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 <pawel@dawidek.net>. 25 * All rights reserved. 26 * Copyright 2013 Martin Matuska <mm@FreeBSD.org>. All rights reserved. 27 * Copyright 2011 Nexenta Systems, Inc. All rights reserved. 28 * Copyright (c) 2012, Joyent, Inc. All rights reserved. 29 * Copyright (c) 2013 by Delphix. All rights reserved. 30 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved. 31 * Copyright (c) 2013 Steven Hartland. All rights reserved. 32 */ 33 34/* 35 * ZFS ioctls. 36 * 37 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage 38 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool. 39 * 40 * There are two ways that we handle ioctls: the legacy way where almost 41 * all of the logic is in the ioctl callback, and the new way where most 42 * of the marshalling is handled in the common entry point, zfsdev_ioctl(). 43 * 44 * Non-legacy ioctls should be registered by calling 45 * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked 46 * from userland by lzc_ioctl(). 47 * 48 * The registration arguments are as follows: 49 * 50 * const char *name 51 * The name of the ioctl. This is used for history logging. If the 52 * ioctl returns successfully (the callback returns 0), and allow_log 53 * is true, then a history log entry will be recorded with the input & 54 * output nvlists. The log entry can be printed with "zpool history -i". 55 * 56 * zfs_ioc_t ioc 57 * The ioctl request number, which userland will pass to ioctl(2). 58 * The ioctl numbers can change from release to release, because 59 * the caller (libzfs) must be matched to the kernel. 60 * 61 * zfs_secpolicy_func_t *secpolicy 62 * This function will be called before the zfs_ioc_func_t, to 63 * determine if this operation is permitted. It should return EPERM 64 * on failure, and 0 on success. Checks include determining if the 65 * dataset is visible in this zone, and if the user has either all 66 * zfs privileges in the zone (SYS_MOUNT), or has been granted permission 67 * to do this operation on this dataset with "zfs allow". 68 * 69 * zfs_ioc_namecheck_t namecheck 70 * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool 71 * name, a dataset name, or nothing. If the name is not well-formed, 72 * the ioctl will fail and the callback will not be called. 73 * Therefore, the callback can assume that the name is well-formed 74 * (e.g. is null-terminated, doesn't have more than one '@' character, 75 * doesn't have invalid characters). 76 * 77 * zfs_ioc_poolcheck_t pool_check 78 * This specifies requirements on the pool state. If the pool does 79 * not meet them (is suspended or is readonly), the ioctl will fail 80 * and the callback will not be called. If any checks are specified 81 * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME. 82 * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED | 83 * POOL_CHECK_READONLY). 84 * 85 * boolean_t smush_outnvlist 86 * If smush_outnvlist is true, then the output is presumed to be a 87 * list of errors, and it will be "smushed" down to fit into the 88 * caller's buffer, by removing some entries and replacing them with a 89 * single "N_MORE_ERRORS" entry indicating how many were removed. See 90 * nvlist_smush() for details. If smush_outnvlist is false, and the 91 * outnvlist does not fit into the userland-provided buffer, then the 92 * ioctl will fail with ENOMEM. 93 * 94 * zfs_ioc_func_t *func 95 * The callback function that will perform the operation. 96 * 97 * The callback should return 0 on success, or an error number on 98 * failure. If the function fails, the userland ioctl will return -1, 99 * and errno will be set to the callback's return value. The callback 100 * will be called with the following arguments: 101 * 102 * const char *name 103 * The name of the pool or dataset to operate on, from 104 * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the 105 * expected type (pool, dataset, or none). 106 * 107 * nvlist_t *innvl 108 * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or 109 * NULL if no input nvlist was provided. Changes to this nvlist are 110 * ignored. If the input nvlist could not be deserialized, the 111 * ioctl will fail and the callback will not be called. 112 * 113 * nvlist_t *outnvl 114 * The output nvlist, initially empty. The callback can fill it in, 115 * and it will be returned to userland by serializing it into 116 * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization 117 * fails (e.g. because the caller didn't supply a large enough 118 * buffer), then the overall ioctl will fail. See the 119 * 'smush_nvlist' argument above for additional behaviors. 120 * 121 * There are two typical uses of the output nvlist: 122 * - To return state, e.g. property values. In this case, 123 * smush_outnvlist should be false. If the buffer was not large 124 * enough, the caller will reallocate a larger buffer and try 125 * the ioctl again. 126 * 127 * - To return multiple errors from an ioctl which makes on-disk 128 * changes. In this case, smush_outnvlist should be true. 129 * Ioctls which make on-disk modifications should generally not 130 * use the outnvl if they succeed, because the caller can not 131 * distinguish between the operation failing, and 132 * deserialization failing. 133 */ 134 135#include <sys/types.h> 136#include <sys/param.h> 137#include <sys/systm.h> 138#include <sys/conf.h> 139#include <sys/kernel.h> 140#include <sys/lock.h> 141#include <sys/malloc.h> 142#include <sys/mutex.h> 143#include <sys/proc.h> 144#include <sys/errno.h> 145#include <sys/uio.h> 146#include <sys/buf.h> 147#include <sys/file.h> 148#include <sys/kmem.h> 149#include <sys/conf.h> 150#include <sys/cmn_err.h> 151#include <sys/stat.h> 152#include <sys/zfs_ioctl.h> 153#include <sys/zfs_vfsops.h> 154#include <sys/zfs_znode.h> 155#include <sys/zap.h> 156#include <sys/spa.h> 157#include <sys/spa_impl.h> 158#include <sys/vdev.h> 159#include <sys/dmu.h> 160#include <sys/dsl_dir.h> 161#include <sys/dsl_dataset.h> 162#include <sys/dsl_prop.h> 163#include <sys/dsl_deleg.h> 164#include <sys/dmu_objset.h> 165#include <sys/dmu_impl.h> 166#include <sys/dmu_tx.h> 167#include <sys/sunddi.h> 168#include <sys/policy.h> 169#include <sys/zone.h> 170#include <sys/nvpair.h> 171#include <sys/mount.h> 172#include <sys/taskqueue.h> 173#include <sys/sdt.h> 174#include <sys/varargs.h> 175#include <sys/fs/zfs.h> 176#include <sys/zfs_ctldir.h> 177#include <sys/zfs_dir.h> 178#include <sys/zfs_onexit.h> 179#include <sys/zvol.h> 180#include <sys/dsl_scan.h> 181#include <sys/dmu_objset.h> 182#include <sys/dmu_send.h> 183#include <sys/dsl_destroy.h> 184#include <sys/dsl_userhold.h> 185#include <sys/zfeature.h> 186 187#include "zfs_namecheck.h" 188#include "zfs_prop.h" 189#include "zfs_deleg.h" 190#include "zfs_comutil.h" 191#include "zfs_ioctl_compat.h" 192 193CTASSERT(sizeof(zfs_cmd_t) < IOCPARM_MAX); 194 195static int snapshot_list_prefetch; 196SYSCTL_DECL(_vfs_zfs); 197TUNABLE_INT("vfs.zfs.snapshot_list_prefetch", &snapshot_list_prefetch); 198SYSCTL_INT(_vfs_zfs, OID_AUTO, snapshot_list_prefetch, CTLFLAG_RW, 199 &snapshot_list_prefetch, 0, "Prefetch data when listing snapshots"); 200 201static struct cdev *zfsdev; 202 203extern void zfs_init(void); 204extern void zfs_fini(void); 205 206uint_t zfs_fsyncer_key; 207extern uint_t rrw_tsd_key; 208static uint_t zfs_allow_log_key; 209 210typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *); 211typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *); 212typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *); 213 214typedef enum { 215 NO_NAME, 216 POOL_NAME, 217 DATASET_NAME 218} zfs_ioc_namecheck_t; 219 220typedef enum { 221 POOL_CHECK_NONE = 1 << 0, 222 POOL_CHECK_SUSPENDED = 1 << 1, 223 POOL_CHECK_READONLY = 1 << 2, 224} zfs_ioc_poolcheck_t; 225 226typedef struct zfs_ioc_vec { 227 zfs_ioc_legacy_func_t *zvec_legacy_func; 228 zfs_ioc_func_t *zvec_func; 229 zfs_secpolicy_func_t *zvec_secpolicy; 230 zfs_ioc_namecheck_t zvec_namecheck; 231 boolean_t zvec_allow_log; 232 zfs_ioc_poolcheck_t zvec_pool_check; 233 boolean_t zvec_smush_outnvlist; 234 const char *zvec_name; 235} zfs_ioc_vec_t; 236 237/* This array is indexed by zfs_userquota_prop_t */ 238static const char *userquota_perms[] = { 239 ZFS_DELEG_PERM_USERUSED, 240 ZFS_DELEG_PERM_USERQUOTA, 241 ZFS_DELEG_PERM_GROUPUSED, 242 ZFS_DELEG_PERM_GROUPQUOTA, 243}; 244 245static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc); 246static int zfs_check_settable(const char *name, nvpair_t *property, 247 cred_t *cr); 248static int zfs_check_clearable(char *dataset, nvlist_t *props, 249 nvlist_t **errors); 250static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *, 251 boolean_t *); 252int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *); 253static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp); 254 255static void zfsdev_close(void *data); 256 257static int zfs_prop_activate_feature(spa_t *spa, zfeature_info_t *feature); 258 259/* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */ 260void 261__dprintf(const char *file, const char *func, int line, const char *fmt, ...) 262{ 263 const char *newfile; 264 char buf[512]; 265 va_list adx; 266 267 /* 268 * Get rid of annoying "../common/" prefix to filename. 269 */ 270 newfile = strrchr(file, '/'); 271 if (newfile != NULL) { 272 newfile = newfile + 1; /* Get rid of leading / */ 273 } else { 274 newfile = file; 275 } 276 277 va_start(adx, fmt); 278 (void) vsnprintf(buf, sizeof (buf), fmt, adx); 279 va_end(adx); 280 281 /* 282 * To get this data, use the zfs-dprintf probe as so: 283 * dtrace -q -n 'zfs-dprintf \ 284 * /stringof(arg0) == "dbuf.c"/ \ 285 * {printf("%s: %s", stringof(arg1), stringof(arg3))}' 286 * arg0 = file name 287 * arg1 = function name 288 * arg2 = line number 289 * arg3 = message 290 */ 291 DTRACE_PROBE4(zfs__dprintf, 292 char *, newfile, char *, func, int, line, char *, buf); 293} 294 295static void 296history_str_free(char *buf) 297{ 298 kmem_free(buf, HIS_MAX_RECORD_LEN); 299} 300 301static char * 302history_str_get(zfs_cmd_t *zc) 303{ 304 char *buf; 305 306 if (zc->zc_history == 0) 307 return (NULL); 308 309 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP); 310 if (copyinstr((void *)(uintptr_t)zc->zc_history, 311 buf, HIS_MAX_RECORD_LEN, NULL) != 0) { 312 history_str_free(buf); 313 return (NULL); 314 } 315 316 buf[HIS_MAX_RECORD_LEN -1] = '\0'; 317 318 return (buf); 319} 320 321/* 322 * Check to see if the named dataset is currently defined as bootable 323 */ 324static boolean_t 325zfs_is_bootfs(const char *name) 326{ 327 objset_t *os; 328 329 if (dmu_objset_hold(name, FTAG, &os) == 0) { 330 boolean_t ret; 331 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os))); 332 dmu_objset_rele(os, FTAG); 333 return (ret); 334 } 335 return (B_FALSE); 336} 337 338/* 339 * Return non-zero if the spa version is less than requested version. 340 */ 341static int 342zfs_earlier_version(const char *name, int version) 343{ 344 spa_t *spa; 345 346 if (spa_open(name, &spa, FTAG) == 0) { 347 if (spa_version(spa) < version) { 348 spa_close(spa, FTAG); 349 return (1); 350 } 351 spa_close(spa, FTAG); 352 } 353 return (0); 354} 355 356/* 357 * Return TRUE if the ZPL version is less than requested version. 358 */ 359static boolean_t 360zpl_earlier_version(const char *name, int version) 361{ 362 objset_t *os; 363 boolean_t rc = B_TRUE; 364 365 if (dmu_objset_hold(name, FTAG, &os) == 0) { 366 uint64_t zplversion; 367 368 if (dmu_objset_type(os) != DMU_OST_ZFS) { 369 dmu_objset_rele(os, FTAG); 370 return (B_TRUE); 371 } 372 /* XXX reading from non-owned objset */ 373 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0) 374 rc = zplversion < version; 375 dmu_objset_rele(os, FTAG); 376 } 377 return (rc); 378} 379 380static void 381zfs_log_history(zfs_cmd_t *zc) 382{ 383 spa_t *spa; 384 char *buf; 385 386 if ((buf = history_str_get(zc)) == NULL) 387 return; 388 389 if (spa_open(zc->zc_name, &spa, FTAG) == 0) { 390 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY) 391 (void) spa_history_log(spa, buf); 392 spa_close(spa, FTAG); 393 } 394 history_str_free(buf); 395} 396 397/* 398 * Policy for top-level read operations (list pools). Requires no privileges, 399 * and can be used in the local zone, as there is no associated dataset. 400 */ 401/* ARGSUSED */ 402static int 403zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 404{ 405 return (0); 406} 407 408/* 409 * Policy for dataset read operations (list children, get statistics). Requires 410 * no privileges, but must be visible in the local zone. 411 */ 412/* ARGSUSED */ 413static int 414zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 415{ 416 if (INGLOBALZONE(curthread) || 417 zone_dataset_visible(zc->zc_name, NULL)) 418 return (0); 419 420 return (SET_ERROR(ENOENT)); 421} 422 423static int 424zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr) 425{ 426 int writable = 1; 427 428 /* 429 * The dataset must be visible by this zone -- check this first 430 * so they don't see EPERM on something they shouldn't know about. 431 */ 432 if (!INGLOBALZONE(curthread) && 433 !zone_dataset_visible(dataset, &writable)) 434 return (SET_ERROR(ENOENT)); 435 436 if (INGLOBALZONE(curthread)) { 437 /* 438 * If the fs is zoned, only root can access it from the 439 * global zone. 440 */ 441 if (secpolicy_zfs(cr) && zoned) 442 return (SET_ERROR(EPERM)); 443 } else { 444 /* 445 * If we are in a local zone, the 'zoned' property must be set. 446 */ 447 if (!zoned) 448 return (SET_ERROR(EPERM)); 449 450 /* must be writable by this zone */ 451 if (!writable) 452 return (SET_ERROR(EPERM)); 453 } 454 return (0); 455} 456 457static int 458zfs_dozonecheck(const char *dataset, cred_t *cr) 459{ 460 uint64_t zoned; 461 462 if (dsl_prop_get_integer(dataset, "jailed", &zoned, NULL)) 463 return (SET_ERROR(ENOENT)); 464 465 return (zfs_dozonecheck_impl(dataset, zoned, cr)); 466} 467 468static int 469zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr) 470{ 471 uint64_t zoned; 472 473 if (dsl_prop_get_int_ds(ds, "jailed", &zoned)) 474 return (SET_ERROR(ENOENT)); 475 476 return (zfs_dozonecheck_impl(dataset, zoned, cr)); 477} 478 479static int 480zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds, 481 const char *perm, cred_t *cr) 482{ 483 int error; 484 485 error = zfs_dozonecheck_ds(name, ds, cr); 486 if (error == 0) { 487 error = secpolicy_zfs(cr); 488 if (error != 0) 489 error = dsl_deleg_access_impl(ds, perm, cr); 490 } 491 return (error); 492} 493 494static int 495zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr) 496{ 497 int error; 498 dsl_dataset_t *ds; 499 dsl_pool_t *dp; 500 501 error = dsl_pool_hold(name, FTAG, &dp); 502 if (error != 0) 503 return (error); 504 505 error = dsl_dataset_hold(dp, name, FTAG, &ds); 506 if (error != 0) { 507 dsl_pool_rele(dp, FTAG); 508 return (error); 509 } 510 511 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr); 512 513 dsl_dataset_rele(ds, FTAG); 514 dsl_pool_rele(dp, FTAG); 515 return (error); 516} 517 518#ifdef SECLABEL 519/* 520 * Policy for setting the security label property. 521 * 522 * Returns 0 for success, non-zero for access and other errors. 523 */ 524static int 525zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr) 526{ 527 char ds_hexsl[MAXNAMELEN]; 528 bslabel_t ds_sl, new_sl; 529 boolean_t new_default = FALSE; 530 uint64_t zoned; 531 int needed_priv = -1; 532 int error; 533 534 /* First get the existing dataset label. */ 535 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL), 536 1, sizeof (ds_hexsl), &ds_hexsl, NULL); 537 if (error != 0) 538 return (SET_ERROR(EPERM)); 539 540 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0) 541 new_default = TRUE; 542 543 /* The label must be translatable */ 544 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0)) 545 return (SET_ERROR(EINVAL)); 546 547 /* 548 * In a non-global zone, disallow attempts to set a label that 549 * doesn't match that of the zone; otherwise no other checks 550 * are needed. 551 */ 552 if (!INGLOBALZONE(curproc)) { 553 if (new_default || !blequal(&new_sl, CR_SL(CRED()))) 554 return (SET_ERROR(EPERM)); 555 return (0); 556 } 557 558 /* 559 * For global-zone datasets (i.e., those whose zoned property is 560 * "off", verify that the specified new label is valid for the 561 * global zone. 562 */ 563 if (dsl_prop_get_integer(name, 564 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL)) 565 return (SET_ERROR(EPERM)); 566 if (!zoned) { 567 if (zfs_check_global_label(name, strval) != 0) 568 return (SET_ERROR(EPERM)); 569 } 570 571 /* 572 * If the existing dataset label is nondefault, check if the 573 * dataset is mounted (label cannot be changed while mounted). 574 * Get the zfsvfs; if there isn't one, then the dataset isn't 575 * mounted (or isn't a dataset, doesn't exist, ...). 576 */ 577 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) { 578 objset_t *os; 579 static char *setsl_tag = "setsl_tag"; 580 581 /* 582 * Try to own the dataset; abort if there is any error, 583 * (e.g., already mounted, in use, or other error). 584 */ 585 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE, 586 setsl_tag, &os); 587 if (error != 0) 588 return (SET_ERROR(EPERM)); 589 590 dmu_objset_disown(os, setsl_tag); 591 592 if (new_default) { 593 needed_priv = PRIV_FILE_DOWNGRADE_SL; 594 goto out_check; 595 } 596 597 if (hexstr_to_label(strval, &new_sl) != 0) 598 return (SET_ERROR(EPERM)); 599 600 if (blstrictdom(&ds_sl, &new_sl)) 601 needed_priv = PRIV_FILE_DOWNGRADE_SL; 602 else if (blstrictdom(&new_sl, &ds_sl)) 603 needed_priv = PRIV_FILE_UPGRADE_SL; 604 } else { 605 /* dataset currently has a default label */ 606 if (!new_default) 607 needed_priv = PRIV_FILE_UPGRADE_SL; 608 } 609 610out_check: 611 if (needed_priv != -1) 612 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL)); 613 return (0); 614} 615#endif /* SECLABEL */ 616 617static int 618zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval, 619 cred_t *cr) 620{ 621 char *strval; 622 623 /* 624 * Check permissions for special properties. 625 */ 626 switch (prop) { 627 case ZFS_PROP_ZONED: 628 /* 629 * Disallow setting of 'zoned' from within a local zone. 630 */ 631 if (!INGLOBALZONE(curthread)) 632 return (SET_ERROR(EPERM)); 633 break; 634 635 case ZFS_PROP_QUOTA: 636 if (!INGLOBALZONE(curthread)) { 637 uint64_t zoned; 638 char setpoint[MAXNAMELEN]; 639 /* 640 * Unprivileged users are allowed to modify the 641 * quota on things *under* (ie. contained by) 642 * the thing they own. 643 */ 644 if (dsl_prop_get_integer(dsname, "jailed", &zoned, 645 setpoint)) 646 return (SET_ERROR(EPERM)); 647 if (!zoned || strlen(dsname) <= strlen(setpoint)) 648 return (SET_ERROR(EPERM)); 649 } 650 break; 651 652 case ZFS_PROP_MLSLABEL: 653#ifdef SECLABEL 654 if (!is_system_labeled()) 655 return (SET_ERROR(EPERM)); 656 657 if (nvpair_value_string(propval, &strval) == 0) { 658 int err; 659 660 err = zfs_set_slabel_policy(dsname, strval, CRED()); 661 if (err != 0) 662 return (err); 663 } 664#else 665 return (EOPNOTSUPP); 666#endif 667 break; 668 } 669 670 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr)); 671} 672 673/* ARGSUSED */ 674static int 675zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 676{ 677 int error; 678 679 error = zfs_dozonecheck(zc->zc_name, cr); 680 if (error != 0) 681 return (error); 682 683 /* 684 * permission to set permissions will be evaluated later in 685 * dsl_deleg_can_allow() 686 */ 687 return (0); 688} 689 690/* ARGSUSED */ 691static int 692zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 693{ 694 return (zfs_secpolicy_write_perms(zc->zc_name, 695 ZFS_DELEG_PERM_ROLLBACK, cr)); 696} 697 698/* ARGSUSED */ 699static int 700zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 701{ 702 dsl_pool_t *dp; 703 dsl_dataset_t *ds; 704 char *cp; 705 int error; 706 707 /* 708 * Generate the current snapshot name from the given objsetid, then 709 * use that name for the secpolicy/zone checks. 710 */ 711 cp = strchr(zc->zc_name, '@'); 712 if (cp == NULL) 713 return (SET_ERROR(EINVAL)); 714 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 715 if (error != 0) 716 return (error); 717 718 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds); 719 if (error != 0) { 720 dsl_pool_rele(dp, FTAG); 721 return (error); 722 } 723 724 dsl_dataset_name(ds, zc->zc_name); 725 726 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds, 727 ZFS_DELEG_PERM_SEND, cr); 728 dsl_dataset_rele(ds, FTAG); 729 dsl_pool_rele(dp, FTAG); 730 731 return (error); 732} 733 734/* ARGSUSED */ 735static int 736zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 737{ 738 return (zfs_secpolicy_write_perms(zc->zc_name, 739 ZFS_DELEG_PERM_SEND, cr)); 740} 741 742/* ARGSUSED */ 743static int 744zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 745{ 746 vnode_t *vp; 747 int error; 748 749 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, 750 NO_FOLLOW, NULL, &vp)) != 0) 751 return (error); 752 753 /* Now make sure mntpnt and dataset are ZFS */ 754 755 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 || 756 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), 757 zc->zc_name) != 0)) { 758 VN_RELE(vp); 759 return (SET_ERROR(EPERM)); 760 } 761 762 VN_RELE(vp); 763 return (dsl_deleg_access(zc->zc_name, 764 ZFS_DELEG_PERM_SHARE, cr)); 765} 766 767int 768zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 769{ 770 if (!INGLOBALZONE(curthread)) 771 return (SET_ERROR(EPERM)); 772 773 if (secpolicy_nfs(cr) == 0) { 774 return (0); 775 } else { 776 return (zfs_secpolicy_deleg_share(zc, innvl, cr)); 777 } 778} 779 780int 781zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 782{ 783 if (!INGLOBALZONE(curthread)) 784 return (SET_ERROR(EPERM)); 785 786 if (secpolicy_smb(cr) == 0) { 787 return (0); 788 } else { 789 return (zfs_secpolicy_deleg_share(zc, innvl, cr)); 790 } 791} 792 793static int 794zfs_get_parent(const char *datasetname, char *parent, int parentsize) 795{ 796 char *cp; 797 798 /* 799 * Remove the @bla or /bla from the end of the name to get the parent. 