aboutsummaryrefslogtreecommitdiff
path: root/sys/kern/vfs_lookup.c
blob: d05e1658a0048449410fa5543a113345410c0181 (plain) (blame)
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
/*-
 * SPDX-License-Identifier: BSD-3-Clause
 *
 * Copyright (c) 1982, 1986, 1989, 1993
 *	The Regents of the University of California.  All rights reserved.
 * (c) UNIX System Laboratories, Inc.
 * All or some portions of this file are derived from material licensed
 * to the University of California by American Telephone and Telegraph
 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
 * the permission of UNIX System Laboratories, Inc.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. Neither the name of the University nor the names of its contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 *
 *	@(#)vfs_lookup.c	8.4 (Berkeley) 2/16/94
 */

#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");

#include "opt_capsicum.h"
#include "opt_ktrace.h"

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/dirent.h>
#include <sys/kernel.h>
#include <sys/capsicum.h>
#include <sys/fcntl.h>
#include <sys/jail.h>
#include <sys/lock.h>
#include <sys/mutex.h>
#include <sys/namei.h>
#include <sys/vnode.h>
#include <sys/mount.h>
#include <sys/filedesc.h>
#include <sys/proc.h>
#include <sys/sdt.h>
#include <sys/syscallsubr.h>
#include <sys/sysctl.h>
#ifdef KTRACE
#include <sys/ktrace.h>
#endif
#ifdef INVARIANTS
#include <machine/_inttypes.h>
#endif

#include <security/audit/audit.h>
#include <security/mac/mac_framework.h>

#include <vm/uma.h>

#define	NAMEI_DIAGNOSTIC 1
#undef NAMEI_DIAGNOSTIC

SDT_PROVIDER_DEFINE(vfs);
SDT_PROBE_DEFINE4(vfs, namei, lookup, entry, "struct vnode *", "char *",
    "unsigned long", "bool");
SDT_PROBE_DEFINE4(vfs, namei, lookup, return, "int", "struct vnode *", "bool",
    "struct nameidata");

/* Allocation zone for namei. */
uma_zone_t namei_zone;

/* Placeholder vnode for mp traversal. */
static struct vnode *vp_crossmp;

static int
crossmp_vop_islocked(struct vop_islocked_args *ap)
{

	return (LK_SHARED);
}

static int
crossmp_vop_lock1(struct vop_lock1_args *ap)
{
	struct vnode *vp;
	struct lock *lk __diagused;
	const char *file __diagused;
	int flags, line __diagused;

	vp = ap->a_vp;
	lk = vp->v_vnlock;
	flags = ap->a_flags;
	file = ap->a_file;
	line = ap->a_line;

	if ((flags & LK_SHARED) == 0)
		panic("invalid lock request for crossmp");

	WITNESS_CHECKORDER(&lk->lock_object, LOP_NEWORDER, file, line,
	    flags & LK_INTERLOCK ? &VI_MTX(vp)->lock_object : NULL);
	WITNESS_LOCK(&lk->lock_object, 0, file, line);
	if ((flags & LK_INTERLOCK) != 0)
		VI_UNLOCK(vp);
	LOCK_LOG_LOCK("SLOCK", &lk->lock_object, 0, 0, ap->a_file, line);
	return (0);
}

static int
crossmp_vop_unlock(struct vop_unlock_args *ap)
{
	struct vnode *vp;
	struct lock *lk __diagused;

	vp = ap->a_vp;
	lk = vp->v_vnlock;

	WITNESS_UNLOCK(&lk->lock_object, 0, LOCK_FILE, LOCK_LINE);
	LOCK_LOG_LOCK("SUNLOCK", &lk->lock_object, 0, 0, LOCK_FILE,
	    LOCK_LINE);
	return (0);
}

static struct vop_vector crossmp_vnodeops = {
	.vop_default =		&default_vnodeops,
	.vop_islocked =		crossmp_vop_islocked,
	.vop_lock1 =		crossmp_vop_lock1,
	.vop_unlock =		crossmp_vop_unlock,
};
/*
 * VFS_VOP_VECTOR_REGISTER(crossmp_vnodeops) is not used here since the vnode
 * gets allocated early. See nameiinit for the direct call below.
 */

struct nameicap_tracker {
	struct vnode *dp;
	TAILQ_ENTRY(nameicap_tracker) nm_link;
};

/* Zone for cap mode tracker elements used for dotdot capability checks. */
MALLOC_DEFINE(M_NAMEITRACKER, "namei_tracker", "namei tracking for dotdot");

static void
nameiinit(void *dummy __unused)
{

	namei_zone = uma_zcreate("NAMEI", MAXPATHLEN, NULL, NULL, NULL, NULL,
	    UMA_ALIGN_PTR, 0);
	vfs_vector_op_register(&crossmp_vnodeops);
	getnewvnode("crossmp", NULL, &crossmp_vnodeops, &vp_crossmp);
}
SYSINIT(vfs, SI_SUB_VFS, SI_ORDER_SECOND, nameiinit, NULL);

static int lookup_cap_dotdot = 1;
SYSCTL_INT(_vfs, OID_AUTO, lookup_cap_dotdot, CTLFLAG_RWTUN,
    &lookup_cap_dotdot, 0,
    "enables \"..\" components in path lookup in capability mode");
static int lookup_cap_dotdot_nonlocal = 1;
SYSCTL_INT(_vfs, OID_AUTO, lookup_cap_dotdot_nonlocal, CTLFLAG_RWTUN,
    &lookup_cap_dotdot_nonlocal, 0,
    "enables \"..\" components in path lookup in capability mode "
    "on non-local mount");

static void
nameicap_tracker_add(struct nameidata *ndp, struct vnode *dp)
{
	struct nameicap_tracker *nt;

	if ((ndp->ni_lcf & NI_LCF_CAP_DOTDOT) == 0 || dp->v_type != VDIR)
		return;
	nt = TAILQ_LAST(&ndp->ni_cap_tracker, nameicap_tracker_head);
	if (nt != NULL && nt->dp == dp)
		return;
	nt = malloc(sizeof(*nt), M_NAMEITRACKER, M_WAITOK);
	vhold(dp);
	nt->dp = dp;
	TAILQ_INSERT_TAIL(&ndp->ni_cap_tracker, nt, nm_link);
}

static void
nameicap_cleanup_from(struct nameidata *ndp, struct nameicap_tracker *first)
{
	struct nameicap_tracker *nt, *nt1;

	nt = first;
	TAILQ_FOREACH_FROM_SAFE(nt, &ndp->ni_cap_tracker, nm_link, nt1) {
		TAILQ_REMOVE(&ndp->ni_cap_tracker, nt, nm_link);
		vdrop(nt->dp);
		free(nt, M_NAMEITRACKER);
	}
}

static void
nameicap_cleanup(struct nameidata *ndp)
{
	KASSERT(TAILQ_EMPTY(&ndp->ni_cap_tracker) ||
	    (ndp->ni_lcf & NI_LCF_CAP_DOTDOT) != 0, ("not strictrelative"));
	nameicap_cleanup_from(ndp, NULL);
}

