zfs_ioctl.c revision 324009
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 vfs_ref(vfsp); 3044 break; 3045 } 3046 } 3047 mtx_unlock(&mountlist_mtx); 3048 return (vfsp); 3049} 3050 3051/* ARGSUSED */ 3052static void 3053zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) 3054{ 3055 zfs_creat_t *zct = arg; 3056 3057 zfs_create_fs(os, cr, zct->zct_zplprops, tx); 3058} 3059 3060#define ZFS_PROP_UNDEFINED ((uint64_t)-1) 3061 3062/* 3063 * inputs: 3064 * os parent objset pointer (NULL if root fs) 3065 * fuids_ok fuids allowed in this version of the spa? 3066 * sa_ok SAs allowed in this version of the spa? 3067 * createprops list of properties requested by creator 3068 * 3069 * outputs: 3070 * zplprops values for the zplprops we attach to the master node object 3071 * is_ci true if requested file system will be purely case-insensitive 3072 * 3073 * Determine the settings for utf8only, normalization and 3074 * casesensitivity. Specific values may have been requested by the 3075 * creator and/or we can inherit values from the parent dataset. If 3076 * the file system is of too early a vintage, a creator can not 3077 * request settings for these properties, even if the requested 3078 * setting is the default value. We don't actually want to create dsl 3079 * properties for these, so remove them from the source nvlist after 3080 * processing. 3081 */ 3082static int 3083zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver, 3084 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops, 3085 nvlist_t *zplprops, boolean_t *is_ci) 3086{ 3087 uint64_t sense = ZFS_PROP_UNDEFINED; 3088 uint64_t norm = ZFS_PROP_UNDEFINED; 3089 uint64_t u8 = ZFS_PROP_UNDEFINED; 3090 3091 ASSERT(zplprops != NULL); 3092 3093 /* 3094 * Pull out creator prop choices, if any. 3095 */ 3096 if (createprops) { 3097 (void) nvlist_lookup_uint64(createprops, 3098 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver); 3099 (void) nvlist_lookup_uint64(createprops, 3100 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm); 3101 (void) nvlist_remove_all(createprops, 3102 zfs_prop_to_name(ZFS_PROP_NORMALIZE)); 3103 (void) nvlist_lookup_uint64(createprops, 3104 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8); 3105 (void) nvlist_remove_all(createprops, 3106 zfs_prop_to_name(ZFS_PROP_UTF8ONLY)); 3107 (void) nvlist_lookup_uint64(createprops, 3108 zfs_prop_to_name(ZFS_PROP_CASE), &sense); 3109 (void) nvlist_remove_all(createprops, 3110 zfs_prop_to_name(ZFS_PROP_CASE)); 3111 } 3112 3113 /* 3114 * If the zpl version requested is whacky or the file system 3115 * or pool is version is too "young" to support normalization 3116 * and the creator tried to set a value for one of the props, 3117 * error out. 3118 */ 3119 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) || 3120 (zplver >= ZPL_VERSION_FUID && !fuids_ok) || 3121 (zplver >= ZPL_VERSION_SA && !sa_ok) || 3122 (zplver < ZPL_VERSION_NORMALIZATION && 3123 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED || 3124 sense != ZFS_PROP_UNDEFINED))) 3125 return (SET_ERROR(ENOTSUP)); 3126 3127 /* 3128 * Put the version in the zplprops 3129 */ 3130 VERIFY(nvlist_add_uint64(zplprops, 3131 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0); 3132 3133 if (norm == ZFS_PROP_UNDEFINED) 3134 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0); 3135 VERIFY(nvlist_add_uint64(zplprops, 3136 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0); 3137 3138 /* 3139 * If we're normalizing, names must always be valid UTF-8 strings. 3140 */ 3141 if (norm) 3142 u8 = 1; 3143 if (u8 == ZFS_PROP_UNDEFINED) 3144 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0); 3145 VERIFY(nvlist_add_uint64(zplprops, 3146 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0); 3147 3148 if (sense == ZFS_PROP_UNDEFINED) 3149 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0); 3150 VERIFY(nvlist_add_uint64(zplprops, 3151 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0); 3152 3153 if (is_ci) 3154 *is_ci = (sense == ZFS_CASE_INSENSITIVE); 3155 3156 return (0); 3157} 3158 3159static int 3160zfs_fill_zplprops(const char *dataset, nvlist_t *createprops, 3161 nvlist_t *zplprops, boolean_t *is_ci) 3162{ 3163 boolean_t fuids_ok, sa_ok; 3164 uint64_t zplver = ZPL_VERSION; 3165 objset_t *os = NULL; 3166 char parentname[ZFS_MAX_DATASET_NAME_LEN]; 3167 char *cp; 3168 spa_t *spa; 3169 uint64_t spa_vers; 3170 int error; 3171 3172 (void) strlcpy(parentname, dataset, sizeof (parentname)); 3173 cp = strrchr(parentname, '/'); 3174 ASSERT(cp != NULL); 3175 cp[0] = '\0'; 3176 3177 if ((error = spa_open(dataset, &spa, FTAG)) != 0) 3178 return (error); 3179 3180 spa_vers = spa_version(spa); 3181 spa_close(spa, FTAG); 3182 3183 zplver = zfs_zpl_version_map(spa_vers); 3184 fuids_ok = (zplver >= ZPL_VERSION_FUID); 3185 sa_ok = (zplver >= ZPL_VERSION_SA); 3186 3187 /* 3188 * Open parent object set so we can inherit zplprop values. 3189 */ 3190 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0) 3191 return (error); 3192 3193 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops, 3194 zplprops, is_ci); 3195 dmu_objset_rele(os, FTAG); 3196 return (error); 3197} 3198 3199static int 3200zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops, 3201 nvlist_t *zplprops, boolean_t *is_ci) 3202{ 3203 boolean_t fuids_ok; 3204 boolean_t sa_ok; 3205 uint64_t zplver = ZPL_VERSION; 3206 int error; 3207 3208 zplver = zfs_zpl_version_map(spa_vers); 3209 fuids_ok = (zplver >= ZPL_VERSION_FUID); 3210 sa_ok = (zplver >= ZPL_VERSION_SA); 3211 3212 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok, 3213 createprops, zplprops, is_ci); 3214 return (error); 3215} 3216 3217/* 3218 * innvl: { 3219 * "type" -> dmu_objset_type_t (int32) 3220 * (optional) "props" -> { prop -> value } 3221 * } 3222 * 3223 * outnvl: propname -> error code (int32) 3224 */ 3225static int 3226zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3227{ 3228 int error = 0; 3229 zfs_creat_t zct = { 0 }; 3230 nvlist_t *nvprops = NULL; 3231 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx); 3232 int32_t type32; 3233 dmu_objset_type_t type; 3234 boolean_t is_insensitive = B_FALSE; 3235 3236 if (nvlist_lookup_int32(innvl, "type", &type32) != 0) 3237 return (SET_ERROR(EINVAL)); 3238 type = type32; 3239 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops); 3240 3241 switch (type) { 3242 case DMU_OST_ZFS: 3243 cbfunc = zfs_create_cb; 3244 break; 3245 3246 case DMU_OST_ZVOL: 3247 cbfunc = zvol_create_cb; 3248 break; 3249 3250 default: 3251 cbfunc = NULL; 3252 break; 3253 } 3254 if (strchr(fsname, '@') || 3255 strchr(fsname, '%')) 3256 return (SET_ERROR(EINVAL)); 3257 3258 zct.zct_props = nvprops; 3259 3260 if (cbfunc == NULL) 3261 return (SET_ERROR(EINVAL)); 3262 3263 if (type == DMU_OST_ZVOL) { 3264 uint64_t volsize, volblocksize; 3265 3266 if (nvprops == NULL) 3267 return (SET_ERROR(EINVAL)); 3268 if (nvlist_lookup_uint64(nvprops, 3269 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0) 3270 return (SET_ERROR(EINVAL)); 3271 3272 if ((error = nvlist_lookup_uint64(nvprops, 3273 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), 3274 &volblocksize)) != 0 && error != ENOENT) 3275 return (SET_ERROR(EINVAL)); 3276 3277 if (error != 0) 3278 volblocksize = zfs_prop_default_numeric( 3279 ZFS_PROP_VOLBLOCKSIZE); 3280 3281 if ((error = zvol_check_volblocksize( 3282 volblocksize)) != 0 || 3283 (error = zvol_check_volsize(volsize, 3284 volblocksize)) != 0) 3285 return (error); 3286 } else if (type == DMU_OST_ZFS) { 3287 int error; 3288 3289 /* 3290 * We have to have normalization and 3291 * case-folding flags correct when we do the 3292 * file system creation, so go figure them out 3293 * now. 3294 */ 3295 VERIFY(nvlist_alloc(&zct.zct_zplprops, 3296 NV_UNIQUE_NAME, KM_SLEEP) == 0); 3297 error = zfs_fill_zplprops(fsname, nvprops, 3298 zct.zct_zplprops, &is_insensitive); 3299 if (error != 0) { 3300 nvlist_free(zct.zct_zplprops); 3301 return (error); 3302 } 3303 } 3304 3305 error = dmu_objset_create(fsname, type, 3306 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct); 3307 nvlist_free(zct.zct_zplprops); 3308 3309 /* 3310 * It would be nice to do this atomically. 3311 */ 3312 if (error == 0) { 3313 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL, 3314 nvprops, outnvl); 3315 if (error != 0) 3316 (void) dsl_destroy_head(fsname); 3317 } 3318#ifdef __FreeBSD__ 3319 if (error == 0 && type == DMU_OST_ZVOL) 3320 zvol_create_minors(fsname); 3321#endif 3322 return (error); 3323} 3324 3325/* 3326 * innvl: { 3327 * "origin" -> name of origin snapshot 3328 * (optional) "props" -> { prop -> value } 3329 * } 3330 * 3331 * outnvl: propname -> error code (int32) 3332 */ 3333static int 3334zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3335{ 3336 int error = 0; 3337 nvlist_t *nvprops = NULL; 3338 char *origin_name; 3339 3340 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0) 3341 return (SET_ERROR(EINVAL)); 3342 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops); 3343 3344 if (strchr(fsname, '@') || 3345 strchr(fsname, '%')) 3346 return (SET_ERROR(EINVAL)); 3347 3348 if (dataset_namecheck(origin_name, NULL, NULL) != 0) 3349 return (SET_ERROR(EINVAL)); 3350 error = dmu_objset_clone(fsname, origin_name); 3351 if (error != 0) 3352 return (error); 3353 3354 /* 3355 * It would be nice to do this atomically. 3356 */ 3357 if (error == 0) { 3358 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL, 3359 nvprops, outnvl); 3360 if (error != 0) 3361 (void) dsl_destroy_head(fsname); 3362 } 3363#ifdef __FreeBSD__ 3364 if (error == 0) 3365 zvol_create_minors(fsname); 3366#endif 3367 return (error); 3368} 3369 3370/* 3371 * innvl: { 3372 * "snaps" -> { snapshot1, snapshot2 } 3373 * (optional) "props" -> { prop -> value (string) } 3374 * } 3375 * 3376 * outnvl: snapshot -> error code (int32) 3377 */ 3378static int 3379zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3380{ 3381 nvlist_t *snaps; 3382 nvlist_t *props = NULL; 3383 int error, poollen; 3384 nvpair_t *pair; 3385 3386 (void) nvlist_lookup_nvlist(innvl, "props", &props); 3387 if ((error = zfs_check_userprops(poolname, props)) != 0) 3388 return (error); 3389 3390 if (!nvlist_empty(props) && 3391 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS)) 3392 return (SET_ERROR(ENOTSUP)); 3393 3394 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 3395 return (SET_ERROR(EINVAL)); 3396 poollen = strlen(poolname); 3397 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 3398 pair = nvlist_next_nvpair(snaps, pair)) { 3399 const char *name = nvpair_name(pair); 3400 const char *cp = strchr(name, '@'); 3401 3402 /* 3403 * The snap name must contain an @, and the part after it must 3404 * contain only valid characters. 3405 */ 3406 if (cp == NULL || 3407 zfs_component_namecheck(cp + 1, NULL, NULL) != 0) 3408 return (SET_ERROR(EINVAL)); 3409 3410 /* 3411 * The snap must be in the specified pool. 3412 */ 3413 if (strncmp(name, poolname, poollen) != 0 || 3414 (name[poollen] != '/' && name[poollen] != '@')) 3415 return (SET_ERROR(EXDEV)); 3416 3417 /* This must be the only snap of this fs. */ 3418 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair); 3419 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) { 3420 if (strncmp(name, nvpair_name(pair2), cp - name + 1) 3421 == 0) { 3422 return (SET_ERROR(EXDEV)); 3423 } 3424 } 3425 } 3426 3427 error = dsl_dataset_snapshot(snaps, props, outnvl); 3428 return (error); 3429} 3430 3431/* 3432 * innvl: "message" -> string 3433 */ 3434/* ARGSUSED */ 3435static int 3436zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl) 3437{ 3438 char *message; 3439 spa_t *spa; 3440 int error; 3441 char *poolname; 3442 3443 /* 3444 * The poolname in the ioctl is not set, we get it from the TSD, 3445 * which was set at the end of the last successful ioctl that allows 3446 * logging. The secpolicy func already checked that it is set. 3447 * Only one log ioctl is allowed after each successful ioctl, so 3448 * we clear the TSD here. 3449 */ 3450 poolname = tsd_get(zfs_allow_log_key); 3451 (void) tsd_set(zfs_allow_log_key, NULL); 3452 error = spa_open(poolname, &spa, FTAG); 3453 strfree(poolname); 3454 if (error != 0) 3455 return (error); 3456 3457 if (nvlist_lookup_string(innvl, "message", &message) != 0) { 3458 spa_close(spa, FTAG); 3459 return (SET_ERROR(EINVAL)); 3460 } 3461 3462 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) { 3463 spa_close(spa, FTAG); 3464 return (SET_ERROR(ENOTSUP)); 3465 } 3466 3467 error = spa_history_log(spa, message); 3468 spa_close(spa, FTAG); 3469 return (error); 3470} 3471 3472#ifdef __FreeBSD__ 3473static int 3474zfs_ioc_nextboot(const char *unused, nvlist_t *innvl, nvlist_t *outnvl) 3475{ 3476 char name[MAXNAMELEN]; 3477 spa_t *spa; 3478 vdev_t *vd; 3479 char *command; 3480 uint64_t pool_guid; 3481 uint64_t vdev_guid; 3482 int error; 3483 3484 if (nvlist_lookup_uint64(innvl, 3485 ZPOOL_CONFIG_POOL_GUID, &pool_guid) != 0) 3486 return (EINVAL); 3487 if (nvlist_lookup_uint64(innvl, 3488 ZPOOL_CONFIG_GUID, &vdev_guid) != 0) 3489 return (EINVAL); 3490 if (nvlist_lookup_string(innvl, 3491 "command", &command) != 0) 3492 return (EINVAL); 3493 3494 mutex_enter(&spa_namespace_lock); 3495 spa = spa_by_guid(pool_guid, vdev_guid); 3496 if (spa != NULL) 3497 strcpy(name, spa_name(spa)); 3498 mutex_exit(&spa_namespace_lock); 3499 if (spa == NULL) 3500 return (ENOENT); 3501 3502 if ((error = spa_open(name, &spa, FTAG)) != 0) 3503 return (error); 3504 spa_vdev_state_enter(spa, SCL_ALL); 3505 vd = spa_lookup_by_guid(spa, vdev_guid, B_TRUE); 3506 if (vd == NULL) { 3507 (void) spa_vdev_state_exit(spa, NULL, ENXIO); 3508 spa_close(spa, FTAG); 3509 return (ENODEV); 3510 } 3511 error = vdev_label_write_pad2(vd, command, strlen(command)); 3512 (void) spa_vdev_state_exit(spa, NULL, 0); 3513 txg_wait_synced(spa->spa_dsl_pool, 0); 3514 spa_close(spa, FTAG); 3515 return (error); 3516} 3517#endif 3518 3519/* 3520 * The dp_config_rwlock must not be held when calling this, because the 3521 * unmount may need to write out data. 3522 * 3523 * This function is best-effort. Callers must deal gracefully if it 3524 * remains mounted (or is remounted after this call). 