zfs_ioctl.c revision 307125
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; 210 211typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *); 212typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *); 213typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *); 214 215typedef enum { 216 NO_NAME, 217 POOL_NAME, 218 DATASET_NAME 219} zfs_ioc_namecheck_t; 220 221typedef enum { 222 POOL_CHECK_NONE = 1 << 0, 223 POOL_CHECK_SUSPENDED = 1 << 1, 224 POOL_CHECK_READONLY = 1 << 2, 225} zfs_ioc_poolcheck_t; 226 227typedef struct zfs_ioc_vec { 228 zfs_ioc_legacy_func_t *zvec_legacy_func; 229 zfs_ioc_func_t *zvec_func; 230 zfs_secpolicy_func_t *zvec_secpolicy; 231 zfs_ioc_namecheck_t zvec_namecheck; 232 boolean_t zvec_allow_log; 233 zfs_ioc_poolcheck_t zvec_pool_check; 234 boolean_t zvec_smush_outnvlist; 235 const char *zvec_name; 236} zfs_ioc_vec_t; 237 238/* This array is indexed by zfs_userquota_prop_t */ 239static const char *userquota_perms[] = { 240 ZFS_DELEG_PERM_USERUSED, 241 ZFS_DELEG_PERM_USERQUOTA, 242 ZFS_DELEG_PERM_GROUPUSED, 243 ZFS_DELEG_PERM_GROUPQUOTA, 244}; 245 246static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc); 247static int zfs_check_settable(const char *name, nvpair_t *property, 248 cred_t *cr); 249static int zfs_check_clearable(char *dataset, nvlist_t *props, 250 nvlist_t **errors); 251static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *, 252 boolean_t *); 253int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *); 254static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp); 255 256static void zfsdev_close(void *data); 257 258static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature); 259 260/* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */ 261void 262__dprintf(const char *file, const char *func, int line, const char *fmt, ...) 263{ 264 const char *newfile; 265 char buf[512]; 266 va_list adx; 267 268 /* 269 * Get rid of annoying "../common/" prefix to filename. 270 */ 271 newfile = strrchr(file, '/'); 272 if (newfile != NULL) { 273 newfile = newfile + 1; /* Get rid of leading / */ 274 } else { 275 newfile = file; 276 } 277 278 va_start(adx, fmt); 279 (void) vsnprintf(buf, sizeof (buf), fmt, adx); 280 va_end(adx); 281 282 /* 283 * To get this data, use the zfs-dprintf probe as so: 284 * dtrace -q -n 'zfs-dprintf \ 285 * /stringof(arg0) == "dbuf.c"/ \ 286 * {printf("%s: %s", stringof(arg1), stringof(arg3))}' 287 * arg0 = file name 288 * arg1 = function name 289 * arg2 = line number 290 * arg3 = message 291 */ 292 DTRACE_PROBE4(zfs__dprintf, 293 char *, newfile, char *, func, int, line, char *, buf); 294} 295 296static void 297history_str_free(char *buf) 298{ 299 kmem_free(buf, HIS_MAX_RECORD_LEN); 300} 301 302static char * 303history_str_get(zfs_cmd_t *zc) 304{ 305 char *buf; 306 307 if (zc->zc_history == 0) 308 return (NULL); 309 310 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP); 311 if (copyinstr((void *)(uintptr_t)zc->zc_history, 312 buf, HIS_MAX_RECORD_LEN, NULL) != 0) { 313 history_str_free(buf); 314 return (NULL); 315 } 316 317 buf[HIS_MAX_RECORD_LEN -1] = '\0'; 318 319 return (buf); 320} 321 322/* 323 * Check to see if the named dataset is currently defined as bootable 324 */ 325static boolean_t 326zfs_is_bootfs(const char *name) 327{ 328 objset_t *os; 329 330 if (dmu_objset_hold(name, FTAG, &os) == 0) { 331 boolean_t ret; 332 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os))); 333 dmu_objset_rele(os, FTAG); 334 return (ret); 335 } 336 return (B_FALSE); 337} 338 339/* 340 * Return non-zero if the spa version is less than requested version. 341 */ 342static int 343zfs_earlier_version(const char *name, int version) 344{ 345 spa_t *spa; 346 347 if (spa_open(name, &spa, FTAG) == 0) { 348 if (spa_version(spa) < version) { 349 spa_close(spa, FTAG); 350 return (1); 351 } 352 spa_close(spa, FTAG); 353 } 354 return (0); 355} 356 357/* 358 * Return TRUE if the ZPL version is less than requested version. 359 */ 360static boolean_t 361zpl_earlier_version(const char *name, int version) 362{ 363 objset_t *os; 364 boolean_t rc = B_TRUE; 365 366 if (dmu_objset_hold(name, FTAG, &os) == 0) { 367 uint64_t zplversion; 368 369 if (dmu_objset_type(os) != DMU_OST_ZFS) { 370 dmu_objset_rele(os, FTAG); 371 return (B_TRUE); 372 } 373 /* XXX reading from non-owned objset */ 374 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0) 375 rc = zplversion < version; 376 dmu_objset_rele(os, FTAG); 377 } 378 return (rc); 379} 380 381static void 382zfs_log_history(zfs_cmd_t *zc) 383{ 384 spa_t *spa; 385 char *buf; 386 387 if ((buf = history_str_get(zc)) == NULL) 388 return; 389 390 if (spa_open(zc->zc_name, &spa, FTAG) == 0) { 391 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY) 392 (void) spa_history_log(spa, buf); 393 spa_close(spa, FTAG); 394 } 395 history_str_free(buf); 396} 397 398/* 399 * Policy for top-level read operations (list pools). Requires no privileges, 400 * and can be used in the local zone, as there is no associated dataset. 401 */ 402/* ARGSUSED */ 403static int 404zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 405{ 406 return (0); 407} 408 409/* 410 * Policy for dataset read operations (list children, get statistics). Requires 411 * no privileges, but must be visible in the local zone. 412 */ 413/* ARGSUSED */ 414static int 415zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 416{ 417 if (INGLOBALZONE(curthread) || 418 zone_dataset_visible(zc->zc_name, NULL)) 419 return (0); 420 421 return (SET_ERROR(ENOENT)); 422} 423 424static int 425zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr) 426{ 427 int writable = 1; 428 429 /* 430 * The dataset must be visible by this zone -- check this first 431 * so they don't see EPERM on something they shouldn't know about. 432 */ 433 if (!INGLOBALZONE(curthread) && 434 !zone_dataset_visible(dataset, &writable)) 435 return (SET_ERROR(ENOENT)); 436 437 if (INGLOBALZONE(curthread)) { 438 /* 439 * If the fs is zoned, only root can access it from the 440 * global zone. 441 */ 442 if (secpolicy_zfs(cr) && zoned) 443 return (SET_ERROR(EPERM)); 444 } else { 445 /* 446 * If we are in a local zone, the 'zoned' property must be set. 447 */ 448 if (!zoned) 449 return (SET_ERROR(EPERM)); 450 451 /* must be writable by this zone */ 452 if (!writable) 453 return (SET_ERROR(EPERM)); 454 } 455 return (0); 456} 457 458static int 459zfs_dozonecheck(const char *dataset, cred_t *cr) 460{ 461 uint64_t zoned; 462 463 if (dsl_prop_get_integer(dataset, "jailed", &zoned, NULL)) 464 return (SET_ERROR(ENOENT)); 465 466 return (zfs_dozonecheck_impl(dataset, zoned, cr)); 467} 468 469static int 470zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr) 471{ 472 uint64_t zoned; 473 474 if (dsl_prop_get_int_ds(ds, "jailed", &zoned)) 475 return (SET_ERROR(ENOENT)); 476 477 return (zfs_dozonecheck_impl(dataset, zoned, cr)); 478} 479 480static int 481zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds, 482 const char *perm, cred_t *cr) 483{ 484 int error; 485 486 error = zfs_dozonecheck_ds(name, ds, cr); 487 if (error == 0) { 488 error = secpolicy_zfs(cr); 489 if (error != 0) 490 error = dsl_deleg_access_impl(ds, perm, cr); 491 } 492 return (error); 493} 494 495static int 496zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr) 497{ 498 int error; 499 dsl_dataset_t *ds; 500 dsl_pool_t *dp; 501 502 /* 503 * First do a quick check for root in the global zone, which 504 * is allowed to do all write_perms. This ensures that zfs_ioc_* 505 * will get to handle nonexistent datasets. 506 */ 507 if (INGLOBALZONE(curthread) && secpolicy_zfs(cr) == 0) 508 return (0); 509 510 error = dsl_pool_hold(name, FTAG, &dp); 511 if (error != 0) 512 return (error); 513 514 error = dsl_dataset_hold(dp, name, FTAG, &ds); 515 if (error != 0) { 516 dsl_pool_rele(dp, FTAG); 517 return (error); 518 } 519 520 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr); 521 522 dsl_dataset_rele(ds, FTAG); 523 dsl_pool_rele(dp, FTAG); 524 return (error); 525} 526 527#ifdef SECLABEL 528/* 529 * Policy for setting the security label property. 530 * 531 * Returns 0 for success, non-zero for access and other errors. 532 */ 533static int 534zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr) 535{ 536 char ds_hexsl[MAXNAMELEN]; 537 bslabel_t ds_sl, new_sl; 538 boolean_t new_default = FALSE; 539 uint64_t zoned; 540 int needed_priv = -1; 541 int error; 542 543 /* First get the existing dataset label. */ 544 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL), 545 1, sizeof (ds_hexsl), &ds_hexsl, NULL); 546 if (error != 0) 547 return (SET_ERROR(EPERM)); 548 549 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0) 550 new_default = TRUE; 551 552 /* The label must be translatable */ 553 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0)) 554 return (SET_ERROR(EINVAL)); 555 556 /* 557 * In a non-global zone, disallow attempts to set a label that 558 * doesn't match that of the zone; otherwise no other checks 559 * are needed. 560 */ 561 if (!INGLOBALZONE(curproc)) { 562 if (new_default || !blequal(&new_sl, CR_SL(CRED()))) 563 return (SET_ERROR(EPERM)); 564 return (0); 565 } 566 567 /* 568 * For global-zone datasets (i.e., those whose zoned property is 569 * "off", verify that the specified new label is valid for the 570 * global zone. 571 */ 572 if (dsl_prop_get_integer(name, 573 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL)) 574 return (SET_ERROR(EPERM)); 575 if (!zoned) { 576 if (zfs_check_global_label(name, strval) != 0) 577 return (SET_ERROR(EPERM)); 578 } 579 580 /* 581 * If the existing dataset label is nondefault, check if the 582 * dataset is mounted (label cannot be changed while mounted). 583 * Get the zfsvfs; if there isn't one, then the dataset isn't 584 * mounted (or isn't a dataset, doesn't exist, ...). 585 */ 586 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) { 587 objset_t *os; 588 static char *setsl_tag = "setsl_tag"; 589 590 /* 591 * Try to own the dataset; abort if there is any error, 592 * (e.g., already mounted, in use, or other error). 593 */ 594 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE, 595 setsl_tag, &os); 596 if (error != 0) 597 return (SET_ERROR(EPERM)); 598 599 dmu_objset_disown(os, setsl_tag); 600 601 if (new_default) { 602 needed_priv = PRIV_FILE_DOWNGRADE_SL; 603 goto out_check; 604 } 605 606 if (hexstr_to_label(strval, &new_sl) != 0) 607 return (SET_ERROR(EPERM)); 608 609 if (blstrictdom(&ds_sl, &new_sl)) 610 needed_priv = PRIV_FILE_DOWNGRADE_SL; 611 else if (blstrictdom(&new_sl, &ds_sl)) 612 needed_priv = PRIV_FILE_UPGRADE_SL; 613 } else { 614 /* dataset currently has a default label */ 615 if (!new_default) 616 needed_priv = PRIV_FILE_UPGRADE_SL; 617 } 618 619out_check: 620 if (needed_priv != -1) 621 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL)); 622 return (0); 623} 624#endif /* SECLABEL */ 625 626static int 627zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval, 628 cred_t *cr) 629{ 630 char *strval; 631 632 /* 633 * Check permissions for special properties. 634 */ 635 switch (prop) { 636 case ZFS_PROP_ZONED: 637 /* 638 * Disallow setting of 'zoned' from within a local zone. 639 */ 640 if (!INGLOBALZONE(curthread)) 641 return (SET_ERROR(EPERM)); 642 break; 643 644 case ZFS_PROP_QUOTA: 645 case ZFS_PROP_FILESYSTEM_LIMIT: 646 case ZFS_PROP_SNAPSHOT_LIMIT: 647 if (!INGLOBALZONE(curthread)) { 648 uint64_t zoned; 649 char setpoint[ZFS_MAX_DATASET_NAME_LEN]; 650 /* 651 * Unprivileged users are allowed to modify the 652 * limit on things *under* (ie. contained by) 653 * the thing they own. 654 */ 655 if (dsl_prop_get_integer(dsname, "jailed", &zoned, 656 setpoint)) 657 return (SET_ERROR(EPERM)); 658 if (!zoned || strlen(dsname) <= strlen(setpoint)) 659 return (SET_ERROR(EPERM)); 660 } 661 break; 662 663 case ZFS_PROP_MLSLABEL: 664#ifdef SECLABEL 665 if (!is_system_labeled()) 666 return (SET_ERROR(EPERM)); 667 668 if (nvpair_value_string(propval, &strval) == 0) { 669 int err; 670 671 err = zfs_set_slabel_policy(dsname, strval, CRED()); 672 if (err != 0) 673 return (err); 674 } 675#else 676 return (EOPNOTSUPP); 677#endif 678 break; 679 } 680 681 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr)); 682} 683 684/* ARGSUSED */ 685static int 686zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 687{ 688 int error; 689 690 error = zfs_dozonecheck(zc->zc_name, cr); 691 if (error != 0) 692 return (error); 693 694 /* 695 * permission to set permissions will be evaluated later in 696 * dsl_deleg_can_allow() 697 */ 698 return (0); 699} 700 701/* ARGSUSED */ 702static int 703zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 704{ 705 return (zfs_secpolicy_write_perms(zc->zc_name, 706 ZFS_DELEG_PERM_ROLLBACK, cr)); 707} 708 709/* ARGSUSED */ 710static int 711zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 712{ 713 dsl_pool_t *dp; 714 dsl_dataset_t *ds; 715 char *cp; 716 int error; 717 718 /* 719 * Generate the current snapshot name from the given objsetid, then 720 * use that name for the secpolicy/zone checks. 721 */ 722 cp = strchr(zc->zc_name, '@'); 723 if (cp == NULL) 724 return (SET_ERROR(EINVAL)); 725 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 726 if (error != 0) 727 return (error); 728 729 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds); 730 if (error != 0) { 731 dsl_pool_rele(dp, FTAG); 732 return (error); 733 } 734 735 dsl_dataset_name(ds, zc->zc_name); 736 737 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds, 738 ZFS_DELEG_PERM_SEND, cr); 739 dsl_dataset_rele(ds, FTAG); 740 dsl_pool_rele(dp, FTAG); 741 742 return (error); 743} 744 745/* ARGSUSED */ 746static int 747zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 748{ 749 return (zfs_secpolicy_write_perms(zc->zc_name, 750 ZFS_DELEG_PERM_SEND, cr)); 751} 752 753/* ARGSUSED */ 754static int 755zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 756{ 757 vnode_t *vp; 758 int error; 759 760 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, 761 NO_FOLLOW, NULL, &vp)) != 0) 762 return (error); 763 764 /* Now make sure mntpnt and dataset are ZFS */ 765 766 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 || 767 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), 768 zc->zc_name) != 0)) { 769 VN_RELE(vp); 770 return (SET_ERROR(EPERM)); 771 } 772 773 VN_RELE(vp); 774 return (dsl_deleg_access(zc->zc_name, 775 ZFS_DELEG_PERM_SHARE, cr)); 776} 777 778int 779zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 780{ 781 if (!INGLOBALZONE(curthread)) 782 return (SET_ERROR(EPERM)); 783 784 if (secpolicy_nfs(cr) == 0) { 785 return (0); 786 } else { 787 return (zfs_secpolicy_deleg_share(zc, innvl, cr)); 788 } 789} 790 791int 792zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 793{ 794 if (!INGLOBALZONE(curthread)) 795 return (SET_ERROR(EPERM)); 796 797 if (secpolicy_smb(cr) == 0) { 798 return (0); 799 } else { 800 return (zfs_secpolicy_deleg_share(zc, innvl, cr)); 801 } 802} 803 804static int 805zfs_get_parent(const char *datasetname, char *parent, int parentsize) 806{ 807 char *cp; 808 809 /* 810 * Remove the @bla or /bla from the end of the name to get the parent. 811 */ 812 (void) strncpy(parent, datasetname, parentsize); 813 cp = strrchr(parent, '@'); 814 if (cp != NULL) { 815 cp[0] = '\0'; 816 } else { 817 cp = strrchr(parent, '/'); 818 if (cp == NULL) 819 return (SET_ERROR(ENOENT)); 820 cp[0] = '\0'; 821 } 822 823 return (0); 824} 825 826int 827zfs_secpolicy_destroy_perms(const char *name, cred_t *cr) 828{ 829 int error; 830 831 if ((error = zfs_secpolicy_write_perms(name, 832 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 833 return (error); 834 835 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr)); 836} 837 838/* ARGSUSED */ 839static int 840zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 841{ 842 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr)); 843} 844 845/* 846 * Destroying snapshots with delegated permissions requires 847 * descendant mount and destroy permissions. 848 */ 849/* ARGSUSED */ 850static int 851zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 852{ 853 nvlist_t *snaps; 854 nvpair_t *pair, *nextpair; 855 int error = 0; 856 857 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 858 return (SET_ERROR(EINVAL)); 859 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 860 pair = nextpair) { 861 nextpair = nvlist_next_nvpair(snaps, pair); 862 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr); 863 if (error == ENOENT) { 864 /* 865 * Ignore any snapshots that don't exist (we consider 866 * them "already destroyed"). Remove the name from the 867 * nvl here in case the snapshot is created between 868 * now and when we try to destroy it (in which case 869 * we don't want to destroy it since we haven't 870 * checked for permission). 871 */ 872 fnvlist_remove_nvpair(snaps, pair); 873 error = 0; 874 } 875 if (error != 0) 876 break; 877 } 878 879 return (error); 880} 881 882int 883zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr) 884{ 885 char parentname[ZFS_MAX_DATASET_NAME_LEN]; 886 int error; 887 888 if ((error = zfs_secpolicy_write_perms(from, 889 ZFS_DELEG_PERM_RENAME, cr)) != 0) 890 return (error); 891 892 if ((error = zfs_secpolicy_write_perms(from, 893 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 894 return (error); 895 896 if ((error = zfs_get_parent(to, parentname, 897 sizeof (parentname))) != 0) 898 return (error); 899 900 if ((error = zfs_secpolicy_write_perms(parentname, 901 ZFS_DELEG_PERM_CREATE, cr)) != 0) 902 return (error); 903 904 if ((error = zfs_secpolicy_write_perms(parentname, 905 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 906 return (error); 907 908 return (error); 909} 910 911/* ARGSUSED */ 912static int 913zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 914{ 915 char *at = NULL; 916 int error; 917 918 if ((zc->zc_cookie & 1) != 0) { 919 /* 920 * This is recursive rename, so the starting snapshot might 921 * not exist. Check file system or volume permission instead. 922 */ 923 at = strchr(zc->zc_name, '@'); 924 if (at == NULL) 925 return (EINVAL); 926 *at = '\0'; 927 } 928 929 error = zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr); 930 931 if (at != NULL) 932 *at = '@'; 933 934 return (error); 935} 936 937/* ARGSUSED */ 938static int 939zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 940{ 941 dsl_pool_t *dp; 942 dsl_dataset_t *clone; 943 int error; 944 945 error = zfs_secpolicy_write_perms(zc->zc_name, 946 ZFS_DELEG_PERM_PROMOTE, cr); 947 if (error != 0) 948 return (error); 949 950 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 951 if (error != 0) 952 return (error); 953 954 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone); 955 956 if (error == 0) { 957 char parentname[ZFS_MAX_DATASET_NAME_LEN]; 958 dsl_dataset_t *origin = NULL; 959 dsl_dir_t *dd; 960 dd = clone->ds_dir; 961 962 error = dsl_dataset_hold_obj(dd->dd_pool, 963 dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin); 964 if (error != 0) { 965 dsl_dataset_rele(clone, FTAG); 966 dsl_pool_rele(dp, FTAG); 967 return (error); 968 } 969 970 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone, 971 ZFS_DELEG_PERM_MOUNT, cr); 972 973 dsl_dataset_name(origin, parentname); 974 if (error == 0) { 975 error = zfs_secpolicy_write_perms_ds(parentname, origin, 976 ZFS_DELEG_PERM_PROMOTE, cr); 977 } 978 dsl_dataset_rele(clone, FTAG); 979 dsl_dataset_rele(origin, FTAG); 980 } 981 dsl_pool_rele(dp, FTAG); 982 return (error); 983} 984 985/* ARGSUSED */ 986static int 987zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 988{ 989 int error; 990 991 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 992 ZFS_DELEG_PERM_RECEIVE, cr)) != 0) 993 return (error); 994 995 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 996 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 997 return (error); 998 999 return (zfs_secpolicy_write_perms(zc->zc_name, 1000 ZFS_DELEG_PERM_CREATE, cr)); 1001} 1002 1003int 1004zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr) 1005{ 1006 return (zfs_secpolicy_write_perms(name, 1007 ZFS_DELEG_PERM_SNAPSHOT, cr)); 1008} 1009 1010/* 1011 * Check for permission to create each snapshot in the nvlist. 1012 */ 1013/* ARGSUSED */ 1014static int 1015zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1016{ 1017 nvlist_t *snaps; 1018 int error; 1019 nvpair_t *pair; 1020 1021 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 1022 return (SET_ERROR(EINVAL)); 1023 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 1024 pair = nvlist_next_nvpair(snaps, pair)) { 1025 char *name = nvpair_name(pair); 1026 char *atp = strchr(name, '@'); 1027 1028 if (atp == NULL) { 1029 error = SET_ERROR(EINVAL); 1030 break; 1031 } 1032 *atp = '\0'; 1033 error = zfs_secpolicy_snapshot_perms(name, cr); 1034 *atp = '@'; 1035 if (error != 0) 1036 break; 1037 } 1038 return (error); 1039} 1040 1041/* 1042 * Check for permission to create each snapshot in the nvlist. 1043 */ 1044/* ARGSUSED */ 1045static int 1046zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1047{ 1048 int error = 0; 1049 1050 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 1051 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 1052 char *name = nvpair_name(pair); 1053 char *hashp = strchr(name, '#'); 1054 1055 if (hashp == NULL) { 1056 error = SET_ERROR(EINVAL); 1057 break; 1058 } 1059 *hashp = '\0'; 1060 error = zfs_secpolicy_write_perms(name, 1061 ZFS_DELEG_PERM_BOOKMARK, cr); 1062 *hashp = '#'; 1063 if (error != 0) 1064 break; 1065 } 1066 return (error); 1067} 1068 1069/* ARGSUSED */ 1070static int 1071zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1072{ 1073 nvpair_t *pair, *nextpair; 1074 int error = 0; 1075 1076 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL; 1077 pair = nextpair) { 1078 char *name = nvpair_name(pair); 1079 char *hashp = strchr(name, '#'); 1080 nextpair = nvlist_next_nvpair(innvl, pair); 1081 1082 if (hashp == NULL) { 1083 error = SET_ERROR(EINVAL); 1084 break; 1085 } 1086 1087 *hashp = '\0'; 1088 error = zfs_secpolicy_write_perms(name, 1089 ZFS_DELEG_PERM_DESTROY, cr); 1090 *hashp = '#'; 1091 if (error == ENOENT) { 1092 /* 1093 * Ignore any filesystems that don't exist (we consider 1094 * their bookmarks "already destroyed"). Remove 1095 * the name from the nvl here in case the filesystem 1096 * is created between now and when we try to destroy 1097 * the bookmark (in which case we don't want to 1098 * destroy it since we haven't checked for permission). 