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
|
// Copyright 2020 The gVisor Authors.
//
// 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 fifo provides the implementation of data-link layer endpoints that
// wrap another endpoint and queues all outbound packets and asynchronously
// dispatches them to the lower endpoint.
package fifo
import (
"gvisor.dev/gvisor/pkg/sleep"
"gvisor.dev/gvisor/pkg/sync"
"gvisor.dev/gvisor/pkg/tcpip"
"gvisor.dev/gvisor/pkg/tcpip/header"
"gvisor.dev/gvisor/pkg/tcpip/stack"
)
// endpoint represents a LinkEndpoint which implements a FIFO queue for all
// outgoing packets. endpoint can have 1 or more underlying queueDispatchers.
// All outgoing packets are consistenly hashed to a single underlying queue
// using the PacketBuffer.Hash if set, otherwise all packets are queued to the
// first queue to avoid reordering in case of missing hash.
type endpoint struct {
dispatcher stack.NetworkDispatcher
lower stack.LinkEndpoint
wg sync.WaitGroup
dispatchers []*queueDispatcher
}
// queueDispatcher is responsible for dispatching all outbound packets in its
// queue. It will also smartly batch packets when possible and write them
// through the lower LinkEndpoint.
type queueDispatcher struct {
lower stack.LinkEndpoint
q *packetBufferQueue
newPacketWaker sleep.Waker
closeWaker sleep.Waker
}
// New creates a new fifo link endpoint with the n queues with maximum
// capacity of queueLen.
func New(lower stack.LinkEndpoint, n int, queueLen int) stack.LinkEndpoint {
e := &endpoint{
lower: lower,
}
// Create the required dispatchers
for i := 0; i < n; i++ {
qd := &queueDispatcher{
q: &packetBufferQueue{limit: queueLen},
lower: lower,
}
e.dispatchers = append(e.dispatchers, qd)
e.wg.Add(1)
go func() {
defer e.wg.Done()
qd.dispatchLoop()
}()
}
return e
}
func (q *queueDispatcher) dispatchLoop() {
const newPacketWakerID = 1
const closeWakerID = 2
s := sleep.Sleeper{}
s.AddWaker(&q.newPacketWaker, newPacketWakerID)
s.AddWaker(&q.closeWaker, closeWakerID)
defer s.Done()
const batchSize = 32
var batch stack.PacketBufferList
for {
id, ok := s.Fetch(true)
if ok && id == closeWakerID {
return
}
for pkt := q.q.dequeue(); pkt != nil; pkt = q.q.dequeue() {
batch.PushBack(pkt)
if batch.Len() < batchSize && !q.q.empty() {
continue
}
// We pass a protocol of zero here because each packet carries its
// NetworkProtocol.
q.lower.WritePackets(stack.RouteInfo{}, nil /* gso */, batch, 0 /* protocol */)
for pkt := batch.Front(); pkt != nil; pkt = pkt.Next() {
batch.Remove(pkt)
}
batch.Reset()
}
}
}
// DeliverNetworkPacket implements stack.NetworkDispatcher.DeliverNetworkPacket.
func (e *endpoint) DeliverNetworkPacket(remote, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) {
e.dispatcher.DeliverNetworkPacket(remote, local, protocol, pkt)
}
// DeliverOutboundPacket implements stack.NetworkDispatcher.DeliverOutboundPacket.
func (e *endpoint) DeliverOutboundPacket(remote, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) {
e.dispatcher.DeliverOutboundPacket(remote, local, protocol, pkt)
}
// Attach implements stack.LinkEndpoint.Attach.
func (e *endpoint) Attach(dispatcher stack.NetworkDispatcher) {
e.dispatcher = dispatcher
e.lower.Attach(e)
}
// IsAttached implements stack.LinkEndpoint.IsAttached.
func (e *endpoint) IsAttached() bool {
return e.dispatcher != nil
}
// MTU implements stack.LinkEndpoint.MTU.
func (e *endpoint) MTU() uint32 {
return e.lower.MTU()
}
// Capabilities implements stack.LinkEndpoint.Capabilities.
func (e *endpoint) Capabilities() stack.LinkEndpointCapabilities {
return e.lower.Capabilities()
}
// MaxHeaderLength implements stack.LinkEndpoint.MaxHeaderLength.
func (e *endpoint) MaxHeaderLength() uint16 {
return e.lower.MaxHeaderLength()
}
// LinkAddress implements stack.LinkEndpoint.LinkAddress.
func (e *endpoint) LinkAddress() tcpip.LinkAddress {
return e.lower.LinkAddress()
}
// GSOMaxSize returns the maximum GSO packet size.
func (e *endpoint) GSOMaxSize() uint32 {
if gso, ok := e.lower.(stack.GSOEndpoint); ok {
return gso.GSOMaxSize()
}
return 0
}
// WritePacket implements stack.LinkEndpoint.WritePacket.
func (e *endpoint) WritePacket(r stack.RouteInfo, gso *stack.GSO, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) *tcpip.Error {
// WritePacket caller's do not set the following fields in PacketBuffer
// so we populate them here.
pkt.EgressRoute = r
pkt.GSOOptions = gso
pkt.NetworkProtocolNumber = protocol
d := e.dispatchers[int(pkt.Hash)%len(e.dispatchers)]
if !d.q.enqueue(pkt) {
return tcpip.ErrNoBufferSpace
}
d.newPacketWaker.Assert()
return nil
}
// WritePackets implements stack.LinkEndpoint.WritePackets.
func (e *endpoint) WritePackets(r stack.RouteInfo, gso *stack.GSO, pkts stack.PacketBufferList, protocol tcpip.NetworkProtocolNumber) (int, *tcpip.Error) {
enqueued := 0
for pkt := pkts.Front(); pkt != nil; {
pkt.EgressRoute = r
pkt.GSOOptions = gso
pkt.NetworkProtocolNumber = protocol
d := e.dispatchers[int(pkt.Hash)%len(e.dispatchers)]
nxt := pkt.Next()
if !d.q.enqueue(pkt) {
if enqueued > 0 {
d.newPacketWaker.Assert()
}
return enqueued, tcpip.ErrNoBufferSpace
}
pkt = nxt
enqueued++
d.newPacketWaker.Assert()
}
return enqueued, nil
}
// Wait implements stack.LinkEndpoint.Wait.
func (e *endpoint) Wait() {
e.lower.Wait()
// The linkEP is gone. Teardown the outbound dispatcher goroutines.
for i := range e.dispatchers {
e.dispatchers[i].closeWaker.Assert()
}
e.wg.Wait()
}
// ARPHardwareType implements stack.LinkEndpoint.ARPHardwareType
func (e *endpoint) ARPHardwareType() header.ARPHardwareType {
return e.lower.ARPHardwareType()
}
// AddHeader implements stack.LinkEndpoint.AddHeader.
func (e *endpoint) AddHeader(local, remote tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) {
e.lower.AddHeader(local, remote, protocol, pkt)
}
|
