zfs_ioctl.c revision 296528
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 2014 Xin Li <delphij@FreeBSD.org>. All rights reserved. 28 * Copyright 2015, OmniTI Computer Consulting, Inc. All rights reserved. 29 * Copyright 2015 Nexenta Systems, Inc. All rights reserved. 30 * Copyright (c) 2014, Joyent, Inc. All rights reserved. 31 * Copyright (c) 2011, 2015 by Delphix. All rights reserved. 32 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved. 33 * Copyright (c) 2013 Steven Hartland. All rights reserved. 34 * Copyright (c) 2014 Integros [integros.com] 35 */ 36 37/* 38 * ZFS ioctls. 39 * 40 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage 41 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool. 42 * 43 * There are two ways that we handle ioctls: the legacy way where almost 44 * all of the logic is in the ioctl callback, and the new way where most 45 * of the marshalling is handled in the common entry point, zfsdev_ioctl(). 46 * 47 * Non-legacy ioctls should be registered by calling 48 * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked 49 * from userland by lzc_ioctl(). 50 * 51 * The registration arguments are as follows: 52 * 53 * const char *name 54 * The name of the ioctl. This is used for history logging. If the 55 * ioctl returns successfully (the callback returns 0), and allow_log 56 * is true, then a history log entry will be recorded with the input & 57 * output nvlists. The log entry can be printed with "zpool history -i". 58 * 59 * zfs_ioc_t ioc 60 * The ioctl request number, which userland will pass to ioctl(2). 61 * The ioctl numbers can change from release to release, because 62 * the caller (libzfs) must be matched to the kernel. 63 * 64 * zfs_secpolicy_func_t *secpolicy 65 * This function will be called before the zfs_ioc_func_t, to 66 * determine if this operation is permitted. It should return EPERM 67 * on failure, and 0 on success. Checks include determining if the 68 * dataset is visible in this zone, and if the user has either all 69 * zfs privileges in the zone (SYS_MOUNT), or has been granted permission 70 * to do this operation on this dataset with "zfs allow". 71 * 72 * zfs_ioc_namecheck_t namecheck 73 * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool 74 * name, a dataset name, or nothing. If the name is not well-formed, 75 * the ioctl will fail and the callback will not be called. 76 * Therefore, the callback can assume that the name is well-formed 77 * (e.g. is null-terminated, doesn't have more than one '@' character, 78 * doesn't have invalid characters). 79 * 80 * zfs_ioc_poolcheck_t pool_check 81 * This specifies requirements on the pool state. If the pool does 82 * not meet them (is suspended or is readonly), the ioctl will fail 83 * and the callback will not be called. If any checks are specified 84 * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME. 85 * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED | 86 * POOL_CHECK_READONLY). 87 * 88 * boolean_t smush_outnvlist 89 * If smush_outnvlist is true, then the output is presumed to be a 90 * list of errors, and it will be "smushed" down to fit into the 91 * caller's buffer, by removing some entries and replacing them with a 92 * single "N_MORE_ERRORS" entry indicating how many were removed. See 93 * nvlist_smush() for details. If smush_outnvlist is false, and the 94 * outnvlist does not fit into the userland-provided buffer, then the 95 * ioctl will fail with ENOMEM. 96 * 97 * zfs_ioc_func_t *func 98 * The callback function that will perform the operation. 99 * 100 * The callback should return 0 on success, or an error number on 101 * failure. If the function fails, the userland ioctl will return -1, 102 * and errno will be set to the callback's return value. The callback 103 * will be called with the following arguments: 104 * 105 * const char *name 106 * The name of the pool or dataset to operate on, from 107 * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the 108 * expected type (pool, dataset, or none). 109 * 110 * nvlist_t *innvl 111 * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or 112 * NULL if no input nvlist was provided. Changes to this nvlist are 113 * ignored. If the input nvlist could not be deserialized, the 114 * ioctl will fail and the callback will not be called. 115 * 116 * nvlist_t *outnvl 117 * The output nvlist, initially empty. The callback can fill it in, 118 * and it will be returned to userland by serializing it into 119 * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization 120 * fails (e.g. because the caller didn't supply a large enough 121 * buffer), then the overall ioctl will fail. See the 122 * 'smush_nvlist' argument above for additional behaviors. 123 * 124 * There are two typical uses of the output nvlist: 125 * - To return state, e.g. property values. In this case, 126 * smush_outnvlist should be false. If the buffer was not large 127 * enough, the caller will reallocate a larger buffer and try 128 * the ioctl again. 129 * 130 * - To return multiple errors from an ioctl which makes on-disk 131 * changes. In this case, smush_outnvlist should be true. 132 * Ioctls which make on-disk modifications should generally not 133 * use the outnvl if they succeed, because the caller can not 134 * distinguish between the operation failing, and 135 * deserialization failing. 136 */ 137#ifdef __FreeBSD__ 138#include "opt_kstack_pages.h" 139#endif 140 141#include <sys/types.h> 142#include <sys/param.h> 143#include <sys/systm.h> 144#include <sys/conf.h> 145#include <sys/kernel.h> 146#include <sys/lock.h> 147#include <sys/malloc.h> 148#include <sys/mutex.h> 149#include <sys/proc.h> 150#include <sys/errno.h> 151#include <sys/uio.h> 152#include <sys/buf.h> 153#include <sys/file.h> 154#include <sys/kmem.h> 155#include <sys/conf.h> 156#include <sys/cmn_err.h> 157#include <sys/stat.h> 158#include <sys/zfs_ioctl.h> 159#include <sys/zfs_vfsops.h> 160#include <sys/zfs_znode.h> 161#include <sys/zap.h> 162#include <sys/spa.h> 163#include <sys/spa_impl.h> 164#include <sys/vdev.h> 165#include <sys/dmu.h> 166#include <sys/dsl_dir.h> 167#include <sys/dsl_dataset.h> 168#include <sys/dsl_prop.h> 169#include <sys/dsl_deleg.h> 170#include <sys/dmu_objset.h> 171#include <sys/dmu_impl.h> 172#include <sys/dmu_tx.h> 173#include <sys/sunddi.h> 174#include <sys/policy.h> 175#include <sys/zone.h> 176#include <sys/nvpair.h> 177#include <sys/mount.h> 178#include <sys/taskqueue.h> 179#include <sys/sdt.h> 180#include <sys/varargs.h> 181#include <sys/fs/zfs.h> 182#include <sys/zfs_ctldir.h> 183#include <sys/zfs_dir.h> 184#include <sys/zfs_onexit.h> 185#include <sys/zvol.h> 186#include <sys/dsl_scan.h> 187#include <sys/dmu_objset.h> 188#include <sys/dmu_send.h> 189#include <sys/dsl_destroy.h> 190#include <sys/dsl_bookmark.h> 191#include <sys/dsl_userhold.h> 192#include <sys/zfeature.h> 193#include <sys/zio_checksum.h> 194 195#include "zfs_namecheck.h" 196#include "zfs_prop.h" 197#include "zfs_deleg.h" 198#include "zfs_comutil.h" 199#include "zfs_ioctl_compat.h" 200 201CTASSERT(sizeof(zfs_cmd_t) < IOCPARM_MAX); 202 203static struct cdev *zfsdev; 204 205extern void zfs_init(void); 206extern void zfs_fini(void); 207 208uint_t zfs_fsyncer_key; 209extern uint_t rrw_tsd_key; 210static uint_t zfs_allow_log_key; 211extern uint_t zfs_geom_probe_vdev_key; 212 213typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *); 214typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *); 215typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *); 216 217typedef enum { 218 NO_NAME, 219 POOL_NAME, 220 DATASET_NAME 221} zfs_ioc_namecheck_t; 222 223typedef enum { 224 POOL_CHECK_NONE = 1 << 0, 225 POOL_CHECK_SUSPENDED = 1 << 1, 226 POOL_CHECK_READONLY = 1 << 2, 227} zfs_ioc_poolcheck_t; 228 229typedef struct zfs_ioc_vec { 230 zfs_ioc_legacy_func_t *zvec_legacy_func; 231 zfs_ioc_func_t *zvec_func; 232 zfs_secpolicy_func_t *zvec_secpolicy; 233 zfs_ioc_namecheck_t zvec_namecheck; 234 boolean_t zvec_allow_log; 235 zfs_ioc_poolcheck_t zvec_pool_check; 236 boolean_t zvec_smush_outnvlist; 237 const char *zvec_name; 238} zfs_ioc_vec_t; 239 240/* This array is indexed by zfs_userquota_prop_t */ 241static const char *userquota_perms[] = { 242 ZFS_DELEG_PERM_USERUSED, 243 ZFS_DELEG_PERM_USERQUOTA, 244 ZFS_DELEG_PERM_GROUPUSED, 245 ZFS_DELEG_PERM_GROUPQUOTA, 246}; 247 248static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc); 249static int zfs_check_settable(const char *name, nvpair_t *property, 250 cred_t *cr); 251static int zfs_check_clearable(char *dataset, nvlist_t *props, 252 nvlist_t **errors); 253static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *, 254 boolean_t *); 255int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *); 256static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp); 257 258static void zfsdev_close(void *data); 259 260static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature); 261 262/* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */ 263void 264__dprintf(const char *file, const char *func, int line, const char *fmt, ...) 265{ 266 const char *newfile; 267 char buf[512]; 268 va_list adx; 269 270 /* 271 * Get rid of annoying "../common/" prefix to filename. 272 */ 273 newfile = strrchr(file, '/'); 274 if (newfile != NULL) { 275 newfile = newfile + 1; /* Get rid of leading / */ 276 } else { 277 newfile = file; 278 } 279 280 va_start(adx, fmt); 281 (void) vsnprintf(buf, sizeof (buf), fmt, adx); 282 va_end(adx); 283 284 /* 285 * To get this data, use the zfs-dprintf probe as so: 286 * dtrace -q -n 'zfs-dprintf \ 287 * /stringof(arg0) == "dbuf.c"/ \ 288 * {printf("%s: %s", stringof(arg1), stringof(arg3))}' 289 * arg0 = file name 290 * arg1 = function name 291 * arg2 = line number 292 * arg3 = message 293 */ 294 DTRACE_PROBE4(zfs__dprintf, 295 char *, newfile, char *, func, int, line, char *, buf); 296} 297 298static void 299history_str_free(char *buf) 300{ 301 kmem_free(buf, HIS_MAX_RECORD_LEN); 302} 303 304static char * 305history_str_get(zfs_cmd_t *zc) 306{ 307 char *buf; 308 309 if (zc->zc_history == 0) 310 return (NULL); 311 312 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP); 313 if (copyinstr((void *)(uintptr_t)zc->zc_history, 314 buf, HIS_MAX_RECORD_LEN, NULL) != 0) { 315 history_str_free(buf); 316 return (NULL); 317 } 318 319 buf[HIS_MAX_RECORD_LEN -1] = '\0'; 320 321 return (buf); 322} 323 324/* 325 * Check to see if the named dataset is currently defined as bootable 326 */ 327static boolean_t 328zfs_is_bootfs(const char *name) 329{ 330 objset_t *os; 331 332 if (dmu_objset_hold(name, FTAG, &os) == 0) { 333 boolean_t ret; 334 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os))); 335 dmu_objset_rele(os, FTAG); 336 return (ret); 337 } 338 return (B_FALSE); 339} 340 341/* 342 * Return non-zero if the spa version is less than requested version. 343 */ 344static int 345zfs_earlier_version(const char *name, int version) 346{ 347 spa_t *spa; 348 349 if (spa_open(name, &spa, FTAG) == 0) { 350 if (spa_version(spa) < version) { 351 spa_close(spa, FTAG); 352 return (1); 353 } 354 spa_close(spa, FTAG); 355 } 356 return (0); 357} 358 359/* 360 * Return TRUE if the ZPL version is less than requested version. 361 */ 362static boolean_t 363zpl_earlier_version(const char *name, int version) 364{ 365 objset_t *os; 366 boolean_t rc = B_TRUE; 367 368 if (dmu_objset_hold(name, FTAG, &os) == 0) { 369 uint64_t zplversion; 370 371 if (dmu_objset_type(os) != DMU_OST_ZFS) { 372 dmu_objset_rele(os, FTAG); 373 return (B_TRUE); 374 } 375 /* XXX reading from non-owned objset */ 376 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0) 377 rc = zplversion < version; 378 dmu_objset_rele(os, FTAG); 379 } 380 return (rc); 381} 382 383static void 384zfs_log_history(zfs_cmd_t *zc) 385{ 386 spa_t *spa; 387 char *buf; 388 389 if ((buf = history_str_get(zc)) == NULL) 390 return; 391 392 if (spa_open(zc->zc_name, &spa, FTAG) == 0) { 393 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY) 394 (void) spa_history_log(spa, buf); 395 spa_close(spa, FTAG); 396 } 397 history_str_free(buf); 398} 399 400/* 401 * Policy for top-level read operations (list pools). Requires no privileges, 402 * and can be used in the local zone, as there is no associated dataset. 403 */ 404/* ARGSUSED */ 405static int 406zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 407{ 408 return (0); 409} 410 411/* 412 * Policy for dataset read operations (list children, get statistics). Requires 413 * no privileges, but must be visible in the local zone. 414 */ 415/* ARGSUSED */ 416static int 417zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 418{ 419 if (INGLOBALZONE(curthread) || 420 zone_dataset_visible(zc->zc_name, NULL)) 421 return (0); 422 423 return (SET_ERROR(ENOENT)); 424} 425 426static int 427zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr) 428{ 429 int writable = 1; 430 431 /* 432 * The dataset must be visible by this zone -- check this first 433 * so they don't see EPERM on something they shouldn't know about. 434 */ 435 if (!INGLOBALZONE(curthread) && 436 !zone_dataset_visible(dataset, &writable)) 437 return (SET_ERROR(ENOENT)); 438 439 if (INGLOBALZONE(curthread)) { 440 /* 441 * If the fs is zoned, only root can access it from the 442 * global zone. 443 */ 444 if (secpolicy_zfs(cr) && zoned) 445 return (SET_ERROR(EPERM)); 446 } else { 447 /* 448 * If we are in a local zone, the 'zoned' property must be set. 449 */ 450 if (!zoned) 451 return (SET_ERROR(EPERM)); 452 453 /* must be writable by this zone */ 454 if (!writable) 455 return (SET_ERROR(EPERM)); 456 } 457 return (0); 458} 459 460static int 461zfs_dozonecheck(const char *dataset, cred_t *cr) 462{ 463 uint64_t zoned; 464 465 if (dsl_prop_get_integer(dataset, "jailed", &zoned, NULL)) 466 return (SET_ERROR(ENOENT)); 467 468 return (zfs_dozonecheck_impl(dataset, zoned, cr)); 469} 470 471static int 472zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr) 473{ 474 uint64_t zoned; 475 476 if (dsl_prop_get_int_ds(ds, "jailed", &zoned)) 477 return (SET_ERROR(ENOENT)); 478 479 return (zfs_dozonecheck_impl(dataset, zoned, cr)); 480} 481 482static int 483zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds, 484 const char *perm, cred_t *cr) 485{ 486 int error; 487 488 error = zfs_dozonecheck_ds(name, ds, cr); 489 if (error == 0) { 490 error = secpolicy_zfs(cr); 491 if (error != 0) 492 error = dsl_deleg_access_impl(ds, perm, cr); 493 } 494 return (error); 495} 496 497static int 498zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr) 499{ 500 int error; 501 dsl_dataset_t *ds; 502 dsl_pool_t *dp; 503 504 error = dsl_pool_hold(name, FTAG, &dp); 505 if (error != 0) 506 return (error); 507 508 error = dsl_dataset_hold(dp, name, FTAG, &ds); 509 if (error != 0) { 510 dsl_pool_rele(dp, FTAG); 511 return (error); 512 } 513 514 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr); 515 516 dsl_dataset_rele(ds, FTAG); 517 dsl_pool_rele(dp, FTAG); 518 return (error); 519} 520 521#ifdef SECLABEL 522/* 523 * Policy for setting the security label property. 524 * 525 * Returns 0 for success, non-zero for access and other errors. 526 */ 527static int 528zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr) 529{ 530 char ds_hexsl[MAXNAMELEN]; 531 bslabel_t ds_sl, new_sl; 532 boolean_t new_default = FALSE; 533 uint64_t zoned; 534 int needed_priv = -1; 535 int error; 536 537 /* First get the existing dataset label. */ 538 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL), 539 1, sizeof (ds_hexsl), &ds_hexsl, NULL); 540 if (error != 0) 541 return (SET_ERROR(EPERM)); 542 543 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0) 544 new_default = TRUE; 545 546 /* The label must be translatable */ 547 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0)) 548 return (SET_ERROR(EINVAL)); 549 550 /* 551 * In a non-global zone, disallow attempts to set a label that 552 * doesn't match that of the zone; otherwise no other checks 553 * are needed. 554 */ 555 if (!INGLOBALZONE(curproc)) { 556 if (new_default || !blequal(&new_sl, CR_SL(CRED()))) 557 return (SET_ERROR(EPERM)); 558 return (0); 559 } 560 561 /* 562 * For global-zone datasets (i.e., those whose zoned property is 563 * "off", verify that the specified new label is valid for the 564 * global zone. 565 */ 566 if (dsl_prop_get_integer(name, 567 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL)) 568 return (SET_ERROR(EPERM)); 569 if (!zoned) { 570 if (zfs_check_global_label(name, strval) != 0) 571 return (SET_ERROR(EPERM)); 572 } 573 574 /* 575 * If the existing dataset label is nondefault, check if the 576 * dataset is mounted (label cannot be changed while mounted). 577 * Get the zfsvfs; if there isn't one, then the dataset isn't 578 * mounted (or isn't a dataset, doesn't exist, ...). 579 */ 580 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) { 581 objset_t *os; 582 static char *setsl_tag = "setsl_tag"; 583 584 /* 585 * Try to own the dataset; abort if there is any error, 586 * (e.g., already mounted, in use, or other error). 587 */ 588 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE, 589 setsl_tag, &os); 590 if (error != 0) 591 return (SET_ERROR(EPERM)); 592 593 dmu_objset_disown(os, setsl_tag); 594 595 if (new_default) { 596 needed_priv = PRIV_FILE_DOWNGRADE_SL; 597 goto out_check; 598 } 599 600 if (hexstr_to_label(strval, &new_sl) != 0) 601 return (SET_ERROR(EPERM)); 602 603 if (blstrictdom(&ds_sl, &new_sl)) 604 needed_priv = PRIV_FILE_DOWNGRADE_SL; 605 else if (blstrictdom(&new_sl, &ds_sl)) 606 needed_priv = PRIV_FILE_UPGRADE_SL; 607 } else { 608 /* dataset currently has a default label */ 609 if (!new_default) 610 needed_priv = PRIV_FILE_UPGRADE_SL; 611 } 612 613out_check: 614 if (needed_priv != -1) 615 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL)); 616 return (0); 617} 618#endif /* SECLABEL */ 619 620static int 621zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval, 622 cred_t *cr) 623{ 624 char *strval; 625 626 /* 627 * Check permissions for special properties. 628 */ 629 switch (prop) { 630 case ZFS_PROP_ZONED: 631 /* 632 * Disallow setting of 'zoned' from within a local zone. 633 */ 634 if (!INGLOBALZONE(curthread)) 635 return (SET_ERROR(EPERM)); 636 break; 637 638 case ZFS_PROP_QUOTA: 639 case ZFS_PROP_FILESYSTEM_LIMIT: 640 case ZFS_PROP_SNAPSHOT_LIMIT: 641 if (!INGLOBALZONE(curthread)) { 642 uint64_t zoned; 643 char setpoint[MAXNAMELEN]; 644 /* 645 * Unprivileged users are allowed to modify the 646 * limit on things *under* (ie. contained by) 647 * the thing they own. 648 */ 649 if (dsl_prop_get_integer(dsname, "jailed", &zoned, 650 setpoint)) 651 return (SET_ERROR(EPERM)); 652 if (!zoned || strlen(dsname) <= strlen(setpoint)) 653 return (SET_ERROR(EPERM)); 654 } 655 break; 656 657 case ZFS_PROP_MLSLABEL: 658#ifdef SECLABEL 659 if (!is_system_labeled()) 660 return (SET_ERROR(EPERM)); 661 662 if (nvpair_value_string(propval, &strval) == 0) { 663 int err; 664 665 err = zfs_set_slabel_policy(dsname, strval, CRED()); 666 if (err != 0) 667 return (err); 668 } 669#else 670 return (EOPNOTSUPP); 671#endif 672 break; 673 } 674 675 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr)); 676} 677 678/* ARGSUSED */ 679static int 680zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 681{ 682 int error; 683 684 error = zfs_dozonecheck(zc->zc_name, cr); 685 if (error != 0) 686 return (error); 687 688 /* 689 * permission to set permissions will be evaluated later in 690 * dsl_deleg_can_allow() 691 */ 692 return (0); 693} 694 695/* ARGSUSED */ 696static int 697zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 698{ 699 return (zfs_secpolicy_write_perms(zc->zc_name, 700 ZFS_DELEG_PERM_ROLLBACK, cr)); 701} 702 703/* ARGSUSED */ 704static int 705zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 706{ 707 dsl_pool_t *dp; 708 dsl_dataset_t *ds; 709 char *cp; 710 int error; 711 712 /* 713 * Generate the current snapshot name from the given objsetid, then 714 * use that name for the secpolicy/zone checks. 715 */ 716 cp = strchr(zc->zc_name, '@'); 717 if (cp == NULL) 718 return (SET_ERROR(EINVAL)); 719 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 720 if (error != 0) 721 return (error); 722 723 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds); 724 if (error != 0) { 725 dsl_pool_rele(dp, FTAG); 726 return (error); 727 } 728 729 dsl_dataset_name(ds, zc->zc_name); 730 731 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds, 732 ZFS_DELEG_PERM_SEND, cr); 733 dsl_dataset_rele(ds, FTAG); 734 dsl_pool_rele(dp, FTAG); 735 736 return (error); 737} 738 739/* ARGSUSED */ 740static int 741zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 742{ 743 return (zfs_secpolicy_write_perms(zc->zc_name, 744 ZFS_DELEG_PERM_SEND, cr)); 745} 746 747/* ARGSUSED */ 748static int 749zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 750{ 751 vnode_t *vp; 752 int error; 753 754 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, 755 NO_FOLLOW, NULL, &vp)) != 0) 756 return (error); 757 758 /* Now make sure mntpnt and dataset are ZFS */ 759 760 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 || 761 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), 762 zc->zc_name) != 0)) { 763 VN_RELE(vp); 764 return (SET_ERROR(EPERM)); 765 } 766 767 VN_RELE(vp); 768 return (dsl_deleg_access(zc->zc_name, 769 ZFS_DELEG_PERM_SHARE, cr)); 770} 771 772int 773zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 774{ 775 if (!INGLOBALZONE(curthread)) 776 return (SET_ERROR(EPERM)); 777 778 if (secpolicy_nfs(cr) == 0) { 779 return (0); 780 } else { 781 return (zfs_secpolicy_deleg_share(zc, innvl, cr)); 782 } 783} 784 785int 786zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 787{ 788 if (!INGLOBALZONE(curthread)) 789 return (SET_ERROR(EPERM)); 790 791 if (secpolicy_smb(cr) == 0) { 792 return (0); 793 } else { 794 return (zfs_secpolicy_deleg_share(zc, innvl, cr)); 795 } 796} 797 798static int 799zfs_get_parent(const char *datasetname, char *parent, int parentsize) 800{ 801 char *cp; 802 803 /* 804 * Remove the @bla or /bla from the end of the name to get the parent. 805 */ 806 (void) strncpy(parent, datasetname, parentsize); 807 cp = strrchr(parent, '@'); 808 if (cp != NULL) { 809 cp[0] = '\0'; 810 } else { 811 cp = strrchr(parent, '/'); 812 if (cp == NULL) 813 return (SET_ERROR(ENOENT)); 814 cp[0] = '\0'; 815 } 816 817 return (0); 818} 819 820int 821zfs_secpolicy_destroy_perms(const char *name, cred_t *cr) 822{ 823 int error; 824 825 if ((error = zfs_secpolicy_write_perms(name, 826 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 827 return (error); 828 829 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr)); 830} 831 832/* ARGSUSED */ 833static int 834zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 835{ 836 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr)); 837} 838 839/* 840 * Destroying snapshots with delegated permissions requires 841 * descendant mount and destroy permissions. 