Files
artifactapi/internal/provider/ansible/local.go
T
unkin-agent a00e45c3a7 Add local ansible Galaxy repositories (#157)
Internal Ansible collections can only be installed from galaxy.ansible.com through the remote; there is nowhere to publish them for ansible-galaxy.

- add the Galaxy v3 publish endpoint (multipart, base64 parts) with an import task that is already completed
- read collection_info from MANIFEST.json under size and decompression caps, rejecting escaping paths and links
- serve collection, paginated versions (root-relative links, as Galaxy does), version detail and artifact downloads
- reject an existing namespace.name:version with 409
- add docs/ansible/local.md, a UI snippet and an e2e-docker case driving real ansible-galaxy

requires unkin/artifactapi#137

Reviewed-on: #157
Co-authored-by: unkin-agent <unkin-agent@unkin.net>
Co-committed-by: unkin-agent <unkin-agent@unkin.net>
2026-10-18 18:01:09 +11:00

348 lines
11 KiB
Go

package ansible
import (
"archive/tar"
"compress/gzip"
"encoding/json"
"errors"
"fmt"
"io"
"net/url"
"path"
"regexp"
"strconv"
"strings"
"time"
)
// Local ansible repos speak the Galaxy v3 publish and install API. Per version:
// artifacts/<ns>-<name>-<version>.tar.gz collection tarball (immutable)
// _index/<ns>/<name>/<version> Entry JSON; gates visibility
const (
IndexPrefix = "_index/"
ArtifactPrefix = "artifacts/"
PublishPath = "api/v3/artifacts/collections/"
)
var (
NameRe = regexp.MustCompile(`^[a-z][a-z0-9_]{0,63}$`)
VersionRe = regexp.MustCompile(`^(0|[1-9]\d*)\.(0|[1-9]\d*)\.(0|[1-9]\d*)(-[0-9A-Za-z.-]+)?(\+[0-9A-Za-z.-]+)?$`)
)
var (
MaxManifestBytes int64 = 1 << 20
MaxScanBytes int64 = 512 << 20
)
// Entry is one published collection version.
type Entry struct {
Namespace string `json:"namespace"`
Name string `json:"name"`
Version string `json:"version"`
SHA256 string `json:"sha256"`
Size int64 `json:"size"`
Metadata json.RawMessage `json:"metadata"`
CreatedAt time.Time `json:"created_at"`
}
func (e Entry) Filename() string { return e.Namespace + "-" + e.Name + "-" + e.Version + ".tar.gz" }
// IndexPath is the entry's hidden index file.
func (e Entry) IndexPath() string { return IndexPrefix + e.Namespace + "/" + e.Name + "/" + e.Version }
// TaskID names the import task returned on publish.
func (e Entry) TaskID() string { return e.Namespace + "-" + e.Name + "-" + e.Version }
// ParseCollection reads collection_info from a collection tarball's top-level
// MANIFEST.json. Every member is checked, so a tarball whose paths or links
// escape the collection root is rejected even after the manifest.
func ParseCollection(r io.Reader) (*Entry, error) {
zr, err := gzip.NewReader(r)
if err != nil {
return nil, fmt.Errorf("not a gzip tarball: %w", err)
}
scan := &io.LimitedReader{R: zr, N: MaxScanBytes}
tr := tar.NewReader(scan)
var e *Entry
for {
hdr, err := tr.Next()
if scan.N <= 0 {
return nil, errors.New("collection exceeds decompressed size limit")
}
if err == io.EOF {
break
}
if err != nil {
return nil, fmt.Errorf("read tarball: %w", err)
}
name, err := memberPath(hdr)
if err != nil {
return nil, err
}
if name != "MANIFEST.json" || hdr.Typeflag != tar.TypeReg {
continue
}
if e != nil {
return nil, errors.New("duplicate MANIFEST.json")
}
if e, err = parseManifest(io.LimitReader(tr, MaxManifestBytes+1)); err != nil {
return nil, err
}
}
if e == nil {
return nil, errors.New("no MANIFEST.json in collection tarball")
}
return e, nil
}
// memberPath returns hdr's cleaned name, rejecting absolute or escaping
// paths and links, and anything but files, directories and symlinks.
func memberPath(hdr *tar.Header) (string, error) {
name := path.Clean(strings.TrimPrefix(hdr.Name, "./"))
if escapes(name) {
