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 = { 0 }; 1894 (void) strcpy(zc.zc_name, dsname); 1895 (void) zfs_ioc_userspace_upgrade(&zc); 1896 } 1897 break; 1898 } 1899 1900 default: 1901 err = -1; 1902 } 1903 1904 return (err); 1905 } 1906 1907 /* 1908 * This function is best effort. If it fails to set any of the given properties, 1909 * it continues to set as many as it can and returns the first error 1910 * encountered. If the caller provides a non-NULL errlist, it also gives the 1911 * complete list of names of all the properties it failed to set along with the 1912 * corresponding error numbers. The caller is responsible for freeing the 1913 * returned errlist. 1914 * 1915 * If every property is set successfully, zero is returned and the list pointed 1916 * at by errlist is NULL. 1917 */ 1918 int 1919 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl, 1920 nvlist_t **errlist) 1921 { 1922 nvpair_t *pair; 1923 nvpair_t *propval; 1924 int err = 0; 1925 int rv = 0; 1926 uint64_t intval; 1927 char *strval; 1928 nvlist_t *genericnvl; 1929 nvlist_t *errors; 1930 nvlist_t *retrynvl; 1931 1932 VERIFY(nvlist_alloc(&genericnvl, NV_UNIQUE_NAME, KM_SLEEP) == 0); 1933 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0); 1934 VERIFY(nvlist_alloc(&retrynvl, NV_UNIQUE_NAME, KM_SLEEP) == 0); 1935 1936 retry: 1937 pair = NULL; 1938 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { 1939 const char *propname = nvpair_name(pair); 1940 zfs_prop_t prop = zfs_name_to_prop(propname); 1941 1942 /* decode the property value */ 1943 propval = pair; 1944 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 1945 nvlist_t *attrs; 1946 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 1947 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 1948 &propval) == 0); 1949 } 1950 1951 /* Validate value type */ 1952 if (prop == ZPROP_INVAL) { 1953 if (zfs_prop_user(propname)) { 1954 if (nvpair_type(propval) != DATA_TYPE_STRING) 1955 err = EINVAL; 1956 } else if (zfs_prop_userquota(propname)) { 1957 if (nvpair_type(propval) != 1958 DATA_TYPE_UINT64_ARRAY) 1959 err = EINVAL; 1960 } 1961 } else { 1962 if (nvpair_type(propval) == DATA_TYPE_STRING) { 1963 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING) 1964 err = EINVAL; 1965 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) { 1966 const char *unused; 1967 1968 VERIFY(nvpair_value_uint64(propval, 1969 &intval) == 0); 1970 1971 switch (zfs_prop_get_type(prop)) { 1972 case PROP_TYPE_NUMBER: 1973 break; 1974 case PROP_TYPE_STRING: 1975 err = EINVAL; 1976 break; 1977 case PROP_TYPE_INDEX: 1978 if (zfs_prop_index_to_string(prop, 1979 intval, &unused) != 0) 1980 err = EINVAL; 1981 break; 1982 default: 1983 cmn_err(CE_PANIC, 1984 "unknown property type"); 1985 } 1986 } else { 1987 err = EINVAL; 1988 } 1989 } 1990 1991 /* Validate permissions */ 1992 if (err == 0) 1993 err = zfs_check_settable(dsname, pair, CRED()); 1994 1995 if (err == 0) { 1996 err = zfs_prop_set_special(dsname, source, pair); 1997 if (err == -1) { 1998 /* 1999 * For better performance we build up a list of 2000 * properties to set in a single transaction. 2001 */ 2002 err = nvlist_add_nvpair(genericnvl, pair); 2003 } else if (err != 0 && nvl != retrynvl) { 2004 /* 2005 * This may be a spurious error caused by 2006 * receiving quota and reservation out of order. 2007 * Try again in a second pass. 2008 */ 2009 err = nvlist_add_nvpair(retrynvl, pair); 2010 } 2011 } 2012 2013 if (err != 0) 2014 VERIFY(nvlist_add_int32(errors, propname, err) == 0); 2015 } 2016 2017 if (nvl != retrynvl && !nvlist_empty(retrynvl)) { 2018 nvl = retrynvl; 2019 goto retry; 2020 } 2021 2022 if (!nvlist_empty(genericnvl) && 2023 dsl_props_set(dsname, source, genericnvl) != 0) { 2024 /* 2025 * If this fails, we still want to set as many properties as we 2026 * can, so try setting them individually. 2027 */ 2028 pair = NULL; 2029 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) { 2030 const char *propname = nvpair_name(pair); 2031 2032 propval = pair; 2033 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2034 nvlist_t *attrs; 2035 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 2036 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2037 &propval) == 0); 2038 } 2039 2040 if (nvpair_type(propval) == DATA_TYPE_STRING) { 2041 VERIFY(nvpair_value_string(propval, 2042 &strval) == 0); 2043 err = dsl_prop_set(dsname, propname, source, 1, 2044 strlen(strval) + 1, strval); 2045 } else { 2046 VERIFY(nvpair_value_uint64(propval, 2047 &intval) == 0); 2048 err = dsl_prop_set(dsname, propname, source, 8, 2049 1, &intval); 2050 } 2051 2052 if (err != 0) { 2053 VERIFY(nvlist_add_int32(errors, propname, 2054 err) == 0); 2055 } 2056 } 2057 } 2058 nvlist_free(genericnvl); 2059 nvlist_free(retrynvl); 2060 2061 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) { 2062 nvlist_free(errors); 2063 errors = NULL; 2064 } else { 2065 VERIFY(nvpair_value_int32(pair, &rv) == 0); 2066 } 2067 2068 if (errlist == NULL) 2069 nvlist_free(errors); 2070 else 2071 *errlist = errors; 2072 2073 return (rv); 2074 } 2075 2076 /* 2077 * Check that all the properties are valid user properties. 2078 */ 2079 static int 2080 zfs_check_userprops(char *fsname, nvlist_t *nvl) 2081 { 2082 nvpair_t *pair = NULL; 2083 int error = 0; 2084 2085 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { 2086 const char *propname = nvpair_name(pair); 2087 char *valstr; 2088 2089 if (!zfs_prop_user(propname) || 2090 nvpair_type(pair) != DATA_TYPE_STRING) 2091 return (EINVAL); 2092 2093 if (error = zfs_secpolicy_write_perms(fsname, 2094 ZFS_DELEG_PERM_USERPROP, CRED())) 2095 return (error); 2096 2097 if (strlen(propname) >= ZAP_MAXNAMELEN) 2098 return (ENAMETOOLONG); 2099 2100 VERIFY(nvpair_value_string(pair, &valstr) == 0); 2101 if (strlen(valstr) >= ZAP_MAXVALUELEN) 2102 return (E2BIG); 2103 } 2104 return (0); 2105 } 2106 2107 static void 2108 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops) 2109 { 2110 nvpair_t *pair; 2111 2112 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2113 2114 pair = NULL; 2115 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) { 2116 if (nvlist_exists(skipped, nvpair_name(pair))) 2117 continue; 2118 2119 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0); 2120 } 2121 } 2122 2123 static int 2124 clear_received_props(objset_t *os, const char *fs, nvlist_t *props, 2125 nvlist_t *skipped) 2126 { 2127 int err = 0; 2128 nvlist_t *cleared_props = NULL; 2129 props_skip(props, skipped, &cleared_props); 2130 if (!nvlist_empty(cleared_props)) { 2131 /* 2132 * Acts on local properties until the dataset has received 2133 * properties at least once on or after SPA_VERSION_RECVD_PROPS. 2134 */ 2135 zprop_source_t flags = (ZPROP_SRC_NONE | 2136 (dsl_prop_get_hasrecvd(os) ? ZPROP_SRC_RECEIVED : 0)); 2137 err = zfs_set_prop_nvlist(fs, flags, cleared_props, NULL); 2138 } 2139 nvlist_free(cleared_props); 2140 return (err); 2141 } 2142 2143 /* 2144 * inputs: 2145 * zc_name name of filesystem 2146 * zc_value name of property to set 2147 * zc_nvlist_src{_size} nvlist of properties to apply 2148 * zc_cookie received properties flag 2149 * 2150 * outputs: 2151 * zc_nvlist_dst{_size} error for each unapplied received property 2152 */ 2153 static int 2154 zfs_ioc_set_prop(zfs_cmd_t *zc) 2155 { 2156 nvlist_t *nvl; 2157 boolean_t received = zc->zc_cookie; 2158 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED : 2159 ZPROP_SRC_LOCAL); 2160 nvlist_t *errors = NULL; 2161 int error; 2162 2163 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2164 zc->zc_iflags, &nvl)) != 0) 2165 return (error); 2166 2167 if (received) { 2168 nvlist_t *origprops; 2169 objset_t *os; 2170 2171 if (dmu_objset_hold(zc->zc_name, FTAG, &os) == 0) { 2172 if (dsl_prop_get_received(os, &origprops) == 0) { 2173 (void) clear_received_props(os, 2174 zc->zc_name, origprops, nvl); 2175 nvlist_free(origprops); 2176 } 2177 2178 dsl_prop_set_hasrecvd(os); 2179 dmu_objset_rele(os, FTAG); 2180 } 2181 } 2182 2183 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, &errors); 2184 2185 if (zc->zc_nvlist_dst != NULL && errors != NULL) { 2186 (void) put_nvlist(zc, errors); 2187 } 2188 2189 nvlist_free(errors); 2190 nvlist_free(nvl); 2191 return (error); 2192 } 2193 2194 /* 2195 * inputs: 2196 * zc_name name of filesystem 2197 * zc_value name of property to inherit 2198 * zc_cookie revert to received value if TRUE 2199 * 2200 * outputs: none 2201 */ 2202 static int 2203 zfs_ioc_inherit_prop(zfs_cmd_t *zc) 2204 { 2205 const char *propname = zc->zc_value; 2206 zfs_prop_t prop = zfs_name_to_prop(propname); 2207 boolean_t received = zc->zc_cookie; 2208 zprop_source_t source = (received 2209 ? ZPROP_SRC_NONE /* revert to received value, if any */ 2210 : ZPROP_SRC_INHERITED); /* explicitly inherit */ 2211 2212 if (received) { 2213 nvlist_t *dummy; 2214 nvpair_t *pair; 2215 zprop_type_t type; 2216 int err; 2217 2218 /* 2219 * zfs_prop_set_special() expects properties in the form of an 2220 * nvpair with type info. 2221 */ 2222 if (prop == ZPROP_INVAL) { 2223 if (!zfs_prop_user(propname)) 2224 return (EINVAL); 2225 2226 type = PROP_TYPE_STRING; 2227 } else { 2228 type = zfs_prop_get_type(prop); 2229 } 2230 2231 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2232 2233 switch (type) { 2234 case PROP_TYPE_STRING: 2235 VERIFY(0 == nvlist_add_string(dummy, propname, "")); 2236 break; 2237 case PROP_TYPE_NUMBER: 2238 case PROP_TYPE_INDEX: 2239 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0)); 2240 break; 2241 default: 2242 nvlist_free(dummy); 2243 return (EINVAL); 2244 } 2245 2246 pair = nvlist_next_nvpair(dummy, NULL); 2247 err = zfs_prop_set_special(zc->zc_name, source, pair); 2248 nvlist_free(dummy); 2249 if (err != -1) 2250 return (err); /* special property already handled */ 2251 } else { 2252 /* 2253 * Only check this in the non-received case. We want to allow 2254 * 'inherit -S' to revert non-inheritable properties like quota 2255 * and reservation to the received or default values even though 2256 * they are not considered inheritable. 