summaryrefslogtreecommitdiff
path: root/pkg/specgen/generate/kube/kube.go
blob: fb563f9354718668c4931018119de7611c2917a6 (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
package kube

import (
	"context"
	"encoding/json"
	"fmt"
	"net"
	"strings"

	"github.com/containers/common/libimage"
	"github.com/containers/common/pkg/parse"
	"github.com/containers/common/pkg/secrets"
	ann "github.com/containers/podman/v3/pkg/annotations"
	"github.com/containers/podman/v3/pkg/specgen"
	"github.com/containers/podman/v3/pkg/specgen/generate"
	"github.com/containers/podman/v3/pkg/util"
	spec "github.com/opencontainers/runtime-spec/specs-go"
	"github.com/pkg/errors"
	v1 "k8s.io/api/core/v1"
	"k8s.io/apimachinery/pkg/api/resource"
)

func ToPodGen(ctx context.Context, podName string, podYAML *v1.PodTemplateSpec) (*specgen.PodSpecGenerator, error) {
	p := specgen.NewPodSpecGenerator()
	p.Name = podName
	p.Labels = podYAML.ObjectMeta.Labels
	// Kube pods must share {ipc, net, uts} by default
	p.SharedNamespaces = append(p.SharedNamespaces, "ipc")
	p.SharedNamespaces = append(p.SharedNamespaces, "net")
	p.SharedNamespaces = append(p.SharedNamespaces, "uts")
	// TODO we only configure Process namespace. We also need to account for Host{IPC,Network,PID}
	// which is not currently possible with pod create
	if podYAML.Spec.ShareProcessNamespace != nil && *podYAML.Spec.ShareProcessNamespace {
		p.SharedNamespaces = append(p.SharedNamespaces, "pid")
	}
	p.Hostname = podYAML.Spec.Hostname
	if p.Hostname == "" {
		p.Hostname = podName
	}
	if podYAML.Spec.HostNetwork {
		p.NetNS.NSMode = specgen.Host
	}
	if podYAML.Spec.HostAliases != nil {
		hosts := make([]string, 0, len(podYAML.Spec.HostAliases))
		for _, hostAlias := range podYAML.Spec.HostAliases {
			for _, host := range hostAlias.Hostnames {
				hosts = append(hosts, host+":"+hostAlias.IP)
			}
		}
		p.HostAdd = hosts
	}
	podPorts := getPodPorts(podYAML.Spec.Containers)
	p.PortMappings = podPorts

	if dnsConfig := podYAML.Spec.DNSConfig; dnsConfig != nil {
		// name servers
		if dnsServers := dnsConfig.Nameservers; len(dnsServers) > 0 {
			servers := make([]net.IP, 0)
			for _, server := range dnsServers {
				servers = append(servers, net.ParseIP(server))
			}
			p.DNSServer = servers
		}
		// search domains
		if domains := dnsConfig.Searches; len(domains) > 0 {
			p.DNSSearch = domains
		}
		// dns options
		if options := dnsConfig.Options; len(options) > 0 {
			dnsOptions := make([]string, 0)
			for _, opts := range options {
				d := opts.Name
				if opts.Value != nil {
					d += ":" + *opts.Value
				}
				dnsOptions = append(dnsOptions, d)
			}
		}
	}
	return p, nil
}

type CtrSpecGenOptions struct {
	// Container as read from the pod yaml
	Container v1.Container
	// Image available to use (pulled or found local)
	Image *libimage.Image
	// Volumes for all containers
	Volumes map[string]*KubeVolume
	// PodID of the parent pod
	PodID string
	// PodName of the parent pod
	PodName string
	// PodInfraID as the infrastructure container id
	PodInfraID string
	// ConfigMaps the configuration maps for environment variables
	ConfigMaps []v1.ConfigMap
	// SeccompPaths for finding the seccomp profile path
	SeccompPaths *KubeSeccompPaths
	// RestartPolicy defines the restart policy of the container
	RestartPolicy string
	// NetNSIsHost tells the container to use the host netns
	NetNSIsHost bool
	// SecretManager to access the secrets
	SecretsManager *secrets.SecretsManager
	// LogDriver which should be used for the container
	LogDriver string
	// Labels define key-value pairs of metadata
	Labels map[string]string
}

func ToSpecGen(ctx context.Context, opts *CtrSpecGenOptions) (*specgen.SpecGenerator, error) {
	s := specgen.NewSpecGenerator(opts.Container.Image, false)

