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  * Portions Copyright 2011 Martin Matuska
  25  * Copyright 2011 Nexenta Systems, Inc.  All rights reserved.
  26  * Copyright (c) 2012, Joyent, Inc. All rights reserved.
  27  * Copyright (c) 2013 by Delphix. All rights reserved.
  28  * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
  29  * Copyright (c) 2013 Steven Hartland. All rights reserved.
  30  */
  31 
  32 /*
  33  * ZFS ioctls.
  34  *
  35  * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
  36  * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
  37  *
  38  * There are two ways that we handle ioctls: the legacy way where almost
  39  * all of the logic is in the ioctl callback, and the new way where most
  40  * of the marshalling is handled in the common entry point, zfsdev_ioctl().
  41  *
  42  * Non-legacy ioctls should be registered by calling
  43  * zfs_ioctl_register() from zfs_ioctl_init().  The ioctl is invoked
  44  * from userland by lzc_ioctl().
  45  *
  46  * The registration arguments are as follows:
  47  *
  48  * const char *name
  49  *   The name of the ioctl.  This is used for history logging.  If the
  50  *   ioctl returns successfully (the callback returns 0), and allow_log
  51  *   is true, then a history log entry will be recorded with the input &
  52  *   output nvlists.  The log entry can be printed with "zpool history -i".
  53  *
  54  * zfs_ioc_t ioc
  55  *   The ioctl request number, which userland will pass to ioctl(2).
  56  *   The ioctl numbers can change from release to release, because
  57  *   the caller (libzfs) must be matched to the kernel.
  58  *
  59  * zfs_secpolicy_func_t *secpolicy
  60  *   This function will be called before the zfs_ioc_func_t, to
  61  *   determine if this operation is permitted.  It should return EPERM
  62  *   on failure, and 0 on success.  Checks include determining if the
  63  *   dataset is visible in this zone, and if the user has either all
  64  *   zfs privileges in the zone (SYS_MOUNT), or has been granted permission
  65  *   to do this operation on this dataset with "zfs allow".
  66  *
  67  * zfs_ioc_namecheck_t namecheck
  68  *   This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
  69  *   name, a dataset name, or nothing.  If the name is not well-formed,
  70  *   the ioctl will fail and the callback will not be called.
  71  *   Therefore, the callback can assume that the name is well-formed
  72  *   (e.g. is null-terminated, doesn't have more than one '@' character,
  73  *   doesn't have invalid characters).
  74  *
  75  * zfs_ioc_poolcheck_t pool_check
  76  *   This specifies requirements on the pool state.  If the pool does
  77  *   not meet them (is suspended or is readonly), the ioctl will fail
  78  *   and the callback will not be called.  If any checks are specified
  79  *   (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
  80  *   Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED |
  81  *   POOL_CHECK_READONLY).
  82  *
  83  * boolean_t smush_outnvlist
  84  *   If smush_outnvlist is true, then the output is presumed to be a
  85  *   list of errors, and it will be "smushed" down to fit into the
  86  *   caller's buffer, by removing some entries and replacing them with a
  87  *   single "N_MORE_ERRORS" entry indicating how many were removed.  See
  88  *   nvlist_smush() for details.  If smush_outnvlist is false, and the
  89  *   outnvlist does not fit into the userland-provided buffer, then the
  90  *   ioctl will fail with ENOMEM.
  91  *
  92  * zfs_ioc_func_t *func
  93  *   The callback function that will perform the operation.
  94  *
  95  *   The callback should return 0 on success, or an error number on
  96  *   failure.  If the function fails, the userland ioctl will return -1,
  97  *   and errno will be set to the callback's return value.  The callback
  98  *   will be called with the following arguments:
  99  *
 100  *   const char *name
 101  *     The name of the pool or dataset to operate on, from
 102  *     zfs_cmd_t:zc_name.  The 'namecheck' argument specifies the
 103  *     expected type (pool, dataset, or none).
 104  *
 105  *   nvlist_t *innvl
 106  *     The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src.  Or
 107  *     NULL if no input nvlist was provided.  Changes to this nvlist are
 108  *     ignored.  If the input nvlist could not be deserialized, the
 109  *     ioctl will fail and the callback will not be called.
 110  *
 111  *   nvlist_t *outnvl
 112  *     The output nvlist, initially empty.  The callback can fill it in,
 113  *     and it will be returned to userland by serializing it into
 114  *     zfs_cmd_t:zc_nvlist_dst.  If it is non-empty, and serialization
 115  *     fails (e.g. because the caller didn't supply a large enough
 116  *     buffer), then the overall ioctl will fail.  See the
 117  *     'smush_nvlist' argument above for additional behaviors.
 118  *
 119  *     There are two typical uses of the output nvlist:
 120  *       - To return state, e.g. property values.  In this case,
 121  *         smush_outnvlist should be false.  If the buffer was not large
 122  *         enough, the caller will reallocate a larger buffer and try
 123  *         the ioctl again.
 124  *
 125  *       - To return multiple errors from an ioctl which makes on-disk
 126  *         changes.  In this case, smush_outnvlist should be true.
 127  *         Ioctls which make on-disk modifications should generally not
 128  *         use the outnvl if they succeed, because the caller can not
 129  *         distinguish between the operation failing, and
 130  *         deserialization failing.
 131  */
 132 
 133 #include <sys/types.h>
 134 #include <sys/param.h>
 135 #include <sys/errno.h>
 136 #include <sys/uio.h>
 137 #include <sys/buf.h>
 138 #include <sys/modctl.h>
 139 #include <sys/open.h>
 140 #include <sys/file.h>
 141 #include <sys/kmem.h>
 142 #include <sys/conf.h>
 143 #include <sys/cmn_err.h>
 144 #include <sys/stat.h>
 145 #include <sys/zfs_ioctl.h>
 146 #include <sys/zfs_vfsops.h>
 147 #include <sys/zfs_znode.h>
 148 #include <sys/zap.h>
 149 #include <sys/spa.h>
 150 #include <sys/spa_impl.h>
 151 #include <sys/vdev.h>
 152 #include <sys/priv_impl.h>
 153 #include <sys/dmu.h>
 154 #include <sys/dsl_dir.h>
 155 #include <sys/dsl_dataset.h>
 156 #include <sys/dsl_prop.h>
 157 #include <sys/dsl_deleg.h>
 158 #include <sys/dmu_objset.h>
 159 #include <sys/dmu_impl.h>
 160 #include <sys/dmu_tx.h>
 161 #include <sys/ddi.h>
 162 #include <sys/sunddi.h>
 163 #include <sys/sunldi.h>
 164 #include <sys/policy.h>
 165 #include <sys/zone.h>
 166 #include <sys/nvpair.h>
 167 #include <sys/pathname.h>
 168 #include <sys/mount.h>
 169 #include <sys/sdt.h>
 170 #include <sys/fs/zfs.h>
 171 #include <sys/zfs_ctldir.h>
 172 #include <sys/zfs_dir.h>
 173 #include <sys/zfs_onexit.h>
 174 #include <sys/zvol.h>
 175 #include <sys/dsl_scan.h>
 176 #include <sharefs/share.h>
 177 #include <sys/dmu_objset.h>
 178 #include <sys/dmu_send.h>
 179 #include <sys/dsl_destroy.h>
 180 #include <sys/dsl_userhold.h>
 181 #include <sys/zfeature.h>
 182 
 183 #include "zfs_namecheck.h"
 184 #include "zfs_prop.h"
 185 #include "zfs_deleg.h"
 186 #include "zfs_comutil.h"
 187 
 188 extern struct modlfs zfs_modlfs;
 189 
 190 extern void zfs_init(void);
 191 extern void zfs_fini(void);
 192 
 193 ldi_ident_t zfs_li = NULL;
 194 dev_info_t *zfs_dip;
 195 
 196 uint_t zfs_fsyncer_key;
 197 extern uint_t rrw_tsd_key;
 198 static uint_t zfs_allow_log_key;
 199 
 200 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *);
 201 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *);
 202 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *);
 203 
 204 typedef enum {
 205         NO_NAME,
 206         POOL_NAME,
 207         DATASET_NAME
 208 } zfs_ioc_namecheck_t;
 209 
 210 typedef enum {
 211         POOL_CHECK_NONE         = 1 << 0,
 212         POOL_CHECK_SUSPENDED    = 1 << 1,
 213         POOL_CHECK_READONLY     = 1 << 2,
 214 } zfs_ioc_poolcheck_t;
 215 
 216 typedef struct zfs_ioc_vec {
 217         zfs_ioc_legacy_func_t   *zvec_legacy_func;
 218         zfs_ioc_func_t          *zvec_func;
 219         zfs_secpolicy_func_t    *zvec_secpolicy;
 220         zfs_ioc_namecheck_t     zvec_namecheck;
 221         boolean_t               zvec_allow_log;
 222         zfs_ioc_poolcheck_t     zvec_pool_check;
 223         boolean_t               zvec_smush_outnvlist;
 224         const char              *zvec_name;
 225 } zfs_ioc_vec_t;
 226 
 227 /* This array is indexed by zfs_userquota_prop_t */
 228 static const char *userquota_perms[] = {
 229         ZFS_DELEG_PERM_USERUSED,
 230         ZFS_DELEG_PERM_USERQUOTA,
 231         ZFS_DELEG_PERM_GROUPUSED,
 232         ZFS_DELEG_PERM_GROUPQUOTA,
 233 };
 234 
 235 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
 236 static int zfs_check_settable(const char *name, nvpair_t *property,
 237     cred_t *cr);
 238 static int zfs_check_clearable(char *dataset, nvlist_t *props,
 239     nvlist_t **errors);
 240 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
 241     boolean_t *);
 242 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
 243 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
 244 
 245 static int zfs_prop_activate_feature(spa_t *spa, zfeature_info_t *feature);
 246 
 247 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
 248 void
 249 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
 250 {
 251         const char *newfile;
 252         char buf[512];
 253         va_list adx;
 254 
 255         /*
 256          * Get rid of annoying "../common/" prefix to filename.
