zfs_ioctl.c revision 310510
1/* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 22/* 23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 24 * Copyright (c) 2011-2012 Pawel Jakub Dawidek. All rights reserved. 25 * Copyright 2013 Martin Matuska <mm@FreeBSD.org>. All rights reserved. 26 * Copyright 2014 Xin Li <delphij@FreeBSD.org>. All rights reserved. 27 * Copyright 2015, OmniTI Computer Consulting, Inc. All rights reserved. 28 * Copyright 2015 Nexenta Systems, Inc. All rights reserved. 29 * Copyright (c) 2014, 2016 Joyent, Inc. All rights reserved. 30 * Copyright (c) 2011, 2015 by Delphix. All rights reserved. 31 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved. 32 * Copyright (c) 2013 Steven Hartland. All rights reserved. 33 * Copyright (c) 2014 Integros [integros.com] 34 */ 35 36/* 37 * ZFS ioctls. 38 * 39 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage 40 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool. 41 * 42 * There are two ways that we handle ioctls: the legacy way where almost 43 * all of the logic is in the ioctl callback, and the new way where most 44 * of the marshalling is handled in the common entry point, zfsdev_ioctl(). 45 * 46 * Non-legacy ioctls should be registered by calling 47 * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked 48 * from userland by lzc_ioctl(). 49 * 50 * The registration arguments are as follows: 51 * 52 * const char *name 53 * The name of the ioctl. This is used for history logging. If the 54 * ioctl returns successfully (the callback returns 0), and allow_log 55 * is true, then a history log entry will be recorded with the input & 56 * output nvlists. The log entry can be printed with "zpool history -i". 57 * 58 * zfs_ioc_t ioc 59 * The ioctl request number, which userland will pass to ioctl(2). 60 * The ioctl numbers can change from release to release, because 61 * the caller (libzfs) must be matched to the kernel. 62 * 63 * zfs_secpolicy_func_t *secpolicy 64 * This function will be called before the zfs_ioc_func_t, to 65 * determine if this operation is permitted. It should return EPERM 66 * on failure, and 0 on success. Checks include determining if the 67 * dataset is visible in this zone, and if the user has either all 68 * zfs privileges in the zone (SYS_MOUNT), or has been granted permission 69 * to do this operation on this dataset with "zfs allow". 70 * 71 * zfs_ioc_namecheck_t namecheck 72 * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool 73 * name, a dataset name, or nothing. If the name is not well-formed, 74 * the ioctl will fail and the callback will not be called. 75 * Therefore, the callback can assume that the name is well-formed 76 * (e.g. is null-terminated, doesn't have more than one '@' character, 77 * doesn't have invalid characters). 78 * 79 * zfs_ioc_poolcheck_t pool_check 80 * This specifies requirements on the pool state. If the pool does 81 * not meet them (is suspended or is readonly), the ioctl will fail 82 * and the callback will not be called. If any checks are specified 83 * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME. 84 * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED | 85 * POOL_CHECK_READONLY). 86 * 87 * boolean_t smush_outnvlist 88 * If smush_outnvlist is true, then the output is presumed to be a 89 * list of errors, and it will be "smushed" down to fit into the 90 * caller's buffer, by removing some entries and replacing them with a 91 * single "N_MORE_ERRORS" entry indicating how many were removed. See 92 * nvlist_smush() for details. If smush_outnvlist is false, and the 93 * outnvlist does not fit into the userland-provided buffer, then the 94 * ioctl will fail with ENOMEM. 95 * 96 * zfs_ioc_func_t *func 97 * The callback function that will perform the operation. 98 * 99 * The callback should return 0 on success, or an error number on 100 * failure. If the function fails, the userland ioctl will return -1, 101 * and errno will be set to the callback's return value. The callback 102 * will be called with the following arguments: 103 * 104 * const char *name 105 * The name of the pool or dataset to operate on, from 106 * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the 107 * expected type (pool, dataset, or none). 108 * 109 * nvlist_t *innvl 110 * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or 111 * NULL if no input nvlist was provided. Changes to this nvlist are 112 * ignored. If the input nvlist could not be deserialized, the 113 * ioctl will fail and the callback will not be called. 114 * 115 * nvlist_t *outnvl 116 * The output nvlist, initially empty. The callback can fill it in, 117 * and it will be returned to userland by serializing it into 118 * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization 119 * fails (e.g. because the caller didn't supply a large enough 120 * buffer), then the overall ioctl will fail. See the 121 * 'smush_nvlist' argument above for additional behaviors. 122 * 123 * There are two typical uses of the output nvlist: 124 * - To return state, e.g. property values. In this case, 125 * smush_outnvlist should be false. If the buffer was not large 126 * enough, the caller will reallocate a larger buffer and try 127 * the ioctl again. 128 * 129 * - To return multiple errors from an ioctl which makes on-disk 130 * changes. In this case, smush_outnvlist should be true. 131 * Ioctls which make on-disk modifications should generally not 132 * use the outnvl if they succeed, because the caller can not 133 * distinguish between the operation failing, and 134 * deserialization failing. 135 */ 136#ifdef __FreeBSD__ 137#include "opt_kstack_pages.h" 138#endif 139 140#include <sys/types.h> 141#include <sys/param.h> 142#include <sys/systm.h> 143#include <sys/conf.h> 144#include <sys/kernel.h> 145#include <sys/lock.h> 146#include <sys/malloc.h> 147#include <sys/mutex.h> 148#include <sys/proc.h> 149#include <sys/errno.h> 150#include <sys/uio.h> 151#include <sys/buf.h> 152#include <sys/file.h> 153#include <sys/kmem.h> 154#include <sys/conf.h> 155#include <sys/cmn_err.h> 156#include <sys/stat.h> 157#include <sys/zfs_ioctl.h> 158#include <sys/zfs_vfsops.h> 159#include <sys/zfs_znode.h> 160#include <sys/zap.h> 161#include <sys/spa.h> 162#include <sys/spa_impl.h> 163#include <sys/vdev.h> 164#include <sys/dmu.h> 165#include <sys/dsl_dir.h> 166#include <sys/dsl_dataset.h> 167#include <sys/dsl_prop.h> 168#include <sys/dsl_deleg.h> 169#include <sys/dmu_objset.h> 170#include <sys/dmu_impl.h> 171#include <sys/dmu_tx.h> 172#include <sys/sunddi.h> 173#include <sys/policy.h> 174#include <sys/zone.h> 175#include <sys/nvpair.h> 176#include <sys/mount.h> 177#include <sys/taskqueue.h> 178#include <sys/sdt.h> 179#include <sys/varargs.h> 180#include <sys/fs/zfs.h> 181#include <sys/zfs_ctldir.h> 182#include <sys/zfs_dir.h> 183#include <sys/zfs_onexit.h> 184#include <sys/zvol.h> 185#include <sys/dsl_scan.h> 186#include <sys/dmu_objset.h> 187#include <sys/dmu_send.h> 188#include <sys/dsl_destroy.h> 189#include <sys/dsl_bookmark.h> 190#include <sys/dsl_userhold.h> 191#include <sys/zfeature.h> 192#include <sys/zio_checksum.h> 193 194#include "zfs_namecheck.h" 195#include "zfs_prop.h" 196#include "zfs_deleg.h" 197#include "zfs_comutil.h" 198#include "zfs_ioctl_compat.h" 199 200CTASSERT(sizeof(zfs_cmd_t) < IOCPARM_MAX); 201 202static struct cdev *zfsdev; 203 204extern void zfs_init(void); 205extern void zfs_fini(void); 206 207uint_t zfs_fsyncer_key; 208extern uint_t rrw_tsd_key; 209static uint_t zfs_allow_log_key; 210extern uint_t zfs_geom_probe_vdev_key; 211 212typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *); 213typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *); 214typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *); 215 216typedef enum { 217 NO_NAME, 218 POOL_NAME, 219 DATASET_NAME 220} zfs_ioc_namecheck_t; 221 222typedef enum { 223 POOL_CHECK_NONE = 1 << 0, 224 POOL_CHECK_SUSPENDED = 1 << 1, 225 POOL_CHECK_READONLY = 1 << 2, 226} zfs_ioc_poolcheck_t; 227 228typedef struct zfs_ioc_vec { 229 zfs_ioc_legacy_func_t *zvec_legacy_func; 230 zfs_ioc_func_t *zvec_func; 231 zfs_secpolicy_func_t *zvec_secpolicy; 232 zfs_ioc_namecheck_t zvec_namecheck; 233 boolean_t zvec_allow_log; 234 zfs_ioc_poolcheck_t zvec_pool_check; 235 boolean_t zvec_smush_outnvlist; 236 const char *zvec_name; 237} zfs_ioc_vec_t; 238 239/* This array is indexed by zfs_userquota_prop_t */ 240static const char *userquota_perms[] = { 241 ZFS_DELEG_PERM_USERUSED, 242 ZFS_DELEG_PERM_USERQUOTA, 243 ZFS_DELEG_PERM_GROUPUSED, 244 ZFS_DELEG_PERM_GROUPQUOTA, 245}; 246 247static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc); 248static int zfs_check_settable(const char *name, nvpair_t *property, 249 cred_t *cr); 250static int zfs_check_clearable(char *dataset, nvlist_t *props, 251 nvlist_t **errors); 252static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *, 253 boolean_t *); 254int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *); 255static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp); 256 257static void zfsdev_close(void *data); 258 259static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature); 260 261/* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */ 262void 263__dprintf(const char *file, const char *func, int line, const char *fmt, ...) 264{ 265 const char *newfile; 266 char buf[512]; 267 va_list adx; 268 269 /* 270 * Get rid of annoying "../common/" prefix to filename. 271 */ 272 newfile = strrchr(file, '/'); 273 if (newfile != NULL) { 274 newfile = newfile + 1; /* Get rid of leading / */ 275 } else { 276 newfile = file; 277 } 278 279 va_start(adx, fmt); 280 (void) vsnprintf(buf, sizeof (buf), fmt, adx); 281 va_end(adx); 282 283 /* 284 * To get this data, use the zfs-dprintf probe as so: 285 * dtrace -q -n 'zfs-dprintf \ 286 * /stringof(arg0) == "dbuf.c"/ \ 287 * {printf("%s: %s", stringof(arg1), stringof(arg3))}' 288 * arg0 = file name 289 * arg1 = function name 290 * arg2 = line number 291 * arg3 = message 292 */ 293 DTRACE_PROBE4(zfs__dprintf, 294 char *, newfile, char *, func, int, line, char *, buf); 295} 296 297static void 298history_str_free(char *buf) 299{ 300 kmem_free(buf, HIS_MAX_RECORD_LEN); 301} 302 303static char * 304history_str_get(zfs_cmd_t *zc) 305{ 306 char *buf; 307 308 if (zc->zc_history == 0) 309 return (NULL); 310 311 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP); 312 if (copyinstr((void *)(uintptr_t)zc->zc_history, 313 buf, HIS_MAX_RECORD_LEN, NULL) != 0) { 314 history_str_free(buf); 315 return (NULL); 316 } 317 318 buf[HIS_MAX_RECORD_LEN -1] = '\0'; 319 320 return (buf); 321} 322 323/* 324 * Check to see if the named dataset is currently defined as bootable 325 */ 326static boolean_t 327zfs_is_bootfs(const char *name) 328{ 329 objset_t *os; 330 331 if (dmu_objset_hold(name, FTAG, &os) == 0) { 332 boolean_t ret; 333 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os))); 334 dmu_objset_rele(os, FTAG); 335 return (ret); 336 } 337 return (B_FALSE); 338} 339 340/* 341 * Return non-zero if the spa version is less than requested version. 342 */ 343static int 344zfs_earlier_version(const char *name, int version) 345{ 346 spa_t *spa; 347 348 if (spa_open(name, &spa, FTAG) == 0) { 349 if (spa_version(spa) < version) { 350 spa_close(spa, FTAG); 351 return (1); 352 } 353 spa_close(spa, FTAG); 354 } 355 return (0); 356} 357 358/* 359 * Return TRUE if the ZPL version is less than requested version. 360 */ 361static boolean_t 362zpl_earlier_version(const char *name, int version) 363{ 364 objset_t *os; 365 boolean_t rc = B_TRUE; 366 367 if (dmu_objset_hold(name, FTAG, &os) == 0) { 368 uint64_t zplversion; 369 370 if (dmu_objset_type(os) != DMU_OST_ZFS) { 371 dmu_objset_rele(os, FTAG); 372 return (B_TRUE); 373 } 374 /* XXX reading from non-owned objset */ 375 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0) 376 rc = zplversion < version; 377 dmu_objset_rele(os, FTAG); 378 } 379 return (rc); 380} 381 382static void 383zfs_log_history(zfs_cmd_t *zc) 384{ 385 spa_t *spa; 386 char *buf; 387 388 if ((buf = history_str_get(zc)) == NULL) 389 return; 390 391 if (spa_open(zc->zc_name, &spa, FTAG) == 0) { 392 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY) 393 (void) spa_history_log(spa, buf); 394 spa_close(spa, FTAG); 395 } 396 history_str_free(buf); 397} 398 399/* 400 * Policy for top-level read operations (list pools). Requires no privileges, 401 * and can be used in the local zone, as there is no associated dataset. 402 */ 403/* ARGSUSED */ 404static int 405zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 406{ 407 return (0); 408} 409 410/* 411 * Policy for dataset read operations (list children, get statistics). Requires 412 * no privileges, but must be visible in the local zone. 413 */ 414/* ARGSUSED */ 415static int 416zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 417{ 418 if (INGLOBALZONE(curthread) || 419 zone_dataset_visible(zc->zc_name, NULL)) 420 return (0); 421 422 return (SET_ERROR(ENOENT)); 423} 424 425static int 426zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr) 427{ 428 int writable = 1; 429 430 /* 431 * The dataset must be visible by this zone -- check this first 432 * so they don't see EPERM on something they shouldn't know about. 433 */ 434 if (!INGLOBALZONE(curthread) && 435 !zone_dataset_visible(dataset, &writable)) 436 return (SET_ERROR(ENOENT)); 437 438 if (INGLOBALZONE(curthread)) { 439 /* 440 * If the fs is zoned, only root can access it from the 441 * global zone. 442 */ 443 if (secpolicy_zfs(cr) && zoned) 444 return (SET_ERROR(EPERM)); 445 } else { 446 /* 447 * If we are in a local zone, the 'zoned' property must be set. 448 */ 449 if (!zoned) 450 return (SET_ERROR(EPERM)); 451 452 /* must be writable by this zone */ 453 if (!writable) 454 return (SET_ERROR(EPERM)); 455 } 456 return (0); 457} 458 459static int 460zfs_dozonecheck(const char *dataset, cred_t *cr) 461{ 462 uint64_t zoned; 463 464 if (dsl_prop_get_integer(dataset, "jailed", &zoned, NULL)) 465 return (SET_ERROR(ENOENT)); 466 467 return (zfs_dozonecheck_impl(dataset, zoned, cr)); 468} 469 470static int 471zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr) 472{ 473 uint64_t zoned; 474 475 if (dsl_prop_get_int_ds(ds, "jailed", &zoned)) 476 return (SET_ERROR(ENOENT)); 477 478 return (zfs_dozonecheck_impl(dataset, zoned, cr)); 479} 480 481static int 482zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds, 483 const char *perm, cred_t *cr) 484{ 485 int error; 486 487 error = zfs_dozonecheck_ds(name, ds, cr); 488 if (error == 0) { 489 error = secpolicy_zfs(cr); 490 if (error != 0) 491 error = dsl_deleg_access_impl(ds, perm, cr); 492 } 493 return (error); 494} 495 496static int 497zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr) 498{ 499 int error; 500 dsl_dataset_t *ds; 501 dsl_pool_t *dp; 502 503 /* 504 * First do a quick check for root in the global zone, which 505 * is allowed to do all write_perms. This ensures that zfs_ioc_* 506 * will get to handle nonexistent datasets. 507 */ 508 if (INGLOBALZONE(curthread) && secpolicy_zfs(cr) == 0) 509 return (0); 510 511 error = dsl_pool_hold(name, FTAG, &dp); 512 if (error != 0) 513 return (error); 514 515 error = dsl_dataset_hold(dp, name, FTAG, &ds); 516 if (error != 0) { 517 dsl_pool_rele(dp, FTAG); 518 return (error); 519 } 520 521 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr); 522 523 dsl_dataset_rele(ds, FTAG); 524 dsl_pool_rele(dp, FTAG); 525 return (error); 526} 527 528#ifdef SECLABEL 529/* 530 * Policy for setting the security label property. 531 * 532 * Returns 0 for success, non-zero for access and other errors. 533 */ 534static int 535zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr) 536{ 537 char ds_hexsl[MAXNAMELEN]; 538 bslabel_t ds_sl, new_sl; 539 boolean_t new_default = FALSE; 540 uint64_t zoned; 541 int needed_priv = -1; 542 int error; 543 544 /* First get the existing dataset label. */ 545 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL), 546 1, sizeof (ds_hexsl), &ds_hexsl, NULL); 547 if (error != 0) 548 return (SET_ERROR(EPERM)); 549 550 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0) 551 new_default = TRUE; 552 553 /* The label must be translatable */ 554 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0)) 555 return (SET_ERROR(EINVAL)); 556 557 /* 558 * In a non-global zone, disallow attempts to set a label that 559 * doesn't match that of the zone; otherwise no other checks 560 * are needed. 561 */ 562 if (!INGLOBALZONE(curproc)) { 563 if (new_default || !blequal(&new_sl, CR_SL(CRED()))) 564 return (SET_ERROR(EPERM)); 565 return (0); 566 } 567 568 /* 569 * For global-zone datasets (i.e., those whose zoned property is 570 * "off", verify that the specified new label is valid for the 571 * global zone. 572 */ 573 if (dsl_prop_get_integer(name, 574 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL)) 575 return (SET_ERROR(EPERM)); 576 if (!zoned) { 577 if (zfs_check_global_label(name, strval) != 0) 578 return (SET_ERROR(EPERM)); 579 } 580 581 /* 582 * If the existing dataset label is nondefault, check if the 583 * dataset is mounted (label cannot be changed while mounted). 584 * Get the zfsvfs; if there isn't one, then the dataset isn't 585 * mounted (or isn't a dataset, doesn't exist, ...). 586 */ 587 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) { 588 objset_t *os; 589 static char *setsl_tag = "setsl_tag"; 590 591 /* 592 * Try to own the dataset; abort if there is any error, 593 * (e.g., already mounted, in use, or other error). 594 */ 595 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE, 596 setsl_tag, &os); 597 if (error != 0) 598 return (SET_ERROR(EPERM)); 599 600 dmu_objset_disown(os, setsl_tag); 601 602 if (new_default) { 603 needed_priv = PRIV_FILE_DOWNGRADE_SL; 604 goto out_check; 605 } 606 607 if (hexstr_to_label(strval, &new_sl) != 0) 608 return (SET_ERROR(EPERM)); 609 610 if (blstrictdom(&ds_sl, &new_sl)) 611 needed_priv = PRIV_FILE_DOWNGRADE_SL; 612 else if (blstrictdom(&new_sl, &ds_sl)) 613 needed_priv = PRIV_FILE_UPGRADE_SL; 614 } else { 615 /* dataset currently has a default label */ 616 if (!new_default) 617 needed_priv = PRIV_FILE_UPGRADE_SL; 618 } 619 620out_check: 621 if (needed_priv != -1) 622 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL)); 623 return (0); 624} 625#endif /* SECLABEL */ 626 627static int 628zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval, 629 cred_t *cr) 630{ 631 char *strval; 632 633 /* 634 * Check permissions for special properties. 635 */ 636 switch (prop) { 637 case ZFS_PROP_ZONED: 638 /* 639 * Disallow setting of 'zoned' from within a local zone. 640 */ 641 if (!INGLOBALZONE(curthread)) 642 return (SET_ERROR(EPERM)); 643 break; 644 645 case ZFS_PROP_QUOTA: 646 case ZFS_PROP_FILESYSTEM_LIMIT: 647 case ZFS_PROP_SNAPSHOT_LIMIT: 648 if (!INGLOBALZONE(curthread)) { 649 uint64_t zoned; 650 char setpoint[ZFS_MAX_DATASET_NAME_LEN]; 651 /* 652 * Unprivileged users are allowed to modify the 653 * limit on things *under* (ie. contained by) 654 * the thing they own. 655 */ 656 if (dsl_prop_get_integer(dsname, "jailed", &zoned, 657 setpoint)) 658 return (SET_ERROR(EPERM)); 659 if (!zoned || strlen(dsname) <= strlen(setpoint)) 660 return (SET_ERROR(EPERM)); 661 } 662 break; 663 664 case ZFS_PROP_MLSLABEL: 665#ifdef SECLABEL 666 if (!is_system_labeled()) 667 return (SET_ERROR(EPERM)); 668 669 if (nvpair_value_string(propval, &strval) == 0) { 670 int err; 671 672 err = zfs_set_slabel_policy(dsname, strval, CRED()); 673 if (err != 0) 674 return (err); 675 } 676#else 677 return (EOPNOTSUPP); 678#endif 679 break; 680 } 681 682 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr)); 683} 684 685/* ARGSUSED */ 686static int 687zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 688{ 689 int error; 690 691 error = zfs_dozonecheck(zc->zc_name, cr); 692 if (error != 0) 693 return (error); 694 695 /* 696 * permission to set permissions will be evaluated later in 697 * dsl_deleg_can_allow() 698 */ 699 return (0); 700} 701 702/* ARGSUSED */ 703static int 704zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 705{ 706 return (zfs_secpolicy_write_perms(zc->zc_name, 707 ZFS_DELEG_PERM_ROLLBACK, cr)); 708} 709 710/* ARGSUSED */ 711static int 712zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 713{ 714 dsl_pool_t *dp; 715 dsl_dataset_t *ds; 716 char *cp; 717 int error; 718 719 /* 720 * Generate the current snapshot name from the given objsetid, then 721 * use that name for the secpolicy/zone checks. 722 */ 723 cp = strchr(zc->zc_name, '@'); 724 if (cp == NULL) 725 return (SET_ERROR(EINVAL)); 726 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 727 if (error != 0) 728 return (error); 729 730 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds); 731 if (error != 0) { 732 dsl_pool_rele(dp, FTAG); 733 return (error); 734 } 735 736 dsl_dataset_name(ds, zc->zc_name); 737 738 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds, 739 ZFS_DELEG_PERM_SEND, cr); 740 dsl_dataset_rele(ds, FTAG); 741 dsl_pool_rele(dp, FTAG); 742 743 return (error); 744} 745 746/* ARGSUSED */ 747static int 748zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 749{ 750 return (zfs_secpolicy_write_perms(zc->zc_name, 751 ZFS_DELEG_PERM_SEND, cr)); 752} 753 754/* ARGSUSED */ 755static int 756zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 757{ 758 vnode_t *vp; 759 int error; 760 761 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, 762 NO_FOLLOW, NULL, &vp)) != 0) 763 return (error); 764 765 /* Now make sure mntpnt and dataset are ZFS */ 766 767 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 || 768 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), 769 zc->zc_name) != 0)) { 770 VN_RELE(vp); 771 return (SET_ERROR(EPERM)); 772 } 773 774 VN_RELE(vp); 775 return (dsl_deleg_access(zc->zc_name, 776 ZFS_DELEG_PERM_SHARE, cr)); 777} 778 779int 780zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 781{ 782 if (!INGLOBALZONE(curthread)) 783 return (SET_ERROR(EPERM)); 784 785 if (secpolicy_nfs(cr) == 0) { 786 return (0); 787 } else { 788 return (zfs_secpolicy_deleg_share(zc, innvl, cr)); 789 } 790} 791 792int 793zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 794{ 795 if (!INGLOBALZONE(curthread)) 796 return (SET_ERROR(EPERM)); 797 798 if (secpolicy_smb(cr) == 0) { 799 return (0); 800 } else { 801 return (zfs_secpolicy_deleg_share(zc, innvl, cr)); 802 } 803} 804 805static int 806zfs_get_parent(const char *datasetname, char *parent, int parentsize) 807{ 808 char *cp; 809 810 /* 811 * Remove the @bla or /bla from the end of the name to get the parent. 812 */ 813 (void) strncpy(parent, datasetname, parentsize); 814 cp = strrchr(parent, '@'); 815 if (cp != NULL) { 816 cp[0] = '\0'; 817 } else { 818 cp = strrchr(parent, '/'); 819 if (cp == NULL) 820 return (SET_ERROR(ENOENT)); 821 cp[0] = '\0'; 822 } 823 824 return (0); 825} 826 827int 828zfs_secpolicy_destroy_perms(const char *name, cred_t *cr) 829{ 830 int error; 831 832 if ((error = zfs_secpolicy_write_perms(name, 833 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 834 return (error); 835 836 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr)); 837} 838 839/* ARGSUSED */ 840static int 841zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 842{ 843 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr)); 844} 845 846/* 847 * Destroying snapshots with delegated permissions requires 848 * descendant mount and destroy permissions. 