1// Copyright 2013 The Go Authors. All rights reserved.
  2// Use of this source code is governed by a BSD-style
  3// license that can be found in the LICENSE file.
  4
  5package ssh
  6
  7import (
  8	"errors"
  9	"fmt"
 10	"io"
 11	"log"
 12	"net"
 13	"slices"
 14	"strings"
 15	"sync"
 16)
 17
 18// debugHandshake, if set, prints messages sent and received.  Key
 19// exchange messages are printed as if DH were used, so the debug
 20// messages are wrong when using ECDH.
 21const debugHandshake = false
 22
 23// chanSize sets the amount of buffering SSH connections. This is
 24// primarily for testing: setting chanSize=0 uncovers deadlocks more
 25// quickly.
 26const chanSize = 16
 27
 28// maxPendingPackets sets the maximum number of packets to queue while waiting
 29// for KEX to complete. This limits the total pending data to maxPendingPackets
 30// * maxPacket bytes, which is ~16.8MB.
 31const maxPendingPackets = 64
 32
 33// keyingTransport is a packet based transport that supports key
 34// changes. It need not be thread-safe. It should pass through
 35// msgNewKeys in both directions.
 36type keyingTransport interface {
 37	packetConn
 38
 39	// prepareKeyChange sets up a key change. The key change for a
 40	// direction will be effected if a msgNewKeys message is sent
 41	// or received.
 42	prepareKeyChange(*NegotiatedAlgorithms, *kexResult) error
 43
 44	// setStrictMode sets the strict KEX mode, notably triggering
 45	// sequence number resets on sending or receiving msgNewKeys.
 46	// If the sequence number is already > 1 when setStrictMode
 47	// is called, an error is returned.
 48	setStrictMode() error
 49
 50	// setInitialKEXDone indicates to the transport that the initial key exchange
 51	// was completed
 52	setInitialKEXDone()
 53}
 54
 55// handshakeTransport implements rekeying on top of a keyingTransport
 56// and offers a thread-safe writePacket() interface.
 57type handshakeTransport struct {
 58	conn   keyingTransport
 59	config *Config
 60
 61	serverVersion []byte
 62	clientVersion []byte
 63
 64	// hostKeys is non-empty if we are the server. In that case,
 65	// it contains all host keys that can be used to sign the
 66	// connection.
 67	hostKeys []Signer
 68
 69	// publicKeyAuthAlgorithms is non-empty if we are the server. In that case,
 70	// it contains the supported client public key authentication algorithms.
 71	publicKeyAuthAlgorithms []string
 72
 73	// hostKeyAlgorithms is non-empty if we are the client. In that case,
 74	// we accept these key types from the server as host key.
 75	hostKeyAlgorithms []string
 76
 77	// On read error, incoming is closed, and readError is set.
 78	incoming  chan []byte
 79	readError error
 80
 81	mu sync.Mutex
 82	// Condition for the above mutex. It is used to notify a completed key
 83	// exchange or a write failure. Writes can wait for this condition while a
 84	// key exchange is in progress.
 85	writeCond      *sync.Cond
 86	writeError     error
 87	sentInitPacket []byte
 88	sentInitMsg    *kexInitMsg
 89	// Used to queue writes when a key exchange is in progress. The length is
 90	// limited by pendingPacketsSize. Once full, writes will block until the key
 91	// exchange is completed or an error occurs. If not empty, it is emptied
 92	// all at once when the key exchange is completed in kexLoop.
 93	pendingPackets   [][]byte
 94	writePacketsLeft uint32
 95	writeBytesLeft   int64
 96	userAuthComplete bool // whether the user authentication phase is complete
 97
 98	// If the read loop wants to schedule a kex, it pings this
 99	// channel, and the write loop will send out a kex
100	// message.
101	requestKex chan struct{}
102
103	// If the other side requests or confirms a kex, its kexInit
104	// packet is sent here for the write loop to find it.
105	startKex    chan *pendingKex
106	kexLoopDone chan struct{} // closed (with writeError non-nil) when kexLoop exits
107
108	// data for host key checking
109	hostKeyCallback HostKeyCallback
110	dialAddress     string
111	remoteAddr      net.Addr
112
113	// bannerCallback is non-empty if we are the client and it has been set in
114	// ClientConfig. In that case it is called during the user authentication
115	// dance to handle a custom server's message.
