-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathclient_local_windows.go
1376 lines (1223 loc) · 42.6 KB
/
client_local_windows.go
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
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
package libcontainerd // import "github.com/docker/docker/libcontainerd"
import (
"context"
"encoding/json"
"fmt"
"io/ioutil"
"os"
"path"
"path/filepath"
"regexp"
"strings"
"sync"
"syscall"
"time"
"github.com/Microsoft/hcsshim"
opengcs "github.com/Microsoft/opengcs/client"
"github.com/containerd/containerd"
"github.com/containerd/containerd/cio"
"github.com/docker/docker/pkg/sysinfo"
"github.com/docker/docker/pkg/system"
specs "github.com/opencontainers/runtime-spec/specs-go"
"github.com/pkg/errors"
"github.com/sirupsen/logrus"
"golang.org/x/sys/windows"
)
const InitProcessName = "init"
type process struct {
id string
pid int
hcsProcess hcsshim.Process
}
type container struct {
sync.Mutex
// The ociSpec is required, as client.Create() needs a spec, but can
// be called from the RestartManager context which does not otherwise
// have access to the Spec
ociSpec *specs.Spec
isWindows bool
hcsContainer hcsshim.Container
id string
status Status
exitedAt time.Time
exitCode uint32
waitCh chan struct{}
init *process
execs map[string]*process
terminateInvoked bool
}
// Win32 error codes that are used for various workarounds
// These really should be ALL_CAPS to match golangs syscall library and standard
// Win32 error conventions, but golint insists on CamelCase.
const (
CoEClassstring = syscall.Errno(0x800401F3) // Invalid class string
ErrorNoNetwork = syscall.Errno(1222) // The network is not present or not started
ErrorBadPathname = syscall.Errno(161) // The specified path is invalid
ErrorInvalidObject = syscall.Errno(0x800710D8) // The object identifier does not represent a valid object
)
// defaultOwner is a tag passed to HCS to allow it to differentiate between
// container creator management stacks. We hard code "docker" in the case
// of docker.
const defaultOwner = "docker"
type client struct {
sync.Mutex
stateDir string
backend Backend
logger *logrus.Entry
eventQ queue
containers map[string]*container
}
// NewClient creates a new local executor for windows
func NewClient(ctx context.Context, cli *containerd.Client, stateDir, ns string, b Backend) (Client, error) {
c := &client{
stateDir: stateDir,
backend: b,
logger: logrus.WithField("module", "libcontainerd").WithField("module", "libcontainerd").WithField("namespace", ns),
containers: make(map[string]*container),
}
return c, nil
}
func (c *client) Version(ctx context.Context) (containerd.Version, error) {
return containerd.Version{}, errors.New("not implemented on Windows")
}
// Create is the entrypoint to create a container from a spec.
// Table below shows the fields required for HCS JSON calling parameters,
// where if not populated, is omitted.
// +-----------------+--------------------------------------------+---------------------------------------------------+
// | | Isolation=Process | Isolation=Hyper-V |
// +-----------------+--------------------------------------------+---------------------------------------------------+
// | VolumePath | \\?\\Volume{GUIDa} | |
// | LayerFolderPath | %root%\windowsfilter\containerID | |
// | Layers[] | ID=GUIDb;Path=%root%\windowsfilter\layerID | ID=GUIDb;Path=%root%\windowsfilter\layerID |
// | HvRuntime | | ImagePath=%root%\BaseLayerID\UtilityVM |
// +-----------------+--------------------------------------------+---------------------------------------------------+
//
// Isolation=Process example:
//
// {
// "SystemType": "Container",
// "Name": "5e0055c814a6005b8e57ac59f9a522066e0af12b48b3c26a9416e23907698776",
// "Owner": "docker",
// "VolumePath": "\\\\\\\\?\\\\Volume{66d1ef4c-7a00-11e6-8948-00155ddbef9d}",
