summaryrefslogtreecommitdiff
path: root/libpod/networking_linux.go
blob: 5db749a4db5bdb68a76a10d15904b7b13d8d3a16 (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
// +build linux

package libpod

import (
	"bytes"
	"crypto/rand"
	"fmt"
	"io"
	"io/ioutil"
	"net"
	"os"
	"os/exec"
	"path/filepath"
	"strconv"
	"strings"
	"syscall"
	"time"

	cnitypes "github.com/containernetworking/cni/pkg/types/current"
	"github.com/containernetworking/plugins/pkg/ns"
	"github.com/containers/libpod/pkg/errorhandling"
	"github.com/containers/libpod/pkg/netns"
	"github.com/containers/libpod/pkg/rootless"
	"github.com/containers/libpod/pkg/rootlessport"
	"github.com/cri-o/ocicni/pkg/ocicni"
	"github.com/pkg/errors"
	"github.com/sirupsen/logrus"
	"github.com/vishvananda/netlink"
	"golang.org/x/sys/unix"
)

// Get an OCICNI network config
func (r *Runtime) getPodNetwork(id, name, nsPath string, networks []string, ports []ocicni.PortMapping, staticIP net.IP, staticMAC net.HardwareAddr) ocicni.PodNetwork {
	var networkKey string
	if len(networks) > 0 {
		// This is inconsistent for >1 network, but it's probably the
		// best we can do.
		networkKey = networks[0]
	} else {
		networkKey = r.netPlugin.GetDefaultNetworkName()
	}
	network := ocicni.PodNetwork{
		Name:      name,
		Namespace: name, // TODO is there something else we should put here? We don't know about Kube namespaces
		ID:        id,
		NetNS:     nsPath,
		RuntimeConfig: map[string]ocicni.RuntimeConfig{
			networkKey: {PortMappings: ports},
		},
	}

	// If we have extra networks, add them
	if len(networks) > 0 {
		network.Networks = make([]ocicni.NetAttachment, len(networks))
		for i, netName := range networks {
			network.Networks[i].Name = netName
		}
	}

	if staticIP != nil || staticMAC != nil {
		// For static IP or MAC, we need to populate networks even if
		// it's just the default.
		if len(networks) == 0 {
			// If len(networks) == 0 this is guaranteed to be the
			// default network.
			network.Networks = []ocicni.NetAttachment{{Name: networkKey}}
		}
		var rt ocicni.RuntimeConfig = ocicni.RuntimeConfig{PortMappings: ports}
		if staticIP != nil {
			rt.IP = staticIP.String()
		}
		if staticMAC != nil {
			rt.MAC = staticMAC.String()
		}
		network.RuntimeConfig = map[string]ocicni.RuntimeConfig{
			networkKey: rt,
		}
	}

	return network
}

// Create and configure a new network namespace for a container
func (r *Runtime) configureNetNS(ctr *Container, ctrNS ns.NetNS) ([]*cnitypes.Result, error) {
	var requestedIP net.IP
	if ctr.requestedIP != nil {
		requestedIP = ctr.requestedIP
		// cancel request for a specific IP in case the container is reused later
		ctr.requestedIP = nil
	} else {
		requestedIP = ctr.config.StaticIP
	}

	var requestedMAC net.HardwareAddr
	if ctr.requestedMAC != nil {
		requestedMAC = ctr.requestedMAC
		// cancel request for a specific MAC in case the container is reused later
		ctr.requestedMAC = nil
	} else {
		requestedMAC = ctr.config.StaticMAC
	}

	podNetwork := r.getPodNetwork(ctr.ID(), ctr.Name(), ctrNS.Path(), ctr.config.Networks, ctr.config.PortMappings, requestedIP, requestedMAC)

	results, err := r.netPlugin.SetUpPod(podNetwork)
	if err != nil {
		return nil, errors.Wrapf(err, "error configuring network namespace for container %s", ctr.ID())
	}
	defer func() {
		if err != nil {
			if err2 := r.netPlugin.TearDownPod(podNetwork); err2 != nil {
				logrus.Errorf("Error tearing down partially created network namespace for container %s: %v", ctr.ID(), err2)
			}
		}
	}()

	networkStatus := make([]*cnitypes.Result, 0)
	for idx, r := range results {
		logrus.Debugf("[%d] CNI result: %v", idx, r.Result.String())
		resultCurrent, err := cnitypes.GetResult(r.Result)
		if err != nil {
			return nil, errors.Wrapf(err, "error parsing CNI plugin result %q: %v", r.Result.String(), err)
		}
		networkStatus = append(networkStatus, resultCurrent)
	}

