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 /* 23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 24 * Copyright (c) 2011-2012 Pawel Jakub Dawidek. All rights reserved. 25 * Portions Copyright 2011 Martin Matuska 26 * Copyright 2015, OmniTI Computer Consulting, Inc. All rights reserved. 27 * Copyright 2015 Nexenta Systems, Inc. All rights reserved. 28 * Copyright (c) 2014, 2016 Joyent, Inc. All rights reserved. 29 * Copyright (c) 2011, 2016 by Delphix. All rights reserved. 30 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved. 31 * Copyright (c) 2013 Steven Hartland. All rights reserved. 32 * Copyright (c) 2014 Integros [integros.com] 33 * Copyright 2016 Toomas Soome <tsoome@me.com> 34 * Copyright 2017 RackTop Systems. 35 */ 36 37 /* 38 * ZFS ioctls. 39 * 40 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage 41 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool. 42 * 43 * There are two ways that we handle ioctls: the legacy way where almost 44 * all of the logic is in the ioctl callback, and the new way where most 45 * of the marshalling is handled in the common entry point, zfsdev_ioctl(). 46 * 47 * Non-legacy ioctls should be registered by calling 48 * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked 49 * from userland by lzc_ioctl(). 50 * 51 * The registration arguments are as follows: 52 * 53 * const char *name 54 * The name of the ioctl. This is used for history logging. If the 55 * ioctl returns successfully (the callback returns 0), and allow_log 56 * is true, then a history log entry will be recorded with the input & 57 * output nvlists. The log entry can be printed with "zpool history -i". 58 * 59 * zfs_ioc_t ioc 60 * The ioctl request number, which userland will pass to ioctl(2). 61 * The ioctl numbers can change from release to release, because 62 * the caller (libzfs) must be matched to the kernel. 63 * 64 * zfs_secpolicy_func_t *secpolicy 65 * This function will be called before the zfs_ioc_func_t, to 66 * determine if this operation is permitted. It should return EPERM 67 * on failure, and 0 on success. Checks include determining if the 68 * dataset is visible in this zone, and if the user has either all 69 * zfs privileges in the zone (SYS_MOUNT), or has been granted permission 70 * to do this operation on this dataset with "zfs allow". 71 * 72 * zfs_ioc_namecheck_t namecheck 73 * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool 74 * name, a dataset name, or nothing. If the name is not well-formed, 75 * the ioctl will fail and the callback will not be called. 76 * Therefore, the callback can assume that the name is well-formed 77 * (e.g. is null-terminated, doesn't have more than one '@' character, 78 * doesn't have invalid characters). 79 * 80 * zfs_ioc_poolcheck_t pool_check 81 * This specifies requirements on the pool state. If the pool does 82 * not meet them (is suspended or is readonly), the ioctl will fail 83 * and the callback will not be called. If any checks are specified 84 * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME. 85 * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED | 86 * POOL_CHECK_READONLY). 87 * 88 * boolean_t smush_outnvlist 89 * If smush_outnvlist is true, then the output is presumed to be a 90 * list of errors, and it will be "smushed" down to fit into the 91 * caller's buffer, by removing some entries and replacing them with a 92 * single "N_MORE_ERRORS" entry indicating how many were removed. See 93 * nvlist_smush() for details. If smush_outnvlist is false, and the 94 * outnvlist does not fit into the userland-provided buffer, then the 95 * ioctl will fail with ENOMEM. 96 * 97 * zfs_ioc_func_t *func 98 * The callback function that will perform the operation. 99 * 100 * The callback should return 0 on success, or an error number on 101 * failure. If the function fails, the userland ioctl will return -1, 102 * and errno will be set to the callback's return value. The callback 103 * will be called with the following arguments: 104 * 105 * const char *name 106 * The name of the pool or dataset to operate on, from 107 * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the 108 * expected type (pool, dataset, or none). 109 * 110 * nvlist_t *innvl 111 * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or 112 * NULL if no input nvlist was provided. Changes to this nvlist are 113 * ignored. If the input nvlist could not be deserialized, the 114 * ioctl will fail and the callback will not be called. 115 * 116 * nvlist_t *outnvl 117 * The output nvlist, initially empty. The callback can fill it in, 118 * and it will be returned to userland by serializing it into 119 * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization 120 * fails (e.g. because the caller didn't supply a large enough 121 * buffer), then the overall ioctl will fail. See the 122 * 'smush_nvlist' argument above for additional behaviors. 123 * 124 * There are two typical uses of the output nvlist: 125 * - To return state, e.g. property values. In this case, 126 * smush_outnvlist should be false. If the buffer was not large 127 * enough, the caller will reallocate a larger buffer and try 128 * the ioctl again. 129 * 130 * - To return multiple errors from an ioctl which makes on-disk 131 * changes. In this case, smush_outnvlist should be true. 132 * Ioctls which make on-disk modifications should generally not 133 * use the outnvl if they succeed, because the caller can not 134 * distinguish between the operation failing, and 135 * deserialization failing. 136 */ 137 138 #include <sys/types.h> 139 #include <sys/param.h> 140 #include <sys/errno.h> 141 #include <sys/uio.h> 142 #include <sys/buf.h> 143 #include <sys/modctl.h> 144 #include <sys/open.h> 145 #include <sys/file.h> 146 #include <sys/kmem.h> 147 #include <sys/conf.h> 148 #include <sys/cmn_err.h> 149 #include <sys/stat.h> 150 #include <sys/zfs_ioctl.h> 151 #include <sys/zfs_vfsops.h> 152 #include <sys/zfs_znode.h> 153 #include <sys/zap.h> 154 #include <sys/spa.h> 155 #include <sys/spa_impl.h> 156 #include <sys/vdev.h> 157 #include <sys/priv_impl.h> 158 #include <sys/dmu.h> 159 #include <sys/dsl_dir.h> 160 #include <sys/dsl_dataset.h> 161 #include <sys/dsl_prop.h> 162 #include <sys/dsl_deleg.h> 163 #include <sys/dmu_objset.h> 164 #include <sys/dmu_impl.h> 165 #include <sys/dmu_tx.h> 166 #include <sys/ddi.h> 167 #include <sys/sunddi.h> 168 #include <sys/sunldi.h> 169 #include <sys/policy.h> 170 #include <sys/zone.h> 171 #include <sys/nvpair.h> 172 #include <sys/pathname.h> 173 #include <sys/mount.h> 174 #include <sys/sdt.h> 175 #include <sys/fs/zfs.h> 176 #include <sys/zfs_ctldir.h> 177 #include <sys/zfs_dir.h> 178 #include <sys/zfs_onexit.h> 179 #include <sys/zvol.h> 180 #include <sys/dsl_scan.h> 181 #include <sharefs/share.h> 182 #include <sys/dmu_objset.h> 183 #include <sys/dmu_send.h> 184 #include <sys/dsl_destroy.h> 185 #include <sys/dsl_bookmark.h> 186 #include <sys/dsl_userhold.h> 187 #include <sys/zfeature.h> 188 #include <sys/zcp.h> 189 #include <sys/zio_checksum.h> 190 191 #include "zfs_namecheck.h" 192 #include "zfs_prop.h" 193 #include "zfs_deleg.h" 194 #include "zfs_comutil.h" 195 196 #include "lua.h" 197 #include "lauxlib.h" 198 199 extern struct modlfs zfs_modlfs; 200 201 extern void zfs_init(void); 202 extern void zfs_fini(void); 203 204 ldi_ident_t zfs_li = NULL; 205 dev_info_t *zfs_dip; 206 207 uint_t zfs_fsyncer_key; 208 extern uint_t rrw_tsd_key; 209 static uint_t zfs_allow_log_key; 210 211 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *); 212 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *); 213 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *); 214 215 typedef enum { 216 NO_NAME, 217 POOL_NAME, 218 DATASET_NAME 219 } zfs_ioc_namecheck_t; 220 221 typedef enum { 222 POOL_CHECK_NONE = 1 << 0, 223 POOL_CHECK_SUSPENDED = 1 << 1, 224 POOL_CHECK_READONLY = 1 << 2, 225 } zfs_ioc_poolcheck_t; 226 227 typedef struct zfs_ioc_vec { 228 zfs_ioc_legacy_func_t *zvec_legacy_func; 229 zfs_ioc_func_t *zvec_func; 230 zfs_secpolicy_func_t *zvec_secpolicy; 231 zfs_ioc_namecheck_t zvec_namecheck; 232 boolean_t zvec_allow_log; 233 zfs_ioc_poolcheck_t zvec_pool_check; 234 boolean_t zvec_smush_outnvlist; 235 const char *zvec_name; 236 } zfs_ioc_vec_t; 237 238 /* This array is indexed by zfs_userquota_prop_t */ 239 static const char *userquota_perms[] = { 240 ZFS_DELEG_PERM_USERUSED, 241 ZFS_DELEG_PERM_USERQUOTA, 242 ZFS_DELEG_PERM_GROUPUSED, 243 ZFS_DELEG_PERM_GROUPQUOTA, 244 }; 245 246 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc); 247 static int zfs_check_settable(const char *name, nvpair_t *property, 248 cred_t *cr); 249 static int zfs_check_clearable(char *dataset, nvlist_t *props, 250 nvlist_t **errors); 251 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *, 252 boolean_t *); 253 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *); 254 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp); 255 256 static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature); 257 258 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */ 259 void 260 __dprintf(const char *file, const char *func, int line, const char *fmt, ...) 261 { 262 const char *newfile; 263 char buf[512]; 264 va_list adx; 265 266 /* 267 * Get rid of annoying "../common/" prefix to filename. 268 */ 269 newfile = strrchr(file, '/'); 270 if (newfile != NULL) { 271 newfile = newfile + 1; /* Get rid of leading / */ 272 } else { 273 newfile = file; 274 } 275 276 va_start(adx, fmt); 277 (void) vsnprintf(buf, sizeof (buf), fmt, adx); 278 va_end(adx); 279 280 /* 281 * To get this data, use the zfs-dprintf probe as so: 282 * dtrace -q -n 'zfs-dprintf \ 283 * /stringof(arg0) == "dbuf.c"/ \ 284 * {printf("%s: %s", stringof(arg1), stringof(arg3))}' 285 * arg0 = file name 286 * arg1 = function name 287 * arg2 = line number 288 * arg3 = message 289 */ 290 DTRACE_PROBE4(zfs__dprintf, 291 char *, newfile, char *, func, int, line, char *, buf); 292 } 293 294 static void 295 history_str_free(char *buf) 296 { 297 kmem_free(buf, HIS_MAX_RECORD_LEN); 298 } 299 300 static char * 301 history_str_get(zfs_cmd_t *zc) 302 { 303 char *buf; 304 305 if (zc->zc_history == NULL) 306 return (NULL); 307 308 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP); 309 if (copyinstr((void *)(uintptr_t)zc->zc_history, 310 buf, HIS_MAX_RECORD_LEN, NULL) != 0) { 311 history_str_free(buf); 312 return (NULL); 313 } 314 315 buf[HIS_MAX_RECORD_LEN -1] = '\0'; 316 317 return (buf); 318 } 319 320 /* 321 * Check to see if the named dataset is currently defined as bootable 322 */ 323 static boolean_t 324 zfs_is_bootfs(const char *name) 325 { 326 objset_t *os; 327 328 if (dmu_objset_hold(name, FTAG, &os) == 0) { 329 boolean_t ret; 330 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os))); 331 dmu_objset_rele(os, FTAG); 332 return (ret); 333 } 334 return (B_FALSE); 335 } 336 337 /* 338 * Return non-zero if the spa version is less than requested version. 339 */ 340 static int 341 zfs_earlier_version(const char *name, int version) 342 { 343 spa_t *spa; 344 345 if (spa_open(name, &spa, FTAG) == 0) { 346 if (spa_version(spa) < version) { 347 spa_close(spa, FTAG); 348 return (1); 349 } 350 spa_close(spa, FTAG); 351 } 352 return (0); 353 } 354 355 /* 356 * Return TRUE if the ZPL version is less than requested version. 357 */ 358 static boolean_t 359 zpl_earlier_version(const char *name, int version) 360 { 361 objset_t *os; 362 boolean_t rc = B_TRUE; 363 364 if (dmu_objset_hold(name, FTAG, &os) == 0) { 365 uint64_t zplversion; 366 367 if (dmu_objset_type(os) != DMU_OST_ZFS) { 368 dmu_objset_rele(os, FTAG); 369 return (B_TRUE); 370 } 371 /* XXX reading from non-owned objset */ 372 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0) 373 rc = zplversion < version; 374 dmu_objset_rele(os, FTAG); 375 } 376 return (rc); 377 } 378 379 static void 380 zfs_log_history(zfs_cmd_t *zc) 381 { 382 spa_t *spa; 383 char *buf; 384 385 if ((buf = history_str_get(zc)) == NULL) 386 return; 387 388 if (spa_open(zc->zc_name, &spa, FTAG) == 0) { 389 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY) 390 (void) spa_history_log(spa, buf); 391 spa_close(spa, FTAG); 392 } 393 history_str_free(buf); 394 } 395 396 /* 397 * Policy for top-level read operations (list pools). Requires no privileges, 398 * and can be used in the local zone, as there is no associated dataset. 399 */ 400 /* ARGSUSED */ 401 static int 402 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 403 { 404 return (0); 405 } 406 407 /* 408 * Policy for dataset read operations (list children, get statistics). Requires 409 * no privileges, but must be visible in the local zone. 410 */ 411 /* ARGSUSED */ 412 static int 413 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 414 { 415 if (INGLOBALZONE(curproc) || 416 zone_dataset_visible(zc->zc_name, NULL)) 417 return (0); 418 419 return (SET_ERROR(ENOENT)); 420 } 421 422 static int 423 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr) 424 { 425 int writable = 1; 426 427 /* 428 * The dataset must be visible by this zone -- check this first 429 * so they don't see EPERM on something they shouldn't know about. 430 */ 431 if (!INGLOBALZONE(curproc) && 432 !zone_dataset_visible(dataset, &writable)) 433 return (SET_ERROR(ENOENT)); 434 435 if (INGLOBALZONE(curproc)) { 436 /* 437 * If the fs is zoned, only root can access it from the 438 * global zone. 439 */ 440 if (secpolicy_zfs(cr) && zoned) 441 return (SET_ERROR(EPERM)); 442 } else { 443 /* 444 * If we are in a local zone, the 'zoned' property must be set. 445 */ 446 if (!zoned) 447 return (SET_ERROR(EPERM)); 448 449 /* must be writable by this zone */ 450 if (!writable) 451 return (SET_ERROR(EPERM)); 452 } 453 return (0); 454 } 455 456 static int 457 zfs_dozonecheck(const char *dataset, cred_t *cr) 458 { 459 uint64_t zoned; 460 461 if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL)) 462 return (SET_ERROR(ENOENT)); 463 464 return (zfs_dozonecheck_impl(dataset, zoned, cr)); 465 } 466 467 static int 468 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr) 469 { 470 uint64_t zoned; 471 472 if (dsl_prop_get_int_ds(ds, "zoned", &zoned)) 473 return (SET_ERROR(ENOENT)); 474 475 return (zfs_dozonecheck_impl(dataset, zoned, cr)); 476 } 477 478 static int 479 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds, 480 const char *perm, cred_t *cr) 481 { 482 int error; 483 484 error = zfs_dozonecheck_ds(name, ds, cr); 485 if (error == 0) { 486 error = secpolicy_zfs(cr); 487 if (error != 0) 488 error = dsl_deleg_access_impl(ds, perm, cr); 489 } 490 return (error); 491 } 492 493 static int 494 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr) 495 { 496 int error; 497 dsl_dataset_t *ds; 498 dsl_pool_t *dp; 499 500 /* 501 * First do a quick check for root in the global zone, which 502 * is allowed to do all write_perms. This ensures that zfs_ioc_* 503 * will get to handle nonexistent datasets. 504 */ 505 if (INGLOBALZONE(curproc) && secpolicy_zfs(cr) == 0) 506 return (0); 507 508 error = dsl_pool_hold(name, FTAG, &dp); 509 if (error != 0) 510 return (error); 511 512 error = dsl_dataset_hold(dp, name, FTAG, &ds); 513 if (error != 0) { 514 dsl_pool_rele(dp, FTAG); 515 return (error); 516 } 517 518 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr); 519 520 dsl_dataset_rele(ds, FTAG); 521 dsl_pool_rele(dp, FTAG); 522 return (error); 523 } 524 525 /* 526 * Policy for setting the security label property. 527 * 528 * Returns 0 for success, non-zero for access and other errors. 529 */ 530 static int 531 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr) 532 { 533 char ds_hexsl[MAXNAMELEN]; 534 bslabel_t ds_sl, new_sl; 535 boolean_t new_default = FALSE; 536 uint64_t zoned; 537 int needed_priv = -1; 538 int error; 539 540 /* First get the existing dataset label. */ 541 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL), 542 1, sizeof (ds_hexsl), &ds_hexsl, NULL); 543 if (error != 0) 544 return (SET_ERROR(EPERM)); 545 546 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0) 547 new_default = TRUE; 548 549 /* The label must be translatable */ 550 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0)) 551 return (SET_ERROR(EINVAL)); 552 553 /* 554 * In a non-global zone, disallow attempts to set a label that 555 * doesn't match that of the zone; otherwise no other checks 556 * are needed. 557 */ 558 if (!INGLOBALZONE(curproc)) { 559 if (new_default || !blequal(&new_sl, CR_SL(CRED()))) 560 return (SET_ERROR(EPERM)); 561 return (0); 562 } 563 564 /* 565 * For global-zone datasets (i.e., those whose zoned property is 566 * "off", verify that the specified new label is valid for the 567 * global zone. 568 */ 569 if (dsl_prop_get_integer(name, 570 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL)) 571 return (SET_ERROR(EPERM)); 572 if (!zoned) { 573 if (zfs_check_global_label(name, strval) != 0) 574 return (SET_ERROR(EPERM)); 575 } 576 577 /* 578 * If the existing dataset label is nondefault, check if the 579 * dataset is mounted (label cannot be changed while mounted). 580 * Get the zfsvfs; if there isn't one, then the dataset isn't 581 * mounted (or isn't a dataset, doesn't exist, ...). 582 */ 583 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) { 584 objset_t *os; 585 static char *setsl_tag = "setsl_tag"; 586 587 /* 588 * Try to own the dataset; abort if there is any error, 589 * (e.g., already mounted, in use, or other error). 590 */ 591 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE, 592 setsl_tag, &os); 593 if (error != 0) 594 return (SET_ERROR(EPERM)); 595 596 dmu_objset_disown(os, setsl_tag); 597 598 if (new_default) { 599 needed_priv = PRIV_FILE_DOWNGRADE_SL; 600 goto out_check; 601 } 602 603 if (hexstr_to_label(strval, &new_sl) != 0) 604 return (SET_ERROR(EPERM)); 605 606 if (blstrictdom(&ds_sl, &new_sl)) 607 needed_priv = PRIV_FILE_DOWNGRADE_SL; 608 else if (blstrictdom(&new_sl, &ds_sl)) 609 needed_priv = PRIV_FILE_UPGRADE_SL; 610 } else { 611 /* dataset currently has a default label */ 612 if (!new_default) 613 needed_priv = PRIV_FILE_UPGRADE_SL; 614 } 615 616 out_check: 617 if (needed_priv != -1) 618 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL)); 619 return (0); 620 } 621 622 static int 623 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval, 624 cred_t *cr) 625 { 626 char *strval; 627 628 /* 629 * Check permissions for special properties. 630 */ 631 switch (prop) { 632 case ZFS_PROP_ZONED: 633 /* 634 * Disallow setting of 'zoned' from within a local zone. 635 */ 636 if (!INGLOBALZONE(curproc)) 637 return (SET_ERROR(EPERM)); 638 break; 639 640 case ZFS_PROP_QUOTA: 641 case ZFS_PROP_FILESYSTEM_LIMIT: 642 case ZFS_PROP_SNAPSHOT_LIMIT: 643 if (!INGLOBALZONE(curproc)) { 644 uint64_t zoned; 645 char setpoint[ZFS_MAX_DATASET_NAME_LEN]; 646 /* 647 * Unprivileged users are allowed to modify the 648 * limit on things *under* (ie. contained by) 649 * the thing they own. 650 */ 651 if (dsl_prop_get_integer(dsname, "zoned", &zoned, 652 setpoint)) 653 return (SET_ERROR(EPERM)); 654 if (!zoned || strlen(dsname) <= strlen(setpoint)) 655 return (SET_ERROR(EPERM)); 656 } 657 break; 658 659 case ZFS_PROP_MLSLABEL: 660 if (!is_system_labeled()) 661 return (SET_ERROR(EPERM)); 662 663 if (nvpair_value_string(propval, &strval) == 0) { 664 int err; 665 666 err = zfs_set_slabel_policy(dsname, strval, CRED()); 667 if (err != 0) 668 return (err); 669 } 670 break; 671 } 672 673 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr)); 674 } 675 676 /* ARGSUSED */ 677 static int 678 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 679 { 680 int error; 681 682 error = zfs_dozonecheck(zc->zc_name, cr); 683 if (error != 0) 684 return (error); 685 686 /* 687 * permission to set permissions will be evaluated later in 688 * dsl_deleg_can_allow() 689 */ 690 return (0); 691 } 692 693 /* ARGSUSED */ 694 static int 695 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 696 { 697 return (zfs_secpolicy_write_perms(zc->zc_name, 698 ZFS_DELEG_PERM_ROLLBACK, cr)); 699 } 700 701 /* ARGSUSED */ 702 static int 703 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 704 { 705 dsl_pool_t *dp; 706 dsl_dataset_t *ds; 707 char *cp; 708 int error; 709 710 /* 711 * Generate the current snapshot name from the given objsetid, then 712 * use that name for the secpolicy/zone checks. 713 */ 714 cp = strchr(zc->zc_name, '@'); 715 if (cp == NULL) 716 return (SET_ERROR(EINVAL)); 717 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 718 if (error != 0) 719 return (error); 720 721 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds); 722 if (error != 0) { 723 dsl_pool_rele(dp, FTAG); 724 return (error); 725 } 726 727 dsl_dataset_name(ds, zc->zc_name); 728 729 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds, 730 ZFS_DELEG_PERM_SEND, cr); 731 dsl_dataset_rele(ds, FTAG); 732 dsl_pool_rele(dp, FTAG); 733 734 return (error); 735 } 736 737 /* ARGSUSED */ 738 static int 739 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 740 { 741 return (zfs_secpolicy_write_perms(zc->zc_name, 742 ZFS_DELEG_PERM_SEND, cr)); 743 } 744 745 /* ARGSUSED */ 746 static int 747 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 748 { 749 vnode_t *vp; 750 int error; 751 752 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, 753 NO_FOLLOW, NULL, &vp)) != 0) 754 return (error); 755 756 /* Now make sure mntpnt and dataset are ZFS */ 757 758 if (vp->v_vfsp->vfs_fstype != zfsfstype || 759 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), 760 zc->zc_name) != 0)) { 761 VN_RELE(vp); 762 return (SET_ERROR(EPERM)); 763 } 764 765 VN_RELE(vp); 766 return (dsl_deleg_access(zc->zc_name, 767 ZFS_DELEG_PERM_SHARE, cr)); 768 } 769 770 int 771 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 772 { 773 if (!INGLOBALZONE(curproc)) 774 return (SET_ERROR(EPERM)); 775 776 if (secpolicy_nfs(cr) == 0) { 777 return (0); 778 } else { 779 return (zfs_secpolicy_deleg_share(zc, innvl, cr)); 780 } 781 } 782 783 int 784 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 785 { 786 if (!INGLOBALZONE(curproc)) 787 return (SET_ERROR(EPERM)); 788 789 if (secpolicy_smb(cr) == 0) { 790 return (0); 791 } else { 792 return (zfs_secpolicy_deleg_share(zc, innvl, cr)); 793 } 794 } 795 796 static int 797 zfs_get_parent(const char *datasetname, char *parent, int parentsize) 798 { 799 char *cp; 800 801 /* 802 * Remove the @bla or /bla from the end of the name to get the parent. 