it checks the flag immediately.

At the time the upgrading process starts to sleep, all the other process
should be aware that they need to check for deadlock during their lock
acquisitions, so the upgrading process does not have to check for deadlocks
when it gets woken up. Thus it should never receive deadlock error report.
---
 src/backend/commands/repack.c   |   4 +-
 src/backend/storage/lmgr/lmgr.c |  35 ++++++++---
 src/backend/storage/lmgr/lock.c | 107 +++++++++++++++++++++++++++++++-
 src/backend/storage/lmgr/proc.c |  32 +++++++---
 src/include/storage/lmgr.h      |   1 +
 src/include/storage/lock.h      |   5 +-
 src/include/storage/proc.h      |   4 +-
 7 files changed, 166 insertions(+), 22 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index 58e3867246f..310b2a65099 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -3055,8 +3055,10 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Acquire AccessExclusiveLock on the table, its TOAST relation (if there
 	 * is one), all its indexes, so that we can swap the files.
+	 *
+	 * TODO The same for indexes and TOAST?
 	 */
-	LockRelationOid(old_table_oid, AccessExclusiveLock);
+	LockRelationOidUpgrade(old_table_oid, AccessExclusiveLock);
 
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
diff --git a/src/backend/storage/lmgr/lmgr.c b/src/backend/storage/lmgr/lmgr.c
index 2ccf7237fee..0b103485d54 100644
--- a/src/backend/storage/lmgr/lmgr.c
+++ b/src/backend/storage/lmgr/lmgr.c
@@ -58,6 +58,8 @@ typedef struct XactLockTableWaitInfo
 	const ItemPointerData *ctid;
 } XactLockTableWaitInfo;
 
+static void LockRelationOidCommon(Oid relid, LOCKMODE lockmode,
+								  bool isUpgrade);
 static void XactLockTableWaitErrorCb(void *arg);
 
 /*
@@ -105,6 +107,21 @@ SetLocktagRelationOid(LOCKTAG *tag, Oid relid)
  */
 void
 LockRelationOid(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, false);
+}
+
+/*
+ * Like above, but upgrade an existing lock.
+ */
+void
+LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, true);
+}
+
+static void
+LockRelationOidCommon(Oid relid, LOCKMODE lockmode, bool isUpgrade)
 {
 	LOCKTAG		tag;
 	LOCALLOCK  *locallock;
@@ -113,7 +130,7 @@ LockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, isUpgrade);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages, so that we
@@ -157,7 +174,7 @@ ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -191,7 +208,7 @@ LockRelationId(LockRelId *relid, LOCKMODE lockmode)
 	SET_LOCKTAG_RELATION(tag, relid->dbId, relid->relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -254,7 +271,7 @@ LockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -286,7 +303,7 @@ ConditionalLockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -591,7 +608,7 @@ ConditionalLockTuple(Relation relation, const ItemPointerData *tid, LOCKMODE loc
 					  ItemPointerGetOffsetNumber(tid));
 
 	return (LockAcquireExtended(&tag, lockmode, false, true, true, NULL,
-								logLockFailure) != LOCKACQUIRE_NOT_AVAIL);
+								logLockFailure, false) != LOCKACQUIRE_NOT_AVAIL);
 }
 
 /*
@@ -749,7 +766,7 @@ ConditionalXactLockTableWait(TransactionId xid, bool logLockFailure)
 		SET_LOCKTAG_TRANSACTION(tag, xid);
 
 		if (LockAcquireExtended(&tag, ShareLock, false, true, true, NULL,
-								logLockFailure)
+								logLockFailure, false)
 			== LOCKACQUIRE_NOT_AVAIL)
 			return false;
 
@@ -1043,7 +1060,7 @@ ConditionalLockDatabaseObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -1123,7 +1140,7 @@ ConditionalLockSharedObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
diff --git a/src/backend/storage/lmgr/lock.c b/src/backend/storage/lmgr/lock.c
index c221fe96889..bcbbf36c091 100644
--- a/src/backend/storage/lmgr/lock.c
+++ b/src/backend/storage/lmgr/lock.c
@@ -810,7 +810,7 @@ LockAcquire(const LOCKTAG *locktag,
 			bool dontWait)
 {
 	return LockAcquireExtended(locktag, lockmode, sessionLock, dontWait,
-							   true, NULL, false);
+							   true, NULL, false, false);
 }
 
 /*
@@ -829,6 +829,9 @@ LockAcquire(const LOCKTAG *locktag,
  *
  * logLockFailure indicates whether to log details when a lock acquisition
  * fails with dontWait = true.
+ *
+ * isUpgrade should be true if the backend already holds this lock in a lower
+ * mode.
  */
 LockAcquireResult
 LockAcquireExtended(const LOCKTAG *locktag,
@@ -837,7 +840,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 					bool dontWait,
 					bool reportMemoryError,
 					LOCALLOCK **locallockp,
-					bool logLockFailure)
+					bool logLockFailure,
+					bool isUpgrade)
 {
 	LOCKMETHODID lockmethodid = locktag->locktag_lockmethodid;
 	LockMethod	lockMethodTable;
@@ -1119,7 +1123,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		 * case, because JoinWaitQueue() may discover that we can acquire the
 		 * lock immediately after all.
 		 */
-		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait);
+		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait,
+								   isUpgrade);
 	}
 
 	if (waitResult == PROC_WAIT_STATUS_ERROR)
@@ -1225,8 +1230,85 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		Assert(!dontWait);
 		PROCLOCK_PRINT("LockAcquire: sleeping on lock", proclock);
 		LOCK_PRINT("LockAcquire: sleeping on lock", lock, lockmode);
+
+		/*
+		 * Lock upgrade can introduce deadlock, therefore enforce special
+		 * behavior of other processes that deal with this lock. In
+		 * particular, any waiter that sees upgradedBy set is expected to
+		 * perform deadlock check as soon as it's woken up. (Since we are
+		 * already in the queue, the deadlock detector has all the information
+		 * it needs.)
+		 */
+		if (isUpgrade)
+		{
+			dlist_iter	iter;
+
+			/*
+			 * There should not be multiple upgrades at the same time. XXX
+			 * ERROR ?
+			 */
+			Assert(lock->upgradedBy == NULL);
+
+			lock->upgradedBy = MyProc;
+
+			/*
+			 * Now, before I sleep myself, wake up all the existing waiters
+			 * (except for me), so they check for deadlock.
+			 */
+			dclist_foreach(iter, &lock->waitProcs)
+			{
+				PGPROC	   *proc = dlist_container(PGPROC, waitLink,
+												   iter.cur);
+
+				if (proc != MyProc)
+					SetLatch(&proc->procLatch);
+			}
+		}
 		LWLockRelease(partitionLock);
 
+		if (!isUpgrade)
+		{
+			HASH_SEQ_STATUS status;
+			LOCALLOCK	*llock;
+			bool	check_deadlock = false;
+
+			/*
+			 * Likewise, new waiters should check the isUpgrade flag and
+			 * engage deadlock detector if needed. The lock being acquired
+			 * should already be in the hash.
+			 *
+			 * Note that, to make deadlock detection happen as soon as
+			 * possible (i.e. regardless deadlock timeout), we check the other
+			 * locks too. XXX There might be ways to get into the deadlock
+			 * indirectly, but that needs more analysis. As long as the
+			 * upgrading process skips deadlock detection altogether, the
+			 * worst consequence of missing some lock wait cycle here is that
+			 * the other process will be kicked-off no sooner than after its
+			 * deadlock timeout has elapsed. (Which in turn means that the
+			 * lock upgrade will take longer than expected.)
+			 */
+			hash_seq_init(&status, LockMethodLocalHash);
+			while ((llock = (LOCALLOCK *) hash_seq_search(&status)) != NULL)
+			{
+				LOCK	*mylock = llock->lock;
+
+				if (mylock && mylock->upgradedBy)
+				{
+					check_deadlock = true;
+					break;
+				}
+			}
+			/* Check for deadlock if needed. */
+			if (check_deadlock)
+			{
+				DeadLockState	deadlock_state;
+
+				deadlock_state = CheckDeadLock();
+				if (deadlock_state == DS_HARD_DEADLOCK)
+					DeadLockReport();
+			}
+		}
+
 		waitResult = WaitOnLock(locallock, owner);
 
 		/*
@@ -1245,6 +1327,18 @@ LockAcquireExtended(const LOCKTAG *locktag,
 			DeadLockReport();
 			/* DeadLockReport() will not return */
 		}
+
+		/*
+		 * If finishing the lock upgrade, we should not get into a deadlock
+		 * anymore, so let others know that they do not have to care either.
+		 */
+		if (isUpgrade)
+		{
+			LWLockAcquire(partitionLock, LW_EXCLUSIVE);
+			Assert(lock->upgradedBy != NULL);
+			lock->upgradedBy = NULL;
+			LWLockRelease(partitionLock);
+		}
 	}
 	else
 		LWLockRelease(partitionLock);
@@ -1322,6 +1416,7 @@ SetupLockInTable(LockMethod lockMethodTable, PGPROC *proc,
 		lock->nGranted = 0;
 		MemSet(lock->requested, 0, sizeof(int) * MAX_LOCKMODES);
 		MemSet(lock->granted, 0, sizeof(int) * MAX_LOCKMODES);
+		lock->upgradedBy = NULL;
 		LOCK_PRINT("LockAcquire: new", lock, lockmode);
 	}
 	else
@@ -1768,6 +1863,12 @@ CleanUpLock(LOCK *lock, PROCLOCK *proclock,
 			elog(PANIC, "proclock table corrupted");
 	}
 
+	/*
+	 * Was this backend upgrading the lock?
+	 */
+	if (lock->upgradedBy == MyProc)
+		lock->upgradedBy = NULL;
+
 	if (lock->nRequested == 0)
 	{
 		/*
diff --git a/src/backend/storage/lmgr/proc.c b/src/backend/storage/lmgr/proc.c
index 1ac25068d62..7b31370db70 100644
--- a/src/backend/storage/lmgr/proc.c
+++ b/src/backend/storage/lmgr/proc.c
@@ -95,7 +95,6 @@ static volatile sig_atomic_t got_deadlock_timeout;
 static void RemoveProcFromArray(int code, Datum arg);
 static void ProcKill(int code, Datum arg);
 static void AuxiliaryProcKill(int code, Datum arg);
-static DeadLockState CheckDeadLock(void);
 
 
 /*
@@ -1143,7 +1142,8 @@ AuxiliaryPidGetProc(int pid)
  * NOTES: The process queue is now a priority queue for locking.
  */
 ProcWaitStatus
-JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
+JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait,
+			  bool isUpgrade)
 {
 	LOCKMODE	lockmode = locallock->tag.mode;
 	LOCK	   *lock = locallock->lock;
@@ -1230,8 +1230,12 @@ JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
 			/* Must he wait for me? */
 			if (lockMethodTable->conflictTab[proc->waitLockMode] & myHeldLocks)
 			{
-				/* Must I wait for him ? */
-				if (lockMethodTable->conflictTab[lockmode] & proc->heldLocks)
+				/*
+				 * Must I wait for him? I don't want a deadlock during lock
+				 * upgrade - other processes should fail on it.
+				 */
+				if ((lockMethodTable->conflictTab[lockmode] & proc->heldLocks) &&
+					!isUpgrade)
 				{
 					/*
 					 * Yes, so we have a deadlock.  Easiest way to clean up
@@ -1461,8 +1465,22 @@ ProcSleep(LOCALLOCK *locallock)
 			(void) WaitLatch(MyLatch, WL_LATCH_SET | WL_EXIT_ON_PM_DEATH, 0,
 							 PG_WAIT_LOCK | locallock->tag.lock.locktag_type);
 			ResetLatch(MyLatch);
-			/* check for deadlocks first, as that's probably log-worthy */
-			if (got_deadlock_timeout)
+			/*
+			 * Check for deadlocks first, as that's probably log-worthy.
+			 *
+			 * Do not wait for the timeout if the lock is being upgraded since
+			 * the risk of deadlock is higher now. However, do not check for
+			 * deadlock if the lock is being upgraded by this process - other
+			 * processes should take care.
+			 *
+			 * TODO Possible optimization: if this is the only lock of the
+			 * backend and if it did not have any weaker lock on the table so
+			 * far, it should be safe to skip the deadlock check. However, to
+			 * evaluate the situation, we need to take fast-path locks into
+			 * account.
+			 */
+			if ((got_deadlock_timeout || lock->upgradedBy) &&
+				lock->upgradedBy != MyProc)
 			{
 				deadlock_state = CheckDeadLock();
 				got_deadlock_timeout = false;
@@ -1819,7 +1837,7 @@ ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock)
  * not, just return.  If we have a real deadlock, remove ourselves from the
  * lock's wait queue.
  */
-static DeadLockState
+DeadLockState
 CheckDeadLock(void)
 {
 	int			i;
diff --git a/src/include/storage/lmgr.h b/src/include/storage/lmgr.h
index 2a985ce5e15..eddbe1e8622 100644
--- a/src/include/storage/lmgr.h
+++ b/src/include/storage/lmgr.h
@@ -38,6 +38,7 @@ extern void RelationInitLockInfo(Relation relation);
 
 /* Lock a relation */
 extern void LockRelationOid(Oid relid, LOCKMODE lockmode);
+extern void LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode);
 extern void LockRelationId(LockRelId *relid, LOCKMODE lockmode);
 extern bool ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode);
 extern void UnlockRelationId(LockRelId *relid, LOCKMODE lockmode);
diff --git a/src/include/storage/lock.h b/src/include/storage/lock.h
index ee3cb1dc203..080b69eddfd 100644
--- a/src/include/storage/lock.h
+++ b/src/include/storage/lock.h
@@ -131,6 +131,7 @@ typedef const LockMethodData *LockMethod;
  * nRequested -- total requested locks of all types.
  * granted -- count of each lock type currently granted on the lock.
  * nGranted -- total granted locks of all types.
+ * upgrader -- process that intends to upgrade the lock
  *
  * Note: these counts count 1 for each backend.  Internally to a backend,
  * there may be multiple grabs on a particular lock, but this is not reflected
@@ -150,6 +151,7 @@ typedef struct LOCK
 	int			nRequested;		/* total of requested[] array */
 	int			granted[MAX_LOCKMODES]; /* counts of granted locks */
 	int			nGranted;		/* total of granted[] array */
+	PGPROC		*upgradedBy;	/* is lock being upgraded by this process? */
 } LOCK;
 
 #define LOCK_LOCKMETHOD(lock) ((LOCKMETHODID) (lock).tag.locktag_lockmethodid)
@@ -390,7 +392,8 @@ extern LockAcquireResult LockAcquireExtended(const LOCKTAG *locktag,
 											 bool dontWait,
 											 bool reportMemoryError,
 											 LOCALLOCK **locallockp,
-											 bool logLockFailure);
+											 bool logLockFailure,
+											 bool isUpgrade);
 extern void AbortStrongLockAcquire(void);
 extern void MarkLockClear(LOCALLOCK *locallock);
 extern bool LockRelease(const LOCKTAG *locktag,
diff --git a/src/include/storage/proc.h b/src/include/storage/proc.h
index 3e1d1fad5f9..c4bc3a24044 100644
--- a/src/include/storage/proc.h
+++ b/src/include/storage/proc.h
@@ -563,10 +563,12 @@ extern bool HaveNFreeProcs(int n, int *nfree);
 extern void ProcReleaseLocks(bool isCommit);
 
 extern ProcWaitStatus JoinWaitQueue(LOCALLOCK *locallock,
-									LockMethod lockMethodTable, bool dontWait);
+									LockMethod lockMethodTable, bool dontWait,
+									bool isUpgrade);
 extern ProcWaitStatus ProcSleep(LOCALLOCK *locallock);
 extern void ProcWakeup(PGPROC *proc, ProcWaitStatus waitStatus);
 extern void ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock);
+extern DeadLockState CheckDeadLock(void);
 extern void CheckDeadLockAlert(void);
 extern void LockErrorCleanup(void);
 extern void GetLockHoldersAndWaiters(LOCALLOCK *locallock,
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 966+ messages in thread

* [PATCH] Lock upgrade without deadlocks.
@ 2026-04-17 13:13  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 966+ messages in thread

From: Antonin Houska @ 2026-04-17 13:13 UTC (permalink / raw)

For REPACK (CONCURRENTLY), it's essential to upgrade ShareUpdateExclusive lock
to AccessExclusiveLock for the final stage of table processing. However, we
cannot release the earlier before requesting the latter, because if another
process ran DDL on the table in between, REPACK would have to cancel its
transaction, and waste possibly a lot of work. Without releasing
ShareUpdateExclusive lock, the upgrade can cause deadlock, and it's not really
deterministic whether REPACK or the other process will get canceled.

This patch adds a new field to the LOCK structure which tells which process is
trying to upgrade the lock - this field is set right before the upgrade. All
processes waiting in the lock queue are then woken up to check the flag.

When another process tries to get the lock after that, it checks this field,
and if it's set, it checks for deadlocks even if deadlock timeout hasn't
expired yet. If the process is already in the lock's queue and sleeping, the
lock upgrading process wakes it up so it checks the flag immediately.

At the time the upgrading process starts to sleep, all the other process
should be aware that they need to check for deadlock during their lock
acquisitions, so the upgrading process does not have to check for deadlocks
when it gets woken up. Thus it should never receive deadlock error report.
---
 src/backend/commands/repack.c   |   4 +-
 src/backend/storage/lmgr/lmgr.c |  35 ++++++++---
 src/backend/storage/lmgr/lock.c | 107 +++++++++++++++++++++++++++++++-
 src/backend/storage/lmgr/proc.c |  32 +++++++---
 src/include/storage/lmgr.h      |   1 +
 src/include/storage/lock.h      |   5 +-
 src/include/storage/proc.h      |   4 +-
 7 files changed, 166 insertions(+), 22 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index 58e3867246f..310b2a65099 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -3055,8 +3055,10 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Acquire AccessExclusiveLock on the table, its TOAST relation (if there
 	 * is one), all its indexes, so that we can swap the files.
+	 *
+	 * TODO The same for indexes and TOAST?
 	 */
-	LockRelationOid(old_table_oid, AccessExclusiveLock);
+	LockRelationOidUpgrade(old_table_oid, AccessExclusiveLock);
 
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
diff --git a/src/backend/storage/lmgr/lmgr.c b/src/backend/storage/lmgr/lmgr.c
index 2ccf7237fee..0b103485d54 100644
--- a/src/backend/storage/lmgr/lmgr.c
+++ b/src/backend/storage/lmgr/lmgr.c
@@ -58,6 +58,8 @@ typedef struct XactLockTableWaitInfo
 	const ItemPointerData *ctid;
 } XactLockTableWaitInfo;
 
+static void LockRelationOidCommon(Oid relid, LOCKMODE lockmode,
+								  bool isUpgrade);
 static void XactLockTableWaitErrorCb(void *arg);
 
 /*
@@ -105,6 +107,21 @@ SetLocktagRelationOid(LOCKTAG *tag, Oid relid)
  */
 void
 LockRelationOid(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, false);
+}
+
+/*
+ * Like above, but upgrade an existing lock.
+ */
+void
+LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, true);
+}
+
+static void
+LockRelationOidCommon(Oid relid, LOCKMODE lockmode, bool isUpgrade)
 {
 	LOCKTAG		tag;
 	LOCALLOCK  *locallock;
@@ -113,7 +130,7 @@ LockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, isUpgrade);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages, so that we
@@ -157,7 +174,7 @@ ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -191,7 +208,7 @@ LockRelationId(LockRelId *relid, LOCKMODE lockmode)
 	SET_LOCKTAG_RELATION(tag, relid->dbId, relid->relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -254,7 +271,7 @@ LockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -286,7 +303,7 @@ ConditionalLockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -591,7 +608,7 @@ ConditionalLockTuple(Relation relation, const ItemPointerData *tid, LOCKMODE loc
 					  ItemPointerGetOffsetNumber(tid));
 
 	return (LockAcquireExtended(&tag, lockmode, false, true, true, NULL,
-								logLockFailure) != LOCKACQUIRE_NOT_AVAIL);
+								logLockFailure, false) != LOCKACQUIRE_NOT_AVAIL);
 }
 
 /*
@@ -749,7 +766,7 @@ ConditionalXactLockTableWait(TransactionId xid, bool logLockFailure)
 		SET_LOCKTAG_TRANSACTION(tag, xid);
 
 		if (LockAcquireExtended(&tag, ShareLock, false, true, true, NULL,
-								logLockFailure)
+								logLockFailure, false)
 			== LOCKACQUIRE_NOT_AVAIL)
 			return false;
 
@@ -1043,7 +1060,7 @@ ConditionalLockDatabaseObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -1123,7 +1140,7 @@ ConditionalLockSharedObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
diff --git a/src/backend/storage/lmgr/lock.c b/src/backend/storage/lmgr/lock.c
index c221fe96889..bcbbf36c091 100644
--- a/src/backend/storage/lmgr/lock.c
+++ b/src/backend/storage/lmgr/lock.c
@@ -810,7 +810,7 @@ LockAcquire(const LOCKTAG *locktag,
 			bool dontWait)
 {
 	return LockAcquireExtended(locktag, lockmode, sessionLock, dontWait,
-							   true, NULL, false);
+							   true, NULL, false, false);
 }
 
 /*
@@ -829,6 +829,9 @@ LockAcquire(const LOCKTAG *locktag,
  *
  * logLockFailure indicates whether to log details when a lock acquisition
  * fails with dontWait = true.
+ *
+ * isUpgrade should be true if the backend already holds this lock in a lower
+ * mode.
  */
 LockAcquireResult
 LockAcquireExtended(const LOCKTAG *locktag,
@@ -837,7 +840,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 					bool dontWait,
 					bool reportMemoryError,
 					LOCALLOCK **locallockp,
-					bool logLockFailure)
+					bool logLockFailure,
+					bool isUpgrade)
 {
 	LOCKMETHODID lockmethodid = locktag->locktag_lockmethodid;
 	LockMethod	lockMethodTable;
@@ -1119,7 +1123,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		 * case, because JoinWaitQueue() may discover that we can acquire the
 		 * lock immediately after all.
 		 */
-		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait);
+		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait,
+								   isUpgrade);
 	}
 
 	if (waitResult == PROC_WAIT_STATUS_ERROR)
@@ -1225,8 +1230,85 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		Assert(!dontWait);
 		PROCLOCK_PRINT("LockAcquire: sleeping on lock", proclock);
 		LOCK_PRINT("LockAcquire: sleeping on lock", lock, lockmode);
+
+		/*
+		 * Lock upgrade can introduce deadlock, therefore enforce special
+		 * behavior of other processes that deal with this lock. In
+		 * particular, any waiter that sees upgradedBy set is expected to
+		 * perform deadlock check as soon as it's woken up. (Since we are
+		 * already in the queue, the deadlock detector has all the information
+		 * it needs.)
+		 */
+		if (isUpgrade)
+		{
+			dlist_iter	iter;
+
+			/*
+			 * There should not be multiple upgrades at the same time. XXX
+			 * ERROR ?
+			 */
+			Assert(lock->upgradedBy == NULL);
+
+			lock->upgradedBy = MyProc;
+
+			/*
+			 * Now, before I sleep myself, wake up all the existing waiters
+			 * (except for me), so they check for deadlock.
+			 */
+			dclist_foreach(iter, &lock->waitProcs)
+			{
+				PGPROC	   *proc = dlist_container(PGPROC, waitLink,
+												   iter.cur);
+
+				if (proc != MyProc)
+					SetLatch(&proc->procLatch);
+			}
+		}
 		LWLockRelease(partitionLock);
 
+		if (!isUpgrade)
+		{
+			HASH_SEQ_STATUS status;
+			LOCALLOCK	*llock;
+			bool	check_deadlock = false;
+
+			/*
+			 * Likewise, new waiters should check the isUpgrade flag and
+			 * engage deadlock detector if needed. The lock being acquired
+			 * should already be in the hash.
+			 *
+			 * Note that, to make deadlock detection happen as soon as
+			 * possible (i.e. regardless deadlock timeout), we check the other
+			 * locks too. XXX There might be ways to get into the deadlock
+			 * indirectly, but that needs more analysis. As long as the
+			 * upgrading process skips deadlock detection altogether, the
+			 * worst consequence of missing some lock wait cycle here is that
+			 * the other process will be kicked-off no sooner than after its
+			 * deadlock timeout has elapsed. (Which in turn means that the
+			 * lock upgrade will take longer than expected.)
+			 */
+			hash_seq_init(&status, LockMethodLocalHash);
+			while ((llock = (LOCALLOCK *) hash_seq_search(&status)) != NULL)
+			{
+				LOCK	*mylock = llock->lock;
+
+				if (mylock && mylock->upgradedBy)
+				{
+					check_deadlock = true;
+					break;
+				}
+			}
+			/* Check for deadlock if needed. */
+			if (check_deadlock)
+			{
+				DeadLockState	deadlock_state;
+
+				deadlock_state = CheckDeadLock();
+				if (deadlock_state == DS_HARD_DEADLOCK)
+					DeadLockReport();
+			}
+		}
+
 		waitResult = WaitOnLock(locallock, owner);
 
 		/*
@@ -1245,6 +1327,18 @@ LockAcquireExtended(const LOCKTAG *locktag,
 			DeadLockReport();
 			/* DeadLockReport() will not return */
 		}
+
+		/*
+		 * If finishing the lock upgrade, we should not get into a deadlock
+		 * anymore, so let others know that they do not have to care either.
+		 */
+		if (isUpgrade)
+		{
+			LWLockAcquire(partitionLock, LW_EXCLUSIVE);
+			Assert(lock->upgradedBy != NULL);
+			lock->upgradedBy = NULL;
+			LWLockRelease(partitionLock);
+		}
 	}
 	else
 		LWLockRelease(partitionLock);
@@ -1322,6 +1416,7 @@ SetupLockInTable(LockMethod lockMethodTable, PGPROC *proc,
 		lock->nGranted = 0;
 		MemSet(lock->requested, 0, sizeof(int) * MAX_LOCKMODES);
 		MemSet(lock->granted, 0, sizeof(int) * MAX_LOCKMODES);
+		lock->upgradedBy = NULL;
 		LOCK_PRINT("LockAcquire: new", lock, lockmode);
 	}
 	else
@@ -1768,6 +1863,12 @@ CleanUpLock(LOCK *lock, PROCLOCK *proclock,
 			elog(PANIC, "proclock table corrupted");
 	}
 
+	/*
+	 * Was this backend upgrading the lock?
+	 */
+	if (lock->upgradedBy == MyProc)
+		lock->upgradedBy = NULL;
+
 	if (lock->nRequested == 0)
 	{
 		/*
diff --git a/src/backend/storage/lmgr/proc.c b/src/backend/storage/lmgr/proc.c
index 1ac25068d62..7b31370db70 100644
--- a/src/backend/storage/lmgr/proc.c
+++ b/src/backend/storage/lmgr/proc.c
@@ -95,7 +95,6 @@ static volatile sig_atomic_t got_deadlock_timeout;
 static void RemoveProcFromArray(int code, Datum arg);
 static void ProcKill(int code, Datum arg);
 static void AuxiliaryProcKill(int code, Datum arg);
-static DeadLockState CheckDeadLock(void);
 
 
 /*
@@ -1143,7 +1142,8 @@ AuxiliaryPidGetProc(int pid)
  * NOTES: The process queue is now a priority queue for locking.
  */
 ProcWaitStatus
-JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
+JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait,
+			  bool isUpgrade)
 {
 	LOCKMODE	lockmode = locallock->tag.mode;
 	LOCK	   *lock = locallock->lock;
@@ -1230,8 +1230,12 @@ JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
 			/* Must he wait for me? */
 			if (lockMethodTable->conflictTab[proc->waitLockMode] & myHeldLocks)
 			{
-				/* Must I wait for him ? */
-				if (lockMethodTable->conflictTab[lockmode] & proc->heldLocks)
+				/*
+				 * Must I wait for him? I don't want a deadlock during lock
+				 * upgrade - other processes should fail on it.
+				 */
+				if ((lockMethodTable->conflictTab[lockmode] & proc->heldLocks) &&
+					!isUpgrade)
 				{
 					/*
 					 * Yes, so we have a deadlock.  Easiest way to clean up
@@ -1461,8 +1465,22 @@ ProcSleep(LOCALLOCK *locallock)
 			(void) WaitLatch(MyLatch, WL_LATCH_SET | WL_EXIT_ON_PM_DEATH, 0,
 							 PG_WAIT_LOCK | locallock->tag.lock.locktag_type);
 			ResetLatch(MyLatch);
-			/* check for deadlocks first, as that's probably log-worthy */
-			if (got_deadlock_timeout)
+			/*
+			 * Check for deadlocks first, as that's probably log-worthy.
+			 *
+			 * Do not wait for the timeout if the lock is being upgraded since
+			 * the risk of deadlock is higher now. However, do not check for
+			 * deadlock if the lock is being upgraded by this process - other
+			 * processes should take care.
+			 *
+			 * TODO Possible optimization: if this is the only lock of the
+			 * backend and if it did not have any weaker lock on the table so
+			 * far, it should be safe to skip the deadlock check. However, to
+			 * evaluate the situation, we need to take fast-path locks into
+			 * account.
+			 */
+			if ((got_deadlock_timeout || lock->upgradedBy) &&
+				lock->upgradedBy != MyProc)
 			{
 				deadlock_state = CheckDeadLock();
 				got_deadlock_timeout = false;
@@ -1819,7 +1837,7 @@ ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock)
  * not, just return.  If we have a real deadlock, remove ourselves from the
  * lock's wait queue.
  */
-static DeadLockState
+DeadLockState
 CheckDeadLock(void)
 {
 	int			i;
diff --git a/src/include/storage/lmgr.h b/src/include/storage/lmgr.h
index 2a985ce5e15..eddbe1e8622 100644
--- a/src/include/storage/lmgr.h
+++ b/src/include/storage/lmgr.h
@@ -38,6 +38,7 @@ extern void RelationInitLockInfo(Relation relation);
 
 /* Lock a relation */
 extern void LockRelationOid(Oid relid, LOCKMODE lockmode);
+extern void LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode);
 extern void LockRelationId(LockRelId *relid, LOCKMODE lockmode);
 extern bool ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode);
 extern void UnlockRelationId(LockRelId *relid, LOCKMODE lockmode);
diff --git a/src/include/storage/lock.h b/src/include/storage/lock.h
index ee3cb1dc203..080b69eddfd 100644
--- a/src/include/storage/lock.h
+++ b/src/include/storage/lock.h
@@ -131,6 +131,7 @@ typedef const LockMethodData *LockMethod;
  * nRequested -- total requested locks of all types.
  * granted -- count of each lock type currently granted on the lock.
  * nGranted -- total granted locks of all types.
+ * upgrader -- process that intends to upgrade the lock
  *
  * Note: these counts count 1 for each backend.  Internally to a backend,
  * there may be multiple grabs on a particular lock, but this is not reflected
@@ -150,6 +151,7 @@ typedef struct LOCK
 	int			nRequested;		/* total of requested[] array */
 	int			granted[MAX_LOCKMODES]; /* counts of granted locks */
 	int			nGranted;		/* total of granted[] array */
+	PGPROC		*upgradedBy;	/* is lock being upgraded by this process? */
 } LOCK;
 
 #define LOCK_LOCKMETHOD(lock) ((LOCKMETHODID) (lock).tag.locktag_lockmethodid)
@@ -390,7 +392,8 @@ extern LockAcquireResult LockAcquireExtended(const LOCKTAG *locktag,
 											 bool dontWait,
 											 bool reportMemoryError,
 											 LOCALLOCK **locallockp,
-											 bool logLockFailure);
+											 bool logLockFailure,
+											 bool isUpgrade);
 extern void AbortStrongLockAcquire(void);
 extern void MarkLockClear(LOCALLOCK *locallock);
 extern bool LockRelease(const LOCKTAG *locktag,
diff --git a/src/include/storage/proc.h b/src/include/storage/proc.h
index 3e1d1fad5f9..c4bc3a24044 100644
--- a/src/include/storage/proc.h
+++ b/src/include/storage/proc.h
@@ -563,10 +563,12 @@ extern bool HaveNFreeProcs(int n, int *nfree);
 extern void ProcReleaseLocks(bool isCommit);
 
 extern ProcWaitStatus JoinWaitQueue(LOCALLOCK *locallock,
-									LockMethod lockMethodTable, bool dontWait);
+									LockMethod lockMethodTable, bool dontWait,
+									bool isUpgrade);
 extern ProcWaitStatus ProcSleep(LOCALLOCK *locallock);
 extern void ProcWakeup(PGPROC *proc, ProcWaitStatus waitStatus);
 extern void ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock);
+extern DeadLockState CheckDeadLock(void);
 extern void CheckDeadLockAlert(void);
 extern void LockErrorCleanup(void);
 extern void GetLockHoldersAndWaiters(LOCALLOCK *locallock,
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 966+ messages in thread

* [PATCH] Lock upgrade without deadlocks.
@ 2026-04-17 13:13  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 966+ messages in thread

From: Antonin Houska @ 2026-04-17 13:13 UTC (permalink / raw)

For REPACK (CONCURRENTLY), it's essential to upgrade ShareUpdateExclusive lock
to AccessExclusiveLock for the final stage of table processing. However, we
cannot release the earlier before requesting the latter, because if another
process ran DDL on the table in between, REPACK would have to cancel its
transaction, and waste possibly a lot of work. Without releasing
ShareUpdateExclusive lock, the upgrade can cause deadlock, and it's not really
deterministic whether REPACK or the other process will get canceled.

This patch adds a new field to the LOCK structure which tells which process is
trying to upgrade the lock - this field is set right before the upgrade. All
processes waiting in the lock queue are then woken up to check the flag.

When another process tries to get the lock after that, it checks this field,
and if it's set, it checks for deadlocks even if deadlock timeout hasn't
expired yet. If the process is already in the lock's queue and sleeping, the
lock upgrading process wakes it up so it checks the flag immediately.

At the time the upgrading process starts to sleep, all the other process
should be aware that they need to check for deadlock during their lock
acquisitions, so the upgrading process does not have to check for deadlocks
when it gets woken up. Thus it should never receive deadlock error report.
---
 src/backend/commands/repack.c   |   4 +-
 src/backend/storage/lmgr/lmgr.c |  35 ++++++++---
 src/backend/storage/lmgr/lock.c | 107 +++++++++++++++++++++++++++++++-
 src/backend/storage/lmgr/proc.c |  32 +++++++---
 src/include/storage/lmgr.h      |   1 +
 src/include/storage/lock.h      |   5 +-
 src/include/storage/proc.h      |   4 +-
 7 files changed, 166 insertions(+), 22 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index 58e3867246f..310b2a65099 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -3055,8 +3055,10 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Acquire AccessExclusiveLock on the table, its TOAST relation (if there
 	 * is one), all its indexes, so that we can swap the files.
+	 *
+	 * TODO The same for indexes and TOAST?
 	 */
-	LockRelationOid(old_table_oid, AccessExclusiveLock);
+	LockRelationOidUpgrade(old_table_oid, AccessExclusiveLock);
 
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
diff --git a/src/backend/storage/lmgr/lmgr.c b/src/backend/storage/lmgr/lmgr.c
index 2ccf7237fee..0b103485d54 100644
--- a/src/backend/storage/lmgr/lmgr.c
+++ b/src/backend/storage/lmgr/lmgr.c
@@ -58,6 +58,8 @@ typedef struct XactLockTableWaitInfo
 	const ItemPointerData *ctid;
 } XactLockTableWaitInfo;
 
+static void LockRelationOidCommon(Oid relid, LOCKMODE lockmode,
+								  bool isUpgrade);
 static void XactLockTableWaitErrorCb(void *arg);
 
 /*
@@ -105,6 +107,21 @@ SetLocktagRelationOid(LOCKTAG *tag, Oid relid)
  */
 void
 LockRelationOid(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, false);
+}
+
+/*
+ * Like above, but upgrade an existing lock.
+ */
+void
+LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, true);
+}
+
+static void
+LockRelationOidCommon(Oid relid, LOCKMODE lockmode, bool isUpgrade)
 {
 	LOCKTAG		tag;
 	LOCALLOCK  *locallock;
@@ -113,7 +130,7 @@ LockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, isUpgrade);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages, so that we
@@ -157,7 +174,7 @@ ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -191,7 +208,7 @@ LockRelationId(LockRelId *relid, LOCKMODE lockmode)
 	SET_LOCKTAG_RELATION(tag, relid->dbId, relid->relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -254,7 +271,7 @@ LockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -286,7 +303,7 @@ ConditionalLockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -591,7 +608,7 @@ ConditionalLockTuple(Relation relation, const ItemPointerData *tid, LOCKMODE loc
 					  ItemPointerGetOffsetNumber(tid));
 
 	return (LockAcquireExtended(&tag, lockmode, false, true, true, NULL,
-								logLockFailure) != LOCKACQUIRE_NOT_AVAIL);
+								logLockFailure, false) != LOCKACQUIRE_NOT_AVAIL);
 }
 
 /*
@@ -749,7 +766,7 @@ ConditionalXactLockTableWait(TransactionId xid, bool logLockFailure)
 		SET_LOCKTAG_TRANSACTION(tag, xid);
 
 		if (LockAcquireExtended(&tag, ShareLock, false, true, true, NULL,
-								logLockFailure)
+								logLockFailure, false)
 			== LOCKACQUIRE_NOT_AVAIL)
 			return false;
 
@@ -1043,7 +1060,7 @@ ConditionalLockDatabaseObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -1123,7 +1140,7 @@ ConditionalLockSharedObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
diff --git a/src/backend/storage/lmgr/lock.c b/src/backend/storage/lmgr/lock.c
index c221fe96889..bcbbf36c091 100644
--- a/src/backend/storage/lmgr/lock.c
+++ b/src/backend/storage/lmgr/lock.c
@@ -810,7 +810,7 @@ LockAcquire(const LOCKTAG *locktag,
 			bool dontWait)
 {
 	return LockAcquireExtended(locktag, lockmode, sessionLock, dontWait,
-							   true, NULL, false);
+							   true, NULL, false, false);
 }
 
 /*
@@ -829,6 +829,9 @@ LockAcquire(const LOCKTAG *locktag,
  *
  * logLockFailure indicates whether to log details when a lock acquisition
  * fails with dontWait = true.
+ *
+ * isUpgrade should be true if the backend already holds this lock in a lower
+ * mode.
  */
 LockAcquireResult
 LockAcquireExtended(const LOCKTAG *locktag,
@@ -837,7 +840,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 					bool dontWait,
 					bool reportMemoryError,
 					LOCALLOCK **locallockp,
-					bool logLockFailure)
+					bool logLockFailure,
+					bool isUpgrade)
 {
 	LOCKMETHODID lockmethodid = locktag->locktag_lockmethodid;
 	LockMethod	lockMethodTable;
@@ -1119,7 +1123,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		 * case, because JoinWaitQueue() may discover that we can acquire the
 		 * lock immediately after all.
 		 */
-		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait);
+		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait,
+								   isUpgrade);
 	}
 
 	if (waitResult == PROC_WAIT_STATUS_ERROR)
@@ -1225,8 +1230,85 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		Assert(!dontWait);
 		PROCLOCK_PRINT("LockAcquire: sleeping on lock", proclock);
 		LOCK_PRINT("LockAcquire: sleeping on lock", lock, lockmode);
+
+		/*
+		 * Lock upgrade can introduce deadlock, therefore enforce special
+		 * behavior of other processes that deal with this lock. In
+		 * particular, any waiter that sees upgradedBy set is expected to
+		 * perform deadlock check as soon as it's woken up. (Since we are
+		 * already in the queue, the deadlock detector has all the information
+		 * it needs.)
+		 */
+		if (isUpgrade)
+		{
+			dlist_iter	iter;
+
+			/*
+			 * There should not be multiple upgrades at the same time. XXX
+			 * ERROR ?
+			 */
+			Assert(lock->upgradedBy == NULL);
+
+			lock->upgradedBy = MyProc;
+
+			/*
+			 * Now, before I sleep myself, wake up all the existing waiters
+			 * (except for me), so they check for deadlock.
+			 */
+			dclist_foreach(iter, &lock->waitProcs)
+			{
+				PGPROC	   *proc = dlist_container(PGPROC, waitLink,
+												   iter.cur);
+
+				if (proc != MyProc)
+					SetLatch(&proc->procLatch);
+			}
+		}
 		LWLockRelease(partitionLock);
 
+		if (!isUpgrade)
+		{
+			HASH_SEQ_STATUS status;
+			LOCALLOCK	*llock;
+			bool	check_deadlock = false;
+
+			/*
+			 * Likewise, new waiters should check the isUpgrade flag and
+			 * engage deadlock detector if needed. The lock being acquired
+			 * should already be in the hash.
+			 *
+			 * Note that, to make deadlock detection happen as soon as
+			 * possible (i.e. regardless deadlock timeout), we check the other
+			 * locks too. XXX There might be ways to get into the deadlock
+			 * indirectly, but that needs more analysis. As long as the
+			 * upgrading process skips deadlock detection altogether, the
+			 * worst consequence of missing some lock wait cycle here is that
+			 * the other process will be kicked-off no sooner than after its
+			 * deadlock timeout has elapsed. (Which in turn means that the
+			 * lock upgrade will take longer than expected.)
+			 */
+			hash_seq_init(&status, LockMethodLocalHash);
+			while ((llock = (LOCALLOCK *) hash_seq_search(&status)) != NULL)
+			{
+				LOCK	*mylock = llock->lock;
+
+				if (mylock && mylock->upgradedBy)
+				{
+					check_deadlock = true;
+					break;
+				}
+			}
+			/* Check for deadlock if needed. */
+			if (check_deadlock)
+			{
+				DeadLockState	deadlock_state;
+
+				deadlock_state = CheckDeadLock();
+				if (deadlock_state == DS_HARD_DEADLOCK)
+					DeadLockReport();
+			}
+		}
+
 		waitResult = WaitOnLock(locallock, owner);
 
 		/*
@@ -1245,6 +1327,18 @@ LockAcquireExtended(const LOCKTAG *locktag,
 			DeadLockReport();
 			/* DeadLockReport() will not return */
 		}
+
+		/*
+		 * If finishing the lock upgrade, we should not get into a deadlock
+		 * anymore, so let others know that they do not have to care either.
+		 */
+		if (isUpgrade)
+		{
+			LWLockAcquire(partitionLock, LW_EXCLUSIVE);
+			Assert(lock->upgradedBy != NULL);
+			lock->upgradedBy = NULL;
+			LWLockRelease(partitionLock);
+		}
 	}
 	else
 		LWLockRelease(partitionLock);
@@ -1322,6 +1416,7 @@ SetupLockInTable(LockMethod lockMethodTable, PGPROC *proc,
 		lock->nGranted = 0;
 		MemSet(lock->requested, 0, sizeof(int) * MAX_LOCKMODES);
 		MemSet(lock->granted, 0, sizeof(int) * MAX_LOCKMODES);
+		lock->upgradedBy = NULL;
 		LOCK_PRINT("LockAcquire: new", lock, lockmode);
 	}
 	else
@@ -1768,6 +1863,12 @@ CleanUpLock(LOCK *lock, PROCLOCK *proclock,
 			elog(PANIC, "proclock table corrupted");
 	}
 
+	/*
+	 * Was this backend upgrading the lock?
+	 */
+	if (lock->upgradedBy == MyProc)
+		lock->upgradedBy = NULL;
+
 	if (lock->nRequested == 0)
 	{
 		/*
diff --git a/src/backend/storage/lmgr/proc.c b/src/backend/storage/lmgr/proc.c
index 1ac25068d62..7b31370db70 100644
--- a/src/backend/storage/lmgr/proc.c
+++ b/src/backend/storage/lmgr/proc.c
@@ -95,7 +95,6 @@ static volatile sig_atomic_t got_deadlock_timeout;
 static void RemoveProcFromArray(int code, Datum arg);
 static void ProcKill(int code, Datum arg);
 static void AuxiliaryProcKill(int code, Datum arg);
-static DeadLockState CheckDeadLock(void);
 
 
 /*
@@ -1143,7 +1142,8 @@ AuxiliaryPidGetProc(int pid)
  * NOTES: The process queue is now a priority queue for locking.
  */
 ProcWaitStatus
-JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
+JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait,
+			  bool isUpgrade)
 {
 	LOCKMODE	lockmode = locallock->tag.mode;
 	LOCK	   *lock = locallock->lock;
@@ -1230,8 +1230,12 @@ JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
 			/* Must he wait for me? */
 			if (lockMethodTable->conflictTab[proc->waitLockMode] & myHeldLocks)
 			{
-				/* Must I wait for him ? */
-				if (lockMethodTable->conflictTab[lockmode] & proc->heldLocks)
+				/*
+				 * Must I wait for him? I don't want a deadlock during lock
+				 * upgrade - other processes should fail on it.
+				 */
+				if ((lockMethodTable->conflictTab[lockmode] & proc->heldLocks) &&
+					!isUpgrade)
 				{
 					/*
 					 * Yes, so we have a deadlock.  Easiest way to clean up
@@ -1461,8 +1465,22 @@ ProcSleep(LOCALLOCK *locallock)
 			(void) WaitLatch(MyLatch, WL_LATCH_SET | WL_EXIT_ON_PM_DEATH, 0,
 							 PG_WAIT_LOCK | locallock->tag.lock.locktag_type);
 			ResetLatch(MyLatch);
-			/* check for deadlocks first, as that's probably log-worthy */
-			if (got_deadlock_timeout)
+			/*
+			 * Check for deadlocks first, as that's probably log-worthy.
+			 *
+			 * Do not wait for the timeout if the lock is being upgraded since
+			 * the risk of deadlock is higher now. However, do not check for
+			 * deadlock if the lock is being upgraded by this process - other
+			 * processes should take care.
+			 *
+			 * TODO Possible optimization: if this is the only lock of the
+			 * backend and if it did not have any weaker lock on the table so
+			 * far, it should be safe to skip the deadlock check. However, to
+			 * evaluate the situation, we need to take fast-path locks into
+			 * account.
+			 */
+			if ((got_deadlock_timeout || lock->upgradedBy) &&
+				lock->upgradedBy != MyProc)
 			{
 				deadlock_state = CheckDeadLock();
 				got_deadlock_timeout = false;
@@ -1819,7 +1837,7 @@ ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock)
  * not, just return.  If we have a real deadlock, remove ourselves from the
  * lock's wait queue.
  */
-static DeadLockState
+DeadLockState
 CheckDeadLock(void)
 {
 	int			i;
diff --git a/src/include/storage/lmgr.h b/src/include/storage/lmgr.h
index 2a985ce5e15..eddbe1e8622 100644
--- a/src/include/storage/lmgr.h
+++ b/src/include/storage/lmgr.h
@@ -38,6 +38,7 @@ extern void RelationInitLockInfo(Relation relation);
 
 /* Lock a relation */
 extern void LockRelationOid(Oid relid, LOCKMODE lockmode);
+extern void LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode);
 extern void LockRelationId(LockRelId *relid, LOCKMODE lockmode);
 extern bool ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode);
 extern void UnlockRelationId(LockRelId *relid, LOCKMODE lockmode);
diff --git a/src/include/storage/lock.h b/src/include/storage/lock.h
index ee3cb1dc203..080b69eddfd 100644
--- a/src/include/storage/lock.h
+++ b/src/include/storage/lock.h
@@ -131,6 +131,7 @@ typedef const LockMethodData *LockMethod;
  * nRequested -- total requested locks of all types.
  * granted -- count of each lock type currently granted on the lock.
  * nGranted -- total granted locks of all types.
+ * upgrader -- process that intends to upgrade the lock
  *
  * Note: these counts count 1 for each backend.  Internally to a backend,
  * there may be multiple grabs on a particular lock, but this is not reflected
@@ -150,6 +151,7 @@ typedef struct LOCK
 	int			nRequested;		/* total of requested[] array */
 	int			granted[MAX_LOCKMODES]; /* counts of granted locks */
 	int			nGranted;		/* total of granted[] array */
+	PGPROC		*upgradedBy;	/* is lock being upgraded by this process? */
 } LOCK;
 
 #define LOCK_LOCKMETHOD(lock) ((LOCKMETHODID) (lock).tag.locktag_lockmethodid)
@@ -390,7 +392,8 @@ extern LockAcquireResult LockAcquireExtended(const LOCKTAG *locktag,
 											 bool dontWait,
 											 bool reportMemoryError,
 											 LOCALLOCK **locallockp,
-											 bool logLockFailure);
+											 bool logLockFailure,
+											 bool isUpgrade);
 extern void AbortStrongLockAcquire(void);
 extern void MarkLockClear(LOCALLOCK *locallock);
 extern bool LockRelease(const LOCKTAG *locktag,
diff --git a/src/include/storage/proc.h b/src/include/storage/proc.h
index 3e1d1fad5f9..c4bc3a24044 100644
--- a/src/include/storage/proc.h
+++ b/src/include/storage/proc.h
@@ -563,10 +563,12 @@ extern bool HaveNFreeProcs(int n, int *nfree);
 extern void ProcReleaseLocks(bool isCommit);
 
 extern ProcWaitStatus JoinWaitQueue(LOCALLOCK *locallock,
-									LockMethod lockMethodTable, bool dontWait);
+									LockMethod lockMethodTable, bool dontWait,
+									bool isUpgrade);
 extern ProcWaitStatus ProcSleep(LOCALLOCK *locallock);
 extern void ProcWakeup(PGPROC *proc, ProcWaitStatus waitStatus);
 extern void ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock);
+extern DeadLockState CheckDeadLock(void);
 extern void CheckDeadLockAlert(void);
 extern void LockErrorCleanup(void);
 extern void GetLockHoldersAndWaiters(LOCALLOCK *locallock,
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 966+ messages in thread

* [PATCH] Lock upgrade without deadlocks.
@ 2026-04-17 13:13  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 966+ messages in thread

From: Antonin Houska @ 2026-04-17 13:13 UTC (permalink / raw)

For REPACK (CONCURRENTLY), it's essential to upgrade ShareUpdateExclusive lock
to AccessExclusiveLock for the final stage of table processing. However, we
cannot release the earlier before requesting the latter, because if another
process ran DDL on the table in between, REPACK would have to cancel its
transaction, and waste possibly a lot of work. Without releasing
ShareUpdateExclusive lock, the upgrade can cause deadlock, and it's not really
deterministic whether REPACK or the other process will get canceled.

This patch adds a new field to the LOCK structure which tells which process is
trying to upgrade the lock - this field is set right before the upgrade. All
processes waiting in the lock queue are then woken up to check the flag.

When another process tries to get the lock after that, it checks this field,
and if it's set, it checks for deadlocks even if deadlock timeout hasn't
expired yet. If the process is already in the lock's queue and sleeping, the
lock upgrading process wakes it up so it checks the flag immediately.

At the time the upgrading process starts to sleep, all the other process
should be aware that they need to check for deadlock during their lock
acquisitions, so the upgrading process does not have to check for deadlocks
when it gets woken up. Thus it should never receive deadlock error report.
---
 src/backend/commands/repack.c   |   4 +-
 src/backend/storage/lmgr/lmgr.c |  35 ++++++++---
 src/backend/storage/lmgr/lock.c | 107 +++++++++++++++++++++++++++++++-
 src/backend/storage/lmgr/proc.c |  32 +++++++---
 src/include/storage/lmgr.h      |   1 +
 src/include/storage/lock.h      |   5 +-
 src/include/storage/proc.h      |   4 +-
 7 files changed, 166 insertions(+), 22 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index 58e3867246f..310b2a65099 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -3055,8 +3055,10 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Acquire AccessExclusiveLock on the table, its TOAST relation (if there
 	 * is one), all its indexes, so that we can swap the files.
+	 *
+	 * TODO The same for indexes and TOAST?
 	 */
-	LockRelationOid(old_table_oid, AccessExclusiveLock);
+	LockRelationOidUpgrade(old_table_oid, AccessExclusiveLock);
 
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
diff --git a/src/backend/storage/lmgr/lmgr.c b/src/backend/storage/lmgr/lmgr.c
index 2ccf7237fee..0b103485d54 100644
--- a/src/backend/storage/lmgr/lmgr.c
+++ b/src/backend/storage/lmgr/lmgr.c
@@ -58,6 +58,8 @@ typedef struct XactLockTableWaitInfo
 	const ItemPointerData *ctid;
 } XactLockTableWaitInfo;
 
+static void LockRelationOidCommon(Oid relid, LOCKMODE lockmode,
+								  bool isUpgrade);
 static void XactLockTableWaitErrorCb(void *arg);
 
 /*
@@ -105,6 +107,21 @@ SetLocktagRelationOid(LOCKTAG *tag, Oid relid)
  */
 void
 LockRelationOid(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, false);
+}
+
+/*
+ * Like above, but upgrade an existing lock.
+ */
+void
+LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, true);
+}
+
+static void
+LockRelationOidCommon(Oid relid, LOCKMODE lockmode, bool isUpgrade)
 {
 	LOCKTAG		tag;
 	LOCALLOCK  *locallock;
@@ -113,7 +130,7 @@ LockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, isUpgrade);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages, so that we
@@ -157,7 +174,7 @@ ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -191,7 +208,7 @@ LockRelationId(LockRelId *relid, LOCKMODE lockmode)
 	SET_LOCKTAG_RELATION(tag, relid->dbId, relid->relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -254,7 +271,7 @@ LockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -286,7 +303,7 @@ ConditionalLockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -591,7 +608,7 @@ ConditionalLockTuple(Relation relation, const ItemPointerData *tid, LOCKMODE loc
 					  ItemPointerGetOffsetNumber(tid));
 
 	return (LockAcquireExtended(&tag, lockmode, false, true, true, NULL,
-								logLockFailure) != LOCKACQUIRE_NOT_AVAIL);
+								logLockFailure, false) != LOCKACQUIRE_NOT_AVAIL);
 }
 
 /*
@@ -749,7 +766,7 @@ ConditionalXactLockTableWait(TransactionId xid, bool logLockFailure)
 		SET_LOCKTAG_TRANSACTION(tag, xid);
 
 		if (LockAcquireExtended(&tag, ShareLock, false, true, true, NULL,
-								logLockFailure)
+								logLockFailure, false)
 			== LOCKACQUIRE_NOT_AVAIL)
 			return false;
 
@@ -1043,7 +1060,7 @@ ConditionalLockDatabaseObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -1123,7 +1140,7 @@ ConditionalLockSharedObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
diff --git a/src/backend/storage/lmgr/lock.c b/src/backend/storage/lmgr/lock.c
index c221fe96889..bcbbf36c091 100644
--- a/src/backend/storage/lmgr/lock.c
+++ b/src/backend/storage/lmgr/lock.c
@@ -810,7 +810,7 @@ LockAcquire(const LOCKTAG *locktag,
 			bool dontWait)
 {
 	return LockAcquireExtended(locktag, lockmode, sessionLock, dontWait,
-							   true, NULL, false);
+							   true, NULL, false, false);
 }
 
 /*
@@ -829,6 +829,9 @@ LockAcquire(const LOCKTAG *locktag,
  *
  * logLockFailure indicates whether to log details when a lock acquisition
  * fails with dontWait = true.
+ *
+ * isUpgrade should be true if the backend already holds this lock in a lower
+ * mode.
  */
 LockAcquireResult
 LockAcquireExtended(const LOCKTAG *locktag,
@@ -837,7 +840,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 					bool dontWait,
 					bool reportMemoryError,
 					LOCALLOCK **locallockp,
-					bool logLockFailure)
+					bool logLockFailure,
+					bool isUpgrade)
 {
 	LOCKMETHODID lockmethodid = locktag->locktag_lockmethodid;
 	LockMethod	lockMethodTable;
@@ -1119,7 +1123,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		 * case, because JoinWaitQueue() may discover that we can acquire the
 		 * lock immediately after all.
 		 */
-		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait);
+		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait,
+								   isUpgrade);
 	}
 
 	if (waitResult == PROC_WAIT_STATUS_ERROR)
@@ -1225,8 +1230,85 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		Assert(!dontWait);
 		PROCLOCK_PRINT("LockAcquire: sleeping on lock", proclock);
 		LOCK_PRINT("LockAcquire: sleeping on lock", lock, lockmode);
+
+		/*
+		 * Lock upgrade can introduce deadlock, therefore enforce special
+		 * behavior of other processes that deal with this lock. In
+		 * particular, any waiter that sees upgradedBy set is expected to
+		 * perform deadlock check as soon as it's woken up. (Since we are
+		 * already in the queue, the deadlock detector has all the information
+		 * it needs.)
+		 */
+		if (isUpgrade)
+		{
+			dlist_iter	iter;
+
+			/*
+			 * There should not be multiple upgrades at the same time. XXX
+			 * ERROR ?
+			 */
+			Assert(lock->upgradedBy == NULL);
+
+			lock->upgradedBy = MyProc;
+
+			/*
+			 * Now, before I sleep myself, wake up all the existing waiters
+			 * (except for me), so they check for deadlock.
+			 */
+			dclist_foreach(iter, &lock->waitProcs)
+			{
+				PGPROC	   *proc = dlist_container(PGPROC, waitLink,
+												   iter.cur);
+
+				if (proc != MyProc)
+					SetLatch(&proc->procLatch);
+			}
+		}
 		LWLockRelease(partitionLock);
 
+		if (!isUpgrade)
+		{
+			HASH_SEQ_STATUS status;
+			LOCALLOCK	*llock;
+			bool	check_deadlock = false;
+
+			/*
+			 * Likewise, new waiters should check the isUpgrade flag and
+			 * engage deadlock detector if needed. The lock being acquired
+			 * should already be in the hash.
+			 *
+			 * Note that, to make deadlock detection happen as soon as
+			 * possible (i.e. regardless deadlock timeout), we check the other
+			 * locks too. XXX There might be ways to get into the deadlock
+			 * indirectly, but that needs more analysis. As long as the
+			 * upgrading process skips deadlock detection altogether, the
+			 * worst consequence of missing some lock wait cycle here is that
+			 * the other process will be kicked-off no sooner than after its
+			 * deadlock timeout has elapsed. (Which in turn means that the
+			 * lock upgrade will take longer than expected.)
+			 */
+			hash_seq_init(&status, LockMethodLocalHash);
+			while ((llock = (LOCALLOCK *) hash_seq_search(&status)) != NULL)
+			{
+				LOCK	*mylock = llock->lock;
+
+				if (mylock && mylock->upgradedBy)
+				{
+					check_deadlock = true;
+					break;
+				}
+			}
+			/* Check for deadlock if needed. */
+			if (check_deadlock)
+			{
+				DeadLockState	deadlock_state;
+
+				deadlock_state = CheckDeadLock();
+				if (deadlock_state == DS_HARD_DEADLOCK)
+					DeadLockReport();
+			}
+		}
+
 		waitResult = WaitOnLock(locallock, owner);
 
 		/*
@@ -1245,6 +1327,18 @@ LockAcquireExtended(const LOCKTAG *locktag,
 			DeadLockReport();
 			/* DeadLockReport() will not return */
 		}
+
+		/*
+		 * If finishing the lock upgrade, we should not get into a deadlock
+		 * anymore, so let others know that they do not have to care either.
+		 */
+		if (isUpgrade)
+		{
+			LWLockAcquire(partitionLock, LW_EXCLUSIVE);
+			Assert(lock->upgradedBy != NULL);
+			lock->upgradedBy = NULL;
+			LWLockRelease(partitionLock);
+		}
 	}
 	else
 		LWLockRelease(partitionLock);
@@ -1322,6 +1416,7 @@ SetupLockInTable(LockMethod lockMethodTable, PGPROC *proc,
 		lock->nGranted = 0;
 		MemSet(lock->requested, 0, sizeof(int) * MAX_LOCKMODES);
 		MemSet(lock->granted, 0, sizeof(int) * MAX_LOCKMODES);
+		lock->upgradedBy = NULL;
 		LOCK_PRINT("LockAcquire: new", lock, lockmode);
 	}
 	else
@@ -1768,6 +1863,12 @@ CleanUpLock(LOCK *lock, PROCLOCK *proclock,
 			elog(PANIC, "proclock table corrupted");
 	}
 
+	/*
+	 * Was this backend upgrading the lock?
+	 */
+	if (lock->upgradedBy == MyProc)
+		lock->upgradedBy = NULL;
+
 	if (lock->nRequested == 0)
 	{
 		/*
diff --git a/src/backend/storage/lmgr/proc.c b/src/backend/storage/lmgr/proc.c
index 1ac25068d62..7b31370db70 100644
--- a/src/backend/storage/lmgr/proc.c
+++ b/src/backend/storage/lmgr/proc.c
@@ -95,7 +95,6 @@ static volatile sig_atomic_t got_deadlock_timeout;
 static void RemoveProcFromArray(int code, Datum arg);
 static void ProcKill(int code, Datum arg);
 static void AuxiliaryProcKill(int code, Datum arg);
-static DeadLockState CheckDeadLock(void);
 
 
 /*
@@ -1143,7 +1142,8 @@ AuxiliaryPidGetProc(int pid)
  * NOTES: The process queue is now a priority queue for locking.
  */
 ProcWaitStatus
-JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
+JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait,
+			  bool isUpgrade)
 {
 	LOCKMODE	lockmode = locallock->tag.mode;
 	LOCK	   *lock = locallock->lock;
@@ -1230,8 +1230,12 @@ JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
 			/* Must he wait for me? */
 			if (lockMethodTable->conflictTab[proc->waitLockMode] & myHeldLocks)
 			{
-				/* Must I wait for him ? */
-				if (lockMethodTable->conflictTab[lockmode] & proc->heldLocks)
+				/*
+				 * Must I wait for him? I don't want a deadlock during lock
+				 * upgrade - other processes should fail on it.
+				 */
+				if ((lockMethodTable->conflictTab[lockmode] & proc->heldLocks) &&
+					!isUpgrade)
 				{
 					/*
 					 * Yes, so we have a deadlock.  Easiest way to clean up
@@ -1461,8 +1465,22 @@ ProcSleep(LOCALLOCK *locallock)
 			(void) WaitLatch(MyLatch, WL_LATCH_SET | WL_EXIT_ON_PM_DEATH, 0,
 							 PG_WAIT_LOCK | locallock->tag.lock.locktag_type);
 			ResetLatch(MyLatch);
-			/* check for deadlocks first, as that's probably log-worthy */
-			if (got_deadlock_timeout)
+			/*
+			 * Check for deadlocks first, as that's probably log-worthy.
+			 *
+			 * Do not wait for the timeout if the lock is being upgraded since
+			 * the risk of deadlock is higher now. However, do not check for
+			 * deadlock if the lock is being upgraded by this process - other
+			 * processes should take care.
+			 *
+			 * TODO Possible optimization: if this is the only lock of the
+			 * backend and if it did not have any weaker lock on the table so
+			 * far, it should be safe to skip the deadlock check. However, to
+			 * evaluate the situation, we need to take fast-path locks into
+			 * account.
+			 */
+			if ((got_deadlock_timeout || lock->upgradedBy) &&
+				lock->upgradedBy != MyProc)
 			{
 				deadlock_state = CheckDeadLock();
 				got_deadlock_timeout = false;
@@ -1819,7 +1837,7 @@ ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock)
  * not, just return.  If we have a real deadlock, remove ourselves from the
  * lock's wait queue.
  */
-static DeadLockState
+DeadLockState
 CheckDeadLock(void)
 {
 	int			i;
diff --git a/src/include/storage/lmgr.h b/src/include/storage/lmgr.h
index 2a985ce5e15..eddbe1e8622 100644
--- a/src/include/storage/lmgr.h
+++ b/src/include/storage/lmgr.h
@@ -38,6 +38,7 @@ extern void RelationInitLockInfo(Relation relation);
 
 /* Lock a relation */
 extern void LockRelationOid(Oid relid, LOCKMODE lockmode);
+extern void LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode);
 extern void LockRelationId(LockRelId *relid, LOCKMODE lockmode);
 extern bool ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode);
 extern void UnlockRelationId(LockRelId *relid, LOCKMODE lockmode);
diff --git a/src/include/storage/lock.h b/src/include/storage/lock.h
index ee3cb1dc203..080b69eddfd 100644
--- a/src/include/storage/lock.h
+++ b/src/include/storage/lock.h
@@ -131,6 +131,7 @@ typedef const LockMethodData *LockMethod;
  * nRequested -- total requested locks of all types.
  * granted -- count of each lock type currently granted on the lock.
  * nGranted -- total granted locks of all types.
+ * upgrader -- process that intends to upgrade the lock
  *
  * Note: these counts count 1 for each backend.  Internally to a backend,
  * there may be multiple grabs on a particular lock, but this is not reflected
@@ -150,6 +151,7 @@ typedef struct LOCK
 	int			nRequested;		/* total of requested[] array */
 	int			granted[MAX_LOCKMODES]; /* counts of granted locks */
 	int			nGranted;		/* total of granted[] array */
+	PGPROC		*upgradedBy;	/* is lock being upgraded by this process? */
 } LOCK;
 
 #define LOCK_LOCKMETHOD(lock) ((LOCKMETHODID) (lock).tag.locktag_lockmethodid)
@@ -390,7 +392,8 @@ extern LockAcquireResult LockAcquireExtended(const LOCKTAG *locktag,
 											 bool dontWait,
 											 bool reportMemoryError,
 											 LOCALLOCK **locallockp,
-											 bool logLockFailure);
+											 bool logLockFailure,
+											 bool isUpgrade);
 extern void AbortStrongLockAcquire(void);
 extern void MarkLockClear(LOCALLOCK *locallock);
 extern bool LockRelease(const LOCKTAG *locktag,
diff --git a/src/include/storage/proc.h b/src/include/storage/proc.h
index 3e1d1fad5f9..c4bc3a24044 100644
--- a/src/include/storage/proc.h
+++ b/src/include/storage/proc.h
@@ -563,10 +563,12 @@ extern bool HaveNFreeProcs(int n, int *nfree);
 extern void ProcReleaseLocks(bool isCommit);
 
 extern ProcWaitStatus JoinWaitQueue(LOCALLOCK *locallock,
-									LockMethod lockMethodTable, bool dontWait);
+									LockMethod lockMethodTable, bool dontWait,
+									bool isUpgrade);
 extern ProcWaitStatus ProcSleep(LOCALLOCK *locallock);
 extern void ProcWakeup(PGPROC *proc, ProcWaitStatus waitStatus);
 extern void ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock);
+extern DeadLockState CheckDeadLock(void);
 extern void CheckDeadLockAlert(void);
 extern void LockErrorCleanup(void);
 extern void GetLockHoldersAndWaiters(LOCALLOCK *locallock,
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 966+ messages in thread

* [PATCH] Lock upgrade without deadlocks.
@ 2026-04-17 13:13  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 966+ messages in thread

From: Antonin Houska @ 2026-04-17 13:13 UTC (permalink / raw)

For REPACK (CONCURRENTLY), it's essential to upgrade ShareUpdateExclusive lock
to AccessExclusiveLock for the final stage of table processing. However, we
cannot release the earlier before requesting the latter, because if another
process ran DDL on the table in between, REPACK would have to cancel its
transaction, and waste possibly a lot of work. Without releasing
ShareUpdateExclusive lock, the upgrade can cause deadlock, and it's not really
deterministic whether REPACK or the other process will get canceled.

