zfs_ioctl.c revision 269419
1/* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 22/* 23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 24 * Copyright (c) 2011-2012 Pawel Jakub Dawidek <pawel@dawidek.net>. 25 * All rights reserved. 26 * Copyright 2013 Martin Matuska <mm@FreeBSD.org>. All rights reserved. 27 * Copyright 2014 Xin Li <delphij@FreeBSD.org>. All rights reserved. 28 * Copyright 2011 Nexenta Systems, Inc. All rights reserved. 29 * Copyright (c) 2014, Joyent, Inc. All rights reserved. 30 * Copyright (c) 2011, 2014 by Delphix. All rights reserved. 31 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved. 32 * Copyright (c) 2013 Steven Hartland. All rights reserved. 33 * Copyright (c) 2014, Nexenta Systems, Inc. All rights reserved. 34 */ 35 36/* 37 * ZFS ioctls. 38 * 39 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage 40 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool. 41 * 42 * There are two ways that we handle ioctls: the legacy way where almost 43 * all of the logic is in the ioctl callback, and the new way where most 44 * of the marshalling is handled in the common entry point, zfsdev_ioctl(). 45 * 46 * Non-legacy ioctls should be registered by calling 47 * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked 48 * from userland by lzc_ioctl(). 49 * 50 * The registration arguments are as follows: 51 * 52 * const char *name 53 * The name of the ioctl. This is used for history logging. If the 54 * ioctl returns successfully (the callback returns 0), and allow_log 55 * is true, then a history log entry will be recorded with the input & 56 * output nvlists. The log entry can be printed with "zpool history -i". 57 * 58 * zfs_ioc_t ioc 59 * The ioctl request number, which userland will pass to ioctl(2). 60 * The ioctl numbers can change from release to release, because 61 * the caller (libzfs) must be matched to the kernel. 62 * 63 * zfs_secpolicy_func_t *secpolicy 64 * This function will be called before the zfs_ioc_func_t, to 65 * determine if this operation is permitted. It should return EPERM 66 * on failure, and 0 on success. Checks include determining if the 67 * dataset is visible in this zone, and if the user has either all 68 * zfs privileges in the zone (SYS_MOUNT), or has been granted permission 69 * to do this operation on this dataset with "zfs allow". 70 * 71 * zfs_ioc_namecheck_t namecheck 72 * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool 73 * name, a dataset name, or nothing. If the name is not well-formed, 74 * the ioctl will fail and the callback will not be called. 75 * Therefore, the callback can assume that the name is well-formed 76 * (e.g. is null-terminated, doesn't have more than one '@' character, 77 * doesn't have invalid characters). 78 * 79 * zfs_ioc_poolcheck_t pool_check 80 * This specifies requirements on the pool state. If the pool does 81 * not meet them (is suspended or is readonly), the ioctl will fail 82 * and the callback will not be called. If any checks are specified 83 * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME. 84 * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED | 85 * POOL_CHECK_READONLY). 86 * 87 * boolean_t smush_outnvlist 88 * If smush_outnvlist is true, then the output is presumed to be a 89 * list of errors, and it will be "smushed" down to fit into the 90 * caller's buffer, by removing some entries and replacing them with a 91 * single "N_MORE_ERRORS" entry indicating how many were removed. See 92 * nvlist_smush() for details. If smush_outnvlist is false, and the 93 * outnvlist does not fit into the userland-provided buffer, then the 94 * ioctl will fail with ENOMEM. 95 * 96 * zfs_ioc_func_t *func 97 * The callback function that will perform the operation. 98 * 99 * The callback should return 0 on success, or an error number on 100 * failure. If the function fails, the userland ioctl will return -1, 101 * and errno will be set to the callback's return value. The callback 102 * will be called with the following arguments: 103 * 104 * const char *name 105 * The name of the pool or dataset to operate on, from 106 * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the 107 * expected type (pool, dataset, or none). 108 * 109 * nvlist_t *innvl 110 * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or 111 * NULL if no input nvlist was provided. Changes to this nvlist are 112 * ignored. If the input nvlist could not be deserialized, the 113 * ioctl will fail and the callback will not be called. 114 * 115 * nvlist_t *outnvl 116 * The output nvlist, initially empty. The callback can fill it in, 117 * and it will be returned to userland by serializing it into 118 * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization 119 * fails (e.g. because the caller didn't supply a large enough 120 * buffer), then the overall ioctl will fail. See the 121 * 'smush_nvlist' argument above for additional behaviors. 122 * 123 * There are two typical uses of the output nvlist: 124 * - To return state, e.g. property values. In this case, 125 * smush_outnvlist should be false. If the buffer was not large 126 * enough, the caller will reallocate a larger buffer and try 127 * the ioctl again. 128 * 129 * - To return multiple errors from an ioctl which makes on-disk 130 * changes. In this case, smush_outnvlist should be true. 131 * Ioctls which make on-disk modifications should generally not 132 * use the outnvl if they succeed, because the caller can not 133 * distinguish between the operation failing, and 134 * deserialization failing. 135 */ 136 137#include <sys/types.h> 138#include <sys/param.h> 139#include <sys/systm.h> 140#include <sys/conf.h> 141#include <sys/kernel.h> 142#include <sys/lock.h> 143#include <sys/malloc.h> 144#include <sys/mutex.h> 145#include <sys/proc.h> 146#include <sys/errno.h> 147#include <sys/uio.h> 148#include <sys/buf.h> 149#include <sys/file.h> 150#include <sys/kmem.h> 151#include <sys/conf.h> 152#include <sys/cmn_err.h> 153#include <sys/stat.h> 154#include <sys/zfs_ioctl.h> 155#include <sys/zfs_vfsops.h> 156#include <sys/zfs_znode.h> 157#include <sys/zap.h> 158#include <sys/spa.h> 159#include <sys/spa_impl.h> 160#include <sys/vdev.h> 161#include <sys/dmu.h> 162#include <sys/dsl_dir.h> 163#include <sys/dsl_dataset.h> 164#include <sys/dsl_prop.h> 165#include <sys/dsl_deleg.h> 166#include <sys/dmu_objset.h> 167#include <sys/dmu_impl.h> 168#include <sys/dmu_tx.h> 169#include <sys/sunddi.h> 170#include <sys/policy.h> 171#include <sys/zone.h> 172#include <sys/nvpair.h> 173#include <sys/mount.h> 174#include <sys/taskqueue.h> 175#include <sys/sdt.h> 176#include <sys/varargs.h> 177#include <sys/fs/zfs.h> 178#include <sys/zfs_ctldir.h> 179#include <sys/zfs_dir.h> 180#include <sys/zfs_onexit.h> 181#include <sys/zvol.h> 182#include <sys/dsl_scan.h> 183#include <sys/dmu_objset.h> 184#include <sys/dmu_send.h> 185#include <sys/dsl_destroy.h> 186#include <sys/dsl_bookmark.h> 187#include <sys/dsl_userhold.h> 188#include <sys/zfeature.h> 189 190#include "zfs_namecheck.h" 191#include "zfs_prop.h" 192#include "zfs_deleg.h" 193#include "zfs_comutil.h" 194#include "zfs_ioctl_compat.h" 195 196CTASSERT(sizeof(zfs_cmd_t) < IOCPARM_MAX); 197 198static int snapshot_list_prefetch; 199SYSCTL_DECL(_vfs_zfs); 200TUNABLE_INT("vfs.zfs.snapshot_list_prefetch", &snapshot_list_prefetch); 201SYSCTL_INT(_vfs_zfs, OID_AUTO, snapshot_list_prefetch, CTLFLAG_RW, 202 &snapshot_list_prefetch, 0, "Prefetch data when listing snapshots"); 203 204static struct cdev *zfsdev; 205 206extern void zfs_init(void); 207extern void zfs_fini(void); 208 209uint_t zfs_fsyncer_key; 210extern uint_t rrw_tsd_key; 211static uint_t zfs_allow_log_key; 212 213typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *); 214typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *); 215typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *); 216 217typedef enum { 218 NO_NAME, 219 POOL_NAME, 220 DATASET_NAME 221} zfs_ioc_namecheck_t; 222 223typedef enum { 224 POOL_CHECK_NONE = 1 << 0, 225 POOL_CHECK_SUSPENDED = 1 << 1, 226 POOL_CHECK_READONLY = 1 << 2, 227} zfs_ioc_poolcheck_t; 228 229typedef struct zfs_ioc_vec { 230 zfs_ioc_legacy_func_t *zvec_legacy_func; 231 zfs_ioc_func_t *zvec_func; 232 zfs_secpolicy_func_t *zvec_secpolicy; 233 zfs_ioc_namecheck_t zvec_namecheck; 234 boolean_t zvec_allow_log; 235 zfs_ioc_poolcheck_t zvec_pool_check; 236 boolean_t zvec_smush_outnvlist; 237 const char *zvec_name; 238} zfs_ioc_vec_t; 239 240/* This array is indexed by zfs_userquota_prop_t */ 241static const char *userquota_perms[] = { 242 ZFS_DELEG_PERM_USERUSED, 243 ZFS_DELEG_PERM_USERQUOTA, 244 ZFS_DELEG_PERM_GROUPUSED, 245 ZFS_DELEG_PERM_GROUPQUOTA, 246}; 247 248static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc); 249static int zfs_check_settable(const char *name, nvpair_t *property, 250 cred_t *cr); 251static int zfs_check_clearable(char *dataset, nvlist_t *props, 252 nvlist_t **errors); 253static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *, 254 boolean_t *); 255int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *); 256static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp); 257 258static void zfsdev_close(void *data); 259 260static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature); 261 262/* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */ 263void 264__dprintf(const char *file, const char *func, int line, const char *fmt, ...) 265{ 266 const char *newfile; 267 char buf[512]; 268 va_list adx; 269 270 /* 271 * Get rid of annoying "../common/" prefix to filename. 272 */ 273 newfile = strrchr(file, '/'); 274 if (newfile != NULL) { 275 newfile = newfile + 1; /* Get rid of leading / */ 276 } else { 277 newfile = file; 278 } 279 280 va_start(adx, fmt); 281 (void) vsnprintf(buf, sizeof (buf), fmt, adx); 282 va_end(adx); 283 284 /* 285 * To get this data, use the zfs-dprintf probe as so: 286 * dtrace -q -n 'zfs-dprintf \ 287 * /stringof(arg0) == "dbuf.c"/ \ 288 * {printf("%s: %s", stringof(arg1), stringof(arg3))}' 289 * arg0 = file name 290 * arg1 = function name 291 * arg2 = line number 292 * arg3 = message 293 */ 294 DTRACE_PROBE4(zfs__dprintf, 295 char *, newfile, char *, func, int, line, char *, buf); 296} 297 298static void 299history_str_free(char *buf) 300{ 301 kmem_free(buf, HIS_MAX_RECORD_LEN); 302} 303 304static char * 305history_str_get(zfs_cmd_t *zc) 306{ 307 char *buf; 308 309 if (zc->zc_history == 0) 310 return (NULL); 311 312 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP); 313 if (copyinstr((void *)(uintptr_t)zc->zc_history, 314 buf, HIS_MAX_RECORD_LEN, NULL) != 0) { 315 history_str_free(buf); 316 return (NULL); 317 } 318 319 buf[HIS_MAX_RECORD_LEN -1] = '\0'; 320 321 return (buf); 322} 323 324/* 325 * Check to see if the named dataset is currently defined as bootable 326 */ 327static boolean_t 328zfs_is_bootfs(const char *name) 329{ 330 objset_t *os; 331 332 if (dmu_objset_hold(name, FTAG, &os) == 0) { 333 boolean_t ret; 334 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os))); 335 dmu_objset_rele(os, FTAG); 336 return (ret); 337 } 338 return (B_FALSE); 339} 340 341/* 342 * Return non-zero if the spa version is less than requested version. 343 */ 344static int 345zfs_earlier_version(const char *name, int version) 346{ 347 spa_t *spa; 348 349 if (spa_open(name, &spa, FTAG) == 0) { 350 if (spa_version(spa) < version) { 351 spa_close(spa, FTAG); 352 return (1); 353 } 354 spa_close(spa, FTAG); 355 } 356 return (0); 357} 358 359/* 360 * Return TRUE if the ZPL version is less than requested version. 361 */ 362static boolean_t 363zpl_earlier_version(const char *name, int version) 364{ 365 objset_t *os; 366 boolean_t rc = B_TRUE; 367 368 if (dmu_objset_hold(name, FTAG, &os) == 0) { 369 uint64_t zplversion; 370 371 if (dmu_objset_type(os) != DMU_OST_ZFS) { 372 dmu_objset_rele(os, FTAG); 373 return (B_TRUE); 374 } 375 /* XXX reading from non-owned objset */ 376 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0) 377 rc = zplversion < version; 378 dmu_objset_rele(os, FTAG); 379 } 380 return (rc); 381} 382 383static void 384zfs_log_history(zfs_cmd_t *zc) 385{ 386 spa_t *spa; 387 char *buf; 388 389 if ((buf = history_str_get(zc)) == NULL) 390 return; 391 392 if (spa_open(zc->zc_name, &spa, FTAG) == 0) { 393 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY) 394 (void) spa_history_log(spa, buf); 395 spa_close(spa, FTAG); 396 } 397 history_str_free(buf); 398} 399 400/* 401 * Policy for top-level read operations (list pools). Requires no privileges, 402 * and can be used in the local zone, as there is no associated dataset. 403 */ 404/* ARGSUSED */ 405static int 406zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 407{ 408 return (0); 409} 410 411/* 412 * Policy for dataset read operations (list children, get statistics). Requires 413 * no privileges, but must be visible in the local zone. 414 */ 415/* ARGSUSED */ 416static int 417zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 418{ 419 if (INGLOBALZONE(curthread) || 420 zone_dataset_visible(zc->zc_name, NULL)) 421 return (0); 422 423 return (SET_ERROR(ENOENT)); 424} 425 426static int 427zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr) 428{ 429 int writable = 1; 430 431 /* 432 * The dataset must be visible by this zone -- check this first 433 * so they don't see EPERM on something they shouldn't know about. 434 */ 435 if (!INGLOBALZONE(curthread) && 436 !zone_dataset_visible(dataset, &writable)) 437 return (SET_ERROR(ENOENT)); 438 439 if (INGLOBALZONE(curthread)) { 440 /* 441 * If the fs is zoned, only root can access it from the 442 * global zone. 443 */ 444 if (secpolicy_zfs(cr) && zoned) 445 return (SET_ERROR(EPERM)); 446 } else { 447 /* 448 * If we are in a local zone, the 'zoned' property must be set. 449 */ 450 if (!zoned) 451 return (SET_ERROR(EPERM)); 452 453 /* must be writable by this zone */ 454 if (!writable) 455 return (SET_ERROR(EPERM)); 456 } 457 return (0); 458} 459 460static int 461zfs_dozonecheck(const char *dataset, cred_t *cr) 462{ 463 uint64_t zoned; 464 465 if (dsl_prop_get_integer(dataset, "jailed", &zoned, NULL)) 466 return (SET_ERROR(ENOENT)); 467 468 return (zfs_dozonecheck_impl(dataset, zoned, cr)); 469} 470 471static int 472zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr) 473{ 474 uint64_t zoned; 475 476 if (dsl_prop_get_int_ds(ds, "jailed", &zoned)) 477 return (SET_ERROR(ENOENT)); 478 479 return (zfs_dozonecheck_impl(dataset, zoned, cr)); 480} 481 482static int 483zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds, 484 const char *perm, cred_t *cr) 485{ 486 int error; 487 488 error = zfs_dozonecheck_ds(name, ds, cr); 489 if (error == 0) { 490 error = secpolicy_zfs(cr); 491 if (error != 0) 492 error = dsl_deleg_access_impl(ds, perm, cr); 493 } 494 return (error); 495} 496 497static int 498zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr) 499{ 500 int error; 501 dsl_dataset_t *ds; 502 dsl_pool_t *dp; 503 504 error = dsl_pool_hold(name, FTAG, &dp); 505 if (error != 0) 506 return (error); 507 508 error = dsl_dataset_hold(dp, name, FTAG, &ds); 509 if (error != 0) { 510 dsl_pool_rele(dp, FTAG); 511 return (error); 512 } 513 514 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr); 515 516 dsl_dataset_rele(ds, FTAG); 517 dsl_pool_rele(dp, FTAG); 518 return (error); 519} 520 521#ifdef SECLABEL 522/* 523 * Policy for setting the security label property. 524 * 525 * Returns 0 for success, non-zero for access and other errors. 526 */ 527static int 528zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr) 529{ 530 char ds_hexsl[MAXNAMELEN]; 531 bslabel_t ds_sl, new_sl; 532 boolean_t new_default = FALSE; 533 uint64_t zoned; 534 int needed_priv = -1; 535 int error; 536 537 /* First get the existing dataset label. */ 538 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL), 539 1, sizeof (ds_hexsl), &ds_hexsl, NULL); 540 if (error != 0) 541 return (SET_ERROR(EPERM)); 542 543 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0) 544 new_default = TRUE; 545 546 /* The label must be translatable */ 547 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0)) 548 return (SET_ERROR(EINVAL)); 549 550 /* 551 * In a non-global zone, disallow attempts to set a label that 552 * doesn't match that of the zone; otherwise no other checks 553 * are needed. 554 */ 555 if (!INGLOBALZONE(curproc)) { 556 if (new_default || !blequal(&new_sl, CR_SL(CRED()))) 557 return (SET_ERROR(EPERM)); 558 return (0); 559 } 560 561 /* 562 * For global-zone datasets (i.e., those whose zoned property is 563 * "off", verify that the specified new label is valid for the 564 * global zone. 565 */ 566 if (dsl_prop_get_integer(name, 567 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL)) 568 return (SET_ERROR(EPERM)); 569 if (!zoned) { 570 if (zfs_check_global_label(name, strval) != 0) 571 return (SET_ERROR(EPERM)); 572 } 573 574 /* 575 * If the existing dataset label is nondefault, check if the 576 * dataset is mounted (label cannot be changed while mounted). 577 * Get the zfsvfs; if there isn't one, then the dataset isn't 578 * mounted (or isn't a dataset, doesn't exist, ...). 579 */ 580 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) { 581 objset_t *os; 582 static char *setsl_tag = "setsl_tag"; 583 584 /* 585 * Try to own the dataset; abort if there is any error, 586 * (e.g., already mounted, in use, or other error). 587 */ 588 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE, 589 setsl_tag, &os); 590 if (error != 0) 591 return (SET_ERROR(EPERM)); 592 593 dmu_objset_disown(os, setsl_tag); 594 595 if (new_default) { 596 needed_priv = PRIV_FILE_DOWNGRADE_SL; 597 goto out_check; 598 } 599 600 if (hexstr_to_label(strval, &new_sl) != 0) 601 return (SET_ERROR(EPERM)); 602 603 if (blstrictdom(&ds_sl, &new_sl)) 604 needed_priv = PRIV_FILE_DOWNGRADE_SL; 605 else if (blstrictdom(&new_sl, &ds_sl)) 606 needed_priv = PRIV_FILE_UPGRADE_SL; 607 } else { 608 /* dataset currently has a default label */ 609 if (!new_default) 610 needed_priv = PRIV_FILE_UPGRADE_SL; 611 } 612 613out_check: 614 if (needed_priv != -1) 615 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL)); 616 return (0); 617} 618#endif /* SECLABEL */ 619 620static int 621zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval, 622 cred_t *cr) 623{ 624 char *strval; 625 626 /* 627 * Check permissions for special properties. 628 */ 629 switch (prop) { 630 case ZFS_PROP_ZONED: 631 /* 632 * Disallow setting of 'zoned' from within a local zone. 633 */ 634 if (!INGLOBALZONE(curthread)) 635 return (SET_ERROR(EPERM)); 636 break; 637 638 case ZFS_PROP_QUOTA: 639 case ZFS_PROP_FILESYSTEM_LIMIT: 640 case ZFS_PROP_SNAPSHOT_LIMIT: 641 if (!INGLOBALZONE(curthread)) { 642 uint64_t zoned; 643 char setpoint[MAXNAMELEN]; 644 /* 645 * Unprivileged users are allowed to modify the 646 * limit on things *under* (ie. contained by) 647 * the thing they own. 648 */ 649 if (dsl_prop_get_integer(dsname, "jailed", &zoned, 650 setpoint)) 651 return (SET_ERROR(EPERM)); 652 if (!zoned || strlen(dsname) <= strlen(setpoint)) 653 return (SET_ERROR(EPERM)); 654 } 655 break; 656 657 case ZFS_PROP_MLSLABEL: 658#ifdef SECLABEL 659 if (!is_system_labeled()) 660 return (SET_ERROR(EPERM)); 661 662 if (nvpair_value_string(propval, &strval) == 0) { 663 int err; 664 665 err = zfs_set_slabel_policy(dsname, strval, CRED()); 666 if (err != 0) 667 return (err); 668 } 669#else 670 return (EOPNOTSUPP); 671#endif 672 break; 673 } 674 675 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr)); 676} 677 678/* ARGSUSED */ 679static int 680zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 681{ 682 int error; 683 684 error = zfs_dozonecheck(zc->zc_name, cr); 685 if (error != 0) 686 return (error); 687 688 /* 689 * permission to set permissions will be evaluated later in 690 * dsl_deleg_can_allow() 691 */ 692 return (0); 693} 694 695/* ARGSUSED */ 696static int 697zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 698{ 699 return (zfs_secpolicy_write_perms(zc->zc_name, 700 ZFS_DELEG_PERM_ROLLBACK, cr)); 701} 702 703/* ARGSUSED */ 704static int 705zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 706{ 707 dsl_pool_t *dp; 708 dsl_dataset_t *ds; 709 char *cp; 710 int error; 711 712 /* 713 * Generate the current snapshot name from the given objsetid, then 714 * use that name for the secpolicy/zone checks. 715 */ 716 cp = strchr(zc->zc_name, '@'); 717 if (cp == NULL) 718 return (SET_ERROR(EINVAL)); 719 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 720 if (error != 0) 721 return (error); 722 723 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds); 724 if (error != 0) { 725 dsl_pool_rele(dp, FTAG); 726 return (error); 727 } 728 729 dsl_dataset_name(ds, zc->zc_name); 730 731 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds, 732 ZFS_DELEG_PERM_SEND, cr); 733 dsl_dataset_rele(ds, FTAG); 734 dsl_pool_rele(dp, FTAG); 735 736 return (error); 737} 738 739/* ARGSUSED */ 740static int 741zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 742{ 743 return (zfs_secpolicy_write_perms(zc->zc_name, 744 ZFS_DELEG_PERM_SEND, cr)); 745} 746 747/* ARGSUSED */ 748static int 749zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 750{ 751 vnode_t *vp; 752 int error; 753 754 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, 755 NO_FOLLOW, NULL, &vp)) != 0) 756 return (error); 757 758 /* Now make sure mntpnt and dataset are ZFS */ 759 760 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 || 761 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), 762 zc->zc_name) != 0)) { 763 VN_RELE(vp); 764 return (SET_ERROR(EPERM)); 765 } 766 767 VN_RELE(vp); 768 return (dsl_deleg_access(zc->zc_name, 769 ZFS_DELEG_PERM_SHARE, cr)); 770} 771 772int 773zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 774{ 775 if (!INGLOBALZONE(curthread)) 776 return (SET_ERROR(EPERM)); 777 778 if (secpolicy_nfs(cr) == 0) { 779 return (0); 780 } else { 781 return (zfs_secpolicy_deleg_share(zc, innvl, cr)); 782 } 783} 784 785int 786zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 787{ 788 if (!INGLOBALZONE(curthread)) 789 return (SET_ERROR(EPERM)); 790 791 if (secpolicy_smb(cr) == 0) { 792 return (0); 793 } else { 794 return (zfs_secpolicy_deleg_share(zc, innvl, cr)); 795 } 796} 797 798static int 799zfs_get_parent(const char *datasetname, char *parent, int parentsize) 800{ 801 char *cp; 802 803 /* 804 * Remove the @bla or /bla from the end of the name to get the parent. 