zfs_ioctl.c revision 307053
1/* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 22/* 23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 24 * Copyright (c) 2011-2012 Pawel Jakub Dawidek. All rights reserved. 25 * Copyright 2013 Martin Matuska <mm@FreeBSD.org>. All rights reserved. 26 * Copyright 2014 Xin Li <delphij@FreeBSD.org>. All rights reserved. 27 * Copyright 2015, OmniTI Computer Consulting, Inc. All rights reserved. 28 * Copyright 2015 Nexenta Systems, Inc. All rights reserved. 29 * Copyright (c) 2014, Joyent, Inc. All rights reserved. 30 * Copyright (c) 2011, 2015 by Delphix. All rights reserved. 31 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved. 32 * Copyright (c) 2013 Steven Hartland. All rights reserved. 33 * Copyright (c) 2014 Integros [integros.com] 34 */ 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 vfs_t *vfsp; 1432 int error; 1433 1434 error = dmu_objset_hold(dsname, FTAG, &os); 1435 if (error != 0) 1436 return (error); 1437 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1438 dmu_objset_rele(os, FTAG); 1439 return (SET_ERROR(EINVAL)); 1440 } 1441 1442 mutex_enter(&os->os_user_ptr_lock); 1443 *zfvp = dmu_objset_get_user(os); 1444 if (*zfvp) { 1445 vfsp = (*zfvp)->z_vfs; 1446 vfs_ref(vfsp); 1447 } else { 1448 error = SET_ERROR(ESRCH); 1449 } 1450 mutex_exit(&os->os_user_ptr_lock); 1451 dmu_objset_rele(os, FTAG); 1452 if (error == 0) { 1453 error = vfs_busy(vfsp, 0); 1454 vfs_rel(vfsp); 1455 if (error != 0) { 1456 *zfvp = NULL; 1457 error = SET_ERROR(ESRCH); 1458 } 1459 } 1460 return (error); 1461} 1462 1463/* 1464 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which 1465 * case its z_vfs will be NULL, and it will be opened as the owner. 1466 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER, 1467 * which prevents all vnode ops from running. 1468 */ 1469static int 1470zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer) 1471{ 1472 int error = 0; 1473 1474 if (getzfsvfs(name, zfvp) != 0) 1475 error = zfsvfs_create(name, zfvp); 1476 if (error == 0) { 1477 rrm_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER : 1478 RW_READER, tag); 1479 if ((*zfvp)->z_unmounted) { 1480 /* 1481 * XXX we could probably try again, since the unmounting 1482 * thread should be just about to disassociate the 1483 * objset from the zfsvfs. 1484 */ 1485 rrm_exit(&(*zfvp)->z_teardown_lock, tag); 1486 return (SET_ERROR(EBUSY)); 1487 } 1488 } 1489 return (error); 1490} 1491 1492static void 1493zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag) 1494{ 1495 rrm_exit(&zfsvfs->z_teardown_lock, tag); 1496 1497 if (zfsvfs->z_vfs) { 1498#ifdef illumos 1499 VFS_RELE(zfsvfs->z_vfs); 1500#else 1501 vfs_unbusy(zfsvfs->z_vfs); 1502#endif 1503 } else { 1504 dmu_objset_disown(zfsvfs->z_os, zfsvfs); 1505 zfsvfs_free(zfsvfs); 1506 } 1507} 1508 1509static int 1510zfs_ioc_pool_create(zfs_cmd_t *zc) 1511{ 1512 int error; 1513 nvlist_t *config, *props = NULL; 1514 nvlist_t *rootprops = NULL; 1515 nvlist_t *zplprops = NULL; 1516 1517 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1518 zc->zc_iflags, &config)) 1519 return (error); 1520 1521 if (zc->zc_nvlist_src_size != 0 && (error = 1522 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1523 zc->zc_iflags, &props))) { 1524 nvlist_free(config); 1525 return (error); 1526 } 1527 1528 if (props) { 1529 nvlist_t *nvl = NULL; 1530 uint64_t version = SPA_VERSION; 1531 1532 (void) nvlist_lookup_uint64(props, 1533 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version); 1534 if (!SPA_VERSION_IS_SUPPORTED(version)) { 1535 error = SET_ERROR(EINVAL); 1536 goto pool_props_bad; 1537 } 1538 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl); 1539 if (nvl) { 1540 error = nvlist_dup(nvl, &rootprops, KM_SLEEP); 1541 if (error != 0) { 1542 nvlist_free(config); 1543 nvlist_free(props); 1544 return (error); 1545 } 1546 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS); 1547 } 1548 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 1549 error = zfs_fill_zplprops_root(version, rootprops, 1550 zplprops, NULL); 1551 if (error != 0) 1552 goto pool_props_bad; 1553 } 1554 1555 error = spa_create(zc->zc_name, config, props, zplprops); 1556 1557 /* 1558 * Set the remaining root properties 1559 */ 1560 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name, 1561 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0) 1562 (void) spa_destroy(zc->zc_name); 1563 1564pool_props_bad: 1565 nvlist_free(rootprops); 1566 nvlist_free(zplprops); 1567 nvlist_free(config); 1568 nvlist_free(props); 1569 1570 return (error); 1571} 1572 1573static int 1574zfs_ioc_pool_destroy(zfs_cmd_t *zc) 1575{ 1576 int error; 1577 zfs_log_history(zc); 1578 error = spa_destroy(zc->zc_name); 1579 if (error == 0) 1580 zvol_remove_minors(zc->zc_name); 1581 return (error); 1582} 1583 1584static int 1585zfs_ioc_pool_import(zfs_cmd_t *zc) 1586{ 1587 nvlist_t *config, *props = NULL; 1588 uint64_t guid; 1589 int error; 1590 1591 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1592 zc->zc_iflags, &config)) != 0) 1593 return (error); 1594 1595 if (zc->zc_nvlist_src_size != 0 && (error = 1596 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1597 zc->zc_iflags, &props))) { 1598 nvlist_free(config); 1599 return (error); 1600 } 1601 1602 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 || 1603 guid != zc->zc_guid) 1604 error = SET_ERROR(EINVAL); 1605 else 1606 error = spa_import(zc->zc_name, config, props, zc->zc_cookie); 1607 1608 if (zc->zc_nvlist_dst != 0) { 1609 int err; 1610 1611 if ((err = put_nvlist(zc, config)) != 0) 1612 error = err; 1613 } 1614 1615 nvlist_free(config); 1616 1617 nvlist_free(props); 1618 1619 return (error); 1620} 1621 1622static int 1623zfs_ioc_pool_export(zfs_cmd_t *zc) 1624{ 1625 int error; 1626 boolean_t force = (boolean_t)zc->zc_cookie; 1627 boolean_t hardforce = (boolean_t)zc->zc_guid; 1628 1629 zfs_log_history(zc); 1630 error = spa_export(zc->zc_name, NULL, force, hardforce); 1631 if (error == 0) 1632 zvol_remove_minors(zc->zc_name); 1633 return (error); 1634} 1635 1636static int 1637zfs_ioc_pool_configs(zfs_cmd_t *zc) 1638{ 1639 nvlist_t *configs; 1640 int error; 1641 1642 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL) 1643 return (SET_ERROR(EEXIST)); 1644 1645 error = put_nvlist(zc, configs); 1646 1647 nvlist_free(configs); 1648 1649 return (error); 1650} 1651 1652/* 1653 * inputs: 1654 * zc_name name of the pool 1655 * 1656 * outputs: 1657 * zc_cookie real errno 1658 * zc_nvlist_dst config nvlist 1659 * zc_nvlist_dst_size size of config nvlist 1660 */ 1661static int 1662zfs_ioc_pool_stats(zfs_cmd_t *zc) 1663{ 1664 nvlist_t *config; 1665 int error; 1666 int ret = 0; 1667 1668 error = spa_get_stats(zc->zc_name, &config, zc->zc_value, 1669 sizeof (zc->zc_value)); 1670 1671 if (config != NULL) { 1672 ret = put_nvlist(zc, config); 1673 nvlist_free(config); 1674 1675 /* 1676 * The config may be present even if 'error' is non-zero. 1677 * In this case we return success, and preserve the real errno 1678 * in 'zc_cookie'. 1679 */ 1680 zc->zc_cookie = error; 1681 } else { 1682 ret = error; 1683 } 1684 1685 return (ret); 1686} 1687 1688/* 1689 * Try to import the given pool, returning pool stats as appropriate so that 1690 * user land knows which devices are available and overall pool health. 1691 */ 1692static int 1693zfs_ioc_pool_tryimport(zfs_cmd_t *zc) 1694{ 1695 nvlist_t *tryconfig, *config; 1696 int error; 1697 1698 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1699 zc->zc_iflags, &tryconfig)) != 0) 1700 return (error); 1701 1702 config = spa_tryimport(tryconfig); 1703 1704 nvlist_free(tryconfig); 1705 1706 if (config == NULL) 1707 return (SET_ERROR(EINVAL)); 1708 1709 error = put_nvlist(zc, config); 1710 nvlist_free(config); 1711 1712 return (error); 1713} 1714 1715/* 1716 * inputs: 1717 * zc_name name of the pool 1718 * zc_cookie scan func (pool_scan_func_t) 1719 */ 1720static int 1721zfs_ioc_pool_scan(zfs_cmd_t *zc) 1722{ 1723 spa_t *spa; 1724 int error; 1725 1726 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1727 return (error); 1728 1729 if (zc->zc_cookie == POOL_SCAN_NONE) 1730 error = spa_scan_stop(spa); 1731 else 1732 error = spa_scan(spa, zc->zc_cookie); 1733 1734 spa_close(spa, FTAG); 1735 1736 return (error); 1737} 1738 1739static int 1740zfs_ioc_pool_freeze(zfs_cmd_t *zc) 1741{ 1742 spa_t *spa; 1743 int error; 1744 1745 error = spa_open(zc->zc_name, &spa, FTAG); 1746 if (error == 0) { 1747 spa_freeze(spa); 1748 spa_close(spa, FTAG); 1749 } 1750 return (error); 1751} 1752 1753static int 1754zfs_ioc_pool_upgrade(zfs_cmd_t *zc) 1755{ 1756 spa_t *spa; 1757 int error; 1758 1759 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1760 return (error); 1761 1762 if (zc->zc_cookie < spa_version(spa) || 1763 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) { 1764 spa_close(spa, FTAG); 1765 return (SET_ERROR(EINVAL)); 1766 } 1767 1768 spa_upgrade(spa, zc->zc_cookie); 1769 spa_close(spa, FTAG); 1770 1771 return (error); 1772} 1773 1774static int 1775zfs_ioc_pool_get_history(zfs_cmd_t *zc) 1776{ 1777 spa_t *spa; 1778 char *hist_buf; 1779 uint64_t size; 1780 int error; 1781 1782 if ((size = zc->zc_history_len) == 0) 1783 return (SET_ERROR(EINVAL)); 1784 1785 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1786 return (error); 1787 1788 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) { 1789 spa_close(spa, FTAG); 1790 return (SET_ERROR(ENOTSUP)); 1791 } 1792 1793 hist_buf = kmem_alloc(size, KM_SLEEP); 1794 if ((error = spa_history_get(spa, &zc->zc_history_offset, 1795 &zc->zc_history_len, hist_buf)) == 0) { 1796 error = ddi_copyout(hist_buf, 1797 (void *)(uintptr_t)zc->zc_history, 1798 zc->zc_history_len, zc->zc_iflags); 1799 } 1800 1801 spa_close(spa, FTAG); 1802 kmem_free(hist_buf, size); 1803 return (error); 1804} 1805 1806static int 1807zfs_ioc_pool_reguid(zfs_cmd_t *zc) 1808{ 1809 spa_t *spa; 1810 int error; 1811 1812 error = spa_open(zc->zc_name, &spa, FTAG); 1813 if (error == 0) { 1814 error = spa_change_guid(spa); 1815 spa_close(spa, FTAG); 1816 } 1817 return (error); 1818} 1819 1820static int 1821zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc) 1822{ 1823 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value)); 1824} 1825 1826/* 1827 * inputs: 1828 * zc_name name of filesystem 1829 * zc_obj object to find 1830 * 1831 * outputs: 1832 * zc_value name of object 1833 */ 1834static int 1835zfs_ioc_obj_to_path(zfs_cmd_t *zc) 1836{ 1837 objset_t *os; 1838 int error; 1839 1840 /* XXX reading from objset not owned */ 1841 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0) 1842 return (error); 1843 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1844 dmu_objset_rele(os, FTAG); 1845 return (SET_ERROR(EINVAL)); 1846 } 1847 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value, 1848 sizeof (zc->zc_value)); 1849 dmu_objset_rele(os, FTAG); 1850 1851 return (error); 1852} 1853 1854/* 1855 * inputs: 1856 * zc_name name of filesystem 1857 * zc_obj object to find 1858 * 1859 * outputs: 1860 * zc_stat stats on object 1861 * zc_value path to object 1862 */ 1863static int 1864zfs_ioc_obj_to_stats(zfs_cmd_t *zc) 1865{ 1866 objset_t *os; 1867 int error; 1868 1869 /* XXX reading from objset not owned */ 1870 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0) 1871 return (error); 1872 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1873 dmu_objset_rele(os, FTAG); 1874 return (SET_ERROR(EINVAL)); 1875 } 1876 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value, 1877 sizeof (zc->zc_value)); 1878 dmu_objset_rele(os, FTAG); 1879 1880 return (error); 1881} 1882 1883static int 1884zfs_ioc_vdev_add(zfs_cmd_t *zc) 1885{ 1886 spa_t *spa; 1887 int error; 1888 nvlist_t *config, **l2cache, **spares; 1889 uint_t nl2cache = 0, nspares = 0; 1890 1891 error = spa_open(zc->zc_name, &spa, FTAG); 1892 if (error != 0) 1893 return (error); 1894 1895 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1896 zc->zc_iflags, &config); 1897 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE, 1898 &l2cache, &nl2cache); 1899 1900 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES, 1901 &spares, &nspares); 1902 1903#ifdef illumos 1904 /* 1905 * A root pool with concatenated devices is not supported. 1906 * Thus, can not add a device to a root pool. 1907 * 1908 * Intent log device can not be added to a rootpool because 1909 * during mountroot, zil is replayed, a seperated log device 1910 * can not be accessed during the mountroot time. 1911 * 1912 * l2cache and spare devices are ok to be added to a rootpool. 1913 */ 1914 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) { 1915 nvlist_free(config); 1916 spa_close(spa, FTAG); 1917 return (SET_ERROR(EDOM)); 1918 } 1919#endif /* illumos */ 1920 1921 if (error == 0) { 1922 error = spa_vdev_add(spa, config); 1923 nvlist_free(config); 1924 } 1925 spa_close(spa, FTAG); 1926 return (error); 1927} 1928 1929/* 1930 * inputs: 1931 * zc_name name of the pool 1932 * zc_nvlist_conf nvlist of devices to remove 1933 * zc_cookie to stop the remove? 1934 */ 1935static int 1936zfs_ioc_vdev_remove(zfs_cmd_t *zc) 1937{ 1938 spa_t *spa; 1939 int error; 1940 1941 error = spa_open(zc->zc_name, &spa, FTAG); 1942 if (error != 0) 1943 return (error); 1944 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE); 1945 spa_close(spa, FTAG); 1946 return (error); 1947} 1948 1949static int 1950zfs_ioc_vdev_set_state(zfs_cmd_t *zc) 1951{ 1952 spa_t *spa; 1953 int error; 1954 vdev_state_t newstate = VDEV_STATE_UNKNOWN; 1955 1956 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1957 return (error); 1958 switch (zc->zc_cookie) { 1959 case VDEV_STATE_ONLINE: 1960 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate); 1961 break; 1962 1963 case VDEV_STATE_OFFLINE: 1964 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj); 1965 break; 1966 1967 case VDEV_STATE_FAULTED: 1968 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED && 1969 zc->zc_obj != VDEV_AUX_EXTERNAL) 1970 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED; 1971 1972 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj); 1973 break; 1974 1975 case VDEV_STATE_DEGRADED: 1976 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED && 1977 zc->zc_obj != VDEV_AUX_EXTERNAL) 1978 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED; 1979 1980 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj); 1981 break; 1982 1983 default: 1984 error = SET_ERROR(EINVAL); 1985 } 1986 zc->zc_cookie = newstate; 1987 spa_close(spa, FTAG); 1988 return (error); 1989} 1990 1991static int 1992zfs_ioc_vdev_attach(zfs_cmd_t *zc) 1993{ 1994 spa_t *spa; 1995 int replacing = zc->zc_cookie; 1996 nvlist_t *config; 1997 int error; 1998 1999 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 2000 return (error); 2001 2002 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 2003 zc->zc_iflags, &config)) == 0) { 2004 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing); 2005 nvlist_free(config); 2006 } 2007 2008 spa_close(spa, FTAG); 2009 return (error); 2010} 2011 2012static int 2013zfs_ioc_vdev_detach(zfs_cmd_t *zc) 2014{ 2015 spa_t *spa; 2016 int error; 2017 2018 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 2019 return (error); 2020 2021 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE); 2022 2023 spa_close(spa, FTAG); 2024 return (error); 2025} 2026 2027static int 2028zfs_ioc_vdev_split(zfs_cmd_t *zc) 2029{ 2030 spa_t *spa; 2031 nvlist_t *config, *props = NULL; 2032 int error; 2033 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT); 2034 2035 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 2036 return (error); 2037 2038 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 2039 zc->zc_iflags, &config)) { 2040 spa_close(spa, FTAG); 2041 return (error); 2042 } 2043 2044 if (zc->zc_nvlist_src_size != 0 && (error = 2045 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2046 zc->zc_iflags, &props))) { 2047 spa_close(spa, FTAG); 2048 nvlist_free(config); 2049 return (error); 2050 } 2051 2052 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp); 2053 2054 spa_close(spa, FTAG); 2055 2056 nvlist_free(config); 2057 nvlist_free(props); 2058 2059 return (error); 2060} 2061 2062static int 2063zfs_ioc_vdev_setpath(zfs_cmd_t *zc) 2064{ 2065 spa_t *spa; 2066 char *path = zc->zc_value; 2067 uint64_t guid = zc->zc_guid; 2068 int error; 2069 2070 error = spa_open(zc->zc_name, &spa, FTAG); 2071 if (error != 0) 2072 return (error); 2073 2074 error = spa_vdev_setpath(spa, guid, path); 2075 spa_close(spa, FTAG); 2076 return (error); 2077} 2078 2079static int 2080zfs_ioc_vdev_setfru(zfs_cmd_t *zc) 2081{ 2082 spa_t *spa; 2083 char *fru = zc->zc_value; 2084 uint64_t guid = zc->zc_guid; 2085 int error; 2086 2087 error = spa_open(zc->zc_name, &spa, FTAG); 2088 if (error != 0) 2089 return (error); 2090 2091 error = spa_vdev_setfru(spa, guid, fru); 2092 spa_close(spa, FTAG); 2093 return (error); 2094} 2095 2096static int 2097zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os) 2098{ 2099 int error = 0; 2100 nvlist_t *nv; 2101 2102 dmu_objset_fast_stat(os, &zc->zc_objset_stats); 2103 2104 if (zc->zc_nvlist_dst != 0 && 2105 (error = dsl_prop_get_all(os, &nv)) == 0) { 2106 dmu_objset_stats(os, nv); 2107 /* 2108 * NB: zvol_get_stats() will read the objset contents, 2109 * which we aren't supposed to do with a 2110 * DS_MODE_USER hold, because it could be 2111 * inconsistent. So this is a bit of a workaround... 2112 * XXX reading with out owning 2113 */ 2114 if (!zc->zc_objset_stats.dds_inconsistent && 2115 dmu_objset_type(os) == DMU_OST_ZVOL) { 2116 error = zvol_get_stats(os, nv); 2117 if (error == EIO) 2118 return (error); 2119 VERIFY0(error); 2120 } 2121 error = put_nvlist(zc, nv); 2122 nvlist_free(nv); 2123 } 2124 2125 return (error); 2126} 2127 2128/* 2129 * inputs: 2130 * zc_name name of filesystem 2131 * zc_nvlist_dst_size size of buffer for property nvlist 2132 * 2133 * outputs: 2134 * zc_objset_stats stats 2135 * zc_nvlist_dst property nvlist 2136 * zc_nvlist_dst_size size of property nvlist 2137 */ 2138static int 2139zfs_ioc_objset_stats(zfs_cmd_t *zc) 2140{ 2141 objset_t *os; 2142 int error; 2143 2144 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 2145 if (error == 0) { 2146 error = zfs_ioc_objset_stats_impl(zc, os); 2147 dmu_objset_rele(os, FTAG); 2148 } 2149 2150 if (error == ENOMEM) 2151 error = 0; 2152 return (error); 2153} 2154 2155/* 2156 * inputs: 2157 * zc_name name of filesystem 2158 * zc_nvlist_dst_size size of buffer for property nvlist 2159 * 2160 * outputs: 2161 * zc_nvlist_dst received property nvlist 2162 * zc_nvlist_dst_size size of received property nvlist 2163 * 2164 * Gets received properties (distinct from local properties on or after 2165 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from 2166 * local property values. 