803 */ 804 (void) strncpy(parent, datasetname, parentsize); 805 cp = strrchr(parent, '@'); 806 if (cp != NULL) { 807 cp[0] = '\0'; 808 } else { 809 cp = strrchr(parent, '/'); 810 if (cp == NULL) 811 return (SET_ERROR(ENOENT)); 812 cp[0] = '\0'; 813 } 814 815 return (0); 816 } 817 818 int 819 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr) 820 { 821 int error; 822 823 if ((error = zfs_secpolicy_write_perms(name, 824 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 825 return (error); 826 827 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr)); 828 } 829 830 /* ARGSUSED */ 831 static int 832 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 833 { 834 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr)); 835 } 836 837 /* 838 * Destroying snapshots with delegated permissions requires 839 * descendant mount and destroy permissions. 840 */ 841 /* ARGSUSED */ 842 static int 843 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 844 { 845 nvlist_t *snaps; 846 nvpair_t *pair, *nextpair; 847 int error = 0; 848 849 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 850 return (SET_ERROR(EINVAL)); 851 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 852 pair = nextpair) { 853 nextpair = nvlist_next_nvpair(snaps, pair); 854 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr); 855 if (error == ENOENT) { 856 /* 857 * Ignore any snapshots that don't exist (we consider 858 * them "already destroyed"). Remove the name from the 859 * nvl here in case the snapshot is created between 860 * now and when we try to destroy it (in which case 861 * we don't want to destroy it since we haven't 862 * checked for permission). 863 */ 864 fnvlist_remove_nvpair(snaps, pair); 865 error = 0; 866 } 867 if (error != 0) 868 break; 869 } 870 871 return (error); 872 } 873 874 int 875 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr) 876 { 877 char parentname[ZFS_MAX_DATASET_NAME_LEN]; 878 int error; 879 880 if ((error = zfs_secpolicy_write_perms(from, 881 ZFS_DELEG_PERM_RENAME, cr)) != 0) 882 return (error); 883 884 if ((error = zfs_secpolicy_write_perms(from, 885 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 886 return (error); 887 888 if ((error = zfs_get_parent(to, parentname, 889 sizeof (parentname))) != 0) 890 return (error); 891 892 if ((error = zfs_secpolicy_write_perms(parentname, 893 ZFS_DELEG_PERM_CREATE, cr)) != 0) 894 return (error); 895 896 if ((error = zfs_secpolicy_write_perms(parentname, 897 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 898 return (error); 899 900 return (error); 901 } 902 903 /* ARGSUSED */ 904 static int 905 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 906 { 907 return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr)); 908 } 909 910 /* ARGSUSED */ 911 static int 912 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 913 { 914 dsl_pool_t *dp; 915 dsl_dataset_t *clone; 916 int error; 917 918 error = zfs_secpolicy_write_perms(zc->zc_name, 919 ZFS_DELEG_PERM_PROMOTE, cr); 920 if (error != 0) 921 return (error); 922 923 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 924 if (error != 0) 925 return (error); 926 927 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone); 928 929 if (error == 0) { 930 char parentname[ZFS_MAX_DATASET_NAME_LEN]; 931 dsl_dataset_t *origin = NULL; 932 dsl_dir_t *dd; 933 dd = clone->ds_dir; 934 935 error = dsl_dataset_hold_obj(dd->dd_pool, 936 dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin); 937 if (error != 0) { 938 dsl_dataset_rele(clone, FTAG); 939 dsl_pool_rele(dp, FTAG); 940 return (error); 941 } 942 943 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone, 944 ZFS_DELEG_PERM_MOUNT, cr); 945 946 dsl_dataset_name(origin, parentname); 947 if (error == 0) { 948 error = zfs_secpolicy_write_perms_ds(parentname, origin, 949 ZFS_DELEG_PERM_PROMOTE, cr); 950 } 951 dsl_dataset_rele(clone, FTAG); 952 dsl_dataset_rele(origin, FTAG); 953 } 954 dsl_pool_rele(dp, FTAG); 955 return (error); 956 } 957 958 /* ARGSUSED */ 959 static int 960 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 961 { 962 int error; 963 964 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 965 ZFS_DELEG_PERM_RECEIVE, cr)) != 0) 966 return (error); 967 968 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 969 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 970 return (error); 971 972 return (zfs_secpolicy_write_perms(zc->zc_name, 973 ZFS_DELEG_PERM_CREATE, cr)); 974 } 975 976 int 977 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr) 978 { 979 return (zfs_secpolicy_write_perms(name, 980 ZFS_DELEG_PERM_SNAPSHOT, cr)); 981 } 982 983 /* 984 * Check for permission to create each snapshot in the nvlist. 985 */ 986 /* ARGSUSED */ 987 static int 988 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 989 { 990 nvlist_t *snaps; 991 int error = 0; 992 nvpair_t *pair; 993 994 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 995 return (SET_ERROR(EINVAL)); 996 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 997 pair = nvlist_next_nvpair(snaps, pair)) { 998 char *name = nvpair_name(pair); 999 char *atp = strchr(name, '@'); 1000 1001 if (atp == NULL) { 1002 error = SET_ERROR(EINVAL); 1003 break; 1004 } 1005 *atp = '\0'; 1006 error = zfs_secpolicy_snapshot_perms(name, cr); 1007 *atp = '@'; 1008 if (error != 0) 1009 break; 1010 } 1011 return (error); 1012 } 1013 1014 /* 1015 * Check for permission to create each snapshot in the nvlist. 1016 */ 1017 /* ARGSUSED */ 1018 static int 1019 zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1020 { 1021 int error = 0; 1022 1023 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 1024 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 1025 char *name = nvpair_name(pair); 1026 char *hashp = strchr(name, '#'); 1027 1028 if (hashp == NULL) { 1029 error = SET_ERROR(EINVAL); 1030 break; 1031 } 1032 *hashp = '\0'; 1033 error = zfs_secpolicy_write_perms(name, 1034 ZFS_DELEG_PERM_BOOKMARK, cr); 1035 *hashp = '#'; 1036 if (error != 0) 1037 break; 1038 } 1039 return (error); 1040 } 1041 1042 /* ARGSUSED */ 1043 static int 1044 zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1045 { 1046 nvpair_t *pair, *nextpair; 1047 int error = 0; 1048 1049 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL; 1050 pair = nextpair) { 1051 char *name = nvpair_name(pair); 1052 char *hashp = strchr(name, '#'); 1053 nextpair = nvlist_next_nvpair(innvl, pair); 1054 1055 if (hashp == NULL) { 1056 error = SET_ERROR(EINVAL); 1057 break; 1058 } 1059 1060 *hashp = '\0'; 1061 error = zfs_secpolicy_write_perms(name, 1062 ZFS_DELEG_PERM_DESTROY, cr); 1063 *hashp = '#'; 1064 if (error == ENOENT) { 1065 /* 1066 * Ignore any filesystems that don't exist (we consider 1067 * their bookmarks "already destroyed"). Remove 1068 * the name from the nvl here in case the filesystem 1069 * is created between now and when we try to destroy 1070 * the bookmark (in which case we don't want to 1071 * destroy it since we haven't checked for permission). 1072 */ 1073 fnvlist_remove_nvpair(innvl, pair); 1074 error = 0; 1075 } 1076 if (error != 0) 1077 break; 1078 } 1079 1080 return (error); 1081 } 1082 1083 /* ARGSUSED */ 1084 static int 1085 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1086 { 1087 /* 1088 * Even root must have a proper TSD so that we know what pool 1089 * to log to. 1090 */ 1091 if (tsd_get(zfs_allow_log_key) == NULL) 1092 return (SET_ERROR(EPERM)); 1093 return (0); 1094 } 1095 1096 static int 1097 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1098 { 1099 char parentname[ZFS_MAX_DATASET_NAME_LEN]; 1100 int error; 1101 char *origin; 1102 1103 if ((error = zfs_get_parent(zc->zc_name, parentname, 1104 sizeof (parentname))) != 0) 1105 return (error); 1106 1107 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 && 1108 (error = zfs_secpolicy_write_perms(origin, 1109 ZFS_DELEG_PERM_CLONE, cr)) != 0) 1110 return (error); 1111 1112 if ((error = zfs_secpolicy_write_perms(parentname, 1113 ZFS_DELEG_PERM_CREATE, cr)) != 0) 1114 return (error); 1115 1116 return (zfs_secpolicy_write_perms(parentname, 1117 ZFS_DELEG_PERM_MOUNT, cr)); 1118 } 1119 1120 /* 1121 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires 1122 * SYS_CONFIG privilege, which is not available in a local zone. 1123 */ 1124 /* ARGSUSED */ 1125 static int 1126 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1127 { 1128 if (secpolicy_sys_config(cr, B_FALSE) != 0) 1129 return (SET_ERROR(EPERM)); 1130 1131 return (0); 1132 } 1133 1134 /* 1135 * Policy for object to name lookups. 1136 */ 1137 /* ARGSUSED */ 1138 static int 1139 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1140 { 1141 int error; 1142 1143 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0) 1144 return (0); 1145 1146 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr); 1147 return (error); 1148 } 1149 1150 /* 1151 * Policy for fault injection. Requires all privileges. 1152 */ 1153 /* ARGSUSED */ 1154 static int 1155 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1156 { 1157 return (secpolicy_zinject(cr)); 1158 } 1159 1160 /* ARGSUSED */ 1161 static int 1162 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1163 { 1164 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value); 1165 1166 if (prop == ZPROP_INVAL) { 1167 if (!zfs_prop_user(zc->zc_value)) 1168 return (SET_ERROR(EINVAL)); 1169 return (zfs_secpolicy_write_perms(zc->zc_name, 1170 ZFS_DELEG_PERM_USERPROP, cr)); 1171 } else { 1172 return (zfs_secpolicy_setprop(zc->zc_name, prop, 1173 NULL, cr)); 1174 } 1175 } 1176 1177 static int 1178 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1179 { 1180 int err = zfs_secpolicy_read(zc, innvl, cr); 1181 if (err) 1182 return (err); 1183 1184 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 1185 return (SET_ERROR(EINVAL)); 1186 1187 if (zc->zc_value[0] == 0) { 1188 /* 1189 * They are asking about a posix uid/gid. If it's 1190 * themself, allow it. 1191 */ 1192 if (zc->zc_objset_type == ZFS_PROP_USERUSED || 1193 zc->zc_objset_type == ZFS_PROP_USERQUOTA) { 1194 if (zc->zc_guid == crgetuid(cr)) 1195 return (0); 1196 } else { 1197 if (groupmember(zc->zc_guid, cr)) 1198 return (0); 1199 } 1200 } 1201 1202 return (zfs_secpolicy_write_perms(zc->zc_name, 1203 userquota_perms[zc->zc_objset_type], cr)); 1204 } 1205 1206 static int 1207 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1208 { 1209 int err = zfs_secpolicy_read(zc, innvl, cr); 1210 if (err) 1211 return (err); 1212 1213 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 1214 return (SET_ERROR(EINVAL)); 1215 1216 return (zfs_secpolicy_write_perms(zc->zc_name, 1217 userquota_perms[zc->zc_objset_type], cr)); 1218 } 1219 1220 /* ARGSUSED */ 1221 static int 1222 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1223 { 1224 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION, 1225 NULL, cr)); 1226 } 1227 1228 /* ARGSUSED */ 1229 static int 1230 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1231 { 1232 nvpair_t *pair; 1233 nvlist_t *holds; 1234 int error; 1235 1236 error = nvlist_lookup_nvlist(innvl, "holds", &holds); 1237 if (error != 0) 1238 return (SET_ERROR(EINVAL)); 1239 1240 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL; 1241 pair = nvlist_next_nvpair(holds, pair)) { 1242 char fsname[ZFS_MAX_DATASET_NAME_LEN]; 1243 error = dmu_fsname(nvpair_name(pair), fsname); 1244 if (error != 0) 1245 return (error); 1246 error = zfs_secpolicy_write_perms(fsname, 1247 ZFS_DELEG_PERM_HOLD, cr); 1248 if (error != 0) 1249 return (error); 1250 } 1251 return (0); 1252 } 1253 1254 /* ARGSUSED */ 1255 static int 1256 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1257 { 1258 nvpair_t *pair; 1259 int error; 1260 1261 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL; 1262 pair = nvlist_next_nvpair(innvl, pair)) { 1263 char fsname[ZFS_MAX_DATASET_NAME_LEN]; 1264 error = dmu_fsname(nvpair_name(pair), fsname); 1265 if (error != 0) 1266 return (error); 1267 error = zfs_secpolicy_write_perms(fsname, 1268 ZFS_DELEG_PERM_RELEASE, cr); 1269 if (error != 0) 1270 return (error); 1271 } 1272 return (0); 1273 } 1274 1275 /* 1276 * Policy for allowing temporary snapshots to be taken or released 1277 */ 1278 static int 1279 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1280 { 1281 /* 1282 * A temporary snapshot is the same as a snapshot, 1283 * hold, destroy and release all rolled into one. 1284 * Delegated diff alone is sufficient that we allow this. 1285 */ 1286 int error; 1287 1288 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 1289 ZFS_DELEG_PERM_DIFF, cr)) == 0) 1290 return (0); 1291 1292 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr); 1293 if (error == 0) 1294 error = zfs_secpolicy_hold(zc, innvl, cr); 1295 if (error == 0) 1296 error = zfs_secpolicy_release(zc, innvl, cr); 1297 if (error == 0) 1298 error = zfs_secpolicy_destroy(zc, innvl, cr); 1299 return (error); 1300 } 1301 1302 /* 1303 * Returns the nvlist as specified by the user in the zfs_cmd_t. 1304 */ 1305 static int 1306 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp) 1307 { 1308 char *packed; 1309 int error; 1310 nvlist_t *list = NULL; 1311 1312 /* 1313 * Read in and unpack the user-supplied nvlist. 1314 */ 1315 if (size == 0) 1316 return (SET_ERROR(EINVAL)); 1317 1318 packed = kmem_alloc(size, KM_SLEEP); 1319 1320 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size, 1321 iflag)) != 0) { 1322 kmem_free(packed, size); 1323 return (SET_ERROR(EFAULT)); 1324 } 1325 1326 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) { 1327 kmem_free(packed, size); 1328 return (error); 1329 } 1330 1331 kmem_free(packed, size); 1332 1333 *nvp = list; 1334 return (0); 1335 } 1336 1337 /* 1338 * Reduce the size of this nvlist until it can be serialized in 'max' bytes. 1339 * Entries will be removed from the end of the nvlist, and one int32 entry 1340 * named "N_MORE_ERRORS" will be added indicating how many entries were 1341 * removed. 1342 */ 1343 static int 1344 nvlist_smush(nvlist_t *errors, size_t max) 1345 { 1346 size_t size; 1347 1348 size = fnvlist_size(errors); 1349 1350 if (size > max) { 1351 nvpair_t *more_errors; 1352 int n = 0; 1353 1354 if (max < 1024) 1355 return (SET_ERROR(ENOMEM)); 1356 1357 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0); 1358 more_errors = nvlist_prev_nvpair(errors, NULL); 1359 1360 do { 1361 nvpair_t *pair = nvlist_prev_nvpair(errors, 1362 more_errors); 1363 fnvlist_remove_nvpair(errors, pair); 1364 n++; 1365 size = fnvlist_size(errors); 1366 } while (size > max); 1367 1368 fnvlist_remove_nvpair(errors, more_errors); 1369 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n); 1370 ASSERT3U(fnvlist_size(errors), <=, max); 1371 } 1372 1373 return (0); 1374 } 1375 1376 static int 1377 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl) 1378 { 1379 char *packed = NULL; 1380 int error = 0; 1381 size_t size; 1382 1383 size = fnvlist_size(nvl); 1384 1385 if (size > zc->zc_nvlist_dst_size) { 1386 error = SET_ERROR(ENOMEM); 1387 } else { 1388 packed = fnvlist_pack(nvl, &size); 1389 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst, 1390 size, zc->zc_iflags) != 0) 1391 error = SET_ERROR(EFAULT); 1392 fnvlist_pack_free(packed, size); 1393 } 1394 1395 zc->zc_nvlist_dst_size = size; 1396 zc->zc_nvlist_dst_filled = B_TRUE; 1397 return (error); 1398 } 1399 1400 int 1401 getzfsvfs_impl(objset_t *os, zfsvfs_t **zfvp) 1402 { 1403 int error = 0; 1404 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1405 return (SET_ERROR(EINVAL)); 1406 } 1407 1408 mutex_enter(&os->os_user_ptr_lock); 1409 *zfvp = dmu_objset_get_user(os); 1410 if (*zfvp) { 1411 VFS_HOLD((*zfvp)->z_vfs); 1412 } else { 1413 error = SET_ERROR(ESRCH); 1414 } 1415 mutex_exit(&os->os_user_ptr_lock); 1416 return (error); 1417 } 1418 1419 static int 1420 getzfsvfs(const char *dsname, zfsvfs_t **zfvp) 1421 { 1422 objset_t *os; 1423 int error; 1424 1425 error = dmu_objset_hold(dsname, FTAG, &os); 1426 if (error != 0) 1427 return (error); 1428 1429 error = getzfsvfs_impl(os, zfvp); 1430 dmu_objset_rele(os, FTAG); 1431 return (error); 1432 } 1433 1434 /* 1435 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which 1436 * case its z_vfs will be NULL, and it will be opened as the owner. 1437 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER, 1438 * which prevents all vnode ops from running. 1439 */ 1440 static int 1441 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer) 1442 { 1443 int error = 0; 1444 1445 if (getzfsvfs(name, zfvp) != 0) 1446 error = zfsvfs_create(name, zfvp); 1447 if (error == 0) { 1448 rrm_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER : 1449 RW_READER, tag); 1450 if ((*zfvp)->z_unmounted) { 1451 /* 1452 * XXX we could probably try again, since the unmounting 1453 * thread should be just about to disassociate the 1454 * objset from the zfsvfs. 1455 */ 1456 rrm_exit(&(*zfvp)->z_teardown_lock, tag); 1457 return (SET_ERROR(EBUSY)); 1458 } 1459 } 1460 return (error); 1461 } 1462 1463 static void 1464 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag) 1465 { 1466 rrm_exit(&zfsvfs->z_teardown_lock, tag); 1467 1468 if (zfsvfs->z_vfs) { 1469 VFS_RELE(zfsvfs->z_vfs); 1470 } else { 1471 dmu_objset_disown(zfsvfs->z_os, zfsvfs); 1472 zfsvfs_free(zfsvfs); 1473 } 1474 } 1475 1476 static int 1477 zfs_ioc_pool_create(zfs_cmd_t *zc) 1478 { 1479 int error; 1480 nvlist_t *config, *props = NULL; 1481 nvlist_t *rootprops = NULL; 1482 nvlist_t *zplprops = NULL; 1483 1484 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1485 zc->zc_iflags, &config)) 1486 return (error); 1487 1488 if (zc->zc_nvlist_src_size != 0 && (error = 1489 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1490 zc->zc_iflags, &props))) { 1491 nvlist_free(config); 1492 return (error); 1493 } 1494 1495 if (props) { 1496 nvlist_t *nvl = NULL; 1497 uint64_t version = SPA_VERSION; 1498 1499 (void) nvlist_lookup_uint64(props, 1500 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version); 1501 if (!SPA_VERSION_IS_SUPPORTED(version)) { 1502 error = SET_ERROR(EINVAL); 1503 goto pool_props_bad; 1504 } 1505 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl); 1506 if (nvl) { 1507 error = nvlist_dup(nvl, &rootprops, KM_SLEEP); 1508 if (error != 0) { 1509 nvlist_free(config); 1510 nvlist_free(props); 1511 return (error); 1512 } 1513 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS); 1514 } 1515 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 1516 error = zfs_fill_zplprops_root(version, rootprops, 1517 zplprops, NULL); 1518 if (error != 0) 1519 goto pool_props_bad; 1520 } 1521 1522 error = spa_create(zc->zc_name, config, props, zplprops); 1523 1524 /* 1525 * Set the remaining root properties 1526 */ 1527 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name, 1528 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0) 1529 (void) spa_destroy(zc->zc_name); 1530 1531 pool_props_bad: 1532 nvlist_free(rootprops); 1533 nvlist_free(zplprops); 1534 nvlist_free(config); 1535 nvlist_free(props); 1536 1537 return (error); 1538 } 1539 1540 static int 1541 zfs_ioc_pool_destroy(zfs_cmd_t *zc) 1542 { 1543 int error; 1544 zfs_log_history(zc); 1545 error = spa_destroy(zc->zc_name); 1546 if (error == 0) 1547 zvol_remove_minors(zc->zc_name); 1548 return (error); 1549 } 1550 1551 static int 1552 zfs_ioc_pool_import(zfs_cmd_t *zc) 1553 { 1554 nvlist_t *config, *props = NULL; 1555 uint64_t guid; 1556 int error; 1557 1558 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1559 zc->zc_iflags, &config)) != 0) 1560 return (error); 1561 1562 if (zc->zc_nvlist_src_size != 0 && (error = 1563 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1564 zc->zc_iflags, &props))) { 1565 nvlist_free(config); 1566 return (error); 1567 } 1568 1569 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 || 1570 guid != zc->zc_guid) 1571 error = SET_ERROR(EINVAL); 1572 else 1573 error = spa_import(zc->zc_name, config, props, zc->zc_cookie); 1574 1575 if (zc->zc_nvlist_dst != 0) { 1576 int err; 1577 1578 if ((err = put_nvlist(zc, config)) != 0) 1579 error = err; 1580 } 1581 1582 nvlist_free(config); 1583 1584 nvlist_free(props); 1585 1586 return (error); 1587 } 1588 1589 static int 1590 zfs_ioc_pool_export(zfs_cmd_t *zc) 1591 { 1592 int error; 1593 boolean_t force = (boolean_t)zc->zc_cookie; 1594 boolean_t hardforce = (boolean_t)zc->zc_guid; 1595 1596 zfs_log_history(zc); 1597 error = spa_export(zc->zc_name, NULL, force, hardforce); 1598 if (error == 0) 1599 zvol_remove_minors(zc->zc_name); 1600 return (error); 1601 } 1602 1603 static int 1604 zfs_ioc_pool_configs(zfs_cmd_t *zc) 1605 { 1606 nvlist_t *configs; 1607 int error; 1608 1609 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL) 1610 return (SET_ERROR(EEXIST)); 1611 1612 error = put_nvlist(zc, configs); 1613 1614 nvlist_free(configs); 1615 1616 return (error); 1617 } 1618 1619 /* 1620 * inputs: 1621 * zc_name name of the pool 1622 * 1623 * outputs: 1624 * zc_cookie real errno 1625 * zc_nvlist_dst config nvlist 1626 * zc_nvlist_dst_size size of config nvlist 1627 */ 1628 static int 1629 zfs_ioc_pool_stats(zfs_cmd_t *zc) 1630 { 1631 nvlist_t *config; 1632 int error; 1633 int ret = 0; 1634 1635 error = spa_get_stats(zc->zc_name, &config, zc->zc_value, 1636 sizeof (zc->zc_value)); 1637 1638 if (config != NULL) { 1639 ret = put_nvlist(zc, config); 1640 nvlist_free(config); 1641 1642 /* 1643 * The config may be present even if 'error' is non-zero. 1644 * In this case we return success, and preserve the real errno 1645 * in 'zc_cookie'. 1646 */ 1647 zc->zc_cookie = error; 1648 } else { 1649 ret = error; 1650 } 1651 1652 return (ret); 1653 } 1654 1655 /* 1656 * Try to import the given pool, returning pool stats as appropriate so that 1657 * user land knows which devices are available and overall pool health. 