internal/sshd/sshd.go
572 lines · 17250 bytes
1// Package sshd implements the embedded SSH listener: public-key auth against
2// registered keys, then dispatch to git transport or control commands.
3package sshd
4
5import (
6 "context"
7 "crypto/ed25519"
8 "crypto/rand"
9 "encoding/base64"
10 "encoding/pem"
11 "errors"
12 "fmt"
13 "io"
14 "log/slog"
15 "maps"
16 "net"
17 "os"
18 "path/filepath"
19 "slices"
20 "strconv"
21 "sync"
22 "sync/atomic"
23 "time"
24
25 "golang.org/x/crypto/ssh"
26
27 "gitbay.org/gitbay/internal/config"
28 "gitbay.org/gitbay/internal/control"
29 "gitbay.org/gitbay/internal/gitutil"
30 "gitbay.org/gitbay/internal/hookd"
31 "gitbay.org/gitbay/internal/policy"
32 "gitbay.org/gitbay/internal/protocol"
33 "gitbay.org/gitbay/internal/store"
34)
35
36type Server struct {
37 cfg config.Config
38 st *store.Store
39 sshCfg *ssh.ServerConfig
40 authLimiter *rateLimiter
41 sessions sync.WaitGroup // accepted connections still being served
42 mu sync.Mutex
43 conns map[*conn]struct{}
44 stopping chan struct{} // closed by Stop
45 stopOnce sync.Once
46}
47
48// conn is one accepted connection and how many sessions it is running.
49// A CLI's shared connection sits idle between commands; on shutdown an
50// idle connection is closed at once and only a session mid-command is
51// waited for (#141).
52type conn struct {
53 net net.Conn
54 active atomic.Int32
55 // keyID and userID are the key that authenticated the connection and
56 // its account: 0 before the handshake and for an unregistered key.
57 // Guarded by Server.mu.
58 keyID, userID int64
59 revoked chan struct{} // closed by cut
60 cutOnce sync.Once
61}
62
63// cut ends the connection because its key was revoked: a git transport
64// on it is killed, and every other command loses its channel.
65func (c *conn) cut() {
66 c.cutOnce.Do(func() { close(c.revoked) })
67 c.net.Close()
68}
69
70func New(cfg config.Config, st *store.Store) (*Server, error) {
71 s := &Server{cfg: cfg, st: st, authLimiter: newRateLimiter(cfg.Limits.SSHAuthRate, time.Minute), conns: map[*conn]struct{}{}, stopping: make(chan struct{})}
72
73 sc := &ssh.ServerConfig{
74 PublicKeyCallback: s.authenticate,
75 ServerVersion: "SSH-2.0-gitbayd",
76 }
77 signers, err := loadHostKeys(cfg)
78 if err != nil {
79 return nil, err
80 }
81 for _, sg := range signers {
82 sc.AddHostKey(sg)
83 }
84 s.sshCfg = sc
85 st.OnRevoke(s.revoke)
86 return s, nil
87}
88
89// loadHostKeys loads the configured host keys, or generates an ed25519 key
90// under server.root/ssh/ when none are configured.
91func loadHostKeys(cfg config.Config) ([]ssh.Signer, error) {
92 paths := cfg.SSH.HostKeys
93 if len(paths) == 0 {
94 p := filepath.Join(cfg.Server.Root, "ssh", "host_ed25519")
95 if _, err := os.Stat(p); errors.Is(err, os.ErrNotExist) {
96 if err := generateHostKey(p); err != nil {
97 return nil, fmt.Errorf("generating host key: %w", err)
98 }
99 slog.Info("generated ssh host key", "path", p)
100 }
101 paths = []string{p}
102 }
103 var signers []ssh.Signer
104 for _, p := range paths {
105 raw, err := os.ReadFile(p)
106 if err != nil {
107 return nil, fmt.Errorf("host key %s: %w", p, err)
108 }
109 sg, err := ssh.ParsePrivateKey(raw)
110 if err != nil {
111 return nil, fmt.Errorf("host key %s: %w", p, err)
112 }
113 signers = append(signers, sg)
114 }
115 return signers, nil
116}
117
118func generateHostKey(path string) error {
119 if err := os.MkdirAll(filepath.Dir(path), 0o700); err != nil {
120 return err
121 }
122 _, priv, err := ed25519.GenerateKey(rand.Reader)
123 if err != nil {
124 return err
125 }
126 block, err := ssh.MarshalPrivateKey(priv, "")
127 if err != nil {
128 return err
129 }
130 return os.WriteFile(path, pem.EncodeToMemory(block), 0o600)
131}
132
133// authenticate resolves the presented key to a registered account. The SSH
134// username is ignored; identity comes from the key alone. When registration
135// is open or invite-based, unknown keys are admitted to run exactly one
136// command: register.