 257          */
 258         newfile = strrchr(file, '/');
 259         if (newfile != NULL) {
 260                 newfile = newfile + 1; /* Get rid of leading / */
 261         } else {
 262                 newfile = file;
 263         }
 264 
 265         va_start(adx, fmt);
 266         (void) vsnprintf(buf, sizeof (buf), fmt, adx);
 267         va_end(adx);
 268 
 269         /*
 270          * To get this data, use the zfs-dprintf probe as so:
 271          * dtrace -q -n 'zfs-dprintf \
 272          *      /stringof(arg0) == "dbuf.c"/ \
 273          *      {printf("%s: %s", stringof(arg1), stringof(arg3))}'
 274          * arg0 = file name
 275          * arg1 = function name
 276          * arg2 = line number
 277          * arg3 = message
 278          */
 279         DTRACE_PROBE4(zfs__dprintf,
 280             char *, newfile, char *, func, int, line, char *, buf);
 281 }
 282 
 283 static void
 284 history_str_free(char *buf)
 285 {
 286         kmem_free(buf, HIS_MAX_RECORD_LEN);
 287 }
 288 
 289 static char *
 290 history_str_get(zfs_cmd_t *zc)
 291 {
 292         char *buf;
 293 
 294         if (zc->zc_history == NULL)
 295                 return (NULL);
 296 
 297         buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
 298         if (copyinstr((void *)(uintptr_t)zc->zc_history,
 299             buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
 300                 history_str_free(buf);
 301                 return (NULL);
 302         }
 303 
 304         buf[HIS_MAX_RECORD_LEN -1] = '\0';
 305 
 306         return (buf);
 307 }
 308 
 309 /*
 310  * Check to see if the named dataset is currently defined as bootable
 311  */
 312 static boolean_t
 313 zfs_is_bootfs(const char *name)
 314 {
 315         objset_t *os;
 316 
 317         if (dmu_objset_hold(name, FTAG, &os) == 0) {
 318                 boolean_t ret;
 319                 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
 320                 dmu_objset_rele(os, FTAG);
 321                 return (ret);
 322         }
 323         return (B_FALSE);
 324 }
 325 
 326 /*
 327  * Return non-zero if the spa version is less than requested version.
 328  */
 329 static int
 330 zfs_earlier_version(const char *name, int version)
 331 {
 332         spa_t *spa;
 333 
 334         if (spa_open(name, &spa, FTAG) == 0) {
 335                 if (spa_version(spa) < version) {
 336                         spa_close(spa, FTAG);
 337                         return (1);
 338                 }
 339                 spa_close(spa, FTAG);
 340         }
 341         return (0);
 342 }
 343 
 344 /*
 345  * Return TRUE if the ZPL version is less than requested version.
 346  */
 347 static boolean_t
 348 zpl_earlier_version(const char *name, int version)
 349 {
 350         objset_t *os;
 351         boolean_t rc = B_TRUE;
 352 
 353         if (dmu_objset_hold(name, FTAG, &os) == 0) {
 354                 uint64_t zplversion;
 355 
 356                 if (dmu_objset_type(os) != DMU_OST_ZFS) {
 357                         dmu_objset_rele(os, FTAG);
 358                         return (B_TRUE);
 359                 }
 360                 /* XXX reading from non-owned objset */
 361                 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
 362                         rc = zplversion < version;
 363                 dmu_objset_rele(os, FTAG);
 364         }
 365         return (rc);
 366 }
 367 
 368 static void
 369 zfs_log_history(zfs_cmd_t *zc)
 370 {
 371         spa_t *spa;
 372         char *buf;
 373 
 374         if ((buf = history_str_get(zc)) == NULL)
 375                 return;
 376 
 377         if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
 378                 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
 379                         (void) spa_history_log(spa, buf);
 380                 spa_close(spa, FTAG);
 381         }
 382         history_str_free(buf);
 383 }
 384 
 385 /*
 386  * Policy for top-level read operations (list pools).  Requires no privileges,
 387  * and can be used in the local zone, as there is no associated dataset.
 388  */
 389 /* ARGSUSED */
 390 static int
 391 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 392 {
 393         return (0);
 394 }
 395 
 396 /*
 397  * Policy for dataset read operations (list children, get statistics).  Requires
 398  * no privileges, but must be visible in the local zone.
 399  */
 400 /* ARGSUSED */
 401 static int
 402 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 403 {
 404         if (INGLOBALZONE(curproc) ||
 405             zone_dataset_visible(zc->zc_name, NULL))
 406                 return (0);
 407 
 408         return (SET_ERROR(ENOENT));
 409 }
 410 
 411 static int
 412 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
 413 {
 414         int writable = 1;
 415 
 416         /*
 417          * The dataset must be visible by this zone -- check this first
 418          * so they don't see EPERM on something they shouldn't know about.
 419          */
 420         if (!INGLOBALZONE(curproc) &&
 421             !zone_dataset_visible(dataset, &writable))
 422                 return (SET_ERROR(ENOENT));
 423 
 424         if (INGLOBALZONE(curproc)) {
 425                 /*
 426                  * If the fs is zoned, only root can access it from the
 427                  * global zone.
 428                  */
 429                 if (secpolicy_zfs(cr) && zoned)
 430                         return (SET_ERROR(EPERM));
 431         } else {
 432                 /*
 433                  * If we are in a local zone, the 'zoned' property must be set.
 434                  */
 435                 if (!zoned)
 436                         return (SET_ERROR(EPERM));
 437 
 438                 /* must be writable by this zone */
 439                 if (!writable)
 440                         return (SET_ERROR(EPERM));
 441         }
 442         return (0);
 443 }
 444 
 445 static int
 446 zfs_dozonecheck(const char *dataset, cred_t *cr)
 447 {
 448         uint64_t zoned;
 449 
 450         if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
 451                 return (SET_ERROR(ENOENT));
 452 
 453         return (zfs_dozonecheck_impl(dataset, zoned, cr));
 454 }
 455 
 456 static int
 457 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
 458 {
 459         uint64_t zoned;
 460 
 461         if (dsl_prop_get_int_ds(ds, "zoned", &zoned))
 462                 return (SET_ERROR(ENOENT));
 463 
 464         return (zfs_dozonecheck_impl(dataset, zoned, cr));
 465 }
 466 
 467 static int
 468 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
 469     const char *perm, cred_t *cr)
 470 {
 471         int error;
 472 
 473         error = zfs_dozonecheck_ds(name, ds, cr);
 474         if (error == 0) {
 475                 error = secpolicy_zfs(cr);
 476                 if (error != 0)
 477                         error = dsl_deleg_access_impl(ds, perm, cr);
 478         }
 479         return (error);
 480 }
 481 
 482 static int
 483 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
 484 {
 485         int error;
 486         dsl_dataset_t *ds;
 487         dsl_pool_t *dp;
 488 
 489         error = dsl_pool_hold(name, FTAG, &dp);
 490         if (error != 0)
 491                 return (error);
 492 
 493         error = dsl_dataset_hold(dp, name, FTAG, &ds);
 494         if (error != 0) {
 495                 dsl_pool_rele(dp, FTAG);
 496                 return (error);
 497         }
 498 
 499         error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
 500 
 501         dsl_dataset_rele(ds, FTAG);
 502         dsl_pool_rele(dp, FTAG);
 503         return (error);
 504 }
 505 
 506 /*
 507  * Policy for setting the security label property.
 508  *
 509  * Returns 0 for success, non-zero for access and other errors.
 510  */
 511 static int
 512 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
 513 {
 514         char            ds_hexsl[MAXNAMELEN];
 515         bslabel_t       ds_sl, new_sl;
 516         boolean_t       new_default = FALSE;
 517         uint64_t        zoned;
 518         int             needed_priv = -1;
 519         int             error;
 520 
 521         /* First get the existing dataset label. */
 522         error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
 523             1, sizeof (ds_hexsl), &ds_hexsl, NULL);
 524         if (error != 0)
 525                 return (SET_ERROR(EPERM));
 526 
 527         if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
 528                 new_default = TRUE;
 529 
 530         /* The label must be translatable */
 531         if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
 532                 return (SET_ERROR(EINVAL));
 533 
 534         /*
 535          * In a non-global zone, disallow attempts to set a label that
 536          * doesn't match that of the zone; otherwise no other checks
 537          * are needed.
 538          */
 539         if (!INGLOBALZONE(curproc)) {
 540                 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
 541                         return (SET_ERROR(EPERM));
 542                 return (0);
 543         }
 544 
 545         /*
 546          * For global-zone datasets (i.e., those whose zoned property is
 547          * "off", verify that the specified new label is valid for the
 548          * global zone.
 549          */
 550         if (dsl_prop_get_integer(name,
 551             zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
 552                 return (SET_ERROR(EPERM));
 553         if (!zoned) {
 554                 if (zfs_check_global_label(name, strval) != 0)
 555                         return (SET_ERROR(EPERM));
 556         }
 557 
 558         /*
 559          * If the existing dataset label is nondefault, check if the
 560          * dataset is mounted (label cannot be changed while mounted).
 561          * Get the zfsvfs; if there isn't one, then the dataset isn't
 562          * mounted (or isn't a dataset, doesn't exist, ...).
 563          */
 564         if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
 565                 objset_t *os;
 566                 static char *setsl_tag = "setsl_tag";
 567 
 568                 /*
 569                  * Try to own the dataset; abort if there is any error,
 570                  * (e.g., already mounted, in use, or other error).
 571                  */
 572                 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
 573                     setsl_tag, &os);
 574                 if (error != 0)
 575                         return (SET_ERROR(EPERM));
 576 
 577                 dmu_objset_disown(os, setsl_tag);
 578 
 579                 if (new_default) {
 580                         needed_priv = PRIV_FILE_DOWNGRADE_SL;
 581                         goto out_check;
 582                 }
 583 
 584                 if (hexstr_to_label(strval, &new_sl) != 0)
 585                         return (SET_ERROR(EPERM));
 586 
 587                 if (blstrictdom(&ds_sl, &new_sl))
 588                         needed_priv = PRIV_FILE_DOWNGRADE_SL;
 589                 else if (blstrictdom(&new_sl, &ds_sl))
 590                         needed_priv = PRIV_FILE_UPGRADE_SL;
 591         } else {
 592                 /* dataset currently has a default label */
 593                 if (!new_default)
 594                         needed_priv = PRIV_FILE_UPGRADE_SL;
 595         }
 596 
 597 out_check:
 598         if (needed_priv != -1)
 599                 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
 600         return (0);
 601 }
 602 
 603 static int
 604 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
 605     cred_t *cr)
 606 {
 607         char *strval;
 608 
 609         /*
 610          * Check permissions for special properties.
