aboutsummaryrefslogtreecommitdiff
path: root/libpod/oci_conmon_linux.go
blob: 2914bd1a19f9921676702ef0b888acddde1e5c8d (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
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
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
// +build linux

package libpod

import (
	"bufio"
	"bytes"
	"context"
	"fmt"
	"io"
	"io/ioutil"
	"net"
	"net/http"
	"os"
	"os/exec"
	"path/filepath"
	"runtime"
	"strconv"
	"strings"
	"sync"
	"syscall"
	"text/template"
	"time"

	"github.com/containers/common/pkg/config"
	conmonConfig "github.com/containers/conmon/runner/config"
	"github.com/containers/podman/v3/libpod/define"
	"github.com/containers/podman/v3/libpod/logs"
	"github.com/containers/podman/v3/pkg/cgroups"
	"github.com/containers/podman/v3/pkg/checkpoint/crutils"
	"github.com/containers/podman/v3/pkg/errorhandling"
	"github.com/containers/podman/v3/pkg/rootless"
	"github.com/containers/podman/v3/pkg/util"
	"github.com/containers/podman/v3/utils"
	"github.com/containers/storage/pkg/homedir"
	pmount "github.com/containers/storage/pkg/mount"
	"github.com/coreos/go-systemd/v22/activation"
	"github.com/coreos/go-systemd/v22/daemon"
	spec "github.com/opencontainers/runtime-spec/specs-go"
	"github.com/opencontainers/selinux/go-selinux"
	"github.com/opencontainers/selinux/go-selinux/label"
	"github.com/pkg/errors"
	"github.com/sirupsen/logrus"
	"golang.org/x/sys/unix"
)

const (
	// This is Conmon's STDIO_BUF_SIZE. I don't believe we have access to it
	// directly from the Go code, so const it here
	bufferSize = conmonConfig.BufSize
)

// ConmonOCIRuntime is an OCI runtime managed by Conmon.
// TODO: Make all calls to OCI runtime have a timeout.
type ConmonOCIRuntime struct {
	name              string
	path              string
	conmonPath        string
	conmonEnv         []string
	tmpDir            string
	exitsDir          string
	logSizeMax        int64
	noPivot           bool
	reservePorts      bool
	runtimeFlags      []string
	supportsJSON      bool
	supportsKVM       bool
	supportsNoCgroups bool
	sdNotify          bool
	enableKeyring     bool
}

// Make a new Conmon-based OCI runtime with the given options.
// Conmon will wrap the given OCI runtime, which can be `runc`, `crun`, or
// any runtime with a runc-compatible CLI.
// The first path that points to a valid executable will be used.
// Deliberately private. Someone should not be able to construct this outside of
// libpod.
func newConmonOCIRuntime(name string, paths []string, conmonPath string, runtimeFlags []string, runtimeCfg *config.Config) (OCIRuntime, error) {
	if name == "" {
		return nil, errors.Wrapf(define.ErrInvalidArg, "the OCI runtime must be provided a non-empty name")
	}

	// Make lookup tables for runtime support
	supportsJSON := make(map[string]bool, len(runtimeCfg.Engine.RuntimeSupportsJSON))
	supportsNoCgroups := make(map[string]bool, len(runtimeCfg.Engine.RuntimeSupportsNoCgroups))
	supportsKVM := make(map[string]bool, len(runtimeCfg.Engine.RuntimeSupportsKVM))
	for _, r := range runtimeCfg.Engine.RuntimeSupportsJSON {
		supportsJSON[r] = true
	}
	for _, r := range runtimeCfg.Engine.RuntimeSupportsNoCgroups {
		supportsNoCgroups[r] = true
	}
	for _, r := range runtimeCfg.Engine.RuntimeSupportsKVM {
		supportsKVM[r] = true
	}

	runtime := new(ConmonOCIRuntime)
	runtime.name = name
	runtime.conmonPath = conmonPath
	runtime.runtimeFlags = runtimeFlags

	runtime.conmonEnv = runtimeCfg.Engine.ConmonEnvVars
	runtime.tmpDir = runtimeCfg.Engine.TmpDir
	runtime.logSizeMax = runtimeCfg.Containers.LogSizeMax
	runtime.noPivot = runtimeCfg.Engine.NoPivotRoot
	runtime.reservePorts = runtimeCfg.Engine.EnablePortReservation
	runtime.sdNotify = runtimeCfg.Engine.SDNotify
	runtime.enableKeyring = runtimeCfg.Containers.EnableKeyring

	// TODO: probe OCI runtime for feature and enable automatically if
	// available.

	base := filepath.Base(name)
	runtime.supportsJSON = supportsJSON[base]
	runtime.supportsNoCgroups = supportsNoCgroups[base]
	runtime.supportsKVM = supportsKVM[base]

	foundPath := false
	for _, path := range paths {
		stat, err := os.Stat(path)
		if err != nil {
			if os.IsNotExist(err) {
				continue
			}
			return nil, errors.Wrapf(err, "cannot stat OCI runtime %s path", name)
		}
		if !stat.Mode().IsRegular() {
			continue
		}
		foundPath = true
		logrus.Tracef("found runtime %q", runtime.path)
		runtime.path = path
		break
	}

	// Search the $PATH as last fallback
	if !foundPath {
		if foundRuntime, err := exec.LookPath(name); err == nil {
			foundPath = true
			runtime.path = foundRuntime
			logrus.Debugf("using runtime %q from $PATH: %q", name, foundRuntime)
		}
	}

	if !foundPath {
		return nil, errors.Wrapf(define.ErrInvalidArg, "no valid executable found for OCI runtime %s", name)
	}

	runtime.exitsDir = filepath.Join(runtime.tmpDir, "exits")

	// Create the exit files and attach sockets directories
	if err := os.MkdirAll(runtime.exitsDir, 0750); err != nil {
		// The directory is allowed to exist
		if !os.IsExist(err) {
			return nil, errors.Wrapf(err, "error creating OCI runtime exit files directory")
		}
	}
	return runtime, nil
}

// Name returns the name of the runtime being wrapped by Conmon.
func (r *ConmonOCIRuntime) Name() string {
	return r.name
}

// Path returns the path of the OCI runtime being wrapped by Conmon.
func (r *ConmonOCIRuntime) Path() string {
	return r.path
}

// hasCurrentUserMapped checks whether the current user is mapped inside the container user namespace
func hasCurrentUserMapped(ctr *Container) bool {
	if len(ctr.config.IDMappings.UIDMap) == 0 && len(ctr.config.IDMappings.GIDMap) == 0 {
		return true
	}
	uid := os.Geteuid()
	for _, m := range ctr.config.IDMappings.UIDMap {
		if uid >= m.HostID && uid < m.HostID+m.Size {
			return true
		}
	}
	return false
}

// CreateContainer creates a container.
func (r *ConmonOCIRuntime) CreateContainer(ctr *Container, restoreOptions *ContainerCheckpointOptions) error {
	// always make the run dir accessible to the current user so that the PID files can be read without
	// being in the rootless user namespace.
	if err := makeAccessible(ctr.state.RunDir, 0, 0); err != nil {
		return err
	}
	if !hasCurrentUserMapped(ctr) {
		for _, i := range []string{ctr.state.RunDir, ctr.runtime.config.Engine.TmpDir, ctr.config.StaticDir, ctr.state.Mountpoint, ctr.runtime.config.Engine.VolumePath} {
			if err := makeAccessible(i, ctr.RootUID(), ctr.RootGID()); err != nil {
				return err
			}
		}

