func (ls *layerStore) applyTar(tx MetadataTransaction, ts io.Reader, parent string, layer *roLayer) error {
digester := digest.Canonical.New()
tr := io.TeeReader(ts, digester.Hash())
tsw, err := tx.TarSplitWriter(true)
if err != nil {
return err
}
metaPacker := storage.NewJSONPacker(tsw)
defer tsw.Close()
// we're passing nil here for the file putter, because the ApplyDiff will
// handle the extraction of the archive
rdr, err := asm.NewInputTarStream(tr, metaPacker, nil)
if err != nil {
return err
}
applySize, err := ls.driver.ApplyDiff(layer.cacheID, parent, archive.Reader(rdr))
if err != nil {
return err
}
// Discard trailing data but ensure metadata is picked up to reconstruct stream
io.Copy(ioutil.Discard, rdr) // ignore error as reader may be closed
layer.size = applySize
layer.diffID = DiffID(digester.Digest())
logrus.Debugf("Applied tar %s to %s, size: %d", layer.diffID, layer.cacheID, applySize)
return nil
}
func (c *checksums) checksumForGraphIDNoTarsplit(id, parent, newTarDataPath string) (diffID layer.DiffID, size int64, err error) {
rawarchive, err := c.driver.TarStream(id, parent)
if err != nil {
return
}
defer rawarchive.Close()
f, err := os.Create(newTarDataPath)
if err != nil {
return
}
defer f.Close()
mfz := gzip.NewWriter(f)
defer mfz.Close()
metaPacker := storage.NewJSONPacker(mfz)
packerCounter := &packSizeCounter{metaPacker, &size}
archive, err := asm.NewInputTarStream(rawarchive, packerCounter, nil)
if err != nil {
return
}
dgst, err := digest.FromReader(archive)
if err != nil {
return
}
diffID = layer.DiffID(dgst)
return
}
func (graph *Graph) disassembleAndApplyTarLayer(id, parent string, layerData archive.ArchiveReader, root string) (size int64, err error) {
// this is saving the tar-split metadata
mf, err := os.OpenFile(filepath.Join(root, tarDataFileName), os.O_CREATE|os.O_WRONLY|os.O_TRUNC, os.FileMode(0600))
if err != nil {
return 0, err
}
mfz := gzip.NewWriter(mf)
metaPacker := storage.NewJSONPacker(mfz)
defer mf.Close()
defer mfz.Close()
inflatedLayerData, err := archive.DecompressStream(layerData)
if err != nil {
return 0, err
}
// we're passing nil here for the file putter, because the ApplyDiff will
// handle the extraction of the archive
rdr, err := asm.NewInputTarStream(inflatedLayerData, metaPacker, nil)
if err != nil {
return 0, err
}
if size, err = graph.driver.ApplyDiff(id, parent, archive.ArchiveReader(rdr)); err != nil {
return 0, err
}
return
}
func BenchmarkAsm(b *testing.B) {
for i := 0; i < b.N; i++ {
for _, tc := range testCases {
func() {
fh, err := os.Open(tc.path)
if err != nil {
b.Fatal(err)
}
defer fh.Close()
gzRdr, err := gzip.NewReader(fh)
if err != nil {
b.Fatal(err)
}
defer gzRdr.Close()
// Setup where we'll store the metadata
w := bytes.NewBuffer([]byte{})
sp := storage.NewJSONPacker(w)
fgp := storage.NewBufferFileGetPutter()
// wrap the disassembly stream
tarStream, err := NewInputTarStream(gzRdr, sp, fgp)
if err != nil {
b.Fatal(err)
}
// read it all to the bit bucket
i1, err := io.Copy(ioutil.Discard, tarStream)
if err != nil {
b.Fatal(err)
}
r := bytes.NewBuffer(w.Bytes())
sup := storage.NewJSONUnpacker(r)
// and reuse the fgp that we Put the payloads to.