 611          */
 612         switch (prop) {
 613         case ZFS_PROP_ZONED:
 614                 /*
 615                  * Disallow setting of 'zoned' from within a local zone.
 616                  */
 617                 if (!INGLOBALZONE(curproc))
 618                         return (SET_ERROR(EPERM));
 619                 break;
 620 
 621         case ZFS_PROP_QUOTA:
 622                 if (!INGLOBALZONE(curproc)) {
 623                         uint64_t zoned;
 624                         char setpoint[MAXNAMELEN];
 625                         /*
 626                          * Unprivileged users are allowed to modify the
 627                          * quota on things *under* (ie. contained by)
 628                          * the thing they own.
 629                          */
 630                         if (dsl_prop_get_integer(dsname, "zoned", &zoned,
 631                             setpoint))
 632                                 return (SET_ERROR(EPERM));
 633                         if (!zoned || strlen(dsname) <= strlen(setpoint))
 634                                 return (SET_ERROR(EPERM));
 635                 }
 636                 break;
 637 
 638         case ZFS_PROP_MLSLABEL:
 639                 if (!is_system_labeled())
 640                         return (SET_ERROR(EPERM));
 641 
 642                 if (nvpair_value_string(propval, &strval) == 0) {
 643                         int err;
 644 
 645                         err = zfs_set_slabel_policy(dsname, strval, CRED());
 646                         if (err != 0)
 647                                 return (err);
 648                 }
 649                 break;
 650         }
 651 
 652         return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
 653 }
 654 
 655 /* ARGSUSED */
 656 static int
 657 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 658 {
 659         int error;
 660 
 661         error = zfs_dozonecheck(zc->zc_name, cr);
 662         if (error != 0)
 663                 return (error);
 664 
 665         /*
 666          * permission to set permissions will be evaluated later in
 667          * dsl_deleg_can_allow()
 668          */
 669         return (0);
 670 }
 671 
 672 /* ARGSUSED */
 673 static int
 674 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 675 {
 676         return (zfs_secpolicy_write_perms(zc->zc_name,
 677             ZFS_DELEG_PERM_ROLLBACK, cr));
 678 }
 679 
 680 /* ARGSUSED */
 681 static int
 682 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 683 {
 684         dsl_pool_t *dp;
 685         dsl_dataset_t *ds;
 686         char *cp;
 687         int error;
 688 
 689         /*
 690          * Generate the current snapshot name from the given objsetid, then
 691          * use that name for the secpolicy/zone checks.
 692          */
 693         cp = strchr(zc->zc_name, '@');
 694         if (cp == NULL)
 695                 return (SET_ERROR(EINVAL));
 696         error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
 697         if (error != 0)
 698                 return (error);
 699 
 700         error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
 701         if (error != 0) {
 702                 dsl_pool_rele(dp, FTAG);
 703                 return (error);
 704         }
 705 
 706         dsl_dataset_name(ds, zc->zc_name);
 707 
 708         error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
 709             ZFS_DELEG_PERM_SEND, cr);
 710         dsl_dataset_rele(ds, FTAG);
 711         dsl_pool_rele(dp, FTAG);
 712 
 713         return (error);
 714 }
 715 
 716 /* ARGSUSED */
 717 static int
 718 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 719 {
 720         return (zfs_secpolicy_write_perms(zc->zc_name,
 721             ZFS_DELEG_PERM_SEND, cr));
 722 }
 723 
 724 /* ARGSUSED */
 725 static int
 726 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 727 {
 728         vnode_t *vp;
 729         int error;
 730 
 731         if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
 732             NO_FOLLOW, NULL, &vp)) != 0)
 733                 return (error);
 734 
 735         /* Now make sure mntpnt and dataset are ZFS */
 736 
 737         if (vp->v_vfsp->vfs_fstype != zfsfstype ||
 738             (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
 739             zc->zc_name) != 0)) {
 740                 VN_RELE(vp);
 741                 return (SET_ERROR(EPERM));
 742         }
 743 
 744         VN_RELE(vp);
 745         return (dsl_deleg_access(zc->zc_name,
 746             ZFS_DELEG_PERM_SHARE, cr));
 747 }
 748 
 749 int
 750 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 751 {
 752         if (!INGLOBALZONE(curproc))
 753                 return (SET_ERROR(EPERM));
 754 
 755         if (secpolicy_nfs(cr) == 0) {
 756                 return (0);
 757         } else {
 758                 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
 759         }
 760 }
 761 
 762 int
 763 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 764 {
 765         if (!INGLOBALZONE(curproc))
 766                 return (SET_ERROR(EPERM));
 767 
 768         if (secpolicy_smb(cr) == 0) {
 769                 return (0);
 770         } else {
 771                 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
 772         }
 773 }
 774 
 775 static int
 776 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
 777 {
 778         char *cp;
 779 
 780         /*
 781          * Remove the @bla or /bla from the end of the name to get the parent.
 782          */
 783         (void) strncpy(parent, datasetname, parentsize);
 784         cp = strrchr(parent, '@');
 785         if (cp != NULL) {
 786                 cp[0] = '\0';
 787         } else {
 788                 cp = strrchr(parent, '/');
 789                 if (cp == NULL)
 790                         return (SET_ERROR(ENOENT));
 791                 cp[0] = '\0';
 792         }
 793 
 794         return (0);
 795 }
 796 
 797 int
 798 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
 799 {
 800         int error;
 801 
 802         if ((error = zfs_secpolicy_write_perms(name,
 803             ZFS_DELEG_PERM_MOUNT, cr)) != 0)
 804                 return (error);
 805 
 806         return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
 807 }
 808 
 809 /* ARGSUSED */
 810 static int
 811 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 812 {
 813         return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
 814 }
 815 
 816 /*
 817  * Destroying snapshots with delegated permissions requires
 818  * descendant mount and destroy permissions.
 819  */
 820 /* ARGSUSED */
 821 static int
 822 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 823 {
 824         nvlist_t *snaps;
 825         nvpair_t *pair, *nextpair;
 826         int error = 0;
 827 
 828         if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
 829                 return (SET_ERROR(EINVAL));
 830         for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
 831             pair = nextpair) {
 832                 dsl_pool_t *dp;
 833                 dsl_dataset_t *ds;
 834 
 835                 error = dsl_pool_hold(nvpair_name(pair), FTAG, &dp);
 836                 if (error != 0)
 837                         break;
 838                 nextpair = nvlist_next_nvpair(snaps, pair);
 839                 error = dsl_dataset_hold(dp, nvpair_name(pair), FTAG, &ds);
 840                 if (error == 0)
 841                         dsl_dataset_rele(ds, FTAG);
 842                 dsl_pool_rele(dp, FTAG);
 843 
 844                 if (error == 0) {
 845                         error = zfs_secpolicy_destroy_perms(nvpair_name(pair),
 846                             cr);
 847                 } else if (error == ENOENT) {
 848                         /*
 849                          * Ignore any snapshots that don't exist (we consider
 850                          * them "already destroyed").  Remove the name from the
 851                          * nvl here in case the snapshot is created between
 852                          * now and when we try to destroy it (in which case
 853                          * we don't want to destroy it since we haven't
 854                          * checked for permission).
 855                          */
 856                         fnvlist_remove_nvpair(snaps, pair);
 857                         error = 0;
 858                 }
 859                 if (error != 0)
 860                         break;
 861         }
 862 
 863         return (error);
 864 }
 865 
 866 int
 867 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
 868 {
 869         char    parentname[MAXNAMELEN];
 870         int     error;
 871 
 872         if ((error = zfs_secpolicy_write_perms(from,
 873             ZFS_DELEG_PERM_RENAME, cr)) != 0)
 874                 return (error);
 875 
 876         if ((error = zfs_secpolicy_write_perms(from,
 877             ZFS_DELEG_PERM_MOUNT, cr)) != 0)
 878                 return (error);
 879 
 880         if ((error = zfs_get_parent(to, parentname,
 881             sizeof (parentname))) != 0)
 882                 return (error);
 883 
 884         if ((error = zfs_secpolicy_write_perms(parentname,
 885             ZFS_DELEG_PERM_CREATE, cr)) != 0)
 886                 return (error);
 887 
 888         if ((error = zfs_secpolicy_write_perms(parentname,
 889             ZFS_DELEG_PERM_MOUNT, cr)) != 0)
 890                 return (error);
 891 
 892         return (error);
 893 }
 894 
 895 /* ARGSUSED */
 896 static int
 897 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 898 {
 899         return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
 900 }
 901 
 902 /* ARGSUSED */
 903 static int
 904 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 905 {
 906         dsl_pool_t *dp;
 907         dsl_dataset_t *clone;
 908         int error;
 909 
 910         error = zfs_secpolicy_write_perms(zc->zc_name,
 911             ZFS_DELEG_PERM_PROMOTE, cr);
 912         if (error != 0)
 913                 return (error);
 914 
 915         error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
 916         if (error != 0)
 917                 return (error);
 918 
 919         error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
 920 
 921         if (error == 0) {
 922                 char parentname[MAXNAMELEN];
 923                 dsl_dataset_t *origin = NULL;
 924                 dsl_dir_t *dd;
 925                 dd = clone->ds_dir;
 926 
 927                 error = dsl_dataset_hold_obj(dd->dd_pool,
 928                     dd->dd_phys->dd_origin_obj, FTAG, &origin);
 929                 if (error != 0) {
 930                         dsl_dataset_rele(clone, FTAG);
 931                         dsl_pool_rele(dp, FTAG);
 932                         return (error);
 933                 }
 934 
 935                 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
 936                     ZFS_DELEG_PERM_MOUNT, cr);
 937 
 938                 dsl_dataset_name(origin, parentname);
 939                 if (error == 0) {
 940                         error = zfs_secpolicy_write_perms_ds(parentname, origin,
 941                             ZFS_DELEG_PERM_PROMOTE, cr);
 942                 }
 943                 dsl_dataset_rele(clone, FTAG);
 944                 dsl_dataset_rele(origin, FTAG);
 945         }
 946         dsl_pool_rele(dp, FTAG);
 947         return (error);
 948 }
 949 
 950 /* ARGSUSED */
 951 static int
 952 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 953 {
 954         int error;
 955 
 956         if ((error = zfs_secpolicy_write_perms(zc->zc_name,
 957             ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
 958                 return (error);
 959 
 960         if ((error = zfs_secpolicy_write_perms(zc->zc_name,
 961             ZFS_DELEG_PERM_MOUNT, cr)) != 0)
 962                 return (error);
 963 
 964         return (zfs_secpolicy_write_perms(zc->zc_name,
 965             ZFS_DELEG_PERM_CREATE, cr));
 966 }
 967 
 968 int
 969 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
 970 {
 971         return (zfs_secpolicy_write_perms(name,
 972             ZFS_DELEG_PERM_SNAPSHOT, cr));
 973 }
 974 
 975 /*
 976  * Check for permission to create each snapshot in the nvlist.
