zfs_ioctl.c revision 297108
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 */ 35 36/* 37 * ZFS ioctls. 38 * 39 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage 40 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool. 41 * 42 * There are two ways that we handle ioctls: the legacy way where almost 43 * all of the logic is in the ioctl callback, and the new way where most 44 * of the marshalling is handled in the common entry point, zfsdev_ioctl(). 45 * 46 * Non-legacy ioctls should be registered by calling 47 * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked 48 * from userland by lzc_ioctl(). 49 * 50 * The registration arguments are as follows: 51 * 52 * const char *name 53 * The name of the ioctl. This is used for history logging. If the 54 * ioctl returns successfully (the callback returns 0), and allow_log 55 * is true, then a history log entry will be recorded with the input & 56 * output nvlists. The log entry can be printed with "zpool history -i". 57 * 58 * zfs_ioc_t ioc 59 * The ioctl request number, which userland will pass to ioctl(2). 60 * The ioctl numbers can change from release to release, because 61 * the caller (libzfs) must be matched to the kernel. 62 * 63 * zfs_secpolicy_func_t *secpolicy 64 * This function will be called before the zfs_ioc_func_t, to 65 * determine if this operation is permitted. It should return EPERM 66 * on failure, and 0 on success. Checks include determining if the 67 * dataset is visible in this zone, and if the user has either all 68 * zfs privileges in the zone (SYS_MOUNT), or has been granted permission 69 * to do this operation on this dataset with "zfs allow". 70 * 71 * zfs_ioc_namecheck_t namecheck 72 * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool 73 * name, a dataset name, or nothing. If the name is not well-formed, 74 * the ioctl will fail and the callback will not be called. 75 * Therefore, the callback can assume that the name is well-formed 76 * (e.g. is null-terminated, doesn't have more than one '@' character, 77 * doesn't have invalid characters). 78 * 79 * zfs_ioc_poolcheck_t pool_check 80 * This specifies requirements on the pool state. If the pool does 81 * not meet them (is suspended or is readonly), the ioctl will fail 82 * and the callback will not be called. If any checks are specified 83 * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME. 84 * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED | 85 * POOL_CHECK_READONLY). 86 * 87 * boolean_t smush_outnvlist 88 * If smush_outnvlist is true, then the output is presumed to be a 89 * list of errors, and it will be "smushed" down to fit into the 90 * caller's buffer, by removing some entries and replacing them with a 91 * single "N_MORE_ERRORS" entry indicating how many were removed. See 92 * nvlist_smush() for details. If smush_outnvlist is false, and the 93 * outnvlist does not fit into the userland-provided buffer, then the 94 * ioctl will fail with ENOMEM. 95 * 96 * zfs_ioc_func_t *func 97 * The callback function that will perform the operation. 98 * 99 * The callback should return 0 on success, or an error number on 100 * failure. If the function fails, the userland ioctl will return -1, 101 * and errno will be set to the callback's return value. The callback 102 * will be called with the following arguments: 103 * 104 * const char *name 105 * The name of the pool or dataset to operate on, from 106 * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the 107 * expected type (pool, dataset, or none). 108 * 109 * nvlist_t *innvl 110 * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or 111 * NULL if no input nvlist was provided. Changes to this nvlist are 112 * ignored. If the input nvlist could not be deserialized, the 113 * ioctl will fail and the callback will not be called. 114 * 115 * nvlist_t *outnvl 116 * The output nvlist, initially empty. The callback can fill it in, 117 * and it will be returned to userland by serializing it into 118 * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization 119 * fails (e.g. because the caller didn't supply a large enough 120 * buffer), then the overall ioctl will fail. See the 121 * 'smush_nvlist' argument above for additional behaviors. 122 * 123 * There are two typical uses of the output nvlist: 124 * - To return state, e.g. property values. In this case, 125 * smush_outnvlist should be false. If the buffer was not large 126 * enough, the caller will reallocate a larger buffer and try 127 * the ioctl again. 128 * 129 * - To return multiple errors from an ioctl which makes on-disk 130 * changes. In this case, smush_outnvlist should be true. 131 * Ioctls which make on-disk modifications should generally not 132 * use the outnvl if they succeed, because the caller can not 133 * distinguish between the operation failing, and 134 * deserialization failing. 135 */ 136#ifdef __FreeBSD__ 137#include "opt_kstack_pages.h" 138#endif 139 140#include <sys/types.h> 141#include <sys/param.h> 142#include <sys/systm.h> 143#include <sys/conf.h> 144#include <sys/kernel.h> 145#include <sys/lock.h> 146#include <sys/malloc.h> 147#include <sys/mutex.h> 148#include <sys/proc.h> 149#include <sys/errno.h> 150#include <sys/uio.h> 151#include <sys/buf.h> 152#include <sys/file.h> 153#include <sys/kmem.h> 154#include <sys/conf.h> 155#include <sys/cmn_err.h> 156#include <sys/stat.h> 157#include <sys/zfs_ioctl.h> 158#include <sys/zfs_vfsops.h> 159#include <sys/zfs_znode.h> 160#include <sys/zap.h> 161#include <sys/spa.h> 162#include <sys/spa_impl.h> 163#include <sys/vdev.h> 164#include <sys/dmu.h> 165#include <sys/dsl_dir.h> 166#include <sys/dsl_dataset.h> 167#include <sys/dsl_prop.h> 168#include <sys/dsl_deleg.h> 169#include <sys/dmu_objset.h> 170#include <sys/dmu_impl.h> 171#include <sys/dmu_tx.h> 172#include <sys/sunddi.h> 173#include <sys/policy.h> 174#include <sys/zone.h> 175#include <sys/nvpair.h> 176#include <sys/mount.h> 177#include <sys/taskqueue.h> 178#include <sys/sdt.h> 179#include <sys/varargs.h> 180#include <sys/fs/zfs.h> 181#include <sys/zfs_ctldir.h> 182#include <sys/zfs_dir.h> 183#include <sys/zfs_onexit.h> 184#include <sys/zvol.h> 185#include <sys/dsl_scan.h> 186#include <sys/dmu_objset.h> 187#include <sys/dmu_send.h> 188#include <sys/dsl_destroy.h> 189#include <sys/dsl_bookmark.h> 190#include <sys/dsl_userhold.h> 191#include <sys/zfeature.h> 192#include <sys/zio_checksum.h> 193 194#include "zfs_namecheck.h" 195#include "zfs_prop.h" 196#include "zfs_deleg.h" 197#include "zfs_comutil.h" 198#include "zfs_ioctl_compat.h" 199 200CTASSERT(sizeof(zfs_cmd_t) < IOCPARM_MAX); 201 202static struct cdev *zfsdev; 203 204extern void zfs_init(void); 205extern void zfs_fini(void); 206 207uint_t zfs_fsyncer_key; 208extern uint_t rrw_tsd_key; 209static uint_t zfs_allow_log_key; 210 211typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *); 212typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *); 213typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *); 214 215typedef enum { 216 NO_NAME, 217 POOL_NAME, 218 DATASET_NAME 219} zfs_ioc_namecheck_t; 220 221typedef enum { 222 POOL_CHECK_NONE = 1 << 0, 223 POOL_CHECK_SUSPENDED = 1 << 1, 224 POOL_CHECK_READONLY = 1 << 2, 225} zfs_ioc_poolcheck_t; 226 227typedef struct zfs_ioc_vec { 228 zfs_ioc_legacy_func_t *zvec_legacy_func; 229 zfs_ioc_func_t *zvec_func; 230 zfs_secpolicy_func_t *zvec_secpolicy; 231 zfs_ioc_namecheck_t zvec_namecheck; 232 boolean_t zvec_allow_log; 233 zfs_ioc_poolcheck_t zvec_pool_check; 234 boolean_t zvec_smush_outnvlist; 235 const char *zvec_name; 236} zfs_ioc_vec_t; 237 238/* This array is indexed by zfs_userquota_prop_t */ 239static const char *userquota_perms[] = { 240 ZFS_DELEG_PERM_USERUSED, 241 ZFS_DELEG_PERM_USERQUOTA, 242 ZFS_DELEG_PERM_GROUPUSED, 243 ZFS_DELEG_PERM_GROUPQUOTA, 244}; 245 246static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc); 247static int zfs_check_settable(const char *name, nvpair_t *property, 248 cred_t *cr); 249static int zfs_check_clearable(char *dataset, nvlist_t *props, 250 nvlist_t **errors); 251static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *, 252 boolean_t *); 253int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *); 254static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp); 255 256static void zfsdev_close(void *data); 257 258static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature); 259 260/* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */ 261void 262__dprintf(const char *file, const char *func, int line, const char *fmt, ...) 263{ 264 const char *newfile; 265 char buf[512]; 266 va_list adx; 267 268 /* 269 * Get rid of annoying "../common/" prefix to filename. 270 */ 271 newfile = strrchr(file, '/'); 272 if (newfile != NULL) { 273 newfile = newfile + 1; /* Get rid of leading / */ 274 } else { 275 newfile = file; 276 } 277 278 va_start(adx, fmt); 279 (void) vsnprintf(buf, sizeof (buf), fmt, adx); 280 va_end(adx); 281 282 /* 283 * To get this data, use the zfs-dprintf probe as so: 284 * dtrace -q -n 'zfs-dprintf \ 285 * /stringof(arg0) == "dbuf.c"/ \ 286 * {printf("%s: %s", stringof(arg1), stringof(arg3))}' 287 * arg0 = file name 288 * arg1 = function name 289 * arg2 = line number 290 * arg3 = message 291 */ 292 DTRACE_PROBE4(zfs__dprintf, 293 char *, newfile, char *, func, int, line, char *, buf); 294} 295 296static void 297history_str_free(char *buf) 298{ 299 kmem_free(buf, HIS_MAX_RECORD_LEN); 300} 301 302static char * 303history_str_get(zfs_cmd_t *zc) 304{ 305 char *buf; 306 307 if (zc->zc_history == 0) 308 return (NULL); 309 310 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP); 311 if (copyinstr((void *)(uintptr_t)zc->zc_history, 312 buf, HIS_MAX_RECORD_LEN, NULL) != 0) { 313 history_str_free(buf); 314 return (NULL); 315 } 316 317 buf[HIS_MAX_RECORD_LEN -1] = '\0'; 318 319 return (buf); 320} 321 322/* 323 * Check to see if the named dataset is currently defined as bootable 324 */ 325static boolean_t 326zfs_is_bootfs(const char *name) 327{ 328 objset_t *os; 329 330 if (dmu_objset_hold(name, FTAG, &os) == 0) { 331 boolean_t ret; 332 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os))); 333 dmu_objset_rele(os, FTAG); 334 return (ret); 335 } 336 return (B_FALSE); 337} 338 339/* 340 * Return non-zero if the spa version is less than requested version. 341 */ 342static int 343zfs_earlier_version(const char *name, int version) 344{ 345 spa_t *spa; 346 347 if (spa_open(name, &spa, FTAG) == 0) { 348 if (spa_version(spa) < version) { 349 spa_close(spa, FTAG); 350 return (1); 351 } 352 spa_close(spa, FTAG); 353 } 354 return (0); 355} 356 357/* 358 * Return TRUE if the ZPL version is less than requested version. 359 */ 360static boolean_t 361zpl_earlier_version(const char *name, int version) 362{ 363 objset_t *os; 364 boolean_t rc = B_TRUE; 365 366 if (dmu_objset_hold(name, FTAG, &os) == 0) { 367 uint64_t zplversion; 368 369 if (dmu_objset_type(os) != DMU_OST_ZFS) { 370 dmu_objset_rele(os, FTAG); 371 return (B_TRUE); 372 } 373 /* XXX reading from non-owned objset */ 374 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0) 375 rc = zplversion < version; 376 dmu_objset_rele(os, FTAG); 377 } 378 return (rc); 379} 380 381static void 382zfs_log_history(zfs_cmd_t *zc) 383{ 384 spa_t *spa; 385 char *buf; 386 387 if ((buf = history_str_get(zc)) == NULL) 388 return; 389 390 if (spa_open(zc->zc_name, &spa, FTAG) == 0) { 391 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY) 392 (void) spa_history_log(spa, buf); 393 spa_close(spa, FTAG); 394 } 395 history_str_free(buf); 396} 397 398/* 399 * Policy for top-level read operations (list pools). Requires no privileges, 400 * and can be used in the local zone, as there is no associated dataset. 401 */ 402/* ARGSUSED */ 403static int 404zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 405{ 406 return (0); 407} 408 409/* 410 * Policy for dataset read operations (list children, get statistics). Requires 411 * no privileges, but must be visible in the local zone. 412 */ 413/* ARGSUSED */ 414static int 415zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 416{ 417 if (INGLOBALZONE(curthread) || 418 zone_dataset_visible(zc->zc_name, NULL)) 419 return (0); 420 421 return (SET_ERROR(ENOENT)); 422} 423 424static int 425zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr) 426{ 427 int writable = 1; 428 429 /* 430 * The dataset must be visible by this zone -- check this first 431 * so they don't see EPERM on something they shouldn't know about. 432 */ 433 if (!INGLOBALZONE(curthread) && 434 !zone_dataset_visible(dataset, &writable)) 435 return (SET_ERROR(ENOENT)); 436 437 if (INGLOBALZONE(curthread)) { 438 /* 439 * If the fs is zoned, only root can access it from the 440 * global zone. 441 */ 442 if (secpolicy_zfs(cr) && zoned) 443 return (SET_ERROR(EPERM)); 444 } else { 445 /* 446 * If we are in a local zone, the 'zoned' property must be set. 447 */ 448 if (!zoned) 449 return (SET_ERROR(EPERM)); 450 451 /* must be writable by this zone */ 452 if (!writable) 453 return (SET_ERROR(EPERM)); 454 } 455 return (0); 456} 457 458static int 459zfs_dozonecheck(const char *dataset, cred_t *cr) 460{ 461 uint64_t zoned; 462 463 if (dsl_prop_get_integer(dataset, "jailed", &zoned, NULL)) 464 return (SET_ERROR(ENOENT)); 465 466 return (zfs_dozonecheck_impl(dataset, zoned, cr)); 467} 468 469static int 470zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr) 471{ 472 uint64_t zoned; 473 474 if (dsl_prop_get_int_ds(ds, "jailed", &zoned)) 475 return (SET_ERROR(ENOENT)); 476 477 return (zfs_dozonecheck_impl(dataset, zoned, cr)); 478} 479 480static int 481zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds, 482 const char *perm, cred_t *cr) 483{ 484 int error; 485 486 error = zfs_dozonecheck_ds(name, ds, cr); 487 if (error == 0) { 488 error = secpolicy_zfs(cr); 489 if (error != 0) 490 error = dsl_deleg_access_impl(ds, perm, cr); 491 } 492 return (error); 493} 494 495static int 496zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr) 497{ 498 int error; 499 dsl_dataset_t *ds; 500 dsl_pool_t *dp; 501 502 error = dsl_pool_hold(name, FTAG, &dp); 503 if (error != 0) 504 return (error); 505 506 error = dsl_dataset_hold(dp, name, FTAG, &ds); 507 if (error != 0) { 508 dsl_pool_rele(dp, FTAG); 509 return (error); 510 } 511 512 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr); 513 514 dsl_dataset_rele(ds, FTAG); 515 dsl_pool_rele(dp, FTAG); 516 return (error); 517} 518 519#ifdef SECLABEL 520/* 521 * Policy for setting the security label property. 522 * 523 * Returns 0 for success, non-zero for access and other errors. 524 */ 525static int 526zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr) 527{ 528 char ds_hexsl[MAXNAMELEN]; 529 bslabel_t ds_sl, new_sl; 530 boolean_t new_default = FALSE; 531 uint64_t zoned; 532 int needed_priv = -1; 533 int error; 534 535 /* First get the existing dataset label. */ 536 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL), 537 1, sizeof (ds_hexsl), &ds_hexsl, NULL); 538 if (error != 0) 539 return (SET_ERROR(EPERM)); 540 541 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0) 542 new_default = TRUE; 543 544 /* The label must be translatable */ 545 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0)) 546 return (SET_ERROR(EINVAL)); 547 548 /* 549 * In a non-global zone, disallow attempts to set a label that 550 * doesn't match that of the zone; otherwise no other checks 551 * are needed. 552 */ 553 if (!INGLOBALZONE(curproc)) { 554 if (new_default || !blequal(&new_sl, CR_SL(CRED()))) 555 return (SET_ERROR(EPERM)); 556 return (0); 557 } 558 559 /* 560 * For global-zone datasets (i.e., those whose zoned property is 561 * "off", verify that the specified new label is valid for the 562 * global zone. 563 */ 564 if (dsl_prop_get_integer(name, 565 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL)) 566 return (SET_ERROR(EPERM)); 567 if (!zoned) { 568 if (zfs_check_global_label(name, strval) != 0) 569 return (SET_ERROR(EPERM)); 570 } 571 572 /* 573 * If the existing dataset label is nondefault, check if the 574 * dataset is mounted (label cannot be changed while mounted). 575 * Get the zfsvfs; if there isn't one, then the dataset isn't 576 * mounted (or isn't a dataset, doesn't exist, ...). 577 */ 578 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) { 579 objset_t *os; 580 static char *setsl_tag = "setsl_tag"; 581 582 /* 583 * Try to own the dataset; abort if there is any error, 584 * (e.g., already mounted, in use, or other error). 585 */ 586 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE, 587 setsl_tag, &os); 588 if (error != 0) 589 return (SET_ERROR(EPERM)); 590 591 dmu_objset_disown(os, setsl_tag); 592 593 if (new_default) { 594 needed_priv = PRIV_FILE_DOWNGRADE_SL; 595 goto out_check; 596 } 597 598 if (hexstr_to_label(strval, &new_sl) != 0) 599 return (SET_ERROR(EPERM)); 600 601 if (blstrictdom(&ds_sl, &new_sl)) 602 needed_priv = PRIV_FILE_DOWNGRADE_SL; 603 else if (blstrictdom(&new_sl, &ds_sl)) 604 needed_priv = PRIV_FILE_UPGRADE_SL; 605 } else { 606 /* dataset currently has a default label */ 607 if (!new_default) 608 needed_priv = PRIV_FILE_UPGRADE_SL; 609 } 610 611out_check: 612 if (needed_priv != -1) 613 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL)); 614 return (0); 615} 616#endif /* SECLABEL */ 617 618static int 619zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval, 620 cred_t *cr) 621{ 622 char *strval; 623 624 /* 625 * Check permissions for special properties. 626 */ 627 switch (prop) { 628 case ZFS_PROP_ZONED: 629 /* 630 * Disallow setting of 'zoned' from within a local zone. 631 */ 632 if (!INGLOBALZONE(curthread)) 633 return (SET_ERROR(EPERM)); 634 break; 635 636 case ZFS_PROP_QUOTA: 637 case ZFS_PROP_FILESYSTEM_LIMIT: 638 case ZFS_PROP_SNAPSHOT_LIMIT: 639 if (!INGLOBALZONE(curthread)) { 640 uint64_t zoned; 641 char setpoint[MAXNAMELEN]; 642 /* 643 * Unprivileged users are allowed to modify the 644 * limit on things *under* (ie. contained by) 645 * the thing they own. 646 */ 647 if (dsl_prop_get_integer(dsname, "jailed", &zoned, 648 setpoint)) 649 return (SET_ERROR(EPERM)); 650 if (!zoned || strlen(dsname) <= strlen(setpoint)) 651 return (SET_ERROR(EPERM)); 652 } 653 break; 654 655 case ZFS_PROP_MLSLABEL: 656#ifdef SECLABEL 657 if (!is_system_labeled()) 658 return (SET_ERROR(EPERM)); 659 660 if (nvpair_value_string(propval, &strval) == 0) { 661 int err; 662 663 err = zfs_set_slabel_policy(dsname, strval, CRED()); 664 if (err != 0) 665 return (err); 666 } 667#else 668 return (EOPNOTSUPP); 669#endif 670 break; 671 } 672 673 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr)); 674} 675 676/* ARGSUSED */ 677static int 678zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 679{ 680 int error; 681 682 error = zfs_dozonecheck(zc->zc_name, cr); 683 if (error != 0) 684 return (error); 685 686 /* 687 * permission to set permissions will be evaluated later in 688 * dsl_deleg_can_allow() 689 */ 690 return (0); 691} 692 693/* ARGSUSED */ 694static int 695zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 696{ 697 return (zfs_secpolicy_write_perms(zc->zc_name, 698 ZFS_DELEG_PERM_ROLLBACK, cr)); 699} 700 701/* ARGSUSED */ 702static int 703zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 704{ 705 dsl_pool_t *dp; 706 dsl_dataset_t *ds; 707 char *cp; 708 int error; 709 710 /* 711 * Generate the current snapshot name from the given objsetid, then 712 * use that name for the secpolicy/zone checks. 713 */ 714 cp = strchr(zc->zc_name, '@'); 715 if (cp == NULL) 716 return (SET_ERROR(EINVAL)); 717 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 718 if (error != 0) 719 return (error); 720 721 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds); 722 if (error != 0) { 723 dsl_pool_rele(dp, FTAG); 724 return (error); 725 } 726 727 dsl_dataset_name(ds, zc->zc_name); 728 729 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds, 730 ZFS_DELEG_PERM_SEND, cr); 731 dsl_dataset_rele(ds, FTAG); 732 dsl_pool_rele(dp, FTAG); 733 734 return (error); 735} 736 737/* ARGSUSED */ 738static int 739zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 740{ 741 return (zfs_secpolicy_write_perms(zc->zc_name, 742 ZFS_DELEG_PERM_SEND, cr)); 743} 744 745/* ARGSUSED */ 746static int 747zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 748{ 749 vnode_t *vp; 750 int error; 751 752 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, 753 NO_FOLLOW, NULL, &vp)) != 0) 754 return (error); 755 756 /* Now make sure mntpnt and dataset are ZFS */ 757 758 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 || 759 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), 760 zc->zc_name) != 0)) { 761 VN_RELE(vp); 762 return (SET_ERROR(EPERM)); 763 } 764 765 VN_RELE(vp); 766 return (dsl_deleg_access(zc->zc_name, 767 ZFS_DELEG_PERM_SHARE, cr)); 768} 769 770int 771zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 772{ 773 if (!INGLOBALZONE(curthread)) 774 return (SET_ERROR(EPERM)); 775 776 if (secpolicy_nfs(cr) == 0) { 777 return (0); 778 } else { 779 return (zfs_secpolicy_deleg_share(zc, innvl, cr)); 780 } 781} 782 783int 784zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 785{ 786 if (!INGLOBALZONE(curthread)) 787 return (SET_ERROR(EPERM)); 788 789 if (secpolicy_smb(cr) == 0) { 790 return (0); 791 } else { 792 return (zfs_secpolicy_deleg_share(zc, innvl, cr)); 793 } 794} 795 796static int 797zfs_get_parent(const char *datasetname, char *parent, int parentsize) 798{ 799 char *cp; 800 801 /* 802 * Remove the @bla or /bla from the end of the name to get the parent. 803 */ 804 (void) strncpy(parent, datasetname, parentsize); 805 cp = strrchr(parent, '@'); 806 if (cp != NULL) { 807 cp[0] = '\0'; 808 } else { 809 cp = strrchr(parent, '/'); 810 if (cp == NULL) 811 return (SET_ERROR(ENOENT)); 812 cp[0] = '\0'; 813 } 814 815 return (0); 816} 817 818int 819zfs_secpolicy_destroy_perms(const char *name, cred_t *cr) 820{ 821 int error; 822 823 if ((error = zfs_secpolicy_write_perms(name, 824 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 825 return (error); 826 827 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr)); 828} 829 830/* ARGSUSED */ 831static int 832zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 833{ 834 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr)); 835} 836 837/* 838 * Destroying snapshots with delegated permissions requires 839 * descendant mount and destroy permissions. 840 */ 841/* ARGSUSED */ 842static int 843zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 844{ 845 nvlist_t *snaps; 846 nvpair_t *pair, *nextpair; 847 int error = 0; 848 849 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 850 return (SET_ERROR(EINVAL)); 851 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 852 pair = nextpair) { 853 nextpair = nvlist_next_nvpair(snaps, pair); 854 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr); 855 if (error == ENOENT) { 856 /* 857 * Ignore any snapshots that don't exist (we consider 858 * them "already destroyed"). Remove the name from the 859 * nvl here in case the snapshot is created between 860 * now and when we try to destroy it (in which case 861 * we don't want to destroy it since we haven't 862 * checked for permission). 