diff --git a/br/pkg/lightning/config/BUILD.bazel b/br/pkg/lightning/config/BUILD.bazel
index dbb7d17d6e317..40a265ebeaf2d 100644
--- a/br/pkg/lightning/config/BUILD.bazel
+++ b/br/pkg/lightning/config/BUILD.bazel
@@ -43,7 +43,7 @@ go_test(
     ],
     embed = [":config"],
     flaky = True,
-    shard_count = 47,
+    shard_count = 48,
     deps = [
         "//br/pkg/lightning/common",
         "@com_github_burntsushi_toml//:toml",
diff --git a/br/pkg/lightning/config/config.go b/br/pkg/lightning/config/config.go
index 2ab49db444285..329f564ddee7d 100644
--- a/br/pkg/lightning/config/config.go
+++ b/br/pkg/lightning/config/config.go
@@ -870,6 +870,11 @@ func (m *MydumperRuntime) adjust() error {
 	if err := m.CSV.adjust(); err != nil {
 		return err
 	}
+	if m.StrictFormat && len(m.CSV.Terminator) == 0 {
+		return common.ErrInvalidConfig.GenWithStack(
+			`mydumper.strict-format can not be used with empty mydumper.csv.terminator. Please set mydumper.csv.terminator to a non-empty value like "\r\n"`)
+	}
+
 	for _, rule := range m.FileRouters {
 		if filepath.IsAbs(rule.Path) {
 			relPath, err := filepath.Rel(m.SourceDir, rule.Path)
diff --git a/br/pkg/lightning/config/config_test.go b/br/pkg/lightning/config/config_test.go
index 0452121c1cb3c..d148b471b549f 100644
--- a/br/pkg/lightning/config/config_test.go
+++ b/br/pkg/lightning/config/config_test.go
@@ -82,6 +82,30 @@ func TestAdjustPdAddrAndPort(t *testing.T) {
 	require.Equal(t, "123.45.67.89:1234", cfg.TiDB.PdAddr)
 }
 
+func TestStrictFormat(t *testing.T) {
+	ts, host, port := startMockServer(t, http.StatusOK,
+		`{"port":4444,"advertise-address":"","path":"123.45.67.89:1234,56.78.90.12:3456"}`,
+	)
+	defer ts.Close()
+
+	cfg := NewConfig()
+	cfg.TiDB.Host = host
+	cfg.TiDB.StatusPort = port
+	cfg.Mydumper.SourceDir = "."
+	cfg.TikvImporter.Backend = BackendLocal
+	cfg.TikvImporter.SortedKVDir = "."
+	cfg.TiDB.DistSQLScanConcurrency = 1
+	cfg.Mydumper.StrictFormat = true
+
+	err := cfg.Adjust(context.Background())
+	require.ErrorContains(t, err, "mydumper.strict-format can not be used with empty mydumper.csv.terminator")
+	t.Log(err.Error())
+
+	cfg.Mydumper.CSV.Terminator = "\r\n"
+	err = cfg.Adjust(context.Background())
+	require.NoError(t, err)
+}
+
 func TestPausePDSchedulerScope(t *testing.T) {
 	ts, host, port := startMockServer(t, http.StatusOK,
 		`{"port":4444,"advertise-address":"","path":"123.45.67.89:1234,56.78.90.12:3456"}`,
diff --git a/br/pkg/lightning/importer/chunk_process.go b/br/pkg/lightning/importer/chunk_process.go
index f8d56c823e595..c9510548b866b 100644
--- a/br/pkg/lightning/importer/chunk_process.go
+++ b/br/pkg/lightning/importer/chunk_process.go
@@ -724,7 +724,7 @@ func (cr *chunkProcessor) deliverLoop(
 					rc.status.FinishedFileSize.Add(delta)
 				}
 			} else {
-				deliverLogger.Warn("offset go back", zap.Int64("curr", highOffset),
+				deliverLogger.Error("offset go back", zap.Int64("curr", highOffset),
 					zap.Int64("start", lowOffset))
 			}
 		}
diff --git a/br/pkg/lightning/mydump/csv_parser.go b/br/pkg/lightning/mydump/csv_parser.go
index 5d202689a0884..5be4a0a9a5c59 100644
--- a/br/pkg/lightning/mydump/csv_parser.go
+++ b/br/pkg/lightning/mydump/csv_parser.go
@@ -98,7 +98,8 @@ type field struct {
 	quoted  bool
 }
 
-// NewCSVParser creates a CSV parser.