1658 */ 1659 static int 1660 zfs_ioc_pool_tryimport(zfs_cmd_t *zc) 1661 { 1662 nvlist_t *tryconfig, *config; 1663 int error; 1664 1665 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1666 zc->zc_iflags, &tryconfig)) != 0) 1667 return (error); 1668 1669 config = spa_tryimport(tryconfig); 1670 1671 nvlist_free(tryconfig); 1672 1673 if (config == NULL) 1674 return (SET_ERROR(EINVAL)); 1675 1676 error = put_nvlist(zc, config); 1677 nvlist_free(config); 1678 1679 return (error); 1680 } 1681 1682 /* 1683 * inputs: 1684 * zc_name name of the pool 1685 * zc_cookie scan func (pool_scan_func_t) 1686 */ 1687 static int 1688 zfs_ioc_pool_scan(zfs_cmd_t *zc) 1689 { 1690 spa_t *spa; 1691 int error; 1692 1693 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1694 return (error); 1695 1696 if (zc->zc_cookie == POOL_SCAN_NONE) 1697 error = spa_scan_stop(spa); 1698 else 1699 error = spa_scan(spa, zc->zc_cookie); 1700 1701 spa_close(spa, FTAG); 1702 1703 return (error); 1704 } 1705 1706 static int 1707 zfs_ioc_pool_freeze(zfs_cmd_t *zc) 1708 { 1709 spa_t *spa; 1710 int error; 1711 1712 error = spa_open(zc->zc_name, &spa, FTAG); 1713 if (error == 0) { 1714 spa_freeze(spa); 1715 spa_close(spa, FTAG); 1716 } 1717 return (error); 1718 } 1719 1720 static int 1721 zfs_ioc_pool_upgrade(zfs_cmd_t *zc) 1722 { 1723 spa_t *spa; 1724 int error; 1725 1726 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1727 return (error); 1728 1729 if (zc->zc_cookie < spa_version(spa) || 1730 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) { 1731 spa_close(spa, FTAG); 1732 return (SET_ERROR(EINVAL)); 1733 } 1734 1735 spa_upgrade(spa, zc->zc_cookie); 1736 spa_close(spa, FTAG); 1737 1738 return (error); 1739 } 1740 1741 static int 1742 zfs_ioc_pool_get_history(zfs_cmd_t *zc) 1743 { 1744 spa_t *spa; 1745 char *hist_buf; 1746 uint64_t size; 1747 int error; 1748 1749 if ((size = zc->zc_history_len) == 0) 1750 return (SET_ERROR(EINVAL)); 1751 1752 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1753 return (error); 1754 1755 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) { 1756 spa_close(spa, FTAG); 1757 return (SET_ERROR(ENOTSUP)); 1758 } 1759 1760 hist_buf = kmem_alloc(size, KM_SLEEP); 1761 if ((error = spa_history_get(spa, &zc->zc_history_offset, 1762 &zc->zc_history_len, hist_buf)) == 0) { 1763 error = ddi_copyout(hist_buf, 1764 (void *)(uintptr_t)zc->zc_history, 1765 zc->zc_history_len, zc->zc_iflags); 1766 } 1767 1768 spa_close(spa, FTAG); 1769 kmem_free(hist_buf, size); 1770 return (error); 1771 } 1772 1773 static int 1774 zfs_ioc_pool_reguid(zfs_cmd_t *zc) 1775 { 1776 spa_t *spa; 1777 int error; 1778 1779 error = spa_open(zc->zc_name, &spa, FTAG); 1780 if (error == 0) { 1781 error = spa_change_guid(spa); 1782 spa_close(spa, FTAG); 1783 } 1784 return (error); 1785 } 1786 1787 static int 1788 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc) 1789 { 1790 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value)); 1791 } 1792 1793 /* 1794 * inputs: 1795 * zc_name name of filesystem 1796 * zc_obj object to find 1797 * 1798 * outputs: 1799 * zc_value name of object 1800 */ 1801 static int 1802 zfs_ioc_obj_to_path(zfs_cmd_t *zc) 1803 { 1804 objset_t *os; 1805 int error; 1806 1807 /* XXX reading from objset not owned */ 1808 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0) 1809 return (error); 1810 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1811 dmu_objset_rele(os, FTAG); 1812 return (SET_ERROR(EINVAL)); 1813 } 1814 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value, 1815 sizeof (zc->zc_value)); 1816 dmu_objset_rele(os, FTAG); 1817 1818 return (error); 1819 } 1820 1821 /* 1822 * inputs: 1823 * zc_name name of filesystem 1824 * zc_obj object to find 1825 * 1826 * outputs: 1827 * zc_stat stats on object 1828 * zc_value path to object 1829 */ 1830 static int 1831 zfs_ioc_obj_to_stats(zfs_cmd_t *zc) 1832 { 1833 objset_t *os; 1834 int error; 1835 1836 /* XXX reading from objset not owned */ 1837 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0) 1838 return (error); 1839 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1840 dmu_objset_rele(os, FTAG); 1841 return (SET_ERROR(EINVAL)); 1842 } 1843 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value, 1844 sizeof (zc->zc_value)); 1845 dmu_objset_rele(os, FTAG); 1846 1847 return (error); 1848 } 1849 1850 static int 1851 zfs_ioc_vdev_add(zfs_cmd_t *zc) 1852 { 1853 spa_t *spa; 1854 int error; 1855 nvlist_t *config, **l2cache, **spares; 1856 uint_t nl2cache = 0, nspares = 0; 1857 1858 error = spa_open(zc->zc_name, &spa, FTAG); 1859 if (error != 0) 1860 return (error); 1861 1862 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1863 zc->zc_iflags, &config); 1864 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE, 1865 &l2cache, &nl2cache); 1866 1867 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES, 1868 &spares, &nspares); 1869 1870 /* 1871 * A root pool with concatenated devices is not supported. 1872 * Thus, can not add a device to a root pool. 1873 * 1874 * Intent log device can not be added to a rootpool because 1875 * during mountroot, zil is replayed, a seperated log device 1876 * can not be accessed during the mountroot time. 1877 * 1878 * l2cache and spare devices are ok to be added to a rootpool. 1879 */ 1880 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) { 1881 nvlist_free(config); 1882 spa_close(spa, FTAG); 1883 return (SET_ERROR(EDOM)); 1884 } 1885 1886 if (error == 0) { 1887 error = spa_vdev_add(spa, config); 1888 nvlist_free(config); 1889 } 1890 spa_close(spa, FTAG); 1891 return (error); 1892 } 1893 1894 /* 1895 * inputs: 1896 * zc_name name of the pool 1897 * zc_nvlist_conf nvlist of devices to remove 1898 * zc_cookie to stop the remove? 1899 */ 1900 static int 1901 zfs_ioc_vdev_remove(zfs_cmd_t *zc) 1902 { 1903 spa_t *spa; 1904 int error; 1905 1906 error = spa_open(zc->zc_name, &spa, FTAG); 1907 if (error != 0) 1908 return (error); 1909 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE); 1910 spa_close(spa, FTAG); 1911 return (error); 1912 } 1913 1914 static int 1915 zfs_ioc_vdev_set_state(zfs_cmd_t *zc) 1916 { 1917 spa_t *spa; 1918 int error; 1919 vdev_state_t newstate = VDEV_STATE_UNKNOWN; 1920 1921 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1922 return (error); 1923 switch (zc->zc_cookie) { 1924 case VDEV_STATE_ONLINE: 1925 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate); 1926 break; 1927 1928 case VDEV_STATE_OFFLINE: 1929 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj); 1930 break; 1931 1932 case VDEV_STATE_FAULTED: 1933 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED && 1934 zc->zc_obj != VDEV_AUX_EXTERNAL) 1935 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED; 1936 1937 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj); 1938 break; 1939 1940 case VDEV_STATE_DEGRADED: 1941 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED && 1942 zc->zc_obj != VDEV_AUX_EXTERNAL) 1943 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED; 1944 1945 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj); 1946 break; 1947 1948 default: 1949 error = SET_ERROR(EINVAL); 1950 } 1951 zc->zc_cookie = newstate; 1952 spa_close(spa, FTAG); 1953 return (error); 1954 } 1955 1956 static int 1957 zfs_ioc_vdev_attach(zfs_cmd_t *zc) 1958 { 1959 spa_t *spa; 1960 int replacing = zc->zc_cookie; 1961 nvlist_t *config; 1962 int error; 1963 1964 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1965 return (error); 1966 1967 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1968 zc->zc_iflags, &config)) == 0) { 1969 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing); 1970 nvlist_free(config); 1971 } 1972 1973 spa_close(spa, FTAG); 1974 return (error); 1975 } 1976 1977 static int 1978 zfs_ioc_vdev_detach(zfs_cmd_t *zc) 1979 { 1980 spa_t *spa; 1981 int error; 1982 1983 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1984 return (error); 1985 1986 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE); 1987 1988 spa_close(spa, FTAG); 1989 return (error); 1990 } 1991 1992 static int 1993 zfs_ioc_vdev_split(zfs_cmd_t *zc) 1994 { 1995 spa_t *spa; 1996 nvlist_t *config, *props = NULL; 1997 int error; 1998 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT); 1999 2000 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 2001 return (error); 2002 2003 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 2004 zc->zc_iflags, &config)) { 2005 spa_close(spa, FTAG); 2006 return (error); 2007 } 2008 2009 if (zc->zc_nvlist_src_size != 0 && (error = 2010 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2011 zc->zc_iflags, &props))) { 2012 spa_close(spa, FTAG); 2013 nvlist_free(config); 2014 return (error); 2015 } 2016 2017 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp); 2018 2019 spa_close(spa, FTAG); 2020 2021 nvlist_free(config); 2022 nvlist_free(props); 2023 2024 return (error); 2025 } 2026 2027 static int 2028 zfs_ioc_vdev_setpath(zfs_cmd_t *zc) 2029 { 2030 spa_t *spa; 2031 char *path = zc->zc_value; 2032 uint64_t guid = zc->zc_guid; 2033 int error; 2034 2035 error = spa_open(zc->zc_name, &spa, FTAG); 2036 if (error != 0) 2037 return (error); 2038 2039 error = spa_vdev_setpath(spa, guid, path); 2040 spa_close(spa, FTAG); 2041 return (error); 2042 } 2043 2044 static int 2045 zfs_ioc_vdev_setfru(zfs_cmd_t *zc) 2046 { 2047 spa_t *spa; 2048 char *fru = zc->zc_value; 2049 uint64_t guid = zc->zc_guid; 2050 int error; 2051 2052 error = spa_open(zc->zc_name, &spa, FTAG); 2053 if (error != 0) 2054 return (error); 2055 2056 error = spa_vdev_setfru(spa, guid, fru); 2057 spa_close(spa, FTAG); 2058 return (error); 2059 } 2060 2061 static int 2062 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os) 2063 { 2064 int error = 0; 2065 nvlist_t *nv; 2066 2067 dmu_objset_fast_stat(os, &zc->zc_objset_stats); 2068 2069 if (zc->zc_nvlist_dst != 0 && 2070 (error = dsl_prop_get_all(os, &nv)) == 0) { 2071 dmu_objset_stats(os, nv); 2072 /* 2073 * NB: zvol_get_stats() will read the objset contents, 2074 * which we aren't supposed to do with a 2075 * DS_MODE_USER hold, because it could be 2076 * inconsistent. So this is a bit of a workaround... 2077 * XXX reading with out owning 2078 */ 2079 if (!zc->zc_objset_stats.dds_inconsistent && 2080 dmu_objset_type(os) == DMU_OST_ZVOL) { 2081 error = zvol_get_stats(os, nv); 2082 if (error == EIO) 2083 return (error); 2084 VERIFY0(error); 2085 } 2086 error = put_nvlist(zc, nv); 2087 nvlist_free(nv); 2088 } 2089 2090 return (error); 2091 } 2092 2093 /* 2094 * inputs: 2095 * zc_name name of filesystem 2096 * zc_nvlist_dst_size size of buffer for property nvlist 2097 * 2098 * outputs: 2099 * zc_objset_stats stats 2100 * zc_nvlist_dst property nvlist 2101 * zc_nvlist_dst_size size of property nvlist 2102 */ 2103 static int 2104 zfs_ioc_objset_stats(zfs_cmd_t *zc) 2105 { 2106 objset_t *os; 2107 int error; 2108 2109 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 2110 if (error == 0) { 2111 error = zfs_ioc_objset_stats_impl(zc, os); 2112 dmu_objset_rele(os, FTAG); 2113 } 2114 2115 return (error); 2116 } 2117 2118 /* 2119 * inputs: 2120 * zc_name name of filesystem 2121 * zc_nvlist_dst_size size of buffer for property nvlist 2122 * 2123 * outputs: 2124 * zc_nvlist_dst received property nvlist 2125 * zc_nvlist_dst_size size of received property nvlist 2126 * 2127 * Gets received properties (distinct from local properties on or after 2128 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from 2129 * local property values. 2130 */ 2131 static int 2132 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc) 2133 { 2134 int error = 0; 2135 nvlist_t *nv; 2136 2137 /* 2138 * Without this check, we would return local property values if the 2139 * caller has not already received properties on or after 2140 * SPA_VERSION_RECVD_PROPS. 2141 */ 2142 if (!dsl_prop_get_hasrecvd(zc->zc_name)) 2143 return (SET_ERROR(ENOTSUP)); 2144 2145 if (zc->zc_nvlist_dst != 0 && 2146 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) { 2147 error = put_nvlist(zc, nv); 2148 nvlist_free(nv); 2149 } 2150 2151 return (error); 2152 } 2153 2154 static int 2155 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop) 2156 { 2157 uint64_t value; 2158 int error; 2159 2160 /* 2161 * zfs_get_zplprop() will either find a value or give us 2162 * the default value (if there is one). 2163 */ 2164 if ((error = zfs_get_zplprop(os, prop, &value)) != 0) 2165 return (error); 2166 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0); 2167 return (0); 2168 } 2169 2170 /* 2171 * inputs: 2172 * zc_name name of filesystem 2173 * zc_nvlist_dst_size size of buffer for zpl property nvlist 2174 * 2175 * outputs: 2176 * zc_nvlist_dst zpl property nvlist 2177 * zc_nvlist_dst_size size of zpl property nvlist 2178 */ 2179 static int 2180 zfs_ioc_objset_zplprops(zfs_cmd_t *zc) 2181 { 2182 objset_t *os; 2183 int err; 2184 2185 /* XXX reading without owning */ 2186 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os)) 2187 return (err); 2188 2189 dmu_objset_fast_stat(os, &zc->zc_objset_stats); 2190 2191 /* 2192 * NB: nvl_add_zplprop() will read the objset contents, 2193 * which we aren't supposed to do with a DS_MODE_USER 2194 * hold, because it could be inconsistent. 2195 */ 2196 if (zc->zc_nvlist_dst != NULL && 2197 !zc->zc_objset_stats.dds_inconsistent && 2198 dmu_objset_type(os) == DMU_OST_ZFS) { 2199 nvlist_t *nv; 2200 2201 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2202 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 && 2203 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 && 2204 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 && 2205 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0) 2206 err = put_nvlist(zc, nv); 2207 nvlist_free(nv); 2208 } else { 2209 err = SET_ERROR(ENOENT); 2210 } 2211 dmu_objset_rele(os, FTAG); 2212 return (err); 2213 } 2214 2215 static boolean_t 2216 dataset_name_hidden(const char *name) 2217 { 2218 /* 2219 * Skip over datasets that are not visible in this zone, 2220 * internal datasets (which have a $ in their name), and 2221 * temporary datasets (which have a % in their name). 2222 */ 2223 if (strchr(name, '$') != NULL) 2224 return (B_TRUE); 2225 if (strchr(name, '%') != NULL) 2226 return (B_TRUE); 2227 if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL)) 2228 return (B_TRUE); 2229 return (B_FALSE); 2230 } 2231 2232 /* 2233 * inputs: 2234 * zc_name name of filesystem 2235 * zc_cookie zap cursor 2236 * zc_nvlist_dst_size size of buffer for property nvlist 2237 * 2238 * outputs: 2239 * zc_name name of next filesystem 2240 * zc_cookie zap cursor 2241 * zc_objset_stats stats 2242 * zc_nvlist_dst property nvlist 2243 * zc_nvlist_dst_size size of property nvlist 2244 */ 2245 static int 2246 zfs_ioc_dataset_list_next(zfs_cmd_t *zc) 2247 { 2248 objset_t *os; 2249 int error; 2250 char *p; 2251 size_t orig_len = strlen(zc->zc_name); 2252 2253 top: 2254 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) { 2255 if (error == ENOENT) 2256 error = SET_ERROR(ESRCH); 2257 return (error); 2258 } 2259 2260 p = strrchr(zc->zc_name, '/'); 2261 if (p == NULL || p[1] != '\0') 2262 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name)); 2263 p = zc->zc_name + strlen(zc->zc_name); 2264 2265 do { 2266 error = dmu_dir_list_next(os, 2267 sizeof (zc->zc_name) - (p - zc->zc_name), p, 2268 NULL, &zc->zc_cookie); 2269 if (error == ENOENT) 2270 error = SET_ERROR(ESRCH); 2271 } while (error == 0 && dataset_name_hidden(zc->zc_name)); 2272 dmu_objset_rele(os, FTAG); 2273 2274 /* 2275 * If it's an internal dataset (ie. with a '$' in its name), 2276 * don't try to get stats for it, otherwise we'll return ENOENT. 2277 */ 2278 if (error == 0 && strchr(zc->zc_name, '$') == NULL) { 2279 error = zfs_ioc_objset_stats(zc); /* fill in the stats */ 2280 if (error == ENOENT) { 2281 /* We lost a race with destroy, get the next one. */ 2282 zc->zc_name[orig_len] = '\0'; 2283 goto top; 2284 } 2285 } 2286 return (error); 2287 } 2288 2289 /* 2290 * inputs: 2291 * zc_name name of filesystem 2292 * zc_cookie zap cursor 2293 * zc_nvlist_dst_size size of buffer for property nvlist 2294 * zc_simple when set, only name is requested 2295 * 2296 * outputs: 2297 * zc_name name of next snapshot 2298 * zc_objset_stats stats 2299 * zc_nvlist_dst property nvlist 2300 * zc_nvlist_dst_size size of property nvlist 2301 */ 2302 static int 2303 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc) 2304 { 2305 objset_t *os; 2306 int error; 2307 2308 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 2309 if (error != 0) { 2310 return (error == ENOENT ? ESRCH : error); 2311 } 2312 2313 /* 2314 * A dataset name of maximum length cannot have any snapshots, 2315 * so exit immediately. 2316 */ 2317 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= 2318 ZFS_MAX_DATASET_NAME_LEN) { 2319 dmu_objset_rele(os, FTAG); 2320 return (SET_ERROR(ESRCH)); 2321 } 2322 2323 error = dmu_snapshot_list_next(os, 2324 sizeof (zc->zc_name) - strlen(zc->zc_name), 2325 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie, 2326 NULL); 2327 2328 if (error == 0 && !zc->zc_simple) { 2329 dsl_dataset_t *ds; 2330 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool; 2331 2332 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds); 2333 if (error == 0) { 2334 objset_t *ossnap; 2335 2336 error = dmu_objset_from_ds(ds, &ossnap); 2337 if (error == 0) 2338 error = zfs_ioc_objset_stats_impl(zc, ossnap); 2339 dsl_dataset_rele(ds, FTAG); 2340 } 2341 } else if (error == ENOENT) { 2342 error = SET_ERROR(ESRCH); 2343 } 2344 2345 dmu_objset_rele(os, FTAG); 2346 /* if we failed, undo the @ that we tacked on to zc_name */ 2347 if (error != 0) 2348 *strchr(zc->zc_name, '@') = '\0'; 2349 return (error); 2350 } 2351 2352 static int 2353 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair) 2354 { 2355 const char *propname = nvpair_name(pair); 2356 uint64_t *valary; 2357 unsigned int vallen; 2358 const char *domain; 2359 char *dash; 2360 zfs_userquota_prop_t type; 2361 uint64_t rid; 2362 uint64_t quota; 2363 zfsvfs_t *zfsvfs; 2364 int err; 2365 2366 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2367 nvlist_t *attrs; 2368 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 2369 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2370 &pair) != 0) 2371 return (SET_ERROR(EINVAL)); 2372 } 2373 2374 /* 2375 * A correctly constructed propname is encoded as 2376 * userquota@<rid>-<domain>. 2377 */ 2378 if ((dash = strchr(propname, '-')) == NULL || 2379 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 || 2380 vallen != 3) 2381 return (SET_ERROR(EINVAL)); 2382 2383 domain = dash + 1; 2384 type = valary[0]; 2385 rid = valary[1]; 2386 quota = valary[2]; 2387 2388 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE); 2389 if (err == 0) { 2390 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota); 2391 zfsvfs_rele(zfsvfs, FTAG); 2392 } 2393 2394 return (err); 2395 } 2396 2397 /* 2398 * If the named property is one that has a special function to set its value, 2399 * return 0 on success and a positive error code on failure; otherwise if it is 2400 * not one of the special properties handled by this function, return -1. 2401 * 2402 * XXX: It would be better for callers of the property interface if we handled 2403 * these special cases in dsl_prop.c (in the dsl layer). 2404 */ 2405 static int 2406 zfs_prop_set_special(const char *dsname, zprop_source_t source, 2407 nvpair_t *pair) 2408 { 2409 const char *propname = nvpair_name(pair); 2410 zfs_prop_t prop = zfs_name_to_prop(propname); 2411 uint64_t intval; 2412 int err = -1; 2413 2414 if (prop == ZPROP_INVAL) { 2415 if (zfs_prop_userquota(propname)) 2416 return (zfs_prop_set_userquota(dsname, pair)); 2417 return (-1); 2418 } 2419 2420 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2421 nvlist_t *attrs; 2422 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 2423 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2424 &pair) == 0); 2425 } 2426 2427 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING) 2428 return (-1); 2429 2430 VERIFY(0 == nvpair_value_uint64(pair, &intval)); 2431 2432 switch (prop) { 2433 case ZFS_PROP_QUOTA: 2434 err = dsl_dir_set_quota(dsname, source, intval); 2435 break; 2436 case ZFS_PROP_REFQUOTA: 2437 err = dsl_dataset_set_refquota(dsname, source, intval); 2438 break; 2439 case ZFS_PROP_FILESYSTEM_LIMIT: 2440 case ZFS_PROP_SNAPSHOT_LIMIT: 2441 if (intval == UINT64_MAX) { 2442 /* clearing the limit, just do it */ 2443 err = 0; 2444 } else { 2445 err = dsl_dir_activate_fs_ss_limit(dsname); 2446 } 2447 /* 2448 * Set err to -1 to force the zfs_set_prop_nvlist code down the 2449 * default path to set the value in the nvlist. 2450 */ 2451 if (err == 0) 2452 err = -1; 2453 break; 2454 case ZFS_PROP_RESERVATION: 2455 err = dsl_dir_set_reservation(dsname, source, intval); 2456 break; 2457 case ZFS_PROP_REFRESERVATION: 2458 err = dsl_dataset_set_refreservation(dsname, source, intval); 2459 break; 2460 case ZFS_PROP_VOLSIZE: 2461 err = zvol_set_volsize(dsname, intval); 2462 break; 2463 case ZFS_PROP_VERSION: 2464 { 2465 zfsvfs_t *zfsvfs; 2466 2467 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0) 2468 break; 2469 2470 err = zfs_set_version(zfsvfs, intval); 2471 zfsvfs_rele(zfsvfs, FTAG); 2472 2473 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) { 2474 zfs_cmd_t *zc; 2475 2476 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP); 2477 (void) strcpy(zc->zc_name, dsname); 2478 (void) zfs_ioc_userspace_upgrade(zc); 2479 kmem_free(zc, sizeof (zfs_cmd_t)); 2480 } 2481 break; 2482 } 2483 default: 2484 err = -1; 2485 } 2486 2487 return (err); 2488 } 2489 2490 /* 2491 * This function is best effort. If it fails to set any of the given properties, 2492 * it continues to set as many as it can and returns the last error 2493 * encountered. If the caller provides a non-NULL errlist, it will be filled in 2494 * with the list of names of all the properties that failed along with the 2495 * corresponding error numbers. 2496 * 2497 * If every property is set successfully, zero is returned and errlist is not 2498 * modified. 