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 2009 Sun Microsystems, Inc. All rights reserved. 23 * Use is subject to license terms. 24 */ 25 26 #include <sys/types.h> 27 #include <sys/param.h> 28 #include <sys/errno.h> 29 #include <sys/uio.h> 30 #include <sys/buf.h> 31 #include <sys/modctl.h> 32 #include <sys/open.h> 33 #include <sys/file.h> 34 #include <sys/kmem.h> 35 #include <sys/conf.h> 36 #include <sys/cmn_err.h> 37 #include <sys/stat.h> 38 #include <sys/zfs_ioctl.h> 39 #include <sys/zfs_vfsops.h> 40 #include <sys/zfs_znode.h> 41 #include <sys/zap.h> 42 #include <sys/spa.h> 43 #include <sys/spa_impl.h> 44 #include <sys/vdev.h> 45 #include <sys/priv_impl.h> 46 #include <sys/dmu.h> 47 #include <sys/dsl_dir.h> 48 #include <sys/dsl_dataset.h> 49 #include <sys/dsl_prop.h> 50 #include <sys/dsl_deleg.h> 51 #include <sys/dmu_objset.h> 52 #include <sys/ddi.h> 53 #include <sys/sunddi.h> 54 #include <sys/sunldi.h> 55 #include <sys/policy.h> 56 #include <sys/zone.h> 57 #include <sys/nvpair.h> 58 #include <sys/pathname.h> 59 #include <sys/mount.h> 60 #include <sys/sdt.h> 61 #include <sys/fs/zfs.h> 62 #include <sys/zfs_ctldir.h> 63 #include <sys/zfs_dir.h> 64 #include <sys/zvol.h> 65 #include <sharefs/share.h> 66 #include <sys/dmu_objset.h> 67 68 #include "zfs_namecheck.h" 69 #include "zfs_prop.h" 70 #include "zfs_deleg.h" 71 72 extern struct modlfs zfs_modlfs; 73 74 extern void zfs_init(void); 75 extern void zfs_fini(void); 76 77 ldi_ident_t zfs_li = NULL; 78 dev_info_t *zfs_dip; 79 80 typedef int zfs_ioc_func_t(zfs_cmd_t *); 81 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, cred_t *); 82 83 typedef enum { 84 NO_NAME, 85 POOL_NAME, 86 DATASET_NAME 87 } zfs_ioc_namecheck_t; 88 89 typedef struct zfs_ioc_vec { 90 zfs_ioc_func_t *zvec_func; 91 zfs_secpolicy_func_t *zvec_secpolicy; 92 zfs_ioc_namecheck_t zvec_namecheck; 93 boolean_t zvec_his_log; 94 boolean_t zvec_pool_check; 95 } zfs_ioc_vec_t; 96 97 /* This array is indexed by zfs_userquota_prop_t */ 98 static const char *userquota_perms[] = { 99 ZFS_DELEG_PERM_USERUSED, 100 ZFS_DELEG_PERM_USERQUOTA, 101 ZFS_DELEG_PERM_GROUPUSED, 102 ZFS_DELEG_PERM_GROUPQUOTA, 103 }; 104 105 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc); 106 static int zfs_check_settable(const char *name, nvpair_t *property, 107 cred_t *cr); 108 static int zfs_check_clearable(char *dataset, nvlist_t *props, 109 nvlist_t **errors); 110 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *, 111 boolean_t *); 112 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t **); 113 114 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */ 115 void 116 __dprintf(const char *file, const char *func, int line, const char *fmt, ...) 117 { 118 const char *newfile; 119 char buf[256]; 120 va_list adx; 121 122 /* 123 * Get rid of annoying "../common/" prefix to filename. 124 */ 125 newfile = strrchr(file, '/'); 126 if (newfile != NULL) { 127 newfile = newfile + 1; /* Get rid of leading / */ 128 } else { 129 newfile = file; 130 } 131 132 va_start(adx, fmt); 133 (void) vsnprintf(buf, sizeof (buf), fmt, adx); 134 va_end(adx); 135 136 /* 137 * To get this data, use the zfs-dprintf probe as so: 138 * dtrace -q -n 'zfs-dprintf \ 139 * /stringof(arg0) == "dbuf.c"/ \ 140 * {printf("%s: %s", stringof(arg1), stringof(arg3))}' 141 * arg0 = file name 142 * arg1 = function name 143 * arg2 = line number 144 * arg3 = message 145 */ 146 DTRACE_PROBE4(zfs__dprintf, 147 char *, newfile, char *, func, int, line, char *, buf); 148 } 149 150 static void 151 history_str_free(char *buf) 152 { 153 kmem_free(buf, HIS_MAX_RECORD_LEN); 154 } 155 156 static char * 157 history_str_get(zfs_cmd_t *zc) 158 { 159 char *buf; 160 161 if (zc->zc_history == NULL) 162 return (NULL); 163 164 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP); 165 if (copyinstr((void *)(uintptr_t)zc->zc_history, 166 buf, HIS_MAX_RECORD_LEN, NULL) != 0) { 167 history_str_free(buf); 168 return (NULL); 169 } 170 171 buf[HIS_MAX_RECORD_LEN -1] = '\0'; 172 173 return (buf); 174 } 175 176 /* 177 * Check to see if the named dataset is currently defined as bootable 178 */ 179 static boolean_t 180 zfs_is_bootfs(const char *name) 181 { 182 objset_t *os; 183 184 if (dmu_objset_hold(name, FTAG, &os) == 0) { 185 boolean_t ret; 186 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os))); 187 dmu_objset_rele(os, FTAG); 188 return (ret); 189 } 190 return (B_FALSE); 191 } 192 193 /* 194 * zfs_earlier_version 195 * 196 * Return non-zero if the spa version is less than requested version. 197 */ 198 static int 199 zfs_earlier_version(const char *name, int version) 200 { 201 spa_t *spa; 202 203 if (spa_open(name, &spa, FTAG) == 0) { 204 if (spa_version(spa) < version) { 205 spa_close(spa, FTAG); 206 return (1); 207 } 208 spa_close(spa, FTAG); 209 } 210 return (0); 211 } 212 213 /* 214 * zpl_earlier_version 215 * 216 * Return TRUE if the ZPL version is less than requested version. 217 */ 218 static boolean_t 219 zpl_earlier_version(const char *name, int version) 220 { 221 objset_t *os; 222 boolean_t rc = B_TRUE; 223 224 if (dmu_objset_hold(name, FTAG, &os) == 0) { 225 uint64_t zplversion; 226 227 if (dmu_objset_type(os) != DMU_OST_ZFS) { 228 dmu_objset_rele(os, FTAG); 229 return (B_TRUE); 230 } 231 /* XXX reading from non-owned objset */ 232 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0) 233 rc = zplversion < version; 234 dmu_objset_rele(os, FTAG); 235 } 236 return (rc); 237 } 238 239 static void 240 zfs_log_history(zfs_cmd_t *zc) 241 { 242 spa_t *spa; 243 char *buf; 244 245 if ((buf = history_str_get(zc)) == NULL) 246 return; 247 248 if (spa_open(zc->zc_name, &spa, FTAG) == 0) { 249 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY) 250 (void) spa_history_log(spa, buf, LOG_CMD_NORMAL); 251 spa_close(spa, FTAG); 252 } 253 history_str_free(buf); 254 } 255 256 /* 257 * Policy for top-level read operations (list pools). Requires no privileges, 258 * and can be used in the local zone, as there is no associated dataset. 259 */ 260 /* ARGSUSED */ 261 static int 262 zfs_secpolicy_none(zfs_cmd_t *zc, cred_t *cr) 263 { 264 return (0); 265 } 266 267 /* 268 * Policy for dataset read operations (list children, get statistics). Requires 269 * no privileges, but must be visible in the local zone. 270 */ 271 /* ARGSUSED */ 272 static int 273 zfs_secpolicy_read(zfs_cmd_t *zc, cred_t *cr) 274 { 275 if (INGLOBALZONE(curproc) || 276 zone_dataset_visible(zc->zc_name, NULL)) 277 return (0); 278 279 return (ENOENT); 280 } 281 282 static int 283 zfs_dozonecheck(const char *dataset, cred_t *cr) 284 { 285 uint64_t zoned; 286 int writable = 1; 287 288 /* 289 * The dataset must be visible by this zone -- check this first 290 * so they don't see EPERM on something they shouldn't know about. 291 */ 292 if (!INGLOBALZONE(curproc) && 293 !zone_dataset_visible(dataset, &writable)) 294 return (ENOENT); 295 296 if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL)) 297 return (ENOENT); 298 299 if (INGLOBALZONE(curproc)) { 300 /* 301 * If the fs is zoned, only root can access it from the 302 * global zone. 303 */ 304 if (secpolicy_zfs(cr) && zoned) 305 return (EPERM); 306 } else { 307 /* 308 * If we are in a local zone, the 'zoned' property must be set. 309 */ 310 if (!zoned) 311 return (EPERM); 312 313 /* must be writable by this zone */ 314 if (!writable) 315 return (EPERM); 316 } 317 return (0); 318 } 319 320 int 321 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr) 322 { 323 int error; 324 325 error = zfs_dozonecheck(name, cr); 326 if (error == 0) { 327 error = secpolicy_zfs(cr); 328 if (error) 329 error = dsl_deleg_access(name, perm, cr); 330 } 331 return (error); 332 } 333 334 /* 335 * Policy for setting the security label property. 336 * 337 * Returns 0 for success, non-zero for access and other errors. 338 */ 339 static int 340 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr) 341 { 342 char ds_hexsl[MAXNAMELEN]; 343 bslabel_t ds_sl, new_sl; 344 boolean_t new_default = FALSE; 345 uint64_t zoned; 346 int needed_priv = -1; 347 int error; 348 349 /* First get the existing dataset label. */ 350 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL), 351 1, sizeof (ds_hexsl), &ds_hexsl, NULL); 352 if (error) 353 return (EPERM); 354 355 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0) 356 new_default = TRUE; 357 358 /* The label must be translatable */ 359 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0)) 360 return (EINVAL); 361 362 /* 363 * In a non-global zone, disallow attempts to set a label that 364 * doesn't match that of the zone; otherwise no other checks 365 * are needed. 366 */ 367 if (!INGLOBALZONE(curproc)) { 368 if (new_default || !blequal(&new_sl, CR_SL(CRED()))) 369 return (EPERM); 370 return (0); 371 } 372 373 /* 374 * For global-zone datasets (i.e., those whose zoned property is 375 * "off", verify that the specified new label is valid for the 376 * global zone. 377 */ 378 if (dsl_prop_get_integer(name, 379 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL)) 380 return (EPERM); 381 if (!zoned) { 382 if (zfs_check_global_label(name, strval) != 0) 383 return (EPERM); 384 } 385 386 /* 387 * If the existing dataset label is nondefault, check if the 388 * dataset is mounted (label cannot be changed while mounted). 389 * Get the zfsvfs; if there isn't one, then the dataset isn't 390 * mounted (or isn't a dataset, doesn't exist, ...). 391 */ 392 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) { 393 objset_t *os; 394 static char *setsl_tag = "setsl_tag"; 395 396 /* 397 * Try to own the dataset; abort if there is any error, 398 * (e.g., already mounted, in use, or other error). 399 */ 400 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE, 401 setsl_tag, &os); 402 if (error) 403 return (EPERM); 404 405 dmu_objset_disown(os, setsl_tag); 406 407 if (new_default) { 408 needed_priv = PRIV_FILE_DOWNGRADE_SL; 409 goto out_check; 410 } 411 412 if (hexstr_to_label(strval, &new_sl) != 0) 413 return (EPERM); 414 415 if (blstrictdom(&ds_sl, &new_sl)) 416 needed_priv = PRIV_FILE_DOWNGRADE_SL; 417 else if (blstrictdom(&new_sl, &ds_sl)) 418 needed_priv = PRIV_FILE_UPGRADE_SL; 419 } else { 420 /* dataset currently has a default label */ 421 if (!new_default) 422 needed_priv = PRIV_FILE_UPGRADE_SL; 423 } 424 425 out_check: 426 if (needed_priv != -1) 427 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL)); 428 return (0); 429 } 430 431 static int 432 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval, 433 cred_t *cr) 434 { 435 char *strval; 436 437 /* 438 * Check permissions for special properties. 439 */ 440 switch (prop) { 441 case ZFS_PROP_ZONED: 442 /* 443 * Disallow setting of 'zoned' from within a local zone. 444 */ 445 if (!INGLOBALZONE(curproc)) 446 return (EPERM); 447 break; 448 449 case ZFS_PROP_QUOTA: 450 if (!INGLOBALZONE(curproc)) { 451 uint64_t zoned; 452 char setpoint[MAXNAMELEN]; 453 /* 454 * Unprivileged users are allowed to modify the 455 * quota on things *under* (ie. contained by) 456 * the thing they own. 457 */ 458 if (dsl_prop_get_integer(dsname, "zoned", &zoned, 459 setpoint)) 460 return (EPERM); 461 if (!zoned || strlen(dsname) <= strlen(setpoint)) 462 return (EPERM); 463 } 464 break; 465 466 case ZFS_PROP_MLSLABEL: 467 if (!is_system_labeled()) 468 return (EPERM); 469 470 if (nvpair_value_string(propval, &strval) == 0) { 471 int err; 472 473 err = zfs_set_slabel_policy(dsname, strval, CRED()); 474 if (err != 0) 475 return (err); 476 } 477 break; 478 } 479 480 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr)); 481 } 482 483 int 484 zfs_secpolicy_fsacl(zfs_cmd_t *zc, cred_t *cr) 485 { 486 int error; 487 488 error = zfs_dozonecheck(zc->zc_name, cr); 489 if (error) 490 return (error); 491 492 /* 493 * permission to set permissions will be evaluated later in 494 * dsl_deleg_can_allow() 495 */ 496 return (0); 497 } 498 499 int 500 zfs_secpolicy_rollback(zfs_cmd_t *zc, cred_t *cr) 501 { 502 return (zfs_secpolicy_write_perms(zc->zc_name, 503 ZFS_DELEG_PERM_ROLLBACK, cr)); 504 } 505 506 int 507 zfs_secpolicy_send(zfs_cmd_t *zc, cred_t *cr) 508 { 509 return (zfs_secpolicy_write_perms(zc->zc_name, 510 ZFS_DELEG_PERM_SEND, cr)); 511 } 512 513 static int 514 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, cred_t *cr) 515 { 516 vnode_t *vp; 517 int error; 518 519 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, 520 NO_FOLLOW, NULL, &vp)) != 0) 521 return (error); 522 523 /* Now make sure mntpnt and dataset are ZFS */ 524 525 if (vp->v_vfsp->vfs_fstype != zfsfstype || 526 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), 527 zc->zc_name) != 0)) { 528 VN_RELE(vp); 529 return (EPERM); 530 } 531 532 VN_RELE(vp); 533 return (dsl_deleg_access(zc->zc_name, 534 ZFS_DELEG_PERM_SHARE, cr)); 535 } 536 537 int 538 zfs_secpolicy_share(zfs_cmd_t *zc, cred_t *cr) 539 { 540 if (!INGLOBALZONE(curproc)) 541 return (EPERM); 542 543 if (secpolicy_nfs(cr) == 0) { 544 return (0); 545 } else { 546 return (zfs_secpolicy_deleg_share(zc, cr)); 547 } 548 } 549 550 int 551 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, cred_t *cr) 552 { 553 if (!INGLOBALZONE(curproc)) 554 return (EPERM); 555 556 if (secpolicy_smb(cr) == 0) { 557 return (0); 558 } else { 559 return (zfs_secpolicy_deleg_share(zc, cr)); 560 } 561 } 562 563 static int 564 zfs_get_parent(const char *datasetname, char *parent, int parentsize) 565 { 566 char *cp; 567 568 /* 569 * Remove the @bla or /bla from the end of the name to get the parent. 