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 */ 25 26 /* 27 * Copyright (c) 2011-2012 Pawel Jakub Dawidek. All rights reserved. 28 * Portions Copyright 2011 Martin Matuska 29 * Copyright 2015, OmniTI Computer Consulting, Inc. All rights reserved. 30 * Copyright 2018 Nexenta Systems, Inc. All rights reserved. 31 * Copyright 2019 Joyent, Inc. 32 * Copyright (c) 2011, 2017 by Delphix. All rights reserved. 33 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved. 34 * Copyright (c) 2013 Steven Hartland. All rights reserved. 35 * Copyright (c) 2014 Integros [integros.com] 36 * Copyright 2016 Toomas Soome <tsoome@me.com> 37 * Copyright (c) 2017, loli10K <ezomori.nozomu@gmail.com>. All rights reserved. 38 * Copyright 2017 RackTop Systems. 39 * Copyright (c) 2017, Datto, Inc. All rights reserved. 40 */ 41 42 /* 43 * ZFS ioctls. 44 * 45 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage 46 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool. 47 * 48 * There are two ways that we handle ioctls: the legacy way where almost 49 * all of the logic is in the ioctl callback, and the new way where most 50 * of the marshalling is handled in the common entry point, zfsdev_ioctl(). 51 * 52 * Non-legacy ioctls should be registered by calling 53 * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked 54 * from userland by lzc_ioctl(). 55 * 56 * The registration arguments are as follows: 57 * 58 * const char *name 59 * The name of the ioctl. This is used for history logging. If the 60 * ioctl returns successfully (the callback returns 0), and allow_log 61 * is true, then a history log entry will be recorded with the input & 62 * output nvlists. The log entry can be printed with "zpool history -i". 63 * 64 * zfs_ioc_t ioc 65 * The ioctl request number, which userland will pass to ioctl(2). 66 * We want newer versions of libzfs and libzfs_core to run against 67 * existing zfs kernel modules (i.e. a deferred reboot after an update). 68 * Therefore the ioctl numbers cannot change from release to release. 69 * 70 * zfs_secpolicy_func_t *secpolicy 71 * This function will be called before the zfs_ioc_func_t, to 72 * determine if this operation is permitted. It should return EPERM 73 * on failure, and 0 on success. Checks include determining if the 74 * dataset is visible in this zone, and if the user has either all 75 * zfs privileges in the zone (SYS_MOUNT), or has been granted permission 76 * to do this operation on this dataset with "zfs allow". 77 * 78 * zfs_ioc_namecheck_t namecheck 79 * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool 80 * name, a dataset name, or nothing. If the name is not well-formed, 81 * the ioctl will fail and the callback will not be called. 82 * Therefore, the callback can assume that the name is well-formed 83 * (e.g. is null-terminated, doesn't have more than one '@' character, 84 * doesn't have invalid characters). 85 * 86 * zfs_ioc_poolcheck_t pool_check 87 * This specifies requirements on the pool state. If the pool does 88 * not meet them (is suspended or is readonly), the ioctl will fail 89 * and the callback will not be called. If any checks are specified 90 * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME. 91 * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED | 92 * POOL_CHECK_READONLY). 93 * 94 * zfs_ioc_key_t *nvl_keys 95 * The list of expected/allowable innvl input keys. This list is used 96 * to validate the nvlist input to the ioctl. 97 * 98 * boolean_t smush_outnvlist 99 * If smush_outnvlist is true, then the output is presumed to be a 100 * list of errors, and it will be "smushed" down to fit into the 101 * caller's buffer, by removing some entries and replacing them with a 102 * single "N_MORE_ERRORS" entry indicating how many were removed. See 103 * nvlist_smush() for details. If smush_outnvlist is false, and the 104 * outnvlist does not fit into the userland-provided buffer, then the 105 * ioctl will fail with ENOMEM. 106 * 107 * zfs_ioc_func_t *func 108 * The callback function that will perform the operation. 109 * 110 * The callback should return 0 on success, or an error number on 111 * failure. If the function fails, the userland ioctl will return -1, 112 * and errno will be set to the callback's return value. The callback 113 * will be called with the following arguments: 114 * 115 * const char *name 116 * The name of the pool or dataset to operate on, from 117 * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the 118 * expected type (pool, dataset, or none). 119 * 120 * nvlist_t *innvl 121 * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or 122 * NULL if no input nvlist was provided. Changes to this nvlist are 123 * ignored. If the input nvlist could not be deserialized, the 124 * ioctl will fail and the callback will not be called. 125 * 126 * nvlist_t *outnvl 127 * The output nvlist, initially empty. The callback can fill it in, 128 * and it will be returned to userland by serializing it into 129 * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization 130 * fails (e.g. because the caller didn't supply a large enough 131 * buffer), then the overall ioctl will fail. See the 132 * 'smush_nvlist' argument above for additional behaviors. 133 * 134 * There are two typical uses of the output nvlist: 135 * - To return state, e.g. property values. In this case, 136 * smush_outnvlist should be false. If the buffer was not large 137 * enough, the caller will reallocate a larger buffer and try 138 * the ioctl again. 139 * 140 * - To return multiple errors from an ioctl which makes on-disk 141 * changes. In this case, smush_outnvlist should be true. 142 * Ioctls which make on-disk modifications should generally not 143 * use the outnvl if they succeed, because the caller can not 144 * distinguish between the operation failing, and 145 * deserialization failing. 146 * 147 * IOCTL Interface Errors 148 * 149 * The following ioctl input errors can be returned: 150 * ZFS_ERR_IOC_CMD_UNAVAIL the ioctl number is not supported by kernel 151 * ZFS_ERR_IOC_ARG_UNAVAIL an input argument is not supported by kernel 152 * ZFS_ERR_IOC_ARG_REQUIRED a required input argument is missing 153 * ZFS_ERR_IOC_ARG_BADTYPE an input argument has an invalid type 154 */ 155 156 #include <sys/types.h> 157 #include <sys/param.h> 158 #include <sys/errno.h> 159 #include <sys/uio.h> 160 #include <sys/buf.h> 161 #include <sys/modctl.h> 162 #include <sys/open.h> 163 #include <sys/file.h> 164 #include <sys/kmem.h> 165 #include <sys/conf.h> 166 #include <sys/cmn_err.h> 167 #include <sys/stat.h> 168 #include <sys/zfs_ioctl.h> 169 #include <sys/zfs_vfsops.h> 170 #include <sys/zfs_znode.h> 171 #include <sys/zap.h> 172 #include <sys/spa.h> 173 #include <sys/spa_impl.h> 174 #include <sys/vdev.h> 175 #include <sys/priv_impl.h> 176 #include <sys/dmu.h> 177 #include <sys/dsl_dir.h> 178 #include <sys/dsl_dataset.h> 179 #include <sys/dsl_prop.h> 180 #include <sys/dsl_deleg.h> 181 #include <sys/dmu_objset.h> 182 #include <sys/dmu_impl.h> 183 #include <sys/dmu_tx.h> 184 #include <sys/ddi.h> 185 #include <sys/sunddi.h> 186 #include <sys/sunldi.h> 187 #include <sys/policy.h> 188 #include <sys/zone.h> 189 #include <sys/nvpair.h> 190 #include <sys/pathname.h> 191 #include <sys/mount.h> 192 #include <sys/sdt.h> 193 #include <sys/fs/zfs.h> 194 #include <sys/zfs_ctldir.h> 195 #include <sys/zfs_dir.h> 196 #include <sys/zfs_onexit.h> 197 #include <sys/zvol.h> 198 #include <sys/dsl_scan.h> 199 #include <sharefs/share.h> 200 #include <sys/dmu_objset.h> 201 #include <sys/dmu_recv.h> 202 #include <sys/dmu_send.h> 203 #include <sys/dsl_destroy.h> 204 #include <sys/dsl_bookmark.h> 205 #include <sys/dsl_userhold.h> 206 #include <sys/zfeature.h> 207 #include <sys/zcp.h> 208 #include <sys/zio_checksum.h> 209 #include <sys/vdev_removal.h> 210 #include <sys/vdev_impl.h> 211 #include <sys/vdev_initialize.h> 212 #include <sys/vdev_trim.h> 213 #include <sys/dsl_crypt.h> 214 215 #include "zfs_namecheck.h" 216 #include "zfs_prop.h" 217 #include "zfs_deleg.h" 218 #include "zfs_comutil.h" 219 220 #include "lua.h" 221 #include "lauxlib.h" 222 223 extern struct modlfs zfs_modlfs; 224 225 extern void zfs_init(void); 226 extern void zfs_fini(void); 227 228 ldi_ident_t zfs_li = NULL; 229 dev_info_t *zfs_dip; 230 231 uint_t zfs_fsyncer_key; 232 extern uint_t rrw_tsd_key; 233 static uint_t zfs_allow_log_key; 234 235 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *); 236 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *); 237 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *); 238 239 /* 240 * IOC Keys are used to document and validate user->kernel interface inputs. 241 * See zfs_keys_recv_new for an example declaration. Any key name that is not 242 * listed will be rejected as input. 243 * 244 * The keyname 'optional' is always allowed, and must be an nvlist if present. 245 * Arguments which older kernels can safely ignore can be placed under the 246 * "optional" key. 247 * 248 * When adding new keys to an existing ioc for new functionality, consider: 249 * - adding an entry into zfs_sysfs.c zfs_features[] list 250 * - updating the libzfs_input_check.c test utility 251 * 252 * Note: in the ZK_WILDCARDLIST case, the name serves as documentation 253 * for the expected name (bookmark, snapshot, property, etc) but there 254 * is no validation in the preflight zfs_check_input_nvpairs() check. 255 */ 256 typedef enum { 257 ZK_OPTIONAL = 1 << 0, /* pair is optional */ 258 ZK_WILDCARDLIST = 1 << 1, /* one or more unspecified key names */ 259 } ioc_key_flag_t; 260 261 /* DATA_TYPE_ANY is used when zkey_type can vary. */ 262 #define DATA_TYPE_ANY DATA_TYPE_UNKNOWN 263 264 typedef struct zfs_ioc_key { 265 const char *zkey_name; 266 data_type_t zkey_type; 267 ioc_key_flag_t zkey_flags; 268 } zfs_ioc_key_t; 269 270 typedef enum { 271 NO_NAME, 272 POOL_NAME, 273 DATASET_NAME 274 } zfs_ioc_namecheck_t; 275 276 typedef enum { 277 POOL_CHECK_NONE = 1 << 0, 278 POOL_CHECK_SUSPENDED = 1 << 1, 279 POOL_CHECK_READONLY = 1 << 2, 280 } zfs_ioc_poolcheck_t; 281 282 typedef struct zfs_ioc_vec { 283 zfs_ioc_legacy_func_t *zvec_legacy_func; 284 zfs_ioc_func_t *zvec_func; 285 zfs_secpolicy_func_t *zvec_secpolicy; 286 zfs_ioc_namecheck_t zvec_namecheck; 287 boolean_t zvec_allow_log; 288 zfs_ioc_poolcheck_t zvec_pool_check; 289 boolean_t zvec_smush_outnvlist; 290 const char *zvec_name; 291 const zfs_ioc_key_t *zvec_nvl_keys; 292 size_t zvec_nvl_key_count; 293 } zfs_ioc_vec_t; 294 295 /* This array is indexed by zfs_userquota_prop_t */ 296 static const char *userquota_perms[] = { 297 ZFS_DELEG_PERM_USERUSED, 298 ZFS_DELEG_PERM_USERQUOTA, 299 ZFS_DELEG_PERM_GROUPUSED, 300 ZFS_DELEG_PERM_GROUPQUOTA, 301 ZFS_DELEG_PERM_USEROBJUSED, 302 ZFS_DELEG_PERM_USEROBJQUOTA, 303 ZFS_DELEG_PERM_GROUPOBJUSED, 304 ZFS_DELEG_PERM_GROUPOBJQUOTA, 305 ZFS_DELEG_PERM_PROJECTUSED, 306 ZFS_DELEG_PERM_PROJECTQUOTA, 307 ZFS_DELEG_PERM_PROJECTOBJUSED, 308 ZFS_DELEG_PERM_PROJECTOBJQUOTA, 309 }; 310 311 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc); 312 static int zfs_ioc_id_quota_upgrade(zfs_cmd_t *zc); 313 static int zfs_check_settable(const char *name, nvpair_t *property, 314 cred_t *cr); 315 static int zfs_check_clearable(char *dataset, nvlist_t *props, 316 nvlist_t **errors); 317 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *, 318 boolean_t *); 319 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *); 320 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp); 321 322 static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature); 323 324 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */ 325 void 326 __dprintf(const char *file, const char *func, int line, const char *fmt, ...) 327 { 328 const char *newfile; 329 char buf[512]; 330 va_list adx; 331 332 /* 333 * Get rid of annoying "../common/" prefix to filename. 334 */ 335 newfile = strrchr(file, '/'); 336 if (newfile != NULL) { 337 newfile = newfile + 1; /* Get rid of leading / */ 338 } else { 339 newfile = file; 340 } 341 342 va_start(adx, fmt); 343 (void) vsnprintf(buf, sizeof (buf), fmt, adx); 344 va_end(adx); 345 346 /* 347 * To get this data, use the zfs-dprintf probe as so: 348 * dtrace -q -n 'zfs-dprintf \ 349 * /stringof(arg0) == "dbuf.c"/ \ 350 * {printf("%s: %s", stringof(arg1), stringof(arg3))}' 351 * arg0 = file name 352 * arg1 = function name 353 * arg2 = line number 354 * arg3 = message 355 */ 356 DTRACE_PROBE4(zfs__dprintf, 357 char *, newfile, char *, func, int, line, char *, buf); 358 } 359 360 static void 361 history_str_free(char *buf) 362 { 363 kmem_free(buf, HIS_MAX_RECORD_LEN); 364 } 365 366 static char * 367 history_str_get(zfs_cmd_t *zc) 368 { 369 char *buf; 370 371 if (zc->zc_history == 0) 372 return (NULL); 373 374 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP); 375 if (copyinstr((void *)(uintptr_t)zc->zc_history, 376 buf, HIS_MAX_RECORD_LEN, NULL) != 0) { 377 history_str_free(buf); 378 return (NULL); 379 } 380 381 buf[HIS_MAX_RECORD_LEN -1] = '\0'; 382 383 return (buf); 384 } 385 386 /* 387 * Check to see if the named dataset is currently defined as bootable 388 */ 389 static boolean_t 390 zfs_is_bootfs(const char *name) 391 { 392 objset_t *os; 393 394 if (dmu_objset_hold(name, FTAG, &os) == 0) { 395 boolean_t ret; 396 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os))); 397 dmu_objset_rele(os, FTAG); 398 return (ret); 399 } 400 return (B_FALSE); 401 } 402 403 /* 404 * Return non-zero if the spa version is less than requested version. 405 */ 406 static int 407 zfs_earlier_version(const char *name, int version) 408 { 409 spa_t *spa; 410 411 if (spa_open(name, &spa, FTAG) == 0) { 412 if (spa_version(spa) < version) { 413 spa_close(spa, FTAG); 414 return (1); 415 } 416 spa_close(spa, FTAG); 417 } 418 return (0); 419 } 420 421 /* 422 * Return TRUE if the ZPL version is less than requested version. 423 */ 424 static boolean_t 425 zpl_earlier_version(const char *name, int version) 426 { 427 objset_t *os; 428 boolean_t rc = B_TRUE; 429 430 if (dmu_objset_hold(name, FTAG, &os) == 0) { 431 uint64_t zplversion; 432 433 if (dmu_objset_type(os) != DMU_OST_ZFS) { 434 dmu_objset_rele(os, FTAG); 435 return (B_TRUE); 436 } 437 /* XXX reading from non-owned objset */ 438 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0) 439 rc = zplversion < version; 440 dmu_objset_rele(os, FTAG); 441 } 442 return (rc); 443 } 444 445 static void 446 zfs_log_history(zfs_cmd_t *zc) 447 { 448 spa_t *spa; 449 char *buf; 450 451 if ((buf = history_str_get(zc)) == NULL) 452 return; 453 454 if (spa_open(zc->zc_name, &spa, FTAG) == 0) { 455 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY) 456 (void) spa_history_log(spa, buf); 457 spa_close(spa, FTAG); 458 } 459 history_str_free(buf); 460 } 461 462 /* 463 * Policy for top-level read operations (list pools). Requires no privileges, 464 * and can be used in the local zone, as there is no associated dataset. 465 */ 466 /* ARGSUSED */ 467 static int 468 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 469 { 470 return (0); 471 } 472 473 /* 474 * Policy for dataset read operations (list children, get statistics). Requires 475 * no privileges, but must be visible in the local zone. 476 */ 477 /* ARGSUSED */ 478 static int 479 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 480 { 481 if (INGLOBALZONE(curproc) || 482 zone_dataset_visible(zc->zc_name, NULL)) 483 return (0); 484 485 return (SET_ERROR(ENOENT)); 486 } 487 488 static int 489 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr) 490 { 491 int writable = 1; 492 493 /* 494 * The dataset must be visible by this zone -- check this first 495 * so they don't see EPERM on something they shouldn't know about. 496 */ 497 if (!INGLOBALZONE(curproc) && 498 !zone_dataset_visible(dataset, &writable)) 499 return (SET_ERROR(ENOENT)); 500 501 if (INGLOBALZONE(curproc)) { 502 /* 503 * If the fs is zoned, only root can access it from the 504 * global zone. 505 */ 506 if (secpolicy_zfs(cr) && zoned) 507 return (SET_ERROR(EPERM)); 508 } else { 509 /* 510 * If we are in a local zone, the 'zoned' property must be set. 511 */ 512 if (!zoned) 513 return (SET_ERROR(EPERM)); 514 515 /* must be writable by this zone */ 516 if (!writable) 517 return (SET_ERROR(EPERM)); 518 } 519 return (0); 520 } 521 522 static int 523 zfs_dozonecheck(const char *dataset, cred_t *cr) 524 { 525 uint64_t zoned; 526 527 if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL)) 528 return (SET_ERROR(ENOENT)); 529 530 return (zfs_dozonecheck_impl(dataset, zoned, cr)); 531 } 532 533 static int 534 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr) 535 { 536 uint64_t zoned; 537 538 if (dsl_prop_get_int_ds(ds, "zoned", &zoned)) 539 return (SET_ERROR(ENOENT)); 540 541 return (zfs_dozonecheck_impl(dataset, zoned, cr)); 542 } 543 544 static int 545 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds, 546 const char *perm, cred_t *cr) 547 { 548 int error; 549 550 error = zfs_dozonecheck_ds(name, ds, cr); 551 if (error == 0) { 552 error = secpolicy_zfs(cr); 553 if (error != 0) 554 error = dsl_deleg_access_impl(ds, perm, cr); 555 } 556 return (error); 557 } 558 559 static int 560 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr) 561 { 562 int error; 563 dsl_dataset_t *ds; 564 dsl_pool_t *dp; 565 566 /* 567 * First do a quick check for root in the global zone, which 568 * is allowed to do all write_perms. This ensures that zfs_ioc_* 569 * will get to handle nonexistent datasets. 570 */ 571 if (INGLOBALZONE(curproc) && secpolicy_zfs(cr) == 0) 572 return (0); 573 574 error = dsl_pool_hold(name, FTAG, &dp); 575 if (error != 0) 576 return (error); 577 578 error = dsl_dataset_hold(dp, name, FTAG, &ds); 579 if (error != 0) { 580 dsl_pool_rele(dp, FTAG); 581 return (error); 582 } 583 584 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr); 585 586 dsl_dataset_rele(ds, FTAG); 587 dsl_pool_rele(dp, FTAG); 588 return (error); 589 } 590 591 /* 592 * Policy for setting the security label property. 593 * 594 * Returns 0 for success, non-zero for access and other errors. 595 */ 596 static int 597 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr) 598 { 599 char ds_hexsl[MAXNAMELEN]; 600 bslabel_t ds_sl, new_sl; 601 boolean_t new_default = FALSE; 602 uint64_t zoned; 603 int needed_priv = -1; 604 int error; 605 606 /* First get the existing dataset label. */ 607 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL), 608 1, sizeof (ds_hexsl), &ds_hexsl, NULL); 609 if (error != 0) 610 return (SET_ERROR(EPERM)); 611 612 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0) 613 new_default = TRUE; 614 615 /* The label must be translatable */ 616 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0)) 617 return (SET_ERROR(EINVAL)); 618 619 /* 620 * In a non-global zone, disallow attempts to set a label that 621 * doesn't match that of the zone; otherwise no other checks 622 * are needed. 623 */ 624 if (!INGLOBALZONE(curproc)) { 625 if (new_default || !blequal(&new_sl, CR_SL(CRED()))) 626 return (SET_ERROR(EPERM)); 627 return (0); 628 } 629 630 /* 631 * For global-zone datasets (i.e., those whose zoned property is 632 * "off", verify that the specified new label is valid for the 633 * global zone. 634 */ 635 if (dsl_prop_get_integer(name, 636 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL)) 637 return (SET_ERROR(EPERM)); 638 if (!zoned) { 639 if (zfs_check_global_label(name, strval) != 0) 640 return (SET_ERROR(EPERM)); 641 } 642 643 /* 644 * If the existing dataset label is nondefault, check if the 645 * dataset is mounted (label cannot be changed while mounted). 646 * Get the zfsvfs; if there isn't one, then the dataset isn't 647 * mounted (or isn't a dataset, doesn't exist, ...). 648 */ 649 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) { 650 objset_t *os; 651 static char *setsl_tag = "setsl_tag"; 652 653 /* 654 * Try to own the dataset; abort if there is any error, 655 * (e.g., already mounted, in use, or other error). 656 */ 657 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE, B_TRUE, 658 setsl_tag, &os); 659 if (error != 0) 660 return (SET_ERROR(EPERM)); 661 662 dmu_objset_disown(os, B_TRUE, setsl_tag); 663 664 if (new_default) { 665 needed_priv = PRIV_FILE_DOWNGRADE_SL; 666 goto out_check; 667 } 668 669 if (hexstr_to_label(strval, &new_sl) != 0) 670 return (SET_ERROR(EPERM)); 671 672 if (blstrictdom(&ds_sl, &new_sl)) 673 needed_priv = PRIV_FILE_DOWNGRADE_SL; 674 else if (blstrictdom(&new_sl, &ds_sl)) 675 needed_priv = PRIV_FILE_UPGRADE_SL; 676 } else { 677 /* dataset currently has a default label */ 678 if (!new_default) 679 needed_priv = PRIV_FILE_UPGRADE_SL; 680 } 681 682 out_check: 683 if (needed_priv != -1) 684 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL)); 685 return (0); 686 } 687 688 static int 689 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval, 690 cred_t *cr) 691 { 692 char *strval; 693 694 /* 695 * Check permissions for special properties. 696 */ 697 switch (prop) { 698 case ZFS_PROP_ZONED: 699 /* 700 * Disallow setting of 'zoned' from within a local zone. 701 */ 702 if (!INGLOBALZONE(curproc)) 703 return (SET_ERROR(EPERM)); 704 break; 705 706 case ZFS_PROP_QUOTA: 707 case ZFS_PROP_FILESYSTEM_LIMIT: 708 case ZFS_PROP_SNAPSHOT_LIMIT: 709 if (!INGLOBALZONE(curproc)) { 710 uint64_t zoned; 711 char setpoint[ZFS_MAX_DATASET_NAME_LEN]; 712 /* 713 * Unprivileged users are allowed to modify the 714 * limit on things *under* (ie. contained by) 715 * the thing they own. 716 */ 717 if (dsl_prop_get_integer(dsname, "zoned", &zoned, 718 setpoint)) 719 return (SET_ERROR(EPERM)); 720 if (!zoned || strlen(dsname) <= strlen(setpoint)) 721 return (SET_ERROR(EPERM)); 722 } 723 break; 724 725 case ZFS_PROP_MLSLABEL: 726 if (!is_system_labeled()) 727 return (SET_ERROR(EPERM)); 728 729 if (nvpair_value_string(propval, &strval) == 0) { 730 int err; 731 732 err = zfs_set_slabel_policy(dsname, strval, CRED()); 733 if (err != 0) 734 return (err); 735 } 736 break; 737 } 738 739 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr)); 740 } 741 742 /* ARGSUSED */ 743 static int 744 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 745 { 746 int error; 747 748 error = zfs_dozonecheck(zc->zc_name, cr); 749 if (error != 0) 750 return (error); 751 752 /* 753 * permission to set permissions will be evaluated later in 754 * dsl_deleg_can_allow() 755 */ 756 return (0); 757 } 758 759 /* ARGSUSED */ 760 static int 761 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 762 { 763 return (zfs_secpolicy_write_perms(zc->zc_name, 764 ZFS_DELEG_PERM_ROLLBACK, cr)); 765 } 766 767 /* ARGSUSED */ 768 static int 769 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 770 { 771 dsl_pool_t *dp; 772 dsl_dataset_t *ds; 773 char *cp; 774 int error; 775 776 /* 777 * Generate the current snapshot name from the given objsetid, then 778 * use that name for the secpolicy/zone checks. 779 */ 780 cp = strchr(zc->zc_name, '@'); 781 if (cp == NULL) 782 return (SET_ERROR(EINVAL)); 783 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 784 if (error != 0) 785 return (error); 786 787 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds); 788 if (error != 0) { 789 dsl_pool_rele(dp, FTAG); 790 return (error); 791 } 792 793 dsl_dataset_name(ds, zc->zc_name); 794 795 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds, 796 ZFS_DELEG_PERM_SEND, cr); 797 dsl_dataset_rele(ds, FTAG); 798 dsl_pool_rele(dp, FTAG); 799 800 return (error); 801 } 802 803 /* ARGSUSED */ 804 static int 805 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 806 { 807 return (zfs_secpolicy_write_perms(zc->zc_name, 808 ZFS_DELEG_PERM_SEND, cr)); 809 } 810 811 /* ARGSUSED */ 812 static int 813 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 814 { 815 vnode_t *vp; 816 int error; 817 818 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, 819 NO_FOLLOW, NULL, &vp)) != 0) 820 return (error); 821 822 /* Now make sure mntpnt and dataset are ZFS */ 823 824 if (vp->v_vfsp->vfs_fstype != zfsfstype || 825 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), 826 zc->zc_name) != 0)) { 827 VN_RELE(vp); 828 return (SET_ERROR(EPERM)); 829 } 830 831 VN_RELE(vp); 832 return (dsl_deleg_access(zc->zc_name, 833 ZFS_DELEG_PERM_SHARE, cr)); 834 } 835 836 int 837 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 838 { 839 if (secpolicy_nfs(cr) == 0) { 840 return (0); 841 } else { 842 return (zfs_secpolicy_deleg_share(zc, innvl, cr)); 843 } 844 } 845 846 int 847 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 848 { 849 if (secpolicy_smb(cr) == 0) { 850 return (0); 851 } else { 852 return (zfs_secpolicy_deleg_share(zc, innvl, cr)); 853 } 854 } 855 856 static int 857 zfs_get_parent(const char *datasetname, char *parent, int parentsize) 858 { 859 char *cp; 860 861 /* 862 * Remove the @bla or /bla from the end of the name to get the parent. 863 */ 864 (void) strncpy(parent, datasetname, parentsize); 865 cp = strrchr(parent, '@'); 866 if (cp != NULL) { 867 cp[0] = '\0'; 868 } else { 869 cp = strrchr(parent, '/'); 870 if (cp == NULL) 871 return (SET_ERROR(ENOENT)); 872 cp[0] = '\0'; 873 } 874 875 return (0); 876 } 877 878 int 879 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr) 880 { 881 int error; 882 883 if ((error = zfs_secpolicy_write_perms(name, 884 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 885 return (error); 886 887 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr)); 888 } 889 890 /* ARGSUSED */ 891 static int 892 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 893 { 894 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr)); 895 } 896 897 /* 898 * Destroying snapshots with delegated permissions requires 899 * descendant mount and destroy permissions. 900 */ 901 /* ARGSUSED */ 902 static int 903 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 904 { 905 nvlist_t *snaps; 906 nvpair_t *pair, *nextpair; 907 int error = 0; 908 909 snaps = fnvlist_lookup_nvlist(innvl, "snaps"); 910 911 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 912 pair = nextpair) { 913 nextpair = nvlist_next_nvpair(snaps, pair); 914 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr); 915 if (error == ENOENT) { 916 /* 917 * Ignore any snapshots that don't exist (we consider 918 * them "already destroyed"). Remove the name from the 919 * nvl here in case the snapshot is created between 920 * now and when we try to destroy it (in which case 921 * we don't want to destroy it since we haven't 922 * checked for permission). 