		// if we are running a non privileged container, be sure to umount some kernel paths so they are not
		// bind mounted inside the container at all.
		if !ctr.config.Privileged && !rootless.IsRootless() {
			ch := make(chan error)
			go func() {
				runtime.LockOSThread()
				err := func() error {
					fd, err := os.Open(fmt.Sprintf("/proc/%d/task/%d/ns/mnt", os.Getpid(), unix.Gettid()))
					if err != nil {
						return err
					}
					defer errorhandling.CloseQuiet(fd)

					// create a new mountns on the current thread
					if err = unix.Unshare(unix.CLONE_NEWNS); err != nil {
						return err
					}
					defer func() {
						if err := unix.Setns(int(fd.Fd()), unix.CLONE_NEWNS); err != nil {
							logrus.Errorf("unable to clone new namespace: %q", err)
						}
					}()

					// don't spread our mounts around.  We are setting only /sys to be slave
					// so that the cleanup process is still able to umount the storage and the
					// changes are propagated to the host.
					err = unix.Mount("/sys", "/sys", "none", unix.MS_REC|unix.MS_SLAVE, "")
					if err != nil {
						return errors.Wrapf(err, "cannot make /sys slave")
					}

					mounts, err := pmount.GetMounts()
					if err != nil {
						return err
					}
					for _, m := range mounts {
						if !strings.HasPrefix(m.Mountpoint, "/sys/kernel") {
							continue
						}
						err = unix.Unmount(m.Mountpoint, 0)
						if err != nil && !os.IsNotExist(err) {
							return errors.Wrapf(err, "cannot unmount %s", m.Mountpoint)
						}
					}
					return r.createOCIContainer(ctr, restoreOptions)
				}()
				ch <- err
			}()
			err := <-ch
			return err
		}
	}
	return r.createOCIContainer(ctr, restoreOptions)
}

// UpdateContainerStatus retrieves the current status of the container from the
// runtime. It updates the container's state but does not save it.
// If useRuntime is false, we will not directly hit runc to see the container's
// status, but will instead only check for the existence of the conmon exit file
// and update state to stopped if it exists.
func (r *ConmonOCIRuntime) UpdateContainerStatus(ctr *Container) error {
	exitFile, err := r.ExitFilePath(ctr)
	if err != nil {
		return err
	}

	runtimeDir, err := util.GetRuntimeDir()
	if err != nil {
		return err
	}

	// Store old state so we know if we were already stopped
	oldState := ctr.state.State

	state := new(spec.State)

	cmd := exec.Command(r.path, "state", ctr.ID())
	cmd.Env = append(cmd.Env, fmt.Sprintf("XDG_RUNTIME_DIR=%s", runtimeDir))

	outPipe, err := cmd.StdoutPipe()
	if err != nil {
		return errors.Wrapf(err, "getting stdout pipe")
	}
	errPipe, err := cmd.StderrPipe()
	if err != nil {
		return errors.Wrapf(err, "getting stderr pipe")
	}

	if err := cmd.Start(); err != nil {
		out, err2 := ioutil.ReadAll(errPipe)
		if err2 != nil {
			return errors.Wrapf(err, "error getting container %s state", ctr.ID())
		}
		if strings.Contains(string(out), "does not exist") {
			if err := ctr.removeConmonFiles(); err != nil {
				logrus.Debugf("unable to remove conmon files for container %s", ctr.ID())
			}
			ctr.state.ExitCode = -1
			ctr.state.FinishedTime = time.Now()
			ctr.state.State = define.ContainerStateExited
			return nil
		}
		return errors.Wrapf(err, "error getting container %s state. stderr/out: %s", ctr.ID(), out)
	}
	defer func() {
		_ = cmd.Wait()
	}()

	if err := errPipe.Close(); err != nil {
		return err
	}
	out, err := ioutil.ReadAll(outPipe)
	if err != nil {
		return errors.Wrapf(err, "error reading stdout: %s", ctr.ID())
	}
	if err := json.NewDecoder(bytes.NewBuffer(out)).Decode(state); err != nil {
		return errors.Wrapf(err, "error decoding container status for container %s", ctr.ID())
	}
	ctr.state.PID = state.Pid

	switch state.Status {
	case "created":
		ctr.state.State = define.ContainerStateCreated
	case "paused":
		ctr.state.State = define.ContainerStatePaused
	case "running":
		ctr.state.State = define.ContainerStateRunning
	case "stopped":
		ctr.state.State = define.ContainerStateStopped
	default:
		return errors.Wrapf(define.ErrInternal, "unrecognized status returned by runtime for container %s: %s",
			ctr.ID(), state.Status)
	}

	// Only grab exit status if we were not already stopped
	// If we were, it should already be in the database
	if ctr.state.State == define.ContainerStateStopped && oldState != define.ContainerStateStopped {
		var fi os.FileInfo
		chWait := make(chan error)
		defer close(chWait)

		_, err := WaitForFile(exitFile, chWait, time.Second*5)
		if err == nil {
			fi, err = os.Stat(exitFile)
		}
		if err != nil {
			ctr.state.ExitCode = -1
			ctr.state.FinishedTime = time.Now()
			logrus.Errorf("No exit file for container %s found: %v", ctr.ID(), err)
			return nil
		}

		return ctr.handleExitFile(exitFile, fi)
	}

	return nil
}

// StartContainer starts the given container.
// Sets time the container was started, but does not save it.
func (r *ConmonOCIRuntime) StartContainer(ctr *Container) error {
	// TODO: streams should probably *not* be our STDIN/OUT/ERR - redirect to buffers?
	runtimeDir, err := util.GetRuntimeDir()
	if err != nil {
		return err
	}
	env := []string{fmt.Sprintf("XDG_RUNTIME_DIR=%s", runtimeDir)}
	if ctr.config.SdNotifyMode == define.SdNotifyModeContainer {
		if notify, ok := os.LookupEnv("NOTIFY_SOCKET"); ok {
			env = append(env, fmt.Sprintf("NOTIFY_SOCKET=%s", notify))
		}
	}
	if path, ok := os.LookupEnv("PATH"); ok {
		env = append(env, fmt.Sprintf("PATH=%s", path))
	}
	if err := utils.ExecCmdWithStdStreams(os.Stdin, os.Stdout, os.Stderr, env, r.path, append(r.runtimeFlags, "start", ctr.ID())...); err != nil {
		return err
	}

	ctr.state.StartedTime = time.Now()

	return nil
}

// KillContainer sends the given signal to the given container.
// If all is set, send to all PIDs in the container.
// All is only supported if the container created cgroups.
func (r *ConmonOCIRuntime) KillContainer(ctr *Container, signal uint, all bool) error {
	logrus.Debugf("Sending signal %d to container %s", signal, ctr.ID())
	runtimeDir, err := util.GetRuntimeDir()
	if err != nil {
		return err
	}
	env := []string{fmt.Sprintf("XDG_RUNTIME_DIR=%s", runtimeDir)}
	var args []string
	args = append(args, r.runtimeFlags...)
	if all {
		args = append(args, "kill", "--all", ctr.ID(), fmt.Sprintf("%d", signal))
	} else {
		args = append(args, "kill", ctr.ID(), fmt.Sprintf("%d", signal))
	}
	if err := utils.ExecCmdWithStdStreams(os.Stdin, os.Stdout, os.Stderr, env, r.path, args...); err != nil {
		return errors.Wrapf(err, "error sending signal to container %s", ctr.ID())
	}

	return nil
}

// StopContainer stops a container, first using its given stop signal (or
// SIGTERM if no signal was specified), then using SIGKILL.
// Timeout is given in seconds. If timeout is 0, the container will be
// immediately kill with SIGKILL.
// Does not set finished time for container, assumes you will run updateStatus
// after to pull the exit code.
func (r *ConmonOCIRuntime) StopContainer(ctr *Container, timeout uint, all bool) error {
	logrus.Debugf("Stopping container %s (PID %d)", ctr.ID(), ctr.state.PID)