116	bannerCallback BannerCallback
117
118	// Algorithms agreed in the last key exchange.
119	algorithms *NegotiatedAlgorithms
120
121	// Counters exclusively owned by readLoop.
122	readPacketsLeft uint32
123	readBytesLeft   int64
124
125	// The session ID or nil if first kex did not complete yet.
126	sessionID []byte
127
128	// strictMode indicates if the other side of the handshake indicated
129	// that we should be following the strict KEX protocol restrictions.
130	strictMode bool
131}
132
133type pendingKex struct {
134	otherInit []byte
135	done      chan error
136}
137
138func newHandshakeTransport(conn keyingTransport, config *Config, clientVersion, serverVersion []byte) *handshakeTransport {
139	t := &handshakeTransport{
140		conn:          conn,
141		serverVersion: serverVersion,
142		clientVersion: clientVersion,
143		incoming:      make(chan []byte, chanSize),
144		requestKex:    make(chan struct{}, 1),
145		startKex:      make(chan *pendingKex),
146		kexLoopDone:   make(chan struct{}),
147
148		config: config,
149	}
150	t.writeCond = sync.NewCond(&t.mu)
151	t.resetReadThresholds()
152	t.resetWriteThresholds()
153
154	// We always start with a mandatory key exchange.
155	t.requestKex <- struct{}{}
156	return t
157}
158
159func newClientTransport(conn keyingTransport, clientVersion, serverVersion []byte, config *ClientConfig, dialAddr string, addr net.Addr) *handshakeTransport {
160	t := newHandshakeTransport(conn, &config.Config, clientVersion, serverVersion)
161	t.dialAddress = dialAddr
162	t.remoteAddr = addr
163	t.hostKeyCallback = config.HostKeyCallback
164	t.bannerCallback = config.BannerCallback
165	if config.HostKeyAlgorithms != nil {
166		t.hostKeyAlgorithms = config.HostKeyAlgorithms
167	} else {
168		t.hostKeyAlgorithms = defaultHostKeyAlgos
169	}
170	go t.readLoop()
171	go t.kexLoop()
172	return t
173}
174
175func newServerTransport(conn keyingTransport, clientVersion, serverVersion []byte, config *ServerConfig) *handshakeTransport {
176	t := newHandshakeTransport(conn, &config.Config, clientVersion, serverVersion)
177	t.hostKeys = config.hostKeys
178	t.publicKeyAuthAlgorithms = config.PublicKeyAuthAlgorithms
179	go t.readLoop()
180	go t.kexLoop()
181	return t
182}
183
184func (t *handshakeTransport) getSessionID() []byte {
185	return t.sessionID
186}
187
188func (t *handshakeTransport) getAlgorithms() NegotiatedAlgorithms {
189	return *t.algorithms
190}
191
192// waitSession waits for the session to be established. This should be
193// the first thing to call after instantiating handshakeTransport.
194func (t *handshakeTransport) waitSession() error {
195	p, err := t.readPacket()
196	if err != nil {
197		return err
198	}
199	if p[0] != msgNewKeys {
200		return fmt.Errorf("ssh: first packet should be msgNewKeys")
201	}
202
203	return nil
204}
205
206func (t *handshakeTransport) id() string {
207	if len(t.hostKeys) > 0 {
208		return "server"
209	}
210	return "client"
211}
212
213func (t *handshakeTransport) printPacket(p []byte, write bool) {
214	action := "got"
215	if write {
216		action = "sent"
217	}
218
219	if p[0] == msgChannelData || p[0] == msgChannelExtendedData {
220		log.Printf("%s %s data (packet %d bytes)", t.id(), action, len(p))
221	} else {
222		msg, err := decode(p)
223		log.Printf("%s %s %T %v (%v)", t.id(), action, msg, msg, err)
224	}
225}
226
227func (t *handshakeTransport) readPacket() ([]byte, error) {
228	p, ok := <-t.incoming
229	if !ok {
230		return nil, t.readError
231	}
232	return p, nil
233}
234
235func (t *handshakeTransport) readLoop() {
236	first := true
237	for {
238		p, err := t.readOnePacket(first)
239		first = false
240		if err != nil {
241			t.readError = err
242			close(t.incoming)
243			break
244		}
245		// If this is the first kex, and strict KEX mode is enabled,
246		// we don't ignore any messages, as they may be used to manipulate
247		// the packet sequence numbers.