2499 */ 2500 int 2501 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl, 2502 nvlist_t *errlist) 2503 { 2504 nvpair_t *pair; 2505 nvpair_t *propval; 2506 int rv = 0; 2507 uint64_t intval; 2508 char *strval; 2509 nvlist_t *genericnvl = fnvlist_alloc(); 2510 nvlist_t *retrynvl = fnvlist_alloc(); 2511 2512 retry: 2513 pair = NULL; 2514 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { 2515 const char *propname = nvpair_name(pair); 2516 zfs_prop_t prop = zfs_name_to_prop(propname); 2517 int err = 0; 2518 2519 /* decode the property value */ 2520 propval = pair; 2521 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2522 nvlist_t *attrs; 2523 attrs = fnvpair_value_nvlist(pair); 2524 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2525 &propval) != 0) 2526 err = SET_ERROR(EINVAL); 2527 } 2528 2529 /* Validate value type */ 2530 if (err == 0 && prop == ZPROP_INVAL) { 2531 if (zfs_prop_user(propname)) { 2532 if (nvpair_type(propval) != DATA_TYPE_STRING) 2533 err = SET_ERROR(EINVAL); 2534 } else if (zfs_prop_userquota(propname)) { 2535 if (nvpair_type(propval) != 2536 DATA_TYPE_UINT64_ARRAY) 2537 err = SET_ERROR(EINVAL); 2538 } else { 2539 err = SET_ERROR(EINVAL); 2540 } 2541 } else if (err == 0) { 2542 if (nvpair_type(propval) == DATA_TYPE_STRING) { 2543 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING) 2544 err = SET_ERROR(EINVAL); 2545 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) { 2546 const char *unused; 2547 2548 intval = fnvpair_value_uint64(propval); 2549 2550 switch (zfs_prop_get_type(prop)) { 2551 case PROP_TYPE_NUMBER: 2552 break; 2553 case PROP_TYPE_STRING: 2554 err = SET_ERROR(EINVAL); 2555 break; 2556 case PROP_TYPE_INDEX: 2557 if (zfs_prop_index_to_string(prop, 2558 intval, &unused) != 0) 2559 err = SET_ERROR(EINVAL); 2560 break; 2561 default: 2562 cmn_err(CE_PANIC, 2563 "unknown property type"); 2564 } 2565 } else { 2566 err = SET_ERROR(EINVAL); 2567 } 2568 } 2569 2570 /* Validate permissions */ 2571 if (err == 0) 2572 err = zfs_check_settable(dsname, pair, CRED()); 2573 2574 if (err == 0) { 2575 err = zfs_prop_set_special(dsname, source, pair); 2576 if (err == -1) { 2577 /* 2578 * For better performance we build up a list of 2579 * properties to set in a single transaction. 2580 */ 2581 err = nvlist_add_nvpair(genericnvl, pair); 2582 } else if (err != 0 && nvl != retrynvl) { 2583 /* 2584 * This may be a spurious error caused by 2585 * receiving quota and reservation out of order. 2586 * Try again in a second pass. 2587 */ 2588 err = nvlist_add_nvpair(retrynvl, pair); 2589 } 2590 } 2591 2592 if (err != 0) { 2593 if (errlist != NULL) 2594 fnvlist_add_int32(errlist, propname, err); 2595 rv = err; 2596 } 2597 } 2598 2599 if (nvl != retrynvl && !nvlist_empty(retrynvl)) { 2600 nvl = retrynvl; 2601 goto retry; 2602 } 2603 2604 if (!nvlist_empty(genericnvl) && 2605 dsl_props_set(dsname, source, genericnvl) != 0) { 2606 /* 2607 * If this fails, we still want to set as many properties as we 2608 * can, so try setting them individually. 2609 */ 2610 pair = NULL; 2611 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) { 2612 const char *propname = nvpair_name(pair); 2613 int err = 0; 2614 2615 propval = pair; 2616 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2617 nvlist_t *attrs; 2618 attrs = fnvpair_value_nvlist(pair); 2619 propval = fnvlist_lookup_nvpair(attrs, 2620 ZPROP_VALUE); 2621 } 2622 2623 if (nvpair_type(propval) == DATA_TYPE_STRING) { 2624 strval = fnvpair_value_string(propval); 2625 err = dsl_prop_set_string(dsname, propname, 2626 source, strval); 2627 } else { 2628 intval = fnvpair_value_uint64(propval); 2629 err = dsl_prop_set_int(dsname, propname, source, 2630 intval); 2631 } 2632 2633 if (err != 0) { 2634 if (errlist != NULL) { 2635 fnvlist_add_int32(errlist, propname, 2636 err); 2637 } 2638 rv = err; 2639 } 2640 } 2641 } 2642 nvlist_free(genericnvl); 2643 nvlist_free(retrynvl); 2644 2645 return (rv); 2646 } 2647 2648 /* 2649 * Check that all the properties are valid user properties. 2650 */ 2651 static int 2652 zfs_check_userprops(const char *fsname, nvlist_t *nvl) 2653 { 2654 nvpair_t *pair = NULL; 2655 int error = 0; 2656 2657 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { 2658 const char *propname = nvpair_name(pair); 2659 2660 if (!zfs_prop_user(propname) || 2661 nvpair_type(pair) != DATA_TYPE_STRING) 2662 return (SET_ERROR(EINVAL)); 2663 2664 if (error = zfs_secpolicy_write_perms(fsname, 2665 ZFS_DELEG_PERM_USERPROP, CRED())) 2666 return (error); 2667 2668 if (strlen(propname) >= ZAP_MAXNAMELEN) 2669 return (SET_ERROR(ENAMETOOLONG)); 2670 2671 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN) 2672 return (E2BIG); 2673 } 2674 return (0); 2675 } 2676 2677 static void 2678 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops) 2679 { 2680 nvpair_t *pair; 2681 2682 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2683 2684 pair = NULL; 2685 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) { 2686 if (nvlist_exists(skipped, nvpair_name(pair))) 2687 continue; 2688 2689 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0); 2690 } 2691 } 2692 2693 static int 2694 clear_received_props(const char *dsname, nvlist_t *props, 2695 nvlist_t *skipped) 2696 { 2697 int err = 0; 2698 nvlist_t *cleared_props = NULL; 2699 props_skip(props, skipped, &cleared_props); 2700 if (!nvlist_empty(cleared_props)) { 2701 /* 2702 * Acts on local properties until the dataset has received 2703 * properties at least once on or after SPA_VERSION_RECVD_PROPS. 2704 */ 2705 zprop_source_t flags = (ZPROP_SRC_NONE | 2706 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0)); 2707 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL); 2708 } 2709 nvlist_free(cleared_props); 2710 return (err); 2711 } 2712 2713 /* 2714 * inputs: 2715 * zc_name name of filesystem 2716 * zc_value name of property to set 2717 * zc_nvlist_src{_size} nvlist of properties to apply 2718 * zc_cookie received properties flag 2719 * 2720 * outputs: 2721 * zc_nvlist_dst{_size} error for each unapplied received property 2722 */ 2723 static int 2724 zfs_ioc_set_prop(zfs_cmd_t *zc) 2725 { 2726 nvlist_t *nvl; 2727 boolean_t received = zc->zc_cookie; 2728 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED : 2729 ZPROP_SRC_LOCAL); 2730 nvlist_t *errors; 2731 int error; 2732 2733 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2734 zc->zc_iflags, &nvl)) != 0) 2735 return (error); 2736 2737 if (received) { 2738 nvlist_t *origprops; 2739 2740 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) { 2741 (void) clear_received_props(zc->zc_name, 2742 origprops, nvl); 2743 nvlist_free(origprops); 2744 } 2745 2746 error = dsl_prop_set_hasrecvd(zc->zc_name); 2747 } 2748 2749 errors = fnvlist_alloc(); 2750 if (error == 0) 2751 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors); 2752 2753 if (zc->zc_nvlist_dst != NULL && errors != NULL) { 2754 (void) put_nvlist(zc, errors); 2755 } 2756 2757 nvlist_free(errors); 2758 nvlist_free(nvl); 2759 return (error); 2760 } 2761 2762 /* 2763 * inputs: 2764 * zc_name name of filesystem 2765 * zc_value name of property to inherit 2766 * zc_cookie revert to received value if TRUE 2767 * 2768 * outputs: none 2769 */ 2770 static int 2771 zfs_ioc_inherit_prop(zfs_cmd_t *zc) 2772 { 2773 const char *propname = zc->zc_value; 2774 zfs_prop_t prop = zfs_name_to_prop(propname); 2775 boolean_t received = zc->zc_cookie; 2776 zprop_source_t source = (received 2777 ? ZPROP_SRC_NONE /* revert to received value, if any */ 2778 : ZPROP_SRC_INHERITED); /* explicitly inherit */ 2779 2780 if (received) { 2781 nvlist_t *dummy; 2782 nvpair_t *pair; 2783 zprop_type_t type; 2784 int err; 2785 2786 /* 2787 * zfs_prop_set_special() expects properties in the form of an 2788 * nvpair with type info. 2789 */ 2790 if (prop == ZPROP_INVAL) { 2791 if (!zfs_prop_user(propname)) 2792 return (SET_ERROR(EINVAL)); 2793 2794 type = PROP_TYPE_STRING; 2795 } else if (prop == ZFS_PROP_VOLSIZE || 2796 prop == ZFS_PROP_VERSION) { 2797 return (SET_ERROR(EINVAL)); 2798 } else { 2799 type = zfs_prop_get_type(prop); 2800 } 2801 2802 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2803 2804 switch (type) { 2805 case PROP_TYPE_STRING: 2806 VERIFY(0 == nvlist_add_string(dummy, propname, "")); 2807 break; 2808 case PROP_TYPE_NUMBER: 2809 case PROP_TYPE_INDEX: 2810 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0)); 2811 break; 2812 default: 2813 nvlist_free(dummy); 2814 return (SET_ERROR(EINVAL)); 2815 } 2816 2817 pair = nvlist_next_nvpair(dummy, NULL); 2818 err = zfs_prop_set_special(zc->zc_name, source, pair); 2819 nvlist_free(dummy); 2820 if (err != -1) 2821 return (err); /* special property already handled */ 2822 } else { 2823 /* 2824 * Only check this in the non-received case. We want to allow 2825 * 'inherit -S' to revert non-inheritable properties like quota 2826 * and reservation to the received or default values even though 2827 * they are not considered inheritable. 2828 */ 2829 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop)) 2830 return (SET_ERROR(EINVAL)); 2831 } 2832 2833 /* property name has been validated by zfs_secpolicy_inherit_prop() */ 2834 return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source)); 2835 } 2836 2837 static int 2838 zfs_ioc_pool_set_props(zfs_cmd_t *zc) 2839 { 2840 nvlist_t *props; 2841 spa_t *spa; 2842 int error; 2843 nvpair_t *pair; 2844 2845 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2846 zc->zc_iflags, &props)) 2847 return (error); 2848 2849 /* 2850 * If the only property is the configfile, then just do a spa_lookup() 2851 * to handle the faulted case. 2852 */ 2853 pair = nvlist_next_nvpair(props, NULL); 2854 if (pair != NULL && strcmp(nvpair_name(pair), 2855 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 && 2856 nvlist_next_nvpair(props, pair) == NULL) { 2857 mutex_enter(&spa_namespace_lock); 2858 if ((spa = spa_lookup(zc->zc_name)) != NULL) { 2859 spa_configfile_set(spa, props, B_FALSE); 2860 spa_config_sync(spa, B_FALSE, B_TRUE); 2861 } 2862 mutex_exit(&spa_namespace_lock); 2863 if (spa != NULL) { 2864 nvlist_free(props); 2865 return (0); 2866 } 2867 } 2868 2869 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { 2870 nvlist_free(props); 2871 return (error); 2872 } 2873 2874 error = spa_prop_set(spa, props); 2875 2876 nvlist_free(props); 2877 spa_close(spa, FTAG); 2878 2879 return (error); 2880 } 2881 2882 static int 2883 zfs_ioc_pool_get_props(zfs_cmd_t *zc) 2884 { 2885 spa_t *spa; 2886 int error; 2887 nvlist_t *nvp = NULL; 2888 2889 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { 2890 /* 2891 * If the pool is faulted, there may be properties we can still 2892 * get (such as altroot and cachefile), so attempt to get them 2893 * anyway. 2894 */ 2895 mutex_enter(&spa_namespace_lock); 2896 if ((spa = spa_lookup(zc->zc_name)) != NULL) 2897 error = spa_prop_get(spa, &nvp); 2898 mutex_exit(&spa_namespace_lock); 2899 } else { 2900 error = spa_prop_get(spa, &nvp); 2901 spa_close(spa, FTAG); 2902 } 2903 2904 if (error == 0 && zc->zc_nvlist_dst != NULL) 2905 error = put_nvlist(zc, nvp); 2906 else 2907 error = SET_ERROR(EFAULT); 2908 2909 nvlist_free(nvp); 2910 return (error); 2911 } 2912 2913 /* 2914 * inputs: 2915 * zc_name name of filesystem 2916 * zc_nvlist_src{_size} nvlist of delegated permissions 2917 * zc_perm_action allow/unallow flag 2918 * 2919 * outputs: none 2920 */ 2921 static int 2922 zfs_ioc_set_fsacl(zfs_cmd_t *zc) 2923 { 2924 int error; 2925 nvlist_t *fsaclnv = NULL; 2926 2927 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2928 zc->zc_iflags, &fsaclnv)) != 0) 2929 return (error); 2930 2931 /* 2932 * Verify nvlist is constructed correctly 2933 */ 2934 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) { 2935 nvlist_free(fsaclnv); 2936 return (SET_ERROR(EINVAL)); 2937 } 2938 2939 /* 2940 * If we don't have PRIV_SYS_MOUNT, then validate 2941 * that user is allowed to hand out each permission in 2942 * the nvlist(s) 2943 */ 2944 2945 error = secpolicy_zfs(CRED()); 2946 if (error != 0) { 2947 if (zc->zc_perm_action == B_FALSE) { 2948 error = dsl_deleg_can_allow(zc->zc_name, 2949 fsaclnv, CRED()); 2950 } else { 2951 error = dsl_deleg_can_unallow(zc->zc_name, 2952 fsaclnv, CRED()); 2953 } 2954 } 2955 2956 if (error == 0) 2957 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action); 2958 2959 nvlist_free(fsaclnv); 2960 return (error); 2961 } 2962 2963 /* 2964 * inputs: 2965 * zc_name name of filesystem 2966 * 2967 * outputs: 2968 * zc_nvlist_src{_size} nvlist of delegated permissions 2969 */ 2970 static int 2971 zfs_ioc_get_fsacl(zfs_cmd_t *zc) 2972 { 2973 nvlist_t *nvp; 2974 int error; 2975 2976 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) { 2977 error = put_nvlist(zc, nvp); 2978 nvlist_free(nvp); 2979 } 2980 2981 return (error); 2982 } 2983 2984 /* 2985 * Search the vfs list for a specified resource. Returns a pointer to it 2986 * or NULL if no suitable entry is found. The caller of this routine 2987 * is responsible for releasing the returned vfs pointer. 2988 */ 2989 static vfs_t * 2990 zfs_get_vfs(const char *resource) 2991 { 2992 struct vfs *vfsp; 2993 struct vfs *vfs_found = NULL; 2994 2995 vfs_list_read_lock(); 2996 vfsp = rootvfs; 2997 do { 2998 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) { 2999 VFS_HOLD(vfsp); 3000 vfs_found = vfsp; 3001 break; 3002 } 3003 vfsp = vfsp->vfs_next; 3004 } while (vfsp != rootvfs); 3005 vfs_list_unlock(); 3006 return (vfs_found); 3007 } 3008 3009 /* ARGSUSED */ 3010 static void 3011 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) 3012 { 3013 zfs_creat_t *zct = arg; 3014 3015 zfs_create_fs(os, cr, zct->zct_zplprops, tx); 3016 } 3017 3018 #define ZFS_PROP_UNDEFINED ((uint64_t)-1) 3019 3020 /* 3021 * inputs: 3022 * os parent objset pointer (NULL if root fs) 3023 * fuids_ok fuids allowed in this version of the spa? 3024 * sa_ok SAs allowed in this version of the spa? 3025 * createprops list of properties requested by creator 3026 * 3027 * outputs: 3028 * zplprops values for the zplprops we attach to the master node object 3029 * is_ci true if requested file system will be purely case-insensitive 3030 * 3031 * Determine the settings for utf8only, normalization and 3032 * casesensitivity. Specific values may have been requested by the 3033 * creator and/or we can inherit values from the parent dataset. If 3034 * the file system is of too early a vintage, a creator can not 3035 * request settings for these properties, even if the requested 3036 * setting is the default value. We don't actually want to create dsl 3037 * properties for these, so remove them from the source nvlist after 3038 * processing. 3039 */ 3040 static int 3041 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver, 3042 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops, 3043 nvlist_t *zplprops, boolean_t *is_ci) 3044 { 3045 uint64_t sense = ZFS_PROP_UNDEFINED; 3046 uint64_t norm = ZFS_PROP_UNDEFINED; 3047 uint64_t u8 = ZFS_PROP_UNDEFINED; 3048 3049 ASSERT(zplprops != NULL); 3050 3051 if (os != NULL && os->os_phys->os_type != DMU_OST_ZFS) 3052 return (SET_ERROR(EINVAL)); 3053 3054 /* 3055 * Pull out creator prop choices, if any. 3056 */ 3057 if (createprops) { 3058 (void) nvlist_lookup_uint64(createprops, 3059 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver); 3060 (void) nvlist_lookup_uint64(createprops, 3061 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm); 3062 (void) nvlist_remove_all(createprops, 3063 zfs_prop_to_name(ZFS_PROP_NORMALIZE)); 3064 (void) nvlist_lookup_uint64(createprops, 3065 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8); 3066 (void) nvlist_remove_all(createprops, 3067 zfs_prop_to_name(ZFS_PROP_UTF8ONLY)); 3068 (void) nvlist_lookup_uint64(createprops, 3069 zfs_prop_to_name(ZFS_PROP_CASE), &sense); 3070 (void) nvlist_remove_all(createprops, 3071 zfs_prop_to_name(ZFS_PROP_CASE)); 3072 } 3073 3074 /* 3075 * If the zpl version requested is whacky or the file system 3076 * or pool is version is too "young" to support normalization 3077 * and the creator tried to set a value for one of the props, 3078 * error out. 3079 */ 3080 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) || 3081 (zplver >= ZPL_VERSION_FUID && !fuids_ok) || 3082 (zplver >= ZPL_VERSION_SA && !sa_ok) || 3083 (zplver < ZPL_VERSION_NORMALIZATION && 3084 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED || 3085 sense != ZFS_PROP_UNDEFINED))) 3086 return (SET_ERROR(ENOTSUP)); 3087 3088 /* 3089 * Put the version in the zplprops 3090 */ 3091 VERIFY(nvlist_add_uint64(zplprops, 3092 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0); 3093 3094 if (norm == ZFS_PROP_UNDEFINED) 3095 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0); 3096 VERIFY(nvlist_add_uint64(zplprops, 3097 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0); 3098 3099 /* 3100 * If we're normalizing, names must always be valid UTF-8 strings. 3101 */ 3102 if (norm) 3103 u8 = 1; 3104 if (u8 == ZFS_PROP_UNDEFINED) 3105 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0); 3106 VERIFY(nvlist_add_uint64(zplprops, 3107 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0); 3108 3109 if (sense == ZFS_PROP_UNDEFINED) 3110 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0); 3111 VERIFY(nvlist_add_uint64(zplprops, 3112 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0); 3113 3114 if (is_ci) 3115 *is_ci = (sense == ZFS_CASE_INSENSITIVE); 3116 3117 return (0); 3118 } 3119 3120 static int 3121 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops, 3122 nvlist_t *zplprops, boolean_t *is_ci) 3123 { 3124 boolean_t fuids_ok, sa_ok; 3125 uint64_t zplver = ZPL_VERSION; 3126 objset_t *os = NULL; 3127 char parentname[ZFS_MAX_DATASET_NAME_LEN]; 3128 char *cp; 3129 spa_t *spa; 3130 uint64_t spa_vers; 3131 int error; 3132 3133 (void) strlcpy(parentname, dataset, sizeof (parentname)); 3134 cp = strrchr(parentname, '/'); 3135 ASSERT(cp != NULL); 3136 cp[0] = '\0'; 3137 3138 if ((error = spa_open(dataset, &spa, FTAG)) != 0) 3139 return (error); 3140 3141 spa_vers = spa_version(spa); 3142 spa_close(spa, FTAG); 3143 3144 zplver = zfs_zpl_version_map(spa_vers); 3145 fuids_ok = (zplver >= ZPL_VERSION_FUID); 3146 sa_ok = (zplver >= ZPL_VERSION_SA); 3147 3148 /* 3149 * Open parent object set so we can inherit zplprop values. 3150 */ 3151 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0) 3152 return (error); 3153 3154 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops, 3155 zplprops, is_ci); 3156 dmu_objset_rele(os, FTAG); 3157 return (error); 3158 } 3159 3160 static int 3161 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops, 3162 nvlist_t *zplprops, boolean_t *is_ci) 3163 { 3164 boolean_t fuids_ok; 3165 boolean_t sa_ok; 3166 uint64_t zplver = ZPL_VERSION; 3167 int error; 3168 3169 zplver = zfs_zpl_version_map(spa_vers); 3170 fuids_ok = (zplver >= ZPL_VERSION_FUID); 3171 sa_ok = (zplver >= ZPL_VERSION_SA); 3172 3173 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok, 3174 createprops, zplprops, is_ci); 3175 return (error); 3176 } 3177 3178 /* 3179 * innvl: { 3180 * "type" -> dmu_objset_type_t (int32) 3181 * (optional) "props" -> { prop -> value } 3182 * } 3183 * 3184 * outnvl: propname -> error code (int32) 3185 */ 3186 static int 3187 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3188 { 3189 int error = 0; 3190 zfs_creat_t zct = { 0 }; 3191 nvlist_t *nvprops = NULL; 3192 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx); 3193 int32_t type32; 3194 dmu_objset_type_t type; 3195 boolean_t is_insensitive = B_FALSE; 3196 3197 if (nvlist_lookup_int32(innvl, "type", &type32) != 0) 3198 return (SET_ERROR(EINVAL)); 3199 type = type32; 3200 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops); 3201 3202 switch (type) { 3203 case DMU_OST_ZFS: 3204 cbfunc = zfs_create_cb; 3205 break; 3206 3207 case DMU_OST_ZVOL: 3208 cbfunc = zvol_create_cb; 3209 break; 3210 3211 default: 3212 cbfunc = NULL; 3213 break; 3214 } 3215 if (strchr(fsname, '@') || 3216 strchr(fsname, '%')) 3217 return (SET_ERROR(EINVAL)); 3218 3219 zct.zct_props = nvprops; 3220 3221 if (cbfunc == NULL) 3222 return (SET_ERROR(EINVAL)); 3223 3224 if (type == DMU_OST_ZVOL) { 3225 uint64_t volsize, volblocksize; 3226 3227 if (nvprops == NULL) 3228 return (SET_ERROR(EINVAL)); 3229 if (nvlist_lookup_uint64(nvprops, 3230 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0) 3231 return (SET_ERROR(EINVAL)); 3232 3233 if ((error = nvlist_lookup_uint64(nvprops, 3234 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), 3235 &volblocksize)) != 0 && error != ENOENT) 3236 return (SET_ERROR(EINVAL)); 3237 3238 if (error != 0) 3239 volblocksize = zfs_prop_default_numeric( 3240 ZFS_PROP_VOLBLOCKSIZE); 3241 3242 if ((error = zvol_check_volblocksize( 3243 volblocksize)) != 0 || 3244 (error = zvol_check_volsize(volsize, 3245 volblocksize)) != 0) 3246 return (error); 3247 } else if (type == DMU_OST_ZFS) { 3248 int error; 3249 3250 /* 3251 * We have to have normalization and 3252 * case-folding flags correct when we do the 3253 * file system creation, so go figure them out 3254 * now. 3255 */ 3256 VERIFY(nvlist_alloc(&zct.zct_zplprops, 3257 NV_UNIQUE_NAME, KM_SLEEP) == 0); 3258 error = zfs_fill_zplprops(fsname, nvprops, 3259 zct.zct_zplprops, &is_insensitive); 3260 if (error != 0) { 3261 nvlist_free(zct.zct_zplprops); 3262 return (error); 3263 } 3264 } 3265 3266 error = dmu_objset_create(fsname, type, 3267 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct); 3268 nvlist_free(zct.zct_zplprops); 3269 3270 /* 3271 * It would be nice to do this atomically. 3272 */ 3273 if (error == 0) { 3274 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL, 3275 nvprops, outnvl); 3276 if (error != 0) 3277 (void) dsl_destroy_head(fsname); 3278 } 3279 return (error); 3280 } 3281 3282 /* 3283 * innvl: { 3284 * "origin" -> name of origin snapshot 3285 * (optional) "props" -> { prop -> value } 3286 * } 3287 * 3288 * outnvl: propname -> error code (int32) 3289 */ 3290 static int 3291 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3292 { 3293 int error = 0; 3294 nvlist_t *nvprops = NULL; 3295 char *origin_name; 3296 3297 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0) 3298 return (SET_ERROR(EINVAL)); 3299 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops); 3300 3301 if (strchr(fsname, '@') || 3302 strchr(fsname, '%')) 3303 return (SET_ERROR(EINVAL)); 3304 3305 if (dataset_namecheck(origin_name, NULL, NULL) != 0) 3306 return (SET_ERROR(EINVAL)); 3307 error = dmu_objset_clone(fsname, origin_name); 3308 if (error != 0) 3309 return (error); 3310 3311 /* 3312 * It would be nice to do this atomically. 