3525 * 3526 * Returns 0 if the argument is not a snapshot, or it is not currently a 3527 * filesystem, or we were able to unmount it. Returns error code otherwise. 3528 */ 3529int 3530zfs_unmount_snap(const char *snapname) 3531{ 3532 vfs_t *vfsp; 3533 zfsvfs_t *zfsvfs; 3534#ifdef illumos 3535 int err; 3536#endif 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#ifdef illumos 3549 err = vn_vfswlock(vfsp->vfs_vnodecovered); 3550 VFS_RELE(vfsp); 3551 if (err != 0) 3552 return (SET_ERROR(err)); 3553#endif 3554 3555 /* 3556 * Always force the unmount for snapshots. 3557 */ 3558#ifdef illumos 3559 (void) dounmount(vfsp, MS_FORCE, kcred); 3560#else 3561 (void) dounmount(vfsp, MS_FORCE, curthread); 3562#endif 3563 return (0); 3564} 3565 3566/* ARGSUSED */ 3567static int 3568zfs_unmount_snap_cb(const char *snapname, void *arg) 3569{ 3570 return (zfs_unmount_snap(snapname)); 3571} 3572 3573/* 3574 * When a clone is destroyed, its origin may also need to be destroyed, 3575 * in which case it must be unmounted. This routine will do that unmount 3576 * if necessary. 3577 */ 3578void 3579zfs_destroy_unmount_origin(const char *fsname) 3580{ 3581 int error; 3582 objset_t *os; 3583 dsl_dataset_t *ds; 3584 3585 error = dmu_objset_hold(fsname, FTAG, &os); 3586 if (error != 0) 3587 return; 3588 ds = dmu_objset_ds(os); 3589 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) { 3590 char originname[ZFS_MAX_DATASET_NAME_LEN]; 3591 dsl_dataset_name(ds->ds_prev, originname); 3592 dmu_objset_rele(os, FTAG); 3593 (void) zfs_unmount_snap(originname); 3594 } else { 3595 dmu_objset_rele(os, FTAG); 3596 } 3597} 3598 3599/* 3600 * innvl: { 3601 * "snaps" -> { snapshot1, snapshot2 } 3602 * (optional boolean) "defer" 3603 * } 3604 * 3605 * outnvl: snapshot -> error code (int32) 3606 * 3607 */ 3608/* ARGSUSED */ 3609static int 3610zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3611{ 3612 int error, poollen; 3613 nvlist_t *snaps; 3614 nvpair_t *pair; 3615 boolean_t defer; 3616 3617 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 3618 return (SET_ERROR(EINVAL)); 3619 defer = nvlist_exists(innvl, "defer"); 3620 3621 poollen = strlen(poolname); 3622 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 3623 pair = nvlist_next_nvpair(snaps, pair)) { 3624 const char *name = nvpair_name(pair); 3625 3626 /* 3627 * The snap must be in the specified pool to prevent the 3628 * invalid removal of zvol minors below. 3629 */ 3630 if (strncmp(name, poolname, poollen) != 0 || 3631 (name[poollen] != '/' && name[poollen] != '@')) 3632 return (SET_ERROR(EXDEV)); 3633 3634 error = zfs_unmount_snap(name); 3635 if (error != 0) 3636 return (error); 3637#if defined(__FreeBSD__) 3638 zvol_remove_minors(name); 3639#endif 3640 } 3641 3642 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl)); 3643} 3644 3645/* 3646 * Create bookmarks. Bookmark names are of the form <fs>#<bmark>. 3647 * All bookmarks must be in the same pool. 3648 * 3649 * innvl: { 3650 * bookmark1 -> snapshot1, bookmark2 -> snapshot2 3651 * } 3652 * 3653 * outnvl: bookmark -> error code (int32) 3654 * 3655 */ 3656/* ARGSUSED */ 3657static int 3658zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3659{ 3660 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 3661 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 3662 char *snap_name; 3663 3664 /* 3665 * Verify the snapshot argument. 3666 */ 3667 if (nvpair_value_string(pair, &snap_name) != 0) 3668 return (SET_ERROR(EINVAL)); 3669 3670 3671 /* Verify that the keys (bookmarks) are unique */ 3672 for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair); 3673 pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) { 3674 if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0) 3675 return (SET_ERROR(EINVAL)); 3676 } 3677 } 3678 3679 return (dsl_bookmark_create(innvl, outnvl)); 3680} 3681 3682/* 3683 * innvl: { 3684 * property 1, property 2, ... 3685 * } 3686 * 3687 * outnvl: { 3688 * bookmark name 1 -> { property 1, property 2, ... }, 3689 * bookmark name 2 -> { property 1, property 2, ... } 3690 * } 3691 * 3692 */ 3693static int 3694zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3695{ 3696 return (dsl_get_bookmarks(fsname, innvl, outnvl)); 3697} 3698 3699/* 3700 * innvl: { 3701 * bookmark name 1, bookmark name 2 3702 * } 3703 * 3704 * outnvl: bookmark -> error code (int32) 3705 * 3706 */ 3707static int 3708zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl, 3709 nvlist_t *outnvl) 3710{ 3711 int error, poollen; 3712 3713 poollen = strlen(poolname); 3714 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 3715 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 3716 const char *name = nvpair_name(pair); 3717 const char *cp = strchr(name, '#'); 3718 3719 /* 3720 * The bookmark name must contain an #, and the part after it 3721 * must contain only valid characters. 3722 */ 3723 if (cp == NULL || 3724 zfs_component_namecheck(cp + 1, NULL, NULL) != 0) 3725 return (SET_ERROR(EINVAL)); 3726 3727 /* 3728 * The bookmark must be in the specified pool. 3729 */ 3730 if (strncmp(name, poolname, poollen) != 0 || 3731 (name[poollen] != '/' && name[poollen] != '#')) 3732 return (SET_ERROR(EXDEV)); 3733 } 3734 3735 error = dsl_bookmark_destroy(innvl, outnvl); 3736 return (error); 3737} 3738 3739/* 3740 * inputs: 3741 * zc_name name of dataset to destroy 3742 * zc_objset_type type of objset 3743 * zc_defer_destroy mark for deferred destroy 3744 * 3745 * outputs: none 3746 */ 3747static int 3748zfs_ioc_destroy(zfs_cmd_t *zc) 3749{ 3750 int err; 3751 3752 if (zc->zc_objset_type == DMU_OST_ZFS) { 3753 err = zfs_unmount_snap(zc->zc_name); 3754 if (err != 0) 3755 return (err); 3756 } 3757 3758 if (strchr(zc->zc_name, '@')) 3759 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy); 3760 else 3761 err = dsl_destroy_head(zc->zc_name); 3762 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0) 3763#ifdef __FreeBSD__ 3764 zvol_remove_minors(zc->zc_name); 3765#else 3766 (void) zvol_remove_minor(zc->zc_name); 3767#endif 3768 return (err); 3769} 3770 3771/* 3772 * fsname is name of dataset to rollback (to most recent snapshot) 3773 * 3774 * innvl is not used. 3775 * 3776 * outnvl: "target" -> name of most recent snapshot 3777 * } 3778 */ 3779/* ARGSUSED */ 3780static int 3781zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl) 3782{ 3783 zfsvfs_t *zfsvfs; 3784 int error; 3785 3786 if (getzfsvfs(fsname, &zfsvfs) == 0) { 3787 dsl_dataset_t *ds; 3788 3789 ds = dmu_objset_ds(zfsvfs->z_os); 3790 error = zfs_suspend_fs(zfsvfs); 3791 if (error == 0) { 3792 int resume_err; 3793 3794 error = dsl_dataset_rollback(fsname, zfsvfs, outnvl); 3795 resume_err = zfs_resume_fs(zfsvfs, ds); 3796 error = error ? error : resume_err; 3797 } 3798#ifdef illumos 3799 VFS_RELE(zfsvfs->z_vfs); 3800#else 3801 vfs_unbusy(zfsvfs->z_vfs); 3802#endif 3803 } else { 3804 error = dsl_dataset_rollback(fsname, NULL, outnvl); 3805 } 3806 return (error); 3807} 3808 3809static int 3810recursive_unmount(const char *fsname, void *arg) 3811{ 3812 const char *snapname = arg; 3813 char fullname[ZFS_MAX_DATASET_NAME_LEN]; 3814 3815 (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname); 3816 return (zfs_unmount_snap(fullname)); 3817} 3818 3819/* 3820 * inputs: 3821 * zc_name old name of dataset 3822 * zc_value new name of dataset 3823 * zc_cookie recursive flag (only valid for snapshots) 3824 * 3825 * outputs: none 3826 */ 3827static int 3828zfs_ioc_rename(zfs_cmd_t *zc) 3829{ 3830 boolean_t recursive = zc->zc_cookie & 1; 3831 char *at; 3832 boolean_t allow_mounted = B_TRUE; 3833 3834#ifdef __FreeBSD__ 3835 allow_mounted = (zc->zc_cookie & 2) != 0; 3836#endif 3837 3838 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0'; 3839 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 3840 strchr(zc->zc_value, '%')) 3841 return (SET_ERROR(EINVAL)); 3842 3843 at = strchr(zc->zc_name, '@'); 3844 if (at != NULL) { 3845 /* snaps must be in same fs */ 3846 int error; 3847 3848 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1)) 3849 return (SET_ERROR(EXDEV)); 3850 *at = '\0'; 3851 if (zc->zc_objset_type == DMU_OST_ZFS && !allow_mounted) { 3852 error = dmu_objset_find(zc->zc_name, 3853 recursive_unmount, at + 1, 3854 recursive ? DS_FIND_CHILDREN : 0); 3855 if (error != 0) { 3856 *at = '@'; 3857 return (error); 3858 } 3859 } 3860 error = dsl_dataset_rename_snapshot(zc->zc_name, 3861 at + 1, strchr(zc->zc_value, '@') + 1, recursive); 3862 *at = '@'; 3863 3864 return (error); 3865 } else { 3866#ifdef illumos 3867 if (zc->zc_objset_type == DMU_OST_ZVOL) 3868 (void) zvol_remove_minor(zc->zc_name); 3869#endif 3870 return (dsl_dir_rename(zc->zc_name, zc->zc_value)); 3871 } 3872} 3873 3874static int 3875zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr) 3876{ 3877 const char *propname = nvpair_name(pair); 3878 boolean_t issnap = (strchr(dsname, '@') != NULL); 3879 zfs_prop_t prop = zfs_name_to_prop(propname); 3880 uint64_t intval; 3881 int err; 3882 3883 if (prop == ZPROP_INVAL) { 3884 if (zfs_prop_user(propname)) { 3885 if (err = zfs_secpolicy_write_perms(dsname, 3886 ZFS_DELEG_PERM_USERPROP, cr)) 3887 return (err); 3888 return (0); 3889 } 3890 3891 if (!issnap && zfs_prop_userquota(propname)) { 3892 const char *perm = NULL; 3893 const char *uq_prefix = 3894 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA]; 3895 const char *gq_prefix = 3896 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA]; 3897 3898 if (strncmp(propname, uq_prefix, 3899 strlen(uq_prefix)) == 0) { 3900 perm = ZFS_DELEG_PERM_USERQUOTA; 3901 } else if (strncmp(propname, gq_prefix, 3902 strlen(gq_prefix)) == 0) { 3903 perm = ZFS_DELEG_PERM_GROUPQUOTA; 3904 } else { 3905 /* USERUSED and GROUPUSED are read-only */ 3906 return (SET_ERROR(EINVAL)); 3907 } 3908 3909 if (err = zfs_secpolicy_write_perms(dsname, perm, cr)) 3910 return (err); 3911 return (0); 3912 } 3913 3914 return (SET_ERROR(EINVAL)); 3915 } 3916 3917 if (issnap) 3918 return (SET_ERROR(EINVAL)); 3919 3920 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 3921 /* 3922 * dsl_prop_get_all_impl() returns properties in this 3923 * format. 3924 */ 3925 nvlist_t *attrs; 3926 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 3927 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 3928 &pair) == 0); 3929 } 3930 3931 /* 3932 * Check that this value is valid for this pool version 3933 */ 3934 switch (prop) { 3935 case ZFS_PROP_COMPRESSION: 3936 /* 3937 * If the user specified gzip compression, make sure 3938 * the SPA supports it. We ignore any errors here since 3939 * we'll catch them later. 3940 */ 3941 if (nvpair_value_uint64(pair, &intval) == 0) { 3942 if (intval >= ZIO_COMPRESS_GZIP_1 && 3943 intval <= ZIO_COMPRESS_GZIP_9 && 3944 zfs_earlier_version(dsname, 3945 SPA_VERSION_GZIP_COMPRESSION)) { 3946 return (SET_ERROR(ENOTSUP)); 3947 } 3948 3949 if (intval == ZIO_COMPRESS_ZLE && 3950 zfs_earlier_version(dsname, 3951 SPA_VERSION_ZLE_COMPRESSION)) 3952 return (SET_ERROR(ENOTSUP)); 3953 3954 if (intval == ZIO_COMPRESS_LZ4) { 3955 spa_t *spa; 3956 3957 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 3958 return (err); 3959 3960 if (!spa_feature_is_enabled(spa, 3961 SPA_FEATURE_LZ4_COMPRESS)) { 3962 spa_close(spa, FTAG); 3963 return (SET_ERROR(ENOTSUP)); 3964 } 3965 spa_close(spa, FTAG); 3966 } 3967 3968 /* 3969 * If this is a bootable dataset then 3970 * verify that the compression algorithm 3971 * is supported for booting. We must return 3972 * something other than ENOTSUP since it 3973 * implies a downrev pool version. 3974 */ 3975 if (zfs_is_bootfs(dsname) && 3976 !BOOTFS_COMPRESS_VALID(intval)) { 3977 return (SET_ERROR(ERANGE)); 3978 } 3979 } 3980 break; 3981 3982 case ZFS_PROP_COPIES: 3983 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS)) 3984 return (SET_ERROR(ENOTSUP)); 3985 break; 3986 3987 case ZFS_PROP_RECORDSIZE: 3988 /* Record sizes above 128k need the feature to be enabled */ 3989 if (nvpair_value_uint64(pair, &intval) == 0 && 3990 intval > SPA_OLD_MAXBLOCKSIZE) { 3991 spa_t *spa; 3992 3993 /* 3994 * If this is a bootable dataset then 3995 * the we don't allow large (>128K) blocks, 3996 * because GRUB doesn't support them. 3997 */ 3998 if (zfs_is_bootfs(dsname) && 3999 intval > SPA_OLD_MAXBLOCKSIZE) { 4000 return (SET_ERROR(ERANGE)); 4001 } 4002 4003 /* 4004 * We don't allow setting the property above 1MB, 4005 * unless the tunable has been changed. 