// "IgnoreFlushesDuringBoot": true,
// "LayerFolderPath": "C:\\\\control\\\\windowsfilter\\\\5e0055c814a6005b8e57ac59f9a522066e0af12b48b3c26a9416e23907698776",
// "Layers": [{
// "ID": "18955d65-d45a-557b-bf1c-49d6dfefc526",
// "Path": "C:\\\\control\\\\windowsfilter\\\\65bf96e5760a09edf1790cb229e2dfb2dbd0fcdc0bf7451bae099106bfbfea0c"
// }],
// "HostName": "5e0055c814a6",
// "MappedDirectories": [],
// "HvPartition": false,
// "EndpointList": ["eef2649d-bb17-4d53-9937-295a8efe6f2c"],
//}
//
// Isolation=Hyper-V example:
//
//{
// "SystemType": "Container",
// "Name": "475c2c58933b72687a88a441e7e0ca4bd72d76413c5f9d5031fee83b98f6045d",
// "Owner": "docker",
// "IgnoreFlushesDuringBoot": true,
// "Layers": [{
// "ID": "18955d65-d45a-557b-bf1c-49d6dfefc526",
// "Path": "C:\\\\control\\\\windowsfilter\\\\65bf96e5760a09edf1790cb229e2dfb2dbd0fcdc0bf7451bae099106bfbfea0c"
// }],
// "HostName": "475c2c58933b",
// "MappedDirectories": [],
// "HvPartition": true,
// "EndpointList": ["e1bb1e61-d56f-405e-b75d-fd520cefa0cb"],
// "DNSSearchList": "a.com,b.com,c.com",
// "HvRuntime": {
// "ImagePath": "C:\\\\control\\\\windowsfilter\\\\65bf96e5760a09edf1790cb229e2dfb2dbd0fcdc0bf7451bae099106bfbfea0c\\\\UtilityVM"
// },
//}
func (c *client) Create(_ context.Context, id string, spec *specs.Spec, runtimeOptions interface{}) error {
if ctr := c.getContainer(id); ctr != nil {
return errors.WithStack(newConflictError("id already in use"))
}
// spec.Linux must be nil for Windows containers, but spec.Windows
// will be filled in regardless of container platform. This is a
// temporary workaround due to LCOW requiring layer folder paths,
// which are stored under spec.Windows.
//
// TODO: @darrenstahlmsft fix this once the OCI spec is updated to
// support layer folder paths for LCOW
if spec.Linux == nil {
return c.createWindows(id, spec, runtimeOptions)
}
return c.createLinux(id, spec, runtimeOptions)
}
func (c *client) createWindows(id string, spec *specs.Spec, runtimeOptions interface{}) error {
logger := c.logger.WithField("container", id)
configuration := &hcsshim.ContainerConfig{
SystemType: "Container",
Name: id,
Owner: defaultOwner,
IgnoreFlushesDuringBoot: spec.Windows.IgnoreFlushesDuringBoot,
HostName: spec.Hostname,
HvPartition: false,
}
c.extractResourcesFromSpec(spec, configuration)
if spec.Windows.Resources != nil {
if spec.Windows.Resources.Storage != nil {
if spec.Windows.Resources.Storage.Bps != nil {
configuration.StorageBandwidthMaximum = *spec.Windows.Resources.Storage.Bps
}
if spec.Windows.Resources.Storage.Iops != nil {
configuration.StorageIOPSMaximum = *spec.Windows.Resources.Storage.Iops
}
}
}
if spec.Windows.HyperV != nil {
configuration.HvPartition = true
}
if spec.Windows.Network != nil {
configuration.EndpointList = spec.Windows.Network.EndpointList
configuration.AllowUnqualifiedDNSQuery = spec.Windows.Network.AllowUnqualifiedDNSQuery
if spec.Windows.Network.DNSSearchList != nil {
configuration.DNSSearchList = strings.Join(spec.Windows.Network.DNSSearchList, ",")
}
configuration.NetworkSharedContainerName = spec.Windows.Network.NetworkSharedContainerName
}
if cs, ok := spec.Windows.CredentialSpec.(string); ok {
configuration.Credentials = cs
}
// We must have least two layers in the spec, the bottom one being a
// base image, the top one being the RW layer.
if spec.Windows.LayerFolders == nil || len(spec.Windows.LayerFolders) < 2 {
return fmt.Errorf("OCI spec is invalid - at least two LayerFolders must be supplied to the runtime")
}
// Strip off the top-most layer as that's passed in separately to HCS
configuration.LayerFolderPath = spec.Windows.LayerFolders[len(spec.Windows.LayerFolders)-1]
layerFolders := spec.Windows.LayerFolders[:len(spec.Windows.LayerFolders)-1]
if configuration.HvPartition {
// We don't currently support setting the utility VM image explicitly.
