Training courses

Kernel and Embedded Linux

Bootlin training courses

Embedded Linux, kernel,
Yocto Project, Buildroot, real-time,
graphics, boot time, debugging...

Bootlin logo

Elixir Cross Referencer

   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
/*
 * CDDL HEADER START
 *
 * The contents of this file are subject to the terms of the
 * Common Development and Distribution License (the "License").
 * You may not use this file except in compliance with the License.
 *
 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
 * or http://www.opensolaris.org/os/licensing.
 * See the License for the specific language governing permissions
 * and limitations under the License.
 *
 * When distributing Covered Code, include this CDDL HEADER in each
 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
 * If applicable, add the following below this CDDL HEADER, with the
 * fields enclosed by brackets "[]" replaced with your own identifying
 * information: Portions Copyright [yyyy] [name of copyright owner]
 *
 * CDDL HEADER END
 */

/*
 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
 * Copyright (c) 2013, 2018 by Delphix. All rights reserved.
 * Copyright (c) 2016, 2017 Intel Corporation.
 * Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com>.
 */

/*
 * Functions to convert between a list of vdevs and an nvlist representing the
 * configuration.  Each entry in the list can be one of:
 *
 * 	Device vdevs
 * 		disk=(path=..., devid=...)
 * 		file=(path=...)
 *
 * 	Group vdevs
 * 		raidz[1|2]=(...)
 * 		mirror=(...)
 *
 * 	Hot spares
 *
 * While the underlying implementation supports it, group vdevs cannot contain
 * other group vdevs.  All userland verification of devices is contained within
 * this file.  If successful, the nvlist returned can be passed directly to the
 * kernel; we've done as much verification as possible in userland.
 *
 * Hot spares are a special case, and passed down as an array of disk vdevs, at
 * the same level as the root of the vdev tree.
 *
 * The only function exported by this file is 'make_root_vdev'.  The
 * function performs several passes:
 *
 * 	1. Construct the vdev specification.  Performs syntax validation and
 *         makes sure each device is valid.
 * 	2. Check for devices in use.  Using libblkid to make sure that no
 *         devices are also in use.  Some can be overridden using the 'force'
 *         flag, others cannot.
 * 	3. Check for replication errors if the 'force' flag is not specified.
 *         validates that the replication level is consistent across the
 *         entire pool.
 * 	4. Call libzfs to label any whole disks with an EFI label.
 */

#include <assert.h>
#include <ctype.h>
#include <errno.h>
#include <fcntl.h>
#include <libintl.h>
#include <libnvpair.h>
#include <libzutil.h>
#include <limits.h>
#include <sys/spa.h>
#include <stdio.h>
#include <string.h>
#include <unistd.h>
#include "zpool_util.h"
#include <sys/zfs_context.h>
#include <sys/stat.h>

/*
 * For any given vdev specification, we can have multiple errors.  The
 * vdev_error() function keeps track of whether we have seen an error yet, and
 * prints out a header if its the first error we've seen.
 */
boolean_t error_seen;
boolean_t is_force;




/*PRINTFLIKE1*/
void
vdev_error(const char *fmt, ...)
{
	va_list ap;

	if (!error_seen) {
		(void) fprintf(stderr, gettext("invalid vdev specification\n"));
		if (!is_force)
			(void) fprintf(stderr, gettext("use '-f' to override "
			    "the following errors:\n"));
		else
			(void) fprintf(stderr, gettext("the following errors "
			    "must be manually repaired:\n"));
		error_seen = B_TRUE;
	}

	va_start(ap, fmt);
	(void) vfprintf(stderr, fmt, ap);
	va_end(ap);
}

/*
 * Check that a file is valid.  All we can do in this case is check that it's
 * not in use by another pool, and not in use by swap.
 */
int
check_file(const char *file, boolean_t force, boolean_t isspare)
{
	char  *name;
	int fd;
	int ret = 0;
	pool_state_t state;
	boolean_t inuse;

	if ((fd = open(file, O_RDONLY)) < 0)
		return (0);

	if (zpool_in_use(g_zfs, fd, &state, &name, &inuse) == 0 && inuse) {
		const char *desc;

		switch (state) {
		case POOL_STATE_ACTIVE:
			desc = gettext("active");
			break;

		case POOL_STATE_EXPORTED:
			desc = gettext("exported");
			break;

		case POOL_STATE_POTENTIALLY_ACTIVE:
			desc = gettext("potentially active");
			break;

		default:
			desc = gettext("unknown");
			break;
		}

		/*
		 * Allow hot spares to be shared between pools.
		 */
		if (state == POOL_STATE_SPARE && isspare) {
			free(name);
			(void) close(fd);
			return (0);
		}

		if (state == POOL_STATE_ACTIVE ||
		    state == POOL_STATE_SPARE || !force) {
			switch (state) {
			case POOL_STATE_SPARE:
				vdev_error(gettext("%s is reserved as a hot "
				    "spare for pool %s\n"), file, name);
				break;
			default:
				vdev_error(gettext("%s is part of %s pool "
				    "'%s'\n"), file, desc, name);
				break;
			}
			ret = -1;
		}

		free(name);
	}

	(void) close(fd);
	return (ret);
}

/*
 * This may be a shorthand device path or it could be total gibberish.
 * Check to see if it is a known device available in zfs_vdev_paths.
 * As part of this check, see if we've been given an entire disk
 * (minus the slice number).
 */
static int
is_shorthand_path(const char *arg, char *path, size_t path_size,
    struct stat64 *statbuf, boolean_t *wholedisk)
{
	int error;

	error = zfs_resolve_shortname(arg, path, path_size);
	if (error == 0) {
		*wholedisk = zfs_dev_is_whole_disk(path);
		if (*wholedisk || (stat64(path, statbuf) == 0))
			return (0);
	}

	strlcpy(path, arg, path_size);
	memset(statbuf, 0, sizeof (*statbuf));
	*wholedisk = B_FALSE;