1099 */ 1100 fnvlist_remove_nvpair(innvl, pair); 1101 error = 0; 1102 } 1103 if (error != 0) 1104 break; 1105 } 1106 1107 return (error); 1108} 1109 1110/* ARGSUSED */ 1111static int 1112zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1113{ 1114 /* 1115 * Even root must have a proper TSD so that we know what pool 1116 * to log to. 1117 */ 1118 if (tsd_get(zfs_allow_log_key) == NULL) 1119 return (SET_ERROR(EPERM)); 1120 return (0); 1121} 1122 1123static int 1124zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1125{ 1126 char parentname[ZFS_MAX_DATASET_NAME_LEN]; 1127 int error; 1128 char *origin; 1129 1130 if ((error = zfs_get_parent(zc->zc_name, parentname, 1131 sizeof (parentname))) != 0) 1132 return (error); 1133 1134 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 && 1135 (error = zfs_secpolicy_write_perms(origin, 1136 ZFS_DELEG_PERM_CLONE, cr)) != 0) 1137 return (error); 1138 1139 if ((error = zfs_secpolicy_write_perms(parentname, 1140 ZFS_DELEG_PERM_CREATE, cr)) != 0) 1141 return (error); 1142 1143 return (zfs_secpolicy_write_perms(parentname, 1144 ZFS_DELEG_PERM_MOUNT, cr)); 1145} 1146 1147/* 1148 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires 1149 * SYS_CONFIG privilege, which is not available in a local zone. 1150 */ 1151/* ARGSUSED */ 1152static int 1153zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1154{ 1155 if (secpolicy_sys_config(cr, B_FALSE) != 0) 1156 return (SET_ERROR(EPERM)); 1157 1158 return (0); 1159} 1160 1161/* 1162 * Policy for object to name lookups. 1163 */ 1164/* ARGSUSED */ 1165static int 1166zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1167{ 1168 int error; 1169 1170 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0) 1171 return (0); 1172 1173 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr); 1174 return (error); 1175} 1176 1177/* 1178 * Policy for fault injection. Requires all privileges. 1179 */ 1180/* ARGSUSED */ 1181static int 1182zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1183{ 1184 return (secpolicy_zinject(cr)); 1185} 1186 1187/* ARGSUSED */ 1188static int 1189zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1190{ 1191 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value); 1192 1193 if (prop == ZPROP_INVAL) { 1194 if (!zfs_prop_user(zc->zc_value)) 1195 return (SET_ERROR(EINVAL)); 1196 return (zfs_secpolicy_write_perms(zc->zc_name, 1197 ZFS_DELEG_PERM_USERPROP, cr)); 1198 } else { 1199 return (zfs_secpolicy_setprop(zc->zc_name, prop, 1200 NULL, cr)); 1201 } 1202} 1203 1204static int 1205zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1206{ 1207 int err = zfs_secpolicy_read(zc, innvl, cr); 1208 if (err) 1209 return (err); 1210 1211 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 1212 return (SET_ERROR(EINVAL)); 1213 1214 if (zc->zc_value[0] == 0) { 1215 /* 1216 * They are asking about a posix uid/gid. If it's 1217 * themself, allow it. 1218 */ 1219 if (zc->zc_objset_type == ZFS_PROP_USERUSED || 1220 zc->zc_objset_type == ZFS_PROP_USERQUOTA) { 1221 if (zc->zc_guid == crgetuid(cr)) 1222 return (0); 1223 } else { 1224 if (groupmember(zc->zc_guid, cr)) 1225 return (0); 1226 } 1227 } 1228 1229 return (zfs_secpolicy_write_perms(zc->zc_name, 1230 userquota_perms[zc->zc_objset_type], cr)); 1231} 1232 1233static int 1234zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1235{ 1236 int err = zfs_secpolicy_read(zc, innvl, cr); 1237 if (err) 1238 return (err); 1239 1240 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 1241 return (SET_ERROR(EINVAL)); 1242 1243 return (zfs_secpolicy_write_perms(zc->zc_name, 1244 userquota_perms[zc->zc_objset_type], cr)); 1245} 1246 1247/* ARGSUSED */ 1248static int 1249zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1250{ 1251 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION, 1252 NULL, cr)); 1253} 1254 1255/* ARGSUSED */ 1256static int 1257zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1258{ 1259 nvpair_t *pair; 1260 nvlist_t *holds; 1261 int error; 1262 1263 error = nvlist_lookup_nvlist(innvl, "holds", &holds); 1264 if (error != 0) 1265 return (SET_ERROR(EINVAL)); 1266 1267 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL; 1268 pair = nvlist_next_nvpair(holds, pair)) { 1269 char fsname[ZFS_MAX_DATASET_NAME_LEN]; 1270 error = dmu_fsname(nvpair_name(pair), fsname); 1271 if (error != 0) 1272 return (error); 1273 error = zfs_secpolicy_write_perms(fsname, 1274 ZFS_DELEG_PERM_HOLD, cr); 1275 if (error != 0) 1276 return (error); 1277 } 1278 return (0); 1279} 1280 1281/* ARGSUSED */ 1282static int 1283zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1284{ 1285 nvpair_t *pair; 1286 int error; 1287 1288 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL; 1289 pair = nvlist_next_nvpair(innvl, pair)) { 1290 char fsname[ZFS_MAX_DATASET_NAME_LEN]; 1291 error = dmu_fsname(nvpair_name(pair), fsname); 1292 if (error != 0) 1293 return (error); 1294 error = zfs_secpolicy_write_perms(fsname, 1295 ZFS_DELEG_PERM_RELEASE, cr); 1296 if (error != 0) 1297 return (error); 1298 } 1299 return (0); 1300} 1301 1302/* 1303 * Policy for allowing temporary snapshots to be taken or released 1304 */ 1305static int 1306zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1307{ 1308 /* 1309 * A temporary snapshot is the same as a snapshot, 1310 * hold, destroy and release all rolled into one. 1311 * Delegated diff alone is sufficient that we allow this. 1312 */ 1313 int error; 1314 1315 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 1316 ZFS_DELEG_PERM_DIFF, cr)) == 0) 1317 return (0); 1318 1319 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr); 1320 if (error == 0) 1321 error = zfs_secpolicy_hold(zc, innvl, cr); 1322 if (error == 0) 1323 error = zfs_secpolicy_release(zc, innvl, cr); 1324 if (error == 0) 1325 error = zfs_secpolicy_destroy(zc, innvl, cr); 1326 return (error); 1327} 1328 1329/* 1330 * Returns the nvlist as specified by the user in the zfs_cmd_t. 1331 */ 1332static int 1333get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp) 1334{ 1335 char *packed; 1336 int error; 1337 nvlist_t *list = NULL; 1338 1339 /* 1340 * Read in and unpack the user-supplied nvlist. 1341 */ 1342 if (size == 0) 1343 return (SET_ERROR(EINVAL)); 1344 1345 packed = kmem_alloc(size, KM_SLEEP); 1346 1347 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size, 1348 iflag)) != 0) { 1349 kmem_free(packed, size); 1350 return (SET_ERROR(EFAULT)); 1351 } 1352 1353 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) { 1354 kmem_free(packed, size); 1355 return (error); 1356 } 1357 1358 kmem_free(packed, size); 1359 1360 *nvp = list; 1361 return (0); 1362} 1363 1364/* 1365 * Reduce the size of this nvlist until it can be serialized in 'max' bytes. 1366 * Entries will be removed from the end of the nvlist, and one int32 entry 1367 * named "N_MORE_ERRORS" will be added indicating how many entries were 1368 * removed. 1369 */ 1370static int 1371nvlist_smush(nvlist_t *errors, size_t max) 1372{ 1373 size_t size; 1374 1375 size = fnvlist_size(errors); 1376 1377 if (size > max) { 1378 nvpair_t *more_errors; 1379 int n = 0; 1380 1381 if (max < 1024) 1382 return (SET_ERROR(ENOMEM)); 1383 1384 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0); 1385 more_errors = nvlist_prev_nvpair(errors, NULL); 1386 1387 do { 1388 nvpair_t *pair = nvlist_prev_nvpair(errors, 1389 more_errors); 1390 fnvlist_remove_nvpair(errors, pair); 1391 n++; 1392 size = fnvlist_size(errors); 1393 } while (size > max); 1394 1395 fnvlist_remove_nvpair(errors, more_errors); 1396 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n); 1397 ASSERT3U(fnvlist_size(errors), <=, max); 1398 } 1399 1400 return (0); 1401} 1402 1403static int 1404put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl) 1405{ 1406 char *packed = NULL; 1407 int error = 0; 1408 size_t size; 1409 1410 size = fnvlist_size(nvl); 1411 1412 if (size > zc->zc_nvlist_dst_size) { 1413 /* 1414 * Solaris returns ENOMEM here, because even if an error is 1415 * returned from an ioctl(2), new zc_nvlist_dst_size will be 1416 * passed to the userland. This is not the case for FreeBSD. 1417 * We need to return 0, so the kernel will copy the 1418 * zc_nvlist_dst_size back and the userland can discover that a 1419 * bigger buffer is needed. 1420 */ 1421 error = 0; 1422 } else { 1423 packed = fnvlist_pack(nvl, &size); 1424 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst, 1425 size, zc->zc_iflags) != 0) 1426 error = SET_ERROR(EFAULT); 1427 fnvlist_pack_free(packed, size); 1428 } 1429 1430 zc->zc_nvlist_dst_size = size; 1431 zc->zc_nvlist_dst_filled = B_TRUE; 1432 return (error); 1433} 1434 1435static int 1436getzfsvfs(const char *dsname, zfsvfs_t **zfvp) 1437{ 1438 objset_t *os; 1439 vfs_t *vfsp; 1440 int error; 1441 1442 error = dmu_objset_hold(dsname, FTAG, &os); 1443 if (error != 0) 1444 return (error); 1445 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1446 dmu_objset_rele(os, FTAG); 1447 return (SET_ERROR(EINVAL)); 1448 } 1449 1450 mutex_enter(&os->os_user_ptr_lock); 1451 *zfvp = dmu_objset_get_user(os); 1452 if (*zfvp) { 1453 vfsp = (*zfvp)->z_vfs; 1454 vfs_ref(vfsp); 1455 } else { 1456 error = SET_ERROR(ESRCH); 1457 } 1458 mutex_exit(&os->os_user_ptr_lock); 1459 dmu_objset_rele(os, FTAG); 1460 if (error == 0) { 1461 error = vfs_busy(vfsp, 0); 1462 vfs_rel(vfsp); 1463 if (error != 0) { 1464 *zfvp = NULL; 1465 error = SET_ERROR(ESRCH); 1466 } 1467 } 1468 return (error); 1469} 1470 1471/* 1472 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which 1473 * case its z_vfs will be NULL, and it will be opened as the owner. 1474 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER, 1475 * which prevents all vnode ops from running. 1476 */ 1477static int 1478zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer) 1479{ 1480 int error = 0; 1481 1482 if (getzfsvfs(name, zfvp) != 0) 1483 error = zfsvfs_create(name, zfvp); 1484 if (error == 0) { 1485 rrm_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER : 1486 RW_READER, tag); 1487 if ((*zfvp)->z_unmounted) { 1488 /* 1489 * XXX we could probably try again, since the unmounting 1490 * thread should be just about to disassociate the 1491 * objset from the zfsvfs. 1492 */ 1493 rrm_exit(&(*zfvp)->z_teardown_lock, tag); 1494 return (SET_ERROR(EBUSY)); 1495 } 1496 } 1497 return (error); 1498} 1499 1500static void 1501zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag) 1502{ 1503 rrm_exit(&zfsvfs->z_teardown_lock, tag); 1504 1505 if (zfsvfs->z_vfs) { 1506#ifdef illumos 1507 VFS_RELE(zfsvfs->z_vfs); 1508#else 1509 vfs_unbusy(zfsvfs->z_vfs); 1510#endif 1511 } else { 1512 dmu_objset_disown(zfsvfs->z_os, zfsvfs); 1513 zfsvfs_free(zfsvfs); 1514 } 1515} 1516 1517static int 1518zfs_ioc_pool_create(zfs_cmd_t *zc) 1519{ 1520 int error; 1521 nvlist_t *config, *props = NULL; 1522 nvlist_t *rootprops = NULL; 1523 nvlist_t *zplprops = NULL; 1524 1525 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1526 zc->zc_iflags, &config)) 1527 return (error); 1528 1529 if (zc->zc_nvlist_src_size != 0 && (error = 1530 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1531 zc->zc_iflags, &props))) { 1532 nvlist_free(config); 1533 return (error); 1534 } 1535 1536 if (props) { 1537 nvlist_t *nvl = NULL; 1538 uint64_t version = SPA_VERSION; 1539 1540 (void) nvlist_lookup_uint64(props, 1541 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version); 1542 if (!SPA_VERSION_IS_SUPPORTED(version)) { 1543 error = SET_ERROR(EINVAL); 1544 goto pool_props_bad; 1545 } 1546 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl); 1547 if (nvl) { 1548 error = nvlist_dup(nvl, &rootprops, KM_SLEEP); 1549 if (error != 0) { 1550 nvlist_free(config); 1551 nvlist_free(props); 1552 return (error); 1553 } 1554 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS); 1555 } 1556 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 1557 error = zfs_fill_zplprops_root(version, rootprops, 1558 zplprops, NULL); 1559 if (error != 0) 1560 goto pool_props_bad; 1561 } 1562 1563 error = spa_create(zc->zc_name, config, props, zplprops); 1564 1565 /* 1566 * Set the remaining root properties 1567 */ 1568 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name, 1569 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0) 1570 (void) spa_destroy(zc->zc_name); 1571 1572pool_props_bad: 1573 nvlist_free(rootprops); 1574 nvlist_free(zplprops); 1575 nvlist_free(config); 1576 nvlist_free(props); 1577 1578 return (error); 1579} 1580 1581static int 1582zfs_ioc_pool_destroy(zfs_cmd_t *zc) 1583{ 1584 int error; 1585 zfs_log_history(zc); 1586 error = spa_destroy(zc->zc_name); 1587 if (error == 0) 1588 zvol_remove_minors(zc->zc_name); 1589 return (error); 1590} 1591 1592static int 1593zfs_ioc_pool_import(zfs_cmd_t *zc) 1594{ 1595 nvlist_t *config, *props = NULL; 1596 uint64_t guid; 1597 int error; 1598 1599 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1600 zc->zc_iflags, &config)) != 0) 1601 return (error); 1602 1603 if (zc->zc_nvlist_src_size != 0 && (error = 1604 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1605 zc->zc_iflags, &props))) { 1606 nvlist_free(config); 1607 return (error); 1608 } 1609 1610 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 || 1611 guid != zc->zc_guid) 1612 error = SET_ERROR(EINVAL); 1613 else 1614 error = spa_import(zc->zc_name, config, props, zc->zc_cookie); 1615 1616 if (zc->zc_nvlist_dst != 0) { 1617 int err; 1618 1619 if ((err = put_nvlist(zc, config)) != 0) 1620 error = err; 1621 } 1622 1623 nvlist_free(config); 1624 1625 nvlist_free(props); 1626 1627 return (error); 1628} 1629 1630static int 1631zfs_ioc_pool_export(zfs_cmd_t *zc) 1632{ 1633 int error; 1634 boolean_t force = (boolean_t)zc->zc_cookie; 1635 boolean_t hardforce = (boolean_t)zc->zc_guid; 1636 1637 zfs_log_history(zc); 1638 error = spa_export(zc->zc_name, NULL, force, hardforce); 1639 if (error == 0) 1640 zvol_remove_minors(zc->zc_name); 1641 return (error); 1642} 1643 1644static int 1645zfs_ioc_pool_configs(zfs_cmd_t *zc) 1646{ 1647 nvlist_t *configs; 1648 int error; 1649 1650 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL) 1651 return (SET_ERROR(EEXIST)); 1652 1653 error = put_nvlist(zc, configs); 1654 1655 nvlist_free(configs); 1656 1657 return (error); 1658} 1659 1660/* 1661 * inputs: 1662 * zc_name name of the pool 1663 * 1664 * outputs: 1665 * zc_cookie real errno 1666 * zc_nvlist_dst config nvlist 1667 * zc_nvlist_dst_size size of config nvlist 1668 */ 1669static int 1670zfs_ioc_pool_stats(zfs_cmd_t *zc) 1671{ 1672 nvlist_t *config; 1673 int error; 1674 int ret = 0; 1675 1676 error = spa_get_stats(zc->zc_name, &config, zc->zc_value, 1677 sizeof (zc->zc_value)); 1678 1679 if (config != NULL) { 1680 ret = put_nvlist(zc, config); 1681 nvlist_free(config); 1682 1683 /* 1684 * The config may be present even if 'error' is non-zero. 1685 * In this case we return success, and preserve the real errno 1686 * in 'zc_cookie'. 1687 */ 1688 zc->zc_cookie = error; 1689 } else { 1690 ret = error; 1691 } 1692 1693 return (ret); 1694} 1695 1696/* 1697 * Try to import the given pool, returning pool stats as appropriate so that 1698 * user land knows which devices are available and overall pool health. 1699 */ 1700static int 1701zfs_ioc_pool_tryimport(zfs_cmd_t *zc) 1702{ 1703 nvlist_t *tryconfig, *config; 1704 int error; 1705 1706 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1707 zc->zc_iflags, &tryconfig)) != 0) 1708 return (error); 1709 1710 config = spa_tryimport(tryconfig); 1711 1712 nvlist_free(tryconfig); 1713 1714 if (config == NULL) 1715 return (SET_ERROR(EINVAL)); 1716 1717 error = put_nvlist(zc, config); 1718 nvlist_free(config); 1719 1720 return (error); 1721} 1722 1723/* 1724 * inputs: 1725 * zc_name name of the pool 1726 * zc_cookie scan func (pool_scan_func_t) 1727 */ 1728static int 1729zfs_ioc_pool_scan(zfs_cmd_t *zc) 1730{ 1731 spa_t *spa; 1732 int error; 1733 1734 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1735 return (error); 1736 1737 if (zc->zc_cookie == POOL_SCAN_NONE) 1738 error = spa_scan_stop(spa); 1739 else 1740 error = spa_scan(spa, zc->zc_cookie); 1741 1742 spa_close(spa, FTAG); 1743 1744 return (error); 1745} 1746 1747static int 1748zfs_ioc_pool_freeze(zfs_cmd_t *zc) 1749{ 1750 spa_t *spa; 1751 int error; 1752 1753 error = spa_open(zc->zc_name, &spa, FTAG); 1754 if (error == 0) { 1755 spa_freeze(spa); 1756 spa_close(spa, FTAG); 1757 } 1758 return (error); 1759} 1760 1761static int 1762zfs_ioc_pool_upgrade(zfs_cmd_t *zc) 1763{ 1764 spa_t *spa; 1765 int error; 1766 1767 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1768 return (error); 1769 1770 if (zc->zc_cookie < spa_version(spa) || 1771 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) { 1772 spa_close(spa, FTAG); 1773 return (SET_ERROR(EINVAL)); 1774 } 1775 1776 spa_upgrade(spa, zc->zc_cookie); 1777 spa_close(spa, FTAG); 1778 1779 return (error); 1780} 1781 1782static int 1783zfs_ioc_pool_get_history(zfs_cmd_t *zc) 1784{ 1785 spa_t *spa; 1786 char *hist_buf; 1787 uint64_t size; 1788 int error; 1789 1790 if ((size = zc->zc_history_len) == 0) 1791 return (SET_ERROR(EINVAL)); 1792 1793 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1794 return (error); 1795 1796 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) { 1797 spa_close(spa, FTAG); 1798 return (SET_ERROR(ENOTSUP)); 1799 } 1800 1801 hist_buf = kmem_alloc(size, KM_SLEEP); 1802 if ((error = spa_history_get(spa, &zc->zc_history_offset, 1803 &zc->zc_history_len, hist_buf)) == 0) { 1804 error = ddi_copyout(hist_buf, 1805 (void *)(uintptr_t)zc->zc_history, 1806 zc->zc_history_len, zc->zc_iflags); 1807 } 1808 1809 spa_close(spa, FTAG); 1810 kmem_free(hist_buf, size); 1811 return (error); 1812} 1813 1814static int 1815zfs_ioc_pool_reguid(zfs_cmd_t *zc) 1816{ 1817 spa_t *spa; 1818 int error; 1819 1820 error = spa_open(zc->zc_name, &spa, FTAG); 1821 if (error == 0) { 1822 error = spa_change_guid(spa); 1823 spa_close(spa, FTAG); 1824 } 1825 return (error); 1826} 1827 1828static int 1829zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc) 1830{ 1831 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value)); 1832} 1833 1834/* 1835 * inputs: 1836 * zc_name name of filesystem 1837 * zc_obj object to find 1838 * 1839 * outputs: 1840 * zc_value name of object 1841 */ 1842static int 1843zfs_ioc_obj_to_path(zfs_cmd_t *zc) 1844{ 1845 objset_t *os; 1846 int error; 1847 1848 /* XXX reading from objset not owned */ 1849 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0) 1850 return (error); 1851 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1852 dmu_objset_rele(os, FTAG); 1853 return (SET_ERROR(EINVAL)); 1854 } 1855 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value, 1856 sizeof (zc->zc_value)); 1857 dmu_objset_rele(os, FTAG); 1858 1859 return (error); 1860} 1861 1862/* 1863 * inputs: 1864 * zc_name name of filesystem 1865 * zc_obj object to find 1866 * 1867 * outputs: 1868 * zc_stat stats on object 1869 * zc_value path to object 1870 */ 1871static int 1872zfs_ioc_obj_to_stats(zfs_cmd_t *zc) 1873{ 1874 objset_t *os; 1875 int error; 1876 1877 /* XXX reading from objset not owned */ 1878 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0) 1879 return (error); 1880 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1881 dmu_objset_rele(os, FTAG); 1882 return (SET_ERROR(EINVAL)); 1883 } 1884 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value, 1885 sizeof (zc->zc_value)); 1886 dmu_objset_rele(os, FTAG); 1887 1888 return (error); 1889} 1890 1891static int 1892zfs_ioc_vdev_add(zfs_cmd_t *zc) 1893{ 1894 spa_t *spa; 1895 int error; 1896 nvlist_t *config, **l2cache, **spares; 1897 uint_t nl2cache = 0, nspares = 0; 1898 1899 error = spa_open(zc->zc_name, &spa, FTAG); 1900 if (error != 0) 1901 return (error); 1902 1903 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1904 zc->zc_iflags, &config); 1905 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE, 1906 &l2cache, &nl2cache); 1907 1908 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES, 1909 &spares, &nspares); 1910 1911#ifdef illumos 1912 /* 1913 * A root pool with concatenated devices is not supported. 1914 * Thus, can not add a device to a root pool. 1915 * 1916 * Intent log device can not be added to a rootpool because 1917 * during mountroot, zil is replayed, a seperated log device 1918 * can not be accessed during the mountroot time. 1919 * 1920 * l2cache and spare devices are ok to be added to a rootpool. 1921 */ 1922 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) { 1923 nvlist_free(config); 1924 spa_close(spa, FTAG); 1925 return (SET_ERROR(EDOM)); 1926 } 1927#endif /* illumos */ 1928 1929 if (error == 0) { 1930 error = spa_vdev_add(spa, config); 1931 nvlist_free(config); 1932 } 1933 spa_close(spa, FTAG); 1934 return (error); 1935} 1936 1937/* 1938 * inputs: 1939 * zc_name name of the pool 1940 * zc_nvlist_conf nvlist of devices to remove 1941 * zc_cookie to stop the remove? 