 977  */
 978 /* ARGSUSED */
 979 static int
 980 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 981 {
 982         nvlist_t *snaps;
 983         int error = 0;
 984         nvpair_t *pair;
 985 
 986         if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
 987                 return (SET_ERROR(EINVAL));
 988         for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
 989             pair = nvlist_next_nvpair(snaps, pair)) {
 990                 char *name = nvpair_name(pair);
 991                 char *atp = strchr(name, '@');
 992 
 993                 if (atp == NULL) {
 994                         error = SET_ERROR(EINVAL);
 995                         break;
 996                 }
 997                 *atp = '\0';
 998                 error = zfs_secpolicy_snapshot_perms(name, cr);
 999                 *atp = '@';
1000                 if (error != 0)
1001                         break;
1002         }
1003         return (error);
1004 }
1005 
1006 /* ARGSUSED */
1007 static int
1008 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1009 {
1010         /*
1011          * Even root must have a proper TSD so that we know what pool
1012          * to log to.
1013          */
1014         if (tsd_get(zfs_allow_log_key) == NULL)
1015                 return (SET_ERROR(EPERM));
1016         return (0);
1017 }
1018 
1019 static int
1020 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1021 {
1022         char    parentname[MAXNAMELEN];
1023         int     error;
1024         char    *origin;
1025 
1026         if ((error = zfs_get_parent(zc->zc_name, parentname,
1027             sizeof (parentname))) != 0)
1028                 return (error);
1029 
1030         if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1031             (error = zfs_secpolicy_write_perms(origin,
1032             ZFS_DELEG_PERM_CLONE, cr)) != 0)
1033                 return (error);
1034 
1035         if ((error = zfs_secpolicy_write_perms(parentname,
1036             ZFS_DELEG_PERM_CREATE, cr)) != 0)
1037                 return (error);
1038 
1039         return (zfs_secpolicy_write_perms(parentname,
1040             ZFS_DELEG_PERM_MOUNT, cr));
1041 }
1042 
1043 /*
1044  * Policy for pool operations - create/destroy pools, add vdevs, etc.  Requires
1045  * SYS_CONFIG privilege, which is not available in a local zone.
1046  */
1047 /* ARGSUSED */
1048 static int
1049 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1050 {
1051         if (secpolicy_sys_config(cr, B_FALSE) != 0)
1052                 return (SET_ERROR(EPERM));
1053 
1054         return (0);
1055 }
1056 
1057 /*
1058  * Policy for object to name lookups.
1059  */
1060 /* ARGSUSED */
1061 static int
1062 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1063 {
1064         int error;
1065 
1066         if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1067                 return (0);
1068 
1069         error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1070         return (error);
1071 }
1072 
1073 /*
1074  * Policy for fault injection.  Requires all privileges.
1075  */
1076 /* ARGSUSED */
1077 static int
1078 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1079 {
1080         return (secpolicy_zinject(cr));
1081 }
1082 
1083 /* ARGSUSED */
1084 static int
1085 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1086 {
1087         zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1088 
1089         if (prop == ZPROP_INVAL) {
1090                 if (!zfs_prop_user(zc->zc_value))
1091                         return (SET_ERROR(EINVAL));
1092                 return (zfs_secpolicy_write_perms(zc->zc_name,
1093                     ZFS_DELEG_PERM_USERPROP, cr));
1094         } else {
1095                 return (zfs_secpolicy_setprop(zc->zc_name, prop,
1096                     NULL, cr));
1097         }
1098 }
1099 
1100 static int
1101 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1102 {
1103         int err = zfs_secpolicy_read(zc, innvl, cr);
1104         if (err)
1105                 return (err);
1106 
1107         if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1108                 return (SET_ERROR(EINVAL));
1109 
1110         if (zc->zc_value[0] == 0) {
1111                 /*
1112                  * They are asking about a posix uid/gid.  If it's
1113                  * themself, allow it.
1114                  */
1115                 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1116                     zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
1117                         if (zc->zc_guid == crgetuid(cr))
1118                                 return (0);
1119                 } else {
1120                         if (groupmember(zc->zc_guid, cr))
1121                                 return (0);
1122                 }
1123         }
1124 
1125         return (zfs_secpolicy_write_perms(zc->zc_name,
1126             userquota_perms[zc->zc_objset_type], cr));
1127 }
1128 
1129 static int
1130 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1131 {
1132         int err = zfs_secpolicy_read(zc, innvl, cr);
1133         if (err)
1134                 return (err);
1135 
1136         if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1137                 return (SET_ERROR(EINVAL));
1138 
1139         return (zfs_secpolicy_write_perms(zc->zc_name,
1140             userquota_perms[zc->zc_objset_type], cr));
1141 }
1142 
1143 /* ARGSUSED */
1144 static int
1145 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1146 {
1147         return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1148             NULL, cr));
1149 }
1150 
1151 /* ARGSUSED */
1152 static int
1153 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1154 {
1155         nvpair_t *pair;
1156         nvlist_t *holds;
1157         int error;
1158 
1159         error = nvlist_lookup_nvlist(innvl, "holds", &holds);
1160         if (error != 0)
1161                 return (SET_ERROR(EINVAL));
1162 
1163         for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
1164             pair = nvlist_next_nvpair(holds, pair)) {
1165                 char fsname[MAXNAMELEN];
1166                 error = dmu_fsname(nvpair_name(pair), fsname);
1167                 if (error != 0)
1168                         return (error);
1169                 error = zfs_secpolicy_write_perms(fsname,
1170                     ZFS_DELEG_PERM_HOLD, cr);
1171                 if (error != 0)
1172                         return (error);
1173         }
1174         return (0);
1175 }
1176 
1177 /* ARGSUSED */
1178 static int
1179 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1180 {
1181         nvpair_t *pair;
1182         int error;
1183 
1184         for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1185             pair = nvlist_next_nvpair(innvl, pair)) {
1186                 char fsname[MAXNAMELEN];
1187                 error = dmu_fsname(nvpair_name(pair), fsname);
1188                 if (error != 0)
1189                         return (error);
1190                 error = zfs_secpolicy_write_perms(fsname,
1191                     ZFS_DELEG_PERM_RELEASE, cr);
1192                 if (error != 0)
1193                         return (error);
1194         }
1195         return (0);
1196 }
1197 
1198 /*
1199  * Policy for allowing temporary snapshots to be taken or released
1200  */
1201 static int
1202 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1203 {
1204         /*
1205          * A temporary snapshot is the same as a snapshot,
1206          * hold, destroy and release all rolled into one.
1207          * Delegated diff alone is sufficient that we allow this.
1208          */
1209         int error;
1210 
1211         if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1212             ZFS_DELEG_PERM_DIFF, cr)) == 0)
1213                 return (0);
1214 
1215         error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1216         if (error == 0)
1217                 error = zfs_secpolicy_hold(zc, innvl, cr);
1218         if (error == 0)
1219                 error = zfs_secpolicy_release(zc, innvl, cr);
1220         if (error == 0)
1221                 error = zfs_secpolicy_destroy(zc, innvl, cr);
1222         return (error);
1223 }
1224 
1225 /*
1226  * Returns the nvlist as specified by the user in the zfs_cmd_t.
1227  */
1228 static int
1229 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1230 {
1231         char *packed;
1232         int error;
1233         nvlist_t *list = NULL;
1234 
1235         /*
1236          * Read in and unpack the user-supplied nvlist.
1237          */
1238         if (size == 0)
1239                 return (SET_ERROR(EINVAL));
1240 
1241         packed = kmem_alloc(size, KM_SLEEP);
1242 
1243         if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1244             iflag)) != 0) {
1245                 kmem_free(packed, size);
1246                 return (error);
1247         }
1248 
1249         if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1250                 kmem_free(packed, size);
1251                 return (error);
1252         }
1253 
1254         kmem_free(packed, size);
1255 
1256         *nvp = list;
1257         return (0);
1258 }
1259 
1260 /*
1261  * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1262  * Entries will be removed from the end of the nvlist, and one int32 entry
1263  * named "N_MORE_ERRORS" will be added indicating how many entries were
1264  * removed.
1265  */
1266 static int
1267 nvlist_smush(nvlist_t *errors, size_t max)
1268 {
1269         size_t size;
1270 
1271         size = fnvlist_size(errors);
1272 
1273         if (size > max) {
1274                 nvpair_t *more_errors;
1275                 int n = 0;
1276 
1277                 if (max < 1024)
1278                         return (SET_ERROR(ENOMEM));
1279 
1280                 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1281                 more_errors = nvlist_prev_nvpair(errors, NULL);
1282 
1283                 do {
1284                         nvpair_t *pair = nvlist_prev_nvpair(errors,
1285                             more_errors);
1286                         fnvlist_remove_nvpair(errors, pair);
1287                         n++;
1288                         size = fnvlist_size(errors);
1289                 } while (size > max);
1290 
1291                 fnvlist_remove_nvpair(errors, more_errors);
1292                 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1293                 ASSERT3U(fnvlist_size(errors), <=, max);
1294         }
1295 
1296         return (0);
1297 }
1298 
1299 static int
1300 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1301 {
1302         char *packed = NULL;
1303         int error = 0;
1304         size_t size;
1305 
1306         size = fnvlist_size(nvl);
1307 
1308         if (size > zc->zc_nvlist_dst_size) {
1309                 error = SET_ERROR(ENOMEM);
1310         } else {
1311                 packed = fnvlist_pack(nvl, &size);
1312                 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1313                     size, zc->zc_iflags) != 0)
1314                         error = SET_ERROR(EFAULT);
1315                 fnvlist_pack_free(packed, size);
1316         }
1317 
1318         zc->zc_nvlist_dst_size = size;
1319         zc->zc_nvlist_dst_filled = B_TRUE;
1320         return (error);
1321 }
1322 
1323 static int
1324 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1325 {
1326         objset_t *os;
1327         int error;
1328 
1329         error = dmu_objset_hold(dsname, FTAG, &os);
1330         if (error != 0)
1331                 return (error);
1332         if (dmu_objset_type(os) != DMU_OST_ZFS) {
1333                 dmu_objset_rele(os, FTAG);
1334                 return (SET_ERROR(EINVAL));
1335         }
1336 
1337         mutex_enter(&os->os_user_ptr_lock);
1338         *zfvp = dmu_objset_get_user(os);
1339         if (*zfvp) {
1340                 VFS_HOLD((*zfvp)->z_vfs);
1341         } else {
1342                 error = SET_ERROR(ESRCH);
1343         }
1344         mutex_exit(&os->os_user_ptr_lock);
1345         dmu_objset_rele(os, FTAG);
1346         return (error);
1347 }
1348 
1349 /*
1350  * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1351  * case its z_vfs will be NULL, and it will be opened as the owner.
