zfs_ioctl.c revision 268658
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 * Copyright (c) 2014, Nexenta Systems, Inc. All rights reserved. 33 */ 34 35/* 36 * ZFS ioctls. 37 * 38 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage 39 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool. 40 * 41 * There are two ways that we handle ioctls: the legacy way where almost 42 * all of the logic is in the ioctl callback, and the new way where most 43 * of the marshalling is handled in the common entry point, zfsdev_ioctl(). 44 * 45 * Non-legacy ioctls should be registered by calling 46 * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked 47 * from userland by lzc_ioctl(). 48 * 49 * The registration arguments are as follows: 50 * 51 * const char *name 52 * The name of the ioctl. This is used for history logging. If the 53 * ioctl returns successfully (the callback returns 0), and allow_log 54 * is true, then a history log entry will be recorded with the input & 55 * output nvlists. The log entry can be printed with "zpool history -i". 56 * 57 * zfs_ioc_t ioc 58 * The ioctl request number, which userland will pass to ioctl(2). 59 * The ioctl numbers can change from release to release, because 60 * the caller (libzfs) must be matched to the kernel. 61 * 62 * zfs_secpolicy_func_t *secpolicy 63 * This function will be called before the zfs_ioc_func_t, to 64 * determine if this operation is permitted. It should return EPERM 65 * on failure, and 0 on success. Checks include determining if the 66 * dataset is visible in this zone, and if the user has either all 67 * zfs privileges in the zone (SYS_MOUNT), or has been granted permission 68 * to do this operation on this dataset with "zfs allow". 69 * 70 * zfs_ioc_namecheck_t namecheck 71 * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool 72 * name, a dataset name, or nothing. If the name is not well-formed, 73 * the ioctl will fail and the callback will not be called. 74 * Therefore, the callback can assume that the name is well-formed 75 * (e.g. is null-terminated, doesn't have more than one '@' character, 76 * doesn't have invalid characters). 77 * 78 * zfs_ioc_poolcheck_t pool_check 79 * This specifies requirements on the pool state. If the pool does 80 * not meet them (is suspended or is readonly), the ioctl will fail 81 * and the callback will not be called. If any checks are specified 82 * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME. 83 * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED | 84 * POOL_CHECK_READONLY). 85 * 86 * boolean_t smush_outnvlist 87 * If smush_outnvlist is true, then the output is presumed to be a 88 * list of errors, and it will be "smushed" down to fit into the 89 * caller's buffer, by removing some entries and replacing them with a 90 * single "N_MORE_ERRORS" entry indicating how many were removed. See 91 * nvlist_smush() for details. If smush_outnvlist is false, and the 92 * outnvlist does not fit into the userland-provided buffer, then the 93 * ioctl will fail with ENOMEM. 94 * 95 * zfs_ioc_func_t *func 96 * The callback function that will perform the operation. 97 * 98 * The callback should return 0 on success, or an error number on 99 * failure. If the function fails, the userland ioctl will return -1, 100 * and errno will be set to the callback's return value. The callback 101 * will be called with the following arguments: 102 * 103 * const char *name 104 * The name of the pool or dataset to operate on, from 105 * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the 106 * expected type (pool, dataset, or none). 107 * 108 * nvlist_t *innvl 109 * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or 110 * NULL if no input nvlist was provided. Changes to this nvlist are 111 * ignored. If the input nvlist could not be deserialized, the 112 * ioctl will fail and the callback will not be called. 113 * 114 * nvlist_t *outnvl 115 * The output nvlist, initially empty. The callback can fill it in, 116 * and it will be returned to userland by serializing it into 117 * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization 118 * fails (e.g. because the caller didn't supply a large enough 119 * buffer), then the overall ioctl will fail. See the 120 * 'smush_nvlist' argument above for additional behaviors. 121 * 122 * There are two typical uses of the output nvlist: 123 * - To return state, e.g. property values. In this case, 124 * smush_outnvlist should be false. If the buffer was not large 125 * enough, the caller will reallocate a larger buffer and try 126 * the ioctl again. 127 * 128 * - To return multiple errors from an ioctl which makes on-disk 129 * changes. In this case, smush_outnvlist should be true. 130 * Ioctls which make on-disk modifications should generally not 131 * use the outnvl if they succeed, because the caller can not 132 * distinguish between the operation failing, and 133 * deserialization failing. 134 */ 135 136#include <sys/types.h> 137#include <sys/param.h> 138#include <sys/systm.h> 139#include <sys/conf.h> 140#include <sys/kernel.h> 141#include <sys/lock.h> 142#include <sys/malloc.h> 143#include <sys/mutex.h> 144#include <sys/proc.h> 145#include <sys/errno.h> 146#include <sys/uio.h> 147#include <sys/buf.h> 148#include <sys/file.h> 149#include <sys/kmem.h> 150#include <sys/conf.h> 151#include <sys/cmn_err.h> 152#include <sys/stat.h> 153#include <sys/zfs_ioctl.h> 154#include <sys/zfs_vfsops.h> 155#include <sys/zfs_znode.h> 156#include <sys/zap.h> 157#include <sys/spa.h> 158#include <sys/spa_impl.h> 159#include <sys/vdev.h> 160#include <sys/dmu.h> 161#include <sys/dsl_dir.h> 162#include <sys/dsl_dataset.h> 163#include <sys/dsl_prop.h> 164#include <sys/dsl_deleg.h> 165#include <sys/dmu_objset.h> 166#include <sys/dmu_impl.h> 167#include <sys/dmu_tx.h> 168#include <sys/sunddi.h> 169#include <sys/policy.h> 170#include <sys/zone.h> 171#include <sys/nvpair.h> 172#include <sys/mount.h> 173#include <sys/taskqueue.h> 174#include <sys/sdt.h> 175#include <sys/varargs.h> 176#include <sys/fs/zfs.h> 177#include <sys/zfs_ctldir.h> 178#include <sys/zfs_dir.h> 179#include <sys/zfs_onexit.h> 180#include <sys/zvol.h> 181#include <sys/dsl_scan.h> 182#include <sys/dmu_objset.h> 183#include <sys/dmu_send.h> 184#include <sys/dsl_destroy.h> 185#include <sys/dsl_bookmark.h> 186#include <sys/dsl_userhold.h> 187#include <sys/zfeature.h> 188 189#include "zfs_namecheck.h" 190#include "zfs_prop.h" 191#include "zfs_deleg.h" 192#include "zfs_comutil.h" 193#include "zfs_ioctl_compat.h" 194 195CTASSERT(sizeof(zfs_cmd_t) < IOCPARM_MAX); 196 197static int snapshot_list_prefetch; 198SYSCTL_DECL(_vfs_zfs); 199TUNABLE_INT("vfs.zfs.snapshot_list_prefetch", &snapshot_list_prefetch); 200SYSCTL_INT(_vfs_zfs, OID_AUTO, snapshot_list_prefetch, CTLFLAG_RW, 201 &snapshot_list_prefetch, 0, "Prefetch data when listing snapshots"); 202 203static struct cdev *zfsdev; 204 205extern void zfs_init(void); 206extern void zfs_fini(void); 207 208uint_t zfs_fsyncer_key; 209extern uint_t rrw_tsd_key; 210static uint_t zfs_allow_log_key; 211 212typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *); 213typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *); 214typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *); 215 216typedef enum { 217 NO_NAME, 218 POOL_NAME, 219 DATASET_NAME 220} zfs_ioc_namecheck_t; 221 222typedef enum { 223 POOL_CHECK_NONE = 1 << 0, 224 POOL_CHECK_SUSPENDED = 1 << 1, 225 POOL_CHECK_READONLY = 1 << 2, 226} zfs_ioc_poolcheck_t; 227 228typedef struct zfs_ioc_vec { 229 zfs_ioc_legacy_func_t *zvec_legacy_func; 230 zfs_ioc_func_t *zvec_func; 231 zfs_secpolicy_func_t *zvec_secpolicy; 232 zfs_ioc_namecheck_t zvec_namecheck; 233 boolean_t zvec_allow_log; 234 zfs_ioc_poolcheck_t zvec_pool_check; 235 boolean_t zvec_smush_outnvlist; 236 const char *zvec_name; 237} zfs_ioc_vec_t; 238 239/* This array is indexed by zfs_userquota_prop_t */ 240static const char *userquota_perms[] = { 241 ZFS_DELEG_PERM_USERUSED, 242 ZFS_DELEG_PERM_USERQUOTA, 243 ZFS_DELEG_PERM_GROUPUSED, 244 ZFS_DELEG_PERM_GROUPQUOTA, 245}; 246 247static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc); 248static int zfs_check_settable(const char *name, nvpair_t *property, 249 cred_t *cr); 250static int zfs_check_clearable(char *dataset, nvlist_t *props, 251 nvlist_t **errors); 252static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *, 253 boolean_t *); 254int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *); 255static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp); 256 257static void zfsdev_close(void *data); 258 259static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature); 260 261/* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */ 262void 263__dprintf(const char *file, const char *func, int line, const char *fmt, ...) 264{ 265 const char *newfile; 266 char buf[512]; 267 va_list adx; 268 269 /* 270 * Get rid of annoying "../common/" prefix to filename. 271 */ 272 newfile = strrchr(file, '/'); 273 if (newfile != NULL) { 274 newfile = newfile + 1; /* Get rid of leading / */ 275 } else { 276 newfile = file; 277 } 278 279 va_start(adx, fmt); 280 (void) vsnprintf(buf, sizeof (buf), fmt, adx); 281 va_end(adx); 282 283 /* 284 * To get this data, use the zfs-dprintf probe as so: 285 * dtrace -q -n 'zfs-dprintf \ 286 * /stringof(arg0) == "dbuf.c"/ \ 287 * {printf("%s: %s", stringof(arg1), stringof(arg3))}' 288 * arg0 = file name 289 * arg1 = function name 290 * arg2 = line number 291 * arg3 = message 292 */ 293 DTRACE_PROBE4(zfs__dprintf, 294 char *, newfile, char *, func, int, line, char *, buf); 295} 296 297static void 298history_str_free(char *buf) 299{ 300 kmem_free(buf, HIS_MAX_RECORD_LEN); 301} 302 303static char * 304history_str_get(zfs_cmd_t *zc) 305{ 306 char *buf; 307 308 if (zc->zc_history == 0) 309 return (NULL); 310 311 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP); 312 if (copyinstr((void *)(uintptr_t)zc->zc_history, 313 buf, HIS_MAX_RECORD_LEN, NULL) != 0) { 314 history_str_free(buf); 315 return (NULL); 316 } 317 318 buf[HIS_MAX_RECORD_LEN -1] = '\0'; 319 320 return (buf); 321} 322 323/* 324 * Check to see if the named dataset is currently defined as bootable 325 */ 326static boolean_t 327zfs_is_bootfs(const char *name) 328{ 329 objset_t *os; 330 331 if (dmu_objset_hold(name, FTAG, &os) == 0) { 332 boolean_t ret; 333 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os))); 334 dmu_objset_rele(os, FTAG); 335 return (ret); 336 } 337 return (B_FALSE); 338} 339 340/* 341 * Return non-zero if the spa version is less than requested version. 342 */ 343static int 344zfs_earlier_version(const char *name, int version) 345{ 346 spa_t *spa; 347 348 if (spa_open(name, &spa, FTAG) == 0) { 349 if (spa_version(spa) < version) { 350 spa_close(spa, FTAG); 351 return (1); 352 } 353 spa_close(spa, FTAG); 354 } 355 return (0); 356} 357 358/* 359 * Return TRUE if the ZPL version is less than requested version. 360 */ 361static boolean_t 362zpl_earlier_version(const char *name, int version) 363{ 364 objset_t *os; 365 boolean_t rc = B_TRUE; 366 367 if (dmu_objset_hold(name, FTAG, &os) == 0) { 368 uint64_t zplversion; 369 370 if (dmu_objset_type(os) != DMU_OST_ZFS) { 371 dmu_objset_rele(os, FTAG); 372 return (B_TRUE); 373 } 374 /* XXX reading from non-owned objset */ 375 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0) 376 rc = zplversion < version; 377 dmu_objset_rele(os, FTAG); 378 } 379 return (rc); 380} 381 382static void 383zfs_log_history(zfs_cmd_t *zc) 384{ 385 spa_t *spa; 386 char *buf; 387 388 if ((buf = history_str_get(zc)) == NULL) 389 return; 390 391 if (spa_open(zc->zc_name, &spa, FTAG) == 0) { 392 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY) 393 (void) spa_history_log(spa, buf); 394 spa_close(spa, FTAG); 395 } 396 history_str_free(buf); 397} 398 399/* 400 * Policy for top-level read operations (list pools). Requires no privileges, 401 * and can be used in the local zone, as there is no associated dataset. 402 */ 403/* ARGSUSED */ 404static int 405zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 406{ 407 return (0); 408} 409 410/* 411 * Policy for dataset read operations (list children, get statistics). Requires 412 * no privileges, but must be visible in the local zone. 413 */ 414/* ARGSUSED */ 415static int 416zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 417{ 418 if (INGLOBALZONE(curthread) || 419 zone_dataset_visible(zc->zc_name, NULL)) 420 return (0); 421 422 return (SET_ERROR(ENOENT)); 423} 424 425static int 426zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr) 427{ 428 int writable = 1; 429 430 /* 431 * The dataset must be visible by this zone -- check this first 432 * so they don't see EPERM on something they shouldn't know about. 433 */ 434 if (!INGLOBALZONE(curthread) && 435 !zone_dataset_visible(dataset, &writable)) 436 return (SET_ERROR(ENOENT)); 437 438 if (INGLOBALZONE(curthread)) { 439 /* 440 * If the fs is zoned, only root can access it from the 441 * global zone. 442 */ 443 if (secpolicy_zfs(cr) && zoned) 444 return (SET_ERROR(EPERM)); 445 } else { 446 /* 447 * If we are in a local zone, the 'zoned' property must be set. 448 */ 449 if (!zoned) 450 return (SET_ERROR(EPERM)); 451 452 /* must be writable by this zone */ 453 if (!writable) 454 return (SET_ERROR(EPERM)); 455 } 456 return (0); 457} 458 459static int 460zfs_dozonecheck(const char *dataset, cred_t *cr) 461{ 462 uint64_t zoned; 463 464 if (dsl_prop_get_integer(dataset, "jailed", &zoned, NULL)) 465 return (SET_ERROR(ENOENT)); 466 467 return (zfs_dozonecheck_impl(dataset, zoned, cr)); 468} 469 470static int 471zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr) 472{ 473 uint64_t zoned; 474 475 if (dsl_prop_get_int_ds(ds, "jailed", &zoned)) 476 return (SET_ERROR(ENOENT)); 477 478 return (zfs_dozonecheck_impl(dataset, zoned, cr)); 479} 480 481static int 482zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds, 483 const char *perm, cred_t *cr) 484{ 485 int error; 486 487 error = zfs_dozonecheck_ds(name, ds, cr); 488 if (error == 0) { 489 error = secpolicy_zfs(cr); 490 if (error != 0) 491 error = dsl_deleg_access_impl(ds, perm, cr); 492 } 493 return (error); 494} 495 496static int 497zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr) 498{ 499 int error; 500 dsl_dataset_t *ds; 501 dsl_pool_t *dp; 502 503 error = dsl_pool_hold(name, FTAG, &dp); 504 if (error != 0) 505 return (error); 506 507 error = dsl_dataset_hold(dp, name, FTAG, &ds); 508 if (error != 0) { 509 dsl_pool_rele(dp, FTAG); 510 return (error); 511 } 512 513 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr); 514 515 dsl_dataset_rele(ds, FTAG); 516 dsl_pool_rele(dp, FTAG); 517 return (error); 518} 519 520#ifdef SECLABEL 521/* 522 * Policy for setting the security label property. 523 * 524 * Returns 0 for success, non-zero for access and other errors. 525 */ 526static int 527zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr) 528{ 529 char ds_hexsl[MAXNAMELEN]; 530 bslabel_t ds_sl, new_sl; 531 boolean_t new_default = FALSE; 532 uint64_t zoned; 533 int needed_priv = -1; 534 int error; 535 536 /* First get the existing dataset label. */ 537 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL), 538 1, sizeof (ds_hexsl), &ds_hexsl, NULL); 539 if (error != 0) 540 return (SET_ERROR(EPERM)); 541 542 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0) 543 new_default = TRUE; 544 545 /* The label must be translatable */ 546 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0)) 547 return (SET_ERROR(EINVAL)); 548 549 /* 550 * In a non-global zone, disallow attempts to set a label that 551 * doesn't match that of the zone; otherwise no other checks 552 * are needed. 553 */ 554 if (!INGLOBALZONE(curproc)) { 555 if (new_default || !blequal(&new_sl, CR_SL(CRED()))) 556 return (SET_ERROR(EPERM)); 557 return (0); 558 } 559 560 /* 561 * For global-zone datasets (i.e., those whose zoned property is 562 * "off", verify that the specified new label is valid for the 563 * global zone. 564 */ 565 if (dsl_prop_get_integer(name, 566 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL)) 567 return (SET_ERROR(EPERM)); 568 if (!zoned) { 569 if (zfs_check_global_label(name, strval) != 0) 570 return (SET_ERROR(EPERM)); 571 } 572 573 /* 574 * If the existing dataset label is nondefault, check if the 575 * dataset is mounted (label cannot be changed while mounted). 576 * Get the zfsvfs; if there isn't one, then the dataset isn't 577 * mounted (or isn't a dataset, doesn't exist, ...). 578 */ 579 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) { 580 objset_t *os; 581 static char *setsl_tag = "setsl_tag"; 582 583 /* 584 * Try to own the dataset; abort if there is any error, 585 * (e.g., already mounted, in use, or other error). 586 */ 587 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE, 588 setsl_tag, &os); 589 if (error != 0) 590 return (SET_ERROR(EPERM)); 591 592 dmu_objset_disown(os, setsl_tag); 593 594 if (new_default) { 595 needed_priv = PRIV_FILE_DOWNGRADE_SL; 596 goto out_check; 597 } 598 599 if (hexstr_to_label(strval, &new_sl) != 0) 600 return (SET_ERROR(EPERM)); 601 602 if (blstrictdom(&ds_sl, &new_sl)) 603 needed_priv = PRIV_FILE_DOWNGRADE_SL; 604 else if (blstrictdom(&new_sl, &ds_sl)) 605 needed_priv = PRIV_FILE_UPGRADE_SL; 606 } else { 607 /* dataset currently has a default label */ 608 if (!new_default) 609 needed_priv = PRIV_FILE_UPGRADE_SL; 610 } 611 612out_check: 613 if (needed_priv != -1) 614 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL)); 615 return (0); 616} 617#endif /* SECLABEL */ 618 619static int 620zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval, 621 cred_t *cr) 622{ 623 char *strval; 624 625 /* 626 * Check permissions for special properties. 627 */ 628 switch (prop) { 629 case ZFS_PROP_ZONED: 630 /* 631 * Disallow setting of 'zoned' from within a local zone. 632 */ 633 if (!INGLOBALZONE(curthread)) 634 return (SET_ERROR(EPERM)); 635 break; 636 637 case ZFS_PROP_QUOTA: 638 case ZFS_PROP_FILESYSTEM_LIMIT: 639 case ZFS_PROP_SNAPSHOT_LIMIT: 640 if (!INGLOBALZONE(curthread)) { 641 uint64_t zoned; 642 char setpoint[MAXNAMELEN]; 643 /* 644 * Unprivileged users are allowed to modify the 645 * limit on things *under* (ie. contained by) 646 * the thing they own. 647 */ 648 if (dsl_prop_get_integer(dsname, "jailed", &zoned, 649 setpoint)) 650 return (SET_ERROR(EPERM)); 651 if (!zoned || strlen(dsname) <= strlen(setpoint)) 652 return (SET_ERROR(EPERM)); 653 } 654 break; 655 656 case ZFS_PROP_MLSLABEL: 657#ifdef SECLABEL 658 if (!is_system_labeled()) 659 return (SET_ERROR(EPERM)); 660 661 if (nvpair_value_string(propval, &strval) == 0) { 662 int err; 663 664 err = zfs_set_slabel_policy(dsname, strval, CRED()); 665 if (err != 0) 666 return (err); 667 } 668#else 669 return (EOPNOTSUPP); 670#endif 671 break; 672 } 673 674 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr)); 675} 676 677/* ARGSUSED */ 678static int 679zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 680{ 681 int error; 682 683 error = zfs_dozonecheck(zc->zc_name, cr); 684 if (error != 0) 685 return (error); 686 687 /* 688 * permission to set permissions will be evaluated later in 689 * dsl_deleg_can_allow() 690 */ 691 return (0); 692} 693 694/* ARGSUSED */ 695static int 696zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 697{ 698 return (zfs_secpolicy_write_perms(zc->zc_name, 699 ZFS_DELEG_PERM_ROLLBACK, cr)); 700} 701 702/* ARGSUSED */ 703static int 704zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 705{ 706 dsl_pool_t *dp; 707 dsl_dataset_t *ds; 708 char *cp; 709 int error; 710 711 /* 712 * Generate the current snapshot name from the given objsetid, then 713 * use that name for the secpolicy/zone checks. 714 */ 715 cp = strchr(zc->zc_name, '@'); 716 if (cp == NULL) 717 return (SET_ERROR(EINVAL)); 718 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 719 if (error != 0) 720 return (error); 721 722 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds); 723 if (error != 0) { 724 dsl_pool_rele(dp, FTAG); 725 return (error); 726 } 727 728 dsl_dataset_name(ds, zc->zc_name); 729 730 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds, 731 ZFS_DELEG_PERM_SEND, cr); 732 dsl_dataset_rele(ds, FTAG); 733 dsl_pool_rele(dp, FTAG); 734 735 return (error); 736} 737 738/* ARGSUSED */ 739static int 740zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 741{ 742 return (zfs_secpolicy_write_perms(zc->zc_name, 743 ZFS_DELEG_PERM_SEND, cr)); 744} 745 746/* ARGSUSED */ 747static int 748zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 749{ 750 vnode_t *vp; 751 int error; 752 753 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, 754 NO_FOLLOW, NULL, &vp)) != 0) 755 return (error); 756 757 /* Now make sure mntpnt and dataset are ZFS */ 758 759 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 || 760 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), 761 zc->zc_name) != 0)) { 762 VN_RELE(vp); 763 return (SET_ERROR(EPERM)); 764 } 765 766 VN_RELE(vp); 767 return (dsl_deleg_access(zc->zc_name, 768 ZFS_DELEG_PERM_SHARE, cr)); 769} 770 771int 772zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 773{ 774 if (!INGLOBALZONE(curthread)) 775 return (SET_ERROR(EPERM)); 776 777 if (secpolicy_nfs(cr) == 0) { 778 return (0); 779 } else { 780 return (zfs_secpolicy_deleg_share(zc, innvl, cr)); 781 } 782} 783 784int 785zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 786{ 787 if (!INGLOBALZONE(curthread)) 788 return (SET_ERROR(EPERM)); 789 790 if (secpolicy_smb(cr) == 0) { 791 return (0); 792 } else { 793 return (zfs_secpolicy_deleg_share(zc, innvl, cr)); 794 } 795} 796 797static int 798zfs_get_parent(const char *datasetname, char *parent, int parentsize) 799{ 800 char *cp; 801 802 /* 803 * Remove the @bla or /bla from the end of the name to get the parent. 