summaryrefslogtreecommitdiffhomepage
path: root/pkg/tcpip/transport/udp/endpoint.go
blob: 840e95302f98dd7965e06e4609b4afb65724aae0 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
// Copyright 2018 Google Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
//     http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

package udp

import (
	"math"
	"sync"

	"gvisor.googlesource.com/gvisor/pkg/sleep"
	"gvisor.googlesource.com/gvisor/pkg/tcpip"
	"gvisor.googlesource.com/gvisor/pkg/tcpip/buffer"
	"gvisor.googlesource.com/gvisor/pkg/tcpip/header"
	"gvisor.googlesource.com/gvisor/pkg/tcpip/stack"
	"gvisor.googlesource.com/gvisor/pkg/waiter"
)

// +stateify savable
type udpPacket struct {
	udpPacketEntry
	senderAddress tcpip.FullAddress
	data          buffer.VectorisedView `state:".(buffer.VectorisedView)"`
	timestamp     int64
	hasTimestamp  bool
	// views is used as buffer for data when its length is large
	// enough to store a VectorisedView.
	views [8]buffer.View `state:"nosave"`
}

type endpointState int

const (
	stateInitial endpointState = iota
	stateBound
	stateConnected
	stateClosed
)

// endpoint represents a UDP endpoint. This struct serves as the interface
// between users of the endpoint and the protocol implementation; it is legal to
// have concurrent goroutines make calls into the endpoint, they are properly
// synchronized.
//
// +stateify savable
type endpoint struct {
	// The following fields are initialized at creation time and do not
	// change throughout the lifetime of the endpoint.
	stack       *stack.Stack `state:"manual"`
	netProto    tcpip.NetworkProtocolNumber
	waiterQueue *waiter.Queue

	// The following fields are used to manage the receive queue, and are
	// protected by rcvMu.
	rcvMu         sync.Mutex `state:"nosave"`
	rcvReady      bool
	rcvList       udpPacketList
	rcvBufSizeMax int `state:".(int)"`
	rcvBufSize    int
	rcvClosed     bool
	rcvTimestamp  bool

	// The following fields are protected by the mu mutex.
	mu           sync.RWMutex `state:"nosave"`
	sndBufSize   int
	id           stack.TransportEndpointID
	state        endpointState
	bindNICID    tcpip.NICID
	regNICID     tcpip.NICID
	route        stack.Route `state:"manual"`
	dstPort      uint16
	v6only       bool
	multicastTTL uint8

	// shutdownFlags represent the current shutdown state of the endpoint.
	shutdownFlags tcpip.ShutdownFlags

	// multicastMemberships that need to be remvoed when the endpoint is
	// closed. Protected by the mu mutex.
	multicastMemberships []multicastMembership

	// effectiveNetProtos contains the network protocols actually in use. In
	// most cases it will only contain "netProto", but in cases like IPv6
	// endpoints with v6only set to false, this could include multiple
	// protocols (e.g., IPv6 and IPv4) or a single different protocol (e.g.,
	// IPv4 when IPv6 endpoint is bound or connected to an IPv4 mapped
	// address).
	effectiveNetProtos []tcpip.NetworkProtocolNumber
}

type multicastMembership struct {
	nicID         tcpip.NICID
	multicastAddr tcpip.Address
}

func newEndpoint(stack *stack.Stack, netProto tcpip.NetworkProtocolNumber, waiterQueue *waiter.Queue) *endpoint {
	return &endpoint{
		stack:       stack,
		netProto:    netProto,
		waiterQueue: waiterQueue,
		// RFC 1075 section 5.4 recommends a TTL of 1 for membership
		// requests.
		//
		// RFC 5135 4.2.1 appears to assume that IGMP messages have a
		// TTL of 1.
		//
		// RFC 5135 Appendix A defines TTL=1: A multicast source that
		// wants its traffic to not traverse a router (e.g., leave a
		// home network) may find it useful to send traffic with IP
		// TTL=1.
		//
		// Linux defaults to TTL=1.
		multicastTTL:  1,
		rcvBufSizeMax: 32 * 1024,
		sndBufSize:    32 * 1024,
	}
}