863 */ 864 fnvlist_remove_nvpair(snaps, pair); 865 error = 0; 866 } 867 if (error != 0) 868 break; 869 } 870 871 return (error); 872} 873 874int 875zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr) 876{ 877 char parentname[MAXNAMELEN]; 878 int error; 879 880 if ((error = zfs_secpolicy_write_perms(from, 881 ZFS_DELEG_PERM_RENAME, cr)) != 0) 882 return (error); 883 884 if ((error = zfs_secpolicy_write_perms(from, 885 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 886 return (error); 887 888 if ((error = zfs_get_parent(to, parentname, 889 sizeof (parentname))) != 0) 890 return (error); 891 892 if ((error = zfs_secpolicy_write_perms(parentname, 893 ZFS_DELEG_PERM_CREATE, cr)) != 0) 894 return (error); 895 896 if ((error = zfs_secpolicy_write_perms(parentname, 897 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 898 return (error); 899 900 return (error); 901} 902 903/* ARGSUSED */ 904static int 905zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 906{ 907 char *at = NULL; 908 int error; 909 910 if ((zc->zc_cookie & 1) != 0) { 911 /* 912 * This is recursive rename, so the starting snapshot might 913 * not exist. Check file system or volume permission instead. 914 */ 915 at = strchr(zc->zc_name, '@'); 916 if (at == NULL) 917 return (EINVAL); 918 *at = '\0'; 919 } 920 921 error = zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr); 922 923 if (at != NULL) 924 *at = '@'; 925 926 return (error); 927} 928 929/* ARGSUSED */ 930static int 931zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 932{ 933 dsl_pool_t *dp; 934 dsl_dataset_t *clone; 935 int error; 936 937 error = zfs_secpolicy_write_perms(zc->zc_name, 938 ZFS_DELEG_PERM_PROMOTE, cr); 939 if (error != 0) 940 return (error); 941 942 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 943 if (error != 0) 944 return (error); 945 946 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone); 947 948 if (error == 0) { 949 char parentname[MAXNAMELEN]; 950 dsl_dataset_t *origin = NULL; 951 dsl_dir_t *dd; 952 dd = clone->ds_dir; 953 954 error = dsl_dataset_hold_obj(dd->dd_pool, 955 dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin); 956 if (error != 0) { 957 dsl_dataset_rele(clone, FTAG); 958 dsl_pool_rele(dp, FTAG); 959 return (error); 960 } 961 962 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone, 963 ZFS_DELEG_PERM_MOUNT, cr); 964 965 dsl_dataset_name(origin, parentname); 966 if (error == 0) { 967 error = zfs_secpolicy_write_perms_ds(parentname, origin, 968 ZFS_DELEG_PERM_PROMOTE, cr); 969 } 970 dsl_dataset_rele(clone, FTAG); 971 dsl_dataset_rele(origin, FTAG); 972 } 973 dsl_pool_rele(dp, FTAG); 974 return (error); 975} 976 977/* ARGSUSED */ 978static int 979zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 980{ 981 int error; 982 983 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 984 ZFS_DELEG_PERM_RECEIVE, cr)) != 0) 985 return (error); 986 987 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 988 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 989 return (error); 990 991 return (zfs_secpolicy_write_perms(zc->zc_name, 992 ZFS_DELEG_PERM_CREATE, cr)); 993} 994 995int 996zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr) 997{ 998 return (zfs_secpolicy_write_perms(name, 999 ZFS_DELEG_PERM_SNAPSHOT, cr)); 1000} 1001 1002/* 1003 * Check for permission to create each snapshot in the nvlist. 1004 */ 1005/* ARGSUSED */ 1006static int 1007zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1008{ 1009 nvlist_t *snaps; 1010 int error; 1011 nvpair_t *pair; 1012 1013 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 1014 return (SET_ERROR(EINVAL)); 1015 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 1016 pair = nvlist_next_nvpair(snaps, pair)) { 1017 char *name = nvpair_name(pair); 1018 char *atp = strchr(name, '@'); 1019 1020 if (atp == NULL) { 1021 error = SET_ERROR(EINVAL); 1022 break; 1023 } 1024 *atp = '\0'; 1025 error = zfs_secpolicy_snapshot_perms(name, cr); 1026 *atp = '@'; 1027 if (error != 0) 1028 break; 1029 } 1030 return (error); 1031} 1032 1033/* 1034 * Check for permission to create each snapshot in the nvlist. 1035 */ 1036/* ARGSUSED */ 1037static int 1038zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1039{ 1040 int error = 0; 1041 1042 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 1043 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 1044 char *name = nvpair_name(pair); 1045 char *hashp = strchr(name, '#'); 1046 1047 if (hashp == NULL) { 1048 error = SET_ERROR(EINVAL); 1049 break; 1050 } 1051 *hashp = '\0'; 1052 error = zfs_secpolicy_write_perms(name, 1053 ZFS_DELEG_PERM_BOOKMARK, cr); 1054 *hashp = '#'; 1055 if (error != 0) 1056 break; 1057 } 1058 return (error); 1059} 1060 1061/* ARGSUSED */ 1062static int 1063zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1064{ 1065 nvpair_t *pair, *nextpair; 1066 int error = 0; 1067 1068 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL; 1069 pair = nextpair) { 1070 char *name = nvpair_name(pair); 1071 char *hashp = strchr(name, '#'); 1072 nextpair = nvlist_next_nvpair(innvl, pair); 1073 1074 if (hashp == NULL) { 1075 error = SET_ERROR(EINVAL); 1076 break; 1077 } 1078 1079 *hashp = '\0'; 1080 error = zfs_secpolicy_write_perms(name, 1081 ZFS_DELEG_PERM_DESTROY, cr); 1082 *hashp = '#'; 1083 if (error == ENOENT) { 1084 /* 1085 * Ignore any filesystems that don't exist (we consider 1086 * their bookmarks "already destroyed"). Remove 1087 * the name from the nvl here in case the filesystem 1088 * is created between now and when we try to destroy 1089 * the bookmark (in which case we don't want to 1090 * destroy it since we haven't checked for permission). 1091 */ 1092 fnvlist_remove_nvpair(innvl, pair); 1093 error = 0; 1094 } 1095 if (error != 0) 1096 break; 1097 } 1098 1099 return (error); 1100} 1101 1102/* ARGSUSED */ 1103static int 1104zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1105{ 1106 /* 1107 * Even root must have a proper TSD so that we know what pool 1108 * to log to. 1109 */ 1110 if (tsd_get(zfs_allow_log_key) == NULL) 1111 return (SET_ERROR(EPERM)); 1112 return (0); 1113} 1114 1115static int 1116zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1117{ 1118 char parentname[MAXNAMELEN]; 1119 int error; 1120 char *origin; 1121 1122 if ((error = zfs_get_parent(zc->zc_name, parentname, 1123 sizeof (parentname))) != 0) 1124 return (error); 1125 1126 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 && 1127 (error = zfs_secpolicy_write_perms(origin, 1128 ZFS_DELEG_PERM_CLONE, cr)) != 0) 1129 return (error); 1130 1131 if ((error = zfs_secpolicy_write_perms(parentname, 1132 ZFS_DELEG_PERM_CREATE, cr)) != 0) 1133 return (error); 1134 1135 return (zfs_secpolicy_write_perms(parentname, 1136 ZFS_DELEG_PERM_MOUNT, cr)); 1137} 1138 1139/* 1140 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires 1141 * SYS_CONFIG privilege, which is not available in a local zone. 1142 */ 1143/* ARGSUSED */ 1144static int 1145zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1146{ 1147 if (secpolicy_sys_config(cr, B_FALSE) != 0) 1148 return (SET_ERROR(EPERM)); 1149 1150 return (0); 1151} 1152 1153/* 1154 * Policy for object to name lookups. 1155 */ 1156/* ARGSUSED */ 1157static int 1158zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1159{ 1160 int error; 1161 1162 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0) 1163 return (0); 1164 1165 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr); 1166 return (error); 1167} 1168 1169/* 1170 * Policy for fault injection. Requires all privileges. 1171 */ 1172/* ARGSUSED */ 1173static int 1174zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1175{ 1176 return (secpolicy_zinject(cr)); 1177} 1178 1179/* ARGSUSED */ 1180static int 1181zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1182{ 1183 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value); 1184 1185 if (prop == ZPROP_INVAL) { 1186 if (!zfs_prop_user(zc->zc_value)) 1187 return (SET_ERROR(EINVAL)); 1188 return (zfs_secpolicy_write_perms(zc->zc_name, 1189 ZFS_DELEG_PERM_USERPROP, cr)); 1190 } else { 1191 return (zfs_secpolicy_setprop(zc->zc_name, prop, 1192 NULL, cr)); 1193 } 1194} 1195 1196static int 1197zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1198{ 1199 int err = zfs_secpolicy_read(zc, innvl, cr); 1200 if (err) 1201 return (err); 1202 1203 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 1204 return (SET_ERROR(EINVAL)); 1205 1206 if (zc->zc_value[0] == 0) { 1207 /* 1208 * They are asking about a posix uid/gid. If it's 1209 * themself, allow it. 1210 */ 1211 if (zc->zc_objset_type == ZFS_PROP_USERUSED || 1212 zc->zc_objset_type == ZFS_PROP_USERQUOTA) { 1213 if (zc->zc_guid == crgetuid(cr)) 1214 return (0); 1215 } else { 1216 if (groupmember(zc->zc_guid, cr)) 1217 return (0); 1218 } 1219 } 1220 1221 return (zfs_secpolicy_write_perms(zc->zc_name, 1222 userquota_perms[zc->zc_objset_type], cr)); 1223} 1224 1225static int 1226zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1227{ 1228 int err = zfs_secpolicy_read(zc, innvl, cr); 1229 if (err) 1230 return (err); 1231 1232 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 1233 return (SET_ERROR(EINVAL)); 1234 1235 return (zfs_secpolicy_write_perms(zc->zc_name, 1236 userquota_perms[zc->zc_objset_type], cr)); 1237} 1238 1239/* ARGSUSED */ 1240static int 1241zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1242{ 1243 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION, 1244 NULL, cr)); 1245} 1246 1247/* ARGSUSED */ 1248static int 1249zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1250{ 1251 nvpair_t *pair; 1252 nvlist_t *holds; 1253 int error; 1254 1255 error = nvlist_lookup_nvlist(innvl, "holds", &holds); 1256 if (error != 0) 1257 return (SET_ERROR(EINVAL)); 1258 1259 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL; 1260 pair = nvlist_next_nvpair(holds, pair)) { 1261 char fsname[MAXNAMELEN]; 1262 error = dmu_fsname(nvpair_name(pair), fsname); 1263 if (error != 0) 1264 return (error); 1265 error = zfs_secpolicy_write_perms(fsname, 1266 ZFS_DELEG_PERM_HOLD, cr); 1267 if (error != 0) 1268 return (error); 1269 } 1270 return (0); 1271} 1272 1273/* ARGSUSED */ 1274static int 1275zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1276{ 1277 nvpair_t *pair; 1278 int error; 1279 1280 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL; 1281 pair = nvlist_next_nvpair(innvl, pair)) { 1282 char fsname[MAXNAMELEN]; 1283 error = dmu_fsname(nvpair_name(pair), fsname); 1284 if (error != 0) 1285 return (error); 1286 error = zfs_secpolicy_write_perms(fsname, 1287 ZFS_DELEG_PERM_RELEASE, cr); 1288 if (error != 0) 1289 return (error); 1290 } 1291 return (0); 1292} 1293 1294/* 1295 * Policy for allowing temporary snapshots to be taken or released 1296 */ 1297static int 1298zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1299{ 1300 /* 1301 * A temporary snapshot is the same as a snapshot, 1302 * hold, destroy and release all rolled into one. 1303 * Delegated diff alone is sufficient that we allow this. 1304 */ 1305 int error; 1306 1307 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 1308 ZFS_DELEG_PERM_DIFF, cr)) == 0) 1309 return (0); 1310 1311 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr); 1312 if (error == 0) 1313 error = zfs_secpolicy_hold(zc, innvl, cr); 1314 if (error == 0) 1315 error = zfs_secpolicy_release(zc, innvl, cr); 1316 if (error == 0) 1317 error = zfs_secpolicy_destroy(zc, innvl, cr); 1318 return (error); 1319} 1320 1321/* 1322 * Returns the nvlist as specified by the user in the zfs_cmd_t. 1323 */ 1324static int 1325get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp) 1326{ 1327 char *packed; 1328 int error; 1329 nvlist_t *list = NULL; 1330 1331 /* 1332 * Read in and unpack the user-supplied nvlist. 1333 */ 1334 if (size == 0) 1335 return (SET_ERROR(EINVAL)); 1336 1337 packed = kmem_alloc(size, KM_SLEEP); 1338 1339 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size, 1340 iflag)) != 0) { 1341 kmem_free(packed, size); 1342 return (SET_ERROR(EFAULT)); 1343 } 1344 1345 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) { 1346 kmem_free(packed, size); 1347 return (error); 1348 } 1349 1350 kmem_free(packed, size); 1351 1352 *nvp = list; 1353 return (0); 1354} 1355 1356/* 1357 * Reduce the size of this nvlist until it can be serialized in 'max' bytes. 1358 * Entries will be removed from the end of the nvlist, and one int32 entry 1359 * named "N_MORE_ERRORS" will be added indicating how many entries were 1360 * removed. 1361 */ 1362static int 1363nvlist_smush(nvlist_t *errors, size_t max) 1364{ 1365 size_t size; 1366 1367 size = fnvlist_size(errors); 1368 1369 if (size > max) { 1370 nvpair_t *more_errors; 1371 int n = 0; 1372 1373 if (max < 1024) 1374 return (SET_ERROR(ENOMEM)); 1375 1376 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0); 1377 more_errors = nvlist_prev_nvpair(errors, NULL); 1378 1379 do { 1380 nvpair_t *pair = nvlist_prev_nvpair(errors, 1381 more_errors); 1382 fnvlist_remove_nvpair(errors, pair); 1383 n++; 1384 size = fnvlist_size(errors); 1385 } while (size > max); 1386 1387 fnvlist_remove_nvpair(errors, more_errors); 1388 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n); 1389 ASSERT3U(fnvlist_size(errors), <=, max); 1390 } 1391 1392 return (0); 1393} 1394 1395static int 1396put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl) 1397{ 1398 char *packed = NULL; 1399 int error = 0; 1400 size_t size; 1401 1402 size = fnvlist_size(nvl); 1403 1404 if (size > zc->zc_nvlist_dst_size) { 1405 /* 1406 * Solaris returns ENOMEM here, because even if an error is 1407 * returned from an ioctl(2), new zc_nvlist_dst_size will be 1408 * passed to the userland. This is not the case for FreeBSD. 1409 * We need to return 0, so the kernel will copy the 1410 * zc_nvlist_dst_size back and the userland can discover that a 1411 * bigger buffer is needed. 1412 */ 1413 error = 0; 1414 } else { 1415 packed = fnvlist_pack(nvl, &size); 1416 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst, 1417 size, zc->zc_iflags) != 0) 1418 error = SET_ERROR(EFAULT); 1419 fnvlist_pack_free(packed, size); 1420 } 1421 1422 zc->zc_nvlist_dst_size = size; 1423 zc->zc_nvlist_dst_filled = B_TRUE; 1424 return (error); 1425} 1426 1427static int 1428getzfsvfs(const char *dsname, zfsvfs_t **zfvp) 1429{ 1430 objset_t *os; 1431 int error; 1432 1433 error = dmu_objset_hold(dsname, FTAG, &os); 1434 if (error != 0) 1435 return (error); 1436 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1437 dmu_objset_rele(os, FTAG); 1438 return (SET_ERROR(EINVAL)); 1439 } 1440 1441 mutex_enter(&os->os_user_ptr_lock); 1442 *zfvp = dmu_objset_get_user(os); 1443 if (*zfvp) { 1444 VFS_HOLD((*zfvp)->z_vfs); 1445 } else { 1446 error = SET_ERROR(ESRCH); 1447 } 1448 mutex_exit(&os->os_user_ptr_lock); 1449 dmu_objset_rele(os, FTAG); 1450 return (error); 1451} 1452 1453/* 1454 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which 1455 * case its z_vfs will be NULL, and it will be opened as the owner. 1456 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER, 1457 * which prevents all vnode ops from running. 1458 */ 1459static int 1460zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer) 1461{ 1462 int error = 0; 1463 1464 if (getzfsvfs(name, zfvp) != 0) 1465 error = zfsvfs_create(name, zfvp); 1466 if (error == 0) { 1467 rrm_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER : 1468 RW_READER, tag); 1469 if ((*zfvp)->z_unmounted) { 1470 /* 1471 * XXX we could probably try again, since the unmounting 1472 * thread should be just about to disassociate the 1473 * objset from the zfsvfs. 1474 */ 1475 rrm_exit(&(*zfvp)->z_teardown_lock, tag); 1476 return (SET_ERROR(EBUSY)); 1477 } 1478 } 1479 return (error); 1480} 1481 1482static void 1483zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag) 1484{ 1485 rrm_exit(&zfsvfs->z_teardown_lock, tag); 1486 1487 if (zfsvfs->z_vfs) { 1488 VFS_RELE(zfsvfs->z_vfs); 1489 } else { 1490 dmu_objset_disown(zfsvfs->z_os, zfsvfs); 1491 zfsvfs_free(zfsvfs); 1492 } 1493} 1494 1495static int 1496zfs_ioc_pool_create(zfs_cmd_t *zc) 1497{ 1498 int error; 1499 nvlist_t *config, *props = NULL; 1500 nvlist_t *rootprops = NULL; 1501 nvlist_t *zplprops = NULL; 1502 1503 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1504 zc->zc_iflags, &config)) 1505 return (error); 1506 1507 if (zc->zc_nvlist_src_size != 0 && (error = 1508 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1509 zc->zc_iflags, &props))) { 1510 nvlist_free(config); 1511 return (error); 1512 } 1513 1514 if (props) { 1515 nvlist_t *nvl = NULL; 1516 uint64_t version = SPA_VERSION; 1517 1518 (void) nvlist_lookup_uint64(props, 1519 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version); 1520 if (!SPA_VERSION_IS_SUPPORTED(version)) { 1521 error = SET_ERROR(EINVAL); 1522 goto pool_props_bad; 1523 } 1524 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl); 1525 if (nvl) { 1526 error = nvlist_dup(nvl, &rootprops, KM_SLEEP); 1527 if (error != 0) { 1528 nvlist_free(config); 1529 nvlist_free(props); 1530 return (error); 1531 } 1532 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS); 1533 } 1534 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 1535 error = zfs_fill_zplprops_root(version, rootprops, 1536 zplprops, NULL); 1537 if (error != 0) 1538 goto pool_props_bad; 1539 } 1540 1541 error = spa_create(zc->zc_name, config, props, zplprops); 1542 1543 /* 1544 * Set the remaining root properties 1545 */ 1546 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name, 1547 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0) 1548 (void) spa_destroy(zc->zc_name); 1549 1550pool_props_bad: 1551 nvlist_free(rootprops); 1552 nvlist_free(zplprops); 1553 nvlist_free(config); 1554 nvlist_free(props); 1555 1556 return (error); 1557} 1558 1559static int 1560zfs_ioc_pool_destroy(zfs_cmd_t *zc) 1561{ 1562 int error; 1563 zfs_log_history(zc); 1564 error = spa_destroy(zc->zc_name); 1565 if (error == 0) 1566 zvol_remove_minors(zc->zc_name); 1567 return (error); 1568} 1569 1570static int 1571zfs_ioc_pool_import(zfs_cmd_t *zc) 1572{ 1573 nvlist_t *config, *props = NULL; 1574 uint64_t guid; 1575 int error; 1576 1577 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1578 zc->zc_iflags, &config)) != 0) 1579 return (error); 1580 1581 if (zc->zc_nvlist_src_size != 0 && (error = 1582 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1583 zc->zc_iflags, &props))) { 1584 nvlist_free(config); 1585 return (error); 1586 } 1587 1588 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 || 1589 guid != zc->zc_guid) 1590 error = SET_ERROR(EINVAL); 1591 else 1592 error = spa_import(zc->zc_name, config, props, zc->zc_cookie); 1593 1594 if (zc->zc_nvlist_dst != 0) { 1595 int err; 1596 1597 if ((err = put_nvlist(zc, config)) != 0) 1598 error = err; 1599 } 1600 1601 nvlist_free(config); 1602 1603 if (props) 1604 nvlist_free(props); 1605 1606 return (error); 1607} 1608 1609static int 1610zfs_ioc_pool_export(zfs_cmd_t *zc) 1611{ 1612 int error; 1613 boolean_t force = (boolean_t)zc->zc_cookie; 1614 boolean_t hardforce = (boolean_t)zc->zc_guid; 1615 1616 zfs_log_history(zc); 1617 error = spa_export(zc->zc_name, NULL, force, hardforce); 1618 if (error == 0) 1619 zvol_remove_minors(zc->zc_name); 1620 return (error); 1621} 1622 1623static int 1624zfs_ioc_pool_configs(zfs_cmd_t *zc) 1625{ 1626 nvlist_t *configs; 1627 int error; 1628 1629 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL) 1630 return (SET_ERROR(EEXIST)); 1631 1632 error = put_nvlist(zc, configs); 1633 1634 nvlist_free(configs); 1635 1636 return (error); 1637} 1638 1639/* 1640 * inputs: 1641 * zc_name name of the pool 1642 * 1643 * outputs: 1644 * zc_cookie real errno 1645 * zc_nvlist_dst config nvlist 1646 * zc_nvlist_dst_size size of config nvlist 1647 */ 1648static int 1649zfs_ioc_pool_stats(zfs_cmd_t *zc) 1650{ 1651 nvlist_t *config; 1652 int error; 1653 int ret = 0; 1654 1655 error = spa_get_stats(zc->zc_name, &config, zc->zc_value, 1656 sizeof (zc->zc_value)); 1657 1658 if (config != NULL) { 1659 ret = put_nvlist(zc, config); 1660 nvlist_free(config); 1661 1662 /* 1663 * The config may be present even if 'error' is non-zero. 1664 * In this case we return success, and preserve the real errno 1665 * in 'zc_cookie'. 1666 */ 1667 zc->zc_cookie = error; 1668 } else { 1669 ret = error; 1670 } 1671 1672 return (ret); 1673} 1674 1675/* 1676 * Try to import the given pool, returning pool stats as appropriate so that 1677 * user land knows which devices are available and overall pool health. 