570 */ 571 (void) strncpy(parent, datasetname, parentsize); 572 cp = strrchr(parent, '@'); 573 if (cp != NULL) { 574 cp[0] = '\0'; 575 } else { 576 cp = strrchr(parent, '/'); 577 if (cp == NULL) 578 return (ENOENT); 579 cp[0] = '\0'; 580 } 581 582 return (0); 583 } 584 585 int 586 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr) 587 { 588 int error; 589 590 if ((error = zfs_secpolicy_write_perms(name, 591 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 592 return (error); 593 594 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr)); 595 } 596 597 static int 598 zfs_secpolicy_destroy(zfs_cmd_t *zc, cred_t *cr) 599 { 600 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr)); 601 } 602 603 /* 604 * Must have sys_config privilege to check the iscsi permission 605 */ 606 /* ARGSUSED */ 607 static int 608 zfs_secpolicy_iscsi(zfs_cmd_t *zc, cred_t *cr) 609 { 610 return (secpolicy_zfs(cr)); 611 } 612 613 int 614 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr) 615 { 616 char parentname[MAXNAMELEN]; 617 int error; 618 619 if ((error = zfs_secpolicy_write_perms(from, 620 ZFS_DELEG_PERM_RENAME, cr)) != 0) 621 return (error); 622 623 if ((error = zfs_secpolicy_write_perms(from, 624 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 625 return (error); 626 627 if ((error = zfs_get_parent(to, parentname, 628 sizeof (parentname))) != 0) 629 return (error); 630 631 if ((error = zfs_secpolicy_write_perms(parentname, 632 ZFS_DELEG_PERM_CREATE, cr)) != 0) 633 return (error); 634 635 if ((error = zfs_secpolicy_write_perms(parentname, 636 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 637 return (error); 638 639 return (error); 640 } 641 642 static int 643 zfs_secpolicy_rename(zfs_cmd_t *zc, cred_t *cr) 644 { 645 return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr)); 646 } 647 648 static int 649 zfs_secpolicy_promote(zfs_cmd_t *zc, cred_t *cr) 650 { 651 char parentname[MAXNAMELEN]; 652 objset_t *clone; 653 int error; 654 655 error = zfs_secpolicy_write_perms(zc->zc_name, 656 ZFS_DELEG_PERM_PROMOTE, cr); 657 if (error) 658 return (error); 659 660 error = dmu_objset_hold(zc->zc_name, FTAG, &clone); 661 662 if (error == 0) { 663 dsl_dataset_t *pclone = NULL; 664 dsl_dir_t *dd; 665 dd = clone->os_dsl_dataset->ds_dir; 666 667 rw_enter(&dd->dd_pool->dp_config_rwlock, RW_READER); 668 error = dsl_dataset_hold_obj(dd->dd_pool, 669 dd->dd_phys->dd_origin_obj, FTAG, &pclone); 670 rw_exit(&dd->dd_pool->dp_config_rwlock); 671 if (error) { 672 dmu_objset_rele(clone, FTAG); 673 return (error); 674 } 675 676 error = zfs_secpolicy_write_perms(zc->zc_name, 677 ZFS_DELEG_PERM_MOUNT, cr); 678 679 dsl_dataset_name(pclone, parentname); 680 dmu_objset_rele(clone, FTAG); 681 dsl_dataset_rele(pclone, FTAG); 682 if (error == 0) 683 error = zfs_secpolicy_write_perms(parentname, 684 ZFS_DELEG_PERM_PROMOTE, cr); 685 } 686 return (error); 687 } 688 689 static int 690 zfs_secpolicy_receive(zfs_cmd_t *zc, cred_t *cr) 691 { 692 int error; 693 694 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 695 ZFS_DELEG_PERM_RECEIVE, cr)) != 0) 696 return (error); 697 698 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 699 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 700 return (error); 701 702 return (zfs_secpolicy_write_perms(zc->zc_name, 703 ZFS_DELEG_PERM_CREATE, cr)); 704 } 705 706 int 707 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr) 708 { 709 return (zfs_secpolicy_write_perms(name, 710 ZFS_DELEG_PERM_SNAPSHOT, cr)); 711 } 712 713 static int 714 zfs_secpolicy_snapshot(zfs_cmd_t *zc, cred_t *cr) 715 { 716 717 return (zfs_secpolicy_snapshot_perms(zc->zc_name, cr)); 718 } 719 720 static int 721 zfs_secpolicy_create(zfs_cmd_t *zc, cred_t *cr) 722 { 723 char parentname[MAXNAMELEN]; 724 int error; 725 726 if ((error = zfs_get_parent(zc->zc_name, parentname, 727 sizeof (parentname))) != 0) 728 return (error); 729 730 if (zc->zc_value[0] != '\0') { 731 if ((error = zfs_secpolicy_write_perms(zc->zc_value, 732 ZFS_DELEG_PERM_CLONE, cr)) != 0) 733 return (error); 734 } 735 736 if ((error = zfs_secpolicy_write_perms(parentname, 737 ZFS_DELEG_PERM_CREATE, cr)) != 0) 738 return (error); 739 740 error = zfs_secpolicy_write_perms(parentname, 741 ZFS_DELEG_PERM_MOUNT, cr); 742 743 return (error); 744 } 745 746 static int 747 zfs_secpolicy_umount(zfs_cmd_t *zc, cred_t *cr) 748 { 749 int error; 750 751 error = secpolicy_fs_unmount(cr, NULL); 752 if (error) { 753 error = dsl_deleg_access(zc->zc_name, ZFS_DELEG_PERM_MOUNT, cr); 754 } 755 return (error); 756 } 757 758 /* 759 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires 760 * SYS_CONFIG privilege, which is not available in a local zone. 761 */ 762 /* ARGSUSED */ 763 static int 764 zfs_secpolicy_config(zfs_cmd_t *zc, cred_t *cr) 765 { 766 if (secpolicy_sys_config(cr, B_FALSE) != 0) 767 return (EPERM); 768 769 return (0); 770 } 771 772 /* 773 * Policy for fault injection. Requires all privileges. 774 */ 775 /* ARGSUSED */ 776 static int 777 zfs_secpolicy_inject(zfs_cmd_t *zc, cred_t *cr) 778 { 779 return (secpolicy_zinject(cr)); 780 } 781 782 static int 783 zfs_secpolicy_inherit(zfs_cmd_t *zc, cred_t *cr) 784 { 785 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value); 786 787 if (prop == ZPROP_INVAL) { 788 if (!zfs_prop_user(zc->zc_value)) 789 return (EINVAL); 790 return (zfs_secpolicy_write_perms(zc->zc_name, 791 ZFS_DELEG_PERM_USERPROP, cr)); 792 } else { 793 return (zfs_secpolicy_setprop(zc->zc_name, prop, 794 NULL, cr)); 795 } 796 } 797 798 static int 799 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, cred_t *cr) 800 { 801 int err = zfs_secpolicy_read(zc, cr); 802 if (err) 803 return (err); 804 805 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 806 return (EINVAL); 807 808 if (zc->zc_value[0] == 0) { 809 /* 810 * They are asking about a posix uid/gid. If it's 811 * themself, allow it. 812 */ 813 if (zc->zc_objset_type == ZFS_PROP_USERUSED || 814 zc->zc_objset_type == ZFS_PROP_USERQUOTA) { 815 if (zc->zc_guid == crgetuid(cr)) 816 return (0); 817 } else { 818 if (groupmember(zc->zc_guid, cr)) 819 return (0); 820 } 821 } 822 823 return (zfs_secpolicy_write_perms(zc->zc_name, 824 userquota_perms[zc->zc_objset_type], cr)); 825 } 826 827 static int 828 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, cred_t *cr) 829 { 830 int err = zfs_secpolicy_read(zc, cr); 831 if (err) 832 return (err); 833 834 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 835 return (EINVAL); 836 837 return (zfs_secpolicy_write_perms(zc->zc_name, 838 userquota_perms[zc->zc_objset_type], cr)); 839 } 840 841 static int 842 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, cred_t *cr) 843 { 844 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION, 845 NULL, cr)); 846 } 847 848 static int 849 zfs_secpolicy_hold(zfs_cmd_t *zc, cred_t *cr) 850 { 851 return (zfs_secpolicy_write_perms(zc->zc_name, 852 ZFS_DELEG_PERM_HOLD, cr)); 853 } 854 855 static int 856 zfs_secpolicy_release(zfs_cmd_t *zc, cred_t *cr) 857 { 858 return (zfs_secpolicy_write_perms(zc->zc_name, 859 ZFS_DELEG_PERM_RELEASE, cr)); 860 } 861 862 /* 863 * Returns the nvlist as specified by the user in the zfs_cmd_t. 864 */ 865 static int 866 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp) 867 { 868 char *packed; 869 int error; 870 nvlist_t *list = NULL; 871 872 /* 873 * Read in and unpack the user-supplied nvlist. 874 */ 875 if (size == 0) 876 return (EINVAL); 877 878 packed = kmem_alloc(size, KM_SLEEP); 879 880 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size, 881 iflag)) != 0) { 882 kmem_free(packed, size); 883 return (error); 884 } 885 886 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) { 887 kmem_free(packed, size); 888 return (error); 889 } 890 891 kmem_free(packed, size); 892 893 *nvp = list; 894 return (0); 895 } 896 897 static int 898 fit_error_list(zfs_cmd_t *zc, nvlist_t **errors) 899 { 900 size_t size; 901 902 VERIFY(nvlist_size(*errors, &size, NV_ENCODE_NATIVE) == 0); 903 904 if (size > zc->zc_nvlist_dst_size) { 905 nvpair_t *more_errors; 906 int n = 0; 907 908 if (zc->zc_nvlist_dst_size < 1024) 909 return (ENOMEM); 910 911 VERIFY(nvlist_add_int32(*errors, ZPROP_N_MORE_ERRORS, 0) == 0); 912 more_errors = nvlist_prev_nvpair(*errors, NULL); 913 914 do { 915 nvpair_t *pair = nvlist_prev_nvpair(*errors, 916 more_errors); 917 VERIFY(nvlist_remove_nvpair(*errors, pair) == 0); 918 n++; 919 VERIFY(nvlist_size(*errors, &size, 920 NV_ENCODE_NATIVE) == 0); 921 } while (size > zc->zc_nvlist_dst_size); 922 923 VERIFY(nvlist_remove_nvpair(*errors, more_errors) == 0); 924 VERIFY(nvlist_add_int32(*errors, ZPROP_N_MORE_ERRORS, n) == 0); 925 ASSERT(nvlist_size(*errors, &size, NV_ENCODE_NATIVE) == 0); 926 ASSERT(size <= zc->zc_nvlist_dst_size); 927 } 928 929 return (0); 930 } 931 932 static int 933 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl) 934 { 935 char *packed = NULL; 936 size_t size; 937 int error; 938 939 VERIFY(nvlist_size(nvl, &size, NV_ENCODE_NATIVE) == 0); 940 941 if (size > zc->zc_nvlist_dst_size) { 942 error = ENOMEM; 943 } else { 944 packed = kmem_alloc(size, KM_SLEEP); 945 VERIFY(nvlist_pack(nvl, &packed, &size, NV_ENCODE_NATIVE, 946 KM_SLEEP) == 0); 947 error = ddi_copyout(packed, 948 (void *)(uintptr_t)zc->zc_nvlist_dst, size, zc->zc_iflags); 949 kmem_free(packed, size); 950 } 951 952 zc->zc_nvlist_dst_size = size; 953 return (error); 954 } 955 956 static int 957 getzfsvfs(const char *dsname, zfsvfs_t **zvp) 958 { 959 objset_t *os; 960 int error; 961 962 error = dmu_objset_hold(dsname, FTAG, &os); 963 if (error) 964 return (error); 965 if (dmu_objset_type(os) != DMU_OST_ZFS) { 966 dmu_objset_rele(os, FTAG); 967 return (EINVAL); 968 } 969 970 mutex_enter(&os->os_user_ptr_lock); 971 *zvp = dmu_objset_get_user(os); 972 if (*zvp) { 973 VFS_HOLD((*zvp)->z_vfs); 974 } else { 975 error = ESRCH; 976 } 977 mutex_exit(&os->os_user_ptr_lock); 978 dmu_objset_rele(os, FTAG); 979 return (error); 980 } 981 982 /* 983 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which 984 * case its z_vfs will be NULL, and it will be opened as the owner. 985 */ 986 static int 987 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zvp) 988 { 989 int error = 0; 990 991 if (getzfsvfs(name, zvp) != 0) 992 error = zfsvfs_create(name, zvp); 993 if (error == 0) { 994 rrw_enter(&(*zvp)->z_teardown_lock, RW_READER, tag); 995 if ((*zvp)->z_unmounted) { 996 /* 997 * XXX we could probably try again, since the unmounting 998 * thread should be just about to disassociate the 999 * objset from the zfsvfs. 1000 */ 1001 rrw_exit(&(*zvp)->z_teardown_lock, tag); 1002 return (EBUSY); 1003 } 1004 } 1005 return (error); 1006 } 1007 1008 static void 1009 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag) 1010 { 1011 rrw_exit(&zfsvfs->z_teardown_lock, tag); 1012 1013 if (zfsvfs->z_vfs) { 1014 VFS_RELE(zfsvfs->z_vfs); 1015 } else { 1016 dmu_objset_disown(zfsvfs->z_os, zfsvfs); 1017 zfsvfs_free(zfsvfs); 1018 } 1019 } 1020 1021 static int 1022 zfs_ioc_pool_create(zfs_cmd_t *zc) 1023 { 1024 int error; 1025 nvlist_t *config, *props = NULL; 1026 nvlist_t *rootprops = NULL; 1027 nvlist_t *zplprops = NULL; 1028 char *buf; 1029 1030 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1031 zc->zc_iflags, &config)) 1032 return (error); 1033 1034 if (zc->zc_nvlist_src_size != 0 && (error = 1035 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1036 zc->zc_iflags, &props))) { 1037 nvlist_free(config); 1038 return (error); 1039 } 1040 1041 if (props) { 1042 nvlist_t *nvl = NULL; 1043 uint64_t version = SPA_VERSION; 1044 1045 (void) nvlist_lookup_uint64(props, 1046 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version); 1047 if (version < SPA_VERSION_INITIAL || version > SPA_VERSION) { 1048 error = EINVAL; 1049 goto pool_props_bad; 1050 } 1051 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl); 1052 if (nvl) { 1053 error = nvlist_dup(nvl, &rootprops, KM_SLEEP); 1054 if (error != 0) { 1055 nvlist_free(config); 1056 nvlist_free(props); 1057 return (error); 1058 } 1059 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS); 1060 } 1061 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 1062 error = zfs_fill_zplprops_root(version, rootprops, 1063 zplprops, NULL); 1064 if (error) 1065 goto pool_props_bad; 1066 } 1067 1068 buf = history_str_get(zc); 1069 1070 error = spa_create(zc->zc_name, config, props, buf, zplprops); 1071 1072 /* 1073 * Set the remaining root properties 1074 */ 1075 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name, 1076 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0) 1077 (void) spa_destroy(zc->zc_name); 1078 1079 if (buf != NULL) 1080 history_str_free(buf); 1081 1082 pool_props_bad: 1083 nvlist_free(rootprops); 1084 nvlist_free(zplprops); 1085 nvlist_free(config); 1086 nvlist_free(props); 1087 1088 return (error); 1089 } 1090 1091 static int 1092 zfs_ioc_pool_destroy(zfs_cmd_t *zc) 1093 { 1094 int error; 1095 zfs_log_history(zc); 1096 error = spa_destroy(zc->zc_name); 1097 if (error == 0) 1098 zvol_remove_minors(zc->zc_name); 1099 return (error); 1100 } 1101 1102 static int 1103 zfs_ioc_pool_import(zfs_cmd_t *zc) 1104 { 1105 nvlist_t *config, *props = NULL; 1106 uint64_t guid; 1107 int error; 1108 1109 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1110 zc->zc_iflags, &config)) != 0) 1111 return (error); 1112 1113 if (zc->zc_nvlist_src_size != 0 && (error = 1114 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1115 zc->zc_iflags, &props))) { 1116 nvlist_free(config); 1117 return (error); 1118 } 1119 1120 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 || 1121 guid != zc->zc_guid) 1122 error = EINVAL; 1123 else if (zc->zc_cookie) 1124 error = spa_import_verbatim(zc->zc_name, config, props); 1125 else 1126 error = spa_import(zc->zc_name, config, props); 1127 1128 if (zc->zc_nvlist_dst != 0) 1129 (void) put_nvlist(zc, config); 1130 1131 nvlist_free(config); 1132 1133 if (props) 1134 nvlist_free(props); 1135 1136 return (error); 1137 } 1138 1139 static int 1140 zfs_ioc_pool_export(zfs_cmd_t *zc) 1141 { 1142 int error; 1143 boolean_t force = (boolean_t)zc->zc_cookie; 1144 boolean_t hardforce = (boolean_t)zc->zc_guid; 1145 1146 zfs_log_history(zc); 1147 error = spa_export(zc->zc_name, NULL, force, hardforce); 1148 if (error == 0) 1149 zvol_remove_minors(zc->zc_name); 1150 return (error); 1151 } 1152 1153 static int 1154 zfs_ioc_pool_configs(zfs_cmd_t *zc) 1155 { 1156 nvlist_t *configs; 1157 int error; 1158 1159 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL) 1160 return (EEXIST); 1161 1162 error = put_nvlist(zc, configs); 1163 1164 nvlist_free(configs); 1165 1166 return (error); 1167 } 1168 1169 static int 1170 zfs_ioc_pool_stats(zfs_cmd_t *zc) 1171 { 1172 nvlist_t *config; 1173 int error; 1174 int ret = 0; 1175 1176 error = spa_get_stats(zc->zc_name, &config, zc->zc_value, 1177 sizeof (zc->zc_value)); 1178 1179 if (config != NULL) { 1180 ret = put_nvlist(zc, config); 1181 nvlist_free(config); 1182 1183 /* 1184 * The config may be present even if 'error' is non-zero. 1185 * In this case we return success, and preserve the real errno 1186 * in 'zc_cookie'. 1187 */ 1188 zc->zc_cookie = error; 1189 } else { 1190 ret = error; 1191 } 1192 1193 return (ret); 1194 } 1195 1196 /* 1197 * Try to import the given pool, returning pool stats as appropriate so that 1198 * user land knows which devices are available and overall pool health. 