// NewConnectedEndpoint creates a new endpoint in the connected state using the
// provided route.
func NewConnectedEndpoint(stack *stack.Stack, r *stack.Route, id stack.TransportEndpointID, waiterQueue *waiter.Queue) (tcpip.Endpoint, *tcpip.Error) {
	ep := newEndpoint(stack, r.NetProto, waiterQueue)

	// Register new endpoint so that packets are routed to it.
	if err := stack.RegisterTransportEndpoint(r.NICID(), []tcpip.NetworkProtocolNumber{r.NetProto}, ProtocolNumber, id, ep); err != nil {
		ep.Close()
		return nil, err
	}

	ep.id = id
	ep.route = r.Clone()
	ep.dstPort = id.RemotePort
	ep.regNICID = r.NICID()

	ep.state = stateConnected

	return ep, nil
}

// Close puts the endpoint in a closed state and frees all resources
// associated with it.
func (e *endpoint) Close() {
	e.mu.Lock()
	e.shutdownFlags = tcpip.ShutdownRead | tcpip.ShutdownWrite

	switch e.state {
	case stateBound, stateConnected:
		e.stack.UnregisterTransportEndpoint(e.regNICID, e.effectiveNetProtos, ProtocolNumber, e.id)
		e.stack.ReleasePort(e.effectiveNetProtos, ProtocolNumber, e.id.LocalAddress, e.id.LocalPort)
	}

	for _, mem := range e.multicastMemberships {
		e.stack.LeaveGroup(e.netProto, mem.nicID, mem.multicastAddr)
	}
	e.multicastMemberships = nil

	// Close the receive list and drain it.
	e.rcvMu.Lock()
	e.rcvClosed = true
	e.rcvBufSize = 0
	for !e.rcvList.Empty() {
		p := e.rcvList.Front()
		e.rcvList.Remove(p)
	}
	e.rcvMu.Unlock()

	e.route.Release()

	// Update the state.
	e.state = stateClosed

	e.mu.Unlock()

	e.waiterQueue.Notify(waiter.EventHUp | waiter.EventErr | waiter.EventIn | waiter.EventOut)
}

// Read reads data from the endpoint. This method does not block if
// there is no data pending.
func (e *endpoint) Read(addr *tcpip.FullAddress) (buffer.View, tcpip.ControlMessages, *tcpip.Error) {
	e.rcvMu.Lock()

	if e.rcvList.Empty() {
		err := tcpip.ErrWouldBlock
		if e.rcvClosed {
			err = tcpip.ErrClosedForReceive
		}
		e.rcvMu.Unlock()
		return buffer.View{}, tcpip.ControlMessages{}, err
	}

	p := e.rcvList.Front()
	e.rcvList.Remove(p)
	e.rcvBufSize -= p.data.Size()
	ts := e.rcvTimestamp

	e.rcvMu.Unlock()

	if addr != nil {
		*addr = p.senderAddress
	}

	if ts && !p.hasTimestamp {
		// Linux uses the current time.
		p.timestamp = e.stack.NowNanoseconds()
	}

	return p.data.ToView(), tcpip.ControlMessages{HasTimestamp: ts, Timestamp: p.timestamp}, nil
}

// prepareForWrite prepares the endpoint for sending data. In particular, it
// binds it if it's still in the initial state. To do so, it must first
// reacquire the mutex in exclusive mode.
//
// Returns true for retry if preparation should be retried.
func (e *endpoint) prepareForWrite(to *tcpip.FullAddress) (retry bool, err *tcpip.Error) {
	switch e.state {
	case stateInitial:
	case stateConnected:
		return false, nil

	case stateBound:
		if to == nil {
			return false, tcpip.ErrDestinationRequired
		}
		return false, nil
	default:
		return false, tcpip.ErrInvalidEndpointState
	}

	e.mu.RUnlock()
	defer e.mu.RLock()

	e.mu.Lock()
	defer e.mu.Unlock()

	// The state changed when we released the shared locked and re-acquired
	// it in exclusive mode. Try again.
	if e.state != stateInitial {
		return true, nil
	}

