summaryrefslogtreecommitdiffhomepage
path: root/pkg/shim/v2/service.go
blob: c67b1bebab2a7d90889ffddd78201465e5e62ea1 (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
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
// 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
//
//     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 v2

import (
	"context"
	"fmt"
	"io/ioutil"
	"os"
	"os/exec"
	"path/filepath"
	"sync"
	"syscall"
	"time"

	"github.com/BurntSushi/toml"
	"github.com/containerd/cgroups"
	metrics "github.com/containerd/cgroups/stats/v1"
	"github.com/containerd/console"
	eventstypes "github.com/containerd/containerd/api/events"
	"github.com/containerd/containerd/api/types/task"
	"github.com/containerd/containerd/errdefs"
	"github.com/containerd/containerd/events"
	"github.com/containerd/containerd/log"
	"github.com/containerd/containerd/mount"
	"github.com/containerd/containerd/namespaces"
	"github.com/containerd/containerd/runtime"
	"github.com/containerd/containerd/runtime/linux/runctypes"
	rproc "github.com/containerd/containerd/runtime/proc"
	"github.com/containerd/containerd/runtime/v2/shim"
	taskAPI "github.com/containerd/containerd/runtime/v2/task"
	runtimeoptions "github.com/containerd/cri/pkg/api/runtimeoptions/v1"
	"github.com/containerd/typeurl"
	ptypes "github.com/gogo/protobuf/types"
	"golang.org/x/sys/unix"

	"gvisor.dev/gvisor/pkg/shim/runsc"
	"gvisor.dev/gvisor/pkg/shim/v1/proc"
	"gvisor.dev/gvisor/pkg/shim/v1/utils"
	"gvisor.dev/gvisor/pkg/shim/v2/options"
	"gvisor.dev/gvisor/runsc/specutils"
)

var (
	empty   = &ptypes.Empty{}
	bufPool = sync.Pool{
		New: func() interface{} {
			buffer := make([]byte, 32<<10)
			return &buffer
		},
	}
)

var _ = (taskAPI.TaskService)(&service{})

// configFile is the default config file name. For containerd 1.2,
// we assume that a config.toml should exist in the runtime root.
const configFile = "config.toml"

// New returns a new shim service that can be used via GRPC.
func New(ctx context.Context, id string, publisher events.Publisher) (shim.Shim, error) {
	ep, err := newOOMEpoller(publisher)
	if err != nil {
		return nil, err
	}
	ctx, cancel := context.WithCancel(ctx)
	go ep.run(ctx)
	s := &service{
		id:        id,
		context:   ctx,
		processes: make(map[string]rproc.Process),
		events:    make(chan interface{}, 128),
		ec:        proc.ExitCh,
		oomPoller: ep,
		cancel:    cancel,
	}
	go s.processExits()
	runsc.Monitor = shim.Default
	if err := s.initPlatform(); err != nil {
		cancel()
		return nil, fmt.Errorf("failed to initialized platform behavior: %w", err)
	}
	go s.forward(publisher)
	return s, nil
}

// service is the shim implementation of a remote shim over GRPC.
type service struct {
	mu sync.Mutex

	context   context.Context
	task      rproc.Process
	processes map[string]rproc.Process
	events    chan interface{}
	platform  rproc.Platform
	opts      options.Options
	ec        chan proc.Exit
	oomPoller *epoller

	id     string
	bundle string
	cancel func()
}

func newCommand(ctx context.Context, containerdBinary, containerdAddress string) (*exec.Cmd, error) {
	ns, err := namespaces.NamespaceRequired(ctx)
	if err != nil {
		return nil, err
	}
	self, err := os.Executable()
	if err != nil {
		return nil, err
	}
	cwd, err := os.Getwd()
	if err != nil {
		return nil, err
	}
	args := []string{
		"-namespace", ns,
		"-address", containerdAddress,
		"-publish-binary", containerdBinary,
	}
	cmd := exec.Command(self, args...)
	cmd.Dir = cwd
	cmd.Env = append(os.Environ(), "GOMAXPROCS=2")
	cmd.SysProcAttr = &syscall.SysProcAttr{
		Setpgid: true,
	}
	return cmd, nil
}