	return (error);
}

/*
 * Determine if the given path is a hot spare within the given configuration.
 * If no configuration is given we rely solely on the label.
 */
static boolean_t
is_spare(nvlist_t *config, const char *path)
{
	int fd;
	pool_state_t state;
	char *name = NULL;
	nvlist_t *label;
	uint64_t guid, spareguid;
	nvlist_t *nvroot;
	nvlist_t **spares;
	uint_t i, nspares;
	boolean_t inuse;

	if ((fd = open(path, O_RDONLY|O_DIRECT)) < 0)
		return (B_FALSE);

	if (zpool_in_use(g_zfs, fd, &state, &name, &inuse) != 0 ||
	    !inuse ||
	    state != POOL_STATE_SPARE ||
	    zpool_read_label(fd, &label, NULL) != 0) {
		free(name);
		(void) close(fd);
		return (B_FALSE);
	}
	free(name);
	(void) close(fd);

	if (config == NULL) {
		nvlist_free(label);
		return (B_TRUE);
	}

	verify(nvlist_lookup_uint64(label, ZPOOL_CONFIG_GUID, &guid) == 0);
	nvlist_free(label);

	verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
	    &nvroot) == 0);
	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
	    &spares, &nspares) == 0) {
		for (i = 0; i < nspares; i++) {
			verify(nvlist_lookup_uint64(spares[i],
			    ZPOOL_CONFIG_GUID, &spareguid) == 0);
			if (spareguid == guid)
				return (B_TRUE);
		}
	}

	return (B_FALSE);
}

/*
 * Create a leaf vdev.  Determine if this is a file or a device.  If it's a
 * device, fill in the device id to make a complete nvlist.  Valid forms for a
 * leaf vdev are:
 *
 *	/dev/xxx	Complete disk path
 *	/xxx		Full path to file
 *	xxx		Shorthand for <zfs_vdev_paths>/xxx
 */
static nvlist_t *
make_leaf_vdev(nvlist_t *props, const char *arg, uint64_t is_log)
{
	char path[MAXPATHLEN];
	struct stat64 statbuf;
	nvlist_t *vdev = NULL;
	char *type = NULL;
	boolean_t wholedisk = B_FALSE;
	uint64_t ashift = 0;
	int err;

	/*
	 * Determine what type of vdev this is, and put the full path into
	 * 'path'.  We detect whether this is a device of file afterwards by
	 * checking the st_mode of the file.
	 */
	if (arg[0] == '/') {
		/*
		 * Complete device or file path.  Exact type is determined by
		 * examining the file descriptor afterwards.  Symbolic links
		 * are resolved to their real paths to determine whole disk
		 * and S_ISBLK/S_ISREG type checks.  However, we are careful
		 * to store the given path as ZPOOL_CONFIG_PATH to ensure we
		 * can leverage udev's persistent device labels.
		 */
		if (realpath(arg, path) == NULL) {
			(void) fprintf(stderr,
			    gettext("cannot resolve path '%s'\n"), arg);
			return (NULL);
		}

		wholedisk = zfs_dev_is_whole_disk(path);
		if (!wholedisk && (stat64(path, &statbuf) != 0)) {
			(void) fprintf(stderr,
			    gettext("cannot open '%s': %s\n"),
			    path, strerror(errno));
			return (NULL);
		}

		/* After whole disk check restore original passed path */
		strlcpy(path, arg, sizeof (path));
	} else {
		err = is_shorthand_path(arg, path, sizeof (path),
		    &statbuf, &wholedisk);
		if (err != 0) {
			/*
			 * If we got ENOENT, then the user gave us
			 * gibberish, so try to direct them with a
			 * reasonable error message.  Otherwise,
			 * regurgitate strerror() since it's the best we
			 * can do.
			 */
			if (err == ENOENT) {
				(void) fprintf(stderr,
				    gettext("cannot open '%s': no such "
				    "device in %s\n"), arg, DISK_ROOT);
				(void) fprintf(stderr,
				    gettext("must be a full path or "
				    "shorthand device name\n"));
				return (NULL);
			} else {
				(void) fprintf(stderr,
				    gettext("cannot open '%s': %s\n"),
				    path, strerror(errno));
				return (NULL);
			}
		}
	}

	/*
	 * Determine whether this is a device or a file.
	 */
	if (wholedisk || S_ISBLK(statbuf.st_mode)) {
		type = VDEV_TYPE_DISK;
	} else if (S_ISREG(statbuf.st_mode)) {
		type = VDEV_TYPE_FILE;
	} else {
		(void) fprintf(stderr, gettext("cannot use '%s': must be a "
		    "block device or regular file\n"), path);
		return (NULL);
	}

	/*
	 * Finally, we have the complete device or file, and we know that it is
	 * acceptable to use.  Construct the nvlist to describe this vdev.  All
	 * vdevs have a 'path' element, and devices also have a 'devid' element.
	 */
	verify(nvlist_alloc(&vdev, NV_UNIQUE_NAME, 0) == 0);
	verify(nvlist_add_string(vdev, ZPOOL_CONFIG_PATH, path) == 0);
	verify(nvlist_add_string(vdev, ZPOOL_CONFIG_TYPE, type) == 0);
	verify(nvlist_add_uint64(vdev, ZPOOL_CONFIG_IS_LOG, is_log) == 0);
	if (is_log)
		verify(nvlist_add_string(vdev, ZPOOL_CONFIG_ALLOCATION_BIAS,
		    VDEV_ALLOC_BIAS_LOG) == 0);
	if (strcmp(type, VDEV_TYPE_DISK) == 0)
		verify(nvlist_add_uint64(vdev, ZPOOL_CONFIG_WHOLE_DISK,
		    (uint64_t)wholedisk) == 0);