923 */ 924 fnvlist_remove_nvpair(snaps, pair); 925 error = 0; 926 } 927 if (error != 0) 928 break; 929 } 930 931 return (error); 932 } 933 934 int 935 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr) 936 { 937 char parentname[ZFS_MAX_DATASET_NAME_LEN]; 938 int error; 939 940 if ((error = zfs_secpolicy_write_perms(from, 941 ZFS_DELEG_PERM_RENAME, cr)) != 0) 942 return (error); 943 944 if ((error = zfs_secpolicy_write_perms(from, 945 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 946 return (error); 947 948 if ((error = zfs_get_parent(to, parentname, 949 sizeof (parentname))) != 0) 950 return (error); 951 952 if ((error = zfs_secpolicy_write_perms(parentname, 953 ZFS_DELEG_PERM_CREATE, cr)) != 0) 954 return (error); 955 956 if ((error = zfs_secpolicy_write_perms(parentname, 957 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 958 return (error); 959 960 return (error); 961 } 962 963 /* ARGSUSED */ 964 static int 965 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 966 { 967 return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr)); 968 } 969 970 /* ARGSUSED */ 971 static int 972 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 973 { 974 dsl_pool_t *dp; 975 dsl_dataset_t *clone; 976 int error; 977 978 error = zfs_secpolicy_write_perms(zc->zc_name, 979 ZFS_DELEG_PERM_PROMOTE, cr); 980 if (error != 0) 981 return (error); 982 983 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 984 if (error != 0) 985 return (error); 986 987 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone); 988 989 if (error == 0) { 990 char parentname[ZFS_MAX_DATASET_NAME_LEN]; 991 dsl_dataset_t *origin = NULL; 992 dsl_dir_t *dd; 993 dd = clone->ds_dir; 994 995 error = dsl_dataset_hold_obj(dd->dd_pool, 996 dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin); 997 if (error != 0) { 998 dsl_dataset_rele(clone, FTAG); 999 dsl_pool_rele(dp, FTAG); 1000 return (error); 1001 } 1002 1003 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone, 1004 ZFS_DELEG_PERM_MOUNT, cr); 1005 1006 dsl_dataset_name(origin, parentname); 1007 if (error == 0) { 1008 error = zfs_secpolicy_write_perms_ds(parentname, origin, 1009 ZFS_DELEG_PERM_PROMOTE, cr); 1010 } 1011 dsl_dataset_rele(clone, FTAG); 1012 dsl_dataset_rele(origin, FTAG); 1013 } 1014 dsl_pool_rele(dp, FTAG); 1015 return (error); 1016 } 1017 1018 /* ARGSUSED */ 1019 static int 1020 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1021 { 1022 int error; 1023 1024 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 1025 ZFS_DELEG_PERM_RECEIVE, cr)) != 0) 1026 return (error); 1027 1028 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 1029 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 1030 return (error); 1031 1032 return (zfs_secpolicy_write_perms(zc->zc_name, 1033 ZFS_DELEG_PERM_CREATE, cr)); 1034 } 1035 1036 int 1037 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr) 1038 { 1039 return (zfs_secpolicy_write_perms(name, 1040 ZFS_DELEG_PERM_SNAPSHOT, cr)); 1041 } 1042 1043 /* 1044 * Check for permission to create each snapshot in the nvlist. 1045 */ 1046 /* ARGSUSED */ 1047 static int 1048 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1049 { 1050 nvlist_t *snaps; 1051 int error = 0; 1052 nvpair_t *pair; 1053 1054 snaps = fnvlist_lookup_nvlist(innvl, "snaps"); 1055 1056 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 1057 pair = nvlist_next_nvpair(snaps, pair)) { 1058 char *name = nvpair_name(pair); 1059 char *atp = strchr(name, '@'); 1060 1061 if (atp == NULL) { 1062 error = SET_ERROR(EINVAL); 1063 break; 1064 } 1065 *atp = '\0'; 1066 error = zfs_secpolicy_snapshot_perms(name, cr); 1067 *atp = '@'; 1068 if (error != 0) 1069 break; 1070 } 1071 return (error); 1072 } 1073 1074 /* 1075 * Check for permission to create each bookmark in the nvlist. 1076 */ 1077 /* ARGSUSED */ 1078 static int 1079 zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1080 { 1081 int error = 0; 1082 1083 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 1084 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 1085 char *name = nvpair_name(pair); 1086 char *hashp = strchr(name, '#'); 1087 1088 if (hashp == NULL) { 1089 error = SET_ERROR(EINVAL); 1090 break; 1091 } 1092 *hashp = '\0'; 1093 error = zfs_secpolicy_write_perms(name, 1094 ZFS_DELEG_PERM_BOOKMARK, cr); 1095 *hashp = '#'; 1096 if (error != 0) 1097 break; 1098 } 1099 return (error); 1100 } 1101 1102 /* ARGSUSED */ 1103 static int 1104 zfs_secpolicy_remap(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1105 { 1106 return (zfs_secpolicy_write_perms(zc->zc_name, 1107 ZFS_DELEG_PERM_REMAP, cr)); 1108 } 1109 1110 /* ARGSUSED */ 1111 static int 1112 zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1113 { 1114 nvpair_t *pair, *nextpair; 1115 int error = 0; 1116 1117 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL; 1118 pair = nextpair) { 1119 char *name = nvpair_name(pair); 1120 char *hashp = strchr(name, '#'); 1121 nextpair = nvlist_next_nvpair(innvl, pair); 1122 1123 if (hashp == NULL) { 1124 error = SET_ERROR(EINVAL); 1125 break; 1126 } 1127 1128 *hashp = '\0'; 1129 error = zfs_secpolicy_write_perms(name, 1130 ZFS_DELEG_PERM_DESTROY, cr); 1131 *hashp = '#'; 1132 if (error == ENOENT) { 1133 /* 1134 * Ignore any filesystems that don't exist (we consider 1135 * their bookmarks "already destroyed"). Remove 1136 * the name from the nvl here in case the filesystem 1137 * is created between now and when we try to destroy 1138 * the bookmark (in which case we don't want to 1139 * destroy it since we haven't checked for permission). 1140 */ 1141 fnvlist_remove_nvpair(innvl, pair); 1142 error = 0; 1143 } 1144 if (error != 0) 1145 break; 1146 } 1147 1148 return (error); 1149 } 1150 1151 /* ARGSUSED */ 1152 static int 1153 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1154 { 1155 /* 1156 * Even root must have a proper TSD so that we know what pool 1157 * to log to. 1158 */ 1159 if (tsd_get(zfs_allow_log_key) == NULL) 1160 return (SET_ERROR(EPERM)); 1161 return (0); 1162 } 1163 1164 static int 1165 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1166 { 1167 char parentname[ZFS_MAX_DATASET_NAME_LEN]; 1168 int error; 1169 char *origin; 1170 1171 if ((error = zfs_get_parent(zc->zc_name, parentname, 1172 sizeof (parentname))) != 0) 1173 return (error); 1174 1175 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 && 1176 (error = zfs_secpolicy_write_perms(origin, 1177 ZFS_DELEG_PERM_CLONE, cr)) != 0) 1178 return (error); 1179 1180 if ((error = zfs_secpolicy_write_perms(parentname, 1181 ZFS_DELEG_PERM_CREATE, cr)) != 0) 1182 return (error); 1183 1184 return (zfs_secpolicy_write_perms(parentname, 1185 ZFS_DELEG_PERM_MOUNT, cr)); 1186 } 1187 1188 /* 1189 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires 1190 * SYS_CONFIG privilege, which is not available in a local zone. 1191 */ 1192 /* ARGSUSED */ 1193 static int 1194 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1195 { 1196 if (secpolicy_sys_config(cr, B_FALSE) != 0) 1197 return (SET_ERROR(EPERM)); 1198 1199 return (0); 1200 } 1201 1202 /* 1203 * Policy for object to name lookups. 1204 */ 1205 /* ARGSUSED */ 1206 static int 1207 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1208 { 1209 int error; 1210 1211 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0) 1212 return (0); 1213 1214 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr); 1215 return (error); 1216 } 1217 1218 /* 1219 * Policy for fault injection. Requires all privileges. 1220 */ 1221 /* ARGSUSED */ 1222 static int 1223 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1224 { 1225 return (secpolicy_zinject(cr)); 1226 } 1227 1228 /* ARGSUSED */ 1229 static int 1230 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1231 { 1232 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value); 1233 1234 if (prop == ZPROP_INVAL) { 1235 if (!zfs_prop_user(zc->zc_value)) 1236 return (SET_ERROR(EINVAL)); 1237 return (zfs_secpolicy_write_perms(zc->zc_name, 1238 ZFS_DELEG_PERM_USERPROP, cr)); 1239 } else { 1240 return (zfs_secpolicy_setprop(zc->zc_name, prop, 1241 NULL, cr)); 1242 } 1243 } 1244 1245 static int 1246 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1247 { 1248 int err = zfs_secpolicy_read(zc, innvl, cr); 1249 if (err) 1250 return (err); 1251 1252 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 1253 return (SET_ERROR(EINVAL)); 1254 1255 if (zc->zc_value[0] == 0) { 1256 /* 1257 * They are asking about a posix uid/gid. If it's 1258 * themself, allow it. 1259 */ 1260 if (zc->zc_objset_type == ZFS_PROP_USERUSED || 1261 zc->zc_objset_type == ZFS_PROP_USERQUOTA || 1262 zc->zc_objset_type == ZFS_PROP_USEROBJUSED || 1263 zc->zc_objset_type == ZFS_PROP_USEROBJQUOTA) { 1264 if (zc->zc_guid == crgetuid(cr)) 1265 return (0); 1266 } else if (zc->zc_objset_type == ZFS_PROP_GROUPUSED || 1267 zc->zc_objset_type == ZFS_PROP_GROUPQUOTA || 1268 zc->zc_objset_type == ZFS_PROP_GROUPOBJUSED || 1269 zc->zc_objset_type == ZFS_PROP_GROUPOBJQUOTA) { 1270 if (groupmember(zc->zc_guid, cr)) 1271 return (0); 1272 } 1273 /* else is for project quota/used */ 1274 } 1275 1276 return (zfs_secpolicy_write_perms(zc->zc_name, 1277 userquota_perms[zc->zc_objset_type], cr)); 1278 } 1279 1280 static int 1281 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1282 { 1283 int err = zfs_secpolicy_read(zc, innvl, cr); 1284 if (err) 1285 return (err); 1286 1287 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 1288 return (SET_ERROR(EINVAL)); 1289 1290 return (zfs_secpolicy_write_perms(zc->zc_name, 1291 userquota_perms[zc->zc_objset_type], cr)); 1292 } 1293 1294 /* ARGSUSED */ 1295 static int 1296 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1297 { 1298 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION, 1299 NULL, cr)); 1300 } 1301 1302 /* ARGSUSED */ 1303 static int 1304 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1305 { 1306 nvpair_t *pair; 1307 nvlist_t *holds; 1308 int error; 1309 1310 holds = fnvlist_lookup_nvlist(innvl, "holds"); 1311 1312 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL; 1313 pair = nvlist_next_nvpair(holds, pair)) { 1314 char fsname[ZFS_MAX_DATASET_NAME_LEN]; 1315 error = dmu_fsname(nvpair_name(pair), fsname); 1316 if (error != 0) 1317 return (error); 1318 error = zfs_secpolicy_write_perms(fsname, 1319 ZFS_DELEG_PERM_HOLD, cr); 1320 if (error != 0) 1321 return (error); 1322 } 1323 return (0); 1324 } 1325 1326 /* ARGSUSED */ 1327 static int 1328 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1329 { 1330 nvpair_t *pair; 1331 int error; 1332 1333 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL; 1334 pair = nvlist_next_nvpair(innvl, pair)) { 1335 char fsname[ZFS_MAX_DATASET_NAME_LEN]; 1336 error = dmu_fsname(nvpair_name(pair), fsname); 1337 if (error != 0) 1338 return (error); 1339 error = zfs_secpolicy_write_perms(fsname, 1340 ZFS_DELEG_PERM_RELEASE, cr); 1341 if (error != 0) 1342 return (error); 1343 } 1344 return (0); 1345 } 1346 1347 /* ARGSUSED */ 1348 static int 1349 zfs_secpolicy_load_key(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1350 { 1351 return (zfs_secpolicy_write_perms(zc->zc_name, 1352 ZFS_DELEG_PERM_LOAD_KEY, cr)); 1353 } 1354 1355 /* ARGSUSED */ 1356 static int 1357 zfs_secpolicy_change_key(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1358 { 1359 return (zfs_secpolicy_write_perms(zc->zc_name, 1360 ZFS_DELEG_PERM_CHANGE_KEY, cr)); 1361 } 1362 1363 /* 1364 * Policy for allowing temporary snapshots to be taken or released 1365 */ 1366 static int 1367 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1368 { 1369 /* 1370 * A temporary snapshot is the same as a snapshot, 1371 * hold, destroy and release all rolled into one. 1372 * Delegated diff alone is sufficient that we allow this. 1373 */ 1374 int error; 1375 1376 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 1377 ZFS_DELEG_PERM_DIFF, cr)) == 0) 1378 return (0); 1379 1380 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr); 1381 1382 if (innvl != NULL) { 1383 if (error == 0) 1384 error = zfs_secpolicy_hold(zc, innvl, cr); 1385 if (error == 0) 1386 error = zfs_secpolicy_release(zc, innvl, cr); 1387 if (error == 0) 1388 error = zfs_secpolicy_destroy(zc, innvl, cr); 1389 } 1390 return (error); 1391 } 1392 1393 /* 1394 * Returns the nvlist as specified by the user in the zfs_cmd_t. 1395 */ 1396 static int 1397 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp) 1398 { 1399 char *packed; 1400 int error; 1401 nvlist_t *list = NULL; 1402 1403 /* 1404 * Read in and unpack the user-supplied nvlist. 1405 */ 1406 if (size == 0) 1407 return (SET_ERROR(EINVAL)); 1408 1409 packed = kmem_alloc(size, KM_SLEEP); 1410 1411 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size, 1412 iflag)) != 0) { 1413 kmem_free(packed, size); 1414 return (SET_ERROR(EFAULT)); 1415 } 1416 1417 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) { 1418 kmem_free(packed, size); 1419 return (error); 1420 } 1421 1422 kmem_free(packed, size); 1423 1424 *nvp = list; 1425 return (0); 1426 } 1427 1428 /* 1429 * Reduce the size of this nvlist until it can be serialized in 'max' bytes. 1430 * Entries will be removed from the end of the nvlist, and one int32 entry 1431 * named "N_MORE_ERRORS" will be added indicating how many entries were 1432 * removed. 1433 */ 1434 static int 1435 nvlist_smush(nvlist_t *errors, size_t max) 1436 { 1437 size_t size; 1438 1439 size = fnvlist_size(errors); 1440 1441 if (size > max) { 1442 nvpair_t *more_errors; 1443 int n = 0; 1444 1445 if (max < 1024) 1446 return (SET_ERROR(ENOMEM)); 1447 1448 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0); 1449 more_errors = nvlist_prev_nvpair(errors, NULL); 1450 1451 do { 1452 nvpair_t *pair = nvlist_prev_nvpair(errors, 1453 more_errors); 1454 fnvlist_remove_nvpair(errors, pair); 1455 n++; 1456 size = fnvlist_size(errors); 1457 } while (size > max); 1458 1459 fnvlist_remove_nvpair(errors, more_errors); 1460 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n); 1461 ASSERT3U(fnvlist_size(errors), <=, max); 1462 } 1463 1464 return (0); 1465 } 1466 1467 static int 1468 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl) 1469 { 1470 char *packed = NULL; 1471 int error = 0; 1472 size_t size; 1473 1474 size = fnvlist_size(nvl); 1475 1476 if (size > zc->zc_nvlist_dst_size) { 1477 error = SET_ERROR(ENOMEM); 1478 } else { 1479 packed = fnvlist_pack(nvl, &size); 1480 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst, 1481 size, zc->zc_iflags) != 0) 1482 error = SET_ERROR(EFAULT); 1483 fnvlist_pack_free(packed, size); 1484 } 1485 1486 zc->zc_nvlist_dst_size = size; 1487 zc->zc_nvlist_dst_filled = B_TRUE; 1488 return (error); 1489 } 1490 1491 int 1492 getzfsvfs_impl(objset_t *os, zfsvfs_t **zfvp) 1493 { 1494 int error = 0; 1495 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1496 return (SET_ERROR(EINVAL)); 1497 } 1498 1499 mutex_enter(&os->os_user_ptr_lock); 1500 *zfvp = dmu_objset_get_user(os); 1501 if (*zfvp) { 1502 VFS_HOLD((*zfvp)->z_vfs); 1503 } else { 1504 error = SET_ERROR(ESRCH); 1505 } 1506 mutex_exit(&os->os_user_ptr_lock); 1507 return (error); 1508 } 1509 1510 int 1511 getzfsvfs(const char *dsname, zfsvfs_t **zfvp) 1512 { 1513 objset_t *os; 1514 int error; 1515 1516 error = dmu_objset_hold(dsname, FTAG, &os); 1517 if (error != 0) 1518 return (error); 1519 1520 error = getzfsvfs_impl(os, zfvp); 1521 dmu_objset_rele(os, FTAG); 1522 return (error); 1523 } 1524 1525 /* 1526 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which 1527 * case its z_vfs will be NULL, and it will be opened as the owner. 1528 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER, 1529 * which prevents all vnode ops from running. 1530 */ 1531 static int 1532 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer) 1533 { 1534 int error = 0; 1535 1536 if (getzfsvfs(name, zfvp) != 0) 1537 error = zfsvfs_create(name, B_FALSE, zfvp); 1538 if (error == 0) { 1539 rrm_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER : 1540 RW_READER, tag); 1541 if ((*zfvp)->z_unmounted) { 1542 /* 1543 * XXX we could probably try again, since the unmounting 1544 * thread should be just about to disassociate the 1545 * objset from the zfsvfs. 1546 */ 1547 rrm_exit(&(*zfvp)->z_teardown_lock, tag); 1548 return (SET_ERROR(EBUSY)); 1549 } 1550 } 1551 return (error); 1552 } 1553 1554 static void 1555 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag) 1556 { 1557 rrm_exit(&zfsvfs->z_teardown_lock, tag); 1558 1559 if (zfsvfs->z_vfs) { 1560 VFS_RELE(zfsvfs->z_vfs); 1561 } else { 1562 dmu_objset_disown(zfsvfs->z_os, B_TRUE, zfsvfs); 1563 zfsvfs_free(zfsvfs); 1564 } 1565 } 1566 1567 static int 1568 zfs_ioc_pool_create(zfs_cmd_t *zc) 1569 { 1570 int error; 1571 nvlist_t *config, *props = NULL; 1572 nvlist_t *rootprops = NULL; 1573 nvlist_t *zplprops = NULL; 1574 char *spa_name = zc->zc_name; 1575 dsl_crypto_params_t *dcp = NULL; 1576 1577 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1578 zc->zc_iflags, &config)) 1579 return (error); 1580 1581 if (zc->zc_nvlist_src_size != 0 && (error = 1582 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1583 zc->zc_iflags, &props))) { 1584 nvlist_free(config); 1585 return (error); 1586 } 1587 1588 if (props) { 1589 nvlist_t *nvl = NULL; 1590 nvlist_t *hidden_args = NULL; 1591 uint64_t version = SPA_VERSION; 1592 char *tname; 1593 1594 (void) nvlist_lookup_uint64(props, 1595 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version); 1596 if (!SPA_VERSION_IS_SUPPORTED(version)) { 1597 error = SET_ERROR(EINVAL); 1598 goto pool_props_bad; 1599 } 1600 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl); 1601 if (nvl) { 1602 error = nvlist_dup(nvl, &rootprops, KM_SLEEP); 1603 if (error != 0) { 1604 nvlist_free(config); 1605 nvlist_free(props); 1606 return (error); 1607 } 1608 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS); 1609 } 1610 1611 (void) nvlist_lookup_nvlist(props, ZPOOL_HIDDEN_ARGS, 1612 &hidden_args); 1613 error = dsl_crypto_params_create_nvlist(DCP_CMD_NONE, 1614 rootprops, hidden_args, &dcp); 1615 if (error != 0) { 1616 nvlist_free(config); 1617 nvlist_free(props); 1618 return (error); 1619 } 1620 (void) nvlist_remove_all(props, ZPOOL_HIDDEN_ARGS); 1621 1622 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 1623 error = zfs_fill_zplprops_root(version, rootprops, 1624 zplprops, NULL); 1625 if (error != 0) 1626 goto pool_props_bad; 1627 1628 if (nvlist_lookup_string(props, 1629 zpool_prop_to_name(ZPOOL_PROP_TNAME), &tname) == 0) 1630 spa_name = tname; 1631 } 1632 1633 error = spa_create(zc->zc_name, config, props, zplprops, dcp); 1634 1635 /* 1636 * Set the remaining root properties 1637 */ 1638 if (!error && (error = zfs_set_prop_nvlist(spa_name, 1639 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0) 1640 (void) spa_destroy(spa_name); 1641 1642 pool_props_bad: 1643 nvlist_free(rootprops); 1644 nvlist_free(zplprops); 1645 nvlist_free(config); 1646 nvlist_free(props); 1647 dsl_crypto_params_free(dcp, !!error); 1648 1649 return (error); 1650 } 1651 1652 static int 1653 zfs_ioc_pool_destroy(zfs_cmd_t *zc) 1654 { 1655 int error; 1656 zfs_log_history(zc); 1657 error = spa_destroy(zc->zc_name); 1658 if (error == 0) 1659 zvol_remove_minors(zc->zc_name); 1660 return (error); 1661 } 1662 1663 static int 1664 zfs_ioc_pool_import(zfs_cmd_t *zc) 1665 { 1666 nvlist_t *config, *props = NULL; 1667 uint64_t guid; 1668 int error; 1669 1670 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1671 zc->zc_iflags, &config)) != 0) 1672 return (error); 1673 1674 if (zc->zc_nvlist_src_size != 0 && (error = 1675 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1676 zc->zc_iflags, &props))) { 1677 nvlist_free(config); 1678 return (error); 1679 } 1680 1681 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 || 1682 guid != zc->zc_guid) 1683 error = SET_ERROR(EINVAL); 1684 else 1685 error = spa_import(zc->zc_name, config, props, zc->zc_cookie); 1686 1687 if (zc->zc_nvlist_dst != 0) { 1688 int err; 1689 1690 if ((err = put_nvlist(zc, config)) != 0) 1691 error = err; 1692 } 1693 1694 nvlist_free(config); 1695 1696 nvlist_free(props); 1697 1698 return (error); 1699 } 1700 1701 static int 1702 zfs_ioc_pool_export(zfs_cmd_t *zc) 1703 { 1704 int error; 1705 boolean_t force = (boolean_t)zc->zc_cookie; 1706 boolean_t hardforce = (boolean_t)zc->zc_guid; 1707 1708 zfs_log_history(zc); 1709 error = spa_export(zc->zc_name, NULL, force, hardforce); 1710 if (error == 0) 1711 zvol_remove_minors(zc->zc_name); 1712 return (error); 1713 } 1714 1715 static int 1716 zfs_ioc_pool_configs(zfs_cmd_t *zc) 1717 { 1718 nvlist_t *configs; 1719 int error; 1720 1721 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL) 1722 return (SET_ERROR(EEXIST)); 1723 1724 error = put_nvlist(zc, configs); 1725 1726 nvlist_free(configs); 1727 1728 return (error); 1729 } 1730 1731 /* 1732 * inputs: 1733 * zc_name name of the pool 1734 * 1735 * outputs: 1736 * zc_cookie real errno 1737 * zc_nvlist_dst config nvlist 1738 * zc_nvlist_dst_size size of config nvlist 1739 */ 1740 static int 1741 zfs_ioc_pool_stats(zfs_cmd_t *zc) 1742 { 1743 nvlist_t *config; 1744 int error; 1745 int ret = 0; 1746 1747 error = spa_get_stats(zc->zc_name, &config, zc->zc_value, 1748 sizeof (zc->zc_value)); 1749 1750 if (config != NULL) { 1751 ret = put_nvlist(zc, config); 1752 nvlist_free(config); 1753 1754 /* 1755 * The config may be present even if 'error' is non-zero. 1756 * In this case we return success, and preserve the real errno 1757 * in 'zc_cookie'. 1758 */ 1759 zc->zc_cookie = error; 1760 } else { 1761 ret = error; 1762 } 1763 1764 return (ret); 1765 } 1766 1767 /* 1768 * Try to import the given pool, returning pool stats as appropriate so that 1769 * user land knows which devices are available and overall pool health. 1770 */ 1771 static int 1772 zfs_ioc_pool_tryimport(zfs_cmd_t *zc) 1773 { 1774 nvlist_t *tryconfig, *config; 1775 int error; 1776 1777 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1778 zc->zc_iflags, &tryconfig)) != 0) 1779 return (error); 1780 1781 config = spa_tryimport(tryconfig); 1782 1783 nvlist_free(tryconfig); 1784 1785 if (config == NULL) 1786 return (SET_ERROR(EINVAL)); 1787 1788 error = put_nvlist(zc, config); 1789 nvlist_free(config); 1790 1791 return (error); 1792 } 1793 1794 /* 1795 * inputs: 1796 * zc_name name of the pool 1797 * zc_cookie scan func (pool_scan_func_t) 1798 * zc_flags scrub pause/resume flag (pool_scrub_cmd_t) 1799 */ 1800 static int 1801 zfs_ioc_pool_scan(zfs_cmd_t *zc) 1802 { 1803 spa_t *spa; 1804 int error; 1805 1806 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1807 return (error); 1808 1809 if (zc->zc_flags >= POOL_SCRUB_FLAGS_END) 1810 return (SET_ERROR(EINVAL)); 1811 1812 if (zc->zc_flags == POOL_SCRUB_PAUSE) 1813 error = spa_scrub_pause_resume(spa, POOL_SCRUB_PAUSE); 1814 else if (zc->zc_cookie == POOL_SCAN_NONE) 1815 error = spa_scan_stop(spa); 1816 else 1817 error = spa_scan(spa, zc->zc_cookie); 1818 1819 spa_close(spa, FTAG); 1820 1821 return (error); 1822 } 1823 1824 static int 1825 zfs_ioc_pool_freeze(zfs_cmd_t *zc) 1826 { 1827 spa_t *spa; 1828 int error; 1829 1830 error = spa_open(zc->zc_name, &spa, FTAG); 1831 if (error == 0) { 1832 spa_freeze(spa); 1833 spa_close(spa, FTAG); 1834 } 1835 return (error); 1836 } 1837 1838 static int 1839 zfs_ioc_pool_upgrade(zfs_cmd_t *zc) 1840 { 1841 spa_t *spa; 1842 int error; 1843 1844 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1845 return (error); 1846 1847 if (zc->zc_cookie < spa_version(spa) || 1848 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) { 1849 spa_close(spa, FTAG); 1850 return (SET_ERROR(EINVAL)); 1851 } 1852 1853 spa_upgrade(spa, zc->zc_cookie); 1854 spa_close(spa, FTAG); 1855 1856 return (error); 1857 } 1858 1859 static int 1860 zfs_ioc_pool_get_history(zfs_cmd_t *zc) 1861 { 1862 spa_t *spa; 1863 char *hist_buf; 1864 uint64_t size; 1865 int error; 1866 1867 if ((size = zc->zc_history_len) == 0) 1868 return (SET_ERROR(EINVAL)); 1869 1870 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1871 return (error); 1872 1873 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) { 1874 spa_close(spa, FTAG); 1875 return (SET_ERROR(ENOTSUP)); 1876 } 1877 1878 hist_buf = kmem_alloc(size, KM_SLEEP); 1879 if ((error = spa_history_get(spa, &zc->zc_history_offset, 1880 &zc->zc_history_len, hist_buf)) == 0) { 1881 error = ddi_copyout(hist_buf, 1882 (void *)(uintptr_t)zc->zc_history, 1883 zc->zc_history_len, zc->zc_iflags); 1884 } 1885 1886 spa_close(spa, FTAG); 1887 kmem_free(hist_buf, size); 1888 return (error); 1889 } 1890 1891 static int 1892 zfs_ioc_pool_reguid(zfs_cmd_t *zc) 1893 { 1894 spa_t *spa; 1895 int error; 1896 1897 error = spa_open(zc->zc_name, &spa, FTAG); 1898 if (error == 0) { 1899 error = spa_change_guid(spa); 1900 spa_close(spa, FTAG); 1901 } 1902 return (error); 1903 } 1904 1905 static int 1906 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc) 1907 { 1908 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value)); 1909 } 1910 1911 /* 1912 * inputs: 1913 * zc_name name of filesystem 1914 * zc_obj object to find 1915 * 1916 * outputs: 1917 * zc_value name of object 1918 */ 1919 static int 1920 zfs_ioc_obj_to_path(zfs_cmd_t *zc) 1921 { 1922 objset_t *os; 1923 int error; 1924 1925 /* XXX reading from objset not owned */ 1926 if ((error = dmu_objset_hold_flags(zc->zc_name, B_TRUE, 1927 FTAG, &os)) != 0) 1928 return (error); 1929 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1930 dmu_objset_rele_flags(os, B_TRUE, FTAG); 1931 return (SET_ERROR(EINVAL)); 1932 } 1933 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value, 1934 sizeof (zc->zc_value)); 1935 dmu_objset_rele_flags(os, B_TRUE, FTAG); 1936 1937 return (error); 1938 } 1939 1940 /* 1941 * inputs: 1942 * zc_name name of filesystem 1943 * zc_obj object to find 1944 * 1945 * outputs: 1946 * zc_stat stats on object 1947 * zc_value path to object 1948 */ 1949 static int 1950 zfs_ioc_obj_to_stats(zfs_cmd_t *zc) 1951 { 1952 objset_t *os; 1953 int error; 1954 1955 /* XXX reading from objset not owned */ 1956 if ((error = dmu_objset_hold_flags(zc->zc_name, B_TRUE, 1957 FTAG, &os)) != 0) 1958 return (error); 1959 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1960 dmu_objset_rele_flags(os, B_TRUE, FTAG); 1961 return (SET_ERROR(EINVAL)); 1962 } 1963 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value, 1964 sizeof (zc->zc_value)); 1965 dmu_objset_rele_flags(os, B_TRUE, FTAG); 1966 1967 return (error); 1968 } 1969 1970 static int 1971 zfs_ioc_vdev_add(zfs_cmd_t *zc) 1972 { 1973 spa_t *spa; 1974 int error; 1975 nvlist_t *config; 1976 1977 error = spa_open(zc->zc_name, &spa, FTAG); 1978 if (error != 0) 1979 return (error); 1980 1981 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1982 zc->zc_iflags, &config); 1983 1984 if (error == 0) { 1985 error = spa_vdev_add(spa, config); 1986 nvlist_free(config); 1987 } 1988 spa_close(spa, FTAG); 1989 return (error); 1990 } 1991 1992 /* 1993 * inputs: 1994 * zc_name name of the pool 1995 * zc_guid guid of vdev to remove 1996 * zc_cookie cancel removal 1997 */ 