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
|
// Copyright 2019 Google LLC
//
// 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.
// +build linux,amd64
package fdbased
import (
"encoding/binary"
"fmt"
"syscall"
"unsafe"
"golang.org/x/sys/unix"
"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/link/rawfile"
)
const (
tPacketAlignment = uintptr(16)
tpStatusKernel = 0
tpStatusUser = 1
tpStatusCopy = 2
tpStatusLosing = 4
)
// We overallocate the frame size to accommodate space for the
// TPacketHdr+RawSockAddrLinkLayer+MAC header and any padding.
//
// Memory allocated for the ring buffer: tpBlockSize * tpBlockNR = 2 MiB
//
// NOTE: Frames need to be aligned at 16 byte boundaries.
const (
tpFrameSize = 65536 + 128
tpBlockSize = tpFrameSize * 32
tpBlockNR = 1
tpFrameNR = (tpBlockSize * tpBlockNR) / tpFrameSize
)
// tPacketAlign aligns the pointer v at a tPacketAlignment boundary. Direct
// translation of the TPACKET_ALIGN macro in <linux/if_packet.h>.
func tPacketAlign(v uintptr) uintptr {
return (v + tPacketAlignment - 1) & uintptr(^(tPacketAlignment - 1))
}
// tPacketHdrlen is the TPACKET_HDRLEN variable defined in <linux/if_packet.h>.
var tPacketHdrlen = tPacketAlign(unsafe.Sizeof(tPacketHdr{}) + unsafe.Sizeof(syscall.RawSockaddrLinklayer{}))
// tPacketReq is the tpacket_req structure as described in
// https://www.kernel.org/doc/Documentation/networking/packet_mmap.txt
type tPacketReq struct {
tpBlockSize uint32
tpBlockNR uint32
tpFrameSize uint32
tpFrameNR uint32
}
// tPacketHdr is tpacket_hdr structure as described in <linux/if_packet.h>
type tPacketHdr []byte
const (
tpStatusOffset = 0
tpLenOffset = 8
tpSnapLenOffset = 12
tpMacOffset = 16
tpNetOffset = 18
tpSecOffset = 20
tpUSecOffset = 24
)
func (t tPacketHdr) tpStatus() uint32 {
return binary.LittleEndian.Uint32(t[tpStatusOffset:])
}
func (t tPacketHdr) setTPStatus(status uint32) {
binary.LittleEndian.PutUint32(t[tpStatusOffset:], status)
}
func (t tPacketHdr) tpLen() uint32 {
return binary.LittleEndian.Uint32(t[tpLenOffset:])
}
func (t tPacketHdr) tpSnapLen() uint32 {
return binary.LittleEndian.Uint32(t[tpSnapLenOffset:])
}
func (t tPacketHdr) tpMac() uint16 {
return binary.LittleEndian.Uint16(t[tpMacOffset:])
}
func (t tPacketHdr) tpNet() uint16 {
return binary.LittleEndian.Uint16(t[tpNetOffset:])
}
func (t tPacketHdr) tpSec() uint32 {
return binary.LittleEndian.Uint32(t[tpSecOffset:])
}
func (t tPacketHdr) tpUSec() uint32 {
return binary.LittleEndian.Uint32(t[tpUSecOffset:])
}
func (t tPacketHdr) Payload() []byte {
return t[uint32(t.tpMac()) : uint32(t.tpMac())+t.tpSnapLen()]
}
func (e *endpoint) setupPacketRXRing() error {
tReq := tPacketReq{
tpBlockSize: uint32(tpBlockSize),
tpBlockNR: uint32(tpBlockNR),
tpFrameSize: uint32(tpFrameSize),
tpFrameNR: uint32(tpFrameNR),
}
// Setup PACKET_RX_RING.
if err := setsockopt(e.fd, syscall.SOL_PACKET, syscall.PACKET_RX_RING, unsafe.Pointer(&tReq), unsafe.Sizeof(tReq)); err != nil {
return fmt.Errorf("failed to enable PACKET_RX_RING: %v", err)
}
// Let's mmap the blocks.
sz := tpBlockSize * tpBlockNR
buf, err := syscall.Mmap(e.fd, 0, sz, syscall.PROT_READ|syscall.PROT_WRITE, syscall.MAP_SHARED)
if err != nil {
return fmt.Errorf("syscall.Mmap(...,0, %v, ...) failed = %v", sz, err)
}
e.ringBuffer = buf
return nil
}
func (e *endpoint) readMMappedPacket() ([]byte, *tcpip.Error) {
hdr := (tPacketHdr)(e.ringBuffer[0+e.ringOffset*tpFrameSize:])
for (hdr.tpStatus() & tpStatusUser) == 0 {
event := rawfile.PollEvent{
FD: int32(e.fd),
Events: unix.POLLIN | unix.POLLERR,
}
_, errno := rawfile.BlockingPoll(&event, 1, -1)
if errno != 0 {
if errno == syscall.EINTR {
continue
}
return nil, rawfile.TranslateErrno(errno)
}
if hdr.tpStatus()&tpStatusCopy != 0 {
continue
}
if hdr.tpStatus()&tpStatusLosing != 0 {
continue
}
}
// Copy out the packet from the mmapped frame to a locally owned buffer.
pkt := make([]byte, hdr.tpSnapLen())
copy(pkt, hdr.Payload())
// Release packet to kernel.
hdr.setTPStatus(tpStatusKernel)
e.ringOffset = (e.ringOffset + 1) % tpFrameNR
return pkt, nil
}
// packetMMapDispatch reads packets from an mmaped ring buffer and dispatches
// them to the network stack.
func (e *endpoint) packetMMapDispatch() (bool, *tcpip.Error) {
pkt, err := e.readMMappedPacket()
if err != nil {
return false, err
}
var (
p tcpip.NetworkProtocolNumber
remote, local tcpip.LinkAddress
)
if e.hdrSize > 0 {
eth := header.Ethernet(pkt)
p = eth.Type()
remote = eth.SourceAddress()
local = eth.DestinationAddress()
} else {
// We don't get any indication of what the packet is, so try to guess
// if it's an IPv4 or IPv6 packet.
switch header.IPVersion(pkt) {
case header.IPv4Version:
p = header.IPv4ProtocolNumber
case header.IPv6Version:
p = header.IPv6ProtocolNumber
default:
return true, nil
}
}
pkt = pkt[e.hdrSize:]
e.dispatcher.DeliverNetworkPacket(e, remote, local, p, buffer.NewVectorisedView(len(pkt), []buffer.View{buffer.View(pkt)}))
return true, nil
}
func setsockopt(fd, level, name int, val unsafe.Pointer, vallen uintptr) error {
if _, _, errno := syscall.Syscall6(syscall.SYS_SETSOCKOPT, uintptr(fd), uintptr(level), uintptr(name), uintptr(val), vallen, 0); errno != 0 {
return error(errno)
}
return nil
}
|
