zfs_ioctl.c revision 265744
1/* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 22/* 23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 24 * Copyright (c) 2011-2012 Pawel Jakub Dawidek <pawel@dawidek.net>. 25 * All rights reserved. 26 * Copyright 2013 Martin Matuska <mm@FreeBSD.org>. All rights reserved. 27 * Copyright 2011 Nexenta Systems, Inc. All rights reserved. 28 * Copyright (c) 2014, Joyent, Inc. All rights reserved. 29 * Copyright (c) 2013 by Delphix. All rights reserved. 30 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved. 31 * Copyright (c) 2013 Steven Hartland. All rights reserved. 32 */ 33 34/* 35 * ZFS ioctls. 36 * 37 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage 38 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool. 39 * 40 * There are two ways that we handle ioctls: the legacy way where almost 41 * all of the logic is in the ioctl callback, and the new way where most 42 * of the marshalling is handled in the common entry point, zfsdev_ioctl(). 43 * 44 * Non-legacy ioctls should be registered by calling 45 * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked 46 * from userland by lzc_ioctl(). 47 * 48 * The registration arguments are as follows: 49 * 50 * const char *name 51 * The name of the ioctl. This is used for history logging. If the 52 * ioctl returns successfully (the callback returns 0), and allow_log 53 * is true, then a history log entry will be recorded with the input & 54 * output nvlists. The log entry can be printed with "zpool history -i". 55 * 56 * zfs_ioc_t ioc 57 * The ioctl request number, which userland will pass to ioctl(2). 58 * The ioctl numbers can change from release to release, because 59 * the caller (libzfs) must be matched to the kernel. 60 * 61 * zfs_secpolicy_func_t *secpolicy 62 * This function will be called before the zfs_ioc_func_t, to 63 * determine if this operation is permitted. It should return EPERM 64 * on failure, and 0 on success. Checks include determining if the 65 * dataset is visible in this zone, and if the user has either all 66 * zfs privileges in the zone (SYS_MOUNT), or has been granted permission 67 * to do this operation on this dataset with "zfs allow". 68 * 69 * zfs_ioc_namecheck_t namecheck 70 * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool 71 * name, a dataset name, or nothing. If the name is not well-formed, 72 * the ioctl will fail and the callback will not be called. 73 * Therefore, the callback can assume that the name is well-formed 74 * (e.g. is null-terminated, doesn't have more than one '@' character, 75 * doesn't have invalid characters). 76 * 77 * zfs_ioc_poolcheck_t pool_check 78 * This specifies requirements on the pool state. If the pool does 79 * not meet them (is suspended or is readonly), the ioctl will fail 80 * and the callback will not be called. If any checks are specified 81 * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME. 82 * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED | 83 * POOL_CHECK_READONLY). 84 * 85 * boolean_t smush_outnvlist 86 * If smush_outnvlist is true, then the output is presumed to be a 87 * list of errors, and it will be "smushed" down to fit into the 88 * caller's buffer, by removing some entries and replacing them with a 89 * single "N_MORE_ERRORS" entry indicating how many were removed. See 90 * nvlist_smush() for details. If smush_outnvlist is false, and the 91 * outnvlist does not fit into the userland-provided buffer, then the 92 * ioctl will fail with ENOMEM. 93 * 94 * zfs_ioc_func_t *func 95 * The callback function that will perform the operation. 96 * 97 * The callback should return 0 on success, or an error number on 98 * failure. If the function fails, the userland ioctl will return -1, 99 * and errno will be set to the callback's return value. The callback 100 * will be called with the following arguments: 101 * 102 * const char *name 103 * The name of the pool or dataset to operate on, from 104 * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the 105 * expected type (pool, dataset, or none). 106 * 107 * nvlist_t *innvl 108 * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or 109 * NULL if no input nvlist was provided. Changes to this nvlist are 110 * ignored. If the input nvlist could not be deserialized, the 111 * ioctl will fail and the callback will not be called. 112 * 113 * nvlist_t *outnvl 114 * The output nvlist, initially empty. The callback can fill it in, 115 * and it will be returned to userland by serializing it into 116 * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization 117 * fails (e.g. because the caller didn't supply a large enough 118 * buffer), then the overall ioctl will fail. See the 119 * 'smush_nvlist' argument above for additional behaviors. 120 * 121 * There are two typical uses of the output nvlist: 122 * - To return state, e.g. property values. In this case, 123 * smush_outnvlist should be false. If the buffer was not large 124 * enough, the caller will reallocate a larger buffer and try 125 * the ioctl again. 126 * 127 * - To return multiple errors from an ioctl which makes on-disk 128 * changes. In this case, smush_outnvlist should be true. 129 * Ioctls which make on-disk modifications should generally not 130 * use the outnvl if they succeed, because the caller can not 131 * distinguish between the operation failing, and 132 * deserialization failing. 133 */ 134 135#include <sys/types.h> 136#include <sys/param.h> 137#include <sys/systm.h> 138#include <sys/conf.h> 139#include <sys/kernel.h> 140#include <sys/lock.h> 141#include <sys/malloc.h> 142#include <sys/mutex.h> 143#include <sys/proc.h> 144#include <sys/errno.h> 145#include <sys/uio.h> 146#include <sys/buf.h> 147#include <sys/file.h> 148#include <sys/kmem.h> 149#include <sys/conf.h> 150#include <sys/cmn_err.h> 151#include <sys/stat.h> 152#include <sys/zfs_ioctl.h> 153#include <sys/zfs_vfsops.h> 154#include <sys/zfs_znode.h> 155#include <sys/zap.h> 156#include <sys/spa.h> 157#include <sys/spa_impl.h> 158#include <sys/vdev.h> 159#include <sys/dmu.h> 160#include <sys/dsl_dir.h> 161#include <sys/dsl_dataset.h> 162#include <sys/dsl_prop.h> 163#include <sys/dsl_deleg.h> 164#include <sys/dmu_objset.h> 165#include <sys/dmu_impl.h> 166#include <sys/dmu_tx.h> 167#include <sys/sunddi.h> 168#include <sys/policy.h> 169#include <sys/zone.h> 170#include <sys/nvpair.h> 171#include <sys/mount.h> 172#include <sys/taskqueue.h> 173#include <sys/sdt.h> 174#include <sys/varargs.h> 175#include <sys/fs/zfs.h> 176#include <sys/zfs_ctldir.h> 177#include <sys/zfs_dir.h> 178#include <sys/zfs_onexit.h> 179#include <sys/zvol.h> 180#include <sys/dsl_scan.h> 181#include <sys/dmu_objset.h> 182#include <sys/dmu_send.h> 183#include <sys/dsl_destroy.h> 184#include <sys/dsl_bookmark.h> 185#include <sys/dsl_userhold.h> 186#include <sys/zfeature.h> 187 188#include "zfs_namecheck.h" 189#include "zfs_prop.h" 190#include "zfs_deleg.h" 191#include "zfs_comutil.h" 192#include "zfs_ioctl_compat.h" 193 194CTASSERT(sizeof(zfs_cmd_t) < IOCPARM_MAX); 195 196static int snapshot_list_prefetch; 197SYSCTL_DECL(_vfs_zfs); 198TUNABLE_INT("vfs.zfs.snapshot_list_prefetch", &snapshot_list_prefetch); 199SYSCTL_INT(_vfs_zfs, OID_AUTO, snapshot_list_prefetch, CTLFLAG_RW, 200 &snapshot_list_prefetch, 0, "Prefetch data when listing snapshots"); 201 202static struct cdev *zfsdev; 203 204extern void zfs_init(void); 205extern void zfs_fini(void); 206 207uint_t zfs_fsyncer_key; 208extern uint_t rrw_tsd_key; 209static uint_t zfs_allow_log_key; 210 211typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *); 212typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *); 213typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *); 214 215typedef enum { 216 NO_NAME, 217 POOL_NAME, 218 DATASET_NAME 219} zfs_ioc_namecheck_t; 220 221typedef enum { 222 POOL_CHECK_NONE = 1 << 0, 223 POOL_CHECK_SUSPENDED = 1 << 1, 224 POOL_CHECK_READONLY = 1 << 2, 225} zfs_ioc_poolcheck_t; 226 227typedef struct zfs_ioc_vec { 228 zfs_ioc_legacy_func_t *zvec_legacy_func; 229 zfs_ioc_func_t *zvec_func; 230 zfs_secpolicy_func_t *zvec_secpolicy; 231 zfs_ioc_namecheck_t zvec_namecheck; 232 boolean_t zvec_allow_log; 233 zfs_ioc_poolcheck_t zvec_pool_check; 234 boolean_t zvec_smush_outnvlist; 235 const char *zvec_name; 236} zfs_ioc_vec_t; 237 238/* This array is indexed by zfs_userquota_prop_t */ 239static const char *userquota_perms[] = { 240 ZFS_DELEG_PERM_USERUSED, 241 ZFS_DELEG_PERM_USERQUOTA, 242 ZFS_DELEG_PERM_GROUPUSED, 243 ZFS_DELEG_PERM_GROUPQUOTA, 244}; 245 246static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc); 247static int zfs_check_settable(const char *name, nvpair_t *property, 248 cred_t *cr); 249static int zfs_check_clearable(char *dataset, nvlist_t *props, 250 nvlist_t **errors); 251static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *, 252 boolean_t *); 253int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *); 254static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp); 255 256static void zfsdev_close(void *data); 257 258static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature); 259 260/* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */ 261void 262__dprintf(const char *file, const char *func, int line, const char *fmt, ...) 263{ 264 const char *newfile; 265 char buf[512]; 266 va_list adx; 267 268 /* 269 * Get rid of annoying "../common/" prefix to filename. 270 */ 271 newfile = strrchr(file, '/'); 272 if (newfile != NULL) { 273 newfile = newfile + 1; /* Get rid of leading / */ 274 } else { 275 newfile = file; 276 } 277 278 va_start(adx, fmt); 279 (void) vsnprintf(buf, sizeof (buf), fmt, adx); 280 va_end(adx); 281 282 /* 283 * To get this data, use the zfs-dprintf probe as so: 284 * dtrace -q -n 'zfs-dprintf \ 285 * /stringof(arg0) == "dbuf.c"/ \ 286 * {printf("%s: %s", stringof(arg1), stringof(arg3))}' 287 * arg0 = file name 288 * arg1 = function name 289 * arg2 = line number 290 * arg3 = message 291 */ 292 DTRACE_PROBE4(zfs__dprintf, 293 char *, newfile, char *, func, int, line, char *, buf); 294} 295 296static void 297history_str_free(char *buf) 298{ 299 kmem_free(buf, HIS_MAX_RECORD_LEN); 300} 301 302static char * 303history_str_get(zfs_cmd_t *zc) 304{ 305 char *buf; 306 307 if (zc->zc_history == 0) 308 return (NULL); 309 310 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP); 311 if (copyinstr((void *)(uintptr_t)zc->zc_history, 312 buf, HIS_MAX_RECORD_LEN, NULL) != 0) { 313 history_str_free(buf); 314 return (NULL); 315 } 316 317 buf[HIS_MAX_RECORD_LEN -1] = '\0'; 318 319 return (buf); 320} 321 322/* 323 * Check to see if the named dataset is currently defined as bootable 324 */ 325static boolean_t 326zfs_is_bootfs(const char *name) 327{ 328 objset_t *os; 329 330 if (dmu_objset_hold(name, FTAG, &os) == 0) { 331 boolean_t ret; 332 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os))); 333 dmu_objset_rele(os, FTAG); 334 return (ret); 335 } 336 return (B_FALSE); 337} 338 339/* 340 * Return non-zero if the spa version is less than requested version. 341 */ 342static int 343zfs_earlier_version(const char *name, int version) 344{ 345 spa_t *spa; 346 347 if (spa_open(name, &spa, FTAG) == 0) { 348 if (spa_version(spa) < version) { 349 spa_close(spa, FTAG); 350 return (1); 351 } 352 spa_close(spa, FTAG); 353 } 354 return (0); 355} 356 357/* 358 * Return TRUE if the ZPL version is less than requested version. 359 */ 360static boolean_t 361zpl_earlier_version(const char *name, int version) 362{ 363 objset_t *os; 364 boolean_t rc = B_TRUE; 365 366 if (dmu_objset_hold(name, FTAG, &os) == 0) { 367 uint64_t zplversion; 368 369 if (dmu_objset_type(os) != DMU_OST_ZFS) { 370 dmu_objset_rele(os, FTAG); 371 return (B_TRUE); 372 } 373 /* XXX reading from non-owned objset */ 374 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0) 375 rc = zplversion < version; 376 dmu_objset_rele(os, FTAG); 377 } 378 return (rc); 379} 380 381static void 382zfs_log_history(zfs_cmd_t *zc) 383{ 384 spa_t *spa; 385 char *buf; 386 387 if ((buf = history_str_get(zc)) == NULL) 388 return; 389 390 if (spa_open(zc->zc_name, &spa, FTAG) == 0) { 391 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY) 392 (void) spa_history_log(spa, buf); 393 spa_close(spa, FTAG); 394 } 395 history_str_free(buf); 396} 397 398/* 399 * Policy for top-level read operations (list pools). Requires no privileges, 400 * and can be used in the local zone, as there is no associated dataset. 401 */ 402/* ARGSUSED */ 403static int 404zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 405{ 406 return (0); 407} 408 409/* 410 * Policy for dataset read operations (list children, get statistics). Requires 411 * no privileges, but must be visible in the local zone. 412 */ 413/* ARGSUSED */ 414static int 415zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 416{ 417 if (INGLOBALZONE(curthread) || 418 zone_dataset_visible(zc->zc_name, NULL)) 419 return (0); 420 421 return (SET_ERROR(ENOENT)); 422} 423 424static int 425zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr) 426{ 427 int writable = 1; 428 429 /* 430 * The dataset must be visible by this zone -- check this first 431 * so they don't see EPERM on something they shouldn't know about. 432 */ 433 if (!INGLOBALZONE(curthread) && 434 !zone_dataset_visible(dataset, &writable)) 435 return (SET_ERROR(ENOENT)); 436 437 if (INGLOBALZONE(curthread)) { 438 /* 439 * If the fs is zoned, only root can access it from the 440 * global zone. 441 */ 442 if (secpolicy_zfs(cr) && zoned) 443 return (SET_ERROR(EPERM)); 444 } else { 445 /* 446 * If we are in a local zone, the 'zoned' property must be set. 447 */ 448 if (!zoned) 449 return (SET_ERROR(EPERM)); 450 451 /* must be writable by this zone */ 452 if (!writable) 453 return (SET_ERROR(EPERM)); 454 } 455 return (0); 456} 457 458static int 459zfs_dozonecheck(const char *dataset, cred_t *cr) 460{ 461 uint64_t zoned; 462 463 if (dsl_prop_get_integer(dataset, "jailed", &zoned, NULL)) 464 return (SET_ERROR(ENOENT)); 465 466 return (zfs_dozonecheck_impl(dataset, zoned, cr)); 467} 468 469static int 470zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr) 471{ 472 uint64_t zoned; 473 474 if (dsl_prop_get_int_ds(ds, "jailed", &zoned)) 475 return (SET_ERROR(ENOENT)); 476 477 return (zfs_dozonecheck_impl(dataset, zoned, cr)); 478} 479 480static int 481zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds, 482 const char *perm, cred_t *cr) 483{ 484 int error; 485 486 error = zfs_dozonecheck_ds(name, ds, cr); 487 if (error == 0) { 488 error = secpolicy_zfs(cr); 489 if (error != 0) 490 error = dsl_deleg_access_impl(ds, perm, cr); 491 } 492 return (error); 493} 494 495static int 496zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr) 497{ 498 int error; 499 dsl_dataset_t *ds; 500 dsl_pool_t *dp; 501 502 error = dsl_pool_hold(name, FTAG, &dp); 503 if (error != 0) 504 return (error); 505 506 error = dsl_dataset_hold(dp, name, FTAG, &ds); 507 if (error != 0) { 508 dsl_pool_rele(dp, FTAG); 509 return (error); 510 } 511 512 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr); 513 514 dsl_dataset_rele(ds, FTAG); 515 dsl_pool_rele(dp, FTAG); 516 return (error); 517} 518 519#ifdef SECLABEL 520/* 521 * Policy for setting the security label property. 522 * 523 * Returns 0 for success, non-zero for access and other errors. 524 */ 525static int 526zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr) 527{ 528 char ds_hexsl[MAXNAMELEN]; 529 bslabel_t ds_sl, new_sl; 530 boolean_t new_default = FALSE; 531 uint64_t zoned; 532 int needed_priv = -1; 533 int error; 534 535 /* First get the existing dataset label. */ 536 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL), 537 1, sizeof (ds_hexsl), &ds_hexsl, NULL); 538 if (error != 0) 539 return (SET_ERROR(EPERM)); 540 541 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0) 542 new_default = TRUE; 543 544 /* The label must be translatable */ 545 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0)) 546 return (SET_ERROR(EINVAL)); 547 548 /* 549 * In a non-global zone, disallow attempts to set a label that 550 * doesn't match that of the zone; otherwise no other checks 551 * are needed. 552 */ 553 if (!INGLOBALZONE(curproc)) { 554 if (new_default || !blequal(&new_sl, CR_SL(CRED()))) 555 return (SET_ERROR(EPERM)); 556 return (0); 557 } 558 559 /* 560 * For global-zone datasets (i.e., those whose zoned property is 561 * "off", verify that the specified new label is valid for the 562 * global zone. 563 */ 564 if (dsl_prop_get_integer(name, 565 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL)) 566 return (SET_ERROR(EPERM)); 567 if (!zoned) { 568 if (zfs_check_global_label(name, strval) != 0) 569 return (SET_ERROR(EPERM)); 570 } 571 572 /* 573 * If the existing dataset label is nondefault, check if the 574 * dataset is mounted (label cannot be changed while mounted). 575 * Get the zfsvfs; if there isn't one, then the dataset isn't 576 * mounted (or isn't a dataset, doesn't exist, ...). 577 */ 578 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) { 579 objset_t *os; 580 static char *setsl_tag = "setsl_tag"; 581 582 /* 583 * Try to own the dataset; abort if there is any error, 584 * (e.g., already mounted, in use, or other error). 585 */ 586 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE, 587 setsl_tag, &os); 588 if (error != 0) 589 return (SET_ERROR(EPERM)); 590 591 dmu_objset_disown(os, setsl_tag); 592 593 if (new_default) { 594 needed_priv = PRIV_FILE_DOWNGRADE_SL; 595 goto out_check; 596 } 597 598 if (hexstr_to_label(strval, &new_sl) != 0) 599 return (SET_ERROR(EPERM)); 600 601 if (blstrictdom(&ds_sl, &new_sl)) 602 needed_priv = PRIV_FILE_DOWNGRADE_SL; 603 else if (blstrictdom(&new_sl, &ds_sl)) 604 needed_priv = PRIV_FILE_UPGRADE_SL; 605 } else { 606 /* dataset currently has a default label */ 607 if (!new_default) 608 needed_priv = PRIV_FILE_UPGRADE_SL; 609 } 610 611out_check: 612 if (needed_priv != -1) 613 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL)); 614 return (0); 615} 616#endif /* SECLABEL */ 617 618static int 619zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval, 620 cred_t *cr) 621{ 622 char *strval; 623 624 /* 625 * Check permissions for special properties. 626 */ 627 switch (prop) { 628 case ZFS_PROP_ZONED: 629 /* 630 * Disallow setting of 'zoned' from within a local zone. 631 */ 632 if (!INGLOBALZONE(curthread)) 633 return (SET_ERROR(EPERM)); 634 break; 635 636 case ZFS_PROP_QUOTA: 637 case ZFS_PROP_FILESYSTEM_LIMIT: 638 case ZFS_PROP_SNAPSHOT_LIMIT: 639 if (!INGLOBALZONE(curthread)) { 640 uint64_t zoned; 641 char setpoint[MAXNAMELEN]; 642 /* 643 * Unprivileged users are allowed to modify the 644 * limit on things *under* (ie. contained by) 645 * the thing they own. 646 */ 647 if (dsl_prop_get_integer(dsname, "jailed", &zoned, 648 setpoint)) 649 return (SET_ERROR(EPERM)); 650 if (!zoned || strlen(dsname) <= strlen(setpoint)) 651 return (SET_ERROR(EPERM)); 652 } 653 break; 654 655 case ZFS_PROP_MLSLABEL: 656#ifdef SECLABEL 657 if (!is_system_labeled()) 658 return (SET_ERROR(EPERM)); 659 660 if (nvpair_value_string(propval, &strval) == 0) { 661 int err; 662 663 err = zfs_set_slabel_policy(dsname, strval, CRED()); 664 if (err != 0) 665 return (err); 666 } 667#else 668 return (EOPNOTSUPP); 669#endif 670 break; 671 } 672 673 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr)); 674} 675 676/* ARGSUSED */ 677static int 678zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 679{ 680 int error; 681 682 error = zfs_dozonecheck(zc->zc_name, cr); 683 if (error != 0) 684 return (error); 685 686 /* 687 * permission to set permissions will be evaluated later in 688 * dsl_deleg_can_allow() 689 */ 690 return (0); 691} 692 693/* ARGSUSED */ 694static int 695zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 696{ 697 return (zfs_secpolicy_write_perms(zc->zc_name, 698 ZFS_DELEG_PERM_ROLLBACK, cr)); 699} 700 701/* ARGSUSED */ 702static int 703zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 704{ 705 dsl_pool_t *dp; 706 dsl_dataset_t *ds; 707 char *cp; 708 int error; 709 710 /* 711 * Generate the current snapshot name from the given objsetid, then 712 * use that name for the secpolicy/zone checks. 713 */ 714 cp = strchr(zc->zc_name, '@'); 715 if (cp == NULL) 716 return (SET_ERROR(EINVAL)); 717 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 718 if (error != 0) 719 return (error); 720 721 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds); 722 if (error != 0) { 723 dsl_pool_rele(dp, FTAG); 724 return (error); 725 } 726 727 dsl_dataset_name(ds, zc->zc_name); 728 729 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds, 730 ZFS_DELEG_PERM_SEND, cr); 731 dsl_dataset_rele(ds, FTAG); 732 dsl_pool_rele(dp, FTAG); 733 734 return (error); 735} 736 737/* ARGSUSED */ 738static int 739zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 740{ 741 return (zfs_secpolicy_write_perms(zc->zc_name, 742 ZFS_DELEG_PERM_SEND, cr)); 743} 744 745/* ARGSUSED */ 746static int 747zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 748{ 749 vnode_t *vp; 750 int error; 751 752 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, 753 NO_FOLLOW, NULL, &vp)) != 0) 754 return (error); 755 756 /* Now make sure mntpnt and dataset are ZFS */ 757 758 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 || 759 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), 760 zc->zc_name) != 0)) { 761 VN_RELE(vp); 762 return (SET_ERROR(EPERM)); 763 } 764 765 VN_RELE(vp); 766 return (dsl_deleg_access(zc->zc_name, 767 ZFS_DELEG_PERM_SHARE, cr)); 768} 769 770int 771zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 772{ 773 if (!INGLOBALZONE(curthread)) 774 return (SET_ERROR(EPERM)); 775 776 if (secpolicy_nfs(cr) == 0) { 777 return (0); 778 } else { 779 return (zfs_secpolicy_deleg_share(zc, innvl, cr)); 780 } 781} 782 783int 784zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 785{ 786 if (!INGLOBALZONE(curthread)) 787 return (SET_ERROR(EPERM)); 788 789 if (secpolicy_smb(cr) == 0) { 790 return (0); 791 } else { 792 return (zfs_secpolicy_deleg_share(zc, innvl, cr)); 793 } 794} 795 796static int 797zfs_get_parent(const char *datasetname, char *parent, int parentsize) 798{ 799 char *cp; 800 801 /* 802 * Remove the @bla or /bla from the end of the name to get the parent. 