This patch adds a new field to the LOCK structure which tells which process is
trying to upgrade the lock - this field is set right before the upgrade. All
processes waiting in the lock queue are then woken up to check the flag.

When another process tries to get the lock after that, it checks this field,
and if it's set, it checks for deadlocks even if deadlock timeout hasn't
expired yet. If the process is already in the lock's queue and sleeping, the
lock upgrading process wakes it up so it checks the flag immediately.

At the time the upgrading process starts to sleep, all the other process
should be aware that they need to check for deadlock during their lock
acquisitions, so the upgrading process does not have to check for deadlocks
when it gets woken up. Thus it should never receive deadlock error report.
---
 src/backend/commands/repack.c   |   4 +-
 src/backend/storage/lmgr/lmgr.c |  35 ++++++++---
 src/backend/storage/lmgr/lock.c | 107 +++++++++++++++++++++++++++++++-
 src/backend/storage/lmgr/proc.c |  32 +++++++---
 src/include/storage/lmgr.h      |   1 +
 src/include/storage/lock.h      |   5 +-
 src/include/storage/proc.h      |   4 +-
 7 files changed, 166 insertions(+), 22 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index 58e3867246f..310b2a65099 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -3055,8 +3055,10 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Acquire AccessExclusiveLock on the table, its TOAST relation (if there
 	 * is one), all its indexes, so that we can swap the files.
+	 *
+	 * TODO The same for indexes and TOAST?
 	 */
-	LockRelationOid(old_table_oid, AccessExclusiveLock);
+	LockRelationOidUpgrade(old_table_oid, AccessExclusiveLock);
 
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
diff --git a/src/backend/storage/lmgr/lmgr.c b/src/backend/storage/lmgr/lmgr.c
index 2ccf7237fee..0b103485d54 100644
--- a/src/backend/storage/lmgr/lmgr.c
+++ b/src/backend/storage/lmgr/lmgr.c
@@ -58,6 +58,8 @@ typedef struct XactLockTableWaitInfo
 	const ItemPointerData *ctid;
 } XactLockTableWaitInfo;
 
+static void LockRelationOidCommon(Oid relid, LOCKMODE lockmode,
+								  bool isUpgrade);
 static void XactLockTableWaitErrorCb(void *arg);
 
 /*
@@ -105,6 +107,21 @@ SetLocktagRelationOid(LOCKTAG *tag, Oid relid)
  */
 void
 LockRelationOid(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, false);
+}
+
+/*
+ * Like above, but upgrade an existing lock.
+ */
+void
+LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, true);
+}
+
+static void
+LockRelationOidCommon(Oid relid, LOCKMODE lockmode, bool isUpgrade)
 {
 	LOCKTAG		tag;
 	LOCALLOCK  *locallock;
@@ -113,7 +130,7 @@ LockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, isUpgrade);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages, so that we
@@ -157,7 +174,7 @@ ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -191,7 +208,7 @@ LockRelationId(LockRelId *relid, LOCKMODE lockmode)
 	SET_LOCKTAG_RELATION(tag, relid->dbId, relid->relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -254,7 +271,7 @@ LockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -286,7 +303,7 @@ ConditionalLockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -591,7 +608,7 @@ ConditionalLockTuple(Relation relation, const ItemPointerData *tid, LOCKMODE loc
 					  ItemPointerGetOffsetNumber(tid));
 
 	return (LockAcquireExtended(&tag, lockmode, false, true, true, NULL,
-								logLockFailure) != LOCKACQUIRE_NOT_AVAIL);
+								logLockFailure, false) != LOCKACQUIRE_NOT_AVAIL);
 }
 
 /*
@@ -749,7 +766,7 @@ ConditionalXactLockTableWait(TransactionId xid, bool logLockFailure)
 		SET_LOCKTAG_TRANSACTION(tag, xid);
 
 		if (LockAcquireExtended(&tag, ShareLock, false, true, true, NULL,
-								logLockFailure)
+								logLockFailure, false)
 			== LOCKACQUIRE_NOT_AVAIL)
 			return false;
 
@@ -1043,7 +1060,7 @@ ConditionalLockDatabaseObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -1123,7 +1140,7 @@ ConditionalLockSharedObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
diff --git a/src/backend/storage/lmgr/lock.c b/src/backend/storage/lmgr/lock.c
index c221fe96889..bcbbf36c091 100644
--- a/src/backend/storage/lmgr/lock.c
+++ b/src/backend/storage/lmgr/lock.c
@@ -810,7 +810,7 @@ LockAcquire(const LOCKTAG *locktag,
 			bool dontWait)
 {
 	return LockAcquireExtended(locktag, lockmode, sessionLock, dontWait,
-							   true, NULL, false);
+							   true, NULL, false, false);
 }
 
 /*
@@ -829,6 +829,9 @@ LockAcquire(const LOCKTAG *locktag,
  *
  * logLockFailure indicates whether to log details when a lock acquisition
  * fails with dontWait = true.
+ *
+ * isUpgrade should be true if the backend already holds this lock in a lower
+ * mode.
  */
 LockAcquireResult
 LockAcquireExtended(const LOCKTAG *locktag,
@@ -837,7 +840,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 					bool dontWait,
 					bool reportMemoryError,
 					LOCALLOCK **locallockp,
-					bool logLockFailure)
+					bool logLockFailure,
+					bool isUpgrade)
 {
 	LOCKMETHODID lockmethodid = locktag->locktag_lockmethodid;
 	LockMethod	lockMethodTable;
@@ -1119,7 +1123,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		 * case, because JoinWaitQueue() may discover that we can acquire the
 		 * lock immediately after all.
 		 */
-		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait);
+		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait,
+								   isUpgrade);
 	}
 
 	if (waitResult == PROC_WAIT_STATUS_ERROR)
@@ -1225,8 +1230,85 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		Assert(!dontWait);
 		PROCLOCK_PRINT("LockAcquire: sleeping on lock", proclock);
 		LOCK_PRINT("LockAcquire: sleeping on lock", lock, lockmode);
+
+		/*
+		 * Lock upgrade can introduce deadlock, therefore enforce special
+		 * behavior of other processes that deal with this lock. In
+		 * particular, any waiter that sees upgradedBy set is expected to
+		 * perform deadlock check as soon as it's woken up. (Since we are
+		 * already in the queue, the deadlock detector has all the information
+		 * it needs.)
+		 */
+		if (isUpgrade)
+		{
+			dlist_iter	iter;
+
+			/*
+			 * There should not be multiple upgrades at the same time. XXX
+			 * ERROR ?
+			 */
+			Assert(lock->upgradedBy == NULL);
+
+			lock->upgradedBy = MyProc;
+
+			/*
+			 * Now, before I sleep myself, wake up all the existing waiters
+			 * (except for me), so they check for deadlock.
+			 */
+			dclist_foreach(iter, &lock->waitProcs)
+			{
+				PGPROC	   *proc = dlist_container(PGPROC, waitLink,
+												   iter.cur);
+
+				if (proc != MyProc)
+					SetLatch(&proc->procLatch);
+			}
+		}
 		LWLockRelease(partitionLock);
 
+		if (!isUpgrade)
+		{
+			HASH_SEQ_STATUS status;
+			LOCALLOCK	*llock;
+			bool	check_deadlock = false;
+
+			/*
+			 * Likewise, new waiters should check the isUpgrade flag and
+			 * engage deadlock detector if needed. The lock being acquired
+			 * should already be in the hash.
+			 *
+			 * Note that, to make deadlock detection happen as soon as
+			 * possible (i.e. regardless deadlock timeout), we check the other
+			 * locks too. XXX There might be ways to get into the deadlock
+			 * indirectly, but that needs more analysis. As long as the
+			 * upgrading process skips deadlock detection altogether, the
+			 * worst consequence of missing some lock wait cycle here is that
+			 * the other process will be kicked-off no sooner than after its
+			 * deadlock timeout has elapsed. (Which in turn means that the
+			 * lock upgrade will take longer than expected.)
+			 */
+			hash_seq_init(&status, LockMethodLocalHash);
+			while ((llock = (LOCALLOCK *) hash_seq_search(&status)) != NULL)
+			{
+				LOCK	*mylock = llock->lock;
+
+				if (mylock && mylock->upgradedBy)
+				{
+					check_deadlock = true;
+					break;
+				}
+			}
+			/* Check for deadlock if needed. */
+			if (check_deadlock)
+			{
+				DeadLockState	deadlock_state;
+
+				deadlock_state = CheckDeadLock();
+				if (deadlock_state == DS_HARD_DEADLOCK)
+					DeadLockReport();
+			}
+		}
+
 		waitResult = WaitOnLock(locallock, owner);
 
 		/*
@@ -1245,6 +1327,18 @@ LockAcquireExtended(const LOCKTAG *locktag,
 			DeadLockReport();
 			/* DeadLockReport() will not return */
 		}
+
+		/*
+		 * If finishing the lock upgrade, we should not get into a deadlock
+		 * anymore, so let others know that they do not have to care either.
+		 */
+		if (isUpgrade)
+		{
+			LWLockAcquire(partitionLock, LW_EXCLUSIVE);
+			Assert(lock->upgradedBy != NULL);
+			lock->upgradedBy = NULL;
+			LWLockRelease(partitionLock);
+		}
 	}
 	else
 		LWLockRelease(partitionLock);
@@ -1322,6 +1416,7 @@ SetupLockInTable(LockMethod lockMethodTable, PGPROC *proc,
 		lock->nGranted = 0;
 		MemSet(lock->requested, 0, sizeof(int) * MAX_LOCKMODES);
 		MemSet(lock->granted, 0, sizeof(int) * MAX_LOCKMODES);
+		lock->upgradedBy = NULL;
 		LOCK_PRINT("LockAcquire: new", lock, lockmode);
 	}
 	else
@@ -1768,6 +1863,12 @@ CleanUpLock(LOCK *lock, PROCLOCK *proclock,
 			elog(PANIC, "proclock table corrupted");
 	}
 
+	/*
+	 * Was this backend upgrading the lock?
+	 */
+	if (lock->upgradedBy == MyProc)
+		lock->upgradedBy = NULL;
+
 	if (lock->nRequested == 0)
 	{
 		/*
diff --git a/src/backend/storage/lmgr/proc.c b/src/backend/storage/lmgr/proc.c
index 1ac25068d62..7b31370db70 100644
--- a/src/backend/storage/lmgr/proc.c
+++ b/src/backend/storage/lmgr/proc.c
@@ -95,7 +95,6 @@ static volatile sig_atomic_t got_deadlock_timeout;
 static void RemoveProcFromArray(int code, Datum arg);
 static void ProcKill(int code, Datum arg);
 static void AuxiliaryProcKill(int code, Datum arg);
-static DeadLockState CheckDeadLock(void);
 
 
 /*
@@ -1143,7 +1142,8 @@ AuxiliaryPidGetProc(int pid)
  * NOTES: The process queue is now a priority queue for locking.
  */
 ProcWaitStatus
-JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
+JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait,
+			  bool isUpgrade)
 {
 	LOCKMODE	lockmode = locallock->tag.mode;
 	LOCK	   *lock = locallock->lock;
@@ -1230,8 +1230,12 @@ JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
 			/* Must he wait for me? */
 			if (lockMethodTable->conflictTab[proc->waitLockMode] & myHeldLocks)
 			{
-				/* Must I wait for him ? */
-				if (lockMethodTable->conflictTab[lockmode] & proc->heldLocks)
+				/*
+				 * Must I wait for him? I don't want a deadlock during lock
+				 * upgrade - other processes should fail on it.
+				 */
+				if ((lockMethodTable->conflictTab[lockmode] & proc->heldLocks) &&
+					!isUpgrade)
 				{
 					/*
 					 * Yes, so we have a deadlock.  Easiest way to clean up
@@ -1461,8 +1465,22 @@ ProcSleep(LOCALLOCK *locallock)
 			(void) WaitLatch(MyLatch, WL_LATCH_SET | WL_EXIT_ON_PM_DEATH, 0,
 							 PG_WAIT_LOCK | locallock->tag.lock.locktag_type);
 			ResetLatch(MyLatch);
-			/* check for deadlocks first, as that's probably log-worthy */
-			if (got_deadlock_timeout)
+			/*
+			 * Check for deadlocks first, as that's probably log-worthy.
+			 *
+			 * Do not wait for the timeout if the lock is being upgraded since
+			 * the risk of deadlock is higher now. However, do not check for
+			 * deadlock if the lock is being upgraded by this process - other
+			 * processes should take care.
+			 *
+			 * TODO Possible optimization: if this is the only lock of the
+			 * backend and if it did not have any weaker lock on the table so
+			 * far, it should be safe to skip the deadlock check. However, to
+			 * evaluate the situation, we need to take fast-path locks into
+			 * account.
+			 */
+			if ((got_deadlock_timeout || lock->upgradedBy) &&
+				lock->upgradedBy != MyProc)
 			{
 				deadlock_state = CheckDeadLock();
 				got_deadlock_timeout = false;
@@ -1819,7 +1837,7 @@ ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock)
  * not, just return.  If we have a real deadlock, remove ourselves from the
  * lock's wait queue.
  */
-static DeadLockState
+DeadLockState
 CheckDeadLock(void)
 {
 	int			i;
diff --git a/src/include/storage/lmgr.h b/src/include/storage/lmgr.h
index 2a985ce5e15..eddbe1e8622 100644
--- a/src/include/storage/lmgr.h
+++ b/src/include/storage/lmgr.h
@@ -38,6 +38,7 @@ extern void RelationInitLockInfo(Relation relation);
 
 /* Lock a relation */
 extern void LockRelationOid(Oid relid, LOCKMODE lockmode);
+extern void LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode);
 extern void LockRelationId(LockRelId *relid, LOCKMODE lockmode);
 extern bool ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode);
 extern void UnlockRelationId(LockRelId *relid, LOCKMODE lockmode);
diff --git a/src/include/storage/lock.h b/src/include/storage/lock.h
index ee3cb1dc203..080b69eddfd 100644
--- a/src/include/storage/lock.h
+++ b/src/include/storage/lock.h
@@ -131,6 +131,7 @@ typedef const LockMethodData *LockMethod;
  * nRequested -- total requested locks of all types.
  * granted -- count of each lock type currently granted on the lock.
  * nGranted -- total granted locks of all types.
+ * upgrader -- process that intends to upgrade the lock
  *
  * Note: these counts count 1 for each backend.  Internally to a backend,
  * there may be multiple grabs on a particular lock, but this is not reflected
@@ -150,6 +151,7 @@ typedef struct LOCK
 	int			nRequested;		/* total of requested[] array */
 	int			granted[MAX_LOCKMODES]; /* counts of granted locks */
 	int			nGranted;		/* total of granted[] array */
+	PGPROC		*upgradedBy;	/* is lock being upgraded by this process? */
 } LOCK;
 
 #define LOCK_LOCKMETHOD(lock) ((LOCKMETHODID) (lock).tag.locktag_lockmethodid)
@@ -390,7 +392,8 @@ extern LockAcquireResult LockAcquireExtended(const LOCKTAG *locktag,
 											 bool dontWait,
 											 bool reportMemoryError,
 											 LOCALLOCK **locallockp,
-											 bool logLockFailure);
+											 bool logLockFailure,
+											 bool isUpgrade);
 extern void AbortStrongLockAcquire(void);
 extern void MarkLockClear(LOCALLOCK *locallock);
 extern bool LockRelease(const LOCKTAG *locktag,
diff --git a/src/include/storage/proc.h b/src/include/storage/proc.h
index 3e1d1fad5f9..c4bc3a24044 100644
--- a/src/include/storage/proc.h
+++ b/src/include/storage/proc.h
@@ -563,10 +563,12 @@ extern bool HaveNFreeProcs(int n, int *nfree);
 extern void ProcReleaseLocks(bool isCommit);
 
 extern ProcWaitStatus JoinWaitQueue(LOCALLOCK *locallock,
-									LockMethod lockMethodTable, bool dontWait);
+									LockMethod lockMethodTable, bool dontWait,
+									bool isUpgrade);
 extern ProcWaitStatus ProcSleep(LOCALLOCK *locallock);
 extern void ProcWakeup(PGPROC *proc, ProcWaitStatus waitStatus);
 extern void ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock);
+extern DeadLockState CheckDeadLock(void);
 extern void CheckDeadLockAlert(void);
 extern void LockErrorCleanup(void);
 extern void GetLockHoldersAndWaiters(LOCALLOCK *locallock,
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 966+ messages in thread

* [PATCH] Lock upgrade without deadlocks.
@ 2026-04-17 13:13  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 966+ messages in thread

From: Antonin Houska @ 2026-04-17 13:13 UTC (permalink / raw)

For REPACK (CONCURRENTLY), it's essential to upgrade ShareUpdateExclusive lock
to AccessExclusiveLock for the final stage of table processing. However, we
cannot release the earlier before requesting the latter, because if another
process ran DDL on the table in between, REPACK would have to cancel its
transaction, and waste possibly a lot of work. Without releasing
ShareUpdateExclusive lock, the upgrade can cause deadlock, and it's not really
deterministic whether REPACK or the other process will get canceled.

This patch adds a new field to the LOCK structure which tells which process is
trying to upgrade the lock - this field is set right before the upgrade. All
processes waiting in the lock queue are then woken up to check the flag.

When another process tries to get the lock after that, it checks this field,
and if it's set, it checks for deadlocks even if deadlock timeout hasn't
expired yet. If the process is already in the lock's queue and sleeping, the
lock upgrading process wakes it up so it checks the flag immediately.

At the time the upgrading process starts to sleep, all the other process
should be aware that they need to check for deadlock during their lock
acquisitions, so the upgrading process does not have to check for deadlocks
when it gets woken up. Thus it should never receive deadlock error report.
---
 src/backend/commands/repack.c   |   4 +-
 src/backend/storage/lmgr/lmgr.c |  35 ++++++++---
 src/backend/storage/lmgr/lock.c | 107 +++++++++++++++++++++++++++++++-
 src/backend/storage/lmgr/proc.c |  32 +++++++---
 src/include/storage/lmgr.h      |   1 +
 src/include/storage/lock.h      |   5 +-
 src/include/storage/proc.h      |   4 +-
 7 files changed, 166 insertions(+), 22 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index 58e3867246f..310b2a65099 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -3055,8 +3055,10 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Acquire AccessExclusiveLock on the table, its TOAST relation (if there
 	 * is one), all its indexes, so that we can swap the files.
+	 *
+	 * TODO The same for indexes and TOAST?
 	 */
-	LockRelationOid(old_table_oid, AccessExclusiveLock);
+	LockRelationOidUpgrade(old_table_oid, AccessExclusiveLock);
 
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
diff --git a/src/backend/storage/lmgr/lmgr.c b/src/backend/storage/lmgr/lmgr.c
index 2ccf7237fee..0b103485d54 100644
--- a/src/backend/storage/lmgr/lmgr.c
+++ b/src/backend/storage/lmgr/lmgr.c
@@ -58,6 +58,8 @@ typedef struct XactLockTableWaitInfo
 	const ItemPointerData *ctid;
 } XactLockTableWaitInfo;
 
+static void LockRelationOidCommon(Oid relid, LOCKMODE lockmode,
+								  bool isUpgrade);
 static void XactLockTableWaitErrorCb(void *arg);
 
 /*
@@ -105,6 +107,21 @@ SetLocktagRelationOid(LOCKTAG *tag, Oid relid)
  */
 void
 LockRelationOid(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, false);
+}
+
+/*
+ * Like above, but upgrade an existing lock.
+ */
+void
+LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, true);
+}
+
+static void
+LockRelationOidCommon(Oid relid, LOCKMODE lockmode, bool isUpgrade)
 {
 	LOCKTAG		tag;
 	LOCALLOCK  *locallock;
@@ -113,7 +130,7 @@ LockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, isUpgrade);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages, so that we
@@ -157,7 +174,7 @@ ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -191,7 +208,7 @@ LockRelationId(LockRelId *relid, LOCKMODE lockmode)
 	SET_LOCKTAG_RELATION(tag, relid->dbId, relid->relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -254,7 +271,7 @@ LockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -286,7 +303,7 @@ ConditionalLockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -591,7 +608,7 @@ ConditionalLockTuple(Relation relation, const ItemPointerData *tid, LOCKMODE loc
 					  ItemPointerGetOffsetNumber(tid));
 
 	return (LockAcquireExtended(&tag, lockmode, false, true, true, NULL,
-								logLockFailure) != LOCKACQUIRE_NOT_AVAIL);
+								logLockFailure, false) != LOCKACQUIRE_NOT_AVAIL);
 }
 
 /*
@@ -749,7 +766,7 @@ ConditionalXactLockTableWait(TransactionId xid, bool logLockFailure)
 		SET_LOCKTAG_TRANSACTION(tag, xid);
 
 		if (LockAcquireExtended(&tag, ShareLock, false, true, true, NULL,
-								logLockFailure)
+								logLockFailure, false)
 			== LOCKACQUIRE_NOT_AVAIL)
 			return false;
 
@@ -1043,7 +1060,7 @@ ConditionalLockDatabaseObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -1123,7 +1140,7 @@ ConditionalLockSharedObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
diff --git a/src/backend/storage/lmgr/lock.c b/src/backend/storage/lmgr/lock.c
index c221fe96889..bcbbf36c091 100644
--- a/src/backend/storage/lmgr/lock.c
+++ b/src/backend/storage/lmgr/lock.c
@@ -810,7 +810,7 @@ LockAcquire(const LOCKTAG *locktag,
 			bool dontWait)
 {
 	return LockAcquireExtended(locktag, lockmode, sessionLock, dontWait,
-							   true, NULL, false);
+							   true, NULL, false, false);
 }
 
 /*
@@ -829,6 +829,9 @@ LockAcquire(const LOCKTAG *locktag,
  *
  * logLockFailure indicates whether to log details when a lock acquisition
  * fails with dontWait = true.
+ *
+ * isUpgrade should be true if the backend already holds this lock in a lower
+ * mode.
  */
 LockAcquireResult
 LockAcquireExtended(const LOCKTAG *locktag,
@@ -837,7 +840,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 					bool dontWait,
 					bool reportMemoryError,
 					LOCALLOCK **locallockp,
-					bool logLockFailure)
+					bool logLockFailure,
+					bool isUpgrade)
 {
 	LOCKMETHODID lockmethodid = locktag->locktag_lockmethodid;
 	LockMethod	lockMethodTable;
@@ -1119,7 +1123,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		 * case, because JoinWaitQueue() may discover that we can acquire the
 		 * lock immediately after all.
 		 */
-		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait);
+		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait,
+								   isUpgrade);
 	}
 
 	if (waitResult == PROC_WAIT_STATUS_ERROR)
@@ -1225,8 +1230,85 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		Assert(!dontWait);
 		PROCLOCK_PRINT("LockAcquire: sleeping on lock", proclock);
 		LOCK_PRINT("LockAcquire: sleeping on lock", lock, lockmode);
+
+		/*
+		 * Lock upgrade can introduce deadlock, therefore enforce special
+		 * behavior of other processes that deal with this lock. In
+		 * particular, any waiter that sees upgradedBy set is expected to
+		 * perform deadlock check as soon as it's woken up. (Since we are
+		 * already in the queue, the deadlock detector has all the information
+		 * it needs.)
+		 */
+		if (isUpgrade)
+		{
+			dlist_iter	iter;
+
+			/*
+			 * There should not be multiple upgrades at the same time. XXX
+			 * ERROR ?
+			 */
+			Assert(lock->upgradedBy == NULL);
+
+			lock->upgradedBy = MyProc;
+
+			/*
+			 * Now, before I sleep myself, wake up all the existing waiters
+			 * (except for me), so they check for deadlock.
+			 */
+			dclist_foreach(iter, &lock->waitProcs)
+			{
+				PGPROC	   *proc = dlist_container(PGPROC, waitLink,
+												   iter.cur);
+
+				if (proc != MyProc)
+					SetLatch(&proc->procLatch);
+			}
+		}
 		LWLockRelease(partitionLock);
 
+		if (!isUpgrade)
+		{
+			HASH_SEQ_STATUS status;
+			LOCALLOCK	*llock;
+			bool	check_deadlock = false;
+
+			/*
+			 * Likewise, new waiters should check the isUpgrade flag and
+			 * engage deadlock detector if needed. The lock being acquired
+			 * should already be in the hash.
+			 *
+			 * Note that, to make deadlock detection happen as soon as
+			 * possible (i.e. regardless deadlock timeout), we check the other
+			 * locks too. XXX There might be ways to get into the deadlock
+			 * indirectly, but that needs more analysis. As long as the
+			 * upgrading process skips deadlock detection altogether, the
+			 * worst consequence of missing some lock wait cycle here is that
+			 * the other process will be kicked-off no sooner than after its
+			 * deadlock timeout has elapsed. (Which in turn means that the
+			 * lock upgrade will take longer than expected.)
+			 */
+			hash_seq_init(&status, LockMethodLocalHash);
+			while ((llock = (LOCALLOCK *) hash_seq_search(&status)) != NULL)
+			{
+				LOCK	*mylock = llock->lock;
+
+				if (mylock && mylock->upgradedBy)
+				{
+					check_deadlock = true;
+					break;
+				}
+			}
+			/* Check for deadlock if needed. */
+			if (check_deadlock)
+			{
+				DeadLockState	deadlock_state;
+
+				deadlock_state = CheckDeadLock();
+				if (deadlock_state == DS_HARD_DEADLOCK)
+					DeadLockReport();
+			}
+		}
+
 		waitResult = WaitOnLock(locallock, owner);
 
 		/*
@@ -1245,6 +1327,18 @@ LockAcquireExtended(const LOCKTAG *locktag,
 			DeadLockReport();
 			/* DeadLockReport() will not return */
 		}
+
+		/*
+		 * If finishing the lock upgrade, we should not get into a deadlock
+		 * anymore, so let others know that they do not have to care either.
+		 */
+		if (isUpgrade)
+		{
+			LWLockAcquire(partitionLock, LW_EXCLUSIVE);
+			Assert(lock->upgradedBy != NULL);
+			lock->upgradedBy = NULL;
+			LWLockRelease(partitionLock);
+		}
 	}
 	else
 		LWLockRelease(partitionLock);
@@ -1322,6 +1416,7 @@ SetupLockInTable(LockMethod lockMethodTable, PGPROC *proc,
 		lock->nGranted = 0;
 		MemSet(lock->requested, 0, sizeof(int) * MAX_LOCKMODES);
 		MemSet(lock->granted, 0, sizeof(int) * MAX_LOCKMODES);
+		lock->upgradedBy = NULL;
 		LOCK_PRINT("LockAcquire: new", lock, lockmode);
 	}
 	else
@@ -1768,6 +1863,12 @@ CleanUpLock(LOCK *lock, PROCLOCK *proclock,
 			elog(PANIC, "proclock table corrupted");
 	}
 
+	/*
+	 * Was this backend upgrading the lock?
+	 */
+	if (lock->upgradedBy == MyProc)
+		lock->upgradedBy = NULL;
+
 	if (lock->nRequested == 0)
 	{
 		/*
diff --git a/src/backend/storage/lmgr/proc.c b/src/backend/storage/lmgr/proc.c
index 1ac25068d62..7b31370db70 100644
--- a/src/backend/storage/lmgr/proc.c
+++ b/src/backend/storage/lmgr/proc.c
@@ -95,7 +95,6 @@ static volatile sig_atomic_t got_deadlock_timeout;
 static void RemoveProcFromArray(int code, Datum arg);
 static void ProcKill(int code, Datum arg);
 static void AuxiliaryProcKill(int code, Datum arg);
-static DeadLockState CheckDeadLock(void);
 
 
 /*
@@ -1143,7 +1142,8 @@ AuxiliaryPidGetProc(int pid)
  * NOTES: The process queue is now a priority queue for locking.
  */
 ProcWaitStatus
-JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
+JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait,
+			  bool isUpgrade)
 {
 	LOCKMODE	lockmode = locallock->tag.mode;
 	LOCK	   *lock = locallock->lock;
@@ -1230,8 +1230,12 @@ JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
 			/* Must he wait for me? */
 			if (lockMethodTable->conflictTab[proc->waitLockMode] & myHeldLocks)
 			{
-				/* Must I wait for him ? */
-				if (lockMethodTable->conflictTab[lockmode] & proc->heldLocks)
+				/*
+				 * Must I wait for him? I don't want a deadlock during lock
+				 * upgrade - other processes should fail on it.
+				 */
+				if ((lockMethodTable->conflictTab[lockmode] & proc->heldLocks) &&
+					!isUpgrade)
 				{
 					/*
 					 * Yes, so we have a deadlock.  Easiest way to clean up
@@ -1461,8 +1465,22 @@ ProcSleep(LOCALLOCK *locallock)
 			(void) WaitLatch(MyLatch, WL_LATCH_SET | WL_EXIT_ON_PM_DEATH, 0,
 							 PG_WAIT_LOCK | locallock->tag.lock.locktag_type);
 			ResetLatch(MyLatch);
-			/* check for deadlocks first, as that's probably log-worthy */
-			if (got_deadlock_timeout)
+			/*
+			 * Check for deadlocks first, as that's probably log-worthy.
+			 *
+			 * Do not wait for the timeout if the lock is being upgraded since
+			 * the risk of deadlock is higher now. However, do not check for
+			 * deadlock if the lock is being upgraded by this process - other
+			 * processes should take care.
+			 *
+			 * TODO Possible optimization: if this is the only lock of the
+			 * backend and if it did not have any weaker lock on the table so
+			 * far, it should be safe to skip the deadlock check. However, to
+			 * evaluate the situation, we need to take fast-path locks into
+			 * account.
+			 */
+			if ((got_deadlock_timeout || lock->upgradedBy) &&
+				lock->upgradedBy != MyProc)
 			{
 				deadlock_state = CheckDeadLock();
 				got_deadlock_timeout = false;
@@ -1819,7 +1837,7 @@ ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock)
  * not, just return.  If we have a real deadlock, remove ourselves from the
  * lock's wait queue.
  */
-static DeadLockState
+DeadLockState
 CheckDeadLock(void)
 {
 	int			i;
diff --git a/src/include/storage/lmgr.h b/src/include/storage/lmgr.h
index 2a985ce5e15..eddbe1e8622 100644
--- a/src/include/storage/lmgr.h
+++ b/src/include/storage/lmgr.h
@@ -38,6 +38,7 @@ extern void RelationInitLockInfo(Relation relation);
 
 /* Lock a relation */
 extern void LockRelationOid(Oid relid, LOCKMODE lockmode);
+extern void LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode);
 extern void LockRelationId(LockRelId *relid, LOCKMODE lockmode);
 extern bool ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode);
 extern void UnlockRelationId(LockRelId *relid, LOCKMODE lockmode);
diff --git a/src/include/storage/lock.h b/src/include/storage/lock.h
index ee3cb1dc203..080b69eddfd 100644
--- a/src/include/storage/lock.h
+++ b/src/include/storage/lock.h
@@ -131,6 +131,7 @@ typedef const LockMethodData *LockMethod;
  * nRequested -- total requested locks of all types.
  * granted -- count of each lock type currently granted on the lock.
  * nGranted -- total granted locks of all types.
+ * upgrader -- process that intends to upgrade the lock
  *
  * Note: these counts count 1 for each backend.  Internally to a backend,
  * there may be multiple grabs on a particular lock, but this is not reflected
@@ -150,6 +151,7 @@ typedef struct LOCK
 	int			nRequested;		/* total of requested[] array */
 	int			granted[MAX_LOCKMODES]; /* counts of granted locks */
 	int			nGranted;		/* total of granted[] array */
+	PGPROC		*upgradedBy;	/* is lock being upgraded by this process? */
 } LOCK;
 
 #define LOCK_LOCKMETHOD(lock) ((LOCKMETHODID) (lock).tag.locktag_lockmethodid)
@@ -390,7 +392,8 @@ extern LockAcquireResult LockAcquireExtended(const LOCKTAG *locktag,
 											 bool dontWait,
 											 bool reportMemoryError,
 											 LOCALLOCK **locallockp,
-											 bool logLockFailure);
+											 bool logLockFailure,
+											 bool isUpgrade);
 extern void AbortStrongLockAcquire(void);
 extern void MarkLockClear(LOCALLOCK *locallock);
 extern bool LockRelease(const LOCKTAG *locktag,
diff --git a/src/include/storage/proc.h b/src/include/storage/proc.h
index 3e1d1fad5f9..c4bc3a24044 100644
--- a/src/include/storage/proc.h
+++ b/src/include/storage/proc.h
@@ -563,10 +563,12 @@ extern bool HaveNFreeProcs(int n, int *nfree);
 extern void ProcReleaseLocks(bool isCommit);
 
 extern ProcWaitStatus JoinWaitQueue(LOCALLOCK *locallock,
-									LockMethod lockMethodTable, bool dontWait);
+									LockMethod lockMethodTable, bool dontWait,
+									bool isUpgrade);
 extern ProcWaitStatus ProcSleep(LOCALLOCK *locallock);
 extern void ProcWakeup(PGPROC *proc, ProcWaitStatus waitStatus);
 extern void ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock);
+extern DeadLockState CheckDeadLock(void);
 extern void CheckDeadLockAlert(void);
 extern void LockErrorCleanup(void);
 extern void GetLockHoldersAndWaiters(LOCALLOCK *locallock,
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 966+ messages in thread

* [PATCH] Lock upgrade without deadlocks.
@ 2026-04-17 13:13  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 966+ messages in thread

From: Antonin Houska @ 2026-04-17 13:13 UTC (permalink / raw)

For REPACK (CONCURRENTLY), it's essential to upgrade ShareUpdateExclusive lock
to AccessExclusiveLock for the final stage of table processing. However, we
cannot release the earlier before requesting the latter, because if another
process ran DDL on the table in between, REPACK would have to cancel its
transaction, and waste possibly a lot of work. Without releasing
ShareUpdateExclusive lock, the upgrade can cause deadlock, and it's not really
deterministic whether REPACK or the other process will get canceled.

This patch adds a new field to the LOCK structure which tells which process is
trying to upgrade the lock - this field is set right before the upgrade. All
processes waiting in the lock queue are then woken up to check the flag.

When another process tries to get the lock after that, it checks this field,
and if it's set, it checks for deadlocks even if deadlock timeout hasn't
expired yet. If the process is already in the lock's queue and sleeping, the
lock upgrading process wakes it up so it checks the flag immediately.

At the time the upgrading process starts to sleep, all the other process
should be aware that they need to check for deadlock during their lock
acquisitions, so the upgrading process does not have to check for deadlocks
when it gets woken up. Thus it should never receive deadlock error report.
---
 src/backend/commands/repack.c   |   4 +-
 src/backend/storage/lmgr/lmgr.c |  35 ++++++++---
 src/backend/storage/lmgr/lock.c | 107 +++++++++++++++++++++++++++++++-
 src/backend/storage/lmgr/proc.c |  32 +++++++---
 src/include/storage/lmgr.h      |   1 +
 src/include/storage/lock.h      |   5 +-
 src/include/storage/proc.h      |   4 +-
 7 files changed, 166 insertions(+), 22 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index 58e3867246f..310b2a65099 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -3055,8 +3055,10 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Acquire AccessExclusiveLock on the table, its TOAST relation (if there
 	 * is one), all its indexes, so that we can swap the files.
+	 *
+	 * TODO The same for indexes and TOAST?
 	 */
-	LockRelationOid(old_table_oid, AccessExclusiveLock);
+	LockRelationOidUpgrade(old_table_oid, AccessExclusiveLock);
 
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
diff --git a/src/backend/storage/lmgr/lmgr.c b/src/backend/storage/lmgr/lmgr.c
index 2ccf7237fee..0b103485d54 100644
--- a/src/backend/storage/lmgr/lmgr.c
+++ b/src/backend/storage/lmgr/lmgr.c
@@ -58,6 +58,8 @@ typedef struct XactLockTableWaitInfo
 	const ItemPointerData *ctid;
 } XactLockTableWaitInfo;
 
+static void LockRelationOidCommon(Oid relid, LOCKMODE lockmode,
+								  bool isUpgrade);
 static void XactLockTableWaitErrorCb(void *arg);
 
 /*
@@ -105,6 +107,21 @@ SetLocktagRelationOid(LOCKTAG *tag, Oid relid)
  */
 void
 LockRelationOid(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, false);
+}
+
+/*
+ * Like above, but upgrade an existing lock.
+ */
+void
+LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, true);
+}
+
+static void
+LockRelationOidCommon(Oid relid, LOCKMODE lockmode, bool isUpgrade)
 {
 	LOCKTAG		tag;
 	LOCALLOCK  *locallock;
@@ -113,7 +130,7 @@ LockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, isUpgrade);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages, so that we
@@ -157,7 +174,7 @@ ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -191,7 +208,7 @@ LockRelationId(LockRelId *relid, LOCKMODE lockmode)
 	SET_LOCKTAG_RELATION(tag, relid->dbId, relid->relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -254,7 +271,7 @@ LockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -286,7 +303,7 @@ ConditionalLockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -591,7 +608,7 @@ ConditionalLockTuple(Relation relation, const ItemPointerData *tid, LOCKMODE loc
 					  ItemPointerGetOffsetNumber(tid));
 
 	return (LockAcquireExtended(&tag, lockmode, false, true, true, NULL,
-								logLockFailure) != LOCKACQUIRE_NOT_AVAIL);
+								logLockFailure, false) != LOCKACQUIRE_NOT_AVAIL);
 }
 
 /*
@@ -749,7 +766,7 @@ ConditionalXactLockTableWait(TransactionId xid, bool logLockFailure)
 		SET_LOCKTAG_TRANSACTION(tag, xid);
 
 		if (LockAcquireExtended(&tag, ShareLock, false, true, true, NULL,
-								logLockFailure)
+								logLockFailure, false)
 			== LOCKACQUIRE_NOT_AVAIL)
 			return false;
 
@@ -1043,7 +1060,7 @@ ConditionalLockDatabaseObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -1123,7 +1140,7 @@ ConditionalLockSharedObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
diff --git a/src/backend/storage/lmgr/lock.c b/src/backend/storage/lmgr/lock.c
index c221fe96889..bcbbf36c091 100644
--- a/src/backend/storage/lmgr/lock.c
+++ b/src/backend/storage/lmgr/lock.c
@@ -810,7 +810,7 @@ LockAcquire(const LOCKTAG *locktag,
 			bool dontWait)
 {
 	return LockAcquireExtended(locktag, lockmode, sessionLock, dontWait,
-							   true, NULL, false);
+							   true, NULL, false, false);
 }
 
 /*
@@ -829,6 +829,9 @@ LockAcquire(const LOCKTAG *locktag,
  *
  * logLockFailure indicates whether to log details when a lock acquisition
  * fails with dontWait = true.
+ *
+ * isUpgrade should be true if the backend already holds this lock in a lower
+ * mode.
  */
 LockAcquireResult
 LockAcquireExtended(const LOCKTAG *locktag,
@@ -837,7 +840,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 					bool dontWait,
 					bool reportMemoryError,
 					LOCALLOCK **locallockp,
-					bool logLockFailure)
+					bool logLockFailure,
+					bool isUpgrade)
 {
 	LOCKMETHODID lockmethodid = locktag->locktag_lockmethodid;
 	LockMethod	lockMethodTable;
@@ -1119,7 +1123,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		 * case, because JoinWaitQueue() may discover that we can acquire the
 		 * lock immediately after all.
 		 */
-		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait);
+		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait,
+								   isUpgrade);
 	}
 
 	if (waitResult == PROC_WAIT_STATUS_ERROR)
@@ -1225,8 +1230,85 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		Assert(!dontWait);
 		PROCLOCK_PRINT("LockAcquire: sleeping on lock", proclock);
 		LOCK_PRINT("LockAcquire: sleeping on lock", lock, lockmode);
+
+		/*
+		 * Lock upgrade can introduce deadlock, therefore enforce special
+		 * behavior of other processes that deal with this lock. In
+		 * particular, any waiter that sees upgradedBy set is expected to
+		 * perform deadlock check as soon as it's woken up. (Since we are
+		 * already in the queue, the deadlock detector has all the information
+		 * it needs.)
+		 */
+		if (isUpgrade)
+		{
+			dlist_iter	iter;
+
+			/*
+			 * There should not be multiple upgrades at the same time. XXX
+			 * ERROR ?
+			 */
+			Assert(lock->upgradedBy == NULL);
+
+			lock->upgradedBy = MyProc;
+
+			/*
+			 * Now, before I sleep myself, wake up all the existing waiters
+			 * (except for me), so they check for deadlock.
+			 */
+			dclist_foreach(iter, &lock->waitProcs)
+			{
+				PGPROC	   *proc = dlist_container(PGPROC, waitLink,
+												   iter.cur);
+
+				if (proc != MyProc)
+					SetLatch(&proc->procLatch);
+			}
+		}
 		LWLockRelease(partitionLock);
 
+		if (!isUpgrade)
+		{
+			HASH_SEQ_STATUS status;
+			LOCALLOCK	*llock;
+			bool	check_deadlock = false;
+
+			/*
+			 * Likewise, new waiters should check the isUpgrade flag and
+			 * engage deadlock detector if needed. The lock being acquired
+			 * should already be in the hash.
+			 *
+			 * Note that, to make deadlock detection happen as soon as
+			 * possible (i.e. regardless deadlock timeout), we check the other
+			 * locks too. XXX There might be ways to get into the deadlock
+			 * indirectly, but that needs more analysis. As long as the
+			 * upgrading process skips deadlock detection altogether, the
+			 * worst consequence of missing some lock wait cycle here is that
+			 * the other process will be kicked-off no sooner than after its
+			 * deadlock timeout has elapsed. (Which in turn means that the
+			 * lock upgrade will take longer than expected.)
+			 */
+			hash_seq_init(&status, LockMethodLocalHash);
+			while ((llock = (LOCALLOCK *) hash_seq_search(&status)) != NULL)
+			{
+				LOCK	*mylock = llock->lock;
+
+				if (mylock && mylock->upgradedBy)
+				{
+					check_deadlock = true;
+					break;
+				}
+			}
+			/* Check for deadlock if needed. */
+			if (check_deadlock)
+			{
+				DeadLockState	deadlock_state;
+
+				deadlock_state = CheckDeadLock();
+				if (deadlock_state == DS_HARD_DEADLOCK)
+					DeadLockReport();
+			}
+		}
+
 		waitResult = WaitOnLock(locallock, owner);
 
 		/*
@@ -1245,6 +1327,18 @@ LockAcquireExtended(const LOCKTAG *locktag,
 			DeadLockReport();
 			/* DeadLockReport() will not return */
 		}
+
+		/*
+		 * If finishing the lock upgrade, we should not get into a deadlock
+		 * anymore, so let others know that they do not have to care either.
+		 */
+		if (isUpgrade)
+		{
+			LWLockAcquire(partitionLock, LW_EXCLUSIVE);
+			Assert(lock->upgradedBy != NULL);
+			lock->upgradedBy = NULL;
+			LWLockRelease(partitionLock);
+		}
 	}
 	else
 		LWLockRelease(partitionLock);
@@ -1322,6 +1416,7 @@ SetupLockInTable(LockMethod lockMethodTable, PGPROC *proc,
 		lock->nGranted = 0;
 		MemSet(lock->requested, 0, sizeof(int) * MAX_LOCKMODES);
 		MemSet(lock->granted, 0, sizeof(int) * MAX_LOCKMODES);
+		lock->upgradedBy = NULL;
 		LOCK_PRINT("LockAcquire: new", lock, lockmode);
 	}
 	else
@@ -1768,6 +1863,12 @@ CleanUpLock(LOCK *lock, PROCLOCK *proclock,
 			elog(PANIC, "proclock table corrupted");
 	}
 
+	/*
+	 * Was this backend upgrading the lock?
+	 */
+	if (lock->upgradedBy == MyProc)
+		lock->upgradedBy = NULL;
+
 	if (lock->nRequested == 0)
 	{
 		/*
diff --git a/src/backend/storage/lmgr/proc.c b/src/backend/storage/lmgr/proc.c
index 1ac25068d62..7b31370db70 100644
--- a/src/backend/storage/lmgr/proc.c
+++ b/src/backend/storage/lmgr/proc.c
@@ -95,7 +95,6 @@ static volatile sig_atomic_t got_deadlock_timeout;
 static void RemoveProcFromArray(int code, Datum arg);
 static void ProcKill(int code, Datum arg);
 static void AuxiliaryProcKill(int code, Datum arg);
-static DeadLockState CheckDeadLock(void);
 
 
 /*
@@ -1143,7 +1142,8 @@ AuxiliaryPidGetProc(int pid)
  * NOTES: The process queue is now a priority queue for locking.
  */
 ProcWaitStatus
-JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
+JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait,
+			  bool isUpgrade)
 {
 	LOCKMODE	lockmode = locallock->tag.mode;
 	LOCK	   *lock = locallock->lock;
@@ -1230,8 +1230,12 @@ JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
 			/* Must he wait for me? */
 			if (lockMethodTable->conflictTab[proc->waitLockMode] & myHeldLocks)
 			{
-				/* Must I wait for him ? */
-				if (lockMethodTable->conflictTab[lockmode] & proc->heldLocks)
+				/*
+				 * Must I wait for him? I don't want a deadlock during lock
+				 * upgrade - other processes should fail on it.
+				 */
+				if ((lockMethodTable->conflictTab[lockmode] & proc->heldLocks) &&
+					!isUpgrade)
 				{
 					/*
 					 * Yes, so we have a deadlock.  Easiest way to clean up
@@ -1461,8 +1465,22 @@ ProcSleep(LOCALLOCK *locallock)
 			(void) WaitLatch(MyLatch, WL_LATCH_SET | WL_EXIT_ON_PM_DEATH, 0,
 							 PG_WAIT_LOCK | locallock->tag.lock.locktag_type);
 			ResetLatch(MyLatch);
-			/* check for deadlocks first, as that's probably log-worthy */
-			if (got_deadlock_timeout)
+			/*
+			 * Check for deadlocks first, as that's probably log-worthy.
+			 *
+			 * Do not wait for the timeout if the lock is being upgraded since
+			 * the risk of deadlock is higher now. However, do not check for
+			 * deadlock if the lock is being upgraded by this process - other
+			 * processes should take care.
+			 *
+			 * TODO Possible optimization: if this is the only lock of the
+			 * backend and if it did not have any weaker lock on the table so
+			 * far, it should be safe to skip the deadlock check. However, to
+			 * evaluate the situation, we need to take fast-path locks into
+			 * account.
+			 */
+			if ((got_deadlock_timeout || lock->upgradedBy) &&
+				lock->upgradedBy != MyProc)
 			{
 				deadlock_state = CheckDeadLock();
 				got_deadlock_timeout = false;
@@ -1819,7 +1837,7 @@ ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock)
  * not, just return.  If we have a real deadlock, remove ourselves from the
  * lock's wait queue.
  */
-static DeadLockState
+DeadLockState
 CheckDeadLock(void)
 {
 	int			i;
diff --git a/src/include/storage/lmgr.h b/src/include/storage/lmgr.h
index 2a985ce5e15..eddbe1e8622 100644
--- a/src/include/storage/lmgr.h
+++ b/src/include/storage/lmgr.h
@@ -38,6 +38,7 @@ extern void RelationInitLockInfo(Relation relation);
 
 /* Lock a relation */
 extern void LockRelationOid(Oid relid, LOCKMODE lockmode);
+extern void LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode);
 extern void LockRelationId(LockRelId *relid, LOCKMODE lockmode);
 extern bool ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode);
 extern void UnlockRelationId(LockRelId *relid, LOCKMODE lockmode);
diff --git a/src/include/storage/lock.h b/src/include/storage/lock.h
index ee3cb1dc203..080b69eddfd 100644
--- a/src/include/storage/lock.h
+++ b/src/include/storage/lock.h
@@ -131,6 +131,7 @@ typedef const LockMethodData *LockMethod;
  * nRequested -- total requested locks of all types.
  * granted -- count of each lock type currently granted on the lock.
  * nGranted -- total granted locks of all types.
+ * upgrader -- process that intends to upgrade the lock
  *
  * Note: these counts count 1 for each backend.  Internally to a backend,
  * there may be multiple grabs on a particular lock, but this is not reflected
@@ -150,6 +151,7 @@ typedef struct LOCK
 	int			nRequested;		/* total of requested[] array */
 	int			granted[MAX_LOCKMODES]; /* counts of granted locks */
 	int			nGranted;		/* total of granted[] array */
+	PGPROC		*upgradedBy;	/* is lock being upgraded by this process? */
 } LOCK;
 
 #define LOCK_LOCKMETHOD(lock) ((LOCKMETHODID) (lock).tag.locktag_lockmethodid)
@@ -390,7 +392,8 @@ extern LockAcquireResult LockAcquireExtended(const LOCKTAG *locktag,
 											 bool dontWait,
 											 bool reportMemoryError,
 											 LOCALLOCK **locallockp,
-											 bool logLockFailure);
+											 bool logLockFailure,
+											 bool isUpgrade);
 extern void AbortStrongLockAcquire(void);
 extern void MarkLockClear(LOCALLOCK *locallock);
 extern bool LockRelease(const LOCKTAG *locktag,
diff --git a/src/include/storage/proc.h b/src/include/storage/proc.h
index 3e1d1fad5f9..c4bc3a24044 100644
--- a/src/include/storage/proc.h
+++ b/src/include/storage/proc.h
@@ -563,10 +563,12 @@ extern bool HaveNFreeProcs(int n, int *nfree);
 extern void ProcReleaseLocks(bool isCommit);
 
 extern ProcWaitStatus JoinWaitQueue(LOCALLOCK *locallock,
-									LockMethod lockMethodTable, bool dontWait);
+									LockMethod lockMethodTable, bool dontWait,
+									bool isUpgrade);
 extern ProcWaitStatus ProcSleep(LOCALLOCK *locallock);
 extern void ProcWakeup(PGPROC *proc, ProcWaitStatus waitStatus);
 extern void ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock);
+extern DeadLockState CheckDeadLock(void);
 extern void CheckDeadLockAlert(void);
 extern void LockErrorCleanup(void);
 extern void GetLockHoldersAndWaiters(LOCALLOCK *locallock,
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 966+ messages in thread

* [PATCH] Lock upgrade without deadlocks.
@ 2026-04-17 13:13  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 966+ messages in thread

From: Antonin Houska @ 2026-04-17 13:13 UTC (permalink / raw)

For REPACK (CONCURRENTLY), it's essential to upgrade ShareUpdateExclusive lock
to AccessExclusiveLock for the final stage of table processing. However, we
cannot release the earlier before requesting the latter, because if another
process ran DDL on the table in between, REPACK would have to cancel its
transaction, and waste possibly a lot of work. Without releasing
ShareUpdateExclusive lock, the upgrade can cause deadlock, and it's not really
deterministic whether REPACK or the other process will get canceled.

This patch adds a new field to the LOCK structure which tells which process is
trying to upgrade the lock - this field is set right before the upgrade. All
processes waiting in the lock queue are then woken up to check the flag.

When another process tries to get the lock after that, it checks this field,
and if it's set, it checks for deadlocks even if deadlock timeout hasn't
expired yet. If the process is already in the lock's queue and sleeping, the
lock upgrading process wakes it up so it checks the flag immediately.