2257 */ 2258 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop)) 2259 return (EINVAL); 2260 } 2261 2262 /* the property name has been validated by zfs_secpolicy_inherit() */ 2263 return (dsl_prop_set(zc->zc_name, zc->zc_value, source, 0, 0, NULL)); 2264 } 2265 2266 static int 2267 zfs_ioc_pool_set_props(zfs_cmd_t *zc) 2268 { 2269 nvlist_t *props; 2270 spa_t *spa; 2271 int error; 2272 nvpair_t *pair; 2273 2274 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2275 zc->zc_iflags, &props)) 2276 return (error); 2277 2278 /* 2279 * If the only property is the configfile, then just do a spa_lookup() 2280 * to handle the faulted case. 2281 */ 2282 pair = nvlist_next_nvpair(props, NULL); 2283 if (pair != NULL && strcmp(nvpair_name(pair), 2284 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 && 2285 nvlist_next_nvpair(props, pair) == NULL) { 2286 mutex_enter(&spa_namespace_lock); 2287 if ((spa = spa_lookup(zc->zc_name)) != NULL) { 2288 spa_configfile_set(spa, props, B_FALSE); 2289 spa_config_sync(spa, B_FALSE, B_TRUE); 2290 } 2291 mutex_exit(&spa_namespace_lock); 2292 if (spa != NULL) { 2293 nvlist_free(props); 2294 return (0); 2295 } 2296 } 2297 2298 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { 2299 nvlist_free(props); 2300 return (error); 2301 } 2302 2303 error = spa_prop_set(spa, props); 2304 2305 nvlist_free(props); 2306 spa_close(spa, FTAG); 2307 2308 return (error); 2309 } 2310 2311 static int 2312 zfs_ioc_pool_get_props(zfs_cmd_t *zc) 2313 { 2314 spa_t *spa; 2315 int error; 2316 nvlist_t *nvp = NULL; 2317 2318 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { 2319 /* 2320 * If the pool is faulted, there may be properties we can still 2321 * get (such as altroot and cachefile), so attempt to get them 2322 * anyway. 2323 */ 2324 mutex_enter(&spa_namespace_lock); 2325 if ((spa = spa_lookup(zc->zc_name)) != NULL) 2326 error = spa_prop_get(spa, &nvp); 2327 mutex_exit(&spa_namespace_lock); 2328 } else { 2329 error = spa_prop_get(spa, &nvp); 2330 spa_close(spa, FTAG); 2331 } 2332 2333 if (error == 0 && zc->zc_nvlist_dst != NULL) 2334 error = put_nvlist(zc, nvp); 2335 else 2336 error = EFAULT; 2337 2338 nvlist_free(nvp); 2339 return (error); 2340 } 2341 2342 static int 2343 zfs_ioc_iscsi_perm_check(zfs_cmd_t *zc) 2344 { 2345 nvlist_t *nvp; 2346 int error; 2347 uint32_t uid; 2348 uint32_t gid; 2349 uint32_t *groups; 2350 uint_t group_cnt; 2351 cred_t *usercred; 2352 2353 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2354 zc->zc_iflags, &nvp)) != 0) { 2355 return (error); 2356 } 2357 2358 if ((error = nvlist_lookup_uint32(nvp, 2359 ZFS_DELEG_PERM_UID, &uid)) != 0) { 2360 nvlist_free(nvp); 2361 return (EPERM); 2362 } 2363 2364 if ((error = nvlist_lookup_uint32(nvp, 2365 ZFS_DELEG_PERM_GID, &gid)) != 0) { 2366 nvlist_free(nvp); 2367 return (EPERM); 2368 } 2369 2370 if ((error = nvlist_lookup_uint32_array(nvp, ZFS_DELEG_PERM_GROUPS, 2371 &groups, &group_cnt)) != 0) { 2372 nvlist_free(nvp); 2373 return (EPERM); 2374 } 2375 usercred = cralloc(); 2376 if ((crsetugid(usercred, uid, gid) != 0) || 2377 (crsetgroups(usercred, group_cnt, (gid_t *)groups) != 0)) { 2378 nvlist_free(nvp); 2379 crfree(usercred); 2380 return (EPERM); 2381 } 2382 nvlist_free(nvp); 2383 error = dsl_deleg_access(zc->zc_name, 2384 zfs_prop_to_name(ZFS_PROP_SHAREISCSI), usercred); 2385 crfree(usercred); 2386 return (error); 2387 } 2388 2389 /* 2390 * inputs: 2391 * zc_name name of filesystem 2392 * zc_nvlist_src{_size} nvlist of delegated permissions 2393 * zc_perm_action allow/unallow flag 2394 * 2395 * outputs: none 2396 */ 2397 static int 2398 zfs_ioc_set_fsacl(zfs_cmd_t *zc) 2399 { 2400 int error; 2401 nvlist_t *fsaclnv = NULL; 2402 2403 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2404 zc->zc_iflags, &fsaclnv)) != 0) 2405 return (error); 2406 2407 /* 2408 * Verify nvlist is constructed correctly 2409 */ 2410 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) { 2411 nvlist_free(fsaclnv); 2412 return (EINVAL); 2413 } 2414 2415 /* 2416 * If we don't have PRIV_SYS_MOUNT, then validate 2417 * that user is allowed to hand out each permission in 2418 * the nvlist(s) 2419 */ 2420 2421 error = secpolicy_zfs(CRED()); 2422 if (error) { 2423 if (zc->zc_perm_action == B_FALSE) { 2424 error = dsl_deleg_can_allow(zc->zc_name, 2425 fsaclnv, CRED()); 2426 } else { 2427 error = dsl_deleg_can_unallow(zc->zc_name, 2428 fsaclnv, CRED()); 2429 } 2430 } 2431 2432 if (error == 0) 2433 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action); 2434 2435 nvlist_free(fsaclnv); 2436 return (error); 2437 } 2438 2439 /* 2440 * inputs: 2441 * zc_name name of filesystem 2442 * 2443 * outputs: 2444 * zc_nvlist_src{_size} nvlist of delegated permissions 2445 */ 2446 static int 2447 zfs_ioc_get_fsacl(zfs_cmd_t *zc) 2448 { 2449 nvlist_t *nvp; 2450 int error; 2451 2452 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) { 2453 error = put_nvlist(zc, nvp); 2454 nvlist_free(nvp); 2455 } 2456 2457 return (error); 2458 } 2459 2460 /* 2461 * Search the vfs list for a specified resource. Returns a pointer to it 2462 * or NULL if no suitable entry is found. The caller of this routine 2463 * is responsible for releasing the returned vfs pointer. 2464 */ 2465 static vfs_t * 2466 zfs_get_vfs(const char *resource) 2467 { 2468 struct vfs *vfsp; 2469 struct vfs *vfs_found = NULL; 2470 2471 vfs_list_read_lock(); 2472 vfsp = rootvfs; 2473 do { 2474 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) { 2475 VFS_HOLD(vfsp); 2476 vfs_found = vfsp; 2477 break; 2478 } 2479 vfsp = vfsp->vfs_next; 2480 } while (vfsp != rootvfs); 2481 vfs_list_unlock(); 2482 return (vfs_found); 2483 } 2484 2485 /* ARGSUSED */ 2486 static void 2487 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) 2488 { 2489 zfs_creat_t *zct = arg; 2490 2491 zfs_create_fs(os, cr, zct->zct_zplprops, tx); 2492 } 2493 2494 #define ZFS_PROP_UNDEFINED ((uint64_t)-1) 2495 2496 /* 2497 * inputs: 2498 * createprops list of properties requested by creator 2499 * default_zplver zpl version to use if unspecified in createprops 2500 * fuids_ok fuids allowed in this version of the spa? 2501 * os parent objset pointer (NULL if root fs) 2502 * 2503 * outputs: 2504 * zplprops values for the zplprops we attach to the master node object 2505 * is_ci true if requested file system will be purely case-insensitive 2506 * 2507 * Determine the settings for utf8only, normalization and 2508 * casesensitivity. Specific values may have been requested by the 2509 * creator and/or we can inherit values from the parent dataset. If 2510 * the file system is of too early a vintage, a creator can not 2511 * request settings for these properties, even if the requested 2512 * setting is the default value. We don't actually want to create dsl 2513 * properties for these, so remove them from the source nvlist after 2514 * processing. 2515 */ 2516 static int 2517 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver, 2518 boolean_t fuids_ok, nvlist_t *createprops, nvlist_t *zplprops, 2519 boolean_t *is_ci) 2520 { 2521 uint64_t sense = ZFS_PROP_UNDEFINED; 2522 uint64_t norm = ZFS_PROP_UNDEFINED; 2523 uint64_t u8 = ZFS_PROP_UNDEFINED; 2524 2525 ASSERT(zplprops != NULL); 2526 2527 /* 2528 * Pull out creator prop choices, if any. 2529 */ 2530 if (createprops) { 2531 (void) nvlist_lookup_uint64(createprops, 2532 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver); 2533 (void) nvlist_lookup_uint64(createprops, 2534 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm); 2535 (void) nvlist_remove_all(createprops, 2536 zfs_prop_to_name(ZFS_PROP_NORMALIZE)); 2537 (void) nvlist_lookup_uint64(createprops, 2538 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8); 2539 (void) nvlist_remove_all(createprops, 2540 zfs_prop_to_name(ZFS_PROP_UTF8ONLY)); 2541 (void) nvlist_lookup_uint64(createprops, 2542 zfs_prop_to_name(ZFS_PROP_CASE), &sense); 2543 (void) nvlist_remove_all(createprops, 2544 zfs_prop_to_name(ZFS_PROP_CASE)); 2545 } 2546 2547 /* 2548 * If the zpl version requested is whacky or the file system 2549 * or pool is version is too "young" to support normalization 2550 * and the creator tried to set a value for one of the props, 2551 * error out. 2552 */ 2553 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) || 2554 (zplver >= ZPL_VERSION_FUID && !fuids_ok) || 2555 (zplver < ZPL_VERSION_NORMALIZATION && 2556 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED || 2557 sense != ZFS_PROP_UNDEFINED))) 2558 return (ENOTSUP); 2559 2560 /* 2561 * Put the version in the zplprops 2562 */ 2563 VERIFY(nvlist_add_uint64(zplprops, 2564 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0); 2565 2566 if (norm == ZFS_PROP_UNDEFINED) 2567 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0); 2568 VERIFY(nvlist_add_uint64(zplprops, 2569 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0); 2570 2571 /* 2572 * If we're normalizing, names must always be valid UTF-8 strings. 