804 */ 805 (void) strncpy(parent, datasetname, parentsize); 806 cp = strrchr(parent, '@'); 807 if (cp != NULL) { 808 cp[0] = '\0'; 809 } else { 810 cp = strrchr(parent, '/'); 811 if (cp == NULL) 812 return (SET_ERROR(ENOENT)); 813 cp[0] = '\0'; 814 } 815 816 return (0); 817} 818 819int 820zfs_secpolicy_destroy_perms(const char *name, cred_t *cr) 821{ 822 int error; 823 824 if ((error = zfs_secpolicy_write_perms(name, 825 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 826 return (error); 827 828 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr)); 829} 830 831/* ARGSUSED */ 832static int 833zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 834{ 835 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr)); 836} 837 838/* 839 * Destroying snapshots with delegated permissions requires 840 * descendant mount and destroy permissions. 841 */ 842/* ARGSUSED */ 843static int 844zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 845{ 846 nvlist_t *snaps; 847 nvpair_t *pair, *nextpair; 848 int error = 0; 849 850 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 851 return (SET_ERROR(EINVAL)); 852 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 853 pair = nextpair) { 854 nextpair = nvlist_next_nvpair(snaps, pair); 855 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr); 856 if (error == ENOENT) { 857 /* 858 * Ignore any snapshots that don't exist (we consider 859 * them "already destroyed"). Remove the name from the 860 * nvl here in case the snapshot is created between 861 * now and when we try to destroy it (in which case 862 * we don't want to destroy it since we haven't 863 * checked for permission). 864 */ 865 fnvlist_remove_nvpair(snaps, pair); 866 error = 0; 867 } 868 if (error != 0) 869 break; 870 } 871 872 return (error); 873} 874 875int 876zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr) 877{ 878 char parentname[MAXNAMELEN]; 879 int error; 880 881 if ((error = zfs_secpolicy_write_perms(from, 882 ZFS_DELEG_PERM_RENAME, cr)) != 0) 883 return (error); 884 885 if ((error = zfs_secpolicy_write_perms(from, 886 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 887 return (error); 888 889 if ((error = zfs_get_parent(to, parentname, 890 sizeof (parentname))) != 0) 891 return (error); 892 893 if ((error = zfs_secpolicy_write_perms(parentname, 894 ZFS_DELEG_PERM_CREATE, cr)) != 0) 895 return (error); 896 897 if ((error = zfs_secpolicy_write_perms(parentname, 898 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 899 return (error); 900 901 return (error); 902} 903 904/* ARGSUSED */ 905static int 906zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 907{ 908 char *at = NULL; 909 int error; 910 911 if ((zc->zc_cookie & 1) != 0) { 912 /* 913 * This is recursive rename, so the starting snapshot might 914 * not exist. Check file system or volume permission instead. 915 */ 916 at = strchr(zc->zc_name, '@'); 917 if (at == NULL) 918 return (EINVAL); 919 *at = '\0'; 920 } 921 922 error = zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr); 923 924 if (at != NULL) 925 *at = '@'; 926 927 return (error); 928} 929 930/* ARGSUSED */ 931static int 932zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 933{ 934 dsl_pool_t *dp; 935 dsl_dataset_t *clone; 936 int error; 937 938 error = zfs_secpolicy_write_perms(zc->zc_name, 939 ZFS_DELEG_PERM_PROMOTE, cr); 940 if (error != 0) 941 return (error); 942 943 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 944 if (error != 0) 945 return (error); 946 947 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone); 948 949 if (error == 0) { 950 char parentname[MAXNAMELEN]; 951 dsl_dataset_t *origin = NULL; 952 dsl_dir_t *dd; 953 dd = clone->ds_dir; 954 955 error = dsl_dataset_hold_obj(dd->dd_pool, 956 dd->dd_phys->dd_origin_obj, FTAG, &origin); 957 if (error != 0) { 958 dsl_dataset_rele(clone, FTAG); 959 dsl_pool_rele(dp, FTAG); 960 return (error); 961 } 962 963 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone, 964 ZFS_DELEG_PERM_MOUNT, cr); 965 966 dsl_dataset_name(origin, parentname); 967 if (error == 0) { 968 error = zfs_secpolicy_write_perms_ds(parentname, origin, 969 ZFS_DELEG_PERM_PROMOTE, cr); 970 } 971 dsl_dataset_rele(clone, FTAG); 972 dsl_dataset_rele(origin, FTAG); 973 } 974 dsl_pool_rele(dp, FTAG); 975 return (error); 976} 977 978/* ARGSUSED */ 979static int 980zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 981{ 982 int error; 983 984 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 985 ZFS_DELEG_PERM_RECEIVE, cr)) != 0) 986 return (error); 987 988 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 989 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 990 return (error); 991 992 return (zfs_secpolicy_write_perms(zc->zc_name, 993 ZFS_DELEG_PERM_CREATE, cr)); 994} 995 996int 997zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr) 998{ 999 return (zfs_secpolicy_write_perms(name, 1000 ZFS_DELEG_PERM_SNAPSHOT, cr)); 1001} 1002 1003/* 1004 * Check for permission to create each snapshot in the nvlist. 1005 */ 1006/* ARGSUSED */ 1007static int 1008zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1009{ 1010 nvlist_t *snaps; 1011 int error; 1012 nvpair_t *pair; 1013 1014 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 1015 return (SET_ERROR(EINVAL)); 1016 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 1017 pair = nvlist_next_nvpair(snaps, pair)) { 1018 char *name = nvpair_name(pair); 1019 char *atp = strchr(name, '@'); 1020 1021 if (atp == NULL) { 1022 error = SET_ERROR(EINVAL); 1023 break; 1024 } 1025 *atp = '\0'; 1026 error = zfs_secpolicy_snapshot_perms(name, cr); 1027 *atp = '@'; 1028 if (error != 0) 1029 break; 1030 } 1031 return (error); 1032} 1033 1034/* 1035 * Check for permission to create each snapshot in the nvlist. 1036 */ 1037/* ARGSUSED */ 1038static int 1039zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1040{ 1041 int error = 0; 1042 1043 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 1044 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 1045 char *name = nvpair_name(pair); 1046 char *hashp = strchr(name, '#'); 1047 1048 if (hashp == NULL) { 1049 error = SET_ERROR(EINVAL); 1050 break; 1051 } 1052 *hashp = '\0'; 1053 error = zfs_secpolicy_write_perms(name, 1054 ZFS_DELEG_PERM_BOOKMARK, cr); 1055 *hashp = '#'; 1056 if (error != 0) 1057 break; 1058 } 1059 return (error); 1060} 1061 1062/* ARGSUSED */ 1063static int 1064zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1065{ 1066 nvpair_t *pair, *nextpair; 1067 int error = 0; 1068 1069 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL; 1070 pair = nextpair) { 1071 char *name = nvpair_name(pair); 1072 char *hashp = strchr(name, '#'); 1073 nextpair = nvlist_next_nvpair(innvl, pair); 1074 1075 if (hashp == NULL) { 1076 error = SET_ERROR(EINVAL); 1077 break; 1078 } 1079 1080 *hashp = '\0'; 1081 error = zfs_secpolicy_write_perms(name, 1082 ZFS_DELEG_PERM_DESTROY, cr); 1083 *hashp = '#'; 1084 if (error == ENOENT) { 1085 /* 1086 * Ignore any filesystems that don't exist (we consider 1087 * their bookmarks "already destroyed"). Remove 1088 * the name from the nvl here in case the filesystem 1089 * is created between now and when we try to destroy 1090 * the bookmark (in which case we don't want to 1091 * destroy it since we haven't checked for permission). 1092 */ 1093 fnvlist_remove_nvpair(innvl, pair); 1094 error = 0; 1095 } 1096 if (error != 0) 1097 break; 1098 } 1099 1100 return (error); 1101} 1102 1103/* ARGSUSED */ 1104static int 1105zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1106{ 1107 /* 1108 * Even root must have a proper TSD so that we know what pool 1109 * to log to. 1110 */ 1111 if (tsd_get(zfs_allow_log_key) == NULL) 1112 return (SET_ERROR(EPERM)); 1113 return (0); 1114} 1115 1116static int 1117zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1118{ 1119 char parentname[MAXNAMELEN]; 1120 int error; 1121 char *origin; 1122 1123 if ((error = zfs_get_parent(zc->zc_name, parentname, 1124 sizeof (parentname))) != 0) 1125 return (error); 1126 1127 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 && 1128 (error = zfs_secpolicy_write_perms(origin, 1129 ZFS_DELEG_PERM_CLONE, cr)) != 0) 1130 return (error); 1131 1132 if ((error = zfs_secpolicy_write_perms(parentname, 1133 ZFS_DELEG_PERM_CREATE, cr)) != 0) 1134 return (error); 1135 1136 return (zfs_secpolicy_write_perms(parentname, 1137 ZFS_DELEG_PERM_MOUNT, cr)); 1138} 1139 1140/* 1141 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires 1142 * SYS_CONFIG privilege, which is not available in a local zone. 1143 */ 1144/* ARGSUSED */ 1145static int 1146zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1147{ 1148 if (secpolicy_sys_config(cr, B_FALSE) != 0) 1149 return (SET_ERROR(EPERM)); 1150 1151 return (0); 1152} 1153 1154/* 1155 * Policy for object to name lookups. 1156 */ 1157/* ARGSUSED */ 1158static int 1159zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1160{ 1161 int error; 1162 1163 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0) 1164 return (0); 1165 1166 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr); 1167 return (error); 1168} 1169 1170/* 1171 * Policy for fault injection. Requires all privileges. 1172 */ 1173/* ARGSUSED */ 1174static int 1175zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1176{ 1177 return (secpolicy_zinject(cr)); 1178} 1179 1180/* ARGSUSED */ 1181static int 1182zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1183{ 1184 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value); 1185 1186 if (prop == ZPROP_INVAL) { 1187 if (!zfs_prop_user(zc->zc_value)) 1188 return (SET_ERROR(EINVAL)); 1189 return (zfs_secpolicy_write_perms(zc->zc_name, 1190 ZFS_DELEG_PERM_USERPROP, cr)); 1191 } else { 1192 return (zfs_secpolicy_setprop(zc->zc_name, prop, 1193 NULL, cr)); 1194 } 1195} 1196 1197static int 1198zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1199{ 1200 int err = zfs_secpolicy_read(zc, innvl, cr); 1201 if (err) 1202 return (err); 1203 1204 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 1205 return (SET_ERROR(EINVAL)); 1206 1207 if (zc->zc_value[0] == 0) { 1208 /* 1209 * They are asking about a posix uid/gid. If it's 1210 * themself, allow it. 1211 */ 1212 if (zc->zc_objset_type == ZFS_PROP_USERUSED || 1213 zc->zc_objset_type == ZFS_PROP_USERQUOTA) { 1214 if (zc->zc_guid == crgetuid(cr)) 1215 return (0); 1216 } else { 1217 if (groupmember(zc->zc_guid, cr)) 1218 return (0); 1219 } 1220 } 1221 1222 return (zfs_secpolicy_write_perms(zc->zc_name, 1223 userquota_perms[zc->zc_objset_type], cr)); 1224} 1225 1226static int 1227zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1228{ 1229 int err = zfs_secpolicy_read(zc, innvl, cr); 1230 if (err) 1231 return (err); 1232 1233 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 1234 return (SET_ERROR(EINVAL)); 1235 1236 return (zfs_secpolicy_write_perms(zc->zc_name, 1237 userquota_perms[zc->zc_objset_type], cr)); 1238} 1239 1240/* ARGSUSED */ 1241static int 1242zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1243{ 1244 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION, 1245 NULL, cr)); 1246} 1247 1248/* ARGSUSED */ 1249static int 1250zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1251{ 1252 nvpair_t *pair; 1253 nvlist_t *holds; 1254 int error; 1255 1256 error = nvlist_lookup_nvlist(innvl, "holds", &holds); 1257 if (error != 0) 1258 return (SET_ERROR(EINVAL)); 1259 1260 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL; 1261 pair = nvlist_next_nvpair(holds, pair)) { 1262 char fsname[MAXNAMELEN]; 1263 error = dmu_fsname(nvpair_name(pair), fsname); 1264 if (error != 0) 1265 return (error); 1266 error = zfs_secpolicy_write_perms(fsname, 1267 ZFS_DELEG_PERM_HOLD, cr); 1268 if (error != 0) 1269 return (error); 1270 } 1271 return (0); 1272} 1273 1274/* ARGSUSED */ 1275static int 1276zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1277{ 1278 nvpair_t *pair; 1279 int error; 1280 1281 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL; 1282 pair = nvlist_next_nvpair(innvl, pair)) { 1283 char fsname[MAXNAMELEN]; 1284 error = dmu_fsname(nvpair_name(pair), fsname); 1285 if (error != 0) 1286 return (error); 1287 error = zfs_secpolicy_write_perms(fsname, 1288 ZFS_DELEG_PERM_RELEASE, cr); 1289 if (error != 0) 1290 return (error); 1291 } 1292 return (0); 1293} 1294 1295/* 1296 * Policy for allowing temporary snapshots to be taken or released 1297 */ 1298static int 1299zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1300{ 1301 /* 1302 * A temporary snapshot is the same as a snapshot, 1303 * hold, destroy and release all rolled into one. 1304 * Delegated diff alone is sufficient that we allow this. 1305 */ 1306 int error; 1307 1308 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 1309 ZFS_DELEG_PERM_DIFF, cr)) == 0) 1310 return (0); 1311 1312 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr); 1313 if (error == 0) 1314 error = zfs_secpolicy_hold(zc, innvl, cr); 1315 if (error == 0) 1316 error = zfs_secpolicy_release(zc, innvl, cr); 1317 if (error == 0) 1318 error = zfs_secpolicy_destroy(zc, innvl, cr); 1319 return (error); 1320} 1321 1322/* 1323 * Returns the nvlist as specified by the user in the zfs_cmd_t. 1324 */ 1325static int 1326get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp) 1327{ 1328 char *packed; 1329 int error; 1330 nvlist_t *list = NULL; 1331 1332 /* 1333 * Read in and unpack the user-supplied nvlist. 1334 */ 1335 if (size == 0) 1336 return (SET_ERROR(EINVAL)); 1337 1338 packed = kmem_alloc(size, KM_SLEEP); 1339 1340 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size, 1341 iflag)) != 0) { 1342 kmem_free(packed, size); 1343 return (error); 1344 } 1345 1346 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) { 1347 kmem_free(packed, size); 1348 return (error); 1349 } 1350 1351 kmem_free(packed, size); 1352 1353 *nvp = list; 1354 return (0); 1355} 1356 1357/* 1358 * Reduce the size of this nvlist until it can be serialized in 'max' bytes. 1359 * Entries will be removed from the end of the nvlist, and one int32 entry 1360 * named "N_MORE_ERRORS" will be added indicating how many entries were 1361 * removed. 1362 */ 1363static int 1364nvlist_smush(nvlist_t *errors, size_t max) 1365{ 1366 size_t size; 1367 1368 size = fnvlist_size(errors); 1369 1370 if (size > max) { 1371 nvpair_t *more_errors; 1372 int n = 0; 1373 1374 if (max < 1024) 1375 return (SET_ERROR(ENOMEM)); 1376 1377 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0); 1378 more_errors = nvlist_prev_nvpair(errors, NULL); 1379 1380 do { 1381 nvpair_t *pair = nvlist_prev_nvpair(errors, 1382 more_errors); 1383 fnvlist_remove_nvpair(errors, pair); 1384 n++; 1385 size = fnvlist_size(errors); 1386 } while (size > max); 1387 1388 fnvlist_remove_nvpair(errors, more_errors); 1389 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n); 1390 ASSERT3U(fnvlist_size(errors), <=, max); 1391 } 1392 1393 return (0); 1394} 1395 1396static int 1397put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl) 1398{ 1399 char *packed = NULL; 1400 int error = 0; 1401 size_t size; 1402 1403 size = fnvlist_size(nvl); 1404 1405 if (size > zc->zc_nvlist_dst_size) { 1406 /* 1407 * Solaris returns ENOMEM here, because even if an error is 1408 * returned from an ioctl(2), new zc_nvlist_dst_size will be 1409 * passed to the userland. This is not the case for FreeBSD. 1410 * We need to return 0, so the kernel will copy the 1411 * zc_nvlist_dst_size back and the userland can discover that a 1412 * bigger buffer is needed. 1413 */ 1414 error = 0; 1415 } else { 1416 packed = fnvlist_pack(nvl, &size); 1417 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst, 1418 size, zc->zc_iflags) != 0) 1419 error = SET_ERROR(EFAULT); 1420 fnvlist_pack_free(packed, size); 1421 } 1422 1423 zc->zc_nvlist_dst_size = size; 1424 zc->zc_nvlist_dst_filled = B_TRUE; 1425 return (error); 1426} 1427 1428static int 1429getzfsvfs(const char *dsname, zfsvfs_t **zfvp) 1430{ 1431 objset_t *os; 1432 int error; 1433 1434 error = dmu_objset_hold(dsname, FTAG, &os); 1435 if (error != 0) 1436 return (error); 1437 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1438 dmu_objset_rele(os, FTAG); 1439 return (SET_ERROR(EINVAL)); 1440 } 1441 1442 mutex_enter(&os->os_user_ptr_lock); 1443 *zfvp = dmu_objset_get_user(os); 1444 if (*zfvp) { 1445 VFS_HOLD((*zfvp)->z_vfs); 1446 } else { 1447 error = SET_ERROR(ESRCH); 1448 } 1449 mutex_exit(&os->os_user_ptr_lock); 1450 dmu_objset_rele(os, FTAG); 1451 return (error); 1452} 1453 1454/* 1455 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which 1456 * case its z_vfs will be NULL, and it will be opened as the owner. 1457 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER, 1458 * which prevents all vnode ops from running. 1459 */ 1460static int 1461zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer) 1462{ 1463 int error = 0; 1464 1465 if (getzfsvfs(name, zfvp) != 0) 1466 error = zfsvfs_create(name, zfvp); 1467 if (error == 0) { 1468 rrw_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER : 1469 RW_READER, tag); 1470 if ((*zfvp)->z_unmounted) { 1471 /* 1472 * XXX we could probably try again, since the unmounting 1473 * thread should be just about to disassociate the 1474 * objset from the zfsvfs. 1475 */ 1476 rrw_exit(&(*zfvp)->z_teardown_lock, tag); 1477 return (SET_ERROR(EBUSY)); 1478 } 1479 } 1480 return (error); 1481} 1482 1483static void 1484zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag) 1485{ 1486 rrw_exit(&zfsvfs->z_teardown_lock, tag); 1487 1488 if (zfsvfs->z_vfs) { 1489 VFS_RELE(zfsvfs->z_vfs); 1490 } else { 1491 dmu_objset_disown(zfsvfs->z_os, zfsvfs); 1492 zfsvfs_free(zfsvfs); 1493 } 1494} 1495 1496static int 1497zfs_ioc_pool_create(zfs_cmd_t *zc) 1498{ 1499 int error; 1500 nvlist_t *config, *props = NULL; 1501 nvlist_t *rootprops = NULL; 1502 nvlist_t *zplprops = NULL; 1503 1504 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1505 zc->zc_iflags, &config)) 1506 return (error); 1507 1508 if (zc->zc_nvlist_src_size != 0 && (error = 1509 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1510 zc->zc_iflags, &props))) { 1511 nvlist_free(config); 1512 return (error); 1513 } 1514 1515 if (props) { 1516 nvlist_t *nvl = NULL; 1517 uint64_t version = SPA_VERSION; 1518 1519 (void) nvlist_lookup_uint64(props, 1520 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version); 1521 if (!SPA_VERSION_IS_SUPPORTED(version)) { 1522 error = SET_ERROR(EINVAL); 1523 goto pool_props_bad; 1524 } 1525 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl); 1526 if (nvl) { 1527 error = nvlist_dup(nvl, &rootprops, KM_SLEEP); 1528 if (error != 0) { 1529 nvlist_free(config); 1530 nvlist_free(props); 1531 return (error); 1532 } 1533 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS); 1534 } 1535 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 1536 error = zfs_fill_zplprops_root(version, rootprops, 1537 zplprops, NULL); 1538 if (error != 0) 1539 goto pool_props_bad; 1540 } 1541 1542 error = spa_create(zc->zc_name, config, props, zplprops); 1543 1544 /* 1545 * Set the remaining root properties 1546 */ 1547 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name, 1548 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0) 1549 (void) spa_destroy(zc->zc_name); 1550 1551pool_props_bad: 1552 nvlist_free(rootprops); 1553 nvlist_free(zplprops); 1554 nvlist_free(config); 1555 nvlist_free(props); 1556 1557 return (error); 1558} 1559 1560static int 1561zfs_ioc_pool_destroy(zfs_cmd_t *zc) 1562{ 1563 int error; 1564 zfs_log_history(zc); 1565 error = spa_destroy(zc->zc_name); 1566 if (error == 0) 1567 zvol_remove_minors(zc->zc_name); 1568 return (error); 1569} 1570 1571static int 1572zfs_ioc_pool_import(zfs_cmd_t *zc) 1573{ 1574 nvlist_t *config, *props = NULL; 1575 uint64_t guid; 1576 int error; 1577 1578 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1579 zc->zc_iflags, &config)) != 0) 1580 return (error); 1581 1582 if (zc->zc_nvlist_src_size != 0 && (error = 1583 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1584 zc->zc_iflags, &props))) { 1585 nvlist_free(config); 1586 return (error); 1587 } 1588 1589 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 || 1590 guid != zc->zc_guid) 1591 error = SET_ERROR(EINVAL); 1592 else 1593 error = spa_import(zc->zc_name, config, props, zc->zc_cookie); 1594 1595 if (zc->zc_nvlist_dst != 0) { 1596 int err; 1597 1598 if ((err = put_nvlist(zc, config)) != 0) 1599 error = err; 1600 } 1601 1602 nvlist_free(config); 1603 1604 if (props) 1605 nvlist_free(props); 1606 1607 return (error); 1608} 1609 1610static int 1611zfs_ioc_pool_export(zfs_cmd_t *zc) 1612{ 1613 int error; 1614 boolean_t force = (boolean_t)zc->zc_cookie; 1615 boolean_t hardforce = (boolean_t)zc->zc_guid; 1616 1617 zfs_log_history(zc); 1618 error = spa_export(zc->zc_name, NULL, force, hardforce); 1619 if (error == 0) 1620 zvol_remove_minors(zc->zc_name); 1621 return (error); 1622} 1623 1624static int 1625zfs_ioc_pool_configs(zfs_cmd_t *zc) 1626{ 1627 nvlist_t *configs; 1628 int error; 1629 1630 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL) 1631 return (SET_ERROR(EEXIST)); 1632 1633 error = put_nvlist(zc, configs); 1634 1635 nvlist_free(configs); 1636 1637 return (error); 1638} 1639 1640/* 1641 * inputs: 1642 * zc_name name of the pool 1643 * 1644 * outputs: 1645 * zc_cookie real errno 1646 * zc_nvlist_dst config nvlist 1647 * zc_nvlist_dst_size size of config nvlist 1648 */ 1649static int 1650zfs_ioc_pool_stats(zfs_cmd_t *zc) 1651{ 1652 nvlist_t *config; 1653 int error; 1654 int ret = 0; 1655 1656 error = spa_get_stats(zc->zc_name, &config, zc->zc_value, 1657 sizeof (zc->zc_value)); 1658 1659 if (config != NULL) { 1660 ret = put_nvlist(zc, config); 1661 nvlist_free(config); 1662 1663 /* 1664 * The config may be present even if 'error' is non-zero. 1665 * In this case we return success, and preserve the real errno 1666 * in 'zc_cookie'. 