800 */ 801 (void) strncpy(parent, datasetname, parentsize); 802 cp = strrchr(parent, '@'); 803 if (cp != NULL) { 804 cp[0] = '\0'; 805 } else { 806 cp = strrchr(parent, '/'); 807 if (cp == NULL) 808 return (SET_ERROR(ENOENT)); 809 cp[0] = '\0'; 810 } 811 812 return (0); 813} 814 815int 816zfs_secpolicy_destroy_perms(const char *name, cred_t *cr) 817{ 818 int error; 819 820 if ((error = zfs_secpolicy_write_perms(name, 821 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 822 return (error); 823 824 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr)); 825} 826 827/* ARGSUSED */ 828static int 829zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 830{ 831 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr)); 832} 833 834/* 835 * Destroying snapshots with delegated permissions requires 836 * descendant mount and destroy permissions. 837 */ 838/* ARGSUSED */ 839static int 840zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 841{ 842 nvlist_t *snaps; 843 nvpair_t *pair, *nextpair; 844 int error = 0; 845 846 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 847 return (SET_ERROR(EINVAL)); 848 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 849 pair = nextpair) { 850 dsl_pool_t *dp; 851 dsl_dataset_t *ds; 852 853 error = dsl_pool_hold(nvpair_name(pair), FTAG, &dp); 854 if (error != 0) 855 break; 856 nextpair = nvlist_next_nvpair(snaps, pair); 857 error = dsl_dataset_hold(dp, nvpair_name(pair), FTAG, &ds); 858 if (error == 0) 859 dsl_dataset_rele(ds, FTAG); 860 dsl_pool_rele(dp, FTAG); 861 862 if (error == 0) { 863 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), 864 cr); 865 } else if (error == ENOENT) { 866 /* 867 * Ignore any snapshots that don't exist (we consider 868 * them "already destroyed"). Remove the name from the 869 * nvl here in case the snapshot is created between 870 * now and when we try to destroy it (in which case 871 * we don't want to destroy it since we haven't 872 * checked for permission). 873 */ 874 fnvlist_remove_nvpair(snaps, pair); 875 error = 0; 876 } 877 if (error != 0) 878 break; 879 } 880 881 return (error); 882} 883 884int 885zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr) 886{ 887 char parentname[MAXNAMELEN]; 888 int error; 889 890 if ((error = zfs_secpolicy_write_perms(from, 891 ZFS_DELEG_PERM_RENAME, cr)) != 0) 892 return (error); 893 894 if ((error = zfs_secpolicy_write_perms(from, 895 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 896 return (error); 897 898 if ((error = zfs_get_parent(to, parentname, 899 sizeof (parentname))) != 0) 900 return (error); 901 902 if ((error = zfs_secpolicy_write_perms(parentname, 903 ZFS_DELEG_PERM_CREATE, cr)) != 0) 904 return (error); 905 906 if ((error = zfs_secpolicy_write_perms(parentname, 907 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 908 return (error); 909 910 return (error); 911} 912 913/* ARGSUSED */ 914static int 915zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 916{ 917 char *at = NULL; 918 int error; 919 920 if ((zc->zc_cookie & 1) != 0) { 921 /* 922 * This is recursive rename, so the starting snapshot might 923 * not exist. Check file system or volume permission instead. 924 */ 925 at = strchr(zc->zc_name, '@'); 926 if (at == NULL) 927 return (EINVAL); 928 *at = '\0'; 929 } 930 931 error = zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr); 932 933 if (at != NULL) 934 *at = '@'; 935 936 return (error); 937} 938 939/* ARGSUSED */ 940static int 941zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 942{ 943 dsl_pool_t *dp; 944 dsl_dataset_t *clone; 945 int error; 946 947 error = zfs_secpolicy_write_perms(zc->zc_name, 948 ZFS_DELEG_PERM_PROMOTE, cr); 949 if (error != 0) 950 return (error); 951 952 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 953 if (error != 0) 954 return (error); 955 956 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone); 957 958 if (error == 0) { 959 char parentname[MAXNAMELEN]; 960 dsl_dataset_t *origin = NULL; 961 dsl_dir_t *dd; 962 dd = clone->ds_dir; 963 964 error = dsl_dataset_hold_obj(dd->dd_pool, 965 dd->dd_phys->dd_origin_obj, FTAG, &origin); 966 if (error != 0) { 967 dsl_dataset_rele(clone, FTAG); 968 dsl_pool_rele(dp, FTAG); 969 return (error); 970 } 971 972 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone, 973 ZFS_DELEG_PERM_MOUNT, cr); 974 975 dsl_dataset_name(origin, parentname); 976 if (error == 0) { 977 error = zfs_secpolicy_write_perms_ds(parentname, origin, 978 ZFS_DELEG_PERM_PROMOTE, cr); 979 } 980 dsl_dataset_rele(clone, FTAG); 981 dsl_dataset_rele(origin, FTAG); 982 } 983 dsl_pool_rele(dp, FTAG); 984 return (error); 985} 986 987/* ARGSUSED */ 988static int 989zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 990{ 991 int error; 992 993 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 994 ZFS_DELEG_PERM_RECEIVE, cr)) != 0) 995 return (error); 996 997 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 998 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 999 return (error); 1000 1001 return (zfs_secpolicy_write_perms(zc->zc_name, 1002 ZFS_DELEG_PERM_CREATE, cr)); 1003} 1004 1005int 1006zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr) 1007{ 1008 return (zfs_secpolicy_write_perms(name, 1009 ZFS_DELEG_PERM_SNAPSHOT, cr)); 1010} 1011 1012/* 1013 * Check for permission to create each snapshot in the nvlist. 1014 */ 1015/* ARGSUSED */ 1016static int 1017zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1018{ 1019 nvlist_t *snaps; 1020 int error; 1021 nvpair_t *pair; 1022 1023 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 1024 return (SET_ERROR(EINVAL)); 1025 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 1026 pair = nvlist_next_nvpair(snaps, pair)) { 1027 char *name = nvpair_name(pair); 1028 char *atp = strchr(name, '@'); 1029 1030 if (atp == NULL) { 1031 error = SET_ERROR(EINVAL); 1032 break; 1033 } 1034 *atp = '\0'; 1035 error = zfs_secpolicy_snapshot_perms(name, cr); 1036 *atp = '@'; 1037 if (error != 0) 1038 break; 1039 } 1040 return (error); 1041} 1042 1043/* ARGSUSED */ 1044static int 1045zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1046{ 1047 /* 1048 * Even root must have a proper TSD so that we know what pool 1049 * to log to. 1050 */ 1051 if (tsd_get(zfs_allow_log_key) == NULL) 1052 return (SET_ERROR(EPERM)); 1053 return (0); 1054} 1055 1056static int 1057zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1058{ 1059 char parentname[MAXNAMELEN]; 1060 int error; 1061 char *origin; 1062 1063 if ((error = zfs_get_parent(zc->zc_name, parentname, 1064 sizeof (parentname))) != 0) 1065 return (error); 1066 1067 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 && 1068 (error = zfs_secpolicy_write_perms(origin, 1069 ZFS_DELEG_PERM_CLONE, cr)) != 0) 1070 return (error); 1071 1072 if ((error = zfs_secpolicy_write_perms(parentname, 1073 ZFS_DELEG_PERM_CREATE, cr)) != 0) 1074 return (error); 1075 1076 return (zfs_secpolicy_write_perms(parentname, 1077 ZFS_DELEG_PERM_MOUNT, cr)); 1078} 1079 1080/* 1081 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires 1082 * SYS_CONFIG privilege, which is not available in a local zone. 1083 */ 1084/* ARGSUSED */ 1085static int 1086zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1087{ 1088 if (secpolicy_sys_config(cr, B_FALSE) != 0) 1089 return (SET_ERROR(EPERM)); 1090 1091 return (0); 1092} 1093 1094/* 1095 * Policy for object to name lookups. 1096 */ 1097/* ARGSUSED */ 1098static int 1099zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1100{ 1101 int error; 1102 1103 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0) 1104 return (0); 1105 1106 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr); 1107 return (error); 1108} 1109 1110/* 1111 * Policy for fault injection. Requires all privileges. 1112 */ 1113/* ARGSUSED */ 1114static int 1115zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1116{ 1117 return (secpolicy_zinject(cr)); 1118} 1119 1120/* ARGSUSED */ 1121static int 1122zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1123{ 1124 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value); 1125 1126 if (prop == ZPROP_INVAL) { 1127 if (!zfs_prop_user(zc->zc_value)) 1128 return (SET_ERROR(EINVAL)); 1129 return (zfs_secpolicy_write_perms(zc->zc_name, 1130 ZFS_DELEG_PERM_USERPROP, cr)); 1131 } else { 1132 return (zfs_secpolicy_setprop(zc->zc_name, prop, 1133 NULL, cr)); 1134 } 1135} 1136 1137static int 1138zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1139{ 1140 int err = zfs_secpolicy_read(zc, innvl, cr); 1141 if (err) 1142 return (err); 1143 1144 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 1145 return (SET_ERROR(EINVAL)); 1146 1147 if (zc->zc_value[0] == 0) { 1148 /* 1149 * They are asking about a posix uid/gid. If it's 1150 * themself, allow it. 1151 */ 1152 if (zc->zc_objset_type == ZFS_PROP_USERUSED || 1153 zc->zc_objset_type == ZFS_PROP_USERQUOTA) { 1154 if (zc->zc_guid == crgetuid(cr)) 1155 return (0); 1156 } else { 1157 if (groupmember(zc->zc_guid, cr)) 1158 return (0); 1159 } 1160 } 1161 1162 return (zfs_secpolicy_write_perms(zc->zc_name, 1163 userquota_perms[zc->zc_objset_type], cr)); 1164} 1165 1166static int 1167zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1168{ 1169 int err = zfs_secpolicy_read(zc, innvl, cr); 1170 if (err) 1171 return (err); 1172 1173 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 1174 return (SET_ERROR(EINVAL)); 1175 1176 return (zfs_secpolicy_write_perms(zc->zc_name, 1177 userquota_perms[zc->zc_objset_type], cr)); 1178} 1179 1180/* ARGSUSED */ 1181static int 1182zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1183{ 1184 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION, 1185 NULL, cr)); 1186} 1187 1188/* ARGSUSED */ 1189static int 1190zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1191{ 1192 nvpair_t *pair; 1193 nvlist_t *holds; 1194 int error; 1195 1196 error = nvlist_lookup_nvlist(innvl, "holds", &holds); 1197 if (error != 0) 1198 return (SET_ERROR(EINVAL)); 1199 1200 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL; 1201 pair = nvlist_next_nvpair(holds, pair)) { 1202 char fsname[MAXNAMELEN]; 1203 error = dmu_fsname(nvpair_name(pair), fsname); 1204 if (error != 0) 1205 return (error); 1206 error = zfs_secpolicy_write_perms(fsname, 1207 ZFS_DELEG_PERM_HOLD, cr); 1208 if (error != 0) 1209 return (error); 1210 } 1211 return (0); 1212} 1213 1214/* ARGSUSED */ 1215static int 1216zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1217{ 1218 nvpair_t *pair; 1219 int error; 1220 1221 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL; 1222 pair = nvlist_next_nvpair(innvl, pair)) { 1223 char fsname[MAXNAMELEN]; 1224 error = dmu_fsname(nvpair_name(pair), fsname); 1225 if (error != 0) 1226 return (error); 1227 error = zfs_secpolicy_write_perms(fsname, 1228 ZFS_DELEG_PERM_RELEASE, cr); 1229 if (error != 0) 1230 return (error); 1231 } 1232 return (0); 1233} 1234 1235/* 1236 * Policy for allowing temporary snapshots to be taken or released 1237 */ 1238static int 1239zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1240{ 1241 /* 1242 * A temporary snapshot is the same as a snapshot, 1243 * hold, destroy and release all rolled into one. 1244 * Delegated diff alone is sufficient that we allow this. 1245 */ 1246 int error; 1247 1248 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 1249 ZFS_DELEG_PERM_DIFF, cr)) == 0) 1250 return (0); 1251 1252 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr); 1253 if (error == 0) 1254 error = zfs_secpolicy_hold(zc, innvl, cr); 1255 if (error == 0) 1256 error = zfs_secpolicy_release(zc, innvl, cr); 1257 if (error == 0) 1258 error = zfs_secpolicy_destroy(zc, innvl, cr); 1259 return (error); 1260} 1261 1262/* 1263 * Returns the nvlist as specified by the user in the zfs_cmd_t. 1264 */ 1265static int 1266get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp) 1267{ 1268 char *packed; 1269 int error; 1270 nvlist_t *list = NULL; 1271 1272 /* 1273 * Read in and unpack the user-supplied nvlist. 1274 */ 1275 if (size == 0) 1276 return (SET_ERROR(EINVAL)); 1277 1278 packed = kmem_alloc(size, KM_SLEEP); 1279 1280 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size, 1281 iflag)) != 0) { 1282 kmem_free(packed, size); 1283 return (error); 1284 } 1285 1286 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) { 1287 kmem_free(packed, size); 1288 return (error); 1289 } 1290 1291 kmem_free(packed, size); 1292 1293 *nvp = list; 1294 return (0); 1295} 1296 1297/* 1298 * Reduce the size of this nvlist until it can be serialized in 'max' bytes. 1299 * Entries will be removed from the end of the nvlist, and one int32 entry 1300 * named "N_MORE_ERRORS" will be added indicating how many entries were 1301 * removed. 1302 */ 1303static int 1304nvlist_smush(nvlist_t *errors, size_t max) 1305{ 1306 size_t size; 1307 1308 size = fnvlist_size(errors); 1309 1310 if (size > max) { 1311 nvpair_t *more_errors; 1312 int n = 0; 1313 1314 if (max < 1024) 1315 return (SET_ERROR(ENOMEM)); 1316 1317 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0); 1318 more_errors = nvlist_prev_nvpair(errors, NULL); 1319 1320 do { 1321 nvpair_t *pair = nvlist_prev_nvpair(errors, 1322 more_errors); 1323 fnvlist_remove_nvpair(errors, pair); 1324 n++; 1325 size = fnvlist_size(errors); 1326 } while (size > max); 1327 1328 fnvlist_remove_nvpair(errors, more_errors); 1329 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n); 1330 ASSERT3U(fnvlist_size(errors), <=, max); 1331 } 1332 1333 return (0); 1334} 1335 1336static int 1337put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl) 1338{ 1339 char *packed = NULL; 1340 int error = 0; 1341 size_t size; 1342 1343 size = fnvlist_size(nvl); 1344 1345 if (size > zc->zc_nvlist_dst_size) { 1346 /* 1347 * Solaris returns ENOMEM here, because even if an error is 1348 * returned from an ioctl(2), new zc_nvlist_dst_size will be 1349 * passed to the userland. This is not the case for FreeBSD. 1350 * We need to return 0, so the kernel will copy the 1351 * zc_nvlist_dst_size back and the userland can discover that a 1352 * bigger buffer is needed. 1353 */ 1354 error = 0; 1355 } else { 1356 packed = fnvlist_pack(nvl, &size); 1357 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst, 1358 size, zc->zc_iflags) != 0) 1359 error = SET_ERROR(EFAULT); 1360 fnvlist_pack_free(packed, size); 1361 } 1362 1363 zc->zc_nvlist_dst_size = size; 1364 zc->zc_nvlist_dst_filled = B_TRUE; 1365 return (error); 1366} 1367 1368static int 1369getzfsvfs(const char *dsname, zfsvfs_t **zfvp) 1370{ 1371 objset_t *os; 1372 int error; 1373 1374 error = dmu_objset_hold(dsname, FTAG, &os); 1375 if (error != 0) 1376 return (error); 1377 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1378 dmu_objset_rele(os, FTAG); 1379 return (SET_ERROR(EINVAL)); 1380 } 1381 1382 mutex_enter(&os->os_user_ptr_lock); 1383 *zfvp = dmu_objset_get_user(os); 1384 if (*zfvp) { 1385 VFS_HOLD((*zfvp)->z_vfs); 1386 } else { 1387 error = SET_ERROR(ESRCH); 1388 } 1389 mutex_exit(&os->os_user_ptr_lock); 1390 dmu_objset_rele(os, FTAG); 1391 return (error); 1392} 1393 1394/* 1395 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which 1396 * case its z_vfs will be NULL, and it will be opened as the owner. 1397 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER, 1398 * which prevents all vnode ops from running. 1399 */ 1400static int 1401zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer) 1402{ 1403 int error = 0; 1404 1405 if (getzfsvfs(name, zfvp) != 0) 1406 error = zfsvfs_create(name, zfvp); 1407 if (error == 0) { 1408 rrw_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER : 1409 RW_READER, tag); 1410 if ((*zfvp)->z_unmounted) { 1411 /* 1412 * XXX we could probably try again, since the unmounting 1413 * thread should be just about to disassociate the 1414 * objset from the zfsvfs. 1415 */ 1416 rrw_exit(&(*zfvp)->z_teardown_lock, tag); 1417 return (SET_ERROR(EBUSY)); 1418 } 1419 } 1420 return (error); 1421} 1422 1423static void 1424zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag) 1425{ 1426 rrw_exit(&zfsvfs->z_teardown_lock, tag); 1427 1428 if (zfsvfs->z_vfs) { 1429 VFS_RELE(zfsvfs->z_vfs); 1430 } else { 1431 dmu_objset_disown(zfsvfs->z_os, zfsvfs); 1432 zfsvfs_free(zfsvfs); 1433 } 1434} 1435 1436static int 1437zfs_ioc_pool_create(zfs_cmd_t *zc) 1438{ 1439 int error; 1440 nvlist_t *config, *props = NULL; 1441 nvlist_t *rootprops = NULL; 1442 nvlist_t *zplprops = NULL; 1443 1444 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1445 zc->zc_iflags, &config)) 1446 return (error); 1447 1448 if (zc->zc_nvlist_src_size != 0 && (error = 1449 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1450 zc->zc_iflags, &props))) { 1451 nvlist_free(config); 1452 return (error); 1453 } 1454 1455 if (props) { 1456 nvlist_t *nvl = NULL; 1457 uint64_t version = SPA_VERSION; 1458 1459 (void) nvlist_lookup_uint64(props, 1460 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version); 1461 if (!SPA_VERSION_IS_SUPPORTED(version)) { 1462 error = SET_ERROR(EINVAL); 1463 goto pool_props_bad; 1464 } 1465 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl); 1466 if (nvl) { 1467 error = nvlist_dup(nvl, &rootprops, KM_SLEEP); 1468 if (error != 0) { 1469 nvlist_free(config); 1470 nvlist_free(props); 1471 return (error); 1472 } 1473 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS); 1474 } 1475 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 1476 error = zfs_fill_zplprops_root(version, rootprops, 1477 zplprops, NULL); 1478 if (error != 0) 1479 goto pool_props_bad; 1480 } 1481 1482 error = spa_create(zc->zc_name, config, props, zplprops); 1483 1484 /* 1485 * Set the remaining root properties 1486 */ 1487 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name, 1488 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0) 1489 (void) spa_destroy(zc->zc_name); 1490 1491pool_props_bad: 1492 nvlist_free(rootprops); 1493 nvlist_free(zplprops); 1494 nvlist_free(config); 1495 nvlist_free(props); 1496 1497 return (error); 1498} 1499 1500static int 1501zfs_ioc_pool_destroy(zfs_cmd_t *zc) 1502{ 1503 int error; 1504 zfs_log_history(zc); 1505 error = spa_destroy(zc->zc_name); 1506 if (error == 0) 1507 zvol_remove_minors(zc->zc_name); 1508 return (error); 1509} 1510 1511static int 1512zfs_ioc_pool_import(zfs_cmd_t *zc) 1513{ 1514 nvlist_t *config, *props = NULL; 1515 uint64_t guid; 1516 int error; 1517 1518 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1519 zc->zc_iflags, &config)) != 0) 1520 return (error); 1521 1522 if (zc->zc_nvlist_src_size != 0 && (error = 1523 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1524 zc->zc_iflags, &props))) { 1525 nvlist_free(config); 1526 return (error); 1527 } 1528 1529 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 || 1530 guid != zc->zc_guid) 1531 error = SET_ERROR(EINVAL); 1532 else 1533 error = spa_import(zc->zc_name, config, props, zc->zc_cookie); 1534 1535 if (zc->zc_nvlist_dst != 0) { 1536 int err; 1537 1538 if ((err = put_nvlist(zc, config)) != 0) 1539 error = err; 1540 } 1541 1542 nvlist_free(config); 1543 1544 if (props) 1545 nvlist_free(props); 1546 1547 return (error); 1548} 1549 1550static int 1551zfs_ioc_pool_export(zfs_cmd_t *zc) 1552{ 1553 int error; 1554 boolean_t force = (boolean_t)zc->zc_cookie; 1555 boolean_t hardforce = (boolean_t)zc->zc_guid; 1556 1557 zfs_log_history(zc); 1558 error = spa_export(zc->zc_name, NULL, force, hardforce); 1559 if (error == 0) 1560 zvol_remove_minors(zc->zc_name); 1561 return (error); 1562} 1563 1564static int 1565zfs_ioc_pool_configs(zfs_cmd_t *zc) 1566{ 1567 nvlist_t *configs; 1568 int error; 1569 1570 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL) 1571 return (SET_ERROR(EEXIST)); 1572 1573 error = put_nvlist(zc, configs); 1574 1575 nvlist_free(configs); 1576 1577 return (error); 1578} 1579 1580/* 1581 * inputs: 1582 * zc_name name of the pool 1583 * 1584 * outputs: 1585 * zc_cookie real errno 1586 * zc_nvlist_dst config nvlist 1587 * zc_nvlist_dst_size size of config nvlist 1588 */ 1589static int 1590zfs_ioc_pool_stats(zfs_cmd_t *zc) 1591{ 1592 nvlist_t *config; 1593 int error; 1594 int ret = 0; 1595 1596 error = spa_get_stats(zc->zc_name, &config, zc->zc_value, 1597 sizeof (zc->zc_value)); 1598 1599 if (config != NULL) { 1600 ret = put_nvlist(zc, config); 1601 nvlist_free(config); 1602 1603 /* 1604 * The config may be present even if 'error' is non-zero. 1605 * In this case we return success, and preserve the real errno 1606 * in 'zc_cookie'. 1607 */ 1608 zc->zc_cookie = error; 1609 } else { 1610 ret = error; 1611 } 1612 1613 return (ret); 1614} 1615 1616/* 1617 * Try to import the given pool, returning pool stats as appropriate so that 1618 * user land knows which devices are available and overall pool health. 