1 /* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 /* 22 * Copyright 2007 Sun Microsystems, Inc. All rights reserved. 23 * Use is subject to license terms. 24 */ 25 26 #pragma ident "%Z%%M% %I% %E% SMI" 27 28 #include <sys/types.h> 29 #include <sys/param.h> 30 #include <sys/errno.h> 31 #include <sys/uio.h> 32 #include <sys/buf.h> 33 #include <sys/modctl.h> 34 #include <sys/open.h> 35 #include <sys/file.h> 36 #include <sys/kmem.h> 37 #include <sys/conf.h> 38 #include <sys/cmn_err.h> 39 #include <sys/stat.h> 40 #include <sys/zfs_ioctl.h> 41 #include <sys/zfs_i18n.h> 42 #include <sys/zfs_znode.h> 43 #include <sys/zap.h> 44 #include <sys/spa.h> 45 #include <sys/spa_impl.h> 46 #include <sys/vdev.h> 47 #include <sys/vdev_impl.h> 48 #include <sys/dmu.h> 49 #include <sys/dsl_dir.h> 50 #include <sys/dsl_dataset.h> 51 #include <sys/dsl_prop.h> 52 #include <sys/dsl_deleg.h> 53 #include <sys/dmu_objset.h> 54 #include <sys/ddi.h> 55 #include <sys/sunddi.h> 56 #include <sys/sunldi.h> 57 #include <sys/policy.h> 58 #include <sys/zone.h> 59 #include <sys/nvpair.h> 60 #include <sys/pathname.h> 61 #include <sys/mount.h> 62 #include <sys/sdt.h> 63 #include <sys/fs/zfs.h> 64 #include <sys/zfs_ctldir.h> 65 #include <sys/zfs_dir.h> 66 #include <sys/zvol.h> 67 #include <sharefs/share.h> 68 #include <sys/zfs_znode.h> 69 #include <sys/zfs_vfsops.h> 70 #include <sys/dmu_objset.h> 71 72 #include "zfs_namecheck.h" 73 #include "zfs_prop.h" 74 #include "zfs_deleg.h" 75 76 extern struct modlfs zfs_modlfs; 77 78 extern void zfs_init(void); 79 extern void zfs_fini(void); 80 81 ldi_ident_t zfs_li = NULL; 82 dev_info_t *zfs_dip; 83 84 typedef int zfs_ioc_func_t(zfs_cmd_t *); 85 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, cred_t *); 86 87 typedef struct zfs_ioc_vec { 88 zfs_ioc_func_t *zvec_func; 89 zfs_secpolicy_func_t *zvec_secpolicy; 90 enum { 91 NO_NAME, 92 POOL_NAME, 93 DATASET_NAME 94 } zvec_namecheck; 95 boolean_t zvec_his_log; 96 } zfs_ioc_vec_t; 97 98 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */ 99 void 100 __dprintf(const char *file, const char *func, int line, const char *fmt, ...) 101 { 102 const char *newfile; 103 char buf[256]; 104 va_list adx; 105 106 /* 107 * Get rid of annoying "../common/" prefix to filename. 108 */ 109 newfile = strrchr(file, '/'); 110 if (newfile != NULL) { 111 newfile = newfile + 1; /* Get rid of leading / */ 112 } else { 113 newfile = file; 114 } 115 116 va_start(adx, fmt); 117 (void) vsnprintf(buf, sizeof (buf), fmt, adx); 118 va_end(adx); 119 120 /* 121 * To get this data, use the zfs-dprintf probe as so: 122 * dtrace -q -n 'zfs-dprintf \ 123 * /stringof(arg0) == "dbuf.c"/ \ 124 * {printf("%s: %s", stringof(arg1), stringof(arg3))}' 125 * arg0 = file name 126 * arg1 = function name 127 * arg2 = line number 128 * arg3 = message 129 */ 130 DTRACE_PROBE4(zfs__dprintf, 131 char *, newfile, char *, func, int, line, char *, buf); 132 } 133 134 static void 135 history_str_free(char *buf) 136 { 137 kmem_free(buf, HIS_MAX_RECORD_LEN); 138 } 139 140 static char * 141 history_str_get(zfs_cmd_t *zc) 142 { 143 char *buf; 144 145 if (zc->zc_history == NULL) 146 return (NULL); 147 148 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP); 149 if (copyinstr((void *)(uintptr_t)zc->zc_history, 150 buf, HIS_MAX_RECORD_LEN, NULL) != 0) { 151 history_str_free(buf); 152 return (NULL); 153 } 154 155 buf[HIS_MAX_RECORD_LEN -1] = '\0'; 156 157 return (buf); 158 } 159 160 /* 161 * zfs_check_version 162 * 163 * Return non-zero if the spa version is less than requested version. 164 */ 165 static int 166 zfs_check_version(const char *name, int version) 167 { 168 169 spa_t *spa; 170 171 if (spa_open(name, &spa, FTAG) == 0) { 172 if (spa_version(spa) < version) { 173 spa_close(spa, FTAG); 174 return (1); 175 } 176 spa_close(spa, FTAG); 177 } 178 return (0); 179 } 180 181 static void 182 zfs_log_history(zfs_cmd_t *zc) 183 { 184 spa_t *spa; 185 char *buf; 186 187 if ((buf = history_str_get(zc)) == NULL) 188 return; 189 190 if (spa_open(zc->zc_name, &spa, FTAG) == 0) { 191 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY) 192 (void) spa_history_log(spa, buf, LOG_CMD_NORMAL); 193 spa_close(spa, FTAG); 194 } 195 history_str_free(buf); 196 } 197 198 /* 199 * Policy for top-level read operations (list pools). Requires no privileges, 200 * and can be used in the local zone, as there is no associated dataset. 201 */ 202 /* ARGSUSED */ 203 static int 204 zfs_secpolicy_none(zfs_cmd_t *zc, cred_t *cr) 205 { 206 return (0); 207 } 208 209 /* 210 * Policy for dataset read operations (list children, get statistics). Requires 211 * no privileges, but must be visible in the local zone. 212 */ 213 /* ARGSUSED */ 214 static int 215 zfs_secpolicy_read(zfs_cmd_t *zc, cred_t *cr) 216 { 217 if (INGLOBALZONE(curproc) || 218 zone_dataset_visible(zc->zc_name, NULL)) 219 return (0); 220 221 return (ENOENT); 222 } 223 224 static int 225 zfs_dozonecheck(const char *dataset, cred_t *cr) 226 { 227 uint64_t zoned; 228 int writable = 1; 229 230 /* 231 * The dataset must be visible by this zone -- check this first 232 * so they don't see EPERM on something they shouldn't know about. 233 */ 234 if (!INGLOBALZONE(curproc) && 235 !zone_dataset_visible(dataset, &writable)) 236 return (ENOENT); 237 238 if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL)) 239 return (ENOENT); 240 241 if (INGLOBALZONE(curproc)) { 242 /* 243 * If the fs is zoned, only root can access it from the 244 * global zone. 245 */ 246 if (secpolicy_zfs(cr) && zoned) 247 return (EPERM); 248 } else { 249 /* 250 * If we are in a local zone, the 'zoned' property must be set. 251 */ 252 if (!zoned) 253 return (EPERM); 254 255 /* must be writable by this zone */ 256 if (!writable) 257 return (EPERM); 258 } 259 return (0); 260 } 261 262 int 263 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr) 264 { 265 int error; 266 267 error = zfs_dozonecheck(name, cr); 268 if (error == 0) { 269 error = secpolicy_zfs(cr); 270 if (error) 271 error = dsl_deleg_access(name, perm, cr); 272 } 273 return (error); 274 } 275 276 static int 277 zfs_secpolicy_setprop(const char *name, zfs_prop_t prop, cred_t *cr) 278 { 279 /* 280 * Check permissions for special properties. 281 */ 282 switch (prop) { 283 case ZFS_PROP_ZONED: 284 /* 285 * Disallow setting of 'zoned' from within a local zone. 286 */ 287 if (!INGLOBALZONE(curproc)) 288 return (EPERM); 289 break; 290 291 case ZFS_PROP_QUOTA: 292 if (!INGLOBALZONE(curproc)) { 293 uint64_t zoned; 294 char setpoint[MAXNAMELEN]; 295 /* 296 * Unprivileged users are allowed to modify the 297 * quota on things *under* (ie. contained by) 298 * the thing they own. 299 */ 300 if (dsl_prop_get_integer(name, "zoned", &zoned, 301 setpoint)) 302 return (EPERM); 303 if (!zoned || strlen(name) <= strlen(setpoint)) 304 return (EPERM); 305 } 306 break; 307 } 308 309 return (zfs_secpolicy_write_perms(name, zfs_prop_to_name(prop), cr)); 310 } 311 312 int 313 zfs_secpolicy_fsacl(zfs_cmd_t *zc, cred_t *cr) 314 { 315 int error; 316 317 error = zfs_dozonecheck(zc->zc_name, cr); 318 if (error) 319 return (error); 320 321 /* 322 * permission to set permissions will be evaluated later in 323 * dsl_deleg_can_allow() 324 */ 325 return (0); 326 } 327 328 int 329 zfs_secpolicy_rollback(zfs_cmd_t *zc, cred_t *cr) 330 { 331 int error; 332 error = zfs_secpolicy_write_perms(zc->zc_name, 333 ZFS_DELEG_PERM_ROLLBACK, cr); 334 if (error == 0) 335 error = zfs_secpolicy_write_perms(zc->zc_name, 336 ZFS_DELEG_PERM_MOUNT, cr); 337 return (error); 338 } 339 340 int 341 zfs_secpolicy_send(zfs_cmd_t *zc, cred_t *cr) 342 { 343 return (zfs_secpolicy_write_perms(zc->zc_name, 344 ZFS_DELEG_PERM_SEND, cr)); 345 } 346 347 int 348 zfs_secpolicy_share(zfs_cmd_t *zc, cred_t *cr) 349 { 350 if (!INGLOBALZONE(curproc)) 351 return (EPERM); 352 353 if (secpolicy_nfs(cr) == 0) { 354 return (0); 355 } else { 356 vnode_t *vp; 357 int error; 358 359 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, 360 NO_FOLLOW, NULL, &vp)) != 0) 361 return (error); 362 363 /* Now make sure mntpnt and dataset are ZFS */ 364 365 if (vp->v_vfsp->vfs_fstype != zfsfstype || 366 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), 367 zc->zc_name) != 0)) { 368 VN_RELE(vp); 369 return (EPERM); 370 } 371 372 VN_RELE(vp); 373 return (dsl_deleg_access(zc->zc_name, 374 ZFS_DELEG_PERM_SHARE, cr)); 375 } 376 } 377 378 static int 379 zfs_get_parent(const char *datasetname, char *parent, int parentsize) 380 { 381 char *cp; 382 383 /* 384 * Remove the @bla or /bla from the end of the name to get the parent. 