	// The state is still 'initial', so try to bind the endpoint.
	if err := e.bindLocked(tcpip.FullAddress{}, nil); err != nil {
		return false, err
	}

	return true, nil
}

// Write writes data to the endpoint's peer. This method does not block
// if the data cannot be written.
func (e *endpoint) Write(p tcpip.Payload, opts tcpip.WriteOptions) (uintptr, <-chan struct{}, *tcpip.Error) {
	// MSG_MORE is unimplemented. (This also means that MSG_EOR is a no-op.)
	if opts.More {
		return 0, nil, tcpip.ErrInvalidOptionValue
	}

	if p.Size() > math.MaxUint16 {
		// Payload can't possibly fit in a packet.
		return 0, nil, tcpip.ErrMessageTooLong
	}

	to := opts.To

	e.mu.RLock()
	defer e.mu.RUnlock()

	// If we've shutdown with SHUT_WR we are in an invalid state for sending.
	if e.shutdownFlags&tcpip.ShutdownWrite != 0 {
		return 0, nil, tcpip.ErrClosedForSend
	}

	// Prepare for write.
	for {
		retry, err := e.prepareForWrite(to)
		if err != nil {
			return 0, nil, err
		}

		if !retry {
			break
		}
	}

	var route *stack.Route
	var dstPort uint16
	if to == nil {
		route = &e.route
		dstPort = e.dstPort

		if route.IsResolutionRequired() {
			// Promote lock to exclusive if using a shared route, given that it may need to
			// change in Route.Resolve() call below.
			e.mu.RUnlock()
			defer e.mu.RLock()

			e.mu.Lock()
			defer e.mu.Unlock()

			// Recheck state after lock was re-acquired.
			if e.state != stateConnected {
				return 0, nil, tcpip.ErrInvalidEndpointState
			}
		}
	} else {
		// Reject destination address if it goes through a different
		// NIC than the endpoint was bound to.
		nicid := to.NIC
		if e.bindNICID != 0 {
			if nicid != 0 && nicid != e.bindNICID {
				return 0, nil, tcpip.ErrNoRoute
			}

			nicid = e.bindNICID
		}

		toCopy := *to
		to = &toCopy
		netProto, err := e.checkV4Mapped(to, false)
		if err != nil {
			return 0, nil, err
		}

		// Find the enpoint.
		r, err := e.stack.FindRoute(nicid, e.id.LocalAddress, to.Addr, netProto)
		if err != nil {
			return 0, nil, err
		}
		defer r.Release()

		route = &r
		dstPort = to.Port
	}

	if route.IsResolutionRequired() {
		waker := &sleep.Waker{}
		if ch, err := route.Resolve(waker); err != nil {
			if err == tcpip.ErrWouldBlock {
				// Link address needs to be resolved. Resolution was triggered the background.
				// Better luck next time.
				route.RemoveWaker(waker)
				return 0, ch, tcpip.ErrNoLinkAddress
			}
			return 0, nil, err
		}
	}

	v, err := p.Get(p.Size())
	if err != nil {
		return 0, nil, err
	}

	ttl := route.DefaultTTL()
	if header.IsV4MulticastAddress(route.RemoteAddress) || header.IsV6MulticastAddress(route.RemoteAddress) {
		ttl = e.multicastTTL
	}

	if err := sendUDP(route, buffer.View(v).ToVectorisedView(), e.id.LocalPort, dstPort, ttl); err != nil {
		return 0, nil, err
	}
	return uintptr(len(v)), nil, nil
}

// Peek only returns data from a single datagram, so do nothing here.
func (e *endpoint) Peek([][]byte) (uintptr, tcpip.ControlMessages, *tcpip.Error) {
	return 0, tcpip.ControlMessages{}, nil
}

// SetSockOpt sets a socket option. Currently not supported.
func (e *endpoint) SetSockOpt(opt interface{}) *tcpip.Error {
	// TODO: Actually implement this.
	switch v := opt.(type) {
	case tcpip.V6OnlyOption:
		// We only recognize this option on v6 endpoints.
		if e.netProto != header.IPv6ProtocolNumber {
			return tcpip.ErrInvalidEndpointState
		}

		e.mu.Lock()
		defer e.mu.Unlock()