+// NewCSVParser creates a CSV parser. The ownership of the reader is transferred
+// to the parser.
 func NewCSVParser(
 	ctx context.Context,
 	cfg *config.CSVConfig,
diff --git a/br/pkg/lightning/mydump/region.go b/br/pkg/lightning/mydump/region.go
index 0e24530f2364e..dd5a48075f35e 100644
--- a/br/pkg/lightning/mydump/region.go
+++ b/br/pkg/lightning/mydump/region.go
@@ -417,6 +417,7 @@ func SplitLargeCSV(
 		// Create a utf8mb4 convertor to encode and decode data with the charset of CSV files.
 		charsetConvertor, err := NewCharsetConvertor(cfg.DataCharacterSet, cfg.DataInvalidCharReplace)
 		if err != nil {
+			_ = r.Close()
 			return nil, nil, err
 		}
 		parser, err := NewCSVParser(ctx, &cfg.CSV, r, cfg.ReadBlockSize, cfg.IOWorkers, true, charsetConvertor)
@@ -424,6 +425,7 @@ func SplitLargeCSV(
 			return nil, nil, err
 		}
 		if err = parser.ReadColumns(); err != nil {
+			_ = parser.Close()
 			return nil, nil, err
 		}
 		if cfg.CSV.HeaderSchemaMatch {
@@ -434,6 +436,7 @@ func SplitLargeCSV(
 		if endOffset > dataFile.FileMeta.FileSize {
 			endOffset = dataFile.FileMeta.FileSize
 		}
+		_ = parser.Close()
 	}
 	divisor := int64(cfg.ColumnCnt)
 	for {
@@ -447,6 +450,7 @@ func SplitLargeCSV(
 			// Create a utf8mb4 convertor to encode and decode data with the charset of CSV files.
 			charsetConvertor, err := NewCharsetConvertor(cfg.DataCharacterSet, cfg.DataInvalidCharReplace)
 			if err != nil {
+				_ = r.Close()
 				return nil, nil, err
 			}
 			parser, err := NewCSVParser(ctx, &cfg.CSV, r, cfg.ReadBlockSize, cfg.IOWorkers, false, charsetConvertor)
@@ -454,11 +458,13 @@ func SplitLargeCSV(
 				return nil, nil, err
 			}
 			if err = parser.SetPos(endOffset, 0); err != nil {
+				_ = parser.Close()
 				return nil, nil, err
 			}
 			_, pos, err := parser.ReadUntilTerminator()
 			if err != nil {
 				if !errors.ErrorEqual(err, io.EOF) {
+					_ = parser.Close()
 					return nil, nil, err
 				}
 				log.FromContext(ctx).Warn("file contains no terminator at end",
@@ -467,7 +473,7 @@ func SplitLargeCSV(
 				pos = dataFile.FileMeta.FileSize
 			}
 			endOffset = pos
-			parser.Close()
+			_ = parser.Close()
 		}
 		regions = append(regions,
 			&TableRegion{
diff --git a/br/pkg/lightning/mydump/region_test.go b/br/pkg/lightning/mydump/region_test.go
index d947f1783242c..032eefe77badf 100644
--- a/br/pkg/lightning/mydump/region_test.go
+++ b/br/pkg/lightning/mydump/region_test.go
@@ -487,3 +487,98 @@ func TestSplitLargeFileOnlyOneChunk(t *testing.T) {
 		require.Equal(t, columns, regions[i].Chunk.Columns)
 	}
 }
+
+func TestSplitLargeFileSeekInsideCRLF(t *testing.T) {
+	ctx := context.Background()
+	meta := &MDTableMeta{
+		DB:   "csv",
