fix: stop artifact uploads depending on the cache server reaching Gitea (#1216)

Cache v2 makes the cache server the `ACTIONS_RESULTS_URL` origin, so artifact calls arrived there and were proxied on to Gitea, failing whenever it could not reach the instance.

- Artifact calls are answered with a redirect, so the cache server opens no connection to Gitea. A scheme change or an untrusted instance is still proxied, but there the cache server is the runner itself, which already reaches Gitea.
- Failures answer in twirp, not an empty `502` that clients report as `Unexpected end of JSON input`.
- `cache.v2: false` really points artifacts at Gitea now.
- Cache reservations are bound to the job that made them, so two jobs saving one key cannot commit against each other's upload, and a retry after a lost answer no longer fails a saved entry.
- The toolkit patch, which edits the GitHub-host check out of an action's bundle, was left in the shared checkout where a job running with `runner.patch_actions: false` could inherit it. It is put back after the job's copy.
- `exec` names an origin for the cache v2 it advertises, and masks its runtime token.

Behaviour changes: `no_proxy` no longer exempts `cache.external_server`, and `cache.enabled: false` also stops external registration.

Fixes https://gitea.com/gitea/runner/issues/1208
Fixes https://gitea.com/gitea/runner/issues/1211

Assisted by Claude (Opus 5).

Reviewed-on: https://gitea.com/gitea/runner/pulls/1216
Reviewed-by: bircni <bircni@icloud.com>
Co-authored-by: silverwind <me@silverwind.io>

act/artifactcache/handler.go

@@ -603,12 +603,12 @@ func (h *Handler) bearerAuth(handler httprouter.Handle) httprouter.Handle {
 		h.logger.Debugf("%s %s", r.Method, r.URL.Path)
 		token := bearerToken(r)
 		if token == "" {
-			h.responseJSON(w, r, http.StatusUnauthorized, errors.New("missing bearer token"))
+			h.unauthorized(w, r, errors.New("missing bearer token"))
 			return
 		}
 		cred, ok := h.lookupCredential(token)
 		if !ok {
-			h.responseJSON(w, r, http.StatusUnauthorized, errors.New("unknown bearer token"))
+			h.unauthorized(w, r, errors.New("unknown bearer token"))
 			return
 		}
 		ctx := context.WithValue(r.Context(), credKey{}, cred)
@@ -617,6 +617,14 @@ func (h *Handler) bearerAuth(handler httprouter.Handle) httprouter.Handle {
 	}
 }
 
+func (h *Handler) unauthorized(w http.ResponseWriter, r *http.Request, err error) {
+	if strings.HasPrefix(r.URL.Path, cacheServiceV2Path) {
+		h.twirpError(w, r, twirpUnauthenticated, err)
+		return
+	}
+	h.responseJSON(w, r, http.StatusUnauthorized, err)
+}
+
 // signedAuth authenticates a signed URL. purpose separates the flavours of URL the
 // handler hands out, so one cannot be replayed as another; see computeSignature.
 func (h *Handler) signedAuth(purpose string, handler httprouter.Handle) httprouter.Handle {
@@ -720,6 +728,12 @@ func (h *Handler) internalRevoke(w http.ResponseWriter, r *http.Request, _ httpr
 	h.responseJSON(w, r, http.StatusOK)
 }
 
+// hashedToken fingerprints a job's bearer, so a reservation can tell its own retry from another job.
+func hashedToken(token string) string {
+	sum := sha256.Sum256([]byte(token))
+	return hex.EncodeToString(sum[:])
+}
+
 func bearerToken(r *http.Request) string {
 	auth := r.Header.Get("Authorization")
 	const prefix = "Bearer "
@@ -805,8 +819,8 @@ func findExactCache(db *bolthold.Store, repo, key, version string, complete bool
 	}
 	cache := &Cache{}
 	err := db.FindOne(cache,
-		bolthold.Where("Repo").Eq(repo).
-			And("Key").Eq(key).
+		bolthold.Where("Key").Eq(key).Index("Key").
+			And("Repo").Eq(repo).
 			And("Version").Eq(version).
 			And("Complete").Eq(complete).
 			SortBy(sortBy).Reverse())

act/artifactcache/handler_v2.go

@@ -39,8 +39,8 @@ const (
 
 	blobUploadURLTTL = time.Hour
 
-	// twirpInternal is the only error code that is not the client's fault.
-	twirpInternal = "internal"
+	twirpInternal        = "internal"
+	twirpUnauthenticated = "unauthenticated"
 )
 