// TODO @swernli/jhowardmsft circa RS5, this may be re-locatable.
if spec.Windows.HyperV.UtilityVMPath != "" {
return errors.New("runtime does not support an explicit utility VM path for Hyper-V containers")
}
// Find the upper-most utility VM image.
var uvmImagePath string
for _, path := range layerFolders {
fullPath := filepath.Join(path, "UtilityVM")
_, err := os.Stat(fullPath)
if err == nil {
uvmImagePath = fullPath
break
}
if !os.IsNotExist(err) {
return err
}
}
if uvmImagePath == "" {
return errors.New("utility VM image could not be found")
}
configuration.HvRuntime = &hcsshim.HvRuntime{ImagePath: uvmImagePath}
if spec.Root.Path != "" {
return errors.New("OCI spec is invalid - Root.Path must be omitted for a Hyper-V container")
}
} else {
const volumeGUIDRegex = `^\\\\\?\\(Volume)\{{0,1}[0-9a-fA-F]{8}\-[0-9a-fA-F]{4}\-[0-9a-fA-F]{4}\-[0-9a-fA-F]{4}\-[0-9a-fA-F]{12}(\}){0,1}\}\\$`
if _, err := regexp.MatchString(volumeGUIDRegex, spec.Root.Path); err != nil {
return fmt.Errorf(`OCI spec is invalid - Root.Path '%s' must be a volume GUID path in the format '\\?\Volume{GUID}\'`, spec.Root.Path)
}
// HCS API requires the trailing backslash to be removed
configuration.VolumePath = spec.Root.Path[:len(spec.Root.Path)-1]
}
if spec.Root.Readonly {
return errors.New(`OCI spec is invalid - Root.Readonly must not be set on Windows`)
}
for _, layerPath := range layerFolders {
_, filename := filepath.Split(layerPath)
g, err := hcsshim.NameToGuid(filename)
if err != nil {
return err
}
configuration.Layers = append(configuration.Layers, hcsshim.Layer{
ID: g.ToString(),
Path: layerPath,
})
}
// Add the mounts (volumes, bind mounts etc) to the structure
var mds []hcsshim.MappedDir
var mps []hcsshim.MappedPipe
for _, mount := range spec.Mounts {
const pipePrefix = `\\.\pipe\`
if mount.Type != "" {
return fmt.Errorf("OCI spec is invalid - Mount.Type '%s' must not be set", mount.Type)
}
if strings.HasPrefix(mount.Destination, pipePrefix) {
mp := hcsshim.MappedPipe{
HostPath: mount.Source,
ContainerPipeName: mount.Destination[len(pipePrefix):],
}
mps = append(mps, mp)
} else {
md := hcsshim.MappedDir{
HostPath: mount.Source,
ContainerPath: mount.Destination,
ReadOnly: false,
}
for _, o := range mount.Options {
if strings.ToLower(o) == "ro" {
md.ReadOnly = true
}
}
mds = append(mds, md)
}
}
configuration.MappedDirectories = mds
if len(mps) > 0 && system.GetOSVersion().Build < 16299 { // RS3
return errors.New("named pipe mounts are not supported on this version of Windows")
}
configuration.MappedPipes = mps
if len(spec.Windows.Devices) > 0 {
// Add any device assignments
if configuration.HvPartition {
return errors.New("device assignment is not supported for HyperV containers")
}
if system.GetOSVersion().Build < 17763 { // RS5
return errors.New("device assignment requires Windows builds RS5 (17763+) or later")
}
for _, d := range spec.Windows.Devices {
configuration.AssignedDevices = append(configuration.AssignedDevices, hcsshim.AssignedDevice{InterfaceClassGUID: d.ID})
}
}
hcsContainer, err := hcsshim.CreateContainer(id, configuration)
if err != nil {
return err
}
// Construct a container object for calling start on it.
