1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331
|
package manifest
import (
"encoding/json"
"errors"
"fmt"
"strings"
"time"
"github.com/containers/image/v5/docker/reference"
"github.com/containers/image/v5/internal/manifest"
"github.com/containers/image/v5/internal/set"
"github.com/containers/image/v5/types"
"github.com/containers/storage/pkg/regexp"
"github.com/docker/docker/api/types/versions"
"github.com/opencontainers/go-digest"
"golang.org/x/exp/slices"
)
// Schema1FSLayers is an entry of the "fsLayers" array in docker/distribution schema 1.
type Schema1FSLayers struct {
BlobSum digest.Digest `json:"blobSum"`
}
// Schema1History is an entry of the "history" array in docker/distribution schema 1.
type Schema1History struct {
V1Compatibility string `json:"v1Compatibility"`
}
// Schema1 is a manifest in docker/distribution schema 1.
type Schema1 struct {
Name string `json:"name"`
Tag string `json:"tag"`
Architecture string `json:"architecture"`
FSLayers []Schema1FSLayers `json:"fsLayers"`
History []Schema1History `json:"history"` // Keep this in sync with ExtractedV1Compatibility!
ExtractedV1Compatibility []Schema1V1Compatibility `json:"-"` // Keep this in sync with History! Does not contain the full config (Schema2V1Image)
SchemaVersion int `json:"schemaVersion"`
}
type schema1V1CompatibilityContainerConfig struct {
Cmd []string
}
// Schema1V1Compatibility is a v1Compatibility in docker/distribution schema 1.
type Schema1V1Compatibility struct {
ID string `json:"id"`
Parent string `json:"parent,omitempty"`
Comment string `json:"comment,omitempty"`
Created time.Time `json:"created"`
ContainerConfig schema1V1CompatibilityContainerConfig `json:"container_config,omitempty"`
Author string `json:"author,omitempty"`
ThrowAway bool `json:"throwaway,omitempty"`
}
// Schema1FromManifest creates a Schema1 manifest instance from a manifest blob.
// (NOTE: The instance is not necessary a literal representation of the original blob,
// layers with duplicate IDs are eliminated.)
func Schema1FromManifest(manifestBlob []byte) (*Schema1, error) {
s1 := Schema1{}
if err := json.Unmarshal(manifestBlob, &s1); err != nil {
return nil, err
}
if s1.SchemaVersion != 1 {
return nil, fmt.Errorf("unsupported schema version %d", s1.SchemaVersion)
}
if err := manifest.ValidateUnambiguousManifestFormat(manifestBlob, DockerV2Schema1SignedMediaType,
manifest.AllowedFieldFSLayers|manifest.AllowedFieldHistory); err != nil {
return nil, err
}
if err := s1.initialize(); err != nil {
return nil, err
}
if err := s1.fixManifestLayers(); err != nil {
return nil, err
}
return &s1, nil
}
// Schema1FromComponents creates an Schema1 manifest instance from the supplied data.
func Schema1FromComponents(ref reference.Named, fsLayers []Schema1FSLayers, history []Schema1History, architecture string) (*Schema1, error) {
var name, tag string
if ref != nil { // Well, what to do if it _is_ nil? Most consumers actually don't use these fields nowadays, so we might as well try not supplying them.
name = reference.Path(ref)
if tagged, ok := ref.(reference.NamedTagged); ok {
tag = tagged.Tag()
}
}
s1 := Schema1{
Name: name,
Tag: tag,
Architecture: architecture,
FSLayers: fsLayers,
History: history,
SchemaVersion: 1,
}
if err := s1.initialize(); err != nil {
return nil, err
}
return &s1, nil
}
// Schema1Clone creates a copy of the supplied Schema1 manifest.
func Schema1Clone(src *Schema1) *Schema1 {
copy := *src
return ©
}
// initialize initializes ExtractedV1Compatibility and verifies invariants, so that the rest of this code can assume a minimally healthy manifest.
func (m *Schema1) initialize() error {
if len(m.FSLayers) != len(m.History) {
return errors.New("length of history not equal to number of layers")
}
if len(m.FSLayers) == 0 {
return errors.New("no FSLayers in manifest")
}
m.ExtractedV1Compatibility = make([]Schema1V1Compatibility, len(m.History))
for i, h := range m.History {
if err := json.Unmarshal([]byte(h.V1Compatibility), &m.ExtractedV1Compatibility[i]); err != nil {
return fmt.Errorf("parsing v2s1 history entry %d: %w", i, err)
}
}
return nil
}
// ConfigInfo returns a complete BlobInfo for the separate config object, or a BlobInfo{Digest:""} if there isn't a separate object.
func (m *Schema1) ConfigInfo() types.BlobInfo {
return types.BlobInfo{}
}
// LayerInfos returns a list of LayerInfos of layers referenced by this image, in order (the root layer first, and then successive layered layers).