1199 */ 1200 static int 1201 zfs_ioc_pool_tryimport(zfs_cmd_t *zc) 1202 { 1203 nvlist_t *tryconfig, *config; 1204 int error; 1205 1206 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1207 zc->zc_iflags, &tryconfig)) != 0) 1208 return (error); 1209 1210 config = spa_tryimport(tryconfig); 1211 1212 nvlist_free(tryconfig); 1213 1214 if (config == NULL) 1215 return (EINVAL); 1216 1217 error = put_nvlist(zc, config); 1218 nvlist_free(config); 1219 1220 return (error); 1221 } 1222 1223 static int 1224 zfs_ioc_pool_scrub(zfs_cmd_t *zc) 1225 { 1226 spa_t *spa; 1227 int error; 1228 1229 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1230 return (error); 1231 1232 error = spa_scrub(spa, zc->zc_cookie); 1233 1234 spa_close(spa, FTAG); 1235 1236 return (error); 1237 } 1238 1239 static int 1240 zfs_ioc_pool_freeze(zfs_cmd_t *zc) 1241 { 1242 spa_t *spa; 1243 int error; 1244 1245 error = spa_open(zc->zc_name, &spa, FTAG); 1246 if (error == 0) { 1247 spa_freeze(spa); 1248 spa_close(spa, FTAG); 1249 } 1250 return (error); 1251 } 1252 1253 static int 1254 zfs_ioc_pool_upgrade(zfs_cmd_t *zc) 1255 { 1256 spa_t *spa; 1257 int error; 1258 1259 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1260 return (error); 1261 1262 if (zc->zc_cookie < spa_version(spa) || zc->zc_cookie > SPA_VERSION) { 1263 spa_close(spa, FTAG); 1264 return (EINVAL); 1265 } 1266 1267 spa_upgrade(spa, zc->zc_cookie); 1268 spa_close(spa, FTAG); 1269 1270 return (error); 1271 } 1272 1273 static int 1274 zfs_ioc_pool_get_history(zfs_cmd_t *zc) 1275 { 1276 spa_t *spa; 1277 char *hist_buf; 1278 uint64_t size; 1279 int error; 1280 1281 if ((size = zc->zc_history_len) == 0) 1282 return (EINVAL); 1283 1284 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1285 return (error); 1286 1287 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) { 1288 spa_close(spa, FTAG); 1289 return (ENOTSUP); 1290 } 1291 1292 hist_buf = kmem_alloc(size, KM_SLEEP); 1293 if ((error = spa_history_get(spa, &zc->zc_history_offset, 1294 &zc->zc_history_len, hist_buf)) == 0) { 1295 error = ddi_copyout(hist_buf, 1296 (void *)(uintptr_t)zc->zc_history, 1297 zc->zc_history_len, zc->zc_iflags); 1298 } 1299 1300 spa_close(spa, FTAG); 1301 kmem_free(hist_buf, size); 1302 return (error); 1303 } 1304 1305 static int 1306 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc) 1307 { 1308 int error; 1309 1310 if (error = dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value)) 1311 return (error); 1312 1313 return (0); 1314 } 1315 1316 /* 1317 * inputs: 1318 * zc_name name of filesystem 1319 * zc_obj object to find 1320 * 1321 * outputs: 1322 * zc_value name of object 1323 */ 1324 static int 1325 zfs_ioc_obj_to_path(zfs_cmd_t *zc) 1326 { 1327 objset_t *os; 1328 int error; 1329 1330 /* XXX reading from objset not owned */ 1331 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0) 1332 return (error); 1333 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1334 dmu_objset_rele(os, FTAG); 1335 return (EINVAL); 1336 } 1337 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value, 1338 sizeof (zc->zc_value)); 1339 dmu_objset_rele(os, FTAG); 1340 1341 return (error); 1342 } 1343 1344 static int 1345 zfs_ioc_vdev_add(zfs_cmd_t *zc) 1346 { 1347 spa_t *spa; 1348 int error; 1349 nvlist_t *config, **l2cache, **spares; 1350 uint_t nl2cache = 0, nspares = 0; 1351 1352 error = spa_open(zc->zc_name, &spa, FTAG); 1353 if (error != 0) 1354 return (error); 1355 1356 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1357 zc->zc_iflags, &config); 1358 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE, 1359 &l2cache, &nl2cache); 1360 1361 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES, 1362 &spares, &nspares); 1363 1364 /* 1365 * A root pool with concatenated devices is not supported. 1366 * Thus, can not add a device to a root pool. 1367 * 1368 * Intent log device can not be added to a rootpool because 1369 * during mountroot, zil is replayed, a seperated log device 1370 * can not be accessed during the mountroot time. 1371 * 1372 * l2cache and spare devices are ok to be added to a rootpool. 1373 */ 1374 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) { 1375 spa_close(spa, FTAG); 1376 return (EDOM); 1377 } 1378 1379 if (error == 0) { 1380 error = spa_vdev_add(spa, config); 1381 nvlist_free(config); 1382 } 1383 spa_close(spa, FTAG); 1384 return (error); 1385 } 1386 1387 static int 1388 zfs_ioc_vdev_remove(zfs_cmd_t *zc) 1389 { 1390 spa_t *spa; 1391 int error; 1392 1393 error = spa_open(zc->zc_name, &spa, FTAG); 1394 if (error != 0) 1395 return (error); 1396 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE); 1397 spa_close(spa, FTAG); 1398 return (error); 1399 } 1400 1401 static int 1402 zfs_ioc_vdev_set_state(zfs_cmd_t *zc) 1403 { 1404 spa_t *spa; 1405 int error; 1406 vdev_state_t newstate = VDEV_STATE_UNKNOWN; 1407 1408 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1409 return (error); 1410 switch (zc->zc_cookie) { 1411 case VDEV_STATE_ONLINE: 1412 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate); 1413 break; 1414 1415 case VDEV_STATE_OFFLINE: 1416 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj); 1417 break; 1418 1419 case VDEV_STATE_FAULTED: 1420 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED && 1421 zc->zc_obj != VDEV_AUX_EXTERNAL) 1422 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED; 1423 1424 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj); 1425 break; 1426 1427 case VDEV_STATE_DEGRADED: 1428 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED && 1429 zc->zc_obj != VDEV_AUX_EXTERNAL) 1430 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED; 1431 1432 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj); 1433 break; 1434 1435 default: 1436 error = EINVAL; 1437 } 1438 zc->zc_cookie = newstate; 1439 spa_close(spa, FTAG); 1440 return (error); 1441 } 1442 1443 static int 1444 zfs_ioc_vdev_attach(zfs_cmd_t *zc) 1445 { 1446 spa_t *spa; 1447 int replacing = zc->zc_cookie; 1448 nvlist_t *config; 1449 int error; 1450 1451 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1452 return (error); 1453 1454 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1455 zc->zc_iflags, &config)) == 0) { 1456 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing); 1457 nvlist_free(config); 1458 } 1459 1460 spa_close(spa, FTAG); 1461 return (error); 1462 } 1463 1464 static int 1465 zfs_ioc_vdev_detach(zfs_cmd_t *zc) 1466 { 1467 spa_t *spa; 1468 int error; 1469 1470 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1471 return (error); 1472 1473 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE); 1474 1475 spa_close(spa, FTAG); 1476 return (error); 1477 } 1478 1479 static int 1480 zfs_ioc_vdev_setpath(zfs_cmd_t *zc) 1481 { 1482 spa_t *spa; 1483 char *path = zc->zc_value; 1484 uint64_t guid = zc->zc_guid; 1485 int error; 1486 1487 error = spa_open(zc->zc_name, &spa, FTAG); 1488 if (error != 0) 1489 return (error); 1490 1491 error = spa_vdev_setpath(spa, guid, path); 1492 spa_close(spa, FTAG); 1493 return (error); 1494 } 1495 1496 static int 1497 zfs_ioc_vdev_setfru(zfs_cmd_t *zc) 1498 { 1499 spa_t *spa; 1500 char *fru = zc->zc_value; 1501 uint64_t guid = zc->zc_guid; 1502 int error; 1503 1504 error = spa_open(zc->zc_name, &spa, FTAG); 1505 if (error != 0) 1506 return (error); 1507 1508 error = spa_vdev_setfru(spa, guid, fru); 1509 spa_close(spa, FTAG); 1510 return (error); 1511 } 1512 1513 /* 1514 * inputs: 1515 * zc_name name of filesystem 1516 * zc_nvlist_dst_size size of buffer for property nvlist 1517 * 1518 * outputs: 1519 * zc_objset_stats stats 1520 * zc_nvlist_dst property nvlist 1521 * zc_nvlist_dst_size size of property nvlist 1522 */ 1523 static int 1524 zfs_ioc_objset_stats(zfs_cmd_t *zc) 1525 { 1526 objset_t *os = NULL; 1527 int error; 1528 nvlist_t *nv; 1529 1530 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) 1531 return (error); 1532 1533 dmu_objset_fast_stat(os, &zc->zc_objset_stats); 1534 1535 if (zc->zc_nvlist_dst != 0 && 1536 (error = dsl_prop_get_all(os, &nv)) == 0) { 1537 dmu_objset_stats(os, nv); 1538 /* 1539 * NB: zvol_get_stats() will read the objset contents, 1540 * which we aren't supposed to do with a 1541 * DS_MODE_USER hold, because it could be 1542 * inconsistent. So this is a bit of a workaround... 1543 * XXX reading with out owning 1544 */ 1545 if (!zc->zc_objset_stats.dds_inconsistent) { 1546 if (dmu_objset_type(os) == DMU_OST_ZVOL) 1547 VERIFY(zvol_get_stats(os, nv) == 0); 1548 } 1549 error = put_nvlist(zc, nv); 1550 nvlist_free(nv); 1551 } 1552 1553 dmu_objset_rele(os, FTAG); 1554 return (error); 1555 } 1556 1557 /* 1558 * inputs: 1559 * zc_name name of filesystem 1560 * zc_nvlist_dst_size size of buffer for property nvlist 1561 * 1562 * outputs: 1563 * zc_nvlist_dst received property nvlist 1564 * zc_nvlist_dst_size size of received property nvlist 1565 * 1566 * Gets received properties (distinct from local properties on or after 1567 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from 1568 * local property values. 1569 */ 1570 static int 1571 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc) 1572 { 1573 objset_t *os = NULL; 1574 int error; 1575 nvlist_t *nv; 1576 1577 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) 1578 return (error); 1579 1580 /* 1581 * Without this check, we would return local property values if the 1582 * caller has not already received properties on or after 1583 * SPA_VERSION_RECVD_PROPS. 1584 */ 1585 if (!dsl_prop_get_hasrecvd(os)) { 1586 dmu_objset_rele(os, FTAG); 1587 return (ENOTSUP); 1588 } 1589 1590 if (zc->zc_nvlist_dst != 0 && 1591 (error = dsl_prop_get_received(os, &nv)) == 0) { 1592 error = put_nvlist(zc, nv); 1593 nvlist_free(nv); 1594 } 1595 1596 dmu_objset_rele(os, FTAG); 1597 return (error); 1598 } 1599 1600 static int 1601 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop) 1602 { 1603 uint64_t value; 1604 int error; 1605 1606 /* 1607 * zfs_get_zplprop() will either find a value or give us 1608 * the default value (if there is one). 1609 */ 1610 if ((error = zfs_get_zplprop(os, prop, &value)) != 0) 1611 return (error); 1612 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0); 1613 return (0); 1614 } 1615 1616 /* 1617 * inputs: 1618 * zc_name name of filesystem 1619 * zc_nvlist_dst_size size of buffer for zpl property nvlist 1620 * 1621 * outputs: 1622 * zc_nvlist_dst zpl property nvlist 1623 * zc_nvlist_dst_size size of zpl property nvlist 1624 */ 1625 static int 1626 zfs_ioc_objset_zplprops(zfs_cmd_t *zc) 1627 { 1628 objset_t *os; 1629 int err; 1630 1631 /* XXX reading without owning */ 1632 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os)) 1633 return (err); 1634 1635 dmu_objset_fast_stat(os, &zc->zc_objset_stats); 1636 1637 /* 1638 * NB: nvl_add_zplprop() will read the objset contents, 1639 * which we aren't supposed to do with a DS_MODE_USER 1640 * hold, because it could be inconsistent. 1641 */ 1642 if (zc->zc_nvlist_dst != NULL && 1643 !zc->zc_objset_stats.dds_inconsistent && 1644 dmu_objset_type(os) == DMU_OST_ZFS) { 1645 nvlist_t *nv; 1646 1647 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0); 1648 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 && 1649 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 && 1650 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 && 1651 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0) 1652 err = put_nvlist(zc, nv); 1653 nvlist_free(nv); 1654 } else { 1655 err = ENOENT; 1656 } 1657 dmu_objset_rele(os, FTAG); 1658 return (err); 1659 } 1660 1661 static boolean_t 1662 dataset_name_hidden(const char *name) 1663 { 1664 /* 1665 * Skip over datasets that are not visible in this zone, 1666 * internal datasets (which have a $ in their name), and 1667 * temporary datasets (which have a % in their name). 1668 */ 1669 if (strchr(name, '$') != NULL) 1670 return (B_TRUE); 1671 if (strchr(name, '%') != NULL) 1672 return (B_TRUE); 1673 if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL)) 1674 return (B_TRUE); 1675 return (B_FALSE); 1676 } 1677 1678 /* 1679 * inputs: 1680 * zc_name name of filesystem 1681 * zc_cookie zap cursor 1682 * zc_nvlist_dst_size size of buffer for property nvlist 1683 * 1684 * outputs: 1685 * zc_name name of next filesystem 1686 * zc_cookie zap cursor 1687 * zc_objset_stats stats 1688 * zc_nvlist_dst property nvlist 1689 * zc_nvlist_dst_size size of property nvlist 1690 */ 1691 static int 1692 zfs_ioc_dataset_list_next(zfs_cmd_t *zc) 1693 { 1694 objset_t *os; 1695 int error; 1696 char *p; 1697 1698 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) { 1699 if (error == ENOENT) 1700 error = ESRCH; 1701 return (error); 1702 } 1703 1704 p = strrchr(zc->zc_name, '/'); 1705 if (p == NULL || p[1] != '\0') 1706 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name)); 1707 p = zc->zc_name + strlen(zc->zc_name); 1708 1709 /* 1710 * Pre-fetch the datasets. dmu_objset_prefetch() always returns 0 1711 * but is not declared void because its called by dmu_objset_find(). 1712 */ 1713 if (zc->zc_cookie == 0) { 1714 uint64_t cookie = 0; 1715 int len = sizeof (zc->zc_name) - (p - zc->zc_name); 1716 1717 while (dmu_dir_list_next(os, len, p, NULL, &cookie) == 0) 1718 (void) dmu_objset_prefetch(p, NULL); 1719 } 1720 1721 do { 1722 error = dmu_dir_list_next(os, 1723 sizeof (zc->zc_name) - (p - zc->zc_name), p, 1724 NULL, &zc->zc_cookie); 1725 if (error == ENOENT) 1726 error = ESRCH; 1727 } while (error == 0 && dataset_name_hidden(zc->zc_name) && 1728 !(zc->zc_iflags & FKIOCTL)); 1729 dmu_objset_rele(os, FTAG); 1730 1731 /* 1732 * If it's an internal dataset (ie. with a '$' in its name), 1733 * don't try to get stats for it, otherwise we'll return ENOENT. 1734 */ 1735 if (error == 0 && strchr(zc->zc_name, '$') == NULL) 1736 error = zfs_ioc_objset_stats(zc); /* fill in the stats */ 1737 return (error); 1738 } 1739 1740 /* 1741 * inputs: 1742 * zc_name name of filesystem 1743 * zc_cookie zap cursor 1744 * zc_nvlist_dst_size size of buffer for property nvlist 1745 * 1746 * outputs: 1747 * zc_name name of next snapshot 1748 * zc_objset_stats stats 1749 * zc_nvlist_dst property nvlist 1750 * zc_nvlist_dst_size size of property nvlist 1751 */ 1752 static int 1753 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc) 1754 { 1755 objset_t *os; 1756 int error; 1757 1758 if (zc->zc_cookie == 0) 1759 (void) dmu_objset_find(zc->zc_name, dmu_objset_prefetch, 1760 NULL, DS_FIND_SNAPSHOTS); 1761 1762 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 1763 if (error) 1764 return (error == ENOENT ? ESRCH : error); 1765 1766 /* 1767 * A dataset name of maximum length cannot have any snapshots, 1768 * so exit immediately. 1769 */ 1770 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) { 1771 dmu_objset_rele(os, FTAG); 1772 return (ESRCH); 1773 } 1774 1775 error = dmu_snapshot_list_next(os, 1776 sizeof (zc->zc_name) - strlen(zc->zc_name), 1777 zc->zc_name + strlen(zc->zc_name), NULL, &zc->zc_cookie, NULL); 1778 dmu_objset_rele(os, FTAG); 1779 if (error == 0) 1780 error = zfs_ioc_objset_stats(zc); /* fill in the stats */ 1781 else if (error == ENOENT) 1782 error = ESRCH; 1783 1784 /* if we failed, undo the @ that we tacked on to zc_name */ 1785 if (error) 1786 *strchr(zc->zc_name, '@') = '\0'; 1787 return (error); 1788 } 1789 1790 static int 1791 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair) 1792 { 1793 const char *propname = nvpair_name(pair); 1794 uint64_t *valary; 1795 unsigned int vallen; 1796 const char *domain; 1797 zfs_userquota_prop_t type; 1798 uint64_t rid; 1799 uint64_t quota; 1800 zfsvfs_t *zfsvfs; 1801 int err; 1802 1803 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 1804 nvlist_t *attrs; 1805 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 1806 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 1807 &pair) == 0); 1808 } 1809 1810 VERIFY(nvpair_value_uint64_array(pair, &valary, &vallen) == 0); 1811 VERIFY(vallen == 3); 1812 type = valary[0]; 1813 rid = valary[1]; 1814 quota = valary[2]; 1815 /* 1816 * The propname is encoded as 1817 * userquota@<rid>-<domain>. 