803 */ 804 (void) strncpy(parent, datasetname, parentsize); 805 cp = strrchr(parent, '@'); 806 if (cp != NULL) { 807 cp[0] = '\0'; 808 } else { 809 cp = strrchr(parent, '/'); 810 if (cp == NULL) 811 return (SET_ERROR(ENOENT)); 812 cp[0] = '\0'; 813 } 814 815 return (0); 816} 817 818int 819zfs_secpolicy_destroy_perms(const char *name, cred_t *cr) 820{ 821 int error; 822 823 if ((error = zfs_secpolicy_write_perms(name, 824 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 825 return (error); 826 827 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr)); 828} 829 830/* ARGSUSED */ 831static int 832zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 833{ 834 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr)); 835} 836 837/* 838 * Destroying snapshots with delegated permissions requires 839 * descendant mount and destroy permissions. 840 */ 841/* ARGSUSED */ 842static int 843zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 844{ 845 nvlist_t *snaps; 846 nvpair_t *pair, *nextpair; 847 int error = 0; 848 849 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 850 return (SET_ERROR(EINVAL)); 851 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 852 pair = nextpair) { 853 nextpair = nvlist_next_nvpair(snaps, pair); 854 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr); 855 if (error == ENOENT) { 856 /* 857 * Ignore any snapshots that don't exist (we consider 858 * them "already destroyed"). Remove the name from the 859 * nvl here in case the snapshot is created between 860 * now and when we try to destroy it (in which case 861 * we don't want to destroy it since we haven't 862 * checked for permission). 863 */ 864 fnvlist_remove_nvpair(snaps, pair); 865 error = 0; 866 } 867 if (error != 0) 868 break; 869 } 870 871 return (error); 872} 873 874int 875zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr) 876{ 877 char parentname[MAXNAMELEN]; 878 int error; 879 880 if ((error = zfs_secpolicy_write_perms(from, 881 ZFS_DELEG_PERM_RENAME, cr)) != 0) 882 return (error); 883 884 if ((error = zfs_secpolicy_write_perms(from, 885 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 886 return (error); 887 888 if ((error = zfs_get_parent(to, parentname, 889 sizeof (parentname))) != 0) 890 return (error); 891 892 if ((error = zfs_secpolicy_write_perms(parentname, 893 ZFS_DELEG_PERM_CREATE, cr)) != 0) 894 return (error); 895 896 if ((error = zfs_secpolicy_write_perms(parentname, 897 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 898 return (error); 899 900 return (error); 901} 902 903/* ARGSUSED */ 904static int 905zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 906{ 907 char *at = NULL; 908 int error; 909 910 if ((zc->zc_cookie & 1) != 0) { 911 /* 912 * This is recursive rename, so the starting snapshot might 913 * not exist. Check file system or volume permission instead. 914 */ 915 at = strchr(zc->zc_name, '@'); 916 if (at == NULL) 917 return (EINVAL); 918 *at = '\0'; 919 } 920 921 error = zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr); 922 923 if (at != NULL) 924 *at = '@'; 925 926 return (error); 927} 928 929/* ARGSUSED */ 930static int 931zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 932{ 933 dsl_pool_t *dp; 934 dsl_dataset_t *clone; 935 int error; 936 937 error = zfs_secpolicy_write_perms(zc->zc_name, 938 ZFS_DELEG_PERM_PROMOTE, cr); 939 if (error != 0) 940 return (error); 941 942 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 943 if (error != 0) 944 return (error); 945 946 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone); 947 948 if (error == 0) { 949 char parentname[MAXNAMELEN]; 950 dsl_dataset_t *origin = NULL; 951 dsl_dir_t *dd; 952 dd = clone->ds_dir; 953 954 error = dsl_dataset_hold_obj(dd->dd_pool, 955 dd->dd_phys->dd_origin_obj, FTAG, &origin); 956 if (error != 0) { 957 dsl_dataset_rele(clone, FTAG); 958 dsl_pool_rele(dp, FTAG); 959 return (error); 960 } 961 962 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone, 963 ZFS_DELEG_PERM_MOUNT, cr); 964 965 dsl_dataset_name(origin, parentname); 966 if (error == 0) { 967 error = zfs_secpolicy_write_perms_ds(parentname, origin, 968 ZFS_DELEG_PERM_PROMOTE, cr); 969 } 970 dsl_dataset_rele(clone, FTAG); 971 dsl_dataset_rele(origin, FTAG); 972 } 973 dsl_pool_rele(dp, FTAG); 974 return (error); 975} 976 977/* ARGSUSED */ 978static int 979zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 980{ 981 int error; 982 983 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 984 ZFS_DELEG_PERM_RECEIVE, cr)) != 0) 985 return (error); 986 987 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 988 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 989 return (error); 990 991 return (zfs_secpolicy_write_perms(zc->zc_name, 992 ZFS_DELEG_PERM_CREATE, cr)); 993} 994 995int 996zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr) 997{ 998 return (zfs_secpolicy_write_perms(name, 999 ZFS_DELEG_PERM_SNAPSHOT, cr)); 1000} 1001 1002/* 1003 * Check for permission to create each snapshot in the nvlist. 1004 */ 1005/* ARGSUSED */ 1006static int 1007zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1008{ 1009 nvlist_t *snaps; 1010 int error; 1011 nvpair_t *pair; 1012 1013 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 1014 return (SET_ERROR(EINVAL)); 1015 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 1016 pair = nvlist_next_nvpair(snaps, pair)) { 1017 char *name = nvpair_name(pair); 1018 char *atp = strchr(name, '@'); 1019 1020 if (atp == NULL) { 1021 error = SET_ERROR(EINVAL); 1022 break; 1023 } 1024 *atp = '\0'; 1025 error = zfs_secpolicy_snapshot_perms(name, cr); 1026 *atp = '@'; 1027 if (error != 0) 1028 break; 1029 } 1030 return (error); 1031} 1032 1033/* 1034 * Check for permission to create each snapshot in the nvlist. 1035 */ 1036/* ARGSUSED */ 1037static int 1038zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1039{ 1040 int error = 0; 1041 1042 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 1043 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 1044 char *name = nvpair_name(pair); 1045 char *hashp = strchr(name, '#'); 1046 1047 if (hashp == NULL) { 1048 error = SET_ERROR(EINVAL); 1049 break; 1050 } 1051 *hashp = '\0'; 1052 error = zfs_secpolicy_write_perms(name, 1053 ZFS_DELEG_PERM_BOOKMARK, cr); 1054 *hashp = '#'; 1055 if (error != 0) 1056 break; 1057 } 1058 return (error); 1059} 1060 1061/* ARGSUSED */ 1062static int 1063zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1064{ 1065 nvpair_t *pair, *nextpair; 1066 int error = 0; 1067 1068 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL; 1069 pair = nextpair) { 1070 char *name = nvpair_name(pair); 1071 char *hashp = strchr(name, '#'); 1072 nextpair = nvlist_next_nvpair(innvl, pair); 1073 1074 if (hashp == NULL) { 1075 error = SET_ERROR(EINVAL); 1076 break; 1077 } 1078 1079 *hashp = '\0'; 1080 error = zfs_secpolicy_write_perms(name, 1081 ZFS_DELEG_PERM_DESTROY, cr); 1082 *hashp = '#'; 1083 if (error == ENOENT) { 1084 /* 1085 * Ignore any filesystems that don't exist (we consider 1086 * their bookmarks "already destroyed"). Remove 1087 * the name from the nvl here in case the filesystem 1088 * is created between now and when we try to destroy 1089 * the bookmark (in which case we don't want to 1090 * destroy it since we haven't checked for permission). 1091 */ 1092 fnvlist_remove_nvpair(innvl, pair); 1093 error = 0; 1094 } 1095 if (error != 0) 1096 break; 1097 } 1098 1099 return (error); 1100} 1101 1102/* ARGSUSED */ 1103static int 1104zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1105{ 1106 /* 1107 * Even root must have a proper TSD so that we know what pool 1108 * to log to. 1109 */ 1110 if (tsd_get(zfs_allow_log_key) == NULL) 1111 return (SET_ERROR(EPERM)); 1112 return (0); 1113} 1114 1115static int 1116zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1117{ 1118 char parentname[MAXNAMELEN]; 1119 int error; 1120 char *origin; 1121 1122 if ((error = zfs_get_parent(zc->zc_name, parentname, 1123 sizeof (parentname))) != 0) 1124 return (error); 1125 1126 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 && 1127 (error = zfs_secpolicy_write_perms(origin, 1128 ZFS_DELEG_PERM_CLONE, cr)) != 0) 1129 return (error); 1130 1131 if ((error = zfs_secpolicy_write_perms(parentname, 1132 ZFS_DELEG_PERM_CREATE, cr)) != 0) 1133 return (error); 1134 1135 return (zfs_secpolicy_write_perms(parentname, 1136 ZFS_DELEG_PERM_MOUNT, cr)); 1137} 1138 1139/* 1140 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires 1141 * SYS_CONFIG privilege, which is not available in a local zone. 1142 */ 1143/* ARGSUSED */ 1144static int 1145zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1146{ 1147 if (secpolicy_sys_config(cr, B_FALSE) != 0) 1148 return (SET_ERROR(EPERM)); 1149 1150 return (0); 1151} 1152 1153/* 1154 * Policy for object to name lookups. 1155 */ 1156/* ARGSUSED */ 1157static int 1158zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1159{ 1160 int error; 1161 1162 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0) 1163 return (0); 1164 1165 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr); 1166 return (error); 1167} 1168 1169/* 1170 * Policy for fault injection. Requires all privileges. 1171 */ 1172/* ARGSUSED */ 1173static int 1174zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1175{ 1176 return (secpolicy_zinject(cr)); 1177} 1178 1179/* ARGSUSED */ 1180static int 1181zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1182{ 1183 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value); 1184 1185 if (prop == ZPROP_INVAL) { 1186 if (!zfs_prop_user(zc->zc_value)) 1187 return (SET_ERROR(EINVAL)); 1188 return (zfs_secpolicy_write_perms(zc->zc_name, 1189 ZFS_DELEG_PERM_USERPROP, cr)); 1190 } else { 1191 return (zfs_secpolicy_setprop(zc->zc_name, prop, 1192 NULL, cr)); 1193 } 1194} 1195 1196static int 1197zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1198{ 1199 int err = zfs_secpolicy_read(zc, innvl, cr); 1200 if (err) 1201 return (err); 1202 1203 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 1204 return (SET_ERROR(EINVAL)); 1205 1206 if (zc->zc_value[0] == 0) { 1207 /* 1208 * They are asking about a posix uid/gid. If it's 1209 * themself, allow it. 1210 */ 1211 if (zc->zc_objset_type == ZFS_PROP_USERUSED || 1212 zc->zc_objset_type == ZFS_PROP_USERQUOTA) { 1213 if (zc->zc_guid == crgetuid(cr)) 1214 return (0); 1215 } else { 1216 if (groupmember(zc->zc_guid, cr)) 1217 return (0); 1218 } 1219 } 1220 1221 return (zfs_secpolicy_write_perms(zc->zc_name, 1222 userquota_perms[zc->zc_objset_type], cr)); 1223} 1224 1225static int 1226zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1227{ 1228 int err = zfs_secpolicy_read(zc, innvl, cr); 1229 if (err) 1230 return (err); 1231 1232 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 1233 return (SET_ERROR(EINVAL)); 1234 1235 return (zfs_secpolicy_write_perms(zc->zc_name, 1236 userquota_perms[zc->zc_objset_type], cr)); 1237} 1238 1239/* ARGSUSED */ 1240static int 1241zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1242{ 1243 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION, 1244 NULL, cr)); 1245} 1246 1247/* ARGSUSED */ 1248static int 1249zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1250{ 1251 nvpair_t *pair; 1252 nvlist_t *holds; 1253 int error; 1254 1255 error = nvlist_lookup_nvlist(innvl, "holds", &holds); 1256 if (error != 0) 1257 return (SET_ERROR(EINVAL)); 1258 1259 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL; 1260 pair = nvlist_next_nvpair(holds, pair)) { 1261 char fsname[MAXNAMELEN]; 1262 error = dmu_fsname(nvpair_name(pair), fsname); 1263 if (error != 0) 1264 return (error); 1265 error = zfs_secpolicy_write_perms(fsname, 1266 ZFS_DELEG_PERM_HOLD, cr); 1267 if (error != 0) 1268 return (error); 1269 } 1270 return (0); 1271} 1272 1273/* ARGSUSED */ 1274static int 1275zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1276{ 1277 nvpair_t *pair; 1278 int error; 1279 1280 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL; 1281 pair = nvlist_next_nvpair(innvl, pair)) { 1282 char fsname[MAXNAMELEN]; 1283 error = dmu_fsname(nvpair_name(pair), fsname); 1284 if (error != 0) 1285 return (error); 1286 error = zfs_secpolicy_write_perms(fsname, 1287 ZFS_DELEG_PERM_RELEASE, cr); 1288 if (error != 0) 1289 return (error); 1290 } 1291 return (0); 1292} 1293 1294/* 1295 * Policy for allowing temporary snapshots to be taken or released 1296 */ 1297static int 1298zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1299{ 1300 /* 1301 * A temporary snapshot is the same as a snapshot, 1302 * hold, destroy and release all rolled into one. 1303 * Delegated diff alone is sufficient that we allow this. 1304 */ 1305 int error; 1306 1307 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 1308 ZFS_DELEG_PERM_DIFF, cr)) == 0) 1309 return (0); 1310 1311 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr); 1312 if (error == 0) 1313 error = zfs_secpolicy_hold(zc, innvl, cr); 1314 if (error == 0) 1315 error = zfs_secpolicy_release(zc, innvl, cr); 1316 if (error == 0) 1317 error = zfs_secpolicy_destroy(zc, innvl, cr); 1318 return (error); 1319} 1320 1321/* 1322 * Returns the nvlist as specified by the user in the zfs_cmd_t. 1323 */ 1324static int 1325get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp) 1326{ 1327 char *packed; 1328 int error; 1329 nvlist_t *list = NULL; 1330 1331 /* 1332 * Read in and unpack the user-supplied nvlist. 1333 */ 1334 if (size == 0) 1335 return (SET_ERROR(EINVAL)); 1336 1337 packed = kmem_alloc(size, KM_SLEEP); 1338 1339 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size, 1340 iflag)) != 0) { 1341 kmem_free(packed, size); 1342 return (error); 1343 } 1344 1345 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) { 1346 kmem_free(packed, size); 1347 return (error); 1348 } 1349 1350 kmem_free(packed, size); 1351 1352 *nvp = list; 1353 return (0); 1354} 1355 1356/* 1357 * Reduce the size of this nvlist until it can be serialized in 'max' bytes. 1358 * Entries will be removed from the end of the nvlist, and one int32 entry 1359 * named "N_MORE_ERRORS" will be added indicating how many entries were 1360 * removed. 1361 */ 1362static int 1363nvlist_smush(nvlist_t *errors, size_t max) 1364{ 1365 size_t size; 1366 1367 size = fnvlist_size(errors); 1368 1369 if (size > max) { 1370 nvpair_t *more_errors; 1371 int n = 0; 1372 1373 if (max < 1024) 1374 return (SET_ERROR(ENOMEM)); 1375 1376 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0); 1377 more_errors = nvlist_prev_nvpair(errors, NULL); 1378 1379 do { 1380 nvpair_t *pair = nvlist_prev_nvpair(errors, 1381 more_errors); 1382 fnvlist_remove_nvpair(errors, pair); 1383 n++; 1384 size = fnvlist_size(errors); 1385 } while (size > max); 1386 1387 fnvlist_remove_nvpair(errors, more_errors); 1388 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n); 1389 ASSERT3U(fnvlist_size(errors), <=, max); 1390 } 1391 1392 return (0); 1393} 1394 1395static int 1396put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl) 1397{ 1398 char *packed = NULL; 1399 int error = 0; 1400 size_t size; 1401 1402 size = fnvlist_size(nvl); 1403 1404 if (size > zc->zc_nvlist_dst_size) { 1405 /* 1406 * Solaris returns ENOMEM here, because even if an error is 1407 * returned from an ioctl(2), new zc_nvlist_dst_size will be 1408 * passed to the userland. This is not the case for FreeBSD. 1409 * We need to return 0, so the kernel will copy the 1410 * zc_nvlist_dst_size back and the userland can discover that a 1411 * bigger buffer is needed. 1412 */ 1413 error = 0; 1414 } else { 1415 packed = fnvlist_pack(nvl, &size); 1416 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst, 1417 size, zc->zc_iflags) != 0) 1418 error = SET_ERROR(EFAULT); 1419 fnvlist_pack_free(packed, size); 1420 } 1421 1422 zc->zc_nvlist_dst_size = size; 1423 zc->zc_nvlist_dst_filled = B_TRUE; 1424 return (error); 1425} 1426 1427static int 1428getzfsvfs(const char *dsname, zfsvfs_t **zfvp) 1429{ 1430 objset_t *os; 1431 int error; 1432 1433 error = dmu_objset_hold(dsname, FTAG, &os); 1434 if (error != 0) 1435 return (error); 1436 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1437 dmu_objset_rele(os, FTAG); 1438 return (SET_ERROR(EINVAL)); 1439 } 1440 1441 mutex_enter(&os->os_user_ptr_lock); 1442 *zfvp = dmu_objset_get_user(os); 1443 if (*zfvp) { 1444 VFS_HOLD((*zfvp)->z_vfs); 1445 } else { 1446 error = SET_ERROR(ESRCH); 1447 } 1448 mutex_exit(&os->os_user_ptr_lock); 1449 dmu_objset_rele(os, FTAG); 1450 return (error); 1451} 1452 1453/* 1454 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which 1455 * case its z_vfs will be NULL, and it will be opened as the owner. 1456 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER, 1457 * which prevents all vnode ops from running. 1458 */ 1459static int 1460zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer) 1461{ 1462 int error = 0; 1463 1464 if (getzfsvfs(name, zfvp) != 0) 1465 error = zfsvfs_create(name, zfvp); 1466 if (error == 0) { 1467 rrw_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER : 1468 RW_READER, tag); 1469 if ((*zfvp)->z_unmounted) { 1470 /* 1471 * XXX we could probably try again, since the unmounting 1472 * thread should be just about to disassociate the 1473 * objset from the zfsvfs. 1474 */ 1475 rrw_exit(&(*zfvp)->z_teardown_lock, tag); 1476 return (SET_ERROR(EBUSY)); 1477 } 1478 } 1479 return (error); 1480} 1481 1482static void 1483zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag) 1484{ 1485 rrw_exit(&zfsvfs->z_teardown_lock, tag); 1486 1487 if (zfsvfs->z_vfs) { 1488 VFS_RELE(zfsvfs->z_vfs); 1489 } else { 1490 dmu_objset_disown(zfsvfs->z_os, zfsvfs); 1491 zfsvfs_free(zfsvfs); 1492 } 1493} 1494 1495static int 1496zfs_ioc_pool_create(zfs_cmd_t *zc) 1497{ 1498 int error; 1499 nvlist_t *config, *props = NULL; 1500 nvlist_t *rootprops = NULL; 1501 nvlist_t *zplprops = NULL; 1502 1503 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1504 zc->zc_iflags, &config)) 1505 return (error); 1506 1507 if (zc->zc_nvlist_src_size != 0 && (error = 1508 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1509 zc->zc_iflags, &props))) { 1510 nvlist_free(config); 1511 return (error); 1512 } 1513 1514 if (props) { 1515 nvlist_t *nvl = NULL; 1516 uint64_t version = SPA_VERSION; 1517 1518 (void) nvlist_lookup_uint64(props, 1519 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version); 1520 if (!SPA_VERSION_IS_SUPPORTED(version)) { 1521 error = SET_ERROR(EINVAL); 1522 goto pool_props_bad; 1523 } 1524 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl); 1525 if (nvl) { 1526 error = nvlist_dup(nvl, &rootprops, KM_SLEEP); 1527 if (error != 0) { 1528 nvlist_free(config); 1529 nvlist_free(props); 1530 return (error); 1531 } 1532 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS); 1533 } 1534 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 1535 error = zfs_fill_zplprops_root(version, rootprops, 1536 zplprops, NULL); 1537 if (error != 0) 1538 goto pool_props_bad; 1539 } 1540 1541 error = spa_create(zc->zc_name, config, props, zplprops); 1542 1543 /* 1544 * Set the remaining root properties 1545 */ 1546 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name, 1547 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0) 1548 (void) spa_destroy(zc->zc_name); 1549 1550pool_props_bad: 1551 nvlist_free(rootprops); 1552 nvlist_free(zplprops); 1553 nvlist_free(config); 1554 nvlist_free(props); 1555 1556 return (error); 1557} 1558 1559static int 1560zfs_ioc_pool_destroy(zfs_cmd_t *zc) 1561{ 1562 int error; 1563 zfs_log_history(zc); 1564 error = spa_destroy(zc->zc_name); 1565 if (error == 0) 1566 zvol_remove_minors(zc->zc_name); 1567 return (error); 1568} 1569 1570static int 1571zfs_ioc_pool_import(zfs_cmd_t *zc) 1572{ 1573 nvlist_t *config, *props = NULL; 1574 uint64_t guid; 1575 int error; 1576 1577 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1578 zc->zc_iflags, &config)) != 0) 1579 return (error); 1580 1581 if (zc->zc_nvlist_src_size != 0 && (error = 1582 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1583 zc->zc_iflags, &props))) { 1584 nvlist_free(config); 1585 return (error); 1586 } 1587 1588 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 || 1589 guid != zc->zc_guid) 1590 error = SET_ERROR(EINVAL); 1591 else 1592 error = spa_import(zc->zc_name, config, props, zc->zc_cookie); 1593 1594 if (zc->zc_nvlist_dst != 0) { 1595 int err; 1596 1597 if ((err = put_nvlist(zc, config)) != 0) 1598 error = err; 1599 } 1600 1601 nvlist_free(config); 1602 1603 if (props) 1604 nvlist_free(props); 1605 1606 return (error); 1607} 1608 1609static int 1610zfs_ioc_pool_export(zfs_cmd_t *zc) 1611{ 1612 int error; 1613 boolean_t force = (boolean_t)zc->zc_cookie; 1614 boolean_t hardforce = (boolean_t)zc->zc_guid; 1615 1616 zfs_log_history(zc); 1617 error = spa_export(zc->zc_name, NULL, force, hardforce); 1618 if (error == 0) 1619 zvol_remove_minors(zc->zc_name); 1620 return (error); 1621} 1622 1623static int 1624zfs_ioc_pool_configs(zfs_cmd_t *zc) 1625{ 1626 nvlist_t *configs; 1627 int error; 1628 1629 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL) 1630 return (SET_ERROR(EEXIST)); 1631 1632 error = put_nvlist(zc, configs); 1633 1634 nvlist_free(configs); 1635 1636 return (error); 1637} 1638 1639/* 1640 * inputs: 1641 * zc_name name of the pool 1642 * 1643 * outputs: 1644 * zc_cookie real errno 1645 * zc_nvlist_dst config nvlist 1646 * zc_nvlist_dst_size size of config nvlist 1647 */ 1648static int 1649zfs_ioc_pool_stats(zfs_cmd_t *zc) 1650{ 1651 nvlist_t *config; 1652 int error; 1653 int ret = 0; 1654 1655 error = spa_get_stats(zc->zc_name, &config, zc->zc_value, 1656 sizeof (zc->zc_value)); 1657 1658 if (config != NULL) { 1659 ret = put_nvlist(zc, config); 1660 nvlist_free(config); 1661 1662 /* 1663 * The config may be present even if 'error' is non-zero. 1664 * In this case we return success, and preserve the real errno 1665 * in 'zc_cookie'. 