 func (h *Handler) registerV2Routes(router *httprouter.Router) {
@@ -56,7 +56,7 @@ func (h *Handler) v2CreateCacheEntry(w http.ResponseWriter, r *http.Request, _ h
 	cred := credFromContext(r.Context())
 	req, err := decodeTwirpRequest[v2CreateRequest](r)
 	if err != nil {
-		h.twirpError(w, r, "malformed_request", err)
+		h.twirpError(w, r, "malformed", err)
 		return
 	}
 	if req.Key == "" || req.Version == "" {
@@ -83,11 +83,30 @@ func (h *Handler) v2CreateCacheEntry(w http.ResponseWriter, r *http.Request, _ h
 		return
 	}
 
+	// A second live reservation is finalized by whichever job calls last, against the other's upload.
+	owner := hashedToken(bearerToken(r))
+	if pending, err := findExactCache(db, cred.Repo, req.Key, req.Version, false); err != nil {
+		h.twirpError(w, r, twirpInternal, err)
+		return
+	} else if pending != nil && pending.UsedAt > time.Now().Add(-uploadStallTimeout).Unix() {
+		if pending.Owner != owner {
+			h.twirpNotOK(w, r)
+			return
+		}
+		h.touch(db, pending)                                // still uploading, so it must not go stale under the sweep
+		h.responseJSON(w, r, http.StatusOK, map[string]any{ // this job retrying its own reservation
+			"ok":                true,
+			"signed_upload_url": h.signedURL(cred, blobPath, blobUploadPurpose, pending.ID, time.Now().Add(blobUploadURLTTL)),
+		})
+		return
+	}
+
 	now := time.Now().Unix()
 	cache := &Cache{
 		Repo:      cred.Repo,
 		Key:       req.Key,
 		Version:   req.Version,
+		Owner:     owner,
 		Size:      -1, // the size is only known at finalize time
 		CreatedAt: now,
 		UsedAt:    now,
@@ -107,7 +126,7 @@ func (h *Handler) v2FinalizeCacheEntryUpload(w http.ResponseWriter, r *http.Requ
 	cred := credFromContext(r.Context())
 	req, err := decodeTwirpRequest[v2FinalizeRequest](r)
 	if err != nil {
-		h.twirpError(w, r, "malformed_request", err)
+		h.twirpError(w, r, "malformed", err)
 		return
 	}
 
@@ -119,6 +138,13 @@ func (h *Handler) v2FinalizeCacheEntryUpload(w http.ResponseWriter, r *http.Requ
 	defer db.Close()
 
 	cache, err := findExactCache(db, cred.Repo, req.Key, req.Version, false)
+	if err == nil && cache != nil && cache.Owner != hashedToken(bearerToken(r)) {
+		cache = nil // not the reservation this job made, so not this job's to commit
+	}
+	if err == nil && cache == nil {
+		// A retry whose first response was lost finds the entry already committed.
+		cache, err = findExactCache(db, cred.Repo, req.Key, req.Version, true)
+	}
 	if err != nil {
 		h.twirpError(w, r, twirpInternal, err)
 		return
@@ -127,13 +153,15 @@ func (h *Handler) v2FinalizeCacheEntryUpload(w http.ResponseWriter, r *http.Requ
 		h.twirpNotOK(w, r)
 		return
 	}
-	db.Close() // commitCache needs the store closed
 
-	cache.Size = int64(cmp.Or(req.SizeBytes, req.SizeBytesCamel))
-	if err := h.commitCache(cache); err != nil {
-		h.logger.Errorf("finalize cache %d (%s): %v", cache.ID, cache.Key, err)
-		h.twirpNotOK(w, r)
-		return
+	if !cache.Complete {
+		db.Close() // commitCache needs the store closed
+		cache.Size = int64(cmp.Or(req.SizeBytes, req.SizeBytesCamel))
+		if err := h.commitCache(cache); err != nil {
+			h.logger.Errorf("finalize cache %d (%s): %v", cache.ID, cache.Key, err)
+			h.twirpNotOK(w, r)
+			return
+		}
 	}
 
 	h.responseJSON(w, r, http.StatusOK, map[string]any{
@@ -147,7 +175,7 @@ func (h *Handler) v2GetCacheEntryDownloadURL(w http.ResponseWriter, r *http.Requ
 	cred := credFromContext(r.Context())
 	req, err := decodeTwirpRequest[v2DownloadRequest](r)
 	if err != nil {
-		h.twirpError(w, r, "malformed_request", err)
+		h.twirpError(w, r, "malformed", err)
 		return
 	}
 
