banger/internal/imagepull/flatten.go
Thales Maciel 78376ba6ec
Phase 1: imagepull package — pull, flatten, ext4
New internal/imagepull/ subpackage. Three concerns, each
independently testable:

Pull (imagepull.go):
 - github.com/google/go-containerregistry's remote.Image with the
   linux/amd64 platform pinned. Anonymous pulls only for v1.
 - Layer blobs cached on disk via cache.NewFilesystemCache under
   <cacheDir>/blobs/sha256/<hex> — OCI-standard layout so
   skopeo/crane could co-exist later.
 - Eagerly touches every layer once so network errors surface at
   Pull time, not deep in Flatten.

Flatten (flatten.go):
 - Replays layers oldest-first into destDir.
 - Whiteout-aware: .wh.<name> deletes the named entry,
   .wh..wh..opq wipes the parent directory's contents from prior
   layers.
 - Path-traversal hardening mirrored from kernelcat extractTar:
   reject .., absolute paths, and symlinks/hardlinks whose
   resolved target escapes destDir.
 - Handles tar.TypeReg, TypeDir, TypeSymlink, TypeLink. Skips
   device/fifo nodes silently (need privilege; udev/devtmpfs
   handles them in the guest).

BuildExt4 (ext4.go):
 - Truncates outFile to sizeBytes, then runs `mkfs.ext4 -F -d
   <srcDir> -E root_owner=0:0`. No mount, no sudo, no loopback.
 - 64 MiB floor; callers handle real sizing with content-aware
   headroom.
 - File ownership in the resulting ext4 reflects srcDir's on-disk
   ownership — runner's uid/gid since extraction was unprivileged.
   Documented in package doc as a Phase A v1 limitation; Phase B
   will add a debugfs- or tar2ext4-based ownership fixup.

paths.Layout gains OCICacheDir at $XDG_CACHE_HOME/banger/oci/,
ensured at startup alongside the other dirs.

Tests use go-containerregistry's in-process registry to push and
pull synthetic multi-layer images. Cover: layer caching round-trip,
whiteout + opaque-marker handling, path-traversal rejection, unsafe
symlink rejection, real mkfs.ext4 round-trip (skipped if mkfs.ext4
absent), and tiny-size rejection.

go-containerregistry v0.21.5 added as a direct dep, plus its
transitive closure (containerd/stargz, opencontainers/go-digest,
docker/cli config helpers, etc).