/*
 * For dotdot lookups in capability mode, only allow the component
 * lookup to succeed if the resulting directory was already traversed
 * during the operation.  This catches situations where already
 * traversed directory is moved to different parent, and then we walk
 * over it with dotdots.
 *
 * Also allow to force failure of dotdot lookups for non-local
 * filesystems, where external agents might assist local lookups to
 * escape the compartment.
 */
static int
nameicap_check_dotdot(struct nameidata *ndp, struct vnode *dp)
{
	struct nameicap_tracker *nt;
	struct mount *mp;

	if (dp == NULL || dp->v_type != VDIR || (ndp->ni_lcf &
	    NI_LCF_STRICTRELATIVE) == 0)
		return (0);
	if ((ndp->ni_lcf & NI_LCF_CAP_DOTDOT) == 0)
		return (ENOTCAPABLE);
	mp = dp->v_mount;
	if (lookup_cap_dotdot_nonlocal == 0 && mp != NULL &&
	    (mp->mnt_flag & MNT_LOCAL) == 0)
		return (ENOTCAPABLE);
	TAILQ_FOREACH_REVERSE(nt, &ndp->ni_cap_tracker, nameicap_tracker_head,
	    nm_link) {
		if (dp == nt->dp) {
			nt = TAILQ_NEXT(nt, nm_link);
			if (nt != NULL)
				nameicap_cleanup_from(ndp, nt);
			return (0);
		}
	}
	return (ENOTCAPABLE);
}

static void
namei_cleanup_cnp(struct componentname *cnp)
{

	uma_zfree(namei_zone, cnp->cn_pnbuf);
#ifdef DIAGNOSTIC
	cnp->cn_pnbuf = NULL;
	cnp->cn_nameptr = NULL;
#endif
}

static int
namei_handle_root(struct nameidata *ndp, struct vnode **dpp)
{
	struct componentname *cnp;

	cnp = &ndp->ni_cnd;
	if ((ndp->ni_lcf & NI_LCF_STRICTRELATIVE) != 0) {
#ifdef KTRACE
		if (KTRPOINT(curthread, KTR_CAPFAIL))
			ktrcapfail(CAPFAIL_LOOKUP, NULL, NULL);
#endif
		return (ENOTCAPABLE);
	}
	while (*(cnp->cn_nameptr) == '/') {
		cnp->cn_nameptr++;
		ndp->ni_pathlen--;
	}
	*dpp = ndp->ni_rootdir;
	vrefact(*dpp);
	return (0);
}

static int
namei_setup(struct nameidata *ndp, struct vnode **dpp, struct pwd **pwdp)
{
	struct componentname *cnp;
	struct file *dfp;
	struct thread *td;
	struct pwd *pwd;
	cap_rights_t rights;
	int error;
	bool startdir_used;

	cnp = &ndp->ni_cnd;
	td = curthread;

	startdir_used = false;
	*pwdp = NULL;
	*dpp = NULL;

#ifdef CAPABILITY_MODE
	/*
	 * In capability mode, lookups must be restricted to happen in
	 * the subtree with the root specified by the file descriptor:
	 * - The root must be real file descriptor, not the pseudo-descriptor
	 *   AT_FDCWD.
	 * - The passed path must be relative and not absolute.
	 * - If lookup_cap_dotdot is disabled, path must not contain the
	 *   '..' components.
	 * - If lookup_cap_dotdot is enabled, we verify that all '..'
	 *   components lookups result in the directories which were
	 *   previously walked by us, which prevents an escape from
	 *   the relative root.
	 */
	if (IN_CAPABILITY_MODE(td) && (cnp->cn_flags & NOCAPCHECK) == 0) {
		ndp->ni_lcf |= NI_LCF_STRICTRELATIVE;
		ndp->ni_resflags |= NIRES_STRICTREL;
		if (ndp->ni_dirfd == AT_FDCWD) {
#ifdef KTRACE
			if (KTRPOINT(td, KTR_CAPFAIL))
				ktrcapfail(CAPFAIL_LOOKUP, NULL, NULL);
#endif
			return (ECAPMODE);
		}
	}
#endif
	error = 0;

	/*
	 * Get starting point for the translation.
	 */
	pwd = pwd_hold(td);
	/*
	 * The reference on ni_rootdir is acquired in the block below to avoid
	 * back-to-back atomics for absolute lookups.
	 */
	ndp->ni_rootdir = pwd->pwd_rdir;
	ndp->ni_topdir = pwd->pwd_jdir;

	if (cnp->cn_pnbuf[0] == '/') {
		ndp->ni_resflags |= NIRES_ABS;
		error = namei_handle_root(ndp, dpp);
	} else {
		if (ndp->ni_startdir != NULL) {
			*dpp = ndp->ni_startdir;
			startdir_used = true;
		} else if (ndp->ni_dirfd == AT_FDCWD) {
			*dpp = pwd->pwd_cdir;
			vrefact(*dpp);
		} else {
			rights = *ndp->ni_rightsneeded;
			cap_rights_set_one(&rights, CAP_LOOKUP);

			if (cnp->cn_flags & AUDITVNODE1)
				AUDIT_ARG_ATFD1(ndp->ni_dirfd);
			if (cnp->cn_flags & AUDITVNODE2)
				AUDIT_ARG_ATFD2(ndp->ni_dirfd);
			/*
			 * Effectively inlined fgetvp_rights, because
			 * we need to inspect the file as well as
			 * grabbing the vnode.  No check for O_PATH,
			 * files to implement its semantic.
			 */
			error = fget_cap(td, ndp->ni_dirfd, &rights,
			    &dfp, &ndp->ni_filecaps);
			if (error != 0) {
				/*
				 * Preserve the error; it should either be EBADF
				 * or capability-related, both of which can be
				 * safely returned to the caller.
				 */
			} else {
				if (dfp->f_ops == &badfileops) {
					error = EBADF;
				} else if (dfp->f_vnode == NULL) {
					error = ENOTDIR;
				} else {
					*dpp = dfp->f_vnode;
					vref(*dpp);

					if ((dfp->f_flag & FSEARCH) != 0)
						cnp->cn_flags |= NOEXECCHECK;
				}
				fdrop(dfp, td);
			}
#ifdef CAPABILITIES
			/*
			 * If file descriptor doesn't have all rights,
			 * all lookups relative to it must also be
			 * strictly relative.
			 */
			CAP_ALL(&rights);
			if (!cap_rights_contains(&ndp->ni_filecaps.fc_rights,
			    &rights) ||
			    ndp->ni_filecaps.fc_fcntls != CAP_FCNTL_ALL ||
			    ndp->ni_filecaps.fc_nioctls != -1) {
				ndp->ni_lcf |= NI_LCF_STRICTRELATIVE;
				ndp->ni_resflags |= NIRES_STRICTREL;
			}
#endif
		}
		if (error == 0 && (*dpp)->v_type != VDIR &&
		    (cnp->cn_pnbuf[0] != '\0' ||
		    (cnp->cn_flags & EMPTYPATH) == 0))
			error = ENOTDIR;
	}
	if (error == 0 && (cnp->cn_flags & RBENEATH) != 0) {
		if (cnp->cn_pnbuf[0] == '/') {
			error = ENOTCAPABLE;
		} else if ((ndp->ni_lcf & NI_LCF_STRICTRELATIVE) == 0) {
			ndp->ni_lcf |= NI_LCF_STRICTRELATIVE |
			    NI_LCF_CAP_DOTDOT;
		}
	}

	/*
	 * If we are auditing the kernel pathname, save the user pathname.
	 */
	if (cnp->cn_flags & AUDITVNODE1)
		AUDIT_ARG_UPATH1_VP(td, ndp->ni_rootdir, *dpp, cnp->cn_pnbuf);
	if (cnp->cn_flags & AUDITVNODE2)
		AUDIT_ARG_UPATH2_VP(td, ndp->ni_rootdir, *dpp, cnp->cn_pnbuf);
	if (ndp->ni_startdir != NULL && !startdir_used)
		vrele(ndp->ni_startdir);
	if (error != 0) {
		if (*dpp != NULL)
			vrele(*dpp);
		pwd_drop(pwd);
		return (error);
	}
	if ((ndp->ni_lcf & NI_LCF_STRICTRELATIVE) != 0 &&
	    lookup_cap_dotdot != 0)
		ndp->ni_lcf |= NI_LCF_CAP_DOTDOT;
	SDT_PROBE4(vfs, namei, lookup, entry, *dpp, cnp->cn_pnbuf,
	    cnp->cn_flags, false);
	*pwdp = pwd;
	return (0);
}

static int
namei_getpath(struct nameidata *ndp)
{
	struct componentname *cnp;
	int error;

	cnp = &ndp->ni_cnd;