@@ -230,8 +258,11 @@ func (h *Handler) twirpNotOK(w http.ResponseWriter, r *http.Request) {
 func (h *Handler) twirpError(w http.ResponseWriter, r *http.Request, code string, err error) {
 	h.logger.Debugf("%s %s: %v", r.Method, r.URL.Path, err)
 	status := http.StatusBadRequest
-	if code == twirpInternal {
+	switch code {
+	case twirpInternal:
 		status = http.StatusInternalServerError
+	case twirpUnauthenticated:
+		status = http.StatusUnauthorized
 	}
 	h.responseJSON(w, r, status, map[string]any{"code": code, "msg": err.Error()})
 }

act/artifactcache/handler_v2_test.go

@@ -10,6 +10,7 @@ import (
 	"fmt"
 	"io"
 	"net/http"
+	"net/url"
 	"strconv"
 	"strings"
 	"testing"
@@ -141,6 +142,17 @@ func TestCacheServiceV2BlockUpload(t *testing.T) {
 	}
 	list.WriteString(`</BlockList>`)
 	require.Equal(t, http.StatusCreated, putBlob(t, uploadURL+"&comp=blocklist", list.Bytes()))
+	require.Equal(t, http.StatusCreated, putBlob(t, uploadURL+"&comp=blocklist", list.Bytes()),
+		"a client that lost the first answer retries the list it already sent")
+
+	var reordered bytes.Buffer
+	reordered.WriteString(`<?xml version="1.0" encoding="utf-8"?><BlockList>`)
+	for _, blockID := range []string{order[1], order[0], order[2]} {
+		fmt.Fprintf(&reordered, "<Latest>%s</Latest>", blockID)
+	}
+	reordered.WriteString(`</BlockList>`)
+	assert.Equal(t, http.StatusInternalServerError, putBlob(t, uploadURL+"&comp=blocklist", reordered.Bytes()),
+		"the blocks are already assembled, so a different order would silently disagree with them")
 