func (s *service) StartShim(ctx context.Context, id, containerdBinary, containerdAddress string) (string, error) {
	cmd, err := newCommand(ctx, containerdBinary, containerdAddress)
	if err != nil {
		return "", err
	}
	address, err := shim.SocketAddress(ctx, id)
	if err != nil {
		return "", err
	}
	socket, err := shim.NewSocket(address)
	if err != nil {
		return "", err
	}
	defer socket.Close()
	f, err := socket.File()
	if err != nil {
		return "", err
	}
	defer f.Close()

	cmd.ExtraFiles = append(cmd.ExtraFiles, f)

	if err := cmd.Start(); err != nil {
		return "", err
	}
	defer func() {
		if err != nil {
			cmd.Process.Kill()
		}
	}()
	// make sure to wait after start
	go cmd.Wait()
	if err := shim.WritePidFile("shim.pid", cmd.Process.Pid); err != nil {
		return "", err
	}
	if err := shim.WriteAddress("address", address); err != nil {
		return "", err
	}
	if err := shim.SetScore(cmd.Process.Pid); err != nil {
		return "", fmt.Errorf("failed to set OOM Score on shim: %w", err)
	}
	return address, nil
}

func (s *service) Cleanup(ctx context.Context) (*taskAPI.DeleteResponse, error) {
	path, err := os.Getwd()
	if err != nil {
		return nil, err
	}
	ns, err := namespaces.NamespaceRequired(ctx)
	if err != nil {
		return nil, err
	}
	runtime, err := s.readRuntime(path)
	if err != nil {
		return nil, err
	}
	r := proc.NewRunsc(s.opts.Root, path, ns, runtime, nil)
	if err := r.Delete(ctx, s.id, &runsc.DeleteOpts{
		Force: true,
	}); err != nil {
		log.L.Printf("failed to remove runc container: %v", err)
	}
	if err := mount.UnmountAll(filepath.Join(path, "rootfs"), 0); err != nil {
		log.L.Printf("failed to cleanup rootfs mount: %v", err)
	}
	return &taskAPI.DeleteResponse{
		ExitedAt:   time.Now(),
		ExitStatus: 128 + uint32(unix.SIGKILL),
	}, nil
}

func (s *service) readRuntime(path string) (string, error) {
	data, err := ioutil.ReadFile(filepath.Join(path, "runtime"))
	if err != nil {
		return "", err
	}
	return string(data), nil
}

func (s *service) writeRuntime(path, runtime string) error {
	return ioutil.WriteFile(filepath.Join(path, "runtime"), []byte(runtime), 0600)
}

// Create creates a new initial process and container with the underlying OCI
// runtime.
func (s *service) Create(ctx context.Context, r *taskAPI.CreateTaskRequest) (_ *taskAPI.CreateTaskResponse, err error) {
	s.mu.Lock()
	defer s.mu.Unlock()

	ns, err := namespaces.NamespaceRequired(ctx)
	if err != nil {
		return nil, fmt.Errorf("create namespace: %w", err)
	}

	// Read from root for now.
	var opts options.Options
	if r.Options != nil {
		v, err := typeurl.UnmarshalAny(r.Options)
		if err != nil {
			return nil, err
		}
		var path string
		switch o := v.(type) {
		case *runctypes.CreateOptions: // containerd 1.2.x
			opts.IoUid = o.IoUid
			opts.IoGid = o.IoGid
			opts.ShimCgroup = o.ShimCgroup
		case *runctypes.RuncOptions: // containerd 1.2.x
			root := proc.RunscRoot
			if o.RuntimeRoot != "" {
				root = o.RuntimeRoot
			}

