zfs_ioctl.c revision 267992
1/* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 22/* 23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 24 * Copyright (c) 2011-2012 Pawel Jakub Dawidek <pawel@dawidek.net>. 25 * All rights reserved. 26 * Copyright 2013 Martin Matuska <mm@FreeBSD.org>. All rights reserved. 27 * Copyright 2011 Nexenta Systems, Inc. All rights reserved. 28 * Copyright (c) 2014, Joyent, Inc. All rights reserved. 29 * Copyright (c) 2013 by Delphix. All rights reserved. 30 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved. 31 * Copyright (c) 2013 Steven Hartland. All rights reserved. 32 */ 33 34/* 35 * ZFS ioctls. 36 * 37 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage 38 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool. 39 * 40 * There are two ways that we handle ioctls: the legacy way where almost 41 * all of the logic is in the ioctl callback, and the new way where most 42 * of the marshalling is handled in the common entry point, zfsdev_ioctl(). 43 * 44 * Non-legacy ioctls should be registered by calling 45 * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked 46 * from userland by lzc_ioctl(). 47 * 48 * The registration arguments are as follows: 49 * 50 * const char *name 51 * The name of the ioctl. This is used for history logging. If the 52 * ioctl returns successfully (the callback returns 0), and allow_log 53 * is true, then a history log entry will be recorded with the input & 54 * output nvlists. The log entry can be printed with "zpool history -i". 55 * 56 * zfs_ioc_t ioc 57 * The ioctl request number, which userland will pass to ioctl(2). 58 * The ioctl numbers can change from release to release, because 59 * the caller (libzfs) must be matched to the kernel. 60 * 61 * zfs_secpolicy_func_t *secpolicy 62 * This function will be called before the zfs_ioc_func_t, to 63 * determine if this operation is permitted. It should return EPERM 64 * on failure, and 0 on success. Checks include determining if the 65 * dataset is visible in this zone, and if the user has either all 66 * zfs privileges in the zone (SYS_MOUNT), or has been granted permission 67 * to do this operation on this dataset with "zfs allow". 68 * 69 * zfs_ioc_namecheck_t namecheck 70 * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool 71 * name, a dataset name, or nothing. If the name is not well-formed, 72 * the ioctl will fail and the callback will not be called. 73 * Therefore, the callback can assume that the name is well-formed 74 * (e.g. is null-terminated, doesn't have more than one '@' character, 75 * doesn't have invalid characters). 76 * 77 * zfs_ioc_poolcheck_t pool_check 78 * This specifies requirements on the pool state. If the pool does 79 * not meet them (is suspended or is readonly), the ioctl will fail 80 * and the callback will not be called. If any checks are specified 81 * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME. 82 * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED | 83 * POOL_CHECK_READONLY). 84 * 85 * boolean_t smush_outnvlist 86 * If smush_outnvlist is true, then the output is presumed to be a 87 * list of errors, and it will be "smushed" down to fit into the 88 * caller's buffer, by removing some entries and replacing them with a 89 * single "N_MORE_ERRORS" entry indicating how many were removed. See 90 * nvlist_smush() for details. If smush_outnvlist is false, and the 91 * outnvlist does not fit into the userland-provided buffer, then the 92 * ioctl will fail with ENOMEM. 93 * 94 * zfs_ioc_func_t *func 95 * The callback function that will perform the operation. 96 * 97 * The callback should return 0 on success, or an error number on 98 * failure. If the function fails, the userland ioctl will return -1, 99 * and errno will be set to the callback's return value. The callback 100 * will be called with the following arguments: 101 * 102 * const char *name 103 * The name of the pool or dataset to operate on, from 104 * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the 105 * expected type (pool, dataset, or none). 106 * 107 * nvlist_t *innvl 108 * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or 109 * NULL if no input nvlist was provided. Changes to this nvlist are 110 * ignored. If the input nvlist could not be deserialized, the 111 * ioctl will fail and the callback will not be called. 112 * 113 * nvlist_t *outnvl 114 * The output nvlist, initially empty. The callback can fill it in, 115 * and it will be returned to userland by serializing it into 116 * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization 117 * fails (e.g. because the caller didn't supply a large enough 118 * buffer), then the overall ioctl will fail. See the 119 * 'smush_nvlist' argument above for additional behaviors. 120 * 121 * There are two typical uses of the output nvlist: 122 * - To return state, e.g. property values. In this case, 123 * smush_outnvlist should be false. If the buffer was not large 124 * enough, the caller will reallocate a larger buffer and try 125 * the ioctl again. 126 * 127 * - To return multiple errors from an ioctl which makes on-disk 128 * changes. In this case, smush_outnvlist should be true. 129 * Ioctls which make on-disk modifications should generally not 130 * use the outnvl if they succeed, because the caller can not 131 * distinguish between the operation failing, and 132 * deserialization failing. 133 */ 134 135#include <sys/types.h> 136#include <sys/param.h> 137#include <sys/systm.h> 138#include <sys/conf.h> 139#include <sys/kernel.h> 140#include <sys/lock.h> 141#include <sys/malloc.h> 142#include <sys/mutex.h> 143#include <sys/proc.h> 144#include <sys/errno.h> 145#include <sys/uio.h> 146#include <sys/buf.h> 147#include <sys/file.h> 148#include <sys/kmem.h> 149#include <sys/conf.h> 150#include <sys/cmn_err.h> 151#include <sys/stat.h> 152#include <sys/zfs_ioctl.h> 153#include <sys/zfs_vfsops.h> 154#include <sys/zfs_znode.h> 155#include <sys/zap.h> 156#include <sys/spa.h> 157#include <sys/spa_impl.h> 158#include <sys/vdev.h> 159#include <sys/dmu.h> 160#include <sys/dsl_dir.h> 161#include <sys/dsl_dataset.h> 162#include <sys/dsl_prop.h> 163#include <sys/dsl_deleg.h> 164#include <sys/dmu_objset.h> 165#include <sys/dmu_impl.h> 166#include <sys/dmu_tx.h> 167#include <sys/sunddi.h> 168#include <sys/policy.h> 169#include <sys/zone.h> 170#include <sys/nvpair.h> 171#include <sys/mount.h> 172#include <sys/taskqueue.h> 173#include <sys/sdt.h> 174#include <sys/varargs.h> 175#include <sys/fs/zfs.h> 176#include <sys/zfs_ctldir.h> 177#include <sys/zfs_dir.h> 178#include <sys/zfs_onexit.h> 179#include <sys/zvol.h> 180#include <sys/dsl_scan.h> 181#include <sys/dmu_objset.h> 182#include <sys/dmu_send.h> 183#include <sys/dsl_destroy.h> 184#include <sys/dsl_bookmark.h> 185#include <sys/dsl_userhold.h> 186#include <sys/zfeature.h> 187 188#include "zfs_namecheck.h" 189#include "zfs_prop.h" 190#include "zfs_deleg.h" 191#include "zfs_comutil.h" 192#include "zfs_ioctl_compat.h" 193 194CTASSERT(sizeof(zfs_cmd_t) < IOCPARM_MAX); 195 196static int snapshot_list_prefetch; 197SYSCTL_DECL(_vfs_zfs); 198SYSCTL_INT(_vfs_zfs, OID_AUTO, snapshot_list_prefetch, CTLFLAG_RWTUN, 199 &snapshot_list_prefetch, 0, "Prefetch data when listing snapshots"); 200 201static struct cdev *zfsdev; 202 203extern void zfs_init(void); 204extern void zfs_fini(void); 205 206uint_t zfs_fsyncer_key; 207extern uint_t rrw_tsd_key; 208static uint_t zfs_allow_log_key; 209 210typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *); 211typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *); 212typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *); 213 214typedef enum { 215 NO_NAME, 216 POOL_NAME, 217 DATASET_NAME 218} zfs_ioc_namecheck_t; 219 220typedef enum { 221 POOL_CHECK_NONE = 1 << 0, 222 POOL_CHECK_SUSPENDED = 1 << 1, 223 POOL_CHECK_READONLY = 1 << 2, 224} zfs_ioc_poolcheck_t; 225 226typedef struct zfs_ioc_vec { 227 zfs_ioc_legacy_func_t *zvec_legacy_func; 228 zfs_ioc_func_t *zvec_func; 229 zfs_secpolicy_func_t *zvec_secpolicy; 230 zfs_ioc_namecheck_t zvec_namecheck; 231 boolean_t zvec_allow_log; 232 zfs_ioc_poolcheck_t zvec_pool_check; 233 boolean_t zvec_smush_outnvlist; 234 const char *zvec_name; 235} zfs_ioc_vec_t; 236 237/* This array is indexed by zfs_userquota_prop_t */ 238static const char *userquota_perms[] = { 239 ZFS_DELEG_PERM_USERUSED, 240 ZFS_DELEG_PERM_USERQUOTA, 241 ZFS_DELEG_PERM_GROUPUSED, 242 ZFS_DELEG_PERM_GROUPQUOTA, 243}; 244 245static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc); 246static int zfs_check_settable(const char *name, nvpair_t *property, 247 cred_t *cr); 248static int zfs_check_clearable(char *dataset, nvlist_t *props, 249 nvlist_t **errors); 250static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *, 251 boolean_t *); 252int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *); 253static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp); 254 255static void zfsdev_close(void *data); 256 257static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature); 258 259/* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */ 260void 261__dprintf(const char *file, const char *func, int line, const char *fmt, ...) 262{ 263 const char *newfile; 264 char buf[512]; 265 va_list adx; 266 267 /* 268 * Get rid of annoying "../common/" prefix to filename. 269 */ 270 newfile = strrchr(file, '/'); 271 if (newfile != NULL) { 272 newfile = newfile + 1; /* Get rid of leading / */ 273 } else { 274 newfile = file; 275 } 276 277 va_start(adx, fmt); 278 (void) vsnprintf(buf, sizeof (buf), fmt, adx); 279 va_end(adx); 280 281 /* 282 * To get this data, use the zfs-dprintf probe as so: 283 * dtrace -q -n 'zfs-dprintf \ 284 * /stringof(arg0) == "dbuf.c"/ \ 285 * {printf("%s: %s", stringof(arg1), stringof(arg3))}' 286 * arg0 = file name 287 * arg1 = function name 288 * arg2 = line number 289 * arg3 = message 290 */ 291 DTRACE_PROBE4(zfs__dprintf, 292 char *, newfile, char *, func, int, line, char *, buf); 293} 294 295static void 296history_str_free(char *buf) 297{ 298 kmem_free(buf, HIS_MAX_RECORD_LEN); 299} 300 301static char * 302history_str_get(zfs_cmd_t *zc) 303{ 304 char *buf; 305 306 if (zc->zc_history == 0) 307 return (NULL); 308 309 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP); 310 if (copyinstr((void *)(uintptr_t)zc->zc_history, 311 buf, HIS_MAX_RECORD_LEN, NULL) != 0) { 312 history_str_free(buf); 313 return (NULL); 314 } 315 316 buf[HIS_MAX_RECORD_LEN -1] = '\0'; 317 318 return (buf); 319} 320 321/* 322 * Check to see if the named dataset is currently defined as bootable 323 */ 324static boolean_t 325zfs_is_bootfs(const char *name) 326{ 327 objset_t *os; 328 329 if (dmu_objset_hold(name, FTAG, &os) == 0) { 330 boolean_t ret; 331 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os))); 332 dmu_objset_rele(os, FTAG); 333 return (ret); 334 } 335 return (B_FALSE); 336} 337 338/* 339 * Return non-zero if the spa version is less than requested version. 340 */ 341static int 342zfs_earlier_version(const char *name, int version) 343{ 344 spa_t *spa; 345 346 if (spa_open(name, &spa, FTAG) == 0) { 347 if (spa_version(spa) < version) { 348 spa_close(spa, FTAG); 349 return (1); 350 } 351 spa_close(spa, FTAG); 352 } 353 return (0); 354} 355 356/* 357 * Return TRUE if the ZPL version is less than requested version. 358 */ 359static boolean_t 360zpl_earlier_version(const char *name, int version) 361{ 362 objset_t *os; 363 boolean_t rc = B_TRUE; 364 365 if (dmu_objset_hold(name, FTAG, &os) == 0) { 366 uint64_t zplversion; 367 368 if (dmu_objset_type(os) != DMU_OST_ZFS) { 369 dmu_objset_rele(os, FTAG); 370 return (B_TRUE); 371 } 372 /* XXX reading from non-owned objset */ 373 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0) 374 rc = zplversion < version; 375 dmu_objset_rele(os, FTAG); 376 } 377 return (rc); 378} 379 380static void 381zfs_log_history(zfs_cmd_t *zc) 382{ 383 spa_t *spa; 384 char *buf; 385 386 if ((buf = history_str_get(zc)) == NULL) 387 return; 388 389 if (spa_open(zc->zc_name, &spa, FTAG) == 0) { 390 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY) 391 (void) spa_history_log(spa, buf); 392 spa_close(spa, FTAG); 393 } 394 history_str_free(buf); 395} 396 397/* 398 * Policy for top-level read operations (list pools). Requires no privileges, 399 * and can be used in the local zone, as there is no associated dataset. 400 */ 401/* ARGSUSED */ 402static int 403zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 404{ 405 return (0); 406} 407 408/* 409 * Policy for dataset read operations (list children, get statistics). Requires 410 * no privileges, but must be visible in the local zone. 411 */ 412/* ARGSUSED */ 413static int 414zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 415{ 416 if (INGLOBALZONE(curthread) || 417 zone_dataset_visible(zc->zc_name, NULL)) 418 return (0); 419 420 return (SET_ERROR(ENOENT)); 421} 422 423static int 424zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr) 425{ 426 int writable = 1; 427 428 /* 429 * The dataset must be visible by this zone -- check this first 430 * so they don't see EPERM on something they shouldn't know about. 431 */ 432 if (!INGLOBALZONE(curthread) && 433 !zone_dataset_visible(dataset, &writable)) 434 return (SET_ERROR(ENOENT)); 435 436 if (INGLOBALZONE(curthread)) { 437 /* 438 * If the fs is zoned, only root can access it from the 439 * global zone. 440 */ 441 if (secpolicy_zfs(cr) && zoned) 442 return (SET_ERROR(EPERM)); 443 } else { 444 /* 445 * If we are in a local zone, the 'zoned' property must be set. 446 */ 447 if (!zoned) 448 return (SET_ERROR(EPERM)); 449 450 /* must be writable by this zone */ 451 if (!writable) 452 return (SET_ERROR(EPERM)); 453 } 454 return (0); 455} 456 457static int 458zfs_dozonecheck(const char *dataset, cred_t *cr) 459{ 460 uint64_t zoned; 461 462 if (dsl_prop_get_integer(dataset, "jailed", &zoned, NULL)) 463 return (SET_ERROR(ENOENT)); 464 465 return (zfs_dozonecheck_impl(dataset, zoned, cr)); 466} 467 468static int 469zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr) 470{ 471 uint64_t zoned; 472 473 if (dsl_prop_get_int_ds(ds, "jailed", &zoned)) 474 return (SET_ERROR(ENOENT)); 475 476 return (zfs_dozonecheck_impl(dataset, zoned, cr)); 477} 478 479static int 480zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds, 481 const char *perm, cred_t *cr) 482{ 483 int error; 484 485 error = zfs_dozonecheck_ds(name, ds, cr); 486 if (error == 0) { 487 error = secpolicy_zfs(cr); 488 if (error != 0) 489 error = dsl_deleg_access_impl(ds, perm, cr); 490 } 491 return (error); 492} 493 494static int 495zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr) 496{ 497 int error; 498 dsl_dataset_t *ds; 499 dsl_pool_t *dp; 500 501 error = dsl_pool_hold(name, FTAG, &dp); 502 if (error != 0) 503 return (error); 504 505 error = dsl_dataset_hold(dp, name, FTAG, &ds); 506 if (error != 0) { 507 dsl_pool_rele(dp, FTAG); 508 return (error); 509 } 510 511 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr); 512 513 dsl_dataset_rele(ds, FTAG); 514 dsl_pool_rele(dp, FTAG); 515 return (error); 516} 517 518#ifdef SECLABEL 519/* 520 * Policy for setting the security label property. 521 * 522 * Returns 0 for success, non-zero for access and other errors. 523 */ 524static int 525zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr) 526{ 527 char ds_hexsl[MAXNAMELEN]; 528 bslabel_t ds_sl, new_sl; 529 boolean_t new_default = FALSE; 530 uint64_t zoned; 531 int needed_priv = -1; 532 int error; 533 534 /* First get the existing dataset label. */ 535 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL), 536 1, sizeof (ds_hexsl), &ds_hexsl, NULL); 537 if (error != 0) 538 return (SET_ERROR(EPERM)); 539 540 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0) 541 new_default = TRUE; 542 543 /* The label must be translatable */ 544 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0)) 545 return (SET_ERROR(EINVAL)); 546 547 /* 548 * In a non-global zone, disallow attempts to set a label that 549 * doesn't match that of the zone; otherwise no other checks 550 * are needed. 551 */ 552 if (!INGLOBALZONE(curproc)) { 553 if (new_default || !blequal(&new_sl, CR_SL(CRED()))) 554 return (SET_ERROR(EPERM)); 555 return (0); 556 } 557 558 /* 559 * For global-zone datasets (i.e., those whose zoned property is 560 * "off", verify that the specified new label is valid for the 561 * global zone. 562 */ 563 if (dsl_prop_get_integer(name, 564 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL)) 565 return (SET_ERROR(EPERM)); 566 if (!zoned) { 567 if (zfs_check_global_label(name, strval) != 0) 568 return (SET_ERROR(EPERM)); 569 } 570 571 /* 572 * If the existing dataset label is nondefault, check if the 573 * dataset is mounted (label cannot be changed while mounted). 574 * Get the zfsvfs; if there isn't one, then the dataset isn't 575 * mounted (or isn't a dataset, doesn't exist, ...). 576 */ 577 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) { 578 objset_t *os; 579 static char *setsl_tag = "setsl_tag"; 580 581 /* 582 * Try to own the dataset; abort if there is any error, 583 * (e.g., already mounted, in use, or other error). 584 */ 585 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE, 586 setsl_tag, &os); 587 if (error != 0) 588 return (SET_ERROR(EPERM)); 589 590 dmu_objset_disown(os, setsl_tag); 591 592 if (new_default) { 593 needed_priv = PRIV_FILE_DOWNGRADE_SL; 594 goto out_check; 595 } 596 597 if (hexstr_to_label(strval, &new_sl) != 0) 598 return (SET_ERROR(EPERM)); 599 600 if (blstrictdom(&ds_sl, &new_sl)) 601 needed_priv = PRIV_FILE_DOWNGRADE_SL; 602 else if (blstrictdom(&new_sl, &ds_sl)) 603 needed_priv = PRIV_FILE_UPGRADE_SL; 604 } else { 605 /* dataset currently has a default label */ 606 if (!new_default) 607 needed_priv = PRIV_FILE_UPGRADE_SL; 608 } 609 610out_check: 611 if (needed_priv != -1) 612 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL)); 613 return (0); 614} 615#endif /* SECLABEL */ 616 617static int 618zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval, 619 cred_t *cr) 620{ 621 char *strval; 622 623 /* 624 * Check permissions for special properties. 625 */ 626 switch (prop) { 627 case ZFS_PROP_ZONED: 628 /* 629 * Disallow setting of 'zoned' from within a local zone. 630 */ 631 if (!INGLOBALZONE(curthread)) 632 return (SET_ERROR(EPERM)); 633 break; 634 635 case ZFS_PROP_QUOTA: 636 case ZFS_PROP_FILESYSTEM_LIMIT: 637 case ZFS_PROP_SNAPSHOT_LIMIT: 638 if (!INGLOBALZONE(curthread)) { 639 uint64_t zoned; 640 char setpoint[MAXNAMELEN]; 641 /* 642 * Unprivileged users are allowed to modify the 643 * limit on things *under* (ie. contained by) 644 * the thing they own. 645 */ 646 if (dsl_prop_get_integer(dsname, "jailed", &zoned, 647 setpoint)) 648 return (SET_ERROR(EPERM)); 649 if (!zoned || strlen(dsname) <= strlen(setpoint)) 650 return (SET_ERROR(EPERM)); 651 } 652 break; 653 654 case ZFS_PROP_MLSLABEL: 655#ifdef SECLABEL 656 if (!is_system_labeled()) 657 return (SET_ERROR(EPERM)); 658 659 if (nvpair_value_string(propval, &strval) == 0) { 660 int err; 661 662 err = zfs_set_slabel_policy(dsname, strval, CRED()); 663 if (err != 0) 664 return (err); 665 } 666#else 667 return (EOPNOTSUPP); 668#endif 669 break; 670 } 671 672 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr)); 673} 674 675/* ARGSUSED */ 676static int 677zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 678{ 679 int error; 680 681 error = zfs_dozonecheck(zc->zc_name, cr); 682 if (error != 0) 683 return (error); 684 685 /* 686 * permission to set permissions will be evaluated later in 687 * dsl_deleg_can_allow() 688 */ 689 return (0); 690} 691 692/* ARGSUSED */ 693static int 694zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 695{ 696 return (zfs_secpolicy_write_perms(zc->zc_name, 697 ZFS_DELEG_PERM_ROLLBACK, cr)); 698} 699 700/* ARGSUSED */ 701static int 702zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 703{ 704 dsl_pool_t *dp; 705 dsl_dataset_t *ds; 706 char *cp; 707 int error; 708 709 /* 710 * Generate the current snapshot name from the given objsetid, then 711 * use that name for the secpolicy/zone checks. 712 */ 713 cp = strchr(zc->zc_name, '@'); 714 if (cp == NULL) 715 return (SET_ERROR(EINVAL)); 716 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 717 if (error != 0) 718 return (error); 719 720 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds); 721 if (error != 0) { 722 dsl_pool_rele(dp, FTAG); 723 return (error); 724 } 725 726 dsl_dataset_name(ds, zc->zc_name); 727 728 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds, 729 ZFS_DELEG_PERM_SEND, cr); 730 dsl_dataset_rele(ds, FTAG); 731 dsl_pool_rele(dp, FTAG); 732 733 return (error); 734} 735 736/* ARGSUSED */ 737static int 738zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 739{ 740 return (zfs_secpolicy_write_perms(zc->zc_name, 741 ZFS_DELEG_PERM_SEND, cr)); 742} 743 744/* ARGSUSED */ 745static int 746zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 747{ 748 vnode_t *vp; 749 int error; 750 751 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, 752 NO_FOLLOW, NULL, &vp)) != 0) 753 return (error); 754 755 /* Now make sure mntpnt and dataset are ZFS */ 756 757 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 || 758 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), 759 zc->zc_name) != 0)) { 760 VN_RELE(vp); 761 return (SET_ERROR(EPERM)); 762 } 763 764 VN_RELE(vp); 765 return (dsl_deleg_access(zc->zc_name, 766 ZFS_DELEG_PERM_SHARE, cr)); 767} 768 769int 770zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 771{ 772 if (!INGLOBALZONE(curthread)) 773 return (SET_ERROR(EPERM)); 774 775 if (secpolicy_nfs(cr) == 0) { 776 return (0); 777 } else { 778 return (zfs_secpolicy_deleg_share(zc, innvl, cr)); 779 } 780} 781 782int 783zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 784{ 785 if (!INGLOBALZONE(curthread)) 786 return (SET_ERROR(EPERM)); 787 788 if (secpolicy_smb(cr) == 0) { 789 return (0); 790 } else { 791 return (zfs_secpolicy_deleg_share(zc, innvl, cr)); 792 } 793} 794 795static int 796zfs_get_parent(const char *datasetname, char *parent, int parentsize) 797{ 798 char *cp; 799 800 /* 801 * Remove the @bla or /bla from the end of the name to get the parent. 