 	finalized := v2Call(t, handler, testClient, "FinalizeCacheEntryUpload", map[string]any{
 		"key": "blocks", "version": "v1", "size_bytes": len("hello world!"),
@@ -241,3 +253,101 @@ func TestCacheServiceV2Lookups(t *testing.T) {
 		assert.Equal(t, false, got["ok"])
 	})
 }
+
+func TestCacheServiceV2RefusalsAreTwirp(t *testing.T) {
+	handler := newTestHandler(t, Policy{})
+
+	tests := []struct {
+		name   string
+		token  string
+		body   string
+		status int
+		code   string
+	}{
+		{"no bearer at all", "", `{"key":"k","version":"v"}`, http.StatusUnauthorized, twirpUnauthenticated},
+		{"a bearer nobody registered", "not-a-job", `{"key":"k","version":"v"}`, http.StatusUnauthorized, twirpUnauthenticated},
+		{"a body that is not the request", testToken, `{`, http.StatusBadRequest, "malformed"},
+		{"a request missing its key", testToken, `{"version":"v"}`, http.StatusBadRequest, "invalid_argument"},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			req, err := http.NewRequestWithContext(t.Context(), http.MethodPost,
+				handler.ExternalURL()+cacheServiceV2Path+"/CreateCacheEntry", strings.NewReader(tt.body))
+			require.NoError(t, err)
+			if tt.token != "" {
+				req.Header.Set("Authorization", "Bearer "+tt.token)
+			}
+			resp, err := http.DefaultClient.Do(req)
+			require.NoError(t, err)
+			defer resp.Body.Close()
+
+			require.Equal(t, tt.status, resp.StatusCode)
+			got := map[string]string{}
+			require.NoError(t, json.UnmarshalRead(resp.Body, &got))
+			assert.Equal(t, tt.code, got["code"])
+			assert.NotEmpty(t, got["msg"])
+		})
+	}
+}
+
+func TestCacheServiceV2ConcurrentSaveAndRetries(t *testing.T) {
+	handler := newTestHandler(t, Policy{})
+	content := []byte("the cached archive")
+
+	const otherJob = "another-jobs-token"
+	defer handler.RegisterJob(otherJob, JobCredential{Repo: testRepo})()
+	reserve := func(client *http.Client) map[string]any {
+		return v2Call(t, handler, client, "CreateCacheEntry", map[string]any{"key": "shared", "version": "v1"})
+	}
+	first := reserve(testClient)
+	require.Equal(t, true, first["ok"])
+	retry := reserve(testClient)
+	assert.Equal(t, true, retry["ok"], "a job retrying its own reservation is not a conflict")
+	entryOf := func(reserved map[string]any) string {
+		parsed, err := url.Parse(reserved["signed_upload_url"].(string))
+		require.NoError(t, err)
+		return parsed.Path
+	}
+	assert.Equal(t, entryOf(first), entryOf(retry), "the same reservation, however freshly its URL is signed")
+	assert.Equal(t, false, reserve(&http.Client{Transport: &bearerTransport{token: otherJob}})["ok"],
+		"another job's reservation would be finalized against the first upload")
+
+	otherClient := &http.Client{Transport: &bearerTransport{token: otherJob}}
+	reserved := v2Call(t, handler, otherClient, "CreateCacheEntry", map[string]any{"key": "owned", "version": "v1"})
+	require.Equal(t, true, reserved["ok"])
+	require.Equal(t, http.StatusCreated, putBlob(t, reserved["signed_upload_url"].(string), content))
+	assert.Equal(t, false, v2Call(t, handler, testClient, "FinalizeCacheEntryUpload",
+		map[string]any{"key": "owned", "version": "v1", "size_bytes": strconv.Itoa(len(content))})["ok"],
+		"another job's upload is not this job's to finalize")
+
+	base := handler.ExternalURL() + apiPath
+	reserveV1, err := json.Marshal(&Request{Key: "v1-made", Version: "v1", Size: int64(len(content))})
+	require.NoError(t, err)
+	resp, err := testClient.Post(base+"/caches", "application/json", bytes.NewReader(reserveV1))
+	require.NoError(t, err)
+	var madeByV1 struct {
+		CacheID uint64 `json:"cacheId"`
+	}
+	require.NoError(t, json.UnmarshalRead(resp.Body, &madeByV1))
+	resp.Body.Close()
+	upload, err := http.NewRequestWithContext(t.Context(), http.MethodPatch,
+		fmt.Sprintf("%s/caches/%d", base, madeByV1.CacheID), bytes.NewReader(content))
+	require.NoError(t, err)
+	upload.Header.Set("Content-Range", fmt.Sprintf("bytes 0-%d/*", len(content)-1))
+	resp, err = testClient.Do(upload)
+	require.NoError(t, err)
+	resp.Body.Close()
+	require.Equal(t, http.StatusOK, resp.StatusCode)
+
+	assert.Equal(t, false, v2Call(t, handler, testClient, "FinalizeCacheEntryUpload",
+		map[string]any{"key": "v1-made", "version": "v1", "size_bytes": strconv.Itoa(len(content))})["ok"],
+		"a reservation this job did not make through v2 is not its to commit")
+
+	finalized, _ := saveV2(t, handler, "deps", "v1", content)
+	require.Equal(t, true, finalized["ok"])
+	again := v2Call(t, handler, testClient, "FinalizeCacheEntryUpload", map[string]any{
+		"key": "deps", "version": "v1", "size_bytes": strconv.Itoa(len(content)),
+	})
+	assert.Equal(t, true, again["ok"], "a retry whose first answer was lost must not report a failed save")
+	assert.Equal(t, finalized["entry_id"], again["entry_id"])
+}

act/artifactcache/model.go

@@ -16,6 +16,7 @@ type Cache struct {
 	Key       string `json:"key" boltholdIndex:"Key"`
 	Version   string `json:"version" boltholdIndex:"Version"`
 	Size      int64  `json:"cacheSize"`
+	Owner     string `json:"owner"` // hashed bearer, so a job's own retry is not another job's reservation
 	Complete  bool   `json:"complete" boltholdIndex:"Complete"`
 	UsedAt    int64  `json:"usedAt" boltholdIndex:"UsedAt"`
 	CreatedAt int64  `json:"createdAt" boltholdIndex:"CreatedAt"`