// The Digest field is guaranteed to be provided; Size may be -1.
// WARNING: The list may contain duplicates, and they are semantically relevant.
func (m *Schema1) LayerInfos() []LayerInfo {
layers := make([]LayerInfo, len(m.FSLayers))
for i, layer := range m.FSLayers { // NOTE: This includes empty layers (where m.History.V1Compatibility->ThrowAway)
layers[(len(m.FSLayers)-1)-i] = LayerInfo{
BlobInfo: types.BlobInfo{Digest: layer.BlobSum, Size: -1},
EmptyLayer: m.ExtractedV1Compatibility[i].ThrowAway,
}
}
return layers
}
// UpdateLayerInfos replaces the original layers with the specified BlobInfos (size+digest+urls), in order (the root layer first, and then successive layered layers)
func (m *Schema1) UpdateLayerInfos(layerInfos []types.BlobInfo) error {
// Our LayerInfos includes empty layers (where m.ExtractedV1Compatibility[].ThrowAway), so expect them to be included here as well.
if len(m.FSLayers) != len(layerInfos) {
return fmt.Errorf("Error preparing updated manifest: layer count changed from %d to %d", len(m.FSLayers), len(layerInfos))
}
m.FSLayers = make([]Schema1FSLayers, len(layerInfos))
for i, info := range layerInfos {
// (docker push) sets up m.ExtractedV1Compatibility[].{Id,Parent} based on values of info.Digest,
// but (docker pull) ignores them in favor of computing DiffIDs from uncompressed data, except verifying the child->parent links and uniqueness.
// So, we don't bother recomputing the IDs in m.History.V1Compatibility.
m.FSLayers[(len(layerInfos)-1)-i].BlobSum = info.Digest
if info.CryptoOperation != types.PreserveOriginalCrypto {
return fmt.Errorf("encryption change (for layer %q) is not supported in schema1 manifests", info.Digest)
}
}
return nil
}
// Serialize returns the manifest in a blob format.
// NOTE: Serialize() does not in general reproduce the original blob if this object was loaded from one, even if no modifications were made!
func (m *Schema1) Serialize() ([]byte, error) {
// docker/distribution requires a signature even if the incoming data uses the nominally unsigned DockerV2Schema1MediaType.
unsigned, err := json.Marshal(*m)
if err != nil {
return nil, err
}
return AddDummyV2S1Signature(unsigned)
}
// fixManifestLayers, after validating the supplied manifest
// (to use correctly-formatted IDs, and to not have non-consecutive ID collisions in m.History),
// modifies manifest to only have one entry for each layer ID in m.History (deleting the older duplicates,
// both from m.History and m.FSLayers).
// Note that even after this succeeds, m.FSLayers may contain duplicate entries
// (for Dockerfile operations which change the configuration but not the filesystem).
func (m *Schema1) fixManifestLayers() error {
// m.initialize() has verified that len(m.FSLayers) == len(m.History)
for _, compat := range m.ExtractedV1Compatibility {
if err := validateV1ID(compat.ID); err != nil {
return err
}
}
if m.ExtractedV1Compatibility[len(m.ExtractedV1Compatibility)-1].Parent != "" {
return errors.New("Invalid parent ID in the base layer of the image")
}
// check general duplicates to error instead of a deadlock
idmap := set.New[string]()
var lastID string
for _, img := range m.ExtractedV1Compatibility {
// skip IDs that appear after each other, we handle those later
if img.ID != lastID && idmap.Contains(img.ID) {
return fmt.Errorf("ID %+v appears multiple times in manifest", img.ID)
}
lastID = img.ID
idmap.Add(lastID)
}
// backwards loop so that we keep the remaining indexes after removing items
for i := len(m.ExtractedV1Compatibility) - 2; i >= 0; i-- {
if m.ExtractedV1Compatibility[i].ID == m.ExtractedV1Compatibility[i+1].ID { // repeated ID. remove and continue
m.FSLayers = slices.Delete(m.FSLayers, i, i+1)
m.History = slices.Delete(m.History, i, i+1)
m.ExtractedV1Compatibility = slices.Delete(m.ExtractedV1Compatibility, i, i+1)
} else if m.ExtractedV1Compatibility[i].Parent != m.ExtractedV1Compatibility[i+1].ID {
return fmt.Errorf("Invalid parent ID. Expected %v, got %v", m.ExtractedV1Compatibility[i+1].ID, m.ExtractedV1Compatibility[i].Parent)
}
}
return nil
}
var validHex = regexp.Delayed(`^([a-f0-9]{64})$`)
func validateV1ID(id string) error {
if ok := validHex.MatchString(id); !ok {
return fmt.Errorf("image ID %q is invalid", id)
}
return nil
}
// Inspect returns various information for (skopeo inspect) parsed from the manifest and configuration.