849 */ 850/* ARGSUSED */ 851static int 852zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 853{ 854 nvlist_t *snaps; 855 nvpair_t *pair, *nextpair; 856 int error = 0; 857 858 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 859 return (SET_ERROR(EINVAL)); 860 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 861 pair = nextpair) { 862 nextpair = nvlist_next_nvpair(snaps, pair); 863 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr); 864 if (error == ENOENT) { 865 /* 866 * Ignore any snapshots that don't exist (we consider 867 * them "already destroyed"). Remove the name from the 868 * nvl here in case the snapshot is created between 869 * now and when we try to destroy it (in which case 870 * we don't want to destroy it since we haven't 871 * checked for permission). 872 */ 873 fnvlist_remove_nvpair(snaps, pair); 874 error = 0; 875 } 876 if (error != 0) 877 break; 878 } 879 880 return (error); 881} 882 883int 884zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr) 885{ 886 char parentname[ZFS_MAX_DATASET_NAME_LEN]; 887 int error; 888 889 if ((error = zfs_secpolicy_write_perms(from, 890 ZFS_DELEG_PERM_RENAME, cr)) != 0) 891 return (error); 892 893 if ((error = zfs_secpolicy_write_perms(from, 894 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 895 return (error); 896 897 if ((error = zfs_get_parent(to, parentname, 898 sizeof (parentname))) != 0) 899 return (error); 900 901 if ((error = zfs_secpolicy_write_perms(parentname, 902 ZFS_DELEG_PERM_CREATE, cr)) != 0) 903 return (error); 904 905 if ((error = zfs_secpolicy_write_perms(parentname, 906 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 907 return (error); 908 909 return (error); 910} 911 912/* ARGSUSED */ 913static int 914zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 915{ 916 char *at = NULL; 917 int error; 918 919 if ((zc->zc_cookie & 1) != 0) { 920 /* 921 * This is recursive rename, so the starting snapshot might 922 * not exist. Check file system or volume permission instead. 923 */ 924 at = strchr(zc->zc_name, '@'); 925 if (at == NULL) 926 return (EINVAL); 927 *at = '\0'; 928 } 929 930 error = zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr); 931 932 if (at != NULL) 933 *at = '@'; 934 935 return (error); 936} 937 938/* ARGSUSED */ 939static int 940zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 941{ 942 dsl_pool_t *dp; 943 dsl_dataset_t *clone; 944 int error; 945 946 error = zfs_secpolicy_write_perms(zc->zc_name, 947 ZFS_DELEG_PERM_PROMOTE, cr); 948 if (error != 0) 949 return (error); 950 951 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 952 if (error != 0) 953 return (error); 954 955 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone); 956 957 if (error == 0) { 958 char parentname[ZFS_MAX_DATASET_NAME_LEN]; 959 dsl_dataset_t *origin = NULL; 960 dsl_dir_t *dd; 961 dd = clone->ds_dir; 962 963 error = dsl_dataset_hold_obj(dd->dd_pool, 964 dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin); 965 if (error != 0) { 966 dsl_dataset_rele(clone, FTAG); 967 dsl_pool_rele(dp, FTAG); 968 return (error); 969 } 970 971 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone, 972 ZFS_DELEG_PERM_MOUNT, cr); 973 974 dsl_dataset_name(origin, parentname); 975 if (error == 0) { 976 error = zfs_secpolicy_write_perms_ds(parentname, origin, 977 ZFS_DELEG_PERM_PROMOTE, cr); 978 } 979 dsl_dataset_rele(clone, FTAG); 980 dsl_dataset_rele(origin, FTAG); 981 } 982 dsl_pool_rele(dp, FTAG); 983 return (error); 984} 985 986/* ARGSUSED */ 987static int 988zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 989{ 990 int error; 991 992 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 993 ZFS_DELEG_PERM_RECEIVE, cr)) != 0) 994 return (error); 995 996 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 997 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 998 return (error); 999 1000 return (zfs_secpolicy_write_perms(zc->zc_name, 1001 ZFS_DELEG_PERM_CREATE, cr)); 1002} 1003 1004int 1005zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr) 1006{ 1007 return (zfs_secpolicy_write_perms(name, 1008 ZFS_DELEG_PERM_SNAPSHOT, cr)); 1009} 1010 1011/* 1012 * Check for permission to create each snapshot in the nvlist. 1013 */ 1014/* ARGSUSED */ 1015static int 1016zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1017{ 1018 nvlist_t *snaps; 1019 int error; 1020 nvpair_t *pair; 1021 1022 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 1023 return (SET_ERROR(EINVAL)); 1024 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 1025 pair = nvlist_next_nvpair(snaps, pair)) { 1026 char *name = nvpair_name(pair); 1027 char *atp = strchr(name, '@'); 1028 1029 if (atp == NULL) { 1030 error = SET_ERROR(EINVAL); 1031 break; 1032 } 1033 *atp = '\0'; 1034 error = zfs_secpolicy_snapshot_perms(name, cr); 1035 *atp = '@'; 1036 if (error != 0) 1037 break; 1038 } 1039 return (error); 1040} 1041 1042/* 1043 * Check for permission to create each snapshot in the nvlist. 1044 */ 1045/* ARGSUSED */ 1046static int 1047zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1048{ 1049 int error = 0; 1050 1051 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 1052 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 1053 char *name = nvpair_name(pair); 1054 char *hashp = strchr(name, '#'); 1055 1056 if (hashp == NULL) { 1057 error = SET_ERROR(EINVAL); 1058 break; 1059 } 1060 *hashp = '\0'; 1061 error = zfs_secpolicy_write_perms(name, 1062 ZFS_DELEG_PERM_BOOKMARK, cr); 1063 *hashp = '#'; 1064 if (error != 0) 1065 break; 1066 } 1067 return (error); 1068} 1069 1070/* ARGSUSED */ 1071static int 1072zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1073{ 1074 nvpair_t *pair, *nextpair; 1075 int error = 0; 1076 1077 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL; 1078 pair = nextpair) { 1079 char *name = nvpair_name(pair); 1080 char *hashp = strchr(name, '#'); 1081 nextpair = nvlist_next_nvpair(innvl, pair); 1082 1083 if (hashp == NULL) { 1084 error = SET_ERROR(EINVAL); 1085 break; 1086 } 1087 1088 *hashp = '\0'; 1089 error = zfs_secpolicy_write_perms(name, 1090 ZFS_DELEG_PERM_DESTROY, cr); 1091 *hashp = '#'; 1092 if (error == ENOENT) { 1093 /* 1094 * Ignore any filesystems that don't exist (we consider 1095 * their bookmarks "already destroyed"). Remove 1096 * the name from the nvl here in case the filesystem 1097 * is created between now and when we try to destroy 1098 * the bookmark (in which case we don't want to 1099 * destroy it since we haven't checked for permission). 1100 */ 1101 fnvlist_remove_nvpair(innvl, pair); 1102 error = 0; 1103 } 1104 if (error != 0) 1105 break; 1106 } 1107 1108 return (error); 1109} 1110 1111/* ARGSUSED */ 1112static int 1113zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1114{ 1115 /* 1116 * Even root must have a proper TSD so that we know what pool 1117 * to log to. 1118 */ 1119 if (tsd_get(zfs_allow_log_key) == NULL) 1120 return (SET_ERROR(EPERM)); 1121 return (0); 1122} 1123 1124static int 1125zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1126{ 1127 char parentname[ZFS_MAX_DATASET_NAME_LEN]; 1128 int error; 1129 char *origin; 1130 1131 if ((error = zfs_get_parent(zc->zc_name, parentname, 1132 sizeof (parentname))) != 0) 1133 return (error); 1134 1135 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 && 1136 (error = zfs_secpolicy_write_perms(origin, 1137 ZFS_DELEG_PERM_CLONE, cr)) != 0) 1138 return (error); 1139 1140 if ((error = zfs_secpolicy_write_perms(parentname, 1141 ZFS_DELEG_PERM_CREATE, cr)) != 0) 1142 return (error); 1143 1144 return (zfs_secpolicy_write_perms(parentname, 1145 ZFS_DELEG_PERM_MOUNT, cr)); 1146} 1147 1148/* 1149 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires 1150 * SYS_CONFIG privilege, which is not available in a local zone. 1151 */ 1152/* ARGSUSED */ 1153static int 1154zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1155{ 1156 if (secpolicy_sys_config(cr, B_FALSE) != 0) 1157 return (SET_ERROR(EPERM)); 1158 1159 return (0); 1160} 1161 1162/* 1163 * Policy for object to name lookups. 1164 */ 1165/* ARGSUSED */ 1166static int 1167zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1168{ 1169 int error; 1170 1171 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0) 1172 return (0); 1173 1174 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr); 1175 return (error); 1176} 1177 1178/* 1179 * Policy for fault injection. Requires all privileges. 1180 */ 1181/* ARGSUSED */ 1182static int 1183zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1184{ 1185 return (secpolicy_zinject(cr)); 1186} 1187 1188/* ARGSUSED */ 1189static int 1190zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1191{ 1192 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value); 1193 1194 if (prop == ZPROP_INVAL) { 1195 if (!zfs_prop_user(zc->zc_value)) 1196 return (SET_ERROR(EINVAL)); 1197 return (zfs_secpolicy_write_perms(zc->zc_name, 1198 ZFS_DELEG_PERM_USERPROP, cr)); 1199 } else { 1200 return (zfs_secpolicy_setprop(zc->zc_name, prop, 1201 NULL, cr)); 1202 } 1203} 1204 1205static int 1206zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1207{ 1208 int err = zfs_secpolicy_read(zc, innvl, cr); 1209 if (err) 1210 return (err); 1211 1212 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 1213 return (SET_ERROR(EINVAL)); 1214 1215 if (zc->zc_value[0] == 0) { 1216 /* 1217 * They are asking about a posix uid/gid. If it's 1218 * themself, allow it. 1219 */ 1220 if (zc->zc_objset_type == ZFS_PROP_USERUSED || 1221 zc->zc_objset_type == ZFS_PROP_USERQUOTA) { 1222 if (zc->zc_guid == crgetuid(cr)) 1223 return (0); 1224 } else { 1225 if (groupmember(zc->zc_guid, cr)) 1226 return (0); 1227 } 1228 } 1229 1230 return (zfs_secpolicy_write_perms(zc->zc_name, 1231 userquota_perms[zc->zc_objset_type], cr)); 1232} 1233 1234static int 1235zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1236{ 1237 int err = zfs_secpolicy_read(zc, innvl, cr); 1238 if (err) 1239 return (err); 1240 1241 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 1242 return (SET_ERROR(EINVAL)); 1243 1244 return (zfs_secpolicy_write_perms(zc->zc_name, 1245 userquota_perms[zc->zc_objset_type], cr)); 1246} 1247 1248/* ARGSUSED */ 1249static int 1250zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1251{ 1252 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION, 1253 NULL, cr)); 1254} 1255 1256/* ARGSUSED */ 1257static int 1258zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1259{ 1260 nvpair_t *pair; 1261 nvlist_t *holds; 1262 int error; 1263 1264 error = nvlist_lookup_nvlist(innvl, "holds", &holds); 1265 if (error != 0) 1266 return (SET_ERROR(EINVAL)); 1267 1268 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL; 1269 pair = nvlist_next_nvpair(holds, pair)) { 1270 char fsname[ZFS_MAX_DATASET_NAME_LEN]; 1271 error = dmu_fsname(nvpair_name(pair), fsname); 1272 if (error != 0) 1273 return (error); 1274 error = zfs_secpolicy_write_perms(fsname, 1275 ZFS_DELEG_PERM_HOLD, cr); 1276 if (error != 0) 1277 return (error); 1278 } 1279 return (0); 1280} 1281 1282/* ARGSUSED */ 1283static int 1284zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1285{ 1286 nvpair_t *pair; 1287 int error; 1288 1289 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL; 1290 pair = nvlist_next_nvpair(innvl, pair)) { 1291 char fsname[ZFS_MAX_DATASET_NAME_LEN]; 1292 error = dmu_fsname(nvpair_name(pair), fsname); 1293 if (error != 0) 1294 return (error); 1295 error = zfs_secpolicy_write_perms(fsname, 1296 ZFS_DELEG_PERM_RELEASE, cr); 1297 if (error != 0) 1298 return (error); 1299 } 1300 return (0); 1301} 1302 1303/* 1304 * Policy for allowing temporary snapshots to be taken or released 1305 */ 1306static int 1307zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1308{ 1309 /* 1310 * A temporary snapshot is the same as a snapshot, 1311 * hold, destroy and release all rolled into one. 1312 * Delegated diff alone is sufficient that we allow this. 1313 */ 1314 int error; 1315 1316 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 1317 ZFS_DELEG_PERM_DIFF, cr)) == 0) 1318 return (0); 1319 1320 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr); 1321 if (error == 0) 1322 error = zfs_secpolicy_hold(zc, innvl, cr); 1323 if (error == 0) 1324 error = zfs_secpolicy_release(zc, innvl, cr); 1325 if (error == 0) 1326 error = zfs_secpolicy_destroy(zc, innvl, cr); 1327 return (error); 1328} 1329 1330/* 1331 * Returns the nvlist as specified by the user in the zfs_cmd_t. 1332 */ 1333static int 1334get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp) 1335{ 1336 char *packed; 1337 int error; 1338 nvlist_t *list = NULL; 1339 1340 /* 1341 * Read in and unpack the user-supplied nvlist. 1342 */ 1343 if (size == 0) 1344 return (SET_ERROR(EINVAL)); 1345 1346 packed = kmem_alloc(size, KM_SLEEP); 1347 1348 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size, 1349 iflag)) != 0) { 1350 kmem_free(packed, size); 1351 return (SET_ERROR(EFAULT)); 1352 } 1353 1354 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) { 1355 kmem_free(packed, size); 1356 return (error); 1357 } 1358 1359 kmem_free(packed, size); 1360 1361 *nvp = list; 1362 return (0); 1363} 1364 1365/* 1366 * Reduce the size of this nvlist until it can be serialized in 'max' bytes. 1367 * Entries will be removed from the end of the nvlist, and one int32 entry 1368 * named "N_MORE_ERRORS" will be added indicating how many entries were 1369 * removed. 1370 */ 1371static int 1372nvlist_smush(nvlist_t *errors, size_t max) 1373{ 1374 size_t size; 1375 1376 size = fnvlist_size(errors); 1377 1378 if (size > max) { 1379 nvpair_t *more_errors; 1380 int n = 0; 1381 1382 if (max < 1024) 1383 return (SET_ERROR(ENOMEM)); 1384 1385 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0); 1386 more_errors = nvlist_prev_nvpair(errors, NULL); 1387 1388 do { 1389 nvpair_t *pair = nvlist_prev_nvpair(errors, 1390 more_errors); 1391 fnvlist_remove_nvpair(errors, pair); 1392 n++; 1393 size = fnvlist_size(errors); 1394 } while (size > max); 1395 1396 fnvlist_remove_nvpair(errors, more_errors); 1397 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n); 1398 ASSERT3U(fnvlist_size(errors), <=, max); 1399 } 1400 1401 return (0); 1402} 1403 1404static int 1405put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl) 1406{ 1407 char *packed = NULL; 1408 int error = 0; 1409 size_t size; 1410 1411 size = fnvlist_size(nvl); 1412 1413 if (size > zc->zc_nvlist_dst_size) { 1414 /* 1415 * Solaris returns ENOMEM here, because even if an error is 1416 * returned from an ioctl(2), new zc_nvlist_dst_size will be 1417 * passed to the userland. This is not the case for FreeBSD. 1418 * We need to return 0, so the kernel will copy the 1419 * zc_nvlist_dst_size back and the userland can discover that a 1420 * bigger buffer is needed. 1421 */ 1422 error = 0; 1423 } else { 1424 packed = fnvlist_pack(nvl, &size); 1425 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst, 1426 size, zc->zc_iflags) != 0) 1427 error = SET_ERROR(EFAULT); 1428 fnvlist_pack_free(packed, size); 1429 } 1430 1431 zc->zc_nvlist_dst_size = size; 1432 zc->zc_nvlist_dst_filled = B_TRUE; 1433 return (error); 1434} 1435 1436static int 1437getzfsvfs(const char *dsname, zfsvfs_t **zfvp) 1438{ 1439 objset_t *os; 1440 vfs_t *vfsp; 1441 int error; 1442 1443 error = dmu_objset_hold(dsname, FTAG, &os); 1444 if (error != 0) 1445 return (error); 1446 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1447 dmu_objset_rele(os, FTAG); 1448 return (SET_ERROR(EINVAL)); 1449 } 1450 1451 mutex_enter(&os->os_user_ptr_lock); 1452 *zfvp = dmu_objset_get_user(os); 1453 if (*zfvp) { 1454 vfsp = (*zfvp)->z_vfs; 1455 vfs_ref(vfsp); 1456 } else { 1457 error = SET_ERROR(ESRCH); 1458 } 1459 mutex_exit(&os->os_user_ptr_lock); 1460 dmu_objset_rele(os, FTAG); 1461 if (error == 0) { 1462 error = vfs_busy(vfsp, 0); 1463 vfs_rel(vfsp); 1464 if (error != 0) { 1465 *zfvp = NULL; 1466 error = SET_ERROR(ESRCH); 1467 } 1468 } 1469 return (error); 1470} 1471 1472/* 1473 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which 1474 * case its z_vfs will be NULL, and it will be opened as the owner. 1475 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER, 1476 * which prevents all vnode ops from running. 1477 */ 1478static int 1479zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer) 1480{ 1481 int error = 0; 1482 1483 if (getzfsvfs(name, zfvp) != 0) 1484 error = zfsvfs_create(name, zfvp); 1485 if (error == 0) { 1486 rrm_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER : 1487 RW_READER, tag); 1488 if ((*zfvp)->z_unmounted) { 1489 /* 1490 * XXX we could probably try again, since the unmounting 1491 * thread should be just about to disassociate the 1492 * objset from the zfsvfs. 1493 */ 1494 rrm_exit(&(*zfvp)->z_teardown_lock, tag); 1495 return (SET_ERROR(EBUSY)); 1496 } 1497 } 1498 return (error); 1499} 1500 1501static void 1502zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag) 1503{ 1504 rrm_exit(&zfsvfs->z_teardown_lock, tag); 1505 1506 if (zfsvfs->z_vfs) { 1507#ifdef illumos 1508 VFS_RELE(zfsvfs->z_vfs); 1509#else 1510 vfs_unbusy(zfsvfs->z_vfs); 1511#endif 1512 } else { 1513 dmu_objset_disown(zfsvfs->z_os, zfsvfs); 1514 zfsvfs_free(zfsvfs); 1515 } 1516} 1517 1518static int 1519zfs_ioc_pool_create(zfs_cmd_t *zc) 1520{ 1521 int error; 1522 nvlist_t *config, *props = NULL; 1523 nvlist_t *rootprops = NULL; 1524 nvlist_t *zplprops = NULL; 1525 1526 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1527 zc->zc_iflags, &config)) 1528 return (error); 1529 1530 if (zc->zc_nvlist_src_size != 0 && (error = 1531 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1532 zc->zc_iflags, &props))) { 1533 nvlist_free(config); 1534 return (error); 1535 } 1536 1537 if (props) { 1538 nvlist_t *nvl = NULL; 1539 uint64_t version = SPA_VERSION; 1540 1541 (void) nvlist_lookup_uint64(props, 1542 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version); 1543 if (!SPA_VERSION_IS_SUPPORTED(version)) { 1544 error = SET_ERROR(EINVAL); 1545 goto pool_props_bad; 1546 } 1547 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl); 1548 if (nvl) { 1549 error = nvlist_dup(nvl, &rootprops, KM_SLEEP); 1550 if (error != 0) { 1551 nvlist_free(config); 1552 nvlist_free(props); 1553 return (error); 1554 } 1555 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS); 1556 } 1557 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 1558 error = zfs_fill_zplprops_root(version, rootprops, 1559 zplprops, NULL); 1560 if (error != 0) 1561 goto pool_props_bad; 1562 } 1563 1564 error = spa_create(zc->zc_name, config, props, zplprops); 1565 1566 /* 1567 * Set the remaining root properties 1568 */ 1569 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name, 1570 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0) 1571 (void) spa_destroy(zc->zc_name); 1572 1573pool_props_bad: 1574 nvlist_free(rootprops); 1575 nvlist_free(zplprops); 1576 nvlist_free(config); 1577 nvlist_free(props); 1578 1579 return (error); 1580} 1581 1582static int 1583zfs_ioc_pool_destroy(zfs_cmd_t *zc) 1584{ 1585 int error; 1586 zfs_log_history(zc); 1587 error = spa_destroy(zc->zc_name); 1588 if (error == 0) 1589 zvol_remove_minors(zc->zc_name); 1590 return (error); 1591} 1592 1593static int 1594zfs_ioc_pool_import(zfs_cmd_t *zc) 1595{ 1596 nvlist_t *config, *props = NULL; 1597 uint64_t guid; 1598 int error; 1599 1600 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1601 zc->zc_iflags, &config)) != 0) 1602 return (error); 1603 1604 if (zc->zc_nvlist_src_size != 0 && (error = 1605 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1606 zc->zc_iflags, &props))) { 1607 nvlist_free(config); 1608 return (error); 1609 } 1610 1611 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 || 1612 guid != zc->zc_guid) 1613 error = SET_ERROR(EINVAL); 1614 else 1615 error = spa_import(zc->zc_name, config, props, zc->zc_cookie); 1616 1617 if (zc->zc_nvlist_dst != 0) { 1618 int err; 1619 1620 if ((err = put_nvlist(zc, config)) != 0) 1621 error = err; 1622 } 1623 1624 nvlist_free(config); 1625 1626 nvlist_free(props); 1627 1628 return (error); 1629} 1630 1631static int 1632zfs_ioc_pool_export(zfs_cmd_t *zc) 1633{ 1634 int error; 1635 boolean_t force = (boolean_t)zc->zc_cookie; 1636 boolean_t hardforce = (boolean_t)zc->zc_guid; 1637 1638 zfs_log_history(zc); 1639 error = spa_export(zc->zc_name, NULL, force, hardforce); 1640 if (error == 0) 1641 zvol_remove_minors(zc->zc_name); 1642 return (error); 1643} 1644 1645static int 1646zfs_ioc_pool_configs(zfs_cmd_t *zc) 1647{ 1648 nvlist_t *configs; 1649 int error; 1650 1651 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL) 1652 return (SET_ERROR(EEXIST)); 1653 1654 error = put_nvlist(zc, configs); 1655 1656 nvlist_free(configs); 1657 1658 return (error); 1659} 1660 1661/* 1662 * inputs: 1663 * zc_name name of the pool 1664 * 1665 * outputs: 1666 * zc_cookie real errno 1667 * zc_nvlist_dst config nvlist 1668 * zc_nvlist_dst_size size of config nvlist 1669 */ 1670static int 1671zfs_ioc_pool_stats(zfs_cmd_t *zc) 1672{ 1673 nvlist_t *config; 1674 int error; 1675 int ret = 0; 1676 1677 error = spa_get_stats(zc->zc_name, &config, zc->zc_value, 1678 sizeof (zc->zc_value)); 1679 1680 if (config != NULL) { 1681 ret = put_nvlist(zc, config); 1682 nvlist_free(config); 1683 1684 /* 1685 * The config may be present even if 'error' is non-zero. 1686 * In this case we return success, and preserve the real errno 1687 * in 'zc_cookie'. 1688 */ 1689 zc->zc_cookie = error; 1690 } else { 1691 ret = error; 1692 } 1693 1694 return (ret); 1695} 1696 1697/* 1698 * Try to import the given pool, returning pool stats as appropriate so that 1699 * user land knows which devices are available and overall pool health. 