802 */ 803 (void) strncpy(parent, datasetname, parentsize); 804 cp = strrchr(parent, '@'); 805 if (cp != NULL) { 806 cp[0] = '\0'; 807 } else { 808 cp = strrchr(parent, '/'); 809 if (cp == NULL) 810 return (SET_ERROR(ENOENT)); 811 cp[0] = '\0'; 812 } 813 814 return (0); 815} 816 817int 818zfs_secpolicy_destroy_perms(const char *name, cred_t *cr) 819{ 820 int error; 821 822 if ((error = zfs_secpolicy_write_perms(name, 823 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 824 return (error); 825 826 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr)); 827} 828 829/* ARGSUSED */ 830static int 831zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 832{ 833 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr)); 834} 835 836/* 837 * Destroying snapshots with delegated permissions requires 838 * descendant mount and destroy permissions. 839 */ 840/* ARGSUSED */ 841static int 842zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 843{ 844 nvlist_t *snaps; 845 nvpair_t *pair, *nextpair; 846 int error = 0; 847 848 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 849 return (SET_ERROR(EINVAL)); 850 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 851 pair = nextpair) { 852 nextpair = nvlist_next_nvpair(snaps, pair); 853 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr); 854 if (error == ENOENT) { 855 /* 856 * Ignore any snapshots that don't exist (we consider 857 * them "already destroyed"). Remove the name from the 858 * nvl here in case the snapshot is created between 859 * now and when we try to destroy it (in which case 860 * we don't want to destroy it since we haven't 861 * checked for permission). 862 */ 863 fnvlist_remove_nvpair(snaps, pair); 864 error = 0; 865 } 866 if (error != 0) 867 break; 868 } 869 870 return (error); 871} 872 873int 874zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr) 875{ 876 char parentname[MAXNAMELEN]; 877 int error; 878 879 if ((error = zfs_secpolicy_write_perms(from, 880 ZFS_DELEG_PERM_RENAME, cr)) != 0) 881 return (error); 882 883 if ((error = zfs_secpolicy_write_perms(from, 884 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 885 return (error); 886 887 if ((error = zfs_get_parent(to, parentname, 888 sizeof (parentname))) != 0) 889 return (error); 890 891 if ((error = zfs_secpolicy_write_perms(parentname, 892 ZFS_DELEG_PERM_CREATE, cr)) != 0) 893 return (error); 894 895 if ((error = zfs_secpolicy_write_perms(parentname, 896 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 897 return (error); 898 899 return (error); 900} 901 902/* ARGSUSED */ 903static int 904zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 905{ 906 char *at = NULL; 907 int error; 908 909 if ((zc->zc_cookie & 1) != 0) { 910 /* 911 * This is recursive rename, so the starting snapshot might 912 * not exist. Check file system or volume permission instead. 913 */ 914 at = strchr(zc->zc_name, '@'); 915 if (at == NULL) 916 return (EINVAL); 917 *at = '\0'; 918 } 919 920 error = zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr); 921 922 if (at != NULL) 923 *at = '@'; 924 925 return (error); 926} 927 928/* ARGSUSED */ 929static int 930zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 931{ 932 dsl_pool_t *dp; 933 dsl_dataset_t *clone; 934 int error; 935 936 error = zfs_secpolicy_write_perms(zc->zc_name, 937 ZFS_DELEG_PERM_PROMOTE, cr); 938 if (error != 0) 939 return (error); 940 941 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 942 if (error != 0) 943 return (error); 944 945 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone); 946 947 if (error == 0) { 948 char parentname[MAXNAMELEN]; 949 dsl_dataset_t *origin = NULL; 950 dsl_dir_t *dd; 951 dd = clone->ds_dir; 952 953 error = dsl_dataset_hold_obj(dd->dd_pool, 954 dd->dd_phys->dd_origin_obj, FTAG, &origin); 955 if (error != 0) { 956 dsl_dataset_rele(clone, FTAG); 957 dsl_pool_rele(dp, FTAG); 958 return (error); 959 } 960 961 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone, 962 ZFS_DELEG_PERM_MOUNT, cr); 963 964 dsl_dataset_name(origin, parentname); 965 if (error == 0) { 966 error = zfs_secpolicy_write_perms_ds(parentname, origin, 967 ZFS_DELEG_PERM_PROMOTE, cr); 968 } 969 dsl_dataset_rele(clone, FTAG); 970 dsl_dataset_rele(origin, FTAG); 971 } 972 dsl_pool_rele(dp, FTAG); 973 return (error); 974} 975 976/* ARGSUSED */ 977static int 978zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 979{ 980 int error; 981 982 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 983 ZFS_DELEG_PERM_RECEIVE, cr)) != 0) 984 return (error); 985 986 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 987 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 988 return (error); 989 990 return (zfs_secpolicy_write_perms(zc->zc_name, 991 ZFS_DELEG_PERM_CREATE, cr)); 992} 993 994int 995zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr) 996{ 997 return (zfs_secpolicy_write_perms(name, 998 ZFS_DELEG_PERM_SNAPSHOT, cr)); 999} 1000 1001/* 1002 * Check for permission to create each snapshot in the nvlist. 1003 */ 1004/* ARGSUSED */ 1005static int 1006zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1007{ 1008 nvlist_t *snaps; 1009 int error; 1010 nvpair_t *pair; 1011 1012 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 1013 return (SET_ERROR(EINVAL)); 1014 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 1015 pair = nvlist_next_nvpair(snaps, pair)) { 1016 char *name = nvpair_name(pair); 1017 char *atp = strchr(name, '@'); 1018 1019 if (atp == NULL) { 1020 error = SET_ERROR(EINVAL); 1021 break; 1022 } 1023 *atp = '\0'; 1024 error = zfs_secpolicy_snapshot_perms(name, cr); 1025 *atp = '@'; 1026 if (error != 0) 1027 break; 1028 } 1029 return (error); 1030} 1031 1032/* 1033 * Check for permission to create each snapshot in the nvlist. 1034 */ 1035/* ARGSUSED */ 1036static int 1037zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1038{ 1039 int error = 0; 1040 1041 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 1042 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 1043 char *name = nvpair_name(pair); 1044 char *hashp = strchr(name, '#'); 1045 1046 if (hashp == NULL) { 1047 error = SET_ERROR(EINVAL); 1048 break; 1049 } 1050 *hashp = '\0'; 1051 error = zfs_secpolicy_write_perms(name, 1052 ZFS_DELEG_PERM_BOOKMARK, cr); 1053 *hashp = '#'; 1054 if (error != 0) 1055 break; 1056 } 1057 return (error); 1058} 1059 1060/* ARGSUSED */ 1061static int 1062zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1063{ 1064 nvpair_t *pair, *nextpair; 1065 int error = 0; 1066 1067 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL; 1068 pair = nextpair) { 1069 char *name = nvpair_name(pair); 1070 char *hashp = strchr(name, '#'); 1071 nextpair = nvlist_next_nvpair(innvl, pair); 1072 1073 if (hashp == NULL) { 1074 error = SET_ERROR(EINVAL); 1075 break; 1076 } 1077 1078 *hashp = '\0'; 1079 error = zfs_secpolicy_write_perms(name, 1080 ZFS_DELEG_PERM_DESTROY, cr); 1081 *hashp = '#'; 1082 if (error == ENOENT) { 1083 /* 1084 * Ignore any filesystems that don't exist (we consider 1085 * their bookmarks "already destroyed"). Remove 1086 * the name from the nvl here in case the filesystem 1087 * is created between now and when we try to destroy 1088 * the bookmark (in which case we don't want to 1089 * destroy it since we haven't checked for permission). 1090 */ 1091 fnvlist_remove_nvpair(innvl, pair); 1092 error = 0; 1093 } 1094 if (error != 0) 1095 break; 1096 } 1097 1098 return (error); 1099} 1100 1101/* ARGSUSED */ 1102static int 1103zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1104{ 1105 /* 1106 * Even root must have a proper TSD so that we know what pool 1107 * to log to. 1108 */ 1109 if (tsd_get(zfs_allow_log_key) == NULL) 1110 return (SET_ERROR(EPERM)); 1111 return (0); 1112} 1113 1114static int 1115zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1116{ 1117 char parentname[MAXNAMELEN]; 1118 int error; 1119 char *origin; 1120 1121 if ((error = zfs_get_parent(zc->zc_name, parentname, 1122 sizeof (parentname))) != 0) 1123 return (error); 1124 1125 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 && 1126 (error = zfs_secpolicy_write_perms(origin, 1127 ZFS_DELEG_PERM_CLONE, cr)) != 0) 1128 return (error); 1129 1130 if ((error = zfs_secpolicy_write_perms(parentname, 1131 ZFS_DELEG_PERM_CREATE, cr)) != 0) 1132 return (error); 1133 1134 return (zfs_secpolicy_write_perms(parentname, 1135 ZFS_DELEG_PERM_MOUNT, cr)); 1136} 1137 1138/* 1139 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires 1140 * SYS_CONFIG privilege, which is not available in a local zone. 1141 */ 1142/* ARGSUSED */ 1143static int 1144zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1145{ 1146 if (secpolicy_sys_config(cr, B_FALSE) != 0) 1147 return (SET_ERROR(EPERM)); 1148 1149 return (0); 1150} 1151 1152/* 1153 * Policy for object to name lookups. 1154 */ 1155/* ARGSUSED */ 1156static int 1157zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1158{ 1159 int error; 1160 1161 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0) 1162 return (0); 1163 1164 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr); 1165 return (error); 1166} 1167 1168/* 1169 * Policy for fault injection. Requires all privileges. 1170 */ 1171/* ARGSUSED */ 1172static int 1173zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1174{ 1175 return (secpolicy_zinject(cr)); 1176} 1177 1178/* ARGSUSED */ 1179static int 1180zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1181{ 1182 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value); 1183 1184 if (prop == ZPROP_INVAL) { 1185 if (!zfs_prop_user(zc->zc_value)) 1186 return (SET_ERROR(EINVAL)); 1187 return (zfs_secpolicy_write_perms(zc->zc_name, 1188 ZFS_DELEG_PERM_USERPROP, cr)); 1189 } else { 1190 return (zfs_secpolicy_setprop(zc->zc_name, prop, 1191 NULL, cr)); 1192 } 1193} 1194 1195static int 1196zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1197{ 1198 int err = zfs_secpolicy_read(zc, innvl, cr); 1199 if (err) 1200 return (err); 1201 1202 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 1203 return (SET_ERROR(EINVAL)); 1204 1205 if (zc->zc_value[0] == 0) { 1206 /* 1207 * They are asking about a posix uid/gid. If it's 1208 * themself, allow it. 1209 */ 1210 if (zc->zc_objset_type == ZFS_PROP_USERUSED || 1211 zc->zc_objset_type == ZFS_PROP_USERQUOTA) { 1212 if (zc->zc_guid == crgetuid(cr)) 1213 return (0); 1214 } else { 1215 if (groupmember(zc->zc_guid, cr)) 1216 return (0); 1217 } 1218 } 1219 1220 return (zfs_secpolicy_write_perms(zc->zc_name, 1221 userquota_perms[zc->zc_objset_type], cr)); 1222} 1223 1224static int 1225zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1226{ 1227 int err = zfs_secpolicy_read(zc, innvl, cr); 1228 if (err) 1229 return (err); 1230 1231 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 1232 return (SET_ERROR(EINVAL)); 1233 1234 return (zfs_secpolicy_write_perms(zc->zc_name, 1235 userquota_perms[zc->zc_objset_type], cr)); 1236} 1237 1238/* ARGSUSED */ 1239static int 1240zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1241{ 1242 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION, 1243 NULL, cr)); 1244} 1245 1246/* ARGSUSED */ 1247static int 1248zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1249{ 1250 nvpair_t *pair; 1251 nvlist_t *holds; 1252 int error; 1253 1254 error = nvlist_lookup_nvlist(innvl, "holds", &holds); 1255 if (error != 0) 1256 return (SET_ERROR(EINVAL)); 1257 1258 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL; 1259 pair = nvlist_next_nvpair(holds, pair)) { 1260 char fsname[MAXNAMELEN]; 1261 error = dmu_fsname(nvpair_name(pair), fsname); 1262 if (error != 0) 1263 return (error); 1264 error = zfs_secpolicy_write_perms(fsname, 1265 ZFS_DELEG_PERM_HOLD, cr); 1266 if (error != 0) 1267 return (error); 1268 } 1269 return (0); 1270} 1271 1272/* ARGSUSED */ 1273static int 1274zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1275{ 1276 nvpair_t *pair; 1277 int error; 1278 1279 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL; 1280 pair = nvlist_next_nvpair(innvl, pair)) { 1281 char fsname[MAXNAMELEN]; 1282 error = dmu_fsname(nvpair_name(pair), fsname); 1283 if (error != 0) 1284 return (error); 1285 error = zfs_secpolicy_write_perms(fsname, 1286 ZFS_DELEG_PERM_RELEASE, cr); 1287 if (error != 0) 1288 return (error); 1289 } 1290 return (0); 1291} 1292 1293/* 1294 * Policy for allowing temporary snapshots to be taken or released 1295 */ 1296static int 1297zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1298{ 1299 /* 1300 * A temporary snapshot is the same as a snapshot, 1301 * hold, destroy and release all rolled into one. 1302 * Delegated diff alone is sufficient that we allow this. 1303 */ 1304 int error; 1305 1306 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 1307 ZFS_DELEG_PERM_DIFF, cr)) == 0) 1308 return (0); 1309 1310 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr); 1311 if (error == 0) 1312 error = zfs_secpolicy_hold(zc, innvl, cr); 1313 if (error == 0) 1314 error = zfs_secpolicy_release(zc, innvl, cr); 1315 if (error == 0) 1316 error = zfs_secpolicy_destroy(zc, innvl, cr); 1317 return (error); 1318} 1319 1320/* 1321 * Returns the nvlist as specified by the user in the zfs_cmd_t. 1322 */ 1323static int 1324get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp) 1325{ 1326 char *packed; 1327 int error; 1328 nvlist_t *list = NULL; 1329 1330 /* 1331 * Read in and unpack the user-supplied nvlist. 1332 */ 1333 if (size == 0) 1334 return (SET_ERROR(EINVAL)); 1335 1336 packed = kmem_alloc(size, KM_SLEEP); 1337 1338 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size, 1339 iflag)) != 0) { 1340 kmem_free(packed, size); 1341 return (error); 1342 } 1343 1344 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) { 1345 kmem_free(packed, size); 1346 return (error); 1347 } 1348 1349 kmem_free(packed, size); 1350 1351 *nvp = list; 1352 return (0); 1353} 1354 1355/* 1356 * Reduce the size of this nvlist until it can be serialized in 'max' bytes. 1357 * Entries will be removed from the end of the nvlist, and one int32 entry 1358 * named "N_MORE_ERRORS" will be added indicating how many entries were 1359 * removed. 1360 */ 1361static int 1362nvlist_smush(nvlist_t *errors, size_t max) 1363{ 1364 size_t size; 1365 1366 size = fnvlist_size(errors); 1367 1368 if (size > max) { 1369 nvpair_t *more_errors; 1370 int n = 0; 1371 1372 if (max < 1024) 1373 return (SET_ERROR(ENOMEM)); 1374 1375 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0); 1376 more_errors = nvlist_prev_nvpair(errors, NULL); 1377 1378 do { 1379 nvpair_t *pair = nvlist_prev_nvpair(errors, 1380 more_errors); 1381 fnvlist_remove_nvpair(errors, pair); 1382 n++; 1383 size = fnvlist_size(errors); 1384 } while (size > max); 1385 1386 fnvlist_remove_nvpair(errors, more_errors); 1387 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n); 1388 ASSERT3U(fnvlist_size(errors), <=, max); 1389 } 1390 1391 return (0); 1392} 1393 1394static int 1395put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl) 1396{ 1397 char *packed = NULL; 1398 int error = 0; 1399 size_t size; 1400 1401 size = fnvlist_size(nvl); 1402 1403 if (size > zc->zc_nvlist_dst_size) { 1404 /* 1405 * Solaris returns ENOMEM here, because even if an error is 1406 * returned from an ioctl(2), new zc_nvlist_dst_size will be 1407 * passed to the userland. This is not the case for FreeBSD. 1408 * We need to return 0, so the kernel will copy the 1409 * zc_nvlist_dst_size back and the userland can discover that a 1410 * bigger buffer is needed. 1411 */ 1412 error = 0; 1413 } else { 1414 packed = fnvlist_pack(nvl, &size); 1415 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst, 1416 size, zc->zc_iflags) != 0) 1417 error = SET_ERROR(EFAULT); 1418 fnvlist_pack_free(packed, size); 1419 } 1420 1421 zc->zc_nvlist_dst_size = size; 1422 zc->zc_nvlist_dst_filled = B_TRUE; 1423 return (error); 1424} 1425 1426static int 1427getzfsvfs(const char *dsname, zfsvfs_t **zfvp) 1428{ 1429 objset_t *os; 1430 int error; 1431 1432 error = dmu_objset_hold(dsname, FTAG, &os); 1433 if (error != 0) 1434 return (error); 1435 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1436 dmu_objset_rele(os, FTAG); 1437 return (SET_ERROR(EINVAL)); 1438 } 1439 1440 mutex_enter(&os->os_user_ptr_lock); 1441 *zfvp = dmu_objset_get_user(os); 1442 if (*zfvp) { 1443 VFS_HOLD((*zfvp)->z_vfs); 1444 } else { 1445 error = SET_ERROR(ESRCH); 1446 } 1447 mutex_exit(&os->os_user_ptr_lock); 1448 dmu_objset_rele(os, FTAG); 1449 return (error); 1450} 1451 1452/* 1453 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which 1454 * case its z_vfs will be NULL, and it will be opened as the owner. 1455 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER, 1456 * which prevents all vnode ops from running. 1457 */ 1458static int 1459zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer) 1460{ 1461 int error = 0; 1462 1463 if (getzfsvfs(name, zfvp) != 0) 1464 error = zfsvfs_create(name, zfvp); 1465 if (error == 0) { 1466 rrw_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER : 1467 RW_READER, tag); 1468 if ((*zfvp)->z_unmounted) { 1469 /* 1470 * XXX we could probably try again, since the unmounting 1471 * thread should be just about to disassociate the 1472 * objset from the zfsvfs. 1473 */ 1474 rrw_exit(&(*zfvp)->z_teardown_lock, tag); 1475 return (SET_ERROR(EBUSY)); 1476 } 1477 } 1478 return (error); 1479} 1480 1481static void 1482zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag) 1483{ 1484 rrw_exit(&zfsvfs->z_teardown_lock, tag); 1485 1486 if (zfsvfs->z_vfs) { 1487 VFS_RELE(zfsvfs->z_vfs); 1488 } else { 1489 dmu_objset_disown(zfsvfs->z_os, zfsvfs); 1490 zfsvfs_free(zfsvfs); 1491 } 1492} 1493 1494static int 1495zfs_ioc_pool_create(zfs_cmd_t *zc) 1496{ 1497 int error; 1498 nvlist_t *config, *props = NULL; 1499 nvlist_t *rootprops = NULL; 1500 nvlist_t *zplprops = NULL; 1501 1502 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1503 zc->zc_iflags, &config)) 1504 return (error); 1505 1506 if (zc->zc_nvlist_src_size != 0 && (error = 1507 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1508 zc->zc_iflags, &props))) { 1509 nvlist_free(config); 1510 return (error); 1511 } 1512 1513 if (props) { 1514 nvlist_t *nvl = NULL; 1515 uint64_t version = SPA_VERSION; 1516 1517 (void) nvlist_lookup_uint64(props, 1518 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version); 1519 if (!SPA_VERSION_IS_SUPPORTED(version)) { 1520 error = SET_ERROR(EINVAL); 1521 goto pool_props_bad; 1522 } 1523 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl); 1524 if (nvl) { 1525 error = nvlist_dup(nvl, &rootprops, KM_SLEEP); 1526 if (error != 0) { 1527 nvlist_free(config); 1528 nvlist_free(props); 1529 return (error); 1530 } 1531 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS); 1532 } 1533 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 1534 error = zfs_fill_zplprops_root(version, rootprops, 1535 zplprops, NULL); 1536 if (error != 0) 1537 goto pool_props_bad; 1538 } 1539 1540 error = spa_create(zc->zc_name, config, props, zplprops); 1541 1542 /* 1543 * Set the remaining root properties 1544 */ 1545 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name, 1546 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0) 1547 (void) spa_destroy(zc->zc_name); 1548 1549pool_props_bad: 1550 nvlist_free(rootprops); 1551 nvlist_free(zplprops); 1552 nvlist_free(config); 1553 nvlist_free(props); 1554 1555 return (error); 1556} 1557 1558static int 1559zfs_ioc_pool_destroy(zfs_cmd_t *zc) 1560{ 1561 int error; 1562 zfs_log_history(zc); 1563 error = spa_destroy(zc->zc_name); 1564 if (error == 0) 1565 zvol_remove_minors(zc->zc_name); 1566 return (error); 1567} 1568 1569static int 1570zfs_ioc_pool_import(zfs_cmd_t *zc) 1571{ 1572 nvlist_t *config, *props = NULL; 1573 uint64_t guid; 1574 int error; 1575 1576 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1577 zc->zc_iflags, &config)) != 0) 1578 return (error); 1579 1580 if (zc->zc_nvlist_src_size != 0 && (error = 1581 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1582 zc->zc_iflags, &props))) { 1583 nvlist_free(config); 1584 return (error); 1585 } 1586 1587 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 || 1588 guid != zc->zc_guid) 1589 error = SET_ERROR(EINVAL); 1590 else 1591 error = spa_import(zc->zc_name, config, props, zc->zc_cookie); 1592 1593 if (zc->zc_nvlist_dst != 0) { 1594 int err; 1595 1596 if ((err = put_nvlist(zc, config)) != 0) 1597 error = err; 1598 } 1599 1600 nvlist_free(config); 1601 1602 if (props) 1603 nvlist_free(props); 1604 1605 return (error); 1606} 1607 1608static int 1609zfs_ioc_pool_export(zfs_cmd_t *zc) 1610{ 1611 int error; 1612 boolean_t force = (boolean_t)zc->zc_cookie; 1613 boolean_t hardforce = (boolean_t)zc->zc_guid; 1614 1615 zfs_log_history(zc); 1616 error = spa_export(zc->zc_name, NULL, force, hardforce); 1617 if (error == 0) 1618 zvol_remove_minors(zc->zc_name); 1619 return (error); 1620} 1621 1622static int 1623zfs_ioc_pool_configs(zfs_cmd_t *zc) 1624{ 1625 nvlist_t *configs; 1626 int error; 1627 1628 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL) 1629 return (SET_ERROR(EEXIST)); 1630 1631 error = put_nvlist(zc, configs); 1632 1633 nvlist_free(configs); 1634 1635 return (error); 1636} 1637 1638/* 1639 * inputs: 1640 * zc_name name of the pool 1641 * 1642 * outputs: 1643 * zc_cookie real errno 1644 * zc_nvlist_dst config nvlist 1645 * zc_nvlist_dst_size size of config nvlist 1646 */ 1647static int 1648zfs_ioc_pool_stats(zfs_cmd_t *zc) 1649{ 1650 nvlist_t *config; 1651 int error; 1652 int ret = 0; 1653 1654 error = spa_get_stats(zc->zc_name, &config, zc->zc_value, 1655 sizeof (zc->zc_value)); 1656 1657 if (config != NULL) { 1658 ret = put_nvlist(zc, config); 1659 nvlist_free(config); 1660 1661 /* 1662 * The config may be present even if 'error' is non-zero. 1663 * In this case we return success, and preserve the real errno 1664 * in 'zc_cookie'. 