rc := NewOutputTarStream(fgp, sup)
i2, err := io.Copy(ioutil.Discard, rc)
if err != nil {
b.Fatal(err)
}
if i1 != i2 {
b.Errorf("%s: input(%d) and ouput(%d) byte count didn't match", tc.path, i1, i2)
}
}()
}
}
}
func CommandDisasm(c *cli.Context) {
if len(c.Args()) != 1 {
logrus.Fatalf("please specify tar to be disabled <NAME|->")
}
if len(c.String("output")) == 0 {
logrus.Fatalf("--output filename must be set")
}
// Set up the tar input stream
var inputStream io.Reader
if c.Args()[0] == "-" {
inputStream = os.Stdin
} else {
fh, err := os.Open(c.Args()[0])
if err != nil {
logrus.Fatal(err)
}
defer fh.Close()
inputStream = fh
}
// Set up the metadata storage
mf, err := os.OpenFile(c.String("output"), os.O_CREATE|os.O_WRONLY|os.O_TRUNC, os.FileMode(0600))
if err != nil {
logrus.Fatal(err)
}
defer mf.Close()
mfz := gzip.NewWriter(mf)
defer mfz.Close()
metaPacker := storage.NewJSONPacker(mfz)
// we're passing nil here for the file putter, because the ApplyDiff will
// handle the extraction of the archive
its, err := asm.NewInputTarStream(inputStream, metaPacker, nil)
if err != nil {
logrus.Fatal(err)
}
i, err := io.Copy(os.Stdout, its)
if err != nil {
logrus.Fatal(err)
}
logrus.Infof("created %s from %s (read %d bytes)", c.String("output"), c.Args()[0], i)
}
func (graph *Graph) disassembleAndApplyTarLayer(img *image.Image, layerData io.Reader, root string) (err error) {
var ar io.Reader
if graph.tarSplitDisabled {
ar = layerData
} else {
// this is saving the tar-split metadata
mf, err := os.OpenFile(filepath.Join(root, tarDataFileName), os.O_CREATE|os.O_WRONLY|os.O_TRUNC, os.FileMode(0600))
if err != nil {
return err
}
mfz := gzip.NewWriter(mf)
metaPacker := storage.NewJSONPacker(mfz)
defer mf.Close()
defer mfz.Close()
inflatedLayerData, err := archive.DecompressStream(layerData)
if err != nil {
return err
}
// we're passing nil here for the file putter, because the ApplyDiff will
// handle the extraction of the archive
rdr, err := asm.NewInputTarStream(inflatedLayerData, metaPacker, nil)
if err != nil {
return err
}
ar = archive.Reader(rdr)
}
if img.Size, err = graph.driver.ApplyDiff(img.ID, img.Parent, ar); err != nil {
return err
}
return nil
}
func writeTarSplitFile(name string, tarContent []byte) error {
f, err := os.OpenFile(name, os.O_TRUNC|os.O_CREATE|os.O_WRONLY, 0644)
if err != nil {
return err
}
defer f.Close()
fz := gzip.NewWriter(f)
metaPacker := storage.NewJSONPacker(fz)
defer fz.Close()
rdr, err := asm.NewInputTarStream(bytes.NewReader(tarContent), metaPacker, nil)
if err != nil {
return err
}
if _, err := io.Copy(ioutil.Discard, rdr); err != nil {
return err
}
return nil
}
func TestTarStream(t *testing.T) {
var (
expectedSum = "1eb237ff69bca6e22789ecb05b45d35ca307adbd"
expectedSize int64 = 10240
)
fh, err := os.Open("./testdata/t.tar.gz")
if err != nil {
t.Fatal(err)
}
defer fh.Close()
gzRdr, err := gzip.NewReader(fh)
if err != nil {
t.Fatal(err)
}
defer gzRdr.Close()
// Setup where we'll store the metadata
w := bytes.NewBuffer([]byte{})
sp := storage.NewJSONPacker(w)
fgp := storage.NewBufferFileGetPutter()
// wrap the disassembly stream
tarStream, err := NewInputTarStream(gzRdr, sp, fgp)
if err != nil {
t.Fatal(err)
}
// get a sum of the stream after it has passed through to ensure it's the same.