	// pod name should be non-empty for Deployment objects to be able to create
	// multiple pods having containers with unique names
	if len(opts.PodName) < 1 {
		return nil, errors.Errorf("got empty pod name on container creation when playing kube")
	}

	s.Name = fmt.Sprintf("%s-%s", opts.PodName, opts.Container.Name)

	s.Terminal = opts.Container.TTY

	s.Pod = opts.PodID

	s.LogConfiguration = &specgen.LogConfig{
		Driver: opts.LogDriver,
	}

	setupSecurityContext(s, opts.Container)

	// Since we prefix the container name with pod name to work-around the uniqueness requirement,
	// the seccomp profile should reference the actual container name from the YAML
	// but apply to the containers with the prefixed name
	s.SeccompProfilePath = opts.SeccompPaths.FindForContainer(opts.Container.Name)

	s.ResourceLimits = &spec.LinuxResources{}
	milliCPU, err := quantityToInt64(opts.Container.Resources.Limits.Cpu())
	if err != nil {
		return nil, errors.Wrap(err, "Failed to set CPU quota")
	}
	if milliCPU > 0 {
		period, quota := util.CoresToPeriodAndQuota(float64(milliCPU) / 1000)
		s.ResourceLimits.CPU = &spec.LinuxCPU{
			Quota:  &quota,
			Period: &period,
		}
	}

	limit, err := quantityToInt64(opts.Container.Resources.Limits.Memory())
	if err != nil {
		return nil, errors.Wrap(err, "Failed to set memory limit")
	}

	memoryRes, err := quantityToInt64(opts.Container.Resources.Requests.Memory())
	if err != nil {
		return nil, errors.Wrap(err, "Failed to set memory reservation")
	}

	if limit > 0 || memoryRes > 0 {
		s.ResourceLimits.Memory = &spec.LinuxMemory{}
	}

	if limit > 0 {
		s.ResourceLimits.Memory.Limit = &limit
	}

	if memoryRes > 0 {
		s.ResourceLimits.Memory.Reservation = &memoryRes
	}

	// TODO: We don't understand why specgen does not take of this, but
	// integration tests clearly pointed out that it was required.
	imageData, err := opts.Image.Inspect(ctx, false)
	if err != nil {
		return nil, err
	}
	s.WorkDir = "/"
	// Entrypoint/Command handling is based off of
	// https://kubernetes.io/docs/tasks/inject-data-application/define-command-argument-container/#notes
	if imageData != nil && imageData.Config != nil {
		if imageData.Config.WorkingDir != "" {
			s.WorkDir = imageData.Config.WorkingDir
		}
		if s.User == "" {
			s.User = imageData.Config.User
		}

		exposed, err := generate.GenExposedPorts(imageData.Config.ExposedPorts)
		if err != nil {
			return nil, err
		}

		for k, v := range s.Expose {
			exposed[k] = v
		}
		s.Expose = exposed
		// Pull entrypoint and cmd from image
		s.Entrypoint = imageData.Config.Entrypoint
		s.Command = imageData.Config.Cmd
		s.Labels = imageData.Config.Labels
		if len(imageData.Config.StopSignal) > 0 {
			stopSignal, err := util.ParseSignal(imageData.Config.StopSignal)
			if err != nil {
				return nil, err
			}
			s.StopSignal = &stopSignal
		}
	}
	// If only the yaml.Command is specified, set it as the entrypoint and drop the image Cmd
	if len(opts.Container.Command) != 0 {
		s.Entrypoint = opts.Container.Command
		s.Command = []string{}
	}
	// Only override the cmd field if yaml.Args is specified
	// Keep the image entrypoint, or the yaml.command if specified
	if len(opts.Container.Args) != 0 {
		s.Command = opts.Container.Args
	}

	// FIXME,
	// we are currently ignoring imageData.Config.ExposedPorts
	if opts.Container.WorkingDir != "" {
		s.WorkDir = opts.Container.WorkingDir
	}

	annotations := make(map[string]string)
	if opts.PodInfraID != "" {
		annotations[ann.SandboxID] = opts.PodInfraID
		annotations[ann.ContainerType] = ann.ContainerTypeContainer
	}
	s.Annotations = annotations