1700 */ 1701static int 1702zfs_ioc_pool_tryimport(zfs_cmd_t *zc) 1703{ 1704 nvlist_t *tryconfig, *config; 1705 int error; 1706 1707 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1708 zc->zc_iflags, &tryconfig)) != 0) 1709 return (error); 1710 1711 config = spa_tryimport(tryconfig); 1712 1713 nvlist_free(tryconfig); 1714 1715 if (config == NULL) 1716 return (SET_ERROR(EINVAL)); 1717 1718 error = put_nvlist(zc, config); 1719 nvlist_free(config); 1720 1721 return (error); 1722} 1723 1724/* 1725 * inputs: 1726 * zc_name name of the pool 1727 * zc_cookie scan func (pool_scan_func_t) 1728 */ 1729static int 1730zfs_ioc_pool_scan(zfs_cmd_t *zc) 1731{ 1732 spa_t *spa; 1733 int error; 1734 1735 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1736 return (error); 1737 1738 if (zc->zc_cookie == POOL_SCAN_NONE) 1739 error = spa_scan_stop(spa); 1740 else 1741 error = spa_scan(spa, zc->zc_cookie); 1742 1743 spa_close(spa, FTAG); 1744 1745 return (error); 1746} 1747 1748static int 1749zfs_ioc_pool_freeze(zfs_cmd_t *zc) 1750{ 1751 spa_t *spa; 1752 int error; 1753 1754 error = spa_open(zc->zc_name, &spa, FTAG); 1755 if (error == 0) { 1756 spa_freeze(spa); 1757 spa_close(spa, FTAG); 1758 } 1759 return (error); 1760} 1761 1762static int 1763zfs_ioc_pool_upgrade(zfs_cmd_t *zc) 1764{ 1765 spa_t *spa; 1766 int error; 1767 1768 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1769 return (error); 1770 1771 if (zc->zc_cookie < spa_version(spa) || 1772 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) { 1773 spa_close(spa, FTAG); 1774 return (SET_ERROR(EINVAL)); 1775 } 1776 1777 spa_upgrade(spa, zc->zc_cookie); 1778 spa_close(spa, FTAG); 1779 1780 return (error); 1781} 1782 1783static int 1784zfs_ioc_pool_get_history(zfs_cmd_t *zc) 1785{ 1786 spa_t *spa; 1787 char *hist_buf; 1788 uint64_t size; 1789 int error; 1790 1791 if ((size = zc->zc_history_len) == 0) 1792 return (SET_ERROR(EINVAL)); 1793 1794 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1795 return (error); 1796 1797 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) { 1798 spa_close(spa, FTAG); 1799 return (SET_ERROR(ENOTSUP)); 1800 } 1801 1802 hist_buf = kmem_alloc(size, KM_SLEEP); 1803 if ((error = spa_history_get(spa, &zc->zc_history_offset, 1804 &zc->zc_history_len, hist_buf)) == 0) { 1805 error = ddi_copyout(hist_buf, 1806 (void *)(uintptr_t)zc->zc_history, 1807 zc->zc_history_len, zc->zc_iflags); 1808 } 1809 1810 spa_close(spa, FTAG); 1811 kmem_free(hist_buf, size); 1812 return (error); 1813} 1814 1815static int 1816zfs_ioc_pool_reguid(zfs_cmd_t *zc) 1817{ 1818 spa_t *spa; 1819 int error; 1820 1821 error = spa_open(zc->zc_name, &spa, FTAG); 1822 if (error == 0) { 1823 error = spa_change_guid(spa); 1824 spa_close(spa, FTAG); 1825 } 1826 return (error); 1827} 1828 1829static int 1830zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc) 1831{ 1832 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value)); 1833} 1834 1835/* 1836 * inputs: 1837 * zc_name name of filesystem 1838 * zc_obj object to find 1839 * 1840 * outputs: 1841 * zc_value name of object 1842 */ 1843static int 1844zfs_ioc_obj_to_path(zfs_cmd_t *zc) 1845{ 1846 objset_t *os; 1847 int error; 1848 1849 /* XXX reading from objset not owned */ 1850 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0) 1851 return (error); 1852 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1853 dmu_objset_rele(os, FTAG); 1854 return (SET_ERROR(EINVAL)); 1855 } 1856 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value, 1857 sizeof (zc->zc_value)); 1858 dmu_objset_rele(os, FTAG); 1859 1860 return (error); 1861} 1862 1863/* 1864 * inputs: 1865 * zc_name name of filesystem 1866 * zc_obj object to find 1867 * 1868 * outputs: 1869 * zc_stat stats on object 1870 * zc_value path to object 1871 */ 1872static int 1873zfs_ioc_obj_to_stats(zfs_cmd_t *zc) 1874{ 1875 objset_t *os; 1876 int error; 1877 1878 /* XXX reading from objset not owned */ 1879 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0) 1880 return (error); 1881 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1882 dmu_objset_rele(os, FTAG); 1883 return (SET_ERROR(EINVAL)); 1884 } 1885 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value, 1886 sizeof (zc->zc_value)); 1887 dmu_objset_rele(os, FTAG); 1888 1889 return (error); 1890} 1891 1892static int 1893zfs_ioc_vdev_add(zfs_cmd_t *zc) 1894{ 1895 spa_t *spa; 1896 int error; 1897 nvlist_t *config, **l2cache, **spares; 1898 uint_t nl2cache = 0, nspares = 0; 1899 1900 error = spa_open(zc->zc_name, &spa, FTAG); 1901 if (error != 0) 1902 return (error); 1903 1904 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1905 zc->zc_iflags, &config); 1906 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE, 1907 &l2cache, &nl2cache); 1908 1909 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES, 1910 &spares, &nspares); 1911 1912#ifdef illumos 1913 /* 1914 * A root pool with concatenated devices is not supported. 1915 * Thus, can not add a device to a root pool. 1916 * 1917 * Intent log device can not be added to a rootpool because 1918 * during mountroot, zil is replayed, a seperated log device 1919 * can not be accessed during the mountroot time. 1920 * 1921 * l2cache and spare devices are ok to be added to a rootpool. 1922 */ 1923 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) { 1924 nvlist_free(config); 1925 spa_close(spa, FTAG); 1926 return (SET_ERROR(EDOM)); 1927 } 1928#endif /* illumos */ 1929 1930 if (error == 0) { 1931 error = spa_vdev_add(spa, config); 1932 nvlist_free(config); 1933 } 1934 spa_close(spa, FTAG); 1935 return (error); 1936} 1937 1938/* 1939 * inputs: 1940 * zc_name name of the pool 1941 * zc_nvlist_conf nvlist of devices to remove 1942 * zc_cookie to stop the remove? 1943 */ 1944static int 1945zfs_ioc_vdev_remove(zfs_cmd_t *zc) 1946{ 1947 spa_t *spa; 1948 int error; 1949 1950 error = spa_open(zc->zc_name, &spa, FTAG); 1951 if (error != 0) 1952 return (error); 1953 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE); 1954 spa_close(spa, FTAG); 1955 return (error); 1956} 1957 1958static int 1959zfs_ioc_vdev_set_state(zfs_cmd_t *zc) 1960{ 1961 spa_t *spa; 1962 int error; 1963 vdev_state_t newstate = VDEV_STATE_UNKNOWN; 1964 1965 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1966 return (error); 1967 switch (zc->zc_cookie) { 1968 case VDEV_STATE_ONLINE: 1969 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate); 1970 break; 1971 1972 case VDEV_STATE_OFFLINE: 1973 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj); 1974 break; 1975 1976 case VDEV_STATE_FAULTED: 1977 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED && 1978 zc->zc_obj != VDEV_AUX_EXTERNAL) 1979 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED; 1980 1981 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj); 1982 break; 1983 1984 case VDEV_STATE_DEGRADED: 1985 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED && 1986 zc->zc_obj != VDEV_AUX_EXTERNAL) 1987 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED; 1988 1989 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj); 1990 break; 1991 1992 default: 1993 error = SET_ERROR(EINVAL); 1994 } 1995 zc->zc_cookie = newstate; 1996 spa_close(spa, FTAG); 1997 return (error); 1998} 1999 2000static int 2001zfs_ioc_vdev_attach(zfs_cmd_t *zc) 2002{ 2003 spa_t *spa; 2004 int replacing = zc->zc_cookie; 2005 nvlist_t *config; 2006 int error; 2007 2008 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 2009 return (error); 2010 2011 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 2012 zc->zc_iflags, &config)) == 0) { 2013 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing); 2014 nvlist_free(config); 2015 } 2016 2017 spa_close(spa, FTAG); 2018 return (error); 2019} 2020 2021static int 2022zfs_ioc_vdev_detach(zfs_cmd_t *zc) 2023{ 2024 spa_t *spa; 2025 int error; 2026 2027 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 2028 return (error); 2029 2030 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE); 2031 2032 spa_close(spa, FTAG); 2033 return (error); 2034} 2035 2036static int 2037zfs_ioc_vdev_split(zfs_cmd_t *zc) 2038{ 2039 spa_t *spa; 2040 nvlist_t *config, *props = NULL; 2041 int error; 2042 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT); 2043 2044 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 2045 return (error); 2046 2047 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 2048 zc->zc_iflags, &config)) { 2049 spa_close(spa, FTAG); 2050 return (error); 2051 } 2052 2053 if (zc->zc_nvlist_src_size != 0 && (error = 2054 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2055 zc->zc_iflags, &props))) { 2056 spa_close(spa, FTAG); 2057 nvlist_free(config); 2058 return (error); 2059 } 2060 2061 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp); 2062 2063 spa_close(spa, FTAG); 2064 2065 nvlist_free(config); 2066 nvlist_free(props); 2067 2068 return (error); 2069} 2070 2071static int 2072zfs_ioc_vdev_setpath(zfs_cmd_t *zc) 2073{ 2074 spa_t *spa; 2075 char *path = zc->zc_value; 2076 uint64_t guid = zc->zc_guid; 2077 int error; 2078 2079 error = spa_open(zc->zc_name, &spa, FTAG); 2080 if (error != 0) 2081 return (error); 2082 2083 error = spa_vdev_setpath(spa, guid, path); 2084 spa_close(spa, FTAG); 2085 return (error); 2086} 2087 2088static int 2089zfs_ioc_vdev_setfru(zfs_cmd_t *zc) 2090{ 2091 spa_t *spa; 2092 char *fru = zc->zc_value; 2093 uint64_t guid = zc->zc_guid; 2094 int error; 2095 2096 error = spa_open(zc->zc_name, &spa, FTAG); 2097 if (error != 0) 2098 return (error); 2099 2100 error = spa_vdev_setfru(spa, guid, fru); 2101 spa_close(spa, FTAG); 2102 return (error); 2103} 2104 2105static int 2106zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os) 2107{ 2108 int error = 0; 2109 nvlist_t *nv; 2110 2111 dmu_objset_fast_stat(os, &zc->zc_objset_stats); 2112 2113 if (zc->zc_nvlist_dst != 0 && 2114 (error = dsl_prop_get_all(os, &nv)) == 0) { 2115 dmu_objset_stats(os, nv); 2116 /* 2117 * NB: zvol_get_stats() will read the objset contents, 2118 * which we aren't supposed to do with a 2119 * DS_MODE_USER hold, because it could be 2120 * inconsistent. So this is a bit of a workaround... 2121 * XXX reading with out owning 2122 */ 2123 if (!zc->zc_objset_stats.dds_inconsistent && 2124 dmu_objset_type(os) == DMU_OST_ZVOL) { 2125 error = zvol_get_stats(os, nv); 2126 if (error == EIO) 2127 return (error); 2128 VERIFY0(error); 2129 } 2130 error = put_nvlist(zc, nv); 2131 nvlist_free(nv); 2132 } 2133 2134 return (error); 2135} 2136 2137/* 2138 * inputs: 2139 * zc_name name of filesystem 2140 * zc_nvlist_dst_size size of buffer for property nvlist 2141 * 2142 * outputs: 2143 * zc_objset_stats stats 2144 * zc_nvlist_dst property nvlist 2145 * zc_nvlist_dst_size size of property nvlist 2146 */ 2147static int 2148zfs_ioc_objset_stats(zfs_cmd_t *zc) 2149{ 2150 objset_t *os; 2151 int error; 2152 2153 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 2154 if (error == 0) { 2155 error = zfs_ioc_objset_stats_impl(zc, os); 2156 dmu_objset_rele(os, FTAG); 2157 } 2158 2159 if (error == ENOMEM) 2160 error = 0; 2161 return (error); 2162} 2163 2164/* 2165 * inputs: 2166 * zc_name name of filesystem 2167 * zc_nvlist_dst_size size of buffer for property nvlist 2168 * 2169 * outputs: 2170 * zc_nvlist_dst received property nvlist 2171 * zc_nvlist_dst_size size of received property nvlist 2172 * 2173 * Gets received properties (distinct from local properties on or after 2174 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from 2175 * local property values. 2176 */ 2177static int 2178zfs_ioc_objset_recvd_props(zfs_cmd_t *zc) 2179{ 2180 int error = 0; 2181 nvlist_t *nv; 2182 2183 /* 2184 * Without this check, we would return local property values if the 2185 * caller has not already received properties on or after 2186 * SPA_VERSION_RECVD_PROPS. 2187 */ 2188 if (!dsl_prop_get_hasrecvd(zc->zc_name)) 2189 return (SET_ERROR(ENOTSUP)); 2190 2191 if (zc->zc_nvlist_dst != 0 && 2192 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) { 2193 error = put_nvlist(zc, nv); 2194 nvlist_free(nv); 2195 } 2196 2197 return (error); 2198} 2199 2200static int 2201nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop) 2202{ 2203 uint64_t value; 2204 int error; 2205 2206 /* 2207 * zfs_get_zplprop() will either find a value or give us 2208 * the default value (if there is one). 2209 */ 2210 if ((error = zfs_get_zplprop(os, prop, &value)) != 0) 2211 return (error); 2212 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0); 2213 return (0); 2214} 2215 2216/* 2217 * inputs: 2218 * zc_name name of filesystem 2219 * zc_nvlist_dst_size size of buffer for zpl property nvlist 2220 * 2221 * outputs: 2222 * zc_nvlist_dst zpl property nvlist 2223 * zc_nvlist_dst_size size of zpl property nvlist 2224 */ 2225static int 2226zfs_ioc_objset_zplprops(zfs_cmd_t *zc) 2227{ 2228 objset_t *os; 2229 int err; 2230 2231 /* XXX reading without owning */ 2232 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os)) 2233 return (err); 2234 2235 dmu_objset_fast_stat(os, &zc->zc_objset_stats); 2236 2237 /* 2238 * NB: nvl_add_zplprop() will read the objset contents, 2239 * which we aren't supposed to do with a DS_MODE_USER 2240 * hold, because it could be inconsistent. 2241 */ 2242 if (zc->zc_nvlist_dst != 0 && 2243 !zc->zc_objset_stats.dds_inconsistent && 2244 dmu_objset_type(os) == DMU_OST_ZFS) { 2245 nvlist_t *nv; 2246 2247 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2248 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 && 2249 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 && 2250 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 && 2251 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0) 2252 err = put_nvlist(zc, nv); 2253 nvlist_free(nv); 2254 } else { 2255 err = SET_ERROR(ENOENT); 2256 } 2257 dmu_objset_rele(os, FTAG); 2258 return (err); 2259} 2260 2261boolean_t 2262dataset_name_hidden(const char *name) 2263{ 2264 /* 2265 * Skip over datasets that are not visible in this zone, 2266 * internal datasets (which have a $ in their name), and 2267 * temporary datasets (which have a % in their name). 2268 */ 2269 if (strchr(name, '$') != NULL) 2270 return (B_TRUE); 2271 if (strchr(name, '%') != NULL) 2272 return (B_TRUE); 2273 if (!INGLOBALZONE(curthread) && !zone_dataset_visible(name, NULL)) 2274 return (B_TRUE); 2275 return (B_FALSE); 2276} 2277 2278/* 2279 * inputs: 2280 * zc_name name of filesystem 2281 * zc_cookie zap cursor 2282 * zc_nvlist_dst_size size of buffer for property nvlist 2283 * 2284 * outputs: 2285 * zc_name name of next filesystem 2286 * zc_cookie zap cursor 2287 * zc_objset_stats stats 2288 * zc_nvlist_dst property nvlist 2289 * zc_nvlist_dst_size size of property nvlist 2290 */ 2291static int 2292zfs_ioc_dataset_list_next(zfs_cmd_t *zc) 2293{ 2294 objset_t *os; 2295 int error; 2296 char *p; 2297 size_t orig_len = strlen(zc->zc_name); 2298 2299top: 2300 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) { 2301 if (error == ENOENT) 2302 error = SET_ERROR(ESRCH); 2303 return (error); 2304 } 2305 2306 p = strrchr(zc->zc_name, '/'); 2307 if (p == NULL || p[1] != '\0') 2308 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name)); 2309 p = zc->zc_name + strlen(zc->zc_name); 2310 2311 do { 2312 error = dmu_dir_list_next(os, 2313 sizeof (zc->zc_name) - (p - zc->zc_name), p, 2314 NULL, &zc->zc_cookie); 2315 if (error == ENOENT) 2316 error = SET_ERROR(ESRCH); 2317 } while (error == 0 && dataset_name_hidden(zc->zc_name)); 2318 dmu_objset_rele(os, FTAG); 2319 2320 /* 2321 * If it's an internal dataset (ie. with a '$' in its name), 2322 * don't try to get stats for it, otherwise we'll return ENOENT. 2323 */ 2324 if (error == 0 && strchr(zc->zc_name, '$') == NULL) { 2325 error = zfs_ioc_objset_stats(zc); /* fill in the stats */ 2326 if (error == ENOENT) { 2327 /* We lost a race with destroy, get the next one. */ 2328 zc->zc_name[orig_len] = '\0'; 2329 goto top; 2330 } 2331 } 2332 return (error); 2333} 2334 2335/* 2336 * inputs: 2337 * zc_name name of filesystem 2338 * zc_cookie zap cursor 2339 * zc_nvlist_dst_size size of buffer for property nvlist 2340 * zc_simple when set, only name is requested 2341 * 2342 * outputs: 2343 * zc_name name of next snapshot 2344 * zc_objset_stats stats 2345 * zc_nvlist_dst property nvlist 2346 * zc_nvlist_dst_size size of property nvlist 2347 */ 2348static int 2349zfs_ioc_snapshot_list_next(zfs_cmd_t *zc) 2350{ 2351 objset_t *os; 2352 int error; 2353 2354 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 2355 if (error != 0) { 2356 return (error == ENOENT ? ESRCH : error); 2357 } 2358 2359 /* 2360 * A dataset name of maximum length cannot have any snapshots, 2361 * so exit immediately. 2362 */ 2363 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= 2364 ZFS_MAX_DATASET_NAME_LEN) { 2365 dmu_objset_rele(os, FTAG); 2366 return (SET_ERROR(ESRCH)); 2367 } 2368 2369 error = dmu_snapshot_list_next(os, 2370 sizeof (zc->zc_name) - strlen(zc->zc_name), 2371 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie, 2372 NULL); 2373 2374 if (error == 0 && !zc->zc_simple) { 2375 dsl_dataset_t *ds; 2376 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool; 2377 2378 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds); 2379 if (error == 0) { 2380 objset_t *ossnap; 2381 2382 error = dmu_objset_from_ds(ds, &ossnap); 2383 if (error == 0) 2384 error = zfs_ioc_objset_stats_impl(zc, ossnap); 2385 dsl_dataset_rele(ds, FTAG); 2386 } 2387 } else if (error == ENOENT) { 2388 error = SET_ERROR(ESRCH); 2389 } 2390 2391 dmu_objset_rele(os, FTAG); 2392 /* if we failed, undo the @ that we tacked on to zc_name */ 2393 if (error != 0) 2394 *strchr(zc->zc_name, '@') = '\0'; 2395 return (error); 2396} 2397 2398static int 2399zfs_prop_set_userquota(const char *dsname, nvpair_t *pair) 2400{ 2401 const char *propname = nvpair_name(pair); 2402 uint64_t *valary; 2403 unsigned int vallen; 2404 const char *domain; 2405 char *dash; 2406 zfs_userquota_prop_t type; 2407 uint64_t rid; 2408 uint64_t quota; 2409 zfsvfs_t *zfsvfs; 2410 int err; 2411 2412 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2413 nvlist_t *attrs; 2414 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 2415 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2416 &pair) != 0) 2417 return (SET_ERROR(EINVAL)); 2418 } 2419 2420 /* 2421 * A correctly constructed propname is encoded as 2422 * userquota@<rid>-<domain>. 2423 */ 2424 if ((dash = strchr(propname, '-')) == NULL || 2425 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 || 2426 vallen != 3) 2427 return (SET_ERROR(EINVAL)); 2428 2429 domain = dash + 1; 2430 type = valary[0]; 2431 rid = valary[1]; 2432 quota = valary[2]; 2433 2434 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE); 2435 if (err == 0) { 2436 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota); 2437 zfsvfs_rele(zfsvfs, FTAG); 2438 } 2439 2440 return (err); 2441} 2442 2443/* 2444 * If the named property is one that has a special function to set its value, 2445 * return 0 on success and a positive error code on failure; otherwise if it is 2446 * not one of the special properties handled by this function, return -1. 2447 * 2448 * XXX: It would be better for callers of the property interface if we handled 2449 * these special cases in dsl_prop.c (in the dsl layer). 2450 */ 2451static int 2452zfs_prop_set_special(const char *dsname, zprop_source_t source, 2453 nvpair_t *pair) 2454{ 2455 const char *propname = nvpair_name(pair); 2456 zfs_prop_t prop = zfs_name_to_prop(propname); 2457 uint64_t intval; 2458 int err = -1; 2459 2460 if (prop == ZPROP_INVAL) { 2461 if (zfs_prop_userquota(propname)) 2462 return (zfs_prop_set_userquota(dsname, pair)); 2463 return (-1); 2464 } 2465 2466 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2467 nvlist_t *attrs; 2468 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 2469 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2470 &pair) == 0); 2471 } 2472 2473 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING) 2474 return (-1); 2475 2476 VERIFY(0 == nvpair_value_uint64(pair, &intval)); 2477 2478 switch (prop) { 2479 case ZFS_PROP_QUOTA: 2480 err = dsl_dir_set_quota(dsname, source, intval); 2481 break; 2482 case ZFS_PROP_REFQUOTA: 2483 err = dsl_dataset_set_refquota(dsname, source, intval); 2484 break; 2485 case ZFS_PROP_FILESYSTEM_LIMIT: 2486 case ZFS_PROP_SNAPSHOT_LIMIT: 2487 if (intval == UINT64_MAX) { 2488 /* clearing the limit, just do it */ 2489 err = 0; 2490 } else { 2491 err = dsl_dir_activate_fs_ss_limit(dsname); 2492 } 2493 /* 2494 * Set err to -1 to force the zfs_set_prop_nvlist code down the 2495 * default path to set the value in the nvlist. 2496 */ 2497 if (err == 0) 2498 err = -1; 2499 break; 2500 case ZFS_PROP_RESERVATION: 2501 err = dsl_dir_set_reservation(dsname, source, intval); 2502 break; 2503 case ZFS_PROP_REFRESERVATION: 2504 err = dsl_dataset_set_refreservation(dsname, source, intval); 2505 break; 2506 case ZFS_PROP_VOLSIZE: 2507 err = zvol_set_volsize(dsname, intval); 2508 break; 2509 case ZFS_PROP_VERSION: 2510 { 2511 zfsvfs_t *zfsvfs; 2512 2513 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0) 2514 break; 2515 2516 err = zfs_set_version(zfsvfs, intval); 2517 zfsvfs_rele(zfsvfs, FTAG); 2518 2519 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) { 2520 zfs_cmd_t *zc; 2521 2522 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP); 2523 (void) strcpy(zc->zc_name, dsname); 2524 (void) zfs_ioc_userspace_upgrade(zc); 2525 kmem_free(zc, sizeof (zfs_cmd_t)); 2526 } 2527 break; 2528 } 2529 default: 2530 err = -1; 2531 } 2532 2533 return (err); 2534} 2535 2536/* 2537 * This function is best effort. If it fails to set any of the given properties, 2538 * it continues to set as many as it can and returns the last error 2539 * encountered. If the caller provides a non-NULL errlist, it will be filled in 2540 * with the list of names of all the properties that failed along with the 2541 * corresponding error numbers. 2542 * 2543 * If every property is set successfully, zero is returned and errlist is not 2544 * modified. 2545 */ 2546int 2547zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl, 2548 nvlist_t *errlist) 2549{ 2550 nvpair_t *pair; 2551 nvpair_t *propval; 2552 int rv = 0; 2553 uint64_t intval; 2554 char *strval; 2555 nvlist_t *genericnvl = fnvlist_alloc(); 2556 nvlist_t *retrynvl = fnvlist_alloc(); 2557 2558retry: 2559 pair = NULL; 2560 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { 2561 const char *propname = nvpair_name(pair); 2562 zfs_prop_t prop = zfs_name_to_prop(propname); 2563 int err = 0; 2564 2565 /* decode the property value */ 2566 propval = pair; 2567 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2568 nvlist_t *attrs; 2569 attrs = fnvpair_value_nvlist(pair); 2570 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2571 &propval) != 0) 2572 err = SET_ERROR(EINVAL); 2573 } 2574 2575 /* Validate value type */ 2576 if (err == 0 && prop == ZPROP_INVAL) { 2577 if (zfs_prop_user(propname)) { 2578 if (nvpair_type(propval) != DATA_TYPE_STRING) 2579 err = SET_ERROR(EINVAL); 2580 } else if (zfs_prop_userquota(propname)) { 2581 if (nvpair_type(propval) != 2582 DATA_TYPE_UINT64_ARRAY) 2583 err = SET_ERROR(EINVAL); 2584 } else { 2585 err = SET_ERROR(EINVAL); 2586 } 2587 } else if (err == 0) { 2588 if (nvpair_type(propval) == DATA_TYPE_STRING) { 2589 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING) 2590 err = SET_ERROR(EINVAL); 2591 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) { 2592 const char *unused; 2593 2594 intval = fnvpair_value_uint64(propval); 2595 2596 switch (zfs_prop_get_type(prop)) { 2597 case PROP_TYPE_NUMBER: 2598 break; 2599 case PROP_TYPE_STRING: 2600 err = SET_ERROR(EINVAL); 2601 break; 2602 case PROP_TYPE_INDEX: 2603 if (zfs_prop_index_to_string(prop, 2604 intval, &unused) != 0) 2605 err = SET_ERROR(EINVAL); 2606 break; 2607 default: 2608 cmn_err(CE_PANIC, 2609 "unknown property type"); 2610 } 2611 } else { 2612 err = SET_ERROR(EINVAL); 2613 } 2614 } 2615 2616 /* Validate permissions */ 2617 if (err == 0) 2618 err = zfs_check_settable(dsname, pair, CRED()); 2619 2620 if (err == 0) { 2621 err = zfs_prop_set_special(dsname, source, pair); 2622 if (err == -1) { 2623 /* 2624 * For better performance we build up a list of 2625 * properties to set in a single transaction. 2626 */ 2627 err = nvlist_add_nvpair(genericnvl, pair); 2628 } else if (err != 0 && nvl != retrynvl) { 2629 /* 2630 * This may be a spurious error caused by 2631 * receiving quota and reservation out of order. 2632 * Try again in a second pass. 2633 */ 2634 err = nvlist_add_nvpair(retrynvl, pair); 2635 } 2636 } 2637 2638 if (err != 0) { 2639 if (errlist != NULL) 2640 fnvlist_add_int32(errlist, propname, err); 2641 rv = err; 2642 } 2643 } 2644 2645 if (nvl != retrynvl && !nvlist_empty(retrynvl)) { 2646 nvl = retrynvl; 2647 goto retry; 2648 } 2649 2650 if (!nvlist_empty(genericnvl) && 2651 dsl_props_set(dsname, source, genericnvl) != 0) { 2652 /* 2653 * If this fails, we still want to set as many properties as we 2654 * can, so try setting them individually. 