ctr := &container{
id: id,
execs: make(map[string]*process),
isWindows: true,
ociSpec: spec,
hcsContainer: hcsContainer,
status: StatusCreated,
waitCh: make(chan struct{}),
}
logger.Debug("starting container")
if err = hcsContainer.Start(); err != nil {
c.logger.WithError(err).Error("failed to start container")
ctr.Lock()
if err := c.terminateContainer(ctr); err != nil {
c.logger.WithError(err).Error("failed to cleanup after a failed Start")
} else {
c.logger.Debug("cleaned up after failed Start by calling Terminate")
}
ctr.Unlock()
return err
}
c.Lock()
c.containers[id] = ctr
c.Unlock()
logger.Debug("createWindows() completed successfully")
return nil
}
func (c *client) createLinux(id string, spec *specs.Spec, runtimeOptions interface{}) error {
logrus.Debugf("libcontainerd: createLinux(): containerId %s ", id)
logger := c.logger.WithField("container", id)
if runtimeOptions == nil {
return fmt.Errorf("lcow option must be supplied to the runtime")
}
lcowConfig, ok := runtimeOptions.(*opengcs.Config)
if !ok {
return fmt.Errorf("lcow option must be supplied to the runtime")
}
configuration := &hcsshim.ContainerConfig{
HvPartition: true,
Name: id,
SystemType: "container",
ContainerType: "linux",
Owner: defaultOwner,
TerminateOnLastHandleClosed: true,
}
if lcowConfig.ActualMode == opengcs.ModeActualVhdx {
configuration.HvRuntime = &hcsshim.HvRuntime{
ImagePath: lcowConfig.Vhdx,
BootSource: "Vhd",
WritableBootSource: false,
}
} else {
configuration.HvRuntime = &hcsshim.HvRuntime{
ImagePath: lcowConfig.KirdPath,
LinuxKernelFile: lcowConfig.KernelFile,
LinuxInitrdFile: lcowConfig.InitrdFile,
LinuxBootParameters: lcowConfig.BootParameters,
}
}
if spec.Windows == nil {
return fmt.Errorf("spec.Windows must not be nil for LCOW containers")
}
c.extractResourcesFromSpec(spec, configuration)
// We must have least one layer in the spec
if spec.Windows.LayerFolders == nil || len(spec.Windows.LayerFolders) == 0 {
return fmt.Errorf("OCI spec is invalid - at least one LayerFolders must be supplied to the runtime")
}
// Strip off the top-most layer as that's passed in separately to HCS
configuration.LayerFolderPath = spec.Windows.LayerFolders[len(spec.Windows.LayerFolders)-1]
layerFolders := spec.Windows.LayerFolders[:len(spec.Windows.LayerFolders)-1]
for _, layerPath := range layerFolders {
_, filename := filepath.Split(layerPath)
g, err := hcsshim.NameToGuid(filename)
if err != nil {
return err
}
configuration.Layers = append(configuration.Layers, hcsshim.Layer{
ID: g.ToString(),
Path: filepath.Join(layerPath, "layer.vhd"),
})
}
if spec.Windows.Network != nil {
configuration.EndpointList = spec.Windows.Network.EndpointList
configuration.AllowUnqualifiedDNSQuery = spec.Windows.Network.AllowUnqualifiedDNSQuery
if spec.Windows.Network.DNSSearchList != nil {
configuration.DNSSearchList = strings.Join(spec.Windows.Network.DNSSearchList, ",")
}
configuration.NetworkSharedContainerName = spec.Windows.Network.NetworkSharedContainerName
}
// Add the mounts (volumes, bind mounts etc) to the structure. We have to do
// some translation for both the mapped directories passed into HCS and in
// the spec.
//
// For HCS, we only pass in the mounts from the spec which are type "bind".
// Further, the "ContainerPath" field (which is a little mis-leadingly
// named when it applies to the utility VM rather than the container in the
// utility VM) is moved to under /tmp/gcs/<ID>/binds, where this is passed
// by the caller through a 'uvmpath' option.
//
// We do similar translation for the mounts in the spec by stripping out
// the uvmpath option, and translating the Source path to the location in the
// utility VM calculated above.
//
// From inside the utility VM, you would see a 9p mount such as in the following
// where a host folder has been mapped to /target. The line with /tmp/gcs/<ID>/binds
// specifically:
//
// / # mount
// rootfs on / type rootfs (rw,size=463736k,nr_inodes=115934)
// proc on /proc type proc (rw,relatime)
// sysfs on /sys type sysfs (rw,relatime)
// udev on /dev type devtmpfs (rw,relatime,size=498100k,nr_inodes=124525,mode=755)
// tmpfs on /run type tmpfs (rw,relatime)
// cgroup on /sys/fs/cgroup type cgroup (rw,relatime,cpuset,cpu,cpuacct,blkio,memory,devices,freezer,net_cls,perf_event,net_prio,hugetlb,pids,rdma)
// mqueue on /dev/mqueue type mqueue (rw,relatime)