1998 static int 1999 zfs_ioc_vdev_remove(zfs_cmd_t *zc) 2000 { 2001 spa_t *spa; 2002 int error; 2003 2004 error = spa_open(zc->zc_name, &spa, FTAG); 2005 if (error != 0) 2006 return (error); 2007 if (zc->zc_cookie != 0) { 2008 error = spa_vdev_remove_cancel(spa); 2009 } else { 2010 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE); 2011 } 2012 spa_close(spa, FTAG); 2013 return (error); 2014 } 2015 2016 static int 2017 zfs_ioc_vdev_set_state(zfs_cmd_t *zc) 2018 { 2019 spa_t *spa; 2020 int error; 2021 vdev_state_t newstate = VDEV_STATE_UNKNOWN; 2022 2023 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 2024 return (error); 2025 switch (zc->zc_cookie) { 2026 case VDEV_STATE_ONLINE: 2027 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate); 2028 break; 2029 2030 case VDEV_STATE_OFFLINE: 2031 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj); 2032 break; 2033 2034 case VDEV_STATE_FAULTED: 2035 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED && 2036 zc->zc_obj != VDEV_AUX_EXTERNAL) 2037 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED; 2038 2039 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj); 2040 break; 2041 2042 case VDEV_STATE_DEGRADED: 2043 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED && 2044 zc->zc_obj != VDEV_AUX_EXTERNAL) 2045 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED; 2046 2047 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj); 2048 break; 2049 2050 default: 2051 error = SET_ERROR(EINVAL); 2052 } 2053 zc->zc_cookie = newstate; 2054 spa_close(spa, FTAG); 2055 return (error); 2056 } 2057 2058 static int 2059 zfs_ioc_vdev_attach(zfs_cmd_t *zc) 2060 { 2061 spa_t *spa; 2062 int replacing = zc->zc_cookie; 2063 nvlist_t *config; 2064 int error; 2065 2066 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 2067 return (error); 2068 2069 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 2070 zc->zc_iflags, &config)) == 0) { 2071 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing); 2072 nvlist_free(config); 2073 } 2074 2075 spa_close(spa, FTAG); 2076 return (error); 2077 } 2078 2079 static int 2080 zfs_ioc_vdev_detach(zfs_cmd_t *zc) 2081 { 2082 spa_t *spa; 2083 int error; 2084 2085 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 2086 return (error); 2087 2088 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE); 2089 2090 spa_close(spa, FTAG); 2091 return (error); 2092 } 2093 2094 static int 2095 zfs_ioc_vdev_split(zfs_cmd_t *zc) 2096 { 2097 spa_t *spa; 2098 nvlist_t *config, *props = NULL; 2099 int error; 2100 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT); 2101 2102 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 2103 return (error); 2104 2105 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 2106 zc->zc_iflags, &config)) { 2107 spa_close(spa, FTAG); 2108 return (error); 2109 } 2110 2111 if (zc->zc_nvlist_src_size != 0 && (error = 2112 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2113 zc->zc_iflags, &props))) { 2114 spa_close(spa, FTAG); 2115 nvlist_free(config); 2116 return (error); 2117 } 2118 2119 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp); 2120 2121 spa_close(spa, FTAG); 2122 2123 nvlist_free(config); 2124 nvlist_free(props); 2125 2126 return (error); 2127 } 2128 2129 static int 2130 zfs_ioc_vdev_setpath(zfs_cmd_t *zc) 2131 { 2132 spa_t *spa; 2133 char *path = zc->zc_value; 2134 uint64_t guid = zc->zc_guid; 2135 int error; 2136 2137 error = spa_open(zc->zc_name, &spa, FTAG); 2138 if (error != 0) 2139 return (error); 2140 2141 error = spa_vdev_setpath(spa, guid, path); 2142 spa_close(spa, FTAG); 2143 return (error); 2144 } 2145 2146 static int 2147 zfs_ioc_vdev_setfru(zfs_cmd_t *zc) 2148 { 2149 spa_t *spa; 2150 char *fru = zc->zc_value; 2151 uint64_t guid = zc->zc_guid; 2152 int error; 2153 2154 error = spa_open(zc->zc_name, &spa, FTAG); 2155 if (error != 0) 2156 return (error); 2157 2158 error = spa_vdev_setfru(spa, guid, fru); 2159 spa_close(spa, FTAG); 2160 return (error); 2161 } 2162 2163 static int 2164 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os) 2165 { 2166 int error = 0; 2167 nvlist_t *nv; 2168 2169 dmu_objset_fast_stat(os, &zc->zc_objset_stats); 2170 2171 if (zc->zc_nvlist_dst != 0 && 2172 (error = dsl_prop_get_all(os, &nv)) == 0) { 2173 dmu_objset_stats(os, nv); 2174 /* 2175 * NB: zvol_get_stats() will read the objset contents, 2176 * which we aren't supposed to do with a 2177 * DS_MODE_USER hold, because it could be 2178 * inconsistent. So this is a bit of a workaround... 2179 * XXX reading with out owning 2180 */ 2181 if (!zc->zc_objset_stats.dds_inconsistent && 2182 dmu_objset_type(os) == DMU_OST_ZVOL) { 2183 error = zvol_get_stats(os, nv); 2184 if (error == EIO) 2185 return (error); 2186 VERIFY0(error); 2187 } 2188 error = put_nvlist(zc, nv); 2189 nvlist_free(nv); 2190 } 2191 2192 return (error); 2193 } 2194 2195 /* 2196 * inputs: 2197 * zc_name name of filesystem 2198 * zc_nvlist_dst_size size of buffer for property nvlist 2199 * 2200 * outputs: 2201 * zc_objset_stats stats 2202 * zc_nvlist_dst property nvlist 2203 * zc_nvlist_dst_size size of property nvlist 2204 */ 2205 static int 2206 zfs_ioc_objset_stats(zfs_cmd_t *zc) 2207 { 2208 objset_t *os; 2209 int error; 2210 2211 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 2212 if (error == 0) { 2213 error = zfs_ioc_objset_stats_impl(zc, os); 2214 dmu_objset_rele(os, FTAG); 2215 } 2216 2217 return (error); 2218 } 2219 2220 /* 2221 * inputs: 2222 * zc_name name of filesystem 2223 * zc_nvlist_dst_size size of buffer for property nvlist 2224 * 2225 * outputs: 2226 * zc_nvlist_dst received property nvlist 2227 * zc_nvlist_dst_size size of received property nvlist 2228 * 2229 * Gets received properties (distinct from local properties on or after 2230 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from 2231 * local property values. 2232 */ 2233 static int 2234 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc) 2235 { 2236 int error = 0; 2237 nvlist_t *nv; 2238 2239 /* 2240 * Without this check, we would return local property values if the 2241 * caller has not already received properties on or after 2242 * SPA_VERSION_RECVD_PROPS. 2243 */ 2244 if (!dsl_prop_get_hasrecvd(zc->zc_name)) 2245 return (SET_ERROR(ENOTSUP)); 2246 2247 if (zc->zc_nvlist_dst != 0 && 2248 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) { 2249 error = put_nvlist(zc, nv); 2250 nvlist_free(nv); 2251 } 2252 2253 return (error); 2254 } 2255 2256 static int 2257 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop) 2258 { 2259 uint64_t value; 2260 int error; 2261 2262 /* 2263 * zfs_get_zplprop() will either find a value or give us 2264 * the default value (if there is one). 2265 */ 2266 if ((error = zfs_get_zplprop(os, prop, &value)) != 0) 2267 return (error); 2268 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0); 2269 return (0); 2270 } 2271 2272 /* 2273 * inputs: 2274 * zc_name name of filesystem 2275 * zc_nvlist_dst_size size of buffer for zpl property nvlist 2276 * 2277 * outputs: 2278 * zc_nvlist_dst zpl property nvlist 2279 * zc_nvlist_dst_size size of zpl property nvlist 2280 */ 2281 static int 2282 zfs_ioc_objset_zplprops(zfs_cmd_t *zc) 2283 { 2284 objset_t *os; 2285 int err; 2286 2287 /* XXX reading without owning */ 2288 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os)) 2289 return (err); 2290 2291 dmu_objset_fast_stat(os, &zc->zc_objset_stats); 2292 2293 /* 2294 * NB: nvl_add_zplprop() will read the objset contents, 2295 * which we aren't supposed to do with a DS_MODE_USER 2296 * hold, because it could be inconsistent. 2297 */ 2298 if (zc->zc_nvlist_dst != 0 && 2299 !zc->zc_objset_stats.dds_inconsistent && 2300 dmu_objset_type(os) == DMU_OST_ZFS) { 2301 nvlist_t *nv; 2302 2303 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2304 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 && 2305 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 && 2306 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 && 2307 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0) 2308 err = put_nvlist(zc, nv); 2309 nvlist_free(nv); 2310 } else { 2311 err = SET_ERROR(ENOENT); 2312 } 2313 dmu_objset_rele(os, FTAG); 2314 return (err); 2315 } 2316 2317 static boolean_t 2318 dataset_name_hidden(const char *name) 2319 { 2320 /* 2321 * Skip over datasets that are not visible in this zone, 2322 * internal datasets (which have a $ in their name), and 2323 * temporary datasets (which have a % in their name). 2324 */ 2325 if (strchr(name, '$') != NULL) 2326 return (B_TRUE); 2327 if (strchr(name, '%') != NULL) 2328 return (B_TRUE); 2329 if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL)) 2330 return (B_TRUE); 2331 return (B_FALSE); 2332 } 2333 2334 /* 2335 * inputs: 2336 * zc_name name of filesystem 2337 * zc_cookie zap cursor 2338 * zc_nvlist_dst_size size of buffer for property nvlist 2339 * 2340 * outputs: 2341 * zc_name name of next filesystem 2342 * zc_cookie zap cursor 2343 * zc_objset_stats stats 2344 * zc_nvlist_dst property nvlist 2345 * zc_nvlist_dst_size size of property nvlist 2346 */ 2347 static int 2348 zfs_ioc_dataset_list_next(zfs_cmd_t *zc) 2349 { 2350 objset_t *os; 2351 int error; 2352 char *p; 2353 size_t orig_len = strlen(zc->zc_name); 2354 2355 top: 2356 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) { 2357 if (error == ENOENT) 2358 error = SET_ERROR(ESRCH); 2359 return (error); 2360 } 2361 2362 p = strrchr(zc->zc_name, '/'); 2363 if (p == NULL || p[1] != '\0') 2364 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name)); 2365 p = zc->zc_name + strlen(zc->zc_name); 2366 2367 do { 2368 error = dmu_dir_list_next(os, 2369 sizeof (zc->zc_name) - (p - zc->zc_name), p, 2370 NULL, &zc->zc_cookie); 2371 if (error == ENOENT) 2372 error = SET_ERROR(ESRCH); 2373 } while (error == 0 && dataset_name_hidden(zc->zc_name)); 2374 dmu_objset_rele(os, FTAG); 2375 2376 /* 2377 * If it's an internal dataset (ie. with a '$' in its name), 2378 * don't try to get stats for it, otherwise we'll return ENOENT. 2379 */ 2380 if (error == 0 && strchr(zc->zc_name, '$') == NULL) { 2381 error = zfs_ioc_objset_stats(zc); /* fill in the stats */ 2382 if (error == ENOENT) { 2383 /* We lost a race with destroy, get the next one. */ 2384 zc->zc_name[orig_len] = '\0'; 2385 goto top; 2386 } 2387 } 2388 return (error); 2389 } 2390 2391 /* 2392 * inputs: 2393 * zc_name name of filesystem 2394 * zc_cookie zap cursor 2395 * zc_nvlist_dst_size size of buffer for property nvlist 2396 * zc_simple when set, only name is requested 2397 * 2398 * outputs: 2399 * zc_name name of next snapshot 2400 * zc_objset_stats stats 2401 * zc_nvlist_dst property nvlist 2402 * zc_nvlist_dst_size size of property nvlist 2403 */ 2404 static int 2405 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc) 2406 { 2407 objset_t *os; 2408 int error; 2409 2410 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 2411 if (error != 0) { 2412 return (error == ENOENT ? ESRCH : error); 2413 } 2414 2415 /* 2416 * A dataset name of maximum length cannot have any snapshots, 2417 * so exit immediately. 2418 */ 2419 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= 2420 ZFS_MAX_DATASET_NAME_LEN) { 2421 dmu_objset_rele(os, FTAG); 2422 return (SET_ERROR(ESRCH)); 2423 } 2424 2425 error = dmu_snapshot_list_next(os, 2426 sizeof (zc->zc_name) - strlen(zc->zc_name), 2427 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie, 2428 NULL); 2429 2430 if (error == 0 && !zc->zc_simple) { 2431 dsl_dataset_t *ds; 2432 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool; 2433 2434 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds); 2435 if (error == 0) { 2436 objset_t *ossnap; 2437 2438 error = dmu_objset_from_ds(ds, &ossnap); 2439 if (error == 0) 2440 error = zfs_ioc_objset_stats_impl(zc, ossnap); 2441 dsl_dataset_rele(ds, FTAG); 2442 } 2443 } else if (error == ENOENT) { 2444 error = SET_ERROR(ESRCH); 2445 } 2446 2447 dmu_objset_rele(os, FTAG); 2448 /* if we failed, undo the @ that we tacked on to zc_name */ 2449 if (error != 0) 2450 *strchr(zc->zc_name, '@') = '\0'; 2451 return (error); 2452 } 2453 2454 static int 2455 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair) 2456 { 2457 const char *propname = nvpair_name(pair); 2458 uint64_t *valary; 2459 unsigned int vallen; 2460 const char *domain; 2461 char *dash; 2462 zfs_userquota_prop_t type; 2463 uint64_t rid; 2464 uint64_t quota; 2465 zfsvfs_t *zfsvfs; 2466 int err; 2467 2468 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2469 nvlist_t *attrs; 2470 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 2471 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2472 &pair) != 0) 2473 return (SET_ERROR(EINVAL)); 2474 } 2475 2476 /* 2477 * A correctly constructed propname is encoded as 2478 * userquota@<rid>-<domain>. 2479 */ 2480 if ((dash = strchr(propname, '-')) == NULL || 2481 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 || 2482 vallen != 3) 2483 return (SET_ERROR(EINVAL)); 2484 2485 domain = dash + 1; 2486 type = valary[0]; 2487 rid = valary[1]; 2488 quota = valary[2]; 2489 2490 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE); 2491 if (err == 0) { 2492 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota); 2493 zfsvfs_rele(zfsvfs, FTAG); 2494 } 2495 2496 return (err); 2497 } 2498 2499 /* 2500 * If the named property is one that has a special function to set its value, 2501 * return 0 on success and a positive error code on failure; otherwise if it is 2502 * not one of the special properties handled by this function, return -1. 2503 * 2504 * XXX: It would be better for callers of the property interface if we handled 2505 * these special cases in dsl_prop.c (in the dsl layer). 2506 */ 2507 static int 2508 zfs_prop_set_special(const char *dsname, zprop_source_t source, 2509 nvpair_t *pair) 2510 { 2511 const char *propname = nvpair_name(pair); 2512 zfs_prop_t prop = zfs_name_to_prop(propname); 2513 uint64_t intval = 0; 2514 char *strval = NULL; 2515 int err = -1; 2516 2517 if (prop == ZPROP_INVAL) { 2518 if (zfs_prop_userquota(propname)) 2519 return (zfs_prop_set_userquota(dsname, pair)); 2520 return (-1); 2521 } 2522 2523 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2524 nvlist_t *attrs; 2525 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 2526 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2527 &pair) == 0); 2528 } 2529 2530 /* all special properties are numeric except for keylocation */ 2531 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING) { 2532 strval = fnvpair_value_string(pair); 2533 } else { 2534 intval = fnvpair_value_uint64(pair); 2535 } 2536 2537 switch (prop) { 2538 case ZFS_PROP_QUOTA: 2539 err = dsl_dir_set_quota(dsname, source, intval); 2540 break; 2541 case ZFS_PROP_REFQUOTA: 2542 err = dsl_dataset_set_refquota(dsname, source, intval); 2543 break; 2544 case ZFS_PROP_FILESYSTEM_LIMIT: 2545 case ZFS_PROP_SNAPSHOT_LIMIT: 2546 if (intval == UINT64_MAX) { 2547 /* clearing the limit, just do it */ 2548 err = 0; 2549 } else { 2550 err = dsl_dir_activate_fs_ss_limit(dsname); 2551 } 2552 /* 2553 * Set err to -1 to force the zfs_set_prop_nvlist code down the 2554 * default path to set the value in the nvlist. 2555 */ 2556 if (err == 0) 2557 err = -1; 2558 break; 2559 case ZFS_PROP_KEYLOCATION: 2560 err = dsl_crypto_can_set_keylocation(dsname, strval); 2561 2562 /* 2563 * Set err to -1 to force the zfs_set_prop_nvlist code down the 2564 * default path to set the value in the nvlist. 2565 */ 2566 if (err == 0) 2567 err = -1; 2568 break; 2569 case ZFS_PROP_RESERVATION: 2570 err = dsl_dir_set_reservation(dsname, source, intval); 2571 break; 2572 case ZFS_PROP_REFRESERVATION: 2573 err = dsl_dataset_set_refreservation(dsname, source, intval); 2574 break; 2575 case ZFS_PROP_VOLSIZE: 2576 err = zvol_set_volsize(dsname, intval); 2577 break; 2578 case ZFS_PROP_VERSION: 2579 { 2580 zfsvfs_t *zfsvfs; 2581 2582 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0) 2583 break; 2584 2585 err = zfs_set_version(zfsvfs, intval); 2586 zfsvfs_rele(zfsvfs, FTAG); 2587 2588 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) { 2589 zfs_cmd_t *zc; 2590 2591 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP); 2592 (void) strcpy(zc->zc_name, dsname); 2593 (void) zfs_ioc_userspace_upgrade(zc); 2594 (void) zfs_ioc_id_quota_upgrade(zc); 2595 kmem_free(zc, sizeof (zfs_cmd_t)); 2596 } 2597 break; 2598 } 2599 default: 2600 err = -1; 2601 } 2602 2603 return (err); 2604 } 2605 2606 /* 2607 * This function is best effort. If it fails to set any of the given properties, 2608 * it continues to set as many as it can and returns the last error 2609 * encountered. If the caller provides a non-NULL errlist, it will be filled in 2610 * with the list of names of all the properties that failed along with the 2611 * corresponding error numbers. 2612 * 2613 * If every property is set successfully, zero is returned and errlist is not 2614 * modified. 2615 */ 2616 int 2617 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl, 2618 nvlist_t *errlist) 2619 { 2620 nvpair_t *pair; 2621 nvpair_t *propval; 2622 int rv = 0; 2623 uint64_t intval; 2624 char *strval; 2625 nvlist_t *genericnvl = fnvlist_alloc(); 2626 nvlist_t *retrynvl = fnvlist_alloc(); 2627 2628 retry: 2629 pair = NULL; 2630 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { 2631 const char *propname = nvpair_name(pair); 2632 zfs_prop_t prop = zfs_name_to_prop(propname); 2633 int err = 0; 2634 2635 /* decode the property value */ 2636 propval = pair; 2637 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2638 nvlist_t *attrs; 2639 attrs = fnvpair_value_nvlist(pair); 2640 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2641 &propval) != 0) 2642 err = SET_ERROR(EINVAL); 2643 } 2644 2645 /* Validate value type */ 2646 if (err == 0 && source == ZPROP_SRC_INHERITED) { 2647 /* inherited properties are expected to be booleans */ 2648 if (nvpair_type(propval) != DATA_TYPE_BOOLEAN) 2649 err = SET_ERROR(EINVAL); 2650 } else if (err == 0 && prop == ZPROP_INVAL) { 2651 if (zfs_prop_user(propname)) { 2652 if (nvpair_type(propval) != DATA_TYPE_STRING) 2653 err = SET_ERROR(EINVAL); 2654 } else if (zfs_prop_userquota(propname)) { 2655 if (nvpair_type(propval) != 2656 DATA_TYPE_UINT64_ARRAY) 2657 err = SET_ERROR(EINVAL); 2658 } else { 2659 err = SET_ERROR(EINVAL); 2660 } 2661 } else if (err == 0) { 2662 if (nvpair_type(propval) == DATA_TYPE_STRING) { 2663 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING) 2664 err = SET_ERROR(EINVAL); 2665 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) { 2666 const char *unused; 2667 2668 intval = fnvpair_value_uint64(propval); 2669 2670 switch (zfs_prop_get_type(prop)) { 2671 case PROP_TYPE_NUMBER: 2672 break; 2673 case PROP_TYPE_STRING: 2674 err = SET_ERROR(EINVAL); 2675 break; 2676 case PROP_TYPE_INDEX: 2677 if (zfs_prop_index_to_string(prop, 2678 intval, &unused) != 0) 2679 err = SET_ERROR(EINVAL); 2680 break; 2681 default: 2682 cmn_err(CE_PANIC, 2683 "unknown property type"); 2684 } 2685 } else { 2686 err = SET_ERROR(EINVAL); 2687 } 2688 } 2689 2690 /* Validate permissions */ 2691 if (err == 0) 2692 err = zfs_check_settable(dsname, pair, CRED()); 2693 2694 if (err == 0) { 2695 if (source == ZPROP_SRC_INHERITED) 2696 err = -1; /* does not need special handling */ 2697 else 2698 err = zfs_prop_set_special(dsname, source, 2699 pair); 2700 if (err == -1) { 2701 /* 2702 * For better performance we build up a list of 2703 * properties to set in a single transaction. 2704 */ 2705 err = nvlist_add_nvpair(genericnvl, pair); 2706 } else if (err != 0 && nvl != retrynvl) { 2707 /* 2708 * This may be a spurious error caused by 2709 * receiving quota and reservation out of order. 2710 * Try again in a second pass. 2711 */ 2712 err = nvlist_add_nvpair(retrynvl, pair); 2713 } 2714 } 2715 2716 if (err != 0) { 2717 if (errlist != NULL) 2718 fnvlist_add_int32(errlist, propname, err); 2719 rv = err; 2720 } 2721 } 2722 2723 if (nvl != retrynvl && !nvlist_empty(retrynvl)) { 2724 nvl = retrynvl; 2725 goto retry; 2726 } 2727 2728 if (!nvlist_empty(genericnvl) && 2729 dsl_props_set(dsname, source, genericnvl) != 0) { 2730 /* 2731 * If this fails, we still want to set as many properties as we 2732 * can, so try setting them individually. 2733 */ 2734 pair = NULL; 2735 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) { 2736 const char *propname = nvpair_name(pair); 2737 int err = 0; 2738 2739 propval = pair; 2740 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2741 nvlist_t *attrs; 2742 attrs = fnvpair_value_nvlist(pair); 2743 propval = fnvlist_lookup_nvpair(attrs, 2744 ZPROP_VALUE); 2745 } 2746 2747 if (nvpair_type(propval) == DATA_TYPE_STRING) { 2748 strval = fnvpair_value_string(propval); 2749 err = dsl_prop_set_string(dsname, propname, 2750 source, strval); 2751 } else if (nvpair_type(propval) == DATA_TYPE_BOOLEAN) { 2752 err = dsl_prop_inherit(dsname, propname, 2753 source); 2754 } else { 2755 intval = fnvpair_value_uint64(propval); 2756 err = dsl_prop_set_int(dsname, propname, source, 2757 intval); 2758 } 2759 2760 if (err != 0) { 2761 if (errlist != NULL) { 2762 fnvlist_add_int32(errlist, propname, 2763 err); 2764 } 2765 rv = err; 2766 } 2767 } 2768 } 2769 nvlist_free(genericnvl); 2770 nvlist_free(retrynvl); 2771 2772 return (rv); 2773 } 2774 2775 /* 2776 * Check that all the properties are valid user properties. 2777 */ 2778 static int 2779 zfs_check_userprops(const char *fsname, nvlist_t *nvl) 2780 { 2781 nvpair_t *pair = NULL; 2782 int error = 0; 2783 2784 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { 2785 const char *propname = nvpair_name(pair); 2786 2787 if (!zfs_prop_user(propname) || 2788 nvpair_type(pair) != DATA_TYPE_STRING) 2789 return (SET_ERROR(EINVAL)); 2790 2791 if (error = zfs_secpolicy_write_perms(fsname, 2792 ZFS_DELEG_PERM_USERPROP, CRED())) 2793 return (error); 2794 2795 if (strlen(propname) >= ZAP_MAXNAMELEN) 2796 return (SET_ERROR(ENAMETOOLONG)); 2797 2798 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN) 2799 return (E2BIG); 2800 } 2801 return (0); 2802 } 2803 2804 static void 2805 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops) 2806 { 2807 nvpair_t *pair; 2808 2809 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2810 2811 pair = NULL; 2812 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) { 2813 if (nvlist_exists(skipped, nvpair_name(pair))) 2814 continue; 2815 2816 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0); 2817 } 2818 } 2819 2820 static int 2821 clear_received_props(const char *dsname, nvlist_t *props, 2822 nvlist_t *skipped) 2823 { 2824 int err = 0; 2825 nvlist_t *cleared_props = NULL; 2826 props_skip(props, skipped, &cleared_props); 2827 if (!nvlist_empty(cleared_props)) { 2828 /* 2829 * Acts on local properties until the dataset has received 2830 * properties at least once on or after SPA_VERSION_RECVD_PROPS. 2831 */ 2832 zprop_source_t flags = (ZPROP_SRC_NONE | 2833 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0)); 2834 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL); 2835 } 2836 nvlist_free(cleared_props); 2837 return (err); 2838 } 2839 2840 /* 2841 * inputs: 2842 * zc_name name of filesystem 2843 * zc_value name of property to set 2844 * zc_nvlist_src{_size} nvlist of properties to apply 2845 * zc_cookie received properties flag 2846 * 2847 * outputs: 2848 * zc_nvlist_dst{_size} error for each unapplied received property 2849 */ 2850 static int 2851 zfs_ioc_set_prop(zfs_cmd_t *zc) 2852 { 2853 nvlist_t *nvl; 2854 boolean_t received = zc->zc_cookie; 2855 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED : 2856 ZPROP_SRC_LOCAL); 2857 nvlist_t *errors; 2858 int error; 2859 2860 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2861 zc->zc_iflags, &nvl)) != 0) 2862 return (error); 2863 2864 if (received) { 2865 nvlist_t *origprops; 2866 2867 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) { 2868 (void) clear_received_props(zc->zc_name, 2869 origprops, nvl); 2870 nvlist_free(origprops); 2871 } 2872 2873 error = dsl_prop_set_hasrecvd(zc->zc_name); 2874 } 2875 2876 errors = fnvlist_alloc(); 2877 if (error == 0) 2878 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors); 2879 2880 if (zc->zc_nvlist_dst != 0 && errors != NULL) { 2881 (void) put_nvlist(zc, errors); 2882 } 2883 2884 nvlist_free(errors); 2885 nvlist_free(nvl); 2886 return (error); 2887 } 2888 2889 /* 2890 * inputs: 2891 * zc_name name of filesystem 2892 * zc_value name of property to inherit 2893 * zc_cookie revert to received value if TRUE 2894 * 2895 * outputs: none 2896 */ 2897 static int 2898 zfs_ioc_inherit_prop(zfs_cmd_t *zc) 2899 { 2900 const char *propname = zc->zc_value; 2901 zfs_prop_t prop = zfs_name_to_prop(propname); 2902 boolean_t received = zc->zc_cookie; 2903 zprop_source_t source = (received 2904 ? ZPROP_SRC_NONE /* revert to received value, if any */ 2905 : ZPROP_SRC_INHERITED); /* explicitly inherit */ 2906 2907 if (received) { 2908 nvlist_t *dummy; 2909 nvpair_t *pair; 2910 zprop_type_t type; 2911 int err; 2912 2913 /* 2914 * zfs_prop_set_special() expects properties in the form of an 2915 * nvpair with type info. 2916 */ 2917 if (prop == ZPROP_INVAL) { 2918 if (!zfs_prop_user(propname)) 2919 return (SET_ERROR(EINVAL)); 2920 2921 type = PROP_TYPE_STRING; 2922 } else if (prop == ZFS_PROP_VOLSIZE || 2923 prop == ZFS_PROP_VERSION) { 2924 return (SET_ERROR(EINVAL)); 2925 } else { 2926 type = zfs_prop_get_type(prop); 2927 } 2928 2929 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2930 2931 switch (type) { 2932 case PROP_TYPE_STRING: 2933 VERIFY(0 == nvlist_add_string(dummy, propname, "")); 2934 break; 2935 case PROP_TYPE_NUMBER: 2936 case PROP_TYPE_INDEX: 2937 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0)); 2938 break; 2939 default: 2940 nvlist_free(dummy); 2941 return (SET_ERROR(EINVAL)); 2942 } 2943 2944 pair = nvlist_next_nvpair(dummy, NULL); 2945 err = zfs_prop_set_special(zc->zc_name, source, pair); 2946 nvlist_free(dummy); 2947 if (err != -1) 2948 return (err); /* special property already handled */ 2949 } else { 2950 /* 2951 * Only check this in the non-received case. We want to allow 2952 * 'inherit -S' to revert non-inheritable properties like quota 2953 * and reservation to the received or default values even though 2954 * they are not considered inheritable. 