334 lines
8.3 KiB
Go
334 lines
8.3 KiB
Go
package exif
|
|
|
|
import (
|
|
"bufio"
|
|
"bytes"
|
|
"errors"
|
|
"fmt"
|
|
"io"
|
|
"os"
|
|
|
|
"encoding/binary"
|
|
"io/ioutil"
|
|
|
|
"github.com/dsoprea/go-logging"
|
|
|
|
"github.com/dsoprea/go-exif/v3/common"
|
|
)
|
|
|
|
const (
|
|
// ExifAddressableAreaStart is the absolute offset in the file that all
|
|
// offsets are relative to.
|
|
ExifAddressableAreaStart = uint32(0x0)
|
|
|
|
// ExifDefaultFirstIfdOffset is essentially the number of bytes in addition
|
|
// to `ExifAddressableAreaStart` that you have to move in order to escape
|
|
// the rest of the header and get to the earliest point where we can put
|
|
// stuff (which has to be the first IFD). This is the size of the header
|
|
// sequence containing the two-character byte-order, two-character fixed-
|
|
// bytes, and the four bytes describing the first-IFD offset.
|
|
ExifDefaultFirstIfdOffset = uint32(2 + 2 + 4)
|
|
)
|
|
|
|
const (
|
|
// ExifSignatureLength is the number of bytes in the EXIF signature (which
|
|
// customarily includes the first IFD offset).
|
|
ExifSignatureLength = 8
|
|
)
|
|
|
|
var (
|
|
exifLogger = log.NewLogger("exif.exif")
|
|
|
|
ExifBigEndianSignature = [4]byte{'M', 'M', 0x00, 0x2a}
|
|
ExifLittleEndianSignature = [4]byte{'I', 'I', 0x2a, 0x00}
|
|
)
|
|
|
|
var (
|
|
ErrNoExif = errors.New("no exif data")
|
|
ErrExifHeaderError = errors.New("exif header error")
|
|
)
|
|
|
|
// SearchAndExtractExif searches for an EXIF blob in the byte-slice.
|
|
func SearchAndExtractExif(data []byte) (rawExif []byte, err error) {
|
|
defer func() {
|
|
if state := recover(); state != nil {
|
|
err = log.Wrap(state.(error))
|
|
}
|
|
}()
|
|
|
|
b := bytes.NewBuffer(data)
|
|
|
|
rawExif, err = SearchAndExtractExifWithReader(b)
|
|
if err != nil {
|
|
if err == ErrNoExif {
|
|
return nil, err
|
|
}
|
|
|
|
log.Panic(err)
|
|
}
|
|
|
|
return rawExif, nil
|
|
}
|
|
|
|
// SearchAndExtractExifN searches for an EXIF blob in the byte-slice, but skips
|
|
// the given number of EXIF blocks first. This is a forensics tool that helps
|
|
// identify multiple EXIF blocks in a file.
|
|
func SearchAndExtractExifN(data []byte, n int) (rawExif []byte, err error) {
|
|
defer func() {
|
|
if state := recover(); state != nil {
|
|
err = log.Wrap(state.(error))
|
|
}
|
|
}()
|
|
|
|
skips := 0
|
|
totalDiscarded := 0
|
|
for {
|
|
b := bytes.NewBuffer(data)
|
|
|
|
var discarded int
|
|
|
|
rawExif, discarded, err = searchAndExtractExifWithReaderWithDiscarded(b)
|
|
if err != nil {
|
|
if err == ErrNoExif {
|
|
return nil, err
|
|
}
|
|
|
|
log.Panic(err)
|
|
}
|
|
|
|
exifLogger.Debugf(nil, "Read EXIF block (%d).", skips)
|
|
|
|
totalDiscarded += discarded
|
|
|
|
if skips >= n {
|
|
exifLogger.Debugf(nil, "Reached requested EXIF block (%d).", n)
|
|
break
|
|
}
|
|
|
|
nextOffset := discarded + 1
|
|
exifLogger.Debugf(nil, "Skipping EXIF block (%d) by seeking to position (%d).", skips, nextOffset)
|
|
|
|
data = data[nextOffset:]
|
|
skips++
|
|
}
|
|
|
|
exifLogger.Debugf(nil, "Found EXIF blob (%d) bytes from initial position.", totalDiscarded)
|
|
return rawExif, nil
|
|
}
|
|
|
|
// searchAndExtractExifWithReaderWithDiscarded searches for an EXIF blob using
|
|
// an `io.Reader`. We can't know how much long the EXIF data is without parsing
|
|
// it, so this will likely grab up a lot of the image-data, too.