	return networkStatus, nil
}

// Create and configure a new network namespace for a container
func (r *Runtime) createNetNS(ctr *Container) (n ns.NetNS, q []*cnitypes.Result, err error) {
	ctrNS, err := netns.NewNS()
	if err != nil {
		return nil, nil, errors.Wrapf(err, "error creating network namespace for container %s", ctr.ID())
	}
	defer func() {
		if err != nil {
			if err2 := netns.UnmountNS(ctrNS); err2 != nil {
				logrus.Errorf("Error unmounting partially created network namespace for container %s: %v", ctr.ID(), err2)
			}
			if err2 := ctrNS.Close(); err2 != nil {
				logrus.Errorf("Error closing partially created network namespace for container %s: %v", ctr.ID(), err2)
			}
		}
	}()

	logrus.Debugf("Made network namespace at %s for container %s", ctrNS.Path(), ctr.ID())

	networkStatus := []*cnitypes.Result{}
	if !rootless.IsRootless() {
		networkStatus, err = r.configureNetNS(ctr, ctrNS)
	}
	return ctrNS, networkStatus, err
}

func checkSlirpFlags(path string) (bool, bool, bool, error) {
	cmd := exec.Command(path, "--help")
	out, err := cmd.CombinedOutput()
	if err != nil {
		return false, false, false, errors.Wrapf(err, "slirp4netns %q", out)
	}
	return strings.Contains(string(out), "--disable-host-loopback"), strings.Contains(string(out), "--mtu"), strings.Contains(string(out), "--enable-sandbox"), nil
}

// Configure the network namespace for a rootless container
func (r *Runtime) setupRootlessNetNS(ctr *Container) (err error) {
	path := r.config.NetworkCmdPath

	if path == "" {
		var err error
		path, err = exec.LookPath("slirp4netns")
		if err != nil {
			logrus.Errorf("could not find slirp4netns, the network namespace won't be configured: %v", err)
			return nil
		}
	}

	syncR, syncW, err := os.Pipe()
	if err != nil {
		return errors.Wrapf(err, "failed to open pipe")
	}
	defer errorhandling.CloseQuiet(syncR)
	defer errorhandling.CloseQuiet(syncW)

	havePortMapping := len(ctr.Config().PortMappings) > 0
	logPath := filepath.Join(ctr.runtime.config.TmpDir, fmt.Sprintf("slirp4netns-%s.log", ctr.config.ID))

	cmdArgs := []string{}
	dhp, mtu, sandbox, err := checkSlirpFlags(path)
	if err != nil {
		return errors.Wrapf(err, "error checking slirp4netns binary %s: %q", path, err)
	}
	if dhp {
		cmdArgs = append(cmdArgs, "--disable-host-loopback")
	}
	if mtu {
		cmdArgs = append(cmdArgs, "--mtu", "65520")
	}
	if sandbox {
		cmdArgs = append(cmdArgs, "--enable-sandbox")
	}

	// the slirp4netns arguments being passed are describes as follows:
	// from the slirp4netns documentation: https://github.com/rootless-containers/slirp4netns
	// -c, --configure Brings up the tap interface
	// -e, --exit-fd=FD specify the FD for terminating slirp4netns
	// -r, --ready-fd=FD specify the FD to write to when the initialization steps are finished
	cmdArgs = append(cmdArgs, "-c", "-e", "3", "-r", "4")
	netnsPath := ""
	if !ctr.config.PostConfigureNetNS {
		ctr.rootlessSlirpSyncR, ctr.rootlessSlirpSyncW, err = os.Pipe()
		if err != nil {
			return errors.Wrapf(err, "failed to create rootless network sync pipe")
		}
		netnsPath = ctr.state.NetNS.Path()
		cmdArgs = append(cmdArgs, "--netns-type=path", netnsPath, "tap0")
	} else {
		defer errorhandling.CloseQuiet(ctr.rootlessSlirpSyncR)
		defer errorhandling.CloseQuiet(ctr.rootlessSlirpSyncW)
		netnsPath = fmt.Sprintf("/proc/%d/ns/net", ctr.state.PID)
		// we don't use --netns-path here (unavailable for slirp4netns < v0.4)
		cmdArgs = append(cmdArgs, fmt.Sprintf("%d", ctr.state.PID), "tap0")
	}

	cmd := exec.Command(path, cmdArgs...)
	logrus.Debugf("slirp4netns command: %s", strings.Join(cmd.Args, " "))
	cmd.SysProcAttr = &syscall.SysProcAttr{
		Setpgid: true,
	}

