aboutsummaryrefslogtreecommitdiff
path: root/libpod/runtime_ctr.go
blob: 0c8d3edab373e1a7382f9555498e171a004b9d22 (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
package libpod

import (
	"context"
	"os"
	"path"
	"path/filepath"
	"strings"
	"time"

	"github.com/containers/libpod/libpod/events"
	"github.com/containers/libpod/libpod/image"
	"github.com/containers/libpod/pkg/rootless"
	"github.com/containers/storage"
	"github.com/containers/storage/pkg/stringid"
	spec "github.com/opencontainers/runtime-spec/specs-go"
	opentracing "github.com/opentracing/opentracing-go"
	"github.com/pkg/errors"
	"github.com/sirupsen/logrus"
)

// CtrRemoveTimeout is the default number of seconds to wait after stopping a container
// before sending the kill signal
const CtrRemoveTimeout = 10

// Contains the public Runtime API for containers

// A CtrCreateOption is a functional option which alters the Container created
// by NewContainer
type CtrCreateOption func(*Container) error

// ContainerFilter is a function to determine whether a container is included
// in command output. Containers to be outputted are tested using the function.
// A true return will include the container, a false return will exclude it.
type ContainerFilter func(*Container) bool

// NewContainer creates a new container from a given OCI config
func (r *Runtime) NewContainer(ctx context.Context, rSpec *spec.Spec, options ...CtrCreateOption) (c *Container, err error) {
	r.lock.Lock()
	defer r.lock.Unlock()
	if !r.valid {
		return nil, ErrRuntimeStopped
	}
	return r.newContainer(ctx, rSpec, options...)
}

func (r *Runtime) newContainer(ctx context.Context, rSpec *spec.Spec, options ...CtrCreateOption) (c *Container, err error) {
	span, _ := opentracing.StartSpanFromContext(ctx, "newContainer")
	span.SetTag("type", "runtime")
	defer span.Finish()

	if rSpec == nil {
		return nil, errors.Wrapf(ErrInvalidArg, "must provide a valid runtime spec to create container")
	}

	ctr := new(Container)
	ctr.config = new(ContainerConfig)
	ctr.state = new(ContainerState)

	ctr.config.ID = stringid.GenerateNonCryptoID()

	ctr.config.Spec = new(spec.Spec)
	if err := JSONDeepCopy(rSpec, ctr.config.Spec); err != nil {
		return nil, errors.Wrapf(err, "error copying runtime spec while creating container")
	}
	ctr.config.CreatedTime = time.Now()

	ctr.config.ShmSize = DefaultShmSize

	ctr.state.BindMounts = make(map[string]string)

	ctr.config.StopTimeout = CtrRemoveTimeout

	ctr.config.OCIRuntime = r.config.OCIRuntime

	// Set namespace based on current runtime namespace
	// Do so before options run so they can override it
	if r.config.Namespace != "" {
		ctr.config.Namespace = r.config.Namespace
	}

	ctr.runtime = r
	for _, option := range options {
		if err := option(ctr); err != nil {
			return nil, errors.Wrapf(err, "error running container create option")
		}
	}

	// Allocate a lock for the container
	lock, err := r.lockManager.AllocateLock()
	if err != nil {
		return nil, errors.Wrapf(err, "error allocating lock for new container")
	}
	ctr.lock = lock
	ctr.config.LockID = ctr.lock.ID()
	logrus.Debugf("Allocated lock %d for container %s", ctr.lock.ID(), ctr.ID())

	ctr.valid = true
	ctr.state.State = ContainerStateConfigured
	ctr.runtime = r

	var pod *Pod
	if ctr.config.Pod != "" {
		// Get the pod from state
		pod, err = r.state.Pod(ctr.config.Pod)
		if err != nil {
			return nil, errors.Wrapf(err, "cannot add container %s to pod %s", ctr.ID(), ctr.config.Pod)
		}
	}