805 */ 806 (void) strncpy(parent, datasetname, parentsize); 807 cp = strrchr(parent, '@'); 808 if (cp != NULL) { 809 cp[0] = '\0'; 810 } else { 811 cp = strrchr(parent, '/'); 812 if (cp == NULL) 813 return (SET_ERROR(ENOENT)); 814 cp[0] = '\0'; 815 } 816 817 return (0); 818} 819 820int 821zfs_secpolicy_destroy_perms(const char *name, cred_t *cr) 822{ 823 int error; 824 825 if ((error = zfs_secpolicy_write_perms(name, 826 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 827 return (error); 828 829 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr)); 830} 831 832/* ARGSUSED */ 833static int 834zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 835{ 836 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr)); 837} 838 839/* 840 * Destroying snapshots with delegated permissions requires 841 * descendant mount and destroy permissions. 842 */ 843/* ARGSUSED */ 844static int 845zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 846{ 847 nvlist_t *snaps; 848 nvpair_t *pair, *nextpair; 849 int error = 0; 850 851 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 852 return (SET_ERROR(EINVAL)); 853 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 854 pair = nextpair) { 855 nextpair = nvlist_next_nvpair(snaps, pair); 856 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr); 857 if (error == ENOENT) { 858 /* 859 * Ignore any snapshots that don't exist (we consider 860 * them "already destroyed"). Remove the name from the 861 * nvl here in case the snapshot is created between 862 * now and when we try to destroy it (in which case 863 * we don't want to destroy it since we haven't 864 * checked for permission). 865 */ 866 fnvlist_remove_nvpair(snaps, pair); 867 error = 0; 868 } 869 if (error != 0) 870 break; 871 } 872 873 return (error); 874} 875 876int 877zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr) 878{ 879 char parentname[MAXNAMELEN]; 880 int error; 881 882 if ((error = zfs_secpolicy_write_perms(from, 883 ZFS_DELEG_PERM_RENAME, cr)) != 0) 884 return (error); 885 886 if ((error = zfs_secpolicy_write_perms(from, 887 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 888 return (error); 889 890 if ((error = zfs_get_parent(to, parentname, 891 sizeof (parentname))) != 0) 892 return (error); 893 894 if ((error = zfs_secpolicy_write_perms(parentname, 895 ZFS_DELEG_PERM_CREATE, cr)) != 0) 896 return (error); 897 898 if ((error = zfs_secpolicy_write_perms(parentname, 899 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 900 return (error); 901 902 return (error); 903} 904 905/* ARGSUSED */ 906static int 907zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 908{ 909 char *at = NULL; 910 int error; 911 912 if ((zc->zc_cookie & 1) != 0) { 913 /* 914 * This is recursive rename, so the starting snapshot might 915 * not exist. Check file system or volume permission instead. 916 */ 917 at = strchr(zc->zc_name, '@'); 918 if (at == NULL) 919 return (EINVAL); 920 *at = '\0'; 921 } 922 923 error = zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr); 924 925 if (at != NULL) 926 *at = '@'; 927 928 return (error); 929} 930 931/* ARGSUSED */ 932static int 933zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 934{ 935 dsl_pool_t *dp; 936 dsl_dataset_t *clone; 937 int error; 938 939 error = zfs_secpolicy_write_perms(zc->zc_name, 940 ZFS_DELEG_PERM_PROMOTE, cr); 941 if (error != 0) 942 return (error); 943 944 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 945 if (error != 0) 946 return (error); 947 948 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone); 949 950 if (error == 0) { 951 char parentname[MAXNAMELEN]; 952 dsl_dataset_t *origin = NULL; 953 dsl_dir_t *dd; 954 dd = clone->ds_dir; 955 956 error = dsl_dataset_hold_obj(dd->dd_pool, 957 dd->dd_phys->dd_origin_obj, FTAG, &origin); 958 if (error != 0) { 959 dsl_dataset_rele(clone, FTAG); 960 dsl_pool_rele(dp, FTAG); 961 return (error); 962 } 963 964 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone, 965 ZFS_DELEG_PERM_MOUNT, cr); 966 967 dsl_dataset_name(origin, parentname); 968 if (error == 0) { 969 error = zfs_secpolicy_write_perms_ds(parentname, origin, 970 ZFS_DELEG_PERM_PROMOTE, cr); 971 } 972 dsl_dataset_rele(clone, FTAG); 973 dsl_dataset_rele(origin, FTAG); 974 } 975 dsl_pool_rele(dp, FTAG); 976 return (error); 977} 978 979/* ARGSUSED */ 980static int 981zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 982{ 983 int error; 984 985 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 986 ZFS_DELEG_PERM_RECEIVE, cr)) != 0) 987 return (error); 988 989 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 990 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 991 return (error); 992 993 return (zfs_secpolicy_write_perms(zc->zc_name, 994 ZFS_DELEG_PERM_CREATE, cr)); 995} 996 997int 998zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr) 999{ 1000 return (zfs_secpolicy_write_perms(name, 1001 ZFS_DELEG_PERM_SNAPSHOT, cr)); 1002} 1003 1004/* 1005 * Check for permission to create each snapshot in the nvlist. 1006 */ 1007/* ARGSUSED */ 1008static int 1009zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1010{ 1011 nvlist_t *snaps; 1012 int error; 1013 nvpair_t *pair; 1014 1015 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 1016 return (SET_ERROR(EINVAL)); 1017 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 1018 pair = nvlist_next_nvpair(snaps, pair)) { 1019 char *name = nvpair_name(pair); 1020 char *atp = strchr(name, '@'); 1021 1022 if (atp == NULL) { 1023 error = SET_ERROR(EINVAL); 1024 break; 1025 } 1026 *atp = '\0'; 1027 error = zfs_secpolicy_snapshot_perms(name, cr); 1028 *atp = '@'; 1029 if (error != 0) 1030 break; 1031 } 1032 return (error); 1033} 1034 1035/* 1036 * Check for permission to create each snapshot in the nvlist. 1037 */ 1038/* ARGSUSED */ 1039static int 1040zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1041{ 1042 int error = 0; 1043 1044 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 1045 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 1046 char *name = nvpair_name(pair); 1047 char *hashp = strchr(name, '#'); 1048 1049 if (hashp == NULL) { 1050 error = SET_ERROR(EINVAL); 1051 break; 1052 } 1053 *hashp = '\0'; 1054 error = zfs_secpolicy_write_perms(name, 1055 ZFS_DELEG_PERM_BOOKMARK, cr); 1056 *hashp = '#'; 1057 if (error != 0) 1058 break; 1059 } 1060 return (error); 1061} 1062 1063/* ARGSUSED */ 1064static int 1065zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1066{ 1067 nvpair_t *pair, *nextpair; 1068 int error = 0; 1069 1070 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL; 1071 pair = nextpair) { 1072 char *name = nvpair_name(pair); 1073 char *hashp = strchr(name, '#'); 1074 nextpair = nvlist_next_nvpair(innvl, pair); 1075 1076 if (hashp == NULL) { 1077 error = SET_ERROR(EINVAL); 1078 break; 1079 } 1080 1081 *hashp = '\0'; 1082 error = zfs_secpolicy_write_perms(name, 1083 ZFS_DELEG_PERM_DESTROY, cr); 1084 *hashp = '#'; 1085 if (error == ENOENT) { 1086 /* 1087 * Ignore any filesystems that don't exist (we consider 1088 * their bookmarks "already destroyed"). Remove 1089 * the name from the nvl here in case the filesystem 1090 * is created between now and when we try to destroy 1091 * the bookmark (in which case we don't want to 1092 * destroy it since we haven't checked for permission). 1093 */ 1094 fnvlist_remove_nvpair(innvl, pair); 1095 error = 0; 1096 } 1097 if (error != 0) 1098 break; 1099 } 1100 1101 return (error); 1102} 1103 1104/* ARGSUSED */ 1105static int 1106zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1107{ 1108 /* 1109 * Even root must have a proper TSD so that we know what pool 1110 * to log to. 1111 */ 1112 if (tsd_get(zfs_allow_log_key) == NULL) 1113 return (SET_ERROR(EPERM)); 1114 return (0); 1115} 1116 1117static int 1118zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1119{ 1120 char parentname[MAXNAMELEN]; 1121 int error; 1122 char *origin; 1123 1124 if ((error = zfs_get_parent(zc->zc_name, parentname, 1125 sizeof (parentname))) != 0) 1126 return (error); 1127 1128 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 && 1129 (error = zfs_secpolicy_write_perms(origin, 1130 ZFS_DELEG_PERM_CLONE, cr)) != 0) 1131 return (error); 1132 1133 if ((error = zfs_secpolicy_write_perms(parentname, 1134 ZFS_DELEG_PERM_CREATE, cr)) != 0) 1135 return (error); 1136 1137 return (zfs_secpolicy_write_perms(parentname, 1138 ZFS_DELEG_PERM_MOUNT, cr)); 1139} 1140 1141/* 1142 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires 1143 * SYS_CONFIG privilege, which is not available in a local zone. 1144 */ 1145/* ARGSUSED */ 1146static int 1147zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1148{ 1149 if (secpolicy_sys_config(cr, B_FALSE) != 0) 1150 return (SET_ERROR(EPERM)); 1151 1152 return (0); 1153} 1154 1155/* 1156 * Policy for object to name lookups. 1157 */ 1158/* ARGSUSED */ 1159static int 1160zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1161{ 1162 int error; 1163 1164 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0) 1165 return (0); 1166 1167 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr); 1168 return (error); 1169} 1170 1171/* 1172 * Policy for fault injection. Requires all privileges. 1173 */ 1174/* ARGSUSED */ 1175static int 1176zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1177{ 1178 return (secpolicy_zinject(cr)); 1179} 1180 1181/* ARGSUSED */ 1182static int 1183zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1184{ 1185 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value); 1186 1187 if (prop == ZPROP_INVAL) { 1188 if (!zfs_prop_user(zc->zc_value)) 1189 return (SET_ERROR(EINVAL)); 1190 return (zfs_secpolicy_write_perms(zc->zc_name, 1191 ZFS_DELEG_PERM_USERPROP, cr)); 1192 } else { 1193 return (zfs_secpolicy_setprop(zc->zc_name, prop, 1194 NULL, cr)); 1195 } 1196} 1197 1198static int 1199zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1200{ 1201 int err = zfs_secpolicy_read(zc, innvl, cr); 1202 if (err) 1203 return (err); 1204 1205 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 1206 return (SET_ERROR(EINVAL)); 1207 1208 if (zc->zc_value[0] == 0) { 1209 /* 1210 * They are asking about a posix uid/gid. If it's 1211 * themself, allow it. 1212 */ 1213 if (zc->zc_objset_type == ZFS_PROP_USERUSED || 1214 zc->zc_objset_type == ZFS_PROP_USERQUOTA) { 1215 if (zc->zc_guid == crgetuid(cr)) 1216 return (0); 1217 } else { 1218 if (groupmember(zc->zc_guid, cr)) 1219 return (0); 1220 } 1221 } 1222 1223 return (zfs_secpolicy_write_perms(zc->zc_name, 1224 userquota_perms[zc->zc_objset_type], cr)); 1225} 1226 1227static int 1228zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1229{ 1230 int err = zfs_secpolicy_read(zc, innvl, cr); 1231 if (err) 1232 return (err); 1233 1234 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 1235 return (SET_ERROR(EINVAL)); 1236 1237 return (zfs_secpolicy_write_perms(zc->zc_name, 1238 userquota_perms[zc->zc_objset_type], cr)); 1239} 1240 1241/* ARGSUSED */ 1242static int 1243zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1244{ 1245 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION, 1246 NULL, cr)); 1247} 1248 1249/* ARGSUSED */ 1250static int 1251zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1252{ 1253 nvpair_t *pair; 1254 nvlist_t *holds; 1255 int error; 1256 1257 error = nvlist_lookup_nvlist(innvl, "holds", &holds); 1258 if (error != 0) 1259 return (SET_ERROR(EINVAL)); 1260 1261 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL; 1262 pair = nvlist_next_nvpair(holds, pair)) { 1263 char fsname[MAXNAMELEN]; 1264 error = dmu_fsname(nvpair_name(pair), fsname); 1265 if (error != 0) 1266 return (error); 1267 error = zfs_secpolicy_write_perms(fsname, 1268 ZFS_DELEG_PERM_HOLD, cr); 1269 if (error != 0) 1270 return (error); 1271 } 1272 return (0); 1273} 1274 1275/* ARGSUSED */ 1276static int 1277zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1278{ 1279 nvpair_t *pair; 1280 int error; 1281 1282 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL; 1283 pair = nvlist_next_nvpair(innvl, pair)) { 1284 char fsname[MAXNAMELEN]; 1285 error = dmu_fsname(nvpair_name(pair), fsname); 1286 if (error != 0) 1287 return (error); 1288 error = zfs_secpolicy_write_perms(fsname, 1289 ZFS_DELEG_PERM_RELEASE, cr); 1290 if (error != 0) 1291 return (error); 1292 } 1293 return (0); 1294} 1295 1296/* 1297 * Policy for allowing temporary snapshots to be taken or released 1298 */ 1299static int 1300zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1301{ 1302 /* 1303 * A temporary snapshot is the same as a snapshot, 1304 * hold, destroy and release all rolled into one. 1305 * Delegated diff alone is sufficient that we allow this. 1306 */ 1307 int error; 1308 1309 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 1310 ZFS_DELEG_PERM_DIFF, cr)) == 0) 1311 return (0); 1312 1313 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr); 1314 if (error == 0) 1315 error = zfs_secpolicy_hold(zc, innvl, cr); 1316 if (error == 0) 1317 error = zfs_secpolicy_release(zc, innvl, cr); 1318 if (error == 0) 1319 error = zfs_secpolicy_destroy(zc, innvl, cr); 1320 return (error); 1321} 1322 1323/* 1324 * Returns the nvlist as specified by the user in the zfs_cmd_t. 1325 */ 1326static int 1327get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp) 1328{ 1329 char *packed; 1330 int error; 1331 nvlist_t *list = NULL; 1332 1333 /* 1334 * Read in and unpack the user-supplied nvlist. 1335 */ 1336 if (size == 0) 1337 return (SET_ERROR(EINVAL)); 1338 1339 packed = kmem_alloc(size, KM_SLEEP); 1340 1341 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size, 1342 iflag)) != 0) { 1343 kmem_free(packed, size); 1344 return (error); 1345 } 1346 1347 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) { 1348 kmem_free(packed, size); 1349 return (error); 1350 } 1351 1352 kmem_free(packed, size); 1353 1354 *nvp = list; 1355 return (0); 1356} 1357 1358/* 1359 * Reduce the size of this nvlist until it can be serialized in 'max' bytes. 1360 * Entries will be removed from the end of the nvlist, and one int32 entry 1361 * named "N_MORE_ERRORS" will be added indicating how many entries were 1362 * removed. 1363 */ 1364static int 1365nvlist_smush(nvlist_t *errors, size_t max) 1366{ 1367 size_t size; 1368 1369 size = fnvlist_size(errors); 1370 1371 if (size > max) { 1372 nvpair_t *more_errors; 1373 int n = 0; 1374 1375 if (max < 1024) 1376 return (SET_ERROR(ENOMEM)); 1377 1378 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0); 1379 more_errors = nvlist_prev_nvpair(errors, NULL); 1380 1381 do { 1382 nvpair_t *pair = nvlist_prev_nvpair(errors, 1383 more_errors); 1384 fnvlist_remove_nvpair(errors, pair); 1385 n++; 1386 size = fnvlist_size(errors); 1387 } while (size > max); 1388 1389 fnvlist_remove_nvpair(errors, more_errors); 1390 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n); 1391 ASSERT3U(fnvlist_size(errors), <=, max); 1392 } 1393 1394 return (0); 1395} 1396 1397static int 1398put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl) 1399{ 1400 char *packed = NULL; 1401 int error = 0; 1402 size_t size; 1403 1404 size = fnvlist_size(nvl); 1405 1406 if (size > zc->zc_nvlist_dst_size) { 1407 /* 1408 * Solaris returns ENOMEM here, because even if an error is 1409 * returned from an ioctl(2), new zc_nvlist_dst_size will be 1410 * passed to the userland. This is not the case for FreeBSD. 1411 * We need to return 0, so the kernel will copy the 1412 * zc_nvlist_dst_size back and the userland can discover that a 1413 * bigger buffer is needed. 1414 */ 1415 error = 0; 1416 } else { 1417 packed = fnvlist_pack(nvl, &size); 1418 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst, 1419 size, zc->zc_iflags) != 0) 1420 error = SET_ERROR(EFAULT); 1421 fnvlist_pack_free(packed, size); 1422 } 1423 1424 zc->zc_nvlist_dst_size = size; 1425 zc->zc_nvlist_dst_filled = B_TRUE; 1426 return (error); 1427} 1428 1429static int 1430getzfsvfs(const char *dsname, zfsvfs_t **zfvp) 1431{ 1432 objset_t *os; 1433 int error; 1434 1435 error = dmu_objset_hold(dsname, FTAG, &os); 1436 if (error != 0) 1437 return (error); 1438 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1439 dmu_objset_rele(os, FTAG); 1440 return (SET_ERROR(EINVAL)); 1441 } 1442 1443 mutex_enter(&os->os_user_ptr_lock); 1444 *zfvp = dmu_objset_get_user(os); 1445 if (*zfvp) { 1446 VFS_HOLD((*zfvp)->z_vfs); 1447 } else { 1448 error = SET_ERROR(ESRCH); 1449 } 1450 mutex_exit(&os->os_user_ptr_lock); 1451 dmu_objset_rele(os, FTAG); 1452 return (error); 1453} 1454 1455/* 1456 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which 1457 * case its z_vfs will be NULL, and it will be opened as the owner. 1458 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER, 1459 * which prevents all vnode ops from running. 1460 */ 1461static int 1462zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer) 1463{ 1464 int error = 0; 1465 1466 if (getzfsvfs(name, zfvp) != 0) 1467 error = zfsvfs_create(name, zfvp); 1468 if (error == 0) { 1469 rrm_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER : 1470 RW_READER, tag); 1471 if ((*zfvp)->z_unmounted) { 1472 /* 1473 * XXX we could probably try again, since the unmounting 1474 * thread should be just about to disassociate the 1475 * objset from the zfsvfs. 1476 */ 1477 rrm_exit(&(*zfvp)->z_teardown_lock, tag); 1478 return (SET_ERROR(EBUSY)); 1479 } 1480 } 1481 return (error); 1482} 1483 1484static void 1485zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag) 1486{ 1487 rrm_exit(&zfsvfs->z_teardown_lock, tag); 1488 1489 if (zfsvfs->z_vfs) { 1490 VFS_RELE(zfsvfs->z_vfs); 1491 } else { 1492 dmu_objset_disown(zfsvfs->z_os, zfsvfs); 1493 zfsvfs_free(zfsvfs); 1494 } 1495} 1496 1497static int 1498zfs_ioc_pool_create(zfs_cmd_t *zc) 1499{ 1500 int error; 1501 nvlist_t *config, *props = NULL; 1502 nvlist_t *rootprops = NULL; 1503 nvlist_t *zplprops = NULL; 1504 1505 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1506 zc->zc_iflags, &config)) 1507 return (error); 1508 1509 if (zc->zc_nvlist_src_size != 0 && (error = 1510 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1511 zc->zc_iflags, &props))) { 1512 nvlist_free(config); 1513 return (error); 1514 } 1515 1516 if (props) { 1517 nvlist_t *nvl = NULL; 1518 uint64_t version = SPA_VERSION; 1519 1520 (void) nvlist_lookup_uint64(props, 1521 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version); 1522 if (!SPA_VERSION_IS_SUPPORTED(version)) { 1523 error = SET_ERROR(EINVAL); 1524 goto pool_props_bad; 1525 } 1526 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl); 1527 if (nvl) { 1528 error = nvlist_dup(nvl, &rootprops, KM_SLEEP); 1529 if (error != 0) { 1530 nvlist_free(config); 1531 nvlist_free(props); 1532 return (error); 1533 } 1534 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS); 1535 } 1536 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 1537 error = zfs_fill_zplprops_root(version, rootprops, 1538 zplprops, NULL); 1539 if (error != 0) 1540 goto pool_props_bad; 1541 } 1542 1543 error = spa_create(zc->zc_name, config, props, zplprops); 1544 1545 /* 1546 * Set the remaining root properties 1547 */ 1548 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name, 1549 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0) 1550 (void) spa_destroy(zc->zc_name); 1551 1552pool_props_bad: 1553 nvlist_free(rootprops); 1554 nvlist_free(zplprops); 1555 nvlist_free(config); 1556 nvlist_free(props); 1557 1558 return (error); 1559} 1560 1561static int 1562zfs_ioc_pool_destroy(zfs_cmd_t *zc) 1563{ 1564 int error; 1565 zfs_log_history(zc); 1566 error = spa_destroy(zc->zc_name); 1567 if (error == 0) 1568 zvol_remove_minors(zc->zc_name); 1569 return (error); 1570} 1571 1572static int 1573zfs_ioc_pool_import(zfs_cmd_t *zc) 1574{ 1575 nvlist_t *config, *props = NULL; 1576 uint64_t guid; 1577 int error; 1578 1579 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1580 zc->zc_iflags, &config)) != 0) 1581 return (error); 1582 1583 if (zc->zc_nvlist_src_size != 0 && (error = 1584 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1585 zc->zc_iflags, &props))) { 1586 nvlist_free(config); 1587 return (error); 1588 } 1589 1590 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 || 1591 guid != zc->zc_guid) 1592 error = SET_ERROR(EINVAL); 1593 else 1594 error = spa_import(zc->zc_name, config, props, zc->zc_cookie); 1595 1596 if (zc->zc_nvlist_dst != 0) { 1597 int err; 1598 1599 if ((err = put_nvlist(zc, config)) != 0) 1600 error = err; 1601 } 1602 1603 nvlist_free(config); 1604 1605 if (props) 1606 nvlist_free(props); 1607 1608 return (error); 1609} 1610 1611static int 1612zfs_ioc_pool_export(zfs_cmd_t *zc) 1613{ 1614 int error; 1615 boolean_t force = (boolean_t)zc->zc_cookie; 1616 boolean_t hardforce = (boolean_t)zc->zc_guid; 1617 1618 zfs_log_history(zc); 1619 error = spa_export(zc->zc_name, NULL, force, hardforce); 1620 if (error == 0) 1621 zvol_remove_minors(zc->zc_name); 1622 return (error); 1623} 1624 1625static int 1626zfs_ioc_pool_configs(zfs_cmd_t *zc) 1627{ 1628 nvlist_t *configs; 1629 int error; 1630 1631 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL) 1632 return (SET_ERROR(EEXIST)); 1633 1634 error = put_nvlist(zc, configs); 1635 1636 nvlist_free(configs); 1637 1638 return (error); 1639} 1640 1641/* 1642 * inputs: 1643 * zc_name name of the pool 1644 * 1645 * outputs: 1646 * zc_cookie real errno 1647 * zc_nvlist_dst config nvlist 1648 * zc_nvlist_dst_size size of config nvlist 1649 */ 1650static int 1651zfs_ioc_pool_stats(zfs_cmd_t *zc) 1652{ 1653 nvlist_t *config; 1654 int error; 1655 int ret = 0; 1656 1657 error = spa_get_stats(zc->zc_name, &config, zc->zc_value, 1658 sizeof (zc->zc_value)); 1659 1660 if (config != NULL) { 1661 ret = put_nvlist(zc, config); 1662 nvlist_free(config); 1663 1664 /* 1665 * The config may be present even if 'error' is non-zero. 1666 * In this case we return success, and preserve the real errno 1667 * in 'zc_cookie'. 