	// Ping the container to see if it's alive
	// If it's not, it's already stopped, return
	err := unix.Kill(ctr.state.PID, 0)
	if err == unix.ESRCH {
		return nil
	}

	stopSignal := ctr.config.StopSignal
	if stopSignal == 0 {
		stopSignal = uint(syscall.SIGTERM)
	}

	if timeout > 0 {
		if err := r.KillContainer(ctr, stopSignal, all); err != nil {
			// Is the container gone?
			// If so, it probably died between the first check and
			// our sending the signal
			// The container is stopped, so exit cleanly
			err := unix.Kill(ctr.state.PID, 0)
			if err == unix.ESRCH {
				return nil
			}

			return err
		}

		if err := waitContainerStop(ctr, time.Duration(timeout)*time.Second); err != nil {
			logrus.Infof("Timed out stopping container %s, resorting to SIGKILL: %v", ctr.ID(), err)
		} else {
			// No error, the container is dead
			return nil
		}
	}

	if err := r.KillContainer(ctr, 9, all); err != nil {
		// Again, check if the container is gone. If it is, exit cleanly.
		err := unix.Kill(ctr.state.PID, 0)
		if err == unix.ESRCH {
			return nil
		}

		return errors.Wrapf(err, "error sending SIGKILL to container %s", ctr.ID())
	}

	// Give runtime a few seconds to make it happen
	if err := waitContainerStop(ctr, killContainerTimeout); err != nil {
		return err
	}

	return nil
}

// DeleteContainer deletes a container from the OCI runtime.
func (r *ConmonOCIRuntime) DeleteContainer(ctr *Container) error {
	runtimeDir, err := util.GetRuntimeDir()
	if err != nil {
		return err
	}
	env := []string{fmt.Sprintf("XDG_RUNTIME_DIR=%s", runtimeDir)}
	return utils.ExecCmdWithStdStreams(os.Stdin, os.Stdout, os.Stderr, env, r.path, append(r.runtimeFlags, "delete", "--force", ctr.ID())...)
}

// PauseContainer pauses the given container.
func (r *ConmonOCIRuntime) PauseContainer(ctr *Container) error {
	runtimeDir, err := util.GetRuntimeDir()
	if err != nil {
		return err
	}
	env := []string{fmt.Sprintf("XDG_RUNTIME_DIR=%s", runtimeDir)}
	return utils.ExecCmdWithStdStreams(os.Stdin, os.Stdout, os.Stderr, env, r.path, append(r.runtimeFlags, "pause", ctr.ID())...)
}

// UnpauseContainer unpauses the given container.
func (r *ConmonOCIRuntime) UnpauseContainer(ctr *Container) error {
	runtimeDir, err := util.GetRuntimeDir()
	if err != nil {
		return err
	}
	env := []string{fmt.Sprintf("XDG_RUNTIME_DIR=%s", runtimeDir)}
	return utils.ExecCmdWithStdStreams(os.Stdin, os.Stdout, os.Stderr, env, r.path, append(r.runtimeFlags, "resume", ctr.ID())...)
}

// HTTPAttach performs an attach for the HTTP API.
// The caller must handle closing the HTTP connection after this returns.
// The cancel channel is not closed; it is up to the caller to do so after
// this function returns.
// If this is a container with a terminal, we will stream raw. If it is not, we
// will stream with an 8-byte header to multiplex STDOUT and STDERR.
// Returns any errors that occurred, and whether the connection was successfully
// hijacked before that error occurred.
func (r *ConmonOCIRuntime) HTTPAttach(ctr *Container, req *http.Request, w http.ResponseWriter, streams *HTTPAttachStreams, detachKeys *string, cancel <-chan bool, hijackDone chan<- bool, streamAttach, streamLogs bool) (deferredErr error) {
	isTerminal := false
	if ctr.config.Spec.Process != nil {
		isTerminal = ctr.config.Spec.Process.Terminal
	}

	if streams != nil {
		if !streams.Stdin && !streams.Stdout && !streams.Stderr {
			return errors.Wrapf(define.ErrInvalidArg, "must specify at least one stream to attach to")
		}
	}

	attachSock, err := r.AttachSocketPath(ctr)
	if err != nil {
		return err
	}

	var conn *net.UnixConn
	if streamAttach {
		newConn, err := openUnixSocket(attachSock)
		if err != nil {
			return errors.Wrapf(err, "failed to connect to container's attach socket: %v", attachSock)
		}
		conn = newConn
		defer func() {
			if err := conn.Close(); err != nil {
				logrus.Errorf("unable to close container %s attach socket: %q", ctr.ID(), err)
			}
		}()

		logrus.Debugf("Successfully connected to container %s attach socket %s", ctr.ID(), attachSock)
	}

	detachString := ctr.runtime.config.Engine.DetachKeys
	if detachKeys != nil {
		detachString = *detachKeys
	}
	detach, err := processDetachKeys(detachString)
	if err != nil {
		return err
	}

	attachStdout := true
	attachStderr := true
	attachStdin := true
	if streams != nil {
		attachStdout = streams.Stdout
		attachStderr = streams.Stderr
		attachStdin = streams.Stdin
	}

	logrus.Debugf("Going to hijack container %s attach connection", ctr.ID())

	// Alright, let's hijack.
	hijacker, ok := w.(http.Hijacker)
	if !ok {
		return errors.Errorf("unable to hijack connection")
	}

	httpCon, httpBuf, err := hijacker.Hijack()
	if err != nil {
		return errors.Wrapf(err, "error hijacking connection")
	}

	hijackDone <- true

	writeHijackHeader(req, httpBuf)

	// Force a flush after the header is written.
	if err := httpBuf.Flush(); err != nil {
		return errors.Wrapf(err, "error flushing HTTP hijack header")
	}

	defer func() {
		hijackWriteErrorAndClose(deferredErr, ctr.ID(), isTerminal, httpCon, httpBuf)
	}()

	logrus.Debugf("Hijack for container %s attach session done, ready to stream", ctr.ID())

	// TODO: This is gross. Really, really gross.
	// I want to say we should read all the logs into an array before
	// calling this, in container_api.go, but that could take a lot of
	// memory...
	// On the whole, we need to figure out a better way of doing this,
	// though.
	logSize := 0
	if streamLogs {
		logrus.Debugf("Will stream logs for container %s attach session", ctr.ID())