842 */ 843/* ARGSUSED */ 844static int 845zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 846{ 847 nvlist_t *snaps; 848 nvpair_t *pair, *nextpair; 849 int error = 0; 850 851 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 852 return (SET_ERROR(EINVAL)); 853 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 854 pair = nextpair) { 855 nextpair = nvlist_next_nvpair(snaps, pair); 856 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr); 857 if (error == ENOENT) { 858 /* 859 * Ignore any snapshots that don't exist (we consider 860 * them "already destroyed"). Remove the name from the 861 * nvl here in case the snapshot is created between 862 * now and when we try to destroy it (in which case 863 * we don't want to destroy it since we haven't 864 * checked for permission). 865 */ 866 fnvlist_remove_nvpair(snaps, pair); 867 error = 0; 868 } 869 if (error != 0) 870 break; 871 } 872 873 return (error); 874} 875 876int 877zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr) 878{ 879 char parentname[MAXNAMELEN]; 880 int error; 881 882 if ((error = zfs_secpolicy_write_perms(from, 883 ZFS_DELEG_PERM_RENAME, cr)) != 0) 884 return (error); 885 886 if ((error = zfs_secpolicy_write_perms(from, 887 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 888 return (error); 889 890 if ((error = zfs_get_parent(to, parentname, 891 sizeof (parentname))) != 0) 892 return (error); 893 894 if ((error = zfs_secpolicy_write_perms(parentname, 895 ZFS_DELEG_PERM_CREATE, cr)) != 0) 896 return (error); 897 898 if ((error = zfs_secpolicy_write_perms(parentname, 899 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 900 return (error); 901 902 return (error); 903} 904 905/* ARGSUSED */ 906static int 907zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 908{ 909 char *at = NULL; 910 int error; 911 912 if ((zc->zc_cookie & 1) != 0) { 913 /* 914 * This is recursive rename, so the starting snapshot might 915 * not exist. Check file system or volume permission instead. 916 */ 917 at = strchr(zc->zc_name, '@'); 918 if (at == NULL) 919 return (EINVAL); 920 *at = '\0'; 921 } 922 923 error = zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr); 924 925 if (at != NULL) 926 *at = '@'; 927 928 return (error); 929} 930 931/* ARGSUSED */ 932static int 933zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 934{ 935 dsl_pool_t *dp; 936 dsl_dataset_t *clone; 937 int error; 938 939 error = zfs_secpolicy_write_perms(zc->zc_name, 940 ZFS_DELEG_PERM_PROMOTE, cr); 941 if (error != 0) 942 return (error); 943 944 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 945 if (error != 0) 946 return (error); 947 948 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone); 949 950 if (error == 0) { 951 char parentname[MAXNAMELEN]; 952 dsl_dataset_t *origin = NULL; 953 dsl_dir_t *dd; 954 dd = clone->ds_dir; 955 956 error = dsl_dataset_hold_obj(dd->dd_pool, 957 dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin); 958 if (error != 0) { 959 dsl_dataset_rele(clone, FTAG); 960 dsl_pool_rele(dp, FTAG); 961 return (error); 962 } 963 964 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone, 965 ZFS_DELEG_PERM_MOUNT, cr); 966 967 dsl_dataset_name(origin, parentname); 968 if (error == 0) { 969 error = zfs_secpolicy_write_perms_ds(parentname, origin, 970 ZFS_DELEG_PERM_PROMOTE, cr); 971 } 972 dsl_dataset_rele(clone, FTAG); 973 dsl_dataset_rele(origin, FTAG); 974 } 975 dsl_pool_rele(dp, FTAG); 976 return (error); 977} 978 979/* ARGSUSED */ 980static int 981zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 982{ 983 int error; 984 985 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 986 ZFS_DELEG_PERM_RECEIVE, cr)) != 0) 987 return (error); 988 989 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 990 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 991 return (error); 992 993 return (zfs_secpolicy_write_perms(zc->zc_name, 994 ZFS_DELEG_PERM_CREATE, cr)); 995} 996 997int 998zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr) 999{ 1000 return (zfs_secpolicy_write_perms(name, 1001 ZFS_DELEG_PERM_SNAPSHOT, cr)); 1002} 1003 1004/* 1005 * Check for permission to create each snapshot in the nvlist. 1006 */ 1007/* ARGSUSED */ 1008static int 1009zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1010{ 1011 nvlist_t *snaps; 1012 int error; 1013 nvpair_t *pair; 1014 1015 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 1016 return (SET_ERROR(EINVAL)); 1017 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 1018 pair = nvlist_next_nvpair(snaps, pair)) { 1019 char *name = nvpair_name(pair); 1020 char *atp = strchr(name, '@'); 1021 1022 if (atp == NULL) { 1023 error = SET_ERROR(EINVAL); 1024 break; 1025 } 1026 *atp = '\0'; 1027 error = zfs_secpolicy_snapshot_perms(name, cr); 1028 *atp = '@'; 1029 if (error != 0) 1030 break; 1031 } 1032 return (error); 1033} 1034 1035/* 1036 * Check for permission to create each snapshot in the nvlist. 1037 */ 1038/* ARGSUSED */ 1039static int 1040zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1041{ 1042 int error = 0; 1043 1044 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 1045 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 1046 char *name = nvpair_name(pair); 1047 char *hashp = strchr(name, '#'); 1048 1049 if (hashp == NULL) { 1050 error = SET_ERROR(EINVAL); 1051 break; 1052 } 1053 *hashp = '\0'; 1054 error = zfs_secpolicy_write_perms(name, 1055 ZFS_DELEG_PERM_BOOKMARK, cr); 1056 *hashp = '#'; 1057 if (error != 0) 1058 break; 1059 } 1060 return (error); 1061} 1062 1063/* ARGSUSED */ 1064static int 1065zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1066{ 1067 nvpair_t *pair, *nextpair; 1068 int error = 0; 1069 1070 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL; 1071 pair = nextpair) { 1072 char *name = nvpair_name(pair); 1073 char *hashp = strchr(name, '#'); 1074 nextpair = nvlist_next_nvpair(innvl, pair); 1075 1076 if (hashp == NULL) { 1077 error = SET_ERROR(EINVAL); 1078 break; 1079 } 1080 1081 *hashp = '\0'; 1082 error = zfs_secpolicy_write_perms(name, 1083 ZFS_DELEG_PERM_DESTROY, cr); 1084 *hashp = '#'; 1085 if (error == ENOENT) { 1086 /* 1087 * Ignore any filesystems that don't exist (we consider 1088 * their bookmarks "already destroyed"). Remove 1089 * the name from the nvl here in case the filesystem 1090 * is created between now and when we try to destroy 1091 * the bookmark (in which case we don't want to 1092 * destroy it since we haven't checked for permission). 1093 */ 1094 fnvlist_remove_nvpair(innvl, pair); 1095 error = 0; 1096 } 1097 if (error != 0) 1098 break; 1099 } 1100 1101 return (error); 1102} 1103 1104/* ARGSUSED */ 1105static int 1106zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1107{ 1108 /* 1109 * Even root must have a proper TSD so that we know what pool 1110 * to log to. 1111 */ 1112 if (tsd_get(zfs_allow_log_key) == NULL) 1113 return (SET_ERROR(EPERM)); 1114 return (0); 1115} 1116 1117static int 1118zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1119{ 1120 char parentname[MAXNAMELEN]; 1121 int error; 1122 char *origin; 1123 1124 if ((error = zfs_get_parent(zc->zc_name, parentname, 1125 sizeof (parentname))) != 0) 1126 return (error); 1127 1128 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 && 1129 (error = zfs_secpolicy_write_perms(origin, 1130 ZFS_DELEG_PERM_CLONE, cr)) != 0) 1131 return (error); 1132 1133 if ((error = zfs_secpolicy_write_perms(parentname, 1134 ZFS_DELEG_PERM_CREATE, cr)) != 0) 1135 return (error); 1136 1137 return (zfs_secpolicy_write_perms(parentname, 1138 ZFS_DELEG_PERM_MOUNT, cr)); 1139} 1140 1141/* 1142 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires 1143 * SYS_CONFIG privilege, which is not available in a local zone. 1144 */ 1145/* ARGSUSED */ 1146static int 1147zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1148{ 1149 if (secpolicy_sys_config(cr, B_FALSE) != 0) 1150 return (SET_ERROR(EPERM)); 1151 1152 return (0); 1153} 1154 1155/* 1156 * Policy for object to name lookups. 1157 */ 1158/* ARGSUSED */ 1159static int 1160zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1161{ 1162 int error; 1163 1164 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0) 1165 return (0); 1166 1167 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr); 1168 return (error); 1169} 1170 1171/* 1172 * Policy for fault injection. Requires all privileges. 1173 */ 1174/* ARGSUSED */ 1175static int 1176zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1177{ 1178 return (secpolicy_zinject(cr)); 1179} 1180 1181/* ARGSUSED */ 1182static int 1183zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1184{ 1185 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value); 1186 1187 if (prop == ZPROP_INVAL) { 1188 if (!zfs_prop_user(zc->zc_value)) 1189 return (SET_ERROR(EINVAL)); 1190 return (zfs_secpolicy_write_perms(zc->zc_name, 1191 ZFS_DELEG_PERM_USERPROP, cr)); 1192 } else { 1193 return (zfs_secpolicy_setprop(zc->zc_name, prop, 1194 NULL, cr)); 1195 } 1196} 1197 1198static int 1199zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1200{ 1201 int err = zfs_secpolicy_read(zc, innvl, cr); 1202 if (err) 1203 return (err); 1204 1205 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 1206 return (SET_ERROR(EINVAL)); 1207 1208 if (zc->zc_value[0] == 0) { 1209 /* 1210 * They are asking about a posix uid/gid. If it's 1211 * themself, allow it. 1212 */ 1213 if (zc->zc_objset_type == ZFS_PROP_USERUSED || 1214 zc->zc_objset_type == ZFS_PROP_USERQUOTA) { 1215 if (zc->zc_guid == crgetuid(cr)) 1216 return (0); 1217 } else { 1218 if (groupmember(zc->zc_guid, cr)) 1219 return (0); 1220 } 1221 } 1222 1223 return (zfs_secpolicy_write_perms(zc->zc_name, 1224 userquota_perms[zc->zc_objset_type], cr)); 1225} 1226 1227static int 1228zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1229{ 1230 int err = zfs_secpolicy_read(zc, innvl, cr); 1231 if (err) 1232 return (err); 1233 1234 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 1235 return (SET_ERROR(EINVAL)); 1236 1237 return (zfs_secpolicy_write_perms(zc->zc_name, 1238 userquota_perms[zc->zc_objset_type], cr)); 1239} 1240 1241/* ARGSUSED */ 1242static int 1243zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1244{ 1245 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION, 1246 NULL, cr)); 1247} 1248 1249/* ARGSUSED */ 1250static int 1251zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1252{ 1253 nvpair_t *pair; 1254 nvlist_t *holds; 1255 int error; 1256 1257 error = nvlist_lookup_nvlist(innvl, "holds", &holds); 1258 if (error != 0) 1259 return (SET_ERROR(EINVAL)); 1260 1261 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL; 1262 pair = nvlist_next_nvpair(holds, pair)) { 1263 char fsname[MAXNAMELEN]; 1264 error = dmu_fsname(nvpair_name(pair), fsname); 1265 if (error != 0) 1266 return (error); 1267 error = zfs_secpolicy_write_perms(fsname, 1268 ZFS_DELEG_PERM_HOLD, cr); 1269 if (error != 0) 1270 return (error); 1271 } 1272 return (0); 1273} 1274 1275/* ARGSUSED */ 1276static int 1277zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1278{ 1279 nvpair_t *pair; 1280 int error; 1281 1282 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL; 1283 pair = nvlist_next_nvpair(innvl, pair)) { 1284 char fsname[MAXNAMELEN]; 1285 error = dmu_fsname(nvpair_name(pair), fsname); 1286 if (error != 0) 1287 return (error); 1288 error = zfs_secpolicy_write_perms(fsname, 1289 ZFS_DELEG_PERM_RELEASE, cr); 1290 if (error != 0) 1291 return (error); 1292 } 1293 return (0); 1294} 1295 1296/* 1297 * Policy for allowing temporary snapshots to be taken or released 1298 */ 1299static int 1300zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1301{ 1302 /* 1303 * A temporary snapshot is the same as a snapshot, 1304 * hold, destroy and release all rolled into one. 1305 * Delegated diff alone is sufficient that we allow this. 1306 */ 1307 int error; 1308 1309 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 1310 ZFS_DELEG_PERM_DIFF, cr)) == 0) 1311 return (0); 1312 1313 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr); 1314 if (error == 0) 1315 error = zfs_secpolicy_hold(zc, innvl, cr); 1316 if (error == 0) 1317 error = zfs_secpolicy_release(zc, innvl, cr); 1318 if (error == 0) 1319 error = zfs_secpolicy_destroy(zc, innvl, cr); 1320 return (error); 1321} 1322 1323/* 1324 * Returns the nvlist as specified by the user in the zfs_cmd_t. 1325 */ 1326static int 1327get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp) 1328{ 1329 char *packed; 1330 int error; 1331 nvlist_t *list = NULL; 1332 1333 /* 1334 * Read in and unpack the user-supplied nvlist. 1335 */ 1336 if (size == 0) 1337 return (SET_ERROR(EINVAL)); 1338 1339 packed = kmem_alloc(size, KM_SLEEP); 1340 1341 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size, 1342 iflag)) != 0) { 1343 kmem_free(packed, size); 1344 return (SET_ERROR(EFAULT)); 1345 } 1346 1347 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) { 1348 kmem_free(packed, size); 1349 return (error); 1350 } 1351 1352 kmem_free(packed, size); 1353 1354 *nvp = list; 1355 return (0); 1356} 1357 1358/* 1359 * Reduce the size of this nvlist until it can be serialized in 'max' bytes. 1360 * Entries will be removed from the end of the nvlist, and one int32 entry 1361 * named "N_MORE_ERRORS" will be added indicating how many entries were 1362 * removed. 1363 */ 1364static int 1365nvlist_smush(nvlist_t *errors, size_t max) 1366{ 1367 size_t size; 1368 1369 size = fnvlist_size(errors); 1370 1371 if (size > max) { 1372 nvpair_t *more_errors; 1373 int n = 0; 1374 1375 if (max < 1024) 1376 return (SET_ERROR(ENOMEM)); 1377 1378 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0); 1379 more_errors = nvlist_prev_nvpair(errors, NULL); 1380 1381 do { 1382 nvpair_t *pair = nvlist_prev_nvpair(errors, 1383 more_errors); 1384 fnvlist_remove_nvpair(errors, pair); 1385 n++; 1386 size = fnvlist_size(errors); 1387 } while (size > max); 1388 1389 fnvlist_remove_nvpair(errors, more_errors); 1390 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n); 1391 ASSERT3U(fnvlist_size(errors), <=, max); 1392 } 1393 1394 return (0); 1395} 1396 1397static int 1398put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl) 1399{ 1400 char *packed = NULL; 1401 int error = 0; 1402 size_t size; 1403 1404 size = fnvlist_size(nvl); 1405 1406 if (size > zc->zc_nvlist_dst_size) { 1407 /* 1408 * Solaris returns ENOMEM here, because even if an error is 1409 * returned from an ioctl(2), new zc_nvlist_dst_size will be 1410 * passed to the userland. This is not the case for FreeBSD. 1411 * We need to return 0, so the kernel will copy the 1412 * zc_nvlist_dst_size back and the userland can discover that a 1413 * bigger buffer is needed. 1414 */ 1415 error = 0; 1416 } else { 1417 packed = fnvlist_pack(nvl, &size); 1418 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst, 1419 size, zc->zc_iflags) != 0) 1420 error = SET_ERROR(EFAULT); 1421 fnvlist_pack_free(packed, size); 1422 } 1423 1424 zc->zc_nvlist_dst_size = size; 1425 zc->zc_nvlist_dst_filled = B_TRUE; 1426 return (error); 1427} 1428 1429static int 1430getzfsvfs(const char *dsname, zfsvfs_t **zfvp) 1431{ 1432 objset_t *os; 1433 int error; 1434 1435 error = dmu_objset_hold(dsname, FTAG, &os); 1436 if (error != 0) 1437 return (error); 1438 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1439 dmu_objset_rele(os, FTAG); 1440 return (SET_ERROR(EINVAL)); 1441 } 1442 1443 mutex_enter(&os->os_user_ptr_lock); 1444 *zfvp = dmu_objset_get_user(os); 1445 if (*zfvp) { 1446 VFS_HOLD((*zfvp)->z_vfs); 1447 } else { 1448 error = SET_ERROR(ESRCH); 1449 } 1450 mutex_exit(&os->os_user_ptr_lock); 1451 dmu_objset_rele(os, FTAG); 1452 return (error); 1453} 1454 1455/* 1456 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which 1457 * case its z_vfs will be NULL, and it will be opened as the owner. 1458 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER, 1459 * which prevents all vnode ops from running. 1460 */ 1461static int 1462zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer) 1463{ 1464 int error = 0; 1465 1466 if (getzfsvfs(name, zfvp) != 0) 1467 error = zfsvfs_create(name, zfvp); 1468 if (error == 0) { 1469 rrm_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER : 1470 RW_READER, tag); 1471 if ((*zfvp)->z_unmounted) { 1472 /* 1473 * XXX we could probably try again, since the unmounting 1474 * thread should be just about to disassociate the 1475 * objset from the zfsvfs. 1476 */ 1477 rrm_exit(&(*zfvp)->z_teardown_lock, tag); 1478 return (SET_ERROR(EBUSY)); 1479 } 1480 } 1481 return (error); 1482} 1483 1484static void 1485zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag) 1486{ 1487 rrm_exit(&zfsvfs->z_teardown_lock, tag); 1488 1489 if (zfsvfs->z_vfs) { 1490 VFS_RELE(zfsvfs->z_vfs); 1491 } else { 1492 dmu_objset_disown(zfsvfs->z_os, zfsvfs); 1493 zfsvfs_free(zfsvfs); 1494 } 1495} 1496 1497static int 1498zfs_ioc_pool_create(zfs_cmd_t *zc) 1499{ 1500 int error; 1501 nvlist_t *config, *props = NULL; 1502 nvlist_t *rootprops = NULL; 1503 nvlist_t *zplprops = NULL; 1504 1505 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1506 zc->zc_iflags, &config)) 1507 return (error); 1508 1509 if (zc->zc_nvlist_src_size != 0 && (error = 1510 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1511 zc->zc_iflags, &props))) { 1512 nvlist_free(config); 1513 return (error); 1514 } 1515 1516 if (props) { 1517 nvlist_t *nvl = NULL; 1518 uint64_t version = SPA_VERSION; 1519 1520 (void) nvlist_lookup_uint64(props, 1521 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version); 1522 if (!SPA_VERSION_IS_SUPPORTED(version)) { 1523 error = SET_ERROR(EINVAL); 1524 goto pool_props_bad; 1525 } 1526 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl); 1527 if (nvl) { 1528 error = nvlist_dup(nvl, &rootprops, KM_SLEEP); 1529 if (error != 0) { 1530 nvlist_free(config); 1531 nvlist_free(props); 1532 return (error); 1533 } 1534 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS); 1535 } 1536 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 1537 error = zfs_fill_zplprops_root(version, rootprops, 1538 zplprops, NULL); 1539 if (error != 0) 1540 goto pool_props_bad; 1541 } 1542 1543 error = spa_create(zc->zc_name, config, props, zplprops); 1544 1545 /* 1546 * Set the remaining root properties 1547 */ 1548 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name, 1549 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0) 1550 (void) spa_destroy(zc->zc_name); 1551 1552pool_props_bad: 1553 nvlist_free(rootprops); 1554 nvlist_free(zplprops); 1555 nvlist_free(config); 1556 nvlist_free(props); 1557 1558 return (error); 1559} 1560 1561static int 1562zfs_ioc_pool_destroy(zfs_cmd_t *zc) 1563{ 1564 int error; 1565 zfs_log_history(zc); 1566 error = spa_destroy(zc->zc_name); 1567 if (error == 0) 1568 zvol_remove_minors(zc->zc_name); 1569 return (error); 1570} 1571 1572static int 1573zfs_ioc_pool_import(zfs_cmd_t *zc) 1574{ 1575 nvlist_t *config, *props = NULL; 1576 uint64_t guid; 1577 int error; 1578 1579 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1580 zc->zc_iflags, &config)) != 0) 1581 return (error); 1582 1583 if (zc->zc_nvlist_src_size != 0 && (error = 1584 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1585 zc->zc_iflags, &props))) { 1586 nvlist_free(config); 1587 return (error); 1588 } 1589 1590 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 || 1591 guid != zc->zc_guid) 1592 error = SET_ERROR(EINVAL); 1593 else 1594 error = spa_import(zc->zc_name, config, props, zc->zc_cookie); 1595 1596 if (zc->zc_nvlist_dst != 0) { 1597 int err; 1598 1599 if ((err = put_nvlist(zc, config)) != 0) 1600 error = err; 1601 } 1602 1603 nvlist_free(config); 1604 1605 nvlist_free(props); 1606 1607 return (error); 1608} 1609 1610static int 1611zfs_ioc_pool_export(zfs_cmd_t *zc) 1612{ 1613 int error; 1614 boolean_t force = (boolean_t)zc->zc_cookie; 1615 boolean_t hardforce = (boolean_t)zc->zc_guid; 1616 1617 zfs_log_history(zc); 1618 error = spa_export(zc->zc_name, NULL, force, hardforce); 1619 if (error == 0) 1620 zvol_remove_minors(zc->zc_name); 1621 return (error); 1622} 1623 1624static int 1625zfs_ioc_pool_configs(zfs_cmd_t *zc) 1626{ 1627 nvlist_t *configs; 1628 int error; 1629 1630 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL) 1631 return (SET_ERROR(EEXIST)); 1632 1633 error = put_nvlist(zc, configs); 1634 1635 nvlist_free(configs); 1636 1637 return (error); 1638} 1639 1640/* 1641 * inputs: 1642 * zc_name name of the pool 1643 * 1644 * outputs: 1645 * zc_cookie real errno 1646 * zc_nvlist_dst config nvlist 1647 * zc_nvlist_dst_size size of config nvlist 1648 */ 1649static int 1650zfs_ioc_pool_stats(zfs_cmd_t *zc) 1651{ 1652 nvlist_t *config; 1653 int error; 1654 int ret = 0; 1655 1656 error = spa_get_stats(zc->zc_name, &config, zc->zc_value, 1657 sizeof (zc->zc_value)); 1658 1659 if (config != NULL) { 1660 ret = put_nvlist(zc, config); 1661 nvlist_free(config); 1662 1663 /* 1664 * The config may be present even if 'error' is non-zero. 1665 * In this case we return success, and preserve the real errno 1666 * in 'zc_cookie'. 1667 */ 1668 zc->zc_cookie = error; 1669 } else { 1670 ret = error; 1671 } 1672 1673 return (ret); 1674} 1675 1676/* 1677 * Try to import the given pool, returning pool stats as appropriate so that 1678 * user land knows which devices are available and overall pool health. 