zfs_ioctl.c revision 329484
1/* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 22/* 23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 24 * Copyright (c) 2011-2012 Pawel Jakub Dawidek. All rights reserved. 25 * Copyright 2013 Martin Matuska <mm@FreeBSD.org>. All rights reserved. 26 * Copyright 2014 Xin Li <delphij@FreeBSD.org>. All rights reserved. 27 * Copyright 2015, OmniTI Computer Consulting, Inc. All rights reserved. 28 * Copyright 2015 Nexenta Systems, Inc. All rights reserved. 29 * Copyright (c) 2014, 2016 Joyent, Inc. All rights reserved. 30 * Copyright (c) 2011, 2017 by Delphix. All rights reserved. 31 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved. 32 * Copyright (c) 2013 Steven Hartland. All rights reserved. 33 * Copyright (c) 2014 Integros [integros.com] 34 * Copyright 2017 RackTop Systems. 35 * Copyright (c) 2017 Datto Inc. 36 */ 37 38/* 39 * ZFS ioctls. 40 * 41 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage 42 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool. 43 * 44 * There are two ways that we handle ioctls: the legacy way where almost 45 * all of the logic is in the ioctl callback, and the new way where most 46 * of the marshalling is handled in the common entry point, zfsdev_ioctl(). 47 * 48 * Non-legacy ioctls should be registered by calling 49 * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked 50 * from userland by lzc_ioctl(). 51 * 52 * The registration arguments are as follows: 53 * 54 * const char *name 55 * The name of the ioctl. This is used for history logging. If the 56 * ioctl returns successfully (the callback returns 0), and allow_log 57 * is true, then a history log entry will be recorded with the input & 58 * output nvlists. The log entry can be printed with "zpool history -i". 59 * 60 * zfs_ioc_t ioc 61 * The ioctl request number, which userland will pass to ioctl(2). 62 * The ioctl numbers can change from release to release, because 63 * the caller (libzfs) must be matched to the kernel. 64 * 65 * zfs_secpolicy_func_t *secpolicy 66 * This function will be called before the zfs_ioc_func_t, to 67 * determine if this operation is permitted. It should return EPERM 68 * on failure, and 0 on success. Checks include determining if the 69 * dataset is visible in this zone, and if the user has either all 70 * zfs privileges in the zone (SYS_MOUNT), or has been granted permission 71 * to do this operation on this dataset with "zfs allow". 72 * 73 * zfs_ioc_namecheck_t namecheck 74 * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool 75 * name, a dataset name, or nothing. If the name is not well-formed, 76 * the ioctl will fail and the callback will not be called. 77 * Therefore, the callback can assume that the name is well-formed 78 * (e.g. is null-terminated, doesn't have more than one '@' character, 79 * doesn't have invalid characters). 80 * 81 * zfs_ioc_poolcheck_t pool_check 82 * This specifies requirements on the pool state. If the pool does 83 * not meet them (is suspended or is readonly), the ioctl will fail 84 * and the callback will not be called. If any checks are specified 85 * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME. 86 * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED | 87 * POOL_CHECK_READONLY). 88 * 89 * boolean_t smush_outnvlist 90 * If smush_outnvlist is true, then the output is presumed to be a 91 * list of errors, and it will be "smushed" down to fit into the 92 * caller's buffer, by removing some entries and replacing them with a 93 * single "N_MORE_ERRORS" entry indicating how many were removed. See 94 * nvlist_smush() for details. If smush_outnvlist is false, and the 95 * outnvlist does not fit into the userland-provided buffer, then the 96 * ioctl will fail with ENOMEM. 97 * 98 * zfs_ioc_func_t *func 99 * The callback function that will perform the operation. 100 * 101 * The callback should return 0 on success, or an error number on 102 * failure. If the function fails, the userland ioctl will return -1, 103 * and errno will be set to the callback's return value. The callback 104 * will be called with the following arguments: 105 * 106 * const char *name 107 * The name of the pool or dataset to operate on, from 108 * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the 109 * expected type (pool, dataset, or none). 110 * 111 * nvlist_t *innvl 112 * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or 113 * NULL if no input nvlist was provided. Changes to this nvlist are 114 * ignored. If the input nvlist could not be deserialized, the 115 * ioctl will fail and the callback will not be called. 116 * 117 * nvlist_t *outnvl 118 * The output nvlist, initially empty. The callback can fill it in, 119 * and it will be returned to userland by serializing it into 120 * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization 121 * fails (e.g. because the caller didn't supply a large enough 122 * buffer), then the overall ioctl will fail. See the 123 * 'smush_nvlist' argument above for additional behaviors. 124 * 125 * There are two typical uses of the output nvlist: 126 * - To return state, e.g. property values. In this case, 127 * smush_outnvlist should be false. If the buffer was not large 128 * enough, the caller will reallocate a larger buffer and try 129 * the ioctl again. 130 * 131 * - To return multiple errors from an ioctl which makes on-disk 132 * changes. In this case, smush_outnvlist should be true. 133 * Ioctls which make on-disk modifications should generally not 134 * use the outnvl if they succeed, because the caller can not 135 * distinguish between the operation failing, and 136 * deserialization failing. 137 */ 138#ifdef __FreeBSD__ 139#include "opt_kstack_pages.h" 140#endif 141 142#include <sys/types.h> 143#include <sys/param.h> 144#include <sys/systm.h> 145#include <sys/conf.h> 146#include <sys/kernel.h> 147#include <sys/lock.h> 148#include <sys/malloc.h> 149#include <sys/mutex.h> 150#include <sys/proc.h> 151#include <sys/errno.h> 152#include <sys/uio.h> 153#include <sys/buf.h> 154#include <sys/file.h> 155#include <sys/kmem.h> 156#include <sys/conf.h> 157#include <sys/cmn_err.h> 158#include <sys/stat.h> 159#include <sys/zfs_ioctl.h> 160#include <sys/zfs_vfsops.h> 161#include <sys/zfs_znode.h> 162#include <sys/zap.h> 163#include <sys/spa.h> 164#include <sys/spa_impl.h> 165#include <sys/vdev.h> 166#include <sys/dmu.h> 167#include <sys/dsl_dir.h> 168#include <sys/dsl_dataset.h> 169#include <sys/dsl_prop.h> 170#include <sys/dsl_deleg.h> 171#include <sys/dmu_objset.h> 172#include <sys/dmu_impl.h> 173#include <sys/dmu_tx.h> 174#include <sys/sunddi.h> 175#include <sys/policy.h> 176#include <sys/zone.h> 177#include <sys/nvpair.h> 178#include <sys/mount.h> 179#include <sys/taskqueue.h> 180#include <sys/sdt.h> 181#include <sys/varargs.h> 182#include <sys/fs/zfs.h> 183#include <sys/zfs_ctldir.h> 184#include <sys/zfs_dir.h> 185#include <sys/zfs_onexit.h> 186#include <sys/zvol.h> 187#include <sys/dsl_scan.h> 188#include <sys/dmu_objset.h> 189#include <sys/dmu_send.h> 190#include <sys/dsl_destroy.h> 191#include <sys/dsl_bookmark.h> 192#include <sys/dsl_userhold.h> 193#include <sys/zfeature.h> 194#include <sys/zcp.h> 195#include <sys/zio_checksum.h> 196 197#include "zfs_namecheck.h" 198#include "zfs_prop.h" 199#include "zfs_deleg.h" 200#include "zfs_comutil.h" 201#include "zfs_ioctl_compat.h" 202 203#include "lua.h" 204#include "lauxlib.h" 205 206CTASSERT(sizeof(zfs_cmd_t) < IOCPARM_MAX); 207static struct cdev *zfsdev; 208 209extern void zfs_init(void); 210extern void zfs_fini(void); 211 212uint_t zfs_fsyncer_key; 213extern uint_t rrw_tsd_key; 214static uint_t zfs_allow_log_key; 215extern uint_t zfs_geom_probe_vdev_key; 216 217typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *); 218typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *); 219typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *); 220 221typedef enum { 222 NO_NAME, 223 POOL_NAME, 224 DATASET_NAME 225} zfs_ioc_namecheck_t; 226 227typedef enum { 228 POOL_CHECK_NONE = 1 << 0, 229 POOL_CHECK_SUSPENDED = 1 << 1, 230 POOL_CHECK_READONLY = 1 << 2, 231} zfs_ioc_poolcheck_t; 232 233typedef struct zfs_ioc_vec { 234 zfs_ioc_legacy_func_t *zvec_legacy_func; 235 zfs_ioc_func_t *zvec_func; 236 zfs_secpolicy_func_t *zvec_secpolicy; 237 zfs_ioc_namecheck_t zvec_namecheck; 238 boolean_t zvec_allow_log; 239 zfs_ioc_poolcheck_t zvec_pool_check; 240 boolean_t zvec_smush_outnvlist; 241 const char *zvec_name; 242} zfs_ioc_vec_t; 243 244/* This array is indexed by zfs_userquota_prop_t */ 245static const char *userquota_perms[] = { 246 ZFS_DELEG_PERM_USERUSED, 247 ZFS_DELEG_PERM_USERQUOTA, 248 ZFS_DELEG_PERM_GROUPUSED, 249 ZFS_DELEG_PERM_GROUPQUOTA, 250}; 251 252static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc); 253static int zfs_check_settable(const char *name, nvpair_t *property, 254 cred_t *cr); 255static int zfs_check_clearable(char *dataset, nvlist_t *props, 256 nvlist_t **errors); 257static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *, 258 boolean_t *); 259int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *); 260static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp); 261 262static void zfsdev_close(void *data); 263 264static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature); 265 266/* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */ 267void 268__dprintf(const char *file, const char *func, int line, const char *fmt, ...) 269{ 270 const char *newfile; 271 char buf[512]; 272 va_list adx; 273 274 /* 275 * Get rid of annoying "../common/" prefix to filename. 276 */ 277 newfile = strrchr(file, '/'); 278 if (newfile != NULL) { 279 newfile = newfile + 1; /* Get rid of leading / */ 280 } else { 281 newfile = file; 282 } 283 284 va_start(adx, fmt); 285 (void) vsnprintf(buf, sizeof (buf), fmt, adx); 286 va_end(adx); 287 288 /* 289 * To get this data, use the zfs-dprintf probe as so: 290 * dtrace -q -n 'zfs-dprintf \ 291 * /stringof(arg0) == "dbuf.c"/ \ 292 * {printf("%s: %s", stringof(arg1), stringof(arg3))}' 293 * arg0 = file name 294 * arg1 = function name 295 * arg2 = line number 296 * arg3 = message 297 */ 298 DTRACE_PROBE4(zfs__dprintf, 299 char *, newfile, char *, func, int, line, char *, buf); 300} 301 302static void 303history_str_free(char *buf) 304{ 305 kmem_free(buf, HIS_MAX_RECORD_LEN); 306} 307 308static char * 309history_str_get(zfs_cmd_t *zc) 310{ 311 char *buf; 312 313 if (zc->zc_history == 0) 314 return (NULL); 315 316 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP); 317 if (copyinstr((void *)(uintptr_t)zc->zc_history, 318 buf, HIS_MAX_RECORD_LEN, NULL) != 0) { 319 history_str_free(buf); 320 return (NULL); 321 } 322 323 buf[HIS_MAX_RECORD_LEN -1] = '\0'; 324 325 return (buf); 326} 327 328/* 329 * Check to see if the named dataset is currently defined as bootable 330 */ 331static boolean_t 332zfs_is_bootfs(const char *name) 333{ 334 objset_t *os; 335 336 if (dmu_objset_hold(name, FTAG, &os) == 0) { 337 boolean_t ret; 338 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os))); 339 dmu_objset_rele(os, FTAG); 340 return (ret); 341 } 342 return (B_FALSE); 343} 344 345/* 346 * Return non-zero if the spa version is less than requested version. 347 */ 348static int 349zfs_earlier_version(const char *name, int version) 350{ 351 spa_t *spa; 352 353 if (spa_open(name, &spa, FTAG) == 0) { 354 if (spa_version(spa) < version) { 355 spa_close(spa, FTAG); 356 return (1); 357 } 358 spa_close(spa, FTAG); 359 } 360 return (0); 361} 362 363/* 364 * Return TRUE if the ZPL version is less than requested version. 365 */ 366static boolean_t 367zpl_earlier_version(const char *name, int version) 368{ 369 objset_t *os; 370 boolean_t rc = B_TRUE; 371 372 if (dmu_objset_hold(name, FTAG, &os) == 0) { 373 uint64_t zplversion; 374 375 if (dmu_objset_type(os) != DMU_OST_ZFS) { 376 dmu_objset_rele(os, FTAG); 377 return (B_TRUE); 378 } 379 /* XXX reading from non-owned objset */ 380 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0) 381 rc = zplversion < version; 382 dmu_objset_rele(os, FTAG); 383 } 384 return (rc); 385} 386 387static void 388zfs_log_history(zfs_cmd_t *zc) 389{ 390 spa_t *spa; 391 char *buf; 392 393 if ((buf = history_str_get(zc)) == NULL) 394 return; 395 396 if (spa_open(zc->zc_name, &spa, FTAG) == 0) { 397 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY) 398 (void) spa_history_log(spa, buf); 399 spa_close(spa, FTAG); 400 } 401 history_str_free(buf); 402} 403 404/* 405 * Policy for top-level read operations (list pools). Requires no privileges, 406 * and can be used in the local zone, as there is no associated dataset. 407 */ 408/* ARGSUSED */ 409static int 410zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 411{ 412 return (0); 413} 414 415/* 416 * Policy for dataset read operations (list children, get statistics). Requires 417 * no privileges, but must be visible in the local zone. 418 */ 419/* ARGSUSED */ 420static int 421zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 422{ 423 if (INGLOBALZONE(curthread) || 424 zone_dataset_visible(zc->zc_name, NULL)) 425 return (0); 426 427 return (SET_ERROR(ENOENT)); 428} 429 430static int 431zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr) 432{ 433 int writable = 1; 434 435 /* 436 * The dataset must be visible by this zone -- check this first 437 * so they don't see EPERM on something they shouldn't know about. 438 */ 439 if (!INGLOBALZONE(curthread) && 440 !zone_dataset_visible(dataset, &writable)) 441 return (SET_ERROR(ENOENT)); 442 443 if (INGLOBALZONE(curthread)) { 444 /* 445 * If the fs is zoned, only root can access it from the 446 * global zone. 447 */ 448 if (secpolicy_zfs(cr) && zoned) 449 return (SET_ERROR(EPERM)); 450 } else { 451 /* 452 * If we are in a local zone, the 'zoned' property must be set. 453 */ 454 if (!zoned) 455 return (SET_ERROR(EPERM)); 456 457 /* must be writable by this zone */ 458 if (!writable) 459 return (SET_ERROR(EPERM)); 460 } 461 return (0); 462} 463 464static int 465zfs_dozonecheck(const char *dataset, cred_t *cr) 466{ 467 uint64_t zoned; 468 469 if (dsl_prop_get_integer(dataset, "jailed", &zoned, NULL)) 470 return (SET_ERROR(ENOENT)); 471 472 return (zfs_dozonecheck_impl(dataset, zoned, cr)); 473} 474 475static int 476zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr) 477{ 478 uint64_t zoned; 479 480 if (dsl_prop_get_int_ds(ds, "jailed", &zoned)) 481 return (SET_ERROR(ENOENT)); 482 483 return (zfs_dozonecheck_impl(dataset, zoned, cr)); 484} 485 486static int 487zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds, 488 const char *perm, cred_t *cr) 489{ 490 int error; 491 492 error = zfs_dozonecheck_ds(name, ds, cr); 493 if (error == 0) { 494 error = secpolicy_zfs(cr); 495 if (error != 0) 496 error = dsl_deleg_access_impl(ds, perm, cr); 497 } 498 return (error); 499} 500 501static int 502zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr) 503{ 504 int error; 505 dsl_dataset_t *ds; 506 dsl_pool_t *dp; 507 508 /* 509 * First do a quick check for root in the global zone, which 510 * is allowed to do all write_perms. This ensures that zfs_ioc_* 511 * will get to handle nonexistent datasets. 512 */ 513 if (INGLOBALZONE(curthread) && secpolicy_zfs(cr) == 0) 514 return (0); 515 516 error = dsl_pool_hold(name, FTAG, &dp); 517 if (error != 0) 518 return (error); 519 520 error = dsl_dataset_hold(dp, name, FTAG, &ds); 521 if (error != 0) { 522 dsl_pool_rele(dp, FTAG); 523 return (error); 524 } 525 526 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr); 527 528 dsl_dataset_rele(ds, FTAG); 529 dsl_pool_rele(dp, FTAG); 530 return (error); 531} 532 533#ifdef SECLABEL 534/* 535 * Policy for setting the security label property. 536 * 537 * Returns 0 for success, non-zero for access and other errors. 538 */ 539static int 540zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr) 541{ 542 char ds_hexsl[MAXNAMELEN]; 543 bslabel_t ds_sl, new_sl; 544 boolean_t new_default = FALSE; 545 uint64_t zoned; 546 int needed_priv = -1; 547 int error; 548 549 /* First get the existing dataset label. */ 550 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL), 551 1, sizeof (ds_hexsl), &ds_hexsl, NULL); 552 if (error != 0) 553 return (SET_ERROR(EPERM)); 554 555 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0) 556 new_default = TRUE; 557 558 /* The label must be translatable */ 559 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0)) 560 return (SET_ERROR(EINVAL)); 561 562 /* 563 * In a non-global zone, disallow attempts to set a label that 564 * doesn't match that of the zone; otherwise no other checks 565 * are needed. 566 */ 567 if (!INGLOBALZONE(curproc)) { 568 if (new_default || !blequal(&new_sl, CR_SL(CRED()))) 569 return (SET_ERROR(EPERM)); 570 return (0); 571 } 572 573 /* 574 * For global-zone datasets (i.e., those whose zoned property is 575 * "off", verify that the specified new label is valid for the 576 * global zone. 577 */ 578 if (dsl_prop_get_integer(name, 579 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL)) 580 return (SET_ERROR(EPERM)); 581 if (!zoned) { 582 if (zfs_check_global_label(name, strval) != 0) 583 return (SET_ERROR(EPERM)); 584 } 585 586 /* 587 * If the existing dataset label is nondefault, check if the 588 * dataset is mounted (label cannot be changed while mounted). 589 * Get the zfsvfs; if there isn't one, then the dataset isn't 590 * mounted (or isn't a dataset, doesn't exist, ...). 591 */ 592 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) { 593 objset_t *os; 594 static char *setsl_tag = "setsl_tag"; 595 596 /* 597 * Try to own the dataset; abort if there is any error, 598 * (e.g., already mounted, in use, or other error). 599 */ 600 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE, 601 setsl_tag, &os); 602 if (error != 0) 603 return (SET_ERROR(EPERM)); 604 605 dmu_objset_disown(os, setsl_tag); 606 607 if (new_default) { 608 needed_priv = PRIV_FILE_DOWNGRADE_SL; 609 goto out_check; 610 } 611 612 if (hexstr_to_label(strval, &new_sl) != 0) 613 return (SET_ERROR(EPERM)); 614 615 if (blstrictdom(&ds_sl, &new_sl)) 616 needed_priv = PRIV_FILE_DOWNGRADE_SL; 617 else if (blstrictdom(&new_sl, &ds_sl)) 618 needed_priv = PRIV_FILE_UPGRADE_SL; 619 } else { 620 /* dataset currently has a default label */ 621 if (!new_default) 622 needed_priv = PRIV_FILE_UPGRADE_SL; 623 } 624 625out_check: 626 if (needed_priv != -1) 627 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL)); 628 return (0); 629} 630#endif /* SECLABEL */ 631 632static int 633zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval, 634 cred_t *cr) 635{ 636 char *strval; 637 638 /* 639 * Check permissions for special properties. 640 */ 641 switch (prop) { 642 case ZFS_PROP_ZONED: 643 /* 644 * Disallow setting of 'zoned' from within a local zone. 645 */ 646 if (!INGLOBALZONE(curthread)) 647 return (SET_ERROR(EPERM)); 648 break; 649 650 case ZFS_PROP_QUOTA: 651 case ZFS_PROP_FILESYSTEM_LIMIT: 652 case ZFS_PROP_SNAPSHOT_LIMIT: 653 if (!INGLOBALZONE(curthread)) { 654 uint64_t zoned; 655 char setpoint[ZFS_MAX_DATASET_NAME_LEN]; 656 /* 657 * Unprivileged users are allowed to modify the 658 * limit on things *under* (ie. contained by) 659 * the thing they own. 660 */ 661 if (dsl_prop_get_integer(dsname, "jailed", &zoned, 662 setpoint)) 663 return (SET_ERROR(EPERM)); 664 if (!zoned || strlen(dsname) <= strlen(setpoint)) 665 return (SET_ERROR(EPERM)); 666 } 667 break; 668 669 case ZFS_PROP_MLSLABEL: 670#ifdef SECLABEL 671 if (!is_system_labeled()) 672 return (SET_ERROR(EPERM)); 673 674 if (nvpair_value_string(propval, &strval) == 0) { 675 int err; 676 677 err = zfs_set_slabel_policy(dsname, strval, CRED()); 678 if (err != 0) 679 return (err); 680 } 681#else 682 return (EOPNOTSUPP); 683#endif 684 break; 685 } 686 687 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr)); 688} 689 690/* ARGSUSED */ 691static int 692zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 693{ 694 int error; 695 696 error = zfs_dozonecheck(zc->zc_name, cr); 697 if (error != 0) 698 return (error); 699 700 /* 701 * permission to set permissions will be evaluated later in 702 * dsl_deleg_can_allow() 703 */ 704 return (0); 705} 706 707/* ARGSUSED */ 708static int 709zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 710{ 711 return (zfs_secpolicy_write_perms(zc->zc_name, 712 ZFS_DELEG_PERM_ROLLBACK, cr)); 713} 714 715/* ARGSUSED */ 716static int 717zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 718{ 719 dsl_pool_t *dp; 720 dsl_dataset_t *ds; 721 char *cp; 722 int error; 723 724 /* 725 * Generate the current snapshot name from the given objsetid, then 726 * use that name for the secpolicy/zone checks. 727 */ 728 cp = strchr(zc->zc_name, '@'); 729 if (cp == NULL) 730 return (SET_ERROR(EINVAL)); 731 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 732 if (error != 0) 733 return (error); 734 735 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds); 736 if (error != 0) { 737 dsl_pool_rele(dp, FTAG); 738 return (error); 739 } 740 741 dsl_dataset_name(ds, zc->zc_name); 742 743 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds, 744 ZFS_DELEG_PERM_SEND, cr); 745 dsl_dataset_rele(ds, FTAG); 746 dsl_pool_rele(dp, FTAG); 747 748 return (error); 749} 750 751/* ARGSUSED */ 752static int 753zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 754{ 755 return (zfs_secpolicy_write_perms(zc->zc_name, 756 ZFS_DELEG_PERM_SEND, cr)); 757} 758 759/* ARGSUSED */ 760static int 761zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 762{ 763 vnode_t *vp; 764 int error; 765 766 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, 767 NO_FOLLOW, NULL, &vp)) != 0) 768 return (error); 769 770 /* Now make sure mntpnt and dataset are ZFS */ 771 772 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 || 773 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), 774 zc->zc_name) != 0)) { 775 VN_RELE(vp); 776 return (SET_ERROR(EPERM)); 777 } 778 779 VN_RELE(vp); 780 return (dsl_deleg_access(zc->zc_name, 781 ZFS_DELEG_PERM_SHARE, cr)); 782} 783 784int 785zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 786{ 787 if (!INGLOBALZONE(curthread)) 788 return (SET_ERROR(EPERM)); 789 790 if (secpolicy_nfs(cr) == 0) { 791 return (0); 792 } else { 793 return (zfs_secpolicy_deleg_share(zc, innvl, cr)); 794 } 795} 796 797int 798zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 799{ 800 if (!INGLOBALZONE(curthread)) 801 return (SET_ERROR(EPERM)); 802 803 if (secpolicy_smb(cr) == 0) { 804 return (0); 805 } else { 806 return (zfs_secpolicy_deleg_share(zc, innvl, cr)); 807 } 808} 809 810static int 811zfs_get_parent(const char *datasetname, char *parent, int parentsize) 812{ 813 char *cp; 814 815 /* 816 * Remove the @bla or /bla from the end of the name to get the parent. 817 */ 818 (void) strncpy(parent, datasetname, parentsize); 819 cp = strrchr(parent, '@'); 820 if (cp != NULL) { 821 cp[0] = '\0'; 822 } else { 823 cp = strrchr(parent, '/'); 824 if (cp == NULL) 825 return (SET_ERROR(ENOENT)); 826 cp[0] = '\0'; 827 } 828 829 return (0); 830} 831 832int 833zfs_secpolicy_destroy_perms(const char *name, cred_t *cr) 834{ 835 int error; 836 837 if ((error = zfs_secpolicy_write_perms(name, 838 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 839 return (error); 840 841 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr)); 842} 843 844/* ARGSUSED */ 845static int 846zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 847{ 848 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr)); 849} 850 851/* 852 * Destroying snapshots with delegated permissions requires 853 * descendant mount and destroy permissions. 