1667 */ 1668 zc->zc_cookie = error; 1669 } else { 1670 ret = error; 1671 } 1672 1673 return (ret); 1674} 1675 1676/* 1677 * Try to import the given pool, returning pool stats as appropriate so that 1678 * user land knows which devices are available and overall pool health. 1679 */ 1680static int 1681zfs_ioc_pool_tryimport(zfs_cmd_t *zc) 1682{ 1683 nvlist_t *tryconfig, *config; 1684 int error; 1685 1686 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1687 zc->zc_iflags, &tryconfig)) != 0) 1688 return (error); 1689 1690 config = spa_tryimport(tryconfig); 1691 1692 nvlist_free(tryconfig); 1693 1694 if (config == NULL) 1695 return (SET_ERROR(EINVAL)); 1696 1697 error = put_nvlist(zc, config); 1698 nvlist_free(config); 1699 1700 return (error); 1701} 1702 1703/* 1704 * inputs: 1705 * zc_name name of the pool 1706 * zc_cookie scan func (pool_scan_func_t) 1707 */ 1708static int 1709zfs_ioc_pool_scan(zfs_cmd_t *zc) 1710{ 1711 spa_t *spa; 1712 int error; 1713 1714 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1715 return (error); 1716 1717 if (zc->zc_cookie == POOL_SCAN_NONE) 1718 error = spa_scan_stop(spa); 1719 else 1720 error = spa_scan(spa, zc->zc_cookie); 1721 1722 spa_close(spa, FTAG); 1723 1724 return (error); 1725} 1726 1727static int 1728zfs_ioc_pool_freeze(zfs_cmd_t *zc) 1729{ 1730 spa_t *spa; 1731 int error; 1732 1733 error = spa_open(zc->zc_name, &spa, FTAG); 1734 if (error == 0) { 1735 spa_freeze(spa); 1736 spa_close(spa, FTAG); 1737 } 1738 return (error); 1739} 1740 1741static int 1742zfs_ioc_pool_upgrade(zfs_cmd_t *zc) 1743{ 1744 spa_t *spa; 1745 int error; 1746 1747 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1748 return (error); 1749 1750 if (zc->zc_cookie < spa_version(spa) || 1751 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) { 1752 spa_close(spa, FTAG); 1753 return (SET_ERROR(EINVAL)); 1754 } 1755 1756 spa_upgrade(spa, zc->zc_cookie); 1757 spa_close(spa, FTAG); 1758 1759 return (error); 1760} 1761 1762static int 1763zfs_ioc_pool_get_history(zfs_cmd_t *zc) 1764{ 1765 spa_t *spa; 1766 char *hist_buf; 1767 uint64_t size; 1768 int error; 1769 1770 if ((size = zc->zc_history_len) == 0) 1771 return (SET_ERROR(EINVAL)); 1772 1773 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1774 return (error); 1775 1776 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) { 1777 spa_close(spa, FTAG); 1778 return (SET_ERROR(ENOTSUP)); 1779 } 1780 1781 hist_buf = kmem_alloc(size, KM_SLEEP); 1782 if ((error = spa_history_get(spa, &zc->zc_history_offset, 1783 &zc->zc_history_len, hist_buf)) == 0) { 1784 error = ddi_copyout(hist_buf, 1785 (void *)(uintptr_t)zc->zc_history, 1786 zc->zc_history_len, zc->zc_iflags); 1787 } 1788 1789 spa_close(spa, FTAG); 1790 kmem_free(hist_buf, size); 1791 return (error); 1792} 1793 1794static int 1795zfs_ioc_pool_reguid(zfs_cmd_t *zc) 1796{ 1797 spa_t *spa; 1798 int error; 1799 1800 error = spa_open(zc->zc_name, &spa, FTAG); 1801 if (error == 0) { 1802 error = spa_change_guid(spa); 1803 spa_close(spa, FTAG); 1804 } 1805 return (error); 1806} 1807 1808static int 1809zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc) 1810{ 1811 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value)); 1812} 1813 1814/* 1815 * inputs: 1816 * zc_name name of filesystem 1817 * zc_obj object to find 1818 * 1819 * outputs: 1820 * zc_value name of object 1821 */ 1822static int 1823zfs_ioc_obj_to_path(zfs_cmd_t *zc) 1824{ 1825 objset_t *os; 1826 int error; 1827 1828 /* XXX reading from objset not owned */ 1829 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0) 1830 return (error); 1831 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1832 dmu_objset_rele(os, FTAG); 1833 return (SET_ERROR(EINVAL)); 1834 } 1835 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value, 1836 sizeof (zc->zc_value)); 1837 dmu_objset_rele(os, FTAG); 1838 1839 return (error); 1840} 1841 1842/* 1843 * inputs: 1844 * zc_name name of filesystem 1845 * zc_obj object to find 1846 * 1847 * outputs: 1848 * zc_stat stats on object 1849 * zc_value path to object 1850 */ 1851static int 1852zfs_ioc_obj_to_stats(zfs_cmd_t *zc) 1853{ 1854 objset_t *os; 1855 int error; 1856 1857 /* XXX reading from objset not owned */ 1858 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0) 1859 return (error); 1860 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1861 dmu_objset_rele(os, FTAG); 1862 return (SET_ERROR(EINVAL)); 1863 } 1864 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value, 1865 sizeof (zc->zc_value)); 1866 dmu_objset_rele(os, FTAG); 1867 1868 return (error); 1869} 1870 1871static int 1872zfs_ioc_vdev_add(zfs_cmd_t *zc) 1873{ 1874 spa_t *spa; 1875 int error; 1876 nvlist_t *config, **l2cache, **spares; 1877 uint_t nl2cache = 0, nspares = 0; 1878 1879 error = spa_open(zc->zc_name, &spa, FTAG); 1880 if (error != 0) 1881 return (error); 1882 1883 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1884 zc->zc_iflags, &config); 1885 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE, 1886 &l2cache, &nl2cache); 1887 1888 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES, 1889 &spares, &nspares); 1890 1891#ifdef illumos 1892 /* 1893 * A root pool with concatenated devices is not supported. 1894 * Thus, can not add a device to a root pool. 1895 * 1896 * Intent log device can not be added to a rootpool because 1897 * during mountroot, zil is replayed, a seperated log device 1898 * can not be accessed during the mountroot time. 1899 * 1900 * l2cache and spare devices are ok to be added to a rootpool. 1901 */ 1902 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) { 1903 nvlist_free(config); 1904 spa_close(spa, FTAG); 1905 return (SET_ERROR(EDOM)); 1906 } 1907#endif /* illumos */ 1908 1909 if (error == 0) { 1910 error = spa_vdev_add(spa, config); 1911 nvlist_free(config); 1912 } 1913 spa_close(spa, FTAG); 1914 return (error); 1915} 1916 1917/* 1918 * inputs: 1919 * zc_name name of the pool 1920 * zc_nvlist_conf nvlist of devices to remove 1921 * zc_cookie to stop the remove? 1922 */ 1923static int 1924zfs_ioc_vdev_remove(zfs_cmd_t *zc) 1925{ 1926 spa_t *spa; 1927 int error; 1928 1929 error = spa_open(zc->zc_name, &spa, FTAG); 1930 if (error != 0) 1931 return (error); 1932 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE); 1933 spa_close(spa, FTAG); 1934 return (error); 1935} 1936 1937static int 1938zfs_ioc_vdev_set_state(zfs_cmd_t *zc) 1939{ 1940 spa_t *spa; 1941 int error; 1942 vdev_state_t newstate = VDEV_STATE_UNKNOWN; 1943 1944 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1945 return (error); 1946 switch (zc->zc_cookie) { 1947 case VDEV_STATE_ONLINE: 1948 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate); 1949 break; 1950 1951 case VDEV_STATE_OFFLINE: 1952 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj); 1953 break; 1954 1955 case VDEV_STATE_FAULTED: 1956 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED && 1957 zc->zc_obj != VDEV_AUX_EXTERNAL) 1958 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED; 1959 1960 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj); 1961 break; 1962 1963 case VDEV_STATE_DEGRADED: 1964 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED && 1965 zc->zc_obj != VDEV_AUX_EXTERNAL) 1966 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED; 1967 1968 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj); 1969 break; 1970 1971 default: 1972 error = SET_ERROR(EINVAL); 1973 } 1974 zc->zc_cookie = newstate; 1975 spa_close(spa, FTAG); 1976 return (error); 1977} 1978 1979static int 1980zfs_ioc_vdev_attach(zfs_cmd_t *zc) 1981{ 1982 spa_t *spa; 1983 int replacing = zc->zc_cookie; 1984 nvlist_t *config; 1985 int error; 1986 1987 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1988 return (error); 1989 1990 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1991 zc->zc_iflags, &config)) == 0) { 1992 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing); 1993 nvlist_free(config); 1994 } 1995 1996 spa_close(spa, FTAG); 1997 return (error); 1998} 1999 2000static int 2001zfs_ioc_vdev_detach(zfs_cmd_t *zc) 2002{ 2003 spa_t *spa; 2004 int error; 2005 2006 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 2007 return (error); 2008 2009 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE); 2010 2011 spa_close(spa, FTAG); 2012 return (error); 2013} 2014 2015static int 2016zfs_ioc_vdev_split(zfs_cmd_t *zc) 2017{ 2018 spa_t *spa; 2019 nvlist_t *config, *props = NULL; 2020 int error; 2021 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT); 2022 2023 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 2024 return (error); 2025 2026 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 2027 zc->zc_iflags, &config)) { 2028 spa_close(spa, FTAG); 2029 return (error); 2030 } 2031 2032 if (zc->zc_nvlist_src_size != 0 && (error = 2033 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2034 zc->zc_iflags, &props))) { 2035 spa_close(spa, FTAG); 2036 nvlist_free(config); 2037 return (error); 2038 } 2039 2040 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp); 2041 2042 spa_close(spa, FTAG); 2043 2044 nvlist_free(config); 2045 nvlist_free(props); 2046 2047 return (error); 2048} 2049 2050static int 2051zfs_ioc_vdev_setpath(zfs_cmd_t *zc) 2052{ 2053 spa_t *spa; 2054 char *path = zc->zc_value; 2055 uint64_t guid = zc->zc_guid; 2056 int error; 2057 2058 error = spa_open(zc->zc_name, &spa, FTAG); 2059 if (error != 0) 2060 return (error); 2061 2062 error = spa_vdev_setpath(spa, guid, path); 2063 spa_close(spa, FTAG); 2064 return (error); 2065} 2066 2067static int 2068zfs_ioc_vdev_setfru(zfs_cmd_t *zc) 2069{ 2070 spa_t *spa; 2071 char *fru = zc->zc_value; 2072 uint64_t guid = zc->zc_guid; 2073 int error; 2074 2075 error = spa_open(zc->zc_name, &spa, FTAG); 2076 if (error != 0) 2077 return (error); 2078 2079 error = spa_vdev_setfru(spa, guid, fru); 2080 spa_close(spa, FTAG); 2081 return (error); 2082} 2083 2084static int 2085zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os) 2086{ 2087 int error = 0; 2088 nvlist_t *nv; 2089 2090 dmu_objset_fast_stat(os, &zc->zc_objset_stats); 2091 2092 if (zc->zc_nvlist_dst != 0 && 2093 (error = dsl_prop_get_all(os, &nv)) == 0) { 2094 dmu_objset_stats(os, nv); 2095 /* 2096 * NB: zvol_get_stats() will read the objset contents, 2097 * which we aren't supposed to do with a 2098 * DS_MODE_USER hold, because it could be 2099 * inconsistent. So this is a bit of a workaround... 2100 * XXX reading with out owning 2101 */ 2102 if (!zc->zc_objset_stats.dds_inconsistent && 2103 dmu_objset_type(os) == DMU_OST_ZVOL) { 2104 error = zvol_get_stats(os, nv); 2105 if (error == EIO) 2106 return (error); 2107 VERIFY0(error); 2108 } 2109 error = put_nvlist(zc, nv); 2110 nvlist_free(nv); 2111 } 2112 2113 return (error); 2114} 2115 2116/* 2117 * inputs: 2118 * zc_name name of filesystem 2119 * zc_nvlist_dst_size size of buffer for property nvlist 2120 * 2121 * outputs: 2122 * zc_objset_stats stats 2123 * zc_nvlist_dst property nvlist 2124 * zc_nvlist_dst_size size of property nvlist 2125 */ 2126static int 2127zfs_ioc_objset_stats(zfs_cmd_t *zc) 2128{ 2129 objset_t *os; 2130 int error; 2131 2132 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 2133 if (error == 0) { 2134 error = zfs_ioc_objset_stats_impl(zc, os); 2135 dmu_objset_rele(os, FTAG); 2136 } 2137 2138 if (error == ENOMEM) 2139 error = 0; 2140 return (error); 2141} 2142 2143/* 2144 * inputs: 2145 * zc_name name of filesystem 2146 * zc_nvlist_dst_size size of buffer for property nvlist 2147 * 2148 * outputs: 2149 * zc_nvlist_dst received property nvlist 2150 * zc_nvlist_dst_size size of received property nvlist 2151 * 2152 * Gets received properties (distinct from local properties on or after 2153 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from 2154 * local property values. 2155 */ 2156static int 2157zfs_ioc_objset_recvd_props(zfs_cmd_t *zc) 2158{ 2159 int error = 0; 2160 nvlist_t *nv; 2161 2162 /* 2163 * Without this check, we would return local property values if the 2164 * caller has not already received properties on or after 2165 * SPA_VERSION_RECVD_PROPS. 2166 */ 2167 if (!dsl_prop_get_hasrecvd(zc->zc_name)) 2168 return (SET_ERROR(ENOTSUP)); 2169 2170 if (zc->zc_nvlist_dst != 0 && 2171 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) { 2172 error = put_nvlist(zc, nv); 2173 nvlist_free(nv); 2174 } 2175 2176 return (error); 2177} 2178 2179static int 2180nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop) 2181{ 2182 uint64_t value; 2183 int error; 2184 2185 /* 2186 * zfs_get_zplprop() will either find a value or give us 2187 * the default value (if there is one). 2188 */ 2189 if ((error = zfs_get_zplprop(os, prop, &value)) != 0) 2190 return (error); 2191 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0); 2192 return (0); 2193} 2194 2195/* 2196 * inputs: 2197 * zc_name name of filesystem 2198 * zc_nvlist_dst_size size of buffer for zpl property nvlist 2199 * 2200 * outputs: 2201 * zc_nvlist_dst zpl property nvlist 2202 * zc_nvlist_dst_size size of zpl property nvlist 2203 */ 2204static int 2205zfs_ioc_objset_zplprops(zfs_cmd_t *zc) 2206{ 2207 objset_t *os; 2208 int err; 2209 2210 /* XXX reading without owning */ 2211 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os)) 2212 return (err); 2213 2214 dmu_objset_fast_stat(os, &zc->zc_objset_stats); 2215 2216 /* 2217 * NB: nvl_add_zplprop() will read the objset contents, 2218 * which we aren't supposed to do with a DS_MODE_USER 2219 * hold, because it could be inconsistent. 2220 */ 2221 if (zc->zc_nvlist_dst != 0 && 2222 !zc->zc_objset_stats.dds_inconsistent && 2223 dmu_objset_type(os) == DMU_OST_ZFS) { 2224 nvlist_t *nv; 2225 2226 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2227 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 && 2228 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 && 2229 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 && 2230 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0) 2231 err = put_nvlist(zc, nv); 2232 nvlist_free(nv); 2233 } else { 2234 err = SET_ERROR(ENOENT); 2235 } 2236 dmu_objset_rele(os, FTAG); 2237 return (err); 2238} 2239 2240boolean_t 2241dataset_name_hidden(const char *name) 2242{ 2243 /* 2244 * Skip over datasets that are not visible in this zone, 2245 * internal datasets (which have a $ in their name), and 2246 * temporary datasets (which have a % in their name). 2247 */ 2248 if (strchr(name, '$') != NULL) 2249 return (B_TRUE); 2250 if (strchr(name, '%') != NULL) 2251 return (B_TRUE); 2252 if (!INGLOBALZONE(curthread) && !zone_dataset_visible(name, NULL)) 2253 return (B_TRUE); 2254 return (B_FALSE); 2255} 2256 2257/* 2258 * inputs: 2259 * zc_name name of filesystem 2260 * zc_cookie zap cursor 2261 * zc_nvlist_dst_size size of buffer for property nvlist 2262 * 2263 * outputs: 2264 * zc_name name of next filesystem 2265 * zc_cookie zap cursor 2266 * zc_objset_stats stats 2267 * zc_nvlist_dst property nvlist 2268 * zc_nvlist_dst_size size of property nvlist 2269 */ 2270static int 2271zfs_ioc_dataset_list_next(zfs_cmd_t *zc) 2272{ 2273 objset_t *os; 2274 int error; 2275 char *p; 2276 size_t orig_len = strlen(zc->zc_name); 2277 2278top: 2279 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) { 2280 if (error == ENOENT) 2281 error = SET_ERROR(ESRCH); 2282 return (error); 2283 } 2284 2285 p = strrchr(zc->zc_name, '/'); 2286 if (p == NULL || p[1] != '\0') 2287 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name)); 2288 p = zc->zc_name + strlen(zc->zc_name); 2289 2290 do { 2291 error = dmu_dir_list_next(os, 2292 sizeof (zc->zc_name) - (p - zc->zc_name), p, 2293 NULL, &zc->zc_cookie); 2294 if (error == ENOENT) 2295 error = SET_ERROR(ESRCH); 2296 } while (error == 0 && dataset_name_hidden(zc->zc_name)); 2297 dmu_objset_rele(os, FTAG); 2298 2299 /* 2300 * If it's an internal dataset (ie. with a '$' in its name), 2301 * don't try to get stats for it, otherwise we'll return ENOENT. 2302 */ 2303 if (error == 0 && strchr(zc->zc_name, '$') == NULL) { 2304 error = zfs_ioc_objset_stats(zc); /* fill in the stats */ 2305 if (error == ENOENT) { 2306 /* We lost a race with destroy, get the next one. */ 2307 zc->zc_name[orig_len] = '\0'; 2308 goto top; 2309 } 2310 } 2311 return (error); 2312} 2313 2314/* 2315 * inputs: 2316 * zc_name name of filesystem 2317 * zc_cookie zap cursor 2318 * zc_nvlist_dst_size size of buffer for property nvlist 2319 * zc_simple when set, only name is requested 2320 * 2321 * outputs: 2322 * zc_name name of next snapshot 2323 * zc_objset_stats stats 2324 * zc_nvlist_dst property nvlist 2325 * zc_nvlist_dst_size size of property nvlist 2326 */ 2327static int 2328zfs_ioc_snapshot_list_next(zfs_cmd_t *zc) 2329{ 2330 objset_t *os; 2331 int error; 2332 2333 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 2334 if (error != 0) { 2335 return (error == ENOENT ? ESRCH : error); 2336 } 2337 2338 /* 2339 * A dataset name of maximum length cannot have any snapshots, 2340 * so exit immediately. 2341 */ 2342 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) { 2343 dmu_objset_rele(os, FTAG); 2344 return (SET_ERROR(ESRCH)); 2345 } 2346 2347 error = dmu_snapshot_list_next(os, 2348 sizeof (zc->zc_name) - strlen(zc->zc_name), 2349 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie, 2350 NULL); 2351 2352 if (error == 0 && !zc->zc_simple) { 2353 dsl_dataset_t *ds; 2354 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool; 2355 2356 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds); 2357 if (error == 0) { 2358 objset_t *ossnap; 2359 2360 error = dmu_objset_from_ds(ds, &ossnap); 2361 if (error == 0) 2362 error = zfs_ioc_objset_stats_impl(zc, ossnap); 2363 dsl_dataset_rele(ds, FTAG); 2364 } 2365 } else if (error == ENOENT) { 2366 error = SET_ERROR(ESRCH); 2367 } 2368 2369 dmu_objset_rele(os, FTAG); 2370 /* if we failed, undo the @ that we tacked on to zc_name */ 2371 if (error != 0) 2372 *strchr(zc->zc_name, '@') = '\0'; 2373 return (error); 2374} 2375 2376static int 2377zfs_prop_set_userquota(const char *dsname, nvpair_t *pair) 2378{ 2379 const char *propname = nvpair_name(pair); 2380 uint64_t *valary; 2381 unsigned int vallen; 2382 const char *domain; 2383 char *dash; 2384 zfs_userquota_prop_t type; 2385 uint64_t rid; 2386 uint64_t quota; 2387 zfsvfs_t *zfsvfs; 2388 int err; 2389 2390 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2391 nvlist_t *attrs; 2392 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 2393 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2394 &pair) != 0) 2395 return (SET_ERROR(EINVAL)); 2396 } 2397 2398 /* 2399 * A correctly constructed propname is encoded as 2400 * userquota@<rid>-<domain>. 2401 */ 2402 if ((dash = strchr(propname, '-')) == NULL || 2403 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 || 2404 vallen != 3) 2405 return (SET_ERROR(EINVAL)); 2406 2407 domain = dash + 1; 2408 type = valary[0]; 2409 rid = valary[1]; 2410 quota = valary[2]; 2411 2412 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE); 2413 if (err == 0) { 2414 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota); 2415 zfsvfs_rele(zfsvfs, FTAG); 2416 } 2417 2418 return (err); 2419} 2420 2421/* 2422 * If the named property is one that has a special function to set its value, 2423 * return 0 on success and a positive error code on failure; otherwise if it is 2424 * not one of the special properties handled by this function, return -1. 2425 * 2426 * XXX: It would be better for callers of the property interface if we handled 2427 * these special cases in dsl_prop.c (in the dsl layer). 2428 */ 2429static int 2430zfs_prop_set_special(const char *dsname, zprop_source_t source, 2431 nvpair_t *pair) 2432{ 2433 const char *propname = nvpair_name(pair); 2434 zfs_prop_t prop = zfs_name_to_prop(propname); 2435 uint64_t intval; 2436 int err; 2437 2438 if (prop == ZPROP_INVAL) { 2439 if (zfs_prop_userquota(propname)) 2440 return (zfs_prop_set_userquota(dsname, pair)); 2441 return (-1); 2442 } 2443 2444 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2445 nvlist_t *attrs; 2446 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 2447 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2448 &pair) == 0); 2449 } 2450 2451 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING) 2452 return (-1); 2453 2454 VERIFY(0 == nvpair_value_uint64(pair, &intval)); 2455 2456 switch (prop) { 2457 case ZFS_PROP_QUOTA: 2458 err = dsl_dir_set_quota(dsname, source, intval); 2459 break; 2460 case ZFS_PROP_REFQUOTA: 2461 err = dsl_dataset_set_refquota(dsname, source, intval); 2462 break; 2463 case ZFS_PROP_FILESYSTEM_LIMIT: 2464 case ZFS_PROP_SNAPSHOT_LIMIT: 2465 if (intval == UINT64_MAX) { 2466 /* clearing the limit, just do it */ 2467 err = 0; 2468 } else { 2469 err = dsl_dir_activate_fs_ss_limit(dsname); 2470 } 2471 /* 2472 * Set err to -1 to force the zfs_set_prop_nvlist code down the 2473 * default path to set the value in the nvlist. 2474 */ 2475 if (err == 0) 2476 err = -1; 2477 break; 2478 case ZFS_PROP_RESERVATION: 2479 err = dsl_dir_set_reservation(dsname, source, intval); 2480 break; 2481 case ZFS_PROP_REFRESERVATION: 2482 err = dsl_dataset_set_refreservation(dsname, source, intval); 2483 break; 2484 case ZFS_PROP_VOLSIZE: 2485 err = zvol_set_volsize(dsname, ddi_driver_major(zfs_dip), 2486 intval); 2487 break; 2488 case ZFS_PROP_VERSION: 2489 { 2490 zfsvfs_t *zfsvfs; 2491 2492 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0) 2493 break; 2494 2495 err = zfs_set_version(zfsvfs, intval); 2496 zfsvfs_rele(zfsvfs, FTAG); 2497 2498 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) { 2499 zfs_cmd_t *zc; 2500 2501 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP); 2502 (void) strcpy(zc->zc_name, dsname); 2503 (void) zfs_ioc_userspace_upgrade(zc); 2504 kmem_free(zc, sizeof (zfs_cmd_t)); 2505 } 2506 break; 2507 } 2508 default: 2509 err = -1; 2510 } 2511 2512 return (err); 2513} 2514 2515/* 2516 * This function is best effort. If it fails to set any of the given properties, 2517 * it continues to set as many as it can and returns the last error 2518 * encountered. If the caller provides a non-NULL errlist, it will be filled in 2519 * with the list of names of all the properties that failed along with the 2520 * corresponding error numbers. 