1679 */ 1680static int 1681zfs_ioc_pool_tryimport(zfs_cmd_t *zc) 1682{ 1683 nvlist_t *tryconfig, *config; 1684 int error; 1685 1686 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1687 zc->zc_iflags, &tryconfig)) != 0) 1688 return (error); 1689 1690 config = spa_tryimport(tryconfig); 1691 1692 nvlist_free(tryconfig); 1693 1694 if (config == NULL) 1695 return (SET_ERROR(EINVAL)); 1696 1697 error = put_nvlist(zc, config); 1698 nvlist_free(config); 1699 1700 return (error); 1701} 1702 1703/* 1704 * inputs: 1705 * zc_name name of the pool 1706 * zc_cookie scan func (pool_scan_func_t) 1707 */ 1708static int 1709zfs_ioc_pool_scan(zfs_cmd_t *zc) 1710{ 1711 spa_t *spa; 1712 int error; 1713 1714 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1715 return (error); 1716 1717 if (zc->zc_cookie == POOL_SCAN_NONE) 1718 error = spa_scan_stop(spa); 1719 else 1720 error = spa_scan(spa, zc->zc_cookie); 1721 1722 spa_close(spa, FTAG); 1723 1724 return (error); 1725} 1726 1727static int 1728zfs_ioc_pool_freeze(zfs_cmd_t *zc) 1729{ 1730 spa_t *spa; 1731 int error; 1732 1733 error = spa_open(zc->zc_name, &spa, FTAG); 1734 if (error == 0) { 1735 spa_freeze(spa); 1736 spa_close(spa, FTAG); 1737 } 1738 return (error); 1739} 1740 1741static int 1742zfs_ioc_pool_upgrade(zfs_cmd_t *zc) 1743{ 1744 spa_t *spa; 1745 int error; 1746 1747 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1748 return (error); 1749 1750 if (zc->zc_cookie < spa_version(spa) || 1751 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) { 1752 spa_close(spa, FTAG); 1753 return (SET_ERROR(EINVAL)); 1754 } 1755 1756 spa_upgrade(spa, zc->zc_cookie); 1757 spa_close(spa, FTAG); 1758 1759 return (error); 1760} 1761 1762static int 1763zfs_ioc_pool_get_history(zfs_cmd_t *zc) 1764{ 1765 spa_t *spa; 1766 char *hist_buf; 1767 uint64_t size; 1768 int error; 1769 1770 if ((size = zc->zc_history_len) == 0) 1771 return (SET_ERROR(EINVAL)); 1772 1773 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1774 return (error); 1775 1776 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) { 1777 spa_close(spa, FTAG); 1778 return (SET_ERROR(ENOTSUP)); 1779 } 1780 1781 hist_buf = kmem_alloc(size, KM_SLEEP); 1782 if ((error = spa_history_get(spa, &zc->zc_history_offset, 1783 &zc->zc_history_len, hist_buf)) == 0) { 1784 error = ddi_copyout(hist_buf, 1785 (void *)(uintptr_t)zc->zc_history, 1786 zc->zc_history_len, zc->zc_iflags); 1787 } 1788 1789 spa_close(spa, FTAG); 1790 kmem_free(hist_buf, size); 1791 return (error); 1792} 1793 1794static int 1795zfs_ioc_pool_reguid(zfs_cmd_t *zc) 1796{ 1797 spa_t *spa; 1798 int error; 1799 1800 error = spa_open(zc->zc_name, &spa, FTAG); 1801 if (error == 0) { 1802 error = spa_change_guid(spa); 1803 spa_close(spa, FTAG); 1804 } 1805 return (error); 1806} 1807 1808static int 1809zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc) 1810{ 1811 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value)); 1812} 1813 1814/* 1815 * inputs: 1816 * zc_name name of filesystem 1817 * zc_obj object to find 1818 * 1819 * outputs: 1820 * zc_value name of object 1821 */ 1822static int 1823zfs_ioc_obj_to_path(zfs_cmd_t *zc) 1824{ 1825 objset_t *os; 1826 int error; 1827 1828 /* XXX reading from objset not owned */ 1829 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0) 1830 return (error); 1831 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1832 dmu_objset_rele(os, FTAG); 1833 return (SET_ERROR(EINVAL)); 1834 } 1835 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value, 1836 sizeof (zc->zc_value)); 1837 dmu_objset_rele(os, FTAG); 1838 1839 return (error); 1840} 1841 1842/* 1843 * inputs: 1844 * zc_name name of filesystem 1845 * zc_obj object to find 1846 * 1847 * outputs: 1848 * zc_stat stats on object 1849 * zc_value path to object 1850 */ 1851static int 1852zfs_ioc_obj_to_stats(zfs_cmd_t *zc) 1853{ 1854 objset_t *os; 1855 int error; 1856 1857 /* XXX reading from objset not owned */ 1858 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0) 1859 return (error); 1860 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1861 dmu_objset_rele(os, FTAG); 1862 return (SET_ERROR(EINVAL)); 1863 } 1864 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value, 1865 sizeof (zc->zc_value)); 1866 dmu_objset_rele(os, FTAG); 1867 1868 return (error); 1869} 1870 1871static int 1872zfs_ioc_vdev_add(zfs_cmd_t *zc) 1873{ 1874 spa_t *spa; 1875 int error; 1876 nvlist_t *config, **l2cache, **spares; 1877 uint_t nl2cache = 0, nspares = 0; 1878 1879 error = spa_open(zc->zc_name, &spa, FTAG); 1880 if (error != 0) 1881 return (error); 1882 1883 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1884 zc->zc_iflags, &config); 1885 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE, 1886 &l2cache, &nl2cache); 1887 1888 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES, 1889 &spares, &nspares); 1890 1891#ifdef illumos 1892 /* 1893 * A root pool with concatenated devices is not supported. 1894 * Thus, can not add a device to a root pool. 1895 * 1896 * Intent log device can not be added to a rootpool because 1897 * during mountroot, zil is replayed, a seperated log device 1898 * can not be accessed during the mountroot time. 1899 * 1900 * l2cache and spare devices are ok to be added to a rootpool. 1901 */ 1902 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) { 1903 nvlist_free(config); 1904 spa_close(spa, FTAG); 1905 return (SET_ERROR(EDOM)); 1906 } 1907#endif /* illumos */ 1908 1909 if (error == 0) { 1910 error = spa_vdev_add(spa, config); 1911 nvlist_free(config); 1912 } 1913 spa_close(spa, FTAG); 1914 return (error); 1915} 1916 1917/* 1918 * inputs: 1919 * zc_name name of the pool 1920 * zc_nvlist_conf nvlist of devices to remove 1921 * zc_cookie to stop the remove? 1922 */ 1923static int 1924zfs_ioc_vdev_remove(zfs_cmd_t *zc) 1925{ 1926 spa_t *spa; 1927 int error; 1928 1929 error = spa_open(zc->zc_name, &spa, FTAG); 1930 if (error != 0) 1931 return (error); 1932 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE); 1933 spa_close(spa, FTAG); 1934 return (error); 1935} 1936 1937static int 1938zfs_ioc_vdev_set_state(zfs_cmd_t *zc) 1939{ 1940 spa_t *spa; 1941 int error; 1942 vdev_state_t newstate = VDEV_STATE_UNKNOWN; 1943 1944 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1945 return (error); 1946 switch (zc->zc_cookie) { 1947 case VDEV_STATE_ONLINE: 1948 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate); 1949 break; 1950 1951 case VDEV_STATE_OFFLINE: 1952 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj); 1953 break; 1954 1955 case VDEV_STATE_FAULTED: 1956 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED && 1957 zc->zc_obj != VDEV_AUX_EXTERNAL) 1958 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED; 1959 1960 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj); 1961 break; 1962 1963 case VDEV_STATE_DEGRADED: 1964 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED && 1965 zc->zc_obj != VDEV_AUX_EXTERNAL) 1966 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED; 1967 1968 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj); 1969 break; 1970 1971 default: 1972 error = SET_ERROR(EINVAL); 1973 } 1974 zc->zc_cookie = newstate; 1975 spa_close(spa, FTAG); 1976 return (error); 1977} 1978 1979static int 1980zfs_ioc_vdev_attach(zfs_cmd_t *zc) 1981{ 1982 spa_t *spa; 1983 int replacing = zc->zc_cookie; 1984 nvlist_t *config; 1985 int error; 1986 1987 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1988 return (error); 1989 1990 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1991 zc->zc_iflags, &config)) == 0) { 1992 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing); 1993 nvlist_free(config); 1994 } 1995 1996 spa_close(spa, FTAG); 1997 return (error); 1998} 1999 2000static int 2001zfs_ioc_vdev_detach(zfs_cmd_t *zc) 2002{ 2003 spa_t *spa; 2004 int error; 2005 2006 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 2007 return (error); 2008 2009 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE); 2010 2011 spa_close(spa, FTAG); 2012 return (error); 2013} 2014 2015static int 2016zfs_ioc_vdev_split(zfs_cmd_t *zc) 2017{ 2018 spa_t *spa; 2019 nvlist_t *config, *props = NULL; 2020 int error; 2021 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT); 2022 2023 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 2024 return (error); 2025 2026 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 2027 zc->zc_iflags, &config)) { 2028 spa_close(spa, FTAG); 2029 return (error); 2030 } 2031 2032 if (zc->zc_nvlist_src_size != 0 && (error = 2033 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2034 zc->zc_iflags, &props))) { 2035 spa_close(spa, FTAG); 2036 nvlist_free(config); 2037 return (error); 2038 } 2039 2040 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp); 2041 2042 spa_close(spa, FTAG); 2043 2044 nvlist_free(config); 2045 nvlist_free(props); 2046 2047 return (error); 2048} 2049 2050static int 2051zfs_ioc_vdev_setpath(zfs_cmd_t *zc) 2052{ 2053 spa_t *spa; 2054 char *path = zc->zc_value; 2055 uint64_t guid = zc->zc_guid; 2056 int error; 2057 2058 error = spa_open(zc->zc_name, &spa, FTAG); 2059 if (error != 0) 2060 return (error); 2061 2062 error = spa_vdev_setpath(spa, guid, path); 2063 spa_close(spa, FTAG); 2064 return (error); 2065} 2066 2067static int 2068zfs_ioc_vdev_setfru(zfs_cmd_t *zc) 2069{ 2070 spa_t *spa; 2071 char *fru = zc->zc_value; 2072 uint64_t guid = zc->zc_guid; 2073 int error; 2074 2075 error = spa_open(zc->zc_name, &spa, FTAG); 2076 if (error != 0) 2077 return (error); 2078 2079 error = spa_vdev_setfru(spa, guid, fru); 2080 spa_close(spa, FTAG); 2081 return (error); 2082} 2083 2084static int 2085zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os) 2086{ 2087 int error = 0; 2088 nvlist_t *nv; 2089 2090 dmu_objset_fast_stat(os, &zc->zc_objset_stats); 2091 2092 if (zc->zc_nvlist_dst != 0 && 2093 (error = dsl_prop_get_all(os, &nv)) == 0) { 2094 dmu_objset_stats(os, nv); 2095 /* 2096 * NB: zvol_get_stats() will read the objset contents, 2097 * which we aren't supposed to do with a 2098 * DS_MODE_USER hold, because it could be 2099 * inconsistent. So this is a bit of a workaround... 2100 * XXX reading with out owning 2101 */ 2102 if (!zc->zc_objset_stats.dds_inconsistent && 2103 dmu_objset_type(os) == DMU_OST_ZVOL) { 2104 error = zvol_get_stats(os, nv); 2105 if (error == EIO) 2106 return (error); 2107 VERIFY0(error); 2108 } 2109 error = put_nvlist(zc, nv); 2110 nvlist_free(nv); 2111 } 2112 2113 return (error); 2114} 2115 2116/* 2117 * inputs: 2118 * zc_name name of filesystem 2119 * zc_nvlist_dst_size size of buffer for property nvlist 2120 * 2121 * outputs: 2122 * zc_objset_stats stats 2123 * zc_nvlist_dst property nvlist 2124 * zc_nvlist_dst_size size of property nvlist 2125 */ 2126static int 2127zfs_ioc_objset_stats(zfs_cmd_t *zc) 2128{ 2129 objset_t *os; 2130 int error; 2131 2132 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 2133 if (error == 0) { 2134 error = zfs_ioc_objset_stats_impl(zc, os); 2135 dmu_objset_rele(os, FTAG); 2136 } 2137 2138 if (error == ENOMEM) 2139 error = 0; 2140 return (error); 2141} 2142 2143/* 2144 * inputs: 2145 * zc_name name of filesystem 2146 * zc_nvlist_dst_size size of buffer for property nvlist 2147 * 2148 * outputs: 2149 * zc_nvlist_dst received property nvlist 2150 * zc_nvlist_dst_size size of received property nvlist 2151 * 2152 * Gets received properties (distinct from local properties on or after 2153 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from 2154 * local property values. 2155 */ 2156static int 2157zfs_ioc_objset_recvd_props(zfs_cmd_t *zc) 2158{ 2159 int error = 0; 2160 nvlist_t *nv; 2161 2162 /* 2163 * Without this check, we would return local property values if the 2164 * caller has not already received properties on or after 2165 * SPA_VERSION_RECVD_PROPS. 2166 */ 2167 if (!dsl_prop_get_hasrecvd(zc->zc_name)) 2168 return (SET_ERROR(ENOTSUP)); 2169 2170 if (zc->zc_nvlist_dst != 0 && 2171 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) { 2172 error = put_nvlist(zc, nv); 2173 nvlist_free(nv); 2174 } 2175 2176 return (error); 2177} 2178 2179static int 2180nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop) 2181{ 2182 uint64_t value; 2183 int error; 2184 2185 /* 2186 * zfs_get_zplprop() will either find a value or give us 2187 * the default value (if there is one). 2188 */ 2189 if ((error = zfs_get_zplprop(os, prop, &value)) != 0) 2190 return (error); 2191 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0); 2192 return (0); 2193} 2194 2195/* 2196 * inputs: 2197 * zc_name name of filesystem 2198 * zc_nvlist_dst_size size of buffer for zpl property nvlist 2199 * 2200 * outputs: 2201 * zc_nvlist_dst zpl property nvlist 2202 * zc_nvlist_dst_size size of zpl property nvlist 2203 */ 2204static int 2205zfs_ioc_objset_zplprops(zfs_cmd_t *zc) 2206{ 2207 objset_t *os; 2208 int err; 2209 2210 /* XXX reading without owning */ 2211 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os)) 2212 return (err); 2213 2214 dmu_objset_fast_stat(os, &zc->zc_objset_stats); 2215 2216 /* 2217 * NB: nvl_add_zplprop() will read the objset contents, 2218 * which we aren't supposed to do with a DS_MODE_USER 2219 * hold, because it could be inconsistent. 2220 */ 2221 if (zc->zc_nvlist_dst != 0 && 2222 !zc->zc_objset_stats.dds_inconsistent && 2223 dmu_objset_type(os) == DMU_OST_ZFS) { 2224 nvlist_t *nv; 2225 2226 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2227 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 && 2228 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 && 2229 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 && 2230 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0) 2231 err = put_nvlist(zc, nv); 2232 nvlist_free(nv); 2233 } else { 2234 err = SET_ERROR(ENOENT); 2235 } 2236 dmu_objset_rele(os, FTAG); 2237 return (err); 2238} 2239 2240boolean_t 2241dataset_name_hidden(const char *name) 2242{ 2243 /* 2244 * Skip over datasets that are not visible in this zone, 2245 * internal datasets (which have a $ in their name), and 2246 * temporary datasets (which have a % in their name). 2247 */ 2248 if (strchr(name, '$') != NULL) 2249 return (B_TRUE); 2250 if (strchr(name, '%') != NULL) 2251 return (B_TRUE); 2252 if (!INGLOBALZONE(curthread) && !zone_dataset_visible(name, NULL)) 2253 return (B_TRUE); 2254 return (B_FALSE); 2255} 2256 2257/* 2258 * inputs: 2259 * zc_name name of filesystem 2260 * zc_cookie zap cursor 2261 * zc_nvlist_dst_size size of buffer for property nvlist 2262 * 2263 * outputs: 2264 * zc_name name of next filesystem 2265 * zc_cookie zap cursor 2266 * zc_objset_stats stats 2267 * zc_nvlist_dst property nvlist 2268 * zc_nvlist_dst_size size of property nvlist 2269 */ 2270static int 2271zfs_ioc_dataset_list_next(zfs_cmd_t *zc) 2272{ 2273 objset_t *os; 2274 int error; 2275 char *p; 2276 size_t orig_len = strlen(zc->zc_name); 2277 2278top: 2279 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) { 2280 if (error == ENOENT) 2281 error = SET_ERROR(ESRCH); 2282 return (error); 2283 } 2284 2285 p = strrchr(zc->zc_name, '/'); 2286 if (p == NULL || p[1] != '\0') 2287 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name)); 2288 p = zc->zc_name + strlen(zc->zc_name); 2289 2290 do { 2291 error = dmu_dir_list_next(os, 2292 sizeof (zc->zc_name) - (p - zc->zc_name), p, 2293 NULL, &zc->zc_cookie); 2294 if (error == ENOENT) 2295 error = SET_ERROR(ESRCH); 2296 } while (error == 0 && dataset_name_hidden(zc->zc_name)); 2297 dmu_objset_rele(os, FTAG); 2298 2299 /* 2300 * If it's an internal dataset (ie. with a '$' in its name), 2301 * don't try to get stats for it, otherwise we'll return ENOENT. 2302 */ 2303 if (error == 0 && strchr(zc->zc_name, '$') == NULL) { 2304 error = zfs_ioc_objset_stats(zc); /* fill in the stats */ 2305 if (error == ENOENT) { 2306 /* We lost a race with destroy, get the next one. */ 2307 zc->zc_name[orig_len] = '\0'; 2308 goto top; 2309 } 2310 } 2311 return (error); 2312} 2313 2314/* 2315 * inputs: 2316 * zc_name name of filesystem 2317 * zc_cookie zap cursor 2318 * zc_nvlist_dst_size size of buffer for property nvlist 2319 * zc_simple when set, only name is requested 2320 * 2321 * outputs: 2322 * zc_name name of next snapshot 2323 * zc_objset_stats stats 2324 * zc_nvlist_dst property nvlist 2325 * zc_nvlist_dst_size size of property nvlist 2326 */ 2327static int 2328zfs_ioc_snapshot_list_next(zfs_cmd_t *zc) 2329{ 2330 objset_t *os; 2331 int error; 2332 2333 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 2334 if (error != 0) { 2335 return (error == ENOENT ? ESRCH : error); 2336 } 2337 2338 /* 2339 * A dataset name of maximum length cannot have any snapshots, 2340 * so exit immediately. 2341 */ 2342 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) { 2343 dmu_objset_rele(os, FTAG); 2344 return (SET_ERROR(ESRCH)); 2345 } 2346 2347 error = dmu_snapshot_list_next(os, 2348 sizeof (zc->zc_name) - strlen(zc->zc_name), 2349 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie, 2350 NULL); 2351 2352 if (error == 0 && !zc->zc_simple) { 2353 dsl_dataset_t *ds; 2354 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool; 2355 2356 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds); 2357 if (error == 0) { 2358 objset_t *ossnap; 2359 2360 error = dmu_objset_from_ds(ds, &ossnap); 2361 if (error == 0) 2362 error = zfs_ioc_objset_stats_impl(zc, ossnap); 2363 dsl_dataset_rele(ds, FTAG); 2364 } 2365 } else if (error == ENOENT) { 2366 error = SET_ERROR(ESRCH); 2367 } 2368 2369 dmu_objset_rele(os, FTAG); 2370 /* if we failed, undo the @ that we tacked on to zc_name */ 2371 if (error != 0) 2372 *strchr(zc->zc_name, '@') = '\0'; 2373 return (error); 2374} 2375 2376static int 2377zfs_prop_set_userquota(const char *dsname, nvpair_t *pair) 2378{ 2379 const char *propname = nvpair_name(pair); 2380 uint64_t *valary; 2381 unsigned int vallen; 2382 const char *domain; 2383 char *dash; 2384 zfs_userquota_prop_t type; 2385 uint64_t rid; 2386 uint64_t quota; 2387 zfsvfs_t *zfsvfs; 2388 int err; 2389 2390 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2391 nvlist_t *attrs; 2392 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 2393 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2394 &pair) != 0) 2395 return (SET_ERROR(EINVAL)); 2396 } 2397 2398 /* 2399 * A correctly constructed propname is encoded as 2400 * userquota@<rid>-<domain>. 2401 */ 2402 if ((dash = strchr(propname, '-')) == NULL || 2403 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 || 2404 vallen != 3) 2405 return (SET_ERROR(EINVAL)); 2406 2407 domain = dash + 1; 2408 type = valary[0]; 2409 rid = valary[1]; 2410 quota = valary[2]; 2411 2412 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE); 2413 if (err == 0) { 2414 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota); 2415 zfsvfs_rele(zfsvfs, FTAG); 2416 } 2417 2418 return (err); 2419} 2420 2421/* 2422 * If the named property is one that has a special function to set its value, 2423 * return 0 on success and a positive error code on failure; otherwise if it is 2424 * not one of the special properties handled by this function, return -1. 2425 * 2426 * XXX: It would be better for callers of the property interface if we handled 2427 * these special cases in dsl_prop.c (in the dsl layer). 2428 */ 2429static int 2430zfs_prop_set_special(const char *dsname, zprop_source_t source, 2431 nvpair_t *pair) 2432{ 2433 const char *propname = nvpair_name(pair); 2434 zfs_prop_t prop = zfs_name_to_prop(propname); 2435 uint64_t intval; 2436 int err = -1; 2437 2438 if (prop == ZPROP_INVAL) { 2439 if (zfs_prop_userquota(propname)) 2440 return (zfs_prop_set_userquota(dsname, pair)); 2441 return (-1); 2442 } 2443 2444 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2445 nvlist_t *attrs; 2446 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 2447 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2448 &pair) == 0); 2449 } 2450 2451 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING) 2452 return (-1); 2453 2454 VERIFY(0 == nvpair_value_uint64(pair, &intval)); 2455 2456 switch (prop) { 2457 case ZFS_PROP_QUOTA: 2458 err = dsl_dir_set_quota(dsname, source, intval); 2459 break; 2460 case ZFS_PROP_REFQUOTA: 2461 err = dsl_dataset_set_refquota(dsname, source, intval); 2462 break; 2463 case ZFS_PROP_FILESYSTEM_LIMIT: 2464 case ZFS_PROP_SNAPSHOT_LIMIT: 2465 if (intval == UINT64_MAX) { 2466 /* clearing the limit, just do it */ 2467 err = 0; 2468 } else { 2469 err = dsl_dir_activate_fs_ss_limit(dsname); 2470 } 2471 /* 2472 * Set err to -1 to force the zfs_set_prop_nvlist code down the 2473 * default path to set the value in the nvlist. 2474 */ 2475 if (err == 0) 2476 err = -1; 2477 break; 2478 case ZFS_PROP_RESERVATION: 2479 err = dsl_dir_set_reservation(dsname, source, intval); 2480 break; 2481 case ZFS_PROP_REFRESERVATION: 2482 err = dsl_dataset_set_refreservation(dsname, source, intval); 2483 break; 2484 case ZFS_PROP_VOLSIZE: 2485 err = zvol_set_volsize(dsname, intval); 2486 break; 2487 case ZFS_PROP_VERSION: 2488 { 2489 zfsvfs_t *zfsvfs; 2490 2491 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0) 2492 break; 2493 2494 err = zfs_set_version(zfsvfs, intval); 2495 zfsvfs_rele(zfsvfs, FTAG); 2496 2497 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) { 2498 zfs_cmd_t *zc; 2499 2500 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP); 2501 (void) strcpy(zc->zc_name, dsname); 2502 (void) zfs_ioc_userspace_upgrade(zc); 2503 kmem_free(zc, sizeof (zfs_cmd_t)); 2504 } 2505 break; 2506 } 2507 default: 2508 err = -1; 2509 } 2510 2511 return (err); 2512} 2513 2514/* 2515 * This function is best effort. If it fails to set any of the given properties, 2516 * it continues to set as many as it can and returns the last error 2517 * encountered. If the caller provides a non-NULL errlist, it will be filled in 2518 * with the list of names of all the properties that failed along with the 2519 * corresponding error numbers. 2520 * 2521 * If every property is set successfully, zero is returned and errlist is not 2522 * modified. 2523 */ 2524int 2525zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl, 2526 nvlist_t *errlist) 2527{ 2528 nvpair_t *pair; 2529 nvpair_t *propval; 2530 int rv = 0; 2531 uint64_t intval; 2532 char *strval; 2533 nvlist_t *genericnvl = fnvlist_alloc(); 2534 nvlist_t *retrynvl = fnvlist_alloc(); 2535 2536retry: 2537 pair = NULL; 2538 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { 2539 const char *propname = nvpair_name(pair); 2540 zfs_prop_t prop = zfs_name_to_prop(propname); 2541 int err = 0; 2542 2543 /* decode the property value */ 2544 propval = pair; 2545 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2546 nvlist_t *attrs; 2547 attrs = fnvpair_value_nvlist(pair); 2548 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2549 &propval) != 0) 2550 err = SET_ERROR(EINVAL); 2551 } 2552 2553 /* Validate value type */ 2554 if (err == 0 && prop == ZPROP_INVAL) { 2555 if (zfs_prop_user(propname)) { 2556 if (nvpair_type(propval) != DATA_TYPE_STRING) 2557 err = SET_ERROR(EINVAL); 2558 } else if (zfs_prop_userquota(propname)) { 2559 if (nvpair_type(propval) != 2560 DATA_TYPE_UINT64_ARRAY) 2561 err = SET_ERROR(EINVAL); 2562 } else { 2563 err = SET_ERROR(EINVAL); 2564 } 2565 } else if (err == 0) { 2566 if (nvpair_type(propval) == DATA_TYPE_STRING) { 2567 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING) 2568 err = SET_ERROR(EINVAL); 2569 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) { 2570 const char *unused; 2571 2572 intval = fnvpair_value_uint64(propval); 2573 2574 switch (zfs_prop_get_type(prop)) { 2575 case PROP_TYPE_NUMBER: 2576 break; 2577 case PROP_TYPE_STRING: 2578 err = SET_ERROR(EINVAL); 2579 break; 2580 case PROP_TYPE_INDEX: 2581 if (zfs_prop_index_to_string(prop, 2582 intval, &unused) != 0) 2583 err = SET_ERROR(EINVAL); 2584 break; 2585 default: 2586 cmn_err(CE_PANIC, 2587 "unknown property type"); 2588 } 2589 } else { 2590 err = SET_ERROR(EINVAL); 2591 } 2592 } 2593 2594 /* Validate permissions */ 2595 if (err == 0) 2596 err = zfs_check_settable(dsname, pair, CRED()); 2597 2598 if (err == 0) { 2599 err = zfs_prop_set_special(dsname, source, pair); 2600 if (err == -1) { 2601 /* 2602 * For better performance we build up a list of 2603 * properties to set in a single transaction. 2604 */ 2605 err = nvlist_add_nvpair(genericnvl, pair); 2606 } else if (err != 0 && nvl != retrynvl) { 2607 /* 2608 * This may be a spurious error caused by 2609 * receiving quota and reservation out of order. 