1619 */ 1620static int 1621zfs_ioc_pool_tryimport(zfs_cmd_t *zc) 1622{ 1623 nvlist_t *tryconfig, *config; 1624 int error; 1625 1626 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1627 zc->zc_iflags, &tryconfig)) != 0) 1628 return (error); 1629 1630 config = spa_tryimport(tryconfig); 1631 1632 nvlist_free(tryconfig); 1633 1634 if (config == NULL) 1635 return (SET_ERROR(EINVAL)); 1636 1637 error = put_nvlist(zc, config); 1638 nvlist_free(config); 1639 1640 return (error); 1641} 1642 1643/* 1644 * inputs: 1645 * zc_name name of the pool 1646 * zc_cookie scan func (pool_scan_func_t) 1647 */ 1648static int 1649zfs_ioc_pool_scan(zfs_cmd_t *zc) 1650{ 1651 spa_t *spa; 1652 int error; 1653 1654 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1655 return (error); 1656 1657 if (zc->zc_cookie == POOL_SCAN_NONE) 1658 error = spa_scan_stop(spa); 1659 else 1660 error = spa_scan(spa, zc->zc_cookie); 1661 1662 spa_close(spa, FTAG); 1663 1664 return (error); 1665} 1666 1667static int 1668zfs_ioc_pool_freeze(zfs_cmd_t *zc) 1669{ 1670 spa_t *spa; 1671 int error; 1672 1673 error = spa_open(zc->zc_name, &spa, FTAG); 1674 if (error == 0) { 1675 spa_freeze(spa); 1676 spa_close(spa, FTAG); 1677 } 1678 return (error); 1679} 1680 1681static int 1682zfs_ioc_pool_upgrade(zfs_cmd_t *zc) 1683{ 1684 spa_t *spa; 1685 int error; 1686 1687 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1688 return (error); 1689 1690 if (zc->zc_cookie < spa_version(spa) || 1691 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) { 1692 spa_close(spa, FTAG); 1693 return (SET_ERROR(EINVAL)); 1694 } 1695 1696 spa_upgrade(spa, zc->zc_cookie); 1697 spa_close(spa, FTAG); 1698 1699 return (error); 1700} 1701 1702static int 1703zfs_ioc_pool_get_history(zfs_cmd_t *zc) 1704{ 1705 spa_t *spa; 1706 char *hist_buf; 1707 uint64_t size; 1708 int error; 1709 1710 if ((size = zc->zc_history_len) == 0) 1711 return (SET_ERROR(EINVAL)); 1712 1713 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1714 return (error); 1715 1716 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) { 1717 spa_close(spa, FTAG); 1718 return (SET_ERROR(ENOTSUP)); 1719 } 1720 1721 hist_buf = kmem_alloc(size, KM_SLEEP); 1722 if ((error = spa_history_get(spa, &zc->zc_history_offset, 1723 &zc->zc_history_len, hist_buf)) == 0) { 1724 error = ddi_copyout(hist_buf, 1725 (void *)(uintptr_t)zc->zc_history, 1726 zc->zc_history_len, zc->zc_iflags); 1727 } 1728 1729 spa_close(spa, FTAG); 1730 kmem_free(hist_buf, size); 1731 return (error); 1732} 1733 1734static int 1735zfs_ioc_pool_reguid(zfs_cmd_t *zc) 1736{ 1737 spa_t *spa; 1738 int error; 1739 1740 error = spa_open(zc->zc_name, &spa, FTAG); 1741 if (error == 0) { 1742 error = spa_change_guid(spa); 1743 spa_close(spa, FTAG); 1744 } 1745 return (error); 1746} 1747 1748static int 1749zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc) 1750{ 1751 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value)); 1752} 1753 1754/* 1755 * inputs: 1756 * zc_name name of filesystem 1757 * zc_obj object to find 1758 * 1759 * outputs: 1760 * zc_value name of object 1761 */ 1762static int 1763zfs_ioc_obj_to_path(zfs_cmd_t *zc) 1764{ 1765 objset_t *os; 1766 int error; 1767 1768 /* XXX reading from objset not owned */ 1769 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0) 1770 return (error); 1771 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1772 dmu_objset_rele(os, FTAG); 1773 return (SET_ERROR(EINVAL)); 1774 } 1775 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value, 1776 sizeof (zc->zc_value)); 1777 dmu_objset_rele(os, FTAG); 1778 1779 return (error); 1780} 1781 1782/* 1783 * inputs: 1784 * zc_name name of filesystem 1785 * zc_obj object to find 1786 * 1787 * outputs: 1788 * zc_stat stats on object 1789 * zc_value path to object 1790 */ 1791static int 1792zfs_ioc_obj_to_stats(zfs_cmd_t *zc) 1793{ 1794 objset_t *os; 1795 int error; 1796 1797 /* XXX reading from objset not owned */ 1798 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0) 1799 return (error); 1800 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1801 dmu_objset_rele(os, FTAG); 1802 return (SET_ERROR(EINVAL)); 1803 } 1804 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value, 1805 sizeof (zc->zc_value)); 1806 dmu_objset_rele(os, FTAG); 1807 1808 return (error); 1809} 1810 1811static int 1812zfs_ioc_vdev_add(zfs_cmd_t *zc) 1813{ 1814 spa_t *spa; 1815 int error; 1816 nvlist_t *config, **l2cache, **spares; 1817 uint_t nl2cache = 0, nspares = 0; 1818 1819 error = spa_open(zc->zc_name, &spa, FTAG); 1820 if (error != 0) 1821 return (error); 1822 1823 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1824 zc->zc_iflags, &config); 1825 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE, 1826 &l2cache, &nl2cache); 1827 1828 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES, 1829 &spares, &nspares); 1830 1831#ifdef illumos 1832 /* 1833 * A root pool with concatenated devices is not supported. 1834 * Thus, can not add a device to a root pool. 1835 * 1836 * Intent log device can not be added to a rootpool because 1837 * during mountroot, zil is replayed, a seperated log device 1838 * can not be accessed during the mountroot time. 1839 * 1840 * l2cache and spare devices are ok to be added to a rootpool. 1841 */ 1842 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) { 1843 nvlist_free(config); 1844 spa_close(spa, FTAG); 1845 return (SET_ERROR(EDOM)); 1846 } 1847#endif /* illumos */ 1848 1849 if (error == 0) { 1850 error = spa_vdev_add(spa, config); 1851 nvlist_free(config); 1852 } 1853 spa_close(spa, FTAG); 1854 return (error); 1855} 1856 1857/* 1858 * inputs: 1859 * zc_name name of the pool 1860 * zc_nvlist_conf nvlist of devices to remove 1861 * zc_cookie to stop the remove? 1862 */ 1863static int 1864zfs_ioc_vdev_remove(zfs_cmd_t *zc) 1865{ 1866 spa_t *spa; 1867 int error; 1868 1869 error = spa_open(zc->zc_name, &spa, FTAG); 1870 if (error != 0) 1871 return (error); 1872 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE); 1873 spa_close(spa, FTAG); 1874 return (error); 1875} 1876 1877static int 1878zfs_ioc_vdev_set_state(zfs_cmd_t *zc) 1879{ 1880 spa_t *spa; 1881 int error; 1882 vdev_state_t newstate = VDEV_STATE_UNKNOWN; 1883 1884 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1885 return (error); 1886 switch (zc->zc_cookie) { 1887 case VDEV_STATE_ONLINE: 1888 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate); 1889 break; 1890 1891 case VDEV_STATE_OFFLINE: 1892 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj); 1893 break; 1894 1895 case VDEV_STATE_FAULTED: 1896 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED && 1897 zc->zc_obj != VDEV_AUX_EXTERNAL) 1898 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED; 1899 1900 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj); 1901 break; 1902 1903 case VDEV_STATE_DEGRADED: 1904 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED && 1905 zc->zc_obj != VDEV_AUX_EXTERNAL) 1906 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED; 1907 1908 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj); 1909 break; 1910 1911 default: 1912 error = SET_ERROR(EINVAL); 1913 } 1914 zc->zc_cookie = newstate; 1915 spa_close(spa, FTAG); 1916 return (error); 1917} 1918 1919static int 1920zfs_ioc_vdev_attach(zfs_cmd_t *zc) 1921{ 1922 spa_t *spa; 1923 int replacing = zc->zc_cookie; 1924 nvlist_t *config; 1925 int error; 1926 1927 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1928 return (error); 1929 1930 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1931 zc->zc_iflags, &config)) == 0) { 1932 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing); 1933 nvlist_free(config); 1934 } 1935 1936 spa_close(spa, FTAG); 1937 return (error); 1938} 1939 1940static int 1941zfs_ioc_vdev_detach(zfs_cmd_t *zc) 1942{ 1943 spa_t *spa; 1944 int error; 1945 1946 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1947 return (error); 1948 1949 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE); 1950 1951 spa_close(spa, FTAG); 1952 return (error); 1953} 1954 1955static int 1956zfs_ioc_vdev_split(zfs_cmd_t *zc) 1957{ 1958 spa_t *spa; 1959 nvlist_t *config, *props = NULL; 1960 int error; 1961 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT); 1962 1963 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1964 return (error); 1965 1966 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1967 zc->zc_iflags, &config)) { 1968 spa_close(spa, FTAG); 1969 return (error); 1970 } 1971 1972 if (zc->zc_nvlist_src_size != 0 && (error = 1973 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1974 zc->zc_iflags, &props))) { 1975 spa_close(spa, FTAG); 1976 nvlist_free(config); 1977 return (error); 1978 } 1979 1980 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp); 1981 1982 spa_close(spa, FTAG); 1983 1984 nvlist_free(config); 1985 nvlist_free(props); 1986 1987 return (error); 1988} 1989 1990static int 1991zfs_ioc_vdev_setpath(zfs_cmd_t *zc) 1992{ 1993 spa_t *spa; 1994 char *path = zc->zc_value; 1995 uint64_t guid = zc->zc_guid; 1996 int error; 1997 1998 error = spa_open(zc->zc_name, &spa, FTAG); 1999 if (error != 0) 2000 return (error); 2001 2002 error = spa_vdev_setpath(spa, guid, path); 2003 spa_close(spa, FTAG); 2004 return (error); 2005} 2006 2007static int 2008zfs_ioc_vdev_setfru(zfs_cmd_t *zc) 2009{ 2010 spa_t *spa; 2011 char *fru = zc->zc_value; 2012 uint64_t guid = zc->zc_guid; 2013 int error; 2014 2015 error = spa_open(zc->zc_name, &spa, FTAG); 2016 if (error != 0) 2017 return (error); 2018 2019 error = spa_vdev_setfru(spa, guid, fru); 2020 spa_close(spa, FTAG); 2021 return (error); 2022} 2023 2024static int 2025zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os) 2026{ 2027 int error = 0; 2028 nvlist_t *nv; 2029 2030 dmu_objset_fast_stat(os, &zc->zc_objset_stats); 2031 2032 if (zc->zc_nvlist_dst != 0 && 2033 (error = dsl_prop_get_all(os, &nv)) == 0) { 2034 dmu_objset_stats(os, nv); 2035 /* 2036 * NB: zvol_get_stats() will read the objset contents, 2037 * which we aren't supposed to do with a 2038 * DS_MODE_USER hold, because it could be 2039 * inconsistent. So this is a bit of a workaround... 2040 * XXX reading with out owning 2041 */ 2042 if (!zc->zc_objset_stats.dds_inconsistent && 2043 dmu_objset_type(os) == DMU_OST_ZVOL) { 2044 error = zvol_get_stats(os, nv); 2045 if (error == EIO) 2046 return (error); 2047 VERIFY0(error); 2048 } 2049 error = put_nvlist(zc, nv); 2050 nvlist_free(nv); 2051 } 2052 2053 return (error); 2054} 2055 2056/* 2057 * inputs: 2058 * zc_name name of filesystem 2059 * zc_nvlist_dst_size size of buffer for property nvlist 2060 * 2061 * outputs: 2062 * zc_objset_stats stats 2063 * zc_nvlist_dst property nvlist 2064 * zc_nvlist_dst_size size of property nvlist 2065 */ 2066static int 2067zfs_ioc_objset_stats(zfs_cmd_t *zc) 2068{ 2069 objset_t *os; 2070 int error; 2071 2072 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 2073 if (error == 0) { 2074 error = zfs_ioc_objset_stats_impl(zc, os); 2075 dmu_objset_rele(os, FTAG); 2076 } 2077 2078 if (error == ENOMEM) 2079 error = 0; 2080 return (error); 2081} 2082 2083/* 2084 * inputs: 2085 * zc_name name of filesystem 2086 * zc_nvlist_dst_size size of buffer for property nvlist 2087 * 2088 * outputs: 2089 * zc_nvlist_dst received property nvlist 2090 * zc_nvlist_dst_size size of received property nvlist 2091 * 2092 * Gets received properties (distinct from local properties on or after 2093 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from 2094 * local property values. 2095 */ 2096static int 2097zfs_ioc_objset_recvd_props(zfs_cmd_t *zc) 2098{ 2099 int error = 0; 2100 nvlist_t *nv; 2101 2102 /* 2103 * Without this check, we would return local property values if the 2104 * caller has not already received properties on or after 2105 * SPA_VERSION_RECVD_PROPS. 2106 */ 2107 if (!dsl_prop_get_hasrecvd(zc->zc_name)) 2108 return (SET_ERROR(ENOTSUP)); 2109 2110 if (zc->zc_nvlist_dst != 0 && 2111 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) { 2112 error = put_nvlist(zc, nv); 2113 nvlist_free(nv); 2114 } 2115 2116 return (error); 2117} 2118 2119static int 2120nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop) 2121{ 2122 uint64_t value; 2123 int error; 2124 2125 /* 2126 * zfs_get_zplprop() will either find a value or give us 2127 * the default value (if there is one). 2128 */ 2129 if ((error = zfs_get_zplprop(os, prop, &value)) != 0) 2130 return (error); 2131 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0); 2132 return (0); 2133} 2134 2135/* 2136 * inputs: 2137 * zc_name name of filesystem 2138 * zc_nvlist_dst_size size of buffer for zpl property nvlist 2139 * 2140 * outputs: 2141 * zc_nvlist_dst zpl property nvlist 2142 * zc_nvlist_dst_size size of zpl property nvlist 2143 */ 2144static int 2145zfs_ioc_objset_zplprops(zfs_cmd_t *zc) 2146{ 2147 objset_t *os; 2148 int err; 2149 2150 /* XXX reading without owning */ 2151 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os)) 2152 return (err); 2153 2154 dmu_objset_fast_stat(os, &zc->zc_objset_stats); 2155 2156 /* 2157 * NB: nvl_add_zplprop() will read the objset contents, 2158 * which we aren't supposed to do with a DS_MODE_USER 2159 * hold, because it could be inconsistent. 2160 */ 2161 if (zc->zc_nvlist_dst != 0 && 2162 !zc->zc_objset_stats.dds_inconsistent && 2163 dmu_objset_type(os) == DMU_OST_ZFS) { 2164 nvlist_t *nv; 2165 2166 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2167 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 && 2168 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 && 2169 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 && 2170 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0) 2171 err = put_nvlist(zc, nv); 2172 nvlist_free(nv); 2173 } else { 2174 err = SET_ERROR(ENOENT); 2175 } 2176 dmu_objset_rele(os, FTAG); 2177 return (err); 2178} 2179 2180boolean_t 2181dataset_name_hidden(const char *name) 2182{ 2183 /* 2184 * Skip over datasets that are not visible in this zone, 2185 * internal datasets (which have a $ in their name), and 2186 * temporary datasets (which have a % in their name). 2187 */ 2188 if (strchr(name, '$') != NULL) 2189 return (B_TRUE); 2190 if (strchr(name, '%') != NULL) 2191 return (B_TRUE); 2192 if (!INGLOBALZONE(curthread) && !zone_dataset_visible(name, NULL)) 2193 return (B_TRUE); 2194 return (B_FALSE); 2195} 2196 2197/* 2198 * inputs: 2199 * zc_name name of filesystem 2200 * zc_cookie zap cursor 2201 * zc_nvlist_dst_size size of buffer for property nvlist 2202 * 2203 * outputs: 2204 * zc_name name of next filesystem 2205 * zc_cookie zap cursor 2206 * zc_objset_stats stats 2207 * zc_nvlist_dst property nvlist 2208 * zc_nvlist_dst_size size of property nvlist 2209 */ 2210static int 2211zfs_ioc_dataset_list_next(zfs_cmd_t *zc) 2212{ 2213 objset_t *os; 2214 int error; 2215 char *p; 2216 size_t orig_len = strlen(zc->zc_name); 2217 2218top: 2219 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) { 2220 if (error == ENOENT) 2221 error = SET_ERROR(ESRCH); 2222 return (error); 2223 } 2224 2225 p = strrchr(zc->zc_name, '/'); 2226 if (p == NULL || p[1] != '\0') 2227 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name)); 2228 p = zc->zc_name + strlen(zc->zc_name); 2229 2230 do { 2231 error = dmu_dir_list_next(os, 2232 sizeof (zc->zc_name) - (p - zc->zc_name), p, 2233 NULL, &zc->zc_cookie); 2234 if (error == ENOENT) 2235 error = SET_ERROR(ESRCH); 2236 } while (error == 0 && dataset_name_hidden(zc->zc_name)); 2237 dmu_objset_rele(os, FTAG); 2238 2239 /* 2240 * If it's an internal dataset (ie. with a '$' in its name), 2241 * don't try to get stats for it, otherwise we'll return ENOENT. 2242 */ 2243 if (error == 0 && strchr(zc->zc_name, '$') == NULL) { 2244 error = zfs_ioc_objset_stats(zc); /* fill in the stats */ 2245 if (error == ENOENT) { 2246 /* We lost a race with destroy, get the next one. */ 2247 zc->zc_name[orig_len] = '\0'; 2248 goto top; 2249 } 2250 } 2251 return (error); 2252} 2253 2254/* 2255 * inputs: 2256 * zc_name name of filesystem 2257 * zc_cookie zap cursor 2258 * zc_nvlist_dst_size size of buffer for property nvlist 2259 * zc_simple when set, only name is requested 2260 * 2261 * outputs: 2262 * zc_name name of next snapshot 2263 * zc_objset_stats stats 2264 * zc_nvlist_dst property nvlist 2265 * zc_nvlist_dst_size size of property nvlist 2266 */ 2267static int 2268zfs_ioc_snapshot_list_next(zfs_cmd_t *zc) 2269{ 2270 objset_t *os; 2271 int error; 2272 2273 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 2274 if (error != 0) { 2275 return (error == ENOENT ? ESRCH : error); 2276 } 2277 2278 /* 2279 * A dataset name of maximum length cannot have any snapshots, 2280 * so exit immediately. 2281 */ 2282 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) { 2283 dmu_objset_rele(os, FTAG); 2284 return (SET_ERROR(ESRCH)); 2285 } 2286 2287 error = dmu_snapshot_list_next(os, 2288 sizeof (zc->zc_name) - strlen(zc->zc_name), 2289 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie, 2290 NULL); 2291 2292 if (error == 0 && !zc->zc_simple) { 2293 dsl_dataset_t *ds; 2294 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool; 2295 2296 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds); 2297 if (error == 0) { 2298 objset_t *ossnap; 2299 2300 error = dmu_objset_from_ds(ds, &ossnap); 2301 if (error == 0) 2302 error = zfs_ioc_objset_stats_impl(zc, ossnap); 2303 dsl_dataset_rele(ds, FTAG); 2304 } 2305 } else if (error == ENOENT) { 2306 error = SET_ERROR(ESRCH); 2307 } 2308 2309 dmu_objset_rele(os, FTAG); 2310 /* if we failed, undo the @ that we tacked on to zc_name */ 2311 if (error != 0) 2312 *strchr(zc->zc_name, '@') = '\0'; 2313 return (error); 2314} 2315 2316static int 2317zfs_prop_set_userquota(const char *dsname, nvpair_t *pair) 2318{ 2319 const char *propname = nvpair_name(pair); 2320 uint64_t *valary; 2321 unsigned int vallen; 2322 const char *domain; 2323 char *dash; 2324 zfs_userquota_prop_t type; 2325 uint64_t rid; 2326 uint64_t quota; 2327 zfsvfs_t *zfsvfs; 2328 int err; 2329 2330 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2331 nvlist_t *attrs; 2332 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 2333 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2334 &pair) != 0) 2335 return (SET_ERROR(EINVAL)); 2336 } 2337 2338 /* 2339 * A correctly constructed propname is encoded as 2340 * userquota@<rid>-<domain>. 2341 */ 2342 if ((dash = strchr(propname, '-')) == NULL || 2343 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 || 2344 vallen != 3) 2345 return (SET_ERROR(EINVAL)); 2346 2347 domain = dash + 1; 2348 type = valary[0]; 2349 rid = valary[1]; 2350 quota = valary[2]; 2351 2352 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE); 2353 if (err == 0) { 2354 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota); 2355 zfsvfs_rele(zfsvfs, FTAG); 2356 } 2357 2358 return (err); 2359} 2360 2361/* 2362 * If the named property is one that has a special function to set its value, 2363 * return 0 on success and a positive error code on failure; otherwise if it is 2364 * not one of the special properties handled by this function, return -1. 2365 * 2366 * XXX: It would be better for callers of the property interface if we handled 2367 * these special cases in dsl_prop.c (in the dsl layer). 2368 */ 2369static int 2370zfs_prop_set_special(const char *dsname, zprop_source_t source, 2371 nvpair_t *pair) 2372{ 2373 const char *propname = nvpair_name(pair); 2374 zfs_prop_t prop = zfs_name_to_prop(propname); 2375 uint64_t intval; 2376 int err; 2377 2378 if (prop == ZPROP_INVAL) { 2379 if (zfs_prop_userquota(propname)) 2380 return (zfs_prop_set_userquota(dsname, pair)); 2381 return (-1); 2382 } 2383 2384 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2385 nvlist_t *attrs; 2386 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 2387 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2388 &pair) == 0); 2389 } 2390 2391 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING) 2392 return (-1); 2393 2394 VERIFY(0 == nvpair_value_uint64(pair, &intval)); 2395 2396 switch (prop) { 2397 case ZFS_PROP_QUOTA: 2398 err = dsl_dir_set_quota(dsname, source, intval); 2399 break; 2400 case ZFS_PROP_REFQUOTA: 2401 err = dsl_dataset_set_refquota(dsname, source, intval); 2402 break; 2403 case ZFS_PROP_RESERVATION: 2404 err = dsl_dir_set_reservation(dsname, source, intval); 2405 break; 2406 case ZFS_PROP_REFRESERVATION: 2407 err = dsl_dataset_set_refreservation(dsname, source, intval); 2408 break; 2409 case ZFS_PROP_VOLSIZE: 2410 err = zvol_set_volsize(dsname, ddi_driver_major(zfs_dip), 2411 intval); 2412 break; 2413 case ZFS_PROP_VERSION: 2414 { 2415 zfsvfs_t *zfsvfs; 2416 2417 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0) 2418 break; 2419 2420 err = zfs_set_version(zfsvfs, intval); 2421 zfsvfs_rele(zfsvfs, FTAG); 2422 2423 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) { 2424 zfs_cmd_t *zc; 2425 2426 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP); 2427 (void) strcpy(zc->zc_name, dsname); 2428 (void) zfs_ioc_userspace_upgrade(zc); 2429 kmem_free(zc, sizeof (zfs_cmd_t)); 2430 } 2431 break; 2432 } 2433 case ZFS_PROP_COMPRESSION: 2434 { 2435 if (intval == ZIO_COMPRESS_LZ4) { 2436 zfeature_info_t *feature = 2437 &spa_feature_table[SPA_FEATURE_LZ4_COMPRESS]; 2438 spa_t *spa; 2439 2440 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 2441 return (err); 2442 2443 /* 2444 * Setting the LZ4 compression algorithm activates 2445 * the feature. 