		// Get all logs for the container
		logChan := make(chan *logs.LogLine)
		logOpts := new(logs.LogOptions)
		logOpts.Tail = -1
		logOpts.WaitGroup = new(sync.WaitGroup)
		errChan := make(chan error)
		go func() {
			var err error
			// In non-terminal mode we need to prepend with the
			// stream header.
			logrus.Debugf("Writing logs for container %s to HTTP attach", ctr.ID())
			for logLine := range logChan {
				if !isTerminal {
					device := logLine.Device
					var header []byte
					headerLen := uint32(len(logLine.Msg))
					logSize += len(logLine.Msg)
					switch strings.ToLower(device) {
					case "stdin":
						header = makeHTTPAttachHeader(0, headerLen)
					case "stdout":
						header = makeHTTPAttachHeader(1, headerLen)
					case "stderr":
						header = makeHTTPAttachHeader(2, headerLen)
					default:
						logrus.Errorf("Unknown device for log line: %s", device)
						header = makeHTTPAttachHeader(1, headerLen)
					}
					_, err = httpBuf.Write(header)
					if err != nil {
						break
					}
				}
				_, err = httpBuf.Write([]byte(logLine.Msg))
				if err != nil {
					break
				}
				_, err = httpBuf.Write([]byte("\n"))
				if err != nil {
					break
				}
				err = httpBuf.Flush()
				if err != nil {
					break
				}
			}
			errChan <- err
		}()
		go func() {
			logOpts.WaitGroup.Wait()
			close(logChan)
		}()
		if err := ctr.ReadLog(context.Background(), logOpts, logChan); err != nil {
			return err
		}
		logrus.Debugf("Done reading logs for container %s, %d bytes", ctr.ID(), logSize)
		if err := <-errChan; err != nil {
			return err
		}
	}
	if !streamAttach {
		logrus.Debugf("Done streaming logs for container %s attach, exiting as attach streaming not requested", ctr.ID())
		return nil
	}

	logrus.Debugf("Forwarding attach output for container %s", ctr.ID())

	stdoutChan := make(chan error)
	stdinChan := make(chan error)

	// Handle STDOUT/STDERR
	go func() {
		var err error
		if isTerminal {
			// Hack: return immediately if attachStdout not set to
			// emulate Docker.
			// Basically, when terminal is set, STDERR goes nowhere.
			// Everything does over STDOUT.
			// Therefore, if not attaching STDOUT - we'll never copy
			// anything from here.
			logrus.Debugf("Performing terminal HTTP attach for container %s", ctr.ID())
			if attachStdout {
				err = httpAttachTerminalCopy(conn, httpBuf, ctr.ID())
			}
		} else {
			logrus.Debugf("Performing non-terminal HTTP attach for container %s", ctr.ID())
			err = httpAttachNonTerminalCopy(conn, httpBuf, ctr.ID(), attachStdin, attachStdout, attachStderr)
		}
		stdoutChan <- err
		logrus.Debugf("STDOUT/ERR copy completed")
	}()
	// Next, STDIN. Avoid entirely if attachStdin unset.
	if attachStdin {
		go func() {
			_, err := utils.CopyDetachable(conn, httpBuf, detach)
			logrus.Debugf("STDIN copy completed")
			stdinChan <- err
		}()
	}

	for {
		select {
		case err := <-stdoutChan:
			if err != nil {
				return err
			}

			return nil
		case err := <-stdinChan:
			if err != nil {
				return err
			}
		case <-cancel:
			return nil
		}
	}
}

// isRetryable returns whether the error was caused by a blocked syscall or the
// specified operation on a non blocking file descriptor wasn't ready for completion.
func isRetryable(err error) bool {
	if errno, isErrno := errors.Cause(err).(syscall.Errno); isErrno {
		return errno == syscall.EINTR || errno == syscall.EAGAIN
	}
	return false
}

// openControlFile opens the terminal control file.
func openControlFile(ctr *Container, parentDir string) (*os.File, error) {
	controlPath := filepath.Join(parentDir, "ctl")
	for i := 0; i < 600; i++ {
		controlFile, err := os.OpenFile(controlPath, unix.O_WRONLY|unix.O_NONBLOCK, 0)
		if err == nil {
			return controlFile, err
		}
		if !isRetryable(err) {
			return nil, errors.Wrapf(err, "could not open ctl file for terminal resize for container %s", ctr.ID())
		}
		time.Sleep(time.Second / 10)
	}
	return nil, errors.Errorf("timeout waiting for %q", controlPath)
}

// AttachResize resizes the terminal used by the given container.
func (r *ConmonOCIRuntime) AttachResize(ctr *Container, newSize define.TerminalSize) error {
	controlFile, err := openControlFile(ctr, ctr.bundlePath())
	if err != nil {
		return err
	}
	defer controlFile.Close()

	logrus.Debugf("Received a resize event for container %s: %+v", ctr.ID(), newSize)
	if _, err = fmt.Fprintf(controlFile, "%d %d %d\n", 1, newSize.Height, newSize.Width); err != nil {
		return errors.Wrapf(err, "failed to write to ctl file to resize terminal")
	}

	return nil
}

// CheckpointContainer checkpoints the given container.
func (r *ConmonOCIRuntime) CheckpointContainer(ctr *Container, options ContainerCheckpointOptions) error {
	if err := label.SetSocketLabel(ctr.ProcessLabel()); err != nil {
		return err
	}
	// imagePath is used by CRIU to store the actual checkpoint files
	imagePath := ctr.CheckpointPath()
	if options.PreCheckPoint {
		imagePath = ctr.PreCheckPointPath()
	}
	// workPath will be used to store dump.log and stats-dump
	workPath := ctr.bundlePath()
	logrus.Debugf("Writing checkpoint to %s", imagePath)
	logrus.Debugf("Writing checkpoint logs to %s", workPath)
	logrus.Debugf("Pre-dump the container %t", options.PreCheckPoint)
	args := []string{}
	args = append(args, r.runtimeFlags...)
	args = append(args, "checkpoint")
	args = append(args, "--image-path")
	args = append(args, imagePath)
	args = append(args, "--work-path")
	args = append(args, workPath)
	if options.KeepRunning {
		args = append(args, "--leave-running")
	}
	if options.TCPEstablished {
		args = append(args, "--tcp-established")
	}
	if !options.PreCheckPoint && options.KeepRunning {
		args = append(args, "--leave-running")
	}
	if options.PreCheckPoint {
		args = append(args, "--pre-dump")
	}
	if !options.PreCheckPoint && options.WithPrevious {
		args = append(
			args,
			"--parent-path",
			filepath.Join("..", preCheckpointDir),
		)
	}
	runtimeDir, err := util.GetRuntimeDir()
	if err != nil {
		return err
	}
	if err = os.Setenv("XDG_RUNTIME_DIR", runtimeDir); err != nil {
		return errors.Wrapf(err, "cannot set XDG_RUNTIME_DIR")
	}
	args = append(args, ctr.ID())
	logrus.Debugf("the args to checkpoint: %s %s", r.path, strings.Join(args, " "))
	return utils.ExecCmdWithStdStreams(os.Stdin, os.Stdout, os.Stderr, nil, r.path, args...)
}

func (r *ConmonOCIRuntime) CheckConmonRunning(ctr *Container) (bool, error) {
	if ctr.state.ConmonPID == 0 {
		// If the container is running or paused, assume Conmon is
		// running. We didn't record Conmon PID on some old versions, so
		// that is likely what's going on...
		// Unusual enough that we should print a warning message though.
		if ctr.ensureState(define.ContainerStateRunning, define.ContainerStatePaused) {
			logrus.Warnf("Conmon PID is not set, but container is running!")
			return true, nil
		}
		// Container's not running, so conmon PID being unset is
		// expected. Conmon is not running.
		return false, nil
	}

	// We have a conmon PID. Ping it with signal 0.
	if err := unix.Kill(ctr.state.ConmonPID, 0); err != nil {
		if err == unix.ESRCH {
			return false, nil
		}
		return false, errors.Wrapf(err, "error pinging container %s conmon with signal 0", ctr.ID())
	}
	return true, nil
}

// SupportsCheckpoint checks if the OCI runtime supports checkpointing
// containers.
func (r *ConmonOCIRuntime) SupportsCheckpoint() bool {
	return crutils.CRRuntimeSupportsCheckpointRestore(r.path)
}

