Sign in

Search issue titles and bodies, change requests and comments. Private repositories you cannot read are not included.

vole
Clone

Read only, no credentials:

https://rodent.software/vole.git

Write over SSH with a registered key:

git clone ssh://git@rodent.software:2222/vole.git

A free and open-source forge that lives in your Git repo.

Phase 4: the runner signs the result it reports

22528ac3da576387a8311969641a7b5783b04924 · Keenan Nemetz · 17d ago · parent 86fe5bb4

Phase 4: the runner signs the result it reports

An instance-key signature is the server witnessing itself. It cannot be checked
without trusting the box that produced it, and it says nothing about the party that
ran the steps — so a record signed that way would have been paperwork. The runner
already holds a key, the one it connects with, so the runner is what signs.

At finish the runner signs a claim: run, repo, ref, SHA, workflow blob, the digest of
the steps pinned when the job was queued, exit code, the log's {oid, size}, the
artifacts it published, and its own finishing time. The server rebuilds the same claim
from its own row plus what the runner reports, verifies the signature, and stores the
payload and signature verbatim — so a verifier re-checks the bytes that were signed
rather than re-deriving them from a format that might have drifted.

Who the signer is comes from the channel, not the claim: the account is the one the
SSH key resolved to, and the signature must verify under one of that account's keys.
A runner may therefore sign with a different key from the one it connected with — an
agent-held key to connect, a file to sign — but not as somebody else.

Refusals, all before anything is recorded, because half a record is worse than none:
an unsigned finish (an unattributed result is what the queue already had), a signature
by another account's key, and a signature over a different claim. The run page shows
`attested by <name> · key <fingerprint>`, and `vole runner` now pings first and refuses
to start work when the key it will sign with is not one its account holds — discovering
that after a ten-minute build is a wasted build.

`steps_sha` is recorded at enqueue and travels with the lease, so the claim covers the
steps that ran and not merely the workflow file they came from.

What this does not buy: truth. A runner can still lie about an exit code. It buys
accountability — the record is attributable to a party other than the instance.
Whether a required check must be attested by a particular runner (an ACL key defaulting
to the instance's own) is the next policy question and is deliberately not answered
here; any admin of the repository may still supply the signature.

Verified: TestAFinishedRunBelongsToTheRunnerThatSignsIt drives the real protocol over
SSH — the lease carries what the claim names, ping names the account and its keys,
unsigned and wrong-account and wrong-claim finishes are all refused with nothing
recorded, the correct one is accepted, and the stored payload still verifies under the
stored signature; TestClaimRendersOneWay pins the byte form (including artifact
ordering and an empty log) and TestClaimRefusesToForgeItself refuses a field carrying a
newline; TestStepDigestFollowsTheSteps shows the digest covering the steps themselves;
the cmd tests cover key discovery from VOLE_SSH_COMMAND/GIT_SSH_COMMAND, the missing-key
message, a non-key file, and the pre-work refusal; and the end-to-end test that runs the
real binary asserts the claim names the run, its log and its exit, and that the run page
says who attested it. scripts/ci.sh green (gofmt, vet, build, markdown links, go test
-race ./...).