4006 */ 4007 if (intval > zfs_max_recordsize || 4008 intval > SPA_MAXBLOCKSIZE) 4009 return (SET_ERROR(ERANGE)); 4010 4011 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 4012 return (err); 4013 4014 if (!spa_feature_is_enabled(spa, 4015 SPA_FEATURE_LARGE_BLOCKS)) { 4016 spa_close(spa, FTAG); 4017 return (SET_ERROR(ENOTSUP)); 4018 } 4019 spa_close(spa, FTAG); 4020 } 4021 break; 4022 4023 case ZFS_PROP_SHARESMB: 4024 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID)) 4025 return (SET_ERROR(ENOTSUP)); 4026 break; 4027 4028 case ZFS_PROP_ACLINHERIT: 4029 if (nvpair_type(pair) == DATA_TYPE_UINT64 && 4030 nvpair_value_uint64(pair, &intval) == 0) { 4031 if (intval == ZFS_ACL_PASSTHROUGH_X && 4032 zfs_earlier_version(dsname, 4033 SPA_VERSION_PASSTHROUGH_X)) 4034 return (SET_ERROR(ENOTSUP)); 4035 } 4036 break; 4037 4038 case ZFS_PROP_CHECKSUM: 4039 case ZFS_PROP_DEDUP: 4040 { 4041 spa_feature_t feature; 4042 spa_t *spa; 4043 4044 /* dedup feature version checks */ 4045 if (prop == ZFS_PROP_DEDUP && 4046 zfs_earlier_version(dsname, SPA_VERSION_DEDUP)) 4047 return (SET_ERROR(ENOTSUP)); 4048 4049 if (nvpair_value_uint64(pair, &intval) != 0) 4050 return (SET_ERROR(EINVAL)); 4051 4052 /* check prop value is enabled in features */ 4053 feature = zio_checksum_to_feature(intval & ZIO_CHECKSUM_MASK); 4054 if (feature == SPA_FEATURE_NONE) 4055 break; 4056 4057 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 4058 return (err); 4059 /* 4060 * Salted checksums are not supported on root pools. 4061 */ 4062 if (spa_bootfs(spa) != 0 && 4063 intval < ZIO_CHECKSUM_FUNCTIONS && 4064 (zio_checksum_table[intval].ci_flags & 4065 ZCHECKSUM_FLAG_SALTED)) { 4066 spa_close(spa, FTAG); 4067 return (SET_ERROR(ERANGE)); 4068 } 4069 if (!spa_feature_is_enabled(spa, feature)) { 4070 spa_close(spa, FTAG); 4071 return (SET_ERROR(ENOTSUP)); 4072 } 4073 spa_close(spa, FTAG); 4074 break; 4075 } 4076 } 4077 4078 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED())); 4079} 4080 4081/* 4082 * Checks for a race condition to make sure we don't increment a feature flag 4083 * multiple times. 4084 */ 4085static int 4086zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx) 4087{ 4088 spa_t *spa = dmu_tx_pool(tx)->dp_spa; 4089 spa_feature_t *featurep = arg; 4090 4091 if (!spa_feature_is_active(spa, *featurep)) 4092 return (0); 4093 else 4094 return (SET_ERROR(EBUSY)); 4095} 4096 4097/* 4098 * The callback invoked on feature activation in the sync task caused by 4099 * zfs_prop_activate_feature. 4100 */ 4101static void 4102zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx) 4103{ 4104 spa_t *spa = dmu_tx_pool(tx)->dp_spa; 4105 spa_feature_t *featurep = arg; 4106 4107 spa_feature_incr(spa, *featurep, tx); 4108} 4109 4110/* 4111 * Activates a feature on a pool in response to a property setting. This 4112 * creates a new sync task which modifies the pool to reflect the feature 4113 * as being active. 4114 */ 4115static int 4116zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature) 4117{ 4118 int err; 4119 4120 /* EBUSY here indicates that the feature is already active */ 4121 err = dsl_sync_task(spa_name(spa), 4122 zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync, 4123 &feature, 2, ZFS_SPACE_CHECK_RESERVED); 4124 4125 if (err != 0 && err != EBUSY) 4126 return (err); 4127 else 4128 return (0); 4129} 4130 4131/* 4132 * Removes properties from the given props list that fail permission checks 4133 * needed to clear them and to restore them in case of a receive error. For each 4134 * property, make sure we have both set and inherit permissions. 4135 * 4136 * Returns the first error encountered if any permission checks fail. If the 4137 * caller provides a non-NULL errlist, it also gives the complete list of names 4138 * of all the properties that failed a permission check along with the 4139 * corresponding error numbers. The caller is responsible for freeing the 4140 * returned errlist. 4141 * 4142 * If every property checks out successfully, zero is returned and the list 4143 * pointed at by errlist is NULL. 4144 */ 4145static int 4146zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist) 4147{ 4148 zfs_cmd_t *zc; 4149 nvpair_t *pair, *next_pair; 4150 nvlist_t *errors; 4151 int err, rv = 0; 4152 4153 if (props == NULL) 4154 return (0); 4155 4156 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0); 4157 4158 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP); 4159 (void) strcpy(zc->zc_name, dataset); 4160 pair = nvlist_next_nvpair(props, NULL); 4161 while (pair != NULL) { 4162 next_pair = nvlist_next_nvpair(props, pair); 4163 4164 (void) strcpy(zc->zc_value, nvpair_name(pair)); 4165 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 || 4166 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) { 4167 VERIFY(nvlist_remove_nvpair(props, pair) == 0); 4168 VERIFY(nvlist_add_int32(errors, 4169 zc->zc_value, err) == 0); 4170 } 4171 pair = next_pair; 4172 } 4173 kmem_free(zc, sizeof (zfs_cmd_t)); 4174 4175 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) { 4176 nvlist_free(errors); 4177 errors = NULL; 4178 } else { 4179 VERIFY(nvpair_value_int32(pair, &rv) == 0); 4180 } 4181 4182 if (errlist == NULL) 4183 nvlist_free(errors); 4184 else 4185 *errlist = errors; 4186 4187 return (rv); 4188} 4189 4190static boolean_t 4191propval_equals(nvpair_t *p1, nvpair_t *p2) 4192{ 4193 if (nvpair_type(p1) == DATA_TYPE_NVLIST) { 4194 /* dsl_prop_get_all_impl() format */ 4195 nvlist_t *attrs; 4196 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0); 4197 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 4198 &p1) == 0); 4199 } 4200 4201 if (nvpair_type(p2) == DATA_TYPE_NVLIST) { 4202 nvlist_t *attrs; 4203 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0); 4204 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 4205 &p2) == 0); 4206 } 4207 4208 if (nvpair_type(p1) != nvpair_type(p2)) 4209 return (B_FALSE); 4210 4211 if (nvpair_type(p1) == DATA_TYPE_STRING) { 4212 char *valstr1, *valstr2; 4213 4214 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0); 4215 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0); 4216 return (strcmp(valstr1, valstr2) == 0); 4217 } else { 4218 uint64_t intval1, intval2; 4219 4220 VERIFY(nvpair_value_uint64(p1, &intval1) == 0); 4221 VERIFY(nvpair_value_uint64(p2, &intval2) == 0); 4222 return (intval1 == intval2); 4223 } 4224} 4225 4226/* 4227 * Remove properties from props if they are not going to change (as determined 4228 * by comparison with origprops). Remove them from origprops as well, since we 4229 * do not need to clear or restore properties that won't change. 4230 */ 4231static void 4232props_reduce(nvlist_t *props, nvlist_t *origprops) 4233{ 4234 nvpair_t *pair, *next_pair; 4235 4236 if (origprops == NULL) 4237 return; /* all props need to be received */ 4238 4239 pair = nvlist_next_nvpair(props, NULL); 4240 while (pair != NULL) { 4241 const char *propname = nvpair_name(pair); 4242 nvpair_t *match; 4243 4244 next_pair = nvlist_next_nvpair(props, pair); 4245 4246 if ((nvlist_lookup_nvpair(origprops, propname, 4247 &match) != 0) || !propval_equals(pair, match)) 4248 goto next; /* need to set received value */ 4249 4250 /* don't clear the existing received value */ 4251 (void) nvlist_remove_nvpair(origprops, match); 4252 /* don't bother receiving the property */ 4253 (void) nvlist_remove_nvpair(props, pair); 4254next: 4255 pair = next_pair; 4256 } 4257} 4258 4259/* 4260 * Extract properties that cannot be set PRIOR to the receipt of a dataset. 4261 * For example, refquota cannot be set until after the receipt of a dataset, 4262 * because in replication streams, an older/earlier snapshot may exceed the 4263 * refquota. We want to receive the older/earlier snapshot, but setting 4264 * refquota pre-receipt will set the dsl's ACTUAL quota, which will prevent 4265 * the older/earlier snapshot from being received (with EDQUOT). 4266 * 4267 * The ZFS test "zfs_receive_011_pos" demonstrates such a scenario. 4268 * 4269 * libzfs will need to be judicious handling errors encountered by props 4270 * extracted by this function. 4271 */ 4272static nvlist_t * 4273extract_delay_props(nvlist_t *props) 4274{ 4275 nvlist_t *delayprops; 4276 nvpair_t *nvp, *tmp; 4277 static const zfs_prop_t delayable[] = { ZFS_PROP_REFQUOTA, 0 }; 4278 int i; 4279 4280 VERIFY(nvlist_alloc(&delayprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 4281 4282 for (nvp = nvlist_next_nvpair(props, NULL); nvp != NULL; 4283 nvp = nvlist_next_nvpair(props, nvp)) { 4284 /* 4285 * strcmp() is safe because zfs_prop_to_name() always returns 4286 * a bounded string. 4287 */ 4288 for (i = 0; delayable[i] != 0; i++) { 4289 if (strcmp(zfs_prop_to_name(delayable[i]), 4290 nvpair_name(nvp)) == 0) { 4291 break; 4292 } 4293 } 4294 if (delayable[i] != 0) { 4295 tmp = nvlist_prev_nvpair(props, nvp); 4296 VERIFY(nvlist_add_nvpair(delayprops, nvp) == 0); 4297 VERIFY(nvlist_remove_nvpair(props, nvp) == 0); 4298 nvp = tmp; 4299 } 4300 } 4301 4302 if (nvlist_empty(delayprops)) { 4303 nvlist_free(delayprops); 4304 delayprops = NULL; 4305 } 4306 return (delayprops); 4307} 4308 4309#ifdef DEBUG 4310static boolean_t zfs_ioc_recv_inject_err; 4311#endif 4312 4313/* 4314 * inputs: 4315 * zc_name name of containing filesystem 4316 * zc_nvlist_src{_size} nvlist of properties to apply 4317 * zc_value name of snapshot to create 4318 * zc_string name of clone origin (if DRR_FLAG_CLONE) 4319 * zc_cookie file descriptor to recv from 4320 * zc_begin_record the BEGIN record of the stream (not byteswapped) 4321 * zc_guid force flag 4322 * zc_cleanup_fd cleanup-on-exit file descriptor 4323 * zc_action_handle handle for this guid/ds mapping (or zero on first call) 4324 * zc_resumable if data is incomplete assume sender will resume 4325 * 4326 * outputs: 4327 * zc_cookie number of bytes read 4328 * zc_nvlist_dst{_size} error for each unapplied received property 4329 * zc_obj zprop_errflags_t 4330 * zc_action_handle handle for this guid/ds mapping 4331 */ 4332static int 4333zfs_ioc_recv(zfs_cmd_t *zc) 4334{ 4335 file_t *fp; 4336 dmu_recv_cookie_t drc; 4337 boolean_t force = (boolean_t)zc->zc_guid; 4338 int fd; 4339 int error = 0; 4340 int props_error = 0; 4341 nvlist_t *errors; 4342 offset_t off; 4343 nvlist_t *props = NULL; /* sent properties */ 4344 nvlist_t *origprops = NULL; /* existing properties */ 4345 nvlist_t *delayprops = NULL; /* sent properties applied post-receive */ 4346 char *origin = NULL; 4347 char *tosnap; 4348 char tofs[ZFS_MAX_DATASET_NAME_LEN]; 4349 cap_rights_t rights; 4350 boolean_t first_recvd_props = B_FALSE; 4351 4352 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 4353 strchr(zc->zc_value, '@') == NULL || 4354 strchr(zc->zc_value, '%')) 4355 return (SET_ERROR(EINVAL)); 4356 4357 (void) strcpy(tofs, zc->zc_value); 4358 tosnap = strchr(tofs, '@'); 4359 *tosnap++ = '\0'; 4360 4361 if (zc->zc_nvlist_src != 0 && 4362 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 4363 zc->zc_iflags, &props)) != 0) 4364 return (error); 4365 4366 fd = zc->zc_cookie; 4367#ifdef illumos 4368 fp = getf(fd); 4369#else 4370 fget_read(curthread, fd, cap_rights_init(&rights, CAP_PREAD), &fp); 4371#endif 4372 if (fp == NULL) { 4373 nvlist_free(props); 4374 return (SET_ERROR(EBADF)); 4375 } 4376 4377 errors = fnvlist_alloc(); 4378 4379 if (zc->zc_string[0]) 4380 origin = zc->zc_string; 4381 4382 error = dmu_recv_begin(tofs, tosnap, 4383 &zc->zc_begin_record, force, zc->zc_resumable, origin, &drc); 4384 if (error != 0) 4385 goto out; 4386 4387 /* 4388 * Set properties before we receive the stream so that they are applied 4389 * to the new data. Note that we must call dmu_recv_stream() if 4390 * dmu_recv_begin() succeeds. 4391 */ 4392 if (props != NULL && !drc.drc_newfs) { 4393 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >= 4394 SPA_VERSION_RECVD_PROPS && 4395 !dsl_prop_get_hasrecvd(tofs)) 4396 first_recvd_props = B_TRUE; 4397 4398 /* 4399 * If new received properties are supplied, they are to 4400 * completely replace the existing received properties, so stash 4401 * away the existing ones. 4402 */ 4403 if (dsl_prop_get_received(tofs, &origprops) == 0) { 4404 nvlist_t *errlist = NULL; 4405 /* 4406 * Don't bother writing a property if its value won't 4407 * change (and avoid the unnecessary security checks). 4408 * 4409 * The first receive after SPA_VERSION_RECVD_PROPS is a 4410 * special case where we blow away all local properties 4411 * regardless. 4412 */ 4413 if (!first_recvd_props) 4414 props_reduce(props, origprops); 4415 if (zfs_check_clearable(tofs, origprops, &errlist) != 0) 4416 (void) nvlist_merge(errors, errlist, 0); 4417 nvlist_free(errlist); 4418 4419 if (clear_received_props(tofs, origprops, 4420 first_recvd_props ? NULL : props) != 0) 4421 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 4422 } else { 4423 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 4424 } 4425 } 4426 4427 if (props != NULL) { 4428 props_error = dsl_prop_set_hasrecvd(tofs); 4429 4430 if (props_error == 0) { 4431 delayprops = extract_delay_props(props); 4432 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED, 4433 props, errors); 4434 } 4435 } 4436 4437 off = fp->f_offset; 4438 error = dmu_recv_stream(&drc, fp, &off, zc->zc_cleanup_fd, 4439 &zc->zc_action_handle); 4440 4441 if (error == 0) { 4442 zfsvfs_t *zfsvfs = NULL; 4443 4444 if (getzfsvfs(tofs, &zfsvfs) == 0) { 4445 /* online recv */ 4446 dsl_dataset_t *ds; 4447 int end_err; 4448 4449 ds = dmu_objset_ds(zfsvfs->z_os); 4450 error = zfs_suspend_fs(zfsvfs); 4451 /* 4452 * If the suspend fails, then the recv_end will 4453 * likely also fail, and clean up after itself. 