		// We only allow this to be set when we're in the initial state.
		if e.state != stateInitial {
			return tcpip.ErrInvalidEndpointState
		}

		e.v6only = v != 0

	case tcpip.TimestampOption:
		e.rcvMu.Lock()
		e.rcvTimestamp = v != 0
		e.rcvMu.Unlock()

	case tcpip.MulticastTTLOption:
		e.mu.Lock()
		e.multicastTTL = uint8(v)
		e.mu.Unlock()

	case tcpip.AddMembershipOption:
		nicID := v.NIC
		if v.InterfaceAddr != header.IPv4Any {
			nicID = e.stack.CheckLocalAddress(nicID, e.netProto, v.InterfaceAddr)
		}
		if nicID == 0 {
			return tcpip.ErrNoRoute
		}

		// TODO: check that v.MulticastAddr is a multicast address.
		if err := e.stack.JoinGroup(e.netProto, nicID, v.MulticastAddr); err != nil {
			return err
		}

		e.mu.Lock()
		defer e.mu.Unlock()

		e.multicastMemberships = append(e.multicastMemberships, multicastMembership{nicID, v.MulticastAddr})

	case tcpip.RemoveMembershipOption:
		nicID := v.NIC
		if v.InterfaceAddr != header.IPv4Any {
			nicID = e.stack.CheckLocalAddress(nicID, e.netProto, v.InterfaceAddr)
		}
		if nicID == 0 {
			return tcpip.ErrNoRoute
		}

		// TODO: check that v.MulticastAddr is a multicast address.
		if err := e.stack.LeaveGroup(e.netProto, nicID, v.MulticastAddr); err != nil {
			return err
		}

		e.mu.Lock()
		defer e.mu.Unlock()
		for i, mem := range e.multicastMemberships {
			if mem.nicID == nicID && mem.multicastAddr == v.MulticastAddr {
				// Only remove the first match, so that each added membership above is
				// paired with exactly 1 removal.
				e.multicastMemberships[i] = e.multicastMemberships[len(e.multicastMemberships)-1]
				e.multicastMemberships = e.multicastMemberships[:len(e.multicastMemberships)-1]
				break
			}
		}
	}
	return nil
}

// GetSockOpt implements tcpip.Endpoint.GetSockOpt.
func (e *endpoint) GetSockOpt(opt interface{}) *tcpip.Error {
	switch o := opt.(type) {
	case tcpip.ErrorOption:
		return nil

	case *tcpip.SendBufferSizeOption:
		e.mu.Lock()
		*o = tcpip.SendBufferSizeOption(e.sndBufSize)
		e.mu.Unlock()
		return nil

	case *tcpip.ReceiveBufferSizeOption:
		e.rcvMu.Lock()
		*o = tcpip.ReceiveBufferSizeOption(e.rcvBufSizeMax)
		e.rcvMu.Unlock()
		return nil

	case *tcpip.V6OnlyOption:
		// We only recognize this option on v6 endpoints.
		if e.netProto != header.IPv6ProtocolNumber {
			return tcpip.ErrUnknownProtocolOption
		}

		e.mu.Lock()
		v := e.v6only
		e.mu.Unlock()

		*o = 0
		if v {
			*o = 1
		}
		return nil

	case *tcpip.ReceiveQueueSizeOption:
		e.rcvMu.Lock()
		if e.rcvList.Empty() {
			*o = 0
		} else {
			p := e.rcvList.Front()
			*o = tcpip.ReceiveQueueSizeOption(p.data.Size())
		}
		e.rcvMu.Unlock()
		return nil

	case *tcpip.TimestampOption:
		e.rcvMu.Lock()
		*o = 0
		if e.rcvTimestamp {
			*o = 1
		}
		e.rcvMu.Unlock()

	case *tcpip.MulticastTTLOption:
		e.mu.Lock()
		*o = tcpip.MulticastTTLOption(e.multicastTTL)
		e.mu.Unlock()
		return nil
	}

	return tcpip.ErrUnknownProtocolOption
}

// sendUDP sends a UDP segment via the provided network endpoint and under the
// provided identity.
func sendUDP(r *stack.Route, data buffer.VectorisedView, localPort, remotePort uint16, ttl uint8) *tcpip.Error {
	// Allocate a buffer for the UDP header.
	hdr := buffer.NewPrependable(header.UDPMinimumSize + int(r.MaxHeaderLength()))