cmd/vole/runner.go +169−16

@@ -3,6 +3,7 @@ package main
3 3 import (
4 4 "bufio"
5 5 "context"
6 "crypto/ed25519"
6 7 "encoding/json"
7 8 "errors"
8 9 "fmt"
@@ -11,24 +12,34 @@ import (
11 12 "os/exec"
12 13 "path"
13 14 "path/filepath"
15 "slices"
14 16 "sort"
15 17 "strings"
16 18
17 19 "github.com/rodent-software/vole/internal/releases"
20 "github.com/rodent-software/vole/internal/runs"
21 "golang.org/x/crypto/ssh"
18 22 "sync"
19 23 "syscall"
20 24 "time"
21 25 )
22 26
23 27 func runnerUsage() string {
24 return `usage: vole runner --remote ssh://git@host[:port] [--once] [--dir path]
28 return `usage: vole runner --remote ssh://git@host[:port] [--once] [--dir path] [--key path]
25 29
26 30 Leases CI jobs from a Vole instance and runs them. Authentication is the SSH key
27 31 you would push with, so there is no token to configure. A runner may only run jobs
28 32 for repositories where its account has admin.
29 33
34 Every finished attempt is signed with --key, because the runner is the party that
35 did the work: an unattributed result is what the queue already had. The key must
36 belong to the account the connection authenticated as, and is checked on connect
37 rather than after the build.
38
30 39 --once run one job, then exit (useful for a container that is scheduled per job)
31 --dir where to keep checkouts (default: os.UserCacheDir()/vole-runner)`
40 --dir where to keep checkouts (default: os.UserCacheDir()/vole-runner)
41 --key private key to sign results with (default: -i from $VOLE_SSH_COMMAND or
42 $GIT_SSH_COMMAND, else ~/.ssh/id_ed25519)`
32 43 }
33 44
34 45 type runnerStep struct {
@@ -38,13 +49,19 @@ type runnerStep struct {
38 49 }
39 50
40 51 type runnerJob struct {
41 ID string `json:"id"`
42 Repo string `json:"repo"`
43 SHA string `json:"sha"`
44 Branch string `json:"branch"`
45 Workflow string `json:"workflow"`
46 Job string `json:"job"`
47 Steps []runnerStep `json:"steps"`
52 ID string `json:"id"`
53 Repo string `json:"repo"`
54 SHA string `json:"sha"`
55 // Ref, WorkflowSHA and StepsSHA are part of what this runner signs when the
56 // attempt ends: the exact revision, the exact workflow file, and the digest of
57 // the steps that were pinned when the job was queued.
58 Ref string `json:"ref"`
59 WorkflowSHA string `json:"workflow_sha"`
60 StepsSHA string `json:"steps_sha"`
61 Branch string `json:"branch"`
62 Workflow string `json:"workflow"`
63 Job string `json:"job"`
64 Steps []runnerStep `json:"steps"`
48 65 // Artifacts are path patterns, expanded in the checkout after the steps pass.
49 66 Artifacts []string `json:"artifacts,omitempty"`
50 67 // Tag is set for a tag-triggered run; TagAnnotated says whether it carries
@@ -81,6 +98,11 @@ type runnerRequest struct {
81 98 LogSize int64 `json:"log_size,omitempty"`
82 99 // Artifacts ride the finish so the result and the outputs are one message.
83 100 Artifacts []runnerArtifact `json:"artifacts,omitempty"`
101 // Finished is this runner's own unix time for the end of the attempt, and
102 // Signature covers the claim built from this message and the job: what was run,
103 // what came of it, and the objects it left behind.
104 Finished int64 `json:"finished,omitempty"`
105 Signature []byte `json:"signature,omitempty"`
84 106 }
85 107
86 108 type runnerReply struct {
@@ -89,6 +111,12 @@ type runnerReply struct {
89 111 Run *runnerJob `json:"run,omitempty"`
90 112 OK bool `json:"ok"`
91 113 Lease int `json:"lease,omitempty"`
114 // The account this session authenticated as and the fingerprints of its keys,
115 // which is how a runner finds out before it starts work whether the key it will
116 // sign with is one the server will accept.
117 Account string `json:"account,omitempty"`
118 Name string `json:"name,omitempty"`
119 Keys []string `json:"keys,omitempty"`
92 120 }
93 121
94 122 // cmdRunner leases and executes CI jobs. The protocol runs over an SSH channel
@@ -96,7 +124,7 @@ type runnerReply struct {
96 124 // is: no separate credential, nothing stored.
97 125 func cmdRunner(args []string) error {
98 126 f, err := parseFlags(args, flagSpec{long: "remote", short: "r"}, flagSpec{long: "dir"},
99 flagSpec{long: "once", boolean: true})
127 flagSpec{long: "key"}, flagSpec{long: "once", boolean: true})
100 128 if err != nil {
101 129 return err
102 130 }
@@ -126,7 +154,27 @@ func cmdRunner(args []string) error {
126 154 }
127 155 defer sess.close()
128 156
129 fmt.Fprintf(os.Stderr, "vole runner: connected to %s, waiting for work\n", remote)
157 key, fingerprint, err := loadSigningKey(f.one("key"))
158 if err != nil {
159 return err
160 }
161 sess.key = key
162
163 // Ask who the server thinks this is, and refuse to start work if the key we sign
164 // with is not one that account can answer for: learning that after a long build
165 // is a wasted build.
166 who, err := sess.call(runnerRequest{Op: "ping"})
167 if err != nil {
168 return err
169 }
170 if who.Error != "" {
171 return errors.New(who.Error)
172 }
173 if err := checkSigningKey(who, fingerprint); err != nil {
174 return err
175 }
176 fmt.Fprintf(os.Stderr, "vole runner: connected as %s, signing with %s; waiting for work\n",
177 who.Name, fingerprint)
130 178 for {
131 179 rep, err := sess.call(runnerRequest{Op: "lease"})
132 180 if err != nil {
@@ -162,6 +210,83 @@ type runnerSession struct {
162 210 enc *json.Encoder
163 211 dec *json.Decoder
164 212 mu sync.Mutex