4454 */ 4455 end_err = dmu_recv_end(&drc, zfsvfs); 4456 if (error == 0) 4457 error = zfs_resume_fs(zfsvfs, ds); 4458 error = error ? error : end_err; 4459#ifdef illumos 4460 VFS_RELE(zfsvfs->z_vfs); 4461#else 4462 vfs_unbusy(zfsvfs->z_vfs); 4463#endif 4464 } else { 4465 error = dmu_recv_end(&drc, NULL); 4466 } 4467 4468 /* Set delayed properties now, after we're done receiving. */ 4469 if (delayprops != NULL && error == 0) { 4470 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED, 4471 delayprops, errors); 4472 } 4473 } 4474 4475 if (delayprops != NULL) { 4476 /* 4477 * Merge delayed props back in with initial props, in case 4478 * we're DEBUG and zfs_ioc_recv_inject_err is set (which means 4479 * we have to make sure clear_received_props() includes 4480 * the delayed properties). 4481 * 4482 * Since zfs_ioc_recv_inject_err is only in DEBUG kernels, 4483 * using ASSERT() will be just like a VERIFY. 4484 */ 4485 ASSERT(nvlist_merge(props, delayprops, 0) == 0); 4486 nvlist_free(delayprops); 4487 } 4488 4489 /* 4490 * Now that all props, initial and delayed, are set, report the prop 4491 * errors to the caller. 4492 */ 4493 if (zc->zc_nvlist_dst_size != 0 && 4494 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 || 4495 put_nvlist(zc, errors) != 0)) { 4496 /* 4497 * Caller made zc->zc_nvlist_dst less than the minimum expected 4498 * size or supplied an invalid address. 4499 */ 4500 props_error = SET_ERROR(EINVAL); 4501 } 4502 4503 zc->zc_cookie = off - fp->f_offset; 4504 if (off >= 0 && off <= MAXOFFSET_T) 4505 fp->f_offset = off; 4506 4507#ifdef DEBUG 4508 if (zfs_ioc_recv_inject_err) { 4509 zfs_ioc_recv_inject_err = B_FALSE; 4510 error = 1; 4511 } 4512#endif 4513 4514#ifdef __FreeBSD__ 4515 if (error == 0) 4516 zvol_create_minors(tofs); 4517#endif 4518 4519 /* 4520 * On error, restore the original props. 4521 */ 4522 if (error != 0 && props != NULL && !drc.drc_newfs) { 4523 if (clear_received_props(tofs, props, NULL) != 0) { 4524 /* 4525 * We failed to clear the received properties. 4526 * Since we may have left a $recvd value on the 4527 * system, we can't clear the $hasrecvd flag. 4528 */ 4529 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4530 } else if (first_recvd_props) { 4531 dsl_prop_unset_hasrecvd(tofs); 4532 } 4533 4534 if (origprops == NULL && !drc.drc_newfs) { 4535 /* We failed to stash the original properties. */ 4536 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4537 } 4538 4539 /* 4540 * dsl_props_set() will not convert RECEIVED to LOCAL on or 4541 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL 4542 * explictly if we're restoring local properties cleared in the 4543 * first new-style receive. 4544 */ 4545 if (origprops != NULL && 4546 zfs_set_prop_nvlist(tofs, (first_recvd_props ? 4547 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED), 4548 origprops, NULL) != 0) { 4549 /* 4550 * We stashed the original properties but failed to 4551 * restore them. 4552 */ 4553 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4554 } 4555 } 4556out: 4557 nvlist_free(props); 4558 nvlist_free(origprops); 4559 nvlist_free(errors); 4560 releasef(fd); 4561 4562 if (error == 0) 4563 error = props_error; 4564 4565 return (error); 4566} 4567 4568/* 4569 * inputs: 4570 * zc_name name of snapshot to send 4571 * zc_cookie file descriptor to send stream to 4572 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj) 4573 * zc_sendobj objsetid of snapshot to send 4574 * zc_fromobj objsetid of incremental fromsnap (may be zero) 4575 * zc_guid if set, estimate size of stream only. zc_cookie is ignored. 4576 * output size in zc_objset_type. 4577 * zc_flags lzc_send_flags 4578 * 4579 * outputs: 4580 * zc_objset_type estimated size, if zc_guid is set 4581 */ 4582static int 4583zfs_ioc_send(zfs_cmd_t *zc) 4584{ 4585 int error; 4586 offset_t off; 4587 boolean_t estimate = (zc->zc_guid != 0); 4588 boolean_t embedok = (zc->zc_flags & 0x1); 4589 boolean_t large_block_ok = (zc->zc_flags & 0x2); 4590 4591 if (zc->zc_obj != 0) { 4592 dsl_pool_t *dp; 4593 dsl_dataset_t *tosnap; 4594 4595 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4596 if (error != 0) 4597 return (error); 4598 4599 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap); 4600 if (error != 0) { 4601 dsl_pool_rele(dp, FTAG); 4602 return (error); 4603 } 4604 4605 if (dsl_dir_is_clone(tosnap->ds_dir)) 4606 zc->zc_fromobj = 4607 dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj; 4608 dsl_dataset_rele(tosnap, FTAG); 4609 dsl_pool_rele(dp, FTAG); 4610 } 4611 4612 if (estimate) { 4613 dsl_pool_t *dp; 4614 dsl_dataset_t *tosnap; 4615 dsl_dataset_t *fromsnap = NULL; 4616 4617 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4618 if (error != 0) 4619 return (error); 4620 4621 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap); 4622 if (error != 0) { 4623 dsl_pool_rele(dp, FTAG); 4624 return (error); 4625 } 4626 4627 if (zc->zc_fromobj != 0) { 4628 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj, 4629 FTAG, &fromsnap); 4630 if (error != 0) { 4631 dsl_dataset_rele(tosnap, FTAG); 4632 dsl_pool_rele(dp, FTAG); 4633 return (error); 4634 } 4635 } 4636 4637 error = dmu_send_estimate(tosnap, fromsnap, 4638 &zc->zc_objset_type); 4639 4640 if (fromsnap != NULL) 4641 dsl_dataset_rele(fromsnap, FTAG); 4642 dsl_dataset_rele(tosnap, FTAG); 4643 dsl_pool_rele(dp, FTAG); 4644 } else { 4645 file_t *fp; 4646 cap_rights_t rights; 4647 4648#ifdef illumos 4649 fp = getf(zc->zc_cookie); 4650#else 4651 fget_write(curthread, zc->zc_cookie, 4652 cap_rights_init(&rights, CAP_WRITE), &fp); 4653#endif 4654 if (fp == NULL) 4655 return (SET_ERROR(EBADF)); 4656 4657 off = fp->f_offset; 4658 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj, 4659 zc->zc_fromobj, embedok, large_block_ok, 4660#ifdef illumos 4661 zc->zc_cookie, fp->f_vnode, &off); 4662#else 4663 zc->zc_cookie, fp, &off); 4664#endif 4665 4666 if (off >= 0 && off <= MAXOFFSET_T) 4667 fp->f_offset = off; 4668 releasef(zc->zc_cookie); 4669 } 4670 return (error); 4671} 4672 4673/* 4674 * inputs: 4675 * zc_name name of snapshot on which to report progress 4676 * zc_cookie file descriptor of send stream 4677 * 4678 * outputs: 4679 * zc_cookie number of bytes written in send stream thus far 4680 */ 4681static int 4682zfs_ioc_send_progress(zfs_cmd_t *zc) 4683{ 4684 dsl_pool_t *dp; 4685 dsl_dataset_t *ds; 4686 dmu_sendarg_t *dsp = NULL; 4687 int error; 4688 4689 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4690 if (error != 0) 4691 return (error); 4692 4693 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds); 4694 if (error != 0) { 4695 dsl_pool_rele(dp, FTAG); 4696 return (error); 4697 } 4698 4699 mutex_enter(&ds->ds_sendstream_lock); 4700 4701 /* 4702 * Iterate over all the send streams currently active on this dataset. 4703 * If there's one which matches the specified file descriptor _and_ the 4704 * stream was started by the current process, return the progress of 4705 * that stream. 4706 */ 4707 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL; 4708 dsp = list_next(&ds->ds_sendstreams, dsp)) { 4709 if (dsp->dsa_outfd == zc->zc_cookie && 4710 dsp->dsa_proc == curproc) 4711 break; 4712 } 4713 4714 if (dsp != NULL) 4715 zc->zc_cookie = *(dsp->dsa_off); 4716 else 4717 error = SET_ERROR(ENOENT); 4718 4719 mutex_exit(&ds->ds_sendstream_lock); 4720 dsl_dataset_rele(ds, FTAG); 4721 dsl_pool_rele(dp, FTAG); 4722 return (error); 4723} 4724 4725static int 4726zfs_ioc_inject_fault(zfs_cmd_t *zc) 4727{ 4728 int id, error; 4729 4730 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id, 4731 &zc->zc_inject_record); 4732 4733 if (error == 0) 4734 zc->zc_guid = (uint64_t)id; 4735 4736 return (error); 4737} 4738 4739static int 4740zfs_ioc_clear_fault(zfs_cmd_t *zc) 4741{ 4742 return (zio_clear_fault((int)zc->zc_guid)); 4743} 4744 4745static int 4746zfs_ioc_inject_list_next(zfs_cmd_t *zc) 4747{ 4748 int id = (int)zc->zc_guid; 4749 int error; 4750 4751 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name), 4752 &zc->zc_inject_record); 4753 4754 zc->zc_guid = id; 4755 4756 return (error); 4757} 4758 4759static int 4760zfs_ioc_error_log(zfs_cmd_t *zc) 4761{ 4762 spa_t *spa; 4763 int error; 4764 size_t count = (size_t)zc->zc_nvlist_dst_size; 4765 4766 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 4767 return (error); 4768 4769 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst, 4770 &count); 4771 if (error == 0) 4772 zc->zc_nvlist_dst_size = count; 4773 else 4774 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa); 4775 4776 spa_close(spa, FTAG); 4777 4778 return (error); 4779} 4780 4781static int 4782zfs_ioc_clear(zfs_cmd_t *zc) 4783{ 4784 spa_t *spa; 4785 vdev_t *vd; 4786 int error; 4787 4788 /* 4789 * On zpool clear we also fix up missing slogs 4790 */ 4791 mutex_enter(&spa_namespace_lock); 4792 spa = spa_lookup(zc->zc_name); 4793 if (spa == NULL) { 4794 mutex_exit(&spa_namespace_lock); 4795 return (SET_ERROR(EIO)); 4796 } 4797 if (spa_get_log_state(spa) == SPA_LOG_MISSING) { 4798 /* we need to let spa_open/spa_load clear the chains */ 4799 spa_set_log_state(spa, SPA_LOG_CLEAR); 4800 } 4801 spa->spa_last_open_failed = 0; 4802 mutex_exit(&spa_namespace_lock); 4803 4804 if (zc->zc_cookie & ZPOOL_NO_REWIND) { 4805 error = spa_open(zc->zc_name, &spa, FTAG); 4806 } else { 4807 nvlist_t *policy; 4808 nvlist_t *config = NULL; 4809 4810 if (zc->zc_nvlist_src == 0) 4811 return (SET_ERROR(EINVAL)); 4812 4813 if ((error = get_nvlist(zc->zc_nvlist_src, 4814 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) { 4815 error = spa_open_rewind(zc->zc_name, &spa, FTAG, 4816 policy, &config); 4817 if (config != NULL) { 4818 int err; 4819 4820 if ((err = put_nvlist(zc, config)) != 0) 4821 error = err; 4822 nvlist_free(config); 4823 } 4824 nvlist_free(policy); 4825 } 4826 } 4827 4828 if (error != 0) 4829 return (error); 4830 4831 spa_vdev_state_enter(spa, SCL_NONE); 4832 4833 if (zc->zc_guid == 0) { 4834 vd = NULL; 4835 } else { 4836 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE); 4837 if (vd == NULL) { 4838 (void) spa_vdev_state_exit(spa, NULL, ENODEV); 4839 spa_close(spa, FTAG); 4840 return (SET_ERROR(ENODEV)); 4841 } 4842 } 4843 4844 vdev_clear(spa, vd); 4845 4846 (void) spa_vdev_state_exit(spa, NULL, 0); 4847 4848 /* 4849 * Resume any suspended I/Os. 4850 */ 4851 if (zio_resume(spa) != 0) 4852 error = SET_ERROR(EIO); 4853 4854 spa_close(spa, FTAG); 4855 4856 return (error); 4857} 4858 4859static int 4860zfs_ioc_pool_reopen(zfs_cmd_t *zc) 4861{ 4862 spa_t *spa; 4863 int error; 4864 4865 error = spa_open(zc->zc_name, &spa, FTAG); 4866 if (error != 0) 4867 return (error); 4868 4869 spa_vdev_state_enter(spa, SCL_NONE); 4870 4871 /* 4872 * If a resilver is already in progress then set the 4873 * spa_scrub_reopen flag to B_TRUE so that we don't restart 4874 * the scan as a side effect of the reopen. Otherwise, let 4875 * vdev_open() decided if a resilver is required. 4876 */ 4877 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool); 4878 vdev_reopen(spa->spa_root_vdev); 4879 spa->spa_scrub_reopen = B_FALSE; 4880 4881 (void) spa_vdev_state_exit(spa, NULL, 0); 4882 spa_close(spa, FTAG); 4883 return (0); 4884} 4885/* 4886 * inputs: 4887 * zc_name name of filesystem 4888 * zc_value name of origin snapshot 4889 * 4890 * outputs: 4891 * zc_string name of conflicting snapshot, if there is one 4892 */ 4893static int 4894zfs_ioc_promote(zfs_cmd_t *zc) 4895{ 4896 char *cp; 4897 4898 /* 4899 * We don't need to unmount *all* the origin fs's snapshots, but 4900 * it's easier. 4901 */ 4902 cp = strchr(zc->zc_value, '@'); 4903 if (cp) 4904 *cp = '\0'; 4905 (void) dmu_objset_find(zc->zc_value, 4906 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS); 4907 return (dsl_dataset_promote(zc->zc_name, zc->zc_string)); 4908} 4909 4910/* 4911 * Retrieve a single {user|group}{used|quota}@... property. 