1665 */ 1666 zc->zc_cookie = error; 1667 } else { 1668 ret = error; 1669 } 1670 1671 return (ret); 1672} 1673 1674/* 1675 * Try to import the given pool, returning pool stats as appropriate so that 1676 * user land knows which devices are available and overall pool health. 1677 */ 1678static int 1679zfs_ioc_pool_tryimport(zfs_cmd_t *zc) 1680{ 1681 nvlist_t *tryconfig, *config; 1682 int error; 1683 1684 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1685 zc->zc_iflags, &tryconfig)) != 0) 1686 return (error); 1687 1688 config = spa_tryimport(tryconfig); 1689 1690 nvlist_free(tryconfig); 1691 1692 if (config == NULL) 1693 return (SET_ERROR(EINVAL)); 1694 1695 error = put_nvlist(zc, config); 1696 nvlist_free(config); 1697 1698 return (error); 1699} 1700 1701/* 1702 * inputs: 1703 * zc_name name of the pool 1704 * zc_cookie scan func (pool_scan_func_t) 1705 */ 1706static int 1707zfs_ioc_pool_scan(zfs_cmd_t *zc) 1708{ 1709 spa_t *spa; 1710 int error; 1711 1712 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1713 return (error); 1714 1715 if (zc->zc_cookie == POOL_SCAN_NONE) 1716 error = spa_scan_stop(spa); 1717 else 1718 error = spa_scan(spa, zc->zc_cookie); 1719 1720 spa_close(spa, FTAG); 1721 1722 return (error); 1723} 1724 1725static int 1726zfs_ioc_pool_freeze(zfs_cmd_t *zc) 1727{ 1728 spa_t *spa; 1729 int error; 1730 1731 error = spa_open(zc->zc_name, &spa, FTAG); 1732 if (error == 0) { 1733 spa_freeze(spa); 1734 spa_close(spa, FTAG); 1735 } 1736 return (error); 1737} 1738 1739static int 1740zfs_ioc_pool_upgrade(zfs_cmd_t *zc) 1741{ 1742 spa_t *spa; 1743 int error; 1744 1745 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1746 return (error); 1747 1748 if (zc->zc_cookie < spa_version(spa) || 1749 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) { 1750 spa_close(spa, FTAG); 1751 return (SET_ERROR(EINVAL)); 1752 } 1753 1754 spa_upgrade(spa, zc->zc_cookie); 1755 spa_close(spa, FTAG); 1756 1757 return (error); 1758} 1759 1760static int 1761zfs_ioc_pool_get_history(zfs_cmd_t *zc) 1762{ 1763 spa_t *spa; 1764 char *hist_buf; 1765 uint64_t size; 1766 int error; 1767 1768 if ((size = zc->zc_history_len) == 0) 1769 return (SET_ERROR(EINVAL)); 1770 1771 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1772 return (error); 1773 1774 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) { 1775 spa_close(spa, FTAG); 1776 return (SET_ERROR(ENOTSUP)); 1777 } 1778 1779 hist_buf = kmem_alloc(size, KM_SLEEP); 1780 if ((error = spa_history_get(spa, &zc->zc_history_offset, 1781 &zc->zc_history_len, hist_buf)) == 0) { 1782 error = ddi_copyout(hist_buf, 1783 (void *)(uintptr_t)zc->zc_history, 1784 zc->zc_history_len, zc->zc_iflags); 1785 } 1786 1787 spa_close(spa, FTAG); 1788 kmem_free(hist_buf, size); 1789 return (error); 1790} 1791 1792static int 1793zfs_ioc_pool_reguid(zfs_cmd_t *zc) 1794{ 1795 spa_t *spa; 1796 int error; 1797 1798 error = spa_open(zc->zc_name, &spa, FTAG); 1799 if (error == 0) { 1800 error = spa_change_guid(spa); 1801 spa_close(spa, FTAG); 1802 } 1803 return (error); 1804} 1805 1806static int 1807zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc) 1808{ 1809 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value)); 1810} 1811 1812/* 1813 * inputs: 1814 * zc_name name of filesystem 1815 * zc_obj object to find 1816 * 1817 * outputs: 1818 * zc_value name of object 1819 */ 1820static int 1821zfs_ioc_obj_to_path(zfs_cmd_t *zc) 1822{ 1823 objset_t *os; 1824 int error; 1825 1826 /* XXX reading from objset not owned */ 1827 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0) 1828 return (error); 1829 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1830 dmu_objset_rele(os, FTAG); 1831 return (SET_ERROR(EINVAL)); 1832 } 1833 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value, 1834 sizeof (zc->zc_value)); 1835 dmu_objset_rele(os, FTAG); 1836 1837 return (error); 1838} 1839 1840/* 1841 * inputs: 1842 * zc_name name of filesystem 1843 * zc_obj object to find 1844 * 1845 * outputs: 1846 * zc_stat stats on object 1847 * zc_value path to object 1848 */ 1849static int 1850zfs_ioc_obj_to_stats(zfs_cmd_t *zc) 1851{ 1852 objset_t *os; 1853 int error; 1854 1855 /* XXX reading from objset not owned */ 1856 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0) 1857 return (error); 1858 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1859 dmu_objset_rele(os, FTAG); 1860 return (SET_ERROR(EINVAL)); 1861 } 1862 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value, 1863 sizeof (zc->zc_value)); 1864 dmu_objset_rele(os, FTAG); 1865 1866 return (error); 1867} 1868 1869static int 1870zfs_ioc_vdev_add(zfs_cmd_t *zc) 1871{ 1872 spa_t *spa; 1873 int error; 1874 nvlist_t *config, **l2cache, **spares; 1875 uint_t nl2cache = 0, nspares = 0; 1876 1877 error = spa_open(zc->zc_name, &spa, FTAG); 1878 if (error != 0) 1879 return (error); 1880 1881 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1882 zc->zc_iflags, &config); 1883 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE, 1884 &l2cache, &nl2cache); 1885 1886 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES, 1887 &spares, &nspares); 1888 1889#ifdef illumos 1890 /* 1891 * A root pool with concatenated devices is not supported. 1892 * Thus, can not add a device to a root pool. 1893 * 1894 * Intent log device can not be added to a rootpool because 1895 * during mountroot, zil is replayed, a seperated log device 1896 * can not be accessed during the mountroot time. 1897 * 1898 * l2cache and spare devices are ok to be added to a rootpool. 1899 */ 1900 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) { 1901 nvlist_free(config); 1902 spa_close(spa, FTAG); 1903 return (SET_ERROR(EDOM)); 1904 } 1905#endif /* illumos */ 1906 1907 if (error == 0) { 1908 error = spa_vdev_add(spa, config); 1909 nvlist_free(config); 1910 } 1911 spa_close(spa, FTAG); 1912 return (error); 1913} 1914 1915/* 1916 * inputs: 1917 * zc_name name of the pool 1918 * zc_nvlist_conf nvlist of devices to remove 1919 * zc_cookie to stop the remove? 1920 */ 1921static int 1922zfs_ioc_vdev_remove(zfs_cmd_t *zc) 1923{ 1924 spa_t *spa; 1925 int error; 1926 1927 error = spa_open(zc->zc_name, &spa, FTAG); 1928 if (error != 0) 1929 return (error); 1930 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE); 1931 spa_close(spa, FTAG); 1932 return (error); 1933} 1934 1935static int 1936zfs_ioc_vdev_set_state(zfs_cmd_t *zc) 1937{ 1938 spa_t *spa; 1939 int error; 1940 vdev_state_t newstate = VDEV_STATE_UNKNOWN; 1941 1942 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1943 return (error); 1944 switch (zc->zc_cookie) { 1945 case VDEV_STATE_ONLINE: 1946 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate); 1947 break; 1948 1949 case VDEV_STATE_OFFLINE: 1950 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj); 1951 break; 1952 1953 case VDEV_STATE_FAULTED: 1954 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED && 1955 zc->zc_obj != VDEV_AUX_EXTERNAL) 1956 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED; 1957 1958 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj); 1959 break; 1960 1961 case VDEV_STATE_DEGRADED: 1962 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED && 1963 zc->zc_obj != VDEV_AUX_EXTERNAL) 1964 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED; 1965 1966 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj); 1967 break; 1968 1969 default: 1970 error = SET_ERROR(EINVAL); 1971 } 1972 zc->zc_cookie = newstate; 1973 spa_close(spa, FTAG); 1974 return (error); 1975} 1976 1977static int 1978zfs_ioc_vdev_attach(zfs_cmd_t *zc) 1979{ 1980 spa_t *spa; 1981 int replacing = zc->zc_cookie; 1982 nvlist_t *config; 1983 int error; 1984 1985 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1986 return (error); 1987 1988 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1989 zc->zc_iflags, &config)) == 0) { 1990 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing); 1991 nvlist_free(config); 1992 } 1993 1994 spa_close(spa, FTAG); 1995 return (error); 1996} 1997 1998static int 1999zfs_ioc_vdev_detach(zfs_cmd_t *zc) 2000{ 2001 spa_t *spa; 2002 int error; 2003 2004 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 2005 return (error); 2006 2007 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE); 2008 2009 spa_close(spa, FTAG); 2010 return (error); 2011} 2012 2013static int 2014zfs_ioc_vdev_split(zfs_cmd_t *zc) 2015{ 2016 spa_t *spa; 2017 nvlist_t *config, *props = NULL; 2018 int error; 2019 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT); 2020 2021 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 2022 return (error); 2023 2024 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 2025 zc->zc_iflags, &config)) { 2026 spa_close(spa, FTAG); 2027 return (error); 2028 } 2029 2030 if (zc->zc_nvlist_src_size != 0 && (error = 2031 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2032 zc->zc_iflags, &props))) { 2033 spa_close(spa, FTAG); 2034 nvlist_free(config); 2035 return (error); 2036 } 2037 2038 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp); 2039 2040 spa_close(spa, FTAG); 2041 2042 nvlist_free(config); 2043 nvlist_free(props); 2044 2045 return (error); 2046} 2047 2048static int 2049zfs_ioc_vdev_setpath(zfs_cmd_t *zc) 2050{ 2051 spa_t *spa; 2052 char *path = zc->zc_value; 2053 uint64_t guid = zc->zc_guid; 2054 int error; 2055 2056 error = spa_open(zc->zc_name, &spa, FTAG); 2057 if (error != 0) 2058 return (error); 2059 2060 error = spa_vdev_setpath(spa, guid, path); 2061 spa_close(spa, FTAG); 2062 return (error); 2063} 2064 2065static int 2066zfs_ioc_vdev_setfru(zfs_cmd_t *zc) 2067{ 2068 spa_t *spa; 2069 char *fru = zc->zc_value; 2070 uint64_t guid = zc->zc_guid; 2071 int error; 2072 2073 error = spa_open(zc->zc_name, &spa, FTAG); 2074 if (error != 0) 2075 return (error); 2076 2077 error = spa_vdev_setfru(spa, guid, fru); 2078 spa_close(spa, FTAG); 2079 return (error); 2080} 2081 2082static int 2083zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os) 2084{ 2085 int error = 0; 2086 nvlist_t *nv; 2087 2088 dmu_objset_fast_stat(os, &zc->zc_objset_stats); 2089 2090 if (zc->zc_nvlist_dst != 0 && 2091 (error = dsl_prop_get_all(os, &nv)) == 0) { 2092 dmu_objset_stats(os, nv); 2093 /* 2094 * NB: zvol_get_stats() will read the objset contents, 2095 * which we aren't supposed to do with a 2096 * DS_MODE_USER hold, because it could be 2097 * inconsistent. So this is a bit of a workaround... 2098 * XXX reading with out owning 2099 */ 2100 if (!zc->zc_objset_stats.dds_inconsistent && 2101 dmu_objset_type(os) == DMU_OST_ZVOL) { 2102 error = zvol_get_stats(os, nv); 2103 if (error == EIO) 2104 return (error); 2105 VERIFY0(error); 2106 } 2107 error = put_nvlist(zc, nv); 2108 nvlist_free(nv); 2109 } 2110 2111 return (error); 2112} 2113 2114/* 2115 * inputs: 2116 * zc_name name of filesystem 2117 * zc_nvlist_dst_size size of buffer for property nvlist 2118 * 2119 * outputs: 2120 * zc_objset_stats stats 2121 * zc_nvlist_dst property nvlist 2122 * zc_nvlist_dst_size size of property nvlist 2123 */ 2124static int 2125zfs_ioc_objset_stats(zfs_cmd_t *zc) 2126{ 2127 objset_t *os; 2128 int error; 2129 2130 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 2131 if (error == 0) { 2132 error = zfs_ioc_objset_stats_impl(zc, os); 2133 dmu_objset_rele(os, FTAG); 2134 } 2135 2136 if (error == ENOMEM) 2137 error = 0; 2138 return (error); 2139} 2140 2141/* 2142 * inputs: 2143 * zc_name name of filesystem 2144 * zc_nvlist_dst_size size of buffer for property nvlist 2145 * 2146 * outputs: 2147 * zc_nvlist_dst received property nvlist 2148 * zc_nvlist_dst_size size of received property nvlist 2149 * 2150 * Gets received properties (distinct from local properties on or after 2151 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from 2152 * local property values. 2153 */ 2154static int 2155zfs_ioc_objset_recvd_props(zfs_cmd_t *zc) 2156{ 2157 int error = 0; 2158 nvlist_t *nv; 2159 2160 /* 2161 * Without this check, we would return local property values if the 2162 * caller has not already received properties on or after 2163 * SPA_VERSION_RECVD_PROPS. 2164 */ 2165 if (!dsl_prop_get_hasrecvd(zc->zc_name)) 2166 return (SET_ERROR(ENOTSUP)); 2167 2168 if (zc->zc_nvlist_dst != 0 && 2169 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) { 2170 error = put_nvlist(zc, nv); 2171 nvlist_free(nv); 2172 } 2173 2174 return (error); 2175} 2176 2177static int 2178nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop) 2179{ 2180 uint64_t value; 2181 int error; 2182 2183 /* 2184 * zfs_get_zplprop() will either find a value or give us 2185 * the default value (if there is one). 2186 */ 2187 if ((error = zfs_get_zplprop(os, prop, &value)) != 0) 2188 return (error); 2189 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0); 2190 return (0); 2191} 2192 2193/* 2194 * inputs: 2195 * zc_name name of filesystem 2196 * zc_nvlist_dst_size size of buffer for zpl property nvlist 2197 * 2198 * outputs: 2199 * zc_nvlist_dst zpl property nvlist 2200 * zc_nvlist_dst_size size of zpl property nvlist 2201 */ 2202static int 2203zfs_ioc_objset_zplprops(zfs_cmd_t *zc) 2204{ 2205 objset_t *os; 2206 int err; 2207 2208 /* XXX reading without owning */ 2209 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os)) 2210 return (err); 2211 2212 dmu_objset_fast_stat(os, &zc->zc_objset_stats); 2213 2214 /* 2215 * NB: nvl_add_zplprop() will read the objset contents, 2216 * which we aren't supposed to do with a DS_MODE_USER 2217 * hold, because it could be inconsistent. 2218 */ 2219 if (zc->zc_nvlist_dst != 0 && 2220 !zc->zc_objset_stats.dds_inconsistent && 2221 dmu_objset_type(os) == DMU_OST_ZFS) { 2222 nvlist_t *nv; 2223 2224 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2225 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 && 2226 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 && 2227 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 && 2228 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0) 2229 err = put_nvlist(zc, nv); 2230 nvlist_free(nv); 2231 } else { 2232 err = SET_ERROR(ENOENT); 2233 } 2234 dmu_objset_rele(os, FTAG); 2235 return (err); 2236} 2237 2238boolean_t 2239dataset_name_hidden(const char *name) 2240{ 2241 /* 2242 * Skip over datasets that are not visible in this zone, 2243 * internal datasets (which have a $ in their name), and 2244 * temporary datasets (which have a % in their name). 2245 */ 2246 if (strchr(name, '$') != NULL) 2247 return (B_TRUE); 2248 if (strchr(name, '%') != NULL) 2249 return (B_TRUE); 2250 if (!INGLOBALZONE(curthread) && !zone_dataset_visible(name, NULL)) 2251 return (B_TRUE); 2252 return (B_FALSE); 2253} 2254 2255/* 2256 * inputs: 2257 * zc_name name of filesystem 2258 * zc_cookie zap cursor 2259 * zc_nvlist_dst_size size of buffer for property nvlist 2260 * 2261 * outputs: 2262 * zc_name name of next filesystem 2263 * zc_cookie zap cursor 2264 * zc_objset_stats stats 2265 * zc_nvlist_dst property nvlist 2266 * zc_nvlist_dst_size size of property nvlist 2267 */ 2268static int 2269zfs_ioc_dataset_list_next(zfs_cmd_t *zc) 2270{ 2271 objset_t *os; 2272 int error; 2273 char *p; 2274 size_t orig_len = strlen(zc->zc_name); 2275 2276top: 2277 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) { 2278 if (error == ENOENT) 2279 error = SET_ERROR(ESRCH); 2280 return (error); 2281 } 2282 2283 p = strrchr(zc->zc_name, '/'); 2284 if (p == NULL || p[1] != '\0') 2285 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name)); 2286 p = zc->zc_name + strlen(zc->zc_name); 2287 2288 do { 2289 error = dmu_dir_list_next(os, 2290 sizeof (zc->zc_name) - (p - zc->zc_name), p, 2291 NULL, &zc->zc_cookie); 2292 if (error == ENOENT) 2293 error = SET_ERROR(ESRCH); 2294 } while (error == 0 && dataset_name_hidden(zc->zc_name)); 2295 dmu_objset_rele(os, FTAG); 2296 2297 /* 2298 * If it's an internal dataset (ie. with a '$' in its name), 2299 * don't try to get stats for it, otherwise we'll return ENOENT. 2300 */ 2301 if (error == 0 && strchr(zc->zc_name, '$') == NULL) { 2302 error = zfs_ioc_objset_stats(zc); /* fill in the stats */ 2303 if (error == ENOENT) { 2304 /* We lost a race with destroy, get the next one. */ 2305 zc->zc_name[orig_len] = '\0'; 2306 goto top; 2307 } 2308 } 2309 return (error); 2310} 2311 2312/* 2313 * inputs: 2314 * zc_name name of filesystem 2315 * zc_cookie zap cursor 2316 * zc_nvlist_dst_size size of buffer for property nvlist 2317 * zc_simple when set, only name is requested 2318 * 2319 * outputs: 2320 * zc_name name of next snapshot 2321 * zc_objset_stats stats 2322 * zc_nvlist_dst property nvlist 2323 * zc_nvlist_dst_size size of property nvlist 2324 */ 2325static int 2326zfs_ioc_snapshot_list_next(zfs_cmd_t *zc) 2327{ 2328 objset_t *os; 2329 int error; 2330 2331 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 2332 if (error != 0) { 2333 return (error == ENOENT ? ESRCH : error); 2334 } 2335 2336 /* 2337 * A dataset name of maximum length cannot have any snapshots, 2338 * so exit immediately. 2339 */ 2340 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) { 2341 dmu_objset_rele(os, FTAG); 2342 return (SET_ERROR(ESRCH)); 2343 } 2344 2345 error = dmu_snapshot_list_next(os, 2346 sizeof (zc->zc_name) - strlen(zc->zc_name), 2347 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie, 2348 NULL); 2349 2350 if (error == 0 && !zc->zc_simple) { 2351 dsl_dataset_t *ds; 2352 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool; 2353 2354 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds); 2355 if (error == 0) { 2356 objset_t *ossnap; 2357 2358 error = dmu_objset_from_ds(ds, &ossnap); 2359 if (error == 0) 2360 error = zfs_ioc_objset_stats_impl(zc, ossnap); 2361 dsl_dataset_rele(ds, FTAG); 2362 } 2363 } else if (error == ENOENT) { 2364 error = SET_ERROR(ESRCH); 2365 } 2366 2367 dmu_objset_rele(os, FTAG); 2368 /* if we failed, undo the @ that we tacked on to zc_name */ 2369 if (error != 0) 2370 *strchr(zc->zc_name, '@') = '\0'; 2371 return (error); 2372} 2373 2374static int 2375zfs_prop_set_userquota(const char *dsname, nvpair_t *pair) 2376{ 2377 const char *propname = nvpair_name(pair); 2378 uint64_t *valary; 2379 unsigned int vallen; 2380 const char *domain; 2381 char *dash; 2382 zfs_userquota_prop_t type; 2383 uint64_t rid; 2384 uint64_t quota; 2385 zfsvfs_t *zfsvfs; 2386 int err; 2387 2388 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2389 nvlist_t *attrs; 2390 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 2391 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2392 &pair) != 0) 2393 return (SET_ERROR(EINVAL)); 2394 } 2395 2396 /* 2397 * A correctly constructed propname is encoded as 2398 * userquota@<rid>-<domain>. 2399 */ 2400 if ((dash = strchr(propname, '-')) == NULL || 2401 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 || 2402 vallen != 3) 2403 return (SET_ERROR(EINVAL)); 2404 2405 domain = dash + 1; 2406 type = valary[0]; 2407 rid = valary[1]; 2408 quota = valary[2]; 2409 2410 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE); 2411 if (err == 0) { 2412 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota); 2413 zfsvfs_rele(zfsvfs, FTAG); 2414 } 2415 2416 return (err); 2417} 2418 2419/* 2420 * If the named property is one that has a special function to set its value, 2421 * return 0 on success and a positive error code on failure; otherwise if it is 2422 * not one of the special properties handled by this function, return -1. 2423 * 2424 * XXX: It would be better for callers of the property interface if we handled 2425 * these special cases in dsl_prop.c (in the dsl layer). 2426 */ 2427static int 2428zfs_prop_set_special(const char *dsname, zprop_source_t source, 2429 nvpair_t *pair) 2430{ 2431 const char *propname = nvpair_name(pair); 2432 zfs_prop_t prop = zfs_name_to_prop(propname); 2433 uint64_t intval; 2434 int err; 2435 2436 if (prop == ZPROP_INVAL) { 2437 if (zfs_prop_userquota(propname)) 2438 return (zfs_prop_set_userquota(dsname, pair)); 2439 return (-1); 2440 } 2441 2442 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2443 nvlist_t *attrs; 2444 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 2445 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2446 &pair) == 0); 2447 } 2448 2449 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING) 2450 return (-1); 2451 2452 VERIFY(0 == nvpair_value_uint64(pair, &intval)); 2453 2454 switch (prop) { 2455 case ZFS_PROP_QUOTA: 2456 err = dsl_dir_set_quota(dsname, source, intval); 2457 break; 2458 case ZFS_PROP_REFQUOTA: 2459 err = dsl_dataset_set_refquota(dsname, source, intval); 2460 break; 2461 case ZFS_PROP_FILESYSTEM_LIMIT: 2462 case ZFS_PROP_SNAPSHOT_LIMIT: 2463 if (intval == UINT64_MAX) { 2464 /* clearing the limit, just do it */ 2465 err = 0; 2466 } else { 2467 err = dsl_dir_activate_fs_ss_limit(dsname); 2468 } 2469 /* 2470 * Set err to -1 to force the zfs_set_prop_nvlist code down the 2471 * default path to set the value in the nvlist. 2472 */ 2473 if (err == 0) 2474 err = -1; 2475 break; 2476 case ZFS_PROP_RESERVATION: 2477 err = dsl_dir_set_reservation(dsname, source, intval); 2478 break; 2479 case ZFS_PROP_REFRESERVATION: 2480 err = dsl_dataset_set_refreservation(dsname, source, intval); 2481 break; 2482 case ZFS_PROP_VOLSIZE: 2483 err = zvol_set_volsize(dsname, ddi_driver_major(zfs_dip), 2484 intval); 2485 break; 2486 case ZFS_PROP_VERSION: 2487 { 2488 zfsvfs_t *zfsvfs; 2489 2490 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0) 2491 break; 2492 2493 err = zfs_set_version(zfsvfs, intval); 2494 zfsvfs_rele(zfsvfs, FTAG); 2495 2496 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) { 2497 zfs_cmd_t *zc; 2498 2499 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP); 2500 (void) strcpy(zc->zc_name, dsname); 2501 (void) zfs_ioc_userspace_upgrade(zc); 2502 kmem_free(zc, sizeof (zfs_cmd_t)); 2503 } 2504 break; 2505 } 2506 case ZFS_PROP_COMPRESSION: 2507 { 2508 if (intval == ZIO_COMPRESS_LZ4) { 2509 spa_t *spa; 2510 2511 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 2512 return (err); 2513 2514 /* 2515 * Setting the LZ4 compression algorithm activates 2516 * the feature. 