2573 */ 2574 if (norm) 2575 u8 = 1; 2576 if (u8 == ZFS_PROP_UNDEFINED) 2577 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0); 2578 VERIFY(nvlist_add_uint64(zplprops, 2579 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0); 2580 2581 if (sense == ZFS_PROP_UNDEFINED) 2582 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0); 2583 VERIFY(nvlist_add_uint64(zplprops, 2584 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0); 2585 2586 if (is_ci) 2587 *is_ci = (sense == ZFS_CASE_INSENSITIVE); 2588 2589 return (0); 2590 } 2591 2592 static int 2593 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops, 2594 nvlist_t *zplprops, boolean_t *is_ci) 2595 { 2596 boolean_t fuids_ok = B_TRUE; 2597 uint64_t zplver = ZPL_VERSION; 2598 objset_t *os = NULL; 2599 char parentname[MAXNAMELEN]; 2600 char *cp; 2601 int error; 2602 2603 (void) strlcpy(parentname, dataset, sizeof (parentname)); 2604 cp = strrchr(parentname, '/'); 2605 ASSERT(cp != NULL); 2606 cp[0] = '\0'; 2607 2608 if (zfs_earlier_version(dataset, SPA_VERSION_USERSPACE)) 2609 zplver = ZPL_VERSION_USERSPACE - 1; 2610 if (zfs_earlier_version(dataset, SPA_VERSION_FUID)) { 2611 zplver = ZPL_VERSION_FUID - 1; 2612 fuids_ok = B_FALSE; 2613 } 2614 2615 /* 2616 * Open parent object set so we can inherit zplprop values. 2617 */ 2618 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0) 2619 return (error); 2620 2621 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, createprops, 2622 zplprops, is_ci); 2623 dmu_objset_rele(os, FTAG); 2624 return (error); 2625 } 2626 2627 static int 2628 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops, 2629 nvlist_t *zplprops, boolean_t *is_ci) 2630 { 2631 boolean_t fuids_ok = B_TRUE; 2632 uint64_t zplver = ZPL_VERSION; 2633 int error; 2634 2635 if (spa_vers < SPA_VERSION_FUID) { 2636 zplver = ZPL_VERSION_FUID - 1; 2637 fuids_ok = B_FALSE; 2638 } 2639 2640 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, createprops, 2641 zplprops, is_ci); 2642 return (error); 2643 } 2644 2645 /* 2646 * inputs: 2647 * zc_objset_type type of objset to create (fs vs zvol) 2648 * zc_name name of new objset 2649 * zc_value name of snapshot to clone from (may be empty) 2650 * zc_nvlist_src{_size} nvlist of properties to apply 2651 * 2652 * outputs: none 2653 */ 2654 static int 2655 zfs_ioc_create(zfs_cmd_t *zc) 2656 { 2657 objset_t *clone; 2658 int error = 0; 2659 zfs_creat_t zct; 2660 nvlist_t *nvprops = NULL; 2661 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx); 2662 dmu_objset_type_t type = zc->zc_objset_type; 2663 2664 switch (type) { 2665 2666 case DMU_OST_ZFS: 2667 cbfunc = zfs_create_cb; 2668 break; 2669 2670 case DMU_OST_ZVOL: 2671 cbfunc = zvol_create_cb; 2672 break; 2673 2674 default: 2675 cbfunc = NULL; 2676 break; 2677 } 2678 if (strchr(zc->zc_name, '@') || 2679 strchr(zc->zc_name, '%')) 2680 return (EINVAL); 2681 2682 if (zc->zc_nvlist_src != NULL && 2683 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2684 zc->zc_iflags, &nvprops)) != 0) 2685 return (error); 2686 2687 zct.zct_zplprops = NULL; 2688 zct.zct_props = nvprops; 2689 2690 if (zc->zc_value[0] != '\0') { 2691 /* 2692 * We're creating a clone of an existing snapshot. 2693 */ 2694 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0'; 2695 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0) { 2696 nvlist_free(nvprops); 2697 return (EINVAL); 2698 } 2699 2700 error = dmu_objset_hold(zc->zc_value, FTAG, &clone); 2701 if (error) { 2702 nvlist_free(nvprops); 2703 return (error); 2704 } 2705 2706 error = dmu_objset_clone(zc->zc_name, dmu_objset_ds(clone), 0); 2707 dmu_objset_rele(clone, FTAG); 2708 if (error) { 2709 nvlist_free(nvprops); 2710 return (error); 2711 } 2712 } else { 2713 boolean_t is_insensitive = B_FALSE; 2714 2715 if (cbfunc == NULL) { 2716 nvlist_free(nvprops); 2717 return (EINVAL); 2718 } 2719 2720 if (type == DMU_OST_ZVOL) { 2721 uint64_t volsize, volblocksize; 2722 2723 if (nvprops == NULL || 2724 nvlist_lookup_uint64(nvprops, 2725 zfs_prop_to_name(ZFS_PROP_VOLSIZE), 2726 &volsize) != 0) { 2727 nvlist_free(nvprops); 2728 return (EINVAL); 2729 } 2730 2731 if ((error = nvlist_lookup_uint64(nvprops, 2732 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), 2733 &volblocksize)) != 0 && error != ENOENT) { 2734 nvlist_free(nvprops); 2735 return (EINVAL); 2736 } 2737 2738 if (error != 0) 2739 volblocksize = zfs_prop_default_numeric( 2740 ZFS_PROP_VOLBLOCKSIZE); 2741 2742 if ((error = zvol_check_volblocksize( 2743 volblocksize)) != 0 || 2744 (error = zvol_check_volsize(volsize, 2745 volblocksize)) != 0) { 2746 nvlist_free(nvprops); 2747 return (error); 2748 } 2749 } else if (type == DMU_OST_ZFS) { 2750 int error; 2751 2752 /* 2753 * We have to have normalization and 2754 * case-folding flags correct when we do the 2755 * file system creation, so go figure them out 2756 * now. 2757 */ 2758 VERIFY(nvlist_alloc(&zct.zct_zplprops, 2759 NV_UNIQUE_NAME, KM_SLEEP) == 0); 2760 error = zfs_fill_zplprops(zc->zc_name, nvprops, 2761 zct.zct_zplprops, &is_insensitive); 2762 if (error != 0) { 2763 nvlist_free(nvprops); 2764 nvlist_free(zct.zct_zplprops); 2765 return (error); 2766 } 2767 } 2768 error = dmu_objset_create(zc->zc_name, type, 2769 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct); 2770 nvlist_free(zct.zct_zplprops); 2771 } 2772 2773 /* 2774 * It would be nice to do this atomically. 2775 */ 2776 if (error == 0) { 2777 error = zfs_set_prop_nvlist(zc->zc_name, ZPROP_SRC_LOCAL, 2778 nvprops, NULL); 2779 if (error != 0) 2780 (void) dmu_objset_destroy(zc->zc_name, B_FALSE); 2781 } 2782 nvlist_free(nvprops); 2783 return (error); 2784 } 2785 2786 /* 2787 * inputs: 2788 * zc_name name of filesystem 2789 * zc_value short name of snapshot 2790 * zc_cookie recursive flag 2791 * zc_nvlist_src[_size] property list 2792 * 2793 * outputs: 2794 * zc_value short snapname (i.e. part after the '@') 2795 */ 2796 static int 2797 zfs_ioc_snapshot(zfs_cmd_t *zc) 2798 { 2799 nvlist_t *nvprops = NULL; 2800 int error; 2801 boolean_t recursive = zc->zc_cookie; 2802 2803 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0) 2804 return (EINVAL); 2805 2806 if (zc->zc_nvlist_src != NULL && 2807 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2808 zc->zc_iflags, &nvprops)) != 0) 2809 return (error); 2810 2811 error = zfs_check_userprops(zc->zc_name, nvprops); 2812 if (error) 2813 goto out; 2814 2815 if (!nvlist_empty(nvprops) && 2816 zfs_earlier_version(zc->zc_name, SPA_VERSION_SNAP_PROPS)) { 2817 error = ENOTSUP; 2818 goto out; 2819 } 2820 2821 error = dmu_objset_snapshot(zc->zc_name, zc->zc_value, 2822 nvprops, recursive); 2823 2824 out: 2825 nvlist_free(nvprops); 2826 return (error); 2827 } 2828 2829 int 2830 zfs_unmount_snap(char *name, void *arg) 2831 { 2832 vfs_t *vfsp = NULL; 2833 2834 if (arg) { 2835 char *snapname = arg; 2836 int len = strlen(name) + strlen(snapname) + 2; 2837 char *buf = kmem_alloc(len, KM_SLEEP); 2838 2839 (void) strcpy(buf, name); 2840 (void) strcat(buf, "@"); 2841 (void) strcat(buf, snapname); 2842 vfsp = zfs_get_vfs(buf); 2843 kmem_free(buf, len); 2844 } else if (strchr(name, '@')) { 2845 vfsp = zfs_get_vfs(name); 2846 } 2847 2848 if (vfsp) { 2849 /* 2850 * Always force the unmount for snapshots. 2851 */ 2852 int flag = MS_FORCE; 2853 int err; 2854 2855 if ((err = vn_vfswlock(vfsp->vfs_vnodecovered)) != 0) { 2856 VFS_RELE(vfsp); 2857 return (err); 2858 } 2859 VFS_RELE(vfsp); 2860 if ((err = dounmount(vfsp, flag, kcred)) != 0) 2861 return (err); 2862 } 2863 return (0); 2864 } 2865 2866 /* 2867 * inputs: 2868 * zc_name name of filesystem 2869 * zc_value short name of snapshot 2870 * zc_defer_destroy mark for deferred destroy 2871 * 2872 * outputs: none 2873 */ 2874 static int 2875 zfs_ioc_destroy_snaps(zfs_cmd_t *zc) 2876 { 2877 int err; 2878 2879 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0) 2880 return (EINVAL); 2881 err = dmu_objset_find(zc->zc_name, 2882 zfs_unmount_snap, zc->zc_value, DS_FIND_CHILDREN); 2883 if (err) 2884 return (err); 2885 return (dmu_snapshots_destroy(zc->zc_name, zc->zc_value, 2886 zc->zc_defer_destroy)); 2887 } 2888 2889 /* 2890 * inputs: 2891 * zc_name name of dataset to destroy 2892 * zc_objset_type type of objset 2893 * zc_defer_destroy mark for deferred destroy 2894 * 2895 * outputs: none 2896 */ 2897 static int 2898 zfs_ioc_destroy(zfs_cmd_t *zc) 2899 { 2900 int err; 2901 if (strchr(zc->zc_name, '@') && zc->zc_objset_type == DMU_OST_ZFS) { 2902 err = zfs_unmount_snap(zc->zc_name, NULL); 2903 if (err) 2904 return (err); 2905 } 2906 2907 err = dmu_objset_destroy(zc->zc_name, zc->zc_defer_destroy); 2908 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0) 2909 (void) zvol_remove_minor(zc->zc_name); 2910 return (err); 2911 } 2912 2913 /* 2914 * inputs: 2915 * zc_name name of dataset to rollback (to most recent snapshot) 2916 * 2917 * outputs: none 2918 */ 2919 static int 2920 zfs_ioc_rollback(zfs_cmd_t *zc) 2921 { 2922 dsl_dataset_t *ds, *clone; 2923 int error; 2924 zfsvfs_t *zfsvfs; 2925 char *clone_name; 2926 2927 error = dsl_dataset_hold(zc->zc_name, FTAG, &ds); 2928 if (error) 2929 return (error); 2930 2931 /* must not be a snapshot */ 2932 if (dsl_dataset_is_snapshot(ds)) { 2933 dsl_dataset_rele(ds, FTAG); 2934 return (EINVAL); 2935 } 2936 2937 /* must have a most recent snapshot */ 2938 if (ds->ds_phys->ds_prev_snap_txg < TXG_INITIAL) { 2939 dsl_dataset_rele(ds, FTAG); 2940 return (EINVAL); 2941 } 2942 2943 /* 2944 * Create clone of most recent snapshot. 