4912 * 4913 * inputs: 4914 * zc_name name of filesystem 4915 * zc_objset_type zfs_userquota_prop_t 4916 * zc_value domain name (eg. "S-1-234-567-89") 4917 * zc_guid RID/UID/GID 4918 * 4919 * outputs: 4920 * zc_cookie property value 4921 */ 4922static int 4923zfs_ioc_userspace_one(zfs_cmd_t *zc) 4924{ 4925 zfsvfs_t *zfsvfs; 4926 int error; 4927 4928 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 4929 return (SET_ERROR(EINVAL)); 4930 4931 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE); 4932 if (error != 0) 4933 return (error); 4934 4935 error = zfs_userspace_one(zfsvfs, 4936 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie); 4937 zfsvfs_rele(zfsvfs, FTAG); 4938 4939 return (error); 4940} 4941 4942/* 4943 * inputs: 4944 * zc_name name of filesystem 4945 * zc_cookie zap cursor 4946 * zc_objset_type zfs_userquota_prop_t 4947 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist) 4948 * 4949 * outputs: 4950 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t) 4951 * zc_cookie zap cursor 4952 */ 4953static int 4954zfs_ioc_userspace_many(zfs_cmd_t *zc) 4955{ 4956 zfsvfs_t *zfsvfs; 4957 int bufsize = zc->zc_nvlist_dst_size; 4958 4959 if (bufsize <= 0) 4960 return (SET_ERROR(ENOMEM)); 4961 4962 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE); 4963 if (error != 0) 4964 return (error); 4965 4966 void *buf = kmem_alloc(bufsize, KM_SLEEP); 4967 4968 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie, 4969 buf, &zc->zc_nvlist_dst_size); 4970 4971 if (error == 0) { 4972 error = ddi_copyout(buf, 4973 (void *)(uintptr_t)zc->zc_nvlist_dst, 4974 zc->zc_nvlist_dst_size, zc->zc_iflags); 4975 } 4976 kmem_free(buf, bufsize); 4977 zfsvfs_rele(zfsvfs, FTAG); 4978 4979 return (error); 4980} 4981 4982/* 4983 * inputs: 4984 * zc_name name of filesystem 4985 * 4986 * outputs: 4987 * none 4988 */ 4989static int 4990zfs_ioc_userspace_upgrade(zfs_cmd_t *zc) 4991{ 4992 objset_t *os; 4993 int error = 0; 4994 zfsvfs_t *zfsvfs; 4995 4996 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) { 4997 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) { 4998 /* 4999 * If userused is not enabled, it may be because the 5000 * objset needs to be closed & reopened (to grow the 5001 * objset_phys_t). Suspend/resume the fs will do that. 5002 */ 5003 dsl_dataset_t *ds; 5004 5005 ds = dmu_objset_ds(zfsvfs->z_os); 5006 error = zfs_suspend_fs(zfsvfs); 5007 if (error == 0) { 5008 dmu_objset_refresh_ownership(zfsvfs->z_os, 5009 zfsvfs); 5010 error = zfs_resume_fs(zfsvfs, ds); 5011 } 5012 } 5013 if (error == 0) 5014 error = dmu_objset_userspace_upgrade(zfsvfs->z_os); 5015#ifdef illumos 5016 VFS_RELE(zfsvfs->z_vfs); 5017#else 5018 vfs_unbusy(zfsvfs->z_vfs); 5019#endif 5020 } else { 5021 /* XXX kind of reading contents without owning */ 5022 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 5023 if (error != 0) 5024 return (error); 5025 5026 error = dmu_objset_userspace_upgrade(os); 5027 dmu_objset_rele(os, FTAG); 5028 } 5029 5030 return (error); 5031} 5032 5033#ifdef illumos 5034/* 5035 * We don't want to have a hard dependency 5036 * against some special symbols in sharefs 5037 * nfs, and smbsrv. Determine them if needed when 5038 * the first file system is shared. 5039 * Neither sharefs, nfs or smbsrv are unloadable modules. 5040 */ 5041int (*znfsexport_fs)(void *arg); 5042int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t); 5043int (*zsmbexport_fs)(void *arg, boolean_t add_share); 5044 5045int zfs_nfsshare_inited; 5046int zfs_smbshare_inited; 5047 5048ddi_modhandle_t nfs_mod; 5049ddi_modhandle_t sharefs_mod; 5050ddi_modhandle_t smbsrv_mod; 5051#endif /* illumos */ 5052kmutex_t zfs_share_lock; 5053 5054#ifdef illumos 5055static int 5056zfs_init_sharefs() 5057{ 5058 int error; 5059 5060 ASSERT(MUTEX_HELD(&zfs_share_lock)); 5061 /* Both NFS and SMB shares also require sharetab support. */ 5062 if (sharefs_mod == NULL && ((sharefs_mod = 5063 ddi_modopen("fs/sharefs", 5064 KRTLD_MODE_FIRST, &error)) == NULL)) { 5065 return (SET_ERROR(ENOSYS)); 5066 } 5067 if (zshare_fs == NULL && ((zshare_fs = 5068 (int (*)(enum sharefs_sys_op, share_t *, uint32_t)) 5069 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) { 5070 return (SET_ERROR(ENOSYS)); 5071 } 5072 return (0); 5073} 5074#endif /* illumos */ 5075 5076static int 5077zfs_ioc_share(zfs_cmd_t *zc) 5078{ 5079#ifdef illumos 5080 int error; 5081 int opcode; 5082 5083 switch (zc->zc_share.z_sharetype) { 5084 case ZFS_SHARE_NFS: 5085 case ZFS_UNSHARE_NFS: 5086 if (zfs_nfsshare_inited == 0) { 5087 mutex_enter(&zfs_share_lock); 5088 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs", 5089 KRTLD_MODE_FIRST, &error)) == NULL)) { 5090 mutex_exit(&zfs_share_lock); 5091 return (SET_ERROR(ENOSYS)); 5092 } 5093 if (znfsexport_fs == NULL && 5094 ((znfsexport_fs = (int (*)(void *)) 5095 ddi_modsym(nfs_mod, 5096 "nfs_export", &error)) == NULL)) { 5097 mutex_exit(&zfs_share_lock); 5098 return (SET_ERROR(ENOSYS)); 5099 } 5100 error = zfs_init_sharefs(); 5101 if (error != 0) { 5102 mutex_exit(&zfs_share_lock); 5103 return (SET_ERROR(ENOSYS)); 5104 } 5105 zfs_nfsshare_inited = 1; 5106 mutex_exit(&zfs_share_lock); 5107 } 5108 break; 5109 case ZFS_SHARE_SMB: 5110 case ZFS_UNSHARE_SMB: 5111 if (zfs_smbshare_inited == 0) { 5112 mutex_enter(&zfs_share_lock); 5113 if (smbsrv_mod == NULL && ((smbsrv_mod = 5114 ddi_modopen("drv/smbsrv", 5115 KRTLD_MODE_FIRST, &error)) == NULL)) { 5116 mutex_exit(&zfs_share_lock); 5117 return (SET_ERROR(ENOSYS)); 5118 } 5119 if (zsmbexport_fs == NULL && ((zsmbexport_fs = 5120 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod, 5121 "smb_server_share", &error)) == NULL)) { 5122 mutex_exit(&zfs_share_lock); 5123 return (SET_ERROR(ENOSYS)); 5124 } 5125 error = zfs_init_sharefs(); 5126 if (error != 0) { 5127 mutex_exit(&zfs_share_lock); 5128 return (SET_ERROR(ENOSYS)); 5129 } 5130 zfs_smbshare_inited = 1; 5131 mutex_exit(&zfs_share_lock); 5132 } 5133 break; 5134 default: 5135 return (SET_ERROR(EINVAL)); 5136 } 5137 5138 switch (zc->zc_share.z_sharetype) { 5139 case ZFS_SHARE_NFS: 5140 case ZFS_UNSHARE_NFS: 5141 if (error = 5142 znfsexport_fs((void *) 5143 (uintptr_t)zc->zc_share.z_exportdata)) 5144 return (error); 5145 break; 5146 case ZFS_SHARE_SMB: 5147 case ZFS_UNSHARE_SMB: 5148 if (error = zsmbexport_fs((void *) 5149 (uintptr_t)zc->zc_share.z_exportdata, 5150 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ? 5151 B_TRUE: B_FALSE)) { 5152 return (error); 5153 } 5154 break; 5155 } 5156 5157 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS || 5158 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ? 5159 SHAREFS_ADD : SHAREFS_REMOVE; 5160 5161 /* 5162 * Add or remove share from sharetab 5163 */ 5164 error = zshare_fs(opcode, 5165 (void *)(uintptr_t)zc->zc_share.z_sharedata, 5166 zc->zc_share.z_sharemax); 5167 5168 return (error); 5169 5170#else /* !illumos */ 5171 return (ENOSYS); 5172#endif /* illumos */ 5173} 5174 5175ace_t full_access[] = { 5176 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0} 5177}; 5178 5179/* 5180 * inputs: 5181 * zc_name name of containing filesystem 5182 * zc_obj object # beyond which we want next in-use object # 5183 * 5184 * outputs: 5185 * zc_obj next in-use object # 5186 */ 5187static int 5188zfs_ioc_next_obj(zfs_cmd_t *zc) 5189{ 5190 objset_t *os = NULL; 5191 int error; 5192 5193 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 5194 if (error != 0) 5195 return (error); 5196 5197 error = dmu_object_next(os, &zc->zc_obj, B_FALSE, 5198 dsl_dataset_phys(os->os_dsl_dataset)->ds_prev_snap_txg); 5199 5200 dmu_objset_rele(os, FTAG); 5201 return (error); 5202} 5203 5204/* 5205 * inputs: 5206 * zc_name name of filesystem 5207 * zc_value prefix name for snapshot 5208 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process 5209 * 5210 * outputs: 5211 * zc_value short name of new snapshot 5212 */ 5213static int 5214zfs_ioc_tmp_snapshot(zfs_cmd_t *zc) 5215{ 5216 char *snap_name; 5217 char *hold_name; 5218 int error; 5219 minor_t minor; 5220 5221 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor); 5222 if (error != 0) 5223 return (error); 5224 5225 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value, 5226 (u_longlong_t)ddi_get_lbolt64()); 5227 hold_name = kmem_asprintf("%%%s", zc->zc_value); 5228 5229 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor, 5230 hold_name); 5231 if (error == 0) 5232 (void) strcpy(zc->zc_value, snap_name); 5233 strfree(snap_name); 5234 strfree(hold_name); 5235 zfs_onexit_fd_rele(zc->zc_cleanup_fd); 5236 return (error); 5237} 5238 5239/* 5240 * inputs: 5241 * zc_name name of "to" snapshot 5242 * zc_value name of "from" snapshot 5243 * zc_cookie file descriptor to write diff data on 5244 * 5245 * outputs: 5246 * dmu_diff_record_t's to the file descriptor 5247 */ 5248static int 5249zfs_ioc_diff(zfs_cmd_t *zc) 5250{ 5251 file_t *fp; 5252 cap_rights_t rights; 5253 offset_t off; 5254 int error; 5255 5256#ifdef illumos 5257 fp = getf(zc->zc_cookie); 5258#else 5259 fget_write(curthread, zc->zc_cookie, 5260 cap_rights_init(&rights, CAP_WRITE), &fp); 5261#endif 5262 if (fp == NULL) 5263 return (SET_ERROR(EBADF)); 5264 5265 off = fp->f_offset; 5266 5267#ifdef illumos 5268 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off); 5269#else 5270 error = dmu_diff(zc->zc_name, zc->zc_value, fp, &off); 5271#endif 5272 5273 if (off >= 0 && off <= MAXOFFSET_T) 5274 fp->f_offset = off; 5275 releasef(zc->zc_cookie); 5276 5277 return (error); 5278} 5279 5280#ifdef illumos 5281/* 5282 * Remove all ACL files in shares dir 5283 */ 5284static int 5285zfs_smb_acl_purge(znode_t *dzp) 5286{ 5287 zap_cursor_t zc; 5288 zap_attribute_t zap; 5289 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 5290 int error; 5291 5292 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id); 5293 (error = zap_cursor_retrieve(&zc, &zap)) == 0; 5294 zap_cursor_advance(&zc)) { 5295 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred, 5296 NULL, 0)) != 0) 5297 break; 5298 } 5299 zap_cursor_fini(&zc); 5300 return (error); 5301} 5302#endif /* illumos */ 5303 5304static int 5305zfs_ioc_smb_acl(zfs_cmd_t *zc) 5306{ 5307#ifdef illumos 5308 vnode_t *vp; 5309 znode_t *dzp; 5310 vnode_t *resourcevp = NULL; 5311 znode_t *sharedir; 5312 zfsvfs_t *zfsvfs; 5313 nvlist_t *nvlist; 5314 char *src, *target; 5315 vattr_t vattr; 5316 vsecattr_t vsec; 5317 int error = 0; 5318 5319 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, 5320 NO_FOLLOW, NULL, &vp)) != 0) 5321 return (error); 5322 5323 /* Now make sure mntpnt and dataset are ZFS */ 5324 5325 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 || 5326 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), 5327 zc->zc_name) != 0)) { 5328 VN_RELE(vp); 5329 return (SET_ERROR(EINVAL)); 5330 } 5331 5332 dzp = VTOZ(vp); 5333 zfsvfs = dzp->z_zfsvfs; 5334 ZFS_ENTER(zfsvfs); 5335 5336 /* 5337 * Create share dir if its missing. 5338 */ 5339 mutex_enter(&zfsvfs->z_lock); 5340 if (zfsvfs->z_shares_dir == 0) { 5341 dmu_tx_t *tx; 5342 5343 tx = dmu_tx_create(zfsvfs->z_os); 5344 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE, 5345 ZFS_SHARES_DIR); 5346 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); 5347 error = dmu_tx_assign(tx, TXG_WAIT); 5348 if (error != 0) { 5349 dmu_tx_abort(tx); 5350 } else { 5351 error = zfs_create_share_dir(zfsvfs, tx); 5352 dmu_tx_commit(tx); 5353 } 5354 if (error != 0) { 5355 mutex_exit(&zfsvfs->z_lock); 5356 VN_RELE(vp); 5357 ZFS_EXIT(zfsvfs); 5358 return (error); 5359 } 5360 } 5361 mutex_exit(&zfsvfs->z_lock); 5362 5363 ASSERT(zfsvfs->z_shares_dir); 5364 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) { 5365 VN_RELE(vp); 5366 ZFS_EXIT(zfsvfs); 5367 return (error); 5368 } 5369 5370 switch (zc->zc_cookie) { 5371 case ZFS_SMB_ACL_ADD: 5372 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; 5373 vattr.va_type = VREG; 5374 vattr.va_mode = S_IFREG|0777; 5375 vattr.va_uid = 0; 5376 vattr.va_gid = 0; 5377 5378 vsec.vsa_mask = VSA_ACE; 5379 vsec.vsa_aclentp = &full_access; 5380 vsec.vsa_aclentsz = sizeof (full_access); 5381 vsec.vsa_aclcnt = 1; 5382 5383 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string, 5384 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec); 5385 if (resourcevp) 5386 VN_RELE(resourcevp); 5387 break; 5388 5389 case ZFS_SMB_ACL_REMOVE: 5390 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred, 5391 NULL, 0); 5392 break; 5393 5394 case ZFS_SMB_ACL_RENAME: 5395 if ((error = get_nvlist(zc->zc_nvlist_src, 5396 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) { 5397 VN_RELE(vp); 5398 VN_RELE(ZTOV(sharedir)); 5399 ZFS_EXIT(zfsvfs); 5400 return (error); 5401 } 5402 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) || 5403 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET, 5404 &target)) { 5405 VN_RELE(vp); 5406 VN_RELE(ZTOV(sharedir)); 5407 ZFS_EXIT(zfsvfs); 5408 nvlist_free(nvlist); 5409 return (error); 5410 } 5411 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target, 5412 kcred, NULL, 0); 5413 nvlist_free(nvlist); 5414 break; 5415 5416 case ZFS_SMB_ACL_PURGE: 5417 error = zfs_smb_acl_purge(sharedir); 5418 break; 5419 5420 default: 5421 error = SET_ERROR(EINVAL); 5422 break; 5423 } 5424 5425 VN_RELE(vp); 5426 VN_RELE(ZTOV(sharedir)); 5427 5428 ZFS_EXIT(zfsvfs); 5429 5430 return (error); 5431#else /* !illumos */ 5432 return (EOPNOTSUPP); 5433#endif /* illumos */ 5434} 5435 5436/* 5437 * innvl: { 5438 * "holds" -> { snapname -> holdname (string), ... } 5439 * (optional) "cleanup_fd" -> fd (int32) 5440 * } 5441 * 5442 * outnvl: { 5443 * snapname -> error value (int32) 5444 * ... 5445 * } 5446 */ 5447/* ARGSUSED */ 5448static int 5449zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist) 5450{ 5451 nvpair_t *pair; 5452 nvlist_t *holds; 5453 int cleanup_fd = -1; 5454 int error; 5455 minor_t minor = 0; 5456 5457 error = nvlist_lookup_nvlist(args, "holds", &holds); 5458 if (error != 0) 5459 return (SET_ERROR(EINVAL)); 5460 5461 /* make sure the user didn't pass us any invalid (empty) tags */ 5462 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL; 5463 pair = nvlist_next_nvpair(holds, pair)) { 5464 char *htag; 5465 5466 error = nvpair_value_string(pair, &htag); 5467 if (error != 0) 5468 return (SET_ERROR(error)); 5469 5470 if (strlen(htag) == 0) 5471 return (SET_ERROR(EINVAL)); 5472 } 5473 5474 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) { 5475 error = zfs_onexit_fd_hold(cleanup_fd, &minor); 5476 if (error != 0) 5477 return (error); 5478 } 5479 5480 error = dsl_dataset_user_hold(holds, minor, errlist); 5481 if (minor != 0) 5482 zfs_onexit_fd_rele(cleanup_fd); 5483 return (error); 5484} 5485 5486/* 5487 * innvl is not used. 5488 * 5489 * outnvl: { 5490 * holdname -> time added (uint64 seconds since epoch) 5491 * ... 