	// workaround for https://github.com/rootless-containers/slirp4netns/pull/153
	if sandbox {
		cmd.SysProcAttr.Cloneflags = syscall.CLONE_NEWNS
		cmd.SysProcAttr.Unshareflags = syscall.CLONE_NEWNS
	}

	// Leak one end of the pipe in slirp4netns, the other will be sent to conmon
	cmd.ExtraFiles = append(cmd.ExtraFiles, ctr.rootlessSlirpSyncR, syncW)

	logFile, err := os.Create(logPath)
	if err != nil {
		return errors.Wrapf(err, "failed to open slirp4netns log file %s", logPath)
	}
	defer logFile.Close()
	// Unlink immediately the file so we won't need to worry about cleaning it up later.
	// It is still accessible through the open fd logFile.
	if err := os.Remove(logPath); err != nil {
		return errors.Wrapf(err, "delete file %s", logPath)
	}
	cmd.Stdout = logFile
	cmd.Stderr = logFile
	if err := cmd.Start(); err != nil {
		return errors.Wrapf(err, "failed to start slirp4netns process")
	}
	defer func() {
		if err := cmd.Process.Release(); err != nil {
			logrus.Errorf("unable to release command process: %q", err)
		}
	}()

	if err := waitForSync(syncR, cmd, logFile, 1*time.Second); err != nil {
		return err
	}

	if havePortMapping {
		return r.setupRootlessPortMapping(ctr, netnsPath)
	}
	return nil
}

func waitForSync(syncR *os.File, cmd *exec.Cmd, logFile io.ReadSeeker, timeout time.Duration) error {
	prog := filepath.Base(cmd.Path)
	if len(cmd.Args) > 0 {
		prog = cmd.Args[0]
	}
	b := make([]byte, 16)
	for {
		if err := syncR.SetDeadline(time.Now().Add(timeout)); err != nil {
			return errors.Wrapf(err, "error setting %s pipe timeout", prog)
		}
		// FIXME: return err as soon as proc exits, without waiting for timeout
		if _, err := syncR.Read(b); err == nil {
			break
		} else {
			if os.IsTimeout(err) {
				// Check if the process is still running.
				var status syscall.WaitStatus
				pid, err := syscall.Wait4(cmd.Process.Pid, &status, syscall.WNOHANG, nil)
				if err != nil {
					return errors.Wrapf(err, "failed to read %s process status", prog)
				}
				if pid != cmd.Process.Pid {
					continue
				}
				if status.Exited() {
					// Seek at the beginning of the file and read all its content
					if _, err := logFile.Seek(0, 0); err != nil {
						logrus.Errorf("could not seek log file: %q", err)
					}
					logContent, err := ioutil.ReadAll(logFile)
					if err != nil {
						return errors.Wrapf(err, "%s failed", prog)
					}
					return errors.Errorf("%s failed: %q", prog, logContent)
				}
				if status.Signaled() {
					return errors.Errorf("%s killed by signal", prog)
				}
				continue
			}
			return errors.Wrapf(err, "failed to read from %s sync pipe", prog)
		}
	}
	return nil
}

func (r *Runtime) setupRootlessPortMapping(ctr *Container, netnsPath string) (err error) {
	syncR, syncW, err := os.Pipe()
	if err != nil {
		return errors.Wrapf(err, "failed to open pipe")
	}
	defer errorhandling.CloseQuiet(syncR)
	defer errorhandling.CloseQuiet(syncW)

	logPath := filepath.Join(ctr.runtime.config.TmpDir, fmt.Sprintf("rootlessport-%s.log", ctr.config.ID))
	logFile, err := os.Create(logPath)
	if err != nil {
		return errors.Wrapf(err, "failed to open rootlessport log file %s", logPath)
	}
	defer logFile.Close()
	// Unlink immediately the file so we won't need to worry about cleaning it up later.
	// It is still accessible through the open fd logFile.
	if err := os.Remove(logPath); err != nil {
		return errors.Wrapf(err, "delete file %s", logPath)
	}