2446 */ 2447 if (!spa_feature_is_active(spa, feature)) { 2448 if ((err = zfs_prop_activate_feature(spa, 2449 feature)) != 0) { 2450 spa_close(spa, FTAG); 2451 return (err); 2452 } 2453 } 2454 2455 spa_close(spa, FTAG); 2456 } 2457 /* 2458 * We still want the default set action to be performed in the 2459 * caller, we only performed zfeature settings here. 2460 */ 2461 err = -1; 2462 break; 2463 } 2464 2465 default: 2466 err = -1; 2467 } 2468 2469 return (err); 2470} 2471 2472/* 2473 * This function is best effort. If it fails to set any of the given properties, 2474 * it continues to set as many as it can and returns the last error 2475 * encountered. If the caller provides a non-NULL errlist, it will be filled in 2476 * with the list of names of all the properties that failed along with the 2477 * corresponding error numbers. 2478 * 2479 * If every property is set successfully, zero is returned and errlist is not 2480 * modified. 2481 */ 2482int 2483zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl, 2484 nvlist_t *errlist) 2485{ 2486 nvpair_t *pair; 2487 nvpair_t *propval; 2488 int rv = 0; 2489 uint64_t intval; 2490 char *strval; 2491 nvlist_t *genericnvl = fnvlist_alloc(); 2492 nvlist_t *retrynvl = fnvlist_alloc(); 2493 2494retry: 2495 pair = NULL; 2496 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { 2497 const char *propname = nvpair_name(pair); 2498 zfs_prop_t prop = zfs_name_to_prop(propname); 2499 int err = 0; 2500 2501 /* decode the property value */ 2502 propval = pair; 2503 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2504 nvlist_t *attrs; 2505 attrs = fnvpair_value_nvlist(pair); 2506 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2507 &propval) != 0) 2508 err = SET_ERROR(EINVAL); 2509 } 2510 2511 /* Validate value type */ 2512 if (err == 0 && prop == ZPROP_INVAL) { 2513 if (zfs_prop_user(propname)) { 2514 if (nvpair_type(propval) != DATA_TYPE_STRING) 2515 err = SET_ERROR(EINVAL); 2516 } else if (zfs_prop_userquota(propname)) { 2517 if (nvpair_type(propval) != 2518 DATA_TYPE_UINT64_ARRAY) 2519 err = SET_ERROR(EINVAL); 2520 } else { 2521 err = SET_ERROR(EINVAL); 2522 } 2523 } else if (err == 0) { 2524 if (nvpair_type(propval) == DATA_TYPE_STRING) { 2525 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING) 2526 err = SET_ERROR(EINVAL); 2527 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) { 2528 const char *unused; 2529 2530 intval = fnvpair_value_uint64(propval); 2531 2532 switch (zfs_prop_get_type(prop)) { 2533 case PROP_TYPE_NUMBER: 2534 break; 2535 case PROP_TYPE_STRING: 2536 err = SET_ERROR(EINVAL); 2537 break; 2538 case PROP_TYPE_INDEX: 2539 if (zfs_prop_index_to_string(prop, 2540 intval, &unused) != 0) 2541 err = SET_ERROR(EINVAL); 2542 break; 2543 default: 2544 cmn_err(CE_PANIC, 2545 "unknown property type"); 2546 } 2547 } else { 2548 err = SET_ERROR(EINVAL); 2549 } 2550 } 2551 2552 /* Validate permissions */ 2553 if (err == 0) 2554 err = zfs_check_settable(dsname, pair, CRED()); 2555 2556 if (err == 0) { 2557 err = zfs_prop_set_special(dsname, source, pair); 2558 if (err == -1) { 2559 /* 2560 * For better performance we build up a list of 2561 * properties to set in a single transaction. 2562 */ 2563 err = nvlist_add_nvpair(genericnvl, pair); 2564 } else if (err != 0 && nvl != retrynvl) { 2565 /* 2566 * This may be a spurious error caused by 2567 * receiving quota and reservation out of order. 2568 * Try again in a second pass. 2569 */ 2570 err = nvlist_add_nvpair(retrynvl, pair); 2571 } 2572 } 2573 2574 if (err != 0) { 2575 if (errlist != NULL) 2576 fnvlist_add_int32(errlist, propname, err); 2577 rv = err; 2578 } 2579 } 2580 2581 if (nvl != retrynvl && !nvlist_empty(retrynvl)) { 2582 nvl = retrynvl; 2583 goto retry; 2584 } 2585 2586 if (!nvlist_empty(genericnvl) && 2587 dsl_props_set(dsname, source, genericnvl) != 0) { 2588 /* 2589 * If this fails, we still want to set as many properties as we 2590 * can, so try setting them individually. 2591 */ 2592 pair = NULL; 2593 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) { 2594 const char *propname = nvpair_name(pair); 2595 int err = 0; 2596 2597 propval = pair; 2598 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2599 nvlist_t *attrs; 2600 attrs = fnvpair_value_nvlist(pair); 2601 propval = fnvlist_lookup_nvpair(attrs, 2602 ZPROP_VALUE); 2603 } 2604 2605 if (nvpair_type(propval) == DATA_TYPE_STRING) { 2606 strval = fnvpair_value_string(propval); 2607 err = dsl_prop_set_string(dsname, propname, 2608 source, strval); 2609 } else { 2610 intval = fnvpair_value_uint64(propval); 2611 err = dsl_prop_set_int(dsname, propname, source, 2612 intval); 2613 } 2614 2615 if (err != 0) { 2616 if (errlist != NULL) { 2617 fnvlist_add_int32(errlist, propname, 2618 err); 2619 } 2620 rv = err; 2621 } 2622 } 2623 } 2624 nvlist_free(genericnvl); 2625 nvlist_free(retrynvl); 2626 2627 return (rv); 2628} 2629 2630/* 2631 * Check that all the properties are valid user properties. 2632 */ 2633static int 2634zfs_check_userprops(const char *fsname, nvlist_t *nvl) 2635{ 2636 nvpair_t *pair = NULL; 2637 int error = 0; 2638 2639 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { 2640 const char *propname = nvpair_name(pair); 2641 char *valstr; 2642 2643 if (!zfs_prop_user(propname) || 2644 nvpair_type(pair) != DATA_TYPE_STRING) 2645 return (SET_ERROR(EINVAL)); 2646 2647 if (error = zfs_secpolicy_write_perms(fsname, 2648 ZFS_DELEG_PERM_USERPROP, CRED())) 2649 return (error); 2650 2651 if (strlen(propname) >= ZAP_MAXNAMELEN) 2652 return (SET_ERROR(ENAMETOOLONG)); 2653 2654 VERIFY(nvpair_value_string(pair, &valstr) == 0); 2655 if (strlen(valstr) >= ZAP_MAXVALUELEN) 2656 return (E2BIG); 2657 } 2658 return (0); 2659} 2660 2661static void 2662props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops) 2663{ 2664 nvpair_t *pair; 2665 2666 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2667 2668 pair = NULL; 2669 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) { 2670 if (nvlist_exists(skipped, nvpair_name(pair))) 2671 continue; 2672 2673 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0); 2674 } 2675} 2676 2677static int 2678clear_received_props(const char *dsname, nvlist_t *props, 2679 nvlist_t *skipped) 2680{ 2681 int err = 0; 2682 nvlist_t *cleared_props = NULL; 2683 props_skip(props, skipped, &cleared_props); 2684 if (!nvlist_empty(cleared_props)) { 2685 /* 2686 * Acts on local properties until the dataset has received 2687 * properties at least once on or after SPA_VERSION_RECVD_PROPS. 2688 */ 2689 zprop_source_t flags = (ZPROP_SRC_NONE | 2690 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0)); 2691 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL); 2692 } 2693 nvlist_free(cleared_props); 2694 return (err); 2695} 2696 2697/* 2698 * inputs: 2699 * zc_name name of filesystem 2700 * zc_value name of property to set 2701 * zc_nvlist_src{_size} nvlist of properties to apply 2702 * zc_cookie received properties flag 2703 * 2704 * outputs: 2705 * zc_nvlist_dst{_size} error for each unapplied received property 2706 */ 2707static int 2708zfs_ioc_set_prop(zfs_cmd_t *zc) 2709{ 2710 nvlist_t *nvl; 2711 boolean_t received = zc->zc_cookie; 2712 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED : 2713 ZPROP_SRC_LOCAL); 2714 nvlist_t *errors; 2715 int error; 2716 2717 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2718 zc->zc_iflags, &nvl)) != 0) 2719 return (error); 2720 2721 if (received) { 2722 nvlist_t *origprops; 2723 2724 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) { 2725 (void) clear_received_props(zc->zc_name, 2726 origprops, nvl); 2727 nvlist_free(origprops); 2728 } 2729 2730 error = dsl_prop_set_hasrecvd(zc->zc_name); 2731 } 2732 2733 errors = fnvlist_alloc(); 2734 if (error == 0) 2735 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors); 2736 2737 if (zc->zc_nvlist_dst != 0 && errors != NULL) { 2738 (void) put_nvlist(zc, errors); 2739 } 2740 2741 nvlist_free(errors); 2742 nvlist_free(nvl); 2743 return (error); 2744} 2745 2746/* 2747 * inputs: 2748 * zc_name name of filesystem 2749 * zc_value name of property to inherit 2750 * zc_cookie revert to received value if TRUE 2751 * 2752 * outputs: none 2753 */ 2754static int 2755zfs_ioc_inherit_prop(zfs_cmd_t *zc) 2756{ 2757 const char *propname = zc->zc_value; 2758 zfs_prop_t prop = zfs_name_to_prop(propname); 2759 boolean_t received = zc->zc_cookie; 2760 zprop_source_t source = (received 2761 ? ZPROP_SRC_NONE /* revert to received value, if any */ 2762 : ZPROP_SRC_INHERITED); /* explicitly inherit */ 2763 2764 if (received) { 2765 nvlist_t *dummy; 2766 nvpair_t *pair; 2767 zprop_type_t type; 2768 int err; 2769 2770 /* 2771 * zfs_prop_set_special() expects properties in the form of an 2772 * nvpair with type info. 2773 */ 2774 if (prop == ZPROP_INVAL) { 2775 if (!zfs_prop_user(propname)) 2776 return (SET_ERROR(EINVAL)); 2777 2778 type = PROP_TYPE_STRING; 2779 } else if (prop == ZFS_PROP_VOLSIZE || 2780 prop == ZFS_PROP_VERSION) { 2781 return (SET_ERROR(EINVAL)); 2782 } else { 2783 type = zfs_prop_get_type(prop); 2784 } 2785 2786 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2787 2788 switch (type) { 2789 case PROP_TYPE_STRING: 2790 VERIFY(0 == nvlist_add_string(dummy, propname, "")); 2791 break; 2792 case PROP_TYPE_NUMBER: 2793 case PROP_TYPE_INDEX: 2794 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0)); 2795 break; 2796 default: 2797 nvlist_free(dummy); 2798 return (SET_ERROR(EINVAL)); 2799 } 2800 2801 pair = nvlist_next_nvpair(dummy, NULL); 2802 err = zfs_prop_set_special(zc->zc_name, source, pair); 2803 nvlist_free(dummy); 2804 if (err != -1) 2805 return (err); /* special property already handled */ 2806 } else { 2807 /* 2808 * Only check this in the non-received case. We want to allow 2809 * 'inherit -S' to revert non-inheritable properties like quota 2810 * and reservation to the received or default values even though 2811 * they are not considered inheritable. 2812 */ 2813 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop)) 2814 return (SET_ERROR(EINVAL)); 2815 } 2816 2817 /* property name has been validated by zfs_secpolicy_inherit_prop() */ 2818 return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source)); 2819} 2820 2821static int 2822zfs_ioc_pool_set_props(zfs_cmd_t *zc) 2823{ 2824 nvlist_t *props; 2825 spa_t *spa; 2826 int error; 2827 nvpair_t *pair; 2828 2829 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2830 zc->zc_iflags, &props)) 2831 return (error); 2832 2833 /* 2834 * If the only property is the configfile, then just do a spa_lookup() 2835 * to handle the faulted case. 2836 */ 2837 pair = nvlist_next_nvpair(props, NULL); 2838 if (pair != NULL && strcmp(nvpair_name(pair), 2839 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 && 2840 nvlist_next_nvpair(props, pair) == NULL) { 2841 mutex_enter(&spa_namespace_lock); 2842 if ((spa = spa_lookup(zc->zc_name)) != NULL) { 2843 spa_configfile_set(spa, props, B_FALSE); 2844 spa_config_sync(spa, B_FALSE, B_TRUE); 2845 } 2846 mutex_exit(&spa_namespace_lock); 2847 if (spa != NULL) { 2848 nvlist_free(props); 2849 return (0); 2850 } 2851 } 2852 2853 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { 2854 nvlist_free(props); 2855 return (error); 2856 } 2857 2858 error = spa_prop_set(spa, props); 2859 2860 nvlist_free(props); 2861 spa_close(spa, FTAG); 2862 2863 return (error); 2864} 2865 2866static int 2867zfs_ioc_pool_get_props(zfs_cmd_t *zc) 2868{ 2869 spa_t *spa; 2870 int error; 2871 nvlist_t *nvp = NULL; 2872 2873 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { 2874 /* 2875 * If the pool is faulted, there may be properties we can still 2876 * get (such as altroot and cachefile), so attempt to get them 2877 * anyway. 2878 */ 2879 mutex_enter(&spa_namespace_lock); 2880 if ((spa = spa_lookup(zc->zc_name)) != NULL) 2881 error = spa_prop_get(spa, &nvp); 2882 mutex_exit(&spa_namespace_lock); 2883 } else { 2884 error = spa_prop_get(spa, &nvp); 2885 spa_close(spa, FTAG); 2886 } 2887 2888 if (error == 0 && zc->zc_nvlist_dst != 0) 2889 error = put_nvlist(zc, nvp); 2890 else 2891 error = SET_ERROR(EFAULT); 2892 2893 nvlist_free(nvp); 2894 return (error); 2895} 2896 2897/* 2898 * inputs: 2899 * zc_name name of filesystem 2900 * zc_nvlist_src{_size} nvlist of delegated permissions 2901 * zc_perm_action allow/unallow flag 2902 * 2903 * outputs: none 2904 */ 2905static int 2906zfs_ioc_set_fsacl(zfs_cmd_t *zc) 2907{ 2908 int error; 2909 nvlist_t *fsaclnv = NULL; 2910 2911 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2912 zc->zc_iflags, &fsaclnv)) != 0) 2913 return (error); 2914 2915 /* 2916 * Verify nvlist is constructed correctly 2917 */ 2918 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) { 2919 nvlist_free(fsaclnv); 2920 return (SET_ERROR(EINVAL)); 2921 } 2922 2923 /* 2924 * If we don't have PRIV_SYS_MOUNT, then validate 2925 * that user is allowed to hand out each permission in 2926 * the nvlist(s) 2927 */ 2928 2929 error = secpolicy_zfs(CRED()); 2930 if (error != 0) { 2931 if (zc->zc_perm_action == B_FALSE) { 2932 error = dsl_deleg_can_allow(zc->zc_name, 2933 fsaclnv, CRED()); 2934 } else { 2935 error = dsl_deleg_can_unallow(zc->zc_name, 2936 fsaclnv, CRED()); 2937 } 2938 } 2939 2940 if (error == 0) 2941 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action); 2942 2943 nvlist_free(fsaclnv); 2944 return (error); 2945} 2946 2947/* 2948 * inputs: 2949 * zc_name name of filesystem 2950 * 2951 * outputs: 2952 * zc_nvlist_src{_size} nvlist of delegated permissions 2953 */ 2954static int 2955zfs_ioc_get_fsacl(zfs_cmd_t *zc) 2956{ 2957 nvlist_t *nvp; 2958 int error; 2959 2960 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) { 2961 error = put_nvlist(zc, nvp); 2962 nvlist_free(nvp); 2963 } 2964 2965 return (error); 2966} 2967 2968/* 2969 * Search the vfs list for a specified resource. Returns a pointer to it 2970 * or NULL if no suitable entry is found. The caller of this routine 2971 * is responsible for releasing the returned vfs pointer. 2972 */ 2973static vfs_t * 2974zfs_get_vfs(const char *resource) 2975{ 2976 vfs_t *vfsp; 2977 2978 mtx_lock(&mountlist_mtx); 2979 TAILQ_FOREACH(vfsp, &mountlist, mnt_list) { 2980 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) { 2981 VFS_HOLD(vfsp); 2982 break; 2983 } 2984 } 2985 mtx_unlock(&mountlist_mtx); 2986 return (vfsp); 2987} 2988 2989/* ARGSUSED */ 2990static void 2991zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) 2992{ 2993 zfs_creat_t *zct = arg; 2994 2995 zfs_create_fs(os, cr, zct->zct_zplprops, tx); 2996} 2997 2998#define ZFS_PROP_UNDEFINED ((uint64_t)-1) 2999 3000/* 3001 * inputs: 3002 * os parent objset pointer (NULL if root fs) 3003 * fuids_ok fuids allowed in this version of the spa? 3004 * sa_ok SAs allowed in this version of the spa? 3005 * createprops list of properties requested by creator 3006 * 3007 * outputs: 3008 * zplprops values for the zplprops we attach to the master node object 3009 * is_ci true if requested file system will be purely case-insensitive 3010 * 3011 * Determine the settings for utf8only, normalization and 3012 * casesensitivity. Specific values may have been requested by the 3013 * creator and/or we can inherit values from the parent dataset. If 3014 * the file system is of too early a vintage, a creator can not 3015 * request settings for these properties, even if the requested 3016 * setting is the default value. We don't actually want to create dsl 3017 * properties for these, so remove them from the source nvlist after 3018 * processing. 3019 */ 3020static int 3021zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver, 3022 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops, 3023 nvlist_t *zplprops, boolean_t *is_ci) 3024{ 3025 uint64_t sense = ZFS_PROP_UNDEFINED; 3026 uint64_t norm = ZFS_PROP_UNDEFINED; 3027 uint64_t u8 = ZFS_PROP_UNDEFINED; 3028 3029 ASSERT(zplprops != NULL); 3030 3031 /* 3032 * Pull out creator prop choices, if any. 3033 */ 3034 if (createprops) { 3035 (void) nvlist_lookup_uint64(createprops, 3036 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver); 3037 (void) nvlist_lookup_uint64(createprops, 3038 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm); 3039 (void) nvlist_remove_all(createprops, 3040 zfs_prop_to_name(ZFS_PROP_NORMALIZE)); 3041 (void) nvlist_lookup_uint64(createprops, 3042 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8); 3043 (void) nvlist_remove_all(createprops, 3044 zfs_prop_to_name(ZFS_PROP_UTF8ONLY)); 3045 (void) nvlist_lookup_uint64(createprops, 3046 zfs_prop_to_name(ZFS_PROP_CASE), &sense); 3047 (void) nvlist_remove_all(createprops, 3048 zfs_prop_to_name(ZFS_PROP_CASE)); 3049 } 3050 3051 /* 3052 * If the zpl version requested is whacky or the file system 3053 * or pool is version is too "young" to support normalization 3054 * and the creator tried to set a value for one of the props, 3055 * error out. 3056 */ 3057 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) || 3058 (zplver >= ZPL_VERSION_FUID && !fuids_ok) || 3059 (zplver >= ZPL_VERSION_SA && !sa_ok) || 3060 (zplver < ZPL_VERSION_NORMALIZATION && 3061 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED || 3062 sense != ZFS_PROP_UNDEFINED))) 3063 return (SET_ERROR(ENOTSUP)); 3064 3065 /* 3066 * Put the version in the zplprops 3067 */ 3068 VERIFY(nvlist_add_uint64(zplprops, 3069 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0); 3070 3071 if (norm == ZFS_PROP_UNDEFINED) 3072 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0); 3073 VERIFY(nvlist_add_uint64(zplprops, 3074 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0); 3075 3076 /* 3077 * If we're normalizing, names must always be valid UTF-8 strings. 3078 */ 3079 if (norm) 3080 u8 = 1; 3081 if (u8 == ZFS_PROP_UNDEFINED) 3082 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0); 3083 VERIFY(nvlist_add_uint64(zplprops, 3084 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0); 3085 3086 if (sense == ZFS_PROP_UNDEFINED) 3087 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0); 3088 VERIFY(nvlist_add_uint64(zplprops, 3089 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0); 3090 3091 if (is_ci) 3092 *is_ci = (sense == ZFS_CASE_INSENSITIVE); 3093 3094 return (0); 3095} 3096 3097static int 3098zfs_fill_zplprops(const char *dataset, nvlist_t *createprops, 3099 nvlist_t *zplprops, boolean_t *is_ci) 3100{ 3101 boolean_t fuids_ok, sa_ok; 3102 uint64_t zplver = ZPL_VERSION; 3103 objset_t *os = NULL; 3104 char parentname[MAXNAMELEN]; 3105 char *cp; 3106 spa_t *spa; 3107 uint64_t spa_vers; 3108 int error; 3109 3110 (void) strlcpy(parentname, dataset, sizeof (parentname)); 3111 cp = strrchr(parentname, '/'); 3112 ASSERT(cp != NULL); 3113 cp[0] = '\0'; 3114 3115 if ((error = spa_open(dataset, &spa, FTAG)) != 0) 3116 return (error); 3117 3118 spa_vers = spa_version(spa); 3119 spa_close(spa, FTAG); 3120 3121 zplver = zfs_zpl_version_map(spa_vers); 3122 fuids_ok = (zplver >= ZPL_VERSION_FUID); 3123 sa_ok = (zplver >= ZPL_VERSION_SA); 3124 3125 /* 3126 * Open parent object set so we can inherit zplprop values. 3127 */ 3128 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0) 3129 return (error); 3130 3131 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops, 3132 zplprops, is_ci); 3133 dmu_objset_rele(os, FTAG); 3134 return (error); 3135} 3136 3137static int 3138zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops, 3139 nvlist_t *zplprops, boolean_t *is_ci) 3140{ 3141 boolean_t fuids_ok; 3142 boolean_t sa_ok; 3143 uint64_t zplver = ZPL_VERSION; 3144 int error; 3145 3146 zplver = zfs_zpl_version_map(spa_vers); 3147 fuids_ok = (zplver >= ZPL_VERSION_FUID); 3148 sa_ok = (zplver >= ZPL_VERSION_SA); 3149 3150 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok, 3151 createprops, zplprops, is_ci); 3152 return (error); 3153} 3154 3155/* 3156 * innvl: { 3157 * "type" -> dmu_objset_type_t (int32) 3158 * (optional) "props" -> { prop -> value } 3159 * } 3160 * 3161 * outnvl: propname -> error code (int32) 3162 */ 3163static int 3164zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3165{ 3166 int error = 0; 3167 zfs_creat_t zct = { 0 }; 3168 nvlist_t *nvprops = NULL; 3169 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx); 3170 int32_t type32; 3171 dmu_objset_type_t type; 3172 boolean_t is_insensitive = B_FALSE; 3173 3174 if (nvlist_lookup_int32(innvl, "type", &type32) != 0) 3175 return (SET_ERROR(EINVAL)); 3176 type = type32; 3177 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops); 3178 3179 switch (type) { 3180 case DMU_OST_ZFS: 3181 cbfunc = zfs_create_cb; 3182 break; 3183 3184 case DMU_OST_ZVOL: 3185 cbfunc = zvol_create_cb; 3186 break; 3187 3188 default: 3189 cbfunc = NULL; 3190 break; 3191 } 3192 if (strchr(fsname, '@') || 3193 strchr(fsname, '%')) 3194 return (SET_ERROR(EINVAL)); 3195 3196 zct.zct_props = nvprops; 3197 3198 if (cbfunc == NULL) 3199 return (SET_ERROR(EINVAL)); 3200 3201 if (type == DMU_OST_ZVOL) { 3202 uint64_t volsize, volblocksize; 3203 3204 if (nvprops == NULL) 3205 return (SET_ERROR(EINVAL)); 3206 if (nvlist_lookup_uint64(nvprops, 3207 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0) 3208 return (SET_ERROR(EINVAL)); 3209 3210 if ((error = nvlist_lookup_uint64(nvprops, 3211 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), 3212 &volblocksize)) != 0 && error != ENOENT) 3213 return (SET_ERROR(EINVAL)); 3214 3215 if (error != 0) 3216 volblocksize = zfs_prop_default_numeric( 3217 ZFS_PROP_VOLBLOCKSIZE); 3218 3219 if ((error = zvol_check_volblocksize( 3220 volblocksize)) != 0 || 3221 (error = zvol_check_volsize(volsize, 3222 volblocksize)) != 0) 3223 return (error); 3224 } else if (type == DMU_OST_ZFS) { 3225 int error; 3226 3227 /* 3228 * We have to have normalization and 3229 * case-folding flags correct when we do the 3230 * file system creation, so go figure them out 3231 * now. 3232 */ 3233 VERIFY(nvlist_alloc(&zct.zct_zplprops, 3234 NV_UNIQUE_NAME, KM_SLEEP) == 0); 3235 error = zfs_fill_zplprops(fsname, nvprops, 3236 zct.zct_zplprops, &is_insensitive); 3237 if (error != 0) { 3238 nvlist_free(zct.zct_zplprops); 3239 return (error); 3240 } 3241 } 3242 3243 error = dmu_objset_create(fsname, type, 3244 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct); 3245 nvlist_free(zct.zct_zplprops); 3246 3247 /* 3248 * It would be nice to do this atomically. 