1 // Copyright 2017 The Go Authors. All rights reserved. 2 // Use of this source code is governed by a BSD-style 3 // license that can be found in the LICENSE file. 4 5 package buildid 6 7 import ( 8 "bytes" 9 "crypto/sha256" 10 "fmt" 11 "io" 12 ) 13 14 // FindAndHash reads all of r and returns the offsets of occurrences of id. 15 // While reading, findAndHash also computes and returns 16 // a hash of the content of r, but with occurrences of id replaced by zeros. 17 // FindAndHash reads bufSize bytes from r at a time. 18 // If bufSize == 0, FindAndHash uses a reasonable default. 19 func FindAndHash(r io.Reader, id string, bufSize int) (matches []int64, hash [32]byte, err error) { 20 if bufSize == 0 { 21 bufSize = 31 * 1024 // bufSize+little will likely fit in 32 kB 22 } 23 if len(id) > bufSize { 24 return nil, [32]byte{}, fmt.Errorf("buildid.FindAndHash: buffer too small") 25 } 26 zeros := make([]byte, len(id)) 27 idBytes := []byte(id) 28 29 // The strategy is to read the file through buf, looking for id, 30 // but we need to worry about what happens if id is broken up 31 // and returned in parts by two different reads. 32 // We allocate a tiny buffer (at least len(id)) and a big buffer (bufSize bytes) 33 // next to each other in memory and then copy the tail of 34 // one read into the tiny buffer before reading new data into the big buffer. 35 // The search for id is over the entire tiny+big buffer. 36 tiny := (len(id) + 127) &^ 127 // round up to 128-aligned 37 buf := make([]byte, tiny+bufSize) 38 h := sha256.New() 39 start := tiny 40 for offset := int64(0); ; { 41 // The file offset maintained by the loop corresponds to &buf[tiny]. 42 // buf[start:tiny] is left over from previous iteration. 43 // After reading n bytes into buf[tiny:], we process buf[start:tiny+n]. 44 n, err := io.ReadFull(r, buf[tiny:]) 45 if err != io.ErrUnexpectedEOF && err != io.EOF && err != nil { 46 return nil, [32]byte{}, err 47 } 48 49 // Process any matches. 50 for { 51 i := bytes.Index(buf[start:tiny+n], idBytes) 52 if i < 0 { 53 break 54 } 55 matches = append(matches, offset+int64(start+i-tiny)) 56 h.Write(buf[start : start+i]) 57 h.Write(zeros) 58 start += i + len(id) 59 } 60 if n < bufSize { 61 // Did not fill buffer, must be at end of file. 62 h.Write(buf[start : tiny+n]) 63 break 64 } 65 66 // Process all but final tiny bytes of buf (bufSize = len(buf)-tiny). 67 // Note that start > len(buf)-tiny is possible, if the search above 68 // found an id ending in the final tiny fringe. That's OK. 69 if start < len(buf)-tiny { 70 h.Write(buf[start : len(buf)-tiny]) 71 start = len(buf) - tiny 72 } 73 74 // Slide ending tiny-sized fringe to beginning of buffer. 75 copy(buf[0:], buf[bufSize:]) 76 start -= bufSize 77 offset += int64(bufSize) 78 } 79 h.Sum(hash[:0]) 80 return matches, hash, nil 81 } 82 83 func Rewrite(w io.WriterAt, pos []int64, id string) error { 84 b := []byte(id) 85 for _, p := range pos { 86 if _, err := w.WriteAt(b, p); err != nil { 87 return err 88 } 89 } 90 return nil 91 } 92