At the time the upgrading process starts to sleep, all the other process
should be aware that they need to check for deadlock during their lock
acquisitions, so the upgrading process does not have to check for deadlocks
when it gets woken up. Thus it should never receive deadlock error report.
---
 src/backend/commands/repack.c   |   4 +-
 src/backend/storage/lmgr/lmgr.c |  35 ++++++++---
 src/backend/storage/lmgr/lock.c | 107 +++++++++++++++++++++++++++++++-
 src/backend/storage/lmgr/proc.c |  32 +++++++---
 src/include/storage/lmgr.h      |   1 +
 src/include/storage/lock.h      |   5 +-
 src/include/storage/proc.h      |   4 +-
 7 files changed, 166 insertions(+), 22 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index 58e3867246f..310b2a65099 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -3055,8 +3055,10 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Acquire AccessExclusiveLock on the table, its TOAST relation (if there
 	 * is one), all its indexes, so that we can swap the files.
+	 *
+	 * TODO The same for indexes and TOAST?
 	 */
-	LockRelationOid(old_table_oid, AccessExclusiveLock);
+	LockRelationOidUpgrade(old_table_oid, AccessExclusiveLock);
 
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
diff --git a/src/backend/storage/lmgr/lmgr.c b/src/backend/storage/lmgr/lmgr.c
index 2ccf7237fee..0b103485d54 100644
--- a/src/backend/storage/lmgr/lmgr.c
+++ b/src/backend/storage/lmgr/lmgr.c
@@ -58,6 +58,8 @@ typedef struct XactLockTableWaitInfo
 	const ItemPointerData *ctid;
 } XactLockTableWaitInfo;
 
+static void LockRelationOidCommon(Oid relid, LOCKMODE lockmode,
+								  bool isUpgrade);
 static void XactLockTableWaitErrorCb(void *arg);
 
 /*
@@ -105,6 +107,21 @@ SetLocktagRelationOid(LOCKTAG *tag, Oid relid)
  */
 void
 LockRelationOid(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, false);
+}
+
+/*
+ * Like above, but upgrade an existing lock.
+ */
+void
+LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, true);
+}
+
+static void
+LockRelationOidCommon(Oid relid, LOCKMODE lockmode, bool isUpgrade)
 {
 	LOCKTAG		tag;
 	LOCALLOCK  *locallock;
@@ -113,7 +130,7 @@ LockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, isUpgrade);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages, so that we
@@ -157,7 +174,7 @@ ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -191,7 +208,7 @@ LockRelationId(LockRelId *relid, LOCKMODE lockmode)
 	SET_LOCKTAG_RELATION(tag, relid->dbId, relid->relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -254,7 +271,7 @@ LockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -286,7 +303,7 @@ ConditionalLockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -591,7 +608,7 @@ ConditionalLockTuple(Relation relation, const ItemPointerData *tid, LOCKMODE loc
 					  ItemPointerGetOffsetNumber(tid));
 
 	return (LockAcquireExtended(&tag, lockmode, false, true, true, NULL,
-								logLockFailure) != LOCKACQUIRE_NOT_AVAIL);
+								logLockFailure, false) != LOCKACQUIRE_NOT_AVAIL);
 }
 
 /*
@@ -749,7 +766,7 @@ ConditionalXactLockTableWait(TransactionId xid, bool logLockFailure)
 		SET_LOCKTAG_TRANSACTION(tag, xid);
 
 		if (LockAcquireExtended(&tag, ShareLock, false, true, true, NULL,
-								logLockFailure)
+								logLockFailure, false)
 			== LOCKACQUIRE_NOT_AVAIL)
 			return false;
 
@@ -1043,7 +1060,7 @@ ConditionalLockDatabaseObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -1123,7 +1140,7 @@ ConditionalLockSharedObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
diff --git a/src/backend/storage/lmgr/lock.c b/src/backend/storage/lmgr/lock.c
index c221fe96889..bcbbf36c091 100644
--- a/src/backend/storage/lmgr/lock.c
+++ b/src/backend/storage/lmgr/lock.c
@@ -810,7 +810,7 @@ LockAcquire(const LOCKTAG *locktag,
 			bool dontWait)
 {
 	return LockAcquireExtended(locktag, lockmode, sessionLock, dontWait,
-							   true, NULL, false);
+							   true, NULL, false, false);
 }
 
 /*
@@ -829,6 +829,9 @@ LockAcquire(const LOCKTAG *locktag,
  *
  * logLockFailure indicates whether to log details when a lock acquisition
  * fails with dontWait = true.
+ *
+ * isUpgrade should be true if the backend already holds this lock in a lower
+ * mode.
  */
 LockAcquireResult
 LockAcquireExtended(const LOCKTAG *locktag,
@@ -837,7 +840,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 					bool dontWait,
 					bool reportMemoryError,
 					LOCALLOCK **locallockp,
-					bool logLockFailure)
+					bool logLockFailure,
+					bool isUpgrade)
 {
 	LOCKMETHODID lockmethodid = locktag->locktag_lockmethodid;
 	LockMethod	lockMethodTable;
@@ -1119,7 +1123,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		 * case, because JoinWaitQueue() may discover that we can acquire the
 		 * lock immediately after all.
 		 */
-		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait);
+		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait,
+								   isUpgrade);
 	}
 
 	if (waitResult == PROC_WAIT_STATUS_ERROR)
@@ -1225,8 +1230,85 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		Assert(!dontWait);
 		PROCLOCK_PRINT("LockAcquire: sleeping on lock", proclock);
 		LOCK_PRINT("LockAcquire: sleeping on lock", lock, lockmode);
+
+		/*
+		 * Lock upgrade can introduce deadlock, therefore enforce special
+		 * behavior of other processes that deal with this lock. In
+		 * particular, any waiter that sees upgradedBy set is expected to
+		 * perform deadlock check as soon as it's woken up. (Since we are
+		 * already in the queue, the deadlock detector has all the information
+		 * it needs.)
+		 */
+		if (isUpgrade)
+		{
+			dlist_iter	iter;
+
+			/*
+			 * There should not be multiple upgrades at the same time. XXX
+			 * ERROR ?
+			 */
+			Assert(lock->upgradedBy == NULL);
+
+			lock->upgradedBy = MyProc;
+
+			/*
+			 * Now, before I sleep myself, wake up all the existing waiters
+			 * (except for me), so they check for deadlock.
+			 */
+			dclist_foreach(iter, &lock->waitProcs)
+			{
+				PGPROC	   *proc = dlist_container(PGPROC, waitLink,
+												   iter.cur);
+
+				if (proc != MyProc)
+					SetLatch(&proc->procLatch);
+			}
+		}
 		LWLockRelease(partitionLock);
 
+		if (!isUpgrade)
+		{
+			HASH_SEQ_STATUS status;
+			LOCALLOCK	*llock;
+			bool	check_deadlock = false;
+
+			/*
+			 * Likewise, new waiters should check the isUpgrade flag and
+			 * engage deadlock detector if needed. The lock being acquired
+			 * should already be in the hash.
+			 *
+			 * Note that, to make deadlock detection happen as soon as
+			 * possible (i.e. regardless deadlock timeout), we check the other
+			 * locks too. XXX There might be ways to get into the deadlock
+			 * indirectly, but that needs more analysis. As long as the
+			 * upgrading process skips deadlock detection altogether, the
+			 * worst consequence of missing some lock wait cycle here is that
+			 * the other process will be kicked-off no sooner than after its
+			 * deadlock timeout has elapsed. (Which in turn means that the
+			 * lock upgrade will take longer than expected.)
+			 */
+			hash_seq_init(&status, LockMethodLocalHash);
+			while ((llock = (LOCALLOCK *) hash_seq_search(&status)) != NULL)
+			{
+				LOCK	*mylock = llock->lock;
+
+				if (mylock && mylock->upgradedBy)
+				{
+					check_deadlock = true;
+					break;
+				}
+			}
+			/* Check for deadlock if needed. */
+			if (check_deadlock)
+			{
+				DeadLockState	deadlock_state;
+
+				deadlock_state = CheckDeadLock();
+				if (deadlock_state == DS_HARD_DEADLOCK)
+					DeadLockReport();
+			}
+		}
+
 		waitResult = WaitOnLock(locallock, owner);
 
 		/*
@@ -1245,6 +1327,18 @@ LockAcquireExtended(const LOCKTAG *locktag,
 			DeadLockReport();
 			/* DeadLockReport() will not return */
 		}
+
+		/*
+		 * If finishing the lock upgrade, we should not get into a deadlock
+		 * anymore, so let others know that they do not have to care either.
+		 */
+		if (isUpgrade)
+		{
+			LWLockAcquire(partitionLock, LW_EXCLUSIVE);
+			Assert(lock->upgradedBy != NULL);
+			lock->upgradedBy = NULL;
+			LWLockRelease(partitionLock);
+		}
 	}
 	else
 		LWLockRelease(partitionLock);
@@ -1322,6 +1416,7 @@ SetupLockInTable(LockMethod lockMethodTable, PGPROC *proc,
 		lock->nGranted = 0;
 		MemSet(lock->requested, 0, sizeof(int) * MAX_LOCKMODES);
 		MemSet(lock->granted, 0, sizeof(int) * MAX_LOCKMODES);
+		lock->upgradedBy = NULL;
 		LOCK_PRINT("LockAcquire: new", lock, lockmode);
 	}
 	else
@@ -1768,6 +1863,12 @@ CleanUpLock(LOCK *lock, PROCLOCK *proclock,
 			elog(PANIC, "proclock table corrupted");
 	}
 
+	/*
+	 * Was this backend upgrading the lock?
+	 */
+	if (lock->upgradedBy == MyProc)
+		lock->upgradedBy = NULL;
+
 	if (lock->nRequested == 0)
 	{
 		/*
diff --git a/src/backend/storage/lmgr/proc.c b/src/backend/storage/lmgr/proc.c
index 1ac25068d62..7b31370db70 100644
--- a/src/backend/storage/lmgr/proc.c
+++ b/src/backend/storage/lmgr/proc.c
@@ -95,7 +95,6 @@ static volatile sig_atomic_t got_deadlock_timeout;
 static void RemoveProcFromArray(int code, Datum arg);
 static void ProcKill(int code, Datum arg);
 static void AuxiliaryProcKill(int code, Datum arg);
-static DeadLockState CheckDeadLock(void);
 
 
 /*
@@ -1143,7 +1142,8 @@ AuxiliaryPidGetProc(int pid)
  * NOTES: The process queue is now a priority queue for locking.
  */
 ProcWaitStatus
-JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
+JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait,
+			  bool isUpgrade)
 {
 	LOCKMODE	lockmode = locallock->tag.mode;
 	LOCK	   *lock = locallock->lock;
@@ -1230,8 +1230,12 @@ JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
 			/* Must he wait for me? */
 			if (lockMethodTable->conflictTab[proc->waitLockMode] & myHeldLocks)
 			{
-				/* Must I wait for him ? */
-				if (lockMethodTable->conflictTab[lockmode] & proc->heldLocks)
+				/*
+				 * Must I wait for him? I don't want a deadlock during lock
+				 * upgrade - other processes should fail on it.
+				 */
+				if ((lockMethodTable->conflictTab[lockmode] & proc->heldLocks) &&
+					!isUpgrade)
 				{
 					/*
 					 * Yes, so we have a deadlock.  Easiest way to clean up
@@ -1461,8 +1465,22 @@ ProcSleep(LOCALLOCK *locallock)
 			(void) WaitLatch(MyLatch, WL_LATCH_SET | WL_EXIT_ON_PM_DEATH, 0,
 							 PG_WAIT_LOCK | locallock->tag.lock.locktag_type);
 			ResetLatch(MyLatch);
-			/* check for deadlocks first, as that's probably log-worthy */
-			if (got_deadlock_timeout)
+			/*
+			 * Check for deadlocks first, as that's probably log-worthy.
+			 *
+			 * Do not wait for the timeout if the lock is being upgraded since
+			 * the risk of deadlock is higher now. However, do not check for
+			 * deadlock if the lock is being upgraded by this process - other
+			 * processes should take care.
+			 *
+			 * TODO Possible optimization: if this is the only lock of the
+			 * backend and if it did not have any weaker lock on the table so
+			 * far, it should be safe to skip the deadlock check. However, to
+			 * evaluate the situation, we need to take fast-path locks into
+			 * account.
+			 */
+			if ((got_deadlock_timeout || lock->upgradedBy) &&
+				lock->upgradedBy != MyProc)
 			{
 				deadlock_state = CheckDeadLock();
 				got_deadlock_timeout = false;
@@ -1819,7 +1837,7 @@ ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock)
  * not, just return.  If we have a real deadlock, remove ourselves from the
  * lock's wait queue.
  */
-static DeadLockState
+DeadLockState
 CheckDeadLock(void)
 {
 	int			i;
diff --git a/src/include/storage/lmgr.h b/src/include/storage/lmgr.h
index 2a985ce5e15..eddbe1e8622 100644
--- a/src/include/storage/lmgr.h
+++ b/src/include/storage/lmgr.h
@@ -38,6 +38,7 @@ extern void RelationInitLockInfo(Relation relation);
 
 /* Lock a relation */
 extern void LockRelationOid(Oid relid, LOCKMODE lockmode);
+extern void LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode);
 extern void LockRelationId(LockRelId *relid, LOCKMODE lockmode);
 extern bool ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode);
 extern void UnlockRelationId(LockRelId *relid, LOCKMODE lockmode);
diff --git a/src/include/storage/lock.h b/src/include/storage/lock.h
index ee3cb1dc203..080b69eddfd 100644
--- a/src/include/storage/lock.h
+++ b/src/include/storage/lock.h
@@ -131,6 +131,7 @@ typedef const LockMethodData *LockMethod;
  * nRequested -- total requested locks of all types.
  * granted -- count of each lock type currently granted on the lock.
  * nGranted -- total granted locks of all types.
+ * upgrader -- process that intends to upgrade the lock
  *
  * Note: these counts count 1 for each backend.  Internally to a backend,
  * there may be multiple grabs on a particular lock, but this is not reflected
@@ -150,6 +151,7 @@ typedef struct LOCK
 	int			nRequested;		/* total of requested[] array */
 	int			granted[MAX_LOCKMODES]; /* counts of granted locks */
 	int			nGranted;		/* total of granted[] array */
+	PGPROC		*upgradedBy;	/* is lock being upgraded by this process? */
 } LOCK;
 
 #define LOCK_LOCKMETHOD(lock) ((LOCKMETHODID) (lock).tag.locktag_lockmethodid)
@@ -390,7 +392,8 @@ extern LockAcquireResult LockAcquireExtended(const LOCKTAG *locktag,
 											 bool dontWait,
 											 bool reportMemoryError,
 											 LOCALLOCK **locallockp,
-											 bool logLockFailure);
+											 bool logLockFailure,
+											 bool isUpgrade);
 extern void AbortStrongLockAcquire(void);
 extern void MarkLockClear(LOCALLOCK *locallock);
 extern bool LockRelease(const LOCKTAG *locktag,
diff --git a/src/include/storage/proc.h b/src/include/storage/proc.h
index 3e1d1fad5f9..c4bc3a24044 100644
--- a/src/include/storage/proc.h
+++ b/src/include/storage/proc.h
@@ -563,10 +563,12 @@ extern bool HaveNFreeProcs(int n, int *nfree);
 extern void ProcReleaseLocks(bool isCommit);
 
 extern ProcWaitStatus JoinWaitQueue(LOCALLOCK *locallock,
-									LockMethod lockMethodTable, bool dontWait);
+									LockMethod lockMethodTable, bool dontWait,
+									bool isUpgrade);
 extern ProcWaitStatus ProcSleep(LOCALLOCK *locallock);
 extern void ProcWakeup(PGPROC *proc, ProcWaitStatus waitStatus);
 extern void ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock);
+extern DeadLockState CheckDeadLock(void);
 extern void CheckDeadLockAlert(void);
 extern void LockErrorCleanup(void);
 extern void GetLockHoldersAndWaiters(LOCALLOCK *locallock,
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 966+ messages in thread

* [PATCH] Lock upgrade without deadlocks.
@ 2026-04-17 13:13  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 966+ messages in thread

From: Antonin Houska @ 2026-04-17 13:13 UTC (permalink / raw)

For REPACK (CONCURRENTLY), it's essential to upgrade ShareUpdateExclusive lock
to AccessExclusiveLock for the final stage of table processing. However, we
cannot release the earlier before requesting the latter, because if another
process ran DDL on the table in between, REPACK would have to cancel its
transaction, and waste possibly a lot of work. Without releasing
ShareUpdateExclusive lock, the upgrade can cause deadlock, and it's not really
deterministic whether REPACK or the other process will get canceled.

This patch adds a new field to the LOCK structure which tells which process is
trying to upgrade the lock - this field is set right before the upgrade. All
processes waiting in the lock queue are then woken up to check the flag.

When another process tries to get the lock after that, it checks this field,
and if it's set, it checks for deadlocks even if deadlock timeout hasn't
expired yet. If the process is already in the lock's queue and sleeping, the
lock upgrading process wakes it up so it checks the flag immediately.

At the time the upgrading process starts to sleep, all the other process
should be aware that they need to check for deadlock during their lock
acquisitions, so the upgrading process does not have to check for deadlocks
when it gets woken up. Thus it should never receive deadlock error report.
---
 src/backend/commands/repack.c   |   4 +-
 src/backend/storage/lmgr/lmgr.c |  35 ++++++++---
 src/backend/storage/lmgr/lock.c | 107 +++++++++++++++++++++++++++++++-
 src/backend/storage/lmgr/proc.c |  32 +++++++---
 src/include/storage/lmgr.h      |   1 +
 src/include/storage/lock.h      |   5 +-
 src/include/storage/proc.h      |   4 +-
 7 files changed, 166 insertions(+), 22 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index 58e3867246f..310b2a65099 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -3055,8 +3055,10 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Acquire AccessExclusiveLock on the table, its TOAST relation (if there
 	 * is one), all its indexes, so that we can swap the files.
+	 *
+	 * TODO The same for indexes and TOAST?
 	 */
-	LockRelationOid(old_table_oid, AccessExclusiveLock);
+	LockRelationOidUpgrade(old_table_oid, AccessExclusiveLock);
 
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
diff --git a/src/backend/storage/lmgr/lmgr.c b/src/backend/storage/lmgr/lmgr.c
index 2ccf7237fee..0b103485d54 100644
--- a/src/backend/storage/lmgr/lmgr.c
+++ b/src/backend/storage/lmgr/lmgr.c
@@ -58,6 +58,8 @@ typedef struct XactLockTableWaitInfo
 	const ItemPointerData *ctid;
 } XactLockTableWaitInfo;
 
+static void LockRelationOidCommon(Oid relid, LOCKMODE lockmode,
+								  bool isUpgrade);
 static void XactLockTableWaitErrorCb(void *arg);
 
 /*
@@ -105,6 +107,21 @@ SetLocktagRelationOid(LOCKTAG *tag, Oid relid)
  */
 void
 LockRelationOid(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, false);
+}
+
+/*
+ * Like above, but upgrade an existing lock.
+ */
+void
+LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, true);
+}
+
+static void
+LockRelationOidCommon(Oid relid, LOCKMODE lockmode, bool isUpgrade)
 {
 	LOCKTAG		tag;
 	LOCALLOCK  *locallock;
@@ -113,7 +130,7 @@ LockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, isUpgrade);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages, so that we
@@ -157,7 +174,7 @@ ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -191,7 +208,7 @@ LockRelationId(LockRelId *relid, LOCKMODE lockmode)
 	SET_LOCKTAG_RELATION(tag, relid->dbId, relid->relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -254,7 +271,7 @@ LockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -286,7 +303,7 @@ ConditionalLockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -591,7 +608,7 @@ ConditionalLockTuple(Relation relation, const ItemPointerData *tid, LOCKMODE loc
 					  ItemPointerGetOffsetNumber(tid));
 
 	return (LockAcquireExtended(&tag, lockmode, false, true, true, NULL,
-								logLockFailure) != LOCKACQUIRE_NOT_AVAIL);
+								logLockFailure, false) != LOCKACQUIRE_NOT_AVAIL);
 }
 
 /*
@@ -749,7 +766,7 @@ ConditionalXactLockTableWait(TransactionId xid, bool logLockFailure)
 		SET_LOCKTAG_TRANSACTION(tag, xid);
 
 		if (LockAcquireExtended(&tag, ShareLock, false, true, true, NULL,
-								logLockFailure)
+								logLockFailure, false)
 			== LOCKACQUIRE_NOT_AVAIL)
 			return false;
 
@@ -1043,7 +1060,7 @@ ConditionalLockDatabaseObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -1123,7 +1140,7 @@ ConditionalLockSharedObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
diff --git a/src/backend/storage/lmgr/lock.c b/src/backend/storage/lmgr/lock.c
index c221fe96889..bcbbf36c091 100644
--- a/src/backend/storage/lmgr/lock.c
+++ b/src/backend/storage/lmgr/lock.c
@@ -810,7 +810,7 @@ LockAcquire(const LOCKTAG *locktag,
 			bool dontWait)
 {
 	return LockAcquireExtended(locktag, lockmode, sessionLock, dontWait,
-							   true, NULL, false);
+							   true, NULL, false, false);
 }
 
 /*
@@ -829,6 +829,9 @@ LockAcquire(const LOCKTAG *locktag,
  *
  * logLockFailure indicates whether to log details when a lock acquisition
  * fails with dontWait = true.
+ *
+ * isUpgrade should be true if the backend already holds this lock in a lower
+ * mode.
  */
 LockAcquireResult
 LockAcquireExtended(const LOCKTAG *locktag,
@@ -837,7 +840,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 					bool dontWait,
 					bool reportMemoryError,
 					LOCALLOCK **locallockp,
-					bool logLockFailure)
+					bool logLockFailure,
+					bool isUpgrade)
 {
 	LOCKMETHODID lockmethodid = locktag->locktag_lockmethodid;
 	LockMethod	lockMethodTable;
@@ -1119,7 +1123,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		 * case, because JoinWaitQueue() may discover that we can acquire the
 		 * lock immediately after all.
 		 */
-		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait);
+		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait,
+								   isUpgrade);
 	}
 
 	if (waitResult == PROC_WAIT_STATUS_ERROR)
@@ -1225,8 +1230,85 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		Assert(!dontWait);
 		PROCLOCK_PRINT("LockAcquire: sleeping on lock", proclock);
 		LOCK_PRINT("LockAcquire: sleeping on lock", lock, lockmode);
+
+		/*
+		 * Lock upgrade can introduce deadlock, therefore enforce special
+		 * behavior of other processes that deal with this lock. In
+		 * particular, any waiter that sees upgradedBy set is expected to
+		 * perform deadlock check as soon as it's woken up. (Since we are
+		 * already in the queue, the deadlock detector has all the information
+		 * it needs.)
+		 */
+		if (isUpgrade)
+		{
+			dlist_iter	iter;
+
+			/*
+			 * There should not be multiple upgrades at the same time. XXX
+			 * ERROR ?
+			 */
+			Assert(lock->upgradedBy == NULL);
+
+			lock->upgradedBy = MyProc;
+
+			/*
+			 * Now, before I sleep myself, wake up all the existing waiters
+			 * (except for me), so they check for deadlock.
+			 */
+			dclist_foreach(iter, &lock->waitProcs)
+			{
+				PGPROC	   *proc = dlist_container(PGPROC, waitLink,
+												   iter.cur);
+
+				if (proc != MyProc)
+					SetLatch(&proc->procLatch);
+			}
+		}
 		LWLockRelease(partitionLock);
 
+		if (!isUpgrade)
+		{
+			HASH_SEQ_STATUS status;
+			LOCALLOCK	*llock;
+			bool	check_deadlock = false;
+
+			/*
+			 * Likewise, new waiters should check the isUpgrade flag and
+			 * engage deadlock detector if needed. The lock being acquired
+			 * should already be in the hash.
+			 *
+			 * Note that, to make deadlock detection happen as soon as
+			 * possible (i.e. regardless deadlock timeout), we check the other
+			 * locks too. XXX There might be ways to get into the deadlock
+			 * indirectly, but that needs more analysis. As long as the
+			 * upgrading process skips deadlock detection altogether, the
+			 * worst consequence of missing some lock wait cycle here is that
+			 * the other process will be kicked-off no sooner than after its
+			 * deadlock timeout has elapsed. (Which in turn means that the
+			 * lock upgrade will take longer than expected.)
+			 */
+			hash_seq_init(&status, LockMethodLocalHash);
+			while ((llock = (LOCALLOCK *) hash_seq_search(&status)) != NULL)
+			{
+				LOCK	*mylock = llock->lock;
+
+				if (mylock && mylock->upgradedBy)
+				{
+					check_deadlock = true;
+					break;
+				}
+			}
+			/* Check for deadlock if needed. */
+			if (check_deadlock)
+			{
+				DeadLockState	deadlock_state;
+
+				deadlock_state = CheckDeadLock();
+				if (deadlock_state == DS_HARD_DEADLOCK)
+					DeadLockReport();
+			}
+		}
+
 		waitResult = WaitOnLock(locallock, owner);
 
 		/*
@@ -1245,6 +1327,18 @@ LockAcquireExtended(const LOCKTAG *locktag,
 			DeadLockReport();
 			/* DeadLockReport() will not return */
 		}
+
+		/*
+		 * If finishing the lock upgrade, we should not get into a deadlock
+		 * anymore, so let others know that they do not have to care either.
+		 */
+		if (isUpgrade)
+		{
+			LWLockAcquire(partitionLock, LW_EXCLUSIVE);
+			Assert(lock->upgradedBy != NULL);
+			lock->upgradedBy = NULL;
+			LWLockRelease(partitionLock);
+		}
 	}
 	else
 		LWLockRelease(partitionLock);
@@ -1322,6 +1416,7 @@ SetupLockInTable(LockMethod lockMethodTable, PGPROC *proc,
 		lock->nGranted = 0;
 		MemSet(lock->requested, 0, sizeof(int) * MAX_LOCKMODES);
 		MemSet(lock->granted, 0, sizeof(int) * MAX_LOCKMODES);
+		lock->upgradedBy = NULL;
 		LOCK_PRINT("LockAcquire: new", lock, lockmode);
 	}
 	else
@@ -1768,6 +1863,12 @@ CleanUpLock(LOCK *lock, PROCLOCK *proclock,
 			elog(PANIC, "proclock table corrupted");
 	}
 
+	/*
+	 * Was this backend upgrading the lock?
+	 */
+	if (lock->upgradedBy == MyProc)
+		lock->upgradedBy = NULL;
+
 	if (lock->nRequested == 0)
 	{
 		/*
diff --git a/src/backend/storage/lmgr/proc.c b/src/backend/storage/lmgr/proc.c
index 1ac25068d62..7b31370db70 100644
--- a/src/backend/storage/lmgr/proc.c
+++ b/src/backend/storage/lmgr/proc.c
@@ -95,7 +95,6 @@ static volatile sig_atomic_t got_deadlock_timeout;
 static void RemoveProcFromArray(int code, Datum arg);
 static void ProcKill(int code, Datum arg);
 static void AuxiliaryProcKill(int code, Datum arg);
-static DeadLockState CheckDeadLock(void);
 
 
 /*
@@ -1143,7 +1142,8 @@ AuxiliaryPidGetProc(int pid)
  * NOTES: The process queue is now a priority queue for locking.
  */
 ProcWaitStatus
-JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
+JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait,
+			  bool isUpgrade)
 {
 	LOCKMODE	lockmode = locallock->tag.mode;
 	LOCK	   *lock = locallock->lock;
@@ -1230,8 +1230,12 @@ JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
 			/* Must he wait for me? */
 			if (lockMethodTable->conflictTab[proc->waitLockMode] & myHeldLocks)
 			{
-				/* Must I wait for him ? */
-				if (lockMethodTable->conflictTab[lockmode] & proc->heldLocks)
+				/*
+				 * Must I wait for him? I don't want a deadlock during lock
+				 * upgrade - other processes should fail on it.
+				 */
+				if ((lockMethodTable->conflictTab[lockmode] & proc->heldLocks) &&
+					!isUpgrade)
 				{
 					/*
 					 * Yes, so we have a deadlock.  Easiest way to clean up
@@ -1461,8 +1465,22 @@ ProcSleep(LOCALLOCK *locallock)
 			(void) WaitLatch(MyLatch, WL_LATCH_SET | WL_EXIT_ON_PM_DEATH, 0,
 							 PG_WAIT_LOCK | locallock->tag.lock.locktag_type);
 			ResetLatch(MyLatch);
-			/* check for deadlocks first, as that's probably log-worthy */
-			if (got_deadlock_timeout)
+			/*
+			 * Check for deadlocks first, as that's probably log-worthy.
+			 *
+			 * Do not wait for the timeout if the lock is being upgraded since
+			 * the risk of deadlock is higher now. However, do not check for
+			 * deadlock if the lock is being upgraded by this process - other
+			 * processes should take care.
+			 *
+			 * TODO Possible optimization: if this is the only lock of the
+			 * backend and if it did not have any weaker lock on the table so
+			 * far, it should be safe to skip the deadlock check. However, to
+			 * evaluate the situation, we need to take fast-path locks into
+			 * account.
+			 */
+			if ((got_deadlock_timeout || lock->upgradedBy) &&
+				lock->upgradedBy != MyProc)
 			{
 				deadlock_state = CheckDeadLock();
 				got_deadlock_timeout = false;
@@ -1819,7 +1837,7 @@ ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock)
  * not, just return.  If we have a real deadlock, remove ourselves from the
  * lock's wait queue.
  */
-static DeadLockState
+DeadLockState
 CheckDeadLock(void)
 {
 	int			i;
diff --git a/src/include/storage/lmgr.h b/src/include/storage/lmgr.h
index 2a985ce5e15..eddbe1e8622 100644
--- a/src/include/storage/lmgr.h
+++ b/src/include/storage/lmgr.h
@@ -38,6 +38,7 @@ extern void RelationInitLockInfo(Relation relation);
 
 /* Lock a relation */
 extern void LockRelationOid(Oid relid, LOCKMODE lockmode);
+extern void LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode);
 extern void LockRelationId(LockRelId *relid, LOCKMODE lockmode);
 extern bool ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode);
 extern void UnlockRelationId(LockRelId *relid, LOCKMODE lockmode);
diff --git a/src/include/storage/lock.h b/src/include/storage/lock.h
index ee3cb1dc203..080b69eddfd 100644
--- a/src/include/storage/lock.h
+++ b/src/include/storage/lock.h
@@ -131,6 +131,7 @@ typedef const LockMethodData *LockMethod;
  * nRequested -- total requested locks of all types.
  * granted -- count of each lock type currently granted on the lock.
  * nGranted -- total granted locks of all types.
+ * upgrader -- process that intends to upgrade the lock
  *
  * Note: these counts count 1 for each backend.  Internally to a backend,
  * there may be multiple grabs on a particular lock, but this is not reflected
@@ -150,6 +151,7 @@ typedef struct LOCK
 	int			nRequested;		/* total of requested[] array */
 	int			granted[MAX_LOCKMODES]; /* counts of granted locks */
 	int			nGranted;		/* total of granted[] array */
+	PGPROC		*upgradedBy;	/* is lock being upgraded by this process? */
 } LOCK;
 
 #define LOCK_LOCKMETHOD(lock) ((LOCKMETHODID) (lock).tag.locktag_lockmethodid)
@@ -390,7 +392,8 @@ extern LockAcquireResult LockAcquireExtended(const LOCKTAG *locktag,
 											 bool dontWait,
 											 bool reportMemoryError,
 											 LOCALLOCK **locallockp,
-											 bool logLockFailure);
+											 bool logLockFailure,
+											 bool isUpgrade);
 extern void AbortStrongLockAcquire(void);
 extern void MarkLockClear(LOCALLOCK *locallock);
 extern bool LockRelease(const LOCKTAG *locktag,
diff --git a/src/include/storage/proc.h b/src/include/storage/proc.h
index 3e1d1fad5f9..c4bc3a24044 100644
--- a/src/include/storage/proc.h
+++ b/src/include/storage/proc.h
@@ -563,10 +563,12 @@ extern bool HaveNFreeProcs(int n, int *nfree);
 extern void ProcReleaseLocks(bool isCommit);
 
 extern ProcWaitStatus JoinWaitQueue(LOCALLOCK *locallock,
-									LockMethod lockMethodTable, bool dontWait);
+									LockMethod lockMethodTable, bool dontWait,
+									bool isUpgrade);
 extern ProcWaitStatus ProcSleep(LOCALLOCK *locallock);
 extern void ProcWakeup(PGPROC *proc, ProcWaitStatus waitStatus);
 extern void ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock);
+extern DeadLockState CheckDeadLock(void);
 extern void CheckDeadLockAlert(void);
 extern void LockErrorCleanup(void);
 extern void GetLockHoldersAndWaiters(LOCALLOCK *locallock,
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 966+ messages in thread

* [PATCH] Lock upgrade without deadlocks.
@ 2026-04-17 13:13  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 966+ messages in thread

From: Antonin Houska @ 2026-04-17 13:13 UTC (permalink / raw)

For REPACK (CONCURRENTLY), it's essential to upgrade ShareUpdateExclusive lock
to AccessExclusiveLock for the final stage of table processing. However, we
cannot release the earlier before requesting the latter, because if another
process ran DDL on the table in between, REPACK would have to cancel its
transaction, and waste possibly a lot of work. Without releasing
ShareUpdateExclusive lock, the upgrade can cause deadlock, and it's not really
deterministic whether REPACK or the other process will get canceled.

This patch adds a new field to the LOCK structure which tells which process is
trying to upgrade the lock - this field is set right before the upgrade. All
processes waiting in the lock queue are then woken up to check the flag.

When another process tries to get the lock after that, it checks this field,
and if it's set, it checks for deadlocks even if deadlock timeout hasn't
expired yet. If the process is already in the lock's queue and sleeping, the
lock upgrading process wakes it up so it checks the flag immediately.

At the time the upgrading process starts to sleep, all the other process
should be aware that they need to check for deadlock during their lock
acquisitions, so the upgrading process does not have to check for deadlocks
when it gets woken up. Thus it should never receive deadlock error report.
---
 src/backend/commands/repack.c   |   4 +-
 src/backend/storage/lmgr/lmgr.c |  35 ++++++++---
 src/backend/storage/lmgr/lock.c | 107 +++++++++++++++++++++++++++++++-
 src/backend/storage/lmgr/proc.c |  32 +++++++---
 src/include/storage/lmgr.h      |   1 +
 src/include/storage/lock.h      |   5 +-
 src/include/storage/proc.h      |   4 +-
 7 files changed, 166 insertions(+), 22 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index 58e3867246f..310b2a65099 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -3055,8 +3055,10 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Acquire AccessExclusiveLock on the table, its TOAST relation (if there
 	 * is one), all its indexes, so that we can swap the files.
+	 *
+	 * TODO The same for indexes and TOAST?
 	 */
-	LockRelationOid(old_table_oid, AccessExclusiveLock);
+	LockRelationOidUpgrade(old_table_oid, AccessExclusiveLock);
 
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
diff --git a/src/backend/storage/lmgr/lmgr.c b/src/backend/storage/lmgr/lmgr.c
index 2ccf7237fee..0b103485d54 100644
--- a/src/backend/storage/lmgr/lmgr.c
+++ b/src/backend/storage/lmgr/lmgr.c
@@ -58,6 +58,8 @@ typedef struct XactLockTableWaitInfo
 	const ItemPointerData *ctid;
 } XactLockTableWaitInfo;
 
+static void LockRelationOidCommon(Oid relid, LOCKMODE lockmode,
+								  bool isUpgrade);
 static void XactLockTableWaitErrorCb(void *arg);
 
 /*
@@ -105,6 +107,21 @@ SetLocktagRelationOid(LOCKTAG *tag, Oid relid)
  */
 void
 LockRelationOid(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, false);
+}
+
+/*
+ * Like above, but upgrade an existing lock.
+ */
+void
+LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, true);
+}
+
+static void
+LockRelationOidCommon(Oid relid, LOCKMODE lockmode, bool isUpgrade)
 {
 	LOCKTAG		tag;
 	LOCALLOCK  *locallock;
@@ -113,7 +130,7 @@ LockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, isUpgrade);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages, so that we
@@ -157,7 +174,7 @@ ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -191,7 +208,7 @@ LockRelationId(LockRelId *relid, LOCKMODE lockmode)
 	SET_LOCKTAG_RELATION(tag, relid->dbId, relid->relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -254,7 +271,7 @@ LockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -286,7 +303,7 @@ ConditionalLockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -591,7 +608,7 @@ ConditionalLockTuple(Relation relation, const ItemPointerData *tid, LOCKMODE loc
 					  ItemPointerGetOffsetNumber(tid));
 
 	return (LockAcquireExtended(&tag, lockmode, false, true, true, NULL,
-								logLockFailure) != LOCKACQUIRE_NOT_AVAIL);
+								logLockFailure, false) != LOCKACQUIRE_NOT_AVAIL);
 }
 
 /*
@@ -749,7 +766,7 @@ ConditionalXactLockTableWait(TransactionId xid, bool logLockFailure)
 		SET_LOCKTAG_TRANSACTION(tag, xid);
 
 		if (LockAcquireExtended(&tag, ShareLock, false, true, true, NULL,
-								logLockFailure)
+								logLockFailure, false)
 			== LOCKACQUIRE_NOT_AVAIL)
 			return false;
 
@@ -1043,7 +1060,7 @@ ConditionalLockDatabaseObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -1123,7 +1140,7 @@ ConditionalLockSharedObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
diff --git a/src/backend/storage/lmgr/lock.c b/src/backend/storage/lmgr/lock.c
index c221fe96889..bcbbf36c091 100644
--- a/src/backend/storage/lmgr/lock.c
+++ b/src/backend/storage/lmgr/lock.c
@@ -810,7 +810,7 @@ LockAcquire(const LOCKTAG *locktag,
 			bool dontWait)
 {
 	return LockAcquireExtended(locktag, lockmode, sessionLock, dontWait,
-							   true, NULL, false);
+							   true, NULL, false, false);
 }
 
 /*
@@ -829,6 +829,9 @@ LockAcquire(const LOCKTAG *locktag,
  *
  * logLockFailure indicates whether to log details when a lock acquisition
  * fails with dontWait = true.
+ *
+ * isUpgrade should be true if the backend already holds this lock in a lower
+ * mode.
  */
 LockAcquireResult
 LockAcquireExtended(const LOCKTAG *locktag,
@@ -837,7 +840,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 					bool dontWait,
 					bool reportMemoryError,
 					LOCALLOCK **locallockp,
-					bool logLockFailure)
+					bool logLockFailure,
+					bool isUpgrade)
 {
 	LOCKMETHODID lockmethodid = locktag->locktag_lockmethodid;
 	LockMethod	lockMethodTable;
@@ -1119,7 +1123,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		 * case, because JoinWaitQueue() may discover that we can acquire the
 		 * lock immediately after all.
 		 */
-		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait);
+		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait,
+								   isUpgrade);
 	}
 
 	if (waitResult == PROC_WAIT_STATUS_ERROR)
@@ -1225,8 +1230,85 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		Assert(!dontWait);
 		PROCLOCK_PRINT("LockAcquire: sleeping on lock", proclock);
 		LOCK_PRINT("LockAcquire: sleeping on lock", lock, lockmode);
+
+		/*
+		 * Lock upgrade can introduce deadlock, therefore enforce special
+		 * behavior of other processes that deal with this lock. In
+		 * particular, any waiter that sees upgradedBy set is expected to
+		 * perform deadlock check as soon as it's woken up. (Since we are
+		 * already in the queue, the deadlock detector has all the information
+		 * it needs.)
+		 */
+		if (isUpgrade)
+		{
+			dlist_iter	iter;
+
+			/*
+			 * There should not be multiple upgrades at the same time. XXX
+			 * ERROR ?
+			 */
+			Assert(lock->upgradedBy == NULL);
+
+			lock->upgradedBy = MyProc;
+
+			/*
+			 * Now, before I sleep myself, wake up all the existing waiters
+			 * (except for me), so they check for deadlock.
+			 */
+			dclist_foreach(iter, &lock->waitProcs)
+			{
+				PGPROC	   *proc = dlist_container(PGPROC, waitLink,
+												   iter.cur);
+
+				if (proc != MyProc)
+					SetLatch(&proc->procLatch);
+			}
+		}
 		LWLockRelease(partitionLock);
 
+		if (!isUpgrade)
+		{
+			HASH_SEQ_STATUS status;
+			LOCALLOCK	*llock;
+			bool	check_deadlock = false;
+
+			/*
+			 * Likewise, new waiters should check the isUpgrade flag and
+			 * engage deadlock detector if needed. The lock being acquired
+			 * should already be in the hash.
+			 *
+			 * Note that, to make deadlock detection happen as soon as
+			 * possible (i.e. regardless deadlock timeout), we check the other
+			 * locks too. XXX There might be ways to get into the deadlock
+			 * indirectly, but that needs more analysis. As long as the
+			 * upgrading process skips deadlock detection altogether, the
+			 * worst consequence of missing some lock wait cycle here is that
+			 * the other process will be kicked-off no sooner than after its
+			 * deadlock timeout has elapsed. (Which in turn means that the
+			 * lock upgrade will take longer than expected.)
+			 */
+			hash_seq_init(&status, LockMethodLocalHash);
+			while ((llock = (LOCALLOCK *) hash_seq_search(&status)) != NULL)
+			{
+				LOCK	*mylock = llock->lock;
+
+				if (mylock && mylock->upgradedBy)
+				{
+					check_deadlock = true;
+					break;
+				}
+			}
+			/* Check for deadlock if needed. */
+			if (check_deadlock)
+			{
+				DeadLockState	deadlock_state;
+
+				deadlock_state = CheckDeadLock();
+				if (deadlock_state == DS_HARD_DEADLOCK)
+					DeadLockReport();
+			}
+		}
+
 		waitResult = WaitOnLock(locallock, owner);
 
 		/*
@@ -1245,6 +1327,18 @@ LockAcquireExtended(const LOCKTAG *locktag,
 			DeadLockReport();
 			/* DeadLockReport() will not return */
 		}
+
+		/*
+		 * If finishing the lock upgrade, we should not get into a deadlock
+		 * anymore, so let others know that they do not have to care either.
+		 */
+		if (isUpgrade)
+		{
+			LWLockAcquire(partitionLock, LW_EXCLUSIVE);
+			Assert(lock->upgradedBy != NULL);
+			lock->upgradedBy = NULL;
+			LWLockRelease(partitionLock);
+		}
 	}
 	else
 		LWLockRelease(partitionLock);
@@ -1322,6 +1416,7 @@ SetupLockInTable(LockMethod lockMethodTable, PGPROC *proc,
 		lock->nGranted = 0;
 		MemSet(lock->requested, 0, sizeof(int) * MAX_LOCKMODES);
 		MemSet(lock->granted, 0, sizeof(int) * MAX_LOCKMODES);
+		lock->upgradedBy = NULL;
 		LOCK_PRINT("LockAcquire: new", lock, lockmode);
 	}
 	else
@@ -1768,6 +1863,12 @@ CleanUpLock(LOCK *lock, PROCLOCK *proclock,
 			elog(PANIC, "proclock table corrupted");
 	}
 
+	/*
+	 * Was this backend upgrading the lock?
+	 */
+	if (lock->upgradedBy == MyProc)
+		lock->upgradedBy = NULL;
+
 	if (lock->nRequested == 0)
 	{
 		/*
diff --git a/src/backend/storage/lmgr/proc.c b/src/backend/storage/lmgr/proc.c
index 1ac25068d62..7b31370db70 100644
--- a/src/backend/storage/lmgr/proc.c
+++ b/src/backend/storage/lmgr/proc.c
@@ -95,7 +95,6 @@ static volatile sig_atomic_t got_deadlock_timeout;
 static void RemoveProcFromArray(int code, Datum arg);
 static void ProcKill(int code, Datum arg);
 static void AuxiliaryProcKill(int code, Datum arg);
-static DeadLockState CheckDeadLock(void);
 
 
 /*
@@ -1143,7 +1142,8 @@ AuxiliaryPidGetProc(int pid)
  * NOTES: The process queue is now a priority queue for locking.
  */
 ProcWaitStatus
-JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
+JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait,
+			  bool isUpgrade)
 {
 	LOCKMODE	lockmode = locallock->tag.mode;
 	LOCK	   *lock = locallock->lock;
@@ -1230,8 +1230,12 @@ JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
 			/* Must he wait for me? */
 			if (lockMethodTable->conflictTab[proc->waitLockMode] & myHeldLocks)
 			{
-				/* Must I wait for him ? */
-				if (lockMethodTable->conflictTab[lockmode] & proc->heldLocks)
+				/*
+				 * Must I wait for him? I don't want a deadlock during lock
+				 * upgrade - other processes should fail on it.
+				 */
+				if ((lockMethodTable->conflictTab[lockmode] & proc->heldLocks) &&
+					!isUpgrade)
 				{
 					/*
 					 * Yes, so we have a deadlock.  Easiest way to clean up
@@ -1461,8 +1465,22 @@ ProcSleep(LOCALLOCK *locallock)
 			(void) WaitLatch(MyLatch, WL_LATCH_SET | WL_EXIT_ON_PM_DEATH, 0,
 							 PG_WAIT_LOCK | locallock->tag.lock.locktag_type);
 			ResetLatch(MyLatch);
-			/* check for deadlocks first, as that's probably log-worthy */
-			if (got_deadlock_timeout)
+			/*
+			 * Check for deadlocks first, as that's probably log-worthy.
+			 *
+			 * Do not wait for the timeout if the lock is being upgraded since
+			 * the risk of deadlock is higher now. However, do not check for
+			 * deadlock if the lock is being upgraded by this process - other
+			 * processes should take care.
+			 *
+			 * TODO Possible optimization: if this is the only lock of the
+			 * backend and if it did not have any weaker lock on the table so
+			 * far, it should be safe to skip the deadlock check. However, to
+			 * evaluate the situation, we need to take fast-path locks into
+			 * account.
+			 */
+			if ((got_deadlock_timeout || lock->upgradedBy) &&
+				lock->upgradedBy != MyProc)
 			{
 				deadlock_state = CheckDeadLock();
 				got_deadlock_timeout = false;
@@ -1819,7 +1837,7 @@ ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock)
  * not, just return.  If we have a real deadlock, remove ourselves from the
  * lock's wait queue.
  */
-static DeadLockState
+DeadLockState
 CheckDeadLock(void)
 {
 	int			i;
diff --git a/src/include/storage/lmgr.h b/src/include/storage/lmgr.h
index 2a985ce5e15..eddbe1e8622 100644
--- a/src/include/storage/lmgr.h
+++ b/src/include/storage/lmgr.h
@@ -38,6 +38,7 @@ extern void RelationInitLockInfo(Relation relation);
 
 /* Lock a relation */
 extern void LockRelationOid(Oid relid, LOCKMODE lockmode);
+extern void LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode);
 extern void LockRelationId(LockRelId *relid, LOCKMODE lockmode);
 extern bool ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode);
 extern void UnlockRelationId(LockRelId *relid, LOCKMODE lockmode);
diff --git a/src/include/storage/lock.h b/src/include/storage/lock.h
index ee3cb1dc203..080b69eddfd 100644
--- a/src/include/storage/lock.h
+++ b/src/include/storage/lock.h
@@ -131,6 +131,7 @@ typedef const LockMethodData *LockMethod;
  * nRequested -- total requested locks of all types.
  * granted -- count of each lock type currently granted on the lock.
  * nGranted -- total granted locks of all types.
+ * upgrader -- process that intends to upgrade the lock
  *
  * Note: these counts count 1 for each backend.  Internally to a backend,
  * there may be multiple grabs on a particular lock, but this is not reflected
@@ -150,6 +151,7 @@ typedef struct LOCK
 	int			nRequested;		/* total of requested[] array */
 	int			granted[MAX_LOCKMODES]; /* counts of granted locks */
 	int			nGranted;		/* total of granted[] array */
+	PGPROC		*upgradedBy;	/* is lock being upgraded by this process? */
 } LOCK;
 
 #define LOCK_LOCKMETHOD(lock) ((LOCKMETHODID) (lock).tag.locktag_lockmethodid)
@@ -390,7 +392,8 @@ extern LockAcquireResult LockAcquireExtended(const LOCKTAG *locktag,
 											 bool dontWait,
 											 bool reportMemoryError,
 											 LOCALLOCK **locallockp,
-											 bool logLockFailure);
+											 bool logLockFailure,
+											 bool isUpgrade);
 extern void AbortStrongLockAcquire(void);
 extern void MarkLockClear(LOCALLOCK *locallock);
 extern bool LockRelease(const LOCKTAG *locktag,
diff --git a/src/include/storage/proc.h b/src/include/storage/proc.h
index 3e1d1fad5f9..c4bc3a24044 100644
--- a/src/include/storage/proc.h
+++ b/src/include/storage/proc.h
@@ -563,10 +563,12 @@ extern bool HaveNFreeProcs(int n, int *nfree);
 extern void ProcReleaseLocks(bool isCommit);
 
 extern ProcWaitStatus JoinWaitQueue(LOCALLOCK *locallock,
-									LockMethod lockMethodTable, bool dontWait);
+									LockMethod lockMethodTable, bool dontWait,
+									bool isUpgrade);
 extern ProcWaitStatus ProcSleep(LOCALLOCK *locallock);
 extern void ProcWakeup(PGPROC *proc, ProcWaitStatus waitStatus);
 extern void ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock);
+extern DeadLockState CheckDeadLock(void);
 extern void CheckDeadLockAlert(void);
 extern void LockErrorCleanup(void);
 extern void GetLockHoldersAndWaiters(LOCALLOCK *locallock,
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 966+ messages in thread

* [PATCH] Lock upgrade without deadlocks.
@ 2026-04-17 13:13  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 966+ messages in thread

From: Antonin Houska @ 2026-04-17 13:13 UTC (permalink / raw)

For REPACK (CONCURRENTLY), it's essential to upgrade ShareUpdateExclusive lock
to AccessExclusiveLock for the final stage of table processing. However, we
cannot release the earlier before requesting the latter, because if another
process ran DDL on the table in between, REPACK would have to cancel its
transaction, and waste possibly a lot of work. Without releasing
ShareUpdateExclusive lock, the upgrade can cause deadlock, and it's not really
deterministic whether REPACK or the other process will get canceled.

This patch adds a new field to the LOCK structure which tells which process is
trying to upgrade the lock - this field is set right before the upgrade. All
processes waiting in the lock queue are then woken up to check the flag.

When another process tries to get the lock after that, it checks this field,
and if it's set, it checks for deadlocks even if deadlock timeout hasn't
expired yet. If the process is already in the lock's queue and sleeping, the
lock upgrading process wakes it up so it checks the flag immediately.

At the time the upgrading process starts to sleep, all the other process
should be aware that they need to check for deadlock during their lock
acquisitions, so the upgrading process does not have to check for deadlocks
when it gets woken up. Thus it should never receive deadlock error report.
---
 src/backend/commands/repack.c   |   4 +-
 src/backend/storage/lmgr/lmgr.c |  35 ++++++++---
 src/backend/storage/lmgr/lock.c | 107 +++++++++++++++++++++++++++++++-
 src/backend/storage/lmgr/proc.c |  32 +++++++---
 src/include/storage/lmgr.h      |   1 +
 src/include/storage/lock.h      |   5 +-
 src/include/storage/proc.h      |   4 +-
 7 files changed, 166 insertions(+), 22 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index 58e3867246f..310b2a65099 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -3055,8 +3055,10 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Acquire AccessExclusiveLock on the table, its TOAST relation (if there
 	 * is one), all its indexes, so that we can swap the files.
+	 *
+	 * TODO The same for indexes and TOAST?
 	 */
-	LockRelationOid(old_table_oid, AccessExclusiveLock);
+	LockRelationOidUpgrade(old_table_oid, AccessExclusiveLock);
 
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
diff --git a/src/backend/storage/lmgr/lmgr.c b/src/backend/storage/lmgr/lmgr.c
index 2ccf7237fee..0b103485d54 100644
--- a/src/backend/storage/lmgr/lmgr.c
+++ b/src/backend/storage/lmgr/lmgr.c
@@ -58,6 +58,8 @@ typedef struct XactLockTableWaitInfo
 	const ItemPointerData *ctid;
 } XactLockTableWaitInfo;
 
+static void LockRelationOidCommon(Oid relid, LOCKMODE lockmode,
+								  bool isUpgrade);
 static void XactLockTableWaitErrorCb(void *arg);
 
 /*
@@ -105,6 +107,21 @@ SetLocktagRelationOid(LOCKTAG *tag, Oid relid)
  */
 void
 LockRelationOid(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, false);
+}
+
+/*
+ * Like above, but upgrade an existing lock.
+ */
+void
+LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, true);
+}
+
+static void
+LockRelationOidCommon(Oid relid, LOCKMODE lockmode, bool isUpgrade)
 {
 	LOCKTAG		tag;
 	LOCALLOCK  *locallock;
@@ -113,7 +130,7 @@ LockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, isUpgrade);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages, so that we
@@ -157,7 +174,7 @@ ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -191,7 +208,7 @@ LockRelationId(LockRelId *relid, LOCKMODE lockmode)
 	SET_LOCKTAG_RELATION(tag, relid->dbId, relid->relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -254,7 +271,7 @@ LockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -286,7 +303,7 @@ ConditionalLockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -591,7 +608,7 @@ ConditionalLockTuple(Relation relation, const ItemPointerData *tid, LOCKMODE loc
 					  ItemPointerGetOffsetNumber(tid));
 
 	return (LockAcquireExtended(&tag, lockmode, false, true, true, NULL,
-								logLockFailure) != LOCKACQUIRE_NOT_AVAIL);
+								logLockFailure, false) != LOCKACQUIRE_NOT_AVAIL);
 }
 
 /*
@@ -749,7 +766,7 @@ ConditionalXactLockTableWait(TransactionId xid, bool logLockFailure)
 		SET_LOCKTAG_TRANSACTION(tag, xid);
 
 		if (LockAcquireExtended(&tag, ShareLock, false, true, true, NULL,
-								logLockFailure)
+								logLockFailure, false)
 			== LOCKACQUIRE_NOT_AVAIL)
 			return false;
 
@@ -1043,7 +1060,7 @@ ConditionalLockDatabaseObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -1123,7 +1140,7 @@ ConditionalLockSharedObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
diff --git a/src/backend/storage/lmgr/lock.c b/src/backend/storage/lmgr/lock.c
index c221fe96889..bcbbf36c091 100644
--- a/src/backend/storage/lmgr/lock.c
+++ b/src/backend/storage/lmgr/lock.c
@@ -810,7 +810,7 @@ LockAcquire(const LOCKTAG *locktag,
 			bool dontWait)
 {
 	return LockAcquireExtended(locktag, lockmode, sessionLock, dontWait,
-							   true, NULL, false);
+							   true, NULL, false, false);
 }
 
 /*
@@ -829,6 +829,9 @@ LockAcquire(const LOCKTAG *locktag,
  *
  * logLockFailure indicates whether to log details when a lock acquisition
  * fails with dontWait = true.
+ *
+ * isUpgrade should be true if the backend already holds this lock in a lower
+ * mode.
  */
 LockAcquireResult
 LockAcquireExtended(const LOCKTAG *locktag,
@@ -837,7 +840,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 					bool dontWait,
 					bool reportMemoryError,
 					LOCALLOCK **locallockp,
-					bool logLockFailure)
+					bool logLockFailure,
+					bool isUpgrade)
 {
 	LOCKMETHODID lockmethodid = locktag->locktag_lockmethodid;
 	LockMethod	lockMethodTable;
@@ -1119,7 +1123,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		 * case, because JoinWaitQueue() may discover that we can acquire the
 		 * lock immediately after all.
 		 */
-		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait);
+		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait,
+								   isUpgrade);
 	}
 
 	if (waitResult == PROC_WAIT_STATUS_ERROR)
@@ -1225,8 +1230,85 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		Assert(!dontWait);
 		PROCLOCK_PRINT("LockAcquire: sleeping on lock", proclock);
 		LOCK_PRINT("LockAcquire: sleeping on lock", lock, lockmode);
+
+		/*
+		 * Lock upgrade can introduce deadlock, therefore enforce special
+		 * behavior of other processes that deal with this lock. In
+		 * particular, any waiter that sees upgradedBy set is expected to
+		 * perform deadlock check as soon as it's woken up. (Since we are
+		 * already in the queue, the deadlock detector has all the information
+		 * it needs.)
+		 */
+		if (isUpgrade)
+		{
+			dlist_iter	iter;
+
+			/*
+			 * There should not be multiple upgrades at the same time. XXX
+			 * ERROR ?
+			 */
+			Assert(lock->upgradedBy == NULL);
+
+			lock->upgradedBy = MyProc;
+
+			/*
+			 * Now, before I sleep myself, wake up all the existing waiters
+			 * (except for me), so they check for deadlock.
+			 */
+			dclist_foreach(iter, &lock->waitProcs)
+			{
+				PGPROC	   *proc = dlist_container(PGPROC, waitLink,
+												   iter.cur);
+
+				if (proc != MyProc)
+					SetLatch(&proc->procLatch);
+			}
+		}
 		LWLockRelease(partitionLock);
 
+		if (!isUpgrade)
+		{
+			HASH_SEQ_STATUS status;
+			LOCALLOCK	*llock;
+			bool	check_deadlock = false;
+
+			/*
+			 * Likewise, new waiters should check the isUpgrade flag and
+			 * engage deadlock detector if needed. The lock being acquired
+			 * should already be in the hash.
+			 *
+			 * Note that, to make deadlock detection happen as soon as
+			 * possible (i.e. regardless deadlock timeout), we check the other
+			 * locks too. XXX There might be ways to get into the deadlock
+			 * indirectly, but that needs more analysis. As long as the
+			 * upgrading process skips deadlock detection altogether, the
+			 * worst consequence of missing some lock wait cycle here is that
+			 * the other process will be kicked-off no sooner than after its
+			 * deadlock timeout has elapsed. (Which in turn means that the
+			 * lock upgrade will take longer than expected.)
+			 */
+			hash_seq_init(&status, LockMethodLocalHash);
+			while ((llock = (LOCALLOCK *) hash_seq_search(&status)) != NULL)
+			{
+				LOCK	*mylock = llock->lock;
+
+				if (mylock && mylock->upgradedBy)
+				{
+					check_deadlock = true;
+					break;
+				}
+			}
+			/* Check for deadlock if needed. */
+			if (check_deadlock)
+			{
+				DeadLockState	deadlock_state;
+
+				deadlock_state = CheckDeadLock();
+				if (deadlock_state == DS_HARD_DEADLOCK)
+					DeadLockReport();
+			}
+		}
+
 		waitResult = WaitOnLock(locallock, owner);
 
 		/*
@@ -1245,6 +1327,18 @@ LockAcquireExtended(const LOCKTAG *locktag,
 			DeadLockReport();
 			/* DeadLockReport() will not return */
 		}
+
+		/*
+		 * If finishing the lock upgrade, we should not get into a deadlock
+		 * anymore, so let others know that they do not have to care either.
+		 */
+		if (isUpgrade)
+		{
+			LWLockAcquire(partitionLock, LW_EXCLUSIVE);
+			Assert(lock->upgradedBy != NULL);
+			lock->upgradedBy = NULL;
+			LWLockRelease(partitionLock);
+		}
 	}
 	else
 		LWLockRelease(partitionLock);
@@ -1322,6 +1416,7 @@ SetupLockInTable(LockMethod lockMethodTable, PGPROC *proc,
 		lock->nGranted = 0;
 		MemSet(lock->requested, 0, sizeof(int) * MAX_LOCKMODES);
 		MemSet(lock->granted, 0, sizeof(int) * MAX_LOCKMODES);
+		lock->upgradedBy = NULL;
 		LOCK_PRINT("LockAcquire: new", lock, lockmode);
 	}
 	else
@@ -1768,6 +1863,12 @@ CleanUpLock(LOCK *lock, PROCLOCK *proclock,
 			elog(PANIC, "proclock table corrupted");
 	}
 
+	/*
+	 * Was this backend upgrading the lock?
+	 */
+	if (lock->upgradedBy == MyProc)
+		lock->upgradedBy = NULL;
+
 	if (lock->nRequested == 0)
 	{
 		/*
diff --git a/src/backend/storage/lmgr/proc.c b/src/backend/storage/lmgr/proc.c
index 1ac25068d62..7b31370db70 100644
--- a/src/backend/storage/lmgr/proc.c
+++ b/src/backend/storage/lmgr/proc.c
@@ -95,7 +95,6 @@ static volatile sig_atomic_t got_deadlock_timeout;
 static void RemoveProcFromArray(int code, Datum arg);
 static void ProcKill(int code, Datum arg);
 static void AuxiliaryProcKill(int code, Datum arg);
-static DeadLockState CheckDeadLock(void);
 
 
 /*
@@ -1143,7 +1142,8 @@ AuxiliaryPidGetProc(int pid)
  * NOTES: The process queue is now a priority queue for locking.
  */
 ProcWaitStatus
-JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
+JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait,
+			  bool isUpgrade)
 {
 	LOCKMODE	lockmode = locallock->tag.mode;
 	LOCK	   *lock = locallock->lock;
@@ -1230,8 +1230,12 @@ JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
 			/* Must he wait for me? */
 			if (lockMethodTable->conflictTab[proc->waitLockMode] & myHeldLocks)
 			{
-				/* Must I wait for him ? */
-				if (lockMethodTable->conflictTab[lockmode] & proc->heldLocks)
+				/*
+				 * Must I wait for him? I don't want a deadlock during lock
+				 * upgrade - other processes should fail on it.
+				 */
+				if ((lockMethodTable->conflictTab[lockmode] & proc->heldLocks) &&
+					!isUpgrade)
 				{
 					/*
 					 * Yes, so we have a deadlock.  Easiest way to clean up
@@ -1461,8 +1465,22 @@ ProcSleep(LOCALLOCK *locallock)
 			(void) WaitLatch(MyLatch, WL_LATCH_SET | WL_EXIT_ON_PM_DEATH, 0,
 							 PG_WAIT_LOCK | locallock->tag.lock.locktag_type);
 			ResetLatch(MyLatch);
-			/* check for deadlocks first, as that's probably log-worthy */
-			if (got_deadlock_timeout)
+			/*
+			 * Check for deadlocks first, as that's probably log-worthy.
+			 *
+			 * Do not wait for the timeout if the lock is being upgraded since
+			 * the risk of deadlock is higher now. However, do not check for
+			 * deadlock if the lock is being upgraded by this process - other
+			 * processes should take care.
+			 *
+			 * TODO Possible optimization: if this is the only lock of the
+			 * backend and if it did not have any weaker lock on the table so
+			 * far, it should be safe to skip the deadlock check. However, to
+			 * evaluate the situation, we need to take fast-path locks into
+			 * account.
+			 */
+			if ((got_deadlock_timeout || lock->upgradedBy) &&
+				lock->upgradedBy != MyProc)
 			{
 				deadlock_state = CheckDeadLock();
 				got_deadlock_timeout = false;
@@ -1819,7 +1837,7 @@ ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock)
  * not, just return.  If we have a real deadlock, remove ourselves from the
  * lock's wait queue.
  */
-static DeadLockState
+DeadLockState
 CheckDeadLock(void)
 {
 	int			i;
diff --git a/src/include/storage/lmgr.h b/src/include/storage/lmgr.h
index 2a985ce5e15..eddbe1e8622 100644
--- a/src/include/storage/lmgr.h
+++ b/src/include/storage/lmgr.h
@@ -38,6 +38,7 @@ extern void RelationInitLockInfo(Relation relation);
 
 /* Lock a relation */
 extern void LockRelationOid(Oid relid, LOCKMODE lockmode);
+extern void LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode);
 extern void LockRelationId(LockRelId *relid, LOCKMODE lockmode);
 extern bool ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode);
 extern void UnlockRelationId(LockRelId *relid, LOCKMODE lockmode);
diff --git a/src/include/storage/lock.h b/src/include/storage/lock.h
index ee3cb1dc203..080b69eddfd 100644
--- a/src/include/storage/lock.h
+++ b/src/include/storage/lock.h
@@ -131,6 +131,7 @@ typedef const LockMethodData *LockMethod;
  * nRequested -- total requested locks of all types.
  * granted -- count of each lock type currently granted on the lock.
  * nGranted -- total granted locks of all types.
+ * upgrader -- process that intends to upgrade the lock
  *
  * Note: these counts count 1 for each backend.  Internally to a backend,
  * there may be multiple grabs on a particular lock, but this is not reflected
@@ -150,6 +151,7 @@ typedef struct LOCK
 	int			nRequested;		/* total of requested[] array */
 	int			granted[MAX_LOCKMODES]; /* counts of granted locks */
 	int			nGranted;		/* total of granted[] array */
+	PGPROC		*upgradedBy;	/* is lock being upgraded by this process? */
 } LOCK;
 
 #define LOCK_LOCKMETHOD(lock) ((LOCKMETHODID) (lock).tag.locktag_lockmethodid)
@@ -390,7 +392,8 @@ extern LockAcquireResult LockAcquireExtended(const LOCKTAG *locktag,
 											 bool dontWait,
 											 bool reportMemoryError,
 											 LOCALLOCK **locallockp,
-											 bool logLockFailure);
+											 bool logLockFailure,
+											 bool isUpgrade);
 extern void AbortStrongLockAcquire(void);
 extern void MarkLockClear(LOCALLOCK *locallock);
 extern bool LockRelease(const LOCKTAG *locktag,
diff --git a/src/include/storage/proc.h b/src/include/storage/proc.h
index 3e1d1fad5f9..c4bc3a24044 100644
--- a/src/include/storage/proc.h
+++ b/src/include/storage/proc.h
@@ -563,10 +563,12 @@ extern bool HaveNFreeProcs(int n, int *nfree);
 extern void ProcReleaseLocks(bool isCommit);
 
 extern ProcWaitStatus JoinWaitQueue(LOCALLOCK *locallock,
-									LockMethod lockMethodTable, bool dontWait);
+									LockMethod lockMethodTable, bool dontWait,
+									bool isUpgrade);
 extern ProcWaitStatus ProcSleep(LOCALLOCK *locallock);
 extern void ProcWakeup(PGPROC *proc, ProcWaitStatus waitStatus);
 extern void ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock);
+extern DeadLockState CheckDeadLock(void);
 extern void CheckDeadLockAlert(void);
 extern void LockErrorCleanup(void);
 extern void GetLockHoldersAndWaiters(LOCALLOCK *locallock,
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 966+ messages in thread

* [PATCH] Lock upgrade without deadlocks.
@ 2026-04-17 13:13  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 966+ messages in thread

From: Antonin Houska @ 2026-04-17 13:13 UTC (permalink / raw)

For REPACK (CONCURRENTLY), it's essential to upgrade ShareUpdateExclusive lock
to AccessExclusiveLock for the final stage of table processing. However, we
cannot release the earlier before requesting the latter, because if another
process ran DDL on the table in between, REPACK would have to cancel its
transaction, and waste possibly a lot of work. Without releasing
ShareUpdateExclusive lock, the upgrade can cause deadlock, and it's not really
deterministic whether REPACK or the other process will get canceled.