2517 */ 2518 if (!spa_feature_is_active(spa, 2519 SPA_FEATURE_LZ4_COMPRESS)) { 2520 if ((err = zfs_prop_activate_feature(spa, 2521 SPA_FEATURE_LZ4_COMPRESS)) != 0) { 2522 spa_close(spa, FTAG); 2523 return (err); 2524 } 2525 } 2526 2527 spa_close(spa, FTAG); 2528 } 2529 /* 2530 * We still want the default set action to be performed in the 2531 * caller, we only performed zfeature settings here. 2532 */ 2533 err = -1; 2534 break; 2535 } 2536 2537 default: 2538 err = -1; 2539 } 2540 2541 return (err); 2542} 2543 2544/* 2545 * This function is best effort. If it fails to set any of the given properties, 2546 * it continues to set as many as it can and returns the last error 2547 * encountered. If the caller provides a non-NULL errlist, it will be filled in 2548 * with the list of names of all the properties that failed along with the 2549 * corresponding error numbers. 2550 * 2551 * If every property is set successfully, zero is returned and errlist is not 2552 * modified. 2553 */ 2554int 2555zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl, 2556 nvlist_t *errlist) 2557{ 2558 nvpair_t *pair; 2559 nvpair_t *propval; 2560 int rv = 0; 2561 uint64_t intval; 2562 char *strval; 2563 nvlist_t *genericnvl = fnvlist_alloc(); 2564 nvlist_t *retrynvl = fnvlist_alloc(); 2565 2566retry: 2567 pair = NULL; 2568 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { 2569 const char *propname = nvpair_name(pair); 2570 zfs_prop_t prop = zfs_name_to_prop(propname); 2571 int err = 0; 2572 2573 /* decode the property value */ 2574 propval = pair; 2575 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2576 nvlist_t *attrs; 2577 attrs = fnvpair_value_nvlist(pair); 2578 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2579 &propval) != 0) 2580 err = SET_ERROR(EINVAL); 2581 } 2582 2583 /* Validate value type */ 2584 if (err == 0 && prop == ZPROP_INVAL) { 2585 if (zfs_prop_user(propname)) { 2586 if (nvpair_type(propval) != DATA_TYPE_STRING) 2587 err = SET_ERROR(EINVAL); 2588 } else if (zfs_prop_userquota(propname)) { 2589 if (nvpair_type(propval) != 2590 DATA_TYPE_UINT64_ARRAY) 2591 err = SET_ERROR(EINVAL); 2592 } else { 2593 err = SET_ERROR(EINVAL); 2594 } 2595 } else if (err == 0) { 2596 if (nvpair_type(propval) == DATA_TYPE_STRING) { 2597 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING) 2598 err = SET_ERROR(EINVAL); 2599 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) { 2600 const char *unused; 2601 2602 intval = fnvpair_value_uint64(propval); 2603 2604 switch (zfs_prop_get_type(prop)) { 2605 case PROP_TYPE_NUMBER: 2606 break; 2607 case PROP_TYPE_STRING: 2608 err = SET_ERROR(EINVAL); 2609 break; 2610 case PROP_TYPE_INDEX: 2611 if (zfs_prop_index_to_string(prop, 2612 intval, &unused) != 0) 2613 err = SET_ERROR(EINVAL); 2614 break; 2615 default: 2616 cmn_err(CE_PANIC, 2617 "unknown property type"); 2618 } 2619 } else { 2620 err = SET_ERROR(EINVAL); 2621 } 2622 } 2623 2624 /* Validate permissions */ 2625 if (err == 0) 2626 err = zfs_check_settable(dsname, pair, CRED()); 2627 2628 if (err == 0) { 2629 err = zfs_prop_set_special(dsname, source, pair); 2630 if (err == -1) { 2631 /* 2632 * For better performance we build up a list of 2633 * properties to set in a single transaction. 2634 */ 2635 err = nvlist_add_nvpair(genericnvl, pair); 2636 } else if (err != 0 && nvl != retrynvl) { 2637 /* 2638 * This may be a spurious error caused by 2639 * receiving quota and reservation out of order. 2640 * Try again in a second pass. 2641 */ 2642 err = nvlist_add_nvpair(retrynvl, pair); 2643 } 2644 } 2645 2646 if (err != 0) { 2647 if (errlist != NULL) 2648 fnvlist_add_int32(errlist, propname, err); 2649 rv = err; 2650 } 2651 } 2652 2653 if (nvl != retrynvl && !nvlist_empty(retrynvl)) { 2654 nvl = retrynvl; 2655 goto retry; 2656 } 2657 2658 if (!nvlist_empty(genericnvl) && 2659 dsl_props_set(dsname, source, genericnvl) != 0) { 2660 /* 2661 * If this fails, we still want to set as many properties as we 2662 * can, so try setting them individually. 2663 */ 2664 pair = NULL; 2665 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) { 2666 const char *propname = nvpair_name(pair); 2667 int err = 0; 2668 2669 propval = pair; 2670 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2671 nvlist_t *attrs; 2672 attrs = fnvpair_value_nvlist(pair); 2673 propval = fnvlist_lookup_nvpair(attrs, 2674 ZPROP_VALUE); 2675 } 2676 2677 if (nvpair_type(propval) == DATA_TYPE_STRING) { 2678 strval = fnvpair_value_string(propval); 2679 err = dsl_prop_set_string(dsname, propname, 2680 source, strval); 2681 } else { 2682 intval = fnvpair_value_uint64(propval); 2683 err = dsl_prop_set_int(dsname, propname, source, 2684 intval); 2685 } 2686 2687 if (err != 0) { 2688 if (errlist != NULL) { 2689 fnvlist_add_int32(errlist, propname, 2690 err); 2691 } 2692 rv = err; 2693 } 2694 } 2695 } 2696 nvlist_free(genericnvl); 2697 nvlist_free(retrynvl); 2698 2699 return (rv); 2700} 2701 2702/* 2703 * Check that all the properties are valid user properties. 2704 */ 2705static int 2706zfs_check_userprops(const char *fsname, nvlist_t *nvl) 2707{ 2708 nvpair_t *pair = NULL; 2709 int error = 0; 2710 2711 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { 2712 const char *propname = nvpair_name(pair); 2713 2714 if (!zfs_prop_user(propname) || 2715 nvpair_type(pair) != DATA_TYPE_STRING) 2716 return (SET_ERROR(EINVAL)); 2717 2718 if (error = zfs_secpolicy_write_perms(fsname, 2719 ZFS_DELEG_PERM_USERPROP, CRED())) 2720 return (error); 2721 2722 if (strlen(propname) >= ZAP_MAXNAMELEN) 2723 return (SET_ERROR(ENAMETOOLONG)); 2724 2725 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN) 2726 return (E2BIG); 2727 } 2728 return (0); 2729} 2730 2731static void 2732props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops) 2733{ 2734 nvpair_t *pair; 2735 2736 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2737 2738 pair = NULL; 2739 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) { 2740 if (nvlist_exists(skipped, nvpair_name(pair))) 2741 continue; 2742 2743 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0); 2744 } 2745} 2746 2747static int 2748clear_received_props(const char *dsname, nvlist_t *props, 2749 nvlist_t *skipped) 2750{ 2751 int err = 0; 2752 nvlist_t *cleared_props = NULL; 2753 props_skip(props, skipped, &cleared_props); 2754 if (!nvlist_empty(cleared_props)) { 2755 /* 2756 * Acts on local properties until the dataset has received 2757 * properties at least once on or after SPA_VERSION_RECVD_PROPS. 2758 */ 2759 zprop_source_t flags = (ZPROP_SRC_NONE | 2760 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0)); 2761 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL); 2762 } 2763 nvlist_free(cleared_props); 2764 return (err); 2765} 2766 2767/* 2768 * inputs: 2769 * zc_name name of filesystem 2770 * zc_value name of property to set 2771 * zc_nvlist_src{_size} nvlist of properties to apply 2772 * zc_cookie received properties flag 2773 * 2774 * outputs: 2775 * zc_nvlist_dst{_size} error for each unapplied received property 2776 */ 2777static int 2778zfs_ioc_set_prop(zfs_cmd_t *zc) 2779{ 2780 nvlist_t *nvl; 2781 boolean_t received = zc->zc_cookie; 2782 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED : 2783 ZPROP_SRC_LOCAL); 2784 nvlist_t *errors; 2785 int error; 2786 2787 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2788 zc->zc_iflags, &nvl)) != 0) 2789 return (error); 2790 2791 if (received) { 2792 nvlist_t *origprops; 2793 2794 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) { 2795 (void) clear_received_props(zc->zc_name, 2796 origprops, nvl); 2797 nvlist_free(origprops); 2798 } 2799 2800 error = dsl_prop_set_hasrecvd(zc->zc_name); 2801 } 2802 2803 errors = fnvlist_alloc(); 2804 if (error == 0) 2805 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors); 2806 2807 if (zc->zc_nvlist_dst != 0 && errors != NULL) { 2808 (void) put_nvlist(zc, errors); 2809 } 2810 2811 nvlist_free(errors); 2812 nvlist_free(nvl); 2813 return (error); 2814} 2815 2816/* 2817 * inputs: 2818 * zc_name name of filesystem 2819 * zc_value name of property to inherit 2820 * zc_cookie revert to received value if TRUE 2821 * 2822 * outputs: none 2823 */ 2824static int 2825zfs_ioc_inherit_prop(zfs_cmd_t *zc) 2826{ 2827 const char *propname = zc->zc_value; 2828 zfs_prop_t prop = zfs_name_to_prop(propname); 2829 boolean_t received = zc->zc_cookie; 2830 zprop_source_t source = (received 2831 ? ZPROP_SRC_NONE /* revert to received value, if any */ 2832 : ZPROP_SRC_INHERITED); /* explicitly inherit */ 2833 2834 if (received) { 2835 nvlist_t *dummy; 2836 nvpair_t *pair; 2837 zprop_type_t type; 2838 int err; 2839 2840 /* 2841 * zfs_prop_set_special() expects properties in the form of an 2842 * nvpair with type info. 2843 */ 2844 if (prop == ZPROP_INVAL) { 2845 if (!zfs_prop_user(propname)) 2846 return (SET_ERROR(EINVAL)); 2847 2848 type = PROP_TYPE_STRING; 2849 } else if (prop == ZFS_PROP_VOLSIZE || 2850 prop == ZFS_PROP_VERSION) { 2851 return (SET_ERROR(EINVAL)); 2852 } else { 2853 type = zfs_prop_get_type(prop); 2854 } 2855 2856 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2857 2858 switch (type) { 2859 case PROP_TYPE_STRING: 2860 VERIFY(0 == nvlist_add_string(dummy, propname, "")); 2861 break; 2862 case PROP_TYPE_NUMBER: 2863 case PROP_TYPE_INDEX: 2864 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0)); 2865 break; 2866 default: 2867 nvlist_free(dummy); 2868 return (SET_ERROR(EINVAL)); 2869 } 2870 2871 pair = nvlist_next_nvpair(dummy, NULL); 2872 err = zfs_prop_set_special(zc->zc_name, source, pair); 2873 nvlist_free(dummy); 2874 if (err != -1) 2875 return (err); /* special property already handled */ 2876 } else { 2877 /* 2878 * Only check this in the non-received case. We want to allow 2879 * 'inherit -S' to revert non-inheritable properties like quota 2880 * and reservation to the received or default values even though 2881 * they are not considered inheritable. 2882 */ 2883 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop)) 2884 return (SET_ERROR(EINVAL)); 2885 } 2886 2887 /* property name has been validated by zfs_secpolicy_inherit_prop() */ 2888 return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source)); 2889} 2890 2891static int 2892zfs_ioc_pool_set_props(zfs_cmd_t *zc) 2893{ 2894 nvlist_t *props; 2895 spa_t *spa; 2896 int error; 2897 nvpair_t *pair; 2898 2899 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2900 zc->zc_iflags, &props)) 2901 return (error); 2902 2903 /* 2904 * If the only property is the configfile, then just do a spa_lookup() 2905 * to handle the faulted case. 2906 */ 2907 pair = nvlist_next_nvpair(props, NULL); 2908 if (pair != NULL && strcmp(nvpair_name(pair), 2909 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 && 2910 nvlist_next_nvpair(props, pair) == NULL) { 2911 mutex_enter(&spa_namespace_lock); 2912 if ((spa = spa_lookup(zc->zc_name)) != NULL) { 2913 spa_configfile_set(spa, props, B_FALSE); 2914 spa_config_sync(spa, B_FALSE, B_TRUE); 2915 } 2916 mutex_exit(&spa_namespace_lock); 2917 if (spa != NULL) { 2918 nvlist_free(props); 2919 return (0); 2920 } 2921 } 2922 2923 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { 2924 nvlist_free(props); 2925 return (error); 2926 } 2927 2928 error = spa_prop_set(spa, props); 2929 2930 nvlist_free(props); 2931 spa_close(spa, FTAG); 2932 2933 return (error); 2934} 2935 2936static int 2937zfs_ioc_pool_get_props(zfs_cmd_t *zc) 2938{ 2939 spa_t *spa; 2940 int error; 2941 nvlist_t *nvp = NULL; 2942 2943 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { 2944 /* 2945 * If the pool is faulted, there may be properties we can still 2946 * get (such as altroot and cachefile), so attempt to get them 2947 * anyway. 2948 */ 2949 mutex_enter(&spa_namespace_lock); 2950 if ((spa = spa_lookup(zc->zc_name)) != NULL) 2951 error = spa_prop_get(spa, &nvp); 2952 mutex_exit(&spa_namespace_lock); 2953 } else { 2954 error = spa_prop_get(spa, &nvp); 2955 spa_close(spa, FTAG); 2956 } 2957 2958 if (error == 0 && zc->zc_nvlist_dst != 0) 2959 error = put_nvlist(zc, nvp); 2960 else 2961 error = SET_ERROR(EFAULT); 2962 2963 nvlist_free(nvp); 2964 return (error); 2965} 2966 2967/* 2968 * inputs: 2969 * zc_name name of filesystem 2970 * zc_nvlist_src{_size} nvlist of delegated permissions 2971 * zc_perm_action allow/unallow flag 2972 * 2973 * outputs: none 2974 */ 2975static int 2976zfs_ioc_set_fsacl(zfs_cmd_t *zc) 2977{ 2978 int error; 2979 nvlist_t *fsaclnv = NULL; 2980 2981 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2982 zc->zc_iflags, &fsaclnv)) != 0) 2983 return (error); 2984 2985 /* 2986 * Verify nvlist is constructed correctly 2987 */ 2988 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) { 2989 nvlist_free(fsaclnv); 2990 return (SET_ERROR(EINVAL)); 2991 } 2992 2993 /* 2994 * If we don't have PRIV_SYS_MOUNT, then validate 2995 * that user is allowed to hand out each permission in 2996 * the nvlist(s) 2997 */ 2998 2999 error = secpolicy_zfs(CRED()); 3000 if (error != 0) { 3001 if (zc->zc_perm_action == B_FALSE) { 3002 error = dsl_deleg_can_allow(zc->zc_name, 3003 fsaclnv, CRED()); 3004 } else { 3005 error = dsl_deleg_can_unallow(zc->zc_name, 3006 fsaclnv, CRED()); 3007 } 3008 } 3009 3010 if (error == 0) 3011 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action); 3012 3013 nvlist_free(fsaclnv); 3014 return (error); 3015} 3016 3017/* 3018 * inputs: 3019 * zc_name name of filesystem 3020 * 3021 * outputs: 3022 * zc_nvlist_src{_size} nvlist of delegated permissions 3023 */ 3024static int 3025zfs_ioc_get_fsacl(zfs_cmd_t *zc) 3026{ 3027 nvlist_t *nvp; 3028 int error; 3029 3030 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) { 3031 error = put_nvlist(zc, nvp); 3032 nvlist_free(nvp); 3033 } 3034 3035 return (error); 3036} 3037 3038/* 3039 * Search the vfs list for a specified resource. Returns a pointer to it 3040 * or NULL if no suitable entry is found. The caller of this routine 3041 * is responsible for releasing the returned vfs pointer. 3042 */ 3043static vfs_t * 3044zfs_get_vfs(const char *resource) 3045{ 3046 vfs_t *vfsp; 3047 3048 mtx_lock(&mountlist_mtx); 3049 TAILQ_FOREACH(vfsp, &mountlist, mnt_list) { 3050 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) { 3051 VFS_HOLD(vfsp); 3052 break; 3053 } 3054 } 3055 mtx_unlock(&mountlist_mtx); 3056 return (vfsp); 3057} 3058 3059/* ARGSUSED */ 3060static void 3061zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) 3062{ 3063 zfs_creat_t *zct = arg; 3064 3065 zfs_create_fs(os, cr, zct->zct_zplprops, tx); 3066} 3067 3068#define ZFS_PROP_UNDEFINED ((uint64_t)-1) 3069 3070/* 3071 * inputs: 3072 * os parent objset pointer (NULL if root fs) 3073 * fuids_ok fuids allowed in this version of the spa? 3074 * sa_ok SAs allowed in this version of the spa? 3075 * createprops list of properties requested by creator 3076 * 3077 * outputs: 3078 * zplprops values for the zplprops we attach to the master node object 3079 * is_ci true if requested file system will be purely case-insensitive 3080 * 3081 * Determine the settings for utf8only, normalization and 3082 * casesensitivity. Specific values may have been requested by the 3083 * creator and/or we can inherit values from the parent dataset. If 3084 * the file system is of too early a vintage, a creator can not 3085 * request settings for these properties, even if the requested 3086 * setting is the default value. We don't actually want to create dsl 3087 * properties for these, so remove them from the source nvlist after 3088 * processing. 3089 */ 3090static int 3091zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver, 3092 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops, 3093 nvlist_t *zplprops, boolean_t *is_ci) 3094{ 3095 uint64_t sense = ZFS_PROP_UNDEFINED; 3096 uint64_t norm = ZFS_PROP_UNDEFINED; 3097 uint64_t u8 = ZFS_PROP_UNDEFINED; 3098 3099 ASSERT(zplprops != NULL); 3100 3101 /* 3102 * Pull out creator prop choices, if any. 3103 */ 3104 if (createprops) { 3105 (void) nvlist_lookup_uint64(createprops, 3106 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver); 3107 (void) nvlist_lookup_uint64(createprops, 3108 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm); 3109 (void) nvlist_remove_all(createprops, 3110 zfs_prop_to_name(ZFS_PROP_NORMALIZE)); 3111 (void) nvlist_lookup_uint64(createprops, 3112 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8); 3113 (void) nvlist_remove_all(createprops, 3114 zfs_prop_to_name(ZFS_PROP_UTF8ONLY)); 3115 (void) nvlist_lookup_uint64(createprops, 3116 zfs_prop_to_name(ZFS_PROP_CASE), &sense); 3117 (void) nvlist_remove_all(createprops, 3118 zfs_prop_to_name(ZFS_PROP_CASE)); 3119 } 3120 3121 /* 3122 * If the zpl version requested is whacky or the file system 3123 * or pool is version is too "young" to support normalization 3124 * and the creator tried to set a value for one of the props, 3125 * error out. 3126 */ 3127 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) || 3128 (zplver >= ZPL_VERSION_FUID && !fuids_ok) || 3129 (zplver >= ZPL_VERSION_SA && !sa_ok) || 3130 (zplver < ZPL_VERSION_NORMALIZATION && 3131 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED || 3132 sense != ZFS_PROP_UNDEFINED))) 3133 return (SET_ERROR(ENOTSUP)); 3134 3135 /* 3136 * Put the version in the zplprops 3137 */ 3138 VERIFY(nvlist_add_uint64(zplprops, 3139 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0); 3140 3141 if (norm == ZFS_PROP_UNDEFINED) 3142 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0); 3143 VERIFY(nvlist_add_uint64(zplprops, 3144 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0); 3145 3146 /* 3147 * If we're normalizing, names must always be valid UTF-8 strings. 3148 */ 3149 if (norm) 3150 u8 = 1; 3151 if (u8 == ZFS_PROP_UNDEFINED) 3152 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0); 3153 VERIFY(nvlist_add_uint64(zplprops, 3154 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0); 3155 3156 if (sense == ZFS_PROP_UNDEFINED) 3157 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0); 3158 VERIFY(nvlist_add_uint64(zplprops, 3159 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0); 3160 3161 if (is_ci) 3162 *is_ci = (sense == ZFS_CASE_INSENSITIVE); 3163 3164 return (0); 3165} 3166 3167static int 3168zfs_fill_zplprops(const char *dataset, nvlist_t *createprops, 3169 nvlist_t *zplprops, boolean_t *is_ci) 3170{ 3171 boolean_t fuids_ok, sa_ok; 3172 uint64_t zplver = ZPL_VERSION; 3173 objset_t *os = NULL; 3174 char parentname[MAXNAMELEN]; 3175 char *cp; 3176 spa_t *spa; 3177 uint64_t spa_vers; 3178 int error; 3179 3180 (void) strlcpy(parentname, dataset, sizeof (parentname)); 3181 cp = strrchr(parentname, '/'); 3182 ASSERT(cp != NULL); 3183 cp[0] = '\0'; 3184 3185 if ((error = spa_open(dataset, &spa, FTAG)) != 0) 3186 return (error); 3187 3188 spa_vers = spa_version(spa); 3189 spa_close(spa, FTAG); 3190 3191 zplver = zfs_zpl_version_map(spa_vers); 3192 fuids_ok = (zplver >= ZPL_VERSION_FUID); 3193 sa_ok = (zplver >= ZPL_VERSION_SA); 3194 3195 /* 3196 * Open parent object set so we can inherit zplprop values. 3197 */ 3198 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0) 3199 return (error); 3200 3201 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops, 3202 zplprops, is_ci); 3203 dmu_objset_rele(os, FTAG); 3204 return (error); 3205} 3206 3207static int 3208zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops, 3209 nvlist_t *zplprops, boolean_t *is_ci) 3210{ 3211 boolean_t fuids_ok; 3212 boolean_t sa_ok; 3213 uint64_t zplver = ZPL_VERSION; 3214 int error; 3215 3216 zplver = zfs_zpl_version_map(spa_vers); 3217 fuids_ok = (zplver >= ZPL_VERSION_FUID); 3218 sa_ok = (zplver >= ZPL_VERSION_SA); 3219 3220 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok, 3221 createprops, zplprops, is_ci); 3222 return (error); 3223} 3224 3225/* 3226 * innvl: { 3227 * "type" -> dmu_objset_type_t (int32) 3228 * (optional) "props" -> { prop -> value } 3229 * } 3230 * 3231 * outnvl: propname -> error code (int32) 3232 */ 3233static int 3234zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3235{ 3236 int error = 0; 3237 zfs_creat_t zct = { 0 }; 3238 nvlist_t *nvprops = NULL; 3239 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx); 3240 int32_t type32; 3241 dmu_objset_type_t type; 3242 boolean_t is_insensitive = B_FALSE; 3243 3244 if (nvlist_lookup_int32(innvl, "type", &type32) != 0) 3245 return (SET_ERROR(EINVAL)); 3246 type = type32; 3247 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops); 3248 3249 switch (type) { 3250 case DMU_OST_ZFS: 3251 cbfunc = zfs_create_cb; 3252 break; 3253 3254 case DMU_OST_ZVOL: 3255 cbfunc = zvol_create_cb; 3256 break; 3257 3258 default: 3259 cbfunc = NULL; 3260 break; 3261 } 3262 if (strchr(fsname, '@') || 3263 strchr(fsname, '%')) 3264 return (SET_ERROR(EINVAL)); 3265 3266 zct.zct_props = nvprops; 3267 3268 if (cbfunc == NULL) 3269 return (SET_ERROR(EINVAL)); 3270 3271 if (type == DMU_OST_ZVOL) { 3272 uint64_t volsize, volblocksize; 3273 3274 if (nvprops == NULL) 3275 return (SET_ERROR(EINVAL)); 3276 if (nvlist_lookup_uint64(nvprops, 3277 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0) 3278 return (SET_ERROR(EINVAL)); 3279 3280 if ((error = nvlist_lookup_uint64(nvprops, 3281 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), 3282 &volblocksize)) != 0 && error != ENOENT) 3283 return (SET_ERROR(EINVAL)); 3284 3285 if (error != 0) 3286 volblocksize = zfs_prop_default_numeric( 3287 ZFS_PROP_VOLBLOCKSIZE); 3288 3289 if ((error = zvol_check_volblocksize( 3290 volblocksize)) != 0 || 3291 (error = zvol_check_volsize(volsize, 3292 volblocksize)) != 0) 3293 return (error); 3294 } else if (type == DMU_OST_ZFS) { 3295 int error; 3296 3297 /* 3298 * We have to have normalization and 3299 * case-folding flags correct when we do the 3300 * file system creation, so go figure them out 3301 * now. 3302 */ 3303 VERIFY(nvlist_alloc(&zct.zct_zplprops, 3304 NV_UNIQUE_NAME, KM_SLEEP) == 0); 3305 error = zfs_fill_zplprops(fsname, nvprops, 3306 zct.zct_zplprops, &is_insensitive); 3307 if (error != 0) { 3308 nvlist_free(zct.zct_zplprops); 3309 return (error); 3310 } 3311 } 3312 3313 error = dmu_objset_create(fsname, type, 3314 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct); 3315 nvlist_free(zct.zct_zplprops); 3316 3317 /* 3318 * It would be nice to do this atomically. 