1352  * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1353  * which prevents all vnode ops from running.
1354  */
1355 static int
1356 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1357 {
1358         int error = 0;
1359 
1360         if (getzfsvfs(name, zfvp) != 0)
1361                 error = zfsvfs_create(name, zfvp);
1362         if (error == 0) {
1363                 rrw_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER :
1364                     RW_READER, tag);
1365                 if ((*zfvp)->z_unmounted) {
1366                         /*
1367                          * XXX we could probably try again, since the unmounting
1368                          * thread should be just about to disassociate the
1369                          * objset from the zfsvfs.
1370                          */
1371                         rrw_exit(&(*zfvp)->z_teardown_lock, tag);
1372                         return (SET_ERROR(EBUSY));
1373                 }
1374         }
1375         return (error);
1376 }
1377 
1378 static void
1379 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1380 {
1381         rrw_exit(&zfsvfs->z_teardown_lock, tag);
1382 
1383         if (zfsvfs->z_vfs) {
1384                 VFS_RELE(zfsvfs->z_vfs);
1385         } else {
1386                 dmu_objset_disown(zfsvfs->z_os, zfsvfs);
1387                 zfsvfs_free(zfsvfs);
1388         }
1389 }
1390 
1391 static int
1392 zfs_ioc_pool_create(zfs_cmd_t *zc)
1393 {
1394         int error;
1395         nvlist_t *config, *props = NULL;
1396         nvlist_t *rootprops = NULL;
1397         nvlist_t *zplprops = NULL;
1398 
1399         if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1400             zc->zc_iflags, &config))
1401                 return (error);
1402 
1403         if (zc->zc_nvlist_src_size != 0 && (error =
1404             get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1405             zc->zc_iflags, &props))) {
1406                 nvlist_free(config);
1407                 return (error);
1408         }
1409 
1410         if (props) {
1411                 nvlist_t *nvl = NULL;
1412                 uint64_t version = SPA_VERSION;
1413 
1414                 (void) nvlist_lookup_uint64(props,
1415                     zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1416                 if (!SPA_VERSION_IS_SUPPORTED(version)) {
1417                         error = SET_ERROR(EINVAL);
1418                         goto pool_props_bad;
1419                 }
1420                 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1421                 if (nvl) {
1422                         error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1423                         if (error != 0) {
1424                                 nvlist_free(config);
1425                                 nvlist_free(props);
1426                                 return (error);
1427                         }
1428                         (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1429                 }
1430                 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1431                 error = zfs_fill_zplprops_root(version, rootprops,
1432                     zplprops, NULL);
1433                 if (error != 0)
1434                         goto pool_props_bad;
1435         }
1436 
1437         error = spa_create(zc->zc_name, config, props, zplprops);
1438 
1439         /*
1440          * Set the remaining root properties
1441          */
1442         if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1443             ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1444                 (void) spa_destroy(zc->zc_name);
1445 
1446 pool_props_bad:
1447         nvlist_free(rootprops);
1448         nvlist_free(zplprops);
1449         nvlist_free(config);
1450         nvlist_free(props);
1451 
1452         return (error);
1453 }
1454 
1455 static int
1456 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1457 {
1458         int error;
1459         zfs_log_history(zc);
1460         error = spa_destroy(zc->zc_name);
1461         if (error == 0)
1462                 zvol_remove_minors(zc->zc_name);
1463         return (error);
1464 }
1465 
1466 static int
1467 zfs_ioc_pool_import(zfs_cmd_t *zc)
1468 {
1469         nvlist_t *config, *props = NULL;
1470         uint64_t guid;
1471         int error;
1472 
1473         if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1474             zc->zc_iflags, &config)) != 0)
1475                 return (error);
1476 
1477         if (zc->zc_nvlist_src_size != 0 && (error =
1478             get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1479             zc->zc_iflags, &props))) {
1480                 nvlist_free(config);
1481                 return (error);
1482         }
1483 
1484         if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1485             guid != zc->zc_guid)
1486                 error = SET_ERROR(EINVAL);
1487         else
1488                 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1489 
1490         if (zc->zc_nvlist_dst != 0) {
1491                 int err;
1492 
1493                 if ((err = put_nvlist(zc, config)) != 0)
1494                         error = err;
1495         }
1496 
1497         nvlist_free(config);
1498 
1499         if (props)
1500                 nvlist_free(props);
1501 
1502         return (error);
1503 }
1504 
1505 static int
1506 zfs_ioc_pool_export(zfs_cmd_t *zc)
1507 {
1508         int error;
1509         boolean_t force = (boolean_t)zc->zc_cookie;
1510         boolean_t hardforce = (boolean_t)zc->zc_guid;
1511 
1512         zfs_log_history(zc);
1513         error = spa_export(zc->zc_name, NULL, force, hardforce);
1514         if (error == 0)
1515                 zvol_remove_minors(zc->zc_name);
1516         return (error);
1517 }
1518 
1519 static int
1520 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1521 {
1522         nvlist_t *configs;
1523         int error;
1524 
1525         if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1526                 return (SET_ERROR(EEXIST));
1527 
1528         error = put_nvlist(zc, configs);
1529 
1530         nvlist_free(configs);
1531 
1532         return (error);
1533 }
1534 
1535 /*
1536  * inputs:
1537  * zc_name              name of the pool
1538  *
1539  * outputs:
1540  * zc_cookie            real errno
1541  * zc_nvlist_dst        config nvlist
1542  * zc_nvlist_dst_size   size of config nvlist
1543  */
1544 static int
1545 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1546 {
1547         nvlist_t *config;
1548         int error;
1549         int ret = 0;
1550 
1551         error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1552             sizeof (zc->zc_value));
1553 
1554         if (config != NULL) {
1555                 ret = put_nvlist(zc, config);
1556                 nvlist_free(config);
1557 
1558                 /*
1559                  * The config may be present even if 'error' is non-zero.
1560                  * In this case we return success, and preserve the real errno
1561                  * in 'zc_cookie'.
1562                  */
1563                 zc->zc_cookie = error;
1564         } else {
1565                 ret = error;
1566         }
1567 
1568         return (ret);
1569 }
1570 
1571 /*
1572  * Try to import the given pool, returning pool stats as appropriate so that
1573  * user land knows which devices are available and overall pool health.
1574  */
1575 static int
1576 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1577 {
1578         nvlist_t *tryconfig, *config;
1579         int error;
1580 
1581         if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1582             zc->zc_iflags, &tryconfig)) != 0)
1583                 return (error);
1584 
1585         config = spa_tryimport(tryconfig);
1586 
1587         nvlist_free(tryconfig);
1588 
1589         if (config == NULL)
1590                 return (SET_ERROR(EINVAL));
1591 
1592         error = put_nvlist(zc, config);
1593         nvlist_free(config);
1594 
1595         return (error);
1596 }
1597 
1598 /*
1599  * inputs:
1600  * zc_name              name of the pool
1601  * zc_cookie            scan func (pool_scan_func_t)
1602  */
1603 static int
1604 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1605 {
1606         spa_t *spa;
1607         int error;
1608 
1609         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1610                 return (error);
1611 
1612         if (zc->zc_cookie == POOL_SCAN_NONE)
1613                 error = spa_scan_stop(spa);
1614         else
1615                 error = spa_scan(spa, zc->zc_cookie);
1616 
1617         spa_close(spa, FTAG);
1618 
1619         return (error);
1620 }
1621 
1622 static int
1623 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1624 {
1625         spa_t *spa;
1626         int error;
1627 
1628         error = spa_open(zc->zc_name, &spa, FTAG);
1629         if (error == 0) {
1630                 spa_freeze(spa);
1631                 spa_close(spa, FTAG);
1632         }
1633         return (error);
1634 }
1635 
1636 static int
1637 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1638 {
1639         spa_t *spa;
1640         int error;
1641 
1642         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1643                 return (error);
1644 
1645         if (zc->zc_cookie < spa_version(spa) ||
1646             !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1647                 spa_close(spa, FTAG);
1648                 return (SET_ERROR(EINVAL));
1649         }
1650 
1651         spa_upgrade(spa, zc->zc_cookie);
1652         spa_close(spa, FTAG);
1653 
1654         return (error);
1655 }
1656 
1657 static int
1658 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1659 {
1660         spa_t *spa;
1661         char *hist_buf;
1662         uint64_t size;
1663         int error;
1664 
1665         if ((size = zc->zc_history_len) == 0)
1666                 return (SET_ERROR(EINVAL));
1667 
1668         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1669                 return (error);
1670 
1671         if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1672                 spa_close(spa, FTAG);
1673                 return (SET_ERROR(ENOTSUP));
1674         }
1675 
1676         hist_buf = kmem_alloc(size, KM_SLEEP);
1677         if ((error = spa_history_get(spa, &zc->zc_history_offset,
1678             &zc->zc_history_len, hist_buf)) == 0) {
1679                 error = ddi_copyout(hist_buf,
1680                     (void *)(uintptr_t)zc->zc_history,
1681                     zc->zc_history_len, zc->zc_iflags);
1682         }
1683 
1684         spa_close(spa, FTAG);
1685         kmem_free(hist_buf, size);
1686         return (error);
1687 }
1688 
1689 static int
1690 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1691 {
1692         spa_t *spa;
1693         int error;
1694 
1695         error = spa_open(zc->zc_name, &spa, FTAG);
1696         if (error == 0) {
1697                 error = spa_change_guid(spa);
1698                 spa_close(spa, FTAG);
1699         }
1700         return (error);
1701 }
1702 
1703 static int
1704 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1705 {
1706         return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1707 }
1708 
1709 /*
1710  * inputs:
1711  * zc_name              name of filesystem
1712  * zc_obj               object to find
1713  *
1714  * outputs:
1715  * zc_value             name of object
1716  */
1717 static int
1718 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1719 {
1720         objset_t *os;
1721         int error;
1722 
1723         /* XXX reading from objset not owned */
1724         if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1725                 return (error);
1726         if (dmu_objset_type(os) != DMU_OST_ZFS) {
1727                 dmu_objset_rele(os, FTAG);
1728                 return (SET_ERROR(EINVAL));
1729         }
1730         error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1731             sizeof (zc->zc_value));
1732         dmu_objset_rele(os, FTAG);
1733 
1734         return (error);
1735 }
1736 
1737 /*
1738  * inputs:
1739  * zc_name              name of filesystem
1740  * zc_obj               object to find
1741  *
1742  * outputs:
1743  * zc_stat              stats on object
1744  * zc_value             path to object
1745  */
1746 static int
1747 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1748 {
1749         objset_t *os;
1750         int error;
1751 
1752         /* XXX reading from objset not owned */
1753         if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1754                 return (error);
1755         if (dmu_objset_type(os) != DMU_OST_ZFS) {
1756                 dmu_objset_rele(os, FTAG);
1757                 return (SET_ERROR(EINVAL));
1758         }
1759         error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1760             sizeof (zc->zc_value));
1761         dmu_objset_rele(os, FTAG);
1762 
1763         return (error);
1764 }
1765 
1766 static int
1767 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1768 {
1769         spa_t *spa;
1770         int error;
1771         nvlist_t *config, **l2cache, **spares;
1772         uint_t nl2cache = 0, nspares = 0;
1773 
1774         error = spa_open(zc->zc_name, &spa, FTAG);
1775         if (error != 0)
1776                 return (error);
1777 
1778         error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1779             zc->zc_iflags, &config);
1780         (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1781             &l2cache, &nl2cache);
1782 
1783         (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1784             &spares, &nspares);
1785 
1786         /*
1787          * A root pool with concatenated devices is not supported.