2655 */ 2656 pair = NULL; 2657 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) { 2658 const char *propname = nvpair_name(pair); 2659 int err = 0; 2660 2661 propval = pair; 2662 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2663 nvlist_t *attrs; 2664 attrs = fnvpair_value_nvlist(pair); 2665 propval = fnvlist_lookup_nvpair(attrs, 2666 ZPROP_VALUE); 2667 } 2668 2669 if (nvpair_type(propval) == DATA_TYPE_STRING) { 2670 strval = fnvpair_value_string(propval); 2671 err = dsl_prop_set_string(dsname, propname, 2672 source, strval); 2673 } else { 2674 intval = fnvpair_value_uint64(propval); 2675 err = dsl_prop_set_int(dsname, propname, source, 2676 intval); 2677 } 2678 2679 if (err != 0) { 2680 if (errlist != NULL) { 2681 fnvlist_add_int32(errlist, propname, 2682 err); 2683 } 2684 rv = err; 2685 } 2686 } 2687 } 2688 nvlist_free(genericnvl); 2689 nvlist_free(retrynvl); 2690 2691 return (rv); 2692} 2693 2694/* 2695 * Check that all the properties are valid user properties. 2696 */ 2697static int 2698zfs_check_userprops(const char *fsname, nvlist_t *nvl) 2699{ 2700 nvpair_t *pair = NULL; 2701 int error = 0; 2702 2703 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { 2704 const char *propname = nvpair_name(pair); 2705 2706 if (!zfs_prop_user(propname) || 2707 nvpair_type(pair) != DATA_TYPE_STRING) 2708 return (SET_ERROR(EINVAL)); 2709 2710 if (error = zfs_secpolicy_write_perms(fsname, 2711 ZFS_DELEG_PERM_USERPROP, CRED())) 2712 return (error); 2713 2714 if (strlen(propname) >= ZAP_MAXNAMELEN) 2715 return (SET_ERROR(ENAMETOOLONG)); 2716 2717 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN) 2718 return (E2BIG); 2719 } 2720 return (0); 2721} 2722 2723static void 2724props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops) 2725{ 2726 nvpair_t *pair; 2727 2728 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2729 2730 pair = NULL; 2731 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) { 2732 if (nvlist_exists(skipped, nvpair_name(pair))) 2733 continue; 2734 2735 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0); 2736 } 2737} 2738 2739static int 2740clear_received_props(const char *dsname, nvlist_t *props, 2741 nvlist_t *skipped) 2742{ 2743 int err = 0; 2744 nvlist_t *cleared_props = NULL; 2745 props_skip(props, skipped, &cleared_props); 2746 if (!nvlist_empty(cleared_props)) { 2747 /* 2748 * Acts on local properties until the dataset has received 2749 * properties at least once on or after SPA_VERSION_RECVD_PROPS. 2750 */ 2751 zprop_source_t flags = (ZPROP_SRC_NONE | 2752 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0)); 2753 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL); 2754 } 2755 nvlist_free(cleared_props); 2756 return (err); 2757} 2758 2759/* 2760 * inputs: 2761 * zc_name name of filesystem 2762 * zc_value name of property to set 2763 * zc_nvlist_src{_size} nvlist of properties to apply 2764 * zc_cookie received properties flag 2765 * 2766 * outputs: 2767 * zc_nvlist_dst{_size} error for each unapplied received property 2768 */ 2769static int 2770zfs_ioc_set_prop(zfs_cmd_t *zc) 2771{ 2772 nvlist_t *nvl; 2773 boolean_t received = zc->zc_cookie; 2774 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED : 2775 ZPROP_SRC_LOCAL); 2776 nvlist_t *errors; 2777 int error; 2778 2779 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2780 zc->zc_iflags, &nvl)) != 0) 2781 return (error); 2782 2783 if (received) { 2784 nvlist_t *origprops; 2785 2786 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) { 2787 (void) clear_received_props(zc->zc_name, 2788 origprops, nvl); 2789 nvlist_free(origprops); 2790 } 2791 2792 error = dsl_prop_set_hasrecvd(zc->zc_name); 2793 } 2794 2795 errors = fnvlist_alloc(); 2796 if (error == 0) 2797 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors); 2798 2799 if (zc->zc_nvlist_dst != 0 && errors != NULL) { 2800 (void) put_nvlist(zc, errors); 2801 } 2802 2803 nvlist_free(errors); 2804 nvlist_free(nvl); 2805 return (error); 2806} 2807 2808/* 2809 * inputs: 2810 * zc_name name of filesystem 2811 * zc_value name of property to inherit 2812 * zc_cookie revert to received value if TRUE 2813 * 2814 * outputs: none 2815 */ 2816static int 2817zfs_ioc_inherit_prop(zfs_cmd_t *zc) 2818{ 2819 const char *propname = zc->zc_value; 2820 zfs_prop_t prop = zfs_name_to_prop(propname); 2821 boolean_t received = zc->zc_cookie; 2822 zprop_source_t source = (received 2823 ? ZPROP_SRC_NONE /* revert to received value, if any */ 2824 : ZPROP_SRC_INHERITED); /* explicitly inherit */ 2825 2826 if (received) { 2827 nvlist_t *dummy; 2828 nvpair_t *pair; 2829 zprop_type_t type; 2830 int err; 2831 2832 /* 2833 * zfs_prop_set_special() expects properties in the form of an 2834 * nvpair with type info. 2835 */ 2836 if (prop == ZPROP_INVAL) { 2837 if (!zfs_prop_user(propname)) 2838 return (SET_ERROR(EINVAL)); 2839 2840 type = PROP_TYPE_STRING; 2841 } else if (prop == ZFS_PROP_VOLSIZE || 2842 prop == ZFS_PROP_VERSION) { 2843 return (SET_ERROR(EINVAL)); 2844 } else { 2845 type = zfs_prop_get_type(prop); 2846 } 2847 2848 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2849 2850 switch (type) { 2851 case PROP_TYPE_STRING: 2852 VERIFY(0 == nvlist_add_string(dummy, propname, "")); 2853 break; 2854 case PROP_TYPE_NUMBER: 2855 case PROP_TYPE_INDEX: 2856 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0)); 2857 break; 2858 default: 2859 nvlist_free(dummy); 2860 return (SET_ERROR(EINVAL)); 2861 } 2862 2863 pair = nvlist_next_nvpair(dummy, NULL); 2864 err = zfs_prop_set_special(zc->zc_name, source, pair); 2865 nvlist_free(dummy); 2866 if (err != -1) 2867 return (err); /* special property already handled */ 2868 } else { 2869 /* 2870 * Only check this in the non-received case. We want to allow 2871 * 'inherit -S' to revert non-inheritable properties like quota 2872 * and reservation to the received or default values even though 2873 * they are not considered inheritable. 2874 */ 2875 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop)) 2876 return (SET_ERROR(EINVAL)); 2877 } 2878 2879 /* property name has been validated by zfs_secpolicy_inherit_prop() */ 2880 return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source)); 2881} 2882 2883static int 2884zfs_ioc_pool_set_props(zfs_cmd_t *zc) 2885{ 2886 nvlist_t *props; 2887 spa_t *spa; 2888 int error; 2889 nvpair_t *pair; 2890 2891 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2892 zc->zc_iflags, &props)) 2893 return (error); 2894 2895 /* 2896 * If the only property is the configfile, then just do a spa_lookup() 2897 * to handle the faulted case. 2898 */ 2899 pair = nvlist_next_nvpair(props, NULL); 2900 if (pair != NULL && strcmp(nvpair_name(pair), 2901 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 && 2902 nvlist_next_nvpair(props, pair) == NULL) { 2903 mutex_enter(&spa_namespace_lock); 2904 if ((spa = spa_lookup(zc->zc_name)) != NULL) { 2905 spa_configfile_set(spa, props, B_FALSE); 2906 spa_config_sync(spa, B_FALSE, B_TRUE); 2907 } 2908 mutex_exit(&spa_namespace_lock); 2909 if (spa != NULL) { 2910 nvlist_free(props); 2911 return (0); 2912 } 2913 } 2914 2915 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { 2916 nvlist_free(props); 2917 return (error); 2918 } 2919 2920 error = spa_prop_set(spa, props); 2921 2922 nvlist_free(props); 2923 spa_close(spa, FTAG); 2924 2925 return (error); 2926} 2927 2928static int 2929zfs_ioc_pool_get_props(zfs_cmd_t *zc) 2930{ 2931 spa_t *spa; 2932 int error; 2933 nvlist_t *nvp = NULL; 2934 2935 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { 2936 /* 2937 * If the pool is faulted, there may be properties we can still 2938 * get (such as altroot and cachefile), so attempt to get them 2939 * anyway. 2940 */ 2941 mutex_enter(&spa_namespace_lock); 2942 if ((spa = spa_lookup(zc->zc_name)) != NULL) 2943 error = spa_prop_get(spa, &nvp); 2944 mutex_exit(&spa_namespace_lock); 2945 } else { 2946 error = spa_prop_get(spa, &nvp); 2947 spa_close(spa, FTAG); 2948 } 2949 2950 if (error == 0 && zc->zc_nvlist_dst != 0) 2951 error = put_nvlist(zc, nvp); 2952 else 2953 error = SET_ERROR(EFAULT); 2954 2955 nvlist_free(nvp); 2956 return (error); 2957} 2958 2959/* 2960 * inputs: 2961 * zc_name name of filesystem 2962 * zc_nvlist_src{_size} nvlist of delegated permissions 2963 * zc_perm_action allow/unallow flag 2964 * 2965 * outputs: none 2966 */ 2967static int 2968zfs_ioc_set_fsacl(zfs_cmd_t *zc) 2969{ 2970 int error; 2971 nvlist_t *fsaclnv = NULL; 2972 2973 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2974 zc->zc_iflags, &fsaclnv)) != 0) 2975 return (error); 2976 2977 /* 2978 * Verify nvlist is constructed correctly 2979 */ 2980 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) { 2981 nvlist_free(fsaclnv); 2982 return (SET_ERROR(EINVAL)); 2983 } 2984 2985 /* 2986 * If we don't have PRIV_SYS_MOUNT, then validate 2987 * that user is allowed to hand out each permission in 2988 * the nvlist(s) 2989 */ 2990 2991 error = secpolicy_zfs(CRED()); 2992 if (error != 0) { 2993 if (zc->zc_perm_action == B_FALSE) { 2994 error = dsl_deleg_can_allow(zc->zc_name, 2995 fsaclnv, CRED()); 2996 } else { 2997 error = dsl_deleg_can_unallow(zc->zc_name, 2998 fsaclnv, CRED()); 2999 } 3000 } 3001 3002 if (error == 0) 3003 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action); 3004 3005 nvlist_free(fsaclnv); 3006 return (error); 3007} 3008 3009/* 3010 * inputs: 3011 * zc_name name of filesystem 3012 * 3013 * outputs: 3014 * zc_nvlist_src{_size} nvlist of delegated permissions 3015 */ 3016static int 3017zfs_ioc_get_fsacl(zfs_cmd_t *zc) 3018{ 3019 nvlist_t *nvp; 3020 int error; 3021 3022 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) { 3023 error = put_nvlist(zc, nvp); 3024 nvlist_free(nvp); 3025 } 3026 3027 return (error); 3028} 3029 3030/* 3031 * Search the vfs list for a specified resource. Returns a pointer to it 3032 * or NULL if no suitable entry is found. The caller of this routine 3033 * is responsible for releasing the returned vfs pointer. 3034 */ 3035static vfs_t * 3036zfs_get_vfs(const char *resource) 3037{ 3038 vfs_t *vfsp; 3039 3040 mtx_lock(&mountlist_mtx); 3041 TAILQ_FOREACH(vfsp, &mountlist, mnt_list) { 3042 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) { 3043 if (vfs_busy(vfsp, MBF_MNTLSTLOCK) != 0) 3044 vfsp = NULL; 3045 break; 3046 } 3047 } 3048 if (vfsp == NULL) 3049 mtx_unlock(&mountlist_mtx); 3050 return (vfsp); 3051} 3052 3053/* ARGSUSED */ 3054static void 3055zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) 3056{ 3057 zfs_creat_t *zct = arg; 3058 3059 zfs_create_fs(os, cr, zct->zct_zplprops, tx); 3060} 3061 3062#define ZFS_PROP_UNDEFINED ((uint64_t)-1) 3063 3064/* 3065 * inputs: 3066 * os parent objset pointer (NULL if root fs) 3067 * fuids_ok fuids allowed in this version of the spa? 3068 * sa_ok SAs allowed in this version of the spa? 3069 * createprops list of properties requested by creator 3070 * 3071 * outputs: 3072 * zplprops values for the zplprops we attach to the master node object 3073 * is_ci true if requested file system will be purely case-insensitive 3074 * 3075 * Determine the settings for utf8only, normalization and 3076 * casesensitivity. Specific values may have been requested by the 3077 * creator and/or we can inherit values from the parent dataset. If 3078 * the file system is of too early a vintage, a creator can not 3079 * request settings for these properties, even if the requested 3080 * setting is the default value. We don't actually want to create dsl 3081 * properties for these, so remove them from the source nvlist after 3082 * processing. 3083 */ 3084static int 3085zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver, 3086 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops, 3087 nvlist_t *zplprops, boolean_t *is_ci) 3088{ 3089 uint64_t sense = ZFS_PROP_UNDEFINED; 3090 uint64_t norm = ZFS_PROP_UNDEFINED; 3091 uint64_t u8 = ZFS_PROP_UNDEFINED; 3092 3093 ASSERT(zplprops != NULL); 3094 3095 /* 3096 * Pull out creator prop choices, if any. 3097 */ 3098 if (createprops) { 3099 (void) nvlist_lookup_uint64(createprops, 3100 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver); 3101 (void) nvlist_lookup_uint64(createprops, 3102 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm); 3103 (void) nvlist_remove_all(createprops, 3104 zfs_prop_to_name(ZFS_PROP_NORMALIZE)); 3105 (void) nvlist_lookup_uint64(createprops, 3106 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8); 3107 (void) nvlist_remove_all(createprops, 3108 zfs_prop_to_name(ZFS_PROP_UTF8ONLY)); 3109 (void) nvlist_lookup_uint64(createprops, 3110 zfs_prop_to_name(ZFS_PROP_CASE), &sense); 3111 (void) nvlist_remove_all(createprops, 3112 zfs_prop_to_name(ZFS_PROP_CASE)); 3113 } 3114 3115 /* 3116 * If the zpl version requested is whacky or the file system 3117 * or pool is version is too "young" to support normalization 3118 * and the creator tried to set a value for one of the props, 3119 * error out. 3120 */ 3121 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) || 3122 (zplver >= ZPL_VERSION_FUID && !fuids_ok) || 3123 (zplver >= ZPL_VERSION_SA && !sa_ok) || 3124 (zplver < ZPL_VERSION_NORMALIZATION && 3125 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED || 3126 sense != ZFS_PROP_UNDEFINED))) 3127 return (SET_ERROR(ENOTSUP)); 3128 3129 /* 3130 * Put the version in the zplprops 3131 */ 3132 VERIFY(nvlist_add_uint64(zplprops, 3133 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0); 3134 3135 if (norm == ZFS_PROP_UNDEFINED) 3136 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0); 3137 VERIFY(nvlist_add_uint64(zplprops, 3138 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0); 3139 3140 /* 3141 * If we're normalizing, names must always be valid UTF-8 strings. 3142 */ 3143 if (norm) 3144 u8 = 1; 3145 if (u8 == ZFS_PROP_UNDEFINED) 3146 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0); 3147 VERIFY(nvlist_add_uint64(zplprops, 3148 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0); 3149 3150 if (sense == ZFS_PROP_UNDEFINED) 3151 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0); 3152 VERIFY(nvlist_add_uint64(zplprops, 3153 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0); 3154 3155 if (is_ci) 3156 *is_ci = (sense == ZFS_CASE_INSENSITIVE); 3157 3158 return (0); 3159} 3160 3161static int 3162zfs_fill_zplprops(const char *dataset, nvlist_t *createprops, 3163 nvlist_t *zplprops, boolean_t *is_ci) 3164{ 3165 boolean_t fuids_ok, sa_ok; 3166 uint64_t zplver = ZPL_VERSION; 3167 objset_t *os = NULL; 3168 char parentname[ZFS_MAX_DATASET_NAME_LEN]; 3169 char *cp; 3170 spa_t *spa; 3171 uint64_t spa_vers; 3172 int error; 3173 3174 (void) strlcpy(parentname, dataset, sizeof (parentname)); 3175 cp = strrchr(parentname, '/'); 3176 ASSERT(cp != NULL); 3177 cp[0] = '\0'; 3178 3179 if ((error = spa_open(dataset, &spa, FTAG)) != 0) 3180 return (error); 3181 3182 spa_vers = spa_version(spa); 3183 spa_close(spa, FTAG); 3184 3185 zplver = zfs_zpl_version_map(spa_vers); 3186 fuids_ok = (zplver >= ZPL_VERSION_FUID); 3187 sa_ok = (zplver >= ZPL_VERSION_SA); 3188 3189 /* 3190 * Open parent object set so we can inherit zplprop values. 3191 */ 3192 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0) 3193 return (error); 3194 3195 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops, 3196 zplprops, is_ci); 3197 dmu_objset_rele(os, FTAG); 3198 return (error); 3199} 3200 3201static int 3202zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops, 3203 nvlist_t *zplprops, boolean_t *is_ci) 3204{ 3205 boolean_t fuids_ok; 3206 boolean_t sa_ok; 3207 uint64_t zplver = ZPL_VERSION; 3208 int error; 3209 3210 zplver = zfs_zpl_version_map(spa_vers); 3211 fuids_ok = (zplver >= ZPL_VERSION_FUID); 3212 sa_ok = (zplver >= ZPL_VERSION_SA); 3213 3214 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok, 3215 createprops, zplprops, is_ci); 3216 return (error); 3217} 3218 3219/* 3220 * innvl: { 3221 * "type" -> dmu_objset_type_t (int32) 3222 * (optional) "props" -> { prop -> value } 3223 * } 3224 * 3225 * outnvl: propname -> error code (int32) 3226 */ 3227static int 3228zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3229{ 3230 int error = 0; 3231 zfs_creat_t zct = { 0 }; 3232 nvlist_t *nvprops = NULL; 3233 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx); 3234 int32_t type32; 3235 dmu_objset_type_t type; 3236 boolean_t is_insensitive = B_FALSE; 3237 3238 if (nvlist_lookup_int32(innvl, "type", &type32) != 0) 3239 return (SET_ERROR(EINVAL)); 3240 type = type32; 3241 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops); 3242 3243 switch (type) { 3244 case DMU_OST_ZFS: 3245 cbfunc = zfs_create_cb; 3246 break; 3247 3248 case DMU_OST_ZVOL: 3249 cbfunc = zvol_create_cb; 3250 break; 3251 3252 default: 3253 cbfunc = NULL; 3254 break; 3255 } 3256 if (strchr(fsname, '@') || 3257 strchr(fsname, '%')) 3258 return (SET_ERROR(EINVAL)); 3259 3260 zct.zct_props = nvprops; 3261 3262 if (cbfunc == NULL) 3263 return (SET_ERROR(EINVAL)); 3264 3265 if (type == DMU_OST_ZVOL) { 3266 uint64_t volsize, volblocksize; 3267 3268 if (nvprops == NULL) 3269 return (SET_ERROR(EINVAL)); 3270 if (nvlist_lookup_uint64(nvprops, 3271 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0) 3272 return (SET_ERROR(EINVAL)); 3273 3274 if ((error = nvlist_lookup_uint64(nvprops, 3275 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), 3276 &volblocksize)) != 0 && error != ENOENT) 3277 return (SET_ERROR(EINVAL)); 3278 3279 if (error != 0) 3280 volblocksize = zfs_prop_default_numeric( 3281 ZFS_PROP_VOLBLOCKSIZE); 3282 3283 if ((error = zvol_check_volblocksize( 3284 volblocksize)) != 0 || 3285 (error = zvol_check_volsize(volsize, 3286 volblocksize)) != 0) 3287 return (error); 3288 } else if (type == DMU_OST_ZFS) { 3289 int error; 3290 3291 /* 3292 * We have to have normalization and 3293 * case-folding flags correct when we do the 3294 * file system creation, so go figure them out 3295 * now. 3296 */ 3297 VERIFY(nvlist_alloc(&zct.zct_zplprops, 3298 NV_UNIQUE_NAME, KM_SLEEP) == 0); 3299 error = zfs_fill_zplprops(fsname, nvprops, 3300 zct.zct_zplprops, &is_insensitive); 3301 if (error != 0) { 3302 nvlist_free(zct.zct_zplprops); 3303 return (error); 3304 } 3305 } 3306 3307 error = dmu_objset_create(fsname, type, 3308 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct); 3309 nvlist_free(zct.zct_zplprops); 3310 3311 /* 3312 * It would be nice to do this atomically. 3313 */ 3314 if (error == 0) { 3315 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL, 3316 nvprops, outnvl); 3317 if (error != 0) 3318 (void) dsl_destroy_head(fsname); 3319 } 3320#ifdef __FreeBSD__ 3321 if (error == 0 && type == DMU_OST_ZVOL) 3322 zvol_create_minors(fsname); 3323#endif 3324 return (error); 3325} 3326 3327/* 3328 * innvl: { 3329 * "origin" -> name of origin snapshot 3330 * (optional) "props" -> { prop -> value } 3331 * } 3332 * 3333 * outnvl: propname -> error code (int32) 3334 */ 3335static int 3336zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3337{ 3338 int error = 0; 3339 nvlist_t *nvprops = NULL; 3340 char *origin_name; 3341 3342 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0) 3343 return (SET_ERROR(EINVAL)); 3344 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops); 3345 3346 if (strchr(fsname, '@') || 3347 strchr(fsname, '%')) 3348 return (SET_ERROR(EINVAL)); 3349 3350 if (dataset_namecheck(origin_name, NULL, NULL) != 0) 3351 return (SET_ERROR(EINVAL)); 3352 error = dmu_objset_clone(fsname, origin_name); 3353 if (error != 0) 3354 return (error); 3355 3356 /* 3357 * It would be nice to do this atomically. 3358 */ 3359 if (error == 0) { 3360 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL, 3361 nvprops, outnvl); 3362 if (error != 0) 3363 (void) dsl_destroy_head(fsname); 3364 } 3365#ifdef __FreeBSD__ 3366 if (error == 0) 3367 zvol_create_minors(fsname); 3368#endif 3369 return (error); 3370} 3371 3372/* 3373 * innvl: { 3374 * "snaps" -> { snapshot1, snapshot2 } 3375 * (optional) "props" -> { prop -> value (string) } 3376 * } 3377 * 3378 * outnvl: snapshot -> error code (int32) 3379 */ 3380static int 3381zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3382{ 3383 nvlist_t *snaps; 3384 nvlist_t *props = NULL; 3385 int error, poollen; 3386 nvpair_t *pair; 3387 3388 (void) nvlist_lookup_nvlist(innvl, "props", &props); 3389 if ((error = zfs_check_userprops(poolname, props)) != 0) 3390 return (error); 3391 3392 if (!nvlist_empty(props) && 3393 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS)) 3394 return (SET_ERROR(ENOTSUP)); 3395 3396 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 3397 return (SET_ERROR(EINVAL)); 3398 poollen = strlen(poolname); 3399 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 3400 pair = nvlist_next_nvpair(snaps, pair)) { 3401 const char *name = nvpair_name(pair); 3402 const char *cp = strchr(name, '@'); 3403 3404 /* 3405 * The snap name must contain an @, and the part after it must 3406 * contain only valid characters. 3407 */ 3408 if (cp == NULL || 3409 zfs_component_namecheck(cp + 1, NULL, NULL) != 0) 3410 return (SET_ERROR(EINVAL)); 3411 3412 /* 3413 * The snap must be in the specified pool. 3414 */ 3415 if (strncmp(name, poolname, poollen) != 0 || 3416 (name[poollen] != '/' && name[poollen] != '@')) 3417 return (SET_ERROR(EXDEV)); 3418 3419 /* This must be the only snap of this fs. */ 3420 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair); 3421 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) { 3422 if (strncmp(name, nvpair_name(pair2), cp - name + 1) 3423 == 0) { 3424 return (SET_ERROR(EXDEV)); 3425 } 3426 } 3427 } 3428 3429 error = dsl_dataset_snapshot(snaps, props, outnvl); 3430 return (error); 3431} 3432 3433/* 3434 * innvl: "message" -> string 3435 */ 3436/* ARGSUSED */ 3437static int 3438zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl) 3439{ 3440 char *message; 3441 spa_t *spa; 3442 int error; 3443 char *poolname; 3444 3445 /* 3446 * The poolname in the ioctl is not set, we get it from the TSD, 3447 * which was set at the end of the last successful ioctl that allows 3448 * logging. The secpolicy func already checked that it is set. 3449 * Only one log ioctl is allowed after each successful ioctl, so 3450 * we clear the TSD here. 3451 */ 3452 poolname = tsd_get(zfs_allow_log_key); 3453 (void) tsd_set(zfs_allow_log_key, NULL); 3454 error = spa_open(poolname, &spa, FTAG); 3455 strfree(poolname); 3456 if (error != 0) 3457 return (error); 3458 3459 if (nvlist_lookup_string(innvl, "message", &message) != 0) { 3460 spa_close(spa, FTAG); 3461 return (SET_ERROR(EINVAL)); 3462 } 3463 3464 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) { 3465 spa_close(spa, FTAG); 3466 return (SET_ERROR(ENOTSUP)); 3467 } 3468 3469 error = spa_history_log(spa, message); 3470 spa_close(spa, FTAG); 3471 return (error); 3472} 3473 3474#ifdef __FreeBSD__ 3475static int 3476zfs_ioc_nextboot(const char *unused, nvlist_t *innvl, nvlist_t *outnvl) 3477{ 3478 char name[MAXNAMELEN]; 3479 spa_t *spa; 3480 vdev_t *vd; 3481 char *command; 3482 uint64_t pool_guid; 3483 uint64_t vdev_guid; 3484 int error; 3485 3486 if (nvlist_lookup_uint64(innvl, 3487 ZPOOL_CONFIG_POOL_GUID, &pool_guid) != 0) 3488 return (EINVAL); 3489 if (nvlist_lookup_uint64(innvl, 3490 ZPOOL_CONFIG_GUID, &vdev_guid) != 0) 3491 return (EINVAL); 3492 if (nvlist_lookup_string(innvl, 3493 "command", &command) != 0) 3494 return (EINVAL); 3495 3496 mutex_enter(&spa_namespace_lock); 3497 spa = spa_by_guid(pool_guid, vdev_guid); 3498 if (spa != NULL) 3499 strcpy(name, spa_name(spa)); 3500 mutex_exit(&spa_namespace_lock); 3501 if (spa == NULL) 3502 return (ENOENT); 3503 3504 if ((error = spa_open(name, &spa, FTAG)) != 0) 3505 return (error); 3506 spa_vdev_state_enter(spa, SCL_ALL); 3507 vd = spa_lookup_by_guid(spa, vdev_guid, B_TRUE); 3508 if (vd == NULL) { 3509 (void) spa_vdev_state_exit(spa, NULL, ENXIO); 3510 spa_close(spa, FTAG); 3511 return (ENODEV); 3512 } 3513 error = vdev_label_write_pad2(vd, command, strlen(command)); 3514 (void) spa_vdev_state_exit(spa, NULL, 0); 3515 txg_wait_synced(spa->spa_dsl_pool, 0); 3516 spa_close(spa, FTAG); 3517 return (error); 3518} 3519#endif 3520 3521/* 3522 * The dp_config_rwlock must not be held when calling this, because the 3523 * unmount may need to write out data. 3524 * 3525 * This function is best-effort. Callers must deal gracefully if it 3526 * remains mounted (or is remounted after this call). 