// devpts on /dev/pts type devpts (rw,relatime,mode=600,ptmxmode=000)
// /binds/b3ea9126d67702173647ece2744f7c11181c0150e9890fc9a431849838033edc/target on /binds/b3ea9126d67702173647ece2744f7c11181c0150e9890fc9a431849838033edc/target type 9p (rw,sync,dirsync,relatime,trans=fd,rfdno=6,wfdno=6)
// /dev/pmem0 on /tmp/gcs/b3ea9126d67702173647ece2744f7c11181c0150e9890fc9a431849838033edc/layer0 type ext4 (ro,relatime,block_validity,delalloc,norecovery,barrier,dax,user_xattr,acl)
// /dev/sda on /tmp/gcs/b3ea9126d67702173647ece2744f7c11181c0150e9890fc9a431849838033edc/scratch type ext4 (rw,relatime,block_validity,delalloc,barrier,user_xattr,acl)
// overlay on /tmp/gcs/b3ea9126d67702173647ece2744f7c11181c0150e9890fc9a431849838033edc/rootfs type overlay (rw,relatime,lowerdir=/tmp/base/:/tmp/gcs/b3ea9126d67702173647ece2744f7c11181c0150e9890fc9a431849838033edc/layer0,upperdir=/tmp/gcs/b3ea9126d67702173647ece2744f7c11181c0150e9890fc9a431849838033edc/scratch/upper,workdir=/tmp/gcs/b3ea9126d67702173647ece2744f7c11181c0150e9890fc9a431849838033edc/scratch/work)
//
// /tmp/gcs/b3ea9126d67702173647ece2744f7c11181c0150e9890fc9a431849838033edc # ls -l
// total 16
// drwx------ 3 0 0 60 Sep 7 18:54 binds
// -rw-r--r-- 1 0 0 3345 Sep 7 18:54 config.json
// drwxr-xr-x 10 0 0 4096 Sep 6 17:26 layer0
// drwxr-xr-x 1 0 0 4096 Sep 7 18:54 rootfs
// drwxr-xr-x 5 0 0 4096 Sep 7 18:54 scratch
//
// /tmp/gcs/b3ea9126d67702173647ece2744f7c11181c0150e9890fc9a431849838033edc # ls -l binds
// total 0
// drwxrwxrwt 2 0 0 4096 Sep 7 16:51 target
mds := []hcsshim.MappedDir{}
specMounts := []specs.Mount{}
for _, mount := range spec.Mounts {
specMount := mount
if mount.Type == "bind" {
// Strip out the uvmpath from the options
updatedOptions := []string{}
uvmPath := ""
readonly := false
for _, opt := range mount.Options {
dropOption := false
elements := strings.SplitN(opt, "=", 2)
switch elements[0] {
case "uvmpath":
uvmPath = elements[1]
dropOption = true
case "rw":
case "ro":
readonly = true
case "rbind":
default:
return fmt.Errorf("unsupported option %q", opt)
}
if !dropOption {
updatedOptions = append(updatedOptions, opt)
}
}
mount.Options = updatedOptions
if uvmPath == "" {
return fmt.Errorf("no uvmpath for bind mount %+v", mount)
}
md := hcsshim.MappedDir{
HostPath: mount.Source,
ContainerPath: path.Join(uvmPath, mount.Destination),
CreateInUtilityVM: true,
ReadOnly: readonly,
}
// If we are 1803/RS4+ enable LinuxMetadata support by default
if system.GetOSVersion().Build >= 17134 {
md.LinuxMetadata = true
}
mds = append(mds, md)
specMount.Source = path.Join(uvmPath, mount.Destination)
}
specMounts = append(specMounts, specMount)
}
configuration.MappedDirectories = mds
hcsContainer, err := hcsshim.CreateContainer(id, configuration)
if err != nil {
return err
}
spec.Mounts = specMounts
// Construct a container object for calling start on it.
ctr := &container{
id: id,
execs: make(map[string]*process),
isWindows: false,
ociSpec: spec,
hcsContainer: hcsContainer,
status: StatusCreated,
waitCh: make(chan struct{}),
}
// Start the container.
logger.Debug("starting container")
if err = hcsContainer.Start(); err != nil {
c.logger.WithError(err).Error("failed to start container")
ctr.debugGCS()
ctr.Lock()
if err := c.terminateContainer(ctr); err != nil {
c.logger.WithError(err).Error("failed to cleanup after a failed Start")
} else {
c.logger.Debug("cleaned up after failed Start by calling Terminate")
}
ctr.Unlock()
return err
}
ctr.debugGCS()
c.Lock()
c.containers[id] = ctr
c.Unlock()
c.eventQ.append(id, func() {
ei := EventInfo{
ContainerID: id,
}
c.logger.WithFields(logrus.Fields{
"container": ctr.id,
"event": EventCreate,
}).Info("sending event")
err := c.backend.ProcessEvent(id, EventCreate, ei)
if err != nil {
c.logger.WithError(err).WithFields(logrus.Fields{
"container": id,
"event": EventCreate,
}).Error("failed to process event")
}
})
logger.Debug("createLinux() completed successfully")
return nil
}
func (c *client) extractResourcesFromSpec(spec *specs.Spec, configuration *hcsshim.ContainerConfig) {
if spec.Windows.Resources != nil {
if spec.Windows.Resources.CPU != nil {
if spec.Windows.Resources.CPU.Count != nil {
// This check is being done here rather than in adaptContainerSettings
// because we don't want to update the HostConfig in case this container
// is moved to a host with more CPUs than this one.