1678 */ 1679static int 1680zfs_ioc_pool_tryimport(zfs_cmd_t *zc) 1681{ 1682 nvlist_t *tryconfig, *config; 1683 int error; 1684 1685 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1686 zc->zc_iflags, &tryconfig)) != 0) 1687 return (error); 1688 1689 config = spa_tryimport(tryconfig); 1690 1691 nvlist_free(tryconfig); 1692 1693 if (config == NULL) 1694 return (SET_ERROR(EINVAL)); 1695 1696 error = put_nvlist(zc, config); 1697 nvlist_free(config); 1698 1699 return (error); 1700} 1701 1702/* 1703 * inputs: 1704 * zc_name name of the pool 1705 * zc_cookie scan func (pool_scan_func_t) 1706 */ 1707static int 1708zfs_ioc_pool_scan(zfs_cmd_t *zc) 1709{ 1710 spa_t *spa; 1711 int error; 1712 1713 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1714 return (error); 1715 1716 if (zc->zc_cookie == POOL_SCAN_NONE) 1717 error = spa_scan_stop(spa); 1718 else 1719 error = spa_scan(spa, zc->zc_cookie); 1720 1721 spa_close(spa, FTAG); 1722 1723 return (error); 1724} 1725 1726static int 1727zfs_ioc_pool_freeze(zfs_cmd_t *zc) 1728{ 1729 spa_t *spa; 1730 int error; 1731 1732 error = spa_open(zc->zc_name, &spa, FTAG); 1733 if (error == 0) { 1734 spa_freeze(spa); 1735 spa_close(spa, FTAG); 1736 } 1737 return (error); 1738} 1739 1740static int 1741zfs_ioc_pool_upgrade(zfs_cmd_t *zc) 1742{ 1743 spa_t *spa; 1744 int error; 1745 1746 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1747 return (error); 1748 1749 if (zc->zc_cookie < spa_version(spa) || 1750 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) { 1751 spa_close(spa, FTAG); 1752 return (SET_ERROR(EINVAL)); 1753 } 1754 1755 spa_upgrade(spa, zc->zc_cookie); 1756 spa_close(spa, FTAG); 1757 1758 return (error); 1759} 1760 1761static int 1762zfs_ioc_pool_get_history(zfs_cmd_t *zc) 1763{ 1764 spa_t *spa; 1765 char *hist_buf; 1766 uint64_t size; 1767 int error; 1768 1769 if ((size = zc->zc_history_len) == 0) 1770 return (SET_ERROR(EINVAL)); 1771 1772 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1773 return (error); 1774 1775 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) { 1776 spa_close(spa, FTAG); 1777 return (SET_ERROR(ENOTSUP)); 1778 } 1779 1780 hist_buf = kmem_alloc(size, KM_SLEEP); 1781 if ((error = spa_history_get(spa, &zc->zc_history_offset, 1782 &zc->zc_history_len, hist_buf)) == 0) { 1783 error = ddi_copyout(hist_buf, 1784 (void *)(uintptr_t)zc->zc_history, 1785 zc->zc_history_len, zc->zc_iflags); 1786 } 1787 1788 spa_close(spa, FTAG); 1789 kmem_free(hist_buf, size); 1790 return (error); 1791} 1792 1793static int 1794zfs_ioc_pool_reguid(zfs_cmd_t *zc) 1795{ 1796 spa_t *spa; 1797 int error; 1798 1799 error = spa_open(zc->zc_name, &spa, FTAG); 1800 if (error == 0) { 1801 error = spa_change_guid(spa); 1802 spa_close(spa, FTAG); 1803 } 1804 return (error); 1805} 1806 1807static int 1808zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc) 1809{ 1810 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value)); 1811} 1812 1813/* 1814 * inputs: 1815 * zc_name name of filesystem 1816 * zc_obj object to find 1817 * 1818 * outputs: 1819 * zc_value name of object 1820 */ 1821static int 1822zfs_ioc_obj_to_path(zfs_cmd_t *zc) 1823{ 1824 objset_t *os; 1825 int error; 1826 1827 /* XXX reading from objset not owned */ 1828 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0) 1829 return (error); 1830 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1831 dmu_objset_rele(os, FTAG); 1832 return (SET_ERROR(EINVAL)); 1833 } 1834 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value, 1835 sizeof (zc->zc_value)); 1836 dmu_objset_rele(os, FTAG); 1837 1838 return (error); 1839} 1840 1841/* 1842 * inputs: 1843 * zc_name name of filesystem 1844 * zc_obj object to find 1845 * 1846 * outputs: 1847 * zc_stat stats on object 1848 * zc_value path to object 1849 */ 1850static int 1851zfs_ioc_obj_to_stats(zfs_cmd_t *zc) 1852{ 1853 objset_t *os; 1854 int error; 1855 1856 /* XXX reading from objset not owned */ 1857 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0) 1858 return (error); 1859 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1860 dmu_objset_rele(os, FTAG); 1861 return (SET_ERROR(EINVAL)); 1862 } 1863 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value, 1864 sizeof (zc->zc_value)); 1865 dmu_objset_rele(os, FTAG); 1866 1867 return (error); 1868} 1869 1870static int 1871zfs_ioc_vdev_add(zfs_cmd_t *zc) 1872{ 1873 spa_t *spa; 1874 int error; 1875 nvlist_t *config, **l2cache, **spares; 1876 uint_t nl2cache = 0, nspares = 0; 1877 1878 error = spa_open(zc->zc_name, &spa, FTAG); 1879 if (error != 0) 1880 return (error); 1881 1882 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1883 zc->zc_iflags, &config); 1884 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE, 1885 &l2cache, &nl2cache); 1886 1887 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES, 1888 &spares, &nspares); 1889 1890#ifdef illumos 1891 /* 1892 * A root pool with concatenated devices is not supported. 1893 * Thus, can not add a device to a root pool. 1894 * 1895 * Intent log device can not be added to a rootpool because 1896 * during mountroot, zil is replayed, a seperated log device 1897 * can not be accessed during the mountroot time. 1898 * 1899 * l2cache and spare devices are ok to be added to a rootpool. 1900 */ 1901 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) { 1902 nvlist_free(config); 1903 spa_close(spa, FTAG); 1904 return (SET_ERROR(EDOM)); 1905 } 1906#endif /* illumos */ 1907 1908 if (error == 0) { 1909 error = spa_vdev_add(spa, config); 1910 nvlist_free(config); 1911 } 1912 spa_close(spa, FTAG); 1913 return (error); 1914} 1915 1916/* 1917 * inputs: 1918 * zc_name name of the pool 1919 * zc_nvlist_conf nvlist of devices to remove 1920 * zc_cookie to stop the remove? 1921 */ 1922static int 1923zfs_ioc_vdev_remove(zfs_cmd_t *zc) 1924{ 1925 spa_t *spa; 1926 int error; 1927 1928 error = spa_open(zc->zc_name, &spa, FTAG); 1929 if (error != 0) 1930 return (error); 1931 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE); 1932 spa_close(spa, FTAG); 1933 return (error); 1934} 1935 1936static int 1937zfs_ioc_vdev_set_state(zfs_cmd_t *zc) 1938{ 1939 spa_t *spa; 1940 int error; 1941 vdev_state_t newstate = VDEV_STATE_UNKNOWN; 1942 1943 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1944 return (error); 1945 switch (zc->zc_cookie) { 1946 case VDEV_STATE_ONLINE: 1947 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate); 1948 break; 1949 1950 case VDEV_STATE_OFFLINE: 1951 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj); 1952 break; 1953 1954 case VDEV_STATE_FAULTED: 1955 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED && 1956 zc->zc_obj != VDEV_AUX_EXTERNAL) 1957 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED; 1958 1959 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj); 1960 break; 1961 1962 case VDEV_STATE_DEGRADED: 1963 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED && 1964 zc->zc_obj != VDEV_AUX_EXTERNAL) 1965 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED; 1966 1967 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj); 1968 break; 1969 1970 default: 1971 error = SET_ERROR(EINVAL); 1972 } 1973 zc->zc_cookie = newstate; 1974 spa_close(spa, FTAG); 1975 return (error); 1976} 1977 1978static int 1979zfs_ioc_vdev_attach(zfs_cmd_t *zc) 1980{ 1981 spa_t *spa; 1982 int replacing = zc->zc_cookie; 1983 nvlist_t *config; 1984 int error; 1985 1986 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1987 return (error); 1988 1989 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1990 zc->zc_iflags, &config)) == 0) { 1991 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing); 1992 nvlist_free(config); 1993 } 1994 1995 spa_close(spa, FTAG); 1996 return (error); 1997} 1998 1999static int 2000zfs_ioc_vdev_detach(zfs_cmd_t *zc) 2001{ 2002 spa_t *spa; 2003 int error; 2004 2005 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 2006 return (error); 2007 2008 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE); 2009 2010 spa_close(spa, FTAG); 2011 return (error); 2012} 2013 2014static int 2015zfs_ioc_vdev_split(zfs_cmd_t *zc) 2016{ 2017 spa_t *spa; 2018 nvlist_t *config, *props = NULL; 2019 int error; 2020 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT); 2021 2022 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 2023 return (error); 2024 2025 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 2026 zc->zc_iflags, &config)) { 2027 spa_close(spa, FTAG); 2028 return (error); 2029 } 2030 2031 if (zc->zc_nvlist_src_size != 0 && (error = 2032 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2033 zc->zc_iflags, &props))) { 2034 spa_close(spa, FTAG); 2035 nvlist_free(config); 2036 return (error); 2037 } 2038 2039 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp); 2040 2041 spa_close(spa, FTAG); 2042 2043 nvlist_free(config); 2044 nvlist_free(props); 2045 2046 return (error); 2047} 2048 2049static int 2050zfs_ioc_vdev_setpath(zfs_cmd_t *zc) 2051{ 2052 spa_t *spa; 2053 char *path = zc->zc_value; 2054 uint64_t guid = zc->zc_guid; 2055 int error; 2056 2057 error = spa_open(zc->zc_name, &spa, FTAG); 2058 if (error != 0) 2059 return (error); 2060 2061 error = spa_vdev_setpath(spa, guid, path); 2062 spa_close(spa, FTAG); 2063 return (error); 2064} 2065 2066static int 2067zfs_ioc_vdev_setfru(zfs_cmd_t *zc) 2068{ 2069 spa_t *spa; 2070 char *fru = zc->zc_value; 2071 uint64_t guid = zc->zc_guid; 2072 int error; 2073 2074 error = spa_open(zc->zc_name, &spa, FTAG); 2075 if (error != 0) 2076 return (error); 2077 2078 error = spa_vdev_setfru(spa, guid, fru); 2079 spa_close(spa, FTAG); 2080 return (error); 2081} 2082 2083static int 2084zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os) 2085{ 2086 int error = 0; 2087 nvlist_t *nv; 2088 2089 dmu_objset_fast_stat(os, &zc->zc_objset_stats); 2090 2091 if (zc->zc_nvlist_dst != 0 && 2092 (error = dsl_prop_get_all(os, &nv)) == 0) { 2093 dmu_objset_stats(os, nv); 2094 /* 2095 * NB: zvol_get_stats() will read the objset contents, 2096 * which we aren't supposed to do with a 2097 * DS_MODE_USER hold, because it could be 2098 * inconsistent. So this is a bit of a workaround... 2099 * XXX reading with out owning 2100 */ 2101 if (!zc->zc_objset_stats.dds_inconsistent && 2102 dmu_objset_type(os) == DMU_OST_ZVOL) { 2103 error = zvol_get_stats(os, nv); 2104 if (error == EIO) 2105 return (error); 2106 VERIFY0(error); 2107 } 2108 error = put_nvlist(zc, nv); 2109 nvlist_free(nv); 2110 } 2111 2112 return (error); 2113} 2114 2115/* 2116 * inputs: 2117 * zc_name name of filesystem 2118 * zc_nvlist_dst_size size of buffer for property nvlist 2119 * 2120 * outputs: 2121 * zc_objset_stats stats 2122 * zc_nvlist_dst property nvlist 2123 * zc_nvlist_dst_size size of property nvlist 2124 */ 2125static int 2126zfs_ioc_objset_stats(zfs_cmd_t *zc) 2127{ 2128 objset_t *os; 2129 int error; 2130 2131 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 2132 if (error == 0) { 2133 error = zfs_ioc_objset_stats_impl(zc, os); 2134 dmu_objset_rele(os, FTAG); 2135 } 2136 2137 if (error == ENOMEM) 2138 error = 0; 2139 return (error); 2140} 2141 2142/* 2143 * inputs: 2144 * zc_name name of filesystem 2145 * zc_nvlist_dst_size size of buffer for property nvlist 2146 * 2147 * outputs: 2148 * zc_nvlist_dst received property nvlist 2149 * zc_nvlist_dst_size size of received property nvlist 2150 * 2151 * Gets received properties (distinct from local properties on or after 2152 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from 2153 * local property values. 2154 */ 2155static int 2156zfs_ioc_objset_recvd_props(zfs_cmd_t *zc) 2157{ 2158 int error = 0; 2159 nvlist_t *nv; 2160 2161 /* 2162 * Without this check, we would return local property values if the 2163 * caller has not already received properties on or after 2164 * SPA_VERSION_RECVD_PROPS. 2165 */ 2166 if (!dsl_prop_get_hasrecvd(zc->zc_name)) 2167 return (SET_ERROR(ENOTSUP)); 2168 2169 if (zc->zc_nvlist_dst != 0 && 2170 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) { 2171 error = put_nvlist(zc, nv); 2172 nvlist_free(nv); 2173 } 2174 2175 return (error); 2176} 2177 2178static int 2179nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop) 2180{ 2181 uint64_t value; 2182 int error; 2183 2184 /* 2185 * zfs_get_zplprop() will either find a value or give us 2186 * the default value (if there is one). 2187 */ 2188 if ((error = zfs_get_zplprop(os, prop, &value)) != 0) 2189 return (error); 2190 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0); 2191 return (0); 2192} 2193 2194/* 2195 * inputs: 2196 * zc_name name of filesystem 2197 * zc_nvlist_dst_size size of buffer for zpl property nvlist 2198 * 2199 * outputs: 2200 * zc_nvlist_dst zpl property nvlist 2201 * zc_nvlist_dst_size size of zpl property nvlist 2202 */ 2203static int 2204zfs_ioc_objset_zplprops(zfs_cmd_t *zc) 2205{ 2206 objset_t *os; 2207 int err; 2208 2209 /* XXX reading without owning */ 2210 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os)) 2211 return (err); 2212 2213 dmu_objset_fast_stat(os, &zc->zc_objset_stats); 2214 2215 /* 2216 * NB: nvl_add_zplprop() will read the objset contents, 2217 * which we aren't supposed to do with a DS_MODE_USER 2218 * hold, because it could be inconsistent. 2219 */ 2220 if (zc->zc_nvlist_dst != 0 && 2221 !zc->zc_objset_stats.dds_inconsistent && 2222 dmu_objset_type(os) == DMU_OST_ZFS) { 2223 nvlist_t *nv; 2224 2225 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2226 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 && 2227 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 && 2228 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 && 2229 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0) 2230 err = put_nvlist(zc, nv); 2231 nvlist_free(nv); 2232 } else { 2233 err = SET_ERROR(ENOENT); 2234 } 2235 dmu_objset_rele(os, FTAG); 2236 return (err); 2237} 2238 2239boolean_t 2240dataset_name_hidden(const char *name) 2241{ 2242 /* 2243 * Skip over datasets that are not visible in this zone, 2244 * internal datasets (which have a $ in their name), and 2245 * temporary datasets (which have a % in their name). 2246 */ 2247 if (strchr(name, '$') != NULL) 2248 return (B_TRUE); 2249 if (strchr(name, '%') != NULL) 2250 return (B_TRUE); 2251 if (!INGLOBALZONE(curthread) && !zone_dataset_visible(name, NULL)) 2252 return (B_TRUE); 2253 return (B_FALSE); 2254} 2255 2256/* 2257 * inputs: 2258 * zc_name name of filesystem 2259 * zc_cookie zap cursor 2260 * zc_nvlist_dst_size size of buffer for property nvlist 2261 * 2262 * outputs: 2263 * zc_name name of next filesystem 2264 * zc_cookie zap cursor 2265 * zc_objset_stats stats 2266 * zc_nvlist_dst property nvlist 2267 * zc_nvlist_dst_size size of property nvlist 2268 */ 2269static int 2270zfs_ioc_dataset_list_next(zfs_cmd_t *zc) 2271{ 2272 objset_t *os; 2273 int error; 2274 char *p; 2275 size_t orig_len = strlen(zc->zc_name); 2276 2277top: 2278 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) { 2279 if (error == ENOENT) 2280 error = SET_ERROR(ESRCH); 2281 return (error); 2282 } 2283 2284 p = strrchr(zc->zc_name, '/'); 2285 if (p == NULL || p[1] != '\0') 2286 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name)); 2287 p = zc->zc_name + strlen(zc->zc_name); 2288 2289 do { 2290 error = dmu_dir_list_next(os, 2291 sizeof (zc->zc_name) - (p - zc->zc_name), p, 2292 NULL, &zc->zc_cookie); 2293 if (error == ENOENT) 2294 error = SET_ERROR(ESRCH); 2295 } while (error == 0 && dataset_name_hidden(zc->zc_name)); 2296 dmu_objset_rele(os, FTAG); 2297 2298 /* 2299 * If it's an internal dataset (ie. with a '$' in its name), 2300 * don't try to get stats for it, otherwise we'll return ENOENT. 2301 */ 2302 if (error == 0 && strchr(zc->zc_name, '$') == NULL) { 2303 error = zfs_ioc_objset_stats(zc); /* fill in the stats */ 2304 if (error == ENOENT) { 2305 /* We lost a race with destroy, get the next one. */ 2306 zc->zc_name[orig_len] = '\0'; 2307 goto top; 2308 } 2309 } 2310 return (error); 2311} 2312 2313/* 2314 * inputs: 2315 * zc_name name of filesystem 2316 * zc_cookie zap cursor 2317 * zc_nvlist_dst_size size of buffer for property nvlist 2318 * zc_simple when set, only name is requested 2319 * 2320 * outputs: 2321 * zc_name name of next snapshot 2322 * zc_objset_stats stats 2323 * zc_nvlist_dst property nvlist 2324 * zc_nvlist_dst_size size of property nvlist 2325 */ 2326static int 2327zfs_ioc_snapshot_list_next(zfs_cmd_t *zc) 2328{ 2329 objset_t *os; 2330 int error; 2331 2332 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 2333 if (error != 0) { 2334 return (error == ENOENT ? ESRCH : error); 2335 } 2336 2337 /* 2338 * A dataset name of maximum length cannot have any snapshots, 2339 * so exit immediately. 2340 */ 2341 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) { 2342 dmu_objset_rele(os, FTAG); 2343 return (SET_ERROR(ESRCH)); 2344 } 2345 2346 error = dmu_snapshot_list_next(os, 2347 sizeof (zc->zc_name) - strlen(zc->zc_name), 2348 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie, 2349 NULL); 2350 2351 if (error == 0 && !zc->zc_simple) { 2352 dsl_dataset_t *ds; 2353 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool; 2354 2355 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds); 2356 if (error == 0) { 2357 objset_t *ossnap; 2358 2359 error = dmu_objset_from_ds(ds, &ossnap); 2360 if (error == 0) 2361 error = zfs_ioc_objset_stats_impl(zc, ossnap); 2362 dsl_dataset_rele(ds, FTAG); 2363 } 2364 } else if (error == ENOENT) { 2365 error = SET_ERROR(ESRCH); 2366 } 2367 2368 dmu_objset_rele(os, FTAG); 2369 /* if we failed, undo the @ that we tacked on to zc_name */ 2370 if (error != 0) 2371 *strchr(zc->zc_name, '@') = '\0'; 2372 return (error); 2373} 2374 2375static int 2376zfs_prop_set_userquota(const char *dsname, nvpair_t *pair) 2377{ 2378 const char *propname = nvpair_name(pair); 2379 uint64_t *valary; 2380 unsigned int vallen; 2381 const char *domain; 2382 char *dash; 2383 zfs_userquota_prop_t type; 2384 uint64_t rid; 2385 uint64_t quota; 2386 zfsvfs_t *zfsvfs; 2387 int err; 2388 2389 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2390 nvlist_t *attrs; 2391 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 2392 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2393 &pair) != 0) 2394 return (SET_ERROR(EINVAL)); 2395 } 2396 2397 /* 2398 * A correctly constructed propname is encoded as 2399 * userquota@<rid>-<domain>. 2400 */ 2401 if ((dash = strchr(propname, '-')) == NULL || 2402 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 || 2403 vallen != 3) 2404 return (SET_ERROR(EINVAL)); 2405 2406 domain = dash + 1; 2407 type = valary[0]; 2408 rid = valary[1]; 2409 quota = valary[2]; 2410 2411 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE); 2412 if (err == 0) { 2413 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota); 2414 zfsvfs_rele(zfsvfs, FTAG); 2415 } 2416 2417 return (err); 2418} 2419 2420/* 2421 * If the named property is one that has a special function to set its value, 2422 * return 0 on success and a positive error code on failure; otherwise if it is 2423 * not one of the special properties handled by this function, return -1. 2424 * 2425 * XXX: It would be better for callers of the property interface if we handled 2426 * these special cases in dsl_prop.c (in the dsl layer). 2427 */ 2428static int 2429zfs_prop_set_special(const char *dsname, zprop_source_t source, 2430 nvpair_t *pair) 2431{ 2432 const char *propname = nvpair_name(pair); 2433 zfs_prop_t prop = zfs_name_to_prop(propname); 2434 uint64_t intval; 2435 int err = -1; 2436 2437 if (prop == ZPROP_INVAL) { 2438 if (zfs_prop_userquota(propname)) 2439 return (zfs_prop_set_userquota(dsname, pair)); 2440 return (-1); 2441 } 2442 2443 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2444 nvlist_t *attrs; 2445 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 2446 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2447 &pair) == 0); 2448 } 2449 2450 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING) 2451 return (-1); 2452 2453 VERIFY(0 == nvpair_value_uint64(pair, &intval)); 2454 2455 switch (prop) { 2456 case ZFS_PROP_QUOTA: 2457 err = dsl_dir_set_quota(dsname, source, intval); 2458 break; 2459 case ZFS_PROP_REFQUOTA: 2460 err = dsl_dataset_set_refquota(dsname, source, intval); 2461 break; 2462 case ZFS_PROP_FILESYSTEM_LIMIT: 2463 case ZFS_PROP_SNAPSHOT_LIMIT: 2464 if (intval == UINT64_MAX) { 2465 /* clearing the limit, just do it */ 2466 err = 0; 2467 } else { 2468 err = dsl_dir_activate_fs_ss_limit(dsname); 2469 } 2470 /* 2471 * Set err to -1 to force the zfs_set_prop_nvlist code down the 2472 * default path to set the value in the nvlist. 2473 */ 2474 if (err == 0) 2475 err = -1; 2476 break; 2477 case ZFS_PROP_RESERVATION: 2478 err = dsl_dir_set_reservation(dsname, source, intval); 2479 break; 2480 case ZFS_PROP_REFRESERVATION: 2481 err = dsl_dataset_set_refreservation(dsname, source, intval); 2482 break; 2483 case ZFS_PROP_VOLSIZE: 2484 err = zvol_set_volsize(dsname, intval); 2485 break; 2486 case ZFS_PROP_VERSION: 2487 { 2488 zfsvfs_t *zfsvfs; 2489 2490 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0) 2491 break; 2492 2493 err = zfs_set_version(zfsvfs, intval); 2494 zfsvfs_rele(zfsvfs, FTAG); 2495 2496 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) { 2497 zfs_cmd_t *zc; 2498 2499 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP); 2500 (void) strcpy(zc->zc_name, dsname); 2501 (void) zfs_ioc_userspace_upgrade(zc); 2502 kmem_free(zc, sizeof (zfs_cmd_t)); 2503 } 2504 break; 2505 } 2506 default: 2507 err = -1; 2508 } 2509 2510 return (err); 2511} 2512 2513/* 2514 * This function is best effort. If it fails to set any of the given properties, 2515 * it continues to set as many as it can and returns the last error 2516 * encountered. If the caller provides a non-NULL errlist, it will be filled in 2517 * with the list of names of all the properties that failed along with the 2518 * corresponding error numbers. 2519 * 2520 * If every property is set successfully, zero is returned and errlist is not 2521 * modified. 2522 */ 2523int 2524zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl, 2525 nvlist_t *errlist) 2526{ 2527 nvpair_t *pair; 2528 nvpair_t *propval; 2529 int rv = 0; 2530 uint64_t intval; 2531 char *strval; 2532 nvlist_t *genericnvl = fnvlist_alloc(); 2533 nvlist_t *retrynvl = fnvlist_alloc(); 2534 2535retry: 2536 pair = NULL; 2537 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { 2538 const char *propname = nvpair_name(pair); 2539 zfs_prop_t prop = zfs_name_to_prop(propname); 2540 int err = 0; 2541 2542 /* decode the property value */ 2543 propval = pair; 2544 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2545 nvlist_t *attrs; 2546 attrs = fnvpair_value_nvlist(pair); 2547 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2548 &propval) != 0) 2549 err = SET_ERROR(EINVAL); 2550 } 2551 2552 /* Validate value type */ 2553 if (err == 0 && prop == ZPROP_INVAL) { 2554 if (zfs_prop_user(propname)) { 2555 if (nvpair_type(propval) != DATA_TYPE_STRING) 2556 err = SET_ERROR(EINVAL); 2557 } else if (zfs_prop_userquota(propname)) { 2558 if (nvpair_type(propval) != 2559 DATA_TYPE_UINT64_ARRAY) 2560 err = SET_ERROR(EINVAL); 2561 } else { 2562 err = SET_ERROR(EINVAL); 2563 } 2564 } else if (err == 0) { 2565 if (nvpair_type(propval) == DATA_TYPE_STRING) { 2566 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING) 2567 err = SET_ERROR(EINVAL); 2568 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) { 2569 const char *unused; 2570 2571 intval = fnvpair_value_uint64(propval); 2572 2573 switch (zfs_prop_get_type(prop)) { 2574 case PROP_TYPE_NUMBER: 2575 break; 2576 case PROP_TYPE_STRING: 2577 err = SET_ERROR(EINVAL); 2578 break; 2579 case PROP_TYPE_INDEX: 2580 if (zfs_prop_index_to_string(prop, 2581 intval, &unused) != 0) 2582 err = SET_ERROR(EINVAL); 2583 break; 2584 default: 2585 cmn_err(CE_PANIC, 2586 "unknown property type"); 2587 } 2588 } else { 2589 err = SET_ERROR(EINVAL); 2590 } 2591 } 2592 2593 /* Validate permissions */ 2594 if (err == 0) 2595 err = zfs_check_settable(dsname, pair, CRED()); 2596 2597 if (err == 0) { 2598 err = zfs_prop_set_special(dsname, source, pair); 2599 if (err == -1) { 2600 /* 2601 * For better performance we build up a list of 2602 * properties to set in a single transaction. 2603 */ 2604 err = nvlist_add_nvpair(genericnvl, pair); 2605 } else if (err != 0 && nvl != retrynvl) { 2606 /* 2607 * This may be a spurious error caused by 2608 * receiving quota and reservation out of order. 2609 * Try again in a second pass. 2610 */ 2611 err = nvlist_add_nvpair(retrynvl, pair); 2612 } 2613 } 2614 2615 if (err != 0) { 2616 if (errlist != NULL) 2617 fnvlist_add_int32(errlist, propname, err); 2618 rv = err; 2619 } 2620 } 2621 2622 if (nvl != retrynvl && !nvlist_empty(retrynvl)) { 2623 nvl = retrynvl; 2624 goto retry; 2625 } 2626 2627 if (!nvlist_empty(genericnvl) && 2628 dsl_props_set(dsname, source, genericnvl) != 0) { 2629 /* 2630 * If this fails, we still want to set as many properties as we 2631 * can, so try setting them individually. 