executor: fix data race in IndexNestedLoopHashJoin (#55824) (#55911)

close #49692

executor/index_lookup_hash_join.go

@@ -19,6 +19,7 @@ import (
 	"fmt"
 	"hash"
 	"hash/fnv"
+	"runtime"
 	"runtime/trace"
 	"sync"
 	"sync/atomic"
@@ -74,7 +75,11 @@ type IndexNestedLoopHashJoin struct {
 	prepared bool
 	// panicErr records the error generated by panic recover. This is introduced to
 	// return the actual error message instead of `context cancelled` to the client.
-	panicErr      error
+	panicErr struct {
+		sync.Mutex
+		atomic.Bool
+		error
+	}
 	ctxWithCancel context.Context
 }
 
@@ -186,13 +191,22 @@ func (e *IndexNestedLoopHashJoin) finishJoinWorkers(r interface{}) {
 	if r != nil {
 		e.IndexLookUpJoin.finished.Store(true)
 		err := errors.New(fmt.Sprintf("%v", r))
+
+		if !e.panicErr.Load() {
+			e.panicErr.Lock()
+			if !e.panicErr.Load() {
+				e.panicErr.error = err
+				e.panicErr.Store(true)
+			}
+			e.panicErr.Unlock()
+		}
+
 		if !e.keepOuterOrder {
 			e.resultCh <- &indexHashJoinResult{err: err}
 		} else {
 			task := &indexHashJoinTask{err: err}
 			e.taskCh <- task
 		}
-		e.panicErr = err
 		if e.cancelFunc != nil {
 			e.cancelFunc()
 		}
@@ -226,7 +240,10 @@ func (e *IndexNestedLoopHashJoin) Next(ctx context.Context, req *chunk.Chunk) er
 func (e *IndexNestedLoopHashJoin) runInOrder(ctx context.Context, req *chunk.Chunk) error {
 	for {
 		if e.isDryUpTasks(ctx) {
-			return e.panicErr
+			if e.panicErr.Load() {
+				return e.panicErr.error
+			}
+			return nil
 		}
 		if e.curTask.err != nil {
 			return e.curTask.err
@@ -282,7 +299,17 @@ func (e *IndexNestedLoopHashJoin) getResultFromChannel(ctx context.Context, resu
 			return nil, result.err
 		}
 	case <-ctx.Done():
-		err := e.panicErr
+		failpoint.Inject("TestIssue49692", func() {
+			for !e.panicErr.Load() {
+				runtime.Gosched()
+			}
+		})
+
+		err := error(nil)
+		if e.panicErr.Load() {
+			err = e.panicErr.error
+		}
+
 		if err == nil {
 			err = ctx.Err()
 		}

executor/join_test.go

@@ -2751,6 +2751,12 @@ func TestIssue30211(t *testing.T) {
 	tk.MustExec("drop table if exists t1, t2;")
 	tk.MustExec("create table t1(a int, index(a));")
 	tk.MustExec("create table t2(a int, index(a));")
+	fpName2 := "github.com/pingcap/tidb/executor/TestIssue49692"
+	require.NoError(t, failpoint.Enable(fpName2, `return`))
+	defer func() {
+		require.NoError(t, failpoint.Disable(fpName2))
+	}()
+
 	func() {
 		fpName := "github.com/pingcap/tidb/executor/TestIssue30211"
 		require.NoError(t, failpoint.Enable(fpName, `panic("TestIssue30211 IndexJoinPanic")`))