summaryrefslogtreecommitdiff
path: root/vendor/github.com/projectatomic/buildah/run.go
blob: 2ce5add39c6a4cd61d6e5fdc54732ba39e53f435 (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
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
package buildah

import (
	"bufio"
	"bytes"
	"encoding/json"
	"fmt"
	"io"
	"io/ioutil"
	"net"
	"os"
	"os/exec"
	"path/filepath"
	"runtime"
	"strconv"
	"strings"
	"sync"
	"syscall"
	"time"

	"github.com/containernetworking/cni/libcni"
	"github.com/containers/libpod/pkg/secrets"
	"github.com/containers/storage/pkg/idtools"
	"github.com/containers/storage/pkg/ioutils"
	"github.com/containers/storage/pkg/reexec"
	units "github.com/docker/go-units"
	digest "github.com/opencontainers/go-digest"
	"github.com/opencontainers/runtime-spec/specs-go"
	"github.com/opencontainers/runtime-tools/generate"
	"github.com/opencontainers/selinux/go-selinux/label"
	"github.com/pkg/errors"
	"github.com/projectatomic/buildah/bind"
	"github.com/projectatomic/buildah/chroot"
	"github.com/projectatomic/buildah/util"
	"github.com/sirupsen/logrus"
	"golang.org/x/crypto/ssh/terminal"
	"golang.org/x/sys/unix"
)

const (
	// DefaultWorkingDir is used if none was specified.
	DefaultWorkingDir = "/"
	// runUsingRuntimeCommand is a command we use as a key for reexec
	runUsingRuntimeCommand = Package + "-oci-runtime"
)

// TerminalPolicy takes the value DefaultTerminal, WithoutTerminal, or WithTerminal.
type TerminalPolicy int

const (
	// DefaultTerminal indicates that this Run invocation should be
	// connected to a pseudoterminal if we're connected to a terminal.
	DefaultTerminal TerminalPolicy = iota
	// WithoutTerminal indicates that this Run invocation should NOT be
	// connected to a pseudoterminal.
	WithoutTerminal
	// WithTerminal indicates that this Run invocation should be connected
	// to a pseudoterminal.
	WithTerminal
)

// String converts a TerminalPoliicy into a string.
func (t TerminalPolicy) String() string {
	switch t {
	case DefaultTerminal:
		return "DefaultTerminal"
	case WithoutTerminal:
		return "WithoutTerminal"
	case WithTerminal:
		return "WithTerminal"
	}
	return fmt.Sprintf("unrecognized terminal setting %d", t)
}

// NamespaceOption controls how we set up a namespace when launching processes.
type NamespaceOption struct {
	// Name specifies the type of namespace, typically matching one of the
	// ...Namespace constants defined in
	// github.com/opencontainers/runtime-spec/specs-go.
	Name string
	// Host is used to force our processes to use the host's namespace of
	// this type.
	Host bool
	// Path is the path of the namespace to attach our process to, if Host
	// is not set.  If Host is not set and Path is also empty, a new
	// namespace will be created for the process that we're starting.
	// If Name is specs.NetworkNamespace, if Path doesn't look like an
	// absolute path, it is treated as a comma-separated list of CNI
	// configuration names which will be selected from among all of the CNI
	// network configurations which we find.
	Path string
}

// NamespaceOptions provides some helper methods for a slice of NamespaceOption
// structs.
type NamespaceOptions []NamespaceOption

// IDMappingOptions controls how we set up UID/GID mapping when we set up a
// user namespace.
type IDMappingOptions struct {
	HostUIDMapping bool
	HostGIDMapping bool
	UIDMap         []specs.LinuxIDMapping
	GIDMap         []specs.LinuxIDMapping
}

// Isolation provides a way to specify whether we're supposed to use a proper
// OCI runtime, or some other method for running commands.
type Isolation int

const (
	// IsolationDefault is whatever we think will work best.
	IsolationDefault Isolation = iota
	// IsolationOCI is a proper OCI runtime.
	IsolationOCI
	// IsolationChroot is a more chroot-like environment: less isolation,
	// but with fewer requirements.
	IsolationChroot
	// IsolationOCIRootless is a proper OCI runtime in rootless mode.
	IsolationOCIRootless
)

// String converts a Isolation into a string.
func (i Isolation) String() string {
	switch i {
	case IsolationDefault:
		return "IsolationDefault"
	case IsolationOCI:
		return "IsolationOCI"
	case IsolationChroot:
		return "IsolationChroot"
	case IsolationOCIRootless:
		return "IsolationOCIRootless"
	}
	return fmt.Sprintf("unrecognized isolation type %d", i)
}

// RunOptions can be used to alter how a command is run in the container.
type RunOptions struct {
	// Hostname is the hostname we set for the running container.
	Hostname string
	// Isolation is either IsolationDefault, IsolationOCI, IsolationChroot, or IsolationOCIRootless.
	Isolation Isolation
	// Runtime is the name of the runtime to run.  It should accept the
	// same arguments that runc does, and produce similar output.
	Runtime string
	// Args adds global arguments for the runtime.
	Args []string
	// Mounts are additional mount points which we want to provide.
	Mounts []specs.Mount
	// Env is additional environment variables to set.
	Env []string
	// User is the user as whom to run the command.
	User string
	// WorkingDir is an override for the working directory.
	WorkingDir string
	// Shell is default shell to run in a container.
	Shell string
	// Cmd is an override for the configured default command.
	Cmd []string
	// Entrypoint is an override for the configured entry point.
	Entrypoint []string
	// NamespaceOptions controls how we set up the namespaces for the process.
	NamespaceOptions NamespaceOptions
	// ConfigureNetwork controls whether or not network interfaces and
	// routing are configured for a new network namespace (i.e., when not
	// joining another's namespace and not just using the host's
	// namespace), effectively deciding whether or not the process has a
	// usable network.
	ConfigureNetwork NetworkConfigurationPolicy
	// CNIPluginPath is the location of CNI plugin helpers, if they should be
	// run from a location other than the default location.
	CNIPluginPath string
	// CNIConfigDir is the location of CNI configuration files, if the files in
	// the default configuration directory shouldn't be used.
	CNIConfigDir string
	// Terminal provides a way to specify whether or not the command should
	// be run with a pseudoterminal.  By default (DefaultTerminal), a
	// terminal is used if os.Stdout is connected to a terminal, but that
	// decision can be overridden by specifying either WithTerminal or
	// WithoutTerminal.
	Terminal TerminalPolicy
	// TerminalSize provides a way to set the number of rows and columns in
	// a pseudo-terminal, if we create one, and Stdin/Stdout/Stderr aren't
	// connected to a terminal.
	TerminalSize *specs.Box
	// The stdin/stdout/stderr descriptors to use.  If set to nil, the
	// corresponding files in the "os" package are used as defaults.
	Stdin  io.Reader `json:"-"`
	Stdout io.Writer `json:"-"`
	Stderr io.Writer `json:"-"`
	// Quiet tells the run to turn off output to stdout.
	Quiet bool
	// AddCapabilities is a list of capabilities to add to the default set.
	AddCapabilities []string
	// DropCapabilities is a list of capabilities to remove from the default set,
	// after processing the AddCapabilities set.  If a capability appears in both
	// lists, it will be dropped.
	DropCapabilities []string
}

// DefaultNamespaceOptions returns the default namespace settings from the
// runtime-tools generator library.
func DefaultNamespaceOptions() (NamespaceOptions, error) {
	options := NamespaceOptions{
		{Name: string(specs.CgroupNamespace), Host: true},
		{Name: string(specs.IPCNamespace), Host: true},
		{Name: string(specs.MountNamespace), Host: true},
		{Name: string(specs.NetworkNamespace), Host: true},
		{Name: string(specs.PIDNamespace), Host: true},
		{Name: string(specs.UserNamespace), Host: true},
		{Name: string(specs.UTSNamespace), Host: true},
	}
	g, err := generate.New("linux")
	if err != nil {
		return options, err
	}
	spec := g.Config
	if spec.Linux != nil {
		for _, ns := range spec.Linux.Namespaces {
			options.AddOrReplace(NamespaceOption{
				Name: string(ns.Type),
				Path: ns.Path,
			})
		}
	}
	return options, nil
}

// Find the configuration for the namespace of the given type.  If there are
// duplicates, find the _last_ one of the type, since we assume it was appended
// more recently.
func (n *NamespaceOptions) Find(namespace string) *NamespaceOption {
	for i := range *n {
		j := len(*n) - 1 - i
		if (*n)[j].Name == namespace {
			return &((*n)[j])
		}
	}
	return nil
}