1668 */ 1669 zc->zc_cookie = error; 1670 } else { 1671 ret = error; 1672 } 1673 1674 return (ret); 1675} 1676 1677/* 1678 * Try to import the given pool, returning pool stats as appropriate so that 1679 * user land knows which devices are available and overall pool health. 1680 */ 1681static int 1682zfs_ioc_pool_tryimport(zfs_cmd_t *zc) 1683{ 1684 nvlist_t *tryconfig, *config; 1685 int error; 1686 1687 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1688 zc->zc_iflags, &tryconfig)) != 0) 1689 return (error); 1690 1691 config = spa_tryimport(tryconfig); 1692 1693 nvlist_free(tryconfig); 1694 1695 if (config == NULL) 1696 return (SET_ERROR(EINVAL)); 1697 1698 error = put_nvlist(zc, config); 1699 nvlist_free(config); 1700 1701 return (error); 1702} 1703 1704/* 1705 * inputs: 1706 * zc_name name of the pool 1707 * zc_cookie scan func (pool_scan_func_t) 1708 */ 1709static int 1710zfs_ioc_pool_scan(zfs_cmd_t *zc) 1711{ 1712 spa_t *spa; 1713 int error; 1714 1715 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1716 return (error); 1717 1718 if (zc->zc_cookie == POOL_SCAN_NONE) 1719 error = spa_scan_stop(spa); 1720 else 1721 error = spa_scan(spa, zc->zc_cookie); 1722 1723 spa_close(spa, FTAG); 1724 1725 return (error); 1726} 1727 1728static int 1729zfs_ioc_pool_freeze(zfs_cmd_t *zc) 1730{ 1731 spa_t *spa; 1732 int error; 1733 1734 error = spa_open(zc->zc_name, &spa, FTAG); 1735 if (error == 0) { 1736 spa_freeze(spa); 1737 spa_close(spa, FTAG); 1738 } 1739 return (error); 1740} 1741 1742static int 1743zfs_ioc_pool_upgrade(zfs_cmd_t *zc) 1744{ 1745 spa_t *spa; 1746 int error; 1747 1748 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1749 return (error); 1750 1751 if (zc->zc_cookie < spa_version(spa) || 1752 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) { 1753 spa_close(spa, FTAG); 1754 return (SET_ERROR(EINVAL)); 1755 } 1756 1757 spa_upgrade(spa, zc->zc_cookie); 1758 spa_close(spa, FTAG); 1759 1760 return (error); 1761} 1762 1763static int 1764zfs_ioc_pool_get_history(zfs_cmd_t *zc) 1765{ 1766 spa_t *spa; 1767 char *hist_buf; 1768 uint64_t size; 1769 int error; 1770 1771 if ((size = zc->zc_history_len) == 0) 1772 return (SET_ERROR(EINVAL)); 1773 1774 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1775 return (error); 1776 1777 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) { 1778 spa_close(spa, FTAG); 1779 return (SET_ERROR(ENOTSUP)); 1780 } 1781 1782 hist_buf = kmem_alloc(size, KM_SLEEP); 1783 if ((error = spa_history_get(spa, &zc->zc_history_offset, 1784 &zc->zc_history_len, hist_buf)) == 0) { 1785 error = ddi_copyout(hist_buf, 1786 (void *)(uintptr_t)zc->zc_history, 1787 zc->zc_history_len, zc->zc_iflags); 1788 } 1789 1790 spa_close(spa, FTAG); 1791 kmem_free(hist_buf, size); 1792 return (error); 1793} 1794 1795static int 1796zfs_ioc_pool_reguid(zfs_cmd_t *zc) 1797{ 1798 spa_t *spa; 1799 int error; 1800 1801 error = spa_open(zc->zc_name, &spa, FTAG); 1802 if (error == 0) { 1803 error = spa_change_guid(spa); 1804 spa_close(spa, FTAG); 1805 } 1806 return (error); 1807} 1808 1809static int 1810zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc) 1811{ 1812 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value)); 1813} 1814 1815/* 1816 * inputs: 1817 * zc_name name of filesystem 1818 * zc_obj object to find 1819 * 1820 * outputs: 1821 * zc_value name of object 1822 */ 1823static int 1824zfs_ioc_obj_to_path(zfs_cmd_t *zc) 1825{ 1826 objset_t *os; 1827 int error; 1828 1829 /* XXX reading from objset not owned */ 1830 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0) 1831 return (error); 1832 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1833 dmu_objset_rele(os, FTAG); 1834 return (SET_ERROR(EINVAL)); 1835 } 1836 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value, 1837 sizeof (zc->zc_value)); 1838 dmu_objset_rele(os, FTAG); 1839 1840 return (error); 1841} 1842 1843/* 1844 * inputs: 1845 * zc_name name of filesystem 1846 * zc_obj object to find 1847 * 1848 * outputs: 1849 * zc_stat stats on object 1850 * zc_value path to object 1851 */ 1852static int 1853zfs_ioc_obj_to_stats(zfs_cmd_t *zc) 1854{ 1855 objset_t *os; 1856 int error; 1857 1858 /* XXX reading from objset not owned */ 1859 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0) 1860 return (error); 1861 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1862 dmu_objset_rele(os, FTAG); 1863 return (SET_ERROR(EINVAL)); 1864 } 1865 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value, 1866 sizeof (zc->zc_value)); 1867 dmu_objset_rele(os, FTAG); 1868 1869 return (error); 1870} 1871 1872static int 1873zfs_ioc_vdev_add(zfs_cmd_t *zc) 1874{ 1875 spa_t *spa; 1876 int error; 1877 nvlist_t *config, **l2cache, **spares; 1878 uint_t nl2cache = 0, nspares = 0; 1879 1880 error = spa_open(zc->zc_name, &spa, FTAG); 1881 if (error != 0) 1882 return (error); 1883 1884 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1885 zc->zc_iflags, &config); 1886 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE, 1887 &l2cache, &nl2cache); 1888 1889 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES, 1890 &spares, &nspares); 1891 1892#ifdef illumos 1893 /* 1894 * A root pool with concatenated devices is not supported. 1895 * Thus, can not add a device to a root pool. 1896 * 1897 * Intent log device can not be added to a rootpool because 1898 * during mountroot, zil is replayed, a seperated log device 1899 * can not be accessed during the mountroot time. 1900 * 1901 * l2cache and spare devices are ok to be added to a rootpool. 1902 */ 1903 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) { 1904 nvlist_free(config); 1905 spa_close(spa, FTAG); 1906 return (SET_ERROR(EDOM)); 1907 } 1908#endif /* illumos */ 1909 1910 if (error == 0) { 1911 error = spa_vdev_add(spa, config); 1912 nvlist_free(config); 1913 } 1914 spa_close(spa, FTAG); 1915 return (error); 1916} 1917 1918/* 1919 * inputs: 1920 * zc_name name of the pool 1921 * zc_nvlist_conf nvlist of devices to remove 1922 * zc_cookie to stop the remove? 1923 */ 1924static int 1925zfs_ioc_vdev_remove(zfs_cmd_t *zc) 1926{ 1927 spa_t *spa; 1928 int error; 1929 1930 error = spa_open(zc->zc_name, &spa, FTAG); 1931 if (error != 0) 1932 return (error); 1933 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE); 1934 spa_close(spa, FTAG); 1935 return (error); 1936} 1937 1938static int 1939zfs_ioc_vdev_set_state(zfs_cmd_t *zc) 1940{ 1941 spa_t *spa; 1942 int error; 1943 vdev_state_t newstate = VDEV_STATE_UNKNOWN; 1944 1945 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1946 return (error); 1947 switch (zc->zc_cookie) { 1948 case VDEV_STATE_ONLINE: 1949 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate); 1950 break; 1951 1952 case VDEV_STATE_OFFLINE: 1953 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj); 1954 break; 1955 1956 case VDEV_STATE_FAULTED: 1957 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED && 1958 zc->zc_obj != VDEV_AUX_EXTERNAL) 1959 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED; 1960 1961 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj); 1962 break; 1963 1964 case VDEV_STATE_DEGRADED: 1965 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED && 1966 zc->zc_obj != VDEV_AUX_EXTERNAL) 1967 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED; 1968 1969 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj); 1970 break; 1971 1972 default: 1973 error = SET_ERROR(EINVAL); 1974 } 1975 zc->zc_cookie = newstate; 1976 spa_close(spa, FTAG); 1977 return (error); 1978} 1979 1980static int 1981zfs_ioc_vdev_attach(zfs_cmd_t *zc) 1982{ 1983 spa_t *spa; 1984 int replacing = zc->zc_cookie; 1985 nvlist_t *config; 1986 int error; 1987 1988 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1989 return (error); 1990 1991 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1992 zc->zc_iflags, &config)) == 0) { 1993 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing); 1994 nvlist_free(config); 1995 } 1996 1997 spa_close(spa, FTAG); 1998 return (error); 1999} 2000 2001static int 2002zfs_ioc_vdev_detach(zfs_cmd_t *zc) 2003{ 2004 spa_t *spa; 2005 int error; 2006 2007 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 2008 return (error); 2009 2010 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE); 2011 2012 spa_close(spa, FTAG); 2013 return (error); 2014} 2015 2016static int 2017zfs_ioc_vdev_split(zfs_cmd_t *zc) 2018{ 2019 spa_t *spa; 2020 nvlist_t *config, *props = NULL; 2021 int error; 2022 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT); 2023 2024 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 2025 return (error); 2026 2027 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 2028 zc->zc_iflags, &config)) { 2029 spa_close(spa, FTAG); 2030 return (error); 2031 } 2032 2033 if (zc->zc_nvlist_src_size != 0 && (error = 2034 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2035 zc->zc_iflags, &props))) { 2036 spa_close(spa, FTAG); 2037 nvlist_free(config); 2038 return (error); 2039 } 2040 2041 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp); 2042 2043 spa_close(spa, FTAG); 2044 2045 nvlist_free(config); 2046 nvlist_free(props); 2047 2048 return (error); 2049} 2050 2051static int 2052zfs_ioc_vdev_setpath(zfs_cmd_t *zc) 2053{ 2054 spa_t *spa; 2055 char *path = zc->zc_value; 2056 uint64_t guid = zc->zc_guid; 2057 int error; 2058 2059 error = spa_open(zc->zc_name, &spa, FTAG); 2060 if (error != 0) 2061 return (error); 2062 2063 error = spa_vdev_setpath(spa, guid, path); 2064 spa_close(spa, FTAG); 2065 return (error); 2066} 2067 2068static int 2069zfs_ioc_vdev_setfru(zfs_cmd_t *zc) 2070{ 2071 spa_t *spa; 2072 char *fru = zc->zc_value; 2073 uint64_t guid = zc->zc_guid; 2074 int error; 2075 2076 error = spa_open(zc->zc_name, &spa, FTAG); 2077 if (error != 0) 2078 return (error); 2079 2080 error = spa_vdev_setfru(spa, guid, fru); 2081 spa_close(spa, FTAG); 2082 return (error); 2083} 2084 2085static int 2086zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os) 2087{ 2088 int error = 0; 2089 nvlist_t *nv; 2090 2091 dmu_objset_fast_stat(os, &zc->zc_objset_stats); 2092 2093 if (zc->zc_nvlist_dst != 0 && 2094 (error = dsl_prop_get_all(os, &nv)) == 0) { 2095 dmu_objset_stats(os, nv); 2096 /* 2097 * NB: zvol_get_stats() will read the objset contents, 2098 * which we aren't supposed to do with a 2099 * DS_MODE_USER hold, because it could be 2100 * inconsistent. So this is a bit of a workaround... 2101 * XXX reading with out owning 2102 */ 2103 if (!zc->zc_objset_stats.dds_inconsistent && 2104 dmu_objset_type(os) == DMU_OST_ZVOL) { 2105 error = zvol_get_stats(os, nv); 2106 if (error == EIO) 2107 return (error); 2108 VERIFY0(error); 2109 } 2110 error = put_nvlist(zc, nv); 2111 nvlist_free(nv); 2112 } 2113 2114 return (error); 2115} 2116 2117/* 2118 * inputs: 2119 * zc_name name of filesystem 2120 * zc_nvlist_dst_size size of buffer for property nvlist 2121 * 2122 * outputs: 2123 * zc_objset_stats stats 2124 * zc_nvlist_dst property nvlist 2125 * zc_nvlist_dst_size size of property nvlist 2126 */ 2127static int 2128zfs_ioc_objset_stats(zfs_cmd_t *zc) 2129{ 2130 objset_t *os; 2131 int error; 2132 2133 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 2134 if (error == 0) { 2135 error = zfs_ioc_objset_stats_impl(zc, os); 2136 dmu_objset_rele(os, FTAG); 2137 } 2138 2139 if (error == ENOMEM) 2140 error = 0; 2141 return (error); 2142} 2143 2144/* 2145 * inputs: 2146 * zc_name name of filesystem 2147 * zc_nvlist_dst_size size of buffer for property nvlist 2148 * 2149 * outputs: 2150 * zc_nvlist_dst received property nvlist 2151 * zc_nvlist_dst_size size of received property nvlist 2152 * 2153 * Gets received properties (distinct from local properties on or after 2154 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from 2155 * local property values. 2156 */ 2157static int 2158zfs_ioc_objset_recvd_props(zfs_cmd_t *zc) 2159{ 2160 int error = 0; 2161 nvlist_t *nv; 2162 2163 /* 2164 * Without this check, we would return local property values if the 2165 * caller has not already received properties on or after 2166 * SPA_VERSION_RECVD_PROPS. 2167 */ 2168 if (!dsl_prop_get_hasrecvd(zc->zc_name)) 2169 return (SET_ERROR(ENOTSUP)); 2170 2171 if (zc->zc_nvlist_dst != 0 && 2172 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) { 2173 error = put_nvlist(zc, nv); 2174 nvlist_free(nv); 2175 } 2176 2177 return (error); 2178} 2179 2180static int 2181nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop) 2182{ 2183 uint64_t value; 2184 int error; 2185 2186 /* 2187 * zfs_get_zplprop() will either find a value or give us 2188 * the default value (if there is one). 2189 */ 2190 if ((error = zfs_get_zplprop(os, prop, &value)) != 0) 2191 return (error); 2192 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0); 2193 return (0); 2194} 2195 2196/* 2197 * inputs: 2198 * zc_name name of filesystem 2199 * zc_nvlist_dst_size size of buffer for zpl property nvlist 2200 * 2201 * outputs: 2202 * zc_nvlist_dst zpl property nvlist 2203 * zc_nvlist_dst_size size of zpl property nvlist 2204 */ 2205static int 2206zfs_ioc_objset_zplprops(zfs_cmd_t *zc) 2207{ 2208 objset_t *os; 2209 int err; 2210 2211 /* XXX reading without owning */ 2212 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os)) 2213 return (err); 2214 2215 dmu_objset_fast_stat(os, &zc->zc_objset_stats); 2216 2217 /* 2218 * NB: nvl_add_zplprop() will read the objset contents, 2219 * which we aren't supposed to do with a DS_MODE_USER 2220 * hold, because it could be inconsistent. 2221 */ 2222 if (zc->zc_nvlist_dst != 0 && 2223 !zc->zc_objset_stats.dds_inconsistent && 2224 dmu_objset_type(os) == DMU_OST_ZFS) { 2225 nvlist_t *nv; 2226 2227 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2228 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 && 2229 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 && 2230 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 && 2231 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0) 2232 err = put_nvlist(zc, nv); 2233 nvlist_free(nv); 2234 } else { 2235 err = SET_ERROR(ENOENT); 2236 } 2237 dmu_objset_rele(os, FTAG); 2238 return (err); 2239} 2240 2241boolean_t 2242dataset_name_hidden(const char *name) 2243{ 2244 /* 2245 * Skip over datasets that are not visible in this zone, 2246 * internal datasets (which have a $ in their name), and 2247 * temporary datasets (which have a % in their name). 2248 */ 2249 if (strchr(name, '$') != NULL) 2250 return (B_TRUE); 2251 if (strchr(name, '%') != NULL) 2252 return (B_TRUE); 2253 if (!INGLOBALZONE(curthread) && !zone_dataset_visible(name, NULL)) 2254 return (B_TRUE); 2255 return (B_FALSE); 2256} 2257 2258/* 2259 * inputs: 2260 * zc_name name of filesystem 2261 * zc_cookie zap cursor 2262 * zc_nvlist_dst_size size of buffer for property nvlist 2263 * 2264 * outputs: 2265 * zc_name name of next filesystem 2266 * zc_cookie zap cursor 2267 * zc_objset_stats stats 2268 * zc_nvlist_dst property nvlist 2269 * zc_nvlist_dst_size size of property nvlist 2270 */ 2271static int 2272zfs_ioc_dataset_list_next(zfs_cmd_t *zc) 2273{ 2274 objset_t *os; 2275 int error; 2276 char *p; 2277 size_t orig_len = strlen(zc->zc_name); 2278 2279top: 2280 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) { 2281 if (error == ENOENT) 2282 error = SET_ERROR(ESRCH); 2283 return (error); 2284 } 2285 2286 p = strrchr(zc->zc_name, '/'); 2287 if (p == NULL || p[1] != '\0') 2288 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name)); 2289 p = zc->zc_name + strlen(zc->zc_name); 2290 2291 do { 2292 error = dmu_dir_list_next(os, 2293 sizeof (zc->zc_name) - (p - zc->zc_name), p, 2294 NULL, &zc->zc_cookie); 2295 if (error == ENOENT) 2296 error = SET_ERROR(ESRCH); 2297 } while (error == 0 && dataset_name_hidden(zc->zc_name)); 2298 dmu_objset_rele(os, FTAG); 2299 2300 /* 2301 * If it's an internal dataset (ie. with a '$' in its name), 2302 * don't try to get stats for it, otherwise we'll return ENOENT. 2303 */ 2304 if (error == 0 && strchr(zc->zc_name, '$') == NULL) { 2305 error = zfs_ioc_objset_stats(zc); /* fill in the stats */ 2306 if (error == ENOENT) { 2307 /* We lost a race with destroy, get the next one. */ 2308 zc->zc_name[orig_len] = '\0'; 2309 goto top; 2310 } 2311 } 2312 return (error); 2313} 2314 2315/* 2316 * inputs: 2317 * zc_name name of filesystem 2318 * zc_cookie zap cursor 2319 * zc_nvlist_dst_size size of buffer for property nvlist 2320 * zc_simple when set, only name is requested 2321 * 2322 * outputs: 2323 * zc_name name of next snapshot 2324 * zc_objset_stats stats 2325 * zc_nvlist_dst property nvlist 2326 * zc_nvlist_dst_size size of property nvlist 2327 */ 2328static int 2329zfs_ioc_snapshot_list_next(zfs_cmd_t *zc) 2330{ 2331 objset_t *os; 2332 int error; 2333 2334 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 2335 if (error != 0) { 2336 return (error == ENOENT ? ESRCH : error); 2337 } 2338 2339 /* 2340 * A dataset name of maximum length cannot have any snapshots, 2341 * so exit immediately. 2342 */ 2343 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) { 2344 dmu_objset_rele(os, FTAG); 2345 return (SET_ERROR(ESRCH)); 2346 } 2347 2348 error = dmu_snapshot_list_next(os, 2349 sizeof (zc->zc_name) - strlen(zc->zc_name), 2350 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie, 2351 NULL); 2352 2353 if (error == 0 && !zc->zc_simple) { 2354 dsl_dataset_t *ds; 2355 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool; 2356 2357 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds); 2358 if (error == 0) { 2359 objset_t *ossnap; 2360 2361 error = dmu_objset_from_ds(ds, &ossnap); 2362 if (error == 0) 2363 error = zfs_ioc_objset_stats_impl(zc, ossnap); 2364 dsl_dataset_rele(ds, FTAG); 2365 } 2366 } else if (error == ENOENT) { 2367 error = SET_ERROR(ESRCH); 2368 } 2369 2370 dmu_objset_rele(os, FTAG); 2371 /* if we failed, undo the @ that we tacked on to zc_name */ 2372 if (error != 0) 2373 *strchr(zc->zc_name, '@') = '\0'; 2374 return (error); 2375} 2376 2377static int 2378zfs_prop_set_userquota(const char *dsname, nvpair_t *pair) 2379{ 2380 const char *propname = nvpair_name(pair); 2381 uint64_t *valary; 2382 unsigned int vallen; 2383 const char *domain; 2384 char *dash; 2385 zfs_userquota_prop_t type; 2386 uint64_t rid; 2387 uint64_t quota; 2388 zfsvfs_t *zfsvfs; 2389 int err; 2390 2391 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2392 nvlist_t *attrs; 2393 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 2394 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2395 &pair) != 0) 2396 return (SET_ERROR(EINVAL)); 2397 } 2398 2399 /* 2400 * A correctly constructed propname is encoded as 2401 * userquota@<rid>-<domain>. 2402 */ 2403 if ((dash = strchr(propname, '-')) == NULL || 2404 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 || 2405 vallen != 3) 2406 return (SET_ERROR(EINVAL)); 2407 2408 domain = dash + 1; 2409 type = valary[0]; 2410 rid = valary[1]; 2411 quota = valary[2]; 2412 2413 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE); 2414 if (err == 0) { 2415 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota); 2416 zfsvfs_rele(zfsvfs, FTAG); 2417 } 2418 2419 return (err); 2420} 2421 2422/* 2423 * If the named property is one that has a special function to set its value, 2424 * return 0 on success and a positive error code on failure; otherwise if it is 2425 * not one of the special properties handled by this function, return -1. 2426 * 2427 * XXX: It would be better for callers of the property interface if we handled 2428 * these special cases in dsl_prop.c (in the dsl layer). 2429 */ 2430static int 2431zfs_prop_set_special(const char *dsname, zprop_source_t source, 2432 nvpair_t *pair) 2433{ 2434 const char *propname = nvpair_name(pair); 2435 zfs_prop_t prop = zfs_name_to_prop(propname); 2436 uint64_t intval; 2437 int err; 2438 2439 if (prop == ZPROP_INVAL) { 2440 if (zfs_prop_userquota(propname)) 2441 return (zfs_prop_set_userquota(dsname, pair)); 2442 return (-1); 2443 } 2444 2445 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2446 nvlist_t *attrs; 2447 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 2448 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2449 &pair) == 0); 2450 } 2451 2452 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING) 2453 return (-1); 2454 2455 VERIFY(0 == nvpair_value_uint64(pair, &intval)); 2456 2457 switch (prop) { 2458 case ZFS_PROP_QUOTA: 2459 err = dsl_dir_set_quota(dsname, source, intval); 2460 break; 2461 case ZFS_PROP_REFQUOTA: 2462 err = dsl_dataset_set_refquota(dsname, source, intval); 2463 break; 2464 case ZFS_PROP_FILESYSTEM_LIMIT: 2465 case ZFS_PROP_SNAPSHOT_LIMIT: 2466 if (intval == UINT64_MAX) { 2467 /* clearing the limit, just do it */ 2468 err = 0; 2469 } else { 2470 err = dsl_dir_activate_fs_ss_limit(dsname); 2471 } 2472 /* 2473 * Set err to -1 to force the zfs_set_prop_nvlist code down the 2474 * default path to set the value in the nvlist. 2475 */ 2476 if (err == 0) 2477 err = -1; 2478 break; 2479 case ZFS_PROP_RESERVATION: 2480 err = dsl_dir_set_reservation(dsname, source, intval); 2481 break; 2482 case ZFS_PROP_REFRESERVATION: 2483 err = dsl_dataset_set_refreservation(dsname, source, intval); 2484 break; 2485 case ZFS_PROP_VOLSIZE: 2486 err = zvol_set_volsize(dsname, ddi_driver_major(zfs_dip), 2487 intval); 2488 break; 2489 case ZFS_PROP_VERSION: 2490 { 2491 zfsvfs_t *zfsvfs; 2492 2493 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0) 2494 break; 2495 2496 err = zfs_set_version(zfsvfs, intval); 2497 zfsvfs_rele(zfsvfs, FTAG); 2498 2499 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) { 2500 zfs_cmd_t *zc; 2501 2502 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP); 2503 (void) strcpy(zc->zc_name, dsname); 2504 (void) zfs_ioc_userspace_upgrade(zc); 2505 kmem_free(zc, sizeof (zfs_cmd_t)); 2506 } 2507 break; 2508 } 2509 default: 2510 err = -1; 2511 } 2512 2513 return (err); 2514} 2515 2516/* 2517 * This function is best effort. If it fails to set any of the given properties, 2518 * it continues to set as many as it can and returns the last error 2519 * encountered. If the caller provides a non-NULL errlist, it will be filled in 2520 * with the list of names of all the properties that failed along with the 2521 * corresponding error numbers. 