+		Name: "large_csv_seek_inside_crlf",
+	}
+
+	dir := t.TempDir()
+
+	fileName := "test.csv"
+	filePath := filepath.Join(dir, fileName)
+
+	content := []byte("1\r\n2\r\n3\r\n4\r\n")
+	err := os.WriteFile(filePath, content, 0o644)
+	require.NoError(t, err)
+
+	dataFileInfo, err := os.Stat(filePath)
+	require.NoError(t, err)
+	fileSize := dataFileInfo.Size()
+	fileInfo := FileInfo{FileMeta: SourceFileMeta{Path: fileName, Type: SourceTypeCSV, FileSize: fileSize}}
+	ioWorker := worker.NewPool(context.Background(), 4, "io")
+
+	store, err := storage.NewLocalStorage(dir)
+	require.NoError(t, err)
+
+	// if we don't set terminator, it will get the wrong result
+
+	cfg := &config.Config{
+		Mydumper: config.MydumperRuntime{
+			ReadBlockSize: config.ReadBlockSize,
+			CSV: config.CSVConfig{
+				Separator: ",",
+			},
+			StrictFormat:  true,
+			Filter:        []string{"*.*"},
+			MaxRegionSize: 2,
+		},
+	}
+	divideConfig := NewDataDivideConfig(cfg, 1, ioWorker, store, meta)
+
+	// in fact this is the wrong result, just to show the bug. pos mismatch with
+	// offsets. and we might read more rows than expected because we use == rather
+	// than >= to stop reading.
+	offsets := [][]int64{{0, 3}, {3, 6}, {6, 9}, {9, 12}}
+	pos := []int64{2, 5, 8, 11}
+
+	regions, _, err := SplitLargeCSV(context.Background(), divideConfig, fileInfo)
+	require.NoError(t, err)
+	require.Len(t, regions, len(offsets))
+	for i := range offsets {
+		require.Equal(t, offsets[i][0], regions[i].Chunk.Offset)
+		require.Equal(t, offsets[i][1], regions[i].Chunk.EndOffset)
+	}
+
+	file, err := os.Open(filePath)
+	require.NoError(t, err)
+	parser, err := NewCSVParser(ctx, &cfg.Mydumper.CSV, file, 128, ioWorker, false, nil)
+	require.NoError(t, err)
+
+	for parser.ReadRow() == nil {
+		p, _ := parser.Pos()
+		require.Equal(t, pos[0], p)
+		pos = pos[1:]
+	}
+	require.NoError(t, parser.Close())
+
+	// set terminator to "\r\n"
+
+	cfg.Mydumper.CSV.Terminator = "\r\n"
+	divideConfig = NewDataDivideConfig(cfg, 1, ioWorker, store, meta)
+	// pos is contained in expectedOffsets
+	expectedOffsets := [][]int64{{0, 6}, {6, 12}}
+	pos = []int64{3, 6, 9, 12}
+
+	regions, _, err = SplitLargeCSV(context.Background(), divideConfig, fileInfo)
+	require.NoError(t, err)
+	require.Len(t, regions, len(expectedOffsets))
+	for i := range expectedOffsets {
+		require.Equal(t, expectedOffsets[i][0], regions[i].Chunk.Offset)
+		require.Equal(t, expectedOffsets[i][1], regions[i].Chunk.EndOffset)
+	}
+
+	file, err = os.Open(filePath)
+	require.NoError(t, err)
+	parser, err = NewCSVParser(ctx, &cfg.Mydumper.CSV, file, 128, ioWorker, false, nil)
+	require.NoError(t, err)
+
+	for parser.ReadRow() == nil {
+		p, _ := parser.Pos()
+		require.Equal(t, pos[0], p)