	if ctr.config.Name == "" {
		name, err := r.generateName()
		if err != nil {
			return nil, err
		}

		ctr.config.Name = name
	}

	// Check CGroup parent sanity, and set it if it was not set
	switch r.config.CgroupManager {
	case CgroupfsCgroupsManager:
		if ctr.config.CgroupParent == "" {
			if pod != nil && pod.config.UsePodCgroup {
				podCgroup, err := pod.CgroupPath()
				if err != nil {
					return nil, errors.Wrapf(err, "error retrieving pod %s cgroup", pod.ID())
				}
				if podCgroup == "" {
					return nil, errors.Wrapf(ErrInternal, "pod %s cgroup is not set", pod.ID())
				}
				ctr.config.CgroupParent = podCgroup
			} else {
				ctr.config.CgroupParent = CgroupfsDefaultCgroupParent
			}
		} else if strings.HasSuffix(path.Base(ctr.config.CgroupParent), ".slice") {
			return nil, errors.Wrapf(ErrInvalidArg, "systemd slice received as cgroup parent when using cgroupfs")
		}
	case SystemdCgroupsManager:
		if ctr.config.CgroupParent == "" {
			if pod != nil && pod.config.UsePodCgroup {
				podCgroup, err := pod.CgroupPath()
				if err != nil {
					return nil, errors.Wrapf(err, "error retrieving pod %s cgroup", pod.ID())
				}
				ctr.config.CgroupParent = podCgroup
			} else {
				ctr.config.CgroupParent = SystemdDefaultCgroupParent
			}
		} else if len(ctr.config.CgroupParent) < 6 || !strings.HasSuffix(path.Base(ctr.config.CgroupParent), ".slice") {
			return nil, errors.Wrapf(ErrInvalidArg, "did not receive systemd slice as cgroup parent when using systemd to manage cgroups")
		}
	default:
		return nil, errors.Wrapf(ErrInvalidArg, "unsupported CGroup manager: %s - cannot validate cgroup parent", r.config.CgroupManager)
	}

	// Set up storage for the container
	if err := ctr.setupStorage(ctx); err != nil {
		return nil, err
	}
	defer func() {
		if err != nil {
			if err2 := ctr.teardownStorage(); err2 != nil {
				logrus.Errorf("Error removing partially-created container root filesystem: %s", err2)
			}
		}
	}()

	if rootless.IsRootless() && ctr.config.ConmonPidFile == "" {
		ctr.config.ConmonPidFile = filepath.Join(ctr.state.RunDir, "conmon.pid")
	}

	// Go through named volumes and add them.
	// If they don't exist they will be created using basic options.
	for _, vol := range ctr.config.NamedVolumes {
		// Check if it exists already
		_, err := r.state.Volume(vol.Name)
		if err == nil {
			// The volume exists, we're good
			continue
		} else if errors.Cause(err) != ErrNoSuchVolume {
			return nil, errors.Wrapf(err, "error retrieving named volume %s for new container", vol.Name)
		}

		logrus.Debugf("Creating new volume %s for container", vol.Name)

		// The volume does not exist, so we need to create it.
		newVol, err := r.newVolume(ctx, WithVolumeName(vol.Name), withSetCtrSpecific(),
			WithVolumeUID(ctr.RootUID()), WithVolumeGID(ctr.RootGID()))
		if err != nil {
			return nil, errors.Wrapf(err, "error creating named volume %q", vol.Name)
		}

		if err := ctr.copyWithTarFromImage(vol.Dest, newVol.MountPoint()); err != nil && !os.IsNotExist(err) {
			return nil, errors.Wrapf(err, "Failed to copy content into new volume mount %q", vol.Name)
		}
	}

	if ctr.config.LogPath == "" && ctr.config.LogDriver != JournaldLogging {
		ctr.config.LogPath = filepath.Join(ctr.config.StaticDir, "ctr.log")
	}

	if !MountExists(ctr.config.Spec.Mounts, "/dev/shm") && ctr.config.ShmDir == "" {
		ctr.config.ShmDir = filepath.Join(ctr.bundlePath(), "shm")
		if err := os.MkdirAll(ctr.config.ShmDir, 0700); err != nil {
			if !os.IsExist(err) {
				return nil, errors.Wrapf(err, "unable to create shm %q dir", ctr.config.ShmDir)
			}
		}
		ctr.config.Mounts = append(ctr.config.Mounts, ctr.config.ShmDir)
	}