3319 */ 3320 if (error == 0) { 3321 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL, 3322 nvprops, outnvl); 3323 if (error != 0) 3324 (void) dsl_destroy_head(fsname); 3325 } 3326#ifdef __FreeBSD__ 3327 if (error == 0 && type == DMU_OST_ZVOL) 3328 zvol_create_minors(fsname); 3329#endif 3330 return (error); 3331} 3332 3333/* 3334 * innvl: { 3335 * "origin" -> name of origin snapshot 3336 * (optional) "props" -> { prop -> value } 3337 * } 3338 * 3339 * outnvl: propname -> error code (int32) 3340 */ 3341static int 3342zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3343{ 3344 int error = 0; 3345 nvlist_t *nvprops = NULL; 3346 char *origin_name; 3347 3348 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0) 3349 return (SET_ERROR(EINVAL)); 3350 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops); 3351 3352 if (strchr(fsname, '@') || 3353 strchr(fsname, '%')) 3354 return (SET_ERROR(EINVAL)); 3355 3356 if (dataset_namecheck(origin_name, NULL, NULL) != 0) 3357 return (SET_ERROR(EINVAL)); 3358 error = dmu_objset_clone(fsname, origin_name); 3359 if (error != 0) 3360 return (error); 3361 3362 /* 3363 * It would be nice to do this atomically. 3364 */ 3365 if (error == 0) { 3366 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL, 3367 nvprops, outnvl); 3368 if (error != 0) 3369 (void) dsl_destroy_head(fsname); 3370 } 3371#ifdef __FreeBSD__ 3372 if (error == 0) 3373 zvol_create_minors(fsname); 3374#endif 3375 return (error); 3376} 3377 3378/* 3379 * innvl: { 3380 * "snaps" -> { snapshot1, snapshot2 } 3381 * (optional) "props" -> { prop -> value (string) } 3382 * } 3383 * 3384 * outnvl: snapshot -> error code (int32) 3385 */ 3386static int 3387zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3388{ 3389 nvlist_t *snaps; 3390 nvlist_t *props = NULL; 3391 int error, poollen; 3392 nvpair_t *pair; 3393 3394 (void) nvlist_lookup_nvlist(innvl, "props", &props); 3395 if ((error = zfs_check_userprops(poolname, props)) != 0) 3396 return (error); 3397 3398 if (!nvlist_empty(props) && 3399 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS)) 3400 return (SET_ERROR(ENOTSUP)); 3401 3402 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 3403 return (SET_ERROR(EINVAL)); 3404 poollen = strlen(poolname); 3405 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 3406 pair = nvlist_next_nvpair(snaps, pair)) { 3407 const char *name = nvpair_name(pair); 3408 const char *cp = strchr(name, '@'); 3409 3410 /* 3411 * The snap name must contain an @, and the part after it must 3412 * contain only valid characters. 3413 */ 3414 if (cp == NULL || 3415 zfs_component_namecheck(cp + 1, NULL, NULL) != 0) 3416 return (SET_ERROR(EINVAL)); 3417 3418 /* 3419 * The snap must be in the specified pool. 3420 */ 3421 if (strncmp(name, poolname, poollen) != 0 || 3422 (name[poollen] != '/' && name[poollen] != '@')) 3423 return (SET_ERROR(EXDEV)); 3424 3425 /* This must be the only snap of this fs. */ 3426 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair); 3427 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) { 3428 if (strncmp(name, nvpair_name(pair2), cp - name + 1) 3429 == 0) { 3430 return (SET_ERROR(EXDEV)); 3431 } 3432 } 3433 } 3434 3435 error = dsl_dataset_snapshot(snaps, props, outnvl); 3436 return (error); 3437} 3438 3439/* 3440 * innvl: "message" -> string 3441 */ 3442/* ARGSUSED */ 3443static int 3444zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl) 3445{ 3446 char *message; 3447 spa_t *spa; 3448 int error; 3449 char *poolname; 3450 3451 /* 3452 * The poolname in the ioctl is not set, we get it from the TSD, 3453 * which was set at the end of the last successful ioctl that allows 3454 * logging. The secpolicy func already checked that it is set. 3455 * Only one log ioctl is allowed after each successful ioctl, so 3456 * we clear the TSD here. 3457 */ 3458 poolname = tsd_get(zfs_allow_log_key); 3459 (void) tsd_set(zfs_allow_log_key, NULL); 3460 error = spa_open(poolname, &spa, FTAG); 3461 strfree(poolname); 3462 if (error != 0) 3463 return (error); 3464 3465 if (nvlist_lookup_string(innvl, "message", &message) != 0) { 3466 spa_close(spa, FTAG); 3467 return (SET_ERROR(EINVAL)); 3468 } 3469 3470 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) { 3471 spa_close(spa, FTAG); 3472 return (SET_ERROR(ENOTSUP)); 3473 } 3474 3475 error = spa_history_log(spa, message); 3476 spa_close(spa, FTAG); 3477 return (error); 3478} 3479 3480/* 3481 * The dp_config_rwlock must not be held when calling this, because the 3482 * unmount may need to write out data. 3483 * 3484 * This function is best-effort. Callers must deal gracefully if it 3485 * remains mounted (or is remounted after this call). 3486 * 3487 * Returns 0 if the argument is not a snapshot, or it is not currently a 3488 * filesystem, or we were able to unmount it. Returns error code otherwise. 3489 */ 3490int 3491zfs_unmount_snap(const char *snapname) 3492{ 3493 vfs_t *vfsp; 3494 zfsvfs_t *zfsvfs; 3495 int err; 3496 3497 if (strchr(snapname, '@') == NULL) 3498 return (0); 3499 3500 vfsp = zfs_get_vfs(snapname); 3501 if (vfsp == NULL) 3502 return (0); 3503 3504 zfsvfs = vfsp->vfs_data; 3505 ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os))); 3506 3507 err = vn_vfswlock(vfsp->vfs_vnodecovered); 3508 VFS_RELE(vfsp); 3509 if (err != 0) 3510 return (SET_ERROR(err)); 3511 3512 /* 3513 * Always force the unmount for snapshots. 3514 */ 3515 3516#ifdef illumos 3517 (void) dounmount(vfsp, MS_FORCE, kcred); 3518#else 3519 mtx_lock(&Giant); /* dounmount() */ 3520 (void) dounmount(vfsp, MS_FORCE, curthread); 3521 mtx_unlock(&Giant); /* dounmount() */ 3522#endif 3523 return (0); 3524} 3525 3526/* ARGSUSED */ 3527static int 3528zfs_unmount_snap_cb(const char *snapname, void *arg) 3529{ 3530 return (zfs_unmount_snap(snapname)); 3531} 3532 3533/* 3534 * When a clone is destroyed, its origin may also need to be destroyed, 3535 * in which case it must be unmounted. This routine will do that unmount 3536 * if necessary. 3537 */ 3538void 3539zfs_destroy_unmount_origin(const char *fsname) 3540{ 3541 int error; 3542 objset_t *os; 3543 dsl_dataset_t *ds; 3544 3545 error = dmu_objset_hold(fsname, FTAG, &os); 3546 if (error != 0) 3547 return; 3548 ds = dmu_objset_ds(os); 3549 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) { 3550 char originname[MAXNAMELEN]; 3551 dsl_dataset_name(ds->ds_prev, originname); 3552 dmu_objset_rele(os, FTAG); 3553 (void) zfs_unmount_snap(originname); 3554 } else { 3555 dmu_objset_rele(os, FTAG); 3556 } 3557} 3558 3559/* 3560 * innvl: { 3561 * "snaps" -> { snapshot1, snapshot2 } 3562 * (optional boolean) "defer" 3563 * } 3564 * 3565 * outnvl: snapshot -> error code (int32) 3566 * 3567 */ 3568/* ARGSUSED */ 3569static int 3570zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3571{ 3572 nvlist_t *snaps; 3573 nvpair_t *pair; 3574 boolean_t defer; 3575 3576 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 3577 return (SET_ERROR(EINVAL)); 3578 defer = nvlist_exists(innvl, "defer"); 3579 3580 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 3581 pair = nvlist_next_nvpair(snaps, pair)) { 3582 (void) zfs_unmount_snap(nvpair_name(pair)); 3583 } 3584 3585 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl)); 3586} 3587 3588/* 3589 * Create bookmarks. Bookmark names are of the form <fs>#<bmark>. 3590 * All bookmarks must be in the same pool. 3591 * 3592 * innvl: { 3593 * bookmark1 -> snapshot1, bookmark2 -> snapshot2 3594 * } 3595 * 3596 * outnvl: bookmark -> error code (int32) 3597 * 3598 */ 3599/* ARGSUSED */ 3600static int 3601zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3602{ 3603 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 3604 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 3605 char *snap_name; 3606 3607 /* 3608 * Verify the snapshot argument. 3609 */ 3610 if (nvpair_value_string(pair, &snap_name) != 0) 3611 return (SET_ERROR(EINVAL)); 3612 3613 3614 /* Verify that the keys (bookmarks) are unique */ 3615 for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair); 3616 pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) { 3617 if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0) 3618 return (SET_ERROR(EINVAL)); 3619 } 3620 } 3621 3622 return (dsl_bookmark_create(innvl, outnvl)); 3623} 3624 3625/* 3626 * innvl: { 3627 * property 1, property 2, ... 3628 * } 3629 * 3630 * outnvl: { 3631 * bookmark name 1 -> { property 1, property 2, ... }, 3632 * bookmark name 2 -> { property 1, property 2, ... } 3633 * } 3634 * 3635 */ 3636static int 3637zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3638{ 3639 return (dsl_get_bookmarks(fsname, innvl, outnvl)); 3640} 3641 3642/* 3643 * innvl: { 3644 * bookmark name 1, bookmark name 2 3645 * } 3646 * 3647 * outnvl: bookmark -> error code (int32) 3648 * 3649 */ 3650static int 3651zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl, 3652 nvlist_t *outnvl) 3653{ 3654 int error, poollen; 3655 3656 poollen = strlen(poolname); 3657 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 3658 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 3659 const char *name = nvpair_name(pair); 3660 const char *cp = strchr(name, '#'); 3661 3662 /* 3663 * The bookmark name must contain an #, and the part after it 3664 * must contain only valid characters. 3665 */ 3666 if (cp == NULL || 3667 zfs_component_namecheck(cp + 1, NULL, NULL) != 0) 3668 return (SET_ERROR(EINVAL)); 3669 3670 /* 3671 * The bookmark must be in the specified pool. 3672 */ 3673 if (strncmp(name, poolname, poollen) != 0 || 3674 (name[poollen] != '/' && name[poollen] != '#')) 3675 return (SET_ERROR(EXDEV)); 3676 (void) zvol_remove_minor(name); 3677 } 3678 3679 error = dsl_bookmark_destroy(innvl, outnvl); 3680 return (error); 3681} 3682 3683/* 3684 * inputs: 3685 * zc_name name of dataset to destroy 3686 * zc_objset_type type of objset 3687 * zc_defer_destroy mark for deferred destroy 3688 * 3689 * outputs: none 3690 */ 3691static int 3692zfs_ioc_destroy(zfs_cmd_t *zc) 3693{ 3694 int err; 3695 3696 if (zc->zc_objset_type == DMU_OST_ZFS) { 3697 err = zfs_unmount_snap(zc->zc_name); 3698 if (err != 0) 3699 return (err); 3700 } 3701 3702 if (strchr(zc->zc_name, '@')) 3703 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy); 3704 else 3705 err = dsl_destroy_head(zc->zc_name); 3706 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0) 3707 (void) zvol_remove_minor(zc->zc_name); 3708 return (err); 3709} 3710 3711/* 3712 * fsname is name of dataset to rollback (to most recent snapshot) 3713 * 3714 * innvl is not used. 3715 * 3716 * outnvl: "target" -> name of most recent snapshot 3717 * } 3718 */ 3719/* ARGSUSED */ 3720static int 3721zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl) 3722{ 3723 zfsvfs_t *zfsvfs; 3724 int error; 3725 3726 if (getzfsvfs(fsname, &zfsvfs) == 0) { 3727 error = zfs_suspend_fs(zfsvfs); 3728 if (error == 0) { 3729 int resume_err; 3730 3731 error = dsl_dataset_rollback(fsname, zfsvfs, outnvl); 3732 resume_err = zfs_resume_fs(zfsvfs, fsname); 3733 error = error ? error : resume_err; 3734 } 3735 VFS_RELE(zfsvfs->z_vfs); 3736 } else { 3737 error = dsl_dataset_rollback(fsname, NULL, outnvl); 3738 } 3739 return (error); 3740} 3741 3742static int 3743recursive_unmount(const char *fsname, void *arg) 3744{ 3745 const char *snapname = arg; 3746 char fullname[MAXNAMELEN]; 3747 3748 (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname); 3749 return (zfs_unmount_snap(fullname)); 3750} 3751 3752/* 3753 * inputs: 3754 * zc_name old name of dataset 3755 * zc_value new name of dataset 3756 * zc_cookie recursive flag (only valid for snapshots) 3757 * 3758 * outputs: none 3759 */ 3760static int 3761zfs_ioc_rename(zfs_cmd_t *zc) 3762{ 3763 boolean_t recursive = zc->zc_cookie & 1; 3764#ifdef __FreeBSD__ 3765 boolean_t allow_mounted = zc->zc_cookie & 2; 3766#endif 3767 char *at; 3768 3769 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0'; 3770 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 3771 strchr(zc->zc_value, '%')) 3772 return (SET_ERROR(EINVAL)); 3773 3774 at = strchr(zc->zc_name, '@'); 3775 if (at != NULL) { 3776 /* snaps must be in same fs */ 3777 int error; 3778 3779 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1)) 3780 return (SET_ERROR(EXDEV)); 3781 *at = '\0'; 3782#ifdef illumos 3783 if (zc->zc_objset_type == DMU_OST_ZFS) { 3784#else 3785 if (zc->zc_objset_type == DMU_OST_ZFS && allow_mounted) { 3786#endif 3787 error = dmu_objset_find(zc->zc_name, 3788 recursive_unmount, at + 1, 3789 recursive ? DS_FIND_CHILDREN : 0); 3790 if (error != 0) { 3791 *at = '@'; 3792 return (error); 3793 } 3794 } 3795 error = dsl_dataset_rename_snapshot(zc->zc_name, 3796 at + 1, strchr(zc->zc_value, '@') + 1, recursive); 3797 *at = '@'; 3798 3799 return (error); 3800 } else { 3801#ifdef illumos 3802 if (zc->zc_objset_type == DMU_OST_ZVOL) 3803 (void) zvol_remove_minor(zc->zc_name); 3804#endif 3805 return (dsl_dir_rename(zc->zc_name, zc->zc_value)); 3806 } 3807} 3808 3809static int 3810zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr) 3811{ 3812 const char *propname = nvpair_name(pair); 3813 boolean_t issnap = (strchr(dsname, '@') != NULL); 3814 zfs_prop_t prop = zfs_name_to_prop(propname); 3815 uint64_t intval; 3816 int err; 3817 3818 if (prop == ZPROP_INVAL) { 3819 if (zfs_prop_user(propname)) { 3820 if (err = zfs_secpolicy_write_perms(dsname, 3821 ZFS_DELEG_PERM_USERPROP, cr)) 3822 return (err); 3823 return (0); 3824 } 3825 3826 if (!issnap && zfs_prop_userquota(propname)) { 3827 const char *perm = NULL; 3828 const char *uq_prefix = 3829 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA]; 3830 const char *gq_prefix = 3831 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA]; 3832 3833 if (strncmp(propname, uq_prefix, 3834 strlen(uq_prefix)) == 0) { 3835 perm = ZFS_DELEG_PERM_USERQUOTA; 3836 } else if (strncmp(propname, gq_prefix, 3837 strlen(gq_prefix)) == 0) { 3838 perm = ZFS_DELEG_PERM_GROUPQUOTA; 3839 } else { 3840 /* USERUSED and GROUPUSED are read-only */ 3841 return (SET_ERROR(EINVAL)); 3842 } 3843 3844 if (err = zfs_secpolicy_write_perms(dsname, perm, cr)) 3845 return (err); 3846 return (0); 3847 } 3848 3849 return (SET_ERROR(EINVAL)); 3850 } 3851 3852 if (issnap) 3853 return (SET_ERROR(EINVAL)); 3854 3855 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 3856 /* 3857 * dsl_prop_get_all_impl() returns properties in this 3858 * format. 3859 */ 3860 nvlist_t *attrs; 3861 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 3862 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 3863 &pair) == 0); 3864 } 3865 3866 /* 3867 * Check that this value is valid for this pool version 3868 */ 3869 switch (prop) { 3870 case ZFS_PROP_COMPRESSION: 3871 /* 3872 * If the user specified gzip compression, make sure 3873 * the SPA supports it. We ignore any errors here since 3874 * we'll catch them later. 3875 */ 3876 if (nvpair_type(pair) == DATA_TYPE_UINT64 && 3877 nvpair_value_uint64(pair, &intval) == 0) { 3878 if (intval >= ZIO_COMPRESS_GZIP_1 && 3879 intval <= ZIO_COMPRESS_GZIP_9 && 3880 zfs_earlier_version(dsname, 3881 SPA_VERSION_GZIP_COMPRESSION)) { 3882 return (SET_ERROR(ENOTSUP)); 3883 } 3884 3885 if (intval == ZIO_COMPRESS_ZLE && 3886 zfs_earlier_version(dsname, 3887 SPA_VERSION_ZLE_COMPRESSION)) 3888 return (SET_ERROR(ENOTSUP)); 3889 3890 if (intval == ZIO_COMPRESS_LZ4) { 3891 spa_t *spa; 3892 3893 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 3894 return (err); 3895 3896 if (!spa_feature_is_enabled(spa, 3897 SPA_FEATURE_LZ4_COMPRESS)) { 3898 spa_close(spa, FTAG); 3899 return (SET_ERROR(ENOTSUP)); 3900 } 3901 spa_close(spa, FTAG); 3902 } 3903 3904 /* 3905 * If this is a bootable dataset then 3906 * verify that the compression algorithm 3907 * is supported for booting. We must return 3908 * something other than ENOTSUP since it 3909 * implies a downrev pool version. 3910 */ 3911 if (zfs_is_bootfs(dsname) && 3912 !BOOTFS_COMPRESS_VALID(intval)) { 3913 return (SET_ERROR(ERANGE)); 3914 } 3915 } 3916 break; 3917 3918 case ZFS_PROP_COPIES: 3919 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS)) 3920 return (SET_ERROR(ENOTSUP)); 3921 break; 3922 3923 case ZFS_PROP_DEDUP: 3924 if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP)) 3925 return (SET_ERROR(ENOTSUP)); 3926 break; 3927 3928 case ZFS_PROP_SHARESMB: 3929 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID)) 3930 return (SET_ERROR(ENOTSUP)); 3931 break; 3932 3933 case ZFS_PROP_ACLINHERIT: 3934 if (nvpair_type(pair) == DATA_TYPE_UINT64 && 3935 nvpair_value_uint64(pair, &intval) == 0) { 3936 if (intval == ZFS_ACL_PASSTHROUGH_X && 3937 zfs_earlier_version(dsname, 3938 SPA_VERSION_PASSTHROUGH_X)) 3939 return (SET_ERROR(ENOTSUP)); 3940 } 3941 break; 3942 } 3943 3944 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED())); 3945} 3946 3947/* 3948 * Checks for a race condition to make sure we don't increment a feature flag 3949 * multiple times. 3950 */ 3951static int 3952zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx) 3953{ 3954 spa_t *spa = dmu_tx_pool(tx)->dp_spa; 3955 spa_feature_t *featurep = arg; 3956 3957 if (!spa_feature_is_active(spa, *featurep)) 3958 return (0); 3959 else 3960 return (SET_ERROR(EBUSY)); 3961} 3962 3963/* 3964 * The callback invoked on feature activation in the sync task caused by 3965 * zfs_prop_activate_feature. 3966 */ 3967static void 3968zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx) 3969{ 3970 spa_t *spa = dmu_tx_pool(tx)->dp_spa; 3971 spa_feature_t *featurep = arg; 3972 3973 spa_feature_incr(spa, *featurep, tx); 3974} 3975 3976/* 3977 * Activates a feature on a pool in response to a property setting. This 3978 * creates a new sync task which modifies the pool to reflect the feature 3979 * as being active. 3980 */ 3981static int 3982zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature) 3983{ 3984 int err; 3985 3986 /* EBUSY here indicates that the feature is already active */ 3987 err = dsl_sync_task(spa_name(spa), 3988 zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync, 3989 &feature, 2); 3990 3991 if (err != 0 && err != EBUSY) 3992 return (err); 3993 else 3994 return (0); 3995} 3996 3997/* 3998 * Removes properties from the given props list that fail permission checks 3999 * needed to clear them and to restore them in case of a receive error. For each 4000 * property, make sure we have both set and inherit permissions. 4001 * 4002 * Returns the first error encountered if any permission checks fail. If the 4003 * caller provides a non-NULL errlist, it also gives the complete list of names 4004 * of all the properties that failed a permission check along with the 4005 * corresponding error numbers. The caller is responsible for freeing the 4006 * returned errlist. 4007 * 4008 * If every property checks out successfully, zero is returned and the list 4009 * pointed at by errlist is NULL. 4010 */ 4011static int 4012zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist) 4013{ 4014 zfs_cmd_t *zc; 4015 nvpair_t *pair, *next_pair; 4016 nvlist_t *errors; 4017 int err, rv = 0; 4018 4019 if (props == NULL) 4020 return (0); 4021 4022 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0); 4023 4024 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP); 4025 (void) strcpy(zc->zc_name, dataset); 4026 pair = nvlist_next_nvpair(props, NULL); 4027 while (pair != NULL) { 4028 next_pair = nvlist_next_nvpair(props, pair); 4029 4030 (void) strcpy(zc->zc_value, nvpair_name(pair)); 4031 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 || 4032 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) { 4033 VERIFY(nvlist_remove_nvpair(props, pair) == 0); 4034 VERIFY(nvlist_add_int32(errors, 4035 zc->zc_value, err) == 0); 4036 } 4037 pair = next_pair; 4038 } 4039 kmem_free(zc, sizeof (zfs_cmd_t)); 4040 4041 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) { 4042 nvlist_free(errors); 4043 errors = NULL; 4044 } else { 4045 VERIFY(nvpair_value_int32(pair, &rv) == 0); 4046 } 4047 4048 if (errlist == NULL) 4049 nvlist_free(errors); 4050 else 4051 *errlist = errors; 4052 4053 return (rv); 4054} 4055 4056static boolean_t 4057propval_equals(nvpair_t *p1, nvpair_t *p2) 4058{ 4059 if (nvpair_type(p1) == DATA_TYPE_NVLIST) { 4060 /* dsl_prop_get_all_impl() format */ 4061 nvlist_t *attrs; 4062 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0); 4063 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 4064 &p1) == 0); 4065 } 4066 4067 if (nvpair_type(p2) == DATA_TYPE_NVLIST) { 4068 nvlist_t *attrs; 4069 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0); 4070 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 4071 &p2) == 0); 4072 } 4073 4074 if (nvpair_type(p1) != nvpair_type(p2)) 4075 return (B_FALSE); 4076 4077 if (nvpair_type(p1) == DATA_TYPE_STRING) { 4078 char *valstr1, *valstr2; 4079 4080 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0); 4081 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0); 4082 return (strcmp(valstr1, valstr2) == 0); 4083 } else { 4084 uint64_t intval1, intval2; 4085 4086 VERIFY(nvpair_value_uint64(p1, &intval1) == 0); 4087 VERIFY(nvpair_value_uint64(p2, &intval2) == 0); 4088 return (intval1 == intval2); 4089 } 4090} 4091 4092/* 4093 * Remove properties from props if they are not going to change (as determined 4094 * by comparison with origprops). Remove them from origprops as well, since we 4095 * do not need to clear or restore properties that won't change. 