	// Initialize the header.
	udp := header.UDP(hdr.Prepend(header.UDPMinimumSize))

	length := uint16(hdr.UsedLength() + data.Size())
	udp.Encode(&header.UDPFields{
		SrcPort: localPort,
		DstPort: remotePort,
		Length:  length,
	})

	// Only calculate the checksum if offloading isn't supported.
	if r.Capabilities()&stack.CapabilityChecksumOffload == 0 {
		xsum := r.PseudoHeaderChecksum(ProtocolNumber)
		for _, v := range data.Views() {
			xsum = header.Checksum(v, xsum)
		}
		udp.SetChecksum(^udp.CalculateChecksum(xsum, length))
	}

	// Track count of packets sent.
	r.Stats().UDP.PacketsSent.Increment()

	return r.WritePacket(hdr, data, ProtocolNumber, ttl)
}

func (e *endpoint) checkV4Mapped(addr *tcpip.FullAddress, allowMismatch bool) (tcpip.NetworkProtocolNumber, *tcpip.Error) {
	netProto := e.netProto
	if header.IsV4MappedAddress(addr.Addr) {
		// Fail if using a v4 mapped address on a v6only endpoint.
		if e.v6only {
			return 0, tcpip.ErrNoRoute
		}

		netProto = header.IPv4ProtocolNumber
		addr.Addr = addr.Addr[header.IPv6AddressSize-header.IPv4AddressSize:]
		if addr.Addr == "\x00\x00\x00\x00" {
			addr.Addr = ""
		}

		// Fail if we are bound to an IPv6 address.
		if !allowMismatch && len(e.id.LocalAddress) == 16 {
			return 0, tcpip.ErrNetworkUnreachable
		}
	}

	// Fail if we're bound to an address length different from the one we're
	// checking.
	if l := len(e.id.LocalAddress); l != 0 && l != len(addr.Addr) {
		return 0, tcpip.ErrInvalidEndpointState
	}

	return netProto, nil
}

// Connect connects the endpoint to its peer. Specifying a NIC is optional.
func (e *endpoint) Connect(addr tcpip.FullAddress) *tcpip.Error {
	if addr.Port == 0 {
		// We don't support connecting to port zero.
		return tcpip.ErrInvalidEndpointState
	}

	e.mu.Lock()
	defer e.mu.Unlock()

	nicid := addr.NIC
	var localPort uint16
	switch e.state {
	case stateInitial:
	case stateBound, stateConnected:
		localPort = e.id.LocalPort
		if e.bindNICID == 0 {
			break
		}

		if nicid != 0 && nicid != e.bindNICID {
			return tcpip.ErrInvalidEndpointState
		}

		nicid = e.bindNICID
	default:
		return tcpip.ErrInvalidEndpointState
	}

	netProto, err := e.checkV4Mapped(&addr, false)
	if err != nil {
		return err
	}

	// Find a route to the desired destination.
	r, err := e.stack.FindRoute(nicid, e.id.LocalAddress, addr.Addr, netProto)
	if err != nil {
		return err
	}
	defer r.Release()

	id := stack.TransportEndpointID{
		LocalAddress:  r.LocalAddress,
		LocalPort:     localPort,
		RemotePort:    addr.Port,
		RemoteAddress: r.RemoteAddress,
	}

	// Even if we're connected, this endpoint can still be used to send
	// packets on a different network protocol, so we register both even if
	// v6only is set to false and this is an ipv6 endpoint.
	netProtos := []tcpip.NetworkProtocolNumber{netProto}
	if netProto == header.IPv6ProtocolNumber && !e.v6only {
		netProtos = []tcpip.NetworkProtocolNumber{
			header.IPv4ProtocolNumber,
			header.IPv6ProtocolNumber,
		}
	}

	id, err = e.registerWithStack(nicid, netProtos, id)
	if err != nil {
		return err
	}