854 */ 855/* ARGSUSED */ 856static int 857zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 858{ 859 nvlist_t *snaps; 860 nvpair_t *pair, *nextpair; 861 int error = 0; 862 863 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 864 return (SET_ERROR(EINVAL)); 865 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 866 pair = nextpair) { 867 nextpair = nvlist_next_nvpair(snaps, pair); 868 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr); 869 if (error == ENOENT) { 870 /* 871 * Ignore any snapshots that don't exist (we consider 872 * them "already destroyed"). Remove the name from the 873 * nvl here in case the snapshot is created between 874 * now and when we try to destroy it (in which case 875 * we don't want to destroy it since we haven't 876 * checked for permission). 877 */ 878 fnvlist_remove_nvpair(snaps, pair); 879 error = 0; 880 } 881 if (error != 0) 882 break; 883 } 884 885 return (error); 886} 887 888int 889zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr) 890{ 891 char parentname[ZFS_MAX_DATASET_NAME_LEN]; 892 int error; 893 894 if ((error = zfs_secpolicy_write_perms(from, 895 ZFS_DELEG_PERM_RENAME, cr)) != 0) 896 return (error); 897 898 if ((error = zfs_secpolicy_write_perms(from, 899 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 900 return (error); 901 902 if ((error = zfs_get_parent(to, parentname, 903 sizeof (parentname))) != 0) 904 return (error); 905 906 if ((error = zfs_secpolicy_write_perms(parentname, 907 ZFS_DELEG_PERM_CREATE, cr)) != 0) 908 return (error); 909 910 if ((error = zfs_secpolicy_write_perms(parentname, 911 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 912 return (error); 913 914 return (error); 915} 916 917/* ARGSUSED */ 918static int 919zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 920{ 921 char *at = NULL; 922 int error; 923 924 if ((zc->zc_cookie & 1) != 0) { 925 /* 926 * This is recursive rename, so the starting snapshot might 927 * not exist. Check file system or volume permission instead. 928 */ 929 at = strchr(zc->zc_name, '@'); 930 if (at == NULL) 931 return (EINVAL); 932 *at = '\0'; 933 } 934 935 error = zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr); 936 937 if (at != NULL) 938 *at = '@'; 939 940 return (error); 941} 942 943/* ARGSUSED */ 944static int 945zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 946{ 947 dsl_pool_t *dp; 948 dsl_dataset_t *clone; 949 int error; 950 951 error = zfs_secpolicy_write_perms(zc->zc_name, 952 ZFS_DELEG_PERM_PROMOTE, cr); 953 if (error != 0) 954 return (error); 955 956 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 957 if (error != 0) 958 return (error); 959 960 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone); 961 962 if (error == 0) { 963 char parentname[ZFS_MAX_DATASET_NAME_LEN]; 964 dsl_dataset_t *origin = NULL; 965 dsl_dir_t *dd; 966 dd = clone->ds_dir; 967 968 error = dsl_dataset_hold_obj(dd->dd_pool, 969 dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin); 970 if (error != 0) { 971 dsl_dataset_rele(clone, FTAG); 972 dsl_pool_rele(dp, FTAG); 973 return (error); 974 } 975 976 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone, 977 ZFS_DELEG_PERM_MOUNT, cr); 978 979 dsl_dataset_name(origin, parentname); 980 if (error == 0) { 981 error = zfs_secpolicy_write_perms_ds(parentname, origin, 982 ZFS_DELEG_PERM_PROMOTE, cr); 983 } 984 dsl_dataset_rele(clone, FTAG); 985 dsl_dataset_rele(origin, FTAG); 986 } 987 dsl_pool_rele(dp, FTAG); 988 return (error); 989} 990 991/* ARGSUSED */ 992static int 993zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 994{ 995 int error; 996 997 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 998 ZFS_DELEG_PERM_RECEIVE, cr)) != 0) 999 return (error); 1000 1001 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 1002 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 1003 return (error); 1004 1005 return (zfs_secpolicy_write_perms(zc->zc_name, 1006 ZFS_DELEG_PERM_CREATE, cr)); 1007} 1008 1009int 1010zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr) 1011{ 1012 return (zfs_secpolicy_write_perms(name, 1013 ZFS_DELEG_PERM_SNAPSHOT, cr)); 1014} 1015 1016/* 1017 * Check for permission to create each snapshot in the nvlist. 1018 */ 1019/* ARGSUSED */ 1020static int 1021zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1022{ 1023 nvlist_t *snaps; 1024 int error; 1025 nvpair_t *pair; 1026 1027 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 1028 return (SET_ERROR(EINVAL)); 1029 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 1030 pair = nvlist_next_nvpair(snaps, pair)) { 1031 char *name = nvpair_name(pair); 1032 char *atp = strchr(name, '@'); 1033 1034 if (atp == NULL) { 1035 error = SET_ERROR(EINVAL); 1036 break; 1037 } 1038 *atp = '\0'; 1039 error = zfs_secpolicy_snapshot_perms(name, cr); 1040 *atp = '@'; 1041 if (error != 0) 1042 break; 1043 } 1044 return (error); 1045} 1046 1047/* 1048 * Check for permission to create each snapshot in the nvlist. 1049 */ 1050/* ARGSUSED */ 1051static int 1052zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1053{ 1054 int error = 0; 1055 1056 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 1057 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 1058 char *name = nvpair_name(pair); 1059 char *hashp = strchr(name, '#'); 1060 1061 if (hashp == NULL) { 1062 error = SET_ERROR(EINVAL); 1063 break; 1064 } 1065 *hashp = '\0'; 1066 error = zfs_secpolicy_write_perms(name, 1067 ZFS_DELEG_PERM_BOOKMARK, cr); 1068 *hashp = '#'; 1069 if (error != 0) 1070 break; 1071 } 1072 return (error); 1073} 1074 1075/* ARGSUSED */ 1076static int 1077zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1078{ 1079 nvpair_t *pair, *nextpair; 1080 int error = 0; 1081 1082 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL; 1083 pair = nextpair) { 1084 char *name = nvpair_name(pair); 1085 char *hashp = strchr(name, '#'); 1086 nextpair = nvlist_next_nvpair(innvl, pair); 1087 1088 if (hashp == NULL) { 1089 error = SET_ERROR(EINVAL); 1090 break; 1091 } 1092 1093 *hashp = '\0'; 1094 error = zfs_secpolicy_write_perms(name, 1095 ZFS_DELEG_PERM_DESTROY, cr); 1096 *hashp = '#'; 1097 if (error == ENOENT) { 1098 /* 1099 * Ignore any filesystems that don't exist (we consider 1100 * their bookmarks "already destroyed"). Remove 1101 * the name from the nvl here in case the filesystem 1102 * is created between now and when we try to destroy 1103 * the bookmark (in which case we don't want to 1104 * destroy it since we haven't checked for permission). 1105 */ 1106 fnvlist_remove_nvpair(innvl, pair); 1107 error = 0; 1108 } 1109 if (error != 0) 1110 break; 1111 } 1112 1113 return (error); 1114} 1115 1116/* ARGSUSED */ 1117static int 1118zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1119{ 1120 /* 1121 * Even root must have a proper TSD so that we know what pool 1122 * to log to. 1123 */ 1124 if (tsd_get(zfs_allow_log_key) == NULL) 1125 return (SET_ERROR(EPERM)); 1126 return (0); 1127} 1128 1129static int 1130zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1131{ 1132 char parentname[ZFS_MAX_DATASET_NAME_LEN]; 1133 int error; 1134 char *origin; 1135 1136 if ((error = zfs_get_parent(zc->zc_name, parentname, 1137 sizeof (parentname))) != 0) 1138 return (error); 1139 1140 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 && 1141 (error = zfs_secpolicy_write_perms(origin, 1142 ZFS_DELEG_PERM_CLONE, cr)) != 0) 1143 return (error); 1144 1145 if ((error = zfs_secpolicy_write_perms(parentname, 1146 ZFS_DELEG_PERM_CREATE, cr)) != 0) 1147 return (error); 1148 1149 return (zfs_secpolicy_write_perms(parentname, 1150 ZFS_DELEG_PERM_MOUNT, cr)); 1151} 1152 1153/* 1154 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires 1155 * SYS_CONFIG privilege, which is not available in a local zone. 1156 */ 1157/* ARGSUSED */ 1158static int 1159zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1160{ 1161 if (secpolicy_sys_config(cr, B_FALSE) != 0) 1162 return (SET_ERROR(EPERM)); 1163 1164 return (0); 1165} 1166 1167/* 1168 * Policy for object to name lookups. 1169 */ 1170/* ARGSUSED */ 1171static int 1172zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1173{ 1174 int error; 1175 1176 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0) 1177 return (0); 1178 1179 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr); 1180 return (error); 1181} 1182 1183/* 1184 * Policy for fault injection. Requires all privileges. 1185 */ 1186/* ARGSUSED */ 1187static int 1188zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1189{ 1190 return (secpolicy_zinject(cr)); 1191} 1192 1193/* ARGSUSED */ 1194static int 1195zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1196{ 1197 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value); 1198 1199 if (prop == ZPROP_INVAL) { 1200 if (!zfs_prop_user(zc->zc_value)) 1201 return (SET_ERROR(EINVAL)); 1202 return (zfs_secpolicy_write_perms(zc->zc_name, 1203 ZFS_DELEG_PERM_USERPROP, cr)); 1204 } else { 1205 return (zfs_secpolicy_setprop(zc->zc_name, prop, 1206 NULL, cr)); 1207 } 1208} 1209 1210static int 1211zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1212{ 1213 int err = zfs_secpolicy_read(zc, innvl, cr); 1214 if (err) 1215 return (err); 1216 1217 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 1218 return (SET_ERROR(EINVAL)); 1219 1220 if (zc->zc_value[0] == 0) { 1221 /* 1222 * They are asking about a posix uid/gid. If it's 1223 * themself, allow it. 1224 */ 1225 if (zc->zc_objset_type == ZFS_PROP_USERUSED || 1226 zc->zc_objset_type == ZFS_PROP_USERQUOTA) { 1227 if (zc->zc_guid == crgetuid(cr)) 1228 return (0); 1229 } else { 1230 if (groupmember(zc->zc_guid, cr)) 1231 return (0); 1232 } 1233 } 1234 1235 return (zfs_secpolicy_write_perms(zc->zc_name, 1236 userquota_perms[zc->zc_objset_type], cr)); 1237} 1238 1239static int 1240zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1241{ 1242 int err = zfs_secpolicy_read(zc, innvl, cr); 1243 if (err) 1244 return (err); 1245 1246 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 1247 return (SET_ERROR(EINVAL)); 1248 1249 return (zfs_secpolicy_write_perms(zc->zc_name, 1250 userquota_perms[zc->zc_objset_type], cr)); 1251} 1252 1253/* ARGSUSED */ 1254static int 1255zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1256{ 1257 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION, 1258 NULL, cr)); 1259} 1260 1261/* ARGSUSED */ 1262static int 1263zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1264{ 1265 nvpair_t *pair; 1266 nvlist_t *holds; 1267 int error; 1268 1269 error = nvlist_lookup_nvlist(innvl, "holds", &holds); 1270 if (error != 0) 1271 return (SET_ERROR(EINVAL)); 1272 1273 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL; 1274 pair = nvlist_next_nvpair(holds, pair)) { 1275 char fsname[ZFS_MAX_DATASET_NAME_LEN]; 1276 error = dmu_fsname(nvpair_name(pair), fsname); 1277 if (error != 0) 1278 return (error); 1279 error = zfs_secpolicy_write_perms(fsname, 1280 ZFS_DELEG_PERM_HOLD, cr); 1281 if (error != 0) 1282 return (error); 1283 } 1284 return (0); 1285} 1286 1287/* ARGSUSED */ 1288static int 1289zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1290{ 1291 nvpair_t *pair; 1292 int error; 1293 1294 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL; 1295 pair = nvlist_next_nvpair(innvl, pair)) { 1296 char fsname[ZFS_MAX_DATASET_NAME_LEN]; 1297 error = dmu_fsname(nvpair_name(pair), fsname); 1298 if (error != 0) 1299 return (error); 1300 error = zfs_secpolicy_write_perms(fsname, 1301 ZFS_DELEG_PERM_RELEASE, cr); 1302 if (error != 0) 1303 return (error); 1304 } 1305 return (0); 1306} 1307 1308/* 1309 * Policy for allowing temporary snapshots to be taken or released 1310 */ 1311static int 1312zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1313{ 1314 /* 1315 * A temporary snapshot is the same as a snapshot, 1316 * hold, destroy and release all rolled into one. 1317 * Delegated diff alone is sufficient that we allow this. 1318 */ 1319 int error; 1320 1321 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 1322 ZFS_DELEG_PERM_DIFF, cr)) == 0) 1323 return (0); 1324 1325 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr); 1326 if (error == 0) 1327 error = zfs_secpolicy_hold(zc, innvl, cr); 1328 if (error == 0) 1329 error = zfs_secpolicy_release(zc, innvl, cr); 1330 if (error == 0) 1331 error = zfs_secpolicy_destroy(zc, innvl, cr); 1332 return (error); 1333} 1334 1335/* 1336 * Returns the nvlist as specified by the user in the zfs_cmd_t. 1337 */ 1338static int 1339get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp) 1340{ 1341 char *packed; 1342 int error; 1343 nvlist_t *list = NULL; 1344 1345 /* 1346 * Read in and unpack the user-supplied nvlist. 1347 */ 1348 if (size == 0) 1349 return (SET_ERROR(EINVAL)); 1350 1351 packed = kmem_alloc(size, KM_SLEEP); 1352 1353 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size, 1354 iflag)) != 0) { 1355 kmem_free(packed, size); 1356 return (SET_ERROR(EFAULT)); 1357 } 1358 1359 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) { 1360 kmem_free(packed, size); 1361 return (error); 1362 } 1363 1364 kmem_free(packed, size); 1365 1366 *nvp = list; 1367 return (0); 1368} 1369 1370/* 1371 * Reduce the size of this nvlist until it can be serialized in 'max' bytes. 1372 * Entries will be removed from the end of the nvlist, and one int32 entry 1373 * named "N_MORE_ERRORS" will be added indicating how many entries were 1374 * removed. 1375 */ 1376static int 1377nvlist_smush(nvlist_t *errors, size_t max) 1378{ 1379 size_t size; 1380 1381 size = fnvlist_size(errors); 1382 1383 if (size > max) { 1384 nvpair_t *more_errors; 1385 int n = 0; 1386 1387 if (max < 1024) 1388 return (SET_ERROR(ENOMEM)); 1389 1390 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0); 1391 more_errors = nvlist_prev_nvpair(errors, NULL); 1392 1393 do { 1394 nvpair_t *pair = nvlist_prev_nvpair(errors, 1395 more_errors); 1396 fnvlist_remove_nvpair(errors, pair); 1397 n++; 1398 size = fnvlist_size(errors); 1399 } while (size > max); 1400 1401 fnvlist_remove_nvpair(errors, more_errors); 1402 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n); 1403 ASSERT3U(fnvlist_size(errors), <=, max); 1404 } 1405 1406 return (0); 1407} 1408 1409static int 1410put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl) 1411{ 1412 char *packed = NULL; 1413 int error = 0; 1414 size_t size; 1415 1416 size = fnvlist_size(nvl); 1417 1418 if (size > zc->zc_nvlist_dst_size) { 1419 /* 1420 * Solaris returns ENOMEM here, because even if an error is 1421 * returned from an ioctl(2), new zc_nvlist_dst_size will be 1422 * passed to the userland. This is not the case for FreeBSD. 1423 * We need to return 0, so the kernel will copy the 1424 * zc_nvlist_dst_size back and the userland can discover that a 1425 * bigger buffer is needed. 1426 */ 1427 error = 0; 1428 } else { 1429 packed = fnvlist_pack(nvl, &size); 1430 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst, 1431 size, zc->zc_iflags) != 0) 1432 error = SET_ERROR(EFAULT); 1433 fnvlist_pack_free(packed, size); 1434 } 1435 1436 zc->zc_nvlist_dst_size = size; 1437 zc->zc_nvlist_dst_filled = B_TRUE; 1438 return (error); 1439} 1440 1441int 1442getzfsvfs_impl(objset_t *os, zfsvfs_t **zfvp) 1443{ 1444 vfs_t *vfsp; 1445 int error = 0; 1446 1447 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1448 return (SET_ERROR(EINVAL)); 1449 } 1450 1451 mutex_enter(&os->os_user_ptr_lock); 1452 *zfvp = dmu_objset_get_user(os); 1453 if (*zfvp) { 1454 vfs_ref((*zfvp)->z_vfs); 1455 } else { 1456 error = SET_ERROR(ESRCH); 1457 } 1458 mutex_exit(&os->os_user_ptr_lock); 1459 return (error); 1460} 1461 1462#ifdef illumos 1463int 1464getzfsvfs(const char *dsname, zfsvfs_t **zfvp) 1465{ 1466 objset_t *os; 1467 int error; 1468 1469 error = dmu_objset_hold(dsname, FTAG, &os); 1470 if (error != 0) 1471 return (error); 1472 1473 error = getzfsvfs_impl(os, zfvp); 1474 dmu_objset_rele(os, FTAG); 1475 return (error); 1476} 1477 1478#else 1479 1480static int 1481getzfsvfs_ref(const char *dsname, zfsvfs_t **zfvp) 1482{ 1483 objset_t *os; 1484 int error; 1485 1486 error = dmu_objset_hold(dsname, FTAG, &os); 1487 if (error != 0) 1488 return (error); 1489 1490 error = getzfsvfs_impl(os, zfvp); 1491 dmu_objset_rele(os, FTAG); 1492 return (error); 1493} 1494 1495int 1496getzfsvfs(const char *dsname, zfsvfs_t **zfvp) 1497{ 1498 objset_t *os; 1499 int error; 1500 1501 error = getzfsvfs_ref(dsname, zfvp); 1502 if (error != 0) 1503 return (error); 1504 error = vfs_busy((*zfvp)->z_vfs, 0); 1505 vfs_rel((*zfvp)->z_vfs); 1506 if (error != 0) { 1507 *zfvp = NULL; 1508 error = SET_ERROR(ESRCH); 1509 } 1510 return (error); 1511} 1512#endif 1513 1514/* 1515 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which 1516 * case its z_vfs will be NULL, and it will be opened as the owner. 1517 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER, 1518 * which prevents all vnode ops from running. 1519 */ 1520static int 1521zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer) 1522{ 1523 int error = 0; 1524 1525 if (getzfsvfs(name, zfvp) != 0) 1526 error = zfsvfs_create(name, zfvp); 1527 if (error == 0) { 1528 rrm_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER : 1529 RW_READER, tag); 1530#ifdef illumos 1531 if ((*zfvp)->z_unmounted) { 1532 /* 1533 * XXX we could probably try again, since the unmounting 1534 * thread should be just about to disassociate the 1535 * objset from the zfsvfs. 1536 */ 1537 rrm_exit(&(*zfvp)->z_teardown_lock, tag); 1538 return (SET_ERROR(EBUSY)); 1539 } 1540#else 1541 /* 1542 * vfs_busy() ensures that the filesystem is not and 1543 * can not be unmounted. 1544 */ 1545 ASSERT(!(*zfvp)->z_unmounted); 1546#endif 1547 } 1548 return (error); 1549} 1550 1551static void 1552zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag) 1553{ 1554 rrm_exit(&zfsvfs->z_teardown_lock, tag); 1555 1556 if (zfsvfs->z_vfs) { 1557#ifdef illumos 1558 VFS_RELE(zfsvfs->z_vfs); 1559#else 1560 vfs_unbusy(zfsvfs->z_vfs); 1561#endif 1562 } else { 1563 dmu_objset_disown(zfsvfs->z_os, zfsvfs); 1564 zfsvfs_free(zfsvfs); 1565 } 1566} 1567 1568static int 1569zfs_ioc_pool_create(zfs_cmd_t *zc) 1570{ 1571 int error; 1572 nvlist_t *config, *props = NULL; 1573 nvlist_t *rootprops = NULL; 1574 nvlist_t *zplprops = NULL; 1575 1576 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1577 zc->zc_iflags, &config)) 1578 return (error); 1579 1580 if (zc->zc_nvlist_src_size != 0 && (error = 1581 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1582 zc->zc_iflags, &props))) { 1583 nvlist_free(config); 1584 return (error); 1585 } 1586 1587 if (props) { 1588 nvlist_t *nvl = NULL; 1589 uint64_t version = SPA_VERSION; 1590 1591 (void) nvlist_lookup_uint64(props, 1592 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version); 1593 if (!SPA_VERSION_IS_SUPPORTED(version)) { 1594 error = SET_ERROR(EINVAL); 1595 goto pool_props_bad; 1596 } 1597 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl); 1598 if (nvl) { 1599 error = nvlist_dup(nvl, &rootprops, KM_SLEEP); 1600 if (error != 0) { 1601 nvlist_free(config); 1602 nvlist_free(props); 1603 return (error); 1604 } 1605 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS); 1606 } 1607 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 1608 error = zfs_fill_zplprops_root(version, rootprops, 1609 zplprops, NULL); 1610 if (error != 0) 1611 goto pool_props_bad; 1612 } 1613 1614 error = spa_create(zc->zc_name, config, props, zplprops); 1615 1616 /* 1617 * Set the remaining root properties 1618 */ 1619 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name, 1620 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0) 1621 (void) spa_destroy(zc->zc_name); 1622 1623pool_props_bad: 1624 nvlist_free(rootprops); 1625 nvlist_free(zplprops); 1626 nvlist_free(config); 1627 nvlist_free(props); 1628 1629 return (error); 1630} 1631 1632static int 1633zfs_ioc_pool_destroy(zfs_cmd_t *zc) 1634{ 1635 int error; 1636 zfs_log_history(zc); 1637 error = spa_destroy(zc->zc_name); 1638 if (error == 0) 1639 zvol_remove_minors(zc->zc_name); 1640 return (error); 1641} 1642 1643static int 1644zfs_ioc_pool_import(zfs_cmd_t *zc) 1645{ 1646 nvlist_t *config, *props = NULL; 1647 uint64_t guid; 1648 int error; 1649 1650 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1651 zc->zc_iflags, &config)) != 0) 1652 return (error); 1653 1654 if (zc->zc_nvlist_src_size != 0 && (error = 1655 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1656 zc->zc_iflags, &props))) { 1657 nvlist_free(config); 1658 return (error); 1659 } 1660 1661 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 || 1662 guid != zc->zc_guid) 1663 error = SET_ERROR(EINVAL); 1664 else 1665 error = spa_import(zc->zc_name, config, props, zc->zc_cookie); 1666 1667 if (zc->zc_nvlist_dst != 0) { 1668 int err; 1669 1670 if ((err = put_nvlist(zc, config)) != 0) 1671 error = err; 1672 } 1673 1674 nvlist_free(config); 1675 1676 nvlist_free(props); 1677 1678 return (error); 1679} 1680 1681static int 1682zfs_ioc_pool_export(zfs_cmd_t *zc) 1683{ 1684 int error; 1685 boolean_t force = (boolean_t)zc->zc_cookie; 1686 boolean_t hardforce = (boolean_t)zc->zc_guid; 1687 1688 zfs_log_history(zc); 1689 error = spa_export(zc->zc_name, NULL, force, hardforce); 1690 if (error == 0) 1691 zvol_remove_minors(zc->zc_name); 1692 return (error); 1693} 1694 1695static int 1696zfs_ioc_pool_configs(zfs_cmd_t *zc) 1697{ 1698 nvlist_t *configs; 1699 int error; 1700 1701 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL) 1702 return (SET_ERROR(EEXIST)); 1703 1704 error = put_nvlist(zc, configs); 1705 1706 nvlist_free(configs); 1707 1708 return (error); 1709} 1710 1711/* 1712 * inputs: 1713 * zc_name name of the pool 1714 * 1715 * outputs: 1716 * zc_cookie real errno 1717 * zc_nvlist_dst config nvlist 1718 * zc_nvlist_dst_size size of config nvlist 1719 */ 1720static int 1721zfs_ioc_pool_stats(zfs_cmd_t *zc) 1722{ 1723 nvlist_t *config; 1724 int error; 1725 int ret = 0; 1726 1727 error = spa_get_stats(zc->zc_name, &config, zc->zc_value, 1728 sizeof (zc->zc_value)); 1729 1730 if (config != NULL) { 1731 ret = put_nvlist(zc, config); 1732 nvlist_free(config); 1733 1734 /* 1735 * The config may be present even if 'error' is non-zero. 1736 * In this case we return success, and preserve the real errno 1737 * in 'zc_cookie'. 1738 */ 1739 zc->zc_cookie = error; 1740 } else { 1741 ret = error; 1742 } 1743 1744 return (ret); 1745} 1746 1747/* 1748 * Try to import the given pool, returning pool stats as appropriate so that 1749 * user land knows which devices are available and overall pool health. 