137func (s *Server) authenticate(meta ssh.ConnMetadata, pub ssh.PublicKey) (*ssh.Permissions, error) {
138 ip := remoteIP(meta.RemoteAddr())
139 if !s.authLimiter.allow(ip) {
140 // One audit entry per throttled window, not per rejected attempt.
141 if s.authLimiter.firstThrottle(ip) {
142 s.st.Audit(0, "auth.throttled", map[string]any{"ip": ip, "rate": s.cfg.Limits.SSHAuthRate})
143 }
144 return nil, fmt.Errorf("too many authentication attempts; try again shortly")
145 }
146 fp := ssh.FingerprintSHA256(pub)
147 key, err := s.st.SSHKeyByFingerprint(fp)
148 if err != nil && !errors.Is(err, store.ErrNotFound) {
149 // The store, not the key, failed. Neither a failure against the
150 // limiter nor "unknown key": a busy database during a restart
151 // would otherwise lock every client out for a minute.
152 slog.Error("ssh auth: key lookup", "err", err)
153 return nil, fmt.Errorf("authentication temporarily unavailable")
154 }
155 if err != nil {
156 if s.cfg.Registration.Mode != "closed" {
157 return &ssh.Permissions{Extensions: map[string]string{
158 "anon-key": base64.StdEncoding.EncodeToString(pub.Marshal()),
159 }}, nil
160 }
161 s.authLimiter.fail(ip)
162 s.st.Audit(0, "auth.failed", map[string]any{"ip": ip, "fingerprint": fp})
163 return nil, fmt.Errorf("unknown key %s", fp)
164 }
165 s.authLimiter.success(ip)
166 return &ssh.Permissions{Extensions: map[string]string{
167 "user-id": strconv.FormatInt(key.UserID, 10),
168 "key-id": strconv.FormatInt(key.ID, 10),
169 }}, nil
170}
171
172// Serve accepts connections on ln until it is closed.
173func (s *Server) Serve(ln net.Listener) error {
174 served := make(chan struct{})
175 defer close(served)
176 go s.sweep(served)
177 for {
178 nc, err := ln.Accept()
179 if err != nil {
180 return err
181 }
182 c := &conn{net: nc, revoked: make(chan struct{})}
183 s.mu.Lock()
184 s.conns[c] = struct{}{}
185 s.mu.Unlock()
186 s.sessions.Add(1)
187 go func() {
188 defer s.sessions.Done()
189 defer func() {
190 s.mu.Lock()
191 delete(s.conns, c)
192 s.mu.Unlock()
193 }()
194 s.handleConn(c)
195 }()
196 }
197}
198
199// revoke closes the connections opened by the keys r names.
200func (s *Server) revoke(r store.Revoked) {
201 var cut []*conn
202 s.mu.Lock()
203 for c := range s.conns {
204 if c.keyID == 0 {
205 continue
206 }
207 if (r.UserID != 0 && c.userID == r.UserID) || slices.Contains(r.KeyIDs, c.keyID) {
208 cut = append(cut, c)
209 }
210 }
211 s.mu.Unlock()
212 for _, c := range cut {
213 c.cut()
214 }
215}
216
217// sweepInterval bounds how long a revocation this process was not told
218// about (gitbayd admin on the host) leaves a connection open.
219const sweepInterval = 15 * time.Second
220
221func (s *Server) sweep(served <-chan struct{}) {
222 t := time.NewTicker(sweepInterval)
223 defer t.Stop()
224 for {
225 select {
226 case <-t.C:
227 s.sweepOnce()
228 case <-served:
229 return
230 case <-s.stopping:
231 return
232 }
233 }
234}
235
236// sweepOnce cuts every connection whose key is no longer live. Only
237// connections whose key was asked about are judged: one that
238// authenticated while the query ran waits for the next sweep.