	// Remove the old registration.
	if e.id.LocalPort != 0 {
		e.stack.UnregisterTransportEndpoint(e.regNICID, e.effectiveNetProtos, ProtocolNumber, e.id)
	}

	e.id = id
	e.route = r.Clone()
	e.dstPort = addr.Port
	e.regNICID = nicid
	e.effectiveNetProtos = netProtos

	e.state = stateConnected

	e.rcvMu.Lock()
	e.rcvReady = true
	e.rcvMu.Unlock()

	return nil
}

// ConnectEndpoint is not supported.
func (*endpoint) ConnectEndpoint(tcpip.Endpoint) *tcpip.Error {
	return tcpip.ErrInvalidEndpointState
}

// Shutdown closes the read and/or write end of the endpoint connection
// to its peer.
func (e *endpoint) Shutdown(flags tcpip.ShutdownFlags) *tcpip.Error {
	e.mu.Lock()
	defer e.mu.Unlock()

	// A socket in the bound state can still receive multicast messages,
	// so we need to notify waiters on shutdown.
	if e.state != stateBound && e.state != stateConnected {
		return tcpip.ErrNotConnected
	}

	e.shutdownFlags |= flags

	if flags&tcpip.ShutdownRead != 0 {
		e.rcvMu.Lock()
		wasClosed := e.rcvClosed
		e.rcvClosed = true
		e.rcvMu.Unlock()

		if !wasClosed {
			e.waiterQueue.Notify(waiter.EventIn)
		}
	}

	return nil
}

// Listen is not supported by UDP, it just fails.
func (*endpoint) Listen(int) *tcpip.Error {
	return tcpip.ErrNotSupported
}

// Accept is not supported by UDP, it just fails.
func (*endpoint) Accept() (tcpip.Endpoint, *waiter.Queue, *tcpip.Error) {
	return nil, nil, tcpip.ErrNotSupported
}

func (e *endpoint) registerWithStack(nicid tcpip.NICID, netProtos []tcpip.NetworkProtocolNumber, id stack.TransportEndpointID) (stack.TransportEndpointID, *tcpip.Error) {
	if e.id.LocalPort == 0 {
		port, err := e.stack.ReservePort(netProtos, ProtocolNumber, id.LocalAddress, id.LocalPort)
		if err != nil {
			return id, err
		}
		id.LocalPort = port
	}

	err := e.stack.RegisterTransportEndpoint(nicid, netProtos, ProtocolNumber, id, e)
	if err != nil {
		e.stack.ReleasePort(netProtos, ProtocolNumber, id.LocalAddress, id.LocalPort)
	}
	return id, err
}

func (e *endpoint) bindLocked(addr tcpip.FullAddress, commit func() *tcpip.Error) *tcpip.Error {
	// Don't allow binding once endpoint is not in the initial state
	// anymore.
	if e.state != stateInitial {
		return tcpip.ErrInvalidEndpointState
	}

	netProto, err := e.checkV4Mapped(&addr, true)
	if err != nil {
		return err
	}

	// Expand netProtos to include v4 and v6 if the caller is binding to a
	// wildcard (empty) address, and this is an IPv6 endpoint with v6only
	// set to false.
	netProtos := []tcpip.NetworkProtocolNumber{netProto}
	if netProto == header.IPv6ProtocolNumber && !e.v6only && addr.Addr == "" {
		netProtos = []tcpip.NetworkProtocolNumber{
			header.IPv6ProtocolNumber,
			header.IPv4ProtocolNumber,
		}
	}

	if len(addr.Addr) != 0 {
		// A local address was specified, verify that it's valid.
		if e.stack.CheckLocalAddress(addr.NIC, netProto, addr.Addr) == 0 {
			return tcpip.ErrBadLocalAddress
		}
	}

	id := stack.TransportEndpointID{
		LocalPort:    addr.Port,
		LocalAddress: addr.Addr,
	}
	id, err = e.registerWithStack(addr.NIC, netProtos, id)
	if err != nil {
		return err
	}
	if commit != nil {
		if err := commit(); err != nil {
			// Unregister, the commit failed.
			e.stack.UnregisterTransportEndpoint(addr.NIC, netProtos, ProtocolNumber, id)
			e.stack.ReleasePort(netProtos, ProtocolNumber, id.LocalAddress, id.LocalPort)
			return err
		}
	}