h0 := sha1.New()
tRdr0 := io.TeeReader(tarStream, h0)
// read it all to the bit bucket
i, err := io.Copy(ioutil.Discard, tRdr0)
if err != nil {
t.Fatal(err)
}
if i != expectedSize {
t.Errorf("size of tar: expected %d; got %d", expectedSize, i)
}
if fmt.Sprintf("%x", h0.Sum(nil)) != expectedSum {
t.Fatalf("checksum of tar: expected %s; got %x", expectedSum, h0.Sum(nil))
}
t.Logf("%s", w.String()) // if we fail, then show the packed info
// If we've made it this far, then we'll turn it around and create a tar
// stream from the packed metadata and buffered file contents.
r := bytes.NewBuffer(w.Bytes())
sup := storage.NewJSONUnpacker(r)
// and reuse the fgp that we Put the payloads to.
rc := NewOutputTarStream(fgp, sup)
h1 := sha1.New()
i, err = io.Copy(h1, rc)
if err != nil {
t.Fatal(err)
}
if i != expectedSize {
t.Errorf("size of output tar: expected %d; got %d", expectedSize, i)
}
if fmt.Sprintf("%x", h1.Sum(nil)) != expectedSum {
t.Fatalf("checksum of output tar: expected %s; got %x", expectedSum, h1.Sum(nil))
}
}
func TestTarStream(t *testing.T) {
testCases := []struct {
path string
expectedSHA1Sum string
expectedSize int64
}{
{"./testdata/t.tar.gz", "1eb237ff69bca6e22789ecb05b45d35ca307adbd", 10240},
{"./testdata/longlink.tar.gz", "d9f6babe107b7247953dff6b5b5ae31a3a880add", 20480},
{"./testdata/fatlonglink.tar.gz", "8537f03f89aeef537382f8b0bb065d93e03b0be8", 26234880},
{"./testdata/iso-8859.tar.gz", "ddafa51cb03c74ec117ab366ee2240d13bba1ec3", 10240},
}
for _, tc := range testCases {
fh, err := os.Open(tc.path)
if err != nil {
t.Fatal(err)
}
defer fh.Close()
gzRdr, err := gzip.NewReader(fh)
if err != nil {
t.Fatal(err)
}
defer gzRdr.Close()
// Setup where we'll store the metadata
w := bytes.NewBuffer([]byte{})
sp := storage.NewJSONPacker(w)
fgp := storage.NewBufferFileGetPutter()
// wrap the disassembly stream
tarStream, err := NewInputTarStream(gzRdr, sp, fgp)
if err != nil {
t.Fatal(err)
}
// get a sum of the stream after it has passed through to ensure it's the same.
h0 := sha1.New()
tRdr0 := io.TeeReader(tarStream, h0)
// read it all to the bit bucket
i, err := io.Copy(ioutil.Discard, tRdr0)
if err != nil {
t.Fatal(err)
}
if i != tc.expectedSize {
t.Errorf("size of tar: expected %d; got %d", tc.expectedSize, i)
}
if fmt.Sprintf("%x", h0.Sum(nil)) != tc.expectedSHA1Sum {
t.Fatalf("checksum of tar: expected %s; got %x", tc.expectedSHA1Sum, h0.Sum(nil))
}
//t.Logf("%s", w.String()) // if we fail, then show the packed info
// If we've made it this far, then we'll turn it around and create a tar
// stream from the packed metadata and buffered file contents.
r := bytes.NewBuffer(w.Bytes())
sup := storage.NewJSONUnpacker(r)
// and reuse the fgp that we Put the payloads to.