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-16 17:22:13 -03:00

201 lines
5.5 KiB
Go

package imagepull
import (
"archive/tar"
"context"
"errors"
"fmt"
"io"
"os"
"path/filepath"
"strings"
)
const (
whiteoutPrefix = ".wh."
// whiteoutOpaque marks the parent directory as opaque: every entry
// from previous layers should be removed, but entries from the
// current layer (siblings of this marker) are preserved.
whiteoutOpaque = ".wh..wh..opq"
)
// Flatten replays the image's layers in oldest-first order into destDir.
// destDir must exist and ideally be empty. Path-traversal members and
// symlink targets that escape destDir are rejected.
//
// File ownership in destDir reflects the running user, not the tar
// header's uid/gid (Phase A v1 limitation; see package docs).
func Flatten(ctx context.Context, img PulledImage, destDir string) error {
absDest, err := filepath.Abs(destDir)
if err != nil {
return err
}
layers, err := img.Image.Layers()
if err != nil {
return fmt.Errorf("read layers: %w", err)
}
for i, layer := range layers {
if err := ctx.Err(); err != nil {
return err
}
if err := applyLayer(layer, absDest); err != nil {
return fmt.Errorf("apply layer %d/%d: %w", i+1, len(layers), err)
}
}
return nil
}
func applyLayer(layer interface {
Uncompressed() (io.ReadCloser, error)
}, dest string) error {
rc, err := layer.Uncompressed()
if err != nil {
return err
}
defer rc.Close()
tr := tar.NewReader(rc)
for {
hdr, err := tr.Next()
if err == io.EOF {
return nil
}
if err != nil {
return fmt.Errorf("read tar entry: %w", err)
}
if err := applyEntry(tr, hdr, dest); err != nil {
return err
}
}
}
func applyEntry(tr *tar.Reader, hdr *tar.Header, dest string) error {
rel := filepath.Clean(hdr.Name)
if rel == "." || rel == string(filepath.Separator) {
return nil
}
if filepath.IsAbs(rel) || rel == ".." || strings.HasPrefix(rel, ".."+string(filepath.Separator)) {
return fmt.Errorf("unsafe path in layer: %q", hdr.Name)
}
base := filepath.Base(rel)
parent := filepath.Dir(rel)
// Whiteouts come in two flavors: opaque-dir markers and per-file
// deletes. Both are resolved relative to the parent directory.
if base == whiteoutOpaque {
parentAbs, err := safeJoin(dest, parent)
if err != nil {
return err
}
return clearDirContents(parentAbs)
}
if strings.HasPrefix(base, whiteoutPrefix) {
target := strings.TrimPrefix(base, whiteoutPrefix)
victim, err := safeJoin(dest, filepath.Join(parent, target))
if err != nil {
return err
}
if err := os.RemoveAll(victim); err != nil && !errors.Is(err, os.ErrNotExist) {
return fmt.Errorf("apply whiteout %s: %w", hdr.Name, err)
}
return nil
}
abs, err := safeJoin(dest, rel)
if err != nil {
return err
}
switch hdr.Typeflag {
case tar.TypeDir:
return os.MkdirAll(abs, 0o755)
case tar.TypeReg:
if err := os.MkdirAll(filepath.Dir(abs), 0o755); err != nil {
return err
}
// Replace any prior file/dir in this slot — later layers
// shadow earlier ones.
if err := os.RemoveAll(abs); err != nil && !errors.Is(err, os.ErrNotExist) {
return err
}
f, err := os.OpenFile(abs, os.O_CREATE|os.O_WRONLY|os.O_TRUNC, os.FileMode(hdr.Mode)|0o600)
if err != nil {
return err
}
if _, err := io.Copy(f, tr); err != nil {
_ = f.Close()
return err
}
return f.Close()
case tar.TypeSymlink:
if err := os.MkdirAll(filepath.Dir(abs), 0o755); err != nil {
return err
}
// Resolve the link target relative to the link's parent and
// require that it stays inside dest. Absolute targets that
// resolve outside dest are also rejected.
resolved := hdr.Linkname
if !filepath.IsAbs(resolved) {
resolved = filepath.Join(filepath.Dir(abs), resolved)
}
resolved = filepath.Clean(resolved)
if resolved != dest && !strings.HasPrefix(resolved, dest+string(filepath.Separator)) {
return fmt.Errorf("unsafe symlink in layer: %q -> %q", hdr.Name, hdr.Linkname)
}
if err := os.RemoveAll(abs); err != nil && !errors.Is(err, os.ErrNotExist) {
return err
}
return os.Symlink(hdr.Linkname, abs)
case tar.TypeLink:
// Hardlink: target must already exist inside dest from this or
// a previous layer, and must not escape.
linkTarget, err := safeJoin(dest, filepath.Clean(hdr.Linkname))
if err != nil {
return err
}
if _, err := os.Lstat(linkTarget); err != nil {
return fmt.Errorf("hardlink target %q missing: %w", hdr.Linkname, err)
}
if err := os.MkdirAll(filepath.Dir(abs), 0o755); err != nil {
return err
}
if err := os.RemoveAll(abs); err != nil && !errors.Is(err, os.ErrNotExist) {
return err
}
return os.Link(linkTarget, abs)
default:
// TypeChar / TypeBlock / TypeFifo / TypeXGlobalHeader / etc.
// Container layers occasionally include /dev nodes — they need
// privilege we don't have. Skip silently; udev/devtmpfs in the
// guest will create them at boot.
return nil
}
}
// safeJoin returns dest+rel after verifying the result lies under dest.
func safeJoin(dest, rel string) (string, error) {
joined := filepath.Join(dest, rel)
if joined != dest && !strings.HasPrefix(joined, dest+string(filepath.Separator)) {
return "", fmt.Errorf("unsafe path: %q escapes %q", rel, dest)
}
return joined, nil
}
// clearDirContents removes every entry under dir but leaves dir itself.
// Used for opaque-whiteout markers.
func clearDirContents(dir string) error {
entries, err := os.ReadDir(dir)
if err != nil {
if errors.Is(err, os.ErrNotExist) {
return os.MkdirAll(dir, 0o755)
}
return err
}
for _, entry := range entries {
if err := os.RemoveAll(filepath.Join(dir, entry.Name())); err != nil {
return err
}
}
return nil
}