	ctr.rootlessPortSyncR, ctr.rootlessPortSyncW, err = os.Pipe()
	if err != nil {
		return errors.Wrapf(err, "failed to create rootless port sync pipe")
	}
	cfg := rootlessport.Config{
		Mappings:  ctr.config.PortMappings,
		NetNSPath: netnsPath,
		ExitFD:    3,
		ReadyFD:   4,
		TmpDir:    ctr.runtime.config.TmpDir,
	}
	cfgJSON, err := json.Marshal(cfg)
	if err != nil {
		return err
	}
	cfgR := bytes.NewReader(cfgJSON)
	var stdout bytes.Buffer
	cmd := exec.Command(fmt.Sprintf("/proc/%d/exe", os.Getpid()))
	cmd.Args = []string{rootlessport.ReexecKey}
	// Leak one end of the pipe in rootlessport process, the other will be sent to conmon
	cmd.ExtraFiles = append(cmd.ExtraFiles, ctr.rootlessPortSyncR, syncW)
	cmd.Stdin = cfgR
	// stdout is for human-readable error, stderr is for debug log
	cmd.Stdout = &stdout
	cmd.Stderr = io.MultiWriter(logFile, &logrusDebugWriter{"rootlessport: "})
	cmd.SysProcAttr = &syscall.SysProcAttr{
		Setpgid: true,
	}
	if err := cmd.Start(); err != nil {
		return errors.Wrapf(err, "failed to start rootlessport process")
	}
	defer func() {
		if err := cmd.Process.Release(); err != nil {
			logrus.Errorf("unable to release rootlessport process: %q", err)
		}
	}()
	if err := waitForSync(syncR, cmd, logFile, 3*time.Second); err != nil {
		stdoutStr := stdout.String()
		if stdoutStr != "" {
			// err contains full debug log and too verbose, so return stdoutStr
			logrus.Debug(err)
			return errors.Errorf("failed to expose ports via rootlessport: %q", stdoutStr)
		}
		return err
	}
	logrus.Debug("rootlessport is ready")
	return nil
}

// Configure the network namespace using the container process
func (r *Runtime) setupNetNS(ctr *Container) (err error) {
	nsProcess := fmt.Sprintf("/proc/%d/ns/net", ctr.state.PID)

	b := make([]byte, 16)

	if _, err := rand.Reader.Read(b); err != nil {
		return errors.Wrapf(err, "failed to generate random netns name")
	}

	nsPath := fmt.Sprintf("/var/run/netns/cni-%x-%x-%x-%x-%x", b[0:4], b[4:6], b[6:8], b[8:10], b[10:])

	if err := os.MkdirAll(filepath.Dir(nsPath), 0711); err != nil {
		return errors.Wrapf(err, "cannot create %s", filepath.Dir(nsPath))
	}

	mountPointFd, err := os.Create(nsPath)
	if err != nil {
		return errors.Wrapf(err, "cannot open %s", nsPath)
	}
	if err := mountPointFd.Close(); err != nil {
		return err
	}

	if err := unix.Mount(nsProcess, nsPath, "none", unix.MS_BIND, ""); err != nil {
		return errors.Wrapf(err, "cannot mount %s", nsPath)
	}

	netNS, err := ns.GetNS(nsPath)
	if err != nil {
		return err
	}
	networkStatus, err := r.configureNetNS(ctr, netNS)

	// Assign NetNS attributes to container
	ctr.state.NetNS = netNS
	ctr.state.NetworkStatus = networkStatus
	return err
}

// Join an existing network namespace
func joinNetNS(path string) (ns.NetNS, error) {
	netNS, err := ns.GetNS(path)
	if err != nil {
		return nil, errors.Wrapf(err, "error retrieving network namespace at %s", path)
	}

	return netNS, nil
}

// Close a network namespace.
// Differs from teardownNetNS() in that it will not attempt to undo the setup of
// the namespace, but will instead only close the open file descriptor
func (r *Runtime) closeNetNS(ctr *Container) error {
	if ctr.state.NetNS == nil {
		// The container has no network namespace, we're set
		return nil
	}

	if err := ctr.state.NetNS.Close(); err != nil {
		return errors.Wrapf(err, "error closing network namespace for container %s", ctr.ID())
	}

	ctr.state.NetNS = nil

	return nil
}

// Tear down a network namespace, undoing all state associated with it.
func (r *Runtime) teardownNetNS(ctr *Container) error {
	if ctr.state.NetNS == nil {
		// The container has no network namespace, we're set
		return nil
	}

	logrus.Debugf("Tearing down network namespace at %s for container %s", ctr.state.NetNS.Path(), ctr.ID())