213 // key signs every finish this session sends. It is the runner's own key, not an
214 // instance secret: the record names the party that did the work.
215 key ed25519.PrivateKey
216 }
217
218 // checkSigningKey is the refusal that has to happen before work, not after it: the
219 // server verifies a finish against the keys of the account the connection
220 // authenticated as, so a runner holding any other key would run a job it cannot then
221 // report.
222 func checkSigningKey(who runnerReply, fingerprint string) error {
223 if slices.Contains(who.Keys, fingerprint) {
224 return nil
225 }
226 return fmt.Errorf("this runner signs with %s, which is not a key of %s; "+
227 "pass --key with a key that account holds (it has %d key(s))", fingerprint, who.Name, len(who.Keys))
228 }
229
230 // loadSigningKey finds and reads the key a runner signs with: --key, then the -i of
231 // the ssh command the session itself is started with, then the usual default. An
232 // agent holds keys but does not sign arbitrary bytes, so a CI runner keeps its key
233 // on disk — which is the one place it already has it.
234 func loadSigningKey(flag string) (ed25519.PrivateKey, string, error) {
235 path := flag
236 if path == "" {
237 path = keyFromSSHCommand()
238 }
239 if path == "" {
240 if home, err := os.UserHomeDir(); err == nil {
241 cand := filepath.Join(home, ".ssh", "id_ed25519")
242 if _, err := os.Stat(cand); err == nil {
243 path = cand
244 }
245 }
246 }
247 if path == "" {
248 return nil, "", errors.New("no signing key: pass --key with the private key of the account this runner connects as")
249 }
250 data, err := os.ReadFile(path)
251 if err != nil {
252 return nil, "", fmt.Errorf("reading the signing key: %w", err)
253 }
254 raw, err := ssh.ParseRawPrivateKey(data)
255 if err != nil {
256 return nil, "", fmt.Errorf("%s: %w (a runner needs an unencrypted key: nothing is there to type a passphrase)", path, err)
257 }
258 var key ed25519.PrivateKey
259 switch k := raw.(type) {
260 case *ed25519.PrivateKey:
261 key = *k
262 case ed25519.PrivateKey:
263 key = k
264 default:
265 return nil, "", fmt.Errorf("%s is a %T key; a runner signs with ed25519", path, raw)
266 }
267 pub, err := ssh.NewPublicKey(key.Public())
268 if err != nil {
269 return nil, "", err
270 }
271 return key, ssh.FingerprintSHA256(pub), nil
272 }
273
274 // keyFromSSHCommand reads -i out of the ssh command this runner is started with, so
275 // `VOLE_SSH_COMMAND="ssh -i /path/ci" vole runner` signs with the same key it
276 // connects with instead of making the operator say it twice.
277 func keyFromSSHCommand() string {
278 for _, env := range []string{"VOLE_SSH_COMMAND", "GIT_SSH_COMMAND"} {
279 fields := strings.Fields(os.Getenv(env))
280 for i, a := range fields {
281 if a == "-i" && i+1 < len(fields) {
282 return fields[i+1]
283 }
284 if strings.HasPrefix(a, "-i") && len(a) > 2 {
285 return a[2:]
286 }
287 }
288 }
289 return ""
165 290 }
166 291
167 292 func dialRunner(remote string) (*runnerSession, error) {
@@ -285,20 +410,48 @@ func (s *runnerSession) run(remote, cache string, job *runnerJob, lease time.Dur
285 410 s.finish(remote, job, log, 0, arts)
286 411 }
287 412
288 // finish uploads this attempt's log and then reports the outcome. The bytes go in
289 // first, the same rule artifacts follow: the server refuses a log it cannot serve,
290 // and the object id is one this runner computed from the content.
413 // finish uploads this attempt's log, signs the result and reports it.
291 414 //
292 // A log that cannot be stored does not lose the result: the run is finished without
293 // one, and the failure is reported on stderr rather than swallowed.
415 // The bytes go in first, the same rule artifacts follow: the server refuses a log it
416 // cannot serve, and the object id is one this runner computed from the content — so
417 // the log is a fact the signature can cover.
418 //
419 // A log that cannot be stored does not lose the result: the claim then names no log
420 // and the failure is reported on stderr rather than swallowed. A claim that cannot be
421 // signed is not sent at all, because the server would refuse it.
294 422 func (s *runnerSession) finish(remote string, job *runnerJob, log *strings.Builder, exit int, arts []runnerArtifact) {
295 423 oid, size, err := uploadLog(remote, job.Repo, log.String())
296 424 if err != nil {
297 425 fmt.Fprintf(os.Stderr, "vole runner: could not upload the log for %s: %v\n", job.ID, err)
298 426 }
427 artifacts := make([]runs.Artifact, 0, len(arts))
428 for _, a := range arts {
429 artifacts = append(artifacts, runs.Artifact{Name: a.Name, OID: a.OID, Size: a.Size})
430 }
431 now := time.Now().UTC()
432 payload, err := runs.Claim{
433 RunID: job.ID,
434 Repo: job.Repo,
435 Ref: job.Ref,
436 SHA: job.SHA,
437 Workflow: job.Workflow,
438 WorkflowSHA: job.WorkflowSHA,
439 Job: job.Job,
440 StepsSHA: job.StepsSHA,
441 Exit: exit,
442 LogOID: oid,
443 LogSize: size,
444 Artifacts: artifacts,
445 Finished: now.Unix(),
446 }.Canonical()
447 if err != nil {
448 fmt.Fprintf(os.Stderr, "vole runner: cannot describe the result of %s: %v\n", job.ID, err)
449 return
450 }
299 451 _, _ = s.call(runnerRequest{
300 452 Op: "finish", Run: job.ID, Exit: exit,
301 453 LogOID: oid, LogSize: size, Artifacts: arts,
454 Finished: now.Unix(), Signature: ed25519.Sign(s.key, payload),
302 455 })
303 456 }
304 457