2955 */ 2956 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop)) 2957 return (SET_ERROR(EINVAL)); 2958 } 2959 2960 /* property name has been validated by zfs_secpolicy_inherit_prop() */ 2961 return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source)); 2962 } 2963 2964 static int 2965 zfs_ioc_pool_set_props(zfs_cmd_t *zc) 2966 { 2967 nvlist_t *props; 2968 spa_t *spa; 2969 int error; 2970 nvpair_t *pair; 2971 2972 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2973 zc->zc_iflags, &props)) 2974 return (error); 2975 2976 /* 2977 * If the only property is the configfile, then just do a spa_lookup() 2978 * to handle the faulted case. 2979 */ 2980 pair = nvlist_next_nvpair(props, NULL); 2981 if (pair != NULL && strcmp(nvpair_name(pair), 2982 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 && 2983 nvlist_next_nvpair(props, pair) == NULL) { 2984 mutex_enter(&spa_namespace_lock); 2985 if ((spa = spa_lookup(zc->zc_name)) != NULL) { 2986 spa_configfile_set(spa, props, B_FALSE); 2987 spa_write_cachefile(spa, B_FALSE, B_TRUE); 2988 } 2989 mutex_exit(&spa_namespace_lock); 2990 if (spa != NULL) { 2991 nvlist_free(props); 2992 return (0); 2993 } 2994 } 2995 2996 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { 2997 nvlist_free(props); 2998 return (error); 2999 } 3000 3001 error = spa_prop_set(spa, props); 3002 3003 nvlist_free(props); 3004 spa_close(spa, FTAG); 3005 3006 return (error); 3007 } 3008 3009 static int 3010 zfs_ioc_pool_get_props(zfs_cmd_t *zc) 3011 { 3012 spa_t *spa; 3013 int error; 3014 nvlist_t *nvp = NULL; 3015 3016 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { 3017 /* 3018 * If the pool is faulted, there may be properties we can still 3019 * get (such as altroot and cachefile), so attempt to get them 3020 * anyway. 3021 */ 3022 mutex_enter(&spa_namespace_lock); 3023 if ((spa = spa_lookup(zc->zc_name)) != NULL) 3024 error = spa_prop_get(spa, &nvp); 3025 mutex_exit(&spa_namespace_lock); 3026 } else { 3027 error = spa_prop_get(spa, &nvp); 3028 spa_close(spa, FTAG); 3029 } 3030 3031 if (error == 0 && zc->zc_nvlist_dst != 0) 3032 error = put_nvlist(zc, nvp); 3033 else 3034 error = SET_ERROR(EFAULT); 3035 3036 nvlist_free(nvp); 3037 return (error); 3038 } 3039 3040 /* 3041 * inputs: 3042 * zc_name name of filesystem 3043 * zc_nvlist_src{_size} nvlist of delegated permissions 3044 * zc_perm_action allow/unallow flag 3045 * 3046 * outputs: none 3047 */ 3048 static int 3049 zfs_ioc_set_fsacl(zfs_cmd_t *zc) 3050 { 3051 int error; 3052 nvlist_t *fsaclnv = NULL; 3053 3054 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 3055 zc->zc_iflags, &fsaclnv)) != 0) 3056 return (error); 3057 3058 /* 3059 * Verify nvlist is constructed correctly 3060 */ 3061 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) { 3062 nvlist_free(fsaclnv); 3063 return (SET_ERROR(EINVAL)); 3064 } 3065 3066 /* 3067 * If we don't have PRIV_SYS_MOUNT, then validate 3068 * that user is allowed to hand out each permission in 3069 * the nvlist(s) 3070 */ 3071 3072 error = secpolicy_zfs(CRED()); 3073 if (error != 0) { 3074 if (zc->zc_perm_action == B_FALSE) { 3075 error = dsl_deleg_can_allow(zc->zc_name, 3076 fsaclnv, CRED()); 3077 } else { 3078 error = dsl_deleg_can_unallow(zc->zc_name, 3079 fsaclnv, CRED()); 3080 } 3081 } 3082 3083 if (error == 0) 3084 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action); 3085 3086 nvlist_free(fsaclnv); 3087 return (error); 3088 } 3089 3090 /* 3091 * inputs: 3092 * zc_name name of filesystem 3093 * 3094 * outputs: 3095 * zc_nvlist_src{_size} nvlist of delegated permissions 3096 */ 3097 static int 3098 zfs_ioc_get_fsacl(zfs_cmd_t *zc) 3099 { 3100 nvlist_t *nvp; 3101 int error; 3102 3103 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) { 3104 error = put_nvlist(zc, nvp); 3105 nvlist_free(nvp); 3106 } 3107 3108 return (error); 3109 } 3110 3111 /* ARGSUSED */ 3112 static void 3113 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) 3114 { 3115 zfs_creat_t *zct = arg; 3116 3117 zfs_create_fs(os, cr, zct->zct_zplprops, tx); 3118 } 3119 3120 #define ZFS_PROP_UNDEFINED ((uint64_t)-1) 3121 3122 /* 3123 * inputs: 3124 * os parent objset pointer (NULL if root fs) 3125 * fuids_ok fuids allowed in this version of the spa? 3126 * sa_ok SAs allowed in this version of the spa? 3127 * createprops list of properties requested by creator 3128 * 3129 * outputs: 3130 * zplprops values for the zplprops we attach to the master node object 3131 * is_ci true if requested file system will be purely case-insensitive 3132 * 3133 * Determine the settings for utf8only, normalization and 3134 * casesensitivity. Specific values may have been requested by the 3135 * creator and/or we can inherit values from the parent dataset. If 3136 * the file system is of too early a vintage, a creator can not 3137 * request settings for these properties, even if the requested 3138 * setting is the default value. We don't actually want to create dsl 3139 * properties for these, so remove them from the source nvlist after 3140 * processing. 3141 */ 3142 static int 3143 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver, 3144 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops, 3145 nvlist_t *zplprops, boolean_t *is_ci) 3146 { 3147 uint64_t sense = ZFS_PROP_UNDEFINED; 3148 uint64_t norm = ZFS_PROP_UNDEFINED; 3149 uint64_t u8 = ZFS_PROP_UNDEFINED; 3150 3151 ASSERT(zplprops != NULL); 3152 3153 if (os != NULL && os->os_phys->os_type != DMU_OST_ZFS) 3154 return (SET_ERROR(EINVAL)); 3155 3156 /* 3157 * Pull out creator prop choices, if any. 3158 */ 3159 if (createprops) { 3160 (void) nvlist_lookup_uint64(createprops, 3161 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver); 3162 (void) nvlist_lookup_uint64(createprops, 3163 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm); 3164 (void) nvlist_remove_all(createprops, 3165 zfs_prop_to_name(ZFS_PROP_NORMALIZE)); 3166 (void) nvlist_lookup_uint64(createprops, 3167 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8); 3168 (void) nvlist_remove_all(createprops, 3169 zfs_prop_to_name(ZFS_PROP_UTF8ONLY)); 3170 (void) nvlist_lookup_uint64(createprops, 3171 zfs_prop_to_name(ZFS_PROP_CASE), &sense); 3172 (void) nvlist_remove_all(createprops, 3173 zfs_prop_to_name(ZFS_PROP_CASE)); 3174 } 3175 3176 /* 3177 * If the zpl version requested is whacky or the file system 3178 * or pool is version is too "young" to support normalization 3179 * and the creator tried to set a value for one of the props, 3180 * error out. 3181 */ 3182 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) || 3183 (zplver >= ZPL_VERSION_FUID && !fuids_ok) || 3184 (zplver >= ZPL_VERSION_SA && !sa_ok) || 3185 (zplver < ZPL_VERSION_NORMALIZATION && 3186 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED || 3187 sense != ZFS_PROP_UNDEFINED))) 3188 return (SET_ERROR(ENOTSUP)); 3189 3190 /* 3191 * Put the version in the zplprops 3192 */ 3193 VERIFY(nvlist_add_uint64(zplprops, 3194 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0); 3195 3196 if (norm == ZFS_PROP_UNDEFINED) 3197 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0); 3198 VERIFY(nvlist_add_uint64(zplprops, 3199 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0); 3200 3201 /* 3202 * If we're normalizing, names must always be valid UTF-8 strings. 3203 */ 3204 if (norm) 3205 u8 = 1; 3206 if (u8 == ZFS_PROP_UNDEFINED) 3207 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0); 3208 VERIFY(nvlist_add_uint64(zplprops, 3209 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0); 3210 3211 if (sense == ZFS_PROP_UNDEFINED) 3212 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0); 3213 VERIFY(nvlist_add_uint64(zplprops, 3214 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0); 3215 3216 if (is_ci) 3217 *is_ci = (sense == ZFS_CASE_INSENSITIVE); 3218 3219 return (0); 3220 } 3221 3222 static int 3223 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops, 3224 nvlist_t *zplprops, boolean_t *is_ci) 3225 { 3226 boolean_t fuids_ok, sa_ok; 3227 uint64_t zplver = ZPL_VERSION; 3228 objset_t *os = NULL; 3229 char parentname[ZFS_MAX_DATASET_NAME_LEN]; 3230 char *cp; 3231 spa_t *spa; 3232 uint64_t spa_vers; 3233 int error; 3234 3235 (void) strlcpy(parentname, dataset, sizeof (parentname)); 3236 cp = strrchr(parentname, '/'); 3237 ASSERT(cp != NULL); 3238 cp[0] = '\0'; 3239 3240 if ((error = spa_open(dataset, &spa, FTAG)) != 0) 3241 return (error); 3242 3243 spa_vers = spa_version(spa); 3244 spa_close(spa, FTAG); 3245 3246 zplver = zfs_zpl_version_map(spa_vers); 3247 fuids_ok = (zplver >= ZPL_VERSION_FUID); 3248 sa_ok = (zplver >= ZPL_VERSION_SA); 3249 3250 /* 3251 * Open parent object set so we can inherit zplprop values. 3252 */ 3253 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0) 3254 return (error); 3255 3256 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops, 3257 zplprops, is_ci); 3258 dmu_objset_rele(os, FTAG); 3259 return (error); 3260 } 3261 3262 static int 3263 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops, 3264 nvlist_t *zplprops, boolean_t *is_ci) 3265 { 3266 boolean_t fuids_ok; 3267 boolean_t sa_ok; 3268 uint64_t zplver = ZPL_VERSION; 3269 int error; 3270 3271 zplver = zfs_zpl_version_map(spa_vers); 3272 fuids_ok = (zplver >= ZPL_VERSION_FUID); 3273 sa_ok = (zplver >= ZPL_VERSION_SA); 3274 3275 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok, 3276 createprops, zplprops, is_ci); 3277 return (error); 3278 } 3279 3280 /* 3281 * innvl: { 3282 * "type" -> dmu_objset_type_t (int32) 3283 * (optional) "props" -> { prop -> value } 3284 * (optional) "hidden_args" -> { "wkeydata" -> value } 3285 * raw uint8_t array of encryption wrapping key data (32 bytes) 3286 * } 3287 * 3288 * outnvl: propname -> error code (int32) 3289 */ 3290 3291 static const zfs_ioc_key_t zfs_keys_create[] = { 3292 {"type", DATA_TYPE_INT32, 0}, 3293 {"props", DATA_TYPE_NVLIST, ZK_OPTIONAL}, 3294 {"hidden_args", DATA_TYPE_NVLIST, ZK_OPTIONAL}, 3295 }; 3296 3297 static int 3298 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3299 { 3300 int error = 0; 3301 zfs_creat_t zct = { 0 }; 3302 nvlist_t *nvprops = NULL; 3303 nvlist_t *hidden_args = NULL; 3304 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx); 3305 dmu_objset_type_t type; 3306 boolean_t is_insensitive = B_FALSE; 3307 dsl_crypto_params_t *dcp = NULL; 3308 3309 type = (dmu_objset_type_t)fnvlist_lookup_int32(innvl, "type"); 3310 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops); 3311 (void) nvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS, &hidden_args); 3312 3313 switch (type) { 3314 case DMU_OST_ZFS: 3315 cbfunc = zfs_create_cb; 3316 break; 3317 3318 case DMU_OST_ZVOL: 3319 cbfunc = zvol_create_cb; 3320 break; 3321 3322 default: 3323 cbfunc = NULL; 3324 break; 3325 } 3326 if (strchr(fsname, '@') || 3327 strchr(fsname, '%')) 3328 return (SET_ERROR(EINVAL)); 3329 3330 zct.zct_props = nvprops; 3331 3332 if (cbfunc == NULL) 3333 return (SET_ERROR(EINVAL)); 3334 3335 if (type == DMU_OST_ZVOL) { 3336 uint64_t volsize, volblocksize; 3337 3338 if (nvprops == NULL) 3339 return (SET_ERROR(EINVAL)); 3340 if (nvlist_lookup_uint64(nvprops, 3341 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0) 3342 return (SET_ERROR(EINVAL)); 3343 3344 if ((error = nvlist_lookup_uint64(nvprops, 3345 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), 3346 &volblocksize)) != 0 && error != ENOENT) 3347 return (SET_ERROR(EINVAL)); 3348 3349 if (error != 0) 3350 volblocksize = zfs_prop_default_numeric( 3351 ZFS_PROP_VOLBLOCKSIZE); 3352 3353 if ((error = zvol_check_volblocksize( 3354 volblocksize)) != 0 || 3355 (error = zvol_check_volsize(volsize, 3356 volblocksize)) != 0) 3357 return (error); 3358 } else if (type == DMU_OST_ZFS) { 3359 int error; 3360 3361 /* 3362 * We have to have normalization and 3363 * case-folding flags correct when we do the 3364 * file system creation, so go figure them out 3365 * now. 3366 */ 3367 VERIFY(nvlist_alloc(&zct.zct_zplprops, 3368 NV_UNIQUE_NAME, KM_SLEEP) == 0); 3369 error = zfs_fill_zplprops(fsname, nvprops, 3370 zct.zct_zplprops, &is_insensitive); 3371 if (error != 0) { 3372 nvlist_free(zct.zct_zplprops); 3373 return (error); 3374 } 3375 } 3376 3377 error = dsl_crypto_params_create_nvlist(DCP_CMD_NONE, nvprops, 3378 hidden_args, &dcp); 3379 if (error != 0) { 3380 nvlist_free(zct.zct_zplprops); 3381 return (error); 3382 } 3383 3384 error = dmu_objset_create(fsname, type, 3385 is_insensitive ? DS_FLAG_CI_DATASET : 0, dcp, cbfunc, &zct); 3386 3387 nvlist_free(zct.zct_zplprops); 3388 dsl_crypto_params_free(dcp, !!error); 3389 3390 /* 3391 * It would be nice to do this atomically. 3392 */ 3393 if (error == 0) { 3394 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL, 3395 nvprops, outnvl); 3396 if (error != 0) 3397 (void) dsl_destroy_head(fsname); 3398 } 3399 return (error); 3400 } 3401 3402 /* 3403 * innvl: { 3404 * "origin" -> name of origin snapshot 3405 * (optional) "props" -> { prop -> value } 3406 * (optional) "hidden_args" -> { "wkeydata" -> value } 3407 * raw uint8_t array of encryption wrapping key data (32 bytes) 3408 * } 3409 * 3410 * outnvl: propname -> error code (int32) 3411 */ 3412 static const zfs_ioc_key_t zfs_keys_clone[] = { 3413 {"origin", DATA_TYPE_STRING, 0}, 3414 {"props", DATA_TYPE_NVLIST, ZK_OPTIONAL}, 3415 {"hidden_args", DATA_TYPE_NVLIST, ZK_OPTIONAL}, 3416 }; 3417 3418 static int 3419 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3420 { 3421 int error = 0; 3422 nvlist_t *nvprops = NULL; 3423 char *origin_name; 3424 3425 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0) 3426 return (SET_ERROR(EINVAL)); 3427 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops); 3428 3429 if (strchr(fsname, '@') || 3430 strchr(fsname, '%')) 3431 return (SET_ERROR(EINVAL)); 3432 3433 if (dataset_namecheck(origin_name, NULL, NULL) != 0) 3434 return (SET_ERROR(EINVAL)); 3435 3436 error = dmu_objset_clone(fsname, origin_name); 3437 3438 /* 3439 * It would be nice to do this atomically. 3440 */ 3441 if (error == 0) { 3442 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL, 3443 nvprops, outnvl); 3444 if (error != 0) 3445 (void) dsl_destroy_head(fsname); 3446 } 3447 return (error); 3448 } 3449 3450 static const zfs_ioc_key_t zfs_keys_remap[] = { 3451 /* no nvl keys */ 3452 }; 3453 3454 /* ARGSUSED */ 3455 static int 3456 zfs_ioc_remap(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3457 { 3458 if (strchr(fsname, '@') || 3459 strchr(fsname, '%')) 3460 return (SET_ERROR(EINVAL)); 3461 3462 return (dmu_objset_remap_indirects(fsname)); 3463 } 3464 3465 /* 3466 * innvl: { 3467 * "snaps" -> { snapshot1, snapshot2 } 3468 * (optional) "props" -> { prop -> value (string) } 3469 * } 3470 * 3471 * outnvl: snapshot -> error code (int32) 3472 */ 3473 static const zfs_ioc_key_t zfs_keys_snapshot[] = { 3474 {"snaps", DATA_TYPE_NVLIST, 0}, 3475 {"props", DATA_TYPE_NVLIST, ZK_OPTIONAL}, 3476 }; 3477 3478 static int 3479 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3480 { 3481 nvlist_t *snaps; 3482 nvlist_t *props = NULL; 3483 int error, poollen; 3484 nvpair_t *pair; 3485 3486 (void) nvlist_lookup_nvlist(innvl, "props", &props); 3487 if ((error = zfs_check_userprops(poolname, props)) != 0) 3488 return (error); 3489 3490 if (!nvlist_empty(props) && 3491 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS)) 3492 return (SET_ERROR(ENOTSUP)); 3493 3494 snaps = fnvlist_lookup_nvlist(innvl, "snaps"); 3495 poollen = strlen(poolname); 3496 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 3497 pair = nvlist_next_nvpair(snaps, pair)) { 3498 const char *name = nvpair_name(pair); 3499 const char *cp = strchr(name, '@'); 3500 3501 /* 3502 * The snap name must contain an @, and the part after it must 3503 * contain only valid characters. 3504 */ 3505 if (cp == NULL || 3506 zfs_component_namecheck(cp + 1, NULL, NULL) != 0) 3507 return (SET_ERROR(EINVAL)); 3508 3509 /* 3510 * The snap must be in the specified pool. 3511 */ 3512 if (strncmp(name, poolname, poollen) != 0 || 3513 (name[poollen] != '/' && name[poollen] != '@')) 3514 return (SET_ERROR(EXDEV)); 3515 3516 /* This must be the only snap of this fs. */ 3517 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair); 3518 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) { 3519 if (strncmp(name, nvpair_name(pair2), cp - name + 1) 3520 == 0) { 3521 return (SET_ERROR(EXDEV)); 3522 } 3523 } 3524 } 3525 3526 error = dsl_dataset_snapshot(snaps, props, outnvl); 3527 return (error); 3528 } 3529 3530 /* 3531 * innvl: "message" -> string 3532 */ 3533 static const zfs_ioc_key_t zfs_keys_log_history[] = { 3534 {"message", DATA_TYPE_STRING, 0}, 3535 }; 3536 3537 /* ARGSUSED */ 3538 static int 3539 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl) 3540 { 3541 char *message; 3542 spa_t *spa; 3543 int error; 3544 char *poolname; 3545 3546 /* 3547 * The poolname in the ioctl is not set, we get it from the TSD, 3548 * which was set at the end of the last successful ioctl that allows 3549 * logging. The secpolicy func already checked that it is set. 3550 * Only one log ioctl is allowed after each successful ioctl, so 3551 * we clear the TSD here. 3552 */ 3553 poolname = tsd_get(zfs_allow_log_key); 3554 (void) tsd_set(zfs_allow_log_key, NULL); 3555 error = spa_open(poolname, &spa, FTAG); 3556 strfree(poolname); 3557 if (error != 0) 3558 return (error); 3559 3560 message = fnvlist_lookup_string(innvl, "message"); 3561 3562 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) { 3563 spa_close(spa, FTAG); 3564 return (SET_ERROR(ENOTSUP)); 3565 } 3566 3567 error = spa_history_log(spa, message); 3568 spa_close(spa, FTAG); 3569 return (error); 3570 } 3571 3572 /* 3573 * This ioctl is used to set the bootenv configuration on the current 3574 * pool. This configuration is stored in the second padding area of the label, 3575 * and it is used by the bootloader(s) to store bootloader and/or system 3576 * specific data. 3577 * The data is stored as nvlist data stream, and is protected by 3578 * an embedded checksum. 3579 * The version can have two possible values: 3580 * VB_RAW: nvlist should have key GRUB_ENVMAP, value DATA_TYPE_STRING. 3581 * VB_NVLIST: nvlist with arbitrary <key, value> pairs. 3582 */ 3583 static const zfs_ioc_key_t zfs_keys_set_bootenv[] = { 3584 {"version", DATA_TYPE_UINT64, 0}, 3585 {"<keys>", DATA_TYPE_ANY, ZK_OPTIONAL | ZK_WILDCARDLIST}, 3586 }; 3587 3588 static int 3589 zfs_ioc_set_bootenv(const char *name, nvlist_t *innvl, 3590 nvlist_t *outnvl __unused) 3591 { 3592 int error; 3593 spa_t *spa; 3594 3595 if ((error = spa_open(name, &spa, FTAG)) != 0) 3596 return (error); 3597 spa_vdev_state_enter(spa, SCL_ALL); 3598 error = vdev_label_write_bootenv(spa->spa_root_vdev, innvl); 3599 (void) spa_vdev_state_exit(spa, NULL, 0); 3600 spa_close(spa, FTAG); 3601 return (error); 3602 } 3603 3604 static const zfs_ioc_key_t zfs_keys_get_bootenv[] = { 3605 /* no nvl keys */ 3606 }; 3607 3608 static int 3609 zfs_ioc_get_bootenv(const char *name, nvlist_t *innvl __unused, 3610 nvlist_t *outnvl) 3611 { 3612 spa_t *spa; 3613 int error; 3614 3615 if ((error = spa_open(name, &spa, FTAG)) != 0) 3616 return (error); 3617 spa_vdev_state_enter(spa, SCL_ALL); 3618 error = vdev_label_read_bootenv(spa->spa_root_vdev, outnvl); 3619 (void) spa_vdev_state_exit(spa, NULL, 0); 3620 spa_close(spa, FTAG); 3621 return (error); 3622 } 3623 3624 /* 3625 * The dp_config_rwlock must not be held when calling this, because the 3626 * unmount may need to write out data. 3627 * 3628 * This function is best-effort. Callers must deal gracefully if it 3629 * remains mounted (or is remounted after this call). 3630 * 3631 * Returns 0 if the argument is not a snapshot, or it is not currently a 3632 * filesystem, or we were able to unmount it. Returns error code otherwise. 3633 */ 3634 void 3635 zfs_unmount_snap(const char *snapname) 3636 { 3637 vfs_t *vfsp = NULL; 3638 zfsvfs_t *zfsvfs = NULL; 3639 3640 if (strchr(snapname, '@') == NULL) 3641 return; 3642 3643 int err = getzfsvfs(snapname, &zfsvfs); 3644 if (err != 0) { 3645 ASSERT3P(zfsvfs, ==, NULL); 3646 return; 3647 } 3648 vfsp = zfsvfs->z_vfs; 3649 3650 ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os))); 3651 3652 err = vn_vfswlock(vfsp->vfs_vnodecovered); 3653 VFS_RELE(vfsp); 3654 if (err != 0) 3655 return; 3656 3657 /* 3658 * Always force the unmount for snapshots. 3659 */ 3660 (void) dounmount(vfsp, MS_FORCE, kcred); 3661 } 3662 3663 /* ARGSUSED */ 3664 static int 3665 zfs_unmount_snap_cb(const char *snapname, void *arg) 3666 { 3667 zfs_unmount_snap(snapname); 3668 return (0); 3669 } 3670 3671 /* 3672 * When a clone is destroyed, its origin may also need to be destroyed, 3673 * in which case it must be unmounted. This routine will do that unmount 3674 * if necessary. 3675 */ 3676 void 3677 zfs_destroy_unmount_origin(const char *fsname) 3678 { 3679 int error; 3680 objset_t *os; 3681 dsl_dataset_t *ds; 3682 3683 error = dmu_objset_hold(fsname, FTAG, &os); 3684 if (error != 0) 3685 return; 3686 ds = dmu_objset_ds(os); 3687 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) { 3688 char originname[ZFS_MAX_DATASET_NAME_LEN]; 3689 dsl_dataset_name(ds->ds_prev, originname); 3690 dmu_objset_rele(os, FTAG); 3691 zfs_unmount_snap(originname); 3692 } else { 3693 dmu_objset_rele(os, FTAG); 3694 } 3695 } 3696 3697 /* 3698 * innvl: { 3699 * "snaps" -> { snapshot1, snapshot2 } 3700 * (optional boolean) "defer" 3701 * } 3702 * 3703 * outnvl: snapshot -> error code (int32) 3704 * 3705 */ 3706 static const zfs_ioc_key_t zfs_keys_destroy_snaps[] = { 3707 {"snaps", DATA_TYPE_NVLIST, 0}, 3708 {"defer", DATA_TYPE_BOOLEAN, ZK_OPTIONAL}, 3709 }; 3710 3711 /* ARGSUSED */ 3712 static int 3713 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3714 { 3715 nvlist_t *snaps; 3716 nvpair_t *pair; 3717 boolean_t defer; 3718 3719 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 3720 return (SET_ERROR(EINVAL)); 3721 defer = nvlist_exists(innvl, "defer"); 3722 3723 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 3724 pair = nvlist_next_nvpair(snaps, pair)) { 3725 zfs_unmount_snap(nvpair_name(pair)); 3726 } 3727 3728 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl)); 3729 } 3730 3731 /* 3732 * Create bookmarks. Bookmark names are of the form <fs>#<bmark>. 3733 * All bookmarks must be in the same pool. 3734 * 3735 * innvl: { 3736 * bookmark1 -> snapshot1, bookmark2 -> snapshot2 3737 * } 3738 * 3739 * outnvl: bookmark -> error code (int32) 3740 * 3741 */ 3742 static const zfs_ioc_key_t zfs_keys_bookmark[] = { 3743 {"<bookmark>...", DATA_TYPE_STRING, ZK_WILDCARDLIST}, 3744 }; 3745 3746 /* ARGSUSED */ 3747 static int 3748 zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3749 { 3750 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 3751 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 3752 char *snap_name; 3753 3754 /* 3755 * Verify the snapshot argument. 3756 */ 3757 if (nvpair_value_string(pair, &snap_name) != 0) 3758 return (SET_ERROR(EINVAL)); 3759 3760 3761 /* Verify that the keys (bookmarks) are unique */ 3762 for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair); 3763 pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) { 3764 if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0) 3765 return (SET_ERROR(EINVAL)); 3766 } 3767 } 3768 3769 return (dsl_bookmark_create(innvl, outnvl)); 3770 } 3771 3772 /* 3773 * innvl: { 3774 * property 1, property 2, ... 3775 * } 3776 * 3777 * outnvl: { 3778 * bookmark name 1 -> { property 1, property 2, ... }, 3779 * bookmark name 2 -> { property 1, property 2, ... } 3780 * } 3781 * 3782 */ 3783 static const zfs_ioc_key_t zfs_keys_get_bookmarks[] = { 3784 {"<property>...", DATA_TYPE_BOOLEAN, ZK_WILDCARDLIST | ZK_OPTIONAL}, 3785 }; 3786 3787 static int 3788 zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3789 { 3790 return (dsl_get_bookmarks(fsname, innvl, outnvl)); 3791 } 3792 3793 /* 3794 * innvl: { 3795 * bookmark name 1, bookmark name 2 3796 * } 3797 * 3798 * outnvl: bookmark -> error code (int32) 3799 * 3800 */ 3801 static const zfs_ioc_key_t zfs_keys_destroy_bookmarks[] = { 3802 {"<bookmark>...", DATA_TYPE_BOOLEAN, ZK_WILDCARDLIST}, 3803 }; 3804 3805 static int 3806 zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl, 3807 nvlist_t *outnvl) 3808 { 3809 int error, poollen; 3810 3811 poollen = strlen(poolname); 3812 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 3813 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 3814 const char *name = nvpair_name(pair); 3815 const char *cp = strchr(name, '#'); 3816 3817 /* 3818 * The bookmark name must contain an #, and the part after it 3819 * must contain only valid characters. 3820 */ 3821 if (cp == NULL || 3822 zfs_component_namecheck(cp + 1, NULL, NULL) != 0) 3823 return (SET_ERROR(EINVAL)); 3824 3825 /* 3826 * The bookmark must be in the specified pool. 3827 */ 3828 if (strncmp(name, poolname, poollen) != 0 || 3829 (name[poollen] != '/' && name[poollen] != '#')) 3830 return (SET_ERROR(EXDEV)); 3831 } 3832 3833 error = dsl_bookmark_destroy(innvl, outnvl); 3834 return (error); 3835 } 3836 3837 static const zfs_ioc_key_t zfs_keys_channel_program[] = { 3838 {"program", DATA_TYPE_STRING, 0}, 3839 {"arg", DATA_TYPE_ANY, 0}, 3840 {"sync", DATA_TYPE_BOOLEAN_VALUE, ZK_OPTIONAL}, 3841 {"instrlimit", DATA_TYPE_UINT64, ZK_OPTIONAL}, 3842 {"memlimit", DATA_TYPE_UINT64, ZK_OPTIONAL}, 3843 }; 3844 3845 static int 3846 zfs_ioc_channel_program(const char *poolname, nvlist_t *innvl, 3847 nvlist_t *outnvl) 3848 { 3849 char *program; 3850 uint64_t instrlimit, memlimit; 3851 boolean_t sync_flag; 3852 nvpair_t *nvarg = NULL; 3853 3854 program = fnvlist_lookup_string(innvl, ZCP_ARG_PROGRAM); 3855 if (0 != nvlist_lookup_boolean_value(innvl, ZCP_ARG_SYNC, &sync_flag)) { 3856 sync_flag = B_TRUE; 3857 } 3858 if (0 != nvlist_lookup_uint64(innvl, ZCP_ARG_INSTRLIMIT, &instrlimit)) { 3859 instrlimit = ZCP_DEFAULT_INSTRLIMIT; 3860 } 3861 if (0 != nvlist_lookup_uint64(innvl, ZCP_ARG_MEMLIMIT, &memlimit)) { 3862 memlimit = ZCP_DEFAULT_MEMLIMIT; 3863 } 3864 nvarg = fnvlist_lookup_nvpair(innvl, ZCP_ARG_ARGLIST); 3865 3866 if (instrlimit == 0 || instrlimit > zfs_lua_max_instrlimit) 3867 return (EINVAL); 3868 if (memlimit == 0 || memlimit > zfs_lua_max_memlimit) 3869 return (EINVAL); 3870 3871 return (zcp_eval(poolname, program, sync_flag, instrlimit, memlimit, 3872 nvarg, outnvl)); 3873 } 3874 3875 /* 3876 * innvl: unused 3877 * outnvl: empty 3878 */ 3879 static const zfs_ioc_key_t zfs_keys_pool_checkpoint[] = { 3880 /* no nvl keys */ 3881 }; 3882 3883 /* ARGSUSED */ 3884 static int 3885 zfs_ioc_pool_checkpoint(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3886 { 3887 return (spa_checkpoint(poolname)); 3888 } 3889 3890 /* 3891 * innvl: unused 3892 * outnvl: empty 3893 */ 3894 static const zfs_ioc_key_t zfs_keys_pool_discard_checkpoint[] = { 3895 /* no nvl keys */ 3896 }; 3897 3898 /* ARGSUSED */ 3899 static int 3900 zfs_ioc_pool_discard_checkpoint(const char *poolname, nvlist_t *innvl, 3901 nvlist_t *outnvl) 3902 { 3903 return (spa_checkpoint_discard(poolname)); 3904 } 3905 3906 /* 3907 * inputs: 3908 * zc_name name of dataset to destroy 3909 * zc_defer_destroy mark for deferred destroy 3910 * 3911 * outputs: none 3912 */ 3913 static int 3914 zfs_ioc_destroy(zfs_cmd_t *zc) 3915 { 3916 objset_t *os; 3917 dmu_objset_type_t ost; 3918 int err; 3919 3920 err = dmu_objset_hold(zc->zc_name, FTAG, &os); 3921 if (err != 0) 3922 return (err); 3923 ost = dmu_objset_type(os); 3924 dmu_objset_rele(os, FTAG); 3925 3926 if (ost == DMU_OST_ZFS) 3927 zfs_unmount_snap(zc->zc_name); 3928 3929 if (strchr(zc->zc_name, '@')) { 3930 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy); 3931 } else { 3932 err = dsl_destroy_head(zc->zc_name); 3933 if (err == EEXIST) { 3934 /* 3935 * It is possible that the given DS may have 3936 * hidden child (%recv) datasets - "leftovers" 3937 * resulting from the previously interrupted 3938 * 'zfs receive'. 