4096 */ 4097static void 4098props_reduce(nvlist_t *props, nvlist_t *origprops) 4099{ 4100 nvpair_t *pair, *next_pair; 4101 4102 if (origprops == NULL) 4103 return; /* all props need to be received */ 4104 4105 pair = nvlist_next_nvpair(props, NULL); 4106 while (pair != NULL) { 4107 const char *propname = nvpair_name(pair); 4108 nvpair_t *match; 4109 4110 next_pair = nvlist_next_nvpair(props, pair); 4111 4112 if ((nvlist_lookup_nvpair(origprops, propname, 4113 &match) != 0) || !propval_equals(pair, match)) 4114 goto next; /* need to set received value */ 4115 4116 /* don't clear the existing received value */ 4117 (void) nvlist_remove_nvpair(origprops, match); 4118 /* don't bother receiving the property */ 4119 (void) nvlist_remove_nvpair(props, pair); 4120next: 4121 pair = next_pair; 4122 } 4123} 4124 4125#ifdef DEBUG 4126static boolean_t zfs_ioc_recv_inject_err; 4127#endif 4128 4129/* 4130 * inputs: 4131 * zc_name name of containing filesystem 4132 * zc_nvlist_src{_size} nvlist of properties to apply 4133 * zc_value name of snapshot to create 4134 * zc_string name of clone origin (if DRR_FLAG_CLONE) 4135 * zc_cookie file descriptor to recv from 4136 * zc_begin_record the BEGIN record of the stream (not byteswapped) 4137 * zc_guid force flag 4138 * zc_cleanup_fd cleanup-on-exit file descriptor 4139 * zc_action_handle handle for this guid/ds mapping (or zero on first call) 4140 * 4141 * outputs: 4142 * zc_cookie number of bytes read 4143 * zc_nvlist_dst{_size} error for each unapplied received property 4144 * zc_obj zprop_errflags_t 4145 * zc_action_handle handle for this guid/ds mapping 4146 */ 4147static int 4148zfs_ioc_recv(zfs_cmd_t *zc) 4149{ 4150 file_t *fp; 4151 dmu_recv_cookie_t drc; 4152 boolean_t force = (boolean_t)zc->zc_guid; 4153 int fd; 4154 int error = 0; 4155 int props_error = 0; 4156 nvlist_t *errors; 4157 offset_t off; 4158 nvlist_t *props = NULL; /* sent properties */ 4159 nvlist_t *origprops = NULL; /* existing properties */ 4160 char *origin = NULL; 4161 char *tosnap; 4162 char tofs[ZFS_MAXNAMELEN]; 4163 cap_rights_t rights; 4164 boolean_t first_recvd_props = B_FALSE; 4165 4166 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 4167 strchr(zc->zc_value, '@') == NULL || 4168 strchr(zc->zc_value, '%')) 4169 return (SET_ERROR(EINVAL)); 4170 4171 (void) strcpy(tofs, zc->zc_value); 4172 tosnap = strchr(tofs, '@'); 4173 *tosnap++ = '\0'; 4174 4175 if (zc->zc_nvlist_src != 0 && 4176 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 4177 zc->zc_iflags, &props)) != 0) 4178 return (error); 4179 4180 fd = zc->zc_cookie; 4181 fp = getf(fd, cap_rights_init(&rights, CAP_PREAD)); 4182 if (fp == NULL) { 4183 nvlist_free(props); 4184 return (SET_ERROR(EBADF)); 4185 } 4186 4187 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0); 4188 4189 if (zc->zc_string[0]) 4190 origin = zc->zc_string; 4191 4192 error = dmu_recv_begin(tofs, tosnap, 4193 &zc->zc_begin_record, force, origin, &drc); 4194 if (error != 0) 4195 goto out; 4196 4197 /* 4198 * Set properties before we receive the stream so that they are applied 4199 * to the new data. Note that we must call dmu_recv_stream() if 4200 * dmu_recv_begin() succeeds. 4201 */ 4202 if (props != NULL && !drc.drc_newfs) { 4203 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >= 4204 SPA_VERSION_RECVD_PROPS && 4205 !dsl_prop_get_hasrecvd(tofs)) 4206 first_recvd_props = B_TRUE; 4207 4208 /* 4209 * If new received properties are supplied, they are to 4210 * completely replace the existing received properties, so stash 4211 * away the existing ones. 4212 */ 4213 if (dsl_prop_get_received(tofs, &origprops) == 0) { 4214 nvlist_t *errlist = NULL; 4215 /* 4216 * Don't bother writing a property if its value won't 4217 * change (and avoid the unnecessary security checks). 4218 * 4219 * The first receive after SPA_VERSION_RECVD_PROPS is a 4220 * special case where we blow away all local properties 4221 * regardless. 4222 */ 4223 if (!first_recvd_props) 4224 props_reduce(props, origprops); 4225 if (zfs_check_clearable(tofs, origprops, &errlist) != 0) 4226 (void) nvlist_merge(errors, errlist, 0); 4227 nvlist_free(errlist); 4228 4229 if (clear_received_props(tofs, origprops, 4230 first_recvd_props ? NULL : props) != 0) 4231 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 4232 } else { 4233 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 4234 } 4235 } 4236 4237 if (props != NULL) { 4238 props_error = dsl_prop_set_hasrecvd(tofs); 4239 4240 if (props_error == 0) { 4241 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED, 4242 props, errors); 4243 } 4244 } 4245 4246 if (zc->zc_nvlist_dst_size != 0 && 4247 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 || 4248 put_nvlist(zc, errors) != 0)) { 4249 /* 4250 * Caller made zc->zc_nvlist_dst less than the minimum expected 4251 * size or supplied an invalid address. 4252 */ 4253 props_error = SET_ERROR(EINVAL); 4254 } 4255 4256 off = fp->f_offset; 4257 error = dmu_recv_stream(&drc, fp, &off, zc->zc_cleanup_fd, 4258 &zc->zc_action_handle); 4259 4260 if (error == 0) { 4261 zfsvfs_t *zfsvfs = NULL; 4262 4263 if (getzfsvfs(tofs, &zfsvfs) == 0) { 4264 /* online recv */ 4265 int end_err; 4266 4267 error = zfs_suspend_fs(zfsvfs); 4268 /* 4269 * If the suspend fails, then the recv_end will 4270 * likely also fail, and clean up after itself. 4271 */ 4272 end_err = dmu_recv_end(&drc, zfsvfs); 4273 if (error == 0) 4274 error = zfs_resume_fs(zfsvfs, tofs); 4275 error = error ? error : end_err; 4276 VFS_RELE(zfsvfs->z_vfs); 4277 } else { 4278 error = dmu_recv_end(&drc, NULL); 4279 } 4280 } 4281 4282 zc->zc_cookie = off - fp->f_offset; 4283 if (off >= 0 && off <= MAXOFFSET_T) 4284 fp->f_offset = off; 4285 4286#ifdef DEBUG 4287 if (zfs_ioc_recv_inject_err) { 4288 zfs_ioc_recv_inject_err = B_FALSE; 4289 error = 1; 4290 } 4291#endif 4292 4293#ifdef __FreeBSD__ 4294 if (error == 0) 4295 zvol_create_minors(tofs); 4296#endif 4297 4298 /* 4299 * On error, restore the original props. 4300 */ 4301 if (error != 0 && props != NULL && !drc.drc_newfs) { 4302 if (clear_received_props(tofs, props, NULL) != 0) { 4303 /* 4304 * We failed to clear the received properties. 4305 * Since we may have left a $recvd value on the 4306 * system, we can't clear the $hasrecvd flag. 4307 */ 4308 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4309 } else if (first_recvd_props) { 4310 dsl_prop_unset_hasrecvd(tofs); 4311 } 4312 4313 if (origprops == NULL && !drc.drc_newfs) { 4314 /* We failed to stash the original properties. */ 4315 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4316 } 4317 4318 /* 4319 * dsl_props_set() will not convert RECEIVED to LOCAL on or 4320 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL 4321 * explictly if we're restoring local properties cleared in the 4322 * first new-style receive. 4323 */ 4324 if (origprops != NULL && 4325 zfs_set_prop_nvlist(tofs, (first_recvd_props ? 4326 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED), 4327 origprops, NULL) != 0) { 4328 /* 4329 * We stashed the original properties but failed to 4330 * restore them. 4331 */ 4332 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4333 } 4334 } 4335out: 4336 nvlist_free(props); 4337 nvlist_free(origprops); 4338 nvlist_free(errors); 4339 releasef(fd); 4340 4341 if (error == 0) 4342 error = props_error; 4343 4344 return (error); 4345} 4346 4347/* 4348 * inputs: 4349 * zc_name name of snapshot to send 4350 * zc_cookie file descriptor to send stream to 4351 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj) 4352 * zc_sendobj objsetid of snapshot to send 4353 * zc_fromobj objsetid of incremental fromsnap (may be zero) 4354 * zc_guid if set, estimate size of stream only. zc_cookie is ignored. 4355 * output size in zc_objset_type. 4356 * 4357 * outputs: 4358 * zc_objset_type estimated size, if zc_guid is set 4359 */ 4360static int 4361zfs_ioc_send(zfs_cmd_t *zc) 4362{ 4363 int error; 4364 offset_t off; 4365 boolean_t estimate = (zc->zc_guid != 0); 4366 4367 if (zc->zc_obj != 0) { 4368 dsl_pool_t *dp; 4369 dsl_dataset_t *tosnap; 4370 4371 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4372 if (error != 0) 4373 return (error); 4374 4375 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap); 4376 if (error != 0) { 4377 dsl_pool_rele(dp, FTAG); 4378 return (error); 4379 } 4380 4381 if (dsl_dir_is_clone(tosnap->ds_dir)) 4382 zc->zc_fromobj = tosnap->ds_dir->dd_phys->dd_origin_obj; 4383 dsl_dataset_rele(tosnap, FTAG); 4384 dsl_pool_rele(dp, FTAG); 4385 } 4386 4387 if (estimate) { 4388 dsl_pool_t *dp; 4389 dsl_dataset_t *tosnap; 4390 dsl_dataset_t *fromsnap = NULL; 4391 4392 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4393 if (error != 0) 4394 return (error); 4395 4396 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap); 4397 if (error != 0) { 4398 dsl_pool_rele(dp, FTAG); 4399 return (error); 4400 } 4401 4402 if (zc->zc_fromobj != 0) { 4403 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj, 4404 FTAG, &fromsnap); 4405 if (error != 0) { 4406 dsl_dataset_rele(tosnap, FTAG); 4407 dsl_pool_rele(dp, FTAG); 4408 return (error); 4409 } 4410 } 4411 4412 error = dmu_send_estimate(tosnap, fromsnap, 4413 &zc->zc_objset_type); 4414 4415 if (fromsnap != NULL) 4416 dsl_dataset_rele(fromsnap, FTAG); 4417 dsl_dataset_rele(tosnap, FTAG); 4418 dsl_pool_rele(dp, FTAG); 4419 } else { 4420 file_t *fp; 4421 cap_rights_t rights; 4422 4423 fp = getf(zc->zc_cookie, 4424 cap_rights_init(&rights, CAP_WRITE)); 4425 if (fp == NULL) 4426 return (SET_ERROR(EBADF)); 4427 4428 off = fp->f_offset; 4429 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj, 4430#ifdef illumos 4431 zc->zc_fromobj, zc->zc_cookie, fp->f_vnode, &off); 4432#else 4433 zc->zc_fromobj, zc->zc_cookie, fp, &off); 4434#endif 4435 4436 if (off >= 0 && off <= MAXOFFSET_T) 4437 fp->f_offset = off; 4438 releasef(zc->zc_cookie); 4439 } 4440 return (error); 4441} 4442 4443/* 4444 * inputs: 4445 * zc_name name of snapshot on which to report progress 4446 * zc_cookie file descriptor of send stream 4447 * 4448 * outputs: 4449 * zc_cookie number of bytes written in send stream thus far 4450 */ 4451static int 4452zfs_ioc_send_progress(zfs_cmd_t *zc) 4453{ 4454 dsl_pool_t *dp; 4455 dsl_dataset_t *ds; 4456 dmu_sendarg_t *dsp = NULL; 4457 int error; 4458 4459 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4460 if (error != 0) 4461 return (error); 4462 4463 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds); 4464 if (error != 0) { 4465 dsl_pool_rele(dp, FTAG); 4466 return (error); 4467 } 4468 4469 mutex_enter(&ds->ds_sendstream_lock); 4470 4471 /* 4472 * Iterate over all the send streams currently active on this dataset. 4473 * If there's one which matches the specified file descriptor _and_ the 4474 * stream was started by the current process, return the progress of 4475 * that stream. 4476 */ 4477 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL; 4478 dsp = list_next(&ds->ds_sendstreams, dsp)) { 4479 if (dsp->dsa_outfd == zc->zc_cookie && 4480 dsp->dsa_proc == curproc) 4481 break; 4482 } 4483 4484 if (dsp != NULL) 4485 zc->zc_cookie = *(dsp->dsa_off); 4486 else 4487 error = SET_ERROR(ENOENT); 4488 4489 mutex_exit(&ds->ds_sendstream_lock); 4490 dsl_dataset_rele(ds, FTAG); 4491 dsl_pool_rele(dp, FTAG); 4492 return (error); 4493} 4494 4495static int 4496zfs_ioc_inject_fault(zfs_cmd_t *zc) 4497{ 4498 int id, error; 4499 4500 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id, 4501 &zc->zc_inject_record); 4502 4503 if (error == 0) 4504 zc->zc_guid = (uint64_t)id; 4505 4506 return (error); 4507} 4508 4509static int 4510zfs_ioc_clear_fault(zfs_cmd_t *zc) 4511{ 4512 return (zio_clear_fault((int)zc->zc_guid)); 4513} 4514 4515static int 4516zfs_ioc_inject_list_next(zfs_cmd_t *zc) 4517{ 4518 int id = (int)zc->zc_guid; 4519 int error; 4520 4521 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name), 4522 &zc->zc_inject_record); 4523 4524 zc->zc_guid = id; 4525 4526 return (error); 4527} 4528 4529static int 4530zfs_ioc_error_log(zfs_cmd_t *zc) 4531{ 4532 spa_t *spa; 4533 int error; 4534 size_t count = (size_t)zc->zc_nvlist_dst_size; 4535 4536 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 4537 return (error); 4538 4539 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst, 4540 &count); 4541 if (error == 0) 4542 zc->zc_nvlist_dst_size = count; 4543 else 4544 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa); 4545 4546 spa_close(spa, FTAG); 4547 4548 return (error); 4549} 4550 4551static int 4552zfs_ioc_clear(zfs_cmd_t *zc) 4553{ 4554 spa_t *spa; 4555 vdev_t *vd; 4556 int error; 4557 4558 /* 4559 * On zpool clear we also fix up missing slogs 4560 */ 4561 mutex_enter(&spa_namespace_lock); 4562 spa = spa_lookup(zc->zc_name); 4563 if (spa == NULL) { 4564 mutex_exit(&spa_namespace_lock); 4565 return (SET_ERROR(EIO)); 4566 } 4567 if (spa_get_log_state(spa) == SPA_LOG_MISSING) { 4568 /* we need to let spa_open/spa_load clear the chains */ 4569 spa_set_log_state(spa, SPA_LOG_CLEAR); 4570 } 4571 spa->spa_last_open_failed = 0; 4572 mutex_exit(&spa_namespace_lock); 4573 4574 if (zc->zc_cookie & ZPOOL_NO_REWIND) { 4575 error = spa_open(zc->zc_name, &spa, FTAG); 4576 } else { 4577 nvlist_t *policy; 4578 nvlist_t *config = NULL; 4579 4580 if (zc->zc_nvlist_src == 0) 4581 return (SET_ERROR(EINVAL)); 4582 4583 if ((error = get_nvlist(zc->zc_nvlist_src, 4584 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) { 4585 error = spa_open_rewind(zc->zc_name, &spa, FTAG, 4586 policy, &config); 4587 if (config != NULL) { 4588 int err; 4589 4590 if ((err = put_nvlist(zc, config)) != 0) 4591 error = err; 4592 nvlist_free(config); 4593 } 4594 nvlist_free(policy); 4595 } 4596 } 4597 4598 if (error != 0) 4599 return (error); 4600 4601 spa_vdev_state_enter(spa, SCL_NONE); 4602 4603 if (zc->zc_guid == 0) { 4604 vd = NULL; 4605 } else { 4606 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE); 4607 if (vd == NULL) { 4608 (void) spa_vdev_state_exit(spa, NULL, ENODEV); 4609 spa_close(spa, FTAG); 4610 return (SET_ERROR(ENODEV)); 4611 } 4612 } 4613 4614 vdev_clear(spa, vd); 4615 4616 (void) spa_vdev_state_exit(spa, NULL, 0); 4617 4618 /* 4619 * Resume any suspended I/Os. 4620 */ 4621 if (zio_resume(spa) != 0) 4622 error = SET_ERROR(EIO); 4623 4624 spa_close(spa, FTAG); 4625 4626 return (error); 4627} 4628 4629static int 4630zfs_ioc_pool_reopen(zfs_cmd_t *zc) 4631{ 4632 spa_t *spa; 4633 int error; 4634 4635 error = spa_open(zc->zc_name, &spa, FTAG); 4636 if (error != 0) 4637 return (error); 4638 4639 spa_vdev_state_enter(spa, SCL_NONE); 4640 4641 /* 4642 * If a resilver is already in progress then set the 4643 * spa_scrub_reopen flag to B_TRUE so that we don't restart 4644 * the scan as a side effect of the reopen. Otherwise, let 4645 * vdev_open() decided if a resilver is required. 4646 */ 4647 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool); 4648 vdev_reopen(spa->spa_root_vdev); 4649 spa->spa_scrub_reopen = B_FALSE; 4650 4651 (void) spa_vdev_state_exit(spa, NULL, 0); 4652 spa_close(spa, FTAG); 4653 return (0); 4654} 4655/* 4656 * inputs: 4657 * zc_name name of filesystem 4658 * zc_value name of origin snapshot 4659 * 4660 * outputs: 4661 * zc_string name of conflicting snapshot, if there is one 4662 */ 4663static int 4664zfs_ioc_promote(zfs_cmd_t *zc) 4665{ 4666 char *cp; 4667 4668 /* 4669 * We don't need to unmount *all* the origin fs's snapshots, but 4670 * it's easier. 4671 */ 4672 cp = strchr(zc->zc_value, '@'); 4673 if (cp) 4674 *cp = '\0'; 4675 (void) dmu_objset_find(zc->zc_value, 4676 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS); 4677 return (dsl_dataset_promote(zc->zc_name, zc->zc_string)); 4678} 4679 4680/* 4681 * Retrieve a single {user|group}{used|quota}@... property. 4682 * 4683 * inputs: 4684 * zc_name name of filesystem 4685 * zc_objset_type zfs_userquota_prop_t 4686 * zc_value domain name (eg. "S-1-234-567-89") 4687 * zc_guid RID/UID/GID 4688 * 4689 * outputs: 4690 * zc_cookie property value 4691 */ 4692static int 4693zfs_ioc_userspace_one(zfs_cmd_t *zc) 4694{ 4695 zfsvfs_t *zfsvfs; 4696 int error; 4697 4698 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 4699 return (SET_ERROR(EINVAL)); 4700 4701 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE); 4702 if (error != 0) 4703 return (error); 4704 4705 error = zfs_userspace_one(zfsvfs, 4706 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie); 4707 zfsvfs_rele(zfsvfs, FTAG); 4708 4709 return (error); 4710} 4711 4712/* 4713 * inputs: 4714 * zc_name name of filesystem 4715 * zc_cookie zap cursor 4716 * zc_objset_type zfs_userquota_prop_t 4717 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist) 4718 * 4719 * outputs: 4720 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t) 4721 * zc_cookie zap cursor 4722 */ 4723static int 4724zfs_ioc_userspace_many(zfs_cmd_t *zc) 4725{ 4726 zfsvfs_t *zfsvfs; 4727 int bufsize = zc->zc_nvlist_dst_size; 4728 4729 if (bufsize <= 0) 4730 return (SET_ERROR(ENOMEM)); 4731 4732 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE); 4733 if (error != 0) 4734 return (error); 4735 4736 void *buf = kmem_alloc(bufsize, KM_SLEEP); 4737 4738 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie, 4739 buf, &zc->zc_nvlist_dst_size); 4740 4741 if (error == 0) { 4742 error = ddi_copyout(buf, 4743 (void *)(uintptr_t)zc->zc_nvlist_dst, 4744 zc->zc_nvlist_dst_size, zc->zc_iflags); 4745 } 4746 kmem_free(buf, bufsize); 4747 zfsvfs_rele(zfsvfs, FTAG); 4748 4749 return (error); 4750} 4751 4752/* 4753 * inputs: 4754 * zc_name name of filesystem 4755 * 4756 * outputs: 4757 * none 4758 */ 4759static int 4760zfs_ioc_userspace_upgrade(zfs_cmd_t *zc) 4761{ 4762 objset_t *os; 4763 int error = 0; 4764 zfsvfs_t *zfsvfs; 4765 4766 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) { 4767 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) { 4768 /* 4769 * If userused is not enabled, it may be because the 4770 * objset needs to be closed & reopened (to grow the 4771 * objset_phys_t). Suspend/resume the fs will do that. 4772 */ 4773 error = zfs_suspend_fs(zfsvfs); 4774 if (error == 0) { 4775 dmu_objset_refresh_ownership(zfsvfs->z_os, 4776 zfsvfs); 4777 error = zfs_resume_fs(zfsvfs, zc->zc_name); 4778 } 4779 } 4780 if (error == 0) 4781 error = dmu_objset_userspace_upgrade(zfsvfs->z_os); 4782 VFS_RELE(zfsvfs->z_vfs); 4783 } else { 4784 /* XXX kind of reading contents without owning */ 4785 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 4786 if (error != 0) 4787 return (error); 4788 4789 error = dmu_objset_userspace_upgrade(os); 4790 dmu_objset_rele(os, FTAG); 4791 } 4792 4793 return (error); 4794} 4795 4796#ifdef sun 4797/* 4798 * We don't want to have a hard dependency 4799 * against some special symbols in sharefs 4800 * nfs, and smbsrv. Determine them if needed when 4801 * the first file system is shared. 4802 * Neither sharefs, nfs or smbsrv are unloadable modules. 4803 */ 4804int (*znfsexport_fs)(void *arg); 4805int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t); 4806int (*zsmbexport_fs)(void *arg, boolean_t add_share); 4807 4808int zfs_nfsshare_inited; 4809int zfs_smbshare_inited; 4810 4811ddi_modhandle_t nfs_mod; 4812ddi_modhandle_t sharefs_mod; 4813ddi_modhandle_t smbsrv_mod; 4814#endif /* sun */ 4815kmutex_t zfs_share_lock; 4816 4817#ifdef sun 4818static int 4819zfs_init_sharefs() 4820{ 4821 int error; 4822 4823 ASSERT(MUTEX_HELD(&zfs_share_lock)); 4824 /* Both NFS and SMB shares also require sharetab support. */ 4825 if (sharefs_mod == NULL && ((sharefs_mod = 4826 ddi_modopen("fs/sharefs", 4827 KRTLD_MODE_FIRST, &error)) == NULL)) { 4828 return (SET_ERROR(ENOSYS)); 4829 } 4830 if (zshare_fs == NULL && ((zshare_fs = 4831 (int (*)(enum sharefs_sys_op, share_t *, uint32_t)) 4832 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) { 4833 return (SET_ERROR(ENOSYS)); 4834 } 4835 return (0); 4836} 4837#endif /* sun */ 4838 4839static int 4840zfs_ioc_share(zfs_cmd_t *zc) 4841{ 4842#ifdef sun 4843 int error; 4844 int opcode; 4845 4846 switch (zc->zc_share.z_sharetype) { 4847 case ZFS_SHARE_NFS: 4848 case ZFS_UNSHARE_NFS: 4849 if (zfs_nfsshare_inited == 0) { 4850 mutex_enter(&zfs_share_lock); 4851 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs", 4852 KRTLD_MODE_FIRST, &error)) == NULL)) { 4853 mutex_exit(&zfs_share_lock); 4854 return (SET_ERROR(ENOSYS)); 4855 } 4856 if (znfsexport_fs == NULL && 4857 ((znfsexport_fs = (int (*)(void *)) 4858 ddi_modsym(nfs_mod, 4859 "nfs_export", &error)) == NULL)) { 4860 mutex_exit(&zfs_share_lock); 4861 return (SET_ERROR(ENOSYS)); 4862 } 4863 error = zfs_init_sharefs(); 4864 if (error != 0) { 4865 mutex_exit(&zfs_share_lock); 4866 return (SET_ERROR(ENOSYS)); 4867 } 4868 zfs_nfsshare_inited = 1; 4869 mutex_exit(&zfs_share_lock); 4870 } 4871 break; 4872 case ZFS_SHARE_SMB: 4873 case ZFS_UNSHARE_SMB: 4874 if (zfs_smbshare_inited == 0) { 4875 mutex_enter(&zfs_share_lock); 4876 if (smbsrv_mod == NULL && ((smbsrv_mod = 4877 ddi_modopen("drv/smbsrv", 4878 KRTLD_MODE_FIRST, &error)) == NULL)) { 4879 mutex_exit(&zfs_share_lock); 4880 return (SET_ERROR(ENOSYS)); 4881 } 4882 if (zsmbexport_fs == NULL && ((zsmbexport_fs = 4883 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod, 4884 "smb_server_share", &error)) == NULL)) { 4885 mutex_exit(&zfs_share_lock); 4886 return (SET_ERROR(ENOSYS)); 4887 } 4888 error = zfs_init_sharefs(); 4889 if (error != 0) { 4890 mutex_exit(&zfs_share_lock); 4891 return (SET_ERROR(ENOSYS)); 4892 } 4893 zfs_smbshare_inited = 1; 4894 mutex_exit(&zfs_share_lock); 4895 } 4896 break; 4897 default: 4898 return (SET_ERROR(EINVAL)); 4899 } 4900 4901 switch (zc->zc_share.z_sharetype) { 4902 case ZFS_SHARE_NFS: 4903 case ZFS_UNSHARE_NFS: 4904 if (error = 4905 znfsexport_fs((void *) 4906 (uintptr_t)zc->zc_share.z_exportdata)) 4907 return (error); 4908 break; 4909 case ZFS_SHARE_SMB: 4910 case ZFS_UNSHARE_SMB: 4911 if (error = zsmbexport_fs((void *) 4912 (uintptr_t)zc->zc_share.z_exportdata, 4913 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ? 