summaryrefslogtreecommitdiffhomepage
path: root/runsc/sandbox/sandbox.go
blob: 07a6bf388a5f22356472cfcc81dbcdb4385b5a8f (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
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
// Copyright 2018 Google Inc.
//
// 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 sandbox creates and manipulates sandboxes.
package sandbox

import (
	"context"
	"fmt"
	"os"
	"os/exec"
	"strconv"
	"syscall"
	"time"

	"github.com/cenkalti/backoff"
	specs "github.com/opencontainers/runtime-spec/specs-go"
	"gvisor.googlesource.com/gvisor/pkg/control/client"
	"gvisor.googlesource.com/gvisor/pkg/control/server"
	"gvisor.googlesource.com/gvisor/pkg/log"
	"gvisor.googlesource.com/gvisor/pkg/sentry/control"
	"gvisor.googlesource.com/gvisor/pkg/sentry/platform/kvm"
	"gvisor.googlesource.com/gvisor/pkg/urpc"
	"gvisor.googlesource.com/gvisor/runsc/boot"
	"gvisor.googlesource.com/gvisor/runsc/console"
	"gvisor.googlesource.com/gvisor/runsc/specutils"
)

// Sandbox wraps a sandbox process.
//
// It is used to start/stop sandbox process (and associated processes like
// gofers), as well as for running and manipulating containers inside a running
// sandbox.
//
// Note: Sandbox must be immutable because a copy of it is saved for each
// container and changes would not be synchronized to all of them.
type Sandbox struct {
	// ID is the id of the sandbox (immutable). By convention, this is the same
	// ID as the first container run in the sandbox.
	ID string `json:"id"`

	// Pid is the pid of the running sandbox (immutable). May be 0 is the sandbox
	// is not running.
	Pid int `json:"pid"`

	// Chroot is the path to the chroot directory that the sandbox process
	// is running in.
	Chroot string `json:"chroot"`
}

// Create creates the sandbox process. The caller must call Destroy() on the
// sandbox.
func Create(id string, spec *specs.Spec, conf *boot.Config, bundleDir, consoleSocket string, ioFiles []*os.File) (*Sandbox, error) {
	s := &Sandbox{ID: id}

	// Create the sandbox process.
	if err := s.createSandboxProcess(spec, conf, bundleDir, consoleSocket, ioFiles); err != nil {
		return nil, err
	}

	// Wait for the control server to come up (or timeout).
	if err := s.waitForCreated(10 * time.Second); err != nil {
		return nil, err
	}

	return s, nil
}

// StartRoot starts running the root container process inside the sandbox.
func (s *Sandbox) StartRoot(spec *specs.Spec, conf *boot.Config) error {
	log.Debugf("Start root sandbox %q, pid: %d", s.ID, s.Pid)
	conn, err := s.sandboxConnect()
	if err != nil {
		return err
	}
	defer conn.Close()

	// Configure the network.
	if err := setupNetwork(conn, s.Pid, spec, conf); err != nil {
		return fmt.Errorf("error setting up network: %v", err)
	}

	// Send a message to the sandbox control server to start the root
	// container.
	if err := conn.Call(boot.RootContainerStart, &s.ID, nil); err != nil {
		return fmt.Errorf("error starting root container %v: %v", spec.Process.Args, err)
	}

	return nil
}

// Start starts running a non-root container inside the sandbox.
func (s *Sandbox) Start(spec *specs.Spec, conf *boot.Config, cid string, goferFiles []*os.File) error {
	for _, f := range goferFiles {
		defer f.Close()
	}

	log.Debugf("Start non-root container sandbox %q, pid: %d", s.ID, s.Pid)
	sandboxConn, err := s.sandboxConnect()
	if err != nil {
		return fmt.Errorf("couldn't connect to sandbox: %v", err)
	}
	defer sandboxConn.Close()

	// The payload must container stdin/stdout/stderr followed by gofer
	// files.
	files := append([]*os.File{os.Stdin, os.Stdout, os.Stderr}, goferFiles...)
	// Start running the container.
	args := boot.StartArgs{
		Spec:        spec,
		Conf:        conf,
		CID:         cid,
		FilePayload: urpc.FilePayload{Files: files},
	}
	if err := sandboxConn.Call(boot.ContainerStart, &args, nil); err != nil {
		return fmt.Errorf("error starting non-root container %v: %v", spec.Process.Args, err)
	}
	return nil
}

