fix: serve the whole results service from the cache server (#1141)

`ACTIONS_RESULTS_URL` names one origin serving every `github.actions.results.api.v1` service. Gitea serves the artifact half and this runner the cache half, so announcing `ACTIONS_CACHE_SERVICE_V2` while that URL pointed at Gitea was a promise the environment could not keep, and `docker buildx` posted its cache calls at Gitea and got a 404. The cache server now forwards the artifact half to the instance each job registers with, so it is the whole results service and jobs are pointed at it. The announcement follows, and the bundle patch follows the cache URL instead. Also fixes three things no JavaScript client reached: camelCase in the v2 responses where the Go clients read proto names, the missing `x-ms-request-id` on blob uploads that panics buildkit, and `cache.external_server` passed through without the trailing slash the v1 client concatenates onto. Tests run the real actions against the services they look for: `actions/cache` over both API versions, the artifact actions up and back down through the forwarding, and `setup-node`. The regression itself is covered by asserting that whatever a job is handed as `ACTIONS_RESULTS_URL` answers a cache service call. Fixes https://gitea.com/gitea/runner/issues/1139 Reviewed-on: https://gitea.com/gitea/runner/pulls/1141 Reviewed-by: bircni <bircni@icloud.com> Co-authored-by: silverwind <2021+silverwind@noreply.gitea.com>
2026-08-06 00:44:22 +02:00 · 2026-08-03 16:41:53 +00:00
parent 55a625f733
commit 3618385b28
15 changed files with 539 additions and 141 deletions
--- a/act/artifactcache/handler.go
+++ b/act/artifactcache/handler.go
@@ -52,7 +52,13 @@ type credKey struct{}
 // poison another repo's cache, even from inside a container that reaches the
 // cache server over the docker bridge network.
 type JobCredential struct {
-	Repo string
+	Repo string `json:"repo"`
+
+	// Results is the instance whose artifact service this server forwards for the job, and
+	// InsecureTLS how the runner reaches it; see results.go. The tags are the wire format a
+	// remote runner registers with.
+	Results     string `json:"results"`
+	InsecureTLS bool   `json:"insecure_tls"`
 }

 // credEntry holds a registered job's credential along with an active
@@ -165,6 +171,7 @@ func StartHandler(dir, outboundIP string, port uint16, internalSecret string, lo
 	router.POST(internalPath+"/register", h.internalAuth(h.internalRegister))
 	router.POST(internalPath+"/revoke", h.internalAuth(h.internalRevoke))
 	h.registerV2Routes(router)
+	router.NotFound = http.HandlerFunc(h.forwardOrNotFound)

 	h.router = router

@@ -211,10 +218,11 @@ func (h *Handler) ExternalURL() string {
 // is only accepted while the job is running.
 //
 // Registrations are reference-counted: if a token is already registered, the
-// existing repo is kept and the refcount is incremented. The entry is
-// removed only when every revoker returned by RegisterJob has been called.
+// credential it was registered with is kept and the refcount is incremented.
+// The entry is removed only when every revoker returned by RegisterJob has
+// been called.
 // This keeps a stray re-registration from silently revoking a live job.
-func (h *Handler) RegisterJob(token, repo string) func() {
+func (h *Handler) RegisterJob(token string, cred JobCredential) func() {
 	if h == nil || token == "" {
 		return func() {}
 	}
@@ -223,7 +231,7 @@ func (h *Handler) RegisterJob(token, repo string) func() {
 		existing.refs++
 	} else {
 		h.creds[token] = &credEntry{
-			cred: JobCredential{Repo: repo},
+			cred: cred,
 			refs: 1,
 		}
 	}
@@ -619,7 +627,7 @@ func (h *Handler) internalAuth(handler httprouter.Handle) httprouter.Handle {

 type internalRegisterBody struct {
 	Token string `json:"token"`
-	Repo  string `json:"repo"`
+	JobCredential
 }

 type internalRevokeBody struct {
@@ -627,6 +635,15 @@ type internalRevokeBody struct {
 }