239func (s *Server) sweepOnce() {
240 asked := map[int64]bool{}
241 s.mu.Lock()
242 for c := range s.conns {
243 if c.keyID != 0 {
244 asked[c.keyID] = true
245 }
246 }
247 s.mu.Unlock()
248 if len(asked) == 0 {
249 return
250 }
251 live, err := s.st.LiveSSHKeys(slices.Collect(maps.Keys(asked)))
252 if err != nil {
253 slog.Error("ssh sweep: key lookup", "err", err)
254 return
255 }
256 var cut []*conn
257 s.mu.Lock()
258 for c := range s.conns {
259 if asked[c.keyID] && !live[c.keyID] {
260 cut = append(cut, c)
261 }
262 }
263 s.mu.Unlock()
264 for _, c := range cut {
265 c.cut()
266 }
267}
268
269// Stop ends the commands that run until something happens (build log
270// --follow), so a shutdown drain waits only for work that finishes. It
271// does not close connections; Shutdown does.
272func (s *Server) Stop() {
273 s.stopOnce.Do(func() { close(s.stopping) })
274}
275
276// Shutdown closes every idle connection, then waits for the ones with a
277// session running, or for ctx. The caller closes the listener first; a
278// push in flight completes rather than being cut mid-pack.
279func (s *Server) Shutdown(ctx context.Context) error {
280 s.Stop()
281 s.mu.Lock()
282 for c := range s.conns {
283 if c.active.Load() == 0 {
284 c.net.Close()
285 }
286 }
287 s.mu.Unlock()
288 done := make(chan struct{})
289 go func() {
290 s.sessions.Wait()
291 close(done)
292 }()
293 select {
294 case <-done:
295 return nil
296 case <-ctx.Done():
297 return ctx.Err()
298 }
299}
300
301func (s *Server) handleConn(c *conn) {
302 defer c.net.Close()
303 sconn, chans, reqs, err := ssh.NewServerConn(c.net, s.sshCfg)
304 if err != nil {
305 return
306 }
307 defer sconn.Close()
308 ext := sconn.Permissions.Extensions
309 s.mu.Lock()
310 c.keyID, _ = strconv.ParseInt(ext["key-id"], 10, 64)
311 c.userID, _ = strconv.ParseInt(ext["user-id"], 10, 64)
312 s.mu.Unlock()
313 go ssh.DiscardRequests(reqs)
314
315 for newCh := range chans {
316 if newCh.ChannelType() != "session" {
317 newCh.Reject(ssh.UnknownChannelType, "only session channels are supported")
318 continue
319 }
320 ch, chReqs, err := newCh.Accept()
321 if err != nil {
322 continue
323 }
324 c.active.Add(1)
325 go func() {
326 defer c.active.Add(-1)
327 s.handleSession(c, sconn, ch, chReqs)
328 }()
329 }
330}
331
332func (s *Server) handleSession(c *conn, sconn *ssh.ServerConn, ch ssh.Channel, reqs <-chan *ssh.Request) {
333 defer ch.Close()
334 var term control.Term
335 for req := range reqs {
336 switch req.Type {
337 case "exec":
338 var payload struct{ Command string }
339 if err := ssh.Unmarshal(req.Payload, &payload); err != nil {
340 req.Reply(false, nil)
341 continue
342 }
343 req.Reply(true, nil)
344 // x/crypto closes reqs when the client closes the channel. That
345 // is how a follow learns nobody is reading: the CLI's shared
346 // connection outlives a Ctrl-C, the channel does not. Stop
347 // ends it too, for a restart.
348 closed := make(chan struct{})
349 go func() {
350 for r := range reqs {
351 r.Reply(false, nil)
352 }
353 close(closed)
354 }()
355 done := make(chan struct{})
356 go func() {
357 select {
358 case <-closed:
359 case <-s.stopping:
360 }
361 close(done)
362 }()
363 code := s.runExec(c, sconn, ch, term, payload.Command, done)
364 sendExit(ch, code)
365 return
366 case "shell":
367 req.Reply(true, nil)
368 fmt.Fprintf(ch, "gitbay control plane: interactive shells are not available.\nTry: ssh %s help\n", s.cfg.Server.SiteURL)
369 sendExit(ch, protocol.ExitUsage)
370 return
371 case "env":
372 var kv struct{ Name, Value string }
373 if ssh.Unmarshal(req.Payload, &kv) == nil && kv.Name == "GITBAY_TERM" {
374 term = control.ParseTerm(kv.Value)
375 }
376 req.Reply(true, nil)
377 case "pty-req":
378 // Harmless; accept and ignore.