2167 */ 2168static int 2169zfs_ioc_objset_recvd_props(zfs_cmd_t *zc) 2170{ 2171 int error = 0; 2172 nvlist_t *nv; 2173 2174 /* 2175 * Without this check, we would return local property values if the 2176 * caller has not already received properties on or after 2177 * SPA_VERSION_RECVD_PROPS. 2178 */ 2179 if (!dsl_prop_get_hasrecvd(zc->zc_name)) 2180 return (SET_ERROR(ENOTSUP)); 2181 2182 if (zc->zc_nvlist_dst != 0 && 2183 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) { 2184 error = put_nvlist(zc, nv); 2185 nvlist_free(nv); 2186 } 2187 2188 return (error); 2189} 2190 2191static int 2192nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop) 2193{ 2194 uint64_t value; 2195 int error; 2196 2197 /* 2198 * zfs_get_zplprop() will either find a value or give us 2199 * the default value (if there is one). 2200 */ 2201 if ((error = zfs_get_zplprop(os, prop, &value)) != 0) 2202 return (error); 2203 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0); 2204 return (0); 2205} 2206 2207/* 2208 * inputs: 2209 * zc_name name of filesystem 2210 * zc_nvlist_dst_size size of buffer for zpl property nvlist 2211 * 2212 * outputs: 2213 * zc_nvlist_dst zpl property nvlist 2214 * zc_nvlist_dst_size size of zpl property nvlist 2215 */ 2216static int 2217zfs_ioc_objset_zplprops(zfs_cmd_t *zc) 2218{ 2219 objset_t *os; 2220 int err; 2221 2222 /* XXX reading without owning */ 2223 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os)) 2224 return (err); 2225 2226 dmu_objset_fast_stat(os, &zc->zc_objset_stats); 2227 2228 /* 2229 * NB: nvl_add_zplprop() will read the objset contents, 2230 * which we aren't supposed to do with a DS_MODE_USER 2231 * hold, because it could be inconsistent. 2232 */ 2233 if (zc->zc_nvlist_dst != 0 && 2234 !zc->zc_objset_stats.dds_inconsistent && 2235 dmu_objset_type(os) == DMU_OST_ZFS) { 2236 nvlist_t *nv; 2237 2238 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2239 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 && 2240 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 && 2241 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 && 2242 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0) 2243 err = put_nvlist(zc, nv); 2244 nvlist_free(nv); 2245 } else { 2246 err = SET_ERROR(ENOENT); 2247 } 2248 dmu_objset_rele(os, FTAG); 2249 return (err); 2250} 2251 2252boolean_t 2253dataset_name_hidden(const char *name) 2254{ 2255 /* 2256 * Skip over datasets that are not visible in this zone, 2257 * internal datasets (which have a $ in their name), and 2258 * temporary datasets (which have a % in their name). 2259 */ 2260 if (strchr(name, '$') != NULL) 2261 return (B_TRUE); 2262 if (strchr(name, '%') != NULL) 2263 return (B_TRUE); 2264 if (!INGLOBALZONE(curthread) && !zone_dataset_visible(name, NULL)) 2265 return (B_TRUE); 2266 return (B_FALSE); 2267} 2268 2269/* 2270 * inputs: 2271 * zc_name name of filesystem 2272 * zc_cookie zap cursor 2273 * zc_nvlist_dst_size size of buffer for property nvlist 2274 * 2275 * outputs: 2276 * zc_name name of next filesystem 2277 * zc_cookie zap cursor 2278 * zc_objset_stats stats 2279 * zc_nvlist_dst property nvlist 2280 * zc_nvlist_dst_size size of property nvlist 2281 */ 2282static int 2283zfs_ioc_dataset_list_next(zfs_cmd_t *zc) 2284{ 2285 objset_t *os; 2286 int error; 2287 char *p; 2288 size_t orig_len = strlen(zc->zc_name); 2289 2290top: 2291 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) { 2292 if (error == ENOENT) 2293 error = SET_ERROR(ESRCH); 2294 return (error); 2295 } 2296 2297 p = strrchr(zc->zc_name, '/'); 2298 if (p == NULL || p[1] != '\0') 2299 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name)); 2300 p = zc->zc_name + strlen(zc->zc_name); 2301 2302 do { 2303 error = dmu_dir_list_next(os, 2304 sizeof (zc->zc_name) - (p - zc->zc_name), p, 2305 NULL, &zc->zc_cookie); 2306 if (error == ENOENT) 2307 error = SET_ERROR(ESRCH); 2308 } while (error == 0 && dataset_name_hidden(zc->zc_name)); 2309 dmu_objset_rele(os, FTAG); 2310 2311 /* 2312 * If it's an internal dataset (ie. with a '$' in its name), 2313 * don't try to get stats for it, otherwise we'll return ENOENT. 2314 */ 2315 if (error == 0 && strchr(zc->zc_name, '$') == NULL) { 2316 error = zfs_ioc_objset_stats(zc); /* fill in the stats */ 2317 if (error == ENOENT) { 2318 /* We lost a race with destroy, get the next one. */ 2319 zc->zc_name[orig_len] = '\0'; 2320 goto top; 2321 } 2322 } 2323 return (error); 2324} 2325 2326/* 2327 * inputs: 2328 * zc_name name of filesystem 2329 * zc_cookie zap cursor 2330 * zc_nvlist_dst_size size of buffer for property nvlist 2331 * zc_simple when set, only name is requested 2332 * 2333 * outputs: 2334 * zc_name name of next snapshot 2335 * zc_objset_stats stats 2336 * zc_nvlist_dst property nvlist 2337 * zc_nvlist_dst_size size of property nvlist 2338 */ 2339static int 2340zfs_ioc_snapshot_list_next(zfs_cmd_t *zc) 2341{ 2342 objset_t *os; 2343 int error; 2344 2345 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 2346 if (error != 0) { 2347 return (error == ENOENT ? ESRCH : error); 2348 } 2349 2350 /* 2351 * A dataset name of maximum length cannot have any snapshots, 2352 * so exit immediately. 2353 */ 2354 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) { 2355 dmu_objset_rele(os, FTAG); 2356 return (SET_ERROR(ESRCH)); 2357 } 2358 2359 error = dmu_snapshot_list_next(os, 2360 sizeof (zc->zc_name) - strlen(zc->zc_name), 2361 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie, 2362 NULL); 2363 2364 if (error == 0 && !zc->zc_simple) { 2365 dsl_dataset_t *ds; 2366 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool; 2367 2368 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds); 2369 if (error == 0) { 2370 objset_t *ossnap; 2371 2372 error = dmu_objset_from_ds(ds, &ossnap); 2373 if (error == 0) 2374 error = zfs_ioc_objset_stats_impl(zc, ossnap); 2375 dsl_dataset_rele(ds, FTAG); 2376 } 2377 } else if (error == ENOENT) { 2378 error = SET_ERROR(ESRCH); 2379 } 2380 2381 dmu_objset_rele(os, FTAG); 2382 /* if we failed, undo the @ that we tacked on to zc_name */ 2383 if (error != 0) 2384 *strchr(zc->zc_name, '@') = '\0'; 2385 return (error); 2386} 2387 2388static int 2389zfs_prop_set_userquota(const char *dsname, nvpair_t *pair) 2390{ 2391 const char *propname = nvpair_name(pair); 2392 uint64_t *valary; 2393 unsigned int vallen; 2394 const char *domain; 2395 char *dash; 2396 zfs_userquota_prop_t type; 2397 uint64_t rid; 2398 uint64_t quota; 2399 zfsvfs_t *zfsvfs; 2400 int err; 2401 2402 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2403 nvlist_t *attrs; 2404 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 2405 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2406 &pair) != 0) 2407 return (SET_ERROR(EINVAL)); 2408 } 2409 2410 /* 2411 * A correctly constructed propname is encoded as 2412 * userquota@<rid>-<domain>. 2413 */ 2414 if ((dash = strchr(propname, '-')) == NULL || 2415 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 || 2416 vallen != 3) 2417 return (SET_ERROR(EINVAL)); 2418 2419 domain = dash + 1; 2420 type = valary[0]; 2421 rid = valary[1]; 2422 quota = valary[2]; 2423 2424 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE); 2425 if (err == 0) { 2426 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota); 2427 zfsvfs_rele(zfsvfs, FTAG); 2428 } 2429 2430 return (err); 2431} 2432 2433/* 2434 * If the named property is one that has a special function to set its value, 2435 * return 0 on success and a positive error code on failure; otherwise if it is 2436 * not one of the special properties handled by this function, return -1. 2437 * 2438 * XXX: It would be better for callers of the property interface if we handled 2439 * these special cases in dsl_prop.c (in the dsl layer). 2440 */ 2441static int 2442zfs_prop_set_special(const char *dsname, zprop_source_t source, 2443 nvpair_t *pair) 2444{ 2445 const char *propname = nvpair_name(pair); 2446 zfs_prop_t prop = zfs_name_to_prop(propname); 2447 uint64_t intval; 2448 int err = -1; 2449 2450 if (prop == ZPROP_INVAL) { 2451 if (zfs_prop_userquota(propname)) 2452 return (zfs_prop_set_userquota(dsname, pair)); 2453 return (-1); 2454 } 2455 2456 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2457 nvlist_t *attrs; 2458 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 2459 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2460 &pair) == 0); 2461 } 2462 2463 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING) 2464 return (-1); 2465 2466 VERIFY(0 == nvpair_value_uint64(pair, &intval)); 2467 2468 switch (prop) { 2469 case ZFS_PROP_QUOTA: 2470 err = dsl_dir_set_quota(dsname, source, intval); 2471 break; 2472 case ZFS_PROP_REFQUOTA: 2473 err = dsl_dataset_set_refquota(dsname, source, intval); 2474 break; 2475 case ZFS_PROP_FILESYSTEM_LIMIT: 2476 case ZFS_PROP_SNAPSHOT_LIMIT: 2477 if (intval == UINT64_MAX) { 2478 /* clearing the limit, just do it */ 2479 err = 0; 2480 } else { 2481 err = dsl_dir_activate_fs_ss_limit(dsname); 2482 } 2483 /* 2484 * Set err to -1 to force the zfs_set_prop_nvlist code down the 2485 * default path to set the value in the nvlist. 2486 */ 2487 if (err == 0) 2488 err = -1; 2489 break; 2490 case ZFS_PROP_RESERVATION: 2491 err = dsl_dir_set_reservation(dsname, source, intval); 2492 break; 2493 case ZFS_PROP_REFRESERVATION: 2494 err = dsl_dataset_set_refreservation(dsname, source, intval); 2495 break; 2496 case ZFS_PROP_VOLSIZE: 2497 err = zvol_set_volsize(dsname, intval); 2498 break; 2499 case ZFS_PROP_VERSION: 2500 { 2501 zfsvfs_t *zfsvfs; 2502 2503 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0) 2504 break; 2505 2506 err = zfs_set_version(zfsvfs, intval); 2507 zfsvfs_rele(zfsvfs, FTAG); 2508 2509 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) { 2510 zfs_cmd_t *zc; 2511 2512 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP); 2513 (void) strcpy(zc->zc_name, dsname); 2514 (void) zfs_ioc_userspace_upgrade(zc); 2515 kmem_free(zc, sizeof (zfs_cmd_t)); 2516 } 2517 break; 2518 } 2519 default: 2520 err = -1; 2521 } 2522 2523 return (err); 2524} 2525 2526/* 2527 * This function is best effort. If it fails to set any of the given properties, 2528 * it continues to set as many as it can and returns the last error 2529 * encountered. If the caller provides a non-NULL errlist, it will be filled in 2530 * with the list of names of all the properties that failed along with the 2531 * corresponding error numbers. 2532 * 2533 * If every property is set successfully, zero is returned and errlist is not 2534 * modified. 2535 */ 2536int 2537zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl, 2538 nvlist_t *errlist) 2539{ 2540 nvpair_t *pair; 2541 nvpair_t *propval; 2542 int rv = 0; 2543 uint64_t intval; 2544 char *strval; 2545 nvlist_t *genericnvl = fnvlist_alloc(); 2546 nvlist_t *retrynvl = fnvlist_alloc(); 2547 2548retry: 2549 pair = NULL; 2550 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { 2551 const char *propname = nvpair_name(pair); 2552 zfs_prop_t prop = zfs_name_to_prop(propname); 2553 int err = 0; 2554 2555 /* decode the property value */ 2556 propval = pair; 2557 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2558 nvlist_t *attrs; 2559 attrs = fnvpair_value_nvlist(pair); 2560 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2561 &propval) != 0) 2562 err = SET_ERROR(EINVAL); 2563 } 2564 2565 /* Validate value type */ 2566 if (err == 0 && prop == ZPROP_INVAL) { 2567 if (zfs_prop_user(propname)) { 2568 if (nvpair_type(propval) != DATA_TYPE_STRING) 2569 err = SET_ERROR(EINVAL); 2570 } else if (zfs_prop_userquota(propname)) { 2571 if (nvpair_type(propval) != 2572 DATA_TYPE_UINT64_ARRAY) 2573 err = SET_ERROR(EINVAL); 2574 } else { 2575 err = SET_ERROR(EINVAL); 2576 } 2577 } else if (err == 0) { 2578 if (nvpair_type(propval) == DATA_TYPE_STRING) { 2579 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING) 2580 err = SET_ERROR(EINVAL); 2581 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) { 2582 const char *unused; 2583 2584 intval = fnvpair_value_uint64(propval); 2585 2586 switch (zfs_prop_get_type(prop)) { 2587 case PROP_TYPE_NUMBER: 2588 break; 2589 case PROP_TYPE_STRING: 2590 err = SET_ERROR(EINVAL); 2591 break; 2592 case PROP_TYPE_INDEX: 2593 if (zfs_prop_index_to_string(prop, 2594 intval, &unused) != 0) 2595 err = SET_ERROR(EINVAL); 2596 break; 2597 default: 2598 cmn_err(CE_PANIC, 2599 "unknown property type"); 2600 } 2601 } else { 2602 err = SET_ERROR(EINVAL); 2603 } 2604 } 2605 2606 /* Validate permissions */ 2607 if (err == 0) 2608 err = zfs_check_settable(dsname, pair, CRED()); 2609 2610 if (err == 0) { 2611 err = zfs_prop_set_special(dsname, source, pair); 2612 if (err == -1) { 2613 /* 2614 * For better performance we build up a list of 2615 * properties to set in a single transaction. 2616 */ 2617 err = nvlist_add_nvpair(genericnvl, pair); 2618 } else if (err != 0 && nvl != retrynvl) { 2619 /* 2620 * This may be a spurious error caused by 2621 * receiving quota and reservation out of order. 2622 * Try again in a second pass. 2623 */ 2624 err = nvlist_add_nvpair(retrynvl, pair); 2625 } 2626 } 2627 2628 if (err != 0) { 2629 if (errlist != NULL) 2630 fnvlist_add_int32(errlist, propname, err); 2631 rv = err; 2632 } 2633 } 2634 2635 if (nvl != retrynvl && !nvlist_empty(retrynvl)) { 2636 nvl = retrynvl; 2637 goto retry; 2638 } 2639 2640 if (!nvlist_empty(genericnvl) && 2641 dsl_props_set(dsname, source, genericnvl) != 0) { 2642 /* 2643 * If this fails, we still want to set as many properties as we 2644 * can, so try setting them individually. 2645 */ 2646 pair = NULL; 2647 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) { 2648 const char *propname = nvpair_name(pair); 2649 int err = 0; 2650 2651 propval = pair; 2652 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2653 nvlist_t *attrs; 2654 attrs = fnvpair_value_nvlist(pair); 2655 propval = fnvlist_lookup_nvpair(attrs, 2656 ZPROP_VALUE); 2657 } 2658 2659 if (nvpair_type(propval) == DATA_TYPE_STRING) { 2660 strval = fnvpair_value_string(propval); 2661 err = dsl_prop_set_string(dsname, propname, 2662 source, strval); 2663 } else { 2664 intval = fnvpair_value_uint64(propval); 2665 err = dsl_prop_set_int(dsname, propname, source, 2666 intval); 2667 } 2668 2669 if (err != 0) { 2670 if (errlist != NULL) { 2671 fnvlist_add_int32(errlist, propname, 2672 err); 2673 } 2674 rv = err; 2675 } 2676 } 2677 } 2678 nvlist_free(genericnvl); 2679 nvlist_free(retrynvl); 2680 2681 return (rv); 2682} 2683 2684/* 2685 * Check that all the properties are valid user properties. 2686 */ 2687static int 2688zfs_check_userprops(const char *fsname, nvlist_t *nvl) 2689{ 2690 nvpair_t *pair = NULL; 2691 int error = 0; 2692 2693 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { 2694 const char *propname = nvpair_name(pair); 2695 2696 if (!zfs_prop_user(propname) || 2697 nvpair_type(pair) != DATA_TYPE_STRING) 2698 return (SET_ERROR(EINVAL)); 2699 2700 if (error = zfs_secpolicy_write_perms(fsname, 2701 ZFS_DELEG_PERM_USERPROP, CRED())) 2702 return (error); 2703 2704 if (strlen(propname) >= ZAP_MAXNAMELEN) 2705 return (SET_ERROR(ENAMETOOLONG)); 2706 2707 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN) 2708 return (E2BIG); 2709 } 2710 return (0); 2711} 2712 2713static void 2714props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops) 2715{ 2716 nvpair_t *pair; 2717 2718 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2719 2720 pair = NULL; 2721 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) { 2722 if (nvlist_exists(skipped, nvpair_name(pair))) 2723 continue; 2724 2725 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0); 2726 } 2727} 2728 2729static int 2730clear_received_props(const char *dsname, nvlist_t *props, 2731 nvlist_t *skipped) 2732{ 2733 int err = 0; 2734 nvlist_t *cleared_props = NULL; 2735 props_skip(props, skipped, &cleared_props); 2736 if (!nvlist_empty(cleared_props)) { 2737 /* 2738 * Acts on local properties until the dataset has received 2739 * properties at least once on or after SPA_VERSION_RECVD_PROPS. 2740 */ 2741 zprop_source_t flags = (ZPROP_SRC_NONE | 2742 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0)); 2743 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL); 2744 } 2745 nvlist_free(cleared_props); 2746 return (err); 2747} 2748 2749/* 2750 * inputs: 2751 * zc_name name of filesystem 2752 * zc_value name of property to set 2753 * zc_nvlist_src{_size} nvlist of properties to apply 2754 * zc_cookie received properties flag 2755 * 2756 * outputs: 2757 * zc_nvlist_dst{_size} error for each unapplied received property 2758 */ 2759static int 2760zfs_ioc_set_prop(zfs_cmd_t *zc) 2761{ 2762 nvlist_t *nvl; 2763 boolean_t received = zc->zc_cookie; 2764 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED : 2765 ZPROP_SRC_LOCAL); 2766 nvlist_t *errors; 2767 int error; 2768 2769 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2770 zc->zc_iflags, &nvl)) != 0) 2771 return (error); 2772 2773 if (received) { 2774 nvlist_t *origprops; 2775 2776 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) { 2777 (void) clear_received_props(zc->zc_name, 2778 origprops, nvl); 2779 nvlist_free(origprops); 2780 } 2781 2782 error = dsl_prop_set_hasrecvd(zc->zc_name); 2783 } 2784 2785 errors = fnvlist_alloc(); 2786 if (error == 0) 2787 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors); 2788 2789 if (zc->zc_nvlist_dst != 0 && errors != NULL) { 2790 (void) put_nvlist(zc, errors); 2791 } 2792 2793 nvlist_free(errors); 2794 nvlist_free(nvl); 2795 return (error); 2796} 2797 2798/* 2799 * inputs: 2800 * zc_name name of filesystem 2801 * zc_value name of property to inherit 2802 * zc_cookie revert to received value if TRUE 2803 * 2804 * outputs: none 2805 */ 2806static int 2807zfs_ioc_inherit_prop(zfs_cmd_t *zc) 2808{ 2809 const char *propname = zc->zc_value; 2810 zfs_prop_t prop = zfs_name_to_prop(propname); 2811 boolean_t received = zc->zc_cookie; 2812 zprop_source_t source = (received 2813 ? ZPROP_SRC_NONE /* revert to received value, if any */ 2814 : ZPROP_SRC_INHERITED); /* explicitly inherit */ 2815 2816 if (received) { 2817 nvlist_t *dummy; 2818 nvpair_t *pair; 2819 zprop_type_t type; 2820 int err; 2821 2822 /* 2823 * zfs_prop_set_special() expects properties in the form of an 2824 * nvpair with type info. 2825 */ 2826 if (prop == ZPROP_INVAL) { 2827 if (!zfs_prop_user(propname)) 2828 return (SET_ERROR(EINVAL)); 2829 2830 type = PROP_TYPE_STRING; 2831 } else if (prop == ZFS_PROP_VOLSIZE || 2832 prop == ZFS_PROP_VERSION) { 2833 return (SET_ERROR(EINVAL)); 2834 } else { 2835 type = zfs_prop_get_type(prop); 2836 } 2837 2838 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2839 2840 switch (type) { 2841 case PROP_TYPE_STRING: 2842 VERIFY(0 == nvlist_add_string(dummy, propname, "")); 2843 break; 2844 case PROP_TYPE_NUMBER: 2845 case PROP_TYPE_INDEX: 2846 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0)); 2847 break; 2848 default: 2849 nvlist_free(dummy); 2850 return (SET_ERROR(EINVAL)); 2851 } 2852 2853 pair = nvlist_next_nvpair(dummy, NULL); 2854 err = zfs_prop_set_special(zc->zc_name, source, pair); 2855 nvlist_free(dummy); 2856 if (err != -1) 2857 return (err); /* special property already handled */ 2858 } else { 2859 /* 2860 * Only check this in the non-received case. We want to allow 2861 * 'inherit -S' to revert non-inheritable properties like quota 2862 * and reservation to the received or default values even though 2863 * they are not considered inheritable. 