			opts.BinaryName = o.Runtime

			path = filepath.Join(root, configFile)
			if _, err := os.Stat(path); err != nil {
				if !os.IsNotExist(err) {
					return nil, fmt.Errorf("stat config file %q: %w", path, err)
				}
				// A config file in runtime root is not required.
				path = ""
			}
		case *runtimeoptions.Options: // containerd 1.3.x+
			if o.ConfigPath == "" {
				break
			}
			if o.TypeUrl != options.OptionType {
				return nil, fmt.Errorf("unsupported runtimeoptions %q", o.TypeUrl)
			}
			path = o.ConfigPath
		default:
			return nil, fmt.Errorf("unsupported option type %q", r.Options.TypeUrl)
		}
		if path != "" {
			if _, err = toml.DecodeFile(path, &opts); err != nil {
				return nil, fmt.Errorf("decode config file %q: %w", path, err)
			}
		}
	}

	var mounts []proc.Mount
	for _, m := range r.Rootfs {
		mounts = append(mounts, proc.Mount{
			Type:    m.Type,
			Source:  m.Source,
			Target:  m.Target,
			Options: m.Options,
		})
	}

	rootfs := filepath.Join(r.Bundle, "rootfs")
	if err := os.Mkdir(rootfs, 0711); err != nil && !os.IsExist(err) {
		return nil, err
	}

	config := &proc.CreateConfig{
		ID:       r.ID,
		Bundle:   r.Bundle,
		Runtime:  opts.BinaryName,
		Rootfs:   mounts,
		Terminal: r.Terminal,
		Stdin:    r.Stdin,
		Stdout:   r.Stdout,
		Stderr:   r.Stderr,
		Options:  r.Options,
	}
	if err := s.writeRuntime(r.Bundle, opts.BinaryName); err != nil {
		return nil, err
	}
	defer func() {
		if err != nil {
			if err := mount.UnmountAll(rootfs, 0); err != nil {
				log.L.Printf("failed to cleanup rootfs mount: %v", err)
			}
		}
	}()
	for _, rm := range mounts {
		m := &mount.Mount{
			Type:    rm.Type,
			Source:  rm.Source,
			Options: rm.Options,
		}
		if err := m.Mount(rootfs); err != nil {
			return nil, fmt.Errorf("failed to mount rootfs component %v: %w", m, err)
		}
	}
	process, err := newInit(
		ctx,
		r.Bundle,
		filepath.Join(r.Bundle, "work"),
		ns,
		s.platform,
		config,
		&opts,
		rootfs,
	)
	if err != nil {
		return nil, errdefs.ToGRPC(err)
	}
	if err := process.Create(ctx, config); err != nil {
		return nil, errdefs.ToGRPC(err)
	}
	// Save the main task id and bundle to the shim for additional
	// requests.
	s.id = r.ID
	s.bundle = r.Bundle

	// Set up OOM notification on the sandbox's cgroup. This is done on
	// sandbox create since the sandbox process will be created here.
	pid := process.Pid()
	if pid > 0 {
		cg, err := cgroups.Load(cgroups.V1, cgroups.PidPath(pid))
		if err != nil {
			return nil, fmt.Errorf("loading cgroup for %d: %w", pid, err)
		}
		if err := s.oomPoller.add(s.id, cg); err != nil {
			return nil, fmt.Errorf("add cg to OOM monitor: %w", err)
		}
	}
	s.task = process
	s.opts = opts
	return &taskAPI.CreateTaskResponse{
		Pid: uint32(process.Pid()),
	}, nil

}

// Start starts a process.
func (s *service) Start(ctx context.Context, r *taskAPI.StartRequest) (*taskAPI.StartResponse, error) {
	p, err := s.getProcess(r.ExecID)
	if err != nil {
		return nil, err
	}
	if err := p.Start(ctx); err != nil {
		return nil, err
	}
	// TODO: Set the cgroup and oom notifications on restore.
	// https://github.com/google/gvisor-containerd-shim/issues/58
	return &taskAPI.StartResponse{
		Pid: uint32(p.Pid()),
	}, nil
}

