aboutsummaryrefslogtreecommitdiff
path: root/libpod/container_internal.go
blob: 0e883588c7bcf6299327249f4967f71678535392 (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
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
package libpod

import (
	"bytes"
	"context"
	"fmt"
	"io"
	"io/ioutil"
	"os"
	"path/filepath"
	"strconv"
	"strings"
	"time"

	"github.com/containers/libpod/libpod/define"
	"github.com/containers/libpod/libpod/events"
	"github.com/containers/libpod/pkg/cgroups"
	"github.com/containers/libpod/pkg/ctime"
	"github.com/containers/libpod/pkg/hooks"
	"github.com/containers/libpod/pkg/hooks/exec"
	"github.com/containers/libpod/pkg/rootless"
	"github.com/containers/storage"
	"github.com/containers/storage/pkg/archive"
	"github.com/containers/storage/pkg/mount"
	securejoin "github.com/cyphar/filepath-securejoin"
	spec "github.com/opencontainers/runtime-spec/specs-go"
	"github.com/opencontainers/runtime-tools/generate"
	"github.com/opencontainers/selinux/go-selinux/label"
	"github.com/opentracing/opentracing-go"
	"github.com/pkg/errors"
	"github.com/sirupsen/logrus"
)

const (
	// name of the directory holding the artifacts
	artifactsDir      = "artifacts"
	execDirPermission = 0755
)

// rootFsSize gets the size of the container's root filesystem
// A container FS is split into two parts.  The first is the top layer, a
// mutable layer, and the rest is the RootFS: the set of immutable layers
// that make up the image on which the container is based.
func (c *Container) rootFsSize() (int64, error) {
	if c.config.Rootfs != "" {
		return 0, nil
	}
	if c.runtime.store == nil {
		return 0, nil
	}

	container, err := c.runtime.store.Container(c.ID())
	if err != nil {
		return 0, err
	}

	// Ignore the size of the top layer.   The top layer is a mutable RW layer
	// and is not considered a part of the rootfs
	rwLayer, err := c.runtime.store.Layer(container.LayerID)
	if err != nil {
		return 0, err
	}
	layer, err := c.runtime.store.Layer(rwLayer.Parent)
	if err != nil {
		return 0, err
	}

	size := int64(0)
	for layer.Parent != "" {
		layerSize, err := c.runtime.store.DiffSize(layer.Parent, layer.ID)
		if err != nil {
			return size, errors.Wrapf(err, "getting diffsize of layer %q and its parent %q", layer.ID, layer.Parent)
		}
		size += layerSize
		layer, err = c.runtime.store.Layer(layer.Parent)
		if err != nil {
			return 0, err
		}
	}
	// Get the size of the last layer.  Has to be outside of the loop
	// because the parent of the last layer is "", and lstore.Get("")
	// will return an error.
	layerSize, err := c.runtime.store.DiffSize(layer.Parent, layer.ID)
	return size + layerSize, err
}

// rwSize gets the size of the mutable top layer of the container.
func (c *Container) rwSize() (int64, error) {
	if c.config.Rootfs != "" {
		var size int64
		err := filepath.Walk(c.config.Rootfs, func(path string, info os.FileInfo, err error) error {
			if err != nil {
				return err
			}
			size += info.Size()
			return nil
		})
		return size, err
	}

	container, err := c.runtime.store.Container(c.ID())
	if err != nil {
		return 0, err
	}

	// The top layer of a container is
	// the only readable/writeable layer, all others are immutable.
	rwLayer, err := c.runtime.store.Layer(container.LayerID)
	if err != nil {
		return 0, err
	}

	// Get the size of the top layer by calculating the size of the diff
	// between the layer and its parent.
	return c.runtime.store.DiffSize(rwLayer.Parent, rwLayer.ID)
}

// bundlePath returns the path to the container's root filesystem - where the OCI spec will be
// placed, amongst other things
func (c *Container) bundlePath() string {
	return c.config.StaticDir
}

// ControlSocketPath returns the path to the containers control socket for things like tty
// resizing
func (c *Container) ControlSocketPath() string {
	return filepath.Join(c.bundlePath(), "ctl")
}

// CheckpointPath returns the path to the directory containing the checkpoint
func (c *Container) CheckpointPath() string {
	return filepath.Join(c.bundlePath(), "checkpoint")
}

// AttachSocketPath retrieves the path of the container's attach socket
func (c *Container) AttachSocketPath() (string, error) {
	return c.ociRuntime.AttachSocketPath(c)
}

// exitFilePath gets the path to the container's exit file
func (c *Container) exitFilePath() (string, error) {
	return c.ociRuntime.ExitFilePath(c)
}

// create a bundle path and associated files for an exec session
func (c *Container) createExecBundle(sessionID string) (err error) {
	bundlePath := c.execBundlePath(sessionID)
	if createErr := os.MkdirAll(bundlePath, execDirPermission); createErr != nil {
		return createErr
	}
	defer func() {
		if err != nil {
			if err2 := os.RemoveAll(bundlePath); err != nil {
				logrus.Warnf("error removing exec bundle after creation caused another error: %v", err2)
			}
		}
	}()
	if err2 := os.MkdirAll(c.execExitFileDir(sessionID), execDirPermission); err2 != nil {
		// The directory is allowed to exist
		if !os.IsExist(err2) {
			err = errors.Wrapf(err2, "error creating OCI runtime exit file path %s", c.execExitFileDir(sessionID))
		}
	}
	return
}

// cleanup an exec session after its done
func (c *Container) cleanupExecBundle(sessionID string) error {
	if err := os.RemoveAll(c.execBundlePath(sessionID)); err != nil && !os.IsNotExist(err) {
		return err
	}

	return c.ociRuntime.ExecContainerCleanup(c, sessionID)
}

// the path to a containers exec session bundle
func (c *Container) execBundlePath(sessionID string) string {
	return filepath.Join(c.bundlePath(), sessionID)
}

// Get PID file path for a container's exec session
func (c *Container) execPidPath(sessionID string) string {
	return filepath.Join(c.execBundlePath(sessionID), "exec_pid")
}

// the log path for an exec session
func (c *Container) execLogPath(sessionID string) string {
	return filepath.Join(c.execBundlePath(sessionID), "exec_log")
}

// the socket conmon creates for an exec session
func (c *Container) execAttachSocketPath(sessionID string) (string, error) {
	return c.ociRuntime.ExecAttachSocketPath(c, sessionID)
}

// execExitFileDir gets the path to the container's exit file
func (c *Container) execExitFileDir(sessionID string) string {
	return filepath.Join(c.execBundlePath(sessionID), "exit")
}

// execOCILog returns the file path for the exec sessions oci log
func (c *Container) execOCILog(sessionID string) string {
	if !c.ociRuntime.SupportsJSONErrors() {
		return ""
	}
	return filepath.Join(c.execBundlePath(sessionID), "oci-log")
}

// readExecExitCode reads the exit file for an exec session and returns
// the exit code
func (c *Container) readExecExitCode(sessionID string) (int, error) {
	exitFile := filepath.Join(c.execExitFileDir(sessionID), c.ID())
	chWait := make(chan error)
	defer close(chWait)

	_, err := WaitForFile(exitFile, chWait, time.Second*5)
	if err != nil {
		return -1, err
	}
	ec, err := ioutil.ReadFile(exitFile)
	if err != nil {
		return -1, err
	}
	ecInt, err := strconv.Atoi(string(ec))
	if err != nil {
		return -1, err
	}
	return ecInt, nil
}

// Wait for the container's exit file to appear.
// When it does, update our state based on it.
func (c *Container) waitForExitFileAndSync() error {
	exitFile, err := c.exitFilePath()
	if err != nil {
		return err
	}

	chWait := make(chan error)
	defer close(chWait)