385 */ 386 (void) strncpy(parent, datasetname, parentsize); 387 cp = strrchr(parent, '@'); 388 if (cp != NULL) { 389 cp[0] = '\0'; 390 } else { 391 cp = strrchr(parent, '/'); 392 if (cp == NULL) 393 return (ENOENT); 394 cp[0] = '\0'; 395 } 396 397 return (0); 398 } 399 400 int 401 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr) 402 { 403 int error; 404 405 if ((error = zfs_secpolicy_write_perms(name, 406 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 407 return (error); 408 409 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr)); 410 } 411 412 static int 413 zfs_secpolicy_destroy(zfs_cmd_t *zc, cred_t *cr) 414 { 415 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr)); 416 } 417 418 /* 419 * Must have sys_config privilege to check the iscsi permission 420 */ 421 /* ARGSUSED */ 422 static int 423 zfs_secpolicy_iscsi(zfs_cmd_t *zc, cred_t *cr) 424 { 425 return (secpolicy_zfs(cr)); 426 } 427 428 int 429 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr) 430 { 431 char parentname[MAXNAMELEN]; 432 int error; 433 434 if ((error = zfs_secpolicy_write_perms(from, 435 ZFS_DELEG_PERM_RENAME, cr)) != 0) 436 return (error); 437 438 if ((error = zfs_secpolicy_write_perms(from, 439 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 440 return (error); 441 442 if ((error = zfs_get_parent(to, parentname, 443 sizeof (parentname))) != 0) 444 return (error); 445 446 if ((error = zfs_secpolicy_write_perms(parentname, 447 ZFS_DELEG_PERM_CREATE, cr)) != 0) 448 return (error); 449 450 if ((error = zfs_secpolicy_write_perms(parentname, 451 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 452 return (error); 453 454 return (error); 455 } 456 457 static int 458 zfs_secpolicy_rename(zfs_cmd_t *zc, cred_t *cr) 459 { 460 return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr)); 461 } 462 463 static int 464 zfs_secpolicy_promote(zfs_cmd_t *zc, cred_t *cr) 465 { 466 char parentname[MAXNAMELEN]; 467 objset_t *clone; 468 int error; 469 470 error = zfs_secpolicy_write_perms(zc->zc_name, 471 ZFS_DELEG_PERM_PROMOTE, cr); 472 if (error) 473 return (error); 474 475 error = dmu_objset_open(zc->zc_name, DMU_OST_ANY, 476 DS_MODE_STANDARD | DS_MODE_READONLY, &clone); 477 478 if (error == 0) { 479 dsl_dataset_t *pclone = NULL; 480 dsl_dir_t *dd; 481 dd = clone->os->os_dsl_dataset->ds_dir; 482 483 rw_enter(&dd->dd_pool->dp_config_rwlock, RW_READER); 484 error = dsl_dataset_open_obj(dd->dd_pool, 485 dd->dd_phys->dd_origin_obj, NULL, 486 DS_MODE_NONE, FTAG, &pclone); 487 rw_exit(&dd->dd_pool->dp_config_rwlock); 488 if (error) { 489 dmu_objset_close(clone); 490 return (error); 491 } 492 493 error = zfs_secpolicy_write_perms(zc->zc_name, 494 ZFS_DELEG_PERM_MOUNT, cr); 495 496 dsl_dataset_name(pclone, parentname); 497 dmu_objset_close(clone); 498 dsl_dataset_close(pclone, DS_MODE_NONE, FTAG); 499 if (error == 0) 500 error = zfs_secpolicy_write_perms(parentname, 501 ZFS_DELEG_PERM_PROMOTE, cr); 502 } 503 return (error); 504 } 505 506 static int 507 zfs_secpolicy_receive(zfs_cmd_t *zc, cred_t *cr) 508 { 509 int error; 510 511 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 512 ZFS_DELEG_PERM_RECEIVE, cr)) != 0) 513 return (error); 514 515 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 516 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 517 return (error); 518 519 return (zfs_secpolicy_write_perms(zc->zc_name, 520 ZFS_DELEG_PERM_CREATE, cr)); 521 } 522 523 int 524 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr) 525 { 526 int error; 527 528 if ((error = zfs_secpolicy_write_perms(name, 529 ZFS_DELEG_PERM_SNAPSHOT, cr)) != 0) 530 return (error); 531 532 error = zfs_secpolicy_write_perms(name, 533 ZFS_DELEG_PERM_MOUNT, cr); 534 535 return (error); 536 } 537 538 static int 539 zfs_secpolicy_snapshot(zfs_cmd_t *zc, cred_t *cr) 540 { 541 542 return (zfs_secpolicy_snapshot_perms(zc->zc_name, cr)); 543 } 544 545 static int 546 zfs_secpolicy_create(zfs_cmd_t *zc, cred_t *cr) 547 { 548 char parentname[MAXNAMELEN]; 549 int error; 550 551 if ((error = zfs_get_parent(zc->zc_name, parentname, 552 sizeof (parentname))) != 0) 553 return (error); 554 555 if (zc->zc_value[0] != '\0') { 556 if ((error = zfs_secpolicy_write_perms(zc->zc_value, 557 ZFS_DELEG_PERM_CLONE, cr)) != 0) 558 return (error); 559 } 560 561 if ((error = zfs_secpolicy_write_perms(parentname, 562 ZFS_DELEG_PERM_CREATE, cr)) != 0) 563 return (error); 564 565 error = zfs_secpolicy_write_perms(parentname, 566 ZFS_DELEG_PERM_MOUNT, cr); 567 568 return (error); 569 } 570 571 static int 572 zfs_secpolicy_umount(zfs_cmd_t *zc, cred_t *cr) 573 { 574 int error; 575 576 error = secpolicy_fs_unmount(cr, NULL); 577 if (error) { 578 error = dsl_deleg_access(zc->zc_name, ZFS_DELEG_PERM_MOUNT, cr); 579 } 580 return (error); 581 } 582 583 /* 584 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires 585 * SYS_CONFIG privilege, which is not available in a local zone. 586 */ 587 /* ARGSUSED */ 588 static int 589 zfs_secpolicy_config(zfs_cmd_t *zc, cred_t *cr) 590 { 591 if (secpolicy_sys_config(cr, B_FALSE) != 0) 592 return (EPERM); 593 594 return (0); 595 } 596 597 /* 598 * Just like zfs_secpolicy_config, except that we will check for 599 * mount permission on the dataset for permission to create/remove 600 * the minor nodes. 601 */ 602 static int 603 zfs_secpolicy_minor(zfs_cmd_t *zc, cred_t *cr) 604 { 605 if (secpolicy_sys_config(cr, B_FALSE) != 0) { 606 return (dsl_deleg_access(zc->zc_name, 607 ZFS_DELEG_PERM_MOUNT, cr)); 608 } 609 610 return (0); 611 } 612 613 /* 614 * Policy for fault injection. Requires all privileges. 615 */ 616 /* ARGSUSED */ 617 static int 618 zfs_secpolicy_inject(zfs_cmd_t *zc, cred_t *cr) 619 { 620 return (secpolicy_zinject(cr)); 621 } 622 623 static int 624 zfs_secpolicy_inherit(zfs_cmd_t *zc, cred_t *cr) 625 { 626 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value); 627 628 if (prop == ZPROP_INVAL) { 629 if (!zfs_prop_user(zc->zc_value)) 630 return (EINVAL); 631 return (zfs_secpolicy_write_perms(zc->zc_name, 632 ZFS_DELEG_PERM_USERPROP, cr)); 633 } else { 634 if (!zfs_prop_inheritable(prop)) 635 return (EINVAL); 636 return (zfs_secpolicy_setprop(zc->zc_name, prop, cr)); 637 } 638 } 639 640 /* 641 * Returns the nvlist as specified by the user in the zfs_cmd_t. 642 */ 643 static int 644 get_nvlist(uint64_t nvl, uint64_t size, nvlist_t **nvp) 645 { 646 char *packed; 647 int error; 648 nvlist_t *list = NULL; 649 650 /* 651 * Read in and unpack the user-supplied nvlist. 652 */ 653 if (size == 0) 654 return (EINVAL); 655 656 packed = kmem_alloc(size, KM_SLEEP); 657 658 if ((error = xcopyin((void *)(uintptr_t)nvl, packed, size)) != 0) { 659 kmem_free(packed, size); 660 return (error); 661 } 662 663 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) { 664 kmem_free(packed, size); 665 return (error); 666 } 667 668 kmem_free(packed, size); 669 670 *nvp = list; 671 return (0); 672 } 673 674 static int 675 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl) 676 { 677 char *packed = NULL; 678 size_t size; 679 int error; 680 681 VERIFY(nvlist_size(nvl, &size, NV_ENCODE_NATIVE) == 0); 682 683 if (size > zc->zc_nvlist_dst_size) { 684 error = ENOMEM; 685 } else { 686 packed = kmem_alloc(size, KM_SLEEP); 687 VERIFY(nvlist_pack(nvl, &packed, &size, NV_ENCODE_NATIVE, 688 KM_SLEEP) == 0); 689 error = xcopyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst, 690 size); 691 kmem_free(packed, size); 692 } 693 694 zc->zc_nvlist_dst_size = size; 695 return (error); 696 } 697 698 static int 699 zfs_ioc_pool_create(zfs_cmd_t *zc) 700 { 701 int error; 702 nvlist_t *config, *props = NULL; 703 char *buf; 704 705 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 706 &config)) 707 return (error); 708 709 if (zc->zc_nvlist_src_size != 0 && (error = 710 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, &props))) { 711 nvlist_free(config); 712 return (error); 713 } 714 715 buf = history_str_get(zc); 716 717 error = spa_create(zc->zc_name, config, props, buf); 718 719 if (buf != NULL) 720 history_str_free(buf); 721 722 nvlist_free(config); 723 724 if (props) 725 nvlist_free(props); 726 727 return (error); 728 } 729 730 static int 731 zfs_ioc_pool_destroy(zfs_cmd_t *zc) 732 { 733 int error; 734 zfs_log_history(zc); 735 error = spa_destroy(zc->zc_name); 736 return (error); 737 } 738 739 static int 740 zfs_ioc_pool_import(zfs_cmd_t *zc) 741 { 742 int error; 743 nvlist_t *config, *props = NULL; 744 uint64_t guid; 745 746 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 747 &config)) != 0) 748 return (error); 749 750 if (zc->zc_nvlist_src_size != 0 && (error = 751 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, &props))) { 752 nvlist_free(config); 753 return (error); 754 } 755 756 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 || 757 guid != zc->zc_guid) 758 error = EINVAL; 759 else 760 error = spa_import(zc->zc_name, config, props); 761 762 nvlist_free(config); 763 764 if (props) 765 nvlist_free(props); 766 767 return (error); 768 } 769 770 static int 771 zfs_ioc_pool_export(zfs_cmd_t *zc) 772 { 773 int error; 774 zfs_log_history(zc); 775 error = spa_export(zc->zc_name, NULL); 776 return (error); 777 } 778 779 static int 780 zfs_ioc_pool_configs(zfs_cmd_t *zc) 781 { 782 nvlist_t *configs; 783 int error; 784 785 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL) 786 return (EEXIST); 787 788 error = put_nvlist(zc, configs); 789 790 nvlist_free(configs); 791 792 return (error); 793 } 794 795 static int 796 zfs_ioc_pool_stats(zfs_cmd_t *zc) 797 { 798 nvlist_t *config; 799 int error; 800 int ret = 0; 801 802 error = spa_get_stats(zc->zc_name, &config, zc->zc_value, 803 sizeof (zc->zc_value)); 804 805 if (config != NULL) { 806 ret = put_nvlist(zc, config); 807 nvlist_free(config); 808 809 /* 810 * The config may be present even if 'error' is non-zero. 811 * In this case we return success, and preserve the real errno 812 * in 'zc_cookie'. 813 */ 814 zc->zc_cookie = error; 815 } else { 816 ret = error; 817 } 818 819 return (ret); 820 } 821 822 /* 823 * Try to import the given pool, returning pool stats as appropriate so that 824 * user land knows which devices are available and overall pool health. 