 // POST /_internal/register
+// ResultsURL is what a job registered with cred should be given as ACTIONS_RESULTS_URL, or "" when
+// the credential names no instance to forward the artifact half to.
+func (h *Handler) ResultsURL(cred JobCredential) string {
+	if h == nil || cred.Results == "" {
+		return ""
+	}
+	return h.ExternalURL()
+}
+
 func (h *Handler) internalRegister(w http.ResponseWriter, r *http.Request, _ httprouter.Params) {
 	var body internalRegisterBody
 	if err := json.NewDecoder(r.Body).Decode(&body); err != nil {
@@ -637,8 +654,9 @@ func (h *Handler) internalRegister(w http.ResponseWriter, r *http.Request, _ htt
 		h.responseJSON(w, r, http.StatusBadRequest, errors.New("token is required"))
 		return
 	}
-	h.RegisterJob(body.Token, body.Repo)
-	h.responseJSON(w, r, http.StatusOK)
+	h.RegisterJob(body.Token, body.JobCredential)
+	// A server too old to forward answers without this, which is how the caller knows.
+	h.responseJSON(w, r, http.StatusOK, map[string]any{"results_url": h.ResultsURL(body.JobCredential)})
 }

 // POST /_internal/revoke
--- a/act/artifactcache/handler_test.go
+++ b/act/artifactcache/handler_test.go
@@ -52,7 +52,7 @@ func TestHandler(t *testing.T) {
 	dir := filepath.Join(t.TempDir(), "artifactcache")
 	handler, err := StartHandler(dir, "", 0, "", nil)
 	require.NoError(t, err)
-	handler.RegisterJob(testToken, testRepo)
+	handler.RegisterJob(testToken, JobCredential{Repo: testRepo})

 	base := fmt.Sprintf("%s%s", handler.ExternalURL(), apiPath)

@@ -890,7 +890,7 @@ func TestHandler_UnregisterRevokes(t *testing.T) {
 	require.NoError(t, err)
 	defer handler.Close()

-	unregister := handler.RegisterJob("tmp-token", testRepo)
+	unregister := handler.RegisterJob("tmp-token", JobCredential{Repo: testRepo})

 	base := handler.ExternalURL() + apiPath
 	req, err := http.NewRequest(http.MethodGet, base+"/cache?keys=x&version=y", nil)
@@ -920,8 +920,8 @@ func TestHandler_CrossRepoIsolation(t *testing.T) {
 	handler, err := StartHandler(dir, "", 0, "", nil)
 	require.NoError(t, err)
 	defer handler.Close()
-	handler.RegisterJob("token-a", "owner/repoA")
-	handler.RegisterJob("token-b", "owner/repoB")
+	handler.RegisterJob("token-a", JobCredential{Repo: "owner/repoA"})
+	handler.RegisterJob("token-b", JobCredential{Repo: "owner/repoB"})

 	base := handler.ExternalURL() + apiPath
 	key := "shared-key"
@@ -986,7 +986,7 @@ func TestHandler_ArtifactSignature(t *testing.T) {
 	handler, err := StartHandler(dir, "", 0, "", nil)
 	require.NoError(t, err)
 	defer handler.Close()
-	handler.RegisterJob(testToken, testRepo)
+	handler.RegisterJob(testToken, JobCredential{Repo: testRepo})

 	base := handler.ExternalURL() + apiPath

@@ -1059,7 +1059,7 @@ func TestHandler_ArtifactSignatureDownload(t *testing.T) {
 	handler, err := StartHandler(dir, "", 0, "", nil)
 	require.NoError(t, err)
 	defer handler.Close()
-	handler.RegisterJob(testToken, testRepo)
+	handler.RegisterJob(testToken, JobCredential{Repo: testRepo})

 	base := handler.ExternalURL() + apiPath
 	key := "download-key"
@@ -1100,8 +1100,8 @@ func TestHandler_RegisterJob_RefCounted(t *testing.T) {
 	require.NoError(t, err)
 	defer handler.Close()