1942 */ 1943static int 1944zfs_ioc_vdev_remove(zfs_cmd_t *zc) 1945{ 1946 spa_t *spa; 1947 int error; 1948 1949 error = spa_open(zc->zc_name, &spa, FTAG); 1950 if (error != 0) 1951 return (error); 1952 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE); 1953 spa_close(spa, FTAG); 1954 return (error); 1955} 1956 1957static int 1958zfs_ioc_vdev_set_state(zfs_cmd_t *zc) 1959{ 1960 spa_t *spa; 1961 int error; 1962 vdev_state_t newstate = VDEV_STATE_UNKNOWN; 1963 1964 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1965 return (error); 1966 switch (zc->zc_cookie) { 1967 case VDEV_STATE_ONLINE: 1968 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate); 1969 break; 1970 1971 case VDEV_STATE_OFFLINE: 1972 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj); 1973 break; 1974 1975 case VDEV_STATE_FAULTED: 1976 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED && 1977 zc->zc_obj != VDEV_AUX_EXTERNAL) 1978 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED; 1979 1980 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj); 1981 break; 1982 1983 case VDEV_STATE_DEGRADED: 1984 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED && 1985 zc->zc_obj != VDEV_AUX_EXTERNAL) 1986 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED; 1987 1988 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj); 1989 break; 1990 1991 default: 1992 error = SET_ERROR(EINVAL); 1993 } 1994 zc->zc_cookie = newstate; 1995 spa_close(spa, FTAG); 1996 return (error); 1997} 1998 1999static int 2000zfs_ioc_vdev_attach(zfs_cmd_t *zc) 2001{ 2002 spa_t *spa; 2003 int replacing = zc->zc_cookie; 2004 nvlist_t *config; 2005 int error; 2006 2007 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 2008 return (error); 2009 2010 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 2011 zc->zc_iflags, &config)) == 0) { 2012 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing); 2013 nvlist_free(config); 2014 } 2015 2016 spa_close(spa, FTAG); 2017 return (error); 2018} 2019 2020static int 2021zfs_ioc_vdev_detach(zfs_cmd_t *zc) 2022{ 2023 spa_t *spa; 2024 int error; 2025 2026 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 2027 return (error); 2028 2029 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE); 2030 2031 spa_close(spa, FTAG); 2032 return (error); 2033} 2034 2035static int 2036zfs_ioc_vdev_split(zfs_cmd_t *zc) 2037{ 2038 spa_t *spa; 2039 nvlist_t *config, *props = NULL; 2040 int error; 2041 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT); 2042 2043 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 2044 return (error); 2045 2046 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 2047 zc->zc_iflags, &config)) { 2048 spa_close(spa, FTAG); 2049 return (error); 2050 } 2051 2052 if (zc->zc_nvlist_src_size != 0 && (error = 2053 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2054 zc->zc_iflags, &props))) { 2055 spa_close(spa, FTAG); 2056 nvlist_free(config); 2057 return (error); 2058 } 2059 2060 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp); 2061 2062 spa_close(spa, FTAG); 2063 2064 nvlist_free(config); 2065 nvlist_free(props); 2066 2067 return (error); 2068} 2069 2070static int 2071zfs_ioc_vdev_setpath(zfs_cmd_t *zc) 2072{ 2073 spa_t *spa; 2074 char *path = zc->zc_value; 2075 uint64_t guid = zc->zc_guid; 2076 int error; 2077 2078 error = spa_open(zc->zc_name, &spa, FTAG); 2079 if (error != 0) 2080 return (error); 2081 2082 error = spa_vdev_setpath(spa, guid, path); 2083 spa_close(spa, FTAG); 2084 return (error); 2085} 2086 2087static int 2088zfs_ioc_vdev_setfru(zfs_cmd_t *zc) 2089{ 2090 spa_t *spa; 2091 char *fru = zc->zc_value; 2092 uint64_t guid = zc->zc_guid; 2093 int error; 2094 2095 error = spa_open(zc->zc_name, &spa, FTAG); 2096 if (error != 0) 2097 return (error); 2098 2099 error = spa_vdev_setfru(spa, guid, fru); 2100 spa_close(spa, FTAG); 2101 return (error); 2102} 2103 2104static int 2105zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os) 2106{ 2107 int error = 0; 2108 nvlist_t *nv; 2109 2110 dmu_objset_fast_stat(os, &zc->zc_objset_stats); 2111 2112 if (zc->zc_nvlist_dst != 0 && 2113 (error = dsl_prop_get_all(os, &nv)) == 0) { 2114 dmu_objset_stats(os, nv); 2115 /* 2116 * NB: zvol_get_stats() will read the objset contents, 2117 * which we aren't supposed to do with a 2118 * DS_MODE_USER hold, because it could be 2119 * inconsistent. So this is a bit of a workaround... 2120 * XXX reading with out owning 2121 */ 2122 if (!zc->zc_objset_stats.dds_inconsistent && 2123 dmu_objset_type(os) == DMU_OST_ZVOL) { 2124 error = zvol_get_stats(os, nv); 2125 if (error == EIO) 2126 return (error); 2127 VERIFY0(error); 2128 } 2129 error = put_nvlist(zc, nv); 2130 nvlist_free(nv); 2131 } 2132 2133 return (error); 2134} 2135 2136/* 2137 * inputs: 2138 * zc_name name of filesystem 2139 * zc_nvlist_dst_size size of buffer for property nvlist 2140 * 2141 * outputs: 2142 * zc_objset_stats stats 2143 * zc_nvlist_dst property nvlist 2144 * zc_nvlist_dst_size size of property nvlist 2145 */ 2146static int 2147zfs_ioc_objset_stats(zfs_cmd_t *zc) 2148{ 2149 objset_t *os; 2150 int error; 2151 2152 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 2153 if (error == 0) { 2154 error = zfs_ioc_objset_stats_impl(zc, os); 2155 dmu_objset_rele(os, FTAG); 2156 } 2157 2158 if (error == ENOMEM) 2159 error = 0; 2160 return (error); 2161} 2162 2163/* 2164 * inputs: 2165 * zc_name name of filesystem 2166 * zc_nvlist_dst_size size of buffer for property nvlist 2167 * 2168 * outputs: 2169 * zc_nvlist_dst received property nvlist 2170 * zc_nvlist_dst_size size of received property nvlist 2171 * 2172 * Gets received properties (distinct from local properties on or after 2173 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from 2174 * local property values. 2175 */ 2176static int 2177zfs_ioc_objset_recvd_props(zfs_cmd_t *zc) 2178{ 2179 int error = 0; 2180 nvlist_t *nv; 2181 2182 /* 2183 * Without this check, we would return local property values if the 2184 * caller has not already received properties on or after 2185 * SPA_VERSION_RECVD_PROPS. 2186 */ 2187 if (!dsl_prop_get_hasrecvd(zc->zc_name)) 2188 return (SET_ERROR(ENOTSUP)); 2189 2190 if (zc->zc_nvlist_dst != 0 && 2191 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) { 2192 error = put_nvlist(zc, nv); 2193 nvlist_free(nv); 2194 } 2195 2196 return (error); 2197} 2198 2199static int 2200nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop) 2201{ 2202 uint64_t value; 2203 int error; 2204 2205 /* 2206 * zfs_get_zplprop() will either find a value or give us 2207 * the default value (if there is one). 2208 */ 2209 if ((error = zfs_get_zplprop(os, prop, &value)) != 0) 2210 return (error); 2211 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0); 2212 return (0); 2213} 2214 2215/* 2216 * inputs: 2217 * zc_name name of filesystem 2218 * zc_nvlist_dst_size size of buffer for zpl property nvlist 2219 * 2220 * outputs: 2221 * zc_nvlist_dst zpl property nvlist 2222 * zc_nvlist_dst_size size of zpl property nvlist 2223 */ 2224static int 2225zfs_ioc_objset_zplprops(zfs_cmd_t *zc) 2226{ 2227 objset_t *os; 2228 int err; 2229 2230 /* XXX reading without owning */ 2231 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os)) 2232 return (err); 2233 2234 dmu_objset_fast_stat(os, &zc->zc_objset_stats); 2235 2236 /* 2237 * NB: nvl_add_zplprop() will read the objset contents, 2238 * which we aren't supposed to do with a DS_MODE_USER 2239 * hold, because it could be inconsistent. 2240 */ 2241 if (zc->zc_nvlist_dst != 0 && 2242 !zc->zc_objset_stats.dds_inconsistent && 2243 dmu_objset_type(os) == DMU_OST_ZFS) { 2244 nvlist_t *nv; 2245 2246 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2247 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 && 2248 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 && 2249 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 && 2250 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0) 2251 err = put_nvlist(zc, nv); 2252 nvlist_free(nv); 2253 } else { 2254 err = SET_ERROR(ENOENT); 2255 } 2256 dmu_objset_rele(os, FTAG); 2257 return (err); 2258} 2259 2260boolean_t 2261dataset_name_hidden(const char *name) 2262{ 2263 /* 2264 * Skip over datasets that are not visible in this zone, 2265 * internal datasets (which have a $ in their name), and 2266 * temporary datasets (which have a % in their name). 2267 */ 2268 if (strchr(name, '$') != NULL) 2269 return (B_TRUE); 2270 if (strchr(name, '%') != NULL) 2271 return (B_TRUE); 2272 if (!INGLOBALZONE(curthread) && !zone_dataset_visible(name, NULL)) 2273 return (B_TRUE); 2274 return (B_FALSE); 2275} 2276 2277/* 2278 * inputs: 2279 * zc_name name of filesystem 2280 * zc_cookie zap cursor 2281 * zc_nvlist_dst_size size of buffer for property nvlist 2282 * 2283 * outputs: 2284 * zc_name name of next filesystem 2285 * zc_cookie zap cursor 2286 * zc_objset_stats stats 2287 * zc_nvlist_dst property nvlist 2288 * zc_nvlist_dst_size size of property nvlist 2289 */ 2290static int 2291zfs_ioc_dataset_list_next(zfs_cmd_t *zc) 2292{ 2293 objset_t *os; 2294 int error; 2295 char *p; 2296 size_t orig_len = strlen(zc->zc_name); 2297 2298top: 2299 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) { 2300 if (error == ENOENT) 2301 error = SET_ERROR(ESRCH); 2302 return (error); 2303 } 2304 2305 p = strrchr(zc->zc_name, '/'); 2306 if (p == NULL || p[1] != '\0') 2307 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name)); 2308 p = zc->zc_name + strlen(zc->zc_name); 2309 2310 do { 2311 error = dmu_dir_list_next(os, 2312 sizeof (zc->zc_name) - (p - zc->zc_name), p, 2313 NULL, &zc->zc_cookie); 2314 if (error == ENOENT) 2315 error = SET_ERROR(ESRCH); 2316 } while (error == 0 && dataset_name_hidden(zc->zc_name)); 2317 dmu_objset_rele(os, FTAG); 2318 2319 /* 2320 * If it's an internal dataset (ie. with a '$' in its name), 2321 * don't try to get stats for it, otherwise we'll return ENOENT. 2322 */ 2323 if (error == 0 && strchr(zc->zc_name, '$') == NULL) { 2324 error = zfs_ioc_objset_stats(zc); /* fill in the stats */ 2325 if (error == ENOENT) { 2326 /* We lost a race with destroy, get the next one. */ 2327 zc->zc_name[orig_len] = '\0'; 2328 goto top; 2329 } 2330 } 2331 return (error); 2332} 2333 2334/* 2335 * inputs: 2336 * zc_name name of filesystem 2337 * zc_cookie zap cursor 2338 * zc_nvlist_dst_size size of buffer for property nvlist 2339 * zc_simple when set, only name is requested 2340 * 2341 * outputs: 2342 * zc_name name of next snapshot 2343 * zc_objset_stats stats 2344 * zc_nvlist_dst property nvlist 2345 * zc_nvlist_dst_size size of property nvlist 2346 */ 2347static int 2348zfs_ioc_snapshot_list_next(zfs_cmd_t *zc) 2349{ 2350 objset_t *os; 2351 int error; 2352 2353 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 2354 if (error != 0) { 2355 return (error == ENOENT ? ESRCH : error); 2356 } 2357 2358 /* 2359 * A dataset name of maximum length cannot have any snapshots, 2360 * so exit immediately. 2361 */ 2362 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= 2363 ZFS_MAX_DATASET_NAME_LEN) { 2364 dmu_objset_rele(os, FTAG); 2365 return (SET_ERROR(ESRCH)); 2366 } 2367 2368 error = dmu_snapshot_list_next(os, 2369 sizeof (zc->zc_name) - strlen(zc->zc_name), 2370 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie, 2371 NULL); 2372 2373 if (error == 0 && !zc->zc_simple) { 2374 dsl_dataset_t *ds; 2375 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool; 2376 2377 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds); 2378 if (error == 0) { 2379 objset_t *ossnap; 2380 2381 error = dmu_objset_from_ds(ds, &ossnap); 2382 if (error == 0) 2383 error = zfs_ioc_objset_stats_impl(zc, ossnap); 2384 dsl_dataset_rele(ds, FTAG); 2385 } 2386 } else if (error == ENOENT) { 2387 error = SET_ERROR(ESRCH); 2388 } 2389 2390 dmu_objset_rele(os, FTAG); 2391 /* if we failed, undo the @ that we tacked on to zc_name */ 2392 if (error != 0) 2393 *strchr(zc->zc_name, '@') = '\0'; 2394 return (error); 2395} 2396 2397static int 2398zfs_prop_set_userquota(const char *dsname, nvpair_t *pair) 2399{ 2400 const char *propname = nvpair_name(pair); 2401 uint64_t *valary; 2402 unsigned int vallen; 2403 const char *domain; 2404 char *dash; 2405 zfs_userquota_prop_t type; 2406 uint64_t rid; 2407 uint64_t quota; 2408 zfsvfs_t *zfsvfs; 2409 int err; 2410 2411 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2412 nvlist_t *attrs; 2413 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 2414 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2415 &pair) != 0) 2416 return (SET_ERROR(EINVAL)); 2417 } 2418 2419 /* 2420 * A correctly constructed propname is encoded as 2421 * userquota@<rid>-<domain>. 2422 */ 2423 if ((dash = strchr(propname, '-')) == NULL || 2424 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 || 2425 vallen != 3) 2426 return (SET_ERROR(EINVAL)); 2427 2428 domain = dash + 1; 2429 type = valary[0]; 2430 rid = valary[1]; 2431 quota = valary[2]; 2432 2433 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE); 2434 if (err == 0) { 2435 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota); 2436 zfsvfs_rele(zfsvfs, FTAG); 2437 } 2438 2439 return (err); 2440} 2441 2442/* 2443 * If the named property is one that has a special function to set its value, 2444 * return 0 on success and a positive error code on failure; otherwise if it is 2445 * not one of the special properties handled by this function, return -1. 2446 * 2447 * XXX: It would be better for callers of the property interface if we handled 2448 * these special cases in dsl_prop.c (in the dsl layer). 2449 */ 2450static int 2451zfs_prop_set_special(const char *dsname, zprop_source_t source, 2452 nvpair_t *pair) 2453{ 2454 const char *propname = nvpair_name(pair); 2455 zfs_prop_t prop = zfs_name_to_prop(propname); 2456 uint64_t intval; 2457 int err = -1; 2458 2459 if (prop == ZPROP_INVAL) { 2460 if (zfs_prop_userquota(propname)) 2461 return (zfs_prop_set_userquota(dsname, pair)); 2462 return (-1); 2463 } 2464 2465 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2466 nvlist_t *attrs; 2467 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 2468 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2469 &pair) == 0); 2470 } 2471 2472 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING) 2473 return (-1); 2474 2475 VERIFY(0 == nvpair_value_uint64(pair, &intval)); 2476 2477 switch (prop) { 2478 case ZFS_PROP_QUOTA: 2479 err = dsl_dir_set_quota(dsname, source, intval); 2480 break; 2481 case ZFS_PROP_REFQUOTA: 2482 err = dsl_dataset_set_refquota(dsname, source, intval); 2483 break; 2484 case ZFS_PROP_FILESYSTEM_LIMIT: 2485 case ZFS_PROP_SNAPSHOT_LIMIT: 2486 if (intval == UINT64_MAX) { 2487 /* clearing the limit, just do it */ 2488 err = 0; 2489 } else { 2490 err = dsl_dir_activate_fs_ss_limit(dsname); 2491 } 2492 /* 2493 * Set err to -1 to force the zfs_set_prop_nvlist code down the 2494 * default path to set the value in the nvlist. 2495 */ 2496 if (err == 0) 2497 err = -1; 2498 break; 2499 case ZFS_PROP_RESERVATION: 2500 err = dsl_dir_set_reservation(dsname, source, intval); 2501 break; 2502 case ZFS_PROP_REFRESERVATION: 2503 err = dsl_dataset_set_refreservation(dsname, source, intval); 2504 break; 2505 case ZFS_PROP_VOLSIZE: 2506 err = zvol_set_volsize(dsname, intval); 2507 break; 2508 case ZFS_PROP_VERSION: 2509 { 2510 zfsvfs_t *zfsvfs; 2511 2512 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0) 2513 break; 2514 2515 err = zfs_set_version(zfsvfs, intval); 2516 zfsvfs_rele(zfsvfs, FTAG); 2517 2518 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) { 2519 zfs_cmd_t *zc; 2520 2521 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP); 2522 (void) strcpy(zc->zc_name, dsname); 2523 (void) zfs_ioc_userspace_upgrade(zc); 2524 kmem_free(zc, sizeof (zfs_cmd_t)); 2525 } 2526 break; 2527 } 2528 default: 2529 err = -1; 2530 } 2531 2532 return (err); 2533} 2534 2535/* 2536 * This function is best effort. If it fails to set any of the given properties, 2537 * it continues to set as many as it can and returns the last error 2538 * encountered. If the caller provides a non-NULL errlist, it will be filled in 2539 * with the list of names of all the properties that failed along with the 2540 * corresponding error numbers. 2541 * 2542 * If every property is set successfully, zero is returned and errlist is not 2543 * modified. 2544 */ 2545int 2546zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl, 2547 nvlist_t *errlist) 2548{ 2549 nvpair_t *pair; 2550 nvpair_t *propval; 2551 int rv = 0; 2552 uint64_t intval; 2553 char *strval; 2554 nvlist_t *genericnvl = fnvlist_alloc(); 2555 nvlist_t *retrynvl = fnvlist_alloc(); 2556 2557retry: 2558 pair = NULL; 2559 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { 2560 const char *propname = nvpair_name(pair); 2561 zfs_prop_t prop = zfs_name_to_prop(propname); 2562 int err = 0; 2563 2564 /* decode the property value */ 2565 propval = pair; 2566 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2567 nvlist_t *attrs; 2568 attrs = fnvpair_value_nvlist(pair); 2569 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2570 &propval) != 0) 2571 err = SET_ERROR(EINVAL); 2572 } 2573 2574 /* Validate value type */ 2575 if (err == 0 && prop == ZPROP_INVAL) { 2576 if (zfs_prop_user(propname)) { 2577 if (nvpair_type(propval) != DATA_TYPE_STRING) 2578 err = SET_ERROR(EINVAL); 2579 } else if (zfs_prop_userquota(propname)) { 2580 if (nvpair_type(propval) != 2581 DATA_TYPE_UINT64_ARRAY) 2582 err = SET_ERROR(EINVAL); 2583 } else { 2584 err = SET_ERROR(EINVAL); 2585 } 2586 } else if (err == 0) { 2587 if (nvpair_type(propval) == DATA_TYPE_STRING) { 2588 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING) 2589 err = SET_ERROR(EINVAL); 2590 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) { 2591 const char *unused; 2592 2593 intval = fnvpair_value_uint64(propval); 2594 2595 switch (zfs_prop_get_type(prop)) { 2596 case PROP_TYPE_NUMBER: 2597 break; 2598 case PROP_TYPE_STRING: 2599 err = SET_ERROR(EINVAL); 2600 break; 2601 case PROP_TYPE_INDEX: 2602 if (zfs_prop_index_to_string(prop, 2603 intval, &unused) != 0) 2604 err = SET_ERROR(EINVAL); 2605 break; 2606 default: 2607 cmn_err(CE_PANIC, 2608 "unknown property type"); 2609 } 2610 } else { 2611 err = SET_ERROR(EINVAL); 2612 } 2613 } 2614 2615 /* Validate permissions */ 2616 if (err == 0) 2617 err = zfs_check_settable(dsname, pair, CRED()); 2618 2619 if (err == 0) { 2620 err = zfs_prop_set_special(dsname, source, pair); 2621 if (err == -1) { 2622 /* 2623 * For better performance we build up a list of 2624 * properties to set in a single transaction. 2625 */ 2626 err = nvlist_add_nvpair(genericnvl, pair); 2627 } else if (err != 0 && nvl != retrynvl) { 2628 /* 2629 * This may be a spurious error caused by 2630 * receiving quota and reservation out of order. 2631 * Try again in a second pass. 2632 */ 2633 err = nvlist_add_nvpair(retrynvl, pair); 2634 } 2635 } 2636 2637 if (err != 0) { 2638 if (errlist != NULL) 2639 fnvlist_add_int32(errlist, propname, err); 2640 rv = err; 2641 } 2642 } 2643 2644 if (nvl != retrynvl && !nvlist_empty(retrynvl)) { 2645 nvl = retrynvl; 2646 goto retry; 2647 } 2648 2649 if (!nvlist_empty(genericnvl) && 2650 dsl_props_set(dsname, source, genericnvl) != 0) { 2651 /* 2652 * If this fails, we still want to set as many properties as we 2653 * can, so try setting them individually. 2654 */ 2655 pair = NULL; 2656 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) { 2657 const char *propname = nvpair_name(pair); 2658 int err = 0; 2659 2660 propval = pair; 2661 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2662 nvlist_t *attrs; 2663 attrs = fnvpair_value_nvlist(pair); 2664 propval = fnvlist_lookup_nvpair(attrs, 2665 ZPROP_VALUE); 2666 } 2667 2668 if (nvpair_type(propval) == DATA_TYPE_STRING) { 2669 strval = fnvpair_value_string(propval); 2670 err = dsl_prop_set_string(dsname, propname, 2671 source, strval); 2672 } else { 2673 intval = fnvpair_value_uint64(propval); 2674 err = dsl_prop_set_int(dsname, propname, source, 2675 intval); 2676 } 2677 2678 if (err != 0) { 2679 if (errlist != NULL) { 2680 fnvlist_add_int32(errlist, propname, 2681 err); 2682 } 2683 rv = err; 2684 } 2685 } 2686 } 2687 nvlist_free(genericnvl); 2688 nvlist_free(retrynvl); 2689 2690 return (rv); 2691} 2692 2693/* 2694 * Check that all the properties are valid user properties. 2695 */ 2696static int 2697zfs_check_userprops(const char *fsname, nvlist_t *nvl) 2698{ 2699 nvpair_t *pair = NULL; 2700 int error = 0; 2701 2702 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { 2703 const char *propname = nvpair_name(pair); 2704 2705 if (!zfs_prop_user(propname) || 2706 nvpair_type(pair) != DATA_TYPE_STRING) 2707 return (SET_ERROR(EINVAL)); 2708 2709 if (error = zfs_secpolicy_write_perms(fsname, 2710 ZFS_DELEG_PERM_USERPROP, CRED())) 2711 return (error); 2712 2713 if (strlen(propname) >= ZAP_MAXNAMELEN) 2714 return (SET_ERROR(ENAMETOOLONG)); 2715 2716 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN) 2717 return (E2BIG); 2718 } 2719 return (0); 2720} 2721 2722static void 2723props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops) 2724{ 2725 nvpair_t *pair; 2726 2727 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2728 2729 pair = NULL; 2730 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) { 2731 if (nvlist_exists(skipped, nvpair_name(pair))) 2732 continue; 2733 2734 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0); 2735 } 2736} 2737 2738static int 2739clear_received_props(const char *dsname, nvlist_t *props, 2740 nvlist_t *skipped) 2741{ 2742 int err = 0; 2743 nvlist_t *cleared_props = NULL; 2744 props_skip(props, skipped, &cleared_props); 2745 if (!nvlist_empty(cleared_props)) { 2746 /* 2747 * Acts on local properties until the dataset has received 2748 * properties at least once on or after SPA_VERSION_RECVD_PROPS. 