func (m *Schema1) Inspect(_ func(types.BlobInfo) ([]byte, error)) (*types.ImageInspectInfo, error) {
s1 := &Schema2V1Image{}
if err := json.Unmarshal([]byte(m.History[0].V1Compatibility), s1); err != nil {
return nil, err
}
layerInfos := m.LayerInfos()
i := &types.ImageInspectInfo{
Tag: m.Tag,
Created: &s1.Created,
DockerVersion: s1.DockerVersion,
Architecture: s1.Architecture,
Variant: s1.Variant,
Os: s1.OS,
Layers: layerInfosToStrings(layerInfos),
LayersData: imgInspectLayersFromLayerInfos(layerInfos),
Author: s1.Author,
}
if s1.Config != nil {
i.Labels = s1.Config.Labels
i.Env = s1.Config.Env
}
return i, nil
}
// ToSchema2Config builds a schema2-style configuration blob using the supplied diffIDs.
func (m *Schema1) ToSchema2Config(diffIDs []digest.Digest) ([]byte, error) {
// Convert the schema 1 compat info into a schema 2 config, constructing some of the fields
// that aren't directly comparable using info from the manifest.
if len(m.History) == 0 {
return nil, errors.New("image has no layers")
}
s1 := Schema2V1Image{}
config := []byte(m.History[0].V1Compatibility)
err := json.Unmarshal(config, &s1)
if err != nil {
return nil, fmt.Errorf("decoding configuration: %w", err)
}
// Images created with versions prior to 1.8.3 require us to re-encode the encoded object,
// adding some fields that aren't "omitempty".
if s1.DockerVersion != "" && versions.LessThan(s1.DockerVersion, "1.8.3") {
config, err = json.Marshal(&s1)
if err != nil {
return nil, fmt.Errorf("re-encoding compat image config %#v: %w", s1, err)
}
}
// Build the history.
convertedHistory := []Schema2History{}
for _, compat := range m.ExtractedV1Compatibility {
hitem := Schema2History{
Created: compat.Created,
CreatedBy: strings.Join(compat.ContainerConfig.Cmd, " "),
Author: compat.Author,
Comment: compat.Comment,
EmptyLayer: compat.ThrowAway,
}
convertedHistory = append([]Schema2History{hitem}, convertedHistory...)
}
// Build the rootfs information. We need the decompressed sums that we've been
// calculating to fill in the DiffIDs. It's expected (but not enforced by us)
// that the number of diffIDs corresponds to the number of non-EmptyLayer
// entries in the history.
rootFS := &Schema2RootFS{
Type: "layers",
DiffIDs: diffIDs,
}
// And now for some raw manipulation.
raw := make(map[string]*json.RawMessage)
err = json.Unmarshal(config, &raw)
if err != nil {
return nil, fmt.Errorf("re-decoding compat image config %#v: %w", s1, err)
}
// Drop some fields.
delete(raw, "id")
delete(raw, "parent")
delete(raw, "parent_id")
delete(raw, "layer_id")
delete(raw, "throwaway")
delete(raw, "Size")
// Add the history and rootfs information.
rootfs, err := json.Marshal(rootFS)
if err != nil {
return nil, fmt.Errorf("error encoding rootfs information %#v: %v", rootFS, err)
}
rawRootfs := json.RawMessage(rootfs)
raw["rootfs"] = &rawRootfs
history, err := json.Marshal(convertedHistory)
if err != nil {
return nil, fmt.Errorf("error encoding history information %#v: %v", convertedHistory, err)
}
rawHistory := json.RawMessage(history)
raw["history"] = &rawHistory
// Encode the result.
config, err = json.Marshal(raw)
if err != nil {
return nil, fmt.Errorf("error re-encoding compat image config %#v: %v", s1, err)
}
return config, nil
}
// ImageID computes an ID which can uniquely identify this image by its contents.
func (m *Schema1) ImageID(diffIDs []digest.Digest) (string, error) {
image, err := m.ToSchema2Config(diffIDs)
if err != nil {
return "", err
}
return digest.FromBytes(image).Hex(), nil
}
|