act/artifactcache/results.go

@@ -5,6 +5,8 @@ package artifactcache
 
 import (
 	"crypto/tls"
+	"errors"
+	"fmt"
 	"net/http"
 	"net/http/httputil"
 	"net/url"
@@ -23,15 +25,33 @@ const artifactServicePath = "/twirp/github.actions.results.api.v1.ArtifactServic
 // 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) {
+	if !strings.HasPrefix(r.URL.Path, artifactServicePath) {
 		http.NotFound(w, r)
 		return
 	}
+	cred, ok := h.lookupCredential(bearerToken(r))
+	if !ok {
+		h.twirpError(w, r, twirpUnauthenticated, errors.New("unknown bearer token"))
+		return
+	}
+	if cred.Results == "" {
+		h.twirpError(w, r, twirpInternal, errors.New("no instance is registered for this job"))
+		return
+	}
 	target, err := url.Parse(strings.TrimSuffix(cred.Results, "/"))
+	if err == nil && (target.Hostname() == "" || (target.Scheme != "http" && target.Scheme != "https")) {
+		err = errors.New("not an absolute http or https URL the job could use")
+	}
 	if err != nil {
 		h.logger.Errorf("artifact service forward to %q: %v", cred.Results, err)
-		w.WriteHeader(http.StatusBadGateway)
+		h.twirpError(w, r, twirpInternal, fmt.Errorf("artifact service address %q is unusable: %w", cred.Results, err))
+		return
+	}
+	if !h.mustProxy(r, cred, target) {
+		redirect := target.JoinPath(r.URL.EscapedPath())
+		redirect.RawQuery = mergeQuery(target.RawQuery, r.URL.RawQuery)
+		h.logger.Debugf("%s %s: redirecting to %s", r.Method, r.URL.Path, target)
+		http.Redirect(w, r, redirect.String(), http.StatusTemporaryRedirect)
 		return
 	}
 	h.logger.Debugf("%s %s: forwarding to %s", r.Method, r.URL.Path, target)
@@ -44,9 +64,9 @@ func (h *Handler) forwardOrNotFound(w http.ResponseWriter, r *http.Request) {
 			// directly and never through here.
 			r.Out.Host = target.Host
 		},
-		ErrorHandler: func(w http.ResponseWriter, _ *http.Request, err error) {
+		ErrorHandler: func(w http.ResponseWriter, r *http.Request, err error) {
 			h.logger.Warnf("artifact service forward to %s: %v", target, err)
-			w.WriteHeader(http.StatusBadGateway)
+			h.twirpError(w, r, twirpInternal, fmt.Errorf("cache server cannot reach the artifact service at %s: %w", target, err))
 		},
 	}
 	if cred.InsecureTLS {
@@ -55,5 +75,27 @@ func (h *Handler) forwardOrNotFound(w http.ResponseWriter, r *http.Request) {
 	proxy.ServeHTTP(w, r)
 }
 
+// A protocol switch loses the bearer or the agent, and an untrusted instance needs skipped verification.
+func (h *Handler) mustProxy(r *http.Request, cred JobCredential, target *url.URL) bool {
+	if cred.InsecureTLS && target.Scheme == "https" {
+		return true
+	}
+	scheme := r.Header.Get("X-Forwarded-Proto")
+	if base, err := url.Parse(cred.PublicURL); err == nil && base.Scheme != "" {
+		scheme = base.Scheme
+	}
+	if scheme == "" {
+		scheme = "http"
+	}
+	return scheme != target.Scheme
+}
+
+func mergeQuery(target, request string) string {
+	if target == "" || request == "" {
+		return target + request
+	}
+	return target + "&" + request
+}
+
 // 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

act/artifactcache/results_test.go

@@ -4,6 +4,8 @@
 package artifactcache
 
 import (
+	"cmp"
+	"encoding/json/v2"
 	"io"
 	"net/http"
 	"net/http/httptest"
@@ -14,8 +16,6 @@ import (
 	"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) {
@@ -23,34 +23,112 @@ func TestFrontResultsService(t *testing.T) {
 		_, _ = io.WriteString(w, `{"ok":true}`)
 	}))
 	defer gitea.Close()
+	selfSigned := httptest.NewTLSServer(gitea.Config.Handler)
+	defer selfSigned.Close()
 
 	handler, err := StartHandler(Options{Dir: t.TempDir(), OutboundIP: "127.0.0.1"})
 	require.NoError(t, err)
 	defer handler.Close()
-	const token = "forward-token"
+	const token, rpc = "forward-token", artifactServicePath + "CreateArtifact"
 