4914 B_TRUE: B_FALSE)) { 4915 return (error); 4916 } 4917 break; 4918 } 4919 4920 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS || 4921 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ? 4922 SHAREFS_ADD : SHAREFS_REMOVE; 4923 4924 /* 4925 * Add or remove share from sharetab 4926 */ 4927 error = zshare_fs(opcode, 4928 (void *)(uintptr_t)zc->zc_share.z_sharedata, 4929 zc->zc_share.z_sharemax); 4930 4931 return (error); 4932 4933#else /* !sun */ 4934 return (ENOSYS); 4935#endif /* !sun */ 4936} 4937 4938ace_t full_access[] = { 4939 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0} 4940}; 4941 4942/* 4943 * inputs: 4944 * zc_name name of containing filesystem 4945 * zc_obj object # beyond which we want next in-use object # 4946 * 4947 * outputs: 4948 * zc_obj next in-use object # 4949 */ 4950static int 4951zfs_ioc_next_obj(zfs_cmd_t *zc) 4952{ 4953 objset_t *os = NULL; 4954 int error; 4955 4956 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 4957 if (error != 0) 4958 return (error); 4959 4960 error = dmu_object_next(os, &zc->zc_obj, B_FALSE, 4961 os->os_dsl_dataset->ds_phys->ds_prev_snap_txg); 4962 4963 dmu_objset_rele(os, FTAG); 4964 return (error); 4965} 4966 4967/* 4968 * inputs: 4969 * zc_name name of filesystem 4970 * zc_value prefix name for snapshot 4971 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process 4972 * 4973 * outputs: 4974 * zc_value short name of new snapshot 4975 */ 4976static int 4977zfs_ioc_tmp_snapshot(zfs_cmd_t *zc) 4978{ 4979 char *snap_name; 4980 char *hold_name; 4981 int error; 4982 minor_t minor; 4983 4984 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor); 4985 if (error != 0) 4986 return (error); 4987 4988 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value, 4989 (u_longlong_t)ddi_get_lbolt64()); 4990 hold_name = kmem_asprintf("%%%s", zc->zc_value); 4991 4992 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor, 4993 hold_name); 4994 if (error == 0) 4995 (void) strcpy(zc->zc_value, snap_name); 4996 strfree(snap_name); 4997 strfree(hold_name); 4998 zfs_onexit_fd_rele(zc->zc_cleanup_fd); 4999 return (error); 5000} 5001 5002/* 5003 * inputs: 5004 * zc_name name of "to" snapshot 5005 * zc_value name of "from" snapshot 5006 * zc_cookie file descriptor to write diff data on 5007 * 5008 * outputs: 5009 * dmu_diff_record_t's to the file descriptor 5010 */ 5011static int 5012zfs_ioc_diff(zfs_cmd_t *zc) 5013{ 5014 file_t *fp; 5015 cap_rights_t rights; 5016 offset_t off; 5017 int error; 5018 5019 fp = getf(zc->zc_cookie, cap_rights_init(&rights, CAP_WRITE)); 5020 if (fp == NULL) 5021 return (SET_ERROR(EBADF)); 5022 5023 off = fp->f_offset; 5024 5025#ifdef illumos 5026 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off); 5027#else 5028 error = dmu_diff(zc->zc_name, zc->zc_value, fp, &off); 5029#endif 5030 5031 if (off >= 0 && off <= MAXOFFSET_T) 5032 fp->f_offset = off; 5033 releasef(zc->zc_cookie); 5034 5035 return (error); 5036} 5037 5038#ifdef sun 5039/* 5040 * Remove all ACL files in shares dir 5041 */ 5042static int 5043zfs_smb_acl_purge(znode_t *dzp) 5044{ 5045 zap_cursor_t zc; 5046 zap_attribute_t zap; 5047 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 5048 int error; 5049 5050 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id); 5051 (error = zap_cursor_retrieve(&zc, &zap)) == 0; 5052 zap_cursor_advance(&zc)) { 5053 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred, 5054 NULL, 0)) != 0) 5055 break; 5056 } 5057 zap_cursor_fini(&zc); 5058 return (error); 5059} 5060#endif /* sun */ 5061 5062static int 5063zfs_ioc_smb_acl(zfs_cmd_t *zc) 5064{ 5065#ifdef sun 5066 vnode_t *vp; 5067 znode_t *dzp; 5068 vnode_t *resourcevp = NULL; 5069 znode_t *sharedir; 5070 zfsvfs_t *zfsvfs; 5071 nvlist_t *nvlist; 5072 char *src, *target; 5073 vattr_t vattr; 5074 vsecattr_t vsec; 5075 int error = 0; 5076 5077 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, 5078 NO_FOLLOW, NULL, &vp)) != 0) 5079 return (error); 5080 5081 /* Now make sure mntpnt and dataset are ZFS */ 5082 5083 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 || 5084 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), 5085 zc->zc_name) != 0)) { 5086 VN_RELE(vp); 5087 return (SET_ERROR(EINVAL)); 5088 } 5089 5090 dzp = VTOZ(vp); 5091 zfsvfs = dzp->z_zfsvfs; 5092 ZFS_ENTER(zfsvfs); 5093 5094 /* 5095 * Create share dir if its missing. 5096 */ 5097 mutex_enter(&zfsvfs->z_lock); 5098 if (zfsvfs->z_shares_dir == 0) { 5099 dmu_tx_t *tx; 5100 5101 tx = dmu_tx_create(zfsvfs->z_os); 5102 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE, 5103 ZFS_SHARES_DIR); 5104 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); 5105 error = dmu_tx_assign(tx, TXG_WAIT); 5106 if (error != 0) { 5107 dmu_tx_abort(tx); 5108 } else { 5109 error = zfs_create_share_dir(zfsvfs, tx); 5110 dmu_tx_commit(tx); 5111 } 5112 if (error != 0) { 5113 mutex_exit(&zfsvfs->z_lock); 5114 VN_RELE(vp); 5115 ZFS_EXIT(zfsvfs); 5116 return (error); 5117 } 5118 } 5119 mutex_exit(&zfsvfs->z_lock); 5120 5121 ASSERT(zfsvfs->z_shares_dir); 5122 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) { 5123 VN_RELE(vp); 5124 ZFS_EXIT(zfsvfs); 5125 return (error); 5126 } 5127 5128 switch (zc->zc_cookie) { 5129 case ZFS_SMB_ACL_ADD: 5130 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; 5131 vattr.va_type = VREG; 5132 vattr.va_mode = S_IFREG|0777; 5133 vattr.va_uid = 0; 5134 vattr.va_gid = 0; 5135 5136 vsec.vsa_mask = VSA_ACE; 5137 vsec.vsa_aclentp = &full_access; 5138 vsec.vsa_aclentsz = sizeof (full_access); 5139 vsec.vsa_aclcnt = 1; 5140 5141 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string, 5142 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec); 5143 if (resourcevp) 5144 VN_RELE(resourcevp); 5145 break; 5146 5147 case ZFS_SMB_ACL_REMOVE: 5148 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred, 5149 NULL, 0); 5150 break; 5151 5152 case ZFS_SMB_ACL_RENAME: 5153 if ((error = get_nvlist(zc->zc_nvlist_src, 5154 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) { 5155 VN_RELE(vp); 5156 ZFS_EXIT(zfsvfs); 5157 return (error); 5158 } 5159 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) || 5160 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET, 5161 &target)) { 5162 VN_RELE(vp); 5163 VN_RELE(ZTOV(sharedir)); 5164 ZFS_EXIT(zfsvfs); 5165 nvlist_free(nvlist); 5166 return (error); 5167 } 5168 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target, 5169 kcred, NULL, 0); 5170 nvlist_free(nvlist); 5171 break; 5172 5173 case ZFS_SMB_ACL_PURGE: 5174 error = zfs_smb_acl_purge(sharedir); 5175 break; 5176 5177 default: 5178 error = SET_ERROR(EINVAL); 5179 break; 5180 } 5181 5182 VN_RELE(vp); 5183 VN_RELE(ZTOV(sharedir)); 5184 5185 ZFS_EXIT(zfsvfs); 5186 5187 return (error); 5188#else /* !sun */ 5189 return (EOPNOTSUPP); 5190#endif /* !sun */ 5191} 5192 5193/* 5194 * innvl: { 5195 * "holds" -> { snapname -> holdname (string), ... } 5196 * (optional) "cleanup_fd" -> fd (int32) 5197 * } 5198 * 5199 * outnvl: { 5200 * snapname -> error value (int32) 5201 * ... 5202 * } 5203 */ 5204/* ARGSUSED */ 5205static int 5206zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist) 5207{ 5208 nvlist_t *holds; 5209 int cleanup_fd = -1; 5210 int error; 5211 minor_t minor = 0; 5212 5213 error = nvlist_lookup_nvlist(args, "holds", &holds); 5214 if (error != 0) 5215 return (SET_ERROR(EINVAL)); 5216 5217 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) { 5218 error = zfs_onexit_fd_hold(cleanup_fd, &minor); 5219 if (error != 0) 5220 return (error); 5221 } 5222 5223 error = dsl_dataset_user_hold(holds, minor, errlist); 5224 if (minor != 0) 5225 zfs_onexit_fd_rele(cleanup_fd); 5226 return (error); 5227} 5228 5229/* 5230 * innvl is not used. 5231 * 5232 * outnvl: { 5233 * holdname -> time added (uint64 seconds since epoch) 5234 * ... 5235 * } 5236 */ 5237/* ARGSUSED */ 5238static int 5239zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl) 5240{ 5241 return (dsl_dataset_get_holds(snapname, outnvl)); 5242} 5243 5244/* 5245 * innvl: { 5246 * snapname -> { holdname, ... } 5247 * ... 5248 * } 5249 * 5250 * outnvl: { 5251 * snapname -> error value (int32) 5252 * ... 5253 * } 5254 */ 5255/* ARGSUSED */ 5256static int 5257zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist) 5258{ 5259 return (dsl_dataset_user_release(holds, errlist)); 5260} 5261 5262/* 5263 * inputs: 5264 * zc_name name of new filesystem or snapshot 5265 * zc_value full name of old snapshot 5266 * 5267 * outputs: 5268 * zc_cookie space in bytes 5269 * zc_objset_type compressed space in bytes 5270 * zc_perm_action uncompressed space in bytes 5271 */ 5272static int 5273zfs_ioc_space_written(zfs_cmd_t *zc) 5274{ 5275 int error; 5276 dsl_pool_t *dp; 5277 dsl_dataset_t *new, *old; 5278 5279 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 5280 if (error != 0) 5281 return (error); 5282 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new); 5283 if (error != 0) { 5284 dsl_pool_rele(dp, FTAG); 5285 return (error); 5286 } 5287 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old); 5288 if (error != 0) { 5289 dsl_dataset_rele(new, FTAG); 5290 dsl_pool_rele(dp, FTAG); 5291 return (error); 5292 } 5293 5294 error = dsl_dataset_space_written(old, new, &zc->zc_cookie, 5295 &zc->zc_objset_type, &zc->zc_perm_action); 5296 dsl_dataset_rele(old, FTAG); 5297 dsl_dataset_rele(new, FTAG); 5298 dsl_pool_rele(dp, FTAG); 5299 return (error); 5300} 5301 5302/* 5303 * innvl: { 5304 * "firstsnap" -> snapshot name 5305 * } 5306 * 5307 * outnvl: { 5308 * "used" -> space in bytes 5309 * "compressed" -> compressed space in bytes 5310 * "uncompressed" -> uncompressed space in bytes 5311 * } 5312 */ 5313static int 5314zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl) 5315{ 5316 int error; 5317 dsl_pool_t *dp; 5318 dsl_dataset_t *new, *old; 5319 char *firstsnap; 5320 uint64_t used, comp, uncomp; 5321 5322 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0) 5323 return (SET_ERROR(EINVAL)); 5324 5325 error = dsl_pool_hold(lastsnap, FTAG, &dp); 5326 if (error != 0) 5327 return (error); 5328 5329 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new); 5330 if (error != 0) { 5331 dsl_pool_rele(dp, FTAG); 5332 return (error); 5333 } 5334 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old); 5335 if (error != 0) { 5336 dsl_dataset_rele(new, FTAG); 5337 dsl_pool_rele(dp, FTAG); 5338 return (error); 5339 } 5340 5341 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp); 5342 dsl_dataset_rele(old, FTAG); 5343 dsl_dataset_rele(new, FTAG); 5344 dsl_pool_rele(dp, FTAG); 5345 fnvlist_add_uint64(outnvl, "used", used); 5346 fnvlist_add_uint64(outnvl, "compressed", comp); 5347 fnvlist_add_uint64(outnvl, "uncompressed", uncomp); 5348 return (error); 5349} 5350 5351static int 5352zfs_ioc_jail(zfs_cmd_t *zc) 5353{ 5354 5355 return (zone_dataset_attach(curthread->td_ucred, zc->zc_name, 5356 (int)zc->zc_jailid)); 5357} 5358 5359static int 5360zfs_ioc_unjail(zfs_cmd_t *zc) 5361{ 5362 5363 return (zone_dataset_detach(curthread->td_ucred, zc->zc_name, 5364 (int)zc->zc_jailid)); 5365} 5366 5367/* 5368 * innvl: { 5369 * "fd" -> file descriptor to write stream to (int32) 5370 * (optional) "fromsnap" -> full snap name to send an incremental from 5371 * } 5372 * 5373 * outnvl is unused 5374 */ 5375/* ARGSUSED */ 5376static int 5377zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl) 5378{ 5379 cap_rights_t rights; 5380 int error; 5381 offset_t off; 5382 char *fromname = NULL; 5383 int fd; 5384 5385 error = nvlist_lookup_int32(innvl, "fd", &fd); 5386 if (error != 0) 5387 return (SET_ERROR(EINVAL)); 5388 5389 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname); 5390 5391 file_t *fp = getf(fd, cap_rights_init(&rights, CAP_READ)); 5392 if (fp == NULL) 5393 return (SET_ERROR(EBADF)); 5394 5395 off = fp->f_offset; 5396#ifdef illumos 5397 error = dmu_send(snapname, fromname, fd, fp->f_vnode, &off); 5398#else 5399 error = dmu_send(snapname, fromname, fd, fp, &off); 5400#endif 5401 5402#ifdef illumos 5403 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0) 5404 fp->f_offset = off; 5405#else 5406 fp->f_offset = off; 5407#endif 5408 5409 releasef(fd); 5410 return (error); 5411} 5412 5413/* 5414 * Determine approximately how large a zfs send stream will be -- the number 5415 * of bytes that will be written to the fd supplied to zfs_ioc_send_new(). 5416 * 5417 * innvl: { 5418 * (optional) "fromsnap" -> full snap name to send an incremental from 5419 * } 5420 * 5421 * outnvl: { 5422 * "space" -> bytes of space (uint64) 5423 * } 5424 */ 5425static int 5426zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl) 5427{ 5428 dsl_pool_t *dp; 5429 dsl_dataset_t *fromsnap = NULL; 5430 dsl_dataset_t *tosnap; 5431 int error; 5432 char *fromname; 5433 uint64_t space; 5434 5435 error = dsl_pool_hold(snapname, FTAG, &dp); 5436 if (error != 0) 5437 return (error); 5438 5439 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap); 5440 if (error != 0) { 5441 dsl_pool_rele(dp, FTAG); 5442 return (error); 5443 } 5444 5445 error = nvlist_lookup_string(innvl, "fromsnap", &fromname); 5446 if (error == 0) { 5447 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap); 5448 if (error != 0) { 5449 dsl_dataset_rele(tosnap, FTAG); 5450 dsl_pool_rele(dp, FTAG); 5451 return (error); 5452 } 5453 } 5454 5455 error = dmu_send_estimate(tosnap, fromsnap, &space); 5456 fnvlist_add_uint64(outnvl, "space", space); 5457 5458 if (fromsnap != NULL) 5459 dsl_dataset_rele(fromsnap, FTAG); 5460 dsl_dataset_rele(tosnap, FTAG); 5461 dsl_pool_rele(dp, FTAG); 5462 return (error); 5463} 5464 5465 5466static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST]; 5467 5468static void 5469zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5470 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck, 5471 boolean_t log_history, zfs_ioc_poolcheck_t pool_check) 5472{ 5473 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST]; 5474 5475 ASSERT3U(ioc, >=, ZFS_IOC_FIRST); 5476 ASSERT3U(ioc, <, ZFS_IOC_LAST); 5477 ASSERT3P(vec->zvec_legacy_func, ==, NULL); 5478 ASSERT3P(vec->zvec_func, ==, NULL); 5479 5480 vec->zvec_legacy_func = func; 5481 vec->zvec_secpolicy = secpolicy; 5482 vec->zvec_namecheck = namecheck; 5483 vec->zvec_allow_log = log_history; 5484 vec->zvec_pool_check = pool_check; 5485} 5486 5487/* 5488 * See the block comment at the beginning of this file for details on 5489 * each argument to this function. 5490 */ 5491static void 5492zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func, 5493 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck, 5494 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist, 5495 boolean_t allow_log) 5496{ 5497 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST]; 5498 5499 ASSERT3U(ioc, >=, ZFS_IOC_FIRST); 5500 ASSERT3U(ioc, <, ZFS_IOC_LAST); 5501 ASSERT3P(vec->zvec_legacy_func, ==, NULL); 5502 ASSERT3P(vec->zvec_func, ==, NULL); 5503 5504 /* if we are logging, the name must be valid */ 5505 ASSERT(!allow_log || namecheck != NO_NAME); 5506 5507 vec->zvec_name = name; 5508 vec->zvec_func = func; 5509 vec->zvec_secpolicy = secpolicy; 5510 vec->zvec_namecheck = namecheck; 5511 vec->zvec_pool_check = pool_check; 5512 vec->zvec_smush_outnvlist = smush_outnvlist; 5513 vec->zvec_allow_log = allow_log; 5514} 5515 5516static void 5517zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5518 zfs_secpolicy_func_t *secpolicy, boolean_t log_history, 5519 zfs_ioc_poolcheck_t pool_check) 5520{ 5521 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5522 POOL_NAME, log_history, pool_check); 5523} 5524 5525static void 5526zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5527 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check) 5528{ 5529 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5530 DATASET_NAME, B_FALSE, pool_check); 5531} 5532 5533static void 5534zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func) 5535{ 5536 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config, 5537 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5538} 5539 5540static void 5541zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5542 zfs_secpolicy_func_t *secpolicy) 5543{ 5544 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5545 NO_NAME, B_FALSE, POOL_CHECK_NONE); 5546} 5547 5548static void 5549zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc, 5550 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy) 5551{ 5552 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5553 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED); 5554} 5555 5556static void 5557zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func) 5558{ 5559 zfs_ioctl_register_dataset_read_secpolicy(ioc, func, 5560 zfs_secpolicy_read); 5561} 5562 5563static void 5564zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5565 zfs_secpolicy_func_t *secpolicy) 5566{ 5567 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5568 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5569} 5570 5571static void 5572zfs_ioctl_init(void) 5573{ 5574 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT, 5575 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME, 5576 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5577 5578 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY, 5579 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME, 5580 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE); 5581 5582 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS, 5583 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME, 5584 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5585 5586 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW, 5587 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME, 5588 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5589 5590 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE, 5591 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME, 5592 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5593 5594 zfs_ioctl_register("create", ZFS_IOC_CREATE, 5595 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME, 5596 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5597 5598 zfs_ioctl_register("clone", ZFS_IOC_CLONE, 5599 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME, 5600 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5601 5602 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS, 5603 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME, 5604 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5605 5606 zfs_ioctl_register("hold", ZFS_IOC_HOLD, 5607 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME, 5608 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5609 zfs_ioctl_register("release", ZFS_IOC_RELEASE, 5610 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME, 5611 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5612 5613 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS, 5614 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME, 5615 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5616 5617 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK, 5618 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, 5619 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE); 5620 5621 zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK, 5622 zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME, 5623 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5624 5625 zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS, 5626 zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME, 5627 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5628 5629 zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS, 5630 zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks, 5631 POOL_NAME, 5632 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5633 5634 /* IOCTLS that use the legacy function signature */ 5635 5636 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze, 5637 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY); 5638 5639 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create, 5640 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5641 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN, 5642 zfs_ioc_pool_scan); 5643 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE, 5644 zfs_ioc_pool_upgrade); 5645 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD, 5646 zfs_ioc_vdev_add); 5647 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE, 5648 zfs_ioc_vdev_remove); 5649 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE, 5650 zfs_ioc_vdev_set_state); 5651 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH, 5652 zfs_ioc_vdev_attach); 5653 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH, 5654 zfs_ioc_vdev_detach); 5655 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH, 5656 zfs_ioc_vdev_setpath); 5657 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU, 5658 zfs_ioc_vdev_setfru); 5659 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS, 5660 zfs_ioc_pool_set_props); 5661 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT, 5662 zfs_ioc_vdev_split); 5663 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID, 5664 zfs_ioc_pool_reguid); 5665 5666 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS, 5667 zfs_ioc_pool_configs, zfs_secpolicy_none); 5668 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT, 5669 zfs_ioc_pool_tryimport, zfs_secpolicy_config); 5670 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT, 5671 zfs_ioc_inject_fault, zfs_secpolicy_inject); 5672 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT, 5673 zfs_ioc_clear_fault, zfs_secpolicy_inject); 5674 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT, 5675 zfs_ioc_inject_list_next, zfs_secpolicy_inject); 5676 5677 /* 5678 * pool destroy, and export don't log the history as part of 5679 * zfsdev_ioctl, but rather zfs_ioc_pool_export 5680 * does the logging of those commands. 