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
|
// Copyright 2018 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.
//go:build linux
// +build linux
package fdbased
import (
"fmt"
"golang.org/x/sys/unix"
"gvisor.dev/gvisor/pkg/tcpip"
"gvisor.dev/gvisor/pkg/tcpip/buffer"
"gvisor.dev/gvisor/pkg/tcpip/header"
"gvisor.dev/gvisor/pkg/tcpip/link/rawfile"
"gvisor.dev/gvisor/pkg/tcpip/stack"
)
// BufConfig defines the shape of the vectorised view used to read packets from the NIC.
var BufConfig = []int{128, 256, 256, 512, 1024, 2048, 4096, 8192, 16384, 32768}
type iovecBuffer struct {
// views are the actual buffers that hold the packet contents.
views []buffer.View
// iovecs are initialized with base pointers/len of the corresponding
// entries in the views defined above, except when GSO is enabled
// (skipsVnetHdr) then the first iovec points to a buffer for the vnet header
// which is stripped before the views are passed up the stack for further
// processing.
iovecs []unix.Iovec
// sizes is an array of buffer sizes for the underlying views. sizes is
// immutable.
sizes []int
// skipsVnetHdr is true if virtioNetHdr is to skipped.
skipsVnetHdr bool
}
func newIovecBuffer(sizes []int, skipsVnetHdr bool) *iovecBuffer {
b := &iovecBuffer{
views: make([]buffer.View, len(sizes)),
sizes: sizes,
skipsVnetHdr: skipsVnetHdr,
}
niov := len(b.views)
if b.skipsVnetHdr {
niov++
}
b.iovecs = make([]unix.Iovec, niov)
return b
}
func (b *iovecBuffer) nextIovecs() []unix.Iovec {
vnetHdrOff := 0
if b.skipsVnetHdr {
var vnetHdr [virtioNetHdrSize]byte
// The kernel adds virtioNetHdr before each packet, but
// we don't use it, so so we allocate a buffer for it,
// add it in iovecs but don't add it in a view.
b.iovecs[0] = unix.Iovec{Base: &vnetHdr[0]}
b.iovecs[0].SetLen(virtioNetHdrSize)
vnetHdrOff++
}
for i := range b.views {
if b.views[i] != nil {
break
}
v := buffer.NewView(b.sizes[i])
b.views[i] = v
b.iovecs[i+vnetHdrOff] = unix.Iovec{Base: &v[0]}
b.iovecs[i+vnetHdrOff].SetLen(len(v))
}
return b.iovecs
}
func (b *iovecBuffer) pullViews(n int) buffer.VectorisedView {
var views []buffer.View
c := 0
if b.skipsVnetHdr {
c += virtioNetHdrSize
if c >= n {
// Nothing in the packet.
return buffer.NewVectorisedView(0, nil)
}
}
for i, v := range b.views {
c += len(v)
if c >= n {
b.views[i].CapLength(len(v) - (c - n))
views = append([]buffer.View(nil), b.views[:i+1]...)
break
}
}
// Remove the first len(views) used views from the state.
for i := range views {
b.views[i] = nil
}
if b.skipsVnetHdr {
// Exclude the size of the vnet header.
n -= virtioNetHdrSize
}
return buffer.NewVectorisedView(n, views)
}
// stopFd is an eventfd used to signal the stop of a dispatcher.
type stopFd struct {
efd int
}
func newStopFd() (stopFd, error) {
efd, err := unix.Eventfd(0, unix.EFD_NONBLOCK)
if err != nil {
return stopFd{efd: -1}, fmt.Errorf("failed to create eventfd: %w", err)
}
return stopFd{efd: efd}, nil
}
// stop writes to the eventfd and notifies the dispatcher to stop. It does not
// block.
func (s *stopFd) stop() {
increment := []byte{1, 0, 0, 0, 0, 0, 0, 0}
if n, err := unix.Write(s.efd, increment); n != len(increment) || err != nil {
// There are two possible errors documented in eventfd(2) for writing:
// 1. We are writing 8 bytes and not 0xffffffffffffff, thus no EINVAL.