3249 */ 3250 if (error == 0) { 3251 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL, 3252 nvprops, outnvl); 3253 if (error != 0) 3254 (void) dsl_destroy_head(fsname); 3255 } 3256#ifdef __FreeBSD__ 3257 if (error == 0 && type == DMU_OST_ZVOL) 3258 zvol_create_minors(fsname); 3259#endif 3260 return (error); 3261} 3262 3263/* 3264 * innvl: { 3265 * "origin" -> name of origin snapshot 3266 * (optional) "props" -> { prop -> value } 3267 * } 3268 * 3269 * outnvl: propname -> error code (int32) 3270 */ 3271static int 3272zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3273{ 3274 int error = 0; 3275 nvlist_t *nvprops = NULL; 3276 char *origin_name; 3277 3278 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0) 3279 return (SET_ERROR(EINVAL)); 3280 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops); 3281 3282 if (strchr(fsname, '@') || 3283 strchr(fsname, '%')) 3284 return (SET_ERROR(EINVAL)); 3285 3286 if (dataset_namecheck(origin_name, NULL, NULL) != 0) 3287 return (SET_ERROR(EINVAL)); 3288 error = dmu_objset_clone(fsname, origin_name); 3289 if (error != 0) 3290 return (error); 3291 3292 /* 3293 * It would be nice to do this atomically. 3294 */ 3295 if (error == 0) { 3296 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL, 3297 nvprops, outnvl); 3298 if (error != 0) 3299 (void) dsl_destroy_head(fsname); 3300 } 3301 return (error); 3302} 3303 3304/* 3305 * innvl: { 3306 * "snaps" -> { snapshot1, snapshot2 } 3307 * (optional) "props" -> { prop -> value (string) } 3308 * } 3309 * 3310 * outnvl: snapshot -> error code (int32) 3311 */ 3312static int 3313zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3314{ 3315 nvlist_t *snaps; 3316 nvlist_t *props = NULL; 3317 int error, poollen; 3318 nvpair_t *pair; 3319 3320 (void) nvlist_lookup_nvlist(innvl, "props", &props); 3321 if ((error = zfs_check_userprops(poolname, props)) != 0) 3322 return (error); 3323 3324 if (!nvlist_empty(props) && 3325 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS)) 3326 return (SET_ERROR(ENOTSUP)); 3327 3328 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 3329 return (SET_ERROR(EINVAL)); 3330 poollen = strlen(poolname); 3331 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 3332 pair = nvlist_next_nvpair(snaps, pair)) { 3333 const char *name = nvpair_name(pair); 3334 const char *cp = strchr(name, '@'); 3335 3336 /* 3337 * The snap name must contain an @, and the part after it must 3338 * contain only valid characters. 3339 */ 3340 if (cp == NULL || snapshot_namecheck(cp + 1, NULL, NULL) != 0) 3341 return (SET_ERROR(EINVAL)); 3342 3343 /* 3344 * The snap must be in the specified pool. 3345 */ 3346 if (strncmp(name, poolname, poollen) != 0 || 3347 (name[poollen] != '/' && name[poollen] != '@')) 3348 return (SET_ERROR(EXDEV)); 3349 3350 /* This must be the only snap of this fs. */ 3351 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair); 3352 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) { 3353 if (strncmp(name, nvpair_name(pair2), cp - name + 1) 3354 == 0) { 3355 return (SET_ERROR(EXDEV)); 3356 } 3357 } 3358 } 3359 3360 error = dsl_dataset_snapshot(snaps, props, outnvl); 3361 return (error); 3362} 3363 3364/* 3365 * innvl: "message" -> string 3366 */ 3367/* ARGSUSED */ 3368static int 3369zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl) 3370{ 3371 char *message; 3372 spa_t *spa; 3373 int error; 3374 char *poolname; 3375 3376 /* 3377 * The poolname in the ioctl is not set, we get it from the TSD, 3378 * which was set at the end of the last successful ioctl that allows 3379 * logging. The secpolicy func already checked that it is set. 3380 * Only one log ioctl is allowed after each successful ioctl, so 3381 * we clear the TSD here. 3382 */ 3383 poolname = tsd_get(zfs_allow_log_key); 3384 (void) tsd_set(zfs_allow_log_key, NULL); 3385 error = spa_open(poolname, &spa, FTAG); 3386 strfree(poolname); 3387 if (error != 0) 3388 return (error); 3389 3390 if (nvlist_lookup_string(innvl, "message", &message) != 0) { 3391 spa_close(spa, FTAG); 3392 return (SET_ERROR(EINVAL)); 3393 } 3394 3395 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) { 3396 spa_close(spa, FTAG); 3397 return (SET_ERROR(ENOTSUP)); 3398 } 3399 3400 error = spa_history_log(spa, message); 3401 spa_close(spa, FTAG); 3402 return (error); 3403} 3404 3405/* 3406 * The dp_config_rwlock must not be held when calling this, because the 3407 * unmount may need to write out data. 3408 * 3409 * This function is best-effort. Callers must deal gracefully if it 3410 * remains mounted (or is remounted after this call). 3411 * 3412 * Returns 0 if the argument is not a snapshot, or it is not currently a 3413 * filesystem, or we were able to unmount it. Returns error code otherwise. 3414 */ 3415int 3416zfs_unmount_snap(const char *snapname) 3417{ 3418 vfs_t *vfsp; 3419 zfsvfs_t *zfsvfs; 3420 int err; 3421 3422 if (strchr(snapname, '@') == NULL) 3423 return (0); 3424 3425 vfsp = zfs_get_vfs(snapname); 3426 if (vfsp == NULL) 3427 return (0); 3428 3429 zfsvfs = vfsp->vfs_data; 3430 ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os))); 3431 3432 err = vn_vfswlock(vfsp->vfs_vnodecovered); 3433 VFS_RELE(vfsp); 3434 if (err != 0) 3435 return (SET_ERROR(err)); 3436 3437 /* 3438 * Always force the unmount for snapshots. 3439 */ 3440 3441#ifdef illumos 3442 (void) dounmount(vfsp, MS_FORCE, kcred); 3443#else 3444 mtx_lock(&Giant); /* dounmount() */ 3445 (void) dounmount(vfsp, MS_FORCE, curthread); 3446 mtx_unlock(&Giant); /* dounmount() */ 3447#endif 3448 return (0); 3449} 3450 3451/* ARGSUSED */ 3452static int 3453zfs_unmount_snap_cb(const char *snapname, void *arg) 3454{ 3455 return (zfs_unmount_snap(snapname)); 3456} 3457 3458/* 3459 * When a clone is destroyed, its origin may also need to be destroyed, 3460 * in which case it must be unmounted. This routine will do that unmount 3461 * if necessary. 3462 */ 3463void 3464zfs_destroy_unmount_origin(const char *fsname) 3465{ 3466 int error; 3467 objset_t *os; 3468 dsl_dataset_t *ds; 3469 3470 error = dmu_objset_hold(fsname, FTAG, &os); 3471 if (error != 0) 3472 return; 3473 ds = dmu_objset_ds(os); 3474 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) { 3475 char originname[MAXNAMELEN]; 3476 dsl_dataset_name(ds->ds_prev, originname); 3477 dmu_objset_rele(os, FTAG); 3478 (void) zfs_unmount_snap(originname); 3479 } else { 3480 dmu_objset_rele(os, FTAG); 3481 } 3482} 3483 3484/* 3485 * innvl: { 3486 * "snaps" -> { snapshot1, snapshot2 } 3487 * (optional boolean) "defer" 3488 * } 3489 * 3490 * outnvl: snapshot -> error code (int32) 3491 * 3492 */ 3493static int 3494zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3495{ 3496 int error, poollen; 3497 nvlist_t *snaps; 3498 nvpair_t *pair; 3499 boolean_t defer; 3500 3501 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 3502 return (SET_ERROR(EINVAL)); 3503 defer = nvlist_exists(innvl, "defer"); 3504 3505 poollen = strlen(poolname); 3506 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 3507 pair = nvlist_next_nvpair(snaps, pair)) { 3508 const char *name = nvpair_name(pair); 3509 3510 /* 3511 * The snap must be in the specified pool. 3512 */ 3513 if (strncmp(name, poolname, poollen) != 0 || 3514 (name[poollen] != '/' && name[poollen] != '@')) 3515 return (SET_ERROR(EXDEV)); 3516 3517 error = zfs_unmount_snap(name); 3518 if (error != 0) 3519 return (error); 3520 (void) zvol_remove_minor(name); 3521 } 3522 3523 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl)); 3524} 3525 3526/* 3527 * inputs: 3528 * zc_name name of dataset to destroy 3529 * zc_objset_type type of objset 3530 * zc_defer_destroy mark for deferred destroy 3531 * 3532 * outputs: none 3533 */ 3534static int 3535zfs_ioc_destroy(zfs_cmd_t *zc) 3536{ 3537 int err; 3538 3539 if (zc->zc_objset_type == DMU_OST_ZFS) { 3540 err = zfs_unmount_snap(zc->zc_name); 3541 if (err != 0) 3542 return (err); 3543 } 3544 3545 if (strchr(zc->zc_name, '@')) 3546 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy); 3547 else 3548 err = dsl_destroy_head(zc->zc_name); 3549 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0) 3550 (void) zvol_remove_minor(zc->zc_name); 3551 return (err); 3552} 3553 3554/* 3555 * fsname is name of dataset to rollback (to most recent snapshot) 3556 * 3557 * innvl is not used. 3558 * 3559 * outnvl: "target" -> name of most recent snapshot 3560 * } 3561 */ 3562/* ARGSUSED */ 3563static int 3564zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl) 3565{ 3566 zfsvfs_t *zfsvfs; 3567 int error; 3568 3569 if (getzfsvfs(fsname, &zfsvfs) == 0) { 3570 error = zfs_suspend_fs(zfsvfs); 3571 if (error == 0) { 3572 int resume_err; 3573 3574 error = dsl_dataset_rollback(fsname, zfsvfs, outnvl); 3575 resume_err = zfs_resume_fs(zfsvfs, fsname); 3576 error = error ? error : resume_err; 3577 } 3578 VFS_RELE(zfsvfs->z_vfs); 3579 } else { 3580 error = dsl_dataset_rollback(fsname, NULL, outnvl); 3581 } 3582 return (error); 3583} 3584 3585static int 3586recursive_unmount(const char *fsname, void *arg) 3587{ 3588 const char *snapname = arg; 3589 char fullname[MAXNAMELEN]; 3590 3591 (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname); 3592 return (zfs_unmount_snap(fullname)); 3593} 3594 3595/* 3596 * inputs: 3597 * zc_name old name of dataset 3598 * zc_value new name of dataset 3599 * zc_cookie recursive flag (only valid for snapshots) 3600 * 3601 * outputs: none 3602 */ 3603static int 3604zfs_ioc_rename(zfs_cmd_t *zc) 3605{ 3606 boolean_t recursive = zc->zc_cookie & 1; 3607#ifdef __FreeBSD__ 3608 boolean_t allow_mounted = zc->zc_cookie & 2; 3609#endif 3610 char *at; 3611 3612 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0'; 3613 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 3614 strchr(zc->zc_value, '%')) 3615 return (SET_ERROR(EINVAL)); 3616 3617 at = strchr(zc->zc_name, '@'); 3618 if (at != NULL) { 3619 /* snaps must be in same fs */ 3620 int error; 3621 3622 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1)) 3623 return (SET_ERROR(EXDEV)); 3624 *at = '\0'; 3625#ifdef illumos 3626 if (zc->zc_objset_type == DMU_OST_ZFS) { 3627#else 3628 if (zc->zc_objset_type == DMU_OST_ZFS && allow_mounted) { 3629#endif 3630 error = dmu_objset_find(zc->zc_name, 3631 recursive_unmount, at + 1, 3632 recursive ? DS_FIND_CHILDREN : 0); 3633 if (error != 0) { 3634 *at = '@'; 3635 return (error); 3636 } 3637 } 3638 error = dsl_dataset_rename_snapshot(zc->zc_name, 3639 at + 1, strchr(zc->zc_value, '@') + 1, recursive); 3640 *at = '@'; 3641 3642 return (error); 3643 } else { 3644#ifdef illumos 3645 if (zc->zc_objset_type == DMU_OST_ZVOL) 3646 (void) zvol_remove_minor(zc->zc_name); 3647#endif 3648 return (dsl_dir_rename(zc->zc_name, zc->zc_value)); 3649 } 3650} 3651 3652static int 3653zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr) 3654{ 3655 const char *propname = nvpair_name(pair); 3656 boolean_t issnap = (strchr(dsname, '@') != NULL); 3657 zfs_prop_t prop = zfs_name_to_prop(propname); 3658 uint64_t intval; 3659 int err; 3660 3661 if (prop == ZPROP_INVAL) { 3662 if (zfs_prop_user(propname)) { 3663 if (err = zfs_secpolicy_write_perms(dsname, 3664 ZFS_DELEG_PERM_USERPROP, cr)) 3665 return (err); 3666 return (0); 3667 } 3668 3669 if (!issnap && zfs_prop_userquota(propname)) { 3670 const char *perm = NULL; 3671 const char *uq_prefix = 3672 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA]; 3673 const char *gq_prefix = 3674 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA]; 3675 3676 if (strncmp(propname, uq_prefix, 3677 strlen(uq_prefix)) == 0) { 3678 perm = ZFS_DELEG_PERM_USERQUOTA; 3679 } else if (strncmp(propname, gq_prefix, 3680 strlen(gq_prefix)) == 0) { 3681 perm = ZFS_DELEG_PERM_GROUPQUOTA; 3682 } else { 3683 /* USERUSED and GROUPUSED are read-only */ 3684 return (SET_ERROR(EINVAL)); 3685 } 3686 3687 if (err = zfs_secpolicy_write_perms(dsname, perm, cr)) 3688 return (err); 3689 return (0); 3690 } 3691 3692 return (SET_ERROR(EINVAL)); 3693 } 3694 3695 if (issnap) 3696 return (SET_ERROR(EINVAL)); 3697 3698 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 3699 /* 3700 * dsl_prop_get_all_impl() returns properties in this 3701 * format. 3702 */ 3703 nvlist_t *attrs; 3704 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 3705 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 3706 &pair) == 0); 3707 } 3708 3709 /* 3710 * Check that this value is valid for this pool version 3711 */ 3712 switch (prop) { 3713 case ZFS_PROP_COMPRESSION: 3714 /* 3715 * If the user specified gzip compression, make sure 3716 * the SPA supports it. We ignore any errors here since 3717 * we'll catch them later. 3718 */ 3719 if (nvpair_type(pair) == DATA_TYPE_UINT64 && 3720 nvpair_value_uint64(pair, &intval) == 0) { 3721 if (intval >= ZIO_COMPRESS_GZIP_1 && 3722 intval <= ZIO_COMPRESS_GZIP_9 && 3723 zfs_earlier_version(dsname, 3724 SPA_VERSION_GZIP_COMPRESSION)) { 3725 return (SET_ERROR(ENOTSUP)); 3726 } 3727 3728 if (intval == ZIO_COMPRESS_ZLE && 3729 zfs_earlier_version(dsname, 3730 SPA_VERSION_ZLE_COMPRESSION)) 3731 return (SET_ERROR(ENOTSUP)); 3732 3733 if (intval == ZIO_COMPRESS_LZ4) { 3734 zfeature_info_t *feature = 3735 &spa_feature_table[ 3736 SPA_FEATURE_LZ4_COMPRESS]; 3737 spa_t *spa; 3738 3739 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 3740 return (err); 3741 3742 if (!spa_feature_is_enabled(spa, feature)) { 3743 spa_close(spa, FTAG); 3744 return (SET_ERROR(ENOTSUP)); 3745 } 3746 spa_close(spa, FTAG); 3747 } 3748 3749 /* 3750 * If this is a bootable dataset then 3751 * verify that the compression algorithm 3752 * is supported for booting. We must return 3753 * something other than ENOTSUP since it 3754 * implies a downrev pool version. 3755 */ 3756 if (zfs_is_bootfs(dsname) && 3757 !BOOTFS_COMPRESS_VALID(intval)) { 3758 return (SET_ERROR(ERANGE)); 3759 } 3760 } 3761 break; 3762 3763 case ZFS_PROP_COPIES: 3764 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS)) 3765 return (SET_ERROR(ENOTSUP)); 3766 break; 3767 3768 case ZFS_PROP_DEDUP: 3769 if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP)) 3770 return (SET_ERROR(ENOTSUP)); 3771 break; 3772 3773 case ZFS_PROP_SHARESMB: 3774 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID)) 3775 return (SET_ERROR(ENOTSUP)); 3776 break; 3777 3778 case ZFS_PROP_ACLINHERIT: 3779 if (nvpair_type(pair) == DATA_TYPE_UINT64 && 3780 nvpair_value_uint64(pair, &intval) == 0) { 3781 if (intval == ZFS_ACL_PASSTHROUGH_X && 3782 zfs_earlier_version(dsname, 3783 SPA_VERSION_PASSTHROUGH_X)) 3784 return (SET_ERROR(ENOTSUP)); 3785 } 3786 break; 3787 } 3788 3789 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED())); 3790} 3791 3792/* 3793 * Checks for a race condition to make sure we don't increment a feature flag 3794 * multiple times. 3795 */ 3796static int 3797zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx) 3798{ 3799 spa_t *spa = dmu_tx_pool(tx)->dp_spa; 3800 zfeature_info_t *feature = arg; 3801 3802 if (!spa_feature_is_active(spa, feature)) 3803 return (0); 3804 else 3805 return (SET_ERROR(EBUSY)); 3806} 3807 3808/* 3809 * The callback invoked on feature activation in the sync task caused by 3810 * zfs_prop_activate_feature. 3811 */ 3812static void 3813zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx) 3814{ 3815 spa_t *spa = dmu_tx_pool(tx)->dp_spa; 3816 zfeature_info_t *feature = arg; 3817 3818 spa_feature_incr(spa, feature, tx); 3819} 3820 3821/* 3822 * Activates a feature on a pool in response to a property setting. This 3823 * creates a new sync task which modifies the pool to reflect the feature 3824 * as being active. 3825 */ 3826static int 3827zfs_prop_activate_feature(spa_t *spa, zfeature_info_t *feature) 3828{ 3829 int err; 3830 3831 /* EBUSY here indicates that the feature is already active */ 3832 err = dsl_sync_task(spa_name(spa), 3833 zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync, 3834 feature, 2); 3835 3836 if (err != 0 && err != EBUSY) 3837 return (err); 3838 else 3839 return (0); 3840} 3841 3842/* 3843 * Removes properties from the given props list that fail permission checks 3844 * needed to clear them and to restore them in case of a receive error. For each 3845 * property, make sure we have both set and inherit permissions. 3846 * 3847 * Returns the first error encountered if any permission checks fail. If the 3848 * caller provides a non-NULL errlist, it also gives the complete list of names 3849 * of all the properties that failed a permission check along with the 3850 * corresponding error numbers. The caller is responsible for freeing the 3851 * returned errlist. 3852 * 3853 * If every property checks out successfully, zero is returned and the list 3854 * pointed at by errlist is NULL. 3855 */ 3856static int 3857zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist) 3858{ 3859 zfs_cmd_t *zc; 3860 nvpair_t *pair, *next_pair; 3861 nvlist_t *errors; 3862 int err, rv = 0; 3863 3864 if (props == NULL) 3865 return (0); 3866 3867 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0); 3868 3869 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP); 3870 (void) strcpy(zc->zc_name, dataset); 3871 pair = nvlist_next_nvpair(props, NULL); 3872 while (pair != NULL) { 3873 next_pair = nvlist_next_nvpair(props, pair); 3874 3875 (void) strcpy(zc->zc_value, nvpair_name(pair)); 3876 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 || 3877 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) { 3878 VERIFY(nvlist_remove_nvpair(props, pair) == 0); 3879 VERIFY(nvlist_add_int32(errors, 3880 zc->zc_value, err) == 0); 3881 } 3882 pair = next_pair; 3883 } 3884 kmem_free(zc, sizeof (zfs_cmd_t)); 3885 3886 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) { 3887 nvlist_free(errors); 3888 errors = NULL; 3889 } else { 3890 VERIFY(nvpair_value_int32(pair, &rv) == 0); 3891 } 3892 3893 if (errlist == NULL) 3894 nvlist_free(errors); 3895 else 3896 *errlist = errors; 3897 3898 return (rv); 3899} 3900 3901static boolean_t 3902propval_equals(nvpair_t *p1, nvpair_t *p2) 3903{ 3904 if (nvpair_type(p1) == DATA_TYPE_NVLIST) { 3905 /* dsl_prop_get_all_impl() format */ 3906 nvlist_t *attrs; 3907 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0); 3908 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 3909 &p1) == 0); 3910 } 3911 3912 if (nvpair_type(p2) == DATA_TYPE_NVLIST) { 3913 nvlist_t *attrs; 3914 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0); 3915 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 3916 &p2) == 0); 3917 } 3918 3919 if (nvpair_type(p1) != nvpair_type(p2)) 3920 return (B_FALSE); 3921 3922 if (nvpair_type(p1) == DATA_TYPE_STRING) { 3923 char *valstr1, *valstr2; 3924 3925 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0); 3926 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0); 3927 return (strcmp(valstr1, valstr2) == 0); 3928 } else { 3929 uint64_t intval1, intval2; 3930 3931 VERIFY(nvpair_value_uint64(p1, &intval1) == 0); 3932 VERIFY(nvpair_value_uint64(p2, &intval2) == 0); 3933 return (intval1 == intval2); 3934 } 3935} 3936 3937/* 3938 * Remove properties from props if they are not going to change (as determined 3939 * by comparison with origprops). Remove them from origprops as well, since we 3940 * do not need to clear or restore properties that won't change. 