	_, err = WaitForFile(exitFile, chWait, time.Second*5)
	if err != nil {
		// Exit file did not appear
		// Reset our state
		c.state.ExitCode = -1
		c.state.FinishedTime = time.Now()
		c.state.State = define.ContainerStateStopped

		if err2 := c.save(); err2 != nil {
			logrus.Errorf("Error saving container %s state: %v", c.ID(), err2)
		}

		return err
	}

	if err := c.checkExitFile(); err != nil {
		return err
	}

	return c.save()
}

// Handle the container exit file.
// The exit file is used to supply container exit time and exit code.
// This assumes the exit file already exists.
func (c *Container) handleExitFile(exitFile string, fi os.FileInfo) error {
	c.state.FinishedTime = ctime.Created(fi)
	statusCodeStr, err := ioutil.ReadFile(exitFile)
	if err != nil {
		return errors.Wrapf(err, "failed to read exit file for container %s", c.ID())
	}
	statusCode, err := strconv.Atoi(string(statusCodeStr))
	if err != nil {
		return errors.Wrapf(err, "error converting exit status code (%q) for container %s to int",
			c.ID(), statusCodeStr)
	}
	c.state.ExitCode = int32(statusCode)

	oomFilePath := filepath.Join(c.bundlePath(), "oom")
	if _, err = os.Stat(oomFilePath); err == nil {
		c.state.OOMKilled = true
	}

	c.state.Exited = true

	// Write an event for the container's death
	c.newContainerExitedEvent(c.state.ExitCode)

	return nil
}

// Handle container restart policy.
// This is called when a container has exited, and was not explicitly stopped by
// an API call to stop the container or pod it is in.
func (c *Container) handleRestartPolicy(ctx context.Context) (restarted bool, err error) {
	// If we did not get a restart policy match, exit immediately.
	// Do the same if we're not a policy that restarts.
	if !c.state.RestartPolicyMatch ||
		c.config.RestartPolicy == RestartPolicyNo ||
		c.config.RestartPolicy == RestartPolicyNone {
		return false, nil
	}

	// If we're RestartPolicyOnFailure, we need to check retries and exit
	// code.
	if c.config.RestartPolicy == RestartPolicyOnFailure {
		if c.state.ExitCode == 0 {
			return false, nil
		}

		// If we don't have a max retries set, continue
		if c.config.RestartRetries > 0 {
			if c.state.RestartCount < c.config.RestartRetries {
				logrus.Debugf("Container %s restart policy trigger: on retry %d (of %d)",
					c.ID(), c.state.RestartCount, c.config.RestartRetries)
			} else {
				logrus.Debugf("Container %s restart policy trigger: retries exhausted", c.ID())
				return false, nil
			}
		}
	}

	logrus.Debugf("Restarting container %s due to restart policy %s", c.ID(), c.config.RestartPolicy)

	// Need to check if dependencies are alive.
	if err = c.checkDependenciesAndHandleError(ctx); err != nil {
		return false, err
	}

	// Is the container running again?
	// If so, we don't have to do anything
	if c.ensureState(define.ContainerStateRunning, define.ContainerStatePaused) {
		return false, nil
	} else if c.state.State == define.ContainerStateUnknown {
		return false, errors.Wrapf(define.ErrInternal, "invalid container state encountered in restart attempt!")
	}

	c.newContainerEvent(events.Restart)

	// Increment restart count
	c.state.RestartCount += 1
	logrus.Debugf("Container %s now on retry %d", c.ID(), c.state.RestartCount)
	if err := c.save(); err != nil {
		return false, err
	}

	defer func() {
		if err != nil {
			if err2 := c.cleanup(ctx); err2 != nil {
				logrus.Errorf("error cleaning up container %s: %v", c.ID(), err2)
			}
		}
	}()
	if err := c.prepare(); err != nil {
		return false, err
	}

	if c.state.State == define.ContainerStateStopped {
		// Reinitialize the container if we need to
		if err := c.reinit(ctx, true); err != nil {
			return false, err
		}
	} else if c.ensureState(define.ContainerStateConfigured, define.ContainerStateExited) {
		// Initialize the container
		if err := c.init(ctx, true); err != nil {
			return false, err
		}
	}
	if err := c.start(); err != nil {
		return false, err
	}
	return true, nil
}

// Ensure that the container is in a specific state or state.
// Returns true if the container is in one of the given states,
// or false otherwise.
func (c *Container) ensureState(states ...define.ContainerStatus) bool {
	for _, state := range states {
		if state == c.state.State {
			return true
		}
	}
	return false
}

// Sync this container with on-disk state and runtime status
// Should only be called with container lock held
// This function should suffice to ensure a container's state is accurate and
// it is valid for use.
func (c *Container) syncContainer() error {
	if err := c.runtime.state.UpdateContainer(c); err != nil {
		return err
	}
	// If runtime knows about the container, update its status in runtime
	// And then save back to disk
	if c.ensureState(define.ContainerStateCreated, define.ContainerStateRunning, define.ContainerStateStopped, define.ContainerStatePaused) {
		oldState := c.state.State

		if err := c.checkExitFile(); err != nil {
			return err
		}

		// Only save back to DB if state changed
		if c.state.State != oldState {
			// Check for a restart policy match
			if c.config.RestartPolicy != RestartPolicyNone && c.config.RestartPolicy != RestartPolicyNo &&
				(oldState == define.ContainerStateRunning || oldState == define.ContainerStatePaused) &&
				(c.state.State == define.ContainerStateStopped || c.state.State == define.ContainerStateExited) &&
				!c.state.StoppedByUser {
				c.state.RestartPolicyMatch = true
			}

			if err := c.save(); err != nil {
				return err
			}
		}
	}

	if !c.valid {
		return errors.Wrapf(define.ErrCtrRemoved, "container %s is not valid", c.ID())
	}

	return nil
}

// Create container root filesystem for use
func (c *Container) setupStorage(ctx context.Context) error {
	span, _ := opentracing.StartSpanFromContext(ctx, "setupStorage")
	span.SetTag("type", "container")
	defer span.Finish()

	if !c.valid {
		return errors.Wrapf(define.ErrCtrRemoved, "container %s is not valid", c.ID())
	}

	if c.state.State != define.ContainerStateConfigured {
		return errors.Wrapf(define.ErrCtrStateInvalid, "container %s must be in Configured state to have storage set up", c.ID())
	}

	// Need both an image ID and image name, plus a bool telling us whether to use the image configuration
	if c.config.Rootfs == "" && (c.config.RootfsImageID == "" || c.config.RootfsImageName == "") {
		return errors.Wrapf(define.ErrInvalidArg, "must provide image ID and image name to use an image")
	}

	options := storage.ContainerOptions{
		IDMappingOptions: storage.IDMappingOptions{
			HostUIDMapping: true,
			HostGIDMapping: true,
		},
		LabelOpts: c.config.LabelOpts,
	}
	if c.restoreFromCheckpoint {
		// If restoring from a checkpoint, the root file-system
		// needs to be mounted with the same SELinux labels as
		// it was mounted previously.
		if options.Flags == nil {
			options.Flags = make(map[string]interface{})
		}
		options.Flags["ProcessLabel"] = c.config.ProcessLabel
		options.Flags["MountLabel"] = c.config.MountLabel
	}
	if c.config.Privileged {
		privOpt := func(opt string) bool {
			for _, privopt := range []string{"nodev", "nosuid", "noexec"} {
				if opt == privopt {
					return true
				}
			}
			return false
		}

		defOptions, err := storage.GetMountOptions(c.runtime.store.GraphDriverName(), c.runtime.store.GraphOptions())
		if err != nil {
			return errors.Wrapf(err, "error getting default mount options")
		}
		var newOptions []string
		for _, opt := range defOptions {
			if !privOpt(opt) {
				newOptions = append(newOptions, opt)
			}
		}
		options.MountOpts = newOptions
	}

	if c.config.Rootfs == "" {
		options.IDMappingOptions = c.config.IDMappings
	}
	containerInfo, err := c.runtime.storageService.CreateContainerStorage(ctx, c.runtime.imageContext, c.config.RootfsImageName, c.config.RootfsImageID, c.config.Name, c.config.ID, options)
	if err != nil {
		return errors.Wrapf(err, "error creating container storage")
	}

	if len(c.config.IDMappings.UIDMap) != 0 || len(c.config.IDMappings.GIDMap) != 0 {
		if err := os.Chown(containerInfo.RunDir, c.RootUID(), c.RootGID()); err != nil {
			return err
		}

		if err := os.Chown(containerInfo.Dir, c.RootUID(), c.RootGID()); err != nil {
			return err
		}
	}

	c.config.ProcessLabel = containerInfo.ProcessLabel
	c.config.MountLabel = containerInfo.MountLabel
	c.config.StaticDir = containerInfo.Dir
	c.state.RunDir = containerInfo.RunDir