2610 * Try again in a second pass. 2611 */ 2612 err = nvlist_add_nvpair(retrynvl, pair); 2613 } 2614 } 2615 2616 if (err != 0) { 2617 if (errlist != NULL) 2618 fnvlist_add_int32(errlist, propname, err); 2619 rv = err; 2620 } 2621 } 2622 2623 if (nvl != retrynvl && !nvlist_empty(retrynvl)) { 2624 nvl = retrynvl; 2625 goto retry; 2626 } 2627 2628 if (!nvlist_empty(genericnvl) && 2629 dsl_props_set(dsname, source, genericnvl) != 0) { 2630 /* 2631 * If this fails, we still want to set as many properties as we 2632 * can, so try setting them individually. 2633 */ 2634 pair = NULL; 2635 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) { 2636 const char *propname = nvpair_name(pair); 2637 int err = 0; 2638 2639 propval = pair; 2640 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2641 nvlist_t *attrs; 2642 attrs = fnvpair_value_nvlist(pair); 2643 propval = fnvlist_lookup_nvpair(attrs, 2644 ZPROP_VALUE); 2645 } 2646 2647 if (nvpair_type(propval) == DATA_TYPE_STRING) { 2648 strval = fnvpair_value_string(propval); 2649 err = dsl_prop_set_string(dsname, propname, 2650 source, strval); 2651 } else { 2652 intval = fnvpair_value_uint64(propval); 2653 err = dsl_prop_set_int(dsname, propname, source, 2654 intval); 2655 } 2656 2657 if (err != 0) { 2658 if (errlist != NULL) { 2659 fnvlist_add_int32(errlist, propname, 2660 err); 2661 } 2662 rv = err; 2663 } 2664 } 2665 } 2666 nvlist_free(genericnvl); 2667 nvlist_free(retrynvl); 2668 2669 return (rv); 2670} 2671 2672/* 2673 * Check that all the properties are valid user properties. 2674 */ 2675static int 2676zfs_check_userprops(const char *fsname, nvlist_t *nvl) 2677{ 2678 nvpair_t *pair = NULL; 2679 int error = 0; 2680 2681 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { 2682 const char *propname = nvpair_name(pair); 2683 2684 if (!zfs_prop_user(propname) || 2685 nvpair_type(pair) != DATA_TYPE_STRING) 2686 return (SET_ERROR(EINVAL)); 2687 2688 if (error = zfs_secpolicy_write_perms(fsname, 2689 ZFS_DELEG_PERM_USERPROP, CRED())) 2690 return (error); 2691 2692 if (strlen(propname) >= ZAP_MAXNAMELEN) 2693 return (SET_ERROR(ENAMETOOLONG)); 2694 2695 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN) 2696 return (E2BIG); 2697 } 2698 return (0); 2699} 2700 2701static void 2702props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops) 2703{ 2704 nvpair_t *pair; 2705 2706 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2707 2708 pair = NULL; 2709 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) { 2710 if (nvlist_exists(skipped, nvpair_name(pair))) 2711 continue; 2712 2713 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0); 2714 } 2715} 2716 2717static int 2718clear_received_props(const char *dsname, nvlist_t *props, 2719 nvlist_t *skipped) 2720{ 2721 int err = 0; 2722 nvlist_t *cleared_props = NULL; 2723 props_skip(props, skipped, &cleared_props); 2724 if (!nvlist_empty(cleared_props)) { 2725 /* 2726 * Acts on local properties until the dataset has received 2727 * properties at least once on or after SPA_VERSION_RECVD_PROPS. 2728 */ 2729 zprop_source_t flags = (ZPROP_SRC_NONE | 2730 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0)); 2731 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL); 2732 } 2733 nvlist_free(cleared_props); 2734 return (err); 2735} 2736 2737/* 2738 * inputs: 2739 * zc_name name of filesystem 2740 * zc_value name of property to set 2741 * zc_nvlist_src{_size} nvlist of properties to apply 2742 * zc_cookie received properties flag 2743 * 2744 * outputs: 2745 * zc_nvlist_dst{_size} error for each unapplied received property 2746 */ 2747static int 2748zfs_ioc_set_prop(zfs_cmd_t *zc) 2749{ 2750 nvlist_t *nvl; 2751 boolean_t received = zc->zc_cookie; 2752 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED : 2753 ZPROP_SRC_LOCAL); 2754 nvlist_t *errors; 2755 int error; 2756 2757 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2758 zc->zc_iflags, &nvl)) != 0) 2759 return (error); 2760 2761 if (received) { 2762 nvlist_t *origprops; 2763 2764 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) { 2765 (void) clear_received_props(zc->zc_name, 2766 origprops, nvl); 2767 nvlist_free(origprops); 2768 } 2769 2770 error = dsl_prop_set_hasrecvd(zc->zc_name); 2771 } 2772 2773 errors = fnvlist_alloc(); 2774 if (error == 0) 2775 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors); 2776 2777 if (zc->zc_nvlist_dst != 0 && errors != NULL) { 2778 (void) put_nvlist(zc, errors); 2779 } 2780 2781 nvlist_free(errors); 2782 nvlist_free(nvl); 2783 return (error); 2784} 2785 2786/* 2787 * inputs: 2788 * zc_name name of filesystem 2789 * zc_value name of property to inherit 2790 * zc_cookie revert to received value if TRUE 2791 * 2792 * outputs: none 2793 */ 2794static int 2795zfs_ioc_inherit_prop(zfs_cmd_t *zc) 2796{ 2797 const char *propname = zc->zc_value; 2798 zfs_prop_t prop = zfs_name_to_prop(propname); 2799 boolean_t received = zc->zc_cookie; 2800 zprop_source_t source = (received 2801 ? ZPROP_SRC_NONE /* revert to received value, if any */ 2802 : ZPROP_SRC_INHERITED); /* explicitly inherit */ 2803 2804 if (received) { 2805 nvlist_t *dummy; 2806 nvpair_t *pair; 2807 zprop_type_t type; 2808 int err; 2809 2810 /* 2811 * zfs_prop_set_special() expects properties in the form of an 2812 * nvpair with type info. 2813 */ 2814 if (prop == ZPROP_INVAL) { 2815 if (!zfs_prop_user(propname)) 2816 return (SET_ERROR(EINVAL)); 2817 2818 type = PROP_TYPE_STRING; 2819 } else if (prop == ZFS_PROP_VOLSIZE || 2820 prop == ZFS_PROP_VERSION) { 2821 return (SET_ERROR(EINVAL)); 2822 } else { 2823 type = zfs_prop_get_type(prop); 2824 } 2825 2826 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2827 2828 switch (type) { 2829 case PROP_TYPE_STRING: 2830 VERIFY(0 == nvlist_add_string(dummy, propname, "")); 2831 break; 2832 case PROP_TYPE_NUMBER: 2833 case PROP_TYPE_INDEX: 2834 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0)); 2835 break; 2836 default: 2837 nvlist_free(dummy); 2838 return (SET_ERROR(EINVAL)); 2839 } 2840 2841 pair = nvlist_next_nvpair(dummy, NULL); 2842 err = zfs_prop_set_special(zc->zc_name, source, pair); 2843 nvlist_free(dummy); 2844 if (err != -1) 2845 return (err); /* special property already handled */ 2846 } else { 2847 /* 2848 * Only check this in the non-received case. We want to allow 2849 * 'inherit -S' to revert non-inheritable properties like quota 2850 * and reservation to the received or default values even though 2851 * they are not considered inheritable. 2852 */ 2853 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop)) 2854 return (SET_ERROR(EINVAL)); 2855 } 2856 2857 /* property name has been validated by zfs_secpolicy_inherit_prop() */ 2858 return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source)); 2859} 2860 2861static int 2862zfs_ioc_pool_set_props(zfs_cmd_t *zc) 2863{ 2864 nvlist_t *props; 2865 spa_t *spa; 2866 int error; 2867 nvpair_t *pair; 2868 2869 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2870 zc->zc_iflags, &props)) 2871 return (error); 2872 2873 /* 2874 * If the only property is the configfile, then just do a spa_lookup() 2875 * to handle the faulted case. 2876 */ 2877 pair = nvlist_next_nvpair(props, NULL); 2878 if (pair != NULL && strcmp(nvpair_name(pair), 2879 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 && 2880 nvlist_next_nvpair(props, pair) == NULL) { 2881 mutex_enter(&spa_namespace_lock); 2882 if ((spa = spa_lookup(zc->zc_name)) != NULL) { 2883 spa_configfile_set(spa, props, B_FALSE); 2884 spa_config_sync(spa, B_FALSE, B_TRUE); 2885 } 2886 mutex_exit(&spa_namespace_lock); 2887 if (spa != NULL) { 2888 nvlist_free(props); 2889 return (0); 2890 } 2891 } 2892 2893 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { 2894 nvlist_free(props); 2895 return (error); 2896 } 2897 2898 error = spa_prop_set(spa, props); 2899 2900 nvlist_free(props); 2901 spa_close(spa, FTAG); 2902 2903 return (error); 2904} 2905 2906static int 2907zfs_ioc_pool_get_props(zfs_cmd_t *zc) 2908{ 2909 spa_t *spa; 2910 int error; 2911 nvlist_t *nvp = NULL; 2912 2913 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { 2914 /* 2915 * If the pool is faulted, there may be properties we can still 2916 * get (such as altroot and cachefile), so attempt to get them 2917 * anyway. 2918 */ 2919 mutex_enter(&spa_namespace_lock); 2920 if ((spa = spa_lookup(zc->zc_name)) != NULL) 2921 error = spa_prop_get(spa, &nvp); 2922 mutex_exit(&spa_namespace_lock); 2923 } else { 2924 error = spa_prop_get(spa, &nvp); 2925 spa_close(spa, FTAG); 2926 } 2927 2928 if (error == 0 && zc->zc_nvlist_dst != 0) 2929 error = put_nvlist(zc, nvp); 2930 else 2931 error = SET_ERROR(EFAULT); 2932 2933 nvlist_free(nvp); 2934 return (error); 2935} 2936 2937/* 2938 * inputs: 2939 * zc_name name of filesystem 2940 * zc_nvlist_src{_size} nvlist of delegated permissions 2941 * zc_perm_action allow/unallow flag 2942 * 2943 * outputs: none 2944 */ 2945static int 2946zfs_ioc_set_fsacl(zfs_cmd_t *zc) 2947{ 2948 int error; 2949 nvlist_t *fsaclnv = NULL; 2950 2951 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2952 zc->zc_iflags, &fsaclnv)) != 0) 2953 return (error); 2954 2955 /* 2956 * Verify nvlist is constructed correctly 2957 */ 2958 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) { 2959 nvlist_free(fsaclnv); 2960 return (SET_ERROR(EINVAL)); 2961 } 2962 2963 /* 2964 * If we don't have PRIV_SYS_MOUNT, then validate 2965 * that user is allowed to hand out each permission in 2966 * the nvlist(s) 2967 */ 2968 2969 error = secpolicy_zfs(CRED()); 2970 if (error != 0) { 2971 if (zc->zc_perm_action == B_FALSE) { 2972 error = dsl_deleg_can_allow(zc->zc_name, 2973 fsaclnv, CRED()); 2974 } else { 2975 error = dsl_deleg_can_unallow(zc->zc_name, 2976 fsaclnv, CRED()); 2977 } 2978 } 2979 2980 if (error == 0) 2981 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action); 2982 2983 nvlist_free(fsaclnv); 2984 return (error); 2985} 2986 2987/* 2988 * inputs: 2989 * zc_name name of filesystem 2990 * 2991 * outputs: 2992 * zc_nvlist_src{_size} nvlist of delegated permissions 2993 */ 2994static int 2995zfs_ioc_get_fsacl(zfs_cmd_t *zc) 2996{ 2997 nvlist_t *nvp; 2998 int error; 2999 3000 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) { 3001 error = put_nvlist(zc, nvp); 3002 nvlist_free(nvp); 3003 } 3004 3005 return (error); 3006} 3007 3008/* 3009 * Search the vfs list for a specified resource. Returns a pointer to it 3010 * or NULL if no suitable entry is found. The caller of this routine 3011 * is responsible for releasing the returned vfs pointer. 3012 */ 3013static vfs_t * 3014zfs_get_vfs(const char *resource) 3015{ 3016 vfs_t *vfsp; 3017 3018 mtx_lock(&mountlist_mtx); 3019 TAILQ_FOREACH(vfsp, &mountlist, mnt_list) { 3020 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) { 3021 VFS_HOLD(vfsp); 3022 break; 3023 } 3024 } 3025 mtx_unlock(&mountlist_mtx); 3026 return (vfsp); 3027} 3028 3029/* ARGSUSED */ 3030static void 3031zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) 3032{ 3033 zfs_creat_t *zct = arg; 3034 3035 zfs_create_fs(os, cr, zct->zct_zplprops, tx); 3036} 3037 3038#define ZFS_PROP_UNDEFINED ((uint64_t)-1) 3039 3040/* 3041 * inputs: 3042 * os parent objset pointer (NULL if root fs) 3043 * fuids_ok fuids allowed in this version of the spa? 3044 * sa_ok SAs allowed in this version of the spa? 3045 * createprops list of properties requested by creator 3046 * 3047 * outputs: 3048 * zplprops values for the zplprops we attach to the master node object 3049 * is_ci true if requested file system will be purely case-insensitive 3050 * 3051 * Determine the settings for utf8only, normalization and 3052 * casesensitivity. Specific values may have been requested by the 3053 * creator and/or we can inherit values from the parent dataset. If 3054 * the file system is of too early a vintage, a creator can not 3055 * request settings for these properties, even if the requested 3056 * setting is the default value. We don't actually want to create dsl 3057 * properties for these, so remove them from the source nvlist after 3058 * processing. 3059 */ 3060static int 3061zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver, 3062 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops, 3063 nvlist_t *zplprops, boolean_t *is_ci) 3064{ 3065 uint64_t sense = ZFS_PROP_UNDEFINED; 3066 uint64_t norm = ZFS_PROP_UNDEFINED; 3067 uint64_t u8 = ZFS_PROP_UNDEFINED; 3068 3069 ASSERT(zplprops != NULL); 3070 3071 /* 3072 * Pull out creator prop choices, if any. 3073 */ 3074 if (createprops) { 3075 (void) nvlist_lookup_uint64(createprops, 3076 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver); 3077 (void) nvlist_lookup_uint64(createprops, 3078 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm); 3079 (void) nvlist_remove_all(createprops, 3080 zfs_prop_to_name(ZFS_PROP_NORMALIZE)); 3081 (void) nvlist_lookup_uint64(createprops, 3082 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8); 3083 (void) nvlist_remove_all(createprops, 3084 zfs_prop_to_name(ZFS_PROP_UTF8ONLY)); 3085 (void) nvlist_lookup_uint64(createprops, 3086 zfs_prop_to_name(ZFS_PROP_CASE), &sense); 3087 (void) nvlist_remove_all(createprops, 3088 zfs_prop_to_name(ZFS_PROP_CASE)); 3089 } 3090 3091 /* 3092 * If the zpl version requested is whacky or the file system 3093 * or pool is version is too "young" to support normalization 3094 * and the creator tried to set a value for one of the props, 3095 * error out. 3096 */ 3097 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) || 3098 (zplver >= ZPL_VERSION_FUID && !fuids_ok) || 3099 (zplver >= ZPL_VERSION_SA && !sa_ok) || 3100 (zplver < ZPL_VERSION_NORMALIZATION && 3101 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED || 3102 sense != ZFS_PROP_UNDEFINED))) 3103 return (SET_ERROR(ENOTSUP)); 3104 3105 /* 3106 * Put the version in the zplprops 3107 */ 3108 VERIFY(nvlist_add_uint64(zplprops, 3109 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0); 3110 3111 if (norm == ZFS_PROP_UNDEFINED) 3112 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0); 3113 VERIFY(nvlist_add_uint64(zplprops, 3114 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0); 3115 3116 /* 3117 * If we're normalizing, names must always be valid UTF-8 strings. 3118 */ 3119 if (norm) 3120 u8 = 1; 3121 if (u8 == ZFS_PROP_UNDEFINED) 3122 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0); 3123 VERIFY(nvlist_add_uint64(zplprops, 3124 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0); 3125 3126 if (sense == ZFS_PROP_UNDEFINED) 3127 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0); 3128 VERIFY(nvlist_add_uint64(zplprops, 3129 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0); 3130 3131 if (is_ci) 3132 *is_ci = (sense == ZFS_CASE_INSENSITIVE); 3133 3134 return (0); 3135} 3136 3137static int 3138zfs_fill_zplprops(const char *dataset, nvlist_t *createprops, 3139 nvlist_t *zplprops, boolean_t *is_ci) 3140{ 3141 boolean_t fuids_ok, sa_ok; 3142 uint64_t zplver = ZPL_VERSION; 3143 objset_t *os = NULL; 3144 char parentname[MAXNAMELEN]; 3145 char *cp; 3146 spa_t *spa; 3147 uint64_t spa_vers; 3148 int error; 3149 3150 (void) strlcpy(parentname, dataset, sizeof (parentname)); 3151 cp = strrchr(parentname, '/'); 3152 ASSERT(cp != NULL); 3153 cp[0] = '\0'; 3154 3155 if ((error = spa_open(dataset, &spa, FTAG)) != 0) 3156 return (error); 3157 3158 spa_vers = spa_version(spa); 3159 spa_close(spa, FTAG); 3160 3161 zplver = zfs_zpl_version_map(spa_vers); 3162 fuids_ok = (zplver >= ZPL_VERSION_FUID); 3163 sa_ok = (zplver >= ZPL_VERSION_SA); 3164 3165 /* 3166 * Open parent object set so we can inherit zplprop values. 3167 */ 3168 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0) 3169 return (error); 3170 3171 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops, 3172 zplprops, is_ci); 3173 dmu_objset_rele(os, FTAG); 3174 return (error); 3175} 3176 3177static int 3178zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops, 3179 nvlist_t *zplprops, boolean_t *is_ci) 3180{ 3181 boolean_t fuids_ok; 3182 boolean_t sa_ok; 3183 uint64_t zplver = ZPL_VERSION; 3184 int error; 3185 3186 zplver = zfs_zpl_version_map(spa_vers); 3187 fuids_ok = (zplver >= ZPL_VERSION_FUID); 3188 sa_ok = (zplver >= ZPL_VERSION_SA); 3189 3190 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok, 3191 createprops, zplprops, is_ci); 3192 return (error); 3193} 3194 3195/* 3196 * innvl: { 3197 * "type" -> dmu_objset_type_t (int32) 3198 * (optional) "props" -> { prop -> value } 3199 * } 3200 * 3201 * outnvl: propname -> error code (int32) 3202 */ 3203static int 3204zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3205{ 3206 int error = 0; 3207 zfs_creat_t zct = { 0 }; 3208 nvlist_t *nvprops = NULL; 3209 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx); 3210 int32_t type32; 3211 dmu_objset_type_t type; 3212 boolean_t is_insensitive = B_FALSE; 3213 3214 if (nvlist_lookup_int32(innvl, "type", &type32) != 0) 3215 return (SET_ERROR(EINVAL)); 3216 type = type32; 3217 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops); 3218 3219 switch (type) { 3220 case DMU_OST_ZFS: 3221 cbfunc = zfs_create_cb; 3222 break; 3223 3224 case DMU_OST_ZVOL: 3225 cbfunc = zvol_create_cb; 3226 break; 3227 3228 default: 3229 cbfunc = NULL; 3230 break; 3231 } 3232 if (strchr(fsname, '@') || 3233 strchr(fsname, '%')) 3234 return (SET_ERROR(EINVAL)); 3235 3236 zct.zct_props = nvprops; 3237 3238 if (cbfunc == NULL) 3239 return (SET_ERROR(EINVAL)); 3240 3241 if (type == DMU_OST_ZVOL) { 3242 uint64_t volsize, volblocksize; 3243 3244 if (nvprops == NULL) 3245 return (SET_ERROR(EINVAL)); 3246 if (nvlist_lookup_uint64(nvprops, 3247 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0) 3248 return (SET_ERROR(EINVAL)); 3249 3250 if ((error = nvlist_lookup_uint64(nvprops, 3251 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), 3252 &volblocksize)) != 0 && error != ENOENT) 3253 return (SET_ERROR(EINVAL)); 3254 3255 if (error != 0) 3256 volblocksize = zfs_prop_default_numeric( 3257 ZFS_PROP_VOLBLOCKSIZE); 3258 3259 if ((error = zvol_check_volblocksize( 3260 volblocksize)) != 0 || 3261 (error = zvol_check_volsize(volsize, 3262 volblocksize)) != 0) 3263 return (error); 3264 } else if (type == DMU_OST_ZFS) { 3265 int error; 3266 3267 /* 3268 * We have to have normalization and 3269 * case-folding flags correct when we do the 3270 * file system creation, so go figure them out 3271 * now. 3272 */ 3273 VERIFY(nvlist_alloc(&zct.zct_zplprops, 3274 NV_UNIQUE_NAME, KM_SLEEP) == 0); 3275 error = zfs_fill_zplprops(fsname, nvprops, 3276 zct.zct_zplprops, &is_insensitive); 3277 if (error != 0) { 3278 nvlist_free(zct.zct_zplprops); 3279 return (error); 3280 } 3281 } 3282 3283 error = dmu_objset_create(fsname, type, 3284 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct); 3285 nvlist_free(zct.zct_zplprops); 3286 3287 /* 3288 * It would be nice to do this atomically. 3289 */ 3290 if (error == 0) { 3291 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL, 3292 nvprops, outnvl); 3293 if (error != 0) 3294 (void) dsl_destroy_head(fsname); 3295 } 3296#ifdef __FreeBSD__ 3297 if (error == 0 && type == DMU_OST_ZVOL) 3298 zvol_create_minors(fsname); 3299#endif 3300 return (error); 3301} 3302 3303/* 3304 * innvl: { 3305 * "origin" -> name of origin snapshot 3306 * (optional) "props" -> { prop -> value } 3307 * } 3308 * 3309 * outnvl: propname -> error code (int32) 3310 */ 3311static int 3312zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3313{ 3314 int error = 0; 3315 nvlist_t *nvprops = NULL; 3316 char *origin_name; 3317 3318 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0) 3319 return (SET_ERROR(EINVAL)); 3320 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops); 3321 3322 if (strchr(fsname, '@') || 3323 strchr(fsname, '%')) 3324 return (SET_ERROR(EINVAL)); 3325 3326 if (dataset_namecheck(origin_name, NULL, NULL) != 0) 3327 return (SET_ERROR(EINVAL)); 3328 error = dmu_objset_clone(fsname, origin_name); 3329 if (error != 0) 3330 return (error); 3331 3332 /* 3333 * It would be nice to do this atomically. 3334 */ 3335 if (error == 0) { 3336 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL, 3337 nvprops, outnvl); 3338 if (error != 0) 3339 (void) dsl_destroy_head(fsname); 3340 } 3341#ifdef __FreeBSD__ 3342 if (error == 0) 3343 zvol_create_minors(fsname); 3344#endif 3345 return (error); 3346} 3347 3348/* 3349 * innvl: { 3350 * "snaps" -> { snapshot1, snapshot2 } 3351 * (optional) "props" -> { prop -> value (string) } 3352 * } 3353 * 3354 * outnvl: snapshot -> error code (int32) 3355 */ 3356static int 3357zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3358{ 3359 nvlist_t *snaps; 3360 nvlist_t *props = NULL; 3361 int error, poollen; 3362 nvpair_t *pair; 3363 3364 (void) nvlist_lookup_nvlist(innvl, "props", &props); 3365 if ((error = zfs_check_userprops(poolname, props)) != 0) 3366 return (error); 3367 3368 if (!nvlist_empty(props) && 3369 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS)) 3370 return (SET_ERROR(ENOTSUP)); 3371 3372 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 3373 return (SET_ERROR(EINVAL)); 3374 poollen = strlen(poolname); 3375 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 3376 pair = nvlist_next_nvpair(snaps, pair)) { 3377 const char *name = nvpair_name(pair); 3378 const char *cp = strchr(name, '@'); 3379 3380 /* 3381 * The snap name must contain an @, and the part after it must 3382 * contain only valid characters. 3383 */ 3384 if (cp == NULL || 3385 zfs_component_namecheck(cp + 1, NULL, NULL) != 0) 3386 return (SET_ERROR(EINVAL)); 3387 3388 /* 3389 * The snap must be in the specified pool. 3390 */ 3391 if (strncmp(name, poolname, poollen) != 0 || 3392 (name[poollen] != '/' && name[poollen] != '@')) 3393 return (SET_ERROR(EXDEV)); 3394 3395 /* This must be the only snap of this fs. */ 3396 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair); 3397 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) { 3398 if (strncmp(name, nvpair_name(pair2), cp - name + 1) 3399 == 0) { 3400 return (SET_ERROR(EXDEV)); 3401 } 3402 } 3403 } 3404 3405 error = dsl_dataset_snapshot(snaps, props, outnvl); 3406 return (error); 3407} 3408 3409/* 3410 * innvl: "message" -> string 3411 */ 3412/* ARGSUSED */ 3413static int 3414zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl) 3415{ 3416 char *message; 3417 spa_t *spa; 3418 int error; 3419 char *poolname; 3420 3421 /* 3422 * The poolname in the ioctl is not set, we get it from the TSD, 3423 * which was set at the end of the last successful ioctl that allows 3424 * logging. The secpolicy func already checked that it is set. 3425 * Only one log ioctl is allowed after each successful ioctl, so 3426 * we clear the TSD here. 3427 */ 3428 poolname = tsd_get(zfs_allow_log_key); 3429 (void) tsd_set(zfs_allow_log_key, NULL); 3430 error = spa_open(poolname, &spa, FTAG); 3431 strfree(poolname); 3432 if (error != 0) 3433 return (error); 3434 3435 if (nvlist_lookup_string(innvl, "message", &message) != 0) { 3436 spa_close(spa, FTAG); 3437 return (SET_ERROR(EINVAL)); 3438 } 3439 3440 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) { 3441 spa_close(spa, FTAG); 3442 return (SET_ERROR(ENOTSUP)); 3443 } 3444 3445 error = spa_history_log(spa, message); 3446 spa_close(spa, FTAG); 3447 return (error); 3448} 3449 3450/* 3451 * The dp_config_rwlock must not be held when calling this, because the 3452 * unmount may need to write out data. 3453 * 3454 * This function is best-effort. Callers must deal gracefully if it 3455 * remains mounted (or is remounted after this call). 