This patch adds a new field to the LOCK structure which tells which process is
trying to upgrade the lock - this field is set right before the upgrade. All
processes waiting in the lock queue are then woken up to check the flag.

When another process tries to get the lock after that, it checks this field,
and if it's set, it checks for deadlocks even if deadlock timeout hasn't
expired yet. If the process is already in the lock's queue and sleeping, the
lock upgrading process wakes it up so it checks the flag immediately.

At the time the upgrading process starts to sleep, all the other process
should be aware that they need to check for deadlock during their lock
acquisitions, so the upgrading process does not have to check for deadlocks
when it gets woken up. Thus it should never receive deadlock error report.
---
 src/backend/commands/repack.c   |   4 +-
 src/backend/storage/lmgr/lmgr.c |  35 ++++++++---
 src/backend/storage/lmgr/lock.c | 107 +++++++++++++++++++++++++++++++-
 src/backend/storage/lmgr/proc.c |  32 +++++++---
 src/include/storage/lmgr.h      |   1 +
 src/include/storage/lock.h      |   5 +-
 src/include/storage/proc.h      |   4 +-
 7 files changed, 166 insertions(+), 22 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index 58e3867246f..310b2a65099 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -3055,8 +3055,10 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Acquire AccessExclusiveLock on the table, its TOAST relation (if there
 	 * is one), all its indexes, so that we can swap the files.
+	 *
+	 * TODO The same for indexes and TOAST?
 	 */
-	LockRelationOid(old_table_oid, AccessExclusiveLock);
+	LockRelationOidUpgrade(old_table_oid, AccessExclusiveLock);
 
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
diff --git a/src/backend/storage/lmgr/lmgr.c b/src/backend/storage/lmgr/lmgr.c
index 2ccf7237fee..0b103485d54 100644
--- a/src/backend/storage/lmgr/lmgr.c
+++ b/src/backend/storage/lmgr/lmgr.c
@@ -58,6 +58,8 @@ typedef struct XactLockTableWaitInfo
 	const ItemPointerData *ctid;
 } XactLockTableWaitInfo;
 
+static void LockRelationOidCommon(Oid relid, LOCKMODE lockmode,
+								  bool isUpgrade);
 static void XactLockTableWaitErrorCb(void *arg);
 
 /*
@@ -105,6 +107,21 @@ SetLocktagRelationOid(LOCKTAG *tag, Oid relid)
  */
 void
 LockRelationOid(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, false);
+}
+
+/*
+ * Like above, but upgrade an existing lock.
+ */
+void
+LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, true);
+}
+
+static void
+LockRelationOidCommon(Oid relid, LOCKMODE lockmode, bool isUpgrade)
 {
 	LOCKTAG		tag;
 	LOCALLOCK  *locallock;
@@ -113,7 +130,7 @@ LockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, isUpgrade);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages, so that we
@@ -157,7 +174,7 @@ ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -191,7 +208,7 @@ LockRelationId(LockRelId *relid, LOCKMODE lockmode)
 	SET_LOCKTAG_RELATION(tag, relid->dbId, relid->relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -254,7 +271,7 @@ LockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -286,7 +303,7 @@ ConditionalLockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -591,7 +608,7 @@ ConditionalLockTuple(Relation relation, const ItemPointerData *tid, LOCKMODE loc
 					  ItemPointerGetOffsetNumber(tid));
 
 	return (LockAcquireExtended(&tag, lockmode, false, true, true, NULL,
-								logLockFailure) != LOCKACQUIRE_NOT_AVAIL);
+								logLockFailure, false) != LOCKACQUIRE_NOT_AVAIL);
 }
 
 /*
@@ -749,7 +766,7 @@ ConditionalXactLockTableWait(TransactionId xid, bool logLockFailure)
 		SET_LOCKTAG_TRANSACTION(tag, xid);
 
 		if (LockAcquireExtended(&tag, ShareLock, false, true, true, NULL,
-								logLockFailure)
+								logLockFailure, false)
 			== LOCKACQUIRE_NOT_AVAIL)
 			return false;
 
@@ -1043,7 +1060,7 @@ ConditionalLockDatabaseObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -1123,7 +1140,7 @@ ConditionalLockSharedObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
diff --git a/src/backend/storage/lmgr/lock.c b/src/backend/storage/lmgr/lock.c
index c221fe96889..bcbbf36c091 100644
--- a/src/backend/storage/lmgr/lock.c
+++ b/src/backend/storage/lmgr/lock.c
@@ -810,7 +810,7 @@ LockAcquire(const LOCKTAG *locktag,
 			bool dontWait)
 {
 	return LockAcquireExtended(locktag, lockmode, sessionLock, dontWait,
-							   true, NULL, false);
+							   true, NULL, false, false);
 }
 
 /*
@@ -829,6 +829,9 @@ LockAcquire(const LOCKTAG *locktag,
  *
  * logLockFailure indicates whether to log details when a lock acquisition
  * fails with dontWait = true.
+ *
+ * isUpgrade should be true if the backend already holds this lock in a lower
+ * mode.
  */
 LockAcquireResult
 LockAcquireExtended(const LOCKTAG *locktag,
@@ -837,7 +840,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 					bool dontWait,
 					bool reportMemoryError,
 					LOCALLOCK **locallockp,
-					bool logLockFailure)
+					bool logLockFailure,
+					bool isUpgrade)
 {
 	LOCKMETHODID lockmethodid = locktag->locktag_lockmethodid;
 	LockMethod	lockMethodTable;
@@ -1119,7 +1123,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		 * case, because JoinWaitQueue() may discover that we can acquire the
 		 * lock immediately after all.
 		 */
-		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait);
+		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait,
+								   isUpgrade);
 	}
 
 	if (waitResult == PROC_WAIT_STATUS_ERROR)
@@ -1225,8 +1230,85 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		Assert(!dontWait);
 		PROCLOCK_PRINT("LockAcquire: sleeping on lock", proclock);
 		LOCK_PRINT("LockAcquire: sleeping on lock", lock, lockmode);
+
+		/*
+		 * Lock upgrade can introduce deadlock, therefore enforce special
+		 * behavior of other processes that deal with this lock. In
+		 * particular, any waiter that sees upgradedBy set is expected to
+		 * perform deadlock check as soon as it's woken up. (Since we are
+		 * already in the queue, the deadlock detector has all the information
+		 * it needs.)
+		 */
+		if (isUpgrade)
+		{
+			dlist_iter	iter;
+
+			/*
+			 * There should not be multiple upgrades at the same time. XXX
+			 * ERROR ?
+			 */
+			Assert(lock->upgradedBy == NULL);
+
+			lock->upgradedBy = MyProc;
+
+			/*
+			 * Now, before I sleep myself, wake up all the existing waiters
+			 * (except for me), so they check for deadlock.
+			 */
+			dclist_foreach(iter, &lock->waitProcs)
+			{
+				PGPROC	   *proc = dlist_container(PGPROC, waitLink,
+												   iter.cur);
+
+				if (proc != MyProc)
+					SetLatch(&proc->procLatch);
+			}
+		}
 		LWLockRelease(partitionLock);
 
+		if (!isUpgrade)
+		{
+			HASH_SEQ_STATUS status;
+			LOCALLOCK	*llock;
+			bool	check_deadlock = false;
+
+			/*
+			 * Likewise, new waiters should check the isUpgrade flag and
+			 * engage deadlock detector if needed. The lock being acquired
+			 * should already be in the hash.
+			 *
+			 * Note that, to make deadlock detection happen as soon as
+			 * possible (i.e. regardless deadlock timeout), we check the other
+			 * locks too. XXX There might be ways to get into the deadlock
+			 * indirectly, but that needs more analysis. As long as the
+			 * upgrading process skips deadlock detection altogether, the
+			 * worst consequence of missing some lock wait cycle here is that
+			 * the other process will be kicked-off no sooner than after its
+			 * deadlock timeout has elapsed. (Which in turn means that the
+			 * lock upgrade will take longer than expected.)
+			 */
+			hash_seq_init(&status, LockMethodLocalHash);
+			while ((llock = (LOCALLOCK *) hash_seq_search(&status)) != NULL)
+			{
+				LOCK	*mylock = llock->lock;
+
+				if (mylock && mylock->upgradedBy)
+				{
+					check_deadlock = true;
+					break;
+				}
+			}
+			/* Check for deadlock if needed. */
+			if (check_deadlock)
+			{
+				DeadLockState	deadlock_state;
+
+				deadlock_state = CheckDeadLock();
+				if (deadlock_state == DS_HARD_DEADLOCK)
+					DeadLockReport();
+			}
+		}
+
 		waitResult = WaitOnLock(locallock, owner);
 
 		/*
@@ -1245,6 +1327,18 @@ LockAcquireExtended(const LOCKTAG *locktag,
 			DeadLockReport();
 			/* DeadLockReport() will not return */
 		}
+
+		/*
+		 * If finishing the lock upgrade, we should not get into a deadlock
+		 * anymore, so let others know that they do not have to care either.
+		 */
+		if (isUpgrade)
+		{
+			LWLockAcquire(partitionLock, LW_EXCLUSIVE);
+			Assert(lock->upgradedBy != NULL);
+			lock->upgradedBy = NULL;
+			LWLockRelease(partitionLock);
+		}
 	}
 	else
 		LWLockRelease(partitionLock);
@@ -1322,6 +1416,7 @@ SetupLockInTable(LockMethod lockMethodTable, PGPROC *proc,
 		lock->nGranted = 0;
 		MemSet(lock->requested, 0, sizeof(int) * MAX_LOCKMODES);
 		MemSet(lock->granted, 0, sizeof(int) * MAX_LOCKMODES);
+		lock->upgradedBy = NULL;
 		LOCK_PRINT("LockAcquire: new", lock, lockmode);
 	}
 	else
@@ -1768,6 +1863,12 @@ CleanUpLock(LOCK *lock, PROCLOCK *proclock,
 			elog(PANIC, "proclock table corrupted");
 	}
 
+	/*
+	 * Was this backend upgrading the lock?
+	 */
+	if (lock->upgradedBy == MyProc)
+		lock->upgradedBy = NULL;
+
 	if (lock->nRequested == 0)
 	{
 		/*
diff --git a/src/backend/storage/lmgr/proc.c b/src/backend/storage/lmgr/proc.c
index 1ac25068d62..7b31370db70 100644
--- a/src/backend/storage/lmgr/proc.c
+++ b/src/backend/storage/lmgr/proc.c
@@ -95,7 +95,6 @@ static volatile sig_atomic_t got_deadlock_timeout;
 static void RemoveProcFromArray(int code, Datum arg);
 static void ProcKill(int code, Datum arg);
 static void AuxiliaryProcKill(int code, Datum arg);
-static DeadLockState CheckDeadLock(void);
 
 
 /*
@@ -1143,7 +1142,8 @@ AuxiliaryPidGetProc(int pid)
  * NOTES: The process queue is now a priority queue for locking.
  */
 ProcWaitStatus
-JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
+JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait,
+			  bool isUpgrade)
 {
 	LOCKMODE	lockmode = locallock->tag.mode;
 	LOCK	   *lock = locallock->lock;
@@ -1230,8 +1230,12 @@ JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
 			/* Must he wait for me? */
 			if (lockMethodTable->conflictTab[proc->waitLockMode] & myHeldLocks)
 			{
-				/* Must I wait for him ? */
-				if (lockMethodTable->conflictTab[lockmode] & proc->heldLocks)
+				/*
+				 * Must I wait for him? I don't want a deadlock during lock
+				 * upgrade - other processes should fail on it.
+				 */
+				if ((lockMethodTable->conflictTab[lockmode] & proc->heldLocks) &&
+					!isUpgrade)
 				{
 					/*
 					 * Yes, so we have a deadlock.  Easiest way to clean up
@@ -1461,8 +1465,22 @@ ProcSleep(LOCALLOCK *locallock)
 			(void) WaitLatch(MyLatch, WL_LATCH_SET | WL_EXIT_ON_PM_DEATH, 0,
 							 PG_WAIT_LOCK | locallock->tag.lock.locktag_type);
 			ResetLatch(MyLatch);
-			/* check for deadlocks first, as that's probably log-worthy */
-			if (got_deadlock_timeout)
+			/*
+			 * Check for deadlocks first, as that's probably log-worthy.
+			 *
+			 * Do not wait for the timeout if the lock is being upgraded since
+			 * the risk of deadlock is higher now. However, do not check for
+			 * deadlock if the lock is being upgraded by this process - other
+			 * processes should take care.
+			 *
+			 * TODO Possible optimization: if this is the only lock of the
+			 * backend and if it did not have any weaker lock on the table so
+			 * far, it should be safe to skip the deadlock check. However, to
+			 * evaluate the situation, we need to take fast-path locks into
+			 * account.
+			 */
+			if ((got_deadlock_timeout || lock->upgradedBy) &&
+				lock->upgradedBy != MyProc)
 			{
 				deadlock_state = CheckDeadLock();
 				got_deadlock_timeout = false;
@@ -1819,7 +1837,7 @@ ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock)
  * not, just return.  If we have a real deadlock, remove ourselves from the
  * lock's wait queue.
  */
-static DeadLockState
+DeadLockState
 CheckDeadLock(void)
 {
 	int			i;
diff --git a/src/include/storage/lmgr.h b/src/include/storage/lmgr.h
index 2a985ce5e15..eddbe1e8622 100644
--- a/src/include/storage/lmgr.h
+++ b/src/include/storage/lmgr.h
@@ -38,6 +38,7 @@ extern void RelationInitLockInfo(Relation relation);
 
 /* Lock a relation */
 extern void LockRelationOid(Oid relid, LOCKMODE lockmode);
+extern void LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode);
 extern void LockRelationId(LockRelId *relid, LOCKMODE lockmode);
 extern bool ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode);
 extern void UnlockRelationId(LockRelId *relid, LOCKMODE lockmode);
diff --git a/src/include/storage/lock.h b/src/include/storage/lock.h
index ee3cb1dc203..080b69eddfd 100644
--- a/src/include/storage/lock.h
+++ b/src/include/storage/lock.h
@@ -131,6 +131,7 @@ typedef const LockMethodData *LockMethod;
  * nRequested -- total requested locks of all types.
  * granted -- count of each lock type currently granted on the lock.
  * nGranted -- total granted locks of all types.
+ * upgrader -- process that intends to upgrade the lock
  *
  * Note: these counts count 1 for each backend.  Internally to a backend,
  * there may be multiple grabs on a particular lock, but this is not reflected
@@ -150,6 +151,7 @@ typedef struct LOCK
 	int			nRequested;		/* total of requested[] array */
 	int			granted[MAX_LOCKMODES]; /* counts of granted locks */
 	int			nGranted;		/* total of granted[] array */
+	PGPROC		*upgradedBy;	/* is lock being upgraded by this process? */
 } LOCK;
 
 #define LOCK_LOCKMETHOD(lock) ((LOCKMETHODID) (lock).tag.locktag_lockmethodid)
@@ -390,7 +392,8 @@ extern LockAcquireResult LockAcquireExtended(const LOCKTAG *locktag,
 											 bool dontWait,
 											 bool reportMemoryError,
 											 LOCALLOCK **locallockp,
-											 bool logLockFailure);
+											 bool logLockFailure,
+											 bool isUpgrade);
 extern void AbortStrongLockAcquire(void);
 extern void MarkLockClear(LOCALLOCK *locallock);
 extern bool LockRelease(const LOCKTAG *locktag,
diff --git a/src/include/storage/proc.h b/src/include/storage/proc.h
index 3e1d1fad5f9..c4bc3a24044 100644
--- a/src/include/storage/proc.h
+++ b/src/include/storage/proc.h
@@ -563,10 +563,12 @@ extern bool HaveNFreeProcs(int n, int *nfree);
 extern void ProcReleaseLocks(bool isCommit);
 
 extern ProcWaitStatus JoinWaitQueue(LOCALLOCK *locallock,
-									LockMethod lockMethodTable, bool dontWait);
+									LockMethod lockMethodTable, bool dontWait,
+									bool isUpgrade);
 extern ProcWaitStatus ProcSleep(LOCALLOCK *locallock);
 extern void ProcWakeup(PGPROC *proc, ProcWaitStatus waitStatus);
 extern void ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock);
+extern DeadLockState CheckDeadLock(void);
 extern void CheckDeadLockAlert(void);
 extern void LockErrorCleanup(void);
 extern void GetLockHoldersAndWaiters(LOCALLOCK *locallock,
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 966+ messages in thread

* [PATCH] Lock upgrade without deadlocks.
@ 2026-04-17 13:13  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 966+ messages in thread

From: Antonin Houska @ 2026-04-17 13:13 UTC (permalink / raw)

For REPACK (CONCURRENTLY), it's essential to upgrade ShareUpdateExclusive lock
to AccessExclusiveLock for the final stage of table processing. However, we
cannot release the earlier before requesting the latter, because if another
process ran DDL on the table in between, REPACK would have to cancel its
transaction, and waste possibly a lot of work. Without releasing
ShareUpdateExclusive lock, the upgrade can cause deadlock, and it's not really
deterministic whether REPACK or the other process will get canceled.

This patch adds a new field to the LOCK structure which tells which process is
trying to upgrade the lock - this field is set right before the upgrade. All
processes waiting in the lock queue are then woken up to check the flag.

When another process tries to get the lock after that, it checks this field,
and if it's set, it checks for deadlocks even if deadlock timeout hasn't
expired yet. If the process is already in the lock's queue and sleeping, the
lock upgrading process wakes it up so it checks the flag immediately.

At the time the upgrading process starts to sleep, all the other process
should be aware that they need to check for deadlock during their lock
acquisitions, so the upgrading process does not have to check for deadlocks
when it gets woken up. Thus it should never receive deadlock error report.
---
 src/backend/commands/repack.c   |   4 +-
 src/backend/storage/lmgr/lmgr.c |  35 ++++++++---
 src/backend/storage/lmgr/lock.c | 107 +++++++++++++++++++++++++++++++-
 src/backend/storage/lmgr/proc.c |  32 +++++++---
 src/include/storage/lmgr.h      |   1 +
 src/include/storage/lock.h      |   5 +-
 src/include/storage/proc.h      |   4 +-
 7 files changed, 166 insertions(+), 22 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index 58e3867246f..310b2a65099 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -3055,8 +3055,10 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Acquire AccessExclusiveLock on the table, its TOAST relation (if there
 	 * is one), all its indexes, so that we can swap the files.
+	 *
+	 * TODO The same for indexes and TOAST?
 	 */
-	LockRelationOid(old_table_oid, AccessExclusiveLock);
+	LockRelationOidUpgrade(old_table_oid, AccessExclusiveLock);
 
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
diff --git a/src/backend/storage/lmgr/lmgr.c b/src/backend/storage/lmgr/lmgr.c
index 2ccf7237fee..0b103485d54 100644
--- a/src/backend/storage/lmgr/lmgr.c
+++ b/src/backend/storage/lmgr/lmgr.c
@@ -58,6 +58,8 @@ typedef struct XactLockTableWaitInfo
 	const ItemPointerData *ctid;
 } XactLockTableWaitInfo;
 
+static void LockRelationOidCommon(Oid relid, LOCKMODE lockmode,
+								  bool isUpgrade);
 static void XactLockTableWaitErrorCb(void *arg);
 
 /*
@@ -105,6 +107,21 @@ SetLocktagRelationOid(LOCKTAG *tag, Oid relid)
  */
 void
 LockRelationOid(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, false);
+}
+
+/*
+ * Like above, but upgrade an existing lock.
+ */
+void
+LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, true);
+}
+
+static void
+LockRelationOidCommon(Oid relid, LOCKMODE lockmode, bool isUpgrade)
 {
 	LOCKTAG		tag;
 	LOCALLOCK  *locallock;
@@ -113,7 +130,7 @@ LockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, isUpgrade);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages, so that we
@@ -157,7 +174,7 @@ ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -191,7 +208,7 @@ LockRelationId(LockRelId *relid, LOCKMODE lockmode)
 	SET_LOCKTAG_RELATION(tag, relid->dbId, relid->relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -254,7 +271,7 @@ LockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -286,7 +303,7 @@ ConditionalLockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -591,7 +608,7 @@ ConditionalLockTuple(Relation relation, const ItemPointerData *tid, LOCKMODE loc
 					  ItemPointerGetOffsetNumber(tid));
 
 	return (LockAcquireExtended(&tag, lockmode, false, true, true, NULL,
-								logLockFailure) != LOCKACQUIRE_NOT_AVAIL);
+								logLockFailure, false) != LOCKACQUIRE_NOT_AVAIL);
 }
 
 /*
@@ -749,7 +766,7 @@ ConditionalXactLockTableWait(TransactionId xid, bool logLockFailure)
 		SET_LOCKTAG_TRANSACTION(tag, xid);
 
 		if (LockAcquireExtended(&tag, ShareLock, false, true, true, NULL,
-								logLockFailure)
+								logLockFailure, false)
 			== LOCKACQUIRE_NOT_AVAIL)
 			return false;
 
@@ -1043,7 +1060,7 @@ ConditionalLockDatabaseObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -1123,7 +1140,7 @@ ConditionalLockSharedObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
diff --git a/src/backend/storage/lmgr/lock.c b/src/backend/storage/lmgr/lock.c
index c221fe96889..bcbbf36c091 100644
--- a/src/backend/storage/lmgr/lock.c
+++ b/src/backend/storage/lmgr/lock.c
@@ -810,7 +810,7 @@ LockAcquire(const LOCKTAG *locktag,
 			bool dontWait)
 {
 	return LockAcquireExtended(locktag, lockmode, sessionLock, dontWait,
-							   true, NULL, false);
+							   true, NULL, false, false);
 }
 
 /*
@@ -829,6 +829,9 @@ LockAcquire(const LOCKTAG *locktag,
  *
  * logLockFailure indicates whether to log details when a lock acquisition
  * fails with dontWait = true.
+ *
+ * isUpgrade should be true if the backend already holds this lock in a lower
+ * mode.
  */
 LockAcquireResult
 LockAcquireExtended(const LOCKTAG *locktag,
@@ -837,7 +840,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 					bool dontWait,
 					bool reportMemoryError,
 					LOCALLOCK **locallockp,
-					bool logLockFailure)
+					bool logLockFailure,
+					bool isUpgrade)
 {
 	LOCKMETHODID lockmethodid = locktag->locktag_lockmethodid;
 	LockMethod	lockMethodTable;
@@ -1119,7 +1123,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		 * case, because JoinWaitQueue() may discover that we can acquire the
 		 * lock immediately after all.
 		 */
-		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait);
+		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait,
+								   isUpgrade);
 	}
 
 	if (waitResult == PROC_WAIT_STATUS_ERROR)
@@ -1225,8 +1230,85 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		Assert(!dontWait);
 		PROCLOCK_PRINT("LockAcquire: sleeping on lock", proclock);
 		LOCK_PRINT("LockAcquire: sleeping on lock", lock, lockmode);
+
+		/*
+		 * Lock upgrade can introduce deadlock, therefore enforce special
+		 * behavior of other processes that deal with this lock. In
+		 * particular, any waiter that sees upgradedBy set is expected to
+		 * perform deadlock check as soon as it's woken up. (Since we are
+		 * already in the queue, the deadlock detector has all the information
+		 * it needs.)
+		 */
+		if (isUpgrade)
+		{
+			dlist_iter	iter;
+
+			/*
+			 * There should not be multiple upgrades at the same time. XXX
+			 * ERROR ?
+			 */
+			Assert(lock->upgradedBy == NULL);
+
+			lock->upgradedBy = MyProc;
+
+			/*
+			 * Now, before I sleep myself, wake up all the existing waiters
+			 * (except for me), so they check for deadlock.
+			 */
+			dclist_foreach(iter, &lock->waitProcs)
+			{
+				PGPROC	   *proc = dlist_container(PGPROC, waitLink,
+												   iter.cur);
+
+				if (proc != MyProc)
+					SetLatch(&proc->procLatch);
+			}
+		}
 		LWLockRelease(partitionLock);
 
+		if (!isUpgrade)
+		{
+			HASH_SEQ_STATUS status;
+			LOCALLOCK	*llock;
+			bool	check_deadlock = false;
+
+			/*
+			 * Likewise, new waiters should check the isUpgrade flag and
+			 * engage deadlock detector if needed. The lock being acquired
+			 * should already be in the hash.
+			 *
+			 * Note that, to make deadlock detection happen as soon as
+			 * possible (i.e. regardless deadlock timeout), we check the other
+			 * locks too. XXX There might be ways to get into the deadlock
+			 * indirectly, but that needs more analysis. As long as the
+			 * upgrading process skips deadlock detection altogether, the
+			 * worst consequence of missing some lock wait cycle here is that
+			 * the other process will be kicked-off no sooner than after its
+			 * deadlock timeout has elapsed. (Which in turn means that the
+			 * lock upgrade will take longer than expected.)
+			 */
+			hash_seq_init(&status, LockMethodLocalHash);
+			while ((llock = (LOCALLOCK *) hash_seq_search(&status)) != NULL)
+			{
+				LOCK	*mylock = llock->lock;
+
+				if (mylock && mylock->upgradedBy)
+				{
+					check_deadlock = true;
+					break;
+				}
+			}
+			/* Check for deadlock if needed. */
+			if (check_deadlock)
+			{
+				DeadLockState	deadlock_state;
+
+				deadlock_state = CheckDeadLock();
+				if (deadlock_state == DS_HARD_DEADLOCK)
+					DeadLockReport();
+			}
+		}
+
 		waitResult = WaitOnLock(locallock, owner);
 
 		/*
@@ -1245,6 +1327,18 @@ LockAcquireExtended(const LOCKTAG *locktag,
 			DeadLockReport();
 			/* DeadLockReport() will not return */
 		}
+
+		/*
+		 * If finishing the lock upgrade, we should not get into a deadlock
+		 * anymore, so let others know that they do not have to care either.
+		 */
+		if (isUpgrade)
+		{
+			LWLockAcquire(partitionLock, LW_EXCLUSIVE);
+			Assert(lock->upgradedBy != NULL);
+			lock->upgradedBy = NULL;
+			LWLockRelease(partitionLock);
+		}
 	}
 	else
 		LWLockRelease(partitionLock);
@@ -1322,6 +1416,7 @@ SetupLockInTable(LockMethod lockMethodTable, PGPROC *proc,
 		lock->nGranted = 0;
 		MemSet(lock->requested, 0, sizeof(int) * MAX_LOCKMODES);
 		MemSet(lock->granted, 0, sizeof(int) * MAX_LOCKMODES);
+		lock->upgradedBy = NULL;
 		LOCK_PRINT("LockAcquire: new", lock, lockmode);
 	}
 	else
@@ -1768,6 +1863,12 @@ CleanUpLock(LOCK *lock, PROCLOCK *proclock,
 			elog(PANIC, "proclock table corrupted");
 	}
 
+	/*
+	 * Was this backend upgrading the lock?
+	 */
+	if (lock->upgradedBy == MyProc)
+		lock->upgradedBy = NULL;
+
 	if (lock->nRequested == 0)
 	{
 		/*
diff --git a/src/backend/storage/lmgr/proc.c b/src/backend/storage/lmgr/proc.c
index 1ac25068d62..7b31370db70 100644
--- a/src/backend/storage/lmgr/proc.c
+++ b/src/backend/storage/lmgr/proc.c
@@ -95,7 +95,6 @@ static volatile sig_atomic_t got_deadlock_timeout;
 static void RemoveProcFromArray(int code, Datum arg);
 static void ProcKill(int code, Datum arg);
 static void AuxiliaryProcKill(int code, Datum arg);
-static DeadLockState CheckDeadLock(void);
 
 
 /*
@@ -1143,7 +1142,8 @@ AuxiliaryPidGetProc(int pid)
  * NOTES: The process queue is now a priority queue for locking.
  */
 ProcWaitStatus
-JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
+JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait,
+			  bool isUpgrade)
 {
 	LOCKMODE	lockmode = locallock->tag.mode;
 	LOCK	   *lock = locallock->lock;
@@ -1230,8 +1230,12 @@ JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
 			/* Must he wait for me? */
 			if (lockMethodTable->conflictTab[proc->waitLockMode] & myHeldLocks)
 			{
-				/* Must I wait for him ? */
-				if (lockMethodTable->conflictTab[lockmode] & proc->heldLocks)
+				/*
+				 * Must I wait for him? I don't want a deadlock during lock
+				 * upgrade - other processes should fail on it.
+				 */
+				if ((lockMethodTable->conflictTab[lockmode] & proc->heldLocks) &&
+					!isUpgrade)
 				{
 					/*
 					 * Yes, so we have a deadlock.  Easiest way to clean up
@@ -1461,8 +1465,22 @@ ProcSleep(LOCALLOCK *locallock)
 			(void) WaitLatch(MyLatch, WL_LATCH_SET | WL_EXIT_ON_PM_DEATH, 0,
 							 PG_WAIT_LOCK | locallock->tag.lock.locktag_type);
 			ResetLatch(MyLatch);
-			/* check for deadlocks first, as that's probably log-worthy */
-			if (got_deadlock_timeout)
+			/*
+			 * Check for deadlocks first, as that's probably log-worthy.
+			 *
+			 * Do not wait for the timeout if the lock is being upgraded since
+			 * the risk of deadlock is higher now. However, do not check for
+			 * deadlock if the lock is being upgraded by this process - other
+			 * processes should take care.
+			 *
+			 * TODO Possible optimization: if this is the only lock of the
+			 * backend and if it did not have any weaker lock on the table so
+			 * far, it should be safe to skip the deadlock check. However, to
+			 * evaluate the situation, we need to take fast-path locks into
+			 * account.
+			 */
+			if ((got_deadlock_timeout || lock->upgradedBy) &&
+				lock->upgradedBy != MyProc)
 			{
 				deadlock_state = CheckDeadLock();
 				got_deadlock_timeout = false;
@@ -1819,7 +1837,7 @@ ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock)
  * not, just return.  If we have a real deadlock, remove ourselves from the
  * lock's wait queue.
  */
-static DeadLockState
+DeadLockState
 CheckDeadLock(void)
 {
 	int			i;
diff --git a/src/include/storage/lmgr.h b/src/include/storage/lmgr.h
index 2a985ce5e15..eddbe1e8622 100644
--- a/src/include/storage/lmgr.h
+++ b/src/include/storage/lmgr.h
@@ -38,6 +38,7 @@ extern void RelationInitLockInfo(Relation relation);
 
 /* Lock a relation */
 extern void LockRelationOid(Oid relid, LOCKMODE lockmode);
+extern void LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode);
 extern void LockRelationId(LockRelId *relid, LOCKMODE lockmode);
 extern bool ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode);
 extern void UnlockRelationId(LockRelId *relid, LOCKMODE lockmode);
diff --git a/src/include/storage/lock.h b/src/include/storage/lock.h
index ee3cb1dc203..080b69eddfd 100644
--- a/src/include/storage/lock.h
+++ b/src/include/storage/lock.h
@@ -131,6 +131,7 @@ typedef const LockMethodData *LockMethod;
  * nRequested -- total requested locks of all types.
  * granted -- count of each lock type currently granted on the lock.
  * nGranted -- total granted locks of all types.
+ * upgrader -- process that intends to upgrade the lock
  *
  * Note: these counts count 1 for each backend.  Internally to a backend,
  * there may be multiple grabs on a particular lock, but this is not reflected
@@ -150,6 +151,7 @@ typedef struct LOCK
 	int			nRequested;		/* total of requested[] array */
 	int			granted[MAX_LOCKMODES]; /* counts of granted locks */
 	int			nGranted;		/* total of granted[] array */
+	PGPROC		*upgradedBy;	/* is lock being upgraded by this process? */
 } LOCK;
 
 #define LOCK_LOCKMETHOD(lock) ((LOCKMETHODID) (lock).tag.locktag_lockmethodid)
@@ -390,7 +392,8 @@ extern LockAcquireResult LockAcquireExtended(const LOCKTAG *locktag,
 											 bool dontWait,
 											 bool reportMemoryError,
 											 LOCALLOCK **locallockp,
-											 bool logLockFailure);
+											 bool logLockFailure,
+											 bool isUpgrade);
 extern void AbortStrongLockAcquire(void);
 extern void MarkLockClear(LOCALLOCK *locallock);
 extern bool LockRelease(const LOCKTAG *locktag,
diff --git a/src/include/storage/proc.h b/src/include/storage/proc.h
index 3e1d1fad5f9..c4bc3a24044 100644
--- a/src/include/storage/proc.h
+++ b/src/include/storage/proc.h
@@ -563,10 +563,12 @@ extern bool HaveNFreeProcs(int n, int *nfree);
 extern void ProcReleaseLocks(bool isCommit);
 
 extern ProcWaitStatus JoinWaitQueue(LOCALLOCK *locallock,
-									LockMethod lockMethodTable, bool dontWait);
+									LockMethod lockMethodTable, bool dontWait,
+									bool isUpgrade);
 extern ProcWaitStatus ProcSleep(LOCALLOCK *locallock);
 extern void ProcWakeup(PGPROC *proc, ProcWaitStatus waitStatus);
 extern void ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock);
+extern DeadLockState CheckDeadLock(void);
 extern void CheckDeadLockAlert(void);
 extern void LockErrorCleanup(void);
 extern void GetLockHoldersAndWaiters(LOCALLOCK *locallock,
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 966+ messages in thread

* [PATCH] Lock upgrade without deadlocks.
@ 2026-04-17 13:13  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 966+ messages in thread

From: Antonin Houska @ 2026-04-17 13:13 UTC (permalink / raw)

For REPACK (CONCURRENTLY), it's essential to upgrade ShareUpdateExclusive lock
to AccessExclusiveLock for the final stage of table processing. However, we
cannot release the earlier before requesting the latter, because if another
process ran DDL on the table in between, REPACK would have to cancel its
transaction, and waste possibly a lot of work. Without releasing
ShareUpdateExclusive lock, the upgrade can cause deadlock, and it's not really
deterministic whether REPACK or the other process will get canceled.

This patch adds a new field to the LOCK structure which tells which process is
trying to upgrade the lock - this field is set right before the upgrade. All
processes waiting in the lock queue are then woken up to check the flag.

When another process tries to get the lock after that, it checks this field,
and if it's set, it checks for deadlocks even if deadlock timeout hasn't
expired yet. If the process is already in the lock's queue and sleeping, the
lock upgrading process wakes it up so it checks the flag immediately.

At the time the upgrading process starts to sleep, all the other process
should be aware that they need to check for deadlock during their lock
acquisitions, so the upgrading process does not have to check for deadlocks
when it gets woken up. Thus it should never receive deadlock error report.
---
 src/backend/commands/repack.c   |   4 +-
 src/backend/storage/lmgr/lmgr.c |  35 ++++++++---
 src/backend/storage/lmgr/lock.c | 107 +++++++++++++++++++++++++++++++-
 src/backend/storage/lmgr/proc.c |  32 +++++++---
 src/include/storage/lmgr.h      |   1 +
 src/include/storage/lock.h      |   5 +-
 src/include/storage/proc.h      |   4 +-
 7 files changed, 166 insertions(+), 22 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index 58e3867246f..310b2a65099 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -3055,8 +3055,10 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Acquire AccessExclusiveLock on the table, its TOAST relation (if there
 	 * is one), all its indexes, so that we can swap the files.
+	 *
+	 * TODO The same for indexes and TOAST?
 	 */
-	LockRelationOid(old_table_oid, AccessExclusiveLock);
+	LockRelationOidUpgrade(old_table_oid, AccessExclusiveLock);
 
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
diff --git a/src/backend/storage/lmgr/lmgr.c b/src/backend/storage/lmgr/lmgr.c
index 2ccf7237fee..0b103485d54 100644
--- a/src/backend/storage/lmgr/lmgr.c
+++ b/src/backend/storage/lmgr/lmgr.c
@@ -58,6 +58,8 @@ typedef struct XactLockTableWaitInfo
 	const ItemPointerData *ctid;
 } XactLockTableWaitInfo;
 
+static void LockRelationOidCommon(Oid relid, LOCKMODE lockmode,
+								  bool isUpgrade);
 static void XactLockTableWaitErrorCb(void *arg);
 
 /*
@@ -105,6 +107,21 @@ SetLocktagRelationOid(LOCKTAG *tag, Oid relid)
  */
 void
 LockRelationOid(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, false);
+}
+
+/*
+ * Like above, but upgrade an existing lock.
+ */
+void
+LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, true);
+}
+
+static void
+LockRelationOidCommon(Oid relid, LOCKMODE lockmode, bool isUpgrade)
 {
 	LOCKTAG		tag;
 	LOCALLOCK  *locallock;
@@ -113,7 +130,7 @@ LockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, isUpgrade);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages, so that we
@@ -157,7 +174,7 @@ ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -191,7 +208,7 @@ LockRelationId(LockRelId *relid, LOCKMODE lockmode)
 	SET_LOCKTAG_RELATION(tag, relid->dbId, relid->relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -254,7 +271,7 @@ LockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -286,7 +303,7 @@ ConditionalLockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -591,7 +608,7 @@ ConditionalLockTuple(Relation relation, const ItemPointerData *tid, LOCKMODE loc
 					  ItemPointerGetOffsetNumber(tid));
 
 	return (LockAcquireExtended(&tag, lockmode, false, true, true, NULL,
-								logLockFailure) != LOCKACQUIRE_NOT_AVAIL);
+								logLockFailure, false) != LOCKACQUIRE_NOT_AVAIL);
 }
 
 /*
@@ -749,7 +766,7 @@ ConditionalXactLockTableWait(TransactionId xid, bool logLockFailure)
 		SET_LOCKTAG_TRANSACTION(tag, xid);
 
 		if (LockAcquireExtended(&tag, ShareLock, false, true, true, NULL,
-								logLockFailure)
+								logLockFailure, false)
 			== LOCKACQUIRE_NOT_AVAIL)
 			return false;
 
@@ -1043,7 +1060,7 @@ ConditionalLockDatabaseObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -1123,7 +1140,7 @@ ConditionalLockSharedObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
diff --git a/src/backend/storage/lmgr/lock.c b/src/backend/storage/lmgr/lock.c
index c221fe96889..bcbbf36c091 100644
--- a/src/backend/storage/lmgr/lock.c
+++ b/src/backend/storage/lmgr/lock.c
@@ -810,7 +810,7 @@ LockAcquire(const LOCKTAG *locktag,
 			bool dontWait)
 {
 	return LockAcquireExtended(locktag, lockmode, sessionLock, dontWait,
-							   true, NULL, false);
+							   true, NULL, false, false);
 }
 
 /*
@@ -829,6 +829,9 @@ LockAcquire(const LOCKTAG *locktag,
  *
  * logLockFailure indicates whether to log details when a lock acquisition
  * fails with dontWait = true.
+ *
+ * isUpgrade should be true if the backend already holds this lock in a lower
+ * mode.
  */
 LockAcquireResult
 LockAcquireExtended(const LOCKTAG *locktag,
@@ -837,7 +840,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 					bool dontWait,
 					bool reportMemoryError,
 					LOCALLOCK **locallockp,
-					bool logLockFailure)
+					bool logLockFailure,
+					bool isUpgrade)
 {
 	LOCKMETHODID lockmethodid = locktag->locktag_lockmethodid;
 	LockMethod	lockMethodTable;
@@ -1119,7 +1123,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		 * case, because JoinWaitQueue() may discover that we can acquire the
 		 * lock immediately after all.
 		 */
-		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait);
+		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait,
+								   isUpgrade);
 	}
 
 	if (waitResult == PROC_WAIT_STATUS_ERROR)
@@ -1225,8 +1230,85 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		Assert(!dontWait);
 		PROCLOCK_PRINT("LockAcquire: sleeping on lock", proclock);
 		LOCK_PRINT("LockAcquire: sleeping on lock", lock, lockmode);
+
+		/*
+		 * Lock upgrade can introduce deadlock, therefore enforce special
+		 * behavior of other processes that deal with this lock. In
+		 * particular, any waiter that sees upgradedBy set is expected to
+		 * perform deadlock check as soon as it's woken up. (Since we are
+		 * already in the queue, the deadlock detector has all the information
+		 * it needs.)
+		 */
+		if (isUpgrade)
+		{
+			dlist_iter	iter;
+
+			/*
+			 * There should not be multiple upgrades at the same time. XXX
+			 * ERROR ?
+			 */
+			Assert(lock->upgradedBy == NULL);
+
+			lock->upgradedBy = MyProc;
+
+			/*
+			 * Now, before I sleep myself, wake up all the existing waiters
+			 * (except for me), so they check for deadlock.
+			 */
+			dclist_foreach(iter, &lock->waitProcs)
+			{
+				PGPROC	   *proc = dlist_container(PGPROC, waitLink,
+												   iter.cur);
+
+				if (proc != MyProc)
+					SetLatch(&proc->procLatch);
+			}
+		}
 		LWLockRelease(partitionLock);
 
+		if (!isUpgrade)
+		{
+			HASH_SEQ_STATUS status;
+			LOCALLOCK	*llock;
+			bool	check_deadlock = false;
+
+			/*
+			 * Likewise, new waiters should check the isUpgrade flag and
+			 * engage deadlock detector if needed. The lock being acquired
+			 * should already be in the hash.
+			 *
+			 * Note that, to make deadlock detection happen as soon as
+			 * possible (i.e. regardless deadlock timeout), we check the other
+			 * locks too. XXX There might be ways to get into the deadlock
+			 * indirectly, but that needs more analysis. As long as the
+			 * upgrading process skips deadlock detection altogether, the
+			 * worst consequence of missing some lock wait cycle here is that
+			 * the other process will be kicked-off no sooner than after its
+			 * deadlock timeout has elapsed. (Which in turn means that the
+			 * lock upgrade will take longer than expected.)
+			 */
+			hash_seq_init(&status, LockMethodLocalHash);
+			while ((llock = (LOCALLOCK *) hash_seq_search(&status)) != NULL)
+			{
+				LOCK	*mylock = llock->lock;
+
+				if (mylock && mylock->upgradedBy)
+				{
+					check_deadlock = true;
+					break;
+				}
+			}
+			/* Check for deadlock if needed. */
+			if (check_deadlock)
+			{
+				DeadLockState	deadlock_state;
+
+				deadlock_state = CheckDeadLock();
+				if (deadlock_state == DS_HARD_DEADLOCK)
+					DeadLockReport();
+			}
+		}
+
 		waitResult = WaitOnLock(locallock, owner);
 
 		/*
@@ -1245,6 +1327,18 @@ LockAcquireExtended(const LOCKTAG *locktag,
 			DeadLockReport();
 			/* DeadLockReport() will not return */
 		}
+
+		/*
+		 * If finishing the lock upgrade, we should not get into a deadlock
+		 * anymore, so let others know that they do not have to care either.
+		 */
+		if (isUpgrade)
+		{
+			LWLockAcquire(partitionLock, LW_EXCLUSIVE);
+			Assert(lock->upgradedBy != NULL);
+			lock->upgradedBy = NULL;
+			LWLockRelease(partitionLock);
+		}
 	}
 	else
 		LWLockRelease(partitionLock);
@@ -1322,6 +1416,7 @@ SetupLockInTable(LockMethod lockMethodTable, PGPROC *proc,
 		lock->nGranted = 0;
 		MemSet(lock->requested, 0, sizeof(int) * MAX_LOCKMODES);
 		MemSet(lock->granted, 0, sizeof(int) * MAX_LOCKMODES);
+		lock->upgradedBy = NULL;
 		LOCK_PRINT("LockAcquire: new", lock, lockmode);
 	}
 	else
@@ -1768,6 +1863,12 @@ CleanUpLock(LOCK *lock, PROCLOCK *proclock,
 			elog(PANIC, "proclock table corrupted");
 	}
 
+	/*
+	 * Was this backend upgrading the lock?
+	 */
+	if (lock->upgradedBy == MyProc)
+		lock->upgradedBy = NULL;
+
 	if (lock->nRequested == 0)
 	{
 		/*
diff --git a/src/backend/storage/lmgr/proc.c b/src/backend/storage/lmgr/proc.c
index 1ac25068d62..7b31370db70 100644
--- a/src/backend/storage/lmgr/proc.c
+++ b/src/backend/storage/lmgr/proc.c
@@ -95,7 +95,6 @@ static volatile sig_atomic_t got_deadlock_timeout;
 static void RemoveProcFromArray(int code, Datum arg);
 static void ProcKill(int code, Datum arg);
 static void AuxiliaryProcKill(int code, Datum arg);
-static DeadLockState CheckDeadLock(void);
 
 
 /*
@@ -1143,7 +1142,8 @@ AuxiliaryPidGetProc(int pid)
  * NOTES: The process queue is now a priority queue for locking.
  */
 ProcWaitStatus
-JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
+JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait,
+			  bool isUpgrade)
 {
 	LOCKMODE	lockmode = locallock->tag.mode;
 	LOCK	   *lock = locallock->lock;
@@ -1230,8 +1230,12 @@ JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
 			/* Must he wait for me? */
 			if (lockMethodTable->conflictTab[proc->waitLockMode] & myHeldLocks)
 			{
-				/* Must I wait for him ? */
-				if (lockMethodTable->conflictTab[lockmode] & proc->heldLocks)
+				/*
+				 * Must I wait for him? I don't want a deadlock during lock
+				 * upgrade - other processes should fail on it.
+				 */
+				if ((lockMethodTable->conflictTab[lockmode] & proc->heldLocks) &&
+					!isUpgrade)
 				{
 					/*
 					 * Yes, so we have a deadlock.  Easiest way to clean up
@@ -1461,8 +1465,22 @@ ProcSleep(LOCALLOCK *locallock)
 			(void) WaitLatch(MyLatch, WL_LATCH_SET | WL_EXIT_ON_PM_DEATH, 0,
 							 PG_WAIT_LOCK | locallock->tag.lock.locktag_type);
 			ResetLatch(MyLatch);
-			/* check for deadlocks first, as that's probably log-worthy */
-			if (got_deadlock_timeout)
+			/*
+			 * Check for deadlocks first, as that's probably log-worthy.
+			 *
+			 * Do not wait for the timeout if the lock is being upgraded since
+			 * the risk of deadlock is higher now. However, do not check for
+			 * deadlock if the lock is being upgraded by this process - other
+			 * processes should take care.
+			 *
+			 * TODO Possible optimization: if this is the only lock of the
+			 * backend and if it did not have any weaker lock on the table so
+			 * far, it should be safe to skip the deadlock check. However, to
+			 * evaluate the situation, we need to take fast-path locks into
+			 * account.
+			 */
+			if ((got_deadlock_timeout || lock->upgradedBy) &&
+				lock->upgradedBy != MyProc)
 			{
 				deadlock_state = CheckDeadLock();
 				got_deadlock_timeout = false;
@@ -1819,7 +1837,7 @@ ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock)
  * not, just return.  If we have a real deadlock, remove ourselves from the
  * lock's wait queue.
  */
-static DeadLockState
+DeadLockState
 CheckDeadLock(void)
 {
 	int			i;
diff --git a/src/include/storage/lmgr.h b/src/include/storage/lmgr.h
index 2a985ce5e15..eddbe1e8622 100644
--- a/src/include/storage/lmgr.h
+++ b/src/include/storage/lmgr.h
@@ -38,6 +38,7 @@ extern void RelationInitLockInfo(Relation relation);
 
 /* Lock a relation */
 extern void LockRelationOid(Oid relid, LOCKMODE lockmode);
+extern void LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode);
 extern void LockRelationId(LockRelId *relid, LOCKMODE lockmode);
 extern bool ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode);
 extern void UnlockRelationId(LockRelId *relid, LOCKMODE lockmode);
diff --git a/src/include/storage/lock.h b/src/include/storage/lock.h
index ee3cb1dc203..080b69eddfd 100644
--- a/src/include/storage/lock.h
+++ b/src/include/storage/lock.h
@@ -131,6 +131,7 @@ typedef const LockMethodData *LockMethod;
  * nRequested -- total requested locks of all types.
  * granted -- count of each lock type currently granted on the lock.
  * nGranted -- total granted locks of all types.
+ * upgrader -- process that intends to upgrade the lock
  *
  * Note: these counts count 1 for each backend.  Internally to a backend,
  * there may be multiple grabs on a particular lock, but this is not reflected
@@ -150,6 +151,7 @@ typedef struct LOCK
 	int			nRequested;		/* total of requested[] array */
 	int			granted[MAX_LOCKMODES]; /* counts of granted locks */
 	int			nGranted;		/* total of granted[] array */
+	PGPROC		*upgradedBy;	/* is lock being upgraded by this process? */
 } LOCK;
 
 #define LOCK_LOCKMETHOD(lock) ((LOCKMETHODID) (lock).tag.locktag_lockmethodid)
@@ -390,7 +392,8 @@ extern LockAcquireResult LockAcquireExtended(const LOCKTAG *locktag,
 											 bool dontWait,
 											 bool reportMemoryError,
 											 LOCALLOCK **locallockp,
-											 bool logLockFailure);
+											 bool logLockFailure,
+											 bool isUpgrade);
 extern void AbortStrongLockAcquire(void);
 extern void MarkLockClear(LOCALLOCK *locallock);
 extern bool LockRelease(const LOCKTAG *locktag,
diff --git a/src/include/storage/proc.h b/src/include/storage/proc.h
index 3e1d1fad5f9..c4bc3a24044 100644
--- a/src/include/storage/proc.h
+++ b/src/include/storage/proc.h
@@ -563,10 +563,12 @@ extern bool HaveNFreeProcs(int n, int *nfree);
 extern void ProcReleaseLocks(bool isCommit);
 
 extern ProcWaitStatus JoinWaitQueue(LOCALLOCK *locallock,
-									LockMethod lockMethodTable, bool dontWait);
+									LockMethod lockMethodTable, bool dontWait,
+									bool isUpgrade);
 extern ProcWaitStatus ProcSleep(LOCALLOCK *locallock);
 extern void ProcWakeup(PGPROC *proc, ProcWaitStatus waitStatus);
 extern void ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock);
+extern DeadLockState CheckDeadLock(void);
 extern void CheckDeadLockAlert(void);
 extern void LockErrorCleanup(void);
 extern void GetLockHoldersAndWaiters(LOCALLOCK *locallock,
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 966+ messages in thread

* [PATCH] Lock upgrade without deadlocks.
@ 2026-04-17 13:13  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 966+ messages in thread

From: Antonin Houska @ 2026-04-17 13:13 UTC (permalink / raw)

For REPACK (CONCURRENTLY), it's essential to upgrade ShareUpdateExclusive lock
to AccessExclusiveLock for the final stage of table processing. However, we
cannot release the earlier before requesting the latter, because if another
process ran DDL on the table in between, REPACK would have to cancel its
transaction, and waste possibly a lot of work. Without releasing
ShareUpdateExclusive lock, the upgrade can cause deadlock, and it's not really
deterministic whether REPACK or the other process will get canceled.

This patch adds a new field to the LOCK structure which tells which process is
trying to upgrade the lock - this field is set right before the upgrade. All
processes waiting in the lock queue are then woken up to check the flag.

When another process tries to get the lock after that, it checks this field,
and if it's set, it checks for deadlocks even if deadlock timeout hasn't
expired yet. If the process is already in the lock's queue and sleeping, the
lock upgrading process wakes it up so it checks the flag immediately.