	// Set the default Entrypoint and Command
	if containerInfo.Config != nil {
		if c.config.Entrypoint == nil {
			c.config.Entrypoint = containerInfo.Config.Config.Entrypoint
		}
		if c.config.Command == nil {
			c.config.Command = containerInfo.Config.Config.Cmd
		}
	}

	artifacts := filepath.Join(c.config.StaticDir, artifactsDir)
	if err := os.MkdirAll(artifacts, 0755); err != nil {
		return errors.Wrapf(err, "error creating artifacts directory %q", artifacts)
	}

	return nil
}

// Tear down a container's storage prior to removal
func (c *Container) teardownStorage() error {
	if c.ensureState(define.ContainerStateRunning, define.ContainerStatePaused) {
		return errors.Wrapf(define.ErrCtrStateInvalid, "cannot remove storage for container %s as it is running or paused", c.ID())
	}

	artifacts := filepath.Join(c.config.StaticDir, artifactsDir)
	if err := os.RemoveAll(artifacts); err != nil {
		return errors.Wrapf(err, "error removing container %s artifacts %q", c.ID(), artifacts)
	}

	if err := c.cleanupStorage(); err != nil {
		return errors.Wrapf(err, "failed to cleanup container %s storage", c.ID())
	}

	if err := c.runtime.storageService.DeleteContainer(c.ID()); err != nil {
		// If the container has already been removed, warn but do not
		// error - we wanted it gone, it is already gone.
		// Potentially another tool using containers/storage already
		// removed it?
		if errors.Cause(err) == storage.ErrNotAContainer || errors.Cause(err) == storage.ErrContainerUnknown {
			logrus.Warnf("Storage for container %s already removed", c.ID())
			return nil
		}

		return errors.Wrapf(err, "error removing container %s root filesystem", c.ID())
	}

	return nil
}

// Reset resets state fields to default values
// It is performed before a refresh and clears the state after a reboot
// It does not save the results - assumes the database will do that for us
func resetState(state *ContainerState) error {
	state.PID = 0
	state.ConmonPID = 0
	state.Mountpoint = ""
	state.Mounted = false
	if state.State != define.ContainerStateExited {
		state.State = define.ContainerStateConfigured
	}
	state.ExecSessions = make(map[string]*ExecSession)
	state.NetworkStatus = nil
	state.BindMounts = make(map[string]string)
	state.StoppedByUser = false
	state.RestartPolicyMatch = false
	state.RestartCount = 0

	return nil
}

// Refresh refreshes the container's state after a restart.
// Refresh cannot perform any operations that would lock another container.
// We cannot guarantee any other container has a valid lock at the time it is
// running.
func (c *Container) refresh() error {
	// Don't need a full sync, but we do need to update from the database to
	// pick up potentially-missing container state
	if err := c.runtime.state.UpdateContainer(c); err != nil {
		return err
	}

	if !c.valid {
		return errors.Wrapf(define.ErrCtrRemoved, "container %s is not valid - may have been removed", c.ID())
	}

	// We need to get the container's temporary directory from c/storage
	// It was lost in the reboot and must be recreated
	dir, err := c.runtime.storageService.GetRunDir(c.ID())
	if err != nil {
		return errors.Wrapf(err, "error retrieving temporary directory for container %s", c.ID())
	}
	c.state.RunDir = dir

	if len(c.config.IDMappings.UIDMap) != 0 || len(c.config.IDMappings.GIDMap) != 0 {
		info, err := os.Stat(c.runtime.config.TmpDir)
		if err != nil {
			return errors.Wrapf(err, "cannot stat `%s`", c.runtime.config.TmpDir)
		}
		if err := os.Chmod(c.runtime.config.TmpDir, info.Mode()|0111); err != nil {
			return errors.Wrapf(err, "cannot chmod `%s`", c.runtime.config.TmpDir)
		}
		root := filepath.Join(c.runtime.config.TmpDir, "containers-root", c.ID())
		if err := os.MkdirAll(root, 0755); err != nil {
			return errors.Wrapf(err, "error creating userNS tmpdir for container %s", c.ID())
		}
		if err := os.Chown(root, c.RootUID(), c.RootGID()); err != nil {
			return err
		}
	}

	// We need to pick up a new lock
	lock, err := c.runtime.lockManager.AllocateAndRetrieveLock(c.config.LockID)
	if err != nil {
		return errors.Wrapf(err, "error acquiring lock %d for container %s", c.config.LockID, c.ID())
	}
	c.lock = lock

	if err := c.save(); err != nil {
		return errors.Wrapf(err, "error refreshing state for container %s", c.ID())
	}

	// Remove ctl and attach files, which may persist across reboot
	if err := c.removeConmonFiles(); err != nil {
		return err
	}

	if rootless.IsRootless() {
		return nil
	}

	return c.refreshCNI()
}

// Remove conmon attach socket and terminal resize FIFO
// This is necessary for restarting containers
func (c *Container) removeConmonFiles() error {
	// Files are allowed to not exist, so ignore ENOENT
	attachFile := filepath.Join(c.bundlePath(), "attach")
	if err := os.Remove(attachFile); err != nil && !os.IsNotExist(err) {
		return errors.Wrapf(err, "error removing container %s attach file", c.ID())
	}

	ctlFile := filepath.Join(c.bundlePath(), "ctl")
	if err := os.Remove(ctlFile); err != nil && !os.IsNotExist(err) {
		return errors.Wrapf(err, "error removing container %s ctl file", c.ID())
	}

	winszFile := filepath.Join(c.bundlePath(), "winsz")
	if err := os.Remove(winszFile); err != nil && !os.IsNotExist(err) {
		return errors.Wrapf(err, "error removing container %s winsz file", c.ID())
	}

	oomFile := filepath.Join(c.bundlePath(), "oom")
	if err := os.Remove(oomFile); err != nil && !os.IsNotExist(err) {
		return errors.Wrapf(err, "error removing container %s OOM file", c.ID())
	}

	// Remove the exit file so we don't leak memory in tmpfs
	exitFile, err := c.exitFilePath()
	if err != nil {
		return err
	}
	if err := os.Remove(exitFile); err != nil && !os.IsNotExist(err) {
		return errors.Wrapf(err, "error removing container %s exit file", c.ID())
	}

	return nil
}

func (c *Container) export(path string) error {
	mountPoint := c.state.Mountpoint
	if !c.state.Mounted {
		containerMount, err := c.runtime.store.Mount(c.ID(), c.config.MountLabel)
		if err != nil {
			return errors.Wrapf(err, "error mounting container %q", c.ID())
		}
		mountPoint = containerMount
		defer func() {
			if _, err := c.runtime.store.Unmount(c.ID(), false); err != nil {
				logrus.Errorf("error unmounting container %q: %v", c.ID(), err)
			}
		}()
	}

	input, err := archive.Tar(mountPoint, archive.Uncompressed)
	if err != nil {
		return errors.Wrapf(err, "error reading container directory %q", c.ID())
	}

	outFile, err := os.Create(path)
	if err != nil {
		return errors.Wrapf(err, "error creating file %q", path)
	}
	defer outFile.Close()

	_, err = io.Copy(outFile, input)
	return err
}

// Get path of artifact with a given name for this container
func (c *Container) getArtifactPath(name string) string {
	return filepath.Join(c.config.StaticDir, artifactsDir, name)
}