248		if !(t.sessionID == nil && t.strictMode) && (p[0] == msgIgnore || p[0] == msgDebug) {
249			continue
250		}
251		t.incoming <- p
252	}
253
254	// Stop writers too.
255	t.recordWriteError(t.readError)
256
257	// Unblock the writer should it wait for this.
258	close(t.startKex)
259
260	// Don't close t.requestKex; it's also written to from writePacket.
261}
262
263func (t *handshakeTransport) pushPacket(p []byte) error {
264	if debugHandshake {
265		t.printPacket(p, true)
266	}
267	return t.conn.writePacket(p)
268}
269
270func (t *handshakeTransport) getWriteError() error {
271	t.mu.Lock()
272	defer t.mu.Unlock()
273	return t.writeError
274}
275
276func (t *handshakeTransport) recordWriteError(err error) {
277	t.mu.Lock()
278	defer t.mu.Unlock()
279	if t.writeError == nil && err != nil {
280		t.writeError = err
281		t.writeCond.Broadcast()
282	}
283}
284
285func (t *handshakeTransport) requestKeyExchange() {
286	select {
287	case t.requestKex <- struct{}{}:
288	default:
289		// something already requested a kex, so do nothing.
290	}
291}
292
293func (t *handshakeTransport) resetWriteThresholds() {
294	t.writePacketsLeft = packetRekeyThreshold
295	if t.config.RekeyThreshold > 0 {
296		t.writeBytesLeft = int64(t.config.RekeyThreshold)
297	} else if t.algorithms != nil {
298		t.writeBytesLeft = t.algorithms.Write.rekeyBytes()
299	} else {
300		t.writeBytesLeft = 1 << 30
301	}
302}
303
304func (t *handshakeTransport) kexLoop() {
305
306write:
307	for t.getWriteError() == nil {
308		var request *pendingKex
309		var sent bool
310
311		for request == nil || !sent {
312			var ok bool
313			select {
314			case request, ok = <-t.startKex:
315				if !ok {
316					break write
317				}
318			case <-t.requestKex:
319				break
320			}
321
322			if !sent {
323				if err := t.sendKexInit(); err != nil {
324					t.recordWriteError(err)
325					break
326				}
327				sent = true
328			}
329		}
330
331		if err := t.getWriteError(); err != nil {
332			if request != nil {
333				request.done <- err
334			}
335			break
336		}
337
338		// We're not servicing t.requestKex, but that is OK:
339		// we never block on sending to t.requestKex.
340
341		// We're not servicing t.startKex, but the remote end
342		// has just sent us a kexInitMsg, so it can't send
343		// another key change request, until we close the done
344		// channel on the pendingKex request.
345
346		err := t.enterKeyExchange(request.otherInit)
347
348		t.mu.Lock()
349		t.writeError = err
350		t.sentInitPacket = nil
351		t.sentInitMsg = nil
352
353		t.resetWriteThresholds()
354
355		// we have completed the key exchange. Since the
356		// reader is still blocked, it is safe to clear out
357		// the requestKex channel. This avoids the situation
358		// where: 1) we consumed our own request for the
359		// initial kex, and 2) the kex from the remote side
360		// caused another send on the requestKex channel,
361	clear:
362		for {
363			select {
364			case <-t.requestKex:
365				//
366			default:
367				break clear
368			}
369		}
370
371		request.done <- t.writeError
372
373		// kex finished. Push packets that we received while
374		// the kex was in progress. Don't look at t.startKex
375		// and don't increment writtenSinceKex: if we trigger
376		// another kex while we are still busy with the last
377		// one, things will become very confusing.
378		for _, p := range t.pendingPackets {
379			t.writeError = t.pushPacket(p)
380			if t.writeError != nil {
381				break
382			}
383		}
384		t.pendingPackets = t.pendingPackets[:0]
385		// Unblock writePacket if waiting for KEX.