At the time the upgrading process starts to sleep, all the other process
should be aware that they need to check for deadlock during their lock
acquisitions, so the upgrading process does not have to check for deadlocks
when it gets woken up. Thus it should never receive deadlock error report.
---
 src/backend/commands/repack.c   |   4 +-
 src/backend/storage/lmgr/lmgr.c |  35 ++++++++---
 src/backend/storage/lmgr/lock.c | 107 +++++++++++++++++++++++++++++++-
 src/backend/storage/lmgr/proc.c |  32 +++++++---
 src/include/storage/lmgr.h      |   1 +
 src/include/storage/lock.h      |   5 +-
 src/include/storage/proc.h      |   4 +-
 7 files changed, 166 insertions(+), 22 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index 58e3867246f..310b2a65099 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -3055,8 +3055,10 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Acquire AccessExclusiveLock on the table, its TOAST relation (if there
 	 * is one), all its indexes, so that we can swap the files.
+	 *
+	 * TODO The same for indexes and TOAST?
 	 */
-	LockRelationOid(old_table_oid, AccessExclusiveLock);
+	LockRelationOidUpgrade(old_table_oid, AccessExclusiveLock);
 
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
diff --git a/src/backend/storage/lmgr/lmgr.c b/src/backend/storage/lmgr/lmgr.c
index 2ccf7237fee..0b103485d54 100644
--- a/src/backend/storage/lmgr/lmgr.c
+++ b/src/backend/storage/lmgr/lmgr.c
@@ -58,6 +58,8 @@ typedef struct XactLockTableWaitInfo
 	const ItemPointerData *ctid;
 } XactLockTableWaitInfo;
 
+static void LockRelationOidCommon(Oid relid, LOCKMODE lockmode,
+								  bool isUpgrade);
 static void XactLockTableWaitErrorCb(void *arg);
 
 /*
@@ -105,6 +107,21 @@ SetLocktagRelationOid(LOCKTAG *tag, Oid relid)
  */
 void
 LockRelationOid(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, false);
+}
+
+/*
+ * Like above, but upgrade an existing lock.
+ */
+void
+LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, true);
+}
+
+static void
+LockRelationOidCommon(Oid relid, LOCKMODE lockmode, bool isUpgrade)
 {
 	LOCKTAG		tag;
 	LOCALLOCK  *locallock;
@@ -113,7 +130,7 @@ LockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, isUpgrade);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages, so that we
@@ -157,7 +174,7 @@ ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -191,7 +208,7 @@ LockRelationId(LockRelId *relid, LOCKMODE lockmode)
 	SET_LOCKTAG_RELATION(tag, relid->dbId, relid->relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -254,7 +271,7 @@ LockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -286,7 +303,7 @@ ConditionalLockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -591,7 +608,7 @@ ConditionalLockTuple(Relation relation, const ItemPointerData *tid, LOCKMODE loc
 					  ItemPointerGetOffsetNumber(tid));
 
 	return (LockAcquireExtended(&tag, lockmode, false, true, true, NULL,
-								logLockFailure) != LOCKACQUIRE_NOT_AVAIL);
+								logLockFailure, false) != LOCKACQUIRE_NOT_AVAIL);
 }
 
 /*
@@ -749,7 +766,7 @@ ConditionalXactLockTableWait(TransactionId xid, bool logLockFailure)
 		SET_LOCKTAG_TRANSACTION(tag, xid);
 
 		if (LockAcquireExtended(&tag, ShareLock, false, true, true, NULL,
-								logLockFailure)
+								logLockFailure, false)
 			== LOCKACQUIRE_NOT_AVAIL)
 			return false;
 
@@ -1043,7 +1060,7 @@ ConditionalLockDatabaseObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -1123,7 +1140,7 @@ ConditionalLockSharedObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
diff --git a/src/backend/storage/lmgr/lock.c b/src/backend/storage/lmgr/lock.c
index c221fe96889..bcbbf36c091 100644
--- a/src/backend/storage/lmgr/lock.c
+++ b/src/backend/storage/lmgr/lock.c
@@ -810,7 +810,7 @@ LockAcquire(const LOCKTAG *locktag,
 			bool dontWait)
 {
 	return LockAcquireExtended(locktag, lockmode, sessionLock, dontWait,
-							   true, NULL, false);
+							   true, NULL, false, false);
 }
 
 /*
@@ -829,6 +829,9 @@ LockAcquire(const LOCKTAG *locktag,
  *
  * logLockFailure indicates whether to log details when a lock acquisition
  * fails with dontWait = true.
+ *
+ * isUpgrade should be true if the backend already holds this lock in a lower
+ * mode.
  */
 LockAcquireResult
 LockAcquireExtended(const LOCKTAG *locktag,
@@ -837,7 +840,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 					bool dontWait,
 					bool reportMemoryError,
 					LOCALLOCK **locallockp,
-					bool logLockFailure)
+					bool logLockFailure,
+					bool isUpgrade)
 {
 	LOCKMETHODID lockmethodid = locktag->locktag_lockmethodid;
 	LockMethod	lockMethodTable;
@@ -1119,7 +1123,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		 * case, because JoinWaitQueue() may discover that we can acquire the
 		 * lock immediately after all.
 		 */
-		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait);
+		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait,
+								   isUpgrade);
 	}
 
 	if (waitResult == PROC_WAIT_STATUS_ERROR)
@@ -1225,8 +1230,85 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		Assert(!dontWait);
 		PROCLOCK_PRINT("LockAcquire: sleeping on lock", proclock);
 		LOCK_PRINT("LockAcquire: sleeping on lock", lock, lockmode);
+
+		/*
+		 * Lock upgrade can introduce deadlock, therefore enforce special
+		 * behavior of other processes that deal with this lock. In
+		 * particular, any waiter that sees upgradedBy set is expected to
+		 * perform deadlock check as soon as it's woken up. (Since we are
+		 * already in the queue, the deadlock detector has all the information
+		 * it needs.)
+		 */
+		if (isUpgrade)
+		{
+			dlist_iter	iter;
+
+			/*
+			 * There should not be multiple upgrades at the same time. XXX
+			 * ERROR ?
+			 */
+			Assert(lock->upgradedBy == NULL);
+
+			lock->upgradedBy = MyProc;
+
+			/*
+			 * Now, before I sleep myself, wake up all the existing waiters
+			 * (except for me), so they check for deadlock.
+			 */
+			dclist_foreach(iter, &lock->waitProcs)
+			{
+				PGPROC	   *proc = dlist_container(PGPROC, waitLink,
+												   iter.cur);
+
+				if (proc != MyProc)
+					SetLatch(&proc->procLatch);
+			}
+		}
 		LWLockRelease(partitionLock);
 
+		if (!isUpgrade)
+		{
+			HASH_SEQ_STATUS status;
+			LOCALLOCK	*llock;
+			bool	check_deadlock = false;
+
+			/*
+			 * Likewise, new waiters should check the isUpgrade flag and
+			 * engage deadlock detector if needed. The lock being acquired
+			 * should already be in the hash.
+			 *
+			 * Note that, to make deadlock detection happen as soon as
+			 * possible (i.e. regardless deadlock timeout), we check the other
+			 * locks too. XXX There might be ways to get into the deadlock
+			 * indirectly, but that needs more analysis. As long as the
+			 * upgrading process skips deadlock detection altogether, the
+			 * worst consequence of missing some lock wait cycle here is that
+			 * the other process will be kicked-off no sooner than after its
+			 * deadlock timeout has elapsed. (Which in turn means that the
+			 * lock upgrade will take longer than expected.)
+			 */
+			hash_seq_init(&status, LockMethodLocalHash);
+			while ((llock = (LOCALLOCK *) hash_seq_search(&status)) != NULL)
+			{
+				LOCK	*mylock = llock->lock;
+
+				if (mylock && mylock->upgradedBy)
+				{
+					check_deadlock = true;
+					break;
+				}
+			}
+			/* Check for deadlock if needed. */
+			if (check_deadlock)
+			{
+				DeadLockState	deadlock_state;
+
+				deadlock_state = CheckDeadLock();
+				if (deadlock_state == DS_HARD_DEADLOCK)
+					DeadLockReport();
+			}
+		}
+
 		waitResult = WaitOnLock(locallock, owner);
 
 		/*
@@ -1245,6 +1327,18 @@ LockAcquireExtended(const LOCKTAG *locktag,
 			DeadLockReport();
 			/* DeadLockReport() will not return */
 		}
+
+		/*
+		 * If finishing the lock upgrade, we should not get into a deadlock
+		 * anymore, so let others know that they do not have to care either.
+		 */
+		if (isUpgrade)
+		{
+			LWLockAcquire(partitionLock, LW_EXCLUSIVE);
+			Assert(lock->upgradedBy != NULL);
+			lock->upgradedBy = NULL;
+			LWLockRelease(partitionLock);
+		}
 	}
 	else
 		LWLockRelease(partitionLock);
@@ -1322,6 +1416,7 @@ SetupLockInTable(LockMethod lockMethodTable, PGPROC *proc,
 		lock->nGranted = 0;
 		MemSet(lock->requested, 0, sizeof(int) * MAX_LOCKMODES);
 		MemSet(lock->granted, 0, sizeof(int) * MAX_LOCKMODES);
+		lock->upgradedBy = NULL;
 		LOCK_PRINT("LockAcquire: new", lock, lockmode);
 	}
 	else
@@ -1768,6 +1863,12 @@ CleanUpLock(LOCK *lock, PROCLOCK *proclock,
 			elog(PANIC, "proclock table corrupted");
 	}
 
+	/*
+	 * Was this backend upgrading the lock?
+	 */
+	if (lock->upgradedBy == MyProc)
+		lock->upgradedBy = NULL;
+
 	if (lock->nRequested == 0)
 	{
 		/*
diff --git a/src/backend/storage/lmgr/proc.c b/src/backend/storage/lmgr/proc.c
index 1ac25068d62..7b31370db70 100644
--- a/src/backend/storage/lmgr/proc.c
+++ b/src/backend/storage/lmgr/proc.c
@@ -95,7 +95,6 @@ static volatile sig_atomic_t got_deadlock_timeout;
 static void RemoveProcFromArray(int code, Datum arg);
 static void ProcKill(int code, Datum arg);
 static void AuxiliaryProcKill(int code, Datum arg);
-static DeadLockState CheckDeadLock(void);
 
 
 /*
@@ -1143,7 +1142,8 @@ AuxiliaryPidGetProc(int pid)
  * NOTES: The process queue is now a priority queue for locking.
  */
 ProcWaitStatus
-JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
+JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait,
+			  bool isUpgrade)
 {
 	LOCKMODE	lockmode = locallock->tag.mode;
 	LOCK	   *lock = locallock->lock;
@@ -1230,8 +1230,12 @@ JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
 			/* Must he wait for me? */
 			if (lockMethodTable->conflictTab[proc->waitLockMode] & myHeldLocks)
 			{
-				/* Must I wait for him ? */
-				if (lockMethodTable->conflictTab[lockmode] & proc->heldLocks)
+				/*
+				 * Must I wait for him? I don't want a deadlock during lock
+				 * upgrade - other processes should fail on it.
+				 */
+				if ((lockMethodTable->conflictTab[lockmode] & proc->heldLocks) &&
+					!isUpgrade)
 				{
 					/*
 					 * Yes, so we have a deadlock.  Easiest way to clean up
@@ -1461,8 +1465,22 @@ ProcSleep(LOCALLOCK *locallock)
 			(void) WaitLatch(MyLatch, WL_LATCH_SET | WL_EXIT_ON_PM_DEATH, 0,
 							 PG_WAIT_LOCK | locallock->tag.lock.locktag_type);
 			ResetLatch(MyLatch);
-			/* check for deadlocks first, as that's probably log-worthy */
-			if (got_deadlock_timeout)
+			/*
+			 * Check for deadlocks first, as that's probably log-worthy.
+			 *
+			 * Do not wait for the timeout if the lock is being upgraded since
+			 * the risk of deadlock is higher now. However, do not check for
+			 * deadlock if the lock is being upgraded by this process - other
+			 * processes should take care.
+			 *
+			 * TODO Possible optimization: if this is the only lock of the
+			 * backend and if it did not have any weaker lock on the table so
+			 * far, it should be safe to skip the deadlock check. However, to
+			 * evaluate the situation, we need to take fast-path locks into
+			 * account.
+			 */
+			if ((got_deadlock_timeout || lock->upgradedBy) &&
+				lock->upgradedBy != MyProc)
 			{
 				deadlock_state = CheckDeadLock();
 				got_deadlock_timeout = false;
@@ -1819,7 +1837,7 @@ ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock)
  * not, just return.  If we have a real deadlock, remove ourselves from the
  * lock's wait queue.
  */
-static DeadLockState
+DeadLockState
 CheckDeadLock(void)
 {
 	int			i;
diff --git a/src/include/storage/lmgr.h b/src/include/storage/lmgr.h
index 2a985ce5e15..eddbe1e8622 100644
--- a/src/include/storage/lmgr.h
+++ b/src/include/storage/lmgr.h
@@ -38,6 +38,7 @@ extern void RelationInitLockInfo(Relation relation);
 
 /* Lock a relation */
 extern void LockRelationOid(Oid relid, LOCKMODE lockmode);
+extern void LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode);
 extern void LockRelationId(LockRelId *relid, LOCKMODE lockmode);
 extern bool ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode);
 extern void UnlockRelationId(LockRelId *relid, LOCKMODE lockmode);
diff --git a/src/include/storage/lock.h b/src/include/storage/lock.h
index ee3cb1dc203..080b69eddfd 100644
--- a/src/include/storage/lock.h
+++ b/src/include/storage/lock.h
@@ -131,6 +131,7 @@ typedef const LockMethodData *LockMethod;
  * nRequested -- total requested locks of all types.
  * granted -- count of each lock type currently granted on the lock.
  * nGranted -- total granted locks of all types.
+ * upgrader -- process that intends to upgrade the lock
  *
  * Note: these counts count 1 for each backend.  Internally to a backend,
  * there may be multiple grabs on a particular lock, but this is not reflected
@@ -150,6 +151,7 @@ typedef struct LOCK
 	int			nRequested;		/* total of requested[] array */
 	int			granted[MAX_LOCKMODES]; /* counts of granted locks */
 	int			nGranted;		/* total of granted[] array */
+	PGPROC		*upgradedBy;	/* is lock being upgraded by this process? */
 } LOCK;
 
 #define LOCK_LOCKMETHOD(lock) ((LOCKMETHODID) (lock).tag.locktag_lockmethodid)
@@ -390,7 +392,8 @@ extern LockAcquireResult LockAcquireExtended(const LOCKTAG *locktag,
 											 bool dontWait,
 											 bool reportMemoryError,
 											 LOCALLOCK **locallockp,
-											 bool logLockFailure);
+											 bool logLockFailure,
+											 bool isUpgrade);
 extern void AbortStrongLockAcquire(void);
 extern void MarkLockClear(LOCALLOCK *locallock);
 extern bool LockRelease(const LOCKTAG *locktag,
diff --git a/src/include/storage/proc.h b/src/include/storage/proc.h
index 3e1d1fad5f9..c4bc3a24044 100644
--- a/src/include/storage/proc.h
+++ b/src/include/storage/proc.h
@@ -563,10 +563,12 @@ extern bool HaveNFreeProcs(int n, int *nfree);
 extern void ProcReleaseLocks(bool isCommit);
 
 extern ProcWaitStatus JoinWaitQueue(LOCALLOCK *locallock,
-									LockMethod lockMethodTable, bool dontWait);
+									LockMethod lockMethodTable, bool dontWait,
+									bool isUpgrade);
 extern ProcWaitStatus ProcSleep(LOCALLOCK *locallock);
 extern void ProcWakeup(PGPROC *proc, ProcWaitStatus waitStatus);
 extern void ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock);
+extern DeadLockState CheckDeadLock(void);
 extern void CheckDeadLockAlert(void);
 extern void LockErrorCleanup(void);
 extern void GetLockHoldersAndWaiters(LOCALLOCK *locallock,
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 966+ messages in thread

* [PATCH] Lock upgrade without deadlocks.
@ 2026-04-17 13:13  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 966+ messages in thread

From: Antonin Houska @ 2026-04-17 13:13 UTC (permalink / raw)

For REPACK (CONCURRENTLY), it's essential to upgrade ShareUpdateExclusive lock
to AccessExclusiveLock for the final stage of table processing. However, we
cannot release the earlier before requesting the latter, because if another
process ran DDL on the table in between, REPACK would have to cancel its
transaction, and waste possibly a lot of work. Without releasing
ShareUpdateExclusive lock, the upgrade can cause deadlock, and it's not really
deterministic whether REPACK or the other process will get canceled.

This patch adds a new field to the LOCK structure which tells which process is
trying to upgrade the lock - this field is set right before the upgrade. All
processes waiting in the lock queue are then woken up to check the flag.

When another process tries to get the lock after that, it checks this field,
and if it's set, it checks for deadlocks even if deadlock timeout hasn't
expired yet. If the process is already in the lock's queue and sleeping, the
lock upgrading process wakes it up so it checks the flag immediately.

At the time the upgrading process starts to sleep, all the other process
should be aware that they need to check for deadlock during their lock
acquisitions, so the upgrading process does not have to check for deadlocks
when it gets woken up. Thus it should never receive deadlock error report.
---
 src/backend/commands/repack.c   |   4 +-
 src/backend/storage/lmgr/lmgr.c |  35 ++++++++---
 src/backend/storage/lmgr/lock.c | 107 +++++++++++++++++++++++++++++++-
 src/backend/storage/lmgr/proc.c |  32 +++++++---
 src/include/storage/lmgr.h      |   1 +
 src/include/storage/lock.h      |   5 +-
 src/include/storage/proc.h      |   4 +-
 7 files changed, 166 insertions(+), 22 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index 58e3867246f..310b2a65099 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -3055,8 +3055,10 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Acquire AccessExclusiveLock on the table, its TOAST relation (if there
 	 * is one), all its indexes, so that we can swap the files.
+	 *
+	 * TODO The same for indexes and TOAST?
 	 */
-	LockRelationOid(old_table_oid, AccessExclusiveLock);
+	LockRelationOidUpgrade(old_table_oid, AccessExclusiveLock);
 
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
diff --git a/src/backend/storage/lmgr/lmgr.c b/src/backend/storage/lmgr/lmgr.c
index 2ccf7237fee..0b103485d54 100644
--- a/src/backend/storage/lmgr/lmgr.c
+++ b/src/backend/storage/lmgr/lmgr.c
@@ -58,6 +58,8 @@ typedef struct XactLockTableWaitInfo
 	const ItemPointerData *ctid;
 } XactLockTableWaitInfo;
 
+static void LockRelationOidCommon(Oid relid, LOCKMODE lockmode,
+								  bool isUpgrade);
 static void XactLockTableWaitErrorCb(void *arg);
 
 /*
@@ -105,6 +107,21 @@ SetLocktagRelationOid(LOCKTAG *tag, Oid relid)
  */
 void
 LockRelationOid(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, false);
+}
+
+/*
+ * Like above, but upgrade an existing lock.
+ */
+void
+LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, true);
+}
+
+static void
+LockRelationOidCommon(Oid relid, LOCKMODE lockmode, bool isUpgrade)
 {
 	LOCKTAG		tag;
 	LOCALLOCK  *locallock;
@@ -113,7 +130,7 @@ LockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, isUpgrade);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages, so that we
@@ -157,7 +174,7 @@ ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -191,7 +208,7 @@ LockRelationId(LockRelId *relid, LOCKMODE lockmode)
 	SET_LOCKTAG_RELATION(tag, relid->dbId, relid->relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -254,7 +271,7 @@ LockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -286,7 +303,7 @@ ConditionalLockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -591,7 +608,7 @@ ConditionalLockTuple(Relation relation, const ItemPointerData *tid, LOCKMODE loc
 					  ItemPointerGetOffsetNumber(tid));
 
 	return (LockAcquireExtended(&tag, lockmode, false, true, true, NULL,
-								logLockFailure) != LOCKACQUIRE_NOT_AVAIL);
+								logLockFailure, false) != LOCKACQUIRE_NOT_AVAIL);
 }
 
 /*
@@ -749,7 +766,7 @@ ConditionalXactLockTableWait(TransactionId xid, bool logLockFailure)
 		SET_LOCKTAG_TRANSACTION(tag, xid);
 
 		if (LockAcquireExtended(&tag, ShareLock, false, true, true, NULL,
-								logLockFailure)
+								logLockFailure, false)
 			== LOCKACQUIRE_NOT_AVAIL)
 			return false;
 
@@ -1043,7 +1060,7 @@ ConditionalLockDatabaseObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -1123,7 +1140,7 @@ ConditionalLockSharedObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
diff --git a/src/backend/storage/lmgr/lock.c b/src/backend/storage/lmgr/lock.c
index c221fe96889..bcbbf36c091 100644
--- a/src/backend/storage/lmgr/lock.c
+++ b/src/backend/storage/lmgr/lock.c
@@ -810,7 +810,7 @@ LockAcquire(const LOCKTAG *locktag,
 			bool dontWait)
 {
 	return LockAcquireExtended(locktag, lockmode, sessionLock, dontWait,
-							   true, NULL, false);
+							   true, NULL, false, false);
 }
 
 /*
@@ -829,6 +829,9 @@ LockAcquire(const LOCKTAG *locktag,
  *
  * logLockFailure indicates whether to log details when a lock acquisition
  * fails with dontWait = true.
+ *
+ * isUpgrade should be true if the backend already holds this lock in a lower
+ * mode.
  */
 LockAcquireResult
 LockAcquireExtended(const LOCKTAG *locktag,
@@ -837,7 +840,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 					bool dontWait,
 					bool reportMemoryError,
 					LOCALLOCK **locallockp,
-					bool logLockFailure)
+					bool logLockFailure,
+					bool isUpgrade)
 {
 	LOCKMETHODID lockmethodid = locktag->locktag_lockmethodid;
 	LockMethod	lockMethodTable;
@@ -1119,7 +1123,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		 * case, because JoinWaitQueue() may discover that we can acquire the
 		 * lock immediately after all.
 		 */
-		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait);
+		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait,
+								   isUpgrade);
 	}
 
 	if (waitResult == PROC_WAIT_STATUS_ERROR)
@@ -1225,8 +1230,85 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		Assert(!dontWait);
 		PROCLOCK_PRINT("LockAcquire: sleeping on lock", proclock);
 		LOCK_PRINT("LockAcquire: sleeping on lock", lock, lockmode);
+
+		/*
+		 * Lock upgrade can introduce deadlock, therefore enforce special
+		 * behavior of other processes that deal with this lock. In
+		 * particular, any waiter that sees upgradedBy set is expected to
+		 * perform deadlock check as soon as it's woken up. (Since we are
+		 * already in the queue, the deadlock detector has all the information
+		 * it needs.)
+		 */
+		if (isUpgrade)
+		{
+			dlist_iter	iter;
+
+			/*
+			 * There should not be multiple upgrades at the same time. XXX
+			 * ERROR ?
+			 */
+			Assert(lock->upgradedBy == NULL);
+
+			lock->upgradedBy = MyProc;
+
+			/*
+			 * Now, before I sleep myself, wake up all the existing waiters
+			 * (except for me), so they check for deadlock.
+			 */
+			dclist_foreach(iter, &lock->waitProcs)
+			{
+				PGPROC	   *proc = dlist_container(PGPROC, waitLink,
+												   iter.cur);
+
+				if (proc != MyProc)
+					SetLatch(&proc->procLatch);
+			}
+		}
 		LWLockRelease(partitionLock);
 
+		if (!isUpgrade)
+		{
+			HASH_SEQ_STATUS status;
+			LOCALLOCK	*llock;
+			bool	check_deadlock = false;
+
+			/*
+			 * Likewise, new waiters should check the isUpgrade flag and
+			 * engage deadlock detector if needed. The lock being acquired
+			 * should already be in the hash.
+			 *
+			 * Note that, to make deadlock detection happen as soon as
+			 * possible (i.e. regardless deadlock timeout), we check the other
+			 * locks too. XXX There might be ways to get into the deadlock
+			 * indirectly, but that needs more analysis. As long as the
+			 * upgrading process skips deadlock detection altogether, the
+			 * worst consequence of missing some lock wait cycle here is that
+			 * the other process will be kicked-off no sooner than after its
+			 * deadlock timeout has elapsed. (Which in turn means that the
+			 * lock upgrade will take longer than expected.)
+			 */
+			hash_seq_init(&status, LockMethodLocalHash);
+			while ((llock = (LOCALLOCK *) hash_seq_search(&status)) != NULL)
+			{
+				LOCK	*mylock = llock->lock;
+
+				if (mylock && mylock->upgradedBy)
+				{
+					check_deadlock = true;
+					break;
+				}
+			}
+			/* Check for deadlock if needed. */
+			if (check_deadlock)
+			{
+				DeadLockState	deadlock_state;
+
+				deadlock_state = CheckDeadLock();
+				if (deadlock_state == DS_HARD_DEADLOCK)
+					DeadLockReport();
+			}
+		}
+
 		waitResult = WaitOnLock(locallock, owner);
 
 		/*
@@ -1245,6 +1327,18 @@ LockAcquireExtended(const LOCKTAG *locktag,
 			DeadLockReport();
 			/* DeadLockReport() will not return */
 		}
+
+		/*
+		 * If finishing the lock upgrade, we should not get into a deadlock
+		 * anymore, so let others know that they do not have to care either.
+		 */
+		if (isUpgrade)
+		{
+			LWLockAcquire(partitionLock, LW_EXCLUSIVE);
+			Assert(lock->upgradedBy != NULL);
+			lock->upgradedBy = NULL;
+			LWLockRelease(partitionLock);
+		}
 	}
 	else
 		LWLockRelease(partitionLock);
@@ -1322,6 +1416,7 @@ SetupLockInTable(LockMethod lockMethodTable, PGPROC *proc,
 		lock->nGranted = 0;
 		MemSet(lock->requested, 0, sizeof(int) * MAX_LOCKMODES);
 		MemSet(lock->granted, 0, sizeof(int) * MAX_LOCKMODES);
+		lock->upgradedBy = NULL;
 		LOCK_PRINT("LockAcquire: new", lock, lockmode);
 	}
 	else
@@ -1768,6 +1863,12 @@ CleanUpLock(LOCK *lock, PROCLOCK *proclock,
 			elog(PANIC, "proclock table corrupted");
 	}
 
+	/*
+	 * Was this backend upgrading the lock?
+	 */
+	if (lock->upgradedBy == MyProc)
+		lock->upgradedBy = NULL;
+
 	if (lock->nRequested == 0)
 	{
 		/*
diff --git a/src/backend/storage/lmgr/proc.c b/src/backend/storage/lmgr/proc.c
index 1ac25068d62..7b31370db70 100644
--- a/src/backend/storage/lmgr/proc.c
+++ b/src/backend/storage/lmgr/proc.c
@@ -95,7 +95,6 @@ static volatile sig_atomic_t got_deadlock_timeout;
 static void RemoveProcFromArray(int code, Datum arg);
 static void ProcKill(int code, Datum arg);
 static void AuxiliaryProcKill(int code, Datum arg);
-static DeadLockState CheckDeadLock(void);
 
 
 /*
@@ -1143,7 +1142,8 @@ AuxiliaryPidGetProc(int pid)
  * NOTES: The process queue is now a priority queue for locking.
  */
 ProcWaitStatus
-JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
+JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait,
+			  bool isUpgrade)
 {
 	LOCKMODE	lockmode = locallock->tag.mode;
 	LOCK	   *lock = locallock->lock;
@@ -1230,8 +1230,12 @@ JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
 			/* Must he wait for me? */
 			if (lockMethodTable->conflictTab[proc->waitLockMode] & myHeldLocks)
 			{
-				/* Must I wait for him ? */
-				if (lockMethodTable->conflictTab[lockmode] & proc->heldLocks)
+				/*
+				 * Must I wait for him? I don't want a deadlock during lock
+				 * upgrade - other processes should fail on it.
+				 */
+				if ((lockMethodTable->conflictTab[lockmode] & proc->heldLocks) &&
+					!isUpgrade)
 				{
 					/*
 					 * Yes, so we have a deadlock.  Easiest way to clean up
@@ -1461,8 +1465,22 @@ ProcSleep(LOCALLOCK *locallock)
 			(void) WaitLatch(MyLatch, WL_LATCH_SET | WL_EXIT_ON_PM_DEATH, 0,
 							 PG_WAIT_LOCK | locallock->tag.lock.locktag_type);
 			ResetLatch(MyLatch);
-			/* check for deadlocks first, as that's probably log-worthy */
-			if (got_deadlock_timeout)
+			/*
+			 * Check for deadlocks first, as that's probably log-worthy.
+			 *
+			 * Do not wait for the timeout if the lock is being upgraded since
+			 * the risk of deadlock is higher now. However, do not check for
+			 * deadlock if the lock is being upgraded by this process - other
+			 * processes should take care.
+			 *
+			 * TODO Possible optimization: if this is the only lock of the
+			 * backend and if it did not have any weaker lock on the table so
+			 * far, it should be safe to skip the deadlock check. However, to
+			 * evaluate the situation, we need to take fast-path locks into
+			 * account.
+			 */
+			if ((got_deadlock_timeout || lock->upgradedBy) &&
+				lock->upgradedBy != MyProc)
 			{
 				deadlock_state = CheckDeadLock();
 				got_deadlock_timeout = false;
@@ -1819,7 +1837,7 @@ ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock)
  * not, just return.  If we have a real deadlock, remove ourselves from the
  * lock's wait queue.
  */
-static DeadLockState
+DeadLockState
 CheckDeadLock(void)
 {
 	int			i;
diff --git a/src/include/storage/lmgr.h b/src/include/storage/lmgr.h
index 2a985ce5e15..eddbe1e8622 100644
--- a/src/include/storage/lmgr.h
+++ b/src/include/storage/lmgr.h
@@ -38,6 +38,7 @@ extern void RelationInitLockInfo(Relation relation);
 
 /* Lock a relation */
 extern void LockRelationOid(Oid relid, LOCKMODE lockmode);
+extern void LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode);
 extern void LockRelationId(LockRelId *relid, LOCKMODE lockmode);
 extern bool ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode);
 extern void UnlockRelationId(LockRelId *relid, LOCKMODE lockmode);
diff --git a/src/include/storage/lock.h b/src/include/storage/lock.h
index ee3cb1dc203..080b69eddfd 100644
--- a/src/include/storage/lock.h
+++ b/src/include/storage/lock.h
@@ -131,6 +131,7 @@ typedef const LockMethodData *LockMethod;
  * nRequested -- total requested locks of all types.
  * granted -- count of each lock type currently granted on the lock.
  * nGranted -- total granted locks of all types.
+ * upgrader -- process that intends to upgrade the lock
  *
  * Note: these counts count 1 for each backend.  Internally to a backend,
  * there may be multiple grabs on a particular lock, but this is not reflected
@@ -150,6 +151,7 @@ typedef struct LOCK
 	int			nRequested;		/* total of requested[] array */
 	int			granted[MAX_LOCKMODES]; /* counts of granted locks */
 	int			nGranted;		/* total of granted[] array */
+	PGPROC		*upgradedBy;	/* is lock being upgraded by this process? */
 } LOCK;
 
 #define LOCK_LOCKMETHOD(lock) ((LOCKMETHODID) (lock).tag.locktag_lockmethodid)
@@ -390,7 +392,8 @@ extern LockAcquireResult LockAcquireExtended(const LOCKTAG *locktag,
 											 bool dontWait,
 											 bool reportMemoryError,
 											 LOCALLOCK **locallockp,
-											 bool logLockFailure);
+											 bool logLockFailure,
+											 bool isUpgrade);
 extern void AbortStrongLockAcquire(void);
 extern void MarkLockClear(LOCALLOCK *locallock);
 extern bool LockRelease(const LOCKTAG *locktag,
diff --git a/src/include/storage/proc.h b/src/include/storage/proc.h
index 3e1d1fad5f9..c4bc3a24044 100644
--- a/src/include/storage/proc.h
+++ b/src/include/storage/proc.h
@@ -563,10 +563,12 @@ extern bool HaveNFreeProcs(int n, int *nfree);
 extern void ProcReleaseLocks(bool isCommit);
 
 extern ProcWaitStatus JoinWaitQueue(LOCALLOCK *locallock,
-									LockMethod lockMethodTable, bool dontWait);
+									LockMethod lockMethodTable, bool dontWait,
+									bool isUpgrade);
 extern ProcWaitStatus ProcSleep(LOCALLOCK *locallock);
 extern void ProcWakeup(PGPROC *proc, ProcWaitStatus waitStatus);
 extern void ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock);
+extern DeadLockState CheckDeadLock(void);
 extern void CheckDeadLockAlert(void);
 extern void LockErrorCleanup(void);
 extern void GetLockHoldersAndWaiters(LOCALLOCK *locallock,
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 966+ messages in thread

* [PATCH] Lock upgrade without deadlocks.
@ 2026-04-17 13:13  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 966+ messages in thread

From: Antonin Houska @ 2026-04-17 13:13 UTC (permalink / raw)

For REPACK (CONCURRENTLY), it's essential to upgrade ShareUpdateExclusive lock
to AccessExclusiveLock for the final stage of table processing. However, we
cannot release the earlier before requesting the latter, because if another
process ran DDL on the table in between, REPACK would have to cancel its
transaction, and waste possibly a lot of work. Without releasing
ShareUpdateExclusive lock, the upgrade can cause deadlock, and it's not really
deterministic whether REPACK or the other process will get canceled.

This patch adds a new field to the LOCK structure which tells which process is
trying to upgrade the lock - this field is set right before the upgrade. All
processes waiting in the lock queue are then woken up to check the flag.

When another process tries to get the lock after that, it checks this field,
and if it's set, it checks for deadlocks even if deadlock timeout hasn't
expired yet. If the process is already in the lock's queue and sleeping, the
lock upgrading process wakes it up so it checks the flag immediately.

At the time the upgrading process starts to sleep, all the other process
should be aware that they need to check for deadlock during their lock
acquisitions, so the upgrading process does not have to check for deadlocks
when it gets woken up. Thus it should never receive deadlock error report.
---
 src/backend/commands/repack.c   |   4 +-
 src/backend/storage/lmgr/lmgr.c |  35 ++++++++---
 src/backend/storage/lmgr/lock.c | 107 +++++++++++++++++++++++++++++++-
 src/backend/storage/lmgr/proc.c |  32 +++++++---
 src/include/storage/lmgr.h      |   1 +
 src/include/storage/lock.h      |   5 +-
 src/include/storage/proc.h      |   4 +-
 7 files changed, 166 insertions(+), 22 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index 58e3867246f..310b2a65099 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -3055,8 +3055,10 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Acquire AccessExclusiveLock on the table, its TOAST relation (if there
 	 * is one), all its indexes, so that we can swap the files.
+	 *
+	 * TODO The same for indexes and TOAST?
 	 */
-	LockRelationOid(old_table_oid, AccessExclusiveLock);
+	LockRelationOidUpgrade(old_table_oid, AccessExclusiveLock);
 
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
diff --git a/src/backend/storage/lmgr/lmgr.c b/src/backend/storage/lmgr/lmgr.c
index 2ccf7237fee..0b103485d54 100644
--- a/src/backend/storage/lmgr/lmgr.c
+++ b/src/backend/storage/lmgr/lmgr.c
@@ -58,6 +58,8 @@ typedef struct XactLockTableWaitInfo
 	const ItemPointerData *ctid;
 } XactLockTableWaitInfo;
 
+static void LockRelationOidCommon(Oid relid, LOCKMODE lockmode,
+								  bool isUpgrade);
 static void XactLockTableWaitErrorCb(void *arg);
 
 /*
@@ -105,6 +107,21 @@ SetLocktagRelationOid(LOCKTAG *tag, Oid relid)
  */
 void
 LockRelationOid(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, false);
+}
+
+/*
+ * Like above, but upgrade an existing lock.
+ */
+void
+LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, true);
+}
+
+static void
+LockRelationOidCommon(Oid relid, LOCKMODE lockmode, bool isUpgrade)
 {
 	LOCKTAG		tag;
 	LOCALLOCK  *locallock;
@@ -113,7 +130,7 @@ LockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, isUpgrade);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages, so that we
@@ -157,7 +174,7 @@ ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -191,7 +208,7 @@ LockRelationId(LockRelId *relid, LOCKMODE lockmode)
 	SET_LOCKTAG_RELATION(tag, relid->dbId, relid->relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -254,7 +271,7 @@ LockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -286,7 +303,7 @@ ConditionalLockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -591,7 +608,7 @@ ConditionalLockTuple(Relation relation, const ItemPointerData *tid, LOCKMODE loc
 					  ItemPointerGetOffsetNumber(tid));
 
 	return (LockAcquireExtended(&tag, lockmode, false, true, true, NULL,
-								logLockFailure) != LOCKACQUIRE_NOT_AVAIL);
+								logLockFailure, false) != LOCKACQUIRE_NOT_AVAIL);
 }
 
 /*
@@ -749,7 +766,7 @@ ConditionalXactLockTableWait(TransactionId xid, bool logLockFailure)
 		SET_LOCKTAG_TRANSACTION(tag, xid);
 
 		if (LockAcquireExtended(&tag, ShareLock, false, true, true, NULL,
-								logLockFailure)
+								logLockFailure, false)
 			== LOCKACQUIRE_NOT_AVAIL)
 			return false;
 
@@ -1043,7 +1060,7 @@ ConditionalLockDatabaseObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -1123,7 +1140,7 @@ ConditionalLockSharedObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
diff --git a/src/backend/storage/lmgr/lock.c b/src/backend/storage/lmgr/lock.c
index c221fe96889..bcbbf36c091 100644
--- a/src/backend/storage/lmgr/lock.c
+++ b/src/backend/storage/lmgr/lock.c
@@ -810,7 +810,7 @@ LockAcquire(const LOCKTAG *locktag,
 			bool dontWait)
 {
 	return LockAcquireExtended(locktag, lockmode, sessionLock, dontWait,
-							   true, NULL, false);
+							   true, NULL, false, false);
 }
 
 /*
@@ -829,6 +829,9 @@ LockAcquire(const LOCKTAG *locktag,
  *
  * logLockFailure indicates whether to log details when a lock acquisition
  * fails with dontWait = true.
+ *
+ * isUpgrade should be true if the backend already holds this lock in a lower
+ * mode.
  */
 LockAcquireResult
 LockAcquireExtended(const LOCKTAG *locktag,
@@ -837,7 +840,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 					bool dontWait,
 					bool reportMemoryError,
 					LOCALLOCK **locallockp,
-					bool logLockFailure)
+					bool logLockFailure,
+					bool isUpgrade)
 {
 	LOCKMETHODID lockmethodid = locktag->locktag_lockmethodid;
 	LockMethod	lockMethodTable;
@@ -1119,7 +1123,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		 * case, because JoinWaitQueue() may discover that we can acquire the
 		 * lock immediately after all.
 		 */
-		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait);
+		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait,
+								   isUpgrade);
 	}
 
 	if (waitResult == PROC_WAIT_STATUS_ERROR)
@@ -1225,8 +1230,85 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		Assert(!dontWait);
 		PROCLOCK_PRINT("LockAcquire: sleeping on lock", proclock);
 		LOCK_PRINT("LockAcquire: sleeping on lock", lock, lockmode);
+
+		/*
+		 * Lock upgrade can introduce deadlock, therefore enforce special
+		 * behavior of other processes that deal with this lock. In
+		 * particular, any waiter that sees upgradedBy set is expected to
+		 * perform deadlock check as soon as it's woken up. (Since we are
+		 * already in the queue, the deadlock detector has all the information
+		 * it needs.)
+		 */
+		if (isUpgrade)
+		{
+			dlist_iter	iter;
+
+			/*
+			 * There should not be multiple upgrades at the same time. XXX
+			 * ERROR ?
+			 */
+			Assert(lock->upgradedBy == NULL);
+
+			lock->upgradedBy = MyProc;
+
+			/*
+			 * Now, before I sleep myself, wake up all the existing waiters
+			 * (except for me), so they check for deadlock.
+			 */
+			dclist_foreach(iter, &lock->waitProcs)
+			{
+				PGPROC	   *proc = dlist_container(PGPROC, waitLink,
+												   iter.cur);
+
+				if (proc != MyProc)
+					SetLatch(&proc->procLatch);
+			}
+		}
 		LWLockRelease(partitionLock);
 
+		if (!isUpgrade)
+		{
+			HASH_SEQ_STATUS status;
+			LOCALLOCK	*llock;
+			bool	check_deadlock = false;
+
+			/*
+			 * Likewise, new waiters should check the isUpgrade flag and
+			 * engage deadlock detector if needed. The lock being acquired
+			 * should already be in the hash.
+			 *
+			 * Note that, to make deadlock detection happen as soon as
+			 * possible (i.e. regardless deadlock timeout), we check the other
+			 * locks too. XXX There might be ways to get into the deadlock
+			 * indirectly, but that needs more analysis. As long as the
+			 * upgrading process skips deadlock detection altogether, the
+			 * worst consequence of missing some lock wait cycle here is that
+			 * the other process will be kicked-off no sooner than after its
+			 * deadlock timeout has elapsed. (Which in turn means that the
+			 * lock upgrade will take longer than expected.)
+			 */
+			hash_seq_init(&status, LockMethodLocalHash);
+			while ((llock = (LOCALLOCK *) hash_seq_search(&status)) != NULL)
+			{
+				LOCK	*mylock = llock->lock;
+
+				if (mylock && mylock->upgradedBy)
+				{
+					check_deadlock = true;
+					break;
+				}
+			}
+			/* Check for deadlock if needed. */
+			if (check_deadlock)
+			{
+				DeadLockState	deadlock_state;
+
+				deadlock_state = CheckDeadLock();
+				if (deadlock_state == DS_HARD_DEADLOCK)
+					DeadLockReport();
+			}
+		}
+
 		waitResult = WaitOnLock(locallock, owner);
 
 		/*
@@ -1245,6 +1327,18 @@ LockAcquireExtended(const LOCKTAG *locktag,
 			DeadLockReport();
 			/* DeadLockReport() will not return */
 		}
+
+		/*
+		 * If finishing the lock upgrade, we should not get into a deadlock
+		 * anymore, so let others know that they do not have to care either.
+		 */
+		if (isUpgrade)
+		{
+			LWLockAcquire(partitionLock, LW_EXCLUSIVE);
+			Assert(lock->upgradedBy != NULL);
+			lock->upgradedBy = NULL;
+			LWLockRelease(partitionLock);
+		}
 	}
 	else
 		LWLockRelease(partitionLock);
@@ -1322,6 +1416,7 @@ SetupLockInTable(LockMethod lockMethodTable, PGPROC *proc,
 		lock->nGranted = 0;
 		MemSet(lock->requested, 0, sizeof(int) * MAX_LOCKMODES);
 		MemSet(lock->granted, 0, sizeof(int) * MAX_LOCKMODES);
+		lock->upgradedBy = NULL;
 		LOCK_PRINT("LockAcquire: new", lock, lockmode);
 	}
 	else
@@ -1768,6 +1863,12 @@ CleanUpLock(LOCK *lock, PROCLOCK *proclock,
 			elog(PANIC, "proclock table corrupted");
 	}
 
+	/*
+	 * Was this backend upgrading the lock?
+	 */
+	if (lock->upgradedBy == MyProc)
+		lock->upgradedBy = NULL;
+
 	if (lock->nRequested == 0)
 	{
 		/*
diff --git a/src/backend/storage/lmgr/proc.c b/src/backend/storage/lmgr/proc.c
index 1ac25068d62..7b31370db70 100644
--- a/src/backend/storage/lmgr/proc.c
+++ b/src/backend/storage/lmgr/proc.c
@@ -95,7 +95,6 @@ static volatile sig_atomic_t got_deadlock_timeout;
 static void RemoveProcFromArray(int code, Datum arg);
 static void ProcKill(int code, Datum arg);
 static void AuxiliaryProcKill(int code, Datum arg);
-static DeadLockState CheckDeadLock(void);
 
 
 /*
@@ -1143,7 +1142,8 @@ AuxiliaryPidGetProc(int pid)
  * NOTES: The process queue is now a priority queue for locking.
  */
 ProcWaitStatus
-JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
+JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait,
+			  bool isUpgrade)
 {
 	LOCKMODE	lockmode = locallock->tag.mode;
 	LOCK	   *lock = locallock->lock;
@@ -1230,8 +1230,12 @@ JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
 			/* Must he wait for me? */
 			if (lockMethodTable->conflictTab[proc->waitLockMode] & myHeldLocks)
 			{
-				/* Must I wait for him ? */
-				if (lockMethodTable->conflictTab[lockmode] & proc->heldLocks)
+				/*
+				 * Must I wait for him? I don't want a deadlock during lock
+				 * upgrade - other processes should fail on it.
+				 */
+				if ((lockMethodTable->conflictTab[lockmode] & proc->heldLocks) &&
+					!isUpgrade)
 				{
 					/*
 					 * Yes, so we have a deadlock.  Easiest way to clean up
@@ -1461,8 +1465,22 @@ ProcSleep(LOCALLOCK *locallock)
 			(void) WaitLatch(MyLatch, WL_LATCH_SET | WL_EXIT_ON_PM_DEATH, 0,
 							 PG_WAIT_LOCK | locallock->tag.lock.locktag_type);
 			ResetLatch(MyLatch);
-			/* check for deadlocks first, as that's probably log-worthy */
-			if (got_deadlock_timeout)
+			/*
+			 * Check for deadlocks first, as that's probably log-worthy.
+			 *
+			 * Do not wait for the timeout if the lock is being upgraded since
+			 * the risk of deadlock is higher now. However, do not check for
+			 * deadlock if the lock is being upgraded by this process - other
+			 * processes should take care.
+			 *
+			 * TODO Possible optimization: if this is the only lock of the
+			 * backend and if it did not have any weaker lock on the table so
+			 * far, it should be safe to skip the deadlock check. However, to
+			 * evaluate the situation, we need to take fast-path locks into
+			 * account.
+			 */
+			if ((got_deadlock_timeout || lock->upgradedBy) &&
+				lock->upgradedBy != MyProc)
 			{
 				deadlock_state = CheckDeadLock();
 				got_deadlock_timeout = false;
@@ -1819,7 +1837,7 @@ ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock)
  * not, just return.  If we have a real deadlock, remove ourselves from the
  * lock's wait queue.
  */
-static DeadLockState
+DeadLockState
 CheckDeadLock(void)
 {
 	int			i;
diff --git a/src/include/storage/lmgr.h b/src/include/storage/lmgr.h
index 2a985ce5e15..eddbe1e8622 100644
--- a/src/include/storage/lmgr.h
+++ b/src/include/storage/lmgr.h
@@ -38,6 +38,7 @@ extern void RelationInitLockInfo(Relation relation);
 
 /* Lock a relation */
 extern void LockRelationOid(Oid relid, LOCKMODE lockmode);
+extern void LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode);
 extern void LockRelationId(LockRelId *relid, LOCKMODE lockmode);
 extern bool ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode);
 extern void UnlockRelationId(LockRelId *relid, LOCKMODE lockmode);
diff --git a/src/include/storage/lock.h b/src/include/storage/lock.h
index ee3cb1dc203..080b69eddfd 100644
--- a/src/include/storage/lock.h
+++ b/src/include/storage/lock.h
@@ -131,6 +131,7 @@ typedef const LockMethodData *LockMethod;
  * nRequested -- total requested locks of all types.
  * granted -- count of each lock type currently granted on the lock.
  * nGranted -- total granted locks of all types.
+ * upgrader -- process that intends to upgrade the lock
  *
  * Note: these counts count 1 for each backend.  Internally to a backend,
  * there may be multiple grabs on a particular lock, but this is not reflected
@@ -150,6 +151,7 @@ typedef struct LOCK
 	int			nRequested;		/* total of requested[] array */
 	int			granted[MAX_LOCKMODES]; /* counts of granted locks */
 	int			nGranted;		/* total of granted[] array */
+	PGPROC		*upgradedBy;	/* is lock being upgraded by this process? */
 } LOCK;
 
 #define LOCK_LOCKMETHOD(lock) ((LOCKMETHODID) (lock).tag.locktag_lockmethodid)
@@ -390,7 +392,8 @@ extern LockAcquireResult LockAcquireExtended(const LOCKTAG *locktag,
 											 bool dontWait,
 											 bool reportMemoryError,
 											 LOCALLOCK **locallockp,
-											 bool logLockFailure);
+											 bool logLockFailure,
+											 bool isUpgrade);
 extern void AbortStrongLockAcquire(void);
 extern void MarkLockClear(LOCALLOCK *locallock);
 extern bool LockRelease(const LOCKTAG *locktag,
diff --git a/src/include/storage/proc.h b/src/include/storage/proc.h
index 3e1d1fad5f9..c4bc3a24044 100644
--- a/src/include/storage/proc.h
+++ b/src/include/storage/proc.h
@@ -563,10 +563,12 @@ extern bool HaveNFreeProcs(int n, int *nfree);
 extern void ProcReleaseLocks(bool isCommit);
 
 extern ProcWaitStatus JoinWaitQueue(LOCALLOCK *locallock,
-									LockMethod lockMethodTable, bool dontWait);
+									LockMethod lockMethodTable, bool dontWait,
+									bool isUpgrade);
 extern ProcWaitStatus ProcSleep(LOCALLOCK *locallock);
 extern void ProcWakeup(PGPROC *proc, ProcWaitStatus waitStatus);
 extern void ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock);
+extern DeadLockState CheckDeadLock(void);
 extern void CheckDeadLockAlert(void);
 extern void LockErrorCleanup(void);
 extern void GetLockHoldersAndWaiters(LOCALLOCK *locallock,
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 966+ messages in thread

* [PATCH] Lock upgrade without deadlocks.
@ 2026-04-17 13:13  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 966+ messages in thread

From: Antonin Houska @ 2026-04-17 13:13 UTC (permalink / raw)

For REPACK (CONCURRENTLY), it's essential to upgrade ShareUpdateExclusive lock
to AccessExclusiveLock for the final stage of table processing. However, we
cannot release the earlier before requesting the latter, because if another
process ran DDL on the table in between, REPACK would have to cancel its
transaction, and waste possibly a lot of work. Without releasing
ShareUpdateExclusive lock, the upgrade can cause deadlock, and it's not really
deterministic whether REPACK or the other process will get canceled.

This patch adds a new field to the LOCK structure which tells which process is
trying to upgrade the lock - this field is set right before the upgrade. All
processes waiting in the lock queue are then woken up to check the flag.

When another process tries to get the lock after that, it checks this field,
and if it's set, it checks for deadlocks even if deadlock timeout hasn't
expired yet. If the process is already in the lock's queue and sleeping, the
lock upgrading process wakes it up so it checks the flag immediately.

At the time the upgrading process starts to sleep, all the other process
should be aware that they need to check for deadlock during their lock
acquisitions, so the upgrading process does not have to check for deadlocks
when it gets woken up. Thus it should never receive deadlock error report.
---
 src/backend/commands/repack.c   |   4 +-
 src/backend/storage/lmgr/lmgr.c |  35 ++++++++---
 src/backend/storage/lmgr/lock.c | 107 +++++++++++++++++++++++++++++++-
 src/backend/storage/lmgr/proc.c |  32 +++++++---
 src/include/storage/lmgr.h      |   1 +
 src/include/storage/lock.h      |   5 +-
 src/include/storage/proc.h      |   4 +-
 7 files changed, 166 insertions(+), 22 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index 58e3867246f..310b2a65099 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -3055,8 +3055,10 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Acquire AccessExclusiveLock on the table, its TOAST relation (if there
 	 * is one), all its indexes, so that we can swap the files.
+	 *
+	 * TODO The same for indexes and TOAST?
 	 */
-	LockRelationOid(old_table_oid, AccessExclusiveLock);
+	LockRelationOidUpgrade(old_table_oid, AccessExclusiveLock);
 
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
diff --git a/src/backend/storage/lmgr/lmgr.c b/src/backend/storage/lmgr/lmgr.c
index 2ccf7237fee..0b103485d54 100644
--- a/src/backend/storage/lmgr/lmgr.c
+++ b/src/backend/storage/lmgr/lmgr.c
@@ -58,6 +58,8 @@ typedef struct XactLockTableWaitInfo
 	const ItemPointerData *ctid;
 } XactLockTableWaitInfo;
 
+static void LockRelationOidCommon(Oid relid, LOCKMODE lockmode,
+								  bool isUpgrade);
 static void XactLockTableWaitErrorCb(void *arg);
 
 /*
@@ -105,6 +107,21 @@ SetLocktagRelationOid(LOCKTAG *tag, Oid relid)
  */
 void
 LockRelationOid(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, false);
+}
+
+/*
+ * Like above, but upgrade an existing lock.
+ */
+void
+LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, true);
+}
+
+static void
+LockRelationOidCommon(Oid relid, LOCKMODE lockmode, bool isUpgrade)
 {
 	LOCKTAG		tag;
 	LOCALLOCK  *locallock;
@@ -113,7 +130,7 @@ LockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, isUpgrade);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages, so that we
@@ -157,7 +174,7 @@ ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -191,7 +208,7 @@ LockRelationId(LockRelId *relid, LOCKMODE lockmode)
 	SET_LOCKTAG_RELATION(tag, relid->dbId, relid->relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -254,7 +271,7 @@ LockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -286,7 +303,7 @@ ConditionalLockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -591,7 +608,7 @@ ConditionalLockTuple(Relation relation, const ItemPointerData *tid, LOCKMODE loc
 					  ItemPointerGetOffsetNumber(tid));
 
 	return (LockAcquireExtended(&tag, lockmode, false, true, true, NULL,
-								logLockFailure) != LOCKACQUIRE_NOT_AVAIL);
+								logLockFailure, false) != LOCKACQUIRE_NOT_AVAIL);
 }
 
 /*
@@ -749,7 +766,7 @@ ConditionalXactLockTableWait(TransactionId xid, bool logLockFailure)
 		SET_LOCKTAG_TRANSACTION(tag, xid);
 
 		if (LockAcquireExtended(&tag, ShareLock, false, true, true, NULL,
-								logLockFailure)
+								logLockFailure, false)
 			== LOCKACQUIRE_NOT_AVAIL)
 			return false;
 
@@ -1043,7 +1060,7 @@ ConditionalLockDatabaseObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -1123,7 +1140,7 @@ ConditionalLockSharedObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
diff --git a/src/backend/storage/lmgr/lock.c b/src/backend/storage/lmgr/lock.c
index c221fe96889..bcbbf36c091 100644
--- a/src/backend/storage/lmgr/lock.c
+++ b/src/backend/storage/lmgr/lock.c
@@ -810,7 +810,7 @@ LockAcquire(const LOCKTAG *locktag,
 			bool dontWait)
 {
 	return LockAcquireExtended(locktag, lockmode, sessionLock, dontWait,
-							   true, NULL, false);
+							   true, NULL, false, false);
 }
 
 /*
@@ -829,6 +829,9 @@ LockAcquire(const LOCKTAG *locktag,
  *
  * logLockFailure indicates whether to log details when a lock acquisition
  * fails with dontWait = true.
+ *
+ * isUpgrade should be true if the backend already holds this lock in a lower
+ * mode.
  */
 LockAcquireResult
 LockAcquireExtended(const LOCKTAG *locktag,
@@ -837,7 +840,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 					bool dontWait,
 					bool reportMemoryError,
 					LOCALLOCK **locallockp,
-					bool logLockFailure)
+					bool logLockFailure,
+					bool isUpgrade)
 {
 	LOCKMETHODID lockmethodid = locktag->locktag_lockmethodid;
 	LockMethod	lockMethodTable;
@@ -1119,7 +1123,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		 * case, because JoinWaitQueue() may discover that we can acquire the
 		 * lock immediately after all.
 		 */
-		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait);
+		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait,
+								   isUpgrade);
 	}
 
 	if (waitResult == PROC_WAIT_STATUS_ERROR)
@@ -1225,8 +1230,85 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		Assert(!dontWait);
 		PROCLOCK_PRINT("LockAcquire: sleeping on lock", proclock);
 		LOCK_PRINT("LockAcquire: sleeping on lock", lock, lockmode);
+
+		/*
+		 * Lock upgrade can introduce deadlock, therefore enforce special
+		 * behavior of other processes that deal with this lock. In
+		 * particular, any waiter that sees upgradedBy set is expected to
+		 * perform deadlock check as soon as it's woken up. (Since we are
+		 * already in the queue, the deadlock detector has all the information
+		 * it needs.)
+		 */
+		if (isUpgrade)
+		{
+			dlist_iter	iter;
+
+			/*
+			 * There should not be multiple upgrades at the same time. XXX
+			 * ERROR ?
+			 */
+			Assert(lock->upgradedBy == NULL);
+
+			lock->upgradedBy = MyProc;
+
+			/*
+			 * Now, before I sleep myself, wake up all the existing waiters
+			 * (except for me), so they check for deadlock.
+			 */
+			dclist_foreach(iter, &lock->waitProcs)
+			{
+				PGPROC	   *proc = dlist_container(PGPROC, waitLink,
+												   iter.cur);
+
+				if (proc != MyProc)
+					SetLatch(&proc->procLatch);
+			}
+		}
 		LWLockRelease(partitionLock);
 
+		if (!isUpgrade)
+		{
+			HASH_SEQ_STATUS status;
+			LOCALLOCK	*llock;
+			bool	check_deadlock = false;
+
+			/*
+			 * Likewise, new waiters should check the isUpgrade flag and
+			 * engage deadlock detector if needed. The lock being acquired
+			 * should already be in the hash.
+			 *
+			 * Note that, to make deadlock detection happen as soon as
+			 * possible (i.e. regardless deadlock timeout), we check the other
+			 * locks too. XXX There might be ways to get into the deadlock
+			 * indirectly, but that needs more analysis. As long as the
+			 * upgrading process skips deadlock detection altogether, the
+			 * worst consequence of missing some lock wait cycle here is that
+			 * the other process will be kicked-off no sooner than after its
+			 * deadlock timeout has elapsed. (Which in turn means that the
+			 * lock upgrade will take longer than expected.)
+			 */
+			hash_seq_init(&status, LockMethodLocalHash);
+			while ((llock = (LOCALLOCK *) hash_seq_search(&status)) != NULL)
+			{
+				LOCK	*mylock = llock->lock;
+
+				if (mylock && mylock->upgradedBy)
+				{
+					check_deadlock = true;
+					break;
+				}
+			}
+			/* Check for deadlock if needed. */
+			if (check_deadlock)
+			{
+				DeadLockState	deadlock_state;
+
+				deadlock_state = CheckDeadLock();
+				if (deadlock_state == DS_HARD_DEADLOCK)
+					DeadLockReport();
+			}
+		}
+
 		waitResult = WaitOnLock(locallock, owner);
 
 		/*
@@ -1245,6 +1327,18 @@ LockAcquireExtended(const LOCKTAG *locktag,
 			DeadLockReport();
 			/* DeadLockReport() will not return */
 		}
+
+		/*
+		 * If finishing the lock upgrade, we should not get into a deadlock
+		 * anymore, so let others know that they do not have to care either.
+		 */
+		if (isUpgrade)
+		{
+			LWLockAcquire(partitionLock, LW_EXCLUSIVE);
+			Assert(lock->upgradedBy != NULL);
+			lock->upgradedBy = NULL;
+			LWLockRelease(partitionLock);
+		}
 	}
 	else
 		LWLockRelease(partitionLock);
@@ -1322,6 +1416,7 @@ SetupLockInTable(LockMethod lockMethodTable, PGPROC *proc,
 		lock->nGranted = 0;
 		MemSet(lock->requested, 0, sizeof(int) * MAX_LOCKMODES);
 		MemSet(lock->granted, 0, sizeof(int) * MAX_LOCKMODES);
+		lock->upgradedBy = NULL;
 		LOCK_PRINT("LockAcquire: new", lock, lockmode);
 	}
 	else
@@ -1768,6 +1863,12 @@ CleanUpLock(LOCK *lock, PROCLOCK *proclock,
 			elog(PANIC, "proclock table corrupted");
 	}
 
+	/*
+	 * Was this backend upgrading the lock?
+	 */
+	if (lock->upgradedBy == MyProc)
+		lock->upgradedBy = NULL;
+
 	if (lock->nRequested == 0)
 	{
 		/*
diff --git a/src/backend/storage/lmgr/proc.c b/src/backend/storage/lmgr/proc.c
index 1ac25068d62..7b31370db70 100644
--- a/src/backend/storage/lmgr/proc.c
+++ b/src/backend/storage/lmgr/proc.c
@@ -95,7 +95,6 @@ static volatile sig_atomic_t got_deadlock_timeout;
 static void RemoveProcFromArray(int code, Datum arg);
 static void ProcKill(int code, Datum arg);
 static void AuxiliaryProcKill(int code, Datum arg);
-static DeadLockState CheckDeadLock(void);
 
 
 /*
@@ -1143,7 +1142,8 @@ AuxiliaryPidGetProc(int pid)
  * NOTES: The process queue is now a priority queue for locking.
  */
 ProcWaitStatus
-JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
+JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait,
+			  bool isUpgrade)
 {
 	LOCKMODE	lockmode = locallock->tag.mode;
 	LOCK	   *lock = locallock->lock;
@@ -1230,8 +1230,12 @@ JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
 			/* Must he wait for me? */
 			if (lockMethodTable->conflictTab[proc->waitLockMode] & myHeldLocks)
 			{
-				/* Must I wait for him ? */
-				if (lockMethodTable->conflictTab[lockmode] & proc->heldLocks)
+				/*
+				 * Must I wait for him? I don't want a deadlock during lock
+				 * upgrade - other processes should fail on it.
+				 */
+				if ((lockMethodTable->conflictTab[lockmode] & proc->heldLocks) &&
+					!isUpgrade)
 				{
 					/*
 					 * Yes, so we have a deadlock.  Easiest way to clean up
@@ -1461,8 +1465,22 @@ ProcSleep(LOCALLOCK *locallock)
 			(void) WaitLatch(MyLatch, WL_LATCH_SET | WL_EXIT_ON_PM_DEATH, 0,
 							 PG_WAIT_LOCK | locallock->tag.lock.locktag_type);
 			ResetLatch(MyLatch);
-			/* check for deadlocks first, as that's probably log-worthy */
-			if (got_deadlock_timeout)
+			/*
+			 * Check for deadlocks first, as that's probably log-worthy.
+			 *
+			 * Do not wait for the timeout if the lock is being upgraded since
+			 * the risk of deadlock is higher now. However, do not check for
+			 * deadlock if the lock is being upgraded by this process - other
+			 * processes should take care.
+			 *
+			 * TODO Possible optimization: if this is the only lock of the
+			 * backend and if it did not have any weaker lock on the table so
+			 * far, it should be safe to skip the deadlock check. However, to
+			 * evaluate the situation, we need to take fast-path locks into
+			 * account.
+			 */
+			if ((got_deadlock_timeout || lock->upgradedBy) &&
+				lock->upgradedBy != MyProc)
 			{
 				deadlock_state = CheckDeadLock();
 				got_deadlock_timeout = false;
@@ -1819,7 +1837,7 @@ ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock)
  * not, just return.  If we have a real deadlock, remove ourselves from the
  * lock's wait queue.
  */
-static DeadLockState
+DeadLockState
 CheckDeadLock(void)
 {
 	int			i;
diff --git a/src/include/storage/lmgr.h b/src/include/storage/lmgr.h
index 2a985ce5e15..eddbe1e8622 100644
--- a/src/include/storage/lmgr.h
+++ b/src/include/storage/lmgr.h
@@ -38,6 +38,7 @@ extern void RelationInitLockInfo(Relation relation);
 
 /* Lock a relation */
 extern void LockRelationOid(Oid relid, LOCKMODE lockmode);
+extern void LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode);
 extern void LockRelationId(LockRelId *relid, LOCKMODE lockmode);
 extern bool ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode);
 extern void UnlockRelationId(LockRelId *relid, LOCKMODE lockmode);
diff --git a/src/include/storage/lock.h b/src/include/storage/lock.h
index ee3cb1dc203..080b69eddfd 100644
--- a/src/include/storage/lock.h
+++ b/src/include/storage/lock.h
@@ -131,6 +131,7 @@ typedef const LockMethodData *LockMethod;
  * nRequested -- total requested locks of all types.
  * granted -- count of each lock type currently granted on the lock.
  * nGranted -- total granted locks of all types.
+ * upgrader -- process that intends to upgrade the lock
  *
  * Note: these counts count 1 for each backend.  Internally to a backend,
  * there may be multiple grabs on a particular lock, but this is not reflected
@@ -150,6 +151,7 @@ typedef struct LOCK
 	int			nRequested;		/* total of requested[] array */
 	int			granted[MAX_LOCKMODES]; /* counts of granted locks */
 	int			nGranted;		/* total of granted[] array */
+	PGPROC		*upgradedBy;	/* is lock being upgraded by this process? */
 } LOCK;
 
 #define LOCK_LOCKMETHOD(lock) ((LOCKMETHODID) (lock).tag.locktag_lockmethodid)
@@ -390,7 +392,8 @@ extern LockAcquireResult LockAcquireExtended(const LOCKTAG *locktag,
 											 bool dontWait,
 											 bool reportMemoryError,
 											 LOCALLOCK **locallockp,
-											 bool logLockFailure);
+											 bool logLockFailure,
+											 bool isUpgrade);
 extern void AbortStrongLockAcquire(void);
 extern void MarkLockClear(LOCALLOCK *locallock);
 extern bool LockRelease(const LOCKTAG *locktag,
diff --git a/src/include/storage/proc.h b/src/include/storage/proc.h
index 3e1d1fad5f9..c4bc3a24044 100644
--- a/src/include/storage/proc.h
+++ b/src/include/storage/proc.h
@@ -563,10 +563,12 @@ extern bool HaveNFreeProcs(int n, int *nfree);
 extern void ProcReleaseLocks(bool isCommit);
 
 extern ProcWaitStatus JoinWaitQueue(LOCALLOCK *locallock,
-									LockMethod lockMethodTable, bool dontWait);
+									LockMethod lockMethodTable, bool dontWait,
+									bool isUpgrade);
 extern ProcWaitStatus ProcSleep(LOCALLOCK *locallock);
 extern void ProcWakeup(PGPROC *proc, ProcWaitStatus waitStatus);
 extern void ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock);
+extern DeadLockState CheckDeadLock(void);
 extern void CheckDeadLockAlert(void);
 extern void LockErrorCleanup(void);
 extern void GetLockHoldersAndWaiters(LOCALLOCK *locallock,
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 966+ messages in thread

* [PATCH] Lock upgrade without deadlocks.
@ 2026-04-17 13:13  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 966+ messages in thread

From: Antonin Houska @ 2026-04-17 13:13 UTC (permalink / raw)

For REPACK (CONCURRENTLY), it's essential to upgrade ShareUpdateExclusive lock
to AccessExclusiveLock for the final stage of table processing. However, we
cannot release the earlier before requesting the latter, because if another
process ran DDL on the table in between, REPACK would have to cancel its
transaction, and waste possibly a lot of work. Without releasing
ShareUpdateExclusive lock, the upgrade can cause deadlock, and it's not really
deterministic whether REPACK or the other process will get canceled.