825 */ 826 static int 827 zfs_ioc_pool_tryimport(zfs_cmd_t *zc) 828 { 829 nvlist_t *tryconfig, *config; 830 int error; 831 832 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 833 &tryconfig)) != 0) 834 return (error); 835 836 config = spa_tryimport(tryconfig); 837 838 nvlist_free(tryconfig); 839 840 if (config == NULL) 841 return (EINVAL); 842 843 error = put_nvlist(zc, config); 844 nvlist_free(config); 845 846 return (error); 847 } 848 849 static int 850 zfs_ioc_pool_scrub(zfs_cmd_t *zc) 851 { 852 spa_t *spa; 853 int error; 854 855 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 856 return (error); 857 858 mutex_enter(&spa_namespace_lock); 859 error = spa_scrub(spa, zc->zc_cookie, B_FALSE); 860 mutex_exit(&spa_namespace_lock); 861 862 spa_close(spa, FTAG); 863 864 return (error); 865 } 866 867 static int 868 zfs_ioc_pool_freeze(zfs_cmd_t *zc) 869 { 870 spa_t *spa; 871 int error; 872 873 error = spa_open(zc->zc_name, &spa, FTAG); 874 if (error == 0) { 875 spa_freeze(spa); 876 spa_close(spa, FTAG); 877 } 878 return (error); 879 } 880 881 static int 882 zfs_ioc_pool_upgrade(zfs_cmd_t *zc) 883 { 884 spa_t *spa; 885 int error; 886 887 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 888 return (error); 889 890 if (zc->zc_cookie < spa_version(spa) || zc->zc_cookie > SPA_VERSION) { 891 spa_close(spa, FTAG); 892 return (EINVAL); 893 } 894 895 spa_upgrade(spa, zc->zc_cookie); 896 spa_close(spa, FTAG); 897 898 return (error); 899 } 900 901 static int 902 zfs_ioc_pool_get_history(zfs_cmd_t *zc) 903 { 904 spa_t *spa; 905 char *hist_buf; 906 uint64_t size; 907 int error; 908 909 if ((size = zc->zc_history_len) == 0) 910 return (EINVAL); 911 912 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 913 return (error); 914 915 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) { 916 spa_close(spa, FTAG); 917 return (ENOTSUP); 918 } 919 920 hist_buf = kmem_alloc(size, KM_SLEEP); 921 if ((error = spa_history_get(spa, &zc->zc_history_offset, 922 &zc->zc_history_len, hist_buf)) == 0) { 923 error = xcopyout(hist_buf, 924 (char *)(uintptr_t)zc->zc_history, 925 zc->zc_history_len); 926 } 927 928 spa_close(spa, FTAG); 929 kmem_free(hist_buf, size); 930 return (error); 931 } 932 933 static int 934 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc) 935 { 936 int error; 937 938 if (error = dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value)) 939 return (error); 940 941 return (0); 942 } 943 944 static int 945 zfs_ioc_obj_to_path(zfs_cmd_t *zc) 946 { 947 objset_t *osp; 948 int error; 949 950 if ((error = dmu_objset_open(zc->zc_name, DMU_OST_ZFS, 951 DS_MODE_NONE | DS_MODE_READONLY, &osp)) != 0) 952 return (error); 953 954 error = zfs_obj_to_path(osp, zc->zc_obj, zc->zc_value, 955 sizeof (zc->zc_value)); 956 dmu_objset_close(osp); 957 958 return (error); 959 } 960 961 static int 962 zfs_ioc_vdev_add(zfs_cmd_t *zc) 963 { 964 spa_t *spa; 965 int error; 966 nvlist_t *config; 967 968 error = spa_open(zc->zc_name, &spa, FTAG); 969 if (error != 0) 970 return (error); 971 972 /* 973 * A root pool with concatenated devices is not supported. 974 * Thus, can not add a device to a root pool with one device. 975 */ 976 if (spa->spa_root_vdev->vdev_children == 1 && spa->spa_bootfs != 0) { 977 spa_close(spa, FTAG); 978 return (EDOM); 979 } 980 981 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 982 &config)) == 0) { 983 error = spa_vdev_add(spa, config); 984 nvlist_free(config); 985 } 986 spa_close(spa, FTAG); 987 return (error); 988 } 989 990 static int 991 zfs_ioc_vdev_remove(zfs_cmd_t *zc) 992 { 993 spa_t *spa; 994 int error; 995 996 error = spa_open(zc->zc_name, &spa, FTAG); 997 if (error != 0) 998 return (error); 999 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE); 1000 spa_close(spa, FTAG); 1001 return (error); 1002 } 1003 1004 static int 1005 zfs_ioc_vdev_set_state(zfs_cmd_t *zc) 1006 { 1007 spa_t *spa; 1008 int error; 1009 vdev_state_t newstate = VDEV_STATE_UNKNOWN; 1010 1011 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1012 return (error); 1013 switch (zc->zc_cookie) { 1014 case VDEV_STATE_ONLINE: 1015 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate); 1016 break; 1017 1018 case VDEV_STATE_OFFLINE: 1019 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj); 1020 break; 1021 1022 case VDEV_STATE_FAULTED: 1023 error = vdev_fault(spa, zc->zc_guid); 1024 break; 1025 1026 case VDEV_STATE_DEGRADED: 1027 error = vdev_degrade(spa, zc->zc_guid); 1028 break; 1029 1030 default: 1031 error = EINVAL; 1032 } 1033 zc->zc_cookie = newstate; 1034 spa_close(spa, FTAG); 1035 return (error); 1036 } 1037 1038 static int 1039 zfs_ioc_vdev_attach(zfs_cmd_t *zc) 1040 { 1041 spa_t *spa; 1042 int replacing = zc->zc_cookie; 1043 nvlist_t *config; 1044 int error; 1045 1046 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1047 return (error); 1048 1049 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1050 &config)) == 0) { 1051 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing); 1052 nvlist_free(config); 1053 } 1054 1055 spa_close(spa, FTAG); 1056 return (error); 1057 } 1058 1059 static int 1060 zfs_ioc_vdev_detach(zfs_cmd_t *zc) 1061 { 1062 spa_t *spa; 1063 int error; 1064 1065 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1066 return (error); 1067 1068 error = spa_vdev_detach(spa, zc->zc_guid, B_FALSE); 1069 1070 spa_close(spa, FTAG); 1071 return (error); 1072 } 1073 1074 static int 1075 zfs_ioc_vdev_setpath(zfs_cmd_t *zc) 1076 { 1077 spa_t *spa; 1078 char *path = zc->zc_value; 1079 uint64_t guid = zc->zc_guid; 1080 int error; 1081 1082 error = spa_open(zc->zc_name, &spa, FTAG); 1083 if (error != 0) 1084 return (error); 1085 1086 error = spa_vdev_setpath(spa, guid, path); 1087 spa_close(spa, FTAG); 1088 return (error); 1089 } 1090 1091 /* 1092 * inputs: 1093 * zc_name name of filesystem 1094 * zc_nvlist_dst_size size of buffer for property nvlist 1095 * 1096 * outputs: 1097 * zc_objset_stats stats 1098 * zc_nvlist_dst property nvlist 1099 * zc_nvlist_dst_size size of property nvlist 1100 * zc_value alternate root 1101 */ 1102 static int 1103 zfs_ioc_objset_stats(zfs_cmd_t *zc) 1104 { 1105 objset_t *os = NULL; 1106 int error; 1107 nvlist_t *nv; 1108 1109 retry: 1110 error = dmu_objset_open(zc->zc_name, DMU_OST_ANY, 1111 DS_MODE_STANDARD | DS_MODE_READONLY, &os); 1112 if (error != 0) { 1113 /* 1114 * This is ugly: dmu_objset_open() can return EBUSY if 1115 * the objset is held exclusively. Fortunately this hold is 1116 * only for a short while, so we retry here. 1117 * This avoids user code having to handle EBUSY, 1118 * for example for a "zfs list". 1119 */ 1120 if (error == EBUSY) { 1121 delay(1); 1122 goto retry; 1123 } 1124 return (error); 1125 } 1126 1127 dmu_objset_fast_stat(os, &zc->zc_objset_stats); 1128 1129 if (zc->zc_nvlist_dst != 0 && 1130 (error = dsl_prop_get_all(os, &nv)) == 0) { 1131 dmu_objset_stats(os, nv); 1132 /* 1133 * NB: zvol_get_stats() will read the objset contents, 1134 * which we aren't supposed to do with a 1135 * DS_MODE_STANDARD open, because it could be 1136 * inconsistent. So this is a bit of a workaround... 1137 */ 1138 if (!zc->zc_objset_stats.dds_inconsistent) { 1139 if (dmu_objset_type(os) == DMU_OST_ZVOL) 1140 VERIFY(zvol_get_stats(os, nv) == 0); 1141 } 1142 error = put_nvlist(zc, nv); 1143 nvlist_free(nv); 1144 } 1145 1146 spa_altroot(dmu_objset_spa(os), zc->zc_value, sizeof (zc->zc_value)); 1147 1148 dmu_objset_close(os); 1149 return (error); 1150 } 1151 1152 /* 1153 * inputs: 1154 * zc_name name of filesystem 1155 * zc_cookie zap cursor 1156 * zc_nvlist_dst_size size of buffer for property nvlist 1157 * 1158 * outputs: 1159 * zc_name name of next filesystem 1160 * zc_objset_stats stats 1161 * zc_nvlist_dst property nvlist 1162 * zc_nvlist_dst_size size of property nvlist 1163 * zc_value alternate root 1164 */ 1165 static int 1166 zfs_ioc_objset_version(zfs_cmd_t *zc) 1167 { 1168 objset_t *os = NULL; 1169 int error; 1170 1171 retry: 1172 error = dmu_objset_open(zc->zc_name, DMU_OST_ANY, 1173 DS_MODE_STANDARD | DS_MODE_READONLY, &os); 1174 if (error != 0) { 1175 /* 1176 * This is ugly: dmu_objset_open() can return EBUSY if 1177 * the objset is held exclusively. Fortunately this hold is 1178 * only for a short while, so we retry here. 1179 * This avoids user code having to handle EBUSY, 1180 * for example for a "zfs list". 1181 */ 1182 if (error == EBUSY) { 1183 delay(1); 1184 goto retry; 1185 } 1186 return (error); 1187 } 1188 1189 dmu_objset_fast_stat(os, &zc->zc_objset_stats); 1190 1191 /* 1192 * NB: zfs_get_version() will read the objset contents, 1193 * which we aren't supposed to do with a 1194 * DS_MODE_STANDARD open, because it could be 1195 * inconsistent. So this is a bit of a workaround... 1196 */ 1197 zc->zc_cookie = 0; 1198 if (!zc->zc_objset_stats.dds_inconsistent) 1199 if (dmu_objset_type(os) == DMU_OST_ZFS) 1200 (void) zfs_get_version(os, &zc->zc_cookie); 1201 1202 dmu_objset_close(os); 1203 return (0); 1204 } 1205 1206 static int 1207 zfs_ioc_dataset_list_next(zfs_cmd_t *zc) 1208 { 1209 objset_t *os; 1210 int error; 1211 char *p; 1212 1213 retry: 1214 error = dmu_objset_open(zc->zc_name, DMU_OST_ANY, 1215 DS_MODE_STANDARD | DS_MODE_READONLY, &os); 1216 if (error != 0) { 1217 /* 1218 * This is ugly: dmu_objset_open() can return EBUSY if 1219 * the objset is held exclusively. Fortunately this hold is 1220 * only for a short while, so we retry here. 1221 * This avoids user code having to handle EBUSY, 1222 * for example for a "zfs list". 