3941 */ 3942static void 3943props_reduce(nvlist_t *props, nvlist_t *origprops) 3944{ 3945 nvpair_t *pair, *next_pair; 3946 3947 if (origprops == NULL) 3948 return; /* all props need to be received */ 3949 3950 pair = nvlist_next_nvpair(props, NULL); 3951 while (pair != NULL) { 3952 const char *propname = nvpair_name(pair); 3953 nvpair_t *match; 3954 3955 next_pair = nvlist_next_nvpair(props, pair); 3956 3957 if ((nvlist_lookup_nvpair(origprops, propname, 3958 &match) != 0) || !propval_equals(pair, match)) 3959 goto next; /* need to set received value */ 3960 3961 /* don't clear the existing received value */ 3962 (void) nvlist_remove_nvpair(origprops, match); 3963 /* don't bother receiving the property */ 3964 (void) nvlist_remove_nvpair(props, pair); 3965next: 3966 pair = next_pair; 3967 } 3968} 3969 3970#ifdef DEBUG 3971static boolean_t zfs_ioc_recv_inject_err; 3972#endif 3973 3974/* 3975 * inputs: 3976 * zc_name name of containing filesystem 3977 * zc_nvlist_src{_size} nvlist of properties to apply 3978 * zc_value name of snapshot to create 3979 * zc_string name of clone origin (if DRR_FLAG_CLONE) 3980 * zc_cookie file descriptor to recv from 3981 * zc_begin_record the BEGIN record of the stream (not byteswapped) 3982 * zc_guid force flag 3983 * zc_cleanup_fd cleanup-on-exit file descriptor 3984 * zc_action_handle handle for this guid/ds mapping (or zero on first call) 3985 * 3986 * outputs: 3987 * zc_cookie number of bytes read 3988 * zc_nvlist_dst{_size} error for each unapplied received property 3989 * zc_obj zprop_errflags_t 3990 * zc_action_handle handle for this guid/ds mapping 3991 */ 3992static int 3993zfs_ioc_recv(zfs_cmd_t *zc) 3994{ 3995 file_t *fp; 3996 dmu_recv_cookie_t drc; 3997 boolean_t force = (boolean_t)zc->zc_guid; 3998 int fd; 3999 int error = 0; 4000 int props_error = 0; 4001 nvlist_t *errors; 4002 offset_t off; 4003 nvlist_t *props = NULL; /* sent properties */ 4004 nvlist_t *origprops = NULL; /* existing properties */ 4005 char *origin = NULL; 4006 char *tosnap; 4007 char tofs[ZFS_MAXNAMELEN]; 4008 cap_rights_t rights; 4009 boolean_t first_recvd_props = B_FALSE; 4010 4011 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 4012 strchr(zc->zc_value, '@') == NULL || 4013 strchr(zc->zc_value, '%')) 4014 return (SET_ERROR(EINVAL)); 4015 4016 (void) strcpy(tofs, zc->zc_value); 4017 tosnap = strchr(tofs, '@'); 4018 *tosnap++ = '\0'; 4019 4020 if (zc->zc_nvlist_src != 0 && 4021 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 4022 zc->zc_iflags, &props)) != 0) 4023 return (error); 4024 4025 fd = zc->zc_cookie; 4026 fp = getf(fd, cap_rights_init(&rights, CAP_PREAD)); 4027 if (fp == NULL) { 4028 nvlist_free(props); 4029 return (SET_ERROR(EBADF)); 4030 } 4031 4032 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0); 4033 4034 if (zc->zc_string[0]) 4035 origin = zc->zc_string; 4036 4037 error = dmu_recv_begin(tofs, tosnap, 4038 &zc->zc_begin_record, force, origin, &drc); 4039 if (error != 0) 4040 goto out; 4041 4042 /* 4043 * Set properties before we receive the stream so that they are applied 4044 * to the new data. Note that we must call dmu_recv_stream() if 4045 * dmu_recv_begin() succeeds. 4046 */ 4047 if (props != NULL && !drc.drc_newfs) { 4048 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >= 4049 SPA_VERSION_RECVD_PROPS && 4050 !dsl_prop_get_hasrecvd(tofs)) 4051 first_recvd_props = B_TRUE; 4052 4053 /* 4054 * If new received properties are supplied, they are to 4055 * completely replace the existing received properties, so stash 4056 * away the existing ones. 4057 */ 4058 if (dsl_prop_get_received(tofs, &origprops) == 0) { 4059 nvlist_t *errlist = NULL; 4060 /* 4061 * Don't bother writing a property if its value won't 4062 * change (and avoid the unnecessary security checks). 4063 * 4064 * The first receive after SPA_VERSION_RECVD_PROPS is a 4065 * special case where we blow away all local properties 4066 * regardless. 4067 */ 4068 if (!first_recvd_props) 4069 props_reduce(props, origprops); 4070 if (zfs_check_clearable(tofs, origprops, &errlist) != 0) 4071 (void) nvlist_merge(errors, errlist, 0); 4072 nvlist_free(errlist); 4073 4074 if (clear_received_props(tofs, origprops, 4075 first_recvd_props ? NULL : props) != 0) 4076 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 4077 } else { 4078 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 4079 } 4080 } 4081 4082 if (props != NULL) { 4083 props_error = dsl_prop_set_hasrecvd(tofs); 4084 4085 if (props_error == 0) { 4086 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED, 4087 props, errors); 4088 } 4089 } 4090 4091 if (zc->zc_nvlist_dst_size != 0 && 4092 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 || 4093 put_nvlist(zc, errors) != 0)) { 4094 /* 4095 * Caller made zc->zc_nvlist_dst less than the minimum expected 4096 * size or supplied an invalid address. 4097 */ 4098 props_error = SET_ERROR(EINVAL); 4099 } 4100 4101 off = fp->f_offset; 4102 error = dmu_recv_stream(&drc, fp, &off, zc->zc_cleanup_fd, 4103 &zc->zc_action_handle); 4104 4105 if (error == 0) { 4106 zfsvfs_t *zfsvfs = NULL; 4107 4108 if (getzfsvfs(tofs, &zfsvfs) == 0) { 4109 /* online recv */ 4110 int end_err; 4111 4112 error = zfs_suspend_fs(zfsvfs); 4113 /* 4114 * If the suspend fails, then the recv_end will 4115 * likely also fail, and clean up after itself. 4116 */ 4117 end_err = dmu_recv_end(&drc, zfsvfs); 4118 if (error == 0) 4119 error = zfs_resume_fs(zfsvfs, tofs); 4120 error = error ? error : end_err; 4121 VFS_RELE(zfsvfs->z_vfs); 4122 } else { 4123 error = dmu_recv_end(&drc, NULL); 4124 } 4125 } 4126 4127 zc->zc_cookie = off - fp->f_offset; 4128 if (off >= 0 && off <= MAXOFFSET_T) 4129 fp->f_offset = off; 4130 4131#ifdef DEBUG 4132 if (zfs_ioc_recv_inject_err) { 4133 zfs_ioc_recv_inject_err = B_FALSE; 4134 error = 1; 4135 } 4136#endif 4137 4138#ifdef __FreeBSD__ 4139 if (error == 0) 4140 zvol_create_minors(tofs); 4141#endif 4142 4143 /* 4144 * On error, restore the original props. 4145 */ 4146 if (error != 0 && props != NULL && !drc.drc_newfs) { 4147 if (clear_received_props(tofs, props, NULL) != 0) { 4148 /* 4149 * We failed to clear the received properties. 4150 * Since we may have left a $recvd value on the 4151 * system, we can't clear the $hasrecvd flag. 4152 */ 4153 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4154 } else if (first_recvd_props) { 4155 dsl_prop_unset_hasrecvd(tofs); 4156 } 4157 4158 if (origprops == NULL && !drc.drc_newfs) { 4159 /* We failed to stash the original properties. */ 4160 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4161 } 4162 4163 /* 4164 * dsl_props_set() will not convert RECEIVED to LOCAL on or 4165 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL 4166 * explictly if we're restoring local properties cleared in the 4167 * first new-style receive. 4168 */ 4169 if (origprops != NULL && 4170 zfs_set_prop_nvlist(tofs, (first_recvd_props ? 4171 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED), 4172 origprops, NULL) != 0) { 4173 /* 4174 * We stashed the original properties but failed to 4175 * restore them. 4176 */ 4177 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4178 } 4179 } 4180out: 4181 nvlist_free(props); 4182 nvlist_free(origprops); 4183 nvlist_free(errors); 4184 releasef(fd); 4185 4186 if (error == 0) 4187 error = props_error; 4188 4189 return (error); 4190} 4191 4192/* 4193 * inputs: 4194 * zc_name name of snapshot to send 4195 * zc_cookie file descriptor to send stream to 4196 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj) 4197 * zc_sendobj objsetid of snapshot to send 4198 * zc_fromobj objsetid of incremental fromsnap (may be zero) 4199 * zc_guid if set, estimate size of stream only. zc_cookie is ignored. 4200 * output size in zc_objset_type. 4201 * 4202 * outputs: none 4203 */ 4204static int 4205zfs_ioc_send(zfs_cmd_t *zc) 4206{ 4207 int error; 4208 offset_t off; 4209 boolean_t estimate = (zc->zc_guid != 0); 4210 4211 if (zc->zc_obj != 0) { 4212 dsl_pool_t *dp; 4213 dsl_dataset_t *tosnap; 4214 4215 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4216 if (error != 0) 4217 return (error); 4218 4219 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap); 4220 if (error != 0) { 4221 dsl_pool_rele(dp, FTAG); 4222 return (error); 4223 } 4224 4225 if (dsl_dir_is_clone(tosnap->ds_dir)) 4226 zc->zc_fromobj = tosnap->ds_dir->dd_phys->dd_origin_obj; 4227 dsl_dataset_rele(tosnap, FTAG); 4228 dsl_pool_rele(dp, FTAG); 4229 } 4230 4231 if (estimate) { 4232 dsl_pool_t *dp; 4233 dsl_dataset_t *tosnap; 4234 dsl_dataset_t *fromsnap = NULL; 4235 4236 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4237 if (error != 0) 4238 return (error); 4239 4240 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap); 4241 if (error != 0) { 4242 dsl_pool_rele(dp, FTAG); 4243 return (error); 4244 } 4245 4246 if (zc->zc_fromobj != 0) { 4247 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj, 4248 FTAG, &fromsnap); 4249 if (error != 0) { 4250 dsl_dataset_rele(tosnap, FTAG); 4251 dsl_pool_rele(dp, FTAG); 4252 return (error); 4253 } 4254 } 4255 4256 error = dmu_send_estimate(tosnap, fromsnap, 4257 &zc->zc_objset_type); 4258 4259 if (fromsnap != NULL) 4260 dsl_dataset_rele(fromsnap, FTAG); 4261 dsl_dataset_rele(tosnap, FTAG); 4262 dsl_pool_rele(dp, FTAG); 4263 } else { 4264 file_t *fp; 4265 cap_rights_t rights; 4266 4267 fp = getf(zc->zc_cookie, 4268 cap_rights_init(&rights, CAP_WRITE)); 4269 if (fp == NULL) 4270 return (SET_ERROR(EBADF)); 4271 4272 off = fp->f_offset; 4273 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj, 4274#ifdef illumos 4275 zc->zc_fromobj, zc->zc_cookie, fp->f_vnode, &off); 4276#else 4277 zc->zc_fromobj, zc->zc_cookie, fp, &off); 4278#endif 4279 4280 if (off >= 0 && off <= MAXOFFSET_T) 4281 fp->f_offset = off; 4282 releasef(zc->zc_cookie); 4283 } 4284 return (error); 4285} 4286 4287/* 4288 * inputs: 4289 * zc_name name of snapshot on which to report progress 4290 * zc_cookie file descriptor of send stream 4291 * 4292 * outputs: 4293 * zc_cookie number of bytes written in send stream thus far 4294 */ 4295static int 4296zfs_ioc_send_progress(zfs_cmd_t *zc) 4297{ 4298 dsl_pool_t *dp; 4299 dsl_dataset_t *ds; 4300 dmu_sendarg_t *dsp = NULL; 4301 int error; 4302 4303 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4304 if (error != 0) 4305 return (error); 4306 4307 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds); 4308 if (error != 0) { 4309 dsl_pool_rele(dp, FTAG); 4310 return (error); 4311 } 4312 4313 mutex_enter(&ds->ds_sendstream_lock); 4314 4315 /* 4316 * Iterate over all the send streams currently active on this dataset. 4317 * If there's one which matches the specified file descriptor _and_ the 4318 * stream was started by the current process, return the progress of 4319 * that stream. 4320 */ 4321 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL; 4322 dsp = list_next(&ds->ds_sendstreams, dsp)) { 4323 if (dsp->dsa_outfd == zc->zc_cookie && 4324 dsp->dsa_proc == curproc) 4325 break; 4326 } 4327 4328 if (dsp != NULL) 4329 zc->zc_cookie = *(dsp->dsa_off); 4330 else 4331 error = SET_ERROR(ENOENT); 4332 4333 mutex_exit(&ds->ds_sendstream_lock); 4334 dsl_dataset_rele(ds, FTAG); 4335 dsl_pool_rele(dp, FTAG); 4336 return (error); 4337} 4338 4339static int 4340zfs_ioc_inject_fault(zfs_cmd_t *zc) 4341{ 4342 int id, error; 4343 4344 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id, 4345 &zc->zc_inject_record); 4346 4347 if (error == 0) 4348 zc->zc_guid = (uint64_t)id; 4349 4350 return (error); 4351} 4352 4353static int 4354zfs_ioc_clear_fault(zfs_cmd_t *zc) 4355{ 4356 return (zio_clear_fault((int)zc->zc_guid)); 4357} 4358 4359static int 4360zfs_ioc_inject_list_next(zfs_cmd_t *zc) 4361{ 4362 int id = (int)zc->zc_guid; 4363 int error; 4364 4365 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name), 4366 &zc->zc_inject_record); 4367 4368 zc->zc_guid = id; 4369 4370 return (error); 4371} 4372 4373static int 4374zfs_ioc_error_log(zfs_cmd_t *zc) 4375{ 4376 spa_t *spa; 4377 int error; 4378 size_t count = (size_t)zc->zc_nvlist_dst_size; 4379 4380 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 4381 return (error); 4382 4383 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst, 4384 &count); 4385 if (error == 0) 4386 zc->zc_nvlist_dst_size = count; 4387 else 4388 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa); 4389 4390 spa_close(spa, FTAG); 4391 4392 return (error); 4393} 4394 4395static int 4396zfs_ioc_clear(zfs_cmd_t *zc) 4397{ 4398 spa_t *spa; 4399 vdev_t *vd; 4400 int error; 4401 4402 /* 4403 * On zpool clear we also fix up missing slogs 4404 */ 4405 mutex_enter(&spa_namespace_lock); 4406 spa = spa_lookup(zc->zc_name); 4407 if (spa == NULL) { 4408 mutex_exit(&spa_namespace_lock); 4409 return (SET_ERROR(EIO)); 4410 } 4411 if (spa_get_log_state(spa) == SPA_LOG_MISSING) { 4412 /* we need to let spa_open/spa_load clear the chains */ 4413 spa_set_log_state(spa, SPA_LOG_CLEAR); 4414 } 4415 spa->spa_last_open_failed = 0; 4416 mutex_exit(&spa_namespace_lock); 4417 4418 if (zc->zc_cookie & ZPOOL_NO_REWIND) { 4419 error = spa_open(zc->zc_name, &spa, FTAG); 4420 } else { 4421 nvlist_t *policy; 4422 nvlist_t *config = NULL; 4423 4424 if (zc->zc_nvlist_src == 0) 4425 return (SET_ERROR(EINVAL)); 4426 4427 if ((error = get_nvlist(zc->zc_nvlist_src, 4428 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) { 4429 error = spa_open_rewind(zc->zc_name, &spa, FTAG, 4430 policy, &config); 4431 if (config != NULL) { 4432 int err; 4433 4434 if ((err = put_nvlist(zc, config)) != 0) 4435 error = err; 4436 nvlist_free(config); 4437 } 4438 nvlist_free(policy); 4439 } 4440 } 4441 4442 if (error != 0) 4443 return (error); 4444 4445 spa_vdev_state_enter(spa, SCL_NONE); 4446 4447 if (zc->zc_guid == 0) { 4448 vd = NULL; 4449 } else { 4450 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE); 4451 if (vd == NULL) { 4452 (void) spa_vdev_state_exit(spa, NULL, ENODEV); 4453 spa_close(spa, FTAG); 4454 return (SET_ERROR(ENODEV)); 4455 } 4456 } 4457 4458 vdev_clear(spa, vd); 4459 4460 (void) spa_vdev_state_exit(spa, NULL, 0); 4461 4462 /* 4463 * Resume any suspended I/Os. 4464 */ 4465 if (zio_resume(spa) != 0) 4466 error = SET_ERROR(EIO); 4467 4468 spa_close(spa, FTAG); 4469 4470 return (error); 4471} 4472 4473static int 4474zfs_ioc_pool_reopen(zfs_cmd_t *zc) 4475{ 4476 spa_t *spa; 4477 int error; 4478 4479 error = spa_open(zc->zc_name, &spa, FTAG); 4480 if (error != 0) 4481 return (error); 4482 4483 spa_vdev_state_enter(spa, SCL_NONE); 4484 4485 /* 4486 * If a resilver is already in progress then set the 4487 * spa_scrub_reopen flag to B_TRUE so that we don't restart 4488 * the scan as a side effect of the reopen. Otherwise, let 4489 * vdev_open() decided if a resilver is required. 4490 */ 4491 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool); 4492 vdev_reopen(spa->spa_root_vdev); 4493 spa->spa_scrub_reopen = B_FALSE; 4494 4495 (void) spa_vdev_state_exit(spa, NULL, 0); 4496 spa_close(spa, FTAG); 4497 return (0); 4498} 4499/* 4500 * inputs: 4501 * zc_name name of filesystem 4502 * zc_value name of origin snapshot 4503 * 4504 * outputs: 4505 * zc_string name of conflicting snapshot, if there is one 4506 */ 4507static int 4508zfs_ioc_promote(zfs_cmd_t *zc) 4509{ 4510 char *cp; 4511 4512 /* 4513 * We don't need to unmount *all* the origin fs's snapshots, but 4514 * it's easier. 4515 */ 4516 cp = strchr(zc->zc_value, '@'); 4517 if (cp) 4518 *cp = '\0'; 4519 (void) dmu_objset_find(zc->zc_value, 4520 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS); 4521 return (dsl_dataset_promote(zc->zc_name, zc->zc_string)); 4522} 4523 4524/* 4525 * Retrieve a single {user|group}{used|quota}@... property. 4526 * 4527 * inputs: 4528 * zc_name name of filesystem 4529 * zc_objset_type zfs_userquota_prop_t 4530 * zc_value domain name (eg. "S-1-234-567-89") 4531 * zc_guid RID/UID/GID 4532 * 4533 * outputs: 4534 * zc_cookie property value 4535 */ 4536static int 4537zfs_ioc_userspace_one(zfs_cmd_t *zc) 4538{ 4539 zfsvfs_t *zfsvfs; 4540 int error; 4541 4542 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 4543 return (SET_ERROR(EINVAL)); 4544 4545 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE); 4546 if (error != 0) 4547 return (error); 4548 4549 error = zfs_userspace_one(zfsvfs, 4550 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie); 4551 zfsvfs_rele(zfsvfs, FTAG); 4552 4553 return (error); 4554} 4555 4556/* 4557 * inputs: 4558 * zc_name name of filesystem 4559 * zc_cookie zap cursor 4560 * zc_objset_type zfs_userquota_prop_t 4561 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist) 4562 * 4563 * outputs: 4564 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t) 4565 * zc_cookie zap cursor 4566 */ 4567static int 4568zfs_ioc_userspace_many(zfs_cmd_t *zc) 4569{ 4570 zfsvfs_t *zfsvfs; 4571 int bufsize = zc->zc_nvlist_dst_size; 4572 4573 if (bufsize <= 0) 4574 return (SET_ERROR(ENOMEM)); 4575 4576 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE); 4577 if (error != 0) 4578 return (error); 4579 4580 void *buf = kmem_alloc(bufsize, KM_SLEEP); 4581 4582 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie, 4583 buf, &zc->zc_nvlist_dst_size); 4584 4585 if (error == 0) { 4586 error = ddi_copyout(buf, 4587 (void *)(uintptr_t)zc->zc_nvlist_dst, 4588 zc->zc_nvlist_dst_size, zc->zc_iflags); 4589 } 4590 kmem_free(buf, bufsize); 4591 zfsvfs_rele(zfsvfs, FTAG); 4592 4593 return (error); 4594} 4595 4596/* 4597 * inputs: 4598 * zc_name name of filesystem 4599 * 4600 * outputs: 4601 * none 4602 */ 4603static int 4604zfs_ioc_userspace_upgrade(zfs_cmd_t *zc) 4605{ 4606 objset_t *os; 4607 int error = 0; 4608 zfsvfs_t *zfsvfs; 4609 4610 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) { 4611 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) { 4612 /* 4613 * If userused is not enabled, it may be because the 4614 * objset needs to be closed & reopened (to grow the 4615 * objset_phys_t). Suspend/resume the fs will do that. 4616 */ 4617 error = zfs_suspend_fs(zfsvfs); 4618 if (error == 0) { 4619 dmu_objset_refresh_ownership(zfsvfs->z_os, 4620 zfsvfs); 4621 error = zfs_resume_fs(zfsvfs, zc->zc_name); 4622 } 4623 } 4624 if (error == 0) 4625 error = dmu_objset_userspace_upgrade(zfsvfs->z_os); 4626 VFS_RELE(zfsvfs->z_vfs); 4627 } else { 4628 /* XXX kind of reading contents without owning */ 4629 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 4630 if (error != 0) 4631 return (error); 4632 4633 error = dmu_objset_userspace_upgrade(os); 4634 dmu_objset_rele(os, FTAG); 4635 } 4636 4637 return (error); 4638} 4639 4640#ifdef sun 4641/* 4642 * We don't want to have a hard dependency 4643 * against some special symbols in sharefs 4644 * nfs, and smbsrv. Determine them if needed when 4645 * the first file system is shared. 4646 * Neither sharefs, nfs or smbsrv are unloadable modules. 4647 */ 4648int (*znfsexport_fs)(void *arg); 4649int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t); 4650int (*zsmbexport_fs)(void *arg, boolean_t add_share); 4651 4652int zfs_nfsshare_inited; 4653int zfs_smbshare_inited; 4654 4655ddi_modhandle_t nfs_mod; 4656ddi_modhandle_t sharefs_mod; 4657ddi_modhandle_t smbsrv_mod; 4658#endif /* sun */ 4659kmutex_t zfs_share_lock; 4660 4661#ifdef sun 4662static int 4663zfs_init_sharefs() 4664{ 4665 int error; 4666 4667 ASSERT(MUTEX_HELD(&zfs_share_lock)); 4668 /* Both NFS and SMB shares also require sharetab support. */ 4669 if (sharefs_mod == NULL && ((sharefs_mod = 4670 ddi_modopen("fs/sharefs", 4671 KRTLD_MODE_FIRST, &error)) == NULL)) { 4672 return (SET_ERROR(ENOSYS)); 4673 } 4674 if (zshare_fs == NULL && ((zshare_fs = 4675 (int (*)(enum sharefs_sys_op, share_t *, uint32_t)) 4676 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) { 4677 return (SET_ERROR(ENOSYS)); 4678 } 4679 return (0); 4680} 4681#endif /* sun */ 4682 4683static int 4684zfs_ioc_share(zfs_cmd_t *zc) 4685{ 4686#ifdef sun 4687 int error; 4688 int opcode; 4689 4690 switch (zc->zc_share.z_sharetype) { 4691 case ZFS_SHARE_NFS: 4692 case ZFS_UNSHARE_NFS: 4693 if (zfs_nfsshare_inited == 0) { 4694 mutex_enter(&zfs_share_lock); 4695 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs", 4696 KRTLD_MODE_FIRST, &error)) == NULL)) { 4697 mutex_exit(&zfs_share_lock); 4698 return (SET_ERROR(ENOSYS)); 4699 } 4700 if (znfsexport_fs == NULL && 4701 ((znfsexport_fs = (int (*)(void *)) 4702 ddi_modsym(nfs_mod, 4703 "nfs_export", &error)) == NULL)) { 4704 mutex_exit(&zfs_share_lock); 4705 return (SET_ERROR(ENOSYS)); 4706 } 4707 error = zfs_init_sharefs(); 4708 if (error != 0) { 4709 mutex_exit(&zfs_share_lock); 4710 return (SET_ERROR(ENOSYS)); 4711 } 4712 zfs_nfsshare_inited = 1; 4713 mutex_exit(&zfs_share_lock); 4714 } 4715 break; 4716 case ZFS_SHARE_SMB: 4717 case ZFS_UNSHARE_SMB: 4718 if (zfs_smbshare_inited == 0) { 4719 mutex_enter(&zfs_share_lock); 4720 if (smbsrv_mod == NULL && ((smbsrv_mod = 4721 ddi_modopen("drv/smbsrv", 4722 KRTLD_MODE_FIRST, &error)) == NULL)) { 4723 mutex_exit(&zfs_share_lock); 4724 return (SET_ERROR(ENOSYS)); 4725 } 4726 if (zsmbexport_fs == NULL && ((zsmbexport_fs = 4727 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod, 4728 "smb_server_share", &error)) == NULL)) { 4729 mutex_exit(&zfs_share_lock); 4730 return (SET_ERROR(ENOSYS)); 4731 } 4732 error = zfs_init_sharefs(); 4733 if (error != 0) { 4734 mutex_exit(&zfs_share_lock); 4735 return (SET_ERROR(ENOSYS)); 4736 } 4737 zfs_smbshare_inited = 1; 4738 mutex_exit(&zfs_share_lock); 4739 } 4740 break; 4741 default: 4742 return (SET_ERROR(EINVAL)); 4743 } 4744 4745 switch (zc->zc_share.z_sharetype) { 4746 case ZFS_SHARE_NFS: 4747 case ZFS_UNSHARE_NFS: 4748 if (error = 4749 znfsexport_fs((void *) 4750 (uintptr_t)zc->zc_share.z_exportdata)) 4751 return (error); 4752 break; 4753 case ZFS_SHARE_SMB: 4754 case ZFS_UNSHARE_SMB: 4755 if (error = zsmbexport_fs((void *) 4756 (uintptr_t)zc->zc_share.z_exportdata, 4757 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ? 4758 B_TRUE: B_FALSE)) { 4759 return (error); 4760 } 4761 break; 4762 } 4763 4764 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS || 4765 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ? 