This patch adds a new field to the LOCK structure which tells which process is
trying to upgrade the lock - this field is set right before the upgrade. All
processes waiting in the lock queue are then woken up to check the flag.

When another process tries to get the lock after that, it checks this field,
and if it's set, it checks for deadlocks even if deadlock timeout hasn't
expired yet. If the process is already in the lock's queue and sleeping, the
lock upgrading process wakes it up so it checks the flag immediately.

At the time the upgrading process starts to sleep, all the other process
should be aware that they need to check for deadlock during their lock
acquisitions, so the upgrading process does not have to check for deadlocks
when it gets woken up. Thus it should never receive deadlock error report.
---
 src/backend/commands/repack.c   |   4 +-
 src/backend/storage/lmgr/lmgr.c |  35 ++++++++---
 src/backend/storage/lmgr/lock.c | 107 +++++++++++++++++++++++++++++++-
 src/backend/storage/lmgr/proc.c |  32 +++++++---
 src/include/storage/lmgr.h      |   1 +
 src/include/storage/lock.h      |   5 +-
 src/include/storage/proc.h      |   4 +-
 7 files changed, 166 insertions(+), 22 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index 58e3867246f..310b2a65099 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -3055,8 +3055,10 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Acquire AccessExclusiveLock on the table, its TOAST relation (if there
 	 * is one), all its indexes, so that we can swap the files.
+	 *
+	 * TODO The same for indexes and TOAST?
 	 */
-	LockRelationOid(old_table_oid, AccessExclusiveLock);
+	LockRelationOidUpgrade(old_table_oid, AccessExclusiveLock);
 
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
diff --git a/src/backend/storage/lmgr/lmgr.c b/src/backend/storage/lmgr/lmgr.c
index 2ccf7237fee..0b103485d54 100644
--- a/src/backend/storage/lmgr/lmgr.c
+++ b/src/backend/storage/lmgr/lmgr.c
@@ -58,6 +58,8 @@ typedef struct XactLockTableWaitInfo
 	const ItemPointerData *ctid;
 } XactLockTableWaitInfo;
 
+static void LockRelationOidCommon(Oid relid, LOCKMODE lockmode,
+								  bool isUpgrade);
 static void XactLockTableWaitErrorCb(void *arg);
 
 /*
@@ -105,6 +107,21 @@ SetLocktagRelationOid(LOCKTAG *tag, Oid relid)
  */
 void
 LockRelationOid(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, false);
+}
+
+/*
+ * Like above, but upgrade an existing lock.
+ */
+void
+LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, true);
+}
+
+static void
+LockRelationOidCommon(Oid relid, LOCKMODE lockmode, bool isUpgrade)
 {
 	LOCKTAG		tag;
 	LOCALLOCK  *locallock;
@@ -113,7 +130,7 @@ LockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, isUpgrade);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages, so that we
@@ -157,7 +174,7 @@ ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -191,7 +208,7 @@ LockRelationId(LockRelId *relid, LOCKMODE lockmode)
 	SET_LOCKTAG_RELATION(tag, relid->dbId, relid->relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -254,7 +271,7 @@ LockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -286,7 +303,7 @@ ConditionalLockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -591,7 +608,7 @@ ConditionalLockTuple(Relation relation, const ItemPointerData *tid, LOCKMODE loc
 					  ItemPointerGetOffsetNumber(tid));
 
 	return (LockAcquireExtended(&tag, lockmode, false, true, true, NULL,
-								logLockFailure) != LOCKACQUIRE_NOT_AVAIL);
+								logLockFailure, false) != LOCKACQUIRE_NOT_AVAIL);
 }
 
 /*
@@ -749,7 +766,7 @@ ConditionalXactLockTableWait(TransactionId xid, bool logLockFailure)
 		SET_LOCKTAG_TRANSACTION(tag, xid);
 
 		if (LockAcquireExtended(&tag, ShareLock, false, true, true, NULL,
-								logLockFailure)
+								logLockFailure, false)
 			== LOCKACQUIRE_NOT_AVAIL)
 			return false;
 
@@ -1043,7 +1060,7 @@ ConditionalLockDatabaseObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -1123,7 +1140,7 @@ ConditionalLockSharedObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
diff --git a/src/backend/storage/lmgr/lock.c b/src/backend/storage/lmgr/lock.c
index c221fe96889..bcbbf36c091 100644
--- a/src/backend/storage/lmgr/lock.c
+++ b/src/backend/storage/lmgr/lock.c
@@ -810,7 +810,7 @@ LockAcquire(const LOCKTAG *locktag,
 			bool dontWait)
 {
 	return LockAcquireExtended(locktag, lockmode, sessionLock, dontWait,
-							   true, NULL, false);
+							   true, NULL, false, false);
 }
 
 /*
@@ -829,6 +829,9 @@ LockAcquire(const LOCKTAG *locktag,
  *
  * logLockFailure indicates whether to log details when a lock acquisition
  * fails with dontWait = true.
+ *
+ * isUpgrade should be true if the backend already holds this lock in a lower
+ * mode.
  */
 LockAcquireResult
 LockAcquireExtended(const LOCKTAG *locktag,
@@ -837,7 +840,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 					bool dontWait,
 					bool reportMemoryError,
 					LOCALLOCK **locallockp,
-					bool logLockFailure)
+					bool logLockFailure,
+					bool isUpgrade)
 {
 	LOCKMETHODID lockmethodid = locktag->locktag_lockmethodid;
 	LockMethod	lockMethodTable;
@@ -1119,7 +1123,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		 * case, because JoinWaitQueue() may discover that we can acquire the
 		 * lock immediately after all.
 		 */
-		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait);
+		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait,
+								   isUpgrade);
 	}
 
 	if (waitResult == PROC_WAIT_STATUS_ERROR)
@@ -1225,8 +1230,85 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		Assert(!dontWait);
 		PROCLOCK_PRINT("LockAcquire: sleeping on lock", proclock);
 		LOCK_PRINT("LockAcquire: sleeping on lock", lock, lockmode);
+
+		/*
+		 * Lock upgrade can introduce deadlock, therefore enforce special
+		 * behavior of other processes that deal with this lock. In
+		 * particular, any waiter that sees upgradedBy set is expected to
+		 * perform deadlock check as soon as it's woken up. (Since we are
+		 * already in the queue, the deadlock detector has all the information
+		 * it needs.)
+		 */
+		if (isUpgrade)
+		{
+			dlist_iter	iter;
+
+			/*
+			 * There should not be multiple upgrades at the same time. XXX
+			 * ERROR ?
+			 */
+			Assert(lock->upgradedBy == NULL);
+
+			lock->upgradedBy = MyProc;
+
+			/*
+			 * Now, before I sleep myself, wake up all the existing waiters
+			 * (except for me), so they check for deadlock.
+			 */
+			dclist_foreach(iter, &lock->waitProcs)
+			{
+				PGPROC	   *proc = dlist_container(PGPROC, waitLink,
+												   iter.cur);
+
+				if (proc != MyProc)
+					SetLatch(&proc->procLatch);
+			}
+		}
 		LWLockRelease(partitionLock);
 
+		if (!isUpgrade)
+		{
+			HASH_SEQ_STATUS status;
+			LOCALLOCK	*llock;
+			bool	check_deadlock = false;
+
+			/*
+			 * Likewise, new waiters should check the isUpgrade flag and
+			 * engage deadlock detector if needed. The lock being acquired
+			 * should already be in the hash.
+			 *
+			 * Note that, to make deadlock detection happen as soon as
+			 * possible (i.e. regardless deadlock timeout), we check the other
+			 * locks too. XXX There might be ways to get into the deadlock
+			 * indirectly, but that needs more analysis. As long as the
+			 * upgrading process skips deadlock detection altogether, the
+			 * worst consequence of missing some lock wait cycle here is that
+			 * the other process will be kicked-off no sooner than after its
+			 * deadlock timeout has elapsed. (Which in turn means that the
+			 * lock upgrade will take longer than expected.)
+			 */
+			hash_seq_init(&status, LockMethodLocalHash);
+			while ((llock = (LOCALLOCK *) hash_seq_search(&status)) != NULL)
+			{
+				LOCK	*mylock = llock->lock;
+
+				if (mylock && mylock->upgradedBy)
+				{
+					check_deadlock = true;
+					break;
+				}
+			}
+			/* Check for deadlock if needed. */
+			if (check_deadlock)
+			{
+				DeadLockState	deadlock_state;
+
+				deadlock_state = CheckDeadLock();
+				if (deadlock_state == DS_HARD_DEADLOCK)
+					DeadLockReport();
+			}
+		}
+
 		waitResult = WaitOnLock(locallock, owner);
 
 		/*
@@ -1245,6 +1327,18 @@ LockAcquireExtended(const LOCKTAG *locktag,
 			DeadLockReport();
 			/* DeadLockReport() will not return */
 		}
+
+		/*
+		 * If finishing the lock upgrade, we should not get into a deadlock
+		 * anymore, so let others know that they do not have to care either.
+		 */
+		if (isUpgrade)
+		{
+			LWLockAcquire(partitionLock, LW_EXCLUSIVE);
+			Assert(lock->upgradedBy != NULL);
+			lock->upgradedBy = NULL;
+			LWLockRelease(partitionLock);
+		}
 	}
 	else
 		LWLockRelease(partitionLock);
@@ -1322,6 +1416,7 @@ SetupLockInTable(LockMethod lockMethodTable, PGPROC *proc,
 		lock->nGranted = 0;
 		MemSet(lock->requested, 0, sizeof(int) * MAX_LOCKMODES);
 		MemSet(lock->granted, 0, sizeof(int) * MAX_LOCKMODES);
+		lock->upgradedBy = NULL;
 		LOCK_PRINT("LockAcquire: new", lock, lockmode);
 	}
 	else
@@ -1768,6 +1863,12 @@ CleanUpLock(LOCK *lock, PROCLOCK *proclock,
 			elog(PANIC, "proclock table corrupted");
 	}
 
+	/*
+	 * Was this backend upgrading the lock?
+	 */
+	if (lock->upgradedBy == MyProc)
+		lock->upgradedBy = NULL;
+
 	if (lock->nRequested == 0)
 	{
 		/*
diff --git a/src/backend/storage/lmgr/proc.c b/src/backend/storage/lmgr/proc.c
index 1ac25068d62..7b31370db70 100644
--- a/src/backend/storage/lmgr/proc.c
+++ b/src/backend/storage/lmgr/proc.c
@@ -95,7 +95,6 @@ static volatile sig_atomic_t got_deadlock_timeout;
 static void RemoveProcFromArray(int code, Datum arg);
 static void ProcKill(int code, Datum arg);
 static void AuxiliaryProcKill(int code, Datum arg);
-static DeadLockState CheckDeadLock(void);
 
 
 /*
@@ -1143,7 +1142,8 @@ AuxiliaryPidGetProc(int pid)
  * NOTES: The process queue is now a priority queue for locking.
  */
 ProcWaitStatus
-JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
+JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait,
+			  bool isUpgrade)
 {
 	LOCKMODE	lockmode = locallock->tag.mode;
 	LOCK	   *lock = locallock->lock;
@@ -1230,8 +1230,12 @@ JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
 			/* Must he wait for me? */
 			if (lockMethodTable->conflictTab[proc->waitLockMode] & myHeldLocks)
 			{
-				/* Must I wait for him ? */
-				if (lockMethodTable->conflictTab[lockmode] & proc->heldLocks)
+				/*
+				 * Must I wait for him? I don't want a deadlock during lock
+				 * upgrade - other processes should fail on it.
+				 */
+				if ((lockMethodTable->conflictTab[lockmode] & proc->heldLocks) &&
+					!isUpgrade)
 				{
 					/*
 					 * Yes, so we have a deadlock.  Easiest way to clean up
@@ -1461,8 +1465,22 @@ ProcSleep(LOCALLOCK *locallock)
 			(void) WaitLatch(MyLatch, WL_LATCH_SET | WL_EXIT_ON_PM_DEATH, 0,
 							 PG_WAIT_LOCK | locallock->tag.lock.locktag_type);
 			ResetLatch(MyLatch);
-			/* check for deadlocks first, as that's probably log-worthy */
-			if (got_deadlock_timeout)
+			/*
+			 * Check for deadlocks first, as that's probably log-worthy.
+			 *
+			 * Do not wait for the timeout if the lock is being upgraded since
+			 * the risk of deadlock is higher now. However, do not check for
+			 * deadlock if the lock is being upgraded by this process - other
+			 * processes should take care.
+			 *
+			 * TODO Possible optimization: if this is the only lock of the
+			 * backend and if it did not have any weaker lock on the table so
+			 * far, it should be safe to skip the deadlock check. However, to
+			 * evaluate the situation, we need to take fast-path locks into
+			 * account.
+			 */
+			if ((got_deadlock_timeout || lock->upgradedBy) &&
+				lock->upgradedBy != MyProc)
 			{
 				deadlock_state = CheckDeadLock();
 				got_deadlock_timeout = false;
@@ -1819,7 +1837,7 @@ ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock)
  * not, just return.  If we have a real deadlock, remove ourselves from the
  * lock's wait queue.
  */
-static DeadLockState
+DeadLockState
 CheckDeadLock(void)
 {
 	int			i;
diff --git a/src/include/storage/lmgr.h b/src/include/storage/lmgr.h
index 2a985ce5e15..eddbe1e8622 100644
--- a/src/include/storage/lmgr.h
+++ b/src/include/storage/lmgr.h
@@ -38,6 +38,7 @@ extern void RelationInitLockInfo(Relation relation);
 
 /* Lock a relation */
 extern void LockRelationOid(Oid relid, LOCKMODE lockmode);
+extern void LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode);
 extern void LockRelationId(LockRelId *relid, LOCKMODE lockmode);
 extern bool ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode);
 extern void UnlockRelationId(LockRelId *relid, LOCKMODE lockmode);
diff --git a/src/include/storage/lock.h b/src/include/storage/lock.h
index ee3cb1dc203..080b69eddfd 100644
--- a/src/include/storage/lock.h
+++ b/src/include/storage/lock.h
@@ -131,6 +131,7 @@ typedef const LockMethodData *LockMethod;
  * nRequested -- total requested locks of all types.
  * granted -- count of each lock type currently granted on the lock.
  * nGranted -- total granted locks of all types.
+ * upgrader -- process that intends to upgrade the lock
  *
  * Note: these counts count 1 for each backend.  Internally to a backend,
  * there may be multiple grabs on a particular lock, but this is not reflected
@@ -150,6 +151,7 @@ typedef struct LOCK
 	int			nRequested;		/* total of requested[] array */
 	int			granted[MAX_LOCKMODES]; /* counts of granted locks */
 	int			nGranted;		/* total of granted[] array */
+	PGPROC		*upgradedBy;	/* is lock being upgraded by this process? */
 } LOCK;
 
 #define LOCK_LOCKMETHOD(lock) ((LOCKMETHODID) (lock).tag.locktag_lockmethodid)
@@ -390,7 +392,8 @@ extern LockAcquireResult LockAcquireExtended(const LOCKTAG *locktag,
 											 bool dontWait,
 											 bool reportMemoryError,
 											 LOCALLOCK **locallockp,
-											 bool logLockFailure);
+											 bool logLockFailure,
+											 bool isUpgrade);
 extern void AbortStrongLockAcquire(void);
 extern void MarkLockClear(LOCALLOCK *locallock);
 extern bool LockRelease(const LOCKTAG *locktag,
diff --git a/src/include/storage/proc.h b/src/include/storage/proc.h
index 3e1d1fad5f9..c4bc3a24044 100644
--- a/src/include/storage/proc.h
+++ b/src/include/storage/proc.h
@@ -563,10 +563,12 @@ extern bool HaveNFreeProcs(int n, int *nfree);
 extern void ProcReleaseLocks(bool isCommit);
 
 extern ProcWaitStatus JoinWaitQueue(LOCALLOCK *locallock,
-									LockMethod lockMethodTable, bool dontWait);
+									LockMethod lockMethodTable, bool dontWait,
+									bool isUpgrade);
 extern ProcWaitStatus ProcSleep(LOCALLOCK *locallock);
 extern void ProcWakeup(PGPROC *proc, ProcWaitStatus waitStatus);
 extern void ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock);
+extern DeadLockState CheckDeadLock(void);
 extern void CheckDeadLockAlert(void);
 extern void LockErrorCleanup(void);
 extern void GetLockHoldersAndWaiters(LOCALLOCK *locallock,
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 966+ messages in thread

* [PATCH] Lock upgrade without deadlocks.
@ 2026-04-17 13:13  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 966+ messages in thread

From: Antonin Houska @ 2026-04-17 13:13 UTC (permalink / raw)

For REPACK (CONCURRENTLY), it's essential to upgrade ShareUpdateExclusive lock
to AccessExclusiveLock for the final stage of table processing. However, we
cannot release the earlier before requesting the latter, because if another
process ran DDL on the table in between, REPACK would have to cancel its
transaction, and waste possibly a lot of work. Without releasing
ShareUpdateExclusive lock, the upgrade can cause deadlock, and it's not really
deterministic whether REPACK or the other process will get canceled.

This patch adds a new field to the LOCK structure which tells which process is
trying to upgrade the lock - this field is set right before the upgrade. All
processes waiting in the lock queue are then woken up to check the flag.

When another process tries to get the lock after that, it checks this field,
and if it's set, it checks for deadlocks even if deadlock timeout hasn't
expired yet. If the process is already in the lock's queue and sleeping, the
lock upgrading process wakes it up so it checks the flag immediately.

At the time the upgrading process starts to sleep, all the other process
should be aware that they need to check for deadlock during their lock
acquisitions, so the upgrading process does not have to check for deadlocks
when it gets woken up. Thus it should never receive deadlock error report.
---
 src/backend/commands/repack.c   |   4 +-
 src/backend/storage/lmgr/lmgr.c |  35 ++++++++---
 src/backend/storage/lmgr/lock.c | 107 +++++++++++++++++++++++++++++++-
 src/backend/storage/lmgr/proc.c |  32 +++++++---
 src/include/storage/lmgr.h      |   1 +
 src/include/storage/lock.h      |   5 +-
 src/include/storage/proc.h      |   4 +-
 7 files changed, 166 insertions(+), 22 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index 58e3867246f..310b2a65099 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -3055,8 +3055,10 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Acquire AccessExclusiveLock on the table, its TOAST relation (if there
 	 * is one), all its indexes, so that we can swap the files.
+	 *
+	 * TODO The same for indexes and TOAST?
 	 */
-	LockRelationOid(old_table_oid, AccessExclusiveLock);
+	LockRelationOidUpgrade(old_table_oid, AccessExclusiveLock);
 
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
diff --git a/src/backend/storage/lmgr/lmgr.c b/src/backend/storage/lmgr/lmgr.c
index 2ccf7237fee..0b103485d54 100644
--- a/src/backend/storage/lmgr/lmgr.c
+++ b/src/backend/storage/lmgr/lmgr.c
@@ -58,6 +58,8 @@ typedef struct XactLockTableWaitInfo
 	const ItemPointerData *ctid;
 } XactLockTableWaitInfo;
 
+static void LockRelationOidCommon(Oid relid, LOCKMODE lockmode,
+								  bool isUpgrade);
 static void XactLockTableWaitErrorCb(void *arg);
 
 /*
@@ -105,6 +107,21 @@ SetLocktagRelationOid(LOCKTAG *tag, Oid relid)
  */
 void
 LockRelationOid(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, false);
+}
+
+/*
+ * Like above, but upgrade an existing lock.
+ */
+void
+LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, true);
+}
+
+static void
+LockRelationOidCommon(Oid relid, LOCKMODE lockmode, bool isUpgrade)
 {
 	LOCKTAG		tag;
 	LOCALLOCK  *locallock;
@@ -113,7 +130,7 @@ LockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, isUpgrade);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages, so that we
@@ -157,7 +174,7 @@ ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -191,7 +208,7 @@ LockRelationId(LockRelId *relid, LOCKMODE lockmode)
 	SET_LOCKTAG_RELATION(tag, relid->dbId, relid->relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -254,7 +271,7 @@ LockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -286,7 +303,7 @@ ConditionalLockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -591,7 +608,7 @@ ConditionalLockTuple(Relation relation, const ItemPointerData *tid, LOCKMODE loc
 					  ItemPointerGetOffsetNumber(tid));
 
 	return (LockAcquireExtended(&tag, lockmode, false, true, true, NULL,
-								logLockFailure) != LOCKACQUIRE_NOT_AVAIL);
+								logLockFailure, false) != LOCKACQUIRE_NOT_AVAIL);
 }
 
 /*
@@ -749,7 +766,7 @@ ConditionalXactLockTableWait(TransactionId xid, bool logLockFailure)
 		SET_LOCKTAG_TRANSACTION(tag, xid);
 
 		if (LockAcquireExtended(&tag, ShareLock, false, true, true, NULL,
-								logLockFailure)
+								logLockFailure, false)
 			== LOCKACQUIRE_NOT_AVAIL)
 			return false;
 
@@ -1043,7 +1060,7 @@ ConditionalLockDatabaseObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -1123,7 +1140,7 @@ ConditionalLockSharedObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
diff --git a/src/backend/storage/lmgr/lock.c b/src/backend/storage/lmgr/lock.c
index c221fe96889..bcbbf36c091 100644
--- a/src/backend/storage/lmgr/lock.c
+++ b/src/backend/storage/lmgr/lock.c
@@ -810,7 +810,7 @@ LockAcquire(const LOCKTAG *locktag,
 			bool dontWait)
 {
 	return LockAcquireExtended(locktag, lockmode, sessionLock, dontWait,
-							   true, NULL, false);
+							   true, NULL, false, false);
 }
 
 /*
@@ -829,6 +829,9 @@ LockAcquire(const LOCKTAG *locktag,
  *
  * logLockFailure indicates whether to log details when a lock acquisition
  * fails with dontWait = true.
+ *
+ * isUpgrade should be true if the backend already holds this lock in a lower
+ * mode.
  */
 LockAcquireResult
 LockAcquireExtended(const LOCKTAG *locktag,
@@ -837,7 +840,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 					bool dontWait,
 					bool reportMemoryError,
 					LOCALLOCK **locallockp,
-					bool logLockFailure)
+					bool logLockFailure,
+					bool isUpgrade)
 {
 	LOCKMETHODID lockmethodid = locktag->locktag_lockmethodid;
 	LockMethod	lockMethodTable;
@@ -1119,7 +1123,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		 * case, because JoinWaitQueue() may discover that we can acquire the
 		 * lock immediately after all.
 		 */
-		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait);
+		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait,
+								   isUpgrade);
 	}
 
 	if (waitResult == PROC_WAIT_STATUS_ERROR)
@@ -1225,8 +1230,85 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		Assert(!dontWait);
 		PROCLOCK_PRINT("LockAcquire: sleeping on lock", proclock);
 		LOCK_PRINT("LockAcquire: sleeping on lock", lock, lockmode);
+
+		/*
+		 * Lock upgrade can introduce deadlock, therefore enforce special
+		 * behavior of other processes that deal with this lock. In
+		 * particular, any waiter that sees upgradedBy set is expected to
+		 * perform deadlock check as soon as it's woken up. (Since we are
+		 * already in the queue, the deadlock detector has all the information
+		 * it needs.)
+		 */
+		if (isUpgrade)
+		{
+			dlist_iter	iter;
+
+			/*
+			 * There should not be multiple upgrades at the same time. XXX
+			 * ERROR ?
+			 */
+			Assert(lock->upgradedBy == NULL);
+
+			lock->upgradedBy = MyProc;
+
+			/*
+			 * Now, before I sleep myself, wake up all the existing waiters
+			 * (except for me), so they check for deadlock.
+			 */
+			dclist_foreach(iter, &lock->waitProcs)
+			{
+				PGPROC	   *proc = dlist_container(PGPROC, waitLink,
+												   iter.cur);
+
+				if (proc != MyProc)
+					SetLatch(&proc->procLatch);
+			}
+		}
 		LWLockRelease(partitionLock);
 
+		if (!isUpgrade)
+		{
+			HASH_SEQ_STATUS status;
+			LOCALLOCK	*llock;
+			bool	check_deadlock = false;
+
+			/*
+			 * Likewise, new waiters should check the isUpgrade flag and
+			 * engage deadlock detector if needed. The lock being acquired
+			 * should already be in the hash.
+			 *
+			 * Note that, to make deadlock detection happen as soon as
+			 * possible (i.e. regardless deadlock timeout), we check the other
+			 * locks too. XXX There might be ways to get into the deadlock
+			 * indirectly, but that needs more analysis. As long as the
+			 * upgrading process skips deadlock detection altogether, the
+			 * worst consequence of missing some lock wait cycle here is that
+			 * the other process will be kicked-off no sooner than after its
+			 * deadlock timeout has elapsed. (Which in turn means that the
+			 * lock upgrade will take longer than expected.)
+			 */
+			hash_seq_init(&status, LockMethodLocalHash);
+			while ((llock = (LOCALLOCK *) hash_seq_search(&status)) != NULL)
+			{
+				LOCK	*mylock = llock->lock;
+
+				if (mylock && mylock->upgradedBy)
+				{
+					check_deadlock = true;
+					break;
+				}
+			}
+			/* Check for deadlock if needed. */
+			if (check_deadlock)
+			{
+				DeadLockState	deadlock_state;
+
+				deadlock_state = CheckDeadLock();
+				if (deadlock_state == DS_HARD_DEADLOCK)
+					DeadLockReport();
+			}
+		}
+
 		waitResult = WaitOnLock(locallock, owner);
 
 		/*
@@ -1245,6 +1327,18 @@ LockAcquireExtended(const LOCKTAG *locktag,
 			DeadLockReport();
 			/* DeadLockReport() will not return */
 		}
+
+		/*
+		 * If finishing the lock upgrade, we should not get into a deadlock
+		 * anymore, so let others know that they do not have to care either.
+		 */
+		if (isUpgrade)
+		{
+			LWLockAcquire(partitionLock, LW_EXCLUSIVE);
+			Assert(lock->upgradedBy != NULL);
+			lock->upgradedBy = NULL;
+			LWLockRelease(partitionLock);
+		}
 	}
 	else
 		LWLockRelease(partitionLock);
@@ -1322,6 +1416,7 @@ SetupLockInTable(LockMethod lockMethodTable, PGPROC *proc,
 		lock->nGranted = 0;
 		MemSet(lock->requested, 0, sizeof(int) * MAX_LOCKMODES);
 		MemSet(lock->granted, 0, sizeof(int) * MAX_LOCKMODES);
+		lock->upgradedBy = NULL;
 		LOCK_PRINT("LockAcquire: new", lock, lockmode);
 	}
 	else
@@ -1768,6 +1863,12 @@ CleanUpLock(LOCK *lock, PROCLOCK *proclock,
 			elog(PANIC, "proclock table corrupted");
 	}
 
+	/*
+	 * Was this backend upgrading the lock?
+	 */
+	if (lock->upgradedBy == MyProc)
+		lock->upgradedBy = NULL;
+
 	if (lock->nRequested == 0)
 	{
 		/*
diff --git a/src/backend/storage/lmgr/proc.c b/src/backend/storage/lmgr/proc.c
index 1ac25068d62..7b31370db70 100644
--- a/src/backend/storage/lmgr/proc.c
+++ b/src/backend/storage/lmgr/proc.c
@@ -95,7 +95,6 @@ static volatile sig_atomic_t got_deadlock_timeout;
 static void RemoveProcFromArray(int code, Datum arg);
 static void ProcKill(int code, Datum arg);
 static void AuxiliaryProcKill(int code, Datum arg);
-static DeadLockState CheckDeadLock(void);
 
 
 /*
@@ -1143,7 +1142,8 @@ AuxiliaryPidGetProc(int pid)
  * NOTES: The process queue is now a priority queue for locking.
  */
 ProcWaitStatus
-JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
+JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait,
+			  bool isUpgrade)
 {
 	LOCKMODE	lockmode = locallock->tag.mode;
 	LOCK	   *lock = locallock->lock;
@@ -1230,8 +1230,12 @@ JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
 			/* Must he wait for me? */
 			if (lockMethodTable->conflictTab[proc->waitLockMode] & myHeldLocks)
 			{
-				/* Must I wait for him ? */
-				if (lockMethodTable->conflictTab[lockmode] & proc->heldLocks)
+				/*
+				 * Must I wait for him? I don't want a deadlock during lock
+				 * upgrade - other processes should fail on it.
+				 */
+				if ((lockMethodTable->conflictTab[lockmode] & proc->heldLocks) &&
+					!isUpgrade)
 				{
 					/*
 					 * Yes, so we have a deadlock.  Easiest way to clean up
@@ -1461,8 +1465,22 @@ ProcSleep(LOCALLOCK *locallock)
 			(void) WaitLatch(MyLatch, WL_LATCH_SET | WL_EXIT_ON_PM_DEATH, 0,
 							 PG_WAIT_LOCK | locallock->tag.lock.locktag_type);
 			ResetLatch(MyLatch);
-			/* check for deadlocks first, as that's probably log-worthy */
-			if (got_deadlock_timeout)
+			/*
+			 * Check for deadlocks first, as that's probably log-worthy.
+			 *
+			 * Do not wait for the timeout if the lock is being upgraded since
+			 * the risk of deadlock is higher now. However, do not check for
+			 * deadlock if the lock is being upgraded by this process - other
+			 * processes should take care.
+			 *
+			 * TODO Possible optimization: if this is the only lock of the
+			 * backend and if it did not have any weaker lock on the table so
+			 * far, it should be safe to skip the deadlock check. However, to
+			 * evaluate the situation, we need to take fast-path locks into
+			 * account.
+			 */
+			if ((got_deadlock_timeout || lock->upgradedBy) &&
+				lock->upgradedBy != MyProc)
 			{
 				deadlock_state = CheckDeadLock();
 				got_deadlock_timeout = false;
@@ -1819,7 +1837,7 @@ ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock)
  * not, just return.  If we have a real deadlock, remove ourselves from the
  * lock's wait queue.
  */
-static DeadLockState
+DeadLockState
 CheckDeadLock(void)
 {
 	int			i;
diff --git a/src/include/storage/lmgr.h b/src/include/storage/lmgr.h
index 2a985ce5e15..eddbe1e8622 100644
--- a/src/include/storage/lmgr.h
+++ b/src/include/storage/lmgr.h
@@ -38,6 +38,7 @@ extern void RelationInitLockInfo(Relation relation);
 
 /* Lock a relation */
 extern void LockRelationOid(Oid relid, LOCKMODE lockmode);
+extern void LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode);
 extern void LockRelationId(LockRelId *relid, LOCKMODE lockmode);
 extern bool ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode);
 extern void UnlockRelationId(LockRelId *relid, LOCKMODE lockmode);
diff --git a/src/include/storage/lock.h b/src/include/storage/lock.h
index ee3cb1dc203..080b69eddfd 100644
--- a/src/include/storage/lock.h
+++ b/src/include/storage/lock.h
@@ -131,6 +131,7 @@ typedef const LockMethodData *LockMethod;
  * nRequested -- total requested locks of all types.
  * granted -- count of each lock type currently granted on the lock.
  * nGranted -- total granted locks of all types.
+ * upgrader -- process that intends to upgrade the lock
  *
  * Note: these counts count 1 for each backend.  Internally to a backend,
  * there may be multiple grabs on a particular lock, but this is not reflected
@@ -150,6 +151,7 @@ typedef struct LOCK
 	int			nRequested;		/* total of requested[] array */
 	int			granted[MAX_LOCKMODES]; /* counts of granted locks */
 	int			nGranted;		/* total of granted[] array */
+	PGPROC		*upgradedBy;	/* is lock being upgraded by this process? */
 } LOCK;
 
 #define LOCK_LOCKMETHOD(lock) ((LOCKMETHODID) (lock).tag.locktag_lockmethodid)
@@ -390,7 +392,8 @@ extern LockAcquireResult LockAcquireExtended(const LOCKTAG *locktag,
 											 bool dontWait,
 											 bool reportMemoryError,
 											 LOCALLOCK **locallockp,
-											 bool logLockFailure);
+											 bool logLockFailure,
+											 bool isUpgrade);
 extern void AbortStrongLockAcquire(void);
 extern void MarkLockClear(LOCALLOCK *locallock);
 extern bool LockRelease(const LOCKTAG *locktag,
diff --git a/src/include/storage/proc.h b/src/include/storage/proc.h
index 3e1d1fad5f9..c4bc3a24044 100644
--- a/src/include/storage/proc.h
+++ b/src/include/storage/proc.h
@@ -563,10 +563,12 @@ extern bool HaveNFreeProcs(int n, int *nfree);
 extern void ProcReleaseLocks(bool isCommit);
 
 extern ProcWaitStatus JoinWaitQueue(LOCALLOCK *locallock,
-									LockMethod lockMethodTable, bool dontWait);
+									LockMethod lockMethodTable, bool dontWait,
+									bool isUpgrade);
 extern ProcWaitStatus ProcSleep(LOCALLOCK *locallock);
 extern void ProcWakeup(PGPROC *proc, ProcWaitStatus waitStatus);
 extern void ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock);
+extern DeadLockState CheckDeadLock(void);
 extern void CheckDeadLockAlert(void);
 extern void LockErrorCleanup(void);
 extern void GetLockHoldersAndWaiters(LOCALLOCK *locallock,
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 966+ messages in thread

* [PATCH] Lock upgrade without deadlocks.
@ 2026-04-17 13:13  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 966+ messages in thread

From: Antonin Houska @ 2026-04-17 13:13 UTC (permalink / raw)

For REPACK (CONCURRENTLY), it's essential to upgrade ShareUpdateExclusive lock
to AccessExclusiveLock for the final stage of table processing. However, we
cannot release the earlier before requesting the latter, because if another
process ran DDL on the table in between, REPACK would have to cancel its
transaction, and waste possibly a lot of work. Without releasing
ShareUpdateExclusive lock, the upgrade can cause deadlock, and it's not really
deterministic whether REPACK or the other process will get canceled.

This patch adds a new field to the LOCK structure which tells which process is
trying to upgrade the lock - this field is set right before the upgrade. All
processes waiting in the lock queue are then woken up to check the flag.

When another process tries to get the lock after that, it checks this field,
and if it's set, it checks for deadlocks even if deadlock timeout hasn't
expired yet. If the process is already in the lock's queue and sleeping, the
lock upgrading process wakes it up so it checks the flag immediately.

At the time the upgrading process starts to sleep, all the other process
should be aware that they need to check for deadlock during their lock
acquisitions, so the upgrading process does not have to check for deadlocks
when it gets woken up. Thus it should never receive deadlock error report.
---
 src/backend/commands/repack.c   |   4 +-
 src/backend/storage/lmgr/lmgr.c |  35 ++++++++---
 src/backend/storage/lmgr/lock.c | 107 +++++++++++++++++++++++++++++++-
 src/backend/storage/lmgr/proc.c |  32 +++++++---
 src/include/storage/lmgr.h      |   1 +
 src/include/storage/lock.h      |   5 +-
 src/include/storage/proc.h      |   4 +-
 7 files changed, 166 insertions(+), 22 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index 58e3867246f..310b2a65099 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -3055,8 +3055,10 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Acquire AccessExclusiveLock on the table, its TOAST relation (if there
 	 * is one), all its indexes, so that we can swap the files.
+	 *
+	 * TODO The same for indexes and TOAST?
 	 */
-	LockRelationOid(old_table_oid, AccessExclusiveLock);
+	LockRelationOidUpgrade(old_table_oid, AccessExclusiveLock);
 
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
diff --git a/src/backend/storage/lmgr/lmgr.c b/src/backend/storage/lmgr/lmgr.c
index 2ccf7237fee..0b103485d54 100644
--- a/src/backend/storage/lmgr/lmgr.c
+++ b/src/backend/storage/lmgr/lmgr.c
@@ -58,6 +58,8 @@ typedef struct XactLockTableWaitInfo
 	const ItemPointerData *ctid;
 } XactLockTableWaitInfo;
 
+static void LockRelationOidCommon(Oid relid, LOCKMODE lockmode,
+								  bool isUpgrade);
 static void XactLockTableWaitErrorCb(void *arg);
 
 /*
@@ -105,6 +107,21 @@ SetLocktagRelationOid(LOCKTAG *tag, Oid relid)
  */
 void
 LockRelationOid(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, false);
+}
+
+/*
+ * Like above, but upgrade an existing lock.
+ */
+void
+LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, true);
+}
+
+static void
+LockRelationOidCommon(Oid relid, LOCKMODE lockmode, bool isUpgrade)
 {
 	LOCKTAG		tag;
 	LOCALLOCK  *locallock;
@@ -113,7 +130,7 @@ LockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, isUpgrade);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages, so that we
@@ -157,7 +174,7 @@ ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -191,7 +208,7 @@ LockRelationId(LockRelId *relid, LOCKMODE lockmode)
 	SET_LOCKTAG_RELATION(tag, relid->dbId, relid->relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -254,7 +271,7 @@ LockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -286,7 +303,7 @@ ConditionalLockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -591,7 +608,7 @@ ConditionalLockTuple(Relation relation, const ItemPointerData *tid, LOCKMODE loc
 					  ItemPointerGetOffsetNumber(tid));
 
 	return (LockAcquireExtended(&tag, lockmode, false, true, true, NULL,
-								logLockFailure) != LOCKACQUIRE_NOT_AVAIL);
+								logLockFailure, false) != LOCKACQUIRE_NOT_AVAIL);
 }
 
 /*
@@ -749,7 +766,7 @@ ConditionalXactLockTableWait(TransactionId xid, bool logLockFailure)
 		SET_LOCKTAG_TRANSACTION(tag, xid);
 
 		if (LockAcquireExtended(&tag, ShareLock, false, true, true, NULL,
-								logLockFailure)
+								logLockFailure, false)
 			== LOCKACQUIRE_NOT_AVAIL)
 			return false;
 
@@ -1043,7 +1060,7 @@ ConditionalLockDatabaseObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -1123,7 +1140,7 @@ ConditionalLockSharedObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
diff --git a/src/backend/storage/lmgr/lock.c b/src/backend/storage/lmgr/lock.c
index c221fe96889..bcbbf36c091 100644
--- a/src/backend/storage/lmgr/lock.c
+++ b/src/backend/storage/lmgr/lock.c
@@ -810,7 +810,7 @@ LockAcquire(const LOCKTAG *locktag,
 			bool dontWait)
 {
 	return LockAcquireExtended(locktag, lockmode, sessionLock, dontWait,
-							   true, NULL, false);
+							   true, NULL, false, false);
 }
 
 /*
@@ -829,6 +829,9 @@ LockAcquire(const LOCKTAG *locktag,
  *
  * logLockFailure indicates whether to log details when a lock acquisition
  * fails with dontWait = true.
+ *
+ * isUpgrade should be true if the backend already holds this lock in a lower
+ * mode.
  */
 LockAcquireResult
 LockAcquireExtended(const LOCKTAG *locktag,
@@ -837,7 +840,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 					bool dontWait,
 					bool reportMemoryError,
 					LOCALLOCK **locallockp,
-					bool logLockFailure)
+					bool logLockFailure,
+					bool isUpgrade)
 {
 	LOCKMETHODID lockmethodid = locktag->locktag_lockmethodid;
 	LockMethod	lockMethodTable;
@@ -1119,7 +1123,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		 * case, because JoinWaitQueue() may discover that we can acquire the
 		 * lock immediately after all.
 		 */
-		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait);
+		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait,
+								   isUpgrade);
 	}
 
 	if (waitResult == PROC_WAIT_STATUS_ERROR)
@@ -1225,8 +1230,85 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		Assert(!dontWait);
 		PROCLOCK_PRINT("LockAcquire: sleeping on lock", proclock);
 		LOCK_PRINT("LockAcquire: sleeping on lock", lock, lockmode);
+
+		/*
+		 * Lock upgrade can introduce deadlock, therefore enforce special
+		 * behavior of other processes that deal with this lock. In
+		 * particular, any waiter that sees upgradedBy set is expected to
+		 * perform deadlock check as soon as it's woken up. (Since we are
+		 * already in the queue, the deadlock detector has all the information
+		 * it needs.)
+		 */
+		if (isUpgrade)
+		{
+			dlist_iter	iter;
+
+			/*
+			 * There should not be multiple upgrades at the same time. XXX
+			 * ERROR ?
+			 */
+			Assert(lock->upgradedBy == NULL);
+
+			lock->upgradedBy = MyProc;
+
+			/*
+			 * Now, before I sleep myself, wake up all the existing waiters
+			 * (except for me), so they check for deadlock.
+			 */
+			dclist_foreach(iter, &lock->waitProcs)
+			{
+				PGPROC	   *proc = dlist_container(PGPROC, waitLink,
+												   iter.cur);
+
+				if (proc != MyProc)
+					SetLatch(&proc->procLatch);
+			}
+		}
 		LWLockRelease(partitionLock);
 
+		if (!isUpgrade)
+		{
+			HASH_SEQ_STATUS status;
+			LOCALLOCK	*llock;
+			bool	check_deadlock = false;
+
+			/*
+			 * Likewise, new waiters should check the isUpgrade flag and
+			 * engage deadlock detector if needed. The lock being acquired
+			 * should already be in the hash.
+			 *
+			 * Note that, to make deadlock detection happen as soon as
+			 * possible (i.e. regardless deadlock timeout), we check the other
+			 * locks too. XXX There might be ways to get into the deadlock
+			 * indirectly, but that needs more analysis. As long as the
+			 * upgrading process skips deadlock detection altogether, the
+			 * worst consequence of missing some lock wait cycle here is that
+			 * the other process will be kicked-off no sooner than after its
+			 * deadlock timeout has elapsed. (Which in turn means that the
+			 * lock upgrade will take longer than expected.)
+			 */
+			hash_seq_init(&status, LockMethodLocalHash);
+			while ((llock = (LOCALLOCK *) hash_seq_search(&status)) != NULL)
+			{
+				LOCK	*mylock = llock->lock;
+
+				if (mylock && mylock->upgradedBy)
+				{
+					check_deadlock = true;
+					break;
+				}
+			}
+			/* Check for deadlock if needed. */
+			if (check_deadlock)
+			{
+				DeadLockState	deadlock_state;
+
+				deadlock_state = CheckDeadLock();
+				if (deadlock_state == DS_HARD_DEADLOCK)
+					DeadLockReport();
+			}
+		}
+
 		waitResult = WaitOnLock(locallock, owner);
 
 		/*
@@ -1245,6 +1327,18 @@ LockAcquireExtended(const LOCKTAG *locktag,
 			DeadLockReport();
 			/* DeadLockReport() will not return */
 		}
+
+		/*
+		 * If finishing the lock upgrade, we should not get into a deadlock
+		 * anymore, so let others know that they do not have to care either.
+		 */
+		if (isUpgrade)
+		{
+			LWLockAcquire(partitionLock, LW_EXCLUSIVE);
+			Assert(lock->upgradedBy != NULL);
+			lock->upgradedBy = NULL;
+			LWLockRelease(partitionLock);
+		}
 	}
 	else
 		LWLockRelease(partitionLock);
@@ -1322,6 +1416,7 @@ SetupLockInTable(LockMethod lockMethodTable, PGPROC *proc,
 		lock->nGranted = 0;
 		MemSet(lock->requested, 0, sizeof(int) * MAX_LOCKMODES);
 		MemSet(lock->granted, 0, sizeof(int) * MAX_LOCKMODES);
+		lock->upgradedBy = NULL;
 		LOCK_PRINT("LockAcquire: new", lock, lockmode);
 	}
 	else
@@ -1768,6 +1863,12 @@ CleanUpLock(LOCK *lock, PROCLOCK *proclock,
 			elog(PANIC, "proclock table corrupted");
 	}
 
+	/*
+	 * Was this backend upgrading the lock?
+	 */
+	if (lock->upgradedBy == MyProc)
+		lock->upgradedBy = NULL;
+
 	if (lock->nRequested == 0)
 	{
 		/*
diff --git a/src/backend/storage/lmgr/proc.c b/src/backend/storage/lmgr/proc.c
index 1ac25068d62..7b31370db70 100644
--- a/src/backend/storage/lmgr/proc.c
+++ b/src/backend/storage/lmgr/proc.c
@@ -95,7 +95,6 @@ static volatile sig_atomic_t got_deadlock_timeout;
 static void RemoveProcFromArray(int code, Datum arg);
 static void ProcKill(int code, Datum arg);
 static void AuxiliaryProcKill(int code, Datum arg);
-static DeadLockState CheckDeadLock(void);
 
 
 /*
@@ -1143,7 +1142,8 @@ AuxiliaryPidGetProc(int pid)
  * NOTES: The process queue is now a priority queue for locking.
  */
 ProcWaitStatus
-JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
+JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait,
+			  bool isUpgrade)
 {
 	LOCKMODE	lockmode = locallock->tag.mode;
 	LOCK	   *lock = locallock->lock;
@@ -1230,8 +1230,12 @@ JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
 			/* Must he wait for me? */
 			if (lockMethodTable->conflictTab[proc->waitLockMode] & myHeldLocks)
 			{
-				/* Must I wait for him ? */
-				if (lockMethodTable->conflictTab[lockmode] & proc->heldLocks)
+				/*
+				 * Must I wait for him? I don't want a deadlock during lock
+				 * upgrade - other processes should fail on it.
+				 */
+				if ((lockMethodTable->conflictTab[lockmode] & proc->heldLocks) &&
+					!isUpgrade)
 				{
 					/*
 					 * Yes, so we have a deadlock.  Easiest way to clean up
@@ -1461,8 +1465,22 @@ ProcSleep(LOCALLOCK *locallock)
 			(void) WaitLatch(MyLatch, WL_LATCH_SET | WL_EXIT_ON_PM_DEATH, 0,
 							 PG_WAIT_LOCK | locallock->tag.lock.locktag_type);
 			ResetLatch(MyLatch);
-			/* check for deadlocks first, as that's probably log-worthy */
-			if (got_deadlock_timeout)
+			/*
+			 * Check for deadlocks first, as that's probably log-worthy.
+			 *
+			 * Do not wait for the timeout if the lock is being upgraded since
+			 * the risk of deadlock is higher now. However, do not check for
+			 * deadlock if the lock is being upgraded by this process - other
+			 * processes should take care.
+			 *
+			 * TODO Possible optimization: if this is the only lock of the
+			 * backend and if it did not have any weaker lock on the table so
+			 * far, it should be safe to skip the deadlock check. However, to
+			 * evaluate the situation, we need to take fast-path locks into
+			 * account.
+			 */
+			if ((got_deadlock_timeout || lock->upgradedBy) &&
+				lock->upgradedBy != MyProc)
 			{
 				deadlock_state = CheckDeadLock();
 				got_deadlock_timeout = false;
@@ -1819,7 +1837,7 @@ ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock)
  * not, just return.  If we have a real deadlock, remove ourselves from the
  * lock's wait queue.
  */
-static DeadLockState
+DeadLockState
 CheckDeadLock(void)
 {
 	int			i;
diff --git a/src/include/storage/lmgr.h b/src/include/storage/lmgr.h
index 2a985ce5e15..eddbe1e8622 100644
--- a/src/include/storage/lmgr.h
+++ b/src/include/storage/lmgr.h
@@ -38,6 +38,7 @@ extern void RelationInitLockInfo(Relation relation);
 
 /* Lock a relation */
 extern void LockRelationOid(Oid relid, LOCKMODE lockmode);
+extern void LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode);
 extern void LockRelationId(LockRelId *relid, LOCKMODE lockmode);
 extern bool ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode);
 extern void UnlockRelationId(LockRelId *relid, LOCKMODE lockmode);
diff --git a/src/include/storage/lock.h b/src/include/storage/lock.h
index ee3cb1dc203..080b69eddfd 100644
--- a/src/include/storage/lock.h
+++ b/src/include/storage/lock.h
@@ -131,6 +131,7 @@ typedef const LockMethodData *LockMethod;
  * nRequested -- total requested locks of all types.
  * granted -- count of each lock type currently granted on the lock.
  * nGranted -- total granted locks of all types.
+ * upgrader -- process that intends to upgrade the lock
  *
  * Note: these counts count 1 for each backend.  Internally to a backend,
  * there may be multiple grabs on a particular lock, but this is not reflected
@@ -150,6 +151,7 @@ typedef struct LOCK
 	int			nRequested;		/* total of requested[] array */
 	int			granted[MAX_LOCKMODES]; /* counts of granted locks */
 	int			nGranted;		/* total of granted[] array */
+	PGPROC		*upgradedBy;	/* is lock being upgraded by this process? */
 } LOCK;
 
 #define LOCK_LOCKMETHOD(lock) ((LOCKMETHODID) (lock).tag.locktag_lockmethodid)
@@ -390,7 +392,8 @@ extern LockAcquireResult LockAcquireExtended(const LOCKTAG *locktag,
 											 bool dontWait,
 											 bool reportMemoryError,
 											 LOCALLOCK **locallockp,
-											 bool logLockFailure);
+											 bool logLockFailure,
+											 bool isUpgrade);
 extern void AbortStrongLockAcquire(void);
 extern void MarkLockClear(LOCALLOCK *locallock);
 extern bool LockRelease(const LOCKTAG *locktag,
diff --git a/src/include/storage/proc.h b/src/include/storage/proc.h
index 3e1d1fad5f9..c4bc3a24044 100644
--- a/src/include/storage/proc.h
+++ b/src/include/storage/proc.h
@@ -563,10 +563,12 @@ extern bool HaveNFreeProcs(int n, int *nfree);
 extern void ProcReleaseLocks(bool isCommit);
 
 extern ProcWaitStatus JoinWaitQueue(LOCALLOCK *locallock,
-									LockMethod lockMethodTable, bool dontWait);
+									LockMethod lockMethodTable, bool dontWait,
+									bool isUpgrade);
 extern ProcWaitStatus ProcSleep(LOCALLOCK *locallock);
 extern void ProcWakeup(PGPROC *proc, ProcWaitStatus waitStatus);
 extern void ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock);
+extern DeadLockState CheckDeadLock(void);
 extern void CheckDeadLockAlert(void);
 extern void LockErrorCleanup(void);
 extern void GetLockHoldersAndWaiters(LOCALLOCK *locallock,
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 966+ messages in thread

* [PATCH] Lock upgrade without deadlocks.
@ 2026-04-17 13:13  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 966+ messages in thread

From: Antonin Houska @ 2026-04-17 13:13 UTC (permalink / raw)

For REPACK (CONCURRENTLY), it's essential to upgrade ShareUpdateExclusive lock
to AccessExclusiveLock for the final stage of table processing. However, we
cannot release the earlier before requesting the latter, because if another
process ran DDL on the table in between, REPACK would have to cancel its
transaction, and waste possibly a lot of work. Without releasing
ShareUpdateExclusive lock, the upgrade can cause deadlock, and it's not really
deterministic whether REPACK or the other process will get canceled.

This patch adds a new field to the LOCK structure which tells which process is
trying to upgrade the lock - this field is set right before the upgrade. All
processes waiting in the lock queue are then woken up to check the flag.

When another process tries to get the lock after that, it checks this field,
and if it's set, it checks for deadlocks even if deadlock timeout hasn't
expired yet. If the process is already in the lock's queue and sleeping, the
lock upgrading process wakes it up so it checks the flag immediately.