// Restore sends the restore call for a container in the sandbox.
func (s *Sandbox) Restore(cid string, spec *specs.Spec, conf *boot.Config, f string) error {
	log.Debugf("Restore sandbox %q", s.ID)

	rf, err := os.Open(f)
	if err != nil {
		return fmt.Errorf("os.Open(%q) failed: %v", f, err)
	}
	defer rf.Close()

	opt := boot.RestoreOpts{
		FilePayload: urpc.FilePayload{
			Files: []*os.File{rf},
		},
		SandboxID: s.ID,
	}

	// If the platform needs a device fd we must pass it in.
	if deviceFile, err := deviceFileForPlatform(conf.Platform); err != nil {
		return err
	} else if deviceFile != nil {
		defer deviceFile.Close()
		opt.FilePayload.Files = append(opt.FilePayload.Files, deviceFile)
	}

	conn, err := s.sandboxConnect()
	if err != nil {
		return err
	}
	defer conn.Close()

	// Configure the network.
	if err := setupNetwork(conn, s.Pid, spec, conf); err != nil {
		return fmt.Errorf("error setting up network: %v", err)
	}

	// Restore the container and start the root container.
	if err := conn.Call(boot.ContainerRestore, &opt, nil); err != nil {
		return fmt.Errorf("error restoring container %q: %v", cid, err)
	}

	return nil
}

// Processes retrieves the list of processes and associated metadata for a
// given container in this sandbox.
func (s *Sandbox) Processes(cid string) ([]*control.Process, error) {
	log.Debugf("Getting processes for container %q in sandbox %q", cid, s.ID)
	conn, err := s.sandboxConnect()
	if err != nil {
		return nil, err
	}
	defer conn.Close()

	var pl []*control.Process
	// TODO: Pass in the container id (cid) here. The sandbox
	// should return process info for only that container.
	if err := conn.Call(boot.ContainerProcesses, nil, &pl); err != nil {
		return nil, fmt.Errorf("error retrieving process data from sandbox: %v", err)
	}
	return pl, nil
}

// Execute runs the specified command in the container. It returns the pid of
// the newly created process.
func (s *Sandbox) Execute(cid string, args *control.ExecArgs) (int32, error) {
	log.Debugf("Executing new process in container %q in sandbox %q", cid, s.ID)
	conn, err := s.sandboxConnect()
	if err != nil {
		return 0, s.connError(err)
	}
	defer conn.Close()

	rpcArgs := &boot.ExecArgs{ExecArgs: *args, CID: cid}

	// Send a message to the sandbox control server to start the container.
	var pid int32
	if err := conn.Call(boot.ContainerExecuteAsync, rpcArgs, &pid); err != nil {
		return 0, fmt.Errorf("error executing in sandbox: %v", err)
	}
	return pid, nil
}

// Event retrieves stats about the sandbox such as memory and CPU utilization.
func (s *Sandbox) Event(cid string) (*boot.Event, error) {
	log.Debugf("Getting events for container %q in sandbox %q", cid, s.ID)
	conn, err := s.sandboxConnect()
	if err != nil {
		return nil, err
	}
	defer conn.Close()

	var e boot.Event
	// TODO: Pass in the container id (cid) here. The sandbox
	// should return events only for that container.
	if err := conn.Call(boot.ContainerEvent, nil, &e); err != nil {
		return nil, fmt.Errorf("error retrieving event data from sandbox: %v", err)
	}
	e.ID = cid
	return &e, nil
}

func (s *Sandbox) sandboxConnect() (*urpc.Client, error) {
	log.Debugf("Connecting to sandbox %q", s.ID)
	conn, err := client.ConnectTo(boot.ControlSocketAddr(s.ID))
	if err != nil {
		return nil, s.connError(err)
	}
	return conn, nil
}

func (s *Sandbox) connError(err error) error {
	return fmt.Errorf("error connecting to control server at pid %d: %v", s.Pid, err)
}

// createSandboxProcess starts the sandbox as a subprocess by running the "boot"
// command, passing in the bundle dir.
func (s *Sandbox) createSandboxProcess(spec *specs.Spec, conf *boot.Config, bundleDir, consoleSocket string, ioFiles []*os.File) error {
	// nextFD is used to get unused FDs that we can pass to the sandbox.  It
	// starts at 3 because 0, 1, and 2 are taken by stdin/out/err.
	nextFD := 3

	binPath, err := specutils.BinPath()
	if err != nil {
		return err
	}
	cmd := exec.Command(binPath, conf.ToFlags()...)
	cmd.SysProcAttr = &syscall.SysProcAttr{}