	/*
	 * Override defaults if custom properties are provided.
	 */
	if (props != NULL) {
		char *value = NULL;

		if (nvlist_lookup_string(props,
		    zpool_prop_to_name(ZPOOL_PROP_ASHIFT), &value) == 0) {
			if (zfs_nicestrtonum(NULL, value, &ashift) != 0) {
				(void) fprintf(stderr,
				    gettext("ashift must be a number.\n"));
				return (NULL);
			}
			if (ashift != 0 &&
			    (ashift < ASHIFT_MIN || ashift > ASHIFT_MAX)) {
				(void) fprintf(stderr,
				    gettext("invalid 'ashift=%" PRIu64 "' "
				    "property: only values between %" PRId32 " "
				    "and %" PRId32 " are allowed.\n"),
				    ashift, ASHIFT_MIN, ASHIFT_MAX);
				return (NULL);
			}
		}
	}

	/*
	 * If the device is known to incorrectly report its physical sector
	 * size explicitly provide the known correct value.
	 */
	if (ashift == 0) {
		int sector_size;

		if (check_sector_size_database(path, &sector_size) == B_TRUE)
			ashift = highbit64(sector_size) - 1;
	}

	if (ashift > 0)
		(void) nvlist_add_uint64(vdev, ZPOOL_CONFIG_ASHIFT, ashift);

	return (vdev);
}

/*
 * Go through and verify the replication level of the pool is consistent.
 * Performs the following checks:
 *
 * 	For the new spec, verifies that devices in mirrors and raidz are the
 * 	same size.
 *
 * 	If the current configuration already has inconsistent replication
 * 	levels, ignore any other potential problems in the new spec.
 *
 * 	Otherwise, make sure that the current spec (if there is one) and the new
 * 	spec have consistent replication levels.
 *
 *	If there is no current spec (create), make sure new spec has at least
 *	one general purpose vdev.
 */
typedef struct replication_level {
	char *zprl_type;
	uint64_t zprl_children;
	uint64_t zprl_parity;
} replication_level_t;

#define	ZPOOL_FUZZ	(16 * 1024 * 1024)

static boolean_t
is_raidz_mirror(replication_level_t *a, replication_level_t *b,
    replication_level_t **raidz, replication_level_t **mirror)
{
	if (strcmp(a->zprl_type, "raidz") == 0 &&
	    strcmp(b->zprl_type, "mirror") == 0) {
		*raidz = a;
		*mirror = b;
		return (B_TRUE);
	}
	return (B_FALSE);
}

/*
 * Given a list of toplevel vdevs, return the current replication level.  If
 * the config is inconsistent, then NULL is returned.  If 'fatal' is set, then
 * an error message will be displayed for each self-inconsistent vdev.
 */
static replication_level_t *
get_replication(nvlist_t *nvroot, boolean_t fatal)
{
	nvlist_t **top;
	uint_t t, toplevels;
	nvlist_t **child;
	uint_t c, children;
	nvlist_t *nv;
	char *type;
	replication_level_t lastrep = {0};
	replication_level_t rep;
	replication_level_t *ret;
	replication_level_t *raidz, *mirror;
	boolean_t dontreport;

	ret = safe_malloc(sizeof (replication_level_t));

	verify(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
	    &top, &toplevels) == 0);

	for (t = 0; t < toplevels; t++) {
		uint64_t is_log = B_FALSE;

		nv = top[t];

		/*
		 * For separate logs we ignore the top level vdev replication
		 * constraints.
		 */
		(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG, &is_log);
		if (is_log)
			continue;

		/* Ignore holes introduced by removing aux devices */
		verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0);
		if (strcmp(type, VDEV_TYPE_HOLE) == 0)
			continue;

		if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
		    &child, &children) != 0) {
			/*
			 * This is a 'file' or 'disk' vdev.
			 */
			rep.zprl_type = type;
			rep.zprl_children = 1;
			rep.zprl_parity = 0;
		} else {
			int64_t vdev_size;

			/*
			 * This is a mirror or RAID-Z vdev.  Go through and make
			 * sure the contents are all the same (files vs. disks),
			 * keeping track of the number of elements in the
			 * process.
			 *
			 * We also check that the size of each vdev (if it can
			 * be determined) is the same.
			 */
			rep.zprl_type = type;
			rep.zprl_children = 0;

			if (strcmp(type, VDEV_TYPE_RAIDZ) == 0) {
				verify(nvlist_lookup_uint64(nv,
				    ZPOOL_CONFIG_NPARITY,
				    &rep.zprl_parity) == 0);
				assert(rep.zprl_parity != 0);
			} else {
				rep.zprl_parity = 0;
			}

			/*
			 * The 'dontreport' variable indicates that we've
			 * already reported an error for this spec, so don't
			 * bother doing it again.
			 */
			type = NULL;
			dontreport = 0;
			vdev_size = -1LL;
			for (c = 0; c < children; c++) {
				nvlist_t *cnv = child[c];
				char *path;
				struct stat64 statbuf;
				int64_t size = -1LL;
				char *childtype;
				int fd, err;

				rep.zprl_children++;

				verify(nvlist_lookup_string(cnv,
				    ZPOOL_CONFIG_TYPE, &childtype) == 0);

				/*
				 * If this is a replacing or spare vdev, then
				 * get the real first child of the vdev: do this
				 * in a loop because replacing and spare vdevs
				 * can be nested.
				 */
				while (strcmp(childtype,
				    VDEV_TYPE_REPLACING) == 0 ||
				    strcmp(childtype, VDEV_TYPE_SPARE) == 0) {
					nvlist_t **rchild;
					uint_t rchildren;

					verify(nvlist_lookup_nvlist_array(cnv,
					    ZPOOL_CONFIG_CHILDREN, &rchild,
					    &rchildren) == 0);
					assert(rchildren == 2);
					cnv = rchild[0];

					verify(nvlist_lookup_string(cnv,
					    ZPOOL_CONFIG_TYPE,
					    &childtype) == 0);
				}

				verify(nvlist_lookup_string(cnv,
				    ZPOOL_CONFIG_PATH, &path) == 0);