|
|
//
|
|
// This function returned the count of preceding bytes.
|
|
func searchAndExtractExifWithReaderWithDiscarded(r io.Reader) (rawExif []byte, discarded int, err error) {
|
|
defer func() {
|
|
if state := recover(); state != nil {
|
|
err = log.Wrap(state.(error))
|
|
}
|
|
}()
|
|
|
|
// Search for the beginning of the EXIF information. The EXIF is near the
|
|
// beginning of most JPEGs, so this likely doesn't have a high cost (at
|
|
// least, again, with JPEGs).
|
|
|
|
br := bufio.NewReader(r)
|
|
|
|
for {
|
|
window, err := br.Peek(ExifSignatureLength)
|
|
if err != nil {
|
|
if err == io.EOF {
|
|
return nil, 0, ErrNoExif
|
|
}
|
|
|
|
log.Panic(err)
|
|
}
|
|
|
|
_, err = ParseExifHeader(window)
|
|
if err != nil {
|
|
if log.Is(err, ErrNoExif) == true {
|
|
// No EXIF. Move forward by one byte.
|
|
|
|
_, err := br.Discard(1)
|
|
log.PanicIf(err)
|
|
|
|
discarded++
|
|
|
|
continue
|
|
}
|
|
|
|
// Some other error.
|
|
log.Panic(err)
|
|
}
|
|
|
|
break
|
|
}
|
|
|
|
exifLogger.Debugf(nil, "Found EXIF blob (%d) bytes from initial position.", discarded)
|
|
|
|
rawExif, err = ioutil.ReadAll(br)
|
|
log.PanicIf(err)
|
|
|
|
return rawExif, discarded, nil
|
|
}
|
|
|
|
// RELEASE(dustin): We should replace the implementation of SearchAndExtractExifWithReader with searchAndExtractExifWithReaderWithDiscarded and drop the latter.
|
|
|
|
// SearchAndExtractExifWithReader searches for an EXIF blob using an
|
|
// `io.Reader`. We can't know how much long the EXIF data is without parsing it,
|
|
// so this will likely grab up a lot of the image-data, too.
|
|
func SearchAndExtractExifWithReader(r io.Reader) (rawExif []byte, err error) {
|
|
defer func() {
|
|
if state := recover(); state != nil {
|
|
err = log.Wrap(state.(error))
|
|
}
|
|
}()
|
|
|
|
rawExif, _, err = searchAndExtractExifWithReaderWithDiscarded(r)
|
|
if err != nil {
|
|
if err == ErrNoExif {
|
|
return nil, err
|
|
}
|
|
|
|
log.Panic(err)
|
|
}
|
|
|
|
return rawExif, nil
|
|
}
|
|
|
|
// SearchFileAndExtractExif returns a slice from the beginning of the EXIF data
|
|
// to the end of the file (it's not practical to try and calculate where the
|
|
// data actually ends).
|
|
func SearchFileAndExtractExif(filepath string) (rawExif []byte, err error) {
|
|
defer func() {
|
|
if state := recover(); state != nil {
|
|
err = log.Wrap(state.(error))
|
|
}
|
|
}()
|
|
|
|
// Open the file.
|
|
|
|
f, err := os.Open(filepath)
|
|
log.PanicIf(err)
|
|
|
|
defer f.Close()
|
|
|
|
rawExif, err = SearchAndExtractExifWithReader(f)
|
|
log.PanicIf(err)
|
|
|
|
return rawExif, nil
|
|
}
|
|
|
|
type ExifHeader struct {
|
|
ByteOrder binary.ByteOrder
|
|
FirstIfdOffset uint32
|
|
}
|
|
|
|
func (eh ExifHeader) String() string {
|
|
return fmt.Sprintf("ExifHeader<BYTE-ORDER=[%v] FIRST-IFD-OFFSET=(0x%02x)>", eh.ByteOrder, eh.FirstIfdOffset)
|
|
}
|
|
|
|
// ParseExifHeader parses the bytes at the very top of the header.
|
|
//
|
|
// This will panic with ErrNoExif on any data errors so that we can double as
|
|
// an EXIF-detection routine.