// Used with Wait() to determine if a container has exited
func (c *Container) isStopped() (bool, error) {
	if !c.batched {
		c.lock.Lock()
		defer c.lock.Unlock()
	}
	err := c.syncContainer()
	if err != nil {
		return true, err
	}

	return !c.ensureState(define.ContainerStateRunning, define.ContainerStatePaused), nil
}

// save container state to the database
func (c *Container) save() error {
	if err := c.runtime.state.SaveContainer(c); err != nil {
		return errors.Wrapf(err, "error saving container %s state", c.ID())
	}
	return nil
}

// Checks the container is in the right state, then initializes the container in preparation to start the container.
// If recursive is true, each of the containers dependencies will be started.
// Otherwise, this function will return with error if there are dependencies of this container that aren't running.
func (c *Container) prepareToStart(ctx context.Context, recursive bool) (err error) {
	// Container must be created or stopped to be started
	if !c.ensureState(define.ContainerStateConfigured, define.ContainerStateCreated, define.ContainerStateStopped, define.ContainerStateExited) {
		return errors.Wrapf(define.ErrCtrStateInvalid, "container %s must be in Created or Stopped state to be started", c.ID())
	}

	if !recursive {
		if err := c.checkDependenciesAndHandleError(ctx); err != nil {
			return err
		}
	} else {
		if err := c.startDependencies(ctx); err != nil {
			return err
		}
	}

	defer func() {
		if err != nil {
			if err2 := c.cleanup(ctx); err2 != nil {
				logrus.Errorf("error cleaning up container %s: %v", c.ID(), err2)
			}
		}
	}()

	if err := c.prepare(); err != nil {
		return err
	}

	if c.state.State == define.ContainerStateStopped {
		// Reinitialize the container if we need to
		if err := c.reinit(ctx, false); err != nil {
			return err
		}
	} else if c.ensureState(define.ContainerStateConfigured, define.ContainerStateExited) {
		// Or initialize it if necessary
		if err := c.init(ctx, false); err != nil {
			return err
		}
	}
	return nil
}

// checks dependencies are running and prints a helpful message
func (c *Container) checkDependenciesAndHandleError(ctx context.Context) error {
	notRunning, err := c.checkDependenciesRunning()
	if err != nil {
		return errors.Wrapf(err, "error checking dependencies for container %s", c.ID())
	}
	if len(notRunning) > 0 {
		depString := strings.Join(notRunning, ",")
		return errors.Wrapf(define.ErrCtrStateInvalid, "some dependencies of container %s are not started: %s", c.ID(), depString)
	}

	return nil
}

// Recursively start all dependencies of a container so the container can be started.
func (c *Container) startDependencies(ctx context.Context) error {
	depCtrIDs := c.Dependencies()
	if len(depCtrIDs) == 0 {
		return nil
	}

	depVisitedCtrs := make(map[string]*Container)
	if err := c.getAllDependencies(depVisitedCtrs); err != nil {
		return errors.Wrapf(err, "error starting dependency for container %s", c.ID())
	}

	// Because of how Go handles passing slices through functions, a slice cannot grow between function calls
	// without clunky syntax. Circumnavigate this by translating the map to a slice for buildContainerGraph
	depCtrs := make([]*Container, 0)
	for _, ctr := range depVisitedCtrs {
		depCtrs = append(depCtrs, ctr)
	}

	// Build a dependency graph of containers
	graph, err := BuildContainerGraph(depCtrs)
	if err != nil {
		return errors.Wrapf(err, "error generating dependency graph for container %s", c.ID())
	}

	// If there are no containers without dependencies, we can't start
	// Error out
	if len(graph.noDepNodes) == 0 {
		// we have no dependencies that need starting, go ahead and return
		if len(graph.nodes) == 0 {
			return nil
		}
		return errors.Wrapf(define.ErrNoSuchCtr, "All dependencies have dependencies of %s", c.ID())
	}

	ctrErrors := make(map[string]error)
	ctrsVisited := make(map[string]bool)

	// Traverse the graph beginning at nodes with no dependencies
	for _, node := range graph.noDepNodes {
		startNode(ctx, node, false, ctrErrors, ctrsVisited, true)
	}

	if len(ctrErrors) > 0 {
		logrus.Errorf("error starting some container dependencies")
		for _, e := range ctrErrors {
			logrus.Errorf("%q", e)
		}
		return errors.Wrapf(define.ErrInternal, "error starting some containers")
	}
	return nil
}

// getAllDependencies is a precursor to starting dependencies.
// To start a container with all of its dependencies, we need to recursively find all dependencies
// a container has, as well as each of those containers' dependencies, and so on
// To do so, keep track of containers already visisted (so there aren't redundant state lookups),
// and recursively search until we have reached the leafs of every dependency node.
// Since we need to start all dependencies for our original container to successfully start, we propegate any errors
// in looking up dependencies.
// Note: this function is currently meant as a robust solution to a narrow problem: start an infra-container when
// a container in the pod is run. It has not been tested for performance past one level, so expansion of recursive start
// must be tested first.
func (c *Container) getAllDependencies(visited map[string]*Container) error {
	depIDs := c.Dependencies()
	if len(depIDs) == 0 {
		return nil
	}
	for _, depID := range depIDs {
		if _, ok := visited[depID]; !ok {
			dep, err := c.runtime.state.Container(depID)
			if err != nil {
				return err
			}
			status, err := dep.State()
			if err != nil {
				return err
			}
			// if the dependency is already running, we can assume its dependencies are also running
			// so no need to add them to those we need to start
			if status != define.ContainerStateRunning {
				visited[depID] = dep
				if err := dep.getAllDependencies(visited); err != nil {
					return err
				}
			}
		}
	}
	return nil
}

// Check if a container's dependencies are running
// Returns a []string containing the IDs of dependencies that are not running
func (c *Container) checkDependenciesRunning() ([]string, error) {
	deps := c.Dependencies()
	notRunning := []string{}

	// We were not passed a set of dependency containers
	// Make it ourselves
	depCtrs := make(map[string]*Container, len(deps))
	for _, dep := range deps {
		// Get the dependency container
		depCtr, err := c.runtime.state.Container(dep)
		if err != nil {
			return nil, errors.Wrapf(err, "error retrieving dependency %s of container %s from state", dep, c.ID())
		}

		// Check the status
		state, err := depCtr.State()
		if err != nil {
			return nil, errors.Wrapf(err, "error retrieving state of dependency %s of container %s", dep, c.ID())
		}
		if state != define.ContainerStateRunning {
			notRunning = append(notRunning, dep)
		}
		depCtrs[dep] = depCtr
	}

	return notRunning, nil
}

func (c *Container) completeNetworkSetup() error {
	netDisabled, err := c.NetworkDisabled()
	if err != nil {
		return err
	}
	if !c.config.PostConfigureNetNS || netDisabled {
		return nil
	}
	if err := c.syncContainer(); err != nil {
		return err
	}
	if c.config.NetMode == "slirp4netns" {
		return c.runtime.setupRootlessNetNS(c)
	}
	return c.runtime.setupNetNS(c)
}

// Initialize a container, creating it in the runtime
func (c *Container) init(ctx context.Context, retainRetries bool) error {
	span, _ := opentracing.StartSpanFromContext(ctx, "init")
	span.SetTag("struct", "container")
	defer span.Finish()

	// Unconditionally remove conmon temporary files.
	// We've been running into far too many issues where they block startup.
	if err := c.removeConmonFiles(); err != nil {
		return err
	}

	// Generate the OCI newSpec
	newSpec, err := c.generateSpec(ctx)
	if err != nil {
		return err
	}

	// Save the OCI newSpec to disk
	if err := c.saveSpec(newSpec); err != nil {
		return err
	}