3456 * 3457 * Returns 0 if the argument is not a snapshot, or it is not currently a 3458 * filesystem, or we were able to unmount it. Returns error code otherwise. 3459 */ 3460int 3461zfs_unmount_snap(const char *snapname) 3462{ 3463 vfs_t *vfsp; 3464 zfsvfs_t *zfsvfs; 3465 int err; 3466 3467 if (strchr(snapname, '@') == NULL) 3468 return (0); 3469 3470 vfsp = zfs_get_vfs(snapname); 3471 if (vfsp == NULL) 3472 return (0); 3473 3474 zfsvfs = vfsp->vfs_data; 3475 ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os))); 3476 3477 err = vn_vfswlock(vfsp->vfs_vnodecovered); 3478 VFS_RELE(vfsp); 3479 if (err != 0) 3480 return (SET_ERROR(err)); 3481 3482 /* 3483 * Always force the unmount for snapshots. 3484 */ 3485 3486#ifdef illumos 3487 (void) dounmount(vfsp, MS_FORCE, kcred); 3488#else 3489 vfs_ref(vfsp); 3490 (void) dounmount(vfsp, MS_FORCE, curthread); 3491#endif 3492 return (0); 3493} 3494 3495/* ARGSUSED */ 3496static int 3497zfs_unmount_snap_cb(const char *snapname, void *arg) 3498{ 3499 return (zfs_unmount_snap(snapname)); 3500} 3501 3502/* 3503 * When a clone is destroyed, its origin may also need to be destroyed, 3504 * in which case it must be unmounted. This routine will do that unmount 3505 * if necessary. 3506 */ 3507void 3508zfs_destroy_unmount_origin(const char *fsname) 3509{ 3510 int error; 3511 objset_t *os; 3512 dsl_dataset_t *ds; 3513 3514 error = dmu_objset_hold(fsname, FTAG, &os); 3515 if (error != 0) 3516 return; 3517 ds = dmu_objset_ds(os); 3518 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) { 3519 char originname[MAXNAMELEN]; 3520 dsl_dataset_name(ds->ds_prev, originname); 3521 dmu_objset_rele(os, FTAG); 3522 (void) zfs_unmount_snap(originname); 3523 } else { 3524 dmu_objset_rele(os, FTAG); 3525 } 3526} 3527 3528/* 3529 * innvl: { 3530 * "snaps" -> { snapshot1, snapshot2 } 3531 * (optional boolean) "defer" 3532 * } 3533 * 3534 * outnvl: snapshot -> error code (int32) 3535 * 3536 */ 3537/* ARGSUSED */ 3538static int 3539zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3540{ 3541 int error, poollen; 3542 nvlist_t *snaps; 3543 nvpair_t *pair; 3544 boolean_t defer; 3545 3546 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 3547 return (SET_ERROR(EINVAL)); 3548 defer = nvlist_exists(innvl, "defer"); 3549 3550 poollen = strlen(poolname); 3551 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 3552 pair = nvlist_next_nvpair(snaps, pair)) { 3553 const char *name = nvpair_name(pair); 3554 3555 /* 3556 * The snap must be in the specified pool to prevent the 3557 * invalid removal of zvol minors below. 3558 */ 3559 if (strncmp(name, poolname, poollen) != 0 || 3560 (name[poollen] != '/' && name[poollen] != '@')) 3561 return (SET_ERROR(EXDEV)); 3562 3563 error = zfs_unmount_snap(name); 3564 if (error != 0) 3565 return (error); 3566#if defined(__FreeBSD__) 3567 zvol_remove_minors(name); 3568#endif 3569 } 3570 3571 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl)); 3572} 3573 3574/* 3575 * Create bookmarks. Bookmark names are of the form <fs>#<bmark>. 3576 * All bookmarks must be in the same pool. 3577 * 3578 * innvl: { 3579 * bookmark1 -> snapshot1, bookmark2 -> snapshot2 3580 * } 3581 * 3582 * outnvl: bookmark -> error code (int32) 3583 * 3584 */ 3585/* ARGSUSED */ 3586static int 3587zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3588{ 3589 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 3590 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 3591 char *snap_name; 3592 3593 /* 3594 * Verify the snapshot argument. 3595 */ 3596 if (nvpair_value_string(pair, &snap_name) != 0) 3597 return (SET_ERROR(EINVAL)); 3598 3599 3600 /* Verify that the keys (bookmarks) are unique */ 3601 for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair); 3602 pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) { 3603 if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0) 3604 return (SET_ERROR(EINVAL)); 3605 } 3606 } 3607 3608 return (dsl_bookmark_create(innvl, outnvl)); 3609} 3610 3611/* 3612 * innvl: { 3613 * property 1, property 2, ... 3614 * } 3615 * 3616 * outnvl: { 3617 * bookmark name 1 -> { property 1, property 2, ... }, 3618 * bookmark name 2 -> { property 1, property 2, ... } 3619 * } 3620 * 3621 */ 3622static int 3623zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3624{ 3625 return (dsl_get_bookmarks(fsname, innvl, outnvl)); 3626} 3627 3628/* 3629 * innvl: { 3630 * bookmark name 1, bookmark name 2 3631 * } 3632 * 3633 * outnvl: bookmark -> error code (int32) 3634 * 3635 */ 3636static int 3637zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl, 3638 nvlist_t *outnvl) 3639{ 3640 int error, poollen; 3641 3642 poollen = strlen(poolname); 3643 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 3644 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 3645 const char *name = nvpair_name(pair); 3646 const char *cp = strchr(name, '#'); 3647 3648 /* 3649 * The bookmark name must contain an #, and the part after it 3650 * must contain only valid characters. 3651 */ 3652 if (cp == NULL || 3653 zfs_component_namecheck(cp + 1, NULL, NULL) != 0) 3654 return (SET_ERROR(EINVAL)); 3655 3656 /* 3657 * The bookmark must be in the specified pool. 3658 */ 3659 if (strncmp(name, poolname, poollen) != 0 || 3660 (name[poollen] != '/' && name[poollen] != '#')) 3661 return (SET_ERROR(EXDEV)); 3662 } 3663 3664 error = dsl_bookmark_destroy(innvl, outnvl); 3665 return (error); 3666} 3667 3668/* 3669 * inputs: 3670 * zc_name name of dataset to destroy 3671 * zc_objset_type type of objset 3672 * zc_defer_destroy mark for deferred destroy 3673 * 3674 * outputs: none 3675 */ 3676static int 3677zfs_ioc_destroy(zfs_cmd_t *zc) 3678{ 3679 int err; 3680 3681 if (zc->zc_objset_type == DMU_OST_ZFS) { 3682 err = zfs_unmount_snap(zc->zc_name); 3683 if (err != 0) 3684 return (err); 3685 } 3686 3687 if (strchr(zc->zc_name, '@')) 3688 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy); 3689 else 3690 err = dsl_destroy_head(zc->zc_name); 3691 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0) 3692#ifdef __FreeBSD__ 3693 zvol_remove_minors(zc->zc_name); 3694#else 3695 (void) zvol_remove_minor(zc->zc_name); 3696#endif 3697 return (err); 3698} 3699 3700/* 3701 * fsname is name of dataset to rollback (to most recent snapshot) 3702 * 3703 * innvl is not used. 3704 * 3705 * outnvl: "target" -> name of most recent snapshot 3706 * } 3707 */ 3708/* ARGSUSED */ 3709static int 3710zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl) 3711{ 3712 zfsvfs_t *zfsvfs; 3713 int error; 3714 3715 if (getzfsvfs(fsname, &zfsvfs) == 0) { 3716 error = zfs_suspend_fs(zfsvfs); 3717 if (error == 0) { 3718 int resume_err; 3719 3720 error = dsl_dataset_rollback(fsname, zfsvfs, outnvl); 3721 resume_err = zfs_resume_fs(zfsvfs, fsname); 3722 error = error ? error : resume_err; 3723 } 3724 VFS_RELE(zfsvfs->z_vfs); 3725 } else { 3726 error = dsl_dataset_rollback(fsname, NULL, outnvl); 3727 } 3728 return (error); 3729} 3730 3731static int 3732recursive_unmount(const char *fsname, void *arg) 3733{ 3734 const char *snapname = arg; 3735 char fullname[MAXNAMELEN]; 3736 3737 (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname); 3738 return (zfs_unmount_snap(fullname)); 3739} 3740 3741/* 3742 * inputs: 3743 * zc_name old name of dataset 3744 * zc_value new name of dataset 3745 * zc_cookie recursive flag (only valid for snapshots) 3746 * 3747 * outputs: none 3748 */ 3749static int 3750zfs_ioc_rename(zfs_cmd_t *zc) 3751{ 3752 boolean_t recursive = zc->zc_cookie & 1; 3753 char *at; 3754 boolean_t allow_mounted = B_TRUE; 3755 3756#ifdef __FreeBSD__ 3757 allow_mounted = (zc->zc_cookie & 2) != 0; 3758#endif 3759 3760 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0'; 3761 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 3762 strchr(zc->zc_value, '%')) 3763 return (SET_ERROR(EINVAL)); 3764 3765 at = strchr(zc->zc_name, '@'); 3766 if (at != NULL) { 3767 /* snaps must be in same fs */ 3768 int error; 3769 3770 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1)) 3771 return (SET_ERROR(EXDEV)); 3772 *at = '\0'; 3773 if (zc->zc_objset_type == DMU_OST_ZFS && allow_mounted) { 3774 error = dmu_objset_find(zc->zc_name, 3775 recursive_unmount, at + 1, 3776 recursive ? DS_FIND_CHILDREN : 0); 3777 if (error != 0) { 3778 *at = '@'; 3779 return (error); 3780 } 3781 } 3782 error = dsl_dataset_rename_snapshot(zc->zc_name, 3783 at + 1, strchr(zc->zc_value, '@') + 1, recursive); 3784 *at = '@'; 3785 3786 return (error); 3787 } else { 3788#ifdef illumos 3789 if (zc->zc_objset_type == DMU_OST_ZVOL) 3790 (void) zvol_remove_minor(zc->zc_name); 3791#endif 3792 return (dsl_dir_rename(zc->zc_name, zc->zc_value)); 3793 } 3794} 3795 3796static int 3797zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr) 3798{ 3799 const char *propname = nvpair_name(pair); 3800 boolean_t issnap = (strchr(dsname, '@') != NULL); 3801 zfs_prop_t prop = zfs_name_to_prop(propname); 3802 uint64_t intval; 3803 int err; 3804 3805 if (prop == ZPROP_INVAL) { 3806 if (zfs_prop_user(propname)) { 3807 if (err = zfs_secpolicy_write_perms(dsname, 3808 ZFS_DELEG_PERM_USERPROP, cr)) 3809 return (err); 3810 return (0); 3811 } 3812 3813 if (!issnap && zfs_prop_userquota(propname)) { 3814 const char *perm = NULL; 3815 const char *uq_prefix = 3816 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA]; 3817 const char *gq_prefix = 3818 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA]; 3819 3820 if (strncmp(propname, uq_prefix, 3821 strlen(uq_prefix)) == 0) { 3822 perm = ZFS_DELEG_PERM_USERQUOTA; 3823 } else if (strncmp(propname, gq_prefix, 3824 strlen(gq_prefix)) == 0) { 3825 perm = ZFS_DELEG_PERM_GROUPQUOTA; 3826 } else { 3827 /* USERUSED and GROUPUSED are read-only */ 3828 return (SET_ERROR(EINVAL)); 3829 } 3830 3831 if (err = zfs_secpolicy_write_perms(dsname, perm, cr)) 3832 return (err); 3833 return (0); 3834 } 3835 3836 return (SET_ERROR(EINVAL)); 3837 } 3838 3839 if (issnap) 3840 return (SET_ERROR(EINVAL)); 3841 3842 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 3843 /* 3844 * dsl_prop_get_all_impl() returns properties in this 3845 * format. 3846 */ 3847 nvlist_t *attrs; 3848 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 3849 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 3850 &pair) == 0); 3851 } 3852 3853 /* 3854 * Check that this value is valid for this pool version 3855 */ 3856 switch (prop) { 3857 case ZFS_PROP_COMPRESSION: 3858 /* 3859 * If the user specified gzip compression, make sure 3860 * the SPA supports it. We ignore any errors here since 3861 * we'll catch them later. 3862 */ 3863 if (nvpair_value_uint64(pair, &intval) == 0) { 3864 if (intval >= ZIO_COMPRESS_GZIP_1 && 3865 intval <= ZIO_COMPRESS_GZIP_9 && 3866 zfs_earlier_version(dsname, 3867 SPA_VERSION_GZIP_COMPRESSION)) { 3868 return (SET_ERROR(ENOTSUP)); 3869 } 3870 3871 if (intval == ZIO_COMPRESS_ZLE && 3872 zfs_earlier_version(dsname, 3873 SPA_VERSION_ZLE_COMPRESSION)) 3874 return (SET_ERROR(ENOTSUP)); 3875 3876 if (intval == ZIO_COMPRESS_LZ4) { 3877 spa_t *spa; 3878 3879 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 3880 return (err); 3881 3882 if (!spa_feature_is_enabled(spa, 3883 SPA_FEATURE_LZ4_COMPRESS)) { 3884 spa_close(spa, FTAG); 3885 return (SET_ERROR(ENOTSUP)); 3886 } 3887 spa_close(spa, FTAG); 3888 } 3889 3890 /* 3891 * If this is a bootable dataset then 3892 * verify that the compression algorithm 3893 * is supported for booting. We must return 3894 * something other than ENOTSUP since it 3895 * implies a downrev pool version. 3896 */ 3897 if (zfs_is_bootfs(dsname) && 3898 !BOOTFS_COMPRESS_VALID(intval)) { 3899 return (SET_ERROR(ERANGE)); 3900 } 3901 } 3902 break; 3903 3904 case ZFS_PROP_COPIES: 3905 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS)) 3906 return (SET_ERROR(ENOTSUP)); 3907 break; 3908 3909 case ZFS_PROP_RECORDSIZE: 3910 /* Record sizes above 128k need the feature to be enabled */ 3911 if (nvpair_value_uint64(pair, &intval) == 0 && 3912 intval > SPA_OLD_MAXBLOCKSIZE) { 3913 spa_t *spa; 3914 3915 /* 3916 * If this is a bootable dataset then 3917 * the we don't allow large (>128K) blocks, 3918 * because GRUB doesn't support them. 3919 */ 3920 if (zfs_is_bootfs(dsname) && 3921 intval > SPA_OLD_MAXBLOCKSIZE) { 3922 return (SET_ERROR(ERANGE)); 3923 } 3924 3925 /* 3926 * We don't allow setting the property above 1MB, 3927 * unless the tunable has been changed. 3928 */ 3929 if (intval > zfs_max_recordsize || 3930 intval > SPA_MAXBLOCKSIZE) 3931 return (SET_ERROR(ERANGE)); 3932 3933 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 3934 return (err); 3935 3936 if (!spa_feature_is_enabled(spa, 3937 SPA_FEATURE_LARGE_BLOCKS)) { 3938 spa_close(spa, FTAG); 3939 return (SET_ERROR(ENOTSUP)); 3940 } 3941 spa_close(spa, FTAG); 3942 } 3943 break; 3944 3945 case ZFS_PROP_SHARESMB: 3946 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID)) 3947 return (SET_ERROR(ENOTSUP)); 3948 break; 3949 3950 case ZFS_PROP_ACLINHERIT: 3951 if (nvpair_type(pair) == DATA_TYPE_UINT64 && 3952 nvpair_value_uint64(pair, &intval) == 0) { 3953 if (intval == ZFS_ACL_PASSTHROUGH_X && 3954 zfs_earlier_version(dsname, 3955 SPA_VERSION_PASSTHROUGH_X)) 3956 return (SET_ERROR(ENOTSUP)); 3957 } 3958 break; 3959 3960 case ZFS_PROP_CHECKSUM: 3961 case ZFS_PROP_DEDUP: 3962 { 3963 spa_feature_t feature; 3964 spa_t *spa; 3965 3966 /* dedup feature version checks */ 3967 if (prop == ZFS_PROP_DEDUP && 3968 zfs_earlier_version(dsname, SPA_VERSION_DEDUP)) 3969 return (SET_ERROR(ENOTSUP)); 3970 3971 if (nvpair_value_uint64(pair, &intval) != 0) 3972 return (SET_ERROR(EINVAL)); 3973 3974 /* check prop value is enabled in features */ 3975 feature = zio_checksum_to_feature(intval & ZIO_CHECKSUM_MASK); 3976 if (feature == SPA_FEATURE_NONE) 3977 break; 3978 3979 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 3980 return (err); 3981 /* 3982 * Salted checksums are not supported on root pools. 3983 */ 3984 if (spa_bootfs(spa) != 0 && 3985 intval < ZIO_CHECKSUM_FUNCTIONS && 3986 (zio_checksum_table[intval].ci_flags & 3987 ZCHECKSUM_FLAG_SALTED)) { 3988 spa_close(spa, FTAG); 3989 return (SET_ERROR(ERANGE)); 3990 } 3991 if (!spa_feature_is_enabled(spa, feature)) { 3992 spa_close(spa, FTAG); 3993 return (SET_ERROR(ENOTSUP)); 3994 } 3995 spa_close(spa, FTAG); 3996 break; 3997 } 3998 } 3999 4000 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED())); 4001} 4002 4003/* 4004 * Checks for a race condition to make sure we don't increment a feature flag 4005 * multiple times. 4006 */ 4007static int 4008zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx) 4009{ 4010 spa_t *spa = dmu_tx_pool(tx)->dp_spa; 4011 spa_feature_t *featurep = arg; 4012 4013 if (!spa_feature_is_active(spa, *featurep)) 4014 return (0); 4015 else 4016 return (SET_ERROR(EBUSY)); 4017} 4018 4019/* 4020 * The callback invoked on feature activation in the sync task caused by 4021 * zfs_prop_activate_feature. 4022 */ 4023static void 4024zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx) 4025{ 4026 spa_t *spa = dmu_tx_pool(tx)->dp_spa; 4027 spa_feature_t *featurep = arg; 4028 4029 spa_feature_incr(spa, *featurep, tx); 4030} 4031 4032/* 4033 * Activates a feature on a pool in response to a property setting. This 4034 * creates a new sync task which modifies the pool to reflect the feature 4035 * as being active. 4036 */ 4037static int 4038zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature) 4039{ 4040 int err; 4041 4042 /* EBUSY here indicates that the feature is already active */ 4043 err = dsl_sync_task(spa_name(spa), 4044 zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync, 4045 &feature, 2, ZFS_SPACE_CHECK_RESERVED); 4046 4047 if (err != 0 && err != EBUSY) 4048 return (err); 4049 else 4050 return (0); 4051} 4052 4053/* 4054 * Removes properties from the given props list that fail permission checks 4055 * needed to clear them and to restore them in case of a receive error. For each 4056 * property, make sure we have both set and inherit permissions. 4057 * 4058 * Returns the first error encountered if any permission checks fail. If the 4059 * caller provides a non-NULL errlist, it also gives the complete list of names 4060 * of all the properties that failed a permission check along with the 4061 * corresponding error numbers. The caller is responsible for freeing the 4062 * returned errlist. 4063 * 4064 * If every property checks out successfully, zero is returned and the list 4065 * pointed at by errlist is NULL. 4066 */ 4067static int 4068zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist) 4069{ 4070 zfs_cmd_t *zc; 4071 nvpair_t *pair, *next_pair; 4072 nvlist_t *errors; 4073 int err, rv = 0; 4074 4075 if (props == NULL) 4076 return (0); 4077 4078 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0); 4079 4080 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP); 4081 (void) strcpy(zc->zc_name, dataset); 4082 pair = nvlist_next_nvpair(props, NULL); 4083 while (pair != NULL) { 4084 next_pair = nvlist_next_nvpair(props, pair); 4085 4086 (void) strcpy(zc->zc_value, nvpair_name(pair)); 4087 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 || 4088 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) { 4089 VERIFY(nvlist_remove_nvpair(props, pair) == 0); 4090 VERIFY(nvlist_add_int32(errors, 4091 zc->zc_value, err) == 0); 4092 } 4093 pair = next_pair; 4094 } 4095 kmem_free(zc, sizeof (zfs_cmd_t)); 4096 4097 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) { 4098 nvlist_free(errors); 4099 errors = NULL; 4100 } else { 4101 VERIFY(nvpair_value_int32(pair, &rv) == 0); 4102 } 4103 4104 if (errlist == NULL) 4105 nvlist_free(errors); 4106 else 4107 *errlist = errors; 4108 4109 return (rv); 4110} 4111 4112static boolean_t 4113propval_equals(nvpair_t *p1, nvpair_t *p2) 4114{ 4115 if (nvpair_type(p1) == DATA_TYPE_NVLIST) { 4116 /* dsl_prop_get_all_impl() format */ 4117 nvlist_t *attrs; 4118 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0); 4119 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 4120 &p1) == 0); 4121 } 4122 4123 if (nvpair_type(p2) == DATA_TYPE_NVLIST) { 4124 nvlist_t *attrs; 4125 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0); 4126 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 4127 &p2) == 0); 4128 } 4129 4130 if (nvpair_type(p1) != nvpair_type(p2)) 4131 return (B_FALSE); 4132 4133 if (nvpair_type(p1) == DATA_TYPE_STRING) { 4134 char *valstr1, *valstr2; 4135 4136 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0); 4137 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0); 4138 return (strcmp(valstr1, valstr2) == 0); 4139 } else { 4140 uint64_t intval1, intval2; 4141 4142 VERIFY(nvpair_value_uint64(p1, &intval1) == 0); 4143 VERIFY(nvpair_value_uint64(p2, &intval2) == 0); 4144 return (intval1 == intval2); 4145 } 4146} 4147 4148/* 4149 * Remove properties from props if they are not going to change (as determined 4150 * by comparison with origprops). Remove them from origprops as well, since we 4151 * do not need to clear or restore properties that won't change. 4152 */ 4153static void 4154props_reduce(nvlist_t *props, nvlist_t *origprops) 4155{ 4156 nvpair_t *pair, *next_pair; 4157 4158 if (origprops == NULL) 4159 return; /* all props need to be received */ 4160 4161 pair = nvlist_next_nvpair(props, NULL); 4162 while (pair != NULL) { 4163 const char *propname = nvpair_name(pair); 4164 nvpair_t *match; 4165 4166 next_pair = nvlist_next_nvpair(props, pair); 4167 4168 if ((nvlist_lookup_nvpair(origprops, propname, 4169 &match) != 0) || !propval_equals(pair, match)) 4170 goto next; /* need to set received value */ 4171 4172 /* don't clear the existing received value */ 4173 (void) nvlist_remove_nvpair(origprops, match); 4174 /* don't bother receiving the property */ 4175 (void) nvlist_remove_nvpair(props, pair); 4176next: 4177 pair = next_pair; 4178 } 4179} 4180 4181/* 4182 * Extract properties that cannot be set PRIOR to the receipt of a dataset. 4183 * For example, refquota cannot be set until after the receipt of a dataset, 4184 * because in replication streams, an older/earlier snapshot may exceed the 4185 * refquota. We want to receive the older/earlier snapshot, but setting 4186 * refquota pre-receipt will set the dsl's ACTUAL quota, which will prevent 4187 * the older/earlier snapshot from being received (with EDQUOT). 4188 * 4189 * The ZFS test "zfs_receive_011_pos" demonstrates such a scenario. 4190 * 4191 * libzfs will need to be judicious handling errors encountered by props 4192 * extracted by this function. 4193 */ 4194static nvlist_t * 4195extract_delay_props(nvlist_t *props) 4196{ 4197 nvlist_t *delayprops; 4198 nvpair_t *nvp, *tmp; 4199 static const zfs_prop_t delayable[] = { ZFS_PROP_REFQUOTA, 0 }; 4200 int i; 4201 4202 VERIFY(nvlist_alloc(&delayprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 4203 4204 for (nvp = nvlist_next_nvpair(props, NULL); nvp != NULL; 4205 nvp = nvlist_next_nvpair(props, nvp)) { 4206 /* 4207 * strcmp() is safe because zfs_prop_to_name() always returns 4208 * a bounded string. 4209 */ 4210 for (i = 0; delayable[i] != 0; i++) { 4211 if (strcmp(zfs_prop_to_name(delayable[i]), 4212 nvpair_name(nvp)) == 0) { 4213 break; 4214 } 4215 } 4216 if (delayable[i] != 0) { 4217 tmp = nvlist_prev_nvpair(props, nvp); 4218 VERIFY(nvlist_add_nvpair(delayprops, nvp) == 0); 4219 VERIFY(nvlist_remove_nvpair(props, nvp) == 0); 4220 nvp = tmp; 4221 } 4222 } 4223 4224 if (nvlist_empty(delayprops)) { 4225 nvlist_free(delayprops); 4226 delayprops = NULL; 4227 } 4228 return (delayprops); 4229} 4230 4231#ifdef DEBUG 4232static boolean_t zfs_ioc_recv_inject_err; 4233#endif 4234 4235/* 4236 * inputs: 4237 * zc_name name of containing filesystem 4238 * zc_nvlist_src{_size} nvlist of properties to apply 4239 * zc_value name of snapshot to create 4240 * zc_string name of clone origin (if DRR_FLAG_CLONE) 4241 * zc_cookie file descriptor to recv from 4242 * zc_begin_record the BEGIN record of the stream (not byteswapped) 4243 * zc_guid force flag 4244 * zc_cleanup_fd cleanup-on-exit file descriptor 4245 * zc_action_handle handle for this guid/ds mapping (or zero on first call) 4246 * zc_resumable if data is incomplete assume sender will resume 4247 * 4248 * outputs: 4249 * zc_cookie number of bytes read 4250 * zc_nvlist_dst{_size} error for each unapplied received property 4251 * zc_obj zprop_errflags_t 4252 * zc_action_handle handle for this guid/ds mapping 4253 */ 4254static int 4255zfs_ioc_recv(zfs_cmd_t *zc) 4256{ 4257 file_t *fp; 4258 dmu_recv_cookie_t drc; 4259 boolean_t force = (boolean_t)zc->zc_guid; 4260 int fd; 4261 int error = 0; 4262 int props_error = 0; 4263 nvlist_t *errors; 4264 offset_t off; 4265 nvlist_t *props = NULL; /* sent properties */ 4266 nvlist_t *origprops = NULL; /* existing properties */ 4267 nvlist_t *delayprops = NULL; /* sent properties applied post-receive */ 4268 char *origin = NULL; 4269 char *tosnap; 4270 char tofs[ZFS_MAXNAMELEN]; 4271 cap_rights_t rights; 4272 boolean_t first_recvd_props = B_FALSE; 4273 4274 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 4275 strchr(zc->zc_value, '@') == NULL || 4276 strchr(zc->zc_value, '%')) 4277 return (SET_ERROR(EINVAL)); 4278 4279 (void) strcpy(tofs, zc->zc_value); 4280 tosnap = strchr(tofs, '@'); 4281 *tosnap++ = '\0'; 4282 4283 if (zc->zc_nvlist_src != 0 && 4284 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 4285 zc->zc_iflags, &props)) != 0) 4286 return (error); 4287 4288 fd = zc->zc_cookie; 4289#ifdef illumos 4290 fp = getf(fd); 4291#else 4292 fget_read(curthread, fd, cap_rights_init(&rights, CAP_PREAD), &fp); 4293#endif 4294 if (fp == NULL) { 4295 nvlist_free(props); 4296 return (SET_ERROR(EBADF)); 4297 } 4298 4299 errors = fnvlist_alloc(); 4300 4301 if (zc->zc_string[0]) 4302 origin = zc->zc_string; 4303 4304 error = dmu_recv_begin(tofs, tosnap, 4305 &zc->zc_begin_record, force, zc->zc_resumable, origin, &drc); 4306 if (error != 0) 4307 goto out; 4308 4309 /* 4310 * Set properties before we receive the stream so that they are applied 4311 * to the new data. Note that we must call dmu_recv_stream() if 4312 * dmu_recv_begin() succeeds. 