2632 */ 2633 pair = NULL; 2634 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) { 2635 const char *propname = nvpair_name(pair); 2636 int err = 0; 2637 2638 propval = pair; 2639 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2640 nvlist_t *attrs; 2641 attrs = fnvpair_value_nvlist(pair); 2642 propval = fnvlist_lookup_nvpair(attrs, 2643 ZPROP_VALUE); 2644 } 2645 2646 if (nvpair_type(propval) == DATA_TYPE_STRING) { 2647 strval = fnvpair_value_string(propval); 2648 err = dsl_prop_set_string(dsname, propname, 2649 source, strval); 2650 } else { 2651 intval = fnvpair_value_uint64(propval); 2652 err = dsl_prop_set_int(dsname, propname, source, 2653 intval); 2654 } 2655 2656 if (err != 0) { 2657 if (errlist != NULL) { 2658 fnvlist_add_int32(errlist, propname, 2659 err); 2660 } 2661 rv = err; 2662 } 2663 } 2664 } 2665 nvlist_free(genericnvl); 2666 nvlist_free(retrynvl); 2667 2668 return (rv); 2669} 2670 2671/* 2672 * Check that all the properties are valid user properties. 2673 */ 2674static int 2675zfs_check_userprops(const char *fsname, nvlist_t *nvl) 2676{ 2677 nvpair_t *pair = NULL; 2678 int error = 0; 2679 2680 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { 2681 const char *propname = nvpair_name(pair); 2682 2683 if (!zfs_prop_user(propname) || 2684 nvpair_type(pair) != DATA_TYPE_STRING) 2685 return (SET_ERROR(EINVAL)); 2686 2687 if (error = zfs_secpolicy_write_perms(fsname, 2688 ZFS_DELEG_PERM_USERPROP, CRED())) 2689 return (error); 2690 2691 if (strlen(propname) >= ZAP_MAXNAMELEN) 2692 return (SET_ERROR(ENAMETOOLONG)); 2693 2694 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN) 2695 return (E2BIG); 2696 } 2697 return (0); 2698} 2699 2700static void 2701props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops) 2702{ 2703 nvpair_t *pair; 2704 2705 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2706 2707 pair = NULL; 2708 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) { 2709 if (nvlist_exists(skipped, nvpair_name(pair))) 2710 continue; 2711 2712 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0); 2713 } 2714} 2715 2716static int 2717clear_received_props(const char *dsname, nvlist_t *props, 2718 nvlist_t *skipped) 2719{ 2720 int err = 0; 2721 nvlist_t *cleared_props = NULL; 2722 props_skip(props, skipped, &cleared_props); 2723 if (!nvlist_empty(cleared_props)) { 2724 /* 2725 * Acts on local properties until the dataset has received 2726 * properties at least once on or after SPA_VERSION_RECVD_PROPS. 2727 */ 2728 zprop_source_t flags = (ZPROP_SRC_NONE | 2729 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0)); 2730 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL); 2731 } 2732 nvlist_free(cleared_props); 2733 return (err); 2734} 2735 2736/* 2737 * inputs: 2738 * zc_name name of filesystem 2739 * zc_value name of property to set 2740 * zc_nvlist_src{_size} nvlist of properties to apply 2741 * zc_cookie received properties flag 2742 * 2743 * outputs: 2744 * zc_nvlist_dst{_size} error for each unapplied received property 2745 */ 2746static int 2747zfs_ioc_set_prop(zfs_cmd_t *zc) 2748{ 2749 nvlist_t *nvl; 2750 boolean_t received = zc->zc_cookie; 2751 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED : 2752 ZPROP_SRC_LOCAL); 2753 nvlist_t *errors; 2754 int error; 2755 2756 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2757 zc->zc_iflags, &nvl)) != 0) 2758 return (error); 2759 2760 if (received) { 2761 nvlist_t *origprops; 2762 2763 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) { 2764 (void) clear_received_props(zc->zc_name, 2765 origprops, nvl); 2766 nvlist_free(origprops); 2767 } 2768 2769 error = dsl_prop_set_hasrecvd(zc->zc_name); 2770 } 2771 2772 errors = fnvlist_alloc(); 2773 if (error == 0) 2774 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors); 2775 2776 if (zc->zc_nvlist_dst != 0 && errors != NULL) { 2777 (void) put_nvlist(zc, errors); 2778 } 2779 2780 nvlist_free(errors); 2781 nvlist_free(nvl); 2782 return (error); 2783} 2784 2785/* 2786 * inputs: 2787 * zc_name name of filesystem 2788 * zc_value name of property to inherit 2789 * zc_cookie revert to received value if TRUE 2790 * 2791 * outputs: none 2792 */ 2793static int 2794zfs_ioc_inherit_prop(zfs_cmd_t *zc) 2795{ 2796 const char *propname = zc->zc_value; 2797 zfs_prop_t prop = zfs_name_to_prop(propname); 2798 boolean_t received = zc->zc_cookie; 2799 zprop_source_t source = (received 2800 ? ZPROP_SRC_NONE /* revert to received value, if any */ 2801 : ZPROP_SRC_INHERITED); /* explicitly inherit */ 2802 2803 if (received) { 2804 nvlist_t *dummy; 2805 nvpair_t *pair; 2806 zprop_type_t type; 2807 int err; 2808 2809 /* 2810 * zfs_prop_set_special() expects properties in the form of an 2811 * nvpair with type info. 2812 */ 2813 if (prop == ZPROP_INVAL) { 2814 if (!zfs_prop_user(propname)) 2815 return (SET_ERROR(EINVAL)); 2816 2817 type = PROP_TYPE_STRING; 2818 } else if (prop == ZFS_PROP_VOLSIZE || 2819 prop == ZFS_PROP_VERSION) { 2820 return (SET_ERROR(EINVAL)); 2821 } else { 2822 type = zfs_prop_get_type(prop); 2823 } 2824 2825 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2826 2827 switch (type) { 2828 case PROP_TYPE_STRING: 2829 VERIFY(0 == nvlist_add_string(dummy, propname, "")); 2830 break; 2831 case PROP_TYPE_NUMBER: 2832 case PROP_TYPE_INDEX: 2833 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0)); 2834 break; 2835 default: 2836 nvlist_free(dummy); 2837 return (SET_ERROR(EINVAL)); 2838 } 2839 2840 pair = nvlist_next_nvpair(dummy, NULL); 2841 err = zfs_prop_set_special(zc->zc_name, source, pair); 2842 nvlist_free(dummy); 2843 if (err != -1) 2844 return (err); /* special property already handled */ 2845 } else { 2846 /* 2847 * Only check this in the non-received case. We want to allow 2848 * 'inherit -S' to revert non-inheritable properties like quota 2849 * and reservation to the received or default values even though 2850 * they are not considered inheritable. 2851 */ 2852 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop)) 2853 return (SET_ERROR(EINVAL)); 2854 } 2855 2856 /* property name has been validated by zfs_secpolicy_inherit_prop() */ 2857 return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source)); 2858} 2859 2860static int 2861zfs_ioc_pool_set_props(zfs_cmd_t *zc) 2862{ 2863 nvlist_t *props; 2864 spa_t *spa; 2865 int error; 2866 nvpair_t *pair; 2867 2868 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2869 zc->zc_iflags, &props)) 2870 return (error); 2871 2872 /* 2873 * If the only property is the configfile, then just do a spa_lookup() 2874 * to handle the faulted case. 2875 */ 2876 pair = nvlist_next_nvpair(props, NULL); 2877 if (pair != NULL && strcmp(nvpair_name(pair), 2878 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 && 2879 nvlist_next_nvpair(props, pair) == NULL) { 2880 mutex_enter(&spa_namespace_lock); 2881 if ((spa = spa_lookup(zc->zc_name)) != NULL) { 2882 spa_configfile_set(spa, props, B_FALSE); 2883 spa_config_sync(spa, B_FALSE, B_TRUE); 2884 } 2885 mutex_exit(&spa_namespace_lock); 2886 if (spa != NULL) { 2887 nvlist_free(props); 2888 return (0); 2889 } 2890 } 2891 2892 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { 2893 nvlist_free(props); 2894 return (error); 2895 } 2896 2897 error = spa_prop_set(spa, props); 2898 2899 nvlist_free(props); 2900 spa_close(spa, FTAG); 2901 2902 return (error); 2903} 2904 2905static int 2906zfs_ioc_pool_get_props(zfs_cmd_t *zc) 2907{ 2908 spa_t *spa; 2909 int error; 2910 nvlist_t *nvp = NULL; 2911 2912 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { 2913 /* 2914 * If the pool is faulted, there may be properties we can still 2915 * get (such as altroot and cachefile), so attempt to get them 2916 * anyway. 2917 */ 2918 mutex_enter(&spa_namespace_lock); 2919 if ((spa = spa_lookup(zc->zc_name)) != NULL) 2920 error = spa_prop_get(spa, &nvp); 2921 mutex_exit(&spa_namespace_lock); 2922 } else { 2923 error = spa_prop_get(spa, &nvp); 2924 spa_close(spa, FTAG); 2925 } 2926 2927 if (error == 0 && zc->zc_nvlist_dst != 0) 2928 error = put_nvlist(zc, nvp); 2929 else 2930 error = SET_ERROR(EFAULT); 2931 2932 nvlist_free(nvp); 2933 return (error); 2934} 2935 2936/* 2937 * inputs: 2938 * zc_name name of filesystem 2939 * zc_nvlist_src{_size} nvlist of delegated permissions 2940 * zc_perm_action allow/unallow flag 2941 * 2942 * outputs: none 2943 */ 2944static int 2945zfs_ioc_set_fsacl(zfs_cmd_t *zc) 2946{ 2947 int error; 2948 nvlist_t *fsaclnv = NULL; 2949 2950 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2951 zc->zc_iflags, &fsaclnv)) != 0) 2952 return (error); 2953 2954 /* 2955 * Verify nvlist is constructed correctly 2956 */ 2957 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) { 2958 nvlist_free(fsaclnv); 2959 return (SET_ERROR(EINVAL)); 2960 } 2961 2962 /* 2963 * If we don't have PRIV_SYS_MOUNT, then validate 2964 * that user is allowed to hand out each permission in 2965 * the nvlist(s) 2966 */ 2967 2968 error = secpolicy_zfs(CRED()); 2969 if (error != 0) { 2970 if (zc->zc_perm_action == B_FALSE) { 2971 error = dsl_deleg_can_allow(zc->zc_name, 2972 fsaclnv, CRED()); 2973 } else { 2974 error = dsl_deleg_can_unallow(zc->zc_name, 2975 fsaclnv, CRED()); 2976 } 2977 } 2978 2979 if (error == 0) 2980 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action); 2981 2982 nvlist_free(fsaclnv); 2983 return (error); 2984} 2985 2986/* 2987 * inputs: 2988 * zc_name name of filesystem 2989 * 2990 * outputs: 2991 * zc_nvlist_src{_size} nvlist of delegated permissions 2992 */ 2993static int 2994zfs_ioc_get_fsacl(zfs_cmd_t *zc) 2995{ 2996 nvlist_t *nvp; 2997 int error; 2998 2999 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) { 3000 error = put_nvlist(zc, nvp); 3001 nvlist_free(nvp); 3002 } 3003 3004 return (error); 3005} 3006 3007/* 3008 * Search the vfs list for a specified resource. Returns a pointer to it 3009 * or NULL if no suitable entry is found. The caller of this routine 3010 * is responsible for releasing the returned vfs pointer. 3011 */ 3012static vfs_t * 3013zfs_get_vfs(const char *resource) 3014{ 3015 vfs_t *vfsp; 3016 3017 mtx_lock(&mountlist_mtx); 3018 TAILQ_FOREACH(vfsp, &mountlist, mnt_list) { 3019 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) { 3020 VFS_HOLD(vfsp); 3021 break; 3022 } 3023 } 3024 mtx_unlock(&mountlist_mtx); 3025 return (vfsp); 3026} 3027 3028/* ARGSUSED */ 3029static void 3030zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) 3031{ 3032 zfs_creat_t *zct = arg; 3033 3034 zfs_create_fs(os, cr, zct->zct_zplprops, tx); 3035} 3036 3037#define ZFS_PROP_UNDEFINED ((uint64_t)-1) 3038 3039/* 3040 * inputs: 3041 * os parent objset pointer (NULL if root fs) 3042 * fuids_ok fuids allowed in this version of the spa? 3043 * sa_ok SAs allowed in this version of the spa? 3044 * createprops list of properties requested by creator 3045 * 3046 * outputs: 3047 * zplprops values for the zplprops we attach to the master node object 3048 * is_ci true if requested file system will be purely case-insensitive 3049 * 3050 * Determine the settings for utf8only, normalization and 3051 * casesensitivity. Specific values may have been requested by the 3052 * creator and/or we can inherit values from the parent dataset. If 3053 * the file system is of too early a vintage, a creator can not 3054 * request settings for these properties, even if the requested 3055 * setting is the default value. We don't actually want to create dsl 3056 * properties for these, so remove them from the source nvlist after 3057 * processing. 3058 */ 3059static int 3060zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver, 3061 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops, 3062 nvlist_t *zplprops, boolean_t *is_ci) 3063{ 3064 uint64_t sense = ZFS_PROP_UNDEFINED; 3065 uint64_t norm = ZFS_PROP_UNDEFINED; 3066 uint64_t u8 = ZFS_PROP_UNDEFINED; 3067 3068 ASSERT(zplprops != NULL); 3069 3070 /* 3071 * Pull out creator prop choices, if any. 3072 */ 3073 if (createprops) { 3074 (void) nvlist_lookup_uint64(createprops, 3075 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver); 3076 (void) nvlist_lookup_uint64(createprops, 3077 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm); 3078 (void) nvlist_remove_all(createprops, 3079 zfs_prop_to_name(ZFS_PROP_NORMALIZE)); 3080 (void) nvlist_lookup_uint64(createprops, 3081 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8); 3082 (void) nvlist_remove_all(createprops, 3083 zfs_prop_to_name(ZFS_PROP_UTF8ONLY)); 3084 (void) nvlist_lookup_uint64(createprops, 3085 zfs_prop_to_name(ZFS_PROP_CASE), &sense); 3086 (void) nvlist_remove_all(createprops, 3087 zfs_prop_to_name(ZFS_PROP_CASE)); 3088 } 3089 3090 /* 3091 * If the zpl version requested is whacky or the file system 3092 * or pool is version is too "young" to support normalization 3093 * and the creator tried to set a value for one of the props, 3094 * error out. 3095 */ 3096 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) || 3097 (zplver >= ZPL_VERSION_FUID && !fuids_ok) || 3098 (zplver >= ZPL_VERSION_SA && !sa_ok) || 3099 (zplver < ZPL_VERSION_NORMALIZATION && 3100 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED || 3101 sense != ZFS_PROP_UNDEFINED))) 3102 return (SET_ERROR(ENOTSUP)); 3103 3104 /* 3105 * Put the version in the zplprops 3106 */ 3107 VERIFY(nvlist_add_uint64(zplprops, 3108 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0); 3109 3110 if (norm == ZFS_PROP_UNDEFINED) 3111 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0); 3112 VERIFY(nvlist_add_uint64(zplprops, 3113 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0); 3114 3115 /* 3116 * If we're normalizing, names must always be valid UTF-8 strings. 3117 */ 3118 if (norm) 3119 u8 = 1; 3120 if (u8 == ZFS_PROP_UNDEFINED) 3121 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0); 3122 VERIFY(nvlist_add_uint64(zplprops, 3123 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0); 3124 3125 if (sense == ZFS_PROP_UNDEFINED) 3126 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0); 3127 VERIFY(nvlist_add_uint64(zplprops, 3128 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0); 3129 3130 if (is_ci) 3131 *is_ci = (sense == ZFS_CASE_INSENSITIVE); 3132 3133 return (0); 3134} 3135 3136static int 3137zfs_fill_zplprops(const char *dataset, nvlist_t *createprops, 3138 nvlist_t *zplprops, boolean_t *is_ci) 3139{ 3140 boolean_t fuids_ok, sa_ok; 3141 uint64_t zplver = ZPL_VERSION; 3142 objset_t *os = NULL; 3143 char parentname[MAXNAMELEN]; 3144 char *cp; 3145 spa_t *spa; 3146 uint64_t spa_vers; 3147 int error; 3148 3149 (void) strlcpy(parentname, dataset, sizeof (parentname)); 3150 cp = strrchr(parentname, '/'); 3151 ASSERT(cp != NULL); 3152 cp[0] = '\0'; 3153 3154 if ((error = spa_open(dataset, &spa, FTAG)) != 0) 3155 return (error); 3156 3157 spa_vers = spa_version(spa); 3158 spa_close(spa, FTAG); 3159 3160 zplver = zfs_zpl_version_map(spa_vers); 3161 fuids_ok = (zplver >= ZPL_VERSION_FUID); 3162 sa_ok = (zplver >= ZPL_VERSION_SA); 3163 3164 /* 3165 * Open parent object set so we can inherit zplprop values. 3166 */ 3167 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0) 3168 return (error); 3169 3170 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops, 3171 zplprops, is_ci); 3172 dmu_objset_rele(os, FTAG); 3173 return (error); 3174} 3175 3176static int 3177zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops, 3178 nvlist_t *zplprops, boolean_t *is_ci) 3179{ 3180 boolean_t fuids_ok; 3181 boolean_t sa_ok; 3182 uint64_t zplver = ZPL_VERSION; 3183 int error; 3184 3185 zplver = zfs_zpl_version_map(spa_vers); 3186 fuids_ok = (zplver >= ZPL_VERSION_FUID); 3187 sa_ok = (zplver >= ZPL_VERSION_SA); 3188 3189 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok, 3190 createprops, zplprops, is_ci); 3191 return (error); 3192} 3193 3194/* 3195 * innvl: { 3196 * "type" -> dmu_objset_type_t (int32) 3197 * (optional) "props" -> { prop -> value } 3198 * } 3199 * 3200 * outnvl: propname -> error code (int32) 3201 */ 3202static int 3203zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3204{ 3205 int error = 0; 3206 zfs_creat_t zct = { 0 }; 3207 nvlist_t *nvprops = NULL; 3208 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx); 3209 int32_t type32; 3210 dmu_objset_type_t type; 3211 boolean_t is_insensitive = B_FALSE; 3212 3213 if (nvlist_lookup_int32(innvl, "type", &type32) != 0) 3214 return (SET_ERROR(EINVAL)); 3215 type = type32; 3216 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops); 3217 3218 switch (type) { 3219 case DMU_OST_ZFS: 3220 cbfunc = zfs_create_cb; 3221 break; 3222 3223 case DMU_OST_ZVOL: 3224 cbfunc = zvol_create_cb; 3225 break; 3226 3227 default: 3228 cbfunc = NULL; 3229 break; 3230 } 3231 if (strchr(fsname, '@') || 3232 strchr(fsname, '%')) 3233 return (SET_ERROR(EINVAL)); 3234 3235 zct.zct_props = nvprops; 3236 3237 if (cbfunc == NULL) 3238 return (SET_ERROR(EINVAL)); 3239 3240 if (type == DMU_OST_ZVOL) { 3241 uint64_t volsize, volblocksize; 3242 3243 if (nvprops == NULL) 3244 return (SET_ERROR(EINVAL)); 3245 if (nvlist_lookup_uint64(nvprops, 3246 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0) 3247 return (SET_ERROR(EINVAL)); 3248 3249 if ((error = nvlist_lookup_uint64(nvprops, 3250 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), 3251 &volblocksize)) != 0 && error != ENOENT) 3252 return (SET_ERROR(EINVAL)); 3253 3254 if (error != 0) 3255 volblocksize = zfs_prop_default_numeric( 3256 ZFS_PROP_VOLBLOCKSIZE); 3257 3258 if ((error = zvol_check_volblocksize( 3259 volblocksize)) != 0 || 3260 (error = zvol_check_volsize(volsize, 3261 volblocksize)) != 0) 3262 return (error); 3263 } else if (type == DMU_OST_ZFS) { 3264 int error; 3265 3266 /* 3267 * We have to have normalization and 3268 * case-folding flags correct when we do the 3269 * file system creation, so go figure them out 3270 * now. 3271 */ 3272 VERIFY(nvlist_alloc(&zct.zct_zplprops, 3273 NV_UNIQUE_NAME, KM_SLEEP) == 0); 3274 error = zfs_fill_zplprops(fsname, nvprops, 3275 zct.zct_zplprops, &is_insensitive); 3276 if (error != 0) { 3277 nvlist_free(zct.zct_zplprops); 3278 return (error); 3279 } 3280 } 3281 3282 error = dmu_objset_create(fsname, type, 3283 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct); 3284 nvlist_free(zct.zct_zplprops); 3285 3286 /* 3287 * It would be nice to do this atomically. 3288 */ 3289 if (error == 0) { 3290 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL, 3291 nvprops, outnvl); 3292 if (error != 0) 3293 (void) dsl_destroy_head(fsname); 3294 } 3295#ifdef __FreeBSD__ 3296 if (error == 0 && type == DMU_OST_ZVOL) 3297 zvol_create_minors(fsname); 3298#endif 3299 return (error); 3300} 3301 3302/* 3303 * innvl: { 3304 * "origin" -> name of origin snapshot 3305 * (optional) "props" -> { prop -> value } 3306 * } 3307 * 3308 * outnvl: propname -> error code (int32) 3309 */ 3310static int 3311zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3312{ 3313 int error = 0; 3314 nvlist_t *nvprops = NULL; 3315 char *origin_name; 3316 3317 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0) 3318 return (SET_ERROR(EINVAL)); 3319 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops); 3320 3321 if (strchr(fsname, '@') || 3322 strchr(fsname, '%')) 3323 return (SET_ERROR(EINVAL)); 3324 3325 if (dataset_namecheck(origin_name, NULL, NULL) != 0) 3326 return (SET_ERROR(EINVAL)); 3327 error = dmu_objset_clone(fsname, origin_name); 3328 if (error != 0) 3329 return (error); 3330 3331 /* 3332 * It would be nice to do this atomically. 3333 */ 3334 if (error == 0) { 3335 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL, 3336 nvprops, outnvl); 3337 if (error != 0) 3338 (void) dsl_destroy_head(fsname); 3339 } 3340#ifdef __FreeBSD__ 3341 if (error == 0) 3342 zvol_create_minors(fsname); 3343#endif 3344 return (error); 3345} 3346 3347/* 3348 * innvl: { 3349 * "snaps" -> { snapshot1, snapshot2 } 3350 * (optional) "props" -> { prop -> value (string) } 3351 * } 3352 * 3353 * outnvl: snapshot -> error code (int32) 3354 */ 3355static int 3356zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3357{ 3358 nvlist_t *snaps; 3359 nvlist_t *props = NULL; 3360 int error, poollen; 3361 nvpair_t *pair; 3362 3363 (void) nvlist_lookup_nvlist(innvl, "props", &props); 3364 if ((error = zfs_check_userprops(poolname, props)) != 0) 3365 return (error); 3366 3367 if (!nvlist_empty(props) && 3368 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS)) 3369 return (SET_ERROR(ENOTSUP)); 3370 3371 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 3372 return (SET_ERROR(EINVAL)); 3373 poollen = strlen(poolname); 3374 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 3375 pair = nvlist_next_nvpair(snaps, pair)) { 3376 const char *name = nvpair_name(pair); 3377 const char *cp = strchr(name, '@'); 3378 3379 /* 3380 * The snap name must contain an @, and the part after it must 3381 * contain only valid characters. 3382 */ 3383 if (cp == NULL || 3384 zfs_component_namecheck(cp + 1, NULL, NULL) != 0) 3385 return (SET_ERROR(EINVAL)); 3386 3387 /* 3388 * The snap must be in the specified pool. 3389 */ 3390 if (strncmp(name, poolname, poollen) != 0 || 3391 (name[poollen] != '/' && name[poollen] != '@')) 3392 return (SET_ERROR(EXDEV)); 3393 3394 /* This must be the only snap of this fs. */ 3395 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair); 3396 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) { 3397 if (strncmp(name, nvpair_name(pair2), cp - name + 1) 3398 == 0) { 3399 return (SET_ERROR(EXDEV)); 3400 } 3401 } 3402 } 3403 3404 error = dsl_dataset_snapshot(snaps, props, outnvl); 3405 return (error); 3406} 3407 3408/* 3409 * innvl: "message" -> string 3410 */ 3411/* ARGSUSED */ 3412static int 3413zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl) 3414{ 3415 char *message; 3416 spa_t *spa; 3417 int error; 3418 char *poolname; 3419 3420 /* 3421 * The poolname in the ioctl is not set, we get it from the TSD, 3422 * which was set at the end of the last successful ioctl that allows 3423 * logging. The secpolicy func already checked that it is set. 3424 * Only one log ioctl is allowed after each successful ioctl, so 3425 * we clear the TSD here. 3426 */ 3427 poolname = tsd_get(zfs_allow_log_key); 3428 (void) tsd_set(zfs_allow_log_key, NULL); 3429 error = spa_open(poolname, &spa, FTAG); 3430 strfree(poolname); 3431 if (error != 0) 3432 return (error); 3433 3434 if (nvlist_lookup_string(innvl, "message", &message) != 0) { 3435 spa_close(spa, FTAG); 3436 return (SET_ERROR(EINVAL)); 3437 } 3438 3439 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) { 3440 spa_close(spa, FTAG); 3441 return (SET_ERROR(ENOTSUP)); 3442 } 3443 3444 error = spa_history_log(spa, message); 3445 spa_close(spa, FTAG); 3446 return (error); 3447} 3448 3449/* 3450 * The dp_config_rwlock must not be held when calling this, because the 3451 * unmount may need to write out data. 3452 * 3453 * This function is best-effort. Callers must deal gracefully if it 3454 * remains mounted (or is remounted after this call). 