2864 */ 2865 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop)) 2866 return (SET_ERROR(EINVAL)); 2867 } 2868 2869 /* property name has been validated by zfs_secpolicy_inherit_prop() */ 2870 return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source)); 2871} 2872 2873static int 2874zfs_ioc_pool_set_props(zfs_cmd_t *zc) 2875{ 2876 nvlist_t *props; 2877 spa_t *spa; 2878 int error; 2879 nvpair_t *pair; 2880 2881 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2882 zc->zc_iflags, &props)) 2883 return (error); 2884 2885 /* 2886 * If the only property is the configfile, then just do a spa_lookup() 2887 * to handle the faulted case. 2888 */ 2889 pair = nvlist_next_nvpair(props, NULL); 2890 if (pair != NULL && strcmp(nvpair_name(pair), 2891 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 && 2892 nvlist_next_nvpair(props, pair) == NULL) { 2893 mutex_enter(&spa_namespace_lock); 2894 if ((spa = spa_lookup(zc->zc_name)) != NULL) { 2895 spa_configfile_set(spa, props, B_FALSE); 2896 spa_config_sync(spa, B_FALSE, B_TRUE); 2897 } 2898 mutex_exit(&spa_namespace_lock); 2899 if (spa != NULL) { 2900 nvlist_free(props); 2901 return (0); 2902 } 2903 } 2904 2905 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { 2906 nvlist_free(props); 2907 return (error); 2908 } 2909 2910 error = spa_prop_set(spa, props); 2911 2912 nvlist_free(props); 2913 spa_close(spa, FTAG); 2914 2915 return (error); 2916} 2917 2918static int 2919zfs_ioc_pool_get_props(zfs_cmd_t *zc) 2920{ 2921 spa_t *spa; 2922 int error; 2923 nvlist_t *nvp = NULL; 2924 2925 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { 2926 /* 2927 * If the pool is faulted, there may be properties we can still 2928 * get (such as altroot and cachefile), so attempt to get them 2929 * anyway. 2930 */ 2931 mutex_enter(&spa_namespace_lock); 2932 if ((spa = spa_lookup(zc->zc_name)) != NULL) 2933 error = spa_prop_get(spa, &nvp); 2934 mutex_exit(&spa_namespace_lock); 2935 } else { 2936 error = spa_prop_get(spa, &nvp); 2937 spa_close(spa, FTAG); 2938 } 2939 2940 if (error == 0 && zc->zc_nvlist_dst != 0) 2941 error = put_nvlist(zc, nvp); 2942 else 2943 error = SET_ERROR(EFAULT); 2944 2945 nvlist_free(nvp); 2946 return (error); 2947} 2948 2949/* 2950 * inputs: 2951 * zc_name name of filesystem 2952 * zc_nvlist_src{_size} nvlist of delegated permissions 2953 * zc_perm_action allow/unallow flag 2954 * 2955 * outputs: none 2956 */ 2957static int 2958zfs_ioc_set_fsacl(zfs_cmd_t *zc) 2959{ 2960 int error; 2961 nvlist_t *fsaclnv = NULL; 2962 2963 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2964 zc->zc_iflags, &fsaclnv)) != 0) 2965 return (error); 2966 2967 /* 2968 * Verify nvlist is constructed correctly 2969 */ 2970 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) { 2971 nvlist_free(fsaclnv); 2972 return (SET_ERROR(EINVAL)); 2973 } 2974 2975 /* 2976 * If we don't have PRIV_SYS_MOUNT, then validate 2977 * that user is allowed to hand out each permission in 2978 * the nvlist(s) 2979 */ 2980 2981 error = secpolicy_zfs(CRED()); 2982 if (error != 0) { 2983 if (zc->zc_perm_action == B_FALSE) { 2984 error = dsl_deleg_can_allow(zc->zc_name, 2985 fsaclnv, CRED()); 2986 } else { 2987 error = dsl_deleg_can_unallow(zc->zc_name, 2988 fsaclnv, CRED()); 2989 } 2990 } 2991 2992 if (error == 0) 2993 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action); 2994 2995 nvlist_free(fsaclnv); 2996 return (error); 2997} 2998 2999/* 3000 * inputs: 3001 * zc_name name of filesystem 3002 * 3003 * outputs: 3004 * zc_nvlist_src{_size} nvlist of delegated permissions 3005 */ 3006static int 3007zfs_ioc_get_fsacl(zfs_cmd_t *zc) 3008{ 3009 nvlist_t *nvp; 3010 int error; 3011 3012 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) { 3013 error = put_nvlist(zc, nvp); 3014 nvlist_free(nvp); 3015 } 3016 3017 return (error); 3018} 3019 3020/* 3021 * Search the vfs list for a specified resource. Returns a pointer to it 3022 * or NULL if no suitable entry is found. The caller of this routine 3023 * is responsible for releasing the returned vfs pointer. 3024 */ 3025static vfs_t * 3026zfs_get_vfs(const char *resource) 3027{ 3028 vfs_t *vfsp; 3029 3030 mtx_lock(&mountlist_mtx); 3031 TAILQ_FOREACH(vfsp, &mountlist, mnt_list) { 3032 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) { 3033 if (vfs_busy(vfsp, MBF_MNTLSTLOCK) != 0) 3034 vfsp = NULL; 3035 break; 3036 } 3037 } 3038 if (vfsp == NULL) 3039 mtx_unlock(&mountlist_mtx); 3040 return (vfsp); 3041} 3042 3043/* ARGSUSED */ 3044static void 3045zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) 3046{ 3047 zfs_creat_t *zct = arg; 3048 3049 zfs_create_fs(os, cr, zct->zct_zplprops, tx); 3050} 3051 3052#define ZFS_PROP_UNDEFINED ((uint64_t)-1) 3053 3054/* 3055 * inputs: 3056 * os parent objset pointer (NULL if root fs) 3057 * fuids_ok fuids allowed in this version of the spa? 3058 * sa_ok SAs allowed in this version of the spa? 3059 * createprops list of properties requested by creator 3060 * 3061 * outputs: 3062 * zplprops values for the zplprops we attach to the master node object 3063 * is_ci true if requested file system will be purely case-insensitive 3064 * 3065 * Determine the settings for utf8only, normalization and 3066 * casesensitivity. Specific values may have been requested by the 3067 * creator and/or we can inherit values from the parent dataset. If 3068 * the file system is of too early a vintage, a creator can not 3069 * request settings for these properties, even if the requested 3070 * setting is the default value. We don't actually want to create dsl 3071 * properties for these, so remove them from the source nvlist after 3072 * processing. 3073 */ 3074static int 3075zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver, 3076 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops, 3077 nvlist_t *zplprops, boolean_t *is_ci) 3078{ 3079 uint64_t sense = ZFS_PROP_UNDEFINED; 3080 uint64_t norm = ZFS_PROP_UNDEFINED; 3081 uint64_t u8 = ZFS_PROP_UNDEFINED; 3082 3083 ASSERT(zplprops != NULL); 3084 3085 /* 3086 * Pull out creator prop choices, if any. 3087 */ 3088 if (createprops) { 3089 (void) nvlist_lookup_uint64(createprops, 3090 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver); 3091 (void) nvlist_lookup_uint64(createprops, 3092 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm); 3093 (void) nvlist_remove_all(createprops, 3094 zfs_prop_to_name(ZFS_PROP_NORMALIZE)); 3095 (void) nvlist_lookup_uint64(createprops, 3096 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8); 3097 (void) nvlist_remove_all(createprops, 3098 zfs_prop_to_name(ZFS_PROP_UTF8ONLY)); 3099 (void) nvlist_lookup_uint64(createprops, 3100 zfs_prop_to_name(ZFS_PROP_CASE), &sense); 3101 (void) nvlist_remove_all(createprops, 3102 zfs_prop_to_name(ZFS_PROP_CASE)); 3103 } 3104 3105 /* 3106 * If the zpl version requested is whacky or the file system 3107 * or pool is version is too "young" to support normalization 3108 * and the creator tried to set a value for one of the props, 3109 * error out. 3110 */ 3111 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) || 3112 (zplver >= ZPL_VERSION_FUID && !fuids_ok) || 3113 (zplver >= ZPL_VERSION_SA && !sa_ok) || 3114 (zplver < ZPL_VERSION_NORMALIZATION && 3115 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED || 3116 sense != ZFS_PROP_UNDEFINED))) 3117 return (SET_ERROR(ENOTSUP)); 3118 3119 /* 3120 * Put the version in the zplprops 3121 */ 3122 VERIFY(nvlist_add_uint64(zplprops, 3123 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0); 3124 3125 if (norm == ZFS_PROP_UNDEFINED) 3126 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0); 3127 VERIFY(nvlist_add_uint64(zplprops, 3128 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0); 3129 3130 /* 3131 * If we're normalizing, names must always be valid UTF-8 strings. 3132 */ 3133 if (norm) 3134 u8 = 1; 3135 if (u8 == ZFS_PROP_UNDEFINED) 3136 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0); 3137 VERIFY(nvlist_add_uint64(zplprops, 3138 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0); 3139 3140 if (sense == ZFS_PROP_UNDEFINED) 3141 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0); 3142 VERIFY(nvlist_add_uint64(zplprops, 3143 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0); 3144 3145 if (is_ci) 3146 *is_ci = (sense == ZFS_CASE_INSENSITIVE); 3147 3148 return (0); 3149} 3150 3151static int 3152zfs_fill_zplprops(const char *dataset, nvlist_t *createprops, 3153 nvlist_t *zplprops, boolean_t *is_ci) 3154{ 3155 boolean_t fuids_ok, sa_ok; 3156 uint64_t zplver = ZPL_VERSION; 3157 objset_t *os = NULL; 3158 char parentname[MAXNAMELEN]; 3159 char *cp; 3160 spa_t *spa; 3161 uint64_t spa_vers; 3162 int error; 3163 3164 (void) strlcpy(parentname, dataset, sizeof (parentname)); 3165 cp = strrchr(parentname, '/'); 3166 ASSERT(cp != NULL); 3167 cp[0] = '\0'; 3168 3169 if ((error = spa_open(dataset, &spa, FTAG)) != 0) 3170 return (error); 3171 3172 spa_vers = spa_version(spa); 3173 spa_close(spa, FTAG); 3174 3175 zplver = zfs_zpl_version_map(spa_vers); 3176 fuids_ok = (zplver >= ZPL_VERSION_FUID); 3177 sa_ok = (zplver >= ZPL_VERSION_SA); 3178 3179 /* 3180 * Open parent object set so we can inherit zplprop values. 3181 */ 3182 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0) 3183 return (error); 3184 3185 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops, 3186 zplprops, is_ci); 3187 dmu_objset_rele(os, FTAG); 3188 return (error); 3189} 3190 3191static int 3192zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops, 3193 nvlist_t *zplprops, boolean_t *is_ci) 3194{ 3195 boolean_t fuids_ok; 3196 boolean_t sa_ok; 3197 uint64_t zplver = ZPL_VERSION; 3198 int error; 3199 3200 zplver = zfs_zpl_version_map(spa_vers); 3201 fuids_ok = (zplver >= ZPL_VERSION_FUID); 3202 sa_ok = (zplver >= ZPL_VERSION_SA); 3203 3204 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok, 3205 createprops, zplprops, is_ci); 3206 return (error); 3207} 3208 3209/* 3210 * innvl: { 3211 * "type" -> dmu_objset_type_t (int32) 3212 * (optional) "props" -> { prop -> value } 3213 * } 3214 * 3215 * outnvl: propname -> error code (int32) 3216 */ 3217static int 3218zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3219{ 3220 int error = 0; 3221 zfs_creat_t zct = { 0 }; 3222 nvlist_t *nvprops = NULL; 3223 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx); 3224 int32_t type32; 3225 dmu_objset_type_t type; 3226 boolean_t is_insensitive = B_FALSE; 3227 3228 if (nvlist_lookup_int32(innvl, "type", &type32) != 0) 3229 return (SET_ERROR(EINVAL)); 3230 type = type32; 3231 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops); 3232 3233 switch (type) { 3234 case DMU_OST_ZFS: 3235 cbfunc = zfs_create_cb; 3236 break; 3237 3238 case DMU_OST_ZVOL: 3239 cbfunc = zvol_create_cb; 3240 break; 3241 3242 default: 3243 cbfunc = NULL; 3244 break; 3245 } 3246 if (strchr(fsname, '@') || 3247 strchr(fsname, '%')) 3248 return (SET_ERROR(EINVAL)); 3249 3250 zct.zct_props = nvprops; 3251 3252 if (cbfunc == NULL) 3253 return (SET_ERROR(EINVAL)); 3254 3255 if (type == DMU_OST_ZVOL) { 3256 uint64_t volsize, volblocksize; 3257 3258 if (nvprops == NULL) 3259 return (SET_ERROR(EINVAL)); 3260 if (nvlist_lookup_uint64(nvprops, 3261 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0) 3262 return (SET_ERROR(EINVAL)); 3263 3264 if ((error = nvlist_lookup_uint64(nvprops, 3265 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), 3266 &volblocksize)) != 0 && error != ENOENT) 3267 return (SET_ERROR(EINVAL)); 3268 3269 if (error != 0) 3270 volblocksize = zfs_prop_default_numeric( 3271 ZFS_PROP_VOLBLOCKSIZE); 3272 3273 if ((error = zvol_check_volblocksize( 3274 volblocksize)) != 0 || 3275 (error = zvol_check_volsize(volsize, 3276 volblocksize)) != 0) 3277 return (error); 3278 } else if (type == DMU_OST_ZFS) { 3279 int error; 3280 3281 /* 3282 * We have to have normalization and 3283 * case-folding flags correct when we do the 3284 * file system creation, so go figure them out 3285 * now. 3286 */ 3287 VERIFY(nvlist_alloc(&zct.zct_zplprops, 3288 NV_UNIQUE_NAME, KM_SLEEP) == 0); 3289 error = zfs_fill_zplprops(fsname, nvprops, 3290 zct.zct_zplprops, &is_insensitive); 3291 if (error != 0) { 3292 nvlist_free(zct.zct_zplprops); 3293 return (error); 3294 } 3295 } 3296 3297 error = dmu_objset_create(fsname, type, 3298 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct); 3299 nvlist_free(zct.zct_zplprops); 3300 3301 /* 3302 * It would be nice to do this atomically. 3303 */ 3304 if (error == 0) { 3305 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL, 3306 nvprops, outnvl); 3307 if (error != 0) 3308 (void) dsl_destroy_head(fsname); 3309 } 3310#ifdef __FreeBSD__ 3311 if (error == 0 && type == DMU_OST_ZVOL) 3312 zvol_create_minors(fsname); 3313#endif 3314 return (error); 3315} 3316 3317/* 3318 * innvl: { 3319 * "origin" -> name of origin snapshot 3320 * (optional) "props" -> { prop -> value } 3321 * } 3322 * 3323 * outnvl: propname -> error code (int32) 3324 */ 3325static int 3326zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3327{ 3328 int error = 0; 3329 nvlist_t *nvprops = NULL; 3330 char *origin_name; 3331 3332 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0) 3333 return (SET_ERROR(EINVAL)); 3334 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops); 3335 3336 if (strchr(fsname, '@') || 3337 strchr(fsname, '%')) 3338 return (SET_ERROR(EINVAL)); 3339 3340 if (dataset_namecheck(origin_name, NULL, NULL) != 0) 3341 return (SET_ERROR(EINVAL)); 3342 error = dmu_objset_clone(fsname, origin_name); 3343 if (error != 0) 3344 return (error); 3345 3346 /* 3347 * It would be nice to do this atomically. 3348 */ 3349 if (error == 0) { 3350 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL, 3351 nvprops, outnvl); 3352 if (error != 0) 3353 (void) dsl_destroy_head(fsname); 3354 } 3355#ifdef __FreeBSD__ 3356 if (error == 0) 3357 zvol_create_minors(fsname); 3358#endif 3359 return (error); 3360} 3361 3362/* 3363 * innvl: { 3364 * "snaps" -> { snapshot1, snapshot2 } 3365 * (optional) "props" -> { prop -> value (string) } 3366 * } 3367 * 3368 * outnvl: snapshot -> error code (int32) 3369 */ 3370static int 3371zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3372{ 3373 nvlist_t *snaps; 3374 nvlist_t *props = NULL; 3375 int error, poollen; 3376 nvpair_t *pair; 3377 3378 (void) nvlist_lookup_nvlist(innvl, "props", &props); 3379 if ((error = zfs_check_userprops(poolname, props)) != 0) 3380 return (error); 3381 3382 if (!nvlist_empty(props) && 3383 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS)) 3384 return (SET_ERROR(ENOTSUP)); 3385 3386 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 3387 return (SET_ERROR(EINVAL)); 3388 poollen = strlen(poolname); 3389 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 3390 pair = nvlist_next_nvpair(snaps, pair)) { 3391 const char *name = nvpair_name(pair); 3392 const char *cp = strchr(name, '@'); 3393 3394 /* 3395 * The snap name must contain an @, and the part after it must 3396 * contain only valid characters. 3397 */ 3398 if (cp == NULL || 3399 zfs_component_namecheck(cp + 1, NULL, NULL) != 0) 3400 return (SET_ERROR(EINVAL)); 3401 3402 /* 3403 * The snap must be in the specified pool. 3404 */ 3405 if (strncmp(name, poolname, poollen) != 0 || 3406 (name[poollen] != '/' && name[poollen] != '@')) 3407 return (SET_ERROR(EXDEV)); 3408 3409 /* This must be the only snap of this fs. */ 3410 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair); 3411 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) { 3412 if (strncmp(name, nvpair_name(pair2), cp - name + 1) 3413 == 0) { 3414 return (SET_ERROR(EXDEV)); 3415 } 3416 } 3417 } 3418 3419 error = dsl_dataset_snapshot(snaps, props, outnvl); 3420 return (error); 3421} 3422 3423/* 3424 * innvl: "message" -> string 3425 */ 3426/* ARGSUSED */ 3427static int 3428zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl) 3429{ 3430 char *message; 3431 spa_t *spa; 3432 int error; 3433 char *poolname; 3434 3435 /* 3436 * The poolname in the ioctl is not set, we get it from the TSD, 3437 * which was set at the end of the last successful ioctl that allows 3438 * logging. The secpolicy func already checked that it is set. 3439 * Only one log ioctl is allowed after each successful ioctl, so 3440 * we clear the TSD here. 3441 */ 3442 poolname = tsd_get(zfs_allow_log_key); 3443 (void) tsd_set(zfs_allow_log_key, NULL); 3444 error = spa_open(poolname, &spa, FTAG); 3445 strfree(poolname); 3446 if (error != 0) 3447 return (error); 3448 3449 if (nvlist_lookup_string(innvl, "message", &message) != 0) { 3450 spa_close(spa, FTAG); 3451 return (SET_ERROR(EINVAL)); 3452 } 3453 3454 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) { 3455 spa_close(spa, FTAG); 3456 return (SET_ERROR(ENOTSUP)); 3457 } 3458 3459 error = spa_history_log(spa, message); 3460 spa_close(spa, FTAG); 3461 return (error); 3462} 3463 3464/* 3465 * The dp_config_rwlock must not be held when calling this, because the 3466 * unmount may need to write out data. 3467 * 3468 * This function is best-effort. Callers must deal gracefully if it 3469 * remains mounted (or is remounted after this call). 