+		pos = pos[1:]
+	}
+	require.NoError(t, parser.Close())
+}
diff --git a/br/tests/lightning_column_permutation/config.toml b/br/tests/lightning_column_permutation/config.toml
index 7bbff7f805a46..845873d84c8b9 100644
--- a/br/tests/lightning_column_permutation/config.toml
+++ b/br/tests/lightning_column_permutation/config.toml
@@ -1,3 +1,6 @@
 [mydumper]
 strict-format = true
 max-region-size = 200
+
+[mydumper.csv]
+terminator = "\n"
diff --git a/pkg/executor/importer/import.go b/pkg/executor/importer/import.go
index 2dd931e828658..6c2e01933b1b8 100644
--- a/pkg/executor/importer/import.go
+++ b/pkg/executor/importer/import.go
@@ -712,6 +712,9 @@ func (p *Plan) initOptions(seCtx sessionctx.Context, options []*plannercore.Load
 	}
 
 	p.adjustOptions()
+	if p.SplitFile && len(p.LinesTerminatedBy) == 0 {
+		return exeerrors.ErrInvalidOptionVal.FastGenByArgs("lines_terminated_by, should not be empty when use split_file")
+	}
 	return nil
 }
 
diff --git a/tests/realtikvtest/importintotest4/split_file_test.go b/tests/realtikvtest/importintotest4/split_file_test.go
index 6d8a5bf7e4665..1e88ba562dea8 100644
--- a/tests/realtikvtest/importintotest4/split_file_test.go
+++ b/tests/realtikvtest/importintotest4/split_file_test.go
@@ -53,7 +53,7 @@ func (s *mockGCSSuite) TestSplitFile() {
 	})
 	// split into 3 engines(subtasks)
 	importSQL := fmt.Sprintf(`import into split_file.t FROM 'gs://split-file/1.csv?endpoint=%s'
-		with split_file, __max_engine_size = '1'`, gcsEndpoint)
+		with split_file, lines_terminated_by = '\n', __max_engine_size = '1'`, gcsEndpoint)
 	result := s.tk.MustQuery(importSQL).Rows()
 	s.Len(result, 1)
 	jobID, err := strconv.Atoi(result[0][0].(string))
@@ -73,7 +73,18 @@ func (s *mockGCSSuite) TestSplitFile() {
 	// skip 1 row
 	s.tk.MustExec("truncate table t")
 	importSQL = fmt.Sprintf(`import into split_file.t FROM 'gs://split-file/1.csv?endpoint=%s'
-		with split_file, skip_rows = 1, __max_engine_size = '1'`, gcsEndpoint)
+		with split_file, lines_terminated_by = '\n', skip_rows = 1, __max_engine_size = '1'`, gcsEndpoint)
 	s.tk.MustQuery(importSQL)
 	s.tk.MustQuery("select * from t").Sort().Check(testkit.Rows(allData[1:]...))
+
+	s.tk.MustExec("create table t2 (a int primary key nonclustered, b varchar(100));")
+	s.server.CreateObject(fakestorage.Object{
+		ObjectAttrs: fakestorage.ObjectAttrs{BucketName: "split-file", Name: "2.csv"},
+		Content:     []byte("1,2\r\n3,4\r\n5,6\r\n7,8\r\n9,10\r\n"),
+	})
+	config.MaxRegionSize = 9
+	importSQL = fmt.Sprintf(`import into split_file.t2 FROM 'gs://split-file/2.csv?endpoint=%s'
+		with split_file, lines_terminated_by='\r\n'`, gcsEndpoint)
+	s.tk.MustQuery(importSQL)
+	s.tk.MustExec("admin check table t2")
 }