// AddOrReplace either adds or replaces the configuration for a given namespace.
func (n *NamespaceOptions) AddOrReplace(options ...NamespaceOption) {
nextOption:
	for _, option := range options {
		for i := range *n {
			j := len(*n) - 1 - i
			if (*n)[j].Name == option.Name {
				(*n)[j] = option
				continue nextOption
			}
		}
		*n = append(*n, option)
	}
}

func addRlimits(ulimit []string, g *generate.Generator) error {
	var (
		ul  *units.Ulimit
		err error
	)

	for _, u := range ulimit {
		if ul, err = units.ParseUlimit(u); err != nil {
			return errors.Wrapf(err, "ulimit option %q requires name=SOFT:HARD, failed to be parsed", u)
		}

		g.AddProcessRlimits("RLIMIT_"+strings.ToUpper(ul.Name), uint64(ul.Hard), uint64(ul.Soft))
	}
	return nil
}

func addHosts(hosts []string, w io.Writer) error {
	buf := bufio.NewWriter(w)
	for _, host := range hosts {
		values := strings.SplitN(host, ":", 2)
		if len(values) != 2 {
			return errors.Errorf("unable to parse host entry %q: incorrect format", host)
		}
		if values[0] == "" {
			return errors.Errorf("hostname in host entry %q is empty", host)
		}
		if values[1] == "" {
			return errors.Errorf("IP address in host entry %q is empty", host)
		}
		fmt.Fprintf(buf, "%s\t%s\n", values[1], values[0])
	}
	return buf.Flush()
}

func addHostsToFile(hosts []string, filename string) error {
	if len(hosts) == 0 {
		return nil
	}
	file, err := os.OpenFile(filename, os.O_APPEND|os.O_WRONLY, os.ModeAppend)
	if err != nil {
		return err
	}
	defer file.Close()
	return addHosts(hosts, file)
}

func addCommonOptsToSpec(commonOpts *CommonBuildOptions, g *generate.Generator) error {
	// Resources - CPU
	if commonOpts.CPUPeriod != 0 {
		g.SetLinuxResourcesCPUPeriod(commonOpts.CPUPeriod)
	}
	if commonOpts.CPUQuota != 0 {
		g.SetLinuxResourcesCPUQuota(commonOpts.CPUQuota)
	}
	if commonOpts.CPUShares != 0 {
		g.SetLinuxResourcesCPUShares(commonOpts.CPUShares)
	}
	if commonOpts.CPUSetCPUs != "" {
		g.SetLinuxResourcesCPUCpus(commonOpts.CPUSetCPUs)
	}
	if commonOpts.CPUSetMems != "" {
		g.SetLinuxResourcesCPUMems(commonOpts.CPUSetMems)
	}

	// Resources - Memory
	if commonOpts.Memory != 0 {
		g.SetLinuxResourcesMemoryLimit(commonOpts.Memory)
	}
	if commonOpts.MemorySwap != 0 {
		g.SetLinuxResourcesMemorySwap(commonOpts.MemorySwap)
	}

	// cgroup membership
	if commonOpts.CgroupParent != "" {
		g.SetLinuxCgroupsPath(commonOpts.CgroupParent)
	}

	// Other process resource limits
	if err := addRlimits(commonOpts.Ulimit, g); err != nil {
		return err
	}

	logrus.Debugf("Resources: %#v", commonOpts)
	return nil
}

func (b *Builder) setupMounts(mountPoint string, spec *specs.Spec, bundlePath string, optionMounts []specs.Mount, bindFiles map[string]string, builtinVolumes, volumeMounts []string, shmSize string, namespaceOptions NamespaceOptions) error {
	// Start building a new list of mounts.
	var mounts []specs.Mount
	haveMount := func(destination string) bool {
		for _, mount := range mounts {
			if mount.Destination == destination {
				// Already have something to mount there.
				return true
			}
		}
		return false
	}

	ipc := namespaceOptions.Find(string(specs.IPCNamespace))
	hostIPC := ipc == nil || ipc.Host
	net := namespaceOptions.Find(string(specs.NetworkNamespace))
	hostNetwork := net == nil || net.Host
	user := namespaceOptions.Find(string(specs.UserNamespace))
	hostUser := user == nil || user.Host

	// Copy mounts from the generated list.
	mountCgroups := true
	specMounts := []specs.Mount{}
	for _, specMount := range spec.Mounts {
		// Override some of the mounts from the generated list if we're doing different things with namespaces.
		if specMount.Destination == "/dev/shm" {
			specMount.Options = []string{"nosuid", "noexec", "nodev", "mode=1777", "size=" + shmSize}
			if hostIPC && !hostUser {
				if _, err := os.Stat("/dev/shm"); err != nil && os.IsNotExist(err) {
					continue
				}
				specMount = specs.Mount{
					Source:      "/dev/shm",
					Type:        "bind",
					Destination: "/dev/shm",
					Options:     []string{bind.NoBindOption, "rbind", "nosuid", "noexec", "nodev"},
				}
			}
		}
		if specMount.Destination == "/dev/mqueue" {
			if hostIPC && !hostUser {
				if _, err := os.Stat("/dev/mqueue"); err != nil && os.IsNotExist(err) {
					continue
				}
				specMount = specs.Mount{
					Source:      "/dev/mqueue",
					Type:        "bind",
					Destination: "/dev/mqueue",
					Options:     []string{bind.NoBindOption, "rbind", "nosuid", "noexec", "nodev"},
				}
			}
		}
		if specMount.Destination == "/sys" {
			if hostNetwork && !hostUser {
				mountCgroups = false
				if _, err := os.Stat("/sys"); err != nil && os.IsNotExist(err) {
					continue
				}
				specMount = specs.Mount{
					Source:      "/sys",
					Type:        "bind",
					Destination: "/sys",
					Options:     []string{bind.NoBindOption, "rbind", "nosuid", "noexec", "nodev", "ro"},
				}
			}
		}
		specMounts = append(specMounts, specMount)
	}

	// Add a mount for the cgroups filesystem, unless we're already
	// recursively bind mounting all of /sys, in which case we shouldn't
	// bother with it.
	sysfsMount := []specs.Mount{}
	if mountCgroups {
		sysfsMount = []specs.Mount{{
			Destination: "/sys/fs/cgroup",
			Type:        "cgroup",
			Source:      "cgroup",
			Options:     []string{bind.NoBindOption, "nosuid", "noexec", "nodev", "relatime", "ro"},
		}}
	}

	// Get the list of files we need to bind into the container.
	bindFileMounts, err := runSetupBoundFiles(bundlePath, bindFiles)
	if err != nil {
		return err
	}

	// After this point we need to know the per-container persistent storage directory.
	cdir, err := b.store.ContainerDirectory(b.ContainerID)
	if err != nil {
		return errors.Wrapf(err, "error determining work directory for container %q", b.ContainerID)
	}

	// Figure out which UID and GID to tell the secrets package to use
	// for files that it creates.
	rootUID, rootGID, err := util.GetHostRootIDs(spec)
	if err != nil {
		return err
	}

	// Get the list of secrets mounts.
	secretMounts := secrets.SecretMountsWithUIDGID(b.MountLabel, cdir, b.DefaultMountsFilePath, cdir, int(rootUID), int(rootGID))

	// Add temporary copies of the contents of volume locations at the
	// volume locations, unless we already have something there.
	copyWithTar := b.copyWithTar(nil, nil)
	builtins, err := runSetupBuiltinVolumes(b.MountLabel, mountPoint, cdir, copyWithTar, builtinVolumes)
	if err != nil {
		return err
	}

	// Get the list of explicitly-specified volume mounts.
	volumes, err := runSetupVolumeMounts(spec.Linux.MountLabel, volumeMounts, optionMounts)
	if err != nil {
		return err
	}

	// Add them all, in the preferred order, except where they conflict with something that was previously added.
	for _, mount := range append(append(append(append(append(volumes, builtins...), secretMounts...), bindFileMounts...), specMounts...), sysfsMount...) {
		if haveMount(mount.Destination) {
			// Already mounting something there, no need to bother with this one.
			continue
		}
		// Add the mount.
		mounts = append(mounts, mount)
	}

	// Set the list in the spec.
	spec.Mounts = mounts
	return nil
}

func runSetupBoundFiles(bundlePath string, bindFiles map[string]string) (mounts []specs.Mount, err error) {
	for dest, src := range bindFiles {
		options := []string{"rbind"}
		if strings.HasPrefix(src, bundlePath) {
			options = append(options, bind.NoBindOption)
		}
		mounts = append(mounts, specs.Mount{
			Source:      src,
			Destination: dest,
			Type:        "bind",
			Options:     options,
		})
	}
	return mounts, nil
}

func runSetupBuiltinVolumes(mountLabel, mountPoint, containerDir string, copyWithTar func(srcPath, dstPath string) error, builtinVolumes []string) ([]specs.Mount, error) {
	var mounts []specs.Mount
	// Add temporary copies of the contents of volume locations at the
	// volume locations, unless we already have something there.
	for _, volume := range builtinVolumes {
		subdir := digest.Canonical.FromString(volume).Hex()
		volumePath := filepath.Join(containerDir, "buildah-volumes", subdir)
		// If we need to, initialize the volume path's initial contents.
		if _, err := os.Stat(volumePath); os.IsNotExist(err) {
			if err = os.MkdirAll(volumePath, 0755); err != nil {
				return nil, errors.Wrapf(err, "error creating directory %q for volume %q", volumePath, volume)
			}
			if err = label.Relabel(volumePath, mountLabel, false); err != nil {
				return nil, errors.Wrapf(err, "error relabeling directory %q for volume %q", volumePath, volume)
			}
			srcPath := filepath.Join(mountPoint, volume)
			if err = copyWithTar(srcPath, volumePath); err != nil && !os.IsNotExist(err) {
				return nil, errors.Wrapf(err, "error populating directory %q for volume %q using contents of %q", volumePath, volume, srcPath)
			}