cpuCount := *spec.Windows.Resources.CPU.Count
hostCPUCount := uint64(sysinfo.NumCPU())
if cpuCount > hostCPUCount {
c.logger.Warnf("Changing requested CPUCount of %d to current number of processors, %d", cpuCount, hostCPUCount)
cpuCount = hostCPUCount
}
configuration.ProcessorCount = uint32(cpuCount)
}
if spec.Windows.Resources.CPU.Shares != nil {
configuration.ProcessorWeight = uint64(*spec.Windows.Resources.CPU.Shares)
}
if spec.Windows.Resources.CPU.Maximum != nil {
configuration.ProcessorMaximum = int64(*spec.Windows.Resources.CPU.Maximum)
}
}
if spec.Windows.Resources.Memory != nil {
if spec.Windows.Resources.Memory.Limit != nil {
configuration.MemoryMaximumInMB = int64(*spec.Windows.Resources.Memory.Limit) / 1024 / 1024
}
}
}
}
func (c *client) Start(_ context.Context, id, _ string, withStdin bool, attachStdio StdioCallback) (int, error) {
ctr := c.getContainer(id)
switch {
case ctr == nil:
return -1, errors.WithStack(newNotFoundError("no such container"))
case ctr.init != nil:
return -1, errors.WithStack(newConflictError("container already started"))
}
logger := c.logger.WithField("container", id)
// Note we always tell HCS to create stdout as it's required
// regardless of '-i' or '-t' options, so that docker can always grab
// the output through logs. We also tell HCS to always create stdin,
// even if it's not used - it will be closed shortly. Stderr is only
// created if it we're not -t.
var (
emulateConsole bool
createStdErrPipe bool
)
if ctr.ociSpec.Process != nil {
emulateConsole = ctr.ociSpec.Process.Terminal
createStdErrPipe = !ctr.ociSpec.Process.Terminal
}
createProcessParms := &hcsshim.ProcessConfig{
EmulateConsole: emulateConsole,
WorkingDirectory: ctr.ociSpec.Process.Cwd,
CreateStdInPipe: true,
CreateStdOutPipe: true,
CreateStdErrPipe: createStdErrPipe,
}
if ctr.ociSpec.Process != nil && ctr.ociSpec.Process.ConsoleSize != nil {
createProcessParms.ConsoleSize[0] = uint(ctr.ociSpec.Process.ConsoleSize.Height)
createProcessParms.ConsoleSize[1] = uint(ctr.ociSpec.Process.ConsoleSize.Width)
}
// Configure the environment for the process
createProcessParms.Environment = setupEnvironmentVariables(ctr.ociSpec.Process.Env)
if ctr.isWindows {
createProcessParms.CommandLine = strings.Join(ctr.ociSpec.Process.Args, " ")
} else {
createProcessParms.CommandArgs = ctr.ociSpec.Process.Args
}
createProcessParms.User = ctr.ociSpec.Process.User.Username
// LCOW requires the raw OCI spec passed through HCS and onwards to
// GCS for the utility VM.
if !ctr.isWindows {
ociBuf, err := json.Marshal(ctr.ociSpec)
if err != nil {
return -1, err
}
ociRaw := json.RawMessage(ociBuf)
createProcessParms.OCISpecification = &ociRaw
}
ctr.Lock()
defer ctr.Unlock()
// Start the command running in the container.