2521 * 2522 * If every property is set successfully, zero is returned and errlist is not 2523 * modified. 2524 */ 2525int 2526zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl, 2527 nvlist_t *errlist) 2528{ 2529 nvpair_t *pair; 2530 nvpair_t *propval; 2531 int rv = 0; 2532 uint64_t intval; 2533 char *strval; 2534 nvlist_t *genericnvl = fnvlist_alloc(); 2535 nvlist_t *retrynvl = fnvlist_alloc(); 2536 2537retry: 2538 pair = NULL; 2539 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { 2540 const char *propname = nvpair_name(pair); 2541 zfs_prop_t prop = zfs_name_to_prop(propname); 2542 int err = 0; 2543 2544 /* decode the property value */ 2545 propval = pair; 2546 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2547 nvlist_t *attrs; 2548 attrs = fnvpair_value_nvlist(pair); 2549 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2550 &propval) != 0) 2551 err = SET_ERROR(EINVAL); 2552 } 2553 2554 /* Validate value type */ 2555 if (err == 0 && prop == ZPROP_INVAL) { 2556 if (zfs_prop_user(propname)) { 2557 if (nvpair_type(propval) != DATA_TYPE_STRING) 2558 err = SET_ERROR(EINVAL); 2559 } else if (zfs_prop_userquota(propname)) { 2560 if (nvpair_type(propval) != 2561 DATA_TYPE_UINT64_ARRAY) 2562 err = SET_ERROR(EINVAL); 2563 } else { 2564 err = SET_ERROR(EINVAL); 2565 } 2566 } else if (err == 0) { 2567 if (nvpair_type(propval) == DATA_TYPE_STRING) { 2568 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING) 2569 err = SET_ERROR(EINVAL); 2570 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) { 2571 const char *unused; 2572 2573 intval = fnvpair_value_uint64(propval); 2574 2575 switch (zfs_prop_get_type(prop)) { 2576 case PROP_TYPE_NUMBER: 2577 break; 2578 case PROP_TYPE_STRING: 2579 err = SET_ERROR(EINVAL); 2580 break; 2581 case PROP_TYPE_INDEX: 2582 if (zfs_prop_index_to_string(prop, 2583 intval, &unused) != 0) 2584 err = SET_ERROR(EINVAL); 2585 break; 2586 default: 2587 cmn_err(CE_PANIC, 2588 "unknown property type"); 2589 } 2590 } else { 2591 err = SET_ERROR(EINVAL); 2592 } 2593 } 2594 2595 /* Validate permissions */ 2596 if (err == 0) 2597 err = zfs_check_settable(dsname, pair, CRED()); 2598 2599 if (err == 0) { 2600 err = zfs_prop_set_special(dsname, source, pair); 2601 if (err == -1) { 2602 /* 2603 * For better performance we build up a list of 2604 * properties to set in a single transaction. 2605 */ 2606 err = nvlist_add_nvpair(genericnvl, pair); 2607 } else if (err != 0 && nvl != retrynvl) { 2608 /* 2609 * This may be a spurious error caused by 2610 * receiving quota and reservation out of order. 2611 * Try again in a second pass. 2612 */ 2613 err = nvlist_add_nvpair(retrynvl, pair); 2614 } 2615 } 2616 2617 if (err != 0) { 2618 if (errlist != NULL) 2619 fnvlist_add_int32(errlist, propname, err); 2620 rv = err; 2621 } 2622 } 2623 2624 if (nvl != retrynvl && !nvlist_empty(retrynvl)) { 2625 nvl = retrynvl; 2626 goto retry; 2627 } 2628 2629 if (!nvlist_empty(genericnvl) && 2630 dsl_props_set(dsname, source, genericnvl) != 0) { 2631 /* 2632 * If this fails, we still want to set as many properties as we 2633 * can, so try setting them individually. 2634 */ 2635 pair = NULL; 2636 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) { 2637 const char *propname = nvpair_name(pair); 2638 int err = 0; 2639 2640 propval = pair; 2641 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2642 nvlist_t *attrs; 2643 attrs = fnvpair_value_nvlist(pair); 2644 propval = fnvlist_lookup_nvpair(attrs, 2645 ZPROP_VALUE); 2646 } 2647 2648 if (nvpair_type(propval) == DATA_TYPE_STRING) { 2649 strval = fnvpair_value_string(propval); 2650 err = dsl_prop_set_string(dsname, propname, 2651 source, strval); 2652 } else { 2653 intval = fnvpair_value_uint64(propval); 2654 err = dsl_prop_set_int(dsname, propname, source, 2655 intval); 2656 } 2657 2658 if (err != 0) { 2659 if (errlist != NULL) { 2660 fnvlist_add_int32(errlist, propname, 2661 err); 2662 } 2663 rv = err; 2664 } 2665 } 2666 } 2667 nvlist_free(genericnvl); 2668 nvlist_free(retrynvl); 2669 2670 return (rv); 2671} 2672 2673/* 2674 * Check that all the properties are valid user properties. 2675 */ 2676static int 2677zfs_check_userprops(const char *fsname, nvlist_t *nvl) 2678{ 2679 nvpair_t *pair = NULL; 2680 int error = 0; 2681 2682 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { 2683 const char *propname = nvpair_name(pair); 2684 2685 if (!zfs_prop_user(propname) || 2686 nvpair_type(pair) != DATA_TYPE_STRING) 2687 return (SET_ERROR(EINVAL)); 2688 2689 if (error = zfs_secpolicy_write_perms(fsname, 2690 ZFS_DELEG_PERM_USERPROP, CRED())) 2691 return (error); 2692 2693 if (strlen(propname) >= ZAP_MAXNAMELEN) 2694 return (SET_ERROR(ENAMETOOLONG)); 2695 2696 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN) 2697 return (E2BIG); 2698 } 2699 return (0); 2700} 2701 2702static void 2703props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops) 2704{ 2705 nvpair_t *pair; 2706 2707 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2708 2709 pair = NULL; 2710 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) { 2711 if (nvlist_exists(skipped, nvpair_name(pair))) 2712 continue; 2713 2714 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0); 2715 } 2716} 2717 2718static int 2719clear_received_props(const char *dsname, nvlist_t *props, 2720 nvlist_t *skipped) 2721{ 2722 int err = 0; 2723 nvlist_t *cleared_props = NULL; 2724 props_skip(props, skipped, &cleared_props); 2725 if (!nvlist_empty(cleared_props)) { 2726 /* 2727 * Acts on local properties until the dataset has received 2728 * properties at least once on or after SPA_VERSION_RECVD_PROPS. 2729 */ 2730 zprop_source_t flags = (ZPROP_SRC_NONE | 2731 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0)); 2732 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL); 2733 } 2734 nvlist_free(cleared_props); 2735 return (err); 2736} 2737 2738/* 2739 * inputs: 2740 * zc_name name of filesystem 2741 * zc_value name of property to set 2742 * zc_nvlist_src{_size} nvlist of properties to apply 2743 * zc_cookie received properties flag 2744 * 2745 * outputs: 2746 * zc_nvlist_dst{_size} error for each unapplied received property 2747 */ 2748static int 2749zfs_ioc_set_prop(zfs_cmd_t *zc) 2750{ 2751 nvlist_t *nvl; 2752 boolean_t received = zc->zc_cookie; 2753 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED : 2754 ZPROP_SRC_LOCAL); 2755 nvlist_t *errors; 2756 int error; 2757 2758 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2759 zc->zc_iflags, &nvl)) != 0) 2760 return (error); 2761 2762 if (received) { 2763 nvlist_t *origprops; 2764 2765 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) { 2766 (void) clear_received_props(zc->zc_name, 2767 origprops, nvl); 2768 nvlist_free(origprops); 2769 } 2770 2771 error = dsl_prop_set_hasrecvd(zc->zc_name); 2772 } 2773 2774 errors = fnvlist_alloc(); 2775 if (error == 0) 2776 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors); 2777 2778 if (zc->zc_nvlist_dst != 0 && errors != NULL) { 2779 (void) put_nvlist(zc, errors); 2780 } 2781 2782 nvlist_free(errors); 2783 nvlist_free(nvl); 2784 return (error); 2785} 2786 2787/* 2788 * inputs: 2789 * zc_name name of filesystem 2790 * zc_value name of property to inherit 2791 * zc_cookie revert to received value if TRUE 2792 * 2793 * outputs: none 2794 */ 2795static int 2796zfs_ioc_inherit_prop(zfs_cmd_t *zc) 2797{ 2798 const char *propname = zc->zc_value; 2799 zfs_prop_t prop = zfs_name_to_prop(propname); 2800 boolean_t received = zc->zc_cookie; 2801 zprop_source_t source = (received 2802 ? ZPROP_SRC_NONE /* revert to received value, if any */ 2803 : ZPROP_SRC_INHERITED); /* explicitly inherit */ 2804 2805 if (received) { 2806 nvlist_t *dummy; 2807 nvpair_t *pair; 2808 zprop_type_t type; 2809 int err; 2810 2811 /* 2812 * zfs_prop_set_special() expects properties in the form of an 2813 * nvpair with type info. 2814 */ 2815 if (prop == ZPROP_INVAL) { 2816 if (!zfs_prop_user(propname)) 2817 return (SET_ERROR(EINVAL)); 2818 2819 type = PROP_TYPE_STRING; 2820 } else if (prop == ZFS_PROP_VOLSIZE || 2821 prop == ZFS_PROP_VERSION) { 2822 return (SET_ERROR(EINVAL)); 2823 } else { 2824 type = zfs_prop_get_type(prop); 2825 } 2826 2827 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2828 2829 switch (type) { 2830 case PROP_TYPE_STRING: 2831 VERIFY(0 == nvlist_add_string(dummy, propname, "")); 2832 break; 2833 case PROP_TYPE_NUMBER: 2834 case PROP_TYPE_INDEX: 2835 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0)); 2836 break; 2837 default: 2838 nvlist_free(dummy); 2839 return (SET_ERROR(EINVAL)); 2840 } 2841 2842 pair = nvlist_next_nvpair(dummy, NULL); 2843 err = zfs_prop_set_special(zc->zc_name, source, pair); 2844 nvlist_free(dummy); 2845 if (err != -1) 2846 return (err); /* special property already handled */ 2847 } else { 2848 /* 2849 * Only check this in the non-received case. We want to allow 2850 * 'inherit -S' to revert non-inheritable properties like quota 2851 * and reservation to the received or default values even though 2852 * they are not considered inheritable. 2853 */ 2854 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop)) 2855 return (SET_ERROR(EINVAL)); 2856 } 2857 2858 /* property name has been validated by zfs_secpolicy_inherit_prop() */ 2859 return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source)); 2860} 2861 2862static int 2863zfs_ioc_pool_set_props(zfs_cmd_t *zc) 2864{ 2865 nvlist_t *props; 2866 spa_t *spa; 2867 int error; 2868 nvpair_t *pair; 2869 2870 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2871 zc->zc_iflags, &props)) 2872 return (error); 2873 2874 /* 2875 * If the only property is the configfile, then just do a spa_lookup() 2876 * to handle the faulted case. 2877 */ 2878 pair = nvlist_next_nvpair(props, NULL); 2879 if (pair != NULL && strcmp(nvpair_name(pair), 2880 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 && 2881 nvlist_next_nvpair(props, pair) == NULL) { 2882 mutex_enter(&spa_namespace_lock); 2883 if ((spa = spa_lookup(zc->zc_name)) != NULL) { 2884 spa_configfile_set(spa, props, B_FALSE); 2885 spa_config_sync(spa, B_FALSE, B_TRUE); 2886 } 2887 mutex_exit(&spa_namespace_lock); 2888 if (spa != NULL) { 2889 nvlist_free(props); 2890 return (0); 2891 } 2892 } 2893 2894 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { 2895 nvlist_free(props); 2896 return (error); 2897 } 2898 2899 error = spa_prop_set(spa, props); 2900 2901 nvlist_free(props); 2902 spa_close(spa, FTAG); 2903 2904 return (error); 2905} 2906 2907static int 2908zfs_ioc_pool_get_props(zfs_cmd_t *zc) 2909{ 2910 spa_t *spa; 2911 int error; 2912 nvlist_t *nvp = NULL; 2913 2914 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { 2915 /* 2916 * If the pool is faulted, there may be properties we can still 2917 * get (such as altroot and cachefile), so attempt to get them 2918 * anyway. 2919 */ 2920 mutex_enter(&spa_namespace_lock); 2921 if ((spa = spa_lookup(zc->zc_name)) != NULL) 2922 error = spa_prop_get(spa, &nvp); 2923 mutex_exit(&spa_namespace_lock); 2924 } else { 2925 error = spa_prop_get(spa, &nvp); 2926 spa_close(spa, FTAG); 2927 } 2928 2929 if (error == 0 && zc->zc_nvlist_dst != 0) 2930 error = put_nvlist(zc, nvp); 2931 else 2932 error = SET_ERROR(EFAULT); 2933 2934 nvlist_free(nvp); 2935 return (error); 2936} 2937 2938/* 2939 * inputs: 2940 * zc_name name of filesystem 2941 * zc_nvlist_src{_size} nvlist of delegated permissions 2942 * zc_perm_action allow/unallow flag 2943 * 2944 * outputs: none 2945 */ 2946static int 2947zfs_ioc_set_fsacl(zfs_cmd_t *zc) 2948{ 2949 int error; 2950 nvlist_t *fsaclnv = NULL; 2951 2952 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2953 zc->zc_iflags, &fsaclnv)) != 0) 2954 return (error); 2955 2956 /* 2957 * Verify nvlist is constructed correctly 2958 */ 2959 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) { 2960 nvlist_free(fsaclnv); 2961 return (SET_ERROR(EINVAL)); 2962 } 2963 2964 /* 2965 * If we don't have PRIV_SYS_MOUNT, then validate 2966 * that user is allowed to hand out each permission in 2967 * the nvlist(s) 2968 */ 2969 2970 error = secpolicy_zfs(CRED()); 2971 if (error != 0) { 2972 if (zc->zc_perm_action == B_FALSE) { 2973 error = dsl_deleg_can_allow(zc->zc_name, 2974 fsaclnv, CRED()); 2975 } else { 2976 error = dsl_deleg_can_unallow(zc->zc_name, 2977 fsaclnv, CRED()); 2978 } 2979 } 2980 2981 if (error == 0) 2982 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action); 2983 2984 nvlist_free(fsaclnv); 2985 return (error); 2986} 2987 2988/* 2989 * inputs: 2990 * zc_name name of filesystem 2991 * 2992 * outputs: 2993 * zc_nvlist_src{_size} nvlist of delegated permissions 2994 */ 2995static int 2996zfs_ioc_get_fsacl(zfs_cmd_t *zc) 2997{ 2998 nvlist_t *nvp; 2999 int error; 3000 3001 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) { 3002 error = put_nvlist(zc, nvp); 3003 nvlist_free(nvp); 3004 } 3005 3006 return (error); 3007} 3008 3009/* 3010 * Search the vfs list for a specified resource. Returns a pointer to it 3011 * or NULL if no suitable entry is found. The caller of this routine 3012 * is responsible for releasing the returned vfs pointer. 3013 */ 3014static vfs_t * 3015zfs_get_vfs(const char *resource) 3016{ 3017 vfs_t *vfsp; 3018 3019 mtx_lock(&mountlist_mtx); 3020 TAILQ_FOREACH(vfsp, &mountlist, mnt_list) { 3021 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) { 3022 VFS_HOLD(vfsp); 3023 break; 3024 } 3025 } 3026 mtx_unlock(&mountlist_mtx); 3027 return (vfsp); 3028} 3029 3030/* ARGSUSED */ 3031static void 3032zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) 3033{ 3034 zfs_creat_t *zct = arg; 3035 3036 zfs_create_fs(os, cr, zct->zct_zplprops, tx); 3037} 3038 3039#define ZFS_PROP_UNDEFINED ((uint64_t)-1) 3040 3041/* 3042 * inputs: 3043 * os parent objset pointer (NULL if root fs) 3044 * fuids_ok fuids allowed in this version of the spa? 3045 * sa_ok SAs allowed in this version of the spa? 3046 * createprops list of properties requested by creator 3047 * 3048 * outputs: 3049 * zplprops values for the zplprops we attach to the master node object 3050 * is_ci true if requested file system will be purely case-insensitive 3051 * 3052 * Determine the settings for utf8only, normalization and 3053 * casesensitivity. Specific values may have been requested by the 3054 * creator and/or we can inherit values from the parent dataset. If 3055 * the file system is of too early a vintage, a creator can not 3056 * request settings for these properties, even if the requested 3057 * setting is the default value. We don't actually want to create dsl 3058 * properties for these, so remove them from the source nvlist after 3059 * processing. 3060 */ 3061static int 3062zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver, 3063 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops, 3064 nvlist_t *zplprops, boolean_t *is_ci) 3065{ 3066 uint64_t sense = ZFS_PROP_UNDEFINED; 3067 uint64_t norm = ZFS_PROP_UNDEFINED; 3068 uint64_t u8 = ZFS_PROP_UNDEFINED; 3069 3070 ASSERT(zplprops != NULL); 3071 3072 /* 3073 * Pull out creator prop choices, if any. 3074 */ 3075 if (createprops) { 3076 (void) nvlist_lookup_uint64(createprops, 3077 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver); 3078 (void) nvlist_lookup_uint64(createprops, 3079 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm); 3080 (void) nvlist_remove_all(createprops, 3081 zfs_prop_to_name(ZFS_PROP_NORMALIZE)); 3082 (void) nvlist_lookup_uint64(createprops, 3083 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8); 3084 (void) nvlist_remove_all(createprops, 3085 zfs_prop_to_name(ZFS_PROP_UTF8ONLY)); 3086 (void) nvlist_lookup_uint64(createprops, 3087 zfs_prop_to_name(ZFS_PROP_CASE), &sense); 3088 (void) nvlist_remove_all(createprops, 3089 zfs_prop_to_name(ZFS_PROP_CASE)); 3090 } 3091 3092 /* 3093 * If the zpl version requested is whacky or the file system 3094 * or pool is version is too "young" to support normalization 3095 * and the creator tried to set a value for one of the props, 3096 * error out. 3097 */ 3098 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) || 3099 (zplver >= ZPL_VERSION_FUID && !fuids_ok) || 3100 (zplver >= ZPL_VERSION_SA && !sa_ok) || 3101 (zplver < ZPL_VERSION_NORMALIZATION && 3102 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED || 3103 sense != ZFS_PROP_UNDEFINED))) 3104 return (SET_ERROR(ENOTSUP)); 3105 3106 /* 3107 * Put the version in the zplprops 3108 */ 3109 VERIFY(nvlist_add_uint64(zplprops, 3110 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0); 3111 3112 if (norm == ZFS_PROP_UNDEFINED) 3113 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0); 3114 VERIFY(nvlist_add_uint64(zplprops, 3115 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0); 3116 3117 /* 3118 * If we're normalizing, names must always be valid UTF-8 strings. 3119 */ 3120 if (norm) 3121 u8 = 1; 3122 if (u8 == ZFS_PROP_UNDEFINED) 3123 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0); 3124 VERIFY(nvlist_add_uint64(zplprops, 3125 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0); 3126 3127 if (sense == ZFS_PROP_UNDEFINED) 3128 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0); 3129 VERIFY(nvlist_add_uint64(zplprops, 3130 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0); 3131 3132 if (is_ci) 3133 *is_ci = (sense == ZFS_CASE_INSENSITIVE); 3134 3135 return (0); 3136} 3137 3138static int 3139zfs_fill_zplprops(const char *dataset, nvlist_t *createprops, 3140 nvlist_t *zplprops, boolean_t *is_ci) 3141{ 3142 boolean_t fuids_ok, sa_ok; 3143 uint64_t zplver = ZPL_VERSION; 3144 objset_t *os = NULL; 3145 char parentname[MAXNAMELEN]; 3146 char *cp; 3147 spa_t *spa; 3148 uint64_t spa_vers; 3149 int error; 3150 3151 (void) strlcpy(parentname, dataset, sizeof (parentname)); 3152 cp = strrchr(parentname, '/'); 3153 ASSERT(cp != NULL); 3154 cp[0] = '\0'; 3155 3156 if ((error = spa_open(dataset, &spa, FTAG)) != 0) 3157 return (error); 3158 3159 spa_vers = spa_version(spa); 3160 spa_close(spa, FTAG); 3161 3162 zplver = zfs_zpl_version_map(spa_vers); 3163 fuids_ok = (zplver >= ZPL_VERSION_FUID); 3164 sa_ok = (zplver >= ZPL_VERSION_SA); 3165 3166 /* 3167 * Open parent object set so we can inherit zplprop values. 3168 */ 3169 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0) 3170 return (error); 3171 3172 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops, 3173 zplprops, is_ci); 3174 dmu_objset_rele(os, FTAG); 3175 return (error); 3176} 3177 3178static int 3179zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops, 3180 nvlist_t *zplprops, boolean_t *is_ci) 3181{ 3182 boolean_t fuids_ok; 3183 boolean_t sa_ok; 3184 uint64_t zplver = ZPL_VERSION; 3185 int error; 3186 3187 zplver = zfs_zpl_version_map(spa_vers); 3188 fuids_ok = (zplver >= ZPL_VERSION_FUID); 3189 sa_ok = (zplver >= ZPL_VERSION_SA); 3190 3191 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok, 3192 createprops, zplprops, is_ci); 3193 return (error); 3194} 3195 3196/* 3197 * innvl: { 3198 * "type" -> dmu_objset_type_t (int32) 3199 * (optional) "props" -> { prop -> value } 3200 * } 3201 * 3202 * outnvl: propname -> error code (int32) 3203 */ 3204static int 3205zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3206{ 3207 int error = 0; 3208 zfs_creat_t zct = { 0 }; 3209 nvlist_t *nvprops = NULL; 3210 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx); 3211 int32_t type32; 3212 dmu_objset_type_t type; 3213 boolean_t is_insensitive = B_FALSE; 3214 3215 if (nvlist_lookup_int32(innvl, "type", &type32) != 0) 3216 return (SET_ERROR(EINVAL)); 3217 type = type32; 3218 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops); 3219 3220 switch (type) { 3221 case DMU_OST_ZFS: 3222 cbfunc = zfs_create_cb; 3223 break; 3224 3225 case DMU_OST_ZVOL: 3226 cbfunc = zvol_create_cb; 3227 break; 3228 3229 default: 3230 cbfunc = NULL; 3231 break; 3232 } 3233 if (strchr(fsname, '@') || 3234 strchr(fsname, '%')) 3235 return (SET_ERROR(EINVAL)); 3236 3237 zct.zct_props = nvprops; 3238 3239 if (cbfunc == NULL) 3240 return (SET_ERROR(EINVAL)); 3241 3242 if (type == DMU_OST_ZVOL) { 3243 uint64_t volsize, volblocksize; 3244 3245 if (nvprops == NULL) 3246 return (SET_ERROR(EINVAL)); 3247 if (nvlist_lookup_uint64(nvprops, 3248 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0) 3249 return (SET_ERROR(EINVAL)); 3250 3251 if ((error = nvlist_lookup_uint64(nvprops, 3252 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), 3253 &volblocksize)) != 0 && error != ENOENT) 3254 return (SET_ERROR(EINVAL)); 3255 3256 if (error != 0) 3257 volblocksize = zfs_prop_default_numeric( 3258 ZFS_PROP_VOLBLOCKSIZE); 3259 3260 if ((error = zvol_check_volblocksize( 3261 volblocksize)) != 0 || 3262 (error = zvol_check_volsize(volsize, 3263 volblocksize)) != 0) 3264 return (error); 3265 } else if (type == DMU_OST_ZFS) { 3266 int error; 3267 3268 /* 3269 * We have to have normalization and 3270 * case-folding flags correct when we do the 3271 * file system creation, so go figure them out 3272 * now. 3273 */ 3274 VERIFY(nvlist_alloc(&zct.zct_zplprops, 3275 NV_UNIQUE_NAME, KM_SLEEP) == 0); 3276 error = zfs_fill_zplprops(fsname, nvprops, 3277 zct.zct_zplprops, &is_insensitive); 3278 if (error != 0) { 3279 nvlist_free(zct.zct_zplprops); 3280 return (error); 3281 } 3282 } 3283 3284 error = dmu_objset_create(fsname, type, 3285 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct); 3286 nvlist_free(zct.zct_zplprops); 3287 3288 /* 3289 * It would be nice to do this atomically. 