3527 * 3528 * Returns 0 if the argument is not a snapshot, or it is not currently a 3529 * filesystem, or we were able to unmount it. Returns error code otherwise. 3530 */ 3531int 3532zfs_unmount_snap(const char *snapname) 3533{ 3534 vfs_t *vfsp; 3535 zfsvfs_t *zfsvfs; 3536 int err; 3537 3538 if (strchr(snapname, '@') == NULL) 3539 return (0); 3540 3541 vfsp = zfs_get_vfs(snapname); 3542 if (vfsp == NULL) 3543 return (0); 3544 3545 zfsvfs = vfsp->vfs_data; 3546 ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os))); 3547 3548 err = vn_vfswlock(vfsp->vfs_vnodecovered); 3549#ifdef illumos 3550 VFS_RELE(vfsp); 3551#else 3552 vfs_unbusy(vfsp); 3553#endif 3554 if (err != 0) 3555 return (SET_ERROR(err)); 3556 3557 /* 3558 * Always force the unmount for snapshots. 3559 */ 3560 3561#ifdef illumos 3562 (void) dounmount(vfsp, MS_FORCE, kcred); 3563#else 3564 vfs_ref(vfsp); 3565 (void) dounmount(vfsp, MS_FORCE, curthread); 3566#endif 3567 return (0); 3568} 3569 3570/* ARGSUSED */ 3571static int 3572zfs_unmount_snap_cb(const char *snapname, void *arg) 3573{ 3574 return (zfs_unmount_snap(snapname)); 3575} 3576 3577/* 3578 * When a clone is destroyed, its origin may also need to be destroyed, 3579 * in which case it must be unmounted. This routine will do that unmount 3580 * if necessary. 3581 */ 3582void 3583zfs_destroy_unmount_origin(const char *fsname) 3584{ 3585 int error; 3586 objset_t *os; 3587 dsl_dataset_t *ds; 3588 3589 error = dmu_objset_hold(fsname, FTAG, &os); 3590 if (error != 0) 3591 return; 3592 ds = dmu_objset_ds(os); 3593 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) { 3594 char originname[ZFS_MAX_DATASET_NAME_LEN]; 3595 dsl_dataset_name(ds->ds_prev, originname); 3596 dmu_objset_rele(os, FTAG); 3597 (void) zfs_unmount_snap(originname); 3598 } else { 3599 dmu_objset_rele(os, FTAG); 3600 } 3601} 3602 3603/* 3604 * innvl: { 3605 * "snaps" -> { snapshot1, snapshot2 } 3606 * (optional boolean) "defer" 3607 * } 3608 * 3609 * outnvl: snapshot -> error code (int32) 3610 * 3611 */ 3612/* ARGSUSED */ 3613static int 3614zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3615{ 3616 int error, poollen; 3617 nvlist_t *snaps; 3618 nvpair_t *pair; 3619 boolean_t defer; 3620 3621 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 3622 return (SET_ERROR(EINVAL)); 3623 defer = nvlist_exists(innvl, "defer"); 3624 3625 poollen = strlen(poolname); 3626 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 3627 pair = nvlist_next_nvpair(snaps, pair)) { 3628 const char *name = nvpair_name(pair); 3629 3630 /* 3631 * The snap must be in the specified pool to prevent the 3632 * invalid removal of zvol minors below. 3633 */ 3634 if (strncmp(name, poolname, poollen) != 0 || 3635 (name[poollen] != '/' && name[poollen] != '@')) 3636 return (SET_ERROR(EXDEV)); 3637 3638 error = zfs_unmount_snap(name); 3639 if (error != 0) 3640 return (error); 3641#if defined(__FreeBSD__) 3642 zvol_remove_minors(name); 3643#endif 3644 } 3645 3646 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl)); 3647} 3648 3649/* 3650 * Create bookmarks. Bookmark names are of the form <fs>#<bmark>. 3651 * All bookmarks must be in the same pool. 3652 * 3653 * innvl: { 3654 * bookmark1 -> snapshot1, bookmark2 -> snapshot2 3655 * } 3656 * 3657 * outnvl: bookmark -> error code (int32) 3658 * 3659 */ 3660/* ARGSUSED */ 3661static int 3662zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3663{ 3664 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 3665 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 3666 char *snap_name; 3667 3668 /* 3669 * Verify the snapshot argument. 3670 */ 3671 if (nvpair_value_string(pair, &snap_name) != 0) 3672 return (SET_ERROR(EINVAL)); 3673 3674 3675 /* Verify that the keys (bookmarks) are unique */ 3676 for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair); 3677 pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) { 3678 if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0) 3679 return (SET_ERROR(EINVAL)); 3680 } 3681 } 3682 3683 return (dsl_bookmark_create(innvl, outnvl)); 3684} 3685 3686/* 3687 * innvl: { 3688 * property 1, property 2, ... 3689 * } 3690 * 3691 * outnvl: { 3692 * bookmark name 1 -> { property 1, property 2, ... }, 3693 * bookmark name 2 -> { property 1, property 2, ... } 3694 * } 3695 * 3696 */ 3697static int 3698zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3699{ 3700 return (dsl_get_bookmarks(fsname, innvl, outnvl)); 3701} 3702 3703/* 3704 * innvl: { 3705 * bookmark name 1, bookmark name 2 3706 * } 3707 * 3708 * outnvl: bookmark -> error code (int32) 3709 * 3710 */ 3711static int 3712zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl, 3713 nvlist_t *outnvl) 3714{ 3715 int error, poollen; 3716 3717 poollen = strlen(poolname); 3718 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 3719 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 3720 const char *name = nvpair_name(pair); 3721 const char *cp = strchr(name, '#'); 3722 3723 /* 3724 * The bookmark name must contain an #, and the part after it 3725 * must contain only valid characters. 3726 */ 3727 if (cp == NULL || 3728 zfs_component_namecheck(cp + 1, NULL, NULL) != 0) 3729 return (SET_ERROR(EINVAL)); 3730 3731 /* 3732 * The bookmark must be in the specified pool. 3733 */ 3734 if (strncmp(name, poolname, poollen) != 0 || 3735 (name[poollen] != '/' && name[poollen] != '#')) 3736 return (SET_ERROR(EXDEV)); 3737 } 3738 3739 error = dsl_bookmark_destroy(innvl, outnvl); 3740 return (error); 3741} 3742 3743/* 3744 * inputs: 3745 * zc_name name of dataset to destroy 3746 * zc_objset_type type of objset 3747 * zc_defer_destroy mark for deferred destroy 3748 * 3749 * outputs: none 3750 */ 3751static int 3752zfs_ioc_destroy(zfs_cmd_t *zc) 3753{ 3754 int err; 3755 3756 if (zc->zc_objset_type == DMU_OST_ZFS) { 3757 err = zfs_unmount_snap(zc->zc_name); 3758 if (err != 0) 3759 return (err); 3760 } 3761 3762 if (strchr(zc->zc_name, '@')) 3763 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy); 3764 else 3765 err = dsl_destroy_head(zc->zc_name); 3766 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0) 3767#ifdef __FreeBSD__ 3768 zvol_remove_minors(zc->zc_name); 3769#else 3770 (void) zvol_remove_minor(zc->zc_name); 3771#endif 3772 return (err); 3773} 3774 3775/* 3776 * fsname is name of dataset to rollback (to most recent snapshot) 3777 * 3778 * innvl is not used. 3779 * 3780 * outnvl: "target" -> name of most recent snapshot 3781 * } 3782 */ 3783/* ARGSUSED */ 3784static int 3785zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl) 3786{ 3787 zfsvfs_t *zfsvfs; 3788 int error; 3789 3790 if (getzfsvfs(fsname, &zfsvfs) == 0) { 3791 dsl_dataset_t *ds; 3792 3793 ds = dmu_objset_ds(zfsvfs->z_os); 3794 error = zfs_suspend_fs(zfsvfs); 3795 if (error == 0) { 3796 int resume_err; 3797 3798 error = dsl_dataset_rollback(fsname, zfsvfs, outnvl); 3799 resume_err = zfs_resume_fs(zfsvfs, ds); 3800 error = error ? error : resume_err; 3801 } 3802#ifdef illumos 3803 VFS_RELE(zfsvfs->z_vfs); 3804#else 3805 vfs_unbusy(zfsvfs->z_vfs); 3806#endif 3807 } else { 3808 error = dsl_dataset_rollback(fsname, NULL, outnvl); 3809 } 3810 return (error); 3811} 3812 3813static int 3814recursive_unmount(const char *fsname, void *arg) 3815{ 3816 const char *snapname = arg; 3817 char fullname[ZFS_MAX_DATASET_NAME_LEN]; 3818 3819 (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname); 3820 return (zfs_unmount_snap(fullname)); 3821} 3822 3823/* 3824 * inputs: 3825 * zc_name old name of dataset 3826 * zc_value new name of dataset 3827 * zc_cookie recursive flag (only valid for snapshots) 3828 * 3829 * outputs: none 3830 */ 3831static int 3832zfs_ioc_rename(zfs_cmd_t *zc) 3833{ 3834 boolean_t recursive = zc->zc_cookie & 1; 3835 char *at; 3836 boolean_t allow_mounted = B_TRUE; 3837 3838#ifdef __FreeBSD__ 3839 allow_mounted = (zc->zc_cookie & 2) != 0; 3840#endif 3841 3842 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0'; 3843 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 3844 strchr(zc->zc_value, '%')) 3845 return (SET_ERROR(EINVAL)); 3846 3847 at = strchr(zc->zc_name, '@'); 3848 if (at != NULL) { 3849 /* snaps must be in same fs */ 3850 int error; 3851 3852 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1)) 3853 return (SET_ERROR(EXDEV)); 3854 *at = '\0'; 3855 if (zc->zc_objset_type == DMU_OST_ZFS && !allow_mounted) { 3856 error = dmu_objset_find(zc->zc_name, 3857 recursive_unmount, at + 1, 3858 recursive ? DS_FIND_CHILDREN : 0); 3859 if (error != 0) { 3860 *at = '@'; 3861 return (error); 3862 } 3863 } 3864 error = dsl_dataset_rename_snapshot(zc->zc_name, 3865 at + 1, strchr(zc->zc_value, '@') + 1, recursive); 3866 *at = '@'; 3867 3868 return (error); 3869 } else { 3870#ifdef illumos 3871 if (zc->zc_objset_type == DMU_OST_ZVOL) 3872 (void) zvol_remove_minor(zc->zc_name); 3873#endif 3874 return (dsl_dir_rename(zc->zc_name, zc->zc_value)); 3875 } 3876} 3877 3878static int 3879zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr) 3880{ 3881 const char *propname = nvpair_name(pair); 3882 boolean_t issnap = (strchr(dsname, '@') != NULL); 3883 zfs_prop_t prop = zfs_name_to_prop(propname); 3884 uint64_t intval; 3885 int err; 3886 3887 if (prop == ZPROP_INVAL) { 3888 if (zfs_prop_user(propname)) { 3889 if (err = zfs_secpolicy_write_perms(dsname, 3890 ZFS_DELEG_PERM_USERPROP, cr)) 3891 return (err); 3892 return (0); 3893 } 3894 3895 if (!issnap && zfs_prop_userquota(propname)) { 3896 const char *perm = NULL; 3897 const char *uq_prefix = 3898 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA]; 3899 const char *gq_prefix = 3900 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA]; 3901 3902 if (strncmp(propname, uq_prefix, 3903 strlen(uq_prefix)) == 0) { 3904 perm = ZFS_DELEG_PERM_USERQUOTA; 3905 } else if (strncmp(propname, gq_prefix, 3906 strlen(gq_prefix)) == 0) { 3907 perm = ZFS_DELEG_PERM_GROUPQUOTA; 3908 } else { 3909 /* USERUSED and GROUPUSED are read-only */ 3910 return (SET_ERROR(EINVAL)); 3911 } 3912 3913 if (err = zfs_secpolicy_write_perms(dsname, perm, cr)) 3914 return (err); 3915 return (0); 3916 } 3917 3918 return (SET_ERROR(EINVAL)); 3919 } 3920 3921 if (issnap) 3922 return (SET_ERROR(EINVAL)); 3923 3924 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 3925 /* 3926 * dsl_prop_get_all_impl() returns properties in this 3927 * format. 3928 */ 3929 nvlist_t *attrs; 3930 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 3931 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 3932 &pair) == 0); 3933 } 3934 3935 /* 3936 * Check that this value is valid for this pool version 3937 */ 3938 switch (prop) { 3939 case ZFS_PROP_COMPRESSION: 3940 /* 3941 * If the user specified gzip compression, make sure 3942 * the SPA supports it. We ignore any errors here since 3943 * we'll catch them later. 3944 */ 3945 if (nvpair_value_uint64(pair, &intval) == 0) { 3946 if (intval >= ZIO_COMPRESS_GZIP_1 && 3947 intval <= ZIO_COMPRESS_GZIP_9 && 3948 zfs_earlier_version(dsname, 3949 SPA_VERSION_GZIP_COMPRESSION)) { 3950 return (SET_ERROR(ENOTSUP)); 3951 } 3952 3953 if (intval == ZIO_COMPRESS_ZLE && 3954 zfs_earlier_version(dsname, 3955 SPA_VERSION_ZLE_COMPRESSION)) 3956 return (SET_ERROR(ENOTSUP)); 3957 3958 if (intval == ZIO_COMPRESS_LZ4) { 3959 spa_t *spa; 3960 3961 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 3962 return (err); 3963 3964 if (!spa_feature_is_enabled(spa, 3965 SPA_FEATURE_LZ4_COMPRESS)) { 3966 spa_close(spa, FTAG); 3967 return (SET_ERROR(ENOTSUP)); 3968 } 3969 spa_close(spa, FTAG); 3970 } 3971 3972 /* 3973 * If this is a bootable dataset then 3974 * verify that the compression algorithm 3975 * is supported for booting. We must return 3976 * something other than ENOTSUP since it 3977 * implies a downrev pool version. 3978 */ 3979 if (zfs_is_bootfs(dsname) && 3980 !BOOTFS_COMPRESS_VALID(intval)) { 3981 return (SET_ERROR(ERANGE)); 3982 } 3983 } 3984 break; 3985 3986 case ZFS_PROP_COPIES: 3987 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS)) 3988 return (SET_ERROR(ENOTSUP)); 3989 break; 3990 3991 case ZFS_PROP_RECORDSIZE: 3992 /* Record sizes above 128k need the feature to be enabled */ 3993 if (nvpair_value_uint64(pair, &intval) == 0 && 3994 intval > SPA_OLD_MAXBLOCKSIZE) { 3995 spa_t *spa; 3996 3997 /* 3998 * If this is a bootable dataset then 3999 * the we don't allow large (>128K) blocks, 4000 * because GRUB doesn't support them. 4001 */ 4002 if (zfs_is_bootfs(dsname) && 4003 intval > SPA_OLD_MAXBLOCKSIZE) { 4004 return (SET_ERROR(ERANGE)); 4005 } 4006 4007 /* 4008 * We don't allow setting the property above 1MB, 4009 * unless the tunable has been changed. 4010 */ 4011 if (intval > zfs_max_recordsize || 4012 intval > SPA_MAXBLOCKSIZE) 4013 return (SET_ERROR(ERANGE)); 4014 4015 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 4016 return (err); 4017 4018 if (!spa_feature_is_enabled(spa, 4019 SPA_FEATURE_LARGE_BLOCKS)) { 4020 spa_close(spa, FTAG); 4021 return (SET_ERROR(ENOTSUP)); 4022 } 4023 spa_close(spa, FTAG); 4024 } 4025 break; 4026 4027 case ZFS_PROP_SHARESMB: 4028 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID)) 4029 return (SET_ERROR(ENOTSUP)); 4030 break; 4031 4032 case ZFS_PROP_ACLINHERIT: 4033 if (nvpair_type(pair) == DATA_TYPE_UINT64 && 4034 nvpair_value_uint64(pair, &intval) == 0) { 4035 if (intval == ZFS_ACL_PASSTHROUGH_X && 4036 zfs_earlier_version(dsname, 4037 SPA_VERSION_PASSTHROUGH_X)) 4038 return (SET_ERROR(ENOTSUP)); 4039 } 4040 break; 4041 4042 case ZFS_PROP_CHECKSUM: 4043 case ZFS_PROP_DEDUP: 4044 { 4045 spa_feature_t feature; 4046 spa_t *spa; 4047 4048 /* dedup feature version checks */ 4049 if (prop == ZFS_PROP_DEDUP && 4050 zfs_earlier_version(dsname, SPA_VERSION_DEDUP)) 4051 return (SET_ERROR(ENOTSUP)); 4052 4053 if (nvpair_value_uint64(pair, &intval) != 0) 4054 return (SET_ERROR(EINVAL)); 4055 4056 /* check prop value is enabled in features */ 4057 feature = zio_checksum_to_feature(intval & ZIO_CHECKSUM_MASK); 4058 if (feature == SPA_FEATURE_NONE) 4059 break; 4060 4061 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 4062 return (err); 4063 /* 4064 * Salted checksums are not supported on root pools. 4065 */ 4066 if (spa_bootfs(spa) != 0 && 4067 intval < ZIO_CHECKSUM_FUNCTIONS && 4068 (zio_checksum_table[intval].ci_flags & 4069 ZCHECKSUM_FLAG_SALTED)) { 4070 spa_close(spa, FTAG); 4071 return (SET_ERROR(ERANGE)); 4072 } 4073 if (!spa_feature_is_enabled(spa, feature)) { 4074 spa_close(spa, FTAG); 4075 return (SET_ERROR(ENOTSUP)); 4076 } 4077 spa_close(spa, FTAG); 4078 break; 4079 } 4080 } 4081 4082 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED())); 4083} 4084 4085/* 4086 * Checks for a race condition to make sure we don't increment a feature flag 4087 * multiple times. 4088 */ 4089static int 4090zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx) 4091{ 4092 spa_t *spa = dmu_tx_pool(tx)->dp_spa; 4093 spa_feature_t *featurep = arg; 4094 4095 if (!spa_feature_is_active(spa, *featurep)) 4096 return (0); 4097 else 4098 return (SET_ERROR(EBUSY)); 4099} 4100 4101/* 4102 * The callback invoked on feature activation in the sync task caused by 4103 * zfs_prop_activate_feature. 4104 */ 4105static void 4106zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx) 4107{ 4108 spa_t *spa = dmu_tx_pool(tx)->dp_spa; 4109 spa_feature_t *featurep = arg; 4110 4111 spa_feature_incr(spa, *featurep, tx); 4112} 4113 4114/* 4115 * Activates a feature on a pool in response to a property setting. This 4116 * creates a new sync task which modifies the pool to reflect the feature 4117 * as being active. 4118 */ 4119static int 4120zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature) 4121{ 4122 int err; 4123 4124 /* EBUSY here indicates that the feature is already active */ 4125 err = dsl_sync_task(spa_name(spa), 4126 zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync, 4127 &feature, 2, ZFS_SPACE_CHECK_RESERVED); 4128 4129 if (err != 0 && err != EBUSY) 4130 return (err); 4131 else 4132 return (0); 4133} 4134 4135/* 4136 * Removes properties from the given props list that fail permission checks 4137 * needed to clear them and to restore them in case of a receive error. For each 4138 * property, make sure we have both set and inherit permissions. 4139 * 4140 * Returns the first error encountered if any permission checks fail. If the 4141 * caller provides a non-NULL errlist, it also gives the complete list of names 4142 * of all the properties that failed a permission check along with the 4143 * corresponding error numbers. The caller is responsible for freeing the 4144 * returned errlist. 4145 * 4146 * If every property checks out successfully, zero is returned and the list 4147 * pointed at by errlist is NULL. 4148 */ 4149static int 4150zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist) 4151{ 4152 zfs_cmd_t *zc; 4153 nvpair_t *pair, *next_pair; 4154 nvlist_t *errors; 4155 int err, rv = 0; 4156 4157 if (props == NULL) 4158 return (0); 4159 4160 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0); 4161 4162 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP); 4163 (void) strcpy(zc->zc_name, dataset); 4164 pair = nvlist_next_nvpair(props, NULL); 4165 while (pair != NULL) { 4166 next_pair = nvlist_next_nvpair(props, pair); 4167 4168 (void) strcpy(zc->zc_value, nvpair_name(pair)); 4169 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 || 4170 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) { 4171 VERIFY(nvlist_remove_nvpair(props, pair) == 0); 4172 VERIFY(nvlist_add_int32(errors, 4173 zc->zc_value, err) == 0); 4174 } 4175 pair = next_pair; 4176 } 4177 kmem_free(zc, sizeof (zfs_cmd_t)); 4178 4179 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) { 4180 nvlist_free(errors); 4181 errors = NULL; 4182 } else { 4183 VERIFY(nvpair_value_int32(pair, &rv) == 0); 4184 } 4185 4186 if (errlist == NULL) 4187 nvlist_free(errors); 4188 else 4189 *errlist = errors; 4190 4191 return (rv); 4192} 4193 4194static boolean_t 4195propval_equals(nvpair_t *p1, nvpair_t *p2) 4196{ 4197 if (nvpair_type(p1) == DATA_TYPE_NVLIST) { 4198 /* dsl_prop_get_all_impl() format */ 4199 nvlist_t *attrs; 4200 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0); 4201 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 4202 &p1) == 0); 4203 } 4204 4205 if (nvpair_type(p2) == DATA_TYPE_NVLIST) { 4206 nvlist_t *attrs; 4207 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0); 4208 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 4209 &p2) == 0); 4210 } 4211 4212 if (nvpair_type(p1) != nvpair_type(p2)) 4213 return (B_FALSE); 4214 4215 if (nvpair_type(p1) == DATA_TYPE_STRING) { 4216 char *valstr1, *valstr2; 4217 4218 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0); 4219 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0); 4220 return (strcmp(valstr1, valstr2) == 0); 4221 } else { 4222 uint64_t intval1, intval2; 4223 4224 VERIFY(nvpair_value_uint64(p1, &intval1) == 0); 4225 VERIFY(nvpair_value_uint64(p2, &intval2) == 0); 4226 return (intval1 == intval2); 4227 } 4228} 4229 4230/* 4231 * Remove properties from props if they are not going to change (as determined 4232 * by comparison with origprops). Remove them from origprops as well, since we 4233 * do not need to clear or restore properties that won't change. 4234 */ 4235static void 4236props_reduce(nvlist_t *props, nvlist_t *origprops) 4237{ 4238 nvpair_t *pair, *next_pair; 4239 4240 if (origprops == NULL) 4241 return; /* all props need to be received */ 4242 4243 pair = nvlist_next_nvpair(props, NULL); 4244 while (pair != NULL) { 4245 const char *propname = nvpair_name(pair); 4246 nvpair_t *match; 4247 4248 next_pair = nvlist_next_nvpair(props, pair); 4249 4250 if ((nvlist_lookup_nvpair(origprops, propname, 4251 &match) != 0) || !propval_equals(pair, match)) 4252 goto next; /* need to set received value */ 4253 4254 /* don't clear the existing received value */ 4255 (void) nvlist_remove_nvpair(origprops, match); 4256 /* don't bother receiving the property */ 4257 (void) nvlist_remove_nvpair(props, pair); 4258next: 4259 pair = next_pair; 4260 } 4261} 4262 4263/* 4264 * Extract properties that cannot be set PRIOR to the receipt of a dataset. 4265 * For example, refquota cannot be set until after the receipt of a dataset, 4266 * because in replication streams, an older/earlier snapshot may exceed the 4267 * refquota. We want to receive the older/earlier snapshot, but setting 4268 * refquota pre-receipt will set the dsl's ACTUAL quota, which will prevent 4269 * the older/earlier snapshot from being received (with EDQUOT). 4270 * 4271 * The ZFS test "zfs_receive_011_pos" demonstrates such a scenario. 4272 * 4273 * libzfs will need to be judicious handling errors encountered by props 4274 * extracted by this function. 4275 */ 4276static nvlist_t * 4277extract_delay_props(nvlist_t *props) 4278{ 4279 nvlist_t *delayprops; 4280 nvpair_t *nvp, *tmp; 4281 static const zfs_prop_t delayable[] = { ZFS_PROP_REFQUOTA, 0 }; 4282 int i; 4283 4284 VERIFY(nvlist_alloc(&delayprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 4285 4286 for (nvp = nvlist_next_nvpair(props, NULL); nvp != NULL; 4287 nvp = nvlist_next_nvpair(props, nvp)) { 4288 /* 4289 * strcmp() is safe because zfs_prop_to_name() always returns 4290 * a bounded string. 