1818 */ 1819 domain = strchr(propname, '-') + 1; 1820 1821 err = zfsvfs_hold(dsname, FTAG, &zfsvfs); 1822 if (err == 0) { 1823 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota); 1824 zfsvfs_rele(zfsvfs, FTAG); 1825 } 1826 1827 return (err); 1828 } 1829 1830 /* 1831 * If the named property is one that has a special function to set its value, 1832 * return 0 on success and a positive error code on failure; otherwise if it is 1833 * not one of the special properties handled by this function, return -1. 1834 * 1835 * XXX: It would be better for callers of the properety interface if we handled 1836 * these special cases in dsl_prop.c (in the dsl layer). 1837 */ 1838 static int 1839 zfs_prop_set_special(const char *dsname, zprop_source_t source, 1840 nvpair_t *pair) 1841 { 1842 const char *propname = nvpair_name(pair); 1843 zfs_prop_t prop = zfs_name_to_prop(propname); 1844 uint64_t intval; 1845 int err; 1846 1847 if (prop == ZPROP_INVAL) { 1848 if (zfs_prop_userquota(propname)) 1849 return (zfs_prop_set_userquota(dsname, pair)); 1850 return (-1); 1851 } 1852 1853 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 1854 nvlist_t *attrs; 1855 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 1856 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 1857 &pair) == 0); 1858 } 1859 1860 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING) 1861 return (-1); 1862 1863 VERIFY(0 == nvpair_value_uint64(pair, &intval)); 1864 1865 switch (prop) { 1866 case ZFS_PROP_QUOTA: 1867 err = dsl_dir_set_quota(dsname, source, intval); 1868 break; 1869 case ZFS_PROP_REFQUOTA: 1870 err = dsl_dataset_set_quota(dsname, source, intval); 1871 break; 1872 case ZFS_PROP_RESERVATION: 1873 err = dsl_dir_set_reservation(dsname, source, intval); 1874 break; 1875 case ZFS_PROP_REFRESERVATION: 1876 err = dsl_dataset_set_reservation(dsname, source, intval); 1877 break; 1878 case ZFS_PROP_VOLSIZE: 1879 err = zvol_set_volsize(dsname, ddi_driver_major(zfs_dip), 1880 intval); 1881 break; 1882 case ZFS_PROP_VERSION: 1883 { 1884 zfsvfs_t *zfsvfs; 1885 1886 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs)) != 0) 1887 break; 1888 1889 err = zfs_set_version(zfsvfs, intval); 1890 zfsvfs_rele(zfsvfs, FTAG); 1891 1892 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) { 1893 zfs_cmd_t *zc; 1894 1895 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP); 1896 (void) strcpy(zc->zc_name, dsname); 1897 (void) zfs_ioc_userspace_upgrade(zc); 1898 kmem_free(zc, sizeof (zfs_cmd_t)); 1899 } 1900 break; 1901 } 1902 1903 default: 1904 err = -1; 1905 } 1906 1907 return (err); 1908 } 1909 1910 /* 1911 * This function is best effort. If it fails to set any of the given properties, 1912 * it continues to set as many as it can and returns the first error 1913 * encountered. If the caller provides a non-NULL errlist, it also gives the 1914 * complete list of names of all the properties it failed to set along with the 1915 * corresponding error numbers. The caller is responsible for freeing the 1916 * returned errlist. 1917 * 1918 * If every property is set successfully, zero is returned and the list pointed 1919 * at by errlist is NULL. 1920 */ 1921 int 1922 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl, 1923 nvlist_t **errlist) 1924 { 1925 nvpair_t *pair; 1926 nvpair_t *propval; 1927 int err = 0; 1928 int rv = 0; 1929 uint64_t intval; 1930 char *strval; 1931 nvlist_t *genericnvl; 1932 nvlist_t *errors; 1933 nvlist_t *retrynvl; 1934 1935 VERIFY(nvlist_alloc(&genericnvl, NV_UNIQUE_NAME, KM_SLEEP) == 0); 1936 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0); 1937 VERIFY(nvlist_alloc(&retrynvl, NV_UNIQUE_NAME, KM_SLEEP) == 0); 1938 1939 retry: 1940 pair = NULL; 1941 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { 1942 const char *propname = nvpair_name(pair); 1943 zfs_prop_t prop = zfs_name_to_prop(propname); 1944 1945 /* decode the property value */ 1946 propval = pair; 1947 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 1948 nvlist_t *attrs; 1949 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 1950 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 1951 &propval) == 0); 1952 } 1953 1954 /* Validate value type */ 1955 if (prop == ZPROP_INVAL) { 1956 if (zfs_prop_user(propname)) { 1957 if (nvpair_type(propval) != DATA_TYPE_STRING) 1958 err = EINVAL; 1959 } else if (zfs_prop_userquota(propname)) { 1960 if (nvpair_type(propval) != 1961 DATA_TYPE_UINT64_ARRAY) 1962 err = EINVAL; 1963 } 1964 } else { 1965 if (nvpair_type(propval) == DATA_TYPE_STRING) { 1966 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING) 1967 err = EINVAL; 1968 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) { 1969 const char *unused; 1970 1971 VERIFY(nvpair_value_uint64(propval, 1972 &intval) == 0); 1973 1974 switch (zfs_prop_get_type(prop)) { 1975 case PROP_TYPE_NUMBER: 1976 break; 1977 case PROP_TYPE_STRING: 1978 err = EINVAL; 1979 break; 1980 case PROP_TYPE_INDEX: 1981 if (zfs_prop_index_to_string(prop, 1982 intval, &unused) != 0) 1983 err = EINVAL; 1984 break; 1985 default: 1986 cmn_err(CE_PANIC, 1987 "unknown property type"); 1988 } 1989 } else { 1990 err = EINVAL; 1991 } 1992 } 1993 1994 /* Validate permissions */ 1995 if (err == 0) 1996 err = zfs_check_settable(dsname, pair, CRED()); 1997 1998 if (err == 0) { 1999 err = zfs_prop_set_special(dsname, source, pair); 2000 if (err == -1) { 2001 /* 2002 * For better performance we build up a list of 2003 * properties to set in a single transaction. 2004 */ 2005 err = nvlist_add_nvpair(genericnvl, pair); 2006 } else if (err != 0 && nvl != retrynvl) { 2007 /* 2008 * This may be a spurious error caused by 2009 * receiving quota and reservation out of order. 2010 * Try again in a second pass. 2011 */ 2012 err = nvlist_add_nvpair(retrynvl, pair); 2013 } 2014 } 2015 2016 if (err != 0) 2017 VERIFY(nvlist_add_int32(errors, propname, err) == 0); 2018 } 2019 2020 if (nvl != retrynvl && !nvlist_empty(retrynvl)) { 2021 nvl = retrynvl; 2022 goto retry; 2023 } 2024 2025 if (!nvlist_empty(genericnvl) && 2026 dsl_props_set(dsname, source, genericnvl) != 0) { 2027 /* 2028 * If this fails, we still want to set as many properties as we 2029 * can, so try setting them individually. 2030 */ 2031 pair = NULL; 2032 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) { 2033 const char *propname = nvpair_name(pair); 2034 2035 propval = pair; 2036 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2037 nvlist_t *attrs; 2038 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 2039 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2040 &propval) == 0); 2041 } 2042 2043 if (nvpair_type(propval) == DATA_TYPE_STRING) { 2044 VERIFY(nvpair_value_string(propval, 2045 &strval) == 0); 2046 err = dsl_prop_set(dsname, propname, source, 1, 2047 strlen(strval) + 1, strval); 2048 } else { 2049 VERIFY(nvpair_value_uint64(propval, 2050 &intval) == 0); 2051 err = dsl_prop_set(dsname, propname, source, 8, 2052 1, &intval); 2053 } 2054 2055 if (err != 0) { 2056 VERIFY(nvlist_add_int32(errors, propname, 2057 err) == 0); 2058 } 2059 } 2060 } 2061 nvlist_free(genericnvl); 2062 nvlist_free(retrynvl); 2063 2064 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) { 2065 nvlist_free(errors); 2066 errors = NULL; 2067 } else { 2068 VERIFY(nvpair_value_int32(pair, &rv) == 0); 2069 } 2070 2071 if (errlist == NULL) 2072 nvlist_free(errors); 2073 else 2074 *errlist = errors; 2075 2076 return (rv); 2077 } 2078 2079 /* 2080 * Check that all the properties are valid user properties. 2081 */ 2082 static int 2083 zfs_check_userprops(char *fsname, nvlist_t *nvl) 2084 { 2085 nvpair_t *pair = NULL; 2086 int error = 0; 2087 2088 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { 2089 const char *propname = nvpair_name(pair); 2090 char *valstr; 2091 2092 if (!zfs_prop_user(propname) || 2093 nvpair_type(pair) != DATA_TYPE_STRING) 2094 return (EINVAL); 2095 2096 if (error = zfs_secpolicy_write_perms(fsname, 2097 ZFS_DELEG_PERM_USERPROP, CRED())) 2098 return (error); 2099 2100 if (strlen(propname) >= ZAP_MAXNAMELEN) 2101 return (ENAMETOOLONG); 2102 2103 VERIFY(nvpair_value_string(pair, &valstr) == 0); 2104 if (strlen(valstr) >= ZAP_MAXVALUELEN) 2105 return (E2BIG); 2106 } 2107 return (0); 2108 } 2109 2110 static void 2111 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops) 2112 { 2113 nvpair_t *pair; 2114 2115 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2116 2117 pair = NULL; 2118 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) { 2119 if (nvlist_exists(skipped, nvpair_name(pair))) 2120 continue; 2121 2122 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0); 2123 } 2124 } 2125 2126 static int 2127 clear_received_props(objset_t *os, const char *fs, nvlist_t *props, 2128 nvlist_t *skipped) 2129 { 2130 int err = 0; 2131 nvlist_t *cleared_props = NULL; 2132 props_skip(props, skipped, &cleared_props); 2133 if (!nvlist_empty(cleared_props)) { 2134 /* 2135 * Acts on local properties until the dataset has received 2136 * properties at least once on or after SPA_VERSION_RECVD_PROPS. 2137 */ 2138 zprop_source_t flags = (ZPROP_SRC_NONE | 2139 (dsl_prop_get_hasrecvd(os) ? ZPROP_SRC_RECEIVED : 0)); 2140 err = zfs_set_prop_nvlist(fs, flags, cleared_props, NULL); 2141 } 2142 nvlist_free(cleared_props); 2143 return (err); 2144 } 2145 2146 /* 2147 * inputs: 2148 * zc_name name of filesystem 2149 * zc_value name of property to set 2150 * zc_nvlist_src{_size} nvlist of properties to apply 2151 * zc_cookie received properties flag 2152 * 2153 * outputs: 2154 * zc_nvlist_dst{_size} error for each unapplied received property 2155 */ 2156 static int 2157 zfs_ioc_set_prop(zfs_cmd_t *zc) 2158 { 2159 nvlist_t *nvl; 2160 boolean_t received = zc->zc_cookie; 2161 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED : 2162 ZPROP_SRC_LOCAL); 2163 nvlist_t *errors = NULL; 2164 int error; 2165 2166 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2167 zc->zc_iflags, &nvl)) != 0) 2168 return (error); 2169 2170 if (received) { 2171 nvlist_t *origprops; 2172 objset_t *os; 2173 2174 if (dmu_objset_hold(zc->zc_name, FTAG, &os) == 0) { 2175 if (dsl_prop_get_received(os, &origprops) == 0) { 2176 (void) clear_received_props(os, 2177 zc->zc_name, origprops, nvl); 2178 nvlist_free(origprops); 2179 } 2180 2181 dsl_prop_set_hasrecvd(os); 2182 dmu_objset_rele(os, FTAG); 2183 } 2184 } 2185 2186 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, &errors); 2187 2188 if (zc->zc_nvlist_dst != NULL && errors != NULL) { 2189 (void) put_nvlist(zc, errors); 2190 } 2191 2192 nvlist_free(errors); 2193 nvlist_free(nvl); 2194 return (error); 2195 } 2196 2197 /* 2198 * inputs: 2199 * zc_name name of filesystem 2200 * zc_value name of property to inherit 2201 * zc_cookie revert to received value if TRUE 2202 * 2203 * outputs: none 2204 */ 2205 static int 2206 zfs_ioc_inherit_prop(zfs_cmd_t *zc) 2207 { 2208 const char *propname = zc->zc_value; 2209 zfs_prop_t prop = zfs_name_to_prop(propname); 2210 boolean_t received = zc->zc_cookie; 2211 zprop_source_t source = (received 2212 ? ZPROP_SRC_NONE /* revert to received value, if any */ 2213 : ZPROP_SRC_INHERITED); /* explicitly inherit */ 2214 2215 if (received) { 2216 nvlist_t *dummy; 2217 nvpair_t *pair; 2218 zprop_type_t type; 2219 int err; 2220 2221 /* 2222 * zfs_prop_set_special() expects properties in the form of an 2223 * nvpair with type info. 2224 */ 2225 if (prop == ZPROP_INVAL) { 2226 if (!zfs_prop_user(propname)) 2227 return (EINVAL); 2228 2229 type = PROP_TYPE_STRING; 2230 } else { 2231 type = zfs_prop_get_type(prop); 2232 } 2233 2234 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2235 2236 switch (type) { 2237 case PROP_TYPE_STRING: 2238 VERIFY(0 == nvlist_add_string(dummy, propname, "")); 2239 break; 2240 case PROP_TYPE_NUMBER: 2241 case PROP_TYPE_INDEX: 2242 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0)); 2243 break; 2244 default: 2245 nvlist_free(dummy); 2246 return (EINVAL); 2247 } 2248 2249 pair = nvlist_next_nvpair(dummy, NULL); 2250 err = zfs_prop_set_special(zc->zc_name, source, pair); 2251 nvlist_free(dummy); 2252 if (err != -1) 2253 return (err); /* special property already handled */ 2254 } else { 2255 /* 2256 * Only check this in the non-received case. We want to allow 2257 * 'inherit -S' to revert non-inheritable properties like quota 2258 * and reservation to the received or default values even though 2259 * they are not considered inheritable. 2260 */ 2261 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop)) 2262 return (EINVAL); 2263 } 2264 2265 /* the property name has been validated by zfs_secpolicy_inherit() */ 2266 return (dsl_prop_set(zc->zc_name, zc->zc_value, source, 0, 0, NULL)); 2267 } 2268 2269 static int 2270 zfs_ioc_pool_set_props(zfs_cmd_t *zc) 2271 { 2272 nvlist_t *props; 2273 spa_t *spa; 2274 int error; 2275 nvpair_t *pair; 2276 2277 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2278 zc->zc_iflags, &props)) 2279 return (error); 2280 2281 /* 2282 * If the only property is the configfile, then just do a spa_lookup() 2283 * to handle the faulted case. 2284 */ 2285 pair = nvlist_next_nvpair(props, NULL); 2286 if (pair != NULL && strcmp(nvpair_name(pair), 2287 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 && 2288 nvlist_next_nvpair(props, pair) == NULL) { 2289 mutex_enter(&spa_namespace_lock); 2290 if ((spa = spa_lookup(zc->zc_name)) != NULL) { 2291 spa_configfile_set(spa, props, B_FALSE); 2292 spa_config_sync(spa, B_FALSE, B_TRUE); 2293 } 2294 mutex_exit(&spa_namespace_lock); 2295 if (spa != NULL) { 2296 nvlist_free(props); 2297 return (0); 2298 } 2299 } 2300 2301 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { 2302 nvlist_free(props); 2303 return (error); 2304 } 2305 2306 error = spa_prop_set(spa, props); 2307 2308 nvlist_free(props); 2309 spa_close(spa, FTAG); 2310 2311 return (error); 2312 } 2313 2314 static int 2315 zfs_ioc_pool_get_props(zfs_cmd_t *zc) 2316 { 2317 spa_t *spa; 2318 int error; 2319 nvlist_t *nvp = NULL; 2320 2321 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { 2322 /* 2323 * If the pool is faulted, there may be properties we can still 2324 * get (such as altroot and cachefile), so attempt to get them 2325 * anyway. 