2749 */ 2750 zprop_source_t flags = (ZPROP_SRC_NONE | 2751 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0)); 2752 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL); 2753 } 2754 nvlist_free(cleared_props); 2755 return (err); 2756} 2757 2758/* 2759 * inputs: 2760 * zc_name name of filesystem 2761 * zc_value name of property to set 2762 * zc_nvlist_src{_size} nvlist of properties to apply 2763 * zc_cookie received properties flag 2764 * 2765 * outputs: 2766 * zc_nvlist_dst{_size} error for each unapplied received property 2767 */ 2768static int 2769zfs_ioc_set_prop(zfs_cmd_t *zc) 2770{ 2771 nvlist_t *nvl; 2772 boolean_t received = zc->zc_cookie; 2773 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED : 2774 ZPROP_SRC_LOCAL); 2775 nvlist_t *errors; 2776 int error; 2777 2778 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2779 zc->zc_iflags, &nvl)) != 0) 2780 return (error); 2781 2782 if (received) { 2783 nvlist_t *origprops; 2784 2785 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) { 2786 (void) clear_received_props(zc->zc_name, 2787 origprops, nvl); 2788 nvlist_free(origprops); 2789 } 2790 2791 error = dsl_prop_set_hasrecvd(zc->zc_name); 2792 } 2793 2794 errors = fnvlist_alloc(); 2795 if (error == 0) 2796 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors); 2797 2798 if (zc->zc_nvlist_dst != 0 && errors != NULL) { 2799 (void) put_nvlist(zc, errors); 2800 } 2801 2802 nvlist_free(errors); 2803 nvlist_free(nvl); 2804 return (error); 2805} 2806 2807/* 2808 * inputs: 2809 * zc_name name of filesystem 2810 * zc_value name of property to inherit 2811 * zc_cookie revert to received value if TRUE 2812 * 2813 * outputs: none 2814 */ 2815static int 2816zfs_ioc_inherit_prop(zfs_cmd_t *zc) 2817{ 2818 const char *propname = zc->zc_value; 2819 zfs_prop_t prop = zfs_name_to_prop(propname); 2820 boolean_t received = zc->zc_cookie; 2821 zprop_source_t source = (received 2822 ? ZPROP_SRC_NONE /* revert to received value, if any */ 2823 : ZPROP_SRC_INHERITED); /* explicitly inherit */ 2824 2825 if (received) { 2826 nvlist_t *dummy; 2827 nvpair_t *pair; 2828 zprop_type_t type; 2829 int err; 2830 2831 /* 2832 * zfs_prop_set_special() expects properties in the form of an 2833 * nvpair with type info. 2834 */ 2835 if (prop == ZPROP_INVAL) { 2836 if (!zfs_prop_user(propname)) 2837 return (SET_ERROR(EINVAL)); 2838 2839 type = PROP_TYPE_STRING; 2840 } else if (prop == ZFS_PROP_VOLSIZE || 2841 prop == ZFS_PROP_VERSION) { 2842 return (SET_ERROR(EINVAL)); 2843 } else { 2844 type = zfs_prop_get_type(prop); 2845 } 2846 2847 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2848 2849 switch (type) { 2850 case PROP_TYPE_STRING: 2851 VERIFY(0 == nvlist_add_string(dummy, propname, "")); 2852 break; 2853 case PROP_TYPE_NUMBER: 2854 case PROP_TYPE_INDEX: 2855 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0)); 2856 break; 2857 default: 2858 nvlist_free(dummy); 2859 return (SET_ERROR(EINVAL)); 2860 } 2861 2862 pair = nvlist_next_nvpair(dummy, NULL); 2863 err = zfs_prop_set_special(zc->zc_name, source, pair); 2864 nvlist_free(dummy); 2865 if (err != -1) 2866 return (err); /* special property already handled */ 2867 } else { 2868 /* 2869 * Only check this in the non-received case. We want to allow 2870 * 'inherit -S' to revert non-inheritable properties like quota 2871 * and reservation to the received or default values even though 2872 * they are not considered inheritable. 2873 */ 2874 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop)) 2875 return (SET_ERROR(EINVAL)); 2876 } 2877 2878 /* property name has been validated by zfs_secpolicy_inherit_prop() */ 2879 return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source)); 2880} 2881 2882static int 2883zfs_ioc_pool_set_props(zfs_cmd_t *zc) 2884{ 2885 nvlist_t *props; 2886 spa_t *spa; 2887 int error; 2888 nvpair_t *pair; 2889 2890 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2891 zc->zc_iflags, &props)) 2892 return (error); 2893 2894 /* 2895 * If the only property is the configfile, then just do a spa_lookup() 2896 * to handle the faulted case. 2897 */ 2898 pair = nvlist_next_nvpair(props, NULL); 2899 if (pair != NULL && strcmp(nvpair_name(pair), 2900 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 && 2901 nvlist_next_nvpair(props, pair) == NULL) { 2902 mutex_enter(&spa_namespace_lock); 2903 if ((spa = spa_lookup(zc->zc_name)) != NULL) { 2904 spa_configfile_set(spa, props, B_FALSE); 2905 spa_config_sync(spa, B_FALSE, B_TRUE); 2906 } 2907 mutex_exit(&spa_namespace_lock); 2908 if (spa != NULL) { 2909 nvlist_free(props); 2910 return (0); 2911 } 2912 } 2913 2914 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { 2915 nvlist_free(props); 2916 return (error); 2917 } 2918 2919 error = spa_prop_set(spa, props); 2920 2921 nvlist_free(props); 2922 spa_close(spa, FTAG); 2923 2924 return (error); 2925} 2926 2927static int 2928zfs_ioc_pool_get_props(zfs_cmd_t *zc) 2929{ 2930 spa_t *spa; 2931 int error; 2932 nvlist_t *nvp = NULL; 2933 2934 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { 2935 /* 2936 * If the pool is faulted, there may be properties we can still 2937 * get (such as altroot and cachefile), so attempt to get them 2938 * anyway. 2939 */ 2940 mutex_enter(&spa_namespace_lock); 2941 if ((spa = spa_lookup(zc->zc_name)) != NULL) 2942 error = spa_prop_get(spa, &nvp); 2943 mutex_exit(&spa_namespace_lock); 2944 } else { 2945 error = spa_prop_get(spa, &nvp); 2946 spa_close(spa, FTAG); 2947 } 2948 2949 if (error == 0 && zc->zc_nvlist_dst != 0) 2950 error = put_nvlist(zc, nvp); 2951 else 2952 error = SET_ERROR(EFAULT); 2953 2954 nvlist_free(nvp); 2955 return (error); 2956} 2957 2958/* 2959 * inputs: 2960 * zc_name name of filesystem 2961 * zc_nvlist_src{_size} nvlist of delegated permissions 2962 * zc_perm_action allow/unallow flag 2963 * 2964 * outputs: none 2965 */ 2966static int 2967zfs_ioc_set_fsacl(zfs_cmd_t *zc) 2968{ 2969 int error; 2970 nvlist_t *fsaclnv = NULL; 2971 2972 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2973 zc->zc_iflags, &fsaclnv)) != 0) 2974 return (error); 2975 2976 /* 2977 * Verify nvlist is constructed correctly 2978 */ 2979 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) { 2980 nvlist_free(fsaclnv); 2981 return (SET_ERROR(EINVAL)); 2982 } 2983 2984 /* 2985 * If we don't have PRIV_SYS_MOUNT, then validate 2986 * that user is allowed to hand out each permission in 2987 * the nvlist(s) 2988 */ 2989 2990 error = secpolicy_zfs(CRED()); 2991 if (error != 0) { 2992 if (zc->zc_perm_action == B_FALSE) { 2993 error = dsl_deleg_can_allow(zc->zc_name, 2994 fsaclnv, CRED()); 2995 } else { 2996 error = dsl_deleg_can_unallow(zc->zc_name, 2997 fsaclnv, CRED()); 2998 } 2999 } 3000 3001 if (error == 0) 3002 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action); 3003 3004 nvlist_free(fsaclnv); 3005 return (error); 3006} 3007 3008/* 3009 * inputs: 3010 * zc_name name of filesystem 3011 * 3012 * outputs: 3013 * zc_nvlist_src{_size} nvlist of delegated permissions 3014 */ 3015static int 3016zfs_ioc_get_fsacl(zfs_cmd_t *zc) 3017{ 3018 nvlist_t *nvp; 3019 int error; 3020 3021 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) { 3022 error = put_nvlist(zc, nvp); 3023 nvlist_free(nvp); 3024 } 3025 3026 return (error); 3027} 3028 3029/* 3030 * Search the vfs list for a specified resource. Returns a pointer to it 3031 * or NULL if no suitable entry is found. The caller of this routine 3032 * is responsible for releasing the returned vfs pointer. 3033 */ 3034static vfs_t * 3035zfs_get_vfs(const char *resource) 3036{ 3037 vfs_t *vfsp; 3038 3039 mtx_lock(&mountlist_mtx); 3040 TAILQ_FOREACH(vfsp, &mountlist, mnt_list) { 3041 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) { 3042 if (vfs_busy(vfsp, MBF_MNTLSTLOCK) != 0) 3043 vfsp = NULL; 3044 break; 3045 } 3046 } 3047 if (vfsp == NULL) 3048 mtx_unlock(&mountlist_mtx); 3049 return (vfsp); 3050} 3051 3052/* ARGSUSED */ 3053static void 3054zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) 3055{ 3056 zfs_creat_t *zct = arg; 3057 3058 zfs_create_fs(os, cr, zct->zct_zplprops, tx); 3059} 3060 3061#define ZFS_PROP_UNDEFINED ((uint64_t)-1) 3062 3063/* 3064 * inputs: 3065 * os parent objset pointer (NULL if root fs) 3066 * fuids_ok fuids allowed in this version of the spa? 3067 * sa_ok SAs allowed in this version of the spa? 3068 * createprops list of properties requested by creator 3069 * 3070 * outputs: 3071 * zplprops values for the zplprops we attach to the master node object 3072 * is_ci true if requested file system will be purely case-insensitive 3073 * 3074 * Determine the settings for utf8only, normalization and 3075 * casesensitivity. Specific values may have been requested by the 3076 * creator and/or we can inherit values from the parent dataset. If 3077 * the file system is of too early a vintage, a creator can not 3078 * request settings for these properties, even if the requested 3079 * setting is the default value. We don't actually want to create dsl 3080 * properties for these, so remove them from the source nvlist after 3081 * processing. 3082 */ 3083static int 3084zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver, 3085 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops, 3086 nvlist_t *zplprops, boolean_t *is_ci) 3087{ 3088 uint64_t sense = ZFS_PROP_UNDEFINED; 3089 uint64_t norm = ZFS_PROP_UNDEFINED; 3090 uint64_t u8 = ZFS_PROP_UNDEFINED; 3091 3092 ASSERT(zplprops != NULL); 3093 3094 /* 3095 * Pull out creator prop choices, if any. 3096 */ 3097 if (createprops) { 3098 (void) nvlist_lookup_uint64(createprops, 3099 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver); 3100 (void) nvlist_lookup_uint64(createprops, 3101 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm); 3102 (void) nvlist_remove_all(createprops, 3103 zfs_prop_to_name(ZFS_PROP_NORMALIZE)); 3104 (void) nvlist_lookup_uint64(createprops, 3105 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8); 3106 (void) nvlist_remove_all(createprops, 3107 zfs_prop_to_name(ZFS_PROP_UTF8ONLY)); 3108 (void) nvlist_lookup_uint64(createprops, 3109 zfs_prop_to_name(ZFS_PROP_CASE), &sense); 3110 (void) nvlist_remove_all(createprops, 3111 zfs_prop_to_name(ZFS_PROP_CASE)); 3112 } 3113 3114 /* 3115 * If the zpl version requested is whacky or the file system 3116 * or pool is version is too "young" to support normalization 3117 * and the creator tried to set a value for one of the props, 3118 * error out. 3119 */ 3120 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) || 3121 (zplver >= ZPL_VERSION_FUID && !fuids_ok) || 3122 (zplver >= ZPL_VERSION_SA && !sa_ok) || 3123 (zplver < ZPL_VERSION_NORMALIZATION && 3124 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED || 3125 sense != ZFS_PROP_UNDEFINED))) 3126 return (SET_ERROR(ENOTSUP)); 3127 3128 /* 3129 * Put the version in the zplprops 3130 */ 3131 VERIFY(nvlist_add_uint64(zplprops, 3132 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0); 3133 3134 if (norm == ZFS_PROP_UNDEFINED) 3135 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0); 3136 VERIFY(nvlist_add_uint64(zplprops, 3137 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0); 3138 3139 /* 3140 * If we're normalizing, names must always be valid UTF-8 strings. 3141 */ 3142 if (norm) 3143 u8 = 1; 3144 if (u8 == ZFS_PROP_UNDEFINED) 3145 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0); 3146 VERIFY(nvlist_add_uint64(zplprops, 3147 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0); 3148 3149 if (sense == ZFS_PROP_UNDEFINED) 3150 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0); 3151 VERIFY(nvlist_add_uint64(zplprops, 3152 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0); 3153 3154 if (is_ci) 3155 *is_ci = (sense == ZFS_CASE_INSENSITIVE); 3156 3157 return (0); 3158} 3159 3160static int 3161zfs_fill_zplprops(const char *dataset, nvlist_t *createprops, 3162 nvlist_t *zplprops, boolean_t *is_ci) 3163{ 3164 boolean_t fuids_ok, sa_ok; 3165 uint64_t zplver = ZPL_VERSION; 3166 objset_t *os = NULL; 3167 char parentname[ZFS_MAX_DATASET_NAME_LEN]; 3168 char *cp; 3169 spa_t *spa; 3170 uint64_t spa_vers; 3171 int error; 3172 3173 (void) strlcpy(parentname, dataset, sizeof (parentname)); 3174 cp = strrchr(parentname, '/'); 3175 ASSERT(cp != NULL); 3176 cp[0] = '\0'; 3177 3178 if ((error = spa_open(dataset, &spa, FTAG)) != 0) 3179 return (error); 3180 3181 spa_vers = spa_version(spa); 3182 spa_close(spa, FTAG); 3183 3184 zplver = zfs_zpl_version_map(spa_vers); 3185 fuids_ok = (zplver >= ZPL_VERSION_FUID); 3186 sa_ok = (zplver >= ZPL_VERSION_SA); 3187 3188 /* 3189 * Open parent object set so we can inherit zplprop values. 3190 */ 3191 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0) 3192 return (error); 3193 3194 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops, 3195 zplprops, is_ci); 3196 dmu_objset_rele(os, FTAG); 3197 return (error); 3198} 3199 3200static int 3201zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops, 3202 nvlist_t *zplprops, boolean_t *is_ci) 3203{ 3204 boolean_t fuids_ok; 3205 boolean_t sa_ok; 3206 uint64_t zplver = ZPL_VERSION; 3207 int error; 3208 3209 zplver = zfs_zpl_version_map(spa_vers); 3210 fuids_ok = (zplver >= ZPL_VERSION_FUID); 3211 sa_ok = (zplver >= ZPL_VERSION_SA); 3212 3213 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok, 3214 createprops, zplprops, is_ci); 3215 return (error); 3216} 3217 3218/* 3219 * innvl: { 3220 * "type" -> dmu_objset_type_t (int32) 3221 * (optional) "props" -> { prop -> value } 3222 * } 3223 * 3224 * outnvl: propname -> error code (int32) 3225 */ 3226static int 3227zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3228{ 3229 int error = 0; 3230 zfs_creat_t zct = { 0 }; 3231 nvlist_t *nvprops = NULL; 3232 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx); 3233 int32_t type32; 3234 dmu_objset_type_t type; 3235 boolean_t is_insensitive = B_FALSE; 3236 3237 if (nvlist_lookup_int32(innvl, "type", &type32) != 0) 3238 return (SET_ERROR(EINVAL)); 3239 type = type32; 3240 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops); 3241 3242 switch (type) { 3243 case DMU_OST_ZFS: 3244 cbfunc = zfs_create_cb; 3245 break; 3246 3247 case DMU_OST_ZVOL: 3248 cbfunc = zvol_create_cb; 3249 break; 3250 3251 default: 3252 cbfunc = NULL; 3253 break; 3254 } 3255 if (strchr(fsname, '@') || 3256 strchr(fsname, '%')) 3257 return (SET_ERROR(EINVAL)); 3258 3259 zct.zct_props = nvprops; 3260 3261 if (cbfunc == NULL) 3262 return (SET_ERROR(EINVAL)); 3263 3264 if (type == DMU_OST_ZVOL) { 3265 uint64_t volsize, volblocksize; 3266 3267 if (nvprops == NULL) 3268 return (SET_ERROR(EINVAL)); 3269 if (nvlist_lookup_uint64(nvprops, 3270 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0) 3271 return (SET_ERROR(EINVAL)); 3272 3273 if ((error = nvlist_lookup_uint64(nvprops, 3274 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), 3275 &volblocksize)) != 0 && error != ENOENT) 3276 return (SET_ERROR(EINVAL)); 3277 3278 if (error != 0) 3279 volblocksize = zfs_prop_default_numeric( 3280 ZFS_PROP_VOLBLOCKSIZE); 3281 3282 if ((error = zvol_check_volblocksize( 3283 volblocksize)) != 0 || 3284 (error = zvol_check_volsize(volsize, 3285 volblocksize)) != 0) 3286 return (error); 3287 } else if (type == DMU_OST_ZFS) { 3288 int error; 3289 3290 /* 3291 * We have to have normalization and 3292 * case-folding flags correct when we do the 3293 * file system creation, so go figure them out 3294 * now. 3295 */ 3296 VERIFY(nvlist_alloc(&zct.zct_zplprops, 3297 NV_UNIQUE_NAME, KM_SLEEP) == 0); 3298 error = zfs_fill_zplprops(fsname, nvprops, 3299 zct.zct_zplprops, &is_insensitive); 3300 if (error != 0) { 3301 nvlist_free(zct.zct_zplprops); 3302 return (error); 3303 } 3304 } 3305 3306 error = dmu_objset_create(fsname, type, 3307 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct); 3308 nvlist_free(zct.zct_zplprops); 3309 3310 /* 3311 * It would be nice to do this atomically. 3312 */ 3313 if (error == 0) { 3314 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL, 3315 nvprops, outnvl); 3316 if (error != 0) 3317 (void) dsl_destroy_head(fsname); 3318 } 3319#ifdef __FreeBSD__ 3320 if (error == 0 && type == DMU_OST_ZVOL) 3321 zvol_create_minors(fsname); 3322#endif 3323 return (error); 3324} 3325 3326/* 3327 * innvl: { 3328 * "origin" -> name of origin snapshot 3329 * (optional) "props" -> { prop -> value } 3330 * } 3331 * 3332 * outnvl: propname -> error code (int32) 3333 */ 3334static int 3335zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3336{ 3337 int error = 0; 3338 nvlist_t *nvprops = NULL; 3339 char *origin_name; 3340 3341 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0) 3342 return (SET_ERROR(EINVAL)); 3343 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops); 3344 3345 if (strchr(fsname, '@') || 3346 strchr(fsname, '%')) 3347 return (SET_ERROR(EINVAL)); 3348 3349 if (dataset_namecheck(origin_name, NULL, NULL) != 0) 3350 return (SET_ERROR(EINVAL)); 3351 error = dmu_objset_clone(fsname, origin_name); 3352 if (error != 0) 3353 return (error); 3354 3355 /* 3356 * It would be nice to do this atomically. 3357 */ 3358 if (error == 0) { 3359 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL, 3360 nvprops, outnvl); 3361 if (error != 0) 3362 (void) dsl_destroy_head(fsname); 3363 } 3364#ifdef __FreeBSD__ 3365 if (error == 0) 3366 zvol_create_minors(fsname); 3367#endif 3368 return (error); 3369} 3370 3371/* 3372 * innvl: { 3373 * "snaps" -> { snapshot1, snapshot2 } 3374 * (optional) "props" -> { prop -> value (string) } 3375 * } 3376 * 3377 * outnvl: snapshot -> error code (int32) 3378 */ 3379static int 3380zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3381{ 3382 nvlist_t *snaps; 3383 nvlist_t *props = NULL; 3384 int error, poollen; 3385 nvpair_t *pair; 3386 3387 (void) nvlist_lookup_nvlist(innvl, "props", &props); 3388 if ((error = zfs_check_userprops(poolname, props)) != 0) 3389 return (error); 3390 3391 if (!nvlist_empty(props) && 3392 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS)) 3393 return (SET_ERROR(ENOTSUP)); 3394 3395 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 3396 return (SET_ERROR(EINVAL)); 3397 poollen = strlen(poolname); 3398 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 3399 pair = nvlist_next_nvpair(snaps, pair)) { 3400 const char *name = nvpair_name(pair); 3401 const char *cp = strchr(name, '@'); 3402 3403 /* 3404 * The snap name must contain an @, and the part after it must 3405 * contain only valid characters. 3406 */ 3407 if (cp == NULL || 3408 zfs_component_namecheck(cp + 1, NULL, NULL) != 0) 3409 return (SET_ERROR(EINVAL)); 3410 3411 /* 3412 * The snap must be in the specified pool. 3413 */ 3414 if (strncmp(name, poolname, poollen) != 0 || 3415 (name[poollen] != '/' && name[poollen] != '@')) 3416 return (SET_ERROR(EXDEV)); 3417 3418 /* This must be the only snap of this fs. */ 3419 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair); 3420 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) { 3421 if (strncmp(name, nvpair_name(pair2), cp - name + 1) 3422 == 0) { 3423 return (SET_ERROR(EXDEV)); 3424 } 3425 } 3426 } 3427 3428 error = dsl_dataset_snapshot(snaps, props, outnvl); 3429 return (error); 3430} 3431 3432/* 3433 * innvl: "message" -> string 3434 */ 3435/* ARGSUSED */ 3436static int 3437zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl) 3438{ 3439 char *message; 3440 spa_t *spa; 3441 int error; 3442 char *poolname; 3443 3444 /* 3445 * The poolname in the ioctl is not set, we get it from the TSD, 3446 * which was set at the end of the last successful ioctl that allows 3447 * logging. The secpolicy func already checked that it is set. 3448 * Only one log ioctl is allowed after each successful ioctl, so 3449 * we clear the TSD here. 3450 */ 3451 poolname = tsd_get(zfs_allow_log_key); 3452 (void) tsd_set(zfs_allow_log_key, NULL); 3453 error = spa_open(poolname, &spa, FTAG); 3454 strfree(poolname); 3455 if (error != 0) 3456 return (error); 3457 3458 if (nvlist_lookup_string(innvl, "message", &message) != 0) { 3459 spa_close(spa, FTAG); 3460 return (SET_ERROR(EINVAL)); 3461 } 3462 3463 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) { 3464 spa_close(spa, FTAG); 3465 return (SET_ERROR(ENOTSUP)); 3466 } 3467 3468 error = spa_history_log(spa, message); 3469 spa_close(spa, FTAG); 3470 return (error); 3471} 3472 3473/* 3474 * The dp_config_rwlock must not be held when calling this, because the 3475 * unmount may need to write out data. 3476 * 3477 * This function is best-effort. Callers must deal gracefully if it 3478 * remains mounted (or is remounted after this call). 3479 * 3480 * Returns 0 if the argument is not a snapshot, or it is not currently a 3481 * filesystem, or we were able to unmount it. Returns error code otherwise. 3482 */ 3483int 3484zfs_unmount_snap(const char *snapname) 3485{ 3486 vfs_t *vfsp; 3487 zfsvfs_t *zfsvfs; 3488 int err; 3489 3490 if (strchr(snapname, '@') == NULL) 3491 return (0); 3492 3493 vfsp = zfs_get_vfs(snapname); 3494 if (vfsp == NULL) 3495 return (0); 3496 3497 zfsvfs = vfsp->vfs_data; 3498 ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os))); 3499 3500 err = vn_vfswlock(vfsp->vfs_vnodecovered); 3501#ifdef illumos 3502 VFS_RELE(vfsp); 3503#else 3504 vfs_unbusy(vfsp); 3505#endif 3506 if (err != 0) 3507 return (SET_ERROR(err)); 3508 3509 /* 3510 * Always force the unmount for snapshots. 3511 */ 3512 3513#ifdef illumos 3514 (void) dounmount(vfsp, MS_FORCE, kcred); 3515#else 3516 vfs_ref(vfsp); 3517 (void) dounmount(vfsp, MS_FORCE, curthread); 3518#endif 3519 return (0); 3520} 3521 3522/* ARGSUSED */ 3523static int 3524zfs_unmount_snap_cb(const char *snapname, void *arg) 3525{ 3526 return (zfs_unmount_snap(snapname)); 3527} 3528 3529/* 3530 * When a clone is destroyed, its origin may also need to be destroyed, 3531 * in which case it must be unmounted. This routine will do that unmount 3532 * if necessary. 3533 */ 3534void 3535zfs_destroy_unmount_origin(const char *fsname) 3536{ 3537 int error; 3538 objset_t *os; 3539 dsl_dataset_t *ds; 3540 3541 error = dmu_objset_hold(fsname, FTAG, &os); 3542 if (error != 0) 3543 return; 3544 ds = dmu_objset_ds(os); 3545 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) { 3546 char originname[ZFS_MAX_DATASET_NAME_LEN]; 3547 dsl_dataset_name(ds->ds_prev, originname); 3548 dmu_objset_rele(os, FTAG); 3549 (void) zfs_unmount_snap(originname); 3550 } else { 3551 dmu_objset_rele(os, FTAG); 3552 } 3553} 3554 3555/* 3556 * innvl: { 3557 * "snaps" -> { snapshot1, snapshot2 } 3558 * (optional boolean) "defer" 3559 * } 3560 * 3561 * outnvl: snapshot -> error code (int32) 3562 * 3563 */ 3564/* ARGSUSED */ 3565static int 3566zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3567{ 3568 int error, poollen; 3569 nvlist_t *snaps; 3570 nvpair_t *pair; 3571 boolean_t defer; 3572 3573 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 3574 return (SET_ERROR(EINVAL)); 3575 defer = nvlist_exists(innvl, "defer"); 3576 3577 poollen = strlen(poolname); 3578 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 3579 pair = nvlist_next_nvpair(snaps, pair)) { 3580 const char *name = nvpair_name(pair); 3581 3582 /* 3583 * The snap must be in the specified pool to prevent the 3584 * invalid removal of zvol minors below. 3585 */ 3586 if (strncmp(name, poolname, poollen) != 0 || 3587 (name[poollen] != '/' && name[poollen] != '@')) 3588 return (SET_ERROR(EXDEV)); 3589 3590 error = zfs_unmount_snap(name); 3591 if (error != 0) 3592 return (error); 3593#if defined(__FreeBSD__) 3594 zvol_remove_minors(name); 3595#endif 3596 } 3597 3598 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl)); 3599} 3600 3601/* 3602 * Create bookmarks. Bookmark names are of the form <fs>#<bmark>. 