// SupportsJSONErrors checks if the OCI runtime supports JSON-formatted error
// messages.
func (r *ConmonOCIRuntime) SupportsJSONErrors() bool {
	return r.supportsJSON
}

// SupportsNoCgroups checks if the OCI runtime supports running containers
// without cgroups (the --cgroup-manager=disabled flag).
func (r *ConmonOCIRuntime) SupportsNoCgroups() bool {
	return r.supportsNoCgroups
}

// SupportsKVM checks if the OCI runtime supports running containers
// without KVM separation
func (r *ConmonOCIRuntime) SupportsKVM() bool {
	return r.supportsKVM
}

// AttachSocketPath is the path to a single container's attach socket.
func (r *ConmonOCIRuntime) AttachSocketPath(ctr *Container) (string, error) {
	if ctr == nil {
		return "", errors.Wrapf(define.ErrInvalidArg, "must provide a valid container to get attach socket path")
	}

	return filepath.Join(ctr.bundlePath(), "attach"), nil
}

// ExitFilePath is the path to a container's exit file.
func (r *ConmonOCIRuntime) ExitFilePath(ctr *Container) (string, error) {
	if ctr == nil {
		return "", errors.Wrapf(define.ErrInvalidArg, "must provide a valid container to get exit file path")
	}
	return filepath.Join(r.exitsDir, ctr.ID()), nil
}

// RuntimeInfo provides information on the runtime.
func (r *ConmonOCIRuntime) RuntimeInfo() (*define.ConmonInfo, *define.OCIRuntimeInfo, error) {
	runtimePackage := packageVersion(r.path)
	conmonPackage := packageVersion(r.conmonPath)
	runtimeVersion, err := r.getOCIRuntimeVersion()
	if err != nil {
		return nil, nil, errors.Wrapf(err, "error getting version of OCI runtime %s", r.name)
	}
	conmonVersion, err := r.getConmonVersion()
	if err != nil {
		return nil, nil, errors.Wrapf(err, "error getting conmon version")
	}

	conmon := define.ConmonInfo{
		Package: conmonPackage,
		Path:    r.conmonPath,
		Version: conmonVersion,
	}
	ocirt := define.OCIRuntimeInfo{
		Name:    r.name,
		Path:    r.path,
		Package: runtimePackage,
		Version: runtimeVersion,
	}
	return &conmon, &ocirt, nil
}

// makeAccessible changes the path permission and each parent directory to have --x--x--x
func makeAccessible(path string, uid, gid int) error {
	for ; path != "/"; path = filepath.Dir(path) {
		st, err := os.Stat(path)
		if err != nil {
			if os.IsNotExist(err) {
				return nil
			}
			return err
		}
		if int(st.Sys().(*syscall.Stat_t).Uid) == uid && int(st.Sys().(*syscall.Stat_t).Gid) == gid {
			continue
		}
		if st.Mode()&0111 != 0111 {
			if err := os.Chmod(path, st.Mode()|0111); err != nil {
				return err
			}
		}
	}
	return nil
}

// Wait for a container which has been sent a signal to stop
func waitContainerStop(ctr *Container, timeout time.Duration) error {
	return waitPidStop(ctr.state.PID, timeout)
}

// Wait for a given PID to stop
func waitPidStop(pid int, timeout time.Duration) error {
	done := make(chan struct{})
	chControl := make(chan struct{})
	go func() {
		for {
			select {
			case <-chControl:
				return
			default:
				if err := unix.Kill(pid, 0); err != nil {
					if err == unix.ESRCH {
						close(done)
						return
					}
					logrus.Errorf("Error pinging PID %d with signal 0: %v", pid, err)
				}
				time.Sleep(100 * time.Millisecond)
			}
		}
	}()
	select {
	case <-done:
		return nil
	case <-time.After(timeout):
		close(chControl)
		return errors.Errorf("given PIDs did not die within timeout")
	}
}

func (r *ConmonOCIRuntime) getLogTag(ctr *Container) (string, error) {
	logTag := ctr.LogTag()
	if logTag == "" {
		return "", nil
	}
	data, err := ctr.inspectLocked(false)
	if err != nil {
		return "", nil
	}
	tmpl, err := template.New("container").Parse(logTag)
	if err != nil {
		return "", errors.Wrapf(err, "template parsing error %s", logTag)
	}
	var b bytes.Buffer
	err = tmpl.Execute(&b, data)
	if err != nil {
		return "", err
	}
	return b.String(), nil
}

// createOCIContainer generates this container's main conmon instance and prepares it for starting
func (r *ConmonOCIRuntime) createOCIContainer(ctr *Container, restoreOptions *ContainerCheckpointOptions) error {
	var stderrBuf bytes.Buffer

	runtimeDir, err := util.GetRuntimeDir()
	if err != nil {
		return err
	}

	parentSyncPipe, childSyncPipe, err := newPipe()
	if err != nil {
		return errors.Wrapf(err, "error creating socket pair")
	}
	defer errorhandling.CloseQuiet(parentSyncPipe)

	childStartPipe, parentStartPipe, err := newPipe()
	if err != nil {
		return errors.Wrapf(err, "error creating socket pair for start pipe")
	}

	defer errorhandling.CloseQuiet(parentStartPipe)

	var ociLog string
	if logrus.GetLevel() != logrus.DebugLevel && r.supportsJSON {
		ociLog = filepath.Join(ctr.state.RunDir, "oci-log")
	}

	logTag, err := r.getLogTag(ctr)
	if err != nil {
		return err
	}

	if ctr.config.CgroupsMode == cgroupSplit {
		if err := utils.MoveUnderCgroupSubtree("supervisor"); err != nil {
			return err
		}
	}

	if ctr.config.SdNotifyMode == define.SdNotifyModeIgnore {
		if err := os.Unsetenv("NOTIFY_SOCKET"); err != nil {
			logrus.Warnf("Error unsetting NOTIFY_SOCKET %v", err)
		}
	}

	pidfile := ctr.config.PidFile
	if pidfile == "" {
		pidfile = filepath.Join(ctr.state.RunDir, "pidfile")
	}

	args := r.sharedConmonArgs(ctr, ctr.ID(), ctr.bundlePath(), pidfile, ctr.LogPath(), r.exitsDir, ociLog, ctr.LogDriver(), logTag)

	if ctr.config.Spec.Process.Terminal {
		args = append(args, "-t")
	} else if ctr.config.Stdin {
		args = append(args, "-i")
	}

	if ctr.config.Timeout > 0 {
		args = append(args, fmt.Sprintf("--timeout=%d", ctr.config.Timeout))
	}

	if !r.enableKeyring {
		args = append(args, "--no-new-keyring")
	}
	if ctr.config.ConmonPidFile != "" {
		args = append(args, "--conmon-pidfile", ctr.config.ConmonPidFile)
	}

	if r.noPivot {
		args = append(args, "--no-pivot")
	}

	if len(ctr.config.ExitCommand) > 0 {
		args = append(args, "--exit-command", ctr.config.ExitCommand[0])
		for _, arg := range ctr.config.ExitCommand[1:] {
			args = append(args, []string{"--exit-command-arg", arg}...)
		}
	}

	if ctr.config.PreserveFDs > 0 {
		args = append(args, formatRuntimeOpts("--preserve-fds", fmt.Sprintf("%d", ctr.config.PreserveFDs))...)
	}

	if restoreOptions != nil {
		args = append(args, "--restore", ctr.CheckpointPath())
		if restoreOptions.TCPEstablished {
			args = append(args, "--runtime-opt", "--tcp-established")
		}
	}

	logrus.WithFields(logrus.Fields{
		"args": args,
	}).Debugf("running conmon: %s", r.conmonPath)

	cmd := exec.Command(r.conmonPath, args...)
	cmd.Dir = ctr.bundlePath()
	cmd.SysProcAttr = &syscall.SysProcAttr{
		Setpgid: true,
	}
	// TODO this is probably a really bad idea for some uses
	// Make this configurable
	cmd.Stdin = os.Stdin
	cmd.Stdout = os.Stdout
	cmd.Stderr = os.Stderr
	if ctr.config.Spec.Process.Terminal {
		cmd.Stderr = &stderrBuf
	}