-	first := handler.RegisterJob("shared", testRepo)
-	second := handler.RegisterJob("shared", testRepo)
+	first := handler.RegisterJob("shared", JobCredential{Repo: testRepo})
+	second := handler.RegisterJob("shared", JobCredential{Repo: testRepo})

 	base := handler.ExternalURL() + apiPath
 	probe := func() int {
@@ -1131,8 +1131,8 @@ func TestHandler_GC_PerRepoDedup(t *testing.T) {
 	handler, err := StartHandler(dir, "", 0, "", nil)
 	require.NoError(t, err)
 	defer handler.Close()
-	handler.RegisterJob("tok-a", "owner/repoA")
-	handler.RegisterJob("tok-b", "owner/repoB")
+	handler.RegisterJob("tok-a", JobCredential{Repo: "owner/repoA"})
+	handler.RegisterJob("tok-b", JobCredential{Repo: "owner/repoB"})

 	key := "shared-dedup-key"
 	version := "c19da02a2bd7e77277f1ac29ab45c09b7d46a4ee758284e26bb3045ad11d9d20"
--- a/act/artifactcache/handler_v2.go
+++ b/act/artifactcache/handler_v2.go
@@ -23,6 +23,9 @@ import (
 // endpoints, and uploads the archive to the returned URL with the Azure blob protocol.
 // Both API versions are served from the same store, so a repository keeps its cache
 // when a workflow moves between action versions.
+//
+// Responses carry the proto field names, which is what Gitea's own results API emits and the only
+// spelling the Go clients parse. The JavaScript toolkit accepts either.
 const (
 	cacheServiceV2Path = "/twirp/github.actions.results.api.v1.CacheService"

@@ -93,8 +96,8 @@ func (h *Handler) v2CreateCacheEntry(w http.ResponseWriter, r *http.Request, _ h
 	}

 	h.responseJSON(w, r, http.StatusOK, map[string]any{
-		"ok":              true,
-		"signedUploadUrl": h.signedURL(blobPath, blobUploadPurpose, cache.ID, time.Now().Add(blobUploadURLTTL)),
+		"ok":                true,
+		"signed_upload_url": h.signedURL(blobPath, blobUploadPurpose, cache.ID, time.Now().Add(blobUploadURLTTL)),
 	})
 }

@@ -134,7 +137,7 @@ func (h *Handler) v2FinalizeCacheEntryUpload(w http.ResponseWriter, r *http.Requ
 	h.responseJSON(w, r, http.StatusOK, map[string]any{
 		"ok": true,
 		// int64 fields travel as strings in the proto JSON mapping.
-		"entryId": strconv.FormatUint(cache.ID, 10),
+		"entry_id": strconv.FormatUint(cache.ID, 10),
 	})
 }

@@ -164,9 +167,9 @@ func (h *Handler) v2GetCacheEntryDownloadURL(w http.ResponseWriter, r *http.Requ
 	}

 	h.responseJSON(w, r, http.StatusOK, map[string]any{
-		"ok":                true,
-		"signedDownloadUrl": h.signedArtifactURL(cache.ID, time.Now().Add(artifactURLTTL)),
-		"matchedKey":        cache.Key,
+		"ok":                  true,
+		"signed_download_url": h.signedArtifactURL(cache.ID, time.Now().Add(artifactURLTTL)),
+		"matched_key":         cache.Key,
 	})
 }

@@ -209,6 +212,8 @@ func (h *Handler) v2UploadBlob(w http.ResponseWriter, r *http.Request, params ht
 		return
 	}

+	// The Azure SDK client dereferences this without checking, so its absence panics the caller.
+	w.Header().Set("x-ms-request-id", strconv.FormatInt(time.Now().UnixNano(), 10))
 	w.WriteHeader(http.StatusCreated)
 }