-	client := &http.Client{Transport: &bearerTransport{token: token}}
-	post := func(path string) int {
+	noRedirect := func(token string) *http.Client {
+		return &http.Client{
+			Transport:     &bearerTransport{token: token},
+			CheckRedirect: func(*http.Request, []*http.Request) error { return http.ErrUseLastResponse },
+		}
+	}
+	post := func(path string) (int, map[string]string) {
 		req, err := http.NewRequestWithContext(t.Context(), http.MethodPost, handler.ExternalURL()+path, nil)
 		require.NoError(t, err)
-		resp, err := client.Do(req)
+		resp, err := noRedirect(token).Do(req)
 		require.NoError(t, err)
-		resp.Body.Close()
-		return resp.StatusCode
+		defer resp.Body.Close()
+		body := map[string]string{}
+		_ = json.UnmarshalRead(resp.Body, &body)
+		return resp.StatusCode, body
 	}
 
-	assert.Equal(t, http.StatusNotFound, post(artifactServicePath+"CreateArtifact"),
-		"an unregistered token is forwarded nowhere")
+	status, body := post(rpc)
+	assert.Equal(t, http.StatusUnauthorized, status, "an unregistered token is told so, not left to parse a 404 page")
+	assert.Equal(t, twirpUnauthenticated, body["code"])
+	assert.Contains(t, body["msg"], "bearer")
+	assert.NotEmpty(t, body["msg"])
 
-	defer handler.RegisterJob(token, JobCredential{Repo: "owner/repo", Results: gitea.URL})()
+	defer handler.RegisterJob(token, JobCredential{Repo: "owner/repo", Results: gitea.URL, PublicURL: "https://cache.example"})()
 
-	assert.Equal(t, http.StatusOK, post(artifactServicePath+"CreateArtifact"))
+	status, _ = post(rpc)
+	require.Equal(t, http.StatusOK, status, "a registered https public URL onto an http instance must proxy, not redirect")
 	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)
+	assert.Equal(t, rpc, 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"))
+	status, _ = post("/twirp/github.actions.results.api.v1.OtherService/Do")
+	assert.Equal(t, http.StatusNotFound, status)
+	status, _ = post("/api/v1/repos/owner/repo")
+	assert.Equal(t, http.StatusNotFound, status)
 	assert.Empty(t, gotPath, "only the artifact service is forwarded")
+
+	tests := []struct {
+		name           string
+		cred           JobCredential
+		forwardedProto string
+		reqQuery       string
+		want           int
+		location       string
+		wantMsg        string
+	}{
+		{name: "an instance the job reaches itself", cred: JobCredential{Results: gitea.URL}, want: http.StatusTemporaryRedirect},
+		{name: "a sub-path instance keeps its prefix", cred: JobCredential{Results: gitea.URL + "/sub"}, want: http.StatusTemporaryRedirect},
+		{name: "an instance with no certificate to distrust", cred: JobCredential{Results: gitea.URL, InsecureTLS: true}, want: http.StatusTemporaryRedirect},
+		{
+			name: "both queries survive", cred: JobCredential{Results: gitea.URL + "?tenant=one"}, reqQuery: "?run=7",
+			want: http.StatusTemporaryRedirect, location: gitea.URL + rpc + "?tenant=one&run=7",
+		},
+		{
+			name: "a trusted https instance is reached by the job itself",
+			cred: JobCredential{Results: selfSigned.URL, PublicURL: "https://cache.example"}, want: http.StatusTemporaryRedirect,
+		},
+		{
+			name: "the job does not share this server's disregard for the certificate",
+			cred: JobCredential{Results: selfSigned.URL, PublicURL: "https://cache.example", InsecureTLS: true}, want: http.StatusOK,
+		},
+		{name: "an http job is not redirected to an https instance", cred: JobCredential{Results: selfSigned.URL}, want: http.StatusInternalServerError},
+		{
+			name: "a runner too old to send its public URL leaves the terminator to say so",
+			cred: JobCredential{Results: gitea.URL}, forwardedProto: "https", want: http.StatusOK,
+		},
+		{
+			name: "a scheme the job could not use is refused before it is sent one",
+			cred: JobCredential{Results: "ftp://gitea.example"}, want: http.StatusInternalServerError, wantMsg: "unusable",
+		},
+		{
+			name: "an address with no host is refused before it is sent one",
+			cred: JobCredential{Results: "http://:3000"}, want: http.StatusInternalServerError, wantMsg: "unusable",
+		},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			defer handler.RegisterJob(tt.name, tt.cred)()
+			req, err := http.NewRequestWithContext(t.Context(), http.MethodPost, handler.ExternalURL()+rpc+tt.reqQuery, nil)
+			require.NoError(t, err)
+			if tt.forwardedProto != "" {
+				req.Header.Set("X-Forwarded-Proto", tt.forwardedProto)
+			}
+			resp, err := noRedirect(tt.name).Do(req)
+			require.NoError(t, err)
+			defer resp.Body.Close()
+
+			require.Equal(t, tt.want, resp.StatusCode)
+			switch tt.want {
+			case http.StatusTemporaryRedirect:
+				assert.Equal(t, cmp.Or(tt.location, tt.cred.Results+rpc), resp.Header.Get("Location"))
+			case http.StatusInternalServerError:
+				body := map[string]string{}
+				require.NoError(t, json.UnmarshalRead(resp.Body, &body))
+				assert.Equal(t, twirpInternal, body["code"])
+				assert.Contains(t, body["msg"], cmp.Or(tt.wantMsg, tt.cred.Results))
+			}
+		})
+	}
 }