At the time the upgrading process starts to sleep, all the other process
should be aware that they need to check for deadlock during their lock
acquisitions, so the upgrading process does not have to check for deadlocks
when it gets woken up. Thus it should never receive deadlock error report.
---
 src/backend/commands/repack.c   |   4 +-
 src/backend/storage/lmgr/lmgr.c |  35 ++++++++---
 src/backend/storage/lmgr/lock.c | 107 +++++++++++++++++++++++++++++++-
 src/backend/storage/lmgr/proc.c |  32 +++++++---
 src/include/storage/lmgr.h      |   1 +
 src/include/storage/lock.h      |   5 +-
 src/include/storage/proc.h      |   4 +-
 7 files changed, 166 insertions(+), 22 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index 58e3867246f..310b2a65099 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -3055,8 +3055,10 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Acquire AccessExclusiveLock on the table, its TOAST relation (if there
 	 * is one), all its indexes, so that we can swap the files.
+	 *
+	 * TODO The same for indexes and TOAST?
 	 */
-	LockRelationOid(old_table_oid, AccessExclusiveLock);
+	LockRelationOidUpgrade(old_table_oid, AccessExclusiveLock);
 
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
diff --git a/src/backend/storage/lmgr/lmgr.c b/src/backend/storage/lmgr/lmgr.c
index 2ccf7237fee..0b103485d54 100644
--- a/src/backend/storage/lmgr/lmgr.c
+++ b/src/backend/storage/lmgr/lmgr.c
@@ -58,6 +58,8 @@ typedef struct XactLockTableWaitInfo
 	const ItemPointerData *ctid;
 } XactLockTableWaitInfo;
 
+static void LockRelationOidCommon(Oid relid, LOCKMODE lockmode,
+								  bool isUpgrade);
 static void XactLockTableWaitErrorCb(void *arg);
 
 /*
@@ -105,6 +107,21 @@ SetLocktagRelationOid(LOCKTAG *tag, Oid relid)
  */
 void
 LockRelationOid(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, false);
+}
+
+/*
+ * Like above, but upgrade an existing lock.
+ */
+void
+LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, true);
+}
+
+static void
+LockRelationOidCommon(Oid relid, LOCKMODE lockmode, bool isUpgrade)
 {
 	LOCKTAG		tag;
 	LOCALLOCK  *locallock;
@@ -113,7 +130,7 @@ LockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, isUpgrade);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages, so that we
@@ -157,7 +174,7 @@ ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -191,7 +208,7 @@ LockRelationId(LockRelId *relid, LOCKMODE lockmode)
 	SET_LOCKTAG_RELATION(tag, relid->dbId, relid->relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -254,7 +271,7 @@ LockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -286,7 +303,7 @@ ConditionalLockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -591,7 +608,7 @@ ConditionalLockTuple(Relation relation, const ItemPointerData *tid, LOCKMODE loc
 					  ItemPointerGetOffsetNumber(tid));
 
 	return (LockAcquireExtended(&tag, lockmode, false, true, true, NULL,
-								logLockFailure) != LOCKACQUIRE_NOT_AVAIL);
+								logLockFailure, false) != LOCKACQUIRE_NOT_AVAIL);
 }
 
 /*
@@ -749,7 +766,7 @@ ConditionalXactLockTableWait(TransactionId xid, bool logLockFailure)
 		SET_LOCKTAG_TRANSACTION(tag, xid);
 
 		if (LockAcquireExtended(&tag, ShareLock, false, true, true, NULL,
-								logLockFailure)
+								logLockFailure, false)
 			== LOCKACQUIRE_NOT_AVAIL)
 			return false;
 
@@ -1043,7 +1060,7 @@ ConditionalLockDatabaseObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -1123,7 +1140,7 @@ ConditionalLockSharedObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
diff --git a/src/backend/storage/lmgr/lock.c b/src/backend/storage/lmgr/lock.c
index c221fe96889..bcbbf36c091 100644
--- a/src/backend/storage/lmgr/lock.c
+++ b/src/backend/storage/lmgr/lock.c
@@ -810,7 +810,7 @@ LockAcquire(const LOCKTAG *locktag,
 			bool dontWait)
 {
 	return LockAcquireExtended(locktag, lockmode, sessionLock, dontWait,
-							   true, NULL, false);
+							   true, NULL, false, false);
 }
 
 /*
@@ -829,6 +829,9 @@ LockAcquire(const LOCKTAG *locktag,
  *
  * logLockFailure indicates whether to log details when a lock acquisition
  * fails with dontWait = true.
+ *
+ * isUpgrade should be true if the backend already holds this lock in a lower
+ * mode.
  */
 LockAcquireResult
 LockAcquireExtended(const LOCKTAG *locktag,
@@ -837,7 +840,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 					bool dontWait,
 					bool reportMemoryError,
 					LOCALLOCK **locallockp,
-					bool logLockFailure)
+					bool logLockFailure,
+					bool isUpgrade)
 {
 	LOCKMETHODID lockmethodid = locktag->locktag_lockmethodid;
 	LockMethod	lockMethodTable;
@@ -1119,7 +1123,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		 * case, because JoinWaitQueue() may discover that we can acquire the
 		 * lock immediately after all.
 		 */
-		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait);
+		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait,
+								   isUpgrade);
 	}
 
 	if (waitResult == PROC_WAIT_STATUS_ERROR)
@@ -1225,8 +1230,85 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		Assert(!dontWait);
 		PROCLOCK_PRINT("LockAcquire: sleeping on lock", proclock);
 		LOCK_PRINT("LockAcquire: sleeping on lock", lock, lockmode);
+
+		/*
+		 * Lock upgrade can introduce deadlock, therefore enforce special
+		 * behavior of other processes that deal with this lock. In
+		 * particular, any waiter that sees upgradedBy set is expected to
+		 * perform deadlock check as soon as it's woken up. (Since we are
+		 * already in the queue, the deadlock detector has all the information
+		 * it needs.)
+		 */
+		if (isUpgrade)
+		{
+			dlist_iter	iter;
+
+			/*
+			 * There should not be multiple upgrades at the same time. XXX
+			 * ERROR ?
+			 */
+			Assert(lock->upgradedBy == NULL);
+
+			lock->upgradedBy = MyProc;
+
+			/*
+			 * Now, before I sleep myself, wake up all the existing waiters
+			 * (except for me), so they check for deadlock.
+			 */
+			dclist_foreach(iter, &lock->waitProcs)
+			{
+				PGPROC	   *proc = dlist_container(PGPROC, waitLink,
+												   iter.cur);
+
+				if (proc != MyProc)
+					SetLatch(&proc->procLatch);
+			}
+		}
 		LWLockRelease(partitionLock);
 
+		if (!isUpgrade)
+		{
+			HASH_SEQ_STATUS status;
+			LOCALLOCK	*llock;
+			bool	check_deadlock = false;
+
+			/*
+			 * Likewise, new waiters should check the isUpgrade flag and
+			 * engage deadlock detector if needed. The lock being acquired
+			 * should already be in the hash.
+			 *
+			 * Note that, to make deadlock detection happen as soon as
+			 * possible (i.e. regardless deadlock timeout), we check the other
+			 * locks too. XXX There might be ways to get into the deadlock
+			 * indirectly, but that needs more analysis. As long as the
+			 * upgrading process skips deadlock detection altogether, the
+			 * worst consequence of missing some lock wait cycle here is that
+			 * the other process will be kicked-off no sooner than after its
+			 * deadlock timeout has elapsed. (Which in turn means that the
+			 * lock upgrade will take longer than expected.)
+			 */
+			hash_seq_init(&status, LockMethodLocalHash);
+			while ((llock = (LOCALLOCK *) hash_seq_search(&status)) != NULL)
+			{
+				LOCK	*mylock = llock->lock;
+
+				if (mylock && mylock->upgradedBy)
+				{
+					check_deadlock = true;
+					break;
+				}
+			}
+			/* Check for deadlock if needed. */
+			if (check_deadlock)
+			{
+				DeadLockState	deadlock_state;
+
+				deadlock_state = CheckDeadLock();
+				if (deadlock_state == DS_HARD_DEADLOCK)
+					DeadLockReport();
+			}
+		}
+
 		waitResult = WaitOnLock(locallock, owner);
 
 		/*
@@ -1245,6 +1327,18 @@ LockAcquireExtended(const LOCKTAG *locktag,
 			DeadLockReport();
 			/* DeadLockReport() will not return */
 		}
+
+		/*
+		 * If finishing the lock upgrade, we should not get into a deadlock
+		 * anymore, so let others know that they do not have to care either.
+		 */
+		if (isUpgrade)
+		{
+			LWLockAcquire(partitionLock, LW_EXCLUSIVE);
+			Assert(lock->upgradedBy != NULL);
+			lock->upgradedBy = NULL;
+			LWLockRelease(partitionLock);
+		}
 	}
 	else
 		LWLockRelease(partitionLock);
@@ -1322,6 +1416,7 @@ SetupLockInTable(LockMethod lockMethodTable, PGPROC *proc,
 		lock->nGranted = 0;
 		MemSet(lock->requested, 0, sizeof(int) * MAX_LOCKMODES);
 		MemSet(lock->granted, 0, sizeof(int) * MAX_LOCKMODES);
+		lock->upgradedBy = NULL;
 		LOCK_PRINT("LockAcquire: new", lock, lockmode);
 	}
 	else
@@ -1768,6 +1863,12 @@ CleanUpLock(LOCK *lock, PROCLOCK *proclock,
 			elog(PANIC, "proclock table corrupted");
 	}
 
+	/*
+	 * Was this backend upgrading the lock?
+	 */
+	if (lock->upgradedBy == MyProc)
+		lock->upgradedBy = NULL;
+
 	if (lock->nRequested == 0)
 	{
 		/*
diff --git a/src/backend/storage/lmgr/proc.c b/src/backend/storage/lmgr/proc.c
index 1ac25068d62..7b31370db70 100644
--- a/src/backend/storage/lmgr/proc.c
+++ b/src/backend/storage/lmgr/proc.c
@@ -95,7 +95,6 @@ static volatile sig_atomic_t got_deadlock_timeout;
 static void RemoveProcFromArray(int code, Datum arg);
 static void ProcKill(int code, Datum arg);
 static void AuxiliaryProcKill(int code, Datum arg);
-static DeadLockState CheckDeadLock(void);
 
 
 /*
@@ -1143,7 +1142,8 @@ AuxiliaryPidGetProc(int pid)
  * NOTES: The process queue is now a priority queue for locking.
  */
 ProcWaitStatus
-JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
+JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait,
+			  bool isUpgrade)
 {
 	LOCKMODE	lockmode = locallock->tag.mode;
 	LOCK	   *lock = locallock->lock;
@@ -1230,8 +1230,12 @@ JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
 			/* Must he wait for me? */
 			if (lockMethodTable->conflictTab[proc->waitLockMode] & myHeldLocks)
 			{
-				/* Must I wait for him ? */
-				if (lockMethodTable->conflictTab[lockmode] & proc->heldLocks)
+				/*
+				 * Must I wait for him? I don't want a deadlock during lock
+				 * upgrade - other processes should fail on it.
+				 */
+				if ((lockMethodTable->conflictTab[lockmode] & proc->heldLocks) &&
+					!isUpgrade)
 				{
 					/*
 					 * Yes, so we have a deadlock.  Easiest way to clean up
@@ -1461,8 +1465,22 @@ ProcSleep(LOCALLOCK *locallock)
 			(void) WaitLatch(MyLatch, WL_LATCH_SET | WL_EXIT_ON_PM_DEATH, 0,
 							 PG_WAIT_LOCK | locallock->tag.lock.locktag_type);
 			ResetLatch(MyLatch);
-			/* check for deadlocks first, as that's probably log-worthy */
-			if (got_deadlock_timeout)
+			/*
+			 * Check for deadlocks first, as that's probably log-worthy.
+			 *
+			 * Do not wait for the timeout if the lock is being upgraded since
+			 * the risk of deadlock is higher now. However, do not check for
+			 * deadlock if the lock is being upgraded by this process - other
+			 * processes should take care.
+			 *
+			 * TODO Possible optimization: if this is the only lock of the
+			 * backend and if it did not have any weaker lock on the table so
+			 * far, it should be safe to skip the deadlock check. However, to
+			 * evaluate the situation, we need to take fast-path locks into
+			 * account.
+			 */
+			if ((got_deadlock_timeout || lock->upgradedBy) &&
+				lock->upgradedBy != MyProc)
 			{
 				deadlock_state = CheckDeadLock();
 				got_deadlock_timeout = false;
@@ -1819,7 +1837,7 @@ ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock)
  * not, just return.  If we have a real deadlock, remove ourselves from the
  * lock's wait queue.
  */
-static DeadLockState
+DeadLockState
 CheckDeadLock(void)
 {
 	int			i;
diff --git a/src/include/storage/lmgr.h b/src/include/storage/lmgr.h
index 2a985ce5e15..eddbe1e8622 100644
--- a/src/include/storage/lmgr.h
+++ b/src/include/storage/lmgr.h
@@ -38,6 +38,7 @@ extern void RelationInitLockInfo(Relation relation);
 
 /* Lock a relation */
 extern void LockRelationOid(Oid relid, LOCKMODE lockmode);
+extern void LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode);
 extern void LockRelationId(LockRelId *relid, LOCKMODE lockmode);
 extern bool ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode);
 extern void UnlockRelationId(LockRelId *relid, LOCKMODE lockmode);
diff --git a/src/include/storage/lock.h b/src/include/storage/lock.h
index ee3cb1dc203..080b69eddfd 100644
--- a/src/include/storage/lock.h
+++ b/src/include/storage/lock.h
@@ -131,6 +131,7 @@ typedef const LockMethodData *LockMethod;
  * nRequested -- total requested locks of all types.
  * granted -- count of each lock type currently granted on the lock.
  * nGranted -- total granted locks of all types.
+ * upgrader -- process that intends to upgrade the lock
  *
  * Note: these counts count 1 for each backend.  Internally to a backend,
  * there may be multiple grabs on a particular lock, but this is not reflected
@@ -150,6 +151,7 @@ typedef struct LOCK
 	int			nRequested;		/* total of requested[] array */
 	int			granted[MAX_LOCKMODES]; /* counts of granted locks */
 	int			nGranted;		/* total of granted[] array */
+	PGPROC		*upgradedBy;	/* is lock being upgraded by this process? */
 } LOCK;
 
 #define LOCK_LOCKMETHOD(lock) ((LOCKMETHODID) (lock).tag.locktag_lockmethodid)
@@ -390,7 +392,8 @@ extern LockAcquireResult LockAcquireExtended(const LOCKTAG *locktag,
 											 bool dontWait,
 											 bool reportMemoryError,
 											 LOCALLOCK **locallockp,
-											 bool logLockFailure);
+											 bool logLockFailure,
+											 bool isUpgrade);
 extern void AbortStrongLockAcquire(void);
 extern void MarkLockClear(LOCALLOCK *locallock);
 extern bool LockRelease(const LOCKTAG *locktag,
diff --git a/src/include/storage/proc.h b/src/include/storage/proc.h
index 3e1d1fad5f9..c4bc3a24044 100644
--- a/src/include/storage/proc.h
+++ b/src/include/storage/proc.h
@@ -563,10 +563,12 @@ extern bool HaveNFreeProcs(int n, int *nfree);
 extern void ProcReleaseLocks(bool isCommit);
 
 extern ProcWaitStatus JoinWaitQueue(LOCALLOCK *locallock,
-									LockMethod lockMethodTable, bool dontWait);
+									LockMethod lockMethodTable, bool dontWait,
+									bool isUpgrade);
 extern ProcWaitStatus ProcSleep(LOCALLOCK *locallock);
 extern void ProcWakeup(PGPROC *proc, ProcWaitStatus waitStatus);
 extern void ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock);
+extern DeadLockState CheckDeadLock(void);
 extern void CheckDeadLockAlert(void);
 extern void LockErrorCleanup(void);
 extern void GetLockHoldersAndWaiters(LOCALLOCK *locallock,
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 966+ messages in thread

* [PATCH] Lock upgrade without deadlocks.
@ 2026-04-17 13:13  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 966+ messages in thread

From: Antonin Houska @ 2026-04-17 13:13 UTC (permalink / raw)

For REPACK (CONCURRENTLY), it's essential to upgrade ShareUpdateExclusive lock
to AccessExclusiveLock for the final stage of table processing. However, we
cannot release the earlier before requesting the latter, because if another
process ran DDL on the table in between, REPACK would have to cancel its
transaction, and waste possibly a lot of work. Without releasing
ShareUpdateExclusive lock, the upgrade can cause deadlock, and it's not really
deterministic whether REPACK or the other process will get canceled.

This patch adds a new field to the LOCK structure which tells which process is
trying to upgrade the lock - this field is set right before the upgrade. All
processes waiting in the lock queue are then woken up to check the flag.

When another process tries to get the lock after that, it checks this field,
and if it's set, it checks for deadlocks even if deadlock timeout hasn't
expired yet. If the process is already in the lock's queue and sleeping, the
lock upgrading process wakes it up so it checks the flag immediately.

At the time the upgrading process starts to sleep, all the other process
should be aware that they need to check for deadlock during their lock
acquisitions, so the upgrading process does not have to check for deadlocks
when it gets woken up. Thus it should never receive deadlock error report.
---
 src/backend/commands/repack.c   |   4 +-
 src/backend/storage/lmgr/lmgr.c |  35 ++++++++---
 src/backend/storage/lmgr/lock.c | 107 +++++++++++++++++++++++++++++++-
 src/backend/storage/lmgr/proc.c |  32 +++++++---
 src/include/storage/lmgr.h      |   1 +
 src/include/storage/lock.h      |   5 +-
 src/include/storage/proc.h      |   4 +-
 7 files changed, 166 insertions(+), 22 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index 58e3867246f..310b2a65099 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -3055,8 +3055,10 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Acquire AccessExclusiveLock on the table, its TOAST relation (if there
 	 * is one), all its indexes, so that we can swap the files.
+	 *
+	 * TODO The same for indexes and TOAST?
 	 */
-	LockRelationOid(old_table_oid, AccessExclusiveLock);
+	LockRelationOidUpgrade(old_table_oid, AccessExclusiveLock);
 
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
diff --git a/src/backend/storage/lmgr/lmgr.c b/src/backend/storage/lmgr/lmgr.c
index 2ccf7237fee..0b103485d54 100644
--- a/src/backend/storage/lmgr/lmgr.c
+++ b/src/backend/storage/lmgr/lmgr.c
@@ -58,6 +58,8 @@ typedef struct XactLockTableWaitInfo
 	const ItemPointerData *ctid;
 } XactLockTableWaitInfo;
 
+static void LockRelationOidCommon(Oid relid, LOCKMODE lockmode,
+								  bool isUpgrade);
 static void XactLockTableWaitErrorCb(void *arg);
 
 /*
@@ -105,6 +107,21 @@ SetLocktagRelationOid(LOCKTAG *tag, Oid relid)
  */
 void
 LockRelationOid(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, false);
+}
+
+/*
+ * Like above, but upgrade an existing lock.
+ */
+void
+LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, true);
+}
+
+static void
+LockRelationOidCommon(Oid relid, LOCKMODE lockmode, bool isUpgrade)
 {
 	LOCKTAG		tag;
 	LOCALLOCK  *locallock;
@@ -113,7 +130,7 @@ LockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, isUpgrade);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages, so that we
@@ -157,7 +174,7 @@ ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -191,7 +208,7 @@ LockRelationId(LockRelId *relid, LOCKMODE lockmode)
 	SET_LOCKTAG_RELATION(tag, relid->dbId, relid->relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -254,7 +271,7 @@ LockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -286,7 +303,7 @@ ConditionalLockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -591,7 +608,7 @@ ConditionalLockTuple(Relation relation, const ItemPointerData *tid, LOCKMODE loc
 					  ItemPointerGetOffsetNumber(tid));
 
 	return (LockAcquireExtended(&tag, lockmode, false, true, true, NULL,
-								logLockFailure) != LOCKACQUIRE_NOT_AVAIL);
+								logLockFailure, false) != LOCKACQUIRE_NOT_AVAIL);
 }
 
 /*
@@ -749,7 +766,7 @@ ConditionalXactLockTableWait(TransactionId xid, bool logLockFailure)
 		SET_LOCKTAG_TRANSACTION(tag, xid);
 
 		if (LockAcquireExtended(&tag, ShareLock, false, true, true, NULL,
-								logLockFailure)
+								logLockFailure, false)
 			== LOCKACQUIRE_NOT_AVAIL)
 			return false;
 
@@ -1043,7 +1060,7 @@ ConditionalLockDatabaseObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -1123,7 +1140,7 @@ ConditionalLockSharedObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
diff --git a/src/backend/storage/lmgr/lock.c b/src/backend/storage/lmgr/lock.c
index c221fe96889..bcbbf36c091 100644
--- a/src/backend/storage/lmgr/lock.c
+++ b/src/backend/storage/lmgr/lock.c
@@ -810,7 +810,7 @@ LockAcquire(const LOCKTAG *locktag,
 			bool dontWait)
 {
 	return LockAcquireExtended(locktag, lockmode, sessionLock, dontWait,
-							   true, NULL, false);
+							   true, NULL, false, false);
 }
 
 /*
@@ -829,6 +829,9 @@ LockAcquire(const LOCKTAG *locktag,
  *
  * logLockFailure indicates whether to log details when a lock acquisition
  * fails with dontWait = true.
+ *
+ * isUpgrade should be true if the backend already holds this lock in a lower
+ * mode.
  */
 LockAcquireResult
 LockAcquireExtended(const LOCKTAG *locktag,
@@ -837,7 +840,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 					bool dontWait,
 					bool reportMemoryError,
 					LOCALLOCK **locallockp,
-					bool logLockFailure)
+					bool logLockFailure,
+					bool isUpgrade)
 {
 	LOCKMETHODID lockmethodid = locktag->locktag_lockmethodid;
 	LockMethod	lockMethodTable;
@@ -1119,7 +1123,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		 * case, because JoinWaitQueue() may discover that we can acquire the
 		 * lock immediately after all.
 		 */
-		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait);
+		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait,
+								   isUpgrade);
 	}
 
 	if (waitResult == PROC_WAIT_STATUS_ERROR)
@@ -1225,8 +1230,85 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		Assert(!dontWait);
 		PROCLOCK_PRINT("LockAcquire: sleeping on lock", proclock);
 		LOCK_PRINT("LockAcquire: sleeping on lock", lock, lockmode);
+
+		/*
+		 * Lock upgrade can introduce deadlock, therefore enforce special
+		 * behavior of other processes that deal with this lock. In
+		 * particular, any waiter that sees upgradedBy set is expected to
+		 * perform deadlock check as soon as it's woken up. (Since we are
+		 * already in the queue, the deadlock detector has all the information
+		 * it needs.)
+		 */
+		if (isUpgrade)
+		{
+			dlist_iter	iter;
+
+			/*
+			 * There should not be multiple upgrades at the same time. XXX
+			 * ERROR ?
+			 */
+			Assert(lock->upgradedBy == NULL);
+
+			lock->upgradedBy = MyProc;
+
+			/*
+			 * Now, before I sleep myself, wake up all the existing waiters
+			 * (except for me), so they check for deadlock.
+			 */
+			dclist_foreach(iter, &lock->waitProcs)
+			{
+				PGPROC	   *proc = dlist_container(PGPROC, waitLink,
+												   iter.cur);
+
+				if (proc != MyProc)
+					SetLatch(&proc->procLatch);
+			}
+		}
 		LWLockRelease(partitionLock);
 
+		if (!isUpgrade)
+		{
+			HASH_SEQ_STATUS status;
+			LOCALLOCK	*llock;
+			bool	check_deadlock = false;
+
+			/*
+			 * Likewise, new waiters should check the isUpgrade flag and
+			 * engage deadlock detector if needed. The lock being acquired
+			 * should already be in the hash.
+			 *
+			 * Note that, to make deadlock detection happen as soon as
+			 * possible (i.e. regardless deadlock timeout), we check the other
+			 * locks too. XXX There might be ways to get into the deadlock
+			 * indirectly, but that needs more analysis. As long as the
+			 * upgrading process skips deadlock detection altogether, the
+			 * worst consequence of missing some lock wait cycle here is that
+			 * the other process will be kicked-off no sooner than after its
+			 * deadlock timeout has elapsed. (Which in turn means that the
+			 * lock upgrade will take longer than expected.)
+			 */
+			hash_seq_init(&status, LockMethodLocalHash);
+			while ((llock = (LOCALLOCK *) hash_seq_search(&status)) != NULL)
+			{
+				LOCK	*mylock = llock->lock;
+
+				if (mylock && mylock->upgradedBy)
+				{
+					check_deadlock = true;
+					break;
+				}
+			}
+			/* Check for deadlock if needed. */
+			if (check_deadlock)
+			{
+				DeadLockState	deadlock_state;
+
+				deadlock_state = CheckDeadLock();
+				if (deadlock_state == DS_HARD_DEADLOCK)
+					DeadLockReport();
+			}
+		}
+
 		waitResult = WaitOnLock(locallock, owner);
 
 		/*
@@ -1245,6 +1327,18 @@ LockAcquireExtended(const LOCKTAG *locktag,
 			DeadLockReport();
 			/* DeadLockReport() will not return */
 		}
+
+		/*
+		 * If finishing the lock upgrade, we should not get into a deadlock
+		 * anymore, so let others know that they do not have to care either.
+		 */
+		if (isUpgrade)
+		{
+			LWLockAcquire(partitionLock, LW_EXCLUSIVE);
+			Assert(lock->upgradedBy != NULL);
+			lock->upgradedBy = NULL;
+			LWLockRelease(partitionLock);
+		}
 	}
 	else
 		LWLockRelease(partitionLock);
@@ -1322,6 +1416,7 @@ SetupLockInTable(LockMethod lockMethodTable, PGPROC *proc,
 		lock->nGranted = 0;
 		MemSet(lock->requested, 0, sizeof(int) * MAX_LOCKMODES);
 		MemSet(lock->granted, 0, sizeof(int) * MAX_LOCKMODES);
+		lock->upgradedBy = NULL;
 		LOCK_PRINT("LockAcquire: new", lock, lockmode);
 	}
 	else
@@ -1768,6 +1863,12 @@ CleanUpLock(LOCK *lock, PROCLOCK *proclock,
 			elog(PANIC, "proclock table corrupted");
 	}
 
+	/*
+	 * Was this backend upgrading the lock?
+	 */
+	if (lock->upgradedBy == MyProc)
+		lock->upgradedBy = NULL;
+
 	if (lock->nRequested == 0)
 	{
 		/*
diff --git a/src/backend/storage/lmgr/proc.c b/src/backend/storage/lmgr/proc.c
index 1ac25068d62..7b31370db70 100644
--- a/src/backend/storage/lmgr/proc.c
+++ b/src/backend/storage/lmgr/proc.c
@@ -95,7 +95,6 @@ static volatile sig_atomic_t got_deadlock_timeout;
 static void RemoveProcFromArray(int code, Datum arg);
 static void ProcKill(int code, Datum arg);
 static void AuxiliaryProcKill(int code, Datum arg);
-static DeadLockState CheckDeadLock(void);
 
 
 /*
@@ -1143,7 +1142,8 @@ AuxiliaryPidGetProc(int pid)
  * NOTES: The process queue is now a priority queue for locking.
  */
 ProcWaitStatus
-JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
+JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait,
+			  bool isUpgrade)
 {
 	LOCKMODE	lockmode = locallock->tag.mode;
 	LOCK	   *lock = locallock->lock;
@@ -1230,8 +1230,12 @@ JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
 			/* Must he wait for me? */
 			if (lockMethodTable->conflictTab[proc->waitLockMode] & myHeldLocks)
 			{
-				/* Must I wait for him ? */
-				if (lockMethodTable->conflictTab[lockmode] & proc->heldLocks)
+				/*
+				 * Must I wait for him? I don't want a deadlock during lock
+				 * upgrade - other processes should fail on it.
+				 */
+				if ((lockMethodTable->conflictTab[lockmode] & proc->heldLocks) &&
+					!isUpgrade)
 				{
 					/*
 					 * Yes, so we have a deadlock.  Easiest way to clean up
@@ -1461,8 +1465,22 @@ ProcSleep(LOCALLOCK *locallock)
 			(void) WaitLatch(MyLatch, WL_LATCH_SET | WL_EXIT_ON_PM_DEATH, 0,
 							 PG_WAIT_LOCK | locallock->tag.lock.locktag_type);
 			ResetLatch(MyLatch);
-			/* check for deadlocks first, as that's probably log-worthy */
-			if (got_deadlock_timeout)
+			/*
+			 * Check for deadlocks first, as that's probably log-worthy.
+			 *
+			 * Do not wait for the timeout if the lock is being upgraded since
+			 * the risk of deadlock is higher now. However, do not check for
+			 * deadlock if the lock is being upgraded by this process - other
+			 * processes should take care.
+			 *
+			 * TODO Possible optimization: if this is the only lock of the
+			 * backend and if it did not have any weaker lock on the table so
+			 * far, it should be safe to skip the deadlock check. However, to
+			 * evaluate the situation, we need to take fast-path locks into
+			 * account.
+			 */
+			if ((got_deadlock_timeout || lock->upgradedBy) &&
+				lock->upgradedBy != MyProc)
 			{
 				deadlock_state = CheckDeadLock();
 				got_deadlock_timeout = false;
@@ -1819,7 +1837,7 @@ ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock)
  * not, just return.  If we have a real deadlock, remove ourselves from the
  * lock's wait queue.
  */
-static DeadLockState
+DeadLockState
 CheckDeadLock(void)
 {
 	int			i;
diff --git a/src/include/storage/lmgr.h b/src/include/storage/lmgr.h
index 2a985ce5e15..eddbe1e8622 100644
--- a/src/include/storage/lmgr.h
+++ b/src/include/storage/lmgr.h
@@ -38,6 +38,7 @@ extern void RelationInitLockInfo(Relation relation);
 
 /* Lock a relation */
 extern void LockRelationOid(Oid relid, LOCKMODE lockmode);
+extern void LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode);
 extern void LockRelationId(LockRelId *relid, LOCKMODE lockmode);
 extern bool ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode);
 extern void UnlockRelationId(LockRelId *relid, LOCKMODE lockmode);
diff --git a/src/include/storage/lock.h b/src/include/storage/lock.h
index ee3cb1dc203..080b69eddfd 100644
--- a/src/include/storage/lock.h
+++ b/src/include/storage/lock.h
@@ -131,6 +131,7 @@ typedef const LockMethodData *LockMethod;
  * nRequested -- total requested locks of all types.
  * granted -- count of each lock type currently granted on the lock.
  * nGranted -- total granted locks of all types.
+ * upgrader -- process that intends to upgrade the lock
  *
  * Note: these counts count 1 for each backend.  Internally to a backend,
  * there may be multiple grabs on a particular lock, but this is not reflected
@@ -150,6 +151,7 @@ typedef struct LOCK
 	int			nRequested;		/* total of requested[] array */
 	int			granted[MAX_LOCKMODES]; /* counts of granted locks */
 	int			nGranted;		/* total of granted[] array */
+	PGPROC		*upgradedBy;	/* is lock being upgraded by this process? */
 } LOCK;
 
 #define LOCK_LOCKMETHOD(lock) ((LOCKMETHODID) (lock).tag.locktag_lockmethodid)
@@ -390,7 +392,8 @@ extern LockAcquireResult LockAcquireExtended(const LOCKTAG *locktag,
 											 bool dontWait,
 											 bool reportMemoryError,
 											 LOCALLOCK **locallockp,
-											 bool logLockFailure);
+											 bool logLockFailure,
+											 bool isUpgrade);
 extern void AbortStrongLockAcquire(void);
 extern void MarkLockClear(LOCALLOCK *locallock);
 extern bool LockRelease(const LOCKTAG *locktag,
diff --git a/src/include/storage/proc.h b/src/include/storage/proc.h
index 3e1d1fad5f9..c4bc3a24044 100644
--- a/src/include/storage/proc.h
+++ b/src/include/storage/proc.h
@@ -563,10 +563,12 @@ extern bool HaveNFreeProcs(int n, int *nfree);
 extern void ProcReleaseLocks(bool isCommit);
 
 extern ProcWaitStatus JoinWaitQueue(LOCALLOCK *locallock,
-									LockMethod lockMethodTable, bool dontWait);
+									LockMethod lockMethodTable, bool dontWait,
+									bool isUpgrade);
 extern ProcWaitStatus ProcSleep(LOCALLOCK *locallock);
 extern void ProcWakeup(PGPROC *proc, ProcWaitStatus waitStatus);
 extern void ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock);
+extern DeadLockState CheckDeadLock(void);
 extern void CheckDeadLockAlert(void);
 extern void LockErrorCleanup(void);
 extern void GetLockHoldersAndWaiters(LOCALLOCK *locallock,
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 966+ messages in thread

* [PATCH] Lock upgrade without deadlocks.
@ 2026-04-17 13:13  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 966+ messages in thread

From: Antonin Houska @ 2026-04-17 13:13 UTC (permalink / raw)

For REPACK (CONCURRENTLY), it's essential to upgrade ShareUpdateExclusive lock
to AccessExclusiveLock for the final stage of table processing. However, we
cannot release the earlier before requesting the latter, because if another
process ran DDL on the table in between, REPACK would have to cancel its
transaction, and waste possibly a lot of work. Without releasing
ShareUpdateExclusive lock, the upgrade can cause deadlock, and it's not really
deterministic whether REPACK or the other process will get canceled.

This patch adds a new field to the LOCK structure which tells which process is
trying to upgrade the lock - this field is set right before the upgrade. All
processes waiting in the lock queue are then woken up to check the flag.

When another process tries to get the lock after that, it checks this field,
and if it's set, it checks for deadlocks even if deadlock timeout hasn't
expired yet. If the process is already in the lock's queue and sleeping, the
lock upgrading process wakes it up so it checks the flag immediately.

At the time the upgrading process starts to sleep, all the other process
should be aware that they need to check for deadlock during their lock
acquisitions, so the upgrading process does not have to check for deadlocks
when it gets woken up. Thus it should never receive deadlock error report.
---
 src/backend/commands/repack.c   |   4 +-
 src/backend/storage/lmgr/lmgr.c |  35 ++++++++---
 src/backend/storage/lmgr/lock.c | 107 +++++++++++++++++++++++++++++++-
 src/backend/storage/lmgr/proc.c |  32 +++++++---
 src/include/storage/lmgr.h      |   1 +
 src/include/storage/lock.h      |   5 +-
 src/include/storage/proc.h      |   4 +-
 7 files changed, 166 insertions(+), 22 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index 58e3867246f..310b2a65099 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -3055,8 +3055,10 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Acquire AccessExclusiveLock on the table, its TOAST relation (if there
 	 * is one), all its indexes, so that we can swap the files.
+	 *
+	 * TODO The same for indexes and TOAST?
 	 */
-	LockRelationOid(old_table_oid, AccessExclusiveLock);
+	LockRelationOidUpgrade(old_table_oid, AccessExclusiveLock);
 
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
diff --git a/src/backend/storage/lmgr/lmgr.c b/src/backend/storage/lmgr/lmgr.c
index 2ccf7237fee..0b103485d54 100644
--- a/src/backend/storage/lmgr/lmgr.c
+++ b/src/backend/storage/lmgr/lmgr.c
@@ -58,6 +58,8 @@ typedef struct XactLockTableWaitInfo
 	const ItemPointerData *ctid;
 } XactLockTableWaitInfo;
 
+static void LockRelationOidCommon(Oid relid, LOCKMODE lockmode,
+								  bool isUpgrade);
 static void XactLockTableWaitErrorCb(void *arg);
 
 /*
@@ -105,6 +107,21 @@ SetLocktagRelationOid(LOCKTAG *tag, Oid relid)
  */
 void
 LockRelationOid(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, false);
+}
+
+/*
+ * Like above, but upgrade an existing lock.
+ */
+void
+LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, true);
+}
+
+static void
+LockRelationOidCommon(Oid relid, LOCKMODE lockmode, bool isUpgrade)
 {
 	LOCKTAG		tag;
 	LOCALLOCK  *locallock;
@@ -113,7 +130,7 @@ LockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, isUpgrade);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages, so that we
@@ -157,7 +174,7 @@ ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -191,7 +208,7 @@ LockRelationId(LockRelId *relid, LOCKMODE lockmode)
 	SET_LOCKTAG_RELATION(tag, relid->dbId, relid->relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -254,7 +271,7 @@ LockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -286,7 +303,7 @@ ConditionalLockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -591,7 +608,7 @@ ConditionalLockTuple(Relation relation, const ItemPointerData *tid, LOCKMODE loc
 					  ItemPointerGetOffsetNumber(tid));
 
 	return (LockAcquireExtended(&tag, lockmode, false, true, true, NULL,
-								logLockFailure) != LOCKACQUIRE_NOT_AVAIL);
+								logLockFailure, false) != LOCKACQUIRE_NOT_AVAIL);
 }
 
 /*
@@ -749,7 +766,7 @@ ConditionalXactLockTableWait(TransactionId xid, bool logLockFailure)
 		SET_LOCKTAG_TRANSACTION(tag, xid);
 
 		if (LockAcquireExtended(&tag, ShareLock, false, true, true, NULL,
-								logLockFailure)
+								logLockFailure, false)
 			== LOCKACQUIRE_NOT_AVAIL)
 			return false;
 
@@ -1043,7 +1060,7 @@ ConditionalLockDatabaseObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -1123,7 +1140,7 @@ ConditionalLockSharedObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
diff --git a/src/backend/storage/lmgr/lock.c b/src/backend/storage/lmgr/lock.c
index c221fe96889..bcbbf36c091 100644
--- a/src/backend/storage/lmgr/lock.c
+++ b/src/backend/storage/lmgr/lock.c
@@ -810,7 +810,7 @@ LockAcquire(const LOCKTAG *locktag,
 			bool dontWait)
 {
 	return LockAcquireExtended(locktag, lockmode, sessionLock, dontWait,
-							   true, NULL, false);
+							   true, NULL, false, false);
 }
 
 /*
@@ -829,6 +829,9 @@ LockAcquire(const LOCKTAG *locktag,
  *
  * logLockFailure indicates whether to log details when a lock acquisition
  * fails with dontWait = true.
+ *
+ * isUpgrade should be true if the backend already holds this lock in a lower
+ * mode.
  */
 LockAcquireResult
 LockAcquireExtended(const LOCKTAG *locktag,
@@ -837,7 +840,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 					bool dontWait,
 					bool reportMemoryError,
 					LOCALLOCK **locallockp,
-					bool logLockFailure)
+					bool logLockFailure,
+					bool isUpgrade)
 {
 	LOCKMETHODID lockmethodid = locktag->locktag_lockmethodid;
 	LockMethod	lockMethodTable;
@@ -1119,7 +1123,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		 * case, because JoinWaitQueue() may discover that we can acquire the
 		 * lock immediately after all.
 		 */
-		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait);
+		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait,
+								   isUpgrade);
 	}
 
 	if (waitResult == PROC_WAIT_STATUS_ERROR)
@@ -1225,8 +1230,85 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		Assert(!dontWait);
 		PROCLOCK_PRINT("LockAcquire: sleeping on lock", proclock);
 		LOCK_PRINT("LockAcquire: sleeping on lock", lock, lockmode);
+
+		/*
+		 * Lock upgrade can introduce deadlock, therefore enforce special
+		 * behavior of other processes that deal with this lock. In
+		 * particular, any waiter that sees upgradedBy set is expected to
+		 * perform deadlock check as soon as it's woken up. (Since we are
+		 * already in the queue, the deadlock detector has all the information
+		 * it needs.)
+		 */
+		if (isUpgrade)
+		{
+			dlist_iter	iter;
+
+			/*
+			 * There should not be multiple upgrades at the same time. XXX
+			 * ERROR ?
+			 */
+			Assert(lock->upgradedBy == NULL);
+
+			lock->upgradedBy = MyProc;
+
+			/*
+			 * Now, before I sleep myself, wake up all the existing waiters
+			 * (except for me), so they check for deadlock.
+			 */
+			dclist_foreach(iter, &lock->waitProcs)
+			{
+				PGPROC	   *proc = dlist_container(PGPROC, waitLink,
+												   iter.cur);
+
+				if (proc != MyProc)
+					SetLatch(&proc->procLatch);
+			}
+		}
 		LWLockRelease(partitionLock);
 
+		if (!isUpgrade)
+		{
+			HASH_SEQ_STATUS status;
+			LOCALLOCK	*llock;
+			bool	check_deadlock = false;
+
+			/*
+			 * Likewise, new waiters should check the isUpgrade flag and
+			 * engage deadlock detector if needed. The lock being acquired
+			 * should already be in the hash.
+			 *
+			 * Note that, to make deadlock detection happen as soon as
+			 * possible (i.e. regardless deadlock timeout), we check the other
+			 * locks too. XXX There might be ways to get into the deadlock
+			 * indirectly, but that needs more analysis. As long as the
+			 * upgrading process skips deadlock detection altogether, the
+			 * worst consequence of missing some lock wait cycle here is that
+			 * the other process will be kicked-off no sooner than after its
+			 * deadlock timeout has elapsed. (Which in turn means that the
+			 * lock upgrade will take longer than expected.)
+			 */
+			hash_seq_init(&status, LockMethodLocalHash);
+			while ((llock = (LOCALLOCK *) hash_seq_search(&status)) != NULL)
+			{
+				LOCK	*mylock = llock->lock;
+
+				if (mylock && mylock->upgradedBy)
+				{
+					check_deadlock = true;
+					break;
+				}
+			}
+			/* Check for deadlock if needed. */
+			if (check_deadlock)
+			{
+				DeadLockState	deadlock_state;
+
+				deadlock_state = CheckDeadLock();
+				if (deadlock_state == DS_HARD_DEADLOCK)
+					DeadLockReport();
+			}
+		}
+
 		waitResult = WaitOnLock(locallock, owner);
 
 		/*
@@ -1245,6 +1327,18 @@ LockAcquireExtended(const LOCKTAG *locktag,
 			DeadLockReport();
 			/* DeadLockReport() will not return */
 		}
+
+		/*
+		 * If finishing the lock upgrade, we should not get into a deadlock
+		 * anymore, so let others know that they do not have to care either.
+		 */
+		if (isUpgrade)
+		{
+			LWLockAcquire(partitionLock, LW_EXCLUSIVE);
+			Assert(lock->upgradedBy != NULL);
+			lock->upgradedBy = NULL;
+			LWLockRelease(partitionLock);
+		}
 	}
 	else
 		LWLockRelease(partitionLock);
@@ -1322,6 +1416,7 @@ SetupLockInTable(LockMethod lockMethodTable, PGPROC *proc,
 		lock->nGranted = 0;
 		MemSet(lock->requested, 0, sizeof(int) * MAX_LOCKMODES);
 		MemSet(lock->granted, 0, sizeof(int) * MAX_LOCKMODES);
+		lock->upgradedBy = NULL;
 		LOCK_PRINT("LockAcquire: new", lock, lockmode);
 	}
 	else
@@ -1768,6 +1863,12 @@ CleanUpLock(LOCK *lock, PROCLOCK *proclock,
 			elog(PANIC, "proclock table corrupted");
 	}
 
+	/*
+	 * Was this backend upgrading the lock?
+	 */
+	if (lock->upgradedBy == MyProc)
+		lock->upgradedBy = NULL;
+
 	if (lock->nRequested == 0)
 	{
 		/*
diff --git a/src/backend/storage/lmgr/proc.c b/src/backend/storage/lmgr/proc.c
index 1ac25068d62..7b31370db70 100644
--- a/src/backend/storage/lmgr/proc.c
+++ b/src/backend/storage/lmgr/proc.c
@@ -95,7 +95,6 @@ static volatile sig_atomic_t got_deadlock_timeout;
 static void RemoveProcFromArray(int code, Datum arg);
 static void ProcKill(int code, Datum arg);
 static void AuxiliaryProcKill(int code, Datum arg);
-static DeadLockState CheckDeadLock(void);
 
 
 /*
@@ -1143,7 +1142,8 @@ AuxiliaryPidGetProc(int pid)
  * NOTES: The process queue is now a priority queue for locking.
  */
 ProcWaitStatus
-JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
+JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait,
+			  bool isUpgrade)
 {
 	LOCKMODE	lockmode = locallock->tag.mode;
 	LOCK	   *lock = locallock->lock;
@@ -1230,8 +1230,12 @@ JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
 			/* Must he wait for me? */
 			if (lockMethodTable->conflictTab[proc->waitLockMode] & myHeldLocks)
 			{
-				/* Must I wait for him ? */
-				if (lockMethodTable->conflictTab[lockmode] & proc->heldLocks)
+				/*
+				 * Must I wait for him? I don't want a deadlock during lock
+				 * upgrade - other processes should fail on it.
+				 */
+				if ((lockMethodTable->conflictTab[lockmode] & proc->heldLocks) &&
+					!isUpgrade)
 				{
 					/*
 					 * Yes, so we have a deadlock.  Easiest way to clean up
@@ -1461,8 +1465,22 @@ ProcSleep(LOCALLOCK *locallock)
 			(void) WaitLatch(MyLatch, WL_LATCH_SET | WL_EXIT_ON_PM_DEATH, 0,
 							 PG_WAIT_LOCK | locallock->tag.lock.locktag_type);
 			ResetLatch(MyLatch);
-			/* check for deadlocks first, as that's probably log-worthy */
-			if (got_deadlock_timeout)
+			/*
+			 * Check for deadlocks first, as that's probably log-worthy.
+			 *
+			 * Do not wait for the timeout if the lock is being upgraded since
+			 * the risk of deadlock is higher now. However, do not check for
+			 * deadlock if the lock is being upgraded by this process - other
+			 * processes should take care.
+			 *
+			 * TODO Possible optimization: if this is the only lock of the
+			 * backend and if it did not have any weaker lock on the table so
+			 * far, it should be safe to skip the deadlock check. However, to
+			 * evaluate the situation, we need to take fast-path locks into
+			 * account.
+			 */
+			if ((got_deadlock_timeout || lock->upgradedBy) &&
+				lock->upgradedBy != MyProc)
 			{
 				deadlock_state = CheckDeadLock();
 				got_deadlock_timeout = false;
@@ -1819,7 +1837,7 @@ ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock)
  * not, just return.  If we have a real deadlock, remove ourselves from the
  * lock's wait queue.
  */
-static DeadLockState
+DeadLockState
 CheckDeadLock(void)
 {
 	int			i;
diff --git a/src/include/storage/lmgr.h b/src/include/storage/lmgr.h
index 2a985ce5e15..eddbe1e8622 100644
--- a/src/include/storage/lmgr.h
+++ b/src/include/storage/lmgr.h
@@ -38,6 +38,7 @@ extern void RelationInitLockInfo(Relation relation);
 
 /* Lock a relation */
 extern void LockRelationOid(Oid relid, LOCKMODE lockmode);
+extern void LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode);
 extern void LockRelationId(LockRelId *relid, LOCKMODE lockmode);
 extern bool ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode);
 extern void UnlockRelationId(LockRelId *relid, LOCKMODE lockmode);
diff --git a/src/include/storage/lock.h b/src/include/storage/lock.h
index ee3cb1dc203..080b69eddfd 100644
--- a/src/include/storage/lock.h
+++ b/src/include/storage/lock.h
@@ -131,6 +131,7 @@ typedef const LockMethodData *LockMethod;
  * nRequested -- total requested locks of all types.
  * granted -- count of each lock type currently granted on the lock.
  * nGranted -- total granted locks of all types.
+ * upgrader -- process that intends to upgrade the lock
  *
  * Note: these counts count 1 for each backend.  Internally to a backend,
  * there may be multiple grabs on a particular lock, but this is not reflected
@@ -150,6 +151,7 @@ typedef struct LOCK
 	int			nRequested;		/* total of requested[] array */
 	int			granted[MAX_LOCKMODES]; /* counts of granted locks */
 	int			nGranted;		/* total of granted[] array */
+	PGPROC		*upgradedBy;	/* is lock being upgraded by this process? */
 } LOCK;
 
 #define LOCK_LOCKMETHOD(lock) ((LOCKMETHODID) (lock).tag.locktag_lockmethodid)
@@ -390,7 +392,8 @@ extern LockAcquireResult LockAcquireExtended(const LOCKTAG *locktag,
 											 bool dontWait,
 											 bool reportMemoryError,
 											 LOCALLOCK **locallockp,
-											 bool logLockFailure);
+											 bool logLockFailure,
+											 bool isUpgrade);
 extern void AbortStrongLockAcquire(void);
 extern void MarkLockClear(LOCALLOCK *locallock);
 extern bool LockRelease(const LOCKTAG *locktag,
diff --git a/src/include/storage/proc.h b/src/include/storage/proc.h
index 3e1d1fad5f9..c4bc3a24044 100644
--- a/src/include/storage/proc.h
+++ b/src/include/storage/proc.h
@@ -563,10 +563,12 @@ extern bool HaveNFreeProcs(int n, int *nfree);
 extern void ProcReleaseLocks(bool isCommit);
 
 extern ProcWaitStatus JoinWaitQueue(LOCALLOCK *locallock,
-									LockMethod lockMethodTable, bool dontWait);
+									LockMethod lockMethodTable, bool dontWait,
+									bool isUpgrade);
 extern ProcWaitStatus ProcSleep(LOCALLOCK *locallock);
 extern void ProcWakeup(PGPROC *proc, ProcWaitStatus waitStatus);
 extern void ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock);
+extern DeadLockState CheckDeadLock(void);
 extern void CheckDeadLockAlert(void);
 extern void LockErrorCleanup(void);
 extern void GetLockHoldersAndWaiters(LOCALLOCK *locallock,
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 966+ messages in thread

* [PATCH] Lock upgrade without deadlocks.
@ 2026-04-17 13:13  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 966+ messages in thread

From: Antonin Houska @ 2026-04-17 13:13 UTC (permalink / raw)

For REPACK (CONCURRENTLY), it's essential to upgrade ShareUpdateExclusive lock
to AccessExclusiveLock for the final stage of table processing. However, we
cannot release the earlier before requesting the latter, because if another
process ran DDL on the table in between, REPACK would have to cancel its
transaction, and waste possibly a lot of work. Without releasing
ShareUpdateExclusive lock, the upgrade can cause deadlock, and it's not really
deterministic whether REPACK or the other process will get canceled.

This patch adds a new field to the LOCK structure which tells which process is
trying to upgrade the lock - this field is set right before the upgrade. All
processes waiting in the lock queue are then woken up to check the flag.

When another process tries to get the lock after that, it checks this field,
and if it's set, it checks for deadlocks even if deadlock timeout hasn't
expired yet. If the process is already in the lock's queue and sleeping, the
lock upgrading process wakes it up so it checks the flag immediately.

At the time the upgrading process starts to sleep, all the other process
should be aware that they need to check for deadlock during their lock
acquisitions, so the upgrading process does not have to check for deadlocks
when it gets woken up. Thus it should never receive deadlock error report.
---
 src/backend/commands/repack.c   |   4 +-
 src/backend/storage/lmgr/lmgr.c |  35 ++++++++---
 src/backend/storage/lmgr/lock.c | 107 +++++++++++++++++++++++++++++++-
 src/backend/storage/lmgr/proc.c |  32 +++++++---
 src/include/storage/lmgr.h      |   1 +
 src/include/storage/lock.h      |   5 +-
 src/include/storage/proc.h      |   4 +-
 7 files changed, 166 insertions(+), 22 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index 58e3867246f..310b2a65099 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -3055,8 +3055,10 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Acquire AccessExclusiveLock on the table, its TOAST relation (if there
 	 * is one), all its indexes, so that we can swap the files.
+	 *
+	 * TODO The same for indexes and TOAST?
 	 */
-	LockRelationOid(old_table_oid, AccessExclusiveLock);
+	LockRelationOidUpgrade(old_table_oid, AccessExclusiveLock);
 
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
diff --git a/src/backend/storage/lmgr/lmgr.c b/src/backend/storage/lmgr/lmgr.c
index 2ccf7237fee..0b103485d54 100644
--- a/src/backend/storage/lmgr/lmgr.c
+++ b/src/backend/storage/lmgr/lmgr.c
@@ -58,6 +58,8 @@ typedef struct XactLockTableWaitInfo
 	const ItemPointerData *ctid;
 } XactLockTableWaitInfo;
 
+static void LockRelationOidCommon(Oid relid, LOCKMODE lockmode,
+								  bool isUpgrade);
 static void XactLockTableWaitErrorCb(void *arg);
 
 /*
@@ -105,6 +107,21 @@ SetLocktagRelationOid(LOCKTAG *tag, Oid relid)
  */
 void
 LockRelationOid(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, false);
+}
+
+/*
+ * Like above, but upgrade an existing lock.
+ */
+void
+LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, true);
+}
+
+static void
+LockRelationOidCommon(Oid relid, LOCKMODE lockmode, bool isUpgrade)
 {
 	LOCKTAG		tag;
 	LOCALLOCK  *locallock;
@@ -113,7 +130,7 @@ LockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, isUpgrade);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages, so that we
@@ -157,7 +174,7 @@ ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -191,7 +208,7 @@ LockRelationId(LockRelId *relid, LOCKMODE lockmode)
 	SET_LOCKTAG_RELATION(tag, relid->dbId, relid->relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -254,7 +271,7 @@ LockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -286,7 +303,7 @@ ConditionalLockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -591,7 +608,7 @@ ConditionalLockTuple(Relation relation, const ItemPointerData *tid, LOCKMODE loc
 					  ItemPointerGetOffsetNumber(tid));
 
 	return (LockAcquireExtended(&tag, lockmode, false, true, true, NULL,
-								logLockFailure) != LOCKACQUIRE_NOT_AVAIL);
+								logLockFailure, false) != LOCKACQUIRE_NOT_AVAIL);
 }
 
 /*
@@ -749,7 +766,7 @@ ConditionalXactLockTableWait(TransactionId xid, bool logLockFailure)
 		SET_LOCKTAG_TRANSACTION(tag, xid);
 
 		if (LockAcquireExtended(&tag, ShareLock, false, true, true, NULL,
-								logLockFailure)
+								logLockFailure, false)
 			== LOCKACQUIRE_NOT_AVAIL)
 			return false;
 
@@ -1043,7 +1060,7 @@ ConditionalLockDatabaseObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -1123,7 +1140,7 @@ ConditionalLockSharedObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
diff --git a/src/backend/storage/lmgr/lock.c b/src/backend/storage/lmgr/lock.c
index c221fe96889..bcbbf36c091 100644
--- a/src/backend/storage/lmgr/lock.c
+++ b/src/backend/storage/lmgr/lock.c
@@ -810,7 +810,7 @@ LockAcquire(const LOCKTAG *locktag,
 			bool dontWait)
 {
 	return LockAcquireExtended(locktag, lockmode, sessionLock, dontWait,
-							   true, NULL, false);
+							   true, NULL, false, false);
 }
 
 /*
@@ -829,6 +829,9 @@ LockAcquire(const LOCKTAG *locktag,
  *
  * logLockFailure indicates whether to log details when a lock acquisition
  * fails with dontWait = true.
+ *
+ * isUpgrade should be true if the backend already holds this lock in a lower
+ * mode.
  */
 LockAcquireResult
 LockAcquireExtended(const LOCKTAG *locktag,
@@ -837,7 +840,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 					bool dontWait,
 					bool reportMemoryError,
 					LOCALLOCK **locallockp,
-					bool logLockFailure)
+					bool logLockFailure,
+					bool isUpgrade)
 {
 	LOCKMETHODID lockmethodid = locktag->locktag_lockmethodid;
 	LockMethod	lockMethodTable;
@@ -1119,7 +1123,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		 * case, because JoinWaitQueue() may discover that we can acquire the
 		 * lock immediately after all.
 		 */
-		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait);
+		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait,
+								   isUpgrade);
 	}
 
 	if (waitResult == PROC_WAIT_STATUS_ERROR)
@@ -1225,8 +1230,85 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		Assert(!dontWait);
 		PROCLOCK_PRINT("LockAcquire: sleeping on lock", proclock);
 		LOCK_PRINT("LockAcquire: sleeping on lock", lock, lockmode);
+
+		/*
+		 * Lock upgrade can introduce deadlock, therefore enforce special
+		 * behavior of other processes that deal with this lock. In
+		 * particular, any waiter that sees upgradedBy set is expected to
+		 * perform deadlock check as soon as it's woken up. (Since we are
+		 * already in the queue, the deadlock detector has all the information
+		 * it needs.)
+		 */
+		if (isUpgrade)
+		{
+			dlist_iter	iter;
+
+			/*
+			 * There should not be multiple upgrades at the same time. XXX
+			 * ERROR ?
+			 */
+			Assert(lock->upgradedBy == NULL);
+
+			lock->upgradedBy = MyProc;
+
+			/*
+			 * Now, before I sleep myself, wake up all the existing waiters
+			 * (except for me), so they check for deadlock.
+			 */
+			dclist_foreach(iter, &lock->waitProcs)
+			{
+				PGPROC	   *proc = dlist_container(PGPROC, waitLink,
+												   iter.cur);
+
+				if (proc != MyProc)
+					SetLatch(&proc->procLatch);
+			}
+		}
 		LWLockRelease(partitionLock);
 
+		if (!isUpgrade)
+		{
+			HASH_SEQ_STATUS status;
+			LOCALLOCK	*llock;
+			bool	check_deadlock = false;
+
+			/*
+			 * Likewise, new waiters should check the isUpgrade flag and
+			 * engage deadlock detector if needed. The lock being acquired
+			 * should already be in the hash.
+			 *
+			 * Note that, to make deadlock detection happen as soon as
+			 * possible (i.e. regardless deadlock timeout), we check the other
+			 * locks too. XXX There might be ways to get into the deadlock
+			 * indirectly, but that needs more analysis. As long as the
+			 * upgrading process skips deadlock detection altogether, the
+			 * worst consequence of missing some lock wait cycle here is that
+			 * the other process will be kicked-off no sooner than after its
+			 * deadlock timeout has elapsed. (Which in turn means that the
+			 * lock upgrade will take longer than expected.)
+			 */
+			hash_seq_init(&status, LockMethodLocalHash);
+			while ((llock = (LOCALLOCK *) hash_seq_search(&status)) != NULL)
+			{
+				LOCK	*mylock = llock->lock;
+
+				if (mylock && mylock->upgradedBy)
+				{
+					check_deadlock = true;
+					break;
+				}
+			}
+			/* Check for deadlock if needed. */
+			if (check_deadlock)
+			{
+				DeadLockState	deadlock_state;
+
+				deadlock_state = CheckDeadLock();
+				if (deadlock_state == DS_HARD_DEADLOCK)
+					DeadLockReport();
+			}
+		}
+
 		waitResult = WaitOnLock(locallock, owner);
 
 		/*
@@ -1245,6 +1327,18 @@ LockAcquireExtended(const LOCKTAG *locktag,
 			DeadLockReport();
 			/* DeadLockReport() will not return */
 		}
+
+		/*
+		 * If finishing the lock upgrade, we should not get into a deadlock
+		 * anymore, so let others know that they do not have to care either.
+		 */
+		if (isUpgrade)
+		{
+			LWLockAcquire(partitionLock, LW_EXCLUSIVE);
+			Assert(lock->upgradedBy != NULL);
+			lock->upgradedBy = NULL;
+			LWLockRelease(partitionLock);
+		}
 	}
 	else
 		LWLockRelease(partitionLock);
@@ -1322,6 +1416,7 @@ SetupLockInTable(LockMethod lockMethodTable, PGPROC *proc,
 		lock->nGranted = 0;
 		MemSet(lock->requested, 0, sizeof(int) * MAX_LOCKMODES);
 		MemSet(lock->granted, 0, sizeof(int) * MAX_LOCKMODES);
+		lock->upgradedBy = NULL;
 		LOCK_PRINT("LockAcquire: new", lock, lockmode);
 	}
 	else
@@ -1768,6 +1863,12 @@ CleanUpLock(LOCK *lock, PROCLOCK *proclock,
 			elog(PANIC, "proclock table corrupted");
 	}
 
+	/*
+	 * Was this backend upgrading the lock?
+	 */
+	if (lock->upgradedBy == MyProc)
+		lock->upgradedBy = NULL;
+
 	if (lock->nRequested == 0)
 	{
 		/*
diff --git a/src/backend/storage/lmgr/proc.c b/src/backend/storage/lmgr/proc.c
index 1ac25068d62..7b31370db70 100644
--- a/src/backend/storage/lmgr/proc.c
+++ b/src/backend/storage/lmgr/proc.c
@@ -95,7 +95,6 @@ static volatile sig_atomic_t got_deadlock_timeout;
 static void RemoveProcFromArray(int code, Datum arg);
 static void ProcKill(int code, Datum arg);
 static void AuxiliaryProcKill(int code, Datum arg);
-static DeadLockState CheckDeadLock(void);
 
 
 /*
@@ -1143,7 +1142,8 @@ AuxiliaryPidGetProc(int pid)
  * NOTES: The process queue is now a priority queue for locking.
  */
 ProcWaitStatus
-JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
+JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait,
+			  bool isUpgrade)
 {
 	LOCKMODE	lockmode = locallock->tag.mode;
 	LOCK	   *lock = locallock->lock;
@@ -1230,8 +1230,12 @@ JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
 			/* Must he wait for me? */
 			if (lockMethodTable->conflictTab[proc->waitLockMode] & myHeldLocks)
 			{
-				/* Must I wait for him ? */
-				if (lockMethodTable->conflictTab[lockmode] & proc->heldLocks)
+				/*
+				 * Must I wait for him? I don't want a deadlock during lock
+				 * upgrade - other processes should fail on it.
+				 */
+				if ((lockMethodTable->conflictTab[lockmode] & proc->heldLocks) &&
+					!isUpgrade)
 				{
 					/*
 					 * Yes, so we have a deadlock.  Easiest way to clean up
@@ -1461,8 +1465,22 @@ ProcSleep(LOCALLOCK *locallock)
 			(void) WaitLatch(MyLatch, WL_LATCH_SET | WL_EXIT_ON_PM_DEATH, 0,
 							 PG_WAIT_LOCK | locallock->tag.lock.locktag_type);
 			ResetLatch(MyLatch);
-			/* check for deadlocks first, as that's probably log-worthy */
-			if (got_deadlock_timeout)
+			/*
+			 * Check for deadlocks first, as that's probably log-worthy.
+			 *
+			 * Do not wait for the timeout if the lock is being upgraded since
+			 * the risk of deadlock is higher now. However, do not check for
+			 * deadlock if the lock is being upgraded by this process - other
+			 * processes should take care.
+			 *
+			 * TODO Possible optimization: if this is the only lock of the
+			 * backend and if it did not have any weaker lock on the table so
+			 * far, it should be safe to skip the deadlock check. However, to
+			 * evaluate the situation, we need to take fast-path locks into
+			 * account.
+			 */
+			if ((got_deadlock_timeout || lock->upgradedBy) &&
+				lock->upgradedBy != MyProc)
 			{
 				deadlock_state = CheckDeadLock();
 				got_deadlock_timeout = false;
@@ -1819,7 +1837,7 @@ ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock)
  * not, just return.  If we have a real deadlock, remove ourselves from the
  * lock's wait queue.
  */
-static DeadLockState
+DeadLockState
 CheckDeadLock(void)
 {
 	int			i;
diff --git a/src/include/storage/lmgr.h b/src/include/storage/lmgr.h
index 2a985ce5e15..eddbe1e8622 100644
--- a/src/include/storage/lmgr.h
+++ b/src/include/storage/lmgr.h
@@ -38,6 +38,7 @@ extern void RelationInitLockInfo(Relation relation);
 
 /* Lock a relation */
 extern void LockRelationOid(Oid relid, LOCKMODE lockmode);
+extern void LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode);
 extern void LockRelationId(LockRelId *relid, LOCKMODE lockmode);
 extern bool ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode);
 extern void UnlockRelationId(LockRelId *relid, LOCKMODE lockmode);
diff --git a/src/include/storage/lock.h b/src/include/storage/lock.h
index ee3cb1dc203..080b69eddfd 100644
--- a/src/include/storage/lock.h
+++ b/src/include/storage/lock.h
@@ -131,6 +131,7 @@ typedef const LockMethodData *LockMethod;
  * nRequested -- total requested locks of all types.
  * granted -- count of each lock type currently granted on the lock.
  * nGranted -- total granted locks of all types.
+ * upgrader -- process that intends to upgrade the lock
  *
  * Note: these counts count 1 for each backend.  Internally to a backend,
  * there may be multiple grabs on a particular lock, but this is not reflected
@@ -150,6 +151,7 @@ typedef struct LOCK
 	int			nRequested;		/* total of requested[] array */
 	int			granted[MAX_LOCKMODES]; /* counts of granted locks */
 	int			nGranted;		/* total of granted[] array */
+	PGPROC		*upgradedBy;	/* is lock being upgraded by this process? */
 } LOCK;
 
 #define LOCK_LOCKMETHOD(lock) ((LOCKMETHODID) (lock).tag.locktag_lockmethodid)
@@ -390,7 +392,8 @@ extern LockAcquireResult LockAcquireExtended(const LOCKTAG *locktag,
 											 bool dontWait,
 											 bool reportMemoryError,
 											 LOCALLOCK **locallockp,
-											 bool logLockFailure);
+											 bool logLockFailure,
+											 bool isUpgrade);
 extern void AbortStrongLockAcquire(void);
 extern void MarkLockClear(LOCALLOCK *locallock);
 extern bool LockRelease(const LOCKTAG *locktag,
diff --git a/src/include/storage/proc.h b/src/include/storage/proc.h
index 3e1d1fad5f9..c4bc3a24044 100644
--- a/src/include/storage/proc.h
+++ b/src/include/storage/proc.h
@@ -563,10 +563,12 @@ extern bool HaveNFreeProcs(int n, int *nfree);
 extern void ProcReleaseLocks(bool isCommit);
 
 extern ProcWaitStatus JoinWaitQueue(LOCALLOCK *locallock,
-									LockMethod lockMethodTable, bool dontWait);
+									LockMethod lockMethodTable, bool dontWait,
+									bool isUpgrade);
 extern ProcWaitStatus ProcSleep(LOCALLOCK *locallock);
 extern void ProcWakeup(PGPROC *proc, ProcWaitStatus waitStatus);
 extern void ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock);
+extern DeadLockState CheckDeadLock(void);
 extern void CheckDeadLockAlert(void);
 extern void LockErrorCleanup(void);
 extern void GetLockHoldersAndWaiters(LOCALLOCK *locallock,
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 966+ messages in thread

* [PATCH] Lock upgrade without deadlocks.
@ 2026-04-17 13:13  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 966+ messages in thread

From: Antonin Houska @ 2026-04-17 13:13 UTC (permalink / raw)

For REPACK (CONCURRENTLY), it's essential to upgrade ShareUpdateExclusive lock
to AccessExclusiveLock for the final stage of table processing. However, we
cannot release the earlier before requesting the latter, because if another
process ran DDL on the table in between, REPACK would have to cancel its
transaction, and waste possibly a lot of work. Without releasing
ShareUpdateExclusive lock, the upgrade can cause deadlock, and it's not really
deterministic whether REPACK or the other process will get canceled.