		}
		// Add the bind mount.
		mounts = append(mounts, specs.Mount{
			Source:      volumePath,
			Destination: volume,
			Type:        "bind",
			Options:     []string{"bind"},
		})
	}
	return mounts, nil
}

func runSetupVolumeMounts(mountLabel string, volumeMounts []string, optionMounts []specs.Mount) ([]specs.Mount, error) {
	var mounts []specs.Mount

	parseMount := func(host, container string, options []string) (specs.Mount, error) {
		var foundrw, foundro, foundz, foundZ bool
		var rootProp string
		for _, opt := range options {
			switch opt {
			case "rw":
				foundrw = true
			case "ro":
				foundro = true
			case "z":
				foundz = true
			case "Z":
				foundZ = true
			case "private", "rprivate", "slave", "rslave", "shared", "rshared":
				rootProp = opt
			}
		}
		if !foundrw && !foundro {
			options = append(options, "rw")
		}
		if foundz {
			if err := label.Relabel(host, mountLabel, true); err != nil {
				return specs.Mount{}, errors.Wrapf(err, "relabeling %q failed", host)
			}
		}
		if foundZ {
			if err := label.Relabel(host, mountLabel, false); err != nil {
				return specs.Mount{}, errors.Wrapf(err, "relabeling %q failed", host)
			}
		}
		if rootProp == "" {
			options = append(options, "private")
		}
		return specs.Mount{
			Destination: container,
			Type:        "bind",
			Source:      host,
			Options:     options,
		}, nil
	}
	// Bind mount volumes specified for this particular Run() invocation
	for _, i := range optionMounts {
		mount, err := parseMount(i.Source, i.Destination, append(i.Options, "rbind"))
		if err != nil {
			return nil, err
		}
		mounts = append(mounts, mount)
	}
	// Bind mount volumes given by the user when the container was created
	for _, i := range volumeMounts {
		var options []string
		spliti := strings.Split(i, ":")
		if len(spliti) > 2 {
			options = strings.Split(spliti[2], ",")
		}
		options = append(options, "rbind")
		mount, err := parseMount(spliti[0], spliti[1], options)
		if err != nil {
			return nil, err
		}
		mounts = append(mounts, mount)
	}
	return mounts, nil
}

// addNetworkConfig copies files from host and sets them up to bind mount into container
func (b *Builder) addNetworkConfig(rdir, hostPath string, chownOpts *idtools.IDPair) (string, error) {
	copyFileWithTar := b.copyFileWithTar(chownOpts, nil)

	cfile := filepath.Join(rdir, filepath.Base(hostPath))

	if err := copyFileWithTar(hostPath, cfile); err != nil {
		return "", errors.Wrapf(err, "error copying %q for container %q", cfile, b.ContainerID)
	}

	if err := label.Relabel(cfile, b.MountLabel, false); err != nil {
		return "", errors.Wrapf(err, "error relabeling %q in container %q", cfile, b.ContainerID)
	}

	return cfile, nil
}

func setupMaskedPaths(g *generate.Generator) {
	for _, mp := range []string{
		"/proc/acpi",
		"/proc/kcore",
		"/proc/keys",
		"/proc/latency_stats",
		"/proc/timer_list",
		"/proc/timer_stats",
		"/proc/sched_debug",
		"/proc/scsi",
		"/sys/firmware",
	} {
		g.AddLinuxMaskedPaths(mp)
	}
}

func setupReadOnlyPaths(g *generate.Generator) {
	for _, rp := range []string{
		"/proc/asound",
		"/proc/bus",
		"/proc/fs",
		"/proc/irq",
		"/proc/sys",
		"/proc/sysrq-trigger",
	} {
		g.AddLinuxReadonlyPaths(rp)
	}
}

func setupCapAdd(g *generate.Generator, caps ...string) error {
	for _, cap := range caps {
		if err := g.AddProcessCapabilityBounding(cap); err != nil {
			return errors.Wrapf(err, "error adding %q to the bounding capability set", cap)
		}
		if err := g.AddProcessCapabilityEffective(cap); err != nil {
			return errors.Wrapf(err, "error adding %q to the effective capability set", cap)
		}
		if err := g.AddProcessCapabilityInheritable(cap); err != nil {
			return errors.Wrapf(err, "error adding %q to the inheritable capability set", cap)
		}
		if err := g.AddProcessCapabilityPermitted(cap); err != nil {
			return errors.Wrapf(err, "error adding %q to the permitted capability set", cap)
		}
		if err := g.AddProcessCapabilityAmbient(cap); err != nil {
			return errors.Wrapf(err, "error adding %q to the ambient capability set", cap)
		}
	}
	return nil
}

func setupCapDrop(g *generate.Generator, caps ...string) error {
	for _, cap := range caps {
		if err := g.DropProcessCapabilityBounding(cap); err != nil {
			return errors.Wrapf(err, "error removing %q from the bounding capability set", cap)
		}
		if err := g.DropProcessCapabilityEffective(cap); err != nil {
			return errors.Wrapf(err, "error removing %q from the effective capability set", cap)
		}
		if err := g.DropProcessCapabilityInheritable(cap); err != nil {
			return errors.Wrapf(err, "error removing %q from the inheritable capability set", cap)
		}
		if err := g.DropProcessCapabilityPermitted(cap); err != nil {
			return errors.Wrapf(err, "error removing %q from the permitted capability set", cap)
		}
		if err := g.DropProcessCapabilityAmbient(cap); err != nil {
			return errors.Wrapf(err, "error removing %q from the ambient capability set", cap)
		}
	}
	return nil
}

func setupCapabilities(g *generate.Generator, firstAdds, firstDrops, secondAdds, secondDrops []string) error {
	g.ClearProcessCapabilities()
	if err := setupCapAdd(g, util.DefaultCapabilities...); err != nil {
		return err
	}
	if err := setupCapAdd(g, firstAdds...); err != nil {
		return err
	}
	if err := setupCapDrop(g, firstDrops...); err != nil {
		return err
	}
	if err := setupCapAdd(g, secondAdds...); err != nil {
		return err
	}
	if err := setupCapDrop(g, secondDrops...); err != nil {
		return err
	}
	return nil
}

func setupTerminal(g *generate.Generator, terminalPolicy TerminalPolicy, terminalSize *specs.Box) {
	switch terminalPolicy {
	case DefaultTerminal:
		onTerminal := terminal.IsTerminal(unix.Stdin) && terminal.IsTerminal(unix.Stdout) && terminal.IsTerminal(unix.Stderr)
		if onTerminal {
			logrus.Debugf("stdio is a terminal, defaulting to using a terminal")
		} else {
			logrus.Debugf("stdio is not a terminal, defaulting to not using a terminal")
		}
		g.SetProcessTerminal(onTerminal)
	case WithTerminal:
		g.SetProcessTerminal(true)
	case WithoutTerminal:
		g.SetProcessTerminal(false)
	}
	if terminalSize != nil {
		g.SetProcessConsoleSize(terminalSize.Width, terminalSize.Height)
	}
}

func setupNamespaces(g *generate.Generator, namespaceOptions NamespaceOptions, idmapOptions IDMappingOptions, policy NetworkConfigurationPolicy) (configureNetwork bool, configureNetworks []string, configureUTS bool, err error) {
	// Set namespace options in the container configuration.
	configureUserns := false
	specifiedNetwork := false
	for _, namespaceOption := range namespaceOptions {
		switch namespaceOption.Name {
		case string(specs.UserNamespace):
			configureUserns = false
			if !namespaceOption.Host && namespaceOption.Path == "" {
				configureUserns = true
			}
		case string(specs.NetworkNamespace):
			specifiedNetwork = true
			configureNetwork = false
			if !namespaceOption.Host && (namespaceOption.Path == "" || !filepath.IsAbs(namespaceOption.Path)) {
				if namespaceOption.Path != "" && !filepath.IsAbs(namespaceOption.Path) {
					configureNetworks = strings.Split(namespaceOption.Path, ",")
					namespaceOption.Path = ""
				}
				configureNetwork = (policy != NetworkDisabled)
			}
		case string(specs.UTSNamespace):
			configureUTS = false
			if !namespaceOption.Host && namespaceOption.Path == "" {
				configureUTS = true
			}
		}
		if namespaceOption.Host {
			if err := g.RemoveLinuxNamespace(namespaceOption.Name); err != nil {
				return false, nil, false, errors.Wrapf(err, "error removing %q namespace for run", namespaceOption.Name)
			}
		} else if err := g.AddOrReplaceLinuxNamespace(namespaceOption.Name, namespaceOption.Path); err != nil {
			if namespaceOption.Path == "" {
				return false, nil, false, errors.Wrapf(err, "error adding new %q namespace for run", namespaceOption.Name)
			}
			return false, nil, false, errors.Wrapf(err, "error adding %q namespace %q for run", namespaceOption.Name, namespaceOption.Path)
		}
	}