2522 * 2523 * If every property is set successfully, zero is returned and errlist is not 2524 * modified. 2525 */ 2526int 2527zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl, 2528 nvlist_t *errlist) 2529{ 2530 nvpair_t *pair; 2531 nvpair_t *propval; 2532 int rv = 0; 2533 uint64_t intval; 2534 char *strval; 2535 nvlist_t *genericnvl = fnvlist_alloc(); 2536 nvlist_t *retrynvl = fnvlist_alloc(); 2537 2538retry: 2539 pair = NULL; 2540 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { 2541 const char *propname = nvpair_name(pair); 2542 zfs_prop_t prop = zfs_name_to_prop(propname); 2543 int err = 0; 2544 2545 /* decode the property value */ 2546 propval = pair; 2547 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2548 nvlist_t *attrs; 2549 attrs = fnvpair_value_nvlist(pair); 2550 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2551 &propval) != 0) 2552 err = SET_ERROR(EINVAL); 2553 } 2554 2555 /* Validate value type */ 2556 if (err == 0 && prop == ZPROP_INVAL) { 2557 if (zfs_prop_user(propname)) { 2558 if (nvpair_type(propval) != DATA_TYPE_STRING) 2559 err = SET_ERROR(EINVAL); 2560 } else if (zfs_prop_userquota(propname)) { 2561 if (nvpair_type(propval) != 2562 DATA_TYPE_UINT64_ARRAY) 2563 err = SET_ERROR(EINVAL); 2564 } else { 2565 err = SET_ERROR(EINVAL); 2566 } 2567 } else if (err == 0) { 2568 if (nvpair_type(propval) == DATA_TYPE_STRING) { 2569 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING) 2570 err = SET_ERROR(EINVAL); 2571 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) { 2572 const char *unused; 2573 2574 intval = fnvpair_value_uint64(propval); 2575 2576 switch (zfs_prop_get_type(prop)) { 2577 case PROP_TYPE_NUMBER: 2578 break; 2579 case PROP_TYPE_STRING: 2580 err = SET_ERROR(EINVAL); 2581 break; 2582 case PROP_TYPE_INDEX: 2583 if (zfs_prop_index_to_string(prop, 2584 intval, &unused) != 0) 2585 err = SET_ERROR(EINVAL); 2586 break; 2587 default: 2588 cmn_err(CE_PANIC, 2589 "unknown property type"); 2590 } 2591 } else { 2592 err = SET_ERROR(EINVAL); 2593 } 2594 } 2595 2596 /* Validate permissions */ 2597 if (err == 0) 2598 err = zfs_check_settable(dsname, pair, CRED()); 2599 2600 if (err == 0) { 2601 err = zfs_prop_set_special(dsname, source, pair); 2602 if (err == -1) { 2603 /* 2604 * For better performance we build up a list of 2605 * properties to set in a single transaction. 2606 */ 2607 err = nvlist_add_nvpair(genericnvl, pair); 2608 } else if (err != 0 && nvl != retrynvl) { 2609 /* 2610 * This may be a spurious error caused by 2611 * receiving quota and reservation out of order. 2612 * Try again in a second pass. 2613 */ 2614 err = nvlist_add_nvpair(retrynvl, pair); 2615 } 2616 } 2617 2618 if (err != 0) { 2619 if (errlist != NULL) 2620 fnvlist_add_int32(errlist, propname, err); 2621 rv = err; 2622 } 2623 } 2624 2625 if (nvl != retrynvl && !nvlist_empty(retrynvl)) { 2626 nvl = retrynvl; 2627 goto retry; 2628 } 2629 2630 if (!nvlist_empty(genericnvl) && 2631 dsl_props_set(dsname, source, genericnvl) != 0) { 2632 /* 2633 * If this fails, we still want to set as many properties as we 2634 * can, so try setting them individually. 2635 */ 2636 pair = NULL; 2637 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) { 2638 const char *propname = nvpair_name(pair); 2639 int err = 0; 2640 2641 propval = pair; 2642 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2643 nvlist_t *attrs; 2644 attrs = fnvpair_value_nvlist(pair); 2645 propval = fnvlist_lookup_nvpair(attrs, 2646 ZPROP_VALUE); 2647 } 2648 2649 if (nvpair_type(propval) == DATA_TYPE_STRING) { 2650 strval = fnvpair_value_string(propval); 2651 err = dsl_prop_set_string(dsname, propname, 2652 source, strval); 2653 } else { 2654 intval = fnvpair_value_uint64(propval); 2655 err = dsl_prop_set_int(dsname, propname, source, 2656 intval); 2657 } 2658 2659 if (err != 0) { 2660 if (errlist != NULL) { 2661 fnvlist_add_int32(errlist, propname, 2662 err); 2663 } 2664 rv = err; 2665 } 2666 } 2667 } 2668 nvlist_free(genericnvl); 2669 nvlist_free(retrynvl); 2670 2671 return (rv); 2672} 2673 2674/* 2675 * Check that all the properties are valid user properties. 2676 */ 2677static int 2678zfs_check_userprops(const char *fsname, nvlist_t *nvl) 2679{ 2680 nvpair_t *pair = NULL; 2681 int error = 0; 2682 2683 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { 2684 const char *propname = nvpair_name(pair); 2685 2686 if (!zfs_prop_user(propname) || 2687 nvpair_type(pair) != DATA_TYPE_STRING) 2688 return (SET_ERROR(EINVAL)); 2689 2690 if (error = zfs_secpolicy_write_perms(fsname, 2691 ZFS_DELEG_PERM_USERPROP, CRED())) 2692 return (error); 2693 2694 if (strlen(propname) >= ZAP_MAXNAMELEN) 2695 return (SET_ERROR(ENAMETOOLONG)); 2696 2697 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN) 2698 return (E2BIG); 2699 } 2700 return (0); 2701} 2702 2703static void 2704props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops) 2705{ 2706 nvpair_t *pair; 2707 2708 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2709 2710 pair = NULL; 2711 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) { 2712 if (nvlist_exists(skipped, nvpair_name(pair))) 2713 continue; 2714 2715 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0); 2716 } 2717} 2718 2719static int 2720clear_received_props(const char *dsname, nvlist_t *props, 2721 nvlist_t *skipped) 2722{ 2723 int err = 0; 2724 nvlist_t *cleared_props = NULL; 2725 props_skip(props, skipped, &cleared_props); 2726 if (!nvlist_empty(cleared_props)) { 2727 /* 2728 * Acts on local properties until the dataset has received 2729 * properties at least once on or after SPA_VERSION_RECVD_PROPS. 2730 */ 2731 zprop_source_t flags = (ZPROP_SRC_NONE | 2732 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0)); 2733 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL); 2734 } 2735 nvlist_free(cleared_props); 2736 return (err); 2737} 2738 2739/* 2740 * inputs: 2741 * zc_name name of filesystem 2742 * zc_value name of property to set 2743 * zc_nvlist_src{_size} nvlist of properties to apply 2744 * zc_cookie received properties flag 2745 * 2746 * outputs: 2747 * zc_nvlist_dst{_size} error for each unapplied received property 2748 */ 2749static int 2750zfs_ioc_set_prop(zfs_cmd_t *zc) 2751{ 2752 nvlist_t *nvl; 2753 boolean_t received = zc->zc_cookie; 2754 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED : 2755 ZPROP_SRC_LOCAL); 2756 nvlist_t *errors; 2757 int error; 2758 2759 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2760 zc->zc_iflags, &nvl)) != 0) 2761 return (error); 2762 2763 if (received) { 2764 nvlist_t *origprops; 2765 2766 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) { 2767 (void) clear_received_props(zc->zc_name, 2768 origprops, nvl); 2769 nvlist_free(origprops); 2770 } 2771 2772 error = dsl_prop_set_hasrecvd(zc->zc_name); 2773 } 2774 2775 errors = fnvlist_alloc(); 2776 if (error == 0) 2777 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors); 2778 2779 if (zc->zc_nvlist_dst != 0 && errors != NULL) { 2780 (void) put_nvlist(zc, errors); 2781 } 2782 2783 nvlist_free(errors); 2784 nvlist_free(nvl); 2785 return (error); 2786} 2787 2788/* 2789 * inputs: 2790 * zc_name name of filesystem 2791 * zc_value name of property to inherit 2792 * zc_cookie revert to received value if TRUE 2793 * 2794 * outputs: none 2795 */ 2796static int 2797zfs_ioc_inherit_prop(zfs_cmd_t *zc) 2798{ 2799 const char *propname = zc->zc_value; 2800 zfs_prop_t prop = zfs_name_to_prop(propname); 2801 boolean_t received = zc->zc_cookie; 2802 zprop_source_t source = (received 2803 ? ZPROP_SRC_NONE /* revert to received value, if any */ 2804 : ZPROP_SRC_INHERITED); /* explicitly inherit */ 2805 2806 if (received) { 2807 nvlist_t *dummy; 2808 nvpair_t *pair; 2809 zprop_type_t type; 2810 int err; 2811 2812 /* 2813 * zfs_prop_set_special() expects properties in the form of an 2814 * nvpair with type info. 2815 */ 2816 if (prop == ZPROP_INVAL) { 2817 if (!zfs_prop_user(propname)) 2818 return (SET_ERROR(EINVAL)); 2819 2820 type = PROP_TYPE_STRING; 2821 } else if (prop == ZFS_PROP_VOLSIZE || 2822 prop == ZFS_PROP_VERSION) { 2823 return (SET_ERROR(EINVAL)); 2824 } else { 2825 type = zfs_prop_get_type(prop); 2826 } 2827 2828 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2829 2830 switch (type) { 2831 case PROP_TYPE_STRING: 2832 VERIFY(0 == nvlist_add_string(dummy, propname, "")); 2833 break; 2834 case PROP_TYPE_NUMBER: 2835 case PROP_TYPE_INDEX: 2836 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0)); 2837 break; 2838 default: 2839 nvlist_free(dummy); 2840 return (SET_ERROR(EINVAL)); 2841 } 2842 2843 pair = nvlist_next_nvpair(dummy, NULL); 2844 err = zfs_prop_set_special(zc->zc_name, source, pair); 2845 nvlist_free(dummy); 2846 if (err != -1) 2847 return (err); /* special property already handled */ 2848 } else { 2849 /* 2850 * Only check this in the non-received case. We want to allow 2851 * 'inherit -S' to revert non-inheritable properties like quota 2852 * and reservation to the received or default values even though 2853 * they are not considered inheritable. 2854 */ 2855 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop)) 2856 return (SET_ERROR(EINVAL)); 2857 } 2858 2859 /* property name has been validated by zfs_secpolicy_inherit_prop() */ 2860 return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source)); 2861} 2862 2863static int 2864zfs_ioc_pool_set_props(zfs_cmd_t *zc) 2865{ 2866 nvlist_t *props; 2867 spa_t *spa; 2868 int error; 2869 nvpair_t *pair; 2870 2871 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2872 zc->zc_iflags, &props)) 2873 return (error); 2874 2875 /* 2876 * If the only property is the configfile, then just do a spa_lookup() 2877 * to handle the faulted case. 2878 */ 2879 pair = nvlist_next_nvpair(props, NULL); 2880 if (pair != NULL && strcmp(nvpair_name(pair), 2881 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 && 2882 nvlist_next_nvpair(props, pair) == NULL) { 2883 mutex_enter(&spa_namespace_lock); 2884 if ((spa = spa_lookup(zc->zc_name)) != NULL) { 2885 spa_configfile_set(spa, props, B_FALSE); 2886 spa_config_sync(spa, B_FALSE, B_TRUE); 2887 } 2888 mutex_exit(&spa_namespace_lock); 2889 if (spa != NULL) { 2890 nvlist_free(props); 2891 return (0); 2892 } 2893 } 2894 2895 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { 2896 nvlist_free(props); 2897 return (error); 2898 } 2899 2900 error = spa_prop_set(spa, props); 2901 2902 nvlist_free(props); 2903 spa_close(spa, FTAG); 2904 2905 return (error); 2906} 2907 2908static int 2909zfs_ioc_pool_get_props(zfs_cmd_t *zc) 2910{ 2911 spa_t *spa; 2912 int error; 2913 nvlist_t *nvp = NULL; 2914 2915 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { 2916 /* 2917 * If the pool is faulted, there may be properties we can still 2918 * get (such as altroot and cachefile), so attempt to get them 2919 * anyway. 2920 */ 2921 mutex_enter(&spa_namespace_lock); 2922 if ((spa = spa_lookup(zc->zc_name)) != NULL) 2923 error = spa_prop_get(spa, &nvp); 2924 mutex_exit(&spa_namespace_lock); 2925 } else { 2926 error = spa_prop_get(spa, &nvp); 2927 spa_close(spa, FTAG); 2928 } 2929 2930 if (error == 0 && zc->zc_nvlist_dst != 0) 2931 error = put_nvlist(zc, nvp); 2932 else 2933 error = SET_ERROR(EFAULT); 2934 2935 nvlist_free(nvp); 2936 return (error); 2937} 2938 2939/* 2940 * inputs: 2941 * zc_name name of filesystem 2942 * zc_nvlist_src{_size} nvlist of delegated permissions 2943 * zc_perm_action allow/unallow flag 2944 * 2945 * outputs: none 2946 */ 2947static int 2948zfs_ioc_set_fsacl(zfs_cmd_t *zc) 2949{ 2950 int error; 2951 nvlist_t *fsaclnv = NULL; 2952 2953 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2954 zc->zc_iflags, &fsaclnv)) != 0) 2955 return (error); 2956 2957 /* 2958 * Verify nvlist is constructed correctly 2959 */ 2960 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) { 2961 nvlist_free(fsaclnv); 2962 return (SET_ERROR(EINVAL)); 2963 } 2964 2965 /* 2966 * If we don't have PRIV_SYS_MOUNT, then validate 2967 * that user is allowed to hand out each permission in 2968 * the nvlist(s) 2969 */ 2970 2971 error = secpolicy_zfs(CRED()); 2972 if (error != 0) { 2973 if (zc->zc_perm_action == B_FALSE) { 2974 error = dsl_deleg_can_allow(zc->zc_name, 2975 fsaclnv, CRED()); 2976 } else { 2977 error = dsl_deleg_can_unallow(zc->zc_name, 2978 fsaclnv, CRED()); 2979 } 2980 } 2981 2982 if (error == 0) 2983 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action); 2984 2985 nvlist_free(fsaclnv); 2986 return (error); 2987} 2988 2989/* 2990 * inputs: 2991 * zc_name name of filesystem 2992 * 2993 * outputs: 2994 * zc_nvlist_src{_size} nvlist of delegated permissions 2995 */ 2996static int 2997zfs_ioc_get_fsacl(zfs_cmd_t *zc) 2998{ 2999 nvlist_t *nvp; 3000 int error; 3001 3002 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) { 3003 error = put_nvlist(zc, nvp); 3004 nvlist_free(nvp); 3005 } 3006 3007 return (error); 3008} 3009 3010/* 3011 * Search the vfs list for a specified resource. Returns a pointer to it 3012 * or NULL if no suitable entry is found. The caller of this routine 3013 * is responsible for releasing the returned vfs pointer. 3014 */ 3015static vfs_t * 3016zfs_get_vfs(const char *resource) 3017{ 3018 vfs_t *vfsp; 3019 3020 mtx_lock(&mountlist_mtx); 3021 TAILQ_FOREACH(vfsp, &mountlist, mnt_list) { 3022 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) { 3023 VFS_HOLD(vfsp); 3024 break; 3025 } 3026 } 3027 mtx_unlock(&mountlist_mtx); 3028 return (vfsp); 3029} 3030 3031/* ARGSUSED */ 3032static void 3033zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) 3034{ 3035 zfs_creat_t *zct = arg; 3036 3037 zfs_create_fs(os, cr, zct->zct_zplprops, tx); 3038} 3039 3040#define ZFS_PROP_UNDEFINED ((uint64_t)-1) 3041 3042/* 3043 * inputs: 3044 * os parent objset pointer (NULL if root fs) 3045 * fuids_ok fuids allowed in this version of the spa? 3046 * sa_ok SAs allowed in this version of the spa? 3047 * createprops list of properties requested by creator 3048 * 3049 * outputs: 3050 * zplprops values for the zplprops we attach to the master node object 3051 * is_ci true if requested file system will be purely case-insensitive 3052 * 3053 * Determine the settings for utf8only, normalization and 3054 * casesensitivity. Specific values may have been requested by the 3055 * creator and/or we can inherit values from the parent dataset. If 3056 * the file system is of too early a vintage, a creator can not 3057 * request settings for these properties, even if the requested 3058 * setting is the default value. We don't actually want to create dsl 3059 * properties for these, so remove them from the source nvlist after 3060 * processing. 3061 */ 3062static int 3063zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver, 3064 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops, 3065 nvlist_t *zplprops, boolean_t *is_ci) 3066{ 3067 uint64_t sense = ZFS_PROP_UNDEFINED; 3068 uint64_t norm = ZFS_PROP_UNDEFINED; 3069 uint64_t u8 = ZFS_PROP_UNDEFINED; 3070 3071 ASSERT(zplprops != NULL); 3072 3073 /* 3074 * Pull out creator prop choices, if any. 3075 */ 3076 if (createprops) { 3077 (void) nvlist_lookup_uint64(createprops, 3078 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver); 3079 (void) nvlist_lookup_uint64(createprops, 3080 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm); 3081 (void) nvlist_remove_all(createprops, 3082 zfs_prop_to_name(ZFS_PROP_NORMALIZE)); 3083 (void) nvlist_lookup_uint64(createprops, 3084 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8); 3085 (void) nvlist_remove_all(createprops, 3086 zfs_prop_to_name(ZFS_PROP_UTF8ONLY)); 3087 (void) nvlist_lookup_uint64(createprops, 3088 zfs_prop_to_name(ZFS_PROP_CASE), &sense); 3089 (void) nvlist_remove_all(createprops, 3090 zfs_prop_to_name(ZFS_PROP_CASE)); 3091 } 3092 3093 /* 3094 * If the zpl version requested is whacky or the file system 3095 * or pool is version is too "young" to support normalization 3096 * and the creator tried to set a value for one of the props, 3097 * error out. 3098 */ 3099 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) || 3100 (zplver >= ZPL_VERSION_FUID && !fuids_ok) || 3101 (zplver >= ZPL_VERSION_SA && !sa_ok) || 3102 (zplver < ZPL_VERSION_NORMALIZATION && 3103 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED || 3104 sense != ZFS_PROP_UNDEFINED))) 3105 return (SET_ERROR(ENOTSUP)); 3106 3107 /* 3108 * Put the version in the zplprops 3109 */ 3110 VERIFY(nvlist_add_uint64(zplprops, 3111 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0); 3112 3113 if (norm == ZFS_PROP_UNDEFINED) 3114 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0); 3115 VERIFY(nvlist_add_uint64(zplprops, 3116 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0); 3117 3118 /* 3119 * If we're normalizing, names must always be valid UTF-8 strings. 3120 */ 3121 if (norm) 3122 u8 = 1; 3123 if (u8 == ZFS_PROP_UNDEFINED) 3124 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0); 3125 VERIFY(nvlist_add_uint64(zplprops, 3126 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0); 3127 3128 if (sense == ZFS_PROP_UNDEFINED) 3129 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0); 3130 VERIFY(nvlist_add_uint64(zplprops, 3131 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0); 3132 3133 if (is_ci) 3134 *is_ci = (sense == ZFS_CASE_INSENSITIVE); 3135 3136 return (0); 3137} 3138 3139static int 3140zfs_fill_zplprops(const char *dataset, nvlist_t *createprops, 3141 nvlist_t *zplprops, boolean_t *is_ci) 3142{ 3143 boolean_t fuids_ok, sa_ok; 3144 uint64_t zplver = ZPL_VERSION; 3145 objset_t *os = NULL; 3146 char parentname[MAXNAMELEN]; 3147 char *cp; 3148 spa_t *spa; 3149 uint64_t spa_vers; 3150 int error; 3151 3152 (void) strlcpy(parentname, dataset, sizeof (parentname)); 3153 cp = strrchr(parentname, '/'); 3154 ASSERT(cp != NULL); 3155 cp[0] = '\0'; 3156 3157 if ((error = spa_open(dataset, &spa, FTAG)) != 0) 3158 return (error); 3159 3160 spa_vers = spa_version(spa); 3161 spa_close(spa, FTAG); 3162 3163 zplver = zfs_zpl_version_map(spa_vers); 3164 fuids_ok = (zplver >= ZPL_VERSION_FUID); 3165 sa_ok = (zplver >= ZPL_VERSION_SA); 3166 3167 /* 3168 * Open parent object set so we can inherit zplprop values. 3169 */ 3170 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0) 3171 return (error); 3172 3173 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops, 3174 zplprops, is_ci); 3175 dmu_objset_rele(os, FTAG); 3176 return (error); 3177} 3178 3179static int 3180zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops, 3181 nvlist_t *zplprops, boolean_t *is_ci) 3182{ 3183 boolean_t fuids_ok; 3184 boolean_t sa_ok; 3185 uint64_t zplver = ZPL_VERSION; 3186 int error; 3187 3188 zplver = zfs_zpl_version_map(spa_vers); 3189 fuids_ok = (zplver >= ZPL_VERSION_FUID); 3190 sa_ok = (zplver >= ZPL_VERSION_SA); 3191 3192 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok, 3193 createprops, zplprops, is_ci); 3194 return (error); 3195} 3196 3197/* 3198 * innvl: { 3199 * "type" -> dmu_objset_type_t (int32) 3200 * (optional) "props" -> { prop -> value } 3201 * } 3202 * 3203 * outnvl: propname -> error code (int32) 3204 */ 3205static int 3206zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3207{ 3208 int error = 0; 3209 zfs_creat_t zct = { 0 }; 3210 nvlist_t *nvprops = NULL; 3211 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx); 3212 int32_t type32; 3213 dmu_objset_type_t type; 3214 boolean_t is_insensitive = B_FALSE; 3215 3216 if (nvlist_lookup_int32(innvl, "type", &type32) != 0) 3217 return (SET_ERROR(EINVAL)); 3218 type = type32; 3219 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops); 3220 3221 switch (type) { 3222 case DMU_OST_ZFS: 3223 cbfunc = zfs_create_cb; 3224 break; 3225 3226 case DMU_OST_ZVOL: 3227 cbfunc = zvol_create_cb; 3228 break; 3229 3230 default: 3231 cbfunc = NULL; 3232 break; 3233 } 3234 if (strchr(fsname, '@') || 3235 strchr(fsname, '%')) 3236 return (SET_ERROR(EINVAL)); 3237 3238 zct.zct_props = nvprops; 3239 3240 if (cbfunc == NULL) 3241 return (SET_ERROR(EINVAL)); 3242 3243 if (type == DMU_OST_ZVOL) { 3244 uint64_t volsize, volblocksize; 3245 3246 if (nvprops == NULL) 3247 return (SET_ERROR(EINVAL)); 3248 if (nvlist_lookup_uint64(nvprops, 3249 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0) 3250 return (SET_ERROR(EINVAL)); 3251 3252 if ((error = nvlist_lookup_uint64(nvprops, 3253 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), 3254 &volblocksize)) != 0 && error != ENOENT) 3255 return (SET_ERROR(EINVAL)); 3256 3257 if (error != 0) 3258 volblocksize = zfs_prop_default_numeric( 3259 ZFS_PROP_VOLBLOCKSIZE); 3260 3261 if ((error = zvol_check_volblocksize( 3262 volblocksize)) != 0 || 3263 (error = zvol_check_volsize(volsize, 3264 volblocksize)) != 0) 3265 return (error); 3266 } else if (type == DMU_OST_ZFS) { 3267 int error; 3268 3269 /* 3270 * We have to have normalization and 3271 * case-folding flags correct when we do the 3272 * file system creation, so go figure them out 3273 * now. 3274 */ 3275 VERIFY(nvlist_alloc(&zct.zct_zplprops, 3276 NV_UNIQUE_NAME, KM_SLEEP) == 0); 3277 error = zfs_fill_zplprops(fsname, nvprops, 3278 zct.zct_zplprops, &is_insensitive); 3279 if (error != 0) { 3280 nvlist_free(zct.zct_zplprops); 3281 return (error); 3282 } 3283 } 3284 3285 error = dmu_objset_create(fsname, type, 3286 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct); 3287 nvlist_free(zct.zct_zplprops); 3288 3289 /* 3290 * It would be nice to do this atomically. 