3939 * 3940 * 6 extra bytes for /%recv 3941 */ 3942 char namebuf[ZFS_MAX_DATASET_NAME_LEN + 6]; 3943 3944 if (snprintf(namebuf, sizeof (namebuf), "%s/%s", 3945 zc->zc_name, recv_clone_name) >= 3946 sizeof (namebuf)) 3947 return (SET_ERROR(EINVAL)); 3948 3949 /* 3950 * Try to remove the hidden child (%recv) and after 3951 * that try to remove the target dataset. 3952 * If the hidden child (%recv) does not exist 3953 * the original error (EEXIST) will be returned 3954 */ 3955 err = dsl_destroy_head(namebuf); 3956 if (err == 0) 3957 err = dsl_destroy_head(zc->zc_name); 3958 else if (err == ENOENT) 3959 err = SET_ERROR(EEXIST); 3960 } 3961 } 3962 if (ost == DMU_OST_ZVOL && err == 0) 3963 (void) zvol_remove_minor(zc->zc_name); 3964 return (err); 3965 } 3966 3967 /* 3968 * innvl: { 3969 * "initialize_command" -> POOL_INITIALIZE_{CANCEL|START|SUSPEND} (uint64) 3970 * "initialize_vdevs": { -> guids to initialize (nvlist) 3971 * "vdev_path_1": vdev_guid_1, (uint64), 3972 * "vdev_path_2": vdev_guid_2, (uint64), 3973 * ... 3974 * }, 3975 * } 3976 * 3977 * outnvl: { 3978 * "initialize_vdevs": { -> initialization errors (nvlist) 3979 * "vdev_path_1": errno, see function body for possible errnos (uint64) 3980 * "vdev_path_2": errno, ... (uint64) 3981 * ... 3982 * } 3983 * } 3984 * 3985 * EINVAL is returned for an unknown command or if any of the provided vdev 3986 * guids have be specified with a type other than uint64. 3987 */ 3988 static const zfs_ioc_key_t zfs_keys_pool_initialize[] = { 3989 {ZPOOL_INITIALIZE_COMMAND, DATA_TYPE_UINT64, 0}, 3990 {ZPOOL_INITIALIZE_VDEVS, DATA_TYPE_NVLIST, 0} 3991 }; 3992 3993 static int 3994 zfs_ioc_pool_initialize(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3995 { 3996 uint64_t cmd_type; 3997 if (nvlist_lookup_uint64(innvl, ZPOOL_INITIALIZE_COMMAND, 3998 &cmd_type) != 0) { 3999 return (SET_ERROR(EINVAL)); 4000 } 4001 4002 if (!(cmd_type == POOL_INITIALIZE_CANCEL || 4003 cmd_type == POOL_INITIALIZE_START || 4004 cmd_type == POOL_INITIALIZE_SUSPEND)) { 4005 return (SET_ERROR(EINVAL)); 4006 } 4007 4008 nvlist_t *vdev_guids; 4009 if (nvlist_lookup_nvlist(innvl, ZPOOL_INITIALIZE_VDEVS, 4010 &vdev_guids) != 0) { 4011 return (SET_ERROR(EINVAL)); 4012 } 4013 4014 for (nvpair_t *pair = nvlist_next_nvpair(vdev_guids, NULL); 4015 pair != NULL; pair = nvlist_next_nvpair(vdev_guids, pair)) { 4016 uint64_t vdev_guid; 4017 if (nvpair_value_uint64(pair, &vdev_guid) != 0) { 4018 return (SET_ERROR(EINVAL)); 4019 } 4020 } 4021 4022 spa_t *spa; 4023 int error = spa_open(poolname, &spa, FTAG); 4024 if (error != 0) 4025 return (error); 4026 4027 nvlist_t *vdev_errlist = fnvlist_alloc(); 4028 int total_errors = spa_vdev_initialize(spa, vdev_guids, cmd_type, 4029 vdev_errlist); 4030 4031 if (fnvlist_size(vdev_errlist) > 0) { 4032 fnvlist_add_nvlist(outnvl, ZPOOL_INITIALIZE_VDEVS, 4033 vdev_errlist); 4034 } 4035 fnvlist_free(vdev_errlist); 4036 4037 spa_close(spa, FTAG); 4038 return (total_errors > 0 ? EINVAL : 0); 4039 } 4040 4041 /* 4042 * innvl: { 4043 * "trim_command" -> POOL_TRIM_{CANCEL|START|SUSPEND} (uint64) 4044 * "trim_vdevs": { -> guids to TRIM (nvlist) 4045 * "vdev_path_1": vdev_guid_1, (uint64), 4046 * "vdev_path_2": vdev_guid_2, (uint64), 4047 * ... 4048 * }, 4049 * "trim_rate" -> Target TRIM rate in bytes/sec. 4050 * "trim_secure" -> Set to request a secure TRIM. 4051 * } 4052 * 4053 * outnvl: { 4054 * "trim_vdevs": { -> TRIM errors (nvlist) 4055 * "vdev_path_1": errno, see function body for possible errnos (uint64) 4056 * "vdev_path_2": errno, ... (uint64) 4057 * ... 4058 * } 4059 * } 4060 * 4061 * EINVAL is returned for an unknown command or if any of the provided vdev 4062 * guids have be specified with a type other than uint64. 4063 */ 4064 static const zfs_ioc_key_t zfs_keys_pool_trim[] = { 4065 {ZPOOL_TRIM_COMMAND, DATA_TYPE_UINT64, 0}, 4066 {ZPOOL_TRIM_VDEVS, DATA_TYPE_NVLIST, 0}, 4067 {ZPOOL_TRIM_RATE, DATA_TYPE_UINT64, ZK_OPTIONAL}, 4068 {ZPOOL_TRIM_SECURE, DATA_TYPE_BOOLEAN_VALUE, ZK_OPTIONAL}, 4069 }; 4070 4071 static int 4072 zfs_ioc_pool_trim(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 4073 { 4074 uint64_t cmd_type; 4075 if (nvlist_lookup_uint64(innvl, ZPOOL_TRIM_COMMAND, &cmd_type) != 0) 4076 return (SET_ERROR(EINVAL)); 4077 4078 if (!(cmd_type == POOL_TRIM_CANCEL || 4079 cmd_type == POOL_TRIM_START || 4080 cmd_type == POOL_TRIM_SUSPEND)) { 4081 return (SET_ERROR(EINVAL)); 4082 } 4083 4084 nvlist_t *vdev_guids; 4085 if (nvlist_lookup_nvlist(innvl, ZPOOL_TRIM_VDEVS, &vdev_guids) != 0) 4086 return (SET_ERROR(EINVAL)); 4087 4088 for (nvpair_t *pair = nvlist_next_nvpair(vdev_guids, NULL); 4089 pair != NULL; pair = nvlist_next_nvpair(vdev_guids, pair)) { 4090 uint64_t vdev_guid; 4091 if (nvpair_value_uint64(pair, &vdev_guid) != 0) { 4092 return (SET_ERROR(EINVAL)); 4093 } 4094 } 4095 4096 /* Optional, defaults to maximum rate when not provided */ 4097 uint64_t rate; 4098 if (nvlist_lookup_uint64(innvl, ZPOOL_TRIM_RATE, &rate) != 0) 4099 rate = 0; 4100 4101 /* Optional, defaults to standard TRIM when not provided */ 4102 boolean_t secure; 4103 if (nvlist_lookup_boolean_value(innvl, ZPOOL_TRIM_SECURE, 4104 &secure) != 0) { 4105 secure = B_FALSE; 4106 } 4107 4108 spa_t *spa; 4109 int error = spa_open(poolname, &spa, FTAG); 4110 if (error != 0) 4111 return (error); 4112 4113 nvlist_t *vdev_errlist = fnvlist_alloc(); 4114 int total_errors = spa_vdev_trim(spa, vdev_guids, cmd_type, 4115 rate, !!zfs_trim_metaslab_skip, secure, vdev_errlist); 4116 4117 if (fnvlist_size(vdev_errlist) > 0) 4118 fnvlist_add_nvlist(outnvl, ZPOOL_TRIM_VDEVS, vdev_errlist); 4119 4120 fnvlist_free(vdev_errlist); 4121 4122 spa_close(spa, FTAG); 4123 return (total_errors > 0 ? EINVAL : 0); 4124 } 4125 4126 /* 4127 * fsname is name of dataset to rollback (to most recent snapshot) 4128 * 4129 * innvl may contain name of expected target snapshot 4130 * 4131 * outnvl: "target" -> name of most recent snapshot 4132 * } 4133 */ 4134 static const zfs_ioc_key_t zfs_keys_rollback[] = { 4135 {"target", DATA_TYPE_STRING, ZK_OPTIONAL}, 4136 }; 4137 4138 /* ARGSUSED */ 4139 static int 4140 zfs_ioc_rollback(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 4141 { 4142 zfsvfs_t *zfsvfs; 4143 char *target = NULL; 4144 int error; 4145 4146 (void) nvlist_lookup_string(innvl, "target", &target); 4147 if (target != NULL) { 4148 const char *cp = strchr(target, '@'); 4149 4150 /* 4151 * The snap name must contain an @, and the part after it must 4152 * contain only valid characters. 4153 */ 4154 if (cp == NULL || 4155 zfs_component_namecheck(cp + 1, NULL, NULL) != 0) 4156 return (SET_ERROR(EINVAL)); 4157 } 4158 4159 if (getzfsvfs(fsname, &zfsvfs) == 0) { 4160 dsl_dataset_t *ds; 4161 4162 ds = dmu_objset_ds(zfsvfs->z_os); 4163 error = zfs_suspend_fs(zfsvfs); 4164 if (error == 0) { 4165 int resume_err; 4166 4167 error = dsl_dataset_rollback(fsname, target, zfsvfs, 4168 outnvl); 4169 resume_err = zfs_resume_fs(zfsvfs, ds); 4170 error = error ? error : resume_err; 4171 } 4172 VFS_RELE(zfsvfs->z_vfs); 4173 } else { 4174 error = dsl_dataset_rollback(fsname, target, NULL, outnvl); 4175 } 4176 return (error); 4177 } 4178 4179 static int 4180 recursive_unmount(const char *fsname, void *arg) 4181 { 4182 const char *snapname = arg; 4183 char fullname[ZFS_MAX_DATASET_NAME_LEN]; 4184 4185 (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname); 4186 zfs_unmount_snap(fullname); 4187 4188 return (0); 4189 } 4190 4191 /* 4192 * inputs: 4193 * zc_name old name of dataset 4194 * zc_value new name of dataset 4195 * zc_cookie recursive flag (only valid for snapshots) 4196 * 4197 * outputs: none 4198 */ 4199 static int 4200 zfs_ioc_rename(zfs_cmd_t *zc) 4201 { 4202 objset_t *os; 4203 dmu_objset_type_t ost; 4204 boolean_t recursive = zc->zc_cookie & 1; 4205 char *at; 4206 int err; 4207 4208 /* "zfs rename" from and to ...%recv datasets should both fail */ 4209 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0'; 4210 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0'; 4211 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0 || 4212 dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 4213 strchr(zc->zc_name, '%') || strchr(zc->zc_value, '%')) 4214 return (SET_ERROR(EINVAL)); 4215 4216 err = dmu_objset_hold(zc->zc_name, FTAG, &os); 4217 if (err != 0) 4218 return (err); 4219 ost = dmu_objset_type(os); 4220 dmu_objset_rele(os, FTAG); 4221 4222 at = strchr(zc->zc_name, '@'); 4223 if (at != NULL) { 4224 /* snaps must be in same fs */ 4225 int error; 4226 4227 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1)) 4228 return (SET_ERROR(EXDEV)); 4229 *at = '\0'; 4230 if (ost == DMU_OST_ZFS) { 4231 error = dmu_objset_find(zc->zc_name, 4232 recursive_unmount, at + 1, 4233 recursive ? DS_FIND_CHILDREN : 0); 4234 if (error != 0) { 4235 *at = '@'; 4236 return (error); 4237 } 4238 } 4239 error = dsl_dataset_rename_snapshot(zc->zc_name, 4240 at + 1, strchr(zc->zc_value, '@') + 1, recursive); 4241 *at = '@'; 4242 4243 return (error); 4244 } else { 4245 if (ost == DMU_OST_ZVOL) 4246 (void) zvol_remove_minor(zc->zc_name); 4247 return (dsl_dir_rename(zc->zc_name, zc->zc_value)); 4248 } 4249 } 4250 4251 static int 4252 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr) 4253 { 4254 const char *propname = nvpair_name(pair); 4255 boolean_t issnap = (strchr(dsname, '@') != NULL); 4256 zfs_prop_t prop = zfs_name_to_prop(propname); 4257 uint64_t intval; 4258 int err; 4259 4260 if (prop == ZPROP_INVAL) { 4261 if (zfs_prop_user(propname)) { 4262 if (err = zfs_secpolicy_write_perms(dsname, 4263 ZFS_DELEG_PERM_USERPROP, cr)) 4264 return (err); 4265 return (0); 4266 } 4267 4268 if (!issnap && zfs_prop_userquota(propname)) { 4269 const char *perm = NULL; 4270 const char *uq_prefix = 4271 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA]; 4272 const char *gq_prefix = 4273 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA]; 4274 const char *uiq_prefix = 4275 zfs_userquota_prop_prefixes[ZFS_PROP_USEROBJQUOTA]; 4276 const char *giq_prefix = 4277 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPOBJQUOTA]; 4278 const char *pq_prefix = 4279 zfs_userquota_prop_prefixes[ZFS_PROP_PROJECTQUOTA]; 4280 const char *piq_prefix = zfs_userquota_prop_prefixes[\ 4281 ZFS_PROP_PROJECTOBJQUOTA]; 4282 4283 if (strncmp(propname, uq_prefix, 4284 strlen(uq_prefix)) == 0) { 4285 perm = ZFS_DELEG_PERM_USERQUOTA; 4286 } else if (strncmp(propname, uiq_prefix, 4287 strlen(uiq_prefix)) == 0) { 4288 perm = ZFS_DELEG_PERM_USEROBJQUOTA; 4289 } else if (strncmp(propname, gq_prefix, 4290 strlen(gq_prefix)) == 0) { 4291 perm = ZFS_DELEG_PERM_GROUPQUOTA; 4292 } else if (strncmp(propname, giq_prefix, 4293 strlen(giq_prefix)) == 0) { 4294 perm = ZFS_DELEG_PERM_GROUPOBJQUOTA; 4295 } else if (strncmp(propname, pq_prefix, 4296 strlen(pq_prefix)) == 0) { 4297 perm = ZFS_DELEG_PERM_PROJECTQUOTA; 4298 } else if (strncmp(propname, piq_prefix, 4299 strlen(piq_prefix)) == 0) { 4300 perm = ZFS_DELEG_PERM_PROJECTOBJQUOTA; 4301 } else { 4302 /* {USER|GROUP|PROJECT}USED are read-only */ 4303 return (SET_ERROR(EINVAL)); 4304 } 4305 4306 if (err = zfs_secpolicy_write_perms(dsname, perm, cr)) 4307 return (err); 4308 return (0); 4309 } 4310 4311 return (SET_ERROR(EINVAL)); 4312 } 4313 4314 if (issnap) 4315 return (SET_ERROR(EINVAL)); 4316 4317 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 4318 /* 4319 * dsl_prop_get_all_impl() returns properties in this 4320 * format. 4321 */ 4322 nvlist_t *attrs; 4323 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 4324 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 4325 &pair) == 0); 4326 } 4327 4328 /* 4329 * Check that this value is valid for this pool version 4330 */ 4331 switch (prop) { 4332 case ZFS_PROP_COMPRESSION: 4333 /* 4334 * If the user specified gzip compression, make sure 4335 * the SPA supports it. We ignore any errors here since 4336 * we'll catch them later. 4337 */ 4338 if (nvpair_value_uint64(pair, &intval) == 0) { 4339 if (intval >= ZIO_COMPRESS_GZIP_1 && 4340 intval <= ZIO_COMPRESS_GZIP_9 && 4341 zfs_earlier_version(dsname, 4342 SPA_VERSION_GZIP_COMPRESSION)) { 4343 return (SET_ERROR(ENOTSUP)); 4344 } 4345 4346 if (intval == ZIO_COMPRESS_ZLE && 4347 zfs_earlier_version(dsname, 4348 SPA_VERSION_ZLE_COMPRESSION)) 4349 return (SET_ERROR(ENOTSUP)); 4350 4351 if (intval == ZIO_COMPRESS_LZ4) { 4352 spa_t *spa; 4353 4354 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 4355 return (err); 4356 4357 if (!spa_feature_is_enabled(spa, 4358 SPA_FEATURE_LZ4_COMPRESS)) { 4359 spa_close(spa, FTAG); 4360 return (SET_ERROR(ENOTSUP)); 4361 } 4362 spa_close(spa, FTAG); 4363 } 4364 4365 /* 4366 * If this is a bootable dataset then 4367 * verify that the compression algorithm 4368 * is supported for booting. We must return 4369 * something other than ENOTSUP since it 4370 * implies a downrev pool version. 4371 */ 4372 if (zfs_is_bootfs(dsname) && 4373 !BOOTFS_COMPRESS_VALID(intval)) { 4374 return (SET_ERROR(ERANGE)); 4375 } 4376 } 4377 break; 4378 4379 case ZFS_PROP_COPIES: 4380 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS)) 4381 return (SET_ERROR(ENOTSUP)); 4382 break; 4383 4384 case ZFS_PROP_RECORDSIZE: 4385 /* Record sizes above 128k need the feature to be enabled */ 4386 if (nvpair_value_uint64(pair, &intval) == 0 && 4387 intval > SPA_OLD_MAXBLOCKSIZE) { 4388 spa_t *spa; 4389 4390 /* 4391 * We don't allow setting the property above 1MB, 4392 * unless the tunable has been changed. 4393 */ 4394 if (intval > zfs_max_recordsize || 4395 intval > SPA_MAXBLOCKSIZE) 4396 return (SET_ERROR(ERANGE)); 4397 4398 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 4399 return (err); 4400 4401 if (!spa_feature_is_enabled(spa, 4402 SPA_FEATURE_LARGE_BLOCKS)) { 4403 spa_close(spa, FTAG); 4404 return (SET_ERROR(ENOTSUP)); 4405 } 4406 spa_close(spa, FTAG); 4407 } 4408 break; 4409 4410 case ZFS_PROP_DNODESIZE: 4411 /* Dnode sizes above 512 need the feature to be enabled */ 4412 if (nvpair_value_uint64(pair, &intval) == 0 && 4413 intval != ZFS_DNSIZE_LEGACY) { 4414 spa_t *spa; 4415 4416 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 4417 return (err); 4418 4419 if (!spa_feature_is_enabled(spa, 4420 SPA_FEATURE_LARGE_DNODE)) { 4421 spa_close(spa, FTAG); 4422 return (SET_ERROR(ENOTSUP)); 4423 } 4424 spa_close(spa, FTAG); 4425 } 4426 break; 4427 4428 case ZFS_PROP_SPECIAL_SMALL_BLOCKS: 4429 /* 4430 * This property could require the allocation classes 4431 * feature to be active for setting, however we allow 4432 * it so that tests of settable properties succeed. 4433 * The CLI will issue a warning in this case. 4434 */ 4435 break; 4436 4437 case ZFS_PROP_SHARESMB: 4438 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID)) 4439 return (SET_ERROR(ENOTSUP)); 4440 break; 4441 4442 case ZFS_PROP_ACLINHERIT: 4443 if (nvpair_type(pair) == DATA_TYPE_UINT64 && 4444 nvpair_value_uint64(pair, &intval) == 0) { 4445 if (intval == ZFS_ACL_PASSTHROUGH_X && 4446 zfs_earlier_version(dsname, 4447 SPA_VERSION_PASSTHROUGH_X)) 4448 return (SET_ERROR(ENOTSUP)); 4449 } 4450 break; 4451 4452 case ZFS_PROP_CHECKSUM: 4453 case ZFS_PROP_DEDUP: 4454 { 4455 spa_feature_t feature; 4456 spa_t *spa; 4457 4458 /* dedup feature version checks */ 4459 if (prop == ZFS_PROP_DEDUP && 4460 zfs_earlier_version(dsname, SPA_VERSION_DEDUP)) 4461 return (SET_ERROR(ENOTSUP)); 4462 4463 if (nvpair_value_uint64(pair, &intval) != 0) 4464 return (SET_ERROR(EINVAL)); 4465 4466 /* check prop value is enabled in features */ 4467 feature = zio_checksum_to_feature(intval & ZIO_CHECKSUM_MASK); 4468 if (feature == SPA_FEATURE_NONE) 4469 break; 4470 4471 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 4472 return (err); 4473 4474 if (!spa_feature_is_enabled(spa, feature)) { 4475 spa_close(spa, FTAG); 4476 return (SET_ERROR(ENOTSUP)); 4477 } 4478 spa_close(spa, FTAG); 4479 break; 4480 } 4481 } 4482 4483 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED())); 4484 } 4485 4486 /* 4487 * Checks for a race condition to make sure we don't increment a feature flag 4488 * multiple times. 4489 */ 4490 static int 4491 zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx) 4492 { 4493 spa_t *spa = dmu_tx_pool(tx)->dp_spa; 4494 spa_feature_t *featurep = arg; 4495 4496 if (!spa_feature_is_active(spa, *featurep)) 4497 return (0); 4498 else 4499 return (SET_ERROR(EBUSY)); 4500 } 4501 4502 /* 4503 * The callback invoked on feature activation in the sync task caused by 4504 * zfs_prop_activate_feature. 4505 */ 4506 static void 4507 zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx) 4508 { 4509 spa_t *spa = dmu_tx_pool(tx)->dp_spa; 4510 spa_feature_t *featurep = arg; 4511 4512 spa_feature_incr(spa, *featurep, tx); 4513 } 4514 4515 /* 4516 * Activates a feature on a pool in response to a property setting. This 4517 * creates a new sync task which modifies the pool to reflect the feature 4518 * as being active. 4519 */ 4520 static int 4521 zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature) 4522 { 4523 int err; 4524 4525 /* EBUSY here indicates that the feature is already active */ 4526 err = dsl_sync_task(spa_name(spa), 4527 zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync, 4528 &feature, 2, ZFS_SPACE_CHECK_RESERVED); 4529 4530 if (err != 0 && err != EBUSY) 4531 return (err); 4532 else 4533 return (0); 4534 } 4535 4536 /* 4537 * Removes properties from the given props list that fail permission checks 4538 * needed to clear them and to restore them in case of a receive error. For each 4539 * property, make sure we have both set and inherit permissions. 4540 * 4541 * Returns the first error encountered if any permission checks fail. If the 4542 * caller provides a non-NULL errlist, it also gives the complete list of names 4543 * of all the properties that failed a permission check along with the 4544 * corresponding error numbers. The caller is responsible for freeing the 4545 * returned errlist. 4546 * 4547 * If every property checks out successfully, zero is returned and the list 4548 * pointed at by errlist is NULL. 4549 */ 4550 static int 4551 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist) 4552 { 4553 zfs_cmd_t *zc; 4554 nvpair_t *pair, *next_pair; 4555 nvlist_t *errors; 4556 int err, rv = 0; 4557 4558 if (props == NULL) 4559 return (0); 4560 4561 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0); 4562 4563 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP); 4564 (void) strcpy(zc->zc_name, dataset); 4565 pair = nvlist_next_nvpair(props, NULL); 4566 while (pair != NULL) { 4567 next_pair = nvlist_next_nvpair(props, pair); 4568 4569 (void) strcpy(zc->zc_value, nvpair_name(pair)); 4570 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 || 4571 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) { 4572 VERIFY(nvlist_remove_nvpair(props, pair) == 0); 4573 VERIFY(nvlist_add_int32(errors, 4574 zc->zc_value, err) == 0); 4575 } 4576 pair = next_pair; 4577 } 4578 kmem_free(zc, sizeof (zfs_cmd_t)); 4579 4580 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) { 4581 nvlist_free(errors); 4582 errors = NULL; 4583 } else { 4584 VERIFY(nvpair_value_int32(pair, &rv) == 0); 4585 } 4586 4587 if (errlist == NULL) 4588 nvlist_free(errors); 4589 else 4590 *errlist = errors; 4591 4592 return (rv); 4593 } 4594 4595 static boolean_t 4596 propval_equals(nvpair_t *p1, nvpair_t *p2) 4597 { 4598 if (nvpair_type(p1) == DATA_TYPE_NVLIST) { 4599 /* dsl_prop_get_all_impl() format */ 4600 nvlist_t *attrs; 4601 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0); 4602 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 4603 &p1) == 0); 4604 } 4605 4606 if (nvpair_type(p2) == DATA_TYPE_NVLIST) { 4607 nvlist_t *attrs; 4608 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0); 4609 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 4610 &p2) == 0); 4611 } 4612 4613 if (nvpair_type(p1) != nvpair_type(p2)) 4614 return (B_FALSE); 4615 4616 if (nvpair_type(p1) == DATA_TYPE_STRING) { 4617 char *valstr1, *valstr2; 4618 4619 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0); 4620 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0); 4621 return (strcmp(valstr1, valstr2) == 0); 4622 } else { 4623 uint64_t intval1, intval2; 4624 4625 VERIFY(nvpair_value_uint64(p1, &intval1) == 0); 4626 VERIFY(nvpair_value_uint64(p2, &intval2) == 0); 4627 return (intval1 == intval2); 4628 } 4629 } 4630 4631 /* 4632 * Remove properties from props if they are not going to change (as determined 4633 * by comparison with origprops). Remove them from origprops as well, since we 4634 * do not need to clear or restore properties that won't change. 4635 */ 4636 static void 4637 props_reduce(nvlist_t *props, nvlist_t *origprops) 4638 { 4639 nvpair_t *pair, *next_pair; 4640 4641 if (origprops == NULL) 4642 return; /* all props need to be received */ 4643 4644 pair = nvlist_next_nvpair(props, NULL); 4645 while (pair != NULL) { 4646 const char *propname = nvpair_name(pair); 4647 nvpair_t *match; 4648 4649 next_pair = nvlist_next_nvpair(props, pair); 4650 4651 if ((nvlist_lookup_nvpair(origprops, propname, 4652 &match) != 0) || !propval_equals(pair, match)) 4653 goto next; /* need to set received value */ 4654 4655 /* don't clear the existing received value */ 4656 (void) nvlist_remove_nvpair(origprops, match); 4657 /* don't bother receiving the property */ 4658 (void) nvlist_remove_nvpair(props, pair); 4659 next: 4660 pair = next_pair; 4661 } 4662 } 4663 4664 /* 4665 * Extract properties that cannot be set PRIOR to the receipt of a dataset. 4666 * For example, refquota cannot be set until after the receipt of a dataset, 4667 * because in replication streams, an older/earlier snapshot may exceed the 4668 * refquota. We want to receive the older/earlier snapshot, but setting 4669 * refquota pre-receipt will set the dsl's ACTUAL quota, which will prevent 4670 * the older/earlier snapshot from being received (with EDQUOT). 4671 * 4672 * The ZFS test "zfs_receive_011_pos" demonstrates such a scenario. 4673 * 4674 * libzfs will need to be judicious handling errors encountered by props 4675 * extracted by this function. 4676 */ 4677 static nvlist_t * 4678 extract_delay_props(nvlist_t *props) 4679 { 4680 nvlist_t *delayprops; 4681 nvpair_t *nvp, *tmp; 4682 static const zfs_prop_t delayable[] = { 4683 ZFS_PROP_REFQUOTA, 4684 ZFS_PROP_KEYLOCATION, 4685 0 4686 }; 4687 int i; 4688 4689 VERIFY(nvlist_alloc(&delayprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 4690 4691 for (nvp = nvlist_next_nvpair(props, NULL); nvp != NULL; 4692 nvp = nvlist_next_nvpair(props, nvp)) { 4693 /* 4694 * strcmp() is safe because zfs_prop_to_name() always returns 4695 * a bounded string. 4696 */ 4697 for (i = 0; delayable[i] != 0; i++) { 4698 if (strcmp(zfs_prop_to_name(delayable[i]), 4699 nvpair_name(nvp)) == 0) { 4700 break; 4701 } 4702 } 4703 if (delayable[i] != 0) { 4704 tmp = nvlist_prev_nvpair(props, nvp); 4705 VERIFY(nvlist_add_nvpair(delayprops, nvp) == 0); 4706 VERIFY(nvlist_remove_nvpair(props, nvp) == 0); 4707 nvp = tmp; 4708 } 4709 } 4710 4711 if (nvlist_empty(delayprops)) { 4712 nvlist_free(delayprops); 4713 delayprops = NULL; 4714 } 4715 return (delayprops); 4716 } 4717 4718 #ifdef DEBUG 4719 static boolean_t zfs_ioc_recv_inject_err; 4720 #endif 4721 4722 /* 4723 * nvlist 'errors' is always allocated. It will contain descriptions of 4724 * encountered errors, if any. It's the callers responsibility to free. 4725 */ 4726 static int 4727 zfs_ioc_recv_impl(char *tofs, char *tosnap, char *origin, nvlist_t *recvprops, 4728 nvlist_t *localprops, nvlist_t *hidden_args, boolean_t force, 4729 boolean_t resumable, int input_fd, dmu_replay_record_t *begin_record, 4730 int cleanup_fd, uint64_t *read_bytes, uint64_t *errflags, 4731 uint64_t *action_handle, nvlist_t **errors) 4732 { 4733 dmu_recv_cookie_t drc; 4734 int error = 0; 4735 int props_error = 0; 4736 offset_t off; 4737 nvlist_t *local_delayprops = NULL; 4738 nvlist_t *recv_delayprops = NULL; 4739 nvlist_t *origprops = NULL; /* existing properties */ 4740 nvlist_t *origrecvd = NULL; /* existing received properties */ 4741 boolean_t first_recvd_props = B_FALSE; 4742 file_t *input_fp; 4743 4744 *read_bytes = 0; 4745 *errflags = 0; 4746 *errors = fnvlist_alloc(); 4747 4748 input_fp = getf(input_fd); 4749 if (input_fp == NULL) 4750 return (SET_ERROR(EBADF)); 4751 4752 error = dmu_recv_begin(tofs, tosnap, begin_record, force, 4753 resumable, localprops, hidden_args, origin, &drc); 4754 if (error != 0) 4755 goto out; 4756 4757 /* 4758 * Set properties before we receive the stream so that they are applied 4759 * to the new data. Note that we must call dmu_recv_stream() if 4760 * dmu_recv_begin() succeeds. 4761 */ 4762 if (recvprops != NULL && !drc.drc_newfs) { 4763 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >= 4764 SPA_VERSION_RECVD_PROPS && 4765 !dsl_prop_get_hasrecvd(tofs)) 4766 first_recvd_props = B_TRUE; 4767 4768 /* 4769 * If new received properties are supplied, they are to 4770 * completely replace the existing received properties, 4771 * so stash away the existing ones. 4772 */ 4773 if (dsl_prop_get_received(tofs, &origrecvd) == 0) { 4774 nvlist_t *errlist = NULL; 4775 /* 4776 * Don't bother writing a property if its value won't 4777 * change (and avoid the unnecessary security checks). 4778 * 4779 * The first receive after SPA_VERSION_RECVD_PROPS is a 4780 * special case where we blow away all local properties 4781 * regardless. 