rc := NewOutputTarStream(fgp, sup)
h1 := sha1.New()
i, err = io.Copy(h1, rc)
if err != nil {
t.Fatal(err)
}
if i != tc.expectedSize {
t.Errorf("size of output tar: expected %d; got %d", tc.expectedSize, i)
}
if fmt.Sprintf("%x", h1.Sum(nil)) != tc.expectedSHA1Sum {
t.Fatalf("checksum of output tar: expected %s; got %x", tc.expectedSHA1Sum, h1.Sum(nil))
}
}
}
func (ls *layerStore) migrateLayer(tx MetadataTransaction, tarDataFile string, layer *roLayer) error {
var ar io.Reader
var tdf *os.File
var err error
if tarDataFile != "" {
tdf, err = os.Open(tarDataFile)
if err != nil {
if !os.IsNotExist(err) {
return err
}
tdf = nil
}
defer tdf.Close()
}
if tdf != nil {
tsw, err := tx.TarSplitWriter()
if err != nil {
return err
}
defer tsw.Close()
uncompressed, err := gzip.NewReader(tdf)
if err != nil {
return err
}
defer uncompressed.Close()
tr := io.TeeReader(uncompressed, tsw)
trc := ioutils.NewReadCloserWrapper(tr, uncompressed.Close)
ar, err = ls.assembleTar(layer.cacheID, trc, &layer.size)
if err != nil {
return err
}
} else {
var graphParent string
if layer.parent != nil {
graphParent = layer.parent.cacheID
}
archiver, err := ls.driver.Diff(layer.cacheID, graphParent)
if err != nil {
return err
}
defer archiver.Close()
tsw, err := tx.TarSplitWriter()
if err != nil {
return err
}
metaPacker := storage.NewJSONPacker(tsw)
packerCounter := &packSizeCounter{metaPacker, &layer.size}
defer tsw.Close()
ar, err = asm.NewInputTarStream(archiver, packerCounter, nil)
if err != nil {
return err
}
}
digester := digest.Canonical.New()
_, err = io.Copy(digester.Hash(), ar)
if err != nil {
return err
}
layer.diffID = DiffID(digester.Digest())
return nil
}
func main() {
flag.Parse()
for _, arg := range flag.Args() {
fh, err := os.Open(arg)
if err != nil {
log.Fatal(err)
}
defer fh.Close()
fi, err := fh.Stat()
if err != nil {
log.Fatal(err)
}
fmt.Printf("inspecting %q (size %dk)\n", fh.Name(), fi.Size()/1024)
packFh, err := ioutil.TempFile("", "packed.")
if err != nil {
log.Fatal(err)
}
defer packFh.Close()
if *flCleanup {
defer os.Remove(packFh.Name())
}
sp := storage.NewJSONPacker(packFh)
fp := storage.NewDiscardFilePutter()
dissam, err := asm.NewInputTarStream(fh, sp, fp)
if err != nil {
log.Fatal(err)
}
var num int
tr := tar.NewReader(dissam)
for {
_, err = tr.Next()
if err != nil {
if err == io.EOF {
break
}
log.Fatal(err)
}
num++
if _, err := io.Copy(ioutil.Discard, tr); err != nil {
log.Fatal(err)
}
}
fmt.Printf(" -- number of files: %d\n", num)
if err := packFh.Sync(); err != nil {
log.Fatal(err)
}
fi, err = packFh.Stat()
if err != nil {
log.Fatal(err)
}
fmt.Printf(" -- size of metadata uncompressed: %dk\n", fi.Size()/1024)
gzPackFh, err := ioutil.TempFile("", "packed.gz.")
if err != nil {
log.Fatal(err)
}
defer gzPackFh.Close()
if *flCleanup {
defer os.Remove(gzPackFh.Name())
}
gzWrtr := gzip.NewWriter(gzPackFh)
if _, err := packFh.Seek(0, 0); err != nil {
log.Fatal(err)
}
if _, err := io.Copy(gzWrtr, packFh); err != nil {
log.Fatal(err)
}
gzWrtr.Close()
if err := gzPackFh.Sync(); err != nil {
log.Fatal(err)
}
fi, err = gzPackFh.Stat()
if err != nil {
log.Fatal(err)
}
fmt.Printf(" -- size of gzip compressed metadata: %dk\n", fi.Size()/1024)
}
}
func CommandChecksize(c *cli.Context) {
if len(c.Args()) == 0 {
logrus.Fatalf("please specify tar archives to check ('-' will check stdin)")
}
for _, arg := range c.Args() {
fh, err := os.Open(arg)
if err != nil {
log.Fatal(err)
}
defer fh.Close()
fi, err := fh.Stat()
if err != nil {
log.Fatal(err)
}
fmt.Printf("inspecting %q (size %dk)\n", fh.Name(), fi.Size()/1024)
packFh, err := ioutil.TempFile("", "packed.")