3603 * All bookmarks must be in the same pool. 3604 * 3605 * innvl: { 3606 * bookmark1 -> snapshot1, bookmark2 -> snapshot2 3607 * } 3608 * 3609 * outnvl: bookmark -> error code (int32) 3610 * 3611 */ 3612/* ARGSUSED */ 3613static int 3614zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3615{ 3616 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 3617 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 3618 char *snap_name; 3619 3620 /* 3621 * Verify the snapshot argument. 3622 */ 3623 if (nvpair_value_string(pair, &snap_name) != 0) 3624 return (SET_ERROR(EINVAL)); 3625 3626 3627 /* Verify that the keys (bookmarks) are unique */ 3628 for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair); 3629 pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) { 3630 if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0) 3631 return (SET_ERROR(EINVAL)); 3632 } 3633 } 3634 3635 return (dsl_bookmark_create(innvl, outnvl)); 3636} 3637 3638/* 3639 * innvl: { 3640 * property 1, property 2, ... 3641 * } 3642 * 3643 * outnvl: { 3644 * bookmark name 1 -> { property 1, property 2, ... }, 3645 * bookmark name 2 -> { property 1, property 2, ... } 3646 * } 3647 * 3648 */ 3649static int 3650zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3651{ 3652 return (dsl_get_bookmarks(fsname, innvl, outnvl)); 3653} 3654 3655/* 3656 * innvl: { 3657 * bookmark name 1, bookmark name 2 3658 * } 3659 * 3660 * outnvl: bookmark -> error code (int32) 3661 * 3662 */ 3663static int 3664zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl, 3665 nvlist_t *outnvl) 3666{ 3667 int error, poollen; 3668 3669 poollen = strlen(poolname); 3670 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 3671 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 3672 const char *name = nvpair_name(pair); 3673 const char *cp = strchr(name, '#'); 3674 3675 /* 3676 * The bookmark name must contain an #, and the part after it 3677 * must contain only valid characters. 3678 */ 3679 if (cp == NULL || 3680 zfs_component_namecheck(cp + 1, NULL, NULL) != 0) 3681 return (SET_ERROR(EINVAL)); 3682 3683 /* 3684 * The bookmark must be in the specified pool. 3685 */ 3686 if (strncmp(name, poolname, poollen) != 0 || 3687 (name[poollen] != '/' && name[poollen] != '#')) 3688 return (SET_ERROR(EXDEV)); 3689 } 3690 3691 error = dsl_bookmark_destroy(innvl, outnvl); 3692 return (error); 3693} 3694 3695/* 3696 * inputs: 3697 * zc_name name of dataset to destroy 3698 * zc_objset_type type of objset 3699 * zc_defer_destroy mark for deferred destroy 3700 * 3701 * outputs: none 3702 */ 3703static int 3704zfs_ioc_destroy(zfs_cmd_t *zc) 3705{ 3706 int err; 3707 3708 if (zc->zc_objset_type == DMU_OST_ZFS) { 3709 err = zfs_unmount_snap(zc->zc_name); 3710 if (err != 0) 3711 return (err); 3712 } 3713 3714 if (strchr(zc->zc_name, '@')) 3715 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy); 3716 else 3717 err = dsl_destroy_head(zc->zc_name); 3718 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0) 3719#ifdef __FreeBSD__ 3720 zvol_remove_minors(zc->zc_name); 3721#else 3722 (void) zvol_remove_minor(zc->zc_name); 3723#endif 3724 return (err); 3725} 3726 3727/* 3728 * fsname is name of dataset to rollback (to most recent snapshot) 3729 * 3730 * innvl is not used. 3731 * 3732 * outnvl: "target" -> name of most recent snapshot 3733 * } 3734 */ 3735/* ARGSUSED */ 3736static int 3737zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl) 3738{ 3739 zfsvfs_t *zfsvfs; 3740 int error; 3741 3742 if (getzfsvfs(fsname, &zfsvfs) == 0) { 3743 error = zfs_suspend_fs(zfsvfs); 3744 if (error == 0) { 3745 int resume_err; 3746 3747 error = dsl_dataset_rollback(fsname, zfsvfs, outnvl); 3748 resume_err = zfs_resume_fs(zfsvfs, fsname); 3749 error = error ? error : resume_err; 3750 } 3751#ifdef illumos 3752 VFS_RELE(zfsvfs->z_vfs); 3753#else 3754 vfs_unbusy(zfsvfs->z_vfs); 3755#endif 3756 } else { 3757 error = dsl_dataset_rollback(fsname, NULL, outnvl); 3758 } 3759 return (error); 3760} 3761 3762static int 3763recursive_unmount(const char *fsname, void *arg) 3764{ 3765 const char *snapname = arg; 3766 char fullname[ZFS_MAX_DATASET_NAME_LEN]; 3767 3768 (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname); 3769 return (zfs_unmount_snap(fullname)); 3770} 3771 3772/* 3773 * inputs: 3774 * zc_name old name of dataset 3775 * zc_value new name of dataset 3776 * zc_cookie recursive flag (only valid for snapshots) 3777 * 3778 * outputs: none 3779 */ 3780static int 3781zfs_ioc_rename(zfs_cmd_t *zc) 3782{ 3783 boolean_t recursive = zc->zc_cookie & 1; 3784 char *at; 3785 boolean_t allow_mounted = B_TRUE; 3786 3787#ifdef __FreeBSD__ 3788 allow_mounted = (zc->zc_cookie & 2) != 0; 3789#endif 3790 3791 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0'; 3792 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 3793 strchr(zc->zc_value, '%')) 3794 return (SET_ERROR(EINVAL)); 3795 3796 at = strchr(zc->zc_name, '@'); 3797 if (at != NULL) { 3798 /* snaps must be in same fs */ 3799 int error; 3800 3801 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1)) 3802 return (SET_ERROR(EXDEV)); 3803 *at = '\0'; 3804 if (zc->zc_objset_type == DMU_OST_ZFS && !allow_mounted) { 3805 error = dmu_objset_find(zc->zc_name, 3806 recursive_unmount, at + 1, 3807 recursive ? DS_FIND_CHILDREN : 0); 3808 if (error != 0) { 3809 *at = '@'; 3810 return (error); 3811 } 3812 } 3813 error = dsl_dataset_rename_snapshot(zc->zc_name, 3814 at + 1, strchr(zc->zc_value, '@') + 1, recursive); 3815 *at = '@'; 3816 3817 return (error); 3818 } else { 3819#ifdef illumos 3820 if (zc->zc_objset_type == DMU_OST_ZVOL) 3821 (void) zvol_remove_minor(zc->zc_name); 3822#endif 3823 return (dsl_dir_rename(zc->zc_name, zc->zc_value)); 3824 } 3825} 3826 3827static int 3828zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr) 3829{ 3830 const char *propname = nvpair_name(pair); 3831 boolean_t issnap = (strchr(dsname, '@') != NULL); 3832 zfs_prop_t prop = zfs_name_to_prop(propname); 3833 uint64_t intval; 3834 int err; 3835 3836 if (prop == ZPROP_INVAL) { 3837 if (zfs_prop_user(propname)) { 3838 if (err = zfs_secpolicy_write_perms(dsname, 3839 ZFS_DELEG_PERM_USERPROP, cr)) 3840 return (err); 3841 return (0); 3842 } 3843 3844 if (!issnap && zfs_prop_userquota(propname)) { 3845 const char *perm = NULL; 3846 const char *uq_prefix = 3847 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA]; 3848 const char *gq_prefix = 3849 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA]; 3850 3851 if (strncmp(propname, uq_prefix, 3852 strlen(uq_prefix)) == 0) { 3853 perm = ZFS_DELEG_PERM_USERQUOTA; 3854 } else if (strncmp(propname, gq_prefix, 3855 strlen(gq_prefix)) == 0) { 3856 perm = ZFS_DELEG_PERM_GROUPQUOTA; 3857 } else { 3858 /* USERUSED and GROUPUSED are read-only */ 3859 return (SET_ERROR(EINVAL)); 3860 } 3861 3862 if (err = zfs_secpolicy_write_perms(dsname, perm, cr)) 3863 return (err); 3864 return (0); 3865 } 3866 3867 return (SET_ERROR(EINVAL)); 3868 } 3869 3870 if (issnap) 3871 return (SET_ERROR(EINVAL)); 3872 3873 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 3874 /* 3875 * dsl_prop_get_all_impl() returns properties in this 3876 * format. 3877 */ 3878 nvlist_t *attrs; 3879 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 3880 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 3881 &pair) == 0); 3882 } 3883 3884 /* 3885 * Check that this value is valid for this pool version 3886 */ 3887 switch (prop) { 3888 case ZFS_PROP_COMPRESSION: 3889 /* 3890 * If the user specified gzip compression, make sure 3891 * the SPA supports it. We ignore any errors here since 3892 * we'll catch them later. 3893 */ 3894 if (nvpair_value_uint64(pair, &intval) == 0) { 3895 if (intval >= ZIO_COMPRESS_GZIP_1 && 3896 intval <= ZIO_COMPRESS_GZIP_9 && 3897 zfs_earlier_version(dsname, 3898 SPA_VERSION_GZIP_COMPRESSION)) { 3899 return (SET_ERROR(ENOTSUP)); 3900 } 3901 3902 if (intval == ZIO_COMPRESS_ZLE && 3903 zfs_earlier_version(dsname, 3904 SPA_VERSION_ZLE_COMPRESSION)) 3905 return (SET_ERROR(ENOTSUP)); 3906 3907 if (intval == ZIO_COMPRESS_LZ4) { 3908 spa_t *spa; 3909 3910 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 3911 return (err); 3912 3913 if (!spa_feature_is_enabled(spa, 3914 SPA_FEATURE_LZ4_COMPRESS)) { 3915 spa_close(spa, FTAG); 3916 return (SET_ERROR(ENOTSUP)); 3917 } 3918 spa_close(spa, FTAG); 3919 } 3920 3921 /* 3922 * If this is a bootable dataset then 3923 * verify that the compression algorithm 3924 * is supported for booting. We must return 3925 * something other than ENOTSUP since it 3926 * implies a downrev pool version. 3927 */ 3928 if (zfs_is_bootfs(dsname) && 3929 !BOOTFS_COMPRESS_VALID(intval)) { 3930 return (SET_ERROR(ERANGE)); 3931 } 3932 } 3933 break; 3934 3935 case ZFS_PROP_COPIES: 3936 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS)) 3937 return (SET_ERROR(ENOTSUP)); 3938 break; 3939 3940 case ZFS_PROP_RECORDSIZE: 3941 /* Record sizes above 128k need the feature to be enabled */ 3942 if (nvpair_value_uint64(pair, &intval) == 0 && 3943 intval > SPA_OLD_MAXBLOCKSIZE) { 3944 spa_t *spa; 3945 3946 /* 3947 * If this is a bootable dataset then 3948 * the we don't allow large (>128K) blocks, 3949 * because GRUB doesn't support them. 3950 */ 3951 if (zfs_is_bootfs(dsname) && 3952 intval > SPA_OLD_MAXBLOCKSIZE) { 3953 return (SET_ERROR(ERANGE)); 3954 } 3955 3956 /* 3957 * We don't allow setting the property above 1MB, 3958 * unless the tunable has been changed. 3959 */ 3960 if (intval > zfs_max_recordsize || 3961 intval > SPA_MAXBLOCKSIZE) 3962 return (SET_ERROR(ERANGE)); 3963 3964 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 3965 return (err); 3966 3967 if (!spa_feature_is_enabled(spa, 3968 SPA_FEATURE_LARGE_BLOCKS)) { 3969 spa_close(spa, FTAG); 3970 return (SET_ERROR(ENOTSUP)); 3971 } 3972 spa_close(spa, FTAG); 3973 } 3974 break; 3975 3976 case ZFS_PROP_SHARESMB: 3977 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID)) 3978 return (SET_ERROR(ENOTSUP)); 3979 break; 3980 3981 case ZFS_PROP_ACLINHERIT: 3982 if (nvpair_type(pair) == DATA_TYPE_UINT64 && 3983 nvpair_value_uint64(pair, &intval) == 0) { 3984 if (intval == ZFS_ACL_PASSTHROUGH_X && 3985 zfs_earlier_version(dsname, 3986 SPA_VERSION_PASSTHROUGH_X)) 3987 return (SET_ERROR(ENOTSUP)); 3988 } 3989 break; 3990 3991 case ZFS_PROP_CHECKSUM: 3992 case ZFS_PROP_DEDUP: 3993 { 3994 spa_feature_t feature; 3995 spa_t *spa; 3996 3997 /* dedup feature version checks */ 3998 if (prop == ZFS_PROP_DEDUP && 3999 zfs_earlier_version(dsname, SPA_VERSION_DEDUP)) 4000 return (SET_ERROR(ENOTSUP)); 4001 4002 if (nvpair_value_uint64(pair, &intval) != 0) 4003 return (SET_ERROR(EINVAL)); 4004 4005 /* check prop value is enabled in features */ 4006 feature = zio_checksum_to_feature(intval & ZIO_CHECKSUM_MASK); 4007 if (feature == SPA_FEATURE_NONE) 4008 break; 4009 4010 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 4011 return (err); 4012 /* 4013 * Salted checksums are not supported on root pools. 4014 */ 4015 if (spa_bootfs(spa) != 0 && 4016 intval < ZIO_CHECKSUM_FUNCTIONS && 4017 (zio_checksum_table[intval].ci_flags & 4018 ZCHECKSUM_FLAG_SALTED)) { 4019 spa_close(spa, FTAG); 4020 return (SET_ERROR(ERANGE)); 4021 } 4022 if (!spa_feature_is_enabled(spa, feature)) { 4023 spa_close(spa, FTAG); 4024 return (SET_ERROR(ENOTSUP)); 4025 } 4026 spa_close(spa, FTAG); 4027 break; 4028 } 4029 } 4030 4031 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED())); 4032} 4033 4034/* 4035 * Checks for a race condition to make sure we don't increment a feature flag 4036 * multiple times. 4037 */ 4038static int 4039zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx) 4040{ 4041 spa_t *spa = dmu_tx_pool(tx)->dp_spa; 4042 spa_feature_t *featurep = arg; 4043 4044 if (!spa_feature_is_active(spa, *featurep)) 4045 return (0); 4046 else 4047 return (SET_ERROR(EBUSY)); 4048} 4049 4050/* 4051 * The callback invoked on feature activation in the sync task caused by 4052 * zfs_prop_activate_feature. 4053 */ 4054static void 4055zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx) 4056{ 4057 spa_t *spa = dmu_tx_pool(tx)->dp_spa; 4058 spa_feature_t *featurep = arg; 4059 4060 spa_feature_incr(spa, *featurep, tx); 4061} 4062 4063/* 4064 * Activates a feature on a pool in response to a property setting. This 4065 * creates a new sync task which modifies the pool to reflect the feature 4066 * as being active. 4067 */ 4068static int 4069zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature) 4070{ 4071 int err; 4072 4073 /* EBUSY here indicates that the feature is already active */ 4074 err = dsl_sync_task(spa_name(spa), 4075 zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync, 4076 &feature, 2, ZFS_SPACE_CHECK_RESERVED); 4077 4078 if (err != 0 && err != EBUSY) 4079 return (err); 4080 else 4081 return (0); 4082} 4083 4084/* 4085 * Removes properties from the given props list that fail permission checks 4086 * needed to clear them and to restore them in case of a receive error. For each 4087 * property, make sure we have both set and inherit permissions. 4088 * 4089 * Returns the first error encountered if any permission checks fail. If the 4090 * caller provides a non-NULL errlist, it also gives the complete list of names 4091 * of all the properties that failed a permission check along with the 4092 * corresponding error numbers. The caller is responsible for freeing the 4093 * returned errlist. 4094 * 4095 * If every property checks out successfully, zero is returned and the list 4096 * pointed at by errlist is NULL. 4097 */ 4098static int 4099zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist) 4100{ 4101 zfs_cmd_t *zc; 4102 nvpair_t *pair, *next_pair; 4103 nvlist_t *errors; 4104 int err, rv = 0; 4105 4106 if (props == NULL) 4107 return (0); 4108 4109 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0); 4110 4111 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP); 4112 (void) strcpy(zc->zc_name, dataset); 4113 pair = nvlist_next_nvpair(props, NULL); 4114 while (pair != NULL) { 4115 next_pair = nvlist_next_nvpair(props, pair); 4116 4117 (void) strcpy(zc->zc_value, nvpair_name(pair)); 4118 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 || 4119 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) { 4120 VERIFY(nvlist_remove_nvpair(props, pair) == 0); 4121 VERIFY(nvlist_add_int32(errors, 4122 zc->zc_value, err) == 0); 4123 } 4124 pair = next_pair; 4125 } 4126 kmem_free(zc, sizeof (zfs_cmd_t)); 4127 4128 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) { 4129 nvlist_free(errors); 4130 errors = NULL; 4131 } else { 4132 VERIFY(nvpair_value_int32(pair, &rv) == 0); 4133 } 4134 4135 if (errlist == NULL) 4136 nvlist_free(errors); 4137 else 4138 *errlist = errors; 4139 4140 return (rv); 4141} 4142 4143static boolean_t 4144propval_equals(nvpair_t *p1, nvpair_t *p2) 4145{ 4146 if (nvpair_type(p1) == DATA_TYPE_NVLIST) { 4147 /* dsl_prop_get_all_impl() format */ 4148 nvlist_t *attrs; 4149 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0); 4150 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 4151 &p1) == 0); 4152 } 4153 4154 if (nvpair_type(p2) == DATA_TYPE_NVLIST) { 4155 nvlist_t *attrs; 4156 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0); 4157 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 4158 &p2) == 0); 4159 } 4160 4161 if (nvpair_type(p1) != nvpair_type(p2)) 4162 return (B_FALSE); 4163 4164 if (nvpair_type(p1) == DATA_TYPE_STRING) { 4165 char *valstr1, *valstr2; 4166 4167 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0); 4168 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0); 4169 return (strcmp(valstr1, valstr2) == 0); 4170 } else { 4171 uint64_t intval1, intval2; 4172 4173 VERIFY(nvpair_value_uint64(p1, &intval1) == 0); 4174 VERIFY(nvpair_value_uint64(p2, &intval2) == 0); 4175 return (intval1 == intval2); 4176 } 4177} 4178 4179/* 4180 * Remove properties from props if they are not going to change (as determined 4181 * by comparison with origprops). Remove them from origprops as well, since we 4182 * do not need to clear or restore properties that won't change. 4183 */ 4184static void 4185props_reduce(nvlist_t *props, nvlist_t *origprops) 4186{ 4187 nvpair_t *pair, *next_pair; 4188 4189 if (origprops == NULL) 4190 return; /* all props need to be received */ 4191 4192 pair = nvlist_next_nvpair(props, NULL); 4193 while (pair != NULL) { 4194 const char *propname = nvpair_name(pair); 4195 nvpair_t *match; 4196 4197 next_pair = nvlist_next_nvpair(props, pair); 4198 4199 if ((nvlist_lookup_nvpair(origprops, propname, 4200 &match) != 0) || !propval_equals(pair, match)) 4201 goto next; /* need to set received value */ 4202 4203 /* don't clear the existing received value */ 4204 (void) nvlist_remove_nvpair(origprops, match); 4205 /* don't bother receiving the property */ 4206 (void) nvlist_remove_nvpair(props, pair); 4207next: 4208 pair = next_pair; 4209 } 4210} 4211 4212/* 4213 * Extract properties that cannot be set PRIOR to the receipt of a dataset. 4214 * For example, refquota cannot be set until after the receipt of a dataset, 4215 * because in replication streams, an older/earlier snapshot may exceed the 4216 * refquota. We want to receive the older/earlier snapshot, but setting 4217 * refquota pre-receipt will set the dsl's ACTUAL quota, which will prevent 4218 * the older/earlier snapshot from being received (with EDQUOT). 4219 * 4220 * The ZFS test "zfs_receive_011_pos" demonstrates such a scenario. 4221 * 4222 * libzfs will need to be judicious handling errors encountered by props 4223 * extracted by this function. 4224 */ 4225static nvlist_t * 4226extract_delay_props(nvlist_t *props) 4227{ 4228 nvlist_t *delayprops; 4229 nvpair_t *nvp, *tmp; 4230 static const zfs_prop_t delayable[] = { ZFS_PROP_REFQUOTA, 0 }; 4231 int i; 4232 4233 VERIFY(nvlist_alloc(&delayprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 4234 4235 for (nvp = nvlist_next_nvpair(props, NULL); nvp != NULL; 4236 nvp = nvlist_next_nvpair(props, nvp)) { 4237 /* 4238 * strcmp() is safe because zfs_prop_to_name() always returns 4239 * a bounded string. 4240 */ 4241 for (i = 0; delayable[i] != 0; i++) { 4242 if (strcmp(zfs_prop_to_name(delayable[i]), 4243 nvpair_name(nvp)) == 0) { 4244 break; 4245 } 4246 } 4247 if (delayable[i] != 0) { 4248 tmp = nvlist_prev_nvpair(props, nvp); 4249 VERIFY(nvlist_add_nvpair(delayprops, nvp) == 0); 4250 VERIFY(nvlist_remove_nvpair(props, nvp) == 0); 4251 nvp = tmp; 4252 } 4253 } 4254 4255 if (nvlist_empty(delayprops)) { 4256 nvlist_free(delayprops); 4257 delayprops = NULL; 4258 } 4259 return (delayprops); 4260} 4261 4262#ifdef DEBUG 4263static boolean_t zfs_ioc_recv_inject_err; 4264#endif 4265 4266/* 4267 * inputs: 4268 * zc_name name of containing filesystem 4269 * zc_nvlist_src{_size} nvlist of properties to apply 4270 * zc_value name of snapshot to create 4271 * zc_string name of clone origin (if DRR_FLAG_CLONE) 4272 * zc_cookie file descriptor to recv from 4273 * zc_begin_record the BEGIN record of the stream (not byteswapped) 4274 * zc_guid force flag 4275 * zc_cleanup_fd cleanup-on-exit file descriptor 4276 * zc_action_handle handle for this guid/ds mapping (or zero on first call) 4277 * zc_resumable if data is incomplete assume sender will resume 4278 * 4279 * outputs: 4280 * zc_cookie number of bytes read 4281 * zc_nvlist_dst{_size} error for each unapplied received property 4282 * zc_obj zprop_errflags_t 4283 * zc_action_handle handle for this guid/ds mapping 4284 */ 4285static int 4286zfs_ioc_recv(zfs_cmd_t *zc) 4287{ 4288 file_t *fp; 4289 dmu_recv_cookie_t drc; 4290 boolean_t force = (boolean_t)zc->zc_guid; 4291 int fd; 4292 int error = 0; 4293 int props_error = 0; 4294 nvlist_t *errors; 4295 offset_t off; 4296 nvlist_t *props = NULL; /* sent properties */ 4297 nvlist_t *origprops = NULL; /* existing properties */ 4298 nvlist_t *delayprops = NULL; /* sent properties applied post-receive */ 4299 char *origin = NULL; 4300 char *tosnap; 4301 char tofs[ZFS_MAX_DATASET_NAME_LEN]; 4302 cap_rights_t rights; 4303 boolean_t first_recvd_props = B_FALSE; 4304 4305 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 4306 strchr(zc->zc_value, '@') == NULL || 4307 strchr(zc->zc_value, '%')) 4308 return (SET_ERROR(EINVAL)); 4309 4310 (void) strcpy(tofs, zc->zc_value); 4311 tosnap = strchr(tofs, '@'); 4312 *tosnap++ = '\0'; 4313 4314 if (zc->zc_nvlist_src != 0 && 4315 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 4316 zc->zc_iflags, &props)) != 0) 4317 return (error); 4318 4319 fd = zc->zc_cookie; 4320#ifdef illumos 4321 fp = getf(fd); 4322#else 4323 fget_read(curthread, fd, cap_rights_init(&rights, CAP_PREAD), &fp); 4324#endif 4325 if (fp == NULL) { 4326 nvlist_free(props); 4327 return (SET_ERROR(EBADF)); 4328 } 4329 4330 errors = fnvlist_alloc(); 4331 4332 if (zc->zc_string[0]) 4333 origin = zc->zc_string; 4334 4335 error = dmu_recv_begin(tofs, tosnap, 4336 &zc->zc_begin_record, force, zc->zc_resumable, origin, &drc); 4337 if (error != 0) 4338 goto out; 4339 4340 /* 4341 * Set properties before we receive the stream so that they are applied 4342 * to the new data. Note that we must call dmu_recv_stream() if 4343 * dmu_recv_begin() succeeds. 4344 */ 4345 if (props != NULL && !drc.drc_newfs) { 4346 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >= 4347 SPA_VERSION_RECVD_PROPS && 4348 !dsl_prop_get_hasrecvd(tofs)) 4349 first_recvd_props = B_TRUE; 4350 4351 /* 4352 * If new received properties are supplied, they are to 4353 * completely replace the existing received properties, so stash 4354 * away the existing ones. 4355 */ 4356 if (dsl_prop_get_received(tofs, &origprops) == 0) { 4357 nvlist_t *errlist = NULL; 4358 /* 4359 * Don't bother writing a property if its value won't 4360 * change (and avoid the unnecessary security checks). 4361 * 4362 * The first receive after SPA_VERSION_RECVD_PROPS is a 4363 * special case where we blow away all local properties 4364 * regardless. 4365 */ 4366 if (!first_recvd_props) 4367 props_reduce(props, origprops); 4368 if (zfs_check_clearable(tofs, origprops, &errlist) != 0) 4369 (void) nvlist_merge(errors, errlist, 0); 4370 nvlist_free(errlist); 4371 4372 if (clear_received_props(tofs, origprops, 4373 first_recvd_props ? NULL : props) != 0) 4374 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 4375 } else { 4376 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 4377 } 4378 } 4379 4380 if (props != NULL) { 4381 props_error = dsl_prop_set_hasrecvd(tofs); 4382 4383 if (props_error == 0) { 4384 delayprops = extract_delay_props(props); 4385 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED, 4386 props, errors); 4387 } 4388 } 4389 4390 off = fp->f_offset; 4391 error = dmu_recv_stream(&drc, fp, &off, zc->zc_cleanup_fd, 4392 &zc->zc_action_handle); 4393 4394 if (error == 0) { 4395 zfsvfs_t *zfsvfs = NULL; 4396 4397 if (getzfsvfs(tofs, &zfsvfs) == 0) { 4398 /* online recv */ 4399 int end_err; 4400 4401 error = zfs_suspend_fs(zfsvfs); 4402 /* 4403 * If the suspend fails, then the recv_end will 4404 * likely also fail, and clean up after itself. 