This patch adds a new field to the LOCK structure which tells which process is
trying to upgrade the lock - this field is set right before the upgrade. All
processes waiting in the lock queue are then woken up to check the flag.

When another process tries to get the lock after that, it checks this field,
and if it's set, it checks for deadlocks even if deadlock timeout hasn't
expired yet. If the process is already in the lock's queue and sleeping, the
lock upgrading process wakes it up so it checks the flag immediately.

At the time the upgrading process starts to sleep, all the other process
should be aware that they need to check for deadlock during their lock
acquisitions, so the upgrading process does not have to check for deadlocks
when it gets woken up. Thus it should never receive deadlock error report.
---
 src/backend/commands/repack.c   |   4 +-
 src/backend/storage/lmgr/lmgr.c |  35 ++++++++---
 src/backend/storage/lmgr/lock.c | 107 +++++++++++++++++++++++++++++++-
 src/backend/storage/lmgr/proc.c |  32 +++++++---
 src/include/storage/lmgr.h      |   1 +
 src/include/storage/lock.h      |   5 +-
 src/include/storage/proc.h      |   4 +-
 7 files changed, 166 insertions(+), 22 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index 58e3867246f..310b2a65099 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -3055,8 +3055,10 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Acquire AccessExclusiveLock on the table, its TOAST relation (if there
 	 * is one), all its indexes, so that we can swap the files.
+	 *
+	 * TODO The same for indexes and TOAST?
 	 */
-	LockRelationOid(old_table_oid, AccessExclusiveLock);
+	LockRelationOidUpgrade(old_table_oid, AccessExclusiveLock);
 
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
diff --git a/src/backend/storage/lmgr/lmgr.c b/src/backend/storage/lmgr/lmgr.c
index 2ccf7237fee..0b103485d54 100644
--- a/src/backend/storage/lmgr/lmgr.c
+++ b/src/backend/storage/lmgr/lmgr.c
@@ -58,6 +58,8 @@ typedef struct XactLockTableWaitInfo
 	const ItemPointerData *ctid;
 } XactLockTableWaitInfo;
 
+static void LockRelationOidCommon(Oid relid, LOCKMODE lockmode,
+								  bool isUpgrade);
 static void XactLockTableWaitErrorCb(void *arg);
 
 /*
@@ -105,6 +107,21 @@ SetLocktagRelationOid(LOCKTAG *tag, Oid relid)
  */
 void
 LockRelationOid(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, false);
+}
+
+/*
+ * Like above, but upgrade an existing lock.
+ */
+void
+LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, true);
+}
+
+static void
+LockRelationOidCommon(Oid relid, LOCKMODE lockmode, bool isUpgrade)
 {
 	LOCKTAG		tag;
 	LOCALLOCK  *locallock;
@@ -113,7 +130,7 @@ LockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, isUpgrade);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages, so that we
@@ -157,7 +174,7 @@ ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -191,7 +208,7 @@ LockRelationId(LockRelId *relid, LOCKMODE lockmode)
 	SET_LOCKTAG_RELATION(tag, relid->dbId, relid->relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -254,7 +271,7 @@ LockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -286,7 +303,7 @@ ConditionalLockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -591,7 +608,7 @@ ConditionalLockTuple(Relation relation, const ItemPointerData *tid, LOCKMODE loc
 					  ItemPointerGetOffsetNumber(tid));
 
 	return (LockAcquireExtended(&tag, lockmode, false, true, true, NULL,
-								logLockFailure) != LOCKACQUIRE_NOT_AVAIL);
+								logLockFailure, false) != LOCKACQUIRE_NOT_AVAIL);
 }
 
 /*
@@ -749,7 +766,7 @@ ConditionalXactLockTableWait(TransactionId xid, bool logLockFailure)
 		SET_LOCKTAG_TRANSACTION(tag, xid);
 
 		if (LockAcquireExtended(&tag, ShareLock, false, true, true, NULL,
-								logLockFailure)
+								logLockFailure, false)
 			== LOCKACQUIRE_NOT_AVAIL)
 			return false;
 
@@ -1043,7 +1060,7 @@ ConditionalLockDatabaseObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -1123,7 +1140,7 @@ ConditionalLockSharedObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
diff --git a/src/backend/storage/lmgr/lock.c b/src/backend/storage/lmgr/lock.c
index c221fe96889..bcbbf36c091 100644
--- a/src/backend/storage/lmgr/lock.c
+++ b/src/backend/storage/lmgr/lock.c
@@ -810,7 +810,7 @@ LockAcquire(const LOCKTAG *locktag,
 			bool dontWait)
 {
 	return LockAcquireExtended(locktag, lockmode, sessionLock, dontWait,
-							   true, NULL, false);
+							   true, NULL, false, false);
 }
 
 /*
@@ -829,6 +829,9 @@ LockAcquire(const LOCKTAG *locktag,
  *
  * logLockFailure indicates whether to log details when a lock acquisition
  * fails with dontWait = true.
+ *
+ * isUpgrade should be true if the backend already holds this lock in a lower
+ * mode.
  */
 LockAcquireResult
 LockAcquireExtended(const LOCKTAG *locktag,
@@ -837,7 +840,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 					bool dontWait,
 					bool reportMemoryError,
 					LOCALLOCK **locallockp,
-					bool logLockFailure)
+					bool logLockFailure,
+					bool isUpgrade)
 {
 	LOCKMETHODID lockmethodid = locktag->locktag_lockmethodid;
 	LockMethod	lockMethodTable;
@@ -1119,7 +1123,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		 * case, because JoinWaitQueue() may discover that we can acquire the
 		 * lock immediately after all.
 		 */
-		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait);
+		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait,
+								   isUpgrade);
 	}
 
 	if (waitResult == PROC_WAIT_STATUS_ERROR)
@@ -1225,8 +1230,85 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		Assert(!dontWait);
 		PROCLOCK_PRINT("LockAcquire: sleeping on lock", proclock);
 		LOCK_PRINT("LockAcquire: sleeping on lock", lock, lockmode);
+
+		/*
+		 * Lock upgrade can introduce deadlock, therefore enforce special
+		 * behavior of other processes that deal with this lock. In
+		 * particular, any waiter that sees upgradedBy set is expected to
+		 * perform deadlock check as soon as it's woken up. (Since we are
+		 * already in the queue, the deadlock detector has all the information
+		 * it needs.)
+		 */
+		if (isUpgrade)
+		{
+			dlist_iter	iter;
+
+			/*
+			 * There should not be multiple upgrades at the same time. XXX
+			 * ERROR ?
+			 */
+			Assert(lock->upgradedBy == NULL);
+
+			lock->upgradedBy = MyProc;
+
+			/*
+			 * Now, before I sleep myself, wake up all the existing waiters
+			 * (except for me), so they check for deadlock.
+			 */
+			dclist_foreach(iter, &lock->waitProcs)
+			{
+				PGPROC	   *proc = dlist_container(PGPROC, waitLink,
+												   iter.cur);
+
+				if (proc != MyProc)
+					SetLatch(&proc->procLatch);
+			}
+		}
 		LWLockRelease(partitionLock);
 
+		if (!isUpgrade)
+		{
+			HASH_SEQ_STATUS status;
+			LOCALLOCK	*llock;
+			bool	check_deadlock = false;
+
+			/*
+			 * Likewise, new waiters should check the isUpgrade flag and
+			 * engage deadlock detector if needed. The lock being acquired
+			 * should already be in the hash.
+			 *
+			 * Note that, to make deadlock detection happen as soon as
+			 * possible (i.e. regardless deadlock timeout), we check the other
+			 * locks too. XXX There might be ways to get into the deadlock
+			 * indirectly, but that needs more analysis. As long as the
+			 * upgrading process skips deadlock detection altogether, the
+			 * worst consequence of missing some lock wait cycle here is that
+			 * the other process will be kicked-off no sooner than after its
+			 * deadlock timeout has elapsed. (Which in turn means that the
+			 * lock upgrade will take longer than expected.)
+			 */
+			hash_seq_init(&status, LockMethodLocalHash);
+			while ((llock = (LOCALLOCK *) hash_seq_search(&status)) != NULL)
+			{
+				LOCK	*mylock = llock->lock;
+
+				if (mylock && mylock->upgradedBy)
+				{
+					check_deadlock = true;
+					break;
+				}
+			}
+			/* Check for deadlock if needed. */
+			if (check_deadlock)
+			{
+				DeadLockState	deadlock_state;
+
+				deadlock_state = CheckDeadLock();
+				if (deadlock_state == DS_HARD_DEADLOCK)
+					DeadLockReport();
+			}
+		}
+
 		waitResult = WaitOnLock(locallock, owner);
 
 		/*
@@ -1245,6 +1327,18 @@ LockAcquireExtended(const LOCKTAG *locktag,
 			DeadLockReport();
 			/* DeadLockReport() will not return */
 		}
+
+		/*
+		 * If finishing the lock upgrade, we should not get into a deadlock
+		 * anymore, so let others know that they do not have to care either.
+		 */
+		if (isUpgrade)
+		{
+			LWLockAcquire(partitionLock, LW_EXCLUSIVE);
+			Assert(lock->upgradedBy != NULL);
+			lock->upgradedBy = NULL;
+			LWLockRelease(partitionLock);
+		}
 	}
 	else
 		LWLockRelease(partitionLock);
@@ -1322,6 +1416,7 @@ SetupLockInTable(LockMethod lockMethodTable, PGPROC *proc,
 		lock->nGranted = 0;
 		MemSet(lock->requested, 0, sizeof(int) * MAX_LOCKMODES);
 		MemSet(lock->granted, 0, sizeof(int) * MAX_LOCKMODES);
+		lock->upgradedBy = NULL;
 		LOCK_PRINT("LockAcquire: new", lock, lockmode);
 	}
 	else
@@ -1768,6 +1863,12 @@ CleanUpLock(LOCK *lock, PROCLOCK *proclock,
 			elog(PANIC, "proclock table corrupted");
 	}
 
+	/*
+	 * Was this backend upgrading the lock?
+	 */
+	if (lock->upgradedBy == MyProc)
+		lock->upgradedBy = NULL;
+
 	if (lock->nRequested == 0)
 	{
 		/*
diff --git a/src/backend/storage/lmgr/proc.c b/src/backend/storage/lmgr/proc.c
index 1ac25068d62..7b31370db70 100644
--- a/src/backend/storage/lmgr/proc.c
+++ b/src/backend/storage/lmgr/proc.c
@@ -95,7 +95,6 @@ static volatile sig_atomic_t got_deadlock_timeout;
 static void RemoveProcFromArray(int code, Datum arg);
 static void ProcKill(int code, Datum arg);
 static void AuxiliaryProcKill(int code, Datum arg);
-static DeadLockState CheckDeadLock(void);
 
 
 /*
@@ -1143,7 +1142,8 @@ AuxiliaryPidGetProc(int pid)
  * NOTES: The process queue is now a priority queue for locking.
  */
 ProcWaitStatus
-JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
+JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait,
+			  bool isUpgrade)
 {
 	LOCKMODE	lockmode = locallock->tag.mode;
 	LOCK	   *lock = locallock->lock;
@@ -1230,8 +1230,12 @@ JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
 			/* Must he wait for me? */
 			if (lockMethodTable->conflictTab[proc->waitLockMode] & myHeldLocks)
 			{
-				/* Must I wait for him ? */
-				if (lockMethodTable->conflictTab[lockmode] & proc->heldLocks)
+				/*
+				 * Must I wait for him? I don't want a deadlock during lock
+				 * upgrade - other processes should fail on it.
+				 */
+				if ((lockMethodTable->conflictTab[lockmode] & proc->heldLocks) &&
+					!isUpgrade)
 				{
 					/*
 					 * Yes, so we have a deadlock.  Easiest way to clean up
@@ -1461,8 +1465,22 @@ ProcSleep(LOCALLOCK *locallock)
 			(void) WaitLatch(MyLatch, WL_LATCH_SET | WL_EXIT_ON_PM_DEATH, 0,
 							 PG_WAIT_LOCK | locallock->tag.lock.locktag_type);
 			ResetLatch(MyLatch);
-			/* check for deadlocks first, as that's probably log-worthy */
-			if (got_deadlock_timeout)
+			/*
+			 * Check for deadlocks first, as that's probably log-worthy.
+			 *
+			 * Do not wait for the timeout if the lock is being upgraded since
+			 * the risk of deadlock is higher now. However, do not check for
+			 * deadlock if the lock is being upgraded by this process - other
+			 * processes should take care.
+			 *
+			 * TODO Possible optimization: if this is the only lock of the
+			 * backend and if it did not have any weaker lock on the table so
+			 * far, it should be safe to skip the deadlock check. However, to
+			 * evaluate the situation, we need to take fast-path locks into
+			 * account.
+			 */
+			if ((got_deadlock_timeout || lock->upgradedBy) &&
+				lock->upgradedBy != MyProc)
 			{
 				deadlock_state = CheckDeadLock();
 				got_deadlock_timeout = false;
@@ -1819,7 +1837,7 @@ ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock)
  * not, just return.  If we have a real deadlock, remove ourselves from the
  * lock's wait queue.
  */
-static DeadLockState
+DeadLockState
 CheckDeadLock(void)
 {
 	int			i;
diff --git a/src/include/storage/lmgr.h b/src/include/storage/lmgr.h
index 2a985ce5e15..eddbe1e8622 100644
--- a/src/include/storage/lmgr.h
+++ b/src/include/storage/lmgr.h
@@ -38,6 +38,7 @@ extern void RelationInitLockInfo(Relation relation);
 
 /* Lock a relation */
 extern void LockRelationOid(Oid relid, LOCKMODE lockmode);
+extern void LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode);
 extern void LockRelationId(LockRelId *relid, LOCKMODE lockmode);
 extern bool ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode);
 extern void UnlockRelationId(LockRelId *relid, LOCKMODE lockmode);
diff --git a/src/include/storage/lock.h b/src/include/storage/lock.h
index ee3cb1dc203..080b69eddfd 100644
--- a/src/include/storage/lock.h
+++ b/src/include/storage/lock.h
@@ -131,6 +131,7 @@ typedef const LockMethodData *LockMethod;
  * nRequested -- total requested locks of all types.
  * granted -- count of each lock type currently granted on the lock.
  * nGranted -- total granted locks of all types.
+ * upgrader -- process that intends to upgrade the lock
  *
  * Note: these counts count 1 for each backend.  Internally to a backend,
  * there may be multiple grabs on a particular lock, but this is not reflected
@@ -150,6 +151,7 @@ typedef struct LOCK
 	int			nRequested;		/* total of requested[] array */
 	int			granted[MAX_LOCKMODES]; /* counts of granted locks */
 	int			nGranted;		/* total of granted[] array */
+	PGPROC		*upgradedBy;	/* is lock being upgraded by this process? */
 } LOCK;
 
 #define LOCK_LOCKMETHOD(lock) ((LOCKMETHODID) (lock).tag.locktag_lockmethodid)
@@ -390,7 +392,8 @@ extern LockAcquireResult LockAcquireExtended(const LOCKTAG *locktag,
 											 bool dontWait,
 											 bool reportMemoryError,
 											 LOCALLOCK **locallockp,
-											 bool logLockFailure);
+											 bool logLockFailure,
+											 bool isUpgrade);
 extern void AbortStrongLockAcquire(void);
 extern void MarkLockClear(LOCALLOCK *locallock);
 extern bool LockRelease(const LOCKTAG *locktag,
diff --git a/src/include/storage/proc.h b/src/include/storage/proc.h
index 3e1d1fad5f9..c4bc3a24044 100644
--- a/src/include/storage/proc.h
+++ b/src/include/storage/proc.h
@@ -563,10 +563,12 @@ extern bool HaveNFreeProcs(int n, int *nfree);
 extern void ProcReleaseLocks(bool isCommit);
 
 extern ProcWaitStatus JoinWaitQueue(LOCALLOCK *locallock,
-									LockMethod lockMethodTable, bool dontWait);
+									LockMethod lockMethodTable, bool dontWait,
+									bool isUpgrade);
 extern ProcWaitStatus ProcSleep(LOCALLOCK *locallock);
 extern void ProcWakeup(PGPROC *proc, ProcWaitStatus waitStatus);
 extern void ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock);
+extern DeadLockState CheckDeadLock(void);
 extern void CheckDeadLockAlert(void);
 extern void LockErrorCleanup(void);
 extern void GetLockHoldersAndWaiters(LOCALLOCK *locallock,
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 966+ messages in thread

* [PATCH] Lock upgrade without deadlocks.
@ 2026-04-17 13:13  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 966+ messages in thread

From: Antonin Houska @ 2026-04-17 13:13 UTC (permalink / raw)

For REPACK (CONCURRENTLY), it's essential to upgrade ShareUpdateExclusive lock
to AccessExclusiveLock for the final stage of table processing. However, we
cannot release the earlier before requesting the latter, because if another
process ran DDL on the table in between, REPACK would have to cancel its
transaction, and waste possibly a lot of work. Without releasing
ShareUpdateExclusive lock, the upgrade can cause deadlock, and it's not really
deterministic whether REPACK or the other process will get canceled.

This patch adds a new field to the LOCK structure which tells which process is
trying to upgrade the lock - this field is set right before the upgrade. All
processes waiting in the lock queue are then woken up to check the flag.

When another process tries to get the lock after that, it checks this field,
and if it's set, it checks for deadlocks even if deadlock timeout hasn't
expired yet. If the process is already in the lock's queue and sleeping, the
lock upgrading process wakes it up so it checks the flag immediately.

At the time the upgrading process starts to sleep, all the other process
should be aware that they need to check for deadlock during their lock
acquisitions, so the upgrading process does not have to check for deadlocks
when it gets woken up. Thus it should never receive deadlock error report.
---
 src/backend/commands/repack.c   |   4 +-
 src/backend/storage/lmgr/lmgr.c |  35 ++++++++---
 src/backend/storage/lmgr/lock.c | 107 +++++++++++++++++++++++++++++++-
 src/backend/storage/lmgr/proc.c |  32 +++++++---
 src/include/storage/lmgr.h      |   1 +
 src/include/storage/lock.h      |   5 +-
 src/include/storage/proc.h      |   4 +-
 7 files changed, 166 insertions(+), 22 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index 58e3867246f..310b2a65099 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -3055,8 +3055,10 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Acquire AccessExclusiveLock on the table, its TOAST relation (if there
 	 * is one), all its indexes, so that we can swap the files.
+	 *
+	 * TODO The same for indexes and TOAST?
 	 */
-	LockRelationOid(old_table_oid, AccessExclusiveLock);
+	LockRelationOidUpgrade(old_table_oid, AccessExclusiveLock);
 
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
diff --git a/src/backend/storage/lmgr/lmgr.c b/src/backend/storage/lmgr/lmgr.c
index 2ccf7237fee..0b103485d54 100644
--- a/src/backend/storage/lmgr/lmgr.c
+++ b/src/backend/storage/lmgr/lmgr.c
@@ -58,6 +58,8 @@ typedef struct XactLockTableWaitInfo
 	const ItemPointerData *ctid;
 } XactLockTableWaitInfo;
 
+static void LockRelationOidCommon(Oid relid, LOCKMODE lockmode,
+								  bool isUpgrade);
 static void XactLockTableWaitErrorCb(void *arg);
 
 /*
@@ -105,6 +107,21 @@ SetLocktagRelationOid(LOCKTAG *tag, Oid relid)
  */
 void
 LockRelationOid(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, false);
+}
+
+/*
+ * Like above, but upgrade an existing lock.
+ */
+void
+LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, true);
+}
+
+static void
+LockRelationOidCommon(Oid relid, LOCKMODE lockmode, bool isUpgrade)
 {
 	LOCKTAG		tag;
 	LOCALLOCK  *locallock;
@@ -113,7 +130,7 @@ LockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, isUpgrade);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages, so that we
@@ -157,7 +174,7 @@ ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -191,7 +208,7 @@ LockRelationId(LockRelId *relid, LOCKMODE lockmode)
 	SET_LOCKTAG_RELATION(tag, relid->dbId, relid->relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -254,7 +271,7 @@ LockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -286,7 +303,7 @@ ConditionalLockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -591,7 +608,7 @@ ConditionalLockTuple(Relation relation, const ItemPointerData *tid, LOCKMODE loc
 					  ItemPointerGetOffsetNumber(tid));
 
 	return (LockAcquireExtended(&tag, lockmode, false, true, true, NULL,
-								logLockFailure) != LOCKACQUIRE_NOT_AVAIL);
+								logLockFailure, false) != LOCKACQUIRE_NOT_AVAIL);
 }
 
 /*
@@ -749,7 +766,7 @@ ConditionalXactLockTableWait(TransactionId xid, bool logLockFailure)
 		SET_LOCKTAG_TRANSACTION(tag, xid);
 
 		if (LockAcquireExtended(&tag, ShareLock, false, true, true, NULL,
-								logLockFailure)
+								logLockFailure, false)
 			== LOCKACQUIRE_NOT_AVAIL)
 			return false;
 
@@ -1043,7 +1060,7 @@ ConditionalLockDatabaseObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -1123,7 +1140,7 @@ ConditionalLockSharedObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
diff --git a/src/backend/storage/lmgr/lock.c b/src/backend/storage/lmgr/lock.c
index c221fe96889..bcbbf36c091 100644
--- a/src/backend/storage/lmgr/lock.c
+++ b/src/backend/storage/lmgr/lock.c
@@ -810,7 +810,7 @@ LockAcquire(const LOCKTAG *locktag,
 			bool dontWait)
 {
 	return LockAcquireExtended(locktag, lockmode, sessionLock, dontWait,
-							   true, NULL, false);
+							   true, NULL, false, false);
 }
 
 /*
@@ -829,6 +829,9 @@ LockAcquire(const LOCKTAG *locktag,
  *
  * logLockFailure indicates whether to log details when a lock acquisition
  * fails with dontWait = true.
+ *
+ * isUpgrade should be true if the backend already holds this lock in a lower
+ * mode.
  */
 LockAcquireResult
 LockAcquireExtended(const LOCKTAG *locktag,
@@ -837,7 +840,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 					bool dontWait,
 					bool reportMemoryError,
 					LOCALLOCK **locallockp,
-					bool logLockFailure)
+					bool logLockFailure,
+					bool isUpgrade)
 {
 	LOCKMETHODID lockmethodid = locktag->locktag_lockmethodid;
 	LockMethod	lockMethodTable;
@@ -1119,7 +1123,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		 * case, because JoinWaitQueue() may discover that we can acquire the
 		 * lock immediately after all.
 		 */
-		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait);
+		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait,
+								   isUpgrade);
 	}
 
 	if (waitResult == PROC_WAIT_STATUS_ERROR)
@@ -1225,8 +1230,85 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		Assert(!dontWait);
 		PROCLOCK_PRINT("LockAcquire: sleeping on lock", proclock);
 		LOCK_PRINT("LockAcquire: sleeping on lock", lock, lockmode);
+
+		/*
+		 * Lock upgrade can introduce deadlock, therefore enforce special
+		 * behavior of other processes that deal with this lock. In
+		 * particular, any waiter that sees upgradedBy set is expected to
+		 * perform deadlock check as soon as it's woken up. (Since we are
+		 * already in the queue, the deadlock detector has all the information
+		 * it needs.)
+		 */
+		if (isUpgrade)
+		{
+			dlist_iter	iter;
+
+			/*
+			 * There should not be multiple upgrades at the same time. XXX
+			 * ERROR ?
+			 */
+			Assert(lock->upgradedBy == NULL);
+
+			lock->upgradedBy = MyProc;
+
+			/*
+			 * Now, before I sleep myself, wake up all the existing waiters
+			 * (except for me), so they check for deadlock.
+			 */
+			dclist_foreach(iter, &lock->waitProcs)
+			{
+				PGPROC	   *proc = dlist_container(PGPROC, waitLink,
+												   iter.cur);
+
+				if (proc != MyProc)
+					SetLatch(&proc->procLatch);
+			}
+		}
 		LWLockRelease(partitionLock);
 
+		if (!isUpgrade)
+		{
+			HASH_SEQ_STATUS status;
+			LOCALLOCK	*llock;
+			bool	check_deadlock = false;
+
+			/*
+			 * Likewise, new waiters should check the isUpgrade flag and
+			 * engage deadlock detector if needed. The lock being acquired
+			 * should already be in the hash.
+			 *
+			 * Note that, to make deadlock detection happen as soon as
+			 * possible (i.e. regardless deadlock timeout), we check the other
+			 * locks too. XXX There might be ways to get into the deadlock
+			 * indirectly, but that needs more analysis. As long as the
+			 * upgrading process skips deadlock detection altogether, the
+			 * worst consequence of missing some lock wait cycle here is that
+			 * the other process will be kicked-off no sooner than after its
+			 * deadlock timeout has elapsed. (Which in turn means that the
+			 * lock upgrade will take longer than expected.)
+			 */
+			hash_seq_init(&status, LockMethodLocalHash);
+			while ((llock = (LOCALLOCK *) hash_seq_search(&status)) != NULL)
+			{
+				LOCK	*mylock = llock->lock;
+
+				if (mylock && mylock->upgradedBy)
+				{
+					check_deadlock = true;
+					break;
+				}
+			}
+			/* Check for deadlock if needed. */
+			if (check_deadlock)
+			{
+				DeadLockState	deadlock_state;
+
+				deadlock_state = CheckDeadLock();
+				if (deadlock_state == DS_HARD_DEADLOCK)
+					DeadLockReport();
+			}
+		}
+
 		waitResult = WaitOnLock(locallock, owner);
 
 		/*
@@ -1245,6 +1327,18 @@ LockAcquireExtended(const LOCKTAG *locktag,
 			DeadLockReport();
 			/* DeadLockReport() will not return */
 		}
+
+		/*
+		 * If finishing the lock upgrade, we should not get into a deadlock
+		 * anymore, so let others know that they do not have to care either.
+		 */
+		if (isUpgrade)
+		{
+			LWLockAcquire(partitionLock, LW_EXCLUSIVE);
+			Assert(lock->upgradedBy != NULL);
+			lock->upgradedBy = NULL;
+			LWLockRelease(partitionLock);
+		}
 	}
 	else
 		LWLockRelease(partitionLock);
@@ -1322,6 +1416,7 @@ SetupLockInTable(LockMethod lockMethodTable, PGPROC *proc,
 		lock->nGranted = 0;
 		MemSet(lock->requested, 0, sizeof(int) * MAX_LOCKMODES);
 		MemSet(lock->granted, 0, sizeof(int) * MAX_LOCKMODES);
+		lock->upgradedBy = NULL;
 		LOCK_PRINT("LockAcquire: new", lock, lockmode);
 	}
 	else
@@ -1768,6 +1863,12 @@ CleanUpLock(LOCK *lock, PROCLOCK *proclock,
 			elog(PANIC, "proclock table corrupted");
 	}
 
+	/*
+	 * Was this backend upgrading the lock?
+	 */
+	if (lock->upgradedBy == MyProc)
+		lock->upgradedBy = NULL;
+
 	if (lock->nRequested == 0)
 	{
 		/*
diff --git a/src/backend/storage/lmgr/proc.c b/src/backend/storage/lmgr/proc.c
index 1ac25068d62..7b31370db70 100644
--- a/src/backend/storage/lmgr/proc.c
+++ b/src/backend/storage/lmgr/proc.c
@@ -95,7 +95,6 @@ static volatile sig_atomic_t got_deadlock_timeout;
 static void RemoveProcFromArray(int code, Datum arg);
 static void ProcKill(int code, Datum arg);
 static void AuxiliaryProcKill(int code, Datum arg);
-static DeadLockState CheckDeadLock(void);
 
 
 /*
@@ -1143,7 +1142,8 @@ AuxiliaryPidGetProc(int pid)
  * NOTES: The process queue is now a priority queue for locking.
  */
 ProcWaitStatus
-JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
+JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait,
+			  bool isUpgrade)
 {
 	LOCKMODE	lockmode = locallock->tag.mode;
 	LOCK	   *lock = locallock->lock;
@@ -1230,8 +1230,12 @@ JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
 			/* Must he wait for me? */
 			if (lockMethodTable->conflictTab[proc->waitLockMode] & myHeldLocks)
 			{
-				/* Must I wait for him ? */
-				if (lockMethodTable->conflictTab[lockmode] & proc->heldLocks)
+				/*
+				 * Must I wait for him? I don't want a deadlock during lock
+				 * upgrade - other processes should fail on it.
+				 */
+				if ((lockMethodTable->conflictTab[lockmode] & proc->heldLocks) &&
+					!isUpgrade)
 				{
 					/*
 					 * Yes, so we have a deadlock.  Easiest way to clean up
@@ -1461,8 +1465,22 @@ ProcSleep(LOCALLOCK *locallock)
 			(void) WaitLatch(MyLatch, WL_LATCH_SET | WL_EXIT_ON_PM_DEATH, 0,
 							 PG_WAIT_LOCK | locallock->tag.lock.locktag_type);
 			ResetLatch(MyLatch);
-			/* check for deadlocks first, as that's probably log-worthy */
-			if (got_deadlock_timeout)
+			/*
+			 * Check for deadlocks first, as that's probably log-worthy.
+			 *
+			 * Do not wait for the timeout if the lock is being upgraded since
+			 * the risk of deadlock is higher now. However, do not check for
+			 * deadlock if the lock is being upgraded by this process - other
+			 * processes should take care.
+			 *
+			 * TODO Possible optimization: if this is the only lock of the
+			 * backend and if it did not have any weaker lock on the table so
+			 * far, it should be safe to skip the deadlock check. However, to
+			 * evaluate the situation, we need to take fast-path locks into
+			 * account.
+			 */
+			if ((got_deadlock_timeout || lock->upgradedBy) &&
+				lock->upgradedBy != MyProc)
 			{
 				deadlock_state = CheckDeadLock();
 				got_deadlock_timeout = false;
@@ -1819,7 +1837,7 @@ ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock)
  * not, just return.  If we have a real deadlock, remove ourselves from the
  * lock's wait queue.
  */
-static DeadLockState
+DeadLockState
 CheckDeadLock(void)
 {
 	int			i;
diff --git a/src/include/storage/lmgr.h b/src/include/storage/lmgr.h
index 2a985ce5e15..eddbe1e8622 100644
--- a/src/include/storage/lmgr.h
+++ b/src/include/storage/lmgr.h
@@ -38,6 +38,7 @@ extern void RelationInitLockInfo(Relation relation);
 
 /* Lock a relation */
 extern void LockRelationOid(Oid relid, LOCKMODE lockmode);
+extern void LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode);
 extern void LockRelationId(LockRelId *relid, LOCKMODE lockmode);
 extern bool ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode);
 extern void UnlockRelationId(LockRelId *relid, LOCKMODE lockmode);
diff --git a/src/include/storage/lock.h b/src/include/storage/lock.h
index ee3cb1dc203..080b69eddfd 100644
--- a/src/include/storage/lock.h
+++ b/src/include/storage/lock.h
@@ -131,6 +131,7 @@ typedef const LockMethodData *LockMethod;
  * nRequested -- total requested locks of all types.
  * granted -- count of each lock type currently granted on the lock.
  * nGranted -- total granted locks of all types.
+ * upgrader -- process that intends to upgrade the lock
  *
  * Note: these counts count 1 for each backend.  Internally to a backend,
  * there may be multiple grabs on a particular lock, but this is not reflected
@@ -150,6 +151,7 @@ typedef struct LOCK
 	int			nRequested;		/* total of requested[] array */
 	int			granted[MAX_LOCKMODES]; /* counts of granted locks */
 	int			nGranted;		/* total of granted[] array */
+	PGPROC		*upgradedBy;	/* is lock being upgraded by this process? */
 } LOCK;
 
 #define LOCK_LOCKMETHOD(lock) ((LOCKMETHODID) (lock).tag.locktag_lockmethodid)
@@ -390,7 +392,8 @@ extern LockAcquireResult LockAcquireExtended(const LOCKTAG *locktag,
 											 bool dontWait,
 											 bool reportMemoryError,
 											 LOCALLOCK **locallockp,
-											 bool logLockFailure);
+											 bool logLockFailure,
+											 bool isUpgrade);
 extern void AbortStrongLockAcquire(void);
 extern void MarkLockClear(LOCALLOCK *locallock);
 extern bool LockRelease(const LOCKTAG *locktag,
diff --git a/src/include/storage/proc.h b/src/include/storage/proc.h
index 3e1d1fad5f9..c4bc3a24044 100644
--- a/src/include/storage/proc.h
+++ b/src/include/storage/proc.h
@@ -563,10 +563,12 @@ extern bool HaveNFreeProcs(int n, int *nfree);
 extern void ProcReleaseLocks(bool isCommit);
 
 extern ProcWaitStatus JoinWaitQueue(LOCALLOCK *locallock,
-									LockMethod lockMethodTable, bool dontWait);
+									LockMethod lockMethodTable, bool dontWait,
+									bool isUpgrade);
 extern ProcWaitStatus ProcSleep(LOCALLOCK *locallock);
 extern void ProcWakeup(PGPROC *proc, ProcWaitStatus waitStatus);
 extern void ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock);
+extern DeadLockState CheckDeadLock(void);
 extern void CheckDeadLockAlert(void);
 extern void LockErrorCleanup(void);
 extern void GetLockHoldersAndWaiters(LOCALLOCK *locallock,
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 966+ messages in thread

* [PATCH] Lock upgrade without deadlocks.
@ 2026-04-17 13:13  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 966+ messages in thread

From: Antonin Houska @ 2026-04-17 13:13 UTC (permalink / raw)

For REPACK (CONCURRENTLY), it's essential to upgrade ShareUpdateExclusive lock
to AccessExclusiveLock for the final stage of table processing. However, we
cannot release the earlier before requesting the latter, because if another
process ran DDL on the table in between, REPACK would have to cancel its
transaction, and waste possibly a lot of work. Without releasing
ShareUpdateExclusive lock, the upgrade can cause deadlock, and it's not really
deterministic whether REPACK or the other process will get canceled.

This patch adds a new field to the LOCK structure which tells which process is
trying to upgrade the lock - this field is set right before the upgrade. All
processes waiting in the lock queue are then woken up to check the flag.

When another process tries to get the lock after that, it checks this field,
and if it's set, it checks for deadlocks even if deadlock timeout hasn't
expired yet. If the process is already in the lock's queue and sleeping, the
lock upgrading process wakes it up so it checks the flag immediately.

At the time the upgrading process starts to sleep, all the other process
should be aware that they need to check for deadlock during their lock
acquisitions, so the upgrading process does not have to check for deadlocks
when it gets woken up. Thus it should never receive deadlock error report.
---
 src/backend/commands/repack.c   |   4 +-
 src/backend/storage/lmgr/lmgr.c |  35 ++++++++---
 src/backend/storage/lmgr/lock.c | 107 +++++++++++++++++++++++++++++++-
 src/backend/storage/lmgr/proc.c |  32 +++++++---
 src/include/storage/lmgr.h      |   1 +
 src/include/storage/lock.h      |   5 +-
 src/include/storage/proc.h      |   4 +-
 7 files changed, 166 insertions(+), 22 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index 58e3867246f..310b2a65099 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -3055,8 +3055,10 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Acquire AccessExclusiveLock on the table, its TOAST relation (if there
 	 * is one), all its indexes, so that we can swap the files.
+	 *
+	 * TODO The same for indexes and TOAST?
 	 */
-	LockRelationOid(old_table_oid, AccessExclusiveLock);
+	LockRelationOidUpgrade(old_table_oid, AccessExclusiveLock);
 
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
diff --git a/src/backend/storage/lmgr/lmgr.c b/src/backend/storage/lmgr/lmgr.c
index 2ccf7237fee..0b103485d54 100644
--- a/src/backend/storage/lmgr/lmgr.c
+++ b/src/backend/storage/lmgr/lmgr.c
@@ -58,6 +58,8 @@ typedef struct XactLockTableWaitInfo
 	const ItemPointerData *ctid;
 } XactLockTableWaitInfo;
 
+static void LockRelationOidCommon(Oid relid, LOCKMODE lockmode,
+								  bool isUpgrade);
 static void XactLockTableWaitErrorCb(void *arg);
 
 /*
@@ -105,6 +107,21 @@ SetLocktagRelationOid(LOCKTAG *tag, Oid relid)
  */
 void
 LockRelationOid(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, false);
+}
+
+/*
+ * Like above, but upgrade an existing lock.
+ */
+void
+LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode)
+{
+	LockRelationOidCommon(relid, lockmode, true);
+}
+
+static void
+LockRelationOidCommon(Oid relid, LOCKMODE lockmode, bool isUpgrade)
 {
 	LOCKTAG		tag;
 	LOCALLOCK  *locallock;
@@ -113,7 +130,7 @@ LockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, isUpgrade);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages, so that we
@@ -157,7 +174,7 @@ ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode)
 	SetLocktagRelationOid(&tag, relid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -191,7 +208,7 @@ LockRelationId(LockRelId *relid, LOCKMODE lockmode)
 	SET_LOCKTAG_RELATION(tag, relid->dbId, relid->relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -254,7 +271,7 @@ LockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock,
-							  false);
+							  false, false);
 
 	/*
 	 * Now that we have the lock, check for invalidation messages; see notes
@@ -286,7 +303,7 @@ ConditionalLockRelation(Relation relation, LOCKMODE lockmode)
 						 relation->rd_lockInfo.lockRelId.relId);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -591,7 +608,7 @@ ConditionalLockTuple(Relation relation, const ItemPointerData *tid, LOCKMODE loc
 					  ItemPointerGetOffsetNumber(tid));
 
 	return (LockAcquireExtended(&tag, lockmode, false, true, true, NULL,
-								logLockFailure) != LOCKACQUIRE_NOT_AVAIL);
+								logLockFailure, false) != LOCKACQUIRE_NOT_AVAIL);
 }
 
 /*
@@ -749,7 +766,7 @@ ConditionalXactLockTableWait(TransactionId xid, bool logLockFailure)
 		SET_LOCKTAG_TRANSACTION(tag, xid);
 
 		if (LockAcquireExtended(&tag, ShareLock, false, true, true, NULL,
-								logLockFailure)
+								logLockFailure, false)
 			== LOCKACQUIRE_NOT_AVAIL)
 			return false;
 
@@ -1043,7 +1060,7 @@ ConditionalLockDatabaseObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
@@ -1123,7 +1140,7 @@ ConditionalLockSharedObject(Oid classid, Oid objid, uint16 objsubid,
 					   objsubid);
 
 	res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock,
-							  false);
+							  false, false);
 
 	if (res == LOCKACQUIRE_NOT_AVAIL)
 		return false;
diff --git a/src/backend/storage/lmgr/lock.c b/src/backend/storage/lmgr/lock.c
index c221fe96889..bcbbf36c091 100644
--- a/src/backend/storage/lmgr/lock.c
+++ b/src/backend/storage/lmgr/lock.c
@@ -810,7 +810,7 @@ LockAcquire(const LOCKTAG *locktag,
 			bool dontWait)
 {
 	return LockAcquireExtended(locktag, lockmode, sessionLock, dontWait,
-							   true, NULL, false);
+							   true, NULL, false, false);
 }
 
 /*
@@ -829,6 +829,9 @@ LockAcquire(const LOCKTAG *locktag,
  *
  * logLockFailure indicates whether to log details when a lock acquisition
  * fails with dontWait = true.
+ *
+ * isUpgrade should be true if the backend already holds this lock in a lower
+ * mode.
  */
 LockAcquireResult
 LockAcquireExtended(const LOCKTAG *locktag,
@@ -837,7 +840,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 					bool dontWait,
 					bool reportMemoryError,
 					LOCALLOCK **locallockp,
-					bool logLockFailure)
+					bool logLockFailure,
+					bool isUpgrade)
 {
 	LOCKMETHODID lockmethodid = locktag->locktag_lockmethodid;
 	LockMethod	lockMethodTable;
@@ -1119,7 +1123,8 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		 * case, because JoinWaitQueue() may discover that we can acquire the
 		 * lock immediately after all.
 		 */
-		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait);
+		waitResult = JoinWaitQueue(locallock, lockMethodTable, dontWait,
+								   isUpgrade);
 	}
 
 	if (waitResult == PROC_WAIT_STATUS_ERROR)
@@ -1225,8 +1230,85 @@ LockAcquireExtended(const LOCKTAG *locktag,
 		Assert(!dontWait);
 		PROCLOCK_PRINT("LockAcquire: sleeping on lock", proclock);
 		LOCK_PRINT("LockAcquire: sleeping on lock", lock, lockmode);
+
+		/*
+		 * Lock upgrade can introduce deadlock, therefore enforce special
+		 * behavior of other processes that deal with this lock. In
+		 * particular, any waiter that sees upgradedBy set is expected to
+		 * perform deadlock check as soon as it's woken up. (Since we are
+		 * already in the queue, the deadlock detector has all the information
+		 * it needs.)
+		 */
+		if (isUpgrade)
+		{
+			dlist_iter	iter;
+
+			/*
+			 * There should not be multiple upgrades at the same time. XXX
+			 * ERROR ?
+			 */
+			Assert(lock->upgradedBy == NULL);
+
+			lock->upgradedBy = MyProc;
+
+			/*
+			 * Now, before I sleep myself, wake up all the existing waiters
+			 * (except for me), so they check for deadlock.
+			 */
+			dclist_foreach(iter, &lock->waitProcs)
+			{
+				PGPROC	   *proc = dlist_container(PGPROC, waitLink,
+												   iter.cur);
+
+				if (proc != MyProc)
+					SetLatch(&proc->procLatch);
+			}
+		}
 		LWLockRelease(partitionLock);
 
+		if (!isUpgrade)
+		{
+			HASH_SEQ_STATUS status;
+			LOCALLOCK	*llock;
+			bool	check_deadlock = false;
+
+			/*
+			 * Likewise, new waiters should check the isUpgrade flag and
+			 * engage deadlock detector if needed. The lock being acquired
+			 * should already be in the hash.
+			 *
+			 * Note that, to make deadlock detection happen as soon as
+			 * possible (i.e. regardless deadlock timeout), we check the other
+			 * locks too. XXX There might be ways to get into the deadlock
+			 * indirectly, but that needs more analysis. As long as the
+			 * upgrading process skips deadlock detection altogether, the
+			 * worst consequence of missing some lock wait cycle here is that
+			 * the other process will be kicked-off no sooner than after its
+			 * deadlock timeout has elapsed. (Which in turn means that the
+			 * lock upgrade will take longer than expected.)
+			 */
+			hash_seq_init(&status, LockMethodLocalHash);
+			while ((llock = (LOCALLOCK *) hash_seq_search(&status)) != NULL)
+			{
+				LOCK	*mylock = llock->lock;
+
+				if (mylock && mylock->upgradedBy)
+				{
+					check_deadlock = true;
+					break;
+				}
+			}
+			/* Check for deadlock if needed. */
+			if (check_deadlock)
+			{
+				DeadLockState	deadlock_state;
+
+				deadlock_state = CheckDeadLock();
+				if (deadlock_state == DS_HARD_DEADLOCK)
+					DeadLockReport();
+			}
+		}
+
 		waitResult = WaitOnLock(locallock, owner);
 
 		/*
@@ -1245,6 +1327,18 @@ LockAcquireExtended(const LOCKTAG *locktag,
 			DeadLockReport();
 			/* DeadLockReport() will not return */
 		}
+
+		/*
+		 * If finishing the lock upgrade, we should not get into a deadlock
+		 * anymore, so let others know that they do not have to care either.
+		 */
+		if (isUpgrade)
+		{
+			LWLockAcquire(partitionLock, LW_EXCLUSIVE);
+			Assert(lock->upgradedBy != NULL);
+			lock->upgradedBy = NULL;
+			LWLockRelease(partitionLock);
+		}
 	}
 	else
 		LWLockRelease(partitionLock);
@@ -1322,6 +1416,7 @@ SetupLockInTable(LockMethod lockMethodTable, PGPROC *proc,
 		lock->nGranted = 0;
 		MemSet(lock->requested, 0, sizeof(int) * MAX_LOCKMODES);
 		MemSet(lock->granted, 0, sizeof(int) * MAX_LOCKMODES);
+		lock->upgradedBy = NULL;
 		LOCK_PRINT("LockAcquire: new", lock, lockmode);
 	}
 	else
@@ -1768,6 +1863,12 @@ CleanUpLock(LOCK *lock, PROCLOCK *proclock,
 			elog(PANIC, "proclock table corrupted");
 	}
 
+	/*
+	 * Was this backend upgrading the lock?
+	 */
+	if (lock->upgradedBy == MyProc)
+		lock->upgradedBy = NULL;
+
 	if (lock->nRequested == 0)
 	{
 		/*
diff --git a/src/backend/storage/lmgr/proc.c b/src/backend/storage/lmgr/proc.c
index 1ac25068d62..7b31370db70 100644
--- a/src/backend/storage/lmgr/proc.c
+++ b/src/backend/storage/lmgr/proc.c
@@ -95,7 +95,6 @@ static volatile sig_atomic_t got_deadlock_timeout;
 static void RemoveProcFromArray(int code, Datum arg);
 static void ProcKill(int code, Datum arg);
 static void AuxiliaryProcKill(int code, Datum arg);
-static DeadLockState CheckDeadLock(void);
 
 
 /*
@@ -1143,7 +1142,8 @@ AuxiliaryPidGetProc(int pid)
  * NOTES: The process queue is now a priority queue for locking.
  */
 ProcWaitStatus
-JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
+JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait,
+			  bool isUpgrade)
 {
 	LOCKMODE	lockmode = locallock->tag.mode;
 	LOCK	   *lock = locallock->lock;
@@ -1230,8 +1230,12 @@ JoinWaitQueue(LOCALLOCK *locallock, LockMethod lockMethodTable, bool dontWait)
 			/* Must he wait for me? */
 			if (lockMethodTable->conflictTab[proc->waitLockMode] & myHeldLocks)
 			{
-				/* Must I wait for him ? */
-				if (lockMethodTable->conflictTab[lockmode] & proc->heldLocks)
+				/*
+				 * Must I wait for him? I don't want a deadlock during lock
+				 * upgrade - other processes should fail on it.
+				 */
+				if ((lockMethodTable->conflictTab[lockmode] & proc->heldLocks) &&
+					!isUpgrade)
 				{
 					/*
 					 * Yes, so we have a deadlock.  Easiest way to clean up
@@ -1461,8 +1465,22 @@ ProcSleep(LOCALLOCK *locallock)
 			(void) WaitLatch(MyLatch, WL_LATCH_SET | WL_EXIT_ON_PM_DEATH, 0,
 							 PG_WAIT_LOCK | locallock->tag.lock.locktag_type);
 			ResetLatch(MyLatch);
-			/* check for deadlocks first, as that's probably log-worthy */
-			if (got_deadlock_timeout)
+			/*
+			 * Check for deadlocks first, as that's probably log-worthy.
+			 *
+			 * Do not wait for the timeout if the lock is being upgraded since
+			 * the risk of deadlock is higher now. However, do not check for
+			 * deadlock if the lock is being upgraded by this process - other
+			 * processes should take care.
+			 *
+			 * TODO Possible optimization: if this is the only lock of the
+			 * backend and if it did not have any weaker lock on the table so
+			 * far, it should be safe to skip the deadlock check. However, to
+			 * evaluate the situation, we need to take fast-path locks into
+			 * account.
+			 */
+			if ((got_deadlock_timeout || lock->upgradedBy) &&
+				lock->upgradedBy != MyProc)
 			{
 				deadlock_state = CheckDeadLock();
 				got_deadlock_timeout = false;
@@ -1819,7 +1837,7 @@ ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock)
  * not, just return.  If we have a real deadlock, remove ourselves from the
  * lock's wait queue.
  */
-static DeadLockState
+DeadLockState
 CheckDeadLock(void)
 {
 	int			i;
diff --git a/src/include/storage/lmgr.h b/src/include/storage/lmgr.h
index 2a985ce5e15..eddbe1e8622 100644
--- a/src/include/storage/lmgr.h
+++ b/src/include/storage/lmgr.h
@@ -38,6 +38,7 @@ extern void RelationInitLockInfo(Relation relation);
 
 /* Lock a relation */
 extern void LockRelationOid(Oid relid, LOCKMODE lockmode);
+extern void LockRelationOidUpgrade(Oid relid, LOCKMODE lockmode);
 extern void LockRelationId(LockRelId *relid, LOCKMODE lockmode);
 extern bool ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode);
 extern void UnlockRelationId(LockRelId *relid, LOCKMODE lockmode);
diff --git a/src/include/storage/lock.h b/src/include/storage/lock.h
index ee3cb1dc203..080b69eddfd 100644
--- a/src/include/storage/lock.h
+++ b/src/include/storage/lock.h
@@ -131,6 +131,7 @@ typedef const LockMethodData *LockMethod;
  * nRequested -- total requested locks of all types.
  * granted -- count of each lock type currently granted on the lock.
  * nGranted -- total granted locks of all types.
+ * upgrader -- process that intends to upgrade the lock
  *
  * Note: these counts count 1 for each backend.  Internally to a backend,
  * there may be multiple grabs on a particular lock, but this is not reflected
@@ -150,6 +151,7 @@ typedef struct LOCK
 	int			nRequested;		/* total of requested[] array */
 	int			granted[MAX_LOCKMODES]; /* counts of granted locks */
 	int			nGranted;		/* total of granted[] array */
+	PGPROC		*upgradedBy;	/* is lock being upgraded by this process? */
 } LOCK;
 
 #define LOCK_LOCKMETHOD(lock) ((LOCKMETHODID) (lock).tag.locktag_lockmethodid)
@@ -390,7 +392,8 @@ extern LockAcquireResult LockAcquireExtended(const LOCKTAG *locktag,
 											 bool dontWait,
 											 bool reportMemoryError,
 											 LOCALLOCK **locallockp,
-											 bool logLockFailure);
+											 bool logLockFailure,
+											 bool isUpgrade);
 extern void AbortStrongLockAcquire(void);
 extern void MarkLockClear(LOCALLOCK *locallock);
 extern bool LockRelease(const LOCKTAG *locktag,
diff --git a/src/include/storage/proc.h b/src/include/storage/proc.h
index 3e1d1fad5f9..c4bc3a24044 100644
--- a/src/include/storage/proc.h
+++ b/src/include/storage/proc.h
@@ -563,10 +563,12 @@ extern bool HaveNFreeProcs(int n, int *nfree);
 extern void ProcReleaseLocks(bool isCommit);
 
 extern ProcWaitStatus JoinWaitQueue(LOCALLOCK *locallock,
-									LockMethod lockMethodTable, bool dontWait);
+									LockMethod lockMethodTable, bool dontWait,
+									bool isUpgrade);
 extern ProcWaitStatus ProcSleep(LOCALLOCK *locallock);
 extern void ProcWakeup(PGPROC *proc, ProcWaitStatus waitStatus);
 extern void ProcLockWakeup(LockMethod lockMethodTable, LOCK *lock);
+extern DeadLockState CheckDeadLock(void);
 extern void CheckDeadLockAlert(void);
 extern void LockErrorCleanup(void);
 extern void GetLockHoldersAndWaiters(LOCALLOCK *locallock,
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 966+ messages in thread