1666 */ 1667 zc->zc_cookie = error; 1668 } else { 1669 ret = error; 1670 } 1671 1672 return (ret); 1673} 1674 1675/* 1676 * Try to import the given pool, returning pool stats as appropriate so that 1677 * user land knows which devices are available and overall pool health. 1678 */ 1679static int 1680zfs_ioc_pool_tryimport(zfs_cmd_t *zc) 1681{ 1682 nvlist_t *tryconfig, *config; 1683 int error; 1684 1685 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1686 zc->zc_iflags, &tryconfig)) != 0) 1687 return (error); 1688 1689 config = spa_tryimport(tryconfig); 1690 1691 nvlist_free(tryconfig); 1692 1693 if (config == NULL) 1694 return (SET_ERROR(EINVAL)); 1695 1696 error = put_nvlist(zc, config); 1697 nvlist_free(config); 1698 1699 return (error); 1700} 1701 1702/* 1703 * inputs: 1704 * zc_name name of the pool 1705 * zc_cookie scan func (pool_scan_func_t) 1706 */ 1707static int 1708zfs_ioc_pool_scan(zfs_cmd_t *zc) 1709{ 1710 spa_t *spa; 1711 int error; 1712 1713 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1714 return (error); 1715 1716 if (zc->zc_cookie == POOL_SCAN_NONE) 1717 error = spa_scan_stop(spa); 1718 else 1719 error = spa_scan(spa, zc->zc_cookie); 1720 1721 spa_close(spa, FTAG); 1722 1723 return (error); 1724} 1725 1726static int 1727zfs_ioc_pool_freeze(zfs_cmd_t *zc) 1728{ 1729 spa_t *spa; 1730 int error; 1731 1732 error = spa_open(zc->zc_name, &spa, FTAG); 1733 if (error == 0) { 1734 spa_freeze(spa); 1735 spa_close(spa, FTAG); 1736 } 1737 return (error); 1738} 1739 1740static int 1741zfs_ioc_pool_upgrade(zfs_cmd_t *zc) 1742{ 1743 spa_t *spa; 1744 int error; 1745 1746 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1747 return (error); 1748 1749 if (zc->zc_cookie < spa_version(spa) || 1750 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) { 1751 spa_close(spa, FTAG); 1752 return (SET_ERROR(EINVAL)); 1753 } 1754 1755 spa_upgrade(spa, zc->zc_cookie); 1756 spa_close(spa, FTAG); 1757 1758 return (error); 1759} 1760 1761static int 1762zfs_ioc_pool_get_history(zfs_cmd_t *zc) 1763{ 1764 spa_t *spa; 1765 char *hist_buf; 1766 uint64_t size; 1767 int error; 1768 1769 if ((size = zc->zc_history_len) == 0) 1770 return (SET_ERROR(EINVAL)); 1771 1772 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1773 return (error); 1774 1775 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) { 1776 spa_close(spa, FTAG); 1777 return (SET_ERROR(ENOTSUP)); 1778 } 1779 1780 hist_buf = kmem_alloc(size, KM_SLEEP); 1781 if ((error = spa_history_get(spa, &zc->zc_history_offset, 1782 &zc->zc_history_len, hist_buf)) == 0) { 1783 error = ddi_copyout(hist_buf, 1784 (void *)(uintptr_t)zc->zc_history, 1785 zc->zc_history_len, zc->zc_iflags); 1786 } 1787 1788 spa_close(spa, FTAG); 1789 kmem_free(hist_buf, size); 1790 return (error); 1791} 1792 1793static int 1794zfs_ioc_pool_reguid(zfs_cmd_t *zc) 1795{ 1796 spa_t *spa; 1797 int error; 1798 1799 error = spa_open(zc->zc_name, &spa, FTAG); 1800 if (error == 0) { 1801 error = spa_change_guid(spa); 1802 spa_close(spa, FTAG); 1803 } 1804 return (error); 1805} 1806 1807static int 1808zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc) 1809{ 1810 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value)); 1811} 1812 1813/* 1814 * inputs: 1815 * zc_name name of filesystem 1816 * zc_obj object to find 1817 * 1818 * outputs: 1819 * zc_value name of object 1820 */ 1821static int 1822zfs_ioc_obj_to_path(zfs_cmd_t *zc) 1823{ 1824 objset_t *os; 1825 int error; 1826 1827 /* XXX reading from objset not owned */ 1828 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0) 1829 return (error); 1830 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1831 dmu_objset_rele(os, FTAG); 1832 return (SET_ERROR(EINVAL)); 1833 } 1834 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value, 1835 sizeof (zc->zc_value)); 1836 dmu_objset_rele(os, FTAG); 1837 1838 return (error); 1839} 1840 1841/* 1842 * inputs: 1843 * zc_name name of filesystem 1844 * zc_obj object to find 1845 * 1846 * outputs: 1847 * zc_stat stats on object 1848 * zc_value path to object 1849 */ 1850static int 1851zfs_ioc_obj_to_stats(zfs_cmd_t *zc) 1852{ 1853 objset_t *os; 1854 int error; 1855 1856 /* XXX reading from objset not owned */ 1857 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0) 1858 return (error); 1859 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1860 dmu_objset_rele(os, FTAG); 1861 return (SET_ERROR(EINVAL)); 1862 } 1863 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value, 1864 sizeof (zc->zc_value)); 1865 dmu_objset_rele(os, FTAG); 1866 1867 return (error); 1868} 1869 1870static int 1871zfs_ioc_vdev_add(zfs_cmd_t *zc) 1872{ 1873 spa_t *spa; 1874 int error; 1875 nvlist_t *config, **l2cache, **spares; 1876 uint_t nl2cache = 0, nspares = 0; 1877 1878 error = spa_open(zc->zc_name, &spa, FTAG); 1879 if (error != 0) 1880 return (error); 1881 1882 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1883 zc->zc_iflags, &config); 1884 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE, 1885 &l2cache, &nl2cache); 1886 1887 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES, 1888 &spares, &nspares); 1889 1890#ifdef illumos 1891 /* 1892 * A root pool with concatenated devices is not supported. 1893 * Thus, can not add a device to a root pool. 1894 * 1895 * Intent log device can not be added to a rootpool because 1896 * during mountroot, zil is replayed, a seperated log device 1897 * can not be accessed during the mountroot time. 1898 * 1899 * l2cache and spare devices are ok to be added to a rootpool. 1900 */ 1901 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) { 1902 nvlist_free(config); 1903 spa_close(spa, FTAG); 1904 return (SET_ERROR(EDOM)); 1905 } 1906#endif /* illumos */ 1907 1908 if (error == 0) { 1909 error = spa_vdev_add(spa, config); 1910 nvlist_free(config); 1911 } 1912 spa_close(spa, FTAG); 1913 return (error); 1914} 1915 1916/* 1917 * inputs: 1918 * zc_name name of the pool 1919 * zc_nvlist_conf nvlist of devices to remove 1920 * zc_cookie to stop the remove? 1921 */ 1922static int 1923zfs_ioc_vdev_remove(zfs_cmd_t *zc) 1924{ 1925 spa_t *spa; 1926 int error; 1927 1928 error = spa_open(zc->zc_name, &spa, FTAG); 1929 if (error != 0) 1930 return (error); 1931 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE); 1932 spa_close(spa, FTAG); 1933 return (error); 1934} 1935 1936static int 1937zfs_ioc_vdev_set_state(zfs_cmd_t *zc) 1938{ 1939 spa_t *spa; 1940 int error; 1941 vdev_state_t newstate = VDEV_STATE_UNKNOWN; 1942 1943 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1944 return (error); 1945 switch (zc->zc_cookie) { 1946 case VDEV_STATE_ONLINE: 1947 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate); 1948 break; 1949 1950 case VDEV_STATE_OFFLINE: 1951 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj); 1952 break; 1953 1954 case VDEV_STATE_FAULTED: 1955 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED && 1956 zc->zc_obj != VDEV_AUX_EXTERNAL) 1957 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED; 1958 1959 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj); 1960 break; 1961 1962 case VDEV_STATE_DEGRADED: 1963 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED && 1964 zc->zc_obj != VDEV_AUX_EXTERNAL) 1965 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED; 1966 1967 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj); 1968 break; 1969 1970 default: 1971 error = SET_ERROR(EINVAL); 1972 } 1973 zc->zc_cookie = newstate; 1974 spa_close(spa, FTAG); 1975 return (error); 1976} 1977 1978static int 1979zfs_ioc_vdev_attach(zfs_cmd_t *zc) 1980{ 1981 spa_t *spa; 1982 int replacing = zc->zc_cookie; 1983 nvlist_t *config; 1984 int error; 1985 1986 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1987 return (error); 1988 1989 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1990 zc->zc_iflags, &config)) == 0) { 1991 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing); 1992 nvlist_free(config); 1993 } 1994 1995 spa_close(spa, FTAG); 1996 return (error); 1997} 1998 1999static int 2000zfs_ioc_vdev_detach(zfs_cmd_t *zc) 2001{ 2002 spa_t *spa; 2003 int error; 2004 2005 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 2006 return (error); 2007 2008 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE); 2009 2010 spa_close(spa, FTAG); 2011 return (error); 2012} 2013 2014static int 2015zfs_ioc_vdev_split(zfs_cmd_t *zc) 2016{ 2017 spa_t *spa; 2018 nvlist_t *config, *props = NULL; 2019 int error; 2020 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT); 2021 2022 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 2023 return (error); 2024 2025 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 2026 zc->zc_iflags, &config)) { 2027 spa_close(spa, FTAG); 2028 return (error); 2029 } 2030 2031 if (zc->zc_nvlist_src_size != 0 && (error = 2032 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2033 zc->zc_iflags, &props))) { 2034 spa_close(spa, FTAG); 2035 nvlist_free(config); 2036 return (error); 2037 } 2038 2039 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp); 2040 2041 spa_close(spa, FTAG); 2042 2043 nvlist_free(config); 2044 nvlist_free(props); 2045 2046 return (error); 2047} 2048 2049static int 2050zfs_ioc_vdev_setpath(zfs_cmd_t *zc) 2051{ 2052 spa_t *spa; 2053 char *path = zc->zc_value; 2054 uint64_t guid = zc->zc_guid; 2055 int error; 2056 2057 error = spa_open(zc->zc_name, &spa, FTAG); 2058 if (error != 0) 2059 return (error); 2060 2061 error = spa_vdev_setpath(spa, guid, path); 2062 spa_close(spa, FTAG); 2063 return (error); 2064} 2065 2066static int 2067zfs_ioc_vdev_setfru(zfs_cmd_t *zc) 2068{ 2069 spa_t *spa; 2070 char *fru = zc->zc_value; 2071 uint64_t guid = zc->zc_guid; 2072 int error; 2073 2074 error = spa_open(zc->zc_name, &spa, FTAG); 2075 if (error != 0) 2076 return (error); 2077 2078 error = spa_vdev_setfru(spa, guid, fru); 2079 spa_close(spa, FTAG); 2080 return (error); 2081} 2082 2083static int 2084zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os) 2085{ 2086 int error = 0; 2087 nvlist_t *nv; 2088 2089 dmu_objset_fast_stat(os, &zc->zc_objset_stats); 2090 2091 if (zc->zc_nvlist_dst != 0 && 2092 (error = dsl_prop_get_all(os, &nv)) == 0) { 2093 dmu_objset_stats(os, nv); 2094 /* 2095 * NB: zvol_get_stats() will read the objset contents, 2096 * which we aren't supposed to do with a 2097 * DS_MODE_USER hold, because it could be 2098 * inconsistent. So this is a bit of a workaround... 2099 * XXX reading with out owning 2100 */ 2101 if (!zc->zc_objset_stats.dds_inconsistent && 2102 dmu_objset_type(os) == DMU_OST_ZVOL) { 2103 error = zvol_get_stats(os, nv); 2104 if (error == EIO) 2105 return (error); 2106 VERIFY0(error); 2107 } 2108 error = put_nvlist(zc, nv); 2109 nvlist_free(nv); 2110 } 2111 2112 return (error); 2113} 2114 2115/* 2116 * inputs: 2117 * zc_name name of filesystem 2118 * zc_nvlist_dst_size size of buffer for property nvlist 2119 * 2120 * outputs: 2121 * zc_objset_stats stats 2122 * zc_nvlist_dst property nvlist 2123 * zc_nvlist_dst_size size of property nvlist 2124 */ 2125static int 2126zfs_ioc_objset_stats(zfs_cmd_t *zc) 2127{ 2128 objset_t *os; 2129 int error; 2130 2131 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 2132 if (error == 0) { 2133 error = zfs_ioc_objset_stats_impl(zc, os); 2134 dmu_objset_rele(os, FTAG); 2135 } 2136 2137 if (error == ENOMEM) 2138 error = 0; 2139 return (error); 2140} 2141 2142/* 2143 * inputs: 2144 * zc_name name of filesystem 2145 * zc_nvlist_dst_size size of buffer for property nvlist 2146 * 2147 * outputs: 2148 * zc_nvlist_dst received property nvlist 2149 * zc_nvlist_dst_size size of received property nvlist 2150 * 2151 * Gets received properties (distinct from local properties on or after 2152 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from 2153 * local property values. 2154 */ 2155static int 2156zfs_ioc_objset_recvd_props(zfs_cmd_t *zc) 2157{ 2158 int error = 0; 2159 nvlist_t *nv; 2160 2161 /* 2162 * Without this check, we would return local property values if the 2163 * caller has not already received properties on or after 2164 * SPA_VERSION_RECVD_PROPS. 2165 */ 2166 if (!dsl_prop_get_hasrecvd(zc->zc_name)) 2167 return (SET_ERROR(ENOTSUP)); 2168 2169 if (zc->zc_nvlist_dst != 0 && 2170 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) { 2171 error = put_nvlist(zc, nv); 2172 nvlist_free(nv); 2173 } 2174 2175 return (error); 2176} 2177 2178static int 2179nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop) 2180{ 2181 uint64_t value; 2182 int error; 2183 2184 /* 2185 * zfs_get_zplprop() will either find a value or give us 2186 * the default value (if there is one). 2187 */ 2188 if ((error = zfs_get_zplprop(os, prop, &value)) != 0) 2189 return (error); 2190 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0); 2191 return (0); 2192} 2193 2194/* 2195 * inputs: 2196 * zc_name name of filesystem 2197 * zc_nvlist_dst_size size of buffer for zpl property nvlist 2198 * 2199 * outputs: 2200 * zc_nvlist_dst zpl property nvlist 2201 * zc_nvlist_dst_size size of zpl property nvlist 2202 */ 2203static int 2204zfs_ioc_objset_zplprops(zfs_cmd_t *zc) 2205{ 2206 objset_t *os; 2207 int err; 2208 2209 /* XXX reading without owning */ 2210 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os)) 2211 return (err); 2212 2213 dmu_objset_fast_stat(os, &zc->zc_objset_stats); 2214 2215 /* 2216 * NB: nvl_add_zplprop() will read the objset contents, 2217 * which we aren't supposed to do with a DS_MODE_USER 2218 * hold, because it could be inconsistent. 2219 */ 2220 if (zc->zc_nvlist_dst != 0 && 2221 !zc->zc_objset_stats.dds_inconsistent && 2222 dmu_objset_type(os) == DMU_OST_ZFS) { 2223 nvlist_t *nv; 2224 2225 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2226 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 && 2227 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 && 2228 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 && 2229 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0) 2230 err = put_nvlist(zc, nv); 2231 nvlist_free(nv); 2232 } else { 2233 err = SET_ERROR(ENOENT); 2234 } 2235 dmu_objset_rele(os, FTAG); 2236 return (err); 2237} 2238 2239boolean_t 2240dataset_name_hidden(const char *name) 2241{ 2242 /* 2243 * Skip over datasets that are not visible in this zone, 2244 * internal datasets (which have a $ in their name), and 2245 * temporary datasets (which have a % in their name). 2246 */ 2247 if (strchr(name, '$') != NULL) 2248 return (B_TRUE); 2249 if (strchr(name, '%') != NULL) 2250 return (B_TRUE); 2251 if (!INGLOBALZONE(curthread) && !zone_dataset_visible(name, NULL)) 2252 return (B_TRUE); 2253 return (B_FALSE); 2254} 2255 2256/* 2257 * inputs: 2258 * zc_name name of filesystem 2259 * zc_cookie zap cursor 2260 * zc_nvlist_dst_size size of buffer for property nvlist 2261 * 2262 * outputs: 2263 * zc_name name of next filesystem 2264 * zc_cookie zap cursor 2265 * zc_objset_stats stats 2266 * zc_nvlist_dst property nvlist 2267 * zc_nvlist_dst_size size of property nvlist 2268 */ 2269static int 2270zfs_ioc_dataset_list_next(zfs_cmd_t *zc) 2271{ 2272 objset_t *os; 2273 int error; 2274 char *p; 2275 size_t orig_len = strlen(zc->zc_name); 2276 2277top: 2278 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) { 2279 if (error == ENOENT) 2280 error = SET_ERROR(ESRCH); 2281 return (error); 2282 } 2283 2284 p = strrchr(zc->zc_name, '/'); 2285 if (p == NULL || p[1] != '\0') 2286 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name)); 2287 p = zc->zc_name + strlen(zc->zc_name); 2288 2289 do { 2290 error = dmu_dir_list_next(os, 2291 sizeof (zc->zc_name) - (p - zc->zc_name), p, 2292 NULL, &zc->zc_cookie); 2293 if (error == ENOENT) 2294 error = SET_ERROR(ESRCH); 2295 } while (error == 0 && dataset_name_hidden(zc->zc_name)); 2296 dmu_objset_rele(os, FTAG); 2297 2298 /* 2299 * If it's an internal dataset (ie. with a '$' in its name), 2300 * don't try to get stats for it, otherwise we'll return ENOENT. 2301 */ 2302 if (error == 0 && strchr(zc->zc_name, '$') == NULL) { 2303 error = zfs_ioc_objset_stats(zc); /* fill in the stats */ 2304 if (error == ENOENT) { 2305 /* We lost a race with destroy, get the next one. */ 2306 zc->zc_name[orig_len] = '\0'; 2307 goto top; 2308 } 2309 } 2310 return (error); 2311} 2312 2313/* 2314 * inputs: 2315 * zc_name name of filesystem 2316 * zc_cookie zap cursor 2317 * zc_nvlist_dst_size size of buffer for property nvlist 2318 * zc_simple when set, only name is requested 2319 * 2320 * outputs: 2321 * zc_name name of next snapshot 2322 * zc_objset_stats stats 2323 * zc_nvlist_dst property nvlist 2324 * zc_nvlist_dst_size size of property nvlist 2325 */ 2326static int 2327zfs_ioc_snapshot_list_next(zfs_cmd_t *zc) 2328{ 2329 objset_t *os; 2330 int error; 2331 2332 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 2333 if (error != 0) { 2334 return (error == ENOENT ? ESRCH : error); 2335 } 2336 2337 /* 2338 * A dataset name of maximum length cannot have any snapshots, 2339 * so exit immediately. 2340 */ 2341 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) { 2342 dmu_objset_rele(os, FTAG); 2343 return (SET_ERROR(ESRCH)); 2344 } 2345 2346 error = dmu_snapshot_list_next(os, 2347 sizeof (zc->zc_name) - strlen(zc->zc_name), 2348 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie, 2349 NULL); 2350 2351 if (error == 0 && !zc->zc_simple) { 2352 dsl_dataset_t *ds; 2353 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool; 2354 2355 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds); 2356 if (error == 0) { 2357 objset_t *ossnap; 2358 2359 error = dmu_objset_from_ds(ds, &ossnap); 2360 if (error == 0) 2361 error = zfs_ioc_objset_stats_impl(zc, ossnap); 2362 dsl_dataset_rele(ds, FTAG); 2363 } 2364 } else if (error == ENOENT) { 2365 error = SET_ERROR(ESRCH); 2366 } 2367 2368 dmu_objset_rele(os, FTAG); 2369 /* if we failed, undo the @ that we tacked on to zc_name */ 2370 if (error != 0) 2371 *strchr(zc->zc_name, '@') = '\0'; 2372 return (error); 2373} 2374 2375static int 2376zfs_prop_set_userquota(const char *dsname, nvpair_t *pair) 2377{ 2378 const char *propname = nvpair_name(pair); 2379 uint64_t *valary; 2380 unsigned int vallen; 2381 const char *domain; 2382 char *dash; 2383 zfs_userquota_prop_t type; 2384 uint64_t rid; 2385 uint64_t quota; 2386 zfsvfs_t *zfsvfs; 2387 int err; 2388 2389 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2390 nvlist_t *attrs; 2391 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 2392 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2393 &pair) != 0) 2394 return (SET_ERROR(EINVAL)); 2395 } 2396 2397 /* 2398 * A correctly constructed propname is encoded as 2399 * userquota@<rid>-<domain>. 2400 */ 2401 if ((dash = strchr(propname, '-')) == NULL || 2402 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 || 2403 vallen != 3) 2404 return (SET_ERROR(EINVAL)); 2405 2406 domain = dash + 1; 2407 type = valary[0]; 2408 rid = valary[1]; 2409 quota = valary[2]; 2410 2411 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE); 2412 if (err == 0) { 2413 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota); 2414 zfsvfs_rele(zfsvfs, FTAG); 2415 } 2416 2417 return (err); 2418} 2419 2420/* 2421 * If the named property is one that has a special function to set its value, 2422 * return 0 on success and a positive error code on failure; otherwise if it is 2423 * not one of the special properties handled by this function, return -1. 2424 * 2425 * XXX: It would be better for callers of the property interface if we handled 2426 * these special cases in dsl_prop.c (in the dsl layer). 2427 */ 2428static int 2429zfs_prop_set_special(const char *dsname, zprop_source_t source, 2430 nvpair_t *pair) 2431{ 2432 const char *propname = nvpair_name(pair); 2433 zfs_prop_t prop = zfs_name_to_prop(propname); 2434 uint64_t intval; 2435 int err; 2436 2437 if (prop == ZPROP_INVAL) { 2438 if (zfs_prop_userquota(propname)) 2439 return (zfs_prop_set_userquota(dsname, pair)); 2440 return (-1); 2441 } 2442 2443 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2444 nvlist_t *attrs; 2445 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 2446 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2447 &pair) == 0); 2448 } 2449 2450 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING) 2451 return (-1); 2452 2453 VERIFY(0 == nvpair_value_uint64(pair, &intval)); 2454 2455 switch (prop) { 2456 case ZFS_PROP_QUOTA: 2457 err = dsl_dir_set_quota(dsname, source, intval); 2458 break; 2459 case ZFS_PROP_REFQUOTA: 2460 err = dsl_dataset_set_refquota(dsname, source, intval); 2461 break; 2462 case ZFS_PROP_FILESYSTEM_LIMIT: 2463 case ZFS_PROP_SNAPSHOT_LIMIT: 2464 if (intval == UINT64_MAX) { 2465 /* clearing the limit, just do it */ 2466 err = 0; 2467 } else { 2468 err = dsl_dir_activate_fs_ss_limit(dsname); 2469 } 2470 /* 2471 * Set err to -1 to force the zfs_set_prop_nvlist code down the 2472 * default path to set the value in the nvlist. 2473 */ 2474 if (err == 0) 2475 err = -1; 2476 break; 2477 case ZFS_PROP_RESERVATION: 2478 err = dsl_dir_set_reservation(dsname, source, intval); 2479 break; 2480 case ZFS_PROP_REFRESERVATION: 2481 err = dsl_dataset_set_refreservation(dsname, source, intval); 2482 break; 2483 case ZFS_PROP_VOLSIZE: 2484 err = zvol_set_volsize(dsname, ddi_driver_major(zfs_dip), 2485 intval); 2486 break; 2487 case ZFS_PROP_VERSION: 2488 { 2489 zfsvfs_t *zfsvfs; 2490 2491 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0) 2492 break; 2493 2494 err = zfs_set_version(zfsvfs, intval); 2495 zfsvfs_rele(zfsvfs, FTAG); 2496 2497 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) { 2498 zfs_cmd_t *zc; 2499 2500 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP); 2501 (void) strcpy(zc->zc_name, dsname); 2502 (void) zfs_ioc_userspace_upgrade(zc); 2503 kmem_free(zc, sizeof (zfs_cmd_t)); 2504 } 2505 break; 2506 } 2507 case ZFS_PROP_COMPRESSION: 2508 { 2509 if (intval == ZIO_COMPRESS_LZ4) { 2510 spa_t *spa; 2511 2512 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 2513 return (err); 2514 2515 /* 2516 * Setting the LZ4 compression algorithm activates 2517 * the feature. 