3470 * 3471 * Returns 0 if the argument is not a snapshot, or it is not currently a 3472 * filesystem, or we were able to unmount it. Returns error code otherwise. 3473 */ 3474int 3475zfs_unmount_snap(const char *snapname) 3476{ 3477 vfs_t *vfsp; 3478 zfsvfs_t *zfsvfs; 3479 int err; 3480 3481 if (strchr(snapname, '@') == NULL) 3482 return (0); 3483 3484 vfsp = zfs_get_vfs(snapname); 3485 if (vfsp == NULL) 3486 return (0); 3487 3488 zfsvfs = vfsp->vfs_data; 3489 ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os))); 3490 3491 err = vn_vfswlock(vfsp->vfs_vnodecovered); 3492#ifdef illumos 3493 VFS_RELE(vfsp); 3494#else 3495 vfs_unbusy(vfsp); 3496#endif 3497 if (err != 0) 3498 return (SET_ERROR(err)); 3499 3500 /* 3501 * Always force the unmount for snapshots. 3502 */ 3503 3504#ifdef illumos 3505 (void) dounmount(vfsp, MS_FORCE, kcred); 3506#else 3507 vfs_ref(vfsp); 3508 (void) dounmount(vfsp, MS_FORCE, curthread); 3509#endif 3510 return (0); 3511} 3512 3513/* ARGSUSED */ 3514static int 3515zfs_unmount_snap_cb(const char *snapname, void *arg) 3516{ 3517 return (zfs_unmount_snap(snapname)); 3518} 3519 3520/* 3521 * When a clone is destroyed, its origin may also need to be destroyed, 3522 * in which case it must be unmounted. This routine will do that unmount 3523 * if necessary. 3524 */ 3525void 3526zfs_destroy_unmount_origin(const char *fsname) 3527{ 3528 int error; 3529 objset_t *os; 3530 dsl_dataset_t *ds; 3531 3532 error = dmu_objset_hold(fsname, FTAG, &os); 3533 if (error != 0) 3534 return; 3535 ds = dmu_objset_ds(os); 3536 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) { 3537 char originname[MAXNAMELEN]; 3538 dsl_dataset_name(ds->ds_prev, originname); 3539 dmu_objset_rele(os, FTAG); 3540 (void) zfs_unmount_snap(originname); 3541 } else { 3542 dmu_objset_rele(os, FTAG); 3543 } 3544} 3545 3546/* 3547 * innvl: { 3548 * "snaps" -> { snapshot1, snapshot2 } 3549 * (optional boolean) "defer" 3550 * } 3551 * 3552 * outnvl: snapshot -> error code (int32) 3553 * 3554 */ 3555/* ARGSUSED */ 3556static int 3557zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3558{ 3559 int error, poollen; 3560 nvlist_t *snaps; 3561 nvpair_t *pair; 3562 boolean_t defer; 3563 3564 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 3565 return (SET_ERROR(EINVAL)); 3566 defer = nvlist_exists(innvl, "defer"); 3567 3568 poollen = strlen(poolname); 3569 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 3570 pair = nvlist_next_nvpair(snaps, pair)) { 3571 const char *name = nvpair_name(pair); 3572 3573 /* 3574 * The snap must be in the specified pool to prevent the 3575 * invalid removal of zvol minors below. 3576 */ 3577 if (strncmp(name, poolname, poollen) != 0 || 3578 (name[poollen] != '/' && name[poollen] != '@')) 3579 return (SET_ERROR(EXDEV)); 3580 3581 error = zfs_unmount_snap(name); 3582 if (error != 0) 3583 return (error); 3584#if defined(__FreeBSD__) 3585 zvol_remove_minors(name); 3586#endif 3587 } 3588 3589 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl)); 3590} 3591 3592/* 3593 * Create bookmarks. Bookmark names are of the form <fs>#<bmark>. 3594 * All bookmarks must be in the same pool. 3595 * 3596 * innvl: { 3597 * bookmark1 -> snapshot1, bookmark2 -> snapshot2 3598 * } 3599 * 3600 * outnvl: bookmark -> error code (int32) 3601 * 3602 */ 3603/* ARGSUSED */ 3604static int 3605zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3606{ 3607 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 3608 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 3609 char *snap_name; 3610 3611 /* 3612 * Verify the snapshot argument. 3613 */ 3614 if (nvpair_value_string(pair, &snap_name) != 0) 3615 return (SET_ERROR(EINVAL)); 3616 3617 3618 /* Verify that the keys (bookmarks) are unique */ 3619 for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair); 3620 pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) { 3621 if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0) 3622 return (SET_ERROR(EINVAL)); 3623 } 3624 } 3625 3626 return (dsl_bookmark_create(innvl, outnvl)); 3627} 3628 3629/* 3630 * innvl: { 3631 * property 1, property 2, ... 3632 * } 3633 * 3634 * outnvl: { 3635 * bookmark name 1 -> { property 1, property 2, ... }, 3636 * bookmark name 2 -> { property 1, property 2, ... } 3637 * } 3638 * 3639 */ 3640static int 3641zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3642{ 3643 return (dsl_get_bookmarks(fsname, innvl, outnvl)); 3644} 3645 3646/* 3647 * innvl: { 3648 * bookmark name 1, bookmark name 2 3649 * } 3650 * 3651 * outnvl: bookmark -> error code (int32) 3652 * 3653 */ 3654static int 3655zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl, 3656 nvlist_t *outnvl) 3657{ 3658 int error, poollen; 3659 3660 poollen = strlen(poolname); 3661 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 3662 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 3663 const char *name = nvpair_name(pair); 3664 const char *cp = strchr(name, '#'); 3665 3666 /* 3667 * The bookmark name must contain an #, and the part after it 3668 * must contain only valid characters. 3669 */ 3670 if (cp == NULL || 3671 zfs_component_namecheck(cp + 1, NULL, NULL) != 0) 3672 return (SET_ERROR(EINVAL)); 3673 3674 /* 3675 * The bookmark must be in the specified pool. 3676 */ 3677 if (strncmp(name, poolname, poollen) != 0 || 3678 (name[poollen] != '/' && name[poollen] != '#')) 3679 return (SET_ERROR(EXDEV)); 3680 } 3681 3682 error = dsl_bookmark_destroy(innvl, outnvl); 3683 return (error); 3684} 3685 3686/* 3687 * inputs: 3688 * zc_name name of dataset to destroy 3689 * zc_objset_type type of objset 3690 * zc_defer_destroy mark for deferred destroy 3691 * 3692 * outputs: none 3693 */ 3694static int 3695zfs_ioc_destroy(zfs_cmd_t *zc) 3696{ 3697 int err; 3698 3699 if (zc->zc_objset_type == DMU_OST_ZFS) { 3700 err = zfs_unmount_snap(zc->zc_name); 3701 if (err != 0) 3702 return (err); 3703 } 3704 3705 if (strchr(zc->zc_name, '@')) 3706 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy); 3707 else 3708 err = dsl_destroy_head(zc->zc_name); 3709 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0) 3710#ifdef __FreeBSD__ 3711 zvol_remove_minors(zc->zc_name); 3712#else 3713 (void) zvol_remove_minor(zc->zc_name); 3714#endif 3715 return (err); 3716} 3717 3718/* 3719 * fsname is name of dataset to rollback (to most recent snapshot) 3720 * 3721 * innvl is not used. 3722 * 3723 * outnvl: "target" -> name of most recent snapshot 3724 * } 3725 */ 3726/* ARGSUSED */ 3727static int 3728zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl) 3729{ 3730 zfsvfs_t *zfsvfs; 3731 int error; 3732 3733 if (getzfsvfs(fsname, &zfsvfs) == 0) { 3734 error = zfs_suspend_fs(zfsvfs); 3735 if (error == 0) { 3736 int resume_err; 3737 3738 error = dsl_dataset_rollback(fsname, zfsvfs, outnvl); 3739 resume_err = zfs_resume_fs(zfsvfs, fsname); 3740 error = error ? error : resume_err; 3741 } 3742#ifdef illumos 3743 VFS_RELE(zfsvfs->z_vfs); 3744#else 3745 vfs_unbusy(zfsvfs->z_vfs); 3746#endif 3747 } else { 3748 error = dsl_dataset_rollback(fsname, NULL, outnvl); 3749 } 3750 return (error); 3751} 3752 3753static int 3754recursive_unmount(const char *fsname, void *arg) 3755{ 3756 const char *snapname = arg; 3757 char fullname[MAXNAMELEN]; 3758 3759 (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname); 3760 return (zfs_unmount_snap(fullname)); 3761} 3762 3763/* 3764 * inputs: 3765 * zc_name old name of dataset 3766 * zc_value new name of dataset 3767 * zc_cookie recursive flag (only valid for snapshots) 3768 * 3769 * outputs: none 3770 */ 3771static int 3772zfs_ioc_rename(zfs_cmd_t *zc) 3773{ 3774 boolean_t recursive = zc->zc_cookie & 1; 3775 char *at; 3776 boolean_t allow_mounted = B_TRUE; 3777 3778#ifdef __FreeBSD__ 3779 allow_mounted = (zc->zc_cookie & 2) != 0; 3780#endif 3781 3782 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0'; 3783 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 3784 strchr(zc->zc_value, '%')) 3785 return (SET_ERROR(EINVAL)); 3786 3787 at = strchr(zc->zc_name, '@'); 3788 if (at != NULL) { 3789 /* snaps must be in same fs */ 3790 int error; 3791 3792 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1)) 3793 return (SET_ERROR(EXDEV)); 3794 *at = '\0'; 3795 if (zc->zc_objset_type == DMU_OST_ZFS && !allow_mounted) { 3796 error = dmu_objset_find(zc->zc_name, 3797 recursive_unmount, at + 1, 3798 recursive ? DS_FIND_CHILDREN : 0); 3799 if (error != 0) { 3800 *at = '@'; 3801 return (error); 3802 } 3803 } 3804 error = dsl_dataset_rename_snapshot(zc->zc_name, 3805 at + 1, strchr(zc->zc_value, '@') + 1, recursive); 3806 *at = '@'; 3807 3808 return (error); 3809 } else { 3810#ifdef illumos 3811 if (zc->zc_objset_type == DMU_OST_ZVOL) 3812 (void) zvol_remove_minor(zc->zc_name); 3813#endif 3814 return (dsl_dir_rename(zc->zc_name, zc->zc_value)); 3815 } 3816} 3817 3818static int 3819zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr) 3820{ 3821 const char *propname = nvpair_name(pair); 3822 boolean_t issnap = (strchr(dsname, '@') != NULL); 3823 zfs_prop_t prop = zfs_name_to_prop(propname); 3824 uint64_t intval; 3825 int err; 3826 3827 if (prop == ZPROP_INVAL) { 3828 if (zfs_prop_user(propname)) { 3829 if (err = zfs_secpolicy_write_perms(dsname, 3830 ZFS_DELEG_PERM_USERPROP, cr)) 3831 return (err); 3832 return (0); 3833 } 3834 3835 if (!issnap && zfs_prop_userquota(propname)) { 3836 const char *perm = NULL; 3837 const char *uq_prefix = 3838 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA]; 3839 const char *gq_prefix = 3840 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA]; 3841 3842 if (strncmp(propname, uq_prefix, 3843 strlen(uq_prefix)) == 0) { 3844 perm = ZFS_DELEG_PERM_USERQUOTA; 3845 } else if (strncmp(propname, gq_prefix, 3846 strlen(gq_prefix)) == 0) { 3847 perm = ZFS_DELEG_PERM_GROUPQUOTA; 3848 } else { 3849 /* USERUSED and GROUPUSED are read-only */ 3850 return (SET_ERROR(EINVAL)); 3851 } 3852 3853 if (err = zfs_secpolicy_write_perms(dsname, perm, cr)) 3854 return (err); 3855 return (0); 3856 } 3857 3858 return (SET_ERROR(EINVAL)); 3859 } 3860 3861 if (issnap) 3862 return (SET_ERROR(EINVAL)); 3863 3864 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 3865 /* 3866 * dsl_prop_get_all_impl() returns properties in this 3867 * format. 3868 */ 3869 nvlist_t *attrs; 3870 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 3871 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 3872 &pair) == 0); 3873 } 3874 3875 /* 3876 * Check that this value is valid for this pool version 3877 */ 3878 switch (prop) { 3879 case ZFS_PROP_COMPRESSION: 3880 /* 3881 * If the user specified gzip compression, make sure 3882 * the SPA supports it. We ignore any errors here since 3883 * we'll catch them later. 3884 */ 3885 if (nvpair_value_uint64(pair, &intval) == 0) { 3886 if (intval >= ZIO_COMPRESS_GZIP_1 && 3887 intval <= ZIO_COMPRESS_GZIP_9 && 3888 zfs_earlier_version(dsname, 3889 SPA_VERSION_GZIP_COMPRESSION)) { 3890 return (SET_ERROR(ENOTSUP)); 3891 } 3892 3893 if (intval == ZIO_COMPRESS_ZLE && 3894 zfs_earlier_version(dsname, 3895 SPA_VERSION_ZLE_COMPRESSION)) 3896 return (SET_ERROR(ENOTSUP)); 3897 3898 if (intval == ZIO_COMPRESS_LZ4) { 3899 spa_t *spa; 3900 3901 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 3902 return (err); 3903 3904 if (!spa_feature_is_enabled(spa, 3905 SPA_FEATURE_LZ4_COMPRESS)) { 3906 spa_close(spa, FTAG); 3907 return (SET_ERROR(ENOTSUP)); 3908 } 3909 spa_close(spa, FTAG); 3910 } 3911 3912 /* 3913 * If this is a bootable dataset then 3914 * verify that the compression algorithm 3915 * is supported for booting. We must return 3916 * something other than ENOTSUP since it 3917 * implies a downrev pool version. 3918 */ 3919 if (zfs_is_bootfs(dsname) && 3920 !BOOTFS_COMPRESS_VALID(intval)) { 3921 return (SET_ERROR(ERANGE)); 3922 } 3923 } 3924 break; 3925 3926 case ZFS_PROP_COPIES: 3927 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS)) 3928 return (SET_ERROR(ENOTSUP)); 3929 break; 3930 3931 case ZFS_PROP_RECORDSIZE: 3932 /* Record sizes above 128k need the feature to be enabled */ 3933 if (nvpair_value_uint64(pair, &intval) == 0 && 3934 intval > SPA_OLD_MAXBLOCKSIZE) { 3935 spa_t *spa; 3936 3937 /* 3938 * If this is a bootable dataset then 3939 * the we don't allow large (>128K) blocks, 3940 * because GRUB doesn't support them. 3941 */ 3942 if (zfs_is_bootfs(dsname) && 3943 intval > SPA_OLD_MAXBLOCKSIZE) { 3944 return (SET_ERROR(ERANGE)); 3945 } 3946 3947 /* 3948 * We don't allow setting the property above 1MB, 3949 * unless the tunable has been changed. 3950 */ 3951 if (intval > zfs_max_recordsize || 3952 intval > SPA_MAXBLOCKSIZE) 3953 return (SET_ERROR(ERANGE)); 3954 3955 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 3956 return (err); 3957 3958 if (!spa_feature_is_enabled(spa, 3959 SPA_FEATURE_LARGE_BLOCKS)) { 3960 spa_close(spa, FTAG); 3961 return (SET_ERROR(ENOTSUP)); 3962 } 3963 spa_close(spa, FTAG); 3964 } 3965 break; 3966 3967 case ZFS_PROP_SHARESMB: 3968 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID)) 3969 return (SET_ERROR(ENOTSUP)); 3970 break; 3971 3972 case ZFS_PROP_ACLINHERIT: 3973 if (nvpair_type(pair) == DATA_TYPE_UINT64 && 3974 nvpair_value_uint64(pair, &intval) == 0) { 3975 if (intval == ZFS_ACL_PASSTHROUGH_X && 3976 zfs_earlier_version(dsname, 3977 SPA_VERSION_PASSTHROUGH_X)) 3978 return (SET_ERROR(ENOTSUP)); 3979 } 3980 break; 3981 3982 case ZFS_PROP_CHECKSUM: 3983 case ZFS_PROP_DEDUP: 3984 { 3985 spa_feature_t feature; 3986 spa_t *spa; 3987 3988 /* dedup feature version checks */ 3989 if (prop == ZFS_PROP_DEDUP && 3990 zfs_earlier_version(dsname, SPA_VERSION_DEDUP)) 3991 return (SET_ERROR(ENOTSUP)); 3992 3993 if (nvpair_value_uint64(pair, &intval) != 0) 3994 return (SET_ERROR(EINVAL)); 3995 3996 /* check prop value is enabled in features */ 3997 feature = zio_checksum_to_feature(intval & ZIO_CHECKSUM_MASK); 3998 if (feature == SPA_FEATURE_NONE) 3999 break; 4000 4001 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 4002 return (err); 4003 /* 4004 * Salted checksums are not supported on root pools. 4005 */ 4006 if (spa_bootfs(spa) != 0 && 4007 intval < ZIO_CHECKSUM_FUNCTIONS && 4008 (zio_checksum_table[intval].ci_flags & 4009 ZCHECKSUM_FLAG_SALTED)) { 4010 spa_close(spa, FTAG); 4011 return (SET_ERROR(ERANGE)); 4012 } 4013 if (!spa_feature_is_enabled(spa, feature)) { 4014 spa_close(spa, FTAG); 4015 return (SET_ERROR(ENOTSUP)); 4016 } 4017 spa_close(spa, FTAG); 4018 break; 4019 } 4020 } 4021 4022 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED())); 4023} 4024 4025/* 4026 * Checks for a race condition to make sure we don't increment a feature flag 4027 * multiple times. 4028 */ 4029static int 4030zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx) 4031{ 4032 spa_t *spa = dmu_tx_pool(tx)->dp_spa; 4033 spa_feature_t *featurep = arg; 4034 4035 if (!spa_feature_is_active(spa, *featurep)) 4036 return (0); 4037 else 4038 return (SET_ERROR(EBUSY)); 4039} 4040 4041/* 4042 * The callback invoked on feature activation in the sync task caused by 4043 * zfs_prop_activate_feature. 4044 */ 4045static void 4046zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx) 4047{ 4048 spa_t *spa = dmu_tx_pool(tx)->dp_spa; 4049 spa_feature_t *featurep = arg; 4050 4051 spa_feature_incr(spa, *featurep, tx); 4052} 4053 4054/* 4055 * Activates a feature on a pool in response to a property setting. This 4056 * creates a new sync task which modifies the pool to reflect the feature 4057 * as being active. 4058 */ 4059static int 4060zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature) 4061{ 4062 int err; 4063 4064 /* EBUSY here indicates that the feature is already active */ 4065 err = dsl_sync_task(spa_name(spa), 4066 zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync, 4067 &feature, 2, ZFS_SPACE_CHECK_RESERVED); 4068 4069 if (err != 0 && err != EBUSY) 4070 return (err); 4071 else 4072 return (0); 4073} 4074 4075/* 4076 * Removes properties from the given props list that fail permission checks 4077 * needed to clear them and to restore them in case of a receive error. For each 4078 * property, make sure we have both set and inherit permissions. 4079 * 4080 * Returns the first error encountered if any permission checks fail. If the 4081 * caller provides a non-NULL errlist, it also gives the complete list of names 4082 * of all the properties that failed a permission check along with the 4083 * corresponding error numbers. The caller is responsible for freeing the 4084 * returned errlist. 4085 * 4086 * If every property checks out successfully, zero is returned and the list 4087 * pointed at by errlist is NULL. 4088 */ 4089static int 4090zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist) 4091{ 4092 zfs_cmd_t *zc; 4093 nvpair_t *pair, *next_pair; 4094 nvlist_t *errors; 4095 int err, rv = 0; 4096 4097 if (props == NULL) 4098 return (0); 4099 4100 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0); 4101 4102 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP); 4103 (void) strcpy(zc->zc_name, dataset); 4104 pair = nvlist_next_nvpair(props, NULL); 4105 while (pair != NULL) { 4106 next_pair = nvlist_next_nvpair(props, pair); 4107 4108 (void) strcpy(zc->zc_value, nvpair_name(pair)); 4109 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 || 4110 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) { 4111 VERIFY(nvlist_remove_nvpair(props, pair) == 0); 4112 VERIFY(nvlist_add_int32(errors, 4113 zc->zc_value, err) == 0); 4114 } 4115 pair = next_pair; 4116 } 4117 kmem_free(zc, sizeof (zfs_cmd_t)); 4118 4119 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) { 4120 nvlist_free(errors); 4121 errors = NULL; 4122 } else { 4123 VERIFY(nvpair_value_int32(pair, &rv) == 0); 4124 } 4125 4126 if (errlist == NULL) 4127 nvlist_free(errors); 4128 else 4129 *errlist = errors; 4130 4131 return (rv); 4132} 4133 4134static boolean_t 4135propval_equals(nvpair_t *p1, nvpair_t *p2) 4136{ 4137 if (nvpair_type(p1) == DATA_TYPE_NVLIST) { 4138 /* dsl_prop_get_all_impl() format */ 4139 nvlist_t *attrs; 4140 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0); 4141 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 4142 &p1) == 0); 4143 } 4144 4145 if (nvpair_type(p2) == DATA_TYPE_NVLIST) { 4146 nvlist_t *attrs; 4147 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0); 4148 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 4149 &p2) == 0); 4150 } 4151 4152 if (nvpair_type(p1) != nvpair_type(p2)) 4153 return (B_FALSE); 4154 4155 if (nvpair_type(p1) == DATA_TYPE_STRING) { 4156 char *valstr1, *valstr2; 4157 4158 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0); 4159 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0); 4160 return (strcmp(valstr1, valstr2) == 0); 4161 } else { 4162 uint64_t intval1, intval2; 4163 4164 VERIFY(nvpair_value_uint64(p1, &intval1) == 0); 4165 VERIFY(nvpair_value_uint64(p2, &intval2) == 0); 4166 return (intval1 == intval2); 4167 } 4168} 4169 4170/* 4171 * Remove properties from props if they are not going to change (as determined 4172 * by comparison with origprops). Remove them from origprops as well, since we 4173 * do not need to clear or restore properties that won't change. 