3290 */ 3291 if (error == 0) { 3292 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL, 3293 nvprops, outnvl); 3294 if (error != 0) 3295 (void) dsl_destroy_head(fsname); 3296 } 3297#ifdef __FreeBSD__ 3298 if (error == 0 && type == DMU_OST_ZVOL) 3299 zvol_create_minors(fsname); 3300#endif 3301 return (error); 3302} 3303 3304/* 3305 * innvl: { 3306 * "origin" -> name of origin snapshot 3307 * (optional) "props" -> { prop -> value } 3308 * } 3309 * 3310 * outnvl: propname -> error code (int32) 3311 */ 3312static int 3313zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3314{ 3315 int error = 0; 3316 nvlist_t *nvprops = NULL; 3317 char *origin_name; 3318 3319 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0) 3320 return (SET_ERROR(EINVAL)); 3321 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops); 3322 3323 if (strchr(fsname, '@') || 3324 strchr(fsname, '%')) 3325 return (SET_ERROR(EINVAL)); 3326 3327 if (dataset_namecheck(origin_name, NULL, NULL) != 0) 3328 return (SET_ERROR(EINVAL)); 3329 error = dmu_objset_clone(fsname, origin_name); 3330 if (error != 0) 3331 return (error); 3332 3333 /* 3334 * It would be nice to do this atomically. 3335 */ 3336 if (error == 0) { 3337 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL, 3338 nvprops, outnvl); 3339 if (error != 0) 3340 (void) dsl_destroy_head(fsname); 3341 } 3342#ifdef __FreeBSD__ 3343 if (error == 0) 3344 zvol_create_minors(fsname); 3345#endif 3346 return (error); 3347} 3348 3349/* 3350 * innvl: { 3351 * "snaps" -> { snapshot1, snapshot2 } 3352 * (optional) "props" -> { prop -> value (string) } 3353 * } 3354 * 3355 * outnvl: snapshot -> error code (int32) 3356 */ 3357static int 3358zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3359{ 3360 nvlist_t *snaps; 3361 nvlist_t *props = NULL; 3362 int error, poollen; 3363 nvpair_t *pair; 3364 3365 (void) nvlist_lookup_nvlist(innvl, "props", &props); 3366 if ((error = zfs_check_userprops(poolname, props)) != 0) 3367 return (error); 3368 3369 if (!nvlist_empty(props) && 3370 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS)) 3371 return (SET_ERROR(ENOTSUP)); 3372 3373 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 3374 return (SET_ERROR(EINVAL)); 3375 poollen = strlen(poolname); 3376 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 3377 pair = nvlist_next_nvpair(snaps, pair)) { 3378 const char *name = nvpair_name(pair); 3379 const char *cp = strchr(name, '@'); 3380 3381 /* 3382 * The snap name must contain an @, and the part after it must 3383 * contain only valid characters. 3384 */ 3385 if (cp == NULL || 3386 zfs_component_namecheck(cp + 1, NULL, NULL) != 0) 3387 return (SET_ERROR(EINVAL)); 3388 3389 /* 3390 * The snap must be in the specified pool. 3391 */ 3392 if (strncmp(name, poolname, poollen) != 0 || 3393 (name[poollen] != '/' && name[poollen] != '@')) 3394 return (SET_ERROR(EXDEV)); 3395 3396 /* This must be the only snap of this fs. */ 3397 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair); 3398 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) { 3399 if (strncmp(name, nvpair_name(pair2), cp - name + 1) 3400 == 0) { 3401 return (SET_ERROR(EXDEV)); 3402 } 3403 } 3404 } 3405 3406 error = dsl_dataset_snapshot(snaps, props, outnvl); 3407 return (error); 3408} 3409 3410/* 3411 * innvl: "message" -> string 3412 */ 3413/* ARGSUSED */ 3414static int 3415zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl) 3416{ 3417 char *message; 3418 spa_t *spa; 3419 int error; 3420 char *poolname; 3421 3422 /* 3423 * The poolname in the ioctl is not set, we get it from the TSD, 3424 * which was set at the end of the last successful ioctl that allows 3425 * logging. The secpolicy func already checked that it is set. 3426 * Only one log ioctl is allowed after each successful ioctl, so 3427 * we clear the TSD here. 3428 */ 3429 poolname = tsd_get(zfs_allow_log_key); 3430 (void) tsd_set(zfs_allow_log_key, NULL); 3431 error = spa_open(poolname, &spa, FTAG); 3432 strfree(poolname); 3433 if (error != 0) 3434 return (error); 3435 3436 if (nvlist_lookup_string(innvl, "message", &message) != 0) { 3437 spa_close(spa, FTAG); 3438 return (SET_ERROR(EINVAL)); 3439 } 3440 3441 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) { 3442 spa_close(spa, FTAG); 3443 return (SET_ERROR(ENOTSUP)); 3444 } 3445 3446 error = spa_history_log(spa, message); 3447 spa_close(spa, FTAG); 3448 return (error); 3449} 3450 3451/* 3452 * The dp_config_rwlock must not be held when calling this, because the 3453 * unmount may need to write out data. 3454 * 3455 * This function is best-effort. Callers must deal gracefully if it 3456 * remains mounted (or is remounted after this call). 3457 * 3458 * Returns 0 if the argument is not a snapshot, or it is not currently a 3459 * filesystem, or we were able to unmount it. Returns error code otherwise. 3460 */ 3461int 3462zfs_unmount_snap(const char *snapname) 3463{ 3464 vfs_t *vfsp; 3465 zfsvfs_t *zfsvfs; 3466 int err; 3467 3468 if (strchr(snapname, '@') == NULL) 3469 return (0); 3470 3471 vfsp = zfs_get_vfs(snapname); 3472 if (vfsp == NULL) 3473 return (0); 3474 3475 zfsvfs = vfsp->vfs_data; 3476 ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os))); 3477 3478 err = vn_vfswlock(vfsp->vfs_vnodecovered); 3479 VFS_RELE(vfsp); 3480 if (err != 0) 3481 return (SET_ERROR(err)); 3482 3483 /* 3484 * Always force the unmount for snapshots. 3485 */ 3486 3487#ifdef illumos 3488 (void) dounmount(vfsp, MS_FORCE, kcred); 3489#else 3490 mtx_lock(&Giant); /* dounmount() */ 3491 (void) dounmount(vfsp, MS_FORCE, curthread); 3492 mtx_unlock(&Giant); /* dounmount() */ 3493#endif 3494 return (0); 3495} 3496 3497/* ARGSUSED */ 3498static int 3499zfs_unmount_snap_cb(const char *snapname, void *arg) 3500{ 3501 return (zfs_unmount_snap(snapname)); 3502} 3503 3504/* 3505 * When a clone is destroyed, its origin may also need to be destroyed, 3506 * in which case it must be unmounted. This routine will do that unmount 3507 * if necessary. 3508 */ 3509void 3510zfs_destroy_unmount_origin(const char *fsname) 3511{ 3512 int error; 3513 objset_t *os; 3514 dsl_dataset_t *ds; 3515 3516 error = dmu_objset_hold(fsname, FTAG, &os); 3517 if (error != 0) 3518 return; 3519 ds = dmu_objset_ds(os); 3520 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) { 3521 char originname[MAXNAMELEN]; 3522 dsl_dataset_name(ds->ds_prev, originname); 3523 dmu_objset_rele(os, FTAG); 3524 (void) zfs_unmount_snap(originname); 3525 } else { 3526 dmu_objset_rele(os, FTAG); 3527 } 3528} 3529 3530/* 3531 * innvl: { 3532 * "snaps" -> { snapshot1, snapshot2 } 3533 * (optional boolean) "defer" 3534 * } 3535 * 3536 * outnvl: snapshot -> error code (int32) 3537 * 3538 */ 3539/* ARGSUSED */ 3540static int 3541zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3542{ 3543 nvlist_t *snaps; 3544 nvpair_t *pair; 3545 boolean_t defer; 3546 3547 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 3548 return (SET_ERROR(EINVAL)); 3549 defer = nvlist_exists(innvl, "defer"); 3550 3551 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 3552 pair = nvlist_next_nvpair(snaps, pair)) { 3553 (void) zfs_unmount_snap(nvpair_name(pair)); 3554 } 3555 3556 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl)); 3557} 3558 3559/* 3560 * Create bookmarks. Bookmark names are of the form <fs>#<bmark>. 3561 * All bookmarks must be in the same pool. 3562 * 3563 * innvl: { 3564 * bookmark1 -> snapshot1, bookmark2 -> snapshot2 3565 * } 3566 * 3567 * outnvl: bookmark -> error code (int32) 3568 * 3569 */ 3570/* ARGSUSED */ 3571static int 3572zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3573{ 3574 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 3575 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 3576 char *snap_name; 3577 3578 /* 3579 * Verify the snapshot argument. 3580 */ 3581 if (nvpair_value_string(pair, &snap_name) != 0) 3582 return (SET_ERROR(EINVAL)); 3583 3584 3585 /* Verify that the keys (bookmarks) are unique */ 3586 for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair); 3587 pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) { 3588 if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0) 3589 return (SET_ERROR(EINVAL)); 3590 } 3591 } 3592 3593 return (dsl_bookmark_create(innvl, outnvl)); 3594} 3595 3596/* 3597 * innvl: { 3598 * property 1, property 2, ... 3599 * } 3600 * 3601 * outnvl: { 3602 * bookmark name 1 -> { property 1, property 2, ... }, 3603 * bookmark name 2 -> { property 1, property 2, ... } 3604 * } 3605 * 3606 */ 3607static int 3608zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3609{ 3610 return (dsl_get_bookmarks(fsname, innvl, outnvl)); 3611} 3612 3613/* 3614 * innvl: { 3615 * bookmark name 1, bookmark name 2 3616 * } 3617 * 3618 * outnvl: bookmark -> error code (int32) 3619 * 3620 */ 3621static int 3622zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl, 3623 nvlist_t *outnvl) 3624{ 3625 int error, poollen; 3626 3627 poollen = strlen(poolname); 3628 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 3629 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 3630 const char *name = nvpair_name(pair); 3631 const char *cp = strchr(name, '#'); 3632 3633 /* 3634 * The bookmark name must contain an #, and the part after it 3635 * must contain only valid characters. 3636 */ 3637 if (cp == NULL || 3638 zfs_component_namecheck(cp + 1, NULL, NULL) != 0) 3639 return (SET_ERROR(EINVAL)); 3640 3641 /* 3642 * The bookmark must be in the specified pool. 3643 */ 3644 if (strncmp(name, poolname, poollen) != 0 || 3645 (name[poollen] != '/' && name[poollen] != '#')) 3646 return (SET_ERROR(EXDEV)); 3647 (void) zvol_remove_minor(name); 3648 } 3649 3650 error = dsl_bookmark_destroy(innvl, outnvl); 3651 return (error); 3652} 3653 3654/* 3655 * inputs: 3656 * zc_name name of dataset to destroy 3657 * zc_objset_type type of objset 3658 * zc_defer_destroy mark for deferred destroy 3659 * 3660 * outputs: none 3661 */ 3662static int 3663zfs_ioc_destroy(zfs_cmd_t *zc) 3664{ 3665 int err; 3666 3667 if (zc->zc_objset_type == DMU_OST_ZFS) { 3668 err = zfs_unmount_snap(zc->zc_name); 3669 if (err != 0) 3670 return (err); 3671 } 3672 3673 if (strchr(zc->zc_name, '@')) 3674 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy); 3675 else 3676 err = dsl_destroy_head(zc->zc_name); 3677 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0) 3678 (void) zvol_remove_minor(zc->zc_name); 3679 return (err); 3680} 3681 3682/* 3683 * fsname is name of dataset to rollback (to most recent snapshot) 3684 * 3685 * innvl is not used. 3686 * 3687 * outnvl: "target" -> name of most recent snapshot 3688 * } 3689 */ 3690/* ARGSUSED */ 3691static int 3692zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl) 3693{ 3694 zfsvfs_t *zfsvfs; 3695 int error; 3696 3697 if (getzfsvfs(fsname, &zfsvfs) == 0) { 3698 error = zfs_suspend_fs(zfsvfs); 3699 if (error == 0) { 3700 int resume_err; 3701 3702 error = dsl_dataset_rollback(fsname, zfsvfs, outnvl); 3703 resume_err = zfs_resume_fs(zfsvfs, fsname); 3704 error = error ? error : resume_err; 3705 } 3706 VFS_RELE(zfsvfs->z_vfs); 3707 } else { 3708 error = dsl_dataset_rollback(fsname, NULL, outnvl); 3709 } 3710 return (error); 3711} 3712 3713static int 3714recursive_unmount(const char *fsname, void *arg) 3715{ 3716 const char *snapname = arg; 3717 char fullname[MAXNAMELEN]; 3718 3719 (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname); 3720 return (zfs_unmount_snap(fullname)); 3721} 3722 3723/* 3724 * inputs: 3725 * zc_name old name of dataset 3726 * zc_value new name of dataset 3727 * zc_cookie recursive flag (only valid for snapshots) 3728 * 3729 * outputs: none 3730 */ 3731static int 3732zfs_ioc_rename(zfs_cmd_t *zc) 3733{ 3734 boolean_t recursive = zc->zc_cookie & 1; 3735#ifdef __FreeBSD__ 3736 boolean_t allow_mounted = zc->zc_cookie & 2; 3737#endif 3738 char *at; 3739 3740 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0'; 3741 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 3742 strchr(zc->zc_value, '%')) 3743 return (SET_ERROR(EINVAL)); 3744 3745 at = strchr(zc->zc_name, '@'); 3746 if (at != NULL) { 3747 /* snaps must be in same fs */ 3748 int error; 3749 3750 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1)) 3751 return (SET_ERROR(EXDEV)); 3752 *at = '\0'; 3753#ifdef illumos 3754 if (zc->zc_objset_type == DMU_OST_ZFS) { 3755#else 3756 if (zc->zc_objset_type == DMU_OST_ZFS && allow_mounted) { 3757#endif 3758 error = dmu_objset_find(zc->zc_name, 3759 recursive_unmount, at + 1, 3760 recursive ? DS_FIND_CHILDREN : 0); 3761 if (error != 0) { 3762 *at = '@'; 3763 return (error); 3764 } 3765 } 3766 error = dsl_dataset_rename_snapshot(zc->zc_name, 3767 at + 1, strchr(zc->zc_value, '@') + 1, recursive); 3768 *at = '@'; 3769 3770 return (error); 3771 } else { 3772#ifdef illumos 3773 if (zc->zc_objset_type == DMU_OST_ZVOL) 3774 (void) zvol_remove_minor(zc->zc_name); 3775#endif 3776 return (dsl_dir_rename(zc->zc_name, zc->zc_value)); 3777 } 3778} 3779 3780static int 3781zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr) 3782{ 3783 const char *propname = nvpair_name(pair); 3784 boolean_t issnap = (strchr(dsname, '@') != NULL); 3785 zfs_prop_t prop = zfs_name_to_prop(propname); 3786 uint64_t intval; 3787 int err; 3788 3789 if (prop == ZPROP_INVAL) { 3790 if (zfs_prop_user(propname)) { 3791 if (err = zfs_secpolicy_write_perms(dsname, 3792 ZFS_DELEG_PERM_USERPROP, cr)) 3793 return (err); 3794 return (0); 3795 } 3796 3797 if (!issnap && zfs_prop_userquota(propname)) { 3798 const char *perm = NULL; 3799 const char *uq_prefix = 3800 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA]; 3801 const char *gq_prefix = 3802 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA]; 3803 3804 if (strncmp(propname, uq_prefix, 3805 strlen(uq_prefix)) == 0) { 3806 perm = ZFS_DELEG_PERM_USERQUOTA; 3807 } else if (strncmp(propname, gq_prefix, 3808 strlen(gq_prefix)) == 0) { 3809 perm = ZFS_DELEG_PERM_GROUPQUOTA; 3810 } else { 3811 /* USERUSED and GROUPUSED are read-only */ 3812 return (SET_ERROR(EINVAL)); 3813 } 3814 3815 if (err = zfs_secpolicy_write_perms(dsname, perm, cr)) 3816 return (err); 3817 return (0); 3818 } 3819 3820 return (SET_ERROR(EINVAL)); 3821 } 3822 3823 if (issnap) 3824 return (SET_ERROR(EINVAL)); 3825 3826 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 3827 /* 3828 * dsl_prop_get_all_impl() returns properties in this 3829 * format. 3830 */ 3831 nvlist_t *attrs; 3832 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 3833 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 3834 &pair) == 0); 3835 } 3836 3837 /* 3838 * Check that this value is valid for this pool version 3839 */ 3840 switch (prop) { 3841 case ZFS_PROP_COMPRESSION: 3842 /* 3843 * If the user specified gzip compression, make sure 3844 * the SPA supports it. We ignore any errors here since 3845 * we'll catch them later. 3846 */ 3847 if (nvpair_type(pair) == DATA_TYPE_UINT64 && 3848 nvpair_value_uint64(pair, &intval) == 0) { 3849 if (intval >= ZIO_COMPRESS_GZIP_1 && 3850 intval <= ZIO_COMPRESS_GZIP_9 && 3851 zfs_earlier_version(dsname, 3852 SPA_VERSION_GZIP_COMPRESSION)) { 3853 return (SET_ERROR(ENOTSUP)); 3854 } 3855 3856 if (intval == ZIO_COMPRESS_ZLE && 3857 zfs_earlier_version(dsname, 3858 SPA_VERSION_ZLE_COMPRESSION)) 3859 return (SET_ERROR(ENOTSUP)); 3860 3861 if (intval == ZIO_COMPRESS_LZ4) { 3862 spa_t *spa; 3863 3864 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 3865 return (err); 3866 3867 if (!spa_feature_is_enabled(spa, 3868 SPA_FEATURE_LZ4_COMPRESS)) { 3869 spa_close(spa, FTAG); 3870 return (SET_ERROR(ENOTSUP)); 3871 } 3872 spa_close(spa, FTAG); 3873 } 3874 3875 /* 3876 * If this is a bootable dataset then 3877 * verify that the compression algorithm 3878 * is supported for booting. We must return 3879 * something other than ENOTSUP since it 3880 * implies a downrev pool version. 3881 */ 3882 if (zfs_is_bootfs(dsname) && 3883 !BOOTFS_COMPRESS_VALID(intval)) { 3884 return (SET_ERROR(ERANGE)); 3885 } 3886 } 3887 break; 3888 3889 case ZFS_PROP_COPIES: 3890 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS)) 3891 return (SET_ERROR(ENOTSUP)); 3892 break; 3893 3894 case ZFS_PROP_DEDUP: 3895 if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP)) 3896 return (SET_ERROR(ENOTSUP)); 3897 break; 3898 3899 case ZFS_PROP_SHARESMB: 3900 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID)) 3901 return (SET_ERROR(ENOTSUP)); 3902 break; 3903 3904 case ZFS_PROP_ACLINHERIT: 3905 if (nvpair_type(pair) == DATA_TYPE_UINT64 && 3906 nvpair_value_uint64(pair, &intval) == 0) { 3907 if (intval == ZFS_ACL_PASSTHROUGH_X && 3908 zfs_earlier_version(dsname, 3909 SPA_VERSION_PASSTHROUGH_X)) 3910 return (SET_ERROR(ENOTSUP)); 3911 } 3912 break; 3913 } 3914 3915 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED())); 3916} 3917 3918/* 3919 * Checks for a race condition to make sure we don't increment a feature flag 3920 * multiple times. 3921 */ 3922static int 3923zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx) 3924{ 3925 spa_t *spa = dmu_tx_pool(tx)->dp_spa; 3926 spa_feature_t *featurep = arg; 3927 3928 if (!spa_feature_is_active(spa, *featurep)) 3929 return (0); 3930 else 3931 return (SET_ERROR(EBUSY)); 3932} 3933 3934/* 3935 * The callback invoked on feature activation in the sync task caused by 3936 * zfs_prop_activate_feature. 3937 */ 3938static void 3939zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx) 3940{ 3941 spa_t *spa = dmu_tx_pool(tx)->dp_spa; 3942 spa_feature_t *featurep = arg; 3943 3944 spa_feature_incr(spa, *featurep, tx); 3945} 3946 3947/* 3948 * Activates a feature on a pool in response to a property setting. This 3949 * creates a new sync task which modifies the pool to reflect the feature 3950 * as being active. 3951 */ 3952static int 3953zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature) 3954{ 3955 int err; 3956 3957 /* EBUSY here indicates that the feature is already active */ 3958 err = dsl_sync_task(spa_name(spa), 3959 zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync, 3960 &feature, 2); 3961 3962 if (err != 0 && err != EBUSY) 3963 return (err); 3964 else 3965 return (0); 3966} 3967 3968/* 3969 * Removes properties from the given props list that fail permission checks 3970 * needed to clear them and to restore them in case of a receive error. For each 3971 * property, make sure we have both set and inherit permissions. 3972 * 3973 * Returns the first error encountered if any permission checks fail. If the 3974 * caller provides a non-NULL errlist, it also gives the complete list of names 3975 * of all the properties that failed a permission check along with the 3976 * corresponding error numbers. The caller is responsible for freeing the 3977 * returned errlist. 3978 * 3979 * If every property checks out successfully, zero is returned and the list 3980 * pointed at by errlist is NULL. 3981 */ 3982static int 3983zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist) 3984{ 3985 zfs_cmd_t *zc; 3986 nvpair_t *pair, *next_pair; 3987 nvlist_t *errors; 3988 int err, rv = 0; 3989 3990 if (props == NULL) 3991 return (0); 3992 3993 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0); 3994 3995 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP); 3996 (void) strcpy(zc->zc_name, dataset); 3997 pair = nvlist_next_nvpair(props, NULL); 3998 while (pair != NULL) { 3999 next_pair = nvlist_next_nvpair(props, pair); 4000 4001 (void) strcpy(zc->zc_value, nvpair_name(pair)); 4002 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 || 4003 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) { 4004 VERIFY(nvlist_remove_nvpair(props, pair) == 0); 4005 VERIFY(nvlist_add_int32(errors, 4006 zc->zc_value, err) == 0); 4007 } 4008 pair = next_pair; 4009 } 4010 kmem_free(zc, sizeof (zfs_cmd_t)); 4011 4012 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) { 4013 nvlist_free(errors); 4014 errors = NULL; 4015 } else { 4016 VERIFY(nvpair_value_int32(pair, &rv) == 0); 4017 } 4018 4019 if (errlist == NULL) 4020 nvlist_free(errors); 4021 else 4022 *errlist = errors; 4023 4024 return (rv); 4025} 4026 4027static boolean_t 4028propval_equals(nvpair_t *p1, nvpair_t *p2) 4029{ 4030 if (nvpair_type(p1) == DATA_TYPE_NVLIST) { 4031 /* dsl_prop_get_all_impl() format */ 4032 nvlist_t *attrs; 4033 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0); 4034 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 4035 &p1) == 0); 4036 } 4037 4038 if (nvpair_type(p2) == DATA_TYPE_NVLIST) { 4039 nvlist_t *attrs; 4040 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0); 4041 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 4042 &p2) == 0); 4043 } 4044 4045 if (nvpair_type(p1) != nvpair_type(p2)) 4046 return (B_FALSE); 4047 4048 if (nvpair_type(p1) == DATA_TYPE_STRING) { 4049 char *valstr1, *valstr2; 4050 4051 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0); 4052 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0); 4053 return (strcmp(valstr1, valstr2) == 0); 4054 } else { 4055 uint64_t intval1, intval2; 4056 4057 VERIFY(nvpair_value_uint64(p1, &intval1) == 0); 4058 VERIFY(nvpair_value_uint64(p2, &intval2) == 0); 4059 return (intval1 == intval2); 4060 } 4061} 4062 4063/* 4064 * Remove properties from props if they are not going to change (as determined 4065 * by comparison with origprops). Remove them from origprops as well, since we 4066 * do not need to clear or restore properties that won't change. 