4766 SHAREFS_ADD : SHAREFS_REMOVE; 4767 4768 /* 4769 * Add or remove share from sharetab 4770 */ 4771 error = zshare_fs(opcode, 4772 (void *)(uintptr_t)zc->zc_share.z_sharedata, 4773 zc->zc_share.z_sharemax); 4774 4775 return (error); 4776 4777#else /* !sun */ 4778 return (ENOSYS); 4779#endif /* !sun */ 4780} 4781 4782ace_t full_access[] = { 4783 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0} 4784}; 4785 4786/* 4787 * inputs: 4788 * zc_name name of containing filesystem 4789 * zc_obj object # beyond which we want next in-use object # 4790 * 4791 * outputs: 4792 * zc_obj next in-use object # 4793 */ 4794static int 4795zfs_ioc_next_obj(zfs_cmd_t *zc) 4796{ 4797 objset_t *os = NULL; 4798 int error; 4799 4800 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 4801 if (error != 0) 4802 return (error); 4803 4804 error = dmu_object_next(os, &zc->zc_obj, B_FALSE, 4805 os->os_dsl_dataset->ds_phys->ds_prev_snap_txg); 4806 4807 dmu_objset_rele(os, FTAG); 4808 return (error); 4809} 4810 4811/* 4812 * inputs: 4813 * zc_name name of filesystem 4814 * zc_value prefix name for snapshot 4815 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process 4816 * 4817 * outputs: 4818 * zc_value short name of new snapshot 4819 */ 4820static int 4821zfs_ioc_tmp_snapshot(zfs_cmd_t *zc) 4822{ 4823 char *snap_name; 4824 char *hold_name; 4825 int error; 4826 minor_t minor; 4827 4828 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor); 4829 if (error != 0) 4830 return (error); 4831 4832 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value, 4833 (u_longlong_t)ddi_get_lbolt64()); 4834 hold_name = kmem_asprintf("%%%s", zc->zc_value); 4835 4836 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor, 4837 hold_name); 4838 if (error == 0) 4839 (void) strcpy(zc->zc_value, snap_name); 4840 strfree(snap_name); 4841 strfree(hold_name); 4842 zfs_onexit_fd_rele(zc->zc_cleanup_fd); 4843 return (error); 4844} 4845 4846/* 4847 * inputs: 4848 * zc_name name of "to" snapshot 4849 * zc_value name of "from" snapshot 4850 * zc_cookie file descriptor to write diff data on 4851 * 4852 * outputs: 4853 * dmu_diff_record_t's to the file descriptor 4854 */ 4855static int 4856zfs_ioc_diff(zfs_cmd_t *zc) 4857{ 4858 file_t *fp; 4859 cap_rights_t rights; 4860 offset_t off; 4861 int error; 4862 4863 fp = getf(zc->zc_cookie, cap_rights_init(&rights, CAP_WRITE)); 4864 if (fp == NULL) 4865 return (SET_ERROR(EBADF)); 4866 4867 off = fp->f_offset; 4868 4869#ifdef illumos 4870 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off); 4871#else 4872 error = dmu_diff(zc->zc_name, zc->zc_value, fp, &off); 4873#endif 4874 4875 if (off >= 0 && off <= MAXOFFSET_T) 4876 fp->f_offset = off; 4877 releasef(zc->zc_cookie); 4878 4879 return (error); 4880} 4881 4882#ifdef sun 4883/* 4884 * Remove all ACL files in shares dir 4885 */ 4886static int 4887zfs_smb_acl_purge(znode_t *dzp) 4888{ 4889 zap_cursor_t zc; 4890 zap_attribute_t zap; 4891 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 4892 int error; 4893 4894 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id); 4895 (error = zap_cursor_retrieve(&zc, &zap)) == 0; 4896 zap_cursor_advance(&zc)) { 4897 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred, 4898 NULL, 0)) != 0) 4899 break; 4900 } 4901 zap_cursor_fini(&zc); 4902 return (error); 4903} 4904#endif /* sun */ 4905 4906static int 4907zfs_ioc_smb_acl(zfs_cmd_t *zc) 4908{ 4909#ifdef sun 4910 vnode_t *vp; 4911 znode_t *dzp; 4912 vnode_t *resourcevp = NULL; 4913 znode_t *sharedir; 4914 zfsvfs_t *zfsvfs; 4915 nvlist_t *nvlist; 4916 char *src, *target; 4917 vattr_t vattr; 4918 vsecattr_t vsec; 4919 int error = 0; 4920 4921 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, 4922 NO_FOLLOW, NULL, &vp)) != 0) 4923 return (error); 4924 4925 /* Now make sure mntpnt and dataset are ZFS */ 4926 4927 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 || 4928 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), 4929 zc->zc_name) != 0)) { 4930 VN_RELE(vp); 4931 return (SET_ERROR(EINVAL)); 4932 } 4933 4934 dzp = VTOZ(vp); 4935 zfsvfs = dzp->z_zfsvfs; 4936 ZFS_ENTER(zfsvfs); 4937 4938 /* 4939 * Create share dir if its missing. 4940 */ 4941 mutex_enter(&zfsvfs->z_lock); 4942 if (zfsvfs->z_shares_dir == 0) { 4943 dmu_tx_t *tx; 4944 4945 tx = dmu_tx_create(zfsvfs->z_os); 4946 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE, 4947 ZFS_SHARES_DIR); 4948 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); 4949 error = dmu_tx_assign(tx, TXG_WAIT); 4950 if (error != 0) { 4951 dmu_tx_abort(tx); 4952 } else { 4953 error = zfs_create_share_dir(zfsvfs, tx); 4954 dmu_tx_commit(tx); 4955 } 4956 if (error != 0) { 4957 mutex_exit(&zfsvfs->z_lock); 4958 VN_RELE(vp); 4959 ZFS_EXIT(zfsvfs); 4960 return (error); 4961 } 4962 } 4963 mutex_exit(&zfsvfs->z_lock); 4964 4965 ASSERT(zfsvfs->z_shares_dir); 4966 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) { 4967 VN_RELE(vp); 4968 ZFS_EXIT(zfsvfs); 4969 return (error); 4970 } 4971 4972 switch (zc->zc_cookie) { 4973 case ZFS_SMB_ACL_ADD: 4974 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; 4975 vattr.va_type = VREG; 4976 vattr.va_mode = S_IFREG|0777; 4977 vattr.va_uid = 0; 4978 vattr.va_gid = 0; 4979 4980 vsec.vsa_mask = VSA_ACE; 4981 vsec.vsa_aclentp = &full_access; 4982 vsec.vsa_aclentsz = sizeof (full_access); 4983 vsec.vsa_aclcnt = 1; 4984 4985 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string, 4986 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec); 4987 if (resourcevp) 4988 VN_RELE(resourcevp); 4989 break; 4990 4991 case ZFS_SMB_ACL_REMOVE: 4992 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred, 4993 NULL, 0); 4994 break; 4995 4996 case ZFS_SMB_ACL_RENAME: 4997 if ((error = get_nvlist(zc->zc_nvlist_src, 4998 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) { 4999 VN_RELE(vp); 5000 ZFS_EXIT(zfsvfs); 5001 return (error); 5002 } 5003 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) || 5004 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET, 5005 &target)) { 5006 VN_RELE(vp); 5007 VN_RELE(ZTOV(sharedir)); 5008 ZFS_EXIT(zfsvfs); 5009 nvlist_free(nvlist); 5010 return (error); 5011 } 5012 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target, 5013 kcred, NULL, 0); 5014 nvlist_free(nvlist); 5015 break; 5016 5017 case ZFS_SMB_ACL_PURGE: 5018 error = zfs_smb_acl_purge(sharedir); 5019 break; 5020 5021 default: 5022 error = SET_ERROR(EINVAL); 5023 break; 5024 } 5025 5026 VN_RELE(vp); 5027 VN_RELE(ZTOV(sharedir)); 5028 5029 ZFS_EXIT(zfsvfs); 5030 5031 return (error); 5032#else /* !sun */ 5033 return (EOPNOTSUPP); 5034#endif /* !sun */ 5035} 5036 5037/* 5038 * innvl: { 5039 * "holds" -> { snapname -> holdname (string), ... } 5040 * (optional) "cleanup_fd" -> fd (int32) 5041 * } 5042 * 5043 * outnvl: { 5044 * snapname -> error value (int32) 5045 * ... 5046 * } 5047 */ 5048/* ARGSUSED */ 5049static int 5050zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist) 5051{ 5052 nvlist_t *holds; 5053 int cleanup_fd = -1; 5054 int error; 5055 minor_t minor = 0; 5056 5057 error = nvlist_lookup_nvlist(args, "holds", &holds); 5058 if (error != 0) 5059 return (SET_ERROR(EINVAL)); 5060 5061 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) { 5062 error = zfs_onexit_fd_hold(cleanup_fd, &minor); 5063 if (error != 0) 5064 return (error); 5065 } 5066 5067 error = dsl_dataset_user_hold(holds, minor, errlist); 5068 if (minor != 0) 5069 zfs_onexit_fd_rele(cleanup_fd); 5070 return (error); 5071} 5072 5073/* 5074 * innvl is not used. 5075 * 5076 * outnvl: { 5077 * holdname -> time added (uint64 seconds since epoch) 5078 * ... 5079 * } 5080 */ 5081/* ARGSUSED */ 5082static int 5083zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl) 5084{ 5085 return (dsl_dataset_get_holds(snapname, outnvl)); 5086} 5087 5088/* 5089 * innvl: { 5090 * snapname -> { holdname, ... } 5091 * ... 5092 * } 5093 * 5094 * outnvl: { 5095 * snapname -> error value (int32) 5096 * ... 5097 * } 5098 */ 5099/* ARGSUSED */ 5100static int 5101zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist) 5102{ 5103 return (dsl_dataset_user_release(holds, errlist)); 5104} 5105 5106/* 5107 * inputs: 5108 * zc_name name of new filesystem or snapshot 5109 * zc_value full name of old snapshot 5110 * 5111 * outputs: 5112 * zc_cookie space in bytes 5113 * zc_objset_type compressed space in bytes 5114 * zc_perm_action uncompressed space in bytes 5115 */ 5116static int 5117zfs_ioc_space_written(zfs_cmd_t *zc) 5118{ 5119 int error; 5120 dsl_pool_t *dp; 5121 dsl_dataset_t *new, *old; 5122 5123 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 5124 if (error != 0) 5125 return (error); 5126 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new); 5127 if (error != 0) { 5128 dsl_pool_rele(dp, FTAG); 5129 return (error); 5130 } 5131 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old); 5132 if (error != 0) { 5133 dsl_dataset_rele(new, FTAG); 5134 dsl_pool_rele(dp, FTAG); 5135 return (error); 5136 } 5137 5138 error = dsl_dataset_space_written(old, new, &zc->zc_cookie, 5139 &zc->zc_objset_type, &zc->zc_perm_action); 5140 dsl_dataset_rele(old, FTAG); 5141 dsl_dataset_rele(new, FTAG); 5142 dsl_pool_rele(dp, FTAG); 5143 return (error); 5144} 5145 5146/* 5147 * innvl: { 5148 * "firstsnap" -> snapshot name 5149 * } 5150 * 5151 * outnvl: { 5152 * "used" -> space in bytes 5153 * "compressed" -> compressed space in bytes 5154 * "uncompressed" -> uncompressed space in bytes 5155 * } 5156 */ 5157static int 5158zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl) 5159{ 5160 int error; 5161 dsl_pool_t *dp; 5162 dsl_dataset_t *new, *old; 5163 char *firstsnap; 5164 uint64_t used, comp, uncomp; 5165 5166 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0) 5167 return (SET_ERROR(EINVAL)); 5168 5169 error = dsl_pool_hold(lastsnap, FTAG, &dp); 5170 if (error != 0) 5171 return (error); 5172 5173 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new); 5174 if (error != 0) { 5175 dsl_pool_rele(dp, FTAG); 5176 return (error); 5177 } 5178 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old); 5179 if (error != 0) { 5180 dsl_dataset_rele(new, FTAG); 5181 dsl_pool_rele(dp, FTAG); 5182 return (error); 5183 } 5184 5185 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp); 5186 dsl_dataset_rele(old, FTAG); 5187 dsl_dataset_rele(new, FTAG); 5188 dsl_pool_rele(dp, FTAG); 5189 fnvlist_add_uint64(outnvl, "used", used); 5190 fnvlist_add_uint64(outnvl, "compressed", comp); 5191 fnvlist_add_uint64(outnvl, "uncompressed", uncomp); 5192 return (error); 5193} 5194 5195static int 5196zfs_ioc_jail(zfs_cmd_t *zc) 5197{ 5198 5199 return (zone_dataset_attach(curthread->td_ucred, zc->zc_name, 5200 (int)zc->zc_jailid)); 5201} 5202 5203static int 5204zfs_ioc_unjail(zfs_cmd_t *zc) 5205{ 5206 5207 return (zone_dataset_detach(curthread->td_ucred, zc->zc_name, 5208 (int)zc->zc_jailid)); 5209} 5210 5211/* 5212 * innvl: { 5213 * "fd" -> file descriptor to write stream to (int32) 5214 * (optional) "fromsnap" -> full snap name to send an incremental from 5215 * } 5216 * 5217 * outnvl is unused 5218 */ 5219/* ARGSUSED */ 5220static int 5221zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl) 5222{ 5223 cap_rights_t rights; 5224 int error; 5225 offset_t off; 5226 char *fromname = NULL; 5227 int fd; 5228 5229 error = nvlist_lookup_int32(innvl, "fd", &fd); 5230 if (error != 0) 5231 return (SET_ERROR(EINVAL)); 5232 5233 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname); 5234 5235 file_t *fp = getf(fd, cap_rights_init(&rights, CAP_READ)); 5236 if (fp == NULL) 5237 return (SET_ERROR(EBADF)); 5238 5239 off = fp->f_offset; 5240#ifdef illumos 5241 error = dmu_send(snapname, fromname, fd, fp->f_vnode, &off); 5242#else 5243 error = dmu_send(snapname, fromname, fd, fp, &off); 5244#endif 5245 5246#ifdef illumos 5247 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0) 5248 fp->f_offset = off; 5249#else 5250 fp->f_offset = off; 5251#endif 5252 5253 releasef(fd); 5254 return (error); 5255} 5256 5257/* 5258 * Determine approximately how large a zfs send stream will be -- the number 5259 * of bytes that will be written to the fd supplied to zfs_ioc_send_new(). 5260 * 5261 * innvl: { 5262 * (optional) "fromsnap" -> full snap name to send an incremental from 5263 * } 5264 * 5265 * outnvl: { 5266 * "space" -> bytes of space (uint64) 5267 * } 5268 */ 5269static int 5270zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl) 5271{ 5272 dsl_pool_t *dp; 5273 dsl_dataset_t *fromsnap = NULL; 5274 dsl_dataset_t *tosnap; 5275 int error; 5276 char *fromname; 5277 uint64_t space; 5278 5279 error = dsl_pool_hold(snapname, FTAG, &dp); 5280 if (error != 0) 5281 return (error); 5282 5283 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap); 5284 if (error != 0) { 5285 dsl_pool_rele(dp, FTAG); 5286 return (error); 5287 } 5288 5289 error = nvlist_lookup_string(innvl, "fromsnap", &fromname); 5290 if (error == 0) { 5291 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap); 5292 if (error != 0) { 5293 dsl_dataset_rele(tosnap, FTAG); 5294 dsl_pool_rele(dp, FTAG); 5295 return (error); 5296 } 5297 } 5298 5299 error = dmu_send_estimate(tosnap, fromsnap, &space); 5300 fnvlist_add_uint64(outnvl, "space", space); 5301 5302 if (fromsnap != NULL) 5303 dsl_dataset_rele(fromsnap, FTAG); 5304 dsl_dataset_rele(tosnap, FTAG); 5305 dsl_pool_rele(dp, FTAG); 5306 return (error); 5307} 5308 5309 5310static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST]; 5311 5312static void 5313zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5314 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck, 5315 boolean_t log_history, zfs_ioc_poolcheck_t pool_check) 5316{ 5317 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST]; 5318 5319 ASSERT3U(ioc, >=, ZFS_IOC_FIRST); 5320 ASSERT3U(ioc, <, ZFS_IOC_LAST); 5321 ASSERT3P(vec->zvec_legacy_func, ==, NULL); 5322 ASSERT3P(vec->zvec_func, ==, NULL); 5323 5324 vec->zvec_legacy_func = func; 5325 vec->zvec_secpolicy = secpolicy; 5326 vec->zvec_namecheck = namecheck; 5327 vec->zvec_allow_log = log_history; 5328 vec->zvec_pool_check = pool_check; 5329} 5330 5331/* 5332 * See the block comment at the beginning of this file for details on 5333 * each argument to this function. 5334 */ 5335static void 5336zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func, 5337 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck, 5338 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist, 5339 boolean_t allow_log) 5340{ 5341 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST]; 5342 5343 ASSERT3U(ioc, >=, ZFS_IOC_FIRST); 5344 ASSERT3U(ioc, <, ZFS_IOC_LAST); 5345 ASSERT3P(vec->zvec_legacy_func, ==, NULL); 5346 ASSERT3P(vec->zvec_func, ==, NULL); 5347 5348 /* if we are logging, the name must be valid */ 5349 ASSERT(!allow_log || namecheck != NO_NAME); 5350 5351 vec->zvec_name = name; 5352 vec->zvec_func = func; 5353 vec->zvec_secpolicy = secpolicy; 5354 vec->zvec_namecheck = namecheck; 5355 vec->zvec_pool_check = pool_check; 5356 vec->zvec_smush_outnvlist = smush_outnvlist; 5357 vec->zvec_allow_log = allow_log; 5358} 5359 5360static void 5361zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5362 zfs_secpolicy_func_t *secpolicy, boolean_t log_history, 5363 zfs_ioc_poolcheck_t pool_check) 5364{ 5365 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5366 POOL_NAME, log_history, pool_check); 5367} 5368 5369static void 5370zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5371 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check) 5372{ 5373 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5374 DATASET_NAME, B_FALSE, pool_check); 5375} 5376 5377static void 5378zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func) 5379{ 5380 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config, 5381 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5382} 5383 5384static void 5385zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5386 zfs_secpolicy_func_t *secpolicy) 5387{ 5388 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5389 NO_NAME, B_FALSE, POOL_CHECK_NONE); 5390} 5391 5392static void 5393zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc, 5394 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy) 5395{ 5396 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5397 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED); 5398} 5399 5400static void 5401zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func) 5402{ 5403 zfs_ioctl_register_dataset_read_secpolicy(ioc, func, 5404 zfs_secpolicy_read); 5405} 5406 5407static void 5408zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5409 zfs_secpolicy_func_t *secpolicy) 5410{ 5411 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5412 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5413} 5414 5415static void 5416zfs_ioctl_init(void) 5417{ 5418 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT, 5419 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME, 5420 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5421 5422 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY, 5423 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME, 5424 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE); 5425 5426 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS, 5427 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME, 5428 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5429 5430 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW, 5431 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME, 5432 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5433 5434 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE, 5435 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME, 5436 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5437 5438 zfs_ioctl_register("create", ZFS_IOC_CREATE, 5439 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME, 5440 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5441 5442 zfs_ioctl_register("clone", ZFS_IOC_CLONE, 5443 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME, 5444 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5445 5446 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS, 5447 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME, 5448 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5449 5450 zfs_ioctl_register("hold", ZFS_IOC_HOLD, 5451 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME, 5452 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5453 zfs_ioctl_register("release", ZFS_IOC_RELEASE, 5454 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME, 5455 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5456 5457 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS, 5458 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME, 5459 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5460 5461 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK, 5462 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, 5463 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE); 5464 5465 /* IOCTLS that use the legacy function signature */ 5466 5467 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze, 5468 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY); 5469 5470 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create, 5471 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5472 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN, 5473 zfs_ioc_pool_scan); 5474 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE, 5475 zfs_ioc_pool_upgrade); 5476 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD, 5477 zfs_ioc_vdev_add); 5478 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE, 5479 zfs_ioc_vdev_remove); 5480 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE, 5481 zfs_ioc_vdev_set_state); 5482 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH, 5483 zfs_ioc_vdev_attach); 5484 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH, 5485 zfs_ioc_vdev_detach); 5486 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH, 5487 zfs_ioc_vdev_setpath); 5488 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU, 5489 zfs_ioc_vdev_setfru); 5490 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS, 5491 zfs_ioc_pool_set_props); 5492 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT, 5493 zfs_ioc_vdev_split); 5494 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID, 5495 zfs_ioc_pool_reguid); 5496 5497 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS, 5498 zfs_ioc_pool_configs, zfs_secpolicy_none); 5499 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT, 5500 zfs_ioc_pool_tryimport, zfs_secpolicy_config); 5501 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT, 5502 zfs_ioc_inject_fault, zfs_secpolicy_inject); 5503 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT, 5504 zfs_ioc_clear_fault, zfs_secpolicy_inject); 5505 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT, 5506 zfs_ioc_inject_list_next, zfs_secpolicy_inject); 5507 5508 /* 5509 * pool destroy, and export don't log the history as part of 5510 * zfsdev_ioctl, but rather zfs_ioc_pool_export 5511 * does the logging of those commands. 