	// Add the container to the state
	// TODO: May be worth looking into recovering from name/ID collisions here
	if ctr.config.Pod != "" {
		// Lock the pod to ensure we can't add containers to pods
		// being removed
		pod.lock.Lock()
		defer pod.lock.Unlock()

		if err := r.state.AddContainerToPod(pod, ctr); err != nil {
			return nil, err
		}
	} else {
		if err := r.state.AddContainer(ctr); err != nil {
			return nil, err
		}
	}
	ctr.newContainerEvent(events.Create)
	return ctr, nil
}

// RemoveContainer removes the given container
// If force is specified, the container will be stopped first
// If removeVolume is specified, named volumes used by the container will
// be removed also if and only if the container is the sole user
// Otherwise, RemoveContainer will return an error if the container is running
func (r *Runtime) RemoveContainer(ctx context.Context, c *Container, force bool, removeVolume bool) error {
	r.lock.Lock()
	defer r.lock.Unlock()
	return r.removeContainer(ctx, c, force, removeVolume, false)
}

// Internal function to remove a container.
// Locks the container, but does not lock the runtime.
// removePod is used only when removing pods. It instructs Podman to ignore
// infra container protections, and *not* remove from the database (as pod
// remove will handle that).
func (r *Runtime) removeContainer(ctx context.Context, c *Container, force bool, removeVolume bool, removePod bool) error {
	span, _ := opentracing.StartSpanFromContext(ctx, "removeContainer")
	span.SetTag("type", "runtime")
	defer span.Finish()

	if !c.valid {
		if ok, _ := r.state.HasContainer(c.ID()); !ok {
			// Container probably already removed
			// Or was never in the runtime to begin with
			return nil
		}
	}

	// We need to lock the pod before we lock the container.
	// To avoid races around removing a container and the pod it is in.
	// Don't need to do this in pod removal case - we're evicting the entire
	// pod.
	var pod *Pod
	var err error
	runtime := c.runtime
	if c.config.Pod != "" && !removePod {
		pod, err = r.state.Pod(c.config.Pod)
		if err != nil {
			return errors.Wrapf(err, "container %s is in pod %s, but pod cannot be retrieved", c.ID(), pod.ID())
		}

		// Lock the pod while we're removing container
		pod.lock.Lock()
		defer pod.lock.Unlock()
		if err := pod.updatePod(); err != nil {
			return err
		}

		infraID := pod.state.InfraContainerID
		if c.ID() == infraID {
			return errors.Errorf("container %s is the infra container of pod %s and cannot be removed without removing the pod", c.ID(), pod.ID())
		}
	}

	// For pod removal, the container is already locked by the caller
	if !removePod {
		c.lock.Lock()
		defer c.lock.Unlock()
	}

	if !r.valid {
		return ErrRuntimeStopped
	}

	// Update the container to get current state
	if err := c.syncContainer(); err != nil {
		return err
	}

	// If we're not force-removing, we need to check if we're in a good
	// state to remove.
	if !force {
		if err := c.checkReadyForRemoval(); err != nil {
			return err
		}
	}

	if c.state.State == ContainerStatePaused {
		if err := c.runtime.ociRuntime.killContainer(c, 9); err != nil {
			return err
		}
		if err := c.unpause(); err != nil {
			return err
		}
		// Need to update container state to make sure we know it's stopped
		if err := c.waitForExitFileAndSync(); err != nil {
			return err
		}
	}