	/*
	 * Get a buffer for the name to be translated, and copy the
	 * name into the buffer.
	 */
	cnp->cn_pnbuf = uma_zalloc(namei_zone, M_WAITOK);
	if (ndp->ni_segflg == UIO_SYSSPACE) {
		error = copystr(ndp->ni_dirp, cnp->cn_pnbuf, MAXPATHLEN,
		    &ndp->ni_pathlen);
	} else {
		error = copyinstr(ndp->ni_dirp, cnp->cn_pnbuf, MAXPATHLEN,
		    &ndp->ni_pathlen);
	}

	if (__predict_false(error != 0))
		return (error);

	cnp->cn_nameptr = cnp->cn_pnbuf;
	return (0);
}

static int
namei_emptypath(struct nameidata *ndp)
{
	struct componentname *cnp;
	struct pwd *pwd;
	struct vnode *dp;
	int error;

	cnp = &ndp->ni_cnd;
	MPASS(*cnp->cn_pnbuf == '\0');
	MPASS((cnp->cn_flags & EMPTYPATH) != 0);
	MPASS((cnp->cn_flags & (LOCKPARENT | WANTPARENT)) == 0);

	ndp->ni_resflags |= NIRES_EMPTYPATH;
	error = namei_setup(ndp, &dp, &pwd);
	if (error != 0) {
		namei_cleanup_cnp(cnp);
		goto errout;
	}

	/*
	 * Usecount on dp already provided by namei_setup.
	 */
	ndp->ni_vp = dp;
	namei_cleanup_cnp(cnp);
	pwd_drop(pwd);
	NDVALIDATE(ndp);
	if ((cnp->cn_flags & LOCKLEAF) != 0) {
		VOP_LOCK(dp, (cnp->cn_flags & LOCKSHARED) != 0 ?
		    LK_SHARED : LK_EXCLUSIVE);
		if (VN_IS_DOOMED(dp)) {
			vput(dp);
			error = ENOENT;
			goto errout;
		}
	}
	SDT_PROBE4(vfs, namei, lookup, return, 0, ndp->ni_vp, false, ndp);
	return (0);

errout:
	SDT_PROBE4(vfs, namei, lookup, return, error, NULL, false, ndp);
	return (error);
}

/*
 * Convert a pathname into a pointer to a locked vnode.
 *
 * The FOLLOW flag is set when symbolic links are to be followed
 * when they occur at the end of the name translation process.
 * Symbolic links are always followed for all other pathname
 * components other than the last.
 *
 * The segflg defines whether the name is to be copied from user
 * space or kernel space.
 *
 * Overall outline of namei:
 *
 *	copy in name
 *	get starting directory
 *	while (!done && !error) {
 *		call lookup to search path.
 *		if symbolic link, massage name in buffer and continue
 *	}
 */
int
namei(struct nameidata *ndp)
{
	char *cp;		/* pointer into pathname argument */
	struct vnode *dp;	/* the directory we are searching */
	struct iovec aiov;		/* uio for reading symbolic links */
	struct componentname *cnp;
	struct thread *td;
	struct pwd *pwd;
	struct uio auio;
	int error, linklen;
	enum cache_fpl_status status;

	cnp = &ndp->ni_cnd;
	td = curthread;
#ifdef INVARIANTS
	KASSERT((ndp->ni_debugflags & NAMEI_DBG_CALLED) == 0,
	    ("%s: repeated call to namei without NDREINIT", __func__));
	KASSERT(ndp->ni_debugflags == NAMEI_DBG_INITED,
	    ("%s: bad debugflags %d", __func__, ndp->ni_debugflags));
	ndp->ni_debugflags |= NAMEI_DBG_CALLED;
	if (ndp->ni_startdir != NULL)
		ndp->ni_debugflags |= NAMEI_DBG_HADSTARTDIR;
	if (cnp->cn_flags & FAILIFEXISTS) {
		KASSERT(cnp->cn_nameiop == CREATE,
		    ("%s: FAILIFEXISTS passed for op %d", __func__, cnp->cn_nameiop));
		/*
		 * The limitation below is to restrict hairy corner cases.
		 */
		KASSERT((cnp->cn_flags & (LOCKPARENT | LOCKLEAF)) == LOCKPARENT,
		    ("%s: FAILIFEXISTS must be passed with LOCKPARENT and without LOCKLEAF",
		    __func__));
	}
	/*
	 * For NDVALIDATE.
	 *
	 * While NDINIT may seem like a more natural place to do it, there are
	 * callers which directly modify flags past invoking init.
	 */
	cnp->cn_origflags = cnp->cn_flags;
#endif
	ndp->ni_cnd.cn_cred = td->td_ucred;
	KASSERT(ndp->ni_resflags == 0, ("%s: garbage in ni_resflags: %x\n",
	    __func__, ndp->ni_resflags));
	KASSERT(cnp->cn_cred && td->td_proc, ("namei: bad cred/proc"));
	KASSERT((cnp->cn_flags & NAMEI_INTERNAL_FLAGS) == 0,
	    ("namei: unexpected flags: %" PRIx64 "\n",
	    cnp->cn_flags & NAMEI_INTERNAL_FLAGS));
	if (cnp->cn_flags & NOCACHE)
		KASSERT(cnp->cn_nameiop != LOOKUP,
		    ("%s: NOCACHE passed with LOOKUP", __func__));
	MPASS(ndp->ni_startdir == NULL || ndp->ni_startdir->v_type == VDIR ||
	    ndp->ni_startdir->v_type == VBAD);

	ndp->ni_lcf = 0;
	ndp->ni_loopcnt = 0;
	ndp->ni_vp = NULL;

	error = namei_getpath(ndp);
	if (__predict_false(error != 0)) {
		namei_cleanup_cnp(cnp);
		SDT_PROBE4(vfs, namei, lookup, return, error, NULL,
		    false, ndp);
		return (error);
	}

#ifdef KTRACE
	if (KTRPOINT(td, KTR_NAMEI)) {
		ktrnamei(cnp->cn_pnbuf);
	}
#endif
	TSNAMEI(curthread->td_proc->p_pid, cnp->cn_pnbuf);

	/*
	 * First try looking up the target without locking any vnodes.
	 *
	 * We may need to start from scratch or pick up where it left off.
	 */
	error = cache_fplookup(ndp, &status, &pwd);
	switch (status) {
	case CACHE_FPL_STATUS_UNSET:
		__assert_unreachable();
		break;
	case CACHE_FPL_STATUS_HANDLED:
		if (error == 0)
			NDVALIDATE(ndp);
		return (error);
	case CACHE_FPL_STATUS_PARTIAL:
		TAILQ_INIT(&ndp->ni_cap_tracker);
		dp = ndp->ni_startdir;
		break;
	case CACHE_FPL_STATUS_DESTROYED:
		ndp->ni_loopcnt = 0;
		error = namei_getpath(ndp);
		if (__predict_false(error != 0)) {
			namei_cleanup_cnp(cnp);
			return (error);
		}
		/* FALLTHROUGH */
	case CACHE_FPL_STATUS_ABORTED:
		TAILQ_INIT(&ndp->ni_cap_tracker);
		MPASS(ndp->ni_lcf == 0);
		if (*cnp->cn_pnbuf == '\0') {
			if ((cnp->cn_flags & EMPTYPATH) != 0) {
				return (namei_emptypath(ndp));
			}
			namei_cleanup_cnp(cnp);
			SDT_PROBE4(vfs, namei, lookup, return, ENOENT, NULL,
			    false, ndp);
			return (ENOENT);
		}
		error = namei_setup(ndp, &dp, &pwd);
		if (error != 0) {
			namei_cleanup_cnp(cnp);
			return (error);
		}
		break;
	}

	/*
	 * Locked lookup.
	 */
	for (;;) {
		ndp->ni_startdir = dp;
		error = lookup(ndp);
		if (error != 0)
			goto out;