newProcess, err := ctr.hcsContainer.CreateProcess(createProcessParms)
if err != nil {
logger.WithError(err).Error("CreateProcess() failed")
return -1, err
}
defer func() {
if err != nil {
if err := newProcess.Kill(); err != nil {
logger.WithError(err).Error("failed to kill process")
}
go func() {
if err := newProcess.Wait(); err != nil {
logger.WithError(err).Error("failed to wait for process")
}
if err := newProcess.Close(); err != nil {
logger.WithError(err).Error("failed to clean process resources")
}
}()
}
}()
p := &process{
hcsProcess: newProcess,
id: InitProcessName,
pid: newProcess.Pid(),
}
logger.WithField("pid", p.pid).Debug("init process started")
dio, err := newIOFromProcess(newProcess, ctr.ociSpec.Process.Terminal)
if err != nil {
logger.WithError(err).Error("failed to get stdio pipes")
return -1, err
}
_, err = attachStdio(dio)
if err != nil {
logger.WithError(err).Error("failed to attache stdio")
return -1, err
}
ctr.status = StatusRunning
ctr.init = p
// Spin up a go routine waiting for exit to handle cleanup
go c.reapProcess(ctr, p)
// Generate the associated event
c.eventQ.append(id, func() {
ei := EventInfo{
ContainerID: id,
ProcessID: InitProcessName,
Pid: uint32(p.pid),
}
c.logger.WithFields(logrus.Fields{
"container": ctr.id,
"event": EventStart,
"event-info": ei,
}).Info("sending event")
err := c.backend.ProcessEvent(ei.ContainerID, EventStart, ei)
if err != nil {
c.logger.WithError(err).WithFields(logrus.Fields{
"container": id,
"event": EventStart,
"event-info": ei,
}).Error("failed to process event")
}
})
logger.Debug("start() completed")
return p.pid, nil
}
func newIOFromProcess(newProcess hcsshim.Process, terminal bool) (*cio.DirectIO, error) {
stdin, stdout, stderr, err := newProcess.Stdio()
if err != nil {
return nil, err
}
dio := cio.NewDirectIO(createStdInCloser(stdin, newProcess), nil, nil, terminal)
// Convert io.ReadClosers to io.Readers
if stdout != nil {
dio.Stdout = ioutil.NopCloser(&autoClosingReader{ReadCloser: stdout})
}
if stderr != nil {
dio.Stderr = ioutil.NopCloser(&autoClosingReader{ReadCloser: stderr})
}
return dio, nil
}
// Exec adds a process in an running container
func (c *client) Exec(ctx context.Context, containerID, processID string, spec *specs.Process, withStdin bool, attachStdio StdioCallback) (int, error) {
ctr := c.getContainer(containerID)
switch {
case ctr == nil:
return -1, errors.WithStack(newNotFoundError("no such container"))
case ctr.hcsContainer == nil:
return -1, errors.WithStack(newInvalidParameterError("container is not running"))
case ctr.execs != nil && ctr.execs[processID] != nil:
return -1, errors.WithStack(newConflictError("id already in use"))
}
logger := c.logger.WithFields(logrus.Fields{
"container": containerID,
"exec": processID,
})
// Note we always tell HCS to
// create stdout as it's required regardless of '-i' or '-t' options, so that
// docker can always grab the output through logs. We also tell HCS to always
// create stdin, even if it's not used - it will be closed shortly. Stderr
// is only created if it we're not -t.
createProcessParms := hcsshim.ProcessConfig{
CreateStdInPipe: true,
CreateStdOutPipe: true,
CreateStdErrPipe: !spec.Terminal,
}
if spec.Terminal {
createProcessParms.EmulateConsole = true
if spec.ConsoleSize != nil {
createProcessParms.ConsoleSize[0] = uint(spec.ConsoleSize.Height)
createProcessParms.ConsoleSize[1] = uint(spec.ConsoleSize.Width)
}
}
// Take working directory from the process to add if it is defined,
// otherwise take from the first process.
if spec.Cwd != "" {
createProcessParms.WorkingDirectory = spec.Cwd
} else {
createProcessParms.WorkingDirectory = ctr.ociSpec.Process.Cwd
}
// Configure the environment for the process
createProcessParms.Environment = setupEnvironmentVariables(spec.Env)
if ctr.isWindows {
createProcessParms.CommandLine = strings.Join(spec.Args, " ")
} else {
createProcessParms.CommandArgs = spec.Args
}
createProcessParms.User = spec.User.Username
logger.Debugf("exec commandLine: %s", createProcessParms.CommandLine)
// Start the command running in the container.