	// 0, 1 and 2 are stdin, stdout and stderr
	conmonEnv, envFiles := r.configureConmonEnv(ctr, runtimeDir)

	var filesToClose []*os.File
	if ctr.config.PreserveFDs > 0 {
		for fd := 3; fd < int(3+ctr.config.PreserveFDs); fd++ {
			f := os.NewFile(uintptr(fd), fmt.Sprintf("fd-%d", fd))
			filesToClose = append(filesToClose, f)
			cmd.ExtraFiles = append(cmd.ExtraFiles, f)
		}
	}

	cmd.Env = r.conmonEnv
	// we don't want to step on users fds they asked to preserve
	// Since 0-2 are used for stdio, start the fds we pass in at preserveFDs+3
	cmd.Env = append(cmd.Env, fmt.Sprintf("_OCI_SYNCPIPE=%d", ctr.config.PreserveFDs+3), fmt.Sprintf("_OCI_STARTPIPE=%d", ctr.config.PreserveFDs+4))
	cmd.Env = append(cmd.Env, conmonEnv...)
	cmd.ExtraFiles = append(cmd.ExtraFiles, childSyncPipe, childStartPipe)
	cmd.ExtraFiles = append(cmd.ExtraFiles, envFiles...)

	if r.reservePorts && !rootless.IsRootless() && !ctr.config.NetMode.IsSlirp4netns() {
		ports, err := bindPorts(ctr.config.PortMappings)
		if err != nil {
			return err
		}

		// Leak the port we bound in the conmon process.  These fd's won't be used
		// by the container and conmon will keep the ports busy so that another
		// process cannot use them.
		cmd.ExtraFiles = append(cmd.ExtraFiles, ports...)
	}

	if ctr.config.NetMode.IsSlirp4netns() || rootless.IsRootless() {
		if ctr.config.PostConfigureNetNS {
			havePortMapping := len(ctr.Config().PortMappings) > 0
			if havePortMapping {
				ctr.rootlessPortSyncR, ctr.rootlessPortSyncW, err = os.Pipe()
				if err != nil {
					return errors.Wrapf(err, "failed to create rootless port sync pipe")
				}
			}
			ctr.rootlessSlirpSyncR, ctr.rootlessSlirpSyncW, err = os.Pipe()
			if err != nil {
				return errors.Wrapf(err, "failed to create rootless network sync pipe")
			}
		} else {
			if ctr.rootlessSlirpSyncR != nil {
				defer errorhandling.CloseQuiet(ctr.rootlessSlirpSyncR)
			}
			if ctr.rootlessSlirpSyncW != nil {
				defer errorhandling.CloseQuiet(ctr.rootlessSlirpSyncW)
			}
		}
		// Leak one end in conmon, the other one will be leaked into slirp4netns
		cmd.ExtraFiles = append(cmd.ExtraFiles, ctr.rootlessSlirpSyncW)

		if ctr.rootlessPortSyncW != nil {
			defer errorhandling.CloseQuiet(ctr.rootlessPortSyncW)
			// Leak one end in conmon, the other one will be leaked into rootlessport
			cmd.ExtraFiles = append(cmd.ExtraFiles, ctr.rootlessPortSyncW)
		}
	}

	err = startCommandGivenSelinux(cmd)
	// regardless of whether we errored or not, we no longer need the children pipes
	childSyncPipe.Close()
	childStartPipe.Close()
	if err != nil {
		return err
	}
	if err := r.moveConmonToCgroupAndSignal(ctr, cmd, parentStartPipe); err != nil {
		return err
	}
	/* Wait for initial setup and fork, and reap child */
	err = cmd.Wait()
	if err != nil {
		return err
	}

	pid, err := readConmonPipeData(parentSyncPipe, ociLog)
	if err != nil {
		if err2 := r.DeleteContainer(ctr); err2 != nil {
			logrus.Errorf("Error removing container %s from runtime after creation failed", ctr.ID())
		}
		return err
	}
	ctr.state.PID = pid

	conmonPID, err := readConmonPidFile(ctr.config.ConmonPidFile)
	if err != nil {
		logrus.Warnf("error reading conmon pid file for container %s: %v", ctr.ID(), err)
	} else if conmonPID > 0 {
		// conmon not having a pid file is a valid state, so don't set it if we don't have it
		logrus.Infof("Got Conmon PID as %d", conmonPID)
		ctr.state.ConmonPID = conmonPID
		if ctr.config.SdNotifyMode != define.SdNotifyModeIgnore {
			if sent, err := daemon.SdNotify(false, fmt.Sprintf("MAINPID=%d", conmonPID)); err != nil {
				logrus.Errorf("Error notifying systemd of Conmon PID: %v", err)
			} else if sent {
				logrus.Debugf("Notify MAINPID sent successfully")
			}
		}
	}

	// These fds were passed down to the runtime.  Close them
	// and not interfere
	for _, f := range filesToClose {
		errorhandling.CloseQuiet(f)
	}

	return nil
}

// configureConmonEnv gets the environment values to add to conmon's exec struct
// TODO this may want to be less hardcoded/more configurable in the future
func (r *ConmonOCIRuntime) configureConmonEnv(ctr *Container, runtimeDir string) ([]string, []*os.File) {
	var env []string
	for _, e := range os.Environ() {
		if strings.HasPrefix(e, "LC_") {
			env = append(env, e)
		}
	}
	env = append(env, fmt.Sprintf("XDG_RUNTIME_DIR=%s", runtimeDir))
	env = append(env, fmt.Sprintf("_CONTAINERS_USERNS_CONFIGURED=%s", os.Getenv("_CONTAINERS_USERNS_CONFIGURED")))
	env = append(env, fmt.Sprintf("_CONTAINERS_ROOTLESS_UID=%s", os.Getenv("_CONTAINERS_ROOTLESS_UID")))
	home := homedir.Get()
	if home != "" {
		env = append(env, fmt.Sprintf("HOME=%s", home))
	}

	extraFiles := make([]*os.File, 0)
	if ctr.config.SdNotifyMode == define.SdNotifyModeContainer {
		if notify, ok := os.LookupEnv("NOTIFY_SOCKET"); ok {
			env = append(env, fmt.Sprintf("NOTIFY_SOCKET=%s", notify))
		}
	}
	if !r.sdNotify {
		if listenfds, ok := os.LookupEnv("LISTEN_FDS"); ok {
			env = append(env, fmt.Sprintf("LISTEN_FDS=%s", listenfds), "LISTEN_PID=1")
			fds := activation.Files(false)
			extraFiles = append(extraFiles, fds...)
		}
	} else {
		logrus.Debug("disabling SD notify")
	}
	return env, extraFiles
}