2518 */ 2519 if (!spa_feature_is_active(spa, 2520 SPA_FEATURE_LZ4_COMPRESS)) { 2521 if ((err = zfs_prop_activate_feature(spa, 2522 SPA_FEATURE_LZ4_COMPRESS)) != 0) { 2523 spa_close(spa, FTAG); 2524 return (err); 2525 } 2526 } 2527 2528 spa_close(spa, FTAG); 2529 } 2530 /* 2531 * We still want the default set action to be performed in the 2532 * caller, we only performed zfeature settings here. 2533 */ 2534 err = -1; 2535 break; 2536 } 2537 2538 default: 2539 err = -1; 2540 } 2541 2542 return (err); 2543} 2544 2545/* 2546 * This function is best effort. If it fails to set any of the given properties, 2547 * it continues to set as many as it can and returns the last error 2548 * encountered. If the caller provides a non-NULL errlist, it will be filled in 2549 * with the list of names of all the properties that failed along with the 2550 * corresponding error numbers. 2551 * 2552 * If every property is set successfully, zero is returned and errlist is not 2553 * modified. 2554 */ 2555int 2556zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl, 2557 nvlist_t *errlist) 2558{ 2559 nvpair_t *pair; 2560 nvpair_t *propval; 2561 int rv = 0; 2562 uint64_t intval; 2563 char *strval; 2564 nvlist_t *genericnvl = fnvlist_alloc(); 2565 nvlist_t *retrynvl = fnvlist_alloc(); 2566 2567retry: 2568 pair = NULL; 2569 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { 2570 const char *propname = nvpair_name(pair); 2571 zfs_prop_t prop = zfs_name_to_prop(propname); 2572 int err = 0; 2573 2574 /* decode the property value */ 2575 propval = pair; 2576 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2577 nvlist_t *attrs; 2578 attrs = fnvpair_value_nvlist(pair); 2579 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2580 &propval) != 0) 2581 err = SET_ERROR(EINVAL); 2582 } 2583 2584 /* Validate value type */ 2585 if (err == 0 && prop == ZPROP_INVAL) { 2586 if (zfs_prop_user(propname)) { 2587 if (nvpair_type(propval) != DATA_TYPE_STRING) 2588 err = SET_ERROR(EINVAL); 2589 } else if (zfs_prop_userquota(propname)) { 2590 if (nvpair_type(propval) != 2591 DATA_TYPE_UINT64_ARRAY) 2592 err = SET_ERROR(EINVAL); 2593 } else { 2594 err = SET_ERROR(EINVAL); 2595 } 2596 } else if (err == 0) { 2597 if (nvpair_type(propval) == DATA_TYPE_STRING) { 2598 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING) 2599 err = SET_ERROR(EINVAL); 2600 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) { 2601 const char *unused; 2602 2603 intval = fnvpair_value_uint64(propval); 2604 2605 switch (zfs_prop_get_type(prop)) { 2606 case PROP_TYPE_NUMBER: 2607 break; 2608 case PROP_TYPE_STRING: 2609 err = SET_ERROR(EINVAL); 2610 break; 2611 case PROP_TYPE_INDEX: 2612 if (zfs_prop_index_to_string(prop, 2613 intval, &unused) != 0) 2614 err = SET_ERROR(EINVAL); 2615 break; 2616 default: 2617 cmn_err(CE_PANIC, 2618 "unknown property type"); 2619 } 2620 } else { 2621 err = SET_ERROR(EINVAL); 2622 } 2623 } 2624 2625 /* Validate permissions */ 2626 if (err == 0) 2627 err = zfs_check_settable(dsname, pair, CRED()); 2628 2629 if (err == 0) { 2630 err = zfs_prop_set_special(dsname, source, pair); 2631 if (err == -1) { 2632 /* 2633 * For better performance we build up a list of 2634 * properties to set in a single transaction. 2635 */ 2636 err = nvlist_add_nvpair(genericnvl, pair); 2637 } else if (err != 0 && nvl != retrynvl) { 2638 /* 2639 * This may be a spurious error caused by 2640 * receiving quota and reservation out of order. 2641 * Try again in a second pass. 2642 */ 2643 err = nvlist_add_nvpair(retrynvl, pair); 2644 } 2645 } 2646 2647 if (err != 0) { 2648 if (errlist != NULL) 2649 fnvlist_add_int32(errlist, propname, err); 2650 rv = err; 2651 } 2652 } 2653 2654 if (nvl != retrynvl && !nvlist_empty(retrynvl)) { 2655 nvl = retrynvl; 2656 goto retry; 2657 } 2658 2659 if (!nvlist_empty(genericnvl) && 2660 dsl_props_set(dsname, source, genericnvl) != 0) { 2661 /* 2662 * If this fails, we still want to set as many properties as we 2663 * can, so try setting them individually. 2664 */ 2665 pair = NULL; 2666 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) { 2667 const char *propname = nvpair_name(pair); 2668 int err = 0; 2669 2670 propval = pair; 2671 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2672 nvlist_t *attrs; 2673 attrs = fnvpair_value_nvlist(pair); 2674 propval = fnvlist_lookup_nvpair(attrs, 2675 ZPROP_VALUE); 2676 } 2677 2678 if (nvpair_type(propval) == DATA_TYPE_STRING) { 2679 strval = fnvpair_value_string(propval); 2680 err = dsl_prop_set_string(dsname, propname, 2681 source, strval); 2682 } else { 2683 intval = fnvpair_value_uint64(propval); 2684 err = dsl_prop_set_int(dsname, propname, source, 2685 intval); 2686 } 2687 2688 if (err != 0) { 2689 if (errlist != NULL) { 2690 fnvlist_add_int32(errlist, propname, 2691 err); 2692 } 2693 rv = err; 2694 } 2695 } 2696 } 2697 nvlist_free(genericnvl); 2698 nvlist_free(retrynvl); 2699 2700 return (rv); 2701} 2702 2703/* 2704 * Check that all the properties are valid user properties. 2705 */ 2706static int 2707zfs_check_userprops(const char *fsname, nvlist_t *nvl) 2708{ 2709 nvpair_t *pair = NULL; 2710 int error = 0; 2711 2712 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { 2713 const char *propname = nvpair_name(pair); 2714 2715 if (!zfs_prop_user(propname) || 2716 nvpair_type(pair) != DATA_TYPE_STRING) 2717 return (SET_ERROR(EINVAL)); 2718 2719 if (error = zfs_secpolicy_write_perms(fsname, 2720 ZFS_DELEG_PERM_USERPROP, CRED())) 2721 return (error); 2722 2723 if (strlen(propname) >= ZAP_MAXNAMELEN) 2724 return (SET_ERROR(ENAMETOOLONG)); 2725 2726 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN) 2727 return (E2BIG); 2728 } 2729 return (0); 2730} 2731 2732static void 2733props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops) 2734{ 2735 nvpair_t *pair; 2736 2737 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2738 2739 pair = NULL; 2740 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) { 2741 if (nvlist_exists(skipped, nvpair_name(pair))) 2742 continue; 2743 2744 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0); 2745 } 2746} 2747 2748static int 2749clear_received_props(const char *dsname, nvlist_t *props, 2750 nvlist_t *skipped) 2751{ 2752 int err = 0; 2753 nvlist_t *cleared_props = NULL; 2754 props_skip(props, skipped, &cleared_props); 2755 if (!nvlist_empty(cleared_props)) { 2756 /* 2757 * Acts on local properties until the dataset has received 2758 * properties at least once on or after SPA_VERSION_RECVD_PROPS. 2759 */ 2760 zprop_source_t flags = (ZPROP_SRC_NONE | 2761 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0)); 2762 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL); 2763 } 2764 nvlist_free(cleared_props); 2765 return (err); 2766} 2767 2768/* 2769 * inputs: 2770 * zc_name name of filesystem 2771 * zc_value name of property to set 2772 * zc_nvlist_src{_size} nvlist of properties to apply 2773 * zc_cookie received properties flag 2774 * 2775 * outputs: 2776 * zc_nvlist_dst{_size} error for each unapplied received property 2777 */ 2778static int 2779zfs_ioc_set_prop(zfs_cmd_t *zc) 2780{ 2781 nvlist_t *nvl; 2782 boolean_t received = zc->zc_cookie; 2783 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED : 2784 ZPROP_SRC_LOCAL); 2785 nvlist_t *errors; 2786 int error; 2787 2788 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2789 zc->zc_iflags, &nvl)) != 0) 2790 return (error); 2791 2792 if (received) { 2793 nvlist_t *origprops; 2794 2795 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) { 2796 (void) clear_received_props(zc->zc_name, 2797 origprops, nvl); 2798 nvlist_free(origprops); 2799 } 2800 2801 error = dsl_prop_set_hasrecvd(zc->zc_name); 2802 } 2803 2804 errors = fnvlist_alloc(); 2805 if (error == 0) 2806 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors); 2807 2808 if (zc->zc_nvlist_dst != 0 && errors != NULL) { 2809 (void) put_nvlist(zc, errors); 2810 } 2811 2812 nvlist_free(errors); 2813 nvlist_free(nvl); 2814 return (error); 2815} 2816 2817/* 2818 * inputs: 2819 * zc_name name of filesystem 2820 * zc_value name of property to inherit 2821 * zc_cookie revert to received value if TRUE 2822 * 2823 * outputs: none 2824 */ 2825static int 2826zfs_ioc_inherit_prop(zfs_cmd_t *zc) 2827{ 2828 const char *propname = zc->zc_value; 2829 zfs_prop_t prop = zfs_name_to_prop(propname); 2830 boolean_t received = zc->zc_cookie; 2831 zprop_source_t source = (received 2832 ? ZPROP_SRC_NONE /* revert to received value, if any */ 2833 : ZPROP_SRC_INHERITED); /* explicitly inherit */ 2834 2835 if (received) { 2836 nvlist_t *dummy; 2837 nvpair_t *pair; 2838 zprop_type_t type; 2839 int err; 2840 2841 /* 2842 * zfs_prop_set_special() expects properties in the form of an 2843 * nvpair with type info. 2844 */ 2845 if (prop == ZPROP_INVAL) { 2846 if (!zfs_prop_user(propname)) 2847 return (SET_ERROR(EINVAL)); 2848 2849 type = PROP_TYPE_STRING; 2850 } else if (prop == ZFS_PROP_VOLSIZE || 2851 prop == ZFS_PROP_VERSION) { 2852 return (SET_ERROR(EINVAL)); 2853 } else { 2854 type = zfs_prop_get_type(prop); 2855 } 2856 2857 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2858 2859 switch (type) { 2860 case PROP_TYPE_STRING: 2861 VERIFY(0 == nvlist_add_string(dummy, propname, "")); 2862 break; 2863 case PROP_TYPE_NUMBER: 2864 case PROP_TYPE_INDEX: 2865 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0)); 2866 break; 2867 default: 2868 nvlist_free(dummy); 2869 return (SET_ERROR(EINVAL)); 2870 } 2871 2872 pair = nvlist_next_nvpair(dummy, NULL); 2873 err = zfs_prop_set_special(zc->zc_name, source, pair); 2874 nvlist_free(dummy); 2875 if (err != -1) 2876 return (err); /* special property already handled */ 2877 } else { 2878 /* 2879 * Only check this in the non-received case. We want to allow 2880 * 'inherit -S' to revert non-inheritable properties like quota 2881 * and reservation to the received or default values even though 2882 * they are not considered inheritable. 2883 */ 2884 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop)) 2885 return (SET_ERROR(EINVAL)); 2886 } 2887 2888 /* property name has been validated by zfs_secpolicy_inherit_prop() */ 2889 return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source)); 2890} 2891 2892static int 2893zfs_ioc_pool_set_props(zfs_cmd_t *zc) 2894{ 2895 nvlist_t *props; 2896 spa_t *spa; 2897 int error; 2898 nvpair_t *pair; 2899 2900 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2901 zc->zc_iflags, &props)) 2902 return (error); 2903 2904 /* 2905 * If the only property is the configfile, then just do a spa_lookup() 2906 * to handle the faulted case. 2907 */ 2908 pair = nvlist_next_nvpair(props, NULL); 2909 if (pair != NULL && strcmp(nvpair_name(pair), 2910 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 && 2911 nvlist_next_nvpair(props, pair) == NULL) { 2912 mutex_enter(&spa_namespace_lock); 2913 if ((spa = spa_lookup(zc->zc_name)) != NULL) { 2914 spa_configfile_set(spa, props, B_FALSE); 2915 spa_config_sync(spa, B_FALSE, B_TRUE); 2916 } 2917 mutex_exit(&spa_namespace_lock); 2918 if (spa != NULL) { 2919 nvlist_free(props); 2920 return (0); 2921 } 2922 } 2923 2924 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { 2925 nvlist_free(props); 2926 return (error); 2927 } 2928 2929 error = spa_prop_set(spa, props); 2930 2931 nvlist_free(props); 2932 spa_close(spa, FTAG); 2933 2934 return (error); 2935} 2936 2937static int 2938zfs_ioc_pool_get_props(zfs_cmd_t *zc) 2939{ 2940 spa_t *spa; 2941 int error; 2942 nvlist_t *nvp = NULL; 2943 2944 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { 2945 /* 2946 * If the pool is faulted, there may be properties we can still 2947 * get (such as altroot and cachefile), so attempt to get them 2948 * anyway. 2949 */ 2950 mutex_enter(&spa_namespace_lock); 2951 if ((spa = spa_lookup(zc->zc_name)) != NULL) 2952 error = spa_prop_get(spa, &nvp); 2953 mutex_exit(&spa_namespace_lock); 2954 } else { 2955 error = spa_prop_get(spa, &nvp); 2956 spa_close(spa, FTAG); 2957 } 2958 2959 if (error == 0 && zc->zc_nvlist_dst != 0) 2960 error = put_nvlist(zc, nvp); 2961 else 2962 error = SET_ERROR(EFAULT); 2963 2964 nvlist_free(nvp); 2965 return (error); 2966} 2967 2968/* 2969 * inputs: 2970 * zc_name name of filesystem 2971 * zc_nvlist_src{_size} nvlist of delegated permissions 2972 * zc_perm_action allow/unallow flag 2973 * 2974 * outputs: none 2975 */ 2976static int 2977zfs_ioc_set_fsacl(zfs_cmd_t *zc) 2978{ 2979 int error; 2980 nvlist_t *fsaclnv = NULL; 2981 2982 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2983 zc->zc_iflags, &fsaclnv)) != 0) 2984 return (error); 2985 2986 /* 2987 * Verify nvlist is constructed correctly 2988 */ 2989 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) { 2990 nvlist_free(fsaclnv); 2991 return (SET_ERROR(EINVAL)); 2992 } 2993 2994 /* 2995 * If we don't have PRIV_SYS_MOUNT, then validate 2996 * that user is allowed to hand out each permission in 2997 * the nvlist(s) 2998 */ 2999 3000 error = secpolicy_zfs(CRED()); 3001 if (error != 0) { 3002 if (zc->zc_perm_action == B_FALSE) { 3003 error = dsl_deleg_can_allow(zc->zc_name, 3004 fsaclnv, CRED()); 3005 } else { 3006 error = dsl_deleg_can_unallow(zc->zc_name, 3007 fsaclnv, CRED()); 3008 } 3009 } 3010 3011 if (error == 0) 3012 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action); 3013 3014 nvlist_free(fsaclnv); 3015 return (error); 3016} 3017 3018/* 3019 * inputs: 3020 * zc_name name of filesystem 3021 * 3022 * outputs: 3023 * zc_nvlist_src{_size} nvlist of delegated permissions 3024 */ 3025static int 3026zfs_ioc_get_fsacl(zfs_cmd_t *zc) 3027{ 3028 nvlist_t *nvp; 3029 int error; 3030 3031 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) { 3032 error = put_nvlist(zc, nvp); 3033 nvlist_free(nvp); 3034 } 3035 3036 return (error); 3037} 3038 3039/* 3040 * Search the vfs list for a specified resource. Returns a pointer to it 3041 * or NULL if no suitable entry is found. The caller of this routine 3042 * is responsible for releasing the returned vfs pointer. 3043 */ 3044static vfs_t * 3045zfs_get_vfs(const char *resource) 3046{ 3047 vfs_t *vfsp; 3048 3049 mtx_lock(&mountlist_mtx); 3050 TAILQ_FOREACH(vfsp, &mountlist, mnt_list) { 3051 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) { 3052 VFS_HOLD(vfsp); 3053 break; 3054 } 3055 } 3056 mtx_unlock(&mountlist_mtx); 3057 return (vfsp); 3058} 3059 3060/* ARGSUSED */ 3061static void 3062zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) 3063{ 3064 zfs_creat_t *zct = arg; 3065 3066 zfs_create_fs(os, cr, zct->zct_zplprops, tx); 3067} 3068 3069#define ZFS_PROP_UNDEFINED ((uint64_t)-1) 3070 3071/* 3072 * inputs: 3073 * os parent objset pointer (NULL if root fs) 3074 * fuids_ok fuids allowed in this version of the spa? 3075 * sa_ok SAs allowed in this version of the spa? 3076 * createprops list of properties requested by creator 3077 * 3078 * outputs: 3079 * zplprops values for the zplprops we attach to the master node object 3080 * is_ci true if requested file system will be purely case-insensitive 3081 * 3082 * Determine the settings for utf8only, normalization and 3083 * casesensitivity. Specific values may have been requested by the 3084 * creator and/or we can inherit values from the parent dataset. If 3085 * the file system is of too early a vintage, a creator can not 3086 * request settings for these properties, even if the requested 3087 * setting is the default value. We don't actually want to create dsl 3088 * properties for these, so remove them from the source nvlist after 3089 * processing. 3090 */ 3091static int 3092zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver, 3093 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops, 3094 nvlist_t *zplprops, boolean_t *is_ci) 3095{ 3096 uint64_t sense = ZFS_PROP_UNDEFINED; 3097 uint64_t norm = ZFS_PROP_UNDEFINED; 3098 uint64_t u8 = ZFS_PROP_UNDEFINED; 3099 3100 ASSERT(zplprops != NULL); 3101 3102 /* 3103 * Pull out creator prop choices, if any. 3104 */ 3105 if (createprops) { 3106 (void) nvlist_lookup_uint64(createprops, 3107 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver); 3108 (void) nvlist_lookup_uint64(createprops, 3109 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm); 3110 (void) nvlist_remove_all(createprops, 3111 zfs_prop_to_name(ZFS_PROP_NORMALIZE)); 3112 (void) nvlist_lookup_uint64(createprops, 3113 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8); 3114 (void) nvlist_remove_all(createprops, 3115 zfs_prop_to_name(ZFS_PROP_UTF8ONLY)); 3116 (void) nvlist_lookup_uint64(createprops, 3117 zfs_prop_to_name(ZFS_PROP_CASE), &sense); 3118 (void) nvlist_remove_all(createprops, 3119 zfs_prop_to_name(ZFS_PROP_CASE)); 3120 } 3121 3122 /* 3123 * If the zpl version requested is whacky or the file system 3124 * or pool is version is too "young" to support normalization 3125 * and the creator tried to set a value for one of the props, 3126 * error out. 3127 */ 3128 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) || 3129 (zplver >= ZPL_VERSION_FUID && !fuids_ok) || 3130 (zplver >= ZPL_VERSION_SA && !sa_ok) || 3131 (zplver < ZPL_VERSION_NORMALIZATION && 3132 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED || 3133 sense != ZFS_PROP_UNDEFINED))) 3134 return (SET_ERROR(ENOTSUP)); 3135 3136 /* 3137 * Put the version in the zplprops 3138 */ 3139 VERIFY(nvlist_add_uint64(zplprops, 3140 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0); 3141 3142 if (norm == ZFS_PROP_UNDEFINED) 3143 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0); 3144 VERIFY(nvlist_add_uint64(zplprops, 3145 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0); 3146 3147 /* 3148 * If we're normalizing, names must always be valid UTF-8 strings. 3149 */ 3150 if (norm) 3151 u8 = 1; 3152 if (u8 == ZFS_PROP_UNDEFINED) 3153 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0); 3154 VERIFY(nvlist_add_uint64(zplprops, 3155 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0); 3156 3157 if (sense == ZFS_PROP_UNDEFINED) 3158 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0); 3159 VERIFY(nvlist_add_uint64(zplprops, 3160 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0); 3161 3162 if (is_ci) 3163 *is_ci = (sense == ZFS_CASE_INSENSITIVE); 3164 3165 return (0); 3166} 3167 3168static int 3169zfs_fill_zplprops(const char *dataset, nvlist_t *createprops, 3170 nvlist_t *zplprops, boolean_t *is_ci) 3171{ 3172 boolean_t fuids_ok, sa_ok; 3173 uint64_t zplver = ZPL_VERSION; 3174 objset_t *os = NULL; 3175 char parentname[MAXNAMELEN]; 3176 char *cp; 3177 spa_t *spa; 3178 uint64_t spa_vers; 3179 int error; 3180 3181 (void) strlcpy(parentname, dataset, sizeof (parentname)); 3182 cp = strrchr(parentname, '/'); 3183 ASSERT(cp != NULL); 3184 cp[0] = '\0'; 3185 3186 if ((error = spa_open(dataset, &spa, FTAG)) != 0) 3187 return (error); 3188 3189 spa_vers = spa_version(spa); 3190 spa_close(spa, FTAG); 3191 3192 zplver = zfs_zpl_version_map(spa_vers); 3193 fuids_ok = (zplver >= ZPL_VERSION_FUID); 3194 sa_ok = (zplver >= ZPL_VERSION_SA); 3195 3196 /* 3197 * Open parent object set so we can inherit zplprop values. 3198 */ 3199 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0) 3200 return (error); 3201 3202 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops, 3203 zplprops, is_ci); 3204 dmu_objset_rele(os, FTAG); 3205 return (error); 3206} 3207 3208static int 3209zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops, 3210 nvlist_t *zplprops, boolean_t *is_ci) 3211{ 3212 boolean_t fuids_ok; 3213 boolean_t sa_ok; 3214 uint64_t zplver = ZPL_VERSION; 3215 int error; 3216 3217 zplver = zfs_zpl_version_map(spa_vers); 3218 fuids_ok = (zplver >= ZPL_VERSION_FUID); 3219 sa_ok = (zplver >= ZPL_VERSION_SA); 3220 3221 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok, 3222 createprops, zplprops, is_ci); 3223 return (error); 3224} 3225 3226/* 3227 * innvl: { 3228 * "type" -> dmu_objset_type_t (int32) 3229 * (optional) "props" -> { prop -> value } 3230 * } 3231 * 3232 * outnvl: propname -> error code (int32) 3233 */ 3234static int 3235zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3236{ 3237 int error = 0; 3238 zfs_creat_t zct = { 0 }; 3239 nvlist_t *nvprops = NULL; 3240 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx); 3241 int32_t type32; 3242 dmu_objset_type_t type; 3243 boolean_t is_insensitive = B_FALSE; 3244 3245 if (nvlist_lookup_int32(innvl, "type", &type32) != 0) 3246 return (SET_ERROR(EINVAL)); 3247 type = type32; 3248 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops); 3249 3250 switch (type) { 3251 case DMU_OST_ZFS: 3252 cbfunc = zfs_create_cb; 3253 break; 3254 3255 case DMU_OST_ZVOL: 3256 cbfunc = zvol_create_cb; 3257 break; 3258 3259 default: 3260 cbfunc = NULL; 3261 break; 3262 } 3263 if (strchr(fsname, '@') || 3264 strchr(fsname, '%')) 3265 return (SET_ERROR(EINVAL)); 3266 3267 zct.zct_props = nvprops; 3268 3269 if (cbfunc == NULL) 3270 return (SET_ERROR(EINVAL)); 3271 3272 if (type == DMU_OST_ZVOL) { 3273 uint64_t volsize, volblocksize; 3274 3275 if (nvprops == NULL) 3276 return (SET_ERROR(EINVAL)); 3277 if (nvlist_lookup_uint64(nvprops, 3278 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0) 3279 return (SET_ERROR(EINVAL)); 3280 3281 if ((error = nvlist_lookup_uint64(nvprops, 3282 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), 3283 &volblocksize)) != 0 && error != ENOENT) 3284 return (SET_ERROR(EINVAL)); 3285 3286 if (error != 0) 3287 volblocksize = zfs_prop_default_numeric( 3288 ZFS_PROP_VOLBLOCKSIZE); 3289 3290 if ((error = zvol_check_volblocksize( 3291 volblocksize)) != 0 || 3292 (error = zvol_check_volsize(volsize, 3293 volblocksize)) != 0) 3294 return (error); 3295 } else if (type == DMU_OST_ZFS) { 3296 int error; 3297 3298 /* 3299 * We have to have normalization and 3300 * case-folding flags correct when we do the 3301 * file system creation, so go figure them out 3302 * now. 3303 */ 3304 VERIFY(nvlist_alloc(&zct.zct_zplprops, 3305 NV_UNIQUE_NAME, KM_SLEEP) == 0); 3306 error = zfs_fill_zplprops(fsname, nvprops, 3307 zct.zct_zplprops, &is_insensitive); 3308 if (error != 0) { 3309 nvlist_free(zct.zct_zplprops); 3310 return (error); 3311 } 3312 } 3313 3314 error = dmu_objset_create(fsname, type, 3315 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct); 3316 nvlist_free(zct.zct_zplprops); 3317 3318 /* 3319 * It would be nice to do this atomically. 