1750 */ 1751static int 1752zfs_ioc_pool_tryimport(zfs_cmd_t *zc) 1753{ 1754 nvlist_t *tryconfig, *config; 1755 int error; 1756 1757 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1758 zc->zc_iflags, &tryconfig)) != 0) 1759 return (error); 1760 1761 config = spa_tryimport(tryconfig); 1762 1763 nvlist_free(tryconfig); 1764 1765 if (config == NULL) 1766 return (SET_ERROR(EINVAL)); 1767 1768 error = put_nvlist(zc, config); 1769 nvlist_free(config); 1770 1771 return (error); 1772} 1773 1774/* 1775 * inputs: 1776 * zc_name name of the pool 1777 * zc_cookie scan func (pool_scan_func_t) 1778 * zc_flags scrub pause/resume flag (pool_scrub_cmd_t) 1779 */ 1780static int 1781zfs_ioc_pool_scan(zfs_cmd_t *zc) 1782{ 1783 spa_t *spa; 1784 int error; 1785 1786 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1787 return (error); 1788 1789 if (zc->zc_flags >= POOL_SCRUB_FLAGS_END) 1790 return (SET_ERROR(EINVAL)); 1791 1792 if (zc->zc_flags == POOL_SCRUB_PAUSE) 1793 error = spa_scrub_pause_resume(spa, POOL_SCRUB_PAUSE); 1794 else if (zc->zc_cookie == POOL_SCAN_NONE) 1795 error = spa_scan_stop(spa); 1796 else 1797 error = spa_scan(spa, zc->zc_cookie); 1798 1799 spa_close(spa, FTAG); 1800 1801 return (error); 1802} 1803 1804static int 1805zfs_ioc_pool_freeze(zfs_cmd_t *zc) 1806{ 1807 spa_t *spa; 1808 int error; 1809 1810 error = spa_open(zc->zc_name, &spa, FTAG); 1811 if (error == 0) { 1812 spa_freeze(spa); 1813 spa_close(spa, FTAG); 1814 } 1815 return (error); 1816} 1817 1818static int 1819zfs_ioc_pool_upgrade(zfs_cmd_t *zc) 1820{ 1821 spa_t *spa; 1822 int error; 1823 1824 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1825 return (error); 1826 1827 if (zc->zc_cookie < spa_version(spa) || 1828 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) { 1829 spa_close(spa, FTAG); 1830 return (SET_ERROR(EINVAL)); 1831 } 1832 1833 spa_upgrade(spa, zc->zc_cookie); 1834 spa_close(spa, FTAG); 1835 1836 return (error); 1837} 1838 1839static int 1840zfs_ioc_pool_get_history(zfs_cmd_t *zc) 1841{ 1842 spa_t *spa; 1843 char *hist_buf; 1844 uint64_t size; 1845 int error; 1846 1847 if ((size = zc->zc_history_len) == 0) 1848 return (SET_ERROR(EINVAL)); 1849 1850 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1851 return (error); 1852 1853 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) { 1854 spa_close(spa, FTAG); 1855 return (SET_ERROR(ENOTSUP)); 1856 } 1857 1858 hist_buf = kmem_alloc(size, KM_SLEEP); 1859 if ((error = spa_history_get(spa, &zc->zc_history_offset, 1860 &zc->zc_history_len, hist_buf)) == 0) { 1861 error = ddi_copyout(hist_buf, 1862 (void *)(uintptr_t)zc->zc_history, 1863 zc->zc_history_len, zc->zc_iflags); 1864 } 1865 1866 spa_close(spa, FTAG); 1867 kmem_free(hist_buf, size); 1868 return (error); 1869} 1870 1871static int 1872zfs_ioc_pool_reguid(zfs_cmd_t *zc) 1873{ 1874 spa_t *spa; 1875 int error; 1876 1877 error = spa_open(zc->zc_name, &spa, FTAG); 1878 if (error == 0) { 1879 error = spa_change_guid(spa); 1880 spa_close(spa, FTAG); 1881 } 1882 return (error); 1883} 1884 1885static int 1886zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc) 1887{ 1888 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value)); 1889} 1890 1891/* 1892 * inputs: 1893 * zc_name name of filesystem 1894 * zc_obj object to find 1895 * 1896 * outputs: 1897 * zc_value name of object 1898 */ 1899static int 1900zfs_ioc_obj_to_path(zfs_cmd_t *zc) 1901{ 1902 objset_t *os; 1903 int error; 1904 1905 /* XXX reading from objset not owned */ 1906 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0) 1907 return (error); 1908 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1909 dmu_objset_rele(os, FTAG); 1910 return (SET_ERROR(EINVAL)); 1911 } 1912 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value, 1913 sizeof (zc->zc_value)); 1914 dmu_objset_rele(os, FTAG); 1915 1916 return (error); 1917} 1918 1919/* 1920 * inputs: 1921 * zc_name name of filesystem 1922 * zc_obj object to find 1923 * 1924 * outputs: 1925 * zc_stat stats on object 1926 * zc_value path to object 1927 */ 1928static int 1929zfs_ioc_obj_to_stats(zfs_cmd_t *zc) 1930{ 1931 objset_t *os; 1932 int error; 1933 1934 /* XXX reading from objset not owned */ 1935 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0) 1936 return (error); 1937 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1938 dmu_objset_rele(os, FTAG); 1939 return (SET_ERROR(EINVAL)); 1940 } 1941 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value, 1942 sizeof (zc->zc_value)); 1943 dmu_objset_rele(os, FTAG); 1944 1945 return (error); 1946} 1947 1948static int 1949zfs_ioc_vdev_add(zfs_cmd_t *zc) 1950{ 1951 spa_t *spa; 1952 int error; 1953 nvlist_t *config, **l2cache, **spares; 1954 uint_t nl2cache = 0, nspares = 0; 1955 1956 error = spa_open(zc->zc_name, &spa, FTAG); 1957 if (error != 0) 1958 return (error); 1959 1960 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1961 zc->zc_iflags, &config); 1962 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE, 1963 &l2cache, &nl2cache); 1964 1965 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES, 1966 &spares, &nspares); 1967 1968#ifdef illumos 1969 /* 1970 * A root pool with concatenated devices is not supported. 1971 * Thus, can not add a device to a root pool. 1972 * 1973 * Intent log device can not be added to a rootpool because 1974 * during mountroot, zil is replayed, a seperated log device 1975 * can not be accessed during the mountroot time. 1976 * 1977 * l2cache and spare devices are ok to be added to a rootpool. 1978 */ 1979 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) { 1980 nvlist_free(config); 1981 spa_close(spa, FTAG); 1982 return (SET_ERROR(EDOM)); 1983 } 1984#endif /* illumos */ 1985 1986 if (error == 0) { 1987 error = spa_vdev_add(spa, config); 1988 nvlist_free(config); 1989 } 1990 spa_close(spa, FTAG); 1991 return (error); 1992} 1993 1994/* 1995 * inputs: 1996 * zc_name name of the pool 1997 * zc_nvlist_conf nvlist of devices to remove 1998 * zc_cookie to stop the remove? 1999 */ 2000static int 2001zfs_ioc_vdev_remove(zfs_cmd_t *zc) 2002{ 2003 spa_t *spa; 2004 int error; 2005 2006 error = spa_open(zc->zc_name, &spa, FTAG); 2007 if (error != 0) 2008 return (error); 2009 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE); 2010 spa_close(spa, FTAG); 2011 return (error); 2012} 2013 2014static int 2015zfs_ioc_vdev_set_state(zfs_cmd_t *zc) 2016{ 2017 spa_t *spa; 2018 int error; 2019 vdev_state_t newstate = VDEV_STATE_UNKNOWN; 2020 2021 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 2022 return (error); 2023 switch (zc->zc_cookie) { 2024 case VDEV_STATE_ONLINE: 2025 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate); 2026 break; 2027 2028 case VDEV_STATE_OFFLINE: 2029 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj); 2030 break; 2031 2032 case VDEV_STATE_FAULTED: 2033 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED && 2034 zc->zc_obj != VDEV_AUX_EXTERNAL) 2035 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED; 2036 2037 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj); 2038 break; 2039 2040 case VDEV_STATE_DEGRADED: 2041 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED && 2042 zc->zc_obj != VDEV_AUX_EXTERNAL) 2043 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED; 2044 2045 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj); 2046 break; 2047 2048 default: 2049 error = SET_ERROR(EINVAL); 2050 } 2051 zc->zc_cookie = newstate; 2052 spa_close(spa, FTAG); 2053 return (error); 2054} 2055 2056static int 2057zfs_ioc_vdev_attach(zfs_cmd_t *zc) 2058{ 2059 spa_t *spa; 2060 int replacing = zc->zc_cookie; 2061 nvlist_t *config; 2062 int error; 2063 2064 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 2065 return (error); 2066 2067 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 2068 zc->zc_iflags, &config)) == 0) { 2069 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing); 2070 nvlist_free(config); 2071 } 2072 2073 spa_close(spa, FTAG); 2074 return (error); 2075} 2076 2077static int 2078zfs_ioc_vdev_detach(zfs_cmd_t *zc) 2079{ 2080 spa_t *spa; 2081 int error; 2082 2083 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 2084 return (error); 2085 2086 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE); 2087 2088 spa_close(spa, FTAG); 2089 return (error); 2090} 2091 2092static int 2093zfs_ioc_vdev_split(zfs_cmd_t *zc) 2094{ 2095 spa_t *spa; 2096 nvlist_t *config, *props = NULL; 2097 int error; 2098 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT); 2099 2100 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 2101 return (error); 2102 2103 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 2104 zc->zc_iflags, &config)) { 2105 spa_close(spa, FTAG); 2106 return (error); 2107 } 2108 2109 if (zc->zc_nvlist_src_size != 0 && (error = 2110 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2111 zc->zc_iflags, &props))) { 2112 spa_close(spa, FTAG); 2113 nvlist_free(config); 2114 return (error); 2115 } 2116 2117 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp); 2118 2119 spa_close(spa, FTAG); 2120 2121 nvlist_free(config); 2122 nvlist_free(props); 2123 2124 return (error); 2125} 2126 2127static int 2128zfs_ioc_vdev_setpath(zfs_cmd_t *zc) 2129{ 2130 spa_t *spa; 2131 char *path = zc->zc_value; 2132 uint64_t guid = zc->zc_guid; 2133 int error; 2134 2135 error = spa_open(zc->zc_name, &spa, FTAG); 2136 if (error != 0) 2137 return (error); 2138 2139 error = spa_vdev_setpath(spa, guid, path); 2140 spa_close(spa, FTAG); 2141 return (error); 2142} 2143 2144static int 2145zfs_ioc_vdev_setfru(zfs_cmd_t *zc) 2146{ 2147 spa_t *spa; 2148 char *fru = zc->zc_value; 2149 uint64_t guid = zc->zc_guid; 2150 int error; 2151 2152 error = spa_open(zc->zc_name, &spa, FTAG); 2153 if (error != 0) 2154 return (error); 2155 2156 error = spa_vdev_setfru(spa, guid, fru); 2157 spa_close(spa, FTAG); 2158 return (error); 2159} 2160 2161static int 2162zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os) 2163{ 2164 int error = 0; 2165 nvlist_t *nv; 2166 2167 dmu_objset_fast_stat(os, &zc->zc_objset_stats); 2168 2169 if (zc->zc_nvlist_dst != 0 && 2170 (error = dsl_prop_get_all(os, &nv)) == 0) { 2171 dmu_objset_stats(os, nv); 2172 /* 2173 * NB: zvol_get_stats() will read the objset contents, 2174 * which we aren't supposed to do with a 2175 * DS_MODE_USER hold, because it could be 2176 * inconsistent. So this is a bit of a workaround... 2177 * XXX reading with out owning 2178 */ 2179 if (!zc->zc_objset_stats.dds_inconsistent && 2180 dmu_objset_type(os) == DMU_OST_ZVOL) { 2181 error = zvol_get_stats(os, nv); 2182 if (error == EIO) 2183 return (error); 2184 VERIFY0(error); 2185 } 2186 error = put_nvlist(zc, nv); 2187 nvlist_free(nv); 2188 } 2189 2190 return (error); 2191} 2192 2193/* 2194 * inputs: 2195 * zc_name name of filesystem 2196 * zc_nvlist_dst_size size of buffer for property nvlist 2197 * 2198 * outputs: 2199 * zc_objset_stats stats 2200 * zc_nvlist_dst property nvlist 2201 * zc_nvlist_dst_size size of property nvlist 2202 */ 2203static int 2204zfs_ioc_objset_stats(zfs_cmd_t *zc) 2205{ 2206 objset_t *os; 2207 int error; 2208 2209 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 2210 if (error == 0) { 2211 error = zfs_ioc_objset_stats_impl(zc, os); 2212 dmu_objset_rele(os, FTAG); 2213 } 2214 2215 if (error == ENOMEM) 2216 error = 0; 2217 return (error); 2218} 2219 2220/* 2221 * inputs: 2222 * zc_name name of filesystem 2223 * zc_nvlist_dst_size size of buffer for property nvlist 2224 * 2225 * outputs: 2226 * zc_nvlist_dst received property nvlist 2227 * zc_nvlist_dst_size size of received property nvlist 2228 * 2229 * Gets received properties (distinct from local properties on or after 2230 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from 2231 * local property values. 2232 */ 2233static int 2234zfs_ioc_objset_recvd_props(zfs_cmd_t *zc) 2235{ 2236 int error = 0; 2237 nvlist_t *nv; 2238 2239 /* 2240 * Without this check, we would return local property values if the 2241 * caller has not already received properties on or after 2242 * SPA_VERSION_RECVD_PROPS. 2243 */ 2244 if (!dsl_prop_get_hasrecvd(zc->zc_name)) 2245 return (SET_ERROR(ENOTSUP)); 2246 2247 if (zc->zc_nvlist_dst != 0 && 2248 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) { 2249 error = put_nvlist(zc, nv); 2250 nvlist_free(nv); 2251 } 2252 2253 return (error); 2254} 2255 2256static int 2257nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop) 2258{ 2259 uint64_t value; 2260 int error; 2261 2262 /* 2263 * zfs_get_zplprop() will either find a value or give us 2264 * the default value (if there is one). 2265 */ 2266 if ((error = zfs_get_zplprop(os, prop, &value)) != 0) 2267 return (error); 2268 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0); 2269 return (0); 2270} 2271 2272/* 2273 * inputs: 2274 * zc_name name of filesystem 2275 * zc_nvlist_dst_size size of buffer for zpl property nvlist 2276 * 2277 * outputs: 2278 * zc_nvlist_dst zpl property nvlist 2279 * zc_nvlist_dst_size size of zpl property nvlist 2280 */ 2281static int 2282zfs_ioc_objset_zplprops(zfs_cmd_t *zc) 2283{ 2284 objset_t *os; 2285 int err; 2286 2287 /* XXX reading without owning */ 2288 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os)) 2289 return (err); 2290 2291 dmu_objset_fast_stat(os, &zc->zc_objset_stats); 2292 2293 /* 2294 * NB: nvl_add_zplprop() will read the objset contents, 2295 * which we aren't supposed to do with a DS_MODE_USER 2296 * hold, because it could be inconsistent. 2297 */ 2298 if (zc->zc_nvlist_dst != 0 && 2299 !zc->zc_objset_stats.dds_inconsistent && 2300 dmu_objset_type(os) == DMU_OST_ZFS) { 2301 nvlist_t *nv; 2302 2303 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2304 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 && 2305 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 && 2306 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 && 2307 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0) 2308 err = put_nvlist(zc, nv); 2309 nvlist_free(nv); 2310 } else { 2311 err = SET_ERROR(ENOENT); 2312 } 2313 dmu_objset_rele(os, FTAG); 2314 return (err); 2315} 2316 2317boolean_t 2318dataset_name_hidden(const char *name) 2319{ 2320 /* 2321 * Skip over datasets that are not visible in this zone, 2322 * internal datasets (which have a $ in their name), and 2323 * temporary datasets (which have a % in their name). 2324 */ 2325 if (strchr(name, '$') != NULL) 2326 return (B_TRUE); 2327 if (strchr(name, '%') != NULL) 2328 return (B_TRUE); 2329 if (!INGLOBALZONE(curthread) && !zone_dataset_visible(name, NULL)) 2330 return (B_TRUE); 2331 return (B_FALSE); 2332} 2333 2334/* 2335 * inputs: 2336 * zc_name name of filesystem 2337 * zc_cookie zap cursor 2338 * zc_nvlist_dst_size size of buffer for property nvlist 2339 * 2340 * outputs: 2341 * zc_name name of next filesystem 2342 * zc_cookie zap cursor 2343 * zc_objset_stats stats 2344 * zc_nvlist_dst property nvlist 2345 * zc_nvlist_dst_size size of property nvlist 2346 */ 2347static int 2348zfs_ioc_dataset_list_next(zfs_cmd_t *zc) 2349{ 2350 objset_t *os; 2351 int error; 2352 char *p; 2353 size_t orig_len = strlen(zc->zc_name); 2354 2355top: 2356 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) { 2357 if (error == ENOENT) 2358 error = SET_ERROR(ESRCH); 2359 return (error); 2360 } 2361 2362 p = strrchr(zc->zc_name, '/'); 2363 if (p == NULL || p[1] != '\0') 2364 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name)); 2365 p = zc->zc_name + strlen(zc->zc_name); 2366 2367 do { 2368 error = dmu_dir_list_next(os, 2369 sizeof (zc->zc_name) - (p - zc->zc_name), p, 2370 NULL, &zc->zc_cookie); 2371 if (error == ENOENT) 2372 error = SET_ERROR(ESRCH); 2373 } while (error == 0 && dataset_name_hidden(zc->zc_name)); 2374 dmu_objset_rele(os, FTAG); 2375 2376 /* 2377 * If it's an internal dataset (ie. with a '$' in its name), 2378 * don't try to get stats for it, otherwise we'll return ENOENT. 2379 */ 2380 if (error == 0 && strchr(zc->zc_name, '$') == NULL) { 2381 error = zfs_ioc_objset_stats(zc); /* fill in the stats */ 2382 if (error == ENOENT) { 2383 /* We lost a race with destroy, get the next one. */ 2384 zc->zc_name[orig_len] = '\0'; 2385 goto top; 2386 } 2387 } 2388 return (error); 2389} 2390 2391/* 2392 * inputs: 2393 * zc_name name of filesystem 2394 * zc_cookie zap cursor 2395 * zc_nvlist_dst_size size of buffer for property nvlist 2396 * zc_simple when set, only name is requested 2397 * 2398 * outputs: 2399 * zc_name name of next snapshot 2400 * zc_objset_stats stats 2401 * zc_nvlist_dst property nvlist 2402 * zc_nvlist_dst_size size of property nvlist 2403 */ 2404static int 2405zfs_ioc_snapshot_list_next(zfs_cmd_t *zc) 2406{ 2407 objset_t *os; 2408 int error; 2409 2410 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 2411 if (error != 0) { 2412 return (error == ENOENT ? ESRCH : error); 2413 } 2414 2415 /* 2416 * A dataset name of maximum length cannot have any snapshots, 2417 * so exit immediately. 2418 */ 2419 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= 2420 ZFS_MAX_DATASET_NAME_LEN) { 2421 dmu_objset_rele(os, FTAG); 2422 return (SET_ERROR(ESRCH)); 2423 } 2424 2425 error = dmu_snapshot_list_next(os, 2426 sizeof (zc->zc_name) - strlen(zc->zc_name), 2427 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie, 2428 NULL); 2429 2430 if (error == 0 && !zc->zc_simple) { 2431 dsl_dataset_t *ds; 2432 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool; 2433 2434 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds); 2435 if (error == 0) { 2436 objset_t *ossnap; 2437 2438 error = dmu_objset_from_ds(ds, &ossnap); 2439 if (error == 0) 2440 error = zfs_ioc_objset_stats_impl(zc, ossnap); 2441 dsl_dataset_rele(ds, FTAG); 2442 } 2443 } else if (error == ENOENT) { 2444 error = SET_ERROR(ESRCH); 2445 } 2446 2447 dmu_objset_rele(os, FTAG); 2448 /* if we failed, undo the @ that we tacked on to zc_name */ 2449 if (error != 0) 2450 *strchr(zc->zc_name, '@') = '\0'; 2451 return (error); 2452} 2453 2454static int 2455zfs_prop_set_userquota(const char *dsname, nvpair_t *pair) 2456{ 2457 const char *propname = nvpair_name(pair); 2458 uint64_t *valary; 2459 unsigned int vallen; 2460 const char *domain; 2461 char *dash; 2462 zfs_userquota_prop_t type; 2463 uint64_t rid; 2464 uint64_t quota; 2465 zfsvfs_t *zfsvfs; 2466 int err; 2467 2468 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2469 nvlist_t *attrs; 2470 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 2471 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2472 &pair) != 0) 2473 return (SET_ERROR(EINVAL)); 2474 } 2475 2476 /* 2477 * A correctly constructed propname is encoded as 2478 * userquota@<rid>-<domain>. 2479 */ 2480 if ((dash = strchr(propname, '-')) == NULL || 2481 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 || 2482 vallen != 3) 2483 return (SET_ERROR(EINVAL)); 2484 2485 domain = dash + 1; 2486 type = valary[0]; 2487 rid = valary[1]; 2488 quota = valary[2]; 2489 2490 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE); 2491 if (err == 0) { 2492 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota); 2493 zfsvfs_rele(zfsvfs, FTAG); 2494 } 2495 2496 return (err); 2497} 2498 2499/* 2500 * If the named property is one that has a special function to set its value, 2501 * return 0 on success and a positive error code on failure; otherwise if it is 2502 * not one of the special properties handled by this function, return -1. 2503 * 2504 * XXX: It would be better for callers of the property interface if we handled 2505 * these special cases in dsl_prop.c (in the dsl layer). 2506 */ 2507static int 2508zfs_prop_set_special(const char *dsname, zprop_source_t source, 2509 nvpair_t *pair) 2510{ 2511 const char *propname = nvpair_name(pair); 2512 zfs_prop_t prop = zfs_name_to_prop(propname); 2513 uint64_t intval; 2514 int err = -1; 2515 2516 if (prop == ZPROP_INVAL) { 2517 if (zfs_prop_userquota(propname)) 2518 return (zfs_prop_set_userquota(dsname, pair)); 2519 return (-1); 2520 } 2521 2522 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2523 nvlist_t *attrs; 2524 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 2525 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2526 &pair) == 0); 2527 } 2528 2529 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING) 2530 return (-1); 2531 2532 VERIFY(0 == nvpair_value_uint64(pair, &intval)); 2533 2534 switch (prop) { 2535 case ZFS_PROP_QUOTA: 2536 err = dsl_dir_set_quota(dsname, source, intval); 2537 break; 2538 case ZFS_PROP_REFQUOTA: 2539 err = dsl_dataset_set_refquota(dsname, source, intval); 2540 break; 2541 case ZFS_PROP_FILESYSTEM_LIMIT: 2542 case ZFS_PROP_SNAPSHOT_LIMIT: 2543 if (intval == UINT64_MAX) { 2544 /* clearing the limit, just do it */ 2545 err = 0; 2546 } else { 2547 err = dsl_dir_activate_fs_ss_limit(dsname); 2548 } 2549 /* 2550 * Set err to -1 to force the zfs_set_prop_nvlist code down the 2551 * default path to set the value in the nvlist. 2552 */ 2553 if (err == 0) 2554 err = -1; 2555 break; 2556 case ZFS_PROP_RESERVATION: 2557 err = dsl_dir_set_reservation(dsname, source, intval); 2558 break; 2559 case ZFS_PROP_REFRESERVATION: 2560 err = dsl_dataset_set_refreservation(dsname, source, intval); 2561 break; 2562 case ZFS_PROP_VOLSIZE: 2563 err = zvol_set_volsize(dsname, intval); 2564 break; 2565 case ZFS_PROP_VERSION: 2566 { 2567 zfsvfs_t *zfsvfs; 2568 2569 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0) 2570 break; 2571 2572 err = zfs_set_version(zfsvfs, intval); 2573 zfsvfs_rele(zfsvfs, FTAG); 2574 2575 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) { 2576 zfs_cmd_t *zc; 2577 2578 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP); 2579 (void) strcpy(zc->zc_name, dsname); 2580 (void) zfs_ioc_userspace_upgrade(zc); 2581 kmem_free(zc, sizeof (zfs_cmd_t)); 2582 } 2583 break; 2584 } 2585 default: 2586 err = -1; 2587 } 2588 2589 return (err); 2590} 2591 2592/* 2593 * This function is best effort. If it fails to set any of the given properties, 2594 * it continues to set as many as it can and returns the last error 2595 * encountered. If the caller provides a non-NULL errlist, it will be filled in 2596 * with the list of names of all the properties that failed along with the 2597 * corresponding error numbers. 2598 * 2599 * If every property is set successfully, zero is returned and errlist is not 2600 * modified. 2601 */ 2602int 2603zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl, 2604 nvlist_t *errlist) 2605{ 2606 nvpair_t *pair; 2607 nvpair_t *propval; 2608 int rv = 0; 2609 uint64_t intval; 2610 char *strval; 2611 nvlist_t *genericnvl = fnvlist_alloc(); 2612 nvlist_t *retrynvl = fnvlist_alloc(); 2613 2614retry: 2615 pair = NULL; 2616 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { 2617 const char *propname = nvpair_name(pair); 2618 zfs_prop_t prop = zfs_name_to_prop(propname); 2619 int err = 0; 2620 2621 /* decode the property value */ 2622 propval = pair; 2623 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2624 nvlist_t *attrs; 2625 attrs = fnvpair_value_nvlist(pair); 2626 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2627 &propval) != 0) 2628 err = SET_ERROR(EINVAL); 2629 } 2630 2631 /* Validate value type */ 2632 if (err == 0 && prop == ZPROP_INVAL) { 2633 if (zfs_prop_user(propname)) { 2634 if (nvpair_type(propval) != DATA_TYPE_STRING) 2635 err = SET_ERROR(EINVAL); 2636 } else if (zfs_prop_userquota(propname)) { 2637 if (nvpair_type(propval) != 2638 DATA_TYPE_UINT64_ARRAY) 2639 err = SET_ERROR(EINVAL); 2640 } else { 2641 err = SET_ERROR(EINVAL); 2642 } 2643 } else if (err == 0) { 2644 if (nvpair_type(propval) == DATA_TYPE_STRING) { 2645 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING) 2646 err = SET_ERROR(EINVAL); 2647 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) { 2648 const char *unused; 2649 2650 intval = fnvpair_value_uint64(propval); 2651 2652 switch (zfs_prop_get_type(prop)) { 2653 case PROP_TYPE_NUMBER: 2654 break; 2655 case PROP_TYPE_STRING: 2656 err = SET_ERROR(EINVAL); 2657 break; 2658 case PROP_TYPE_INDEX: 2659 if (zfs_prop_index_to_string(prop, 2660 intval, &unused) != 0) 2661 err = SET_ERROR(EINVAL); 2662 break; 2663 default: 2664 cmn_err(CE_PANIC, 2665 "unknown property type"); 2666 } 2667 } else { 2668 err = SET_ERROR(EINVAL); 2669 } 2670 } 2671 2672 /* Validate permissions */ 2673 if (err == 0) 2674 err = zfs_check_settable(dsname, pair, CRED()); 2675 2676 if (err == 0) { 2677 err = zfs_prop_set_special(dsname, source, pair); 2678 if (err == -1) { 2679 /* 2680 * For better performance we build up a list of 2681 * properties to set in a single transaction. 2682 */ 2683 err = nvlist_add_nvpair(genericnvl, pair); 2684 } else if (err != 0 && nvl != retrynvl) { 2685 /* 2686 * This may be a spurious error caused by 2687 * receiving quota and reservation out of order. 2688 * Try again in a second pass. 2689 */ 2690 err = nvlist_add_nvpair(retrynvl, pair); 2691 } 2692 } 2693 2694 if (err != 0) { 2695 if (errlist != NULL) 2696 fnvlist_add_int32(errlist, propname, err); 2697 rv = err; 2698 } 2699 } 2700 2701 if (nvl != retrynvl && !nvlist_empty(retrynvl)) { 2702 nvl = retrynvl; 2703 goto retry; 2704 } 2705 2706 if (!nvlist_empty(genericnvl) && 2707 dsl_props_set(dsname, source, genericnvl) != 0) { 2708 /* 2709 * If this fails, we still want to set as many properties as we 2710 * can, so try setting them individually. 