		/*
		 * If not a symbolic link, we're done.
		 */
		if ((cnp->cn_flags & ISSYMLINK) == 0) {
			SDT_PROBE4(vfs, namei, lookup, return, error,
			    (error == 0 ? ndp->ni_vp : NULL), false, ndp);
			if ((cnp->cn_flags & (SAVENAME | SAVESTART)) == 0) {
				namei_cleanup_cnp(cnp);
			} else
				cnp->cn_flags |= HASBUF;
			nameicap_cleanup(ndp);
			pwd_drop(pwd);
			if (error == 0)
				NDVALIDATE(ndp);
			return (error);
		}
		if (ndp->ni_loopcnt++ >= MAXSYMLINKS) {
			error = ELOOP;
			break;
		}
#ifdef MAC
		if ((cnp->cn_flags & NOMACCHECK) == 0) {
			error = mac_vnode_check_readlink(td->td_ucred,
			    ndp->ni_vp);
			if (error != 0)
				break;
		}
#endif
		if (ndp->ni_pathlen > 1)
			cp = uma_zalloc(namei_zone, M_WAITOK);
		else
			cp = cnp->cn_pnbuf;
		aiov.iov_base = cp;
		aiov.iov_len = MAXPATHLEN;
		auio.uio_iov = &aiov;
		auio.uio_iovcnt = 1;
		auio.uio_offset = 0;
		auio.uio_rw = UIO_READ;
		auio.uio_segflg = UIO_SYSSPACE;
		auio.uio_td = td;
		auio.uio_resid = MAXPATHLEN;
		error = VOP_READLINK(ndp->ni_vp, &auio, cnp->cn_cred);
		if (error != 0) {
			if (ndp->ni_pathlen > 1)
				uma_zfree(namei_zone, cp);
			break;
		}
		linklen = MAXPATHLEN - auio.uio_resid;
		if (linklen == 0) {
			if (ndp->ni_pathlen > 1)
				uma_zfree(namei_zone, cp);
			error = ENOENT;
			break;
		}
		if (linklen + ndp->ni_pathlen > MAXPATHLEN) {
			if (ndp->ni_pathlen > 1)
				uma_zfree(namei_zone, cp);
			error = ENAMETOOLONG;
			break;
		}
		if (ndp->ni_pathlen > 1) {
			bcopy(ndp->ni_next, cp + linklen, ndp->ni_pathlen);
			uma_zfree(namei_zone, cnp->cn_pnbuf);
			cnp->cn_pnbuf = cp;
		} else
			cnp->cn_pnbuf[linklen] = '\0';
		ndp->ni_pathlen += linklen;
		vput(ndp->ni_vp);
		dp = ndp->ni_dvp;
		/*
		 * Check if root directory should replace current directory.
		 */
		cnp->cn_nameptr = cnp->cn_pnbuf;
		if (*(cnp->cn_nameptr) == '/') {
			vrele(dp);
			error = namei_handle_root(ndp, &dp);
			if (error != 0)
				goto out;
		}
	}
	vput(ndp->ni_vp);
	ndp->ni_vp = NULL;
	vrele(ndp->ni_dvp);
out:
	MPASS(error != 0);
	SDT_PROBE4(vfs, namei, lookup, return, error, NULL, false, ndp);
	namei_cleanup_cnp(cnp);
	nameicap_cleanup(ndp);
	pwd_drop(pwd);
	return (error);
}

static int
compute_cn_lkflags(struct mount *mp, int lkflags, int cnflags)
{

	if (mp == NULL || ((lkflags & LK_SHARED) &&
	    (!(mp->mnt_kern_flag & MNTK_LOOKUP_SHARED) ||
	    ((cnflags & ISDOTDOT) &&
	    (mp->mnt_kern_flag & MNTK_LOOKUP_EXCL_DOTDOT))))) {
		lkflags &= ~LK_SHARED;
		lkflags |= LK_EXCLUSIVE;
	}
	lkflags |= LK_NODDLKTREAT;
	return (lkflags);
}

static __inline int
needs_exclusive_leaf(struct mount *mp, int flags)
{

	/*
	 * Intermediate nodes can use shared locks, we only need to
	 * force an exclusive lock for leaf nodes.
	 */
	if ((flags & (ISLASTCN | LOCKLEAF)) != (ISLASTCN | LOCKLEAF))
		return (0);

	/* Always use exclusive locks if LOCKSHARED isn't set. */
	if (!(flags & LOCKSHARED))
		return (1);

	/*
	 * For lookups during open(), if the mount point supports
	 * extended shared operations, then use a shared lock for the
	 * leaf node, otherwise use an exclusive lock.
	 */
	if ((flags & ISOPEN) != 0)
		return (!MNT_EXTENDED_SHARED(mp));

	/*
	 * Lookup requests outside of open() that specify LOCKSHARED
	 * only need a shared lock on the leaf vnode.
	 */
	return (0);
}

/*
 * Various filesystems expect to be able to copy a name component with length
 * bounded by NAME_MAX into a directory entry buffer of size MAXNAMLEN.  Make
 * sure that these are the same size.
 */
_Static_assert(MAXNAMLEN == NAME_MAX,
    "MAXNAMLEN and NAME_MAX have different values");

/*
 * Search a pathname.
 * This is a very central and rather complicated routine.
 *
 * The pathname is pointed to by ni_ptr and is of length ni_pathlen.
 * The starting directory is taken from ni_startdir. The pathname is
 * descended until done, or a symbolic link is encountered. The variable
 * ni_more is clear if the path is completed; it is set to one if a
 * symbolic link needing interpretation is encountered.
 *
 * The flag argument is LOOKUP, CREATE, RENAME, or DELETE depending on
 * whether the name is to be looked up, created, renamed, or deleted.
 * When CREATE, RENAME, or DELETE is specified, information usable in
 * creating, renaming, or deleting a directory entry may be calculated.
 * If flag has LOCKPARENT or'ed into it, the parent directory is returned
 * locked. If flag has WANTPARENT or'ed into it, the parent directory is
 * returned unlocked. Otherwise the parent directory is not returned. If
 * the target of the pathname exists and LOCKLEAF is or'ed into the flag
 * the target is returned locked, otherwise it is returned unlocked.
 * When creating or renaming and LOCKPARENT is specified, the target may not
 * be ".".  When deleting and LOCKPARENT is specified, the target may be ".".
 *
 * Overall outline of lookup:
 *
 * dirloop:
 *	identify next component of name at ndp->ni_ptr
 *	handle degenerate case where name is null string
 *	if .. and crossing mount points and on mounted filesys, find parent
 *	call VOP_LOOKUP routine for next component name
 *	    directory vnode returned in ni_dvp, unlocked unless LOCKPARENT set
 *	    component vnode returned in ni_vp (if it exists), locked.
 *	if result vnode is mounted on and crossing mount points,
 *	    find mounted on vnode
 *	if more components of name, do next level at dirloop
 *	return the answer in ni_vp, locked if LOCKLEAF set
 *	    if LOCKPARENT set, return locked parent in ni_dvp
 *	    if WANTPARENT set, return unlocked parent in ni_dvp
 */
int
lookup(struct nameidata *ndp)
{
	char *cp;			/* pointer into pathname argument */
	char *prev_ni_next;		/* saved ndp->ni_next */
	char *nulchar;			/* location of '\0' in cn_pnbuf */
	struct vnode *dp = NULL;	/* the directory we are searching */
	struct vnode *tdp;		/* saved dp */
	struct mount *mp;		/* mount table entry */
	struct prison *pr;
	size_t prev_ni_pathlen;		/* saved ndp->ni_pathlen */
	int docache;			/* == 0 do not cache last component */
	int wantparent;			/* 1 => wantparent or lockparent flag */
	int rdonly;			/* lookup read-only flag bit */
	int error = 0;
	int dpunlocked = 0;		/* dp has already been unlocked */
	int relookup = 0;		/* do not consume the path component */
	struct componentname *cnp = &ndp->ni_cnd;
	int lkflags_save;
	int ni_dvp_unlocked;

