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
|
// Copyright 2018 The containerd Authors.
// Copyright 2019 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
//
// https://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 main
import (
"bytes"
"context"
"flag"
"fmt"
"net"
"os"
"os/exec"
"os/signal"
"path/filepath"
"runtime"
"runtime/debug"
"strings"
"sync"
"syscall"
"time"
"github.com/containerd/containerd/events"
"github.com/containerd/containerd/namespaces"
"github.com/containerd/containerd/runtime/v1/linux/proc"
containerdshim "github.com/containerd/containerd/runtime/v1/shim"
shimapi "github.com/containerd/containerd/runtime/v1/shim/v1"
"github.com/containerd/ttrpc"
"github.com/containerd/typeurl"
ptypes "github.com/gogo/protobuf/types"
"github.com/opencontainers/runc/libcontainer/system"
"github.com/pkg/errors"
"github.com/sirupsen/logrus"
"golang.org/x/sys/unix"
"gvisor.dev/gvisor/pkg/shim/runsc"
"gvisor.dev/gvisor/pkg/shim/v1/shim"
)
var (
debugFlag bool
namespaceFlag string
socketFlag string
addressFlag string
workdirFlag string
runtimeRootFlag string
containerdBinaryFlag string
shimConfigFlag string
)
// ShimConfigPath is the default shim config file path.
const ShimConfigPath = "/etc/containerd/gvisor-containerd-shim.toml"
func init() {
flag.BoolVar(&debugFlag, "debug", false, "enable debug output in logs")
flag.StringVar(&namespaceFlag, "namespace", "", "namespace that owns the shim")
flag.StringVar(&socketFlag, "socket", "", "abstract socket path to serve")
flag.StringVar(&addressFlag, "address", "", "grpc address back to main containerd")
flag.StringVar(&workdirFlag, "workdir", "", "path used to storge large temporary data")
// Containerd default to runc, unless another runtime is explicitly specified.
// We keep the same default to make the default behavior consistent.
flag.StringVar(&runtimeRootFlag, "runtime-root", proc.RuncRoot, "root directory for the runtime")
// currently, the `containerd publish` utility is embedded in the daemon binary.
// The daemon invokes `containerd-shim -containerd-binary ...` with its own os.Executable() path.
flag.StringVar(&containerdBinaryFlag, "containerd-binary", "containerd", "path to containerd binary (used for `containerd publish`)")
flag.StringVar(&shimConfigFlag, "config", ShimConfigPath, "path to the shim configuration file")
flag.Parse()
}
func main() {
// This is a hack. Exec current process to run standard containerd-shim
// if runtime root is not `runsc`. We don't need this for shim v2 api.
if filepath.Base(runtimeRootFlag) != "runsc" {
if err := executeRuncShim(); err != nil {
fmt.Fprintf(os.Stderr, "gvisor-containerd-shim: %s\n", err)
os.Exit(1)
}
}
debug.SetGCPercent(40)
go func() {
for range time.Tick(30 * time.Second) {
debug.FreeOSMemory()
}
}()
if debugFlag {
logrus.SetLevel(logrus.DebugLevel)
}
if os.Getenv("GOMAXPROCS") == "" {
// If GOMAXPROCS hasn't been set, we default to a value of 2 to reduce
// the number of Go stacks present in the shim.
runtime.GOMAXPROCS(2)
}
// Run regular shim if needed.
if err := executeShim(); err != nil {
fmt.Fprintf(os.Stderr, "gvisor-containerd-shim: %s\n", err)
os.Exit(1)
}
}
// executeRuncShim execs current process to a containerd-shim process and
// retains all flags and envs.
func executeRuncShim() error {
c, err := loadConfig(shimConfigFlag)
if err != nil && !os.IsNotExist(err) {
return errors.Wrap(err, "failed to load shim config")
}
shimPath := c.RuncShim
if shimPath == "" {
shimPath, err = exec.LookPath("containerd-shim")
if err != nil {
return errors.Wrapf(err, "lookup containerd-shim")
}
}
args := append([]string{shimPath}, os.Args[1:]...)
