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