	// Check that the container's in a good state to be removed
	if c.state.State == ContainerStateRunning {
		if err := r.ociRuntime.stopContainer(c, c.StopTimeout()); err != nil {
			return errors.Wrapf(err, "cannot remove container %s as it could not be stopped", c.ID())
		}

		// Need to update container state to make sure we know it's stopped
		if err := c.waitForExitFileAndSync(); err != nil {
			return err
		}
	}

	// Check that all of our exec sessions have finished
	if len(c.state.ExecSessions) != 0 {
		if err := r.ociRuntime.execStopContainer(c, c.StopTimeout()); err != nil {
			return err
		}
	}

	// Check that no other containers depend on the container.
	// Only used if not removing a pod - pods guarantee that all
	// deps will be evicted at the same time.
	if !removePod {
		deps, err := r.state.ContainerInUse(c)
		if err != nil {
			return err
		}
		if len(deps) != 0 {
			depsStr := strings.Join(deps, ", ")
			return errors.Wrapf(ErrCtrExists, "container %s has dependent containers which must be removed before it: %s", c.ID(), depsStr)
		}
	}

	var cleanupErr error
	// Remove the container from the state
	if c.config.Pod != "" {
		// If we're removing the pod, the container will be evicted
		// from the state elsewhere
		if !removePod {
			if err := r.state.RemoveContainerFromPod(pod, c); err != nil {
				if cleanupErr == nil {
					cleanupErr = err
				} else {
					logrus.Errorf("removing container from pod: %v", err)
				}
			}
		}
	} else {
		if err := r.state.RemoveContainer(c); err != nil {
			if cleanupErr == nil {
				cleanupErr = err
			} else {
				logrus.Errorf("removing container: %v", err)
			}
		}
	}

	// Set container as invalid so it can no longer be used
	c.valid = false

	// Clean up network namespace, cgroups, mounts
	if err := c.cleanup(ctx); err != nil {
		if cleanupErr == nil {
			cleanupErr = errors.Wrapf(err, "error cleaning up container %s", c.ID())
		} else {
			logrus.Errorf("cleanup network, cgroups, mounts: %v", err)
		}
	}

	// Stop the container's storage
	if err := c.teardownStorage(); err != nil {
		if cleanupErr == nil {
			cleanupErr = err
		} else {
			logrus.Errorf("cleanup storage: %v", err)
		}
	}

	// Delete the container.
	// Not needed in Configured and Exited states, where the container
	// doesn't exist in the runtime
	if c.state.State != ContainerStateConfigured &&
		c.state.State != ContainerStateExited {
		if err := c.delete(ctx); err != nil {
			if cleanupErr == nil {
				cleanupErr = err
			} else {
				logrus.Errorf("delete container: %v", err)
			}
		}
	}

	// Deallocate the container's lock
	if err := c.lock.Free(); err != nil {
		if cleanupErr == nil {
			cleanupErr = errors.Wrapf(err, "error freeing lock for container %s", c.ID())
		} else {
			logrus.Errorf("free container lock: %v", err)
		}
	}

	c.newContainerEvent(events.Remove)

	if !removeVolume {
		return cleanupErr
	}

	for _, v := range c.config.NamedVolumes {
		if volume, err := runtime.state.Volume(v.Name); err == nil {
			if !volume.IsCtrSpecific() {
				continue
			}
			if err := runtime.removeVolume(ctx, volume, false); err != nil && err != ErrNoSuchVolume && err != ErrVolumeBeingUsed {
				logrus.Errorf("cleanup volume (%s): %v", v, err)
			}
		}
	}

	return cleanupErr
}

// GetContainer retrieves a container by its ID
func (r *Runtime) GetContainer(id string) (*Container, error) {
	r.lock.RLock()
	defer r.lock.RUnlock()