1788          * Thus, can not add a device to a root pool.
1789          *
1790          * Intent log device can not be added to a rootpool because
1791          * during mountroot, zil is replayed, a seperated log device
1792          * can not be accessed during the mountroot time.
1793          *
1794          * l2cache and spare devices are ok to be added to a rootpool.
1795          */
1796         if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1797                 nvlist_free(config);
1798                 spa_close(spa, FTAG);
1799                 return (SET_ERROR(EDOM));
1800         }
1801 
1802         if (error == 0) {
1803                 error = spa_vdev_add(spa, config);
1804                 nvlist_free(config);
1805         }
1806         spa_close(spa, FTAG);
1807         return (error);
1808 }
1809 
1810 /*
1811  * inputs:
1812  * zc_name              name of the pool
1813  * zc_nvlist_conf       nvlist of devices to remove
1814  * zc_cookie            to stop the remove?
1815  */
1816 static int
1817 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1818 {
1819         spa_t *spa;
1820         int error;
1821 
1822         error = spa_open(zc->zc_name, &spa, FTAG);
1823         if (error != 0)
1824                 return (error);
1825         error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1826         spa_close(spa, FTAG);
1827         return (error);
1828 }
1829 
1830 static int
1831 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1832 {
1833         spa_t *spa;
1834         int error;
1835         vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1836 
1837         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1838                 return (error);
1839         switch (zc->zc_cookie) {
1840         case VDEV_STATE_ONLINE:
1841                 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1842                 break;
1843 
1844         case VDEV_STATE_OFFLINE:
1845                 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1846                 break;
1847 
1848         case VDEV_STATE_FAULTED:
1849                 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1850                     zc->zc_obj != VDEV_AUX_EXTERNAL)
1851                         zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1852 
1853                 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1854                 break;
1855 
1856         case VDEV_STATE_DEGRADED:
1857                 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1858                     zc->zc_obj != VDEV_AUX_EXTERNAL)
1859                         zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1860 
1861                 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1862                 break;
1863 
1864         default:
1865                 error = SET_ERROR(EINVAL);
1866         }
1867         zc->zc_cookie = newstate;
1868         spa_close(spa, FTAG);
1869         return (error);
1870 }
1871 
1872 static int
1873 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1874 {
1875         spa_t *spa;
1876         int replacing = zc->zc_cookie;
1877         nvlist_t *config;
1878         int error;
1879 
1880         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1881                 return (error);
1882 
1883         if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1884             zc->zc_iflags, &config)) == 0) {
1885                 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1886                 nvlist_free(config);
1887         }
1888 
1889         spa_close(spa, FTAG);
1890         return (error);
1891 }
1892 
1893 static int
1894 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1895 {
1896         spa_t *spa;
1897         int error;
1898 
1899         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1900                 return (error);
1901 
1902         error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
1903 
1904         spa_close(spa, FTAG);
1905         return (error);
1906 }
1907 
1908 static int
1909 zfs_ioc_vdev_split(zfs_cmd_t *zc)
1910 {
1911         spa_t *spa;
1912         nvlist_t *config, *props = NULL;
1913         int error;
1914         boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
1915 
1916         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1917                 return (error);
1918 
1919         if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1920             zc->zc_iflags, &config)) {
1921                 spa_close(spa, FTAG);
1922                 return (error);
1923         }
1924 
1925         if (zc->zc_nvlist_src_size != 0 && (error =
1926             get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1927             zc->zc_iflags, &props))) {
1928                 spa_close(spa, FTAG);
1929                 nvlist_free(config);
1930                 return (error);
1931         }
1932 
1933         error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
1934 
1935         spa_close(spa, FTAG);
1936 
1937         nvlist_free(config);
1938         nvlist_free(props);
1939 
1940         return (error);
1941 }
1942 
1943 static int
1944 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
1945 {
1946         spa_t *spa;
1947         char *path = zc->zc_value;
1948         uint64_t guid = zc->zc_guid;
1949         int error;
1950 
1951         error = spa_open(zc->zc_name, &spa, FTAG);
1952         if (error != 0)
1953                 return (error);
1954 
1955         error = spa_vdev_setpath(spa, guid, path);
1956         spa_close(spa, FTAG);
1957         return (error);
1958 }
1959 
1960 static int
1961 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
1962 {
1963         spa_t *spa;
1964         char *fru = zc->zc_value;
1965         uint64_t guid = zc->zc_guid;
1966         int error;
1967 
1968         error = spa_open(zc->zc_name, &spa, FTAG);
1969         if (error != 0)
1970                 return (error);
1971 
1972         error = spa_vdev_setfru(spa, guid, fru);
1973         spa_close(spa, FTAG);
1974         return (error);
1975 }
1976 
1977 static int
1978 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
1979 {
1980         int error = 0;
1981         nvlist_t *nv;
1982 
1983         dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1984 
1985         if (zc->zc_nvlist_dst != 0 &&
1986             (error = dsl_prop_get_all(os, &nv)) == 0) {
1987                 dmu_objset_stats(os, nv);
1988                 /*
1989                  * NB: zvol_get_stats() will read the objset contents,
1990                  * which we aren't supposed to do with a
1991                  * DS_MODE_USER hold, because it could be
1992                  * inconsistent.  So this is a bit of a workaround...
1993                  * XXX reading with out owning
1994                  */
1995                 if (!zc->zc_objset_stats.dds_inconsistent &&
1996                     dmu_objset_type(os) == DMU_OST_ZVOL) {
1997                         error = zvol_get_stats(os, nv);
1998                         if (error == EIO)
1999                                 return (error);
2000                         VERIFY0(error);
2001                 }
2002                 error = put_nvlist(zc, nv);
2003                 nvlist_free(nv);
2004         }
2005 
2006         return (error);
2007 }
2008 
2009 /*
2010  * inputs:
2011  * zc_name              name of filesystem
2012  * zc_nvlist_dst_size   size of buffer for property nvlist
2013  *
2014  * outputs:
2015  * zc_objset_stats      stats
2016  * zc_nvlist_dst        property nvlist
2017  * zc_nvlist_dst_size   size of property nvlist
2018  */
2019 static int
2020 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2021 {
2022         objset_t *os;
2023         int error;
2024 
2025         error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2026         if (error == 0) {
2027                 error = zfs_ioc_objset_stats_impl(zc, os);
2028                 dmu_objset_rele(os, FTAG);
2029         }
2030 
2031         return (error);
2032 }
2033 
2034 /*
2035  * inputs:
2036  * zc_name              name of filesystem
2037  * zc_nvlist_dst_size   size of buffer for property nvlist
2038  *
2039  * outputs:
2040  * zc_nvlist_dst        received property nvlist
2041  * zc_nvlist_dst_size   size of received property nvlist
2042  *
2043  * Gets received properties (distinct from local properties on or after
2044  * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2045  * local property values.
2046  */
2047 static int
2048 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2049 {
2050         int error = 0;
2051         nvlist_t *nv;
2052 
2053         /*
2054          * Without this check, we would return local property values if the
2055          * caller has not already received properties on or after
2056          * SPA_VERSION_RECVD_PROPS.
2057          */
2058         if (!dsl_prop_get_hasrecvd(zc->zc_name))
2059                 return (SET_ERROR(ENOTSUP));
2060 
2061         if (zc->zc_nvlist_dst != 0 &&
2062             (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2063                 error = put_nvlist(zc, nv);
2064                 nvlist_free(nv);
2065         }
2066 
2067         return (error);
2068 }
2069 
2070 static int
2071 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2072 {
2073         uint64_t value;
2074         int error;
2075 
2076         /*
2077          * zfs_get_zplprop() will either find a value or give us
2078          * the default value (if there is one).
2079          */
2080         if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2081                 return (error);
2082         VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2083         return (0);
2084 }
2085 
2086 /*
2087  * inputs:
2088  * zc_name              name of filesystem
2089  * zc_nvlist_dst_size   size of buffer for zpl property nvlist
2090  *
2091  * outputs:
2092  * zc_nvlist_dst        zpl property nvlist
2093  * zc_nvlist_dst_size   size of zpl property nvlist
2094  */
2095 static int
2096 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2097 {
2098         objset_t *os;
2099         int err;
2100 
2101         /* XXX reading without owning */
2102         if (err = dmu_objset_hold(zc->zc_name, FTAG, &os))
2103                 return (err);
2104 
2105         dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2106 
2107         /*
2108          * NB: nvl_add_zplprop() will read the objset contents,
2109          * which we aren't supposed to do with a DS_MODE_USER
2110          * hold, because it could be inconsistent.