3291 */ 3292 if (error == 0) { 3293 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL, 3294 nvprops, outnvl); 3295 if (error != 0) 3296 (void) dsl_destroy_head(fsname); 3297 } 3298#ifdef __FreeBSD__ 3299 if (error == 0 && type == DMU_OST_ZVOL) 3300 zvol_create_minors(fsname); 3301#endif 3302 return (error); 3303} 3304 3305/* 3306 * innvl: { 3307 * "origin" -> name of origin snapshot 3308 * (optional) "props" -> { prop -> value } 3309 * } 3310 * 3311 * outnvl: propname -> error code (int32) 3312 */ 3313static int 3314zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3315{ 3316 int error = 0; 3317 nvlist_t *nvprops = NULL; 3318 char *origin_name; 3319 3320 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0) 3321 return (SET_ERROR(EINVAL)); 3322 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops); 3323 3324 if (strchr(fsname, '@') || 3325 strchr(fsname, '%')) 3326 return (SET_ERROR(EINVAL)); 3327 3328 if (dataset_namecheck(origin_name, NULL, NULL) != 0) 3329 return (SET_ERROR(EINVAL)); 3330 error = dmu_objset_clone(fsname, origin_name); 3331 if (error != 0) 3332 return (error); 3333 3334 /* 3335 * It would be nice to do this atomically. 3336 */ 3337 if (error == 0) { 3338 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL, 3339 nvprops, outnvl); 3340 if (error != 0) 3341 (void) dsl_destroy_head(fsname); 3342 } 3343#ifdef __FreeBSD__ 3344 if (error == 0) 3345 zvol_create_minors(fsname); 3346#endif 3347 return (error); 3348} 3349 3350/* 3351 * innvl: { 3352 * "snaps" -> { snapshot1, snapshot2 } 3353 * (optional) "props" -> { prop -> value (string) } 3354 * } 3355 * 3356 * outnvl: snapshot -> error code (int32) 3357 */ 3358static int 3359zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3360{ 3361 nvlist_t *snaps; 3362 nvlist_t *props = NULL; 3363 int error, poollen; 3364 nvpair_t *pair; 3365 3366 (void) nvlist_lookup_nvlist(innvl, "props", &props); 3367 if ((error = zfs_check_userprops(poolname, props)) != 0) 3368 return (error); 3369 3370 if (!nvlist_empty(props) && 3371 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS)) 3372 return (SET_ERROR(ENOTSUP)); 3373 3374 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 3375 return (SET_ERROR(EINVAL)); 3376 poollen = strlen(poolname); 3377 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 3378 pair = nvlist_next_nvpair(snaps, pair)) { 3379 const char *name = nvpair_name(pair); 3380 const char *cp = strchr(name, '@'); 3381 3382 /* 3383 * The snap name must contain an @, and the part after it must 3384 * contain only valid characters. 3385 */ 3386 if (cp == NULL || 3387 zfs_component_namecheck(cp + 1, NULL, NULL) != 0) 3388 return (SET_ERROR(EINVAL)); 3389 3390 /* 3391 * The snap must be in the specified pool. 3392 */ 3393 if (strncmp(name, poolname, poollen) != 0 || 3394 (name[poollen] != '/' && name[poollen] != '@')) 3395 return (SET_ERROR(EXDEV)); 3396 3397 /* This must be the only snap of this fs. */ 3398 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair); 3399 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) { 3400 if (strncmp(name, nvpair_name(pair2), cp - name + 1) 3401 == 0) { 3402 return (SET_ERROR(EXDEV)); 3403 } 3404 } 3405 } 3406 3407 error = dsl_dataset_snapshot(snaps, props, outnvl); 3408 return (error); 3409} 3410 3411/* 3412 * innvl: "message" -> string 3413 */ 3414/* ARGSUSED */ 3415static int 3416zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl) 3417{ 3418 char *message; 3419 spa_t *spa; 3420 int error; 3421 char *poolname; 3422 3423 /* 3424 * The poolname in the ioctl is not set, we get it from the TSD, 3425 * which was set at the end of the last successful ioctl that allows 3426 * logging. The secpolicy func already checked that it is set. 3427 * Only one log ioctl is allowed after each successful ioctl, so 3428 * we clear the TSD here. 3429 */ 3430 poolname = tsd_get(zfs_allow_log_key); 3431 (void) tsd_set(zfs_allow_log_key, NULL); 3432 error = spa_open(poolname, &spa, FTAG); 3433 strfree(poolname); 3434 if (error != 0) 3435 return (error); 3436 3437 if (nvlist_lookup_string(innvl, "message", &message) != 0) { 3438 spa_close(spa, FTAG); 3439 return (SET_ERROR(EINVAL)); 3440 } 3441 3442 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) { 3443 spa_close(spa, FTAG); 3444 return (SET_ERROR(ENOTSUP)); 3445 } 3446 3447 error = spa_history_log(spa, message); 3448 spa_close(spa, FTAG); 3449 return (error); 3450} 3451 3452/* 3453 * The dp_config_rwlock must not be held when calling this, because the 3454 * unmount may need to write out data. 3455 * 3456 * This function is best-effort. Callers must deal gracefully if it 3457 * remains mounted (or is remounted after this call). 3458 * 3459 * Returns 0 if the argument is not a snapshot, or it is not currently a 3460 * filesystem, or we were able to unmount it. Returns error code otherwise. 3461 */ 3462int 3463zfs_unmount_snap(const char *snapname) 3464{ 3465 vfs_t *vfsp; 3466 zfsvfs_t *zfsvfs; 3467 int err; 3468 3469 if (strchr(snapname, '@') == NULL) 3470 return (0); 3471 3472 vfsp = zfs_get_vfs(snapname); 3473 if (vfsp == NULL) 3474 return (0); 3475 3476 zfsvfs = vfsp->vfs_data; 3477 ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os))); 3478 3479 err = vn_vfswlock(vfsp->vfs_vnodecovered); 3480 VFS_RELE(vfsp); 3481 if (err != 0) 3482 return (SET_ERROR(err)); 3483 3484 /* 3485 * Always force the unmount for snapshots. 3486 */ 3487 3488#ifdef illumos 3489 (void) dounmount(vfsp, MS_FORCE, kcred); 3490#else 3491 mtx_lock(&Giant); /* dounmount() */ 3492 (void) dounmount(vfsp, MS_FORCE, curthread); 3493 mtx_unlock(&Giant); /* dounmount() */ 3494#endif 3495 return (0); 3496} 3497 3498/* ARGSUSED */ 3499static int 3500zfs_unmount_snap_cb(const char *snapname, void *arg) 3501{ 3502 return (zfs_unmount_snap(snapname)); 3503} 3504 3505/* 3506 * When a clone is destroyed, its origin may also need to be destroyed, 3507 * in which case it must be unmounted. This routine will do that unmount 3508 * if necessary. 3509 */ 3510void 3511zfs_destroy_unmount_origin(const char *fsname) 3512{ 3513 int error; 3514 objset_t *os; 3515 dsl_dataset_t *ds; 3516 3517 error = dmu_objset_hold(fsname, FTAG, &os); 3518 if (error != 0) 3519 return; 3520 ds = dmu_objset_ds(os); 3521 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) { 3522 char originname[MAXNAMELEN]; 3523 dsl_dataset_name(ds->ds_prev, originname); 3524 dmu_objset_rele(os, FTAG); 3525 (void) zfs_unmount_snap(originname); 3526 } else { 3527 dmu_objset_rele(os, FTAG); 3528 } 3529} 3530 3531/* 3532 * innvl: { 3533 * "snaps" -> { snapshot1, snapshot2 } 3534 * (optional boolean) "defer" 3535 * } 3536 * 3537 * outnvl: snapshot -> error code (int32) 3538 * 3539 */ 3540/* ARGSUSED */ 3541static int 3542zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3543{ 3544 nvlist_t *snaps; 3545 nvpair_t *pair; 3546 boolean_t defer; 3547 3548 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 3549 return (SET_ERROR(EINVAL)); 3550 defer = nvlist_exists(innvl, "defer"); 3551 3552 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 3553 pair = nvlist_next_nvpair(snaps, pair)) { 3554 (void) zfs_unmount_snap(nvpair_name(pair)); 3555 } 3556 3557 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl)); 3558} 3559 3560/* 3561 * Create bookmarks. Bookmark names are of the form <fs>#<bmark>. 3562 * All bookmarks must be in the same pool. 3563 * 3564 * innvl: { 3565 * bookmark1 -> snapshot1, bookmark2 -> snapshot2 3566 * } 3567 * 3568 * outnvl: bookmark -> error code (int32) 3569 * 3570 */ 3571/* ARGSUSED */ 3572static int 3573zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3574{ 3575 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 3576 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 3577 char *snap_name; 3578 3579 /* 3580 * Verify the snapshot argument. 3581 */ 3582 if (nvpair_value_string(pair, &snap_name) != 0) 3583 return (SET_ERROR(EINVAL)); 3584 3585 3586 /* Verify that the keys (bookmarks) are unique */ 3587 for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair); 3588 pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) { 3589 if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0) 3590 return (SET_ERROR(EINVAL)); 3591 } 3592 } 3593 3594 return (dsl_bookmark_create(innvl, outnvl)); 3595} 3596 3597/* 3598 * innvl: { 3599 * property 1, property 2, ... 3600 * } 3601 * 3602 * outnvl: { 3603 * bookmark name 1 -> { property 1, property 2, ... }, 3604 * bookmark name 2 -> { property 1, property 2, ... } 3605 * } 3606 * 3607 */ 3608static int 3609zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3610{ 3611 return (dsl_get_bookmarks(fsname, innvl, outnvl)); 3612} 3613 3614/* 3615 * innvl: { 3616 * bookmark name 1, bookmark name 2 3617 * } 3618 * 3619 * outnvl: bookmark -> error code (int32) 3620 * 3621 */ 3622static int 3623zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl, 3624 nvlist_t *outnvl) 3625{ 3626 int error, poollen; 3627 3628 poollen = strlen(poolname); 3629 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 3630 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 3631 const char *name = nvpair_name(pair); 3632 const char *cp = strchr(name, '#'); 3633 3634 /* 3635 * The bookmark name must contain an #, and the part after it 3636 * must contain only valid characters. 3637 */ 3638 if (cp == NULL || 3639 zfs_component_namecheck(cp + 1, NULL, NULL) != 0) 3640 return (SET_ERROR(EINVAL)); 3641 3642 /* 3643 * The bookmark must be in the specified pool. 3644 */ 3645 if (strncmp(name, poolname, poollen) != 0 || 3646 (name[poollen] != '/' && name[poollen] != '#')) 3647 return (SET_ERROR(EXDEV)); 3648 (void) zvol_remove_minor(name); 3649 } 3650 3651 error = dsl_bookmark_destroy(innvl, outnvl); 3652 return (error); 3653} 3654 3655/* 3656 * inputs: 3657 * zc_name name of dataset to destroy 3658 * zc_objset_type type of objset 3659 * zc_defer_destroy mark for deferred destroy 3660 * 3661 * outputs: none 3662 */ 3663static int 3664zfs_ioc_destroy(zfs_cmd_t *zc) 3665{ 3666 int err; 3667 3668 if (zc->zc_objset_type == DMU_OST_ZFS) { 3669 err = zfs_unmount_snap(zc->zc_name); 3670 if (err != 0) 3671 return (err); 3672 } 3673 3674 if (strchr(zc->zc_name, '@')) 3675 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy); 3676 else 3677 err = dsl_destroy_head(zc->zc_name); 3678 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0) 3679 (void) zvol_remove_minor(zc->zc_name); 3680 return (err); 3681} 3682 3683/* 3684 * fsname is name of dataset to rollback (to most recent snapshot) 3685 * 3686 * innvl is not used. 3687 * 3688 * outnvl: "target" -> name of most recent snapshot 3689 * } 3690 */ 3691/* ARGSUSED */ 3692static int 3693zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl) 3694{ 3695 zfsvfs_t *zfsvfs; 3696 int error; 3697 3698 if (getzfsvfs(fsname, &zfsvfs) == 0) { 3699 error = zfs_suspend_fs(zfsvfs); 3700 if (error == 0) { 3701 int resume_err; 3702 3703 error = dsl_dataset_rollback(fsname, zfsvfs, outnvl); 3704 resume_err = zfs_resume_fs(zfsvfs, fsname); 3705 error = error ? error : resume_err; 3706 } 3707 VFS_RELE(zfsvfs->z_vfs); 3708 } else { 3709 error = dsl_dataset_rollback(fsname, NULL, outnvl); 3710 } 3711 return (error); 3712} 3713 3714static int 3715recursive_unmount(const char *fsname, void *arg) 3716{ 3717 const char *snapname = arg; 3718 char fullname[MAXNAMELEN]; 3719 3720 (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname); 3721 return (zfs_unmount_snap(fullname)); 3722} 3723 3724/* 3725 * inputs: 3726 * zc_name old name of dataset 3727 * zc_value new name of dataset 3728 * zc_cookie recursive flag (only valid for snapshots) 3729 * 3730 * outputs: none 3731 */ 3732static int 3733zfs_ioc_rename(zfs_cmd_t *zc) 3734{ 3735 boolean_t recursive = zc->zc_cookie & 1; 3736#ifdef __FreeBSD__ 3737 boolean_t allow_mounted = zc->zc_cookie & 2; 3738#endif 3739 char *at; 3740 3741 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0'; 3742 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 3743 strchr(zc->zc_value, '%')) 3744 return (SET_ERROR(EINVAL)); 3745 3746 at = strchr(zc->zc_name, '@'); 3747 if (at != NULL) { 3748 /* snaps must be in same fs */ 3749 int error; 3750 3751 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1)) 3752 return (SET_ERROR(EXDEV)); 3753 *at = '\0'; 3754#ifdef illumos 3755 if (zc->zc_objset_type == DMU_OST_ZFS) { 3756#else 3757 if (zc->zc_objset_type == DMU_OST_ZFS && allow_mounted) { 3758#endif 3759 error = dmu_objset_find(zc->zc_name, 3760 recursive_unmount, at + 1, 3761 recursive ? DS_FIND_CHILDREN : 0); 3762 if (error != 0) { 3763 *at = '@'; 3764 return (error); 3765 } 3766 } 3767 error = dsl_dataset_rename_snapshot(zc->zc_name, 3768 at + 1, strchr(zc->zc_value, '@') + 1, recursive); 3769 *at = '@'; 3770 3771 return (error); 3772 } else { 3773#ifdef illumos 3774 if (zc->zc_objset_type == DMU_OST_ZVOL) 3775 (void) zvol_remove_minor(zc->zc_name); 3776#endif 3777 return (dsl_dir_rename(zc->zc_name, zc->zc_value)); 3778 } 3779} 3780 3781static int 3782zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr) 3783{ 3784 const char *propname = nvpair_name(pair); 3785 boolean_t issnap = (strchr(dsname, '@') != NULL); 3786 zfs_prop_t prop = zfs_name_to_prop(propname); 3787 uint64_t intval; 3788 int err; 3789 3790 if (prop == ZPROP_INVAL) { 3791 if (zfs_prop_user(propname)) { 3792 if (err = zfs_secpolicy_write_perms(dsname, 3793 ZFS_DELEG_PERM_USERPROP, cr)) 3794 return (err); 3795 return (0); 3796 } 3797 3798 if (!issnap && zfs_prop_userquota(propname)) { 3799 const char *perm = NULL; 3800 const char *uq_prefix = 3801 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA]; 3802 const char *gq_prefix = 3803 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA]; 3804 3805 if (strncmp(propname, uq_prefix, 3806 strlen(uq_prefix)) == 0) { 3807 perm = ZFS_DELEG_PERM_USERQUOTA; 3808 } else if (strncmp(propname, gq_prefix, 3809 strlen(gq_prefix)) == 0) { 3810 perm = ZFS_DELEG_PERM_GROUPQUOTA; 3811 } else { 3812 /* USERUSED and GROUPUSED are read-only */ 3813 return (SET_ERROR(EINVAL)); 3814 } 3815 3816 if (err = zfs_secpolicy_write_perms(dsname, perm, cr)) 3817 return (err); 3818 return (0); 3819 } 3820 3821 return (SET_ERROR(EINVAL)); 3822 } 3823 3824 if (issnap) 3825 return (SET_ERROR(EINVAL)); 3826 3827 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 3828 /* 3829 * dsl_prop_get_all_impl() returns properties in this 3830 * format. 3831 */ 3832 nvlist_t *attrs; 3833 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 3834 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 3835 &pair) == 0); 3836 } 3837 3838 /* 3839 * Check that this value is valid for this pool version 3840 */ 3841 switch (prop) { 3842 case ZFS_PROP_COMPRESSION: 3843 /* 3844 * If the user specified gzip compression, make sure 3845 * the SPA supports it. We ignore any errors here since 3846 * we'll catch them later. 3847 */ 3848 if (nvpair_type(pair) == DATA_TYPE_UINT64 && 3849 nvpair_value_uint64(pair, &intval) == 0) { 3850 if (intval >= ZIO_COMPRESS_GZIP_1 && 3851 intval <= ZIO_COMPRESS_GZIP_9 && 3852 zfs_earlier_version(dsname, 3853 SPA_VERSION_GZIP_COMPRESSION)) { 3854 return (SET_ERROR(ENOTSUP)); 3855 } 3856 3857 if (intval == ZIO_COMPRESS_ZLE && 3858 zfs_earlier_version(dsname, 3859 SPA_VERSION_ZLE_COMPRESSION)) 3860 return (SET_ERROR(ENOTSUP)); 3861 3862 if (intval == ZIO_COMPRESS_LZ4) { 3863 spa_t *spa; 3864 3865 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 3866 return (err); 3867 3868 if (!spa_feature_is_enabled(spa, 3869 SPA_FEATURE_LZ4_COMPRESS)) { 3870 spa_close(spa, FTAG); 3871 return (SET_ERROR(ENOTSUP)); 3872 } 3873 spa_close(spa, FTAG); 3874 } 3875 3876 /* 3877 * If this is a bootable dataset then 3878 * verify that the compression algorithm 3879 * is supported for booting. We must return 3880 * something other than ENOTSUP since it 3881 * implies a downrev pool version. 3882 */ 3883 if (zfs_is_bootfs(dsname) && 3884 !BOOTFS_COMPRESS_VALID(intval)) { 3885 return (SET_ERROR(ERANGE)); 3886 } 3887 } 3888 break; 3889 3890 case ZFS_PROP_COPIES: 3891 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS)) 3892 return (SET_ERROR(ENOTSUP)); 3893 break; 3894 3895 case ZFS_PROP_DEDUP: 3896 if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP)) 3897 return (SET_ERROR(ENOTSUP)); 3898 break; 3899 3900 case ZFS_PROP_SHARESMB: 3901 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID)) 3902 return (SET_ERROR(ENOTSUP)); 3903 break; 3904 3905 case ZFS_PROP_ACLINHERIT: 3906 if (nvpair_type(pair) == DATA_TYPE_UINT64 && 3907 nvpair_value_uint64(pair, &intval) == 0) { 3908 if (intval == ZFS_ACL_PASSTHROUGH_X && 3909 zfs_earlier_version(dsname, 3910 SPA_VERSION_PASSTHROUGH_X)) 3911 return (SET_ERROR(ENOTSUP)); 3912 } 3913 break; 3914 } 3915 3916 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED())); 3917} 3918 3919/* 3920 * Checks for a race condition to make sure we don't increment a feature flag 3921 * multiple times. 3922 */ 3923static int 3924zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx) 3925{ 3926 spa_t *spa = dmu_tx_pool(tx)->dp_spa; 3927 spa_feature_t *featurep = arg; 3928 3929 if (!spa_feature_is_active(spa, *featurep)) 3930 return (0); 3931 else 3932 return (SET_ERROR(EBUSY)); 3933} 3934 3935/* 3936 * The callback invoked on feature activation in the sync task caused by 3937 * zfs_prop_activate_feature. 3938 */ 3939static void 3940zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx) 3941{ 3942 spa_t *spa = dmu_tx_pool(tx)->dp_spa; 3943 spa_feature_t *featurep = arg; 3944 3945 spa_feature_incr(spa, *featurep, tx); 3946} 3947 3948/* 3949 * Activates a feature on a pool in response to a property setting. This 3950 * creates a new sync task which modifies the pool to reflect the feature 3951 * as being active. 3952 */ 3953static int 3954zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature) 3955{ 3956 int err; 3957 3958 /* EBUSY here indicates that the feature is already active */ 3959 err = dsl_sync_task(spa_name(spa), 3960 zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync, 3961 &feature, 2, ZFS_SPACE_CHECK_RESERVED); 3962 3963 if (err != 0 && err != EBUSY) 3964 return (err); 3965 else 3966 return (0); 3967} 3968 3969/* 3970 * Removes properties from the given props list that fail permission checks 3971 * needed to clear them and to restore them in case of a receive error. For each 3972 * property, make sure we have both set and inherit permissions. 3973 * 3974 * Returns the first error encountered if any permission checks fail. If the 3975 * caller provides a non-NULL errlist, it also gives the complete list of names 3976 * of all the properties that failed a permission check along with the 3977 * corresponding error numbers. The caller is responsible for freeing the 3978 * returned errlist. 3979 * 3980 * If every property checks out successfully, zero is returned and the list 3981 * pointed at by errlist is NULL. 3982 */ 3983static int 3984zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist) 3985{ 3986 zfs_cmd_t *zc; 3987 nvpair_t *pair, *next_pair; 3988 nvlist_t *errors; 3989 int err, rv = 0; 3990 3991 if (props == NULL) 3992 return (0); 3993 3994 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0); 3995 3996 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP); 3997 (void) strcpy(zc->zc_name, dataset); 3998 pair = nvlist_next_nvpair(props, NULL); 3999 while (pair != NULL) { 4000 next_pair = nvlist_next_nvpair(props, pair); 4001 4002 (void) strcpy(zc->zc_value, nvpair_name(pair)); 4003 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 || 4004 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) { 4005 VERIFY(nvlist_remove_nvpair(props, pair) == 0); 4006 VERIFY(nvlist_add_int32(errors, 4007 zc->zc_value, err) == 0); 4008 } 4009 pair = next_pair; 4010 } 4011 kmem_free(zc, sizeof (zfs_cmd_t)); 4012 4013 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) { 4014 nvlist_free(errors); 4015 errors = NULL; 4016 } else { 4017 VERIFY(nvpair_value_int32(pair, &rv) == 0); 4018 } 4019 4020 if (errlist == NULL) 4021 nvlist_free(errors); 4022 else 4023 *errlist = errors; 4024 4025 return (rv); 4026} 4027 4028static boolean_t 4029propval_equals(nvpair_t *p1, nvpair_t *p2) 4030{ 4031 if (nvpair_type(p1) == DATA_TYPE_NVLIST) { 4032 /* dsl_prop_get_all_impl() format */ 4033 nvlist_t *attrs; 4034 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0); 4035 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 4036 &p1) == 0); 4037 } 4038 4039 if (nvpair_type(p2) == DATA_TYPE_NVLIST) { 4040 nvlist_t *attrs; 4041 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0); 4042 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 4043 &p2) == 0); 4044 } 4045 4046 if (nvpair_type(p1) != nvpair_type(p2)) 4047 return (B_FALSE); 4048 4049 if (nvpair_type(p1) == DATA_TYPE_STRING) { 4050 char *valstr1, *valstr2; 4051 4052 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0); 4053 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0); 4054 return (strcmp(valstr1, valstr2) == 0); 4055 } else { 4056 uint64_t intval1, intval2; 4057 4058 VERIFY(nvpair_value_uint64(p1, &intval1) == 0); 4059 VERIFY(nvpair_value_uint64(p2, &intval2) == 0); 4060 return (intval1 == intval2); 4061 } 4062} 4063 4064/* 4065 * Remove properties from props if they are not going to change (as determined 4066 * by comparison with origprops). Remove them from origprops as well, since we 4067 * do not need to clear or restore properties that won't change. 