// sharedConmonArgs takes common arguments for exec and create/restore and formats them for the conmon CLI
func (r *ConmonOCIRuntime) sharedConmonArgs(ctr *Container, cuuid, bundlePath, pidPath, logPath, exitDir, ociLogPath, logDriver, logTag string) []string {
	// set the conmon API version to be able to use the correct sync struct keys
	args := []string{
		"--api-version", "1",
		"-c", ctr.ID(),
		"-u", cuuid,
		"-r", r.path,
		"-b", bundlePath,
		"-p", pidPath,
		"-n", ctr.Name(),
		"--exit-dir", exitDir,
		"--full-attach",
	}
	if len(r.runtimeFlags) > 0 {
		rFlags := []string{}
		for _, arg := range r.runtimeFlags {
			rFlags = append(rFlags, "--runtime-arg", arg)
		}
		args = append(args, rFlags...)
	}

	if ctr.CgroupManager() == config.SystemdCgroupsManager && !ctr.config.NoCgroups && ctr.config.CgroupsMode != cgroupSplit {
		args = append(args, "-s")
	}

	var logDriverArg string
	switch logDriver {
	case define.JournaldLogging:
		logDriverArg = define.JournaldLogging
	case define.NoLogging:
		logDriverArg = define.NoLogging
	case define.JSONLogging:
		fallthrough
	//lint:ignore ST1015 the default case has to be here
	default: //nolint-stylecheck
		// No case here should happen except JSONLogging, but keep this here in case the options are extended
		logrus.Errorf("%s logging specified but not supported. Choosing k8s-file logging instead", ctr.LogDriver())
		fallthrough
	case "":
		// to get here, either a user would specify `--log-driver ""`, or this came from another place in libpod
		// since the former case is obscure, and the latter case isn't an error, let's silently fallthrough
		fallthrough
	case define.KubernetesLogging:
		logDriverArg = fmt.Sprintf("%s:%s", define.KubernetesLogging, logPath)
	}

	args = append(args, "-l", logDriverArg)
	logLevel := logrus.GetLevel()
	args = append(args, "--log-level", logLevel.String())

	if logLevel == logrus.DebugLevel {
		logrus.Debugf("%s messages will be logged to syslog", r.conmonPath)
		args = append(args, "--syslog")
	}

	size := r.logSizeMax
	if ctr.config.LogSize > 0 {
		size = ctr.config.LogSize
	}
	if size > 0 {
		args = append(args, "--log-size-max", fmt.Sprintf("%v", size))
	}

	if ociLogPath != "" {
		args = append(args, "--runtime-arg", "--log-format=json", "--runtime-arg", "--log", fmt.Sprintf("--runtime-arg=%s", ociLogPath))
	}
	if logTag != "" {
		args = append(args, "--log-tag", logTag)
	}
	if ctr.config.NoCgroups {
		logrus.Debugf("Running with no CGroups")
		args = append(args, "--runtime-arg", "--cgroup-manager", "--runtime-arg", "disabled")
	}
	return args
}

// startCommandGivenSelinux starts a container ensuring to set the labels of
// the process to make sure SELinux doesn't block conmon communication, if SELinux is enabled
func startCommandGivenSelinux(cmd *exec.Cmd) error {
	if !selinux.GetEnabled() {
		return cmd.Start()
	}
	// Set the label of the conmon process to be level :s0
	// This will allow the container processes to talk to fifo-files
	// passed into the container by conmon
	var (
		plabel string
		con    selinux.Context
		err    error
	)
	plabel, err = selinux.CurrentLabel()
	if err != nil {
		return errors.Wrapf(err, "failed to get current SELinux label")
	}

	con, err = selinux.NewContext(plabel)
	if err != nil {
		return errors.Wrapf(err, "failed to get new context from SELinux label")
	}

	runtime.LockOSThread()
	if con["level"] != "s0" && con["level"] != "" {
		con["level"] = "s0"
		if err = label.SetProcessLabel(con.Get()); err != nil {
			runtime.UnlockOSThread()
			return err
		}
	}
	err = cmd.Start()
	// Ignore error returned from SetProcessLabel("") call,
	// can't recover.
	if labelErr := label.SetProcessLabel(""); labelErr != nil {
		logrus.Errorf("unable to set process label: %q", err)
	}
	runtime.UnlockOSThread()
	return err
}

// moveConmonToCgroupAndSignal gets a container's cgroupParent and moves the conmon process to that cgroup
// it then signals for conmon to start by sending nonce data down the start fd
func (r *ConmonOCIRuntime) moveConmonToCgroupAndSignal(ctr *Container, cmd *exec.Cmd, startFd *os.File) error {
	mustCreateCgroup := true

	if ctr.config.NoCgroups {
		mustCreateCgroup = false
	}

	// If cgroup creation is disabled - just signal.
	switch ctr.config.CgroupsMode {
	case "disabled", "no-conmon", cgroupSplit:
		mustCreateCgroup = false
	}

	// $INVOCATION_ID is set by systemd when running as a service.
	if os.Getenv("INVOCATION_ID") != "" {
		mustCreateCgroup = false
	}

	if mustCreateCgroup {
		// Usually rootless users are not allowed to configure cgroupfs.
		// There are cases though, where it is allowed, e.g. if the cgroup
		// is manually configured and chowned).  Avoid detecting all
		// such cases and simply use a lower log level.
		logLevel := logrus.WarnLevel
		if rootless.IsRootless() {
			logLevel = logrus.InfoLevel
		}
		// TODO: This should be a switch - we are not guaranteed that
		// there are only 2 valid cgroup managers
		cgroupParent := ctr.CgroupParent()
		if ctr.CgroupManager() == config.SystemdCgroupsManager {
			unitName := createUnitName("libpod-conmon", ctr.ID())

			realCgroupParent := cgroupParent
			splitParent := strings.Split(cgroupParent, "/")
			if strings.HasSuffix(cgroupParent, ".slice") && len(splitParent) > 1 {
				realCgroupParent = splitParent[len(splitParent)-1]
			}

			logrus.Infof("Running conmon under slice %s and unitName %s", realCgroupParent, unitName)
			if err := utils.RunUnderSystemdScope(cmd.Process.Pid, realCgroupParent, unitName); err != nil {
				logrus.StandardLogger().Logf(logLevel, "Failed to add conmon to systemd sandbox cgroup: %v", err)
			}
		} else {
			cgroupPath := filepath.Join(ctr.config.CgroupParent, "conmon")
			control, err := cgroups.New(cgroupPath, &spec.LinuxResources{})
			if err != nil {
				logrus.StandardLogger().Logf(logLevel, "Failed to add conmon to cgroupfs sandbox cgroup: %v", err)
			} else if err := control.AddPid(cmd.Process.Pid); err != nil {
				// we need to remove this defer and delete the cgroup once conmon exits
				// maybe need a conmon monitor?
				logrus.StandardLogger().Logf(logLevel, "Failed to add conmon to cgroupfs sandbox cgroup: %v", err)
			}
		}
	}

	/* We set the cgroup, now the child can start creating children */
	if err := writeConmonPipeData(startFd); err != nil {
		return err
	}
	return nil
}

// newPipe creates a unix socket pair for communication.
// Returns two files - first is parent, second is child.
func newPipe() (*os.File, *os.File, error) {
	fds, err := unix.Socketpair(unix.AF_LOCAL, unix.SOCK_SEQPACKET|unix.SOCK_CLOEXEC, 0)
	if err != nil {
		return nil, nil, err
	}
	return os.NewFile(uintptr(fds[1]), "parent"), os.NewFile(uintptr(fds[0]), "child"), nil
}