1223 */ 1224 if (error == EBUSY) { 1225 delay(1); 1226 goto retry; 1227 } 1228 if (error == ENOENT) 1229 error = ESRCH; 1230 return (error); 1231 } 1232 1233 p = strrchr(zc->zc_name, '/'); 1234 if (p == NULL || p[1] != '\0') 1235 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name)); 1236 p = zc->zc_name + strlen(zc->zc_name); 1237 1238 do { 1239 error = dmu_dir_list_next(os, 1240 sizeof (zc->zc_name) - (p - zc->zc_name), p, 1241 NULL, &zc->zc_cookie); 1242 if (error == ENOENT) 1243 error = ESRCH; 1244 } while (error == 0 && !INGLOBALZONE(curproc) && 1245 !zone_dataset_visible(zc->zc_name, NULL)); 1246 1247 /* 1248 * If it's a hidden dataset (ie. with a '$' in its name), don't 1249 * try to get stats for it. Userland will skip over it. 1250 */ 1251 if (error == 0 && strchr(zc->zc_name, '$') == NULL) 1252 error = zfs_ioc_objset_stats(zc); /* fill in the stats */ 1253 1254 dmu_objset_close(os); 1255 return (error); 1256 } 1257 1258 /* 1259 * inputs: 1260 * zc_name name of filesystem 1261 * zc_cookie zap cursor 1262 * zc_nvlist_dst_size size of buffer for property nvlist 1263 * 1264 * outputs: 1265 * zc_name name of next snapshot 1266 * zc_objset_stats stats 1267 * zc_nvlist_dst property nvlist 1268 * zc_nvlist_dst_size size of property nvlist 1269 * zc_value alternate root 1270 */ 1271 static int 1272 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc) 1273 { 1274 objset_t *os; 1275 int error; 1276 1277 retry: 1278 error = dmu_objset_open(zc->zc_name, DMU_OST_ANY, 1279 DS_MODE_STANDARD | DS_MODE_READONLY, &os); 1280 if (error != 0) { 1281 /* 1282 * This is ugly: dmu_objset_open() can return EBUSY if 1283 * the objset is held exclusively. Fortunately this hold is 1284 * only for a short while, so we retry here. 1285 * This avoids user code having to handle EBUSY, 1286 * for example for a "zfs list". 1287 */ 1288 if (error == EBUSY) { 1289 delay(1); 1290 goto retry; 1291 } 1292 if (error == ENOENT) 1293 error = ESRCH; 1294 return (error); 1295 } 1296 1297 /* 1298 * A dataset name of maximum length cannot have any snapshots, 1299 * so exit immediately. 1300 */ 1301 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) { 1302 dmu_objset_close(os); 1303 return (ESRCH); 1304 } 1305 1306 error = dmu_snapshot_list_next(os, 1307 sizeof (zc->zc_name) - strlen(zc->zc_name), 1308 zc->zc_name + strlen(zc->zc_name), NULL, &zc->zc_cookie); 1309 if (error == ENOENT) 1310 error = ESRCH; 1311 1312 if (error == 0) 1313 error = zfs_ioc_objset_stats(zc); /* fill in the stats */ 1314 1315 /* if we failed, undo the @ that we tacked on to zc_name */ 1316 if (error != 0) 1317 *strchr(zc->zc_name, '@') = '\0'; 1318 1319 dmu_objset_close(os); 1320 return (error); 1321 } 1322 1323 static int 1324 zfs_set_prop_nvlist(const char *name, nvlist_t *nvl) 1325 { 1326 nvpair_t *elem; 1327 int error; 1328 uint64_t intval; 1329 char *strval; 1330 1331 /* 1332 * First validate permission to set all of the properties 1333 */ 1334 elem = NULL; 1335 while ((elem = nvlist_next_nvpair(nvl, elem)) != NULL) { 1336 const char *propname = nvpair_name(elem); 1337 zfs_prop_t prop = zfs_name_to_prop(propname); 1338 1339 if (prop == ZPROP_INVAL) { 1340 /* 1341 * If this is a user-defined property, it must be a 1342 * string, and there is no further validation to do. 1343 */ 1344 if (!zfs_prop_user(propname) || 1345 nvpair_type(elem) != DATA_TYPE_STRING) 1346 return (EINVAL); 1347 1348 if (error = zfs_secpolicy_write_perms(name, 1349 ZFS_DELEG_PERM_USERPROP, CRED())) 1350 return (error); 1351 continue; 1352 } 1353 1354 if ((error = zfs_secpolicy_setprop(name, prop, CRED())) != 0) 1355 return (error); 1356 1357 /* 1358 * Check that this value is valid for this pool version 1359 */ 1360 switch (prop) { 1361 case ZFS_PROP_COMPRESSION: 1362 /* 1363 * If the user specified gzip compression, make sure 1364 * the SPA supports it. We ignore any errors here since 1365 * we'll catch them later. 1366 */ 1367 if (nvpair_type(elem) == DATA_TYPE_UINT64 && 1368 nvpair_value_uint64(elem, &intval) == 0 && 1369 intval >= ZIO_COMPRESS_GZIP_1 && 1370 intval <= ZIO_COMPRESS_GZIP_9) { 1371 if (zfs_check_version(name, 1372 SPA_VERSION_GZIP_COMPRESSION)) 1373 return (ENOTSUP); 1374 } 1375 break; 1376 1377 case ZFS_PROP_COPIES: 1378 if (zfs_check_version(name, SPA_VERSION_DITTO_BLOCKS)) 1379 return (ENOTSUP); 1380 break; 1381 case ZFS_PROP_NORMALIZE: 1382 case ZFS_PROP_UTF8ONLY: 1383 case ZFS_PROP_CASE: 1384 if (zfs_check_version(name, SPA_VERSION_NORMALIZATION)) 1385 return (ENOTSUP); 1386 1387 } 1388 if ((error = zfs_secpolicy_setprop(name, prop, CRED())) != 0) 1389 return (error); 1390 } 1391 1392 elem = NULL; 1393 while ((elem = nvlist_next_nvpair(nvl, elem)) != NULL) { 1394 const char *propname = nvpair_name(elem); 1395 zfs_prop_t prop = zfs_name_to_prop(propname); 1396 1397 if (prop == ZPROP_INVAL) { 1398 VERIFY(nvpair_value_string(elem, &strval) == 0); 1399 error = dsl_prop_set(name, propname, 1, 1400 strlen(strval) + 1, strval); 1401 if (error == 0) 1402 continue; 1403 else 1404 return (error); 1405 } 1406 1407 switch (prop) { 1408 case ZFS_PROP_QUOTA: 1409 if ((error = nvpair_value_uint64(elem, &intval)) != 0 || 1410 (error = dsl_dir_set_quota(name, intval)) != 0) 1411 return (error); 1412 break; 1413 1414 case ZFS_PROP_RESERVATION: 1415 if ((error = nvpair_value_uint64(elem, &intval)) != 0 || 1416 (error = dsl_dir_set_reservation(name, 1417 intval)) != 0) 1418 return (error); 1419 break; 1420 1421 case ZFS_PROP_VOLSIZE: 1422 if ((error = nvpair_value_uint64(elem, &intval)) != 0 || 1423 (error = zvol_set_volsize(name, 1424 ddi_driver_major(zfs_dip), intval)) != 0) 1425 return (error); 1426 break; 1427 1428 case ZFS_PROP_VOLBLOCKSIZE: 1429 if ((error = nvpair_value_uint64(elem, &intval)) != 0 || 1430 (error = zvol_set_volblocksize(name, intval)) != 0) 1431 return (error); 1432 break; 1433 1434 case ZFS_PROP_VERSION: 1435 if ((error = nvpair_value_uint64(elem, &intval)) != 0 || 1436 (error = zfs_set_version(name, intval)) != 0) 1437 return (error); 1438 break; 1439 1440 default: 1441 if (nvpair_type(elem) == DATA_TYPE_STRING) { 1442 if (zfs_prop_get_type(prop) != 1443 PROP_TYPE_STRING) 1444 return (EINVAL); 1445 VERIFY(nvpair_value_string(elem, &strval) == 0); 1446 if ((error = dsl_prop_set(name, 1447 nvpair_name(elem), 1, strlen(strval) + 1, 1448 strval)) != 0) 1449 return (error); 1450 } else if (nvpair_type(elem) == DATA_TYPE_UINT64) { 1451 const char *unused; 1452 1453 VERIFY(nvpair_value_uint64(elem, &intval) == 0); 1454 1455 switch (zfs_prop_get_type(prop)) { 1456 case PROP_TYPE_NUMBER: 1457 break; 1458 case PROP_TYPE_STRING: 1459 return (EINVAL); 1460 case PROP_TYPE_INDEX: 1461 if (zfs_prop_index_to_string(prop, 1462 intval, &unused) != 0) 1463 return (EINVAL); 1464 break; 1465 default: 1466 cmn_err(CE_PANIC, 1467 "unknown property type"); 1468 break; 1469 } 1470 1471 if ((error = dsl_prop_set(name, propname, 1472 8, 1, &intval)) != 0) 1473 return (error); 1474 } else { 1475 return (EINVAL); 1476 } 1477 break; 1478 } 1479 } 1480 1481 return (0); 1482 } 1483 1484 /* 1485 * inputs: 1486 * zc_name name of filesystem 1487 * zc_value name of property to inherit 1488 * zc_nvlist_src{_size} nvlist of properties to apply 1489 * 1490 * outputs: none 1491 */ 1492 static int 1493 zfs_ioc_set_prop(zfs_cmd_t *zc) 1494 { 1495 nvlist_t *nvl; 1496 int error; 1497 1498 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1499 &nvl)) != 0) 1500 return (error); 1501 1502 error = zfs_set_prop_nvlist(zc->zc_name, nvl); 1503 1504 nvlist_free(nvl); 1505 return (error); 1506 } 1507 1508 /* 1509 * inputs: 1510 * zc_name name of filesystem 1511 * zc_value name of property to inherit 1512 * 1513 * outputs: none 1514 */ 1515 static int 1516 zfs_ioc_inherit_prop(zfs_cmd_t *zc) 1517 { 1518 /* the property name has been validated by zfs_secpolicy_inherit() */ 1519 return (dsl_prop_set(zc->zc_name, zc->zc_value, 0, 0, NULL)); 1520 } 1521 1522 static int 1523 zfs_ioc_pool_set_props(zfs_cmd_t *zc) 1524 { 1525 nvlist_t *props; 1526 spa_t *spa; 1527 int error; 1528 1529 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1530 &props))) 1531 return (error); 1532 1533 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { 1534 nvlist_free(props); 1535 return (error); 1536 } 1537 1538 error = spa_prop_set(spa, props); 1539 1540 nvlist_free(props); 1541 spa_close(spa, FTAG); 1542 1543 return (error); 1544 } 1545 1546 static int 1547 zfs_ioc_pool_get_props(zfs_cmd_t *zc) 1548 { 1549 spa_t *spa; 1550 int error; 1551 nvlist_t *nvp = NULL; 1552 1553 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1554 return (error); 1555 1556 error = spa_prop_get(spa, &nvp); 1557 1558 if (error == 0 && zc->zc_nvlist_dst != NULL) 1559 error = put_nvlist(zc, nvp); 1560 else 1561 error = EFAULT; 1562 1563 spa_close(spa, FTAG); 1564 1565 if (nvp) 1566 nvlist_free(nvp); 1567 return (error); 1568 } 1569 1570 static int 1571 zfs_ioc_iscsi_perm_check(zfs_cmd_t *zc) 1572 { 1573 nvlist_t *nvp; 1574 int error; 1575 uint32_t uid; 1576 uint32_t gid; 1577 uint32_t *groups; 1578 uint_t group_cnt; 1579 cred_t *usercred; 1580 1581 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1582 &nvp)) != 0) { 1583 return (error); 1584 } 1585 1586 if ((error = nvlist_lookup_uint32(nvp, 1587 