4068 */ 4069static void 4070props_reduce(nvlist_t *props, nvlist_t *origprops) 4071{ 4072 nvpair_t *pair, *next_pair; 4073 4074 if (origprops == NULL) 4075 return; /* all props need to be received */ 4076 4077 pair = nvlist_next_nvpair(props, NULL); 4078 while (pair != NULL) { 4079 const char *propname = nvpair_name(pair); 4080 nvpair_t *match; 4081 4082 next_pair = nvlist_next_nvpair(props, pair); 4083 4084 if ((nvlist_lookup_nvpair(origprops, propname, 4085 &match) != 0) || !propval_equals(pair, match)) 4086 goto next; /* need to set received value */ 4087 4088 /* don't clear the existing received value */ 4089 (void) nvlist_remove_nvpair(origprops, match); 4090 /* don't bother receiving the property */ 4091 (void) nvlist_remove_nvpair(props, pair); 4092next: 4093 pair = next_pair; 4094 } 4095} 4096 4097#ifdef DEBUG 4098static boolean_t zfs_ioc_recv_inject_err; 4099#endif 4100 4101/* 4102 * inputs: 4103 * zc_name name of containing filesystem 4104 * zc_nvlist_src{_size} nvlist of properties to apply 4105 * zc_value name of snapshot to create 4106 * zc_string name of clone origin (if DRR_FLAG_CLONE) 4107 * zc_cookie file descriptor to recv from 4108 * zc_begin_record the BEGIN record of the stream (not byteswapped) 4109 * zc_guid force flag 4110 * zc_cleanup_fd cleanup-on-exit file descriptor 4111 * zc_action_handle handle for this guid/ds mapping (or zero on first call) 4112 * 4113 * outputs: 4114 * zc_cookie number of bytes read 4115 * zc_nvlist_dst{_size} error for each unapplied received property 4116 * zc_obj zprop_errflags_t 4117 * zc_action_handle handle for this guid/ds mapping 4118 */ 4119static int 4120zfs_ioc_recv(zfs_cmd_t *zc) 4121{ 4122 file_t *fp; 4123 dmu_recv_cookie_t drc; 4124 boolean_t force = (boolean_t)zc->zc_guid; 4125 int fd; 4126 int error = 0; 4127 int props_error = 0; 4128 nvlist_t *errors; 4129 offset_t off; 4130 nvlist_t *props = NULL; /* sent properties */ 4131 nvlist_t *origprops = NULL; /* existing properties */ 4132 char *origin = NULL; 4133 char *tosnap; 4134 char tofs[ZFS_MAXNAMELEN]; 4135 cap_rights_t rights; 4136 boolean_t first_recvd_props = B_FALSE; 4137 4138 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 4139 strchr(zc->zc_value, '@') == NULL || 4140 strchr(zc->zc_value, '%')) 4141 return (SET_ERROR(EINVAL)); 4142 4143 (void) strcpy(tofs, zc->zc_value); 4144 tosnap = strchr(tofs, '@'); 4145 *tosnap++ = '\0'; 4146 4147 if (zc->zc_nvlist_src != 0 && 4148 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 4149 zc->zc_iflags, &props)) != 0) 4150 return (error); 4151 4152 fd = zc->zc_cookie; 4153 fp = getf(fd, cap_rights_init(&rights, CAP_PREAD)); 4154 if (fp == NULL) { 4155 nvlist_free(props); 4156 return (SET_ERROR(EBADF)); 4157 } 4158 4159 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0); 4160 4161 if (zc->zc_string[0]) 4162 origin = zc->zc_string; 4163 4164 error = dmu_recv_begin(tofs, tosnap, 4165 &zc->zc_begin_record, force, origin, &drc); 4166 if (error != 0) 4167 goto out; 4168 4169 /* 4170 * Set properties before we receive the stream so that they are applied 4171 * to the new data. Note that we must call dmu_recv_stream() if 4172 * dmu_recv_begin() succeeds. 4173 */ 4174 if (props != NULL && !drc.drc_newfs) { 4175 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >= 4176 SPA_VERSION_RECVD_PROPS && 4177 !dsl_prop_get_hasrecvd(tofs)) 4178 first_recvd_props = B_TRUE; 4179 4180 /* 4181 * If new received properties are supplied, they are to 4182 * completely replace the existing received properties, so stash 4183 * away the existing ones. 4184 */ 4185 if (dsl_prop_get_received(tofs, &origprops) == 0) { 4186 nvlist_t *errlist = NULL; 4187 /* 4188 * Don't bother writing a property if its value won't 4189 * change (and avoid the unnecessary security checks). 4190 * 4191 * The first receive after SPA_VERSION_RECVD_PROPS is a 4192 * special case where we blow away all local properties 4193 * regardless. 4194 */ 4195 if (!first_recvd_props) 4196 props_reduce(props, origprops); 4197 if (zfs_check_clearable(tofs, origprops, &errlist) != 0) 4198 (void) nvlist_merge(errors, errlist, 0); 4199 nvlist_free(errlist); 4200 4201 if (clear_received_props(tofs, origprops, 4202 first_recvd_props ? NULL : props) != 0) 4203 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 4204 } else { 4205 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 4206 } 4207 } 4208 4209 if (props != NULL) { 4210 props_error = dsl_prop_set_hasrecvd(tofs); 4211 4212 if (props_error == 0) { 4213 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED, 4214 props, errors); 4215 } 4216 } 4217 4218 if (zc->zc_nvlist_dst_size != 0 && 4219 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 || 4220 put_nvlist(zc, errors) != 0)) { 4221 /* 4222 * Caller made zc->zc_nvlist_dst less than the minimum expected 4223 * size or supplied an invalid address. 4224 */ 4225 props_error = SET_ERROR(EINVAL); 4226 } 4227 4228 off = fp->f_offset; 4229 error = dmu_recv_stream(&drc, fp, &off, zc->zc_cleanup_fd, 4230 &zc->zc_action_handle); 4231 4232 if (error == 0) { 4233 zfsvfs_t *zfsvfs = NULL; 4234 4235 if (getzfsvfs(tofs, &zfsvfs) == 0) { 4236 /* online recv */ 4237 int end_err; 4238 4239 error = zfs_suspend_fs(zfsvfs); 4240 /* 4241 * If the suspend fails, then the recv_end will 4242 * likely also fail, and clean up after itself. 4243 */ 4244 end_err = dmu_recv_end(&drc, zfsvfs); 4245 if (error == 0) 4246 error = zfs_resume_fs(zfsvfs, tofs); 4247 error = error ? error : end_err; 4248 VFS_RELE(zfsvfs->z_vfs); 4249 } else { 4250 error = dmu_recv_end(&drc, NULL); 4251 } 4252 } 4253 4254 zc->zc_cookie = off - fp->f_offset; 4255 if (off >= 0 && off <= MAXOFFSET_T) 4256 fp->f_offset = off; 4257 4258#ifdef DEBUG 4259 if (zfs_ioc_recv_inject_err) { 4260 zfs_ioc_recv_inject_err = B_FALSE; 4261 error = 1; 4262 } 4263#endif 4264 4265#ifdef __FreeBSD__ 4266 if (error == 0) 4267 zvol_create_minors(tofs); 4268#endif 4269 4270 /* 4271 * On error, restore the original props. 4272 */ 4273 if (error != 0 && props != NULL && !drc.drc_newfs) { 4274 if (clear_received_props(tofs, props, NULL) != 0) { 4275 /* 4276 * We failed to clear the received properties. 4277 * Since we may have left a $recvd value on the 4278 * system, we can't clear the $hasrecvd flag. 4279 */ 4280 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4281 } else if (first_recvd_props) { 4282 dsl_prop_unset_hasrecvd(tofs); 4283 } 4284 4285 if (origprops == NULL && !drc.drc_newfs) { 4286 /* We failed to stash the original properties. */ 4287 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4288 } 4289 4290 /* 4291 * dsl_props_set() will not convert RECEIVED to LOCAL on or 4292 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL 4293 * explictly if we're restoring local properties cleared in the 4294 * first new-style receive. 4295 */ 4296 if (origprops != NULL && 4297 zfs_set_prop_nvlist(tofs, (first_recvd_props ? 4298 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED), 4299 origprops, NULL) != 0) { 4300 /* 4301 * We stashed the original properties but failed to 4302 * restore them. 4303 */ 4304 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4305 } 4306 } 4307out: 4308 nvlist_free(props); 4309 nvlist_free(origprops); 4310 nvlist_free(errors); 4311 releasef(fd); 4312 4313 if (error == 0) 4314 error = props_error; 4315 4316 return (error); 4317} 4318 4319/* 4320 * inputs: 4321 * zc_name name of snapshot to send 4322 * zc_cookie file descriptor to send stream to 4323 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj) 4324 * zc_sendobj objsetid of snapshot to send 4325 * zc_fromobj objsetid of incremental fromsnap (may be zero) 4326 * zc_guid if set, estimate size of stream only. zc_cookie is ignored. 4327 * output size in zc_objset_type. 4328 * zc_flags if =1, WRITE_EMBEDDED records are permitted 4329 * 4330 * outputs: 4331 * zc_objset_type estimated size, if zc_guid is set 4332 */ 4333static int 4334zfs_ioc_send(zfs_cmd_t *zc) 4335{ 4336 int error; 4337 offset_t off; 4338 boolean_t estimate = (zc->zc_guid != 0); 4339 boolean_t embedok = (zc->zc_flags & 0x1); 4340 4341 if (zc->zc_obj != 0) { 4342 dsl_pool_t *dp; 4343 dsl_dataset_t *tosnap; 4344 4345 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4346 if (error != 0) 4347 return (error); 4348 4349 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap); 4350 if (error != 0) { 4351 dsl_pool_rele(dp, FTAG); 4352 return (error); 4353 } 4354 4355 if (dsl_dir_is_clone(tosnap->ds_dir)) 4356 zc->zc_fromobj = tosnap->ds_dir->dd_phys->dd_origin_obj; 4357 dsl_dataset_rele(tosnap, FTAG); 4358 dsl_pool_rele(dp, FTAG); 4359 } 4360 4361 if (estimate) { 4362 dsl_pool_t *dp; 4363 dsl_dataset_t *tosnap; 4364 dsl_dataset_t *fromsnap = NULL; 4365 4366 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4367 if (error != 0) 4368 return (error); 4369 4370 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap); 4371 if (error != 0) { 4372 dsl_pool_rele(dp, FTAG); 4373 return (error); 4374 } 4375 4376 if (zc->zc_fromobj != 0) { 4377 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj, 4378 FTAG, &fromsnap); 4379 if (error != 0) { 4380 dsl_dataset_rele(tosnap, FTAG); 4381 dsl_pool_rele(dp, FTAG); 4382 return (error); 4383 } 4384 } 4385 4386 error = dmu_send_estimate(tosnap, fromsnap, 4387 &zc->zc_objset_type); 4388 4389 if (fromsnap != NULL) 4390 dsl_dataset_rele(fromsnap, FTAG); 4391 dsl_dataset_rele(tosnap, FTAG); 4392 dsl_pool_rele(dp, FTAG); 4393 } else { 4394 file_t *fp; 4395 cap_rights_t rights; 4396 4397 fp = getf(zc->zc_cookie, 4398 cap_rights_init(&rights, CAP_WRITE)); 4399 if (fp == NULL) 4400 return (SET_ERROR(EBADF)); 4401 4402 off = fp->f_offset; 4403 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj, 4404#ifdef illumos 4405 zc->zc_fromobj, embedok, zc->zc_cookie, fp->f_vnode, &off); 4406#else 4407 zc->zc_fromobj, embedok, zc->zc_cookie, fp, &off); 4408#endif 4409 4410 if (off >= 0 && off <= MAXOFFSET_T) 4411 fp->f_offset = off; 4412 releasef(zc->zc_cookie); 4413 } 4414 return (error); 4415} 4416 4417/* 4418 * inputs: 4419 * zc_name name of snapshot on which to report progress 4420 * zc_cookie file descriptor of send stream 4421 * 4422 * outputs: 4423 * zc_cookie number of bytes written in send stream thus far 4424 */ 4425static int 4426zfs_ioc_send_progress(zfs_cmd_t *zc) 4427{ 4428 dsl_pool_t *dp; 4429 dsl_dataset_t *ds; 4430 dmu_sendarg_t *dsp = NULL; 4431 int error; 4432 4433 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4434 if (error != 0) 4435 return (error); 4436 4437 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds); 4438 if (error != 0) { 4439 dsl_pool_rele(dp, FTAG); 4440 return (error); 4441 } 4442 4443 mutex_enter(&ds->ds_sendstream_lock); 4444 4445 /* 4446 * Iterate over all the send streams currently active on this dataset. 4447 * If there's one which matches the specified file descriptor _and_ the 4448 * stream was started by the current process, return the progress of 4449 * that stream. 4450 */ 4451 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL; 4452 dsp = list_next(&ds->ds_sendstreams, dsp)) { 4453 if (dsp->dsa_outfd == zc->zc_cookie && 4454 dsp->dsa_proc == curproc) 4455 break; 4456 } 4457 4458 if (dsp != NULL) 4459 zc->zc_cookie = *(dsp->dsa_off); 4460 else 4461 error = SET_ERROR(ENOENT); 4462 4463 mutex_exit(&ds->ds_sendstream_lock); 4464 dsl_dataset_rele(ds, FTAG); 4465 dsl_pool_rele(dp, FTAG); 4466 return (error); 4467} 4468 4469static int 4470zfs_ioc_inject_fault(zfs_cmd_t *zc) 4471{ 4472 int id, error; 4473 4474 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id, 4475 &zc->zc_inject_record); 4476 4477 if (error == 0) 4478 zc->zc_guid = (uint64_t)id; 4479 4480 return (error); 4481} 4482 4483static int 4484zfs_ioc_clear_fault(zfs_cmd_t *zc) 4485{ 4486 return (zio_clear_fault((int)zc->zc_guid)); 4487} 4488 4489static int 4490zfs_ioc_inject_list_next(zfs_cmd_t *zc) 4491{ 4492 int id = (int)zc->zc_guid; 4493 int error; 4494 4495 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name), 4496 &zc->zc_inject_record); 4497 4498 zc->zc_guid = id; 4499 4500 return (error); 4501} 4502 4503static int 4504zfs_ioc_error_log(zfs_cmd_t *zc) 4505{ 4506 spa_t *spa; 4507 int error; 4508 size_t count = (size_t)zc->zc_nvlist_dst_size; 4509 4510 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 4511 return (error); 4512 4513 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst, 4514 &count); 4515 if (error == 0) 4516 zc->zc_nvlist_dst_size = count; 4517 else 4518 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa); 4519 4520 spa_close(spa, FTAG); 4521 4522 return (error); 4523} 4524 4525static int 4526zfs_ioc_clear(zfs_cmd_t *zc) 4527{ 4528 spa_t *spa; 4529 vdev_t *vd; 4530 int error; 4531 4532 /* 4533 * On zpool clear we also fix up missing slogs 4534 */ 4535 mutex_enter(&spa_namespace_lock); 4536 spa = spa_lookup(zc->zc_name); 4537 if (spa == NULL) { 4538 mutex_exit(&spa_namespace_lock); 4539 return (SET_ERROR(EIO)); 4540 } 4541 if (spa_get_log_state(spa) == SPA_LOG_MISSING) { 4542 /* we need to let spa_open/spa_load clear the chains */ 4543 spa_set_log_state(spa, SPA_LOG_CLEAR); 4544 } 4545 spa->spa_last_open_failed = 0; 4546 mutex_exit(&spa_namespace_lock); 4547 4548 if (zc->zc_cookie & ZPOOL_NO_REWIND) { 4549 error = spa_open(zc->zc_name, &spa, FTAG); 4550 } else { 4551 nvlist_t *policy; 4552 nvlist_t *config = NULL; 4553 4554 if (zc->zc_nvlist_src == 0) 4555 return (SET_ERROR(EINVAL)); 4556 4557 if ((error = get_nvlist(zc->zc_nvlist_src, 4558 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) { 4559 error = spa_open_rewind(zc->zc_name, &spa, FTAG, 4560 policy, &config); 4561 if (config != NULL) { 4562 int err; 4563 4564 if ((err = put_nvlist(zc, config)) != 0) 4565 error = err; 4566 nvlist_free(config); 4567 } 4568 nvlist_free(policy); 4569 } 4570 } 4571 4572 if (error != 0) 4573 return (error); 4574 4575 spa_vdev_state_enter(spa, SCL_NONE); 4576 4577 if (zc->zc_guid == 0) { 4578 vd = NULL; 4579 } else { 4580 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE); 4581 if (vd == NULL) { 4582 (void) spa_vdev_state_exit(spa, NULL, ENODEV); 4583 spa_close(spa, FTAG); 4584 return (SET_ERROR(ENODEV)); 4585 } 4586 } 4587 4588 vdev_clear(spa, vd); 4589 4590 (void) spa_vdev_state_exit(spa, NULL, 0); 4591 4592 /* 4593 * Resume any suspended I/Os. 4594 */ 4595 if (zio_resume(spa) != 0) 4596 error = SET_ERROR(EIO); 4597 4598 spa_close(spa, FTAG); 4599 4600 return (error); 4601} 4602 4603static int 4604zfs_ioc_pool_reopen(zfs_cmd_t *zc) 4605{ 4606 spa_t *spa; 4607 int error; 4608 4609 error = spa_open(zc->zc_name, &spa, FTAG); 4610 if (error != 0) 4611 return (error); 4612 4613 spa_vdev_state_enter(spa, SCL_NONE); 4614 4615 /* 4616 * If a resilver is already in progress then set the 4617 * spa_scrub_reopen flag to B_TRUE so that we don't restart 4618 * the scan as a side effect of the reopen. Otherwise, let 4619 * vdev_open() decided if a resilver is required. 4620 */ 4621 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool); 4622 vdev_reopen(spa->spa_root_vdev); 4623 spa->spa_scrub_reopen = B_FALSE; 4624 4625 (void) spa_vdev_state_exit(spa, NULL, 0); 4626 spa_close(spa, FTAG); 4627 return (0); 4628} 4629/* 4630 * inputs: 4631 * zc_name name of filesystem 4632 * zc_value name of origin snapshot 4633 * 4634 * outputs: 4635 * zc_string name of conflicting snapshot, if there is one 4636 */ 4637static int 4638zfs_ioc_promote(zfs_cmd_t *zc) 4639{ 4640 char *cp; 4641 4642 /* 4643 * We don't need to unmount *all* the origin fs's snapshots, but 4644 * it's easier. 4645 */ 4646 cp = strchr(zc->zc_value, '@'); 4647 if (cp) 4648 *cp = '\0'; 4649 (void) dmu_objset_find(zc->zc_value, 4650 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS); 4651 return (dsl_dataset_promote(zc->zc_name, zc->zc_string)); 4652} 4653 4654/* 4655 * Retrieve a single {user|group}{used|quota}@... property. 4656 * 4657 * inputs: 4658 * zc_name name of filesystem 4659 * zc_objset_type zfs_userquota_prop_t 4660 * zc_value domain name (eg. "S-1-234-567-89") 4661 * zc_guid RID/UID/GID 4662 * 4663 * outputs: 4664 * zc_cookie property value 4665 */ 4666static int 4667zfs_ioc_userspace_one(zfs_cmd_t *zc) 4668{ 4669 zfsvfs_t *zfsvfs; 4670 int error; 4671 4672 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 4673 return (SET_ERROR(EINVAL)); 4674 4675 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE); 4676 if (error != 0) 4677 return (error); 4678 4679 error = zfs_userspace_one(zfsvfs, 4680 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie); 4681 zfsvfs_rele(zfsvfs, FTAG); 4682 4683 return (error); 4684} 4685 4686/* 4687 * inputs: 4688 * zc_name name of filesystem 4689 * zc_cookie zap cursor 4690 * zc_objset_type zfs_userquota_prop_t 4691 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist) 4692 * 4693 * outputs: 4694 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t) 4695 * zc_cookie zap cursor 4696 */ 4697static int 4698zfs_ioc_userspace_many(zfs_cmd_t *zc) 4699{ 4700 zfsvfs_t *zfsvfs; 4701 int bufsize = zc->zc_nvlist_dst_size; 4702 4703 if (bufsize <= 0) 4704 return (SET_ERROR(ENOMEM)); 4705 4706 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE); 4707 if (error != 0) 4708 return (error); 4709 4710 void *buf = kmem_alloc(bufsize, KM_SLEEP); 4711 4712 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie, 4713 buf, &zc->zc_nvlist_dst_size); 4714 4715 if (error == 0) { 4716 error = ddi_copyout(buf, 4717 (void *)(uintptr_t)zc->zc_nvlist_dst, 4718 zc->zc_nvlist_dst_size, zc->zc_iflags); 4719 } 4720 kmem_free(buf, bufsize); 4721 zfsvfs_rele(zfsvfs, FTAG); 4722 4723 return (error); 4724} 4725 4726/* 4727 * inputs: 4728 * zc_name name of filesystem 4729 * 4730 * outputs: 4731 * none 4732 */ 4733static int 4734zfs_ioc_userspace_upgrade(zfs_cmd_t *zc) 4735{ 4736 objset_t *os; 4737 int error = 0; 4738 zfsvfs_t *zfsvfs; 4739 4740 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) { 4741 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) { 4742 /* 4743 * If userused is not enabled, it may be because the 4744 * objset needs to be closed & reopened (to grow the 4745 * objset_phys_t). Suspend/resume the fs will do that. 4746 */ 4747 error = zfs_suspend_fs(zfsvfs); 4748 if (error == 0) { 4749 dmu_objset_refresh_ownership(zfsvfs->z_os, 4750 zfsvfs); 4751 error = zfs_resume_fs(zfsvfs, zc->zc_name); 4752 } 4753 } 4754 if (error == 0) 4755 error = dmu_objset_userspace_upgrade(zfsvfs->z_os); 4756 VFS_RELE(zfsvfs->z_vfs); 4757 } else { 4758 /* XXX kind of reading contents without owning */ 4759 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 4760 if (error != 0) 4761 return (error); 4762 4763 error = dmu_objset_userspace_upgrade(os); 4764 dmu_objset_rele(os, FTAG); 4765 } 4766 4767 return (error); 4768} 4769 4770#ifdef sun 4771/* 4772 * We don't want to have a hard dependency 4773 * against some special symbols in sharefs 4774 * nfs, and smbsrv. Determine them if needed when 4775 * the first file system is shared. 4776 * Neither sharefs, nfs or smbsrv are unloadable modules. 4777 */ 4778int (*znfsexport_fs)(void *arg); 4779int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t); 4780int (*zsmbexport_fs)(void *arg, boolean_t add_share); 4781 4782int zfs_nfsshare_inited; 4783int zfs_smbshare_inited; 4784 4785ddi_modhandle_t nfs_mod; 4786ddi_modhandle_t sharefs_mod; 4787ddi_modhandle_t smbsrv_mod; 4788#endif /* sun */ 4789kmutex_t zfs_share_lock; 4790 4791#ifdef sun 4792static int 4793zfs_init_sharefs() 4794{ 4795 int error; 4796 4797 ASSERT(MUTEX_HELD(&zfs_share_lock)); 4798 /* Both NFS and SMB shares also require sharetab support. */ 4799 if (sharefs_mod == NULL && ((sharefs_mod = 4800 ddi_modopen("fs/sharefs", 4801 KRTLD_MODE_FIRST, &error)) == NULL)) { 4802 return (SET_ERROR(ENOSYS)); 4803 } 4804 if (zshare_fs == NULL && ((zshare_fs = 4805 (int (*)(enum sharefs_sys_op, share_t *, uint32_t)) 4806 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) { 4807 return (SET_ERROR(ENOSYS)); 4808 } 4809 return (0); 4810} 4811#endif /* sun */ 4812 4813static int 4814zfs_ioc_share(zfs_cmd_t *zc) 4815{ 4816#ifdef sun 4817 int error; 4818 int opcode; 4819 4820 switch (zc->zc_share.z_sharetype) { 4821 case ZFS_SHARE_NFS: 4822 case ZFS_UNSHARE_NFS: 4823 if (zfs_nfsshare_inited == 0) { 4824 mutex_enter(&zfs_share_lock); 4825 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs", 4826 KRTLD_MODE_FIRST, &error)) == NULL)) { 4827 mutex_exit(&zfs_share_lock); 4828 return (SET_ERROR(ENOSYS)); 4829 } 4830 if (znfsexport_fs == NULL && 4831 ((znfsexport_fs = (int (*)(void *)) 4832 ddi_modsym(nfs_mod, 4833 "nfs_export", &error)) == NULL)) { 4834 mutex_exit(&zfs_share_lock); 4835 return (SET_ERROR(ENOSYS)); 4836 } 4837 error = zfs_init_sharefs(); 4838 if (error != 0) { 4839 mutex_exit(&zfs_share_lock); 4840 return (SET_ERROR(ENOSYS)); 4841 } 4842 zfs_nfsshare_inited = 1; 4843 mutex_exit(&zfs_share_lock); 4844 } 4845 break; 4846 case ZFS_SHARE_SMB: 4847 case ZFS_UNSHARE_SMB: 4848 if (zfs_smbshare_inited == 0) { 4849 mutex_enter(&zfs_share_lock); 4850 if (smbsrv_mod == NULL && ((smbsrv_mod = 4851 ddi_modopen("drv/smbsrv", 4852 KRTLD_MODE_FIRST, &error)) == NULL)) { 4853 mutex_exit(&zfs_share_lock); 4854 return (SET_ERROR(ENOSYS)); 4855 } 4856 if (zsmbexport_fs == NULL && ((zsmbexport_fs = 4857 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod, 4858 "smb_server_share", &error)) == NULL)) { 4859 mutex_exit(&zfs_share_lock); 4860 return (SET_ERROR(ENOSYS)); 4861 } 4862 error = zfs_init_sharefs(); 4863 if (error != 0) { 4864 mutex_exit(&zfs_share_lock); 4865 return (SET_ERROR(ENOSYS)); 4866 } 4867 zfs_smbshare_inited = 1; 4868 mutex_exit(&zfs_share_lock); 4869 } 4870 break; 4871 default: 4872 return (SET_ERROR(EINVAL)); 4873 } 4874 4875 switch (zc->zc_share.z_sharetype) { 4876 case ZFS_SHARE_NFS: 4877 case ZFS_UNSHARE_NFS: 4878 if (error = 4879 znfsexport_fs((void *) 4880 (uintptr_t)zc->zc_share.z_exportdata)) 4881 return (error); 4882 break; 4883 case ZFS_SHARE_SMB: 4884 case ZFS_UNSHARE_SMB: 4885 if (error = zsmbexport_fs((void *) 4886 (uintptr_t)zc->zc_share.z_exportdata, 4887 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ? 4888 B_TRUE: B_FALSE)) { 4889 return (error); 4890 } 4891 break; 4892 } 4893 4894 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS || 4895 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ? 