2111          */
2112         if (zc->zc_nvlist_dst != NULL &&
2113             !zc->zc_objset_stats.dds_inconsistent &&
2114             dmu_objset_type(os) == DMU_OST_ZFS) {
2115                 nvlist_t *nv;
2116 
2117                 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2118                 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2119                     (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2120                     (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2121                     (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2122                         err = put_nvlist(zc, nv);
2123                 nvlist_free(nv);
2124         } else {
2125                 err = SET_ERROR(ENOENT);
2126         }
2127         dmu_objset_rele(os, FTAG);
2128         return (err);
2129 }
2130 
2131 static boolean_t
2132 dataset_name_hidden(const char *name)
2133 {
2134         /*
2135          * Skip over datasets that are not visible in this zone,
2136          * internal datasets (which have a $ in their name), and
2137          * temporary datasets (which have a % in their name).
2138          */
2139         if (strchr(name, '$') != NULL)
2140                 return (B_TRUE);
2141         if (strchr(name, '%') != NULL)
2142                 return (B_TRUE);
2143         if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
2144                 return (B_TRUE);
2145         return (B_FALSE);
2146 }
2147 
2148 /*
2149  * inputs:
2150  * zc_name              name of filesystem
2151  * zc_cookie            zap cursor
2152  * zc_nvlist_dst_size   size of buffer for property nvlist
2153  *
2154  * outputs:
2155  * zc_name              name of next filesystem
2156  * zc_cookie            zap cursor
2157  * zc_objset_stats      stats
2158  * zc_nvlist_dst        property nvlist
2159  * zc_nvlist_dst_size   size of property nvlist
2160  */
2161 static int
2162 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2163 {
2164         objset_t *os;
2165         int error;
2166         char *p;
2167         size_t orig_len = strlen(zc->zc_name);
2168 
2169 top:
2170         if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) {
2171                 if (error == ENOENT)
2172                         error = SET_ERROR(ESRCH);
2173                 return (error);
2174         }
2175 
2176         p = strrchr(zc->zc_name, '/');
2177         if (p == NULL || p[1] != '\0')
2178                 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2179         p = zc->zc_name + strlen(zc->zc_name);
2180 
2181         do {
2182                 error = dmu_dir_list_next(os,
2183                     sizeof (zc->zc_name) - (p - zc->zc_name), p,
2184                     NULL, &zc->zc_cookie);
2185                 if (error == ENOENT)
2186                         error = SET_ERROR(ESRCH);
2187         } while (error == 0 && dataset_name_hidden(zc->zc_name));
2188         dmu_objset_rele(os, FTAG);
2189 
2190         /*
2191          * If it's an internal dataset (ie. with a '$' in its name),
2192          * don't try to get stats for it, otherwise we'll return ENOENT.
2193          */
2194         if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2195                 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2196                 if (error == ENOENT) {
2197                         /* We lost a race with destroy, get the next one. */
2198                         zc->zc_name[orig_len] = '\0';
2199                         goto top;
2200                 }
2201         }
2202         return (error);
2203 }
2204 
2205 /*
2206  * inputs:
2207  * zc_name              name of filesystem
2208  * zc_cookie            zap cursor
2209  * zc_nvlist_dst_size   size of buffer for property nvlist
2210  *
2211  * outputs:
2212  * zc_name              name of next snapshot
2213  * zc_objset_stats      stats
2214  * zc_nvlist_dst        property nvlist
2215  * zc_nvlist_dst_size   size of property nvlist
2216  */
2217 static int
2218 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2219 {
2220         objset_t *os;
2221         int error;
2222 
2223         error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2224         if (error != 0) {
2225                 return (error == ENOENT ? ESRCH : error);
2226         }
2227 
2228         /*
2229          * A dataset name of maximum length cannot have any snapshots,
2230          * so exit immediately.
2231          */
2232         if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
2233                 dmu_objset_rele(os, FTAG);
2234                 return (SET_ERROR(ESRCH));
2235         }
2236 
2237         error = dmu_snapshot_list_next(os,
2238             sizeof (zc->zc_name) - strlen(zc->zc_name),
2239             zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2240             NULL);
2241 
2242         if (error == 0) {
2243                 dsl_dataset_t *ds;
2244                 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2245 
2246                 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2247                 if (error == 0) {
2248                         objset_t *ossnap;
2249 
2250                         error = dmu_objset_from_ds(ds, &ossnap);
2251                         if (error == 0)
2252                                 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2253                         dsl_dataset_rele(ds, FTAG);
2254                 }
2255         } else if (error == ENOENT) {
2256                 error = SET_ERROR(ESRCH);
2257         }
2258 
2259         dmu_objset_rele(os, FTAG);
2260         /* if we failed, undo the @ that we tacked on to zc_name */
2261         if (error != 0)
2262                 *strchr(zc->zc_name, '@') = '\0';
2263         return (error);
2264 }
2265 
2266 static int
2267 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2268 {
2269         const char *propname = nvpair_name(pair);
2270         uint64_t *valary;
2271         unsigned int vallen;
2272         const char *domain;
2273         char *dash;
2274         zfs_userquota_prop_t type;
2275         uint64_t rid;
2276         uint64_t quota;
2277         zfsvfs_t *zfsvfs;
2278         int err;
2279 
2280         if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2281                 nvlist_t *attrs;
2282                 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2283                 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2284                     &pair) != 0)
2285                         return (SET_ERROR(EINVAL));
2286         }
2287 
2288         /*
2289          * A correctly constructed propname is encoded as
2290          * userquota@<rid>-<domain>.
2291          */
2292         if ((dash = strchr(propname, '-')) == NULL ||
2293             nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2294             vallen != 3)
2295                 return (SET_ERROR(EINVAL));
2296 
2297         domain = dash + 1;
2298         type = valary[0];
2299         rid = valary[1];
2300         quota = valary[2];
2301 
2302         err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2303         if (err == 0) {
2304                 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2305                 zfsvfs_rele(zfsvfs, FTAG);
2306         }
2307 
2308         return (err);
2309 }
2310 
2311 /*
2312  * If the named property is one that has a special function to set its value,
2313  * return 0 on success and a positive error code on failure; otherwise if it is
2314  * not one of the special properties handled by this function, return -1.
2315  *
2316  * XXX: It would be better for callers of the property interface if we handled
2317  * these special cases in dsl_prop.c (in the dsl layer).
2318  */
2319 static int
2320 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2321     nvpair_t *pair)
2322 {
2323         const char *propname = nvpair_name(pair);
2324         zfs_prop_t prop = zfs_name_to_prop(propname);
2325         uint64_t intval;
2326         int err;
2327 
2328         if (prop == ZPROP_INVAL) {
2329                 if (zfs_prop_userquota(propname))
2330                         return (zfs_prop_set_userquota(dsname, pair));
2331                 return (-1);
2332         }
2333 
2334         if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2335                 nvlist_t *attrs;
2336                 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2337                 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2338                     &pair) == 0);
2339         }
2340 
2341         if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2342                 return (-1);
2343 
2344         VERIFY(0 == nvpair_value_uint64(pair, &intval));
2345 
2346         switch (prop) {
2347         case ZFS_PROP_QUOTA:
2348                 err = dsl_dir_set_quota(dsname, source, intval);
2349                 break;
2350         case ZFS_PROP_REFQUOTA:
2351                 err = dsl_dataset_set_refquota(dsname, source, intval);
2352                 break;
2353         case ZFS_PROP_RESERVATION:
2354                 err = dsl_dir_set_reservation(dsname, source, intval);
2355                 break;
2356         case ZFS_PROP_REFRESERVATION:
2357                 err = dsl_dataset_set_refreservation(dsname, source, intval);
2358                 break;
2359         case ZFS_PROP_VOLSIZE:
2360                 err = zvol_set_volsize(dsname, intval);
2361                 break;
2362         case ZFS_PROP_VERSION:
2363         {
2364                 zfsvfs_t *zfsvfs;
2365 
2366                 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2367                         break;
2368 
2369                 err = zfs_set_version(zfsvfs, intval);
2370                 zfsvfs_rele(zfsvfs, FTAG);
2371 
2372                 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2373                         zfs_cmd_t *zc;
2374 
2375                         zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2376                         (void) strcpy(zc->zc_name, dsname);
2377                         (void) zfs_ioc_userspace_upgrade(zc);
2378                         kmem_free(zc, sizeof (zfs_cmd_t));
2379                 }
2380                 break;
2381         }
2382         case ZFS_PROP_COMPRESSION:
2383         {
2384                 if (intval == ZIO_COMPRESS_LZ4) {
2385                         zfeature_info_t *feature =
2386                             &spa_feature_table[SPA_FEATURE_LZ4_COMPRESS];
2387                         spa_t *spa;
2388 
2389                         if ((err = spa_open(dsname, &spa, FTAG)) != 0)
2390                                 return (err);
2391 
2392                         /*
2393                          * Setting the LZ4 compression algorithm activates
2394                          * the feature.
2395                          */
2396                         if (!spa_feature_is_active(spa, feature)) {
2397                                 if ((err = zfs_prop_activate_feature(spa,
2398                                     feature)) != 0) {
2399                                         spa_close(spa, FTAG);
2400                                         return (err);
2401                                 }
2402                         }
2403 
2404                         spa_close(spa, FTAG);
2405                 }
2406                 /*
2407                  * We still want the default set action to be performed in the
2408                  * caller, we only performed zfeature settings here.
2409                  */
2410                 err = -1;
2411                 break;
2412         }
2413 
2414         default:
2415                 err = -1;
2416         }
2417 
2418         return (err);
2419 }
2420 
2421 /*
2422  * This function is best effort. If it fails to set any of the given properties,
2423  * it continues to set as many as it can and returns the last error
2424  * encountered. If the caller provides a non-NULL errlist, it will be filled in
2425  * with the list of names of all the properties that failed along with the
2426  * corresponding error numbers.
2427  *
2428  * If every property is set successfully, zero is returned and errlist is not
2429  * modified.