// Delete deletes the initial process and container.
func (s *service) Delete(ctx context.Context, r *taskAPI.DeleteRequest) (*taskAPI.DeleteResponse, error) {
	p, err := s.getProcess(r.ExecID)
	if err != nil {
		return nil, err
	}
	if p == nil {
		return nil, errdefs.ToGRPCf(errdefs.ErrFailedPrecondition, "container must be created")
	}
	if err := p.Delete(ctx); err != nil {
		return nil, err
	}
	isTask := r.ExecID == ""
	if !isTask {
		s.mu.Lock()
		delete(s.processes, r.ExecID)
		s.mu.Unlock()
	}
	if isTask && s.platform != nil {
		s.platform.Close()
	}
	return &taskAPI.DeleteResponse{
		ExitStatus: uint32(p.ExitStatus()),
		ExitedAt:   p.ExitedAt(),
		Pid:        uint32(p.Pid()),
	}, nil
}

// Exec spawns an additional process inside the container.
func (s *service) Exec(ctx context.Context, r *taskAPI.ExecProcessRequest) (*ptypes.Empty, error) {
	s.mu.Lock()
	p := s.processes[r.ExecID]
	s.mu.Unlock()
	if p != nil {
		return nil, errdefs.ToGRPCf(errdefs.ErrAlreadyExists, "id %s", r.ExecID)
	}
	p = s.task
	if p == nil {
		return nil, errdefs.ToGRPCf(errdefs.ErrFailedPrecondition, "container must be created")
	}
	process, err := p.(*proc.Init).Exec(ctx, s.bundle, &proc.ExecConfig{
		ID:       r.ExecID,
		Terminal: r.Terminal,
		Stdin:    r.Stdin,
		Stdout:   r.Stdout,
		Stderr:   r.Stderr,
		Spec:     r.Spec,
	})
	if err != nil {
		return nil, errdefs.ToGRPC(err)
	}
	s.mu.Lock()
	s.processes[r.ExecID] = process
	s.mu.Unlock()
	return empty, nil
}

// ResizePty resizes the terminal of a process.
func (s *service) ResizePty(ctx context.Context, r *taskAPI.ResizePtyRequest) (*ptypes.Empty, error) {
	p, err := s.getProcess(r.ExecID)
	if err != nil {
		return nil, err
	}
	ws := console.WinSize{
		Width:  uint16(r.Width),
		Height: uint16(r.Height),
	}
	if err := p.Resize(ws); err != nil {
		return nil, errdefs.ToGRPC(err)
	}
	return empty, nil
}

// State returns runtime state information for a process.
func (s *service) State(ctx context.Context, r *taskAPI.StateRequest) (*taskAPI.StateResponse, error) {
	p, err := s.getProcess(r.ExecID)
	if err != nil {
		return nil, err
	}
	st, err := p.Status(ctx)
	if err != nil {
		return nil, err
	}
	status := task.StatusUnknown
	switch st {
	case "created":
		status = task.StatusCreated
	case "running":
		status = task.StatusRunning
	case "stopped":
		status = task.StatusStopped
	}
	sio := p.Stdio()
	return &taskAPI.StateResponse{
		ID:         p.ID(),
		Bundle:     s.bundle,
		Pid:        uint32(p.Pid()),
		Status:     status,
		Stdin:      sio.Stdin,
		Stdout:     sio.Stdout,
		Stderr:     sio.Stderr,
		Terminal:   sio.Terminal,
		ExitStatus: uint32(p.ExitStatus()),
		ExitedAt:   p.ExitedAt(),
	}, nil
}

// Pause the container.
func (s *service) Pause(ctx context.Context, r *taskAPI.PauseRequest) (*ptypes.Empty, error) {
	return empty, errdefs.ToGRPC(errdefs.ErrNotImplemented)
}