ZFS_DELEG_PERM_UID, &uid)) != 0) { 1588 nvlist_free(nvp); 1589 return (EPERM); 1590 } 1591 1592 if ((error = nvlist_lookup_uint32(nvp, 1593 ZFS_DELEG_PERM_GID, &gid)) != 0) { 1594 nvlist_free(nvp); 1595 return (EPERM); 1596 } 1597 1598 if ((error = nvlist_lookup_uint32_array(nvp, ZFS_DELEG_PERM_GROUPS, 1599 &groups, &group_cnt)) != 0) { 1600 nvlist_free(nvp); 1601 return (EPERM); 1602 } 1603 usercred = cralloc(); 1604 if ((crsetugid(usercred, uid, gid) != 0) || 1605 (crsetgroups(usercred, group_cnt, (gid_t *)groups) != 0)) { 1606 nvlist_free(nvp); 1607 crfree(usercred); 1608 return (EPERM); 1609 } 1610 nvlist_free(nvp); 1611 error = dsl_deleg_access(zc->zc_name, 1612 zfs_prop_to_name(ZFS_PROP_SHAREISCSI), usercred); 1613 crfree(usercred); 1614 return (error); 1615 } 1616 1617 /* 1618 * inputs: 1619 * zc_name name of filesystem 1620 * zc_nvlist_src{_size} nvlist of delegated permissions 1621 * zc_perm_action allow/unallow flag 1622 * 1623 * outputs: none 1624 */ 1625 static int 1626 zfs_ioc_set_fsacl(zfs_cmd_t *zc) 1627 { 1628 int error; 1629 nvlist_t *fsaclnv = NULL; 1630 1631 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1632 &fsaclnv)) != 0) 1633 return (error); 1634 1635 /* 1636 * Verify nvlist is constructed correctly 1637 */ 1638 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) { 1639 nvlist_free(fsaclnv); 1640 return (EINVAL); 1641 } 1642 1643 /* 1644 * If we don't have PRIV_SYS_MOUNT, then validate 1645 * that user is allowed to hand out each permission in 1646 * the nvlist(s) 1647 */ 1648 1649 error = secpolicy_zfs(CRED()); 1650 if (error) { 1651 if (zc->zc_perm_action == B_FALSE) { 1652 error = dsl_deleg_can_allow(zc->zc_name, 1653 fsaclnv, CRED()); 1654 } else { 1655 error = dsl_deleg_can_unallow(zc->zc_name, 1656 fsaclnv, CRED()); 1657 } 1658 } 1659 1660 if (error == 0) 1661 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action); 1662 1663 nvlist_free(fsaclnv); 1664 return (error); 1665 } 1666 1667 /* 1668 * inputs: 1669 * zc_name name of filesystem 1670 * 1671 * outputs: 1672 * zc_nvlist_src{_size} nvlist of delegated permissions 1673 */ 1674 static int 1675 zfs_ioc_get_fsacl(zfs_cmd_t *zc) 1676 { 1677 nvlist_t *nvp; 1678 int error; 1679 1680 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) { 1681 error = put_nvlist(zc, nvp); 1682 nvlist_free(nvp); 1683 } 1684 1685 return (error); 1686 } 1687 1688 /* 1689 * inputs: 1690 * zc_name name of volume 1691 * 1692 * outputs: none 1693 */ 1694 static int 1695 zfs_ioc_create_minor(zfs_cmd_t *zc) 1696 { 1697 return (zvol_create_minor(zc->zc_name, ddi_driver_major(zfs_dip))); 1698 } 1699 1700 /* 1701 * inputs: 1702 * zc_name name of volume 1703 * 1704 * outputs: none 1705 */ 1706 static int 1707 zfs_ioc_remove_minor(zfs_cmd_t *zc) 1708 { 1709 return (zvol_remove_minor(zc->zc_name)); 1710 } 1711 1712 /* 1713 * Search the vfs list for a specified resource. Returns a pointer to it 1714 * or NULL if no suitable entry is found. The caller of this routine 1715 * is responsible for releasing the returned vfs pointer. 1716 */ 1717 static vfs_t * 1718 zfs_get_vfs(const char *resource) 1719 { 1720 struct vfs *vfsp; 1721 struct vfs *vfs_found = NULL; 1722 1723 vfs_list_read_lock(); 1724 vfsp = rootvfs; 1725 do { 1726 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) { 1727 VFS_HOLD(vfsp); 1728 vfs_found = vfsp; 1729 break; 1730 } 1731 vfsp = vfsp->vfs_next; 1732 } while (vfsp != rootvfs); 1733 vfs_list_unlock(); 1734 return (vfs_found); 1735 } 1736 1737 /* ARGSUSED */ 1738 static void 1739 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) 1740 { 1741 zfs_creat_t *zct = arg; 1742 uint64_t version; 1743 1744 if (spa_version(dmu_objset_spa(os)) >= SPA_VERSION_FUID) 1745 version = ZPL_VERSION; 1746 else 1747 version = ZPL_VERSION_FUID - 1; 1748 1749 (void) nvlist_lookup_uint64(zct->zct_props, 1750 zfs_prop_to_name(ZFS_PROP_VERSION), &version); 1751 1752 zfs_create_fs(os, cr, version, zct->zct_norm, tx); 1753 } 1754 1755 /* 1756 * zfs_prop_lookup() 1757 * 1758 * Look for the property first in the existing property nvlist. If 1759 * it's already present, you're done. If it's not there, attempt to 1760 * find the property value from a parent dataset. If that fails, fall 1761 * back to the property's default value. In either of these two 1762 * cases, if update is TRUE, add a value for the property to the 1763 * property nvlist. 1764 * 1765 * If the rval pointer is non-NULL, copy the discovered value to rval. 1766 * 1767 * If we get any unexpected errors, bail and return the error number 1768 * to the caller. 1769 * 1770 * If we succeed, return 0. 1771 */ 1772 static int 1773 zfs_prop_lookup(const char *parentname, zfs_prop_t propnum, 1774 nvlist_t *proplist, uint64_t *rval, boolean_t update) 1775 { 1776 const char *propname; 1777 uint64_t value; 1778 int error = ENOENT; 1779 1780 propname = zfs_prop_to_name(propnum); 1781 if (proplist != NULL) 1782 error = nvlist_lookup_uint64(proplist, propname, &value); 1783 if (error == ENOENT) { 1784 error = dsl_prop_get_integer(parentname, propname, 1785 &value, NULL); 1786 if (error == ENOENT) 1787 value = zfs_prop_default_numeric(propnum); 1788 else if (error != 0) 1789 return (error); 1790 if (update) { 1791 ASSERT(proplist != NULL); 1792 error = nvlist_add_uint64(proplist, propname, value); 1793 } 1794 } 1795 if (error == 0 && rval) 1796 *rval = value; 1797 return (error); 1798 } 1799 1800 /* 1801 * zfs_normalization_get 1802 * 1803 * Get the normalization flag value. If the properties have 1804 * non-default values, make sure the pool version is recent enough to 1805 * support these choices. 1806 */ 1807 static int 1808 zfs_normalization_get(const char *dataset, nvlist_t *proplist, int *norm, 1809 boolean_t update) 1810 { 1811 char parentname[MAXNAMELEN]; 1812 char poolname[MAXNAMELEN]; 1813 char *cp; 1814 uint64_t value; 1815 int check = 0; 1816 int error; 1817 1818 ASSERT(norm != NULL); 1819 *norm = 0; 1820 1821 (void) strncpy(parentname, dataset, sizeof (parentname)); 1822 cp = strrchr(parentname, '@'); 1823 if (cp != NULL) { 1824 cp[0] = '\0'; 1825 } else { 1826 cp = strrchr(parentname, '/'); 1827 if (cp == NULL) 1828 return (ENOENT); 1829 cp[0] = '\0'; 1830 } 1831 1832 (void) strncpy(poolname, dataset, sizeof (poolname)); 1833 cp = strchr(poolname, '/'); 1834 if (cp != NULL) 1835 cp[0] = '\0'; 1836 1837 /* 1838 * Make sure pool is of new enough vintage to support normalization. 1839 */ 1840 if (zfs_check_version(poolname, SPA_VERSION_NORMALIZATION)) 1841 return (0); 1842 1843 error = zfs_prop_lookup(parentname, ZFS_PROP_UTF8ONLY, 1844 proplist, &value, update); 1845 if (error != 0) 1846 return (error); 1847 if (value != zfs_prop_default_numeric(ZFS_PROP_UTF8ONLY)) 1848 check = 1; 1849 1850 error = zfs_prop_lookup(parentname, ZFS_PROP_NORMALIZE, 1851 proplist, &value, update); 1852 if (error != 0) 1853 return (error); 1854 if (value != zfs_prop_default_numeric(ZFS_PROP_NORMALIZE)) { 1855 check = 1; 1856 switch ((int)value) { 1857 case ZFS_NORMALIZE_NONE: 1858 break; 1859 case ZFS_NORMALIZE_C: 1860 *norm |= U8_TEXTPREP_NFC; 1861 break; 1862 case ZFS_NORMALIZE_D: 1863 *norm |= U8_TEXTPREP_NFD; 1864 break; 1865 case ZFS_NORMALIZE_KC: 1866 *norm |= U8_TEXTPREP_NFKC; 1867 break; 1868 case ZFS_NORMALIZE_KD: 1869 *norm |= U8_TEXTPREP_NFKD; 1870 break; 1871 default: 1872 ASSERT((int)value >= ZFS_NORMALIZE_NONE); 1873 ASSERT((int)value <= ZFS_NORMALIZE_KD); 1874 break; 1875 } 1876 } 1877 1878 error = zfs_prop_lookup(parentname, ZFS_PROP_CASE, 1879 proplist, &value, update); 1880 if (error != 0) 1881 return (error); 1882 if (value != zfs_prop_default_numeric(ZFS_PROP_CASE)) { 1883 check = 1; 1884 switch ((int)value) { 1885 case ZFS_CASE_SENSITIVE: 1886 break; 1887 case ZFS_CASE_INSENSITIVE: 1888 *norm |= U8_TEXTPREP_TOUPPER; 1889 break; 1890 case ZFS_CASE_MIXED: 1891 *norm |= U8_TEXTPREP_TOUPPER; 1892 break; 1893 default: 1894 ASSERT((int)value >= ZFS_CASE_SENSITIVE); 1895 ASSERT((int)value <= ZFS_CASE_MIXED); 1896 break; 1897 } 1898 } 1899 1900 /* 1901 * At the moment we are disabling non-default values for these 1902 * properties because they cannot be preserved properly with a 1903 * zfs send. 1904 */ 1905 if (check == 1) 1906 return (ENOTSUP); 1907 1908 return (0); 1909 } 1910 1911 /* 1912 * inputs: 1913 * zc_objset_type type of objset to create (fs vs zvol) 1914 * zc_name name of new objset 1915 * zc_value name of snapshot to clone from (may be empty) 1916 * zc_nvlist_src{_size} nvlist of properties to apply 1917 * 1918 * outputs: none 1919 */ 1920 static int 1921 zfs_ioc_create(zfs_cmd_t *zc) 1922 { 1923 objset_t *clone; 1924 int error = 0; 1925 zfs_creat_t zct; 1926 nvlist_t *nvprops = NULL; 1927 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx); 1928 dmu_objset_type_t type = zc->zc_objset_type; 1929 1930 switch (type) { 1931 1932 case DMU_OST_ZFS: 1933 cbfunc = zfs_create_cb; 1934 break; 1935 1936 case DMU_OST_ZVOL: 1937 cbfunc = zvol_create_cb; 1938 break; 1939 1940 default: 1941 cbfunc = NULL; 1942 } 1943 if (strchr(zc->zc_name, '@') || 1944 strchr(zc->zc_name, '%')) 1945 return (EINVAL); 1946 1947 if (zc->zc_nvlist_src != NULL && 1948 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1949 &nvprops)) != 0) 1950 return (error); 1951 1952 zct.zct_norm = 0; 1953 zct.zct_props = nvprops; 1954 1955 if (zc->zc_value[0] != '\0') { 1956 /* 1957 * We're creating a clone of an existing snapshot. 