				/*
				 * If we have a raidz/mirror that combines disks
				 * with files, report it as an error.
				 */
				if (!dontreport && type != NULL &&
				    strcmp(type, childtype) != 0) {
					if (ret != NULL)
						free(ret);
					ret = NULL;
					if (fatal)
						vdev_error(gettext(
						    "mismatched replication "
						    "level: %s contains both "
						    "files and devices\n"),
						    rep.zprl_type);
					else
						return (NULL);
					dontreport = B_TRUE;
				}

				/*
				 * According to stat(2), the value of 'st_size'
				 * is undefined for block devices and character
				 * devices.  But there is no effective way to
				 * determine the real size in userland.
				 *
				 * Instead, we'll take advantage of an
				 * implementation detail of spec_size().  If the
				 * device is currently open, then we (should)
				 * return a valid size.
				 *
				 * If we still don't get a valid size (indicated
				 * by a size of 0 or MAXOFFSET_T), then ignore
				 * this device altogether.
				 */
				if ((fd = open(path, O_RDONLY)) >= 0) {
					err = fstat64_blk(fd, &statbuf);
					(void) close(fd);
				} else {
					err = stat64(path, &statbuf);
				}

				if (err != 0 ||
				    statbuf.st_size == 0 ||
				    statbuf.st_size == MAXOFFSET_T)
					continue;

				size = statbuf.st_size;

				/*
				 * Also make sure that devices and
				 * slices have a consistent size.  If
				 * they differ by a significant amount
				 * (~16MB) then report an error.
				 */
				if (!dontreport &&
				    (vdev_size != -1LL &&
				    (llabs(size - vdev_size) >
				    ZPOOL_FUZZ))) {
					if (ret != NULL)
						free(ret);
					ret = NULL;
					if (fatal)
						vdev_error(gettext(
						    "%s contains devices of "
						    "different sizes\n"),
						    rep.zprl_type);
					else
						return (NULL);
					dontreport = B_TRUE;
				}

				type = childtype;
				vdev_size = size;
			}
		}

		/*
		 * At this point, we have the replication of the last toplevel
		 * vdev in 'rep'.  Compare it to 'lastrep' to see if it is
		 * different.
		 */
		if (lastrep.zprl_type != NULL) {
			if (is_raidz_mirror(&lastrep, &rep, &raidz, &mirror) ||
			    is_raidz_mirror(&rep, &lastrep, &raidz, &mirror)) {
				/*
				 * Accepted raidz and mirror when they can
				 * handle the same number of disk failures.
				 */
				if (raidz->zprl_parity !=
				    mirror->zprl_children - 1) {
					if (ret != NULL)
						free(ret);
					ret = NULL;
					if (fatal)
						vdev_error(gettext(
						    "mismatched replication "
						    "level: "
						    "%s and %s vdevs with "
						    "different redundancy, "
						    "%llu vs. %llu (%llu-way) "
						    "are present\n"),
						    raidz->zprl_type,
						    mirror->zprl_type,
						    raidz->zprl_parity,
						    mirror->zprl_children - 1,
						    mirror->zprl_children);
					else
						return (NULL);
				}
			} else if (strcmp(lastrep.zprl_type, rep.zprl_type) !=
			    0) {
				if (ret != NULL)
					free(ret);
				ret = NULL;
				if (fatal)
					vdev_error(gettext(
					    "mismatched replication level: "
					    "both %s and %s vdevs are "
					    "present\n"),
					    lastrep.zprl_type, rep.zprl_type);
				else
					return (NULL);
			} else if (lastrep.zprl_parity != rep.zprl_parity) {
				if (ret)
					free(ret);
				ret = NULL;
				if (fatal)
					vdev_error(gettext(
					    "mismatched replication level: "
					    "both %llu and %llu device parity "
					    "%s vdevs are present\n"),
					    lastrep.zprl_parity,
					    rep.zprl_parity,
					    rep.zprl_type);
				else
					return (NULL);
			} else if (lastrep.zprl_children != rep.zprl_children) {
				if (ret)
					free(ret);
				ret = NULL;
				if (fatal)
					vdev_error(gettext(
					    "mismatched replication level: "
					    "both %llu-way and %llu-way %s "
					    "vdevs are present\n"),
					    lastrep.zprl_children,
					    rep.zprl_children,
					    rep.zprl_type);
				else
					return (NULL);
			}
		}
		lastrep = rep;
	}

	if (ret != NULL)
		*ret = rep;

	return (ret);
}

/*
 * Check the replication level of the vdev spec against the current pool.  Calls
 * get_replication() to make sure the new spec is self-consistent.  If the pool
 * has a consistent replication level, then we ignore any errors.  Otherwise,
 * report any difference between the two.
 */
static int
check_replication(nvlist_t *config, nvlist_t *newroot)
{
	nvlist_t **child;
	uint_t	children;
	replication_level_t *current = NULL, *new;
	replication_level_t *raidz, *mirror;
	int ret;

	/*
	 * If we have a current pool configuration, check to see if it's
	 * self-consistent.  If not, simply return success.
	 */
	if (config != NULL) {
		nvlist_t *nvroot;

		verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
		    &nvroot) == 0);
		if ((current = get_replication(nvroot, B_FALSE)) == NULL)
			return (0);
	}
	/*
	 * for spares there may be no children, and therefore no
	 * replication level to check
	 */
	if ((nvlist_lookup_nvlist_array(newroot, ZPOOL_CONFIG_CHILDREN,
	    &child, &children) != 0) || (children == 0)) {
		free(current);
		return (0);
	}

	/*
	 * If all we have is logs then there's no replication level to check.
	 */
	if (num_logs(newroot) == children) {
		free(current);
		return (0);
	}

	/*
	 * Get the replication level of the new vdev spec, reporting any
	 * inconsistencies found.
	 */
	if ((new = get_replication(newroot, B_TRUE)) == NULL) {
		free(current);
		return (-1);
	}