// Resume the container.
func (s *service) Resume(ctx context.Context, r *taskAPI.ResumeRequest) (*ptypes.Empty, error) {
	return empty, errdefs.ToGRPC(errdefs.ErrNotImplemented)
}

// Kill a process with the provided signal.
func (s *service) Kill(ctx context.Context, r *taskAPI.KillRequest) (*ptypes.Empty, error) {
	p, err := s.getProcess(r.ExecID)
	if err != nil {
		return nil, err
	}
	if p == nil {
		return nil, errdefs.ToGRPCf(errdefs.ErrFailedPrecondition, "container must be created")
	}
	if err := p.Kill(ctx, r.Signal, r.All); err != nil {
		return nil, errdefs.ToGRPC(err)
	}
	return empty, nil
}

// Pids returns all pids inside the container.
func (s *service) Pids(ctx context.Context, r *taskAPI.PidsRequest) (*taskAPI.PidsResponse, error) {
	pids, err := s.getContainerPids(ctx, r.ID)
	if err != nil {
		return nil, errdefs.ToGRPC(err)
	}
	var processes []*task.ProcessInfo
	for _, pid := range pids {
		pInfo := task.ProcessInfo{
			Pid: pid,
		}
		for _, p := range s.processes {
			if p.Pid() == int(pid) {
				d := &runctypes.ProcessDetails{
					ExecID: p.ID(),
				}
				a, err := typeurl.MarshalAny(d)
				if err != nil {
					return nil, fmt.Errorf("failed to marshal process %d info: %w", pid, err)
				}
				pInfo.Info = a
				break
			}
		}
		processes = append(processes, &pInfo)
	}
	return &taskAPI.PidsResponse{
		Processes: processes,
	}, nil
}

// CloseIO closes the I/O context of a process.
func (s *service) CloseIO(ctx context.Context, r *taskAPI.CloseIORequest) (*ptypes.Empty, error) {
	p, err := s.getProcess(r.ExecID)
	if err != nil {
		return nil, err
	}
	if stdin := p.Stdin(); stdin != nil {
		if err := stdin.Close(); err != nil {
			return nil, fmt.Errorf("close stdin: %w", err)
		}
	}
	return empty, nil
}

// Checkpoint checkpoints the container.
func (s *service) Checkpoint(ctx context.Context, r *taskAPI.CheckpointTaskRequest) (*ptypes.Empty, error) {
	return empty, errdefs.ToGRPC(errdefs.ErrNotImplemented)
}

// Connect returns shim information such as the shim's pid.
func (s *service) Connect(ctx context.Context, r *taskAPI.ConnectRequest) (*taskAPI.ConnectResponse, error) {
	var pid int
	if s.task != nil {
		pid = s.task.Pid()
	}
	return &taskAPI.ConnectResponse{
		ShimPid: uint32(os.Getpid()),
		TaskPid: uint32(pid),
	}, nil
}

func (s *service) Shutdown(ctx context.Context, r *taskAPI.ShutdownRequest) (*ptypes.Empty, error) {
	s.cancel()
	os.Exit(0)
	return empty, nil
}

func (s *service) Stats(ctx context.Context, r *taskAPI.StatsRequest) (*taskAPI.StatsResponse, error) {
	path, err := os.Getwd()
	if err != nil {
		return nil, err
	}
	ns, err := namespaces.NamespaceRequired(ctx)
	if err != nil {
		return nil, err
	}
	runtime, err := s.readRuntime(path)
	if err != nil {
		return nil, err
	}
	rs := proc.NewRunsc(s.opts.Root, path, ns, runtime, nil)
	stats, err := rs.Stats(ctx, s.id)
	if err != nil {
		return nil, err
	}