act/artifactcache/storage.go

@@ -5,8 +5,10 @@
 package artifactcache
 
 import (
+	"bytes"
 	"crypto/sha256"
 	"encoding/hex"
+	"encoding/json/v2"
 	"fmt"
 	"io"
 	"net/http"
@@ -66,6 +68,18 @@ func (s *Storage) WriteBlock(id uint64, blockID string, reader io.Reader) error
 // rename pass is safe because a staged name always carries blockFilePrefix and a target name
 // never does, so no rename can collide with a block not yet moved.
 func (s *Storage) OrderBlocks(id uint64, blockIDs []string) error {
+	wanted, err := json.Marshal(blockIDs) // a block id is client-chosen, so it cannot be a separator
+	if err != nil {
+		return err
+	}
+	list := filepath.Join(s.tempDir(id), blockFilePrefix+"list")
+	if recorded, err := os.ReadFile(list); err == nil {
+		// A retry repeats the list it already sent; another one would reorder what is now staged.
+		if !bytes.Equal(recorded, wanted) {
+			return fmt.Errorf("cache %d was already assembled from a different block list", id)
+		}
+		return nil
+	}
 	for i, blockID := range blockIDs {
 		if err := os.Rename(s.blockName(id, blockID), s.tempName(id, int64(i))); err != nil {
 			if os.IsNotExist(err) {
@@ -74,7 +88,7 @@ func (s *Storage) OrderBlocks(id uint64, blockIDs []string) error {
 			return err
 		}
 	}
-	return nil
+	return os.WriteFile(list, wanted, 0o600)
 }
 
 func (s *Storage) Commit(id uint64, size int64) (int64, error) {

act/runner/action.go

@@ -147,7 +147,7 @@ func maybeCopyToActionDir(ctx context.Context, step actionStep, actionDir, actio
 
 	if !rc.Config.NoActionPatch {
 		// A concurrent job's prepare resets this directory, so patch under the copy's lock.
-		patchActions(ctx, actionScriptPaths(filepath.Join(actionDir, actionPath), step.getActionModel()))
+		defer patchActions(ctx, actionScriptPaths(filepath.Join(actionDir, actionPath), step.getActionModel()))()
 	}
 
 	if err := removeGitIgnore(ctx, actionDir); err != nil {

act/runner/patch_actions.go

@@ -28,8 +28,8 @@ import (
 // of upload-artifact.
 //
 // getCacheServiceURL() then resolves the cache service from ACTIONS_RESULTS_URL alone, where v1
-// reads ACTIONS_CACHE_URL first. Both reads there are given the same preference, which is what
-// keeps the runner out of the artifact path: the results URL still points at Gitea.
+// reads ACTIONS_CACHE_URL first. Both reads there are given the same preference, so a patched cache
+// client reaches the cache server by its own address rather than through the results origin.
 //
 // Either of these landing upstream makes this file deletable:
 //
@@ -105,37 +105,45 @@ func actionScriptPaths(dir string, action *model.Action) []string {
 	return paths
 }
 