	// If we've got mappings, we're going to have to create a user namespace.
	if len(idmapOptions.UIDMap) > 0 || len(idmapOptions.GIDMap) > 0 || configureUserns {
		if err := g.AddOrReplaceLinuxNamespace(specs.UserNamespace, ""); err != nil {
			return false, nil, false, errors.Wrapf(err, "error adding new %q namespace for run", string(specs.UserNamespace))
		}
		hostUidmap, hostGidmap, err := util.GetHostIDMappings("")
		if err != nil {
			return false, nil, false, err
		}
		for _, m := range idmapOptions.UIDMap {
			g.AddLinuxUIDMapping(m.HostID, m.ContainerID, m.Size)
		}
		if len(idmapOptions.UIDMap) == 0 {
			for _, m := range hostUidmap {
				g.AddLinuxUIDMapping(m.ContainerID, m.ContainerID, m.Size)
			}
		}
		for _, m := range idmapOptions.GIDMap {
			g.AddLinuxGIDMapping(m.HostID, m.ContainerID, m.Size)
		}
		if len(idmapOptions.GIDMap) == 0 {
			for _, m := range hostGidmap {
				g.AddLinuxGIDMapping(m.ContainerID, m.ContainerID, m.Size)
			}
		}
		if !specifiedNetwork {
			if err := g.AddOrReplaceLinuxNamespace(specs.NetworkNamespace, ""); err != nil {
				return false, nil, false, errors.Wrapf(err, "error adding new %q namespace for run", string(specs.NetworkNamespace))
			}
			configureNetwork = (policy != NetworkDisabled)
		}
	} else {
		if err := g.RemoveLinuxNamespace(specs.UserNamespace); err != nil {
			return false, nil, false, errors.Wrapf(err, "error removing %q namespace for run", string(specs.UserNamespace))
		}
		if !specifiedNetwork {
			if err := g.RemoveLinuxNamespace(specs.NetworkNamespace); err != nil {
				return false, nil, false, errors.Wrapf(err, "error removing %q namespace for run", string(specs.NetworkNamespace))
			}
		}
	}
	if configureNetwork {
		for name, val := range util.DefaultNetworkSysctl {
			g.AddLinuxSysctl(name, val)
		}
	}
	return configureNetwork, configureNetworks, configureUTS, nil
}

// Search for a command that isn't given as an absolute path using the $PATH
// under the rootfs.  We can't resolve absolute symbolic links without
// chroot()ing, which we may not be able to do, so just accept a link as a
// valid resolution.
func runLookupPath(g *generate.Generator, command []string) []string {
	// Look for the configured $PATH.
	spec := g.Spec()
	envPath := ""
	for i := range spec.Process.Env {
		if strings.HasPrefix(spec.Process.Env[i], "PATH=") {
			envPath = spec.Process.Env[i]
		}
	}
	// If there is no configured $PATH, supply one.
	if envPath == "" {
		defaultPath := "/usr/local/bin:/usr/local/sbin:/usr/bin:/usr/sbin:/bin:/sbin"
		envPath = "PATH=" + defaultPath
		g.AddProcessEnv("PATH", defaultPath)
	}
	// No command, nothing to do.
	if len(command) == 0 {
		return command
	}
	// Command is already an absolute path, use it as-is.
	if filepath.IsAbs(command[0]) {
		return command
	}
	// For each element in the PATH,
	for _, pathEntry := range filepath.SplitList(envPath[5:]) {
		// if it's the empty string, it's ".", which is the Cwd,
		if pathEntry == "" {
			pathEntry = spec.Process.Cwd
		}
		// build the absolute path which it might be,
		candidate := filepath.Join(pathEntry, command[0])
		// check if it's there,
		if fi, err := os.Lstat(filepath.Join(spec.Root.Path, candidate)); fi != nil && err == nil {
			// and if it's not a directory, and either a symlink or executable,
			if !fi.IsDir() && ((fi.Mode()&os.ModeSymlink != 0) || (fi.Mode()&0111 != 0)) {
				// use that.
				return append([]string{candidate}, command[1:]...)
			}
		}
	}
	return command
}

func (b *Builder) configureUIDGID(g *generate.Generator, mountPoint string, options RunOptions) error {
	// Set the user UID/GID/supplemental group list/capabilities lists.
	user, err := b.user(mountPoint, options.User)
	if err != nil {
		return err
	}
	if err := setupCapabilities(g, b.AddCapabilities, b.DropCapabilities, options.AddCapabilities, options.DropCapabilities); err != nil {
		return err
	}
	g.SetProcessUID(user.UID)
	g.SetProcessGID(user.GID)
	for _, gid := range user.AdditionalGids {
		g.AddProcessAdditionalGid(gid)
	}

	// Remove capabilities if not running as root except Bounding set
	if user.UID != 0 {
		bounding := g.Config.Process.Capabilities.Bounding
		g.ClearProcessCapabilities()
		g.Config.Process.Capabilities.Bounding = bounding
	}

	return nil
}

func (b *Builder) configureEnvironment(g *generate.Generator, options RunOptions) {
	g.ClearProcessEnv()
	for _, envSpec := range append(b.Env(), options.Env...) {
		env := strings.SplitN(envSpec, "=", 2)
		if len(env) > 1 {
			g.AddProcessEnv(env[0], env[1])
		}
	}

	for src, dest := range b.Args {
		g.AddProcessEnv(src, dest)
	}
}

func (b *Builder) configureNamespaces(g *generate.Generator, options RunOptions) (bool, []string, error) {
	defaultNamespaceOptions, err := DefaultNamespaceOptions()
	if err != nil {
		return false, nil, err
	}

	namespaceOptions := defaultNamespaceOptions
	namespaceOptions.AddOrReplace(b.NamespaceOptions...)
	namespaceOptions.AddOrReplace(options.NamespaceOptions...)

	networkPolicy := options.ConfigureNetwork
	if networkPolicy == NetworkDefault {
		networkPolicy = b.ConfigureNetwork
	}

	configureNetwork, configureNetworks, configureUTS, err := setupNamespaces(g, namespaceOptions, b.IDMappingOptions, networkPolicy)
	if err != nil {
		return false, nil, err
	}

	if configureUTS {
		if options.Hostname != "" {
			g.SetHostname(options.Hostname)
		} else if b.Hostname() != "" {
			g.SetHostname(b.Hostname())
		}
	} else {
		g.SetHostname("")
	}
	return configureNetwork, configureNetworks, nil
}

// Run runs the specified command in the container's root filesystem.
func (b *Builder) Run(command []string, options RunOptions) error {
	p, err := ioutil.TempDir("", Package)
	if err != nil {
		return err
	}
	// On some hosts like AH, /tmp is a symlink and we need an
	// absolute path.
	path, err := filepath.EvalSymlinks(p)
	if err != nil {
		return err
	}
	logrus.Debugf("using %q to hold bundle data", path)
	defer func() {
		if err2 := os.RemoveAll(path); err2 != nil {
			logrus.Errorf("error removing %q: %v", path, err2)
		}
	}()

	gp, err := generate.New("linux")
	if err != nil {
		return err
	}
	g := &gp

	isolation := options.Isolation
	if isolation == IsolationDefault {
		isolation = b.Isolation
		if isolation == IsolationDefault {
			isolation = IsolationOCI
		}
	}
	if err := checkAndOverrideIsolationOptions(isolation, &options); err != nil {
		return err
	}

	b.configureEnvironment(g, options)

	if b.CommonBuildOpts == nil {
		return errors.Errorf("Invalid format on container you must recreate the container")
	}

	if err := addCommonOptsToSpec(b.CommonBuildOpts, g); err != nil {
		return err
	}

	if options.WorkingDir != "" {
		g.SetProcessCwd(options.WorkingDir)
	} else if b.WorkDir() != "" {
		g.SetProcessCwd(b.WorkDir())
	}
	setupSelinux(g, b.ProcessLabel, b.MountLabel)
	mountPoint, err := b.Mount(b.MountLabel)
	if err != nil {
		return err
	}
	defer func() {
		if err := b.Unmount(); err != nil {
			logrus.Errorf("error unmounting container: %v", err)
		}
	}()
	g.SetRootPath(mountPoint)
	if len(command) > 0 {
		command = runLookupPath(g, command)
		g.SetProcessArgs(command)
	} else {
		g.SetProcessArgs(nil)
	}

	setupMaskedPaths(g)
	setupReadOnlyPaths(g)

	setupTerminal(g, options.Terminal, options.TerminalSize)

	configureNetwork, configureNetworks, err := b.configureNamespaces(g, options)
	if err != nil {
		return err
	}

	if err := b.configureUIDGID(g, mountPoint, options); err != nil {
		return err
	}

	g.SetProcessApparmorProfile(b.CommonBuildOpts.ApparmorProfile)

	// Now grab the spec from the generator.  Set the generator to nil so that future contributors
	// will quickly be able to tell that they're supposed to be modifying the spec directly from here.
	spec := g.Spec()
	g = nil

	// Set the working directory, creating it if we must.
	if spec.Process.Cwd == "" {
		spec.Process.Cwd = DefaultWorkingDir
	}
	logrus.Debugf("ensuring working directory %q exists", filepath.Join(mountPoint, spec.Process.Cwd))
	if err = os.MkdirAll(filepath.Join(mountPoint, spec.Process.Cwd), 0755); err != nil {
		return errors.Wrapf(err, "error ensuring working directory %q exists", spec.Process.Cwd)
	}

	// Set the seccomp configuration using the specified profile name.  Some syscalls are
	// allowed if certain capabilities are to be granted (example: CAP_SYS_CHROOT and chroot),
	// so we sorted out the capabilities lists first.
	if err = setupSeccomp(spec, b.CommonBuildOpts.SeccompProfilePath); err != nil {
		return err
	}