	// rootless containers do not use the CNI plugin
	if !rootless.IsRootless() {
		var requestedIP net.IP
		if ctr.requestedIP != nil {
			requestedIP = ctr.requestedIP
			// cancel request for a specific IP in case the container is reused later
			ctr.requestedIP = nil
		} else {
			requestedIP = ctr.config.StaticIP
		}

		var requestedMAC net.HardwareAddr
		if ctr.requestedMAC != nil {
			requestedMAC = ctr.requestedMAC
			// cancel request for a specific MAC in case the container is reused later
			ctr.requestedMAC = nil
		} else {
			requestedMAC = ctr.config.StaticMAC
		}

		podNetwork := r.getPodNetwork(ctr.ID(), ctr.Name(), ctr.state.NetNS.Path(), ctr.config.Networks, ctr.config.PortMappings, requestedIP, requestedMAC)

		if err := r.netPlugin.TearDownPod(podNetwork); err != nil {
			return errors.Wrapf(err, "error tearing down CNI namespace configuration for container %s", ctr.ID())
		}
	}

	// First unmount the namespace
	if err := netns.UnmountNS(ctr.state.NetNS); err != nil {
		return errors.Wrapf(err, "error unmounting network namespace for container %s", ctr.ID())
	}

	// Now close the open file descriptor
	if err := ctr.state.NetNS.Close(); err != nil {
		return errors.Wrapf(err, "error closing network namespace for container %s", ctr.ID())
	}

	ctr.state.NetNS = nil

	return nil
}

func getContainerNetNS(ctr *Container) (string, error) {
	if ctr.state.NetNS != nil {
		return ctr.state.NetNS.Path(), nil
	}
	if ctr.config.NetNsCtr != "" {
		c, err := ctr.runtime.GetContainer(ctr.config.NetNsCtr)
		if err != nil {
			return "", err
		}
		if err = c.syncContainer(); err != nil {
			return "", err
		}
		return c.state.NetNS.Path(), nil
	}
	return "", nil
}

func getContainerNetIO(ctr *Container) (*netlink.LinkStatistics, error) {
	var netStats *netlink.LinkStatistics
	// rootless v2 cannot seem to resolve its network connection to
	// collect statistics.  For now, we allow stats to at least run
	// by returning nil
	if rootless.IsRootless() {
		return netStats, nil
	}
	netNSPath, netPathErr := getContainerNetNS(ctr)
	if netPathErr != nil {
		return nil, netPathErr
	}
	if netNSPath == "" {
		// If netNSPath is empty, it was set as none, and no netNS was set up
		// this is a valid state and thus return no error, nor any statistics
		return nil, nil
	}
	err := ns.WithNetNSPath(netNSPath, func(_ ns.NetNS) error {
		link, err := netlink.LinkByName(ocicni.DefaultInterfaceName)
		if err != nil {
			return err
		}
		netStats = link.Attrs().Statistics
		return nil
	})
	return netStats, err
}

func (c *Container) getContainerNetworkInfo(data *InspectContainerData) *InspectContainerData {
	if c.state.NetNS != nil && len(c.state.NetworkStatus) > 0 {
		// Report network settings from the first pod network
		result := c.state.NetworkStatus[0]
		// Go through our IP addresses
		for _, ctrIP := range result.IPs {
			ipWithMask := ctrIP.Address.String()
			splitIP := strings.Split(ipWithMask, "/")
			mask, _ := strconv.Atoi(splitIP[1])
			if ctrIP.Version == "4" {
				data.NetworkSettings.IPAddress = splitIP[0]
				data.NetworkSettings.IPPrefixLen = mask
				data.NetworkSettings.Gateway = ctrIP.Gateway.String()
			} else {
				data.NetworkSettings.GlobalIPv6Address = splitIP[0]
				data.NetworkSettings.GlobalIPv6PrefixLen = mask
				data.NetworkSettings.IPv6Gateway = ctrIP.Gateway.String()
			}
		}

		// Set network namespace path
		data.NetworkSettings.SandboxKey = c.state.NetNS.Path()

		// Set MAC address of interface linked with network namespace path
		for _, i := range result.Interfaces {
			if i.Sandbox == data.NetworkSettings.SandboxKey {
				data.NetworkSettings.MacAddress = i.Mac
			}
		}
	}
	return data
}

type logrusDebugWriter struct {
	prefix string
}

func (w *logrusDebugWriter) Write(p []byte) (int, error) {
	logrus.Debugf("%s%s", w.prefix, string(p))
	return len(p), nil
}