2326 */ 2327 mutex_enter(&spa_namespace_lock); 2328 if ((spa = spa_lookup(zc->zc_name)) != NULL) 2329 error = spa_prop_get(spa, &nvp); 2330 mutex_exit(&spa_namespace_lock); 2331 } else { 2332 error = spa_prop_get(spa, &nvp); 2333 spa_close(spa, FTAG); 2334 } 2335 2336 if (error == 0 && zc->zc_nvlist_dst != NULL) 2337 error = put_nvlist(zc, nvp); 2338 else 2339 error = EFAULT; 2340 2341 nvlist_free(nvp); 2342 return (error); 2343 } 2344 2345 static int 2346 zfs_ioc_iscsi_perm_check(zfs_cmd_t *zc) 2347 { 2348 nvlist_t *nvp; 2349 int error; 2350 uint32_t uid; 2351 uint32_t gid; 2352 uint32_t *groups; 2353 uint_t group_cnt; 2354 cred_t *usercred; 2355 2356 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2357 zc->zc_iflags, &nvp)) != 0) { 2358 return (error); 2359 } 2360 2361 if ((error = nvlist_lookup_uint32(nvp, 2362 ZFS_DELEG_PERM_UID, &uid)) != 0) { 2363 nvlist_free(nvp); 2364 return (EPERM); 2365 } 2366 2367 if ((error = nvlist_lookup_uint32(nvp, 2368 ZFS_DELEG_PERM_GID, &gid)) != 0) { 2369 nvlist_free(nvp); 2370 return (EPERM); 2371 } 2372 2373 if ((error = nvlist_lookup_uint32_array(nvp, ZFS_DELEG_PERM_GROUPS, 2374 &groups, &group_cnt)) != 0) { 2375 nvlist_free(nvp); 2376 return (EPERM); 2377 } 2378 usercred = cralloc(); 2379 if ((crsetugid(usercred, uid, gid) != 0) || 2380 (crsetgroups(usercred, group_cnt, (gid_t *)groups) != 0)) { 2381 nvlist_free(nvp); 2382 crfree(usercred); 2383 return (EPERM); 2384 } 2385 nvlist_free(nvp); 2386 error = dsl_deleg_access(zc->zc_name, 2387 zfs_prop_to_name(ZFS_PROP_SHAREISCSI), usercred); 2388 crfree(usercred); 2389 return (error); 2390 } 2391 2392 /* 2393 * inputs: 2394 * zc_name name of filesystem 2395 * zc_nvlist_src{_size} nvlist of delegated permissions 2396 * zc_perm_action allow/unallow flag 2397 * 2398 * outputs: none 2399 */ 2400 static int 2401 zfs_ioc_set_fsacl(zfs_cmd_t *zc) 2402 { 2403 int error; 2404 nvlist_t *fsaclnv = NULL; 2405 2406 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2407 zc->zc_iflags, &fsaclnv)) != 0) 2408 return (error); 2409 2410 /* 2411 * Verify nvlist is constructed correctly 2412 */ 2413 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) { 2414 nvlist_free(fsaclnv); 2415 return (EINVAL); 2416 } 2417 2418 /* 2419 * If we don't have PRIV_SYS_MOUNT, then validate 2420 * that user is allowed to hand out each permission in 2421 * the nvlist(s) 2422 */ 2423 2424 error = secpolicy_zfs(CRED()); 2425 if (error) { 2426 if (zc->zc_perm_action == B_FALSE) { 2427 error = dsl_deleg_can_allow(zc->zc_name, 2428 fsaclnv, CRED()); 2429 } else { 2430 error = dsl_deleg_can_unallow(zc->zc_name, 2431 fsaclnv, CRED()); 2432 } 2433 } 2434 2435 if (error == 0) 2436 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action); 2437 2438 nvlist_free(fsaclnv); 2439 return (error); 2440 } 2441 2442 /* 2443 * inputs: 2444 * zc_name name of filesystem 2445 * 2446 * outputs: 2447 * zc_nvlist_src{_size} nvlist of delegated permissions 2448 */ 2449 static int 2450 zfs_ioc_get_fsacl(zfs_cmd_t *zc) 2451 { 2452 nvlist_t *nvp; 2453 int error; 2454 2455 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) { 2456 error = put_nvlist(zc, nvp); 2457 nvlist_free(nvp); 2458 } 2459 2460 return (error); 2461 } 2462 2463 /* 2464 * Search the vfs list for a specified resource. Returns a pointer to it 2465 * or NULL if no suitable entry is found. The caller of this routine 2466 * is responsible for releasing the returned vfs pointer. 2467 */ 2468 static vfs_t * 2469 zfs_get_vfs(const char *resource) 2470 { 2471 struct vfs *vfsp; 2472 struct vfs *vfs_found = NULL; 2473 2474 vfs_list_read_lock(); 2475 vfsp = rootvfs; 2476 do { 2477 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) { 2478 VFS_HOLD(vfsp); 2479 vfs_found = vfsp; 2480 break; 2481 } 2482 vfsp = vfsp->vfs_next; 2483 } while (vfsp != rootvfs); 2484 vfs_list_unlock(); 2485 return (vfs_found); 2486 } 2487 2488 /* ARGSUSED */ 2489 static void 2490 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) 2491 { 2492 zfs_creat_t *zct = arg; 2493 2494 zfs_create_fs(os, cr, zct->zct_zplprops, tx); 2495 } 2496 2497 #define ZFS_PROP_UNDEFINED ((uint64_t)-1) 2498 2499 /* 2500 * inputs: 2501 * createprops list of properties requested by creator 2502 * default_zplver zpl version to use if unspecified in createprops 2503 * fuids_ok fuids allowed in this version of the spa? 2504 * os parent objset pointer (NULL if root fs) 2505 * 2506 * outputs: 2507 * zplprops values for the zplprops we attach to the master node object 2508 * is_ci true if requested file system will be purely case-insensitive 2509 * 2510 * Determine the settings for utf8only, normalization and 2511 * casesensitivity. Specific values may have been requested by the 2512 * creator and/or we can inherit values from the parent dataset. If 2513 * the file system is of too early a vintage, a creator can not 2514 * request settings for these properties, even if the requested 2515 * setting is the default value. We don't actually want to create dsl 2516 * properties for these, so remove them from the source nvlist after 2517 * processing. 2518 */ 2519 static int 2520 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver, 2521 boolean_t fuids_ok, nvlist_t *createprops, nvlist_t *zplprops, 2522 boolean_t *is_ci) 2523 { 2524 uint64_t sense = ZFS_PROP_UNDEFINED; 2525 uint64_t norm = ZFS_PROP_UNDEFINED; 2526 uint64_t u8 = ZFS_PROP_UNDEFINED; 2527 2528 ASSERT(zplprops != NULL); 2529 2530 /* 2531 * Pull out creator prop choices, if any. 2532 */ 2533 if (createprops) { 2534 (void) nvlist_lookup_uint64(createprops, 2535 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver); 2536 (void) nvlist_lookup_uint64(createprops, 2537 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm); 2538 (void) nvlist_remove_all(createprops, 2539 zfs_prop_to_name(ZFS_PROP_NORMALIZE)); 2540 (void) nvlist_lookup_uint64(createprops, 2541 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8); 2542 (void) nvlist_remove_all(createprops, 2543 zfs_prop_to_name(ZFS_PROP_UTF8ONLY)); 2544 (void) nvlist_lookup_uint64(createprops, 2545 zfs_prop_to_name(ZFS_PROP_CASE), &sense); 2546 (void) nvlist_remove_all(createprops, 2547 zfs_prop_to_name(ZFS_PROP_CASE)); 2548 } 2549 2550 /* 2551 * If the zpl version requested is whacky or the file system 2552 * or pool is version is too "young" to support normalization 2553 * and the creator tried to set a value for one of the props, 2554 * error out. 2555 */ 2556 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) || 2557 (zplver >= ZPL_VERSION_FUID && !fuids_ok) || 2558 (zplver < ZPL_VERSION_NORMALIZATION && 2559 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED || 2560 sense != ZFS_PROP_UNDEFINED))) 2561 return (ENOTSUP); 2562 2563 /* 2564 * Put the version in the zplprops 2565 */ 2566 VERIFY(nvlist_add_uint64(zplprops, 2567 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0); 2568 2569 if (norm == ZFS_PROP_UNDEFINED) 2570 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0); 2571 VERIFY(nvlist_add_uint64(zplprops, 2572 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0); 2573 2574 /* 2575 * If we're normalizing, names must always be valid UTF-8 strings. 2576 */ 2577 if (norm) 2578 u8 = 1; 2579 if (u8 == ZFS_PROP_UNDEFINED) 2580 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0); 2581 VERIFY(nvlist_add_uint64(zplprops, 2582 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0); 2583 2584 if (sense == ZFS_PROP_UNDEFINED) 2585 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0); 2586 VERIFY(nvlist_add_uint64(zplprops, 2587 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0); 2588 2589 if (is_ci) 2590 *is_ci = (sense == ZFS_CASE_INSENSITIVE); 2591 2592 return (0); 2593 } 2594 2595 static int 2596 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops, 2597 nvlist_t *zplprops, boolean_t *is_ci) 2598 { 2599 boolean_t fuids_ok = B_TRUE; 2600 uint64_t zplver = ZPL_VERSION; 2601 objset_t *os = NULL; 2602 char parentname[MAXNAMELEN]; 2603 char *cp; 2604 int error; 2605 2606 (void) strlcpy(parentname, dataset, sizeof (parentname)); 2607 cp = strrchr(parentname, '/'); 2608 ASSERT(cp != NULL); 2609 cp[0] = '\0'; 2610 2611 if (zfs_earlier_version(dataset, SPA_VERSION_USERSPACE)) 2612 zplver = ZPL_VERSION_USERSPACE - 1; 2613 if (zfs_earlier_version(dataset, SPA_VERSION_FUID)) { 2614 zplver = ZPL_VERSION_FUID - 1; 2615 fuids_ok = B_FALSE; 2616 } 2617 2618 /* 2619 * Open parent object set so we can inherit zplprop values. 2620 */ 2621 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0) 2622 return (error); 2623 2624 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, createprops, 2625 zplprops, is_ci); 2626 dmu_objset_rele(os, FTAG); 2627 return (error); 2628 } 2629 2630 static int 2631 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops, 2632 nvlist_t *zplprops, boolean_t *is_ci) 2633 { 2634 boolean_t fuids_ok = B_TRUE; 2635 uint64_t zplver = ZPL_VERSION; 2636 int error; 2637 2638 if (spa_vers < SPA_VERSION_FUID) { 2639 zplver = ZPL_VERSION_FUID - 1; 2640 fuids_ok = B_FALSE; 2641 } 2642 2643 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, createprops, 2644 zplprops, is_ci); 2645 return (error); 2646 } 2647 2648 /* 2649 * inputs: 2650 * zc_objset_type type of objset to create (fs vs zvol) 2651 * zc_name name of new objset 2652 * zc_value name of snapshot to clone from (may be empty) 2653 * zc_nvlist_src{_size} nvlist of properties to apply 2654 * 2655 * outputs: none 2656 */ 2657 static int 2658 zfs_ioc_create(zfs_cmd_t *zc) 2659 { 2660 objset_t *clone; 2661 int error = 0; 2662 zfs_creat_t zct; 2663 nvlist_t *nvprops = NULL; 2664 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx); 2665 dmu_objset_type_t type = zc->zc_objset_type; 2666 2667 switch (type) { 2668 2669 case DMU_OST_ZFS: 2670 cbfunc = zfs_create_cb; 2671 break; 2672 2673 case DMU_OST_ZVOL: 2674 cbfunc = zvol_create_cb; 2675 break; 2676 2677 default: 2678 cbfunc = NULL; 2679 break; 2680 } 2681 if (strchr(zc->zc_name, '@') || 2682 strchr(zc->zc_name, '%')) 2683 return (EINVAL); 2684 2685 if (zc->zc_nvlist_src != NULL && 2686 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2687 zc->zc_iflags, &nvprops)) != 0) 2688 return (error); 2689 2690 zct.zct_zplprops = NULL; 2691 zct.zct_props = nvprops; 2692 2693 if (zc->zc_value[0] != '\0') { 2694 /* 2695 * We're creating a clone of an existing snapshot. 2696 */ 2697 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0'; 2698 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0) { 2699 nvlist_free(nvprops); 2700 return (EINVAL); 2701 } 2702 2703 error = dmu_objset_hold(zc->zc_value, FTAG, &clone); 2704 if (error) { 2705 nvlist_free(nvprops); 2706 return (error); 2707 } 2708 2709 error = dmu_objset_clone(zc->zc_name, dmu_objset_ds(clone), 0); 2710 dmu_objset_rele(clone, FTAG); 2711 if (error) { 2712 nvlist_free(nvprops); 2713 return (error); 2714 } 2715 } else { 2716 boolean_t is_insensitive = B_FALSE; 2717 2718 if (cbfunc == NULL) { 2719 nvlist_free(nvprops); 2720 return (EINVAL); 2721 } 2722 2723 if (type == DMU_OST_ZVOL) { 2724 uint64_t volsize, volblocksize; 2725 2726 if (nvprops == NULL || 2727 nvlist_lookup_uint64(nvprops, 2728 zfs_prop_to_name(ZFS_PROP_VOLSIZE), 2729 &volsize) != 0) { 2730 nvlist_free(nvprops); 2731 return (EINVAL); 2732 } 2733 2734 if ((error = nvlist_lookup_uint64(nvprops, 2735 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), 2736 &volblocksize)) != 0 && error != ENOENT) { 2737 nvlist_free(nvprops); 2738 return (EINVAL); 2739 } 2740 2741 if (error != 0) 2742 volblocksize = zfs_prop_default_numeric( 2743 ZFS_PROP_VOLBLOCKSIZE); 2744 2745 if ((error = zvol_check_volblocksize( 2746 volblocksize)) != 0 || 2747 (error = zvol_check_volsize(volsize, 2748 volblocksize)) != 0) { 2749 nvlist_free(nvprops); 2750 return (error); 2751 } 2752 } else if (type == DMU_OST_ZFS) { 2753 int error; 2754 2755 /* 2756 * We have to have normalization and 2757 * case-folding flags correct when we do the 2758 * file system creation, so go figure them out 2759 * now. 2760 */ 2761 VERIFY(nvlist_alloc(&zct.zct_zplprops, 2762 NV_UNIQUE_NAME, KM_SLEEP) == 0); 2763 error = zfs_fill_zplprops(zc->zc_name, nvprops, 2764 zct.zct_zplprops, &is_insensitive); 2765 if (error != 0) { 2766 nvlist_free(nvprops); 2767 nvlist_free(zct.zct_zplprops); 2768 return (error); 2769 } 2770 } 2771 error = dmu_objset_create(zc->zc_name, type, 2772 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct); 2773 nvlist_free(zct.zct_zplprops); 2774 } 2775 2776 /* 2777 * It would be nice to do this atomically. 2778 */ 2779 if (error == 0) { 2780 error = zfs_set_prop_nvlist(zc->zc_name, ZPROP_SRC_LOCAL, 2781 nvprops, NULL); 2782 if (error != 0) 2783 (void) dmu_objset_destroy(zc->zc_name, B_FALSE); 2784 } 2785 nvlist_free(nvprops); 2786 return (error); 2787 } 2788 2789 /* 2790 * inputs: 2791 * zc_name name of filesystem 2792 * zc_value short name of snapshot 2793 * zc_cookie recursive flag 2794 * zc_nvlist_src[_size] property list 2795 * 2796 * outputs: 2797 * zc_value short snapname (i.e. part after the '@') 2798 */ 2799 static int 2800 zfs_ioc_snapshot(zfs_cmd_t *zc) 2801 { 2802 nvlist_t *nvprops = NULL; 2803 int error; 2804 boolean_t recursive = zc->zc_cookie; 2805 2806 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0) 2807 return (EINVAL); 2808 2809 if (zc->zc_nvlist_src != NULL && 2810 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2811 zc->zc_iflags, &nvprops)) != 0) 2812 return (error); 2813 2814 error = zfs_check_userprops(zc->zc_name, nvprops); 2815 if (error) 2816 goto out; 2817 2818 if (!nvlist_empty(nvprops) && 2819 zfs_earlier_version(zc->zc_name, SPA_VERSION_SNAP_PROPS)) { 2820 error = ENOTSUP; 2821 goto out; 2822 } 2823 2824 error = dmu_objset_snapshot(zc->zc_name, zc->zc_value, 2825 nvprops, recursive); 2826 2827 out: 2828 nvlist_free(nvprops); 2829 return (error); 2830 } 2831 2832 int 2833 zfs_unmount_snap(char *name, void *arg) 2834 { 2835 vfs_t *vfsp = NULL; 2836 2837 if (arg) { 2838 char *snapname = arg; 2839 int len = strlen(name) + strlen(snapname) + 2; 2840 char *buf = kmem_alloc(len, KM_SLEEP); 2841 2842 (void) strcpy(buf, name); 2843 (void) strcat(buf, "@"); 2844 (void) strcat(buf, snapname); 2845 vfsp = zfs_get_vfs(buf); 2846 kmem_free(buf, len); 2847 } else if (strchr(name, '@')) { 2848 vfsp = zfs_get_vfs(name); 2849 } 2850 2851 if (vfsp) { 2852 /* 2853 * Always force the unmount for snapshots. 2854 */ 2855 int flag = MS_FORCE; 2856 int err; 2857 2858 if ((err = vn_vfswlock(vfsp->vfs_vnodecovered)) != 0) { 2859 VFS_RELE(vfsp); 2860 return (err); 2861 } 2862 VFS_RELE(vfsp); 2863 if ((err = dounmount(vfsp, flag, kcred)) != 0) 2864 return (err); 2865 } 2866 return (0); 2867 } 2868 2869 /* 2870 * inputs: 2871 * zc_name name of filesystem 2872 * zc_value short name of snapshot 2873 * zc_defer_destroy mark for deferred destroy 2874 * 2875 * outputs: none 2876 */ 2877 static int 2878 zfs_ioc_destroy_snaps(zfs_cmd_t *zc) 2879 { 2880 int err; 2881 2882 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0) 2883 return (EINVAL); 2884 err = dmu_objset_find(zc->zc_name, 2885 zfs_unmount_snap, zc->zc_value, DS_FIND_CHILDREN); 2886 if (err) 2887 return (err); 2888 return (dmu_snapshots_destroy(zc->zc_name, zc->zc_value, 2889 zc->zc_defer_destroy)); 2890 } 2891 2892 /* 2893 * inputs: 2894 * zc_name name of dataset to destroy 2895 * zc_objset_type type of objset 2896 * zc_defer_destroy mark for deferred destroy 2897 * 2898 * outputs: none 2899 */ 2900 static int 2901 zfs_ioc_destroy(zfs_cmd_t *zc) 2902 { 2903 int err; 2904 if (strchr(zc->zc_name, '@') && zc->zc_objset_type == DMU_OST_ZFS) { 2905 err = zfs_unmount_snap(zc->zc_name, NULL); 2906 if (err) 2907 return (err); 2908 } 2909 2910 err = dmu_objset_destroy(zc->zc_name, zc->zc_defer_destroy); 2911 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0) 2912 (void) zvol_remove_minor(zc->zc_name); 2913 return (err); 2914 } 2915 2916 /* 2917 * inputs: 2918 * zc_name name of dataset to rollback (to most recent snapshot) 2919 * 2920 * outputs: none 2921 */ 2922 static int 2923 zfs_ioc_rollback(zfs_cmd_t *zc) 2924 { 2925 dsl_dataset_t *ds, *clone; 2926 int error; 2927 zfsvfs_t *zfsvfs; 2928 char *clone_name; 2929 2930 error = dsl_dataset_hold(zc->zc_name, FTAG, &ds); 2931 if (error) 2932 return (error); 2933 2934 /* must not be a snapshot */ 2935 if (dsl_dataset_is_snapshot(ds)) { 2936 dsl_dataset_rele(ds, FTAG); 2937 return (EINVAL); 2938 } 2939 2940 /* must have a most recent snapshot */ 2941 if (ds->ds_phys->ds_prev_snap_txg < TXG_INITIAL) { 2942 dsl_dataset_rele(ds, FTAG); 2943 return (EINVAL); 2944 } 2945 2946 /* 2947 * Create clone of most recent snapshot. 