	e.id = id
	e.regNICID = addr.NIC
	e.effectiveNetProtos = netProtos

	// Mark endpoint as bound.
	e.state = stateBound

	e.rcvMu.Lock()
	e.rcvReady = true
	e.rcvMu.Unlock()

	return nil
}

// Bind binds the endpoint to a specific local address and port.
// Specifying a NIC is optional.
func (e *endpoint) Bind(addr tcpip.FullAddress, commit func() *tcpip.Error) *tcpip.Error {
	e.mu.Lock()
	defer e.mu.Unlock()

	err := e.bindLocked(addr, commit)
	if err != nil {
		return err
	}

	e.bindNICID = addr.NIC

	return nil
}

// GetLocalAddress returns the address to which the endpoint is bound.
func (e *endpoint) GetLocalAddress() (tcpip.FullAddress, *tcpip.Error) {
	e.mu.RLock()
	defer e.mu.RUnlock()

	return tcpip.FullAddress{
		NIC:  e.regNICID,
		Addr: e.id.LocalAddress,
		Port: e.id.LocalPort,
	}, nil
}

// GetRemoteAddress returns the address to which the endpoint is connected.
func (e *endpoint) GetRemoteAddress() (tcpip.FullAddress, *tcpip.Error) {
	e.mu.RLock()
	defer e.mu.RUnlock()

	if e.state != stateConnected {
		return tcpip.FullAddress{}, tcpip.ErrNotConnected
	}

	return tcpip.FullAddress{
		NIC:  e.regNICID,
		Addr: e.id.RemoteAddress,
		Port: e.id.RemotePort,
	}, nil
}

// Readiness returns the current readiness of the endpoint. For example, if
// waiter.EventIn is set, the endpoint is immediately readable.
func (e *endpoint) Readiness(mask waiter.EventMask) waiter.EventMask {
	// The endpoint is always writable.
	result := waiter.EventOut & mask

	// Determine if the endpoint is readable if requested.
	if (mask & waiter.EventIn) != 0 {
		e.rcvMu.Lock()
		if !e.rcvList.Empty() || e.rcvClosed {
			result |= waiter.EventIn
		}
		e.rcvMu.Unlock()
	}

	return result
}

// HandlePacket is called by the stack when new packets arrive to this transport
// endpoint.
func (e *endpoint) HandlePacket(r *stack.Route, id stack.TransportEndpointID, vv buffer.VectorisedView) {
	// Get the header then trim it from the view.
	hdr := header.UDP(vv.First())
	if int(hdr.Length()) > vv.Size() {
		// Malformed packet.
		e.stack.Stats().UDP.MalformedPacketsReceived.Increment()
		return
	}

	vv.TrimFront(header.UDPMinimumSize)

	e.rcvMu.Lock()
	e.stack.Stats().UDP.PacketsReceived.Increment()

	// Drop the packet if our buffer is currently full.
	if !e.rcvReady || e.rcvClosed || e.rcvBufSize >= e.rcvBufSizeMax {
		e.stack.Stats().UDP.ReceiveBufferErrors.Increment()
		e.rcvMu.Unlock()
		return
	}

	wasEmpty := e.rcvBufSize == 0

	// Push new packet into receive list and increment the buffer size.
	pkt := &udpPacket{
		senderAddress: tcpip.FullAddress{
			NIC:  r.NICID(),
			Addr: id.RemoteAddress,
			Port: hdr.SourcePort(),
		},
	}
	pkt.data = vv.Clone(pkt.views[:])
	e.rcvList.PushBack(pkt)
	e.rcvBufSize += vv.Size()

	if e.rcvTimestamp {
		pkt.timestamp = e.stack.NowNanoseconds()
		pkt.hasTimestamp = true
	}

	e.rcvMu.Unlock()

	// Notify any waiters that there's data to be read now.
	if wasEmpty {
		e.waiterQueue.Notify(waiter.EventIn)
	}
}

// HandleControlPacket implements stack.TransportEndpoint.HandleControlPacket.
func (e *endpoint) HandleControlPacket(id stack.TransportEndpointID, typ stack.ControlType, extra uint32, vv buffer.VectorisedView) {
}