4405 */ 4406 end_err = dmu_recv_end(&drc, zfsvfs); 4407 if (error == 0) 4408 error = zfs_resume_fs(zfsvfs, tofs); 4409 error = error ? error : end_err; 4410#ifdef illumos 4411 VFS_RELE(zfsvfs->z_vfs); 4412#else 4413 vfs_unbusy(zfsvfs->z_vfs); 4414#endif 4415 } else { 4416 error = dmu_recv_end(&drc, NULL); 4417 } 4418 4419 /* Set delayed properties now, after we're done receiving. */ 4420 if (delayprops != NULL && error == 0) { 4421 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED, 4422 delayprops, errors); 4423 } 4424 } 4425 4426 if (delayprops != NULL) { 4427 /* 4428 * Merge delayed props back in with initial props, in case 4429 * we're DEBUG and zfs_ioc_recv_inject_err is set (which means 4430 * we have to make sure clear_received_props() includes 4431 * the delayed properties). 4432 * 4433 * Since zfs_ioc_recv_inject_err is only in DEBUG kernels, 4434 * using ASSERT() will be just like a VERIFY. 4435 */ 4436 ASSERT(nvlist_merge(props, delayprops, 0) == 0); 4437 nvlist_free(delayprops); 4438 } 4439 4440 /* 4441 * Now that all props, initial and delayed, are set, report the prop 4442 * errors to the caller. 4443 */ 4444 if (zc->zc_nvlist_dst_size != 0 && 4445 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 || 4446 put_nvlist(zc, errors) != 0)) { 4447 /* 4448 * Caller made zc->zc_nvlist_dst less than the minimum expected 4449 * size or supplied an invalid address. 4450 */ 4451 props_error = SET_ERROR(EINVAL); 4452 } 4453 4454 zc->zc_cookie = off - fp->f_offset; 4455 if (off >= 0 && off <= MAXOFFSET_T) 4456 fp->f_offset = off; 4457 4458#ifdef DEBUG 4459 if (zfs_ioc_recv_inject_err) { 4460 zfs_ioc_recv_inject_err = B_FALSE; 4461 error = 1; 4462 } 4463#endif 4464 4465#ifdef __FreeBSD__ 4466 if (error == 0) 4467 zvol_create_minors(tofs); 4468#endif 4469 4470 /* 4471 * On error, restore the original props. 4472 */ 4473 if (error != 0 && props != NULL && !drc.drc_newfs) { 4474 if (clear_received_props(tofs, props, NULL) != 0) { 4475 /* 4476 * We failed to clear the received properties. 4477 * Since we may have left a $recvd value on the 4478 * system, we can't clear the $hasrecvd flag. 4479 */ 4480 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4481 } else if (first_recvd_props) { 4482 dsl_prop_unset_hasrecvd(tofs); 4483 } 4484 4485 if (origprops == NULL && !drc.drc_newfs) { 4486 /* We failed to stash the original properties. */ 4487 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4488 } 4489 4490 /* 4491 * dsl_props_set() will not convert RECEIVED to LOCAL on or 4492 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL 4493 * explictly if we're restoring local properties cleared in the 4494 * first new-style receive. 4495 */ 4496 if (origprops != NULL && 4497 zfs_set_prop_nvlist(tofs, (first_recvd_props ? 4498 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED), 4499 origprops, NULL) != 0) { 4500 /* 4501 * We stashed the original properties but failed to 4502 * restore them. 4503 */ 4504 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4505 } 4506 } 4507out: 4508 nvlist_free(props); 4509 nvlist_free(origprops); 4510 nvlist_free(errors); 4511 releasef(fd); 4512 4513 if (error == 0) 4514 error = props_error; 4515 4516 return (error); 4517} 4518 4519/* 4520 * inputs: 4521 * zc_name name of snapshot to send 4522 * zc_cookie file descriptor to send stream to 4523 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj) 4524 * zc_sendobj objsetid of snapshot to send 4525 * zc_fromobj objsetid of incremental fromsnap (may be zero) 4526 * zc_guid if set, estimate size of stream only. zc_cookie is ignored. 4527 * output size in zc_objset_type. 4528 * zc_flags lzc_send_flags 4529 * 4530 * outputs: 4531 * zc_objset_type estimated size, if zc_guid is set 4532 */ 4533static int 4534zfs_ioc_send(zfs_cmd_t *zc) 4535{ 4536 int error; 4537 offset_t off; 4538 boolean_t estimate = (zc->zc_guid != 0); 4539 boolean_t embedok = (zc->zc_flags & 0x1); 4540 boolean_t large_block_ok = (zc->zc_flags & 0x2); 4541 4542 if (zc->zc_obj != 0) { 4543 dsl_pool_t *dp; 4544 dsl_dataset_t *tosnap; 4545 4546 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4547 if (error != 0) 4548 return (error); 4549 4550 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap); 4551 if (error != 0) { 4552 dsl_pool_rele(dp, FTAG); 4553 return (error); 4554 } 4555 4556 if (dsl_dir_is_clone(tosnap->ds_dir)) 4557 zc->zc_fromobj = 4558 dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj; 4559 dsl_dataset_rele(tosnap, FTAG); 4560 dsl_pool_rele(dp, FTAG); 4561 } 4562 4563 if (estimate) { 4564 dsl_pool_t *dp; 4565 dsl_dataset_t *tosnap; 4566 dsl_dataset_t *fromsnap = NULL; 4567 4568 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4569 if (error != 0) 4570 return (error); 4571 4572 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap); 4573 if (error != 0) { 4574 dsl_pool_rele(dp, FTAG); 4575 return (error); 4576 } 4577 4578 if (zc->zc_fromobj != 0) { 4579 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj, 4580 FTAG, &fromsnap); 4581 if (error != 0) { 4582 dsl_dataset_rele(tosnap, FTAG); 4583 dsl_pool_rele(dp, FTAG); 4584 return (error); 4585 } 4586 } 4587 4588 error = dmu_send_estimate(tosnap, fromsnap, 4589 &zc->zc_objset_type); 4590 4591 if (fromsnap != NULL) 4592 dsl_dataset_rele(fromsnap, FTAG); 4593 dsl_dataset_rele(tosnap, FTAG); 4594 dsl_pool_rele(dp, FTAG); 4595 } else { 4596 file_t *fp; 4597 cap_rights_t rights; 4598 4599#ifdef illumos 4600 fp = getf(zc->zc_cookie); 4601#else 4602 fget_write(curthread, zc->zc_cookie, 4603 cap_rights_init(&rights, CAP_WRITE), &fp); 4604#endif 4605 if (fp == NULL) 4606 return (SET_ERROR(EBADF)); 4607 4608 off = fp->f_offset; 4609 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj, 4610 zc->zc_fromobj, embedok, large_block_ok, 4611#ifdef illumos 4612 zc->zc_cookie, fp->f_vnode, &off); 4613#else 4614 zc->zc_cookie, fp, &off); 4615#endif 4616 4617 if (off >= 0 && off <= MAXOFFSET_T) 4618 fp->f_offset = off; 4619 releasef(zc->zc_cookie); 4620 } 4621 return (error); 4622} 4623 4624/* 4625 * inputs: 4626 * zc_name name of snapshot on which to report progress 4627 * zc_cookie file descriptor of send stream 4628 * 4629 * outputs: 4630 * zc_cookie number of bytes written in send stream thus far 4631 */ 4632static int 4633zfs_ioc_send_progress(zfs_cmd_t *zc) 4634{ 4635 dsl_pool_t *dp; 4636 dsl_dataset_t *ds; 4637 dmu_sendarg_t *dsp = NULL; 4638 int error; 4639 4640 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4641 if (error != 0) 4642 return (error); 4643 4644 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds); 4645 if (error != 0) { 4646 dsl_pool_rele(dp, FTAG); 4647 return (error); 4648 } 4649 4650 mutex_enter(&ds->ds_sendstream_lock); 4651 4652 /* 4653 * Iterate over all the send streams currently active on this dataset. 4654 * If there's one which matches the specified file descriptor _and_ the 4655 * stream was started by the current process, return the progress of 4656 * that stream. 4657 */ 4658 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL; 4659 dsp = list_next(&ds->ds_sendstreams, dsp)) { 4660 if (dsp->dsa_outfd == zc->zc_cookie && 4661 dsp->dsa_proc == curproc) 4662 break; 4663 } 4664 4665 if (dsp != NULL) 4666 zc->zc_cookie = *(dsp->dsa_off); 4667 else 4668 error = SET_ERROR(ENOENT); 4669 4670 mutex_exit(&ds->ds_sendstream_lock); 4671 dsl_dataset_rele(ds, FTAG); 4672 dsl_pool_rele(dp, FTAG); 4673 return (error); 4674} 4675 4676static int 4677zfs_ioc_inject_fault(zfs_cmd_t *zc) 4678{ 4679 int id, error; 4680 4681 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id, 4682 &zc->zc_inject_record); 4683 4684 if (error == 0) 4685 zc->zc_guid = (uint64_t)id; 4686 4687 return (error); 4688} 4689 4690static int 4691zfs_ioc_clear_fault(zfs_cmd_t *zc) 4692{ 4693 return (zio_clear_fault((int)zc->zc_guid)); 4694} 4695 4696static int 4697zfs_ioc_inject_list_next(zfs_cmd_t *zc) 4698{ 4699 int id = (int)zc->zc_guid; 4700 int error; 4701 4702 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name), 4703 &zc->zc_inject_record); 4704 4705 zc->zc_guid = id; 4706 4707 return (error); 4708} 4709 4710static int 4711zfs_ioc_error_log(zfs_cmd_t *zc) 4712{ 4713 spa_t *spa; 4714 int error; 4715 size_t count = (size_t)zc->zc_nvlist_dst_size; 4716 4717 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 4718 return (error); 4719 4720 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst, 4721 &count); 4722 if (error == 0) 4723 zc->zc_nvlist_dst_size = count; 4724 else 4725 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa); 4726 4727 spa_close(spa, FTAG); 4728 4729 return (error); 4730} 4731 4732static int 4733zfs_ioc_clear(zfs_cmd_t *zc) 4734{ 4735 spa_t *spa; 4736 vdev_t *vd; 4737 int error; 4738 4739 /* 4740 * On zpool clear we also fix up missing slogs 4741 */ 4742 mutex_enter(&spa_namespace_lock); 4743 spa = spa_lookup(zc->zc_name); 4744 if (spa == NULL) { 4745 mutex_exit(&spa_namespace_lock); 4746 return (SET_ERROR(EIO)); 4747 } 4748 if (spa_get_log_state(spa) == SPA_LOG_MISSING) { 4749 /* we need to let spa_open/spa_load clear the chains */ 4750 spa_set_log_state(spa, SPA_LOG_CLEAR); 4751 } 4752 spa->spa_last_open_failed = 0; 4753 mutex_exit(&spa_namespace_lock); 4754 4755 if (zc->zc_cookie & ZPOOL_NO_REWIND) { 4756 error = spa_open(zc->zc_name, &spa, FTAG); 4757 } else { 4758 nvlist_t *policy; 4759 nvlist_t *config = NULL; 4760 4761 if (zc->zc_nvlist_src == 0) 4762 return (SET_ERROR(EINVAL)); 4763 4764 if ((error = get_nvlist(zc->zc_nvlist_src, 4765 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) { 4766 error = spa_open_rewind(zc->zc_name, &spa, FTAG, 4767 policy, &config); 4768 if (config != NULL) { 4769 int err; 4770 4771 if ((err = put_nvlist(zc, config)) != 0) 4772 error = err; 4773 nvlist_free(config); 4774 } 4775 nvlist_free(policy); 4776 } 4777 } 4778 4779 if (error != 0) 4780 return (error); 4781 4782 spa_vdev_state_enter(spa, SCL_NONE); 4783 4784 if (zc->zc_guid == 0) { 4785 vd = NULL; 4786 } else { 4787 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE); 4788 if (vd == NULL) { 4789 (void) spa_vdev_state_exit(spa, NULL, ENODEV); 4790 spa_close(spa, FTAG); 4791 return (SET_ERROR(ENODEV)); 4792 } 4793 } 4794 4795 vdev_clear(spa, vd); 4796 4797 (void) spa_vdev_state_exit(spa, NULL, 0); 4798 4799 /* 4800 * Resume any suspended I/Os. 4801 */ 4802 if (zio_resume(spa) != 0) 4803 error = SET_ERROR(EIO); 4804 4805 spa_close(spa, FTAG); 4806 4807 return (error); 4808} 4809 4810static int 4811zfs_ioc_pool_reopen(zfs_cmd_t *zc) 4812{ 4813 spa_t *spa; 4814 int error; 4815 4816 error = spa_open(zc->zc_name, &spa, FTAG); 4817 if (error != 0) 4818 return (error); 4819 4820 spa_vdev_state_enter(spa, SCL_NONE); 4821 4822 /* 4823 * If a resilver is already in progress then set the 4824 * spa_scrub_reopen flag to B_TRUE so that we don't restart 4825 * the scan as a side effect of the reopen. Otherwise, let 4826 * vdev_open() decided if a resilver is required. 4827 */ 4828 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool); 4829 vdev_reopen(spa->spa_root_vdev); 4830 spa->spa_scrub_reopen = B_FALSE; 4831 4832 (void) spa_vdev_state_exit(spa, NULL, 0); 4833 spa_close(spa, FTAG); 4834 return (0); 4835} 4836/* 4837 * inputs: 4838 * zc_name name of filesystem 4839 * zc_value name of origin snapshot 4840 * 4841 * outputs: 4842 * zc_string name of conflicting snapshot, if there is one 4843 */ 4844static int 4845zfs_ioc_promote(zfs_cmd_t *zc) 4846{ 4847 char *cp; 4848 4849 /* 4850 * We don't need to unmount *all* the origin fs's snapshots, but 4851 * it's easier. 4852 */ 4853 cp = strchr(zc->zc_value, '@'); 4854 if (cp) 4855 *cp = '\0'; 4856 (void) dmu_objset_find(zc->zc_value, 4857 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS); 4858 return (dsl_dataset_promote(zc->zc_name, zc->zc_string)); 4859} 4860 4861/* 4862 * Retrieve a single {user|group}{used|quota}@... property. 4863 * 4864 * inputs: 4865 * zc_name name of filesystem 4866 * zc_objset_type zfs_userquota_prop_t 4867 * zc_value domain name (eg. "S-1-234-567-89") 4868 * zc_guid RID/UID/GID 4869 * 4870 * outputs: 4871 * zc_cookie property value 4872 */ 4873static int 4874zfs_ioc_userspace_one(zfs_cmd_t *zc) 4875{ 4876 zfsvfs_t *zfsvfs; 4877 int error; 4878 4879 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 4880 return (SET_ERROR(EINVAL)); 4881 4882 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE); 4883 if (error != 0) 4884 return (error); 4885 4886 error = zfs_userspace_one(zfsvfs, 4887 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie); 4888 zfsvfs_rele(zfsvfs, FTAG); 4889 4890 return (error); 4891} 4892 4893/* 4894 * inputs: 4895 * zc_name name of filesystem 4896 * zc_cookie zap cursor 4897 * zc_objset_type zfs_userquota_prop_t 4898 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist) 4899 * 4900 * outputs: 4901 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t) 4902 * zc_cookie zap cursor 4903 */ 4904static int 4905zfs_ioc_userspace_many(zfs_cmd_t *zc) 4906{ 4907 zfsvfs_t *zfsvfs; 4908 int bufsize = zc->zc_nvlist_dst_size; 4909 4910 if (bufsize <= 0) 4911 return (SET_ERROR(ENOMEM)); 4912 4913 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE); 4914 if (error != 0) 4915 return (error); 4916 4917 void *buf = kmem_alloc(bufsize, KM_SLEEP); 4918 4919 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie, 4920 buf, &zc->zc_nvlist_dst_size); 4921 4922 if (error == 0) { 4923 error = ddi_copyout(buf, 4924 (void *)(uintptr_t)zc->zc_nvlist_dst, 4925 zc->zc_nvlist_dst_size, zc->zc_iflags); 4926 } 4927 kmem_free(buf, bufsize); 4928 zfsvfs_rele(zfsvfs, FTAG); 4929 4930 return (error); 4931} 4932 4933/* 4934 * inputs: 4935 * zc_name name of filesystem 4936 * 4937 * outputs: 4938 * none 4939 */ 4940static int 4941zfs_ioc_userspace_upgrade(zfs_cmd_t *zc) 4942{ 4943 objset_t *os; 4944 int error = 0; 4945 zfsvfs_t *zfsvfs; 4946 4947 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) { 4948 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) { 4949 /* 4950 * If userused is not enabled, it may be because the 4951 * objset needs to be closed & reopened (to grow the 4952 * objset_phys_t). Suspend/resume the fs will do that. 4953 */ 4954 error = zfs_suspend_fs(zfsvfs); 4955 if (error == 0) { 4956 dmu_objset_refresh_ownership(zfsvfs->z_os, 4957 zfsvfs); 4958 error = zfs_resume_fs(zfsvfs, zc->zc_name); 4959 } 4960 } 4961 if (error == 0) 4962 error = dmu_objset_userspace_upgrade(zfsvfs->z_os); 4963#ifdef illumos 4964 VFS_RELE(zfsvfs->z_vfs); 4965#else 4966 vfs_unbusy(zfsvfs->z_vfs); 4967#endif 4968 } else { 4969 /* XXX kind of reading contents without owning */ 4970 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 4971 if (error != 0) 4972 return (error); 4973 4974 error = dmu_objset_userspace_upgrade(os); 4975 dmu_objset_rele(os, FTAG); 4976 } 4977 4978 return (error); 4979} 4980 4981#ifdef illumos 4982/* 4983 * We don't want to have a hard dependency 4984 * against some special symbols in sharefs 4985 * nfs, and smbsrv. Determine them if needed when 4986 * the first file system is shared. 4987 * Neither sharefs, nfs or smbsrv are unloadable modules. 4988 */ 4989int (*znfsexport_fs)(void *arg); 4990int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t); 4991int (*zsmbexport_fs)(void *arg, boolean_t add_share); 4992 4993int zfs_nfsshare_inited; 4994int zfs_smbshare_inited; 4995 4996ddi_modhandle_t nfs_mod; 4997ddi_modhandle_t sharefs_mod; 4998ddi_modhandle_t smbsrv_mod; 4999#endif /* illumos */ 5000kmutex_t zfs_share_lock; 5001 5002#ifdef illumos 5003static int 5004zfs_init_sharefs() 5005{ 5006 int error; 5007 5008 ASSERT(MUTEX_HELD(&zfs_share_lock)); 5009 /* Both NFS and SMB shares also require sharetab support. */ 5010 if (sharefs_mod == NULL && ((sharefs_mod = 5011 ddi_modopen("fs/sharefs", 5012 KRTLD_MODE_FIRST, &error)) == NULL)) { 5013 return (SET_ERROR(ENOSYS)); 5014 } 5015 if (zshare_fs == NULL && ((zshare_fs = 5016 (int (*)(enum sharefs_sys_op, share_t *, uint32_t)) 5017 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) { 5018 return (SET_ERROR(ENOSYS)); 5019 } 5020 return (0); 5021} 5022#endif /* illumos */ 5023 5024static int 5025zfs_ioc_share(zfs_cmd_t *zc) 5026{ 5027#ifdef illumos 5028 int error; 5029 int opcode; 5030 5031 switch (zc->zc_share.z_sharetype) { 5032 case ZFS_SHARE_NFS: 5033 case ZFS_UNSHARE_NFS: 5034 if (zfs_nfsshare_inited == 0) { 5035 mutex_enter(&zfs_share_lock); 5036 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs", 5037 KRTLD_MODE_FIRST, &error)) == NULL)) { 5038 mutex_exit(&zfs_share_lock); 5039 return (SET_ERROR(ENOSYS)); 5040 } 5041 if (znfsexport_fs == NULL && 5042 ((znfsexport_fs = (int (*)(void *)) 5043 ddi_modsym(nfs_mod, 5044 "nfs_export", &error)) == NULL)) { 5045 mutex_exit(&zfs_share_lock); 5046 return (SET_ERROR(ENOSYS)); 5047 } 5048 error = zfs_init_sharefs(); 5049 if (error != 0) { 5050 mutex_exit(&zfs_share_lock); 5051 return (SET_ERROR(ENOSYS)); 5052 } 5053 zfs_nfsshare_inited = 1; 5054 mutex_exit(&zfs_share_lock); 5055 } 5056 break; 5057 case ZFS_SHARE_SMB: 5058 case ZFS_UNSHARE_SMB: 5059 if (zfs_smbshare_inited == 0) { 5060 mutex_enter(&zfs_share_lock); 5061 if (smbsrv_mod == NULL && ((smbsrv_mod = 5062 ddi_modopen("drv/smbsrv", 5063 KRTLD_MODE_FIRST, &error)) == NULL)) { 5064 mutex_exit(&zfs_share_lock); 5065 return (SET_ERROR(ENOSYS)); 5066 } 5067 if (zsmbexport_fs == NULL && ((zsmbexport_fs = 5068 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod, 5069 "smb_server_share", &error)) == NULL)) { 5070 mutex_exit(&zfs_share_lock); 5071 return (SET_ERROR(ENOSYS)); 5072 } 5073 error = zfs_init_sharefs(); 5074 if (error != 0) { 5075 mutex_exit(&zfs_share_lock); 5076 return (SET_ERROR(ENOSYS)); 5077 } 5078 zfs_smbshare_inited = 1; 5079 mutex_exit(&zfs_share_lock); 5080 } 5081 break; 5082 default: 5083 return (SET_ERROR(EINVAL)); 5084 } 5085 5086 switch (zc->zc_share.z_sharetype) { 5087 case ZFS_SHARE_NFS: 5088 case ZFS_UNSHARE_NFS: 5089 if (error = 5090 znfsexport_fs((void *) 5091 (uintptr_t)zc->zc_share.z_exportdata)) 5092 return (error); 5093 break; 5094 case ZFS_SHARE_SMB: 5095 case ZFS_UNSHARE_SMB: 5096 if (error = zsmbexport_fs((void *) 5097 (uintptr_t)zc->zc_share.z_exportdata, 5098 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ? 5099 B_TRUE: B_FALSE)) { 5100 return (error); 5101 } 5102 break; 5103 } 5104 5105 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS || 5106 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ? 5107 SHAREFS_ADD : SHAREFS_REMOVE; 5108 5109 /* 5110 * Add or remove share from sharetab 5111 */ 5112 error = zshare_fs(opcode, 5113 (void *)(uintptr_t)zc->zc_share.z_sharedata, 5114 zc->zc_share.z_sharemax); 5115 5116 return (error); 5117 5118#else /* !illumos */ 5119 return (ENOSYS); 5120#endif /* illumos */ 5121} 5122 5123ace_t full_access[] = { 5124 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0} 5125}; 5126 5127/* 5128 * inputs: 5129 * zc_name name of containing filesystem 5130 * zc_obj object # beyond which we want next in-use object # 5131 * 5132 * outputs: 5133 * zc_obj next in-use object # 5134 */ 5135static int 5136zfs_ioc_next_obj(zfs_cmd_t *zc) 5137{ 5138 objset_t *os = NULL; 5139 int error; 5140 5141 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 5142 if (error != 0) 5143 return (error); 5144 5145 error = dmu_object_next(os, &zc->zc_obj, B_FALSE, 5146 dsl_dataset_phys(os->os_dsl_dataset)->ds_prev_snap_txg); 5147 5148 dmu_objset_rele(os, FTAG); 5149 return (error); 5150} 5151 5152/* 5153 * inputs: 5154 * zc_name name of filesystem 5155 * zc_value prefix name for snapshot 5156 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process 5157 * 5158 * outputs: 5159 * zc_value short name of new snapshot 5160 */ 5161static int 5162zfs_ioc_tmp_snapshot(zfs_cmd_t *zc) 5163{ 5164 char *snap_name; 5165 char *hold_name; 5166 int error; 5167 minor_t minor; 5168 5169 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor); 5170 if (error != 0) 5171 return (error); 5172 5173 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value, 5174 (u_longlong_t)ddi_get_lbolt64()); 5175 hold_name = kmem_asprintf("%%%s", zc->zc_value); 5176 5177 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor, 5178 hold_name); 5179 if (error == 0) 5180 (void) strcpy(zc->zc_value, snap_name); 5181 strfree(snap_name); 5182 strfree(hold_name); 5183 zfs_onexit_fd_rele(zc->zc_cleanup_fd); 5184 return (error); 5185} 5186 5187/* 5188 * inputs: 5189 * zc_name name of "to" snapshot 5190 * zc_value name of "from" snapshot 5191 * zc_cookie file descriptor to write diff data on 5192 * 5193 * outputs: 5194 * dmu_diff_record_t's to the file descriptor 5195 */ 5196static int 5197zfs_ioc_diff(zfs_cmd_t *zc) 5198{ 5199 file_t *fp; 5200 cap_rights_t rights; 5201 offset_t off; 5202 int error; 5203 5204#ifdef illumos 5205 fp = getf(zc->zc_cookie); 5206#else 5207 fget_write(curthread, zc->zc_cookie, 5208 cap_rights_init(&rights, CAP_WRITE), &fp); 5209#endif 5210 if (fp == NULL) 5211 return (SET_ERROR(EBADF)); 5212 5213 off = fp->f_offset; 5214 5215#ifdef illumos 5216 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off); 5217#else 5218 error = dmu_diff(zc->zc_name, zc->zc_value, fp, &off); 5219#endif 5220 5221 if (off >= 0 && off <= MAXOFFSET_T) 5222 fp->f_offset = off; 5223 releasef(zc->zc_cookie); 5224 5225 return (error); 5226} 5227 5228#ifdef illumos 5229/* 5230 * Remove all ACL files in shares dir 5231 */ 5232static int 5233zfs_smb_acl_purge(znode_t *dzp) 5234{ 5235 zap_cursor_t zc; 5236 zap_attribute_t zap; 5237 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 5238 int error; 5239 5240 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id); 5241 (error = zap_cursor_retrieve(&zc, &zap)) == 0; 5242 zap_cursor_advance(&zc)) { 5243 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred, 5244 NULL, 0)) != 0) 5245 break; 5246 } 5247 zap_cursor_fini(&zc); 5248 return (error); 5249} 5250#endif /* illumos */ 5251 5252static int 5253zfs_ioc_smb_acl(zfs_cmd_t *zc) 5254{ 5255#ifdef illumos 5256 vnode_t *vp; 5257 znode_t *dzp; 5258 vnode_t *resourcevp = NULL; 5259 znode_t *sharedir; 5260 zfsvfs_t *zfsvfs; 5261 nvlist_t *nvlist; 5262 char *src, *target; 5263 vattr_t vattr; 5264 vsecattr_t vsec; 5265 int error = 0; 5266 5267 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, 5268 NO_FOLLOW, NULL, &vp)) != 0) 5269 return (error); 5270 5271 /* Now make sure mntpnt and dataset are ZFS */ 5272 5273 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 || 5274 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), 5275 zc->zc_name) != 0)) { 5276 VN_RELE(vp); 5277 return (SET_ERROR(EINVAL)); 5278 } 5279 5280 dzp = VTOZ(vp); 5281 zfsvfs = dzp->z_zfsvfs; 5282 ZFS_ENTER(zfsvfs); 5283 5284 /* 5285 * Create share dir if its missing. 