	/*
	 * Setup: break out flag bits into variables.
	 */
	ni_dvp_unlocked = 0;
	wantparent = cnp->cn_flags & (LOCKPARENT | WANTPARENT);
	KASSERT(cnp->cn_nameiop == LOOKUP || wantparent,
	    ("CREATE, DELETE, RENAME require LOCKPARENT or WANTPARENT."));
	/*
	 * When set to zero, docache causes the last component of the
	 * pathname to be deleted from the cache and the full lookup
	 * of the name to be done (via VOP_CACHEDLOOKUP()). Often
	 * filesystems need some pre-computed values that are made
	 * during the full lookup, for instance UFS sets dp->i_offset.
	 *
	 * The docache variable is set to zero when requested by the
	 * NOCACHE flag and for all modifying operations except CREATE.
	 */
	docache = (cnp->cn_flags & NOCACHE) ^ NOCACHE;
	if (cnp->cn_nameiop == DELETE ||
	    (wantparent && cnp->cn_nameiop != CREATE &&
	     cnp->cn_nameiop != LOOKUP))
		docache = 0;
	rdonly = cnp->cn_flags & RDONLY;
	cnp->cn_flags &= ~ISSYMLINK;
	ndp->ni_dvp = NULL;
	/*
	 * We use shared locks until we hit the parent of the last cn then
	 * we adjust based on the requesting flags.
	 */
	cnp->cn_lkflags = LK_SHARED;
	dp = ndp->ni_startdir;
	ndp->ni_startdir = NULLVP;
	vn_lock(dp,
	    compute_cn_lkflags(dp->v_mount, cnp->cn_lkflags | LK_RETRY,
	    cnp->cn_flags));

dirloop:
	/*
	 * Search a new directory.
	 *
	 * The last component of the filename is left accessible via
	 * cnp->cn_nameptr for callers that need the name. Callers needing
	 * the name set the SAVENAME flag. When done, they assume
	 * responsibility for freeing the pathname buffer.
	 *
	 * Store / as a temporary sentinel so that we only have one character
	 * to test for. Pathnames tend to be short so this should not be
	 * resulting in cache misses.
	 */
	nulchar = &cnp->cn_nameptr[ndp->ni_pathlen - 1];
	KASSERT(*nulchar == '\0',
	    ("%s: expected nul at %p; string [%s]\n", __func__, nulchar,
	    cnp->cn_pnbuf));
	*nulchar = '/';
	for (cp = cnp->cn_nameptr; *cp != '/'; cp++) {
		KASSERT(*cp != '\0',
		    ("%s: encountered unexpected nul; string [%s]\n", __func__,
		    cnp->cn_nameptr));
		continue;
	}
	*nulchar = '\0';
	cnp->cn_namelen = cp - cnp->cn_nameptr;
	if (cnp->cn_namelen > NAME_MAX) {
		error = ENAMETOOLONG;
		goto bad;
	}
#ifdef NAMEI_DIAGNOSTIC
	{ char c = *cp;
	*cp = '\0';
	printf("{%s}: ", cnp->cn_nameptr);
	*cp = c; }
#endif
	prev_ni_pathlen = ndp->ni_pathlen;
	ndp->ni_pathlen -= cnp->cn_namelen;
	KASSERT(ndp->ni_pathlen <= PATH_MAX,
	    ("%s: ni_pathlen underflow to %zd\n", __func__, ndp->ni_pathlen));
	prev_ni_next = ndp->ni_next;
	ndp->ni_next = cp;

	/*
	 * Replace multiple slashes by a single slash and trailing slashes
	 * by a null.  This must be done before VOP_LOOKUP() because some
	 * fs's don't know about trailing slashes.  Remember if there were
	 * trailing slashes to handle symlinks, existing non-directories
	 * and non-existing files that won't be directories specially later.
	 */
	while (*cp == '/' && (cp[1] == '/' || cp[1] == '\0')) {
		cp++;
		ndp->ni_pathlen--;
		if (*cp == '\0') {
			*ndp->ni_next = '\0';
			cnp->cn_flags |= TRAILINGSLASH;
		}
	}
	ndp->ni_next = cp;

	cnp->cn_flags |= MAKEENTRY;
	if (*cp == '\0' && docache == 0)
		cnp->cn_flags &= ~MAKEENTRY;
	if (cnp->cn_namelen == 2 &&
	    cnp->cn_nameptr[1] == '.' && cnp->cn_nameptr[0] == '.')
		cnp->cn_flags |= ISDOTDOT;
	else
		cnp->cn_flags &= ~ISDOTDOT;
	if (*ndp->ni_next == 0)
		cnp->cn_flags |= ISLASTCN;
	else
		cnp->cn_flags &= ~ISLASTCN;

	if ((cnp->cn_flags & ISLASTCN) != 0 &&
	    cnp->cn_namelen == 1 && cnp->cn_nameptr[0] == '.' &&
	    (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)) {
		error = EINVAL;
		goto bad;
	}

	nameicap_tracker_add(ndp, dp);

	/*
	 * Check for degenerate name (e.g. / or "")
	 * which is a way of talking about a directory,
	 * e.g. like "/." or ".".
	 */
	if (cnp->cn_nameptr[0] == '\0') {
		if (dp->v_type != VDIR) {
			error = ENOTDIR;
			goto bad;
		}
		if (cnp->cn_nameiop != LOOKUP) {
			error = EISDIR;
			goto bad;
		}
		if (wantparent) {
			ndp->ni_dvp = dp;
			VREF(dp);
		}
		ndp->ni_vp = dp;

		if (cnp->cn_flags & AUDITVNODE1)
			AUDIT_ARG_VNODE1(dp);
		else if (cnp->cn_flags & AUDITVNODE2)
			AUDIT_ARG_VNODE2(dp);

		if (!(cnp->cn_flags & (LOCKPARENT | LOCKLEAF)))
			VOP_UNLOCK(dp);
		/* XXX This should probably move to the top of function. */
		if (cnp->cn_flags & SAVESTART)
			panic("lookup: SAVESTART");
		goto success;
	}

	/*
	 * Handle "..": five special cases.
	 * 0. If doing a capability lookup and lookup_cap_dotdot is
	 *    disabled, return ENOTCAPABLE.
	 * 1. Return an error if this is the last component of
	 *    the name and the operation is DELETE or RENAME.
	 * 2. If at root directory (e.g. after chroot)
	 *    or at absolute root directory
	 *    then ignore it so can't get out.
	 * 3. If this vnode is the root of a mounted
	 *    filesystem, then replace it with the
	 *    vnode which was mounted on so we take the
	 *    .. in the other filesystem.
	 * 4. If the vnode is the top directory of
	 *    the jail or chroot, don't let them out.
	 * 5. If doing a capability lookup and lookup_cap_dotdot is
	 *    enabled, return ENOTCAPABLE if the lookup would escape
	 *    from the initial file descriptor directory.  Checks are
	 *    done by ensuring that namei() already traversed the
	 *    result of dotdot lookup.
	 */
	if (cnp->cn_flags & ISDOTDOT) {
		if ((ndp->ni_lcf & (NI_LCF_STRICTRELATIVE | NI_LCF_CAP_DOTDOT))
		    == NI_LCF_STRICTRELATIVE) {
#ifdef KTRACE
			if (KTRPOINT(curthread, KTR_CAPFAIL))
				ktrcapfail(CAPFAIL_LOOKUP, NULL, NULL);
#endif
			error = ENOTCAPABLE;
			goto bad;
		}
		if ((cnp->cn_flags & ISLASTCN) != 0 &&
		    (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)) {
			error = EINVAL;
			goto bad;
		}
		for (;;) {
			for (pr = cnp->cn_cred->cr_prison; pr != NULL;
			     pr = pr->pr_parent)
				if (dp == pr->pr_root)
					break;
			if (dp == ndp->ni_rootdir || 
			    dp == ndp->ni_topdir || 
			    dp == rootvnode ||
			    pr != NULL ||
			    ((dp->v_vflag & VV_ROOT) != 0 &&
			     (cnp->cn_flags & NOCROSSMOUNT) != 0)) {
				ndp->ni_dvp = dp;
				ndp->ni_vp = dp;
				VREF(dp);
				goto nextname;
			}
			if ((dp->v_vflag & VV_ROOT) == 0)
				break;
			if (VN_IS_DOOMED(dp)) {	/* forced unmount */
				error = ENOENT;
				goto bad;
			}
			tdp = dp;
			dp = dp->v_mount->mnt_vnodecovered;
			VREF(dp);
			vput(tdp);
			vn_lock(dp,
			    compute_cn_lkflags(dp->v_mount, cnp->cn_lkflags |
			    LK_RETRY, ISDOTDOT));
			error = nameicap_check_dotdot(ndp, dp);
			if (error != 0) {
#ifdef KTRACE
				if (KTRPOINT(curthread, KTR_CAPFAIL))
					ktrcapfail(CAPFAIL_LOOKUP, NULL, NULL);
#endif
				goto bad;
			}
		}
	}