386		t.writeCond.Broadcast()
387		t.mu.Unlock()
388	}
389
390	// Unblock reader.
391	t.conn.Close()
392
393	// drain startKex channel. We don't service t.requestKex
394	// because nobody does blocking sends there.
395	for request := range t.startKex {
396		request.done <- t.getWriteError()
397	}
398
399	// Mark that the loop is done so that Close can return.
400	close(t.kexLoopDone)
401}
402
403// The protocol uses uint32 for packet counters, so we can't let them
404// reach 1<<32.  We will actually read and write more packets than
405// this, though: the other side may send more packets, and after we
406// hit this limit on writing we will send a few more packets for the
407// key exchange itself.
408const packetRekeyThreshold = (1 << 31)
409
410func (t *handshakeTransport) resetReadThresholds() {
411	t.readPacketsLeft = packetRekeyThreshold
412	if t.config.RekeyThreshold > 0 {
413		t.readBytesLeft = int64(t.config.RekeyThreshold)
414	} else if t.algorithms != nil {
415		t.readBytesLeft = t.algorithms.Read.rekeyBytes()
416	} else {
417		t.readBytesLeft = 1 << 30
418	}
419}
420
421func (t *handshakeTransport) readOnePacket(first bool) ([]byte, error) {
422	p, err := t.conn.readPacket()
423	if err != nil {
424		return nil, err
425	}
426
427	if t.readPacketsLeft > 0 {
428		t.readPacketsLeft--
429	} else {
430		t.requestKeyExchange()
431	}
432
433	if t.readBytesLeft > 0 {
434		t.readBytesLeft -= int64(len(p))
435	} else {
436		t.requestKeyExchange()
437	}
438
439	if debugHandshake {
440		t.printPacket(p, false)
441	}
442
443	if first && p[0] != msgKexInit {
444		return nil, fmt.Errorf("ssh: first packet should be msgKexInit")
445	}
446
447	if p[0] != msgKexInit {
448		return p, nil
449	}
450
451	firstKex := t.sessionID == nil
452
453	kex := pendingKex{
454		done:      make(chan error, 1),
455		otherInit: p,
456	}
457	t.startKex <- &kex
458	err = <-kex.done
459
460	if debugHandshake {
461		log.Printf("%s exited key exchange (first %v), err %v", t.id(), firstKex, err)
462	}
463
464	if err != nil {
465		return nil, err
466	}
467
468	t.resetReadThresholds()
469
470	// By default, a key exchange is hidden from higher layers by
471	// translating it into msgIgnore.
472	successPacket := []byte{msgIgnore}
473	if firstKex {
474		// sendKexInit() for the first kex waits for
475		// msgNewKeys so the authentication process is
476		// guaranteed to happen over an encrypted transport.
477		successPacket = []byte{msgNewKeys}
478	}
479
480	return successPacket, nil
481}
482
483const (
484	kexStrictClient = "kex-strict-c-v00@openssh.com"
485	kexStrictServer = "kex-strict-s-v00@openssh.com"
486)
487
488// sendKexInit sends a key change message.
489func (t *handshakeTransport) sendKexInit() error {
490	t.mu.Lock()
491	defer t.mu.Unlock()
492	if t.sentInitMsg != nil {
493		// kexInits may be sent either in response to the other side,
494		// or because our side wants to initiate a key change, so we
495		// may have already sent a kexInit. In that case, don't send a
496		// second kexInit.
497		return nil
498	}
499
500	msg := &kexInitMsg{
501		CiphersClientServer:     t.config.Ciphers,
502		CiphersServerClient:     t.config.Ciphers,
503		MACsClientServer:        t.config.MACs,
504		MACsServerClient:        t.config.MACs,
505		CompressionClientServer: supportedCompressions,
506		CompressionServerClient: supportedCompressions,
507	}
508	io.ReadFull(t.config.Rand, msg.Cookie[:])
509
510	// We mutate the KexAlgos slice, in order to add the kex-strict extension algorithm,
511	// and possibly to add the ext-info extension algorithm. Since the slice may be the
512	// user owned KeyExchanges, we create our own slice in order to avoid using user
513	// owned memory by mistake.