|
|
func ParseExifHeader(data []byte) (eh ExifHeader, err error) {
|
|
defer func() {
|
|
if state := recover(); state != nil {
|
|
err = log.Wrap(state.(error))
|
|
}
|
|
}()
|
|
|
|
// Good reference:
|
|
//
|
|
// CIPA DC-008-2016; JEITA CP-3451D
|
|
// -> http://www.cipa.jp/std/documents/e/DC-008-Translation-2016-E.pdf
|
|
|
|
if len(data) < ExifSignatureLength {
|
|
exifLogger.Warningf(nil, "Not enough data for EXIF header: (%d)", len(data))
|
|
return eh, ErrNoExif
|
|
}
|
|
|
|
if bytes.Equal(data[:4], ExifBigEndianSignature[:]) == true {
|
|
exifLogger.Debugf(nil, "Byte-order is big-endian.")
|
|
eh.ByteOrder = binary.BigEndian
|
|
} else if bytes.Equal(data[:4], ExifLittleEndianSignature[:]) == true {
|
|
eh.ByteOrder = binary.LittleEndian
|
|
exifLogger.Debugf(nil, "Byte-order is little-endian.")
|
|
} else {
|
|
return eh, ErrNoExif
|
|
}
|
|
|
|
eh.FirstIfdOffset = eh.ByteOrder.Uint32(data[4:8])
|
|
|
|
return eh, nil
|
|
}
|
|
|
|
// Visit recursively invokes a callback for every tag.
|
|
func Visit(rootIfdIdentity *exifcommon.IfdIdentity, ifdMapping *exifcommon.IfdMapping, tagIndex *TagIndex, exifData []byte, visitor TagVisitorFn, so *ScanOptions) (eh ExifHeader, furthestOffset uint32, err error) {
|
|
defer func() {
|
|
if state := recover(); state != nil {
|
|
err = log.Wrap(state.(error))
|
|
}
|
|
}()
|
|
|
|
eh, err = ParseExifHeader(exifData)
|
|
log.PanicIf(err)
|
|
|
|
ebs := NewExifReadSeekerWithBytes(exifData)
|
|
ie := NewIfdEnumerate(ifdMapping, tagIndex, ebs, eh.ByteOrder)
|
|
|
|
_, err = ie.Scan(rootIfdIdentity, eh.FirstIfdOffset, visitor, so)
|
|
log.PanicIf(err)
|
|
|
|
furthestOffset = ie.FurthestOffset()
|
|
|
|
return eh, furthestOffset, nil
|
|
}
|
|
|
|
// Collect recursively builds a static structure of all IFDs and tags.
|
|
func Collect(ifdMapping *exifcommon.IfdMapping, tagIndex *TagIndex, exifData []byte) (eh ExifHeader, index IfdIndex, err error) {
|
|
defer func() {
|
|
if state := recover(); state != nil {
|
|
err = log.Wrap(state.(error))
|
|
}
|
|
}()
|
|
|
|
eh, err = ParseExifHeader(exifData)
|
|
log.PanicIf(err)
|
|
|
|
ebs := NewExifReadSeekerWithBytes(exifData)
|
|
ie := NewIfdEnumerate(ifdMapping, tagIndex, ebs, eh.ByteOrder)
|
|
|
|
index, err = ie.Collect(eh.FirstIfdOffset)
|
|
log.PanicIf(err)
|
|
|
|
return eh, index, nil
|
|
}
|
|
|
|
// BuildExifHeader constructs the bytes that go at the front of the stream.
|
|
func BuildExifHeader(byteOrder binary.ByteOrder, firstIfdOffset uint32) (headerBytes []byte, err error) {
|
|
defer func() {
|
|
if state := recover(); state != nil {
|
|
err = log.Wrap(state.(error))
|
|
}
|
|
}()
|
|
|
|
b := new(bytes.Buffer)
|
|
|
|
var signatureBytes []byte
|
|
if byteOrder == binary.BigEndian {
|
|
signatureBytes = ExifBigEndianSignature[:]
|
|
} else {
|
|
signatureBytes = ExifLittleEndianSignature[:]
|
|
}
|
|
|
|
_, err = b.Write(signatureBytes)
|
|
log.PanicIf(err)
|
|
|
|
err = binary.Write(b, byteOrder, firstIfdOffset)
|
|
log.PanicIf(err)
|
|
|
|
return b.Bytes(), nil
|
|
}
|