1958 */ 1959 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0'; 1960 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0) { 1961 nvlist_free(nvprops); 1962 return (EINVAL); 1963 } 1964 1965 error = dmu_objset_open(zc->zc_value, type, 1966 DS_MODE_STANDARD | DS_MODE_READONLY, &clone); 1967 if (error) { 1968 nvlist_free(nvprops); 1969 return (error); 1970 } 1971 error = dmu_objset_create(zc->zc_name, type, clone, NULL, NULL); 1972 if (error) { 1973 dmu_objset_close(clone); 1974 nvlist_free(nvprops); 1975 return (error); 1976 } 1977 /* 1978 * If caller did not provide any properties, allocate 1979 * an nvlist for properties, as we will be adding our set-once 1980 * properties to it. This carries the choices made on the 1981 * original file system into the clone. 1982 */ 1983 if (nvprops == NULL) 1984 VERIFY(nvlist_alloc(&nvprops, 1985 NV_UNIQUE_NAME, KM_SLEEP) == 0); 1986 1987 /* 1988 * We have to have normalization and case-folding 1989 * flags correct when we do the file system creation, 1990 * so go figure them out now. All we really care about 1991 * here is getting these values into the property list. 1992 */ 1993 error = zfs_normalization_get(zc->zc_value, nvprops, 1994 &zct.zct_norm, B_TRUE); 1995 if (error != 0) { 1996 dmu_objset_close(clone); 1997 nvlist_free(nvprops); 1998 return (error); 1999 } 2000 dmu_objset_close(clone); 2001 } else { 2002 if (cbfunc == NULL) { 2003 nvlist_free(nvprops); 2004 return (EINVAL); 2005 } 2006 2007 if (type == DMU_OST_ZVOL) { 2008 uint64_t volsize, volblocksize; 2009 2010 if (nvprops == NULL || 2011 nvlist_lookup_uint64(nvprops, 2012 zfs_prop_to_name(ZFS_PROP_VOLSIZE), 2013 &volsize) != 0) { 2014 nvlist_free(nvprops); 2015 return (EINVAL); 2016 } 2017 2018 if ((error = nvlist_lookup_uint64(nvprops, 2019 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), 2020 &volblocksize)) != 0 && error != ENOENT) { 2021 nvlist_free(nvprops); 2022 return (EINVAL); 2023 } 2024 2025 if (error != 0) 2026 volblocksize = zfs_prop_default_numeric( 2027 ZFS_PROP_VOLBLOCKSIZE); 2028 2029 if ((error = zvol_check_volblocksize( 2030 volblocksize)) != 0 || 2031 (error = zvol_check_volsize(volsize, 2032 volblocksize)) != 0) { 2033 nvlist_free(nvprops); 2034 return (error); 2035 } 2036 } else if (type == DMU_OST_ZFS) { 2037 uint64_t version; 2038 int error; 2039 2040 error = nvlist_lookup_uint64(nvprops, 2041 zfs_prop_to_name(ZFS_PROP_VERSION), &version); 2042 2043 if (error == 0 && (version < ZPL_VERSION_INITIAL || 2044 version > ZPL_VERSION)) { 2045 nvlist_free(nvprops); 2046 return (ENOTSUP); 2047 } else if (error == 0 && version >= ZPL_VERSION_FUID && 2048 zfs_check_version(zc->zc_name, SPA_VERSION_FUID)) { 2049 nvlist_free(nvprops); 2050 return (ENOTSUP); 2051 } 2052 2053 /* 2054 * We have to have normalization and 2055 * case-folding flags correct when we do the 2056 * file system creation, so go figure them out 2057 * now. The final argument to zfs_normalization_get() 2058 * tells that routine not to update the nvprops 2059 * list. 2060 */ 2061 error = zfs_normalization_get(zc->zc_name, nvprops, 2062 &zct.zct_norm, B_FALSE); 2063 if (error != 0) { 2064 nvlist_free(nvprops); 2065 return (error); 2066 } 2067 } 2068 error = dmu_objset_create(zc->zc_name, type, NULL, cbfunc, 2069 &zct); 2070 } 2071 2072 /* 2073 * It would be nice to do this atomically. 2074 */ 2075 if (error == 0) { 2076 if ((error = zfs_set_prop_nvlist(zc->zc_name, nvprops)) != 0) 2077 (void) dmu_objset_destroy(zc->zc_name); 2078 } 2079 2080 nvlist_free(nvprops); 2081 return (error); 2082 } 2083 2084 /* 2085 * inputs: 2086 * zc_name name of filesystem 2087 * zc_value short name of snapshot 2088 * zc_cookie recursive flag 2089 * 2090 * outputs: none 2091 */ 2092 static int 2093 zfs_ioc_snapshot(zfs_cmd_t *zc) 2094 { 2095 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0) 2096 return (EINVAL); 2097 return (dmu_objset_snapshot(zc->zc_name, 2098 zc->zc_value, zc->zc_cookie)); 2099 } 2100 2101 int 2102 zfs_unmount_snap(char *name, void *arg) 2103 { 2104 char *snapname = arg; 2105 char *cp; 2106 vfs_t *vfsp = NULL; 2107 2108 /* 2109 * Snapshots (which are under .zfs control) must be unmounted 2110 * before they can be destroyed. 2111 */ 2112 2113 if (snapname) { 2114 (void) strcat(name, "@"); 2115 (void) strcat(name, snapname); 2116 vfsp = zfs_get_vfs(name); 2117 cp = strchr(name, '@'); 2118 *cp = '\0'; 2119 } else if (strchr(name, '@')) { 2120 vfsp = zfs_get_vfs(name); 2121 } 2122 2123 if (vfsp) { 2124 /* 2125 * Always force the unmount for snapshots. 2126 */ 2127 int flag = MS_FORCE; 2128 int err; 2129 2130 if ((err = vn_vfswlock(vfsp->vfs_vnodecovered)) != 0) { 2131 VFS_RELE(vfsp); 2132 return (err); 2133 } 2134 VFS_RELE(vfsp); 2135 if ((err = dounmount(vfsp, flag, kcred)) != 0) 2136 return (err); 2137 } 2138 return (0); 2139 } 2140 2141 /* 2142 * inputs: 2143 * zc_name name of filesystem 2144 * zc_value short name of snapshot 2145 * 2146 * outputs: none 2147 */ 2148 static int 2149 zfs_ioc_destroy_snaps(zfs_cmd_t *zc) 2150 { 2151 int err; 2152 2153 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0) 2154 return (EINVAL); 2155 err = dmu_objset_find(zc->zc_name, 2156 zfs_unmount_snap, zc->zc_value, DS_FIND_CHILDREN); 2157 if (err) 2158 return (err); 2159 return (dmu_snapshots_destroy(zc->zc_name, zc->zc_value)); 2160 } 2161 2162 /* 2163 * inputs: 2164 * zc_name name of dataset to destroy 2165 * zc_objset_type type of objset 2166 * 2167 * outputs: none 2168 */ 2169 static int 2170 zfs_ioc_destroy(zfs_cmd_t *zc) 2171 { 2172 if (strchr(zc->zc_name, '@') && zc->zc_objset_type == DMU_OST_ZFS) { 2173 int err = zfs_unmount_snap(zc->zc_name, NULL); 2174 if (err) 2175 return (err); 2176 } 2177 2178 return (dmu_objset_destroy(zc->zc_name)); 2179 } 2180 2181 /* 2182 * inputs: 2183 * zc_name name of snapshot to roll back to 2184 * 2185 * outputs: none 2186 */ 2187 static int 2188 zfs_ioc_rollback(zfs_cmd_t *zc) 2189 { 2190 return (dmu_objset_rollback(zc->zc_name)); 2191 } 2192 2193 /* 2194 * inputs: 2195 * zc_name old name of dataset 2196 * zc_value new name of dataset 2197 * zc_cookie recursive flag (only valid for snapshots) 2198 * 2199 * outputs: none 2200 */ 2201 static int 2202 zfs_ioc_rename(zfs_cmd_t *zc) 2203 { 2204 boolean_t recursive = zc->zc_cookie & 1; 2205 2206 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0'; 2207 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 2208 strchr(zc->zc_value, '%')) 2209 return (EINVAL); 2210 2211 /* 2212 * Unmount snapshot unless we're doing a recursive rename, 2213 * in which case the dataset code figures out which snapshots 2214 * to unmount. 2215 */ 2216 if (!recursive && strchr(zc->zc_name, '@') != NULL && 2217 zc->zc_objset_type == DMU_OST_ZFS) { 2218 int err = zfs_unmount_snap(zc->zc_name, NULL); 2219 if (err) 2220 return (err); 2221 } 2222 2223 return (dmu_objset_rename(zc->zc_name, zc->zc_value, recursive)); 2224 } 2225 2226 /* 2227 * inputs: 2228 * zc_name name of containing filesystem 2229 * zc_nvlist_src{_size} nvlist of properties to apply 2230 * zc_value name of snapshot to create 2231 * zc_string name of clone origin (if DRR_FLAG_CLONE) 2232 * zc_cookie file descriptor to recv from 2233 * zc_begin_record the BEGIN record of the stream (not byteswapped) 2234 * zc_guid force flag 2235 * 2236 * outputs: 2237 * zc_cookie number of bytes read 2238 */ 2239 static int 2240 zfs_ioc_recv(zfs_cmd_t *zc) 2241 { 2242 file_t *fp; 2243 objset_t *os; 2244 dmu_recv_cookie_t drc; 2245 zfsvfs_t *zfsvfs = NULL; 2246 boolean_t force = (boolean_t)zc->zc_guid; 2247 int error, fd; 2248 offset_t off; 2249 nvlist_t *props = NULL; 2250 objset_t *origin = NULL; 2251 char *tosnap; 2252 char tofs[ZFS_MAXNAMELEN]; 2253 2254 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 2255 strchr(zc->zc_value, '@') == NULL || 2256 strchr(zc->zc_value, '%')) 2257 return (EINVAL); 2258 2259 (void) strcpy(tofs, zc->zc_value); 2260 tosnap = strchr(tofs, '@'); 2261 *tosnap = '\0'; 2262 tosnap++; 2263 2264 if (zc->zc_nvlist_src != NULL && 2265 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2266 &props)) != 0) 2267 return (error); 2268 2269 fd = zc->zc_cookie; 2270 fp = getf(fd); 2271 if (fp == NULL) { 2272 nvlist_free(props); 2273 return (EBADF); 2274 } 2275 2276 /* 2277 * Get the zfsvfs for the receiving objset. There 2278 * won't be one if we're operating on a zvol, if the 2279 * objset doesn't exist yet, or is not mounted. 2280 */ 2281 2282 error = dmu_objset_open(tofs, DMU_OST_ANY, 2283 DS_MODE_STANDARD | DS_MODE_READONLY, &os); 2284 if (!error) { 2285 if (dmu_objset_type(os) == DMU_OST_ZFS) { 2286 mutex_enter(&os->os->os_user_ptr_lock); 2287 zfsvfs = dmu_objset_get_user(os); 2288 if (zfsvfs != NULL) 2289 VFS_HOLD(zfsvfs->z_vfs); 2290 mutex_exit(&os->os->os_user_ptr_lock); 2291 } 2292 dmu_objset_close(os); 2293 } 2294 2295 if (zc->zc_string[0]) { 2296 error = dmu_objset_open(zc->zc_string, DMU_OST_ANY, 2297 DS_MODE_STANDARD | DS_MODE_READONLY, &origin); 2298 if (error) { 2299 if (zfsvfs != NULL) 2300 VFS_RELE(zfsvfs->z_vfs); 2301 nvlist_free(props); 2302 releasef(fd); 2303 return (error); 2304 } 2305 } 2306 2307 error = dmu_recv_begin(tofs, tosnap, &zc->zc_begin_record, 2308 force, origin, zfsvfs != NULL, &drc); 2309 if (origin) 2310 dmu_objset_close(origin); 2311 if (error) { 2312 if (zfsvfs != NULL) 2313 VFS_RELE(zfsvfs->z_vfs); 2314 nvlist_free(props); 2315 releasef(fd); 2316 return (error); 2317 } 2318 2319 /* 2320 * If properties are supplied, they are to completely replace 2321 * the existing ones; "inherit" any existing properties. 