3320 */ 3321 if (error == 0) { 3322 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL, 3323 nvprops, outnvl); 3324 if (error != 0) 3325 (void) dsl_destroy_head(fsname); 3326 } 3327#ifdef __FreeBSD__ 3328 if (error == 0 && type == DMU_OST_ZVOL) 3329 zvol_create_minors(fsname); 3330#endif 3331 return (error); 3332} 3333 3334/* 3335 * innvl: { 3336 * "origin" -> name of origin snapshot 3337 * (optional) "props" -> { prop -> value } 3338 * } 3339 * 3340 * outnvl: propname -> error code (int32) 3341 */ 3342static int 3343zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3344{ 3345 int error = 0; 3346 nvlist_t *nvprops = NULL; 3347 char *origin_name; 3348 3349 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0) 3350 return (SET_ERROR(EINVAL)); 3351 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops); 3352 3353 if (strchr(fsname, '@') || 3354 strchr(fsname, '%')) 3355 return (SET_ERROR(EINVAL)); 3356 3357 if (dataset_namecheck(origin_name, NULL, NULL) != 0) 3358 return (SET_ERROR(EINVAL)); 3359 error = dmu_objset_clone(fsname, origin_name); 3360 if (error != 0) 3361 return (error); 3362 3363 /* 3364 * It would be nice to do this atomically. 3365 */ 3366 if (error == 0) { 3367 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL, 3368 nvprops, outnvl); 3369 if (error != 0) 3370 (void) dsl_destroy_head(fsname); 3371 } 3372#ifdef __FreeBSD__ 3373 if (error == 0) 3374 zvol_create_minors(fsname); 3375#endif 3376 return (error); 3377} 3378 3379/* 3380 * innvl: { 3381 * "snaps" -> { snapshot1, snapshot2 } 3382 * (optional) "props" -> { prop -> value (string) } 3383 * } 3384 * 3385 * outnvl: snapshot -> error code (int32) 3386 */ 3387static int 3388zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3389{ 3390 nvlist_t *snaps; 3391 nvlist_t *props = NULL; 3392 int error, poollen; 3393 nvpair_t *pair; 3394 3395 (void) nvlist_lookup_nvlist(innvl, "props", &props); 3396 if ((error = zfs_check_userprops(poolname, props)) != 0) 3397 return (error); 3398 3399 if (!nvlist_empty(props) && 3400 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS)) 3401 return (SET_ERROR(ENOTSUP)); 3402 3403 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 3404 return (SET_ERROR(EINVAL)); 3405 poollen = strlen(poolname); 3406 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 3407 pair = nvlist_next_nvpair(snaps, pair)) { 3408 const char *name = nvpair_name(pair); 3409 const char *cp = strchr(name, '@'); 3410 3411 /* 3412 * The snap name must contain an @, and the part after it must 3413 * contain only valid characters. 3414 */ 3415 if (cp == NULL || 3416 zfs_component_namecheck(cp + 1, NULL, NULL) != 0) 3417 return (SET_ERROR(EINVAL)); 3418 3419 /* 3420 * The snap must be in the specified pool. 3421 */ 3422 if (strncmp(name, poolname, poollen) != 0 || 3423 (name[poollen] != '/' && name[poollen] != '@')) 3424 return (SET_ERROR(EXDEV)); 3425 3426 /* This must be the only snap of this fs. */ 3427 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair); 3428 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) { 3429 if (strncmp(name, nvpair_name(pair2), cp - name + 1) 3430 == 0) { 3431 return (SET_ERROR(EXDEV)); 3432 } 3433 } 3434 } 3435 3436 error = dsl_dataset_snapshot(snaps, props, outnvl); 3437 return (error); 3438} 3439 3440/* 3441 * innvl: "message" -> string 3442 */ 3443/* ARGSUSED */ 3444static int 3445zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl) 3446{ 3447 char *message; 3448 spa_t *spa; 3449 int error; 3450 char *poolname; 3451 3452 /* 3453 * The poolname in the ioctl is not set, we get it from the TSD, 3454 * which was set at the end of the last successful ioctl that allows 3455 * logging. The secpolicy func already checked that it is set. 3456 * Only one log ioctl is allowed after each successful ioctl, so 3457 * we clear the TSD here. 3458 */ 3459 poolname = tsd_get(zfs_allow_log_key); 3460 (void) tsd_set(zfs_allow_log_key, NULL); 3461 error = spa_open(poolname, &spa, FTAG); 3462 strfree(poolname); 3463 if (error != 0) 3464 return (error); 3465 3466 if (nvlist_lookup_string(innvl, "message", &message) != 0) { 3467 spa_close(spa, FTAG); 3468 return (SET_ERROR(EINVAL)); 3469 } 3470 3471 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) { 3472 spa_close(spa, FTAG); 3473 return (SET_ERROR(ENOTSUP)); 3474 } 3475 3476 error = spa_history_log(spa, message); 3477 spa_close(spa, FTAG); 3478 return (error); 3479} 3480 3481/* 3482 * The dp_config_rwlock must not be held when calling this, because the 3483 * unmount may need to write out data. 3484 * 3485 * This function is best-effort. Callers must deal gracefully if it 3486 * remains mounted (or is remounted after this call). 3487 * 3488 * Returns 0 if the argument is not a snapshot, or it is not currently a 3489 * filesystem, or we were able to unmount it. Returns error code otherwise. 3490 */ 3491int 3492zfs_unmount_snap(const char *snapname) 3493{ 3494 vfs_t *vfsp; 3495 zfsvfs_t *zfsvfs; 3496 int err; 3497 3498 if (strchr(snapname, '@') == NULL) 3499 return (0); 3500 3501 vfsp = zfs_get_vfs(snapname); 3502 if (vfsp == NULL) 3503 return (0); 3504 3505 zfsvfs = vfsp->vfs_data; 3506 ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os))); 3507 3508 err = vn_vfswlock(vfsp->vfs_vnodecovered); 3509 VFS_RELE(vfsp); 3510 if (err != 0) 3511 return (SET_ERROR(err)); 3512 3513 /* 3514 * Always force the unmount for snapshots. 3515 */ 3516 3517#ifdef illumos 3518 (void) dounmount(vfsp, MS_FORCE, kcred); 3519#else 3520 mtx_lock(&Giant); /* dounmount() */ 3521 (void) dounmount(vfsp, MS_FORCE, curthread); 3522 mtx_unlock(&Giant); /* dounmount() */ 3523#endif 3524 return (0); 3525} 3526 3527/* ARGSUSED */ 3528static int 3529zfs_unmount_snap_cb(const char *snapname, void *arg) 3530{ 3531 return (zfs_unmount_snap(snapname)); 3532} 3533 3534/* 3535 * When a clone is destroyed, its origin may also need to be destroyed, 3536 * in which case it must be unmounted. This routine will do that unmount 3537 * if necessary. 3538 */ 3539void 3540zfs_destroy_unmount_origin(const char *fsname) 3541{ 3542 int error; 3543 objset_t *os; 3544 dsl_dataset_t *ds; 3545 3546 error = dmu_objset_hold(fsname, FTAG, &os); 3547 if (error != 0) 3548 return; 3549 ds = dmu_objset_ds(os); 3550 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) { 3551 char originname[MAXNAMELEN]; 3552 dsl_dataset_name(ds->ds_prev, originname); 3553 dmu_objset_rele(os, FTAG); 3554 (void) zfs_unmount_snap(originname); 3555 } else { 3556 dmu_objset_rele(os, FTAG); 3557 } 3558} 3559 3560/* 3561 * innvl: { 3562 * "snaps" -> { snapshot1, snapshot2 } 3563 * (optional boolean) "defer" 3564 * } 3565 * 3566 * outnvl: snapshot -> error code (int32) 3567 * 3568 */ 3569/* ARGSUSED */ 3570static int 3571zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3572{ 3573 nvlist_t *snaps; 3574 nvpair_t *pair; 3575 boolean_t defer; 3576 3577 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 3578 return (SET_ERROR(EINVAL)); 3579 defer = nvlist_exists(innvl, "defer"); 3580 3581 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 3582 pair = nvlist_next_nvpair(snaps, pair)) { 3583 (void) zfs_unmount_snap(nvpair_name(pair)); 3584 } 3585 3586 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl)); 3587} 3588 3589/* 3590 * Create bookmarks. Bookmark names are of the form <fs>#<bmark>. 3591 * All bookmarks must be in the same pool. 3592 * 3593 * innvl: { 3594 * bookmark1 -> snapshot1, bookmark2 -> snapshot2 3595 * } 3596 * 3597 * outnvl: bookmark -> error code (int32) 3598 * 3599 */ 3600/* ARGSUSED */ 3601static int 3602zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3603{ 3604 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 3605 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 3606 char *snap_name; 3607 3608 /* 3609 * Verify the snapshot argument. 3610 */ 3611 if (nvpair_value_string(pair, &snap_name) != 0) 3612 return (SET_ERROR(EINVAL)); 3613 3614 3615 /* Verify that the keys (bookmarks) are unique */ 3616 for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair); 3617 pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) { 3618 if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0) 3619 return (SET_ERROR(EINVAL)); 3620 } 3621 } 3622 3623 return (dsl_bookmark_create(innvl, outnvl)); 3624} 3625 3626/* 3627 * innvl: { 3628 * property 1, property 2, ... 3629 * } 3630 * 3631 * outnvl: { 3632 * bookmark name 1 -> { property 1, property 2, ... }, 3633 * bookmark name 2 -> { property 1, property 2, ... } 3634 * } 3635 * 3636 */ 3637static int 3638zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3639{ 3640 return (dsl_get_bookmarks(fsname, innvl, outnvl)); 3641} 3642 3643/* 3644 * innvl: { 3645 * bookmark name 1, bookmark name 2 3646 * } 3647 * 3648 * outnvl: bookmark -> error code (int32) 3649 * 3650 */ 3651static int 3652zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl, 3653 nvlist_t *outnvl) 3654{ 3655 int error, poollen; 3656 3657 poollen = strlen(poolname); 3658 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 3659 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 3660 const char *name = nvpair_name(pair); 3661 const char *cp = strchr(name, '#'); 3662 3663 /* 3664 * The bookmark name must contain an #, and the part after it 3665 * must contain only valid characters. 3666 */ 3667 if (cp == NULL || 3668 zfs_component_namecheck(cp + 1, NULL, NULL) != 0) 3669 return (SET_ERROR(EINVAL)); 3670 3671 /* 3672 * The bookmark must be in the specified pool. 3673 */ 3674 if (strncmp(name, poolname, poollen) != 0 || 3675 (name[poollen] != '/' && name[poollen] != '#')) 3676 return (SET_ERROR(EXDEV)); 3677 (void) zvol_remove_minor(name); 3678 } 3679 3680 error = dsl_bookmark_destroy(innvl, outnvl); 3681 return (error); 3682} 3683 3684/* 3685 * inputs: 3686 * zc_name name of dataset to destroy 3687 * zc_objset_type type of objset 3688 * zc_defer_destroy mark for deferred destroy 3689 * 3690 * outputs: none 3691 */ 3692static int 3693zfs_ioc_destroy(zfs_cmd_t *zc) 3694{ 3695 int err; 3696 3697 if (zc->zc_objset_type == DMU_OST_ZFS) { 3698 err = zfs_unmount_snap(zc->zc_name); 3699 if (err != 0) 3700 return (err); 3701 } 3702 3703 if (strchr(zc->zc_name, '@')) 3704 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy); 3705 else 3706 err = dsl_destroy_head(zc->zc_name); 3707 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0) 3708 (void) zvol_remove_minor(zc->zc_name); 3709 return (err); 3710} 3711 3712/* 3713 * fsname is name of dataset to rollback (to most recent snapshot) 3714 * 3715 * innvl is not used. 3716 * 3717 * outnvl: "target" -> name of most recent snapshot 3718 * } 3719 */ 3720/* ARGSUSED */ 3721static int 3722zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl) 3723{ 3724 zfsvfs_t *zfsvfs; 3725 int error; 3726 3727 if (getzfsvfs(fsname, &zfsvfs) == 0) { 3728 error = zfs_suspend_fs(zfsvfs); 3729 if (error == 0) { 3730 int resume_err; 3731 3732 error = dsl_dataset_rollback(fsname, zfsvfs, outnvl); 3733 resume_err = zfs_resume_fs(zfsvfs, fsname); 3734 error = error ? error : resume_err; 3735 } 3736 VFS_RELE(zfsvfs->z_vfs); 3737 } else { 3738 error = dsl_dataset_rollback(fsname, NULL, outnvl); 3739 } 3740 return (error); 3741} 3742 3743static int 3744recursive_unmount(const char *fsname, void *arg) 3745{ 3746 const char *snapname = arg; 3747 char fullname[MAXNAMELEN]; 3748 3749 (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname); 3750 return (zfs_unmount_snap(fullname)); 3751} 3752 3753/* 3754 * inputs: 3755 * zc_name old name of dataset 3756 * zc_value new name of dataset 3757 * zc_cookie recursive flag (only valid for snapshots) 3758 * 3759 * outputs: none 3760 */ 3761static int 3762zfs_ioc_rename(zfs_cmd_t *zc) 3763{ 3764 boolean_t recursive = zc->zc_cookie & 1; 3765#ifdef __FreeBSD__ 3766 boolean_t allow_mounted = zc->zc_cookie & 2; 3767#endif 3768 char *at; 3769 3770 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0'; 3771 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 3772 strchr(zc->zc_value, '%')) 3773 return (SET_ERROR(EINVAL)); 3774 3775 at = strchr(zc->zc_name, '@'); 3776 if (at != NULL) { 3777 /* snaps must be in same fs */ 3778 int error; 3779 3780 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1)) 3781 return (SET_ERROR(EXDEV)); 3782 *at = '\0'; 3783#ifdef illumos 3784 if (zc->zc_objset_type == DMU_OST_ZFS) { 3785#else 3786 if (zc->zc_objset_type == DMU_OST_ZFS && allow_mounted) { 3787#endif 3788 error = dmu_objset_find(zc->zc_name, 3789 recursive_unmount, at + 1, 3790 recursive ? DS_FIND_CHILDREN : 0); 3791 if (error != 0) { 3792 *at = '@'; 3793 return (error); 3794 } 3795 } 3796 error = dsl_dataset_rename_snapshot(zc->zc_name, 3797 at + 1, strchr(zc->zc_value, '@') + 1, recursive); 3798 *at = '@'; 3799 3800 return (error); 3801 } else { 3802#ifdef illumos 3803 if (zc->zc_objset_type == DMU_OST_ZVOL) 3804 (void) zvol_remove_minor(zc->zc_name); 3805#endif 3806 return (dsl_dir_rename(zc->zc_name, zc->zc_value)); 3807 } 3808} 3809 3810static int 3811zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr) 3812{ 3813 const char *propname = nvpair_name(pair); 3814 boolean_t issnap = (strchr(dsname, '@') != NULL); 3815 zfs_prop_t prop = zfs_name_to_prop(propname); 3816 uint64_t intval; 3817 int err; 3818 3819 if (prop == ZPROP_INVAL) { 3820 if (zfs_prop_user(propname)) { 3821 if (err = zfs_secpolicy_write_perms(dsname, 3822 ZFS_DELEG_PERM_USERPROP, cr)) 3823 return (err); 3824 return (0); 3825 } 3826 3827 if (!issnap && zfs_prop_userquota(propname)) { 3828 const char *perm = NULL; 3829 const char *uq_prefix = 3830 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA]; 3831 const char *gq_prefix = 3832 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA]; 3833 3834 if (strncmp(propname, uq_prefix, 3835 strlen(uq_prefix)) == 0) { 3836 perm = ZFS_DELEG_PERM_USERQUOTA; 3837 } else if (strncmp(propname, gq_prefix, 3838 strlen(gq_prefix)) == 0) { 3839 perm = ZFS_DELEG_PERM_GROUPQUOTA; 3840 } else { 3841 /* USERUSED and GROUPUSED are read-only */ 3842 return (SET_ERROR(EINVAL)); 3843 } 3844 3845 if (err = zfs_secpolicy_write_perms(dsname, perm, cr)) 3846 return (err); 3847 return (0); 3848 } 3849 3850 return (SET_ERROR(EINVAL)); 3851 } 3852 3853 if (issnap) 3854 return (SET_ERROR(EINVAL)); 3855 3856 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 3857 /* 3858 * dsl_prop_get_all_impl() returns properties in this 3859 * format. 3860 */ 3861 nvlist_t *attrs; 3862 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 3863 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 3864 &pair) == 0); 3865 } 3866 3867 /* 3868 * Check that this value is valid for this pool version 3869 */ 3870 switch (prop) { 3871 case ZFS_PROP_COMPRESSION: 3872 /* 3873 * If the user specified gzip compression, make sure 3874 * the SPA supports it. We ignore any errors here since 3875 * we'll catch them later. 3876 */ 3877 if (nvpair_type(pair) == DATA_TYPE_UINT64 && 3878 nvpair_value_uint64(pair, &intval) == 0) { 3879 if (intval >= ZIO_COMPRESS_GZIP_1 && 3880 intval <= ZIO_COMPRESS_GZIP_9 && 3881 zfs_earlier_version(dsname, 3882 SPA_VERSION_GZIP_COMPRESSION)) { 3883 return (SET_ERROR(ENOTSUP)); 3884 } 3885 3886 if (intval == ZIO_COMPRESS_ZLE && 3887 zfs_earlier_version(dsname, 3888 SPA_VERSION_ZLE_COMPRESSION)) 3889 return (SET_ERROR(ENOTSUP)); 3890 3891 if (intval == ZIO_COMPRESS_LZ4) { 3892 spa_t *spa; 3893 3894 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 3895 return (err); 3896 3897 if (!spa_feature_is_enabled(spa, 3898 SPA_FEATURE_LZ4_COMPRESS)) { 3899 spa_close(spa, FTAG); 3900 return (SET_ERROR(ENOTSUP)); 3901 } 3902 spa_close(spa, FTAG); 3903 } 3904 3905 /* 3906 * If this is a bootable dataset then 3907 * verify that the compression algorithm 3908 * is supported for booting. We must return 3909 * something other than ENOTSUP since it 3910 * implies a downrev pool version. 3911 */ 3912 if (zfs_is_bootfs(dsname) && 3913 !BOOTFS_COMPRESS_VALID(intval)) { 3914 return (SET_ERROR(ERANGE)); 3915 } 3916 } 3917 break; 3918 3919 case ZFS_PROP_COPIES: 3920 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS)) 3921 return (SET_ERROR(ENOTSUP)); 3922 break; 3923 3924 case ZFS_PROP_DEDUP: 3925 if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP)) 3926 return (SET_ERROR(ENOTSUP)); 3927 break; 3928 3929 case ZFS_PROP_SHARESMB: 3930 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID)) 3931 return (SET_ERROR(ENOTSUP)); 3932 break; 3933 3934 case ZFS_PROP_ACLINHERIT: 3935 if (nvpair_type(pair) == DATA_TYPE_UINT64 && 3936 nvpair_value_uint64(pair, &intval) == 0) { 3937 if (intval == ZFS_ACL_PASSTHROUGH_X && 3938 zfs_earlier_version(dsname, 3939 SPA_VERSION_PASSTHROUGH_X)) 3940 return (SET_ERROR(ENOTSUP)); 3941 } 3942 break; 3943 } 3944 3945 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED())); 3946} 3947 3948/* 3949 * Checks for a race condition to make sure we don't increment a feature flag 3950 * multiple times. 3951 */ 3952static int 3953zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx) 3954{ 3955 spa_t *spa = dmu_tx_pool(tx)->dp_spa; 3956 spa_feature_t *featurep = arg; 3957 3958 if (!spa_feature_is_active(spa, *featurep)) 3959 return (0); 3960 else 3961 return (SET_ERROR(EBUSY)); 3962} 3963 3964/* 3965 * The callback invoked on feature activation in the sync task caused by 3966 * zfs_prop_activate_feature. 3967 */ 3968static void 3969zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx) 3970{ 3971 spa_t *spa = dmu_tx_pool(tx)->dp_spa; 3972 spa_feature_t *featurep = arg; 3973 3974 spa_feature_incr(spa, *featurep, tx); 3975} 3976 3977/* 3978 * Activates a feature on a pool in response to a property setting. This 3979 * creates a new sync task which modifies the pool to reflect the feature 3980 * as being active. 3981 */ 3982static int 3983zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature) 3984{ 3985 int err; 3986 3987 /* EBUSY here indicates that the feature is already active */ 3988 err = dsl_sync_task(spa_name(spa), 3989 zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync, 3990 &feature, 2); 3991 3992 if (err != 0 && err != EBUSY) 3993 return (err); 3994 else 3995 return (0); 3996} 3997 3998/* 3999 * Removes properties from the given props list that fail permission checks 4000 * needed to clear them and to restore them in case of a receive error. For each 4001 * property, make sure we have both set and inherit permissions. 4002 * 4003 * Returns the first error encountered if any permission checks fail. If the 4004 * caller provides a non-NULL errlist, it also gives the complete list of names 4005 * of all the properties that failed a permission check along with the 4006 * corresponding error numbers. The caller is responsible for freeing the 4007 * returned errlist. 4008 * 4009 * If every property checks out successfully, zero is returned and the list 4010 * pointed at by errlist is NULL. 4011 */ 4012static int 4013zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist) 4014{ 4015 zfs_cmd_t *zc; 4016 nvpair_t *pair, *next_pair; 4017 nvlist_t *errors; 4018 int err, rv = 0; 4019 4020 if (props == NULL) 4021 return (0); 4022 4023 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0); 4024 4025 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP); 4026 (void) strcpy(zc->zc_name, dataset); 4027 pair = nvlist_next_nvpair(props, NULL); 4028 while (pair != NULL) { 4029 next_pair = nvlist_next_nvpair(props, pair); 4030 4031 (void) strcpy(zc->zc_value, nvpair_name(pair)); 4032 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 || 4033 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) { 4034 VERIFY(nvlist_remove_nvpair(props, pair) == 0); 4035 VERIFY(nvlist_add_int32(errors, 4036 zc->zc_value, err) == 0); 4037 } 4038 pair = next_pair; 4039 } 4040 kmem_free(zc, sizeof (zfs_cmd_t)); 4041 4042 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) { 4043 nvlist_free(errors); 4044 errors = NULL; 4045 } else { 4046 VERIFY(nvpair_value_int32(pair, &rv) == 0); 4047 } 4048 4049 if (errlist == NULL) 4050 nvlist_free(errors); 4051 else 4052 *errlist = errors; 4053 4054 return (rv); 4055} 4056 4057static boolean_t 4058propval_equals(nvpair_t *p1, nvpair_t *p2) 4059{ 4060 if (nvpair_type(p1) == DATA_TYPE_NVLIST) { 4061 /* dsl_prop_get_all_impl() format */ 4062 nvlist_t *attrs; 4063 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0); 4064 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 4065 &p1) == 0); 4066 } 4067 4068 if (nvpair_type(p2) == DATA_TYPE_NVLIST) { 4069 nvlist_t *attrs; 4070 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0); 4071 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 4072 &p2) == 0); 4073 } 4074 4075 if (nvpair_type(p1) != nvpair_type(p2)) 4076 return (B_FALSE); 4077 4078 if (nvpair_type(p1) == DATA_TYPE_STRING) { 4079 char *valstr1, *valstr2; 4080 4081 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0); 4082 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0); 4083 return (strcmp(valstr1, valstr2) == 0); 4084 } else { 4085 uint64_t intval1, intval2; 4086 4087 VERIFY(nvpair_value_uint64(p1, &intval1) == 0); 4088 VERIFY(nvpair_value_uint64(p2, &intval2) == 0); 4089 return (intval1 == intval2); 4090 } 4091} 4092 4093/* 4094 * Remove properties from props if they are not going to change (as determined 4095 * by comparison with origprops). Remove them from origprops as well, since we 4096 * do not need to clear or restore properties that won't change. 