	// Open the log files to pass to the sandbox as FDs.
	//
	// These flags must come BEFORE the "boot" command in cmd.Args.
	if conf.LogFilename != "" {
		logFile, err := os.OpenFile(conf.LogFilename, os.O_APPEND|os.O_CREATE|os.O_WRONLY, 0644)
		if err != nil {
			return fmt.Errorf("error opening log file %q: %v", conf.LogFilename, err)
		}
		defer logFile.Close()
		cmd.ExtraFiles = append(cmd.ExtraFiles, logFile)
		cmd.Args = append(cmd.Args, "--log-fd="+strconv.Itoa(nextFD))
		nextFD++
	}
	if conf.DebugLogDir != "" {
		debugLogFile, err := specutils.DebugLogFile(conf.DebugLogDir, "boot")
		if err != nil {
			return fmt.Errorf("error opening debug log file in %q: %v", conf.DebugLogDir, err)
		}
		defer debugLogFile.Close()
		cmd.ExtraFiles = append(cmd.ExtraFiles, debugLogFile)
		cmd.Args = append(cmd.Args, "--debug-log-fd="+strconv.Itoa(nextFD))
		nextFD++
	}

	// Add the "boot" command to the args.
	//
	// All flags after this must be for the boot command
	cmd.Args = append(cmd.Args, "boot", "--bundle="+bundleDir)

	consoleEnabled := consoleSocket != ""
	cmd.Args = append(cmd.Args, "--console="+strconv.FormatBool(consoleEnabled))

	// Create a socket for the control server and donate it to the sandbox.
	addr := boot.ControlSocketAddr(s.ID)
	sockFD, err := server.CreateSocket(addr)
	log.Infof("Creating sandbox process with addr: %s", addr[1:]) // skip "\00".
	if err != nil {
		return fmt.Errorf("error creating control server socket for sandbox %q: %v", s.ID, err)
	}
	controllerFile := os.NewFile(uintptr(sockFD), "control_server_socket")
	defer controllerFile.Close()
	cmd.ExtraFiles = append(cmd.ExtraFiles, controllerFile)
	cmd.Args = append(cmd.Args, "--controller-fd="+strconv.Itoa(nextFD))
	nextFD++

	// Open the spec file to donate to the sandbox.
	if conf.SpecFile == "" {
		return fmt.Errorf("conf.SpecFile must be set")
	}
	specFile, err := os.Open(conf.SpecFile)
	if err != nil {
		return fmt.Errorf("error opening spec file %q: %v", conf.SpecFile, err)
	}
	defer specFile.Close()
	cmd.ExtraFiles = append(cmd.ExtraFiles, specFile)
	cmd.Args = append(cmd.Args, "--spec-fd="+strconv.Itoa(nextFD))
	nextFD++

	// If there is a gofer, sends all socket ends to the sandbox.
	for _, f := range ioFiles {
		defer f.Close()
		cmd.ExtraFiles = append(cmd.ExtraFiles, f)
		cmd.Args = append(cmd.Args, "--io-fds="+strconv.Itoa(nextFD))
		nextFD++
	}

	// If the platform needs a device fd we must pass it in.
	if deviceFile, err := deviceFileForPlatform(conf.Platform); err != nil {
		return err
	} else if deviceFile != nil {
		defer deviceFile.Close()
		cmd.ExtraFiles = append(cmd.ExtraFiles, deviceFile)
		cmd.Args = append(cmd.Args, "--device-fd="+strconv.Itoa(nextFD))
		nextFD++
	}

	// Sandbox stdio defaults to current process stdio.
	cmd.Stdin = os.Stdin
	cmd.Stdout = os.Stdout
	cmd.Stderr = os.Stderr

	// If the console control socket file is provided, then create a new
	// pty master/slave pair and set the tty on the sandbox process.
	if consoleEnabled {
		// console.NewWithSocket will send the master on the socket,
		// and return the slave.
		tty, err := console.NewWithSocket(consoleSocket)
		if err != nil {
			return fmt.Errorf("error setting up console with socket %q: %v", consoleSocket, err)
		}
		defer tty.Close()

		cmd.Stdin = tty
		cmd.Stdout = tty
		cmd.Stderr = tty
		cmd.SysProcAttr.Setctty = true
		cmd.SysProcAttr.Ctty = int(tty.Fd())
	}