2322 */ 2323 if (props) { 2324 objset_t *os; 2325 nvlist_t *nv = NULL; 2326 2327 error = dmu_objset_open(tofs, DMU_OST_ANY, 2328 DS_MODE_STANDARD | DS_MODE_READONLY | DS_MODE_INCONSISTENT, 2329 &os); 2330 if (error == 0) { 2331 error = dsl_prop_get_all(os, &nv); 2332 dmu_objset_close(os); 2333 } 2334 if (error == 0) { 2335 nvpair_t *elem; 2336 zfs_cmd_t zc2 = { 0 }; 2337 2338 (void) strcpy(zc2.zc_name, tofs); 2339 for (elem = nvlist_next_nvpair(nv, NULL); elem; 2340 elem = nvlist_next_nvpair(nv, elem)) { 2341 (void) strcpy(zc2.zc_value, nvpair_name(elem)); 2342 if (zfs_secpolicy_inherit(&zc2, CRED()) == 0) 2343 (void) zfs_ioc_inherit_prop(&zc2); 2344 } 2345 } 2346 if (nv) 2347 nvlist_free(nv); 2348 } 2349 2350 /* 2351 * Set properties. Note, we ignore errors. Would be better to 2352 * do best-effort in zfs_set_prop_nvlist, too. 2353 */ 2354 (void) zfs_set_prop_nvlist(tofs, props); 2355 nvlist_free(props); 2356 2357 off = fp->f_offset; 2358 error = dmu_recv_stream(&drc, fp->f_vnode, &off); 2359 2360 if (error == 0) { 2361 if (zfsvfs != NULL) { 2362 char osname[MAXNAMELEN]; 2363 int mode; 2364 2365 (void) zfs_suspend_fs(zfsvfs, osname, &mode); 2366 error = dmu_recv_end(&drc); 2367 error |= zfs_resume_fs(zfsvfs, osname, mode); 2368 } else { 2369 error = dmu_recv_end(&drc); 2370 } 2371 } 2372 if (zfsvfs != NULL) 2373 VFS_RELE(zfsvfs->z_vfs); 2374 2375 zc->zc_cookie = off - fp->f_offset; 2376 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0) 2377 fp->f_offset = off; 2378 2379 releasef(fd); 2380 return (error); 2381 } 2382 2383 /* 2384 * inputs: 2385 * zc_name name of snapshot to send 2386 * zc_value short name of incremental fromsnap (may be empty) 2387 * zc_cookie file descriptor to send stream to 2388 * zc_obj fromorigin flag (mutually exclusive with zc_value) 2389 * 2390 * outputs: none 2391 */ 2392 static int 2393 zfs_ioc_send(zfs_cmd_t *zc) 2394 { 2395 objset_t *fromsnap = NULL; 2396 objset_t *tosnap; 2397 file_t *fp; 2398 int error; 2399 offset_t off; 2400 2401 error = dmu_objset_open(zc->zc_name, DMU_OST_ANY, 2402 DS_MODE_STANDARD | DS_MODE_READONLY, &tosnap); 2403 if (error) 2404 return (error); 2405 2406 if (zc->zc_value[0] != '\0') { 2407 char buf[MAXPATHLEN]; 2408 char *cp; 2409 2410 (void) strncpy(buf, zc->zc_name, sizeof (buf)); 2411 cp = strchr(buf, '@'); 2412 if (cp) 2413 *(cp+1) = 0; 2414 (void) strncat(buf, zc->zc_value, sizeof (buf)); 2415 error = dmu_objset_open(buf, DMU_OST_ANY, 2416 DS_MODE_STANDARD | DS_MODE_READONLY, &fromsnap); 2417 if (error) { 2418 dmu_objset_close(tosnap); 2419 return (error); 2420 } 2421 } 2422 2423 fp = getf(zc->zc_cookie); 2424 if (fp == NULL) { 2425 dmu_objset_close(tosnap); 2426 if (fromsnap) 2427 dmu_objset_close(fromsnap); 2428 return (EBADF); 2429 } 2430 2431 off = fp->f_offset; 2432 error = dmu_sendbackup(tosnap, fromsnap, zc->zc_obj, fp->f_vnode, &off); 2433 2434 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0) 2435 fp->f_offset = off; 2436 releasef(zc->zc_cookie); 2437 if (fromsnap) 2438 dmu_objset_close(fromsnap); 2439 dmu_objset_close(tosnap); 2440 return (error); 2441 } 2442 2443 static int 2444 zfs_ioc_inject_fault(zfs_cmd_t *zc) 2445 { 2446 int id, error; 2447 2448 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id, 2449 &zc->zc_inject_record); 2450 2451 if (error == 0) 2452 zc->zc_guid = (uint64_t)id; 2453 2454 return (error); 2455 } 2456 2457 static int 2458 zfs_ioc_clear_fault(zfs_cmd_t *zc) 2459 { 2460 return (zio_clear_fault((int)zc->zc_guid)); 2461 } 2462 2463 static int 2464 zfs_ioc_inject_list_next(zfs_cmd_t *zc) 2465 { 2466 int id = (int)zc->zc_guid; 2467 int error; 2468 2469 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name), 2470 &zc->zc_inject_record); 2471 2472 zc->zc_guid = id; 2473 2474 return (error); 2475 } 2476 2477 static int 2478 zfs_ioc_error_log(zfs_cmd_t *zc) 2479 { 2480 spa_t *spa; 2481 int error; 2482 size_t count = (size_t)zc->zc_nvlist_dst_size; 2483 2484 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 2485 return (error); 2486 2487 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst, 2488 &count); 2489 if (error == 0) 2490 zc->zc_nvlist_dst_size = count; 2491 else 2492 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa); 2493 2494 spa_close(spa, FTAG); 2495 2496 return (error); 2497 } 2498 2499 static int 2500 zfs_ioc_clear(zfs_cmd_t *zc) 2501 { 2502 spa_t *spa; 2503 vdev_t *vd; 2504 uint64_t txg; 2505 int error; 2506 2507 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 2508 return (error); 2509 2510 /* 2511 * Try to resume any I/Os which may have been suspended 2512 * as a result of a complete pool failure. 2513 */ 2514 if (!list_is_empty(&spa->spa_zio_list)) { 2515 if (zio_vdev_resume_io(spa) != 0) { 2516 spa_close(spa, FTAG); 2517 return (EIO); 2518 } 2519 } 2520 2521 txg = spa_vdev_enter(spa); 2522 2523 if (zc->zc_guid == 0) { 2524 vd = NULL; 2525 } else if ((vd = spa_lookup_by_guid(spa, zc->zc_guid)) == NULL) { 2526 (void) spa_vdev_exit(spa, NULL, txg, ENODEV); 2527 spa_close(spa, FTAG); 2528 return (ENODEV); 2529 } 2530 2531 vdev_clear(spa, vd, B_TRUE); 2532 2533 (void) spa_vdev_exit(spa, NULL, txg, 0); 2534 2535 spa_close(spa, FTAG); 2536 2537 return (0); 2538 } 2539 2540 /* 2541 * inputs: 2542 * zc_name name of filesystem 2543 * zc_value name of origin snapshot 2544 * 2545 * outputs: none 2546 */ 2547 static int 2548 zfs_ioc_promote(zfs_cmd_t *zc) 2549 { 2550 char *cp; 2551 2552 /* 2553 * We don't need to unmount *all* the origin fs's snapshots, but 2554 * it's easier. 2555 */ 2556 cp = strchr(zc->zc_value, '@'); 2557 if (cp) 2558 *cp = '\0'; 2559 (void) dmu_objset_find(zc->zc_value, 2560 zfs_unmount_snap, NULL, DS_FIND_SNAPSHOTS); 2561 return (dsl_dataset_promote(zc->zc_name)); 2562 } 2563 2564 /* 2565 * We don't want to have a hard dependency 2566 * against some special symbols in sharefs 2567 * nfs, and smbsrv. Determine them if needed when 2568 * the first file system is shared. 2569 * Neither sharefs, nfs or smbsrv are unloadable modules. 2570 */ 2571 int (*znfsexport_fs)(void *arg); 2572 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t); 2573 int (*zsmbexport_fs)(void *arg, boolean_t add_share); 2574 2575 int zfs_nfsshare_inited; 2576 int zfs_smbshare_inited; 2577 2578 ddi_modhandle_t nfs_mod; 2579 ddi_modhandle_t sharefs_mod; 2580 ddi_modhandle_t smbsrv_mod; 2581 kmutex_t zfs_share_lock; 2582 2583 static int 2584 zfs_init_sharefs() 2585 { 2586 int error; 2587 2588 ASSERT(MUTEX_HELD(&zfs_share_lock)); 2589 /* Both NFS and SMB shares also require sharetab support. */ 2590 if (sharefs_mod == NULL && ((sharefs_mod = 2591 ddi_modopen("fs/sharefs", 2592 KRTLD_MODE_FIRST, &error)) == NULL)) { 2593 return (ENOSYS); 2594 } 2595 if (zshare_fs == NULL && ((zshare_fs = 2596 (int (*)(enum sharefs_sys_op, share_t *, uint32_t)) 2597 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) { 2598 return (ENOSYS); 2599 } 2600 return (0); 2601 } 2602 2603 static int 2604 zfs_ioc_share(zfs_cmd_t *zc) 2605 { 2606 int error; 2607 int opcode; 2608 2609 switch (zc->zc_share.z_sharetype) { 2610 case ZFS_SHARE_NFS: 2611 case ZFS_UNSHARE_NFS: 2612 if (zfs_nfsshare_inited == 0) { 2613 mutex_enter(&zfs_share_lock); 2614 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs", 2615 KRTLD_MODE_FIRST, &error)) == NULL)) { 2616 mutex_exit(&zfs_share_lock); 2617 return (ENOSYS); 2618 } 2619 if (znfsexport_fs == NULL && 2620 ((znfsexport_fs = (int (*)(void *)) 2621 ddi_modsym(nfs_mod, 2622 "nfs_export", &error)) == NULL)) { 2623 mutex_exit(&zfs_share_lock); 2624 return (ENOSYS); 2625 } 2626 error = zfs_init_sharefs(); 2627 if (error) { 2628 mutex_exit(&zfs_share_lock); 2629 return (ENOSYS); 2630 } 2631 zfs_nfsshare_inited = 1; 2632 mutex_exit(&zfs_share_lock); 2633 } 2634 break; 2635 case ZFS_SHARE_SMB: 2636 case ZFS_UNSHARE_SMB: 2637 if (zfs_smbshare_inited == 0) { 2638 mutex_enter(&zfs_share_lock); 2639 if (smbsrv_mod == NULL && ((smbsrv_mod = 2640 ddi_modopen("drv/smbsrv", 2641 KRTLD_MODE_FIRST, &error)) == NULL)) { 2642 mutex_exit(&zfs_share_lock); 2643 return (ENOSYS); 2644 } 2645 if (zsmbexport_fs == NULL && ((zsmbexport_fs = 2646 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod, 2647 "lmshrd_share_upcall", &error)) == NULL)) { 2648 mutex_exit(&zfs_share_lock); 2649 return (ENOSYS); 2650 } 2651 error = zfs_init_sharefs(); 2652 if (error) { 2653 mutex_exit(&zfs_share_lock); 2654 return (ENOSYS); 2655 } 2656 zfs_smbshare_inited = 1; 2657 mutex_exit(&zfs_share_lock); 2658 } 2659 break; 2660 default: 2661 return (EINVAL); 2662 } 2663 2664 switch (zc->zc_share.z_sharetype) { 2665 case ZFS_SHARE_NFS: 2666 case ZFS_UNSHARE_NFS: 2667 if (error = 2668 znfsexport_fs((void *) 2669 (uintptr_t)zc->zc_share.z_exportdata)) 2670 return (error); 2671 break; 2672 case ZFS_SHARE_SMB: 2673 case ZFS_UNSHARE_SMB: 2674 if (error = zsmbexport_fs((void *) 2675 (uintptr_t)zc->zc_share.z_exportdata, 2676 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ? 2677 B_TRUE : B_FALSE)) { 2678 return (error); 2679 } 2680 break; 2681 } 2682 2683 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS || 2684 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ? 