4291 */ 4292 for (i = 0; delayable[i] != 0; i++) { 4293 if (strcmp(zfs_prop_to_name(delayable[i]), 4294 nvpair_name(nvp)) == 0) { 4295 break; 4296 } 4297 } 4298 if (delayable[i] != 0) { 4299 tmp = nvlist_prev_nvpair(props, nvp); 4300 VERIFY(nvlist_add_nvpair(delayprops, nvp) == 0); 4301 VERIFY(nvlist_remove_nvpair(props, nvp) == 0); 4302 nvp = tmp; 4303 } 4304 } 4305 4306 if (nvlist_empty(delayprops)) { 4307 nvlist_free(delayprops); 4308 delayprops = NULL; 4309 } 4310 return (delayprops); 4311} 4312 4313#ifdef DEBUG 4314static boolean_t zfs_ioc_recv_inject_err; 4315#endif 4316 4317/* 4318 * inputs: 4319 * zc_name name of containing filesystem 4320 * zc_nvlist_src{_size} nvlist of properties to apply 4321 * zc_value name of snapshot to create 4322 * zc_string name of clone origin (if DRR_FLAG_CLONE) 4323 * zc_cookie file descriptor to recv from 4324 * zc_begin_record the BEGIN record of the stream (not byteswapped) 4325 * zc_guid force flag 4326 * zc_cleanup_fd cleanup-on-exit file descriptor 4327 * zc_action_handle handle for this guid/ds mapping (or zero on first call) 4328 * zc_resumable if data is incomplete assume sender will resume 4329 * 4330 * outputs: 4331 * zc_cookie number of bytes read 4332 * zc_nvlist_dst{_size} error for each unapplied received property 4333 * zc_obj zprop_errflags_t 4334 * zc_action_handle handle for this guid/ds mapping 4335 */ 4336static int 4337zfs_ioc_recv(zfs_cmd_t *zc) 4338{ 4339 file_t *fp; 4340 dmu_recv_cookie_t drc; 4341 boolean_t force = (boolean_t)zc->zc_guid; 4342 int fd; 4343 int error = 0; 4344 int props_error = 0; 4345 nvlist_t *errors; 4346 offset_t off; 4347 nvlist_t *props = NULL; /* sent properties */ 4348 nvlist_t *origprops = NULL; /* existing properties */ 4349 nvlist_t *delayprops = NULL; /* sent properties applied post-receive */ 4350 char *origin = NULL; 4351 char *tosnap; 4352 char tofs[ZFS_MAX_DATASET_NAME_LEN]; 4353 cap_rights_t rights; 4354 boolean_t first_recvd_props = B_FALSE; 4355 4356 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 4357 strchr(zc->zc_value, '@') == NULL || 4358 strchr(zc->zc_value, '%')) 4359 return (SET_ERROR(EINVAL)); 4360 4361 (void) strcpy(tofs, zc->zc_value); 4362 tosnap = strchr(tofs, '@'); 4363 *tosnap++ = '\0'; 4364 4365 if (zc->zc_nvlist_src != 0 && 4366 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 4367 zc->zc_iflags, &props)) != 0) 4368 return (error); 4369 4370 fd = zc->zc_cookie; 4371#ifdef illumos 4372 fp = getf(fd); 4373#else 4374 fget_read(curthread, fd, cap_rights_init(&rights, CAP_PREAD), &fp); 4375#endif 4376 if (fp == NULL) { 4377 nvlist_free(props); 4378 return (SET_ERROR(EBADF)); 4379 } 4380 4381 errors = fnvlist_alloc(); 4382 4383 if (zc->zc_string[0]) 4384 origin = zc->zc_string; 4385 4386 error = dmu_recv_begin(tofs, tosnap, 4387 &zc->zc_begin_record, force, zc->zc_resumable, origin, &drc); 4388 if (error != 0) 4389 goto out; 4390 4391 /* 4392 * Set properties before we receive the stream so that they are applied 4393 * to the new data. Note that we must call dmu_recv_stream() if 4394 * dmu_recv_begin() succeeds. 4395 */ 4396 if (props != NULL && !drc.drc_newfs) { 4397 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >= 4398 SPA_VERSION_RECVD_PROPS && 4399 !dsl_prop_get_hasrecvd(tofs)) 4400 first_recvd_props = B_TRUE; 4401 4402 /* 4403 * If new received properties are supplied, they are to 4404 * completely replace the existing received properties, so stash 4405 * away the existing ones. 4406 */ 4407 if (dsl_prop_get_received(tofs, &origprops) == 0) { 4408 nvlist_t *errlist = NULL; 4409 /* 4410 * Don't bother writing a property if its value won't 4411 * change (and avoid the unnecessary security checks). 4412 * 4413 * The first receive after SPA_VERSION_RECVD_PROPS is a 4414 * special case where we blow away all local properties 4415 * regardless. 4416 */ 4417 if (!first_recvd_props) 4418 props_reduce(props, origprops); 4419 if (zfs_check_clearable(tofs, origprops, &errlist) != 0) 4420 (void) nvlist_merge(errors, errlist, 0); 4421 nvlist_free(errlist); 4422 4423 if (clear_received_props(tofs, origprops, 4424 first_recvd_props ? NULL : props) != 0) 4425 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 4426 } else { 4427 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 4428 } 4429 } 4430 4431 if (props != NULL) { 4432 props_error = dsl_prop_set_hasrecvd(tofs); 4433 4434 if (props_error == 0) { 4435 delayprops = extract_delay_props(props); 4436 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED, 4437 props, errors); 4438 } 4439 } 4440 4441 off = fp->f_offset; 4442 error = dmu_recv_stream(&drc, fp, &off, zc->zc_cleanup_fd, 4443 &zc->zc_action_handle); 4444 4445 if (error == 0) { 4446 zfsvfs_t *zfsvfs = NULL; 4447 4448 if (getzfsvfs(tofs, &zfsvfs) == 0) { 4449 /* online recv */ 4450 dsl_dataset_t *ds; 4451 int end_err; 4452 4453 ds = dmu_objset_ds(zfsvfs->z_os); 4454 error = zfs_suspend_fs(zfsvfs); 4455 /* 4456 * If the suspend fails, then the recv_end will 4457 * likely also fail, and clean up after itself. 4458 */ 4459 end_err = dmu_recv_end(&drc, zfsvfs); 4460 if (error == 0) 4461 error = zfs_resume_fs(zfsvfs, ds); 4462 error = error ? error : end_err; 4463#ifdef illumos 4464 VFS_RELE(zfsvfs->z_vfs); 4465#else 4466 vfs_unbusy(zfsvfs->z_vfs); 4467#endif 4468 } else { 4469 error = dmu_recv_end(&drc, NULL); 4470 } 4471 4472 /* Set delayed properties now, after we're done receiving. */ 4473 if (delayprops != NULL && error == 0) { 4474 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED, 4475 delayprops, errors); 4476 } 4477 } 4478 4479 if (delayprops != NULL) { 4480 /* 4481 * Merge delayed props back in with initial props, in case 4482 * we're DEBUG and zfs_ioc_recv_inject_err is set (which means 4483 * we have to make sure clear_received_props() includes 4484 * the delayed properties). 4485 * 4486 * Since zfs_ioc_recv_inject_err is only in DEBUG kernels, 4487 * using ASSERT() will be just like a VERIFY. 4488 */ 4489 ASSERT(nvlist_merge(props, delayprops, 0) == 0); 4490 nvlist_free(delayprops); 4491 } 4492 4493 /* 4494 * Now that all props, initial and delayed, are set, report the prop 4495 * errors to the caller. 4496 */ 4497 if (zc->zc_nvlist_dst_size != 0 && 4498 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 || 4499 put_nvlist(zc, errors) != 0)) { 4500 /* 4501 * Caller made zc->zc_nvlist_dst less than the minimum expected 4502 * size or supplied an invalid address. 4503 */ 4504 props_error = SET_ERROR(EINVAL); 4505 } 4506 4507 zc->zc_cookie = off - fp->f_offset; 4508 if (off >= 0 && off <= MAXOFFSET_T) 4509 fp->f_offset = off; 4510 4511#ifdef DEBUG 4512 if (zfs_ioc_recv_inject_err) { 4513 zfs_ioc_recv_inject_err = B_FALSE; 4514 error = 1; 4515 } 4516#endif 4517 4518#ifdef __FreeBSD__ 4519 if (error == 0) 4520 zvol_create_minors(tofs); 4521#endif 4522 4523 /* 4524 * On error, restore the original props. 4525 */ 4526 if (error != 0 && props != NULL && !drc.drc_newfs) { 4527 if (clear_received_props(tofs, props, NULL) != 0) { 4528 /* 4529 * We failed to clear the received properties. 4530 * Since we may have left a $recvd value on the 4531 * system, we can't clear the $hasrecvd flag. 4532 */ 4533 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4534 } else if (first_recvd_props) { 4535 dsl_prop_unset_hasrecvd(tofs); 4536 } 4537 4538 if (origprops == NULL && !drc.drc_newfs) { 4539 /* We failed to stash the original properties. */ 4540 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4541 } 4542 4543 /* 4544 * dsl_props_set() will not convert RECEIVED to LOCAL on or 4545 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL 4546 * explictly if we're restoring local properties cleared in the 4547 * first new-style receive. 4548 */ 4549 if (origprops != NULL && 4550 zfs_set_prop_nvlist(tofs, (first_recvd_props ? 4551 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED), 4552 origprops, NULL) != 0) { 4553 /* 4554 * We stashed the original properties but failed to 4555 * restore them. 4556 */ 4557 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4558 } 4559 } 4560out: 4561 nvlist_free(props); 4562 nvlist_free(origprops); 4563 nvlist_free(errors); 4564 releasef(fd); 4565 4566 if (error == 0) 4567 error = props_error; 4568 4569 return (error); 4570} 4571 4572/* 4573 * inputs: 4574 * zc_name name of snapshot to send 4575 * zc_cookie file descriptor to send stream to 4576 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj) 4577 * zc_sendobj objsetid of snapshot to send 4578 * zc_fromobj objsetid of incremental fromsnap (may be zero) 4579 * zc_guid if set, estimate size of stream only. zc_cookie is ignored. 4580 * output size in zc_objset_type. 4581 * zc_flags lzc_send_flags 4582 * 4583 * outputs: 4584 * zc_objset_type estimated size, if zc_guid is set 4585 */ 4586static int 4587zfs_ioc_send(zfs_cmd_t *zc) 4588{ 4589 int error; 4590 offset_t off; 4591 boolean_t estimate = (zc->zc_guid != 0); 4592 boolean_t embedok = (zc->zc_flags & 0x1); 4593 boolean_t large_block_ok = (zc->zc_flags & 0x2); 4594 4595 if (zc->zc_obj != 0) { 4596 dsl_pool_t *dp; 4597 dsl_dataset_t *tosnap; 4598 4599 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4600 if (error != 0) 4601 return (error); 4602 4603 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap); 4604 if (error != 0) { 4605 dsl_pool_rele(dp, FTAG); 4606 return (error); 4607 } 4608 4609 if (dsl_dir_is_clone(tosnap->ds_dir)) 4610 zc->zc_fromobj = 4611 dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj; 4612 dsl_dataset_rele(tosnap, FTAG); 4613 dsl_pool_rele(dp, FTAG); 4614 } 4615 4616 if (estimate) { 4617 dsl_pool_t *dp; 4618 dsl_dataset_t *tosnap; 4619 dsl_dataset_t *fromsnap = NULL; 4620 4621 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4622 if (error != 0) 4623 return (error); 4624 4625 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap); 4626 if (error != 0) { 4627 dsl_pool_rele(dp, FTAG); 4628 return (error); 4629 } 4630 4631 if (zc->zc_fromobj != 0) { 4632 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj, 4633 FTAG, &fromsnap); 4634 if (error != 0) { 4635 dsl_dataset_rele(tosnap, FTAG); 4636 dsl_pool_rele(dp, FTAG); 4637 return (error); 4638 } 4639 } 4640 4641 error = dmu_send_estimate(tosnap, fromsnap, 4642 &zc->zc_objset_type); 4643 4644 if (fromsnap != NULL) 4645 dsl_dataset_rele(fromsnap, FTAG); 4646 dsl_dataset_rele(tosnap, FTAG); 4647 dsl_pool_rele(dp, FTAG); 4648 } else { 4649 file_t *fp; 4650 cap_rights_t rights; 4651 4652#ifdef illumos 4653 fp = getf(zc->zc_cookie); 4654#else 4655 fget_write(curthread, zc->zc_cookie, 4656 cap_rights_init(&rights, CAP_WRITE), &fp); 4657#endif 4658 if (fp == NULL) 4659 return (SET_ERROR(EBADF)); 4660 4661 off = fp->f_offset; 4662 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj, 4663 zc->zc_fromobj, embedok, large_block_ok, 4664#ifdef illumos 4665 zc->zc_cookie, fp->f_vnode, &off); 4666#else 4667 zc->zc_cookie, fp, &off); 4668#endif 4669 4670 if (off >= 0 && off <= MAXOFFSET_T) 4671 fp->f_offset = off; 4672 releasef(zc->zc_cookie); 4673 } 4674 return (error); 4675} 4676 4677/* 4678 * inputs: 4679 * zc_name name of snapshot on which to report progress 4680 * zc_cookie file descriptor of send stream 4681 * 4682 * outputs: 4683 * zc_cookie number of bytes written in send stream thus far 4684 */ 4685static int 4686zfs_ioc_send_progress(zfs_cmd_t *zc) 4687{ 4688 dsl_pool_t *dp; 4689 dsl_dataset_t *ds; 4690 dmu_sendarg_t *dsp = NULL; 4691 int error; 4692 4693 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4694 if (error != 0) 4695 return (error); 4696 4697 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds); 4698 if (error != 0) { 4699 dsl_pool_rele(dp, FTAG); 4700 return (error); 4701 } 4702 4703 mutex_enter(&ds->ds_sendstream_lock); 4704 4705 /* 4706 * Iterate over all the send streams currently active on this dataset. 4707 * If there's one which matches the specified file descriptor _and_ the 4708 * stream was started by the current process, return the progress of 4709 * that stream. 4710 */ 4711 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL; 4712 dsp = list_next(&ds->ds_sendstreams, dsp)) { 4713 if (dsp->dsa_outfd == zc->zc_cookie && 4714 dsp->dsa_proc == curproc) 4715 break; 4716 } 4717 4718 if (dsp != NULL) 4719 zc->zc_cookie = *(dsp->dsa_off); 4720 else 4721 error = SET_ERROR(ENOENT); 4722 4723 mutex_exit(&ds->ds_sendstream_lock); 4724 dsl_dataset_rele(ds, FTAG); 4725 dsl_pool_rele(dp, FTAG); 4726 return (error); 4727} 4728 4729static int 4730zfs_ioc_inject_fault(zfs_cmd_t *zc) 4731{ 4732 int id, error; 4733 4734 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id, 4735 &zc->zc_inject_record); 4736 4737 if (error == 0) 4738 zc->zc_guid = (uint64_t)id; 4739 4740 return (error); 4741} 4742 4743static int 4744zfs_ioc_clear_fault(zfs_cmd_t *zc) 4745{ 4746 return (zio_clear_fault((int)zc->zc_guid)); 4747} 4748 4749static int 4750zfs_ioc_inject_list_next(zfs_cmd_t *zc) 4751{ 4752 int id = (int)zc->zc_guid; 4753 int error; 4754 4755 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name), 4756 &zc->zc_inject_record); 4757 4758 zc->zc_guid = id; 4759 4760 return (error); 4761} 4762 4763static int 4764zfs_ioc_error_log(zfs_cmd_t *zc) 4765{ 4766 spa_t *spa; 4767 int error; 4768 size_t count = (size_t)zc->zc_nvlist_dst_size; 4769 4770 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 4771 return (error); 4772 4773 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst, 4774 &count); 4775 if (error == 0) 4776 zc->zc_nvlist_dst_size = count; 4777 else 4778 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa); 4779 4780 spa_close(spa, FTAG); 4781 4782 return (error); 4783} 4784 4785static int 4786zfs_ioc_clear(zfs_cmd_t *zc) 4787{ 4788 spa_t *spa; 4789 vdev_t *vd; 4790 int error; 4791 4792 /* 4793 * On zpool clear we also fix up missing slogs 4794 */ 4795 mutex_enter(&spa_namespace_lock); 4796 spa = spa_lookup(zc->zc_name); 4797 if (spa == NULL) { 4798 mutex_exit(&spa_namespace_lock); 4799 return (SET_ERROR(EIO)); 4800 } 4801 if (spa_get_log_state(spa) == SPA_LOG_MISSING) { 4802 /* we need to let spa_open/spa_load clear the chains */ 4803 spa_set_log_state(spa, SPA_LOG_CLEAR); 4804 } 4805 spa->spa_last_open_failed = 0; 4806 mutex_exit(&spa_namespace_lock); 4807 4808 if (zc->zc_cookie & ZPOOL_NO_REWIND) { 4809 error = spa_open(zc->zc_name, &spa, FTAG); 4810 } else { 4811 nvlist_t *policy; 4812 nvlist_t *config = NULL; 4813 4814 if (zc->zc_nvlist_src == 0) 4815 return (SET_ERROR(EINVAL)); 4816 4817 if ((error = get_nvlist(zc->zc_nvlist_src, 4818 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) { 4819 error = spa_open_rewind(zc->zc_name, &spa, FTAG, 4820 policy, &config); 4821 if (config != NULL) { 4822 int err; 4823 4824 if ((err = put_nvlist(zc, config)) != 0) 4825 error = err; 4826 nvlist_free(config); 4827 } 4828 nvlist_free(policy); 4829 } 4830 } 4831 4832 if (error != 0) 4833 return (error); 4834 4835 spa_vdev_state_enter(spa, SCL_NONE); 4836 4837 if (zc->zc_guid == 0) { 4838 vd = NULL; 4839 } else { 4840 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE); 4841 if (vd == NULL) { 4842 (void) spa_vdev_state_exit(spa, NULL, ENODEV); 4843 spa_close(spa, FTAG); 4844 return (SET_ERROR(ENODEV)); 4845 } 4846 } 4847 4848 vdev_clear(spa, vd); 4849 4850 (void) spa_vdev_state_exit(spa, NULL, 0); 4851 4852 /* 4853 * Resume any suspended I/Os. 4854 */ 4855 if (zio_resume(spa) != 0) 4856 error = SET_ERROR(EIO); 4857 4858 spa_close(spa, FTAG); 4859 4860 return (error); 4861} 4862 4863static int 4864zfs_ioc_pool_reopen(zfs_cmd_t *zc) 4865{ 4866 spa_t *spa; 4867 int error; 4868 4869 error = spa_open(zc->zc_name, &spa, FTAG); 4870 if (error != 0) 4871 return (error); 4872 4873 spa_vdev_state_enter(spa, SCL_NONE); 4874 4875 /* 4876 * If a resilver is already in progress then set the 4877 * spa_scrub_reopen flag to B_TRUE so that we don't restart 4878 * the scan as a side effect of the reopen. Otherwise, let 4879 * vdev_open() decided if a resilver is required. 4880 */ 4881 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool); 4882 vdev_reopen(spa->spa_root_vdev); 4883 spa->spa_scrub_reopen = B_FALSE; 4884 4885 (void) spa_vdev_state_exit(spa, NULL, 0); 4886 spa_close(spa, FTAG); 4887 return (0); 4888} 4889/* 4890 * inputs: 4891 * zc_name name of filesystem 4892 * zc_value name of origin snapshot 4893 * 4894 * outputs: 4895 * zc_string name of conflicting snapshot, if there is one 4896 */ 4897static int 4898zfs_ioc_promote(zfs_cmd_t *zc) 4899{ 4900 char *cp; 4901 4902 /* 4903 * We don't need to unmount *all* the origin fs's snapshots, but 4904 * it's easier. 4905 */ 4906 cp = strchr(zc->zc_value, '@'); 4907 if (cp) 4908 *cp = '\0'; 4909 (void) dmu_objset_find(zc->zc_value, 4910 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS); 4911 return (dsl_dataset_promote(zc->zc_name, zc->zc_string)); 4912} 4913 4914/* 4915 * Retrieve a single {user|group}{used|quota}@... property. 4916 * 4917 * inputs: 4918 * zc_name name of filesystem 4919 * zc_objset_type zfs_userquota_prop_t 4920 * zc_value domain name (eg. "S-1-234-567-89") 4921 * zc_guid RID/UID/GID 4922 * 4923 * outputs: 4924 * zc_cookie property value 4925 */ 4926static int 4927zfs_ioc_userspace_one(zfs_cmd_t *zc) 4928{ 4929 zfsvfs_t *zfsvfs; 4930 int error; 4931 4932 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 4933 return (SET_ERROR(EINVAL)); 4934 4935 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE); 4936 if (error != 0) 4937 return (error); 4938 4939 error = zfs_userspace_one(zfsvfs, 4940 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie); 4941 zfsvfs_rele(zfsvfs, FTAG); 4942 4943 return (error); 4944} 4945 4946/* 4947 * inputs: 4948 * zc_name name of filesystem 4949 * zc_cookie zap cursor 4950 * zc_objset_type zfs_userquota_prop_t 4951 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist) 4952 * 4953 * outputs: 4954 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t) 4955 * zc_cookie zap cursor 4956 */ 4957static int 4958zfs_ioc_userspace_many(zfs_cmd_t *zc) 4959{ 4960 zfsvfs_t *zfsvfs; 4961 int bufsize = zc->zc_nvlist_dst_size; 4962 4963 if (bufsize <= 0) 4964 return (SET_ERROR(ENOMEM)); 4965 4966 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE); 4967 if (error != 0) 4968 return (error); 4969 4970 void *buf = kmem_alloc(bufsize, KM_SLEEP); 4971 4972 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie, 4973 buf, &zc->zc_nvlist_dst_size); 4974 4975 if (error == 0) { 4976 error = ddi_copyout(buf, 4977 (void *)(uintptr_t)zc->zc_nvlist_dst, 4978 zc->zc_nvlist_dst_size, zc->zc_iflags); 4979 } 4980 kmem_free(buf, bufsize); 4981 zfsvfs_rele(zfsvfs, FTAG); 4982 4983 return (error); 4984} 4985 4986/* 4987 * inputs: 4988 * zc_name name of filesystem 4989 * 4990 * outputs: 4991 * none 4992 */ 4993static int 4994zfs_ioc_userspace_upgrade(zfs_cmd_t *zc) 4995{ 4996 objset_t *os; 4997 int error = 0; 4998 zfsvfs_t *zfsvfs; 4999 5000 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) { 5001 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) { 5002 /* 5003 * If userused is not enabled, it may be because the 5004 * objset needs to be closed & reopened (to grow the 5005 * objset_phys_t). Suspend/resume the fs will do that. 5006 */ 5007 dsl_dataset_t *ds; 5008 5009 ds = dmu_objset_ds(zfsvfs->z_os); 5010 error = zfs_suspend_fs(zfsvfs); 5011 if (error == 0) { 5012 dmu_objset_refresh_ownership(zfsvfs->z_os, 5013 zfsvfs); 5014 error = zfs_resume_fs(zfsvfs, ds); 5015 } 5016 } 5017 if (error == 0) 5018 error = dmu_objset_userspace_upgrade(zfsvfs->z_os); 5019#ifdef illumos 5020 VFS_RELE(zfsvfs->z_vfs); 5021#else 5022 vfs_unbusy(zfsvfs->z_vfs); 5023#endif 5024 } else { 5025 /* XXX kind of reading contents without owning */ 5026 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 5027 if (error != 0) 5028 return (error); 5029 5030 error = dmu_objset_userspace_upgrade(os); 5031 dmu_objset_rele(os, FTAG); 5032 } 5033 5034 return (error); 5035} 5036 5037#ifdef illumos 5038/* 5039 * We don't want to have a hard dependency 5040 * against some special symbols in sharefs 5041 * nfs, and smbsrv. Determine them if needed when 5042 * the first file system is shared. 5043 * Neither sharefs, nfs or smbsrv are unloadable modules. 5044 */ 5045int (*znfsexport_fs)(void *arg); 5046int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t); 5047int (*zsmbexport_fs)(void *arg, boolean_t add_share); 5048 5049int zfs_nfsshare_inited; 5050int zfs_smbshare_inited; 5051 5052ddi_modhandle_t nfs_mod; 5053ddi_modhandle_t sharefs_mod; 5054ddi_modhandle_t smbsrv_mod; 5055#endif /* illumos */ 5056kmutex_t zfs_share_lock; 5057 5058#ifdef illumos 5059static int 5060zfs_init_sharefs() 5061{ 5062 int error; 5063 5064 ASSERT(MUTEX_HELD(&zfs_share_lock)); 5065 /* Both NFS and SMB shares also require sharetab support. */ 5066 if (sharefs_mod == NULL && ((sharefs_mod = 5067 ddi_modopen("fs/sharefs", 5068 KRTLD_MODE_FIRST, &error)) == NULL)) { 5069 return (SET_ERROR(ENOSYS)); 5070 } 5071 if (zshare_fs == NULL && ((zshare_fs = 5072 (int (*)(enum sharefs_sys_op, share_t *, uint32_t)) 5073 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) { 5074 return (SET_ERROR(ENOSYS)); 5075 } 5076 return (0); 5077} 5078#endif /* illumos */ 5079 5080static int 5081zfs_ioc_share(zfs_cmd_t *zc) 5082{ 5083#ifdef illumos 5084 int error; 5085 int opcode; 5086 5087 switch (zc->zc_share.z_sharetype) { 5088 case ZFS_SHARE_NFS: 5089 case ZFS_UNSHARE_NFS: 5090 if (zfs_nfsshare_inited == 0) { 5091 mutex_enter(&zfs_share_lock); 5092 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs", 5093 KRTLD_MODE_FIRST, &error)) == NULL)) { 5094 mutex_exit(&zfs_share_lock); 5095 return (SET_ERROR(ENOSYS)); 5096 } 5097 if (znfsexport_fs == NULL && 5098 ((znfsexport_fs = (int (*)(void *)) 5099 ddi_modsym(nfs_mod, 5100 "nfs_export", &error)) == NULL)) { 5101 mutex_exit(&zfs_share_lock); 5102 return (SET_ERROR(ENOSYS)); 5103 } 5104 error = zfs_init_sharefs(); 5105 if (error != 0) { 5106 mutex_exit(&zfs_share_lock); 5107 return (SET_ERROR(ENOSYS)); 5108 } 5109 zfs_nfsshare_inited = 1; 5110 mutex_exit(&zfs_share_lock); 5111 } 5112 break; 5113 case ZFS_SHARE_SMB: 5114 case ZFS_UNSHARE_SMB: 5115 if (zfs_smbshare_inited == 0) { 5116 mutex_enter(&zfs_share_lock); 5117 if (smbsrv_mod == NULL && ((smbsrv_mod = 5118 ddi_modopen("drv/smbsrv", 5119 KRTLD_MODE_FIRST, &error)) == NULL)) { 5120 mutex_exit(&zfs_share_lock); 5121 return (SET_ERROR(ENOSYS)); 5122 } 5123 if (zsmbexport_fs == NULL && ((zsmbexport_fs = 5124 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod, 5125 "smb_server_share", &error)) == NULL)) { 5126 mutex_exit(&zfs_share_lock); 5127 return (SET_ERROR(ENOSYS)); 5128 } 5129 error = zfs_init_sharefs(); 5130 if (error != 0) { 5131 mutex_exit(&zfs_share_lock); 5132 return (SET_ERROR(ENOSYS)); 5133 } 5134 zfs_smbshare_inited = 1; 5135 mutex_exit(&zfs_share_lock); 5136 } 5137 break; 5138 default: 5139 return (SET_ERROR(EINVAL)); 5140 } 5141 5142 switch (zc->zc_share.z_sharetype) { 5143 case ZFS_SHARE_NFS: 5144 case ZFS_UNSHARE_NFS: 5145 if (error = 5146 znfsexport_fs((void *) 5147 (uintptr_t)zc->zc_share.z_exportdata)) 5148 return (error); 5149 break; 5150 case ZFS_SHARE_SMB: 5151 case ZFS_UNSHARE_SMB: 5152 if (error = zsmbexport_fs((void *) 5153 (uintptr_t)zc->zc_share.z_exportdata, 5154 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ? 5155 B_TRUE: B_FALSE)) { 5156 return (error); 5157 } 5158 break; 5159 } 5160 5161 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS || 5162 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ? 