2711 */ 2712 pair = NULL; 2713 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) { 2714 const char *propname = nvpair_name(pair); 2715 int err = 0; 2716 2717 propval = pair; 2718 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2719 nvlist_t *attrs; 2720 attrs = fnvpair_value_nvlist(pair); 2721 propval = fnvlist_lookup_nvpair(attrs, 2722 ZPROP_VALUE); 2723 } 2724 2725 if (nvpair_type(propval) == DATA_TYPE_STRING) { 2726 strval = fnvpair_value_string(propval); 2727 err = dsl_prop_set_string(dsname, propname, 2728 source, strval); 2729 } else { 2730 intval = fnvpair_value_uint64(propval); 2731 err = dsl_prop_set_int(dsname, propname, source, 2732 intval); 2733 } 2734 2735 if (err != 0) { 2736 if (errlist != NULL) { 2737 fnvlist_add_int32(errlist, propname, 2738 err); 2739 } 2740 rv = err; 2741 } 2742 } 2743 } 2744 nvlist_free(genericnvl); 2745 nvlist_free(retrynvl); 2746 2747 return (rv); 2748} 2749 2750/* 2751 * Check that all the properties are valid user properties. 2752 */ 2753static int 2754zfs_check_userprops(const char *fsname, nvlist_t *nvl) 2755{ 2756 nvpair_t *pair = NULL; 2757 int error = 0; 2758 2759 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { 2760 const char *propname = nvpair_name(pair); 2761 2762 if (!zfs_prop_user(propname) || 2763 nvpair_type(pair) != DATA_TYPE_STRING) 2764 return (SET_ERROR(EINVAL)); 2765 2766 if (error = zfs_secpolicy_write_perms(fsname, 2767 ZFS_DELEG_PERM_USERPROP, CRED())) 2768 return (error); 2769 2770 if (strlen(propname) >= ZAP_MAXNAMELEN) 2771 return (SET_ERROR(ENAMETOOLONG)); 2772 2773 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN) 2774 return (E2BIG); 2775 } 2776 return (0); 2777} 2778 2779static void 2780props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops) 2781{ 2782 nvpair_t *pair; 2783 2784 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2785 2786 pair = NULL; 2787 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) { 2788 if (nvlist_exists(skipped, nvpair_name(pair))) 2789 continue; 2790 2791 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0); 2792 } 2793} 2794 2795static int 2796clear_received_props(const char *dsname, nvlist_t *props, 2797 nvlist_t *skipped) 2798{ 2799 int err = 0; 2800 nvlist_t *cleared_props = NULL; 2801 props_skip(props, skipped, &cleared_props); 2802 if (!nvlist_empty(cleared_props)) { 2803 /* 2804 * Acts on local properties until the dataset has received 2805 * properties at least once on or after SPA_VERSION_RECVD_PROPS. 2806 */ 2807 zprop_source_t flags = (ZPROP_SRC_NONE | 2808 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0)); 2809 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL); 2810 } 2811 nvlist_free(cleared_props); 2812 return (err); 2813} 2814 2815/* 2816 * inputs: 2817 * zc_name name of filesystem 2818 * zc_value name of property to set 2819 * zc_nvlist_src{_size} nvlist of properties to apply 2820 * zc_cookie received properties flag 2821 * 2822 * outputs: 2823 * zc_nvlist_dst{_size} error for each unapplied received property 2824 */ 2825static int 2826zfs_ioc_set_prop(zfs_cmd_t *zc) 2827{ 2828 nvlist_t *nvl; 2829 boolean_t received = zc->zc_cookie; 2830 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED : 2831 ZPROP_SRC_LOCAL); 2832 nvlist_t *errors; 2833 int error; 2834 2835 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2836 zc->zc_iflags, &nvl)) != 0) 2837 return (error); 2838 2839 if (received) { 2840 nvlist_t *origprops; 2841 2842 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) { 2843 (void) clear_received_props(zc->zc_name, 2844 origprops, nvl); 2845 nvlist_free(origprops); 2846 } 2847 2848 error = dsl_prop_set_hasrecvd(zc->zc_name); 2849 } 2850 2851 errors = fnvlist_alloc(); 2852 if (error == 0) 2853 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors); 2854 2855 if (zc->zc_nvlist_dst != 0 && errors != NULL) { 2856 (void) put_nvlist(zc, errors); 2857 } 2858 2859 nvlist_free(errors); 2860 nvlist_free(nvl); 2861 return (error); 2862} 2863 2864/* 2865 * inputs: 2866 * zc_name name of filesystem 2867 * zc_value name of property to inherit 2868 * zc_cookie revert to received value if TRUE 2869 * 2870 * outputs: none 2871 */ 2872static int 2873zfs_ioc_inherit_prop(zfs_cmd_t *zc) 2874{ 2875 const char *propname = zc->zc_value; 2876 zfs_prop_t prop = zfs_name_to_prop(propname); 2877 boolean_t received = zc->zc_cookie; 2878 zprop_source_t source = (received 2879 ? ZPROP_SRC_NONE /* revert to received value, if any */ 2880 : ZPROP_SRC_INHERITED); /* explicitly inherit */ 2881 2882 if (received) { 2883 nvlist_t *dummy; 2884 nvpair_t *pair; 2885 zprop_type_t type; 2886 int err; 2887 2888 /* 2889 * zfs_prop_set_special() expects properties in the form of an 2890 * nvpair with type info. 2891 */ 2892 if (prop == ZPROP_INVAL) { 2893 if (!zfs_prop_user(propname)) 2894 return (SET_ERROR(EINVAL)); 2895 2896 type = PROP_TYPE_STRING; 2897 } else if (prop == ZFS_PROP_VOLSIZE || 2898 prop == ZFS_PROP_VERSION) { 2899 return (SET_ERROR(EINVAL)); 2900 } else { 2901 type = zfs_prop_get_type(prop); 2902 } 2903 2904 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2905 2906 switch (type) { 2907 case PROP_TYPE_STRING: 2908 VERIFY(0 == nvlist_add_string(dummy, propname, "")); 2909 break; 2910 case PROP_TYPE_NUMBER: 2911 case PROP_TYPE_INDEX: 2912 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0)); 2913 break; 2914 default: 2915 nvlist_free(dummy); 2916 return (SET_ERROR(EINVAL)); 2917 } 2918 2919 pair = nvlist_next_nvpair(dummy, NULL); 2920 err = zfs_prop_set_special(zc->zc_name, source, pair); 2921 nvlist_free(dummy); 2922 if (err != -1) 2923 return (err); /* special property already handled */ 2924 } else { 2925 /* 2926 * Only check this in the non-received case. We want to allow 2927 * 'inherit -S' to revert non-inheritable properties like quota 2928 * and reservation to the received or default values even though 2929 * they are not considered inheritable. 2930 */ 2931 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop)) 2932 return (SET_ERROR(EINVAL)); 2933 } 2934 2935 /* property name has been validated by zfs_secpolicy_inherit_prop() */ 2936 return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source)); 2937} 2938 2939static int 2940zfs_ioc_pool_set_props(zfs_cmd_t *zc) 2941{ 2942 nvlist_t *props; 2943 spa_t *spa; 2944 int error; 2945 nvpair_t *pair; 2946 2947 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2948 zc->zc_iflags, &props)) 2949 return (error); 2950 2951 /* 2952 * If the only property is the configfile, then just do a spa_lookup() 2953 * to handle the faulted case. 2954 */ 2955 pair = nvlist_next_nvpair(props, NULL); 2956 if (pair != NULL && strcmp(nvpair_name(pair), 2957 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 && 2958 nvlist_next_nvpair(props, pair) == NULL) { 2959 mutex_enter(&spa_namespace_lock); 2960 if ((spa = spa_lookup(zc->zc_name)) != NULL) { 2961 spa_configfile_set(spa, props, B_FALSE); 2962 spa_config_sync(spa, B_FALSE, B_TRUE); 2963 } 2964 mutex_exit(&spa_namespace_lock); 2965 if (spa != NULL) { 2966 nvlist_free(props); 2967 return (0); 2968 } 2969 } 2970 2971 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { 2972 nvlist_free(props); 2973 return (error); 2974 } 2975 2976 error = spa_prop_set(spa, props); 2977 2978 nvlist_free(props); 2979 spa_close(spa, FTAG); 2980 2981 return (error); 2982} 2983 2984static int 2985zfs_ioc_pool_get_props(zfs_cmd_t *zc) 2986{ 2987 spa_t *spa; 2988 int error; 2989 nvlist_t *nvp = NULL; 2990 2991 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { 2992 /* 2993 * If the pool is faulted, there may be properties we can still 2994 * get (such as altroot and cachefile), so attempt to get them 2995 * anyway. 2996 */ 2997 mutex_enter(&spa_namespace_lock); 2998 if ((spa = spa_lookup(zc->zc_name)) != NULL) 2999 error = spa_prop_get(spa, &nvp); 3000 mutex_exit(&spa_namespace_lock); 3001 } else { 3002 error = spa_prop_get(spa, &nvp); 3003 spa_close(spa, FTAG); 3004 } 3005 3006 if (error == 0 && zc->zc_nvlist_dst != 0) 3007 error = put_nvlist(zc, nvp); 3008 else 3009 error = SET_ERROR(EFAULT); 3010 3011 nvlist_free(nvp); 3012 return (error); 3013} 3014 3015/* 3016 * inputs: 3017 * zc_name name of filesystem 3018 * zc_nvlist_src{_size} nvlist of delegated permissions 3019 * zc_perm_action allow/unallow flag 3020 * 3021 * outputs: none 3022 */ 3023static int 3024zfs_ioc_set_fsacl(zfs_cmd_t *zc) 3025{ 3026 int error; 3027 nvlist_t *fsaclnv = NULL; 3028 3029 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 3030 zc->zc_iflags, &fsaclnv)) != 0) 3031 return (error); 3032 3033 /* 3034 * Verify nvlist is constructed correctly 3035 */ 3036 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) { 3037 nvlist_free(fsaclnv); 3038 return (SET_ERROR(EINVAL)); 3039 } 3040 3041 /* 3042 * If we don't have PRIV_SYS_MOUNT, then validate 3043 * that user is allowed to hand out each permission in 3044 * the nvlist(s) 3045 */ 3046 3047 error = secpolicy_zfs(CRED()); 3048 if (error != 0) { 3049 if (zc->zc_perm_action == B_FALSE) { 3050 error = dsl_deleg_can_allow(zc->zc_name, 3051 fsaclnv, CRED()); 3052 } else { 3053 error = dsl_deleg_can_unallow(zc->zc_name, 3054 fsaclnv, CRED()); 3055 } 3056 } 3057 3058 if (error == 0) 3059 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action); 3060 3061 nvlist_free(fsaclnv); 3062 return (error); 3063} 3064 3065/* 3066 * inputs: 3067 * zc_name name of filesystem 3068 * 3069 * outputs: 3070 * zc_nvlist_src{_size} nvlist of delegated permissions 3071 */ 3072static int 3073zfs_ioc_get_fsacl(zfs_cmd_t *zc) 3074{ 3075 nvlist_t *nvp; 3076 int error; 3077 3078 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) { 3079 error = put_nvlist(zc, nvp); 3080 nvlist_free(nvp); 3081 } 3082 3083 return (error); 3084} 3085 3086/* ARGSUSED */ 3087static void 3088zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) 3089{ 3090 zfs_creat_t *zct = arg; 3091 3092 zfs_create_fs(os, cr, zct->zct_zplprops, tx); 3093} 3094 3095#define ZFS_PROP_UNDEFINED ((uint64_t)-1) 3096 3097/* 3098 * inputs: 3099 * os parent objset pointer (NULL if root fs) 3100 * fuids_ok fuids allowed in this version of the spa? 3101 * sa_ok SAs allowed in this version of the spa? 3102 * createprops list of properties requested by creator 3103 * 3104 * outputs: 3105 * zplprops values for the zplprops we attach to the master node object 3106 * is_ci true if requested file system will be purely case-insensitive 3107 * 3108 * Determine the settings for utf8only, normalization and 3109 * casesensitivity. Specific values may have been requested by the 3110 * creator and/or we can inherit values from the parent dataset. If 3111 * the file system is of too early a vintage, a creator can not 3112 * request settings for these properties, even if the requested 3113 * setting is the default value. We don't actually want to create dsl 3114 * properties for these, so remove them from the source nvlist after 3115 * processing. 3116 */ 3117static int 3118zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver, 3119 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops, 3120 nvlist_t *zplprops, boolean_t *is_ci) 3121{ 3122 uint64_t sense = ZFS_PROP_UNDEFINED; 3123 uint64_t norm = ZFS_PROP_UNDEFINED; 3124 uint64_t u8 = ZFS_PROP_UNDEFINED; 3125 3126 ASSERT(zplprops != NULL); 3127 3128 if (os != NULL && os->os_phys->os_type != DMU_OST_ZFS) 3129 return (SET_ERROR(EINVAL)); 3130 3131 /* 3132 * Pull out creator prop choices, if any. 3133 */ 3134 if (createprops) { 3135 (void) nvlist_lookup_uint64(createprops, 3136 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver); 3137 (void) nvlist_lookup_uint64(createprops, 3138 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm); 3139 (void) nvlist_remove_all(createprops, 3140 zfs_prop_to_name(ZFS_PROP_NORMALIZE)); 3141 (void) nvlist_lookup_uint64(createprops, 3142 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8); 3143 (void) nvlist_remove_all(createprops, 3144 zfs_prop_to_name(ZFS_PROP_UTF8ONLY)); 3145 (void) nvlist_lookup_uint64(createprops, 3146 zfs_prop_to_name(ZFS_PROP_CASE), &sense); 3147 (void) nvlist_remove_all(createprops, 3148 zfs_prop_to_name(ZFS_PROP_CASE)); 3149 } 3150 3151 /* 3152 * If the zpl version requested is whacky or the file system 3153 * or pool is version is too "young" to support normalization 3154 * and the creator tried to set a value for one of the props, 3155 * error out. 3156 */ 3157 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) || 3158 (zplver >= ZPL_VERSION_FUID && !fuids_ok) || 3159 (zplver >= ZPL_VERSION_SA && !sa_ok) || 3160 (zplver < ZPL_VERSION_NORMALIZATION && 3161 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED || 3162 sense != ZFS_PROP_UNDEFINED))) 3163 return (SET_ERROR(ENOTSUP)); 3164 3165 /* 3166 * Put the version in the zplprops 3167 */ 3168 VERIFY(nvlist_add_uint64(zplprops, 3169 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0); 3170 3171 if (norm == ZFS_PROP_UNDEFINED) 3172 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0); 3173 VERIFY(nvlist_add_uint64(zplprops, 3174 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0); 3175 3176 /* 3177 * If we're normalizing, names must always be valid UTF-8 strings. 3178 */ 3179 if (norm) 3180 u8 = 1; 3181 if (u8 == ZFS_PROP_UNDEFINED) 3182 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0); 3183 VERIFY(nvlist_add_uint64(zplprops, 3184 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0); 3185 3186 if (sense == ZFS_PROP_UNDEFINED) 3187 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0); 3188 VERIFY(nvlist_add_uint64(zplprops, 3189 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0); 3190 3191 if (is_ci) 3192 *is_ci = (sense == ZFS_CASE_INSENSITIVE); 3193 3194 return (0); 3195} 3196 3197static int 3198zfs_fill_zplprops(const char *dataset, nvlist_t *createprops, 3199 nvlist_t *zplprops, boolean_t *is_ci) 3200{ 3201 boolean_t fuids_ok, sa_ok; 3202 uint64_t zplver = ZPL_VERSION; 3203 objset_t *os = NULL; 3204 char parentname[ZFS_MAX_DATASET_NAME_LEN]; 3205 char *cp; 3206 spa_t *spa; 3207 uint64_t spa_vers; 3208 int error; 3209 3210 (void) strlcpy(parentname, dataset, sizeof (parentname)); 3211 cp = strrchr(parentname, '/'); 3212 ASSERT(cp != NULL); 3213 cp[0] = '\0'; 3214 3215 if ((error = spa_open(dataset, &spa, FTAG)) != 0) 3216 return (error); 3217 3218 spa_vers = spa_version(spa); 3219 spa_close(spa, FTAG); 3220 3221 zplver = zfs_zpl_version_map(spa_vers); 3222 fuids_ok = (zplver >= ZPL_VERSION_FUID); 3223 sa_ok = (zplver >= ZPL_VERSION_SA); 3224 3225 /* 3226 * Open parent object set so we can inherit zplprop values. 3227 */ 3228 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0) 3229 return (error); 3230 3231 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops, 3232 zplprops, is_ci); 3233 dmu_objset_rele(os, FTAG); 3234 return (error); 3235} 3236 3237static int 3238zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops, 3239 nvlist_t *zplprops, boolean_t *is_ci) 3240{ 3241 boolean_t fuids_ok; 3242 boolean_t sa_ok; 3243 uint64_t zplver = ZPL_VERSION; 3244 int error; 3245 3246 zplver = zfs_zpl_version_map(spa_vers); 3247 fuids_ok = (zplver >= ZPL_VERSION_FUID); 3248 sa_ok = (zplver >= ZPL_VERSION_SA); 3249 3250 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok, 3251 createprops, zplprops, is_ci); 3252 return (error); 3253} 3254 3255/* 3256 * innvl: { 3257 * "type" -> dmu_objset_type_t (int32) 3258 * (optional) "props" -> { prop -> value } 3259 * } 3260 * 3261 * outnvl: propname -> error code (int32) 3262 */ 3263static int 3264zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3265{ 3266 int error = 0; 3267 zfs_creat_t zct = { 0 }; 3268 nvlist_t *nvprops = NULL; 3269 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx); 3270 int32_t type32; 3271 dmu_objset_type_t type; 3272 boolean_t is_insensitive = B_FALSE; 3273 3274 if (nvlist_lookup_int32(innvl, "type", &type32) != 0) 3275 return (SET_ERROR(EINVAL)); 3276 type = type32; 3277 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops); 3278 3279 switch (type) { 3280 case DMU_OST_ZFS: 3281 cbfunc = zfs_create_cb; 3282 break; 3283 3284 case DMU_OST_ZVOL: 3285 cbfunc = zvol_create_cb; 3286 break; 3287 3288 default: 3289 cbfunc = NULL; 3290 break; 3291 } 3292 if (strchr(fsname, '@') || 3293 strchr(fsname, '%')) 3294 return (SET_ERROR(EINVAL)); 3295 3296 zct.zct_props = nvprops; 3297 3298 if (cbfunc == NULL) 3299 return (SET_ERROR(EINVAL)); 3300 3301 if (type == DMU_OST_ZVOL) { 3302 uint64_t volsize, volblocksize; 3303 3304 if (nvprops == NULL) 3305 return (SET_ERROR(EINVAL)); 3306 if (nvlist_lookup_uint64(nvprops, 3307 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0) 3308 return (SET_ERROR(EINVAL)); 3309 3310 if ((error = nvlist_lookup_uint64(nvprops, 3311 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), 3312 &volblocksize)) != 0 && error != ENOENT) 3313 return (SET_ERROR(EINVAL)); 3314 3315 if (error != 0) 3316 volblocksize = zfs_prop_default_numeric( 3317 ZFS_PROP_VOLBLOCKSIZE); 3318 3319 if ((error = zvol_check_volblocksize( 3320 volblocksize)) != 0 || 3321 (error = zvol_check_volsize(volsize, 3322 volblocksize)) != 0) 3323 return (error); 3324 } else if (type == DMU_OST_ZFS) { 3325 int error; 3326 3327 /* 3328 * We have to have normalization and 3329 * case-folding flags correct when we do the 3330 * file system creation, so go figure them out 3331 * now. 3332 */ 3333 VERIFY(nvlist_alloc(&zct.zct_zplprops, 3334 NV_UNIQUE_NAME, KM_SLEEP) == 0); 3335 error = zfs_fill_zplprops(fsname, nvprops, 3336 zct.zct_zplprops, &is_insensitive); 3337 if (error != 0) { 3338 nvlist_free(zct.zct_zplprops); 3339 return (error); 3340 } 3341 } 3342 3343 error = dmu_objset_create(fsname, type, 3344 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct); 3345 nvlist_free(zct.zct_zplprops); 3346 3347 /* 3348 * It would be nice to do this atomically. 3349 */ 3350 if (error == 0) { 3351 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL, 3352 nvprops, outnvl); 3353 if (error != 0) 3354 (void) dsl_destroy_head(fsname); 3355 } 3356#ifdef __FreeBSD__ 3357 if (error == 0 && type == DMU_OST_ZVOL) 3358 zvol_create_minors(fsname); 3359#endif 3360 return (error); 3361} 3362 3363/* 3364 * innvl: { 3365 * "origin" -> name of origin snapshot 3366 * (optional) "props" -> { prop -> value } 3367 * } 3368 * 3369 * outnvl: propname -> error code (int32) 3370 */ 3371static int 3372zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3373{ 3374 int error = 0; 3375 nvlist_t *nvprops = NULL; 3376 char *origin_name; 3377 3378 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0) 3379 return (SET_ERROR(EINVAL)); 3380 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops); 3381 3382 if (strchr(fsname, '@') || 3383 strchr(fsname, '%')) 3384 return (SET_ERROR(EINVAL)); 3385 3386 if (dataset_namecheck(origin_name, NULL, NULL) != 0) 3387 return (SET_ERROR(EINVAL)); 3388 error = dmu_objset_clone(fsname, origin_name); 3389 if (error != 0) 3390 return (error); 3391 3392 /* 3393 * It would be nice to do this atomically. 3394 */ 3395 if (error == 0) { 3396 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL, 3397 nvprops, outnvl); 3398 if (error != 0) 3399 (void) dsl_destroy_head(fsname); 3400 } 3401#ifdef __FreeBSD__ 3402 if (error == 0) 3403 zvol_create_minors(fsname); 3404#endif 3405 return (error); 3406} 3407 3408/* 3409 * innvl: { 3410 * "snaps" -> { snapshot1, snapshot2 } 3411 * (optional) "props" -> { prop -> value (string) } 3412 * } 3413 * 3414 * outnvl: snapshot -> error code (int32) 3415 */ 3416static int 3417zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3418{ 3419 nvlist_t *snaps; 3420 nvlist_t *props = NULL; 3421 int error, poollen; 3422 nvpair_t *pair; 3423 3424 (void) nvlist_lookup_nvlist(innvl, "props", &props); 3425 if ((error = zfs_check_userprops(poolname, props)) != 0) 3426 return (error); 3427 3428 if (!nvlist_empty(props) && 3429 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS)) 3430 return (SET_ERROR(ENOTSUP)); 3431 3432 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 3433 return (SET_ERROR(EINVAL)); 3434 poollen = strlen(poolname); 3435 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 3436 pair = nvlist_next_nvpair(snaps, pair)) { 3437 const char *name = nvpair_name(pair); 3438 const char *cp = strchr(name, '@'); 3439 3440 /* 3441 * The snap name must contain an @, and the part after it must 3442 * contain only valid characters. 3443 */ 3444 if (cp == NULL || 3445 zfs_component_namecheck(cp + 1, NULL, NULL) != 0) 3446 return (SET_ERROR(EINVAL)); 3447 3448 /* 3449 * The snap must be in the specified pool. 3450 */ 3451 if (strncmp(name, poolname, poollen) != 0 || 3452 (name[poollen] != '/' && name[poollen] != '@')) 3453 return (SET_ERROR(EXDEV)); 3454 3455 /* This must be the only snap of this fs. */ 3456 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair); 3457 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) { 3458 if (strncmp(name, nvpair_name(pair2), cp - name + 1) 3459 == 0) { 3460 return (SET_ERROR(EXDEV)); 3461 } 3462 } 3463 } 3464 3465 error = dsl_dataset_snapshot(snaps, props, outnvl); 3466 return (error); 3467} 3468 3469/* 3470 * innvl: "message" -> string 3471 */ 3472/* ARGSUSED */ 3473static int 3474zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl) 3475{ 3476 char *message; 3477 spa_t *spa; 3478 int error; 3479 char *poolname; 3480 3481 /* 3482 * The poolname in the ioctl is not set, we get it from the TSD, 3483 * which was set at the end of the last successful ioctl that allows 3484 * logging. The secpolicy func already checked that it is set. 3485 * Only one log ioctl is allowed after each successful ioctl, so 3486 * we clear the TSD here. 3487 */ 3488 poolname = tsd_get(zfs_allow_log_key); 3489 (void) tsd_set(zfs_allow_log_key, NULL); 3490 error = spa_open(poolname, &spa, FTAG); 3491 strfree(poolname); 3492 if (error != 0) 3493 return (error); 3494 3495 if (nvlist_lookup_string(innvl, "message", &message) != 0) { 3496 spa_close(spa, FTAG); 3497 return (SET_ERROR(EINVAL)); 3498 } 3499 3500 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) { 3501 spa_close(spa, FTAG); 3502 return (SET_ERROR(ENOTSUP)); 3503 } 3504 3505 error = spa_history_log(spa, message); 3506 spa_close(spa, FTAG); 3507 return (error); 3508} 3509 3510#ifdef __FreeBSD__ 3511static int 3512zfs_ioc_nextboot(const char *unused, nvlist_t *innvl, nvlist_t *outnvl) 3513{ 3514 char name[MAXNAMELEN]; 3515 spa_t *spa; 3516 vdev_t *vd; 3517 char *command; 3518 uint64_t pool_guid; 3519 uint64_t vdev_guid; 3520 int error; 3521 3522 if (nvlist_lookup_uint64(innvl, 3523 ZPOOL_CONFIG_POOL_GUID, &pool_guid) != 0) 3524 return (EINVAL); 3525 if (nvlist_lookup_uint64(innvl, 3526 ZPOOL_CONFIG_GUID, &vdev_guid) != 0) 3527 return (EINVAL); 3528 if (nvlist_lookup_string(innvl, 3529 "command", &command) != 0) 3530 return (EINVAL); 3531 3532 mutex_enter(&spa_namespace_lock); 3533 spa = spa_by_guid(pool_guid, vdev_guid); 3534 if (spa != NULL) 3535 strcpy(name, spa_name(spa)); 3536 mutex_exit(&spa_namespace_lock); 3537 if (spa == NULL) 3538 return (ENOENT); 3539 3540 if ((error = spa_open(name, &spa, FTAG)) != 0) 3541 return (error); 3542 spa_vdev_state_enter(spa, SCL_ALL); 3543 vd = spa_lookup_by_guid(spa, vdev_guid, B_TRUE); 3544 if (vd == NULL) { 3545 (void) spa_vdev_state_exit(spa, NULL, ENXIO); 3546 spa_close(spa, FTAG); 3547 return (ENODEV); 3548 } 3549 error = vdev_label_write_pad2(vd, command, strlen(command)); 3550 (void) spa_vdev_state_exit(spa, NULL, 0); 3551 txg_wait_synced(spa->spa_dsl_pool, 0); 3552 spa_close(spa, FTAG); 3553 return (error); 3554} 3555#endif 3556 3557/* 3558 * The dp_config_rwlock must not be held when calling this, because the 3559 * unmount may need to write out data. 3560 * 3561 * This function is best-effort. Callers must deal gracefully if it 3562 * remains mounted (or is remounted after this call). 3563 * 3564 * Returns 0 if the argument is not a snapshot, or it is not currently a 3565 * filesystem, or we were able to unmount it. Returns error code otherwise. 3566 */ 3567void 3568zfs_unmount_snap(const char *snapname) 3569{ 3570 vfs_t *vfsp = NULL; 3571 zfsvfs_t *zfsvfs = NULL; 3572 3573 if (strchr(snapname, '@') == NULL) 3574 return; 3575 3576 int err = getzfsvfs_ref(snapname, &zfsvfs); 3577 if (err != 0) { 3578 ASSERT3P(zfsvfs, ==, NULL); 3579 return; 3580 } 3581 vfsp = zfsvfs->z_vfs; 3582 3583 ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os))); 3584 3585#ifdef illumos 3586 err = vn_vfswlock(vfsp->vfs_vnodecovered); 3587 VFS_RELE(vfsp); 3588 if (err != 0) 3589 return; 3590#endif 3591 3592 /* 3593 * Always force the unmount for snapshots. 