2945 */ 2946 clone_name = kmem_asprintf("%s/%%rollback", zc->zc_name); 2947 error = dmu_objset_clone(clone_name, ds->ds_prev, DS_FLAG_INCONSISTENT); 2948 if (error) 2949 goto out; 2950 2951 error = dsl_dataset_own(clone_name, B_TRUE, FTAG, &clone); 2952 if (error) 2953 goto out; 2954 2955 /* 2956 * Do clone swap. 2957 */ 2958 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) { 2959 error = zfs_suspend_fs(zfsvfs); 2960 if (error == 0) { 2961 int resume_err; 2962 2963 if (dsl_dataset_tryown(ds, B_FALSE, FTAG)) { 2964 error = dsl_dataset_clone_swap(clone, ds, 2965 B_TRUE); 2966 dsl_dataset_disown(ds, FTAG); 2967 ds = NULL; 2968 } else { 2969 error = EBUSY; 2970 } 2971 resume_err = zfs_resume_fs(zfsvfs, zc->zc_name); 2972 error = error ? error : resume_err; 2973 } 2974 VFS_RELE(zfsvfs->z_vfs); 2975 } else { 2976 if (dsl_dataset_tryown(ds, B_FALSE, FTAG)) { 2977 error = dsl_dataset_clone_swap(clone, ds, B_TRUE); 2978 dsl_dataset_disown(ds, FTAG); 2979 ds = NULL; 2980 } else { 2981 error = EBUSY; 2982 } 2983 } 2984 2985 /* 2986 * Destroy clone (which also closes it). 2987 */ 2988 (void) dsl_dataset_destroy(clone, FTAG, B_FALSE); 2989 2990 out: 2991 strfree(clone_name); 2992 if (ds) 2993 dsl_dataset_rele(ds, FTAG); 2994 return (error); 2995 } 2996 2997 /* 2998 * inputs: 2999 * zc_name old name of dataset 3000 * zc_value new name of dataset 3001 * zc_cookie recursive flag (only valid for snapshots) 3002 * 3003 * outputs: none 3004 */ 3005 static int 3006 zfs_ioc_rename(zfs_cmd_t *zc) 3007 { 3008 boolean_t recursive = zc->zc_cookie & 1; 3009 3010 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0'; 3011 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 3012 strchr(zc->zc_value, '%')) 3013 return (EINVAL); 3014 3015 /* 3016 * Unmount snapshot unless we're doing a recursive rename, 3017 * in which case the dataset code figures out which snapshots 3018 * to unmount. 3019 */ 3020 if (!recursive && strchr(zc->zc_name, '@') != NULL && 3021 zc->zc_objset_type == DMU_OST_ZFS) { 3022 int err = zfs_unmount_snap(zc->zc_name, NULL); 3023 if (err) 3024 return (err); 3025 } 3026 if (zc->zc_objset_type == DMU_OST_ZVOL) 3027 (void) zvol_remove_minor(zc->zc_name); 3028 return (dmu_objset_rename(zc->zc_name, zc->zc_value, recursive)); 3029 } 3030 3031 static int 3032 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr) 3033 { 3034 const char *propname = nvpair_name(pair); 3035 boolean_t issnap = (strchr(dsname, '@') != NULL); 3036 zfs_prop_t prop = zfs_name_to_prop(propname); 3037 uint64_t intval; 3038 int err; 3039 3040 if (prop == ZPROP_INVAL) { 3041 if (zfs_prop_user(propname)) { 3042 if (err = zfs_secpolicy_write_perms(dsname, 3043 ZFS_DELEG_PERM_USERPROP, cr)) 3044 return (err); 3045 return (0); 3046 } 3047 3048 if (!issnap && zfs_prop_userquota(propname)) { 3049 const char *perm = NULL; 3050 const char *uq_prefix = 3051 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA]; 3052 const char *gq_prefix = 3053 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA]; 3054 3055 if (strncmp(propname, uq_prefix, 3056 strlen(uq_prefix)) == 0) { 3057 perm = ZFS_DELEG_PERM_USERQUOTA; 3058 } else if (strncmp(propname, gq_prefix, 3059 strlen(gq_prefix)) == 0) { 3060 perm = ZFS_DELEG_PERM_GROUPQUOTA; 3061 } else { 3062 /* USERUSED and GROUPUSED are read-only */ 3063 return (EINVAL); 3064 } 3065 3066 if (err = zfs_secpolicy_write_perms(dsname, perm, cr)) 3067 return (err); 3068 return (0); 3069 } 3070 3071 return (EINVAL); 3072 } 3073 3074 if (issnap) 3075 return (EINVAL); 3076 3077 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 3078 /* 3079 * dsl_prop_get_all_impl() returns properties in this 3080 * format. 3081 */ 3082 nvlist_t *attrs; 3083 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 3084 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 3085 &pair) == 0); 3086 } 3087 3088 /* 3089 * Check that this value is valid for this pool version 3090 */ 3091 switch (prop) { 3092 case ZFS_PROP_COMPRESSION: 3093 /* 3094 * If the user specified gzip compression, make sure 3095 * the SPA supports it. We ignore any errors here since 3096 * we'll catch them later. 3097 */ 3098 if (nvpair_type(pair) == DATA_TYPE_UINT64 && 3099 nvpair_value_uint64(pair, &intval) == 0) { 3100 if (intval >= ZIO_COMPRESS_GZIP_1 && 3101 intval <= ZIO_COMPRESS_GZIP_9 && 3102 zfs_earlier_version(dsname, 3103 SPA_VERSION_GZIP_COMPRESSION)) { 3104 return (ENOTSUP); 3105 } 3106 3107 if (intval == ZIO_COMPRESS_ZLE && 3108 zfs_earlier_version(dsname, 3109 SPA_VERSION_ZLE_COMPRESSION)) 3110 return (ENOTSUP); 3111 3112 /* 3113 * If this is a bootable dataset then 3114 * verify that the compression algorithm 3115 * is supported for booting. We must return 3116 * something other than ENOTSUP since it 3117 * implies a downrev pool version. 3118 */ 3119 if (zfs_is_bootfs(dsname) && 3120 !BOOTFS_COMPRESS_VALID(intval)) { 3121 return (ERANGE); 3122 } 3123 } 3124 break; 3125 3126 case ZFS_PROP_COPIES: 3127 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS)) 3128 return (ENOTSUP); 3129 break; 3130 3131 case ZFS_PROP_DEDUP: 3132 if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP)) 3133 return (ENOTSUP); 3134 break; 3135 3136 case ZFS_PROP_SHARESMB: 3137 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID)) 3138 return (ENOTSUP); 3139 break; 3140 3141 case ZFS_PROP_ACLINHERIT: 3142 if (nvpair_type(pair) == DATA_TYPE_UINT64 && 3143 nvpair_value_uint64(pair, &intval) == 0) { 3144 if (intval == ZFS_ACL_PASSTHROUGH_X && 3145 zfs_earlier_version(dsname, 3146 SPA_VERSION_PASSTHROUGH_X)) 3147 return (ENOTSUP); 3148 } 3149 break; 3150 } 3151 3152 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED())); 3153 } 3154 3155 /* 3156 * Removes properties from the given props list that fail permission checks 3157 * needed to clear them and to restore them in case of a receive error. For each 3158 * property, make sure we have both set and inherit permissions. 3159 * 3160 * Returns the first error encountered if any permission checks fail. If the 3161 * caller provides a non-NULL errlist, it also gives the complete list of names 3162 * of all the properties that failed a permission check along with the 3163 * corresponding error numbers. The caller is responsible for freeing the 3164 * returned errlist. 3165 * 3166 * If every property checks out successfully, zero is returned and the list 3167 * pointed at by errlist is NULL. 3168 */ 3169 static int 3170 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist) 3171 { 3172 zfs_cmd_t *zc; 3173 nvpair_t *pair, *next_pair; 3174 nvlist_t *errors; 3175 int err, rv = 0; 3176 3177 if (props == NULL) 3178 return (0); 3179 3180 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0); 3181 3182 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP); 3183 (void) strcpy(zc->zc_name, dataset); 3184 pair = nvlist_next_nvpair(props, NULL); 3185 while (pair != NULL) { 3186 next_pair = nvlist_next_nvpair(props, pair); 3187 3188 (void) strcpy(zc->zc_value, nvpair_name(pair)); 3189 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 || 3190 (err = zfs_secpolicy_inherit(zc, CRED())) != 0) { 3191 VERIFY(nvlist_remove_nvpair(props, pair) == 0); 3192 VERIFY(nvlist_add_int32(errors, 3193 zc->zc_value, err) == 0); 3194 } 3195 pair = next_pair; 3196 } 3197 kmem_free(zc, sizeof (zfs_cmd_t)); 3198 3199 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) { 3200 nvlist_free(errors); 3201 errors = NULL; 3202 } else { 3203 VERIFY(nvpair_value_int32(pair, &rv) == 0); 3204 } 3205 3206 if (errlist == NULL) 3207 nvlist_free(errors); 3208 else 3209 *errlist = errors; 3210 3211 return (rv); 3212 } 3213 3214 static boolean_t 3215 propval_equals(nvpair_t *p1, nvpair_t *p2) 3216 { 3217 if (nvpair_type(p1) == DATA_TYPE_NVLIST) { 3218 /* dsl_prop_get_all_impl() format */ 3219 nvlist_t *attrs; 3220 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0); 3221 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 3222 &p1) == 0); 3223 } 3224 3225 if (nvpair_type(p2) == DATA_TYPE_NVLIST) { 3226 nvlist_t *attrs; 3227 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0); 3228 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 3229 &p2) == 0); 3230 } 3231 3232 if (nvpair_type(p1) != nvpair_type(p2)) 3233 return (B_FALSE); 3234 3235 if (nvpair_type(p1) == DATA_TYPE_STRING) { 3236 char *valstr1, *valstr2; 3237 3238 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0); 3239 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0); 3240 return (strcmp(valstr1, valstr2) == 0); 3241 } else { 3242 uint64_t intval1, intval2; 3243 3244 VERIFY(nvpair_value_uint64(p1, &intval1) == 0); 3245 VERIFY(nvpair_value_uint64(p2, &intval2) == 0); 3246 return (intval1 == intval2); 3247 } 3248 } 3249 3250 /* 3251 * Remove properties from props if they are not going to change (as determined 3252 * by comparison with origprops). Remove them from origprops as well, since we 3253 * do not need to clear or restore properties that won't change. 