5492 * } 5493 */ 5494/* ARGSUSED */ 5495static int 5496zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl) 5497{ 5498 return (dsl_dataset_get_holds(snapname, outnvl)); 5499} 5500 5501/* 5502 * innvl: { 5503 * snapname -> { holdname, ... } 5504 * ... 5505 * } 5506 * 5507 * outnvl: { 5508 * snapname -> error value (int32) 5509 * ... 5510 * } 5511 */ 5512/* ARGSUSED */ 5513static int 5514zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist) 5515{ 5516 return (dsl_dataset_user_release(holds, errlist)); 5517} 5518 5519/* 5520 * inputs: 5521 * zc_name name of new filesystem or snapshot 5522 * zc_value full name of old snapshot 5523 * 5524 * outputs: 5525 * zc_cookie space in bytes 5526 * zc_objset_type compressed space in bytes 5527 * zc_perm_action uncompressed space in bytes 5528 */ 5529static int 5530zfs_ioc_space_written(zfs_cmd_t *zc) 5531{ 5532 int error; 5533 dsl_pool_t *dp; 5534 dsl_dataset_t *new, *old; 5535 5536 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 5537 if (error != 0) 5538 return (error); 5539 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new); 5540 if (error != 0) { 5541 dsl_pool_rele(dp, FTAG); 5542 return (error); 5543 } 5544 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old); 5545 if (error != 0) { 5546 dsl_dataset_rele(new, FTAG); 5547 dsl_pool_rele(dp, FTAG); 5548 return (error); 5549 } 5550 5551 error = dsl_dataset_space_written(old, new, &zc->zc_cookie, 5552 &zc->zc_objset_type, &zc->zc_perm_action); 5553 dsl_dataset_rele(old, FTAG); 5554 dsl_dataset_rele(new, FTAG); 5555 dsl_pool_rele(dp, FTAG); 5556 return (error); 5557} 5558 5559/* 5560 * innvl: { 5561 * "firstsnap" -> snapshot name 5562 * } 5563 * 5564 * outnvl: { 5565 * "used" -> space in bytes 5566 * "compressed" -> compressed space in bytes 5567 * "uncompressed" -> uncompressed space in bytes 5568 * } 5569 */ 5570static int 5571zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl) 5572{ 5573 int error; 5574 dsl_pool_t *dp; 5575 dsl_dataset_t *new, *old; 5576 char *firstsnap; 5577 uint64_t used, comp, uncomp; 5578 5579 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0) 5580 return (SET_ERROR(EINVAL)); 5581 5582 error = dsl_pool_hold(lastsnap, FTAG, &dp); 5583 if (error != 0) 5584 return (error); 5585 5586 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new); 5587 if (error == 0 && !new->ds_is_snapshot) { 5588 dsl_dataset_rele(new, FTAG); 5589 error = SET_ERROR(EINVAL); 5590 } 5591 if (error != 0) { 5592 dsl_pool_rele(dp, FTAG); 5593 return (error); 5594 } 5595 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old); 5596 if (error == 0 && !old->ds_is_snapshot) { 5597 dsl_dataset_rele(old, FTAG); 5598 error = SET_ERROR(EINVAL); 5599 } 5600 if (error != 0) { 5601 dsl_dataset_rele(new, FTAG); 5602 dsl_pool_rele(dp, FTAG); 5603 return (error); 5604 } 5605 5606 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp); 5607 dsl_dataset_rele(old, FTAG); 5608 dsl_dataset_rele(new, FTAG); 5609 dsl_pool_rele(dp, FTAG); 5610 fnvlist_add_uint64(outnvl, "used", used); 5611 fnvlist_add_uint64(outnvl, "compressed", comp); 5612 fnvlist_add_uint64(outnvl, "uncompressed", uncomp); 5613 return (error); 5614} 5615 5616static int 5617zfs_ioc_jail(zfs_cmd_t *zc) 5618{ 5619 5620 return (zone_dataset_attach(curthread->td_ucred, zc->zc_name, 5621 (int)zc->zc_jailid)); 5622} 5623 5624static int 5625zfs_ioc_unjail(zfs_cmd_t *zc) 5626{ 5627 5628 return (zone_dataset_detach(curthread->td_ucred, zc->zc_name, 5629 (int)zc->zc_jailid)); 5630} 5631 5632/* 5633 * innvl: { 5634 * "fd" -> file descriptor to write stream to (int32) 5635 * (optional) "fromsnap" -> full snap name to send an incremental from 5636 * (optional) "largeblockok" -> (value ignored) 5637 * indicates that blocks > 128KB are permitted 5638 * (optional) "embedok" -> (value ignored) 5639 * presence indicates DRR_WRITE_EMBEDDED records are permitted 5640 * (optional) "resume_object" and "resume_offset" -> (uint64) 5641 * if present, resume send stream from specified object and offset. 5642 * } 5643 * 5644 * outnvl is unused 5645 */ 5646/* ARGSUSED */ 5647static int 5648zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl) 5649{ 5650 cap_rights_t rights; 5651 file_t *fp; 5652 int error; 5653 offset_t off; 5654 char *fromname = NULL; 5655 int fd; 5656 boolean_t largeblockok; 5657 boolean_t embedok; 5658 uint64_t resumeobj = 0; 5659 uint64_t resumeoff = 0; 5660 5661 error = nvlist_lookup_int32(innvl, "fd", &fd); 5662 if (error != 0) 5663 return (SET_ERROR(EINVAL)); 5664 5665 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname); 5666 5667 largeblockok = nvlist_exists(innvl, "largeblockok"); 5668 embedok = nvlist_exists(innvl, "embedok"); 5669 5670 (void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj); 5671 (void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff); 5672 5673#ifdef illumos 5674 file_t *fp = getf(fd); 5675#else 5676 fget_write(curthread, fd, cap_rights_init(&rights, CAP_WRITE), &fp); 5677#endif 5678 if (fp == NULL) 5679 return (SET_ERROR(EBADF)); 5680 5681 off = fp->f_offset; 5682 error = dmu_send(snapname, fromname, embedok, largeblockok, fd, 5683#ifdef illumos 5684 resumeobj, resumeoff, fp->f_vnode, &off); 5685#else 5686 resumeobj, resumeoff, fp, &off); 5687#endif 5688 5689#ifdef illumos 5690 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0) 5691 fp->f_offset = off; 5692#else 5693 fp->f_offset = off; 5694#endif 5695 5696 releasef(fd); 5697 return (error); 5698} 5699 5700/* 5701 * Determine approximately how large a zfs send stream will be -- the number 5702 * of bytes that will be written to the fd supplied to zfs_ioc_send_new(). 5703 * 5704 * innvl: { 5705 * (optional) "from" -> full snap or bookmark name to send an incremental 5706 * from 5707 * } 5708 * 5709 * outnvl: { 5710 * "space" -> bytes of space (uint64) 5711 * } 5712 */ 5713static int 5714zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl) 5715{ 5716 dsl_pool_t *dp; 5717 dsl_dataset_t *tosnap; 5718 int error; 5719 char *fromname; 5720 uint64_t space; 5721 5722 error = dsl_pool_hold(snapname, FTAG, &dp); 5723 if (error != 0) 5724 return (error); 5725 5726 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap); 5727 if (error != 0) { 5728 dsl_pool_rele(dp, FTAG); 5729 return (error); 5730 } 5731 5732 error = nvlist_lookup_string(innvl, "from", &fromname); 5733 if (error == 0) { 5734 if (strchr(fromname, '@') != NULL) { 5735 /* 5736 * If from is a snapshot, hold it and use the more 5737 * efficient dmu_send_estimate to estimate send space 5738 * size using deadlists. 5739 */ 5740 dsl_dataset_t *fromsnap; 5741 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap); 5742 if (error != 0) 5743 goto out; 5744 error = dmu_send_estimate(tosnap, fromsnap, &space); 5745 dsl_dataset_rele(fromsnap, FTAG); 5746 } else if (strchr(fromname, '#') != NULL) { 5747 /* 5748 * If from is a bookmark, fetch the creation TXG of the 5749 * snapshot it was created from and use that to find 5750 * blocks that were born after it. 5751 */ 5752 zfs_bookmark_phys_t frombm; 5753 5754 error = dsl_bookmark_lookup(dp, fromname, tosnap, 5755 &frombm); 5756 if (error != 0) 5757 goto out; 5758 error = dmu_send_estimate_from_txg(tosnap, 5759 frombm.zbm_creation_txg, &space); 5760 } else { 5761 /* 5762 * from is not properly formatted as a snapshot or 5763 * bookmark 5764 */ 5765 error = SET_ERROR(EINVAL); 5766 goto out; 5767 } 5768 } else { 5769 // If estimating the size of a full send, use dmu_send_estimate 5770 error = dmu_send_estimate(tosnap, NULL, &space); 5771 } 5772 5773 fnvlist_add_uint64(outnvl, "space", space); 5774 5775out: 5776 dsl_dataset_rele(tosnap, FTAG); 5777 dsl_pool_rele(dp, FTAG); 5778 return (error); 5779} 5780 5781static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST]; 5782 5783static void 5784zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5785 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck, 5786 boolean_t log_history, zfs_ioc_poolcheck_t pool_check) 5787{ 5788 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST]; 5789 5790 ASSERT3U(ioc, >=, ZFS_IOC_FIRST); 5791 ASSERT3U(ioc, <, ZFS_IOC_LAST); 5792 ASSERT3P(vec->zvec_legacy_func, ==, NULL); 5793 ASSERT3P(vec->zvec_func, ==, NULL); 5794 5795 vec->zvec_legacy_func = func; 5796 vec->zvec_secpolicy = secpolicy; 5797 vec->zvec_namecheck = namecheck; 5798 vec->zvec_allow_log = log_history; 5799 vec->zvec_pool_check = pool_check; 5800} 5801 5802/* 5803 * See the block comment at the beginning of this file for details on 5804 * each argument to this function. 5805 */ 5806static void 5807zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func, 5808 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck, 5809 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist, 5810 boolean_t allow_log) 5811{ 5812 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST]; 5813 5814 ASSERT3U(ioc, >=, ZFS_IOC_FIRST); 5815 ASSERT3U(ioc, <, ZFS_IOC_LAST); 5816 ASSERT3P(vec->zvec_legacy_func, ==, NULL); 5817 ASSERT3P(vec->zvec_func, ==, NULL); 5818 5819 /* if we are logging, the name must be valid */ 5820 ASSERT(!allow_log || namecheck != NO_NAME); 5821 5822 vec->zvec_name = name; 5823 vec->zvec_func = func; 5824 vec->zvec_secpolicy = secpolicy; 5825 vec->zvec_namecheck = namecheck; 5826 vec->zvec_pool_check = pool_check; 5827 vec->zvec_smush_outnvlist = smush_outnvlist; 5828 vec->zvec_allow_log = allow_log; 5829} 5830 5831static void 5832zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5833 zfs_secpolicy_func_t *secpolicy, boolean_t log_history, 5834 zfs_ioc_poolcheck_t pool_check) 5835{ 5836 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5837 POOL_NAME, log_history, pool_check); 5838} 5839 5840static void 5841zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5842 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check) 5843{ 5844 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5845 DATASET_NAME, B_FALSE, pool_check); 5846} 5847 5848static void 5849zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func) 5850{ 5851 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config, 5852 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5853} 5854 5855static void 5856zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5857 zfs_secpolicy_func_t *secpolicy) 5858{ 5859 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5860 NO_NAME, B_FALSE, POOL_CHECK_NONE); 5861} 5862 5863static void 5864zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc, 5865 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy) 5866{ 5867 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5868 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED); 5869} 5870 5871static void 5872zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func) 5873{ 5874 zfs_ioctl_register_dataset_read_secpolicy(ioc, func, 5875 zfs_secpolicy_read); 5876} 5877 5878static void 5879zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5880 zfs_secpolicy_func_t *secpolicy) 5881{ 5882 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5883 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5884} 5885 5886static void 5887zfs_ioctl_init(void) 5888{ 5889 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT, 5890 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME, 5891 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5892 5893 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY, 5894 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME, 5895 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE); 5896 5897 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS, 5898 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME, 5899 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5900 5901 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW, 5902 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME, 5903 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5904 5905 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE, 5906 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME, 5907 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5908 5909 zfs_ioctl_register("create", ZFS_IOC_CREATE, 5910 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME, 5911 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5912 5913 zfs_ioctl_register("clone", ZFS_IOC_CLONE, 5914 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME, 5915 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5916 5917 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS, 5918 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME, 5919 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5920 5921 zfs_ioctl_register("hold", ZFS_IOC_HOLD, 5922 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME, 5923 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5924 zfs_ioctl_register("release", ZFS_IOC_RELEASE, 5925 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME, 5926 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5927 5928 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS, 5929 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME, 5930 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5931 5932 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK, 5933 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, 