4313 */ 4314 if (props != NULL && !drc.drc_newfs) { 4315 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >= 4316 SPA_VERSION_RECVD_PROPS && 4317 !dsl_prop_get_hasrecvd(tofs)) 4318 first_recvd_props = B_TRUE; 4319 4320 /* 4321 * If new received properties are supplied, they are to 4322 * completely replace the existing received properties, so stash 4323 * away the existing ones. 4324 */ 4325 if (dsl_prop_get_received(tofs, &origprops) == 0) { 4326 nvlist_t *errlist = NULL; 4327 /* 4328 * Don't bother writing a property if its value won't 4329 * change (and avoid the unnecessary security checks). 4330 * 4331 * The first receive after SPA_VERSION_RECVD_PROPS is a 4332 * special case where we blow away all local properties 4333 * regardless. 4334 */ 4335 if (!first_recvd_props) 4336 props_reduce(props, origprops); 4337 if (zfs_check_clearable(tofs, origprops, &errlist) != 0) 4338 (void) nvlist_merge(errors, errlist, 0); 4339 nvlist_free(errlist); 4340 4341 if (clear_received_props(tofs, origprops, 4342 first_recvd_props ? NULL : props) != 0) 4343 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 4344 } else { 4345 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 4346 } 4347 } 4348 4349 if (props != NULL) { 4350 props_error = dsl_prop_set_hasrecvd(tofs); 4351 4352 if (props_error == 0) { 4353 delayprops = extract_delay_props(props); 4354 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED, 4355 props, errors); 4356 } 4357 } 4358 4359 off = fp->f_offset; 4360 error = dmu_recv_stream(&drc, fp, &off, zc->zc_cleanup_fd, 4361 &zc->zc_action_handle); 4362 4363 if (error == 0) { 4364 zfsvfs_t *zfsvfs = NULL; 4365 4366 if (getzfsvfs(tofs, &zfsvfs) == 0) { 4367 /* online recv */ 4368 int end_err; 4369 4370 error = zfs_suspend_fs(zfsvfs); 4371 /* 4372 * If the suspend fails, then the recv_end will 4373 * likely also fail, and clean up after itself. 4374 */ 4375 end_err = dmu_recv_end(&drc, zfsvfs); 4376 if (error == 0) 4377 error = zfs_resume_fs(zfsvfs, tofs); 4378 error = error ? error : end_err; 4379 VFS_RELE(zfsvfs->z_vfs); 4380 } else { 4381 error = dmu_recv_end(&drc, NULL); 4382 } 4383 4384 /* Set delayed properties now, after we're done receiving. */ 4385 if (delayprops != NULL && error == 0) { 4386 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED, 4387 delayprops, errors); 4388 } 4389 } 4390 4391 if (delayprops != NULL) { 4392 /* 4393 * Merge delayed props back in with initial props, in case 4394 * we're DEBUG and zfs_ioc_recv_inject_err is set (which means 4395 * we have to make sure clear_received_props() includes 4396 * the delayed properties). 4397 * 4398 * Since zfs_ioc_recv_inject_err is only in DEBUG kernels, 4399 * using ASSERT() will be just like a VERIFY. 4400 */ 4401 ASSERT(nvlist_merge(props, delayprops, 0) == 0); 4402 nvlist_free(delayprops); 4403 } 4404 4405 /* 4406 * Now that all props, initial and delayed, are set, report the prop 4407 * errors to the caller. 4408 */ 4409 if (zc->zc_nvlist_dst_size != 0 && 4410 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 || 4411 put_nvlist(zc, errors) != 0)) { 4412 /* 4413 * Caller made zc->zc_nvlist_dst less than the minimum expected 4414 * size or supplied an invalid address. 4415 */ 4416 props_error = SET_ERROR(EINVAL); 4417 } 4418 4419 zc->zc_cookie = off - fp->f_offset; 4420 if (off >= 0 && off <= MAXOFFSET_T) 4421 fp->f_offset = off; 4422 4423#ifdef DEBUG 4424 if (zfs_ioc_recv_inject_err) { 4425 zfs_ioc_recv_inject_err = B_FALSE; 4426 error = 1; 4427 } 4428#endif 4429 4430#ifdef __FreeBSD__ 4431 if (error == 0) 4432 zvol_create_minors(tofs); 4433#endif 4434 4435 /* 4436 * On error, restore the original props. 4437 */ 4438 if (error != 0 && props != NULL && !drc.drc_newfs) { 4439 if (clear_received_props(tofs, props, NULL) != 0) { 4440 /* 4441 * We failed to clear the received properties. 4442 * Since we may have left a $recvd value on the 4443 * system, we can't clear the $hasrecvd flag. 4444 */ 4445 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4446 } else if (first_recvd_props) { 4447 dsl_prop_unset_hasrecvd(tofs); 4448 } 4449 4450 if (origprops == NULL && !drc.drc_newfs) { 4451 /* We failed to stash the original properties. */ 4452 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4453 } 4454 4455 /* 4456 * dsl_props_set() will not convert RECEIVED to LOCAL on or 4457 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL 4458 * explictly if we're restoring local properties cleared in the 4459 * first new-style receive. 4460 */ 4461 if (origprops != NULL && 4462 zfs_set_prop_nvlist(tofs, (first_recvd_props ? 4463 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED), 4464 origprops, NULL) != 0) { 4465 /* 4466 * We stashed the original properties but failed to 4467 * restore them. 4468 */ 4469 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4470 } 4471 } 4472out: 4473 nvlist_free(props); 4474 nvlist_free(origprops); 4475 nvlist_free(errors); 4476 releasef(fd); 4477 4478 if (error == 0) 4479 error = props_error; 4480 4481 return (error); 4482} 4483 4484/* 4485 * inputs: 4486 * zc_name name of snapshot to send 4487 * zc_cookie file descriptor to send stream to 4488 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj) 4489 * zc_sendobj objsetid of snapshot to send 4490 * zc_fromobj objsetid of incremental fromsnap (may be zero) 4491 * zc_guid if set, estimate size of stream only. zc_cookie is ignored. 4492 * output size in zc_objset_type. 4493 * zc_flags lzc_send_flags 4494 * 4495 * outputs: 4496 * zc_objset_type estimated size, if zc_guid is set 4497 */ 4498static int 4499zfs_ioc_send(zfs_cmd_t *zc) 4500{ 4501 int error; 4502 offset_t off; 4503 boolean_t estimate = (zc->zc_guid != 0); 4504 boolean_t embedok = (zc->zc_flags & 0x1); 4505 boolean_t large_block_ok = (zc->zc_flags & 0x2); 4506 4507 if (zc->zc_obj != 0) { 4508 dsl_pool_t *dp; 4509 dsl_dataset_t *tosnap; 4510 4511 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4512 if (error != 0) 4513 return (error); 4514 4515 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap); 4516 if (error != 0) { 4517 dsl_pool_rele(dp, FTAG); 4518 return (error); 4519 } 4520 4521 if (dsl_dir_is_clone(tosnap->ds_dir)) 4522 zc->zc_fromobj = 4523 dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj; 4524 dsl_dataset_rele(tosnap, FTAG); 4525 dsl_pool_rele(dp, FTAG); 4526 } 4527 4528 if (estimate) { 4529 dsl_pool_t *dp; 4530 dsl_dataset_t *tosnap; 4531 dsl_dataset_t *fromsnap = NULL; 4532 4533 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4534 if (error != 0) 4535 return (error); 4536 4537 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap); 4538 if (error != 0) { 4539 dsl_pool_rele(dp, FTAG); 4540 return (error); 4541 } 4542 4543 if (zc->zc_fromobj != 0) { 4544 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj, 4545 FTAG, &fromsnap); 4546 if (error != 0) { 4547 dsl_dataset_rele(tosnap, FTAG); 4548 dsl_pool_rele(dp, FTAG); 4549 return (error); 4550 } 4551 } 4552 4553 error = dmu_send_estimate(tosnap, fromsnap, 4554 &zc->zc_objset_type); 4555 4556 if (fromsnap != NULL) 4557 dsl_dataset_rele(fromsnap, FTAG); 4558 dsl_dataset_rele(tosnap, FTAG); 4559 dsl_pool_rele(dp, FTAG); 4560 } else { 4561 file_t *fp; 4562 cap_rights_t rights; 4563 4564#ifdef illumos 4565 fp = getf(zc->zc_cookie); 4566#else 4567 fget_write(curthread, zc->zc_cookie, 4568 cap_rights_init(&rights, CAP_WRITE), &fp); 4569#endif 4570 if (fp == NULL) 4571 return (SET_ERROR(EBADF)); 4572 4573 off = fp->f_offset; 4574 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj, 4575 zc->zc_fromobj, embedok, large_block_ok, 4576#ifdef illumos 4577 zc->zc_cookie, fp->f_vnode, &off); 4578#else 4579 zc->zc_cookie, fp, &off); 4580#endif 4581 4582 if (off >= 0 && off <= MAXOFFSET_T) 4583 fp->f_offset = off; 4584 releasef(zc->zc_cookie); 4585 } 4586 return (error); 4587} 4588 4589/* 4590 * inputs: 4591 * zc_name name of snapshot on which to report progress 4592 * zc_cookie file descriptor of send stream 4593 * 4594 * outputs: 4595 * zc_cookie number of bytes written in send stream thus far 4596 */ 4597static int 4598zfs_ioc_send_progress(zfs_cmd_t *zc) 4599{ 4600 dsl_pool_t *dp; 4601 dsl_dataset_t *ds; 4602 dmu_sendarg_t *dsp = NULL; 4603 int error; 4604 4605 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4606 if (error != 0) 4607 return (error); 4608 4609 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds); 4610 if (error != 0) { 4611 dsl_pool_rele(dp, FTAG); 4612 return (error); 4613 } 4614 4615 mutex_enter(&ds->ds_sendstream_lock); 4616 4617 /* 4618 * Iterate over all the send streams currently active on this dataset. 4619 * If there's one which matches the specified file descriptor _and_ the 4620 * stream was started by the current process, return the progress of 4621 * that stream. 4622 */ 4623 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL; 4624 dsp = list_next(&ds->ds_sendstreams, dsp)) { 4625 if (dsp->dsa_outfd == zc->zc_cookie && 4626 dsp->dsa_proc == curproc) 4627 break; 4628 } 4629 4630 if (dsp != NULL) 4631 zc->zc_cookie = *(dsp->dsa_off); 4632 else 4633 error = SET_ERROR(ENOENT); 4634 4635 mutex_exit(&ds->ds_sendstream_lock); 4636 dsl_dataset_rele(ds, FTAG); 4637 dsl_pool_rele(dp, FTAG); 4638 return (error); 4639} 4640 4641static int 4642zfs_ioc_inject_fault(zfs_cmd_t *zc) 4643{ 4644 int id, error; 4645 4646 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id, 4647 &zc->zc_inject_record); 4648 4649 if (error == 0) 4650 zc->zc_guid = (uint64_t)id; 4651 4652 return (error); 4653} 4654 4655static int 4656zfs_ioc_clear_fault(zfs_cmd_t *zc) 4657{ 4658 return (zio_clear_fault((int)zc->zc_guid)); 4659} 4660 4661static int 4662zfs_ioc_inject_list_next(zfs_cmd_t *zc) 4663{ 4664 int id = (int)zc->zc_guid; 4665 int error; 4666 4667 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name), 4668 &zc->zc_inject_record); 4669 4670 zc->zc_guid = id; 4671 4672 return (error); 4673} 4674 4675static int 4676zfs_ioc_error_log(zfs_cmd_t *zc) 4677{ 4678 spa_t *spa; 4679 int error; 4680 size_t count = (size_t)zc->zc_nvlist_dst_size; 4681 4682 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 4683 return (error); 4684 4685 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst, 4686 &count); 4687 if (error == 0) 4688 zc->zc_nvlist_dst_size = count; 4689 else 4690 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa); 4691 4692 spa_close(spa, FTAG); 4693 4694 return (error); 4695} 4696 4697static int 4698zfs_ioc_clear(zfs_cmd_t *zc) 4699{ 4700 spa_t *spa; 4701 vdev_t *vd; 4702 int error; 4703 4704 /* 4705 * On zpool clear we also fix up missing slogs 4706 */ 4707 mutex_enter(&spa_namespace_lock); 4708 spa = spa_lookup(zc->zc_name); 4709 if (spa == NULL) { 4710 mutex_exit(&spa_namespace_lock); 4711 return (SET_ERROR(EIO)); 4712 } 4713 if (spa_get_log_state(spa) == SPA_LOG_MISSING) { 4714 /* we need to let spa_open/spa_load clear the chains */ 4715 spa_set_log_state(spa, SPA_LOG_CLEAR); 4716 } 4717 spa->spa_last_open_failed = 0; 4718 mutex_exit(&spa_namespace_lock); 4719 4720 if (zc->zc_cookie & ZPOOL_NO_REWIND) { 4721 error = spa_open(zc->zc_name, &spa, FTAG); 4722 } else { 4723 nvlist_t *policy; 4724 nvlist_t *config = NULL; 4725 4726 if (zc->zc_nvlist_src == 0) 4727 return (SET_ERROR(EINVAL)); 4728 4729 if ((error = get_nvlist(zc->zc_nvlist_src, 4730 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) { 4731 error = spa_open_rewind(zc->zc_name, &spa, FTAG, 4732 policy, &config); 4733 if (config != NULL) { 4734 int err; 4735 4736 if ((err = put_nvlist(zc, config)) != 0) 4737 error = err; 4738 nvlist_free(config); 4739 } 4740 nvlist_free(policy); 4741 } 4742 } 4743 4744 if (error != 0) 4745 return (error); 4746 4747 spa_vdev_state_enter(spa, SCL_NONE); 4748 4749 if (zc->zc_guid == 0) { 4750 vd = NULL; 4751 } else { 4752 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE); 4753 if (vd == NULL) { 4754 (void) spa_vdev_state_exit(spa, NULL, ENODEV); 4755 spa_close(spa, FTAG); 4756 return (SET_ERROR(ENODEV)); 4757 } 4758 } 4759 4760 vdev_clear(spa, vd); 4761 4762 (void) spa_vdev_state_exit(spa, NULL, 0); 4763 4764 /* 4765 * Resume any suspended I/Os. 4766 */ 4767 if (zio_resume(spa) != 0) 4768 error = SET_ERROR(EIO); 4769 4770 spa_close(spa, FTAG); 4771 4772 return (error); 4773} 4774 4775static int 4776zfs_ioc_pool_reopen(zfs_cmd_t *zc) 4777{ 4778 spa_t *spa; 4779 int error; 4780 4781 error = spa_open(zc->zc_name, &spa, FTAG); 4782 if (error != 0) 4783 return (error); 4784 4785 spa_vdev_state_enter(spa, SCL_NONE); 4786 4787 /* 4788 * If a resilver is already in progress then set the 4789 * spa_scrub_reopen flag to B_TRUE so that we don't restart 4790 * the scan as a side effect of the reopen. Otherwise, let 4791 * vdev_open() decided if a resilver is required. 4792 */ 4793 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool); 4794 vdev_reopen(spa->spa_root_vdev); 4795 spa->spa_scrub_reopen = B_FALSE; 4796 4797 (void) spa_vdev_state_exit(spa, NULL, 0); 4798 spa_close(spa, FTAG); 4799 return (0); 4800} 4801/* 4802 * inputs: 4803 * zc_name name of filesystem 4804 * zc_value name of origin snapshot 4805 * 4806 * outputs: 4807 * zc_string name of conflicting snapshot, if there is one 4808 */ 4809static int 4810zfs_ioc_promote(zfs_cmd_t *zc) 4811{ 4812 char *cp; 4813 4814 /* 4815 * We don't need to unmount *all* the origin fs's snapshots, but 4816 * it's easier. 4817 */ 4818 cp = strchr(zc->zc_value, '@'); 4819 if (cp) 4820 *cp = '\0'; 4821 (void) dmu_objset_find(zc->zc_value, 4822 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS); 4823 return (dsl_dataset_promote(zc->zc_name, zc->zc_string)); 4824} 4825 4826/* 4827 * Retrieve a single {user|group}{used|quota}@... property. 4828 * 4829 * inputs: 4830 * zc_name name of filesystem 4831 * zc_objset_type zfs_userquota_prop_t 4832 * zc_value domain name (eg. "S-1-234-567-89") 4833 * zc_guid RID/UID/GID 4834 * 4835 * outputs: 4836 * zc_cookie property value 4837 */ 4838static int 4839zfs_ioc_userspace_one(zfs_cmd_t *zc) 4840{ 4841 zfsvfs_t *zfsvfs; 4842 int error; 4843 4844 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 4845 return (SET_ERROR(EINVAL)); 4846 4847 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE); 4848 if (error != 0) 4849 return (error); 4850 4851 error = zfs_userspace_one(zfsvfs, 4852 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie); 4853 zfsvfs_rele(zfsvfs, FTAG); 4854 4855 return (error); 4856} 4857 4858/* 4859 * inputs: 4860 * zc_name name of filesystem 4861 * zc_cookie zap cursor 4862 * zc_objset_type zfs_userquota_prop_t 4863 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist) 4864 * 4865 * outputs: 4866 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t) 4867 * zc_cookie zap cursor 4868 */ 4869static int 4870zfs_ioc_userspace_many(zfs_cmd_t *zc) 4871{ 4872 zfsvfs_t *zfsvfs; 4873 int bufsize = zc->zc_nvlist_dst_size; 4874 4875 if (bufsize <= 0) 4876 return (SET_ERROR(ENOMEM)); 4877 4878 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE); 4879 if (error != 0) 4880 return (error); 4881 4882 void *buf = kmem_alloc(bufsize, KM_SLEEP); 4883 4884 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie, 4885 buf, &zc->zc_nvlist_dst_size); 4886 4887 if (error == 0) { 4888 error = ddi_copyout(buf, 4889 (void *)(uintptr_t)zc->zc_nvlist_dst, 4890 zc->zc_nvlist_dst_size, zc->zc_iflags); 4891 } 4892 kmem_free(buf, bufsize); 4893 zfsvfs_rele(zfsvfs, FTAG); 4894 4895 return (error); 4896} 4897 4898/* 4899 * inputs: 4900 * zc_name name of filesystem 4901 * 4902 * outputs: 4903 * none 4904 */ 4905static int 4906zfs_ioc_userspace_upgrade(zfs_cmd_t *zc) 4907{ 4908 objset_t *os; 4909 int error = 0; 4910 zfsvfs_t *zfsvfs; 4911 4912 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) { 4913 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) { 4914 /* 4915 * If userused is not enabled, it may be because the 4916 * objset needs to be closed & reopened (to grow the 4917 * objset_phys_t). Suspend/resume the fs will do that. 4918 */ 4919 error = zfs_suspend_fs(zfsvfs); 4920 if (error == 0) { 4921 dmu_objset_refresh_ownership(zfsvfs->z_os, 4922 zfsvfs); 4923 error = zfs_resume_fs(zfsvfs, zc->zc_name); 4924 } 4925 } 4926 if (error == 0) 4927 error = dmu_objset_userspace_upgrade(zfsvfs->z_os); 4928 VFS_RELE(zfsvfs->z_vfs); 4929 } else { 4930 /* XXX kind of reading contents without owning */ 4931 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 4932 if (error != 0) 4933 return (error); 4934 4935 error = dmu_objset_userspace_upgrade(os); 4936 dmu_objset_rele(os, FTAG); 4937 } 4938 4939 return (error); 4940} 4941 4942#ifdef illumos 4943/* 4944 * We don't want to have a hard dependency 4945 * against some special symbols in sharefs 4946 * nfs, and smbsrv. Determine them if needed when 4947 * the first file system is shared. 4948 * Neither sharefs, nfs or smbsrv are unloadable modules. 4949 */ 4950int (*znfsexport_fs)(void *arg); 4951int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t); 4952int (*zsmbexport_fs)(void *arg, boolean_t add_share); 4953 4954int zfs_nfsshare_inited; 4955int zfs_smbshare_inited; 4956 4957ddi_modhandle_t nfs_mod; 4958ddi_modhandle_t sharefs_mod; 4959ddi_modhandle_t smbsrv_mod; 4960#endif /* illumos */ 4961kmutex_t zfs_share_lock; 4962 4963#ifdef illumos 4964static int 4965zfs_init_sharefs() 4966{ 4967 int error; 4968 4969 ASSERT(MUTEX_HELD(&zfs_share_lock)); 4970 /* Both NFS and SMB shares also require sharetab support. */ 4971 if (sharefs_mod == NULL && ((sharefs_mod = 4972 ddi_modopen("fs/sharefs", 4973 KRTLD_MODE_FIRST, &error)) == NULL)) { 4974 return (SET_ERROR(ENOSYS)); 4975 } 4976 if (zshare_fs == NULL && ((zshare_fs = 4977 (int (*)(enum sharefs_sys_op, share_t *, uint32_t)) 4978 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) { 4979 return (SET_ERROR(ENOSYS)); 4980 } 4981 return (0); 4982} 4983#endif /* illumos */ 4984 4985static int 4986zfs_ioc_share(zfs_cmd_t *zc) 4987{ 4988#ifdef illumos 4989 int error; 4990 int opcode; 4991 4992 switch (zc->zc_share.z_sharetype) { 4993 case ZFS_SHARE_NFS: 4994 case ZFS_UNSHARE_NFS: 4995 if (zfs_nfsshare_inited == 0) { 4996 mutex_enter(&zfs_share_lock); 4997 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs", 4998 KRTLD_MODE_FIRST, &error)) == NULL)) { 4999 mutex_exit(&zfs_share_lock); 5000 return (SET_ERROR(ENOSYS)); 5001 } 5002 if (znfsexport_fs == NULL && 5003 ((znfsexport_fs = (int (*)(void *)) 5004 ddi_modsym(nfs_mod, 5005 "nfs_export", &error)) == NULL)) { 5006 mutex_exit(&zfs_share_lock); 5007 return (SET_ERROR(ENOSYS)); 5008 } 5009 error = zfs_init_sharefs(); 5010 if (error != 0) { 5011 mutex_exit(&zfs_share_lock); 5012 return (SET_ERROR(ENOSYS)); 5013 } 5014 zfs_nfsshare_inited = 1; 5015 mutex_exit(&zfs_share_lock); 5016 } 5017 break; 5018 case ZFS_SHARE_SMB: 5019 case ZFS_UNSHARE_SMB: 5020 if (zfs_smbshare_inited == 0) { 5021 mutex_enter(&zfs_share_lock); 5022 if (smbsrv_mod == NULL && ((smbsrv_mod = 5023 ddi_modopen("drv/smbsrv", 5024 KRTLD_MODE_FIRST, &error)) == NULL)) { 5025 mutex_exit(&zfs_share_lock); 5026 return (SET_ERROR(ENOSYS)); 5027 } 5028 if (zsmbexport_fs == NULL && ((zsmbexport_fs = 5029 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod, 5030 "smb_server_share", &error)) == NULL)) { 5031 mutex_exit(&zfs_share_lock); 5032 return (SET_ERROR(ENOSYS)); 5033 } 5034 error = zfs_init_sharefs(); 5035 if (error != 0) { 5036 mutex_exit(&zfs_share_lock); 5037 return (SET_ERROR(ENOSYS)); 5038 } 5039 zfs_smbshare_inited = 1; 5040 mutex_exit(&zfs_share_lock); 5041 } 5042 break; 5043 default: 5044 return (SET_ERROR(EINVAL)); 5045 } 5046 5047 switch (zc->zc_share.z_sharetype) { 5048 case ZFS_SHARE_NFS: 5049 case ZFS_UNSHARE_NFS: 5050 if (error = 5051 znfsexport_fs((void *) 5052 (uintptr_t)zc->zc_share.z_exportdata)) 5053 return (error); 5054 break; 5055 case ZFS_SHARE_SMB: 5056 case ZFS_UNSHARE_SMB: 5057 if (error = zsmbexport_fs((void *) 5058 (uintptr_t)zc->zc_share.z_exportdata, 5059 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ? 5060 B_TRUE: B_FALSE)) { 5061 return (error); 5062 } 5063 break; 5064 } 5065 5066 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS || 5067 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ? 