4174 */ 4175static void 4176props_reduce(nvlist_t *props, nvlist_t *origprops) 4177{ 4178 nvpair_t *pair, *next_pair; 4179 4180 if (origprops == NULL) 4181 return; /* all props need to be received */ 4182 4183 pair = nvlist_next_nvpair(props, NULL); 4184 while (pair != NULL) { 4185 const char *propname = nvpair_name(pair); 4186 nvpair_t *match; 4187 4188 next_pair = nvlist_next_nvpair(props, pair); 4189 4190 if ((nvlist_lookup_nvpair(origprops, propname, 4191 &match) != 0) || !propval_equals(pair, match)) 4192 goto next; /* need to set received value */ 4193 4194 /* don't clear the existing received value */ 4195 (void) nvlist_remove_nvpair(origprops, match); 4196 /* don't bother receiving the property */ 4197 (void) nvlist_remove_nvpair(props, pair); 4198next: 4199 pair = next_pair; 4200 } 4201} 4202 4203/* 4204 * Extract properties that cannot be set PRIOR to the receipt of a dataset. 4205 * For example, refquota cannot be set until after the receipt of a dataset, 4206 * because in replication streams, an older/earlier snapshot may exceed the 4207 * refquota. We want to receive the older/earlier snapshot, but setting 4208 * refquota pre-receipt will set the dsl's ACTUAL quota, which will prevent 4209 * the older/earlier snapshot from being received (with EDQUOT). 4210 * 4211 * The ZFS test "zfs_receive_011_pos" demonstrates such a scenario. 4212 * 4213 * libzfs will need to be judicious handling errors encountered by props 4214 * extracted by this function. 4215 */ 4216static nvlist_t * 4217extract_delay_props(nvlist_t *props) 4218{ 4219 nvlist_t *delayprops; 4220 nvpair_t *nvp, *tmp; 4221 static const zfs_prop_t delayable[] = { ZFS_PROP_REFQUOTA, 0 }; 4222 int i; 4223 4224 VERIFY(nvlist_alloc(&delayprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 4225 4226 for (nvp = nvlist_next_nvpair(props, NULL); nvp != NULL; 4227 nvp = nvlist_next_nvpair(props, nvp)) { 4228 /* 4229 * strcmp() is safe because zfs_prop_to_name() always returns 4230 * a bounded string. 4231 */ 4232 for (i = 0; delayable[i] != 0; i++) { 4233 if (strcmp(zfs_prop_to_name(delayable[i]), 4234 nvpair_name(nvp)) == 0) { 4235 break; 4236 } 4237 } 4238 if (delayable[i] != 0) { 4239 tmp = nvlist_prev_nvpair(props, nvp); 4240 VERIFY(nvlist_add_nvpair(delayprops, nvp) == 0); 4241 VERIFY(nvlist_remove_nvpair(props, nvp) == 0); 4242 nvp = tmp; 4243 } 4244 } 4245 4246 if (nvlist_empty(delayprops)) { 4247 nvlist_free(delayprops); 4248 delayprops = NULL; 4249 } 4250 return (delayprops); 4251} 4252 4253#ifdef DEBUG 4254static boolean_t zfs_ioc_recv_inject_err; 4255#endif 4256 4257/* 4258 * inputs: 4259 * zc_name name of containing filesystem 4260 * zc_nvlist_src{_size} nvlist of properties to apply 4261 * zc_value name of snapshot to create 4262 * zc_string name of clone origin (if DRR_FLAG_CLONE) 4263 * zc_cookie file descriptor to recv from 4264 * zc_begin_record the BEGIN record of the stream (not byteswapped) 4265 * zc_guid force flag 4266 * zc_cleanup_fd cleanup-on-exit file descriptor 4267 * zc_action_handle handle for this guid/ds mapping (or zero on first call) 4268 * zc_resumable if data is incomplete assume sender will resume 4269 * 4270 * outputs: 4271 * zc_cookie number of bytes read 4272 * zc_nvlist_dst{_size} error for each unapplied received property 4273 * zc_obj zprop_errflags_t 4274 * zc_action_handle handle for this guid/ds mapping 4275 */ 4276static int 4277zfs_ioc_recv(zfs_cmd_t *zc) 4278{ 4279 file_t *fp; 4280 dmu_recv_cookie_t drc; 4281 boolean_t force = (boolean_t)zc->zc_guid; 4282 int fd; 4283 int error = 0; 4284 int props_error = 0; 4285 nvlist_t *errors; 4286 offset_t off; 4287 nvlist_t *props = NULL; /* sent properties */ 4288 nvlist_t *origprops = NULL; /* existing properties */ 4289 nvlist_t *delayprops = NULL; /* sent properties applied post-receive */ 4290 char *origin = NULL; 4291 char *tosnap; 4292 char tofs[ZFS_MAXNAMELEN]; 4293 cap_rights_t rights; 4294 boolean_t first_recvd_props = B_FALSE; 4295 4296 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 4297 strchr(zc->zc_value, '@') == NULL || 4298 strchr(zc->zc_value, '%')) 4299 return (SET_ERROR(EINVAL)); 4300 4301 (void) strcpy(tofs, zc->zc_value); 4302 tosnap = strchr(tofs, '@'); 4303 *tosnap++ = '\0'; 4304 4305 if (zc->zc_nvlist_src != 0 && 4306 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 4307 zc->zc_iflags, &props)) != 0) 4308 return (error); 4309 4310 fd = zc->zc_cookie; 4311#ifdef illumos 4312 fp = getf(fd); 4313#else 4314 fget_read(curthread, fd, cap_rights_init(&rights, CAP_PREAD), &fp); 4315#endif 4316 if (fp == NULL) { 4317 nvlist_free(props); 4318 return (SET_ERROR(EBADF)); 4319 } 4320 4321 errors = fnvlist_alloc(); 4322 4323 if (zc->zc_string[0]) 4324 origin = zc->zc_string; 4325 4326 error = dmu_recv_begin(tofs, tosnap, 4327 &zc->zc_begin_record, force, zc->zc_resumable, origin, &drc); 4328 if (error != 0) 4329 goto out; 4330 4331 /* 4332 * Set properties before we receive the stream so that they are applied 4333 * to the new data. Note that we must call dmu_recv_stream() if 4334 * dmu_recv_begin() succeeds. 4335 */ 4336 if (props != NULL && !drc.drc_newfs) { 4337 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >= 4338 SPA_VERSION_RECVD_PROPS && 4339 !dsl_prop_get_hasrecvd(tofs)) 4340 first_recvd_props = B_TRUE; 4341 4342 /* 4343 * If new received properties are supplied, they are to 4344 * completely replace the existing received properties, so stash 4345 * away the existing ones. 4346 */ 4347 if (dsl_prop_get_received(tofs, &origprops) == 0) { 4348 nvlist_t *errlist = NULL; 4349 /* 4350 * Don't bother writing a property if its value won't 4351 * change (and avoid the unnecessary security checks). 4352 * 4353 * The first receive after SPA_VERSION_RECVD_PROPS is a 4354 * special case where we blow away all local properties 4355 * regardless. 4356 */ 4357 if (!first_recvd_props) 4358 props_reduce(props, origprops); 4359 if (zfs_check_clearable(tofs, origprops, &errlist) != 0) 4360 (void) nvlist_merge(errors, errlist, 0); 4361 nvlist_free(errlist); 4362 4363 if (clear_received_props(tofs, origprops, 4364 first_recvd_props ? NULL : props) != 0) 4365 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 4366 } else { 4367 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 4368 } 4369 } 4370 4371 if (props != NULL) { 4372 props_error = dsl_prop_set_hasrecvd(tofs); 4373 4374 if (props_error == 0) { 4375 delayprops = extract_delay_props(props); 4376 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED, 4377 props, errors); 4378 } 4379 } 4380 4381 off = fp->f_offset; 4382 error = dmu_recv_stream(&drc, fp, &off, zc->zc_cleanup_fd, 4383 &zc->zc_action_handle); 4384 4385 if (error == 0) { 4386 zfsvfs_t *zfsvfs = NULL; 4387 4388 if (getzfsvfs(tofs, &zfsvfs) == 0) { 4389 /* online recv */ 4390 int end_err; 4391 4392 error = zfs_suspend_fs(zfsvfs); 4393 /* 4394 * If the suspend fails, then the recv_end will 4395 * likely also fail, and clean up after itself. 4396 */ 4397 end_err = dmu_recv_end(&drc, zfsvfs); 4398 if (error == 0) 4399 error = zfs_resume_fs(zfsvfs, tofs); 4400 error = error ? error : end_err; 4401#ifdef illumos 4402 VFS_RELE(zfsvfs->z_vfs); 4403#else 4404 vfs_unbusy(zfsvfs->z_vfs); 4405#endif 4406 } else { 4407 error = dmu_recv_end(&drc, NULL); 4408 } 4409 4410 /* Set delayed properties now, after we're done receiving. */ 4411 if (delayprops != NULL && error == 0) { 4412 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED, 4413 delayprops, errors); 4414 } 4415 } 4416 4417 if (delayprops != NULL) { 4418 /* 4419 * Merge delayed props back in with initial props, in case 4420 * we're DEBUG and zfs_ioc_recv_inject_err is set (which means 4421 * we have to make sure clear_received_props() includes 4422 * the delayed properties). 4423 * 4424 * Since zfs_ioc_recv_inject_err is only in DEBUG kernels, 4425 * using ASSERT() will be just like a VERIFY. 4426 */ 4427 ASSERT(nvlist_merge(props, delayprops, 0) == 0); 4428 nvlist_free(delayprops); 4429 } 4430 4431 /* 4432 * Now that all props, initial and delayed, are set, report the prop 4433 * errors to the caller. 4434 */ 4435 if (zc->zc_nvlist_dst_size != 0 && 4436 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 || 4437 put_nvlist(zc, errors) != 0)) { 4438 /* 4439 * Caller made zc->zc_nvlist_dst less than the minimum expected 4440 * size or supplied an invalid address. 4441 */ 4442 props_error = SET_ERROR(EINVAL); 4443 } 4444 4445 zc->zc_cookie = off - fp->f_offset; 4446 if (off >= 0 && off <= MAXOFFSET_T) 4447 fp->f_offset = off; 4448 4449#ifdef DEBUG 4450 if (zfs_ioc_recv_inject_err) { 4451 zfs_ioc_recv_inject_err = B_FALSE; 4452 error = 1; 4453 } 4454#endif 4455 4456#ifdef __FreeBSD__ 4457 if (error == 0) 4458 zvol_create_minors(tofs); 4459#endif 4460 4461 /* 4462 * On error, restore the original props. 4463 */ 4464 if (error != 0 && props != NULL && !drc.drc_newfs) { 4465 if (clear_received_props(tofs, props, NULL) != 0) { 4466 /* 4467 * We failed to clear the received properties. 4468 * Since we may have left a $recvd value on the 4469 * system, we can't clear the $hasrecvd flag. 4470 */ 4471 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4472 } else if (first_recvd_props) { 4473 dsl_prop_unset_hasrecvd(tofs); 4474 } 4475 4476 if (origprops == NULL && !drc.drc_newfs) { 4477 /* We failed to stash the original properties. */ 4478 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4479 } 4480 4481 /* 4482 * dsl_props_set() will not convert RECEIVED to LOCAL on or 4483 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL 4484 * explictly if we're restoring local properties cleared in the 4485 * first new-style receive. 4486 */ 4487 if (origprops != NULL && 4488 zfs_set_prop_nvlist(tofs, (first_recvd_props ? 4489 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED), 4490 origprops, NULL) != 0) { 4491 /* 4492 * We stashed the original properties but failed to 4493 * restore them. 4494 */ 4495 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4496 } 4497 } 4498out: 4499 nvlist_free(props); 4500 nvlist_free(origprops); 4501 nvlist_free(errors); 4502 releasef(fd); 4503 4504 if (error == 0) 4505 error = props_error; 4506 4507 return (error); 4508} 4509 4510/* 4511 * inputs: 4512 * zc_name name of snapshot to send 4513 * zc_cookie file descriptor to send stream to 4514 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj) 4515 * zc_sendobj objsetid of snapshot to send 4516 * zc_fromobj objsetid of incremental fromsnap (may be zero) 4517 * zc_guid if set, estimate size of stream only. zc_cookie is ignored. 4518 * output size in zc_objset_type. 4519 * zc_flags lzc_send_flags 4520 * 4521 * outputs: 4522 * zc_objset_type estimated size, if zc_guid is set 4523 */ 4524static int 4525zfs_ioc_send(zfs_cmd_t *zc) 4526{ 4527 int error; 4528 offset_t off; 4529 boolean_t estimate = (zc->zc_guid != 0); 4530 boolean_t embedok = (zc->zc_flags & 0x1); 4531 boolean_t large_block_ok = (zc->zc_flags & 0x2); 4532 4533 if (zc->zc_obj != 0) { 4534 dsl_pool_t *dp; 4535 dsl_dataset_t *tosnap; 4536 4537 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4538 if (error != 0) 4539 return (error); 4540 4541 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap); 4542 if (error != 0) { 4543 dsl_pool_rele(dp, FTAG); 4544 return (error); 4545 } 4546 4547 if (dsl_dir_is_clone(tosnap->ds_dir)) 4548 zc->zc_fromobj = 4549 dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj; 4550 dsl_dataset_rele(tosnap, FTAG); 4551 dsl_pool_rele(dp, FTAG); 4552 } 4553 4554 if (estimate) { 4555 dsl_pool_t *dp; 4556 dsl_dataset_t *tosnap; 4557 dsl_dataset_t *fromsnap = NULL; 4558 4559 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4560 if (error != 0) 4561 return (error); 4562 4563 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap); 4564 if (error != 0) { 4565 dsl_pool_rele(dp, FTAG); 4566 return (error); 4567 } 4568 4569 if (zc->zc_fromobj != 0) { 4570 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj, 4571 FTAG, &fromsnap); 4572 if (error != 0) { 4573 dsl_dataset_rele(tosnap, FTAG); 4574 dsl_pool_rele(dp, FTAG); 4575 return (error); 4576 } 4577 } 4578 4579 error = dmu_send_estimate(tosnap, fromsnap, 4580 &zc->zc_objset_type); 4581 4582 if (fromsnap != NULL) 4583 dsl_dataset_rele(fromsnap, FTAG); 4584 dsl_dataset_rele(tosnap, FTAG); 4585 dsl_pool_rele(dp, FTAG); 4586 } else { 4587 file_t *fp; 4588 cap_rights_t rights; 4589 4590#ifdef illumos 4591 fp = getf(zc->zc_cookie); 4592#else 4593 fget_write(curthread, zc->zc_cookie, 4594 cap_rights_init(&rights, CAP_WRITE), &fp); 4595#endif 4596 if (fp == NULL) 4597 return (SET_ERROR(EBADF)); 4598 4599 off = fp->f_offset; 4600 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj, 4601 zc->zc_fromobj, embedok, large_block_ok, 4602#ifdef illumos 4603 zc->zc_cookie, fp->f_vnode, &off); 4604#else 4605 zc->zc_cookie, fp, &off); 4606#endif 4607 4608 if (off >= 0 && off <= MAXOFFSET_T) 4609 fp->f_offset = off; 4610 releasef(zc->zc_cookie); 4611 } 4612 return (error); 4613} 4614 4615/* 4616 * inputs: 4617 * zc_name name of snapshot on which to report progress 4618 * zc_cookie file descriptor of send stream 4619 * 4620 * outputs: 4621 * zc_cookie number of bytes written in send stream thus far 4622 */ 4623static int 4624zfs_ioc_send_progress(zfs_cmd_t *zc) 4625{ 4626 dsl_pool_t *dp; 4627 dsl_dataset_t *ds; 4628 dmu_sendarg_t *dsp = NULL; 4629 int error; 4630 4631 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4632 if (error != 0) 4633 return (error); 4634 4635 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds); 4636 if (error != 0) { 4637 dsl_pool_rele(dp, FTAG); 4638 return (error); 4639 } 4640 4641 mutex_enter(&ds->ds_sendstream_lock); 4642 4643 /* 4644 * Iterate over all the send streams currently active on this dataset. 4645 * If there's one which matches the specified file descriptor _and_ the 4646 * stream was started by the current process, return the progress of 4647 * that stream. 4648 */ 4649 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL; 4650 dsp = list_next(&ds->ds_sendstreams, dsp)) { 4651 if (dsp->dsa_outfd == zc->zc_cookie && 4652 dsp->dsa_proc == curproc) 4653 break; 4654 } 4655 4656 if (dsp != NULL) 4657 zc->zc_cookie = *(dsp->dsa_off); 4658 else 4659 error = SET_ERROR(ENOENT); 4660 4661 mutex_exit(&ds->ds_sendstream_lock); 4662 dsl_dataset_rele(ds, FTAG); 4663 dsl_pool_rele(dp, FTAG); 4664 return (error); 4665} 4666 4667static int 4668zfs_ioc_inject_fault(zfs_cmd_t *zc) 4669{ 4670 int id, error; 4671 4672 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id, 4673 &zc->zc_inject_record); 4674 4675 if (error == 0) 4676 zc->zc_guid = (uint64_t)id; 4677 4678 return (error); 4679} 4680 4681static int 4682zfs_ioc_clear_fault(zfs_cmd_t *zc) 4683{ 4684 return (zio_clear_fault((int)zc->zc_guid)); 4685} 4686 4687static int 4688zfs_ioc_inject_list_next(zfs_cmd_t *zc) 4689{ 4690 int id = (int)zc->zc_guid; 4691 int error; 4692 4693 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name), 4694 &zc->zc_inject_record); 4695 4696 zc->zc_guid = id; 4697 4698 return (error); 4699} 4700 4701static int 4702zfs_ioc_error_log(zfs_cmd_t *zc) 4703{ 4704 spa_t *spa; 4705 int error; 4706 size_t count = (size_t)zc->zc_nvlist_dst_size; 4707 4708 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 4709 return (error); 4710 4711 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst, 4712 &count); 4713 if (error == 0) 4714 zc->zc_nvlist_dst_size = count; 4715 else 4716 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa); 4717 4718 spa_close(spa, FTAG); 4719 4720 return (error); 4721} 4722 4723static int 4724zfs_ioc_clear(zfs_cmd_t *zc) 4725{ 4726 spa_t *spa; 4727 vdev_t *vd; 4728 int error; 4729 4730 /* 4731 * On zpool clear we also fix up missing slogs 4732 */ 4733 mutex_enter(&spa_namespace_lock); 4734 spa = spa_lookup(zc->zc_name); 4735 if (spa == NULL) { 4736 mutex_exit(&spa_namespace_lock); 4737 return (SET_ERROR(EIO)); 4738 } 4739 if (spa_get_log_state(spa) == SPA_LOG_MISSING) { 4740 /* we need to let spa_open/spa_load clear the chains */ 4741 spa_set_log_state(spa, SPA_LOG_CLEAR); 4742 } 4743 spa->spa_last_open_failed = 0; 4744 mutex_exit(&spa_namespace_lock); 4745 4746 if (zc->zc_cookie & ZPOOL_NO_REWIND) { 4747 error = spa_open(zc->zc_name, &spa, FTAG); 4748 } else { 4749 nvlist_t *policy; 4750 nvlist_t *config = NULL; 4751 4752 if (zc->zc_nvlist_src == 0) 4753 return (SET_ERROR(EINVAL)); 4754 4755 if ((error = get_nvlist(zc->zc_nvlist_src, 4756 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) { 4757 error = spa_open_rewind(zc->zc_name, &spa, FTAG, 4758 policy, &config); 4759 if (config != NULL) { 4760 int err; 4761 4762 if ((err = put_nvlist(zc, config)) != 0) 4763 error = err; 4764 nvlist_free(config); 4765 } 4766 nvlist_free(policy); 4767 } 4768 } 4769 4770 if (error != 0) 4771 return (error); 4772 4773 spa_vdev_state_enter(spa, SCL_NONE); 4774 4775 if (zc->zc_guid == 0) { 4776 vd = NULL; 4777 } else { 4778 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE); 4779 if (vd == NULL) { 4780 (void) spa_vdev_state_exit(spa, NULL, ENODEV); 4781 spa_close(spa, FTAG); 4782 return (SET_ERROR(ENODEV)); 4783 } 4784 } 4785 4786 vdev_clear(spa, vd); 4787 4788 (void) spa_vdev_state_exit(spa, NULL, 0); 4789 4790 /* 4791 * Resume any suspended I/Os. 4792 */ 4793 if (zio_resume(spa) != 0) 4794 error = SET_ERROR(EIO); 4795 4796 spa_close(spa, FTAG); 4797 4798 return (error); 4799} 4800 4801static int 4802zfs_ioc_pool_reopen(zfs_cmd_t *zc) 4803{ 4804 spa_t *spa; 4805 int error; 4806 4807 error = spa_open(zc->zc_name, &spa, FTAG); 4808 if (error != 0) 4809 return (error); 4810 4811 spa_vdev_state_enter(spa, SCL_NONE); 4812 4813 /* 4814 * If a resilver is already in progress then set the 4815 * spa_scrub_reopen flag to B_TRUE so that we don't restart 4816 * the scan as a side effect of the reopen. Otherwise, let 4817 * vdev_open() decided if a resilver is required. 