3254 */ 3255 static void 3256 props_reduce(nvlist_t *props, nvlist_t *origprops) 3257 { 3258 nvpair_t *pair, *next_pair; 3259 3260 if (origprops == NULL) 3261 return; /* all props need to be received */ 3262 3263 pair = nvlist_next_nvpair(props, NULL); 3264 while (pair != NULL) { 3265 const char *propname = nvpair_name(pair); 3266 nvpair_t *match; 3267 3268 next_pair = nvlist_next_nvpair(props, pair); 3269 3270 if ((nvlist_lookup_nvpair(origprops, propname, 3271 &match) != 0) || !propval_equals(pair, match)) 3272 goto next; /* need to set received value */ 3273 3274 /* don't clear the existing received value */ 3275 (void) nvlist_remove_nvpair(origprops, match); 3276 /* don't bother receiving the property */ 3277 (void) nvlist_remove_nvpair(props, pair); 3278 next: 3279 pair = next_pair; 3280 } 3281 } 3282 3283 #ifdef DEBUG 3284 static boolean_t zfs_ioc_recv_inject_err; 3285 #endif 3286 3287 /* 3288 * inputs: 3289 * zc_name name of containing filesystem 3290 * zc_nvlist_src{_size} nvlist of properties to apply 3291 * zc_value name of snapshot to create 3292 * zc_string name of clone origin (if DRR_FLAG_CLONE) 3293 * zc_cookie file descriptor to recv from 3294 * zc_begin_record the BEGIN record of the stream (not byteswapped) 3295 * zc_guid force flag 3296 * 3297 * outputs: 3298 * zc_cookie number of bytes read 3299 * zc_nvlist_dst{_size} error for each unapplied received property 3300 * zc_obj zprop_errflags_t 3301 */ 3302 static int 3303 zfs_ioc_recv(zfs_cmd_t *zc) 3304 { 3305 file_t *fp; 3306 objset_t *os; 3307 dmu_recv_cookie_t drc; 3308 boolean_t force = (boolean_t)zc->zc_guid; 3309 int fd; 3310 int error = 0; 3311 int props_error = 0; 3312 nvlist_t *errors; 3313 offset_t off; 3314 nvlist_t *props = NULL; /* sent properties */ 3315 nvlist_t *origprops = NULL; /* existing properties */ 3316 objset_t *origin = NULL; 3317 char *tosnap; 3318 char tofs[ZFS_MAXNAMELEN]; 3319 boolean_t first_recvd_props = B_FALSE; 3320 3321 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 3322 strchr(zc->zc_value, '@') == NULL || 3323 strchr(zc->zc_value, '%')) 3324 return (EINVAL); 3325 3326 (void) strcpy(tofs, zc->zc_value); 3327 tosnap = strchr(tofs, '@'); 3328 *tosnap++ = '\0'; 3329 3330 if (zc->zc_nvlist_src != NULL && 3331 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 3332 zc->zc_iflags, &props)) != 0) 3333 return (error); 3334 3335 fd = zc->zc_cookie; 3336 fp = getf(fd); 3337 if (fp == NULL) { 3338 nvlist_free(props); 3339 return (EBADF); 3340 } 3341 3342 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0); 3343 3344 if (props && dmu_objset_hold(tofs, FTAG, &os) == 0) { 3345 if ((spa_version(os->os_spa) >= SPA_VERSION_RECVD_PROPS) && 3346 !dsl_prop_get_hasrecvd(os)) { 3347 first_recvd_props = B_TRUE; 3348 } 3349 3350 /* 3351 * If new received properties are supplied, they are to 3352 * completely replace the existing received properties, so stash 3353 * away the existing ones. 3354 */ 3355 if (dsl_prop_get_received(os, &origprops) == 0) { 3356 nvlist_t *errlist = NULL; 3357 /* 3358 * Don't bother writing a property if its value won't 3359 * change (and avoid the unnecessary security checks). 3360 * 3361 * The first receive after SPA_VERSION_RECVD_PROPS is a 3362 * special case where we blow away all local properties 3363 * regardless. 3364 */ 3365 if (!first_recvd_props) 3366 props_reduce(props, origprops); 3367 if (zfs_check_clearable(tofs, origprops, 3368 &errlist) != 0) 3369 (void) nvlist_merge(errors, errlist, 0); 3370 nvlist_free(errlist); 3371 } 3372 3373 dmu_objset_rele(os, FTAG); 3374 } 3375 3376 if (zc->zc_string[0]) { 3377 error = dmu_objset_hold(zc->zc_string, FTAG, &origin); 3378 if (error) 3379 goto out; 3380 } 3381 3382 error = dmu_recv_begin(tofs, tosnap, zc->zc_top_ds, 3383 &zc->zc_begin_record, force, origin, &drc); 3384 if (origin) 3385 dmu_objset_rele(origin, FTAG); 3386 if (error) 3387 goto out; 3388 3389 /* 3390 * Set properties before we receive the stream so that they are applied 3391 * to the new data. Note that we must call dmu_recv_stream() if 3392 * dmu_recv_begin() succeeds. 3393 */ 3394 if (props) { 3395 nvlist_t *errlist; 3396 3397 if (dmu_objset_from_ds(drc.drc_logical_ds, &os) == 0) { 3398 if (drc.drc_newfs) { 3399 if (spa_version(os->os_spa) >= 3400 SPA_VERSION_RECVD_PROPS) 3401 first_recvd_props = B_TRUE; 3402 } else if (origprops != NULL) { 3403 if (clear_received_props(os, tofs, origprops, 3404 first_recvd_props ? NULL : props) != 0) 3405 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 3406 } else { 3407 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 3408 } 3409 dsl_prop_set_hasrecvd(os); 3410 } else if (!drc.drc_newfs) { 3411 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 3412 } 3413 3414 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED, 3415 props, &errlist); 3416 (void) nvlist_merge(errors, errlist, 0); 3417 nvlist_free(errlist); 3418 } 3419 3420 if (fit_error_list(zc, &errors) != 0 || put_nvlist(zc, errors) != 0) { 3421 /* 3422 * Caller made zc->zc_nvlist_dst less than the minimum expected 3423 * size or supplied an invalid address. 3424 */ 3425 props_error = EINVAL; 3426 } 3427 3428 off = fp->f_offset; 3429 error = dmu_recv_stream(&drc, fp->f_vnode, &off); 3430 3431 if (error == 0) { 3432 zfsvfs_t *zfsvfs = NULL; 3433 3434 if (getzfsvfs(tofs, &zfsvfs) == 0) { 3435 /* online recv */ 3436 int end_err; 3437 3438 error = zfs_suspend_fs(zfsvfs); 3439 /* 3440 * If the suspend fails, then the recv_end will 3441 * likely also fail, and clean up after itself. 3442 */ 3443 end_err = dmu_recv_end(&drc); 3444 if (error == 0) { 3445 int resume_err = 3446 zfs_resume_fs(zfsvfs, tofs); 3447 error = error ? error : resume_err; 3448 } 3449 error = error ? error : end_err; 3450 VFS_RELE(zfsvfs->z_vfs); 3451 } else { 3452 error = dmu_recv_end(&drc); 3453 } 3454 } 3455 3456 zc->zc_cookie = off - fp->f_offset; 3457 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0) 3458 fp->f_offset = off; 3459 3460 #ifdef DEBUG 3461 if (zfs_ioc_recv_inject_err) { 3462 zfs_ioc_recv_inject_err = B_FALSE; 3463 error = 1; 3464 } 3465 #endif 3466 /* 3467 * On error, restore the original props. 3468 */ 3469 if (error && props) { 3470 if (dmu_objset_hold(tofs, FTAG, &os) == 0) { 3471 if (clear_received_props(os, tofs, props, NULL) != 0) { 3472 /* 3473 * We failed to clear the received properties. 3474 * Since we may have left a $recvd value on the 3475 * system, we can't clear the $hasrecvd flag. 3476 */ 3477 zc->zc_obj |= ZPROP_ERR_NORESTORE; 3478 } else if (first_recvd_props) { 3479 dsl_prop_unset_hasrecvd(os); 3480 } 3481 dmu_objset_rele(os, FTAG); 3482 } else if (!drc.drc_newfs) { 3483 /* We failed to clear the received properties. */ 3484 zc->zc_obj |= ZPROP_ERR_NORESTORE; 3485 } 3486 3487 if (origprops == NULL && !drc.drc_newfs) { 3488 /* We failed to stash the original properties. */ 3489 zc->zc_obj |= ZPROP_ERR_NORESTORE; 3490 } 3491 3492 /* 3493 * dsl_props_set() will not convert RECEIVED to LOCAL on or 3494 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL 3495 * explictly if we're restoring local properties cleared in the 3496 * first new-style receive. 3497 */ 3498 if (origprops != NULL && 3499 zfs_set_prop_nvlist(tofs, (first_recvd_props ? 