3594 */ 3595#ifdef illumos 3596 (void) dounmount(vfsp, MS_FORCE, kcred); 3597#else 3598 (void) dounmount(vfsp, MS_FORCE, curthread); 3599#endif 3600} 3601 3602/* ARGSUSED */ 3603static int 3604zfs_unmount_snap_cb(const char *snapname, void *arg) 3605{ 3606 zfs_unmount_snap(snapname); 3607 return (0); 3608} 3609 3610/* 3611 * When a clone is destroyed, its origin may also need to be destroyed, 3612 * in which case it must be unmounted. This routine will do that unmount 3613 * if necessary. 3614 */ 3615void 3616zfs_destroy_unmount_origin(const char *fsname) 3617{ 3618 int error; 3619 objset_t *os; 3620 dsl_dataset_t *ds; 3621 3622 error = dmu_objset_hold(fsname, FTAG, &os); 3623 if (error != 0) 3624 return; 3625 ds = dmu_objset_ds(os); 3626 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) { 3627 char originname[ZFS_MAX_DATASET_NAME_LEN]; 3628 dsl_dataset_name(ds->ds_prev, originname); 3629 dmu_objset_rele(os, FTAG); 3630 zfs_unmount_snap(originname); 3631 } else { 3632 dmu_objset_rele(os, FTAG); 3633 } 3634} 3635 3636/* 3637 * innvl: { 3638 * "snaps" -> { snapshot1, snapshot2 } 3639 * (optional boolean) "defer" 3640 * } 3641 * 3642 * outnvl: snapshot -> error code (int32) 3643 * 3644 */ 3645/* ARGSUSED */ 3646static int 3647zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3648{ 3649 int error, poollen; 3650 nvlist_t *snaps; 3651 nvpair_t *pair; 3652 boolean_t defer; 3653 3654 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 3655 return (SET_ERROR(EINVAL)); 3656 defer = nvlist_exists(innvl, "defer"); 3657 3658 poollen = strlen(poolname); 3659 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 3660 pair = nvlist_next_nvpair(snaps, pair)) { 3661 const char *name = nvpair_name(pair); 3662 3663 /* 3664 * The snap must be in the specified pool to prevent the 3665 * invalid removal of zvol minors below. 3666 */ 3667 if (strncmp(name, poolname, poollen) != 0 || 3668 (name[poollen] != '/' && name[poollen] != '@')) 3669 return (SET_ERROR(EXDEV)); 3670 3671 zfs_unmount_snap(nvpair_name(pair)); 3672#if defined(__FreeBSD__) 3673 zvol_remove_minors(name); 3674#endif 3675 } 3676 3677 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl)); 3678} 3679 3680/* 3681 * Create bookmarks. Bookmark names are of the form <fs>#<bmark>. 3682 * All bookmarks must be in the same pool. 3683 * 3684 * innvl: { 3685 * bookmark1 -> snapshot1, bookmark2 -> snapshot2 3686 * } 3687 * 3688 * outnvl: bookmark -> error code (int32) 3689 * 3690 */ 3691/* ARGSUSED */ 3692static int 3693zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3694{ 3695 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 3696 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 3697 char *snap_name; 3698 3699 /* 3700 * Verify the snapshot argument. 3701 */ 3702 if (nvpair_value_string(pair, &snap_name) != 0) 3703 return (SET_ERROR(EINVAL)); 3704 3705 3706 /* Verify that the keys (bookmarks) are unique */ 3707 for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair); 3708 pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) { 3709 if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0) 3710 return (SET_ERROR(EINVAL)); 3711 } 3712 } 3713 3714 return (dsl_bookmark_create(innvl, outnvl)); 3715} 3716 3717/* 3718 * innvl: { 3719 * property 1, property 2, ... 3720 * } 3721 * 3722 * outnvl: { 3723 * bookmark name 1 -> { property 1, property 2, ... }, 3724 * bookmark name 2 -> { property 1, property 2, ... } 3725 * } 3726 * 3727 */ 3728static int 3729zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3730{ 3731 return (dsl_get_bookmarks(fsname, innvl, outnvl)); 3732} 3733 3734/* 3735 * innvl: { 3736 * bookmark name 1, bookmark name 2 3737 * } 3738 * 3739 * outnvl: bookmark -> error code (int32) 3740 * 3741 */ 3742static int 3743zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl, 3744 nvlist_t *outnvl) 3745{ 3746 int error, poollen; 3747 3748 poollen = strlen(poolname); 3749 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 3750 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 3751 const char *name = nvpair_name(pair); 3752 const char *cp = strchr(name, '#'); 3753 3754 /* 3755 * The bookmark name must contain an #, and the part after it 3756 * must contain only valid characters. 3757 */ 3758 if (cp == NULL || 3759 zfs_component_namecheck(cp + 1, NULL, NULL) != 0) 3760 return (SET_ERROR(EINVAL)); 3761 3762 /* 3763 * The bookmark must be in the specified pool. 3764 */ 3765 if (strncmp(name, poolname, poollen) != 0 || 3766 (name[poollen] != '/' && name[poollen] != '#')) 3767 return (SET_ERROR(EXDEV)); 3768 } 3769 3770 error = dsl_bookmark_destroy(innvl, outnvl); 3771 return (error); 3772} 3773 3774static int 3775zfs_ioc_channel_program(const char *poolname, nvlist_t *innvl, 3776 nvlist_t *outnvl) 3777{ 3778 char *program; 3779 uint64_t instrlimit, memlimit; 3780 boolean_t sync_flag; 3781 nvpair_t *nvarg = NULL; 3782 3783 if (0 != nvlist_lookup_string(innvl, ZCP_ARG_PROGRAM, &program)) { 3784 return (EINVAL); 3785 } 3786 if (0 != nvlist_lookup_boolean_value(innvl, ZCP_ARG_SYNC, &sync_flag)) { 3787 sync_flag = B_TRUE; 3788 } 3789 if (0 != nvlist_lookup_uint64(innvl, ZCP_ARG_INSTRLIMIT, &instrlimit)) { 3790 instrlimit = ZCP_DEFAULT_INSTRLIMIT; 3791 } 3792 if (0 != nvlist_lookup_uint64(innvl, ZCP_ARG_MEMLIMIT, &memlimit)) { 3793 memlimit = ZCP_DEFAULT_MEMLIMIT; 3794 } 3795 if (0 != nvlist_lookup_nvpair(innvl, ZCP_ARG_ARGLIST, &nvarg)) { 3796 return (EINVAL); 3797 } 3798 3799 if (instrlimit == 0 || instrlimit > zfs_lua_max_instrlimit) 3800 return (EINVAL); 3801 if (memlimit == 0 || memlimit > zfs_lua_max_memlimit) 3802 return (EINVAL); 3803 3804 return (zcp_eval(poolname, program, sync_flag, instrlimit, memlimit, 3805 nvarg, outnvl)); 3806} 3807 3808/* 3809 * inputs: 3810 * zc_name name of dataset to destroy 3811 * zc_objset_type type of objset 3812 * zc_defer_destroy mark for deferred destroy 3813 * 3814 * outputs: none 3815 */ 3816static int 3817zfs_ioc_destroy(zfs_cmd_t *zc) 3818{ 3819 int err; 3820 3821 if (zc->zc_objset_type == DMU_OST_ZFS) 3822 zfs_unmount_snap(zc->zc_name); 3823 3824 if (strchr(zc->zc_name, '@')) 3825 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy); 3826 else 3827 err = dsl_destroy_head(zc->zc_name); 3828 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0) 3829#ifdef __FreeBSD__ 3830 zvol_remove_minors(zc->zc_name); 3831#else 3832 (void) zvol_remove_minor(zc->zc_name); 3833#endif 3834 return (err); 3835} 3836 3837/* 3838 * fsname is name of dataset to rollback (to most recent snapshot) 3839 * 3840 * innvl may contain name of expected target snapshot 3841 * 3842 * outnvl: "target" -> name of most recent snapshot 3843 * } 3844 */ 3845/* ARGSUSED */ 3846static int 3847zfs_ioc_rollback(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3848{ 3849 zfsvfs_t *zfsvfs; 3850 char *target = NULL; 3851 int error; 3852 3853 (void) nvlist_lookup_string(innvl, "target", &target); 3854 if (target != NULL) { 3855 int fslen = strlen(fsname); 3856 3857 if (strncmp(fsname, target, fslen) != 0) 3858 return (SET_ERROR(EINVAL)); 3859 if (target[fslen] != '@') 3860 return (SET_ERROR(EINVAL)); 3861 } 3862 3863 if (getzfsvfs(fsname, &zfsvfs) == 0) { 3864 dsl_dataset_t *ds; 3865 3866 ds = dmu_objset_ds(zfsvfs->z_os); 3867 error = zfs_suspend_fs(zfsvfs); 3868 if (error == 0) { 3869 int resume_err; 3870 3871 error = dsl_dataset_rollback(fsname, target, zfsvfs, 3872 outnvl); 3873 resume_err = zfs_resume_fs(zfsvfs, ds); 3874 error = error ? error : resume_err; 3875 } 3876#ifdef illumos 3877 VFS_RELE(zfsvfs->z_vfs); 3878#else 3879 vfs_unbusy(zfsvfs->z_vfs); 3880#endif 3881 } else { 3882 error = dsl_dataset_rollback(fsname, target, NULL, outnvl); 3883 } 3884 return (error); 3885} 3886 3887static int 3888recursive_unmount(const char *fsname, void *arg) 3889{ 3890 const char *snapname = arg; 3891 char fullname[ZFS_MAX_DATASET_NAME_LEN]; 3892 3893 (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname); 3894 zfs_unmount_snap(fullname); 3895 3896 return (0); 3897} 3898 3899/* 3900 * inputs: 3901 * zc_name old name of dataset 3902 * zc_value new name of dataset 3903 * zc_cookie recursive flag (only valid for snapshots) 3904 * 3905 * outputs: none 3906 */ 3907static int 3908zfs_ioc_rename(zfs_cmd_t *zc) 3909{ 3910 boolean_t recursive = zc->zc_cookie & 1; 3911 char *at; 3912 boolean_t allow_mounted = B_TRUE; 3913 3914#ifdef __FreeBSD__ 3915 allow_mounted = (zc->zc_cookie & 2) != 0; 3916#endif 3917 3918 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0'; 3919 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 3920 strchr(zc->zc_value, '%')) 3921 return (SET_ERROR(EINVAL)); 3922 3923 at = strchr(zc->zc_name, '@'); 3924 if (at != NULL) { 3925 /* snaps must be in same fs */ 3926 int error; 3927 3928 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1)) 3929 return (SET_ERROR(EXDEV)); 3930 *at = '\0'; 3931 if (zc->zc_objset_type == DMU_OST_ZFS && !allow_mounted) { 3932 error = dmu_objset_find(zc->zc_name, 3933 recursive_unmount, at + 1, 3934 recursive ? DS_FIND_CHILDREN : 0); 3935 if (error != 0) { 3936 *at = '@'; 3937 return (error); 3938 } 3939 } 3940 error = dsl_dataset_rename_snapshot(zc->zc_name, 3941 at + 1, strchr(zc->zc_value, '@') + 1, recursive); 3942 *at = '@'; 3943 3944 return (error); 3945 } else { 3946#ifdef illumos 3947 if (zc->zc_objset_type == DMU_OST_ZVOL) 3948 (void) zvol_remove_minor(zc->zc_name); 3949#endif 3950 return (dsl_dir_rename(zc->zc_name, zc->zc_value)); 3951 } 3952} 3953 3954static int 3955zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr) 3956{ 3957 const char *propname = nvpair_name(pair); 3958 boolean_t issnap = (strchr(dsname, '@') != NULL); 3959 zfs_prop_t prop = zfs_name_to_prop(propname); 3960 uint64_t intval; 3961 int err; 3962 3963 if (prop == ZPROP_INVAL) { 3964 if (zfs_prop_user(propname)) { 3965 if (err = zfs_secpolicy_write_perms(dsname, 3966 ZFS_DELEG_PERM_USERPROP, cr)) 3967 return (err); 3968 return (0); 3969 } 3970 3971 if (!issnap && zfs_prop_userquota(propname)) { 3972 const char *perm = NULL; 3973 const char *uq_prefix = 3974 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA]; 3975 const char *gq_prefix = 3976 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA]; 3977 3978 if (strncmp(propname, uq_prefix, 3979 strlen(uq_prefix)) == 0) { 3980 perm = ZFS_DELEG_PERM_USERQUOTA; 3981 } else if (strncmp(propname, gq_prefix, 3982 strlen(gq_prefix)) == 0) { 3983 perm = ZFS_DELEG_PERM_GROUPQUOTA; 3984 } else { 3985 /* USERUSED and GROUPUSED are read-only */ 3986 return (SET_ERROR(EINVAL)); 3987 } 3988 3989 if (err = zfs_secpolicy_write_perms(dsname, perm, cr)) 3990 return (err); 3991 return (0); 3992 } 3993 3994 return (SET_ERROR(EINVAL)); 3995 } 3996 3997 if (issnap) 3998 return (SET_ERROR(EINVAL)); 3999 4000 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 4001 /* 4002 * dsl_prop_get_all_impl() returns properties in this 4003 * format. 4004 */ 4005 nvlist_t *attrs; 4006 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 4007 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 4008 &pair) == 0); 4009 } 4010 4011 /* 4012 * Check that this value is valid for this pool version 4013 */ 4014 switch (prop) { 4015 case ZFS_PROP_COMPRESSION: 4016 /* 4017 * If the user specified gzip compression, make sure 4018 * the SPA supports it. We ignore any errors here since 4019 * we'll catch them later. 4020 */ 4021 if (nvpair_value_uint64(pair, &intval) == 0) { 4022 if (intval >= ZIO_COMPRESS_GZIP_1 && 4023 intval <= ZIO_COMPRESS_GZIP_9 && 4024 zfs_earlier_version(dsname, 4025 SPA_VERSION_GZIP_COMPRESSION)) { 4026 return (SET_ERROR(ENOTSUP)); 4027 } 4028 4029 if (intval == ZIO_COMPRESS_ZLE && 4030 zfs_earlier_version(dsname, 4031 SPA_VERSION_ZLE_COMPRESSION)) 4032 return (SET_ERROR(ENOTSUP)); 4033 4034 if (intval == ZIO_COMPRESS_LZ4) { 4035 spa_t *spa; 4036 4037 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 4038 return (err); 4039 4040 if (!spa_feature_is_enabled(spa, 4041 SPA_FEATURE_LZ4_COMPRESS)) { 4042 spa_close(spa, FTAG); 4043 return (SET_ERROR(ENOTSUP)); 4044 } 4045 spa_close(spa, FTAG); 4046 } 4047 4048 /* 4049 * If this is a bootable dataset then 4050 * verify that the compression algorithm 4051 * is supported for booting. We must return 4052 * something other than ENOTSUP since it 4053 * implies a downrev pool version. 4054 */ 4055 if (zfs_is_bootfs(dsname) && 4056 !BOOTFS_COMPRESS_VALID(intval)) { 4057 return (SET_ERROR(ERANGE)); 4058 } 4059 } 4060 break; 4061 4062 case ZFS_PROP_COPIES: 4063 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS)) 4064 return (SET_ERROR(ENOTSUP)); 4065 break; 4066 4067 case ZFS_PROP_RECORDSIZE: 4068 /* Record sizes above 128k need the feature to be enabled */ 4069 if (nvpair_value_uint64(pair, &intval) == 0 && 4070 intval > SPA_OLD_MAXBLOCKSIZE) { 4071 spa_t *spa; 4072 4073 /* 4074 * If this is a bootable dataset then 4075 * the we don't allow large (>128K) blocks, 4076 * because GRUB doesn't support them. 4077 */ 4078 if (zfs_is_bootfs(dsname) && 4079 intval > SPA_OLD_MAXBLOCKSIZE) { 4080 return (SET_ERROR(ERANGE)); 4081 } 4082 4083 /* 4084 * We don't allow setting the property above 1MB, 4085 * unless the tunable has been changed. 4086 */ 4087 if (intval > zfs_max_recordsize || 4088 intval > SPA_MAXBLOCKSIZE) 4089 return (SET_ERROR(ERANGE)); 4090 4091 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 4092 return (err); 4093 4094 if (!spa_feature_is_enabled(spa, 4095 SPA_FEATURE_LARGE_BLOCKS)) { 4096 spa_close(spa, FTAG); 4097 return (SET_ERROR(ENOTSUP)); 4098 } 4099 spa_close(spa, FTAG); 4100 } 4101 break; 4102 4103 case ZFS_PROP_SHARESMB: 4104 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID)) 4105 return (SET_ERROR(ENOTSUP)); 4106 break; 4107 4108 case ZFS_PROP_ACLINHERIT: 4109 if (nvpair_type(pair) == DATA_TYPE_UINT64 && 4110 nvpair_value_uint64(pair, &intval) == 0) { 4111 if (intval == ZFS_ACL_PASSTHROUGH_X && 4112 zfs_earlier_version(dsname, 4113 SPA_VERSION_PASSTHROUGH_X)) 4114 return (SET_ERROR(ENOTSUP)); 4115 } 4116 break; 4117 4118 case ZFS_PROP_CHECKSUM: 4119 case ZFS_PROP_DEDUP: 4120 { 4121 spa_feature_t feature; 4122 spa_t *spa; 4123 4124 /* dedup feature version checks */ 4125 if (prop == ZFS_PROP_DEDUP && 4126 zfs_earlier_version(dsname, SPA_VERSION_DEDUP)) 4127 return (SET_ERROR(ENOTSUP)); 4128 4129 if (nvpair_value_uint64(pair, &intval) != 0) 4130 return (SET_ERROR(EINVAL)); 4131 4132 /* check prop value is enabled in features */ 4133 feature = zio_checksum_to_feature(intval & ZIO_CHECKSUM_MASK); 4134 if (feature == SPA_FEATURE_NONE) 4135 break; 4136 4137 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 4138 return (err); 4139 /* 4140 * Salted checksums are not supported on root pools. 4141 */ 4142 if (spa_bootfs(spa) != 0 && 4143 intval < ZIO_CHECKSUM_FUNCTIONS && 4144 (zio_checksum_table[intval].ci_flags & 4145 ZCHECKSUM_FLAG_SALTED)) { 4146 spa_close(spa, FTAG); 4147 return (SET_ERROR(ERANGE)); 4148 } 4149 if (!spa_feature_is_enabled(spa, feature)) { 4150 spa_close(spa, FTAG); 4151 return (SET_ERROR(ENOTSUP)); 4152 } 4153 spa_close(spa, FTAG); 4154 break; 4155 } 4156 } 4157 4158 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED())); 4159} 4160 4161/* 4162 * Checks for a race condition to make sure we don't increment a feature flag 4163 * multiple times. 4164 */ 4165static int 4166zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx) 4167{ 4168 spa_t *spa = dmu_tx_pool(tx)->dp_spa; 4169 spa_feature_t *featurep = arg; 4170 4171 if (!spa_feature_is_active(spa, *featurep)) 4172 return (0); 4173 else 4174 return (SET_ERROR(EBUSY)); 4175} 4176 4177/* 4178 * The callback invoked on feature activation in the sync task caused by 4179 * zfs_prop_activate_feature. 4180 */ 4181static void 4182zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx) 4183{ 4184 spa_t *spa = dmu_tx_pool(tx)->dp_spa; 4185 spa_feature_t *featurep = arg; 4186 4187 spa_feature_incr(spa, *featurep, tx); 4188} 4189 4190/* 4191 * Activates a feature on a pool in response to a property setting. This 4192 * creates a new sync task which modifies the pool to reflect the feature 4193 * as being active. 4194 */ 4195static int 4196zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature) 4197{ 4198 int err; 4199 4200 /* EBUSY here indicates that the feature is already active */ 4201 err = dsl_sync_task(spa_name(spa), 4202 zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync, 4203 &feature, 2, ZFS_SPACE_CHECK_RESERVED); 4204 4205 if (err != 0 && err != EBUSY) 4206 return (err); 4207 else 4208 return (0); 4209} 4210 4211/* 4212 * Removes properties from the given props list that fail permission checks 4213 * needed to clear them and to restore them in case of a receive error. For each 4214 * property, make sure we have both set and inherit permissions. 4215 * 4216 * Returns the first error encountered if any permission checks fail. If the 4217 * caller provides a non-NULL errlist, it also gives the complete list of names 4218 * of all the properties that failed a permission check along with the 4219 * corresponding error numbers. The caller is responsible for freeing the 4220 * returned errlist. 4221 * 4222 * If every property checks out successfully, zero is returned and the list 4223 * pointed at by errlist is NULL. 4224 */ 4225static int 4226zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist) 4227{ 4228 zfs_cmd_t *zc; 4229 nvpair_t *pair, *next_pair; 4230 nvlist_t *errors; 4231 int err, rv = 0; 4232 4233 if (props == NULL) 4234 return (0); 4235 4236 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0); 4237 4238 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP); 4239 (void) strcpy(zc->zc_name, dataset); 4240 pair = nvlist_next_nvpair(props, NULL); 4241 while (pair != NULL) { 4242 next_pair = nvlist_next_nvpair(props, pair); 4243 4244 (void) strcpy(zc->zc_value, nvpair_name(pair)); 4245 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 || 4246 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) { 4247 VERIFY(nvlist_remove_nvpair(props, pair) == 0); 4248 VERIFY(nvlist_add_int32(errors, 4249 zc->zc_value, err) == 0); 4250 } 4251 pair = next_pair; 4252 } 4253 kmem_free(zc, sizeof (zfs_cmd_t)); 4254 4255 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) { 4256 nvlist_free(errors); 4257 errors = NULL; 4258 } else { 4259 VERIFY(nvpair_value_int32(pair, &rv) == 0); 4260 } 4261 4262 if (errlist == NULL) 4263 nvlist_free(errors); 4264 else 4265 *errlist = errors; 4266 4267 return (rv); 4268} 4269 4270static boolean_t 4271propval_equals(nvpair_t *p1, nvpair_t *p2) 4272{ 4273 if (nvpair_type(p1) == DATA_TYPE_NVLIST) { 4274 /* dsl_prop_get_all_impl() format */ 4275 nvlist_t *attrs; 4276 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0); 4277 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 4278 &p1) == 0); 4279 } 4280 4281 if (nvpair_type(p2) == DATA_TYPE_NVLIST) { 4282 nvlist_t *attrs; 4283 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0); 4284 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 4285 &p2) == 0); 4286 } 4287 4288 if (nvpair_type(p1) != nvpair_type(p2)) 4289 return (B_FALSE); 4290 4291 if (nvpair_type(p1) == DATA_TYPE_STRING) { 4292 char *valstr1, *valstr2; 4293 4294 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0); 4295 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0); 4296 return (strcmp(valstr1, valstr2) == 0); 4297 } else { 4298 uint64_t intval1, intval2; 4299 4300 VERIFY(nvpair_value_uint64(p1, &intval1) == 0); 4301 VERIFY(nvpair_value_uint64(p2, &intval2) == 0); 4302 return (intval1 == intval2); 4303 } 4304} 4305 4306/* 4307 * Remove properties from props if they are not going to change (as determined 4308 * by comparison with origprops). Remove them from origprops as well, since we 4309 * do not need to clear or restore properties that won't change. 4310 */ 4311static void 4312props_reduce(nvlist_t *props, nvlist_t *origprops) 4313{ 4314 nvpair_t *pair, *next_pair; 4315 4316 if (origprops == NULL) 4317 return; /* all props need to be received */ 4318 4319 pair = nvlist_next_nvpair(props, NULL); 4320 while (pair != NULL) { 4321 const char *propname = nvpair_name(pair); 4322 nvpair_t *match; 4323 4324 next_pair = nvlist_next_nvpair(props, pair); 4325 4326 if ((nvlist_lookup_nvpair(origprops, propname, 4327 &match) != 0) || !propval_equals(pair, match)) 4328 goto next; /* need to set received value */ 4329 4330 /* don't clear the existing received value */ 4331 (void) nvlist_remove_nvpair(origprops, match); 4332 /* don't bother receiving the property */ 4333 (void) nvlist_remove_nvpair(props, pair); 4334next: 4335 pair = next_pair; 4336 } 4337} 4338 4339/* 4340 * Extract properties that cannot be set PRIOR to the receipt of a dataset. 4341 * For example, refquota cannot be set until after the receipt of a dataset, 4342 * because in replication streams, an older/earlier snapshot may exceed the 4343 * refquota. We want to receive the older/earlier snapshot, but setting 4344 * refquota pre-receipt will set the dsl's ACTUAL quota, which will prevent 4345 * the older/earlier snapshot from being received (with EDQUOT). 4346 * 4347 * The ZFS test "zfs_receive_011_pos" demonstrates such a scenario. 4348 * 4349 * libzfs will need to be judicious handling errors encountered by props 4350 * extracted by this function. 4351 */ 4352static nvlist_t * 4353extract_delay_props(nvlist_t *props) 4354{ 4355 nvlist_t *delayprops; 4356 nvpair_t *nvp, *tmp; 4357 static const zfs_prop_t delayable[] = { ZFS_PROP_REFQUOTA, 0 }; 4358 int i; 4359 4360 VERIFY(nvlist_alloc(&delayprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 4361 4362 for (nvp = nvlist_next_nvpair(props, NULL); nvp != NULL; 4363 nvp = nvlist_next_nvpair(props, nvp)) { 4364 /* 4365 * strcmp() is safe because zfs_prop_to_name() always returns 4366 * a bounded string. 4367 */ 4368 for (i = 0; delayable[i] != 0; i++) { 4369 if (strcmp(zfs_prop_to_name(delayable[i]), 4370 nvpair_name(nvp)) == 0) { 4371 break; 4372 } 4373 } 4374 if (delayable[i] != 0) { 4375 tmp = nvlist_prev_nvpair(props, nvp); 4376 VERIFY(nvlist_add_nvpair(delayprops, nvp) == 0); 4377 VERIFY(nvlist_remove_nvpair(props, nvp) == 0); 4378 nvp = tmp; 4379 } 4380 } 4381 4382 if (nvlist_empty(delayprops)) { 4383 nvlist_free(delayprops); 4384 delayprops = NULL; 4385 } 4386 return (delayprops); 4387} 4388 4389#ifdef DEBUG 4390static boolean_t zfs_ioc_recv_inject_err; 4391#endif 4392 4393/* 4394 * inputs: 4395 * zc_name name of containing filesystem 4396 * zc_nvlist_src{_size} nvlist of properties to apply 4397 * zc_value name of snapshot to create 4398 * zc_string name of clone origin (if DRR_FLAG_CLONE) 4399 * zc_cookie file descriptor to recv from 4400 * zc_begin_record the BEGIN record of the stream (not byteswapped) 4401 * zc_guid force flag 4402 * zc_cleanup_fd cleanup-on-exit file descriptor 4403 * zc_action_handle handle for this guid/ds mapping (or zero on first call) 4404 * zc_resumable if data is incomplete assume sender will resume 4405 * 4406 * outputs: 4407 * zc_cookie number of bytes read 4408 * zc_nvlist_dst{_size} error for each unapplied received property 4409 * zc_obj zprop_errflags_t 4410 * zc_action_handle handle for this guid/ds mapping 4411 */ 4412static int 4413zfs_ioc_recv(zfs_cmd_t *zc) 4414{ 4415 file_t *fp; 4416 dmu_recv_cookie_t drc; 4417 boolean_t force = (boolean_t)zc->zc_guid; 4418 int fd; 4419 int error = 0; 4420 int props_error = 0; 4421 nvlist_t *errors; 4422 offset_t off; 4423 nvlist_t *props = NULL; /* sent properties */ 4424 nvlist_t *origprops = NULL; /* existing properties */ 4425 nvlist_t *delayprops = NULL; /* sent properties applied post-receive */ 4426 char *origin = NULL; 4427 char *tosnap; 4428 char tofs[ZFS_MAX_DATASET_NAME_LEN]; 4429 cap_rights_t rights; 4430 boolean_t first_recvd_props = B_FALSE; 4431 4432 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 4433 strchr(zc->zc_value, '@') == NULL || 4434 strchr(zc->zc_value, '%')) 4435 return (SET_ERROR(EINVAL)); 4436 4437 (void) strcpy(tofs, zc->zc_value); 4438 tosnap = strchr(tofs, '@'); 4439 *tosnap++ = '\0'; 4440 4441 if (zc->zc_nvlist_src != 0 && 4442 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 4443 zc->zc_iflags, &props)) != 0) 4444 return (error); 4445 4446 fd = zc->zc_cookie; 4447#ifdef illumos 4448 fp = getf(fd); 4449#else 4450 fget_read(curthread, fd, cap_rights_init(&rights, CAP_PREAD), &fp); 4451#endif 4452 if (fp == NULL) { 4453 nvlist_free(props); 4454 return (SET_ERROR(EBADF)); 4455 } 4456 4457 errors = fnvlist_alloc(); 4458 4459 if (zc->zc_string[0]) 4460 origin = zc->zc_string; 4461 4462 error = dmu_recv_begin(tofs, tosnap, 4463 &zc->zc_begin_record, force, zc->zc_resumable, origin, &drc); 4464 if (error != 0) 4465 goto out; 4466 4467 /* 4468 * Set properties before we receive the stream so that they are applied 4469 * to the new data. Note that we must call dmu_recv_stream() if 4470 * dmu_recv_begin() succeeds. 