--- a/act/artifactcache/handler_v2_test.go
+++ b/act/artifactcache/handler_v2_test.go
@@ -45,6 +45,11 @@ func putBlob(t *testing.T, url string, content []byte) int {
 	resp, err := http.DefaultClient.Do(req)
 	require.NoError(t, err)
 	defer resp.Body.Close()
+	if resp.StatusCode == http.StatusCreated {
+		// The Azure SDK client dereferences this header without checking, so a blob upload that
+		// omits it panics the caller rather than failing it.
+		require.NotEmpty(t, resp.Header.Get("x-ms-request-id"))
+	}
 	return resp.StatusCode
 }

@@ -67,7 +72,7 @@ func startTestHandler(t *testing.T) *Handler {
 	handler, err := StartHandler(filepath.Join(t.TempDir(), "artifactcache"), "127.0.0.1", 0, "", nil)
 	require.NoError(t, err)
 	t.Cleanup(func() { _ = handler.Close() })
-	handler.RegisterJob(testToken, testRepo)
+	handler.RegisterJob(testToken, JobCredential{Repo: testRepo})
 	return handler
 }

@@ -78,7 +83,7 @@ func saveV2(t *testing.T, handler *Handler, key, version string, content []byte)

 	created := v2Call(t, handler, testClient, "CreateCacheEntry", map[string]any{"key": key, "version": version})
 	require.Equal(t, true, created["ok"])
-	uploadURL, _ = created["signedUploadUrl"].(string)
+	uploadURL, _ = created["signed_upload_url"].(string)
 	require.NotEmpty(t, uploadURL)
 	require.Equal(t, http.StatusCreated, putBlob(t, uploadURL, content))

@@ -100,7 +105,7 @@ func TestCacheServiceV2RoundTrip(t *testing.T) {

 	finalized, uploadURL := saveV2(t, handler, "deps-v1", "abc123", content)
 	require.Equal(t, true, finalized["ok"])
-	assert.NotEmpty(t, finalized["entryId"])
+	assert.NotEmpty(t, finalized["entry_id"])

 	// The upload URL outlives the finalize call, so replaying it must not poison the entry,
 	// and it is an upload URL only: nothing reads a blob back through it.
@@ -112,8 +117,8 @@ func TestCacheServiceV2RoundTrip(t *testing.T) {

 	got := v2Call(t, handler, testClient, "GetCacheEntryDownloadURL", map[string]any{"key": "deps-v1", "version": "abc123"})
 	require.Equal(t, true, got["ok"])
-	assert.Equal(t, "deps-v1", got["matchedKey"])
-	downloadURL, _ := got["signedDownloadUrl"].(string)
+	assert.Equal(t, "deps-v1", got["matched_key"])
+	downloadURL, _ := got["signed_download_url"].(string)
 	require.NotEmpty(t, downloadURL)
 	assert.Equal(t, content, getURL(t, downloadURL))
 }
@@ -124,7 +129,7 @@ func TestCacheServiceV2BlockUpload(t *testing.T) {
 	handler := startTestHandler(t)

 	created := v2Call(t, handler, testClient, "CreateCacheEntry", map[string]any{"key": "blocks", "version": "v1"})
-	uploadURL, _ := created["signedUploadUrl"].(string)
+	uploadURL, _ := created["signed_upload_url"].(string)
 	require.NotEmpty(t, uploadURL)

 	blocks := map[string][]byte{}
@@ -154,7 +159,7 @@ func TestCacheServiceV2BlockUpload(t *testing.T) {

 	got := v2Call(t, handler, testClient, "GetCacheEntryDownloadURL", map[string]any{"key": "blocks", "version": "v1"})
 	require.Equal(t, true, got["ok"])
-	assert.Equal(t, "hello world!", string(getURL(t, got["signedDownloadUrl"].(string))))
+	assert.Equal(t, "hello world!", string(getURL(t, got["signed_download_url"].(string))))
 }

 func TestCacheServiceV2Lookups(t *testing.T) {
@@ -175,7 +180,7 @@ func TestCacheServiceV2Lookups(t *testing.T) {
 				"key": "deps-zzz", field: []string{"deps-"}, "version": "v1",
 			})
 			require.Equal(t, true, got["ok"])
-			assert.Equal(t, "deps-abc", got["matchedKey"])
+			assert.Equal(t, "deps-abc", got["matched_key"])
 		})
 	}