4067 */ 4068static void 4069props_reduce(nvlist_t *props, nvlist_t *origprops) 4070{ 4071 nvpair_t *pair, *next_pair; 4072 4073 if (origprops == NULL) 4074 return; /* all props need to be received */ 4075 4076 pair = nvlist_next_nvpair(props, NULL); 4077 while (pair != NULL) { 4078 const char *propname = nvpair_name(pair); 4079 nvpair_t *match; 4080 4081 next_pair = nvlist_next_nvpair(props, pair); 4082 4083 if ((nvlist_lookup_nvpair(origprops, propname, 4084 &match) != 0) || !propval_equals(pair, match)) 4085 goto next; /* need to set received value */ 4086 4087 /* don't clear the existing received value */ 4088 (void) nvlist_remove_nvpair(origprops, match); 4089 /* don't bother receiving the property */ 4090 (void) nvlist_remove_nvpair(props, pair); 4091next: 4092 pair = next_pair; 4093 } 4094} 4095 4096#ifdef DEBUG 4097static boolean_t zfs_ioc_recv_inject_err; 4098#endif 4099 4100/* 4101 * inputs: 4102 * zc_name name of containing filesystem 4103 * zc_nvlist_src{_size} nvlist of properties to apply 4104 * zc_value name of snapshot to create 4105 * zc_string name of clone origin (if DRR_FLAG_CLONE) 4106 * zc_cookie file descriptor to recv from 4107 * zc_begin_record the BEGIN record of the stream (not byteswapped) 4108 * zc_guid force flag 4109 * zc_cleanup_fd cleanup-on-exit file descriptor 4110 * zc_action_handle handle for this guid/ds mapping (or zero on first call) 4111 * 4112 * outputs: 4113 * zc_cookie number of bytes read 4114 * zc_nvlist_dst{_size} error for each unapplied received property 4115 * zc_obj zprop_errflags_t 4116 * zc_action_handle handle for this guid/ds mapping 4117 */ 4118static int 4119zfs_ioc_recv(zfs_cmd_t *zc) 4120{ 4121 file_t *fp; 4122 dmu_recv_cookie_t drc; 4123 boolean_t force = (boolean_t)zc->zc_guid; 4124 int fd; 4125 int error = 0; 4126 int props_error = 0; 4127 nvlist_t *errors; 4128 offset_t off; 4129 nvlist_t *props = NULL; /* sent properties */ 4130 nvlist_t *origprops = NULL; /* existing properties */ 4131 char *origin = NULL; 4132 char *tosnap; 4133 char tofs[ZFS_MAXNAMELEN]; 4134 cap_rights_t rights; 4135 boolean_t first_recvd_props = B_FALSE; 4136 4137 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 4138 strchr(zc->zc_value, '@') == NULL || 4139 strchr(zc->zc_value, '%')) 4140 return (SET_ERROR(EINVAL)); 4141 4142 (void) strcpy(tofs, zc->zc_value); 4143 tosnap = strchr(tofs, '@'); 4144 *tosnap++ = '\0'; 4145 4146 if (zc->zc_nvlist_src != 0 && 4147 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 4148 zc->zc_iflags, &props)) != 0) 4149 return (error); 4150 4151 fd = zc->zc_cookie; 4152 fp = getf(fd, cap_rights_init(&rights, CAP_PREAD)); 4153 if (fp == NULL) { 4154 nvlist_free(props); 4155 return (SET_ERROR(EBADF)); 4156 } 4157 4158 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0); 4159 4160 if (zc->zc_string[0]) 4161 origin = zc->zc_string; 4162 4163 error = dmu_recv_begin(tofs, tosnap, 4164 &zc->zc_begin_record, force, origin, &drc); 4165 if (error != 0) 4166 goto out; 4167 4168 /* 4169 * Set properties before we receive the stream so that they are applied 4170 * to the new data. Note that we must call dmu_recv_stream() if 4171 * dmu_recv_begin() succeeds. 4172 */ 4173 if (props != NULL && !drc.drc_newfs) { 4174 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >= 4175 SPA_VERSION_RECVD_PROPS && 4176 !dsl_prop_get_hasrecvd(tofs)) 4177 first_recvd_props = B_TRUE; 4178 4179 /* 4180 * If new received properties are supplied, they are to 4181 * completely replace the existing received properties, so stash 4182 * away the existing ones. 4183 */ 4184 if (dsl_prop_get_received(tofs, &origprops) == 0) { 4185 nvlist_t *errlist = NULL; 4186 /* 4187 * Don't bother writing a property if its value won't 4188 * change (and avoid the unnecessary security checks). 4189 * 4190 * The first receive after SPA_VERSION_RECVD_PROPS is a 4191 * special case where we blow away all local properties 4192 * regardless. 4193 */ 4194 if (!first_recvd_props) 4195 props_reduce(props, origprops); 4196 if (zfs_check_clearable(tofs, origprops, &errlist) != 0) 4197 (void) nvlist_merge(errors, errlist, 0); 4198 nvlist_free(errlist); 4199 4200 if (clear_received_props(tofs, origprops, 4201 first_recvd_props ? NULL : props) != 0) 4202 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 4203 } else { 4204 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 4205 } 4206 } 4207 4208 if (props != NULL) { 4209 props_error = dsl_prop_set_hasrecvd(tofs); 4210 4211 if (props_error == 0) { 4212 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED, 4213 props, errors); 4214 } 4215 } 4216 4217 if (zc->zc_nvlist_dst_size != 0 && 4218 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 || 4219 put_nvlist(zc, errors) != 0)) { 4220 /* 4221 * Caller made zc->zc_nvlist_dst less than the minimum expected 4222 * size or supplied an invalid address. 4223 */ 4224 props_error = SET_ERROR(EINVAL); 4225 } 4226 4227 off = fp->f_offset; 4228 error = dmu_recv_stream(&drc, fp, &off, zc->zc_cleanup_fd, 4229 &zc->zc_action_handle); 4230 4231 if (error == 0) { 4232 zfsvfs_t *zfsvfs = NULL; 4233 4234 if (getzfsvfs(tofs, &zfsvfs) == 0) { 4235 /* online recv */ 4236 int end_err; 4237 4238 error = zfs_suspend_fs(zfsvfs); 4239 /* 4240 * If the suspend fails, then the recv_end will 4241 * likely also fail, and clean up after itself. 4242 */ 4243 end_err = dmu_recv_end(&drc, zfsvfs); 4244 if (error == 0) 4245 error = zfs_resume_fs(zfsvfs, tofs); 4246 error = error ? error : end_err; 4247 VFS_RELE(zfsvfs->z_vfs); 4248 } else { 4249 error = dmu_recv_end(&drc, NULL); 4250 } 4251 } 4252 4253 zc->zc_cookie = off - fp->f_offset; 4254 if (off >= 0 && off <= MAXOFFSET_T) 4255 fp->f_offset = off; 4256 4257#ifdef DEBUG 4258 if (zfs_ioc_recv_inject_err) { 4259 zfs_ioc_recv_inject_err = B_FALSE; 4260 error = 1; 4261 } 4262#endif 4263 4264#ifdef __FreeBSD__ 4265 if (error == 0) 4266 zvol_create_minors(tofs); 4267#endif 4268 4269 /* 4270 * On error, restore the original props. 4271 */ 4272 if (error != 0 && props != NULL && !drc.drc_newfs) { 4273 if (clear_received_props(tofs, props, NULL) != 0) { 4274 /* 4275 * We failed to clear the received properties. 4276 * Since we may have left a $recvd value on the 4277 * system, we can't clear the $hasrecvd flag. 4278 */ 4279 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4280 } else if (first_recvd_props) { 4281 dsl_prop_unset_hasrecvd(tofs); 4282 } 4283 4284 if (origprops == NULL && !drc.drc_newfs) { 4285 /* We failed to stash the original properties. */ 4286 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4287 } 4288 4289 /* 4290 * dsl_props_set() will not convert RECEIVED to LOCAL on or 4291 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL 4292 * explictly if we're restoring local properties cleared in the 4293 * first new-style receive. 4294 */ 4295 if (origprops != NULL && 4296 zfs_set_prop_nvlist(tofs, (first_recvd_props ? 4297 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED), 4298 origprops, NULL) != 0) { 4299 /* 4300 * We stashed the original properties but failed to 4301 * restore them. 4302 */ 4303 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4304 } 4305 } 4306out: 4307 nvlist_free(props); 4308 nvlist_free(origprops); 4309 nvlist_free(errors); 4310 releasef(fd); 4311 4312 if (error == 0) 4313 error = props_error; 4314 4315 return (error); 4316} 4317 4318/* 4319 * inputs: 4320 * zc_name name of snapshot to send 4321 * zc_cookie file descriptor to send stream to 4322 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj) 4323 * zc_sendobj objsetid of snapshot to send 4324 * zc_fromobj objsetid of incremental fromsnap (may be zero) 4325 * zc_guid if set, estimate size of stream only. zc_cookie is ignored. 4326 * output size in zc_objset_type. 4327 * zc_flags if =1, WRITE_EMBEDDED records are permitted 4328 * 4329 * outputs: 4330 * zc_objset_type estimated size, if zc_guid is set 4331 */ 4332static int 4333zfs_ioc_send(zfs_cmd_t *zc) 4334{ 4335 int error; 4336 offset_t off; 4337 boolean_t estimate = (zc->zc_guid != 0); 4338 boolean_t embedok = (zc->zc_flags & 0x1); 4339 4340 if (zc->zc_obj != 0) { 4341 dsl_pool_t *dp; 4342 dsl_dataset_t *tosnap; 4343 4344 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4345 if (error != 0) 4346 return (error); 4347 4348 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap); 4349 if (error != 0) { 4350 dsl_pool_rele(dp, FTAG); 4351 return (error); 4352 } 4353 4354 if (dsl_dir_is_clone(tosnap->ds_dir)) 4355 zc->zc_fromobj = tosnap->ds_dir->dd_phys->dd_origin_obj; 4356 dsl_dataset_rele(tosnap, FTAG); 4357 dsl_pool_rele(dp, FTAG); 4358 } 4359 4360 if (estimate) { 4361 dsl_pool_t *dp; 4362 dsl_dataset_t *tosnap; 4363 dsl_dataset_t *fromsnap = NULL; 4364 4365 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4366 if (error != 0) 4367 return (error); 4368 4369 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap); 4370 if (error != 0) { 4371 dsl_pool_rele(dp, FTAG); 4372 return (error); 4373 } 4374 4375 if (zc->zc_fromobj != 0) { 4376 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj, 4377 FTAG, &fromsnap); 4378 if (error != 0) { 4379 dsl_dataset_rele(tosnap, FTAG); 4380 dsl_pool_rele(dp, FTAG); 4381 return (error); 4382 } 4383 } 4384 4385 error = dmu_send_estimate(tosnap, fromsnap, 4386 &zc->zc_objset_type); 4387 4388 if (fromsnap != NULL) 4389 dsl_dataset_rele(fromsnap, FTAG); 4390 dsl_dataset_rele(tosnap, FTAG); 4391 dsl_pool_rele(dp, FTAG); 4392 } else { 4393 file_t *fp; 4394 cap_rights_t rights; 4395 4396 fp = getf(zc->zc_cookie, 4397 cap_rights_init(&rights, CAP_WRITE)); 4398 if (fp == NULL) 4399 return (SET_ERROR(EBADF)); 4400 4401 off = fp->f_offset; 4402 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj, 4403#ifdef illumos 4404 zc->zc_fromobj, embedok, zc->zc_cookie, fp->f_vnode, &off); 4405#else 4406 zc->zc_fromobj, embedok, zc->zc_cookie, fp, &off); 4407#endif 4408 4409 if (off >= 0 && off <= MAXOFFSET_T) 4410 fp->f_offset = off; 4411 releasef(zc->zc_cookie); 4412 } 4413 return (error); 4414} 4415 4416/* 4417 * inputs: 4418 * zc_name name of snapshot on which to report progress 4419 * zc_cookie file descriptor of send stream 4420 * 4421 * outputs: 4422 * zc_cookie number of bytes written in send stream thus far 4423 */ 4424static int 4425zfs_ioc_send_progress(zfs_cmd_t *zc) 4426{ 4427 dsl_pool_t *dp; 4428 dsl_dataset_t *ds; 4429 dmu_sendarg_t *dsp = NULL; 4430 int error; 4431 4432 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4433 if (error != 0) 4434 return (error); 4435 4436 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds); 4437 if (error != 0) { 4438 dsl_pool_rele(dp, FTAG); 4439 return (error); 4440 } 4441 4442 mutex_enter(&ds->ds_sendstream_lock); 4443 4444 /* 4445 * Iterate over all the send streams currently active on this dataset. 4446 * If there's one which matches the specified file descriptor _and_ the 4447 * stream was started by the current process, return the progress of 4448 * that stream. 4449 */ 4450 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL; 4451 dsp = list_next(&ds->ds_sendstreams, dsp)) { 4452 if (dsp->dsa_outfd == zc->zc_cookie && 4453 dsp->dsa_proc == curproc) 4454 break; 4455 } 4456 4457 if (dsp != NULL) 4458 zc->zc_cookie = *(dsp->dsa_off); 4459 else 4460 error = SET_ERROR(ENOENT); 4461 4462 mutex_exit(&ds->ds_sendstream_lock); 4463 dsl_dataset_rele(ds, FTAG); 4464 dsl_pool_rele(dp, FTAG); 4465 return (error); 4466} 4467 4468static int 4469zfs_ioc_inject_fault(zfs_cmd_t *zc) 4470{ 4471 int id, error; 4472 4473 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id, 4474 &zc->zc_inject_record); 4475 4476 if (error == 0) 4477 zc->zc_guid = (uint64_t)id; 4478 4479 return (error); 4480} 4481 4482static int 4483zfs_ioc_clear_fault(zfs_cmd_t *zc) 4484{ 4485 return (zio_clear_fault((int)zc->zc_guid)); 4486} 4487 4488static int 4489zfs_ioc_inject_list_next(zfs_cmd_t *zc) 4490{ 4491 int id = (int)zc->zc_guid; 4492 int error; 4493 4494 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name), 4495 &zc->zc_inject_record); 4496 4497 zc->zc_guid = id; 4498 4499 return (error); 4500} 4501 4502static int 4503zfs_ioc_error_log(zfs_cmd_t *zc) 4504{ 4505 spa_t *spa; 4506 int error; 4507 size_t count = (size_t)zc->zc_nvlist_dst_size; 4508 4509 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 4510 return (error); 4511 4512 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst, 4513 &count); 4514 if (error == 0) 4515 zc->zc_nvlist_dst_size = count; 4516 else 4517 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa); 4518 4519 spa_close(spa, FTAG); 4520 4521 return (error); 4522} 4523 4524static int 4525zfs_ioc_clear(zfs_cmd_t *zc) 4526{ 4527 spa_t *spa; 4528 vdev_t *vd; 4529 int error; 4530 4531 /* 4532 * On zpool clear we also fix up missing slogs 4533 */ 4534 mutex_enter(&spa_namespace_lock); 4535 spa = spa_lookup(zc->zc_name); 4536 if (spa == NULL) { 4537 mutex_exit(&spa_namespace_lock); 4538 return (SET_ERROR(EIO)); 4539 } 4540 if (spa_get_log_state(spa) == SPA_LOG_MISSING) { 4541 /* we need to let spa_open/spa_load clear the chains */ 4542 spa_set_log_state(spa, SPA_LOG_CLEAR); 4543 } 4544 spa->spa_last_open_failed = 0; 4545 mutex_exit(&spa_namespace_lock); 4546 4547 if (zc->zc_cookie & ZPOOL_NO_REWIND) { 4548 error = spa_open(zc->zc_name, &spa, FTAG); 4549 } else { 4550 nvlist_t *policy; 4551 nvlist_t *config = NULL; 4552 4553 if (zc->zc_nvlist_src == 0) 4554 return (SET_ERROR(EINVAL)); 4555 4556 if ((error = get_nvlist(zc->zc_nvlist_src, 4557 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) { 4558 error = spa_open_rewind(zc->zc_name, &spa, FTAG, 4559 policy, &config); 4560 if (config != NULL) { 4561 int err; 4562 4563 if ((err = put_nvlist(zc, config)) != 0) 4564 error = err; 4565 nvlist_free(config); 4566 } 4567 nvlist_free(policy); 4568 } 4569 } 4570 4571 if (error != 0) 4572 return (error); 4573 4574 spa_vdev_state_enter(spa, SCL_NONE); 4575 4576 if (zc->zc_guid == 0) { 4577 vd = NULL; 4578 } else { 4579 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE); 4580 if (vd == NULL) { 4581 (void) spa_vdev_state_exit(spa, NULL, ENODEV); 4582 spa_close(spa, FTAG); 4583 return (SET_ERROR(ENODEV)); 4584 } 4585 } 4586 4587 vdev_clear(spa, vd); 4588 4589 (void) spa_vdev_state_exit(spa, NULL, 0); 4590 4591 /* 4592 * Resume any suspended I/Os. 4593 */ 4594 if (zio_resume(spa) != 0) 4595 error = SET_ERROR(EIO); 4596 4597 spa_close(spa, FTAG); 4598 4599 return (error); 4600} 4601 4602static int 4603zfs_ioc_pool_reopen(zfs_cmd_t *zc) 4604{ 4605 spa_t *spa; 4606 int error; 4607 4608 error = spa_open(zc->zc_name, &spa, FTAG); 4609 if (error != 0) 4610 return (error); 4611 4612 spa_vdev_state_enter(spa, SCL_NONE); 4613 4614 /* 4615 * If a resilver is already in progress then set the 4616 * spa_scrub_reopen flag to B_TRUE so that we don't restart 4617 * the scan as a side effect of the reopen. Otherwise, let 4618 * vdev_open() decided if a resilver is required. 4619 */ 4620 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool); 4621 vdev_reopen(spa->spa_root_vdev); 4622 spa->spa_scrub_reopen = B_FALSE; 4623 4624 (void) spa_vdev_state_exit(spa, NULL, 0); 4625 spa_close(spa, FTAG); 4626 return (0); 4627} 4628/* 4629 * inputs: 4630 * zc_name name of filesystem 4631 * zc_value name of origin snapshot 4632 * 4633 * outputs: 4634 * zc_string name of conflicting snapshot, if there is one 4635 */ 4636static int 4637zfs_ioc_promote(zfs_cmd_t *zc) 4638{ 4639 char *cp; 4640 4641 /* 4642 * We don't need to unmount *all* the origin fs's snapshots, but 4643 * it's easier. 4644 */ 4645 cp = strchr(zc->zc_value, '@'); 4646 if (cp) 4647 *cp = '\0'; 4648 (void) dmu_objset_find(zc->zc_value, 4649 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS); 4650 return (dsl_dataset_promote(zc->zc_name, zc->zc_string)); 4651} 4652 4653/* 4654 * Retrieve a single {user|group}{used|quota}@... property. 4655 * 4656 * inputs: 4657 * zc_name name of filesystem 4658 * zc_objset_type zfs_userquota_prop_t 4659 * zc_value domain name (eg. "S-1-234-567-89") 4660 * zc_guid RID/UID/GID 4661 * 4662 * outputs: 4663 * zc_cookie property value 4664 */ 4665static int 4666zfs_ioc_userspace_one(zfs_cmd_t *zc) 4667{ 4668 zfsvfs_t *zfsvfs; 4669 int error; 4670 4671 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 4672 return (SET_ERROR(EINVAL)); 4673 4674 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE); 4675 if (error != 0) 4676 return (error); 4677 4678 error = zfs_userspace_one(zfsvfs, 4679 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie); 4680 zfsvfs_rele(zfsvfs, FTAG); 4681 4682 return (error); 4683} 4684 4685/* 4686 * inputs: 4687 * zc_name name of filesystem 4688 * zc_cookie zap cursor 4689 * zc_objset_type zfs_userquota_prop_t 4690 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist) 4691 * 4692 * outputs: 4693 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t) 4694 * zc_cookie zap cursor 4695 */ 4696static int 4697zfs_ioc_userspace_many(zfs_cmd_t *zc) 4698{ 4699 zfsvfs_t *zfsvfs; 4700 int bufsize = zc->zc_nvlist_dst_size; 4701 4702 if (bufsize <= 0) 4703 return (SET_ERROR(ENOMEM)); 4704 4705 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE); 4706 if (error != 0) 4707 return (error); 4708 4709 void *buf = kmem_alloc(bufsize, KM_SLEEP); 4710 4711 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie, 4712 buf, &zc->zc_nvlist_dst_size); 4713 4714 if (error == 0) { 4715 error = ddi_copyout(buf, 4716 (void *)(uintptr_t)zc->zc_nvlist_dst, 4717 zc->zc_nvlist_dst_size, zc->zc_iflags); 4718 } 4719 kmem_free(buf, bufsize); 4720 zfsvfs_rele(zfsvfs, FTAG); 4721 4722 return (error); 4723} 4724 4725/* 4726 * inputs: 4727 * zc_name name of filesystem 4728 * 4729 * outputs: 4730 * none 4731 */ 4732static int 4733zfs_ioc_userspace_upgrade(zfs_cmd_t *zc) 4734{ 4735 objset_t *os; 4736 int error = 0; 4737 zfsvfs_t *zfsvfs; 4738 4739 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) { 4740 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) { 4741 /* 4742 * If userused is not enabled, it may be because the 4743 * objset needs to be closed & reopened (to grow the 4744 * objset_phys_t). Suspend/resume the fs will do that. 4745 */ 4746 error = zfs_suspend_fs(zfsvfs); 4747 if (error == 0) { 4748 dmu_objset_refresh_ownership(zfsvfs->z_os, 4749 zfsvfs); 4750 error = zfs_resume_fs(zfsvfs, zc->zc_name); 4751 } 4752 } 4753 if (error == 0) 4754 error = dmu_objset_userspace_upgrade(zfsvfs->z_os); 4755 VFS_RELE(zfsvfs->z_vfs); 4756 } else { 4757 /* XXX kind of reading contents without owning */ 4758 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 4759 if (error != 0) 4760 return (error); 4761 4762 error = dmu_objset_userspace_upgrade(os); 4763 dmu_objset_rele(os, FTAG); 4764 } 4765 4766 return (error); 4767} 4768 4769#ifdef sun 4770/* 4771 * We don't want to have a hard dependency 4772 * against some special symbols in sharefs 4773 * nfs, and smbsrv. Determine them if needed when 4774 * the first file system is shared. 4775 * Neither sharefs, nfs or smbsrv are unloadable modules. 4776 */ 4777int (*znfsexport_fs)(void *arg); 4778int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t); 4779int (*zsmbexport_fs)(void *arg, boolean_t add_share); 4780 4781int zfs_nfsshare_inited; 4782int zfs_smbshare_inited; 4783 4784ddi_modhandle_t nfs_mod; 4785ddi_modhandle_t sharefs_mod; 4786ddi_modhandle_t smbsrv_mod; 4787#endif /* sun */ 4788kmutex_t zfs_share_lock; 4789 4790#ifdef sun 4791static int 4792zfs_init_sharefs() 4793{ 4794 int error; 4795 4796 ASSERT(MUTEX_HELD(&zfs_share_lock)); 4797 /* Both NFS and SMB shares also require sharetab support. */ 4798 if (sharefs_mod == NULL && ((sharefs_mod = 4799 ddi_modopen("fs/sharefs", 4800 KRTLD_MODE_FIRST, &error)) == NULL)) { 4801 return (SET_ERROR(ENOSYS)); 4802 } 4803 if (zshare_fs == NULL && ((zshare_fs = 4804 (int (*)(enum sharefs_sys_op, share_t *, uint32_t)) 4805 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) { 4806 return (SET_ERROR(ENOSYS)); 4807 } 4808 return (0); 4809} 4810#endif /* sun */ 4811 4812static int 4813zfs_ioc_share(zfs_cmd_t *zc) 4814{ 4815#ifdef sun 4816 int error; 4817 int opcode; 4818 4819 switch (zc->zc_share.z_sharetype) { 4820 case ZFS_SHARE_NFS: 4821 case ZFS_UNSHARE_NFS: 4822 if (zfs_nfsshare_inited == 0) { 4823 mutex_enter(&zfs_share_lock); 4824 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs", 4825 KRTLD_MODE_FIRST, &error)) == NULL)) { 4826 mutex_exit(&zfs_share_lock); 4827 return (SET_ERROR(ENOSYS)); 4828 } 4829 if (znfsexport_fs == NULL && 4830 ((znfsexport_fs = (int (*)(void *)) 4831 ddi_modsym(nfs_mod, 4832 "nfs_export", &error)) == NULL)) { 4833 mutex_exit(&zfs_share_lock); 4834 return (SET_ERROR(ENOSYS)); 4835 } 4836 error = zfs_init_sharefs(); 4837 if (error != 0) { 4838 mutex_exit(&zfs_share_lock); 4839 return (SET_ERROR(ENOSYS)); 4840 } 4841 zfs_nfsshare_inited = 1; 4842 mutex_exit(&zfs_share_lock); 4843 } 4844 break; 4845 case ZFS_SHARE_SMB: 4846 case ZFS_UNSHARE_SMB: 4847 if (zfs_smbshare_inited == 0) { 4848 mutex_enter(&zfs_share_lock); 4849 if (smbsrv_mod == NULL && ((smbsrv_mod = 4850 ddi_modopen("drv/smbsrv", 4851 KRTLD_MODE_FIRST, &error)) == NULL)) { 4852 mutex_exit(&zfs_share_lock); 4853 return (SET_ERROR(ENOSYS)); 4854 } 4855 if (zsmbexport_fs == NULL && ((zsmbexport_fs = 4856 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod, 4857 "smb_server_share", &error)) == NULL)) { 4858 mutex_exit(&zfs_share_lock); 4859 return (SET_ERROR(ENOSYS)); 4860 } 4861 error = zfs_init_sharefs(); 4862 if (error != 0) { 4863 mutex_exit(&zfs_share_lock); 4864 return (SET_ERROR(ENOSYS)); 4865 } 4866 zfs_smbshare_inited = 1; 4867 mutex_exit(&zfs_share_lock); 4868 } 4869 break; 4870 default: 4871 return (SET_ERROR(EINVAL)); 4872 } 4873 4874 switch (zc->zc_share.z_sharetype) { 4875 case ZFS_SHARE_NFS: 4876 case ZFS_UNSHARE_NFS: 4877 if (error = 4878 znfsexport_fs((void *) 4879 (uintptr_t)zc->zc_share.z_exportdata)) 4880 return (error); 4881 break; 4882 case ZFS_SHARE_SMB: 4883 case ZFS_UNSHARE_SMB: 4884 if (error = zsmbexport_fs((void *) 4885 (uintptr_t)zc->zc_share.z_exportdata, 4886 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ? 4887 B_TRUE: B_FALSE)) { 4888 return (error); 4889 } 4890 break; 4891 } 4892 4893 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS || 4894 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ? 