5163 SHAREFS_ADD : SHAREFS_REMOVE; 5164 5165 /* 5166 * Add or remove share from sharetab 5167 */ 5168 error = zshare_fs(opcode, 5169 (void *)(uintptr_t)zc->zc_share.z_sharedata, 5170 zc->zc_share.z_sharemax); 5171 5172 return (error); 5173 5174#else /* !illumos */ 5175 return (ENOSYS); 5176#endif /* illumos */ 5177} 5178 5179ace_t full_access[] = { 5180 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0} 5181}; 5182 5183/* 5184 * inputs: 5185 * zc_name name of containing filesystem 5186 * zc_obj object # beyond which we want next in-use object # 5187 * 5188 * outputs: 5189 * zc_obj next in-use object # 5190 */ 5191static int 5192zfs_ioc_next_obj(zfs_cmd_t *zc) 5193{ 5194 objset_t *os = NULL; 5195 int error; 5196 5197 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 5198 if (error != 0) 5199 return (error); 5200 5201 error = dmu_object_next(os, &zc->zc_obj, B_FALSE, 5202 dsl_dataset_phys(os->os_dsl_dataset)->ds_prev_snap_txg); 5203 5204 dmu_objset_rele(os, FTAG); 5205 return (error); 5206} 5207 5208/* 5209 * inputs: 5210 * zc_name name of filesystem 5211 * zc_value prefix name for snapshot 5212 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process 5213 * 5214 * outputs: 5215 * zc_value short name of new snapshot 5216 */ 5217static int 5218zfs_ioc_tmp_snapshot(zfs_cmd_t *zc) 5219{ 5220 char *snap_name; 5221 char *hold_name; 5222 int error; 5223 minor_t minor; 5224 5225 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor); 5226 if (error != 0) 5227 return (error); 5228 5229 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value, 5230 (u_longlong_t)ddi_get_lbolt64()); 5231 hold_name = kmem_asprintf("%%%s", zc->zc_value); 5232 5233 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor, 5234 hold_name); 5235 if (error == 0) 5236 (void) strcpy(zc->zc_value, snap_name); 5237 strfree(snap_name); 5238 strfree(hold_name); 5239 zfs_onexit_fd_rele(zc->zc_cleanup_fd); 5240 return (error); 5241} 5242 5243/* 5244 * inputs: 5245 * zc_name name of "to" snapshot 5246 * zc_value name of "from" snapshot 5247 * zc_cookie file descriptor to write diff data on 5248 * 5249 * outputs: 5250 * dmu_diff_record_t's to the file descriptor 5251 */ 5252static int 5253zfs_ioc_diff(zfs_cmd_t *zc) 5254{ 5255 file_t *fp; 5256 cap_rights_t rights; 5257 offset_t off; 5258 int error; 5259 5260#ifdef illumos 5261 fp = getf(zc->zc_cookie); 5262#else 5263 fget_write(curthread, zc->zc_cookie, 5264 cap_rights_init(&rights, CAP_WRITE), &fp); 5265#endif 5266 if (fp == NULL) 5267 return (SET_ERROR(EBADF)); 5268 5269 off = fp->f_offset; 5270 5271#ifdef illumos 5272 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off); 5273#else 5274 error = dmu_diff(zc->zc_name, zc->zc_value, fp, &off); 5275#endif 5276 5277 if (off >= 0 && off <= MAXOFFSET_T) 5278 fp->f_offset = off; 5279 releasef(zc->zc_cookie); 5280 5281 return (error); 5282} 5283 5284#ifdef illumos 5285/* 5286 * Remove all ACL files in shares dir 5287 */ 5288static int 5289zfs_smb_acl_purge(znode_t *dzp) 5290{ 5291 zap_cursor_t zc; 5292 zap_attribute_t zap; 5293 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 5294 int error; 5295 5296 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id); 5297 (error = zap_cursor_retrieve(&zc, &zap)) == 0; 5298 zap_cursor_advance(&zc)) { 5299 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred, 5300 NULL, 0)) != 0) 5301 break; 5302 } 5303 zap_cursor_fini(&zc); 5304 return (error); 5305} 5306#endif /* illumos */ 5307 5308static int 5309zfs_ioc_smb_acl(zfs_cmd_t *zc) 5310{ 5311#ifdef illumos 5312 vnode_t *vp; 5313 znode_t *dzp; 5314 vnode_t *resourcevp = NULL; 5315 znode_t *sharedir; 5316 zfsvfs_t *zfsvfs; 5317 nvlist_t *nvlist; 5318 char *src, *target; 5319 vattr_t vattr; 5320 vsecattr_t vsec; 5321 int error = 0; 5322 5323 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, 5324 NO_FOLLOW, NULL, &vp)) != 0) 5325 return (error); 5326 5327 /* Now make sure mntpnt and dataset are ZFS */ 5328 5329 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 || 5330 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), 5331 zc->zc_name) != 0)) { 5332 VN_RELE(vp); 5333 return (SET_ERROR(EINVAL)); 5334 } 5335 5336 dzp = VTOZ(vp); 5337 zfsvfs = dzp->z_zfsvfs; 5338 ZFS_ENTER(zfsvfs); 5339 5340 /* 5341 * Create share dir if its missing. 5342 */ 5343 mutex_enter(&zfsvfs->z_lock); 5344 if (zfsvfs->z_shares_dir == 0) { 5345 dmu_tx_t *tx; 5346 5347 tx = dmu_tx_create(zfsvfs->z_os); 5348 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE, 5349 ZFS_SHARES_DIR); 5350 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); 5351 error = dmu_tx_assign(tx, TXG_WAIT); 5352 if (error != 0) { 5353 dmu_tx_abort(tx); 5354 } else { 5355 error = zfs_create_share_dir(zfsvfs, tx); 5356 dmu_tx_commit(tx); 5357 } 5358 if (error != 0) { 5359 mutex_exit(&zfsvfs->z_lock); 5360 VN_RELE(vp); 5361 ZFS_EXIT(zfsvfs); 5362 return (error); 5363 } 5364 } 5365 mutex_exit(&zfsvfs->z_lock); 5366 5367 ASSERT(zfsvfs->z_shares_dir); 5368 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) { 5369 VN_RELE(vp); 5370 ZFS_EXIT(zfsvfs); 5371 return (error); 5372 } 5373 5374 switch (zc->zc_cookie) { 5375 case ZFS_SMB_ACL_ADD: 5376 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; 5377 vattr.va_type = VREG; 5378 vattr.va_mode = S_IFREG|0777; 5379 vattr.va_uid = 0; 5380 vattr.va_gid = 0; 5381 5382 vsec.vsa_mask = VSA_ACE; 5383 vsec.vsa_aclentp = &full_access; 5384 vsec.vsa_aclentsz = sizeof (full_access); 5385 vsec.vsa_aclcnt = 1; 5386 5387 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string, 5388 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec); 5389 if (resourcevp) 5390 VN_RELE(resourcevp); 5391 break; 5392 5393 case ZFS_SMB_ACL_REMOVE: 5394 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred, 5395 NULL, 0); 5396 break; 5397 5398 case ZFS_SMB_ACL_RENAME: 5399 if ((error = get_nvlist(zc->zc_nvlist_src, 5400 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) { 5401 VN_RELE(vp); 5402 VN_RELE(ZTOV(sharedir)); 5403 ZFS_EXIT(zfsvfs); 5404 return (error); 5405 } 5406 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) || 5407 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET, 5408 &target)) { 5409 VN_RELE(vp); 5410 VN_RELE(ZTOV(sharedir)); 5411 ZFS_EXIT(zfsvfs); 5412 nvlist_free(nvlist); 5413 return (error); 5414 } 5415 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target, 5416 kcred, NULL, 0); 5417 nvlist_free(nvlist); 5418 break; 5419 5420 case ZFS_SMB_ACL_PURGE: 5421 error = zfs_smb_acl_purge(sharedir); 5422 break; 5423 5424 default: 5425 error = SET_ERROR(EINVAL); 5426 break; 5427 } 5428 5429 VN_RELE(vp); 5430 VN_RELE(ZTOV(sharedir)); 5431 5432 ZFS_EXIT(zfsvfs); 5433 5434 return (error); 5435#else /* !illumos */ 5436 return (EOPNOTSUPP); 5437#endif /* illumos */ 5438} 5439 5440/* 5441 * innvl: { 5442 * "holds" -> { snapname -> holdname (string), ... } 5443 * (optional) "cleanup_fd" -> fd (int32) 5444 * } 5445 * 5446 * outnvl: { 5447 * snapname -> error value (int32) 5448 * ... 5449 * } 5450 */ 5451/* ARGSUSED */ 5452static int 5453zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist) 5454{ 5455 nvpair_t *pair; 5456 nvlist_t *holds; 5457 int cleanup_fd = -1; 5458 int error; 5459 minor_t minor = 0; 5460 5461 error = nvlist_lookup_nvlist(args, "holds", &holds); 5462 if (error != 0) 5463 return (SET_ERROR(EINVAL)); 5464 5465 /* make sure the user didn't pass us any invalid (empty) tags */ 5466 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL; 5467 pair = nvlist_next_nvpair(holds, pair)) { 5468 char *htag; 5469 5470 error = nvpair_value_string(pair, &htag); 5471 if (error != 0) 5472 return (SET_ERROR(error)); 5473 5474 if (strlen(htag) == 0) 5475 return (SET_ERROR(EINVAL)); 5476 } 5477 5478 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) { 5479 error = zfs_onexit_fd_hold(cleanup_fd, &minor); 5480 if (error != 0) 5481 return (error); 5482 } 5483 5484 error = dsl_dataset_user_hold(holds, minor, errlist); 5485 if (minor != 0) 5486 zfs_onexit_fd_rele(cleanup_fd); 5487 return (error); 5488} 5489 5490/* 5491 * innvl is not used. 5492 * 5493 * outnvl: { 5494 * holdname -> time added (uint64 seconds since epoch) 5495 * ... 5496 * } 5497 */ 5498/* ARGSUSED */ 5499static int 5500zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl) 5501{ 5502 return (dsl_dataset_get_holds(snapname, outnvl)); 5503} 5504 5505/* 5506 * innvl: { 5507 * snapname -> { holdname, ... } 5508 * ... 5509 * } 5510 * 5511 * outnvl: { 5512 * snapname -> error value (int32) 5513 * ... 5514 * } 5515 */ 5516/* ARGSUSED */ 5517static int 5518zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist) 5519{ 5520 return (dsl_dataset_user_release(holds, errlist)); 5521} 5522 5523/* 5524 * inputs: 5525 * zc_name name of new filesystem or snapshot 5526 * zc_value full name of old snapshot 5527 * 5528 * outputs: 5529 * zc_cookie space in bytes 5530 * zc_objset_type compressed space in bytes 5531 * zc_perm_action uncompressed space in bytes 5532 */ 5533static int 5534zfs_ioc_space_written(zfs_cmd_t *zc) 5535{ 5536 int error; 5537 dsl_pool_t *dp; 5538 dsl_dataset_t *new, *old; 5539 5540 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 5541 if (error != 0) 5542 return (error); 5543 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new); 5544 if (error != 0) { 5545 dsl_pool_rele(dp, FTAG); 5546 return (error); 5547 } 5548 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old); 5549 if (error != 0) { 5550 dsl_dataset_rele(new, FTAG); 5551 dsl_pool_rele(dp, FTAG); 5552 return (error); 5553 } 5554 5555 error = dsl_dataset_space_written(old, new, &zc->zc_cookie, 5556 &zc->zc_objset_type, &zc->zc_perm_action); 5557 dsl_dataset_rele(old, FTAG); 5558 dsl_dataset_rele(new, FTAG); 5559 dsl_pool_rele(dp, FTAG); 5560 return (error); 5561} 5562 5563/* 5564 * innvl: { 5565 * "firstsnap" -> snapshot name 5566 * } 5567 * 5568 * outnvl: { 5569 * "used" -> space in bytes 5570 * "compressed" -> compressed space in bytes 5571 * "uncompressed" -> uncompressed space in bytes 5572 * } 5573 */ 5574static int 5575zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl) 5576{ 5577 int error; 5578 dsl_pool_t *dp; 5579 dsl_dataset_t *new, *old; 5580 char *firstsnap; 5581 uint64_t used, comp, uncomp; 5582 5583 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0) 5584 return (SET_ERROR(EINVAL)); 5585 5586 error = dsl_pool_hold(lastsnap, FTAG, &dp); 5587 if (error != 0) 5588 return (error); 5589 5590 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new); 5591 if (error == 0 && !new->ds_is_snapshot) { 5592 dsl_dataset_rele(new, FTAG); 5593 error = SET_ERROR(EINVAL); 5594 } 5595 if (error != 0) { 5596 dsl_pool_rele(dp, FTAG); 5597 return (error); 5598 } 5599 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old); 5600 if (error == 0 && !old->ds_is_snapshot) { 5601 dsl_dataset_rele(old, FTAG); 5602 error = SET_ERROR(EINVAL); 5603 } 5604 if (error != 0) { 5605 dsl_dataset_rele(new, FTAG); 5606 dsl_pool_rele(dp, FTAG); 5607 return (error); 5608 } 5609 5610 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp); 5611 dsl_dataset_rele(old, FTAG); 5612 dsl_dataset_rele(new, FTAG); 5613 dsl_pool_rele(dp, FTAG); 5614 fnvlist_add_uint64(outnvl, "used", used); 5615 fnvlist_add_uint64(outnvl, "compressed", comp); 5616 fnvlist_add_uint64(outnvl, "uncompressed", uncomp); 5617 return (error); 5618} 5619 5620static int 5621zfs_ioc_jail(zfs_cmd_t *zc) 5622{ 5623 5624 return (zone_dataset_attach(curthread->td_ucred, zc->zc_name, 5625 (int)zc->zc_jailid)); 5626} 5627 5628static int 5629zfs_ioc_unjail(zfs_cmd_t *zc) 5630{ 5631 5632 return (zone_dataset_detach(curthread->td_ucred, zc->zc_name, 5633 (int)zc->zc_jailid)); 5634} 5635 5636/* 5637 * innvl: { 5638 * "fd" -> file descriptor to write stream to (int32) 5639 * (optional) "fromsnap" -> full snap name to send an incremental from 5640 * (optional) "largeblockok" -> (value ignored) 5641 * indicates that blocks > 128KB are permitted 5642 * (optional) "embedok" -> (value ignored) 5643 * presence indicates DRR_WRITE_EMBEDDED records are permitted 5644 * (optional) "resume_object" and "resume_offset" -> (uint64) 5645 * if present, resume send stream from specified object and offset. 5646 * } 5647 * 5648 * outnvl is unused 5649 */ 5650/* ARGSUSED */ 5651static int 5652zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl) 5653{ 5654 cap_rights_t rights; 5655 file_t *fp; 5656 int error; 5657 offset_t off; 5658 char *fromname = NULL; 5659 int fd; 5660 boolean_t largeblockok; 5661 boolean_t embedok; 5662 uint64_t resumeobj = 0; 5663 uint64_t resumeoff = 0; 5664 5665 error = nvlist_lookup_int32(innvl, "fd", &fd); 5666 if (error != 0) 5667 return (SET_ERROR(EINVAL)); 5668 5669 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname); 5670 5671 largeblockok = nvlist_exists(innvl, "largeblockok"); 5672 embedok = nvlist_exists(innvl, "embedok"); 5673 5674 (void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj); 5675 (void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff); 5676 5677#ifdef illumos 5678 file_t *fp = getf(fd); 5679#else 5680 fget_write(curthread, fd, cap_rights_init(&rights, CAP_WRITE), &fp); 5681#endif 5682 if (fp == NULL) 5683 return (SET_ERROR(EBADF)); 5684 5685 off = fp->f_offset; 5686 error = dmu_send(snapname, fromname, embedok, largeblockok, fd, 5687#ifdef illumos 5688 resumeobj, resumeoff, fp->f_vnode, &off); 5689#else 5690 resumeobj, resumeoff, fp, &off); 5691#endif 5692 5693#ifdef illumos 5694 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0) 5695 fp->f_offset = off; 5696#else 5697 fp->f_offset = off; 5698#endif 5699 5700 releasef(fd); 5701 return (error); 5702} 5703 5704/* 5705 * Determine approximately how large a zfs send stream will be -- the number 5706 * of bytes that will be written to the fd supplied to zfs_ioc_send_new(). 5707 * 5708 * innvl: { 5709 * (optional) "from" -> full snap or bookmark name to send an incremental 5710 * from 5711 * } 5712 * 5713 * outnvl: { 5714 * "space" -> bytes of space (uint64) 5715 * } 5716 */ 5717static int 5718zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl) 5719{ 5720 dsl_pool_t *dp; 5721 dsl_dataset_t *tosnap; 5722 int error; 5723 char *fromname; 5724 uint64_t space; 5725 5726 error = dsl_pool_hold(snapname, FTAG, &dp); 5727 if (error != 0) 5728 return (error); 5729 5730 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap); 5731 if (error != 0) { 5732 dsl_pool_rele(dp, FTAG); 5733 return (error); 5734 } 5735 5736 error = nvlist_lookup_string(innvl, "from", &fromname); 5737 if (error == 0) { 5738 if (strchr(fromname, '@') != NULL) { 5739 /* 5740 * If from is a snapshot, hold it and use the more 5741 * efficient dmu_send_estimate to estimate send space 5742 * size using deadlists. 5743 */ 5744 dsl_dataset_t *fromsnap; 5745 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap); 5746 if (error != 0) 5747 goto out; 5748 error = dmu_send_estimate(tosnap, fromsnap, &space); 5749 dsl_dataset_rele(fromsnap, FTAG); 5750 } else if (strchr(fromname, '#') != NULL) { 5751 /* 5752 * If from is a bookmark, fetch the creation TXG of the 5753 * snapshot it was created from and use that to find 5754 * blocks that were born after it. 5755 */ 5756 zfs_bookmark_phys_t frombm; 5757 5758 error = dsl_bookmark_lookup(dp, fromname, tosnap, 5759 &frombm); 5760 if (error != 0) 5761 goto out; 5762 error = dmu_send_estimate_from_txg(tosnap, 5763 frombm.zbm_creation_txg, &space); 5764 } else { 5765 /* 5766 * from is not properly formatted as a snapshot or 5767 * bookmark 5768 */ 5769 error = SET_ERROR(EINVAL); 5770 goto out; 5771 } 5772 } else { 5773 // If estimating the size of a full send, use dmu_send_estimate 5774 error = dmu_send_estimate(tosnap, NULL, &space); 5775 } 5776 5777 fnvlist_add_uint64(outnvl, "space", space); 5778 5779out: 5780 dsl_dataset_rele(tosnap, FTAG); 5781 dsl_pool_rele(dp, FTAG); 5782 return (error); 5783} 5784 5785static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST]; 5786 5787static void 5788zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5789 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck, 5790 boolean_t log_history, zfs_ioc_poolcheck_t pool_check) 5791{ 5792 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST]; 5793 5794 ASSERT3U(ioc, >=, ZFS_IOC_FIRST); 5795 ASSERT3U(ioc, <, ZFS_IOC_LAST); 5796 ASSERT3P(vec->zvec_legacy_func, ==, NULL); 5797 ASSERT3P(vec->zvec_func, ==, NULL); 5798 5799 vec->zvec_legacy_func = func; 5800 vec->zvec_secpolicy = secpolicy; 5801 vec->zvec_namecheck = namecheck; 5802 vec->zvec_allow_log = log_history; 5803 vec->zvec_pool_check = pool_check; 5804} 5805 5806/* 5807 * See the block comment at the beginning of this file for details on 5808 * each argument to this function. 5809 */ 5810static void 5811zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func, 5812 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck, 5813 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist, 5814 boolean_t allow_log) 5815{ 5816 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST]; 5817 5818 ASSERT3U(ioc, >=, ZFS_IOC_FIRST); 5819 ASSERT3U(ioc, <, ZFS_IOC_LAST); 5820 ASSERT3P(vec->zvec_legacy_func, ==, NULL); 5821 ASSERT3P(vec->zvec_func, ==, NULL); 5822 5823 /* if we are logging, the name must be valid */ 5824 ASSERT(!allow_log || namecheck != NO_NAME); 5825 5826 vec->zvec_name = name; 5827 vec->zvec_func = func; 5828 vec->zvec_secpolicy = secpolicy; 5829 vec->zvec_namecheck = namecheck; 5830 vec->zvec_pool_check = pool_check; 5831 vec->zvec_smush_outnvlist = smush_outnvlist; 5832 vec->zvec_allow_log = allow_log; 5833} 5834 5835static void 5836zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5837 zfs_secpolicy_func_t *secpolicy, boolean_t log_history, 5838 zfs_ioc_poolcheck_t pool_check) 5839{ 5840 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5841 POOL_NAME, log_history, pool_check); 5842} 5843 5844static void 5845zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5846 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check) 5847{ 5848 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5849 DATASET_NAME, B_FALSE, pool_check); 5850} 5851 5852static void 5853zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func) 5854{ 5855 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config, 5856 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5857} 5858 5859static void 5860zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5861 zfs_secpolicy_func_t *secpolicy) 5862{ 5863 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5864 NO_NAME, B_FALSE, POOL_CHECK_NONE); 5865} 5866 5867static void 5868zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc, 5869 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy) 5870{ 5871 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5872 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED); 5873} 5874 5875static void 5876zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func) 5877{ 5878 zfs_ioctl_register_dataset_read_secpolicy(ioc, func, 5879 zfs_secpolicy_read); 5880} 5881 5882static void 5883zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5884 zfs_secpolicy_func_t *secpolicy) 5885{ 5886 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5887 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5888} 5889 5890static void 5891zfs_ioctl_init(void) 5892{ 5893 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT, 5894 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME, 5895 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5896 5897 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY, 5898 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME, 5899 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE); 5900 5901 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS, 5902 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME, 5903 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5904 5905 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW, 5906 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME, 5907 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5908 5909 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE, 5910 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME, 5911 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5912 5913 zfs_ioctl_register("create", ZFS_IOC_CREATE, 5914 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME, 5915 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5916 5917 zfs_ioctl_register("clone", ZFS_IOC_CLONE, 5918 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME, 5919 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5920 5921 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS, 5922 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME, 5923 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5924 5925 zfs_ioctl_register("hold", ZFS_IOC_HOLD, 5926 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME, 5927 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5928 zfs_ioctl_register("release", ZFS_IOC_RELEASE, 5929 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME, 5930 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5931 5932 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS, 5933 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME, 5934 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5935 5936 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK, 5937 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, 5938 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE); 5939 5940 zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK, 5941 zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME, 5942 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5943 5944 zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS, 5945 zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME, 5946 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5947 5948 zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS, 5949 zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks, 5950 POOL_NAME, 5951 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5952 5953 /* IOCTLS that use the legacy function signature */ 5954 5955 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze, 5956 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY); 5957 5958 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create, 5959 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5960 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN, 5961 zfs_ioc_pool_scan); 5962 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE, 5963 zfs_ioc_pool_upgrade); 5964 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD, 5965 zfs_ioc_vdev_add); 5966 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE, 5967 zfs_ioc_vdev_remove); 5968 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE, 5969 zfs_ioc_vdev_set_state); 5970 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH, 5971 zfs_ioc_vdev_attach); 5972 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH, 5973 zfs_ioc_vdev_detach); 5974 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH, 5975 zfs_ioc_vdev_setpath); 5976 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU, 5977 zfs_ioc_vdev_setfru); 5978 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS, 5979 zfs_ioc_pool_set_props); 5980 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT, 5981 zfs_ioc_vdev_split); 5982 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID, 5983 zfs_ioc_pool_reguid); 5984 5985 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS, 5986 zfs_ioc_pool_configs, zfs_secpolicy_none); 5987 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT, 5988 zfs_ioc_pool_tryimport, zfs_secpolicy_config); 5989 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT, 5990 zfs_ioc_inject_fault, zfs_secpolicy_inject); 5991 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT, 5992 zfs_ioc_clear_fault, zfs_secpolicy_inject); 5993 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT, 5994 zfs_ioc_inject_list_next, zfs_secpolicy_inject); 5995 5996 /* 5997 * pool destroy, and export don't log the history as part of 5998 * zfsdev_ioctl, but rather zfs_ioc_pool_export 5999 * does the logging of those commands. 