!1576 修复多事务并发对同一元组更新加锁的xlog回放导致的数据不一致问题

Merge pull request !1576 from chenxiaobin/fixRelay

src/gausskernel/storage/access/heap/heapam.cpp

@@ -6779,7 +6779,7 @@ failed:
         xlrec.locking_xid = xid;
         xlrec.offnum = ItemPointerGetOffsetNumber(&tuple->t_self);
         xlrec.xid_is_mxact = ((new_infomask & HEAP_XMAX_IS_MULTI) != 0);
-        xlrec.shared_lock = (mode == LockTupleShared);
+        xlrec.shared_lock = (mode == LockTupleShared || mode == LockTupleKeyShare);
         xlrec.infobits_set = ComputeInfobits(new_infomask, tuple->t_data->t_infomask2);
         xlrec.lock_updated = false;
         useOldXlog = t_thrd.proc->workingVersionNum < ENHANCED_TUPLE_LOCK_VERSION_NUM ||
@@ -7321,10 +7321,8 @@ l4:
             xlrec.xid_is_mxact = ((new_infomask & HEAP_XMAX_IS_MULTI) != 0);
             xlrec.infobits_set = ComputeInfobits(new_infomask, new_infomask2);
             xlrec.lock_updated = true;
-            /*
-             * We don't record shared_lock, this field is reserverd for compatibility
-             * This xlog can only be record in new version
-             */
+            xlrec.shared_lock = (mode == LockTupleShared || mode == LockTupleKeyShare);
+
             XLogRegisterData((char *)&xlrec, SizeOfHeapLock);
 
             recptr = XLogInsert(RM_HEAP_ID, XLOG_HEAP_LOCK | XLOG_TUPLE_LOCK_UPGRADE_FLAG);

src/gausskernel/storage/access/redo/redo_heapam.cpp

@@ -725,11 +725,17 @@ void HeapXlogLockOperatorPage(RedoBufferInfo *buffer, void *recorddata, bool isT
             htup->t_infomask2 |= HEAP_KEYS_UPDATED;
         }
     }
-    HeapTupleHeaderClearHotUpdated(htup);
+    /*
+     * Clear relevant update flags, but only if the modified infomask says
+     * there's no update.
+     */
+    if (HEAP_XMAX_IS_LOCKED_ONLY(htup->t_infomask, htup->t_infomask2)) {
+        HeapTupleHeaderClearHotUpdated(htup);
+        /* Make sure there is no forward chain link in t_ctid */
+        ItemPointerSet(&htup->t_ctid, buffer->blockinfo.blkno, xlrec->offnum);
+    }
     HeapTupleHeaderSetXmax(page, htup, xlrec->locking_xid);
     HeapTupleHeaderSetCmax(htup, FirstCommandId, false);
-    /* Make sure there is no forward chain link in t_ctid */
-    ItemPointerSet(&htup->t_ctid, buffer->blockinfo.blkno, xlrec->offnum);
 
     PageSetLSN(page, buffer->lsn);
 }

src/test/ha/ha_schedule_multi_single

@@ -12,3 +12,4 @@ multi_standby_single/xlog_redo_apply_delay
 #multi_standby_single/most_available
 multi_standby_single/failover_with_data
 multi_standby_single/hash_index
+multi_standby_single/consistency.sh

src/test/ha/testcase/multi_standby_single/consistency.sh

@@ -0,0 +1,48 @@
+#!/bin/sh
+
+source ./util.sh
+
+function test_1()
+{
+  set_default
+
+  echo "check 1-sync slaves"
+  check_synchronous_commit "datanode1" 1
+
+  echo "create table consistency_t1 (a int primary key, b int);
+insert into consistency_t1 values (1, 1);
+\parallel on 2
+begin
+update consistency_t1 set b = b + 1 where a = 1;
+perform pg_sleep(2);
+end;
+/
+begin
+perform pg_sleep(1);
+perform * from consistency_t1 where a = 1 for key share;
+perform pg_sleep(3);
+end;
+/
+\parallel off" > consistency_tmp.sql
+
+  gsql -d $db -p $dn1_primary_port -f consistency_tmp.sql
+
+  rm consistency_tmp.sql
+
+  if [ $(gsql -d $db -p $dn1_standby_port -c "select count(*) from consistency_t1 where a = 1;" | grep 0 |wc -l) -eq 0 ]; then
+    echo "consistency check success on dn1_standby"
+  else
+    echo "consistency check failed on dn1_standby"
+    exit 1
+  fi
+
+}
+
+function tear_down() {
+  sleep 1
+  set_default
+  gsql -d $db -p $dn1_primary_port -c "DROP TABLE if exists consistency_t1;"
+}
+
+test_1
+tear_down