3500 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED), 3501 origprops, NULL) != 0) { 3502 /* 3503 * We stashed the original properties but failed to 3504 * restore them. 3505 */ 3506 zc->zc_obj |= ZPROP_ERR_NORESTORE; 3507 } 3508 } 3509 out: 3510 nvlist_free(props); 3511 nvlist_free(origprops); 3512 nvlist_free(errors); 3513 releasef(fd); 3514 3515 if (error == 0) 3516 error = props_error; 3517 3518 return (error); 3519 } 3520 3521 /* 3522 * inputs: 3523 * zc_name name of snapshot to send 3524 * zc_value short name of incremental fromsnap (may be empty) 3525 * zc_cookie file descriptor to send stream to 3526 * zc_obj fromorigin flag (mutually exclusive with zc_value) 3527 * 3528 * outputs: none 3529 */ 3530 static int 3531 zfs_ioc_send(zfs_cmd_t *zc) 3532 { 3533 objset_t *fromsnap = NULL; 3534 objset_t *tosnap; 3535 file_t *fp; 3536 int error; 3537 offset_t off; 3538 3539 error = dmu_objset_hold(zc->zc_name, FTAG, &tosnap); 3540 if (error) 3541 return (error); 3542 3543 if (zc->zc_value[0] != '\0') { 3544 char *buf; 3545 char *cp; 3546 3547 buf = kmem_alloc(MAXPATHLEN, KM_SLEEP); 3548 (void) strncpy(buf, zc->zc_name, MAXPATHLEN); 3549 cp = strchr(buf, '@'); 3550 if (cp) 3551 *(cp+1) = 0; 3552 (void) strncat(buf, zc->zc_value, MAXPATHLEN); 3553 error = dmu_objset_hold(buf, FTAG, &fromsnap); 3554 kmem_free(buf, MAXPATHLEN); 3555 if (error) { 3556 dmu_objset_rele(tosnap, FTAG); 3557 return (error); 3558 } 3559 } 3560 3561 fp = getf(zc->zc_cookie); 3562 if (fp == NULL) { 3563 dmu_objset_rele(tosnap, FTAG); 3564 if (fromsnap) 3565 dmu_objset_rele(fromsnap, FTAG); 3566 return (EBADF); 3567 } 3568 3569 off = fp->f_offset; 3570 error = dmu_sendbackup(tosnap, fromsnap, zc->zc_obj, fp->f_vnode, &off); 3571 3572 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0) 3573 fp->f_offset = off; 3574 releasef(zc->zc_cookie); 3575 if (fromsnap) 3576 dmu_objset_rele(fromsnap, FTAG); 3577 dmu_objset_rele(tosnap, FTAG); 3578 return (error); 3579 } 3580 3581 static int 3582 zfs_ioc_inject_fault(zfs_cmd_t *zc) 3583 { 3584 int id, error; 3585 3586 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id, 3587 &zc->zc_inject_record); 3588 3589 if (error == 0) 3590 zc->zc_guid = (uint64_t)id; 3591 3592 return (error); 3593 } 3594 3595 static int 3596 zfs_ioc_clear_fault(zfs_cmd_t *zc) 3597 { 3598 return (zio_clear_fault((int)zc->zc_guid)); 3599 } 3600 3601 static int 3602 zfs_ioc_inject_list_next(zfs_cmd_t *zc) 3603 { 3604 int id = (int)zc->zc_guid; 3605 int error; 3606 3607 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name), 3608 &zc->zc_inject_record); 3609 3610 zc->zc_guid = id; 3611 3612 return (error); 3613 } 3614 3615 static int 3616 zfs_ioc_error_log(zfs_cmd_t *zc) 3617 { 3618 spa_t *spa; 3619 int error; 3620 size_t count = (size_t)zc->zc_nvlist_dst_size; 3621 3622 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 3623 return (error); 3624 3625 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst, 3626 &count); 3627 if (error == 0) 3628 zc->zc_nvlist_dst_size = count; 3629 else 3630 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa); 3631 3632 spa_close(spa, FTAG); 3633 3634 return (error); 3635 } 3636 3637 static int 3638 zfs_ioc_clear(zfs_cmd_t *zc) 3639 { 3640 spa_t *spa; 3641 vdev_t *vd; 3642 int error; 3643 3644 /* 3645 * On zpool clear we also fix up missing slogs 3646 */ 3647 mutex_enter(&spa_namespace_lock); 3648 spa = spa_lookup(zc->zc_name); 3649 if (spa == NULL) { 3650 mutex_exit(&spa_namespace_lock); 3651 return (EIO); 3652 } 3653 if (spa_get_log_state(spa) == SPA_LOG_MISSING) { 3654 /* we need to let spa_open/spa_load clear the chains */ 3655 spa_set_log_state(spa, SPA_LOG_CLEAR); 3656 } 3657 spa->spa_last_open_failed = 0; 3658 mutex_exit(&spa_namespace_lock); 3659 3660 if (zc->zc_cookie == ZPOOL_NO_REWIND) { 3661 error = spa_open(zc->zc_name, &spa, FTAG); 3662 } else { 3663 nvlist_t *policy; 3664 nvlist_t *config = NULL; 3665 3666 if (zc->zc_nvlist_src == NULL) 3667 return (EINVAL); 3668 3669 if ((error = get_nvlist(zc->zc_nvlist_src, 3670 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) { 3671 error = spa_open_rewind(zc->zc_name, &spa, FTAG, 3672 policy, &config); 3673 if (config != NULL) { 3674 (void) put_nvlist(zc, config); 3675 nvlist_free(config); 3676 } 3677 nvlist_free(policy); 3678 } 3679 } 3680 3681 if (error) 3682 return (error); 3683 3684 spa_vdev_state_enter(spa, SCL_NONE); 3685 3686 if (zc->zc_guid == 0) { 3687 vd = NULL; 3688 } else { 3689 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE); 3690 if (vd == NULL) { 3691 (void) spa_vdev_state_exit(spa, NULL, ENODEV); 3692 spa_close(spa, FTAG); 3693 return (ENODEV); 3694 } 3695 } 3696 3697 vdev_clear(spa, vd); 3698 3699 (void) spa_vdev_state_exit(spa, NULL, 0); 3700 3701 /* 3702 * Resume any suspended I/Os. 3703 */ 3704 if (zio_resume(spa) != 0) 3705 error = EIO; 3706 3707 spa_close(spa, FTAG); 3708 3709 return (error); 3710 } 3711 3712 /* 3713 * inputs: 3714 * zc_name name of filesystem 3715 * zc_value name of origin snapshot 3716 * 3717 * outputs: 3718 * zc_string name of conflicting snapshot, if there is one 3719 */ 3720 static int 3721 zfs_ioc_promote(zfs_cmd_t *zc) 3722 { 3723 char *cp; 3724 3725 /* 3726 * We don't need to unmount *all* the origin fs's snapshots, but 3727 * it's easier. 3728 */ 3729 cp = strchr(zc->zc_value, '@'); 3730 if (cp) 3731 *cp = '\0'; 3732 (void) dmu_objset_find(zc->zc_value, 3733 zfs_unmount_snap, NULL, DS_FIND_SNAPSHOTS); 3734 return (dsl_dataset_promote(zc->zc_name, zc->zc_string)); 3735 } 3736 3737 /* 3738 * Retrieve a single {user|group}{used|quota}@... property. 3739 * 3740 * inputs: 3741 * zc_name name of filesystem 3742 * zc_objset_type zfs_userquota_prop_t 3743 * zc_value domain name (eg. "S-1-234-567-89") 3744 * zc_guid RID/UID/GID 3745 * 3746 * outputs: 3747 * zc_cookie property value 3748 */ 3749 static int 3750 zfs_ioc_userspace_one(zfs_cmd_t *zc) 3751 { 3752 zfsvfs_t *zfsvfs; 3753 int error; 3754 3755 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 3756 return (EINVAL); 3757 3758 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs); 3759 if (error) 3760 return (error); 3761 3762 error = zfs_userspace_one(zfsvfs, 3763 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie); 3764 zfsvfs_rele(zfsvfs, FTAG); 3765 3766 return (error); 3767 } 3768 3769 /* 3770 * inputs: 3771 * zc_name name of filesystem 3772 * zc_cookie zap cursor 3773 * zc_objset_type zfs_userquota_prop_t 3774 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist) 3775 * 3776 * outputs: 3777 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t) 3778 * zc_cookie zap cursor 3779 */ 3780 static int 3781 zfs_ioc_userspace_many(zfs_cmd_t *zc) 3782 { 3783 zfsvfs_t *zfsvfs; 3784 int error; 3785 3786 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs); 3787 if (error) 3788 return (error); 3789 3790 int bufsize = zc->zc_nvlist_dst_size; 3791 void *buf = kmem_alloc(bufsize, KM_SLEEP); 3792 3793 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie, 3794 buf, &zc->zc_nvlist_dst_size); 3795 3796 if (error == 0) { 3797 error = xcopyout(buf, 3798 (void *)(uintptr_t)zc->zc_nvlist_dst, 3799 zc->zc_nvlist_dst_size); 3800 } 3801 kmem_free(buf, bufsize); 3802 zfsvfs_rele(zfsvfs, FTAG); 3803 3804 return (error); 3805 } 3806 3807 /* 3808 * inputs: 3809 * zc_name name of filesystem 3810 * 3811 * outputs: 3812 * none 3813 */ 3814 static int 3815 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc) 3816 { 3817 objset_t *os; 3818 int error; 3819 zfsvfs_t *zfsvfs; 3820 3821 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) { 3822 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) { 3823 /* 3824 * If userused is not enabled, it may be because the 3825 * objset needs to be closed & reopened (to grow the 3826 * objset_phys_t). Suspend/resume the fs will do that. 3827 */ 3828 error = zfs_suspend_fs(zfsvfs); 3829 if (error == 0) 3830 error = zfs_resume_fs(zfsvfs, zc->zc_name); 3831 } 3832 if (error == 0) 3833 error = dmu_objset_userspace_upgrade(zfsvfs->z_os); 3834 VFS_RELE(zfsvfs->z_vfs); 3835 } else { 3836 /* XXX kind of reading contents without owning */ 3837 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 3838 if (error) 3839 return (error); 3840 3841 error = dmu_objset_userspace_upgrade(os); 3842 dmu_objset_rele(os, FTAG); 3843 } 3844 3845 return (error); 3846 } 3847 3848 /* 3849 * We don't want to have a hard dependency 3850 * against some special symbols in sharefs 3851 * nfs, and smbsrv. Determine them if needed when 3852 * the first file system is shared. 3853 * Neither sharefs, nfs or smbsrv are unloadable modules. 3854 */ 3855 int (*znfsexport_fs)(void *arg); 3856 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t); 3857 int (*zsmbexport_fs)(void *arg, boolean_t add_share); 3858 3859 int zfs_nfsshare_inited; 3860 int zfs_smbshare_inited; 3861 3862 ddi_modhandle_t nfs_mod; 3863 ddi_modhandle_t sharefs_mod; 3864 ddi_modhandle_t smbsrv_mod; 3865 kmutex_t zfs_share_lock; 3866 3867 static int 3868 zfs_init_sharefs() 3869 { 3870 int error; 3871 3872 ASSERT(MUTEX_HELD(&zfs_share_lock)); 3873 /* Both NFS and SMB shares also require sharetab support. */ 3874 if (sharefs_mod == NULL && ((sharefs_mod = 3875 ddi_modopen("fs/sharefs", 3876 KRTLD_MODE_FIRST, &error)) == NULL)) { 3877 return (ENOSYS); 3878 } 3879 if (zshare_fs == NULL && ((zshare_fs = 3880 (int (*)(enum sharefs_sys_op, share_t *, uint32_t)) 3881 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) { 3882 return (ENOSYS); 3883 } 3884 return (0); 3885 } 3886 3887 static int 3888 zfs_ioc_share(zfs_cmd_t *zc) 3889 { 3890 int error; 3891 int opcode; 3892 3893 switch (zc->zc_share.z_sharetype) { 3894 case ZFS_SHARE_NFS: 3895 case ZFS_UNSHARE_NFS: 3896 if (zfs_nfsshare_inited == 0) { 3897 mutex_enter(&zfs_share_lock); 3898 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs", 3899 KRTLD_MODE_FIRST, &error)) == NULL)) { 3900 mutex_exit(&zfs_share_lock); 3901 return (ENOSYS); 3902 } 3903 if (znfsexport_fs == NULL && 3904 ((znfsexport_fs = (int (*)(void *)) 3905 ddi_modsym(nfs_mod, 3906 "nfs_export", &error)) == NULL)) { 3907 mutex_exit(&zfs_share_lock); 3908 return (ENOSYS); 3909 } 3910 error = zfs_init_sharefs(); 3911 if (error) { 3912 mutex_exit(&zfs_share_lock); 3913 return (ENOSYS); 3914 } 3915 zfs_nfsshare_inited = 1; 3916 mutex_exit(&zfs_share_lock); 3917 } 3918 break; 3919 case ZFS_SHARE_SMB: 3920 case ZFS_UNSHARE_SMB: 3921 if (zfs_smbshare_inited == 0) { 3922 mutex_enter(&zfs_share_lock); 3923 if (smbsrv_mod == NULL && ((smbsrv_mod = 3924 ddi_modopen("drv/smbsrv", 3925 KRTLD_MODE_FIRST, &error)) == NULL)) { 3926 mutex_exit(&zfs_share_lock); 3927 return (ENOSYS); 3928 } 3929 if (zsmbexport_fs == NULL && ((zsmbexport_fs = 3930 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod, 3931 "smb_server_share", &error)) == NULL)) { 3932 mutex_exit(&zfs_share_lock); 3933 return (ENOSYS); 3934 } 3935 error = zfs_init_sharefs(); 3936 if (error) { 3937 mutex_exit(&zfs_share_lock); 3938 return (ENOSYS); 3939 } 3940 zfs_smbshare_inited = 1; 3941 mutex_exit(&zfs_share_lock); 3942 } 3943 break; 3944 default: 3945 return (EINVAL); 3946 } 3947 3948 switch (zc->zc_share.z_sharetype) { 3949 case ZFS_SHARE_NFS: 3950 case ZFS_UNSHARE_NFS: 3951 if (error = 3952 znfsexport_fs((void *) 3953 (uintptr_t)zc->zc_share.z_exportdata)) 3954 return (error); 3955 break; 3956 case ZFS_SHARE_SMB: 3957 case ZFS_UNSHARE_SMB: 3958 if (error = zsmbexport_fs((void *) 3959 (uintptr_t)zc->zc_share.z_exportdata, 3960 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ? 