379 req.Reply(true, nil)
380 default:
381 req.Reply(false, nil)
382 }
383 }
384}
385
386func sendExit(ch ssh.Channel, code int) {
387 var msg = struct{ Status uint32 }{uint32(code)}
388 ch.SendRequest("exit-status", false, ssh.Marshal(&msg))
389}
390
391func (s *Server) runExec(c *conn, sconn *ssh.ServerConn, ch ssh.Channel, term control.Term, cmdline string, done <-chan struct{}) int {
392 ext := sconn.Permissions.Extensions
393 if blob := ext["anon-key"]; blob != "" {
394 return s.runAnonymous(ch, blob, cmdline)
395 }
396 userID, _ := strconv.ParseInt(ext["user-id"], 10, 64)
397 keyID, _ := strconv.ParseInt(ext["key-id"], 10, 64)
398 // A connection outlives its commands, so the key is read again for
399 // each one: what it may do is what it may do now (#256).
400 key, err := s.st.SSHKeyByID(keyID)
401 if errors.Is(err, store.ErrNotFound) || (err == nil && key.UserID != userID) {
402 fmt.Fprintln(ch.Stderr(), "this key is no longer registered")
403 return protocol.ExitDenied
404 }
405 if err != nil {
406 slog.Error("ssh exec: key lookup", "err", err)
407 fmt.Fprintln(ch.Stderr(), "authentication temporarily unavailable")
408 return protocol.ExitFailure
409 }
410 user, err := s.st.UserByID(userID)
411 if err != nil {
412 fmt.Fprintln(ch.Stderr(), "account no longer exists")
413 return protocol.ExitDenied
414 }
415 _ = s.st.TouchSSHKey(keyID)
416 return Exec(s.cfg, s.st, user, key, term, cmdline, ch, ch, ch.Stderr(), done, s.stopping, c.revoked)
417}
418
419// runAnonymous handles a session from an unregistered key: the register
420// command and nothing else.
421func (s *Server) runAnonymous(ch ssh.Channel, keyB64, cmdline string) int {
422 raw, err := base64.StdEncoding.DecodeString(keyB64)
423 if err != nil {
424 return protocol.ExitFailure
425 }
426 pub, err := ssh.ParsePublicKey(raw)
427 if err != nil {
428 return protocol.ExitFailure
429 }
430 argv, err := protocol.Tokenize(cmdline)
431 if err != nil {
432 fmt.Fprintf(ch.Stderr(), "cannot parse command: %v\n", err)
433 return protocol.ExitUsage
434 }
435 if len(argv) == 0 || argv[0] != "register" {
436 fmt.Fprintf(ch.Stderr(), "this key is not registered here. Create an account with:\n ssh <host> register --username <name> %s\n",
437 map[string]string{"open": "--email <address>", "invite": "--invite <code>"}[s.cfg.Registration.Mode])
438 return protocol.ExitDenied
439 }
440 return control.RunRegister(s.cfg, s.st, pub, argv, ch, ch.Stderr())
441}
442
443// Exec runs one SSH exec command line for an authenticated key. It is the
444// single dispatch path shared by the embedded listener and the system-sshd
445// forced command (gitbayd shell). Closing revoked kills a git transport.
446func Exec(cfg config.Config, st *store.Store, user store.User, key store.SSHKey, term control.Term, cmdline string,
447 stdin io.Reader, stdout, stderr io.Writer, done, stopping, revoked <-chan struct{}) int {
448 if user.Disabled {
449 fmt.Fprintln(stderr, "this account is disabled; contact the instance admin")
450 return protocol.ExitDenied
451 }
452 argv, err := protocol.Tokenize(cmdline)
453 if err != nil {
454 fmt.Fprintf(stderr, "cannot parse command: %v\n", err)
455 return protocol.ExitUsage
456 }
457 if len(argv) > 0 {
458 switch argv[0] {
459 case "git-upload-pack", "git-receive-pack", "git-upload-archive":
460 if user.Pending {
461 fmt.Fprintln(stderr, "your account is not active yet: verify your email first")
462 return protocol.ExitDenied
463 }
464 return runGit(cfg, st, user, key.Scope, argv, stdin, stdout, stderr, revoked)
465 case "git-lfs-authenticate":
466 // Part of the git transport, not the control plane: usable by
467 // git-scoped and deploy keys, with the transports' access rules.