3313 */ 3314 if (error == 0) { 3315 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL, 3316 nvprops, outnvl); 3317 if (error != 0) 3318 (void) dsl_destroy_head(fsname); 3319 } 3320 return (error); 3321 } 3322 3323 /* 3324 * innvl: { 3325 * "snaps" -> { snapshot1, snapshot2 } 3326 * (optional) "props" -> { prop -> value (string) } 3327 * } 3328 * 3329 * outnvl: snapshot -> error code (int32) 3330 */ 3331 static int 3332 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3333 { 3334 nvlist_t *snaps; 3335 nvlist_t *props = NULL; 3336 int error, poollen; 3337 nvpair_t *pair; 3338 3339 (void) nvlist_lookup_nvlist(innvl, "props", &props); 3340 if ((error = zfs_check_userprops(poolname, props)) != 0) 3341 return (error); 3342 3343 if (!nvlist_empty(props) && 3344 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS)) 3345 return (SET_ERROR(ENOTSUP)); 3346 3347 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 3348 return (SET_ERROR(EINVAL)); 3349 poollen = strlen(poolname); 3350 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 3351 pair = nvlist_next_nvpair(snaps, pair)) { 3352 const char *name = nvpair_name(pair); 3353 const char *cp = strchr(name, '@'); 3354 3355 /* 3356 * The snap name must contain an @, and the part after it must 3357 * contain only valid characters. 3358 */ 3359 if (cp == NULL || 3360 zfs_component_namecheck(cp + 1, NULL, NULL) != 0) 3361 return (SET_ERROR(EINVAL)); 3362 3363 /* 3364 * The snap must be in the specified pool. 3365 */ 3366 if (strncmp(name, poolname, poollen) != 0 || 3367 (name[poollen] != '/' && name[poollen] != '@')) 3368 return (SET_ERROR(EXDEV)); 3369 3370 /* This must be the only snap of this fs. */ 3371 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair); 3372 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) { 3373 if (strncmp(name, nvpair_name(pair2), cp - name + 1) 3374 == 0) { 3375 return (SET_ERROR(EXDEV)); 3376 } 3377 } 3378 } 3379 3380 error = dsl_dataset_snapshot(snaps, props, outnvl); 3381 return (error); 3382 } 3383 3384 /* 3385 * innvl: "message" -> string 3386 */ 3387 /* ARGSUSED */ 3388 static int 3389 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl) 3390 { 3391 char *message; 3392 spa_t *spa; 3393 int error; 3394 char *poolname; 3395 3396 /* 3397 * The poolname in the ioctl is not set, we get it from the TSD, 3398 * which was set at the end of the last successful ioctl that allows 3399 * logging. The secpolicy func already checked that it is set. 3400 * Only one log ioctl is allowed after each successful ioctl, so 3401 * we clear the TSD here. 3402 */ 3403 poolname = tsd_get(zfs_allow_log_key); 3404 (void) tsd_set(zfs_allow_log_key, NULL); 3405 error = spa_open(poolname, &spa, FTAG); 3406 strfree(poolname); 3407 if (error != 0) 3408 return (error); 3409 3410 if (nvlist_lookup_string(innvl, "message", &message) != 0) { 3411 spa_close(spa, FTAG); 3412 return (SET_ERROR(EINVAL)); 3413 } 3414 3415 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) { 3416 spa_close(spa, FTAG); 3417 return (SET_ERROR(ENOTSUP)); 3418 } 3419 3420 error = spa_history_log(spa, message); 3421 spa_close(spa, FTAG); 3422 return (error); 3423 } 3424 3425 /* 3426 * The dp_config_rwlock must not be held when calling this, because the 3427 * unmount may need to write out data. 3428 * 3429 * This function is best-effort. Callers must deal gracefully if it 3430 * remains mounted (or is remounted after this call). 3431 * 3432 * Returns 0 if the argument is not a snapshot, or it is not currently a 3433 * filesystem, or we were able to unmount it. Returns error code otherwise. 3434 */ 3435 int 3436 zfs_unmount_snap(const char *snapname) 3437 { 3438 vfs_t *vfsp; 3439 zfsvfs_t *zfsvfs; 3440 int err; 3441 3442 if (strchr(snapname, '@') == NULL) 3443 return (0); 3444 3445 vfsp = zfs_get_vfs(snapname); 3446 if (vfsp == NULL) 3447 return (0); 3448 3449 zfsvfs = vfsp->vfs_data; 3450 ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os))); 3451 3452 err = vn_vfswlock(vfsp->vfs_vnodecovered); 3453 VFS_RELE(vfsp); 3454 if (err != 0) 3455 return (SET_ERROR(err)); 3456 3457 /* 3458 * Always force the unmount for snapshots. 3459 */ 3460 (void) dounmount(vfsp, MS_FORCE, kcred); 3461 return (0); 3462 } 3463 3464 /* ARGSUSED */ 3465 static int 3466 zfs_unmount_snap_cb(const char *snapname, void *arg) 3467 { 3468 return (zfs_unmount_snap(snapname)); 3469 } 3470 3471 /* 3472 * When a clone is destroyed, its origin may also need to be destroyed, 3473 * in which case it must be unmounted. This routine will do that unmount 3474 * if necessary. 3475 */ 3476 void 3477 zfs_destroy_unmount_origin(const char *fsname) 3478 { 3479 int error; 3480 objset_t *os; 3481 dsl_dataset_t *ds; 3482 3483 error = dmu_objset_hold(fsname, FTAG, &os); 3484 if (error != 0) 3485 return; 3486 ds = dmu_objset_ds(os); 3487 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) { 3488 char originname[ZFS_MAX_DATASET_NAME_LEN]; 3489 dsl_dataset_name(ds->ds_prev, originname); 3490 dmu_objset_rele(os, FTAG); 3491 (void) zfs_unmount_snap(originname); 3492 } else { 3493 dmu_objset_rele(os, FTAG); 3494 } 3495 } 3496 3497 /* 3498 * innvl: { 3499 * "snaps" -> { snapshot1, snapshot2 } 3500 * (optional boolean) "defer" 3501 * } 3502 * 3503 * outnvl: snapshot -> error code (int32) 3504 * 3505 */ 3506 /* ARGSUSED */ 3507 static int 3508 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3509 { 3510 nvlist_t *snaps; 3511 nvpair_t *pair; 3512 boolean_t defer; 3513 3514 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 3515 return (SET_ERROR(EINVAL)); 3516 defer = nvlist_exists(innvl, "defer"); 3517 3518 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 3519 pair = nvlist_next_nvpair(snaps, pair)) { 3520 (void) zfs_unmount_snap(nvpair_name(pair)); 3521 } 3522 3523 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl)); 3524 } 3525 3526 /* 3527 * Create bookmarks. Bookmark names are of the form <fs>#<bmark>. 3528 * All bookmarks must be in the same pool. 3529 * 3530 * innvl: { 3531 * bookmark1 -> snapshot1, bookmark2 -> snapshot2 3532 * } 3533 * 3534 * outnvl: bookmark -> error code (int32) 3535 * 3536 */ 3537 /* ARGSUSED */ 3538 static int 3539 zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3540 { 3541 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 3542 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 3543 char *snap_name; 3544 3545 /* 3546 * Verify the snapshot argument. 3547 */ 3548 if (nvpair_value_string(pair, &snap_name) != 0) 3549 return (SET_ERROR(EINVAL)); 3550 3551 3552 /* Verify that the keys (bookmarks) are unique */ 3553 for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair); 3554 pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) { 3555 if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0) 3556 return (SET_ERROR(EINVAL)); 3557 } 3558 } 3559 3560 return (dsl_bookmark_create(innvl, outnvl)); 3561 } 3562 3563 /* 3564 * innvl: { 3565 * property 1, property 2, ... 3566 * } 3567 * 3568 * outnvl: { 3569 * bookmark name 1 -> { property 1, property 2, ... }, 3570 * bookmark name 2 -> { property 1, property 2, ... } 3571 * } 3572 * 3573 */ 3574 static int 3575 zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3576 { 3577 return (dsl_get_bookmarks(fsname, innvl, outnvl)); 3578 } 3579 3580 /* 3581 * innvl: { 3582 * bookmark name 1, bookmark name 2 3583 * } 3584 * 3585 * outnvl: bookmark -> error code (int32) 3586 * 3587 */ 3588 static int 3589 zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl, 3590 nvlist_t *outnvl) 3591 { 3592 int error, poollen; 3593 3594 poollen = strlen(poolname); 3595 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 3596 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 3597 const char *name = nvpair_name(pair); 3598 const char *cp = strchr(name, '#'); 3599 3600 /* 3601 * The bookmark name must contain an #, and the part after it 3602 * must contain only valid characters. 3603 */ 3604 if (cp == NULL || 3605 zfs_component_namecheck(cp + 1, NULL, NULL) != 0) 3606 return (SET_ERROR(EINVAL)); 3607 3608 /* 3609 * The bookmark must be in the specified pool. 3610 */ 3611 if (strncmp(name, poolname, poollen) != 0 || 3612 (name[poollen] != '/' && name[poollen] != '#')) 3613 return (SET_ERROR(EXDEV)); 3614 } 3615 3616 error = dsl_bookmark_destroy(innvl, outnvl); 3617 return (error); 3618 } 3619 3620 static int 3621 zfs_ioc_channel_program(const char *poolname, nvlist_t *innvl, 3622 nvlist_t *outnvl) 3623 { 3624 char *program; 3625 uint64_t instrlimit, memlimit; 3626 nvpair_t *nvarg = NULL; 3627 3628 if (0 != nvlist_lookup_string(innvl, ZCP_ARG_PROGRAM, &program)) { 3629 return (EINVAL); 3630 } 3631 if (0 != nvlist_lookup_uint64(innvl, ZCP_ARG_INSTRLIMIT, &instrlimit)) { 3632 instrlimit = ZCP_DEFAULT_INSTRLIMIT; 3633 } 3634 if (0 != nvlist_lookup_uint64(innvl, ZCP_ARG_MEMLIMIT, &memlimit)) { 3635 memlimit = ZCP_DEFAULT_MEMLIMIT; 3636 } 3637 if (0 != nvlist_lookup_nvpair(innvl, ZCP_ARG_ARGLIST, &nvarg)) { 3638 return (EINVAL); 3639 } 3640 3641 if (instrlimit == 0 || instrlimit > zfs_lua_max_instrlimit) 3642 return (EINVAL); 3643 if (memlimit == 0 || memlimit > ZCP_MAX_MEMLIMIT) 3644 return (EINVAL); 3645 3646 return (zcp_eval(poolname, program, instrlimit, memlimit, 3647 nvarg, outnvl)); 3648 } 3649 3650 /* 3651 * inputs: 3652 * zc_name name of dataset to destroy 3653 * zc_objset_type type of objset 3654 * zc_defer_destroy mark for deferred destroy 3655 * 3656 * outputs: none 3657 */ 3658 static int 3659 zfs_ioc_destroy(zfs_cmd_t *zc) 3660 { 3661 int err; 3662 3663 if (zc->zc_objset_type == DMU_OST_ZFS) { 3664 err = zfs_unmount_snap(zc->zc_name); 3665 if (err != 0) 3666 return (err); 3667 } 3668 3669 if (strchr(zc->zc_name, '@')) 3670 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy); 3671 else 3672 err = dsl_destroy_head(zc->zc_name); 3673 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0) 3674 (void) zvol_remove_minor(zc->zc_name); 3675 return (err); 3676 } 3677 3678 /* 3679 * fsname is name of dataset to rollback (to most recent snapshot) 3680 * 3681 * innvl is not used. 3682 * 3683 * outnvl: "target" -> name of most recent snapshot 3684 * } 3685 */ 3686 /* ARGSUSED */ 3687 static int 3688 zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl) 3689 { 3690 zfsvfs_t *zfsvfs; 3691 int error; 3692 3693 if (getzfsvfs(fsname, &zfsvfs) == 0) { 3694 dsl_dataset_t *ds; 3695 3696 ds = dmu_objset_ds(zfsvfs->z_os); 3697 error = zfs_suspend_fs(zfsvfs); 3698 if (error == 0) { 3699 int resume_err; 3700 3701 error = dsl_dataset_rollback(fsname, zfsvfs, outnvl); 3702 resume_err = zfs_resume_fs(zfsvfs, ds); 3703 error = error ? error : resume_err; 3704 } 3705 VFS_RELE(zfsvfs->z_vfs); 3706 } else { 3707 error = dsl_dataset_rollback(fsname, NULL, outnvl); 3708 } 3709 return (error); 3710 } 3711 3712 static int 3713 recursive_unmount(const char *fsname, void *arg) 3714 { 3715 const char *snapname = arg; 3716 char fullname[ZFS_MAX_DATASET_NAME_LEN]; 3717 3718 (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname); 3719 return (zfs_unmount_snap(fullname)); 3720 } 3721 3722 /* 3723 * inputs: 3724 * zc_name old name of dataset 3725 * zc_value new name of dataset 3726 * zc_cookie recursive flag (only valid for snapshots) 3727 * 3728 * outputs: none 3729 */ 3730 static int 3731 zfs_ioc_rename(zfs_cmd_t *zc) 3732 { 3733 boolean_t recursive = zc->zc_cookie & 1; 3734 char *at; 3735 3736 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0'; 3737 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 3738 strchr(zc->zc_value, '%')) 3739 return (SET_ERROR(EINVAL)); 3740 3741 at = strchr(zc->zc_name, '@'); 3742 if (at != NULL) { 3743 /* snaps must be in same fs */ 3744 int error; 3745 3746 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1)) 3747 return (SET_ERROR(EXDEV)); 3748 *at = '\0'; 3749 if (zc->zc_objset_type == DMU_OST_ZFS) { 3750 error = dmu_objset_find(zc->zc_name, 3751 recursive_unmount, at + 1, 3752 recursive ? DS_FIND_CHILDREN : 0); 3753 if (error != 0) { 3754 *at = '@'; 3755 return (error); 3756 } 3757 } 3758 error = dsl_dataset_rename_snapshot(zc->zc_name, 3759 at + 1, strchr(zc->zc_value, '@') + 1, recursive); 3760 *at = '@'; 3761 3762 return (error); 3763 } else { 3764 if (zc->zc_objset_type == DMU_OST_ZVOL) 3765 (void) zvol_remove_minor(zc->zc_name); 3766 return (dsl_dir_rename(zc->zc_name, zc->zc_value)); 3767 } 3768 } 3769 3770 static int 3771 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr) 3772 { 3773 const char *propname = nvpair_name(pair); 3774 boolean_t issnap = (strchr(dsname, '@') != NULL); 3775 zfs_prop_t prop = zfs_name_to_prop(propname); 3776 uint64_t intval; 3777 int err; 3778 3779 if (prop == ZPROP_INVAL) { 3780 if (zfs_prop_user(propname)) { 3781 if (err = zfs_secpolicy_write_perms(dsname, 3782 ZFS_DELEG_PERM_USERPROP, cr)) 3783 return (err); 3784 return (0); 3785 } 3786 3787 if (!issnap && zfs_prop_userquota(propname)) { 3788 const char *perm = NULL; 3789 const char *uq_prefix = 3790 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA]; 3791 const char *gq_prefix = 3792 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA]; 3793 3794 if (strncmp(propname, uq_prefix, 3795 strlen(uq_prefix)) == 0) { 3796 perm = ZFS_DELEG_PERM_USERQUOTA; 3797 } else if (strncmp(propname, gq_prefix, 3798 strlen(gq_prefix)) == 0) { 3799 perm = ZFS_DELEG_PERM_GROUPQUOTA; 3800 } else { 3801 /* USERUSED and GROUPUSED are read-only */ 3802 return (SET_ERROR(EINVAL)); 3803 } 3804 3805 if (err = zfs_secpolicy_write_perms(dsname, perm, cr)) 3806 return (err); 3807 return (0); 3808 } 3809 3810 return (SET_ERROR(EINVAL)); 3811 } 3812 3813 if (issnap) 3814 return (SET_ERROR(EINVAL)); 3815 3816 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 3817 /* 3818 * dsl_prop_get_all_impl() returns properties in this 3819 * format. 3820 */ 3821 nvlist_t *attrs; 3822 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 3823 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 3824 &pair) == 0); 3825 } 3826 3827 /* 3828 * Check that this value is valid for this pool version 3829 */ 3830 switch (prop) { 3831 case ZFS_PROP_COMPRESSION: 3832 /* 3833 * If the user specified gzip compression, make sure 3834 * the SPA supports it. We ignore any errors here since 3835 * we'll catch them later. 3836 */ 3837 if (nvpair_value_uint64(pair, &intval) == 0) { 3838 if (intval >= ZIO_COMPRESS_GZIP_1 && 3839 intval <= ZIO_COMPRESS_GZIP_9 && 3840 zfs_earlier_version(dsname, 3841 SPA_VERSION_GZIP_COMPRESSION)) { 3842 return (SET_ERROR(ENOTSUP)); 3843 } 3844 3845 if (intval == ZIO_COMPRESS_ZLE && 3846 zfs_earlier_version(dsname, 3847 SPA_VERSION_ZLE_COMPRESSION)) 3848 return (SET_ERROR(ENOTSUP)); 3849 3850 if (intval == ZIO_COMPRESS_LZ4) { 3851 spa_t *spa; 3852 3853 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 3854 return (err); 3855 3856 if (!spa_feature_is_enabled(spa, 3857 SPA_FEATURE_LZ4_COMPRESS)) { 3858 spa_close(spa, FTAG); 3859 return (SET_ERROR(ENOTSUP)); 3860 } 3861 spa_close(spa, FTAG); 3862 } 3863 3864 /* 3865 * If this is a bootable dataset then 3866 * verify that the compression algorithm 3867 * is supported for booting. We must return 3868 * something other than ENOTSUP since it 3869 * implies a downrev pool version. 3870 */ 3871 if (zfs_is_bootfs(dsname) && 3872 !BOOTFS_COMPRESS_VALID(intval)) { 3873 return (SET_ERROR(ERANGE)); 3874 } 3875 } 3876 break; 3877 3878 case ZFS_PROP_COPIES: 3879 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS)) 3880 return (SET_ERROR(ENOTSUP)); 3881 break; 3882 3883 case ZFS_PROP_RECORDSIZE: 3884 /* Record sizes above 128k need the feature to be enabled */ 3885 if (nvpair_value_uint64(pair, &intval) == 0 && 3886 intval > SPA_OLD_MAXBLOCKSIZE) { 3887 spa_t *spa; 3888 3889 /* 3890 * We don't allow setting the property above 1MB, 3891 * unless the tunable has been changed. 3892 */ 3893 if (intval > zfs_max_recordsize || 3894 intval > SPA_MAXBLOCKSIZE) 3895 return (SET_ERROR(ERANGE)); 3896 3897 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 3898 return (err); 3899 3900 if (!spa_feature_is_enabled(spa, 3901 SPA_FEATURE_LARGE_BLOCKS)) { 3902 spa_close(spa, FTAG); 3903 return (SET_ERROR(ENOTSUP)); 3904 } 3905 spa_close(spa, FTAG); 3906 } 3907 break; 3908 3909 case ZFS_PROP_SHARESMB: 3910 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID)) 3911 return (SET_ERROR(ENOTSUP)); 3912 break; 3913 3914 case ZFS_PROP_ACLINHERIT: 3915 if (nvpair_type(pair) == DATA_TYPE_UINT64 && 3916 nvpair_value_uint64(pair, &intval) == 0) { 3917 if (intval == ZFS_ACL_PASSTHROUGH_X && 3918 zfs_earlier_version(dsname, 3919 SPA_VERSION_PASSTHROUGH_X)) 3920 return (SET_ERROR(ENOTSUP)); 3921 } 3922 break; 3923 3924 case ZFS_PROP_CHECKSUM: 3925 case ZFS_PROP_DEDUP: 3926 { 3927 spa_feature_t feature; 3928 spa_t *spa; 3929 3930 /* dedup feature version checks */ 3931 if (prop == ZFS_PROP_DEDUP && 3932 zfs_earlier_version(dsname, SPA_VERSION_DEDUP)) 3933 return (SET_ERROR(ENOTSUP)); 3934 3935 if (nvpair_value_uint64(pair, &intval) != 0) 3936 return (SET_ERROR(EINVAL)); 3937 3938 /* check prop value is enabled in features */ 3939 feature = zio_checksum_to_feature(intval & ZIO_CHECKSUM_MASK); 3940 if (feature == SPA_FEATURE_NONE) 3941 break; 3942 3943 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 3944 return (err); 3945 /* 3946 * Salted checksums are not supported on root pools. 3947 */ 3948 if (spa_bootfs(spa) != 0 && 3949 intval < ZIO_CHECKSUM_FUNCTIONS && 3950 (zio_checksum_table[intval].ci_flags & 3951 ZCHECKSUM_FLAG_SALTED)) { 3952 spa_close(spa, FTAG); 3953 return (SET_ERROR(ERANGE)); 3954 } 3955 if (!spa_feature_is_enabled(spa, feature)) { 3956 spa_close(spa, FTAG); 3957 return (SET_ERROR(ENOTSUP)); 3958 } 3959 spa_close(spa, FTAG); 3960 break; 3961 } 3962 } 3963 3964 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED())); 3965 } 3966 3967 /* 3968 * Checks for a race condition to make sure we don't increment a feature flag 3969 * multiple times. 3970 */ 3971 static int 3972 zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx) 3973 { 3974 spa_t *spa = dmu_tx_pool(tx)->dp_spa; 3975 spa_feature_t *featurep = arg; 3976 3977 if (!spa_feature_is_active(spa, *featurep)) 3978 return (0); 3979 else 3980 return (SET_ERROR(EBUSY)); 3981 } 3982 3983 /* 3984 * The callback invoked on feature activation in the sync task caused by 3985 * zfs_prop_activate_feature. 3986 */ 3987 static void 3988 zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx) 3989 { 3990 spa_t *spa = dmu_tx_pool(tx)->dp_spa; 3991 spa_feature_t *featurep = arg; 3992 3993 spa_feature_incr(spa, *featurep, tx); 3994 } 3995 3996 /* 3997 * Activates a feature on a pool in response to a property setting. This 3998 * creates a new sync task which modifies the pool to reflect the feature 3999 * as being active. 4000 */ 4001 static int 4002 zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature) 4003 { 4004 int err; 4005 4006 /* EBUSY here indicates that the feature is already active */ 4007 err = dsl_sync_task(spa_name(spa), 4008 zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync, 4009 &feature, 2, ZFS_SPACE_CHECK_RESERVED); 4010 4011 if (err != 0 && err != EBUSY) 4012 return (err); 4013 else 4014 return (0); 4015 } 4016 4017 /* 4018 * Removes properties from the given props list that fail permission checks 4019 * needed to clear them and to restore them in case of a receive error. For each 4020 * property, make sure we have both set and inherit permissions. 4021 * 4022 * Returns the first error encountered if any permission checks fail. If the 4023 * caller provides a non-NULL errlist, it also gives the complete list of names 4024 * of all the properties that failed a permission check along with the 4025 * corresponding error numbers. The caller is responsible for freeing the 4026 * returned errlist. 4027 * 4028 * If every property checks out successfully, zero is returned and the list 4029 * pointed at by errlist is NULL. 4030 */ 4031 static int 4032 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist) 4033 { 4034 zfs_cmd_t *zc; 4035 nvpair_t *pair, *next_pair; 4036 nvlist_t *errors; 4037 int err, rv = 0; 4038 4039 if (props == NULL) 4040 return (0); 4041 4042 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0); 4043 4044 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP); 4045 (void) strcpy(zc->zc_name, dataset); 4046 pair = nvlist_next_nvpair(props, NULL); 4047 while (pair != NULL) { 4048 next_pair = nvlist_next_nvpair(props, pair); 4049 4050 (void) strcpy(zc->zc_value, nvpair_name(pair)); 4051 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 || 4052 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) { 4053 VERIFY(nvlist_remove_nvpair(props, pair) == 0); 4054 VERIFY(nvlist_add_int32(errors, 4055 zc->zc_value, err) == 0); 4056 } 4057 pair = next_pair; 4058 } 4059 kmem_free(zc, sizeof (zfs_cmd_t)); 4060 4061 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) { 4062 nvlist_free(errors); 4063 errors = NULL; 4064 } else { 4065 VERIFY(nvpair_value_int32(pair, &rv) == 0); 4066 } 4067 4068 if (errlist == NULL) 4069 nvlist_free(errors); 4070 else 4071 *errlist = errors; 4072 4073 return (rv); 4074 } 4075 4076 static boolean_t 4077 propval_equals(nvpair_t *p1, nvpair_t *p2) 4078 { 4079 if (nvpair_type(p1) == DATA_TYPE_NVLIST) { 4080 /* dsl_prop_get_all_impl() format */ 4081 nvlist_t *attrs; 4082 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0); 4083 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 4084 &p1) == 0); 4085 } 4086 4087 if (nvpair_type(p2) == DATA_TYPE_NVLIST) { 4088 nvlist_t *attrs; 4089 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0); 4090 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 4091 &p2) == 0); 4092 } 4093 4094 if (nvpair_type(p1) != nvpair_type(p2)) 4095 return (B_FALSE); 4096 4097 if (nvpair_type(p1) == DATA_TYPE_STRING) { 4098 char *valstr1, *valstr2; 4099 4100 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0); 4101 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0); 4102 return (strcmp(valstr1, valstr2) == 0); 4103 } else { 4104 uint64_t intval1, intval2; 4105 4106 VERIFY(nvpair_value_uint64(p1, &intval1) == 0); 4107 VERIFY(nvpair_value_uint64(p2, &intval2) == 0); 4108 return (intval1 == intval2); 4109 } 4110 } 4111 4112 /* 4113 * Remove properties from props if they are not going to change (as determined 4114 * by comparison with origprops). Remove them from origprops as well, since we 4115 * do not need to clear or restore properties that won't change. 