3455 * 3456 * Returns 0 if the argument is not a snapshot, or it is not currently a 3457 * filesystem, or we were able to unmount it. Returns error code otherwise. 3458 */ 3459int 3460zfs_unmount_snap(const char *snapname) 3461{ 3462 vfs_t *vfsp; 3463 zfsvfs_t *zfsvfs; 3464 int err; 3465 3466 if (strchr(snapname, '@') == NULL) 3467 return (0); 3468 3469 vfsp = zfs_get_vfs(snapname); 3470 if (vfsp == NULL) 3471 return (0); 3472 3473 zfsvfs = vfsp->vfs_data; 3474 ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os))); 3475 3476 err = vn_vfswlock(vfsp->vfs_vnodecovered); 3477 VFS_RELE(vfsp); 3478 if (err != 0) 3479 return (SET_ERROR(err)); 3480 3481 /* 3482 * Always force the unmount for snapshots. 3483 */ 3484 3485#ifdef illumos 3486 (void) dounmount(vfsp, MS_FORCE, kcred); 3487#else 3488 vfs_ref(vfsp); 3489 (void) dounmount(vfsp, MS_FORCE, curthread); 3490#endif 3491 return (0); 3492} 3493 3494/* ARGSUSED */ 3495static int 3496zfs_unmount_snap_cb(const char *snapname, void *arg) 3497{ 3498 return (zfs_unmount_snap(snapname)); 3499} 3500 3501/* 3502 * When a clone is destroyed, its origin may also need to be destroyed, 3503 * in which case it must be unmounted. This routine will do that unmount 3504 * if necessary. 3505 */ 3506void 3507zfs_destroy_unmount_origin(const char *fsname) 3508{ 3509 int error; 3510 objset_t *os; 3511 dsl_dataset_t *ds; 3512 3513 error = dmu_objset_hold(fsname, FTAG, &os); 3514 if (error != 0) 3515 return; 3516 ds = dmu_objset_ds(os); 3517 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) { 3518 char originname[MAXNAMELEN]; 3519 dsl_dataset_name(ds->ds_prev, originname); 3520 dmu_objset_rele(os, FTAG); 3521 (void) zfs_unmount_snap(originname); 3522 } else { 3523 dmu_objset_rele(os, FTAG); 3524 } 3525} 3526 3527/* 3528 * innvl: { 3529 * "snaps" -> { snapshot1, snapshot2 } 3530 * (optional boolean) "defer" 3531 * } 3532 * 3533 * outnvl: snapshot -> error code (int32) 3534 * 3535 */ 3536/* ARGSUSED */ 3537static int 3538zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3539{ 3540 int error, poollen; 3541 nvlist_t *snaps; 3542 nvpair_t *pair; 3543 boolean_t defer; 3544 3545 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 3546 return (SET_ERROR(EINVAL)); 3547 defer = nvlist_exists(innvl, "defer"); 3548 3549 poollen = strlen(poolname); 3550 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 3551 pair = nvlist_next_nvpair(snaps, pair)) { 3552 const char *name = nvpair_name(pair); 3553 3554 /* 3555 * The snap must be in the specified pool to prevent the 3556 * invalid removal of zvol minors below. 3557 */ 3558 if (strncmp(name, poolname, poollen) != 0 || 3559 (name[poollen] != '/' && name[poollen] != '@')) 3560 return (SET_ERROR(EXDEV)); 3561 3562 error = zfs_unmount_snap(name); 3563 if (error != 0) 3564 return (error); 3565#if defined(__FreeBSD__) 3566 zvol_remove_minors(name); 3567#endif 3568 } 3569 3570 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl)); 3571} 3572 3573/* 3574 * Create bookmarks. Bookmark names are of the form <fs>#<bmark>. 3575 * All bookmarks must be in the same pool. 3576 * 3577 * innvl: { 3578 * bookmark1 -> snapshot1, bookmark2 -> snapshot2 3579 * } 3580 * 3581 * outnvl: bookmark -> error code (int32) 3582 * 3583 */ 3584/* ARGSUSED */ 3585static int 3586zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3587{ 3588 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 3589 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 3590 char *snap_name; 3591 3592 /* 3593 * Verify the snapshot argument. 3594 */ 3595 if (nvpair_value_string(pair, &snap_name) != 0) 3596 return (SET_ERROR(EINVAL)); 3597 3598 3599 /* Verify that the keys (bookmarks) are unique */ 3600 for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair); 3601 pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) { 3602 if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0) 3603 return (SET_ERROR(EINVAL)); 3604 } 3605 } 3606 3607 return (dsl_bookmark_create(innvl, outnvl)); 3608} 3609 3610/* 3611 * innvl: { 3612 * property 1, property 2, ... 3613 * } 3614 * 3615 * outnvl: { 3616 * bookmark name 1 -> { property 1, property 2, ... }, 3617 * bookmark name 2 -> { property 1, property 2, ... } 3618 * } 3619 * 3620 */ 3621static int 3622zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3623{ 3624 return (dsl_get_bookmarks(fsname, innvl, outnvl)); 3625} 3626 3627/* 3628 * innvl: { 3629 * bookmark name 1, bookmark name 2 3630 * } 3631 * 3632 * outnvl: bookmark -> error code (int32) 3633 * 3634 */ 3635static int 3636zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl, 3637 nvlist_t *outnvl) 3638{ 3639 int error, poollen; 3640 3641 poollen = strlen(poolname); 3642 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 3643 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 3644 const char *name = nvpair_name(pair); 3645 const char *cp = strchr(name, '#'); 3646 3647 /* 3648 * The bookmark name must contain an #, and the part after it 3649 * must contain only valid characters. 3650 */ 3651 if (cp == NULL || 3652 zfs_component_namecheck(cp + 1, NULL, NULL) != 0) 3653 return (SET_ERROR(EINVAL)); 3654 3655 /* 3656 * The bookmark must be in the specified pool. 3657 */ 3658 if (strncmp(name, poolname, poollen) != 0 || 3659 (name[poollen] != '/' && name[poollen] != '#')) 3660 return (SET_ERROR(EXDEV)); 3661 } 3662 3663 error = dsl_bookmark_destroy(innvl, outnvl); 3664 return (error); 3665} 3666 3667/* 3668 * inputs: 3669 * zc_name name of dataset to destroy 3670 * zc_objset_type type of objset 3671 * zc_defer_destroy mark for deferred destroy 3672 * 3673 * outputs: none 3674 */ 3675static int 3676zfs_ioc_destroy(zfs_cmd_t *zc) 3677{ 3678 int err; 3679 3680 if (zc->zc_objset_type == DMU_OST_ZFS) { 3681 err = zfs_unmount_snap(zc->zc_name); 3682 if (err != 0) 3683 return (err); 3684 } 3685 3686 if (strchr(zc->zc_name, '@')) 3687 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy); 3688 else 3689 err = dsl_destroy_head(zc->zc_name); 3690 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0) 3691#ifdef __FreeBSD__ 3692 zvol_remove_minors(zc->zc_name); 3693#else 3694 (void) zvol_remove_minor(zc->zc_name); 3695#endif 3696 return (err); 3697} 3698 3699/* 3700 * fsname is name of dataset to rollback (to most recent snapshot) 3701 * 3702 * innvl is not used. 3703 * 3704 * outnvl: "target" -> name of most recent snapshot 3705 * } 3706 */ 3707/* ARGSUSED */ 3708static int 3709zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl) 3710{ 3711 zfsvfs_t *zfsvfs; 3712 int error; 3713 3714 if (getzfsvfs(fsname, &zfsvfs) == 0) { 3715 error = zfs_suspend_fs(zfsvfs); 3716 if (error == 0) { 3717 int resume_err; 3718 3719 error = dsl_dataset_rollback(fsname, zfsvfs, outnvl); 3720 resume_err = zfs_resume_fs(zfsvfs, fsname); 3721 error = error ? error : resume_err; 3722 } 3723 VFS_RELE(zfsvfs->z_vfs); 3724 } else { 3725 error = dsl_dataset_rollback(fsname, NULL, outnvl); 3726 } 3727 return (error); 3728} 3729 3730static int 3731recursive_unmount(const char *fsname, void *arg) 3732{ 3733 const char *snapname = arg; 3734 char fullname[MAXNAMELEN]; 3735 3736 (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname); 3737 return (zfs_unmount_snap(fullname)); 3738} 3739 3740/* 3741 * inputs: 3742 * zc_name old name of dataset 3743 * zc_value new name of dataset 3744 * zc_cookie recursive flag (only valid for snapshots) 3745 * 3746 * outputs: none 3747 */ 3748static int 3749zfs_ioc_rename(zfs_cmd_t *zc) 3750{ 3751 boolean_t recursive = zc->zc_cookie & 1; 3752 char *at; 3753 boolean_t allow_mounted = B_TRUE; 3754 3755#ifdef __FreeBSD__ 3756 allow_mounted = (zc->zc_cookie & 2) != 0; 3757#endif 3758 3759 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0'; 3760 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 3761 strchr(zc->zc_value, '%')) 3762 return (SET_ERROR(EINVAL)); 3763 3764 at = strchr(zc->zc_name, '@'); 3765 if (at != NULL) { 3766 /* snaps must be in same fs */ 3767 int error; 3768 3769 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1)) 3770 return (SET_ERROR(EXDEV)); 3771 *at = '\0'; 3772 if (zc->zc_objset_type == DMU_OST_ZFS && allow_mounted) { 3773 error = dmu_objset_find(zc->zc_name, 3774 recursive_unmount, at + 1, 3775 recursive ? DS_FIND_CHILDREN : 0); 3776 if (error != 0) { 3777 *at = '@'; 3778 return (error); 3779 } 3780 } 3781 error = dsl_dataset_rename_snapshot(zc->zc_name, 3782 at + 1, strchr(zc->zc_value, '@') + 1, recursive); 3783 *at = '@'; 3784 3785 return (error); 3786 } else { 3787#ifdef illumos 3788 if (zc->zc_objset_type == DMU_OST_ZVOL) 3789 (void) zvol_remove_minor(zc->zc_name); 3790#endif 3791 return (dsl_dir_rename(zc->zc_name, zc->zc_value)); 3792 } 3793} 3794 3795static int 3796zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr) 3797{ 3798 const char *propname = nvpair_name(pair); 3799 boolean_t issnap = (strchr(dsname, '@') != NULL); 3800 zfs_prop_t prop = zfs_name_to_prop(propname); 3801 uint64_t intval; 3802 int err; 3803 3804 if (prop == ZPROP_INVAL) { 3805 if (zfs_prop_user(propname)) { 3806 if (err = zfs_secpolicy_write_perms(dsname, 3807 ZFS_DELEG_PERM_USERPROP, cr)) 3808 return (err); 3809 return (0); 3810 } 3811 3812 if (!issnap && zfs_prop_userquota(propname)) { 3813 const char *perm = NULL; 3814 const char *uq_prefix = 3815 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA]; 3816 const char *gq_prefix = 3817 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA]; 3818 3819 if (strncmp(propname, uq_prefix, 3820 strlen(uq_prefix)) == 0) { 3821 perm = ZFS_DELEG_PERM_USERQUOTA; 3822 } else if (strncmp(propname, gq_prefix, 3823 strlen(gq_prefix)) == 0) { 3824 perm = ZFS_DELEG_PERM_GROUPQUOTA; 3825 } else { 3826 /* USERUSED and GROUPUSED are read-only */ 3827 return (SET_ERROR(EINVAL)); 3828 } 3829 3830 if (err = zfs_secpolicy_write_perms(dsname, perm, cr)) 3831 return (err); 3832 return (0); 3833 } 3834 3835 return (SET_ERROR(EINVAL)); 3836 } 3837 3838 if (issnap) 3839 return (SET_ERROR(EINVAL)); 3840 3841 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 3842 /* 3843 * dsl_prop_get_all_impl() returns properties in this 3844 * format. 3845 */ 3846 nvlist_t *attrs; 3847 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 3848 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 3849 &pair) == 0); 3850 } 3851 3852 /* 3853 * Check that this value is valid for this pool version 3854 */ 3855 switch (prop) { 3856 case ZFS_PROP_COMPRESSION: 3857 /* 3858 * If the user specified gzip compression, make sure 3859 * the SPA supports it. We ignore any errors here since 3860 * we'll catch them later. 3861 */ 3862 if (nvpair_value_uint64(pair, &intval) == 0) { 3863 if (intval >= ZIO_COMPRESS_GZIP_1 && 3864 intval <= ZIO_COMPRESS_GZIP_9 && 3865 zfs_earlier_version(dsname, 3866 SPA_VERSION_GZIP_COMPRESSION)) { 3867 return (SET_ERROR(ENOTSUP)); 3868 } 3869 3870 if (intval == ZIO_COMPRESS_ZLE && 3871 zfs_earlier_version(dsname, 3872 SPA_VERSION_ZLE_COMPRESSION)) 3873 return (SET_ERROR(ENOTSUP)); 3874 3875 if (intval == ZIO_COMPRESS_LZ4) { 3876 spa_t *spa; 3877 3878 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 3879 return (err); 3880 3881 if (!spa_feature_is_enabled(spa, 3882 SPA_FEATURE_LZ4_COMPRESS)) { 3883 spa_close(spa, FTAG); 3884 return (SET_ERROR(ENOTSUP)); 3885 } 3886 spa_close(spa, FTAG); 3887 } 3888 3889 /* 3890 * If this is a bootable dataset then 3891 * verify that the compression algorithm 3892 * is supported for booting. We must return 3893 * something other than ENOTSUP since it 3894 * implies a downrev pool version. 3895 */ 3896 if (zfs_is_bootfs(dsname) && 3897 !BOOTFS_COMPRESS_VALID(intval)) { 3898 return (SET_ERROR(ERANGE)); 3899 } 3900 } 3901 break; 3902 3903 case ZFS_PROP_COPIES: 3904 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS)) 3905 return (SET_ERROR(ENOTSUP)); 3906 break; 3907 3908 case ZFS_PROP_RECORDSIZE: 3909 /* Record sizes above 128k need the feature to be enabled */ 3910 if (nvpair_value_uint64(pair, &intval) == 0 && 3911 intval > SPA_OLD_MAXBLOCKSIZE) { 3912 spa_t *spa; 3913 3914 /* 3915 * If this is a bootable dataset then 3916 * the we don't allow large (>128K) blocks, 3917 * because GRUB doesn't support them. 3918 */ 3919 if (zfs_is_bootfs(dsname) && 3920 intval > SPA_OLD_MAXBLOCKSIZE) { 3921 return (SET_ERROR(ERANGE)); 3922 } 3923 3924 /* 3925 * We don't allow setting the property above 1MB, 3926 * unless the tunable has been changed. 3927 */ 3928 if (intval > zfs_max_recordsize || 3929 intval > SPA_MAXBLOCKSIZE) 3930 return (SET_ERROR(ERANGE)); 3931 3932 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 3933 return (err); 3934 3935 if (!spa_feature_is_enabled(spa, 3936 SPA_FEATURE_LARGE_BLOCKS)) { 3937 spa_close(spa, FTAG); 3938 return (SET_ERROR(ENOTSUP)); 3939 } 3940 spa_close(spa, FTAG); 3941 } 3942 break; 3943 3944 case ZFS_PROP_SHARESMB: 3945 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID)) 3946 return (SET_ERROR(ENOTSUP)); 3947 break; 3948 3949 case ZFS_PROP_ACLINHERIT: 3950 if (nvpair_type(pair) == DATA_TYPE_UINT64 && 3951 nvpair_value_uint64(pair, &intval) == 0) { 3952 if (intval == ZFS_ACL_PASSTHROUGH_X && 3953 zfs_earlier_version(dsname, 3954 SPA_VERSION_PASSTHROUGH_X)) 3955 return (SET_ERROR(ENOTSUP)); 3956 } 3957 break; 3958 3959 case ZFS_PROP_CHECKSUM: 3960 case ZFS_PROP_DEDUP: 3961 { 3962 spa_feature_t feature; 3963 spa_t *spa; 3964 3965 /* dedup feature version checks */ 3966 if (prop == ZFS_PROP_DEDUP && 3967 zfs_earlier_version(dsname, SPA_VERSION_DEDUP)) 3968 return (SET_ERROR(ENOTSUP)); 3969 3970 if (nvpair_value_uint64(pair, &intval) != 0) 3971 return (SET_ERROR(EINVAL)); 3972 3973 /* check prop value is enabled in features */ 3974 feature = zio_checksum_to_feature(intval); 3975 if (feature == SPA_FEATURE_NONE) 3976 break; 3977 3978 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 3979 return (err); 3980 /* 3981 * Salted checksums are not supported on root pools. 3982 */ 3983 if (spa_bootfs(spa) != 0 && 3984 intval < ZIO_CHECKSUM_FUNCTIONS && 3985 (zio_checksum_table[intval].ci_flags & 3986 ZCHECKSUM_FLAG_SALTED)) { 3987 spa_close(spa, FTAG); 3988 return (SET_ERROR(ERANGE)); 3989 } 3990 if (!spa_feature_is_enabled(spa, feature)) { 3991 spa_close(spa, FTAG); 3992 return (SET_ERROR(ENOTSUP)); 3993 } 3994 spa_close(spa, FTAG); 3995 break; 3996 } 3997 } 3998 3999 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED())); 4000} 4001 4002/* 4003 * Checks for a race condition to make sure we don't increment a feature flag 4004 * multiple times. 4005 */ 4006static int 4007zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx) 4008{ 4009 spa_t *spa = dmu_tx_pool(tx)->dp_spa; 4010 spa_feature_t *featurep = arg; 4011 4012 if (!spa_feature_is_active(spa, *featurep)) 4013 return (0); 4014 else 4015 return (SET_ERROR(EBUSY)); 4016} 4017 4018/* 4019 * The callback invoked on feature activation in the sync task caused by 4020 * zfs_prop_activate_feature. 4021 */ 4022static void 4023zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx) 4024{ 4025 spa_t *spa = dmu_tx_pool(tx)->dp_spa; 4026 spa_feature_t *featurep = arg; 4027 4028 spa_feature_incr(spa, *featurep, tx); 4029} 4030 4031/* 4032 * Activates a feature on a pool in response to a property setting. This 4033 * creates a new sync task which modifies the pool to reflect the feature 4034 * as being active. 4035 */ 4036static int 4037zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature) 4038{ 4039 int err; 4040 4041 /* EBUSY here indicates that the feature is already active */ 4042 err = dsl_sync_task(spa_name(spa), 4043 zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync, 4044 &feature, 2, ZFS_SPACE_CHECK_RESERVED); 4045 4046 if (err != 0 && err != EBUSY) 4047 return (err); 4048 else 4049 return (0); 4050} 4051 4052/* 4053 * Removes properties from the given props list that fail permission checks 4054 * needed to clear them and to restore them in case of a receive error. For each 4055 * property, make sure we have both set and inherit permissions. 4056 * 4057 * Returns the first error encountered if any permission checks fail. If the 4058 * caller provides a non-NULL errlist, it also gives the complete list of names 4059 * of all the properties that failed a permission check along with the 4060 * corresponding error numbers. The caller is responsible for freeing the 4061 * returned errlist. 4062 * 4063 * If every property checks out successfully, zero is returned and the list 4064 * pointed at by errlist is NULL. 4065 */ 4066static int 4067zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist) 4068{ 4069 zfs_cmd_t *zc; 4070 nvpair_t *pair, *next_pair; 4071 nvlist_t *errors; 4072 int err, rv = 0; 4073 4074 if (props == NULL) 4075 return (0); 4076 4077 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0); 4078 4079 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP); 4080 (void) strcpy(zc->zc_name, dataset); 4081 pair = nvlist_next_nvpair(props, NULL); 4082 while (pair != NULL) { 4083 next_pair = nvlist_next_nvpair(props, pair); 4084 4085 (void) strcpy(zc->zc_value, nvpair_name(pair)); 4086 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 || 4087 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) { 4088 VERIFY(nvlist_remove_nvpair(props, pair) == 0); 4089 VERIFY(nvlist_add_int32(errors, 4090 zc->zc_value, err) == 0); 4091 } 4092 pair = next_pair; 4093 } 4094 kmem_free(zc, sizeof (zfs_cmd_t)); 4095 4096 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) { 4097 nvlist_free(errors); 4098 errors = NULL; 4099 } else { 4100 VERIFY(nvpair_value_int32(pair, &rv) == 0); 4101 } 4102 4103 if (errlist == NULL) 4104 nvlist_free(errors); 4105 else 4106 *errlist = errors; 4107 4108 return (rv); 4109} 4110 4111static boolean_t 4112propval_equals(nvpair_t *p1, nvpair_t *p2) 4113{ 4114 if (nvpair_type(p1) == DATA_TYPE_NVLIST) { 4115 /* dsl_prop_get_all_impl() format */ 4116 nvlist_t *attrs; 4117 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0); 4118 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 4119 &p1) == 0); 4120 } 4121 4122 if (nvpair_type(p2) == DATA_TYPE_NVLIST) { 4123 nvlist_t *attrs; 4124 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0); 4125 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 4126 &p2) == 0); 4127 } 4128 4129 if (nvpair_type(p1) != nvpair_type(p2)) 4130 return (B_FALSE); 4131 4132 if (nvpair_type(p1) == DATA_TYPE_STRING) { 4133 char *valstr1, *valstr2; 4134 4135 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0); 4136 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0); 4137 return (strcmp(valstr1, valstr2) == 0); 4138 } else { 4139 uint64_t intval1, intval2; 4140 4141 VERIFY(nvpair_value_uint64(p1, &intval1) == 0); 4142 VERIFY(nvpair_value_uint64(p2, &intval2) == 0); 4143 return (intval1 == intval2); 4144 } 4145} 4146 4147/* 4148 * Remove properties from props if they are not going to change (as determined 4149 * by comparison with origprops). Remove them from origprops as well, since we 4150 * do not need to clear or restore properties that won't change. 4151 */ 4152static void 4153props_reduce(nvlist_t *props, nvlist_t *origprops) 4154{ 4155 nvpair_t *pair, *next_pair; 4156 4157 if (origprops == NULL) 4158 return; /* all props need to be received */ 4159 4160 pair = nvlist_next_nvpair(props, NULL); 4161 while (pair != NULL) { 4162 const char *propname = nvpair_name(pair); 4163 nvpair_t *match; 4164 4165 next_pair = nvlist_next_nvpair(props, pair); 4166 4167 if ((nvlist_lookup_nvpair(origprops, propname, 4168 &match) != 0) || !propval_equals(pair, match)) 4169 goto next; /* need to set received value */ 4170 4171 /* don't clear the existing received value */ 4172 (void) nvlist_remove_nvpair(origprops, match); 4173 /* don't bother receiving the property */ 4174 (void) nvlist_remove_nvpair(props, pair); 4175next: 4176 pair = next_pair; 4177 } 4178} 4179 4180/* 4181 * Extract properties that cannot be set PRIOR to the receipt of a dataset. 4182 * For example, refquota cannot be set until after the receipt of a dataset, 4183 * because in replication streams, an older/earlier snapshot may exceed the 4184 * refquota. We want to receive the older/earlier snapshot, but setting 4185 * refquota pre-receipt will set the dsl's ACTUAL quota, which will prevent 4186 * the older/earlier snapshot from being received (with EDQUOT). 4187 * 4188 * The ZFS test "zfs_receive_011_pos" demonstrates such a scenario. 4189 * 4190 * libzfs will need to be judicious handling errors encountered by props 4191 * extracted by this function. 4192 */ 4193static nvlist_t * 4194extract_delay_props(nvlist_t *props) 4195{ 4196 nvlist_t *delayprops; 4197 nvpair_t *nvp, *tmp; 4198 static const zfs_prop_t delayable[] = { ZFS_PROP_REFQUOTA, 0 }; 4199 int i; 4200 4201 VERIFY(nvlist_alloc(&delayprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 4202 4203 for (nvp = nvlist_next_nvpair(props, NULL); nvp != NULL; 4204 nvp = nvlist_next_nvpair(props, nvp)) { 4205 /* 4206 * strcmp() is safe because zfs_prop_to_name() always returns 4207 * a bounded string. 4208 */ 4209 for (i = 0; delayable[i] != 0; i++) { 4210 if (strcmp(zfs_prop_to_name(delayable[i]), 4211 nvpair_name(nvp)) == 0) { 4212 break; 4213 } 4214 } 4215 if (delayable[i] != 0) { 4216 tmp = nvlist_prev_nvpair(props, nvp); 4217 VERIFY(nvlist_add_nvpair(delayprops, nvp) == 0); 4218 VERIFY(nvlist_remove_nvpair(props, nvp) == 0); 4219 nvp = tmp; 4220 } 4221 } 4222 4223 if (nvlist_empty(delayprops)) { 4224 nvlist_free(delayprops); 4225 delayprops = NULL; 4226 } 4227 return (delayprops); 4228} 4229 4230#ifdef DEBUG 4231static boolean_t zfs_ioc_recv_inject_err; 4232#endif 4233 4234/* 4235 * inputs: 4236 * zc_name name of containing filesystem 4237 * zc_nvlist_src{_size} nvlist of properties to apply 4238 * zc_value name of snapshot to create 4239 * zc_string name of clone origin (if DRR_FLAG_CLONE) 4240 * zc_cookie file descriptor to recv from 4241 * zc_begin_record the BEGIN record of the stream (not byteswapped) 4242 * zc_guid force flag 4243 * zc_cleanup_fd cleanup-on-exit file descriptor 4244 * zc_action_handle handle for this guid/ds mapping (or zero on first call) 4245 * zc_resumable if data is incomplete assume sender will resume 4246 * 4247 * outputs: 4248 * zc_cookie number of bytes read 4249 * zc_nvlist_dst{_size} error for each unapplied received property 4250 * zc_obj zprop_errflags_t 4251 * zc_action_handle handle for this guid/ds mapping 4252 */ 4253static int 4254zfs_ioc_recv(zfs_cmd_t *zc) 4255{ 4256 file_t *fp; 4257 dmu_recv_cookie_t drc; 4258 boolean_t force = (boolean_t)zc->zc_guid; 4259 int fd; 4260 int error = 0; 4261 int props_error = 0; 4262 nvlist_t *errors; 4263 offset_t off; 4264 nvlist_t *props = NULL; /* sent properties */ 4265 nvlist_t *origprops = NULL; /* existing properties */ 4266 nvlist_t *delayprops = NULL; /* sent properties applied post-receive */ 4267 char *origin = NULL; 4268 char *tosnap; 4269 char tofs[ZFS_MAXNAMELEN]; 4270 cap_rights_t rights; 4271 boolean_t first_recvd_props = B_FALSE; 4272 4273 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 4274 strchr(zc->zc_value, '@') == NULL || 4275 strchr(zc->zc_value, '%')) 4276 return (SET_ERROR(EINVAL)); 4277 4278 (void) strcpy(tofs, zc->zc_value); 4279 tosnap = strchr(tofs, '@'); 4280 *tosnap++ = '\0'; 4281 4282 if (zc->zc_nvlist_src != 0 && 4283 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 4284 zc->zc_iflags, &props)) != 0) 4285 return (error); 4286 4287 fd = zc->zc_cookie; 4288#ifdef illumos 4289 fp = getf(fd); 4290#else 4291 fget_read(curthread, fd, cap_rights_init(&rights, CAP_PREAD), &fp); 4292#endif 4293 if (fp == NULL) { 4294 nvlist_free(props); 4295 return (SET_ERROR(EBADF)); 4296 } 4297 4298 errors = fnvlist_alloc(); 4299 4300 if (zc->zc_string[0]) 4301 origin = zc->zc_string; 4302 4303 error = dmu_recv_begin(tofs, tosnap, 4304 &zc->zc_begin_record, force, zc->zc_resumable, origin, &drc); 4305 if (error != 0) 4306 goto out; 4307 4308 /* 4309 * Set properties before we receive the stream so that they are applied 4310 * to the new data. Note that we must call dmu_recv_stream() if 4311 * dmu_recv_begin() succeeds. 