4896 SHAREFS_ADD : SHAREFS_REMOVE; 4897 4898 /* 4899 * Add or remove share from sharetab 4900 */ 4901 error = zshare_fs(opcode, 4902 (void *)(uintptr_t)zc->zc_share.z_sharedata, 4903 zc->zc_share.z_sharemax); 4904 4905 return (error); 4906 4907#else /* !sun */ 4908 return (ENOSYS); 4909#endif /* !sun */ 4910} 4911 4912ace_t full_access[] = { 4913 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0} 4914}; 4915 4916/* 4917 * inputs: 4918 * zc_name name of containing filesystem 4919 * zc_obj object # beyond which we want next in-use object # 4920 * 4921 * outputs: 4922 * zc_obj next in-use object # 4923 */ 4924static int 4925zfs_ioc_next_obj(zfs_cmd_t *zc) 4926{ 4927 objset_t *os = NULL; 4928 int error; 4929 4930 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 4931 if (error != 0) 4932 return (error); 4933 4934 error = dmu_object_next(os, &zc->zc_obj, B_FALSE, 4935 os->os_dsl_dataset->ds_phys->ds_prev_snap_txg); 4936 4937 dmu_objset_rele(os, FTAG); 4938 return (error); 4939} 4940 4941/* 4942 * inputs: 4943 * zc_name name of filesystem 4944 * zc_value prefix name for snapshot 4945 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process 4946 * 4947 * outputs: 4948 * zc_value short name of new snapshot 4949 */ 4950static int 4951zfs_ioc_tmp_snapshot(zfs_cmd_t *zc) 4952{ 4953 char *snap_name; 4954 char *hold_name; 4955 int error; 4956 minor_t minor; 4957 4958 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor); 4959 if (error != 0) 4960 return (error); 4961 4962 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value, 4963 (u_longlong_t)ddi_get_lbolt64()); 4964 hold_name = kmem_asprintf("%%%s", zc->zc_value); 4965 4966 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor, 4967 hold_name); 4968 if (error == 0) 4969 (void) strcpy(zc->zc_value, snap_name); 4970 strfree(snap_name); 4971 strfree(hold_name); 4972 zfs_onexit_fd_rele(zc->zc_cleanup_fd); 4973 return (error); 4974} 4975 4976/* 4977 * inputs: 4978 * zc_name name of "to" snapshot 4979 * zc_value name of "from" snapshot 4980 * zc_cookie file descriptor to write diff data on 4981 * 4982 * outputs: 4983 * dmu_diff_record_t's to the file descriptor 4984 */ 4985static int 4986zfs_ioc_diff(zfs_cmd_t *zc) 4987{ 4988 file_t *fp; 4989 cap_rights_t rights; 4990 offset_t off; 4991 int error; 4992 4993 fp = getf(zc->zc_cookie, cap_rights_init(&rights, CAP_WRITE)); 4994 if (fp == NULL) 4995 return (SET_ERROR(EBADF)); 4996 4997 off = fp->f_offset; 4998 4999#ifdef illumos 5000 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off); 5001#else 5002 error = dmu_diff(zc->zc_name, zc->zc_value, fp, &off); 5003#endif 5004 5005 if (off >= 0 && off <= MAXOFFSET_T) 5006 fp->f_offset = off; 5007 releasef(zc->zc_cookie); 5008 5009 return (error); 5010} 5011 5012#ifdef sun 5013/* 5014 * Remove all ACL files in shares dir 5015 */ 5016static int 5017zfs_smb_acl_purge(znode_t *dzp) 5018{ 5019 zap_cursor_t zc; 5020 zap_attribute_t zap; 5021 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 5022 int error; 5023 5024 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id); 5025 (error = zap_cursor_retrieve(&zc, &zap)) == 0; 5026 zap_cursor_advance(&zc)) { 5027 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred, 5028 NULL, 0)) != 0) 5029 break; 5030 } 5031 zap_cursor_fini(&zc); 5032 return (error); 5033} 5034#endif /* sun */ 5035 5036static int 5037zfs_ioc_smb_acl(zfs_cmd_t *zc) 5038{ 5039#ifdef sun 5040 vnode_t *vp; 5041 znode_t *dzp; 5042 vnode_t *resourcevp = NULL; 5043 znode_t *sharedir; 5044 zfsvfs_t *zfsvfs; 5045 nvlist_t *nvlist; 5046 char *src, *target; 5047 vattr_t vattr; 5048 vsecattr_t vsec; 5049 int error = 0; 5050 5051 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, 5052 NO_FOLLOW, NULL, &vp)) != 0) 5053 return (error); 5054 5055 /* Now make sure mntpnt and dataset are ZFS */ 5056 5057 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 || 5058 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), 5059 zc->zc_name) != 0)) { 5060 VN_RELE(vp); 5061 return (SET_ERROR(EINVAL)); 5062 } 5063 5064 dzp = VTOZ(vp); 5065 zfsvfs = dzp->z_zfsvfs; 5066 ZFS_ENTER(zfsvfs); 5067 5068 /* 5069 * Create share dir if its missing. 5070 */ 5071 mutex_enter(&zfsvfs->z_lock); 5072 if (zfsvfs->z_shares_dir == 0) { 5073 dmu_tx_t *tx; 5074 5075 tx = dmu_tx_create(zfsvfs->z_os); 5076 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE, 5077 ZFS_SHARES_DIR); 5078 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); 5079 error = dmu_tx_assign(tx, TXG_WAIT); 5080 if (error != 0) { 5081 dmu_tx_abort(tx); 5082 } else { 5083 error = zfs_create_share_dir(zfsvfs, tx); 5084 dmu_tx_commit(tx); 5085 } 5086 if (error != 0) { 5087 mutex_exit(&zfsvfs->z_lock); 5088 VN_RELE(vp); 5089 ZFS_EXIT(zfsvfs); 5090 return (error); 5091 } 5092 } 5093 mutex_exit(&zfsvfs->z_lock); 5094 5095 ASSERT(zfsvfs->z_shares_dir); 5096 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) { 5097 VN_RELE(vp); 5098 ZFS_EXIT(zfsvfs); 5099 return (error); 5100 } 5101 5102 switch (zc->zc_cookie) { 5103 case ZFS_SMB_ACL_ADD: 5104 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; 5105 vattr.va_type = VREG; 5106 vattr.va_mode = S_IFREG|0777; 5107 vattr.va_uid = 0; 5108 vattr.va_gid = 0; 5109 5110 vsec.vsa_mask = VSA_ACE; 5111 vsec.vsa_aclentp = &full_access; 5112 vsec.vsa_aclentsz = sizeof (full_access); 5113 vsec.vsa_aclcnt = 1; 5114 5115 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string, 5116 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec); 5117 if (resourcevp) 5118 VN_RELE(resourcevp); 5119 break; 5120 5121 case ZFS_SMB_ACL_REMOVE: 5122 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred, 5123 NULL, 0); 5124 break; 5125 5126 case ZFS_SMB_ACL_RENAME: 5127 if ((error = get_nvlist(zc->zc_nvlist_src, 5128 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) { 5129 VN_RELE(vp); 5130 ZFS_EXIT(zfsvfs); 5131 return (error); 5132 } 5133 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) || 5134 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET, 5135 &target)) { 5136 VN_RELE(vp); 5137 VN_RELE(ZTOV(sharedir)); 5138 ZFS_EXIT(zfsvfs); 5139 nvlist_free(nvlist); 5140 return (error); 5141 } 5142 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target, 5143 kcred, NULL, 0); 5144 nvlist_free(nvlist); 5145 break; 5146 5147 case ZFS_SMB_ACL_PURGE: 5148 error = zfs_smb_acl_purge(sharedir); 5149 break; 5150 5151 default: 5152 error = SET_ERROR(EINVAL); 5153 break; 5154 } 5155 5156 VN_RELE(vp); 5157 VN_RELE(ZTOV(sharedir)); 5158 5159 ZFS_EXIT(zfsvfs); 5160 5161 return (error); 5162#else /* !sun */ 5163 return (EOPNOTSUPP); 5164#endif /* !sun */ 5165} 5166 5167/* 5168 * innvl: { 5169 * "holds" -> { snapname -> holdname (string), ... } 5170 * (optional) "cleanup_fd" -> fd (int32) 5171 * } 5172 * 5173 * outnvl: { 5174 * snapname -> error value (int32) 5175 * ... 5176 * } 5177 */ 5178/* ARGSUSED */ 5179static int 5180zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist) 5181{ 5182 nvlist_t *holds; 5183 int cleanup_fd = -1; 5184 int error; 5185 minor_t minor = 0; 5186 5187 error = nvlist_lookup_nvlist(args, "holds", &holds); 5188 if (error != 0) 5189 return (SET_ERROR(EINVAL)); 5190 5191 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) { 5192 error = zfs_onexit_fd_hold(cleanup_fd, &minor); 5193 if (error != 0) 5194 return (error); 5195 } 5196 5197 error = dsl_dataset_user_hold(holds, minor, errlist); 5198 if (minor != 0) 5199 zfs_onexit_fd_rele(cleanup_fd); 5200 return (error); 5201} 5202 5203/* 5204 * innvl is not used. 5205 * 5206 * outnvl: { 5207 * holdname -> time added (uint64 seconds since epoch) 5208 * ... 5209 * } 5210 */ 5211/* ARGSUSED */ 5212static int 5213zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl) 5214{ 5215 return (dsl_dataset_get_holds(snapname, outnvl)); 5216} 5217 5218/* 5219 * innvl: { 5220 * snapname -> { holdname, ... } 5221 * ... 5222 * } 5223 * 5224 * outnvl: { 5225 * snapname -> error value (int32) 5226 * ... 5227 * } 5228 */ 5229/* ARGSUSED */ 5230static int 5231zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist) 5232{ 5233 return (dsl_dataset_user_release(holds, errlist)); 5234} 5235 5236/* 5237 * inputs: 5238 * zc_name name of new filesystem or snapshot 5239 * zc_value full name of old snapshot 5240 * 5241 * outputs: 5242 * zc_cookie space in bytes 5243 * zc_objset_type compressed space in bytes 5244 * zc_perm_action uncompressed space in bytes 5245 */ 5246static int 5247zfs_ioc_space_written(zfs_cmd_t *zc) 5248{ 5249 int error; 5250 dsl_pool_t *dp; 5251 dsl_dataset_t *new, *old; 5252 5253 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 5254 if (error != 0) 5255 return (error); 5256 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new); 5257 if (error != 0) { 5258 dsl_pool_rele(dp, FTAG); 5259 return (error); 5260 } 5261 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old); 5262 if (error != 0) { 5263 dsl_dataset_rele(new, FTAG); 5264 dsl_pool_rele(dp, FTAG); 5265 return (error); 5266 } 5267 5268 error = dsl_dataset_space_written(old, new, &zc->zc_cookie, 5269 &zc->zc_objset_type, &zc->zc_perm_action); 5270 dsl_dataset_rele(old, FTAG); 5271 dsl_dataset_rele(new, FTAG); 5272 dsl_pool_rele(dp, FTAG); 5273 return (error); 5274} 5275 5276/* 5277 * innvl: { 5278 * "firstsnap" -> snapshot name 5279 * } 5280 * 5281 * outnvl: { 5282 * "used" -> space in bytes 5283 * "compressed" -> compressed space in bytes 5284 * "uncompressed" -> uncompressed space in bytes 5285 * } 5286 */ 5287static int 5288zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl) 5289{ 5290 int error; 5291 dsl_pool_t *dp; 5292 dsl_dataset_t *new, *old; 5293 char *firstsnap; 5294 uint64_t used, comp, uncomp; 5295 5296 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0) 5297 return (SET_ERROR(EINVAL)); 5298 5299 error = dsl_pool_hold(lastsnap, FTAG, &dp); 5300 if (error != 0) 5301 return (error); 5302 5303 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new); 5304 if (error != 0) { 5305 dsl_pool_rele(dp, FTAG); 5306 return (error); 5307 } 5308 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old); 5309 if (error != 0) { 5310 dsl_dataset_rele(new, FTAG); 5311 dsl_pool_rele(dp, FTAG); 5312 return (error); 5313 } 5314 5315 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp); 5316 dsl_dataset_rele(old, FTAG); 5317 dsl_dataset_rele(new, FTAG); 5318 dsl_pool_rele(dp, FTAG); 5319 fnvlist_add_uint64(outnvl, "used", used); 5320 fnvlist_add_uint64(outnvl, "compressed", comp); 5321 fnvlist_add_uint64(outnvl, "uncompressed", uncomp); 5322 return (error); 5323} 5324 5325static int 5326zfs_ioc_jail(zfs_cmd_t *zc) 5327{ 5328 5329 return (zone_dataset_attach(curthread->td_ucred, zc->zc_name, 5330 (int)zc->zc_jailid)); 5331} 5332 5333static int 5334zfs_ioc_unjail(zfs_cmd_t *zc) 5335{ 5336 5337 return (zone_dataset_detach(curthread->td_ucred, zc->zc_name, 5338 (int)zc->zc_jailid)); 5339} 5340 5341/* 5342 * innvl: { 5343 * "fd" -> file descriptor to write stream to (int32) 5344 * (optional) "fromsnap" -> full snap name to send an incremental from 5345 * (optional) "embedok" -> (value ignored) 5346 * presence indicates DRR_WRITE_EMBEDDED records are permitted 5347 * } 5348 * 5349 * outnvl is unused 5350 */ 5351/* ARGSUSED */ 5352static int 5353zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl) 5354{ 5355 cap_rights_t rights; 5356 int error; 5357 offset_t off; 5358 char *fromname = NULL; 5359 int fd; 5360 boolean_t embedok; 5361 5362 error = nvlist_lookup_int32(innvl, "fd", &fd); 5363 if (error != 0) 5364 return (SET_ERROR(EINVAL)); 5365 5366 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname); 5367 5368 embedok = nvlist_exists(innvl, "embedok"); 5369 5370 file_t *fp = getf(fd, cap_rights_init(&rights, CAP_READ)); 5371 if (fp == NULL) 5372 return (SET_ERROR(EBADF)); 5373 5374 off = fp->f_offset; 5375#ifdef illumos 5376 error = dmu_send(snapname, fromname, embedok, fd, fp->f_vnode, &off); 5377#else 5378 error = dmu_send(snapname, fromname, embedok, fd, fp, &off); 5379#endif 5380 5381#ifdef illumos 5382 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0) 5383 fp->f_offset = off; 5384#else 5385 fp->f_offset = off; 5386#endif 5387 5388 releasef(fd); 5389 return (error); 5390} 5391 5392/* 5393 * Determine approximately how large a zfs send stream will be -- the number 5394 * of bytes that will be written to the fd supplied to zfs_ioc_send_new(). 5395 * 5396 * innvl: { 5397 * (optional) "fromsnap" -> full snap name to send an incremental from 5398 * } 5399 * 5400 * outnvl: { 5401 * "space" -> bytes of space (uint64) 5402 * } 5403 */ 5404static int 5405zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl) 5406{ 5407 dsl_pool_t *dp; 5408 dsl_dataset_t *fromsnap = NULL; 5409 dsl_dataset_t *tosnap; 5410 int error; 5411 char *fromname; 5412 uint64_t space; 5413 5414 error = dsl_pool_hold(snapname, FTAG, &dp); 5415 if (error != 0) 5416 return (error); 5417 5418 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap); 5419 if (error != 0) { 5420 dsl_pool_rele(dp, FTAG); 5421 return (error); 5422 } 5423 5424 error = nvlist_lookup_string(innvl, "fromsnap", &fromname); 5425 if (error == 0) { 5426 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap); 5427 if (error != 0) { 5428 dsl_dataset_rele(tosnap, FTAG); 5429 dsl_pool_rele(dp, FTAG); 5430 return (error); 5431 } 5432 } 5433 5434 error = dmu_send_estimate(tosnap, fromsnap, &space); 5435 fnvlist_add_uint64(outnvl, "space", space); 5436 5437 if (fromsnap != NULL) 5438 dsl_dataset_rele(fromsnap, FTAG); 5439 dsl_dataset_rele(tosnap, FTAG); 5440 dsl_pool_rele(dp, FTAG); 5441 return (error); 5442} 5443 5444 5445static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST]; 5446 5447static void 5448zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5449 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck, 5450 boolean_t log_history, zfs_ioc_poolcheck_t pool_check) 5451{ 5452 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST]; 5453 5454 ASSERT3U(ioc, >=, ZFS_IOC_FIRST); 5455 ASSERT3U(ioc, <, ZFS_IOC_LAST); 5456 ASSERT3P(vec->zvec_legacy_func, ==, NULL); 5457 ASSERT3P(vec->zvec_func, ==, NULL); 5458 5459 vec->zvec_legacy_func = func; 5460 vec->zvec_secpolicy = secpolicy; 5461 vec->zvec_namecheck = namecheck; 5462 vec->zvec_allow_log = log_history; 5463 vec->zvec_pool_check = pool_check; 5464} 5465 5466/* 5467 * See the block comment at the beginning of this file for details on 5468 * each argument to this function. 5469 */ 5470static void 5471zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func, 5472 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck, 5473 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist, 5474 boolean_t allow_log) 5475{ 5476 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST]; 5477 5478 ASSERT3U(ioc, >=, ZFS_IOC_FIRST); 5479 ASSERT3U(ioc, <, ZFS_IOC_LAST); 5480 ASSERT3P(vec->zvec_legacy_func, ==, NULL); 5481 ASSERT3P(vec->zvec_func, ==, NULL); 5482 5483 /* if we are logging, the name must be valid */ 5484 ASSERT(!allow_log || namecheck != NO_NAME); 5485 5486 vec->zvec_name = name; 5487 vec->zvec_func = func; 5488 vec->zvec_secpolicy = secpolicy; 5489 vec->zvec_namecheck = namecheck; 5490 vec->zvec_pool_check = pool_check; 5491 vec->zvec_smush_outnvlist = smush_outnvlist; 5492 vec->zvec_allow_log = allow_log; 5493} 5494 5495static void 5496zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5497 zfs_secpolicy_func_t *secpolicy, boolean_t log_history, 5498 zfs_ioc_poolcheck_t pool_check) 5499{ 5500 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5501 POOL_NAME, log_history, pool_check); 5502} 5503 5504static void 5505zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5506 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check) 5507{ 5508 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5509 DATASET_NAME, B_FALSE, pool_check); 5510} 5511 5512static void 5513zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func) 5514{ 5515 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config, 5516 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5517} 5518 5519static void 5520zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5521 zfs_secpolicy_func_t *secpolicy) 5522{ 5523 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5524 NO_NAME, B_FALSE, POOL_CHECK_NONE); 5525} 5526 5527static void 5528zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc, 5529 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy) 5530{ 5531 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5532 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED); 5533} 5534 5535static void 5536zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func) 5537{ 5538 zfs_ioctl_register_dataset_read_secpolicy(ioc, func, 5539 zfs_secpolicy_read); 5540} 5541 5542static void 5543zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5544 zfs_secpolicy_func_t *secpolicy) 5545{ 5546 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5547 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5548} 5549 5550static void 5551zfs_ioctl_init(void) 5552{ 5553 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT, 5554 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME, 5555 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5556 5557 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY, 5558 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME, 5559 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE); 5560 5561 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS, 5562 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME, 5563 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5564 5565 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW, 5566 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME, 5567 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5568 5569 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE, 5570 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME, 5571 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5572 5573 zfs_ioctl_register("create", ZFS_IOC_CREATE, 5574 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME, 5575 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5576 5577 zfs_ioctl_register("clone", ZFS_IOC_CLONE, 5578 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME, 5579 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5580 5581 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS, 5582 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME, 5583 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5584 5585 zfs_ioctl_register("hold", ZFS_IOC_HOLD, 5586 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME, 5587 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5588 zfs_ioctl_register("release", ZFS_IOC_RELEASE, 5589 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME, 5590 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5591 5592 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS, 5593 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME, 5594 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5595 5596 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK, 5597 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, 5598 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE); 5599 5600 zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK, 5601 zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME, 5602 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5603 5604 zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS, 5605 zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME, 5606 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5607 5608 zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS, 5609 zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks, 5610 POOL_NAME, 5611 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5612 5613 /* IOCTLS that use the legacy function signature */ 5614 5615 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze, 5616 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY); 5617 5618 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create, 5619 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5620 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN, 5621 zfs_ioc_pool_scan); 5622 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE, 5623 zfs_ioc_pool_upgrade); 5624 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD, 5625 zfs_ioc_vdev_add); 5626 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE, 5627 zfs_ioc_vdev_remove); 5628 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE, 5629 zfs_ioc_vdev_set_state); 5630 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH, 5631 zfs_ioc_vdev_attach); 5632 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH, 5633 zfs_ioc_vdev_detach); 5634 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH, 5635 zfs_ioc_vdev_setpath); 5636 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU, 5637 zfs_ioc_vdev_setfru); 5638 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS, 5639 zfs_ioc_pool_set_props); 5640 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT, 5641 zfs_ioc_vdev_split); 5642 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID, 5643 zfs_ioc_pool_reguid); 5644 5645 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS, 5646 zfs_ioc_pool_configs, zfs_secpolicy_none); 5647 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT, 5648 zfs_ioc_pool_tryimport, zfs_secpolicy_config); 5649 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT, 5650 zfs_ioc_inject_fault, zfs_secpolicy_inject); 5651 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT, 5652 zfs_ioc_clear_fault, zfs_secpolicy_inject); 5653 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT, 5654 zfs_ioc_inject_list_next, zfs_secpolicy_inject); 5655 5656 /* 5657 * pool destroy, and export don't log the history as part of 5658 * zfsdev_ioctl, but rather zfs_ioc_pool_export 5659 * does the logging of those commands. 