return "", fmt.Errorf("tarball member %q escapes the collection root", hdr.Name)
}
switch hdr.Typeflag {
case tar.TypeReg, tar.TypeDir:
case tar.TypeSymlink:
if path.IsAbs(hdr.Linkname) || escapes(path.Join(path.Dir(name), hdr.Linkname)) {
return "", fmt.Errorf("tarball symlink %q escapes the collection root", hdr.Name)
}
default:
return "", fmt.Errorf("tarball member %q has unsupported type %q", hdr.Name, hdr.Typeflag)
}
return name, nil
}
func escapes(p string) bool {
return path.IsAbs(p) || p == ".." || strings.HasPrefix(p, "../") || strings.Contains(p, "\\")
}
func parseManifest(r io.Reader) (*Entry, error) {
raw, err := io.ReadAll(r)
if err != nil {
return nil, fmt.Errorf("read MANIFEST.json: %w", err)
}
if int64(len(raw)) > MaxManifestBytes {
return nil, fmt.Errorf("MANIFEST.json exceeds %d bytes", MaxManifestBytes)
}
var doc struct {
CollectionInfo map[string]json.RawMessage `json:"collection_info"`
}
if err := json.Unmarshal(raw, &doc); err != nil || doc.CollectionInfo == nil {
return nil, errors.New("MANIFEST.json has no collection_info")
}
info := doc.CollectionInfo
e := &Entry{}
for k, dst := range map[string]*string{"namespace": &e.Namespace, "name": &e.Name, "version": &e.Version} {
if err := json.Unmarshal(info[k], dst); err != nil {
return nil, fmt.Errorf("collection_info.%s is not a string", k)
}
}
if !NameRe.MatchString(e.Namespace) || !NameRe.MatchString(e.Name) || !VersionRe.MatchString(e.Version) {
return nil, fmt.Errorf("invalid collection %q.%q version %q", e.Namespace, e.Name, e.Version)
}
deps := map[string]string{}
if d := info["dependencies"]; len(d) > 0 && string(d) != "null" {
if err := json.Unmarshal(d, &deps); err != nil {
return nil, errors.New("collection_info.dependencies must map collection names to version ranges")
}
}
for k := range deps {
ns, n, ok := strings.Cut(k, ".")
if !ok || !NameRe.MatchString(ns) || !NameRe.MatchString(n) {
return nil, fmt.Errorf("invalid dependency %q", k)
}
}
info["dependencies"], _ = json.Marshal(deps)
if e.Metadata, err = json.Marshal(info); err != nil {
return nil, err
}
return e, nil
}
// APIPath is a parsed Galaxy v3 request path below a repo.
type APIPath struct {
Kind APIKind
Namespace, Name, Version string
TaskID string
}
type APIKind int
const (
APIUnknown APIKind = iota
APIRoot
APICollection
APIVersions
APIVersion
APIImport
)
// ParseAPIPath parses api/, api/v3/collections/<ns>/<name>/[versions/[<v>/]]
// and api/v3/imports/collections/<id>/, with or without trailing slashes.
func ParseAPIPath(p string) APIPath {
p = strings.TrimSuffix(p, "/")
if p == "api" {
return APIPath{Kind: APIRoot}
}
if id, ok := strings.CutPrefix(p, "api/v3/imports/collections/"); ok && id != "" && !strings.Contains(id, "/") {
return APIPath{Kind: APIImport, TaskID: id}
}
rest, ok := strings.CutPrefix(p, "api/v3/collections/")
if !ok {
return APIPath{}
}
segs := strings.Split(rest, "/")
if len(segs) < 2 || !NameRe.MatchString(segs[0]) || !NameRe.MatchString(segs[1]) {
return APIPath{}
}
ap := APIPath{Namespace: segs[0], Name: segs[1]}
switch {
case len(segs) == 2:
ap.Kind = APICollection
case len(segs) == 3 && segs[2] == "versions":
ap.Kind = APIVersions
case len(segs) == 4 && segs[2] == "versions" && VersionRe.MatchString(segs[3]):
ap.Kind, ap.Version = APIVersion, segs[3]
default:
return APIPath{}
}
return ap
}
// Root is the API root ansible-galaxy discovers.
func Root() map[string]any {
return map[string]any{"available_versions": map[string]string{"v3": "v3/"}}
}
func CollectionPath(ns, name string) string { return "/api/v3/collections/" + ns + "/" + name + "/" }
func VersionPath(ns, name, version string) string {
return CollectionPath(ns, name) + "versions/" + url.PathEscape(version) + "/"
}
// VersionRef is one entry of a versions list.