@@ -190,7 +195,7 @@ func TestCacheServiceV2Lookups(t *testing.T) {
 	t.Run("a prefix of an existing key is still reserved", func(t *testing.T) {
 		reserved := v2Call(t, handler, testClient, "CreateCacheEntry", map[string]any{"key": "deps", "version": "v1"})
 		require.Equal(t, true, reserved["ok"])
-		assert.NotEmpty(t, reserved["signedUploadUrl"])
+		assert.NotEmpty(t, reserved["signed_upload_url"])
 	})

 	t.Run("finalizing without a reservation is not ok", func(t *testing.T) {
@@ -203,7 +208,7 @@ func TestCacheServiceV2Lookups(t *testing.T) {
 	// The size the client declares is what Commit validates the assembled archive against.
 	t.Run("finalizing with the wrong size is not ok", func(t *testing.T) {
 		created := v2Call(t, handler, testClient, "CreateCacheEntry", map[string]any{"key": "wrong-size", "version": "v1"})
-		require.Equal(t, http.StatusCreated, putBlob(t, created["signedUploadUrl"].(string), []byte("four")))
+		require.Equal(t, http.StatusCreated, putBlob(t, created["signed_upload_url"].(string), []byte("four")))

 		got := v2Call(t, handler, testClient, "FinalizeCacheEntryUpload", map[string]any{
 			"key": "wrong-size", "version": "v1", "size_bytes": 99,
@@ -227,7 +232,7 @@ func TestCacheServiceV2Lookups(t *testing.T) {

 	// The cache of one repository must stay invisible to another, as it does for the v1 API.
 	t.Run("another repository sees nothing", func(t *testing.T) {
-		handler.RegisterJob("other-runtime-token", "other/repo")
+		handler.RegisterJob("other-runtime-token", JobCredential{Repo: "other/repo"})
 		otherClient := &http.Client{Transport: &bearerTransport{token: "other-runtime-token"}}