	// With the spec complete, do an OCI create
	if err := c.ociRuntime.CreateContainer(c, nil); err != nil {
		// Fedora 31 is carrying a patch to display improved error
		// messages to better handle the V2 transition. This is NOT
		// upstream in any OCI runtime.
		// TODO: Remove once runc supports cgroupsv2
		if strings.Contains(err.Error(), "this version of runc doesn't work on cgroups v2") {
			logrus.Errorf("oci runtime %q does not support CGroups V2: use system migrate to mitigate", c.ociRuntime.Name())
		}
		return err
	}

	logrus.Debugf("Created container %s in OCI runtime", c.ID())

	c.state.ExitCode = 0
	c.state.Exited = false
	c.state.State = define.ContainerStateCreated
	c.state.StoppedByUser = false
	c.state.RestartPolicyMatch = false

	if !retainRetries {
		c.state.RestartCount = 0
	}

	if err := c.save(); err != nil {
		return err
	}
	if c.config.HealthCheckConfig != nil {
		if err := c.createTimer(); err != nil {
			logrus.Error(err)
		}
	}

	defer c.newContainerEvent(events.Init)
	return c.completeNetworkSetup()
}

// Clean up a container in the OCI runtime.
// Deletes the container in the runtime, and resets its state to Exited.
// The container can be restarted cleanly after this.
func (c *Container) cleanupRuntime(ctx context.Context) error {
	span, _ := opentracing.StartSpanFromContext(ctx, "cleanupRuntime")
	span.SetTag("struct", "container")
	defer span.Finish()

	// If the container is not ContainerStateStopped or
	// ContainerStateCreated, do nothing.
	if !c.ensureState(define.ContainerStateStopped, define.ContainerStateCreated) {
		return nil
	}

	// If necessary, delete attach and ctl files
	if err := c.removeConmonFiles(); err != nil {
		return err
	}

	if err := c.delete(ctx); err != nil {
		return err
	}

	// If we were Stopped, we are now Exited, as we've removed ourself
	// from the runtime.
	// If we were Created, we are now Configured.
	if c.state.State == define.ContainerStateStopped {
		c.state.State = define.ContainerStateExited
	} else if c.state.State == define.ContainerStateCreated {
		c.state.State = define.ContainerStateConfigured
	}

	if c.valid {
		if err := c.save(); err != nil {
			return err
		}
	}

	logrus.Debugf("Successfully cleaned up container %s", c.ID())

	return nil
}

// Reinitialize a container.
// Deletes and recreates a container in the runtime.
// Should only be done on ContainerStateStopped containers.
// Not necessary for ContainerStateExited - the container has already been
// removed from the runtime, so init() can proceed freely.
func (c *Container) reinit(ctx context.Context, retainRetries bool) error {
	span, _ := opentracing.StartSpanFromContext(ctx, "reinit")
	span.SetTag("struct", "container")
	defer span.Finish()

	logrus.Debugf("Recreating container %s in OCI runtime", c.ID())

	if err := c.cleanupRuntime(ctx); err != nil {
		return err
	}

	// Initialize the container again
	return c.init(ctx, retainRetries)
}

// Initialize (if necessary) and start a container
// Performs all necessary steps to start a container that is not running
// Does not lock or check validity
func (c *Container) initAndStart(ctx context.Context) (err error) {
	// If we are ContainerStateUnknown, throw an error
	if c.state.State == define.ContainerStateUnknown {
		return errors.Wrapf(define.ErrCtrStateInvalid, "container %s is in an unknown state", c.ID())
	} else if c.state.State == define.ContainerStateRemoving {
		return errors.Wrapf(define.ErrCtrStateInvalid, "cannot start container %s as it is being removed", c.ID())
	}

	// If we are running, do nothing
	if c.state.State == define.ContainerStateRunning {
		return nil
	}
	// If we are paused, throw an error
	if c.state.State == define.ContainerStatePaused {
		return errors.Wrapf(define.ErrCtrStateInvalid, "cannot start paused container %s", c.ID())
	}

	defer func() {
		if err != nil {
			if err2 := c.cleanup(ctx); err2 != nil {
				logrus.Errorf("error cleaning up container %s: %v", c.ID(), err2)
			}
		}
	}()

	if err := c.prepare(); err != nil {
		return err
	}

	// If we are ContainerStateStopped we need to remove from runtime
	// And reset to ContainerStateConfigured
	if c.state.State == define.ContainerStateStopped {
		logrus.Debugf("Recreating container %s in OCI runtime", c.ID())

		if err := c.reinit(ctx, false); err != nil {
			return err
		}
	} else if c.ensureState(define.ContainerStateConfigured, define.ContainerStateExited) {
		if err := c.init(ctx, false); err != nil {
			return err
		}
	}

	// Now start the container
	return c.start()
}

// Internal, non-locking function to start a container
func (c *Container) start() error {
	if c.config.Spec.Process != nil {
		logrus.Debugf("Starting container %s with command %v", c.ID(), c.config.Spec.Process.Args)
	}

	if err := c.ociRuntime.StartContainer(c); err != nil {
		return err
	}
	logrus.Debugf("Started container %s", c.ID())

	c.state.State = define.ContainerStateRunning

	if c.config.HealthCheckConfig != nil {
		if err := c.updateHealthStatus(HealthCheckStarting); err != nil {
			logrus.Error(err)
		}
		if err := c.startTimer(); err != nil {
			logrus.Error(err)
		}
	}

	defer c.newContainerEvent(events.Start)

	return c.save()
}

// Internal, non-locking function to stop container
func (c *Container) stop(timeout uint) error {
	logrus.Debugf("Stopping ctr %s (timeout %d)", c.ID(), timeout)

	// If the container is running in a PID Namespace, then killing the
	// primary pid is enough to kill the container.  If it is not running in
	// a pid namespace then the OCI Runtime needs to kill ALL processes in
	// the containers cgroup in order to make sure the container is stopped.
	all := !c.hasNamespace(spec.PIDNamespace)
	// We can't use --all if CGroups aren't present.
	// Rootless containers with CGroups v1 and NoCgroups are both cases
	// where this can happen.
	if all {
		if c.config.NoCgroups {
			all = false
		} else if rootless.IsRootless() {
			// Only do this check if we need to
			unified, err := cgroups.IsCgroup2UnifiedMode()
			if err != nil {
				return err
			}
			if !unified {
				all = false
			}
		}
	}

	if err := c.ociRuntime.StopContainer(c, timeout, all); err != nil {
		return err
	}

	c.state.PID = 0
	c.state.ConmonPID = 0
	c.state.StoppedByUser = true
	if err := c.save(); err != nil {
		return errors.Wrapf(err, "error saving container %s state after stopping", c.ID())
	}

	// Wait until we have an exit file, and sync once we do
	if err := c.waitForExitFileAndSync(); err != nil {
		return err
	}

	c.newContainerEvent(events.Stop)

	return nil
}

// Internal, non-locking function to pause a container
func (c *Container) pause() error {
	if c.config.NoCgroups {
		return errors.Wrapf(define.ErrNoCgroups, "cannot pause without using CGroups")
	}

	if rootless.IsRootless() {
		cgroupv2, err := cgroups.IsCgroup2UnifiedMode()
		if err != nil {
			return errors.Wrap(err, "failed to determine cgroupversion")
		}
		if !cgroupv2 {
			return errors.Wrap(define.ErrNoCgroups, "can not pause containers on rootless containers with cgroup V1")
		}
	}

	if err := c.ociRuntime.PauseContainer(c); err != nil {
		// TODO when using docker-py there is some sort of race/incompatibility here
		return err
	}

	logrus.Debugf("Paused container %s", c.ID())

	c.state.State = define.ContainerStatePaused

	return c.save()
}

// Internal, non-locking function to unpause a container
func (c *Container) unpause() error {
	if c.config.NoCgroups {
		return errors.Wrapf(define.ErrNoCgroups, "cannot unpause without using CGroups")
	}

	if err := c.ociRuntime.UnpauseContainer(c); err != nil {
		// TODO when using docker-py there is some sort of race/incompatibility here
		return err
	}

	logrus.Debugf("Unpaused container %s", c.ID())