514	msg.KexAlgos = make([]string, 0, len(t.config.KeyExchanges)+2) // room for kex-strict and ext-info
515	msg.KexAlgos = append(msg.KexAlgos, t.config.KeyExchanges...)
516
517	isServer := len(t.hostKeys) > 0
518	if isServer {
519		for _, k := range t.hostKeys {
520			// If k is a MultiAlgorithmSigner, we restrict the signature
521			// algorithms. If k is a AlgorithmSigner, presume it supports all
522			// signature algorithms associated with the key format. If k is not
523			// an AlgorithmSigner, we can only assume it only supports the
524			// algorithms that matches the key format. (This means that Sign
525			// can't pick a different default).
526			keyFormat := k.PublicKey().Type()
527
528			switch s := k.(type) {
529			case MultiAlgorithmSigner:
530				for _, algo := range algorithmsForKeyFormat(keyFormat) {
531					if slices.Contains(s.Algorithms(), underlyingAlgo(algo)) {
532						msg.ServerHostKeyAlgos = append(msg.ServerHostKeyAlgos, algo)
533					}
534				}
535			case AlgorithmSigner:
536				msg.ServerHostKeyAlgos = append(msg.ServerHostKeyAlgos, algorithmsForKeyFormat(keyFormat)...)
537			default:
538				msg.ServerHostKeyAlgos = append(msg.ServerHostKeyAlgos, keyFormat)
539			}
540		}
541
542		if t.sessionID == nil {
543			msg.KexAlgos = append(msg.KexAlgos, kexStrictServer)
544		}
545	} else {
546		msg.ServerHostKeyAlgos = t.hostKeyAlgorithms
547
548		// As a client we opt in to receiving SSH_MSG_EXT_INFO so we know what
549		// algorithms the server supports for public key authentication. See RFC
550		// 8308, Section 2.1.
551		//
552		// We also send the strict KEX mode extension algorithm, in order to opt
553		// into the strict KEX mode.
554		if firstKeyExchange := t.sessionID == nil; firstKeyExchange {
555			msg.KexAlgos = append(msg.KexAlgos, "ext-info-c")
556			msg.KexAlgos = append(msg.KexAlgos, kexStrictClient)
557		}
558
559	}
560
561	packet := Marshal(msg)
562
563	// writePacket destroys the contents, so save a copy.
564	packetCopy := make([]byte, len(packet))
565	copy(packetCopy, packet)
566
567	if err := t.pushPacket(packetCopy); err != nil {
568		return err
569	}
570
571	t.sentInitMsg = msg
572	t.sentInitPacket = packet
573
574	return nil
575}
576
577var errSendBannerPhase = errors.New("ssh: SendAuthBanner outside of authentication phase")
578
579func (t *handshakeTransport) writePacket(p []byte) error {
580	t.mu.Lock()
581	defer t.mu.Unlock()
582
583	switch p[0] {
584	case msgKexInit:
585		return errors.New("ssh: only handshakeTransport can send kexInit")
586	case msgNewKeys:
587		return errors.New("ssh: only handshakeTransport can send newKeys")
588	case msgUserAuthBanner:
589		if t.userAuthComplete {
590			return errSendBannerPhase
591		}
592	case msgUserAuthSuccess:
593		t.userAuthComplete = true
594	}
595
596	if t.writeError != nil {
597		return t.writeError
598	}
599
600	if t.sentInitMsg != nil {
601		if len(t.pendingPackets) < maxPendingPackets {
602			// Copy the packet so the writer can reuse the buffer.
603			cp := make([]byte, len(p))
604			copy(cp, p)
605			t.pendingPackets = append(t.pendingPackets, cp)
606			return nil
607		}
608		for t.sentInitMsg != nil {
609			// Block and wait for KEX to complete or an error.
610			t.writeCond.Wait()
611			if t.writeError != nil {
612				return t.writeError
613			}
614		}
615	}
616
617	if t.writeBytesLeft > 0 {
618		t.writeBytesLeft -= int64(len(p))
619	} else {
620		t.requestKeyExchange()
621	}
622
623	if t.writePacketsLeft > 0 {
624		t.writePacketsLeft--
625	} else {
626		t.requestKeyExchange()
627	}
628
629	if err := t.pushPacket(p); err != nil {
630		t.writeError = err
631		t.writeCond.Broadcast()
632	}
633
634	return nil
635}
636
637func (t *handshakeTransport) Close() error {
638	// Close the connection. This should cause the readLoop goroutine to wake up
639	// and close t.startKex, which will shut down kexLoop if running.