 		got := v2Call(t, handler, otherClient, "GetCacheEntryDownloadURL", map[string]any{"key": "deps-abc", "version": "v1"})
--- a/act/artifactcache/results.go
+++ b/act/artifactcache/results.go
@@ -0,0 +1,59 @@
+// Copyright 2026 The Gitea Authors. All rights reserved.
+// SPDX-License-Identifier: MIT
+
+package artifactcache
+
+import (
+	"crypto/tls"
+	"net/http"
+	"net/http/httputil"
+	"net/url"
+	"strings"
+)
+
+// The results service is one origin serving every github.actions.results.api.v1 service, and
+// Gitea implements only the artifact half of it. Forwarding that half from here makes this origin
+// the whole service, so ACTIONS_RESULTS_URL can point at it truthfully, which is what the clients
+// this runner cannot patch need, docker buildx among them.
+//
+// The instance to forward to travels with the job registration rather than with configuration, so
+// a cache server shared between runners serves each of their instances.
+const artifactServicePath = "/twirp/github.actions.results.api.v1.ArtifactService/"
+
+// forwardOrNotFound is the router's fallback: the artifact service of the instance the job
+// registered with, and the 404 the router would have written otherwise.
+func (h *Handler) forwardOrNotFound(w http.ResponseWriter, r *http.Request) {
+	cred, ok := h.lookupCredential(bearerToken(r))
+	if !ok || cred.Results == "" || !strings.HasPrefix(r.URL.Path, artifactServicePath) {
+		http.NotFound(w, r)
+		return
+	}
+	target, err := url.Parse(strings.TrimSuffix(cred.Results, "/"))
+	if err != nil {
+		h.logger.Errorf("artifact service forward to %q: %v", cred.Results, err)
+		w.WriteHeader(http.StatusBadGateway)
+		return
+	}
+	h.logger.Debugf("%s %s: forwarding to %s", r.Method, r.URL.Path, target)
+
+	proxy := &httputil.ReverseProxy{
+		Rewrite: func(r *httputil.ProxyRequest) {
+			r.SetURL(target)
+			// Gitea builds the URLs it hands back from this Host, and their scheme from the
+			// connection unless a forwarded header overrides it, so artifact bodies go to Gitea
+			// directly and never through here.
+			r.Out.Host = target.Host
+		},
+		ErrorHandler: func(w http.ResponseWriter, _ *http.Request, err error) {
+			h.logger.Warnf("artifact service forward to %s: %v", target, err)
+			w.WriteHeader(http.StatusBadGateway)
+		},
+	}
+	if cred.InsecureTLS {
+		proxy.Transport = insecureTransport
+	}
+	proxy.ServeHTTP(w, r)
+}
+
+// insecureTransport is shared, because a transport per request would pool no connections.
+var insecureTransport = &http.Transport{TLSClientConfig: &tls.Config{InsecureSkipVerify: true}} //nolint:gosec // the runner reaches its instance on the operator's say-so
--- a/act/artifactcache/results_test.go
+++ b/act/artifactcache/results_test.go
@@ -0,0 +1,56 @@
+// Copyright 2026 The Gitea Authors. All rights reserved.
+// SPDX-License-Identifier: MIT
+
+package artifactcache
+
+import (
+	"io"
+	"net/http"
+	"net/http/httptest"
+	"strings"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+// The artifact half is forwarded under the Host Gitea knows itself by, so the URLs it hands back
+// still point at Gitea, and nothing else is proxied.
+func TestFrontResultsService(t *testing.T) {
+	var gotHost, gotPath, gotProto string
+	gitea := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		gotHost, gotPath, gotProto = r.Host, r.URL.Path, r.Header.Get("X-Forwarded-Proto")
+		_, _ = io.WriteString(w, `{"ok":true}`)
+	}))
+	defer gitea.Close()
+
+	handler, err := StartHandler(t.TempDir(), "127.0.0.1", 0, "", nil)
+	require.NoError(t, err)
+	defer handler.Close()
+	const token = "forward-token"
+
+	client := &http.Client{Transport: &bearerTransport{token: token}}
+	post := func(path string) int {
+		req, err := http.NewRequestWithContext(t.Context(), http.MethodPost, handler.ExternalURL()+path, nil)
+		require.NoError(t, err)
+		resp, err := client.Do(req)
+		require.NoError(t, err)
+		resp.Body.Close()
+		return resp.StatusCode
+	}
+
+	assert.Equal(t, http.StatusNotFound, post(artifactServicePath+"CreateArtifact"),
+		"an unregistered token is forwarded nowhere")
+
+	defer handler.RegisterJob(token, JobCredential{Repo: "owner/repo", Results: gitea.URL})()
+
+	assert.Equal(t, http.StatusOK, post(artifactServicePath+"CreateArtifact"))
+	assert.Equal(t, strings.TrimPrefix(gitea.URL, "http://"), gotHost, "Gitea must see the host it mints its URLs from")
+	assert.Empty(t, gotProto, "a forwarded scheme would make an https Gitea mint http URLs")
+	assert.Equal(t, artifactServicePath+"CreateArtifact", gotPath)
+
+	gotPath = ""
+	assert.Equal(t, http.StatusNotFound, post("/twirp/github.actions.results.api.v1.OtherService/Do"))
+	assert.Equal(t, http.StatusNotFound, post("/api/v1/repos/owner/repo"))
+	assert.Empty(t, gotPath, "only the artifact service is forwarded")
+}