	c.state.State = define.ContainerStateRunning

	return c.save()
}

// Internal, non-locking function to restart a container
func (c *Container) restartWithTimeout(ctx context.Context, timeout uint) (err error) {
	if !c.ensureState(define.ContainerStateConfigured, define.ContainerStateCreated, define.ContainerStateRunning, define.ContainerStateStopped, define.ContainerStateExited) {
		return errors.Wrapf(define.ErrCtrStateInvalid, "unable to restart a container in a paused or unknown state")
	}

	c.newContainerEvent(events.Restart)

	if c.state.State == define.ContainerStateRunning {
		conmonPID := c.state.ConmonPID
		if err := c.stop(timeout); err != nil {
			return err
		}
		// Old versions of conmon have a bug where they create the exit file before
		// closing open file descriptors causing a race condition when restarting
		// containers with open ports since we cannot bind the ports as they're not
		// yet closed by conmon.
		//
		// Killing the old conmon PID is ~okay since it forces the FDs of old conmons
		// to be closed, while it's a NOP for newer versions which should have
		// exited already.
		if conmonPID != 0 {
			// Ignore errors from FindProcess() as conmon could already have exited.
			p, err := os.FindProcess(conmonPID)
			if p != nil && err == nil {
				if err = p.Kill(); err != nil {
					logrus.Debugf("error killing conmon process: %v", err)
				}
			}
		}
	}
	defer func() {
		if err != nil {
			if err2 := c.cleanup(ctx); err2 != nil {
				logrus.Errorf("error cleaning up container %s: %v", c.ID(), err2)
			}
		}
	}()
	if err := c.prepare(); err != nil {
		return err
	}

	if c.state.State == define.ContainerStateStopped {
		// Reinitialize the container if we need to
		if err := c.reinit(ctx, false); err != nil {
			return err
		}
	} else if c.state.State == define.ContainerStateConfigured ||
		c.state.State == define.ContainerStateExited {
		// Initialize the container
		if err := c.init(ctx, false); err != nil {
			return err
		}
	}
	return c.start()
}

// mountStorage sets up the container's root filesystem
// It mounts the image and any other requested mounts
// TODO: Add ability to override mount label so we can use this for Mount() too
// TODO: Can we use this for export? Copying SHM into the export might not be
// good
func (c *Container) mountStorage() (_ string, deferredErr error) {
	var err error
	// Container already mounted, nothing to do
	if c.state.Mounted {
		return c.state.Mountpoint, nil
	}

	mounted, err := mount.Mounted(c.config.ShmDir)
	if err != nil {
		return "", errors.Wrapf(err, "unable to determine if %q is mounted", c.config.ShmDir)
	}

	if !mounted && !MountExists(c.config.Spec.Mounts, "/dev/shm") {
		shmOptions := fmt.Sprintf("mode=1777,size=%d", c.config.ShmSize)
		if err := c.mountSHM(shmOptions); err != nil {
			return "", err
		}
		if err := os.Chown(c.config.ShmDir, c.RootUID(), c.RootGID()); err != nil {
			return "", errors.Wrapf(err, "failed to chown %s", c.config.ShmDir)
		}
		defer func() {
			if deferredErr != nil {
				if err := c.unmountSHM(c.config.ShmDir); err != nil {
					logrus.Errorf("Error unmounting SHM for container %s after mount error: %v", c.ID(), err)
				}
			}
		}()
	}

	// We need to mount the container before volumes - to ensure the copyup
	// works properly.
	mountPoint := c.config.Rootfs
	if mountPoint == "" {
		mountPoint, err = c.mount()
		if err != nil {
			return "", err
		}
		defer func() {
			if deferredErr != nil {
				if err := c.unmount(false); err != nil {
					logrus.Errorf("Error unmounting container %s after mount error: %v", c.ID(), err)
				}
			}
		}()
	}

	// Request a mount of all named volumes
	for _, v := range c.config.NamedVolumes {
		vol, err := c.mountNamedVolume(v, mountPoint)
		if err != nil {
			return "", err
		}
		defer func() {
			if deferredErr == nil {
				return
			}
			vol.lock.Lock()
			if err := vol.unmount(false); err != nil {
				logrus.Errorf("Error unmounting volume %s after error mounting container %s: %v", vol.Name(), c.ID(), err)
			}
			vol.lock.Unlock()
		}()
	}

	return mountPoint, nil
}

// Mount a single named volume into the container.
// If necessary, copy up image contents into the volume.
// Does not verify that the name volume given is actually present in container
// config.
// Returns the volume that was mounted.
func (c *Container) mountNamedVolume(v *ContainerNamedVolume, mountpoint string) (*Volume, error) {
	vol, err := c.runtime.state.Volume(v.Name)
	if err != nil {
		return nil, errors.Wrapf(err, "error retrieving named volume %s for container %s", v.Name, c.ID())
	}

	vol.lock.Lock()
	defer vol.lock.Unlock()
	if vol.needsMount() {
		if err := vol.mount(); err != nil {
			return nil, errors.Wrapf(err, "error mounting volume %s for container %s", vol.Name(), c.ID())
		}
	}
	// The volume may need a copy-up. Check the state.
	if err := vol.update(); err != nil {
		return nil, err
	}
	if vol.state.NeedsCopyUp {
		logrus.Debugf("Copying up contents from container %s to volume %s", c.ID(), vol.Name())
		srcDir, err := securejoin.SecureJoin(mountpoint, v.Dest)
		if err != nil {
			return nil, errors.Wrapf(err, "error calculating destination path to copy up container %s volume %s", c.ID(), vol.Name())
		}
		if err := c.copyWithTarFromImage(srcDir, vol.MountPoint()); err != nil && !os.IsNotExist(err) {
			return nil, errors.Wrapf(err, "error copying content from container %s into volume %s", c.ID(), vol.Name())
		}

		vol.state.NeedsCopyUp = false
		if err := vol.save(); err != nil {
			return nil, err
		}
	}
	return vol, nil
}

// cleanupStorage unmounts and cleans up the container's root filesystem
func (c *Container) cleanupStorage() error {
	if !c.state.Mounted {
		// Already unmounted, do nothing
		logrus.Debugf("Container %s storage is already unmounted, skipping...", c.ID())
		return nil
	}

	var cleanupErr error

	for _, containerMount := range c.config.Mounts {
		if err := c.unmountSHM(containerMount); err != nil {
			if cleanupErr != nil {
				logrus.Errorf("Error unmounting container %s: %v", c.ID(), cleanupErr)
			}
			cleanupErr = err
		}
	}

	if c.config.Rootfs != "" {
		return cleanupErr
	}

	if err := c.unmount(false); err != nil {
		// If the container has already been removed, warn but don't
		// error
		// We still want to be able to kick the container out of the
		// state
		if errors.Cause(err) == storage.ErrNotAContainer || errors.Cause(err) == storage.ErrContainerUnknown || errors.Cause(err) == storage.ErrLayerNotMounted {
			logrus.Errorf("Storage for container %s has been removed", c.ID())
		} else {
			if cleanupErr != nil {
				logrus.Errorf("Error cleaning up container %s storage: %v", c.ID(), cleanupErr)
			}
			cleanupErr = err
		}
	}

	// Request an unmount of all named volumes
	for _, v := range c.config.NamedVolumes {
		vol, err := c.runtime.state.Volume(v.Name)
		if err != nil {
			if cleanupErr != nil {
				logrus.Errorf("Error unmounting container %s: %v", c.ID(), cleanupErr)
			}
			cleanupErr = errors.Wrapf(err, "error retrieving named volume %s for container %s", v.Name, c.ID())