4116 */ 4117 static void 4118 props_reduce(nvlist_t *props, nvlist_t *origprops) 4119 { 4120 nvpair_t *pair, *next_pair; 4121 4122 if (origprops == NULL) 4123 return; /* all props need to be received */ 4124 4125 pair = nvlist_next_nvpair(props, NULL); 4126 while (pair != NULL) { 4127 const char *propname = nvpair_name(pair); 4128 nvpair_t *match; 4129 4130 next_pair = nvlist_next_nvpair(props, pair); 4131 4132 if ((nvlist_lookup_nvpair(origprops, propname, 4133 &match) != 0) || !propval_equals(pair, match)) 4134 goto next; /* need to set received value */ 4135 4136 /* don't clear the existing received value */ 4137 (void) nvlist_remove_nvpair(origprops, match); 4138 /* don't bother receiving the property */ 4139 (void) nvlist_remove_nvpair(props, pair); 4140 next: 4141 pair = next_pair; 4142 } 4143 } 4144 4145 /* 4146 * Extract properties that cannot be set PRIOR to the receipt of a dataset. 4147 * For example, refquota cannot be set until after the receipt of a dataset, 4148 * because in replication streams, an older/earlier snapshot may exceed the 4149 * refquota. We want to receive the older/earlier snapshot, but setting 4150 * refquota pre-receipt will set the dsl's ACTUAL quota, which will prevent 4151 * the older/earlier snapshot from being received (with EDQUOT). 4152 * 4153 * The ZFS test "zfs_receive_011_pos" demonstrates such a scenario. 4154 * 4155 * libzfs will need to be judicious handling errors encountered by props 4156 * extracted by this function. 4157 */ 4158 static nvlist_t * 4159 extract_delay_props(nvlist_t *props) 4160 { 4161 nvlist_t *delayprops; 4162 nvpair_t *nvp, *tmp; 4163 static const zfs_prop_t delayable[] = { ZFS_PROP_REFQUOTA, 0 }; 4164 int i; 4165 4166 VERIFY(nvlist_alloc(&delayprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 4167 4168 for (nvp = nvlist_next_nvpair(props, NULL); nvp != NULL; 4169 nvp = nvlist_next_nvpair(props, nvp)) { 4170 /* 4171 * strcmp() is safe because zfs_prop_to_name() always returns 4172 * a bounded string. 4173 */ 4174 for (i = 0; delayable[i] != 0; i++) { 4175 if (strcmp(zfs_prop_to_name(delayable[i]), 4176 nvpair_name(nvp)) == 0) { 4177 break; 4178 } 4179 } 4180 if (delayable[i] != 0) { 4181 tmp = nvlist_prev_nvpair(props, nvp); 4182 VERIFY(nvlist_add_nvpair(delayprops, nvp) == 0); 4183 VERIFY(nvlist_remove_nvpair(props, nvp) == 0); 4184 nvp = tmp; 4185 } 4186 } 4187 4188 if (nvlist_empty(delayprops)) { 4189 nvlist_free(delayprops); 4190 delayprops = NULL; 4191 } 4192 return (delayprops); 4193 } 4194 4195 #ifdef DEBUG 4196 static boolean_t zfs_ioc_recv_inject_err; 4197 #endif 4198 4199 /* 4200 * inputs: 4201 * zc_name name of containing filesystem 4202 * zc_nvlist_src{_size} nvlist of properties to apply 4203 * zc_value name of snapshot to create 4204 * zc_string name of clone origin (if DRR_FLAG_CLONE) 4205 * zc_cookie file descriptor to recv from 4206 * zc_begin_record the BEGIN record of the stream (not byteswapped) 4207 * zc_guid force flag 4208 * zc_cleanup_fd cleanup-on-exit file descriptor 4209 * zc_action_handle handle for this guid/ds mapping (or zero on first call) 4210 * zc_resumable if data is incomplete assume sender will resume 4211 * 4212 * outputs: 4213 * zc_cookie number of bytes read 4214 * zc_nvlist_dst{_size} error for each unapplied received property 4215 * zc_obj zprop_errflags_t 4216 * zc_action_handle handle for this guid/ds mapping 4217 */ 4218 static int 4219 zfs_ioc_recv(zfs_cmd_t *zc) 4220 { 4221 file_t *fp; 4222 dmu_recv_cookie_t drc; 4223 boolean_t force = (boolean_t)zc->zc_guid; 4224 int fd; 4225 int error = 0; 4226 int props_error = 0; 4227 nvlist_t *errors; 4228 offset_t off; 4229 nvlist_t *props = NULL; /* sent properties */ 4230 nvlist_t *origprops = NULL; /* existing properties */ 4231 nvlist_t *delayprops = NULL; /* sent properties applied post-receive */ 4232 char *origin = NULL; 4233 char *tosnap; 4234 char tofs[ZFS_MAX_DATASET_NAME_LEN]; 4235 boolean_t first_recvd_props = B_FALSE; 4236 4237 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 4238 strchr(zc->zc_value, '@') == NULL || 4239 strchr(zc->zc_value, '%')) 4240 return (SET_ERROR(EINVAL)); 4241 4242 (void) strcpy(tofs, zc->zc_value); 4243 tosnap = strchr(tofs, '@'); 4244 *tosnap++ = '\0'; 4245 4246 if (zc->zc_nvlist_src != NULL && 4247 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 4248 zc->zc_iflags, &props)) != 0) 4249 return (error); 4250 4251 fd = zc->zc_cookie; 4252 fp = getf(fd); 4253 if (fp == NULL) { 4254 nvlist_free(props); 4255 return (SET_ERROR(EBADF)); 4256 } 4257 4258 errors = fnvlist_alloc(); 4259 4260 if (zc->zc_string[0]) 4261 origin = zc->zc_string; 4262 4263 error = dmu_recv_begin(tofs, tosnap, 4264 &zc->zc_begin_record, force, zc->zc_resumable, origin, &drc); 4265 if (error != 0) 4266 goto out; 4267 4268 /* 4269 * Set properties before we receive the stream so that they are applied 4270 * to the new data. Note that we must call dmu_recv_stream() if 4271 * dmu_recv_begin() succeeds. 4272 */ 4273 if (props != NULL && !drc.drc_newfs) { 4274 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >= 4275 SPA_VERSION_RECVD_PROPS && 4276 !dsl_prop_get_hasrecvd(tofs)) 4277 first_recvd_props = B_TRUE; 4278 4279 /* 4280 * If new received properties are supplied, they are to 4281 * completely replace the existing received properties, so stash 4282 * away the existing ones. 4283 */ 4284 if (dsl_prop_get_received(tofs, &origprops) == 0) { 4285 nvlist_t *errlist = NULL; 4286 /* 4287 * Don't bother writing a property if its value won't 4288 * change (and avoid the unnecessary security checks). 4289 * 4290 * The first receive after SPA_VERSION_RECVD_PROPS is a 4291 * special case where we blow away all local properties 4292 * regardless. 4293 */ 4294 if (!first_recvd_props) 4295 props_reduce(props, origprops); 4296 if (zfs_check_clearable(tofs, origprops, &errlist) != 0) 4297 (void) nvlist_merge(errors, errlist, 0); 4298 nvlist_free(errlist); 4299 4300 if (clear_received_props(tofs, origprops, 4301 first_recvd_props ? NULL : props) != 0) 4302 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 4303 } else { 4304 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 4305 } 4306 } 4307 4308 if (props != NULL) { 4309 props_error = dsl_prop_set_hasrecvd(tofs); 4310 4311 if (props_error == 0) { 4312 delayprops = extract_delay_props(props); 4313 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED, 4314 props, errors); 4315 } 4316 } 4317 4318 off = fp->f_offset; 4319 error = dmu_recv_stream(&drc, fp->f_vnode, &off, zc->zc_cleanup_fd, 4320 &zc->zc_action_handle); 4321 4322 if (error == 0) { 4323 zfsvfs_t *zfsvfs = NULL; 4324 4325 if (getzfsvfs(tofs, &zfsvfs) == 0) { 4326 /* online recv */ 4327 dsl_dataset_t *ds; 4328 int end_err; 4329 4330 ds = dmu_objset_ds(zfsvfs->z_os); 4331 error = zfs_suspend_fs(zfsvfs); 4332 /* 4333 * If the suspend fails, then the recv_end will 4334 * likely also fail, and clean up after itself. 4335 */ 4336 end_err = dmu_recv_end(&drc, zfsvfs); 4337 if (error == 0) 4338 error = zfs_resume_fs(zfsvfs, ds); 4339 error = error ? error : end_err; 4340 VFS_RELE(zfsvfs->z_vfs); 4341 } else { 4342 error = dmu_recv_end(&drc, NULL); 4343 } 4344 4345 /* Set delayed properties now, after we're done receiving. */ 4346 if (delayprops != NULL && error == 0) { 4347 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED, 4348 delayprops, errors); 4349 } 4350 } 4351 4352 if (delayprops != NULL) { 4353 /* 4354 * Merge delayed props back in with initial props, in case 4355 * we're DEBUG and zfs_ioc_recv_inject_err is set (which means 4356 * we have to make sure clear_received_props() includes 4357 * the delayed properties). 4358 * 4359 * Since zfs_ioc_recv_inject_err is only in DEBUG kernels, 4360 * using ASSERT() will be just like a VERIFY. 4361 */ 4362 ASSERT(nvlist_merge(props, delayprops, 0) == 0); 4363 nvlist_free(delayprops); 4364 } 4365 4366 /* 4367 * Now that all props, initial and delayed, are set, report the prop 4368 * errors to the caller. 4369 */ 4370 if (zc->zc_nvlist_dst_size != 0 && 4371 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 || 4372 put_nvlist(zc, errors) != 0)) { 4373 /* 4374 * Caller made zc->zc_nvlist_dst less than the minimum expected 4375 * size or supplied an invalid address. 4376 */ 4377 props_error = SET_ERROR(EINVAL); 4378 } 4379 4380 zc->zc_cookie = off - fp->f_offset; 4381 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0) 4382 fp->f_offset = off; 4383 4384 #ifdef DEBUG 4385 if (zfs_ioc_recv_inject_err) { 4386 zfs_ioc_recv_inject_err = B_FALSE; 4387 error = 1; 4388 } 4389 #endif 4390 /* 4391 * On error, restore the original props. 4392 */ 4393 if (error != 0 && props != NULL && !drc.drc_newfs) { 4394 if (clear_received_props(tofs, props, NULL) != 0) { 4395 /* 4396 * We failed to clear the received properties. 4397 * Since we may have left a $recvd value on the 4398 * system, we can't clear the $hasrecvd flag. 4399 */ 4400 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4401 } else if (first_recvd_props) { 4402 dsl_prop_unset_hasrecvd(tofs); 4403 } 4404 4405 if (origprops == NULL && !drc.drc_newfs) { 4406 /* We failed to stash the original properties. */ 4407 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4408 } 4409 4410 /* 4411 * dsl_props_set() will not convert RECEIVED to LOCAL on or 4412 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL 4413 * explictly if we're restoring local properties cleared in the 4414 * first new-style receive. 4415 */ 4416 if (origprops != NULL && 4417 zfs_set_prop_nvlist(tofs, (first_recvd_props ? 4418 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED), 4419 origprops, NULL) != 0) { 4420 /* 4421 * We stashed the original properties but failed to 4422 * restore them. 4423 */ 4424 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4425 } 4426 } 4427 out: 4428 nvlist_free(props); 4429 nvlist_free(origprops); 4430 nvlist_free(errors); 4431 releasef(fd); 4432 4433 if (error == 0) 4434 error = props_error; 4435 4436 return (error); 4437 } 4438 4439 /* 4440 * inputs: 4441 * zc_name name of snapshot to send 4442 * zc_cookie file descriptor to send stream to 4443 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj) 4444 * zc_sendobj objsetid of snapshot to send 4445 * zc_fromobj objsetid of incremental fromsnap (may be zero) 4446 * zc_guid if set, estimate size of stream only. zc_cookie is ignored. 4447 * output size in zc_objset_type. 4448 * zc_flags lzc_send_flags 4449 * 4450 * outputs: 4451 * zc_objset_type estimated size, if zc_guid is set 4452 */ 4453 static int 4454 zfs_ioc_send(zfs_cmd_t *zc) 4455 { 4456 int error; 4457 offset_t off; 4458 boolean_t estimate = (zc->zc_guid != 0); 4459 boolean_t embedok = (zc->zc_flags & 0x1); 4460 boolean_t large_block_ok = (zc->zc_flags & 0x2); 4461 boolean_t compressok = (zc->zc_flags & 0x4); 4462 4463 if (zc->zc_obj != 0) { 4464 dsl_pool_t *dp; 4465 dsl_dataset_t *tosnap; 4466 4467 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4468 if (error != 0) 4469 return (error); 4470 4471 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap); 4472 if (error != 0) { 4473 dsl_pool_rele(dp, FTAG); 4474 return (error); 4475 } 4476 4477 if (dsl_dir_is_clone(tosnap->ds_dir)) 4478 zc->zc_fromobj = 4479 dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj; 4480 dsl_dataset_rele(tosnap, FTAG); 4481 dsl_pool_rele(dp, FTAG); 4482 } 4483 4484 if (estimate) { 4485 dsl_pool_t *dp; 4486 dsl_dataset_t *tosnap; 4487 dsl_dataset_t *fromsnap = NULL; 4488 4489 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4490 if (error != 0) 4491 return (error); 4492 4493 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap); 4494 if (error != 0) { 4495 dsl_pool_rele(dp, FTAG); 4496 return (error); 4497 } 4498 4499 if (zc->zc_fromobj != 0) { 4500 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj, 4501 FTAG, &fromsnap); 4502 if (error != 0) { 4503 dsl_dataset_rele(tosnap, FTAG); 4504 dsl_pool_rele(dp, FTAG); 4505 return (error); 4506 } 4507 } 4508 4509 error = dmu_send_estimate(tosnap, fromsnap, compressok, 4510 &zc->zc_objset_type); 4511 4512 if (fromsnap != NULL) 4513 dsl_dataset_rele(fromsnap, FTAG); 4514 dsl_dataset_rele(tosnap, FTAG); 4515 dsl_pool_rele(dp, FTAG); 4516 } else { 4517 file_t *fp = getf(zc->zc_cookie); 4518 if (fp == NULL) 4519 return (SET_ERROR(EBADF)); 4520 4521 off = fp->f_offset; 4522 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj, 4523 zc->zc_fromobj, embedok, large_block_ok, compressok, 4524 zc->zc_cookie, fp->f_vnode, &off); 4525 4526 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0) 4527 fp->f_offset = off; 4528 releasef(zc->zc_cookie); 4529 } 4530 return (error); 4531 } 4532 4533 /* 4534 * inputs: 4535 * zc_name name of snapshot on which to report progress 4536 * zc_cookie file descriptor of send stream 4537 * 4538 * outputs: 4539 * zc_cookie number of bytes written in send stream thus far 4540 */ 4541 static int 4542 zfs_ioc_send_progress(zfs_cmd_t *zc) 4543 { 4544 dsl_pool_t *dp; 4545 dsl_dataset_t *ds; 4546 dmu_sendarg_t *dsp = NULL; 4547 int error; 4548 4549 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4550 if (error != 0) 4551 return (error); 4552 4553 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds); 4554 if (error != 0) { 4555 dsl_pool_rele(dp, FTAG); 4556 return (error); 4557 } 4558 4559 mutex_enter(&ds->ds_sendstream_lock); 4560 4561 /* 4562 * Iterate over all the send streams currently active on this dataset. 4563 * If there's one which matches the specified file descriptor _and_ the 4564 * stream was started by the current process, return the progress of 4565 * that stream. 4566 */ 4567 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL; 4568 dsp = list_next(&ds->ds_sendstreams, dsp)) { 4569 if (dsp->dsa_outfd == zc->zc_cookie && 4570 dsp->dsa_proc == curproc) 4571 break; 4572 } 4573 4574 if (dsp != NULL) 4575 zc->zc_cookie = *(dsp->dsa_off); 4576 else 4577 error = SET_ERROR(ENOENT); 4578 4579 mutex_exit(&ds->ds_sendstream_lock); 4580 dsl_dataset_rele(ds, FTAG); 4581 dsl_pool_rele(dp, FTAG); 4582 return (error); 4583 } 4584 4585 static int 4586 zfs_ioc_inject_fault(zfs_cmd_t *zc) 4587 { 4588 int id, error; 4589 4590 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id, 4591 &zc->zc_inject_record); 4592 4593 if (error == 0) 4594 zc->zc_guid = (uint64_t)id; 4595 4596 return (error); 4597 } 4598 4599 static int 4600 zfs_ioc_clear_fault(zfs_cmd_t *zc) 4601 { 4602 return (zio_clear_fault((int)zc->zc_guid)); 4603 } 4604 4605 static int 4606 zfs_ioc_inject_list_next(zfs_cmd_t *zc) 4607 { 4608 int id = (int)zc->zc_guid; 4609 int error; 4610 4611 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name), 4612 &zc->zc_inject_record); 4613 4614 zc->zc_guid = id; 4615 4616 return (error); 4617 } 4618 4619 static int 4620 zfs_ioc_error_log(zfs_cmd_t *zc) 4621 { 4622 spa_t *spa; 4623 int error; 4624 size_t count = (size_t)zc->zc_nvlist_dst_size; 4625 4626 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 4627 return (error); 4628 4629 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst, 4630 &count); 4631 if (error == 0) 4632 zc->zc_nvlist_dst_size = count; 4633 else 4634 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa); 4635 4636 spa_close(spa, FTAG); 4637 4638 return (error); 4639 } 4640 4641 static int 4642 zfs_ioc_clear(zfs_cmd_t *zc) 4643 { 4644 spa_t *spa; 4645 vdev_t *vd; 4646 int error; 4647 4648 /* 4649 * On zpool clear we also fix up missing slogs 4650 */ 4651 mutex_enter(&spa_namespace_lock); 4652 spa = spa_lookup(zc->zc_name); 4653 if (spa == NULL) { 4654 mutex_exit(&spa_namespace_lock); 4655 return (SET_ERROR(EIO)); 4656 } 4657 if (spa_get_log_state(spa) == SPA_LOG_MISSING) { 4658 /* we need to let spa_open/spa_load clear the chains */ 4659 spa_set_log_state(spa, SPA_LOG_CLEAR); 4660 } 4661 spa->spa_last_open_failed = 0; 4662 mutex_exit(&spa_namespace_lock); 4663 4664 if (zc->zc_cookie & ZPOOL_NO_REWIND) { 4665 error = spa_open(zc->zc_name, &spa, FTAG); 4666 } else { 4667 nvlist_t *policy; 4668 nvlist_t *config = NULL; 4669 4670 if (zc->zc_nvlist_src == NULL) 4671 return (SET_ERROR(EINVAL)); 4672 4673 if ((error = get_nvlist(zc->zc_nvlist_src, 4674 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) { 4675 error = spa_open_rewind(zc->zc_name, &spa, FTAG, 4676 policy, &config); 4677 if (config != NULL) { 4678 int err; 4679 4680 if ((err = put_nvlist(zc, config)) != 0) 4681 error = err; 4682 nvlist_free(config); 4683 } 4684 nvlist_free(policy); 4685 } 4686 } 4687 4688 if (error != 0) 4689 return (error); 4690 4691 spa_vdev_state_enter(spa, SCL_NONE); 4692 4693 if (zc->zc_guid == 0) { 4694 vd = NULL; 4695 } else { 4696 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE); 4697 if (vd == NULL) { 4698 (void) spa_vdev_state_exit(spa, NULL, ENODEV); 4699 spa_close(spa, FTAG); 4700 return (SET_ERROR(ENODEV)); 4701 } 4702 } 4703 4704 vdev_clear(spa, vd); 4705 4706 (void) spa_vdev_state_exit(spa, NULL, 0); 4707 4708 /* 4709 * Resume any suspended I/Os. 4710 */ 4711 if (zio_resume(spa) != 0) 4712 error = SET_ERROR(EIO); 4713 4714 spa_close(spa, FTAG); 4715 4716 return (error); 4717 } 4718 4719 static int 4720 zfs_ioc_pool_reopen(zfs_cmd_t *zc) 4721 { 4722 spa_t *spa; 4723 int error; 4724 4725 error = spa_open(zc->zc_name, &spa, FTAG); 4726 if (error != 0) 4727 return (error); 4728 4729 spa_vdev_state_enter(spa, SCL_NONE); 4730 4731 /* 4732 * If a resilver is already in progress then set the 4733 * spa_scrub_reopen flag to B_TRUE so that we don't restart 4734 * the scan as a side effect of the reopen. Otherwise, let 4735 * vdev_open() decided if a resilver is required. 4736 */ 4737 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool); 4738 vdev_reopen(spa->spa_root_vdev); 4739 spa->spa_scrub_reopen = B_FALSE; 4740 4741 (void) spa_vdev_state_exit(spa, NULL, 0); 4742 spa_close(spa, FTAG); 4743 return (0); 4744 } 4745 /* 4746 * inputs: 4747 * zc_name name of filesystem 4748 * 4749 * outputs: 4750 * zc_string name of conflicting snapshot, if there is one 4751 */ 4752 static int 4753 zfs_ioc_promote(zfs_cmd_t *zc) 4754 { 4755 dsl_pool_t *dp; 4756 dsl_dataset_t *ds, *ods; 4757 char origin[ZFS_MAX_DATASET_NAME_LEN]; 4758 char *cp; 4759 int error; 4760 4761 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4762 if (error != 0) 4763 return (error); 4764 4765 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds); 4766 if (error != 0) { 4767 dsl_pool_rele(dp, FTAG); 4768 return (error); 4769 } 4770 4771 if (!dsl_dir_is_clone(ds->ds_dir)) { 4772 dsl_dataset_rele(ds, FTAG); 4773 dsl_pool_rele(dp, FTAG); 4774 return (SET_ERROR(EINVAL)); 4775 } 4776 4777 error = dsl_dataset_hold_obj(dp, 4778 dsl_dir_phys(ds->ds_dir)->dd_origin_obj, FTAG, &ods); 4779 if (error != 0) { 4780 dsl_dataset_rele(ds, FTAG); 4781 dsl_pool_rele(dp, FTAG); 4782 return (error); 4783 } 4784 4785 dsl_dataset_name(ods, origin); 4786 dsl_dataset_rele(ods, FTAG); 4787 dsl_dataset_rele(ds, FTAG); 4788 dsl_pool_rele(dp, FTAG); 4789 4790 /* 4791 * We don't need to unmount *all* the origin fs's snapshots, but 4792 * it's easier. 4793 */ 4794 cp = strchr(origin, '@'); 4795 if (cp) 4796 *cp = '\0'; 4797 (void) dmu_objset_find(origin, 4798 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS); 4799 return (dsl_dataset_promote(zc->zc_name, zc->zc_string)); 4800 } 4801 4802 /* 4803 * Retrieve a single {user|group}{used|quota}@... property. 4804 * 4805 * inputs: 4806 * zc_name name of filesystem 4807 * zc_objset_type zfs_userquota_prop_t 4808 * zc_value domain name (eg. "S-1-234-567-89") 4809 * zc_guid RID/UID/GID 4810 * 4811 * outputs: 4812 * zc_cookie property value 4813 */ 4814 static int 4815 zfs_ioc_userspace_one(zfs_cmd_t *zc) 4816 { 4817 zfsvfs_t *zfsvfs; 4818 int error; 4819 4820 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 4821 return (SET_ERROR(EINVAL)); 4822 4823 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE); 4824 if (error != 0) 4825 return (error); 4826 4827 error = zfs_userspace_one(zfsvfs, 4828 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie); 4829 zfsvfs_rele(zfsvfs, FTAG); 4830 4831 return (error); 4832 } 4833 4834 /* 4835 * inputs: 4836 * zc_name name of filesystem 4837 * zc_cookie zap cursor 4838 * zc_objset_type zfs_userquota_prop_t 4839 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist) 4840 * 4841 * outputs: 4842 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t) 4843 * zc_cookie zap cursor 4844 */ 4845 static int 4846 zfs_ioc_userspace_many(zfs_cmd_t *zc) 4847 { 4848 zfsvfs_t *zfsvfs; 4849 int bufsize = zc->zc_nvlist_dst_size; 4850 4851 if (bufsize <= 0) 4852 return (SET_ERROR(ENOMEM)); 4853 4854 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE); 4855 if (error != 0) 4856 return (error); 4857 4858 void *buf = kmem_alloc(bufsize, KM_SLEEP); 4859 4860 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie, 4861 buf, &zc->zc_nvlist_dst_size); 4862 4863 if (error == 0) { 4864 error = xcopyout(buf, 4865 (void *)(uintptr_t)zc->zc_nvlist_dst, 4866 zc->zc_nvlist_dst_size); 4867 } 4868 kmem_free(buf, bufsize); 4869 zfsvfs_rele(zfsvfs, FTAG); 4870 4871 return (error); 4872 } 4873 4874 /* 4875 * inputs: 4876 * zc_name name of filesystem 4877 * 4878 * outputs: 4879 * none 4880 */ 4881 static int 4882 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc) 4883 { 4884 objset_t *os; 4885 int error = 0; 4886 zfsvfs_t *zfsvfs; 4887 4888 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) { 4889 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) { 4890 /* 4891 * If userused is not enabled, it may be because the 4892 * objset needs to be closed & reopened (to grow the 4893 * objset_phys_t). Suspend/resume the fs will do that. 4894 */ 4895 dsl_dataset_t *ds; 4896 4897 ds = dmu_objset_ds(zfsvfs->z_os); 4898 error = zfs_suspend_fs(zfsvfs); 4899 if (error == 0) { 4900 dmu_objset_refresh_ownership(zfsvfs->z_os, 4901 zfsvfs); 4902 error = zfs_resume_fs(zfsvfs, ds); 4903 } 4904 } 4905 if (error == 0) 4906 error = dmu_objset_userspace_upgrade(zfsvfs->z_os); 4907 VFS_RELE(zfsvfs->z_vfs); 4908 } else { 4909 /* XXX kind of reading contents without owning */ 4910 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 4911 if (error != 0) 4912 return (error); 4913 4914 error = dmu_objset_userspace_upgrade(os); 4915 dmu_objset_rele(os, FTAG); 4916 } 4917 4918 return (error); 4919 } 4920 4921 /* 4922 * We don't want to have a hard dependency 4923 * against some special symbols in sharefs 4924 * nfs, and smbsrv. Determine them if needed when 4925 * the first file system is shared. 