// readConmonPidFile attempts to read conmon's pid from its pid file
func readConmonPidFile(pidFile string) (int, error) {
	// Let's try reading the Conmon pid at the same time.
	if pidFile != "" {
		contents, err := ioutil.ReadFile(pidFile)
		if err != nil {
			return -1, err
		}
		// Convert it to an int
		conmonPID, err := strconv.Atoi(string(contents))
		if err != nil {
			return -1, err
		}
		return conmonPID, nil
	}
	return 0, nil
}

// readConmonPipeData attempts to read a syncInfo struct from the pipe
func readConmonPipeData(pipe *os.File, ociLog string) (int, error) {
	// syncInfo is used to return data from monitor process to daemon
	type syncInfo struct {
		Data    int    `json:"data"`
		Message string `json:"message,omitempty"`
	}

	// Wait to get container pid from conmon
	type syncStruct struct {
		si  *syncInfo
		err error
	}
	ch := make(chan syncStruct)
	go func() {
		var si *syncInfo
		rdr := bufio.NewReader(pipe)
		b, err := rdr.ReadBytes('\n')
		if err != nil {
			ch <- syncStruct{err: err}
		}
		if err := json.Unmarshal(b, &si); err != nil {
			ch <- syncStruct{err: err}
			return
		}
		ch <- syncStruct{si: si}
	}()

	data := -1
	select {
	case ss := <-ch:
		if ss.err != nil {
			if ociLog != "" {
				ociLogData, err := ioutil.ReadFile(ociLog)
				if err == nil {
					var ociErr ociError
					if err := json.Unmarshal(ociLogData, &ociErr); err == nil {
						return -1, getOCIRuntimeError(ociErr.Msg)
					}
				}
			}
			return -1, errors.Wrapf(ss.err, "container create failed (no logs from conmon)")
		}
		logrus.Debugf("Received: %d", ss.si.Data)
		if ss.si.Data < 0 {
			if ociLog != "" {
				ociLogData, err := ioutil.ReadFile(ociLog)
				if err == nil {
					var ociErr ociError
					if err := json.Unmarshal(ociLogData, &ociErr); err == nil {
						return ss.si.Data, getOCIRuntimeError(ociErr.Msg)
					}
				}
			}
			// If we failed to parse the JSON errors, then print the output as it is
			if ss.si.Message != "" {
				return ss.si.Data, getOCIRuntimeError(ss.si.Message)
			}
			return ss.si.Data, errors.Wrapf(define.ErrInternal, "container create failed")
		}
		data = ss.si.Data
	case <-time.After(define.ContainerCreateTimeout):
		return -1, errors.Wrapf(define.ErrInternal, "container creation timeout")
	}
	return data, nil
}

// writeConmonPipeData writes nonce data to a pipe
func writeConmonPipeData(pipe *os.File) error {
	someData := []byte{0}
	_, err := pipe.Write(someData)
	return err
}

// formatRuntimeOpts prepends opts passed to it with --runtime-opt for passing to conmon
func formatRuntimeOpts(opts ...string) []string {
	args := make([]string, 0, len(opts)*2)
	for _, o := range opts {
		args = append(args, "--runtime-opt", o)
	}
	return args
}

// getConmonVersion returns a string representation of the conmon version.
func (r *ConmonOCIRuntime) getConmonVersion() (string, error) {
	output, err := utils.ExecCmd(r.conmonPath, "--version")
	if err != nil {
		return "", err
	}
	return strings.TrimSuffix(strings.Replace(output, "\n", ", ", 1), "\n"), nil
}

// getOCIRuntimeVersion returns a string representation of the OCI runtime's
// version.
func (r *ConmonOCIRuntime) getOCIRuntimeVersion() (string, error) {
	output, err := utils.ExecCmd(r.path, "--version")
	if err != nil {
		return "", err
	}
	return strings.TrimSuffix(output, "\n"), nil
}

// Copy data from container to HTTP connection, for terminal attach.
// Container is the container's attach socket connection, http is a buffer for
// the HTTP connection. cid is the ID of the container the attach session is
// running for (used solely for error messages).
func httpAttachTerminalCopy(container *net.UnixConn, http *bufio.ReadWriter, cid string) error {
	buf := make([]byte, bufferSize)
	for {
		numR, err := container.Read(buf)
		logrus.Debugf("Read fd(%d) %d/%d bytes for container %s", int(buf[0]), numR, len(buf), cid)

		if numR > 0 {
			switch buf[0] {
			case AttachPipeStdout:
				// Do nothing
			default:
				logrus.Errorf("Received unexpected attach type %+d, discarding %d bytes", buf[0], numR)
				continue
			}

			numW, err2 := http.Write(buf[1:numR])
			if err2 != nil {
				if err != nil {
					logrus.Errorf("Error reading container %s STDOUT: %v", cid, err)
				}
				return err2
			} else if numW+1 != numR {
				return io.ErrShortWrite
			}
			// We need to force the buffer to write immediately, so
			// there isn't a delay on the terminal side.
			if err2 := http.Flush(); err2 != nil {
				if err != nil {
					logrus.Errorf("Error reading container %s STDOUT: %v", cid, err)
				}
				return err2
			}
		}
		if err != nil {
			if err == io.EOF {
				return nil
			}
			return err
		}
	}
}

// Copy data from a container to an HTTP connection, for non-terminal attach.
// Appends a header to multiplex input.
func httpAttachNonTerminalCopy(container *net.UnixConn, http *bufio.ReadWriter, cid string, stdin, stdout, stderr bool) error {
	buf := make([]byte, bufferSize)
	for {
		numR, err := container.Read(buf)
		if numR > 0 {
			var headerBuf []byte

			// Subtract 1 because we strip the first byte (used for
			// multiplexing by Conmon).
			headerLen := uint32(numR - 1)
			// Practically speaking, we could make this buf[0] - 1,
			// but we need to validate it anyways...
			switch buf[0] {
			case AttachPipeStdin:
				headerBuf = makeHTTPAttachHeader(0, headerLen)
				if !stdin {
					continue
				}
			case AttachPipeStdout:
				if !stdout {
					continue
				}
				headerBuf = makeHTTPAttachHeader(1, headerLen)
			case AttachPipeStderr:
				if !stderr {
					continue
				}
				headerBuf = makeHTTPAttachHeader(2, headerLen)
			default:
				logrus.Errorf("Received unexpected attach type %+d, discarding %d bytes", buf[0], numR)
				continue
			}

			numH, err2 := http.Write(headerBuf)
			if err2 != nil {
				if err != nil {
					logrus.Errorf("Error reading container %s standard streams: %v", cid, err)
				}

				return err2
			}
			// Hardcoding header length is pretty gross, but
			// fast. Should be safe, as this is a fixed part
			// of the protocol.
			if numH != 8 {
				if err != nil {
					logrus.Errorf("Error reading container %s standard streams: %v", cid, err)
				}

				return io.ErrShortWrite
			}

			numW, err2 := http.Write(buf[1:numR])
			if err2 != nil {
				if err != nil {
					logrus.Errorf("Error reading container %s standard streams: %v", cid, err)
				}

				return err2
			} else if numW+1 != numR {
				if err != nil {
					logrus.Errorf("Error reading container %s standard streams: %v", cid, err)
				}

				return io.ErrShortWrite
			}
			// We need to force the buffer to write immediately, so
			// there isn't a delay on the terminal side.
			if err2 := http.Flush(); err2 != nil {
				if err != nil {
					logrus.Errorf("Error reading container %s STDOUT: %v", cid, err)
				}
				return err2
			}
		}
		if err != nil {
			if err == io.EOF {
				return nil
			}

			return err
		}
	}
}