5286 */ 5287 mutex_enter(&zfsvfs->z_lock); 5288 if (zfsvfs->z_shares_dir == 0) { 5289 dmu_tx_t *tx; 5290 5291 tx = dmu_tx_create(zfsvfs->z_os); 5292 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE, 5293 ZFS_SHARES_DIR); 5294 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); 5295 error = dmu_tx_assign(tx, TXG_WAIT); 5296 if (error != 0) { 5297 dmu_tx_abort(tx); 5298 } else { 5299 error = zfs_create_share_dir(zfsvfs, tx); 5300 dmu_tx_commit(tx); 5301 } 5302 if (error != 0) { 5303 mutex_exit(&zfsvfs->z_lock); 5304 VN_RELE(vp); 5305 ZFS_EXIT(zfsvfs); 5306 return (error); 5307 } 5308 } 5309 mutex_exit(&zfsvfs->z_lock); 5310 5311 ASSERT(zfsvfs->z_shares_dir); 5312 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) { 5313 VN_RELE(vp); 5314 ZFS_EXIT(zfsvfs); 5315 return (error); 5316 } 5317 5318 switch (zc->zc_cookie) { 5319 case ZFS_SMB_ACL_ADD: 5320 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; 5321 vattr.va_type = VREG; 5322 vattr.va_mode = S_IFREG|0777; 5323 vattr.va_uid = 0; 5324 vattr.va_gid = 0; 5325 5326 vsec.vsa_mask = VSA_ACE; 5327 vsec.vsa_aclentp = &full_access; 5328 vsec.vsa_aclentsz = sizeof (full_access); 5329 vsec.vsa_aclcnt = 1; 5330 5331 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string, 5332 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec); 5333 if (resourcevp) 5334 VN_RELE(resourcevp); 5335 break; 5336 5337 case ZFS_SMB_ACL_REMOVE: 5338 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred, 5339 NULL, 0); 5340 break; 5341 5342 case ZFS_SMB_ACL_RENAME: 5343 if ((error = get_nvlist(zc->zc_nvlist_src, 5344 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) { 5345 VN_RELE(vp); 5346 VN_RELE(ZTOV(sharedir)); 5347 ZFS_EXIT(zfsvfs); 5348 return (error); 5349 } 5350 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) || 5351 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET, 5352 &target)) { 5353 VN_RELE(vp); 5354 VN_RELE(ZTOV(sharedir)); 5355 ZFS_EXIT(zfsvfs); 5356 nvlist_free(nvlist); 5357 return (error); 5358 } 5359 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target, 5360 kcred, NULL, 0); 5361 nvlist_free(nvlist); 5362 break; 5363 5364 case ZFS_SMB_ACL_PURGE: 5365 error = zfs_smb_acl_purge(sharedir); 5366 break; 5367 5368 default: 5369 error = SET_ERROR(EINVAL); 5370 break; 5371 } 5372 5373 VN_RELE(vp); 5374 VN_RELE(ZTOV(sharedir)); 5375 5376 ZFS_EXIT(zfsvfs); 5377 5378 return (error); 5379#else /* !illumos */ 5380 return (EOPNOTSUPP); 5381#endif /* illumos */ 5382} 5383 5384/* 5385 * innvl: { 5386 * "holds" -> { snapname -> holdname (string), ... } 5387 * (optional) "cleanup_fd" -> fd (int32) 5388 * } 5389 * 5390 * outnvl: { 5391 * snapname -> error value (int32) 5392 * ... 5393 * } 5394 */ 5395/* ARGSUSED */ 5396static int 5397zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist) 5398{ 5399 nvpair_t *pair; 5400 nvlist_t *holds; 5401 int cleanup_fd = -1; 5402 int error; 5403 minor_t minor = 0; 5404 5405 error = nvlist_lookup_nvlist(args, "holds", &holds); 5406 if (error != 0) 5407 return (SET_ERROR(EINVAL)); 5408 5409 /* make sure the user didn't pass us any invalid (empty) tags */ 5410 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL; 5411 pair = nvlist_next_nvpair(holds, pair)) { 5412 char *htag; 5413 5414 error = nvpair_value_string(pair, &htag); 5415 if (error != 0) 5416 return (SET_ERROR(error)); 5417 5418 if (strlen(htag) == 0) 5419 return (SET_ERROR(EINVAL)); 5420 } 5421 5422 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) { 5423 error = zfs_onexit_fd_hold(cleanup_fd, &minor); 5424 if (error != 0) 5425 return (error); 5426 } 5427 5428 error = dsl_dataset_user_hold(holds, minor, errlist); 5429 if (minor != 0) 5430 zfs_onexit_fd_rele(cleanup_fd); 5431 return (error); 5432} 5433 5434/* 5435 * innvl is not used. 5436 * 5437 * outnvl: { 5438 * holdname -> time added (uint64 seconds since epoch) 5439 * ... 5440 * } 5441 */ 5442/* ARGSUSED */ 5443static int 5444zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl) 5445{ 5446 return (dsl_dataset_get_holds(snapname, outnvl)); 5447} 5448 5449/* 5450 * innvl: { 5451 * snapname -> { holdname, ... } 5452 * ... 5453 * } 5454 * 5455 * outnvl: { 5456 * snapname -> error value (int32) 5457 * ... 5458 * } 5459 */ 5460/* ARGSUSED */ 5461static int 5462zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist) 5463{ 5464 return (dsl_dataset_user_release(holds, errlist)); 5465} 5466 5467/* 5468 * inputs: 5469 * zc_name name of new filesystem or snapshot 5470 * zc_value full name of old snapshot 5471 * 5472 * outputs: 5473 * zc_cookie space in bytes 5474 * zc_objset_type compressed space in bytes 5475 * zc_perm_action uncompressed space in bytes 5476 */ 5477static int 5478zfs_ioc_space_written(zfs_cmd_t *zc) 5479{ 5480 int error; 5481 dsl_pool_t *dp; 5482 dsl_dataset_t *new, *old; 5483 5484 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 5485 if (error != 0) 5486 return (error); 5487 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new); 5488 if (error != 0) { 5489 dsl_pool_rele(dp, FTAG); 5490 return (error); 5491 } 5492 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old); 5493 if (error != 0) { 5494 dsl_dataset_rele(new, FTAG); 5495 dsl_pool_rele(dp, FTAG); 5496 return (error); 5497 } 5498 5499 error = dsl_dataset_space_written(old, new, &zc->zc_cookie, 5500 &zc->zc_objset_type, &zc->zc_perm_action); 5501 dsl_dataset_rele(old, FTAG); 5502 dsl_dataset_rele(new, FTAG); 5503 dsl_pool_rele(dp, FTAG); 5504 return (error); 5505} 5506 5507/* 5508 * innvl: { 5509 * "firstsnap" -> snapshot name 5510 * } 5511 * 5512 * outnvl: { 5513 * "used" -> space in bytes 5514 * "compressed" -> compressed space in bytes 5515 * "uncompressed" -> uncompressed space in bytes 5516 * } 5517 */ 5518static int 5519zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl) 5520{ 5521 int error; 5522 dsl_pool_t *dp; 5523 dsl_dataset_t *new, *old; 5524 char *firstsnap; 5525 uint64_t used, comp, uncomp; 5526 5527 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0) 5528 return (SET_ERROR(EINVAL)); 5529 5530 error = dsl_pool_hold(lastsnap, FTAG, &dp); 5531 if (error != 0) 5532 return (error); 5533 5534 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new); 5535 if (error == 0 && !new->ds_is_snapshot) { 5536 dsl_dataset_rele(new, FTAG); 5537 error = SET_ERROR(EINVAL); 5538 } 5539 if (error != 0) { 5540 dsl_pool_rele(dp, FTAG); 5541 return (error); 5542 } 5543 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old); 5544 if (error == 0 && !old->ds_is_snapshot) { 5545 dsl_dataset_rele(old, FTAG); 5546 error = SET_ERROR(EINVAL); 5547 } 5548 if (error != 0) { 5549 dsl_dataset_rele(new, FTAG); 5550 dsl_pool_rele(dp, FTAG); 5551 return (error); 5552 } 5553 5554 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp); 5555 dsl_dataset_rele(old, FTAG); 5556 dsl_dataset_rele(new, FTAG); 5557 dsl_pool_rele(dp, FTAG); 5558 fnvlist_add_uint64(outnvl, "used", used); 5559 fnvlist_add_uint64(outnvl, "compressed", comp); 5560 fnvlist_add_uint64(outnvl, "uncompressed", uncomp); 5561 return (error); 5562} 5563 5564static int 5565zfs_ioc_jail(zfs_cmd_t *zc) 5566{ 5567 5568 return (zone_dataset_attach(curthread->td_ucred, zc->zc_name, 5569 (int)zc->zc_jailid)); 5570} 5571 5572static int 5573zfs_ioc_unjail(zfs_cmd_t *zc) 5574{ 5575 5576 return (zone_dataset_detach(curthread->td_ucred, zc->zc_name, 5577 (int)zc->zc_jailid)); 5578} 5579 5580/* 5581 * innvl: { 5582 * "fd" -> file descriptor to write stream to (int32) 5583 * (optional) "fromsnap" -> full snap name to send an incremental from 5584 * (optional) "largeblockok" -> (value ignored) 5585 * indicates that blocks > 128KB are permitted 5586 * (optional) "embedok" -> (value ignored) 5587 * presence indicates DRR_WRITE_EMBEDDED records are permitted 5588 * (optional) "resume_object" and "resume_offset" -> (uint64) 5589 * if present, resume send stream from specified object and offset. 5590 * } 5591 * 5592 * outnvl is unused 5593 */ 5594/* ARGSUSED */ 5595static int 5596zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl) 5597{ 5598 cap_rights_t rights; 5599 file_t *fp; 5600 int error; 5601 offset_t off; 5602 char *fromname = NULL; 5603 int fd; 5604 boolean_t largeblockok; 5605 boolean_t embedok; 5606 uint64_t resumeobj = 0; 5607 uint64_t resumeoff = 0; 5608 5609 error = nvlist_lookup_int32(innvl, "fd", &fd); 5610 if (error != 0) 5611 return (SET_ERROR(EINVAL)); 5612 5613 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname); 5614 5615 largeblockok = nvlist_exists(innvl, "largeblockok"); 5616 embedok = nvlist_exists(innvl, "embedok"); 5617 5618 (void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj); 5619 (void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff); 5620 5621#ifdef illumos 5622 file_t *fp = getf(fd); 5623#else 5624 fget_write(curthread, fd, cap_rights_init(&rights, CAP_WRITE), &fp); 5625#endif 5626 if (fp == NULL) 5627 return (SET_ERROR(EBADF)); 5628 5629 off = fp->f_offset; 5630 error = dmu_send(snapname, fromname, embedok, largeblockok, fd, 5631#ifdef illumos 5632 resumeobj, resumeoff, fp->f_vnode, &off); 5633#else 5634 resumeobj, resumeoff, fp, &off); 5635#endif 5636 5637#ifdef illumos 5638 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0) 5639 fp->f_offset = off; 5640#else 5641 fp->f_offset = off; 5642#endif 5643 5644 releasef(fd); 5645 return (error); 5646} 5647 5648/* 5649 * Determine approximately how large a zfs send stream will be -- the number 5650 * of bytes that will be written to the fd supplied to zfs_ioc_send_new(). 5651 * 5652 * innvl: { 5653 * (optional) "from" -> full snap or bookmark name to send an incremental 5654 * from 5655 * } 5656 * 5657 * outnvl: { 5658 * "space" -> bytes of space (uint64) 5659 * } 5660 */ 5661static int 5662zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl) 5663{ 5664 dsl_pool_t *dp; 5665 dsl_dataset_t *tosnap; 5666 int error; 5667 char *fromname; 5668 uint64_t space; 5669 5670 error = dsl_pool_hold(snapname, FTAG, &dp); 5671 if (error != 0) 5672 return (error); 5673 5674 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap); 5675 if (error != 0) { 5676 dsl_pool_rele(dp, FTAG); 5677 return (error); 5678 } 5679 5680 error = nvlist_lookup_string(innvl, "from", &fromname); 5681 if (error == 0) { 5682 if (strchr(fromname, '@') != NULL) { 5683 /* 5684 * If from is a snapshot, hold it and use the more 5685 * efficient dmu_send_estimate to estimate send space 5686 * size using deadlists. 5687 */ 5688 dsl_dataset_t *fromsnap; 5689 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap); 5690 if (error != 0) 5691 goto out; 5692 error = dmu_send_estimate(tosnap, fromsnap, &space); 5693 dsl_dataset_rele(fromsnap, FTAG); 5694 } else if (strchr(fromname, '#') != NULL) { 5695 /* 5696 * If from is a bookmark, fetch the creation TXG of the 5697 * snapshot it was created from and use that to find 5698 * blocks that were born after it. 5699 */ 5700 zfs_bookmark_phys_t frombm; 5701 5702 error = dsl_bookmark_lookup(dp, fromname, tosnap, 5703 &frombm); 5704 if (error != 0) 5705 goto out; 5706 error = dmu_send_estimate_from_txg(tosnap, 5707 frombm.zbm_creation_txg, &space); 5708 } else { 5709 /* 5710 * from is not properly formatted as a snapshot or 5711 * bookmark 5712 */ 5713 error = SET_ERROR(EINVAL); 5714 goto out; 5715 } 5716 } else { 5717 // If estimating the size of a full send, use dmu_send_estimate 5718 error = dmu_send_estimate(tosnap, NULL, &space); 5719 } 5720 5721 fnvlist_add_uint64(outnvl, "space", space); 5722 5723out: 5724 dsl_dataset_rele(tosnap, FTAG); 5725 dsl_pool_rele(dp, FTAG); 5726 return (error); 5727} 5728 5729static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST]; 5730 5731static void 5732zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5733 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck, 5734 boolean_t log_history, zfs_ioc_poolcheck_t pool_check) 5735{ 5736 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST]; 5737 5738 ASSERT3U(ioc, >=, ZFS_IOC_FIRST); 5739 ASSERT3U(ioc, <, ZFS_IOC_LAST); 5740 ASSERT3P(vec->zvec_legacy_func, ==, NULL); 5741 ASSERT3P(vec->zvec_func, ==, NULL); 5742 5743 vec->zvec_legacy_func = func; 5744 vec->zvec_secpolicy = secpolicy; 5745 vec->zvec_namecheck = namecheck; 5746 vec->zvec_allow_log = log_history; 5747 vec->zvec_pool_check = pool_check; 5748} 5749 5750/* 5751 * See the block comment at the beginning of this file for details on 5752 * each argument to this function. 5753 */ 5754static void 5755zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func, 5756 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck, 5757 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist, 5758 boolean_t allow_log) 5759{ 5760 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST]; 5761 5762 ASSERT3U(ioc, >=, ZFS_IOC_FIRST); 5763 ASSERT3U(ioc, <, ZFS_IOC_LAST); 5764 ASSERT3P(vec->zvec_legacy_func, ==, NULL); 5765 ASSERT3P(vec->zvec_func, ==, NULL); 5766 5767 /* if we are logging, the name must be valid */ 5768 ASSERT(!allow_log || namecheck != NO_NAME); 5769 5770 vec->zvec_name = name; 5771 vec->zvec_func = func; 5772 vec->zvec_secpolicy = secpolicy; 5773 vec->zvec_namecheck = namecheck; 5774 vec->zvec_pool_check = pool_check; 5775 vec->zvec_smush_outnvlist = smush_outnvlist; 5776 vec->zvec_allow_log = allow_log; 5777} 5778 5779static void 5780zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5781 zfs_secpolicy_func_t *secpolicy, boolean_t log_history, 5782 zfs_ioc_poolcheck_t pool_check) 5783{ 5784 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5785 POOL_NAME, log_history, pool_check); 5786} 5787 5788static void 5789zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5790 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check) 5791{ 5792 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5793 DATASET_NAME, B_FALSE, pool_check); 5794} 5795 5796static void 5797zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func) 5798{ 5799 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config, 5800 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5801} 5802 5803static void 5804zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5805 zfs_secpolicy_func_t *secpolicy) 5806{ 5807 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5808 NO_NAME, B_FALSE, POOL_CHECK_NONE); 5809} 5810 5811static void 5812zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc, 5813 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy) 5814{ 5815 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5816 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED); 5817} 5818 5819static void 5820zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func) 5821{ 5822 zfs_ioctl_register_dataset_read_secpolicy(ioc, func, 5823 zfs_secpolicy_read); 5824} 5825 5826static void 5827zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5828 zfs_secpolicy_func_t *secpolicy) 5829{ 5830 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5831 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5832} 5833 5834static void 5835zfs_ioctl_init(void) 5836{ 5837 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT, 5838 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME, 5839 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5840 5841 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY, 5842 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME, 5843 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE); 5844 5845 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS, 5846 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME, 5847 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5848 5849 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW, 5850 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME, 5851 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5852 5853 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE, 5854 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME, 5855 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5856 5857 zfs_ioctl_register("create", ZFS_IOC_CREATE, 5858 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME, 5859 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5860 5861 zfs_ioctl_register("clone", ZFS_IOC_CLONE, 5862 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME, 5863 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5864 5865 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS, 5866 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME, 5867 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5868 5869 zfs_ioctl_register("hold", ZFS_IOC_HOLD, 5870 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME, 5871 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5872 zfs_ioctl_register("release", ZFS_IOC_RELEASE, 5873 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME, 5874 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5875 5876 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS, 5877 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME, 5878 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5879 5880 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK, 5881 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, 5882 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE); 5883 5884 zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK, 5885 zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME, 5886 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5887 5888 zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS, 5889 zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME, 5890 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5891 5892 zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS, 5893 zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks, 5894 POOL_NAME, 5895 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5896 5897 /* IOCTLS that use the legacy function signature */ 5898 5899 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze, 5900 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY); 5901 5902 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create, 5903 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5904 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN, 5905 zfs_ioc_pool_scan); 5906 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE, 5907 zfs_ioc_pool_upgrade); 5908 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD, 5909 zfs_ioc_vdev_add); 5910 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE, 5911 zfs_ioc_vdev_remove); 5912 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE, 5913 zfs_ioc_vdev_set_state); 5914 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH, 5915 zfs_ioc_vdev_attach); 5916 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH, 5917 zfs_ioc_vdev_detach); 5918 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH, 5919 zfs_ioc_vdev_setpath); 5920 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU, 5921 zfs_ioc_vdev_setfru); 5922 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS, 5923 zfs_ioc_pool_set_props); 5924 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT, 5925 zfs_ioc_vdev_split); 5926 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID, 5927 zfs_ioc_pool_reguid); 5928 5929 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS, 5930 zfs_ioc_pool_configs, zfs_secpolicy_none); 5931 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT, 5932 zfs_ioc_pool_tryimport, zfs_secpolicy_config); 5933 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT, 5934 zfs_ioc_inject_fault, zfs_secpolicy_inject); 5935 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT, 5936 zfs_ioc_clear_fault, zfs_secpolicy_inject); 5937 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT, 5938 zfs_ioc_inject_list_next, zfs_secpolicy_inject); 5939 5940 /* 5941 * pool destroy, and export don't log the history as part of 5942 * zfsdev_ioctl, but rather zfs_ioc_pool_export 5943 * does the logging of those commands. 