	// Environment Variables
	envs := map[string]string{}
	for _, env := range imageData.Config.Env {
		keyval := strings.Split(env, "=")
		envs[keyval[0]] = keyval[1]
	}

	for _, env := range opts.Container.Env {
		value, err := envVarValue(env, opts)
		if err != nil {
			return nil, err
		}

		envs[env.Name] = value
	}
	for _, envFrom := range opts.Container.EnvFrom {
		cmEnvs, err := envVarsFrom(envFrom, opts)
		if err != nil {
			return nil, err
		}

		for k, v := range cmEnvs {
			envs[k] = v
		}
	}
	s.Env = envs

	for _, volume := range opts.Container.VolumeMounts {
		volumeSource, exists := opts.Volumes[volume.Name]
		if !exists {
			return nil, errors.Errorf("Volume mount %s specified for container but not configured in volumes", volume.Name)
		}

		dest, options, err := parseMountPath(volume.MountPath, volume.ReadOnly)
		if err != nil {
			return nil, err
		}

		switch volumeSource.Type {
		case KubeVolumeTypeBindMount:
			mount := spec.Mount{
				Destination: dest,
				Source:      volumeSource.Source,
				Type:        "bind",
				Options:     options,
			}
			s.Mounts = append(s.Mounts, mount)
		case KubeVolumeTypeNamed:
			namedVolume := specgen.NamedVolume{
				Dest:    dest,
				Name:    volumeSource.Source,
				Options: options,
			}
			s.Volumes = append(s.Volumes, &namedVolume)
		default:
			return nil, errors.Errorf("Unsupported volume source type")
		}
	}

	s.RestartPolicy = opts.RestartPolicy

	if opts.NetNSIsHost {
		s.NetNS.NSMode = specgen.Host
	}

	// Add labels that come from kube
	if len(s.Labels) == 0 {
		// If there are no labels, let's use the map that comes
		// from kube
		s.Labels = opts.Labels
	} else {
		// If there are already labels in the map, append the ones
		// obtained from kube
		for k, v := range opts.Labels {
			s.Labels[k] = v
		}
	}

	return s, nil
}

func parseMountPath(mountPath string, readOnly bool) (string, []string, error) {
	options := []string{}
	splitVol := strings.Split(mountPath, ":")
	if len(splitVol) > 2 {
		return "", options, errors.Errorf("%q incorrect volume format, should be ctr-dir[:option]", mountPath)
	}
	dest := splitVol[0]
	if len(splitVol) > 1 {
		options = strings.Split(splitVol[1], ",")
	}
	if err := parse.ValidateVolumeCtrDir(dest); err != nil {
		return "", options, errors.Wrapf(err, "error in parsing MountPath")
	}
	if readOnly {
		options = append(options, "ro")
	}
	return dest, options, nil
}

func setupSecurityContext(s *specgen.SpecGenerator, containerYAML v1.Container) {
	if containerYAML.SecurityContext == nil {
		return
	}
	if containerYAML.SecurityContext.ReadOnlyRootFilesystem != nil {
		s.ReadOnlyFilesystem = *containerYAML.SecurityContext.ReadOnlyRootFilesystem
	}
	if containerYAML.SecurityContext.Privileged != nil {
		s.Privileged = *containerYAML.SecurityContext.Privileged
	}

	if containerYAML.SecurityContext.AllowPrivilegeEscalation != nil {
		s.NoNewPrivileges = !*containerYAML.SecurityContext.AllowPrivilegeEscalation
	}

	if seopt := containerYAML.SecurityContext.SELinuxOptions; seopt != nil {
		if seopt.User != "" {
			s.SelinuxOpts = append(s.SelinuxOpts, fmt.Sprintf("user:%s", seopt.User))
		}
		if seopt.Role != "" {
			s.SelinuxOpts = append(s.SelinuxOpts, fmt.Sprintf("role:%s", seopt.Role))
		}
		if seopt.Type != "" {
			s.SelinuxOpts = append(s.SelinuxOpts, fmt.Sprintf("type:%s", seopt.Type))
		}
		if seopt.Level != "" {
			s.SelinuxOpts = append(s.SelinuxOpts, fmt.Sprintf("level:%s", seopt.Level))
		}
	}
	if caps := containerYAML.SecurityContext.Capabilities; caps != nil {
		for _, capability := range caps.Add {
			s.CapAdd = append(s.CapAdd, string(capability))
		}
		for _, capability := range caps.Drop {
			s.CapDrop = append(s.CapDrop, string(capability))
		}
	}
	if containerYAML.SecurityContext.RunAsUser != nil {
		s.User = fmt.Sprintf("%d", *containerYAML.SecurityContext.RunAsUser)
	}
	if containerYAML.SecurityContext.RunAsGroup != nil {
		if s.User == "" {
			s.User = "0"
		}
		s.User = fmt.Sprintf("%s:%d", s.User, *containerYAML.SecurityContext.RunAsGroup)
	}
}

func quantityToInt64(quantity *resource.Quantity) (int64, error) {
	if i, ok := quantity.AsInt64(); ok {
		return i, nil
	}

	if i, ok := quantity.AsDec().Unscaled(); ok {
		return i, nil
	}

	return 0, errors.Errorf("Quantity cannot be represented as int64: %v", quantity)
}

