fix lightning populate chunks

br/pkg/lightning/mydump/region.go

@@ -268,7 +268,7 @@ func makeSourceFileRegion(
 	}
 	// If a csv file is overlarge, we need to split it into multiple regions.
 	// Note: We can only split a csv file whose format is strict.
-	if isCsvFile && dataFileSize > int64(cfg.Mydumper.MaxRegionSize) && cfg.Mydumper.StrictFormat {
+	if isCsvFile && cfg.Mydumper.StrictFormat && dataFileSize > int64(cfg.Mydumper.MaxRegionSize)*11/10 {
 		_, regions, subFileSizes, err := SplitLargeFile(ctx, meta, cfg, fi, divisor, 0, ioWorkers, store)
 		return regions, subFileSizes, err
 	}
@@ -359,6 +359,9 @@ func SplitLargeFile(
 		columns = parser.Columns()
 		startOffset, _ = parser.Pos()
 		endOffset = startOffset + maxRegionSize
+		if endOffset > dataFile.FileMeta.FileSize {
+			endOffset = dataFile.FileMeta.FileSize
+		}
 	}
 	for {
 		curRowsCnt := (endOffset - startOffset) / divisor

br/pkg/lightning/mydump/region_test.go

@@ -331,3 +331,59 @@ func (s *testMydumpRegionSuite) TestSplitLargeFileWithCustomTerminator(c *C) {
 		c.Assert(regions[i].Chunk.EndOffset, Equals, offsets[i][1])
 	}
 }
+
+func (s *testMydumpRegionSuite) TestSplitLargeFileOnlyOneChunk(c *C) {
+	meta := &MDTableMeta{
+		DB:   "csv",
+		Name: "large_csv_file",
+	}
+	cfg := &config.Config{
+		Mydumper: config.MydumperRuntime{
+			ReadBlockSize: config.ReadBlockSize,
+			CSV: config.CSVConfig{
+				Separator:       ",",
+				Delimiter:       "",
+				Header:          true,
+				TrimLastSep:     false,
+				NotNull:         false,
+				Null:            "NULL",
+				BackslashEscape: true,
+			},
+			StrictFormat:  true,
+			Filter:        []string{"*.*"},
+			MaxRegionSize: 15,
+		},
+	}
+
+	dir := c.MkDir()
+
+	fileName := "test.csv"
+	filePath := filepath.Join(dir, fileName)
+
+	content := []byte("field1,field2\r\n123,456\r\n")
+	err := os.WriteFile(filePath, content, 0o644)
+	c.Assert(err, IsNil)
+
+	dataFileInfo, err := os.Stat(filePath)
+	c.Assert(err, IsNil)
+	fileSize := dataFileInfo.Size()
+	fileInfo := FileInfo{FileMeta: SourceFileMeta{Path: fileName, Type: SourceTypeCSV, FileSize: fileSize}}
+	colCnt := int64(2)
+	columns := []string{"field1", "field2"}
+	prevRowIdxMax := int64(0)
+	ioWorker := worker.NewPool(context.Background(), 4, "io")
+
+	store, err := storage.NewLocalStorage(dir)
+	c.Assert(err, IsNil)
+
+	offsets := [][]int64{{14, 24}}
+
+	_, regions, _, err := SplitLargeFile(context.Background(), meta, cfg, fileInfo, colCnt, prevRowIdxMax, ioWorker, store)
+	c.Assert(err, IsNil)
+	c.Assert(regions, HasLen, len(offsets))
+	for i := range offsets {
+		c.Assert(regions[i].Chunk.Offset, Equals, offsets[i][0])
+		c.Assert(regions[i].Chunk.EndOffset, Equals, offsets[i][1])
+		c.Assert(regions[i].Chunk.Columns, DeepEquals, columns)
+	}
+}

br/pkg/storage/s3.go

@@ -648,6 +648,17 @@ func (r *s3ObjectReader) Close() error {
 	return r.reader.Close()
 }
 
+// eofReader is a io.ReaderClose Reader that always return io.EOF
+type eofReader struct{}
+
+func (r eofReader) Read([]byte) (n int, err error) {
+	return 0, io.EOF
+}
+
+func (r eofReader) Close() error {
+	return nil
+}
+
 // Seek implement the io.Seeker interface.
 //
 // Currently, tidb-lightning depends on this method to read parquet file for s3 storage.
@@ -666,6 +677,18 @@ func (r *s3ObjectReader) Seek(offset int64, whence int) (int64, error) {
 
 	if realOffset == r.pos {
 		return realOffset, nil
+	} else if realOffset >= r.rangeInfo.Size {
+		// See: https://www.w3.org/Protocols/rfc2616/rfc2616-sec14.html#sec14.35
+		// because s3's GetObject interface doesn't all a range that matches zero lenghth data,
+		// so if the position is out of range, we need to always return io.EOF after the seek operation.
+
+		// close current read and open a new one which target offset
+		if err := r.reader.Close(); err != nil {
+			log.L().Warn("close s3 reader failed, will ignore this error", logutil.ShortError(err))
+		}
+
+		r.reader = eofReader{}
+		return r.rangeInfo.Size, nil
 	}
 
 	// if seek ahead no more than 64k, we discard these data

br/pkg/storage/s3_test.go

@@ -740,6 +740,15 @@ func (s *s3Suite) TestOpenSeek(c *C) {
 	c.Assert(err, IsNil)
 	c.Assert(n, Equals, 100)
 	c.Assert(slice, DeepEquals, someRandomBytes[990100:990200])
+
+	// test seek to the file end or bigger positions
+	for _, p := range []int64{1000000, 1000001, 2000000} {
+		offset, err = reader.Seek(p, io.SeekStart)
+		c.Assert(offset, Equals, int64(1000000))
+		c.Assert(err, IsNil)
+		_, err = reader.Read(slice)
+		c.Assert(err, Equals, io.EOF)
+	}
 }
 
 type limitedBytesReader struct {