5944 */ 5945 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy, 5946 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE); 5947 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export, 5948 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE); 5949 5950 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats, 5951 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE); 5952 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props, 5953 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE); 5954 5955 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log, 5956 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_NONE); 5957 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME, 5958 zfs_ioc_dsobj_to_dsname, 5959 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_NONE); 5960 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY, 5961 zfs_ioc_pool_get_history, 5962 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED); 5963 5964 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import, 5965 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5966 5967 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear, 5968 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5969 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen, 5970 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED); 5971 5972 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN, 5973 zfs_ioc_space_written); 5974 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS, 5975 zfs_ioc_objset_recvd_props); 5976 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ, 5977 zfs_ioc_next_obj); 5978 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL, 5979 zfs_ioc_get_fsacl); 5980 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS, 5981 zfs_ioc_objset_stats); 5982 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS, 5983 zfs_ioc_objset_zplprops); 5984 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT, 5985 zfs_ioc_dataset_list_next); 5986 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT, 5987 zfs_ioc_snapshot_list_next); 5988 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS, 5989 zfs_ioc_send_progress); 5990 5991 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF, 5992 zfs_ioc_diff, zfs_secpolicy_diff); 5993 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS, 5994 zfs_ioc_obj_to_stats, zfs_secpolicy_diff); 5995 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH, 5996 zfs_ioc_obj_to_path, zfs_secpolicy_diff); 5997 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE, 5998 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one); 5999 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY, 6000 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many); 6001 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND, 6002 zfs_ioc_send, zfs_secpolicy_send); 6003 6004 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop, 6005 zfs_secpolicy_none); 6006 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy, 6007 zfs_secpolicy_destroy); 6008 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename, 6009 zfs_secpolicy_rename); 6010 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv, 6011 zfs_secpolicy_recv); 6012 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote, 6013 zfs_secpolicy_promote); 6014 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP, 6015 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop); 6016 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl, 6017 zfs_secpolicy_set_fsacl); 6018 6019 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share, 6020 zfs_secpolicy_share, POOL_CHECK_NONE); 6021 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl, 6022 zfs_secpolicy_smb_acl, POOL_CHECK_NONE); 6023 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE, 6024 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade, 6025 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 6026 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT, 6027 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot, 6028 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 6029 6030#ifdef __FreeBSD__ 6031 zfs_ioctl_register_dataset_nolog(ZFS_IOC_JAIL, zfs_ioc_jail, 6032 zfs_secpolicy_config, POOL_CHECK_NONE); 6033 zfs_ioctl_register_dataset_nolog(ZFS_IOC_UNJAIL, zfs_ioc_unjail, 6034 zfs_secpolicy_config, POOL_CHECK_NONE); 6035#endif 6036} 6037 6038int 6039pool_status_check(const char *name, zfs_ioc_namecheck_t type, 6040 zfs_ioc_poolcheck_t check) 6041{ 6042 spa_t *spa; 6043 int error; 6044 6045 ASSERT(type == POOL_NAME || type == DATASET_NAME); 6046 6047 if (check & POOL_CHECK_NONE) 6048 return (0); 6049 6050 error = spa_open(name, &spa, FTAG); 6051 if (error == 0) { 6052 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa)) 6053 error = SET_ERROR(EAGAIN); 6054 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa)) 6055 error = SET_ERROR(EROFS); 6056 spa_close(spa, FTAG); 6057 } 6058 return (error); 6059} 6060 6061/* 6062 * Find a free minor number. 6063 */ 6064minor_t 6065zfsdev_minor_alloc(void) 6066{ 6067 static minor_t last_minor; 6068 minor_t m; 6069 6070 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 6071 6072 for (m = last_minor + 1; m != last_minor; m++) { 6073 if (m > ZFSDEV_MAX_MINOR) 6074 m = 1; 6075 if (ddi_get_soft_state(zfsdev_state, m) == NULL) { 6076 last_minor = m; 6077 return (m); 6078 } 6079 } 6080 6081 return (0); 6082} 6083 6084static int 6085zfs_ctldev_init(struct cdev *devp) 6086{ 6087 minor_t minor; 6088 zfs_soft_state_t *zs; 6089 6090 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 6091 6092 minor = zfsdev_minor_alloc(); 6093 if (minor == 0) 6094 return (SET_ERROR(ENXIO)); 6095 6096 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS) 6097 return (SET_ERROR(EAGAIN)); 6098 6099 devfs_set_cdevpriv((void *)(uintptr_t)minor, zfsdev_close); 6100 6101 zs = ddi_get_soft_state(zfsdev_state, minor); 6102 zs->zss_type = ZSST_CTLDEV; 6103 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data); 6104 6105 return (0); 6106} 6107 6108static void 6109zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor) 6110{ 6111 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 6112 6113 zfs_onexit_destroy(zo); 6114 ddi_soft_state_free(zfsdev_state, minor); 6115} 6116 6117void * 6118zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which) 6119{ 6120 zfs_soft_state_t *zp; 6121 6122 zp = ddi_get_soft_state(zfsdev_state, minor); 6123 if (zp == NULL || zp->zss_type != which) 6124 return (NULL); 6125 6126 return (zp->zss_data); 6127} 6128 6129static int 6130zfsdev_open(struct cdev *devp, int flag, int mode, struct thread *td) 6131{ 6132 int error = 0; 6133 6134#ifdef illumos 6135 if (getminor(*devp) != 0) 6136 return (zvol_open(devp, flag, otyp, cr)); 6137#endif 6138 6139 /* This is the control device. Allocate a new minor if requested. */ 6140 if (flag & FEXCL) { 6141 mutex_enter(&spa_namespace_lock); 6142 error = zfs_ctldev_init(devp); 6143 mutex_exit(&spa_namespace_lock); 6144 } 6145 6146 return (error); 6147} 6148 6149static void 6150zfsdev_close(void *data) 6151{ 6152 zfs_onexit_t *zo; 6153 minor_t minor = (minor_t)(uintptr_t)data; 6154 6155 if (minor == 0) 6156 return; 6157 6158 mutex_enter(&spa_namespace_lock); 6159 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV); 6160 if (zo == NULL) { 6161 mutex_exit(&spa_namespace_lock); 6162 return; 6163 } 6164 zfs_ctldev_destroy(zo, minor); 6165 mutex_exit(&spa_namespace_lock); 6166} 6167 6168static int 6169zfsdev_ioctl(struct cdev *dev, u_long zcmd, caddr_t arg, int flag, 6170 struct thread *td) 6171{ 6172 zfs_cmd_t *zc; 6173 uint_t vecnum; 6174 int error, rc, len; 6175#ifdef illumos 6176 minor_t minor = getminor(dev); 6177#else 6178 zfs_iocparm_t *zc_iocparm; 6179 int cflag, cmd, oldvecnum; 6180 boolean_t newioc, compat; 6181 void *compat_zc = NULL; 6182 cred_t *cr = td->td_ucred; 6183#endif 6184 const zfs_ioc_vec_t *vec; 6185 char *saved_poolname = NULL; 6186 nvlist_t *innvl = NULL; 6187 6188 cflag = ZFS_CMD_COMPAT_NONE; 6189 compat = B_FALSE; 6190 newioc = B_TRUE; /* "new" style (zfs_iocparm_t) ioctl */ 6191 6192 len = IOCPARM_LEN(zcmd); 6193 vecnum = cmd = zcmd & 0xff; 6194 6195 /* 6196 * Check if we are talking to supported older binaries 6197 * and translate zfs_cmd if necessary 6198 */ 6199 if (len != sizeof(zfs_iocparm_t)) { 6200 newioc = B_FALSE; 6201 compat = B_TRUE; 6202 6203 vecnum = cmd; 6204 6205 switch (len) { 6206 case sizeof(zfs_cmd_zcmd_t): 6207 cflag = ZFS_CMD_COMPAT_LZC; 6208 break; 6209 case sizeof(zfs_cmd_deadman_t): 6210 cflag = ZFS_CMD_COMPAT_DEADMAN; 6211 break; 6212 case sizeof(zfs_cmd_v28_t): 6213 cflag = ZFS_CMD_COMPAT_V28; 6214 break; 6215 case sizeof(zfs_cmd_v15_t): 6216 cflag = ZFS_CMD_COMPAT_V15; 6217 vecnum = zfs_ioctl_v15_to_v28[cmd]; 6218 6219 /* 6220 * Return without further handling 6221 * if the command is blacklisted. 6222 */ 6223 if (vecnum == ZFS_IOC_COMPAT_PASS) 6224 return (0); 6225 else if (vecnum == ZFS_IOC_COMPAT_FAIL) 6226 return (ENOTSUP); 6227 break; 6228 default: 6229 return (EINVAL); 6230 } 6231 } 6232 6233#ifdef illumos 6234 vecnum = cmd - ZFS_IOC_FIRST; 6235 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip)); 6236#endif 6237 6238 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0])) 6239 return (SET_ERROR(EINVAL)); 6240 vec = &zfs_ioc_vec[vecnum]; 6241 6242 zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP); 6243 6244#ifdef illumos 6245 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag); 6246 if (error != 0) { 6247 error = SET_ERROR(EFAULT); 6248 goto out; 6249 } 6250#else /* !illumos */ 6251 bzero(zc, sizeof(zfs_cmd_t)); 6252 6253 if (newioc) { 6254 zc_iocparm = (void *)arg; 6255 6256 switch (zc_iocparm->zfs_ioctl_version) { 6257 case ZFS_IOCVER_CURRENT: 6258 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_t)) { 6259 error = SET_ERROR(EINVAL); 6260 goto out; 6261 } 6262 break; 6263 case ZFS_IOCVER_INLANES: 6264 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_inlanes_t)) { 6265 error = SET_ERROR(EFAULT); 6266 goto out; 6267 } 6268 compat = B_TRUE; 6269 cflag = ZFS_CMD_COMPAT_INLANES; 6270 break; 6271 case ZFS_IOCVER_RESUME: 6272 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_resume_t)) { 6273 error = SET_ERROR(EFAULT); 6274 goto out; 6275 } 6276 compat = B_TRUE; 6277 cflag = ZFS_CMD_COMPAT_RESUME; 6278 break; 6279 case ZFS_IOCVER_EDBP: 6280 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_edbp_t)) { 6281 error = SET_ERROR(EFAULT); 6282 goto out; 6283 } 6284 compat = B_TRUE; 6285 cflag = ZFS_CMD_COMPAT_EDBP; 6286 break; 6287 case ZFS_IOCVER_ZCMD: 6288 if (zc_iocparm->zfs_cmd_size > sizeof(zfs_cmd_t) || 6289 zc_iocparm->zfs_cmd_size < sizeof(zfs_cmd_zcmd_t)) { 6290 error = SET_ERROR(EFAULT); 6291 goto out; 6292 } 6293 compat = B_TRUE; 6294 cflag = ZFS_CMD_COMPAT_ZCMD; 6295 break; 6296 default: 6297 error = SET_ERROR(EINVAL); 6298 goto out; 6299 /* NOTREACHED */ 6300 } 6301 6302 if (compat) { 6303 ASSERT(sizeof(zfs_cmd_t) >= zc_iocparm->zfs_cmd_size); 6304 compat_zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP); 6305 bzero(compat_zc, sizeof(zfs_cmd_t)); 6306 6307 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd, 6308 compat_zc, zc_iocparm->zfs_cmd_size, flag); 6309 if (error != 0) { 6310 error = SET_ERROR(EFAULT); 6311 goto out; 6312 } 6313 } else { 6314 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd, 6315 zc, zc_iocparm->zfs_cmd_size, flag); 6316 if (error != 0) { 6317 error = SET_ERROR(EFAULT); 6318 goto out; 6319 } 6320 } 6321 } 6322 6323 if (compat) { 6324 if (newioc) { 6325 ASSERT(compat_zc != NULL); 6326 zfs_cmd_compat_get(zc, compat_zc, cflag); 6327 } else { 6328 ASSERT(compat_zc == NULL); 6329 zfs_cmd_compat_get(zc, arg, cflag); 6330 } 6331 oldvecnum = vecnum; 6332 error = zfs_ioctl_compat_pre(zc, &vecnum, cflag); 6333 if (error != 0) 6334 goto out; 6335 if (oldvecnum != vecnum) 6336 vec = &zfs_ioc_vec[vecnum]; 6337 } 6338#endif /* !illumos */ 6339 6340 zc->zc_iflags = flag & FKIOCTL; 6341 if (zc->zc_nvlist_src_size != 0) { 6342 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 6343 zc->zc_iflags, &innvl); 6344 if (error != 0) 6345 goto out; 6346 } 6347 6348 /* rewrite innvl for backwards compatibility */ 6349 if (compat) 6350 innvl = zfs_ioctl_compat_innvl(zc, innvl, vecnum, cflag); 6351 6352 /* 6353 * Ensure that all pool/dataset names are valid before we pass down to 6354 * the lower layers. 6355 */ 6356 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0'; 6357 switch (vec->zvec_namecheck) { 6358 case POOL_NAME: 6359 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0) 6360 error = SET_ERROR(EINVAL); 6361 else 6362 error = pool_status_check(zc->zc_name, 6363 vec->zvec_namecheck, vec->zvec_pool_check); 6364 break; 6365 6366 case DATASET_NAME: 6367 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0) 6368 error = SET_ERROR(EINVAL); 6369 else 6370 error = pool_status_check(zc->zc_name, 6371 vec->zvec_namecheck, vec->zvec_pool_check); 6372 break; 6373 6374 case NO_NAME: 6375 break; 6376 } 6377 6378 if (error == 0) 6379 error = vec->zvec_secpolicy(zc, innvl, cr); 6380 6381 if (error != 0) 6382 goto out; 6383 6384 /* legacy ioctls can modify zc_name */ 6385 len = strcspn(zc->zc_name, "/@#") + 1; 6386 saved_poolname = kmem_alloc(len, KM_SLEEP); 6387 (void) strlcpy(saved_poolname, zc->zc_name, len); 6388 6389 if (vec->zvec_func != NULL) { 6390 nvlist_t *outnvl; 6391 int puterror = 0; 6392 spa_t *spa; 6393 nvlist_t *lognv = NULL; 6394 6395 ASSERT(vec->zvec_legacy_func == NULL); 6396 6397 /* 6398 * Add the innvl to the lognv before calling the func, 6399 * in case the func changes the innvl. 6400 */ 6401 if (vec->zvec_allow_log) { 6402 lognv = fnvlist_alloc(); 6403 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL, 6404 vec->zvec_name); 6405 if (!nvlist_empty(innvl)) { 6406 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL, 6407 innvl); 6408 } 6409 } 6410 6411 outnvl = fnvlist_alloc(); 6412 error = vec->zvec_func(zc->zc_name, innvl, outnvl); 6413 6414 if (error == 0 && vec->zvec_allow_log && 6415 spa_open(zc->zc_name, &spa, FTAG) == 0) { 6416 if (!nvlist_empty(outnvl)) { 6417 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL, 6418 outnvl); 6419 } 6420 (void) spa_history_log_nvl(spa, lognv); 6421 spa_close(spa, FTAG); 6422 } 6423 fnvlist_free(lognv); 6424 6425 /* rewrite outnvl for backwards compatibility */ 6426 if (compat) 6427 outnvl = zfs_ioctl_compat_outnvl(zc, outnvl, vecnum, 6428 cflag); 6429 6430 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) { 6431 int smusherror = 0; 6432 if (vec->zvec_smush_outnvlist) { 6433 smusherror = nvlist_smush(outnvl, 6434 zc->zc_nvlist_dst_size); 6435 } 6436 if (smusherror == 0) 6437 puterror = put_nvlist(zc, outnvl); 6438 } 6439 6440 if (puterror != 0) 6441 error = puterror; 6442 6443 nvlist_free(outnvl); 6444 } else { 6445 error = vec->zvec_legacy_func(zc); 6446 } 6447 6448out: 6449 nvlist_free(innvl); 6450 6451#ifdef illumos 6452 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag); 6453 if (error == 0 && rc != 0) 6454 error = SET_ERROR(EFAULT); 6455#else 6456 if (compat) { 6457 zfs_ioctl_compat_post(zc, cmd, cflag); 6458 if (newioc) { 6459 ASSERT(compat_zc != NULL); 6460 ASSERT(sizeof(zfs_cmd_t) >= zc_iocparm->zfs_cmd_size); 6461 6462 zfs_cmd_compat_put(zc, compat_zc, vecnum, cflag); 6463 rc = ddi_copyout(compat_zc, 6464 (void *)(uintptr_t)zc_iocparm->zfs_cmd, 6465 zc_iocparm->zfs_cmd_size, flag); 6466 if (error == 0 && rc != 0) 6467 error = SET_ERROR(EFAULT); 6468 kmem_free(compat_zc, sizeof (zfs_cmd_t)); 6469 } else { 6470 zfs_cmd_compat_put(zc, arg, vecnum, cflag); 6471 } 6472 } else { 6473 ASSERT(newioc); 6474 6475 rc = ddi_copyout(zc, (void *)(uintptr_t)zc_iocparm->zfs_cmd, 6476 sizeof (zfs_cmd_t), flag); 6477 if (error == 0 && rc != 0) 6478 error = SET_ERROR(EFAULT); 6479 } 6480#endif 6481 if (error == 0 && vec->zvec_allow_log) { 6482 char *s = tsd_get(zfs_allow_log_key); 6483 if (s != NULL) 6484 strfree(s); 6485 (void) tsd_set(zfs_allow_log_key, saved_poolname); 6486 } else { 6487 if (saved_poolname != NULL) 6488 strfree(saved_poolname); 6489 } 6490 6491 kmem_free(zc, sizeof (zfs_cmd_t)); 6492 return (error); 6493} 6494 6495#ifdef illumos 6496static int 6497zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd) 6498{ 6499 if (cmd != DDI_ATTACH) 6500 return (DDI_FAILURE); 6501 6502 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0, 6503 DDI_PSEUDO, 0) == DDI_FAILURE) 6504 return (DDI_FAILURE); 6505 6506 zfs_dip = dip; 6507 6508 ddi_report_dev(dip); 6509 6510 return (DDI_SUCCESS); 6511} 6512 6513static int 6514zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd) 6515{ 6516 if (spa_busy() || zfs_busy() || zvol_busy()) 6517 return (DDI_FAILURE); 6518 6519 if (cmd != DDI_DETACH) 6520 return (DDI_FAILURE); 6521 6522 zfs_dip = NULL; 6523 6524 ddi_prop_remove_all(dip); 6525 ddi_remove_minor_node(dip, NULL); 6526 6527 return (DDI_SUCCESS); 6528} 6529 6530/*ARGSUSED*/ 6531static int 6532zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result) 6533{ 6534 switch (infocmd) { 6535 case DDI_INFO_DEVT2DEVINFO: 6536 *result = zfs_dip; 6537 return (DDI_SUCCESS); 6538 6539 case DDI_INFO_DEVT2INSTANCE: 6540 *result = (void *)0; 6541 return (DDI_SUCCESS); 6542 } 6543 6544 return (DDI_FAILURE); 6545} 6546#endif /* illumos */ 6547 6548/* 6549 * OK, so this is a little weird. 6550 * 6551 * /dev/zfs is the control node, i.e. minor 0. 6552 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0. 6553 * 6554 * /dev/zfs has basically nothing to do except serve up ioctls, 6555 * so most of the standard driver entry points are in zvol.c. 6556 */ 6557#ifdef illumos 6558static struct cb_ops zfs_cb_ops = { 6559 zfsdev_open, /* open */ 6560 zfsdev_close, /* close */ 6561 zvol_strategy, /* strategy */ 6562 nodev, /* print */ 6563 zvol_dump, /* dump */ 6564 zvol_read, /* read */ 6565 zvol_write, /* write */ 6566 zfsdev_ioctl, /* ioctl */ 6567 nodev, /* devmap */ 6568 nodev, /* mmap */ 6569 nodev, /* segmap */ 6570 nochpoll, /* poll */ 6571 ddi_prop_op, /* prop_op */ 6572 NULL, /* streamtab */ 6573 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */ 6574 CB_REV, /* version */ 6575 nodev, /* async read */ 6576 nodev, /* async write */ 6577}; 6578 6579static struct dev_ops zfs_dev_ops = { 6580 DEVO_REV, /* version */ 6581 0, /* refcnt */ 6582 zfs_info, /* info */ 6583 nulldev, /* identify */ 6584 nulldev, /* probe */ 6585 zfs_attach, /* attach */ 6586 zfs_detach, /* detach */ 6587 nodev, /* reset */ 6588 &zfs_cb_ops, /* driver operations */ 6589 NULL, /* no bus operations */ 6590 NULL, /* power */ 6591 ddi_quiesce_not_needed, /* quiesce */ 6592}; 6593 6594static struct modldrv zfs_modldrv = { 6595 &mod_driverops, 6596 "ZFS storage pool", 6597 &zfs_dev_ops 6598}; 6599 6600static struct modlinkage modlinkage = { 6601 MODREV_1, 6602 (void *)&zfs_modlfs, 6603 (void *)&zfs_modldrv, 6604 NULL 6605}; 6606#endif /* illumos */ 6607 6608static struct cdevsw zfs_cdevsw = { 6609 .d_version = D_VERSION, 6610 .d_open = zfsdev_open, 6611 .d_ioctl = zfsdev_ioctl, 6612 .d_name = ZFS_DEV_NAME 6613}; 6614 6615static void 6616zfs_allow_log_destroy(void *arg) 6617{ 6618 char *poolname = arg; 6619 strfree(poolname); 6620} 6621 6622static void 6623zfsdev_init(void) 6624{ 6625 zfsdev = make_dev(&zfs_cdevsw, 0x0, UID_ROOT, GID_OPERATOR, 0666, 6626 ZFS_DEV_NAME); 6627} 6628 6629static void 6630zfsdev_fini(void) 6631{ 6632 if (zfsdev != NULL) 6633 destroy_dev(zfsdev); 6634} 6635 6636static struct root_hold_token *zfs_root_token; 6637struct proc *zfsproc; 6638 6639#ifdef illumos 6640int 6641_init(void) 6642{ 6643 int error; 6644 6645 spa_init(FREAD | FWRITE); 6646 zfs_init(); 6647 zvol_init(); 6648 zfs_ioctl_init(); 6649 6650 if ((error = mod_install(&modlinkage)) != 0) { 6651 zvol_fini(); 6652 zfs_fini(); 6653 spa_fini(); 6654 return (error); 6655 } 6656 6657 tsd_create(&zfs_fsyncer_key, NULL); 6658 tsd_create(&rrw_tsd_key, rrw_tsd_destroy); 6659 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy); 6660 6661 error = ldi_ident_from_mod(&modlinkage, &zfs_li); 6662 ASSERT(error == 0); 6663 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL); 6664 6665 return (0); 6666} 6667 6668int 6669_fini(void) 6670{ 6671 int error; 6672 6673 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled) 6674 return (SET_ERROR(EBUSY)); 6675 6676 if ((error = mod_remove(&modlinkage)) != 0) 6677 return (error); 6678 6679 zvol_fini(); 6680 zfs_fini(); 6681 spa_fini(); 6682 if (zfs_nfsshare_inited) 6683 (void) ddi_modclose(nfs_mod); 6684 if (zfs_smbshare_inited) 6685 (void) ddi_modclose(smbsrv_mod); 6686 if (zfs_nfsshare_inited || zfs_smbshare_inited) 6687 (void) ddi_modclose(sharefs_mod); 6688 6689 tsd_destroy(&zfs_fsyncer_key); 6690 ldi_ident_release(zfs_li); 6691 zfs_li = NULL; 6692 mutex_destroy(&zfs_share_lock); 6693 6694 return (error); 6695} 6696 6697int 6698_info(struct modinfo *modinfop) 6699{ 6700 return (mod_info(&modlinkage, modinfop)); 6701} 6702#endif /* illumos */ 6703 6704static int zfs__init(void); 6705static int zfs__fini(void); 6706static void zfs_shutdown(void *, int); 6707 6708static eventhandler_tag zfs_shutdown_event_tag; 6709 6710#ifdef __FreeBSD__ 6711#define ZFS_MIN_KSTACK_PAGES 4 6712#endif 6713 6714int 6715zfs__init(void) 6716{ 6717 6718#ifdef __FreeBSD__ 6719#if KSTACK_PAGES < ZFS_MIN_KSTACK_PAGES 6720 printf("ZFS NOTICE: KSTACK_PAGES is %d which could result in stack " 6721 "overflow panic!\nPlease consider adding " 6722 "'options KSTACK_PAGES=%d' to your kernel config\n", KSTACK_PAGES, 6723 ZFS_MIN_KSTACK_PAGES); 6724#endif 6725#endif 6726 zfs_root_token = root_mount_hold("ZFS"); 6727 6728 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL); 6729 6730 spa_init(FREAD | FWRITE); 6731 zfs_init(); 6732 zvol_init(); 6733 zfs_ioctl_init(); 6734 6735 tsd_create(&zfs_fsyncer_key, NULL); 6736 tsd_create(&rrw_tsd_key, rrw_tsd_destroy); 6737 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy); 6738 6739 printf("ZFS storage pool version: features support (" SPA_VERSION_STRING ")\n"); 6740 root_mount_rel(zfs_root_token); 6741 6742 zfsdev_init(); 6743 6744 return (0); 6745} 6746 6747int 6748zfs__fini(void) 6749{ 6750 if (spa_busy() || zfs_busy() || zvol_busy() || 6751 zio_injection_enabled) { 6752 return (EBUSY); 6753 } 6754 6755 zfsdev_fini(); 6756 zvol_fini(); 6757 zfs_fini(); 6758 spa_fini(); 6759 6760 tsd_destroy(&zfs_fsyncer_key); 6761 tsd_destroy(&rrw_tsd_key); 6762 tsd_destroy(&zfs_allow_log_key); 6763 6764 mutex_destroy(&zfs_share_lock); 6765 6766 return (0); 6767} 6768 6769static void 6770zfs_shutdown(void *arg __unused, int howto __unused) 6771{ 6772 6773 /* 6774 * ZFS fini routines can not properly work in a panic-ed system. 6775 */ 6776 if (panicstr == NULL) 6777 (void)zfs__fini(); 6778} 6779 6780 6781static int 6782zfs_modevent(module_t mod, int type, void *unused __unused) 6783{ 6784 int err; 6785 6786 switch (type) { 6787 case MOD_LOAD: 6788 err = zfs__init(); 6789 if (err == 0) 6790 zfs_shutdown_event_tag = EVENTHANDLER_REGISTER( 6791 shutdown_post_sync, zfs_shutdown, NULL, 6792 SHUTDOWN_PRI_FIRST); 6793 return (err); 6794 case MOD_UNLOAD: 6795 err = zfs__fini(); 6796 if (err == 0 && zfs_shutdown_event_tag != NULL) 6797 EVENTHANDLER_DEREGISTER(shutdown_post_sync, 6798 zfs_shutdown_event_tag); 6799 return (err); 6800 case MOD_SHUTDOWN: 6801 return (0); 6802 default: 6803 break; 6804 } 6805 return (EOPNOTSUPP); 6806} 6807 6808static moduledata_t zfs_mod = { 6809 "zfsctrl", 6810 zfs_modevent, 6811 0 6812}; 6813DECLARE_MODULE(zfsctrl, zfs_mod, SI_SUB_VFS, SI_ORDER_ANY); 6814MODULE_VERSION(zfsctrl, 1); 6815MODULE_DEPEND(zfsctrl, opensolaris, 1, 1, 1); 6816MODULE_DEPEND(zfsctrl, krpc, 1, 1, 1); 6817MODULE_DEPEND(zfsctrl, acl_nfs4, 1, 1, 1); 6818