	/*
	 * We now have a segment name to search for, and a directory to search.
	 */
unionlookup:
#ifdef MAC
	error = mac_vnode_check_lookup(cnp->cn_cred, dp, cnp);
	if (error)
		goto bad;
#endif
	ndp->ni_dvp = dp;
	ndp->ni_vp = NULL;
	ASSERT_VOP_LOCKED(dp, "lookup");
	/*
	 * If we have a shared lock we may need to upgrade the lock for the
	 * last operation.
	 */
	if ((cnp->cn_flags & LOCKPARENT) && (cnp->cn_flags & ISLASTCN) &&
	    dp != vp_crossmp && VOP_ISLOCKED(dp) == LK_SHARED)
		vn_lock(dp, LK_UPGRADE|LK_RETRY);
	if (VN_IS_DOOMED(dp)) {
		error = ENOENT;
		goto bad;
	}
	/*
	 * If we're looking up the last component and we need an exclusive
	 * lock, adjust our lkflags.
	 */
	if (needs_exclusive_leaf(dp->v_mount, cnp->cn_flags))
		cnp->cn_lkflags = LK_EXCLUSIVE;
#ifdef NAMEI_DIAGNOSTIC
	vn_printf(dp, "lookup in ");
#endif
	lkflags_save = cnp->cn_lkflags;
	cnp->cn_lkflags = compute_cn_lkflags(dp->v_mount, cnp->cn_lkflags,
	    cnp->cn_flags);
	error = VOP_LOOKUP(dp, &ndp->ni_vp, cnp);
	cnp->cn_lkflags = lkflags_save;
	if (error != 0) {
		KASSERT(ndp->ni_vp == NULL, ("leaf should be empty"));
#ifdef NAMEI_DIAGNOSTIC
		printf("not found\n");
#endif
		if ((error == ENOENT) &&
		    (dp->v_vflag & VV_ROOT) && (dp->v_mount != NULL) &&
		    (dp->v_mount->mnt_flag & MNT_UNION)) {
			tdp = dp;
			dp = dp->v_mount->mnt_vnodecovered;
			VREF(dp);
			vput(tdp);
			vn_lock(dp,
			    compute_cn_lkflags(dp->v_mount, cnp->cn_lkflags |
			    LK_RETRY, cnp->cn_flags));
			nameicap_tracker_add(ndp, dp);
			goto unionlookup;
		}

		if (error == ERELOOKUP) {
			vref(dp);
			ndp->ni_vp = dp;
			error = 0;
			relookup = 1;
			goto good;
		}

		if (error != EJUSTRETURN)
			goto bad;
		/*
		 * At this point, we know we're at the end of the
		 * pathname.  If creating / renaming, we can consider
		 * allowing the file or directory to be created / renamed,
		 * provided we're not on a read-only filesystem.
		 */
		if (rdonly) {
			error = EROFS;
			goto bad;
		}
		/* trailing slash only allowed for directories */
		if ((cnp->cn_flags & TRAILINGSLASH) &&
		    !(cnp->cn_flags & WILLBEDIR)) {
			error = ENOENT;
			goto bad;
		}
		if ((cnp->cn_flags & LOCKPARENT) == 0)
			VOP_UNLOCK(dp);
		/*
		 * We return with ni_vp NULL to indicate that the entry
		 * doesn't currently exist, leaving a pointer to the
		 * (possibly locked) directory vnode in ndp->ni_dvp.
		 */
		if (cnp->cn_flags & SAVESTART) {
			ndp->ni_startdir = ndp->ni_dvp;
			VREF(ndp->ni_startdir);
		}
		goto success;
	}

good:
#ifdef NAMEI_DIAGNOSTIC
	printf("found\n");
#endif
	dp = ndp->ni_vp;

	/*
	 * Check to see if the vnode has been mounted on;
	 * if so find the root of the mounted filesystem.
	 */
	while (dp->v_type == VDIR && (mp = dp->v_mountedhere) &&
	       (cnp->cn_flags & NOCROSSMOUNT) == 0) {
		if (vfs_busy(mp, 0))
			continue;
		vput(dp);
		if (dp != ndp->ni_dvp)
			vput(ndp->ni_dvp);
		else
			vrele(ndp->ni_dvp);
		vrefact(vp_crossmp);
		ndp->ni_dvp = vp_crossmp;
		error = VFS_ROOT(mp, compute_cn_lkflags(mp, cnp->cn_lkflags,
		    cnp->cn_flags), &tdp);
		vfs_unbusy(mp);
		if (vn_lock(vp_crossmp, LK_SHARED | LK_NOWAIT))
			panic("vp_crossmp exclusively locked or reclaimed");
		if (error) {
			dpunlocked = 1;
			goto bad2;
		}
		ndp->ni_vp = dp = tdp;
	}

	/*
	 * Check for symbolic link
	 */
	if ((dp->v_type == VLNK) &&
	    ((cnp->cn_flags & FOLLOW) || (cnp->cn_flags & TRAILINGSLASH) ||
	     *ndp->ni_next == '/')) {
		cnp->cn_flags |= ISSYMLINK;
		if (VN_IS_DOOMED(dp)) {
			/*
			 * We can't know whether the directory was mounted with
			 * NOSYMFOLLOW, so we can't follow safely.
			 */
			error = ENOENT;
			goto bad2;
		}
		if (dp->v_mount->mnt_flag & MNT_NOSYMFOLLOW) {
			error = EACCES;
			goto bad2;
		}
		/*
		 * Symlink code always expects an unlocked dvp.
		 */
		if (ndp->ni_dvp != ndp->ni_vp) {
			VOP_UNLOCK(ndp->ni_dvp);
			ni_dvp_unlocked = 1;
		}
		goto success;
	}

nextname:
	/*
	 * Not a symbolic link that we will follow.  Continue with the
	 * next component if there is any; otherwise, we're done.
	 */
	KASSERT((cnp->cn_flags & ISLASTCN) || *ndp->ni_next == '/',
	    ("lookup: invalid path state."));
	if (relookup) {
		relookup = 0;
		ndp->ni_pathlen = prev_ni_pathlen;
		ndp->ni_next = prev_ni_next;
		if (ndp->ni_dvp != dp)
			vput(ndp->ni_dvp);
		else
			vrele(ndp->ni_dvp);
		goto dirloop;
	}
	if (cnp->cn_flags & ISDOTDOT) {
		error = nameicap_check_dotdot(ndp, ndp->ni_vp);
		if (error != 0) {
#ifdef KTRACE
			if (KTRPOINT(curthread, KTR_CAPFAIL))
				ktrcapfail(CAPFAIL_LOOKUP, NULL, NULL);
#endif
			goto bad2;
		}
	}
	if (*ndp->ni_next == '/') {
		cnp->cn_nameptr = ndp->ni_next;
		while (*cnp->cn_nameptr == '/') {
			cnp->cn_nameptr++;
			ndp->ni_pathlen--;
		}
		if (ndp->ni_dvp != dp)
			vput(ndp->ni_dvp);
		else
			vrele(ndp->ni_dvp);
		goto dirloop;
	}
	/*
	 * If we're processing a path with a trailing slash,
	 * check that the end result is a directory.
	 */
	if ((cnp->cn_flags & TRAILINGSLASH) && dp->v_type != VDIR) {
		error = ENOTDIR;
		goto bad2;
	}
	/*
	 * Disallow directory write attempts on read-only filesystems.
	 */
	if (rdonly &&
	    (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)) {
		error = EROFS;
		goto bad2;
	}
	if (cnp->cn_flags & SAVESTART) {
		ndp->ni_startdir = ndp->ni_dvp;
		VREF(ndp->ni_startdir);
	}
	if (!wantparent) {
		ni_dvp_unlocked = 2;
		if (ndp->ni_dvp != dp)
			vput(ndp->ni_dvp);
		else
			vrele(ndp->ni_dvp);
	} else if ((cnp->cn_flags & LOCKPARENT) == 0 && ndp->ni_dvp != dp) {
		VOP_UNLOCK(ndp->ni_dvp);
		ni_dvp_unlocked = 1;
	}