640	err := t.conn.Close()
641
642	// Wait for the kexLoop goroutine to complete.
643	// At that point we know that the readLoop goroutine is complete too,
644	// because kexLoop itself waits for readLoop to close the startKex channel.
645	<-t.kexLoopDone
646
647	return err
648}
649
650func (t *handshakeTransport) enterKeyExchange(otherInitPacket []byte) error {
651	if debugHandshake {
652		log.Printf("%s entered key exchange", t.id())
653	}
654
655	otherInit := &kexInitMsg{}
656	if err := Unmarshal(otherInitPacket, otherInit); err != nil {
657		return err
658	}
659
660	magics := handshakeMagics{
661		clientVersion: t.clientVersion,
662		serverVersion: t.serverVersion,
663		clientKexInit: otherInitPacket,
664		serverKexInit: t.sentInitPacket,
665	}
666
667	clientInit := otherInit
668	serverInit := t.sentInitMsg
669	isClient := len(t.hostKeys) == 0
670	if isClient {
671		clientInit, serverInit = serverInit, clientInit
672
673		magics.clientKexInit = t.sentInitPacket
674		magics.serverKexInit = otherInitPacket
675	}
676
677	var err error
678	t.algorithms, err = findAgreedAlgorithms(isClient, clientInit, serverInit)
679	if err != nil {
680		return err
681	}
682
683	if t.sessionID == nil && ((isClient && slices.Contains(serverInit.KexAlgos, kexStrictServer)) || (!isClient && slices.Contains(clientInit.KexAlgos, kexStrictClient))) {
684		t.strictMode = true
685		if err := t.conn.setStrictMode(); err != nil {
686			return err
687		}
688	}
689
690	// We don't send FirstKexFollows, but we handle receiving it.
691	//
692	// RFC 4253 section 7 defines the kex and the agreement method for
693	// first_kex_packet_follows. It states that the guessed packet
694	// should be ignored if the "kex algorithm and/or the host
695	// key algorithm is guessed wrong (server and client have
696	// different preferred algorithm), or if any of the other
697	// algorithms cannot be agreed upon". The other algorithms have
698	// already been checked above so the kex algorithm and host key
699	// algorithm are checked here.
700	if otherInit.FirstKexFollows && (clientInit.KexAlgos[0] != serverInit.KexAlgos[0] || clientInit.ServerHostKeyAlgos[0] != serverInit.ServerHostKeyAlgos[0]) {
701		// other side sent a kex message for the wrong algorithm,
702		// which we have to ignore.
703		if _, err := t.conn.readPacket(); err != nil {
704			return err
705		}
706	}
707
708	kex, ok := kexAlgoMap[t.algorithms.KeyExchange]
709	if !ok {
710		return fmt.Errorf("ssh: unexpected key exchange algorithm %v", t.algorithms.KeyExchange)
711	}
712
713	var result *kexResult
714	if len(t.hostKeys) > 0 {
715		result, err = t.server(kex, &magics)
716	} else {
717		result, err = t.client(kex, &magics)
718	}
719
720	if err != nil {
721		return err
722	}
723
724	firstKeyExchange := t.sessionID == nil
725	if firstKeyExchange {
726		t.sessionID = result.H
727	}
728	result.SessionID = t.sessionID
729
730	if err := t.conn.prepareKeyChange(t.algorithms, result); err != nil {
731		return err
732	}
733	if err = t.conn.writePacket([]byte{msgNewKeys}); err != nil {
734		return err
735	}
736
737	// On the server side, after the first SSH_MSG_NEWKEYS, send a SSH_MSG_EXT_INFO
738	// message with the server-sig-algs extension if the client supports it. See
739	// RFC 8308, Sections 2.4 and 3.1, and [PROTOCOL], Section 1.9.