4895 SHAREFS_ADD : SHAREFS_REMOVE; 4896 4897 /* 4898 * Add or remove share from sharetab 4899 */ 4900 error = zshare_fs(opcode, 4901 (void *)(uintptr_t)zc->zc_share.z_sharedata, 4902 zc->zc_share.z_sharemax); 4903 4904 return (error); 4905 4906#else /* !sun */ 4907 return (ENOSYS); 4908#endif /* !sun */ 4909} 4910 4911ace_t full_access[] = { 4912 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0} 4913}; 4914 4915/* 4916 * inputs: 4917 * zc_name name of containing filesystem 4918 * zc_obj object # beyond which we want next in-use object # 4919 * 4920 * outputs: 4921 * zc_obj next in-use object # 4922 */ 4923static int 4924zfs_ioc_next_obj(zfs_cmd_t *zc) 4925{ 4926 objset_t *os = NULL; 4927 int error; 4928 4929 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 4930 if (error != 0) 4931 return (error); 4932 4933 error = dmu_object_next(os, &zc->zc_obj, B_FALSE, 4934 os->os_dsl_dataset->ds_phys->ds_prev_snap_txg); 4935 4936 dmu_objset_rele(os, FTAG); 4937 return (error); 4938} 4939 4940/* 4941 * inputs: 4942 * zc_name name of filesystem 4943 * zc_value prefix name for snapshot 4944 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process 4945 * 4946 * outputs: 4947 * zc_value short name of new snapshot 4948 */ 4949static int 4950zfs_ioc_tmp_snapshot(zfs_cmd_t *zc) 4951{ 4952 char *snap_name; 4953 char *hold_name; 4954 int error; 4955 minor_t minor; 4956 4957 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor); 4958 if (error != 0) 4959 return (error); 4960 4961 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value, 4962 (u_longlong_t)ddi_get_lbolt64()); 4963 hold_name = kmem_asprintf("%%%s", zc->zc_value); 4964 4965 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor, 4966 hold_name); 4967 if (error == 0) 4968 (void) strcpy(zc->zc_value, snap_name); 4969 strfree(snap_name); 4970 strfree(hold_name); 4971 zfs_onexit_fd_rele(zc->zc_cleanup_fd); 4972 return (error); 4973} 4974 4975/* 4976 * inputs: 4977 * zc_name name of "to" snapshot 4978 * zc_value name of "from" snapshot 4979 * zc_cookie file descriptor to write diff data on 4980 * 4981 * outputs: 4982 * dmu_diff_record_t's to the file descriptor 4983 */ 4984static int 4985zfs_ioc_diff(zfs_cmd_t *zc) 4986{ 4987 file_t *fp; 4988 cap_rights_t rights; 4989 offset_t off; 4990 int error; 4991 4992 fp = getf(zc->zc_cookie, cap_rights_init(&rights, CAP_WRITE)); 4993 if (fp == NULL) 4994 return (SET_ERROR(EBADF)); 4995 4996 off = fp->f_offset; 4997 4998#ifdef illumos 4999 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off); 5000#else 5001 error = dmu_diff(zc->zc_name, zc->zc_value, fp, &off); 5002#endif 5003 5004 if (off >= 0 && off <= MAXOFFSET_T) 5005 fp->f_offset = off; 5006 releasef(zc->zc_cookie); 5007 5008 return (error); 5009} 5010 5011#ifdef sun 5012/* 5013 * Remove all ACL files in shares dir 5014 */ 5015static int 5016zfs_smb_acl_purge(znode_t *dzp) 5017{ 5018 zap_cursor_t zc; 5019 zap_attribute_t zap; 5020 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 5021 int error; 5022 5023 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id); 5024 (error = zap_cursor_retrieve(&zc, &zap)) == 0; 5025 zap_cursor_advance(&zc)) { 5026 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred, 5027 NULL, 0)) != 0) 5028 break; 5029 } 5030 zap_cursor_fini(&zc); 5031 return (error); 5032} 5033#endif /* sun */ 5034 5035static int 5036zfs_ioc_smb_acl(zfs_cmd_t *zc) 5037{ 5038#ifdef sun 5039 vnode_t *vp; 5040 znode_t *dzp; 5041 vnode_t *resourcevp = NULL; 5042 znode_t *sharedir; 5043 zfsvfs_t *zfsvfs; 5044 nvlist_t *nvlist; 5045 char *src, *target; 5046 vattr_t vattr; 5047 vsecattr_t vsec; 5048 int error = 0; 5049 5050 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, 5051 NO_FOLLOW, NULL, &vp)) != 0) 5052 return (error); 5053 5054 /* Now make sure mntpnt and dataset are ZFS */ 5055 5056 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 || 5057 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), 5058 zc->zc_name) != 0)) { 5059 VN_RELE(vp); 5060 return (SET_ERROR(EINVAL)); 5061 } 5062 5063 dzp = VTOZ(vp); 5064 zfsvfs = dzp->z_zfsvfs; 5065 ZFS_ENTER(zfsvfs); 5066 5067 /* 5068 * Create share dir if its missing. 5069 */ 5070 mutex_enter(&zfsvfs->z_lock); 5071 if (zfsvfs->z_shares_dir == 0) { 5072 dmu_tx_t *tx; 5073 5074 tx = dmu_tx_create(zfsvfs->z_os); 5075 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE, 5076 ZFS_SHARES_DIR); 5077 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); 5078 error = dmu_tx_assign(tx, TXG_WAIT); 5079 if (error != 0) { 5080 dmu_tx_abort(tx); 5081 } else { 5082 error = zfs_create_share_dir(zfsvfs, tx); 5083 dmu_tx_commit(tx); 5084 } 5085 if (error != 0) { 5086 mutex_exit(&zfsvfs->z_lock); 5087 VN_RELE(vp); 5088 ZFS_EXIT(zfsvfs); 5089 return (error); 5090 } 5091 } 5092 mutex_exit(&zfsvfs->z_lock); 5093 5094 ASSERT(zfsvfs->z_shares_dir); 5095 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) { 5096 VN_RELE(vp); 5097 ZFS_EXIT(zfsvfs); 5098 return (error); 5099 } 5100 5101 switch (zc->zc_cookie) { 5102 case ZFS_SMB_ACL_ADD: 5103 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; 5104 vattr.va_type = VREG; 5105 vattr.va_mode = S_IFREG|0777; 5106 vattr.va_uid = 0; 5107 vattr.va_gid = 0; 5108 5109 vsec.vsa_mask = VSA_ACE; 5110 vsec.vsa_aclentp = &full_access; 5111 vsec.vsa_aclentsz = sizeof (full_access); 5112 vsec.vsa_aclcnt = 1; 5113 5114 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string, 5115 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec); 5116 if (resourcevp) 5117 VN_RELE(resourcevp); 5118 break; 5119 5120 case ZFS_SMB_ACL_REMOVE: 5121 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred, 5122 NULL, 0); 5123 break; 5124 5125 case ZFS_SMB_ACL_RENAME: 5126 if ((error = get_nvlist(zc->zc_nvlist_src, 5127 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) { 5128 VN_RELE(vp); 5129 ZFS_EXIT(zfsvfs); 5130 return (error); 5131 } 5132 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) || 5133 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET, 5134 &target)) { 5135 VN_RELE(vp); 5136 VN_RELE(ZTOV(sharedir)); 5137 ZFS_EXIT(zfsvfs); 5138 nvlist_free(nvlist); 5139 return (error); 5140 } 5141 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target, 5142 kcred, NULL, 0); 5143 nvlist_free(nvlist); 5144 break; 5145 5146 case ZFS_SMB_ACL_PURGE: 5147 error = zfs_smb_acl_purge(sharedir); 5148 break; 5149 5150 default: 5151 error = SET_ERROR(EINVAL); 5152 break; 5153 } 5154 5155 VN_RELE(vp); 5156 VN_RELE(ZTOV(sharedir)); 5157 5158 ZFS_EXIT(zfsvfs); 5159 5160 return (error); 5161#else /* !sun */ 5162 return (EOPNOTSUPP); 5163#endif /* !sun */ 5164} 5165 5166/* 5167 * innvl: { 5168 * "holds" -> { snapname -> holdname (string), ... } 5169 * (optional) "cleanup_fd" -> fd (int32) 5170 * } 5171 * 5172 * outnvl: { 5173 * snapname -> error value (int32) 5174 * ... 5175 * } 5176 */ 5177/* ARGSUSED */ 5178static int 5179zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist) 5180{ 5181 nvlist_t *holds; 5182 int cleanup_fd = -1; 5183 int error; 5184 minor_t minor = 0; 5185 5186 error = nvlist_lookup_nvlist(args, "holds", &holds); 5187 if (error != 0) 5188 return (SET_ERROR(EINVAL)); 5189 5190 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) { 5191 error = zfs_onexit_fd_hold(cleanup_fd, &minor); 5192 if (error != 0) 5193 return (error); 5194 } 5195 5196 error = dsl_dataset_user_hold(holds, minor, errlist); 5197 if (minor != 0) 5198 zfs_onexit_fd_rele(cleanup_fd); 5199 return (error); 5200} 5201 5202/* 5203 * innvl is not used. 5204 * 5205 * outnvl: { 5206 * holdname -> time added (uint64 seconds since epoch) 5207 * ... 5208 * } 5209 */ 5210/* ARGSUSED */ 5211static int 5212zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl) 5213{ 5214 return (dsl_dataset_get_holds(snapname, outnvl)); 5215} 5216 5217/* 5218 * innvl: { 5219 * snapname -> { holdname, ... } 5220 * ... 5221 * } 5222 * 5223 * outnvl: { 5224 * snapname -> error value (int32) 5225 * ... 5226 * } 5227 */ 5228/* ARGSUSED */ 5229static int 5230zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist) 5231{ 5232 return (dsl_dataset_user_release(holds, errlist)); 5233} 5234 5235/* 5236 * inputs: 5237 * zc_name name of new filesystem or snapshot 5238 * zc_value full name of old snapshot 5239 * 5240 * outputs: 5241 * zc_cookie space in bytes 5242 * zc_objset_type compressed space in bytes 5243 * zc_perm_action uncompressed space in bytes 5244 */ 5245static int 5246zfs_ioc_space_written(zfs_cmd_t *zc) 5247{ 5248 int error; 5249 dsl_pool_t *dp; 5250 dsl_dataset_t *new, *old; 5251 5252 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 5253 if (error != 0) 5254 return (error); 5255 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new); 5256 if (error != 0) { 5257 dsl_pool_rele(dp, FTAG); 5258 return (error); 5259 } 5260 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old); 5261 if (error != 0) { 5262 dsl_dataset_rele(new, FTAG); 5263 dsl_pool_rele(dp, FTAG); 5264 return (error); 5265 } 5266 5267 error = dsl_dataset_space_written(old, new, &zc->zc_cookie, 5268 &zc->zc_objset_type, &zc->zc_perm_action); 5269 dsl_dataset_rele(old, FTAG); 5270 dsl_dataset_rele(new, FTAG); 5271 dsl_pool_rele(dp, FTAG); 5272 return (error); 5273} 5274 5275/* 5276 * innvl: { 5277 * "firstsnap" -> snapshot name 5278 * } 5279 * 5280 * outnvl: { 5281 * "used" -> space in bytes 5282 * "compressed" -> compressed space in bytes 5283 * "uncompressed" -> uncompressed space in bytes 5284 * } 5285 */ 5286static int 5287zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl) 5288{ 5289 int error; 5290 dsl_pool_t *dp; 5291 dsl_dataset_t *new, *old; 5292 char *firstsnap; 5293 uint64_t used, comp, uncomp; 5294 5295 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0) 5296 return (SET_ERROR(EINVAL)); 5297 5298 error = dsl_pool_hold(lastsnap, FTAG, &dp); 5299 if (error != 0) 5300 return (error); 5301 5302 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new); 5303 if (error != 0) { 5304 dsl_pool_rele(dp, FTAG); 5305 return (error); 5306 } 5307 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old); 5308 if (error != 0) { 5309 dsl_dataset_rele(new, FTAG); 5310 dsl_pool_rele(dp, FTAG); 5311 return (error); 5312 } 5313 5314 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp); 5315 dsl_dataset_rele(old, FTAG); 5316 dsl_dataset_rele(new, FTAG); 5317 dsl_pool_rele(dp, FTAG); 5318 fnvlist_add_uint64(outnvl, "used", used); 5319 fnvlist_add_uint64(outnvl, "compressed", comp); 5320 fnvlist_add_uint64(outnvl, "uncompressed", uncomp); 5321 return (error); 5322} 5323 5324static int 5325zfs_ioc_jail(zfs_cmd_t *zc) 5326{ 5327 5328 return (zone_dataset_attach(curthread->td_ucred, zc->zc_name, 5329 (int)zc->zc_jailid)); 5330} 5331 5332static int 5333zfs_ioc_unjail(zfs_cmd_t *zc) 5334{ 5335 5336 return (zone_dataset_detach(curthread->td_ucred, zc->zc_name, 5337 (int)zc->zc_jailid)); 5338} 5339 5340/* 5341 * innvl: { 5342 * "fd" -> file descriptor to write stream to (int32) 5343 * (optional) "fromsnap" -> full snap name to send an incremental from 5344 * (optional) "embedok" -> (value ignored) 5345 * presence indicates DRR_WRITE_EMBEDDED records are permitted 5346 * } 5347 * 5348 * outnvl is unused 5349 */ 5350/* ARGSUSED */ 5351static int 5352zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl) 5353{ 5354 cap_rights_t rights; 5355 int error; 5356 offset_t off; 5357 char *fromname = NULL; 5358 int fd; 5359 boolean_t embedok; 5360 5361 error = nvlist_lookup_int32(innvl, "fd", &fd); 5362 if (error != 0) 5363 return (SET_ERROR(EINVAL)); 5364 5365 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname); 5366 5367 embedok = nvlist_exists(innvl, "embedok"); 5368 5369 file_t *fp = getf(fd, cap_rights_init(&rights, CAP_READ)); 5370 if (fp == NULL) 5371 return (SET_ERROR(EBADF)); 5372 5373 off = fp->f_offset; 5374#ifdef illumos 5375 error = dmu_send(snapname, fromname, embedok, fd, fp->f_vnode, &off); 5376#else 5377 error = dmu_send(snapname, fromname, embedok, fd, fp, &off); 5378#endif 5379 5380#ifdef illumos 5381 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0) 5382 fp->f_offset = off; 5383#else 5384 fp->f_offset = off; 5385#endif 5386 5387 releasef(fd); 5388 return (error); 5389} 5390 5391/* 5392 * Determine approximately how large a zfs send stream will be -- the number 5393 * of bytes that will be written to the fd supplied to zfs_ioc_send_new(). 5394 * 5395 * innvl: { 5396 * (optional) "fromsnap" -> full snap name to send an incremental from 5397 * } 5398 * 5399 * outnvl: { 5400 * "space" -> bytes of space (uint64) 5401 * } 5402 */ 5403static int 5404zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl) 5405{ 5406 dsl_pool_t *dp; 5407 dsl_dataset_t *fromsnap = NULL; 5408 dsl_dataset_t *tosnap; 5409 int error; 5410 char *fromname; 5411 uint64_t space; 5412 5413 error = dsl_pool_hold(snapname, FTAG, &dp); 5414 if (error != 0) 5415 return (error); 5416 5417 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap); 5418 if (error != 0) { 5419 dsl_pool_rele(dp, FTAG); 5420 return (error); 5421 } 5422 5423 error = nvlist_lookup_string(innvl, "fromsnap", &fromname); 5424 if (error == 0) { 5425 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap); 5426 if (error != 0) { 5427 dsl_dataset_rele(tosnap, FTAG); 5428 dsl_pool_rele(dp, FTAG); 5429 return (error); 5430 } 5431 } 5432 5433 error = dmu_send_estimate(tosnap, fromsnap, &space); 5434 fnvlist_add_uint64(outnvl, "space", space); 5435 5436 if (fromsnap != NULL) 5437 dsl_dataset_rele(fromsnap, FTAG); 5438 dsl_dataset_rele(tosnap, FTAG); 5439 dsl_pool_rele(dp, FTAG); 5440 return (error); 5441} 5442 5443 5444static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST]; 5445 5446static void 5447zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5448 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck, 5449 boolean_t log_history, zfs_ioc_poolcheck_t pool_check) 5450{ 5451 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST]; 5452 5453 ASSERT3U(ioc, >=, ZFS_IOC_FIRST); 5454 ASSERT3U(ioc, <, ZFS_IOC_LAST); 5455 ASSERT3P(vec->zvec_legacy_func, ==, NULL); 5456 ASSERT3P(vec->zvec_func, ==, NULL); 5457 5458 vec->zvec_legacy_func = func; 5459 vec->zvec_secpolicy = secpolicy; 5460 vec->zvec_namecheck = namecheck; 5461 vec->zvec_allow_log = log_history; 5462 vec->zvec_pool_check = pool_check; 5463} 5464 5465/* 5466 * See the block comment at the beginning of this file for details on 5467 * each argument to this function. 5468 */ 5469static void 5470zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func, 5471 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck, 5472 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist, 5473 boolean_t allow_log) 5474{ 5475 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST]; 5476 5477 ASSERT3U(ioc, >=, ZFS_IOC_FIRST); 5478 ASSERT3U(ioc, <, ZFS_IOC_LAST); 5479 ASSERT3P(vec->zvec_legacy_func, ==, NULL); 5480 ASSERT3P(vec->zvec_func, ==, NULL); 5481 5482 /* if we are logging, the name must be valid */ 5483 ASSERT(!allow_log || namecheck != NO_NAME); 5484 5485 vec->zvec_name = name; 5486 vec->zvec_func = func; 5487 vec->zvec_secpolicy = secpolicy; 5488 vec->zvec_namecheck = namecheck; 5489 vec->zvec_pool_check = pool_check; 5490 vec->zvec_smush_outnvlist = smush_outnvlist; 5491 vec->zvec_allow_log = allow_log; 5492} 5493 5494static void 5495zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5496 zfs_secpolicy_func_t *secpolicy, boolean_t log_history, 5497 zfs_ioc_poolcheck_t pool_check) 5498{ 5499 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5500 POOL_NAME, log_history, pool_check); 5501} 5502 5503static void 5504zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5505 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check) 5506{ 5507 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5508 DATASET_NAME, B_FALSE, pool_check); 5509} 5510 5511static void 5512zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func) 5513{ 5514 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config, 5515 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5516} 5517 5518static void 5519zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5520 zfs_secpolicy_func_t *secpolicy) 5521{ 5522 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5523 NO_NAME, B_FALSE, POOL_CHECK_NONE); 5524} 5525 5526static void 5527zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc, 5528 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy) 5529{ 5530 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5531 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED); 5532} 5533 5534static void 5535zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func) 5536{ 5537 zfs_ioctl_register_dataset_read_secpolicy(ioc, func, 5538 zfs_secpolicy_read); 5539} 5540 5541static void 5542zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5543 zfs_secpolicy_func_t *secpolicy) 5544{ 5545 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5546 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5547} 5548 5549static void 5550zfs_ioctl_init(void) 5551{ 5552 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT, 5553 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME, 5554 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5555 5556 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY, 5557 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME, 5558 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE); 5559 5560 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS, 5561 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME, 5562 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5563 5564 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW, 5565 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME, 5566 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5567 5568 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE, 5569 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME, 5570 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5571 5572 zfs_ioctl_register("create", ZFS_IOC_CREATE, 5573 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME, 5574 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5575 5576 zfs_ioctl_register("clone", ZFS_IOC_CLONE, 5577 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME, 5578 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5579 5580 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS, 5581 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME, 5582 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5583 5584 zfs_ioctl_register("hold", ZFS_IOC_HOLD, 5585 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME, 5586 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5587 zfs_ioctl_register("release", ZFS_IOC_RELEASE, 5588 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME, 5589 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5590 5591 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS, 5592 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME, 5593 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5594 5595 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK, 5596 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, 5597 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE); 5598 5599 zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK, 5600 zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME, 5601 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5602 5603 zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS, 5604 zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME, 5605 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5606 5607 zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS, 5608 zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks, 5609 POOL_NAME, 5610 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5611 5612 /* IOCTLS that use the legacy function signature */ 5613 5614 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze, 5615 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY); 5616 5617 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create, 5618 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5619 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN, 5620 zfs_ioc_pool_scan); 5621 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE, 5622 zfs_ioc_pool_upgrade); 5623 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD, 5624 zfs_ioc_vdev_add); 5625 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE, 5626 zfs_ioc_vdev_remove); 5627 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE, 5628 zfs_ioc_vdev_set_state); 5629 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH, 5630 zfs_ioc_vdev_attach); 5631 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH, 5632 zfs_ioc_vdev_detach); 5633 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH, 5634 zfs_ioc_vdev_setpath); 5635 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU, 5636 zfs_ioc_vdev_setfru); 5637 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS, 5638 zfs_ioc_pool_set_props); 5639 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT, 5640 zfs_ioc_vdev_split); 5641 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID, 5642 zfs_ioc_pool_reguid); 5643 5644 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS, 5645 zfs_ioc_pool_configs, zfs_secpolicy_none); 5646 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT, 5647 zfs_ioc_pool_tryimport, zfs_secpolicy_config); 5648 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT, 5649 zfs_ioc_inject_fault, zfs_secpolicy_inject); 5650 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT, 5651 zfs_ioc_clear_fault, zfs_secpolicy_inject); 5652 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT, 5653 zfs_ioc_inject_list_next, zfs_secpolicy_inject); 5654 5655 /* 5656 * pool destroy, and export don't log the history as part of 5657 * zfsdev_ioctl, but rather zfs_ioc_pool_export 5658 * does the logging of those commands. 