2685 SHAREFS_ADD : SHAREFS_REMOVE; 2686 2687 /* 2688 * Add or remove share from sharetab 2689 */ 2690 error = zshare_fs(opcode, 2691 (void *)(uintptr_t)zc->zc_share.z_sharedata, 2692 zc->zc_share.z_sharemax); 2693 2694 return (error); 2695 2696 } 2697 2698 /* 2699 * pool create, destroy, and export don't log the history as part of 2700 * zfsdev_ioctl, but rather zfs_ioc_pool_create, and zfs_ioc_pool_export 2701 * do the logging of those commands. 2702 */ 2703 static zfs_ioc_vec_t zfs_ioc_vec[] = { 2704 { zfs_ioc_pool_create, zfs_secpolicy_config, POOL_NAME, B_FALSE }, 2705 { zfs_ioc_pool_destroy, zfs_secpolicy_config, POOL_NAME, B_FALSE }, 2706 { zfs_ioc_pool_import, zfs_secpolicy_config, POOL_NAME, B_TRUE }, 2707 { zfs_ioc_pool_export, zfs_secpolicy_config, POOL_NAME, B_FALSE }, 2708 { zfs_ioc_pool_configs, zfs_secpolicy_none, NO_NAME, B_FALSE }, 2709 { zfs_ioc_pool_stats, zfs_secpolicy_read, POOL_NAME, B_FALSE }, 2710 { zfs_ioc_pool_tryimport, zfs_secpolicy_config, NO_NAME, B_FALSE }, 2711 { zfs_ioc_pool_scrub, zfs_secpolicy_config, POOL_NAME, B_TRUE }, 2712 { zfs_ioc_pool_freeze, zfs_secpolicy_config, NO_NAME, B_FALSE }, 2713 { zfs_ioc_pool_upgrade, zfs_secpolicy_config, POOL_NAME, B_TRUE }, 2714 { zfs_ioc_pool_get_history, zfs_secpolicy_config, POOL_NAME, B_FALSE }, 2715 { zfs_ioc_vdev_add, zfs_secpolicy_config, POOL_NAME, B_TRUE }, 2716 { zfs_ioc_vdev_remove, zfs_secpolicy_config, POOL_NAME, B_TRUE }, 2717 { zfs_ioc_vdev_set_state, zfs_secpolicy_config, POOL_NAME, B_TRUE }, 2718 { zfs_ioc_vdev_attach, zfs_secpolicy_config, POOL_NAME, B_TRUE }, 2719 { zfs_ioc_vdev_detach, zfs_secpolicy_config, POOL_NAME, B_TRUE }, 2720 { zfs_ioc_vdev_setpath, zfs_secpolicy_config, POOL_NAME, B_FALSE }, 2721 { zfs_ioc_objset_stats, zfs_secpolicy_read, DATASET_NAME, B_FALSE }, 2722 { zfs_ioc_objset_version, zfs_secpolicy_read, DATASET_NAME, B_FALSE }, 2723 { zfs_ioc_dataset_list_next, zfs_secpolicy_read, 2724 DATASET_NAME, B_FALSE }, 2725 { zfs_ioc_snapshot_list_next, zfs_secpolicy_read, 2726 DATASET_NAME, B_FALSE }, 2727 { zfs_ioc_set_prop, zfs_secpolicy_none, DATASET_NAME, B_TRUE }, 2728 { zfs_ioc_create_minor, zfs_secpolicy_minor, DATASET_NAME, B_FALSE }, 2729 { zfs_ioc_remove_minor, zfs_secpolicy_minor, DATASET_NAME, B_FALSE }, 2730 { zfs_ioc_create, zfs_secpolicy_create, DATASET_NAME, B_TRUE }, 2731 { zfs_ioc_destroy, zfs_secpolicy_destroy, DATASET_NAME, B_TRUE }, 2732 { zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, B_TRUE }, 2733 { zfs_ioc_rename, zfs_secpolicy_rename, DATASET_NAME, B_TRUE }, 2734 { zfs_ioc_recv, zfs_secpolicy_receive, DATASET_NAME, B_TRUE }, 2735 { zfs_ioc_send, zfs_secpolicy_send, DATASET_NAME, B_TRUE }, 2736 { zfs_ioc_inject_fault, zfs_secpolicy_inject, NO_NAME, B_FALSE }, 2737 { zfs_ioc_clear_fault, zfs_secpolicy_inject, NO_NAME, B_FALSE }, 2738 { zfs_ioc_inject_list_next, zfs_secpolicy_inject, NO_NAME, B_FALSE }, 2739 { zfs_ioc_error_log, zfs_secpolicy_inject, POOL_NAME, B_FALSE }, 2740 { zfs_ioc_clear, zfs_secpolicy_config, POOL_NAME, B_TRUE }, 2741 { zfs_ioc_promote, zfs_secpolicy_promote, DATASET_NAME, B_TRUE }, 2742 { zfs_ioc_destroy_snaps, zfs_secpolicy_destroy, DATASET_NAME, B_TRUE }, 2743 { zfs_ioc_snapshot, zfs_secpolicy_snapshot, DATASET_NAME, B_TRUE }, 2744 { zfs_ioc_dsobj_to_dsname, zfs_secpolicy_config, POOL_NAME, B_FALSE }, 2745 { zfs_ioc_obj_to_path, zfs_secpolicy_config, NO_NAME, B_FALSE }, 2746 { zfs_ioc_pool_set_props, zfs_secpolicy_config, POOL_NAME, B_TRUE }, 2747 { zfs_ioc_pool_get_props, zfs_secpolicy_read, POOL_NAME, B_FALSE }, 2748 { zfs_ioc_set_fsacl, zfs_secpolicy_fsacl, DATASET_NAME, B_TRUE }, 2749 { zfs_ioc_get_fsacl, zfs_secpolicy_read, DATASET_NAME, B_FALSE }, 2750 { zfs_ioc_iscsi_perm_check, zfs_secpolicy_iscsi, 2751 DATASET_NAME, B_FALSE }, 2752 { zfs_ioc_share, zfs_secpolicy_share, DATASET_NAME, B_FALSE }, 2753 { zfs_ioc_inherit_prop, zfs_secpolicy_inherit, DATASET_NAME, B_TRUE }, 2754 }; 2755 2756 static int 2757 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp) 2758 { 2759 zfs_cmd_t *zc; 2760 uint_t vec; 2761 int error, rc; 2762 2763 if (getminor(dev) != 0) 2764 return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp)); 2765 2766 vec = cmd - ZFS_IOC; 2767 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip)); 2768 2769 if (vec >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0])) 2770 return (EINVAL); 2771 2772 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP); 2773 2774 error = xcopyin((void *)arg, zc, sizeof (zfs_cmd_t)); 2775 2776 if (error == 0) 2777 error = zfs_ioc_vec[vec].zvec_secpolicy(zc, cr); 2778 2779 /* 2780 * Ensure that all pool/dataset names are valid before we pass down to 2781 * the lower layers. 2782 */ 2783 if (error == 0) { 2784 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0'; 2785 switch (zfs_ioc_vec[vec].zvec_namecheck) { 2786 case POOL_NAME: 2787 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0) 2788 error = EINVAL; 2789 break; 2790 2791 case DATASET_NAME: 2792 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0) 2793 error = EINVAL; 2794 break; 2795 2796 case NO_NAME: 2797 break; 2798 } 2799 } 2800 2801 if (error == 0) 2802 error = zfs_ioc_vec[vec].zvec_func(zc); 2803 2804 rc = xcopyout(zc, (void *)arg, sizeof (zfs_cmd_t)); 2805 if (error == 0) { 2806 error = rc; 2807 if (zfs_ioc_vec[vec].zvec_his_log == B_TRUE) 2808 zfs_log_history(zc); 2809 } 2810 2811 kmem_free(zc, sizeof (zfs_cmd_t)); 2812 return (error); 2813 } 2814 2815 static int 2816 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd) 2817 { 2818 if (cmd != DDI_ATTACH) 2819 return (DDI_FAILURE); 2820 2821 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0, 2822 DDI_PSEUDO, 0) == DDI_FAILURE) 2823 return (DDI_FAILURE); 2824 2825 zfs_dip = dip; 2826 2827 ddi_report_dev(dip); 2828 2829 return (DDI_SUCCESS); 2830 } 2831 2832 static int 2833 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd) 2834 { 2835 if (spa_busy() || zfs_busy() || zvol_busy()) 2836 return (DDI_FAILURE); 2837 2838 if (cmd != DDI_DETACH) 2839 return (DDI_FAILURE); 2840 2841 zfs_dip = NULL; 2842 2843 ddi_prop_remove_all(dip); 2844 ddi_remove_minor_node(dip, NULL); 2845 2846 return (DDI_SUCCESS); 2847 } 2848 2849 /*ARGSUSED*/ 2850 static int 2851 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result) 2852 { 2853 switch (infocmd) { 2854 case DDI_INFO_DEVT2DEVINFO: 2855 *result = zfs_dip; 2856 return (DDI_SUCCESS); 2857 2858 case DDI_INFO_DEVT2INSTANCE: 2859 *result = (void *)0; 2860 return (DDI_SUCCESS); 2861 } 2862 2863 return (DDI_FAILURE); 2864 } 2865 2866 /* 2867 * OK, so this is a little weird. 2868 * 2869 * /dev/zfs is the control node, i.e. minor 0. 2870 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0. 2871 * 2872 * /dev/zfs has basically nothing to do except serve up ioctls, 2873 * so most of the standard driver entry points are in zvol.c. 2874 */ 2875 static struct cb_ops zfs_cb_ops = { 2876 zvol_open, /* open */ 2877 zvol_close, /* close */ 2878 zvol_strategy, /* strategy */ 2879 nodev, /* print */ 2880 nodev, /* dump */ 2881 zvol_read, /* read */ 2882 zvol_write, /* write */ 2883 zfsdev_ioctl, /* ioctl */ 2884 nodev, /* devmap */ 2885 nodev, /* mmap */ 2886 nodev, /* segmap */ 2887 nochpoll, /* poll */ 2888 ddi_prop_op, /* prop_op */ 2889 NULL, /* streamtab */ 2890 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */ 2891 CB_REV, /* version */ 2892 nodev, /* async read */ 2893 nodev, /* async write */ 2894 }; 2895 2896 static struct dev_ops zfs_dev_ops = { 2897 DEVO_REV, /* version */ 2898 0, /* refcnt */ 2899 zfs_info, /* info */ 2900 nulldev, /* identify */ 2901 nulldev, /* probe */ 2902 zfs_attach, /* attach */ 2903 zfs_detach, /* detach */ 2904 nodev, /* reset */ 2905 &zfs_cb_ops, /* driver operations */ 2906 NULL /* no bus operations */ 2907 }; 2908 2909 static struct modldrv zfs_modldrv = { 2910 &mod_driverops, "ZFS storage pool version " SPA_VERSION_STRING, 2911 &zfs_dev_ops 2912 }; 2913 2914 static struct modlinkage modlinkage = { 2915 MODREV_1, 2916 (void *)&zfs_modlfs, 2917 (void *)&zfs_modldrv, 2918 NULL 2919 }; 2920 2921 2922 uint_t zfs_fsyncer_key; 2923 extern uint_t rrw_tsd_key; 2924 2925 int 2926 _init(void) 2927 { 2928 int error; 2929 2930 spa_init(FREAD | FWRITE); 2931 zfs_init(); 2932 zvol_init(); 2933 2934 if ((error = mod_install(&modlinkage)) != 0) { 2935 zvol_fini(); 2936 zfs_fini(); 2937 spa_fini(); 2938 return (error); 2939 } 2940 2941 tsd_create(&zfs_fsyncer_key, NULL); 2942 tsd_create(&rrw_tsd_key, NULL); 2943 2944 error = ldi_ident_from_mod(&modlinkage, &zfs_li); 2945 ASSERT(error == 0); 2946 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL); 2947 2948 return (0); 2949 } 2950 2951 int 2952 _fini(void) 2953 { 2954 int error; 2955 2956 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled) 2957 return (EBUSY); 2958 2959 if ((error = mod_remove(&modlinkage)) != 0) 2960 return (error); 2961 2962 zvol_fini(); 2963 zfs_fini(); 2964 spa_fini(); 2965 if (zfs_nfsshare_inited) 2966 (void) ddi_modclose(nfs_mod); 2967 if (zfs_smbshare_inited) 2968 (void) ddi_modclose(smbsrv_mod); 2969 if (zfs_nfsshare_inited || zfs_smbshare_inited) 2970 (void) ddi_modclose(sharefs_mod); 2971 2972 tsd_destroy(&zfs_fsyncer_key); 2973 ldi_ident_release(zfs_li); 2974 zfs_li = NULL; 2975 mutex_destroy(&zfs_share_lock); 2976 2977 return (error); 2978 } 2979 2980 int 2981 _info(struct modinfo *modinfop) 2982 { 2983 return (mod_info(&modlinkage, modinfop)); 2984 } 2985