	// Detach from this session, otherwise cmd will get SIGHUP and SIGCONT
	// when re-parented.
	cmd.SysProcAttr.Setsid = true

	// nss is the set of namespaces to join or create before starting the sandbox
	// process. IPC and UTS namespaces from the host are not used as they
	// are virtualized inside the sandbox. Be paranoid and run inside an empty
	// namespace for these.
	log.Infof("Sandbox will be started in new IPC and UTS namespaces")
	nss := []specs.LinuxNamespace{
		{Type: specs.IPCNamespace},
		{Type: specs.UTSNamespace},
	}

	if conf.Platform == boot.PlatformPtrace {
		// TODO: Also set a new PID namespace so that we limit
		// access to other host processes.
		log.Infof("Sandbox will be started in the current PID namespace")
	} else {
		log.Infof("Sandbox will be started in a new PID namespace")
		nss = append(nss, specs.LinuxNamespace{Type: specs.PIDNamespace})
	}

	log.Infof("Sandbox will be started in new mount namespace")
	nss = append(nss, specs.LinuxNamespace{Type: specs.MountNamespace})

	// Joins the network namespace if network is enabled. the sandbox talks
	// directly to the host network, which may have been configured in the
	// namespace.
	if ns, ok := specutils.GetNS(specs.NetworkNamespace, spec); ok && conf.Network != boot.NetworkNone {
		log.Infof("Sandbox will be started in the container's network namespace: %+v", ns)
		nss = append(nss, ns)
	} else {
		log.Infof("Sandbox will be started in new network namespace")
		nss = append(nss, specs.LinuxNamespace{Type: specs.NetworkNamespace})
	}

	// User namespace depends on the following options:
	//   - Host network/filesystem: requires to run inside the user namespace
	//       specified in the spec or the current namespace if none is configured.
	if conf.Network == boot.NetworkHost {
		if userns, ok := specutils.GetNS(specs.UserNamespace, spec); ok {
			log.Infof("Sandbox will be started in container's user namespace: %+v", userns)
			nss = append(nss, userns)
			specutils.SetUIDGIDMappings(cmd, spec)
		} else {
			log.Infof("Sandbox will be started in the current user namespace")
		}
		// When running in the caller's defined user namespace, apply the same
		// capabilities to the sandbox process to ensure it abides to the same
		// rules.
		cmd.Args = append(cmd.Args, "--apply-caps=true")

	} else {
		log.Infof("Sandbox will be started in new user namespace")
		nss = append(nss, specs.LinuxNamespace{Type: specs.UserNamespace})

		// If we have CAP_SETUID and CAP_SETGID, then we can also run
		// as user nobody.
		if conf.TestOnlyAllowRunAsCurrentUserWithoutChroot {
			log.Warningf("Running sandbox in test mode as current user (uid=%d gid=%d). This is only safe in tests!", os.Getuid(), os.Getgid())
		} else if specutils.CanSetUIDGID() {
			// Map nobody in the new namespace to nobody in the parent namespace.
			const nobody = 65534
			cmd.SysProcAttr.UidMappings = []syscall.SysProcIDMap{{
				ContainerID: int(nobody),
				HostID:      int(nobody),
				Size:        int(1),
			}}
			cmd.SysProcAttr.GidMappings = []syscall.SysProcIDMap{{
				ContainerID: int(nobody),
				HostID:      int(nobody),
				Size:        int(1),
			}}

			// Set credentials to run as user and group nobody.
			cmd.SysProcAttr.Credential = &syscall.Credential{
				Uid: nobody,
				Gid: nobody,
			}
		} else {
			return fmt.Errorf("can't run sandbox process as user nobody since we don't have CAP_SETUID or CAP_SETGID")
		}

		// If we have CAP_SYS_ADMIN, we can create an empty chroot and
		// bind-mount the executable inside it.
		if conf.TestOnlyAllowRunAsCurrentUserWithoutChroot {
			log.Warningf("Running sandbox in test mode without chroot. This is only safe in tests!")
		} else if specutils.HasCapSysAdmin() {
			log.Infof("Sandbox will be started in minimal chroot")
			chroot, err := setUpChroot()
			if err != nil {
				return fmt.Errorf("error setting up chroot: %v", err)
			}
			cmd.SysProcAttr.Chroot = chroot
			cmd.Args[0] = "/runsc"
			cmd.Path = "/runsc"
		} else {
			return fmt.Errorf("can't run sandbox process in minimal chroot since we don't have CAP_SYS_ADMIN")
		}
	}