	// Figure out who owns files that will appear to be owned by UID/GID 0 in the container.
	rootUID, rootGID, err := util.GetHostRootIDs(spec)
	if err != nil {
		return err
	}
	rootIDPair := &idtools.IDPair{UID: int(rootUID), GID: int(rootGID)}

	hostFile, err := b.addNetworkConfig(path, "/etc/hosts", rootIDPair)
	if err != nil {
		return err
	}
	resolvFile, err := b.addNetworkConfig(path, "/etc/resolv.conf", rootIDPair)
	if err != nil {
		return err
	}

	if err := addHostsToFile(b.CommonBuildOpts.AddHost, hostFile); err != nil {
		return err
	}

	bindFiles := map[string]string{
		"/etc/hosts":       hostFile,
		"/etc/resolv.conf": resolvFile,
	}
	err = b.setupMounts(mountPoint, spec, path, options.Mounts, bindFiles, b.Volumes(), b.CommonBuildOpts.Volumes, b.CommonBuildOpts.ShmSize, append(b.NamespaceOptions, options.NamespaceOptions...))
	if err != nil {
		return errors.Wrapf(err, "error resolving mountpoints for container")
	}

	if options.CNIConfigDir == "" {
		options.CNIConfigDir = b.CNIConfigDir
		if b.CNIConfigDir == "" {
			options.CNIConfigDir = util.DefaultCNIConfigDir
		}
	}
	if options.CNIPluginPath == "" {
		options.CNIPluginPath = b.CNIPluginPath
		if b.CNIPluginPath == "" {
			options.CNIPluginPath = util.DefaultCNIPluginPath
		}
	}

	switch isolation {
	case IsolationOCI:
		// The default is --rootless=auto, which makes troubleshooting a bit harder.
		// rootlessFlag := []string{"--rootless=false"}
		// for _, arg := range options.Args {
		// 	if strings.HasPrefix(arg, "--rootless") {
		// 		rootlessFlag = nil
		// 	}
		// }
		// options.Args = append(options.Args, rootlessFlag...)
		err = b.runUsingRuntimeSubproc(options, configureNetwork, configureNetworks, nil, spec, mountPoint, path, Package+"-"+filepath.Base(path))
	case IsolationChroot:
		err = chroot.RunUsingChroot(spec, path, options.Stdin, options.Stdout, options.Stderr)
	case IsolationOCIRootless:
		if err := setupRootlessSpecChanges(spec, path, rootUID, rootGID); err != nil {
			return err
		}
		rootlessFlag := []string{"--rootless=true"}
		for _, arg := range options.Args {
			if strings.HasPrefix(arg, "--rootless") {
				rootlessFlag = nil
			}
		}
		options.Args = append(options.Args, rootlessFlag...)
		err = b.runUsingRuntimeSubproc(options, configureNetwork, configureNetworks, []string{"--no-new-keyring"}, spec, mountPoint, path, Package+"-"+filepath.Base(path))
	default:
		err = errors.Errorf("don't know how to run this command")
	}
	return err
}

func checkAndOverrideIsolationOptions(isolation Isolation, options *RunOptions) error {
	switch isolation {
	case IsolationOCIRootless:
		if ns := options.NamespaceOptions.Find(string(specs.IPCNamespace)); ns == nil || ns.Host {
			logrus.Debugf("Forcing use of an IPC namespace.")
		}
		options.NamespaceOptions.AddOrReplace(NamespaceOption{Name: string(specs.IPCNamespace)})
		if ns := options.NamespaceOptions.Find(string(specs.NetworkNamespace)); ns != nil && !ns.Host {
			logrus.Debugf("Disabling network namespace.")
		}
		options.NamespaceOptions.AddOrReplace(NamespaceOption{Name: string(specs.NetworkNamespace), Host: true})
		if ns := options.NamespaceOptions.Find(string(specs.PIDNamespace)); ns == nil || ns.Host {
			logrus.Debugf("Forcing use of a PID namespace.")
		}
		options.NamespaceOptions.AddOrReplace(NamespaceOption{Name: string(specs.PIDNamespace), Host: false})
		if ns := options.NamespaceOptions.Find(string(specs.UserNamespace)); ns == nil || ns.Host {
			logrus.Debugf("Forcing use of a user namespace.")
		}
		options.NamespaceOptions.AddOrReplace(NamespaceOption{Name: string(specs.UserNamespace)})
		if ns := options.NamespaceOptions.Find(string(specs.UTSNamespace)); ns != nil && !ns.Host {
			logrus.Debugf("Disabling UTS namespace.")
		}
		options.NamespaceOptions.AddOrReplace(NamespaceOption{Name: string(specs.UTSNamespace), Host: true})
	case IsolationOCI:
		pidns := options.NamespaceOptions.Find(string(specs.PIDNamespace))
		userns := options.NamespaceOptions.Find(string(specs.UserNamespace))
		if (pidns == nil || pidns.Host) && (userns != nil && !userns.Host) {
			return fmt.Errorf("not allowed to mix host PID namespace with container user namespace")
		}
	}
	return nil
}

func setupRootlessSpecChanges(spec *specs.Spec, bundleDir string, rootUID, rootGID uint32) error {
	spec.Hostname = ""
	spec.Process.User.AdditionalGids = nil
	spec.Linux.Resources = nil

	emptyDir := filepath.Join(bundleDir, "empty")
	if err := os.Mkdir(emptyDir, 0); err != nil {
		return errors.Wrapf(err, "error creating %q", emptyDir)
	}

	// Replace /sys with a read-only bind mount.
	mounts := []specs.Mount{
		{
			Source:      "/dev",
			Destination: "/dev",
			Type:        "tmpfs",
			Options:     []string{"private", "strictatime", "noexec", "nosuid", "mode=755", "size=65536k"},
		},
		{
			Source:      "mqueue",
			Destination: "/dev/mqueue",
			Type:        "mqueue",
			Options:     []string{"private", "nodev", "noexec", "nosuid"},
		},
		{
			Source:      "pts",
			Destination: "/dev/pts",
			Type:        "devpts",
			Options:     []string{"private", "noexec", "nosuid", "newinstance", "ptmxmode=0666", "mode=0620"},
		},
		{
			Source:      "shm",
			Destination: "/dev/shm",
			Type:        "tmpfs",
			Options:     []string{"private", "nodev", "noexec", "nosuid", "mode=1777", "size=65536k"},
		},
		{
			Source:      "/proc",
			Destination: "/proc",
			Type:        "proc",
			Options:     []string{"private", "nodev", "noexec", "nosuid"},
		},
		{
			Source:      "/sys",
			Destination: "/sys",
			Type:        "bind",
			Options:     []string{bind.NoBindOption, "rbind", "private", "nodev", "noexec", "nosuid", "ro"},
		},
	}
	// Cover up /sys/fs/cgroup and /sys/fs/selinux, if they exist in our source for /sys.
	if _, err := os.Stat("/sys/fs/cgroup"); err == nil {
		spec.Linux.MaskedPaths = append(spec.Linux.MaskedPaths, "/sys/fs/cgroup")
	}
	if _, err := os.Stat("/sys/fs/selinux"); err == nil {
		spec.Linux.MaskedPaths = append(spec.Linux.MaskedPaths, "/sys/fs/selinux")
	}
	// Keep anything that isn't under /dev, /proc, or /sys.
	for i := range spec.Mounts {
		if spec.Mounts[i].Destination == "/dev" || strings.HasPrefix(spec.Mounts[i].Destination, "/dev/") ||
			spec.Mounts[i].Destination == "/proc" || strings.HasPrefix(spec.Mounts[i].Destination, "/proc/") ||
			spec.Mounts[i].Destination == "/sys" || strings.HasPrefix(spec.Mounts[i].Destination, "/sys/") {
			continue
		}
		mounts = append(mounts, spec.Mounts[i])
	}
	spec.Mounts = mounts
	return nil
}

type runUsingRuntimeSubprocOptions struct {
	Options           RunOptions
	Spec              *specs.Spec
	RootPath          string
	BundlePath        string
	ConfigureNetwork  bool
	ConfigureNetworks []string
	MoreCreateArgs    []string
	ContainerName     string
}

func (b *Builder) runUsingRuntimeSubproc(options RunOptions, configureNetwork bool, configureNetworks, moreCreateArgs []string, spec *specs.Spec, rootPath, bundlePath, containerName string) (err error) {
	var confwg sync.WaitGroup
	config, conferr := json.Marshal(runUsingRuntimeSubprocOptions{
		Options:           options,
		Spec:              spec,
		RootPath:          rootPath,
		BundlePath:        bundlePath,
		ConfigureNetwork:  configureNetwork,
		ConfigureNetworks: configureNetworks,
		MoreCreateArgs:    moreCreateArgs,
		ContainerName:     containerName,
	})
	if conferr != nil {
		return errors.Wrapf(conferr, "error encoding configuration for %q", runUsingRuntimeCommand)
	}
	cmd := reexec.Command(runUsingRuntimeCommand)
	cmd.Dir = bundlePath
	cmd.Stdin = options.Stdin
	if cmd.Stdin == nil {
		cmd.Stdin = os.Stdin
	}
	cmd.Stdout = options.Stdout
	if cmd.Stdout == nil {
		cmd.Stdout = os.Stdout
	}
	cmd.Stderr = options.Stderr
	if cmd.Stderr == nil {
		cmd.Stderr = os.Stderr
	}
	cmd.Env = append(os.Environ(), fmt.Sprintf("LOGLEVEL=%d", logrus.GetLevel()))
	preader, pwriter, err := os.Pipe()
	if err != nil {
		return errors.Wrapf(err, "error creating configuration pipe")
	}
	confwg.Add(1)
	go func() {
		_, conferr = io.Copy(pwriter, bytes.NewReader(config))
		confwg.Done()
	}()
	cmd.ExtraFiles = append([]*os.File{preader}, cmd.ExtraFiles...)
	defer preader.Close()
	defer pwriter.Close()
	err = cmd.Run()
	confwg.Wait()
	if err == nil {
		return conferr
	}
	return err
}