			// We need to try and unmount every volume, so continue
			// if they fail.
			continue
		}

		if vol.needsMount() {
			vol.lock.Lock()
			if err := vol.unmount(false); err != nil {
				if cleanupErr != nil {
					logrus.Errorf("Error unmounting container %s: %v", c.ID(), cleanupErr)
				}
				cleanupErr = errors.Wrapf(err, "error unmounting volume %s for container %s", vol.Name(), c.ID())
			}
			vol.lock.Unlock()
		}
	}

	c.state.Mountpoint = ""
	c.state.Mounted = false

	if c.valid {
		if err := c.save(); err != nil {
			if cleanupErr != nil {
				logrus.Errorf("Error unmounting container %s: %v", c.ID(), cleanupErr)
			}
			cleanupErr = err
		}
	}
	return cleanupErr
}

// Unmount the a container and free its resources
func (c *Container) cleanup(ctx context.Context) error {
	var lastError error

	span, _ := opentracing.StartSpanFromContext(ctx, "cleanup")
	span.SetTag("struct", "container")
	defer span.Finish()

	logrus.Debugf("Cleaning up container %s", c.ID())

	// Remove healthcheck unit/timer file if it execs
	if c.config.HealthCheckConfig != nil {
		if err := c.removeTimer(); err != nil {
			logrus.Errorf("Error removing timer for container %s healthcheck: %v", c.ID(), err)
		}
	}

	// Clean up network namespace, if present
	if err := c.cleanupNetwork(); err != nil {
		lastError = errors.Wrapf(err, "error removing container %s network", c.ID())
	}

	// Unmount storage
	if err := c.cleanupStorage(); err != nil {
		if lastError != nil {
			logrus.Errorf("Error unmounting container %s storage: %v", c.ID(), err)
		} else {
			lastError = errors.Wrapf(err, "error unmounting container %s storage", c.ID())
		}
	}

	// Remove the container from the runtime, if necessary
	if err := c.cleanupRuntime(ctx); err != nil {
		if lastError != nil {
			logrus.Errorf("Error removing container %s from OCI runtime: %v", c.ID(), err)
		} else {
			lastError = err
		}
	}

	return lastError
}

// delete deletes the container and runs any configured poststop
// hooks.
func (c *Container) delete(ctx context.Context) (err error) {
	span, _ := opentracing.StartSpanFromContext(ctx, "delete")
	span.SetTag("struct", "container")
	defer span.Finish()

	if err := c.ociRuntime.DeleteContainer(c); err != nil {
		return errors.Wrapf(err, "error removing container %s from runtime", c.ID())
	}

	if err := c.postDeleteHooks(ctx); err != nil {
		return errors.Wrapf(err, "container %s poststop hooks", c.ID())
	}

	return nil
}

// postDeleteHooks runs the poststop hooks (if any) as specified by
// the OCI Runtime Specification (which requires them to run
// post-delete, despite the stage name).
func (c *Container) postDeleteHooks(ctx context.Context) (err error) {
	span, _ := opentracing.StartSpanFromContext(ctx, "postDeleteHooks")
	span.SetTag("struct", "container")
	defer span.Finish()

	if c.state.ExtensionStageHooks != nil {
		extensionHooks, ok := c.state.ExtensionStageHooks["poststop"]
		if ok {
			state, err := json.Marshal(spec.State{
				Version:     spec.Version,
				ID:          c.ID(),
				Status:      "stopped",
				Bundle:      c.bundlePath(),
				Annotations: c.config.Spec.Annotations,
			})
			if err != nil {
				return err
			}
			for i, hook := range extensionHooks {
				hook := hook
				logrus.Debugf("container %s: invoke poststop hook %d, path %s", c.ID(), i, hook.Path)
				var stderr, stdout bytes.Buffer
				hookErr, err := exec.Run(ctx, &hook, state, &stdout, &stderr, exec.DefaultPostKillTimeout)
				if err != nil {
					logrus.Warnf("container %s: poststop hook %d: %v", c.ID(), i, err)
					if hookErr != err {
						logrus.Debugf("container %s: poststop hook %d (hook error): %v", c.ID(), i, hookErr)
					}
					stdoutString := stdout.String()
					if stdoutString != "" {
						logrus.Debugf("container %s: poststop hook %d: stdout:\n%s", c.ID(), i, stdoutString)
					}
					stderrString := stderr.String()
					if stderrString != "" {
						logrus.Debugf("container %s: poststop hook %d: stderr:\n%s", c.ID(), i, stderrString)
					}
				}
			}
		}
	}

	return nil
}

// writeStringToRundir copies the provided file to the runtimedir
func (c *Container) writeStringToRundir(destFile, output string) (string, error) {
	destFileName := filepath.Join(c.state.RunDir, destFile)

	if err := os.Remove(destFileName); err != nil && !os.IsNotExist(err) {
		return "", errors.Wrapf(err, "error removing %s for container %s", destFile, c.ID())
	}

	f, err := os.Create(destFileName)
	if err != nil {
		return "", errors.Wrapf(err, "unable to create %s", destFileName)
	}
	defer f.Close()
	if err := f.Chown(c.RootUID(), c.RootGID()); err != nil {
		return "", err
	}

	if _, err := f.WriteString(output); err != nil {
		return "", errors.Wrapf(err, "unable to write %s", destFileName)
	}
	// Relabel runDirResolv for the container
	if err := label.Relabel(destFileName, c.config.MountLabel, false); err != nil {
		return "", err
	}

	return filepath.Join(c.state.RunDir, destFile), nil
}

// appendStringToRundir appends the provided string to the runtimedir file
func (c *Container) appendStringToRundir(destFile, output string) (string, error) {
	destFileName := filepath.Join(c.state.RunDir, destFile)

	f, err := os.OpenFile(destFileName, os.O_APPEND|os.O_WRONLY, 0600)
	if err != nil {
		return "", errors.Wrapf(err, "unable to open %s", destFileName)
	}
	defer f.Close()

	if _, err := f.WriteString(output); err != nil {
		return "", errors.Wrapf(err, "unable to write %s", destFileName)
	}

	return filepath.Join(c.state.RunDir, destFile), nil
}

// saveSpec saves the OCI spec to disk, replacing any existing specs for the container
func (c *Container) saveSpec(spec *spec.Spec) error {
	// If the OCI spec already exists, we need to replace it
	// Cannot guarantee some things, e.g. network namespaces, have the same
	// paths
	jsonPath := filepath.Join(c.bundlePath(), "config.json")
	if _, err := os.Stat(jsonPath); err != nil {
		if !os.IsNotExist(err) {
			return errors.Wrapf(err, "error doing stat on container %s spec", c.ID())
		}
		// The spec does not exist, we're fine
	} else {
		// The spec exists, need to remove it
		if err := os.Remove(jsonPath); err != nil {
			return errors.Wrapf(err, "error replacing runtime spec for container %s", c.ID())
		}
	}

	fileJSON, err := json.Marshal(spec)
	if err != nil {
		return errors.Wrapf(err, "error exporting runtime spec for container %s to JSON", c.ID())
	}
	if err := ioutil.WriteFile(jsonPath, fileJSON, 0644); err != nil {
		return errors.Wrapf(err, "error writing runtime spec JSON for container %s to disk", c.ID())
	}

	logrus.Debugf("Created OCI spec for container %s at %s", c.ID(), jsonPath)

	c.state.ConfigPath = jsonPath

	return nil
}

// Warning: precreate hooks may alter 'config' in place.
func (c *Container) setupOCIHooks(ctx context.Context, config *spec.Spec) (extensionStageHooks map[string][]spec.Hook, err error) {
	allHooks := make(map[string][]spec.Hook)
	if c.runtime.config.HooksDir == nil {
		if rootless.IsRootless() {
			return nil, nil
		}
		for _, hDir := range []string{hooks.DefaultDir, hooks.OverrideDir} {
			manager, err := hooks.New(ctx, []string{hDir}, []string{"precreate", "poststop"})
			if err != nil {
				if os.IsNotExist(err) {
					continue
				}
				return nil, err
			}
			ociHooks, err := manager.Hooks(config, c.Spec().Annotations, len(c.config.UserVolumes) > 0)
			if err != nil {
				return nil, err
			}
			if len(ociHooks) > 0 || config.Hooks != nil {
				logrus.Warnf("implicit hook directories are deprecated; set --ociHooks-dir=%q explicitly to continue to load ociHooks from this directory", hDir)
			}
			for i, hook := range ociHooks {
				allHooks[i] = hook
			}
		}
	} else {
		manager, err := hooks.New(ctx, c.runtime.config.HooksDir, []string{"precreate", "poststop"})
		if err != nil {
			return nil, err
		}