-// patchActions edits the toolkit in an action's bundles. The caller holds the action directory's
-// clone lock, which is what keeps another job's checkout from resetting them before the copy.
-func patchActions(ctx context.Context, scripts []string) {
+// patchActions returns a restore: the tree is shared, so a job with patching off gets no edits.
+func patchActions(ctx context.Context, scripts []string) func() {
+	restore := map[string][]byte{}
 	for _, script := range scripts {
-		switch patched, err := patchBundle(script); {
-		case err != nil:
+		original, err := patchBundle(script)
+		if original != nil {
+			restore[script] = original // also when the write failed part way through
+		}
+		if err != nil {
 			common.Logger(ctx).Warnf("actions toolkit: %s left unpatched: %v", script, err)
-		case patched:
-			common.Logger(ctx).Debugf("actions toolkit: patched %s", script)
+		}
+	}
+	return func() {
+		for script, original := range restore {
+			if err := os.WriteFile(script, original, 0o644); err != nil { //nolint:gosec // as the checkout wrote it
+				common.Logger(ctx).Warnf("actions toolkit: %s left patched in the shared copy: %v", script, err)
+			}
 		}
 	}
 }
 
-func patchBundle(script string) (bool, error) {
+// patchBundle returns the bytes it replaced, or nil when it left the bundle alone.
+func patchBundle(script string) ([]byte, error) {
 	info, err := os.Stat(script)
 	if err != nil {
-		return false, err
+		return nil, err
 	}
 	if info.Size() > maxBundleSize {
-		return false, nil
+		return nil, nil
 	}
 	data, err := os.ReadFile(script)
 	if err != nil {
-		return false, err
+		return nil, err
 	}
 	patched, ok := patchedBundle(data)
 	if !ok {
-		return false, nil
+		return nil, nil
 	}
-	// No atomic write needed: every prepare checks the action out and hard resets it.
-	return true, os.WriteFile(script, patched, info.Mode().Perm())
+	return data, os.WriteFile(script, patched, info.Mode().Perm())
 }
 
 // patchedBundle opens the GHES gate, and where the cache toolkit is present, points the cache

act/runner/patch_actions_test.go

@@ -176,16 +176,16 @@ func TestPatchedBundleSurvivesInsideAStringLiteral(t *testing.T) {
 func TestPatchBundleIsIdempotent(t *testing.T) {
 	script := bundleFile(t, gateTSC)
 
-	done, err := patchBundle(script)
+	original, err := patchBundle(script)
 	require.NoError(t, err)
-	require.True(t, done)
+	require.NotNil(t, original)
 	patched, err := os.ReadFile(script)
 	require.NoError(t, err)
 	require.True(t, gateOpened(string(patched)))
 
-	done, err = patchBundle(script)
+	original, err = patchBundle(script)
 	require.NoError(t, err)
-	assert.False(t, done, "a patched bundle is not patched again")
+	assert.Nil(t, original, "a patched bundle is not patched again")
 	again, err := os.ReadFile(script)
 	require.NoError(t, err)
 	assert.Equal(t, string(patched), string(again))
@@ -194,9 +194,9 @@ func TestPatchBundleIsIdempotent(t *testing.T) {
 func TestPatchBundleLeavesOtherActionsAlone(t *testing.T) {
 	script := bundleFile(t, `console.log("checkout")`)
 
-	done, err := patchBundle(script)
+	original, err := patchBundle(script)
 	require.NoError(t, err)
-	assert.False(t, done)
+	assert.Nil(t, original)
 	body, err := os.ReadFile(script)
 	require.NoError(t, err)
 	assert.Equal(t, `console.log("checkout")`, string(body))
@@ -251,6 +251,9 @@ func TestPatchActionsAtTheContainerCopy(t *testing.T) {
 			return nil
 		})
 		require.NoError(t, maybeCopyToActionDir(t.Context(), sar, sar.actionDir(), "sub", "/var/run/act/actions/repo/sub"))
+		source, err := os.ReadFile(script)
+		require.NoError(t, err)
+		assert.Equal(t, gateTSC, string(source), "the shared copy is left as it was found")
 		return copied
 	}