5934 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE); 5935 5936 zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK, 5937 zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME, 5938 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5939 5940 zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS, 5941 zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME, 5942 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5943 5944 zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS, 5945 zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks, 5946 POOL_NAME, 5947 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5948 5949 /* IOCTLS that use the legacy function signature */ 5950 5951 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze, 5952 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY); 5953 5954 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create, 5955 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5956 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN, 5957 zfs_ioc_pool_scan); 5958 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE, 5959 zfs_ioc_pool_upgrade); 5960 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD, 5961 zfs_ioc_vdev_add); 5962 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE, 5963 zfs_ioc_vdev_remove); 5964 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE, 5965 zfs_ioc_vdev_set_state); 5966 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH, 5967 zfs_ioc_vdev_attach); 5968 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH, 5969 zfs_ioc_vdev_detach); 5970 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH, 5971 zfs_ioc_vdev_setpath); 5972 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU, 5973 zfs_ioc_vdev_setfru); 5974 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS, 5975 zfs_ioc_pool_set_props); 5976 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT, 5977 zfs_ioc_vdev_split); 5978 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID, 5979 zfs_ioc_pool_reguid); 5980 5981 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS, 5982 zfs_ioc_pool_configs, zfs_secpolicy_none); 5983 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT, 5984 zfs_ioc_pool_tryimport, zfs_secpolicy_config); 5985 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT, 5986 zfs_ioc_inject_fault, zfs_secpolicy_inject); 5987 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT, 5988 zfs_ioc_clear_fault, zfs_secpolicy_inject); 5989 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT, 5990 zfs_ioc_inject_list_next, zfs_secpolicy_inject); 5991 5992 /* 5993 * pool destroy, and export don't log the history as part of 5994 * zfsdev_ioctl, but rather zfs_ioc_pool_export 5995 * does the logging of those commands. 5996 */ 5997 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy, 5998 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE); 5999 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export, 6000 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE); 6001 6002 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats, 6003 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE); 6004 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props, 6005 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE); 6006 6007 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log, 6008 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_NONE); 6009 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME, 6010 zfs_ioc_dsobj_to_dsname, 6011 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_NONE); 6012 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY, 6013 zfs_ioc_pool_get_history, 6014 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED); 6015 6016 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import, 6017 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 6018 6019 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear, 6020 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 6021 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen, 6022 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED); 6023 6024 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN, 6025 zfs_ioc_space_written); 6026 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS, 6027 zfs_ioc_objset_recvd_props); 6028 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ, 6029 zfs_ioc_next_obj); 6030 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL, 6031 zfs_ioc_get_fsacl); 6032 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS, 6033 zfs_ioc_objset_stats); 6034 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS, 6035 zfs_ioc_objset_zplprops); 6036 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT, 6037 zfs_ioc_dataset_list_next); 6038 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT, 6039 zfs_ioc_snapshot_list_next); 6040 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS, 6041 zfs_ioc_send_progress); 6042 6043 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF, 6044 zfs_ioc_diff, zfs_secpolicy_diff); 6045 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS, 6046 zfs_ioc_obj_to_stats, zfs_secpolicy_diff); 6047 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH, 6048 zfs_ioc_obj_to_path, zfs_secpolicy_diff); 6049 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE, 6050 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one); 6051 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY, 6052 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many); 6053 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND, 6054 zfs_ioc_send, zfs_secpolicy_send); 6055 6056 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop, 6057 zfs_secpolicy_none); 6058 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy, 6059 zfs_secpolicy_destroy); 6060 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename, 6061 zfs_secpolicy_rename); 6062 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv, 6063 zfs_secpolicy_recv); 6064 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote, 6065 zfs_secpolicy_promote); 6066 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP, 6067 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop); 6068 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl, 6069 zfs_secpolicy_set_fsacl); 6070 6071 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share, 6072 zfs_secpolicy_share, POOL_CHECK_NONE); 6073 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl, 6074 zfs_secpolicy_smb_acl, POOL_CHECK_NONE); 6075 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE, 6076 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade, 6077 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 6078 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT, 6079 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot, 6080 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 6081 6082#ifdef __FreeBSD__ 6083 zfs_ioctl_register_dataset_nolog(ZFS_IOC_JAIL, zfs_ioc_jail, 6084 zfs_secpolicy_config, POOL_CHECK_NONE); 6085 zfs_ioctl_register_dataset_nolog(ZFS_IOC_UNJAIL, zfs_ioc_unjail, 6086 zfs_secpolicy_config, POOL_CHECK_NONE); 6087 zfs_ioctl_register("fbsd_nextboot", ZFS_IOC_NEXTBOOT, 6088 zfs_ioc_nextboot, zfs_secpolicy_config, NO_NAME, 6089 POOL_CHECK_NONE, B_FALSE, B_FALSE); 6090#endif 6091} 6092 6093int 6094pool_status_check(const char *name, zfs_ioc_namecheck_t type, 6095 zfs_ioc_poolcheck_t check) 6096{ 6097 spa_t *spa; 6098 int error; 6099 6100 ASSERT(type == POOL_NAME || type == DATASET_NAME); 6101 6102 if (check & POOL_CHECK_NONE) 6103 return (0); 6104 6105 error = spa_open(name, &spa, FTAG); 6106 if (error == 0) { 6107 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa)) 6108 error = SET_ERROR(EAGAIN); 6109 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa)) 6110 error = SET_ERROR(EROFS); 6111 spa_close(spa, FTAG); 6112 } 6113 return (error); 6114} 6115 6116/* 6117 * Find a free minor number. 6118 */ 6119minor_t 6120zfsdev_minor_alloc(void) 6121{ 6122 static minor_t last_minor; 6123 minor_t m; 6124 6125 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 6126 6127 for (m = last_minor + 1; m != last_minor; m++) { 6128 if (m > ZFSDEV_MAX_MINOR) 6129 m = 1; 6130 if (ddi_get_soft_state(zfsdev_state, m) == NULL) { 6131 last_minor = m; 6132 return (m); 6133 } 6134 } 6135 6136 return (0); 6137} 6138 6139static int 6140zfs_ctldev_init(struct cdev *devp) 6141{ 6142 minor_t minor; 6143 zfs_soft_state_t *zs; 6144 6145 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 6146 6147 minor = zfsdev_minor_alloc(); 6148 if (minor == 0) 6149 return (SET_ERROR(ENXIO)); 6150 6151 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS) 6152 return (SET_ERROR(EAGAIN)); 6153 6154 devfs_set_cdevpriv((void *)(uintptr_t)minor, zfsdev_close); 6155 6156 zs = ddi_get_soft_state(zfsdev_state, minor); 6157 zs->zss_type = ZSST_CTLDEV; 6158 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data); 6159 6160 return (0); 6161} 6162 6163static void 6164zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor) 6165{ 6166 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 6167 6168 zfs_onexit_destroy(zo); 6169 ddi_soft_state_free(zfsdev_state, minor); 6170} 6171 6172void * 6173zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which) 6174{ 6175 zfs_soft_state_t *zp; 6176 6177 zp = ddi_get_soft_state(zfsdev_state, minor); 6178 if (zp == NULL || zp->zss_type != which) 6179 return (NULL); 6180 6181 return (zp->zss_data); 6182} 6183 6184static int 6185zfsdev_open(struct cdev *devp, int flag, int mode, struct thread *td) 6186{ 6187 int error = 0; 6188 6189#ifdef illumos 6190 if (getminor(*devp) != 0) 6191 return (zvol_open(devp, flag, otyp, cr)); 6192#endif 6193 6194 /* This is the control device. Allocate a new minor if requested. */ 6195 if (flag & FEXCL) { 6196 mutex_enter(&spa_namespace_lock); 6197 error = zfs_ctldev_init(devp); 6198 mutex_exit(&spa_namespace_lock); 6199 } 6200 6201 return (error); 6202} 6203 6204static void 6205zfsdev_close(void *data) 6206{ 6207 zfs_onexit_t *zo; 6208 minor_t minor = (minor_t)(uintptr_t)data; 6209 6210 if (minor == 0) 6211 return; 6212 6213 mutex_enter(&spa_namespace_lock); 6214 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV); 6215 if (zo == NULL) { 6216 mutex_exit(&spa_namespace_lock); 6217 return; 6218 } 6219 zfs_ctldev_destroy(zo, minor); 6220 mutex_exit(&spa_namespace_lock); 6221} 6222 6223static int 6224zfsdev_ioctl(struct cdev *dev, u_long zcmd, caddr_t arg, int flag, 6225 struct thread *td) 6226{ 6227 zfs_cmd_t *zc; 6228 uint_t vecnum; 6229 int error, rc, len; 6230#ifdef illumos 6231 minor_t minor = getminor(dev); 6232#else 6233 zfs_iocparm_t *zc_iocparm; 6234 int cflag, cmd, oldvecnum; 6235 boolean_t newioc, compat; 6236 void *compat_zc = NULL; 6237 cred_t *cr = td->td_ucred; 6238#endif 6239 const zfs_ioc_vec_t *vec; 6240 char *saved_poolname = NULL; 6241 nvlist_t *innvl = NULL; 6242 6243 cflag = ZFS_CMD_COMPAT_NONE; 6244 compat = B_FALSE; 6245 newioc = B_TRUE; /* "new" style (zfs_iocparm_t) ioctl */ 6246 6247 len = IOCPARM_LEN(zcmd); 6248 vecnum = cmd = zcmd & 0xff; 6249 6250 /* 6251 * Check if we are talking to supported older binaries 6252 * and translate zfs_cmd if necessary 6253 */ 6254 if (len != sizeof(zfs_iocparm_t)) { 6255 newioc = B_FALSE; 6256 compat = B_TRUE; 6257 6258 vecnum = cmd; 6259 6260 switch (len) { 6261 case sizeof(zfs_cmd_zcmd_t): 6262 cflag = ZFS_CMD_COMPAT_LZC; 6263 break; 6264 case sizeof(zfs_cmd_deadman_t): 6265 cflag = ZFS_CMD_COMPAT_DEADMAN; 6266 break; 6267 case sizeof(zfs_cmd_v28_t): 6268 cflag = ZFS_CMD_COMPAT_V28; 6269 break; 6270 case sizeof(zfs_cmd_v15_t): 6271 cflag = ZFS_CMD_COMPAT_V15; 6272 vecnum = zfs_ioctl_v15_to_v28[cmd]; 6273 6274 /* 6275 * Return without further handling 6276 * if the command is blacklisted. 