// read a k8s secret in JSON format from the secret manager
func k8sSecretFromSecretManager(name string, secretsManager *secrets.SecretsManager) (map[string][]byte, error) {
	_, jsonSecret, err := secretsManager.LookupSecretData(name)
	if err != nil {
		return nil, err
	}

	var secrets map[string][]byte
	if err := json.Unmarshal(jsonSecret, &secrets); err != nil {
		return nil, errors.Errorf("Secret %v is not valid JSON: %v", name, err)
	}
	return secrets, nil
}

// envVarsFrom returns all key-value pairs as env vars from a configMap or secret that matches the envFrom setting of a container
func envVarsFrom(envFrom v1.EnvFromSource, opts *CtrSpecGenOptions) (map[string]string, error) {
	envs := map[string]string{}

	if envFrom.ConfigMapRef != nil {
		cmRef := envFrom.ConfigMapRef
		err := errors.Errorf("Configmap %v not found", cmRef.Name)

		for _, c := range opts.ConfigMaps {
			if cmRef.Name == c.Name {
				envs = c.Data
				err = nil
				break
			}
		}

		if err != nil && (cmRef.Optional == nil || !*cmRef.Optional) {
			return nil, err
		}
	}

	if envFrom.SecretRef != nil {
		secRef := envFrom.SecretRef
		secret, err := k8sSecretFromSecretManager(secRef.Name, opts.SecretsManager)
		if err == nil {
			for k, v := range secret {
				envs[k] = string(v)
			}
		} else if secRef.Optional == nil || !*secRef.Optional {
			return nil, err
		}
	}

	return envs, nil
}

// envVarValue returns the environment variable value configured within the container's env setting.
// It gets the value from a configMap or secret if specified, otherwise returns env.Value
func envVarValue(env v1.EnvVar, opts *CtrSpecGenOptions) (string, error) {
	if env.ValueFrom != nil {
		if env.ValueFrom.ConfigMapKeyRef != nil {
			cmKeyRef := env.ValueFrom.ConfigMapKeyRef
			err := errors.Errorf("Cannot set env %v: configmap %v not found", env.Name, cmKeyRef.Name)

			for _, c := range opts.ConfigMaps {
				if cmKeyRef.Name == c.Name {
					if value, ok := c.Data[cmKeyRef.Key]; ok {
						return value, nil
					}
					err = errors.Errorf("Cannot set env %v: key %s not found in configmap %v", env.Name, cmKeyRef.Key, cmKeyRef.Name)
					break
				}
			}
			if cmKeyRef.Optional == nil || !*cmKeyRef.Optional {
				return "", err
			}
			return "", nil
		}

		if env.ValueFrom.SecretKeyRef != nil {
			secKeyRef := env.ValueFrom.SecretKeyRef
			secret, err := k8sSecretFromSecretManager(secKeyRef.Name, opts.SecretsManager)
			if err == nil {
				if val, ok := secret[secKeyRef.Key]; ok {
					return string(val), nil
				}
				err = errors.Errorf("Secret %v has not %v key", secKeyRef.Name, secKeyRef.Key)
			}
			if secKeyRef.Optional == nil || !*secKeyRef.Optional {
				return "", errors.Errorf("Cannot set env %v: %v", env.Name, err)
			}
			return "", nil
		}
	}

	return env.Value, nil
}

// getPodPorts converts a slice of kube container descriptions to an
// array of portmapping
func getPodPorts(containers []v1.Container) []specgen.PortMapping {
	var infraPorts []specgen.PortMapping
	for _, container := range containers {
		for _, p := range container.Ports {
			if p.HostPort != 0 && p.ContainerPort == 0 {
				p.ContainerPort = p.HostPort
			}
			if p.Protocol == "" {
				p.Protocol = "tcp"
			}
			portBinding := specgen.PortMapping{
				HostPort:      uint16(p.HostPort),
				ContainerPort: uint16(p.ContainerPort),
				Protocol:      strings.ToLower(string(p.Protocol)),
				HostIP:        p.HostIP,
			}
			// only hostPort is utilized in podman context, all container ports
			// are accessible inside the shared network namespace
			if p.HostPort != 0 {
				infraPorts = append(infraPorts, portBinding)
			}
		}
	}
	return infraPorts
}