4818 */ 4819 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool); 4820 vdev_reopen(spa->spa_root_vdev); 4821 spa->spa_scrub_reopen = B_FALSE; 4822 4823 (void) spa_vdev_state_exit(spa, NULL, 0); 4824 spa_close(spa, FTAG); 4825 return (0); 4826} 4827/* 4828 * inputs: 4829 * zc_name name of filesystem 4830 * zc_value name of origin snapshot 4831 * 4832 * outputs: 4833 * zc_string name of conflicting snapshot, if there is one 4834 */ 4835static int 4836zfs_ioc_promote(zfs_cmd_t *zc) 4837{ 4838 char *cp; 4839 4840 /* 4841 * We don't need to unmount *all* the origin fs's snapshots, but 4842 * it's easier. 4843 */ 4844 cp = strchr(zc->zc_value, '@'); 4845 if (cp) 4846 *cp = '\0'; 4847 (void) dmu_objset_find(zc->zc_value, 4848 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS); 4849 return (dsl_dataset_promote(zc->zc_name, zc->zc_string)); 4850} 4851 4852/* 4853 * Retrieve a single {user|group}{used|quota}@... property. 4854 * 4855 * inputs: 4856 * zc_name name of filesystem 4857 * zc_objset_type zfs_userquota_prop_t 4858 * zc_value domain name (eg. "S-1-234-567-89") 4859 * zc_guid RID/UID/GID 4860 * 4861 * outputs: 4862 * zc_cookie property value 4863 */ 4864static int 4865zfs_ioc_userspace_one(zfs_cmd_t *zc) 4866{ 4867 zfsvfs_t *zfsvfs; 4868 int error; 4869 4870 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 4871 return (SET_ERROR(EINVAL)); 4872 4873 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE); 4874 if (error != 0) 4875 return (error); 4876 4877 error = zfs_userspace_one(zfsvfs, 4878 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie); 4879 zfsvfs_rele(zfsvfs, FTAG); 4880 4881 return (error); 4882} 4883 4884/* 4885 * inputs: 4886 * zc_name name of filesystem 4887 * zc_cookie zap cursor 4888 * zc_objset_type zfs_userquota_prop_t 4889 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist) 4890 * 4891 * outputs: 4892 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t) 4893 * zc_cookie zap cursor 4894 */ 4895static int 4896zfs_ioc_userspace_many(zfs_cmd_t *zc) 4897{ 4898 zfsvfs_t *zfsvfs; 4899 int bufsize = zc->zc_nvlist_dst_size; 4900 4901 if (bufsize <= 0) 4902 return (SET_ERROR(ENOMEM)); 4903 4904 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE); 4905 if (error != 0) 4906 return (error); 4907 4908 void *buf = kmem_alloc(bufsize, KM_SLEEP); 4909 4910 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie, 4911 buf, &zc->zc_nvlist_dst_size); 4912 4913 if (error == 0) { 4914 error = ddi_copyout(buf, 4915 (void *)(uintptr_t)zc->zc_nvlist_dst, 4916 zc->zc_nvlist_dst_size, zc->zc_iflags); 4917 } 4918 kmem_free(buf, bufsize); 4919 zfsvfs_rele(zfsvfs, FTAG); 4920 4921 return (error); 4922} 4923 4924/* 4925 * inputs: 4926 * zc_name name of filesystem 4927 * 4928 * outputs: 4929 * none 4930 */ 4931static int 4932zfs_ioc_userspace_upgrade(zfs_cmd_t *zc) 4933{ 4934 objset_t *os; 4935 int error = 0; 4936 zfsvfs_t *zfsvfs; 4937 4938 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) { 4939 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) { 4940 /* 4941 * If userused is not enabled, it may be because the 4942 * objset needs to be closed & reopened (to grow the 4943 * objset_phys_t). Suspend/resume the fs will do that. 4944 */ 4945 error = zfs_suspend_fs(zfsvfs); 4946 if (error == 0) { 4947 dmu_objset_refresh_ownership(zfsvfs->z_os, 4948 zfsvfs); 4949 error = zfs_resume_fs(zfsvfs, zc->zc_name); 4950 } 4951 } 4952 if (error == 0) 4953 error = dmu_objset_userspace_upgrade(zfsvfs->z_os); 4954#ifdef illumos 4955 VFS_RELE(zfsvfs->z_vfs); 4956#else 4957 vfs_unbusy(zfsvfs->z_vfs); 4958#endif 4959 } else { 4960 /* XXX kind of reading contents without owning */ 4961 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 4962 if (error != 0) 4963 return (error); 4964 4965 error = dmu_objset_userspace_upgrade(os); 4966 dmu_objset_rele(os, FTAG); 4967 } 4968 4969 return (error); 4970} 4971 4972#ifdef illumos 4973/* 4974 * We don't want to have a hard dependency 4975 * against some special symbols in sharefs 4976 * nfs, and smbsrv. Determine them if needed when 4977 * the first file system is shared. 4978 * Neither sharefs, nfs or smbsrv are unloadable modules. 4979 */ 4980int (*znfsexport_fs)(void *arg); 4981int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t); 4982int (*zsmbexport_fs)(void *arg, boolean_t add_share); 4983 4984int zfs_nfsshare_inited; 4985int zfs_smbshare_inited; 4986 4987ddi_modhandle_t nfs_mod; 4988ddi_modhandle_t sharefs_mod; 4989ddi_modhandle_t smbsrv_mod; 4990#endif /* illumos */ 4991kmutex_t zfs_share_lock; 4992 4993#ifdef illumos 4994static int 4995zfs_init_sharefs() 4996{ 4997 int error; 4998 4999 ASSERT(MUTEX_HELD(&zfs_share_lock)); 5000 /* Both NFS and SMB shares also require sharetab support. */ 5001 if (sharefs_mod == NULL && ((sharefs_mod = 5002 ddi_modopen("fs/sharefs", 5003 KRTLD_MODE_FIRST, &error)) == NULL)) { 5004 return (SET_ERROR(ENOSYS)); 5005 } 5006 if (zshare_fs == NULL && ((zshare_fs = 5007 (int (*)(enum sharefs_sys_op, share_t *, uint32_t)) 5008 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) { 5009 return (SET_ERROR(ENOSYS)); 5010 } 5011 return (0); 5012} 5013#endif /* illumos */ 5014 5015static int 5016zfs_ioc_share(zfs_cmd_t *zc) 5017{ 5018#ifdef illumos 5019 int error; 5020 int opcode; 5021 5022 switch (zc->zc_share.z_sharetype) { 5023 case ZFS_SHARE_NFS: 5024 case ZFS_UNSHARE_NFS: 5025 if (zfs_nfsshare_inited == 0) { 5026 mutex_enter(&zfs_share_lock); 5027 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs", 5028 KRTLD_MODE_FIRST, &error)) == NULL)) { 5029 mutex_exit(&zfs_share_lock); 5030 return (SET_ERROR(ENOSYS)); 5031 } 5032 if (znfsexport_fs == NULL && 5033 ((znfsexport_fs = (int (*)(void *)) 5034 ddi_modsym(nfs_mod, 5035 "nfs_export", &error)) == NULL)) { 5036 mutex_exit(&zfs_share_lock); 5037 return (SET_ERROR(ENOSYS)); 5038 } 5039 error = zfs_init_sharefs(); 5040 if (error != 0) { 5041 mutex_exit(&zfs_share_lock); 5042 return (SET_ERROR(ENOSYS)); 5043 } 5044 zfs_nfsshare_inited = 1; 5045 mutex_exit(&zfs_share_lock); 5046 } 5047 break; 5048 case ZFS_SHARE_SMB: 5049 case ZFS_UNSHARE_SMB: 5050 if (zfs_smbshare_inited == 0) { 5051 mutex_enter(&zfs_share_lock); 5052 if (smbsrv_mod == NULL && ((smbsrv_mod = 5053 ddi_modopen("drv/smbsrv", 5054 KRTLD_MODE_FIRST, &error)) == NULL)) { 5055 mutex_exit(&zfs_share_lock); 5056 return (SET_ERROR(ENOSYS)); 5057 } 5058 if (zsmbexport_fs == NULL && ((zsmbexport_fs = 5059 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod, 5060 "smb_server_share", &error)) == NULL)) { 5061 mutex_exit(&zfs_share_lock); 5062 return (SET_ERROR(ENOSYS)); 5063 } 5064 error = zfs_init_sharefs(); 5065 if (error != 0) { 5066 mutex_exit(&zfs_share_lock); 5067 return (SET_ERROR(ENOSYS)); 5068 } 5069 zfs_smbshare_inited = 1; 5070 mutex_exit(&zfs_share_lock); 5071 } 5072 break; 5073 default: 5074 return (SET_ERROR(EINVAL)); 5075 } 5076 5077 switch (zc->zc_share.z_sharetype) { 5078 case ZFS_SHARE_NFS: 5079 case ZFS_UNSHARE_NFS: 5080 if (error = 5081 znfsexport_fs((void *) 5082 (uintptr_t)zc->zc_share.z_exportdata)) 5083 return (error); 5084 break; 5085 case ZFS_SHARE_SMB: 5086 case ZFS_UNSHARE_SMB: 5087 if (error = zsmbexport_fs((void *) 5088 (uintptr_t)zc->zc_share.z_exportdata, 5089 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ? 5090 B_TRUE: B_FALSE)) { 5091 return (error); 5092 } 5093 break; 5094 } 5095 5096 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS || 5097 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ? 5098 SHAREFS_ADD : SHAREFS_REMOVE; 5099 5100 /* 5101 * Add or remove share from sharetab 5102 */ 5103 error = zshare_fs(opcode, 5104 (void *)(uintptr_t)zc->zc_share.z_sharedata, 5105 zc->zc_share.z_sharemax); 5106 5107 return (error); 5108 5109#else /* !illumos */ 5110 return (ENOSYS); 5111#endif /* illumos */ 5112} 5113 5114ace_t full_access[] = { 5115 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0} 5116}; 5117 5118/* 5119 * inputs: 5120 * zc_name name of containing filesystem 5121 * zc_obj object # beyond which we want next in-use object # 5122 * 5123 * outputs: 5124 * zc_obj next in-use object # 5125 */ 5126static int 5127zfs_ioc_next_obj(zfs_cmd_t *zc) 5128{ 5129 objset_t *os = NULL; 5130 int error; 5131 5132 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 5133 if (error != 0) 5134 return (error); 5135 5136 error = dmu_object_next(os, &zc->zc_obj, B_FALSE, 5137 dsl_dataset_phys(os->os_dsl_dataset)->ds_prev_snap_txg); 5138 5139 dmu_objset_rele(os, FTAG); 5140 return (error); 5141} 5142 5143/* 5144 * inputs: 5145 * zc_name name of filesystem 5146 * zc_value prefix name for snapshot 5147 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process 5148 * 5149 * outputs: 5150 * zc_value short name of new snapshot 5151 */ 5152static int 5153zfs_ioc_tmp_snapshot(zfs_cmd_t *zc) 5154{ 5155 char *snap_name; 5156 char *hold_name; 5157 int error; 5158 minor_t minor; 5159 5160 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor); 5161 if (error != 0) 5162 return (error); 5163 5164 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value, 5165 (u_longlong_t)ddi_get_lbolt64()); 5166 hold_name = kmem_asprintf("%%%s", zc->zc_value); 5167 5168 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor, 5169 hold_name); 5170 if (error == 0) 5171 (void) strcpy(zc->zc_value, snap_name); 5172 strfree(snap_name); 5173 strfree(hold_name); 5174 zfs_onexit_fd_rele(zc->zc_cleanup_fd); 5175 return (error); 5176} 5177 5178/* 5179 * inputs: 5180 * zc_name name of "to" snapshot 5181 * zc_value name of "from" snapshot 5182 * zc_cookie file descriptor to write diff data on 5183 * 5184 * outputs: 5185 * dmu_diff_record_t's to the file descriptor 5186 */ 5187static int 5188zfs_ioc_diff(zfs_cmd_t *zc) 5189{ 5190 file_t *fp; 5191 cap_rights_t rights; 5192 offset_t off; 5193 int error; 5194 5195#ifdef illumos 5196 fp = getf(zc->zc_cookie); 5197#else 5198 fget_write(curthread, zc->zc_cookie, 5199 cap_rights_init(&rights, CAP_WRITE), &fp); 5200#endif 5201 if (fp == NULL) 5202 return (SET_ERROR(EBADF)); 5203 5204 off = fp->f_offset; 5205 5206#ifdef illumos 5207 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off); 5208#else 5209 error = dmu_diff(zc->zc_name, zc->zc_value, fp, &off); 5210#endif 5211 5212 if (off >= 0 && off <= MAXOFFSET_T) 5213 fp->f_offset = off; 5214 releasef(zc->zc_cookie); 5215 5216 return (error); 5217} 5218 5219#ifdef illumos 5220/* 5221 * Remove all ACL files in shares dir 5222 */ 5223static int 5224zfs_smb_acl_purge(znode_t *dzp) 5225{ 5226 zap_cursor_t zc; 5227 zap_attribute_t zap; 5228 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 5229 int error; 5230 5231 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id); 5232 (error = zap_cursor_retrieve(&zc, &zap)) == 0; 5233 zap_cursor_advance(&zc)) { 5234 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred, 5235 NULL, 0)) != 0) 5236 break; 5237 } 5238 zap_cursor_fini(&zc); 5239 return (error); 5240} 5241#endif /* illumos */ 5242 5243static int 5244zfs_ioc_smb_acl(zfs_cmd_t *zc) 5245{ 5246#ifdef illumos 5247 vnode_t *vp; 5248 znode_t *dzp; 5249 vnode_t *resourcevp = NULL; 5250 znode_t *sharedir; 5251 zfsvfs_t *zfsvfs; 5252 nvlist_t *nvlist; 5253 char *src, *target; 5254 vattr_t vattr; 5255 vsecattr_t vsec; 5256 int error = 0; 5257 5258 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, 5259 NO_FOLLOW, NULL, &vp)) != 0) 5260 return (error); 5261 5262 /* Now make sure mntpnt and dataset are ZFS */ 5263 5264 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 || 5265 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), 5266 zc->zc_name) != 0)) { 5267 VN_RELE(vp); 5268 return (SET_ERROR(EINVAL)); 5269 } 5270 5271 dzp = VTOZ(vp); 5272 zfsvfs = dzp->z_zfsvfs; 5273 ZFS_ENTER(zfsvfs); 5274 5275 /* 5276 * Create share dir if its missing. 5277 */ 5278 mutex_enter(&zfsvfs->z_lock); 5279 if (zfsvfs->z_shares_dir == 0) { 5280 dmu_tx_t *tx; 5281 5282 tx = dmu_tx_create(zfsvfs->z_os); 5283 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE, 5284 ZFS_SHARES_DIR); 5285 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); 5286 error = dmu_tx_assign(tx, TXG_WAIT); 5287 if (error != 0) { 5288 dmu_tx_abort(tx); 5289 } else { 5290 error = zfs_create_share_dir(zfsvfs, tx); 5291 dmu_tx_commit(tx); 5292 } 5293 if (error != 0) { 5294 mutex_exit(&zfsvfs->z_lock); 5295 VN_RELE(vp); 5296 ZFS_EXIT(zfsvfs); 5297 return (error); 5298 } 5299 } 5300 mutex_exit(&zfsvfs->z_lock); 5301 5302 ASSERT(zfsvfs->z_shares_dir); 5303 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) { 5304 VN_RELE(vp); 5305 ZFS_EXIT(zfsvfs); 5306 return (error); 5307 } 5308 5309 switch (zc->zc_cookie) { 5310 case ZFS_SMB_ACL_ADD: 5311 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; 5312 vattr.va_type = VREG; 5313 vattr.va_mode = S_IFREG|0777; 5314 vattr.va_uid = 0; 5315 vattr.va_gid = 0; 5316 5317 vsec.vsa_mask = VSA_ACE; 5318 vsec.vsa_aclentp = &full_access; 5319 vsec.vsa_aclentsz = sizeof (full_access); 5320 vsec.vsa_aclcnt = 1; 5321 5322 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string, 5323 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec); 5324 if (resourcevp) 5325 VN_RELE(resourcevp); 5326 break; 5327 5328 case ZFS_SMB_ACL_REMOVE: 5329 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred, 5330 NULL, 0); 5331 break; 5332 5333 case ZFS_SMB_ACL_RENAME: 5334 if ((error = get_nvlist(zc->zc_nvlist_src, 5335 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) { 5336 VN_RELE(vp); 5337 VN_RELE(ZTOV(sharedir)); 5338 ZFS_EXIT(zfsvfs); 5339 return (error); 5340 } 5341 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) || 5342 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET, 5343 &target)) { 5344 VN_RELE(vp); 5345 VN_RELE(ZTOV(sharedir)); 5346 ZFS_EXIT(zfsvfs); 5347 nvlist_free(nvlist); 5348 return (error); 5349 } 5350 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target, 5351 kcred, NULL, 0); 5352 nvlist_free(nvlist); 5353 break; 5354 5355 case ZFS_SMB_ACL_PURGE: 5356 error = zfs_smb_acl_purge(sharedir); 5357 break; 5358 5359 default: 5360 error = SET_ERROR(EINVAL); 5361 break; 5362 } 5363 5364 VN_RELE(vp); 5365 VN_RELE(ZTOV(sharedir)); 5366 5367 ZFS_EXIT(zfsvfs); 5368 5369 return (error); 5370#else /* !illumos */ 5371 return (EOPNOTSUPP); 5372#endif /* illumos */ 5373} 5374 5375/* 5376 * innvl: { 5377 * "holds" -> { snapname -> holdname (string), ... } 5378 * (optional) "cleanup_fd" -> fd (int32) 5379 * } 5380 * 5381 * outnvl: { 5382 * snapname -> error value (int32) 5383 * ... 5384 * } 5385 */ 5386/* ARGSUSED */ 5387static int 5388zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist) 5389{ 5390 nvpair_t *pair; 5391 nvlist_t *holds; 5392 int cleanup_fd = -1; 5393 int error; 5394 minor_t minor = 0; 5395 5396 error = nvlist_lookup_nvlist(args, "holds", &holds); 5397 if (error != 0) 5398 return (SET_ERROR(EINVAL)); 5399 5400 /* make sure the user didn't pass us any invalid (empty) tags */ 5401 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL; 5402 pair = nvlist_next_nvpair(holds, pair)) { 5403 char *htag; 5404 5405 error = nvpair_value_string(pair, &htag); 5406 if (error != 0) 5407 return (SET_ERROR(error)); 5408 5409 if (strlen(htag) == 0) 5410 return (SET_ERROR(EINVAL)); 5411 } 5412 5413 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) { 5414 error = zfs_onexit_fd_hold(cleanup_fd, &minor); 5415 if (error != 0) 5416 return (error); 5417 } 5418 5419 error = dsl_dataset_user_hold(holds, minor, errlist); 5420 if (minor != 0) 5421 zfs_onexit_fd_rele(cleanup_fd); 5422 return (error); 5423} 5424 5425/* 5426 * innvl is not used. 5427 * 5428 * outnvl: { 5429 * holdname -> time added (uint64 seconds since epoch) 5430 * ... 5431 * } 5432 */ 5433/* ARGSUSED */ 5434static int 5435zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl) 5436{ 5437 return (dsl_dataset_get_holds(snapname, outnvl)); 5438} 5439 5440/* 5441 * innvl: { 5442 * snapname -> { holdname, ... } 5443 * ... 5444 * } 5445 * 5446 * outnvl: { 5447 * snapname -> error value (int32) 5448 * ... 5449 * } 5450 */ 5451/* ARGSUSED */ 5452static int 5453zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist) 5454{ 5455 return (dsl_dataset_user_release(holds, errlist)); 5456} 5457 5458/* 5459 * inputs: 5460 * zc_name name of new filesystem or snapshot 5461 * zc_value full name of old snapshot 5462 * 5463 * outputs: 5464 * zc_cookie space in bytes 5465 * zc_objset_type compressed space in bytes 5466 * zc_perm_action uncompressed space in bytes 5467 */ 5468static int 5469zfs_ioc_space_written(zfs_cmd_t *zc) 5470{ 5471 int error; 5472 dsl_pool_t *dp; 5473 dsl_dataset_t *new, *old; 5474 5475 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 5476 if (error != 0) 5477 return (error); 5478 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new); 5479 if (error != 0) { 5480 dsl_pool_rele(dp, FTAG); 5481 return (error); 5482 } 5483 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old); 5484 if (error != 0) { 5485 dsl_dataset_rele(new, FTAG); 5486 dsl_pool_rele(dp, FTAG); 5487 return (error); 5488 } 5489 5490 error = dsl_dataset_space_written(old, new, &zc->zc_cookie, 5491 &zc->zc_objset_type, &zc->zc_perm_action); 5492 dsl_dataset_rele(old, FTAG); 5493 dsl_dataset_rele(new, FTAG); 5494 dsl_pool_rele(dp, FTAG); 5495 return (error); 5496} 5497 5498/* 5499 * innvl: { 5500 * "firstsnap" -> snapshot name 5501 * } 5502 * 5503 * outnvl: { 5504 * "used" -> space in bytes 5505 * "compressed" -> compressed space in bytes 5506 * "uncompressed" -> uncompressed space in bytes 5507 * } 5508 */ 5509static int 5510zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl) 5511{ 5512 int error; 5513 dsl_pool_t *dp; 5514 dsl_dataset_t *new, *old; 5515 char *firstsnap; 5516 uint64_t used, comp, uncomp; 5517 5518 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0) 5519 return (SET_ERROR(EINVAL)); 5520 5521 error = dsl_pool_hold(lastsnap, FTAG, &dp); 5522 if (error != 0) 5523 return (error); 5524 5525 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new); 5526 if (error == 0 && !new->ds_is_snapshot) { 5527 dsl_dataset_rele(new, FTAG); 5528 error = SET_ERROR(EINVAL); 5529 } 5530 if (error != 0) { 5531 dsl_pool_rele(dp, FTAG); 5532 return (error); 5533 } 5534 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old); 5535 if (error == 0 && !old->ds_is_snapshot) { 5536 dsl_dataset_rele(old, FTAG); 5537 error = SET_ERROR(EINVAL); 5538 } 5539 if (error != 0) { 5540 dsl_dataset_rele(new, FTAG); 5541 dsl_pool_rele(dp, FTAG); 5542 return (error); 5543 } 5544 5545 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp); 5546 dsl_dataset_rele(old, FTAG); 5547 dsl_dataset_rele(new, FTAG); 5548 dsl_pool_rele(dp, FTAG); 5549 fnvlist_add_uint64(outnvl, "used", used); 5550 fnvlist_add_uint64(outnvl, "compressed", comp); 5551 fnvlist_add_uint64(outnvl, "uncompressed", uncomp); 5552 return (error); 5553} 5554 5555static int 5556zfs_ioc_jail(zfs_cmd_t *zc) 5557{ 5558 5559 return (zone_dataset_attach(curthread->td_ucred, zc->zc_name, 5560 (int)zc->zc_jailid)); 5561} 5562 5563static int 5564zfs_ioc_unjail(zfs_cmd_t *zc) 5565{ 5566 5567 return (zone_dataset_detach(curthread->td_ucred, zc->zc_name, 5568 (int)zc->zc_jailid)); 5569} 5570 5571/* 5572 * innvl: { 5573 * "fd" -> file descriptor to write stream to (int32) 5574 * (optional) "fromsnap" -> full snap name to send an incremental from 5575 * (optional) "largeblockok" -> (value ignored) 5576 * indicates that blocks > 128KB are permitted 5577 * (optional) "embedok" -> (value ignored) 5578 * presence indicates DRR_WRITE_EMBEDDED records are permitted 5579 * (optional) "resume_object" and "resume_offset" -> (uint64) 5580 * if present, resume send stream from specified object and offset. 