	if !r.valid {
		return nil, ErrRuntimeStopped
	}

	return r.state.Container(id)
}

// HasContainer checks if a container with the given ID is present
func (r *Runtime) HasContainer(id string) (bool, error) {
	r.lock.RLock()
	defer r.lock.RUnlock()

	if !r.valid {
		return false, ErrRuntimeStopped
	}

	return r.state.HasContainer(id)
}

// LookupContainer looks up a container by its name or a partial ID
// If a partial ID is not unique, an error will be returned
func (r *Runtime) LookupContainer(idOrName string) (*Container, error) {
	r.lock.RLock()
	defer r.lock.RUnlock()

	if !r.valid {
		return nil, ErrRuntimeStopped
	}
	return r.state.LookupContainer(idOrName)
}

// GetContainers retrieves all containers from the state
// Filters can be provided which will determine what containers are included in
// the output. Multiple filters are handled by ANDing their output, so only
// containers matching all filters are returned
func (r *Runtime) GetContainers(filters ...ContainerFilter) ([]*Container, error) {
	r.lock.RLock()
	defer r.lock.RUnlock()

	if !r.valid {
		return nil, ErrRuntimeStopped
	}

	ctrs, err := r.state.AllContainers()
	if err != nil {
		return nil, err
	}

	ctrsFiltered := make([]*Container, 0, len(ctrs))

	for _, ctr := range ctrs {
		include := true
		for _, filter := range filters {
			include = include && filter(ctr)
		}

		if include {
			ctrsFiltered = append(ctrsFiltered, ctr)
		}
	}

	return ctrsFiltered, nil
}

// GetAllContainers is a helper function for GetContainers
func (r *Runtime) GetAllContainers() ([]*Container, error) {
	return r.state.AllContainers()
}

// GetRunningContainers is a helper function for GetContainers
func (r *Runtime) GetRunningContainers() ([]*Container, error) {
	running := func(c *Container) bool {
		state, _ := c.State()
		return state == ContainerStateRunning
	}
	return r.GetContainers(running)
}

// GetContainersByList is a helper function for GetContainers
// which takes a []string of container IDs or names
func (r *Runtime) GetContainersByList(containers []string) ([]*Container, error) {
	var ctrs []*Container
	for _, inputContainer := range containers {
		ctr, err := r.LookupContainer(inputContainer)
		if err != nil {
			return ctrs, errors.Wrapf(err, "unable to lookup container %s", inputContainer)
		}
		ctrs = append(ctrs, ctr)
	}
	return ctrs, nil
}

// GetLatestContainer returns a container object of the latest created container.
func (r *Runtime) GetLatestContainer() (*Container, error) {
	lastCreatedIndex := -1
	var lastCreatedTime time.Time
	ctrs, err := r.GetAllContainers()
	if err != nil {
		return nil, errors.Wrapf(err, "unable to find latest container")
	}
	if len(ctrs) == 0 {
		return nil, ErrNoSuchCtr
	}
	for containerIndex, ctr := range ctrs {
		createdTime := ctr.config.CreatedTime
		if createdTime.After(lastCreatedTime) {
			lastCreatedTime = createdTime
			lastCreatedIndex = containerIndex
		}
	}
	return ctrs[lastCreatedIndex], nil
}

// RemoveContainersFromStorage attempt to remove containers from storage that do not exist in libpod database
func (r *Runtime) RemoveContainersFromStorage(ctrs []string) {
	for _, i := range ctrs {
		// if the container does not exist in database, attempt to remove it from storage
		if _, err := r.LookupContainer(i); err != nil && errors.Cause(err) == image.ErrNoSuchCtr {
			r.storageService.UnmountContainerImage(i, true)
			if err := r.storageService.DeleteContainer(i); err != nil && errors.Cause(err) != storage.ErrContainerUnknown {
				logrus.Errorf("Failed to remove container %q from storage: %s", i, err)
			}
		}
	}
}