	/*
	 * Check to see if the new vdev spec matches the replication level of
	 * the current pool.
	 */
	ret = 0;
	if (current != NULL) {
		if (is_raidz_mirror(current, new, &raidz, &mirror) ||
		    is_raidz_mirror(new, current, &raidz, &mirror)) {
			if (raidz->zprl_parity != mirror->zprl_children - 1) {
				vdev_error(gettext(
				    "mismatched replication level: pool and "
				    "new vdev with different redundancy, %s "
				    "and %s vdevs, %llu vs. %llu (%llu-way)\n"),
				    raidz->zprl_type,
				    mirror->zprl_type,
				    raidz->zprl_parity,
				    mirror->zprl_children - 1,
				    mirror->zprl_children);
				ret = -1;
			}
		} else if (strcmp(current->zprl_type, new->zprl_type) != 0) {
			vdev_error(gettext(
			    "mismatched replication level: pool uses %s "
			    "and new vdev is %s\n"),
			    current->zprl_type, new->zprl_type);
			ret = -1;
		} else if (current->zprl_parity != new->zprl_parity) {
			vdev_error(gettext(
			    "mismatched replication level: pool uses %llu "
			    "device parity and new vdev uses %llu\n"),
			    current->zprl_parity, new->zprl_parity);
			ret = -1;
		} else if (current->zprl_children != new->zprl_children) {
			vdev_error(gettext(
			    "mismatched replication level: pool uses %llu-way "
			    "%s and new vdev uses %llu-way %s\n"),
			    current->zprl_children, current->zprl_type,
			    new->zprl_children, new->zprl_type);
			ret = -1;
		}
	}

	free(new);
	if (current != NULL)
		free(current);

	return (ret);
}

static int
zero_label(char *path)
{
	const int size = 4096;
	char buf[size];
	int err, fd;

	if ((fd = open(path, O_WRONLY|O_EXCL)) < 0) {
		(void) fprintf(stderr, gettext("cannot open '%s': %s\n"),
		    path, strerror(errno));
		return (-1);
	}

	memset(buf, 0, size);
	err = write(fd, buf, size);
	(void) fdatasync(fd);
	(void) close(fd);

	if (err == -1) {
		(void) fprintf(stderr, gettext("cannot zero first %d bytes "
		    "of '%s': %s\n"), size, path, strerror(errno));
		return (-1);
	}

	if (err != size) {
		(void) fprintf(stderr, gettext("could only zero %d/%d bytes "
		    "of '%s'\n"), err, size, path);
		return (-1);
	}

	return (0);
}

/*
 * Go through and find any whole disks in the vdev specification, labelling them
 * as appropriate.  When constructing the vdev spec, we were unable to open this
 * device in order to provide a devid.  Now that we have labelled the disk and
 * know that slice 0 is valid, we can construct the devid now.
 *
 * If the disk was already labeled with an EFI label, we will have gotten the
 * devid already (because we were able to open the whole disk).  Otherwise, we
 * need to get the devid after we label the disk.
 */
static int
make_disks(zpool_handle_t *zhp, nvlist_t *nv)
{
	nvlist_t **child;
	uint_t c, children;
	char *type, *path;
	char devpath[MAXPATHLEN];
	char udevpath[MAXPATHLEN];
	uint64_t wholedisk;
	struct stat64 statbuf;
	int is_exclusive = 0;
	int fd;
	int ret;

	verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0);

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
	    &child, &children) != 0) {

		if (strcmp(type, VDEV_TYPE_DISK) != 0)
			return (0);

		/*
		 * We have a disk device.  If this is a whole disk write
		 * out the efi partition table, otherwise write zero's to
		 * the first 4k of the partition.  This is to ensure that
		 * libblkid will not misidentify the partition due to a
		 * magic value left by the previous filesystem.
		 */
		verify(!nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path));
		verify(!nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
		    &wholedisk));

		if (!wholedisk) {
			/*
			 * Update device id string for mpath nodes (Linux only)
			 */
			if (is_mpath_whole_disk(path))
				update_vdev_config_dev_strs(nv);

			if (!is_spare(NULL, path))
				(void) zero_label(path);
			return (0);
		}

		if (realpath(path, devpath) == NULL) {
			ret = errno;
			(void) fprintf(stderr,
			    gettext("cannot resolve path '%s'\n"), path);
			return (ret);
		}

		/*
		 * Remove any previously existing symlink from a udev path to
		 * the device before labeling the disk.  This ensures that
		 * only newly created links are used.  Otherwise there is a
		 * window between when udev deletes and recreates the link
		 * during which access attempts will fail with ENOENT.
		 */
		strlcpy(udevpath, path, MAXPATHLEN);
		(void) zfs_append_partition(udevpath, MAXPATHLEN);

		fd = open(devpath, O_RDWR|O_EXCL);
		if (fd == -1) {
			if (errno == EBUSY)
				is_exclusive = 1;
#ifdef __FreeBSD__
			if (errno == EPERM)
				is_exclusive = 1;
#endif
		} else {
			(void) close(fd);
		}

		/*
		 * If the partition exists, contains a valid spare label,
		 * and is opened exclusively there is no need to partition
		 * it.  Hot spares have already been partitioned and are
		 * held open exclusively by the kernel as a safety measure.
		 *
		 * If the provided path is for a /dev/disk/ device its
		 * symbolic link will be removed, partition table created,
		 * and then block until udev creates the new link.
		 */
		if (!is_exclusive && !is_spare(NULL, udevpath)) {
			char *devnode = strrchr(devpath, '/') + 1;

			ret = strncmp(udevpath, UDISK_ROOT, strlen(UDISK_ROOT));
			if (ret == 0) {
				ret = lstat64(udevpath, &statbuf);
				if (ret == 0 && S_ISLNK(statbuf.st_mode))
					(void) unlink(udevpath);
			}

			/*
			 * When labeling a pool the raw device node name
			 * is provided as it appears under /dev/.
			 */
			if (zpool_label_disk(g_zfs, zhp, devnode) == -1)
				return (-1);