4782 */ 4783 if (!first_recvd_props) 4784 props_reduce(recvprops, origrecvd); 4785 if (zfs_check_clearable(tofs, origrecvd, &errlist) != 0) 4786 (void) nvlist_merge(*errors, errlist, 0); 4787 nvlist_free(errlist); 4788 4789 if (clear_received_props(tofs, origrecvd, 4790 first_recvd_props ? NULL : recvprops) != 0) 4791 *errflags |= ZPROP_ERR_NOCLEAR; 4792 } else { 4793 *errflags |= ZPROP_ERR_NOCLEAR; 4794 } 4795 } 4796 4797 /* 4798 * Stash away existing properties so we can restore them on error unless 4799 * we're doing the first receive after SPA_VERSION_RECVD_PROPS, in which 4800 * case "origrecvd" will take care of that. 4801 */ 4802 if (localprops != NULL && !drc.drc_newfs && !first_recvd_props) { 4803 objset_t *os; 4804 if (dmu_objset_hold(tofs, FTAG, &os) == 0) { 4805 if (dsl_prop_get_all(os, &origprops) != 0) { 4806 *errflags |= ZPROP_ERR_NOCLEAR; 4807 } 4808 dmu_objset_rele(os, FTAG); 4809 } else { 4810 *errflags |= ZPROP_ERR_NOCLEAR; 4811 } 4812 } 4813 4814 if (recvprops != NULL) { 4815 props_error = dsl_prop_set_hasrecvd(tofs); 4816 4817 if (props_error == 0) { 4818 recv_delayprops = extract_delay_props(recvprops); 4819 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED, 4820 recvprops, *errors); 4821 } 4822 } 4823 4824 if (localprops != NULL) { 4825 nvlist_t *oprops = fnvlist_alloc(); 4826 nvlist_t *xprops = fnvlist_alloc(); 4827 nvpair_t *nvp = NULL; 4828 4829 while ((nvp = nvlist_next_nvpair(localprops, nvp)) != NULL) { 4830 if (nvpair_type(nvp) == DATA_TYPE_BOOLEAN) { 4831 /* -x property */ 4832 const char *name = nvpair_name(nvp); 4833 zfs_prop_t prop = zfs_name_to_prop(name); 4834 if (prop != ZPROP_INVAL) { 4835 if (!zfs_prop_inheritable(prop)) 4836 continue; 4837 } else if (!zfs_prop_user(name)) 4838 continue; 4839 fnvlist_add_boolean(xprops, name); 4840 } else { 4841 /* -o property=value */ 4842 fnvlist_add_nvpair(oprops, nvp); 4843 } 4844 } 4845 4846 local_delayprops = extract_delay_props(oprops); 4847 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_LOCAL, 4848 oprops, *errors); 4849 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_INHERITED, 4850 xprops, *errors); 4851 4852 nvlist_free(oprops); 4853 nvlist_free(xprops); 4854 } 4855 4856 off = input_fp->f_offset; 4857 error = dmu_recv_stream(&drc, input_fp->f_vnode, &off, cleanup_fd, 4858 action_handle); 4859 4860 if (error == 0) { 4861 zfsvfs_t *zfsvfs = NULL; 4862 4863 if (getzfsvfs(tofs, &zfsvfs) == 0) { 4864 /* online recv */ 4865 dsl_dataset_t *ds; 4866 int end_err; 4867 4868 ds = dmu_objset_ds(zfsvfs->z_os); 4869 error = zfs_suspend_fs(zfsvfs); 4870 /* 4871 * If the suspend fails, then the recv_end will 4872 * likely also fail, and clean up after itself. 4873 */ 4874 end_err = dmu_recv_end(&drc, zfsvfs); 4875 if (error == 0) 4876 error = zfs_resume_fs(zfsvfs, ds); 4877 error = error ? error : end_err; 4878 VFS_RELE(zfsvfs->z_vfs); 4879 } else { 4880 error = dmu_recv_end(&drc, NULL); 4881 } 4882 4883 /* Set delayed properties now, after we're done receiving. */ 4884 if (recv_delayprops != NULL && error == 0) { 4885 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED, 4886 recv_delayprops, *errors); 4887 } 4888 if (local_delayprops != NULL && error == 0) { 4889 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_LOCAL, 4890 local_delayprops, *errors); 4891 } 4892 } 4893 4894 /* 4895 * Merge delayed props back in with initial props, in case 4896 * we're DEBUG and zfs_ioc_recv_inject_err is set (which means 4897 * we have to make sure clear_received_props() includes 4898 * the delayed properties). 4899 * 4900 * Since zfs_ioc_recv_inject_err is only in DEBUG kernels, 4901 * using ASSERT() will be just like a VERIFY. 4902 */ 4903 if (recv_delayprops != NULL) { 4904 ASSERT(nvlist_merge(recvprops, recv_delayprops, 0) == 0); 4905 nvlist_free(recv_delayprops); 4906 } 4907 if (local_delayprops != NULL) { 4908 ASSERT(nvlist_merge(localprops, local_delayprops, 0) == 0); 4909 nvlist_free(local_delayprops); 4910 } 4911 4912 *read_bytes = off - input_fp->f_offset; 4913 if (VOP_SEEK(input_fp->f_vnode, input_fp->f_offset, &off, NULL) == 0) 4914 input_fp->f_offset = off; 4915 4916 #ifdef DEBUG 4917 if (zfs_ioc_recv_inject_err) { 4918 zfs_ioc_recv_inject_err = B_FALSE; 4919 error = 1; 4920 } 4921 #endif 4922 4923 /* 4924 * On error, restore the original props. 4925 */ 4926 if (error != 0 && recvprops != NULL && !drc.drc_newfs) { 4927 if (clear_received_props(tofs, recvprops, NULL) != 0) { 4928 /* 4929 * We failed to clear the received properties. 4930 * Since we may have left a $recvd value on the 4931 * system, we can't clear the $hasrecvd flag. 4932 */ 4933 *errflags |= ZPROP_ERR_NORESTORE; 4934 } else if (first_recvd_props) { 4935 dsl_prop_unset_hasrecvd(tofs); 4936 } 4937 4938 if (origrecvd == NULL && !drc.drc_newfs) { 4939 /* We failed to stash the original properties. */ 4940 *errflags |= ZPROP_ERR_NORESTORE; 4941 } 4942 4943 /* 4944 * dsl_props_set() will not convert RECEIVED to LOCAL on or 4945 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL 4946 * explicitly if we're restoring local properties cleared in the 4947 * first new-style receive. 4948 */ 4949 if (origrecvd != NULL && 4950 zfs_set_prop_nvlist(tofs, (first_recvd_props ? 4951 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED), 4952 origrecvd, NULL) != 0) { 4953 /* 4954 * We stashed the original properties but failed to 4955 * restore them. 4956 */ 4957 *errflags |= ZPROP_ERR_NORESTORE; 4958 } 4959 } 4960 if (error != 0 && localprops != NULL && !drc.drc_newfs && 4961 !first_recvd_props) { 4962 nvlist_t *setprops; 4963 nvlist_t *inheritprops; 4964 nvpair_t *nvp; 4965 4966 if (origprops == NULL) { 4967 /* We failed to stash the original properties. */ 4968 *errflags |= ZPROP_ERR_NORESTORE; 4969 goto out; 4970 } 4971 4972 /* Restore original props */ 4973 setprops = fnvlist_alloc(); 4974 inheritprops = fnvlist_alloc(); 4975 nvp = NULL; 4976 while ((nvp = nvlist_next_nvpair(localprops, nvp)) != NULL) { 4977 const char *name = nvpair_name(nvp); 4978 const char *source; 4979 nvlist_t *attrs; 4980 4981 if (!nvlist_exists(origprops, name)) { 4982 /* 4983 * Property was not present or was explicitly 4984 * inherited before the receive, restore this. 4985 */ 4986 fnvlist_add_boolean(inheritprops, name); 4987 continue; 4988 } 4989 attrs = fnvlist_lookup_nvlist(origprops, name); 4990 source = fnvlist_lookup_string(attrs, ZPROP_SOURCE); 4991 4992 /* Skip received properties */ 4993 if (strcmp(source, ZPROP_SOURCE_VAL_RECVD) == 0) 4994 continue; 4995 4996 if (strcmp(source, tofs) == 0) { 4997 /* Property was locally set */ 4998 fnvlist_add_nvlist(setprops, name, attrs); 4999 } else { 5000 /* Property was implicitly inherited */ 5001 fnvlist_add_boolean(inheritprops, name); 5002 } 5003 } 5004 5005 if (zfs_set_prop_nvlist(tofs, ZPROP_SRC_LOCAL, setprops, 5006 NULL) != 0) 5007 *errflags |= ZPROP_ERR_NORESTORE; 5008 if (zfs_set_prop_nvlist(tofs, ZPROP_SRC_INHERITED, inheritprops, 5009 NULL) != 0) 5010 *errflags |= ZPROP_ERR_NORESTORE; 5011 5012 nvlist_free(setprops); 5013 nvlist_free(inheritprops); 5014 } 5015 out: 5016 releasef(input_fd); 5017 nvlist_free(origrecvd); 5018 nvlist_free(origprops); 5019 5020 if (error == 0) 5021 error = props_error; 5022 5023 return (error); 5024 } 5025 5026 /* 5027 * inputs: 5028 * zc_name name of containing filesystem 5029 * zc_nvlist_src{_size} nvlist of received properties to apply 5030 * zc_nvlist_conf{_size} nvlist of local properties to apply 5031 * zc_history_offset{_len} nvlist of hidden args { "wkeydata" -> value } 5032 * zc_value name of snapshot to create 5033 * zc_string name of clone origin (if DRR_FLAG_CLONE) 5034 * zc_cookie file descriptor to recv from 5035 * zc_begin_record the BEGIN record of the stream (not byteswapped) 5036 * zc_guid force flag 5037 * zc_cleanup_fd cleanup-on-exit file descriptor 5038 * zc_action_handle handle for this guid/ds mapping (or zero on first call) 5039 * zc_resumable if data is incomplete assume sender will resume 5040 * 5041 * outputs: 5042 * zc_cookie number of bytes read 5043 * zc_nvlist_dst{_size} error for each unapplied received property 5044 * zc_obj zprop_errflags_t 5045 * zc_action_handle handle for this guid/ds mapping 5046 */ 5047 static int 5048 zfs_ioc_recv(zfs_cmd_t *zc) 5049 { 5050 dmu_replay_record_t begin_record; 5051 nvlist_t *errors = NULL; 5052 nvlist_t *recvdprops = NULL; 5053 nvlist_t *localprops = NULL; 5054 nvlist_t *hidden_args = NULL; 5055 char *origin = NULL; 5056 char *tosnap; 5057 char tofs[ZFS_MAX_DATASET_NAME_LEN]; 5058 int error = 0; 5059 5060 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 5061 strchr(zc->zc_value, '@') == NULL || 5062 strchr(zc->zc_value, '%')) 5063 return (SET_ERROR(EINVAL)); 5064 5065 (void) strlcpy(tofs, zc->zc_value, sizeof (tofs)); 5066 tosnap = strchr(tofs, '@'); 5067 *tosnap++ = '\0'; 5068 5069 if (zc->zc_nvlist_src != 0 && 5070 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 5071 zc->zc_iflags, &recvdprops)) != 0) 5072 return (error); 5073 5074 if (zc->zc_nvlist_conf != 0 && 5075 (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 5076 zc->zc_iflags, &localprops)) != 0) 5077 return (error); 5078 5079 if (zc->zc_history_offset != 0 && 5080 (error = get_nvlist(zc->zc_history_offset, zc->zc_history_len, 5081 zc->zc_iflags, &hidden_args)) != 0) 5082 return (error); 5083 5084 if (zc->zc_string[0]) 5085 origin = zc->zc_string; 5086 5087 begin_record.drr_type = DRR_BEGIN; 5088 begin_record.drr_payloadlen = zc->zc_begin_record.drr_payloadlen; 5089 begin_record.drr_u.drr_begin = zc->zc_begin_record.drr_u.drr_begin; 5090 5091 error = zfs_ioc_recv_impl(tofs, tosnap, origin, recvdprops, localprops, 5092 hidden_args, zc->zc_guid, zc->zc_resumable, zc->zc_cookie, 5093 &begin_record, zc->zc_cleanup_fd, &zc->zc_cookie, &zc->zc_obj, 5094 &zc->zc_action_handle, &errors); 5095 nvlist_free(recvdprops); 5096 nvlist_free(localprops); 5097 5098 /* 5099 * Now that all props, initial and delayed, are set, report the prop 5100 * errors to the caller. 5101 */ 5102 if (zc->zc_nvlist_dst_size != 0 && errors != NULL && 5103 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 || 5104 put_nvlist(zc, errors) != 0)) { 5105 /* 5106 * Caller made zc->zc_nvlist_dst less than the minimum expected 5107 * size or supplied an invalid address. 5108 */ 5109 error = SET_ERROR(EINVAL); 5110 } 5111 5112 nvlist_free(errors); 5113 5114 return (error); 5115 } 5116 5117 /* 5118 * inputs: 5119 * zc_name name of snapshot to send 5120 * zc_cookie file descriptor to send stream to 5121 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj) 5122 * zc_sendobj objsetid of snapshot to send 5123 * zc_fromobj objsetid of incremental fromsnap (may be zero) 5124 * zc_guid if set, estimate size of stream only. zc_cookie is ignored. 5125 * output size in zc_objset_type. 5126 * zc_flags lzc_send_flags 5127 * 5128 * outputs: 5129 * zc_objset_type estimated size, if zc_guid is set 5130 */ 5131 static int 5132 zfs_ioc_send(zfs_cmd_t *zc) 5133 { 5134 int error; 5135 offset_t off; 5136 boolean_t estimate = (zc->zc_guid != 0); 5137 boolean_t embedok = (zc->zc_flags & 0x1); 5138 boolean_t large_block_ok = (zc->zc_flags & 0x2); 5139 boolean_t compressok = (zc->zc_flags & 0x4); 5140 boolean_t rawok = (zc->zc_flags & 0x8); 5141 5142 if (zc->zc_obj != 0) { 5143 dsl_pool_t *dp; 5144 dsl_dataset_t *tosnap; 5145 5146 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 5147 if (error != 0) 5148 return (error); 5149 5150 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap); 5151 if (error != 0) { 5152 dsl_pool_rele(dp, FTAG); 5153 return (error); 5154 } 5155 5156 if (dsl_dir_is_clone(tosnap->ds_dir)) 5157 zc->zc_fromobj = 5158 dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj; 5159 dsl_dataset_rele(tosnap, FTAG); 5160 dsl_pool_rele(dp, FTAG); 5161 } 5162 5163 if (estimate) { 5164 dsl_pool_t *dp; 5165 dsl_dataset_t *tosnap; 5166 dsl_dataset_t *fromsnap = NULL; 5167 5168 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 5169 if (error != 0) 5170 return (error); 5171 5172 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, 5173 FTAG, &tosnap); 5174 if (error != 0) { 5175 dsl_pool_rele(dp, FTAG); 5176 return (error); 5177 } 5178 5179 if (zc->zc_fromobj != 0) { 5180 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj, 5181 FTAG, &fromsnap); 5182 if (error != 0) { 5183 dsl_dataset_rele(tosnap, FTAG); 5184 dsl_pool_rele(dp, FTAG); 5185 return (error); 5186 } 5187 } 5188 5189 error = dmu_send_estimate(tosnap, fromsnap, compressok || rawok, 5190 &zc->zc_objset_type); 5191 5192 if (fromsnap != NULL) 5193 dsl_dataset_rele(fromsnap, FTAG); 5194 dsl_dataset_rele(tosnap, FTAG); 5195 dsl_pool_rele(dp, FTAG); 5196 } else { 5197 file_t *fp = getf(zc->zc_cookie); 5198 if (fp == NULL) 5199 return (SET_ERROR(EBADF)); 5200 5201 off = fp->f_offset; 5202 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj, 5203 zc->zc_fromobj, embedok, large_block_ok, compressok, rawok, 5204 zc->zc_cookie, fp->f_vnode, &off); 5205 5206 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0) 5207 fp->f_offset = off; 5208 releasef(zc->zc_cookie); 5209 } 5210 return (error); 5211 } 5212 5213 /* 5214 * inputs: 5215 * zc_name name of snapshot on which to report progress 5216 * zc_cookie file descriptor of send stream 5217 * 5218 * outputs: 5219 * zc_cookie number of bytes written in send stream thus far 5220 */ 5221 static int 5222 zfs_ioc_send_progress(zfs_cmd_t *zc) 5223 { 5224 dsl_pool_t *dp; 5225 dsl_dataset_t *ds; 5226 dmu_sendarg_t *dsp = NULL; 5227 int error; 5228 5229 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 5230 if (error != 0) 5231 return (error); 5232 5233 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds); 5234 if (error != 0) { 5235 dsl_pool_rele(dp, FTAG); 5236 return (error); 5237 } 5238 5239 mutex_enter(&ds->ds_sendstream_lock); 5240 5241 /* 5242 * Iterate over all the send streams currently active on this dataset. 5243 * If there's one which matches the specified file descriptor _and_ the 5244 * stream was started by the current process, return the progress of 5245 * that stream. 5246 */ 5247 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL; 5248 dsp = list_next(&ds->ds_sendstreams, dsp)) { 5249 if (dsp->dsa_outfd == zc->zc_cookie && 5250 dsp->dsa_proc == curproc) 5251 break; 5252 } 5253 5254 if (dsp != NULL) 5255 zc->zc_cookie = *(dsp->dsa_off); 5256 else 5257 error = SET_ERROR(ENOENT); 5258 5259 mutex_exit(&ds->ds_sendstream_lock); 5260 dsl_dataset_rele(ds, FTAG); 5261 dsl_pool_rele(dp, FTAG); 5262 return (error); 5263 } 5264 5265 static int 5266 zfs_ioc_inject_fault(zfs_cmd_t *zc) 5267 { 5268 int id, error; 5269 5270 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id, 5271 &zc->zc_inject_record); 5272 5273 if (error == 0) 5274 zc->zc_guid = (uint64_t)id; 5275 5276 return (error); 5277 } 5278 5279 static int 5280 zfs_ioc_clear_fault(zfs_cmd_t *zc) 5281 { 5282 return (zio_clear_fault((int)zc->zc_guid)); 5283 } 5284 5285 static int 5286 zfs_ioc_inject_list_next(zfs_cmd_t *zc) 5287 { 5288 int id = (int)zc->zc_guid; 5289 int error; 5290 5291 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name), 5292 &zc->zc_inject_record); 5293 5294 zc->zc_guid = id; 5295 5296 return (error); 5297 } 5298 5299 static int 5300 zfs_ioc_error_log(zfs_cmd_t *zc) 5301 { 5302 spa_t *spa; 5303 int error; 5304 size_t count = (size_t)zc->zc_nvlist_dst_size; 5305 5306 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 5307 return (error); 5308 5309 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst, 5310 &count); 5311 if (error == 0) 5312 zc->zc_nvlist_dst_size = count; 5313 else 5314 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa); 5315 5316 spa_close(spa, FTAG); 5317 5318 return (error); 5319 } 5320 5321 static int 5322 zfs_ioc_clear(zfs_cmd_t *zc) 5323 { 5324 spa_t *spa; 5325 vdev_t *vd; 5326 int error; 5327 5328 /* 5329 * On zpool clear we also fix up missing slogs 5330 */ 5331 mutex_enter(&spa_namespace_lock); 5332 spa = spa_lookup(zc->zc_name); 5333 if (spa == NULL) { 5334 mutex_exit(&spa_namespace_lock); 5335 return (SET_ERROR(EIO)); 5336 } 5337 if (spa_get_log_state(spa) == SPA_LOG_MISSING) { 5338 /* we need to let spa_open/spa_load clear the chains */ 5339 spa_set_log_state(spa, SPA_LOG_CLEAR); 5340 } 5341 spa->spa_last_open_failed = 0; 5342 mutex_exit(&spa_namespace_lock); 5343 5344 if (zc->zc_cookie & ZPOOL_NO_REWIND) { 5345 error = spa_open(zc->zc_name, &spa, FTAG); 5346 } else { 5347 nvlist_t *policy; 5348 nvlist_t *config = NULL; 5349 5350 if (zc->zc_nvlist_src == 0) 5351 return (SET_ERROR(EINVAL)); 5352 5353 if ((error = get_nvlist(zc->zc_nvlist_src, 5354 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) { 5355 error = spa_open_rewind(zc->zc_name, &spa, FTAG, 5356 policy, &config); 5357 if (config != NULL) { 5358 int err; 5359 5360 if ((err = put_nvlist(zc, config)) != 0) 5361 error = err; 5362 nvlist_free(config); 5363 } 5364 nvlist_free(policy); 5365 } 5366 } 5367 5368 if (error != 0) 5369 return (error); 5370 5371 /* 5372 * If multihost is enabled, resuming I/O is unsafe as another 5373 * host may have imported the pool. 5374 */ 5375 if (spa_multihost(spa) && spa_suspended(spa)) 5376 return (SET_ERROR(EINVAL)); 5377 5378 spa_vdev_state_enter(spa, SCL_NONE); 5379 5380 if (zc->zc_guid == 0) { 5381 vd = NULL; 5382 } else { 5383 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE); 5384 if (vd == NULL) { 5385 (void) spa_vdev_state_exit(spa, NULL, ENODEV); 5386 spa_close(spa, FTAG); 5387 return (SET_ERROR(ENODEV)); 5388 } 5389 } 5390 5391 vdev_clear(spa, vd); 5392 5393 (void) spa_vdev_state_exit(spa, NULL, 0); 5394 5395 /* 5396 * Resume any suspended I/Os. 5397 */ 5398 if (zio_resume(spa) != 0) 5399 error = SET_ERROR(EIO); 5400 5401 spa_close(spa, FTAG); 5402 5403 return (error); 5404 } 5405 5406 static int 5407 zfs_ioc_pool_reopen(zfs_cmd_t *zc) 5408 { 5409 spa_t *spa; 5410 int error; 5411 5412 error = spa_open(zc->zc_name, &spa, FTAG); 5413 if (error != 0) 5414 return (error); 5415 5416 spa_vdev_state_enter(spa, SCL_NONE); 5417 5418 /* 5419 * If a resilver is already in progress then set the 5420 * spa_scrub_reopen flag to B_TRUE so that we don't restart 5421 * the scan as a side effect of the reopen. Otherwise, let 5422 * vdev_open() decided if a resilver is required. 5423 */ 5424 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool); 5425 vdev_reopen(spa->spa_root_vdev); 5426 spa->spa_scrub_reopen = B_FALSE; 5427 5428 (void) spa_vdev_state_exit(spa, NULL, 0); 5429 spa_close(spa, FTAG); 5430 return (0); 5431 } 5432 /* 5433 * inputs: 5434 * zc_name name of filesystem 5435 * 5436 * outputs: 5437 * zc_string name of conflicting snapshot, if there is one 5438 */ 5439 static int 5440 zfs_ioc_promote(zfs_cmd_t *zc) 5441 { 5442 dsl_pool_t *dp; 5443 dsl_dataset_t *ds, *ods; 5444 char origin[ZFS_MAX_DATASET_NAME_LEN]; 5445 char *cp; 5446 int error; 5447 5448 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0'; 5449 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0 || 5450 strchr(zc->zc_name, '%')) 5451 return (SET_ERROR(EINVAL)); 5452 5453 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 5454 if (error != 0) 5455 return (error); 5456 5457 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds); 5458 if (error != 0) { 5459 dsl_pool_rele(dp, FTAG); 5460 return (error); 5461 } 5462 5463 if (!dsl_dir_is_clone(ds->ds_dir)) { 5464 dsl_dataset_rele(ds, FTAG); 5465 dsl_pool_rele(dp, FTAG); 5466 return (SET_ERROR(EINVAL)); 5467 } 5468 5469 error = dsl_dataset_hold_obj(dp, 5470 dsl_dir_phys(ds->ds_dir)->dd_origin_obj, FTAG, &ods); 5471 if (error != 0) { 5472 dsl_dataset_rele(ds, FTAG); 5473 dsl_pool_rele(dp, FTAG); 5474 return (error); 5475 } 5476 5477 dsl_dataset_name(ods, origin); 5478 dsl_dataset_rele(ods, FTAG); 5479 dsl_dataset_rele(ds, FTAG); 5480 dsl_pool_rele(dp, FTAG); 5481 5482 /* 5483 * We don't need to unmount *all* the origin fs's snapshots, but 5484 * it's easier. 5485 */ 5486 cp = strchr(origin, '@'); 5487 if (cp) 5488 *cp = '\0'; 5489 (void) dmu_objset_find(origin, 5490 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS); 5491 return (dsl_dataset_promote(zc->zc_name, zc->zc_string)); 5492 } 5493 5494 /* 5495 * Retrieve a single {user|group|project}{used|quota}@... property. 5496 * 5497 * inputs: 5498 * zc_name name of filesystem 5499 * zc_objset_type zfs_userquota_prop_t 5500 * zc_value domain name (eg. "S-1-234-567-89") 5501 * zc_guid RID/UID/GID 5502 * 5503 * outputs: 5504 * zc_cookie property value 5505 */ 5506 static int 5507 zfs_ioc_userspace_one(zfs_cmd_t *zc) 5508 { 5509 zfsvfs_t *zfsvfs; 5510 int error; 5511 5512 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 5513 return (SET_ERROR(EINVAL)); 5514 5515 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE); 5516 if (error != 0) 5517 return (error); 5518 5519 error = zfs_userspace_one(zfsvfs, 5520 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie); 5521 zfsvfs_rele(zfsvfs, FTAG); 5522 5523 return (error); 5524 } 5525 5526 /* 5527 * inputs: 5528 * zc_name name of filesystem 5529 * zc_cookie zap cursor 5530 * zc_objset_type zfs_userquota_prop_t 5531 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist) 5532 * 5533 * outputs: 5534 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t) 5535 * zc_cookie zap cursor 5536 */ 5537 static int 5538 zfs_ioc_userspace_many(zfs_cmd_t *zc) 5539 { 5540 zfsvfs_t *zfsvfs; 5541 int bufsize = zc->zc_nvlist_dst_size; 5542 5543 if (bufsize <= 0) 5544 return (SET_ERROR(ENOMEM)); 5545 5546 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE); 5547 if (error != 0) 5548 return (error); 5549 5550 void *buf = kmem_alloc(bufsize, KM_SLEEP); 5551 5552 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie, 5553 buf, &zc->zc_nvlist_dst_size); 5554 5555 if (error == 0) { 5556 error = xcopyout(buf, 5557 (void *)(uintptr_t)zc->zc_nvlist_dst, 5558 zc->zc_nvlist_dst_size); 5559 } 5560 kmem_free(buf, bufsize); 5561 zfsvfs_rele(zfsvfs, FTAG); 5562 5563 return (error); 5564 } 5565 5566 /* 5567 * inputs: 5568 * zc_name name of filesystem 5569 * 5570 * outputs: 5571 * none 5572 */ 5573 static int 5574 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc) 5575 { 5576 objset_t *os; 5577 int error = 0; 5578 zfsvfs_t *zfsvfs; 5579 5580 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) { 5581 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) { 5582 /* 5583 * If userused is not enabled, it may be because the 5584 * objset needs to be closed & reopened (to grow the 5585 * objset_phys_t). Suspend/resume the fs will do that. 5586 */ 5587 dsl_dataset_t *ds, *newds; 5588 5589 ds = dmu_objset_ds(zfsvfs->z_os); 5590 error = zfs_suspend_fs(zfsvfs); 5591 if (error == 0) { 5592 dmu_objset_refresh_ownership(ds, &newds, 5593 B_TRUE, zfsvfs); 5594 error = zfs_resume_fs(zfsvfs, newds); 5595 } 5596 } 5597 if (error == 0) 5598 error = dmu_objset_userspace_upgrade(zfsvfs->z_os); 5599 VFS_RELE(zfsvfs->z_vfs); 5600 } else { 5601 /* XXX kind of reading contents without owning */ 5602 error = dmu_objset_hold_flags(zc->zc_name, B_TRUE, FTAG, &os); 5603 if (error != 0) 5604 return (error); 5605 5606 error = dmu_objset_userspace_upgrade(os); 5607 dmu_objset_rele_flags(os, B_TRUE, FTAG); 5608 } 5609 5610 return (error); 5611 } 5612 5613 /* 5614 * inputs: 5615 * zc_name name of filesystem 5616 * 5617 * outputs: 5618 * none 5619 */ 5620 static int 5621 zfs_ioc_id_quota_upgrade(zfs_cmd_t *zc) 5622 { 5623 objset_t *os; 5624 int error; 5625 5626 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 5627 if (error != 0) 5628 return (error); 5629 5630 if (dmu_objset_userobjspace_upgradable(os) || 5631 dmu_objset_projectquota_upgradable(os)) { 5632 mutex_enter(&os->os_upgrade_lock); 5633 if (os->os_upgrade_id == 0) { 5634 /* clear potential error code and retry */ 5635 os->os_upgrade_status = 0; 5636 mutex_exit(&os->os_upgrade_lock); 5637 5638 dmu_objset_id_quota_upgrade(os); 5639 } else { 5640 mutex_exit(&os->os_upgrade_lock); 5641 } 5642 5643 dsl_pool_rele(dmu_objset_pool(os), FTAG); 5644 5645 taskq_wait_id(os->os_spa->spa_upgrade_taskq, os->os_upgrade_id); 5646 error = os->os_upgrade_status; 5647 } else { 5648 dsl_pool_rele(dmu_objset_pool(os), FTAG); 5649 } 5650 5651 dsl_dataset_rele(dmu_objset_ds(os), FTAG); 5652 5653 return (error); 5654 } 5655 5656 /* 5657 * We don't want to have a hard dependency 5658 * against some special symbols in sharefs 5659 * nfs, and smbsrv. Determine them if needed when 5660 * the first file system is shared. 5661 * Neither sharefs, nfs or smbsrv are unloadable modules. 5662 */ 5663 int (*znfsexport_fs)(void *arg); 5664 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t); 5665 int (*zsmbexport_fs)(void *arg, boolean_t add_share); 5666 5667 int zfs_nfsshare_inited; 5668 int zfs_smbshare_inited; 5669 5670 ddi_modhandle_t nfs_mod; 5671 ddi_modhandle_t sharefs_mod; 5672 ddi_modhandle_t smbsrv_mod; 5673 kmutex_t zfs_share_lock; 5674 5675 static int 5676 zfs_init_sharefs() 5677 { 5678 int error; 5679 5680 ASSERT(MUTEX_HELD(&zfs_share_lock)); 5681 /* Both NFS and SMB shares also require sharetab support. */ 5682 if (sharefs_mod == NULL && ((sharefs_mod = 5683 ddi_modopen("fs/sharefs", 5684 KRTLD_MODE_FIRST, &error)) == NULL)) { 5685 return (SET_ERROR(ENOSYS)); 5686 } 5687 if (zshare_fs == NULL && ((zshare_fs = 5688 (int (*)(enum sharefs_sys_op, share_t *, uint32_t)) 5689 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) { 5690 return (SET_ERROR(ENOSYS)); 5691 } 5692 return (0); 5693 } 5694 5695 static int 5696 zfs_ioc_share(zfs_cmd_t *zc) 5697 { 5698 int error; 5699 int opcode; 5700 5701 switch (zc->zc_share.z_sharetype) { 5702 case ZFS_SHARE_NFS: 5703 case ZFS_UNSHARE_NFS: 5704 if (zfs_nfsshare_inited == 0) { 5705 mutex_enter(&zfs_share_lock); 5706 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs", 5707 KRTLD_MODE_FIRST, &error)) == NULL)) { 5708 mutex_exit(&zfs_share_lock); 5709 return (SET_ERROR(ENOSYS)); 5710 } 5711 if (znfsexport_fs == NULL && 5712 ((znfsexport_fs = (int (*)(void *)) 5713 ddi_modsym(nfs_mod, 5714 "nfs_export", &error)) == NULL)) { 5715 mutex_exit(&zfs_share_lock); 5716 return (SET_ERROR(ENOSYS)); 5717 } 5718 error = zfs_init_sharefs(); 5719 if (error != 0) { 5720 mutex_exit(&zfs_share_lock); 5721 return (SET_ERROR(ENOSYS)); 5722 } 5723 zfs_nfsshare_inited = 1; 5724 mutex_exit(&zfs_share_lock); 5725 } 5726 break; 5727 case ZFS_SHARE_SMB: 5728 case ZFS_UNSHARE_SMB: 5729 if (zfs_smbshare_inited == 0) { 5730 mutex_enter(&zfs_share_lock); 5731 if (smbsrv_mod == NULL && ((smbsrv_mod = 5732 ddi_modopen("drv/smbsrv", 5733 KRTLD_MODE_FIRST, &error)) == NULL)) { 5734 mutex_exit(&zfs_share_lock); 5735 return (SET_ERROR(ENOSYS)); 5736 } 5737 if (zsmbexport_fs == NULL && ((zsmbexport_fs = 5738 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod, 5739 "smb_server_share", &error)) == NULL)) { 5740 mutex_exit(&zfs_share_lock); 5741 return (SET_ERROR(ENOSYS)); 5742 } 5743 error = zfs_init_sharefs(); 5744 if (error != 0) { 5745 mutex_exit(&zfs_share_lock); 5746 return (SET_ERROR(ENOSYS)); 5747 } 5748 zfs_smbshare_inited = 1; 5749 mutex_exit(&zfs_share_lock); 5750 } 5751 break; 5752 default: 5753 return (SET_ERROR(EINVAL)); 5754 } 5755 5756 switch (zc->zc_share.z_sharetype) { 5757 case ZFS_SHARE_NFS: 5758 case ZFS_UNSHARE_NFS: 5759 if (error = 5760 znfsexport_fs((void *) 5761 (uintptr_t)zc->zc_share.z_exportdata)) 5762 return (error); 5763 break; 5764 case ZFS_SHARE_SMB: 5765 case ZFS_UNSHARE_SMB: 5766 if (error = zsmbexport_fs((void *) 5767 (uintptr_t)zc->zc_share.z_exportdata, 5768 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ? 