// 2. stop is only supposed to be called once, it can't reach the limit,
// thus no EAGAIN.
panic(fmt.Sprintf("write(efd) = (%d, %s), want (%d, nil)", n, err, len(increment)))
}
}
// readVDispatcher uses readv() system call to read inbound packets and
// dispatches them.
type readVDispatcher struct {
stopFd
// fd is the file descriptor used to send and receive packets.
fd int
// e is the endpoint this dispatcher is attached to.
e *endpoint
// buf is the iovec buffer that contains the packet contents.
buf *iovecBuffer
}
func newReadVDispatcher(fd int, e *endpoint) (linkDispatcher, error) {
stopFd, err := newStopFd()
if err != nil {
return nil, err
}
d := &readVDispatcher{
stopFd: stopFd,
fd: fd,
e: e,
}
skipsVnetHdr := d.e.gsoKind == stack.HWGSOSupported
d.buf = newIovecBuffer(BufConfig, skipsVnetHdr)
return d, nil
}
// dispatch reads one packet from the file descriptor and dispatches it.
func (d *readVDispatcher) dispatch() (bool, tcpip.Error) {
n, err := rawfile.BlockingReadvUntilStopped(d.efd, d.fd, d.buf.nextIovecs())
if n <= 0 || err != nil {
return false, err
}
pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{
Data: d.buf.pullViews(n),
})
var (
p tcpip.NetworkProtocolNumber
remote, local tcpip.LinkAddress
)
if d.e.hdrSize > 0 {
hdr, ok := pkt.LinkHeader().Consume(d.e.hdrSize)
if !ok {
return false, nil
}
eth := header.Ethernet(hdr)
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.
// IP version information is at the first octet, so pulling up 1 byte.
h, ok := pkt.Data().PullUp(1)
if !ok {
return true, nil
}
switch header.IPVersion(h) {
case header.IPv4Version:
p = header.IPv4ProtocolNumber
case header.IPv6Version:
p = header.IPv6ProtocolNumber
default:
return true, nil
}
}
d.e.dispatcher.DeliverNetworkPacket(remote, local, p, pkt)
return true, nil
}
// recvMMsgDispatcher uses the recvmmsg system call to read inbound packets and
// dispatches them.
type recvMMsgDispatcher struct {
stopFd
// fd is the file descriptor used to send and receive packets.
fd int
// e is the endpoint this dispatcher is attached to.
e *endpoint
// bufs is an array of iovec buffers that contain packet contents.
bufs []*iovecBuffer
// msgHdrs is an array of MMsgHdr objects where each MMsghdr is used to
// reference an array of iovecs in the iovecs field defined above. This
// array is passed as the parameter to recvmmsg call to retrieve
// potentially more than 1 packet per unix.
msgHdrs []rawfile.MMsgHdr
}
const (
// MaxMsgsPerRecv is the maximum number of packets we want to retrieve
// in a single RecvMMsg call.
MaxMsgsPerRecv = 8
)
func newRecvMMsgDispatcher(fd int, e *endpoint) (linkDispatcher, error) {
stopFd, err := newStopFd()
if err != nil {
return nil, err
}
d := &recvMMsgDispatcher{
stopFd: stopFd,
fd: fd,
e: e,
bufs: make([]*iovecBuffer, MaxMsgsPerRecv),
msgHdrs: make([]rawfile.MMsgHdr, MaxMsgsPerRecv),
}
skipsVnetHdr := d.e.gsoKind == stack.HWGSOSupported
for i := range d.bufs {
d.bufs[i] = newIovecBuffer(BufConfig, skipsVnetHdr)
}
return d, nil
}
// recvMMsgDispatch reads more than one packet at a time from the file
// descriptor and dispatches it.
func (d *recvMMsgDispatcher) dispatch() (bool, tcpip.Error) {
// Fill message headers.
for k := range d.msgHdrs {
if d.msgHdrs[k].Msg.Iovlen > 0 {
break
}
iovecs := d.bufs[k].nextIovecs()
iovLen := len(iovecs)
d.msgHdrs[k].Len = 0
d.msgHdrs[k].Msg.Iov = &iovecs[0]
d.msgHdrs[k].Msg.SetIovlen(iovLen)
}
nMsgs, err := rawfile.BlockingRecvMMsgUntilStopped(d.efd, d.fd, d.msgHdrs)
if nMsgs == -1 || err != nil {
return false, err
}
// Process each of received packets.
for k := 0; k < nMsgs; k++ {
n := int(d.msgHdrs[k].Len)
pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{
Data: d.bufs[k].pullViews(n),
})
// Mark that this iovec has been processed.
d.msgHdrs[k].Msg.Iovlen = 0
var (
p tcpip.NetworkProtocolNumber
remote, local tcpip.LinkAddress
)
if d.e.hdrSize > 0 {
hdr, ok := pkt.LinkHeader().Consume(d.e.hdrSize)
if !ok {
return false, nil
}
eth := header.Ethernet(hdr)
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.
// IP version information is at the first octet, so pulling up 1 byte.
h, ok := pkt.Data().PullUp(1)
if !ok {
// Skip this packet.
continue
}
switch header.IPVersion(h) {
case header.IPv4Version:
p = header.IPv4ProtocolNumber
case header.IPv6Version:
p = header.IPv6ProtocolNumber
default:
// Skip this packet.
continue
}
}
d.e.dispatcher.DeliverNetworkPacket(remote, local, p, pkt)
}
return true, nil
}
|