5581 * } 5582 * 5583 * outnvl is unused 5584 */ 5585/* ARGSUSED */ 5586static int 5587zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl) 5588{ 5589 cap_rights_t rights; 5590 file_t *fp; 5591 int error; 5592 offset_t off; 5593 char *fromname = NULL; 5594 int fd; 5595 boolean_t largeblockok; 5596 boolean_t embedok; 5597 uint64_t resumeobj = 0; 5598 uint64_t resumeoff = 0; 5599 5600 error = nvlist_lookup_int32(innvl, "fd", &fd); 5601 if (error != 0) 5602 return (SET_ERROR(EINVAL)); 5603 5604 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname); 5605 5606 largeblockok = nvlist_exists(innvl, "largeblockok"); 5607 embedok = nvlist_exists(innvl, "embedok"); 5608 5609 (void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj); 5610 (void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff); 5611 5612#ifdef illumos 5613 file_t *fp = getf(fd); 5614#else 5615 fget_write(curthread, fd, cap_rights_init(&rights, CAP_WRITE), &fp); 5616#endif 5617 if (fp == NULL) 5618 return (SET_ERROR(EBADF)); 5619 5620 off = fp->f_offset; 5621 error = dmu_send(snapname, fromname, embedok, largeblockok, fd, 5622#ifdef illumos 5623 resumeobj, resumeoff, fp->f_vnode, &off); 5624#else 5625 resumeobj, resumeoff, fp, &off); 5626#endif 5627 5628#ifdef illumos 5629 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0) 5630 fp->f_offset = off; 5631#else 5632 fp->f_offset = off; 5633#endif 5634 5635 releasef(fd); 5636 return (error); 5637} 5638 5639/* 5640 * Determine approximately how large a zfs send stream will be -- the number 5641 * of bytes that will be written to the fd supplied to zfs_ioc_send_new(). 5642 * 5643 * innvl: { 5644 * (optional) "from" -> full snap or bookmark name to send an incremental 5645 * from 5646 * } 5647 * 5648 * outnvl: { 5649 * "space" -> bytes of space (uint64) 5650 * } 5651 */ 5652static int 5653zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl) 5654{ 5655 dsl_pool_t *dp; 5656 dsl_dataset_t *tosnap; 5657 int error; 5658 char *fromname; 5659 uint64_t space; 5660 5661 error = dsl_pool_hold(snapname, FTAG, &dp); 5662 if (error != 0) 5663 return (error); 5664 5665 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap); 5666 if (error != 0) { 5667 dsl_pool_rele(dp, FTAG); 5668 return (error); 5669 } 5670 5671 error = nvlist_lookup_string(innvl, "from", &fromname); 5672 if (error == 0) { 5673 if (strchr(fromname, '@') != NULL) { 5674 /* 5675 * If from is a snapshot, hold it and use the more 5676 * efficient dmu_send_estimate to estimate send space 5677 * size using deadlists. 5678 */ 5679 dsl_dataset_t *fromsnap; 5680 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap); 5681 if (error != 0) 5682 goto out; 5683 error = dmu_send_estimate(tosnap, fromsnap, &space); 5684 dsl_dataset_rele(fromsnap, FTAG); 5685 } else if (strchr(fromname, '#') != NULL) { 5686 /* 5687 * If from is a bookmark, fetch the creation TXG of the 5688 * snapshot it was created from and use that to find 5689 * blocks that were born after it. 5690 */ 5691 zfs_bookmark_phys_t frombm; 5692 5693 error = dsl_bookmark_lookup(dp, fromname, tosnap, 5694 &frombm); 5695 if (error != 0) 5696 goto out; 5697 error = dmu_send_estimate_from_txg(tosnap, 5698 frombm.zbm_creation_txg, &space); 5699 } else { 5700 /* 5701 * from is not properly formatted as a snapshot or 5702 * bookmark 5703 */ 5704 error = SET_ERROR(EINVAL); 5705 goto out; 5706 } 5707 } else { 5708 // If estimating the size of a full send, use dmu_send_estimate 5709 error = dmu_send_estimate(tosnap, NULL, &space); 5710 } 5711 5712 fnvlist_add_uint64(outnvl, "space", space); 5713 5714out: 5715 dsl_dataset_rele(tosnap, FTAG); 5716 dsl_pool_rele(dp, FTAG); 5717 return (error); 5718} 5719 5720static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST]; 5721 5722static void 5723zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5724 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck, 5725 boolean_t log_history, zfs_ioc_poolcheck_t pool_check) 5726{ 5727 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST]; 5728 5729 ASSERT3U(ioc, >=, ZFS_IOC_FIRST); 5730 ASSERT3U(ioc, <, ZFS_IOC_LAST); 5731 ASSERT3P(vec->zvec_legacy_func, ==, NULL); 5732 ASSERT3P(vec->zvec_func, ==, NULL); 5733 5734 vec->zvec_legacy_func = func; 5735 vec->zvec_secpolicy = secpolicy; 5736 vec->zvec_namecheck = namecheck; 5737 vec->zvec_allow_log = log_history; 5738 vec->zvec_pool_check = pool_check; 5739} 5740 5741/* 5742 * See the block comment at the beginning of this file for details on 5743 * each argument to this function. 5744 */ 5745static void 5746zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func, 5747 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck, 5748 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist, 5749 boolean_t allow_log) 5750{ 5751 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST]; 5752 5753 ASSERT3U(ioc, >=, ZFS_IOC_FIRST); 5754 ASSERT3U(ioc, <, ZFS_IOC_LAST); 5755 ASSERT3P(vec->zvec_legacy_func, ==, NULL); 5756 ASSERT3P(vec->zvec_func, ==, NULL); 5757 5758 /* if we are logging, the name must be valid */ 5759 ASSERT(!allow_log || namecheck != NO_NAME); 5760 5761 vec->zvec_name = name; 5762 vec->zvec_func = func; 5763 vec->zvec_secpolicy = secpolicy; 5764 vec->zvec_namecheck = namecheck; 5765 vec->zvec_pool_check = pool_check; 5766 vec->zvec_smush_outnvlist = smush_outnvlist; 5767 vec->zvec_allow_log = allow_log; 5768} 5769 5770static void 5771zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5772 zfs_secpolicy_func_t *secpolicy, boolean_t log_history, 5773 zfs_ioc_poolcheck_t pool_check) 5774{ 5775 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5776 POOL_NAME, log_history, pool_check); 5777} 5778 5779static void 5780zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5781 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check) 5782{ 5783 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5784 DATASET_NAME, B_FALSE, pool_check); 5785} 5786 5787static void 5788zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func) 5789{ 5790 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config, 5791 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5792} 5793 5794static void 5795zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5796 zfs_secpolicy_func_t *secpolicy) 5797{ 5798 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5799 NO_NAME, B_FALSE, POOL_CHECK_NONE); 5800} 5801 5802static void 5803zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc, 5804 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy) 5805{ 5806 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5807 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED); 5808} 5809 5810static void 5811zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func) 5812{ 5813 zfs_ioctl_register_dataset_read_secpolicy(ioc, func, 5814 zfs_secpolicy_read); 5815} 5816 5817static void 5818zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5819 zfs_secpolicy_func_t *secpolicy) 5820{ 5821 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5822 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5823} 5824 5825static void 5826zfs_ioctl_init(void) 5827{ 5828 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT, 5829 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME, 5830 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5831 5832 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY, 5833 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME, 5834 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE); 5835 5836 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS, 5837 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME, 5838 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5839 5840 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW, 5841 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME, 5842 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5843 5844 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE, 5845 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME, 5846 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5847 5848 zfs_ioctl_register("create", ZFS_IOC_CREATE, 5849 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME, 5850 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5851 5852 zfs_ioctl_register("clone", ZFS_IOC_CLONE, 5853 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME, 5854 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5855 5856 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS, 5857 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME, 5858 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5859 5860 zfs_ioctl_register("hold", ZFS_IOC_HOLD, 5861 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME, 5862 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5863 zfs_ioctl_register("release", ZFS_IOC_RELEASE, 5864 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME, 5865 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5866 5867 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS, 5868 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME, 5869 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5870 5871 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK, 5872 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, 5873 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE); 5874 5875 zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK, 5876 zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME, 5877 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5878 5879 zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS, 5880 zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME, 5881 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5882 5883 zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS, 5884 zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks, 5885 POOL_NAME, 5886 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5887 5888 /* IOCTLS that use the legacy function signature */ 5889 5890 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze, 5891 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY); 5892 5893 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create, 5894 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5895 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN, 5896 zfs_ioc_pool_scan); 5897 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE, 5898 zfs_ioc_pool_upgrade); 5899 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD, 5900 zfs_ioc_vdev_add); 5901 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE, 5902 zfs_ioc_vdev_remove); 5903 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE, 5904 zfs_ioc_vdev_set_state); 5905 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH, 5906 zfs_ioc_vdev_attach); 5907 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH, 5908 zfs_ioc_vdev_detach); 5909 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH, 5910 zfs_ioc_vdev_setpath); 5911 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU, 5912 zfs_ioc_vdev_setfru); 5913 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS, 5914 zfs_ioc_pool_set_props); 5915 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT, 5916 zfs_ioc_vdev_split); 5917 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID, 5918 zfs_ioc_pool_reguid); 5919 5920 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS, 5921 zfs_ioc_pool_configs, zfs_secpolicy_none); 5922 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT, 5923 zfs_ioc_pool_tryimport, zfs_secpolicy_config); 5924 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT, 5925 zfs_ioc_inject_fault, zfs_secpolicy_inject); 5926 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT, 5927 zfs_ioc_clear_fault, zfs_secpolicy_inject); 5928 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT, 5929 zfs_ioc_inject_list_next, zfs_secpolicy_inject); 5930 5931 /* 5932 * pool destroy, and export don't log the history as part of 5933 * zfsdev_ioctl, but rather zfs_ioc_pool_export 5934 * does the logging of those commands. 5935 */ 5936 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy, 5937 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE); 5938 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export, 5939 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE); 5940 5941 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats, 5942 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE); 5943 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props, 5944 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE); 5945 5946 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log, 5947 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_NONE); 5948 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME, 5949 zfs_ioc_dsobj_to_dsname, 5950 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_NONE); 5951 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY, 5952 zfs_ioc_pool_get_history, 5953 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED); 5954 5955 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import, 5956 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5957 5958 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear, 5959 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5960 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen, 5961 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED); 5962 5963 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN, 5964 zfs_ioc_space_written); 5965 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS, 5966 zfs_ioc_objset_recvd_props); 5967 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ, 5968 zfs_ioc_next_obj); 5969 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL, 5970 zfs_ioc_get_fsacl); 5971 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS, 5972 zfs_ioc_objset_stats); 5973 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS, 5974 zfs_ioc_objset_zplprops); 5975 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT, 5976 zfs_ioc_dataset_list_next); 5977 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT, 5978 zfs_ioc_snapshot_list_next); 5979 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS, 5980 zfs_ioc_send_progress); 5981 5982 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF, 5983 zfs_ioc_diff, zfs_secpolicy_diff); 5984 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS, 5985 zfs_ioc_obj_to_stats, zfs_secpolicy_diff); 5986 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH, 5987 zfs_ioc_obj_to_path, zfs_secpolicy_diff); 5988 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE, 5989 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one); 5990 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY, 5991 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many); 5992 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND, 5993 zfs_ioc_send, zfs_secpolicy_send); 5994 5995 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop, 5996 zfs_secpolicy_none); 5997 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy, 5998 zfs_secpolicy_destroy); 5999 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename, 6000 zfs_secpolicy_rename); 6001 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv, 6002 zfs_secpolicy_recv); 6003 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote, 6004 zfs_secpolicy_promote); 6005 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP, 6006 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop); 6007 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl, 6008 zfs_secpolicy_set_fsacl); 6009 6010 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share, 6011 zfs_secpolicy_share, POOL_CHECK_NONE); 6012 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl, 6013 zfs_secpolicy_smb_acl, POOL_CHECK_NONE); 6014 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE, 6015 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade, 6016 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 6017 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT, 6018 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot, 6019 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 6020 6021#ifdef __FreeBSD__ 6022 zfs_ioctl_register_dataset_nolog(ZFS_IOC_JAIL, zfs_ioc_jail, 6023 zfs_secpolicy_config, POOL_CHECK_NONE); 6024 zfs_ioctl_register_dataset_nolog(ZFS_IOC_UNJAIL, zfs_ioc_unjail, 6025 zfs_secpolicy_config, POOL_CHECK_NONE); 6026#endif 6027} 6028 6029int 6030pool_status_check(const char *name, zfs_ioc_namecheck_t type, 6031 zfs_ioc_poolcheck_t check) 6032{ 6033 spa_t *spa; 6034 int error; 6035 6036 ASSERT(type == POOL_NAME || type == DATASET_NAME); 6037 6038 if (check & POOL_CHECK_NONE) 6039 return (0); 6040 6041 error = spa_open(name, &spa, FTAG); 6042 if (error == 0) { 6043 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa)) 6044 error = SET_ERROR(EAGAIN); 6045 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa)) 6046 error = SET_ERROR(EROFS); 6047 spa_close(spa, FTAG); 6048 } 6049 return (error); 6050} 6051 6052/* 6053 * Find a free minor number. 