5769 B_TRUE: B_FALSE)) { 5770 return (error); 5771 } 5772 break; 5773 } 5774 5775 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS || 5776 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ? 5777 SHAREFS_ADD : SHAREFS_REMOVE; 5778 5779 /* 5780 * Add or remove share from sharetab 5781 */ 5782 error = zshare_fs(opcode, 5783 (void *)(uintptr_t)zc->zc_share.z_sharedata, 5784 zc->zc_share.z_sharemax); 5785 5786 return (error); 5787 5788 } 5789 5790 ace_t full_access[] = { 5791 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0} 5792 }; 5793 5794 /* 5795 * inputs: 5796 * zc_name name of containing filesystem 5797 * zc_obj object # beyond which we want next in-use object # 5798 * 5799 * outputs: 5800 * zc_obj next in-use object # 5801 */ 5802 static int 5803 zfs_ioc_next_obj(zfs_cmd_t *zc) 5804 { 5805 objset_t *os = NULL; 5806 int error; 5807 5808 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 5809 if (error != 0) 5810 return (error); 5811 5812 error = dmu_object_next(os, &zc->zc_obj, B_FALSE, 5813 dsl_dataset_phys(os->os_dsl_dataset)->ds_prev_snap_txg); 5814 5815 dmu_objset_rele(os, FTAG); 5816 return (error); 5817 } 5818 5819 /* 5820 * inputs: 5821 * zc_name name of filesystem 5822 * zc_value prefix name for snapshot 5823 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process 5824 * 5825 * outputs: 5826 * zc_value short name of new snapshot 5827 */ 5828 static int 5829 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc) 5830 { 5831 char *snap_name; 5832 char *hold_name; 5833 int error; 5834 minor_t minor; 5835 5836 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor); 5837 if (error != 0) 5838 return (error); 5839 5840 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value, 5841 (u_longlong_t)ddi_get_lbolt64()); 5842 hold_name = kmem_asprintf("%%%s", zc->zc_value); 5843 5844 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor, 5845 hold_name); 5846 if (error == 0) 5847 (void) strcpy(zc->zc_value, snap_name); 5848 strfree(snap_name); 5849 strfree(hold_name); 5850 zfs_onexit_fd_rele(zc->zc_cleanup_fd); 5851 return (error); 5852 } 5853 5854 /* 5855 * inputs: 5856 * zc_name name of "to" snapshot 5857 * zc_value name of "from" snapshot 5858 * zc_cookie file descriptor to write diff data on 5859 * 5860 * outputs: 5861 * dmu_diff_record_t's to the file descriptor 5862 */ 5863 static int 5864 zfs_ioc_diff(zfs_cmd_t *zc) 5865 { 5866 file_t *fp; 5867 offset_t off; 5868 int error; 5869 5870 fp = getf(zc->zc_cookie); 5871 if (fp == NULL) 5872 return (SET_ERROR(EBADF)); 5873 5874 off = fp->f_offset; 5875 5876 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off); 5877 5878 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0) 5879 fp->f_offset = off; 5880 releasef(zc->zc_cookie); 5881 5882 return (error); 5883 } 5884 5885 /* 5886 * Remove all ACL files in shares dir 5887 */ 5888 static int 5889 zfs_smb_acl_purge(znode_t *dzp) 5890 { 5891 zap_cursor_t zc; 5892 zap_attribute_t zap; 5893 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 5894 int error; 5895 5896 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id); 5897 (error = zap_cursor_retrieve(&zc, &zap)) == 0; 5898 zap_cursor_advance(&zc)) { 5899 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred, 5900 NULL, 0)) != 0) 5901 break; 5902 } 5903 zap_cursor_fini(&zc); 5904 return (error); 5905 } 5906 5907 static int 5908 zfs_ioc_smb_acl(zfs_cmd_t *zc) 5909 { 5910 vnode_t *vp; 5911 znode_t *dzp; 5912 vnode_t *resourcevp = NULL; 5913 znode_t *sharedir; 5914 zfsvfs_t *zfsvfs; 5915 nvlist_t *nvlist; 5916 char *src, *target; 5917 vattr_t vattr; 5918 vsecattr_t vsec; 5919 int error = 0; 5920 5921 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, 5922 NO_FOLLOW, NULL, &vp)) != 0) 5923 return (error); 5924 5925 /* Now make sure mntpnt and dataset are ZFS */ 5926 5927 if (vp->v_vfsp->vfs_fstype != zfsfstype || 5928 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), 5929 zc->zc_name) != 0)) { 5930 VN_RELE(vp); 5931 return (SET_ERROR(EINVAL)); 5932 } 5933 5934 dzp = VTOZ(vp); 5935 zfsvfs = dzp->z_zfsvfs; 5936 ZFS_ENTER(zfsvfs); 5937 5938 /* 5939 * Create share dir if its missing. 5940 */ 5941 mutex_enter(&zfsvfs->z_lock); 5942 if (zfsvfs->z_shares_dir == 0) { 5943 dmu_tx_t *tx; 5944 5945 tx = dmu_tx_create(zfsvfs->z_os); 5946 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE, 5947 ZFS_SHARES_DIR); 5948 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); 5949 error = dmu_tx_assign(tx, TXG_WAIT); 5950 if (error != 0) { 5951 dmu_tx_abort(tx); 5952 } else { 5953 error = zfs_create_share_dir(zfsvfs, tx); 5954 dmu_tx_commit(tx); 5955 } 5956 if (error != 0) { 5957 mutex_exit(&zfsvfs->z_lock); 5958 VN_RELE(vp); 5959 ZFS_EXIT(zfsvfs); 5960 return (error); 5961 } 5962 } 5963 mutex_exit(&zfsvfs->z_lock); 5964 5965 ASSERT(zfsvfs->z_shares_dir); 5966 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) { 5967 VN_RELE(vp); 5968 ZFS_EXIT(zfsvfs); 5969 return (error); 5970 } 5971 5972 switch (zc->zc_cookie) { 5973 case ZFS_SMB_ACL_ADD: 5974 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; 5975 vattr.va_type = VREG; 5976 vattr.va_mode = S_IFREG|0777; 5977 vattr.va_uid = 0; 5978 vattr.va_gid = 0; 5979 5980 vsec.vsa_mask = VSA_ACE; 5981 vsec.vsa_aclentp = &full_access; 5982 vsec.vsa_aclentsz = sizeof (full_access); 5983 vsec.vsa_aclcnt = 1; 5984 5985 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string, 5986 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec); 5987 if (resourcevp) 5988 VN_RELE(resourcevp); 5989 break; 5990 5991 case ZFS_SMB_ACL_REMOVE: 5992 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred, 5993 NULL, 0); 5994 break; 5995 5996 case ZFS_SMB_ACL_RENAME: 5997 if ((error = get_nvlist(zc->zc_nvlist_src, 5998 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) { 5999 VN_RELE(vp); 6000 VN_RELE(ZTOV(sharedir)); 6001 ZFS_EXIT(zfsvfs); 6002 return (error); 6003 } 6004 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) || 6005 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET, 6006 &target)) { 6007 VN_RELE(vp); 6008 VN_RELE(ZTOV(sharedir)); 6009 ZFS_EXIT(zfsvfs); 6010 nvlist_free(nvlist); 6011 return (error); 6012 } 6013 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target, 6014 kcred, NULL, 0); 6015 nvlist_free(nvlist); 6016 break; 6017 6018 case ZFS_SMB_ACL_PURGE: 6019 error = zfs_smb_acl_purge(sharedir); 6020 break; 6021 6022 default: 6023 error = SET_ERROR(EINVAL); 6024 break; 6025 } 6026 6027 VN_RELE(vp); 6028 VN_RELE(ZTOV(sharedir)); 6029 6030 ZFS_EXIT(zfsvfs); 6031 6032 return (error); 6033 } 6034 6035 /* 6036 * innvl: { 6037 * "holds" -> { snapname -> holdname (string), ... } 6038 * (optional) "cleanup_fd" -> fd (int32) 6039 * } 6040 * 6041 * outnvl: { 6042 * snapname -> error value (int32) 6043 * ... 6044 * } 6045 */ 6046 static const zfs_ioc_key_t zfs_keys_hold[] = { 6047 {"holds", DATA_TYPE_NVLIST, 0}, 6048 {"cleanup_fd", DATA_TYPE_INT32, ZK_OPTIONAL}, 6049 }; 6050 6051 /* ARGSUSED */ 6052 static int 6053 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist) 6054 { 6055 nvpair_t *pair; 6056 nvlist_t *holds; 6057 int cleanup_fd = -1; 6058 int error; 6059 minor_t minor = 0; 6060 6061 holds = fnvlist_lookup_nvlist(args, "holds"); 6062 6063 /* make sure the user didn't pass us any invalid (empty) tags */ 6064 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL; 6065 pair = nvlist_next_nvpair(holds, pair)) { 6066 char *htag; 6067 6068 error = nvpair_value_string(pair, &htag); 6069 if (error != 0) 6070 return (SET_ERROR(error)); 6071 6072 if (strlen(htag) == 0) 6073 return (SET_ERROR(EINVAL)); 6074 } 6075 6076 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) { 6077 error = zfs_onexit_fd_hold(cleanup_fd, &minor); 6078 if (error != 0) 6079 return (error); 6080 } 6081 6082 error = dsl_dataset_user_hold(holds, minor, errlist); 6083 if (minor != 0) 6084 zfs_onexit_fd_rele(cleanup_fd); 6085 return (error); 6086 } 6087 6088 /* 6089 * innvl is not used. 6090 * 6091 * outnvl: { 6092 * holdname -> time added (uint64 seconds since epoch) 6093 * ... 6094 * } 6095 */ 6096 static const zfs_ioc_key_t zfs_keys_get_holds[] = { 6097 /* no nvl keys */ 6098 }; 6099 6100 /* ARGSUSED */ 6101 static int 6102 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl) 6103 { 6104 return (dsl_dataset_get_holds(snapname, outnvl)); 6105 } 6106 6107 /* 6108 * innvl: { 6109 * snapname -> { holdname, ... } 6110 * ... 6111 * } 6112 * 6113 * outnvl: { 6114 * snapname -> error value (int32) 6115 * ... 6116 * } 6117 */ 6118 static const zfs_ioc_key_t zfs_keys_release[] = { 6119 {"<snapname>...", DATA_TYPE_NVLIST, ZK_WILDCARDLIST}, 6120 }; 6121 6122 /* ARGSUSED */ 6123 static int 6124 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist) 6125 { 6126 return (dsl_dataset_user_release(holds, errlist)); 6127 } 6128 6129 /* 6130 * inputs: 6131 * zc_name name of new filesystem or snapshot 6132 * zc_value full name of old snapshot 6133 * 6134 * outputs: 6135 * zc_cookie space in bytes 6136 * zc_objset_type compressed space in bytes 6137 * zc_perm_action uncompressed space in bytes 6138 */ 6139 static int 6140 zfs_ioc_space_written(zfs_cmd_t *zc) 6141 { 6142 int error; 6143 dsl_pool_t *dp; 6144 dsl_dataset_t *new, *old; 6145 6146 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 6147 if (error != 0) 6148 return (error); 6149 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new); 6150 if (error != 0) { 6151 dsl_pool_rele(dp, FTAG); 6152 return (error); 6153 } 6154 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old); 6155 if (error != 0) { 6156 dsl_dataset_rele(new, FTAG); 6157 dsl_pool_rele(dp, FTAG); 6158 return (error); 6159 } 6160 6161 error = dsl_dataset_space_written(old, new, &zc->zc_cookie, 6162 &zc->zc_objset_type, &zc->zc_perm_action); 6163 dsl_dataset_rele(old, FTAG); 6164 dsl_dataset_rele(new, FTAG); 6165 dsl_pool_rele(dp, FTAG); 6166 return (error); 6167 } 6168 6169 /* 6170 * innvl: { 6171 * "firstsnap" -> snapshot name 6172 * } 6173 * 6174 * outnvl: { 6175 * "used" -> space in bytes 6176 * "compressed" -> compressed space in bytes 6177 * "uncompressed" -> uncompressed space in bytes 6178 * } 6179 */ 6180 static const zfs_ioc_key_t zfs_keys_space_snaps[] = { 6181 {"firstsnap", DATA_TYPE_STRING, 0}, 6182 }; 6183 6184 static int 6185 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl) 6186 { 6187 int error; 6188 dsl_pool_t *dp; 6189 dsl_dataset_t *new, *old; 6190 char *firstsnap; 6191 uint64_t used, comp, uncomp; 6192 6193 firstsnap = fnvlist_lookup_string(innvl, "firstsnap"); 6194 6195 error = dsl_pool_hold(lastsnap, FTAG, &dp); 6196 if (error != 0) 6197 return (error); 6198 6199 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new); 6200 if (error == 0 && !new->ds_is_snapshot) { 6201 dsl_dataset_rele(new, FTAG); 6202 error = SET_ERROR(EINVAL); 6203 } 6204 if (error != 0) { 6205 dsl_pool_rele(dp, FTAG); 6206 return (error); 6207 } 6208 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old); 6209 if (error == 0 && !old->ds_is_snapshot) { 6210 dsl_dataset_rele(old, FTAG); 6211 error = SET_ERROR(EINVAL); 6212 } 6213 if (error != 0) { 6214 dsl_dataset_rele(new, FTAG); 6215 dsl_pool_rele(dp, FTAG); 6216 return (error); 6217 } 6218 6219 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp); 6220 dsl_dataset_rele(old, FTAG); 6221 dsl_dataset_rele(new, FTAG); 6222 dsl_pool_rele(dp, FTAG); 6223 fnvlist_add_uint64(outnvl, "used", used); 6224 fnvlist_add_uint64(outnvl, "compressed", comp); 6225 fnvlist_add_uint64(outnvl, "uncompressed", uncomp); 6226 return (error); 6227 } 6228 6229 /* 6230 * innvl: { 6231 * "fd" -> file descriptor to write stream to (int32) 6232 * (optional) "fromsnap" -> full snap name to send an incremental from 6233 * (optional) "largeblockok" -> (value ignored) 6234 * indicates that blocks > 128KB are permitted 6235 * (optional) "embedok" -> (value ignored) 6236 * presence indicates DRR_WRITE_EMBEDDED records are permitted 6237 * (optional) "compressok" -> (value ignored) 6238 * presence indicates compressed DRR_WRITE records are permitted 6239 * (optional) "rawok" -> (value ignored) 6240 * presence indicates raw encrypted records should be used. 6241 * (optional) "resume_object" and "resume_offset" -> (uint64) 6242 * if present, resume send stream from specified object and offset. 6243 * } 6244 * 6245 * outnvl is unused 6246 */ 6247 static const zfs_ioc_key_t zfs_keys_send_new[] = { 6248 {"fd", DATA_TYPE_INT32, 0}, 6249 {"fromsnap", DATA_TYPE_STRING, ZK_OPTIONAL}, 6250 {"largeblockok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL}, 6251 {"embedok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL}, 6252 {"compressok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL}, 6253 {"rawok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL}, 6254 {"resume_object", DATA_TYPE_UINT64, ZK_OPTIONAL}, 6255 {"resume_offset", DATA_TYPE_UINT64, ZK_OPTIONAL}, 6256 }; 6257 6258 /* ARGSUSED */ 6259 static int 6260 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl) 6261 { 6262 int error; 6263 offset_t off; 6264 char *fromname = NULL; 6265 int fd; 6266 boolean_t largeblockok; 6267 boolean_t embedok; 6268 boolean_t compressok; 6269 boolean_t rawok; 6270 uint64_t resumeobj = 0; 6271 uint64_t resumeoff = 0; 6272 6273 fd = fnvlist_lookup_int32(innvl, "fd"); 6274 6275 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname); 6276 6277 largeblockok = nvlist_exists(innvl, "largeblockok"); 6278 embedok = nvlist_exists(innvl, "embedok"); 6279 compressok = nvlist_exists(innvl, "compressok"); 6280 rawok = nvlist_exists(innvl, "rawok"); 6281 6282 (void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj); 6283 (void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff); 6284 6285 file_t *fp = getf(fd); 6286 if (fp == NULL) 6287 return (SET_ERROR(EBADF)); 6288 6289 off = fp->f_offset; 6290 error = dmu_send(snapname, fromname, embedok, largeblockok, compressok, 6291 rawok, fd, resumeobj, resumeoff, fp->f_vnode, &off); 6292 6293 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0) 6294 fp->f_offset = off; 6295 releasef(fd); 6296 return (error); 6297 } 6298 6299 /* 6300 * Determine approximately how large a zfs send stream will be -- the number 6301 * of bytes that will be written to the fd supplied to zfs_ioc_send_new(). 6302 * 6303 * innvl: { 6304 * (optional) "from" -> full snap or bookmark name to send an incremental 6305 * from 6306 * (optional) "largeblockok" -> (value ignored) 6307 * indicates that blocks > 128KB are permitted 6308 * (optional) "embedok" -> (value ignored) 6309 * presence indicates DRR_WRITE_EMBEDDED records are permitted 6310 * (optional) "compressok" -> (value ignored) 6311 * presence indicates compressed DRR_WRITE records are permitted 6312 * } 6313 * 6314 * outnvl: { 6315 * "space" -> bytes of space (uint64) 6316 * } 6317 */ 6318 static const zfs_ioc_key_t zfs_keys_send_space[] = { 6319 {"from", DATA_TYPE_STRING, ZK_OPTIONAL}, 6320 {"fromsnap", DATA_TYPE_STRING, ZK_OPTIONAL}, 6321 {"largeblockok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL}, 6322 {"embedok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL}, 6323 {"compressok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL}, 6324 {"rawok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL}, 6325 }; 6326 6327 static int 6328 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl) 6329 { 6330 dsl_pool_t *dp; 6331 dsl_dataset_t *tosnap; 6332 int error; 6333 char *fromname; 6334 boolean_t compressok; 6335 boolean_t rawok; 6336 uint64_t space; 6337 6338 error = dsl_pool_hold(snapname, FTAG, &dp); 6339 if (error != 0) 6340 return (error); 6341 6342 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap); 6343 if (error != 0) { 6344 dsl_pool_rele(dp, FTAG); 6345 return (error); 6346 } 6347 6348 compressok = nvlist_exists(innvl, "compressok"); 6349 rawok = nvlist_exists(innvl, "rawok"); 6350 6351 error = nvlist_lookup_string(innvl, "from", &fromname); 6352 if (error == 0) { 6353 if (strchr(fromname, '@') != NULL) { 6354 /* 6355 * If from is a snapshot, hold it and use the more 6356 * efficient dmu_send_estimate to estimate send space 6357 * size using deadlists. 6358 */ 6359 dsl_dataset_t *fromsnap; 6360 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap); 6361 if (error != 0) 6362 goto out; 6363 error = dmu_send_estimate(tosnap, fromsnap, 6364 compressok || rawok, &space); 6365 dsl_dataset_rele(fromsnap, FTAG); 6366 } else if (strchr(fromname, '#') != NULL) { 6367 /* 6368 * If from is a bookmark, fetch the creation TXG of the 6369 * snapshot it was created from and use that to find 6370 * blocks that were born after it. 6371 */ 6372 zfs_bookmark_phys_t frombm; 6373 6374 error = dsl_bookmark_lookup(dp, fromname, tosnap, 6375 &frombm); 6376 if (error != 0) 6377 goto out; 6378 error = dmu_send_estimate_from_txg(tosnap, 6379 frombm.zbm_creation_txg, compressok || rawok, 6380 &space); 6381 } else { 6382 /* 6383 * from is not properly formatted as a snapshot or 6384 * bookmark 6385 */ 6386 error = SET_ERROR(EINVAL); 6387 goto out; 6388 } 6389 } else { 6390 /* 6391 * If estimating the size of a full send, use dmu_send_estimate. 6392 */ 6393 error = dmu_send_estimate(tosnap, NULL, compressok || rawok, 6394 &space); 6395 } 6396 6397 fnvlist_add_uint64(outnvl, "space", space); 6398 6399 out: 6400 dsl_dataset_rele(tosnap, FTAG); 6401 dsl_pool_rele(dp, FTAG); 6402 return (error); 6403 } 6404 6405 /* 6406 * Sync the currently open TXG to disk for the specified pool. 6407 * This is somewhat similar to 'zfs_sync()'. 6408 * For cases that do not result in error this ioctl will wait for 6409 * the currently open TXG to commit before returning back to the caller. 6410 * 6411 * innvl: { 6412 * "force" -> when true, force uberblock update even if there is no dirty data. 6413 * In addition this will cause the vdev configuration to be written 6414 * out including updating the zpool cache file. (boolean_t) 6415 * } 6416 * 6417 * onvl is unused 6418 */ 6419 static const zfs_ioc_key_t zfs_keys_pool_sync[] = { 6420 {"force", DATA_TYPE_BOOLEAN_VALUE, 0}, 6421 }; 6422 6423 /* ARGSUSED */ 6424 static int 6425 zfs_ioc_pool_sync(const char *pool, nvlist_t *innvl, nvlist_t *onvl) 6426 { 6427 int err; 6428 boolean_t force = B_FALSE; 6429 spa_t *spa; 6430 6431 if ((err = spa_open(pool, &spa, FTAG)) != 0) 6432 return (err); 6433 6434 if (innvl) 6435 force = fnvlist_lookup_boolean_value(innvl, "force"); 6436 6437 if (force) { 6438 spa_config_enter(spa, SCL_CONFIG, FTAG, RW_WRITER); 6439 vdev_config_dirty(spa->spa_root_vdev); 6440 spa_config_exit(spa, SCL_CONFIG, FTAG); 6441 } 6442 txg_wait_synced(spa_get_dsl(spa), 0); 6443 6444 spa_close(spa, FTAG); 6445 6446 return (err); 6447 } 6448 6449 /* 6450 * Load a user's wrapping key into the kernel. 6451 * innvl: { 6452 * "hidden_args" -> { "wkeydata" -> value } 6453 * raw uint8_t array of encryption wrapping key data (32 bytes) 6454 * (optional) "noop" -> (value ignored) 6455 * presence indicated key should only be verified, not loaded 6456 * } 6457 */ 6458 static const zfs_ioc_key_t zfs_keys_load_key[] = { 6459 {"hidden_args", DATA_TYPE_NVLIST, 0}, 6460 {"noop", DATA_TYPE_BOOLEAN, ZK_OPTIONAL}, 6461 }; 6462 6463 /* ARGSUSED */ 6464 static int 6465 zfs_ioc_load_key(const char *dsname, nvlist_t *innvl, nvlist_t *outnvl) 6466 { 6467 int ret = 0; 6468 dsl_crypto_params_t *dcp = NULL; 6469 nvlist_t *hidden_args; 6470 boolean_t noop = nvlist_exists(innvl, "noop"); 6471 6472 if (strchr(dsname, '@') != NULL || strchr(dsname, '%') != NULL) { 6473 ret = SET_ERROR(EINVAL); 6474 goto error; 6475 } 6476 6477 hidden_args = fnvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS); 6478 6479 ret = dsl_crypto_params_create_nvlist(DCP_CMD_NONE, NULL, 6480 hidden_args, &dcp); 6481 if (ret != 0) 6482 goto error; 6483 6484 ret = spa_keystore_load_wkey(dsname, dcp, noop); 6485 if (ret != 0) 6486 goto error; 6487 6488 dsl_crypto_params_free(dcp, noop); 6489 6490 return (0); 6491 6492 error: 6493 dsl_crypto_params_free(dcp, B_TRUE); 6494 return (ret); 6495 } 6496 6497 /* 6498 * Unload a user's wrapping key from the kernel. 6499 * Both innvl and outnvl are unused. 6500 */ 6501 static const zfs_ioc_key_t zfs_keys_unload_key[] = { 6502 /* no nvl keys */ 6503 }; 6504 6505 /* ARGSUSED */ 6506 static int 6507 zfs_ioc_unload_key(const char *dsname, nvlist_t *innvl, nvlist_t *outnvl) 6508 { 6509 int ret = 0; 6510 6511 if (strchr(dsname, '@') != NULL || strchr(dsname, '%') != NULL) { 6512 ret = (SET_ERROR(EINVAL)); 6513 goto out; 6514 } 6515 6516 ret = spa_keystore_unload_wkey(dsname); 6517 if (ret != 0) 6518 goto out; 6519 6520 out: 6521 return (ret); 6522 } 6523 6524 /* 6525 * Changes a user's wrapping key used to decrypt a dataset. The keyformat, 6526 * keylocation, pbkdf2salt, and pbkdf2iters properties can also be specified 6527 * here to change how the key is derived in userspace. 6528 * 6529 * innvl: { 6530 * "hidden_args" (optional) -> { "wkeydata" -> value } 6531 * raw uint8_t array of new encryption wrapping key data (32 bytes) 6532 * "props" (optional) -> { prop -> value } 6533 * } 6534 * 6535 * outnvl is unused 6536 */ 6537 static const zfs_ioc_key_t zfs_keys_change_key[] = { 6538 {"crypt_cmd", DATA_TYPE_UINT64, ZK_OPTIONAL}, 6539 {"hidden_args", DATA_TYPE_NVLIST, ZK_OPTIONAL}, 6540 {"props", DATA_TYPE_NVLIST, ZK_OPTIONAL}, 6541 }; 6542 6543 /* ARGSUSED */ 6544 static int 6545 zfs_ioc_change_key(const char *dsname, nvlist_t *innvl, nvlist_t *outnvl) 6546 { 6547 int ret; 6548 uint64_t cmd = DCP_CMD_NONE; 6549 dsl_crypto_params_t *dcp = NULL; 6550 nvlist_t *args = NULL, *hidden_args = NULL; 6551 6552 if (strchr(dsname, '@') != NULL || strchr(dsname, '%') != NULL) { 6553 ret = (SET_ERROR(EINVAL)); 6554 goto error; 6555 } 6556 6557 (void) nvlist_lookup_uint64(innvl, "crypt_cmd", &cmd); 6558 (void) nvlist_lookup_nvlist(innvl, "props", &args); 6559 (void) nvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS, &hidden_args); 6560 6561 ret = dsl_crypto_params_create_nvlist(cmd, args, hidden_args, &dcp); 6562 if (ret != 0) 6563 goto error; 6564 6565 ret = spa_keystore_change_key(dsname, dcp); 6566 if (ret != 0) 6567 goto error; 6568 6569 dsl_crypto_params_free(dcp, B_FALSE); 6570 6571 return (0); 6572 6573 error: 6574 dsl_crypto_params_free(dcp, B_TRUE); 6575 return (ret); 6576 } 6577 6578 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST]; 6579 6580 static void 6581 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 6582 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck, 6583 boolean_t log_history, zfs_ioc_poolcheck_t pool_check) 6584 { 6585 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST]; 6586 6587 ASSERT3U(ioc, >=, ZFS_IOC_FIRST); 6588 ASSERT3U(ioc, <, ZFS_IOC_LAST); 6589 ASSERT3P(vec->zvec_legacy_func, ==, NULL); 6590 ASSERT3P(vec->zvec_func, ==, NULL); 6591 6592 vec->zvec_legacy_func = func; 6593 vec->zvec_secpolicy = secpolicy; 6594 vec->zvec_namecheck = namecheck; 6595 vec->zvec_allow_log = log_history; 6596 vec->zvec_pool_check = pool_check; 6597 } 6598 6599 /* 6600 * See the block comment at the beginning of this file for details on 6601 * each argument to this function. 