4312 */ 4313 if (props != NULL && !drc.drc_newfs) { 4314 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >= 4315 SPA_VERSION_RECVD_PROPS && 4316 !dsl_prop_get_hasrecvd(tofs)) 4317 first_recvd_props = B_TRUE; 4318 4319 /* 4320 * If new received properties are supplied, they are to 4321 * completely replace the existing received properties, so stash 4322 * away the existing ones. 4323 */ 4324 if (dsl_prop_get_received(tofs, &origprops) == 0) { 4325 nvlist_t *errlist = NULL; 4326 /* 4327 * Don't bother writing a property if its value won't 4328 * change (and avoid the unnecessary security checks). 4329 * 4330 * The first receive after SPA_VERSION_RECVD_PROPS is a 4331 * special case where we blow away all local properties 4332 * regardless. 4333 */ 4334 if (!first_recvd_props) 4335 props_reduce(props, origprops); 4336 if (zfs_check_clearable(tofs, origprops, &errlist) != 0) 4337 (void) nvlist_merge(errors, errlist, 0); 4338 nvlist_free(errlist); 4339 4340 if (clear_received_props(tofs, origprops, 4341 first_recvd_props ? NULL : props) != 0) 4342 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 4343 } else { 4344 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 4345 } 4346 } 4347 4348 if (props != NULL) { 4349 props_error = dsl_prop_set_hasrecvd(tofs); 4350 4351 if (props_error == 0) { 4352 delayprops = extract_delay_props(props); 4353 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED, 4354 props, errors); 4355 } 4356 } 4357 4358 off = fp->f_offset; 4359 error = dmu_recv_stream(&drc, fp, &off, zc->zc_cleanup_fd, 4360 &zc->zc_action_handle); 4361 4362 if (error == 0) { 4363 zfsvfs_t *zfsvfs = NULL; 4364 4365 if (getzfsvfs(tofs, &zfsvfs) == 0) { 4366 /* online recv */ 4367 int end_err; 4368 4369 error = zfs_suspend_fs(zfsvfs); 4370 /* 4371 * If the suspend fails, then the recv_end will 4372 * likely also fail, and clean up after itself. 4373 */ 4374 end_err = dmu_recv_end(&drc, zfsvfs); 4375 if (error == 0) 4376 error = zfs_resume_fs(zfsvfs, tofs); 4377 error = error ? error : end_err; 4378 VFS_RELE(zfsvfs->z_vfs); 4379 } else { 4380 error = dmu_recv_end(&drc, NULL); 4381 } 4382 4383 /* Set delayed properties now, after we're done receiving. */ 4384 if (delayprops != NULL && error == 0) { 4385 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED, 4386 delayprops, errors); 4387 } 4388 } 4389 4390 if (delayprops != NULL) { 4391 /* 4392 * Merge delayed props back in with initial props, in case 4393 * we're DEBUG and zfs_ioc_recv_inject_err is set (which means 4394 * we have to make sure clear_received_props() includes 4395 * the delayed properties). 4396 * 4397 * Since zfs_ioc_recv_inject_err is only in DEBUG kernels, 4398 * using ASSERT() will be just like a VERIFY. 4399 */ 4400 ASSERT(nvlist_merge(props, delayprops, 0) == 0); 4401 nvlist_free(delayprops); 4402 } 4403 4404 /* 4405 * Now that all props, initial and delayed, are set, report the prop 4406 * errors to the caller. 4407 */ 4408 if (zc->zc_nvlist_dst_size != 0 && 4409 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 || 4410 put_nvlist(zc, errors) != 0)) { 4411 /* 4412 * Caller made zc->zc_nvlist_dst less than the minimum expected 4413 * size or supplied an invalid address. 4414 */ 4415 props_error = SET_ERROR(EINVAL); 4416 } 4417 4418 zc->zc_cookie = off - fp->f_offset; 4419 if (off >= 0 && off <= MAXOFFSET_T) 4420 fp->f_offset = off; 4421 4422#ifdef DEBUG 4423 if (zfs_ioc_recv_inject_err) { 4424 zfs_ioc_recv_inject_err = B_FALSE; 4425 error = 1; 4426 } 4427#endif 4428 4429#ifdef __FreeBSD__ 4430 if (error == 0) 4431 zvol_create_minors(tofs); 4432#endif 4433 4434 /* 4435 * On error, restore the original props. 4436 */ 4437 if (error != 0 && props != NULL && !drc.drc_newfs) { 4438 if (clear_received_props(tofs, props, NULL) != 0) { 4439 /* 4440 * We failed to clear the received properties. 4441 * Since we may have left a $recvd value on the 4442 * system, we can't clear the $hasrecvd flag. 4443 */ 4444 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4445 } else if (first_recvd_props) { 4446 dsl_prop_unset_hasrecvd(tofs); 4447 } 4448 4449 if (origprops == NULL && !drc.drc_newfs) { 4450 /* We failed to stash the original properties. */ 4451 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4452 } 4453 4454 /* 4455 * dsl_props_set() will not convert RECEIVED to LOCAL on or 4456 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL 4457 * explictly if we're restoring local properties cleared in the 4458 * first new-style receive. 4459 */ 4460 if (origprops != NULL && 4461 zfs_set_prop_nvlist(tofs, (first_recvd_props ? 4462 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED), 4463 origprops, NULL) != 0) { 4464 /* 4465 * We stashed the original properties but failed to 4466 * restore them. 4467 */ 4468 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4469 } 4470 } 4471out: 4472 nvlist_free(props); 4473 nvlist_free(origprops); 4474 nvlist_free(errors); 4475 releasef(fd); 4476 4477 if (error == 0) 4478 error = props_error; 4479 4480 return (error); 4481} 4482 4483/* 4484 * inputs: 4485 * zc_name name of snapshot to send 4486 * zc_cookie file descriptor to send stream to 4487 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj) 4488 * zc_sendobj objsetid of snapshot to send 4489 * zc_fromobj objsetid of incremental fromsnap (may be zero) 4490 * zc_guid if set, estimate size of stream only. zc_cookie is ignored. 4491 * output size in zc_objset_type. 4492 * zc_flags lzc_send_flags 4493 * 4494 * outputs: 4495 * zc_objset_type estimated size, if zc_guid is set 4496 */ 4497static int 4498zfs_ioc_send(zfs_cmd_t *zc) 4499{ 4500 int error; 4501 offset_t off; 4502 boolean_t estimate = (zc->zc_guid != 0); 4503 boolean_t embedok = (zc->zc_flags & 0x1); 4504 boolean_t large_block_ok = (zc->zc_flags & 0x2); 4505 4506 if (zc->zc_obj != 0) { 4507 dsl_pool_t *dp; 4508 dsl_dataset_t *tosnap; 4509 4510 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4511 if (error != 0) 4512 return (error); 4513 4514 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap); 4515 if (error != 0) { 4516 dsl_pool_rele(dp, FTAG); 4517 return (error); 4518 } 4519 4520 if (dsl_dir_is_clone(tosnap->ds_dir)) 4521 zc->zc_fromobj = 4522 dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj; 4523 dsl_dataset_rele(tosnap, FTAG); 4524 dsl_pool_rele(dp, FTAG); 4525 } 4526 4527 if (estimate) { 4528 dsl_pool_t *dp; 4529 dsl_dataset_t *tosnap; 4530 dsl_dataset_t *fromsnap = NULL; 4531 4532 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4533 if (error != 0) 4534 return (error); 4535 4536 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap); 4537 if (error != 0) { 4538 dsl_pool_rele(dp, FTAG); 4539 return (error); 4540 } 4541 4542 if (zc->zc_fromobj != 0) { 4543 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj, 4544 FTAG, &fromsnap); 4545 if (error != 0) { 4546 dsl_dataset_rele(tosnap, FTAG); 4547 dsl_pool_rele(dp, FTAG); 4548 return (error); 4549 } 4550 } 4551 4552 error = dmu_send_estimate(tosnap, fromsnap, 4553 &zc->zc_objset_type); 4554 4555 if (fromsnap != NULL) 4556 dsl_dataset_rele(fromsnap, FTAG); 4557 dsl_dataset_rele(tosnap, FTAG); 4558 dsl_pool_rele(dp, FTAG); 4559 } else { 4560 file_t *fp; 4561 cap_rights_t rights; 4562 4563#ifdef illumos 4564 fp = getf(zc->zc_cookie); 4565#else 4566 fget_write(curthread, zc->zc_cookie, 4567 cap_rights_init(&rights, CAP_WRITE), &fp); 4568#endif 4569 if (fp == NULL) 4570 return (SET_ERROR(EBADF)); 4571 4572 off = fp->f_offset; 4573 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj, 4574 zc->zc_fromobj, embedok, large_block_ok, 4575#ifdef illumos 4576 zc->zc_cookie, fp->f_vnode, &off); 4577#else 4578 zc->zc_cookie, fp, &off); 4579#endif 4580 4581 if (off >= 0 && off <= MAXOFFSET_T) 4582 fp->f_offset = off; 4583 releasef(zc->zc_cookie); 4584 } 4585 return (error); 4586} 4587 4588/* 4589 * inputs: 4590 * zc_name name of snapshot on which to report progress 4591 * zc_cookie file descriptor of send stream 4592 * 4593 * outputs: 4594 * zc_cookie number of bytes written in send stream thus far 4595 */ 4596static int 4597zfs_ioc_send_progress(zfs_cmd_t *zc) 4598{ 4599 dsl_pool_t *dp; 4600 dsl_dataset_t *ds; 4601 dmu_sendarg_t *dsp = NULL; 4602 int error; 4603 4604 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4605 if (error != 0) 4606 return (error); 4607 4608 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds); 4609 if (error != 0) { 4610 dsl_pool_rele(dp, FTAG); 4611 return (error); 4612 } 4613 4614 mutex_enter(&ds->ds_sendstream_lock); 4615 4616 /* 4617 * Iterate over all the send streams currently active on this dataset. 4618 * If there's one which matches the specified file descriptor _and_ the 4619 * stream was started by the current process, return the progress of 4620 * that stream. 4621 */ 4622 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL; 4623 dsp = list_next(&ds->ds_sendstreams, dsp)) { 4624 if (dsp->dsa_outfd == zc->zc_cookie && 4625 dsp->dsa_proc == curproc) 4626 break; 4627 } 4628 4629 if (dsp != NULL) 4630 zc->zc_cookie = *(dsp->dsa_off); 4631 else 4632 error = SET_ERROR(ENOENT); 4633 4634 mutex_exit(&ds->ds_sendstream_lock); 4635 dsl_dataset_rele(ds, FTAG); 4636 dsl_pool_rele(dp, FTAG); 4637 return (error); 4638} 4639 4640static int 4641zfs_ioc_inject_fault(zfs_cmd_t *zc) 4642{ 4643 int id, error; 4644 4645 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id, 4646 &zc->zc_inject_record); 4647 4648 if (error == 0) 4649 zc->zc_guid = (uint64_t)id; 4650 4651 return (error); 4652} 4653 4654static int 4655zfs_ioc_clear_fault(zfs_cmd_t *zc) 4656{ 4657 return (zio_clear_fault((int)zc->zc_guid)); 4658} 4659 4660static int 4661zfs_ioc_inject_list_next(zfs_cmd_t *zc) 4662{ 4663 int id = (int)zc->zc_guid; 4664 int error; 4665 4666 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name), 4667 &zc->zc_inject_record); 4668 4669 zc->zc_guid = id; 4670 4671 return (error); 4672} 4673 4674static int 4675zfs_ioc_error_log(zfs_cmd_t *zc) 4676{ 4677 spa_t *spa; 4678 int error; 4679 size_t count = (size_t)zc->zc_nvlist_dst_size; 4680 4681 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 4682 return (error); 4683 4684 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst, 4685 &count); 4686 if (error == 0) 4687 zc->zc_nvlist_dst_size = count; 4688 else 4689 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa); 4690 4691 spa_close(spa, FTAG); 4692 4693 return (error); 4694} 4695 4696static int 4697zfs_ioc_clear(zfs_cmd_t *zc) 4698{ 4699 spa_t *spa; 4700 vdev_t *vd; 4701 int error; 4702 4703 /* 4704 * On zpool clear we also fix up missing slogs 4705 */ 4706 mutex_enter(&spa_namespace_lock); 4707 spa = spa_lookup(zc->zc_name); 4708 if (spa == NULL) { 4709 mutex_exit(&spa_namespace_lock); 4710 return (SET_ERROR(EIO)); 4711 } 4712 if (spa_get_log_state(spa) == SPA_LOG_MISSING) { 4713 /* we need to let spa_open/spa_load clear the chains */ 4714 spa_set_log_state(spa, SPA_LOG_CLEAR); 4715 } 4716 spa->spa_last_open_failed = 0; 4717 mutex_exit(&spa_namespace_lock); 4718 4719 if (zc->zc_cookie & ZPOOL_NO_REWIND) { 4720 error = spa_open(zc->zc_name, &spa, FTAG); 4721 } else { 4722 nvlist_t *policy; 4723 nvlist_t *config = NULL; 4724 4725 if (zc->zc_nvlist_src == 0) 4726 return (SET_ERROR(EINVAL)); 4727 4728 if ((error = get_nvlist(zc->zc_nvlist_src, 4729 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) { 4730 error = spa_open_rewind(zc->zc_name, &spa, FTAG, 4731 policy, &config); 4732 if (config != NULL) { 4733 int err; 4734 4735 if ((err = put_nvlist(zc, config)) != 0) 4736 error = err; 4737 nvlist_free(config); 4738 } 4739 nvlist_free(policy); 4740 } 4741 } 4742 4743 if (error != 0) 4744 return (error); 4745 4746 spa_vdev_state_enter(spa, SCL_NONE); 4747 4748 if (zc->zc_guid == 0) { 4749 vd = NULL; 4750 } else { 4751 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE); 4752 if (vd == NULL) { 4753 (void) spa_vdev_state_exit(spa, NULL, ENODEV); 4754 spa_close(spa, FTAG); 4755 return (SET_ERROR(ENODEV)); 4756 } 4757 } 4758 4759 vdev_clear(spa, vd); 4760 4761 (void) spa_vdev_state_exit(spa, NULL, 0); 4762 4763 /* 4764 * Resume any suspended I/Os. 4765 */ 4766 if (zio_resume(spa) != 0) 4767 error = SET_ERROR(EIO); 4768 4769 spa_close(spa, FTAG); 4770 4771 return (error); 4772} 4773 4774static int 4775zfs_ioc_pool_reopen(zfs_cmd_t *zc) 4776{ 4777 spa_t *spa; 4778 int error; 4779 4780 error = spa_open(zc->zc_name, &spa, FTAG); 4781 if (error != 0) 4782 return (error); 4783 4784 spa_vdev_state_enter(spa, SCL_NONE); 4785 4786 /* 4787 * If a resilver is already in progress then set the 4788 * spa_scrub_reopen flag to B_TRUE so that we don't restart 4789 * the scan as a side effect of the reopen. Otherwise, let 4790 * vdev_open() decided if a resilver is required. 4791 */ 4792 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool); 4793 vdev_reopen(spa->spa_root_vdev); 4794 spa->spa_scrub_reopen = B_FALSE; 4795 4796 (void) spa_vdev_state_exit(spa, NULL, 0); 4797 spa_close(spa, FTAG); 4798 return (0); 4799} 4800/* 4801 * inputs: 4802 * zc_name name of filesystem 4803 * zc_value name of origin snapshot 4804 * 4805 * outputs: 4806 * zc_string name of conflicting snapshot, if there is one 4807 */ 4808static int 4809zfs_ioc_promote(zfs_cmd_t *zc) 4810{ 4811 char *cp; 4812 4813 /* 4814 * We don't need to unmount *all* the origin fs's snapshots, but 4815 * it's easier. 4816 */ 4817 cp = strchr(zc->zc_value, '@'); 4818 if (cp) 4819 *cp = '\0'; 4820 (void) dmu_objset_find(zc->zc_value, 4821 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS); 4822 return (dsl_dataset_promote(zc->zc_name, zc->zc_string)); 4823} 4824 4825/* 4826 * Retrieve a single {user|group}{used|quota}@... property. 4827 * 4828 * inputs: 4829 * zc_name name of filesystem 4830 * zc_objset_type zfs_userquota_prop_t 4831 * zc_value domain name (eg. "S-1-234-567-89") 4832 * zc_guid RID/UID/GID 4833 * 4834 * outputs: 4835 * zc_cookie property value 4836 */ 4837static int 4838zfs_ioc_userspace_one(zfs_cmd_t *zc) 4839{ 4840 zfsvfs_t *zfsvfs; 4841 int error; 4842 4843 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 4844 return (SET_ERROR(EINVAL)); 4845 4846 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE); 4847 if (error != 0) 4848 return (error); 4849 4850 error = zfs_userspace_one(zfsvfs, 4851 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie); 4852 zfsvfs_rele(zfsvfs, FTAG); 4853 4854 return (error); 4855} 4856 4857/* 4858 * inputs: 4859 * zc_name name of filesystem 4860 * zc_cookie zap cursor 4861 * zc_objset_type zfs_userquota_prop_t 4862 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist) 4863 * 4864 * outputs: 4865 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t) 4866 * zc_cookie zap cursor 4867 */ 4868static int 4869zfs_ioc_userspace_many(zfs_cmd_t *zc) 4870{ 4871 zfsvfs_t *zfsvfs; 4872 int bufsize = zc->zc_nvlist_dst_size; 4873 4874 if (bufsize <= 0) 4875 return (SET_ERROR(ENOMEM)); 4876 4877 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE); 4878 if (error != 0) 4879 return (error); 4880 4881 void *buf = kmem_alloc(bufsize, KM_SLEEP); 4882 4883 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie, 4884 buf, &zc->zc_nvlist_dst_size); 4885 4886 if (error == 0) { 4887 error = ddi_copyout(buf, 4888 (void *)(uintptr_t)zc->zc_nvlist_dst, 4889 zc->zc_nvlist_dst_size, zc->zc_iflags); 4890 } 4891 kmem_free(buf, bufsize); 4892 zfsvfs_rele(zfsvfs, FTAG); 4893 4894 return (error); 4895} 4896 4897/* 4898 * inputs: 4899 * zc_name name of filesystem 4900 * 4901 * outputs: 4902 * none 4903 */ 4904static int 4905zfs_ioc_userspace_upgrade(zfs_cmd_t *zc) 4906{ 4907 objset_t *os; 4908 int error = 0; 4909 zfsvfs_t *zfsvfs; 4910 4911 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) { 4912 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) { 4913 /* 4914 * If userused is not enabled, it may be because the 4915 * objset needs to be closed & reopened (to grow the 4916 * objset_phys_t). Suspend/resume the fs will do that. 4917 */ 4918 error = zfs_suspend_fs(zfsvfs); 4919 if (error == 0) { 4920 dmu_objset_refresh_ownership(zfsvfs->z_os, 4921 zfsvfs); 4922 error = zfs_resume_fs(zfsvfs, zc->zc_name); 4923 } 4924 } 4925 if (error == 0) 4926 error = dmu_objset_userspace_upgrade(zfsvfs->z_os); 4927 VFS_RELE(zfsvfs->z_vfs); 4928 } else { 4929 /* XXX kind of reading contents without owning */ 4930 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 4931 if (error != 0) 4932 return (error); 4933 4934 error = dmu_objset_userspace_upgrade(os); 4935 dmu_objset_rele(os, FTAG); 4936 } 4937 4938 return (error); 4939} 4940 4941#ifdef illumos 4942/* 4943 * We don't want to have a hard dependency 4944 * against some special symbols in sharefs 4945 * nfs, and smbsrv. Determine them if needed when 4946 * the first file system is shared. 4947 * Neither sharefs, nfs or smbsrv are unloadable modules. 4948 */ 4949int (*znfsexport_fs)(void *arg); 4950int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t); 4951int (*zsmbexport_fs)(void *arg, boolean_t add_share); 4952 4953int zfs_nfsshare_inited; 4954int zfs_smbshare_inited; 4955 4956ddi_modhandle_t nfs_mod; 4957ddi_modhandle_t sharefs_mod; 4958ddi_modhandle_t smbsrv_mod; 4959#endif /* illumos */ 4960kmutex_t zfs_share_lock; 4961 4962#ifdef illumos 4963static int 4964zfs_init_sharefs() 4965{ 4966 int error; 4967 4968 ASSERT(MUTEX_HELD(&zfs_share_lock)); 4969 /* Both NFS and SMB shares also require sharetab support. */ 4970 if (sharefs_mod == NULL && ((sharefs_mod = 4971 ddi_modopen("fs/sharefs", 4972 KRTLD_MODE_FIRST, &error)) == NULL)) { 4973 return (SET_ERROR(ENOSYS)); 4974 } 4975 if (zshare_fs == NULL && ((zshare_fs = 4976 (int (*)(enum sharefs_sys_op, share_t *, uint32_t)) 4977 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) { 4978 return (SET_ERROR(ENOSYS)); 4979 } 4980 return (0); 4981} 4982#endif /* illumos */ 4983 4984static int 4985zfs_ioc_share(zfs_cmd_t *zc) 4986{ 4987#ifdef illumos 4988 int error; 4989 int opcode; 4990 4991 switch (zc->zc_share.z_sharetype) { 4992 case ZFS_SHARE_NFS: 4993 case ZFS_UNSHARE_NFS: 4994 if (zfs_nfsshare_inited == 0) { 4995 mutex_enter(&zfs_share_lock); 4996 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs", 4997 KRTLD_MODE_FIRST, &error)) == NULL)) { 4998 mutex_exit(&zfs_share_lock); 4999 return (SET_ERROR(ENOSYS)); 5000 } 5001 if (znfsexport_fs == NULL && 5002 ((znfsexport_fs = (int (*)(void *)) 5003 ddi_modsym(nfs_mod, 5004 "nfs_export", &error)) == NULL)) { 5005 mutex_exit(&zfs_share_lock); 5006 return (SET_ERROR(ENOSYS)); 5007 } 5008 error = zfs_init_sharefs(); 5009 if (error != 0) { 5010 mutex_exit(&zfs_share_lock); 5011 return (SET_ERROR(ENOSYS)); 5012 } 5013 zfs_nfsshare_inited = 1; 5014 mutex_exit(&zfs_share_lock); 5015 } 5016 break; 5017 case ZFS_SHARE_SMB: 5018 case ZFS_UNSHARE_SMB: 5019 if (zfs_smbshare_inited == 0) { 5020 mutex_enter(&zfs_share_lock); 5021 if (smbsrv_mod == NULL && ((smbsrv_mod = 5022 ddi_modopen("drv/smbsrv", 5023 KRTLD_MODE_FIRST, &error)) == NULL)) { 5024 mutex_exit(&zfs_share_lock); 5025 return (SET_ERROR(ENOSYS)); 5026 } 5027 if (zsmbexport_fs == NULL && ((zsmbexport_fs = 5028 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod, 5029 "smb_server_share", &error)) == NULL)) { 5030 mutex_exit(&zfs_share_lock); 5031 return (SET_ERROR(ENOSYS)); 5032 } 5033 error = zfs_init_sharefs(); 5034 if (error != 0) { 5035 mutex_exit(&zfs_share_lock); 5036 return (SET_ERROR(ENOSYS)); 5037 } 5038 zfs_smbshare_inited = 1; 5039 mutex_exit(&zfs_share_lock); 5040 } 5041 break; 5042 default: 5043 return (SET_ERROR(EINVAL)); 5044 } 5045 5046 switch (zc->zc_share.z_sharetype) { 5047 case ZFS_SHARE_NFS: 5048 case ZFS_UNSHARE_NFS: 5049 if (error = 5050 znfsexport_fs((void *) 5051 (uintptr_t)zc->zc_share.z_exportdata)) 5052 return (error); 5053 break; 5054 case ZFS_SHARE_SMB: 5055 case ZFS_UNSHARE_SMB: 5056 if (error = zsmbexport_fs((void *) 5057 (uintptr_t)zc->zc_share.z_exportdata, 5058 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ? 5059 B_TRUE: B_FALSE)) { 5060 return (error); 5061 } 5062 break; 5063 } 5064 5065 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS || 5066 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ? 