4097 */ 4098static void 4099props_reduce(nvlist_t *props, nvlist_t *origprops) 4100{ 4101 nvpair_t *pair, *next_pair; 4102 4103 if (origprops == NULL) 4104 return; /* all props need to be received */ 4105 4106 pair = nvlist_next_nvpair(props, NULL); 4107 while (pair != NULL) { 4108 const char *propname = nvpair_name(pair); 4109 nvpair_t *match; 4110 4111 next_pair = nvlist_next_nvpair(props, pair); 4112 4113 if ((nvlist_lookup_nvpair(origprops, propname, 4114 &match) != 0) || !propval_equals(pair, match)) 4115 goto next; /* need to set received value */ 4116 4117 /* don't clear the existing received value */ 4118 (void) nvlist_remove_nvpair(origprops, match); 4119 /* don't bother receiving the property */ 4120 (void) nvlist_remove_nvpair(props, pair); 4121next: 4122 pair = next_pair; 4123 } 4124} 4125 4126#ifdef DEBUG 4127static boolean_t zfs_ioc_recv_inject_err; 4128#endif 4129 4130/* 4131 * inputs: 4132 * zc_name name of containing filesystem 4133 * zc_nvlist_src{_size} nvlist of properties to apply 4134 * zc_value name of snapshot to create 4135 * zc_string name of clone origin (if DRR_FLAG_CLONE) 4136 * zc_cookie file descriptor to recv from 4137 * zc_begin_record the BEGIN record of the stream (not byteswapped) 4138 * zc_guid force flag 4139 * zc_cleanup_fd cleanup-on-exit file descriptor 4140 * zc_action_handle handle for this guid/ds mapping (or zero on first call) 4141 * 4142 * outputs: 4143 * zc_cookie number of bytes read 4144 * zc_nvlist_dst{_size} error for each unapplied received property 4145 * zc_obj zprop_errflags_t 4146 * zc_action_handle handle for this guid/ds mapping 4147 */ 4148static int 4149zfs_ioc_recv(zfs_cmd_t *zc) 4150{ 4151 file_t *fp; 4152 dmu_recv_cookie_t drc; 4153 boolean_t force = (boolean_t)zc->zc_guid; 4154 int fd; 4155 int error = 0; 4156 int props_error = 0; 4157 nvlist_t *errors; 4158 offset_t off; 4159 nvlist_t *props = NULL; /* sent properties */ 4160 nvlist_t *origprops = NULL; /* existing properties */ 4161 char *origin = NULL; 4162 char *tosnap; 4163 char tofs[ZFS_MAXNAMELEN]; 4164 cap_rights_t rights; 4165 boolean_t first_recvd_props = B_FALSE; 4166 4167 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 4168 strchr(zc->zc_value, '@') == NULL || 4169 strchr(zc->zc_value, '%')) 4170 return (SET_ERROR(EINVAL)); 4171 4172 (void) strcpy(tofs, zc->zc_value); 4173 tosnap = strchr(tofs, '@'); 4174 *tosnap++ = '\0'; 4175 4176 if (zc->zc_nvlist_src != 0 && 4177 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 4178 zc->zc_iflags, &props)) != 0) 4179 return (error); 4180 4181 fd = zc->zc_cookie; 4182 fp = getf(fd, cap_rights_init(&rights, CAP_PREAD)); 4183 if (fp == NULL) { 4184 nvlist_free(props); 4185 return (SET_ERROR(EBADF)); 4186 } 4187 4188 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0); 4189 4190 if (zc->zc_string[0]) 4191 origin = zc->zc_string; 4192 4193 error = dmu_recv_begin(tofs, tosnap, 4194 &zc->zc_begin_record, force, origin, &drc); 4195 if (error != 0) 4196 goto out; 4197 4198 /* 4199 * Set properties before we receive the stream so that they are applied 4200 * to the new data. Note that we must call dmu_recv_stream() if 4201 * dmu_recv_begin() succeeds. 4202 */ 4203 if (props != NULL && !drc.drc_newfs) { 4204 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >= 4205 SPA_VERSION_RECVD_PROPS && 4206 !dsl_prop_get_hasrecvd(tofs)) 4207 first_recvd_props = B_TRUE; 4208 4209 /* 4210 * If new received properties are supplied, they are to 4211 * completely replace the existing received properties, so stash 4212 * away the existing ones. 4213 */ 4214 if (dsl_prop_get_received(tofs, &origprops) == 0) { 4215 nvlist_t *errlist = NULL; 4216 /* 4217 * Don't bother writing a property if its value won't 4218 * change (and avoid the unnecessary security checks). 4219 * 4220 * The first receive after SPA_VERSION_RECVD_PROPS is a 4221 * special case where we blow away all local properties 4222 * regardless. 4223 */ 4224 if (!first_recvd_props) 4225 props_reduce(props, origprops); 4226 if (zfs_check_clearable(tofs, origprops, &errlist) != 0) 4227 (void) nvlist_merge(errors, errlist, 0); 4228 nvlist_free(errlist); 4229 4230 if (clear_received_props(tofs, origprops, 4231 first_recvd_props ? NULL : props) != 0) 4232 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 4233 } else { 4234 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 4235 } 4236 } 4237 4238 if (props != NULL) { 4239 props_error = dsl_prop_set_hasrecvd(tofs); 4240 4241 if (props_error == 0) { 4242 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED, 4243 props, errors); 4244 } 4245 } 4246 4247 if (zc->zc_nvlist_dst_size != 0 && 4248 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 || 4249 put_nvlist(zc, errors) != 0)) { 4250 /* 4251 * Caller made zc->zc_nvlist_dst less than the minimum expected 4252 * size or supplied an invalid address. 4253 */ 4254 props_error = SET_ERROR(EINVAL); 4255 } 4256 4257 off = fp->f_offset; 4258 error = dmu_recv_stream(&drc, fp, &off, zc->zc_cleanup_fd, 4259 &zc->zc_action_handle); 4260 4261 if (error == 0) { 4262 zfsvfs_t *zfsvfs = NULL; 4263 4264 if (getzfsvfs(tofs, &zfsvfs) == 0) { 4265 /* online recv */ 4266 int end_err; 4267 4268 error = zfs_suspend_fs(zfsvfs); 4269 /* 4270 * If the suspend fails, then the recv_end will 4271 * likely also fail, and clean up after itself. 4272 */ 4273 end_err = dmu_recv_end(&drc, zfsvfs); 4274 if (error == 0) 4275 error = zfs_resume_fs(zfsvfs, tofs); 4276 error = error ? error : end_err; 4277 VFS_RELE(zfsvfs->z_vfs); 4278 } else { 4279 error = dmu_recv_end(&drc, NULL); 4280 } 4281 } 4282 4283 zc->zc_cookie = off - fp->f_offset; 4284 if (off >= 0 && off <= MAXOFFSET_T) 4285 fp->f_offset = off; 4286 4287#ifdef DEBUG 4288 if (zfs_ioc_recv_inject_err) { 4289 zfs_ioc_recv_inject_err = B_FALSE; 4290 error = 1; 4291 } 4292#endif 4293 4294#ifdef __FreeBSD__ 4295 if (error == 0) 4296 zvol_create_minors(tofs); 4297#endif 4298 4299 /* 4300 * On error, restore the original props. 4301 */ 4302 if (error != 0 && props != NULL && !drc.drc_newfs) { 4303 if (clear_received_props(tofs, props, NULL) != 0) { 4304 /* 4305 * We failed to clear the received properties. 4306 * Since we may have left a $recvd value on the 4307 * system, we can't clear the $hasrecvd flag. 4308 */ 4309 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4310 } else if (first_recvd_props) { 4311 dsl_prop_unset_hasrecvd(tofs); 4312 } 4313 4314 if (origprops == NULL && !drc.drc_newfs) { 4315 /* We failed to stash the original properties. */ 4316 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4317 } 4318 4319 /* 4320 * dsl_props_set() will not convert RECEIVED to LOCAL on or 4321 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL 4322 * explictly if we're restoring local properties cleared in the 4323 * first new-style receive. 4324 */ 4325 if (origprops != NULL && 4326 zfs_set_prop_nvlist(tofs, (first_recvd_props ? 4327 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED), 4328 origprops, NULL) != 0) { 4329 /* 4330 * We stashed the original properties but failed to 4331 * restore them. 4332 */ 4333 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4334 } 4335 } 4336out: 4337 nvlist_free(props); 4338 nvlist_free(origprops); 4339 nvlist_free(errors); 4340 releasef(fd); 4341 4342 if (error == 0) 4343 error = props_error; 4344 4345 return (error); 4346} 4347 4348/* 4349 * inputs: 4350 * zc_name name of snapshot to send 4351 * zc_cookie file descriptor to send stream to 4352 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj) 4353 * zc_sendobj objsetid of snapshot to send 4354 * zc_fromobj objsetid of incremental fromsnap (may be zero) 4355 * zc_guid if set, estimate size of stream only. zc_cookie is ignored. 4356 * output size in zc_objset_type. 4357 * 4358 * outputs: 4359 * zc_objset_type estimated size, if zc_guid is set 4360 */ 4361static int 4362zfs_ioc_send(zfs_cmd_t *zc) 4363{ 4364 int error; 4365 offset_t off; 4366 boolean_t estimate = (zc->zc_guid != 0); 4367 4368 if (zc->zc_obj != 0) { 4369 dsl_pool_t *dp; 4370 dsl_dataset_t *tosnap; 4371 4372 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4373 if (error != 0) 4374 return (error); 4375 4376 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap); 4377 if (error != 0) { 4378 dsl_pool_rele(dp, FTAG); 4379 return (error); 4380 } 4381 4382 if (dsl_dir_is_clone(tosnap->ds_dir)) 4383 zc->zc_fromobj = tosnap->ds_dir->dd_phys->dd_origin_obj; 4384 dsl_dataset_rele(tosnap, FTAG); 4385 dsl_pool_rele(dp, FTAG); 4386 } 4387 4388 if (estimate) { 4389 dsl_pool_t *dp; 4390 dsl_dataset_t *tosnap; 4391 dsl_dataset_t *fromsnap = NULL; 4392 4393 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4394 if (error != 0) 4395 return (error); 4396 4397 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap); 4398 if (error != 0) { 4399 dsl_pool_rele(dp, FTAG); 4400 return (error); 4401 } 4402 4403 if (zc->zc_fromobj != 0) { 4404 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj, 4405 FTAG, &fromsnap); 4406 if (error != 0) { 4407 dsl_dataset_rele(tosnap, FTAG); 4408 dsl_pool_rele(dp, FTAG); 4409 return (error); 4410 } 4411 } 4412 4413 error = dmu_send_estimate(tosnap, fromsnap, 4414 &zc->zc_objset_type); 4415 4416 if (fromsnap != NULL) 4417 dsl_dataset_rele(fromsnap, FTAG); 4418 dsl_dataset_rele(tosnap, FTAG); 4419 dsl_pool_rele(dp, FTAG); 4420 } else { 4421 file_t *fp; 4422 cap_rights_t rights; 4423 4424 fp = getf(zc->zc_cookie, 4425 cap_rights_init(&rights, CAP_WRITE)); 4426 if (fp == NULL) 4427 return (SET_ERROR(EBADF)); 4428 4429 off = fp->f_offset; 4430 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj, 4431#ifdef illumos 4432 zc->zc_fromobj, zc->zc_cookie, fp->f_vnode, &off); 4433#else 4434 zc->zc_fromobj, zc->zc_cookie, fp, &off); 4435#endif 4436 4437 if (off >= 0 && off <= MAXOFFSET_T) 4438 fp->f_offset = off; 4439 releasef(zc->zc_cookie); 4440 } 4441 return (error); 4442} 4443 4444/* 4445 * inputs: 4446 * zc_name name of snapshot on which to report progress 4447 * zc_cookie file descriptor of send stream 4448 * 4449 * outputs: 4450 * zc_cookie number of bytes written in send stream thus far 4451 */ 4452static int 4453zfs_ioc_send_progress(zfs_cmd_t *zc) 4454{ 4455 dsl_pool_t *dp; 4456 dsl_dataset_t *ds; 4457 dmu_sendarg_t *dsp = NULL; 4458 int error; 4459 4460 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4461 if (error != 0) 4462 return (error); 4463 4464 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds); 4465 if (error != 0) { 4466 dsl_pool_rele(dp, FTAG); 4467 return (error); 4468 } 4469 4470 mutex_enter(&ds->ds_sendstream_lock); 4471 4472 /* 4473 * Iterate over all the send streams currently active on this dataset. 4474 * If there's one which matches the specified file descriptor _and_ the 4475 * stream was started by the current process, return the progress of 4476 * that stream. 4477 */ 4478 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL; 4479 dsp = list_next(&ds->ds_sendstreams, dsp)) { 4480 if (dsp->dsa_outfd == zc->zc_cookie && 4481 dsp->dsa_proc == curproc) 4482 break; 4483 } 4484 4485 if (dsp != NULL) 4486 zc->zc_cookie = *(dsp->dsa_off); 4487 else 4488 error = SET_ERROR(ENOENT); 4489 4490 mutex_exit(&ds->ds_sendstream_lock); 4491 dsl_dataset_rele(ds, FTAG); 4492 dsl_pool_rele(dp, FTAG); 4493 return (error); 4494} 4495 4496static int 4497zfs_ioc_inject_fault(zfs_cmd_t *zc) 4498{ 4499 int id, error; 4500 4501 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id, 4502 &zc->zc_inject_record); 4503 4504 if (error == 0) 4505 zc->zc_guid = (uint64_t)id; 4506 4507 return (error); 4508} 4509 4510static int 4511zfs_ioc_clear_fault(zfs_cmd_t *zc) 4512{ 4513 return (zio_clear_fault((int)zc->zc_guid)); 4514} 4515 4516static int 4517zfs_ioc_inject_list_next(zfs_cmd_t *zc) 4518{ 4519 int id = (int)zc->zc_guid; 4520 int error; 4521 4522 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name), 4523 &zc->zc_inject_record); 4524 4525 zc->zc_guid = id; 4526 4527 return (error); 4528} 4529 4530static int 4531zfs_ioc_error_log(zfs_cmd_t *zc) 4532{ 4533 spa_t *spa; 4534 int error; 4535 size_t count = (size_t)zc->zc_nvlist_dst_size; 4536 4537 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 4538 return (error); 4539 4540 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst, 4541 &count); 4542 if (error == 0) 4543 zc->zc_nvlist_dst_size = count; 4544 else 4545 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa); 4546 4547 spa_close(spa, FTAG); 4548 4549 return (error); 4550} 4551 4552static int 4553zfs_ioc_clear(zfs_cmd_t *zc) 4554{ 4555 spa_t *spa; 4556 vdev_t *vd; 4557 int error; 4558 4559 /* 4560 * On zpool clear we also fix up missing slogs 4561 */ 4562 mutex_enter(&spa_namespace_lock); 4563 spa = spa_lookup(zc->zc_name); 4564 if (spa == NULL) { 4565 mutex_exit(&spa_namespace_lock); 4566 return (SET_ERROR(EIO)); 4567 } 4568 if (spa_get_log_state(spa) == SPA_LOG_MISSING) { 4569 /* we need to let spa_open/spa_load clear the chains */ 4570 spa_set_log_state(spa, SPA_LOG_CLEAR); 4571 } 4572 spa->spa_last_open_failed = 0; 4573 mutex_exit(&spa_namespace_lock); 4574 4575 if (zc->zc_cookie & ZPOOL_NO_REWIND) { 4576 error = spa_open(zc->zc_name, &spa, FTAG); 4577 } else { 4578 nvlist_t *policy; 4579 nvlist_t *config = NULL; 4580 4581 if (zc->zc_nvlist_src == 0) 4582 return (SET_ERROR(EINVAL)); 4583 4584 if ((error = get_nvlist(zc->zc_nvlist_src, 4585 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) { 4586 error = spa_open_rewind(zc->zc_name, &spa, FTAG, 4587 policy, &config); 4588 if (config != NULL) { 4589 int err; 4590 4591 if ((err = put_nvlist(zc, config)) != 0) 4592 error = err; 4593 nvlist_free(config); 4594 } 4595 nvlist_free(policy); 4596 } 4597 } 4598 4599 if (error != 0) 4600 return (error); 4601 4602 spa_vdev_state_enter(spa, SCL_NONE); 4603 4604 if (zc->zc_guid == 0) { 4605 vd = NULL; 4606 } else { 4607 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE); 4608 if (vd == NULL) { 4609 (void) spa_vdev_state_exit(spa, NULL, ENODEV); 4610 spa_close(spa, FTAG); 4611 return (SET_ERROR(ENODEV)); 4612 } 4613 } 4614 4615 vdev_clear(spa, vd); 4616 4617 (void) spa_vdev_state_exit(spa, NULL, 0); 4618 4619 /* 4620 * Resume any suspended I/Os. 4621 */ 4622 if (zio_resume(spa) != 0) 4623 error = SET_ERROR(EIO); 4624 4625 spa_close(spa, FTAG); 4626 4627 return (error); 4628} 4629 4630static int 4631zfs_ioc_pool_reopen(zfs_cmd_t *zc) 4632{ 4633 spa_t *spa; 4634 int error; 4635 4636 error = spa_open(zc->zc_name, &spa, FTAG); 4637 if (error != 0) 4638 return (error); 4639 4640 spa_vdev_state_enter(spa, SCL_NONE); 4641 4642 /* 4643 * If a resilver is already in progress then set the 4644 * spa_scrub_reopen flag to B_TRUE so that we don't restart 4645 * the scan as a side effect of the reopen. Otherwise, let 4646 * vdev_open() decided if a resilver is required. 4647 */ 4648 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool); 4649 vdev_reopen(spa->spa_root_vdev); 4650 spa->spa_scrub_reopen = B_FALSE; 4651 4652 (void) spa_vdev_state_exit(spa, NULL, 0); 4653 spa_close(spa, FTAG); 4654 return (0); 4655} 4656/* 4657 * inputs: 4658 * zc_name name of filesystem 4659 * zc_value name of origin snapshot 4660 * 4661 * outputs: 4662 * zc_string name of conflicting snapshot, if there is one 4663 */ 4664static int 4665zfs_ioc_promote(zfs_cmd_t *zc) 4666{ 4667 char *cp; 4668 4669 /* 4670 * We don't need to unmount *all* the origin fs's snapshots, but 4671 * it's easier. 4672 */ 4673 cp = strchr(zc->zc_value, '@'); 4674 if (cp) 4675 *cp = '\0'; 4676 (void) dmu_objset_find(zc->zc_value, 4677 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS); 4678 return (dsl_dataset_promote(zc->zc_name, zc->zc_string)); 4679} 4680 4681/* 4682 * Retrieve a single {user|group}{used|quota}@... property. 4683 * 4684 * inputs: 4685 * zc_name name of filesystem 4686 * zc_objset_type zfs_userquota_prop_t 4687 * zc_value domain name (eg. "S-1-234-567-89") 4688 * zc_guid RID/UID/GID 4689 * 4690 * outputs: 4691 * zc_cookie property value 4692 */ 4693static int 4694zfs_ioc_userspace_one(zfs_cmd_t *zc) 4695{ 4696 zfsvfs_t *zfsvfs; 4697 int error; 4698 4699 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 4700 return (SET_ERROR(EINVAL)); 4701 4702 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE); 4703 if (error != 0) 4704 return (error); 4705 4706 error = zfs_userspace_one(zfsvfs, 4707 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie); 4708 zfsvfs_rele(zfsvfs, FTAG); 4709 4710 return (error); 4711} 4712 4713/* 4714 * inputs: 4715 * zc_name name of filesystem 4716 * zc_cookie zap cursor 4717 * zc_objset_type zfs_userquota_prop_t 4718 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist) 4719 * 4720 * outputs: 4721 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t) 4722 * zc_cookie zap cursor 4723 */ 4724static int 4725zfs_ioc_userspace_many(zfs_cmd_t *zc) 4726{ 4727 zfsvfs_t *zfsvfs; 4728 int bufsize = zc->zc_nvlist_dst_size; 4729 4730 if (bufsize <= 0) 4731 return (SET_ERROR(ENOMEM)); 4732 4733 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE); 4734 if (error != 0) 4735 return (error); 4736 4737 void *buf = kmem_alloc(bufsize, KM_SLEEP); 4738 4739 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie, 4740 buf, &zc->zc_nvlist_dst_size); 4741 4742 if (error == 0) { 4743 error = ddi_copyout(buf, 4744 (void *)(uintptr_t)zc->zc_nvlist_dst, 4745 zc->zc_nvlist_dst_size, zc->zc_iflags); 4746 } 4747 kmem_free(buf, bufsize); 4748 zfsvfs_rele(zfsvfs, FTAG); 4749 4750 return (error); 4751} 4752 4753/* 4754 * inputs: 4755 * zc_name name of filesystem 4756 * 4757 * outputs: 4758 * none 4759 */ 4760static int 4761zfs_ioc_userspace_upgrade(zfs_cmd_t *zc) 4762{ 4763 objset_t *os; 4764 int error = 0; 4765 zfsvfs_t *zfsvfs; 4766 4767 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) { 4768 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) { 4769 /* 4770 * If userused is not enabled, it may be because the 4771 * objset needs to be closed & reopened (to grow the 4772 * objset_phys_t). Suspend/resume the fs will do that. 4773 */ 4774 error = zfs_suspend_fs(zfsvfs); 4775 if (error == 0) { 4776 dmu_objset_refresh_ownership(zfsvfs->z_os, 4777 zfsvfs); 4778 error = zfs_resume_fs(zfsvfs, zc->zc_name); 4779 } 4780 } 4781 if (error == 0) 4782 error = dmu_objset_userspace_upgrade(zfsvfs->z_os); 4783 VFS_RELE(zfsvfs->z_vfs); 4784 } else { 4785 /* XXX kind of reading contents without owning */ 4786 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 4787 if (error != 0) 4788 return (error); 4789 4790 error = dmu_objset_userspace_upgrade(os); 4791 dmu_objset_rele(os, FTAG); 4792 } 4793 4794 return (error); 4795} 4796 4797#ifdef sun 4798/* 4799 * We don't want to have a hard dependency 4800 * against some special symbols in sharefs 4801 * nfs, and smbsrv. Determine them if needed when 4802 * the first file system is shared. 4803 * Neither sharefs, nfs or smbsrv are unloadable modules. 4804 */ 4805int (*znfsexport_fs)(void *arg); 4806int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t); 4807int (*zsmbexport_fs)(void *arg, boolean_t add_share); 4808 4809int zfs_nfsshare_inited; 4810int zfs_smbshare_inited; 4811 4812ddi_modhandle_t nfs_mod; 4813ddi_modhandle_t sharefs_mod; 4814ddi_modhandle_t smbsrv_mod; 4815#endif /* sun */ 4816kmutex_t zfs_share_lock; 4817 4818#ifdef sun 4819static int 4820zfs_init_sharefs() 4821{ 4822 int error; 4823 4824 ASSERT(MUTEX_HELD(&zfs_share_lock)); 4825 /* Both NFS and SMB shares also require sharetab support. */ 4826 if (sharefs_mod == NULL && ((sharefs_mod = 4827 ddi_modopen("fs/sharefs", 4828 KRTLD_MODE_FIRST, &error)) == NULL)) { 4829 return (SET_ERROR(ENOSYS)); 4830 } 4831 if (zshare_fs == NULL && ((zshare_fs = 4832 (int (*)(enum sharefs_sys_op, share_t *, uint32_t)) 4833 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) { 4834 return (SET_ERROR(ENOSYS)); 4835 } 4836 return (0); 4837} 4838#endif /* sun */ 4839 4840static int 4841zfs_ioc_share(zfs_cmd_t *zc) 4842{ 4843#ifdef sun 4844 int error; 4845 int opcode; 4846 4847 switch (zc->zc_share.z_sharetype) { 4848 case ZFS_SHARE_NFS: 4849 case ZFS_UNSHARE_NFS: 4850 if (zfs_nfsshare_inited == 0) { 4851 mutex_enter(&zfs_share_lock); 4852 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs", 4853 KRTLD_MODE_FIRST, &error)) == NULL)) { 4854 mutex_exit(&zfs_share_lock); 4855 return (SET_ERROR(ENOSYS)); 4856 } 4857 if (znfsexport_fs == NULL && 4858 ((znfsexport_fs = (int (*)(void *)) 4859 ddi_modsym(nfs_mod, 4860 "nfs_export", &error)) == NULL)) { 4861 mutex_exit(&zfs_share_lock); 4862 return (SET_ERROR(ENOSYS)); 4863 } 4864 error = zfs_init_sharefs(); 4865 if (error != 0) { 4866 mutex_exit(&zfs_share_lock); 4867 return (SET_ERROR(ENOSYS)); 4868 } 4869 zfs_nfsshare_inited = 1; 4870 mutex_exit(&zfs_share_lock); 4871 } 4872 break; 4873 case ZFS_SHARE_SMB: 4874 case ZFS_UNSHARE_SMB: 4875 if (zfs_smbshare_inited == 0) { 4876 mutex_enter(&zfs_share_lock); 4877 if (smbsrv_mod == NULL && ((smbsrv_mod = 4878 ddi_modopen("drv/smbsrv", 4879 KRTLD_MODE_FIRST, &error)) == NULL)) { 4880 mutex_exit(&zfs_share_lock); 4881 return (SET_ERROR(ENOSYS)); 4882 } 4883 if (zsmbexport_fs == NULL && ((zsmbexport_fs = 4884 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod, 4885 "smb_server_share", &error)) == NULL)) { 4886 mutex_exit(&zfs_share_lock); 4887 return (SET_ERROR(ENOSYS)); 4888 } 4889 error = zfs_init_sharefs(); 4890 if (error != 0) { 4891 mutex_exit(&zfs_share_lock); 4892 return (SET_ERROR(ENOSYS)); 4893 } 4894 zfs_smbshare_inited = 1; 4895 mutex_exit(&zfs_share_lock); 4896 } 4897 break; 4898 default: 4899 return (SET_ERROR(EINVAL)); 4900 } 4901 4902 switch (zc->zc_share.z_sharetype) { 4903 case ZFS_SHARE_NFS: 4904 case ZFS_UNSHARE_NFS: 4905 if (error = 4906 znfsexport_fs((void *) 4907 (uintptr_t)zc->zc_share.z_exportdata)) 4908 return (error); 4909 break; 4910 case ZFS_SHARE_SMB: 4911 case ZFS_UNSHARE_SMB: 4912 if (error = zsmbexport_fs((void *) 4913 (uintptr_t)zc->zc_share.z_exportdata, 4914 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ? 