6054 */ 6055minor_t 6056zfsdev_minor_alloc(void) 6057{ 6058 static minor_t last_minor; 6059 minor_t m; 6060 6061 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 6062 6063 for (m = last_minor + 1; m != last_minor; m++) { 6064 if (m > ZFSDEV_MAX_MINOR) 6065 m = 1; 6066 if (ddi_get_soft_state(zfsdev_state, m) == NULL) { 6067 last_minor = m; 6068 return (m); 6069 } 6070 } 6071 6072 return (0); 6073} 6074 6075static int 6076zfs_ctldev_init(struct cdev *devp) 6077{ 6078 minor_t minor; 6079 zfs_soft_state_t *zs; 6080 6081 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 6082 6083 minor = zfsdev_minor_alloc(); 6084 if (minor == 0) 6085 return (SET_ERROR(ENXIO)); 6086 6087 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS) 6088 return (SET_ERROR(EAGAIN)); 6089 6090 devfs_set_cdevpriv((void *)(uintptr_t)minor, zfsdev_close); 6091 6092 zs = ddi_get_soft_state(zfsdev_state, minor); 6093 zs->zss_type = ZSST_CTLDEV; 6094 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data); 6095 6096 return (0); 6097} 6098 6099static void 6100zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor) 6101{ 6102 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 6103 6104 zfs_onexit_destroy(zo); 6105 ddi_soft_state_free(zfsdev_state, minor); 6106} 6107 6108void * 6109zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which) 6110{ 6111 zfs_soft_state_t *zp; 6112 6113 zp = ddi_get_soft_state(zfsdev_state, minor); 6114 if (zp == NULL || zp->zss_type != which) 6115 return (NULL); 6116 6117 return (zp->zss_data); 6118} 6119 6120static int 6121zfsdev_open(struct cdev *devp, int flag, int mode, struct thread *td) 6122{ 6123 int error = 0; 6124 6125#ifdef illumos 6126 if (getminor(*devp) != 0) 6127 return (zvol_open(devp, flag, otyp, cr)); 6128#endif 6129 6130 /* This is the control device. Allocate a new minor if requested. */ 6131 if (flag & FEXCL) { 6132 mutex_enter(&spa_namespace_lock); 6133 error = zfs_ctldev_init(devp); 6134 mutex_exit(&spa_namespace_lock); 6135 } 6136 6137 return (error); 6138} 6139 6140static void 6141zfsdev_close(void *data) 6142{ 6143 zfs_onexit_t *zo; 6144 minor_t minor = (minor_t)(uintptr_t)data; 6145 6146 if (minor == 0) 6147 return; 6148 6149 mutex_enter(&spa_namespace_lock); 6150 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV); 6151 if (zo == NULL) { 6152 mutex_exit(&spa_namespace_lock); 6153 return; 6154 } 6155 zfs_ctldev_destroy(zo, minor); 6156 mutex_exit(&spa_namespace_lock); 6157} 6158 6159static int 6160zfsdev_ioctl(struct cdev *dev, u_long zcmd, caddr_t arg, int flag, 6161 struct thread *td) 6162{ 6163 zfs_cmd_t *zc; 6164 uint_t vecnum; 6165 int error, rc, len; 6166#ifdef illumos 6167 minor_t minor = getminor(dev); 6168#else 6169 zfs_iocparm_t *zc_iocparm; 6170 int cflag, cmd, oldvecnum; 6171 boolean_t newioc, compat; 6172 void *compat_zc = NULL; 6173 cred_t *cr = td->td_ucred; 6174#endif 6175 const zfs_ioc_vec_t *vec; 6176 char *saved_poolname = NULL; 6177 nvlist_t *innvl = NULL; 6178 6179 cflag = ZFS_CMD_COMPAT_NONE; 6180 compat = B_FALSE; 6181 newioc = B_TRUE; /* "new" style (zfs_iocparm_t) ioctl */ 6182 6183 len = IOCPARM_LEN(zcmd); 6184 vecnum = cmd = zcmd & 0xff; 6185 6186 /* 6187 * Check if we are talking to supported older binaries 6188 * and translate zfs_cmd if necessary 6189 */ 6190 if (len != sizeof(zfs_iocparm_t)) { 6191 newioc = B_FALSE; 6192 compat = B_TRUE; 6193 6194 vecnum = cmd; 6195 6196 switch (len) { 6197 case sizeof(zfs_cmd_zcmd_t): 6198 cflag = ZFS_CMD_COMPAT_LZC; 6199 break; 6200 case sizeof(zfs_cmd_deadman_t): 6201 cflag = ZFS_CMD_COMPAT_DEADMAN; 6202 break; 6203 case sizeof(zfs_cmd_v28_t): 6204 cflag = ZFS_CMD_COMPAT_V28; 6205 break; 6206 case sizeof(zfs_cmd_v15_t): 6207 cflag = ZFS_CMD_COMPAT_V15; 6208 vecnum = zfs_ioctl_v15_to_v28[cmd]; 6209 6210 /* 6211 * Return without further handling 6212 * if the command is blacklisted. 6213 */ 6214 if (vecnum == ZFS_IOC_COMPAT_PASS) 6215 return (0); 6216 else if (vecnum == ZFS_IOC_COMPAT_FAIL) 6217 return (ENOTSUP); 6218 break; 6219 default: 6220 return (EINVAL); 6221 } 6222 } 6223 6224#ifdef illumos 6225 vecnum = cmd - ZFS_IOC_FIRST; 6226 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip)); 6227#endif 6228 6229 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0])) 6230 return (SET_ERROR(EINVAL)); 6231 vec = &zfs_ioc_vec[vecnum]; 6232 6233 zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP); 6234 6235#ifdef illumos 6236 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag); 6237 if (error != 0) { 6238 error = SET_ERROR(EFAULT); 6239 goto out; 6240 } 6241#else /* !illumos */ 6242 bzero(zc, sizeof(zfs_cmd_t)); 6243 6244 if (newioc) { 6245 zc_iocparm = (void *)arg; 6246 6247 switch (zc_iocparm->zfs_ioctl_version) { 6248 case ZFS_IOCVER_CURRENT: 6249 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_t)) { 6250 error = SET_ERROR(EINVAL); 6251 goto out; 6252 } 6253 break; 6254 case ZFS_IOCVER_RESUME: 6255 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_resume_t)) { 6256 error = SET_ERROR(EFAULT); 6257 goto out; 6258 } 6259 compat = B_TRUE; 6260 cflag = ZFS_CMD_COMPAT_RESUME; 6261 break; 6262 case ZFS_IOCVER_EDBP: 6263 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_edbp_t)) { 6264 error = SET_ERROR(EFAULT); 6265 goto out; 6266 } 6267 compat = B_TRUE; 6268 cflag = ZFS_CMD_COMPAT_EDBP; 6269 break; 6270 case ZFS_IOCVER_ZCMD: 6271 if (zc_iocparm->zfs_cmd_size > sizeof(zfs_cmd_t) || 6272 zc_iocparm->zfs_cmd_size < sizeof(zfs_cmd_zcmd_t)) { 6273 error = SET_ERROR(EFAULT); 6274 goto out; 6275 } 6276 compat = B_TRUE; 6277 cflag = ZFS_CMD_COMPAT_ZCMD; 6278 break; 6279 default: 6280 error = SET_ERROR(EINVAL); 6281 goto out; 6282 /* NOTREACHED */ 6283 } 6284 6285 if (compat) { 6286 ASSERT(sizeof(zfs_cmd_t) >= zc_iocparm->zfs_cmd_size); 6287 compat_zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP); 6288 bzero(compat_zc, sizeof(zfs_cmd_t)); 6289 6290 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd, 6291 compat_zc, zc_iocparm->zfs_cmd_size, flag); 6292 if (error != 0) { 6293 error = SET_ERROR(EFAULT); 6294 goto out; 6295 } 6296 } else { 6297 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd, 6298 zc, zc_iocparm->zfs_cmd_size, flag); 6299 if (error != 0) { 6300 error = SET_ERROR(EFAULT); 6301 goto out; 6302 } 6303 } 6304 } 6305 6306 if (compat) { 6307 if (newioc) { 6308 ASSERT(compat_zc != NULL); 6309 zfs_cmd_compat_get(zc, compat_zc, cflag); 6310 } else { 6311 ASSERT(compat_zc == NULL); 6312 zfs_cmd_compat_get(zc, arg, cflag); 6313 } 6314 oldvecnum = vecnum; 6315 error = zfs_ioctl_compat_pre(zc, &vecnum, cflag); 6316 if (error != 0) 6317 goto out; 6318 if (oldvecnum != vecnum) 6319 vec = &zfs_ioc_vec[vecnum]; 6320 } 6321#endif /* !illumos */ 6322 6323 zc->zc_iflags = flag & FKIOCTL; 6324 if (zc->zc_nvlist_src_size != 0) { 6325 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 6326 zc->zc_iflags, &innvl); 6327 if (error != 0) 6328 goto out; 6329 } 6330 6331 /* rewrite innvl for backwards compatibility */ 6332 if (compat) 6333 innvl = zfs_ioctl_compat_innvl(zc, innvl, vecnum, cflag); 6334 6335 /* 6336 * Ensure that all pool/dataset names are valid before we pass down to 6337 * the lower layers. 6338 */ 6339 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0'; 6340 switch (vec->zvec_namecheck) { 6341 case POOL_NAME: 6342 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0) 6343 error = SET_ERROR(EINVAL); 6344 else 6345 error = pool_status_check(zc->zc_name, 6346 vec->zvec_namecheck, vec->zvec_pool_check); 6347 break; 6348 6349 case DATASET_NAME: 6350 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0) 6351 error = SET_ERROR(EINVAL); 6352 else 6353 error = pool_status_check(zc->zc_name, 6354 vec->zvec_namecheck, vec->zvec_pool_check); 6355 break; 6356 6357 case NO_NAME: 6358 break; 6359 } 6360 6361 if (error == 0 && !(flag & FKIOCTL)) 6362 error = vec->zvec_secpolicy(zc, innvl, cr); 6363 6364 if (error != 0) 6365 goto out; 6366 6367 /* legacy ioctls can modify zc_name */ 6368 len = strcspn(zc->zc_name, "/@#") + 1; 6369 saved_poolname = kmem_alloc(len, KM_SLEEP); 6370 (void) strlcpy(saved_poolname, zc->zc_name, len); 6371 6372 if (vec->zvec_func != NULL) { 6373 nvlist_t *outnvl; 6374 int puterror = 0; 6375 spa_t *spa; 6376 nvlist_t *lognv = NULL; 6377 6378 ASSERT(vec->zvec_legacy_func == NULL); 6379 6380 /* 6381 * Add the innvl to the lognv before calling the func, 6382 * in case the func changes the innvl. 6383 */ 6384 if (vec->zvec_allow_log) { 6385 lognv = fnvlist_alloc(); 6386 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL, 6387 vec->zvec_name); 6388 if (!nvlist_empty(innvl)) { 6389 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL, 6390 innvl); 6391 } 6392 } 6393 6394 outnvl = fnvlist_alloc(); 6395 error = vec->zvec_func(zc->zc_name, innvl, outnvl); 6396 6397 if (error == 0 && vec->zvec_allow_log && 6398 spa_open(zc->zc_name, &spa, FTAG) == 0) { 6399 if (!nvlist_empty(outnvl)) { 6400 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL, 6401 outnvl); 6402 } 6403 (void) spa_history_log_nvl(spa, lognv); 6404 spa_close(spa, FTAG); 6405 } 6406 fnvlist_free(lognv); 6407 6408 /* rewrite outnvl for backwards compatibility */ 6409 if (compat) 6410 outnvl = zfs_ioctl_compat_outnvl(zc, outnvl, vecnum, 6411 cflag); 6412 6413 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) { 6414 int smusherror = 0; 6415 if (vec->zvec_smush_outnvlist) { 6416 smusherror = nvlist_smush(outnvl, 6417 zc->zc_nvlist_dst_size); 6418 } 6419 if (smusherror == 0) 6420 puterror = put_nvlist(zc, outnvl); 6421 } 6422 6423 if (puterror != 0) 6424 error = puterror; 6425 6426 nvlist_free(outnvl); 6427 } else { 6428 error = vec->zvec_legacy_func(zc); 6429 } 6430 6431out: 6432 nvlist_free(innvl); 6433 6434#ifdef illumos 6435 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag); 6436 if (error == 0 && rc != 0) 6437 error = SET_ERROR(EFAULT); 6438#else 6439 if (compat) { 6440 zfs_ioctl_compat_post(zc, cmd, cflag); 6441 if (newioc) { 6442 ASSERT(compat_zc != NULL); 6443 ASSERT(sizeof(zfs_cmd_t) >= zc_iocparm->zfs_cmd_size); 6444 6445 zfs_cmd_compat_put(zc, compat_zc, vecnum, cflag); 6446 rc = ddi_copyout(compat_zc, 6447 (void *)(uintptr_t)zc_iocparm->zfs_cmd, 6448 zc_iocparm->zfs_cmd_size, flag); 6449 if (error == 0 && rc != 0) 6450 error = SET_ERROR(EFAULT); 6451 kmem_free(compat_zc, sizeof (zfs_cmd_t)); 6452 } else { 6453 zfs_cmd_compat_put(zc, arg, vecnum, cflag); 6454 } 6455 } else { 6456 ASSERT(newioc); 6457 6458 rc = ddi_copyout(zc, (void *)(uintptr_t)zc_iocparm->zfs_cmd, 6459 sizeof (zfs_cmd_t), flag); 6460 if (error == 0 && rc != 0) 6461 error = SET_ERROR(EFAULT); 6462 } 6463#endif 6464 if (error == 0 && vec->zvec_allow_log) { 6465 char *s = tsd_get(zfs_allow_log_key); 6466 if (s != NULL) 6467 strfree(s); 6468 (void) tsd_set(zfs_allow_log_key, saved_poolname); 6469 } else { 6470 if (saved_poolname != NULL) 6471 strfree(saved_poolname); 6472 } 6473 6474 kmem_free(zc, sizeof (zfs_cmd_t)); 6475 return (error); 6476} 6477 6478#ifdef illumos 6479static int 6480zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd) 6481{ 6482 if (cmd != DDI_ATTACH) 6483 return (DDI_FAILURE); 6484 6485 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0, 6486 DDI_PSEUDO, 0) == DDI_FAILURE) 6487 return (DDI_FAILURE); 6488 6489 zfs_dip = dip; 6490 6491 ddi_report_dev(dip); 6492 6493 return (DDI_SUCCESS); 6494} 6495 6496static int 6497zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd) 6498{ 6499 if (spa_busy() || zfs_busy() || zvol_busy()) 6500 return (DDI_FAILURE); 6501 6502 if (cmd != DDI_DETACH) 6503 return (DDI_FAILURE); 6504 6505 zfs_dip = NULL; 6506 6507 ddi_prop_remove_all(dip); 6508 ddi_remove_minor_node(dip, NULL); 6509 6510 return (DDI_SUCCESS); 6511} 6512 6513/*ARGSUSED*/ 6514static int 6515zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result) 6516{ 6517 switch (infocmd) { 6518 case DDI_INFO_DEVT2DEVINFO: 6519 *result = zfs_dip; 6520 return (DDI_SUCCESS); 6521 6522 case DDI_INFO_DEVT2INSTANCE: 6523 *result = (void *)0; 6524 return (DDI_SUCCESS); 6525 } 6526 6527 return (DDI_FAILURE); 6528} 6529#endif /* illumos */ 6530 6531/* 6532 * OK, so this is a little weird. 6533 * 6534 * /dev/zfs is the control node, i.e. minor 0. 6535 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0. 6536 * 6537 * /dev/zfs has basically nothing to do except serve up ioctls, 6538 * so most of the standard driver entry points are in zvol.c. 6539 */ 6540#ifdef illumos 6541static struct cb_ops zfs_cb_ops = { 6542 zfsdev_open, /* open */ 6543 zfsdev_close, /* close */ 6544 zvol_strategy, /* strategy */ 6545 nodev, /* print */ 6546 zvol_dump, /* dump */ 6547 zvol_read, /* read */ 6548 zvol_write, /* write */ 6549 zfsdev_ioctl, /* ioctl */ 6550 nodev, /* devmap */ 6551 nodev, /* mmap */ 6552 nodev, /* segmap */ 6553 nochpoll, /* poll */ 6554 ddi_prop_op, /* prop_op */ 6555 NULL, /* streamtab */ 6556 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */ 6557 CB_REV, /* version */ 6558 nodev, /* async read */ 6559 nodev, /* async write */ 6560}; 6561 6562static struct dev_ops zfs_dev_ops = { 6563 DEVO_REV, /* version */ 6564 0, /* refcnt */ 6565 zfs_info, /* info */ 6566 nulldev, /* identify */ 6567 nulldev, /* probe */ 6568 zfs_attach, /* attach */ 6569 zfs_detach, /* detach */ 6570 nodev, /* reset */ 6571 &zfs_cb_ops, /* driver operations */ 6572 NULL, /* no bus operations */ 6573 NULL, /* power */ 6574 ddi_quiesce_not_needed, /* quiesce */ 6575}; 6576 6577static struct modldrv zfs_modldrv = { 6578 &mod_driverops, 6579 "ZFS storage pool", 6580 &zfs_dev_ops 6581}; 6582 6583static struct modlinkage modlinkage = { 6584 MODREV_1, 6585 (void *)&zfs_modlfs, 6586 (void *)&zfs_modldrv, 6587 NULL 6588}; 6589#endif /* illumos */ 6590 6591static struct cdevsw zfs_cdevsw = { 6592 .d_version = D_VERSION, 6593 .d_open = zfsdev_open, 6594 .d_ioctl = zfsdev_ioctl, 6595 .d_name = ZFS_DEV_NAME 6596}; 6597 6598static void 6599zfs_allow_log_destroy(void *arg) 6600{ 6601 char *poolname = arg; 6602 strfree(poolname); 6603} 6604 6605static void 6606zfsdev_init(void) 6607{ 6608 zfsdev = make_dev(&zfs_cdevsw, 0x0, UID_ROOT, GID_OPERATOR, 0666, 6609 ZFS_DEV_NAME); 6610} 6611 6612static void 6613zfsdev_fini(void) 6614{ 6615 if (zfsdev != NULL) 6616 destroy_dev(zfsdev); 6617} 6618 6619static struct root_hold_token *zfs_root_token; 6620struct proc *zfsproc; 6621 6622#ifdef illumos 6623int 6624_init(void) 6625{ 6626 int error; 6627 6628 spa_init(FREAD | FWRITE); 6629 zfs_init(); 6630 zvol_init(); 6631 zfs_ioctl_init(); 6632 6633 if ((error = mod_install(&modlinkage)) != 0) { 6634 zvol_fini(); 6635 zfs_fini(); 6636 spa_fini(); 6637 return (error); 6638 } 6639 6640 tsd_create(&zfs_fsyncer_key, NULL); 6641 tsd_create(&rrw_tsd_key, rrw_tsd_destroy); 6642 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy); 6643 6644 error = ldi_ident_from_mod(&modlinkage, &zfs_li); 6645 ASSERT(error == 0); 6646 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL); 6647 6648 return (0); 6649} 6650 6651int 6652_fini(void) 6653{ 6654 int error; 6655 6656 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled) 6657 return (SET_ERROR(EBUSY)); 6658 6659 if ((error = mod_remove(&modlinkage)) != 0) 6660 return (error); 6661 6662 zvol_fini(); 6663 zfs_fini(); 6664 spa_fini(); 6665 if (zfs_nfsshare_inited) 6666 (void) ddi_modclose(nfs_mod); 6667 if (zfs_smbshare_inited) 6668 (void) ddi_modclose(smbsrv_mod); 6669 if (zfs_nfsshare_inited || zfs_smbshare_inited) 6670 (void) ddi_modclose(sharefs_mod); 6671 6672 tsd_destroy(&zfs_fsyncer_key); 6673 ldi_ident_release(zfs_li); 6674 zfs_li = NULL; 6675 mutex_destroy(&zfs_share_lock); 6676 6677 return (error); 6678} 6679 6680int 6681_info(struct modinfo *modinfop) 6682{ 6683 return (mod_info(&modlinkage, modinfop)); 6684} 6685#endif /* illumos */ 6686 6687static int zfs__init(void); 6688static int zfs__fini(void); 6689static void zfs_shutdown(void *, int); 6690 6691static eventhandler_tag zfs_shutdown_event_tag; 6692 6693#ifdef __FreeBSD__ 6694#define ZFS_MIN_KSTACK_PAGES 4 6695#endif 6696 6697int 6698zfs__init(void) 6699{ 6700 6701#ifdef __FreeBSD__ 6702#if KSTACK_PAGES < ZFS_MIN_KSTACK_PAGES 6703 printf("ZFS NOTICE: KSTACK_PAGES is %d which could result in stack " 6704 "overflow panic!\nPlease consider adding " 6705 "'options KSTACK_PAGES=%d' to your kernel config\n", KSTACK_PAGES, 6706 ZFS_MIN_KSTACK_PAGES); 6707#endif 6708#endif 6709 zfs_root_token = root_mount_hold("ZFS"); 6710 6711 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL); 6712 6713 spa_init(FREAD | FWRITE); 6714 zfs_init(); 6715 zvol_init(); 6716 zfs_ioctl_init(); 6717 6718 tsd_create(&zfs_fsyncer_key, NULL); 6719 tsd_create(&rrw_tsd_key, rrw_tsd_destroy); 6720 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy); 6721 6722 printf("ZFS storage pool version: features support (" SPA_VERSION_STRING ")\n"); 6723 root_mount_rel(zfs_root_token); 6724 6725 zfsdev_init(); 6726 6727 return (0); 6728} 6729 6730int 6731zfs__fini(void) 6732{ 6733 if (spa_busy() || zfs_busy() || zvol_busy() || 6734 zio_injection_enabled) { 6735 return (EBUSY); 6736 } 6737 6738 zfsdev_fini(); 6739 zvol_fini(); 6740 zfs_fini(); 6741 spa_fini(); 6742 6743 tsd_destroy(&zfs_fsyncer_key); 6744 tsd_destroy(&rrw_tsd_key); 6745 tsd_destroy(&zfs_allow_log_key); 6746 6747 mutex_destroy(&zfs_share_lock); 6748 6749 return (0); 6750} 6751 6752static void 6753zfs_shutdown(void *arg __unused, int howto __unused) 6754{ 6755 6756 /* 6757 * ZFS fini routines can not properly work in a panic-ed system. 6758 */ 6759 if (panicstr == NULL) 6760 (void)zfs__fini(); 6761} 6762 6763 6764static int 6765zfs_modevent(module_t mod, int type, void *unused __unused) 6766{ 6767 int err; 6768 6769 switch (type) { 6770 case MOD_LOAD: 6771 err = zfs__init(); 6772 if (err == 0) 6773 zfs_shutdown_event_tag = EVENTHANDLER_REGISTER( 6774 shutdown_post_sync, zfs_shutdown, NULL, 6775 SHUTDOWN_PRI_FIRST); 6776 return (err); 6777 case MOD_UNLOAD: 6778 err = zfs__fini(); 6779 if (err == 0 && zfs_shutdown_event_tag != NULL) 6780 EVENTHANDLER_DEREGISTER(shutdown_post_sync, 6781 zfs_shutdown_event_tag); 6782 return (err); 6783 case MOD_SHUTDOWN: 6784 return (0); 6785 default: 6786 break; 6787 } 6788 return (EOPNOTSUPP); 6789} 6790 6791static moduledata_t zfs_mod = { 6792 "zfsctrl", 6793 zfs_modevent, 6794 0 6795}; 6796DECLARE_MODULE(zfsctrl, zfs_mod, SI_SUB_VFS, SI_ORDER_ANY); 6797MODULE_VERSION(zfsctrl, 1); 6798MODULE_DEPEND(zfsctrl, opensolaris, 1, 1, 1); 6799MODULE_DEPEND(zfsctrl, krpc, 1, 1, 1); 6800MODULE_DEPEND(zfsctrl, acl_nfs4, 1, 1, 1); 6801