6277 */ 6278 if (vecnum == ZFS_IOC_COMPAT_PASS) 6279 return (0); 6280 else if (vecnum == ZFS_IOC_COMPAT_FAIL) 6281 return (ENOTSUP); 6282 break; 6283 default: 6284 return (EINVAL); 6285 } 6286 } 6287 6288#ifdef illumos 6289 vecnum = cmd - ZFS_IOC_FIRST; 6290 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip)); 6291#endif 6292 6293 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0])) 6294 return (SET_ERROR(EINVAL)); 6295 vec = &zfs_ioc_vec[vecnum]; 6296 6297 zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP); 6298 6299#ifdef illumos 6300 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag); 6301 if (error != 0) { 6302 error = SET_ERROR(EFAULT); 6303 goto out; 6304 } 6305#else /* !illumos */ 6306 bzero(zc, sizeof(zfs_cmd_t)); 6307 6308 if (newioc) { 6309 zc_iocparm = (void *)arg; 6310 6311 switch (zc_iocparm->zfs_ioctl_version) { 6312 case ZFS_IOCVER_CURRENT: 6313 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_t)) { 6314 error = SET_ERROR(EINVAL); 6315 goto out; 6316 } 6317 break; 6318 case ZFS_IOCVER_INLANES: 6319 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_inlanes_t)) { 6320 error = SET_ERROR(EFAULT); 6321 goto out; 6322 } 6323 compat = B_TRUE; 6324 cflag = ZFS_CMD_COMPAT_INLANES; 6325 break; 6326 case ZFS_IOCVER_RESUME: 6327 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_resume_t)) { 6328 error = SET_ERROR(EFAULT); 6329 goto out; 6330 } 6331 compat = B_TRUE; 6332 cflag = ZFS_CMD_COMPAT_RESUME; 6333 break; 6334 case ZFS_IOCVER_EDBP: 6335 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_edbp_t)) { 6336 error = SET_ERROR(EFAULT); 6337 goto out; 6338 } 6339 compat = B_TRUE; 6340 cflag = ZFS_CMD_COMPAT_EDBP; 6341 break; 6342 case ZFS_IOCVER_ZCMD: 6343 if (zc_iocparm->zfs_cmd_size > sizeof(zfs_cmd_t) || 6344 zc_iocparm->zfs_cmd_size < sizeof(zfs_cmd_zcmd_t)) { 6345 error = SET_ERROR(EFAULT); 6346 goto out; 6347 } 6348 compat = B_TRUE; 6349 cflag = ZFS_CMD_COMPAT_ZCMD; 6350 break; 6351 default: 6352 error = SET_ERROR(EINVAL); 6353 goto out; 6354 /* NOTREACHED */ 6355 } 6356 6357 if (compat) { 6358 ASSERT(sizeof(zfs_cmd_t) >= zc_iocparm->zfs_cmd_size); 6359 compat_zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP); 6360 bzero(compat_zc, sizeof(zfs_cmd_t)); 6361 6362 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd, 6363 compat_zc, zc_iocparm->zfs_cmd_size, flag); 6364 if (error != 0) { 6365 error = SET_ERROR(EFAULT); 6366 goto out; 6367 } 6368 } else { 6369 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd, 6370 zc, zc_iocparm->zfs_cmd_size, flag); 6371 if (error != 0) { 6372 error = SET_ERROR(EFAULT); 6373 goto out; 6374 } 6375 } 6376 } 6377 6378 if (compat) { 6379 if (newioc) { 6380 ASSERT(compat_zc != NULL); 6381 zfs_cmd_compat_get(zc, compat_zc, cflag); 6382 } else { 6383 ASSERT(compat_zc == NULL); 6384 zfs_cmd_compat_get(zc, arg, cflag); 6385 } 6386 oldvecnum = vecnum; 6387 error = zfs_ioctl_compat_pre(zc, &vecnum, cflag); 6388 if (error != 0) 6389 goto out; 6390 if (oldvecnum != vecnum) 6391 vec = &zfs_ioc_vec[vecnum]; 6392 } 6393#endif /* !illumos */ 6394 6395 zc->zc_iflags = flag & FKIOCTL; 6396 if (zc->zc_nvlist_src_size != 0) { 6397 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 6398 zc->zc_iflags, &innvl); 6399 if (error != 0) 6400 goto out; 6401 } 6402 6403 /* rewrite innvl for backwards compatibility */ 6404 if (compat) 6405 innvl = zfs_ioctl_compat_innvl(zc, innvl, vecnum, cflag); 6406 6407 /* 6408 * Ensure that all pool/dataset names are valid before we pass down to 6409 * the lower layers. 6410 */ 6411 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0'; 6412 switch (vec->zvec_namecheck) { 6413 case POOL_NAME: 6414 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0) 6415 error = SET_ERROR(EINVAL); 6416 else 6417 error = pool_status_check(zc->zc_name, 6418 vec->zvec_namecheck, vec->zvec_pool_check); 6419 break; 6420 6421 case DATASET_NAME: 6422 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0) 6423 error = SET_ERROR(EINVAL); 6424 else 6425 error = pool_status_check(zc->zc_name, 6426 vec->zvec_namecheck, vec->zvec_pool_check); 6427 break; 6428 6429 case NO_NAME: 6430 break; 6431 } 6432 6433 if (error == 0) 6434 error = vec->zvec_secpolicy(zc, innvl, cr); 6435 6436 if (error != 0) 6437 goto out; 6438 6439 /* legacy ioctls can modify zc_name */ 6440 len = strcspn(zc->zc_name, "/@#") + 1; 6441 saved_poolname = kmem_alloc(len, KM_SLEEP); 6442 (void) strlcpy(saved_poolname, zc->zc_name, len); 6443 6444 if (vec->zvec_func != NULL) { 6445 nvlist_t *outnvl; 6446 int puterror = 0; 6447 spa_t *spa; 6448 nvlist_t *lognv = NULL; 6449 6450 ASSERT(vec->zvec_legacy_func == NULL); 6451 6452 /* 6453 * Add the innvl to the lognv before calling the func, 6454 * in case the func changes the innvl. 6455 */ 6456 if (vec->zvec_allow_log) { 6457 lognv = fnvlist_alloc(); 6458 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL, 6459 vec->zvec_name); 6460 if (!nvlist_empty(innvl)) { 6461 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL, 6462 innvl); 6463 } 6464 } 6465 6466 outnvl = fnvlist_alloc(); 6467 error = vec->zvec_func(zc->zc_name, innvl, outnvl); 6468 6469 if (error == 0 && vec->zvec_allow_log && 6470 spa_open(zc->zc_name, &spa, FTAG) == 0) { 6471 if (!nvlist_empty(outnvl)) { 6472 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL, 6473 outnvl); 6474 } 6475 (void) spa_history_log_nvl(spa, lognv); 6476 spa_close(spa, FTAG); 6477 } 6478 fnvlist_free(lognv); 6479 6480 /* rewrite outnvl for backwards compatibility */ 6481 if (compat) 6482 outnvl = zfs_ioctl_compat_outnvl(zc, outnvl, vecnum, 6483 cflag); 6484 6485 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) { 6486 int smusherror = 0; 6487 if (vec->zvec_smush_outnvlist) { 6488 smusherror = nvlist_smush(outnvl, 6489 zc->zc_nvlist_dst_size); 6490 } 6491 if (smusherror == 0) 6492 puterror = put_nvlist(zc, outnvl); 6493 } 6494 6495 if (puterror != 0) 6496 error = puterror; 6497 6498 nvlist_free(outnvl); 6499 } else { 6500 error = vec->zvec_legacy_func(zc); 6501 } 6502 6503out: 6504 nvlist_free(innvl); 6505 6506#ifdef illumos 6507 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag); 6508 if (error == 0 && rc != 0) 6509 error = SET_ERROR(EFAULT); 6510#else 6511 if (compat) { 6512 zfs_ioctl_compat_post(zc, cmd, cflag); 6513 if (newioc) { 6514 ASSERT(compat_zc != NULL); 6515 ASSERT(sizeof(zfs_cmd_t) >= zc_iocparm->zfs_cmd_size); 6516 6517 zfs_cmd_compat_put(zc, compat_zc, vecnum, cflag); 6518 rc = ddi_copyout(compat_zc, 6519 (void *)(uintptr_t)zc_iocparm->zfs_cmd, 6520 zc_iocparm->zfs_cmd_size, flag); 6521 if (error == 0 && rc != 0) 6522 error = SET_ERROR(EFAULT); 6523 kmem_free(compat_zc, sizeof (zfs_cmd_t)); 6524 } else { 6525 zfs_cmd_compat_put(zc, arg, vecnum, cflag); 6526 } 6527 } else { 6528 ASSERT(newioc); 6529 6530 rc = ddi_copyout(zc, (void *)(uintptr_t)zc_iocparm->zfs_cmd, 6531 sizeof (zfs_cmd_t), flag); 6532 if (error == 0 && rc != 0) 6533 error = SET_ERROR(EFAULT); 6534 } 6535#endif 6536 if (error == 0 && vec->zvec_allow_log) { 6537 char *s = tsd_get(zfs_allow_log_key); 6538 if (s != NULL) 6539 strfree(s); 6540 (void) tsd_set(zfs_allow_log_key, saved_poolname); 6541 } else { 6542 if (saved_poolname != NULL) 6543 strfree(saved_poolname); 6544 } 6545 6546 kmem_free(zc, sizeof (zfs_cmd_t)); 6547 return (error); 6548} 6549 6550#ifdef illumos 6551static int 6552zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd) 6553{ 6554 if (cmd != DDI_ATTACH) 6555 return (DDI_FAILURE); 6556 6557 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0, 6558 DDI_PSEUDO, 0) == DDI_FAILURE) 6559 return (DDI_FAILURE); 6560 6561 zfs_dip = dip; 6562 6563 ddi_report_dev(dip); 6564 6565 return (DDI_SUCCESS); 6566} 6567 6568static int 6569zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd) 6570{ 6571 if (spa_busy() || zfs_busy() || zvol_busy()) 6572 return (DDI_FAILURE); 6573 6574 if (cmd != DDI_DETACH) 6575 return (DDI_FAILURE); 6576 6577 zfs_dip = NULL; 6578 6579 ddi_prop_remove_all(dip); 6580 ddi_remove_minor_node(dip, NULL); 6581 6582 return (DDI_SUCCESS); 6583} 6584 6585/*ARGSUSED*/ 6586static int 6587zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result) 6588{ 6589 switch (infocmd) { 6590 case DDI_INFO_DEVT2DEVINFO: 6591 *result = zfs_dip; 6592 return (DDI_SUCCESS); 6593 6594 case DDI_INFO_DEVT2INSTANCE: 6595 *result = (void *)0; 6596 return (DDI_SUCCESS); 6597 } 6598 6599 return (DDI_FAILURE); 6600} 6601#endif /* illumos */ 6602 6603/* 6604 * OK, so this is a little weird. 6605 * 6606 * /dev/zfs is the control node, i.e. minor 0. 6607 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0. 6608 * 6609 * /dev/zfs has basically nothing to do except serve up ioctls, 6610 * so most of the standard driver entry points are in zvol.c. 6611 */ 6612#ifdef illumos 6613static struct cb_ops zfs_cb_ops = { 6614 zfsdev_open, /* open */ 6615 zfsdev_close, /* close */ 6616 zvol_strategy, /* strategy */ 6617 nodev, /* print */ 6618 zvol_dump, /* dump */ 6619 zvol_read, /* read */ 6620 zvol_write, /* write */ 6621 zfsdev_ioctl, /* ioctl */ 6622 nodev, /* devmap */ 6623 nodev, /* mmap */ 6624 nodev, /* segmap */ 6625 nochpoll, /* poll */ 6626 ddi_prop_op, /* prop_op */ 6627 NULL, /* streamtab */ 6628 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */ 6629 CB_REV, /* version */ 6630 nodev, /* async read */ 6631 nodev, /* async write */ 6632}; 6633 6634static struct dev_ops zfs_dev_ops = { 6635 DEVO_REV, /* version */ 6636 0, /* refcnt */ 6637 zfs_info, /* info */ 6638 nulldev, /* identify */ 6639 nulldev, /* probe */ 6640 zfs_attach, /* attach */ 6641 zfs_detach, /* detach */ 6642 nodev, /* reset */ 6643 &zfs_cb_ops, /* driver operations */ 6644 NULL, /* no bus operations */ 6645 NULL, /* power */ 6646 ddi_quiesce_not_needed, /* quiesce */ 6647}; 6648 6649static struct modldrv zfs_modldrv = { 6650 &mod_driverops, 6651 "ZFS storage pool", 6652 &zfs_dev_ops 6653}; 6654 6655static struct modlinkage modlinkage = { 6656 MODREV_1, 6657 (void *)&zfs_modlfs, 6658 (void *)&zfs_modldrv, 6659 NULL 6660}; 6661#endif /* illumos */ 6662 6663static struct cdevsw zfs_cdevsw = { 6664 .d_version = D_VERSION, 6665 .d_open = zfsdev_open, 6666 .d_ioctl = zfsdev_ioctl, 6667 .d_name = ZFS_DEV_NAME 6668}; 6669 6670static void 6671zfs_allow_log_destroy(void *arg) 6672{ 6673 char *poolname = arg; 6674 strfree(poolname); 6675} 6676 6677static void 6678zfsdev_init(void) 6679{ 6680 zfsdev = make_dev(&zfs_cdevsw, 0x0, UID_ROOT, GID_OPERATOR, 0666, 6681 ZFS_DEV_NAME); 6682} 6683 6684static void 6685zfsdev_fini(void) 6686{ 6687 if (zfsdev != NULL) 6688 destroy_dev(zfsdev); 6689} 6690 6691static struct root_hold_token *zfs_root_token; 6692struct proc *zfsproc; 6693 6694#ifdef illumos 6695int 6696_init(void) 6697{ 6698 int error; 6699 6700 spa_init(FREAD | FWRITE); 6701 zfs_init(); 6702 zvol_init(); 6703 zfs_ioctl_init(); 6704 6705 if ((error = mod_install(&modlinkage)) != 0) { 6706 zvol_fini(); 6707 zfs_fini(); 6708 spa_fini(); 6709 return (error); 6710 } 6711 6712 tsd_create(&zfs_fsyncer_key, NULL); 6713 tsd_create(&rrw_tsd_key, rrw_tsd_destroy); 6714 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy); 6715 6716 error = ldi_ident_from_mod(&modlinkage, &zfs_li); 6717 ASSERT(error == 0); 6718 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL); 6719 6720 return (0); 6721} 6722 6723int 6724_fini(void) 6725{ 6726 int error; 6727 6728 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled) 6729 return (SET_ERROR(EBUSY)); 6730 6731 if ((error = mod_remove(&modlinkage)) != 0) 6732 return (error); 6733 6734 zvol_fini(); 6735 zfs_fini(); 6736 spa_fini(); 6737 if (zfs_nfsshare_inited) 6738 (void) ddi_modclose(nfs_mod); 6739 if (zfs_smbshare_inited) 6740 (void) ddi_modclose(smbsrv_mod); 6741 if (zfs_nfsshare_inited || zfs_smbshare_inited) 6742 (void) ddi_modclose(sharefs_mod); 6743 6744 tsd_destroy(&zfs_fsyncer_key); 6745 ldi_ident_release(zfs_li); 6746 zfs_li = NULL; 6747 mutex_destroy(&zfs_share_lock); 6748 6749 return (error); 6750} 6751 6752int 6753_info(struct modinfo *modinfop) 6754{ 6755 return (mod_info(&modlinkage, modinfop)); 6756} 6757#endif /* illumos */ 6758 6759static int zfs__init(void); 6760static int zfs__fini(void); 6761static void zfs_shutdown(void *, int); 6762 6763static eventhandler_tag zfs_shutdown_event_tag; 6764 6765#ifdef __FreeBSD__ 6766#define ZFS_MIN_KSTACK_PAGES 4 6767#endif 6768 6769int 6770zfs__init(void) 6771{ 6772 6773#ifdef __FreeBSD__ 6774#if KSTACK_PAGES < ZFS_MIN_KSTACK_PAGES 6775 printf("ZFS NOTICE: KSTACK_PAGES is %d which could result in stack " 6776 "overflow panic!\nPlease consider adding " 6777 "'options KSTACK_PAGES=%d' to your kernel config\n", KSTACK_PAGES, 6778 ZFS_MIN_KSTACK_PAGES); 6779#endif 6780#endif 6781 zfs_root_token = root_mount_hold("ZFS"); 6782 6783 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL); 6784 6785 spa_init(FREAD | FWRITE); 6786 zfs_init(); 6787 zvol_init(); 6788 zfs_ioctl_init(); 6789 6790 tsd_create(&zfs_fsyncer_key, NULL); 6791 tsd_create(&rrw_tsd_key, rrw_tsd_destroy); 6792 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy); 6793 tsd_create(&zfs_geom_probe_vdev_key, NULL); 6794 6795 printf("ZFS storage pool version: features support (" SPA_VERSION_STRING ")\n"); 6796 root_mount_rel(zfs_root_token); 6797 6798 zfsdev_init(); 6799 6800 return (0); 6801} 6802 6803int 6804zfs__fini(void) 6805{ 6806 if (spa_busy() || zfs_busy() || zvol_busy() || 6807 zio_injection_enabled) { 6808 return (EBUSY); 6809 } 6810 6811 zfsdev_fini(); 6812 zvol_fini(); 6813 zfs_fini(); 6814 spa_fini(); 6815 6816 tsd_destroy(&zfs_fsyncer_key); 6817 tsd_destroy(&rrw_tsd_key); 6818 tsd_destroy(&zfs_allow_log_key); 6819 6820 mutex_destroy(&zfs_share_lock); 6821 6822 return (0); 6823} 6824 6825static void 6826zfs_shutdown(void *arg __unused, int howto __unused) 6827{ 6828 6829 /* 6830 * ZFS fini routines can not properly work in a panic-ed system. 6831 */ 6832 if (panicstr == NULL) 6833 (void)zfs__fini(); 6834} 6835 6836 6837static int 6838zfs_modevent(module_t mod, int type, void *unused __unused) 6839{ 6840 int err; 6841 6842 switch (type) { 6843 case MOD_LOAD: 6844 err = zfs__init(); 6845 if (err == 0) 6846 zfs_shutdown_event_tag = EVENTHANDLER_REGISTER( 6847 shutdown_post_sync, zfs_shutdown, NULL, 6848 SHUTDOWN_PRI_FIRST); 6849 return (err); 6850 case MOD_UNLOAD: 6851 err = zfs__fini(); 6852 if (err == 0 && zfs_shutdown_event_tag != NULL) 6853 EVENTHANDLER_DEREGISTER(shutdown_post_sync, 6854 zfs_shutdown_event_tag); 6855 return (err); 6856 case MOD_SHUTDOWN: 6857 return (0); 6858 default: 6859 break; 6860 } 6861 return (EOPNOTSUPP); 6862} 6863 6864static moduledata_t zfs_mod = { 6865 "zfsctrl", 6866 zfs_modevent, 6867 0 6868}; 6869DECLARE_MODULE(zfsctrl, zfs_mod, SI_SUB_VFS, SI_ORDER_ANY); 6870MODULE_VERSION(zfsctrl, 1); 6871MODULE_DEPEND(zfsctrl, opensolaris, 1, 1, 1); 6872MODULE_DEPEND(zfsctrl, krpc, 1, 1, 1); 6873MODULE_DEPEND(zfsctrl, acl_nfs4, 1, 1, 1); 6874