5681 */ 5682 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy, 5683 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE); 5684 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export, 5685 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE); 5686 5687 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats, 5688 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE); 5689 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props, 5690 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE); 5691 5692 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log, 5693 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_NONE); 5694 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME, 5695 zfs_ioc_dsobj_to_dsname, 5696 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_NONE); 5697 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY, 5698 zfs_ioc_pool_get_history, 5699 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED); 5700 5701 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import, 5702 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5703 5704 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear, 5705 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5706 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen, 5707 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED); 5708 5709 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN, 5710 zfs_ioc_space_written); 5711 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS, 5712 zfs_ioc_objset_recvd_props); 5713 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ, 5714 zfs_ioc_next_obj); 5715 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL, 5716 zfs_ioc_get_fsacl); 5717 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS, 5718 zfs_ioc_objset_stats); 5719 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS, 5720 zfs_ioc_objset_zplprops); 5721 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT, 5722 zfs_ioc_dataset_list_next); 5723 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT, 5724 zfs_ioc_snapshot_list_next); 5725 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS, 5726 zfs_ioc_send_progress); 5727 5728 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF, 5729 zfs_ioc_diff, zfs_secpolicy_diff); 5730 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS, 5731 zfs_ioc_obj_to_stats, zfs_secpolicy_diff); 5732 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH, 5733 zfs_ioc_obj_to_path, zfs_secpolicy_diff); 5734 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE, 5735 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one); 5736 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY, 5737 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many); 5738 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND, 5739 zfs_ioc_send, zfs_secpolicy_send); 5740 5741 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop, 5742 zfs_secpolicy_none); 5743 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy, 5744 zfs_secpolicy_destroy); 5745 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename, 5746 zfs_secpolicy_rename); 5747 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv, 5748 zfs_secpolicy_recv); 5749 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote, 5750 zfs_secpolicy_promote); 5751 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP, 5752 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop); 5753 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl, 5754 zfs_secpolicy_set_fsacl); 5755 5756 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share, 5757 zfs_secpolicy_share, POOL_CHECK_NONE); 5758 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl, 5759 zfs_secpolicy_smb_acl, POOL_CHECK_NONE); 5760 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE, 5761 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade, 5762 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5763 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT, 5764 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot, 5765 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5766 5767#ifdef __FreeBSD__ 5768 zfs_ioctl_register_dataset_nolog(ZFS_IOC_JAIL, zfs_ioc_jail, 5769 zfs_secpolicy_config, POOL_CHECK_NONE); 5770 zfs_ioctl_register_dataset_nolog(ZFS_IOC_UNJAIL, zfs_ioc_unjail, 5771 zfs_secpolicy_config, POOL_CHECK_NONE); 5772#endif 5773} 5774 5775int 5776pool_status_check(const char *name, zfs_ioc_namecheck_t type, 5777 zfs_ioc_poolcheck_t check) 5778{ 5779 spa_t *spa; 5780 int error; 5781 5782 ASSERT(type == POOL_NAME || type == DATASET_NAME); 5783 5784 if (check & POOL_CHECK_NONE) 5785 return (0); 5786 5787 error = spa_open(name, &spa, FTAG); 5788 if (error == 0) { 5789 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa)) 5790 error = SET_ERROR(EAGAIN); 5791 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa)) 5792 error = SET_ERROR(EROFS); 5793 spa_close(spa, FTAG); 5794 } 5795 return (error); 5796} 5797 5798/* 5799 * Find a free minor number. 5800 */ 5801minor_t 5802zfsdev_minor_alloc(void) 5803{ 5804 static minor_t last_minor; 5805 minor_t m; 5806 5807 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 5808 5809 for (m = last_minor + 1; m != last_minor; m++) { 5810 if (m > ZFSDEV_MAX_MINOR) 5811 m = 1; 5812 if (ddi_get_soft_state(zfsdev_state, m) == NULL) { 5813 last_minor = m; 5814 return (m); 5815 } 5816 } 5817 5818 return (0); 5819} 5820 5821static int 5822zfs_ctldev_init(struct cdev *devp) 5823{ 5824 minor_t minor; 5825 zfs_soft_state_t *zs; 5826 5827 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 5828 5829 minor = zfsdev_minor_alloc(); 5830 if (minor == 0) 5831 return (SET_ERROR(ENXIO)); 5832 5833 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS) 5834 return (SET_ERROR(EAGAIN)); 5835 5836 devfs_set_cdevpriv((void *)(uintptr_t)minor, zfsdev_close); 5837 5838 zs = ddi_get_soft_state(zfsdev_state, minor); 5839 zs->zss_type = ZSST_CTLDEV; 5840 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data); 5841 5842 return (0); 5843} 5844 5845static void 5846zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor) 5847{ 5848 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 5849 5850 zfs_onexit_destroy(zo); 5851 ddi_soft_state_free(zfsdev_state, minor); 5852} 5853 5854void * 5855zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which) 5856{ 5857 zfs_soft_state_t *zp; 5858 5859 zp = ddi_get_soft_state(zfsdev_state, minor); 5860 if (zp == NULL || zp->zss_type != which) 5861 return (NULL); 5862 5863 return (zp->zss_data); 5864} 5865 5866static int 5867zfsdev_open(struct cdev *devp, int flag, int mode, struct thread *td) 5868{ 5869 int error = 0; 5870 5871#ifdef sun 5872 if (getminor(*devp) != 0) 5873 return (zvol_open(devp, flag, otyp, cr)); 5874#endif 5875 5876 /* This is the control device. Allocate a new minor if requested. */ 5877 if (flag & FEXCL) { 5878 mutex_enter(&spa_namespace_lock); 5879 error = zfs_ctldev_init(devp); 5880 mutex_exit(&spa_namespace_lock); 5881 } 5882 5883 return (error); 5884} 5885 5886static void 5887zfsdev_close(void *data) 5888{ 5889 zfs_onexit_t *zo; 5890 minor_t minor = (minor_t)(uintptr_t)data; 5891 5892 if (minor == 0) 5893 return; 5894 5895 mutex_enter(&spa_namespace_lock); 5896 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV); 5897 if (zo == NULL) { 5898 mutex_exit(&spa_namespace_lock); 5899 return; 5900 } 5901 zfs_ctldev_destroy(zo, minor); 5902 mutex_exit(&spa_namespace_lock); 5903} 5904 5905static int 5906zfsdev_ioctl(struct cdev *dev, u_long zcmd, caddr_t arg, int flag, 5907 struct thread *td) 5908{ 5909 zfs_cmd_t *zc; 5910 uint_t vecnum; 5911 int error, rc, len; 5912#ifdef illumos 5913 minor_t minor = getminor(dev); 5914#else 5915 zfs_iocparm_t *zc_iocparm; 5916 int cflag, cmd, oldvecnum; 5917 boolean_t newioc, compat; 5918 cred_t *cr = td->td_ucred; 5919#endif 5920 const zfs_ioc_vec_t *vec; 5921 char *saved_poolname = NULL; 5922 nvlist_t *innvl = NULL; 5923 5924 cflag = ZFS_CMD_COMPAT_NONE; 5925 compat = B_FALSE; 5926 newioc = B_TRUE; 5927 5928 len = IOCPARM_LEN(zcmd); 5929 cmd = zcmd & 0xff; 5930 5931 /* 5932 * Check if we are talking to supported older binaries 5933 * and translate zfs_cmd if necessary 5934 */ 5935 if (len != sizeof(zfs_iocparm_t)) { 5936 newioc = B_FALSE; 5937 if (len == sizeof(zfs_cmd_t)) { 5938 cflag = ZFS_CMD_COMPAT_LZC; 5939 vecnum = cmd; 5940 } else if (len == sizeof(zfs_cmd_deadman_t)) { 5941 cflag = ZFS_CMD_COMPAT_DEADMAN; 5942 compat = B_TRUE; 5943 vecnum = cmd; 5944 } else if (len == sizeof(zfs_cmd_v28_t)) { 5945 cflag = ZFS_CMD_COMPAT_V28; 5946 compat = B_TRUE; 5947 vecnum = cmd; 5948 } else if (len == sizeof(zfs_cmd_v15_t)) { 5949 cflag = ZFS_CMD_COMPAT_V15; 5950 compat = B_TRUE; 5951 vecnum = zfs_ioctl_v15_to_v28[cmd]; 5952 } else 5953 return (EINVAL); 5954 } else 5955 vecnum = cmd; 5956 5957#ifdef illumos 5958 vecnum = cmd - ZFS_IOC_FIRST; 5959 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip)); 5960#endif 5961 5962 if (compat) { 5963 if (vecnum == ZFS_IOC_COMPAT_PASS) 5964 return (0); 5965 else if (vecnum == ZFS_IOC_COMPAT_FAIL) 5966 return (ENOTSUP); 5967 } 5968 5969 /* 5970 * Check if we have sufficient kernel memory allocated 5971 * for the zfs_cmd_t request. Bail out if not so we 5972 * will not access undefined memory region. 5973 */ 5974 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0])) 5975 return (SET_ERROR(EINVAL)); 5976 vec = &zfs_ioc_vec[vecnum]; 5977 5978#ifdef illumos 5979 zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP); 5980 bzero(zc, sizeof(zfs_cmd_t)); 5981 5982 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag); 5983 if (error != 0) { 5984 error = SET_ERROR(EFAULT); 5985 goto out; 5986 } 5987#else /* !illumos */ 5988 /* 5989 * We don't alloc/free zc only if talking to library ioctl version 2 5990 */ 5991 if (cflag != ZFS_CMD_COMPAT_LZC) { 5992 zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP); 5993 bzero(zc, sizeof(zfs_cmd_t)); 5994 } else { 5995 zc = (void *)arg; 5996 error = 0; 5997 } 5998 5999 if (newioc) { 6000 zc_iocparm = (void *)arg; 6001 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_t)) { 6002 error = SET_ERROR(EFAULT); 6003 goto out; 6004 } 6005 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd, zc, 6006 sizeof(zfs_cmd_t), flag); 6007 if (error != 0) { 6008 error = SET_ERROR(EFAULT); 6009 goto out; 6010 } 6011 } 6012 6013 if (compat) { 6014 zfs_cmd_compat_get(zc, arg, cflag); 6015 oldvecnum = vecnum; 6016 error = zfs_ioctl_compat_pre(zc, &vecnum, cflag); 6017 if (error != 0) 6018 goto out; 6019 if (oldvecnum != vecnum) 6020 vec = &zfs_ioc_vec[vecnum]; 6021 } 6022#endif /* !illumos */ 6023 6024 zc->zc_iflags = flag & FKIOCTL; 6025 if (zc->zc_nvlist_src_size != 0) { 6026 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 6027 zc->zc_iflags, &innvl); 6028 if (error != 0) 6029 goto out; 6030 } 6031 6032 /* rewrite innvl for backwards compatibility */ 6033 if (compat) 6034 innvl = zfs_ioctl_compat_innvl(zc, innvl, vecnum, cflag); 6035 6036 /* 6037 * Ensure that all pool/dataset names are valid before we pass down to 6038 * the lower layers. 6039 */ 6040 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0'; 6041 switch (vec->zvec_namecheck) { 6042 case POOL_NAME: 6043 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0) 6044 error = SET_ERROR(EINVAL); 6045 else 6046 error = pool_status_check(zc->zc_name, 6047 vec->zvec_namecheck, vec->zvec_pool_check); 6048 break; 6049 6050 case DATASET_NAME: 6051 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0) 6052 error = SET_ERROR(EINVAL); 6053 else 6054 error = pool_status_check(zc->zc_name, 6055 vec->zvec_namecheck, vec->zvec_pool_check); 6056 break; 6057 6058 case NO_NAME: 6059 break; 6060 } 6061 6062 if (error == 0 && !(flag & FKIOCTL)) 6063 error = vec->zvec_secpolicy(zc, innvl, cr); 6064 6065 if (error != 0) 6066 goto out; 6067 6068 /* legacy ioctls can modify zc_name */ 6069 len = strcspn(zc->zc_name, "/@#") + 1; 6070 saved_poolname = kmem_alloc(len, KM_SLEEP); 6071 (void) strlcpy(saved_poolname, zc->zc_name, len); 6072 6073 if (vec->zvec_func != NULL) { 6074 nvlist_t *outnvl; 6075 int puterror = 0; 6076 spa_t *spa; 6077 nvlist_t *lognv = NULL; 6078 6079 ASSERT(vec->zvec_legacy_func == NULL); 6080 6081 /* 6082 * Add the innvl to the lognv before calling the func, 6083 * in case the func changes the innvl. 6084 */ 6085 if (vec->zvec_allow_log) { 6086 lognv = fnvlist_alloc(); 6087 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL, 6088 vec->zvec_name); 6089 if (!nvlist_empty(innvl)) { 6090 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL, 6091 innvl); 6092 } 6093 } 6094 6095 outnvl = fnvlist_alloc(); 6096 error = vec->zvec_func(zc->zc_name, innvl, outnvl); 6097 6098 if (error == 0 && vec->zvec_allow_log && 6099 spa_open(zc->zc_name, &spa, FTAG) == 0) { 6100 if (!nvlist_empty(outnvl)) { 6101 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL, 6102 outnvl); 6103 } 6104 (void) spa_history_log_nvl(spa, lognv); 6105 spa_close(spa, FTAG); 6106 } 6107 fnvlist_free(lognv); 6108 6109 /* rewrite outnvl for backwards compatibility */ 6110 if (cflag != ZFS_CMD_COMPAT_NONE && cflag != ZFS_CMD_COMPAT_LZC) 6111 outnvl = zfs_ioctl_compat_outnvl(zc, outnvl, vecnum, 6112 cflag); 6113 6114 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) { 6115 int smusherror = 0; 6116 if (vec->zvec_smush_outnvlist) { 6117 smusherror = nvlist_smush(outnvl, 6118 zc->zc_nvlist_dst_size); 6119 } 6120 if (smusherror == 0) 6121 puterror = put_nvlist(zc, outnvl); 6122 } 6123 6124 if (puterror != 0) 6125 error = puterror; 6126 6127 nvlist_free(outnvl); 6128 } else { 6129 error = vec->zvec_legacy_func(zc); 6130 } 6131 6132out: 6133 nvlist_free(innvl); 6134 6135 if (compat) { 6136 zfs_ioctl_compat_post(zc, cmd, cflag); 6137 zfs_cmd_compat_put(zc, arg, vecnum, cflag); 6138 } 6139 6140#ifdef illumos 6141 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag); 6142 if (error == 0 && rc != 0) 6143 error = SET_ERROR(EFAULT); 6144#else 6145 if (newioc) { 6146 rc = ddi_copyout(zc, (void *)(uintptr_t)zc_iocparm->zfs_cmd, 6147 sizeof (zfs_cmd_t), flag); 6148 if (error == 0 && rc != 0) 6149 error = SET_ERROR(EFAULT); 6150 } 6151#endif 6152 if (error == 0 && vec->zvec_allow_log) { 6153 char *s = tsd_get(zfs_allow_log_key); 6154 if (s != NULL) 6155 strfree(s); 6156 (void) tsd_set(zfs_allow_log_key, saved_poolname); 6157 } else { 6158 if (saved_poolname != NULL) 6159 strfree(saved_poolname); 6160 } 6161 6162#ifdef illumos 6163 kmem_free(zc, sizeof (zfs_cmd_t)); 6164#else 6165 /* 6166 * We don't alloc/free zc only if talking to library ioctl version 2 6167 */ 6168 if (cflag != ZFS_CMD_COMPAT_LZC) 6169 kmem_free(zc, sizeof (zfs_cmd_t)); 6170#endif 6171 return (error); 6172} 6173 6174#ifdef sun 6175static int 6176zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd) 6177{ 6178 if (cmd != DDI_ATTACH) 6179 return (DDI_FAILURE); 6180 6181 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0, 6182 DDI_PSEUDO, 0) == DDI_FAILURE) 6183 return (DDI_FAILURE); 6184 6185 zfs_dip = dip; 6186 6187 ddi_report_dev(dip); 6188 6189 return (DDI_SUCCESS); 6190} 6191 6192static int 6193zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd) 6194{ 6195 if (spa_busy() || zfs_busy() || zvol_busy()) 6196 return (DDI_FAILURE); 6197 6198 if (cmd != DDI_DETACH) 6199 return (DDI_FAILURE); 6200 6201 zfs_dip = NULL; 6202 6203 ddi_prop_remove_all(dip); 6204 ddi_remove_minor_node(dip, NULL); 6205 6206 return (DDI_SUCCESS); 6207} 6208 6209/*ARGSUSED*/ 6210static int 6211zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result) 6212{ 6213 switch (infocmd) { 6214 case DDI_INFO_DEVT2DEVINFO: 6215 *result = zfs_dip; 6216 return (DDI_SUCCESS); 6217 6218 case DDI_INFO_DEVT2INSTANCE: 6219 *result = (void *)0; 6220 return (DDI_SUCCESS); 6221 } 6222 6223 return (DDI_FAILURE); 6224} 6225#endif /* sun */ 6226 6227/* 6228 * OK, so this is a little weird. 6229 * 6230 * /dev/zfs is the control node, i.e. minor 0. 6231 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0. 6232 * 6233 * /dev/zfs has basically nothing to do except serve up ioctls, 6234 * so most of the standard driver entry points are in zvol.c. 6235 */ 6236#ifdef sun 6237static struct cb_ops zfs_cb_ops = { 6238 zfsdev_open, /* open */ 6239 zfsdev_close, /* close */ 6240 zvol_strategy, /* strategy */ 6241 nodev, /* print */ 6242 zvol_dump, /* dump */ 6243 zvol_read, /* read */ 6244 zvol_write, /* write */ 6245 zfsdev_ioctl, /* ioctl */ 6246 nodev, /* devmap */ 6247 nodev, /* mmap */ 6248 nodev, /* segmap */ 6249 nochpoll, /* poll */ 6250 ddi_prop_op, /* prop_op */ 6251 NULL, /* streamtab */ 6252 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */ 6253 CB_REV, /* version */ 6254 nodev, /* async read */ 6255 nodev, /* async write */ 6256}; 6257 6258static struct dev_ops zfs_dev_ops = { 6259 DEVO_REV, /* version */ 6260 0, /* refcnt */ 6261 zfs_info, /* info */ 6262 nulldev, /* identify */ 6263 nulldev, /* probe */ 6264 zfs_attach, /* attach */ 6265 zfs_detach, /* detach */ 6266 nodev, /* reset */ 6267 &zfs_cb_ops, /* driver operations */ 6268 NULL, /* no bus operations */ 6269 NULL, /* power */ 6270 ddi_quiesce_not_needed, /* quiesce */ 6271}; 6272 6273static struct modldrv zfs_modldrv = { 6274 &mod_driverops, 6275 "ZFS storage pool", 6276 &zfs_dev_ops 6277}; 6278 6279static struct modlinkage modlinkage = { 6280 MODREV_1, 6281 (void *)&zfs_modlfs, 6282 (void *)&zfs_modldrv, 6283 NULL 6284}; 6285#endif /* sun */ 6286 6287static struct cdevsw zfs_cdevsw = { 6288 .d_version = D_VERSION, 6289 .d_open = zfsdev_open, 6290 .d_ioctl = zfsdev_ioctl, 6291 .d_name = ZFS_DEV_NAME 6292}; 6293 6294static void 6295zfs_allow_log_destroy(void *arg) 6296{ 6297 char *poolname = arg; 6298 strfree(poolname); 6299} 6300 6301static void 6302zfsdev_init(void) 6303{ 6304 zfsdev = make_dev(&zfs_cdevsw, 0x0, UID_ROOT, GID_OPERATOR, 0666, 6305 ZFS_DEV_NAME); 6306} 6307 6308static void 6309zfsdev_fini(void) 6310{ 6311 if (zfsdev != NULL) 6312 destroy_dev(zfsdev); 6313} 6314 6315static struct root_hold_token *zfs_root_token; 6316struct proc *zfsproc; 6317 6318#ifdef sun 6319int 6320_init(void) 6321{ 6322 int error; 6323 6324 spa_init(FREAD | FWRITE); 6325 zfs_init(); 6326 zvol_init(); 6327 zfs_ioctl_init(); 6328 6329 if ((error = mod_install(&modlinkage)) != 0) { 6330 zvol_fini(); 6331 zfs_fini(); 6332 spa_fini(); 6333 return (error); 6334 } 6335 6336 tsd_create(&zfs_fsyncer_key, NULL); 6337 tsd_create(&rrw_tsd_key, rrw_tsd_destroy); 6338 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy); 6339 6340 error = ldi_ident_from_mod(&modlinkage, &zfs_li); 6341 ASSERT(error == 0); 6342 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL); 6343 6344 return (0); 6345} 6346 6347int 6348_fini(void) 6349{ 6350 int error; 6351 6352 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled) 6353 return (SET_ERROR(EBUSY)); 6354 6355 if ((error = mod_remove(&modlinkage)) != 0) 6356 return (error); 6357 6358 zvol_fini(); 6359 zfs_fini(); 6360 spa_fini(); 6361 if (zfs_nfsshare_inited) 6362 (void) ddi_modclose(nfs_mod); 6363 if (zfs_smbshare_inited) 6364 (void) ddi_modclose(smbsrv_mod); 6365 if (zfs_nfsshare_inited || zfs_smbshare_inited) 6366 (void) ddi_modclose(sharefs_mod); 6367 6368 tsd_destroy(&zfs_fsyncer_key); 6369 ldi_ident_release(zfs_li); 6370 zfs_li = NULL; 6371 mutex_destroy(&zfs_share_lock); 6372 6373 return (error); 6374} 6375 6376int 6377_info(struct modinfo *modinfop) 6378{ 6379 return (mod_info(&modlinkage, modinfop)); 6380} 6381#endif /* sun */ 6382 6383static int zfs__init(void); 6384static int zfs__fini(void); 6385static void zfs_shutdown(void *, int); 6386 6387static eventhandler_tag zfs_shutdown_event_tag; 6388 6389int 6390zfs__init(void) 6391{ 6392 6393 zfs_root_token = root_mount_hold("ZFS"); 6394 6395 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL); 6396 6397 spa_init(FREAD | FWRITE); 6398 zfs_init(); 6399 zvol_init(); 6400 zfs_ioctl_init(); 6401 6402 tsd_create(&zfs_fsyncer_key, NULL); 6403 tsd_create(&rrw_tsd_key, rrw_tsd_destroy); 6404 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy); 6405 6406 printf("ZFS storage pool version: features support (" SPA_VERSION_STRING ")\n"); 6407 root_mount_rel(zfs_root_token); 6408 6409 zfsdev_init(); 6410 6411 return (0); 6412} 6413 6414int 6415zfs__fini(void) 6416{ 6417 if (spa_busy() || zfs_busy() || zvol_busy() || 6418 zio_injection_enabled) { 6419 return (EBUSY); 6420 } 6421 6422 zfsdev_fini(); 6423 zvol_fini(); 6424 zfs_fini(); 6425 spa_fini(); 6426 6427 tsd_destroy(&zfs_fsyncer_key); 6428 tsd_destroy(&rrw_tsd_key); 6429 tsd_destroy(&zfs_allow_log_key); 6430 6431 mutex_destroy(&zfs_share_lock); 6432 6433 return (0); 6434} 6435 6436static void 6437zfs_shutdown(void *arg __unused, int howto __unused) 6438{ 6439 6440 /* 6441 * ZFS fini routines can not properly work in a panic-ed system. 6442 */ 6443 if (panicstr == NULL) 6444 (void)zfs__fini(); 6445} 6446 6447 6448static int 6449zfs_modevent(module_t mod, int type, void *unused __unused) 6450{ 6451 int err; 6452 6453 switch (type) { 6454 case MOD_LOAD: 6455 err = zfs__init(); 6456 if (err == 0) 6457 zfs_shutdown_event_tag = EVENTHANDLER_REGISTER( 6458 shutdown_post_sync, zfs_shutdown, NULL, 6459 SHUTDOWN_PRI_FIRST); 6460 return (err); 6461 case MOD_UNLOAD: 6462 err = zfs__fini(); 6463 if (err == 0 && zfs_shutdown_event_tag != NULL) 6464 EVENTHANDLER_DEREGISTER(shutdown_post_sync, 6465 zfs_shutdown_event_tag); 6466 return (err); 6467 case MOD_SHUTDOWN: 6468 return (0); 6469 default: 6470 break; 6471 } 6472 return (EOPNOTSUPP); 6473} 6474 6475static moduledata_t zfs_mod = { 6476 "zfsctrl", 6477 zfs_modevent, 6478 0 6479}; 6480DECLARE_MODULE(zfsctrl, zfs_mod, SI_SUB_VFS, SI_ORDER_ANY); 6481MODULE_VERSION(zfsctrl, 1); 6482MODULE_DEPEND(zfsctrl, opensolaris, 1, 1, 1); 6483MODULE_DEPEND(zfsctrl, krpc, 1, 1, 1); 6484MODULE_DEPEND(zfsctrl, acl_nfs4, 1, 1, 1); 6485