740	if !isClient && firstKeyExchange && slices.Contains(clientInit.KexAlgos, "ext-info-c") {
741		supportedPubKeyAuthAlgosList := strings.Join(t.publicKeyAuthAlgorithms, ",")
742		extInfo := &extInfoMsg{
743			NumExtensions: 2,
744			Payload:       make([]byte, 0, 4+15+4+len(supportedPubKeyAuthAlgosList)+4+16+4+1),
745		}
746		extInfo.Payload = appendInt(extInfo.Payload, len("server-sig-algs"))
747		extInfo.Payload = append(extInfo.Payload, "server-sig-algs"...)
748		extInfo.Payload = appendInt(extInfo.Payload, len(supportedPubKeyAuthAlgosList))
749		extInfo.Payload = append(extInfo.Payload, supportedPubKeyAuthAlgosList...)
750		extInfo.Payload = appendInt(extInfo.Payload, len("ping@openssh.com"))
751		extInfo.Payload = append(extInfo.Payload, "ping@openssh.com"...)
752		extInfo.Payload = appendInt(extInfo.Payload, 1)
753		extInfo.Payload = append(extInfo.Payload, "0"...)
754		if err := t.conn.writePacket(Marshal(extInfo)); err != nil {
755			return err
756		}
757	}
758
759	if packet, err := t.conn.readPacket(); err != nil {
760		return err
761	} else if packet[0] != msgNewKeys {
762		return unexpectedMessageError(msgNewKeys, packet[0])
763	}
764
765	if firstKeyExchange {
766		// Indicates to the transport that the first key exchange is completed
767		// after receiving SSH_MSG_NEWKEYS.
768		t.conn.setInitialKEXDone()
769	}
770
771	return nil
772}
773
774// algorithmSignerWrapper is an AlgorithmSigner that only supports the default
775// key format algorithm.
776//
777// This is technically a violation of the AlgorithmSigner interface, but it
778// should be unreachable given where we use this. Anyway, at least it returns an
779// error instead of panicing or producing an incorrect signature.
780type algorithmSignerWrapper struct {
781	Signer
782}
783
784func (a algorithmSignerWrapper) SignWithAlgorithm(rand io.Reader, data []byte, algorithm string) (*Signature, error) {
785	if algorithm != underlyingAlgo(a.PublicKey().Type()) {
786		return nil, errors.New("ssh: internal error: algorithmSignerWrapper invoked with non-default algorithm")
787	}
788	return a.Sign(rand, data)
789}
790
791func pickHostKey(hostKeys []Signer, algo string) AlgorithmSigner {
792	for _, k := range hostKeys {
793		if s, ok := k.(MultiAlgorithmSigner); ok {
794			if !slices.Contains(s.Algorithms(), underlyingAlgo(algo)) {
795				continue
796			}
797		}
798
799		if algo == k.PublicKey().Type() {
800			return algorithmSignerWrapper{k}
801		}
802
803		k, ok := k.(AlgorithmSigner)
804		if !ok {
805			continue
806		}
807		for _, a := range algorithmsForKeyFormat(k.PublicKey().Type()) {
808			if algo == a {
809				return k
810			}
811		}
812	}
813	return nil
814}
815
816func (t *handshakeTransport) server(kex kexAlgorithm, magics *handshakeMagics) (*kexResult, error) {
817	hostKey := pickHostKey(t.hostKeys, t.algorithms.HostKey)
818	if hostKey == nil {
819		return nil, errors.New("ssh: internal error: negotiated unsupported signature type")
820	}
821
822	r, err := kex.Server(t.conn, t.config.Rand, magics, hostKey, t.algorithms.HostKey)
823	return r, err
824}
825
826func (t *handshakeTransport) client(kex kexAlgorithm, magics *handshakeMagics) (*kexResult, error) {
827	result, err := kex.Client(t.conn, t.config.Rand, magics)
828	if err != nil {
829		return nil, err
830	}
831
832	hostKey, err := ParsePublicKey(result.HostKey)
833	if err != nil {
834		return nil, err
835	}
836
837	if err := verifyHostKeySignature(hostKey, t.algorithms.HostKey, result); err != nil {
838		return nil, err
839	}
840
841	err = t.hostKeyCallback(t.dialAddress, t.remoteAddr, hostKey)
842	if err != nil {
843		return nil, err
844	}
845
846	return result, nil
847}