2948 */ 2949 clone_name = kmem_asprintf("%s/%%rollback", zc->zc_name); 2950 error = dmu_objset_clone(clone_name, ds->ds_prev, DS_FLAG_INCONSISTENT); 2951 if (error) 2952 goto out; 2953 2954 error = dsl_dataset_own(clone_name, B_TRUE, FTAG, &clone); 2955 if (error) 2956 goto out; 2957 2958 /* 2959 * Do clone swap. 2960 */ 2961 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) { 2962 error = zfs_suspend_fs(zfsvfs); 2963 if (error == 0) { 2964 int resume_err; 2965 2966 if (dsl_dataset_tryown(ds, B_FALSE, FTAG)) { 2967 error = dsl_dataset_clone_swap(clone, ds, 2968 B_TRUE); 2969 dsl_dataset_disown(ds, FTAG); 2970 ds = NULL; 2971 } else { 2972 error = EBUSY; 2973 } 2974 resume_err = zfs_resume_fs(zfsvfs, zc->zc_name); 2975 error = error ? error : resume_err; 2976 } 2977 VFS_RELE(zfsvfs->z_vfs); 2978 } else { 2979 if (dsl_dataset_tryown(ds, B_FALSE, FTAG)) { 2980 error = dsl_dataset_clone_swap(clone, ds, B_TRUE); 2981 dsl_dataset_disown(ds, FTAG); 2982 ds = NULL; 2983 } else { 2984 error = EBUSY; 2985 } 2986 } 2987 2988 /* 2989 * Destroy clone (which also closes it). 2990 */ 2991 (void) dsl_dataset_destroy(clone, FTAG, B_FALSE); 2992 2993 out: 2994 strfree(clone_name); 2995 if (ds) 2996 dsl_dataset_rele(ds, FTAG); 2997 return (error); 2998 } 2999 3000 /* 3001 * inputs: 3002 * zc_name old name of dataset 3003 * zc_value new name of dataset 3004 * zc_cookie recursive flag (only valid for snapshots) 3005 * 3006 * outputs: none 3007 */ 3008 static int 3009 zfs_ioc_rename(zfs_cmd_t *zc) 3010 { 3011 boolean_t recursive = zc->zc_cookie & 1; 3012 3013 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0'; 3014 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 3015 strchr(zc->zc_value, '%')) 3016 return (EINVAL); 3017 3018 /* 3019 * Unmount snapshot unless we're doing a recursive rename, 3020 * in which case the dataset code figures out which snapshots 3021 * to unmount. 3022 */ 3023 if (!recursive && strchr(zc->zc_name, '@') != NULL && 3024 zc->zc_objset_type == DMU_OST_ZFS) { 3025 int err = zfs_unmount_snap(zc->zc_name, NULL); 3026 if (err) 3027 return (err); 3028 } 3029 if (zc->zc_objset_type == DMU_OST_ZVOL) 3030 (void) zvol_remove_minor(zc->zc_name); 3031 return (dmu_objset_rename(zc->zc_name, zc->zc_value, recursive)); 3032 } 3033 3034 static int 3035 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr) 3036 { 3037 const char *propname = nvpair_name(pair); 3038 boolean_t issnap = (strchr(dsname, '@') != NULL); 3039 zfs_prop_t prop = zfs_name_to_prop(propname); 3040 uint64_t intval; 3041 int err; 3042 3043 if (prop == ZPROP_INVAL) { 3044 if (zfs_prop_user(propname)) { 3045 if (err = zfs_secpolicy_write_perms(dsname, 3046 ZFS_DELEG_PERM_USERPROP, cr)) 3047 return (err); 3048 return (0); 3049 } 3050 3051 if (!issnap && zfs_prop_userquota(propname)) { 3052 const char *perm = NULL; 3053 const char *uq_prefix = 3054 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA]; 3055 const char *gq_prefix = 3056 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA]; 3057 3058 if (strncmp(propname, uq_prefix, 3059 strlen(uq_prefix)) == 0) { 3060 perm = ZFS_DELEG_PERM_USERQUOTA; 3061 } else if (strncmp(propname, gq_prefix, 3062 strlen(gq_prefix)) == 0) { 3063 perm = ZFS_DELEG_PERM_GROUPQUOTA; 3064 } else { 3065 /* USERUSED and GROUPUSED are read-only */ 3066 return (EINVAL); 3067 } 3068 3069 if (err = zfs_secpolicy_write_perms(dsname, perm, cr)) 3070 return (err); 3071 return (0); 3072 } 3073 3074 return (EINVAL); 3075 } 3076 3077 if (issnap) 3078 return (EINVAL); 3079 3080 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 3081 /* 3082 * dsl_prop_get_all_impl() returns properties in this 3083 * format. 3084 */ 3085 nvlist_t *attrs; 3086 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 3087 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 3088 &pair) == 0); 3089 } 3090 3091 /* 3092 * Check that this value is valid for this pool version 3093 */ 3094 switch (prop) { 3095 case ZFS_PROP_COMPRESSION: 3096 /* 3097 * If the user specified gzip compression, make sure 3098 * the SPA supports it. We ignore any errors here since 3099 * we'll catch them later. 3100 */ 3101 if (nvpair_type(pair) == DATA_TYPE_UINT64 && 3102 nvpair_value_uint64(pair, &intval) == 0) { 3103 if (intval >= ZIO_COMPRESS_GZIP_1 && 3104 intval <= ZIO_COMPRESS_GZIP_9 && 3105 zfs_earlier_version(dsname, 3106 SPA_VERSION_GZIP_COMPRESSION)) { 3107 return (ENOTSUP); 3108 } 3109 3110 if (intval == ZIO_COMPRESS_ZLE && 3111 zfs_earlier_version(dsname, 3112 SPA_VERSION_ZLE_COMPRESSION)) 3113 return (ENOTSUP); 3114 3115 /* 3116 * If this is a bootable dataset then 3117 * verify that the compression algorithm 3118 * is supported for booting. We must return 3119 * something other than ENOTSUP since it 3120 * implies a downrev pool version. 3121 */ 3122 if (zfs_is_bootfs(dsname) && 3123 !BOOTFS_COMPRESS_VALID(intval)) { 3124 return (ERANGE); 3125 } 3126 } 3127 break; 3128 3129 case ZFS_PROP_COPIES: 3130 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS)) 3131 return (ENOTSUP); 3132 break; 3133 3134 case ZFS_PROP_DEDUP: 3135 if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP)) 3136 return (ENOTSUP); 3137 break; 3138 3139 case ZFS_PROP_SHARESMB: 3140 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID)) 3141 return (ENOTSUP); 3142 break; 3143 3144 case ZFS_PROP_ACLINHERIT: 3145 if (nvpair_type(pair) == DATA_TYPE_UINT64 && 3146 nvpair_value_uint64(pair, &intval) == 0) { 3147 if (intval == ZFS_ACL_PASSTHROUGH_X && 3148 zfs_earlier_version(dsname, 3149 SPA_VERSION_PASSTHROUGH_X)) 3150 return (ENOTSUP); 3151 } 3152 break; 3153 } 3154 3155 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED())); 3156 } 3157 3158 /* 3159 * Removes properties from the given props list that fail permission checks 3160 * needed to clear them and to restore them in case of a receive error. For each 3161 * property, make sure we have both set and inherit permissions. 3162 * 3163 * Returns the first error encountered if any permission checks fail. If the 3164 * caller provides a non-NULL errlist, it also gives the complete list of names 3165 * of all the properties that failed a permission check along with the 3166 * corresponding error numbers. The caller is responsible for freeing the 3167 * returned errlist. 3168 * 3169 * If every property checks out successfully, zero is returned and the list 3170 * pointed at by errlist is NULL. 3171 */ 3172 static int 3173 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist) 3174 { 3175 zfs_cmd_t *zc; 3176 nvpair_t *pair, *next_pair; 3177 nvlist_t *errors; 3178 int err, rv = 0; 3179 3180 if (props == NULL) 3181 return (0); 3182 3183 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0); 3184 3185 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP); 3186 (void) strcpy(zc->zc_name, dataset); 3187 pair = nvlist_next_nvpair(props, NULL); 3188 while (pair != NULL) { 3189 next_pair = nvlist_next_nvpair(props, pair); 3190 3191 (void) strcpy(zc->zc_value, nvpair_name(pair)); 3192 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 || 3193 (err = zfs_secpolicy_inherit(zc, CRED())) != 0) { 3194 VERIFY(nvlist_remove_nvpair(props, pair) == 0); 3195 VERIFY(nvlist_add_int32(errors, 3196 zc->zc_value, err) == 0); 3197 } 3198 pair = next_pair; 3199 } 3200 kmem_free(zc, sizeof (zfs_cmd_t)); 3201 3202 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) { 3203 nvlist_free(errors); 3204 errors = NULL; 3205 } else { 3206 VERIFY(nvpair_value_int32(pair, &rv) == 0); 3207 } 3208 3209 if (errlist == NULL) 3210 nvlist_free(errors); 3211 else 3212 *errlist = errors; 3213 3214 return (rv); 3215 } 3216 3217 static boolean_t 3218 propval_equals(nvpair_t *p1, nvpair_t *p2) 3219 { 3220 if (nvpair_type(p1) == DATA_TYPE_NVLIST) { 3221 /* dsl_prop_get_all_impl() format */ 3222 nvlist_t *attrs; 3223 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0); 3224 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 3225 &p1) == 0); 3226 } 3227 3228 if (nvpair_type(p2) == DATA_TYPE_NVLIST) { 3229 nvlist_t *attrs; 3230 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0); 3231 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 3232 &p2) == 0); 3233 } 3234 3235 if (nvpair_type(p1) != nvpair_type(p2)) 3236 return (B_FALSE); 3237 3238 if (nvpair_type(p1) == DATA_TYPE_STRING) { 3239 char *valstr1, *valstr2; 3240 3241 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0); 3242 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0); 3243 return (strcmp(valstr1, valstr2) == 0); 3244 } else { 3245 uint64_t intval1, intval2; 3246 3247 VERIFY(nvpair_value_uint64(p1, &intval1) == 0); 3248 VERIFY(nvpair_value_uint64(p2, &intval2) == 0); 3249 return (intval1 == intval2); 3250 } 3251 } 3252 3253 /* 3254 * Remove properties from props if they are not going to change (as determined 3255 * by comparison with origprops). Remove them from origprops as well, since we 3256 * do not need to clear or restore properties that won't change. 3257 */ 3258 static void 3259 props_reduce(nvlist_t *props, nvlist_t *origprops) 3260 { 3261 nvpair_t *pair, *next_pair; 3262 3263 if (origprops == NULL) 3264 return; /* all props need to be received */ 3265 3266 pair = nvlist_next_nvpair(props, NULL); 3267 while (pair != NULL) { 3268 const char *propname = nvpair_name(pair); 3269 nvpair_t *match; 3270 3271 next_pair = nvlist_next_nvpair(props, pair); 3272 3273 if ((nvlist_lookup_nvpair(origprops, propname, 3274 &match) != 0) || !propval_equals(pair, match)) 3275 goto next; /* need to set received value */ 3276 3277 /* don't clear the existing received value */ 3278 (void) nvlist_remove_nvpair(origprops, match); 3279 /* don't bother receiving the property */ 3280 (void) nvlist_remove_nvpair(props, pair); 3281 next: 3282 pair = next_pair; 3283 } 3284 } 3285 3286 #ifdef DEBUG 3287 static boolean_t zfs_ioc_recv_inject_err; 3288 #endif 3289 3290 /* 3291 * inputs: 3292 * zc_name name of containing filesystem 3293 * zc_nvlist_src{_size} nvlist of properties to apply 3294 * zc_value name of snapshot to create 3295 * zc_string name of clone origin (if DRR_FLAG_CLONE) 3296 * zc_cookie file descriptor to recv from 3297 * zc_begin_record the BEGIN record of the stream (not byteswapped) 3298 * zc_guid force flag 3299 * 3300 * outputs: 3301 * zc_cookie number of bytes read 3302 * zc_nvlist_dst{_size} error for each unapplied received property 3303 * zc_obj zprop_errflags_t 3304 */ 3305 static int 3306 zfs_ioc_recv(zfs_cmd_t *zc) 3307 { 3308 file_t *fp; 3309 objset_t *os; 3310 dmu_recv_cookie_t drc; 3311 boolean_t force = (boolean_t)zc->zc_guid; 3312 int fd; 3313 int error = 0; 3314 int props_error = 0; 3315 nvlist_t *errors; 3316 offset_t off; 3317 nvlist_t *props = NULL; /* sent properties */ 3318 nvlist_t *origprops = NULL; /* existing properties */ 3319 objset_t *origin = NULL; 3320 char *tosnap; 3321 char tofs[ZFS_MAXNAMELEN]; 3322 boolean_t first_recvd_props = B_FALSE; 3323 3324 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 3325 strchr(zc->zc_value, '@') == NULL || 3326 strchr(zc->zc_value, '%')) 3327 return (EINVAL); 3328 3329 (void) strcpy(tofs, zc->zc_value); 3330 tosnap = strchr(tofs, '@'); 3331 *tosnap++ = '\0'; 3332 3333 if (zc->zc_nvlist_src != NULL && 3334 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 3335 zc->zc_iflags, &props)) != 0) 3336 return (error); 3337 3338 fd = zc->zc_cookie; 3339 fp = getf(fd); 3340 if (fp == NULL) { 3341 nvlist_free(props); 3342 return (EBADF); 3343 } 3344 3345 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0); 3346 3347 if (props && dmu_objset_hold(tofs, FTAG, &os) == 0) { 3348 if ((spa_version(os->os_spa) >= SPA_VERSION_RECVD_PROPS) && 3349 !dsl_prop_get_hasrecvd(os)) { 3350 first_recvd_props = B_TRUE; 3351 } 3352 3353 /* 3354 * If new received properties are supplied, they are to 3355 * completely replace the existing received properties, so stash 3356 * away the existing ones. 3357 */ 3358 if (dsl_prop_get_received(os, &origprops) == 0) { 3359 nvlist_t *errlist = NULL; 3360 /* 3361 * Don't bother writing a property if its value won't 3362 * change (and avoid the unnecessary security checks). 3363 * 3364 * The first receive after SPA_VERSION_RECVD_PROPS is a 3365 * special case where we blow away all local properties 3366 * regardless. 3367 */ 3368 if (!first_recvd_props) 3369 props_reduce(props, origprops); 3370 if (zfs_check_clearable(tofs, origprops, 3371 &errlist) != 0) 3372 (void) nvlist_merge(errors, errlist, 0); 3373 nvlist_free(errlist); 3374 } 3375 3376 dmu_objset_rele(os, FTAG); 3377 } 3378 3379 if (zc->zc_string[0]) { 3380 error = dmu_objset_hold(zc->zc_string, FTAG, &origin); 3381 if (error) 3382 goto out; 3383 } 3384 3385 error = dmu_recv_begin(tofs, tosnap, zc->zc_top_ds, 3386 &zc->zc_begin_record, force, origin, &drc); 3387 if (origin) 3388 dmu_objset_rele(origin, FTAG); 3389 if (error) 3390 goto out; 3391 3392 /* 3393 * Set properties before we receive the stream so that they are applied 3394 * to the new data. Note that we must call dmu_recv_stream() if 3395 * dmu_recv_begin() succeeds. 3396 */ 3397 if (props) { 3398 nvlist_t *errlist; 3399 3400 if (dmu_objset_from_ds(drc.drc_logical_ds, &os) == 0) { 3401 if (drc.drc_newfs) { 3402 if (spa_version(os->os_spa) >= 3403 SPA_VERSION_RECVD_PROPS) 3404 first_recvd_props = B_TRUE; 3405 } else if (origprops != NULL) { 3406 if (clear_received_props(os, tofs, origprops, 3407 first_recvd_props ? NULL : props) != 0) 3408 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 3409 } else { 3410 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 3411 } 3412 dsl_prop_set_hasrecvd(os); 3413 } else if (!drc.drc_newfs) { 3414 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 3415 } 3416 3417 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED, 3418 props, &errlist); 3419 (void) nvlist_merge(errors, errlist, 0); 3420 nvlist_free(errlist); 3421 } 3422 3423 if (fit_error_list(zc, &errors) != 0 || put_nvlist(zc, errors) != 0) { 3424 /* 3425 * Caller made zc->zc_nvlist_dst less than the minimum expected 3426 * size or supplied an invalid address. 3427 */ 3428 props_error = EINVAL; 3429 } 3430 3431 off = fp->f_offset; 3432 error = dmu_recv_stream(&drc, fp->f_vnode, &off); 3433 3434 if (error == 0) { 3435 zfsvfs_t *zfsvfs = NULL; 3436 3437 if (getzfsvfs(tofs, &zfsvfs) == 0) { 3438 /* online recv */ 3439 int end_err; 3440 3441 error = zfs_suspend_fs(zfsvfs); 3442 /* 3443 * If the suspend fails, then the recv_end will 3444 * likely also fail, and clean up after itself. 