Predicate locking in hash indexes.

Hash index searches acquire predicate locks on the primary
page of a bucket. It acquires a lock on both the old and new buckets
for scans that happen concurrently with page splits. During a bucket
split, a predicate lock is copied from the primary page of an old
bucket to the primary page of a new bucket.

Author: Shubham Barai, Amit Kapila
Reviewed by: Amit Kapila, Alexander Korotkov, Thomas Munro
Discussion: https://www.postgresql.org/message-id/flat/CALxAEPvNsM2GTiXdRgaaZ1Pjd1bs+sxfFsf7Ytr+iq+5JJoYXA@mail.gmail.com

Author: feodor

src/backend/access/hash/hash.c

@@ -68,7 +68,7 @@ hashhandler(PG_FUNCTION_ARGS)
 	amroutine->amsearchnulls = false;
 	amroutine->amstorage = false;
 	amroutine->amclusterable = false;
-	amroutine->ampredlocks = false;
+	amroutine->ampredlocks = true;
 	amroutine->amcanparallel = false;
 	amroutine->amkeytype = INT4OID;
 

src/backend/access/hash/hashinsert.c

@@ -22,6 +22,7 @@
 #include "utils/rel.h"
 #include "storage/lwlock.h"
 #include "storage/buf_internals.h"
+#include "storage/predicate.h"
 
 static void _hash_vacuum_one_page(Relation rel, Buffer metabuf, Buffer buf,
 					  RelFileNode hnode);
@@ -88,6 +89,8 @@ _hash_doinsert(Relation rel, IndexTuple itup, Relation heapRel)
 										  &usedmetap);
 	Assert(usedmetap != NULL);
 
+	CheckForSerializableConflictIn(rel, NULL, buf);
+
 	/* remember the primary bucket buffer to release the pin on it at end. */
 	bucket_buf = buf;
 

src/backend/access/hash/hashpage.c

@@ -33,6 +33,7 @@
 #include "miscadmin.h"
 #include "storage/lmgr.h"
 #include "storage/smgr.h"
+#include "storage/predicate.h"
 
 
 static bool _hash_alloc_buckets(Relation rel, BlockNumber firstblock,
@@ -1107,6 +1108,11 @@ _hash_splitbucket(Relation rel,
 	npage = BufferGetPage(nbuf);
 	nopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
 
+	/* Copy the predicate locks from old bucket to new bucket. */
+	PredicateLockPageSplit(rel,
+						   BufferGetBlockNumber(bucket_obuf),
+						   BufferGetBlockNumber(bucket_nbuf));
+
 	/*
 	 * Partition the tuples in the old bucket between the old bucket and the
 	 * new bucket, advancing along the old bucket's overflow bucket chain and

src/backend/access/hash/hashsearch.c

@@ -19,6 +19,7 @@
 #include "miscadmin.h"
 #include "pgstat.h"
 #include "utils/rel.h"
+#include "storage/predicate.h"
 
 static bool _hash_readpage(IndexScanDesc scan, Buffer *bufP,
 			   ScanDirection dir);
@@ -171,6 +172,7 @@ _hash_readnext(IndexScanDesc scan,
 		Assert(BufferIsValid(*bufp));
 
 		LockBuffer(*bufp, BUFFER_LOCK_SHARE);
+		PredicateLockPage(rel, BufferGetBlockNumber(*bufp), scan->xs_snapshot);
 
 		/*
 		 * setting hashso_buc_split to true indicates that we are scanning
@@ -347,6 +349,7 @@ _hash_first(IndexScanDesc scan, ScanDirection dir)
 	so->hashso_sk_hash = hashkey;
 
 	buf = _hash_getbucketbuf_from_hashkey(rel, hashkey, HASH_READ, NULL);
+	PredicateLockPage(rel, BufferGetBlockNumber(buf), scan->xs_snapshot);
 	page = BufferGetPage(buf);
 	TestForOldSnapshot(scan->xs_snapshot, rel, page);
 	opaque = (HashPageOpaque) PageGetSpecialPointer(page);

src/backend/storage/lmgr/README-SSI

@@ -389,6 +389,13 @@ relation is required. Fast update postpones the insertion of tuples into index
 structure by temporarily storing them into pending list. That makes us unable
 to detect r-w conflicts using page-level locks.
 
+    * Hash index searches acquire predicate locks on the primary
+page of a bucket. It acquires a lock on both the old and new buckets
+for scans that happen concurrently with page splits. During a bucket
+split, a predicate lock is copied from the primary page of an old
+bucket to the primary page of a new bucket.
+
+
     * The effects of page splits, overflows, consolidations, and
 removals must be carefully reviewed to ensure that predicate locks
 aren't "lost" during those operations, or kept with pages which could