5660 */ 5661 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy, 5662 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE); 5663 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export, 5664 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE); 5665 5666 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats, 5667 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE); 5668 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props, 5669 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE); 5670 5671 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log, 5672 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_NONE); 5673 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME, 5674 zfs_ioc_dsobj_to_dsname, 5675 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_NONE); 5676 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY, 5677 zfs_ioc_pool_get_history, 5678 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED); 5679 5680 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import, 5681 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5682 5683 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear, 5684 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5685 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen, 5686 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED); 5687 5688 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN, 5689 zfs_ioc_space_written); 5690 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS, 5691 zfs_ioc_objset_recvd_props); 5692 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ, 5693 zfs_ioc_next_obj); 5694 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL, 5695 zfs_ioc_get_fsacl); 5696 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS, 5697 zfs_ioc_objset_stats); 5698 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS, 5699 zfs_ioc_objset_zplprops); 5700 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT, 5701 zfs_ioc_dataset_list_next); 5702 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT, 5703 zfs_ioc_snapshot_list_next); 5704 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS, 5705 zfs_ioc_send_progress); 5706 5707 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF, 5708 zfs_ioc_diff, zfs_secpolicy_diff); 5709 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS, 5710 zfs_ioc_obj_to_stats, zfs_secpolicy_diff); 5711 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH, 5712 zfs_ioc_obj_to_path, zfs_secpolicy_diff); 5713 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE, 5714 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one); 5715 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY, 5716 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many); 5717 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND, 5718 zfs_ioc_send, zfs_secpolicy_send); 5719 5720 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop, 5721 zfs_secpolicy_none); 5722 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy, 5723 zfs_secpolicy_destroy); 5724 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename, 5725 zfs_secpolicy_rename); 5726 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv, 5727 zfs_secpolicy_recv); 5728 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote, 5729 zfs_secpolicy_promote); 5730 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP, 5731 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop); 5732 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl, 5733 zfs_secpolicy_set_fsacl); 5734 5735 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share, 5736 zfs_secpolicy_share, POOL_CHECK_NONE); 5737 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl, 5738 zfs_secpolicy_smb_acl, POOL_CHECK_NONE); 5739 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE, 5740 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade, 5741 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5742 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT, 5743 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot, 5744 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5745 5746#ifdef __FreeBSD__ 5747 zfs_ioctl_register_dataset_nolog(ZFS_IOC_JAIL, zfs_ioc_jail, 5748 zfs_secpolicy_config, POOL_CHECK_NONE); 5749 zfs_ioctl_register_dataset_nolog(ZFS_IOC_UNJAIL, zfs_ioc_unjail, 5750 zfs_secpolicy_config, POOL_CHECK_NONE); 5751#endif 5752} 5753 5754int 5755pool_status_check(const char *name, zfs_ioc_namecheck_t type, 5756 zfs_ioc_poolcheck_t check) 5757{ 5758 spa_t *spa; 5759 int error; 5760 5761 ASSERT(type == POOL_NAME || type == DATASET_NAME); 5762 5763 if (check & POOL_CHECK_NONE) 5764 return (0); 5765 5766 error = spa_open(name, &spa, FTAG); 5767 if (error == 0) { 5768 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa)) 5769 error = SET_ERROR(EAGAIN); 5770 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa)) 5771 error = SET_ERROR(EROFS); 5772 spa_close(spa, FTAG); 5773 } 5774 return (error); 5775} 5776 5777/* 5778 * Find a free minor number. 5779 */ 5780minor_t 5781zfsdev_minor_alloc(void) 5782{ 5783 static minor_t last_minor; 5784 minor_t m; 5785 5786 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 5787 5788 for (m = last_minor + 1; m != last_minor; m++) { 5789 if (m > ZFSDEV_MAX_MINOR) 5790 m = 1; 5791 if (ddi_get_soft_state(zfsdev_state, m) == NULL) { 5792 last_minor = m; 5793 return (m); 5794 } 5795 } 5796 5797 return (0); 5798} 5799 5800static int 5801zfs_ctldev_init(struct cdev *devp) 5802{ 5803 minor_t minor; 5804 zfs_soft_state_t *zs; 5805 5806 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 5807 5808 minor = zfsdev_minor_alloc(); 5809 if (minor == 0) 5810 return (SET_ERROR(ENXIO)); 5811 5812 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS) 5813 return (SET_ERROR(EAGAIN)); 5814 5815 devfs_set_cdevpriv((void *)(uintptr_t)minor, zfsdev_close); 5816 5817 zs = ddi_get_soft_state(zfsdev_state, minor); 5818 zs->zss_type = ZSST_CTLDEV; 5819 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data); 5820 5821 return (0); 5822} 5823 5824static void 5825zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor) 5826{ 5827 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 5828 5829 zfs_onexit_destroy(zo); 5830 ddi_soft_state_free(zfsdev_state, minor); 5831} 5832 5833void * 5834zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which) 5835{ 5836 zfs_soft_state_t *zp; 5837 5838 zp = ddi_get_soft_state(zfsdev_state, minor); 5839 if (zp == NULL || zp->zss_type != which) 5840 return (NULL); 5841 5842 return (zp->zss_data); 5843} 5844 5845static int 5846zfsdev_open(struct cdev *devp, int flag, int mode, struct thread *td) 5847{ 5848 int error = 0; 5849 5850#ifdef sun 5851 if (getminor(*devp) != 0) 5852 return (zvol_open(devp, flag, otyp, cr)); 5853#endif 5854 5855 /* This is the control device. Allocate a new minor if requested. */ 5856 if (flag & FEXCL) { 5857 mutex_enter(&spa_namespace_lock); 5858 error = zfs_ctldev_init(devp); 5859 mutex_exit(&spa_namespace_lock); 5860 } 5861 5862 return (error); 5863} 5864 5865static void 5866zfsdev_close(void *data) 5867{ 5868 zfs_onexit_t *zo; 5869 minor_t minor = (minor_t)(uintptr_t)data; 5870 5871 if (minor == 0) 5872 return; 5873 5874 mutex_enter(&spa_namespace_lock); 5875 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV); 5876 if (zo == NULL) { 5877 mutex_exit(&spa_namespace_lock); 5878 return; 5879 } 5880 zfs_ctldev_destroy(zo, minor); 5881 mutex_exit(&spa_namespace_lock); 5882} 5883 5884static int 5885zfsdev_ioctl(struct cdev *dev, u_long zcmd, caddr_t arg, int flag, 5886 struct thread *td) 5887{ 5888 zfs_cmd_t *zc; 5889 uint_t vecnum; 5890 int error, rc, len; 5891#ifdef illumos 5892 minor_t minor = getminor(dev); 5893#else 5894 zfs_iocparm_t *zc_iocparm; 5895 int cflag, cmd, oldvecnum; 5896 boolean_t newioc, compat; 5897 void *compat_zc = NULL; 5898 cred_t *cr = td->td_ucred; 5899#endif 5900 const zfs_ioc_vec_t *vec; 5901 char *saved_poolname = NULL; 5902 nvlist_t *innvl = NULL; 5903 5904 cflag = ZFS_CMD_COMPAT_NONE; 5905 compat = B_FALSE; 5906 newioc = B_TRUE; /* "new" style (zfs_iocparm_t) ioctl */ 5907 5908 len = IOCPARM_LEN(zcmd); 5909 vecnum = cmd = zcmd & 0xff; 5910 5911 /* 5912 * Check if we are talking to supported older binaries 5913 * and translate zfs_cmd if necessary 5914 */ 5915 if (len != sizeof(zfs_iocparm_t)) { 5916 newioc = B_FALSE; 5917 compat = B_TRUE; 5918 5919 vecnum = cmd; 5920 5921 switch (len) { 5922 case sizeof(zfs_cmd_zcmd_t): 5923 cflag = ZFS_CMD_COMPAT_LZC; 5924 break; 5925 case sizeof(zfs_cmd_deadman_t): 5926 cflag = ZFS_CMD_COMPAT_DEADMAN; 5927 break; 5928 case sizeof(zfs_cmd_v28_t): 5929 cflag = ZFS_CMD_COMPAT_V28; 5930 break; 5931 case sizeof(zfs_cmd_v15_t): 5932 cflag = ZFS_CMD_COMPAT_V15; 5933 vecnum = zfs_ioctl_v15_to_v28[cmd]; 5934 5935 /* 5936 * Return without further handling 5937 * if the command is blacklisted. 5938 */ 5939 if (vecnum == ZFS_IOC_COMPAT_PASS) 5940 return (0); 5941 else if (vecnum == ZFS_IOC_COMPAT_FAIL) 5942 return (ENOTSUP); 5943 break; 5944 default: 5945 return (EINVAL); 5946 } 5947 } 5948 5949#ifdef illumos 5950 vecnum = cmd - ZFS_IOC_FIRST; 5951 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip)); 5952#endif 5953 5954 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0])) 5955 return (SET_ERROR(EINVAL)); 5956 vec = &zfs_ioc_vec[vecnum]; 5957 5958 zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP); 5959 5960#ifdef illumos 5961 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag); 5962 if (error != 0) { 5963 error = SET_ERROR(EFAULT); 5964 goto out; 5965 } 5966#else /* !illumos */ 5967 bzero(zc, sizeof(zfs_cmd_t)); 5968 5969 if (newioc) { 5970 zc_iocparm = (void *)arg; 5971 5972 switch (zc_iocparm->zfs_ioctl_version) { 5973 case ZFS_IOCVER_CURRENT: 5974 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_t)) { 5975 error = SET_ERROR(EINVAL); 5976 goto out; 5977 } 5978 break; 5979 case ZFS_IOCVER_ZCMD: 5980 if (zc_iocparm->zfs_cmd_size > sizeof(zfs_cmd_t) || 5981 zc_iocparm->zfs_cmd_size < sizeof(zfs_cmd_zcmd_t)) { 5982 error = SET_ERROR(EFAULT); 5983 goto out; 5984 } 5985 compat = B_TRUE; 5986 cflag = ZFS_CMD_COMPAT_ZCMD; 5987 break; 5988 default: 5989 error = SET_ERROR(EINVAL); 5990 goto out; 5991 /* NOTREACHED */ 5992 } 5993 5994 if (compat) { 5995 ASSERT(sizeof(zfs_cmd_t) >= zc_iocparm->zfs_cmd_size); 5996 compat_zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP); 5997 bzero(compat_zc, sizeof(zfs_cmd_t)); 5998 5999 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd, 6000 compat_zc, zc_iocparm->zfs_cmd_size, flag); 6001 if (error != 0) { 6002 error = SET_ERROR(EFAULT); 6003 goto out; 6004 } 6005 } else { 6006 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd, 6007 zc, zc_iocparm->zfs_cmd_size, flag); 6008 if (error != 0) { 6009 error = SET_ERROR(EFAULT); 6010 goto out; 6011 } 6012 } 6013 } 6014 6015 if (compat) { 6016 if (newioc) { 6017 ASSERT(compat_zc != NULL); 6018 zfs_cmd_compat_get(zc, compat_zc, cflag); 6019 } else { 6020 ASSERT(compat_zc == NULL); 6021 zfs_cmd_compat_get(zc, arg, cflag); 6022 } 6023 oldvecnum = vecnum; 6024 error = zfs_ioctl_compat_pre(zc, &vecnum, cflag); 6025 if (error != 0) 6026 goto out; 6027 if (oldvecnum != vecnum) 6028 vec = &zfs_ioc_vec[vecnum]; 6029 } 6030#endif /* !illumos */ 6031 6032 zc->zc_iflags = flag & FKIOCTL; 6033 if (zc->zc_nvlist_src_size != 0) { 6034 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 6035 zc->zc_iflags, &innvl); 6036 if (error != 0) 6037 goto out; 6038 } 6039 6040 /* rewrite innvl for backwards compatibility */ 6041 if (compat) 6042 innvl = zfs_ioctl_compat_innvl(zc, innvl, vecnum, cflag); 6043 6044 /* 6045 * Ensure that all pool/dataset names are valid before we pass down to 6046 * the lower layers. 6047 */ 6048 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0'; 6049 switch (vec->zvec_namecheck) { 6050 case POOL_NAME: 6051 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0) 6052 error = SET_ERROR(EINVAL); 6053 else 6054 error = pool_status_check(zc->zc_name, 6055 vec->zvec_namecheck, vec->zvec_pool_check); 6056 break; 6057 6058 case DATASET_NAME: 6059 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0) 6060 error = SET_ERROR(EINVAL); 6061 else 6062 error = pool_status_check(zc->zc_name, 6063 vec->zvec_namecheck, vec->zvec_pool_check); 6064 break; 6065 6066 case NO_NAME: 6067 break; 6068 } 6069 6070 if (error == 0 && !(flag & FKIOCTL)) 6071 error = vec->zvec_secpolicy(zc, innvl, cr); 6072 6073 if (error != 0) 6074 goto out; 6075 6076 /* legacy ioctls can modify zc_name */ 6077 len = strcspn(zc->zc_name, "/@#") + 1; 6078 saved_poolname = kmem_alloc(len, KM_SLEEP); 6079 (void) strlcpy(saved_poolname, zc->zc_name, len); 6080 6081 if (vec->zvec_func != NULL) { 6082 nvlist_t *outnvl; 6083 int puterror = 0; 6084 spa_t *spa; 6085 nvlist_t *lognv = NULL; 6086 6087 ASSERT(vec->zvec_legacy_func == NULL); 6088 6089 /* 6090 * Add the innvl to the lognv before calling the func, 6091 * in case the func changes the innvl. 6092 */ 6093 if (vec->zvec_allow_log) { 6094 lognv = fnvlist_alloc(); 6095 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL, 6096 vec->zvec_name); 6097 if (!nvlist_empty(innvl)) { 6098 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL, 6099 innvl); 6100 } 6101 } 6102 6103 outnvl = fnvlist_alloc(); 6104 error = vec->zvec_func(zc->zc_name, innvl, outnvl); 6105 6106 if (error == 0 && vec->zvec_allow_log && 6107 spa_open(zc->zc_name, &spa, FTAG) == 0) { 6108 if (!nvlist_empty(outnvl)) { 6109 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL, 6110 outnvl); 6111 } 6112 (void) spa_history_log_nvl(spa, lognv); 6113 spa_close(spa, FTAG); 6114 } 6115 fnvlist_free(lognv); 6116 6117 /* rewrite outnvl for backwards compatibility */ 6118 if (compat) 6119 outnvl = zfs_ioctl_compat_outnvl(zc, outnvl, vecnum, 6120 cflag); 6121 6122 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) { 6123 int smusherror = 0; 6124 if (vec->zvec_smush_outnvlist) { 6125 smusherror = nvlist_smush(outnvl, 6126 zc->zc_nvlist_dst_size); 6127 } 6128 if (smusherror == 0) 6129 puterror = put_nvlist(zc, outnvl); 6130 } 6131 6132 if (puterror != 0) 6133 error = puterror; 6134 6135 nvlist_free(outnvl); 6136 } else { 6137 error = vec->zvec_legacy_func(zc); 6138 } 6139 6140out: 6141 nvlist_free(innvl); 6142 6143#ifdef illumos 6144 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag); 6145 if (error == 0 && rc != 0) 6146 error = SET_ERROR(EFAULT); 6147#else 6148 if (compat) { 6149 zfs_ioctl_compat_post(zc, cmd, cflag); 6150 if (newioc) { 6151 ASSERT(compat_zc != NULL); 6152 ASSERT(sizeof(zfs_cmd_t) >= zc_iocparm->zfs_cmd_size); 6153 6154 zfs_cmd_compat_put(zc, compat_zc, vecnum, cflag); 6155 rc = ddi_copyout(compat_zc, 6156 (void *)(uintptr_t)zc_iocparm->zfs_cmd, 6157 zc_iocparm->zfs_cmd_size, flag); 6158 if (error == 0 && rc != 0) 6159 error = SET_ERROR(EFAULT); 6160 kmem_free(compat_zc, sizeof (zfs_cmd_t)); 6161 } else { 6162 zfs_cmd_compat_put(zc, arg, vecnum, cflag); 6163 } 6164 } else { 6165 ASSERT(newioc); 6166 6167 rc = ddi_copyout(zc, (void *)(uintptr_t)zc_iocparm->zfs_cmd, 6168 sizeof (zfs_cmd_t), flag); 6169 if (error == 0 && rc != 0) 6170 error = SET_ERROR(EFAULT); 6171 } 6172#endif 6173 if (error == 0 && vec->zvec_allow_log) { 6174 char *s = tsd_get(zfs_allow_log_key); 6175 if (s != NULL) 6176 strfree(s); 6177 (void) tsd_set(zfs_allow_log_key, saved_poolname); 6178 } else { 6179 if (saved_poolname != NULL) 6180 strfree(saved_poolname); 6181 } 6182 6183 kmem_free(zc, sizeof (zfs_cmd_t)); 6184 return (error); 6185} 6186 6187#ifdef sun 6188static int 6189zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd) 6190{ 6191 if (cmd != DDI_ATTACH) 6192 return (DDI_FAILURE); 6193 6194 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0, 6195 DDI_PSEUDO, 0) == DDI_FAILURE) 6196 return (DDI_FAILURE); 6197 6198 zfs_dip = dip; 6199 6200 ddi_report_dev(dip); 6201 6202 return (DDI_SUCCESS); 6203} 6204 6205static int 6206zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd) 6207{ 6208 if (spa_busy() || zfs_busy() || zvol_busy()) 6209 return (DDI_FAILURE); 6210 6211 if (cmd != DDI_DETACH) 6212 return (DDI_FAILURE); 6213 6214 zfs_dip = NULL; 6215 6216 ddi_prop_remove_all(dip); 6217 ddi_remove_minor_node(dip, NULL); 6218 6219 return (DDI_SUCCESS); 6220} 6221 6222/*ARGSUSED*/ 6223static int 6224zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result) 6225{ 6226 switch (infocmd) { 6227 case DDI_INFO_DEVT2DEVINFO: 6228 *result = zfs_dip; 6229 return (DDI_SUCCESS); 6230 6231 case DDI_INFO_DEVT2INSTANCE: 6232 *result = (void *)0; 6233 return (DDI_SUCCESS); 6234 } 6235 6236 return (DDI_FAILURE); 6237} 6238#endif /* sun */ 6239 6240/* 6241 * OK, so this is a little weird. 6242 * 6243 * /dev/zfs is the control node, i.e. minor 0. 6244 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0. 6245 * 6246 * /dev/zfs has basically nothing to do except serve up ioctls, 6247 * so most of the standard driver entry points are in zvol.c. 6248 */ 6249#ifdef sun 6250static struct cb_ops zfs_cb_ops = { 6251 zfsdev_open, /* open */ 6252 zfsdev_close, /* close */ 6253 zvol_strategy, /* strategy */ 6254 nodev, /* print */ 6255 zvol_dump, /* dump */ 6256 zvol_read, /* read */ 6257 zvol_write, /* write */ 6258 zfsdev_ioctl, /* ioctl */ 6259 nodev, /* devmap */ 6260 nodev, /* mmap */ 6261 nodev, /* segmap */ 6262 nochpoll, /* poll */ 6263 ddi_prop_op, /* prop_op */ 6264 NULL, /* streamtab */ 6265 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */ 6266 CB_REV, /* version */ 6267 nodev, /* async read */ 6268 nodev, /* async write */ 6269}; 6270 6271static struct dev_ops zfs_dev_ops = { 6272 DEVO_REV, /* version */ 6273 0, /* refcnt */ 6274 zfs_info, /* info */ 6275 nulldev, /* identify */ 6276 nulldev, /* probe */ 6277 zfs_attach, /* attach */ 6278 zfs_detach, /* detach */ 6279 nodev, /* reset */ 6280 &zfs_cb_ops, /* driver operations */ 6281 NULL, /* no bus operations */ 6282 NULL, /* power */ 6283 ddi_quiesce_not_needed, /* quiesce */ 6284}; 6285 6286static struct modldrv zfs_modldrv = { 6287 &mod_driverops, 6288 "ZFS storage pool", 6289 &zfs_dev_ops 6290}; 6291 6292static struct modlinkage modlinkage = { 6293 MODREV_1, 6294 (void *)&zfs_modlfs, 6295 (void *)&zfs_modldrv, 6296 NULL 6297}; 6298#endif /* sun */ 6299 6300static struct cdevsw zfs_cdevsw = { 6301 .d_version = D_VERSION, 6302 .d_open = zfsdev_open, 6303 .d_ioctl = zfsdev_ioctl, 6304 .d_name = ZFS_DEV_NAME 6305}; 6306 6307static void 6308zfs_allow_log_destroy(void *arg) 6309{ 6310 char *poolname = arg; 6311 strfree(poolname); 6312} 6313 6314static void 6315zfsdev_init(void) 6316{ 6317 zfsdev = make_dev(&zfs_cdevsw, 0x0, UID_ROOT, GID_OPERATOR, 0666, 6318 ZFS_DEV_NAME); 6319} 6320 6321static void 6322zfsdev_fini(void) 6323{ 6324 if (zfsdev != NULL) 6325 destroy_dev(zfsdev); 6326} 6327 6328static struct root_hold_token *zfs_root_token; 6329struct proc *zfsproc; 6330 6331#ifdef sun 6332int 6333_init(void) 6334{ 6335 int error; 6336 6337 spa_init(FREAD | FWRITE); 6338 zfs_init(); 6339 zvol_init(); 6340 zfs_ioctl_init(); 6341 6342 if ((error = mod_install(&modlinkage)) != 0) { 6343 zvol_fini(); 6344 zfs_fini(); 6345 spa_fini(); 6346 return (error); 6347 } 6348 6349 tsd_create(&zfs_fsyncer_key, NULL); 6350 tsd_create(&rrw_tsd_key, rrw_tsd_destroy); 6351 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy); 6352 6353 error = ldi_ident_from_mod(&modlinkage, &zfs_li); 6354 ASSERT(error == 0); 6355 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL); 6356 6357 return (0); 6358} 6359 6360int 6361_fini(void) 6362{ 6363 int error; 6364 6365 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled) 6366 return (SET_ERROR(EBUSY)); 6367 6368 if ((error = mod_remove(&modlinkage)) != 0) 6369 return (error); 6370 6371 zvol_fini(); 6372 zfs_fini(); 6373 spa_fini(); 6374 if (zfs_nfsshare_inited) 6375 (void) ddi_modclose(nfs_mod); 6376 if (zfs_smbshare_inited) 6377 (void) ddi_modclose(smbsrv_mod); 6378 if (zfs_nfsshare_inited || zfs_smbshare_inited) 6379 (void) ddi_modclose(sharefs_mod); 6380 6381 tsd_destroy(&zfs_fsyncer_key); 6382 ldi_ident_release(zfs_li); 6383 zfs_li = NULL; 6384 mutex_destroy(&zfs_share_lock); 6385 6386 return (error); 6387} 6388 6389int 6390_info(struct modinfo *modinfop) 6391{ 6392 return (mod_info(&modlinkage, modinfop)); 6393} 6394#endif /* sun */ 6395 6396static int zfs__init(void); 6397static int zfs__fini(void); 6398static void zfs_shutdown(void *, int); 6399 6400static eventhandler_tag zfs_shutdown_event_tag; 6401 6402int 6403zfs__init(void) 6404{ 6405 6406 zfs_root_token = root_mount_hold("ZFS"); 6407 6408 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL); 6409 6410 spa_init(FREAD | FWRITE); 6411 zfs_init(); 6412 zvol_init(); 6413 zfs_ioctl_init(); 6414 6415 tsd_create(&zfs_fsyncer_key, NULL); 6416 tsd_create(&rrw_tsd_key, rrw_tsd_destroy); 6417 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy); 6418 6419 printf("ZFS storage pool version: features support (" SPA_VERSION_STRING ")\n"); 6420 root_mount_rel(zfs_root_token); 6421 6422 zfsdev_init(); 6423 6424 return (0); 6425} 6426 6427int 6428zfs__fini(void) 6429{ 6430 if (spa_busy() || zfs_busy() || zvol_busy() || 6431 zio_injection_enabled) { 6432 return (EBUSY); 6433 } 6434 6435 zfsdev_fini(); 6436 zvol_fini(); 6437 zfs_fini(); 6438 spa_fini(); 6439 6440 tsd_destroy(&zfs_fsyncer_key); 6441 tsd_destroy(&rrw_tsd_key); 6442 tsd_destroy(&zfs_allow_log_key); 6443 6444 mutex_destroy(&zfs_share_lock); 6445 6446 return (0); 6447} 6448 6449static void 6450zfs_shutdown(void *arg __unused, int howto __unused) 6451{ 6452 6453 /* 6454 * ZFS fini routines can not properly work in a panic-ed system. 6455 */ 6456 if (panicstr == NULL) 6457 (void)zfs__fini(); 6458} 6459 6460 6461static int 6462zfs_modevent(module_t mod, int type, void *unused __unused) 6463{ 6464 int err; 6465 6466 switch (type) { 6467 case MOD_LOAD: 6468 err = zfs__init(); 6469 if (err == 0) 6470 zfs_shutdown_event_tag = EVENTHANDLER_REGISTER( 6471 shutdown_post_sync, zfs_shutdown, NULL, 6472 SHUTDOWN_PRI_FIRST); 6473 return (err); 6474 case MOD_UNLOAD: 6475 err = zfs__fini(); 6476 if (err == 0 && zfs_shutdown_event_tag != NULL) 6477 EVENTHANDLER_DEREGISTER(shutdown_post_sync, 6478 zfs_shutdown_event_tag); 6479 return (err); 6480 case MOD_SHUTDOWN: 6481 return (0); 6482 default: 6483 break; 6484 } 6485 return (EOPNOTSUPP); 6486} 6487 6488static moduledata_t zfs_mod = { 6489 "zfsctrl", 6490 zfs_modevent, 6491 0 6492}; 6493DECLARE_MODULE(zfsctrl, zfs_mod, SI_SUB_VFS, SI_ORDER_ANY); 6494MODULE_VERSION(zfsctrl, 1); 6495MODULE_DEPEND(zfsctrl, opensolaris, 1, 1, 1); 6496MODULE_DEPEND(zfsctrl, krpc, 1, 1, 1); 6497MODULE_DEPEND(zfsctrl, acl_nfs4, 1, 1, 1); 6498