	if (cnp->cn_flags & AUDITVNODE1)
		AUDIT_ARG_VNODE1(dp);
	else if (cnp->cn_flags & AUDITVNODE2)
		AUDIT_ARG_VNODE2(dp);

	if ((cnp->cn_flags & LOCKLEAF) == 0)
		VOP_UNLOCK(dp);
success:
	/*
	 * FIXME: for lookups which only cross a mount point to fetch the
	 * root vnode, ni_dvp will be set to vp_crossmp. This can be a problem
	 * if either WANTPARENT or LOCKPARENT is set.
	 */
	/*
	 * Because of shared lookup we may have the vnode shared locked, but
	 * the caller may want it to be exclusively locked.
	 */
	if (needs_exclusive_leaf(dp->v_mount, cnp->cn_flags) &&
	    VOP_ISLOCKED(dp) != LK_EXCLUSIVE) {
		vn_lock(dp, LK_UPGRADE | LK_RETRY);
		if (VN_IS_DOOMED(dp)) {
			error = ENOENT;
			goto bad2;
		}
	}
	if (ndp->ni_vp != NULL) {
		if ((cnp->cn_flags & ISDOTDOT) == 0)
			nameicap_tracker_add(ndp, ndp->ni_vp);
		if ((cnp->cn_flags & (FAILIFEXISTS | ISSYMLINK)) == FAILIFEXISTS)
			goto bad_eexist;
	}
	return (0);

bad2:
	if (ni_dvp_unlocked != 2) {
		if (dp != ndp->ni_dvp && !ni_dvp_unlocked)
			vput(ndp->ni_dvp);
		else
			vrele(ndp->ni_dvp);
	}
bad:
	if (!dpunlocked)
		vput(dp);
	ndp->ni_vp = NULL;
	return (error);
bad_eexist:
	/*
	 * FAILIFEXISTS handling.
	 *
	 * XXX namei called with LOCKPARENT but not LOCKLEAF has the strange
	 * behaviour of leaving the vnode unlocked if the target is the same
	 * vnode as the parent.
	 */
	MPASS((cnp->cn_flags & ISSYMLINK) == 0);
	if (ndp->ni_vp == ndp->ni_dvp)
		vrele(ndp->ni_dvp);
	else
		vput(ndp->ni_dvp);
	vrele(ndp->ni_vp);
	ndp->ni_dvp = NULL;
	ndp->ni_vp = NULL;
	NDFREE(ndp, NDF_ONLY_PNBUF);
	return (EEXIST);
}

/*
 * relookup - lookup a path name component
 *    Used by lookup to re-acquire things.
 */
int
relookup(struct vnode *dvp, struct vnode **vpp, struct componentname *cnp)
{
	struct vnode *dp = NULL;		/* the directory we are searching */
	int rdonly;			/* lookup read-only flag bit */
	int error = 0;

	KASSERT(cnp->cn_flags & ISLASTCN,
	    ("relookup: Not given last component."));
	/*
	 * Setup: break out flag bits into variables.
	 */
	KASSERT((cnp->cn_flags & (LOCKPARENT | WANTPARENT)) != 0,
	    ("relookup: parent not wanted"));
	rdonly = cnp->cn_flags & RDONLY;
	cnp->cn_flags &= ~ISSYMLINK;
	dp = dvp;
	cnp->cn_lkflags = LK_EXCLUSIVE;
	vn_lock(dp, LK_EXCLUSIVE | LK_RETRY);

	/*
	 * Search a new directory.
	 *
	 * The last component of the filename is left accessible via
	 * cnp->cn_nameptr for callers that need the name. Callers needing
	 * the name set the SAVENAME flag. When done, they assume
	 * responsibility for freeing the pathname buffer.
	 */
#ifdef NAMEI_DIAGNOSTIC
	printf("{%s}: ", cnp->cn_nameptr);
#endif

	/*
	 * Check for "" which represents the root directory after slash
	 * removal.
	 */
	if (cnp->cn_nameptr[0] == '\0') {
		/*
		 * Support only LOOKUP for "/" because lookup()
		 * can't succeed for CREATE, DELETE and RENAME.
		 */
		KASSERT(cnp->cn_nameiop == LOOKUP, ("nameiop must be LOOKUP"));
		KASSERT(dp->v_type == VDIR, ("dp is not a directory"));

		if (!(cnp->cn_flags & LOCKLEAF))
			VOP_UNLOCK(dp);
		*vpp = dp;
		/* XXX This should probably move to the top of function. */
		if (cnp->cn_flags & SAVESTART)
			panic("lookup: SAVESTART");
		return (0);
	}

	if (cnp->cn_flags & ISDOTDOT)
		panic ("relookup: lookup on dot-dot");

	/*
	 * We now have a segment name to search for, and a directory to search.
	 */
#ifdef NAMEI_DIAGNOSTIC
	vn_printf(dp, "search in ");
#endif
	if ((error = VOP_LOOKUP(dp, vpp, cnp)) != 0) {
		KASSERT(*vpp == NULL, ("leaf should be empty"));
		if (error != EJUSTRETURN)
			goto bad;
		/*
		 * If creating and at end of pathname, then can consider
		 * allowing file to be created.
		 */
		if (rdonly) {
			error = EROFS;
			goto bad;
		}
		/* ASSERT(dvp == ndp->ni_startdir) */
		if (cnp->cn_flags & SAVESTART)
			VREF(dvp);
		if ((cnp->cn_flags & LOCKPARENT) == 0)
			VOP_UNLOCK(dp);
		/*
		 * We return with ni_vp NULL to indicate that the entry
		 * doesn't currently exist, leaving a pointer to the
		 * (possibly locked) directory vnode in ndp->ni_dvp.
		 */
		return (0);
	}

	dp = *vpp;

	/*
	 * Disallow directory write attempts on read-only filesystems.
	 */
	if (rdonly &&
	    (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)) {
		if (dvp == dp)
			vrele(dvp);
		else
			vput(dvp);
		error = EROFS;
		goto bad;
	}
	/*
	 * Set the parent lock/ref state to the requested state.
	 */
	if ((cnp->cn_flags & LOCKPARENT) == 0 && dvp != dp)
		VOP_UNLOCK(dvp);
	/*
	 * Check for symbolic link
	 */
	KASSERT(dp->v_type != VLNK || !(cnp->cn_flags & FOLLOW),
	    ("relookup: symlink found.\n"));

	/* ASSERT(dvp == ndp->ni_startdir) */
	if (cnp->cn_flags & SAVESTART)
		VREF(dvp);

	if ((cnp->cn_flags & LOCKLEAF) == 0)
		VOP_UNLOCK(dp);
	return (0);
bad:
	vput(dp);
	*vpp = NULL;
	return (error);
}

/*
 * Free data allocated by namei(); see namei(9) for details.
 */
void
NDFREE_PNBUF(struct nameidata *ndp)
{

	if ((ndp->ni_cnd.cn_flags & HASBUF) != 0) {
		MPASS((ndp->ni_cnd.cn_flags & (SAVENAME | SAVESTART)) != 0);
		uma_zfree(namei_zone, ndp->ni_cnd.cn_pnbuf);
		ndp->ni_cnd.cn_flags &= ~HASBUF;
	}
}