5067 SHAREFS_ADD : SHAREFS_REMOVE; 5068 5069 /* 5070 * Add or remove share from sharetab 5071 */ 5072 error = zshare_fs(opcode, 5073 (void *)(uintptr_t)zc->zc_share.z_sharedata, 5074 zc->zc_share.z_sharemax); 5075 5076 return (error); 5077 5078#else /* !illumos */ 5079 return (ENOSYS); 5080#endif /* illumos */ 5081} 5082 5083ace_t full_access[] = { 5084 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0} 5085}; 5086 5087/* 5088 * inputs: 5089 * zc_name name of containing filesystem 5090 * zc_obj object # beyond which we want next in-use object # 5091 * 5092 * outputs: 5093 * zc_obj next in-use object # 5094 */ 5095static int 5096zfs_ioc_next_obj(zfs_cmd_t *zc) 5097{ 5098 objset_t *os = NULL; 5099 int error; 5100 5101 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 5102 if (error != 0) 5103 return (error); 5104 5105 error = dmu_object_next(os, &zc->zc_obj, B_FALSE, 5106 dsl_dataset_phys(os->os_dsl_dataset)->ds_prev_snap_txg); 5107 5108 dmu_objset_rele(os, FTAG); 5109 return (error); 5110} 5111 5112/* 5113 * inputs: 5114 * zc_name name of filesystem 5115 * zc_value prefix name for snapshot 5116 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process 5117 * 5118 * outputs: 5119 * zc_value short name of new snapshot 5120 */ 5121static int 5122zfs_ioc_tmp_snapshot(zfs_cmd_t *zc) 5123{ 5124 char *snap_name; 5125 char *hold_name; 5126 int error; 5127 minor_t minor; 5128 5129 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor); 5130 if (error != 0) 5131 return (error); 5132 5133 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value, 5134 (u_longlong_t)ddi_get_lbolt64()); 5135 hold_name = kmem_asprintf("%%%s", zc->zc_value); 5136 5137 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor, 5138 hold_name); 5139 if (error == 0) 5140 (void) strcpy(zc->zc_value, snap_name); 5141 strfree(snap_name); 5142 strfree(hold_name); 5143 zfs_onexit_fd_rele(zc->zc_cleanup_fd); 5144 return (error); 5145} 5146 5147/* 5148 * inputs: 5149 * zc_name name of "to" snapshot 5150 * zc_value name of "from" snapshot 5151 * zc_cookie file descriptor to write diff data on 5152 * 5153 * outputs: 5154 * dmu_diff_record_t's to the file descriptor 5155 */ 5156static int 5157zfs_ioc_diff(zfs_cmd_t *zc) 5158{ 5159 file_t *fp; 5160 cap_rights_t rights; 5161 offset_t off; 5162 int error; 5163 5164#ifdef illumos 5165 fp = getf(zc->zc_cookie); 5166#else 5167 fget_write(curthread, zc->zc_cookie, 5168 cap_rights_init(&rights, CAP_WRITE), &fp); 5169#endif 5170 if (fp == NULL) 5171 return (SET_ERROR(EBADF)); 5172 5173 off = fp->f_offset; 5174 5175#ifdef illumos 5176 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off); 5177#else 5178 error = dmu_diff(zc->zc_name, zc->zc_value, fp, &off); 5179#endif 5180 5181 if (off >= 0 && off <= MAXOFFSET_T) 5182 fp->f_offset = off; 5183 releasef(zc->zc_cookie); 5184 5185 return (error); 5186} 5187 5188#ifdef illumos 5189/* 5190 * Remove all ACL files in shares dir 5191 */ 5192static int 5193zfs_smb_acl_purge(znode_t *dzp) 5194{ 5195 zap_cursor_t zc; 5196 zap_attribute_t zap; 5197 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 5198 int error; 5199 5200 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id); 5201 (error = zap_cursor_retrieve(&zc, &zap)) == 0; 5202 zap_cursor_advance(&zc)) { 5203 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred, 5204 NULL, 0)) != 0) 5205 break; 5206 } 5207 zap_cursor_fini(&zc); 5208 return (error); 5209} 5210#endif /* illumos */ 5211 5212static int 5213zfs_ioc_smb_acl(zfs_cmd_t *zc) 5214{ 5215#ifdef illumos 5216 vnode_t *vp; 5217 znode_t *dzp; 5218 vnode_t *resourcevp = NULL; 5219 znode_t *sharedir; 5220 zfsvfs_t *zfsvfs; 5221 nvlist_t *nvlist; 5222 char *src, *target; 5223 vattr_t vattr; 5224 vsecattr_t vsec; 5225 int error = 0; 5226 5227 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, 5228 NO_FOLLOW, NULL, &vp)) != 0) 5229 return (error); 5230 5231 /* Now make sure mntpnt and dataset are ZFS */ 5232 5233 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 || 5234 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), 5235 zc->zc_name) != 0)) { 5236 VN_RELE(vp); 5237 return (SET_ERROR(EINVAL)); 5238 } 5239 5240 dzp = VTOZ(vp); 5241 zfsvfs = dzp->z_zfsvfs; 5242 ZFS_ENTER(zfsvfs); 5243 5244 /* 5245 * Create share dir if its missing. 5246 */ 5247 mutex_enter(&zfsvfs->z_lock); 5248 if (zfsvfs->z_shares_dir == 0) { 5249 dmu_tx_t *tx; 5250 5251 tx = dmu_tx_create(zfsvfs->z_os); 5252 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE, 5253 ZFS_SHARES_DIR); 5254 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); 5255 error = dmu_tx_assign(tx, TXG_WAIT); 5256 if (error != 0) { 5257 dmu_tx_abort(tx); 5258 } else { 5259 error = zfs_create_share_dir(zfsvfs, tx); 5260 dmu_tx_commit(tx); 5261 } 5262 if (error != 0) { 5263 mutex_exit(&zfsvfs->z_lock); 5264 VN_RELE(vp); 5265 ZFS_EXIT(zfsvfs); 5266 return (error); 5267 } 5268 } 5269 mutex_exit(&zfsvfs->z_lock); 5270 5271 ASSERT(zfsvfs->z_shares_dir); 5272 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) { 5273 VN_RELE(vp); 5274 ZFS_EXIT(zfsvfs); 5275 return (error); 5276 } 5277 5278 switch (zc->zc_cookie) { 5279 case ZFS_SMB_ACL_ADD: 5280 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; 5281 vattr.va_type = VREG; 5282 vattr.va_mode = S_IFREG|0777; 5283 vattr.va_uid = 0; 5284 vattr.va_gid = 0; 5285 5286 vsec.vsa_mask = VSA_ACE; 5287 vsec.vsa_aclentp = &full_access; 5288 vsec.vsa_aclentsz = sizeof (full_access); 5289 vsec.vsa_aclcnt = 1; 5290 5291 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string, 5292 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec); 5293 if (resourcevp) 5294 VN_RELE(resourcevp); 5295 break; 5296 5297 case ZFS_SMB_ACL_REMOVE: 5298 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred, 5299 NULL, 0); 5300 break; 5301 5302 case ZFS_SMB_ACL_RENAME: 5303 if ((error = get_nvlist(zc->zc_nvlist_src, 5304 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) { 5305 VN_RELE(vp); 5306 VN_RELE(ZTOV(sharedir)); 5307 ZFS_EXIT(zfsvfs); 5308 return (error); 5309 } 5310 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) || 5311 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET, 5312 &target)) { 5313 VN_RELE(vp); 5314 VN_RELE(ZTOV(sharedir)); 5315 ZFS_EXIT(zfsvfs); 5316 nvlist_free(nvlist); 5317 return (error); 5318 } 5319 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target, 5320 kcred, NULL, 0); 5321 nvlist_free(nvlist); 5322 break; 5323 5324 case ZFS_SMB_ACL_PURGE: 5325 error = zfs_smb_acl_purge(sharedir); 5326 break; 5327 5328 default: 5329 error = SET_ERROR(EINVAL); 5330 break; 5331 } 5332 5333 VN_RELE(vp); 5334 VN_RELE(ZTOV(sharedir)); 5335 5336 ZFS_EXIT(zfsvfs); 5337 5338 return (error); 5339#else /* !illumos */ 5340 return (EOPNOTSUPP); 5341#endif /* illumos */ 5342} 5343 5344/* 5345 * innvl: { 5346 * "holds" -> { snapname -> holdname (string), ... } 5347 * (optional) "cleanup_fd" -> fd (int32) 5348 * } 5349 * 5350 * outnvl: { 5351 * snapname -> error value (int32) 5352 * ... 5353 * } 5354 */ 5355/* ARGSUSED */ 5356static int 5357zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist) 5358{ 5359 nvpair_t *pair; 5360 nvlist_t *holds; 5361 int cleanup_fd = -1; 5362 int error; 5363 minor_t minor = 0; 5364 5365 error = nvlist_lookup_nvlist(args, "holds", &holds); 5366 if (error != 0) 5367 return (SET_ERROR(EINVAL)); 5368 5369 /* make sure the user didn't pass us any invalid (empty) tags */ 5370 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL; 5371 pair = nvlist_next_nvpair(holds, pair)) { 5372 char *htag; 5373 5374 error = nvpair_value_string(pair, &htag); 5375 if (error != 0) 5376 return (SET_ERROR(error)); 5377 5378 if (strlen(htag) == 0) 5379 return (SET_ERROR(EINVAL)); 5380 } 5381 5382 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) { 5383 error = zfs_onexit_fd_hold(cleanup_fd, &minor); 5384 if (error != 0) 5385 return (error); 5386 } 5387 5388 error = dsl_dataset_user_hold(holds, minor, errlist); 5389 if (minor != 0) 5390 zfs_onexit_fd_rele(cleanup_fd); 5391 return (error); 5392} 5393 5394/* 5395 * innvl is not used. 5396 * 5397 * outnvl: { 5398 * holdname -> time added (uint64 seconds since epoch) 5399 * ... 5400 * } 5401 */ 5402/* ARGSUSED */ 5403static int 5404zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl) 5405{ 5406 return (dsl_dataset_get_holds(snapname, outnvl)); 5407} 5408 5409/* 5410 * innvl: { 5411 * snapname -> { holdname, ... } 5412 * ... 5413 * } 5414 * 5415 * outnvl: { 5416 * snapname -> error value (int32) 5417 * ... 5418 * } 5419 */ 5420/* ARGSUSED */ 5421static int 5422zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist) 5423{ 5424 return (dsl_dataset_user_release(holds, errlist)); 5425} 5426 5427/* 5428 * inputs: 5429 * zc_name name of new filesystem or snapshot 5430 * zc_value full name of old snapshot 5431 * 5432 * outputs: 5433 * zc_cookie space in bytes 5434 * zc_objset_type compressed space in bytes 5435 * zc_perm_action uncompressed space in bytes 5436 */ 5437static int 5438zfs_ioc_space_written(zfs_cmd_t *zc) 5439{ 5440 int error; 5441 dsl_pool_t *dp; 5442 dsl_dataset_t *new, *old; 5443 5444 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 5445 if (error != 0) 5446 return (error); 5447 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new); 5448 if (error != 0) { 5449 dsl_pool_rele(dp, FTAG); 5450 return (error); 5451 } 5452 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old); 5453 if (error != 0) { 5454 dsl_dataset_rele(new, FTAG); 5455 dsl_pool_rele(dp, FTAG); 5456 return (error); 5457 } 5458 5459 error = dsl_dataset_space_written(old, new, &zc->zc_cookie, 5460 &zc->zc_objset_type, &zc->zc_perm_action); 5461 dsl_dataset_rele(old, FTAG); 5462 dsl_dataset_rele(new, FTAG); 5463 dsl_pool_rele(dp, FTAG); 5464 return (error); 5465} 5466 5467/* 5468 * innvl: { 5469 * "firstsnap" -> snapshot name 5470 * } 5471 * 5472 * outnvl: { 5473 * "used" -> space in bytes 5474 * "compressed" -> compressed space in bytes 5475 * "uncompressed" -> uncompressed space in bytes 5476 * } 5477 */ 5478static int 5479zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl) 5480{ 5481 int error; 5482 dsl_pool_t *dp; 5483 dsl_dataset_t *new, *old; 5484 char *firstsnap; 5485 uint64_t used, comp, uncomp; 5486 5487 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0) 5488 return (SET_ERROR(EINVAL)); 5489 5490 error = dsl_pool_hold(lastsnap, FTAG, &dp); 5491 if (error != 0) 5492 return (error); 5493 5494 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new); 5495 if (error == 0 && !new->ds_is_snapshot) { 5496 dsl_dataset_rele(new, FTAG); 5497 error = SET_ERROR(EINVAL); 5498 } 5499 if (error != 0) { 5500 dsl_pool_rele(dp, FTAG); 5501 return (error); 5502 } 5503 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old); 5504 if (error == 0 && !old->ds_is_snapshot) { 5505 dsl_dataset_rele(old, FTAG); 5506 error = SET_ERROR(EINVAL); 5507 } 5508 if (error != 0) { 5509 dsl_dataset_rele(new, FTAG); 5510 dsl_pool_rele(dp, FTAG); 5511 return (error); 5512 } 5513 5514 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp); 5515 dsl_dataset_rele(old, FTAG); 5516 dsl_dataset_rele(new, FTAG); 5517 dsl_pool_rele(dp, FTAG); 5518 fnvlist_add_uint64(outnvl, "used", used); 5519 fnvlist_add_uint64(outnvl, "compressed", comp); 5520 fnvlist_add_uint64(outnvl, "uncompressed", uncomp); 5521 return (error); 5522} 5523 5524static int 5525zfs_ioc_jail(zfs_cmd_t *zc) 5526{ 5527 5528 return (zone_dataset_attach(curthread->td_ucred, zc->zc_name, 5529 (int)zc->zc_jailid)); 5530} 5531 5532static int 5533zfs_ioc_unjail(zfs_cmd_t *zc) 5534{ 5535 5536 return (zone_dataset_detach(curthread->td_ucred, zc->zc_name, 5537 (int)zc->zc_jailid)); 5538} 5539 5540/* 5541 * innvl: { 5542 * "fd" -> file descriptor to write stream to (int32) 5543 * (optional) "fromsnap" -> full snap name to send an incremental from 5544 * (optional) "largeblockok" -> (value ignored) 5545 * indicates that blocks > 128KB are permitted 5546 * (optional) "embedok" -> (value ignored) 5547 * presence indicates DRR_WRITE_EMBEDDED records are permitted 5548 * (optional) "resume_object" and "resume_offset" -> (uint64) 5549 * if present, resume send stream from specified object and offset. 5550 * } 5551 * 5552 * outnvl is unused 5553 */ 5554/* ARGSUSED */ 5555static int 5556zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl) 5557{ 5558 cap_rights_t rights; 5559 file_t *fp; 5560 int error; 5561 offset_t off; 5562 char *fromname = NULL; 5563 int fd; 5564 boolean_t largeblockok; 5565 boolean_t embedok; 5566 uint64_t resumeobj = 0; 5567 uint64_t resumeoff = 0; 5568 5569 error = nvlist_lookup_int32(innvl, "fd", &fd); 5570 if (error != 0) 5571 return (SET_ERROR(EINVAL)); 5572 5573 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname); 5574 5575 largeblockok = nvlist_exists(innvl, "largeblockok"); 5576 embedok = nvlist_exists(innvl, "embedok"); 5577 5578 (void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj); 5579 (void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff); 5580 5581#ifdef illumos 5582 file_t *fp = getf(fd); 5583#else 5584 fget_write(curthread, fd, cap_rights_init(&rights, CAP_WRITE), &fp); 5585#endif 5586 if (fp == NULL) 5587 return (SET_ERROR(EBADF)); 5588 5589 off = fp->f_offset; 5590 error = dmu_send(snapname, fromname, embedok, largeblockok, fd, 5591#ifdef illumos 5592 resumeobj, resumeoff, fp->f_vnode, &off); 5593#else 5594 resumeobj, resumeoff, fp, &off); 5595#endif 5596 5597#ifdef illumos 5598 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0) 5599 fp->f_offset = off; 5600#else 5601 fp->f_offset = off; 5602#endif 5603 5604 releasef(fd); 5605 return (error); 5606} 5607 5608/* 5609 * Determine approximately how large a zfs send stream will be -- the number 5610 * of bytes that will be written to the fd supplied to zfs_ioc_send_new(). 5611 * 5612 * innvl: { 5613 * (optional) "from" -> full snap or bookmark name to send an incremental 5614 * from 5615 * } 5616 * 5617 * outnvl: { 5618 * "space" -> bytes of space (uint64) 5619 * } 5620 */ 5621static int 5622zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl) 5623{ 5624 dsl_pool_t *dp; 5625 dsl_dataset_t *tosnap; 5626 int error; 5627 char *fromname; 5628 uint64_t space; 5629 5630 error = dsl_pool_hold(snapname, FTAG, &dp); 5631 if (error != 0) 5632 return (error); 5633 5634 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap); 5635 if (error != 0) { 5636 dsl_pool_rele(dp, FTAG); 5637 return (error); 5638 } 5639 5640 error = nvlist_lookup_string(innvl, "from", &fromname); 5641 if (error == 0) { 5642 if (strchr(fromname, '@') != NULL) { 5643 /* 5644 * If from is a snapshot, hold it and use the more 5645 * efficient dmu_send_estimate to estimate send space 5646 * size using deadlists. 5647 */ 5648 dsl_dataset_t *fromsnap; 5649 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap); 5650 if (error != 0) 5651 goto out; 5652 error = dmu_send_estimate(tosnap, fromsnap, &space); 5653 dsl_dataset_rele(fromsnap, FTAG); 5654 } else if (strchr(fromname, '#') != NULL) { 5655 /* 5656 * If from is a bookmark, fetch the creation TXG of the 5657 * snapshot it was created from and use that to find 5658 * blocks that were born after it. 5659 */ 5660 zfs_bookmark_phys_t frombm; 5661 5662 error = dsl_bookmark_lookup(dp, fromname, tosnap, 5663 &frombm); 5664 if (error != 0) 5665 goto out; 5666 error = dmu_send_estimate_from_txg(tosnap, 5667 frombm.zbm_creation_txg, &space); 5668 } else { 5669 /* 5670 * from is not properly formatted as a snapshot or 5671 * bookmark 5672 */ 5673 error = SET_ERROR(EINVAL); 5674 goto out; 5675 } 5676 } else { 5677 // If estimating the size of a full send, use dmu_send_estimate 5678 error = dmu_send_estimate(tosnap, NULL, &space); 5679 } 5680 5681 fnvlist_add_uint64(outnvl, "space", space); 5682 5683out: 5684 dsl_dataset_rele(tosnap, FTAG); 5685 dsl_pool_rele(dp, FTAG); 5686 return (error); 5687} 5688 5689static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST]; 5690 5691static void 5692zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5693 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck, 5694 boolean_t log_history, zfs_ioc_poolcheck_t pool_check) 5695{ 5696 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST]; 5697 5698 ASSERT3U(ioc, >=, ZFS_IOC_FIRST); 5699 ASSERT3U(ioc, <, ZFS_IOC_LAST); 5700 ASSERT3P(vec->zvec_legacy_func, ==, NULL); 5701 ASSERT3P(vec->zvec_func, ==, NULL); 5702 5703 vec->zvec_legacy_func = func; 5704 vec->zvec_secpolicy = secpolicy; 5705 vec->zvec_namecheck = namecheck; 5706 vec->zvec_allow_log = log_history; 5707 vec->zvec_pool_check = pool_check; 5708} 5709 5710/* 5711 * See the block comment at the beginning of this file for details on 5712 * each argument to this function. 5713 */ 5714static void 5715zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func, 5716 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck, 5717 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist, 5718 boolean_t allow_log) 5719{ 5720 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST]; 5721 5722 ASSERT3U(ioc, >=, ZFS_IOC_FIRST); 5723 ASSERT3U(ioc, <, ZFS_IOC_LAST); 5724 ASSERT3P(vec->zvec_legacy_func, ==, NULL); 5725 ASSERT3P(vec->zvec_func, ==, NULL); 5726 5727 /* if we are logging, the name must be valid */ 5728 ASSERT(!allow_log || namecheck != NO_NAME); 5729 5730 vec->zvec_name = name; 5731 vec->zvec_func = func; 5732 vec->zvec_secpolicy = secpolicy; 5733 vec->zvec_namecheck = namecheck; 5734 vec->zvec_pool_check = pool_check; 5735 vec->zvec_smush_outnvlist = smush_outnvlist; 5736 vec->zvec_allow_log = allow_log; 5737} 5738 5739static void 5740zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5741 zfs_secpolicy_func_t *secpolicy, boolean_t log_history, 5742 zfs_ioc_poolcheck_t pool_check) 5743{ 5744 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5745 POOL_NAME, log_history, pool_check); 5746} 5747 5748static void 5749zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5750 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check) 5751{ 5752 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5753 DATASET_NAME, B_FALSE, pool_check); 5754} 5755 5756static void 5757zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func) 5758{ 5759 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config, 5760 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5761} 5762 5763static void 5764zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5765 zfs_secpolicy_func_t *secpolicy) 5766{ 5767 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5768 NO_NAME, B_FALSE, POOL_CHECK_NONE); 5769} 5770 5771static void 5772zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc, 5773 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy) 5774{ 5775 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5776 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED); 5777} 5778 5779static void 5780zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func) 5781{ 5782 zfs_ioctl_register_dataset_read_secpolicy(ioc, func, 5783 zfs_secpolicy_read); 5784} 5785 5786static void 5787zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5788 zfs_secpolicy_func_t *secpolicy) 5789{ 5790 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5791 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5792} 5793 5794static void 5795zfs_ioctl_init(void) 5796{ 5797 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT, 5798 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME, 5799 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5800 5801 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY, 5802 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME, 5803 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE); 5804 5805 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS, 5806 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME, 5807 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5808 5809 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW, 5810 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME, 5811 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5812 5813 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE, 5814 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME, 5815 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5816 5817 zfs_ioctl_register("create", ZFS_IOC_CREATE, 5818 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME, 5819 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5820 5821 zfs_ioctl_register("clone", ZFS_IOC_CLONE, 5822 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME, 5823 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5824 5825 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS, 5826 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME, 5827 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5828 5829 zfs_ioctl_register("hold", ZFS_IOC_HOLD, 5830 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME, 5831 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5832 zfs_ioctl_register("release", ZFS_IOC_RELEASE, 5833 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME, 5834 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5835 5836 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS, 5837 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME, 5838 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5839 5840 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK, 5841 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, 5842 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE); 5843 5844 zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK, 5845 zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME, 5846 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5847 5848 zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS, 5849 zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME, 5850 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5851 5852 zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS, 5853 zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks, 5854 POOL_NAME, 5855 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5856 5857 /* IOCTLS that use the legacy function signature */ 5858 5859 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze, 5860 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY); 5861 5862 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create, 5863 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5864 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN, 5865 zfs_ioc_pool_scan); 5866 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE, 5867 zfs_ioc_pool_upgrade); 5868 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD, 5869 zfs_ioc_vdev_add); 5870 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE, 5871 zfs_ioc_vdev_remove); 5872 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE, 5873 zfs_ioc_vdev_set_state); 5874 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH, 5875 zfs_ioc_vdev_attach); 5876 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH, 5877 zfs_ioc_vdev_detach); 5878 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH, 5879 zfs_ioc_vdev_setpath); 5880 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU, 5881 zfs_ioc_vdev_setfru); 5882 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS, 5883 zfs_ioc_pool_set_props); 5884 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT, 5885 zfs_ioc_vdev_split); 5886 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID, 5887 zfs_ioc_pool_reguid); 5888 5889 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS, 5890 zfs_ioc_pool_configs, zfs_secpolicy_none); 5891 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT, 5892 zfs_ioc_pool_tryimport, zfs_secpolicy_config); 5893 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT, 5894 zfs_ioc_inject_fault, zfs_secpolicy_inject); 5895 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT, 5896 zfs_ioc_clear_fault, zfs_secpolicy_inject); 5897 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT, 5898 zfs_ioc_inject_list_next, zfs_secpolicy_inject); 5899 5900 /* 5901 * pool destroy, and export don't log the history as part of 5902 * zfsdev_ioctl, but rather zfs_ioc_pool_export 5903 * does the logging of those commands. 