4915 B_TRUE: B_FALSE)) { 4916 return (error); 4917 } 4918 break; 4919 } 4920 4921 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS || 4922 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ? 4923 SHAREFS_ADD : SHAREFS_REMOVE; 4924 4925 /* 4926 * Add or remove share from sharetab 4927 */ 4928 error = zshare_fs(opcode, 4929 (void *)(uintptr_t)zc->zc_share.z_sharedata, 4930 zc->zc_share.z_sharemax); 4931 4932 return (error); 4933 4934#else /* !sun */ 4935 return (ENOSYS); 4936#endif /* !sun */ 4937} 4938 4939ace_t full_access[] = { 4940 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0} 4941}; 4942 4943/* 4944 * inputs: 4945 * zc_name name of containing filesystem 4946 * zc_obj object # beyond which we want next in-use object # 4947 * 4948 * outputs: 4949 * zc_obj next in-use object # 4950 */ 4951static int 4952zfs_ioc_next_obj(zfs_cmd_t *zc) 4953{ 4954 objset_t *os = NULL; 4955 int error; 4956 4957 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 4958 if (error != 0) 4959 return (error); 4960 4961 error = dmu_object_next(os, &zc->zc_obj, B_FALSE, 4962 os->os_dsl_dataset->ds_phys->ds_prev_snap_txg); 4963 4964 dmu_objset_rele(os, FTAG); 4965 return (error); 4966} 4967 4968/* 4969 * inputs: 4970 * zc_name name of filesystem 4971 * zc_value prefix name for snapshot 4972 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process 4973 * 4974 * outputs: 4975 * zc_value short name of new snapshot 4976 */ 4977static int 4978zfs_ioc_tmp_snapshot(zfs_cmd_t *zc) 4979{ 4980 char *snap_name; 4981 char *hold_name; 4982 int error; 4983 minor_t minor; 4984 4985 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor); 4986 if (error != 0) 4987 return (error); 4988 4989 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value, 4990 (u_longlong_t)ddi_get_lbolt64()); 4991 hold_name = kmem_asprintf("%%%s", zc->zc_value); 4992 4993 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor, 4994 hold_name); 4995 if (error == 0) 4996 (void) strcpy(zc->zc_value, snap_name); 4997 strfree(snap_name); 4998 strfree(hold_name); 4999 zfs_onexit_fd_rele(zc->zc_cleanup_fd); 5000 return (error); 5001} 5002 5003/* 5004 * inputs: 5005 * zc_name name of "to" snapshot 5006 * zc_value name of "from" snapshot 5007 * zc_cookie file descriptor to write diff data on 5008 * 5009 * outputs: 5010 * dmu_diff_record_t's to the file descriptor 5011 */ 5012static int 5013zfs_ioc_diff(zfs_cmd_t *zc) 5014{ 5015 file_t *fp; 5016 cap_rights_t rights; 5017 offset_t off; 5018 int error; 5019 5020 fp = getf(zc->zc_cookie, cap_rights_init(&rights, CAP_WRITE)); 5021 if (fp == NULL) 5022 return (SET_ERROR(EBADF)); 5023 5024 off = fp->f_offset; 5025 5026#ifdef illumos 5027 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off); 5028#else 5029 error = dmu_diff(zc->zc_name, zc->zc_value, fp, &off); 5030#endif 5031 5032 if (off >= 0 && off <= MAXOFFSET_T) 5033 fp->f_offset = off; 5034 releasef(zc->zc_cookie); 5035 5036 return (error); 5037} 5038 5039#ifdef sun 5040/* 5041 * Remove all ACL files in shares dir 5042 */ 5043static int 5044zfs_smb_acl_purge(znode_t *dzp) 5045{ 5046 zap_cursor_t zc; 5047 zap_attribute_t zap; 5048 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 5049 int error; 5050 5051 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id); 5052 (error = zap_cursor_retrieve(&zc, &zap)) == 0; 5053 zap_cursor_advance(&zc)) { 5054 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred, 5055 NULL, 0)) != 0) 5056 break; 5057 } 5058 zap_cursor_fini(&zc); 5059 return (error); 5060} 5061#endif /* sun */ 5062 5063static int 5064zfs_ioc_smb_acl(zfs_cmd_t *zc) 5065{ 5066#ifdef sun 5067 vnode_t *vp; 5068 znode_t *dzp; 5069 vnode_t *resourcevp = NULL; 5070 znode_t *sharedir; 5071 zfsvfs_t *zfsvfs; 5072 nvlist_t *nvlist; 5073 char *src, *target; 5074 vattr_t vattr; 5075 vsecattr_t vsec; 5076 int error = 0; 5077 5078 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, 5079 NO_FOLLOW, NULL, &vp)) != 0) 5080 return (error); 5081 5082 /* Now make sure mntpnt and dataset are ZFS */ 5083 5084 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 || 5085 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), 5086 zc->zc_name) != 0)) { 5087 VN_RELE(vp); 5088 return (SET_ERROR(EINVAL)); 5089 } 5090 5091 dzp = VTOZ(vp); 5092 zfsvfs = dzp->z_zfsvfs; 5093 ZFS_ENTER(zfsvfs); 5094 5095 /* 5096 * Create share dir if its missing. 5097 */ 5098 mutex_enter(&zfsvfs->z_lock); 5099 if (zfsvfs->z_shares_dir == 0) { 5100 dmu_tx_t *tx; 5101 5102 tx = dmu_tx_create(zfsvfs->z_os); 5103 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE, 5104 ZFS_SHARES_DIR); 5105 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); 5106 error = dmu_tx_assign(tx, TXG_WAIT); 5107 if (error != 0) { 5108 dmu_tx_abort(tx); 5109 } else { 5110 error = zfs_create_share_dir(zfsvfs, tx); 5111 dmu_tx_commit(tx); 5112 } 5113 if (error != 0) { 5114 mutex_exit(&zfsvfs->z_lock); 5115 VN_RELE(vp); 5116 ZFS_EXIT(zfsvfs); 5117 return (error); 5118 } 5119 } 5120 mutex_exit(&zfsvfs->z_lock); 5121 5122 ASSERT(zfsvfs->z_shares_dir); 5123 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) { 5124 VN_RELE(vp); 5125 ZFS_EXIT(zfsvfs); 5126 return (error); 5127 } 5128 5129 switch (zc->zc_cookie) { 5130 case ZFS_SMB_ACL_ADD: 5131 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; 5132 vattr.va_type = VREG; 5133 vattr.va_mode = S_IFREG|0777; 5134 vattr.va_uid = 0; 5135 vattr.va_gid = 0; 5136 5137 vsec.vsa_mask = VSA_ACE; 5138 vsec.vsa_aclentp = &full_access; 5139 vsec.vsa_aclentsz = sizeof (full_access); 5140 vsec.vsa_aclcnt = 1; 5141 5142 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string, 5143 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec); 5144 if (resourcevp) 5145 VN_RELE(resourcevp); 5146 break; 5147 5148 case ZFS_SMB_ACL_REMOVE: 5149 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred, 5150 NULL, 0); 5151 break; 5152 5153 case ZFS_SMB_ACL_RENAME: 5154 if ((error = get_nvlist(zc->zc_nvlist_src, 5155 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) { 5156 VN_RELE(vp); 5157 ZFS_EXIT(zfsvfs); 5158 return (error); 5159 } 5160 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) || 5161 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET, 5162 &target)) { 5163 VN_RELE(vp); 5164 VN_RELE(ZTOV(sharedir)); 5165 ZFS_EXIT(zfsvfs); 5166 nvlist_free(nvlist); 5167 return (error); 5168 } 5169 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target, 5170 kcred, NULL, 0); 5171 nvlist_free(nvlist); 5172 break; 5173 5174 case ZFS_SMB_ACL_PURGE: 5175 error = zfs_smb_acl_purge(sharedir); 5176 break; 5177 5178 default: 5179 error = SET_ERROR(EINVAL); 5180 break; 5181 } 5182 5183 VN_RELE(vp); 5184 VN_RELE(ZTOV(sharedir)); 5185 5186 ZFS_EXIT(zfsvfs); 5187 5188 return (error); 5189#else /* !sun */ 5190 return (EOPNOTSUPP); 5191#endif /* !sun */ 5192} 5193 5194/* 5195 * innvl: { 5196 * "holds" -> { snapname -> holdname (string), ... } 5197 * (optional) "cleanup_fd" -> fd (int32) 5198 * } 5199 * 5200 * outnvl: { 5201 * snapname -> error value (int32) 5202 * ... 5203 * } 5204 */ 5205/* ARGSUSED */ 5206static int 5207zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist) 5208{ 5209 nvlist_t *holds; 5210 int cleanup_fd = -1; 5211 int error; 5212 minor_t minor = 0; 5213 5214 error = nvlist_lookup_nvlist(args, "holds", &holds); 5215 if (error != 0) 5216 return (SET_ERROR(EINVAL)); 5217 5218 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) { 5219 error = zfs_onexit_fd_hold(cleanup_fd, &minor); 5220 if (error != 0) 5221 return (error); 5222 } 5223 5224 error = dsl_dataset_user_hold(holds, minor, errlist); 5225 if (minor != 0) 5226 zfs_onexit_fd_rele(cleanup_fd); 5227 return (error); 5228} 5229 5230/* 5231 * innvl is not used. 5232 * 5233 * outnvl: { 5234 * holdname -> time added (uint64 seconds since epoch) 5235 * ... 5236 * } 5237 */ 5238/* ARGSUSED */ 5239static int 5240zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl) 5241{ 5242 return (dsl_dataset_get_holds(snapname, outnvl)); 5243} 5244 5245/* 5246 * innvl: { 5247 * snapname -> { holdname, ... } 5248 * ... 5249 * } 5250 * 5251 * outnvl: { 5252 * snapname -> error value (int32) 5253 * ... 5254 * } 5255 */ 5256/* ARGSUSED */ 5257static int 5258zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist) 5259{ 5260 return (dsl_dataset_user_release(holds, errlist)); 5261} 5262 5263/* 5264 * inputs: 5265 * zc_name name of new filesystem or snapshot 5266 * zc_value full name of old snapshot 5267 * 5268 * outputs: 5269 * zc_cookie space in bytes 5270 * zc_objset_type compressed space in bytes 5271 * zc_perm_action uncompressed space in bytes 5272 */ 5273static int 5274zfs_ioc_space_written(zfs_cmd_t *zc) 5275{ 5276 int error; 5277 dsl_pool_t *dp; 5278 dsl_dataset_t *new, *old; 5279 5280 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 5281 if (error != 0) 5282 return (error); 5283 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new); 5284 if (error != 0) { 5285 dsl_pool_rele(dp, FTAG); 5286 return (error); 5287 } 5288 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old); 5289 if (error != 0) { 5290 dsl_dataset_rele(new, FTAG); 5291 dsl_pool_rele(dp, FTAG); 5292 return (error); 5293 } 5294 5295 error = dsl_dataset_space_written(old, new, &zc->zc_cookie, 5296 &zc->zc_objset_type, &zc->zc_perm_action); 5297 dsl_dataset_rele(old, FTAG); 5298 dsl_dataset_rele(new, FTAG); 5299 dsl_pool_rele(dp, FTAG); 5300 return (error); 5301} 5302 5303/* 5304 * innvl: { 5305 * "firstsnap" -> snapshot name 5306 * } 5307 * 5308 * outnvl: { 5309 * "used" -> space in bytes 5310 * "compressed" -> compressed space in bytes 5311 * "uncompressed" -> uncompressed space in bytes 5312 * } 5313 */ 5314static int 5315zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl) 5316{ 5317 int error; 5318 dsl_pool_t *dp; 5319 dsl_dataset_t *new, *old; 5320 char *firstsnap; 5321 uint64_t used, comp, uncomp; 5322 5323 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0) 5324 return (SET_ERROR(EINVAL)); 5325 5326 error = dsl_pool_hold(lastsnap, FTAG, &dp); 5327 if (error != 0) 5328 return (error); 5329 5330 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new); 5331 if (error != 0) { 5332 dsl_pool_rele(dp, FTAG); 5333 return (error); 5334 } 5335 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old); 5336 if (error != 0) { 5337 dsl_dataset_rele(new, FTAG); 5338 dsl_pool_rele(dp, FTAG); 5339 return (error); 5340 } 5341 5342 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp); 5343 dsl_dataset_rele(old, FTAG); 5344 dsl_dataset_rele(new, FTAG); 5345 dsl_pool_rele(dp, FTAG); 5346 fnvlist_add_uint64(outnvl, "used", used); 5347 fnvlist_add_uint64(outnvl, "compressed", comp); 5348 fnvlist_add_uint64(outnvl, "uncompressed", uncomp); 5349 return (error); 5350} 5351 5352static int 5353zfs_ioc_jail(zfs_cmd_t *zc) 5354{ 5355 5356 return (zone_dataset_attach(curthread->td_ucred, zc->zc_name, 5357 (int)zc->zc_jailid)); 5358} 5359 5360static int 5361zfs_ioc_unjail(zfs_cmd_t *zc) 5362{ 5363 5364 return (zone_dataset_detach(curthread->td_ucred, zc->zc_name, 5365 (int)zc->zc_jailid)); 5366} 5367 5368/* 5369 * innvl: { 5370 * "fd" -> file descriptor to write stream to (int32) 5371 * (optional) "fromsnap" -> full snap name to send an incremental from 5372 * } 5373 * 5374 * outnvl is unused 5375 */ 5376/* ARGSUSED */ 5377static int 5378zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl) 5379{ 5380 cap_rights_t rights; 5381 int error; 5382 offset_t off; 5383 char *fromname = NULL; 5384 int fd; 5385 5386 error = nvlist_lookup_int32(innvl, "fd", &fd); 5387 if (error != 0) 5388 return (SET_ERROR(EINVAL)); 5389 5390 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname); 5391 5392 file_t *fp = getf(fd, cap_rights_init(&rights, CAP_READ)); 5393 if (fp == NULL) 5394 return (SET_ERROR(EBADF)); 5395 5396 off = fp->f_offset; 5397#ifdef illumos 5398 error = dmu_send(snapname, fromname, fd, fp->f_vnode, &off); 5399#else 5400 error = dmu_send(snapname, fromname, fd, fp, &off); 5401#endif 5402 5403#ifdef illumos 5404 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0) 5405 fp->f_offset = off; 5406#else 5407 fp->f_offset = off; 5408#endif 5409 5410 releasef(fd); 5411 return (error); 5412} 5413 5414/* 5415 * Determine approximately how large a zfs send stream will be -- the number 5416 * of bytes that will be written to the fd supplied to zfs_ioc_send_new(). 5417 * 5418 * innvl: { 5419 * (optional) "fromsnap" -> full snap name to send an incremental from 5420 * } 5421 * 5422 * outnvl: { 5423 * "space" -> bytes of space (uint64) 5424 * } 5425 */ 5426static int 5427zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl) 5428{ 5429 dsl_pool_t *dp; 5430 dsl_dataset_t *fromsnap = NULL; 5431 dsl_dataset_t *tosnap; 5432 int error; 5433 char *fromname; 5434 uint64_t space; 5435 5436 error = dsl_pool_hold(snapname, FTAG, &dp); 5437 if (error != 0) 5438 return (error); 5439 5440 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap); 5441 if (error != 0) { 5442 dsl_pool_rele(dp, FTAG); 5443 return (error); 5444 } 5445 5446 error = nvlist_lookup_string(innvl, "fromsnap", &fromname); 5447 if (error == 0) { 5448 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap); 5449 if (error != 0) { 5450 dsl_dataset_rele(tosnap, FTAG); 5451 dsl_pool_rele(dp, FTAG); 5452 return (error); 5453 } 5454 } 5455 5456 error = dmu_send_estimate(tosnap, fromsnap, &space); 5457 fnvlist_add_uint64(outnvl, "space", space); 5458 5459 if (fromsnap != NULL) 5460 dsl_dataset_rele(fromsnap, FTAG); 5461 dsl_dataset_rele(tosnap, FTAG); 5462 dsl_pool_rele(dp, FTAG); 5463 return (error); 5464} 5465 5466 5467static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST]; 5468 5469static void 5470zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5471 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck, 5472 boolean_t log_history, zfs_ioc_poolcheck_t pool_check) 5473{ 5474 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST]; 5475 5476 ASSERT3U(ioc, >=, ZFS_IOC_FIRST); 5477 ASSERT3U(ioc, <, ZFS_IOC_LAST); 5478 ASSERT3P(vec->zvec_legacy_func, ==, NULL); 5479 ASSERT3P(vec->zvec_func, ==, NULL); 5480 5481 vec->zvec_legacy_func = func; 5482 vec->zvec_secpolicy = secpolicy; 5483 vec->zvec_namecheck = namecheck; 5484 vec->zvec_allow_log = log_history; 5485 vec->zvec_pool_check = pool_check; 5486} 5487 5488/* 5489 * See the block comment at the beginning of this file for details on 5490 * each argument to this function. 5491 */ 5492static void 5493zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func, 5494 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck, 5495 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist, 5496 boolean_t allow_log) 5497{ 5498 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST]; 5499 5500 ASSERT3U(ioc, >=, ZFS_IOC_FIRST); 5501 ASSERT3U(ioc, <, ZFS_IOC_LAST); 5502 ASSERT3P(vec->zvec_legacy_func, ==, NULL); 5503 ASSERT3P(vec->zvec_func, ==, NULL); 5504 5505 /* if we are logging, the name must be valid */ 5506 ASSERT(!allow_log || namecheck != NO_NAME); 5507 5508 vec->zvec_name = name; 5509 vec->zvec_func = func; 5510 vec->zvec_secpolicy = secpolicy; 5511 vec->zvec_namecheck = namecheck; 5512 vec->zvec_pool_check = pool_check; 5513 vec->zvec_smush_outnvlist = smush_outnvlist; 5514 vec->zvec_allow_log = allow_log; 5515} 5516 5517static void 5518zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5519 zfs_secpolicy_func_t *secpolicy, boolean_t log_history, 5520 zfs_ioc_poolcheck_t pool_check) 5521{ 5522 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5523 POOL_NAME, log_history, pool_check); 5524} 5525 5526static void 5527zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5528 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check) 5529{ 5530 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5531 DATASET_NAME, B_FALSE, pool_check); 5532} 5533 5534static void 5535zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func) 5536{ 5537 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config, 5538 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5539} 5540 5541static void 5542zfs_ioctl_register_pool_meta(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 NO_NAME, B_FALSE, POOL_CHECK_NONE); 5547} 5548 5549static void 5550zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc, 5551 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy) 5552{ 5553 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5554 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED); 5555} 5556 5557static void 5558zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func) 5559{ 5560 zfs_ioctl_register_dataset_read_secpolicy(ioc, func, 5561 zfs_secpolicy_read); 5562} 5563 5564static void 5565zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5566 zfs_secpolicy_func_t *secpolicy) 5567{ 5568 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5569 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5570} 5571 5572static void 5573zfs_ioctl_init(void) 5574{ 5575 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT, 5576 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME, 5577 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5578 5579 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY, 5580 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME, 5581 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE); 5582 5583 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS, 5584 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME, 5585 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5586 5587 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW, 5588 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME, 5589 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5590 5591 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE, 5592 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME, 5593 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5594 5595 zfs_ioctl_register("create", ZFS_IOC_CREATE, 5596 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME, 5597 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5598 5599 zfs_ioctl_register("clone", ZFS_IOC_CLONE, 5600 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME, 5601 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5602 5603 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS, 5604 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME, 5605 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5606 5607 zfs_ioctl_register("hold", ZFS_IOC_HOLD, 5608 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME, 5609 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5610 zfs_ioctl_register("release", ZFS_IOC_RELEASE, 5611 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME, 5612 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5613 5614 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS, 5615 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME, 5616 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5617 5618 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK, 5619 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, 5620 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE); 5621 5622 zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK, 5623 zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME, 5624 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5625 5626 zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS, 5627 zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME, 5628 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5629 5630 zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS, 5631 zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks, 5632 POOL_NAME, 5633 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5634 5635 /* IOCTLS that use the legacy function signature */ 5636 5637 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze, 5638 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY); 5639 5640 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create, 5641 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5642 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN, 5643 zfs_ioc_pool_scan); 5644 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE, 5645 zfs_ioc_pool_upgrade); 5646 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD, 5647 zfs_ioc_vdev_add); 5648 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE, 5649 zfs_ioc_vdev_remove); 5650 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE, 5651 zfs_ioc_vdev_set_state); 5652 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH, 5653 zfs_ioc_vdev_attach); 5654 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH, 5655 zfs_ioc_vdev_detach); 5656 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH, 5657 zfs_ioc_vdev_setpath); 5658 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU, 5659 zfs_ioc_vdev_setfru); 5660 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS, 5661 zfs_ioc_pool_set_props); 5662 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT, 5663 zfs_ioc_vdev_split); 5664 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID, 5665 zfs_ioc_pool_reguid); 5666 5667 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS, 5668 zfs_ioc_pool_configs, zfs_secpolicy_none); 5669 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT, 5670 zfs_ioc_pool_tryimport, zfs_secpolicy_config); 5671 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT, 5672 zfs_ioc_inject_fault, zfs_secpolicy_inject); 5673 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT, 5674 zfs_ioc_clear_fault, zfs_secpolicy_inject); 5675 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT, 5676 zfs_ioc_inject_list_next, zfs_secpolicy_inject); 5677 5678 /* 5679 * pool destroy, and export don't log the history as part of 5680 * zfsdev_ioctl, but rather zfs_ioc_pool_export 5681 * does the logging of those commands. 