	// gvisor currently (as of 2020-03-03) only returns the total memory
	// usage and current PID value[0]. However, we copy the common fields here
	// so that future updates will propagate correct information.  We're
	// using the metrics.Metrics structure so we're returning the same type
	// as runc.
	//
	// [0]: https://github.com/google/gvisor/blob/277a0d5a1fbe8272d4729c01ee4c6e374d047ebc/runsc/boot/events.go#L61-L81
	data, err := typeurl.MarshalAny(&metrics.Metrics{
		CPU: &metrics.CPUStat{
			Usage: &metrics.CPUUsage{
				Total:  stats.Cpu.Usage.Total,
				Kernel: stats.Cpu.Usage.Kernel,
				User:   stats.Cpu.Usage.User,
				PerCPU: stats.Cpu.Usage.Percpu,
			},
			Throttling: &metrics.Throttle{
				Periods:          stats.Cpu.Throttling.Periods,
				ThrottledPeriods: stats.Cpu.Throttling.ThrottledPeriods,
				ThrottledTime:    stats.Cpu.Throttling.ThrottledTime,
			},
		},
		Memory: &metrics.MemoryStat{
			Cache: stats.Memory.Cache,
			Usage: &metrics.MemoryEntry{
				Limit:   stats.Memory.Usage.Limit,
				Usage:   stats.Memory.Usage.Usage,
				Max:     stats.Memory.Usage.Max,
				Failcnt: stats.Memory.Usage.Failcnt,
			},
			Swap: &metrics.MemoryEntry{
				Limit:   stats.Memory.Swap.Limit,
				Usage:   stats.Memory.Swap.Usage,
				Max:     stats.Memory.Swap.Max,
				Failcnt: stats.Memory.Swap.Failcnt,
			},
			Kernel: &metrics.MemoryEntry{
				Limit:   stats.Memory.Kernel.Limit,
				Usage:   stats.Memory.Kernel.Usage,
				Max:     stats.Memory.Kernel.Max,
				Failcnt: stats.Memory.Kernel.Failcnt,
			},
			KernelTCP: &metrics.MemoryEntry{
				Limit:   stats.Memory.KernelTCP.Limit,
				Usage:   stats.Memory.KernelTCP.Usage,
				Max:     stats.Memory.KernelTCP.Max,
				Failcnt: stats.Memory.KernelTCP.Failcnt,
			},
		},
		Pids: &metrics.PidsStat{
			Current: stats.Pids.Current,
			Limit:   stats.Pids.Limit,
		},
	})
	if err != nil {
		return nil, err
	}
	return &taskAPI.StatsResponse{
		Stats: data,
	}, nil
}

// Update updates a running container.
func (s *service) Update(ctx context.Context, r *taskAPI.UpdateTaskRequest) (*ptypes.Empty, error) {
	return empty, errdefs.ToGRPC(errdefs.ErrNotImplemented)
}

// Wait waits for a process to exit.
func (s *service) Wait(ctx context.Context, r *taskAPI.WaitRequest) (*taskAPI.WaitResponse, error) {
	p, err := s.getProcess(r.ExecID)
	if err != nil {
		return nil, err
	}
	if p == nil {
		return nil, errdefs.ToGRPCf(errdefs.ErrFailedPrecondition, "container must be created")
	}
	p.Wait()

	return &taskAPI.WaitResponse{
		ExitStatus: uint32(p.ExitStatus()),
		ExitedAt:   p.ExitedAt(),
	}, nil
}

func (s *service) processExits() {
	for e := range s.ec {
		s.checkProcesses(e)
	}
}

func (s *service) checkProcesses(e proc.Exit) {
	// TODO(random-liu): Add `shouldKillAll` logic if container pid
	// namespace is supported.
	for _, p := range s.allProcesses() {
		if p.ID() == e.ID {
			if ip, ok := p.(*proc.Init); ok {
				// Ensure all children are killed.
				if err := ip.KillAll(s.context); err != nil {
					log.G(s.context).WithError(err).WithField("id", ip.ID()).
						Error("failed to kill init's children")
				}
			}
			p.SetExited(e.Status)
			s.events <- &eventstypes.TaskExit{
				ContainerID: s.id,
				ID:          p.ID(),
				Pid:         uint32(p.Pid()),
				ExitStatus:  uint32(e.Status),
				ExitedAt:    p.ExitedAt(),
			}
			return
		}
	}
}