5512 */ 5513 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy, 5514 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE); 5515 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export, 5516 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE); 5517 5518 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats, 5519 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE); 5520 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props, 5521 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE); 5522 5523 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log, 5524 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_NONE); 5525 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME, 5526 zfs_ioc_dsobj_to_dsname, 5527 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_NONE); 5528 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY, 5529 zfs_ioc_pool_get_history, 5530 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED); 5531 5532 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import, 5533 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5534 5535 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear, 5536 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5537 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen, 5538 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED); 5539 5540 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN, 5541 zfs_ioc_space_written); 5542 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS, 5543 zfs_ioc_objset_recvd_props); 5544 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ, 5545 zfs_ioc_next_obj); 5546 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL, 5547 zfs_ioc_get_fsacl); 5548 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS, 5549 zfs_ioc_objset_stats); 5550 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS, 5551 zfs_ioc_objset_zplprops); 5552 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT, 5553 zfs_ioc_dataset_list_next); 5554 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT, 5555 zfs_ioc_snapshot_list_next); 5556 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS, 5557 zfs_ioc_send_progress); 5558 5559 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF, 5560 zfs_ioc_diff, zfs_secpolicy_diff); 5561 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS, 5562 zfs_ioc_obj_to_stats, zfs_secpolicy_diff); 5563 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH, 5564 zfs_ioc_obj_to_path, zfs_secpolicy_diff); 5565 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE, 5566 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one); 5567 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY, 5568 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many); 5569 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND, 5570 zfs_ioc_send, zfs_secpolicy_send); 5571 5572 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop, 5573 zfs_secpolicy_none); 5574 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy, 5575 zfs_secpolicy_destroy); 5576 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename, 5577 zfs_secpolicy_rename); 5578 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv, 5579 zfs_secpolicy_recv); 5580 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote, 5581 zfs_secpolicy_promote); 5582 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP, 5583 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop); 5584 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl, 5585 zfs_secpolicy_set_fsacl); 5586 5587 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share, 5588 zfs_secpolicy_share, POOL_CHECK_NONE); 5589 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl, 5590 zfs_secpolicy_smb_acl, POOL_CHECK_NONE); 5591 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE, 5592 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade, 5593 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5594 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT, 5595 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot, 5596 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5597 5598#ifdef __FreeBSD__ 5599 zfs_ioctl_register_dataset_nolog(ZFS_IOC_JAIL, zfs_ioc_jail, 5600 zfs_secpolicy_config, POOL_CHECK_NONE); 5601 zfs_ioctl_register_dataset_nolog(ZFS_IOC_UNJAIL, zfs_ioc_unjail, 5602 zfs_secpolicy_config, POOL_CHECK_NONE); 5603#endif 5604} 5605 5606int 5607pool_status_check(const char *name, zfs_ioc_namecheck_t type, 5608 zfs_ioc_poolcheck_t check) 5609{ 5610 spa_t *spa; 5611 int error; 5612 5613 ASSERT(type == POOL_NAME || type == DATASET_NAME); 5614 5615 if (check & POOL_CHECK_NONE) 5616 return (0); 5617 5618 error = spa_open(name, &spa, FTAG); 5619 if (error == 0) { 5620 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa)) 5621 error = SET_ERROR(EAGAIN); 5622 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa)) 5623 error = SET_ERROR(EROFS); 5624 spa_close(spa, FTAG); 5625 } 5626 return (error); 5627} 5628 5629/* 5630 * Find a free minor number. 5631 */ 5632minor_t 5633zfsdev_minor_alloc(void) 5634{ 5635 static minor_t last_minor; 5636 minor_t m; 5637 5638 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 5639 5640 for (m = last_minor + 1; m != last_minor; m++) { 5641 if (m > ZFSDEV_MAX_MINOR) 5642 m = 1; 5643 if (ddi_get_soft_state(zfsdev_state, m) == NULL) { 5644 last_minor = m; 5645 return (m); 5646 } 5647 } 5648 5649 return (0); 5650} 5651 5652static int 5653zfs_ctldev_init(struct cdev *devp) 5654{ 5655 minor_t minor; 5656 zfs_soft_state_t *zs; 5657 5658 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 5659 5660 minor = zfsdev_minor_alloc(); 5661 if (minor == 0) 5662 return (SET_ERROR(ENXIO)); 5663 5664 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS) 5665 return (SET_ERROR(EAGAIN)); 5666 5667 devfs_set_cdevpriv((void *)(uintptr_t)minor, zfsdev_close); 5668 5669 zs = ddi_get_soft_state(zfsdev_state, minor); 5670 zs->zss_type = ZSST_CTLDEV; 5671 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data); 5672 5673 return (0); 5674} 5675 5676static void 5677zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor) 5678{ 5679 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 5680 5681 zfs_onexit_destroy(zo); 5682 ddi_soft_state_free(zfsdev_state, minor); 5683} 5684 5685void * 5686zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which) 5687{ 5688 zfs_soft_state_t *zp; 5689 5690 zp = ddi_get_soft_state(zfsdev_state, minor); 5691 if (zp == NULL || zp->zss_type != which) 5692 return (NULL); 5693 5694 return (zp->zss_data); 5695} 5696 5697static int 5698zfsdev_open(struct cdev *devp, int flag, int mode, struct thread *td) 5699{ 5700 int error = 0; 5701 5702#ifdef sun 5703 if (getminor(*devp) != 0) 5704 return (zvol_open(devp, flag, otyp, cr)); 5705#endif 5706 5707 /* This is the control device. Allocate a new minor if requested. */ 5708 if (flag & FEXCL) { 5709 mutex_enter(&spa_namespace_lock); 5710 error = zfs_ctldev_init(devp); 5711 mutex_exit(&spa_namespace_lock); 5712 } 5713 5714 return (error); 5715} 5716 5717static void 5718zfsdev_close(void *data) 5719{ 5720 zfs_onexit_t *zo; 5721 minor_t minor = (minor_t)(uintptr_t)data; 5722 5723 if (minor == 0) 5724 return; 5725 5726 mutex_enter(&spa_namespace_lock); 5727 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV); 5728 if (zo == NULL) { 5729 mutex_exit(&spa_namespace_lock); 5730 return; 5731 } 5732 zfs_ctldev_destroy(zo, minor); 5733 mutex_exit(&spa_namespace_lock); 5734} 5735 5736static int 5737zfsdev_ioctl(struct cdev *dev, u_long zcmd, caddr_t arg, int flag, 5738 struct thread *td) 5739{ 5740 zfs_cmd_t *zc; 5741 uint_t vecnum; 5742 int error, rc, len; 5743#ifdef illumos 5744 minor_t minor = getminor(dev); 5745#else 5746 zfs_iocparm_t *zc_iocparm; 5747 int cflag, cmd, oldvecnum; 5748 boolean_t newioc, compat; 5749 cred_t *cr = td->td_ucred; 5750#endif 5751 const zfs_ioc_vec_t *vec; 5752 char *saved_poolname = NULL; 5753 nvlist_t *innvl = NULL; 5754 5755 cflag = ZFS_CMD_COMPAT_NONE; 5756 compat = B_FALSE; 5757 newioc = B_TRUE; 5758 5759 len = IOCPARM_LEN(zcmd); 5760 cmd = zcmd & 0xff; 5761 5762 /* 5763 * Check if we are talking to supported older binaries 5764 * and translate zfs_cmd if necessary 5765 */ 5766 if (len != sizeof(zfs_iocparm_t)) { 5767 newioc = B_FALSE; 5768 if (len == sizeof(zfs_cmd_t)) { 5769 cflag = ZFS_CMD_COMPAT_LZC; 5770 vecnum = cmd; 5771 } else if (len == sizeof(zfs_cmd_deadman_t)) { 5772 cflag = ZFS_CMD_COMPAT_DEADMAN; 5773 compat = B_TRUE; 5774 vecnum = cmd; 5775 } else if (len == sizeof(zfs_cmd_v28_t)) { 5776 cflag = ZFS_CMD_COMPAT_V28; 5777 compat = B_TRUE; 5778 vecnum = cmd; 5779 } else if (len == sizeof(zfs_cmd_v15_t)) { 5780 cflag = ZFS_CMD_COMPAT_V15; 5781 compat = B_TRUE; 5782 vecnum = zfs_ioctl_v15_to_v28[cmd]; 5783 } else 5784 return (EINVAL); 5785 } else 5786 vecnum = cmd; 5787 5788#ifdef illumos 5789 vecnum = cmd - ZFS_IOC_FIRST; 5790 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip)); 5791#endif 5792 5793 if (compat) { 5794 if (vecnum == ZFS_IOC_COMPAT_PASS) 5795 return (0); 5796 else if (vecnum == ZFS_IOC_COMPAT_FAIL) 5797 return (ENOTSUP); 5798 } 5799 5800 /* 5801 * Check if we have sufficient kernel memory allocated 5802 * for the zfs_cmd_t request. Bail out if not so we 5803 * will not access undefined memory region. 5804 */ 5805 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0])) 5806 return (SET_ERROR(EINVAL)); 5807 vec = &zfs_ioc_vec[vecnum]; 5808 5809#ifdef illumos 5810 zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP); 5811 bzero(zc, sizeof(zfs_cmd_t)); 5812 5813 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag); 5814 if (error != 0) { 5815 error = SET_ERROR(EFAULT); 5816 goto out; 5817 } 5818#else /* !illumos */ 5819 /* 5820 * We don't alloc/free zc only if talking to library ioctl version 2 5821 */ 5822 if (cflag != ZFS_CMD_COMPAT_LZC) { 5823 zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP); 5824 bzero(zc, sizeof(zfs_cmd_t)); 5825 } else { 5826 zc = (void *)arg; 5827 error = 0; 5828 } 5829 5830 if (newioc) { 5831 zc_iocparm = (void *)arg; 5832 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_t)) { 5833 error = SET_ERROR(EFAULT); 5834 goto out; 5835 } 5836 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd, zc, 5837 sizeof(zfs_cmd_t), flag); 5838 if (error != 0) { 5839 error = SET_ERROR(EFAULT); 5840 goto out; 5841 } 5842 } 5843 5844 if (compat) { 5845 zfs_cmd_compat_get(zc, arg, cflag); 5846 oldvecnum = vecnum; 5847 error = zfs_ioctl_compat_pre(zc, &vecnum, cflag); 5848 if (error != 0) 5849 goto out; 5850 if (oldvecnum != vecnum) 5851 vec = &zfs_ioc_vec[vecnum]; 5852 } 5853#endif /* !illumos */ 5854 5855 zc->zc_iflags = flag & FKIOCTL; 5856 if (zc->zc_nvlist_src_size != 0) { 5857 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 5858 zc->zc_iflags, &innvl); 5859 if (error != 0) 5860 goto out; 5861 } 5862 5863 /* rewrite innvl for backwards compatibility */ 5864 if (compat) 5865 innvl = zfs_ioctl_compat_innvl(zc, innvl, vecnum, cflag); 5866 5867 /* 5868 * Ensure that all pool/dataset names are valid before we pass down to 5869 * the lower layers. 5870 */ 5871 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0'; 5872 switch (vec->zvec_namecheck) { 5873 case POOL_NAME: 5874 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0) 5875 error = SET_ERROR(EINVAL); 5876 else 5877 error = pool_status_check(zc->zc_name, 5878 vec->zvec_namecheck, vec->zvec_pool_check); 5879 break; 5880 5881 case DATASET_NAME: 5882 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0) 5883 error = SET_ERROR(EINVAL); 5884 else 5885 error = pool_status_check(zc->zc_name, 5886 vec->zvec_namecheck, vec->zvec_pool_check); 5887 break; 5888 5889 case NO_NAME: 5890 break; 5891 } 5892 5893 if (error == 0 && !(flag & FKIOCTL)) 5894 error = vec->zvec_secpolicy(zc, innvl, cr); 5895 5896 if (error != 0) 5897 goto out; 5898 5899 /* legacy ioctls can modify zc_name */ 5900 len = strcspn(zc->zc_name, "/@") + 1; 5901 saved_poolname = kmem_alloc(len, KM_SLEEP); 5902 (void) strlcpy(saved_poolname, zc->zc_name, len); 5903 5904 if (vec->zvec_func != NULL) { 5905 nvlist_t *outnvl; 5906 int puterror = 0; 5907 spa_t *spa; 5908 nvlist_t *lognv = NULL; 5909 5910 ASSERT(vec->zvec_legacy_func == NULL); 5911 5912 /* 5913 * Add the innvl to the lognv before calling the func, 5914 * in case the func changes the innvl. 5915 */ 5916 if (vec->zvec_allow_log) { 5917 lognv = fnvlist_alloc(); 5918 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL, 5919 vec->zvec_name); 5920 if (!nvlist_empty(innvl)) { 5921 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL, 5922 innvl); 5923 } 5924 } 5925 5926 outnvl = fnvlist_alloc(); 5927 error = vec->zvec_func(zc->zc_name, innvl, outnvl); 5928 5929 if (error == 0 && vec->zvec_allow_log && 5930 spa_open(zc->zc_name, &spa, FTAG) == 0) { 5931 if (!nvlist_empty(outnvl)) { 5932 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL, 5933 outnvl); 5934 } 5935 (void) spa_history_log_nvl(spa, lognv); 5936 spa_close(spa, FTAG); 5937 } 5938 fnvlist_free(lognv); 5939 5940 /* rewrite outnvl for backwards compatibility */ 5941 if (cflag != ZFS_CMD_COMPAT_NONE && cflag != ZFS_CMD_COMPAT_LZC) 5942 outnvl = zfs_ioctl_compat_outnvl(zc, outnvl, vecnum, 5943 cflag); 5944 5945 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) { 5946 int smusherror = 0; 5947 if (vec->zvec_smush_outnvlist) { 5948 smusherror = nvlist_smush(outnvl, 5949 zc->zc_nvlist_dst_size); 5950 } 5951 if (smusherror == 0) 5952 puterror = put_nvlist(zc, outnvl); 5953 } 5954 5955 if (puterror != 0) 5956 error = puterror; 5957 5958 nvlist_free(outnvl); 5959 } else { 5960 error = vec->zvec_legacy_func(zc); 5961 } 5962 5963out: 5964 nvlist_free(innvl); 5965 5966 if (compat) { 5967 zfs_ioctl_compat_post(zc, cmd, cflag); 5968 zfs_cmd_compat_put(zc, arg, vecnum, cflag); 5969 } 5970 5971#ifdef illumos 5972 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag); 5973 if (error == 0 && rc != 0) 5974 error = SET_ERROR(EFAULT); 5975#else 5976 if (newioc) { 5977 rc = ddi_copyout(zc, (void *)(uintptr_t)zc_iocparm->zfs_cmd, 5978 sizeof (zfs_cmd_t), flag); 5979 if (error == 0 && rc != 0) 5980 error = SET_ERROR(EFAULT); 5981 } 5982#endif 5983 if (error == 0 && vec->zvec_allow_log) { 5984 char *s = tsd_get(zfs_allow_log_key); 5985 if (s != NULL) 5986 strfree(s); 5987 (void) tsd_set(zfs_allow_log_key, saved_poolname); 5988 } else { 5989 if (saved_poolname != NULL) 5990 strfree(saved_poolname); 5991 } 5992 5993#ifdef illumos 5994 kmem_free(zc, sizeof (zfs_cmd_t)); 5995#else 5996 /* 5997 * We don't alloc/free zc only if talking to library ioctl version 2 5998 */ 5999 if (cflag != ZFS_CMD_COMPAT_LZC) 6000 kmem_free(zc, sizeof (zfs_cmd_t)); 6001#endif 6002 return (error); 6003} 6004 6005#ifdef sun 6006static int 6007zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd) 6008{ 6009 if (cmd != DDI_ATTACH) 6010 return (DDI_FAILURE); 6011 6012 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0, 6013 DDI_PSEUDO, 0) == DDI_FAILURE) 6014 return (DDI_FAILURE); 6015 6016 zfs_dip = dip; 6017 6018 ddi_report_dev(dip); 6019 6020 return (DDI_SUCCESS); 6021} 6022 6023static int 6024zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd) 6025{ 6026 if (spa_busy() || zfs_busy() || zvol_busy()) 6027 return (DDI_FAILURE); 6028 6029 if (cmd != DDI_DETACH) 6030 return (DDI_FAILURE); 6031 6032 zfs_dip = NULL; 6033 6034 ddi_prop_remove_all(dip); 6035 ddi_remove_minor_node(dip, NULL); 6036 6037 return (DDI_SUCCESS); 6038} 6039 6040/*ARGSUSED*/ 6041static int 6042zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result) 6043{ 6044 switch (infocmd) { 6045 case DDI_INFO_DEVT2DEVINFO: 6046 *result = zfs_dip; 6047 return (DDI_SUCCESS); 6048 6049 case DDI_INFO_DEVT2INSTANCE: 6050 *result = (void *)0; 6051 return (DDI_SUCCESS); 6052 } 6053 6054 return (DDI_FAILURE); 6055} 6056#endif /* sun */ 6057 6058/* 6059 * OK, so this is a little weird. 6060 * 6061 * /dev/zfs is the control node, i.e. minor 0. 6062 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0. 6063 * 6064 * /dev/zfs has basically nothing to do except serve up ioctls, 6065 * so most of the standard driver entry points are in zvol.c. 6066 */ 6067#ifdef sun 6068static struct cb_ops zfs_cb_ops = { 6069 zfsdev_open, /* open */ 6070 zfsdev_close, /* close */ 6071 zvol_strategy, /* strategy */ 6072 nodev, /* print */ 6073 zvol_dump, /* dump */ 6074 zvol_read, /* read */ 6075 zvol_write, /* write */ 6076 zfsdev_ioctl, /* ioctl */ 6077 nodev, /* devmap */ 6078 nodev, /* mmap */ 6079 nodev, /* segmap */ 6080 nochpoll, /* poll */ 6081 ddi_prop_op, /* prop_op */ 6082 NULL, /* streamtab */ 6083 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */ 6084 CB_REV, /* version */ 6085 nodev, /* async read */ 6086 nodev, /* async write */ 6087}; 6088 6089static struct dev_ops zfs_dev_ops = { 6090 DEVO_REV, /* version */ 6091 0, /* refcnt */ 6092 zfs_info, /* info */ 6093 nulldev, /* identify */ 6094 nulldev, /* probe */ 6095 zfs_attach, /* attach */ 6096 zfs_detach, /* detach */ 6097 nodev, /* reset */ 6098 &zfs_cb_ops, /* driver operations */ 6099 NULL, /* no bus operations */ 6100 NULL, /* power */ 6101 ddi_quiesce_not_needed, /* quiesce */ 6102}; 6103 6104static struct modldrv zfs_modldrv = { 6105 &mod_driverops, 6106 "ZFS storage pool", 6107 &zfs_dev_ops 6108}; 6109 6110static struct modlinkage modlinkage = { 6111 MODREV_1, 6112 (void *)&zfs_modlfs, 6113 (void *)&zfs_modldrv, 6114 NULL 6115}; 6116#endif /* sun */ 6117 6118static struct cdevsw zfs_cdevsw = { 6119 .d_version = D_VERSION, 6120 .d_open = zfsdev_open, 6121 .d_ioctl = zfsdev_ioctl, 6122 .d_name = ZFS_DEV_NAME 6123}; 6124 6125static void 6126zfs_allow_log_destroy(void *arg) 6127{ 6128 char *poolname = arg; 6129 strfree(poolname); 6130} 6131 6132static void 6133zfsdev_init(void) 6134{ 6135 zfsdev = make_dev(&zfs_cdevsw, 0x0, UID_ROOT, GID_OPERATOR, 0666, 6136 ZFS_DEV_NAME); 6137} 6138 6139static void 6140zfsdev_fini(void) 6141{ 6142 if (zfsdev != NULL) 6143 destroy_dev(zfsdev); 6144} 6145 6146static struct root_hold_token *zfs_root_token; 6147struct proc *zfsproc; 6148 6149#ifdef sun 6150int 6151_init(void) 6152{ 6153 int error; 6154 6155 spa_init(FREAD | FWRITE); 6156 zfs_init(); 6157 zvol_init(); 6158 zfs_ioctl_init(); 6159 6160 if ((error = mod_install(&modlinkage)) != 0) { 6161 zvol_fini(); 6162 zfs_fini(); 6163 spa_fini(); 6164 return (error); 6165 } 6166 6167 tsd_create(&zfs_fsyncer_key, NULL); 6168 tsd_create(&rrw_tsd_key, rrw_tsd_destroy); 6169 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy); 6170 6171 error = ldi_ident_from_mod(&modlinkage, &zfs_li); 6172 ASSERT(error == 0); 6173 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL); 6174 6175 return (0); 6176} 6177 6178int 6179_fini(void) 6180{ 6181 int error; 6182 6183 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled) 6184 return (SET_ERROR(EBUSY)); 6185 6186 if ((error = mod_remove(&modlinkage)) != 0) 6187 return (error); 6188 6189 zvol_fini(); 6190 zfs_fini(); 6191 spa_fini(); 6192 if (zfs_nfsshare_inited) 6193 (void) ddi_modclose(nfs_mod); 6194 if (zfs_smbshare_inited) 6195 (void) ddi_modclose(smbsrv_mod); 6196 if (zfs_nfsshare_inited || zfs_smbshare_inited) 6197 (void) ddi_modclose(sharefs_mod); 6198 6199 tsd_destroy(&zfs_fsyncer_key); 6200 ldi_ident_release(zfs_li); 6201 zfs_li = NULL; 6202 mutex_destroy(&zfs_share_lock); 6203 6204 return (error); 6205} 6206 6207int 6208_info(struct modinfo *modinfop) 6209{ 6210 return (mod_info(&modlinkage, modinfop)); 6211} 6212#endif /* sun */ 6213 6214static int zfs__init(void); 6215static int zfs__fini(void); 6216static void zfs_shutdown(void *, int); 6217 6218static eventhandler_tag zfs_shutdown_event_tag; 6219 6220int 6221zfs__init(void) 6222{ 6223 6224 zfs_root_token = root_mount_hold("ZFS"); 6225 6226 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL); 6227 6228 spa_init(FREAD | FWRITE); 6229 zfs_init(); 6230 zvol_init(); 6231 zfs_ioctl_init(); 6232 6233 tsd_create(&zfs_fsyncer_key, NULL); 6234 tsd_create(&rrw_tsd_key, rrw_tsd_destroy); 6235 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy); 6236 6237 printf("ZFS storage pool version: features support (" SPA_VERSION_STRING ")\n"); 6238 root_mount_rel(zfs_root_token); 6239 6240 zfsdev_init(); 6241 6242 return (0); 6243} 6244 6245int 6246zfs__fini(void) 6247{ 6248 if (spa_busy() || zfs_busy() || zvol_busy() || 6249 zio_injection_enabled) { 6250 return (EBUSY); 6251 } 6252 6253 zfsdev_fini(); 6254 zvol_fini(); 6255 zfs_fini(); 6256 spa_fini(); 6257 6258 tsd_destroy(&zfs_fsyncer_key); 6259 tsd_destroy(&rrw_tsd_key); 6260 tsd_destroy(&zfs_allow_log_key); 6261 6262 mutex_destroy(&zfs_share_lock); 6263 6264 return (0); 6265} 6266 6267static void 6268zfs_shutdown(void *arg __unused, int howto __unused) 6269{ 6270 6271 /* 6272 * ZFS fini routines can not properly work in a panic-ed system. 6273 */ 6274 if (panicstr == NULL) 6275 (void)zfs__fini(); 6276} 6277 6278 6279static int 6280zfs_modevent(module_t mod, int type, void *unused __unused) 6281{ 6282 int err; 6283 6284 switch (type) { 6285 case MOD_LOAD: 6286 err = zfs__init(); 6287 if (err == 0) 6288 zfs_shutdown_event_tag = EVENTHANDLER_REGISTER( 6289 shutdown_post_sync, zfs_shutdown, NULL, 6290 SHUTDOWN_PRI_FIRST); 6291 return (err); 6292 case MOD_UNLOAD: 6293 err = zfs__fini(); 6294 if (err == 0 && zfs_shutdown_event_tag != NULL) 6295 EVENTHANDLER_DEREGISTER(shutdown_post_sync, 6296 zfs_shutdown_event_tag); 6297 return (err); 6298 case MOD_SHUTDOWN: 6299 return (0); 6300 default: 6301 break; 6302 } 6303 return (EOPNOTSUPP); 6304} 6305 6306static moduledata_t zfs_mod = { 6307 "zfsctrl", 6308 zfs_modevent, 6309 0 6310}; 6311DECLARE_MODULE(zfsctrl, zfs_mod, SI_SUB_VFS, SI_ORDER_ANY); 6312MODULE_VERSION(zfsctrl, 1); 6313MODULE_DEPEND(zfsctrl, opensolaris, 1, 1, 1); 6314MODULE_DEPEND(zfsctrl, krpc, 1, 1, 1); 6315MODULE_DEPEND(zfsctrl, acl_nfs4, 1, 1, 1); 6316