4926 * Neither sharefs, nfs or smbsrv are unloadable modules. 4927 */ 4928 int (*znfsexport_fs)(void *arg); 4929 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t); 4930 int (*zsmbexport_fs)(void *arg, boolean_t add_share); 4931 4932 int zfs_nfsshare_inited; 4933 int zfs_smbshare_inited; 4934 4935 ddi_modhandle_t nfs_mod; 4936 ddi_modhandle_t sharefs_mod; 4937 ddi_modhandle_t smbsrv_mod; 4938 kmutex_t zfs_share_lock; 4939 4940 static int 4941 zfs_init_sharefs() 4942 { 4943 int error; 4944 4945 ASSERT(MUTEX_HELD(&zfs_share_lock)); 4946 /* Both NFS and SMB shares also require sharetab support. */ 4947 if (sharefs_mod == NULL && ((sharefs_mod = 4948 ddi_modopen("fs/sharefs", 4949 KRTLD_MODE_FIRST, &error)) == NULL)) { 4950 return (SET_ERROR(ENOSYS)); 4951 } 4952 if (zshare_fs == NULL && ((zshare_fs = 4953 (int (*)(enum sharefs_sys_op, share_t *, uint32_t)) 4954 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) { 4955 return (SET_ERROR(ENOSYS)); 4956 } 4957 return (0); 4958 } 4959 4960 static int 4961 zfs_ioc_share(zfs_cmd_t *zc) 4962 { 4963 int error; 4964 int opcode; 4965 4966 switch (zc->zc_share.z_sharetype) { 4967 case ZFS_SHARE_NFS: 4968 case ZFS_UNSHARE_NFS: 4969 if (zfs_nfsshare_inited == 0) { 4970 mutex_enter(&zfs_share_lock); 4971 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs", 4972 KRTLD_MODE_FIRST, &error)) == NULL)) { 4973 mutex_exit(&zfs_share_lock); 4974 return (SET_ERROR(ENOSYS)); 4975 } 4976 if (znfsexport_fs == NULL && 4977 ((znfsexport_fs = (int (*)(void *)) 4978 ddi_modsym(nfs_mod, 4979 "nfs_export", &error)) == NULL)) { 4980 mutex_exit(&zfs_share_lock); 4981 return (SET_ERROR(ENOSYS)); 4982 } 4983 error = zfs_init_sharefs(); 4984 if (error != 0) { 4985 mutex_exit(&zfs_share_lock); 4986 return (SET_ERROR(ENOSYS)); 4987 } 4988 zfs_nfsshare_inited = 1; 4989 mutex_exit(&zfs_share_lock); 4990 } 4991 break; 4992 case ZFS_SHARE_SMB: 4993 case ZFS_UNSHARE_SMB: 4994 if (zfs_smbshare_inited == 0) { 4995 mutex_enter(&zfs_share_lock); 4996 if (smbsrv_mod == NULL && ((smbsrv_mod = 4997 ddi_modopen("drv/smbsrv", 4998 KRTLD_MODE_FIRST, &error)) == NULL)) { 4999 mutex_exit(&zfs_share_lock); 5000 return (SET_ERROR(ENOSYS)); 5001 } 5002 if (zsmbexport_fs == NULL && ((zsmbexport_fs = 5003 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod, 5004 "smb_server_share", &error)) == NULL)) { 5005 mutex_exit(&zfs_share_lock); 5006 return (SET_ERROR(ENOSYS)); 5007 } 5008 error = zfs_init_sharefs(); 5009 if (error != 0) { 5010 mutex_exit(&zfs_share_lock); 5011 return (SET_ERROR(ENOSYS)); 5012 } 5013 zfs_smbshare_inited = 1; 5014 mutex_exit(&zfs_share_lock); 5015 } 5016 break; 5017 default: 5018 return (SET_ERROR(EINVAL)); 5019 } 5020 5021 switch (zc->zc_share.z_sharetype) { 5022 case ZFS_SHARE_NFS: 5023 case ZFS_UNSHARE_NFS: 5024 if (error = 5025 znfsexport_fs((void *) 5026 (uintptr_t)zc->zc_share.z_exportdata)) 5027 return (error); 5028 break; 5029 case ZFS_SHARE_SMB: 5030 case ZFS_UNSHARE_SMB: 5031 if (error = zsmbexport_fs((void *) 5032 (uintptr_t)zc->zc_share.z_exportdata, 5033 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ? 5034 B_TRUE: B_FALSE)) { 5035 return (error); 5036 } 5037 break; 5038 } 5039 5040 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS || 5041 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ? 5042 SHAREFS_ADD : SHAREFS_REMOVE; 5043 5044 /* 5045 * Add or remove share from sharetab 5046 */ 5047 error = zshare_fs(opcode, 5048 (void *)(uintptr_t)zc->zc_share.z_sharedata, 5049 zc->zc_share.z_sharemax); 5050 5051 return (error); 5052 5053 } 5054 5055 ace_t full_access[] = { 5056 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0} 5057 }; 5058 5059 /* 5060 * inputs: 5061 * zc_name name of containing filesystem 5062 * zc_obj object # beyond which we want next in-use object # 5063 * 5064 * outputs: 5065 * zc_obj next in-use object # 5066 */ 5067 static int 5068 zfs_ioc_next_obj(zfs_cmd_t *zc) 5069 { 5070 objset_t *os = NULL; 5071 int error; 5072 5073 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 5074 if (error != 0) 5075 return (error); 5076 5077 error = dmu_object_next(os, &zc->zc_obj, B_FALSE, 5078 dsl_dataset_phys(os->os_dsl_dataset)->ds_prev_snap_txg); 5079 5080 dmu_objset_rele(os, FTAG); 5081 return (error); 5082 } 5083 5084 /* 5085 * inputs: 5086 * zc_name name of filesystem 5087 * zc_value prefix name for snapshot 5088 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process 5089 * 5090 * outputs: 5091 * zc_value short name of new snapshot 5092 */ 5093 static int 5094 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc) 5095 { 5096 char *snap_name; 5097 char *hold_name; 5098 int error; 5099 minor_t minor; 5100 5101 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor); 5102 if (error != 0) 5103 return (error); 5104 5105 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value, 5106 (u_longlong_t)ddi_get_lbolt64()); 5107 hold_name = kmem_asprintf("%%%s", zc->zc_value); 5108 5109 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor, 5110 hold_name); 5111 if (error == 0) 5112 (void) strcpy(zc->zc_value, snap_name); 5113 strfree(snap_name); 5114 strfree(hold_name); 5115 zfs_onexit_fd_rele(zc->zc_cleanup_fd); 5116 return (error); 5117 } 5118 5119 /* 5120 * inputs: 5121 * zc_name name of "to" snapshot 5122 * zc_value name of "from" snapshot 5123 * zc_cookie file descriptor to write diff data on 5124 * 5125 * outputs: 5126 * dmu_diff_record_t's to the file descriptor 5127 */ 5128 static int 5129 zfs_ioc_diff(zfs_cmd_t *zc) 5130 { 5131 file_t *fp; 5132 offset_t off; 5133 int error; 5134 5135 fp = getf(zc->zc_cookie); 5136 if (fp == NULL) 5137 return (SET_ERROR(EBADF)); 5138 5139 off = fp->f_offset; 5140 5141 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off); 5142 5143 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0) 5144 fp->f_offset = off; 5145 releasef(zc->zc_cookie); 5146 5147 return (error); 5148 } 5149 5150 /* 5151 * Remove all ACL files in shares dir 5152 */ 5153 static int 5154 zfs_smb_acl_purge(znode_t *dzp) 5155 { 5156 zap_cursor_t zc; 5157 zap_attribute_t zap; 5158 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 5159 int error; 5160 5161 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id); 5162 (error = zap_cursor_retrieve(&zc, &zap)) == 0; 5163 zap_cursor_advance(&zc)) { 5164 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred, 5165 NULL, 0)) != 0) 5166 break; 5167 } 5168 zap_cursor_fini(&zc); 5169 return (error); 5170 } 5171 5172 static int 5173 zfs_ioc_smb_acl(zfs_cmd_t *zc) 5174 { 5175 vnode_t *vp; 5176 znode_t *dzp; 5177 vnode_t *resourcevp = NULL; 5178 znode_t *sharedir; 5179 zfsvfs_t *zfsvfs; 5180 nvlist_t *nvlist; 5181 char *src, *target; 5182 vattr_t vattr; 5183 vsecattr_t vsec; 5184 int error = 0; 5185 5186 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, 5187 NO_FOLLOW, NULL, &vp)) != 0) 5188 return (error); 5189 5190 /* Now make sure mntpnt and dataset are ZFS */ 5191 5192 if (vp->v_vfsp->vfs_fstype != zfsfstype || 5193 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), 5194 zc->zc_name) != 0)) { 5195 VN_RELE(vp); 5196 return (SET_ERROR(EINVAL)); 5197 } 5198 5199 dzp = VTOZ(vp); 5200 zfsvfs = dzp->z_zfsvfs; 5201 ZFS_ENTER(zfsvfs); 5202 5203 /* 5204 * Create share dir if its missing. 5205 */ 5206 mutex_enter(&zfsvfs->z_lock); 5207 if (zfsvfs->z_shares_dir == 0) { 5208 dmu_tx_t *tx; 5209 5210 tx = dmu_tx_create(zfsvfs->z_os); 5211 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE, 5212 ZFS_SHARES_DIR); 5213 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); 5214 error = dmu_tx_assign(tx, TXG_WAIT); 5215 if (error != 0) { 5216 dmu_tx_abort(tx); 5217 } else { 5218 error = zfs_create_share_dir(zfsvfs, tx); 5219 dmu_tx_commit(tx); 5220 } 5221 if (error != 0) { 5222 mutex_exit(&zfsvfs->z_lock); 5223 VN_RELE(vp); 5224 ZFS_EXIT(zfsvfs); 5225 return (error); 5226 } 5227 } 5228 mutex_exit(&zfsvfs->z_lock); 5229 5230 ASSERT(zfsvfs->z_shares_dir); 5231 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) { 5232 VN_RELE(vp); 5233 ZFS_EXIT(zfsvfs); 5234 return (error); 5235 } 5236 5237 switch (zc->zc_cookie) { 5238 case ZFS_SMB_ACL_ADD: 5239 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; 5240 vattr.va_type = VREG; 5241 vattr.va_mode = S_IFREG|0777; 5242 vattr.va_uid = 0; 5243 vattr.va_gid = 0; 5244 5245 vsec.vsa_mask = VSA_ACE; 5246 vsec.vsa_aclentp = &full_access; 5247 vsec.vsa_aclentsz = sizeof (full_access); 5248 vsec.vsa_aclcnt = 1; 5249 5250 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string, 5251 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec); 5252 if (resourcevp) 5253 VN_RELE(resourcevp); 5254 break; 5255 5256 case ZFS_SMB_ACL_REMOVE: 5257 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred, 5258 NULL, 0); 5259 break; 5260 5261 case ZFS_SMB_ACL_RENAME: 5262 if ((error = get_nvlist(zc->zc_nvlist_src, 5263 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) { 5264 VN_RELE(vp); 5265 VN_RELE(ZTOV(sharedir)); 5266 ZFS_EXIT(zfsvfs); 5267 return (error); 5268 } 5269 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) || 5270 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET, 5271 &target)) { 5272 VN_RELE(vp); 5273 VN_RELE(ZTOV(sharedir)); 5274 ZFS_EXIT(zfsvfs); 5275 nvlist_free(nvlist); 5276 return (error); 5277 } 5278 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target, 5279 kcred, NULL, 0); 5280 nvlist_free(nvlist); 5281 break; 5282 5283 case ZFS_SMB_ACL_PURGE: 5284 error = zfs_smb_acl_purge(sharedir); 5285 break; 5286 5287 default: 5288 error = SET_ERROR(EINVAL); 5289 break; 5290 } 5291 5292 VN_RELE(vp); 5293 VN_RELE(ZTOV(sharedir)); 5294 5295 ZFS_EXIT(zfsvfs); 5296 5297 return (error); 5298 } 5299 5300 /* 5301 * innvl: { 5302 * "holds" -> { snapname -> holdname (string), ... } 5303 * (optional) "cleanup_fd" -> fd (int32) 5304 * } 5305 * 5306 * outnvl: { 5307 * snapname -> error value (int32) 5308 * ... 5309 * } 5310 */ 5311 /* ARGSUSED */ 5312 static int 5313 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist) 5314 { 5315 nvpair_t *pair; 5316 nvlist_t *holds; 5317 int cleanup_fd = -1; 5318 int error; 5319 minor_t minor = 0; 5320 5321 error = nvlist_lookup_nvlist(args, "holds", &holds); 5322 if (error != 0) 5323 return (SET_ERROR(EINVAL)); 5324 5325 /* make sure the user didn't pass us any invalid (empty) tags */ 5326 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL; 5327 pair = nvlist_next_nvpair(holds, pair)) { 5328 char *htag; 5329 5330 error = nvpair_value_string(pair, &htag); 5331 if (error != 0) 5332 return (SET_ERROR(error)); 5333 5334 if (strlen(htag) == 0) 5335 return (SET_ERROR(EINVAL)); 5336 } 5337 5338 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) { 5339 error = zfs_onexit_fd_hold(cleanup_fd, &minor); 5340 if (error != 0) 5341 return (error); 5342 } 5343 5344 error = dsl_dataset_user_hold(holds, minor, errlist); 5345 if (minor != 0) 5346 zfs_onexit_fd_rele(cleanup_fd); 5347 return (error); 5348 } 5349 5350 /* 5351 * innvl is not used. 5352 * 5353 * outnvl: { 5354 * holdname -> time added (uint64 seconds since epoch) 5355 * ... 5356 * } 5357 */ 5358 /* ARGSUSED */ 5359 static int 5360 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl) 5361 { 5362 return (dsl_dataset_get_holds(snapname, outnvl)); 5363 } 5364 5365 /* 5366 * innvl: { 5367 * snapname -> { holdname, ... } 5368 * ... 5369 * } 5370 * 5371 * outnvl: { 5372 * snapname -> error value (int32) 5373 * ... 5374 * } 5375 */ 5376 /* ARGSUSED */ 5377 static int 5378 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist) 5379 { 5380 return (dsl_dataset_user_release(holds, errlist)); 5381 } 5382 5383 /* 5384 * inputs: 5385 * zc_name name of new filesystem or snapshot 5386 * zc_value full name of old snapshot 5387 * 5388 * outputs: 5389 * zc_cookie space in bytes 5390 * zc_objset_type compressed space in bytes 5391 * zc_perm_action uncompressed space in bytes 5392 */ 5393 static int 5394 zfs_ioc_space_written(zfs_cmd_t *zc) 5395 { 5396 int error; 5397 dsl_pool_t *dp; 5398 dsl_dataset_t *new, *old; 5399 5400 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 5401 if (error != 0) 5402 return (error); 5403 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new); 5404 if (error != 0) { 5405 dsl_pool_rele(dp, FTAG); 5406 return (error); 5407 } 5408 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old); 5409 if (error != 0) { 5410 dsl_dataset_rele(new, FTAG); 5411 dsl_pool_rele(dp, FTAG); 5412 return (error); 5413 } 5414 5415 error = dsl_dataset_space_written(old, new, &zc->zc_cookie, 5416 &zc->zc_objset_type, &zc->zc_perm_action); 5417 dsl_dataset_rele(old, FTAG); 5418 dsl_dataset_rele(new, FTAG); 5419 dsl_pool_rele(dp, FTAG); 5420 return (error); 5421 } 5422 5423 /* 5424 * innvl: { 5425 * "firstsnap" -> snapshot name 5426 * } 5427 * 5428 * outnvl: { 5429 * "used" -> space in bytes 5430 * "compressed" -> compressed space in bytes 5431 * "uncompressed" -> uncompressed space in bytes 5432 * } 5433 */ 5434 static int 5435 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl) 5436 { 5437 int error; 5438 dsl_pool_t *dp; 5439 dsl_dataset_t *new, *old; 5440 char *firstsnap; 5441 uint64_t used, comp, uncomp; 5442 5443 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0) 5444 return (SET_ERROR(EINVAL)); 5445 5446 error = dsl_pool_hold(lastsnap, FTAG, &dp); 5447 if (error != 0) 5448 return (error); 5449 5450 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new); 5451 if (error == 0 && !new->ds_is_snapshot) { 5452 dsl_dataset_rele(new, FTAG); 5453 error = SET_ERROR(EINVAL); 5454 } 5455 if (error != 0) { 5456 dsl_pool_rele(dp, FTAG); 5457 return (error); 5458 } 5459 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old); 5460 if (error == 0 && !old->ds_is_snapshot) { 5461 dsl_dataset_rele(old, FTAG); 5462 error = SET_ERROR(EINVAL); 5463 } 5464 if (error != 0) { 5465 dsl_dataset_rele(new, FTAG); 5466 dsl_pool_rele(dp, FTAG); 5467 return (error); 5468 } 5469 5470 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp); 5471 dsl_dataset_rele(old, FTAG); 5472 dsl_dataset_rele(new, FTAG); 5473 dsl_pool_rele(dp, FTAG); 5474 fnvlist_add_uint64(outnvl, "used", used); 5475 fnvlist_add_uint64(outnvl, "compressed", comp); 5476 fnvlist_add_uint64(outnvl, "uncompressed", uncomp); 5477 return (error); 5478 } 5479 5480 /* 5481 * innvl: { 5482 * "fd" -> file descriptor to write stream to (int32) 5483 * (optional) "fromsnap" -> full snap name to send an incremental from 5484 * (optional) "largeblockok" -> (value ignored) 5485 * indicates that blocks > 128KB are permitted 5486 * (optional) "embedok" -> (value ignored) 5487 * presence indicates DRR_WRITE_EMBEDDED records are permitted 5488 * (optional) "compressok" -> (value ignored) 5489 * presence indicates compressed DRR_WRITE records are permitted 5490 * (optional) "resume_object" and "resume_offset" -> (uint64) 5491 * if present, resume send stream from specified object and offset. 5492 * } 5493 * 5494 * outnvl is unused 5495 */ 5496 /* ARGSUSED */ 5497 static int 5498 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl) 5499 { 5500 int error; 5501 offset_t off; 5502 char *fromname = NULL; 5503 int fd; 5504 boolean_t largeblockok; 5505 boolean_t embedok; 5506 boolean_t compressok; 5507 uint64_t resumeobj = 0; 5508 uint64_t resumeoff = 0; 5509 5510 error = nvlist_lookup_int32(innvl, "fd", &fd); 5511 if (error != 0) 5512 return (SET_ERROR(EINVAL)); 5513 5514 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname); 5515 5516 largeblockok = nvlist_exists(innvl, "largeblockok"); 5517 embedok = nvlist_exists(innvl, "embedok"); 5518 compressok = nvlist_exists(innvl, "compressok"); 5519 5520 (void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj); 5521 (void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff); 5522 5523 file_t *fp = getf(fd); 5524 if (fp == NULL) 5525 return (SET_ERROR(EBADF)); 5526 5527 off = fp->f_offset; 5528 error = dmu_send(snapname, fromname, embedok, largeblockok, compressok, 5529 fd, resumeobj, resumeoff, fp->f_vnode, &off); 5530 5531 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0) 5532 fp->f_offset = off; 5533 releasef(fd); 5534 return (error); 5535 } 5536 5537 /* 5538 * Determine approximately how large a zfs send stream will be -- the number 5539 * of bytes that will be written to the fd supplied to zfs_ioc_send_new(). 5540 * 5541 * innvl: { 5542 * (optional) "from" -> full snap or bookmark name to send an incremental 5543 * from 5544 * (optional) "largeblockok" -> (value ignored) 5545 * indicates that blocks > 128KB are permitted 5546 * (optional) "embedok" -> (value ignored) 5547 * presence indicates DRR_WRITE_EMBEDDED records are permitted 5548 * (optional) "compressok" -> (value ignored) 5549 * presence indicates compressed DRR_WRITE records are permitted 5550 * } 5551 * 5552 * outnvl: { 5553 * "space" -> bytes of space (uint64) 5554 * } 5555 */ 5556 static int 5557 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl) 5558 { 5559 dsl_pool_t *dp; 5560 dsl_dataset_t *tosnap; 5561 int error; 5562 char *fromname; 5563 /* LINTED E_FUNC_SET_NOT_USED */ 5564 boolean_t largeblockok; 5565 /* LINTED E_FUNC_SET_NOT_USED */ 5566 boolean_t embedok; 5567 boolean_t compressok; 5568 uint64_t space; 5569 5570 error = dsl_pool_hold(snapname, FTAG, &dp); 5571 if (error != 0) 5572 return (error); 5573 5574 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap); 5575 if (error != 0) { 5576 dsl_pool_rele(dp, FTAG); 5577 return (error); 5578 } 5579 5580 largeblockok = nvlist_exists(innvl, "largeblockok"); 5581 embedok = nvlist_exists(innvl, "embedok"); 5582 compressok = nvlist_exists(innvl, "compressok"); 5583 5584 error = nvlist_lookup_string(innvl, "from", &fromname); 5585 if (error == 0) { 5586 if (strchr(fromname, '@') != NULL) { 5587 /* 5588 * If from is a snapshot, hold it and use the more 5589 * efficient dmu_send_estimate to estimate send space 5590 * size using deadlists. 5591 */ 5592 dsl_dataset_t *fromsnap; 5593 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap); 5594 if (error != 0) 5595 goto out; 5596 error = dmu_send_estimate(tosnap, fromsnap, compressok, 5597 &space); 5598 dsl_dataset_rele(fromsnap, FTAG); 5599 } else if (strchr(fromname, '#') != NULL) { 5600 /* 5601 * If from is a bookmark, fetch the creation TXG of the 5602 * snapshot it was created from and use that to find 5603 * blocks that were born after it. 5604 */ 5605 zfs_bookmark_phys_t frombm; 5606 5607 error = dsl_bookmark_lookup(dp, fromname, tosnap, 5608 &frombm); 5609 if (error != 0) 5610 goto out; 5611 error = dmu_send_estimate_from_txg(tosnap, 5612 frombm.zbm_creation_txg, compressok, &space); 5613 } else { 5614 /* 5615 * from is not properly formatted as a snapshot or 5616 * bookmark 5617 */ 5618 error = SET_ERROR(EINVAL); 5619 goto out; 5620 } 5621 } else { 5622 // If estimating the size of a full send, use dmu_send_estimate 5623 error = dmu_send_estimate(tosnap, NULL, compressok, &space); 5624 } 5625 5626 fnvlist_add_uint64(outnvl, "space", space); 5627 5628 out: 5629 dsl_dataset_rele(tosnap, FTAG); 5630 dsl_pool_rele(dp, FTAG); 5631 return (error); 5632 } 5633 5634 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST]; 5635 5636 static void 5637 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5638 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck, 5639 boolean_t log_history, zfs_ioc_poolcheck_t pool_check) 5640 { 5641 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST]; 5642 5643 ASSERT3U(ioc, >=, ZFS_IOC_FIRST); 5644 ASSERT3U(ioc, <, ZFS_IOC_LAST); 5645 ASSERT3P(vec->zvec_legacy_func, ==, NULL); 5646 ASSERT3P(vec->zvec_func, ==, NULL); 5647 5648 vec->zvec_legacy_func = func; 5649 vec->zvec_secpolicy = secpolicy; 5650 vec->zvec_namecheck = namecheck; 5651 vec->zvec_allow_log = log_history; 5652 vec->zvec_pool_check = pool_check; 5653 } 5654 5655 /* 5656 * See the block comment at the beginning of this file for details on 5657 * each argument to this function. 