5068 SHAREFS_ADD : SHAREFS_REMOVE; 5069 5070 /* 5071 * Add or remove share from sharetab 5072 */ 5073 error = zshare_fs(opcode, 5074 (void *)(uintptr_t)zc->zc_share.z_sharedata, 5075 zc->zc_share.z_sharemax); 5076 5077 return (error); 5078 5079#else /* !illumos */ 5080 return (ENOSYS); 5081#endif /* illumos */ 5082} 5083 5084ace_t full_access[] = { 5085 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0} 5086}; 5087 5088/* 5089 * inputs: 5090 * zc_name name of containing filesystem 5091 * zc_obj object # beyond which we want next in-use object # 5092 * 5093 * outputs: 5094 * zc_obj next in-use object # 5095 */ 5096static int 5097zfs_ioc_next_obj(zfs_cmd_t *zc) 5098{ 5099 objset_t *os = NULL; 5100 int error; 5101 5102 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 5103 if (error != 0) 5104 return (error); 5105 5106 error = dmu_object_next(os, &zc->zc_obj, B_FALSE, 5107 dsl_dataset_phys(os->os_dsl_dataset)->ds_prev_snap_txg); 5108 5109 dmu_objset_rele(os, FTAG); 5110 return (error); 5111} 5112 5113/* 5114 * inputs: 5115 * zc_name name of filesystem 5116 * zc_value prefix name for snapshot 5117 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process 5118 * 5119 * outputs: 5120 * zc_value short name of new snapshot 5121 */ 5122static int 5123zfs_ioc_tmp_snapshot(zfs_cmd_t *zc) 5124{ 5125 char *snap_name; 5126 char *hold_name; 5127 int error; 5128 minor_t minor; 5129 5130 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor); 5131 if (error != 0) 5132 return (error); 5133 5134 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value, 5135 (u_longlong_t)ddi_get_lbolt64()); 5136 hold_name = kmem_asprintf("%%%s", zc->zc_value); 5137 5138 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor, 5139 hold_name); 5140 if (error == 0) 5141 (void) strcpy(zc->zc_value, snap_name); 5142 strfree(snap_name); 5143 strfree(hold_name); 5144 zfs_onexit_fd_rele(zc->zc_cleanup_fd); 5145 return (error); 5146} 5147 5148/* 5149 * inputs: 5150 * zc_name name of "to" snapshot 5151 * zc_value name of "from" snapshot 5152 * zc_cookie file descriptor to write diff data on 5153 * 5154 * outputs: 5155 * dmu_diff_record_t's to the file descriptor 5156 */ 5157static int 5158zfs_ioc_diff(zfs_cmd_t *zc) 5159{ 5160 file_t *fp; 5161 cap_rights_t rights; 5162 offset_t off; 5163 int error; 5164 5165#ifdef illumos 5166 fp = getf(zc->zc_cookie); 5167#else 5168 fget_write(curthread, zc->zc_cookie, 5169 cap_rights_init(&rights, CAP_WRITE), &fp); 5170#endif 5171 if (fp == NULL) 5172 return (SET_ERROR(EBADF)); 5173 5174 off = fp->f_offset; 5175 5176#ifdef illumos 5177 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off); 5178#else 5179 error = dmu_diff(zc->zc_name, zc->zc_value, fp, &off); 5180#endif 5181 5182 if (off >= 0 && off <= MAXOFFSET_T) 5183 fp->f_offset = off; 5184 releasef(zc->zc_cookie); 5185 5186 return (error); 5187} 5188 5189#ifdef illumos 5190/* 5191 * Remove all ACL files in shares dir 5192 */ 5193static int 5194zfs_smb_acl_purge(znode_t *dzp) 5195{ 5196 zap_cursor_t zc; 5197 zap_attribute_t zap; 5198 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 5199 int error; 5200 5201 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id); 5202 (error = zap_cursor_retrieve(&zc, &zap)) == 0; 5203 zap_cursor_advance(&zc)) { 5204 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred, 5205 NULL, 0)) != 0) 5206 break; 5207 } 5208 zap_cursor_fini(&zc); 5209 return (error); 5210} 5211#endif /* illumos */ 5212 5213static int 5214zfs_ioc_smb_acl(zfs_cmd_t *zc) 5215{ 5216#ifdef illumos 5217 vnode_t *vp; 5218 znode_t *dzp; 5219 vnode_t *resourcevp = NULL; 5220 znode_t *sharedir; 5221 zfsvfs_t *zfsvfs; 5222 nvlist_t *nvlist; 5223 char *src, *target; 5224 vattr_t vattr; 5225 vsecattr_t vsec; 5226 int error = 0; 5227 5228 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, 5229 NO_FOLLOW, NULL, &vp)) != 0) 5230 return (error); 5231 5232 /* Now make sure mntpnt and dataset are ZFS */ 5233 5234 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 || 5235 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), 5236 zc->zc_name) != 0)) { 5237 VN_RELE(vp); 5238 return (SET_ERROR(EINVAL)); 5239 } 5240 5241 dzp = VTOZ(vp); 5242 zfsvfs = dzp->z_zfsvfs; 5243 ZFS_ENTER(zfsvfs); 5244 5245 /* 5246 * Create share dir if its missing. 5247 */ 5248 mutex_enter(&zfsvfs->z_lock); 5249 if (zfsvfs->z_shares_dir == 0) { 5250 dmu_tx_t *tx; 5251 5252 tx = dmu_tx_create(zfsvfs->z_os); 5253 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE, 5254 ZFS_SHARES_DIR); 5255 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); 5256 error = dmu_tx_assign(tx, TXG_WAIT); 5257 if (error != 0) { 5258 dmu_tx_abort(tx); 5259 } else { 5260 error = zfs_create_share_dir(zfsvfs, tx); 5261 dmu_tx_commit(tx); 5262 } 5263 if (error != 0) { 5264 mutex_exit(&zfsvfs->z_lock); 5265 VN_RELE(vp); 5266 ZFS_EXIT(zfsvfs); 5267 return (error); 5268 } 5269 } 5270 mutex_exit(&zfsvfs->z_lock); 5271 5272 ASSERT(zfsvfs->z_shares_dir); 5273 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) { 5274 VN_RELE(vp); 5275 ZFS_EXIT(zfsvfs); 5276 return (error); 5277 } 5278 5279 switch (zc->zc_cookie) { 5280 case ZFS_SMB_ACL_ADD: 5281 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; 5282 vattr.va_type = VREG; 5283 vattr.va_mode = S_IFREG|0777; 5284 vattr.va_uid = 0; 5285 vattr.va_gid = 0; 5286 5287 vsec.vsa_mask = VSA_ACE; 5288 vsec.vsa_aclentp = &full_access; 5289 vsec.vsa_aclentsz = sizeof (full_access); 5290 vsec.vsa_aclcnt = 1; 5291 5292 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string, 5293 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec); 5294 if (resourcevp) 5295 VN_RELE(resourcevp); 5296 break; 5297 5298 case ZFS_SMB_ACL_REMOVE: 5299 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred, 5300 NULL, 0); 5301 break; 5302 5303 case ZFS_SMB_ACL_RENAME: 5304 if ((error = get_nvlist(zc->zc_nvlist_src, 5305 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) { 5306 VN_RELE(vp); 5307 VN_RELE(ZTOV(sharedir)); 5308 ZFS_EXIT(zfsvfs); 5309 return (error); 5310 } 5311 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) || 5312 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET, 5313 &target)) { 5314 VN_RELE(vp); 5315 VN_RELE(ZTOV(sharedir)); 5316 ZFS_EXIT(zfsvfs); 5317 nvlist_free(nvlist); 5318 return (error); 5319 } 5320 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target, 5321 kcred, NULL, 0); 5322 nvlist_free(nvlist); 5323 break; 5324 5325 case ZFS_SMB_ACL_PURGE: 5326 error = zfs_smb_acl_purge(sharedir); 5327 break; 5328 5329 default: 5330 error = SET_ERROR(EINVAL); 5331 break; 5332 } 5333 5334 VN_RELE(vp); 5335 VN_RELE(ZTOV(sharedir)); 5336 5337 ZFS_EXIT(zfsvfs); 5338 5339 return (error); 5340#else /* !illumos */ 5341 return (EOPNOTSUPP); 5342#endif /* illumos */ 5343} 5344 5345/* 5346 * innvl: { 5347 * "holds" -> { snapname -> holdname (string), ... } 5348 * (optional) "cleanup_fd" -> fd (int32) 5349 * } 5350 * 5351 * outnvl: { 5352 * snapname -> error value (int32) 5353 * ... 5354 * } 5355 */ 5356/* ARGSUSED */ 5357static int 5358zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist) 5359{ 5360 nvpair_t *pair; 5361 nvlist_t *holds; 5362 int cleanup_fd = -1; 5363 int error; 5364 minor_t minor = 0; 5365 5366 error = nvlist_lookup_nvlist(args, "holds", &holds); 5367 if (error != 0) 5368 return (SET_ERROR(EINVAL)); 5369 5370 /* make sure the user didn't pass us any invalid (empty) tags */ 5371 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL; 5372 pair = nvlist_next_nvpair(holds, pair)) { 5373 char *htag; 5374 5375 error = nvpair_value_string(pair, &htag); 5376 if (error != 0) 5377 return (SET_ERROR(error)); 5378 5379 if (strlen(htag) == 0) 5380 return (SET_ERROR(EINVAL)); 5381 } 5382 5383 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) { 5384 error = zfs_onexit_fd_hold(cleanup_fd, &minor); 5385 if (error != 0) 5386 return (error); 5387 } 5388 5389 error = dsl_dataset_user_hold(holds, minor, errlist); 5390 if (minor != 0) 5391 zfs_onexit_fd_rele(cleanup_fd); 5392 return (error); 5393} 5394 5395/* 5396 * innvl is not used. 5397 * 5398 * outnvl: { 5399 * holdname -> time added (uint64 seconds since epoch) 5400 * ... 5401 * } 5402 */ 5403/* ARGSUSED */ 5404static int 5405zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl) 5406{ 5407 return (dsl_dataset_get_holds(snapname, outnvl)); 5408} 5409 5410/* 5411 * innvl: { 5412 * snapname -> { holdname, ... } 5413 * ... 5414 * } 5415 * 5416 * outnvl: { 5417 * snapname -> error value (int32) 5418 * ... 5419 * } 5420 */ 5421/* ARGSUSED */ 5422static int 5423zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist) 5424{ 5425 return (dsl_dataset_user_release(holds, errlist)); 5426} 5427 5428/* 5429 * inputs: 5430 * zc_name name of new filesystem or snapshot 5431 * zc_value full name of old snapshot 5432 * 5433 * outputs: 5434 * zc_cookie space in bytes 5435 * zc_objset_type compressed space in bytes 5436 * zc_perm_action uncompressed space in bytes 5437 */ 5438static int 5439zfs_ioc_space_written(zfs_cmd_t *zc) 5440{ 5441 int error; 5442 dsl_pool_t *dp; 5443 dsl_dataset_t *new, *old; 5444 5445 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 5446 if (error != 0) 5447 return (error); 5448 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new); 5449 if (error != 0) { 5450 dsl_pool_rele(dp, FTAG); 5451 return (error); 5452 } 5453 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old); 5454 if (error != 0) { 5455 dsl_dataset_rele(new, FTAG); 5456 dsl_pool_rele(dp, FTAG); 5457 return (error); 5458 } 5459 5460 error = dsl_dataset_space_written(old, new, &zc->zc_cookie, 5461 &zc->zc_objset_type, &zc->zc_perm_action); 5462 dsl_dataset_rele(old, FTAG); 5463 dsl_dataset_rele(new, FTAG); 5464 dsl_pool_rele(dp, FTAG); 5465 return (error); 5466} 5467 5468/* 5469 * innvl: { 5470 * "firstsnap" -> snapshot name 5471 * } 5472 * 5473 * outnvl: { 5474 * "used" -> space in bytes 5475 * "compressed" -> compressed space in bytes 5476 * "uncompressed" -> uncompressed space in bytes 5477 * } 5478 */ 5479static int 5480zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl) 5481{ 5482 int error; 5483 dsl_pool_t *dp; 5484 dsl_dataset_t *new, *old; 5485 char *firstsnap; 5486 uint64_t used, comp, uncomp; 5487 5488 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0) 5489 return (SET_ERROR(EINVAL)); 5490 5491 error = dsl_pool_hold(lastsnap, FTAG, &dp); 5492 if (error != 0) 5493 return (error); 5494 5495 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new); 5496 if (error == 0 && !new->ds_is_snapshot) { 5497 dsl_dataset_rele(new, FTAG); 5498 error = SET_ERROR(EINVAL); 5499 } 5500 if (error != 0) { 5501 dsl_pool_rele(dp, FTAG); 5502 return (error); 5503 } 5504 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old); 5505 if (error == 0 && !old->ds_is_snapshot) { 5506 dsl_dataset_rele(old, FTAG); 5507 error = SET_ERROR(EINVAL); 5508 } 5509 if (error != 0) { 5510 dsl_dataset_rele(new, FTAG); 5511 dsl_pool_rele(dp, FTAG); 5512 return (error); 5513 } 5514 5515 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp); 5516 dsl_dataset_rele(old, FTAG); 5517 dsl_dataset_rele(new, FTAG); 5518 dsl_pool_rele(dp, FTAG); 5519 fnvlist_add_uint64(outnvl, "used", used); 5520 fnvlist_add_uint64(outnvl, "compressed", comp); 5521 fnvlist_add_uint64(outnvl, "uncompressed", uncomp); 5522 return (error); 5523} 5524 5525static int 5526zfs_ioc_jail(zfs_cmd_t *zc) 5527{ 5528 5529 return (zone_dataset_attach(curthread->td_ucred, zc->zc_name, 5530 (int)zc->zc_jailid)); 5531} 5532 5533static int 5534zfs_ioc_unjail(zfs_cmd_t *zc) 5535{ 5536 5537 return (zone_dataset_detach(curthread->td_ucred, zc->zc_name, 5538 (int)zc->zc_jailid)); 5539} 5540 5541/* 5542 * innvl: { 5543 * "fd" -> file descriptor to write stream to (int32) 5544 * (optional) "fromsnap" -> full snap name to send an incremental from 5545 * (optional) "largeblockok" -> (value ignored) 5546 * indicates that blocks > 128KB are permitted 5547 * (optional) "embedok" -> (value ignored) 5548 * presence indicates DRR_WRITE_EMBEDDED records are permitted 5549 * (optional) "resume_object" and "resume_offset" -> (uint64) 5550 * if present, resume send stream from specified object and offset. 5551 * } 5552 * 5553 * outnvl is unused 5554 */ 5555/* ARGSUSED */ 5556static int 5557zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl) 5558{ 5559 cap_rights_t rights; 5560 file_t *fp; 5561 int error; 5562 offset_t off; 5563 char *fromname = NULL; 5564 int fd; 5565 boolean_t largeblockok; 5566 boolean_t embedok; 5567 uint64_t resumeobj = 0; 5568 uint64_t resumeoff = 0; 5569 5570 error = nvlist_lookup_int32(innvl, "fd", &fd); 5571 if (error != 0) 5572 return (SET_ERROR(EINVAL)); 5573 5574 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname); 5575 5576 largeblockok = nvlist_exists(innvl, "largeblockok"); 5577 embedok = nvlist_exists(innvl, "embedok"); 5578 5579 (void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj); 5580 (void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff); 5581 5582#ifdef illumos 5583 file_t *fp = getf(fd); 5584#else 5585 fget_write(curthread, fd, cap_rights_init(&rights, CAP_WRITE), &fp); 5586#endif 5587 if (fp == NULL) 5588 return (SET_ERROR(EBADF)); 5589 5590 off = fp->f_offset; 5591 error = dmu_send(snapname, fromname, embedok, largeblockok, fd, 5592#ifdef illumos 5593 resumeobj, resumeoff, fp->f_vnode, &off); 5594#else 5595 resumeobj, resumeoff, fp, &off); 5596#endif 5597 5598#ifdef illumos 5599 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0) 5600 fp->f_offset = off; 5601#else 5602 fp->f_offset = off; 5603#endif 5604 5605 releasef(fd); 5606 return (error); 5607} 5608 5609/* 5610 * Determine approximately how large a zfs send stream will be -- the number 5611 * of bytes that will be written to the fd supplied to zfs_ioc_send_new(). 5612 * 5613 * innvl: { 5614 * (optional) "from" -> full snap or bookmark name to send an incremental 5615 * from 5616 * } 5617 * 5618 * outnvl: { 5619 * "space" -> bytes of space (uint64) 5620 * } 5621 */ 5622static int 5623zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl) 5624{ 5625 dsl_pool_t *dp; 5626 dsl_dataset_t *tosnap; 5627 int error; 5628 char *fromname; 5629 uint64_t space; 5630 5631 error = dsl_pool_hold(snapname, FTAG, &dp); 5632 if (error != 0) 5633 return (error); 5634 5635 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap); 5636 if (error != 0) { 5637 dsl_pool_rele(dp, FTAG); 5638 return (error); 5639 } 5640 5641 error = nvlist_lookup_string(innvl, "from", &fromname); 5642 if (error == 0) { 5643 if (strchr(fromname, '@') != NULL) { 5644 /* 5645 * If from is a snapshot, hold it and use the more 5646 * efficient dmu_send_estimate to estimate send space 5647 * size using deadlists. 5648 */ 5649 dsl_dataset_t *fromsnap; 5650 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap); 5651 if (error != 0) 5652 goto out; 5653 error = dmu_send_estimate(tosnap, fromsnap, &space); 5654 dsl_dataset_rele(fromsnap, FTAG); 5655 } else if (strchr(fromname, '#') != NULL) { 5656 /* 5657 * If from is a bookmark, fetch the creation TXG of the 5658 * snapshot it was created from and use that to find 5659 * blocks that were born after it. 5660 */ 5661 zfs_bookmark_phys_t frombm; 5662 5663 error = dsl_bookmark_lookup(dp, fromname, tosnap, 5664 &frombm); 5665 if (error != 0) 5666 goto out; 5667 error = dmu_send_estimate_from_txg(tosnap, 5668 frombm.zbm_creation_txg, &space); 5669 } else { 5670 /* 5671 * from is not properly formatted as a snapshot or 5672 * bookmark 5673 */ 5674 error = SET_ERROR(EINVAL); 5675 goto out; 5676 } 5677 } else { 5678 // If estimating the size of a full send, use dmu_send_estimate 5679 error = dmu_send_estimate(tosnap, NULL, &space); 5680 } 5681 5682 fnvlist_add_uint64(outnvl, "space", space); 5683 5684out: 5685 dsl_dataset_rele(tosnap, FTAG); 5686 dsl_pool_rele(dp, FTAG); 5687 return (error); 5688} 5689 5690static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST]; 5691 5692static void 5693zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5694 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck, 5695 boolean_t log_history, zfs_ioc_poolcheck_t pool_check) 5696{ 5697 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST]; 5698 5699 ASSERT3U(ioc, >=, ZFS_IOC_FIRST); 5700 ASSERT3U(ioc, <, ZFS_IOC_LAST); 5701 ASSERT3P(vec->zvec_legacy_func, ==, NULL); 5702 ASSERT3P(vec->zvec_func, ==, NULL); 5703 5704 vec->zvec_legacy_func = func; 5705 vec->zvec_secpolicy = secpolicy; 5706 vec->zvec_namecheck = namecheck; 5707 vec->zvec_allow_log = log_history; 5708 vec->zvec_pool_check = pool_check; 5709} 5710 5711/* 5712 * See the block comment at the beginning of this file for details on 5713 * each argument to this function. 5714 */ 5715static void 5716zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func, 5717 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck, 5718 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist, 5719 boolean_t allow_log) 5720{ 5721 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST]; 5722 5723 ASSERT3U(ioc, >=, ZFS_IOC_FIRST); 5724 ASSERT3U(ioc, <, ZFS_IOC_LAST); 5725 ASSERT3P(vec->zvec_legacy_func, ==, NULL); 5726 ASSERT3P(vec->zvec_func, ==, NULL); 5727 5728 /* if we are logging, the name must be valid */ 5729 ASSERT(!allow_log || namecheck != NO_NAME); 5730 5731 vec->zvec_name = name; 5732 vec->zvec_func = func; 5733 vec->zvec_secpolicy = secpolicy; 5734 vec->zvec_namecheck = namecheck; 5735 vec->zvec_pool_check = pool_check; 5736 vec->zvec_smush_outnvlist = smush_outnvlist; 5737 vec->zvec_allow_log = allow_log; 5738} 5739 5740static void 5741zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5742 zfs_secpolicy_func_t *secpolicy, boolean_t log_history, 5743 zfs_ioc_poolcheck_t pool_check) 5744{ 5745 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5746 POOL_NAME, log_history, pool_check); 5747} 5748 5749static void 5750zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5751 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check) 5752{ 5753 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5754 DATASET_NAME, B_FALSE, pool_check); 5755} 5756 5757static void 5758zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func) 5759{ 5760 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config, 5761 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5762} 5763 5764static void 5765zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5766 zfs_secpolicy_func_t *secpolicy) 5767{ 5768 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5769 NO_NAME, B_FALSE, POOL_CHECK_NONE); 5770} 5771 5772static void 5773zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc, 5774 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy) 5775{ 5776 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5777 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED); 5778} 5779 5780static void 5781zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func) 5782{ 5783 zfs_ioctl_register_dataset_read_secpolicy(ioc, func, 5784 zfs_secpolicy_read); 5785} 5786 5787static void 5788zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5789 zfs_secpolicy_func_t *secpolicy) 5790{ 5791 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5792 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5793} 5794 5795static void 5796zfs_ioctl_init(void) 5797{ 5798 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT, 5799 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME, 5800 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5801 5802 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY, 5803 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME, 5804 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE); 5805 5806 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS, 5807 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME, 5808 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5809 5810 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW, 5811 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME, 5812 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5813 5814 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE, 5815 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME, 5816 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5817 5818 zfs_ioctl_register("create", ZFS_IOC_CREATE, 5819 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME, 5820 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5821 5822 zfs_ioctl_register("clone", ZFS_IOC_CLONE, 5823 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME, 5824 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5825 5826 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS, 5827 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME, 5828 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5829 5830 zfs_ioctl_register("hold", ZFS_IOC_HOLD, 5831 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME, 5832 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5833 zfs_ioctl_register("release", ZFS_IOC_RELEASE, 5834 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME, 5835 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5836 5837 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS, 5838 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME, 5839 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5840 5841 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK, 5842 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, 5843 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE); 5844 5845 zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK, 5846 zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME, 5847 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5848 5849 zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS, 5850 zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME, 5851 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5852 5853 zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS, 5854 zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks, 5855 POOL_NAME, 5856 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5857 5858 /* IOCTLS that use the legacy function signature */ 5859 5860 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze, 5861 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY); 5862 5863 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create, 5864 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5865 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN, 5866 zfs_ioc_pool_scan); 5867 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE, 5868 zfs_ioc_pool_upgrade); 5869 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD, 5870 zfs_ioc_vdev_add); 5871 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE, 5872 zfs_ioc_vdev_remove); 5873 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE, 5874 zfs_ioc_vdev_set_state); 5875 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH, 5876 zfs_ioc_vdev_attach); 5877 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH, 5878 zfs_ioc_vdev_detach); 5879 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH, 5880 zfs_ioc_vdev_setpath); 5881 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU, 5882 zfs_ioc_vdev_setfru); 5883 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS, 5884 zfs_ioc_pool_set_props); 5885 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT, 5886 zfs_ioc_vdev_split); 5887 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID, 5888 zfs_ioc_pool_reguid); 5889 5890 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS, 5891 zfs_ioc_pool_configs, zfs_secpolicy_none); 5892 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT, 5893 zfs_ioc_pool_tryimport, zfs_secpolicy_config); 5894 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT, 5895 zfs_ioc_inject_fault, zfs_secpolicy_inject); 5896 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT, 5897 zfs_ioc_clear_fault, zfs_secpolicy_inject); 5898 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT, 5899 zfs_ioc_inject_list_next, zfs_secpolicy_inject); 5900 5901 /* 5902 * pool destroy, and export don't log the history as part of 5903 * zfsdev_ioctl, but rather zfs_ioc_pool_export 5904 * does the logging of those commands. 