func (s *service) allProcesses() (o []rproc.Process) {
	s.mu.Lock()
	defer s.mu.Unlock()
	for _, p := range s.processes {
		o = append(o, p)
	}
	if s.task != nil {
		o = append(o, s.task)
	}
	return o
}

func (s *service) getContainerPids(ctx context.Context, id string) ([]uint32, error) {
	s.mu.Lock()
	p := s.task
	s.mu.Unlock()
	if p == nil {
		return nil, fmt.Errorf("container must be created: %w", errdefs.ErrFailedPrecondition)
	}
	ps, err := p.(*proc.Init).Runtime().Ps(ctx, id)
	if err != nil {
		return nil, err
	}
	pids := make([]uint32, 0, len(ps))
	for _, pid := range ps {
		pids = append(pids, uint32(pid))
	}
	return pids, nil
}

func (s *service) forward(publisher events.Publisher) {
	for e := range s.events {
		ctx, cancel := context.WithTimeout(s.context, 5*time.Second)
		err := publisher.Publish(ctx, getTopic(e), e)
		cancel()
		if err != nil {
			fmt.Errorf("post event: %w", err)
		}
	}
}

func (s *service) getProcess(execID string) (rproc.Process, error) {
	s.mu.Lock()
	defer s.mu.Unlock()
	if execID == "" {
		return s.task, nil
	}
	p := s.processes[execID]
	if p == nil {
		return nil, errdefs.ToGRPCf(errdefs.ErrNotFound, "process does not exist %s", execID)
	}
	return p, nil
}

func getTopic(e interface{}) string {
	switch e.(type) {
	case *eventstypes.TaskCreate:
		return runtime.TaskCreateEventTopic
	case *eventstypes.TaskStart:
		return runtime.TaskStartEventTopic
	case *eventstypes.TaskOOM:
		return runtime.TaskOOMEventTopic
	case *eventstypes.TaskExit:
		return runtime.TaskExitEventTopic
	case *eventstypes.TaskDelete:
		return runtime.TaskDeleteEventTopic
	case *eventstypes.TaskExecAdded:
		return runtime.TaskExecAddedEventTopic
	case *eventstypes.TaskExecStarted:
		return runtime.TaskExecStartedEventTopic
	default:
		log.L.Printf("no topic for type %#v", e)
	}
	return runtime.TaskUnknownTopic
}

func newInit(ctx context.Context, path, workDir, namespace string, platform rproc.Platform, r *proc.CreateConfig, options *options.Options, rootfs string) (*proc.Init, error) {
	spec, err := utils.ReadSpec(r.Bundle)
	if err != nil {
		return nil, fmt.Errorf("read oci spec: %w", err)
	}
	if err := utils.UpdateVolumeAnnotations(r.Bundle, spec); err != nil {
		return nil, fmt.Errorf("update volume annotations: %w", err)
	}
	runsc.FormatLogPath(r.ID, options.RunscConfig)
	runtime := proc.NewRunsc(options.Root, path, namespace, options.BinaryName, options.RunscConfig)
	p := proc.New(r.ID, runtime, rproc.Stdio{
		Stdin:    r.Stdin,
		Stdout:   r.Stdout,
		Stderr:   r.Stderr,
		Terminal: r.Terminal,
	})
	p.Bundle = r.Bundle
	p.Platform = platform
	p.Rootfs = rootfs
	p.WorkDir = workDir
	p.IoUID = int(options.IoUid)
	p.IoGID = int(options.IoGid)
	p.Sandbox = specutils.SpecContainerType(spec) == specutils.ContainerTypeSandbox
	p.UserLog = utils.UserLogPath(spec)
	p.Monitor = shim.Default
	return p, nil
}