5682 */ 5683 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy, 5684 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE); 5685 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export, 5686 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE); 5687 5688 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats, 5689 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE); 5690 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props, 5691 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE); 5692 5693 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log, 5694 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_NONE); 5695 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME, 5696 zfs_ioc_dsobj_to_dsname, 5697 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_NONE); 5698 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY, 5699 zfs_ioc_pool_get_history, 5700 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED); 5701 5702 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import, 5703 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5704 5705 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear, 5706 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5707 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen, 5708 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED); 5709 5710 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN, 5711 zfs_ioc_space_written); 5712 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS, 5713 zfs_ioc_objset_recvd_props); 5714 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ, 5715 zfs_ioc_next_obj); 5716 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL, 5717 zfs_ioc_get_fsacl); 5718 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS, 5719 zfs_ioc_objset_stats); 5720 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS, 5721 zfs_ioc_objset_zplprops); 5722 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT, 5723 zfs_ioc_dataset_list_next); 5724 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT, 5725 zfs_ioc_snapshot_list_next); 5726 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS, 5727 zfs_ioc_send_progress); 5728 5729 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF, 5730 zfs_ioc_diff, zfs_secpolicy_diff); 5731 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS, 5732 zfs_ioc_obj_to_stats, zfs_secpolicy_diff); 5733 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH, 5734 zfs_ioc_obj_to_path, zfs_secpolicy_diff); 5735 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE, 5736 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one); 5737 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY, 5738 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many); 5739 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND, 5740 zfs_ioc_send, zfs_secpolicy_send); 5741 5742 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop, 5743 zfs_secpolicy_none); 5744 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy, 5745 zfs_secpolicy_destroy); 5746 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename, 5747 zfs_secpolicy_rename); 5748 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv, 5749 zfs_secpolicy_recv); 5750 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote, 5751 zfs_secpolicy_promote); 5752 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP, 5753 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop); 5754 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl, 5755 zfs_secpolicy_set_fsacl); 5756 5757 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share, 5758 zfs_secpolicy_share, POOL_CHECK_NONE); 5759 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl, 5760 zfs_secpolicy_smb_acl, POOL_CHECK_NONE); 5761 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE, 5762 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade, 5763 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5764 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT, 5765 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot, 5766 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5767 5768#ifdef __FreeBSD__ 5769 zfs_ioctl_register_dataset_nolog(ZFS_IOC_JAIL, zfs_ioc_jail, 5770 zfs_secpolicy_config, POOL_CHECK_NONE); 5771 zfs_ioctl_register_dataset_nolog(ZFS_IOC_UNJAIL, zfs_ioc_unjail, 5772 zfs_secpolicy_config, POOL_CHECK_NONE); 5773#endif 5774} 5775 5776int 5777pool_status_check(const char *name, zfs_ioc_namecheck_t type, 5778 zfs_ioc_poolcheck_t check) 5779{ 5780 spa_t *spa; 5781 int error; 5782 5783 ASSERT(type == POOL_NAME || type == DATASET_NAME); 5784 5785 if (check & POOL_CHECK_NONE) 5786 return (0); 5787 5788 error = spa_open(name, &spa, FTAG); 5789 if (error == 0) { 5790 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa)) 5791 error = SET_ERROR(EAGAIN); 5792 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa)) 5793 error = SET_ERROR(EROFS); 5794 spa_close(spa, FTAG); 5795 } 5796 return (error); 5797} 5798 5799/* 5800 * Find a free minor number. 5801 */ 5802minor_t 5803zfsdev_minor_alloc(void) 5804{ 5805 static minor_t last_minor; 5806 minor_t m; 5807 5808 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 5809 5810 for (m = last_minor + 1; m != last_minor; m++) { 5811 if (m > ZFSDEV_MAX_MINOR) 5812 m = 1; 5813 if (ddi_get_soft_state(zfsdev_state, m) == NULL) { 5814 last_minor = m; 5815 return (m); 5816 } 5817 } 5818 5819 return (0); 5820} 5821 5822static int 5823zfs_ctldev_init(struct cdev *devp) 5824{ 5825 minor_t minor; 5826 zfs_soft_state_t *zs; 5827 5828 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 5829 5830 minor = zfsdev_minor_alloc(); 5831 if (minor == 0) 5832 return (SET_ERROR(ENXIO)); 5833 5834 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS) 5835 return (SET_ERROR(EAGAIN)); 5836 5837 devfs_set_cdevpriv((void *)(uintptr_t)minor, zfsdev_close); 5838 5839 zs = ddi_get_soft_state(zfsdev_state, minor); 5840 zs->zss_type = ZSST_CTLDEV; 5841 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data); 5842 5843 return (0); 5844} 5845 5846static void 5847zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor) 5848{ 5849 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 5850 5851 zfs_onexit_destroy(zo); 5852 ddi_soft_state_free(zfsdev_state, minor); 5853} 5854 5855void * 5856zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which) 5857{ 5858 zfs_soft_state_t *zp; 5859 5860 zp = ddi_get_soft_state(zfsdev_state, minor); 5861 if (zp == NULL || zp->zss_type != which) 5862 return (NULL); 5863 5864 return (zp->zss_data); 5865} 5866 5867static int 5868zfsdev_open(struct cdev *devp, int flag, int mode, struct thread *td) 5869{ 5870 int error = 0; 5871 5872#ifdef sun 5873 if (getminor(*devp) != 0) 5874 return (zvol_open(devp, flag, otyp, cr)); 5875#endif 5876 5877 /* This is the control device. Allocate a new minor if requested. */ 5878 if (flag & FEXCL) { 5879 mutex_enter(&spa_namespace_lock); 5880 error = zfs_ctldev_init(devp); 5881 mutex_exit(&spa_namespace_lock); 5882 } 5883 5884 return (error); 5885} 5886 5887static void 5888zfsdev_close(void *data) 5889{ 5890 zfs_onexit_t *zo; 5891 minor_t minor = (minor_t)(uintptr_t)data; 5892 5893 if (minor == 0) 5894 return; 5895 5896 mutex_enter(&spa_namespace_lock); 5897 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV); 5898 if (zo == NULL) { 5899 mutex_exit(&spa_namespace_lock); 5900 return; 5901 } 5902 zfs_ctldev_destroy(zo, minor); 5903 mutex_exit(&spa_namespace_lock); 5904} 5905 5906static int 5907zfsdev_ioctl(struct cdev *dev, u_long zcmd, caddr_t arg, int flag, 5908 struct thread *td) 5909{ 5910 zfs_cmd_t *zc; 5911 uint_t vecnum; 5912 int error, rc, len; 5913#ifdef illumos 5914 minor_t minor = getminor(dev); 5915#else 5916 zfs_iocparm_t *zc_iocparm; 5917 int cflag, cmd, oldvecnum; 5918 boolean_t newioc, compat; 5919 cred_t *cr = td->td_ucred; 5920#endif 5921 const zfs_ioc_vec_t *vec; 5922 char *saved_poolname = NULL; 5923 nvlist_t *innvl = NULL; 5924 5925 cflag = ZFS_CMD_COMPAT_NONE; 5926 compat = B_FALSE; 5927 newioc = B_TRUE; 5928 5929 len = IOCPARM_LEN(zcmd); 5930 cmd = zcmd & 0xff; 5931 5932 /* 5933 * Check if we are talking to supported older binaries 5934 * and translate zfs_cmd if necessary 5935 */ 5936 if (len != sizeof(zfs_iocparm_t)) { 5937 newioc = B_FALSE; 5938 if (len == sizeof(zfs_cmd_t)) { 5939 cflag = ZFS_CMD_COMPAT_LZC; 5940 vecnum = cmd; 5941 } else if (len == sizeof(zfs_cmd_deadman_t)) { 5942 cflag = ZFS_CMD_COMPAT_DEADMAN; 5943 compat = B_TRUE; 5944 vecnum = cmd; 5945 } else if (len == sizeof(zfs_cmd_v28_t)) { 5946 cflag = ZFS_CMD_COMPAT_V28; 5947 compat = B_TRUE; 5948 vecnum = cmd; 5949 } else if (len == sizeof(zfs_cmd_v15_t)) { 5950 cflag = ZFS_CMD_COMPAT_V15; 5951 compat = B_TRUE; 5952 vecnum = zfs_ioctl_v15_to_v28[cmd]; 5953 } else 5954 return (EINVAL); 5955 } else 5956 vecnum = cmd; 5957 5958#ifdef illumos 5959 vecnum = cmd - ZFS_IOC_FIRST; 5960 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip)); 5961#endif 5962 5963 if (compat) { 5964 if (vecnum == ZFS_IOC_COMPAT_PASS) 5965 return (0); 5966 else if (vecnum == ZFS_IOC_COMPAT_FAIL) 5967 return (ENOTSUP); 5968 } 5969 5970 /* 5971 * Check if we have sufficient kernel memory allocated 5972 * for the zfs_cmd_t request. Bail out if not so we 5973 * will not access undefined memory region. 5974 */ 5975 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0])) 5976 return (SET_ERROR(EINVAL)); 5977 vec = &zfs_ioc_vec[vecnum]; 5978 5979#ifdef illumos 5980 zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP); 5981 bzero(zc, sizeof(zfs_cmd_t)); 5982 5983 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag); 5984 if (error != 0) { 5985 error = SET_ERROR(EFAULT); 5986 goto out; 5987 } 5988#else /* !illumos */ 5989 /* 5990 * We don't alloc/free zc only if talking to library ioctl version 2 5991 */ 5992 if (cflag != ZFS_CMD_COMPAT_LZC) { 5993 zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP); 5994 bzero(zc, sizeof(zfs_cmd_t)); 5995 } else { 5996 zc = (void *)arg; 5997 error = 0; 5998 } 5999 6000 if (newioc) { 6001 zc_iocparm = (void *)arg; 6002 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_t)) { 6003 error = SET_ERROR(EFAULT); 6004 goto out; 6005 } 6006 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd, zc, 6007 sizeof(zfs_cmd_t), flag); 6008 if (error != 0) { 6009 error = SET_ERROR(EFAULT); 6010 goto out; 6011 } 6012 } 6013 6014 if (compat) { 6015 zfs_cmd_compat_get(zc, arg, cflag); 6016 oldvecnum = vecnum; 6017 error = zfs_ioctl_compat_pre(zc, &vecnum, cflag); 6018 if (error != 0) 6019 goto out; 6020 if (oldvecnum != vecnum) 6021 vec = &zfs_ioc_vec[vecnum]; 6022 } 6023#endif /* !illumos */ 6024 6025 zc->zc_iflags = flag & FKIOCTL; 6026 if (zc->zc_nvlist_src_size != 0) { 6027 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 6028 zc->zc_iflags, &innvl); 6029 if (error != 0) 6030 goto out; 6031 } 6032 6033 /* rewrite innvl for backwards compatibility */ 6034 if (compat) 6035 innvl = zfs_ioctl_compat_innvl(zc, innvl, vecnum, cflag); 6036 6037 /* 6038 * Ensure that all pool/dataset names are valid before we pass down to 6039 * the lower layers. 6040 */ 6041 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0'; 6042 switch (vec->zvec_namecheck) { 6043 case POOL_NAME: 6044 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0) 6045 error = SET_ERROR(EINVAL); 6046 else 6047 error = pool_status_check(zc->zc_name, 6048 vec->zvec_namecheck, vec->zvec_pool_check); 6049 break; 6050 6051 case DATASET_NAME: 6052 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0) 6053 error = SET_ERROR(EINVAL); 6054 else 6055 error = pool_status_check(zc->zc_name, 6056 vec->zvec_namecheck, vec->zvec_pool_check); 6057 break; 6058 6059 case NO_NAME: 6060 break; 6061 } 6062 6063 if (error == 0 && !(flag & FKIOCTL)) 6064 error = vec->zvec_secpolicy(zc, innvl, cr); 6065 6066 if (error != 0) 6067 goto out; 6068 6069 /* legacy ioctls can modify zc_name */ 6070 len = strcspn(zc->zc_name, "/@#") + 1; 6071 saved_poolname = kmem_alloc(len, KM_SLEEP); 6072 (void) strlcpy(saved_poolname, zc->zc_name, len); 6073 6074 if (vec->zvec_func != NULL) { 6075 nvlist_t *outnvl; 6076 int puterror = 0; 6077 spa_t *spa; 6078 nvlist_t *lognv = NULL; 6079 6080 ASSERT(vec->zvec_legacy_func == NULL); 6081 6082 /* 6083 * Add the innvl to the lognv before calling the func, 6084 * in case the func changes the innvl. 6085 */ 6086 if (vec->zvec_allow_log) { 6087 lognv = fnvlist_alloc(); 6088 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL, 6089 vec->zvec_name); 6090 if (!nvlist_empty(innvl)) { 6091 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL, 6092 innvl); 6093 } 6094 } 6095 6096 outnvl = fnvlist_alloc(); 6097 error = vec->zvec_func(zc->zc_name, innvl, outnvl); 6098 6099 if (error == 0 && vec->zvec_allow_log && 6100 spa_open(zc->zc_name, &spa, FTAG) == 0) { 6101 if (!nvlist_empty(outnvl)) { 6102 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL, 6103 outnvl); 6104 } 6105 (void) spa_history_log_nvl(spa, lognv); 6106 spa_close(spa, FTAG); 6107 } 6108 fnvlist_free(lognv); 6109 6110 /* rewrite outnvl for backwards compatibility */ 6111 if (cflag != ZFS_CMD_COMPAT_NONE && cflag != ZFS_CMD_COMPAT_LZC) 6112 outnvl = zfs_ioctl_compat_outnvl(zc, outnvl, vecnum, 6113 cflag); 6114 6115 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) { 6116 int smusherror = 0; 6117 if (vec->zvec_smush_outnvlist) { 6118 smusherror = nvlist_smush(outnvl, 6119 zc->zc_nvlist_dst_size); 6120 } 6121 if (smusherror == 0) 6122 puterror = put_nvlist(zc, outnvl); 6123 } 6124 6125 if (puterror != 0) 6126 error = puterror; 6127 6128 nvlist_free(outnvl); 6129 } else { 6130 error = vec->zvec_legacy_func(zc); 6131 } 6132 6133out: 6134 nvlist_free(innvl); 6135 6136 if (compat) { 6137 zfs_ioctl_compat_post(zc, cmd, cflag); 6138 zfs_cmd_compat_put(zc, arg, vecnum, cflag); 6139 } 6140 6141#ifdef illumos 6142 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag); 6143 if (error == 0 && rc != 0) 6144 error = SET_ERROR(EFAULT); 6145#else 6146 if (newioc) { 6147 rc = ddi_copyout(zc, (void *)(uintptr_t)zc_iocparm->zfs_cmd, 6148 sizeof (zfs_cmd_t), flag); 6149 if (error == 0 && rc != 0) 6150 error = SET_ERROR(EFAULT); 6151 } 6152#endif 6153 if (error == 0 && vec->zvec_allow_log) { 6154 char *s = tsd_get(zfs_allow_log_key); 6155 if (s != NULL) 6156 strfree(s); 6157 (void) tsd_set(zfs_allow_log_key, saved_poolname); 6158 } else { 6159 if (saved_poolname != NULL) 6160 strfree(saved_poolname); 6161 } 6162 6163#ifdef illumos 6164 kmem_free(zc, sizeof (zfs_cmd_t)); 6165#else 6166 /* 6167 * We don't alloc/free zc only if talking to library ioctl version 2 6168 */ 6169 if (cflag != ZFS_CMD_COMPAT_LZC) 6170 kmem_free(zc, sizeof (zfs_cmd_t)); 6171#endif 6172 return (error); 6173} 6174 6175#ifdef sun 6176static int 6177zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd) 6178{ 6179 if (cmd != DDI_ATTACH) 6180 return (DDI_FAILURE); 6181 6182 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0, 6183 DDI_PSEUDO, 0) == DDI_FAILURE) 6184 return (DDI_FAILURE); 6185 6186 zfs_dip = dip; 6187 6188 ddi_report_dev(dip); 6189 6190 return (DDI_SUCCESS); 6191} 6192 6193static int 6194zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd) 6195{ 6196 if (spa_busy() || zfs_busy() || zvol_busy()) 6197 return (DDI_FAILURE); 6198 6199 if (cmd != DDI_DETACH) 6200 return (DDI_FAILURE); 6201 6202 zfs_dip = NULL; 6203 6204 ddi_prop_remove_all(dip); 6205 ddi_remove_minor_node(dip, NULL); 6206 6207 return (DDI_SUCCESS); 6208} 6209 6210/*ARGSUSED*/ 6211static int 6212zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result) 6213{ 6214 switch (infocmd) { 6215 case DDI_INFO_DEVT2DEVINFO: 6216 *result = zfs_dip; 6217 return (DDI_SUCCESS); 6218 6219 case DDI_INFO_DEVT2INSTANCE: 6220 *result = (void *)0; 6221 return (DDI_SUCCESS); 6222 } 6223 6224 return (DDI_FAILURE); 6225} 6226#endif /* sun */ 6227 6228/* 6229 * OK, so this is a little weird. 6230 * 6231 * /dev/zfs is the control node, i.e. minor 0. 6232 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0. 6233 * 6234 * /dev/zfs has basically nothing to do except serve up ioctls, 6235 * so most of the standard driver entry points are in zvol.c. 6236 */ 6237#ifdef sun 6238static struct cb_ops zfs_cb_ops = { 6239 zfsdev_open, /* open */ 6240 zfsdev_close, /* close */ 6241 zvol_strategy, /* strategy */ 6242 nodev, /* print */ 6243 zvol_dump, /* dump */ 6244 zvol_read, /* read */ 6245 zvol_write, /* write */ 6246 zfsdev_ioctl, /* ioctl */ 6247 nodev, /* devmap */ 6248 nodev, /* mmap */ 6249 nodev, /* segmap */ 6250 nochpoll, /* poll */ 6251 ddi_prop_op, /* prop_op */ 6252 NULL, /* streamtab */ 6253 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */ 6254 CB_REV, /* version */ 6255 nodev, /* async read */ 6256 nodev, /* async write */ 6257}; 6258 6259static struct dev_ops zfs_dev_ops = { 6260 DEVO_REV, /* version */ 6261 0, /* refcnt */ 6262 zfs_info, /* info */ 6263 nulldev, /* identify */ 6264 nulldev, /* probe */ 6265 zfs_attach, /* attach */ 6266 zfs_detach, /* detach */ 6267 nodev, /* reset */ 6268 &zfs_cb_ops, /* driver operations */ 6269 NULL, /* no bus operations */ 6270 NULL, /* power */ 6271 ddi_quiesce_not_needed, /* quiesce */ 6272}; 6273 6274static struct modldrv zfs_modldrv = { 6275 &mod_driverops, 6276 "ZFS storage pool", 6277 &zfs_dev_ops 6278}; 6279 6280static struct modlinkage modlinkage = { 6281 MODREV_1, 6282 (void *)&zfs_modlfs, 6283 (void *)&zfs_modldrv, 6284 NULL 6285}; 6286#endif /* sun */ 6287 6288static struct cdevsw zfs_cdevsw = { 6289 .d_version = D_VERSION, 6290 .d_open = zfsdev_open, 6291 .d_ioctl = zfsdev_ioctl, 6292 .d_name = ZFS_DEV_NAME 6293}; 6294 6295static void 6296zfs_allow_log_destroy(void *arg) 6297{ 6298 char *poolname = arg; 6299 strfree(poolname); 6300} 6301 6302static void 6303zfsdev_init(void) 6304{ 6305 zfsdev = make_dev(&zfs_cdevsw, 0x0, UID_ROOT, GID_OPERATOR, 0666, 6306 ZFS_DEV_NAME); 6307} 6308 6309static void 6310zfsdev_fini(void) 6311{ 6312 if (zfsdev != NULL) 6313 destroy_dev(zfsdev); 6314} 6315 6316static struct root_hold_token *zfs_root_token; 6317struct proc *zfsproc; 6318 6319#ifdef sun 6320int 6321_init(void) 6322{ 6323 int error; 6324 6325 spa_init(FREAD | FWRITE); 6326 zfs_init(); 6327 zvol_init(); 6328 zfs_ioctl_init(); 6329 6330 if ((error = mod_install(&modlinkage)) != 0) { 6331 zvol_fini(); 6332 zfs_fini(); 6333 spa_fini(); 6334 return (error); 6335 } 6336 6337 tsd_create(&zfs_fsyncer_key, NULL); 6338 tsd_create(&rrw_tsd_key, rrw_tsd_destroy); 6339 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy); 6340 6341 error = ldi_ident_from_mod(&modlinkage, &zfs_li); 6342 ASSERT(error == 0); 6343 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL); 6344 6345 return (0); 6346} 6347 6348int 6349_fini(void) 6350{ 6351 int error; 6352 6353 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled) 6354 return (SET_ERROR(EBUSY)); 6355 6356 if ((error = mod_remove(&modlinkage)) != 0) 6357 return (error); 6358 6359 zvol_fini(); 6360 zfs_fini(); 6361 spa_fini(); 6362 if (zfs_nfsshare_inited) 6363 (void) ddi_modclose(nfs_mod); 6364 if (zfs_smbshare_inited) 6365 (void) ddi_modclose(smbsrv_mod); 6366 if (zfs_nfsshare_inited || zfs_smbshare_inited) 6367 (void) ddi_modclose(sharefs_mod); 6368 6369 tsd_destroy(&zfs_fsyncer_key); 6370 ldi_ident_release(zfs_li); 6371 zfs_li = NULL; 6372 mutex_destroy(&zfs_share_lock); 6373 6374 return (error); 6375} 6376 6377int 6378_info(struct modinfo *modinfop) 6379{ 6380 return (mod_info(&modlinkage, modinfop)); 6381} 6382#endif /* sun */ 6383 6384static int zfs__init(void); 6385static int zfs__fini(void); 6386static void zfs_shutdown(void *, int); 6387 6388static eventhandler_tag zfs_shutdown_event_tag; 6389 6390int 6391zfs__init(void) 6392{ 6393 6394 zfs_root_token = root_mount_hold("ZFS"); 6395 6396 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL); 6397 6398 spa_init(FREAD | FWRITE); 6399 zfs_init(); 6400 zvol_init(); 6401 zfs_ioctl_init(); 6402 6403 tsd_create(&zfs_fsyncer_key, NULL); 6404 tsd_create(&rrw_tsd_key, rrw_tsd_destroy); 6405 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy); 6406 6407 printf("ZFS storage pool version: features support (" SPA_VERSION_STRING ")\n"); 6408 root_mount_rel(zfs_root_token); 6409 6410 zfsdev_init(); 6411 6412 return (0); 6413} 6414 6415int 6416zfs__fini(void) 6417{ 6418 if (spa_busy() || zfs_busy() || zvol_busy() || 6419 zio_injection_enabled) { 6420 return (EBUSY); 6421 } 6422 6423 zfsdev_fini(); 6424 zvol_fini(); 6425 zfs_fini(); 6426 spa_fini(); 6427 6428 tsd_destroy(&zfs_fsyncer_key); 6429 tsd_destroy(&rrw_tsd_key); 6430 tsd_destroy(&zfs_allow_log_key); 6431 6432 mutex_destroy(&zfs_share_lock); 6433 6434 return (0); 6435} 6436 6437static void 6438zfs_shutdown(void *arg __unused, int howto __unused) 6439{ 6440 6441 /* 6442 * ZFS fini routines can not properly work in a panic-ed system. 6443 */ 6444 if (panicstr == NULL) 6445 (void)zfs__fini(); 6446} 6447 6448 6449static int 6450zfs_modevent(module_t mod, int type, void *unused __unused) 6451{ 6452 int err; 6453 6454 switch (type) { 6455 case MOD_LOAD: 6456 err = zfs__init(); 6457 if (err == 0) 6458 zfs_shutdown_event_tag = EVENTHANDLER_REGISTER( 6459 shutdown_post_sync, zfs_shutdown, NULL, 6460 SHUTDOWN_PRI_FIRST); 6461 return (err); 6462 case MOD_UNLOAD: 6463 err = zfs__fini(); 6464 if (err == 0 && zfs_shutdown_event_tag != NULL) 6465 EVENTHANDLER_DEREGISTER(shutdown_post_sync, 6466 zfs_shutdown_event_tag); 6467 return (err); 6468 case MOD_SHUTDOWN: 6469 return (0); 6470 default: 6471 break; 6472 } 6473 return (EOPNOTSUPP); 6474} 6475 6476static moduledata_t zfs_mod = { 6477 "zfsctrl", 6478 zfs_modevent, 6479 0 6480}; 6481DECLARE_MODULE(zfsctrl, zfs_mod, SI_SUB_VFS, SI_ORDER_ANY); 6482MODULE_VERSION(zfsctrl, 1); 6483MODULE_DEPEND(zfsctrl, opensolaris, 1, 1, 1); 6484MODULE_DEPEND(zfsctrl, krpc, 1, 1, 1); 6485MODULE_DEPEND(zfsctrl, acl_nfs4, 1, 1, 1); 6486