5659 */ 5660 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy, 5661 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE); 5662 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export, 5663 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE); 5664 5665 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats, 5666 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE); 5667 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props, 5668 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE); 5669 5670 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log, 5671 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_NONE); 5672 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME, 5673 zfs_ioc_dsobj_to_dsname, 5674 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_NONE); 5675 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY, 5676 zfs_ioc_pool_get_history, 5677 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED); 5678 5679 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import, 5680 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5681 5682 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear, 5683 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5684 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen, 5685 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED); 5686 5687 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN, 5688 zfs_ioc_space_written); 5689 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS, 5690 zfs_ioc_objset_recvd_props); 5691 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ, 5692 zfs_ioc_next_obj); 5693 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL, 5694 zfs_ioc_get_fsacl); 5695 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS, 5696 zfs_ioc_objset_stats); 5697 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS, 5698 zfs_ioc_objset_zplprops); 5699 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT, 5700 zfs_ioc_dataset_list_next); 5701 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT, 5702 zfs_ioc_snapshot_list_next); 5703 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS, 5704 zfs_ioc_send_progress); 5705 5706 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF, 5707 zfs_ioc_diff, zfs_secpolicy_diff); 5708 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS, 5709 zfs_ioc_obj_to_stats, zfs_secpolicy_diff); 5710 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH, 5711 zfs_ioc_obj_to_path, zfs_secpolicy_diff); 5712 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE, 5713 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one); 5714 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY, 5715 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many); 5716 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND, 5717 zfs_ioc_send, zfs_secpolicy_send); 5718 5719 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop, 5720 zfs_secpolicy_none); 5721 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy, 5722 zfs_secpolicy_destroy); 5723 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename, 5724 zfs_secpolicy_rename); 5725 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv, 5726 zfs_secpolicy_recv); 5727 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote, 5728 zfs_secpolicy_promote); 5729 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP, 5730 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop); 5731 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl, 5732 zfs_secpolicy_set_fsacl); 5733 5734 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share, 5735 zfs_secpolicy_share, POOL_CHECK_NONE); 5736 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl, 5737 zfs_secpolicy_smb_acl, POOL_CHECK_NONE); 5738 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE, 5739 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade, 5740 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5741 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT, 5742 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot, 5743 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5744 5745#ifdef __FreeBSD__ 5746 zfs_ioctl_register_dataset_nolog(ZFS_IOC_JAIL, zfs_ioc_jail, 5747 zfs_secpolicy_config, POOL_CHECK_NONE); 5748 zfs_ioctl_register_dataset_nolog(ZFS_IOC_UNJAIL, zfs_ioc_unjail, 5749 zfs_secpolicy_config, POOL_CHECK_NONE); 5750#endif 5751} 5752 5753int 5754pool_status_check(const char *name, zfs_ioc_namecheck_t type, 5755 zfs_ioc_poolcheck_t check) 5756{ 5757 spa_t *spa; 5758 int error; 5759 5760 ASSERT(type == POOL_NAME || type == DATASET_NAME); 5761 5762 if (check & POOL_CHECK_NONE) 5763 return (0); 5764 5765 error = spa_open(name, &spa, FTAG); 5766 if (error == 0) { 5767 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa)) 5768 error = SET_ERROR(EAGAIN); 5769 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa)) 5770 error = SET_ERROR(EROFS); 5771 spa_close(spa, FTAG); 5772 } 5773 return (error); 5774} 5775 5776/* 5777 * Find a free minor number. 5778 */ 5779minor_t 5780zfsdev_minor_alloc(void) 5781{ 5782 static minor_t last_minor; 5783 minor_t m; 5784 5785 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 5786 5787 for (m = last_minor + 1; m != last_minor; m++) { 5788 if (m > ZFSDEV_MAX_MINOR) 5789 m = 1; 5790 if (ddi_get_soft_state(zfsdev_state, m) == NULL) { 5791 last_minor = m; 5792 return (m); 5793 } 5794 } 5795 5796 return (0); 5797} 5798 5799static int 5800zfs_ctldev_init(struct cdev *devp) 5801{ 5802 minor_t minor; 5803 zfs_soft_state_t *zs; 5804 5805 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 5806 5807 minor = zfsdev_minor_alloc(); 5808 if (minor == 0) 5809 return (SET_ERROR(ENXIO)); 5810 5811 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS) 5812 return (SET_ERROR(EAGAIN)); 5813 5814 devfs_set_cdevpriv((void *)(uintptr_t)minor, zfsdev_close); 5815 5816 zs = ddi_get_soft_state(zfsdev_state, minor); 5817 zs->zss_type = ZSST_CTLDEV; 5818 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data); 5819 5820 return (0); 5821} 5822 5823static void 5824zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor) 5825{ 5826 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 5827 5828 zfs_onexit_destroy(zo); 5829 ddi_soft_state_free(zfsdev_state, minor); 5830} 5831 5832void * 5833zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which) 5834{ 5835 zfs_soft_state_t *zp; 5836 5837 zp = ddi_get_soft_state(zfsdev_state, minor); 5838 if (zp == NULL || zp->zss_type != which) 5839 return (NULL); 5840 5841 return (zp->zss_data); 5842} 5843 5844static int 5845zfsdev_open(struct cdev *devp, int flag, int mode, struct thread *td) 5846{ 5847 int error = 0; 5848 5849#ifdef sun 5850 if (getminor(*devp) != 0) 5851 return (zvol_open(devp, flag, otyp, cr)); 5852#endif 5853 5854 /* This is the control device. Allocate a new minor if requested. */ 5855 if (flag & FEXCL) { 5856 mutex_enter(&spa_namespace_lock); 5857 error = zfs_ctldev_init(devp); 5858 mutex_exit(&spa_namespace_lock); 5859 } 5860 5861 return (error); 5862} 5863 5864static void 5865zfsdev_close(void *data) 5866{ 5867 zfs_onexit_t *zo; 5868 minor_t minor = (minor_t)(uintptr_t)data; 5869 5870 if (minor == 0) 5871 return; 5872 5873 mutex_enter(&spa_namespace_lock); 5874 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV); 5875 if (zo == NULL) { 5876 mutex_exit(&spa_namespace_lock); 5877 return; 5878 } 5879 zfs_ctldev_destroy(zo, minor); 5880 mutex_exit(&spa_namespace_lock); 5881} 5882 5883static int 5884zfsdev_ioctl(struct cdev *dev, u_long zcmd, caddr_t arg, int flag, 5885 struct thread *td) 5886{ 5887 zfs_cmd_t *zc; 5888 uint_t vecnum; 5889 int error, rc, len; 5890#ifdef illumos 5891 minor_t minor = getminor(dev); 5892#else 5893 zfs_iocparm_t *zc_iocparm; 5894 int cflag, cmd, oldvecnum; 5895 boolean_t newioc, compat; 5896 cred_t *cr = td->td_ucred; 5897#endif 5898 const zfs_ioc_vec_t *vec; 5899 char *saved_poolname = NULL; 5900 nvlist_t *innvl = NULL; 5901 5902 cflag = ZFS_CMD_COMPAT_NONE; 5903 compat = B_FALSE; 5904 newioc = B_TRUE; 5905 5906 len = IOCPARM_LEN(zcmd); 5907 cmd = zcmd & 0xff; 5908 5909 /* 5910 * Check if we are talking to supported older binaries 5911 * and translate zfs_cmd if necessary 5912 */ 5913 if (len != sizeof(zfs_iocparm_t)) { 5914 newioc = B_FALSE; 5915 if (len == sizeof(zfs_cmd_t)) { 5916 cflag = ZFS_CMD_COMPAT_LZC; 5917 vecnum = cmd; 5918 } else if (len == sizeof(zfs_cmd_deadman_t)) { 5919 cflag = ZFS_CMD_COMPAT_DEADMAN; 5920 compat = B_TRUE; 5921 vecnum = cmd; 5922 } else if (len == sizeof(zfs_cmd_v28_t)) { 5923 cflag = ZFS_CMD_COMPAT_V28; 5924 compat = B_TRUE; 5925 vecnum = cmd; 5926 } else if (len == sizeof(zfs_cmd_v15_t)) { 5927 cflag = ZFS_CMD_COMPAT_V15; 5928 compat = B_TRUE; 5929 vecnum = zfs_ioctl_v15_to_v28[cmd]; 5930 } else 5931 return (EINVAL); 5932 } else 5933 vecnum = cmd; 5934 5935#ifdef illumos 5936 vecnum = cmd - ZFS_IOC_FIRST; 5937 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip)); 5938#endif 5939 5940 if (compat) { 5941 if (vecnum == ZFS_IOC_COMPAT_PASS) 5942 return (0); 5943 else if (vecnum == ZFS_IOC_COMPAT_FAIL) 5944 return (ENOTSUP); 5945 } 5946 5947 /* 5948 * Check if we have sufficient kernel memory allocated 5949 * for the zfs_cmd_t request. Bail out if not so we 5950 * will not access undefined memory region. 5951 */ 5952 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0])) 5953 return (SET_ERROR(EINVAL)); 5954 vec = &zfs_ioc_vec[vecnum]; 5955 5956#ifdef illumos 5957 zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP); 5958 bzero(zc, sizeof(zfs_cmd_t)); 5959 5960 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag); 5961 if (error != 0) { 5962 error = SET_ERROR(EFAULT); 5963 goto out; 5964 } 5965#else /* !illumos */ 5966 /* 5967 * We don't alloc/free zc only if talking to library ioctl version 2 5968 */ 5969 if (cflag != ZFS_CMD_COMPAT_LZC) { 5970 zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP); 5971 bzero(zc, sizeof(zfs_cmd_t)); 5972 } else { 5973 zc = (void *)arg; 5974 error = 0; 5975 } 5976 5977 if (newioc) { 5978 zc_iocparm = (void *)arg; 5979 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_t)) { 5980 error = SET_ERROR(EFAULT); 5981 goto out; 5982 } 5983 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd, zc, 5984 sizeof(zfs_cmd_t), flag); 5985 if (error != 0) { 5986 error = SET_ERROR(EFAULT); 5987 goto out; 5988 } 5989 if (zc_iocparm->zfs_ioctl_version != ZFS_IOCVER_CURRENT) { 5990 compat = B_TRUE; 5991 5992 switch (zc_iocparm->zfs_ioctl_version) { 5993 case ZFS_IOCVER_ZCMD: 5994 cflag = ZFS_CMD_COMPAT_ZCMD; 5995 break; 5996 default: 5997 error = SET_ERROR(EINVAL); 5998 goto out; 5999 } 6000 } 6001 } 6002 6003 if (compat) { 6004 zfs_cmd_compat_get(zc, arg, cflag); 6005 oldvecnum = vecnum; 6006 error = zfs_ioctl_compat_pre(zc, &vecnum, cflag); 6007 if (error != 0) 6008 goto out; 6009 if (oldvecnum != vecnum) 6010 vec = &zfs_ioc_vec[vecnum]; 6011 } 6012#endif /* !illumos */ 6013 6014 zc->zc_iflags = flag & FKIOCTL; 6015 if (zc->zc_nvlist_src_size != 0) { 6016 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 6017 zc->zc_iflags, &innvl); 6018 if (error != 0) 6019 goto out; 6020 } 6021 6022 /* rewrite innvl for backwards compatibility */ 6023 if (compat) 6024 innvl = zfs_ioctl_compat_innvl(zc, innvl, vecnum, cflag); 6025 6026 /* 6027 * Ensure that all pool/dataset names are valid before we pass down to 6028 * the lower layers. 6029 */ 6030 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0'; 6031 switch (vec->zvec_namecheck) { 6032 case POOL_NAME: 6033 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0) 6034 error = SET_ERROR(EINVAL); 6035 else 6036 error = pool_status_check(zc->zc_name, 6037 vec->zvec_namecheck, vec->zvec_pool_check); 6038 break; 6039 6040 case DATASET_NAME: 6041 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0) 6042 error = SET_ERROR(EINVAL); 6043 else 6044 error = pool_status_check(zc->zc_name, 6045 vec->zvec_namecheck, vec->zvec_pool_check); 6046 break; 6047 6048 case NO_NAME: 6049 break; 6050 } 6051 6052 if (error == 0 && !(flag & FKIOCTL)) 6053 error = vec->zvec_secpolicy(zc, innvl, cr); 6054 6055 if (error != 0) 6056 goto out; 6057 6058 /* legacy ioctls can modify zc_name */ 6059 len = strcspn(zc->zc_name, "/@#") + 1; 6060 saved_poolname = kmem_alloc(len, KM_SLEEP); 6061 (void) strlcpy(saved_poolname, zc->zc_name, len); 6062 6063 if (vec->zvec_func != NULL) { 6064 nvlist_t *outnvl; 6065 int puterror = 0; 6066 spa_t *spa; 6067 nvlist_t *lognv = NULL; 6068 6069 ASSERT(vec->zvec_legacy_func == NULL); 6070 6071 /* 6072 * Add the innvl to the lognv before calling the func, 6073 * in case the func changes the innvl. 6074 */ 6075 if (vec->zvec_allow_log) { 6076 lognv = fnvlist_alloc(); 6077 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL, 6078 vec->zvec_name); 6079 if (!nvlist_empty(innvl)) { 6080 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL, 6081 innvl); 6082 } 6083 } 6084 6085 outnvl = fnvlist_alloc(); 6086 error = vec->zvec_func(zc->zc_name, innvl, outnvl); 6087 6088 if (error == 0 && vec->zvec_allow_log && 6089 spa_open(zc->zc_name, &spa, FTAG) == 0) { 6090 if (!nvlist_empty(outnvl)) { 6091 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL, 6092 outnvl); 6093 } 6094 (void) spa_history_log_nvl(spa, lognv); 6095 spa_close(spa, FTAG); 6096 } 6097 fnvlist_free(lognv); 6098 6099 /* rewrite outnvl for backwards compatibility */ 6100 if (cflag != ZFS_CMD_COMPAT_NONE && cflag != ZFS_CMD_COMPAT_LZC) 6101 outnvl = zfs_ioctl_compat_outnvl(zc, outnvl, vecnum, 6102 cflag); 6103 6104 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) { 6105 int smusherror = 0; 6106 if (vec->zvec_smush_outnvlist) { 6107 smusherror = nvlist_smush(outnvl, 6108 zc->zc_nvlist_dst_size); 6109 } 6110 if (smusherror == 0) 6111 puterror = put_nvlist(zc, outnvl); 6112 } 6113 6114 if (puterror != 0) 6115 error = puterror; 6116 6117 nvlist_free(outnvl); 6118 } else { 6119 error = vec->zvec_legacy_func(zc); 6120 } 6121 6122out: 6123 nvlist_free(innvl); 6124 6125 if (compat) { 6126 zfs_ioctl_compat_post(zc, cmd, cflag); 6127 zfs_cmd_compat_put(zc, arg, vecnum, cflag); 6128 } 6129 6130#ifdef illumos 6131 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag); 6132 if (error == 0 && rc != 0) 6133 error = SET_ERROR(EFAULT); 6134#else 6135 if (newioc) { 6136 rc = ddi_copyout(zc, (void *)(uintptr_t)zc_iocparm->zfs_cmd, 6137 sizeof (zfs_cmd_t), flag); 6138 if (error == 0 && rc != 0) 6139 error = SET_ERROR(EFAULT); 6140 } 6141#endif 6142 if (error == 0 && vec->zvec_allow_log) { 6143 char *s = tsd_get(zfs_allow_log_key); 6144 if (s != NULL) 6145 strfree(s); 6146 (void) tsd_set(zfs_allow_log_key, saved_poolname); 6147 } else { 6148 if (saved_poolname != NULL) 6149 strfree(saved_poolname); 6150 } 6151 6152#ifdef illumos 6153 kmem_free(zc, sizeof (zfs_cmd_t)); 6154#else 6155 /* 6156 * We don't alloc/free zc only if talking to library ioctl version 2 6157 */ 6158 if (cflag != ZFS_CMD_COMPAT_LZC) 6159 kmem_free(zc, sizeof (zfs_cmd_t)); 6160#endif 6161 return (error); 6162} 6163 6164#ifdef sun 6165static int 6166zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd) 6167{ 6168 if (cmd != DDI_ATTACH) 6169 return (DDI_FAILURE); 6170 6171 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0, 6172 DDI_PSEUDO, 0) == DDI_FAILURE) 6173 return (DDI_FAILURE); 6174 6175 zfs_dip = dip; 6176 6177 ddi_report_dev(dip); 6178 6179 return (DDI_SUCCESS); 6180} 6181 6182static int 6183zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd) 6184{ 6185 if (spa_busy() || zfs_busy() || zvol_busy()) 6186 return (DDI_FAILURE); 6187 6188 if (cmd != DDI_DETACH) 6189 return (DDI_FAILURE); 6190 6191 zfs_dip = NULL; 6192 6193 ddi_prop_remove_all(dip); 6194 ddi_remove_minor_node(dip, NULL); 6195 6196 return (DDI_SUCCESS); 6197} 6198 6199/*ARGSUSED*/ 6200static int 6201zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result) 6202{ 6203 switch (infocmd) { 6204 case DDI_INFO_DEVT2DEVINFO: 6205 *result = zfs_dip; 6206 return (DDI_SUCCESS); 6207 6208 case DDI_INFO_DEVT2INSTANCE: 6209 *result = (void *)0; 6210 return (DDI_SUCCESS); 6211 } 6212 6213 return (DDI_FAILURE); 6214} 6215#endif /* sun */ 6216 6217/* 6218 * OK, so this is a little weird. 6219 * 6220 * /dev/zfs is the control node, i.e. minor 0. 6221 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0. 6222 * 6223 * /dev/zfs has basically nothing to do except serve up ioctls, 6224 * so most of the standard driver entry points are in zvol.c. 6225 */ 6226#ifdef sun 6227static struct cb_ops zfs_cb_ops = { 6228 zfsdev_open, /* open */ 6229 zfsdev_close, /* close */ 6230 zvol_strategy, /* strategy */ 6231 nodev, /* print */ 6232 zvol_dump, /* dump */ 6233 zvol_read, /* read */ 6234 zvol_write, /* write */ 6235 zfsdev_ioctl, /* ioctl */ 6236 nodev, /* devmap */ 6237 nodev, /* mmap */ 6238 nodev, /* segmap */ 6239 nochpoll, /* poll */ 6240 ddi_prop_op, /* prop_op */ 6241 NULL, /* streamtab */ 6242 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */ 6243 CB_REV, /* version */ 6244 nodev, /* async read */ 6245 nodev, /* async write */ 6246}; 6247 6248static struct dev_ops zfs_dev_ops = { 6249 DEVO_REV, /* version */ 6250 0, /* refcnt */ 6251 zfs_info, /* info */ 6252 nulldev, /* identify */ 6253 nulldev, /* probe */ 6254 zfs_attach, /* attach */ 6255 zfs_detach, /* detach */ 6256 nodev, /* reset */ 6257 &zfs_cb_ops, /* driver operations */ 6258 NULL, /* no bus operations */ 6259 NULL, /* power */ 6260 ddi_quiesce_not_needed, /* quiesce */ 6261}; 6262 6263static struct modldrv zfs_modldrv = { 6264 &mod_driverops, 6265 "ZFS storage pool", 6266 &zfs_dev_ops 6267}; 6268 6269static struct modlinkage modlinkage = { 6270 MODREV_1, 6271 (void *)&zfs_modlfs, 6272 (void *)&zfs_modldrv, 6273 NULL 6274}; 6275#endif /* sun */ 6276 6277static struct cdevsw zfs_cdevsw = { 6278 .d_version = D_VERSION, 6279 .d_open = zfsdev_open, 6280 .d_ioctl = zfsdev_ioctl, 6281 .d_name = ZFS_DEV_NAME 6282}; 6283 6284static void 6285zfs_allow_log_destroy(void *arg) 6286{ 6287 char *poolname = arg; 6288 strfree(poolname); 6289} 6290 6291static void 6292zfsdev_init(void) 6293{ 6294 zfsdev = make_dev(&zfs_cdevsw, 0x0, UID_ROOT, GID_OPERATOR, 0666, 6295 ZFS_DEV_NAME); 6296} 6297 6298static void 6299zfsdev_fini(void) 6300{ 6301 if (zfsdev != NULL) 6302 destroy_dev(zfsdev); 6303} 6304 6305static struct root_hold_token *zfs_root_token; 6306struct proc *zfsproc; 6307 6308#ifdef sun 6309int 6310_init(void) 6311{ 6312 int error; 6313 6314 spa_init(FREAD | FWRITE); 6315 zfs_init(); 6316 zvol_init(); 6317 zfs_ioctl_init(); 6318 6319 if ((error = mod_install(&modlinkage)) != 0) { 6320 zvol_fini(); 6321 zfs_fini(); 6322 spa_fini(); 6323 return (error); 6324 } 6325 6326 tsd_create(&zfs_fsyncer_key, NULL); 6327 tsd_create(&rrw_tsd_key, rrw_tsd_destroy); 6328 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy); 6329 6330 error = ldi_ident_from_mod(&modlinkage, &zfs_li); 6331 ASSERT(error == 0); 6332 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL); 6333 6334 return (0); 6335} 6336 6337int 6338_fini(void) 6339{ 6340 int error; 6341 6342 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled) 6343 return (SET_ERROR(EBUSY)); 6344 6345 if ((error = mod_remove(&modlinkage)) != 0) 6346 return (error); 6347 6348 zvol_fini(); 6349 zfs_fini(); 6350 spa_fini(); 6351 if (zfs_nfsshare_inited) 6352 (void) ddi_modclose(nfs_mod); 6353 if (zfs_smbshare_inited) 6354 (void) ddi_modclose(smbsrv_mod); 6355 if (zfs_nfsshare_inited || zfs_smbshare_inited) 6356 (void) ddi_modclose(sharefs_mod); 6357 6358 tsd_destroy(&zfs_fsyncer_key); 6359 ldi_ident_release(zfs_li); 6360 zfs_li = NULL; 6361 mutex_destroy(&zfs_share_lock); 6362 6363 return (error); 6364} 6365 6366int 6367_info(struct modinfo *modinfop) 6368{ 6369 return (mod_info(&modlinkage, modinfop)); 6370} 6371#endif /* sun */ 6372 6373static int zfs__init(void); 6374static int zfs__fini(void); 6375static void zfs_shutdown(void *, int); 6376 6377static eventhandler_tag zfs_shutdown_event_tag; 6378 6379int 6380zfs__init(void) 6381{ 6382 6383 zfs_root_token = root_mount_hold("ZFS"); 6384 6385 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL); 6386 6387 spa_init(FREAD | FWRITE); 6388 zfs_init(); 6389 zvol_init(); 6390 zfs_ioctl_init(); 6391 6392 tsd_create(&zfs_fsyncer_key, NULL); 6393 tsd_create(&rrw_tsd_key, rrw_tsd_destroy); 6394 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy); 6395 6396 printf("ZFS storage pool version: features support (" SPA_VERSION_STRING ")\n"); 6397 root_mount_rel(zfs_root_token); 6398 6399 zfsdev_init(); 6400 6401 return (0); 6402} 6403 6404int 6405zfs__fini(void) 6406{ 6407 if (spa_busy() || zfs_busy() || zvol_busy() || 6408 zio_injection_enabled) { 6409 return (EBUSY); 6410 } 6411 6412 zfsdev_fini(); 6413 zvol_fini(); 6414 zfs_fini(); 6415 spa_fini(); 6416 6417 tsd_destroy(&zfs_fsyncer_key); 6418 tsd_destroy(&rrw_tsd_key); 6419 tsd_destroy(&zfs_allow_log_key); 6420 6421 mutex_destroy(&zfs_share_lock); 6422 6423 return (0); 6424} 6425 6426static void 6427zfs_shutdown(void *arg __unused, int howto __unused) 6428{ 6429 6430 /* 6431 * ZFS fini routines can not properly work in a panic-ed system. 6432 */ 6433 if (panicstr == NULL) 6434 (void)zfs__fini(); 6435} 6436 6437 6438static int 6439zfs_modevent(module_t mod, int type, void *unused __unused) 6440{ 6441 int err; 6442 6443 switch (type) { 6444 case MOD_LOAD: 6445 err = zfs__init(); 6446 if (err == 0) 6447 zfs_shutdown_event_tag = EVENTHANDLER_REGISTER( 6448 shutdown_post_sync, zfs_shutdown, NULL, 6449 SHUTDOWN_PRI_FIRST); 6450 return (err); 6451 case MOD_UNLOAD: 6452 err = zfs__fini(); 6453 if (err == 0 && zfs_shutdown_event_tag != NULL) 6454 EVENTHANDLER_DEREGISTER(shutdown_post_sync, 6455 zfs_shutdown_event_tag); 6456 return (err); 6457 case MOD_SHUTDOWN: 6458 return (0); 6459 default: 6460 break; 6461 } 6462 return (EOPNOTSUPP); 6463} 6464 6465static moduledata_t zfs_mod = { 6466 "zfsctrl", 6467 zfs_modevent, 6468 0 6469}; 6470DECLARE_MODULE(zfsctrl, zfs_mod, SI_SUB_VFS, SI_ORDER_ANY); 6471MODULE_VERSION(zfsctrl, 1); 6472MODULE_DEPEND(zfsctrl, opensolaris, 1, 1, 1); 6473MODULE_DEPEND(zfsctrl, krpc, 1, 1, 1); 6474MODULE_DEPEND(zfsctrl, acl_nfs4, 1, 1, 1); 6475