/*
 * NDFREE_PNBUF replacement for callers that know there is no buffer.
 *
 * This is a hack. Preferably the VFS layer would not produce anything more
 * than it was asked to do. Unfortunately several non-LOOKUP cases can add the
 * HASBUF flag to the result. Even then an interface could be implemented where
 * the caller specifies what they expected to see in the result and what they
 * are going to take care of.
 *
 * In the meantime provide this kludge as a trivial replacement for NDFREE_PNBUF
 * calls scattered throughout the kernel where we know for a fact the flag must not
 * be seen.
 */
#ifdef INVARIANTS
void
NDFREE_NOTHING(struct nameidata *ndp)
{
	struct componentname *cnp;

	cnp = &ndp->ni_cnd;
	KASSERT(cnp->cn_nameiop == LOOKUP, ("%s: got non-LOOKUP op %d\n",
	    __func__, cnp->cn_nameiop));
	KASSERT((cnp->cn_flags & (SAVENAME | HASBUF)) == 0,
	    ("%s: bad flags \%" PRIx64 "\n", __func__, cnp->cn_flags));
}
#endif

void
(NDFREE)(struct nameidata *ndp, const u_int flags)
{
	int unlock_dvp;
	int unlock_vp;

	unlock_dvp = 0;
	unlock_vp = 0;

	if (!(flags & NDF_NO_FREE_PNBUF)) {
		NDFREE_PNBUF(ndp);
	}
	if (!(flags & NDF_NO_VP_UNLOCK) &&
	    (ndp->ni_cnd.cn_flags & LOCKLEAF) && ndp->ni_vp)
		unlock_vp = 1;
	if (!(flags & NDF_NO_DVP_UNLOCK) &&
	    (ndp->ni_cnd.cn_flags & LOCKPARENT) &&
	    ndp->ni_dvp != ndp->ni_vp)
		unlock_dvp = 1;
	if (!(flags & NDF_NO_VP_RELE) && ndp->ni_vp) {
		if (unlock_vp) {
			vput(ndp->ni_vp);
			unlock_vp = 0;
		} else
			vrele(ndp->ni_vp);
		ndp->ni_vp = NULL;
	}
	if (unlock_vp)
		VOP_UNLOCK(ndp->ni_vp);
	if (!(flags & NDF_NO_DVP_RELE) &&
	    (ndp->ni_cnd.cn_flags & (LOCKPARENT|WANTPARENT))) {
		if (unlock_dvp) {
			vput(ndp->ni_dvp);
			unlock_dvp = 0;
		} else
			vrele(ndp->ni_dvp);
		ndp->ni_dvp = NULL;
	}
	if (unlock_dvp)
		VOP_UNLOCK(ndp->ni_dvp);
	if (!(flags & NDF_NO_STARTDIR_RELE) &&
	    (ndp->ni_cnd.cn_flags & SAVESTART)) {
		vrele(ndp->ni_startdir);
		ndp->ni_startdir = NULL;
	}
}

#ifdef INVARIANTS
/*
 * Validate the final state of ndp after the lookup.
 *
 * Historically filesystems were allowed to modify cn_flags. Most notably they
 * can add SAVENAME to the request, resulting in HASBUF and pushing subsequent
 * clean up to the consumer. In practice this seems to only concern != LOOKUP
 * operations.
 *
 * As a step towards stricter API contract this routine validates the state to
 * clean up. Note validation is a work in progress with the intent of becoming
 * stricter over time.
 */
#define NDMODIFYINGFLAGS (LOCKLEAF | LOCKPARENT | WANTPARENT | SAVENAME | SAVESTART | HASBUF)
void
NDVALIDATE(struct nameidata *ndp)
{
	struct componentname *cnp;
	u_int64_t used, orig;

	cnp = &ndp->ni_cnd;
	orig = cnp->cn_origflags;
	used = cnp->cn_flags;
	switch (cnp->cn_nameiop) {
	case LOOKUP:
		/*
		 * For plain lookup we require strict conformance -- nothing
		 * to clean up if it was not requested by the caller.
		 */
		orig &= NDMODIFYINGFLAGS;
		used &= NDMODIFYINGFLAGS;
		if ((orig & (SAVENAME | SAVESTART)) != 0)
			orig |= HASBUF;
		if (orig != used) {
			goto out_mismatch;
		}
		break;
	case CREATE:
	case DELETE:
	case RENAME:
		/*
		 * Some filesystems set SAVENAME to provoke HASBUF, accomodate
		 * for it until it gets fixed.
		 */
		orig &= NDMODIFYINGFLAGS;
		orig |= (SAVENAME | HASBUF);
		used &= NDMODIFYINGFLAGS;
		used |= (SAVENAME | HASBUF);
		if (orig != used) {
			goto out_mismatch;
		}
		break;
	}
	return;
out_mismatch:
	panic("%s: mismatched flags for op %d: added %" PRIx64 ", "
	    "removed %" PRIx64" (%" PRIx64" != %" PRIx64"; stored %" PRIx64" != %" PRIx64")",
	    __func__, cnp->cn_nameiop, used & ~orig, orig &~ used,
	    orig, used, cnp->cn_origflags, cnp->cn_flags);
}
#endif

/*
 * Determine if there is a suitable alternate filename under the specified
 * prefix for the specified path.  If the create flag is set, then the
 * alternate prefix will be used so long as the parent directory exists.
 * This is used by the various compatibility ABIs so that Linux binaries prefer
 * files under /compat/linux for example.  The chosen path (whether under
 * the prefix or under /) is returned in a kernel malloc'd buffer pointed
 * to by pathbuf.  The caller is responsible for free'ing the buffer from
 * the M_TEMP bucket if one is returned.
 */
int
kern_alternate_path(const char *prefix, const char *path, enum uio_seg pathseg,
    char **pathbuf, int create, int dirfd)
{
	struct nameidata nd, ndroot;
	char *ptr, *buf, *cp;
	size_t len, sz;
	int error;

	buf = (char *) malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
	*pathbuf = buf;

	/* Copy the prefix into the new pathname as a starting point. */
	len = strlcpy(buf, prefix, MAXPATHLEN);
	if (len >= MAXPATHLEN) {
		*pathbuf = NULL;
		free(buf, M_TEMP);
		return (EINVAL);
	}
	sz = MAXPATHLEN - len;
	ptr = buf + len;

	/* Append the filename to the prefix. */
	if (pathseg == UIO_SYSSPACE)
		error = copystr(path, ptr, sz, &len);
	else
		error = copyinstr(path, ptr, sz, &len);

	if (error) {
		*pathbuf = NULL;
		free(buf, M_TEMP);
		return (error);
	}

	/* Only use a prefix with absolute pathnames. */
	if (*ptr != '/') {
		error = EINVAL;
		goto keeporig;
	}

	if (dirfd != AT_FDCWD) {
		/*
		 * We want the original because the "prefix" is
		 * included in the already opened dirfd.
		 */
		bcopy(ptr, buf, len);
		return (0);
	}

	/*
	 * We know that there is a / somewhere in this pathname.
	 * Search backwards for it, to find the file's parent dir
	 * to see if it exists in the alternate tree. If it does,
	 * and we want to create a file (cflag is set). We don't
	 * need to worry about the root comparison in this case.
	 */

	if (create) {
		for (cp = &ptr[len] - 1; *cp != '/'; cp--);
		*cp = '\0';

		NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_SYSSPACE, buf);
		error = namei(&nd);
		*cp = '/';
		if (error != 0)
			goto keeporig;
	} else {
		NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_SYSSPACE, buf);

		error = namei(&nd);
		if (error != 0)
			goto keeporig;

		/*
		 * We now compare the vnode of the prefix to the one
		 * vnode asked. If they resolve to be the same, then we
		 * ignore the match so that the real root gets used.
		 * This avoids the problem of traversing "../.." to find the
		 * root directory and never finding it, because "/" resolves
		 * to the emulation root directory. This is expensive :-(
		 */
		NDINIT(&ndroot, LOOKUP, FOLLOW, UIO_SYSSPACE, prefix);

		/* We shouldn't ever get an error from this namei(). */
		error = namei(&ndroot);
		if (error == 0) {
			if (nd.ni_vp == ndroot.ni_vp)
				error = ENOENT;

			NDFREE(&ndroot, NDF_ONLY_PNBUF);
			vrele(ndroot.ni_vp);
		}
	}

	NDFREE(&nd, NDF_ONLY_PNBUF);
	vrele(nd.ni_vp);

keeporig:
	/* If there was an error, use the original path name. */
	if (error)
		bcopy(ptr, buf, len);
	return (error);
}