3445 */ 3446 end_err = dmu_recv_end(&drc); 3447 if (error == 0) { 3448 int resume_err = 3449 zfs_resume_fs(zfsvfs, tofs); 3450 error = error ? error : resume_err; 3451 } 3452 error = error ? error : end_err; 3453 VFS_RELE(zfsvfs->z_vfs); 3454 } else { 3455 error = dmu_recv_end(&drc); 3456 } 3457 } 3458 3459 zc->zc_cookie = off - fp->f_offset; 3460 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0) 3461 fp->f_offset = off; 3462 3463 #ifdef DEBUG 3464 if (zfs_ioc_recv_inject_err) { 3465 zfs_ioc_recv_inject_err = B_FALSE; 3466 error = 1; 3467 } 3468 #endif 3469 /* 3470 * On error, restore the original props. 3471 */ 3472 if (error && props) { 3473 if (dmu_objset_hold(tofs, FTAG, &os) == 0) { 3474 if (clear_received_props(os, tofs, props, NULL) != 0) { 3475 /* 3476 * We failed to clear the received properties. 3477 * Since we may have left a $recvd value on the 3478 * system, we can't clear the $hasrecvd flag. 3479 */ 3480 zc->zc_obj |= ZPROP_ERR_NORESTORE; 3481 } else if (first_recvd_props) { 3482 dsl_prop_unset_hasrecvd(os); 3483 } 3484 dmu_objset_rele(os, FTAG); 3485 } else if (!drc.drc_newfs) { 3486 /* We failed to clear the received properties. */ 3487 zc->zc_obj |= ZPROP_ERR_NORESTORE; 3488 } 3489 3490 if (origprops == NULL && !drc.drc_newfs) { 3491 /* We failed to stash the original properties. */ 3492 zc->zc_obj |= ZPROP_ERR_NORESTORE; 3493 } 3494 3495 /* 3496 * dsl_props_set() will not convert RECEIVED to LOCAL on or 3497 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL 3498 * explictly if we're restoring local properties cleared in the 3499 * first new-style receive. 3500 */ 3501 if (origprops != NULL && 3502 zfs_set_prop_nvlist(tofs, (first_recvd_props ? 3503 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED), 3504 origprops, NULL) != 0) { 3505 /* 3506 * We stashed the original properties but failed to 3507 * restore them. 3508 */ 3509 zc->zc_obj |= ZPROP_ERR_NORESTORE; 3510 } 3511 } 3512 out: 3513 nvlist_free(props); 3514 nvlist_free(origprops); 3515 nvlist_free(errors); 3516 releasef(fd); 3517 3518 if (error == 0) 3519 error = props_error; 3520 3521 return (error); 3522 } 3523 3524 /* 3525 * inputs: 3526 * zc_name name of snapshot to send 3527 * zc_value short name of incremental fromsnap (may be empty) 3528 * zc_cookie file descriptor to send stream to 3529 * zc_obj fromorigin flag (mutually exclusive with zc_value) 3530 * 3531 * outputs: none 3532 */ 3533 static int 3534 zfs_ioc_send(zfs_cmd_t *zc) 3535 { 3536 objset_t *fromsnap = NULL; 3537 objset_t *tosnap; 3538 file_t *fp; 3539 int error; 3540 offset_t off; 3541 3542 error = dmu_objset_hold(zc->zc_name, FTAG, &tosnap); 3543 if (error) 3544 return (error); 3545 3546 if (zc->zc_value[0] != '\0') { 3547 char *buf; 3548 char *cp; 3549 3550 buf = kmem_alloc(MAXPATHLEN, KM_SLEEP); 3551 (void) strncpy(buf, zc->zc_name, MAXPATHLEN); 3552 cp = strchr(buf, '@'); 3553 if (cp) 3554 *(cp+1) = 0; 3555 (void) strncat(buf, zc->zc_value, MAXPATHLEN); 3556 error = dmu_objset_hold(buf, FTAG, &fromsnap); 3557 kmem_free(buf, MAXPATHLEN); 3558 if (error) { 3559 dmu_objset_rele(tosnap, FTAG); 3560 return (error); 3561 } 3562 } 3563 3564 fp = getf(zc->zc_cookie); 3565 if (fp == NULL) { 3566 dmu_objset_rele(tosnap, FTAG); 3567 if (fromsnap) 3568 dmu_objset_rele(fromsnap, FTAG); 3569 return (EBADF); 3570 } 3571 3572 off = fp->f_offset; 3573 error = dmu_sendbackup(tosnap, fromsnap, zc->zc_obj, fp->f_vnode, &off); 3574 3575 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0) 3576 fp->f_offset = off; 3577 releasef(zc->zc_cookie); 3578 if (fromsnap) 3579 dmu_objset_rele(fromsnap, FTAG); 3580 dmu_objset_rele(tosnap, FTAG); 3581 return (error); 3582 } 3583 3584 static int 3585 zfs_ioc_inject_fault(zfs_cmd_t *zc) 3586 { 3587 int id, error; 3588 3589 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id, 3590 &zc->zc_inject_record); 3591 3592 if (error == 0) 3593 zc->zc_guid = (uint64_t)id; 3594 3595 return (error); 3596 } 3597 3598 static int 3599 zfs_ioc_clear_fault(zfs_cmd_t *zc) 3600 { 3601 return (zio_clear_fault((int)zc->zc_guid)); 3602 } 3603 3604 static int 3605 zfs_ioc_inject_list_next(zfs_cmd_t *zc) 3606 { 3607 int id = (int)zc->zc_guid; 3608 int error; 3609 3610 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name), 3611 &zc->zc_inject_record); 3612 3613 zc->zc_guid = id; 3614 3615 return (error); 3616 } 3617 3618 static int 3619 zfs_ioc_error_log(zfs_cmd_t *zc) 3620 { 3621 spa_t *spa; 3622 int error; 3623 size_t count = (size_t)zc->zc_nvlist_dst_size; 3624 3625 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 3626 return (error); 3627 3628 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst, 3629 &count); 3630 if (error == 0) 3631 zc->zc_nvlist_dst_size = count; 3632 else 3633 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa); 3634 3635 spa_close(spa, FTAG); 3636 3637 return (error); 3638 } 3639 3640 static int 3641 zfs_ioc_clear(zfs_cmd_t *zc) 3642 { 3643 spa_t *spa; 3644 vdev_t *vd; 3645 int error; 3646 3647 /* 3648 * On zpool clear we also fix up missing slogs 3649 */ 3650 mutex_enter(&spa_namespace_lock); 3651 spa = spa_lookup(zc->zc_name); 3652 if (spa == NULL) { 3653 mutex_exit(&spa_namespace_lock); 3654 return (EIO); 3655 } 3656 if (spa_get_log_state(spa) == SPA_LOG_MISSING) { 3657 /* we need to let spa_open/spa_load clear the chains */ 3658 spa_set_log_state(spa, SPA_LOG_CLEAR); 3659 } 3660 spa->spa_last_open_failed = 0; 3661 mutex_exit(&spa_namespace_lock); 3662 3663 if (zc->zc_cookie == ZPOOL_NO_REWIND) { 3664 error = spa_open(zc->zc_name, &spa, FTAG); 3665 } else { 3666 nvlist_t *policy; 3667 nvlist_t *config = NULL; 3668 3669 if (zc->zc_nvlist_src == NULL) 3670 return (EINVAL); 3671 3672 if ((error = get_nvlist(zc->zc_nvlist_src, 3673 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) { 3674 error = spa_open_rewind(zc->zc_name, &spa, FTAG, 3675 policy, &config); 3676 if (config != NULL) { 3677 (void) put_nvlist(zc, config); 3678 nvlist_free(config); 3679 } 3680 nvlist_free(policy); 3681 } 3682 } 3683 3684 if (error) 3685 return (error); 3686 3687 spa_vdev_state_enter(spa, SCL_NONE); 3688 3689 if (zc->zc_guid == 0) { 3690 vd = NULL; 3691 } else { 3692 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE); 3693 if (vd == NULL) { 3694 (void) spa_vdev_state_exit(spa, NULL, ENODEV); 3695 spa_close(spa, FTAG); 3696 return (ENODEV); 3697 } 3698 } 3699 3700 vdev_clear(spa, vd); 3701 3702 (void) spa_vdev_state_exit(spa, NULL, 0); 3703 3704 /* 3705 * Resume any suspended I/Os. 3706 */ 3707 if (zio_resume(spa) != 0) 3708 error = EIO; 3709 3710 spa_close(spa, FTAG); 3711 3712 return (error); 3713 } 3714 3715 /* 3716 * inputs: 3717 * zc_name name of filesystem 3718 * zc_value name of origin snapshot 3719 * 3720 * outputs: 3721 * zc_string name of conflicting snapshot, if there is one 3722 */ 3723 static int 3724 zfs_ioc_promote(zfs_cmd_t *zc) 3725 { 3726 char *cp; 3727 3728 /* 3729 * We don't need to unmount *all* the origin fs's snapshots, but 3730 * it's easier. 3731 */ 3732 cp = strchr(zc->zc_value, '@'); 3733 if (cp) 3734 *cp = '\0'; 3735 (void) dmu_objset_find(zc->zc_value, 3736 zfs_unmount_snap, NULL, DS_FIND_SNAPSHOTS); 3737 return (dsl_dataset_promote(zc->zc_name, zc->zc_string)); 3738 } 3739 3740 /* 3741 * Retrieve a single {user|group}{used|quota}@... property. 3742 * 3743 * inputs: 3744 * zc_name name of filesystem 3745 * zc_objset_type zfs_userquota_prop_t 3746 * zc_value domain name (eg. "S-1-234-567-89") 3747 * zc_guid RID/UID/GID 3748 * 3749 * outputs: 3750 * zc_cookie property value 3751 */ 3752 static int 3753 zfs_ioc_userspace_one(zfs_cmd_t *zc) 3754 { 3755 zfsvfs_t *zfsvfs; 3756 int error; 3757 3758 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 3759 return (EINVAL); 3760 3761 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs); 3762 if (error) 3763 return (error); 3764 3765 error = zfs_userspace_one(zfsvfs, 3766 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie); 3767 zfsvfs_rele(zfsvfs, FTAG); 3768 3769 return (error); 3770 } 3771 3772 /* 3773 * inputs: 3774 * zc_name name of filesystem 3775 * zc_cookie zap cursor 3776 * zc_objset_type zfs_userquota_prop_t 3777 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist) 3778 * 3779 * outputs: 3780 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t) 3781 * zc_cookie zap cursor 3782 */ 3783 static int 3784 zfs_ioc_userspace_many(zfs_cmd_t *zc) 3785 { 3786 zfsvfs_t *zfsvfs; 3787 int error; 3788 3789 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs); 3790 if (error) 3791 return (error); 3792 3793 int bufsize = zc->zc_nvlist_dst_size; 3794 void *buf = kmem_alloc(bufsize, KM_SLEEP); 3795 3796 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie, 3797 buf, &zc->zc_nvlist_dst_size); 3798 3799 if (error == 0) { 3800 error = xcopyout(buf, 3801 (void *)(uintptr_t)zc->zc_nvlist_dst, 3802 zc->zc_nvlist_dst_size); 3803 } 3804 kmem_free(buf, bufsize); 3805 zfsvfs_rele(zfsvfs, FTAG); 3806 3807 return (error); 3808 } 3809 3810 /* 3811 * inputs: 3812 * zc_name name of filesystem 3813 * 3814 * outputs: 3815 * none 3816 */ 3817 static int 3818 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc) 3819 { 3820 objset_t *os; 3821 int error; 3822 zfsvfs_t *zfsvfs; 3823 3824 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) { 3825 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) { 3826 /* 3827 * If userused is not enabled, it may be because the 3828 * objset needs to be closed & reopened (to grow the 3829 * objset_phys_t). Suspend/resume the fs will do that. 3830 */ 3831 error = zfs_suspend_fs(zfsvfs); 3832 if (error == 0) 3833 error = zfs_resume_fs(zfsvfs, zc->zc_name); 3834 } 3835 if (error == 0) 3836 error = dmu_objset_userspace_upgrade(zfsvfs->z_os); 3837 VFS_RELE(zfsvfs->z_vfs); 3838 } else { 3839 /* XXX kind of reading contents without owning */ 3840 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 3841 if (error) 3842 return (error); 3843 3844 error = dmu_objset_userspace_upgrade(os); 3845 dmu_objset_rele(os, FTAG); 3846 } 3847 3848 return (error); 3849 } 3850 3851 /* 3852 * We don't want to have a hard dependency 3853 * against some special symbols in sharefs 3854 * nfs, and smbsrv. Determine them if needed when 3855 * the first file system is shared. 3856 * Neither sharefs, nfs or smbsrv are unloadable modules. 3857 */ 3858 int (*znfsexport_fs)(void *arg); 3859 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t); 3860 int (*zsmbexport_fs)(void *arg, boolean_t add_share); 3861 3862 int zfs_nfsshare_inited; 3863 int zfs_smbshare_inited; 3864 3865 ddi_modhandle_t nfs_mod; 3866 ddi_modhandle_t sharefs_mod; 3867 ddi_modhandle_t smbsrv_mod; 3868 kmutex_t zfs_share_lock; 3869 3870 static int 3871 zfs_init_sharefs() 3872 { 3873 int error; 3874 3875 ASSERT(MUTEX_HELD(&zfs_share_lock)); 3876 /* Both NFS and SMB shares also require sharetab support. */ 3877 if (sharefs_mod == NULL && ((sharefs_mod = 3878 ddi_modopen("fs/sharefs", 3879 KRTLD_MODE_FIRST, &error)) == NULL)) { 3880 return (ENOSYS); 3881 } 3882 if (zshare_fs == NULL && ((zshare_fs = 3883 (int (*)(enum sharefs_sys_op, share_t *, uint32_t)) 3884 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) { 3885 return (ENOSYS); 3886 } 3887 return (0); 3888 } 3889 3890 static int 3891 zfs_ioc_share(zfs_cmd_t *zc) 3892 { 3893 int error; 3894 int opcode; 3895 3896 switch (zc->zc_share.z_sharetype) { 3897 case ZFS_SHARE_NFS: 3898 case ZFS_UNSHARE_NFS: 3899 if (zfs_nfsshare_inited == 0) { 3900 mutex_enter(&zfs_share_lock); 3901 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs", 3902 KRTLD_MODE_FIRST, &error)) == NULL)) { 3903 mutex_exit(&zfs_share_lock); 3904 return (ENOSYS); 3905 } 3906 if (znfsexport_fs == NULL && 3907 ((znfsexport_fs = (int (*)(void *)) 3908 ddi_modsym(nfs_mod, 3909 "nfs_export", &error)) == NULL)) { 3910 mutex_exit(&zfs_share_lock); 3911 return (ENOSYS); 3912 } 3913 error = zfs_init_sharefs(); 3914 if (error) { 3915 mutex_exit(&zfs_share_lock); 3916 return (ENOSYS); 3917 } 3918 zfs_nfsshare_inited = 1; 3919 mutex_exit(&zfs_share_lock); 3920 } 3921 break; 3922 case ZFS_SHARE_SMB: 3923 case ZFS_UNSHARE_SMB: 3924 if (zfs_smbshare_inited == 0) { 3925 mutex_enter(&zfs_share_lock); 3926 if (smbsrv_mod == NULL && ((smbsrv_mod = 3927 ddi_modopen("drv/smbsrv", 3928 KRTLD_MODE_FIRST, &error)) == NULL)) { 3929 mutex_exit(&zfs_share_lock); 3930 return (ENOSYS); 3931 } 3932 if (zsmbexport_fs == NULL && ((zsmbexport_fs = 3933 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod, 3934 "smb_server_share", &error)) == NULL)) { 3935 mutex_exit(&zfs_share_lock); 3936 return (ENOSYS); 3937 } 3938 error = zfs_init_sharefs(); 3939 if (error) { 3940 mutex_exit(&zfs_share_lock); 3941 return (ENOSYS); 3942 } 3943 zfs_smbshare_inited = 1; 3944 mutex_exit(&zfs_share_lock); 3945 } 3946 break; 3947 default: 3948 return (EINVAL); 3949 } 3950 3951 switch (zc->zc_share.z_sharetype) { 3952 case ZFS_SHARE_NFS: 3953 case ZFS_UNSHARE_NFS: 3954 if (error = 3955 znfsexport_fs((void *) 3956 (uintptr_t)zc->zc_share.z_exportdata)) 3957 return (error); 3958 break; 3959 case ZFS_SHARE_SMB: 3960 case ZFS_UNSHARE_SMB: 3961 if (error = zsmbexport_fs((void *) 3962 (uintptr_t)zc->zc_share.z_exportdata, 3963 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ? 3964 B_TRUE: B_FALSE)) { 3965 return (error); 3966 } 3967 break; 3968 } 3969 3970 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS || 3971 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ? 3972 SHAREFS_ADD : SHAREFS_REMOVE; 3973 3974 /* 3975 * Add or remove share from sharetab 3976 */ 3977 error = zshare_fs(opcode, 3978 (void *)(uintptr_t)zc->zc_share.z_sharedata, 3979 zc->zc_share.z_sharemax); 3980 3981 return (error); 3982 3983 } 3984 3985 ace_t full_access[] = { 3986 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0} 3987 }; 3988 3989 /* 3990 * Remove all ACL files in shares dir 3991 */ 3992 static int 3993 zfs_smb_acl_purge(znode_t *dzp) 3994 { 3995 zap_cursor_t zc; 3996 zap_attribute_t zap; 3997 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 3998 int error; 3999 4000 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id); 4001 (error = zap_cursor_retrieve(&zc, &zap)) == 0; 4002 zap_cursor_advance(&zc)) { 4003 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred, 4004 NULL, 0)) != 0) 4005 break; 4006 } 4007 zap_cursor_fini(&zc); 4008 return (error); 4009 } 4010 4011 static int 4012 zfs_ioc_smb_acl(zfs_cmd_t *zc) 4013 { 4014 vnode_t *vp; 4015 znode_t *dzp; 4016 vnode_t *resourcevp = NULL; 4017 znode_t *sharedir; 4018 zfsvfs_t *zfsvfs; 4019 nvlist_t *nvlist; 4020 char *src, *target; 4021 vattr_t vattr; 4022 vsecattr_t vsec; 4023 int error = 0; 4024 4025 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, 4026 NO_FOLLOW, NULL, &vp)) != 0) 4027 return (error); 4028 4029 /* Now make sure mntpnt and dataset are ZFS */ 4030 4031 if (vp->v_vfsp->vfs_fstype != zfsfstype || 4032 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), 4033 zc->zc_name) != 0)) { 4034 VN_RELE(vp); 4035 return (EINVAL); 4036 } 4037 4038 dzp = VTOZ(vp); 4039 zfsvfs = dzp->z_zfsvfs; 4040 ZFS_ENTER(zfsvfs); 4041 4042 /* 4043 * Create share dir if its missing. 