468 if user.Pending {
469 fmt.Fprintln(stderr, "your account is not active yet: verify your email first")
470 return protocol.ExitDenied
471 }
472 return runLFSAuthenticate(cfg, st, user, key.Scope, argv, stdout, stderr)
473 }
474 }
475 ctx := &control.Ctx{
476 User: user,
477 Scope: key.Scope,
478 Source: key.Fingerprint,
479 Term: term,
480 Store: st,
481 Cfg: cfg,
482 Stdin: stdin,
483 Stdout: stdout,
484 Stderr: stderr,
485 Done: done,
486 Stopping: stopping,
487 }
488 return control.Dispatch(ctx, argv)
489}
490
491// runGit streams a git transport service after access checks.
492func runGit(cfg config.Config, st *store.Store, user store.User, scope string, argv []string,
493 stdin io.Reader, stdout, stderr io.Writer, revoked <-chan struct{}) int {
494 service := argv[0]
495 if len(argv) != 2 {
496 fmt.Fprintf(stderr, "usage: %s <path>\n", service)
497 return protocol.ExitUsage
498 }
499 write := service == "git-receive-pack"
500
501 repo, err := st.RepoByPath(argv[1])
502 if err != nil {
503 fmt.Fprintln(stderr, "repository not found")
504 return protocol.ExitNotFound
505 }
506 if policy.IsDeployScope(scope) {
507 // A deploy key authorizes by its binding alone: one repository,
508 // its mode, nothing inherited from whoever registered it. Any
509 // mismatch reads as nonexistence, same as the access rules.
510 if !policy.DeployScopeAllows(scope, repo.ID, write) {
511 fmt.Fprintln(stderr, "repository not found")
512 return protocol.ExitNotFound
513 }
514 } else {
515 grant, err := st.AccessRole(repo.ID, user.ID)
516 if err != nil {
517 fmt.Fprintln(stderr, "internal error")
518 return protocol.ExitFailure
519 }
520 if !policy.CanRead(user, repo, grant) {
521 // Same answer as nonexistence: private repos must not be enumerable.
522 fmt.Fprintln(stderr, "repository not found")
523 return protocol.ExitNotFound
524 }
525 if !policy.ScopeAllowsGit(scope, repo.Path(), write) {
526 fmt.Fprintf(stderr, "this key's scope (%s) does not allow %s on %s\n", scope, service, repo.Path())
527 return protocol.ExitDenied
528 }
529 if write && !policy.CanWrite(user, repo, grant) {
530 fmt.Fprintf(stderr, "write access to %s denied\n", repo.Path())
531 return protocol.ExitDenied
532 }
533 }
534 if write && repo.Settings.Archived {
535 fmt.Fprintf(stderr, "%s is archived and read-only\n", repo.Path())
536 return protocol.ExitDenied
537 }
538 if write {
539 if mirrored, err := st.PullMirrored(repo.ID); err == nil && mirrored {
540 fmt.Fprintf(stderr, "%s is a pull mirror: its refs come from the upstream; push there instead\n", repo.Path())
541 return protocol.ExitDenied
542 }
543 }
544
545 dir := control.RepoDir(cfg.Server.Root, repo.OwnerName, repo.Name)
546 env := []string{
547 hookd.EnvSocket + "=" + hookd.SocketPath(cfg.Server.Root),
548 hookd.EnvRepoID + "=" + strconv.FormatInt(repo.ID, 10),
549 hookd.EnvUserID + "=" + strconv.FormatInt(user.ID, 10),
550 hookd.EnvScope + "=" + scope,
551 }
552 // A storage quota on the owner rides the same mechanism as the pack
553 // cap: the pack may be no larger than what the owner has left.
554 maxPack := cfg.Limits.MaxPackBytes
555 if write && repo.OwnerKind == "user" {
556 if limit := control.ByteLimit(st, control.QuotaConfig(cfg), repo.OwnerID); limit > 0 {
557 used := control.OwnedBytes(st, cfg.Server.Root, repo.OwnerID)
558 left := limit - used
559 if left <= 0 {
560 fmt.Fprintf(stderr, "%s's storage quota is used up (%d of %d bytes); delete something, or ask an admin to raise the limit\n", repo.OwnerName, used, limit)
561 return protocol.ExitDenied
562 }
563 if maxPack == 0 || left < maxPack {
564 maxPack = left
565 }
566 }
567 }
568 if err := gitutil.Transport(service, dir, stdin, stdout, stderr, env, maxPack, revoked); err != nil {
569 return protocol.ExitFailure
570 }
571 return protocol.ExitOK
572}