	// Log the fds we are donating to the sandbox process.
	for i, f := range cmd.ExtraFiles {
		log.Debugf("Donating FD %d: %q", i+3, f.Name())
	}

	log.Debugf("Starting sandbox: %s %v", binPath, cmd.Args)
	log.Debugf("SysProcAttr: %+v", cmd.SysProcAttr)
	if err := specutils.StartInNS(cmd, nss); err != nil {
		return err
	}
	s.Pid = cmd.Process.Pid
	log.Infof("Sandbox started, pid: %d", s.Pid)

	return nil
}

// waitForCreated waits for the sandbox subprocess control server to be
// running and for the loader to have been created, at which point the sandbox
// is in Created state.
func (s *Sandbox) waitForCreated(timeout time.Duration) error {
	log.Debugf("Waiting for sandbox %q creation", s.ID)

	ready := func() (bool, error) {
		c, err := client.ConnectTo(boot.ControlSocketAddr(s.ID))
		if err != nil {
			return false, nil
		}
		// It's alive!
		c.Close()
		return true, nil
	}
	if err := specutils.WaitForReady(s.Pid, timeout, ready); err != nil {
		return fmt.Errorf("unexpected error waiting for sandbox %q, err: %v", s.ID, err)
	}
	conn, err := s.sandboxConnect()
	if err != nil {
		return err
	}
	defer conn.Close()

	if err := conn.Call(boot.ContainerWaitForLoader, nil, nil); err != nil {
		return fmt.Errorf("err waiting on loader on sandbox %q, err: %v", s.ID, err)
	}
	return nil
}

// Wait waits for the containerized process to exit, and returns its WaitStatus.
func (s *Sandbox) Wait(cid string) (syscall.WaitStatus, error) {
	log.Debugf("Waiting for container %q in sandbox %q", cid, s.ID)
	var ws syscall.WaitStatus
	conn, err := s.sandboxConnect()
	if err != nil {
		return ws, err
	}
	defer conn.Close()

	if err := conn.Call(boot.ContainerWait, &cid, &ws); err != nil {
		return ws, fmt.Errorf("error waiting on container %q: %v", cid, err)
	}
	return ws, nil
}

// WaitPID waits for process 'pid' in the container's sandbox and returns its
// WaitStatus.
func (s *Sandbox) WaitPID(cid string, pid int32, clearStatus bool) (syscall.WaitStatus, error) {
	log.Debugf("Waiting for PID %d in sandbox %q", pid, s.ID)
	var ws syscall.WaitStatus
	conn, err := s.sandboxConnect()
	if err != nil {
		return ws, err
	}
	defer conn.Close()

	args := &boot.WaitPIDArgs{
		PID:         pid,
		CID:         cid,
		ClearStatus: clearStatus,
	}
	if err := conn.Call(boot.ContainerWaitPID, args, &ws); err != nil {
		return ws, fmt.Errorf("error waiting on PID %d in sandbox %q: %v", pid, s.ID, err)
	}
	return ws, nil
}

// IsRootContainer returns true if the specified container ID belongs to the
// root container.
func (s *Sandbox) IsRootContainer(cid string) bool {
	return s.ID == cid
}

// Destroy frees all resources associated with the sandbox.
// Destroy returns error if any step fails, and the function can be safely retried.
func (s *Sandbox) Destroy() error {
	log.Debugf("Destroy sandbox %q", s.ID)
	if s.Pid != 0 {
		log.Debugf("Killing sandbox %q", s.ID)
		if err := syscall.Kill(s.Pid, syscall.SIGKILL); err != nil && err != syscall.ESRCH {
			return fmt.Errorf("error killing sandbox %q PID %q: %v", s.ID, s.Pid, err)
		}
		if err := s.waitForStopped(); err != nil {
			return fmt.Errorf("error waiting sandbox %q stop: %v", s.ID, err)
		}
	}

	if s.Chroot != "" {
		return tearDownChroot(s.Chroot)
	}

	return nil
}

// Signal sends the signal to a container in the sandbox.
func (s *Sandbox) Signal(cid string, sig syscall.Signal) error {
	log.Debugf("Signal sandbox %q", s.ID)
	conn, err := s.sandboxConnect()
	if err != nil {
		return err
	}
	defer conn.Close()

	args := boot.SignalArgs{
		CID:   cid,
		Signo: int32(sig),
	}
	if err := conn.Call(boot.ContainerSignal, &args, nil); err != nil {
		return fmt.Errorf("err signaling container %q: %v", cid, err)
	}
	return nil
}