5905 */ 5906 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy, 5907 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE); 5908 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export, 5909 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE); 5910 5911 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats, 5912 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE); 5913 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props, 5914 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE); 5915 5916 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log, 5917 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_NONE); 5918 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME, 5919 zfs_ioc_dsobj_to_dsname, 5920 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_NONE); 5921 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY, 5922 zfs_ioc_pool_get_history, 5923 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED); 5924 5925 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import, 5926 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5927 5928 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear, 5929 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5930 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen, 5931 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED); 5932 5933 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN, 5934 zfs_ioc_space_written); 5935 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS, 5936 zfs_ioc_objset_recvd_props); 5937 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ, 5938 zfs_ioc_next_obj); 5939 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL, 5940 zfs_ioc_get_fsacl); 5941 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS, 5942 zfs_ioc_objset_stats); 5943 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS, 5944 zfs_ioc_objset_zplprops); 5945 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT, 5946 zfs_ioc_dataset_list_next); 5947 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT, 5948 zfs_ioc_snapshot_list_next); 5949 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS, 5950 zfs_ioc_send_progress); 5951 5952 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF, 5953 zfs_ioc_diff, zfs_secpolicy_diff); 5954 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS, 5955 zfs_ioc_obj_to_stats, zfs_secpolicy_diff); 5956 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH, 5957 zfs_ioc_obj_to_path, zfs_secpolicy_diff); 5958 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE, 5959 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one); 5960 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY, 5961 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many); 5962 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND, 5963 zfs_ioc_send, zfs_secpolicy_send); 5964 5965 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop, 5966 zfs_secpolicy_none); 5967 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy, 5968 zfs_secpolicy_destroy); 5969 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename, 5970 zfs_secpolicy_rename); 5971 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv, 5972 zfs_secpolicy_recv); 5973 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote, 5974 zfs_secpolicy_promote); 5975 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP, 5976 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop); 5977 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl, 5978 zfs_secpolicy_set_fsacl); 5979 5980 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share, 5981 zfs_secpolicy_share, POOL_CHECK_NONE); 5982 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl, 5983 zfs_secpolicy_smb_acl, POOL_CHECK_NONE); 5984 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE, 5985 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade, 5986 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5987 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT, 5988 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot, 5989 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5990 5991#ifdef __FreeBSD__ 5992 zfs_ioctl_register_dataset_nolog(ZFS_IOC_JAIL, zfs_ioc_jail, 5993 zfs_secpolicy_config, POOL_CHECK_NONE); 5994 zfs_ioctl_register_dataset_nolog(ZFS_IOC_UNJAIL, zfs_ioc_unjail, 5995 zfs_secpolicy_config, POOL_CHECK_NONE); 5996#endif 5997} 5998 5999int 6000pool_status_check(const char *name, zfs_ioc_namecheck_t type, 6001 zfs_ioc_poolcheck_t check) 6002{ 6003 spa_t *spa; 6004 int error; 6005 6006 ASSERT(type == POOL_NAME || type == DATASET_NAME); 6007 6008 if (check & POOL_CHECK_NONE) 6009 return (0); 6010 6011 error = spa_open(name, &spa, FTAG); 6012 if (error == 0) { 6013 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa)) 6014 error = SET_ERROR(EAGAIN); 6015 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa)) 6016 error = SET_ERROR(EROFS); 6017 spa_close(spa, FTAG); 6018 } 6019 return (error); 6020} 6021 6022/* 6023 * Find a free minor number. 6024 */ 6025minor_t 6026zfsdev_minor_alloc(void) 6027{ 6028 static minor_t last_minor; 6029 minor_t m; 6030 6031 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 6032 6033 for (m = last_minor + 1; m != last_minor; m++) { 6034 if (m > ZFSDEV_MAX_MINOR) 6035 m = 1; 6036 if (ddi_get_soft_state(zfsdev_state, m) == NULL) { 6037 last_minor = m; 6038 return (m); 6039 } 6040 } 6041 6042 return (0); 6043} 6044 6045static int 6046zfs_ctldev_init(struct cdev *devp) 6047{ 6048 minor_t minor; 6049 zfs_soft_state_t *zs; 6050 6051 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 6052 6053 minor = zfsdev_minor_alloc(); 6054 if (minor == 0) 6055 return (SET_ERROR(ENXIO)); 6056 6057 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS) 6058 return (SET_ERROR(EAGAIN)); 6059 6060 devfs_set_cdevpriv((void *)(uintptr_t)minor, zfsdev_close); 6061 6062 zs = ddi_get_soft_state(zfsdev_state, minor); 6063 zs->zss_type = ZSST_CTLDEV; 6064 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data); 6065 6066 return (0); 6067} 6068 6069static void 6070zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor) 6071{ 6072 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 6073 6074 zfs_onexit_destroy(zo); 6075 ddi_soft_state_free(zfsdev_state, minor); 6076} 6077 6078void * 6079zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which) 6080{ 6081 zfs_soft_state_t *zp; 6082 6083 zp = ddi_get_soft_state(zfsdev_state, minor); 6084 if (zp == NULL || zp->zss_type != which) 6085 return (NULL); 6086 6087 return (zp->zss_data); 6088} 6089 6090static int 6091zfsdev_open(struct cdev *devp, int flag, int mode, struct thread *td) 6092{ 6093 int error = 0; 6094 6095#ifdef illumos 6096 if (getminor(*devp) != 0) 6097 return (zvol_open(devp, flag, otyp, cr)); 6098#endif 6099 6100 /* This is the control device. Allocate a new minor if requested. */ 6101 if (flag & FEXCL) { 6102 mutex_enter(&spa_namespace_lock); 6103 error = zfs_ctldev_init(devp); 6104 mutex_exit(&spa_namespace_lock); 6105 } 6106 6107 return (error); 6108} 6109 6110static void 6111zfsdev_close(void *data) 6112{ 6113 zfs_onexit_t *zo; 6114 minor_t minor = (minor_t)(uintptr_t)data; 6115 6116 if (minor == 0) 6117 return; 6118 6119 mutex_enter(&spa_namespace_lock); 6120 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV); 6121 if (zo == NULL) { 6122 mutex_exit(&spa_namespace_lock); 6123 return; 6124 } 6125 zfs_ctldev_destroy(zo, minor); 6126 mutex_exit(&spa_namespace_lock); 6127} 6128 6129static int 6130zfsdev_ioctl(struct cdev *dev, u_long zcmd, caddr_t arg, int flag, 6131 struct thread *td) 6132{ 6133 zfs_cmd_t *zc; 6134 uint_t vecnum; 6135 int error, rc, len; 6136#ifdef illumos 6137 minor_t minor = getminor(dev); 6138#else 6139 zfs_iocparm_t *zc_iocparm; 6140 int cflag, cmd, oldvecnum; 6141 boolean_t newioc, compat; 6142 void *compat_zc = NULL; 6143 cred_t *cr = td->td_ucred; 6144#endif 6145 const zfs_ioc_vec_t *vec; 6146 char *saved_poolname = NULL; 6147 nvlist_t *innvl = NULL; 6148 6149 cflag = ZFS_CMD_COMPAT_NONE; 6150 compat = B_FALSE; 6151 newioc = B_TRUE; /* "new" style (zfs_iocparm_t) ioctl */ 6152 6153 len = IOCPARM_LEN(zcmd); 6154 vecnum = cmd = zcmd & 0xff; 6155 6156 /* 6157 * Check if we are talking to supported older binaries 6158 * and translate zfs_cmd if necessary 6159 */ 6160 if (len != sizeof(zfs_iocparm_t)) { 6161 newioc = B_FALSE; 6162 compat = B_TRUE; 6163 6164 vecnum = cmd; 6165 6166 switch (len) { 6167 case sizeof(zfs_cmd_zcmd_t): 6168 cflag = ZFS_CMD_COMPAT_LZC; 6169 break; 6170 case sizeof(zfs_cmd_deadman_t): 6171 cflag = ZFS_CMD_COMPAT_DEADMAN; 6172 break; 6173 case sizeof(zfs_cmd_v28_t): 6174 cflag = ZFS_CMD_COMPAT_V28; 6175 break; 6176 case sizeof(zfs_cmd_v15_t): 6177 cflag = ZFS_CMD_COMPAT_V15; 6178 vecnum = zfs_ioctl_v15_to_v28[cmd]; 6179 6180 /* 6181 * Return without further handling 6182 * if the command is blacklisted. 6183 */ 6184 if (vecnum == ZFS_IOC_COMPAT_PASS) 6185 return (0); 6186 else if (vecnum == ZFS_IOC_COMPAT_FAIL) 6187 return (ENOTSUP); 6188 break; 6189 default: 6190 return (EINVAL); 6191 } 6192 } 6193 6194#ifdef illumos 6195 vecnum = cmd - ZFS_IOC_FIRST; 6196 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip)); 6197#endif 6198 6199 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0])) 6200 return (SET_ERROR(EINVAL)); 6201 vec = &zfs_ioc_vec[vecnum]; 6202 6203 zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP); 6204 6205#ifdef illumos 6206 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag); 6207 if (error != 0) { 6208 error = SET_ERROR(EFAULT); 6209 goto out; 6210 } 6211#else /* !illumos */ 6212 bzero(zc, sizeof(zfs_cmd_t)); 6213 6214 if (newioc) { 6215 zc_iocparm = (void *)arg; 6216 6217 switch (zc_iocparm->zfs_ioctl_version) { 6218 case ZFS_IOCVER_CURRENT: 6219 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_t)) { 6220 error = SET_ERROR(EINVAL); 6221 goto out; 6222 } 6223 break; 6224 case ZFS_IOCVER_EDBP: 6225 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_edbp_t)) { 6226 error = SET_ERROR(EFAULT); 6227 goto out; 6228 } 6229 compat = B_TRUE; 6230 cflag = ZFS_CMD_COMPAT_EDBP; 6231 break; 6232 case ZFS_IOCVER_ZCMD: 6233 if (zc_iocparm->zfs_cmd_size > sizeof(zfs_cmd_t) || 6234 zc_iocparm->zfs_cmd_size < sizeof(zfs_cmd_zcmd_t)) { 6235 error = SET_ERROR(EFAULT); 6236 goto out; 6237 } 6238 compat = B_TRUE; 6239 cflag = ZFS_CMD_COMPAT_ZCMD; 6240 break; 6241 default: 6242 error = SET_ERROR(EINVAL); 6243 goto out; 6244 /* NOTREACHED */ 6245 } 6246 6247 if (compat) { 6248 ASSERT(sizeof(zfs_cmd_t) >= zc_iocparm->zfs_cmd_size); 6249 compat_zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP); 6250 bzero(compat_zc, sizeof(zfs_cmd_t)); 6251 6252 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd, 6253 compat_zc, zc_iocparm->zfs_cmd_size, flag); 6254 if (error != 0) { 6255 error = SET_ERROR(EFAULT); 6256 goto out; 6257 } 6258 } else { 6259 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd, 6260 zc, zc_iocparm->zfs_cmd_size, flag); 6261 if (error != 0) { 6262 error = SET_ERROR(EFAULT); 6263 goto out; 6264 } 6265 } 6266 } 6267 6268 if (compat) { 6269 if (newioc) { 6270 ASSERT(compat_zc != NULL); 6271 zfs_cmd_compat_get(zc, compat_zc, cflag); 6272 } else { 6273 ASSERT(compat_zc == NULL); 6274 zfs_cmd_compat_get(zc, arg, cflag); 6275 } 6276 oldvecnum = vecnum; 6277 error = zfs_ioctl_compat_pre(zc, &vecnum, cflag); 6278 if (error != 0) 6279 goto out; 6280 if (oldvecnum != vecnum) 6281 vec = &zfs_ioc_vec[vecnum]; 6282 } 6283#endif /* !illumos */ 6284 6285 zc->zc_iflags = flag & FKIOCTL; 6286 if (zc->zc_nvlist_src_size != 0) { 6287 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 6288 zc->zc_iflags, &innvl); 6289 if (error != 0) 6290 goto out; 6291 } 6292 6293 /* rewrite innvl for backwards compatibility */ 6294 if (compat) 6295 innvl = zfs_ioctl_compat_innvl(zc, innvl, vecnum, cflag); 6296 6297 /* 6298 * Ensure that all pool/dataset names are valid before we pass down to 6299 * the lower layers. 6300 */ 6301 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0'; 6302 switch (vec->zvec_namecheck) { 6303 case POOL_NAME: 6304 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0) 6305 error = SET_ERROR(EINVAL); 6306 else 6307 error = pool_status_check(zc->zc_name, 6308 vec->zvec_namecheck, vec->zvec_pool_check); 6309 break; 6310 6311 case DATASET_NAME: 6312 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0) 6313 error = SET_ERROR(EINVAL); 6314 else 6315 error = pool_status_check(zc->zc_name, 6316 vec->zvec_namecheck, vec->zvec_pool_check); 6317 break; 6318 6319 case NO_NAME: 6320 break; 6321 } 6322 6323 if (error == 0 && !(flag & FKIOCTL)) 6324 error = vec->zvec_secpolicy(zc, innvl, cr); 6325 6326 if (error != 0) 6327 goto out; 6328 6329 /* legacy ioctls can modify zc_name */ 6330 len = strcspn(zc->zc_name, "/@#") + 1; 6331 saved_poolname = kmem_alloc(len, KM_SLEEP); 6332 (void) strlcpy(saved_poolname, zc->zc_name, len); 6333 6334 if (vec->zvec_func != NULL) { 6335 nvlist_t *outnvl; 6336 int puterror = 0; 6337 spa_t *spa; 6338 nvlist_t *lognv = NULL; 6339 6340 ASSERT(vec->zvec_legacy_func == NULL); 6341 6342 /* 6343 * Add the innvl to the lognv before calling the func, 6344 * in case the func changes the innvl. 6345 */ 6346 if (vec->zvec_allow_log) { 6347 lognv = fnvlist_alloc(); 6348 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL, 6349 vec->zvec_name); 6350 if (!nvlist_empty(innvl)) { 6351 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL, 6352 innvl); 6353 } 6354 } 6355 6356 outnvl = fnvlist_alloc(); 6357 error = vec->zvec_func(zc->zc_name, innvl, outnvl); 6358 6359 if (error == 0 && vec->zvec_allow_log && 6360 spa_open(zc->zc_name, &spa, FTAG) == 0) { 6361 if (!nvlist_empty(outnvl)) { 6362 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL, 6363 outnvl); 6364 } 6365 (void) spa_history_log_nvl(spa, lognv); 6366 spa_close(spa, FTAG); 6367 } 6368 fnvlist_free(lognv); 6369 6370 /* rewrite outnvl for backwards compatibility */ 6371 if (compat) 6372 outnvl = zfs_ioctl_compat_outnvl(zc, outnvl, vecnum, 6373 cflag); 6374 6375 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) { 6376 int smusherror = 0; 6377 if (vec->zvec_smush_outnvlist) { 6378 smusherror = nvlist_smush(outnvl, 6379 zc->zc_nvlist_dst_size); 6380 } 6381 if (smusherror == 0) 6382 puterror = put_nvlist(zc, outnvl); 6383 } 6384 6385 if (puterror != 0) 6386 error = puterror; 6387 6388 nvlist_free(outnvl); 6389 } else { 6390 error = vec->zvec_legacy_func(zc); 6391 } 6392 6393out: 6394 nvlist_free(innvl); 6395 6396#ifdef illumos 6397 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag); 6398 if (error == 0 && rc != 0) 6399 error = SET_ERROR(EFAULT); 6400#else 6401 if (compat) { 6402 zfs_ioctl_compat_post(zc, cmd, cflag); 6403 if (newioc) { 6404 ASSERT(compat_zc != NULL); 6405 ASSERT(sizeof(zfs_cmd_t) >= zc_iocparm->zfs_cmd_size); 6406 6407 zfs_cmd_compat_put(zc, compat_zc, vecnum, cflag); 6408 rc = ddi_copyout(compat_zc, 6409 (void *)(uintptr_t)zc_iocparm->zfs_cmd, 6410 zc_iocparm->zfs_cmd_size, flag); 6411 if (error == 0 && rc != 0) 6412 error = SET_ERROR(EFAULT); 6413 kmem_free(compat_zc, sizeof (zfs_cmd_t)); 6414 } else { 6415 zfs_cmd_compat_put(zc, arg, vecnum, cflag); 6416 } 6417 } else { 6418 ASSERT(newioc); 6419 6420 rc = ddi_copyout(zc, (void *)(uintptr_t)zc_iocparm->zfs_cmd, 6421 sizeof (zfs_cmd_t), flag); 6422 if (error == 0 && rc != 0) 6423 error = SET_ERROR(EFAULT); 6424 } 6425#endif 6426 if (error == 0 && vec->zvec_allow_log) { 6427 char *s = tsd_get(zfs_allow_log_key); 6428 if (s != NULL) 6429 strfree(s); 6430 (void) tsd_set(zfs_allow_log_key, saved_poolname); 6431 } else { 6432 if (saved_poolname != NULL) 6433 strfree(saved_poolname); 6434 } 6435 6436 kmem_free(zc, sizeof (zfs_cmd_t)); 6437 return (error); 6438} 6439 6440#ifdef illumos 6441static int 6442zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd) 6443{ 6444 if (cmd != DDI_ATTACH) 6445 return (DDI_FAILURE); 6446 6447 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0, 6448 DDI_PSEUDO, 0) == DDI_FAILURE) 6449 return (DDI_FAILURE); 6450 6451 zfs_dip = dip; 6452 6453 ddi_report_dev(dip); 6454 6455 return (DDI_SUCCESS); 6456} 6457 6458static int 6459zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd) 6460{ 6461 if (spa_busy() || zfs_busy() || zvol_busy()) 6462 return (DDI_FAILURE); 6463 6464 if (cmd != DDI_DETACH) 6465 return (DDI_FAILURE); 6466 6467 zfs_dip = NULL; 6468 6469 ddi_prop_remove_all(dip); 6470 ddi_remove_minor_node(dip, NULL); 6471 6472 return (DDI_SUCCESS); 6473} 6474 6475/*ARGSUSED*/ 6476static int 6477zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result) 6478{ 6479 switch (infocmd) { 6480 case DDI_INFO_DEVT2DEVINFO: 6481 *result = zfs_dip; 6482 return (DDI_SUCCESS); 6483 6484 case DDI_INFO_DEVT2INSTANCE: 6485 *result = (void *)0; 6486 return (DDI_SUCCESS); 6487 } 6488 6489 return (DDI_FAILURE); 6490} 6491#endif /* illumos */ 6492 6493/* 6494 * OK, so this is a little weird. 6495 * 6496 * /dev/zfs is the control node, i.e. minor 0. 6497 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0. 6498 * 6499 * /dev/zfs has basically nothing to do except serve up ioctls, 6500 * so most of the standard driver entry points are in zvol.c. 6501 */ 6502#ifdef illumos 6503static struct cb_ops zfs_cb_ops = { 6504 zfsdev_open, /* open */ 6505 zfsdev_close, /* close */ 6506 zvol_strategy, /* strategy */ 6507 nodev, /* print */ 6508 zvol_dump, /* dump */ 6509 zvol_read, /* read */ 6510 zvol_write, /* write */ 6511 zfsdev_ioctl, /* ioctl */ 6512 nodev, /* devmap */ 6513 nodev, /* mmap */ 6514 nodev, /* segmap */ 6515 nochpoll, /* poll */ 6516 ddi_prop_op, /* prop_op */ 6517 NULL, /* streamtab */ 6518 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */ 6519 CB_REV, /* version */ 6520 nodev, /* async read */ 6521 nodev, /* async write */ 6522}; 6523 6524static struct dev_ops zfs_dev_ops = { 6525 DEVO_REV, /* version */ 6526 0, /* refcnt */ 6527 zfs_info, /* info */ 6528 nulldev, /* identify */ 6529 nulldev, /* probe */ 6530 zfs_attach, /* attach */ 6531 zfs_detach, /* detach */ 6532 nodev, /* reset */ 6533 &zfs_cb_ops, /* driver operations */ 6534 NULL, /* no bus operations */ 6535 NULL, /* power */ 6536 ddi_quiesce_not_needed, /* quiesce */ 6537}; 6538 6539static struct modldrv zfs_modldrv = { 6540 &mod_driverops, 6541 "ZFS storage pool", 6542 &zfs_dev_ops 6543}; 6544 6545static struct modlinkage modlinkage = { 6546 MODREV_1, 6547 (void *)&zfs_modlfs, 6548 (void *)&zfs_modldrv, 6549 NULL 6550}; 6551#endif /* illumos */ 6552 6553static struct cdevsw zfs_cdevsw = { 6554 .d_version = D_VERSION, 6555 .d_open = zfsdev_open, 6556 .d_ioctl = zfsdev_ioctl, 6557 .d_name = ZFS_DEV_NAME 6558}; 6559 6560static void 6561zfs_allow_log_destroy(void *arg) 6562{ 6563 char *poolname = arg; 6564 strfree(poolname); 6565} 6566 6567static void 6568zfsdev_init(void) 6569{ 6570 zfsdev = make_dev(&zfs_cdevsw, 0x0, UID_ROOT, GID_OPERATOR, 0666, 6571 ZFS_DEV_NAME); 6572} 6573 6574static void 6575zfsdev_fini(void) 6576{ 6577 if (zfsdev != NULL) 6578 destroy_dev(zfsdev); 6579} 6580 6581static struct root_hold_token *zfs_root_token; 6582struct proc *zfsproc; 6583 6584#ifdef illumos 6585int 6586_init(void) 6587{ 6588 int error; 6589 6590 spa_init(FREAD | FWRITE); 6591 zfs_init(); 6592 zvol_init(); 6593 zfs_ioctl_init(); 6594 6595 if ((error = mod_install(&modlinkage)) != 0) { 6596 zvol_fini(); 6597 zfs_fini(); 6598 spa_fini(); 6599 return (error); 6600 } 6601 6602 tsd_create(&zfs_fsyncer_key, NULL); 6603 tsd_create(&rrw_tsd_key, rrw_tsd_destroy); 6604 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy); 6605 6606 error = ldi_ident_from_mod(&modlinkage, &zfs_li); 6607 ASSERT(error == 0); 6608 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL); 6609 6610 return (0); 6611} 6612 6613int 6614_fini(void) 6615{ 6616 int error; 6617 6618 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled) 6619 return (SET_ERROR(EBUSY)); 6620 6621 if ((error = mod_remove(&modlinkage)) != 0) 6622 return (error); 6623 6624 zvol_fini(); 6625 zfs_fini(); 6626 spa_fini(); 6627 if (zfs_nfsshare_inited) 6628 (void) ddi_modclose(nfs_mod); 6629 if (zfs_smbshare_inited) 6630 (void) ddi_modclose(smbsrv_mod); 6631 if (zfs_nfsshare_inited || zfs_smbshare_inited) 6632 (void) ddi_modclose(sharefs_mod); 6633 6634 tsd_destroy(&zfs_fsyncer_key); 6635 ldi_ident_release(zfs_li); 6636 zfs_li = NULL; 6637 mutex_destroy(&zfs_share_lock); 6638 6639 return (error); 6640} 6641 6642int 6643_info(struct modinfo *modinfop) 6644{ 6645 return (mod_info(&modlinkage, modinfop)); 6646} 6647#endif /* illumos */ 6648 6649static int zfs__init(void); 6650static int zfs__fini(void); 6651static void zfs_shutdown(void *, int); 6652 6653static eventhandler_tag zfs_shutdown_event_tag; 6654 6655#ifdef __FreeBSD__ 6656#define ZFS_MIN_KSTACK_PAGES 4 6657#endif 6658 6659int 6660zfs__init(void) 6661{ 6662 6663#ifdef __FreeBSD__ 6664#if KSTACK_PAGES < ZFS_MIN_KSTACK_PAGES 6665 printf("ZFS NOTICE: KSTACK_PAGES is %d which could result in stack " 6666 "overflow panic!\nPlease consider adding " 6667 "'options KSTACK_PAGES=%d' to your kernel config\n", KSTACK_PAGES, 6668 ZFS_MIN_KSTACK_PAGES); 6669#endif 6670#endif 6671 zfs_root_token = root_mount_hold("ZFS"); 6672 6673 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL); 6674 6675 spa_init(FREAD | FWRITE); 6676 zfs_init(); 6677 zvol_init(); 6678 zfs_ioctl_init(); 6679 6680 tsd_create(&zfs_fsyncer_key, NULL); 6681 tsd_create(&rrw_tsd_key, rrw_tsd_destroy); 6682 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy); 6683 tsd_create(&zfs_geom_probe_vdev_key, NULL); 6684 6685 printf("ZFS storage pool version: features support (" SPA_VERSION_STRING ")\n"); 6686 root_mount_rel(zfs_root_token); 6687 6688 zfsdev_init(); 6689 6690 return (0); 6691} 6692 6693int 6694zfs__fini(void) 6695{ 6696 if (spa_busy() || zfs_busy() || zvol_busy() || 6697 zio_injection_enabled) { 6698 return (EBUSY); 6699 } 6700 6701 zfsdev_fini(); 6702 zvol_fini(); 6703 zfs_fini(); 6704 spa_fini(); 6705 6706 tsd_destroy(&zfs_fsyncer_key); 6707 tsd_destroy(&rrw_tsd_key); 6708 tsd_destroy(&zfs_allow_log_key); 6709 6710 mutex_destroy(&zfs_share_lock); 6711 6712 return (0); 6713} 6714 6715static void 6716zfs_shutdown(void *arg __unused, int howto __unused) 6717{ 6718 6719 /* 6720 * ZFS fini routines can not properly work in a panic-ed system. 6721 */ 6722 if (panicstr == NULL) 6723 (void)zfs__fini(); 6724} 6725 6726 6727static int 6728zfs_modevent(module_t mod, int type, void *unused __unused) 6729{ 6730 int err; 6731 6732 switch (type) { 6733 case MOD_LOAD: 6734 err = zfs__init(); 6735 if (err == 0) 6736 zfs_shutdown_event_tag = EVENTHANDLER_REGISTER( 6737 shutdown_post_sync, zfs_shutdown, NULL, 6738 SHUTDOWN_PRI_FIRST); 6739 return (err); 6740 case MOD_UNLOAD: 6741 err = zfs__fini(); 6742 if (err == 0 && zfs_shutdown_event_tag != NULL) 6743 EVENTHANDLER_DEREGISTER(shutdown_post_sync, 6744 zfs_shutdown_event_tag); 6745 return (err); 6746 case MOD_SHUTDOWN: 6747 return (0); 6748 default: 6749 break; 6750 } 6751 return (EOPNOTSUPP); 6752} 6753 6754static moduledata_t zfs_mod = { 6755 "zfsctrl", 6756 zfs_modevent, 6757 0 6758}; 6759DECLARE_MODULE(zfsctrl, zfs_mod, SI_SUB_VFS, SI_ORDER_ANY); 6760MODULE_VERSION(zfsctrl, 1); 6761MODULE_DEPEND(zfsctrl, opensolaris, 1, 1, 1); 6762MODULE_DEPEND(zfsctrl, krpc, 1, 1, 1); 6763MODULE_DEPEND(zfsctrl, acl_nfs4, 1, 1, 1); 6764