4471 */ 4472 if (props != NULL && !drc.drc_newfs) { 4473 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >= 4474 SPA_VERSION_RECVD_PROPS && 4475 !dsl_prop_get_hasrecvd(tofs)) 4476 first_recvd_props = B_TRUE; 4477 4478 /* 4479 * If new received properties are supplied, they are to 4480 * completely replace the existing received properties, so stash 4481 * away the existing ones. 4482 */ 4483 if (dsl_prop_get_received(tofs, &origprops) == 0) { 4484 nvlist_t *errlist = NULL; 4485 /* 4486 * Don't bother writing a property if its value won't 4487 * change (and avoid the unnecessary security checks). 4488 * 4489 * The first receive after SPA_VERSION_RECVD_PROPS is a 4490 * special case where we blow away all local properties 4491 * regardless. 4492 */ 4493 if (!first_recvd_props) 4494 props_reduce(props, origprops); 4495 if (zfs_check_clearable(tofs, origprops, &errlist) != 0) 4496 (void) nvlist_merge(errors, errlist, 0); 4497 nvlist_free(errlist); 4498 4499 if (clear_received_props(tofs, origprops, 4500 first_recvd_props ? NULL : props) != 0) 4501 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 4502 } else { 4503 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 4504 } 4505 } 4506 4507 if (props != NULL) { 4508 props_error = dsl_prop_set_hasrecvd(tofs); 4509 4510 if (props_error == 0) { 4511 delayprops = extract_delay_props(props); 4512 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED, 4513 props, errors); 4514 } 4515 } 4516 4517 off = fp->f_offset; 4518 error = dmu_recv_stream(&drc, fp, &off, zc->zc_cleanup_fd, 4519 &zc->zc_action_handle); 4520 4521 if (error == 0) { 4522 zfsvfs_t *zfsvfs = NULL; 4523 4524 if (getzfsvfs(tofs, &zfsvfs) == 0) { 4525 /* online recv */ 4526 dsl_dataset_t *ds; 4527 int end_err; 4528 4529 ds = dmu_objset_ds(zfsvfs->z_os); 4530 error = zfs_suspend_fs(zfsvfs); 4531 /* 4532 * If the suspend fails, then the recv_end will 4533 * likely also fail, and clean up after itself. 4534 */ 4535 end_err = dmu_recv_end(&drc, zfsvfs); 4536 if (error == 0) 4537 error = zfs_resume_fs(zfsvfs, ds); 4538 error = error ? error : end_err; 4539#ifdef illumos 4540 VFS_RELE(zfsvfs->z_vfs); 4541#else 4542 vfs_unbusy(zfsvfs->z_vfs); 4543#endif 4544 } else { 4545 error = dmu_recv_end(&drc, NULL); 4546 } 4547 4548 /* Set delayed properties now, after we're done receiving. */ 4549 if (delayprops != NULL && error == 0) { 4550 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED, 4551 delayprops, errors); 4552 } 4553 } 4554 4555 if (delayprops != NULL) { 4556 /* 4557 * Merge delayed props back in with initial props, in case 4558 * we're DEBUG and zfs_ioc_recv_inject_err is set (which means 4559 * we have to make sure clear_received_props() includes 4560 * the delayed properties). 4561 * 4562 * Since zfs_ioc_recv_inject_err is only in DEBUG kernels, 4563 * using ASSERT() will be just like a VERIFY. 4564 */ 4565 ASSERT(nvlist_merge(props, delayprops, 0) == 0); 4566 nvlist_free(delayprops); 4567 } 4568 4569 /* 4570 * Now that all props, initial and delayed, are set, report the prop 4571 * errors to the caller. 4572 */ 4573 if (zc->zc_nvlist_dst_size != 0 && 4574 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 || 4575 put_nvlist(zc, errors) != 0)) { 4576 /* 4577 * Caller made zc->zc_nvlist_dst less than the minimum expected 4578 * size or supplied an invalid address. 4579 */ 4580 props_error = SET_ERROR(EINVAL); 4581 } 4582 4583 zc->zc_cookie = off - fp->f_offset; 4584 if (off >= 0 && off <= MAXOFFSET_T) 4585 fp->f_offset = off; 4586 4587#ifdef DEBUG 4588 if (zfs_ioc_recv_inject_err) { 4589 zfs_ioc_recv_inject_err = B_FALSE; 4590 error = 1; 4591 } 4592#endif 4593 4594#ifdef __FreeBSD__ 4595 if (error == 0) 4596 zvol_create_minors(tofs); 4597#endif 4598 4599 /* 4600 * On error, restore the original props. 4601 */ 4602 if (error != 0 && props != NULL && !drc.drc_newfs) { 4603 if (clear_received_props(tofs, props, NULL) != 0) { 4604 /* 4605 * We failed to clear the received properties. 4606 * Since we may have left a $recvd value on the 4607 * system, we can't clear the $hasrecvd flag. 4608 */ 4609 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4610 } else if (first_recvd_props) { 4611 dsl_prop_unset_hasrecvd(tofs); 4612 } 4613 4614 if (origprops == NULL && !drc.drc_newfs) { 4615 /* We failed to stash the original properties. */ 4616 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4617 } 4618 4619 /* 4620 * dsl_props_set() will not convert RECEIVED to LOCAL on or 4621 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL 4622 * explictly if we're restoring local properties cleared in the 4623 * first new-style receive. 4624 */ 4625 if (origprops != NULL && 4626 zfs_set_prop_nvlist(tofs, (first_recvd_props ? 4627 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED), 4628 origprops, NULL) != 0) { 4629 /* 4630 * We stashed the original properties but failed to 4631 * restore them. 4632 */ 4633 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4634 } 4635 } 4636out: 4637 nvlist_free(props); 4638 nvlist_free(origprops); 4639 nvlist_free(errors); 4640 releasef(fd); 4641 4642 if (error == 0) 4643 error = props_error; 4644 4645 return (error); 4646} 4647 4648/* 4649 * inputs: 4650 * zc_name name of snapshot to send 4651 * zc_cookie file descriptor to send stream to 4652 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj) 4653 * zc_sendobj objsetid of snapshot to send 4654 * zc_fromobj objsetid of incremental fromsnap (may be zero) 4655 * zc_guid if set, estimate size of stream only. zc_cookie is ignored. 4656 * output size in zc_objset_type. 4657 * zc_flags lzc_send_flags 4658 * 4659 * outputs: 4660 * zc_objset_type estimated size, if zc_guid is set 4661 */ 4662static int 4663zfs_ioc_send(zfs_cmd_t *zc) 4664{ 4665 int error; 4666 offset_t off; 4667 boolean_t estimate = (zc->zc_guid != 0); 4668 boolean_t embedok = (zc->zc_flags & 0x1); 4669 boolean_t large_block_ok = (zc->zc_flags & 0x2); 4670 boolean_t compressok = (zc->zc_flags & 0x4); 4671 4672 if (zc->zc_obj != 0) { 4673 dsl_pool_t *dp; 4674 dsl_dataset_t *tosnap; 4675 4676 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4677 if (error != 0) 4678 return (error); 4679 4680 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap); 4681 if (error != 0) { 4682 dsl_pool_rele(dp, FTAG); 4683 return (error); 4684 } 4685 4686 if (dsl_dir_is_clone(tosnap->ds_dir)) 4687 zc->zc_fromobj = 4688 dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj; 4689 dsl_dataset_rele(tosnap, FTAG); 4690 dsl_pool_rele(dp, FTAG); 4691 } 4692 4693 if (estimate) { 4694 dsl_pool_t *dp; 4695 dsl_dataset_t *tosnap; 4696 dsl_dataset_t *fromsnap = NULL; 4697 4698 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4699 if (error != 0) 4700 return (error); 4701 4702 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap); 4703 if (error != 0) { 4704 dsl_pool_rele(dp, FTAG); 4705 return (error); 4706 } 4707 4708 if (zc->zc_fromobj != 0) { 4709 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj, 4710 FTAG, &fromsnap); 4711 if (error != 0) { 4712 dsl_dataset_rele(tosnap, FTAG); 4713 dsl_pool_rele(dp, FTAG); 4714 return (error); 4715 } 4716 } 4717 4718 error = dmu_send_estimate(tosnap, fromsnap, compressok, 4719 &zc->zc_objset_type); 4720 4721 if (fromsnap != NULL) 4722 dsl_dataset_rele(fromsnap, FTAG); 4723 dsl_dataset_rele(tosnap, FTAG); 4724 dsl_pool_rele(dp, FTAG); 4725 } else { 4726 file_t *fp; 4727 cap_rights_t rights; 4728 4729#ifdef illumos 4730 fp = getf(zc->zc_cookie); 4731#else 4732 fget_write(curthread, zc->zc_cookie, 4733 cap_rights_init(&rights, CAP_WRITE), &fp); 4734#endif 4735 if (fp == NULL) 4736 return (SET_ERROR(EBADF)); 4737 4738 off = fp->f_offset; 4739 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj, 4740 zc->zc_fromobj, embedok, large_block_ok, compressok, 4741#ifdef illumos 4742 zc->zc_cookie, fp->f_vnode, &off); 4743#else 4744 zc->zc_cookie, fp, &off); 4745#endif 4746 4747 if (off >= 0 && off <= MAXOFFSET_T) 4748 fp->f_offset = off; 4749 releasef(zc->zc_cookie); 4750 } 4751 return (error); 4752} 4753 4754/* 4755 * inputs: 4756 * zc_name name of snapshot on which to report progress 4757 * zc_cookie file descriptor of send stream 4758 * 4759 * outputs: 4760 * zc_cookie number of bytes written in send stream thus far 4761 */ 4762static int 4763zfs_ioc_send_progress(zfs_cmd_t *zc) 4764{ 4765 dsl_pool_t *dp; 4766 dsl_dataset_t *ds; 4767 dmu_sendarg_t *dsp = NULL; 4768 int error; 4769 4770 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4771 if (error != 0) 4772 return (error); 4773 4774 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds); 4775 if (error != 0) { 4776 dsl_pool_rele(dp, FTAG); 4777 return (error); 4778 } 4779 4780 mutex_enter(&ds->ds_sendstream_lock); 4781 4782 /* 4783 * Iterate over all the send streams currently active on this dataset. 4784 * If there's one which matches the specified file descriptor _and_ the 4785 * stream was started by the current process, return the progress of 4786 * that stream. 4787 */ 4788 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL; 4789 dsp = list_next(&ds->ds_sendstreams, dsp)) { 4790 if (dsp->dsa_outfd == zc->zc_cookie && 4791 dsp->dsa_proc == curproc) 4792 break; 4793 } 4794 4795 if (dsp != NULL) 4796 zc->zc_cookie = *(dsp->dsa_off); 4797 else 4798 error = SET_ERROR(ENOENT); 4799 4800 mutex_exit(&ds->ds_sendstream_lock); 4801 dsl_dataset_rele(ds, FTAG); 4802 dsl_pool_rele(dp, FTAG); 4803 return (error); 4804} 4805 4806static int 4807zfs_ioc_inject_fault(zfs_cmd_t *zc) 4808{ 4809 int id, error; 4810 4811 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id, 4812 &zc->zc_inject_record); 4813 4814 if (error == 0) 4815 zc->zc_guid = (uint64_t)id; 4816 4817 return (error); 4818} 4819 4820static int 4821zfs_ioc_clear_fault(zfs_cmd_t *zc) 4822{ 4823 return (zio_clear_fault((int)zc->zc_guid)); 4824} 4825 4826static int 4827zfs_ioc_inject_list_next(zfs_cmd_t *zc) 4828{ 4829 int id = (int)zc->zc_guid; 4830 int error; 4831 4832 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name), 4833 &zc->zc_inject_record); 4834 4835 zc->zc_guid = id; 4836 4837 return (error); 4838} 4839 4840static int 4841zfs_ioc_error_log(zfs_cmd_t *zc) 4842{ 4843 spa_t *spa; 4844 int error; 4845 size_t count = (size_t)zc->zc_nvlist_dst_size; 4846 4847 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 4848 return (error); 4849 4850 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst, 4851 &count); 4852 if (error == 0) 4853 zc->zc_nvlist_dst_size = count; 4854 else 4855 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa); 4856 4857 spa_close(spa, FTAG); 4858 4859 return (error); 4860} 4861 4862static int 4863zfs_ioc_clear(zfs_cmd_t *zc) 4864{ 4865 spa_t *spa; 4866 vdev_t *vd; 4867 int error; 4868 4869 /* 4870 * On zpool clear we also fix up missing slogs 4871 */ 4872 mutex_enter(&spa_namespace_lock); 4873 spa = spa_lookup(zc->zc_name); 4874 if (spa == NULL) { 4875 mutex_exit(&spa_namespace_lock); 4876 return (SET_ERROR(EIO)); 4877 } 4878 if (spa_get_log_state(spa) == SPA_LOG_MISSING) { 4879 /* we need to let spa_open/spa_load clear the chains */ 4880 spa_set_log_state(spa, SPA_LOG_CLEAR); 4881 } 4882 spa->spa_last_open_failed = 0; 4883 mutex_exit(&spa_namespace_lock); 4884 4885 if (zc->zc_cookie & ZPOOL_NO_REWIND) { 4886 error = spa_open(zc->zc_name, &spa, FTAG); 4887 } else { 4888 nvlist_t *policy; 4889 nvlist_t *config = NULL; 4890 4891 if (zc->zc_nvlist_src == 0) 4892 return (SET_ERROR(EINVAL)); 4893 4894 if ((error = get_nvlist(zc->zc_nvlist_src, 4895 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) { 4896 error = spa_open_rewind(zc->zc_name, &spa, FTAG, 4897 policy, &config); 4898 if (config != NULL) { 4899 int err; 4900 4901 if ((err = put_nvlist(zc, config)) != 0) 4902 error = err; 4903 nvlist_free(config); 4904 } 4905 nvlist_free(policy); 4906 } 4907 } 4908 4909 if (error != 0) 4910 return (error); 4911 4912 spa_vdev_state_enter(spa, SCL_NONE); 4913 4914 if (zc->zc_guid == 0) { 4915 vd = NULL; 4916 } else { 4917 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE); 4918 if (vd == NULL) { 4919 (void) spa_vdev_state_exit(spa, NULL, ENODEV); 4920 spa_close(spa, FTAG); 4921 return (SET_ERROR(ENODEV)); 4922 } 4923 } 4924 4925 vdev_clear(spa, vd); 4926 4927 (void) spa_vdev_state_exit(spa, NULL, 0); 4928 4929 /* 4930 * Resume any suspended I/Os. 4931 */ 4932 if (zio_resume(spa) != 0) 4933 error = SET_ERROR(EIO); 4934 4935 spa_close(spa, FTAG); 4936 4937 return (error); 4938} 4939 4940static int 4941zfs_ioc_pool_reopen(zfs_cmd_t *zc) 4942{ 4943 spa_t *spa; 4944 int error; 4945 4946 error = spa_open(zc->zc_name, &spa, FTAG); 4947 if (error != 0) 4948 return (error); 4949 4950 spa_vdev_state_enter(spa, SCL_NONE); 4951 4952 /* 4953 * If a resilver is already in progress then set the 4954 * spa_scrub_reopen flag to B_TRUE so that we don't restart 4955 * the scan as a side effect of the reopen. Otherwise, let 4956 * vdev_open() decided if a resilver is required. 4957 */ 4958 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool); 4959 vdev_reopen(spa->spa_root_vdev); 4960 spa->spa_scrub_reopen = B_FALSE; 4961 4962 (void) spa_vdev_state_exit(spa, NULL, 0); 4963 spa_close(spa, FTAG); 4964 return (0); 4965} 4966/* 4967 * inputs: 4968 * zc_name name of filesystem 4969 * 4970 * outputs: 4971 * zc_string name of conflicting snapshot, if there is one 4972 */ 4973static int 4974zfs_ioc_promote(zfs_cmd_t *zc) 4975{ 4976 dsl_pool_t *dp; 4977 dsl_dataset_t *ds, *ods; 4978 char origin[ZFS_MAX_DATASET_NAME_LEN]; 4979 char *cp; 4980 int error; 4981 4982 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4983 if (error != 0) 4984 return (error); 4985 4986 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds); 4987 if (error != 0) { 4988 dsl_pool_rele(dp, FTAG); 4989 return (error); 4990 } 4991 4992 if (!dsl_dir_is_clone(ds->ds_dir)) { 4993 dsl_dataset_rele(ds, FTAG); 4994 dsl_pool_rele(dp, FTAG); 4995 return (SET_ERROR(EINVAL)); 4996 } 4997 4998 error = dsl_dataset_hold_obj(dp, 4999 dsl_dir_phys(ds->ds_dir)->dd_origin_obj, FTAG, &ods); 5000 if (error != 0) { 5001 dsl_dataset_rele(ds, FTAG); 5002 dsl_pool_rele(dp, FTAG); 5003 return (error); 5004 } 5005 5006 dsl_dataset_name(ods, origin); 5007 dsl_dataset_rele(ods, FTAG); 5008 dsl_dataset_rele(ds, FTAG); 5009 dsl_pool_rele(dp, FTAG); 5010 5011 /* 5012 * We don't need to unmount *all* the origin fs's snapshots, but 5013 * it's easier. 5014 */ 5015 cp = strchr(origin, '@'); 5016 if (cp) 5017 *cp = '\0'; 5018 (void) dmu_objset_find(origin, 5019 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS); 5020 return (dsl_dataset_promote(zc->zc_name, zc->zc_string)); 5021} 5022 5023/* 5024 * Retrieve a single {user|group}{used|quota}@... property. 5025 * 5026 * inputs: 5027 * zc_name name of filesystem 5028 * zc_objset_type zfs_userquota_prop_t 5029 * zc_value domain name (eg. "S-1-234-567-89") 5030 * zc_guid RID/UID/GID 5031 * 5032 * outputs: 5033 * zc_cookie property value 5034 */ 5035static int 5036zfs_ioc_userspace_one(zfs_cmd_t *zc) 5037{ 5038 zfsvfs_t *zfsvfs; 5039 int error; 5040 5041 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 5042 return (SET_ERROR(EINVAL)); 5043 5044 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE); 5045 if (error != 0) 5046 return (error); 5047 5048 error = zfs_userspace_one(zfsvfs, 5049 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie); 5050 zfsvfs_rele(zfsvfs, FTAG); 5051 5052 return (error); 5053} 5054 5055/* 5056 * inputs: 5057 * zc_name name of filesystem 5058 * zc_cookie zap cursor 5059 * zc_objset_type zfs_userquota_prop_t 5060 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist) 5061 * 5062 * outputs: 5063 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t) 5064 * zc_cookie zap cursor 5065 */ 5066static int 5067zfs_ioc_userspace_many(zfs_cmd_t *zc) 5068{ 5069 zfsvfs_t *zfsvfs; 5070 int bufsize = zc->zc_nvlist_dst_size; 5071 5072 if (bufsize <= 0) 5073 return (SET_ERROR(ENOMEM)); 5074 5075 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE); 5076 if (error != 0) 5077 return (error); 5078 5079 void *buf = kmem_alloc(bufsize, KM_SLEEP); 5080 5081 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie, 5082 buf, &zc->zc_nvlist_dst_size); 5083 5084 if (error == 0) { 5085 error = ddi_copyout(buf, 5086 (void *)(uintptr_t)zc->zc_nvlist_dst, 5087 zc->zc_nvlist_dst_size, zc->zc_iflags); 5088 } 5089 kmem_free(buf, bufsize); 5090 zfsvfs_rele(zfsvfs, FTAG); 5091 5092 return (error); 5093} 5094 5095/* 5096 * inputs: 5097 * zc_name name of filesystem 5098 * 5099 * outputs: 5100 * none 5101 */ 5102static int 5103zfs_ioc_userspace_upgrade(zfs_cmd_t *zc) 5104{ 5105 objset_t *os; 5106 int error = 0; 5107 zfsvfs_t *zfsvfs; 5108 5109 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) { 5110 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) { 5111 /* 5112 * If userused is not enabled, it may be because the 5113 * objset needs to be closed & reopened (to grow the 5114 * objset_phys_t). Suspend/resume the fs will do that. 5115 */ 5116 dsl_dataset_t *ds; 5117 5118 ds = dmu_objset_ds(zfsvfs->z_os); 5119 error = zfs_suspend_fs(zfsvfs); 5120 if (error == 0) { 5121 dmu_objset_refresh_ownership(zfsvfs->z_os, 5122 zfsvfs); 5123 error = zfs_resume_fs(zfsvfs, ds); 5124 } 5125 } 5126 if (error == 0) 5127 error = dmu_objset_userspace_upgrade(zfsvfs->z_os); 5128#ifdef illumos 5129 VFS_RELE(zfsvfs->z_vfs); 5130#else 5131 vfs_unbusy(zfsvfs->z_vfs); 5132#endif 5133 } else { 5134 /* XXX kind of reading contents without owning */ 5135 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 5136 if (error != 0) 5137 return (error); 5138 5139 error = dmu_objset_userspace_upgrade(os); 5140 dmu_objset_rele(os, FTAG); 5141 } 5142 5143 return (error); 5144} 5145 5146#ifdef illumos 5147/* 5148 * We don't want to have a hard dependency 5149 * against some special symbols in sharefs 5150 * nfs, and smbsrv. Determine them if needed when 5151 * the first file system is shared. 5152 * Neither sharefs, nfs or smbsrv are unloadable modules. 5153 */ 5154int (*znfsexport_fs)(void *arg); 5155int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t); 5156int (*zsmbexport_fs)(void *arg, boolean_t add_share); 5157 5158int zfs_nfsshare_inited; 5159int zfs_smbshare_inited; 5160 5161ddi_modhandle_t nfs_mod; 5162ddi_modhandle_t sharefs_mod; 5163ddi_modhandle_t smbsrv_mod; 5164#endif /* illumos */ 5165kmutex_t zfs_share_lock; 5166 5167#ifdef illumos 5168static int 5169zfs_init_sharefs() 5170{ 5171 int error; 5172 5173 ASSERT(MUTEX_HELD(&zfs_share_lock)); 5174 /* Both NFS and SMB shares also require sharetab support. */ 5175 if (sharefs_mod == NULL && ((sharefs_mod = 5176 ddi_modopen("fs/sharefs", 5177 KRTLD_MODE_FIRST, &error)) == NULL)) { 5178 return (SET_ERROR(ENOSYS)); 5179 } 5180 if (zshare_fs == NULL && ((zshare_fs = 5181 (int (*)(enum sharefs_sys_op, share_t *, uint32_t)) 5182 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) { 5183 return (SET_ERROR(ENOSYS)); 5184 } 5185 return (0); 5186} 5187#endif /* illumos */ 5188 5189static int 5190zfs_ioc_share(zfs_cmd_t *zc) 5191{ 5192#ifdef illumos 5193 int error; 5194 int opcode; 5195 5196 switch (zc->zc_share.z_sharetype) { 5197 case ZFS_SHARE_NFS: 5198 case ZFS_UNSHARE_NFS: 5199 if (zfs_nfsshare_inited == 0) { 5200 mutex_enter(&zfs_share_lock); 5201 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs", 5202 KRTLD_MODE_FIRST, &error)) == NULL)) { 5203 mutex_exit(&zfs_share_lock); 5204 return (SET_ERROR(ENOSYS)); 5205 } 5206 if (znfsexport_fs == NULL && 5207 ((znfsexport_fs = (int (*)(void *)) 5208 ddi_modsym(nfs_mod, 5209 "nfs_export", &error)) == NULL)) { 5210 mutex_exit(&zfs_share_lock); 5211 return (SET_ERROR(ENOSYS)); 5212 } 5213 error = zfs_init_sharefs(); 5214 if (error != 0) { 5215 mutex_exit(&zfs_share_lock); 5216 return (SET_ERROR(ENOSYS)); 5217 } 5218 zfs_nfsshare_inited = 1; 5219 mutex_exit(&zfs_share_lock); 5220 } 5221 break; 5222 case ZFS_SHARE_SMB: 5223 case ZFS_UNSHARE_SMB: 5224 if (zfs_smbshare_inited == 0) { 5225 mutex_enter(&zfs_share_lock); 5226 if (smbsrv_mod == NULL && ((smbsrv_mod = 5227 ddi_modopen("drv/smbsrv", 5228 KRTLD_MODE_FIRST, &error)) == NULL)) { 5229 mutex_exit(&zfs_share_lock); 5230 return (SET_ERROR(ENOSYS)); 5231 } 5232 if (zsmbexport_fs == NULL && ((zsmbexport_fs = 5233 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod, 5234 "smb_server_share", &error)) == NULL)) { 5235 mutex_exit(&zfs_share_lock); 5236 return (SET_ERROR(ENOSYS)); 5237 } 5238 error = zfs_init_sharefs(); 5239 if (error != 0) { 5240 mutex_exit(&zfs_share_lock); 5241 return (SET_ERROR(ENOSYS)); 5242 } 5243 zfs_smbshare_inited = 1; 5244 mutex_exit(&zfs_share_lock); 5245 } 5246 break; 5247 default: 5248 return (SET_ERROR(EINVAL)); 5249 } 5250 5251 switch (zc->zc_share.z_sharetype) { 5252 case ZFS_SHARE_NFS: 5253 case ZFS_UNSHARE_NFS: 5254 if (error = 5255 znfsexport_fs((void *) 5256 (uintptr_t)zc->zc_share.z_exportdata)) 5257 return (error); 5258 break; 5259 case ZFS_SHARE_SMB: 5260 case ZFS_UNSHARE_SMB: 5261 if (error = zsmbexport_fs((void *) 5262 (uintptr_t)zc->zc_share.z_exportdata, 5263 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ? 5264 B_TRUE: B_FALSE)) { 5265 return (error); 5266 } 5267 break; 5268 } 5269 5270 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS || 5271 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ? 