3961 B_TRUE: B_FALSE)) { 3962 return (error); 3963 } 3964 break; 3965 } 3966 3967 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS || 3968 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ? 3969 SHAREFS_ADD : SHAREFS_REMOVE; 3970 3971 /* 3972 * Add or remove share from sharetab 3973 */ 3974 error = zshare_fs(opcode, 3975 (void *)(uintptr_t)zc->zc_share.z_sharedata, 3976 zc->zc_share.z_sharemax); 3977 3978 return (error); 3979 3980 } 3981 3982 ace_t full_access[] = { 3983 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0} 3984 }; 3985 3986 /* 3987 * Remove all ACL files in shares dir 3988 */ 3989 static int 3990 zfs_smb_acl_purge(znode_t *dzp) 3991 { 3992 zap_cursor_t zc; 3993 zap_attribute_t zap; 3994 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 3995 int error; 3996 3997 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id); 3998 (error = zap_cursor_retrieve(&zc, &zap)) == 0; 3999 zap_cursor_advance(&zc)) { 4000 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred, 4001 NULL, 0)) != 0) 4002 break; 4003 } 4004 zap_cursor_fini(&zc); 4005 return (error); 4006 } 4007 4008 static int 4009 zfs_ioc_smb_acl(zfs_cmd_t *zc) 4010 { 4011 vnode_t *vp; 4012 znode_t *dzp; 4013 vnode_t *resourcevp = NULL; 4014 znode_t *sharedir; 4015 zfsvfs_t *zfsvfs; 4016 nvlist_t *nvlist; 4017 char *src, *target; 4018 vattr_t vattr; 4019 vsecattr_t vsec; 4020 int error = 0; 4021 4022 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, 4023 NO_FOLLOW, NULL, &vp)) != 0) 4024 return (error); 4025 4026 /* Now make sure mntpnt and dataset are ZFS */ 4027 4028 if (vp->v_vfsp->vfs_fstype != zfsfstype || 4029 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), 4030 zc->zc_name) != 0)) { 4031 VN_RELE(vp); 4032 return (EINVAL); 4033 } 4034 4035 dzp = VTOZ(vp); 4036 zfsvfs = dzp->z_zfsvfs; 4037 ZFS_ENTER(zfsvfs); 4038 4039 /* 4040 * Create share dir if its missing. 4041 */ 4042 mutex_enter(&zfsvfs->z_lock); 4043 if (zfsvfs->z_shares_dir == 0) { 4044 dmu_tx_t *tx; 4045 4046 tx = dmu_tx_create(zfsvfs->z_os); 4047 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE, 4048 ZFS_SHARES_DIR); 4049 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); 4050 error = dmu_tx_assign(tx, TXG_WAIT); 4051 if (error) { 4052 dmu_tx_abort(tx); 4053 } else { 4054 error = zfs_create_share_dir(zfsvfs, tx); 4055 dmu_tx_commit(tx); 4056 } 4057 if (error) { 4058 mutex_exit(&zfsvfs->z_lock); 4059 VN_RELE(vp); 4060 ZFS_EXIT(zfsvfs); 4061 return (error); 4062 } 4063 } 4064 mutex_exit(&zfsvfs->z_lock); 4065 4066 ASSERT(zfsvfs->z_shares_dir); 4067 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) { 4068 VN_RELE(vp); 4069 ZFS_EXIT(zfsvfs); 4070 return (error); 4071 } 4072 4073 switch (zc->zc_cookie) { 4074 case ZFS_SMB_ACL_ADD: 4075 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; 4076 vattr.va_type = VREG; 4077 vattr.va_mode = S_IFREG|0777; 4078 vattr.va_uid = 0; 4079 vattr.va_gid = 0; 4080 4081 vsec.vsa_mask = VSA_ACE; 4082 vsec.vsa_aclentp = &full_access; 4083 vsec.vsa_aclentsz = sizeof (full_access); 4084 vsec.vsa_aclcnt = 1; 4085 4086 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string, 4087 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec); 4088 if (resourcevp) 4089 VN_RELE(resourcevp); 4090 break; 4091 4092 case ZFS_SMB_ACL_REMOVE: 4093 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred, 4094 NULL, 0); 4095 break; 4096 4097 case ZFS_SMB_ACL_RENAME: 4098 if ((error = get_nvlist(zc->zc_nvlist_src, 4099 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) { 4100 VN_RELE(vp); 4101 ZFS_EXIT(zfsvfs); 4102 return (error); 4103 } 4104 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) || 4105 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET, 4106 &target)) { 4107 VN_RELE(vp); 4108 VN_RELE(ZTOV(sharedir)); 4109 ZFS_EXIT(zfsvfs); 4110 return (error); 4111 } 4112 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target, 4113 kcred, NULL, 0); 4114 nvlist_free(nvlist); 4115 break; 4116 4117 case ZFS_SMB_ACL_PURGE: 4118 error = zfs_smb_acl_purge(sharedir); 4119 break; 4120 4121 default: 4122 error = EINVAL; 4123 break; 4124 } 4125 4126 VN_RELE(vp); 4127 VN_RELE(ZTOV(sharedir)); 4128 4129 ZFS_EXIT(zfsvfs); 4130 4131 return (error); 4132 } 4133 4134 /* 4135 * inputs: 4136 * zc_name name of filesystem 4137 * zc_value short name of snap 4138 * zc_string user-supplied tag for this reference 4139 * zc_cookie recursive flag 4140 * zc_temphold set if hold is temporary 4141 * 4142 * outputs: none 4143 */ 4144 static int 4145 zfs_ioc_hold(zfs_cmd_t *zc) 4146 { 4147 boolean_t recursive = zc->zc_cookie; 4148 4149 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0) 4150 return (EINVAL); 4151 4152 return (dsl_dataset_user_hold(zc->zc_name, zc->zc_value, 4153 zc->zc_string, recursive, zc->zc_temphold)); 4154 } 4155 4156 /* 4157 * inputs: 4158 * zc_name name of dataset from which we're releasing a user reference 4159 * zc_value short name of snap 4160 * zc_string user-supplied tag for this reference 4161 * zc_cookie recursive flag 4162 * 4163 * outputs: none 4164 */ 4165 static int 4166 zfs_ioc_release(zfs_cmd_t *zc) 4167 { 4168 boolean_t recursive = zc->zc_cookie; 4169 4170 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0) 4171 return (EINVAL); 4172 4173 return (dsl_dataset_user_release(zc->zc_name, zc->zc_value, 4174 zc->zc_string, recursive)); 4175 } 4176 4177 /* 4178 * inputs: 4179 * zc_name name of filesystem 4180 * 4181 * outputs: 4182 * zc_nvlist_src{_size} nvlist of snapshot holds 4183 */ 4184 static int 4185 zfs_ioc_get_holds(zfs_cmd_t *zc) 4186 { 4187 nvlist_t *nvp; 4188 int error; 4189 4190 if ((error = dsl_dataset_get_holds(zc->zc_name, &nvp)) == 0) { 4191 error = put_nvlist(zc, nvp); 4192 nvlist_free(nvp); 4193 } 4194 4195 return (error); 4196 } 4197 4198 /* 4199 * pool create, destroy, and export don't log the history as part of 4200 * zfsdev_ioctl, but rather zfs_ioc_pool_create, and zfs_ioc_pool_export 4201 * do the logging of those commands. 4202 */ 4203 static zfs_ioc_vec_t zfs_ioc_vec[] = { 4204 { zfs_ioc_pool_create, zfs_secpolicy_config, POOL_NAME, B_FALSE, 4205 B_FALSE }, 4206 { zfs_ioc_pool_destroy, zfs_secpolicy_config, POOL_NAME, B_FALSE, 4207 B_FALSE }, 4208 { zfs_ioc_pool_import, zfs_secpolicy_config, POOL_NAME, B_TRUE, 4209 B_FALSE }, 4210 { zfs_ioc_pool_export, zfs_secpolicy_config, POOL_NAME, B_FALSE, 4211 B_FALSE }, 4212 { zfs_ioc_pool_configs, zfs_secpolicy_none, NO_NAME, B_FALSE, 4213 B_FALSE }, 4214 { zfs_ioc_pool_stats, zfs_secpolicy_read, POOL_NAME, B_FALSE, 4215 B_FALSE }, 4216 { zfs_ioc_pool_tryimport, zfs_secpolicy_config, NO_NAME, B_FALSE, 4217 B_FALSE }, 4218 { zfs_ioc_pool_scrub, zfs_secpolicy_config, POOL_NAME, B_TRUE, 4219 B_TRUE }, 4220 { zfs_ioc_pool_freeze, zfs_secpolicy_config, NO_NAME, B_FALSE, 4221 B_FALSE }, 4222 { zfs_ioc_pool_upgrade, zfs_secpolicy_config, POOL_NAME, B_TRUE, 4223 B_TRUE }, 4224 { zfs_ioc_pool_get_history, zfs_secpolicy_config, POOL_NAME, B_FALSE, 4225 B_FALSE }, 4226 { zfs_ioc_vdev_add, zfs_secpolicy_config, POOL_NAME, B_TRUE, 4227 B_TRUE }, 4228 { zfs_ioc_vdev_remove, zfs_secpolicy_config, POOL_NAME, B_TRUE, 4229 B_TRUE }, 4230 { zfs_ioc_vdev_set_state, zfs_secpolicy_config, POOL_NAME, B_TRUE, 4231 B_FALSE }, 4232 { zfs_ioc_vdev_attach, zfs_secpolicy_config, POOL_NAME, B_TRUE, 4233 B_TRUE }, 4234 { zfs_ioc_vdev_detach, zfs_secpolicy_config, POOL_NAME, B_TRUE, 4235 B_TRUE }, 4236 { zfs_ioc_vdev_setpath, zfs_secpolicy_config, POOL_NAME, B_FALSE, 4237 B_TRUE }, 4238 { zfs_ioc_vdev_setfru, zfs_secpolicy_config, POOL_NAME, B_FALSE, 4239 B_TRUE }, 4240 { zfs_ioc_objset_stats, zfs_secpolicy_read, DATASET_NAME, B_FALSE, 4241 B_FALSE }, 4242 { zfs_ioc_objset_zplprops, zfs_secpolicy_read, DATASET_NAME, B_FALSE, 4243 B_FALSE }, 4244 { zfs_ioc_dataset_list_next, zfs_secpolicy_read, DATASET_NAME, B_FALSE, 4245 B_FALSE }, 4246 { zfs_ioc_snapshot_list_next, zfs_secpolicy_read, DATASET_NAME, B_FALSE, 4247 B_FALSE }, 4248 { zfs_ioc_set_prop, zfs_secpolicy_none, DATASET_NAME, B_TRUE, B_TRUE }, 4249 { zfs_ioc_create, zfs_secpolicy_create, DATASET_NAME, B_TRUE, B_TRUE }, 4250 { zfs_ioc_destroy, zfs_secpolicy_destroy, DATASET_NAME, B_TRUE, 4251 B_TRUE}, 4252 { zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, B_TRUE, 4253 B_TRUE }, 4254 { zfs_ioc_rename, zfs_secpolicy_rename, DATASET_NAME, B_TRUE, B_TRUE }, 4255 { zfs_ioc_recv, zfs_secpolicy_receive, DATASET_NAME, B_TRUE, B_TRUE }, 4256 { zfs_ioc_send, zfs_secpolicy_send, DATASET_NAME, B_TRUE, B_FALSE }, 4257 { zfs_ioc_inject_fault, zfs_secpolicy_inject, NO_NAME, B_FALSE, 4258 B_FALSE }, 4259 { zfs_ioc_clear_fault, zfs_secpolicy_inject, NO_NAME, B_FALSE, 4260 B_FALSE }, 4261 { zfs_ioc_inject_list_next, zfs_secpolicy_inject, NO_NAME, B_FALSE, 4262 B_FALSE }, 4263 { zfs_ioc_error_log, zfs_secpolicy_inject, POOL_NAME, B_FALSE, 4264 B_FALSE }, 4265 { zfs_ioc_clear, zfs_secpolicy_config, POOL_NAME, B_TRUE, B_FALSE }, 4266 { zfs_ioc_promote, zfs_secpolicy_promote, DATASET_NAME, B_TRUE, 4267 B_TRUE }, 4268 { zfs_ioc_destroy_snaps, zfs_secpolicy_destroy, DATASET_NAME, B_TRUE, 4269 B_TRUE }, 4270 { zfs_ioc_snapshot, zfs_secpolicy_snapshot, DATASET_NAME, B_TRUE, 4271 B_TRUE }, 4272 { zfs_ioc_dsobj_to_dsname, zfs_secpolicy_config, POOL_NAME, B_FALSE, 4273 B_FALSE }, 4274 { zfs_ioc_obj_to_path, zfs_secpolicy_config, DATASET_NAME, B_FALSE, 4275 B_TRUE }, 4276 { zfs_ioc_pool_set_props, zfs_secpolicy_config, POOL_NAME, B_TRUE, 4277 B_TRUE }, 4278 { zfs_ioc_pool_get_props, zfs_secpolicy_read, POOL_NAME, B_FALSE, 4279 B_FALSE }, 4280 { zfs_ioc_set_fsacl, zfs_secpolicy_fsacl, DATASET_NAME, B_TRUE, 4281 B_TRUE }, 4282 { zfs_ioc_get_fsacl, zfs_secpolicy_read, DATASET_NAME, B_FALSE, 4283 B_FALSE }, 4284 { zfs_ioc_iscsi_perm_check, zfs_secpolicy_iscsi, DATASET_NAME, B_FALSE, 4285 B_FALSE }, 4286 { zfs_ioc_share, zfs_secpolicy_share, DATASET_NAME, B_FALSE, B_FALSE }, 4287 { zfs_ioc_inherit_prop, zfs_secpolicy_inherit, DATASET_NAME, B_TRUE, 4288 B_TRUE }, 4289 { zfs_ioc_smb_acl, zfs_secpolicy_smb_acl, DATASET_NAME, B_FALSE, 4290 B_FALSE }, 4291 { zfs_ioc_userspace_one, zfs_secpolicy_userspace_one, 4292 DATASET_NAME, B_FALSE, B_FALSE }, 4293 { zfs_ioc_userspace_many, zfs_secpolicy_userspace_many, 4294 DATASET_NAME, B_FALSE, B_FALSE }, 4295 { zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade, 4296 DATASET_NAME, B_FALSE, B_TRUE }, 4297 { zfs_ioc_hold, zfs_secpolicy_hold, DATASET_NAME, B_TRUE, B_TRUE }, 4298 { zfs_ioc_release, zfs_secpolicy_release, DATASET_NAME, B_TRUE, 4299 B_TRUE }, 4300 { zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME, B_FALSE, 4301 B_TRUE }, 4302 { zfs_ioc_objset_recvd_props, zfs_secpolicy_read, DATASET_NAME, B_FALSE, 4303 B_FALSE } 4304 }; 4305 4306 int 4307 pool_status_check(const char *name, zfs_ioc_namecheck_t type) 4308 { 4309 spa_t *spa; 4310 int error; 4311 4312 ASSERT(type == POOL_NAME || type == DATASET_NAME); 4313 4314 error = spa_open(name, &spa, FTAG); 4315 if (error == 0) { 4316 if (spa_suspended(spa)) 4317 error = EAGAIN; 4318 spa_close(spa, FTAG); 4319 } 4320 return (error); 4321 } 4322 4323 static int 4324 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp) 4325 { 4326 zfs_cmd_t *zc; 4327 uint_t vec; 4328 int error, rc; 4329 4330 if (getminor(dev) != 0) 4331 return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp)); 4332 4333 vec = cmd - ZFS_IOC; 4334 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip)); 4335 4336 if (vec >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0])) 4337 return (EINVAL); 4338 4339 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP); 4340 4341 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag); 4342 4343 if ((error == 0) && !(flag & FKIOCTL)) 4344 error = zfs_ioc_vec[vec].zvec_secpolicy(zc, cr); 4345 4346 /* 4347 * Ensure that all pool/dataset names are valid before we pass down to 4348 * the lower layers. 4349 */ 4350 if (error == 0) { 4351 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0'; 4352 zc->zc_iflags = flag & FKIOCTL; 4353 switch (zfs_ioc_vec[vec].zvec_namecheck) { 4354 case POOL_NAME: 4355 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0) 4356 error = EINVAL; 4357 if (zfs_ioc_vec[vec].zvec_pool_check) 4358 error = pool_status_check(zc->zc_name, 4359 zfs_ioc_vec[vec].zvec_namecheck); 4360 break; 4361 4362 case DATASET_NAME: 4363 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0) 4364 error = EINVAL; 4365 if (zfs_ioc_vec[vec].zvec_pool_check) 4366 error = pool_status_check(zc->zc_name, 4367 zfs_ioc_vec[vec].zvec_namecheck); 4368 break; 4369 4370 case NO_NAME: 4371 break; 4372 } 4373 } 4374 4375 if (error == 0) 4376 error = zfs_ioc_vec[vec].zvec_func(zc); 4377 4378 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag); 4379 if (error == 0) { 4380 error = rc; 4381 if (zfs_ioc_vec[vec].zvec_his_log) 4382 zfs_log_history(zc); 4383 } 4384 4385 kmem_free(zc, sizeof (zfs_cmd_t)); 4386 return (error); 4387 } 4388 4389 static int 4390 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd) 4391 { 4392 if (cmd != DDI_ATTACH) 4393 return (DDI_FAILURE); 4394 4395 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0, 4396 DDI_PSEUDO, 0) == DDI_FAILURE) 4397 return (DDI_FAILURE); 4398 4399 zfs_dip = dip; 4400 4401 ddi_report_dev(dip); 4402 4403 return (DDI_SUCCESS); 4404 } 4405 4406 static int 4407 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd) 4408 { 4409 if (spa_busy() || zfs_busy() || zvol_busy()) 4410 return (DDI_FAILURE); 4411 4412 if (cmd != DDI_DETACH) 4413 return (DDI_FAILURE); 4414 4415 zfs_dip = NULL; 4416 4417 ddi_prop_remove_all(dip); 4418 ddi_remove_minor_node(dip, NULL); 4419 4420 return (DDI_SUCCESS); 4421 } 4422 4423 /*ARGSUSED*/ 4424 static int 4425 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result) 4426 { 4427 switch (infocmd) { 4428 case DDI_INFO_DEVT2DEVINFO: 4429 *result = zfs_dip; 4430 return (DDI_SUCCESS); 4431 4432 case DDI_INFO_DEVT2INSTANCE: 4433 *result = (void *)0; 4434 return (DDI_SUCCESS); 4435 } 4436 4437 return (DDI_FAILURE); 4438 } 4439 4440 /* 4441 * OK, so this is a little weird. 4442 * 4443 * /dev/zfs is the control node, i.e. minor 0. 4444 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0. 4445 * 4446 * /dev/zfs has basically nothing to do except serve up ioctls, 4447 * so most of the standard driver entry points are in zvol.c. 4448 */ 4449 static struct cb_ops zfs_cb_ops = { 4450 zvol_open, /* open */ 4451 zvol_close, /* close */ 4452 zvol_strategy, /* strategy */ 4453 nodev, /* print */ 4454 zvol_dump, /* dump */ 4455 zvol_read, /* read */ 4456 zvol_write, /* write */ 4457 zfsdev_ioctl, /* ioctl */ 4458 nodev, /* devmap */ 4459 nodev, /* mmap */ 4460 nodev, /* segmap */ 4461 nochpoll, /* poll */ 4462 ddi_prop_op, /* prop_op */ 4463 NULL, /* streamtab */ 4464 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */ 4465 CB_REV, /* version */ 4466 nodev, /* async read */ 4467 nodev, /* async write */ 4468 }; 4469 4470 static struct dev_ops zfs_dev_ops = { 4471 DEVO_REV, /* version */ 4472 0, /* refcnt */ 4473 zfs_info, /* info */ 4474 nulldev, /* identify */ 4475 nulldev, /* probe */ 4476 zfs_attach, /* attach */ 4477 zfs_detach, /* detach */ 4478 nodev, /* reset */ 4479 &zfs_cb_ops, /* driver operations */ 4480 NULL, /* no bus operations */ 4481 NULL, /* power */ 4482 ddi_quiesce_not_needed, /* quiesce */ 4483 }; 4484 4485 static struct modldrv zfs_modldrv = { 4486 &mod_driverops, 4487 "ZFS storage pool", 4488 &zfs_dev_ops 4489 }; 4490 4491 static struct modlinkage modlinkage = { 4492 MODREV_1, 4493 (void *)&zfs_modlfs, 4494 (void *)&zfs_modldrv, 4495 NULL 4496 }; 4497 4498 4499 uint_t zfs_fsyncer_key; 4500 extern uint_t rrw_tsd_key; 4501 4502 int 4503 _init(void) 4504 { 4505 int error; 4506 4507 spa_init(FREAD | FWRITE); 4508 zfs_init(); 4509 zvol_init(); 4510 4511 if ((error = mod_install(&modlinkage)) != 0) { 4512 zvol_fini(); 4513 zfs_fini(); 4514 spa_fini(); 4515 return (error); 4516 } 4517 4518 tsd_create(&zfs_fsyncer_key, NULL); 4519 tsd_create(&rrw_tsd_key, NULL); 4520 4521 error = ldi_ident_from_mod(&modlinkage, &zfs_li); 4522 ASSERT(error == 0); 4523 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL); 4524 4525 return (0); 4526 } 4527 4528 int 4529 _fini(void) 4530 { 4531 int error; 4532 4533 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled) 4534 return (EBUSY); 4535 4536 if ((error = mod_remove(&modlinkage)) != 0) 4537 return (error); 4538 4539 zvol_fini(); 4540 zfs_fini(); 4541 spa_fini(); 4542 if (zfs_nfsshare_inited) 4543 (void) ddi_modclose(nfs_mod); 4544 if (zfs_smbshare_inited) 4545 (void) ddi_modclose(smbsrv_mod); 4546 if (zfs_nfsshare_inited || zfs_smbshare_inited) 4547 (void) ddi_modclose(sharefs_mod); 4548 4549 tsd_destroy(&zfs_fsyncer_key); 4550 ldi_ident_release(zfs_li); 4551 zfs_li = NULL; 4552 mutex_destroy(&zfs_share_lock); 4553 4554 return (error); 4555 } 4556 4557 int 4558 _info(struct modinfo *modinfop) 4559 { 4560 return (mod_info(&modlinkage, modinfop)); 4561 } 4562