if err != nil {
log.Fatal(err)
}
defer packFh.Close()
if !c.Bool("work") {
defer os.Remove(packFh.Name())
} else {
fmt.Printf(" -- working file preserved: %s\n", packFh.Name())
}
sp := storage.NewJSONPacker(packFh)
fp := storage.NewDiscardFilePutter()
dissam, err := asm.NewInputTarStream(fh, sp, fp)
if err != nil {
log.Fatal(err)
}
var num int
tr := tar.NewReader(dissam)
for {
_, err = tr.Next()
if err != nil {
if err == io.EOF {
break
}
log.Fatal(err)
}
num++
if _, err := io.Copy(ioutil.Discard, tr); err != nil {
log.Fatal(err)
}
}
fmt.Printf(" -- number of files: %d\n", num)
if err := packFh.Sync(); err != nil {
log.Fatal(err)
}
fi, err = packFh.Stat()
if err != nil {
log.Fatal(err)
}
fmt.Printf(" -- size of metadata uncompressed: %dk\n", fi.Size()/1024)
gzPackFh, err := ioutil.TempFile("", "packed.gz.")
if err != nil {
log.Fatal(err)
}
defer gzPackFh.Close()
if !c.Bool("work") {
defer os.Remove(gzPackFh.Name())
}
gzWrtr := gzip.NewWriter(gzPackFh)
if _, err := packFh.Seek(0, 0); err != nil {
log.Fatal(err)
}
if _, err := io.Copy(gzWrtr, packFh); err != nil {
log.Fatal(err)
}
gzWrtr.Close()
if err := gzPackFh.Sync(); err != nil {
log.Fatal(err)
}
fi, err = gzPackFh.Stat()
if err != nil {
log.Fatal(err)
}
fmt.Printf(" -- size of gzip compressed metadata: %dk\n", fi.Size()/1024)
}
}
func TestTarStream(t *testing.T) {
for _, tc := range testCases {
fh, err := os.Open(tc.path)
if err != nil {
t.Fatal(err)
}
defer fh.Close()
gzRdr, err := gzip.NewReader(fh)
if err != nil {
t.Fatal(err)
}
defer gzRdr.Close()
// Setup where we'll store the metadata
w := bytes.NewBuffer([]byte{})
sp := storage.NewJSONPacker(w)
fgp := storage.NewBufferFileGetPutter()
// wrap the disassembly stream
tarStream, err := NewInputTarStream(gzRdr, sp, fgp)
if err != nil {
t.Fatal(err)
}
// get a sum of the stream after it has passed through to ensure it's the same.
h0 := sha1.New()
i, err := io.Copy(h0, tarStream)
if err != nil {
t.Fatal(err)
}
if i != tc.expectedSize {
t.Errorf("size of tar: expected %d; got %d", tc.expectedSize, i)
}
if fmt.Sprintf("%x", h0.Sum(nil)) != tc.expectedSHA1Sum {
t.Fatalf("checksum of tar: expected %s; got %x", tc.expectedSHA1Sum, h0.Sum(nil))
}
//t.Logf("%s", w.String()) // if we fail, then show the packed info
// If we've made it this far, then we'll turn it around and create a tar
// stream from the packed metadata and buffered file contents.
r := bytes.NewBuffer(w.Bytes())
sup := storage.NewJSONUnpacker(r)
// and reuse the fgp that we Put the payloads to.
rc := NewOutputTarStream(fgp, sup)
h1 := sha1.New()
i, err = io.Copy(h1, rc)
if err != nil {
t.Fatal(err)
}
if i != tc.expectedSize {
t.Errorf("size of output tar: expected %d; got %d", tc.expectedSize, i)
}
if fmt.Sprintf("%x", h1.Sum(nil)) != tc.expectedSHA1Sum {
t.Fatalf("checksum of output tar: expected %s; got %x", tc.expectedSHA1Sum, h1.Sum(nil))
}
}
}