5904 */ 5905 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy, 5906 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE); 5907 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export, 5908 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE); 5909 5910 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats, 5911 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE); 5912 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props, 5913 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE); 5914 5915 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log, 5916 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_NONE); 5917 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME, 5918 zfs_ioc_dsobj_to_dsname, 5919 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_NONE); 5920 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY, 5921 zfs_ioc_pool_get_history, 5922 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED); 5923 5924 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import, 5925 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5926 5927 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear, 5928 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5929 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen, 5930 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED); 5931 5932 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN, 5933 zfs_ioc_space_written); 5934 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS, 5935 zfs_ioc_objset_recvd_props); 5936 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ, 5937 zfs_ioc_next_obj); 5938 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL, 5939 zfs_ioc_get_fsacl); 5940 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS, 5941 zfs_ioc_objset_stats); 5942 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS, 5943 zfs_ioc_objset_zplprops); 5944 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT, 5945 zfs_ioc_dataset_list_next); 5946 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT, 5947 zfs_ioc_snapshot_list_next); 5948 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS, 5949 zfs_ioc_send_progress); 5950 5951 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF, 5952 zfs_ioc_diff, zfs_secpolicy_diff); 5953 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS, 5954 zfs_ioc_obj_to_stats, zfs_secpolicy_diff); 5955 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH, 5956 zfs_ioc_obj_to_path, zfs_secpolicy_diff); 5957 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE, 5958 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one); 5959 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY, 5960 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many); 5961 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND, 5962 zfs_ioc_send, zfs_secpolicy_send); 5963 5964 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop, 5965 zfs_secpolicy_none); 5966 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy, 5967 zfs_secpolicy_destroy); 5968 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename, 5969 zfs_secpolicy_rename); 5970 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv, 5971 zfs_secpolicy_recv); 5972 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote, 5973 zfs_secpolicy_promote); 5974 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP, 5975 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop); 5976 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl, 5977 zfs_secpolicy_set_fsacl); 5978 5979 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share, 5980 zfs_secpolicy_share, POOL_CHECK_NONE); 5981 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl, 5982 zfs_secpolicy_smb_acl, POOL_CHECK_NONE); 5983 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE, 5984 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade, 5985 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5986 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT, 5987 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot, 5988 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5989 5990#ifdef __FreeBSD__ 5991 zfs_ioctl_register_dataset_nolog(ZFS_IOC_JAIL, zfs_ioc_jail, 5992 zfs_secpolicy_config, POOL_CHECK_NONE); 5993 zfs_ioctl_register_dataset_nolog(ZFS_IOC_UNJAIL, zfs_ioc_unjail, 5994 zfs_secpolicy_config, POOL_CHECK_NONE); 5995#endif 5996} 5997 5998int 5999pool_status_check(const char *name, zfs_ioc_namecheck_t type, 6000 zfs_ioc_poolcheck_t check) 6001{ 6002 spa_t *spa; 6003 int error; 6004 6005 ASSERT(type == POOL_NAME || type == DATASET_NAME); 6006 6007 if (check & POOL_CHECK_NONE) 6008 return (0); 6009 6010 error = spa_open(name, &spa, FTAG); 6011 if (error == 0) { 6012 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa)) 6013 error = SET_ERROR(EAGAIN); 6014 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa)) 6015 error = SET_ERROR(EROFS); 6016 spa_close(spa, FTAG); 6017 } 6018 return (error); 6019} 6020 6021/* 6022 * Find a free minor number. 6023 */ 6024minor_t 6025zfsdev_minor_alloc(void) 6026{ 6027 static minor_t last_minor; 6028 minor_t m; 6029 6030 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 6031 6032 for (m = last_minor + 1; m != last_minor; m++) { 6033 if (m > ZFSDEV_MAX_MINOR) 6034 m = 1; 6035 if (ddi_get_soft_state(zfsdev_state, m) == NULL) { 6036 last_minor = m; 6037 return (m); 6038 } 6039 } 6040 6041 return (0); 6042} 6043 6044static int 6045zfs_ctldev_init(struct cdev *devp) 6046{ 6047 minor_t minor; 6048 zfs_soft_state_t *zs; 6049 6050 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 6051 6052 minor = zfsdev_minor_alloc(); 6053 if (minor == 0) 6054 return (SET_ERROR(ENXIO)); 6055 6056 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS) 6057 return (SET_ERROR(EAGAIN)); 6058 6059 devfs_set_cdevpriv((void *)(uintptr_t)minor, zfsdev_close); 6060 6061 zs = ddi_get_soft_state(zfsdev_state, minor); 6062 zs->zss_type = ZSST_CTLDEV; 6063 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data); 6064 6065 return (0); 6066} 6067 6068static void 6069zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor) 6070{ 6071 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 6072 6073 zfs_onexit_destroy(zo); 6074 ddi_soft_state_free(zfsdev_state, minor); 6075} 6076 6077void * 6078zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which) 6079{ 6080 zfs_soft_state_t *zp; 6081 6082 zp = ddi_get_soft_state(zfsdev_state, minor); 6083 if (zp == NULL || zp->zss_type != which) 6084 return (NULL); 6085 6086 return (zp->zss_data); 6087} 6088 6089static int 6090zfsdev_open(struct cdev *devp, int flag, int mode, struct thread *td) 6091{ 6092 int error = 0; 6093 6094#ifdef illumos 6095 if (getminor(*devp) != 0) 6096 return (zvol_open(devp, flag, otyp, cr)); 6097#endif 6098 6099 /* This is the control device. Allocate a new minor if requested. */ 6100 if (flag & FEXCL) { 6101 mutex_enter(&spa_namespace_lock); 6102 error = zfs_ctldev_init(devp); 6103 mutex_exit(&spa_namespace_lock); 6104 } 6105 6106 return (error); 6107} 6108 6109static void 6110zfsdev_close(void *data) 6111{ 6112 zfs_onexit_t *zo; 6113 minor_t minor = (minor_t)(uintptr_t)data; 6114 6115 if (minor == 0) 6116 return; 6117 6118 mutex_enter(&spa_namespace_lock); 6119 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV); 6120 if (zo == NULL) { 6121 mutex_exit(&spa_namespace_lock); 6122 return; 6123 } 6124 zfs_ctldev_destroy(zo, minor); 6125 mutex_exit(&spa_namespace_lock); 6126} 6127 6128static int 6129zfsdev_ioctl(struct cdev *dev, u_long zcmd, caddr_t arg, int flag, 6130 struct thread *td) 6131{ 6132 zfs_cmd_t *zc; 6133 uint_t vecnum; 6134 int error, rc, len; 6135#ifdef illumos 6136 minor_t minor = getminor(dev); 6137#else 6138 zfs_iocparm_t *zc_iocparm; 6139 int cflag, cmd, oldvecnum; 6140 boolean_t newioc, compat; 6141 void *compat_zc = NULL; 6142 cred_t *cr = td->td_ucred; 6143#endif 6144 const zfs_ioc_vec_t *vec; 6145 char *saved_poolname = NULL; 6146 nvlist_t *innvl = NULL; 6147 6148 cflag = ZFS_CMD_COMPAT_NONE; 6149 compat = B_FALSE; 6150 newioc = B_TRUE; /* "new" style (zfs_iocparm_t) ioctl */ 6151 6152 len = IOCPARM_LEN(zcmd); 6153 vecnum = cmd = zcmd & 0xff; 6154 6155 /* 6156 * Check if we are talking to supported older binaries 6157 * and translate zfs_cmd if necessary 6158 */ 6159 if (len != sizeof(zfs_iocparm_t)) { 6160 newioc = B_FALSE; 6161 compat = B_TRUE; 6162 6163 vecnum = cmd; 6164 6165 switch (len) { 6166 case sizeof(zfs_cmd_zcmd_t): 6167 cflag = ZFS_CMD_COMPAT_LZC; 6168 break; 6169 case sizeof(zfs_cmd_deadman_t): 6170 cflag = ZFS_CMD_COMPAT_DEADMAN; 6171 break; 6172 case sizeof(zfs_cmd_v28_t): 6173 cflag = ZFS_CMD_COMPAT_V28; 6174 break; 6175 case sizeof(zfs_cmd_v15_t): 6176 cflag = ZFS_CMD_COMPAT_V15; 6177 vecnum = zfs_ioctl_v15_to_v28[cmd]; 6178 6179 /* 6180 * Return without further handling 6181 * if the command is blacklisted. 6182 */ 6183 if (vecnum == ZFS_IOC_COMPAT_PASS) 6184 return (0); 6185 else if (vecnum == ZFS_IOC_COMPAT_FAIL) 6186 return (ENOTSUP); 6187 break; 6188 default: 6189 return (EINVAL); 6190 } 6191 } 6192 6193#ifdef illumos 6194 vecnum = cmd - ZFS_IOC_FIRST; 6195 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip)); 6196#endif 6197 6198 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0])) 6199 return (SET_ERROR(EINVAL)); 6200 vec = &zfs_ioc_vec[vecnum]; 6201 6202 zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP); 6203 6204#ifdef illumos 6205 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag); 6206 if (error != 0) { 6207 error = SET_ERROR(EFAULT); 6208 goto out; 6209 } 6210#else /* !illumos */ 6211 bzero(zc, sizeof(zfs_cmd_t)); 6212 6213 if (newioc) { 6214 zc_iocparm = (void *)arg; 6215 6216 switch (zc_iocparm->zfs_ioctl_version) { 6217 case ZFS_IOCVER_CURRENT: 6218 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_t)) { 6219 error = SET_ERROR(EINVAL); 6220 goto out; 6221 } 6222 break; 6223 case ZFS_IOCVER_RESUME: 6224 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_resume_t)) { 6225 error = SET_ERROR(EFAULT); 6226 goto out; 6227 } 6228 compat = B_TRUE; 6229 cflag = ZFS_CMD_COMPAT_RESUME; 6230 break; 6231 case ZFS_IOCVER_EDBP: 6232 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_edbp_t)) { 6233 error = SET_ERROR(EFAULT); 6234 goto out; 6235 } 6236 compat = B_TRUE; 6237 cflag = ZFS_CMD_COMPAT_EDBP; 6238 break; 6239 case ZFS_IOCVER_ZCMD: 6240 if (zc_iocparm->zfs_cmd_size > sizeof(zfs_cmd_t) || 6241 zc_iocparm->zfs_cmd_size < sizeof(zfs_cmd_zcmd_t)) { 6242 error = SET_ERROR(EFAULT); 6243 goto out; 6244 } 6245 compat = B_TRUE; 6246 cflag = ZFS_CMD_COMPAT_ZCMD; 6247 break; 6248 default: 6249 error = SET_ERROR(EINVAL); 6250 goto out; 6251 /* NOTREACHED */ 6252 } 6253 6254 if (compat) { 6255 ASSERT(sizeof(zfs_cmd_t) >= zc_iocparm->zfs_cmd_size); 6256 compat_zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP); 6257 bzero(compat_zc, sizeof(zfs_cmd_t)); 6258 6259 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd, 6260 compat_zc, zc_iocparm->zfs_cmd_size, flag); 6261 if (error != 0) { 6262 error = SET_ERROR(EFAULT); 6263 goto out; 6264 } 6265 } else { 6266 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd, 6267 zc, zc_iocparm->zfs_cmd_size, flag); 6268 if (error != 0) { 6269 error = SET_ERROR(EFAULT); 6270 goto out; 6271 } 6272 } 6273 } 6274 6275 if (compat) { 6276 if (newioc) { 6277 ASSERT(compat_zc != NULL); 6278 zfs_cmd_compat_get(zc, compat_zc, cflag); 6279 } else { 6280 ASSERT(compat_zc == NULL); 6281 zfs_cmd_compat_get(zc, arg, cflag); 6282 } 6283 oldvecnum = vecnum; 6284 error = zfs_ioctl_compat_pre(zc, &vecnum, cflag); 6285 if (error != 0) 6286 goto out; 6287 if (oldvecnum != vecnum) 6288 vec = &zfs_ioc_vec[vecnum]; 6289 } 6290#endif /* !illumos */ 6291 6292 zc->zc_iflags = flag & FKIOCTL; 6293 if (zc->zc_nvlist_src_size != 0) { 6294 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 6295 zc->zc_iflags, &innvl); 6296 if (error != 0) 6297 goto out; 6298 } 6299 6300 /* rewrite innvl for backwards compatibility */ 6301 if (compat) 6302 innvl = zfs_ioctl_compat_innvl(zc, innvl, vecnum, cflag); 6303 6304 /* 6305 * Ensure that all pool/dataset names are valid before we pass down to 6306 * the lower layers. 6307 */ 6308 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0'; 6309 switch (vec->zvec_namecheck) { 6310 case POOL_NAME: 6311 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0) 6312 error = SET_ERROR(EINVAL); 6313 else 6314 error = pool_status_check(zc->zc_name, 6315 vec->zvec_namecheck, vec->zvec_pool_check); 6316 break; 6317 6318 case DATASET_NAME: 6319 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0) 6320 error = SET_ERROR(EINVAL); 6321 else 6322 error = pool_status_check(zc->zc_name, 6323 vec->zvec_namecheck, vec->zvec_pool_check); 6324 break; 6325 6326 case NO_NAME: 6327 break; 6328 } 6329 6330 if (error == 0 && !(flag & FKIOCTL)) 6331 error = vec->zvec_secpolicy(zc, innvl, cr); 6332 6333 if (error != 0) 6334 goto out; 6335 6336 /* legacy ioctls can modify zc_name */ 6337 len = strcspn(zc->zc_name, "/@#") + 1; 6338 saved_poolname = kmem_alloc(len, KM_SLEEP); 6339 (void) strlcpy(saved_poolname, zc->zc_name, len); 6340 6341 if (vec->zvec_func != NULL) { 6342 nvlist_t *outnvl; 6343 int puterror = 0; 6344 spa_t *spa; 6345 nvlist_t *lognv = NULL; 6346 6347 ASSERT(vec->zvec_legacy_func == NULL); 6348 6349 /* 6350 * Add the innvl to the lognv before calling the func, 6351 * in case the func changes the innvl. 6352 */ 6353 if (vec->zvec_allow_log) { 6354 lognv = fnvlist_alloc(); 6355 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL, 6356 vec->zvec_name); 6357 if (!nvlist_empty(innvl)) { 6358 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL, 6359 innvl); 6360 } 6361 } 6362 6363 outnvl = fnvlist_alloc(); 6364 error = vec->zvec_func(zc->zc_name, innvl, outnvl); 6365 6366 if (error == 0 && vec->zvec_allow_log && 6367 spa_open(zc->zc_name, &spa, FTAG) == 0) { 6368 if (!nvlist_empty(outnvl)) { 6369 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL, 6370 outnvl); 6371 } 6372 (void) spa_history_log_nvl(spa, lognv); 6373 spa_close(spa, FTAG); 6374 } 6375 fnvlist_free(lognv); 6376 6377 /* rewrite outnvl for backwards compatibility */ 6378 if (compat) 6379 outnvl = zfs_ioctl_compat_outnvl(zc, outnvl, vecnum, 6380 cflag); 6381 6382 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) { 6383 int smusherror = 0; 6384 if (vec->zvec_smush_outnvlist) { 6385 smusherror = nvlist_smush(outnvl, 6386 zc->zc_nvlist_dst_size); 6387 } 6388 if (smusherror == 0) 6389 puterror = put_nvlist(zc, outnvl); 6390 } 6391 6392 if (puterror != 0) 6393 error = puterror; 6394 6395 nvlist_free(outnvl); 6396 } else { 6397 error = vec->zvec_legacy_func(zc); 6398 } 6399 6400out: 6401 nvlist_free(innvl); 6402 6403#ifdef illumos 6404 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag); 6405 if (error == 0 && rc != 0) 6406 error = SET_ERROR(EFAULT); 6407#else 6408 if (compat) { 6409 zfs_ioctl_compat_post(zc, cmd, cflag); 6410 if (newioc) { 6411 ASSERT(compat_zc != NULL); 6412 ASSERT(sizeof(zfs_cmd_t) >= zc_iocparm->zfs_cmd_size); 6413 6414 zfs_cmd_compat_put(zc, compat_zc, vecnum, cflag); 6415 rc = ddi_copyout(compat_zc, 6416 (void *)(uintptr_t)zc_iocparm->zfs_cmd, 6417 zc_iocparm->zfs_cmd_size, flag); 6418 if (error == 0 && rc != 0) 6419 error = SET_ERROR(EFAULT); 6420 kmem_free(compat_zc, sizeof (zfs_cmd_t)); 6421 } else { 6422 zfs_cmd_compat_put(zc, arg, vecnum, cflag); 6423 } 6424 } else { 6425 ASSERT(newioc); 6426 6427 rc = ddi_copyout(zc, (void *)(uintptr_t)zc_iocparm->zfs_cmd, 6428 sizeof (zfs_cmd_t), flag); 6429 if (error == 0 && rc != 0) 6430 error = SET_ERROR(EFAULT); 6431 } 6432#endif 6433 if (error == 0 && vec->zvec_allow_log) { 6434 char *s = tsd_get(zfs_allow_log_key); 6435 if (s != NULL) 6436 strfree(s); 6437 (void) tsd_set(zfs_allow_log_key, saved_poolname); 6438 } else { 6439 if (saved_poolname != NULL) 6440 strfree(saved_poolname); 6441 } 6442 6443 kmem_free(zc, sizeof (zfs_cmd_t)); 6444 return (error); 6445} 6446 6447#ifdef illumos 6448static int 6449zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd) 6450{ 6451 if (cmd != DDI_ATTACH) 6452 return (DDI_FAILURE); 6453 6454 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0, 6455 DDI_PSEUDO, 0) == DDI_FAILURE) 6456 return (DDI_FAILURE); 6457 6458 zfs_dip = dip; 6459 6460 ddi_report_dev(dip); 6461 6462 return (DDI_SUCCESS); 6463} 6464 6465static int 6466zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd) 6467{ 6468 if (spa_busy() || zfs_busy() || zvol_busy()) 6469 return (DDI_FAILURE); 6470 6471 if (cmd != DDI_DETACH) 6472 return (DDI_FAILURE); 6473 6474 zfs_dip = NULL; 6475 6476 ddi_prop_remove_all(dip); 6477 ddi_remove_minor_node(dip, NULL); 6478 6479 return (DDI_SUCCESS); 6480} 6481 6482/*ARGSUSED*/ 6483static int 6484zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result) 6485{ 6486 switch (infocmd) { 6487 case DDI_INFO_DEVT2DEVINFO: 6488 *result = zfs_dip; 6489 return (DDI_SUCCESS); 6490 6491 case DDI_INFO_DEVT2INSTANCE: 6492 *result = (void *)0; 6493 return (DDI_SUCCESS); 6494 } 6495 6496 return (DDI_FAILURE); 6497} 6498#endif /* illumos */ 6499 6500/* 6501 * OK, so this is a little weird. 6502 * 6503 * /dev/zfs is the control node, i.e. minor 0. 6504 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0. 6505 * 6506 * /dev/zfs has basically nothing to do except serve up ioctls, 6507 * so most of the standard driver entry points are in zvol.c. 6508 */ 6509#ifdef illumos 6510static struct cb_ops zfs_cb_ops = { 6511 zfsdev_open, /* open */ 6512 zfsdev_close, /* close */ 6513 zvol_strategy, /* strategy */ 6514 nodev, /* print */ 6515 zvol_dump, /* dump */ 6516 zvol_read, /* read */ 6517 zvol_write, /* write */ 6518 zfsdev_ioctl, /* ioctl */ 6519 nodev, /* devmap */ 6520 nodev, /* mmap */ 6521 nodev, /* segmap */ 6522 nochpoll, /* poll */ 6523 ddi_prop_op, /* prop_op */ 6524 NULL, /* streamtab */ 6525 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */ 6526 CB_REV, /* version */ 6527 nodev, /* async read */ 6528 nodev, /* async write */ 6529}; 6530 6531static struct dev_ops zfs_dev_ops = { 6532 DEVO_REV, /* version */ 6533 0, /* refcnt */ 6534 zfs_info, /* info */ 6535 nulldev, /* identify */ 6536 nulldev, /* probe */ 6537 zfs_attach, /* attach */ 6538 zfs_detach, /* detach */ 6539 nodev, /* reset */ 6540 &zfs_cb_ops, /* driver operations */ 6541 NULL, /* no bus operations */ 6542 NULL, /* power */ 6543 ddi_quiesce_not_needed, /* quiesce */ 6544}; 6545 6546static struct modldrv zfs_modldrv = { 6547 &mod_driverops, 6548 "ZFS storage pool", 6549 &zfs_dev_ops 6550}; 6551 6552static struct modlinkage modlinkage = { 6553 MODREV_1, 6554 (void *)&zfs_modlfs, 6555 (void *)&zfs_modldrv, 6556 NULL 6557}; 6558#endif /* illumos */ 6559 6560static struct cdevsw zfs_cdevsw = { 6561 .d_version = D_VERSION, 6562 .d_open = zfsdev_open, 6563 .d_ioctl = zfsdev_ioctl, 6564 .d_name = ZFS_DEV_NAME 6565}; 6566 6567static void 6568zfs_allow_log_destroy(void *arg) 6569{ 6570 char *poolname = arg; 6571 strfree(poolname); 6572} 6573 6574static void 6575zfsdev_init(void) 6576{ 6577 zfsdev = make_dev(&zfs_cdevsw, 0x0, UID_ROOT, GID_OPERATOR, 0666, 6578 ZFS_DEV_NAME); 6579} 6580 6581static void 6582zfsdev_fini(void) 6583{ 6584 if (zfsdev != NULL) 6585 destroy_dev(zfsdev); 6586} 6587 6588static struct root_hold_token *zfs_root_token; 6589struct proc *zfsproc; 6590 6591#ifdef illumos 6592int 6593_init(void) 6594{ 6595 int error; 6596 6597 spa_init(FREAD | FWRITE); 6598 zfs_init(); 6599 zvol_init(); 6600 zfs_ioctl_init(); 6601 6602 if ((error = mod_install(&modlinkage)) != 0) { 6603 zvol_fini(); 6604 zfs_fini(); 6605 spa_fini(); 6606 return (error); 6607 } 6608 6609 tsd_create(&zfs_fsyncer_key, NULL); 6610 tsd_create(&rrw_tsd_key, rrw_tsd_destroy); 6611 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy); 6612 6613 error = ldi_ident_from_mod(&modlinkage, &zfs_li); 6614 ASSERT(error == 0); 6615 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL); 6616 6617 return (0); 6618} 6619 6620int 6621_fini(void) 6622{ 6623 int error; 6624 6625 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled) 6626 return (SET_ERROR(EBUSY)); 6627 6628 if ((error = mod_remove(&modlinkage)) != 0) 6629 return (error); 6630 6631 zvol_fini(); 6632 zfs_fini(); 6633 spa_fini(); 6634 if (zfs_nfsshare_inited) 6635 (void) ddi_modclose(nfs_mod); 6636 if (zfs_smbshare_inited) 6637 (void) ddi_modclose(smbsrv_mod); 6638 if (zfs_nfsshare_inited || zfs_smbshare_inited) 6639 (void) ddi_modclose(sharefs_mod); 6640 6641 tsd_destroy(&zfs_fsyncer_key); 6642 ldi_ident_release(zfs_li); 6643 zfs_li = NULL; 6644 mutex_destroy(&zfs_share_lock); 6645 6646 return (error); 6647} 6648 6649int 6650_info(struct modinfo *modinfop) 6651{ 6652 return (mod_info(&modlinkage, modinfop)); 6653} 6654#endif /* illumos */ 6655 6656static int zfs__init(void); 6657static int zfs__fini(void); 6658static void zfs_shutdown(void *, int); 6659 6660static eventhandler_tag zfs_shutdown_event_tag; 6661 6662#ifdef __FreeBSD__ 6663#define ZFS_MIN_KSTACK_PAGES 4 6664#endif 6665 6666int 6667zfs__init(void) 6668{ 6669 6670#ifdef __FreeBSD__ 6671#if KSTACK_PAGES < ZFS_MIN_KSTACK_PAGES 6672 printf("ZFS NOTICE: KSTACK_PAGES is %d which could result in stack " 6673 "overflow panic!\nPlease consider adding " 6674 "'options KSTACK_PAGES=%d' to your kernel config\n", KSTACK_PAGES, 6675 ZFS_MIN_KSTACK_PAGES); 6676#endif 6677#endif 6678 zfs_root_token = root_mount_hold("ZFS"); 6679 6680 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL); 6681 6682 spa_init(FREAD | FWRITE); 6683 zfs_init(); 6684 zvol_init(); 6685 zfs_ioctl_init(); 6686 6687 tsd_create(&zfs_fsyncer_key, NULL); 6688 tsd_create(&rrw_tsd_key, rrw_tsd_destroy); 6689 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy); 6690 6691 printf("ZFS storage pool version: features support (" SPA_VERSION_STRING ")\n"); 6692 root_mount_rel(zfs_root_token); 6693 6694 zfsdev_init(); 6695 6696 return (0); 6697} 6698 6699int 6700zfs__fini(void) 6701{ 6702 if (spa_busy() || zfs_busy() || zvol_busy() || 6703 zio_injection_enabled) { 6704 return (EBUSY); 6705 } 6706 6707 zfsdev_fini(); 6708 zvol_fini(); 6709 zfs_fini(); 6710 spa_fini(); 6711 6712 tsd_destroy(&zfs_fsyncer_key); 6713 tsd_destroy(&rrw_tsd_key); 6714 tsd_destroy(&zfs_allow_log_key); 6715 6716 mutex_destroy(&zfs_share_lock); 6717 6718 return (0); 6719} 6720 6721static void 6722zfs_shutdown(void *arg __unused, int howto __unused) 6723{ 6724 6725 /* 6726 * ZFS fini routines can not properly work in a panic-ed system. 6727 */ 6728 if (panicstr == NULL) 6729 (void)zfs__fini(); 6730} 6731 6732 6733static int 6734zfs_modevent(module_t mod, int type, void *unused __unused) 6735{ 6736 int err; 6737 6738 switch (type) { 6739 case MOD_LOAD: 6740 err = zfs__init(); 6741 if (err == 0) 6742 zfs_shutdown_event_tag = EVENTHANDLER_REGISTER( 6743 shutdown_post_sync, zfs_shutdown, NULL, 6744 SHUTDOWN_PRI_FIRST); 6745 return (err); 6746 case MOD_UNLOAD: 6747 err = zfs__fini(); 6748 if (err == 0 && zfs_shutdown_event_tag != NULL) 6749 EVENTHANDLER_DEREGISTER(shutdown_post_sync, 6750 zfs_shutdown_event_tag); 6751 return (err); 6752 case MOD_SHUTDOWN: 6753 return (0); 6754 default: 6755 break; 6756 } 6757 return (EOPNOTSUPP); 6758} 6759 6760static moduledata_t zfs_mod = { 6761 "zfsctrl", 6762 zfs_modevent, 6763 0 6764}; 6765DECLARE_MODULE(zfsctrl, zfs_mod, SI_SUB_VFS, SI_ORDER_ANY); 6766MODULE_VERSION(zfsctrl, 1); 6767MODULE_DEPEND(zfsctrl, opensolaris, 1, 1, 1); 6768MODULE_DEPEND(zfsctrl, krpc, 1, 1, 1); 6769MODULE_DEPEND(zfsctrl, acl_nfs4, 1, 1, 1); 6770