			/*
			 * Wait for udev to signal the device is available
			 * by the provided path.
			 */
			ret = zpool_label_disk_wait(udevpath, DISK_LABEL_WAIT);
			if (ret) {
				(void) fprintf(stderr,
				    gettext("missing link: %s was "
				    "partitioned but %s is missing\n"),
				    devnode, udevpath);
				return (ret);
			}

			ret = zero_label(udevpath);
			if (ret)
				return (ret);
		}

		/*
		 * Update the path to refer to the partition.  The presence of
		 * the 'whole_disk' field indicates to the CLI that we should
		 * chop off the partition number when displaying the device in
		 * future output.
		 */
		verify(nvlist_add_string(nv, ZPOOL_CONFIG_PATH, udevpath) == 0);

		/*
		 * Update device id strings for whole disks (Linux only)
		 */
		update_vdev_config_dev_strs(nv);

		return (0);
	}

	for (c = 0; c < children; c++)
		if ((ret = make_disks(zhp, child[c])) != 0)
			return (ret);

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
	    &child, &children) == 0)
		for (c = 0; c < children; c++)
			if ((ret = make_disks(zhp, child[c])) != 0)
				return (ret);

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
	    &child, &children) == 0)
		for (c = 0; c < children; c++)
			if ((ret = make_disks(zhp, child[c])) != 0)
				return (ret);

	return (0);
}

/*
 * Go through and find any devices that are in use.  We rely on libdiskmgt for
 * the majority of this task.
 */
static boolean_t
is_device_in_use(nvlist_t *config, nvlist_t *nv, boolean_t force,
    boolean_t replacing, boolean_t isspare)
{
	nvlist_t **child;
	uint_t c, children;
	char *type, *path;
	int ret = 0;
	char buf[MAXPATHLEN];
	uint64_t wholedisk = B_FALSE;
	boolean_t anyinuse = B_FALSE;

	verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0);

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
	    &child, &children) != 0) {

		verify(!nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path));
		if (strcmp(type, VDEV_TYPE_DISK) == 0)
			verify(!nvlist_lookup_uint64(nv,
			    ZPOOL_CONFIG_WHOLE_DISK, &wholedisk));

		/*
		 * As a generic check, we look to see if this is a replace of a
		 * hot spare within the same pool.  If so, we allow it
		 * regardless of what libblkid or zpool_in_use() says.
		 */
		if (replacing) {
			(void) strlcpy(buf, path, sizeof (buf));
			if (wholedisk) {
				ret = zfs_append_partition(buf,  sizeof (buf));
				if (ret == -1)
					return (-1);
			}

			if (is_spare(config, buf))
				return (B_FALSE);
		}

		if (strcmp(type, VDEV_TYPE_DISK) == 0)
			ret = check_device(path, force, isspare, wholedisk);

		else if (strcmp(type, VDEV_TYPE_FILE) == 0)
			ret = check_file(path, force, isspare);

		return (ret != 0);
	}

	for (c = 0; c < children; c++)
		if (is_device_in_use(config, child[c], force, replacing,
		    B_FALSE))
			anyinuse = B_TRUE;

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
	    &child, &children) == 0)
		for (c = 0; c < children; c++)
			if (is_device_in_use(config, child[c], force, replacing,
			    B_TRUE))
				anyinuse = B_TRUE;

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
	    &child, &children) == 0)
		for (c = 0; c < children; c++)
			if (is_device_in_use(config, child[c], force, replacing,
			    B_FALSE))
				anyinuse = B_TRUE;

	return (anyinuse);
}

static const char *
is_grouping(const char *type, int *mindev, int *maxdev)
{
	if (strncmp(type, "raidz", 5) == 0) {
		const char *p = type + 5;
		char *end;
		long nparity;

		if (*p == '\0') {
			nparity = 1;
		} else if (*p == '0') {
			return (NULL); /* no zero prefixes allowed */
		} else {
			errno = 0;
			nparity = strtol(p, &end, 10);
			if (errno != 0 || nparity < 1 || nparity >= 255 ||
			    *end != '\0')
				return (NULL);
		}

		if (mindev != NULL)
			*mindev = nparity + 1;
		if (maxdev != NULL)
			*maxdev = 255;
		return (VDEV_TYPE_RAIDZ);
	}

	if (maxdev != NULL)
		*maxdev = INT_MAX;

	if (strcmp(type, "mirror") == 0) {
		if (mindev != NULL)
			*mindev = 2;
		return (VDEV_TYPE_MIRROR);
	}

	if (strcmp(type, "spare") == 0) {
		if (mindev != NULL)
			*mindev = 1;
		return (VDEV_TYPE_SPARE);
	}

	if (strcmp(type, "log") == 0) {
		if (mindev != NULL)
			*mindev = 1;
		return (VDEV_TYPE_LOG);
	}

	if (strcmp(type, VDEV_ALLOC_BIAS_SPECIAL) == 0 ||
	    strcmp(type, VDEV_ALLOC_BIAS_DEDUP) == 0) {
		if (mindev != NULL)
			*mindev = 1;
		return (type);
	}

	if (strcmp(type, "cache") == 0) {
		if (mindev != NULL)
			*mindev = 1;
		return (VDEV_TYPE_L2CACHE);
	}

	return (NULL);
}

/*
 * Construct a syntactically valid vdev specification,
 * and ensure that all devices and files exist and can be opened.
 * Note: we don't bother freeing anything in the error paths
 * because the program is just going to exit anyway.
 */
static nvlist_t *
construct_spec(nvlist_t *props, int argc, char **argv)
{
	nvlist_t *nvroot, *nv, **top, **spares, **l2cache;
	int t, toplevels, mindev, maxdev, nspares, nlogs, nl2cache;
	const char *type;
	uint64_t is_log, is_special, is_dedup;
	boolean_t seen_logs;

	top = NULL;
	toplevels = 0;
	spares = NULL;
	l2cache = NULL;
	nspares = 0;
	nlogs = 0;
	nl2cache = 0;
	is_log = is_special = is_dedup = B_FALSE;
	seen_logs = B_FALSE;
	nvroot = NULL;

	while (argc > 0) {
		nv = NULL;