6602 */ 6603 static void 6604 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func, 6605 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck, 6606 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist, 6607 boolean_t allow_log, const zfs_ioc_key_t *nvl_keys, size_t num_keys) 6608 { 6609 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST]; 6610 6611 ASSERT3U(ioc, >=, ZFS_IOC_FIRST); 6612 ASSERT3U(ioc, <, ZFS_IOC_LAST); 6613 ASSERT3P(vec->zvec_legacy_func, ==, NULL); 6614 ASSERT3P(vec->zvec_func, ==, NULL); 6615 6616 /* if we are logging, the name must be valid */ 6617 ASSERT(!allow_log || namecheck != NO_NAME); 6618 6619 vec->zvec_name = name; 6620 vec->zvec_func = func; 6621 vec->zvec_secpolicy = secpolicy; 6622 vec->zvec_namecheck = namecheck; 6623 vec->zvec_pool_check = pool_check; 6624 vec->zvec_smush_outnvlist = smush_outnvlist; 6625 vec->zvec_allow_log = allow_log; 6626 vec->zvec_nvl_keys = nvl_keys; 6627 vec->zvec_nvl_key_count = num_keys; 6628 } 6629 6630 static void 6631 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 6632 zfs_secpolicy_func_t *secpolicy, boolean_t log_history, 6633 zfs_ioc_poolcheck_t pool_check) 6634 { 6635 zfs_ioctl_register_legacy(ioc, func, secpolicy, 6636 POOL_NAME, log_history, pool_check); 6637 } 6638 6639 static void 6640 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 6641 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check) 6642 { 6643 zfs_ioctl_register_legacy(ioc, func, secpolicy, 6644 DATASET_NAME, B_FALSE, pool_check); 6645 } 6646 6647 static void 6648 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func) 6649 { 6650 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config, 6651 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 6652 } 6653 6654 static void 6655 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 6656 zfs_secpolicy_func_t *secpolicy) 6657 { 6658 zfs_ioctl_register_legacy(ioc, func, secpolicy, 6659 NO_NAME, B_FALSE, POOL_CHECK_NONE); 6660 } 6661 6662 static void 6663 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc, 6664 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy) 6665 { 6666 zfs_ioctl_register_legacy(ioc, func, secpolicy, 6667 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED); 6668 } 6669 6670 static void 6671 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func) 6672 { 6673 zfs_ioctl_register_dataset_read_secpolicy(ioc, func, 6674 zfs_secpolicy_read); 6675 } 6676 6677 static void 6678 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 6679 zfs_secpolicy_func_t *secpolicy) 6680 { 6681 zfs_ioctl_register_legacy(ioc, func, secpolicy, 6682 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 6683 } 6684 6685 static void 6686 zfs_ioctl_init(void) 6687 { 6688 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT, 6689 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME, 6690 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE, 6691 zfs_keys_snapshot, ARRAY_SIZE(zfs_keys_snapshot)); 6692 6693 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY, 6694 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME, 6695 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE, 6696 zfs_keys_log_history, ARRAY_SIZE(zfs_keys_log_history)); 6697 6698 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS, 6699 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME, 6700 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE, 6701 zfs_keys_space_snaps, ARRAY_SIZE(zfs_keys_space_snaps)); 6702 6703 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW, 6704 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME, 6705 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE, 6706 zfs_keys_send_new, ARRAY_SIZE(zfs_keys_send_new)); 6707 6708 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE, 6709 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME, 6710 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE, 6711 zfs_keys_send_space, ARRAY_SIZE(zfs_keys_send_space)); 6712 6713 zfs_ioctl_register("create", ZFS_IOC_CREATE, 6714 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME, 6715 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE, 6716 zfs_keys_create, ARRAY_SIZE(zfs_keys_create)); 6717 6718 zfs_ioctl_register("clone", ZFS_IOC_CLONE, 6719 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME, 6720 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE, 6721 zfs_keys_clone, ARRAY_SIZE(zfs_keys_clone)); 6722 6723 zfs_ioctl_register("remap", ZFS_IOC_REMAP, 6724 zfs_ioc_remap, zfs_secpolicy_remap, DATASET_NAME, 6725 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE, 6726 zfs_keys_remap, ARRAY_SIZE(zfs_keys_remap)); 6727 6728 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS, 6729 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME, 6730 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE, 6731 zfs_keys_destroy_snaps, ARRAY_SIZE(zfs_keys_destroy_snaps)); 6732 6733 zfs_ioctl_register("hold", ZFS_IOC_HOLD, 6734 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME, 6735 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE, 6736 zfs_keys_hold, ARRAY_SIZE(zfs_keys_hold)); 6737 zfs_ioctl_register("release", ZFS_IOC_RELEASE, 6738 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME, 6739 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE, 6740 zfs_keys_release, ARRAY_SIZE(zfs_keys_release)); 6741 6742 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS, 6743 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME, 6744 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE, 6745 zfs_keys_get_holds, ARRAY_SIZE(zfs_keys_get_holds)); 6746 6747 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK, 6748 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, 6749 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE, 6750 zfs_keys_rollback, ARRAY_SIZE(zfs_keys_rollback)); 6751 6752 zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK, 6753 zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME, 6754 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE, 6755 zfs_keys_bookmark, ARRAY_SIZE(zfs_keys_bookmark)); 6756 6757 zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS, 6758 zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME, 6759 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE, 6760 zfs_keys_get_bookmarks, ARRAY_SIZE(zfs_keys_get_bookmarks)); 6761 6762 zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS, 6763 zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks, 6764 POOL_NAME, 6765 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE, 6766 zfs_keys_destroy_bookmarks, 6767 ARRAY_SIZE(zfs_keys_destroy_bookmarks)); 6768 6769 zfs_ioctl_register("load-key", ZFS_IOC_LOAD_KEY, 6770 zfs_ioc_load_key, zfs_secpolicy_load_key, 6771 DATASET_NAME, POOL_CHECK_SUSPENDED, B_TRUE, B_TRUE, 6772 zfs_keys_load_key, ARRAY_SIZE(zfs_keys_load_key)); 6773 zfs_ioctl_register("unload-key", ZFS_IOC_UNLOAD_KEY, 6774 zfs_ioc_unload_key, zfs_secpolicy_load_key, 6775 DATASET_NAME, POOL_CHECK_SUSPENDED, B_TRUE, B_TRUE, 6776 zfs_keys_unload_key, ARRAY_SIZE(zfs_keys_unload_key)); 6777 zfs_ioctl_register("change-key", ZFS_IOC_CHANGE_KEY, 6778 zfs_ioc_change_key, zfs_secpolicy_change_key, 6779 DATASET_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, 6780 B_TRUE, B_TRUE, zfs_keys_change_key, 6781 ARRAY_SIZE(zfs_keys_change_key)); 6782 6783 zfs_ioctl_register("sync", ZFS_IOC_POOL_SYNC, 6784 zfs_ioc_pool_sync, zfs_secpolicy_none, POOL_NAME, 6785 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE, 6786 zfs_keys_pool_sync, ARRAY_SIZE(zfs_keys_pool_sync)); 6787 6788 zfs_ioctl_register("channel_program", ZFS_IOC_CHANNEL_PROGRAM, 6789 zfs_ioc_channel_program, zfs_secpolicy_config, 6790 POOL_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, 6791 B_TRUE, zfs_keys_channel_program, 6792 ARRAY_SIZE(zfs_keys_channel_program)); 6793 6794 zfs_ioctl_register("zpool_checkpoint", ZFS_IOC_POOL_CHECKPOINT, 6795 zfs_ioc_pool_checkpoint, zfs_secpolicy_config, POOL_NAME, 6796 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE, 6797 zfs_keys_pool_checkpoint, ARRAY_SIZE(zfs_keys_pool_checkpoint)); 6798 6799 zfs_ioctl_register("zpool_discard_checkpoint", 6800 ZFS_IOC_POOL_DISCARD_CHECKPOINT, zfs_ioc_pool_discard_checkpoint, 6801 zfs_secpolicy_config, POOL_NAME, 6802 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE, 6803 zfs_keys_pool_discard_checkpoint, 6804 ARRAY_SIZE(zfs_keys_pool_discard_checkpoint)); 6805 6806 zfs_ioctl_register("initialize", ZFS_IOC_POOL_INITIALIZE, 6807 zfs_ioc_pool_initialize, zfs_secpolicy_config, POOL_NAME, 6808 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE, 6809 zfs_keys_pool_initialize, ARRAY_SIZE(zfs_keys_pool_initialize)); 6810 6811 zfs_ioctl_register("trim", ZFS_IOC_POOL_TRIM, 6812 zfs_ioc_pool_trim, zfs_secpolicy_config, POOL_NAME, 6813 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE, 6814 zfs_keys_pool_trim, ARRAY_SIZE(zfs_keys_pool_trim)); 6815 6816 zfs_ioctl_register("set_bootenv", ZFS_IOC_SET_BOOTENV, 6817 zfs_ioc_set_bootenv, zfs_secpolicy_config, POOL_NAME, 6818 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE, 6819 zfs_keys_set_bootenv, ARRAY_SIZE(zfs_keys_set_bootenv)); 6820 6821 zfs_ioctl_register("get_bootenv", ZFS_IOC_GET_BOOTENV, 6822 zfs_ioc_get_bootenv, zfs_secpolicy_none, POOL_NAME, 6823 POOL_CHECK_SUSPENDED, B_FALSE, B_TRUE, 6824 zfs_keys_get_bootenv, ARRAY_SIZE(zfs_keys_get_bootenv)); 6825 6826 /* IOCTLS that use the legacy function signature */ 6827 6828 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze, 6829 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY); 6830 6831 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create, 6832 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 6833 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN, 6834 zfs_ioc_pool_scan); 6835 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE, 6836 zfs_ioc_pool_upgrade); 6837 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD, 6838 zfs_ioc_vdev_add); 6839 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE, 6840 zfs_ioc_vdev_remove); 6841 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE, 6842 zfs_ioc_vdev_set_state); 6843 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH, 6844 zfs_ioc_vdev_attach); 6845 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH, 6846 zfs_ioc_vdev_detach); 6847 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH, 6848 zfs_ioc_vdev_setpath); 6849 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU, 6850 zfs_ioc_vdev_setfru); 6851 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS, 6852 zfs_ioc_pool_set_props); 6853 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT, 6854 zfs_ioc_vdev_split); 6855 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID, 6856 zfs_ioc_pool_reguid); 6857 6858 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS, 6859 zfs_ioc_pool_configs, zfs_secpolicy_none); 6860 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT, 6861 zfs_ioc_pool_tryimport, zfs_secpolicy_config); 6862 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT, 6863 zfs_ioc_inject_fault, zfs_secpolicy_inject); 6864 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT, 6865 zfs_ioc_clear_fault, zfs_secpolicy_inject); 6866 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT, 6867 zfs_ioc_inject_list_next, zfs_secpolicy_inject); 6868 6869 /* 6870 * pool destroy, and export don't log the history as part of 6871 * zfsdev_ioctl, but rather zfs_ioc_pool_export 6872 * does the logging of those commands. 6873 */ 6874 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy, 6875 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE); 6876 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export, 6877 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE); 6878 6879 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats, 6880 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE); 6881 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props, 6882 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE); 6883 6884 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log, 6885 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED); 6886 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME, 6887 zfs_ioc_dsobj_to_dsname, 6888 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED); 6889 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY, 6890 zfs_ioc_pool_get_history, 6891 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED); 6892 6893 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import, 6894 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 6895 6896 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear, 6897 zfs_secpolicy_config, B_TRUE, POOL_CHECK_READONLY); 6898 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen, 6899 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED); 6900 6901 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN, 6902 zfs_ioc_space_written); 6903 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS, 6904 zfs_ioc_objset_recvd_props); 6905 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ, 6906 zfs_ioc_next_obj); 6907 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL, 6908 zfs_ioc_get_fsacl); 6909 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS, 6910 zfs_ioc_objset_stats); 6911 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS, 6912 zfs_ioc_objset_zplprops); 6913 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT, 6914 zfs_ioc_dataset_list_next); 6915 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT, 6916 zfs_ioc_snapshot_list_next); 6917 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS, 6918 zfs_ioc_send_progress); 6919 6920 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF, 6921 zfs_ioc_diff, zfs_secpolicy_diff); 6922 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS, 6923 zfs_ioc_obj_to_stats, zfs_secpolicy_diff); 6924 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH, 6925 zfs_ioc_obj_to_path, zfs_secpolicy_diff); 6926 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE, 6927 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one); 6928 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY, 6929 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many); 6930 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND, 6931 zfs_ioc_send, zfs_secpolicy_send); 6932 6933 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop, 6934 zfs_secpolicy_none); 6935 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy, 6936 zfs_secpolicy_destroy); 6937 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename, 6938 zfs_secpolicy_rename); 6939 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv, 6940 zfs_secpolicy_recv); 6941 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote, 6942 zfs_secpolicy_promote); 6943 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP, 6944 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop); 6945 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl, 6946 zfs_secpolicy_set_fsacl); 6947 6948 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share, 6949 zfs_secpolicy_share, POOL_CHECK_NONE); 6950 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl, 6951 zfs_secpolicy_smb_acl, POOL_CHECK_NONE); 6952 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE, 6953 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade, 6954 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 6955 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT, 6956 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot, 6957 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 6958 } 6959 6960 /* 6961 * Verify that for non-legacy ioctls the input nvlist 6962 * pairs match against the expected input. 6963 * 6964 * Possible errors are: 6965 * ZFS_ERR_IOC_ARG_UNAVAIL An unrecognized nvpair was encountered 6966 * ZFS_ERR_IOC_ARG_REQUIRED A required nvpair is missing 6967 * ZFS_ERR_IOC_ARG_BADTYPE Invalid type for nvpair 6968 */ 6969 static int 6970 zfs_check_input_nvpairs(nvlist_t *innvl, const zfs_ioc_vec_t *vec) 6971 { 6972 const zfs_ioc_key_t *nvl_keys = vec->zvec_nvl_keys; 6973 boolean_t required_keys_found = B_FALSE; 6974 6975 /* 6976 * examine each input pair 6977 */ 6978 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 6979 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 6980 char *name = nvpair_name(pair); 6981 data_type_t type = nvpair_type(pair); 6982 boolean_t identified = B_FALSE; 6983 6984 /* 6985 * check pair against the documented names and type 6986 */ 6987 for (int k = 0; k < vec->zvec_nvl_key_count; k++) { 6988 /* if not a wild card name, check for an exact match */ 6989 if ((nvl_keys[k].zkey_flags & ZK_WILDCARDLIST) == 0 && 6990 strcmp(nvl_keys[k].zkey_name, name) != 0) 6991 continue; 6992 6993 identified = B_TRUE; 6994 6995 if (nvl_keys[k].zkey_type != DATA_TYPE_ANY && 6996 nvl_keys[k].zkey_type != type) { 6997 return (SET_ERROR(ZFS_ERR_IOC_ARG_BADTYPE)); 6998 } 6999 7000 if (nvl_keys[k].zkey_flags & ZK_OPTIONAL) 7001 continue; 7002 7003 required_keys_found = B_TRUE; 7004 break; 7005 } 7006 7007 /* allow an 'optional' key, everything else is invalid */ 7008 if (!identified && 7009 (strcmp(name, "optional") != 0 || 7010 type != DATA_TYPE_NVLIST)) { 7011 return (SET_ERROR(ZFS_ERR_IOC_ARG_UNAVAIL)); 7012 } 7013 } 7014 7015 /* verify that all required keys were found */ 7016 for (int k = 0; k < vec->zvec_nvl_key_count; k++) { 7017 if (nvl_keys[k].zkey_flags & ZK_OPTIONAL) 7018 continue; 7019 7020 if (nvl_keys[k].zkey_flags & ZK_WILDCARDLIST) { 7021 /* at least one non-optionial key is expected here */ 7022 if (!required_keys_found) 7023 return (SET_ERROR(ZFS_ERR_IOC_ARG_REQUIRED)); 7024 continue; 7025 } 7026 7027 if (!nvlist_exists(innvl, nvl_keys[k].zkey_name)) 7028 return (SET_ERROR(ZFS_ERR_IOC_ARG_REQUIRED)); 7029 } 7030 7031 return (0); 7032 } 7033 7034 int 7035 pool_status_check(const char *name, zfs_ioc_namecheck_t type, 7036 zfs_ioc_poolcheck_t check) 7037 { 7038 spa_t *spa; 7039 int error; 7040 7041 ASSERT(type == POOL_NAME || type == DATASET_NAME); 7042 7043 if (check & POOL_CHECK_NONE) 7044 return (0); 7045 7046 error = spa_open(name, &spa, FTAG); 7047 if (error == 0) { 7048 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa)) 7049 error = SET_ERROR(EAGAIN); 7050 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa)) 7051 error = SET_ERROR(EROFS); 7052 spa_close(spa, FTAG); 7053 } 7054 return (error); 7055 } 7056 7057 /* 7058 * Find a free minor number. 7059 */ 7060 minor_t 7061 zfsdev_minor_alloc(void) 7062 { 7063 static minor_t last_minor; 7064 minor_t m; 7065 7066 ASSERT(MUTEX_HELD(&zfsdev_state_lock)); 7067 7068 for (m = last_minor + 1; m != last_minor; m++) { 7069 if (m > ZFSDEV_MAX_MINOR) 7070 m = 1; 7071 if (ddi_get_soft_state(zfsdev_state, m) == NULL) { 7072 last_minor = m; 7073 return (m); 7074 } 7075 } 7076 7077 return (0); 7078 } 7079 7080 static int 7081 zfs_ctldev_init(dev_t *devp) 7082 { 7083 minor_t minor; 7084 zfs_soft_state_t *zs; 7085 7086 ASSERT(MUTEX_HELD(&zfsdev_state_lock)); 7087 ASSERT(getminor(*devp) == 0); 7088 7089 minor = zfsdev_minor_alloc(); 7090 if (minor == 0) 7091 return (SET_ERROR(ENXIO)); 7092 7093 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS) 7094 return (SET_ERROR(EAGAIN)); 7095 7096 *devp = makedevice(getemajor(*devp), minor); 7097 7098 zs = ddi_get_soft_state(zfsdev_state, minor); 7099 zs->zss_type = ZSST_CTLDEV; 7100 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data); 7101 7102 return (0); 7103 } 7104 7105 static void 7106 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor) 7107 { 7108 ASSERT(MUTEX_HELD(&zfsdev_state_lock)); 7109 7110 zfs_onexit_destroy(zo); 7111 ddi_soft_state_free(zfsdev_state, minor); 7112 } 7113 7114 void * 7115 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which) 7116 { 7117 zfs_soft_state_t *zp; 7118 7119 zp = ddi_get_soft_state(zfsdev_state, minor); 7120 if (zp == NULL || zp->zss_type != which) 7121 return (NULL); 7122 7123 return (zp->zss_data); 7124 } 7125 7126 static int 7127 zfsdev_open(dev_t *devp, int flag, int otyp, cred_t *cr) 7128 { 7129 int error = 0; 7130 7131 if (getminor(*devp) != 0) 7132 return (zvol_open(devp, flag, otyp, cr)); 7133 7134 /* This is the control device. Allocate a new minor if requested. */ 7135 if (flag & FEXCL) { 7136 mutex_enter(&zfsdev_state_lock); 7137 error = zfs_ctldev_init(devp); 7138 mutex_exit(&zfsdev_state_lock); 7139 } 7140 7141 return (error); 7142 } 7143 7144 static int 7145 zfsdev_close(dev_t dev, int flag, int otyp, cred_t *cr) 7146 { 7147 zfs_onexit_t *zo; 7148 minor_t minor = getminor(dev); 7149 7150 if (minor == 0) 7151 return (0); 7152 7153 mutex_enter(&zfsdev_state_lock); 7154 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV); 7155 if (zo == NULL) { 7156 mutex_exit(&zfsdev_state_lock); 7157 return (zvol_close(dev, flag, otyp, cr)); 7158 } 7159 zfs_ctldev_destroy(zo, minor); 7160 mutex_exit(&zfsdev_state_lock); 7161 7162 return (0); 7163 } 7164 7165 static int 7166 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp) 7167 { 7168 zfs_cmd_t *zc; 7169 uint_t vecnum; 7170 int error, rc, len; 7171 minor_t minor = getminor(dev); 7172 const zfs_ioc_vec_t *vec; 7173 char *saved_poolname = NULL; 7174 nvlist_t *innvl = NULL; 7175 7176 if (minor != 0 && 7177 zfsdev_get_soft_state(minor, ZSST_CTLDEV) == NULL) 7178 return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp)); 7179 7180 vecnum = cmd - ZFS_IOC_FIRST; 7181 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip)); 7182 7183 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0])) 7184 return (SET_ERROR(ZFS_ERR_IOC_CMD_UNAVAIL)); 7185 vec = &zfs_ioc_vec[vecnum]; 7186 7187 /* 7188 * The registered ioctl list may be sparse, verify that either 7189 * a normal or legacy handler are registered. 7190 */ 7191 if (vec->zvec_func == NULL && vec->zvec_legacy_func == NULL) 7192 return (SET_ERROR(ZFS_ERR_IOC_CMD_UNAVAIL)); 7193 7194 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP); 7195 7196 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag); 7197 if (error != 0) { 7198 error = SET_ERROR(EFAULT); 7199 goto out; 7200 } 7201 7202 zc->zc_iflags = flag & FKIOCTL; 7203 if (zc->zc_nvlist_src_size != 0) { 7204 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 7205 zc->zc_iflags, &innvl); 7206 if (error != 0) 7207 goto out; 7208 } 7209 7210 /* 7211 * Ensure that all pool/dataset names are valid before we pass down to 7212 * the lower layers. 7213 */ 7214 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0'; 7215 switch (vec->zvec_namecheck) { 7216 case POOL_NAME: 7217 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0) 7218 error = SET_ERROR(EINVAL); 7219 else 7220 error = pool_status_check(zc->zc_name, 7221 vec->zvec_namecheck, vec->zvec_pool_check); 7222 break; 7223 7224 case DATASET_NAME: 7225 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0) 7226 error = SET_ERROR(EINVAL); 7227 else 7228 error = pool_status_check(zc->zc_name, 7229 vec->zvec_namecheck, vec->zvec_pool_check); 7230 break; 7231 7232 case NO_NAME: 7233 break; 7234 } 7235 7236 /* 7237 * Ensure that all input pairs are valid before we pass them down 7238 * to the lower layers. 7239 * 7240 * The vectored functions can use fnvlist_lookup_{type} for any 7241 * required pairs since zfs_check_input_nvpairs() confirmed that 7242 * they exist and are of the correct type. 7243 */ 7244 if (error == 0 && vec->zvec_func != NULL) { 7245 error = zfs_check_input_nvpairs(innvl, vec); 7246 if (error != 0) 7247 goto out; 7248 } 7249 7250 if (error == 0) 7251 error = vec->zvec_secpolicy(zc, innvl, cr); 7252 7253 if (error != 0) 7254 goto out; 7255 7256 /* legacy ioctls can modify zc_name */ 7257 len = strcspn(zc->zc_name, "/@#") + 1; 7258 saved_poolname = kmem_alloc(len, KM_SLEEP); 7259 (void) strlcpy(saved_poolname, zc->zc_name, len); 7260 7261 if (vec->zvec_func != NULL) { 7262 nvlist_t *outnvl; 7263 int puterror = 0; 7264 spa_t *spa; 7265 nvlist_t *lognv = NULL; 7266 7267 ASSERT(vec->zvec_legacy_func == NULL); 7268 7269 /* 7270 * Add the innvl to the lognv before calling the func, 7271 * in case the func changes the innvl. 7272 */ 7273 if (vec->zvec_allow_log) { 7274 lognv = fnvlist_alloc(); 7275 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL, 7276 vec->zvec_name); 7277 if (!nvlist_empty(innvl)) { 7278 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL, 7279 innvl); 7280 } 7281 } 7282 7283 outnvl = fnvlist_alloc(); 7284 error = vec->zvec_func(zc->zc_name, innvl, outnvl); 7285 7286 /* 7287 * Some commands can partially execute, modify state, and still 7288 * return an error. In these cases, attempt to record what 7289 * was modified. 7290 */ 7291 if ((error == 0 || 7292 (cmd == ZFS_IOC_CHANNEL_PROGRAM && error != EINVAL)) && 7293 vec->zvec_allow_log && 7294 spa_open(zc->zc_name, &spa, FTAG) == 0) { 7295 if (!nvlist_empty(outnvl)) { 7296 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL, 7297 outnvl); 7298 } 7299 if (error != 0) { 7300 fnvlist_add_int64(lognv, ZPOOL_HIST_ERRNO, 7301 error); 7302 } 7303 (void) spa_history_log_nvl(spa, lognv); 7304 spa_close(spa, FTAG); 7305 } 7306 fnvlist_free(lognv); 7307 7308 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) { 7309 int smusherror = 0; 7310 if (vec->zvec_smush_outnvlist) { 7311 smusherror = nvlist_smush(outnvl, 7312 zc->zc_nvlist_dst_size); 7313 } 7314 if (smusherror == 0) 7315 puterror = put_nvlist(zc, outnvl); 7316 } 7317 7318 if (puterror != 0) 7319 error = puterror; 7320 7321 nvlist_free(outnvl); 7322 } else { 7323 error = vec->zvec_legacy_func(zc); 7324 } 7325 7326 out: 7327 nvlist_free(innvl); 7328 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag); 7329 if (error == 0 && rc != 0) 7330 error = SET_ERROR(EFAULT); 7331 if (error == 0 && vec->zvec_allow_log) { 7332 char *s = tsd_get(zfs_allow_log_key); 7333 if (s != NULL) 7334 strfree(s); 7335 (void) tsd_set(zfs_allow_log_key, saved_poolname); 7336 } else { 7337 if (saved_poolname != NULL) 7338 strfree(saved_poolname); 7339 } 7340 7341 kmem_free(zc, sizeof (zfs_cmd_t)); 7342 return (error); 7343 } 7344 7345 static int 7346 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd) 7347 { 7348 if (cmd != DDI_ATTACH) 7349 return (DDI_FAILURE); 7350 7351 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0, 7352 DDI_PSEUDO, 0) == DDI_FAILURE) 7353 return (DDI_FAILURE); 7354 7355 zfs_dip = dip; 7356 7357 ddi_report_dev(dip); 7358 7359 return (DDI_SUCCESS); 7360 } 7361 7362 static int 7363 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd) 7364 { 7365 if (spa_busy() || zfs_busy() || zvol_busy()) 7366 return (DDI_FAILURE); 7367 7368 if (cmd != DDI_DETACH) 7369 return (DDI_FAILURE); 7370 7371 zfs_dip = NULL; 7372 7373 ddi_prop_remove_all(dip); 7374 ddi_remove_minor_node(dip, NULL); 7375 7376 return (DDI_SUCCESS); 7377 } 7378 7379 /*ARGSUSED*/ 7380 static int 7381 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result) 7382 { 7383 switch (infocmd) { 7384 case DDI_INFO_DEVT2DEVINFO: 7385 *result = zfs_dip; 7386 return (DDI_SUCCESS); 7387 7388 case DDI_INFO_DEVT2INSTANCE: 7389 *result = (void *)0; 7390 return (DDI_SUCCESS); 7391 } 7392 7393 return (DDI_FAILURE); 7394 } 7395 7396 /* 7397 * OK, so this is a little weird. 7398 * 7399 * /dev/zfs is the control node, i.e. minor 0. 7400 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0. 7401 * 7402 * /dev/zfs has basically nothing to do except serve up ioctls, 7403 * so most of the standard driver entry points are in zvol.c. 7404 */ 7405 static struct cb_ops zfs_cb_ops = { 7406 zfsdev_open, /* open */ 7407 zfsdev_close, /* close */ 7408 zvol_strategy, /* strategy */ 7409 nodev, /* print */ 7410 zvol_dump, /* dump */ 7411 zvol_read, /* read */ 7412 zvol_write, /* write */ 7413 zfsdev_ioctl, /* ioctl */ 7414 nodev, /* devmap */ 7415 nodev, /* mmap */ 7416 nodev, /* segmap */ 7417 nochpoll, /* poll */ 7418 ddi_prop_op, /* prop_op */ 7419 NULL, /* streamtab */ 7420 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */ 7421 CB_REV, /* version */ 7422 nodev, /* async read */ 7423 nodev, /* async write */ 7424 }; 7425 7426 static struct dev_ops zfs_dev_ops = { 7427 DEVO_REV, /* version */ 7428 0, /* refcnt */ 7429 zfs_info, /* info */ 7430 nulldev, /* identify */ 7431 nulldev, /* probe */ 7432 zfs_attach, /* attach */ 7433 zfs_detach, /* detach */ 7434 nodev, /* reset */ 7435 &zfs_cb_ops, /* driver operations */ 7436 NULL, /* no bus operations */ 7437 NULL, /* power */ 7438 ddi_quiesce_not_needed, /* quiesce */ 7439 }; 7440 7441 static struct modldrv zfs_modldrv = { 7442 &mod_driverops, 7443 "ZFS storage pool", 7444 &zfs_dev_ops 7445 }; 7446 7447 static struct modlinkage modlinkage = { 7448 MODREV_1, 7449 (void *)&zfs_modlfs, 7450 (void *)&zfs_modldrv, 7451 NULL 7452 }; 7453 7454 static void 7455 zfs_allow_log_destroy(void *arg) 7456 { 7457 char *poolname = arg; 7458 strfree(poolname); 7459 } 7460 7461 int 7462 _init(void) 7463 { 7464 int error; 7465 7466 spa_init(FREAD | FWRITE); 7467 zfs_init(); 7468 zvol_init(); 7469 zfs_ioctl_init(); 7470 7471 if ((error = mod_install(&modlinkage)) != 0) { 7472 zvol_fini(); 7473 zfs_fini(); 7474 spa_fini(); 7475 return (error); 7476 } 7477 7478 tsd_create(&zfs_fsyncer_key, NULL); 7479 tsd_create(&rrw_tsd_key, rrw_tsd_destroy); 7480 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy); 7481 7482 error = ldi_ident_from_mod(&modlinkage, &zfs_li); 7483 ASSERT(error == 0); 7484 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL); 7485 7486 return (0); 7487 } 7488 7489 int 7490 _fini(void) 7491 { 7492 int error; 7493 7494 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled) 7495 return (SET_ERROR(EBUSY)); 7496 7497 if ((error = mod_remove(&modlinkage)) != 0) 7498 return (error); 7499 7500 zvol_fini(); 7501 zfs_fini(); 7502 spa_fini(); 7503 if (zfs_nfsshare_inited) 7504 (void) ddi_modclose(nfs_mod); 7505 if (zfs_smbshare_inited) 7506 (void) ddi_modclose(smbsrv_mod); 7507 if (zfs_nfsshare_inited || zfs_smbshare_inited) 7508 (void) ddi_modclose(sharefs_mod); 7509 7510 tsd_destroy(&zfs_fsyncer_key); 7511 ldi_ident_release(zfs_li); 7512 zfs_li = NULL; 7513 mutex_destroy(&zfs_share_lock); 7514 7515 return (error); 7516 } 7517 7518 int 7519 _info(struct modinfo *modinfop) 7520 { 7521 return (mod_info(&modlinkage, modinfop)); 7522 } 7523