5658 */ 5659 static void 5660 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func, 5661 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck, 5662 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist, 5663 boolean_t allow_log) 5664 { 5665 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST]; 5666 5667 ASSERT3U(ioc, >=, ZFS_IOC_FIRST); 5668 ASSERT3U(ioc, <, ZFS_IOC_LAST); 5669 ASSERT3P(vec->zvec_legacy_func, ==, NULL); 5670 ASSERT3P(vec->zvec_func, ==, NULL); 5671 5672 /* if we are logging, the name must be valid */ 5673 ASSERT(!allow_log || namecheck != NO_NAME); 5674 5675 vec->zvec_name = name; 5676 vec->zvec_func = func; 5677 vec->zvec_secpolicy = secpolicy; 5678 vec->zvec_namecheck = namecheck; 5679 vec->zvec_pool_check = pool_check; 5680 vec->zvec_smush_outnvlist = smush_outnvlist; 5681 vec->zvec_allow_log = allow_log; 5682 } 5683 5684 static void 5685 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5686 zfs_secpolicy_func_t *secpolicy, boolean_t log_history, 5687 zfs_ioc_poolcheck_t pool_check) 5688 { 5689 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5690 POOL_NAME, log_history, pool_check); 5691 } 5692 5693 static void 5694 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5695 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check) 5696 { 5697 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5698 DATASET_NAME, B_FALSE, pool_check); 5699 } 5700 5701 static void 5702 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func) 5703 { 5704 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config, 5705 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5706 } 5707 5708 static void 5709 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5710 zfs_secpolicy_func_t *secpolicy) 5711 { 5712 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5713 NO_NAME, B_FALSE, POOL_CHECK_NONE); 5714 } 5715 5716 static void 5717 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc, 5718 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy) 5719 { 5720 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5721 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED); 5722 } 5723 5724 static void 5725 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func) 5726 { 5727 zfs_ioctl_register_dataset_read_secpolicy(ioc, func, 5728 zfs_secpolicy_read); 5729 } 5730 5731 static void 5732 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5733 zfs_secpolicy_func_t *secpolicy) 5734 { 5735 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5736 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5737 } 5738 5739 static void 5740 zfs_ioctl_init(void) 5741 { 5742 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT, 5743 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME, 5744 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5745 5746 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY, 5747 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME, 5748 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE); 5749 5750 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS, 5751 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME, 5752 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5753 5754 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW, 5755 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME, 5756 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5757 5758 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE, 5759 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME, 5760 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5761 5762 zfs_ioctl_register("create", ZFS_IOC_CREATE, 5763 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME, 5764 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5765 5766 zfs_ioctl_register("clone", ZFS_IOC_CLONE, 5767 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME, 5768 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5769 5770 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS, 5771 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME, 5772 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5773 5774 zfs_ioctl_register("hold", ZFS_IOC_HOLD, 5775 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME, 5776 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5777 zfs_ioctl_register("release", ZFS_IOC_RELEASE, 5778 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME, 5779 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5780 5781 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS, 5782 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME, 5783 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5784 5785 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK, 5786 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, 5787 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE); 5788 5789 zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK, 5790 zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME, 5791 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5792 5793 zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS, 5794 zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME, 5795 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5796 5797 zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS, 5798 zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks, 5799 POOL_NAME, 5800 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5801 5802 zfs_ioctl_register("channel_program", ZFS_IOC_CHANNEL_PROGRAM, 5803 zfs_ioc_channel_program, zfs_secpolicy_config, 5804 POOL_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, 5805 B_TRUE); 5806 5807 /* IOCTLS that use the legacy function signature */ 5808 5809 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze, 5810 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY); 5811 5812 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create, 5813 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5814 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN, 5815 zfs_ioc_pool_scan); 5816 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE, 5817 zfs_ioc_pool_upgrade); 5818 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD, 5819 zfs_ioc_vdev_add); 5820 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE, 5821 zfs_ioc_vdev_remove); 5822 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE, 5823 zfs_ioc_vdev_set_state); 5824 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH, 5825 zfs_ioc_vdev_attach); 5826 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH, 5827 zfs_ioc_vdev_detach); 5828 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH, 5829 zfs_ioc_vdev_setpath); 5830 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU, 5831 zfs_ioc_vdev_setfru); 5832 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS, 5833 zfs_ioc_pool_set_props); 5834 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT, 5835 zfs_ioc_vdev_split); 5836 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID, 5837 zfs_ioc_pool_reguid); 5838 5839 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS, 5840 zfs_ioc_pool_configs, zfs_secpolicy_none); 5841 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT, 5842 zfs_ioc_pool_tryimport, zfs_secpolicy_config); 5843 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT, 5844 zfs_ioc_inject_fault, zfs_secpolicy_inject); 5845 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT, 5846 zfs_ioc_clear_fault, zfs_secpolicy_inject); 5847 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT, 5848 zfs_ioc_inject_list_next, zfs_secpolicy_inject); 5849 5850 /* 5851 * pool destroy, and export don't log the history as part of 5852 * zfsdev_ioctl, but rather zfs_ioc_pool_export 5853 * does the logging of those commands. 5854 */ 5855 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy, 5856 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE); 5857 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export, 5858 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE); 5859 5860 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats, 5861 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE); 5862 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props, 5863 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE); 5864 5865 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log, 5866 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED); 5867 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME, 5868 zfs_ioc_dsobj_to_dsname, 5869 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED); 5870 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY, 5871 zfs_ioc_pool_get_history, 5872 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED); 5873 5874 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import, 5875 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5876 5877 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear, 5878 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5879 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen, 5880 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED); 5881 5882 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN, 5883 zfs_ioc_space_written); 5884 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS, 5885 zfs_ioc_objset_recvd_props); 5886 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ, 5887 zfs_ioc_next_obj); 5888 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL, 5889 zfs_ioc_get_fsacl); 5890 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS, 5891 zfs_ioc_objset_stats); 5892 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS, 5893 zfs_ioc_objset_zplprops); 5894 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT, 5895 zfs_ioc_dataset_list_next); 5896 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT, 5897 zfs_ioc_snapshot_list_next); 5898 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS, 5899 zfs_ioc_send_progress); 5900 5901 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF, 5902 zfs_ioc_diff, zfs_secpolicy_diff); 5903 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS, 5904 zfs_ioc_obj_to_stats, zfs_secpolicy_diff); 5905 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH, 5906 zfs_ioc_obj_to_path, zfs_secpolicy_diff); 5907 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE, 5908 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one); 5909 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY, 5910 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many); 5911 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND, 5912 zfs_ioc_send, zfs_secpolicy_send); 5913 5914 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop, 5915 zfs_secpolicy_none); 5916 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy, 5917 zfs_secpolicy_destroy); 5918 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename, 5919 zfs_secpolicy_rename); 5920 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv, 5921 zfs_secpolicy_recv); 5922 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote, 5923 zfs_secpolicy_promote); 5924 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP, 5925 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop); 5926 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl, 5927 zfs_secpolicy_set_fsacl); 5928 5929 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share, 5930 zfs_secpolicy_share, POOL_CHECK_NONE); 5931 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl, 5932 zfs_secpolicy_smb_acl, POOL_CHECK_NONE); 5933 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE, 5934 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade, 5935 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5936 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT, 5937 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot, 5938 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5939 } 5940 5941 int 5942 pool_status_check(const char *name, zfs_ioc_namecheck_t type, 5943 zfs_ioc_poolcheck_t check) 5944 { 5945 spa_t *spa; 5946 int error; 5947 5948 ASSERT(type == POOL_NAME || type == DATASET_NAME); 5949 5950 if (check & POOL_CHECK_NONE) 5951 return (0); 5952 5953 error = spa_open(name, &spa, FTAG); 5954 if (error == 0) { 5955 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa)) 5956 error = SET_ERROR(EAGAIN); 5957 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa)) 5958 error = SET_ERROR(EROFS); 5959 spa_close(spa, FTAG); 5960 } 5961 return (error); 5962 } 5963 5964 /* 5965 * Find a free minor number. 5966 */ 5967 minor_t 5968 zfsdev_minor_alloc(void) 5969 { 5970 static minor_t last_minor; 5971 minor_t m; 5972 5973 ASSERT(MUTEX_HELD(&zfsdev_state_lock)); 5974 5975 for (m = last_minor + 1; m != last_minor; m++) { 5976 if (m > ZFSDEV_MAX_MINOR) 5977 m = 1; 5978 if (ddi_get_soft_state(zfsdev_state, m) == NULL) { 5979 last_minor = m; 5980 return (m); 5981 } 5982 } 5983 5984 return (0); 5985 } 5986 5987 static int 5988 zfs_ctldev_init(dev_t *devp) 5989 { 5990 minor_t minor; 5991 zfs_soft_state_t *zs; 5992 5993 ASSERT(MUTEX_HELD(&zfsdev_state_lock)); 5994 ASSERT(getminor(*devp) == 0); 5995 5996 minor = zfsdev_minor_alloc(); 5997 if (minor == 0) 5998 return (SET_ERROR(ENXIO)); 5999 6000 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS) 6001 return (SET_ERROR(EAGAIN)); 6002 6003 *devp = makedevice(getemajor(*devp), minor); 6004 6005 zs = ddi_get_soft_state(zfsdev_state, minor); 6006 zs->zss_type = ZSST_CTLDEV; 6007 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data); 6008 6009 return (0); 6010 } 6011 6012 static void 6013 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor) 6014 { 6015 ASSERT(MUTEX_HELD(&zfsdev_state_lock)); 6016 6017 zfs_onexit_destroy(zo); 6018 ddi_soft_state_free(zfsdev_state, minor); 6019 } 6020 6021 void * 6022 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which) 6023 { 6024 zfs_soft_state_t *zp; 6025 6026 zp = ddi_get_soft_state(zfsdev_state, minor); 6027 if (zp == NULL || zp->zss_type != which) 6028 return (NULL); 6029 6030 return (zp->zss_data); 6031 } 6032 6033 static int 6034 zfsdev_open(dev_t *devp, int flag, int otyp, cred_t *cr) 6035 { 6036 int error = 0; 6037 6038 if (getminor(*devp) != 0) 6039 return (zvol_open(devp, flag, otyp, cr)); 6040 6041 /* This is the control device. Allocate a new minor if requested. */ 6042 if (flag & FEXCL) { 6043 mutex_enter(&zfsdev_state_lock); 6044 error = zfs_ctldev_init(devp); 6045 mutex_exit(&zfsdev_state_lock); 6046 } 6047 6048 return (error); 6049 } 6050 6051 static int 6052 zfsdev_close(dev_t dev, int flag, int otyp, cred_t *cr) 6053 { 6054 zfs_onexit_t *zo; 6055 minor_t minor = getminor(dev); 6056 6057 if (minor == 0) 6058 return (0); 6059 6060 mutex_enter(&zfsdev_state_lock); 6061 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV); 6062 if (zo == NULL) { 6063 mutex_exit(&zfsdev_state_lock); 6064 return (zvol_close(dev, flag, otyp, cr)); 6065 } 6066 zfs_ctldev_destroy(zo, minor); 6067 mutex_exit(&zfsdev_state_lock); 6068 6069 return (0); 6070 } 6071 6072 static int 6073 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp) 6074 { 6075 zfs_cmd_t *zc; 6076 uint_t vecnum; 6077 int error, rc, len; 6078 minor_t minor = getminor(dev); 6079 const zfs_ioc_vec_t *vec; 6080 char *saved_poolname = NULL; 6081 nvlist_t *innvl = NULL; 6082 6083 if (minor != 0 && 6084 zfsdev_get_soft_state(minor, ZSST_CTLDEV) == NULL) 6085 return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp)); 6086 6087 vecnum = cmd - ZFS_IOC_FIRST; 6088 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip)); 6089 6090 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0])) 6091 return (SET_ERROR(EINVAL)); 6092 vec = &zfs_ioc_vec[vecnum]; 6093 6094 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP); 6095 6096 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag); 6097 if (error != 0) { 6098 error = SET_ERROR(EFAULT); 6099 goto out; 6100 } 6101 6102 zc->zc_iflags = flag & FKIOCTL; 6103 if (zc->zc_nvlist_src_size != 0) { 6104 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 6105 zc->zc_iflags, &innvl); 6106 if (error != 0) 6107 goto out; 6108 } 6109 6110 /* 6111 * Ensure that all pool/dataset names are valid before we pass down to 6112 * the lower layers. 6113 */ 6114 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0'; 6115 switch (vec->zvec_namecheck) { 6116 case POOL_NAME: 6117 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0) 6118 error = SET_ERROR(EINVAL); 6119 else 6120 error = pool_status_check(zc->zc_name, 6121 vec->zvec_namecheck, vec->zvec_pool_check); 6122 break; 6123 6124 case DATASET_NAME: 6125 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0) 6126 error = SET_ERROR(EINVAL); 6127 else 6128 error = pool_status_check(zc->zc_name, 6129 vec->zvec_namecheck, vec->zvec_pool_check); 6130 break; 6131 6132 case NO_NAME: 6133 break; 6134 } 6135 6136 6137 if (error == 0) 6138 error = vec->zvec_secpolicy(zc, innvl, cr); 6139 6140 if (error != 0) 6141 goto out; 6142 6143 /* legacy ioctls can modify zc_name */ 6144 len = strcspn(zc->zc_name, "/@#") + 1; 6145 saved_poolname = kmem_alloc(len, KM_SLEEP); 6146 (void) strlcpy(saved_poolname, zc->zc_name, len); 6147 6148 if (vec->zvec_func != NULL) { 6149 nvlist_t *outnvl; 6150 int puterror = 0; 6151 spa_t *spa; 6152 nvlist_t *lognv = NULL; 6153 6154 ASSERT(vec->zvec_legacy_func == NULL); 6155 6156 /* 6157 * Add the innvl to the lognv before calling the func, 6158 * in case the func changes the innvl. 6159 */ 6160 if (vec->zvec_allow_log) { 6161 lognv = fnvlist_alloc(); 6162 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL, 6163 vec->zvec_name); 6164 if (!nvlist_empty(innvl)) { 6165 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL, 6166 innvl); 6167 } 6168 } 6169 6170 outnvl = fnvlist_alloc(); 6171 error = vec->zvec_func(zc->zc_name, innvl, outnvl); 6172 6173 /* 6174 * Some commands can partially execute, modfiy state, and still 6175 * return an error. In these cases, attempt to record what 6176 * was modified. 6177 */ 6178 if ((error == 0 || 6179 (cmd == ZFS_IOC_CHANNEL_PROGRAM && error != EINVAL)) && 6180 vec->zvec_allow_log && 6181 spa_open(zc->zc_name, &spa, FTAG) == 0) { 6182 if (!nvlist_empty(outnvl)) { 6183 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL, 6184 outnvl); 6185 } 6186 if (error != 0) { 6187 fnvlist_add_int64(lognv, ZPOOL_HIST_ERRNO, 6188 error); 6189 } 6190 (void) spa_history_log_nvl(spa, lognv); 6191 spa_close(spa, FTAG); 6192 } 6193 fnvlist_free(lognv); 6194 6195 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) { 6196 int smusherror = 0; 6197 if (vec->zvec_smush_outnvlist) { 6198 smusherror = nvlist_smush(outnvl, 6199 zc->zc_nvlist_dst_size); 6200 } 6201 if (smusherror == 0) 6202 puterror = put_nvlist(zc, outnvl); 6203 } 6204 6205 if (puterror != 0) 6206 error = puterror; 6207 6208 nvlist_free(outnvl); 6209 } else { 6210 error = vec->zvec_legacy_func(zc); 6211 } 6212 6213 out: 6214 nvlist_free(innvl); 6215 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag); 6216 if (error == 0 && rc != 0) 6217 error = SET_ERROR(EFAULT); 6218 if (error == 0 && vec->zvec_allow_log) { 6219 char *s = tsd_get(zfs_allow_log_key); 6220 if (s != NULL) 6221 strfree(s); 6222 (void) tsd_set(zfs_allow_log_key, saved_poolname); 6223 } else { 6224 if (saved_poolname != NULL) 6225 strfree(saved_poolname); 6226 } 6227 6228 kmem_free(zc, sizeof (zfs_cmd_t)); 6229 return (error); 6230 } 6231 6232 static int 6233 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd) 6234 { 6235 if (cmd != DDI_ATTACH) 6236 return (DDI_FAILURE); 6237 6238 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0, 6239 DDI_PSEUDO, 0) == DDI_FAILURE) 6240 return (DDI_FAILURE); 6241 6242 zfs_dip = dip; 6243 6244 ddi_report_dev(dip); 6245 6246 return (DDI_SUCCESS); 6247 } 6248 6249 static int 6250 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd) 6251 { 6252 if (spa_busy() || zfs_busy() || zvol_busy()) 6253 return (DDI_FAILURE); 6254 6255 if (cmd != DDI_DETACH) 6256 return (DDI_FAILURE); 6257 6258 zfs_dip = NULL; 6259 6260 ddi_prop_remove_all(dip); 6261 ddi_remove_minor_node(dip, NULL); 6262 6263 return (DDI_SUCCESS); 6264 } 6265 6266 /*ARGSUSED*/ 6267 static int 6268 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result) 6269 { 6270 switch (infocmd) { 6271 case DDI_INFO_DEVT2DEVINFO: 6272 *result = zfs_dip; 6273 return (DDI_SUCCESS); 6274 6275 case DDI_INFO_DEVT2INSTANCE: 6276 *result = (void *)0; 6277 return (DDI_SUCCESS); 6278 } 6279 6280 return (DDI_FAILURE); 6281 } 6282 6283 /* 6284 * OK, so this is a little weird. 6285 * 6286 * /dev/zfs is the control node, i.e. minor 0. 6287 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0. 6288 * 6289 * /dev/zfs has basically nothing to do except serve up ioctls, 6290 * so most of the standard driver entry points are in zvol.c. 6291 */ 6292 static struct cb_ops zfs_cb_ops = { 6293 zfsdev_open, /* open */ 6294 zfsdev_close, /* close */ 6295 zvol_strategy, /* strategy */ 6296 nodev, /* print */ 6297 zvol_dump, /* dump */ 6298 zvol_read, /* read */ 6299 zvol_write, /* write */ 6300 zfsdev_ioctl, /* ioctl */ 6301 nodev, /* devmap */ 6302 nodev, /* mmap */ 6303 nodev, /* segmap */ 6304 nochpoll, /* poll */ 6305 ddi_prop_op, /* prop_op */ 6306 NULL, /* streamtab */ 6307 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */ 6308 CB_REV, /* version */ 6309 nodev, /* async read */ 6310 nodev, /* async write */ 6311 }; 6312 6313 static struct dev_ops zfs_dev_ops = { 6314 DEVO_REV, /* version */ 6315 0, /* refcnt */ 6316 zfs_info, /* info */ 6317 nulldev, /* identify */ 6318 nulldev, /* probe */ 6319 zfs_attach, /* attach */ 6320 zfs_detach, /* detach */ 6321 nodev, /* reset */ 6322 &zfs_cb_ops, /* driver operations */ 6323 NULL, /* no bus operations */ 6324 NULL, /* power */ 6325 ddi_quiesce_not_needed, /* quiesce */ 6326 }; 6327 6328 static struct modldrv zfs_modldrv = { 6329 &mod_driverops, 6330 "ZFS storage pool", 6331 &zfs_dev_ops 6332 }; 6333 6334 static struct modlinkage modlinkage = { 6335 MODREV_1, 6336 (void *)&zfs_modlfs, 6337 (void *)&zfs_modldrv, 6338 NULL 6339 }; 6340 6341 static void 6342 zfs_allow_log_destroy(void *arg) 6343 { 6344 char *poolname = arg; 6345 strfree(poolname); 6346 } 6347 6348 int 6349 _init(void) 6350 { 6351 int error; 6352 6353 spa_init(FREAD | FWRITE); 6354 zfs_init(); 6355 zvol_init(); 6356 zfs_ioctl_init(); 6357 6358 if ((error = mod_install(&modlinkage)) != 0) { 6359 zvol_fini(); 6360 zfs_fini(); 6361 spa_fini(); 6362 return (error); 6363 } 6364 6365 tsd_create(&zfs_fsyncer_key, NULL); 6366 tsd_create(&rrw_tsd_key, rrw_tsd_destroy); 6367 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy); 6368 6369 error = ldi_ident_from_mod(&modlinkage, &zfs_li); 6370 ASSERT(error == 0); 6371 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL); 6372 6373 return (0); 6374 } 6375 6376 int 6377 _fini(void) 6378 { 6379 int error; 6380 6381 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled) 6382 return (SET_ERROR(EBUSY)); 6383 6384 if ((error = mod_remove(&modlinkage)) != 0) 6385 return (error); 6386 6387 zvol_fini(); 6388 zfs_fini(); 6389 spa_fini(); 6390 if (zfs_nfsshare_inited) 6391 (void) ddi_modclose(nfs_mod); 6392 if (zfs_smbshare_inited) 6393 (void) ddi_modclose(smbsrv_mod); 6394 if (zfs_nfsshare_inited || zfs_smbshare_inited) 6395 (void) ddi_modclose(sharefs_mod); 6396 6397 tsd_destroy(&zfs_fsyncer_key); 6398 ldi_ident_release(zfs_li); 6399 zfs_li = NULL; 6400 mutex_destroy(&zfs_share_lock); 6401 6402 return (error); 6403 } 6404 6405 int 6406 _info(struct modinfo *modinfop) 6407 { 6408 return (mod_info(&modlinkage, modinfop)); 6409 } 6410