		/*
		 * If it's a mirror or raidz, the subsequent arguments are
		 * its leaves -- until we encounter the next mirror or raidz.
		 */
		if ((type = is_grouping(argv[0], &mindev, &maxdev)) != NULL) {
			nvlist_t **child = NULL;
			int c, children = 0;

			if (strcmp(type, VDEV_TYPE_SPARE) == 0) {
				if (spares != NULL) {
					(void) fprintf(stderr,
					    gettext("invalid vdev "
					    "specification: 'spare' can be "
					    "specified only once\n"));
					goto spec_out;
				}
				is_log = is_special = is_dedup = B_FALSE;
			}

			if (strcmp(type, VDEV_TYPE_LOG) == 0) {
				if (seen_logs) {
					(void) fprintf(stderr,
					    gettext("invalid vdev "
					    "specification: 'log' can be "
					    "specified only once\n"));
					goto spec_out;
				}
				seen_logs = B_TRUE;
				is_log = B_TRUE;
				is_special = B_FALSE;
				is_dedup = B_FALSE;
				argc--;
				argv++;
				/*
				 * A log is not a real grouping device.
				 * We just set is_log and continue.
				 */
				continue;
			}

			if (strcmp(type, VDEV_ALLOC_BIAS_SPECIAL) == 0) {
				is_special = B_TRUE;
				is_log = B_FALSE;
				is_dedup = B_FALSE;
				argc--;
				argv++;
				continue;
			}

			if (strcmp(type, VDEV_ALLOC_BIAS_DEDUP) == 0) {
				is_dedup = B_TRUE;
				is_log = B_FALSE;
				is_special = B_FALSE;
				argc--;
				argv++;
				continue;
			}

			if (strcmp(type, VDEV_TYPE_L2CACHE) == 0) {
				if (l2cache != NULL) {
					(void) fprintf(stderr,
					    gettext("invalid vdev "
					    "specification: 'cache' can be "
					    "specified only once\n"));
					goto spec_out;
				}
				is_log = is_special = is_dedup = B_FALSE;
			}

			if (is_log || is_special || is_dedup) {
				if (strcmp(type, VDEV_TYPE_MIRROR) != 0) {
					(void) fprintf(stderr,
					    gettext("invalid vdev "
					    "specification: unsupported '%s' "
					    "device: %s\n"), is_log ? "log" :
					    "special", type);
					goto spec_out;
				}
				nlogs++;
			}

			for (c = 1; c < argc; c++) {
				if (is_grouping(argv[c], NULL, NULL) != NULL)
					break;
				children++;
				child = realloc(child,
				    children * sizeof (nvlist_t *));
				if (child == NULL)
					zpool_no_memory();
				if ((nv = make_leaf_vdev(props, argv[c],
				    B_FALSE)) == NULL) {
					for (c = 0; c < children - 1; c++)
						nvlist_free(child[c]);
					free(child);
					goto spec_out;
				}

				child[children - 1] = nv;
			}

			if (children < mindev) {
				(void) fprintf(stderr, gettext("invalid vdev "
				    "specification: %s requires at least %d "
				    "devices\n"), argv[0], mindev);
				for (c = 0; c < children; c++)
					nvlist_free(child[c]);
				free(child);
				goto spec_out;
			}

			if (children > maxdev) {
				(void) fprintf(stderr, gettext("invalid vdev "
				    "specification: %s supports no more than "
				    "%d devices\n"), argv[0], maxdev);
				for (c = 0; c < children; c++)
					nvlist_free(child[c]);
				free(child);
				goto spec_out;
			}

			argc -= c;
			argv += c;

			if (strcmp(type, VDEV_TYPE_SPARE) == 0) {
				spares = child;
				nspares = children;
				continue;
			} else if (strcmp(type, VDEV_TYPE_L2CACHE) == 0) {
				l2cache = child;
				nl2cache = children;
				continue;
			} else {
				/* create a top-level vdev with children */
				verify(nvlist_alloc(&nv, NV_UNIQUE_NAME,
				    0) == 0);
				verify(nvlist_add_string(nv, ZPOOL_CONFIG_TYPE,
				    type) == 0);
				verify(nvlist_add_uint64(nv,
				    ZPOOL_CONFIG_IS_LOG, is_log) == 0);
				if (is_log)
					verify(nvlist_add_string(nv,
					    ZPOOL_CONFIG_ALLOCATION_BIAS,
					    VDEV_ALLOC_BIAS_LOG) == 0);
				if (is_special) {
					verify(nvlist_add_string(nv,
					    ZPOOL_CONFIG_ALLOCATION_BIAS,
					    VDEV_ALLOC_BIAS_SPECIAL) == 0);
				}
				if (is_dedup) {
					verify(nvlist_add_string(nv,
					    ZPOOL_CONFIG_ALLOCATION_BIAS,
					    VDEV_ALLOC_BIAS_DEDUP) == 0);
				}
				if (strcmp(type, VDEV_TYPE_RAIDZ) == 0) {
					verify(nvlist_add_uint64(nv,
					    ZPOOL_CONFIG_NPARITY,
					    mindev - 1) == 0);
				}
				verify(nvlist_add_nvlist_array(nv,
				    ZPOOL_CONFIG_CHILDREN, child,
				    children) == 0);

				for (c = 0; c < children; c++)
					nvlist_free(child[c]);
				free(child);
			}
		} else {
			/*
			 * We have a device.  Pass off to make_leaf_vdev() to
			 * construct the appropriate nvlist describing the vdev.
			 */
			if ((nv = make_leaf_vdev(props, argv[0],
			    is_log)) == NULL)
				goto spec_out;

			if (is_log)
				nlogs++;
			if (is_special) {
				verify(nvlist_add_string(nv,
				    ZPOOL_CONFIG_ALLOCATION_BIAS,
				    VDEV_ALLOC_BIAS_SPECIAL) == 0);
			}
			if (is_dedup) {
				verify(nvlist_add_string(nv,
				    ZPOOL_CONFIG_ALLOCATION_BIAS,
				    VDEV_ALLOC_BIAS_DEDUP) == 0);
			}
			argc--;
			argv++;
		}

		toplevels++;
		top = realloc(top, toplevels * sizeof (nvlist_t *));
		if (top == NULL)
			zpool_no_memory();
		top[toplevels - 1] = nv;
	}

	if (toplevels == 0 && nspares == 0 && nl2cache == 0) {
		(void) fprintf(stderr, gettext("invalid vdev "
		    "specification: at least one toplevel vdev must be "
		    "specified\n"));
		goto spec_out;
	}

	if (seen_logs && nlogs == 0) {
		(void) fprintf(stderr, gettext("invalid vdev specification: "
		    "log requires at least 1 device\n"));
		goto spec_out;
	}