5272 SHAREFS_ADD : SHAREFS_REMOVE; 5273 5274 /* 5275 * Add or remove share from sharetab 5276 */ 5277 error = zshare_fs(opcode, 5278 (void *)(uintptr_t)zc->zc_share.z_sharedata, 5279 zc->zc_share.z_sharemax); 5280 5281 return (error); 5282 5283#else /* !illumos */ 5284 return (ENOSYS); 5285#endif /* illumos */ 5286} 5287 5288ace_t full_access[] = { 5289 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0} 5290}; 5291 5292/* 5293 * inputs: 5294 * zc_name name of containing filesystem 5295 * zc_obj object # beyond which we want next in-use object # 5296 * 5297 * outputs: 5298 * zc_obj next in-use object # 5299 */ 5300static int 5301zfs_ioc_next_obj(zfs_cmd_t *zc) 5302{ 5303 objset_t *os = NULL; 5304 int error; 5305 5306 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 5307 if (error != 0) 5308 return (error); 5309 5310 error = dmu_object_next(os, &zc->zc_obj, B_FALSE, 5311 dsl_dataset_phys(os->os_dsl_dataset)->ds_prev_snap_txg); 5312 5313 dmu_objset_rele(os, FTAG); 5314 return (error); 5315} 5316 5317/* 5318 * inputs: 5319 * zc_name name of filesystem 5320 * zc_value prefix name for snapshot 5321 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process 5322 * 5323 * outputs: 5324 * zc_value short name of new snapshot 5325 */ 5326static int 5327zfs_ioc_tmp_snapshot(zfs_cmd_t *zc) 5328{ 5329 char *snap_name; 5330 char *hold_name; 5331 int error; 5332 minor_t minor; 5333 5334 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor); 5335 if (error != 0) 5336 return (error); 5337 5338 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value, 5339 (u_longlong_t)ddi_get_lbolt64()); 5340 hold_name = kmem_asprintf("%%%s", zc->zc_value); 5341 5342 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor, 5343 hold_name); 5344 if (error == 0) 5345 (void) strcpy(zc->zc_value, snap_name); 5346 strfree(snap_name); 5347 strfree(hold_name); 5348 zfs_onexit_fd_rele(zc->zc_cleanup_fd); 5349 return (error); 5350} 5351 5352/* 5353 * inputs: 5354 * zc_name name of "to" snapshot 5355 * zc_value name of "from" snapshot 5356 * zc_cookie file descriptor to write diff data on 5357 * 5358 * outputs: 5359 * dmu_diff_record_t's to the file descriptor 5360 */ 5361static int 5362zfs_ioc_diff(zfs_cmd_t *zc) 5363{ 5364 file_t *fp; 5365 cap_rights_t rights; 5366 offset_t off; 5367 int error; 5368 5369#ifdef illumos 5370 fp = getf(zc->zc_cookie); 5371#else 5372 fget_write(curthread, zc->zc_cookie, 5373 cap_rights_init(&rights, CAP_WRITE), &fp); 5374#endif 5375 if (fp == NULL) 5376 return (SET_ERROR(EBADF)); 5377 5378 off = fp->f_offset; 5379 5380#ifdef illumos 5381 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off); 5382#else 5383 error = dmu_diff(zc->zc_name, zc->zc_value, fp, &off); 5384#endif 5385 5386 if (off >= 0 && off <= MAXOFFSET_T) 5387 fp->f_offset = off; 5388 releasef(zc->zc_cookie); 5389 5390 return (error); 5391} 5392 5393#ifdef illumos 5394/* 5395 * Remove all ACL files in shares dir 5396 */ 5397static int 5398zfs_smb_acl_purge(znode_t *dzp) 5399{ 5400 zap_cursor_t zc; 5401 zap_attribute_t zap; 5402 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 5403 int error; 5404 5405 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id); 5406 (error = zap_cursor_retrieve(&zc, &zap)) == 0; 5407 zap_cursor_advance(&zc)) { 5408 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred, 5409 NULL, 0)) != 0) 5410 break; 5411 } 5412 zap_cursor_fini(&zc); 5413 return (error); 5414} 5415#endif /* illumos */ 5416 5417static int 5418zfs_ioc_smb_acl(zfs_cmd_t *zc) 5419{ 5420#ifdef illumos 5421 vnode_t *vp; 5422 znode_t *dzp; 5423 vnode_t *resourcevp = NULL; 5424 znode_t *sharedir; 5425 zfsvfs_t *zfsvfs; 5426 nvlist_t *nvlist; 5427 char *src, *target; 5428 vattr_t vattr; 5429 vsecattr_t vsec; 5430 int error = 0; 5431 5432 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, 5433 NO_FOLLOW, NULL, &vp)) != 0) 5434 return (error); 5435 5436 /* Now make sure mntpnt and dataset are ZFS */ 5437 5438 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 || 5439 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), 5440 zc->zc_name) != 0)) { 5441 VN_RELE(vp); 5442 return (SET_ERROR(EINVAL)); 5443 } 5444 5445 dzp = VTOZ(vp); 5446 zfsvfs = dzp->z_zfsvfs; 5447 ZFS_ENTER(zfsvfs); 5448 5449 /* 5450 * Create share dir if its missing. 5451 */ 5452 mutex_enter(&zfsvfs->z_lock); 5453 if (zfsvfs->z_shares_dir == 0) { 5454 dmu_tx_t *tx; 5455 5456 tx = dmu_tx_create(zfsvfs->z_os); 5457 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE, 5458 ZFS_SHARES_DIR); 5459 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); 5460 error = dmu_tx_assign(tx, TXG_WAIT); 5461 if (error != 0) { 5462 dmu_tx_abort(tx); 5463 } else { 5464 error = zfs_create_share_dir(zfsvfs, tx); 5465 dmu_tx_commit(tx); 5466 } 5467 if (error != 0) { 5468 mutex_exit(&zfsvfs->z_lock); 5469 VN_RELE(vp); 5470 ZFS_EXIT(zfsvfs); 5471 return (error); 5472 } 5473 } 5474 mutex_exit(&zfsvfs->z_lock); 5475 5476 ASSERT(zfsvfs->z_shares_dir); 5477 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) { 5478 VN_RELE(vp); 5479 ZFS_EXIT(zfsvfs); 5480 return (error); 5481 } 5482 5483 switch (zc->zc_cookie) { 5484 case ZFS_SMB_ACL_ADD: 5485 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; 5486 vattr.va_type = VREG; 5487 vattr.va_mode = S_IFREG|0777; 5488 vattr.va_uid = 0; 5489 vattr.va_gid = 0; 5490 5491 vsec.vsa_mask = VSA_ACE; 5492 vsec.vsa_aclentp = &full_access; 5493 vsec.vsa_aclentsz = sizeof (full_access); 5494 vsec.vsa_aclcnt = 1; 5495 5496 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string, 5497 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec); 5498 if (resourcevp) 5499 VN_RELE(resourcevp); 5500 break; 5501 5502 case ZFS_SMB_ACL_REMOVE: 5503 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred, 5504 NULL, 0); 5505 break; 5506 5507 case ZFS_SMB_ACL_RENAME: 5508 if ((error = get_nvlist(zc->zc_nvlist_src, 5509 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) { 5510 VN_RELE(vp); 5511 VN_RELE(ZTOV(sharedir)); 5512 ZFS_EXIT(zfsvfs); 5513 return (error); 5514 } 5515 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) || 5516 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET, 5517 &target)) { 5518 VN_RELE(vp); 5519 VN_RELE(ZTOV(sharedir)); 5520 ZFS_EXIT(zfsvfs); 5521 nvlist_free(nvlist); 5522 return (error); 5523 } 5524 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target, 5525 kcred, NULL, 0); 5526 nvlist_free(nvlist); 5527 break; 5528 5529 case ZFS_SMB_ACL_PURGE: 5530 error = zfs_smb_acl_purge(sharedir); 5531 break; 5532 5533 default: 5534 error = SET_ERROR(EINVAL); 5535 break; 5536 } 5537 5538 VN_RELE(vp); 5539 VN_RELE(ZTOV(sharedir)); 5540 5541 ZFS_EXIT(zfsvfs); 5542 5543 return (error); 5544#else /* !illumos */ 5545 return (EOPNOTSUPP); 5546#endif /* illumos */ 5547} 5548 5549/* 5550 * innvl: { 5551 * "holds" -> { snapname -> holdname (string), ... } 5552 * (optional) "cleanup_fd" -> fd (int32) 5553 * } 5554 * 5555 * outnvl: { 5556 * snapname -> error value (int32) 5557 * ... 5558 * } 5559 */ 5560/* ARGSUSED */ 5561static int 5562zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist) 5563{ 5564 nvpair_t *pair; 5565 nvlist_t *holds; 5566 int cleanup_fd = -1; 5567 int error; 5568 minor_t minor = 0; 5569 5570 error = nvlist_lookup_nvlist(args, "holds", &holds); 5571 if (error != 0) 5572 return (SET_ERROR(EINVAL)); 5573 5574 /* make sure the user didn't pass us any invalid (empty) tags */ 5575 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL; 5576 pair = nvlist_next_nvpair(holds, pair)) { 5577 char *htag; 5578 5579 error = nvpair_value_string(pair, &htag); 5580 if (error != 0) 5581 return (SET_ERROR(error)); 5582 5583 if (strlen(htag) == 0) 5584 return (SET_ERROR(EINVAL)); 5585 } 5586 5587 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) { 5588 error = zfs_onexit_fd_hold(cleanup_fd, &minor); 5589 if (error != 0) 5590 return (error); 5591 } 5592 5593 error = dsl_dataset_user_hold(holds, minor, errlist); 5594 if (minor != 0) 5595 zfs_onexit_fd_rele(cleanup_fd); 5596 return (error); 5597} 5598 5599/* 5600 * innvl is not used. 5601 * 5602 * outnvl: { 5603 * holdname -> time added (uint64 seconds since epoch) 5604 * ... 5605 * } 5606 */ 5607/* ARGSUSED */ 5608static int 5609zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl) 5610{ 5611 return (dsl_dataset_get_holds(snapname, outnvl)); 5612} 5613 5614/* 5615 * innvl: { 5616 * snapname -> { holdname, ... } 5617 * ... 5618 * } 5619 * 5620 * outnvl: { 5621 * snapname -> error value (int32) 5622 * ... 5623 * } 5624 */ 5625/* ARGSUSED */ 5626static int 5627zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist) 5628{ 5629 return (dsl_dataset_user_release(holds, errlist)); 5630} 5631 5632/* 5633 * inputs: 5634 * zc_name name of new filesystem or snapshot 5635 * zc_value full name of old snapshot 5636 * 5637 * outputs: 5638 * zc_cookie space in bytes 5639 * zc_objset_type compressed space in bytes 5640 * zc_perm_action uncompressed space in bytes 5641 */ 5642static int 5643zfs_ioc_space_written(zfs_cmd_t *zc) 5644{ 5645 int error; 5646 dsl_pool_t *dp; 5647 dsl_dataset_t *new, *old; 5648 5649 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 5650 if (error != 0) 5651 return (error); 5652 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new); 5653 if (error != 0) { 5654 dsl_pool_rele(dp, FTAG); 5655 return (error); 5656 } 5657 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old); 5658 if (error != 0) { 5659 dsl_dataset_rele(new, FTAG); 5660 dsl_pool_rele(dp, FTAG); 5661 return (error); 5662 } 5663 5664 error = dsl_dataset_space_written(old, new, &zc->zc_cookie, 5665 &zc->zc_objset_type, &zc->zc_perm_action); 5666 dsl_dataset_rele(old, FTAG); 5667 dsl_dataset_rele(new, FTAG); 5668 dsl_pool_rele(dp, FTAG); 5669 return (error); 5670} 5671 5672/* 5673 * innvl: { 5674 * "firstsnap" -> snapshot name 5675 * } 5676 * 5677 * outnvl: { 5678 * "used" -> space in bytes 5679 * "compressed" -> compressed space in bytes 5680 * "uncompressed" -> uncompressed space in bytes 5681 * } 5682 */ 5683static int 5684zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl) 5685{ 5686 int error; 5687 dsl_pool_t *dp; 5688 dsl_dataset_t *new, *old; 5689 char *firstsnap; 5690 uint64_t used, comp, uncomp; 5691 5692 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0) 5693 return (SET_ERROR(EINVAL)); 5694 5695 error = dsl_pool_hold(lastsnap, FTAG, &dp); 5696 if (error != 0) 5697 return (error); 5698 5699 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new); 5700 if (error == 0 && !new->ds_is_snapshot) { 5701 dsl_dataset_rele(new, FTAG); 5702 error = SET_ERROR(EINVAL); 5703 } 5704 if (error != 0) { 5705 dsl_pool_rele(dp, FTAG); 5706 return (error); 5707 } 5708 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old); 5709 if (error == 0 && !old->ds_is_snapshot) { 5710 dsl_dataset_rele(old, FTAG); 5711 error = SET_ERROR(EINVAL); 5712 } 5713 if (error != 0) { 5714 dsl_dataset_rele(new, FTAG); 5715 dsl_pool_rele(dp, FTAG); 5716 return (error); 5717 } 5718 5719 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp); 5720 dsl_dataset_rele(old, FTAG); 5721 dsl_dataset_rele(new, FTAG); 5722 dsl_pool_rele(dp, FTAG); 5723 fnvlist_add_uint64(outnvl, "used", used); 5724 fnvlist_add_uint64(outnvl, "compressed", comp); 5725 fnvlist_add_uint64(outnvl, "uncompressed", uncomp); 5726 return (error); 5727} 5728 5729static int 5730zfs_ioc_jail(zfs_cmd_t *zc) 5731{ 5732 5733 return (zone_dataset_attach(curthread->td_ucred, zc->zc_name, 5734 (int)zc->zc_jailid)); 5735} 5736 5737static int 5738zfs_ioc_unjail(zfs_cmd_t *zc) 5739{ 5740 5741 return (zone_dataset_detach(curthread->td_ucred, zc->zc_name, 5742 (int)zc->zc_jailid)); 5743} 5744 5745/* 5746 * innvl: { 5747 * "fd" -> file descriptor to write stream to (int32) 5748 * (optional) "fromsnap" -> full snap name to send an incremental from 5749 * (optional) "largeblockok" -> (value ignored) 5750 * indicates that blocks > 128KB are permitted 5751 * (optional) "embedok" -> (value ignored) 5752 * presence indicates DRR_WRITE_EMBEDDED records are permitted 5753 * (optional) "compressok" -> (value ignored) 5754 * presence indicates compressed DRR_WRITE records are permitted 5755 * (optional) "resume_object" and "resume_offset" -> (uint64) 5756 * if present, resume send stream from specified object and offset. 5757 * } 5758 * 5759 * outnvl is unused 5760 */ 5761/* ARGSUSED */ 5762static int 5763zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl) 5764{ 5765 cap_rights_t rights; 5766 file_t *fp; 5767 int error; 5768 offset_t off; 5769 char *fromname = NULL; 5770 int fd; 5771 boolean_t largeblockok; 5772 boolean_t embedok; 5773 boolean_t compressok; 5774 uint64_t resumeobj = 0; 5775 uint64_t resumeoff = 0; 5776 5777 error = nvlist_lookup_int32(innvl, "fd", &fd); 5778 if (error != 0) 5779 return (SET_ERROR(EINVAL)); 5780 5781 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname); 5782 5783 largeblockok = nvlist_exists(innvl, "largeblockok"); 5784 embedok = nvlist_exists(innvl, "embedok"); 5785 compressok = nvlist_exists(innvl, "compressok"); 5786 5787 (void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj); 5788 (void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff); 5789 5790#ifdef illumos 5791 file_t *fp = getf(fd); 5792#else 5793 fget_write(curthread, fd, cap_rights_init(&rights, CAP_WRITE), &fp); 5794#endif 5795 if (fp == NULL) 5796 return (SET_ERROR(EBADF)); 5797 5798 off = fp->f_offset; 5799 error = dmu_send(snapname, fromname, embedok, largeblockok, compressok, 5800#ifdef illumos 5801 fd, resumeobj, resumeoff, fp->f_vnode, &off); 5802#else 5803 fd, resumeobj, resumeoff, fp, &off); 5804#endif 5805 5806#ifdef illumos 5807 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0) 5808 fp->f_offset = off; 5809#else 5810 fp->f_offset = off; 5811#endif 5812 5813 releasef(fd); 5814 return (error); 5815} 5816 5817/* 5818 * Determine approximately how large a zfs send stream will be -- the number 5819 * of bytes that will be written to the fd supplied to zfs_ioc_send_new(). 5820 * 5821 * innvl: { 5822 * (optional) "from" -> full snap or bookmark name to send an incremental 5823 * from 5824 * (optional) "largeblockok" -> (value ignored) 5825 * indicates that blocks > 128KB are permitted 5826 * (optional) "embedok" -> (value ignored) 5827 * presence indicates DRR_WRITE_EMBEDDED records are permitted 5828 * (optional) "compressok" -> (value ignored) 5829 * presence indicates compressed DRR_WRITE records are permitted 5830 * } 5831 * 5832 * outnvl: { 5833 * "space" -> bytes of space (uint64) 5834 * } 5835 */ 5836static int 5837zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl) 5838{ 5839 dsl_pool_t *dp; 5840 dsl_dataset_t *tosnap; 5841 int error; 5842 char *fromname; 5843 boolean_t compressok; 5844 uint64_t space; 5845 5846 error = dsl_pool_hold(snapname, FTAG, &dp); 5847 if (error != 0) 5848 return (error); 5849 5850 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap); 5851 if (error != 0) { 5852 dsl_pool_rele(dp, FTAG); 5853 return (error); 5854 } 5855 5856 compressok = nvlist_exists(innvl, "compressok"); 5857 5858 error = nvlist_lookup_string(innvl, "from", &fromname); 5859 if (error == 0) { 5860 if (strchr(fromname, '@') != NULL) { 5861 /* 5862 * If from is a snapshot, hold it and use the more 5863 * efficient dmu_send_estimate to estimate send space 5864 * size using deadlists. 5865 */ 5866 dsl_dataset_t *fromsnap; 5867 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap); 5868 if (error != 0) 5869 goto out; 5870 error = dmu_send_estimate(tosnap, fromsnap, compressok, 5871 &space); 5872 dsl_dataset_rele(fromsnap, FTAG); 5873 } else if (strchr(fromname, '#') != NULL) { 5874 /* 5875 * If from is a bookmark, fetch the creation TXG of the 5876 * snapshot it was created from and use that to find 5877 * blocks that were born after it. 5878 */ 5879 zfs_bookmark_phys_t frombm; 5880 5881 error = dsl_bookmark_lookup(dp, fromname, tosnap, 5882 &frombm); 5883 if (error != 0) 5884 goto out; 5885 error = dmu_send_estimate_from_txg(tosnap, 5886 frombm.zbm_creation_txg, compressok, &space); 5887 } else { 5888 /* 5889 * from is not properly formatted as a snapshot or 5890 * bookmark 5891 */ 5892 error = SET_ERROR(EINVAL); 5893 goto out; 5894 } 5895 } else { 5896 /* 5897 * If estimating the size of a full send, use dmu_send_estimate. 5898 */ 5899 error = dmu_send_estimate(tosnap, NULL, compressok, &space); 5900 } 5901 5902 fnvlist_add_uint64(outnvl, "space", space); 5903 5904out: 5905 dsl_dataset_rele(tosnap, FTAG); 5906 dsl_pool_rele(dp, FTAG); 5907 return (error); 5908} 5909 5910static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST]; 5911 5912static void 5913zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5914 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck, 5915 boolean_t log_history, zfs_ioc_poolcheck_t pool_check) 5916{ 5917 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST]; 5918 5919 ASSERT3U(ioc, >=, ZFS_IOC_FIRST); 5920 ASSERT3U(ioc, <, ZFS_IOC_LAST); 5921 ASSERT3P(vec->zvec_legacy_func, ==, NULL); 5922 ASSERT3P(vec->zvec_func, ==, NULL); 5923 5924 vec->zvec_legacy_func = func; 5925 vec->zvec_secpolicy = secpolicy; 5926 vec->zvec_namecheck = namecheck; 5927 vec->zvec_allow_log = log_history; 5928 vec->zvec_pool_check = pool_check; 5929} 5930 5931/* 5932 * See the block comment at the beginning of this file for details on 5933 * each argument to this function. 5934 */ 5935static void 5936zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func, 5937 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck, 5938 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist, 5939 boolean_t allow_log) 5940{ 5941 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST]; 5942 5943 ASSERT3U(ioc, >=, ZFS_IOC_FIRST); 5944 ASSERT3U(ioc, <, ZFS_IOC_LAST); 5945 ASSERT3P(vec->zvec_legacy_func, ==, NULL); 5946 ASSERT3P(vec->zvec_func, ==, NULL); 5947 5948 /* if we are logging, the name must be valid */ 5949 ASSERT(!allow_log || namecheck != NO_NAME); 5950 5951 vec->zvec_name = name; 5952 vec->zvec_func = func; 5953 vec->zvec_secpolicy = secpolicy; 5954 vec->zvec_namecheck = namecheck; 5955 vec->zvec_pool_check = pool_check; 5956 vec->zvec_smush_outnvlist = smush_outnvlist; 5957 vec->zvec_allow_log = allow_log; 5958} 5959 5960static void 5961zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5962 zfs_secpolicy_func_t *secpolicy, boolean_t log_history, 5963 zfs_ioc_poolcheck_t pool_check) 5964{ 5965 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5966 POOL_NAME, log_history, pool_check); 5967} 5968 5969static void 5970zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5971 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check) 5972{ 5973 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5974 DATASET_NAME, B_FALSE, pool_check); 5975} 5976 5977static void 5978zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func) 5979{ 5980 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config, 5981 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5982} 5983 5984static void 5985zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5986 zfs_secpolicy_func_t *secpolicy) 5987{ 5988 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5989 NO_NAME, B_FALSE, POOL_CHECK_NONE); 5990} 5991 5992static void 5993zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc, 5994 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy) 5995{ 5996 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5997 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED); 5998} 5999 6000static void 6001zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func) 6002{ 6003 zfs_ioctl_register_dataset_read_secpolicy(ioc, func, 6004 zfs_secpolicy_read); 6005} 6006 6007static void 6008zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 6009 zfs_secpolicy_func_t *secpolicy) 6010{ 6011 zfs_ioctl_register_legacy(ioc, func, secpolicy, 6012 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 6013} 6014 6015static void 6016zfs_ioctl_init(void) 6017{ 6018 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT, 6019 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME, 6020 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 6021 6022 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY, 6023 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME, 6024 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE); 6025 6026 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS, 6027 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME, 6028 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 6029 6030 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW, 6031 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME, 6032 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 6033 6034 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE, 6035 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME, 6036 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 6037 6038 zfs_ioctl_register("create", ZFS_IOC_CREATE, 6039 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME, 6040 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 6041 6042 zfs_ioctl_register("clone", ZFS_IOC_CLONE, 6043 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME, 6044 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 6045 6046 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS, 6047 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME, 6048 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 6049 6050 zfs_ioctl_register("hold", ZFS_IOC_HOLD, 6051 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME, 6052 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 6053 zfs_ioctl_register("release", ZFS_IOC_RELEASE, 6054 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME, 6055 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 6056 6057 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS, 6058 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME, 6059 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 6060 6061 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK, 6062 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, 6063 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE); 6064 6065 zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK, 6066 zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME, 6067 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 6068 6069 zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS, 6070 zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME, 6071 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 6072 6073 zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS, 6074 zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks, 6075 POOL_NAME, 6076 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 6077 6078 zfs_ioctl_register("channel_program", ZFS_IOC_CHANNEL_PROGRAM, 6079 zfs_ioc_channel_program, zfs_secpolicy_config, 6080 POOL_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, 6081 B_TRUE); 6082 6083 /* IOCTLS that use the legacy function signature */ 6084 6085 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze, 6086 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY); 6087 6088 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create, 6089 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 6090 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN, 6091 zfs_ioc_pool_scan); 6092 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE, 6093 zfs_ioc_pool_upgrade); 6094 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD, 6095 zfs_ioc_vdev_add); 6096 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE, 6097 zfs_ioc_vdev_remove); 6098 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE, 6099 zfs_ioc_vdev_set_state); 6100 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH, 6101 zfs_ioc_vdev_attach); 6102 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH, 6103 zfs_ioc_vdev_detach); 6104 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH, 6105 zfs_ioc_vdev_setpath); 6106 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU, 6107 zfs_ioc_vdev_setfru); 6108 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS, 6109 zfs_ioc_pool_set_props); 6110 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT, 6111 zfs_ioc_vdev_split); 6112 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID, 6113 zfs_ioc_pool_reguid); 6114 6115 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS, 6116 zfs_ioc_pool_configs, zfs_secpolicy_none); 6117 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT, 6118 zfs_ioc_pool_tryimport, zfs_secpolicy_config); 6119 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT, 6120 zfs_ioc_inject_fault, zfs_secpolicy_inject); 6121 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT, 6122 zfs_ioc_clear_fault, zfs_secpolicy_inject); 6123 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT, 6124 zfs_ioc_inject_list_next, zfs_secpolicy_inject); 6125 6126 /* 6127 * pool destroy, and export don't log the history as part of 6128 * zfsdev_ioctl, but rather zfs_ioc_pool_export 6129 * does the logging of those commands. 