Make fingerprinting faster (#36073) (#36080)

By using a buffer of the fingerprint's size. This descreases the
amount of syscalls for reading the file.

Also added the benchmark for tracking the function's performance.

(cherry picked from commit 061cb88)

Co-authored-by: Denis <denis.rechkunov@elastic.co>

filebeat/input/filestream/fswatch.go

@@ -18,6 +18,7 @@
 package filestream
 
 import (
+	"bufio"
 	"crypto/sha256"
 	"encoding/hex"
 	"fmt"
@@ -453,7 +454,6 @@ func (s *fileScanner) toFileDescriptor(it *ingestTarget) (fd loginp.FileDescript
 			return fd, fmt.Errorf("filesize of %q is %d bytes, expected at least %d bytes for fingerprinting", fd.Filename, fileSize, minSize)
 		}
 
-		h := sha256.New()
 		file, err := os.Open(it.originalFilename)
 		if err != nil {
 			return fd, fmt.Errorf("failed to open %q for fingerprinting: %w", it.originalFilename, err)
@@ -467,9 +467,10 @@ func (s *fileScanner) toFileDescriptor(it *ingestTarget) (fd loginp.FileDescript
 			}
 		}
 
-		r := io.LimitReader(file, s.cfg.Fingerprint.Length)
-		buf := make([]byte, h.BlockSize())
-		written, err := io.CopyBuffer(h, r, buf)
+		bfile := bufio.NewReaderSize(file, int(s.cfg.Fingerprint.Length))
+		r := io.LimitReader(bfile, s.cfg.Fingerprint.Length)
+		h := sha256.New()
+		written, err := io.Copy(h, r)
 		if err != nil {
 			return fd, fmt.Errorf("failed to compute hash for first %d bytes of %q: %w", s.cfg.Fingerprint.Length, fd.Filename, err)
 		}

filebeat/input/filestream/fswatch_test.go

@@ -757,3 +757,31 @@ func filenames(m map[string]loginp.FileDescriptor) (result string) {
 	}
 	return result
 }
+
+func BenchmarkToFileDescriptor(b *testing.B) {
+	dir := b.TempDir()
+	basename := "created.log"
+	filename := filepath.Join(dir, basename)
+	err := os.WriteFile(filename, []byte(strings.Repeat("a", 1024)), 0777)
+	require.NoError(b, err)
+
+	s := fileScanner{
+		paths: []string{filename},
+		cfg: fileScannerConfig{
+			Fingerprint: fingerprintConfig{
+				Enabled: true,
+				Offset:  0,
+				Length:  1024,
+			},
+		},
+	}
+
+	it, err := s.getIngestTarget(filename)
+	require.NoError(b, err)
+
+	for i := 0; i < b.N; i++ {
+		fd, err := s.toFileDescriptor(&it)
+		require.NoError(b, err)
+		require.Equal(b, "2edc986847e209b4016e141a6dc8716d3207350f416969382d431539bf292e4a", fd.Fingerprint)
+	}
+}