6000 */ 6001 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy, 6002 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE); 6003 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export, 6004 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE); 6005 6006 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats, 6007 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE); 6008 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props, 6009 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE); 6010 6011 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log, 6012 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_NONE); 6013 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME, 6014 zfs_ioc_dsobj_to_dsname, 6015 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_NONE); 6016 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY, 6017 zfs_ioc_pool_get_history, 6018 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED); 6019 6020 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import, 6021 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 6022 6023 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear, 6024 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 6025 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen, 6026 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED); 6027 6028 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN, 6029 zfs_ioc_space_written); 6030 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS, 6031 zfs_ioc_objset_recvd_props); 6032 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ, 6033 zfs_ioc_next_obj); 6034 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL, 6035 zfs_ioc_get_fsacl); 6036 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS, 6037 zfs_ioc_objset_stats); 6038 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS, 6039 zfs_ioc_objset_zplprops); 6040 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT, 6041 zfs_ioc_dataset_list_next); 6042 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT, 6043 zfs_ioc_snapshot_list_next); 6044 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS, 6045 zfs_ioc_send_progress); 6046 6047 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF, 6048 zfs_ioc_diff, zfs_secpolicy_diff); 6049 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS, 6050 zfs_ioc_obj_to_stats, zfs_secpolicy_diff); 6051 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH, 6052 zfs_ioc_obj_to_path, zfs_secpolicy_diff); 6053 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE, 6054 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one); 6055 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY, 6056 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many); 6057 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND, 6058 zfs_ioc_send, zfs_secpolicy_send); 6059 6060 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop, 6061 zfs_secpolicy_none); 6062 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy, 6063 zfs_secpolicy_destroy); 6064 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename, 6065 zfs_secpolicy_rename); 6066 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv, 6067 zfs_secpolicy_recv); 6068 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote, 6069 zfs_secpolicy_promote); 6070 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP, 6071 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop); 6072 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl, 6073 zfs_secpolicy_set_fsacl); 6074 6075 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share, 6076 zfs_secpolicy_share, POOL_CHECK_NONE); 6077 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl, 6078 zfs_secpolicy_smb_acl, POOL_CHECK_NONE); 6079 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE, 6080 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade, 6081 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 6082 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT, 6083 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot, 6084 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 6085 6086#ifdef __FreeBSD__ 6087 zfs_ioctl_register_dataset_nolog(ZFS_IOC_JAIL, zfs_ioc_jail, 6088 zfs_secpolicy_config, POOL_CHECK_NONE); 6089 zfs_ioctl_register_dataset_nolog(ZFS_IOC_UNJAIL, zfs_ioc_unjail, 6090 zfs_secpolicy_config, POOL_CHECK_NONE); 6091 zfs_ioctl_register("fbsd_nextboot", ZFS_IOC_NEXTBOOT, 6092 zfs_ioc_nextboot, zfs_secpolicy_config, NO_NAME, 6093 POOL_CHECK_NONE, B_FALSE, B_FALSE); 6094#endif 6095} 6096 6097int 6098pool_status_check(const char *name, zfs_ioc_namecheck_t type, 6099 zfs_ioc_poolcheck_t check) 6100{ 6101 spa_t *spa; 6102 int error; 6103 6104 ASSERT(type == POOL_NAME || type == DATASET_NAME); 6105 6106 if (check & POOL_CHECK_NONE) 6107 return (0); 6108 6109 error = spa_open(name, &spa, FTAG); 6110 if (error == 0) { 6111 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa)) 6112 error = SET_ERROR(EAGAIN); 6113 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa)) 6114 error = SET_ERROR(EROFS); 6115 spa_close(spa, FTAG); 6116 } 6117 return (error); 6118} 6119 6120/* 6121 * Find a free minor number. 6122 */ 6123minor_t 6124zfsdev_minor_alloc(void) 6125{ 6126 static minor_t last_minor; 6127 minor_t m; 6128 6129 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 6130 6131 for (m = last_minor + 1; m != last_minor; m++) { 6132 if (m > ZFSDEV_MAX_MINOR) 6133 m = 1; 6134 if (ddi_get_soft_state(zfsdev_state, m) == NULL) { 6135 last_minor = m; 6136 return (m); 6137 } 6138 } 6139 6140 return (0); 6141} 6142 6143static int 6144zfs_ctldev_init(struct cdev *devp) 6145{ 6146 minor_t minor; 6147 zfs_soft_state_t *zs; 6148 6149 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 6150 6151 minor = zfsdev_minor_alloc(); 6152 if (minor == 0) 6153 return (SET_ERROR(ENXIO)); 6154 6155 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS) 6156 return (SET_ERROR(EAGAIN)); 6157 6158 devfs_set_cdevpriv((void *)(uintptr_t)minor, zfsdev_close); 6159 6160 zs = ddi_get_soft_state(zfsdev_state, minor); 6161 zs->zss_type = ZSST_CTLDEV; 6162 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data); 6163 6164 return (0); 6165} 6166 6167static void 6168zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor) 6169{ 6170 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 6171 6172 zfs_onexit_destroy(zo); 6173 ddi_soft_state_free(zfsdev_state, minor); 6174} 6175 6176void * 6177zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which) 6178{ 6179 zfs_soft_state_t *zp; 6180 6181 zp = ddi_get_soft_state(zfsdev_state, minor); 6182 if (zp == NULL || zp->zss_type != which) 6183 return (NULL); 6184 6185 return (zp->zss_data); 6186} 6187 6188static int 6189zfsdev_open(struct cdev *devp, int flag, int mode, struct thread *td) 6190{ 6191 int error = 0; 6192 6193#ifdef illumos 6194 if (getminor(*devp) != 0) 6195 return (zvol_open(devp, flag, otyp, cr)); 6196#endif 6197 6198 /* This is the control device. Allocate a new minor if requested. */ 6199 if (flag & FEXCL) { 6200 mutex_enter(&spa_namespace_lock); 6201 error = zfs_ctldev_init(devp); 6202 mutex_exit(&spa_namespace_lock); 6203 } 6204 6205 return (error); 6206} 6207 6208static void 6209zfsdev_close(void *data) 6210{ 6211 zfs_onexit_t *zo; 6212 minor_t minor = (minor_t)(uintptr_t)data; 6213 6214 if (minor == 0) 6215 return; 6216 6217 mutex_enter(&spa_namespace_lock); 6218 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV); 6219 if (zo == NULL) { 6220 mutex_exit(&spa_namespace_lock); 6221 return; 6222 } 6223 zfs_ctldev_destroy(zo, minor); 6224 mutex_exit(&spa_namespace_lock); 6225} 6226 6227static int 6228zfsdev_ioctl(struct cdev *dev, u_long zcmd, caddr_t arg, int flag, 6229 struct thread *td) 6230{ 6231 zfs_cmd_t *zc; 6232 uint_t vecnum; 6233 int error, rc, len; 6234#ifdef illumos 6235 minor_t minor = getminor(dev); 6236#else 6237 zfs_iocparm_t *zc_iocparm; 6238 int cflag, cmd, oldvecnum; 6239 boolean_t newioc, compat; 6240 void *compat_zc = NULL; 6241 cred_t *cr = td->td_ucred; 6242#endif 6243 const zfs_ioc_vec_t *vec; 6244 char *saved_poolname = NULL; 6245 nvlist_t *innvl = NULL; 6246 6247 cflag = ZFS_CMD_COMPAT_NONE; 6248 compat = B_FALSE; 6249 newioc = B_TRUE; /* "new" style (zfs_iocparm_t) ioctl */ 6250 6251 len = IOCPARM_LEN(zcmd); 6252 vecnum = cmd = zcmd & 0xff; 6253 6254 /* 6255 * Check if we are talking to supported older binaries 6256 * and translate zfs_cmd if necessary 6257 */ 6258 if (len != sizeof(zfs_iocparm_t)) { 6259 newioc = B_FALSE; 6260 compat = B_TRUE; 6261 6262 vecnum = cmd; 6263 6264 switch (len) { 6265 case sizeof(zfs_cmd_zcmd_t): 6266 cflag = ZFS_CMD_COMPAT_LZC; 6267 break; 6268 case sizeof(zfs_cmd_deadman_t): 6269 cflag = ZFS_CMD_COMPAT_DEADMAN; 6270 break; 6271 case sizeof(zfs_cmd_v28_t): 6272 cflag = ZFS_CMD_COMPAT_V28; 6273 break; 6274 case sizeof(zfs_cmd_v15_t): 6275 cflag = ZFS_CMD_COMPAT_V15; 6276 vecnum = zfs_ioctl_v15_to_v28[cmd]; 6277 6278 /* 6279 * Return without further handling 6280 * if the command is blacklisted. 6281 */ 6282 if (vecnum == ZFS_IOC_COMPAT_PASS) 6283 return (0); 6284 else if (vecnum == ZFS_IOC_COMPAT_FAIL) 6285 return (ENOTSUP); 6286 break; 6287 default: 6288 return (EINVAL); 6289 } 6290 } 6291 6292#ifdef illumos 6293 vecnum = cmd - ZFS_IOC_FIRST; 6294 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip)); 6295#endif 6296 6297 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0])) 6298 return (SET_ERROR(EINVAL)); 6299 vec = &zfs_ioc_vec[vecnum]; 6300 6301 zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP); 6302 6303#ifdef illumos 6304 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag); 6305 if (error != 0) { 6306 error = SET_ERROR(EFAULT); 6307 goto out; 6308 } 6309#else /* !illumos */ 6310 bzero(zc, sizeof(zfs_cmd_t)); 6311 6312 if (newioc) { 6313 zc_iocparm = (void *)arg; 6314 6315 switch (zc_iocparm->zfs_ioctl_version) { 6316 case ZFS_IOCVER_CURRENT: 6317 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_t)) { 6318 error = SET_ERROR(EINVAL); 6319 goto out; 6320 } 6321 break; 6322 case ZFS_IOCVER_INLANES: 6323 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_inlanes_t)) { 6324 error = SET_ERROR(EFAULT); 6325 goto out; 6326 } 6327 compat = B_TRUE; 6328 cflag = ZFS_CMD_COMPAT_INLANES; 6329 break; 6330 case ZFS_IOCVER_RESUME: 6331 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_resume_t)) { 6332 error = SET_ERROR(EFAULT); 6333 goto out; 6334 } 6335 compat = B_TRUE; 6336 cflag = ZFS_CMD_COMPAT_RESUME; 6337 break; 6338 case ZFS_IOCVER_EDBP: 6339 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_edbp_t)) { 6340 error = SET_ERROR(EFAULT); 6341 goto out; 6342 } 6343 compat = B_TRUE; 6344 cflag = ZFS_CMD_COMPAT_EDBP; 6345 break; 6346 case ZFS_IOCVER_ZCMD: 6347 if (zc_iocparm->zfs_cmd_size > sizeof(zfs_cmd_t) || 6348 zc_iocparm->zfs_cmd_size < sizeof(zfs_cmd_zcmd_t)) { 6349 error = SET_ERROR(EFAULT); 6350 goto out; 6351 } 6352 compat = B_TRUE; 6353 cflag = ZFS_CMD_COMPAT_ZCMD; 6354 break; 6355 default: 6356 error = SET_ERROR(EINVAL); 6357 goto out; 6358 /* NOTREACHED */ 6359 } 6360 6361 if (compat) { 6362 ASSERT(sizeof(zfs_cmd_t) >= zc_iocparm->zfs_cmd_size); 6363 compat_zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP); 6364 bzero(compat_zc, sizeof(zfs_cmd_t)); 6365 6366 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd, 6367 compat_zc, zc_iocparm->zfs_cmd_size, flag); 6368 if (error != 0) { 6369 error = SET_ERROR(EFAULT); 6370 goto out; 6371 } 6372 } else { 6373 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd, 6374 zc, zc_iocparm->zfs_cmd_size, flag); 6375 if (error != 0) { 6376 error = SET_ERROR(EFAULT); 6377 goto out; 6378 } 6379 } 6380 } 6381 6382 if (compat) { 6383 if (newioc) { 6384 ASSERT(compat_zc != NULL); 6385 zfs_cmd_compat_get(zc, compat_zc, cflag); 6386 } else { 6387 ASSERT(compat_zc == NULL); 6388 zfs_cmd_compat_get(zc, arg, cflag); 6389 } 6390 oldvecnum = vecnum; 6391 error = zfs_ioctl_compat_pre(zc, &vecnum, cflag); 6392 if (error != 0) 6393 goto out; 6394 if (oldvecnum != vecnum) 6395 vec = &zfs_ioc_vec[vecnum]; 6396 } 6397#endif /* !illumos */ 6398 6399 zc->zc_iflags = flag & FKIOCTL; 6400 if (zc->zc_nvlist_src_size != 0) { 6401 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 6402 zc->zc_iflags, &innvl); 6403 if (error != 0) 6404 goto out; 6405 } 6406 6407 /* rewrite innvl for backwards compatibility */ 6408 if (compat) 6409 innvl = zfs_ioctl_compat_innvl(zc, innvl, vecnum, cflag); 6410 6411 /* 6412 * Ensure that all pool/dataset names are valid before we pass down to 6413 * the lower layers. 6414 */ 6415 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0'; 6416 switch (vec->zvec_namecheck) { 6417 case POOL_NAME: 6418 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0) 6419 error = SET_ERROR(EINVAL); 6420 else 6421 error = pool_status_check(zc->zc_name, 6422 vec->zvec_namecheck, vec->zvec_pool_check); 6423 break; 6424 6425 case DATASET_NAME: 6426 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0) 6427 error = SET_ERROR(EINVAL); 6428 else 6429 error = pool_status_check(zc->zc_name, 6430 vec->zvec_namecheck, vec->zvec_pool_check); 6431 break; 6432 6433 case NO_NAME: 6434 break; 6435 } 6436 6437 if (error == 0) 6438 error = vec->zvec_secpolicy(zc, innvl, cr); 6439 6440 if (error != 0) 6441 goto out; 6442 6443 /* legacy ioctls can modify zc_name */ 6444 len = strcspn(zc->zc_name, "/@#") + 1; 6445 saved_poolname = kmem_alloc(len, KM_SLEEP); 6446 (void) strlcpy(saved_poolname, zc->zc_name, len); 6447 6448 if (vec->zvec_func != NULL) { 6449 nvlist_t *outnvl; 6450 int puterror = 0; 6451 spa_t *spa; 6452 nvlist_t *lognv = NULL; 6453 6454 ASSERT(vec->zvec_legacy_func == NULL); 6455 6456 /* 6457 * Add the innvl to the lognv before calling the func, 6458 * in case the func changes the innvl. 6459 */ 6460 if (vec->zvec_allow_log) { 6461 lognv = fnvlist_alloc(); 6462 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL, 6463 vec->zvec_name); 6464 if (!nvlist_empty(innvl)) { 6465 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL, 6466 innvl); 6467 } 6468 } 6469 6470 outnvl = fnvlist_alloc(); 6471 error = vec->zvec_func(zc->zc_name, innvl, outnvl); 6472 6473 if (error == 0 && vec->zvec_allow_log && 6474 spa_open(zc->zc_name, &spa, FTAG) == 0) { 6475 if (!nvlist_empty(outnvl)) { 6476 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL, 6477 outnvl); 6478 } 6479 (void) spa_history_log_nvl(spa, lognv); 6480 spa_close(spa, FTAG); 6481 } 6482 fnvlist_free(lognv); 6483 6484 /* rewrite outnvl for backwards compatibility */ 6485 if (compat) 6486 outnvl = zfs_ioctl_compat_outnvl(zc, outnvl, vecnum, 6487 cflag); 6488 6489 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) { 6490 int smusherror = 0; 6491 if (vec->zvec_smush_outnvlist) { 6492 smusherror = nvlist_smush(outnvl, 6493 zc->zc_nvlist_dst_size); 6494 } 6495 if (smusherror == 0) 6496 puterror = put_nvlist(zc, outnvl); 6497 } 6498 6499 if (puterror != 0) 6500 error = puterror; 6501 6502 nvlist_free(outnvl); 6503 } else { 6504 error = vec->zvec_legacy_func(zc); 6505 } 6506 6507out: 6508 nvlist_free(innvl); 6509 6510#ifdef illumos 6511 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag); 6512 if (error == 0 && rc != 0) 6513 error = SET_ERROR(EFAULT); 6514#else 6515 if (compat) { 6516 zfs_ioctl_compat_post(zc, cmd, cflag); 6517 if (newioc) { 6518 ASSERT(compat_zc != NULL); 6519 ASSERT(sizeof(zfs_cmd_t) >= zc_iocparm->zfs_cmd_size); 6520 6521 zfs_cmd_compat_put(zc, compat_zc, vecnum, cflag); 6522 rc = ddi_copyout(compat_zc, 6523 (void *)(uintptr_t)zc_iocparm->zfs_cmd, 6524 zc_iocparm->zfs_cmd_size, flag); 6525 if (error == 0 && rc != 0) 6526 error = SET_ERROR(EFAULT); 6527 kmem_free(compat_zc, sizeof (zfs_cmd_t)); 6528 } else { 6529 zfs_cmd_compat_put(zc, arg, vecnum, cflag); 6530 } 6531 } else { 6532 ASSERT(newioc); 6533 6534 rc = ddi_copyout(zc, (void *)(uintptr_t)zc_iocparm->zfs_cmd, 6535 sizeof (zfs_cmd_t), flag); 6536 if (error == 0 && rc != 0) 6537 error = SET_ERROR(EFAULT); 6538 } 6539#endif 6540 if (error == 0 && vec->zvec_allow_log) { 6541 char *s = tsd_get(zfs_allow_log_key); 6542 if (s != NULL) 6543 strfree(s); 6544 (void) tsd_set(zfs_allow_log_key, saved_poolname); 6545 } else { 6546 if (saved_poolname != NULL) 6547 strfree(saved_poolname); 6548 } 6549 6550 kmem_free(zc, sizeof (zfs_cmd_t)); 6551 return (error); 6552} 6553 6554#ifdef illumos 6555static int 6556zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd) 6557{ 6558 if (cmd != DDI_ATTACH) 6559 return (DDI_FAILURE); 6560 6561 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0, 6562 DDI_PSEUDO, 0) == DDI_FAILURE) 6563 return (DDI_FAILURE); 6564 6565 zfs_dip = dip; 6566 6567 ddi_report_dev(dip); 6568 6569 return (DDI_SUCCESS); 6570} 6571 6572static int 6573zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd) 6574{ 6575 if (spa_busy() || zfs_busy() || zvol_busy()) 6576 return (DDI_FAILURE); 6577 6578 if (cmd != DDI_DETACH) 6579 return (DDI_FAILURE); 6580 6581 zfs_dip = NULL; 6582 6583 ddi_prop_remove_all(dip); 6584 ddi_remove_minor_node(dip, NULL); 6585 6586 return (DDI_SUCCESS); 6587} 6588 6589/*ARGSUSED*/ 6590static int 6591zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result) 6592{ 6593 switch (infocmd) { 6594 case DDI_INFO_DEVT2DEVINFO: 6595 *result = zfs_dip; 6596 return (DDI_SUCCESS); 6597 6598 case DDI_INFO_DEVT2INSTANCE: 6599 *result = (void *)0; 6600 return (DDI_SUCCESS); 6601 } 6602 6603 return (DDI_FAILURE); 6604} 6605#endif /* illumos */ 6606 6607/* 6608 * OK, so this is a little weird. 6609 * 6610 * /dev/zfs is the control node, i.e. minor 0. 6611 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0. 6612 * 6613 * /dev/zfs has basically nothing to do except serve up ioctls, 6614 * so most of the standard driver entry points are in zvol.c. 6615 */ 6616#ifdef illumos 6617static struct cb_ops zfs_cb_ops = { 6618 zfsdev_open, /* open */ 6619 zfsdev_close, /* close */ 6620 zvol_strategy, /* strategy */ 6621 nodev, /* print */ 6622 zvol_dump, /* dump */ 6623 zvol_read, /* read */ 6624 zvol_write, /* write */ 6625 zfsdev_ioctl, /* ioctl */ 6626 nodev, /* devmap */ 6627 nodev, /* mmap */ 6628 nodev, /* segmap */ 6629 nochpoll, /* poll */ 6630 ddi_prop_op, /* prop_op */ 6631 NULL, /* streamtab */ 6632 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */ 6633 CB_REV, /* version */ 6634 nodev, /* async read */ 6635 nodev, /* async write */ 6636}; 6637 6638static struct dev_ops zfs_dev_ops = { 6639 DEVO_REV, /* version */ 6640 0, /* refcnt */ 6641 zfs_info, /* info */ 6642 nulldev, /* identify */ 6643 nulldev, /* probe */ 6644 zfs_attach, /* attach */ 6645 zfs_detach, /* detach */ 6646 nodev, /* reset */ 6647 &zfs_cb_ops, /* driver operations */ 6648 NULL, /* no bus operations */ 6649 NULL, /* power */ 6650 ddi_quiesce_not_needed, /* quiesce */ 6651}; 6652 6653static struct modldrv zfs_modldrv = { 6654 &mod_driverops, 6655 "ZFS storage pool", 6656 &zfs_dev_ops 6657}; 6658 6659static struct modlinkage modlinkage = { 6660 MODREV_1, 6661 (void *)&zfs_modlfs, 6662 (void *)&zfs_modldrv, 6663 NULL 6664}; 6665#endif /* illumos */ 6666 6667static struct cdevsw zfs_cdevsw = { 6668 .d_version = D_VERSION, 6669 .d_open = zfsdev_open, 6670 .d_ioctl = zfsdev_ioctl, 6671 .d_name = ZFS_DEV_NAME 6672}; 6673 6674static void 6675zfs_allow_log_destroy(void *arg) 6676{ 6677 char *poolname = arg; 6678 strfree(poolname); 6679} 6680 6681static void 6682zfsdev_init(void) 6683{ 6684 zfsdev = make_dev(&zfs_cdevsw, 0x0, UID_ROOT, GID_OPERATOR, 0666, 6685 ZFS_DEV_NAME); 6686} 6687 6688static void 6689zfsdev_fini(void) 6690{ 6691 if (zfsdev != NULL) 6692 destroy_dev(zfsdev); 6693} 6694 6695static struct root_hold_token *zfs_root_token; 6696struct proc *zfsproc; 6697 6698#ifdef illumos 6699int 6700_init(void) 6701{ 6702 int error; 6703 6704 spa_init(FREAD | FWRITE); 6705 zfs_init(); 6706 zvol_init(); 6707 zfs_ioctl_init(); 6708 6709 if ((error = mod_install(&modlinkage)) != 0) { 6710 zvol_fini(); 6711 zfs_fini(); 6712 spa_fini(); 6713 return (error); 6714 } 6715 6716 tsd_create(&zfs_fsyncer_key, NULL); 6717 tsd_create(&rrw_tsd_key, rrw_tsd_destroy); 6718 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy); 6719 6720 error = ldi_ident_from_mod(&modlinkage, &zfs_li); 6721 ASSERT(error == 0); 6722 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL); 6723 6724 return (0); 6725} 6726 6727int 6728_fini(void) 6729{ 6730 int error; 6731 6732 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled) 6733 return (SET_ERROR(EBUSY)); 6734 6735 if ((error = mod_remove(&modlinkage)) != 0) 6736 return (error); 6737 6738 zvol_fini(); 6739 zfs_fini(); 6740 spa_fini(); 6741 if (zfs_nfsshare_inited) 6742 (void) ddi_modclose(nfs_mod); 6743 if (zfs_smbshare_inited) 6744 (void) ddi_modclose(smbsrv_mod); 6745 if (zfs_nfsshare_inited || zfs_smbshare_inited) 6746 (void) ddi_modclose(sharefs_mod); 6747 6748 tsd_destroy(&zfs_fsyncer_key); 6749 ldi_ident_release(zfs_li); 6750 zfs_li = NULL; 6751 mutex_destroy(&zfs_share_lock); 6752 6753 return (error); 6754} 6755 6756int 6757_info(struct modinfo *modinfop) 6758{ 6759 return (mod_info(&modlinkage, modinfop)); 6760} 6761#endif /* illumos */ 6762 6763static int zfs__init(void); 6764static int zfs__fini(void); 6765static void zfs_shutdown(void *, int); 6766 6767static eventhandler_tag zfs_shutdown_event_tag; 6768 6769#ifdef __FreeBSD__ 6770#define ZFS_MIN_KSTACK_PAGES 4 6771#endif 6772 6773int 6774zfs__init(void) 6775{ 6776 6777#ifdef __FreeBSD__ 6778#if KSTACK_PAGES < ZFS_MIN_KSTACK_PAGES 6779 printf("ZFS NOTICE: KSTACK_PAGES is %d which could result in stack " 6780 "overflow panic!\nPlease consider adding " 6781 "'options KSTACK_PAGES=%d' to your kernel config\n", KSTACK_PAGES, 6782 ZFS_MIN_KSTACK_PAGES); 6783#endif 6784#endif 6785 zfs_root_token = root_mount_hold("ZFS"); 6786 6787 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL); 6788 6789 spa_init(FREAD | FWRITE); 6790 zfs_init(); 6791 zvol_init(); 6792 zfs_ioctl_init(); 6793 6794 tsd_create(&zfs_fsyncer_key, NULL); 6795 tsd_create(&rrw_tsd_key, rrw_tsd_destroy); 6796 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy); 6797 tsd_create(&zfs_geom_probe_vdev_key, NULL); 6798 6799 printf("ZFS storage pool version: features support (" SPA_VERSION_STRING ")\n"); 6800 root_mount_rel(zfs_root_token); 6801 6802 zfsdev_init(); 6803 6804 return (0); 6805} 6806 6807int 6808zfs__fini(void) 6809{ 6810 if (spa_busy() || zfs_busy() || zvol_busy() || 6811 zio_injection_enabled) { 6812 return (EBUSY); 6813 } 6814 6815 zfsdev_fini(); 6816 zvol_fini(); 6817 zfs_fini(); 6818 spa_fini(); 6819 6820 tsd_destroy(&zfs_fsyncer_key); 6821 tsd_destroy(&rrw_tsd_key); 6822 tsd_destroy(&zfs_allow_log_key); 6823 6824 mutex_destroy(&zfs_share_lock); 6825 6826 return (0); 6827} 6828 6829static void 6830zfs_shutdown(void *arg __unused, int howto __unused) 6831{ 6832 6833 /* 6834 * ZFS fini routines can not properly work in a panic-ed system. 6835 */ 6836 if (panicstr == NULL) 6837 (void)zfs__fini(); 6838} 6839 6840 6841static int 6842zfs_modevent(module_t mod, int type, void *unused __unused) 6843{ 6844 int err; 6845 6846 switch (type) { 6847 case MOD_LOAD: 6848 err = zfs__init(); 6849 if (err == 0) 6850 zfs_shutdown_event_tag = EVENTHANDLER_REGISTER( 6851 shutdown_post_sync, zfs_shutdown, NULL, 6852 SHUTDOWN_PRI_FIRST); 6853 return (err); 6854 case MOD_UNLOAD: 6855 err = zfs__fini(); 6856 if (err == 0 && zfs_shutdown_event_tag != NULL) 6857 EVENTHANDLER_DEREGISTER(shutdown_post_sync, 6858 zfs_shutdown_event_tag); 6859 return (err); 6860 case MOD_SHUTDOWN: 6861 return (0); 6862 default: 6863 break; 6864 } 6865 return (EOPNOTSUPP); 6866} 6867 6868static moduledata_t zfs_mod = { 6869 "zfsctrl", 6870 zfs_modevent, 6871 0 6872}; 6873DECLARE_MODULE(zfsctrl, zfs_mod, SI_SUB_VFS, SI_ORDER_ANY); 6874MODULE_VERSION(zfsctrl, 1); 6875MODULE_DEPEND(zfsctrl, opensolaris, 1, 1, 1); 6876MODULE_DEPEND(zfsctrl, krpc, 1, 1, 1); 6877MODULE_DEPEND(zfsctrl, acl_nfs4, 1, 1, 1); 6878