4044 */ 4045 mutex_enter(&zfsvfs->z_lock); 4046 if (zfsvfs->z_shares_dir == 0) { 4047 dmu_tx_t *tx; 4048 4049 tx = dmu_tx_create(zfsvfs->z_os); 4050 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE, 4051 ZFS_SHARES_DIR); 4052 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); 4053 error = dmu_tx_assign(tx, TXG_WAIT); 4054 if (error) { 4055 dmu_tx_abort(tx); 4056 } else { 4057 error = zfs_create_share_dir(zfsvfs, tx); 4058 dmu_tx_commit(tx); 4059 } 4060 if (error) { 4061 mutex_exit(&zfsvfs->z_lock); 4062 VN_RELE(vp); 4063 ZFS_EXIT(zfsvfs); 4064 return (error); 4065 } 4066 } 4067 mutex_exit(&zfsvfs->z_lock); 4068 4069 ASSERT(zfsvfs->z_shares_dir); 4070 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) { 4071 VN_RELE(vp); 4072 ZFS_EXIT(zfsvfs); 4073 return (error); 4074 } 4075 4076 switch (zc->zc_cookie) { 4077 case ZFS_SMB_ACL_ADD: 4078 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; 4079 vattr.va_type = VREG; 4080 vattr.va_mode = S_IFREG|0777; 4081 vattr.va_uid = 0; 4082 vattr.va_gid = 0; 4083 4084 vsec.vsa_mask = VSA_ACE; 4085 vsec.vsa_aclentp = &full_access; 4086 vsec.vsa_aclentsz = sizeof (full_access); 4087 vsec.vsa_aclcnt = 1; 4088 4089 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string, 4090 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec); 4091 if (resourcevp) 4092 VN_RELE(resourcevp); 4093 break; 4094 4095 case ZFS_SMB_ACL_REMOVE: 4096 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred, 4097 NULL, 0); 4098 break; 4099 4100 case ZFS_SMB_ACL_RENAME: 4101 if ((error = get_nvlist(zc->zc_nvlist_src, 4102 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) { 4103 VN_RELE(vp); 4104 ZFS_EXIT(zfsvfs); 4105 return (error); 4106 } 4107 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) || 4108 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET, 4109 &target)) { 4110 VN_RELE(vp); 4111 VN_RELE(ZTOV(sharedir)); 4112 ZFS_EXIT(zfsvfs); 4113 return (error); 4114 } 4115 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target, 4116 kcred, NULL, 0); 4117 nvlist_free(nvlist); 4118 break; 4119 4120 case ZFS_SMB_ACL_PURGE: 4121 error = zfs_smb_acl_purge(sharedir); 4122 break; 4123 4124 default: 4125 error = EINVAL; 4126 break; 4127 } 4128 4129 VN_RELE(vp); 4130 VN_RELE(ZTOV(sharedir)); 4131 4132 ZFS_EXIT(zfsvfs); 4133 4134 return (error); 4135 } 4136 4137 /* 4138 * inputs: 4139 * zc_name name of filesystem 4140 * zc_value short name of snap 4141 * zc_string user-supplied tag for this reference 4142 * zc_cookie recursive flag 4143 * zc_temphold set if hold is temporary 4144 * 4145 * outputs: none 4146 */ 4147 static int 4148 zfs_ioc_hold(zfs_cmd_t *zc) 4149 { 4150 boolean_t recursive = zc->zc_cookie; 4151 4152 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0) 4153 return (EINVAL); 4154 4155 return (dsl_dataset_user_hold(zc->zc_name, zc->zc_value, 4156 zc->zc_string, recursive, zc->zc_temphold)); 4157 } 4158 4159 /* 4160 * inputs: 4161 * zc_name name of dataset from which we're releasing a user reference 4162 * zc_value short name of snap 4163 * zc_string user-supplied tag for this reference 4164 * zc_cookie recursive flag 4165 * 4166 * outputs: none 4167 */ 4168 static int 4169 zfs_ioc_release(zfs_cmd_t *zc) 4170 { 4171 boolean_t recursive = zc->zc_cookie; 4172 4173 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0) 4174 return (EINVAL); 4175 4176 return (dsl_dataset_user_release(zc->zc_name, zc->zc_value, 4177 zc->zc_string, recursive)); 4178 } 4179 4180 /* 4181 * inputs: 4182 * zc_name name of filesystem 4183 * 4184 * outputs: 4185 * zc_nvlist_src{_size} nvlist of snapshot holds 4186 */ 4187 static int 4188 zfs_ioc_get_holds(zfs_cmd_t *zc) 4189 { 4190 nvlist_t *nvp; 4191 int error; 4192 4193 if ((error = dsl_dataset_get_holds(zc->zc_name, &nvp)) == 0) { 4194 error = put_nvlist(zc, nvp); 4195 nvlist_free(nvp); 4196 } 4197 4198 return (error); 4199 } 4200 4201 /* 4202 * pool create, destroy, and export don't log the history as part of 4203 * zfsdev_ioctl, but rather zfs_ioc_pool_create, and zfs_ioc_pool_export 4204 * do the logging of those commands. 4205 */ 4206 static zfs_ioc_vec_t zfs_ioc_vec[] = { 4207 { zfs_ioc_pool_create, zfs_secpolicy_config, POOL_NAME, B_FALSE, 4208 B_FALSE }, 4209 { zfs_ioc_pool_destroy, zfs_secpolicy_config, POOL_NAME, B_FALSE, 4210 B_FALSE }, 4211 { zfs_ioc_pool_import, zfs_secpolicy_config, POOL_NAME, B_TRUE, 4212 B_FALSE }, 4213 { zfs_ioc_pool_export, zfs_secpolicy_config, POOL_NAME, B_FALSE, 4214 B_FALSE }, 4215 { zfs_ioc_pool_configs, zfs_secpolicy_none, NO_NAME, B_FALSE, 4216 B_FALSE }, 4217 { zfs_ioc_pool_stats, zfs_secpolicy_read, POOL_NAME, B_FALSE, 4218 B_FALSE }, 4219 { zfs_ioc_pool_tryimport, zfs_secpolicy_config, NO_NAME, B_FALSE, 4220 B_FALSE }, 4221 { zfs_ioc_pool_scrub, zfs_secpolicy_config, POOL_NAME, B_TRUE, 4222 B_TRUE }, 4223 { zfs_ioc_pool_freeze, zfs_secpolicy_config, NO_NAME, B_FALSE, 4224 B_FALSE }, 4225 { zfs_ioc_pool_upgrade, zfs_secpolicy_config, POOL_NAME, B_TRUE, 4226 B_TRUE }, 4227 { zfs_ioc_pool_get_history, zfs_secpolicy_config, POOL_NAME, B_FALSE, 4228 B_FALSE }, 4229 { zfs_ioc_vdev_add, zfs_secpolicy_config, POOL_NAME, B_TRUE, 4230 B_TRUE }, 4231 { zfs_ioc_vdev_remove, zfs_secpolicy_config, POOL_NAME, B_TRUE, 4232 B_TRUE }, 4233 { zfs_ioc_vdev_set_state, zfs_secpolicy_config, POOL_NAME, B_TRUE, 4234 B_FALSE }, 4235 { zfs_ioc_vdev_attach, zfs_secpolicy_config, POOL_NAME, B_TRUE, 4236 B_TRUE }, 4237 { zfs_ioc_vdev_detach, zfs_secpolicy_config, POOL_NAME, B_TRUE, 4238 B_TRUE }, 4239 { zfs_ioc_vdev_setpath, zfs_secpolicy_config, POOL_NAME, B_FALSE, 4240 B_TRUE }, 4241 { zfs_ioc_vdev_setfru, zfs_secpolicy_config, POOL_NAME, B_FALSE, 4242 B_TRUE }, 4243 { zfs_ioc_objset_stats, zfs_secpolicy_read, DATASET_NAME, B_FALSE, 4244 B_FALSE }, 4245 { zfs_ioc_objset_zplprops, zfs_secpolicy_read, DATASET_NAME, B_FALSE, 4246 B_FALSE }, 4247 { zfs_ioc_dataset_list_next, zfs_secpolicy_read, DATASET_NAME, B_FALSE, 4248 B_FALSE }, 4249 { zfs_ioc_snapshot_list_next, zfs_secpolicy_read, DATASET_NAME, B_FALSE, 4250 B_FALSE }, 4251 { zfs_ioc_set_prop, zfs_secpolicy_none, DATASET_NAME, B_TRUE, B_TRUE }, 4252 { zfs_ioc_create, zfs_secpolicy_create, DATASET_NAME, B_TRUE, B_TRUE }, 4253 { zfs_ioc_destroy, zfs_secpolicy_destroy, DATASET_NAME, B_TRUE, 4254 B_TRUE}, 4255 { zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, B_TRUE, 4256 B_TRUE }, 4257 { zfs_ioc_rename, zfs_secpolicy_rename, DATASET_NAME, B_TRUE, B_TRUE }, 4258 { zfs_ioc_recv, zfs_secpolicy_receive, DATASET_NAME, B_TRUE, B_TRUE }, 4259 { zfs_ioc_send, zfs_secpolicy_send, DATASET_NAME, B_TRUE, B_FALSE }, 4260 { zfs_ioc_inject_fault, zfs_secpolicy_inject, NO_NAME, B_FALSE, 4261 B_FALSE }, 4262 { zfs_ioc_clear_fault, zfs_secpolicy_inject, NO_NAME, B_FALSE, 4263 B_FALSE }, 4264 { zfs_ioc_inject_list_next, zfs_secpolicy_inject, NO_NAME, B_FALSE, 4265 B_FALSE }, 4266 { zfs_ioc_error_log, zfs_secpolicy_inject, POOL_NAME, B_FALSE, 4267 B_FALSE }, 4268 { zfs_ioc_clear, zfs_secpolicy_config, POOL_NAME, B_TRUE, B_FALSE }, 4269 { zfs_ioc_promote, zfs_secpolicy_promote, DATASET_NAME, B_TRUE, 4270 B_TRUE }, 4271 { zfs_ioc_destroy_snaps, zfs_secpolicy_destroy, DATASET_NAME, B_TRUE, 4272 B_TRUE }, 4273 { zfs_ioc_snapshot, zfs_secpolicy_snapshot, DATASET_NAME, B_TRUE, 4274 B_TRUE }, 4275 { zfs_ioc_dsobj_to_dsname, zfs_secpolicy_config, POOL_NAME, B_FALSE, 4276 B_FALSE }, 4277 { zfs_ioc_obj_to_path, zfs_secpolicy_config, DATASET_NAME, B_FALSE, 4278 B_TRUE }, 4279 { zfs_ioc_pool_set_props, zfs_secpolicy_config, POOL_NAME, B_TRUE, 4280 B_TRUE }, 4281 { zfs_ioc_pool_get_props, zfs_secpolicy_read, POOL_NAME, B_FALSE, 4282 B_FALSE }, 4283 { zfs_ioc_set_fsacl, zfs_secpolicy_fsacl, DATASET_NAME, B_TRUE, 4284 B_TRUE }, 4285 { zfs_ioc_get_fsacl, zfs_secpolicy_read, DATASET_NAME, B_FALSE, 4286 B_FALSE }, 4287 { zfs_ioc_iscsi_perm_check, zfs_secpolicy_iscsi, DATASET_NAME, B_FALSE, 4288 B_FALSE }, 4289 { zfs_ioc_share, zfs_secpolicy_share, DATASET_NAME, B_FALSE, B_FALSE }, 4290 { zfs_ioc_inherit_prop, zfs_secpolicy_inherit, DATASET_NAME, B_TRUE, 4291 B_TRUE }, 4292 { zfs_ioc_smb_acl, zfs_secpolicy_smb_acl, DATASET_NAME, B_FALSE, 4293 B_FALSE }, 4294 { zfs_ioc_userspace_one, zfs_secpolicy_userspace_one, 4295 DATASET_NAME, B_FALSE, B_FALSE }, 4296 { zfs_ioc_userspace_many, zfs_secpolicy_userspace_many, 4297 DATASET_NAME, B_FALSE, B_FALSE }, 4298 { zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade, 4299 DATASET_NAME, B_FALSE, B_TRUE }, 4300 { zfs_ioc_hold, zfs_secpolicy_hold, DATASET_NAME, B_TRUE, B_TRUE }, 4301 { zfs_ioc_release, zfs_secpolicy_release, DATASET_NAME, B_TRUE, 4302 B_TRUE }, 4303 { zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME, B_FALSE, 4304 B_TRUE }, 4305 { zfs_ioc_objset_recvd_props, zfs_secpolicy_read, DATASET_NAME, B_FALSE, 4306 B_FALSE } 4307 }; 4308 4309 int 4310 pool_status_check(const char *name, zfs_ioc_namecheck_t type) 4311 { 4312 spa_t *spa; 4313 int error; 4314 4315 ASSERT(type == POOL_NAME || type == DATASET_NAME); 4316 4317 error = spa_open(name, &spa, FTAG); 4318 if (error == 0) { 4319 if (spa_suspended(spa)) 4320 error = EAGAIN; 4321 spa_close(spa, FTAG); 4322 } 4323 return (error); 4324 } 4325 4326 static int 4327 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp) 4328 { 4329 zfs_cmd_t *zc; 4330 uint_t vec; 4331 int error, rc; 4332 4333 if (getminor(dev) != 0) 4334 return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp)); 4335 4336 vec = cmd - ZFS_IOC; 4337 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip)); 4338 4339 if (vec >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0])) 4340 return (EINVAL); 4341 4342 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP); 4343 4344 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag); 4345 4346 if ((error == 0) && !(flag & FKIOCTL)) 4347 error = zfs_ioc_vec[vec].zvec_secpolicy(zc, cr); 4348 4349 /* 4350 * Ensure that all pool/dataset names are valid before we pass down to 4351 * the lower layers. 4352 */ 4353 if (error == 0) { 4354 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0'; 4355 zc->zc_iflags = flag & FKIOCTL; 4356 switch (zfs_ioc_vec[vec].zvec_namecheck) { 4357 case POOL_NAME: 4358 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0) 4359 error = EINVAL; 4360 if (zfs_ioc_vec[vec].zvec_pool_check) 4361 error = pool_status_check(zc->zc_name, 4362 zfs_ioc_vec[vec].zvec_namecheck); 4363 break; 4364 4365 case DATASET_NAME: 4366 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0) 4367 error = EINVAL; 4368 if (zfs_ioc_vec[vec].zvec_pool_check) 4369 error = pool_status_check(zc->zc_name, 4370 zfs_ioc_vec[vec].zvec_namecheck); 4371 break; 4372 4373 case NO_NAME: 4374 break; 4375 } 4376 } 4377 4378 if (error == 0) 4379 error = zfs_ioc_vec[vec].zvec_func(zc); 4380 4381 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag); 4382 if (error == 0) { 4383 error = rc; 4384 if (zfs_ioc_vec[vec].zvec_his_log) 4385 zfs_log_history(zc); 4386 } 4387 4388 kmem_free(zc, sizeof (zfs_cmd_t)); 4389 return (error); 4390 } 4391 4392 static int 4393 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd) 4394 { 4395 if (cmd != DDI_ATTACH) 4396 return (DDI_FAILURE); 4397 4398 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0, 4399 DDI_PSEUDO, 0) == DDI_FAILURE) 4400 return (DDI_FAILURE); 4401 4402 zfs_dip = dip; 4403 4404 ddi_report_dev(dip); 4405 4406 return (DDI_SUCCESS); 4407 } 4408 4409 static int 4410 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd) 4411 { 4412 if (spa_busy() || zfs_busy() || zvol_busy()) 4413 return (DDI_FAILURE); 4414 4415 if (cmd != DDI_DETACH) 4416 return (DDI_FAILURE); 4417 4418 zfs_dip = NULL; 4419 4420 ddi_prop_remove_all(dip); 4421 ddi_remove_minor_node(dip, NULL); 4422 4423 return (DDI_SUCCESS); 4424 } 4425 4426 /*ARGSUSED*/ 4427 static int 4428 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result) 4429 { 4430 switch (infocmd) { 4431 case DDI_INFO_DEVT2DEVINFO: 4432 *result = zfs_dip; 4433 return (DDI_SUCCESS); 4434 4435 case DDI_INFO_DEVT2INSTANCE: 4436 *result = (void *)0; 4437 return (DDI_SUCCESS); 4438 } 4439 4440 return (DDI_FAILURE); 4441 } 4442 4443 /* 4444 * OK, so this is a little weird. 4445 * 4446 * /dev/zfs is the control node, i.e. minor 0. 4447 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0. 4448 * 4449 * /dev/zfs has basically nothing to do except serve up ioctls, 4450 * so most of the standard driver entry points are in zvol.c. 4451 */ 4452 static struct cb_ops zfs_cb_ops = { 4453 zvol_open, /* open */ 4454 zvol_close, /* close */ 4455 zvol_strategy, /* strategy */ 4456 nodev, /* print */ 4457 zvol_dump, /* dump */ 4458 zvol_read, /* read */ 4459 zvol_write, /* write */ 4460 zfsdev_ioctl, /* ioctl */ 4461 nodev, /* devmap */ 4462 nodev, /* mmap */ 4463 nodev, /* segmap */ 4464 nochpoll, /* poll */ 4465 ddi_prop_op, /* prop_op */ 4466 NULL, /* streamtab */ 4467 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */ 4468 CB_REV, /* version */ 4469 nodev, /* async read */ 4470 nodev, /* async write */ 4471 }; 4472 4473 static struct dev_ops zfs_dev_ops = { 4474 DEVO_REV, /* version */ 4475 0, /* refcnt */ 4476 zfs_info, /* info */ 4477 nulldev, /* identify */ 4478 nulldev, /* probe */ 4479 zfs_attach, /* attach */ 4480 zfs_detach, /* detach */ 4481 nodev, /* reset */ 4482 &zfs_cb_ops, /* driver operations */ 4483 NULL, /* no bus operations */ 4484 NULL, /* power */ 4485 ddi_quiesce_not_needed, /* quiesce */ 4486 }; 4487 4488 static struct modldrv zfs_modldrv = { 4489 &mod_driverops, 4490 "ZFS storage pool", 4491 &zfs_dev_ops 4492 }; 4493 4494 static struct modlinkage modlinkage = { 4495 MODREV_1, 4496 (void *)&zfs_modlfs, 4497 (void *)&zfs_modldrv, 4498 NULL 4499 }; 4500 4501 4502 uint_t zfs_fsyncer_key; 4503 extern uint_t rrw_tsd_key; 4504 4505 int 4506 _init(void) 4507 { 4508 int error; 4509 4510 spa_init(FREAD | FWRITE); 4511 zfs_init(); 4512 zvol_init(); 4513 4514 if ((error = mod_install(&modlinkage)) != 0) { 4515 zvol_fini(); 4516 zfs_fini(); 4517 spa_fini(); 4518 return (error); 4519 } 4520 4521 tsd_create(&zfs_fsyncer_key, NULL); 4522 tsd_create(&rrw_tsd_key, NULL); 4523 4524 error = ldi_ident_from_mod(&modlinkage, &zfs_li); 4525 ASSERT(error == 0); 4526 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL); 4527 4528 return (0); 4529 } 4530 4531 int 4532 _fini(void) 4533 { 4534 int error; 4535 4536 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled) 4537 return (EBUSY); 4538 4539 if ((error = mod_remove(&modlinkage)) != 0) 4540 return (error); 4541 4542 zvol_fini(); 4543 zfs_fini(); 4544 spa_fini(); 4545 if (zfs_nfsshare_inited) 4546 (void) ddi_modclose(nfs_mod); 4547 if (zfs_smbshare_inited) 4548 (void) ddi_modclose(smbsrv_mod); 4549 if (zfs_nfsshare_inited || zfs_smbshare_inited) 4550 (void) ddi_modclose(sharefs_mod); 4551 4552 tsd_destroy(&zfs_fsyncer_key); 4553 ldi_ident_release(zfs_li); 4554 zfs_li = NULL; 4555 mutex_destroy(&zfs_share_lock); 4556 4557 return (error); 4558 } 4559 4560 int 4561 _info(struct modinfo *modinfop) 4562 { 4563 return (mod_info(&modlinkage, modinfop)); 4564 } 4565