	/*
	 * Finally, create nvroot and add all top-level vdevs to it.
	 */
	verify(nvlist_alloc(&nvroot, NV_UNIQUE_NAME, 0) == 0);
	verify(nvlist_add_string(nvroot, ZPOOL_CONFIG_TYPE,
	    VDEV_TYPE_ROOT) == 0);
	verify(nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
	    top, toplevels) == 0);
	if (nspares != 0)
		verify(nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
		    spares, nspares) == 0);
	if (nl2cache != 0)
		verify(nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
		    l2cache, nl2cache) == 0);

spec_out:
	for (t = 0; t < toplevels; t++)
		nvlist_free(top[t]);
	for (t = 0; t < nspares; t++)
		nvlist_free(spares[t]);
	for (t = 0; t < nl2cache; t++)
		nvlist_free(l2cache[t]);

	free(spares);
	free(l2cache);
	free(top);

	return (nvroot);
}

nvlist_t *
split_mirror_vdev(zpool_handle_t *zhp, char *newname, nvlist_t *props,
    splitflags_t flags, int argc, char **argv)
{
	nvlist_t *newroot = NULL, **child;
	uint_t c, children;

	if (argc > 0) {
		if ((newroot = construct_spec(props, argc, argv)) == NULL) {
			(void) fprintf(stderr, gettext("Unable to build a "
			    "pool from the specified devices\n"));
			return (NULL);
		}

		if (!flags.dryrun && make_disks(zhp, newroot) != 0) {
			nvlist_free(newroot);
			return (NULL);
		}

		/* avoid any tricks in the spec */
		verify(nvlist_lookup_nvlist_array(newroot,
		    ZPOOL_CONFIG_CHILDREN, &child, &children) == 0);
		for (c = 0; c < children; c++) {
			char *path;
			const char *type;
			int min, max;

			verify(nvlist_lookup_string(child[c],
			    ZPOOL_CONFIG_PATH, &path) == 0);
			if ((type = is_grouping(path, &min, &max)) != NULL) {
				(void) fprintf(stderr, gettext("Cannot use "
				    "'%s' as a device for splitting\n"), type);
				nvlist_free(newroot);
				return (NULL);
			}
		}
	}

	if (zpool_vdev_split(zhp, newname, &newroot, props, flags) != 0) {
		nvlist_free(newroot);
		return (NULL);
	}

	return (newroot);
}

static int
num_normal_vdevs(nvlist_t *nvroot)
{
	nvlist_t **top;
	uint_t t, toplevels, normal = 0;

	verify(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
	    &top, &toplevels) == 0);

	for (t = 0; t < toplevels; t++) {
		uint64_t log = B_FALSE;

		(void) nvlist_lookup_uint64(top[t], ZPOOL_CONFIG_IS_LOG, &log);
		if (log)
			continue;
		if (nvlist_exists(top[t], ZPOOL_CONFIG_ALLOCATION_BIAS))
			continue;

		normal++;
	}

	return (normal);
}

/*
 * Get and validate the contents of the given vdev specification.  This ensures
 * that the nvlist returned is well-formed, that all the devices exist, and that
 * they are not currently in use by any other known consumer.  The 'poolconfig'
 * parameter is the current configuration of the pool when adding devices
 * existing pool, and is used to perform additional checks, such as changing the
 * replication level of the pool.  It can be 'NULL' to indicate that this is a
 * new pool.  The 'force' flag controls whether devices should be forcefully
 * added, even if they appear in use.
 */
nvlist_t *
make_root_vdev(zpool_handle_t *zhp, nvlist_t *props, int force, int check_rep,
    boolean_t replacing, boolean_t dryrun, int argc, char **argv)
{
	nvlist_t *newroot;
	nvlist_t *poolconfig = NULL;
	is_force = force;

	/*
	 * Construct the vdev specification.  If this is successful, we know
	 * that we have a valid specification, and that all devices can be
	 * opened.
	 */
	if ((newroot = construct_spec(props, argc, argv)) == NULL)
		return (NULL);

	if (zhp && ((poolconfig = zpool_get_config(zhp, NULL)) == NULL)) {
		nvlist_free(newroot);
		return (NULL);
	}

	/*
	 * Validate each device to make sure that it's not shared with another
	 * subsystem.  We do this even if 'force' is set, because there are some
	 * uses (such as a dedicated dump device) that even '-f' cannot
	 * override.
	 */
	if (is_device_in_use(poolconfig, newroot, force, replacing, B_FALSE)) {
		nvlist_free(newroot);
		return (NULL);
	}

	/*
	 * Check the replication level of the given vdevs and report any errors
	 * found.  We include the existing pool spec, if any, as we need to
	 * catch changes against the existing replication level.
	 */
	if (check_rep && check_replication(poolconfig, newroot) != 0) {
		nvlist_free(newroot);
		return (NULL);
	}

	/*
	 * On pool create the new vdev spec must have one normal vdev.
	 */
	if (poolconfig == NULL && num_normal_vdevs(newroot) == 0) {
		vdev_error(gettext("at least one general top-level vdev must "
		    "be specified\n"));
		nvlist_free(newroot);
		return (NULL);
	}

	/*
	 * Run through the vdev specification and label any whole disks found.
	 */
	if (!dryrun && make_disks(zhp, newroot) != 0) {
		nvlist_free(newroot);
		return (NULL);
	}

	return (newroot);
}