func init() {
	reexec.Register(runUsingRuntimeCommand, runUsingRuntimeMain)
}

func runUsingRuntimeMain() {
	var options runUsingRuntimeSubprocOptions
	// Set logging.
	if level := os.Getenv("LOGLEVEL"); level != "" {
		if ll, err := strconv.Atoi(level); err == nil {
			logrus.SetLevel(logrus.Level(ll))
		}
	}
	// Unpack our configuration.
	confPipe := os.NewFile(3, "confpipe")
	if confPipe == nil {
		fmt.Fprintf(os.Stderr, "error reading options pipe\n")
		os.Exit(1)
	}
	defer confPipe.Close()
	if err := json.NewDecoder(confPipe).Decode(&options); err != nil {
		fmt.Fprintf(os.Stderr, "error decoding options: %v\n", err)
		os.Exit(1)
	}
	// Set ourselves up to read the container's exit status.  We're doing this in a child process
	// so that we won't mess with the setting in a caller of the library. This stubs to OS specific
	// calls
	if err := setChildProcess(); err != nil {
		os.Exit(1)
	}
	// Run the container, start to finish.
	status, err := runUsingRuntime(options.Options, options.ConfigureNetwork, options.ConfigureNetworks, options.MoreCreateArgs, options.Spec, options.RootPath, options.BundlePath, options.ContainerName)
	if err != nil {
		fmt.Fprintf(os.Stderr, "error running container: %v\n", err)
		os.Exit(1)
	}
	// Pass the container's exit status back to the caller by exiting with the same status.
	if status.Exited() {
		os.Exit(status.ExitStatus())
	} else if status.Signaled() {
		fmt.Fprintf(os.Stderr, "container exited on %s\n", status.Signal())
		os.Exit(1)
	}
	os.Exit(1)
}

func runUsingRuntime(options RunOptions, configureNetwork bool, configureNetworks, moreCreateArgs []string, spec *specs.Spec, rootPath, bundlePath, containerName string) (wstatus unix.WaitStatus, err error) {
	// Lock the caller to a single OS-level thread.
	runtime.LockOSThread()

	// Set up bind mounts for things that a namespaced user might not be able to get to directly.
	unmountAll, err := bind.SetupIntermediateMountNamespace(spec, bundlePath)
	if unmountAll != nil {
		defer func() {
			if err := unmountAll(); err != nil {
				logrus.Error(err)
			}
		}()
	}
	if err != nil {
		return 1, err
	}

	// Write the runtime configuration.
	specbytes, err := json.Marshal(spec)
	if err != nil {
		return 1, err
	}
	if err = ioutils.AtomicWriteFile(filepath.Join(bundlePath, "config.json"), specbytes, 0600); err != nil {
		return 1, errors.Wrapf(err, "error storing runtime configuration")
	}

	logrus.Debugf("config = %v", string(specbytes))

	// Decide which runtime to use.
	runtime := options.Runtime
	if runtime == "" {
		runtime = util.Runtime()
	}

	// Default to just passing down our stdio.
	getCreateStdio := func() (io.ReadCloser, io.WriteCloser, io.WriteCloser) {
		return os.Stdin, os.Stdout, os.Stderr
	}

	// Figure out how we're doing stdio handling, and create pipes and sockets.
	var stdio sync.WaitGroup
	var consoleListener *net.UnixListener
	var errorFds, closeBeforeReadingErrorFds []int
	stdioPipe := make([][]int, 3)
	copyConsole := false
	copyPipes := false
	finishCopy := make([]int, 2)
	if err = unix.Pipe(finishCopy); err != nil {
		return 1, errors.Wrapf(err, "error creating pipe for notifying to stop stdio")
	}
	finishedCopy := make(chan struct{})
	if spec.Process != nil {
		if spec.Process.Terminal {
			copyConsole = true
			// Create a listening socket for accepting the container's terminal's PTY master.
			socketPath := filepath.Join(bundlePath, "console.sock")
			consoleListener, err = net.ListenUnix("unix", &net.UnixAddr{Name: socketPath, Net: "unix"})
			if err != nil {
				return 1, errors.Wrapf(err, "error creating socket to receive terminal descriptor")
			}
			// Add console socket arguments.
			moreCreateArgs = append(moreCreateArgs, "--console-socket", socketPath)
		} else {
			copyPipes = true
			// Figure out who should own the pipes.
			uid, gid, err := util.GetHostRootIDs(spec)
			if err != nil {
				return 1, err
			}
			// Create stdio pipes.
			if stdioPipe, err = runMakeStdioPipe(int(uid), int(gid)); err != nil {
				return 1, err
			}
			errorFds = []int{stdioPipe[unix.Stdout][0], stdioPipe[unix.Stderr][0]}
			closeBeforeReadingErrorFds = []int{stdioPipe[unix.Stdout][1], stdioPipe[unix.Stderr][1]}
			// Set stdio to our pipes.
			getCreateStdio = func() (io.ReadCloser, io.WriteCloser, io.WriteCloser) {
				stdin := os.NewFile(uintptr(stdioPipe[unix.Stdin][0]), "/dev/stdin")
				stdout := os.NewFile(uintptr(stdioPipe[unix.Stdout][1]), "/dev/stdout")
				stderr := os.NewFile(uintptr(stdioPipe[unix.Stderr][1]), "/dev/stderr")
				return stdin, stdout, stderr
			}
		}
	} else {
		if options.Quiet {
			// Discard stdout.
			getCreateStdio = func() (io.ReadCloser, io.WriteCloser, io.WriteCloser) {
				return os.Stdin, nil, os.Stderr
			}
		}
	}

	// Build the commands that we'll execute.
	pidFile := filepath.Join(bundlePath, "pid")
	args := append(append(append(options.Args, "create", "--bundle", bundlePath, "--pid-file", pidFile), moreCreateArgs...), containerName)
	create := exec.Command(runtime, args...)
	create.Dir = bundlePath
	stdin, stdout, stderr := getCreateStdio()
	create.Stdin, create.Stdout, create.Stderr = stdin, stdout, stderr
	if create.SysProcAttr == nil {
		create.SysProcAttr = &syscall.SysProcAttr{}
	}

	args = append(options.Args, "start", containerName)
	start := exec.Command(runtime, args...)
	start.Dir = bundlePath
	start.Stderr = os.Stderr

	args = append(options.Args, "kill", containerName)
	kill := exec.Command(runtime, args...)
	kill.Dir = bundlePath
	kill.Stderr = os.Stderr

	args = append(options.Args, "delete", containerName)
	del := exec.Command(runtime, args...)
	del.Dir = bundlePath
	del.Stderr = os.Stderr

	// Actually create the container.
	logrus.Debugf("Running %q", create.Args)
	err = create.Run()
	if err != nil {
		return 1, errors.Wrapf(err, "error creating container for %v: %s", spec.Process.Args, runCollectOutput(errorFds, closeBeforeReadingErrorFds))
	}
	defer func() {
		err2 := del.Run()
		if err2 != nil {
			if err == nil {
				err = errors.Wrapf(err2, "error deleting container")
			} else {
				logrus.Infof("error deleting container: %v", err2)
			}
		}
	}()

	// Make sure we read the container's exit status when it exits.
	pidValue, err := ioutil.ReadFile(pidFile)
	if err != nil {
		return 1, errors.Wrapf(err, "error reading pid from %q", pidFile)
	}
	pid, err := strconv.Atoi(strings.TrimSpace(string(pidValue)))
	if err != nil {
		return 1, errors.Wrapf(err, "error parsing pid %s as a number", string(pidValue))
	}
	var reaping sync.WaitGroup
	reaping.Add(1)
	go func() {
		defer reaping.Done()
		var err error
		_, err = unix.Wait4(pid, &wstatus, 0, nil)
		if err != nil {
			wstatus = 0
			logrus.Errorf("error waiting for container child process %d: %v\n", pid, err)
		}
	}()

	if configureNetwork {
		teardown, err := runConfigureNetwork(options, configureNetworks, pid, containerName, spec.Process.Args)
		if teardown != nil {
			defer teardown()
		}
		if err != nil {
			return 1, err
		}
	}

	if copyPipes {
		// We don't need the ends of the pipes that belong to the container.
		stdin.Close()
		if stdout != nil {
			stdout.Close()
		}
		stderr.Close()
	}

	// Handle stdio for the container in the background.
	stdio.Add(1)
	go runCopyStdio(&stdio, copyPipes, stdioPipe, copyConsole, consoleListener, finishCopy, finishedCopy, spec)

	// Start the container.
	logrus.Debugf("Running %q", start.Args)
	err = start.Run()
	if err != nil {
		return 1, errors.Wrapf(err, "error starting container")
	}
	stopped := false
	defer func() {
		if !stopped {
			err2 := kill.Run()
			if err2 != nil {
				if err == nil {
					err = errors.Wrapf(err2, "error stopping container")
				} else {
					logrus.Infof("error stopping container: %v", err2)
				}
			}
		}
	}()

	// Wait for the container to exit.
	for {
		now := time.Now()
		var state specs.State
		args = append(options.Args, "state", containerName)
		stat := exec.Command(runtime, args...)
		stat.Dir = bundlePath
		stat.Stderr = os.Stderr
		stateOutput, stateErr := stat.Output()
		if stateErr != nil {
			return 1, errors.Wrapf(stateErr, "error reading container state")
		}
		if err = json.Unmarshal(stateOutput, &state); err != nil {
			return 1, errors.Wrapf(stateErr, "error parsing container state %q", string(stateOutput))
		}
		switch state.Status {
		case "running":
		case "stopped":
			stopped = true
		default:
			return 1, errors.Errorf("container status unexpectedly changed to %q", state.Status)
		}
		if stopped {
			break
		}
		select {
		case <-finishedCopy:
			stopped = true
		case <-time.After(time.Until(now.Add(100 * time.Millisecond))):
			continue
		}
		if stopped {
			break
		}
	}