// Checkpoint sends the checkpoint call for a container in the sandbox.
// The statefile will be written to f.
func (s *Sandbox) Checkpoint(cid string, f *os.File) error {
	log.Debugf("Checkpoint sandbox %q", s.ID)
	conn, err := s.sandboxConnect()
	if err != nil {
		return err
	}
	defer conn.Close()

	opt := control.SaveOpts{
		FilePayload: urpc.FilePayload{
			Files: []*os.File{f},
		},
	}

	if err := conn.Call(boot.ContainerCheckpoint, &opt, nil); err != nil {
		return fmt.Errorf("err checkpointing container %q: %v", cid, err)
	}
	return nil
}

// Pause sends the pause call for a container in the sandbox.
func (s *Sandbox) Pause(cid string) error {
	log.Debugf("Pause sandbox %q", s.ID)
	conn, err := s.sandboxConnect()
	if err != nil {
		return err
	}
	defer conn.Close()

	if err := conn.Call(boot.ContainerPause, nil, nil); err != nil {
		return fmt.Errorf("err pausing container %q: %v", cid, err)
	}
	return nil
}

// Resume sends the resume call for a container in the sandbox.
func (s *Sandbox) Resume(cid string) error {
	log.Debugf("Resume sandbox %q", s.ID)
	conn, err := s.sandboxConnect()
	if err != nil {
		return err
	}
	defer conn.Close()

	if err := conn.Call(boot.ContainerResume, nil, nil); err != nil {
		return fmt.Errorf("err resuming container %q: %v", cid, err)
	}
	return nil
}

// IsRunning returns true if the sandbox or gofer process is running.
func (s *Sandbox) IsRunning() bool {
	if s.Pid != 0 {
		// Send a signal 0 to the sandbox process.
		if err := syscall.Kill(s.Pid, 0); err == nil {
			// Succeeded, process is running.
			return true
		}
	}
	return false
}

// Stacks collects and returns all stacks for the sandbox.
func (s *Sandbox) Stacks() (string, error) {
	log.Debugf("Stacks sandbox %q", s.ID)
	conn, err := s.sandboxConnect()
	if err != nil {
		return "", err
	}
	defer conn.Close()

	var stacks string
	if err := conn.Call(boot.SandboxStacks, nil, &stacks); err != nil {
		return "", fmt.Errorf("err getting sandbox %q stacks: %v", s.ID, err)
	}
	return stacks, nil
}

// DestroyContainer destroys the given container. If it is the root container,
// then the entire sandbox is destroyed.
func (s *Sandbox) DestroyContainer(cid string) error {
	if s.IsRootContainer(cid) {
		log.Debugf("Destroying root container %q by destroying sandbox", cid)
		return s.Destroy()
	}

	log.Debugf("Destroying container %q in sandbox %q", cid, s.ID)
	conn, err := s.sandboxConnect()
	if err != nil {
		return err
	}
	defer conn.Close()
	if err := conn.Call(boot.ContainerDestroy, &cid, nil); err != nil {
		return fmt.Errorf("error destroying container %q: %v", cid, err)
	}
	return nil
}

func (s *Sandbox) waitForStopped() error {
	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
	defer cancel()
	b := backoff.WithContext(backoff.NewConstantBackOff(100*time.Millisecond), ctx)
	op := func() error {
		if s.IsRunning() {
			return fmt.Errorf("sandbox is still running")
		}
		return nil
	}
	return backoff.Retry(op, b)
}

// deviceFileForPlatform opens the device file for the given platform. If the
// platform does not need a device file, then nil is returned.
func deviceFileForPlatform(p boot.PlatformType) (*os.File, error) {
	var (
		f   *os.File
		err error
	)
	switch p {
	case boot.PlatformKVM:
		f, err = kvm.OpenDevice()
	default:
		return nil, nil
	}
	if err != nil {
		return nil, fmt.Errorf("error opening device file for platform %q: %v", p, err)
	}
	return f, err
}