		allHooks, err = manager.Hooks(config, c.Spec().Annotations, len(c.config.UserVolumes) > 0)
		if err != nil {
			return nil, err
		}
	}

	hookErr, err := exec.RuntimeConfigFilter(ctx, allHooks["precreate"], config, exec.DefaultPostKillTimeout)
	if err != nil {
		logrus.Warnf("container %s: precreate hook: %v", c.ID(), err)
		if hookErr != nil && hookErr != err {
			logrus.Debugf("container %s: precreate hook (hook error): %v", c.ID(), hookErr)
		}
		return nil, err
	}

	return allHooks, nil
}

// mount mounts the container's root filesystem
func (c *Container) mount() (string, error) {
	mountPoint, err := c.runtime.storageService.MountContainerImage(c.ID())
	if err != nil {
		return "", errors.Wrapf(err, "error mounting storage for container %s", c.ID())
	}
	mountPoint, err = filepath.EvalSymlinks(mountPoint)
	if err != nil {
		return "", errors.Wrapf(err, "error resolving storage path for container %s", c.ID())
	}
	if err := os.Chown(mountPoint, c.RootUID(), c.RootGID()); err != nil {
		return "", errors.Wrapf(err, "cannot chown %s to %d:%d", mountPoint, c.RootUID(), c.RootGID())
	}
	return mountPoint, nil
}

// unmount unmounts the container's root filesystem
func (c *Container) unmount(force bool) error {
	// Also unmount storage
	if _, err := c.runtime.storageService.UnmountContainerImage(c.ID(), force); err != nil {
		return errors.Wrapf(err, "error unmounting container %s root filesystem", c.ID())
	}

	return nil
}

// this should be from chrootarchive.
// Container MUST be mounted before calling.
func (c *Container) copyWithTarFromImage(source, dest string) error {
	a := archive.NewDefaultArchiver()

	if err := c.copyOwnerAndPerms(source, dest); err != nil {
		return err
	}
	return a.CopyWithTar(source, dest)
}

// checkReadyForRemoval checks whether the given container is ready to be
// removed.
// These checks are only used if force-remove is not specified.
// If it is, we'll remove the container anyways.
// Returns nil if safe to remove, or an error describing why it's unsafe if not.
func (c *Container) checkReadyForRemoval() error {
	if c.state.State == define.ContainerStateUnknown {
		return errors.Wrapf(define.ErrCtrStateInvalid, "container %s is in invalid state", c.ID())
	}

	if c.ensureState(define.ContainerStateRunning, define.ContainerStatePaused) {
		return errors.Wrapf(define.ErrCtrStateInvalid, "cannot remove container %s as it is %s - running or paused containers cannot be removed without force", c.ID(), c.state.State.String())
	}

	// Reap exec sessions
	if err := c.reapExecSessions(); err != nil {
		return err
	}

	if len(c.state.ExecSessions) != 0 {
		return errors.Wrapf(define.ErrCtrStateInvalid, "cannot remove container %s as it has active exec sessions", c.ID())
	}

	return nil
}

// writeJSONFile marshalls and writes the given data to a JSON file
// in the bundle path
func (c *Container) writeJSONFile(v interface{}, file string) (err error) {
	fileJSON, err := json.MarshalIndent(v, "", "  ")
	if err != nil {
		return errors.Wrapf(err, "error writing JSON to %s for container %s", file, c.ID())
	}
	file = filepath.Join(c.bundlePath(), file)
	if err := ioutil.WriteFile(file, fileJSON, 0644); err != nil {
		return err
	}

	return nil
}

// prepareCheckpointExport writes the config and spec to
// JSON files for later export
func (c *Container) prepareCheckpointExport() (err error) {
	// save live config
	if err := c.writeJSONFile(c.Config(), "config.dump"); err != nil {
		return err
	}

	// save spec
	jsonPath := filepath.Join(c.bundlePath(), "config.json")
	g, err := generate.NewFromFile(jsonPath)
	if err != nil {
		logrus.Debugf("generating spec for container %q failed with %v", c.ID(), err)
		return err
	}
	if err := c.writeJSONFile(g.Config, "spec.dump"); err != nil {
		return err
	}

	return nil
}

// sortUserVolumes sorts the volumes specified for a container
// between named and normal volumes
func (c *Container) sortUserVolumes(ctrSpec *spec.Spec) ([]*ContainerNamedVolume, []spec.Mount) {
	namedUserVolumes := []*ContainerNamedVolume{}
	userMounts := []spec.Mount{}

	// We need to parse all named volumes and mounts into maps, so we don't
	// end up with repeated lookups for each user volume.
	// Map destination to struct, as destination is what is stored in
	// UserVolumes.
	namedVolumes := make(map[string]*ContainerNamedVolume)
	mounts := make(map[string]spec.Mount)
	for _, namedVol := range c.config.NamedVolumes {
		namedVolumes[namedVol.Dest] = namedVol
	}
	for _, mount := range ctrSpec.Mounts {
		mounts[mount.Destination] = mount
	}

	for _, vol := range c.config.UserVolumes {
		if volume, ok := namedVolumes[vol]; ok {
			namedUserVolumes = append(namedUserVolumes, volume)
		} else if mount, ok := mounts[vol]; ok {
			userMounts = append(userMounts, mount)
		} else {
			logrus.Warnf("Could not find mount at destination %q when parsing user volumes for container %s", vol, c.ID())
		}
	}
	return namedUserVolumes, userMounts
}

// Check for an exit file, and handle one if present
func (c *Container) checkExitFile() error {
	// If the container's not running, nothing to do.
	if !c.ensureState(define.ContainerStateRunning, define.ContainerStatePaused) {
		return nil
	}

	exitFile, err := c.exitFilePath()
	if err != nil {
		return err
	}

	// Check for the exit file
	info, err := os.Stat(exitFile)
	if err != nil {
		if os.IsNotExist(err) {
			// Container is still running, no error
			return nil
		}

		return errors.Wrapf(err, "error running stat on container %s exit file", c.ID())
	}

	// Alright, it exists. Transition to Stopped state.
	c.state.State = define.ContainerStateStopped
	c.state.PID = 0
	c.state.ConmonPID = 0

	// Read the exit file to get our stopped time and exit code.
	return c.handleExitFile(exitFile, info)
}

// Reap dead exec sessions
func (c *Container) reapExecSessions() error {
	// Instead of saving once per iteration, use a defer to do it once at
	// the end.
	var lastErr error
	needSave := false
	for id := range c.state.ExecSessions {
		alive, err := c.ociRuntime.ExecUpdateStatus(c, id)
		if err != nil {
			if lastErr != nil {
				logrus.Errorf("Error reaping exec sessions for container %s: %v", c.ID(), lastErr)
			}
			lastErr = err
			continue
		}
		if !alive {
			// Clean up lingering files and remove the exec session
			if err := c.ociRuntime.ExecContainerCleanup(c, id); err != nil {
				return errors.Wrapf(err, "error cleaning up container %s exec session %s files", c.ID(), id)
			}
			delete(c.state.ExecSessions, id)
			needSave = true
		}
	}
	if needSave {
		if err := c.save(); err != nil {
			if lastErr != nil {
				logrus.Errorf("Error reaping exec sessions for container %s: %v", c.ID(), lastErr)
			}
			lastErr = err
		}
	}
	return lastErr
}

func (c *Container) hasNamespace(namespace spec.LinuxNamespaceType) bool {
	if c.config.Spec == nil || c.config.Spec.Linux == nil {
		return false
	}
	for _, n := range c.config.Spec.Linux.Namespaces {
		if n.Type == namespace {
			return true
		}
	}
	return false
}