newProcess, err := ctr.hcsContainer.CreateProcess(&createProcessParms)
if err != nil {
logger.WithError(err).Errorf("exec's CreateProcess() failed")
return -1, err
}
pid := newProcess.Pid()
defer func() {
if err != nil {
if err := newProcess.Kill(); err != nil {
logger.WithError(err).Error("failed to kill process")
}
go func() {
if err := newProcess.Wait(); err != nil {
logger.WithError(err).Error("failed to wait for process")
}
if err := newProcess.Close(); err != nil {
logger.WithError(err).Error("failed to clean process resources")
}
}()
}
}()
dio, err := newIOFromProcess(newProcess, spec.Terminal)
if err != nil {
logger.WithError(err).Error("failed to get stdio pipes")
return -1, err
}
// Tell the engine to attach streams back to the client
_, err = attachStdio(dio)
if err != nil {
return -1, err
}
p := &process{
id: processID,
pid: pid,
hcsProcess: newProcess,
}
// Add the process to the container's list of processes
ctr.Lock()
ctr.execs[processID] = p
ctr.Unlock()
// Spin up a go routine waiting for exit to handle cleanup
go c.reapProcess(ctr, p)
c.eventQ.append(ctr.id, func() {
ei := EventInfo{
ContainerID: ctr.id,
ProcessID: p.id,
Pid: uint32(p.pid),
}
c.logger.WithFields(logrus.Fields{
"container": ctr.id,
"event": EventExecAdded,
"event-info": ei,
}).Info("sending event")
err := c.backend.ProcessEvent(ctr.id, EventExecAdded, ei)
if err != nil {
c.logger.WithError(err).WithFields(logrus.Fields{
"container": ctr.id,
"event": EventExecAdded,
"event-info": ei,
}).Error("failed to process event")
}
err = c.backend.ProcessEvent(ctr.id, EventExecStarted, ei)
if err != nil {
c.logger.WithError(err).WithFields(logrus.Fields{
"container": ctr.id,
"event": EventExecStarted,
"event-info": ei,
}).Error("failed to process event")
}
})
return pid, nil
}
// Signal handles `docker stop` on Windows. While Linux has support for
// the full range of signals, signals aren't really implemented on Windows.
// We fake supporting regular stop and -9 to force kill.
func (c *client) SignalProcess(_ context.Context, containerID, processID string, signal int) error {
ctr, p, err := c.getProcess(containerID, processID)
if err != nil {
return err
}
logger := c.logger.WithFields(logrus.Fields{
"container": containerID,
"process": processID,
"pid": p.pid,
"signal": signal,
})
logger.Debug("Signal()")
if processID == InitProcessName {
if syscall.Signal(signal) == syscall.SIGKILL {
// Terminate the compute system
ctr.Lock()
ctr.terminateInvoked = true
if err := ctr.hcsContainer.Terminate(); err != nil {
if !hcsshim.IsPending(err) {
logger.WithError(err).Error("failed to terminate hccshim container")
}
}
ctr.Unlock()
} else {
// Shut down the container
if err := ctr.hcsContainer.Shutdown(); err != nil {
if !hcsshim.IsPending(err) && !hcsshim.IsAlreadyStopped(err) {
// ignore errors
logger.WithError(err).Error("failed to shutdown hccshim container")
}
}
}
} else {
return p.hcsProcess.Kill()
}
return nil
}
// Resize handles a CLI event to resize an interactive docker run or docker
// exec window.
func (c *client) ResizeTerminal(_ context.Context, containerID, processID string, width, height int) error {
_, p, err := c.getProcess(containerID, processID)
if err != nil {
return err
}
c.logger.WithFields(logrus.Fields{
"container": containerID,
"process": processID,
"height": height,
"width": width,
"pid": p.pid,
}).Debug("resizing")
return p.hcsProcess.ResizeConsole(uint16(width), uint16(height))
}
func (c *client) CloseStdin(_ context.Context, containerID, processID string) error {
_, p, err := c.getProcess(containerID, processID)
if err != nil {
return err
}
return p.hcsProcess.CloseStdin()
}
// Pause handles pause requests for containers
func (c *client) Pause(_ context.Context, containerID string) error {
ctr, _, err := c.getProcess(containerID, InitProcessName)
if err != nil {
return err
}
if ctr.ociSpec.Windows.HyperV == nil {
return errors.New("cannot pause Windows Server Containers")
}
ctr.Lock()
defer ctr.Unlock()
if err = ctr.hcsContainer.Pause(); err != nil {
return err
}
ctr.status = StatusPaused
c.eventQ.append(containerID, func() {
err := c.backend.ProcessEvent(containerID, EventPaused, EventInfo{
ContainerID: containerID,
ProcessID: InitProcessName,
})
c.logger.WithFields(logrus.Fields{
"container": ctr.id,
"event": EventPaused,
}).Info("sending event")
if err != nil {
c.logger.WithError(err).WithFields(logrus.Fields{
"container": containerID,
"event": EventPaused,
}).Error("failed to process event")
}
})
return nil
}