	// Close the writing end of the stop-handling-stdio notification pipe.
	unix.Close(finishCopy[1])
	// Wait for the stdio copy goroutine to flush.
	stdio.Wait()
	// Wait until we finish reading the exit status.
	reaping.Wait()

	return wstatus, nil
}

func runCollectOutput(fds, closeBeforeReadingFds []int) string {
	for _, fd := range closeBeforeReadingFds {
		unix.Close(fd)
	}
	var b bytes.Buffer
	buf := make([]byte, 8192)
	for _, fd := range fds {
		nread, err := unix.Read(fd, buf)
		if err != nil {
			if errno, isErrno := err.(syscall.Errno); isErrno {
				switch errno {
				default:
					logrus.Errorf("error reading from pipe %d: %v", fd, err)
				case syscall.EINTR, syscall.EAGAIN:
				}
			} else {
				logrus.Errorf("unable to wait for data from pipe %d: %v", fd, err)
			}
			continue
		}
		for nread > 0 {
			r := buf[:nread]
			if nwritten, err := b.Write(r); err != nil || nwritten != len(r) {
				if nwritten != len(r) {
					logrus.Errorf("error buffering data from pipe %d: %v", fd, err)
					break
				}
			}
			nread, err = unix.Read(fd, buf)
			if err != nil {
				if errno, isErrno := err.(syscall.Errno); isErrno {
					switch errno {
					default:
						logrus.Errorf("error reading from pipe %d: %v", fd, err)
					case syscall.EINTR, syscall.EAGAIN:
					}
				} else {
					logrus.Errorf("unable to wait for data from pipe %d: %v", fd, err)
				}
				break
			}
		}
	}
	return b.String()
}

func runConfigureNetwork(options RunOptions, configureNetworks []string, pid int, containerName string, command []string) (teardown func(), err error) {
	var netconf, undo []*libcni.NetworkConfigList
	// Scan for CNI configuration files.
	confdir := options.CNIConfigDir
	files, err := libcni.ConfFiles(confdir, []string{".conf"})
	if err != nil {
		return nil, errors.Wrapf(err, "error finding CNI networking configuration files named *.conf in directory %q", confdir)
	}
	lists, err := libcni.ConfFiles(confdir, []string{".conflist"})
	if err != nil {
		return nil, errors.Wrapf(err, "error finding CNI networking configuration list files named *.conflist in directory %q", confdir)
	}
	logrus.Debugf("CNI network configuration file list: %#v", append(files, lists...))
	// Read the CNI configuration files.
	for _, file := range files {
		nc, err := libcni.ConfFromFile(file)
		if err != nil {
			return nil, errors.Wrapf(err, "error loading networking configuration from file %q for %v", file, command)
		}
		if len(configureNetworks) > 0 && nc.Network != nil && (nc.Network.Name == "" || !util.StringInSlice(nc.Network.Name, configureNetworks)) {
			if nc.Network.Name == "" {
				logrus.Debugf("configuration in %q has no name, skipping it", file)
			} else {
				logrus.Debugf("configuration in %q has name %q, skipping it", file, nc.Network.Name)
			}
			continue
		}
		cl, err := libcni.ConfListFromConf(nc)
		if err != nil {
			return nil, errors.Wrapf(err, "error converting networking configuration from file %q for %v", file, command)
		}
		logrus.Debugf("using network configuration from %q", file)
		netconf = append(netconf, cl)
	}
	for _, list := range lists {
		cl, err := libcni.ConfListFromFile(list)
		if err != nil {
			return nil, errors.Wrapf(err, "error loading networking configuration list from file %q for %v", list, command)
		}
		if len(configureNetworks) > 0 && (cl.Name == "" || !util.StringInSlice(cl.Name, configureNetworks)) {
			if cl.Name == "" {
				logrus.Debugf("configuration list in %q has no name, skipping it", list)
			} else {
				logrus.Debugf("configuration list in %q has name %q, skipping it", list, cl.Name)
			}
			continue
		}
		logrus.Debugf("using network configuration list from %q", list)
		netconf = append(netconf, cl)
	}
	// Make sure we can access the container's network namespace,
	// even after it exits, to successfully tear down the
	// interfaces.  Ensure this by opening a handle to the network
	// namespace, and using our copy to both configure and
	// deconfigure it.
	netns := fmt.Sprintf("/proc/%d/ns/net", pid)
	netFD, err := unix.Open(netns, unix.O_RDONLY, 0)
	if err != nil {
		return nil, errors.Wrapf(err, "error opening network namespace for %v", command)
	}
	mynetns := fmt.Sprintf("/proc/%d/fd/%d", unix.Getpid(), netFD)
	// Build our search path for the plugins.
	pluginPaths := strings.Split(options.CNIPluginPath, string(os.PathListSeparator))
	cni := libcni.CNIConfig{Path: pluginPaths}
	// Configure the interfaces.
	rtconf := make(map[*libcni.NetworkConfigList]*libcni.RuntimeConf)
	teardown = func() {
		for _, nc := range undo {
			if err = cni.DelNetworkList(nc, rtconf[nc]); err != nil {
				logrus.Errorf("error cleaning up network %v for %v: %v", rtconf[nc].IfName, command, err)
			}
		}
		unix.Close(netFD)
	}
	for i, nc := range netconf {
		// Build the runtime config for use with this network configuration.
		rtconf[nc] = &libcni.RuntimeConf{
			ContainerID:    containerName,
			NetNS:          mynetns,
			IfName:         fmt.Sprintf("if%d", i),
			Args:           [][2]string{},
			CapabilityArgs: map[string]interface{}{},
		}
		// Bring it up.
		_, err := cni.AddNetworkList(nc, rtconf[nc])
		if err != nil {
			return teardown, errors.Wrapf(err, "error configuring network list %v for %v", rtconf[nc].IfName, command)
		}
		// Add it to the list of networks to take down when the container process exits.
		undo = append([]*libcni.NetworkConfigList{nc}, undo...)
	}
	return teardown, nil
}