6130 */ 6131 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy, 6132 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE); 6133 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export, 6134 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE); 6135 6136 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats, 6137 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE); 6138 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props, 6139 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE); 6140 6141 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log, 6142 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_NONE); 6143 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME, 6144 zfs_ioc_dsobj_to_dsname, 6145 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_NONE); 6146 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY, 6147 zfs_ioc_pool_get_history, 6148 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED); 6149 6150 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import, 6151 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 6152 6153 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear, 6154 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 6155 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen, 6156 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED); 6157 6158 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN, 6159 zfs_ioc_space_written); 6160 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS, 6161 zfs_ioc_objset_recvd_props); 6162 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ, 6163 zfs_ioc_next_obj); 6164 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL, 6165 zfs_ioc_get_fsacl); 6166 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS, 6167 zfs_ioc_objset_stats); 6168 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS, 6169 zfs_ioc_objset_zplprops); 6170 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT, 6171 zfs_ioc_dataset_list_next); 6172 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT, 6173 zfs_ioc_snapshot_list_next); 6174 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS, 6175 zfs_ioc_send_progress); 6176 6177 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF, 6178 zfs_ioc_diff, zfs_secpolicy_diff); 6179 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS, 6180 zfs_ioc_obj_to_stats, zfs_secpolicy_diff); 6181 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH, 6182 zfs_ioc_obj_to_path, zfs_secpolicy_diff); 6183 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE, 6184 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one); 6185 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY, 6186 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many); 6187 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND, 6188 zfs_ioc_send, zfs_secpolicy_send); 6189 6190 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop, 6191 zfs_secpolicy_none); 6192 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy, 6193 zfs_secpolicy_destroy); 6194 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename, 6195 zfs_secpolicy_rename); 6196 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv, 6197 zfs_secpolicy_recv); 6198 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote, 6199 zfs_secpolicy_promote); 6200 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP, 6201 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop); 6202 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl, 6203 zfs_secpolicy_set_fsacl); 6204 6205 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share, 6206 zfs_secpolicy_share, POOL_CHECK_NONE); 6207 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl, 6208 zfs_secpolicy_smb_acl, POOL_CHECK_NONE); 6209 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE, 6210 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade, 6211 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 6212 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT, 6213 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot, 6214 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 6215 6216#ifdef __FreeBSD__ 6217 zfs_ioctl_register_dataset_nolog(ZFS_IOC_JAIL, zfs_ioc_jail, 6218 zfs_secpolicy_config, POOL_CHECK_NONE); 6219 zfs_ioctl_register_dataset_nolog(ZFS_IOC_UNJAIL, zfs_ioc_unjail, 6220 zfs_secpolicy_config, POOL_CHECK_NONE); 6221 zfs_ioctl_register("fbsd_nextboot", ZFS_IOC_NEXTBOOT, 6222 zfs_ioc_nextboot, zfs_secpolicy_config, NO_NAME, 6223 POOL_CHECK_NONE, B_FALSE, B_FALSE); 6224#endif 6225} 6226 6227int 6228pool_status_check(const char *name, zfs_ioc_namecheck_t type, 6229 zfs_ioc_poolcheck_t check) 6230{ 6231 spa_t *spa; 6232 int error; 6233 6234 ASSERT(type == POOL_NAME || type == DATASET_NAME); 6235 6236 if (check & POOL_CHECK_NONE) 6237 return (0); 6238 6239 error = spa_open(name, &spa, FTAG); 6240 if (error == 0) { 6241 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa)) 6242 error = SET_ERROR(EAGAIN); 6243 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa)) 6244 error = SET_ERROR(EROFS); 6245 spa_close(spa, FTAG); 6246 } 6247 return (error); 6248} 6249 6250/* 6251 * Find a free minor number. 6252 */ 6253minor_t 6254zfsdev_minor_alloc(void) 6255{ 6256 static minor_t last_minor; 6257 minor_t m; 6258 6259 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 6260 6261 for (m = last_minor + 1; m != last_minor; m++) { 6262 if (m > ZFSDEV_MAX_MINOR) 6263 m = 1; 6264 if (ddi_get_soft_state(zfsdev_state, m) == NULL) { 6265 last_minor = m; 6266 return (m); 6267 } 6268 } 6269 6270 return (0); 6271} 6272 6273static int 6274zfs_ctldev_init(struct cdev *devp) 6275{ 6276 minor_t minor; 6277 zfs_soft_state_t *zs; 6278 6279 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 6280 6281 minor = zfsdev_minor_alloc(); 6282 if (minor == 0) 6283 return (SET_ERROR(ENXIO)); 6284 6285 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS) 6286 return (SET_ERROR(EAGAIN)); 6287 6288 devfs_set_cdevpriv((void *)(uintptr_t)minor, zfsdev_close); 6289 6290 zs = ddi_get_soft_state(zfsdev_state, minor); 6291 zs->zss_type = ZSST_CTLDEV; 6292 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data); 6293 6294 return (0); 6295} 6296 6297static void 6298zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor) 6299{ 6300 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 6301 6302 zfs_onexit_destroy(zo); 6303 ddi_soft_state_free(zfsdev_state, minor); 6304} 6305 6306void * 6307zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which) 6308{ 6309 zfs_soft_state_t *zp; 6310 6311 zp = ddi_get_soft_state(zfsdev_state, minor); 6312 if (zp == NULL || zp->zss_type != which) 6313 return (NULL); 6314 6315 return (zp->zss_data); 6316} 6317 6318static int 6319zfsdev_open(struct cdev *devp, int flag, int mode, struct thread *td) 6320{ 6321 int error = 0; 6322 6323#ifdef illumos 6324 if (getminor(*devp) != 0) 6325 return (zvol_open(devp, flag, otyp, cr)); 6326#endif 6327 6328 /* This is the control device. Allocate a new minor if requested. */ 6329 if (flag & FEXCL) { 6330 mutex_enter(&spa_namespace_lock); 6331 error = zfs_ctldev_init(devp); 6332 mutex_exit(&spa_namespace_lock); 6333 } 6334 6335 return (error); 6336} 6337 6338static void 6339zfsdev_close(void *data) 6340{ 6341 zfs_onexit_t *zo; 6342 minor_t minor = (minor_t)(uintptr_t)data; 6343 6344 if (minor == 0) 6345 return; 6346 6347 mutex_enter(&spa_namespace_lock); 6348 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV); 6349 if (zo == NULL) { 6350 mutex_exit(&spa_namespace_lock); 6351 return; 6352 } 6353 zfs_ctldev_destroy(zo, minor); 6354 mutex_exit(&spa_namespace_lock); 6355} 6356 6357static int 6358zfsdev_ioctl(struct cdev *dev, u_long zcmd, caddr_t arg, int flag, 6359 struct thread *td) 6360{ 6361 zfs_cmd_t *zc; 6362 uint_t vecnum; 6363 int error, rc, len; 6364#ifdef illumos 6365 minor_t minor = getminor(dev); 6366#else 6367 zfs_iocparm_t *zc_iocparm; 6368 int cflag, cmd, oldvecnum; 6369 boolean_t newioc, compat; 6370 void *compat_zc = NULL; 6371 cred_t *cr = td->td_ucred; 6372#endif 6373 const zfs_ioc_vec_t *vec; 6374 char *saved_poolname = NULL; 6375 nvlist_t *innvl = NULL; 6376 6377 cflag = ZFS_CMD_COMPAT_NONE; 6378 compat = B_FALSE; 6379 newioc = B_TRUE; /* "new" style (zfs_iocparm_t) ioctl */ 6380 6381 len = IOCPARM_LEN(zcmd); 6382 vecnum = cmd = zcmd & 0xff; 6383 6384 /* 6385 * Check if we are talking to supported older binaries 6386 * and translate zfs_cmd if necessary 6387 */ 6388 if (len != sizeof(zfs_iocparm_t)) { 6389 newioc = B_FALSE; 6390 compat = B_TRUE; 6391 6392 vecnum = cmd; 6393 6394 switch (len) { 6395 case sizeof(zfs_cmd_zcmd_t): 6396 cflag = ZFS_CMD_COMPAT_LZC; 6397 break; 6398 case sizeof(zfs_cmd_deadman_t): 6399 cflag = ZFS_CMD_COMPAT_DEADMAN; 6400 break; 6401 case sizeof(zfs_cmd_v28_t): 6402 cflag = ZFS_CMD_COMPAT_V28; 6403 break; 6404 case sizeof(zfs_cmd_v15_t): 6405 cflag = ZFS_CMD_COMPAT_V15; 6406 vecnum = zfs_ioctl_v15_to_v28[cmd]; 6407 6408 /* 6409 * Return without further handling 6410 * if the command is blacklisted. 6411 */ 6412 if (vecnum == ZFS_IOC_COMPAT_PASS) 6413 return (0); 6414 else if (vecnum == ZFS_IOC_COMPAT_FAIL) 6415 return (ENOTSUP); 6416 break; 6417 default: 6418 return (EINVAL); 6419 } 6420 } 6421 6422#ifdef illumos 6423 vecnum = cmd - ZFS_IOC_FIRST; 6424 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip)); 6425#endif 6426 6427 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0])) 6428 return (SET_ERROR(EINVAL)); 6429 vec = &zfs_ioc_vec[vecnum]; 6430 6431 zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP); 6432 6433#ifdef illumos 6434 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag); 6435 if (error != 0) { 6436 error = SET_ERROR(EFAULT); 6437 goto out; 6438 } 6439#else /* !illumos */ 6440 bzero(zc, sizeof(zfs_cmd_t)); 6441 6442 if (newioc) { 6443 zc_iocparm = (void *)arg; 6444 6445 switch (zc_iocparm->zfs_ioctl_version) { 6446 case ZFS_IOCVER_CURRENT: 6447 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_t)) { 6448 error = SET_ERROR(EINVAL); 6449 goto out; 6450 } 6451 break; 6452 case ZFS_IOCVER_INLANES: 6453 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_inlanes_t)) { 6454 error = SET_ERROR(EFAULT); 6455 goto out; 6456 } 6457 compat = B_TRUE; 6458 cflag = ZFS_CMD_COMPAT_INLANES; 6459 break; 6460 case ZFS_IOCVER_RESUME: 6461 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_resume_t)) { 6462 error = SET_ERROR(EFAULT); 6463 goto out; 6464 } 6465 compat = B_TRUE; 6466 cflag = ZFS_CMD_COMPAT_RESUME; 6467 break; 6468 case ZFS_IOCVER_EDBP: 6469 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_edbp_t)) { 6470 error = SET_ERROR(EFAULT); 6471 goto out; 6472 } 6473 compat = B_TRUE; 6474 cflag = ZFS_CMD_COMPAT_EDBP; 6475 break; 6476 case ZFS_IOCVER_ZCMD: 6477 if (zc_iocparm->zfs_cmd_size > sizeof(zfs_cmd_t) || 6478 zc_iocparm->zfs_cmd_size < sizeof(zfs_cmd_zcmd_t)) { 6479 error = SET_ERROR(EFAULT); 6480 goto out; 6481 } 6482 compat = B_TRUE; 6483 cflag = ZFS_CMD_COMPAT_ZCMD; 6484 break; 6485 default: 6486 error = SET_ERROR(EINVAL); 6487 goto out; 6488 /* NOTREACHED */ 6489 } 6490 6491 if (compat) { 6492 ASSERT(sizeof(zfs_cmd_t) >= zc_iocparm->zfs_cmd_size); 6493 compat_zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP); 6494 bzero(compat_zc, sizeof(zfs_cmd_t)); 6495 6496 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd, 6497 compat_zc, zc_iocparm->zfs_cmd_size, flag); 6498 if (error != 0) { 6499 error = SET_ERROR(EFAULT); 6500 goto out; 6501 } 6502 } else { 6503 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd, 6504 zc, zc_iocparm->zfs_cmd_size, flag); 6505 if (error != 0) { 6506 error = SET_ERROR(EFAULT); 6507 goto out; 6508 } 6509 } 6510 } 6511 6512 if (compat) { 6513 if (newioc) { 6514 ASSERT(compat_zc != NULL); 6515 zfs_cmd_compat_get(zc, compat_zc, cflag); 6516 } else { 6517 ASSERT(compat_zc == NULL); 6518 zfs_cmd_compat_get(zc, arg, cflag); 6519 } 6520 oldvecnum = vecnum; 6521 error = zfs_ioctl_compat_pre(zc, &vecnum, cflag); 6522 if (error != 0) 6523 goto out; 6524 if (oldvecnum != vecnum) 6525 vec = &zfs_ioc_vec[vecnum]; 6526 } 6527#endif /* !illumos */ 6528 6529 zc->zc_iflags = flag & FKIOCTL; 6530 if (zc->zc_nvlist_src_size != 0) { 6531 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 6532 zc->zc_iflags, &innvl); 6533 if (error != 0) 6534 goto out; 6535 } 6536 6537 /* rewrite innvl for backwards compatibility */ 6538 if (compat) 6539 innvl = zfs_ioctl_compat_innvl(zc, innvl, vecnum, cflag); 6540 6541 /* 6542 * Ensure that all pool/dataset names are valid before we pass down to 6543 * the lower layers. 6544 */ 6545 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0'; 6546 switch (vec->zvec_namecheck) { 6547 case POOL_NAME: 6548 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0) 6549 error = SET_ERROR(EINVAL); 6550 else 6551 error = pool_status_check(zc->zc_name, 6552 vec->zvec_namecheck, vec->zvec_pool_check); 6553 break; 6554 6555 case DATASET_NAME: 6556 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0) 6557 error = SET_ERROR(EINVAL); 6558 else 6559 error = pool_status_check(zc->zc_name, 6560 vec->zvec_namecheck, vec->zvec_pool_check); 6561 break; 6562 6563 case NO_NAME: 6564 break; 6565 } 6566 6567 if (error == 0) 6568 error = vec->zvec_secpolicy(zc, innvl, cr); 6569 6570 if (error != 0) 6571 goto out; 6572 6573 /* legacy ioctls can modify zc_name */ 6574 len = strcspn(zc->zc_name, "/@#") + 1; 6575 saved_poolname = kmem_alloc(len, KM_SLEEP); 6576 (void) strlcpy(saved_poolname, zc->zc_name, len); 6577 6578 if (vec->zvec_func != NULL) { 6579 nvlist_t *outnvl; 6580 int puterror = 0; 6581 spa_t *spa; 6582 nvlist_t *lognv = NULL; 6583 6584 ASSERT(vec->zvec_legacy_func == NULL); 6585 6586 /* 6587 * Add the innvl to the lognv before calling the func, 6588 * in case the func changes the innvl. 6589 */ 6590 if (vec->zvec_allow_log) { 6591 lognv = fnvlist_alloc(); 6592 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL, 6593 vec->zvec_name); 6594 if (!nvlist_empty(innvl)) { 6595 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL, 6596 innvl); 6597 } 6598 } 6599 6600 outnvl = fnvlist_alloc(); 6601 error = vec->zvec_func(zc->zc_name, innvl, outnvl); 6602 6603 /* 6604 * Some commands can partially execute, modfiy state, and still 6605 * return an error. In these cases, attempt to record what 6606 * was modified. 6607 */ 6608 if ((error == 0 || 6609 (cmd == ZFS_IOC_CHANNEL_PROGRAM && error != EINVAL)) && 6610 vec->zvec_allow_log && 6611 spa_open(zc->zc_name, &spa, FTAG) == 0) { 6612 if (!nvlist_empty(outnvl)) { 6613 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL, 6614 outnvl); 6615 } 6616 if (error != 0) { 6617 fnvlist_add_int64(lognv, ZPOOL_HIST_ERRNO, 6618 error); 6619 } 6620 (void) spa_history_log_nvl(spa, lognv); 6621 spa_close(spa, FTAG); 6622 } 6623 fnvlist_free(lognv); 6624 6625 /* rewrite outnvl for backwards compatibility */ 6626 if (compat) 6627 outnvl = zfs_ioctl_compat_outnvl(zc, outnvl, vecnum, 6628 cflag); 6629 6630 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) { 6631 int smusherror = 0; 6632 if (vec->zvec_smush_outnvlist) { 6633 smusherror = nvlist_smush(outnvl, 6634 zc->zc_nvlist_dst_size); 6635 } 6636 if (smusherror == 0) 6637 puterror = put_nvlist(zc, outnvl); 6638 } 6639 6640 if (puterror != 0) 6641 error = puterror; 6642 6643 nvlist_free(outnvl); 6644 } else { 6645 error = vec->zvec_legacy_func(zc); 6646 } 6647 6648out: 6649 nvlist_free(innvl); 6650 6651#ifdef illumos 6652 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag); 6653 if (error == 0 && rc != 0) 6654 error = SET_ERROR(EFAULT); 6655#else 6656 if (compat) { 6657 zfs_ioctl_compat_post(zc, cmd, cflag); 6658 if (newioc) { 6659 ASSERT(compat_zc != NULL); 6660 ASSERT(sizeof(zfs_cmd_t) >= zc_iocparm->zfs_cmd_size); 6661 6662 zfs_cmd_compat_put(zc, compat_zc, vecnum, cflag); 6663 rc = ddi_copyout(compat_zc, 6664 (void *)(uintptr_t)zc_iocparm->zfs_cmd, 6665 zc_iocparm->zfs_cmd_size, flag); 6666 if (error == 0 && rc != 0) 6667 error = SET_ERROR(EFAULT); 6668 kmem_free(compat_zc, sizeof (zfs_cmd_t)); 6669 } else { 6670 zfs_cmd_compat_put(zc, arg, vecnum, cflag); 6671 } 6672 } else { 6673 ASSERT(newioc); 6674 6675 rc = ddi_copyout(zc, (void *)(uintptr_t)zc_iocparm->zfs_cmd, 6676 sizeof (zfs_cmd_t), flag); 6677 if (error == 0 && rc != 0) 6678 error = SET_ERROR(EFAULT); 6679 } 6680#endif 6681 if (error == 0 && vec->zvec_allow_log) { 6682 char *s = tsd_get(zfs_allow_log_key); 6683 if (s != NULL) 6684 strfree(s); 6685 (void) tsd_set(zfs_allow_log_key, saved_poolname); 6686 } else { 6687 if (saved_poolname != NULL) 6688 strfree(saved_poolname); 6689 } 6690 6691 kmem_free(zc, sizeof (zfs_cmd_t)); 6692 return (error); 6693} 6694 6695#ifdef illumos 6696static int 6697zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd) 6698{ 6699 if (cmd != DDI_ATTACH) 6700 return (DDI_FAILURE); 6701 6702 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0, 6703 DDI_PSEUDO, 0) == DDI_FAILURE) 6704 return (DDI_FAILURE); 6705 6706 zfs_dip = dip; 6707 6708 ddi_report_dev(dip); 6709 6710 return (DDI_SUCCESS); 6711} 6712 6713static int 6714zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd) 6715{ 6716 if (spa_busy() || zfs_busy() || zvol_busy()) 6717 return (DDI_FAILURE); 6718 6719 if (cmd != DDI_DETACH) 6720 return (DDI_FAILURE); 6721 6722 zfs_dip = NULL; 6723 6724 ddi_prop_remove_all(dip); 6725 ddi_remove_minor_node(dip, NULL); 6726 6727 return (DDI_SUCCESS); 6728} 6729 6730/*ARGSUSED*/ 6731static int 6732zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result) 6733{ 6734 switch (infocmd) { 6735 case DDI_INFO_DEVT2DEVINFO: 6736 *result = zfs_dip; 6737 return (DDI_SUCCESS); 6738 6739 case DDI_INFO_DEVT2INSTANCE: 6740 *result = (void *)0; 6741 return (DDI_SUCCESS); 6742 } 6743 6744 return (DDI_FAILURE); 6745} 6746#endif /* illumos */ 6747 6748/* 6749 * OK, so this is a little weird. 6750 * 6751 * /dev/zfs is the control node, i.e. minor 0. 6752 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0. 6753 * 6754 * /dev/zfs has basically nothing to do except serve up ioctls, 6755 * so most of the standard driver entry points are in zvol.c. 6756 */ 6757#ifdef illumos 6758static struct cb_ops zfs_cb_ops = { 6759 zfsdev_open, /* open */ 6760 zfsdev_close, /* close */ 6761 zvol_strategy, /* strategy */ 6762 nodev, /* print */ 6763 zvol_dump, /* dump */ 6764 zvol_read, /* read */ 6765 zvol_write, /* write */ 6766 zfsdev_ioctl, /* ioctl */ 6767 nodev, /* devmap */ 6768 nodev, /* mmap */ 6769 nodev, /* segmap */ 6770 nochpoll, /* poll */ 6771 ddi_prop_op, /* prop_op */ 6772 NULL, /* streamtab */ 6773 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */ 6774 CB_REV, /* version */ 6775 nodev, /* async read */ 6776 nodev, /* async write */ 6777}; 6778 6779static struct dev_ops zfs_dev_ops = { 6780 DEVO_REV, /* version */ 6781 0, /* refcnt */ 6782 zfs_info, /* info */ 6783 nulldev, /* identify */ 6784 nulldev, /* probe */ 6785 zfs_attach, /* attach */ 6786 zfs_detach, /* detach */ 6787 nodev, /* reset */ 6788 &zfs_cb_ops, /* driver operations */ 6789 NULL, /* no bus operations */ 6790 NULL, /* power */ 6791 ddi_quiesce_not_needed, /* quiesce */ 6792}; 6793 6794static struct modldrv zfs_modldrv = { 6795 &mod_driverops, 6796 "ZFS storage pool", 6797 &zfs_dev_ops 6798}; 6799 6800static struct modlinkage modlinkage = { 6801 MODREV_1, 6802 (void *)&zfs_modlfs, 6803 (void *)&zfs_modldrv, 6804 NULL 6805}; 6806#endif /* illumos */ 6807 6808static struct cdevsw zfs_cdevsw = { 6809 .d_version = D_VERSION, 6810 .d_open = zfsdev_open, 6811 .d_ioctl = zfsdev_ioctl, 6812 .d_name = ZFS_DEV_NAME 6813}; 6814 6815static void 6816zfs_allow_log_destroy(void *arg) 6817{ 6818 char *poolname = arg; 6819 strfree(poolname); 6820} 6821 6822static void 6823zfsdev_init(void) 6824{ 6825 zfsdev = make_dev(&zfs_cdevsw, 0x0, UID_ROOT, GID_OPERATOR, 0666, 6826 ZFS_DEV_NAME); 6827} 6828 6829static void 6830zfsdev_fini(void) 6831{ 6832 if (zfsdev != NULL) 6833 destroy_dev(zfsdev); 6834} 6835 6836static struct root_hold_token *zfs_root_token; 6837struct proc *zfsproc; 6838 6839#ifdef illumos 6840int 6841_init(void) 6842{ 6843 int error; 6844 6845 spa_init(FREAD | FWRITE); 6846 zfs_init(); 6847 zvol_init(); 6848 zfs_ioctl_init(); 6849 6850 if ((error = mod_install(&modlinkage)) != 0) { 6851 zvol_fini(); 6852 zfs_fini(); 6853 spa_fini(); 6854 return (error); 6855 } 6856 6857 tsd_create(&zfs_fsyncer_key, NULL); 6858 tsd_create(&rrw_tsd_key, rrw_tsd_destroy); 6859 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy); 6860 6861 error = ldi_ident_from_mod(&modlinkage, &zfs_li); 6862 ASSERT(error == 0); 6863 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL); 6864 6865 return (0); 6866} 6867 6868int 6869_fini(void) 6870{ 6871 int error; 6872 6873 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled) 6874 return (SET_ERROR(EBUSY)); 6875 6876 if ((error = mod_remove(&modlinkage)) != 0) 6877 return (error); 6878 6879 zvol_fini(); 6880 zfs_fini(); 6881 spa_fini(); 6882 if (zfs_nfsshare_inited) 6883 (void) ddi_modclose(nfs_mod); 6884 if (zfs_smbshare_inited) 6885 (void) ddi_modclose(smbsrv_mod); 6886 if (zfs_nfsshare_inited || zfs_smbshare_inited) 6887 (void) ddi_modclose(sharefs_mod); 6888 6889 tsd_destroy(&zfs_fsyncer_key); 6890 ldi_ident_release(zfs_li); 6891 zfs_li = NULL; 6892 mutex_destroy(&zfs_share_lock); 6893 6894 return (error); 6895} 6896 6897int 6898_info(struct modinfo *modinfop) 6899{ 6900 return (mod_info(&modlinkage, modinfop)); 6901} 6902#endif /* illumos */ 6903 6904static int zfs__init(void); 6905static int zfs__fini(void); 6906static void zfs_shutdown(void *, int); 6907 6908static eventhandler_tag zfs_shutdown_event_tag; 6909 6910#ifdef __FreeBSD__ 6911#define ZFS_MIN_KSTACK_PAGES 4 6912#endif 6913 6914int 6915zfs__init(void) 6916{ 6917 6918#ifdef __FreeBSD__ 6919#if KSTACK_PAGES < ZFS_MIN_KSTACK_PAGES 6920 printf("ZFS NOTICE: KSTACK_PAGES is %d which could result in stack " 6921 "overflow panic!\nPlease consider adding " 6922 "'options KSTACK_PAGES=%d' to your kernel config\n", KSTACK_PAGES, 6923 ZFS_MIN_KSTACK_PAGES); 6924#endif 6925#endif 6926 zfs_root_token = root_mount_hold("ZFS"); 6927 6928 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL); 6929 6930 spa_init(FREAD | FWRITE); 6931 zfs_init(); 6932 zvol_init(); 6933 zfs_ioctl_init(); 6934 6935 tsd_create(&zfs_fsyncer_key, NULL); 6936 tsd_create(&rrw_tsd_key, rrw_tsd_destroy); 6937 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy); 6938 tsd_create(&zfs_geom_probe_vdev_key, NULL); 6939 6940 printf("ZFS storage pool version: features support (" SPA_VERSION_STRING ")\n"); 6941 root_mount_rel(zfs_root_token); 6942 6943 zfsdev_init(); 6944 6945 return (0); 6946} 6947 6948int 6949zfs__fini(void) 6950{ 6951 if (spa_busy() || zfs_busy() || zvol_busy() || 6952 zio_injection_enabled) { 6953 return (EBUSY); 6954 } 6955 6956 zfsdev_fini(); 6957 zvol_fini(); 6958 zfs_fini(); 6959 spa_fini(); 6960 6961 tsd_destroy(&zfs_fsyncer_key); 6962 tsd_destroy(&rrw_tsd_key); 6963 tsd_destroy(&zfs_allow_log_key); 6964 6965 mutex_destroy(&zfs_share_lock); 6966 6967 return (0); 6968} 6969 6970static void 6971zfs_shutdown(void *arg __unused, int howto __unused) 6972{ 6973 6974 /* 6975 * ZFS fini routines can not properly work in a panic-ed system. 6976 */ 6977 if (panicstr == NULL) 6978 (void)zfs__fini(); 6979} 6980 6981 6982static int 6983zfs_modevent(module_t mod, int type, void *unused __unused) 6984{ 6985 int err; 6986 6987 switch (type) { 6988 case MOD_LOAD: 6989 err = zfs__init(); 6990 if (err == 0) 6991 zfs_shutdown_event_tag = EVENTHANDLER_REGISTER( 6992 shutdown_post_sync, zfs_shutdown, NULL, 6993 SHUTDOWN_PRI_FIRST); 6994 return (err); 6995 case MOD_UNLOAD: 6996 err = zfs__fini(); 6997 if (err == 0 && zfs_shutdown_event_tag != NULL) 6998 EVENTHANDLER_DEREGISTER(shutdown_post_sync, 6999 zfs_shutdown_event_tag); 7000 return (err); 7001 case MOD_SHUTDOWN: 7002 return (0); 7003 default: 7004 break; 7005 } 7006 return (EOPNOTSUPP); 7007} 7008 7009static moduledata_t zfs_mod = { 7010 "zfsctrl", 7011 zfs_modevent, 7012 0 7013}; 7014DECLARE_MODULE(zfsctrl, zfs_mod, SI_SUB_VFS, SI_ORDER_ANY); 7015MODULE_VERSION(zfsctrl, 1); 7016MODULE_DEPEND(zfsctrl, opensolaris, 1, 1, 1); 7017MODULE_DEPEND(zfsctrl, krpc, 1, 1, 1); 7018MODULE_DEPEND(zfsctrl, acl_nfs4, 1, 1, 1); 7019