type VersionRef struct {
Version string `json:"version"`
CreatedAt string `json:"created_at,omitempty"`
UpdatedAt string `json:"updated_at,omitempty"`
}
func (e Entry) Ref() VersionRef {
t := e.CreatedAt.UTC().Format(time.RFC3339Nano)
return VersionRef{Version: e.Version, CreatedAt: t, UpdatedAt: t}
}
// CollectionDoc describes a collection from its versions; base is the repo
// URL the hrefs are built on.
func CollectionDoc(base, ns, name string, refs []VersionRef) map[string]any {
var created, updated time.Time
for _, r := range refs {
if t, err := time.Parse(time.RFC3339Nano, r.CreatedAt); err == nil && (created.IsZero() || t.Before(created)) {
created = t
}
if t, err := time.Parse(time.RFC3339Nano, r.UpdatedAt); err == nil && t.After(updated) {
updated = t
}
}
return map[string]any{
"href": base + CollectionPath(ns, name),
"namespace": ns,
"name": name,
"deprecated": false,
"versions_url": base + CollectionPath(ns, name) + "versions/",
"created_at": formatTime(created),
"updated_at": formatTime(updated),
}
}
func formatTime(t time.Time) string {
if t.IsZero() {
return ""
}
return t.UTC().Format(time.RFC3339Nano)
}
// Page reads limit/offset as the remote forwards them, defaulting to 100/0.
func Page(q url.Values) (limit, offset int) {
limit, offset = maxPageLimit, 0
if v, ok := pageLimit(q.Get("limit")); ok {
limit, _ = strconv.Atoi(v)
}
if v, ok := pageOffset(q.Get("offset")); ok {
offset, _ = strconv.Atoi(v)
}
return limit, offset
}
// VersionsPage is one page of refs in Galaxy v3 list form. Links are
// root-relative like Galaxy's, since ansible-galaxy splices them onto the
// server URL.
func VersionsPage(base, ns, name string, refs []VersionRef, limit, offset int) map[string]any {
list := rootPath(base) + CollectionPath(ns, name) + "versions/"
link := func(off int) any {
return list + "?" + url.Values{"limit": {strconv.Itoa(limit)}, "offset": {strconv.Itoa(off)}}.Encode()
}
offset = min(offset, len(refs))
var previous, next any
if offset > 0 {
previous = link(max(offset-limit, 0))
}
if offset+limit < len(refs) {
next = link(offset + limit)
}
data := []map[string]any{}
for _, r := range refs[offset:min(offset+limit, len(refs))] {
data = append(data, map[string]any{
"version": r.Version,
"href": base + VersionPath(ns, name, r.Version),
"created_at": r.CreatedAt,
"updated_at": r.UpdatedAt,
"marks": []string{},
})
}
return map[string]any{
"meta": map[string]any{"count": len(refs)},
"links": map[string]any{"first": link(0), "previous": previous, "next": next, "last": link(max(len(refs)-1, 0) / limit * limit)},
"data": data,
}
}
// rootPath drops the scheme and host from base.
func rootPath(base string) string {
if u, err := url.Parse(base); err == nil {
return u.EscapedPath()
}
return base
}
// VersionDoc is a version's detail; the tarball is served from the repo at
// artifacts/<filename>.
func VersionDoc(base string, e Entry) map[string]any {
t := e.CreatedAt.UTC().Format(time.RFC3339Nano)
return map[string]any{
"href": base + VersionPath(e.Namespace, e.Name, e.Version),
"namespace": map[string]string{"name": e.Namespace},
"collection": map[string]string{"name": e.Name},
"name": e.Name,
"version": e.Version,
"download_url": base + "/" + ArtifactPrefix + e.Filename(),
"artifact": map[string]any{"filename": e.Filename(), "sha256": e.SHA256, "size": e.Size},
"metadata": e.Metadata,
"signatures": []string{},
"created_at": t,
"updated_at": t,
}
}
// ImportTask is a finished import: publishing completes synchronously.
func ImportTask(e Entry) map[string]any {
t := e.CreatedAt.UTC().Format(time.RFC3339Nano)
return map[string]any{
"id": e.TaskID(), "state": "completed", "created_at": t, "started_at": t, "finished_at": t,
"messages": []any{}, "error": nil,
}
}