if err := syscall.Exec(shimPath, args, os.Environ()); err != nil {
return errors.Wrapf(err, "exec containerd-shim %q", shimPath)
}
return nil
}
func executeShim() error {
// start handling signals as soon as possible so that things are properly reaped
// or if runtime exits before we hit the handler
signals, err := setupSignals()
if err != nil {
return err
}
dump := make(chan os.Signal, 32)
signal.Notify(dump, syscall.SIGUSR1)
path, err := os.Getwd()
if err != nil {
return err
}
server, err := ttrpc.NewServer(ttrpc.WithServerHandshaker(ttrpc.UnixSocketRequireSameUser()))
if err != nil {
return errors.Wrap(err, "failed creating server")
}
c, err := loadConfig(shimConfigFlag)
if err != nil && !os.IsNotExist(err) {
return errors.Wrap(err, "failed to load shim config")
}
sv, err := shim.NewService(
shim.Config{
Path: path,
Namespace: namespaceFlag,
WorkDir: workdirFlag,
RuntimeRoot: runtimeRootFlag,
RunscConfig: c.RunscConfig,
},
&remoteEventsPublisher{address: addressFlag},
)
if err != nil {
return err
}
logrus.Debug("registering ttrpc server")
shimapi.RegisterShimService(server, sv)
socket := socketFlag
if err := serve(server, socket); err != nil {
return err
}
logger := logrus.WithFields(logrus.Fields{
"pid": os.Getpid(),
"path": path,
"namespace": namespaceFlag,
})
go func() {
for range dump {
dumpStacks(logger)
}
}()
return handleSignals(logger, signals, server, sv)
}
// serve serves the ttrpc API over a unix socket at the provided path
// this function does not block
func serve(server *ttrpc.Server, path string) error {
var (
l net.Listener
err error
)
if path == "" {
l, err = net.FileListener(os.NewFile(3, "socket"))
path = "[inherited from parent]"
} else {
if len(path) > 106 {
return errors.Errorf("%q: unix socket path too long (> 106)", path)
}
l, err = net.Listen("unix", "\x00"+path)
}
if err != nil {
return err
}
logrus.WithField("socket", path).Debug("serving api on unix socket")
go func() {
defer l.Close()
if err := server.Serve(context.Background(), l); err != nil &&
!strings.Contains(err.Error(), "use of closed network connection") {
logrus.WithError(err).Fatal("gvisor-containerd-shim: ttrpc server failure")
}
}()
return nil
}
// setupSignals creates a new signal handler for all signals and sets the shim as a
// sub-reaper so that the container processes are reparented
func setupSignals() (chan os.Signal, error) {
signals := make(chan os.Signal, 32)
signal.Notify(signals, unix.SIGTERM, unix.SIGINT, unix.SIGCHLD, unix.SIGPIPE)
// make sure runc is setup to use the monitor
// for waiting on processes
// TODO(random-liu): Move shim/reaper.go to a separate package.
runsc.Monitor = containerdshim.Default
// set the shim as the subreaper for all orphaned processes created by the container
if err := system.SetSubreaper(1); err != nil {
return nil, err
}
return signals, nil
}
func handleSignals(logger *logrus.Entry, signals chan os.Signal, server *ttrpc.Server, sv *shim.Service) error {
var (
termOnce sync.Once
done = make(chan struct{})
)
for {
select {
case <-done:
return nil
case s := <-signals:
switch s {
case unix.SIGCHLD:
if err := containerdshim.Reap(); err != nil {
logger.WithError(err).Error("reap exit status")
}
case unix.SIGTERM, unix.SIGINT:
go termOnce.Do(func() {
ctx := context.TODO()
if err := server.Shutdown(ctx); err != nil {
logger.WithError(err).Error("failed to shutdown server")
}
// Ensure our child is dead if any
sv.Kill(ctx, &shimapi.KillRequest{
Signal: uint32(syscall.SIGKILL),
All: true,
})
sv.Delete(context.Background(), &ptypes.Empty{})
close(done)
})
case unix.SIGPIPE:
}
}
}
}
func dumpStacks(logger *logrus.Entry) {
var (
buf []byte
stackSize int
)
bufferLen := 16384
for stackSize == len(buf) {
buf = make([]byte, bufferLen)
stackSize = runtime.Stack(buf, true)
bufferLen *= 2
}
buf = buf[:stackSize]
logger.Infof("=== BEGIN goroutine stack dump ===\n%s\n=== END goroutine stack dump ===", buf)
}
type remoteEventsPublisher struct {
address string
}
func (l *remoteEventsPublisher) Publish(ctx context.Context, topic string, event events.Event) error {
ns, _ := namespaces.Namespace(ctx)
encoded, err := typeurl.MarshalAny(event)
if err != nil {
return err
}
data, err := encoded.Marshal()
if err != nil {
return err
}
cmd := exec.CommandContext(ctx, containerdBinaryFlag, "--address", l.address, "publish", "--topic", topic, "--namespace", ns)
cmd.Stdin = bytes.NewReader(data)
c, err := containerdshim.Default.Start(cmd)
if err != nil {
return err
}
status, err := containerdshim.Default.Wait(cmd, c)
if err != nil {
return err
}
if status != 0 {
return errors.New("failed to publish event")
}
return nil
}
|