func runCopyStdio(stdio *sync.WaitGroup, copyPipes bool, stdioPipe [][]int, copyConsole bool, consoleListener *net.UnixListener, finishCopy []int, finishedCopy chan struct{}, spec *specs.Spec) {
	defer func() {
		unix.Close(finishCopy[0])
		if copyPipes {
			unix.Close(stdioPipe[unix.Stdin][1])
			unix.Close(stdioPipe[unix.Stdout][0])
			unix.Close(stdioPipe[unix.Stderr][0])
		}
		stdio.Done()
		finishedCopy <- struct{}{}
	}()
	// Map describing where data on an incoming descriptor should go.
	relayMap := make(map[int]int)
	// Map describing incoming and outgoing descriptors.
	readDesc := make(map[int]string)
	writeDesc := make(map[int]string)
	// Buffers.
	relayBuffer := make(map[int]*bytes.Buffer)
	// Set up the terminal descriptor or pipes for polling.
	if copyConsole {
		// Accept a connection over our listening socket.
		fd, err := runAcceptTerminal(consoleListener, spec.Process.ConsoleSize)
		if err != nil {
			logrus.Errorf("%v", err)
			return
		}
		terminalFD := fd
		// Input from our stdin, output from the terminal descriptor.
		relayMap[unix.Stdin] = terminalFD
		readDesc[unix.Stdin] = "stdin"
		relayBuffer[terminalFD] = new(bytes.Buffer)
		writeDesc[terminalFD] = "container terminal input"
		relayMap[terminalFD] = unix.Stdout
		readDesc[terminalFD] = "container terminal output"
		relayBuffer[unix.Stdout] = new(bytes.Buffer)
		writeDesc[unix.Stdout] = "output"
		// Set our terminal's mode to raw, to pass handling of special
		// terminal input to the terminal in the container.
		if terminal.IsTerminal(unix.Stdin) {
			if state, err := terminal.MakeRaw(unix.Stdin); err != nil {
				logrus.Warnf("error setting terminal state: %v", err)
			} else {
				defer func() {
					if err = terminal.Restore(unix.Stdin, state); err != nil {
						logrus.Errorf("unable to restore terminal state: %v", err)
					}
				}()
			}
		}
	}
	if copyPipes {
		// Input from our stdin, output from the stdout and stderr pipes.
		relayMap[unix.Stdin] = stdioPipe[unix.Stdin][1]
		readDesc[unix.Stdin] = "stdin"
		relayBuffer[stdioPipe[unix.Stdin][1]] = new(bytes.Buffer)
		writeDesc[stdioPipe[unix.Stdin][1]] = "container stdin"
		relayMap[stdioPipe[unix.Stdout][0]] = unix.Stdout
		readDesc[stdioPipe[unix.Stdout][0]] = "container stdout"
		relayBuffer[unix.Stdout] = new(bytes.Buffer)
		writeDesc[unix.Stdout] = "stdout"
		relayMap[stdioPipe[unix.Stderr][0]] = unix.Stderr
		readDesc[stdioPipe[unix.Stderr][0]] = "container stderr"
		relayBuffer[unix.Stderr] = new(bytes.Buffer)
		writeDesc[unix.Stderr] = "stderr"
	}
	// Set our reading descriptors to non-blocking.
	for rfd, wfd := range relayMap {
		if err := unix.SetNonblock(rfd, true); err != nil {
			logrus.Errorf("error setting %s to nonblocking: %v", readDesc[rfd], err)
			return
		}
		if err := unix.SetNonblock(wfd, false); err != nil {
			logrus.Errorf("error setting descriptor %d (%s) blocking: %v", wfd, writeDesc[wfd], err)
		}
	}
	// A helper that returns false if err is an error that would cause us
	// to give up.
	logIfNotRetryable := func(err error, what string) (retry bool) {
		if err == nil {
			return true
		}
		if errno, isErrno := err.(syscall.Errno); isErrno {
			switch errno {
			case syscall.EINTR, syscall.EAGAIN:
				return true
			}
		}
		logrus.Error(what)
		return false
	}
	// Pass data back and forth.
	pollTimeout := -1
	for len(relayMap) > 0 {
		// Start building the list of descriptors to poll.
		pollFds := make([]unix.PollFd, 0, len(relayMap)+1)
		// Poll for a notification that we should stop handling stdio.
		pollFds = append(pollFds, unix.PollFd{Fd: int32(finishCopy[0]), Events: unix.POLLIN | unix.POLLHUP})
		// Poll on our reading descriptors.
		for rfd := range relayMap {
			pollFds = append(pollFds, unix.PollFd{Fd: int32(rfd), Events: unix.POLLIN | unix.POLLHUP})
		}
		buf := make([]byte, 8192)
		// Wait for new data from any input descriptor, or a notification that we're done.
		_, err := unix.Poll(pollFds, pollTimeout)
		if !logIfNotRetryable(err, fmt.Sprintf("error waiting for stdio/terminal data to relay: %v", err)) {
			return
		}
		removes := make(map[int]struct{})
		for _, pollFd := range pollFds {
			// If this descriptor's just been closed from the other end, mark it for
			// removal from the set that we're checking for.
			if pollFd.Revents&unix.POLLHUP == unix.POLLHUP {
				removes[int(pollFd.Fd)] = struct{}{}
			}
			// If the descriptor was closed elsewhere, remove it from our list.
			if pollFd.Revents&unix.POLLNVAL != 0 {
				logrus.Debugf("error polling descriptor %s: closed?", readDesc[int(pollFd.Fd)])
				removes[int(pollFd.Fd)] = struct{}{}
			}
			// If the POLLIN flag isn't set, then there's no data to be read from this descriptor.
			if pollFd.Revents&unix.POLLIN == 0 {
				// If we're using pipes and it's our stdin and it's closed, close the writing
				// end of the corresponding pipe.
				if copyPipes && int(pollFd.Fd) == unix.Stdin && pollFd.Revents&unix.POLLHUP != 0 {
					unix.Close(stdioPipe[unix.Stdin][1])
					stdioPipe[unix.Stdin][1] = -1
				}
				continue
			}
			// Read whatever there is to be read.
			readFD := int(pollFd.Fd)
			writeFD, needToRelay := relayMap[readFD]
			if needToRelay {
				n, err := unix.Read(readFD, buf)
				if !logIfNotRetryable(err, fmt.Sprintf("unable to read %s data: %v", readDesc[readFD], err)) {
					return
				}
				// If it's zero-length on our stdin and we're
				// using pipes, it's an EOF, so close the stdin
				// pipe's writing end.
				if n == 0 && copyPipes && int(pollFd.Fd) == unix.Stdin {
					unix.Close(stdioPipe[unix.Stdin][1])
					stdioPipe[unix.Stdin][1] = -1
				}
				if n > 0 {
					// Buffer the data in case we get blocked on where they need to go.
					nwritten, err := relayBuffer[writeFD].Write(buf[:n])
					if err != nil {
						logrus.Debugf("buffer: %v", err)
						continue
					}
					if nwritten != n {
						logrus.Debugf("buffer: expected to buffer %d bytes, wrote %d", n, nwritten)
						continue
					}
					// If this is the last of the data we'll be able to read from this
					// descriptor, read all that there is to read.
					for pollFd.Revents&unix.POLLHUP == unix.POLLHUP {
						nr, err := unix.Read(readFD, buf)
						logIfNotRetryable(err, fmt.Sprintf("read %s: %v", readDesc[readFD], err))
						if nr <= 0 {
							break
						}
						nwritten, err := relayBuffer[writeFD].Write(buf[:nr])
						if err != nil {
							logrus.Debugf("buffer: %v", err)
							break
						}
						if nwritten != nr {
							logrus.Debugf("buffer: expected to buffer %d bytes, wrote %d", nr, nwritten)
							break
						}
					}
				}
			}
		}
		// Try to drain the output buffers.  Set the default timeout
		// for the next poll() to 100ms if we still have data to write.
		pollTimeout = -1
		for writeFD := range relayBuffer {
			if relayBuffer[writeFD].Len() > 0 {
				n, err := unix.Write(writeFD, relayBuffer[writeFD].Bytes())
				if !logIfNotRetryable(err, fmt.Sprintf("unable to write %s data: %v", writeDesc[writeFD], err)) {
					return
				}
				if n > 0 {
					relayBuffer[writeFD].Next(n)
				}
			}
			if relayBuffer[writeFD].Len() > 0 {
				pollTimeout = 100
			}
		}
		// Remove any descriptors which we don't need to poll any more from the poll descriptor list.
		for remove := range removes {
			delete(relayMap, remove)
		}
		// If the we-can-return pipe had anything for us, we're done.
		for _, pollFd := range pollFds {
			if int(pollFd.Fd) == finishCopy[0] && pollFd.Revents != 0 {
				// The pipe is closed, indicating that we can stop now.
				return
			}
		}
	}
}

func runAcceptTerminal(consoleListener *net.UnixListener, terminalSize *specs.Box) (int, error) {
	defer consoleListener.Close()
	c, err := consoleListener.AcceptUnix()
	if err != nil {
		return -1, errors.Wrapf(err, "error accepting socket descriptor connection")
	}
	defer c.Close()
	// Expect a control message over our new connection.
	b := make([]byte, 8192)
	oob := make([]byte, 8192)
	n, oobn, _, _, err := c.ReadMsgUnix(b, oob)
	if err != nil {
		return -1, errors.Wrapf(err, "error reading socket descriptor: %v")
	}
	if n > 0 {
		logrus.Debugf("socket descriptor is for %q", string(b[:n]))
	}
	if oobn > len(oob) {
		return -1, errors.Errorf("too much out-of-bounds data (%d bytes)", oobn)
	}
	// Parse the control message.
	scm, err := unix.ParseSocketControlMessage(oob[:oobn])
	if err != nil {
		return -1, errors.Wrapf(err, "error parsing out-of-bound data as a socket control message")
	}
	logrus.Debugf("control messages: %v", scm)
	// Expect to get a descriptor.
	terminalFD := -1
	for i := range scm {
		fds, err := unix.ParseUnixRights(&scm[i])
		if err != nil {
			return -1, errors.Wrapf(err, "error parsing unix rights control message: %v")
		}
		logrus.Debugf("fds: %v", fds)
		if len(fds) == 0 {
			continue
		}
		terminalFD = fds[0]
		break
	}
	if terminalFD == -1 {
		return -1, errors.Errorf("unable to read terminal descriptor")
	}
	// Set the pseudoterminal's size to the configured size, or our own.
	winsize := &unix.Winsize{}
	if terminalSize != nil {
		// Use configured sizes.
		winsize.Row = uint16(terminalSize.Height)
		winsize.Col = uint16(terminalSize.Width)
	} else {
		if terminal.IsTerminal(unix.Stdin) {
			// Use the size of our terminal.
			if winsize, err = unix.IoctlGetWinsize(unix.Stdin, unix.TIOCGWINSZ); err != nil {
				logrus.Warnf("error reading size of controlling terminal: %v", err)
				winsize.Row = 0
				winsize.Col = 0
			}
		}
	}
	if winsize.Row != 0 && winsize.Col != 0 {
		if err = unix.IoctlSetWinsize(terminalFD, unix.TIOCSWINSZ, winsize); err != nil {
			logrus.Warnf("error setting size of container pseudoterminal: %v", err)
		}
		// FIXME - if we're connected to a terminal, we should
		// be passing the updated terminal size down when we
		// receive a SIGWINCH.
	}
	return terminalFD, nil
}

// Create pipes to use for relaying stdio.
func runMakeStdioPipe(uid, gid int) ([][]int, error) {
	stdioPipe := make([][]int, 3)
	for i := range stdioPipe {
		stdioPipe[i] = make([]int, 2)
		if err := unix.Pipe(stdioPipe[i]); err != nil {
			return nil, errors.Wrapf(err, "error creating pipe for container FD %d", i)
		}
	}
	if err := unix.Fchown(stdioPipe[unix.Stdin][0], uid, gid); err != nil {
		return nil, errors.Wrapf(err, "error setting owner of stdin pipe descriptor")
	}
	if err := unix.Fchown(stdioPipe[unix.Stdout][1], uid, gid); err != nil {
		return nil, errors.Wrapf(err, "error setting owner of stdout pipe descriptor")
	}
	if err := unix.Fchown(stdioPipe[unix.Stderr][1], uid, gid); err != nil {
		return nil, errors.Wrapf(err, "error setting owner of stderr pipe descriptor")
	}
	return stdioPipe, nil
}