2430  */
2431 int
2432 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2433     nvlist_t *errlist)
2434 {
2435         nvpair_t *pair;
2436         nvpair_t *propval;
2437         int rv = 0;
2438         uint64_t intval;
2439         char *strval;
2440         nvlist_t *genericnvl = fnvlist_alloc();
2441         nvlist_t *retrynvl = fnvlist_alloc();
2442 
2443 retry:
2444         pair = NULL;
2445         while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2446                 const char *propname = nvpair_name(pair);
2447                 zfs_prop_t prop = zfs_name_to_prop(propname);
2448                 int err = 0;
2449 
2450                 /* decode the property value */
2451                 propval = pair;
2452                 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2453                         nvlist_t *attrs;
2454                         attrs = fnvpair_value_nvlist(pair);
2455                         if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2456                             &propval) != 0)
2457                                 err = SET_ERROR(EINVAL);
2458                 }
2459 
2460                 /* Validate value type */
2461                 if (err == 0 && prop == ZPROP_INVAL) {
2462                         if (zfs_prop_user(propname)) {
2463                                 if (nvpair_type(propval) != DATA_TYPE_STRING)
2464                                         err = SET_ERROR(EINVAL);
2465                         } else if (zfs_prop_userquota(propname)) {
2466                                 if (nvpair_type(propval) !=
2467                                     DATA_TYPE_UINT64_ARRAY)
2468                                         err = SET_ERROR(EINVAL);
2469                         } else {
2470                                 err = SET_ERROR(EINVAL);
2471                         }
2472                 } else if (err == 0) {
2473                         if (nvpair_type(propval) == DATA_TYPE_STRING) {
2474                                 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2475                                         err = SET_ERROR(EINVAL);
2476                         } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2477                                 const char *unused;
2478 
2479                                 intval = fnvpair_value_uint64(propval);
2480 
2481                                 switch (zfs_prop_get_type(prop)) {
2482                                 case PROP_TYPE_NUMBER:
2483                                         break;
2484                                 case PROP_TYPE_STRING:
2485                                         err = SET_ERROR(EINVAL);
2486                                         break;
2487                                 case PROP_TYPE_INDEX:
2488                                         if (zfs_prop_index_to_string(prop,
2489                                             intval, &unused) != 0)
2490                                                 err = SET_ERROR(EINVAL);
2491                                         break;
2492                                 default:
2493                                         cmn_err(CE_PANIC,
2494                                             "unknown property type");
2495                                 }
2496                         } else {
2497                                 err = SET_ERROR(EINVAL);
2498                         }
2499                 }
2500 
2501                 /* Validate permissions */
2502                 if (err == 0)
2503                         err = zfs_check_settable(dsname, pair, CRED());
2504 
2505                 if (err == 0) {
2506                         err = zfs_prop_set_special(dsname, source, pair);
2507                         if (err == -1) {
2508                                 /*
2509                                  * For better performance we build up a list of
2510                                  * properties to set in a single transaction.
2511                                  */
2512                                 err = nvlist_add_nvpair(genericnvl, pair);
2513                         } else if (err != 0 && nvl != retrynvl) {
2514                                 /*
2515                                  * This may be a spurious error caused by
2516                                  * receiving quota and reservation out of order.
2517                                  * Try again in a second pass.
2518                                  */
2519                                 err = nvlist_add_nvpair(retrynvl, pair);
2520                         }
2521                 }
2522 
2523                 if (err != 0) {
2524                         if (errlist != NULL)
2525                                 fnvlist_add_int32(errlist, propname, err);
2526                         rv = err;
2527                 }
2528         }
2529 
2530         if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2531                 nvl = retrynvl;
2532                 goto retry;
2533         }
2534 
2535         if (!nvlist_empty(genericnvl) &&
2536             dsl_props_set(dsname, source, genericnvl) != 0) {
2537                 /*
2538                  * If this fails, we still want to set as many properties as we
2539                  * can, so try setting them individually.
2540                  */
2541                 pair = NULL;
2542                 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2543                         const char *propname = nvpair_name(pair);
2544                         int err = 0;
2545 
2546                         propval = pair;
2547                         if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2548                                 nvlist_t *attrs;
2549                                 attrs = fnvpair_value_nvlist(pair);
2550                                 propval = fnvlist_lookup_nvpair(attrs,
2551                                     ZPROP_VALUE);
2552                         }
2553 
2554                         if (nvpair_type(propval) == DATA_TYPE_STRING) {
2555                                 strval = fnvpair_value_string(propval);
2556                                 err = dsl_prop_set_string(dsname, propname,
2557                                     source, strval);
2558                         } else {
2559                                 intval = fnvpair_value_uint64(propval);
2560                                 err = dsl_prop_set_int(dsname, propname, source,
2561                                     intval);
2562                         }
2563 
2564                         if (err != 0) {
2565                                 if (errlist != NULL) {
2566                                         fnvlist_add_int32(errlist, propname,
2567                                             err);
2568                                 }
2569                                 rv = err;
2570                         }
2571                 }
2572         }
2573         nvlist_free(genericnvl);
2574         nvlist_free(retrynvl);
2575 
2576         return (rv);
2577 }
2578 
2579 /*
2580  * Check that all the properties are valid user properties.
2581  */
2582 static int
2583 zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2584 {
2585         nvpair_t *pair = NULL;
2586         int error = 0;
2587 
2588         while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2589                 const char *propname = nvpair_name(pair);
2590                 char *valstr;
2591 
2592                 if (!zfs_prop_user(propname) ||
2593                     nvpair_type(pair) != DATA_TYPE_STRING)
2594                         return (SET_ERROR(EINVAL));
2595 
2596                 if (error = zfs_secpolicy_write_perms(fsname,
2597                     ZFS_DELEG_PERM_USERPROP, CRED()))
2598                         return (error);
2599 
2600                 if (strlen(propname) >= ZAP_MAXNAMELEN)
2601                         return (SET_ERROR(ENAMETOOLONG));
2602 
2603                 VERIFY(nvpair_value_string(pair, &valstr) == 0);
2604                 if (strlen(valstr) >= ZAP_MAXVALUELEN)
2605                         return (E2BIG);
2606         }
2607         return (0);
2608 }
2609 
2610 static void
2611 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2612 {
2613         nvpair_t *pair;
2614 
2615         VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2616 
2617         pair = NULL;
2618         while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2619                 if (nvlist_exists(skipped, nvpair_name(pair)))
2620                         continue;
2621 
2622                 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2623         }
2624 }
2625 
2626 static int
2627 clear_received_props(const char *dsname, nvlist_t *props,
2628     nvlist_t *skipped)
2629 {
2630         int err = 0;
2631         nvlist_t *cleared_props = NULL;
2632         props_skip(props, skipped, &cleared_props);
2633         if (!nvlist_empty(cleared_props)) {
2634                 /*
2635                  * Acts on local properties until the dataset has received
2636                  * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2637                  */
2638                 zprop_source_t flags = (ZPROP_SRC_NONE |
2639                     (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2640                 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2641         }
2642         nvlist_free(cleared_props);
2643         return (err);
2644 }
2645 
2646 /*
2647  * inputs:
2648  * zc_name              name of filesystem
2649  * zc_value             name of property to set
2650  * zc_nvlist_src{_size} nvlist of properties to apply
2651  * zc_cookie            received properties flag
2652  *
2653  * outputs:
2654  * zc_nvlist_dst{_size} error for each unapplied received property
2655  */
2656 static int
2657 zfs_ioc_set_prop(zfs_cmd_t *zc)
2658 {
2659         nvlist_t *nvl;
2660         boolean_t received = zc->zc_cookie;
2661         zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2662             ZPROP_SRC_LOCAL);
2663         nvlist_t *errors;
2664         int error;
2665 
2666         if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2667             zc->zc_iflags, &nvl)) != 0)
2668                 return (error);
2669 
2670         if (received) {
2671                 nvlist_t *origprops;
2672 
2673                 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2674                         (void) clear_received_props(zc->zc_name,
2675                             origprops, nvl);
2676                         nvlist_free(origprops);
2677                 }
2678 
2679                 error = dsl_prop_set_hasrecvd(zc->zc_name);
2680         }
2681 
2682         errors = fnvlist_alloc();
2683         if (error == 0)
2684                 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2685 
2686         if (zc->zc_nvlist_dst != NULL && errors != NULL) {
2687                 (void) put_nvlist(zc, errors);
2688         }
2689 
2690         nvlist_free(errors);
2691         nvlist_free(nvl);
2692         return (error);
2693 }
2694 
2695 /*
2696  * inputs:
2697  * zc_name              name of filesystem
2698  * zc_value             name of property to inherit
2699  * zc_cookie            revert to received value if TRUE
2700  *
2701  * outputs:             none
2702  */
2703 static int
2704 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2705 {
2706         const char *propname = zc->zc_value;
2707         zfs_prop_t prop = zfs_name_to_prop(propname);
2708         boolean_t received = zc->zc_cookie;
2709         zprop_source_t source = (received
2710             ? ZPROP_SRC_NONE            /* revert to received value, if any */
2711             : ZPROP_SRC_INHERITED);     /* explicitly inherit */
2712 
2713         if (received) {
2714                 nvlist_t *dummy;
2715                 nvpair_t *pair;
2716                 zprop_type_t type;
2717                 int err;
2718 
2719                 /*
2720                  * zfs_prop_set_special() expects properties in the form of an
2721                  * nvpair with type info.
2722                  */
2723                 if (prop == ZPROP_INVAL) {
2724                         if (!zfs_prop_user(propname))
2725                                 return (SET_ERROR(EINVAL));
2726 
2727                         type = PROP_TYPE_STRING;
2728                 } else if (prop == ZFS_PROP_VOLSIZE ||
2729                     prop == ZFS_PROP_VERSION) {
2730                         return (SET_ERROR(EINVAL));
2731                 } else {
2732                         type = zfs_prop_get_type(prop);
2733                 }
2734 
2735                 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2736 
2737                 switch (type) {
2738                 case PROP_TYPE_STRING:
2739                         VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2740                         break;
2741                 case PROP_TYPE_NUMBER:
2742                 case PROP_TYPE_INDEX:
2743                         VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2744                         break;
2745                 default:
2746                         nvlist_free(dummy);
2747                         return (SET_ERROR(EINVAL));
2748                 }
2749 
2750                 pair = nvlist_next_nvpair(dummy, NULL);
2751                 err = zfs_prop_set_special(zc->zc_name, source, pair);
2752                 nvlist_free(dummy);
2753                 if (err != -1)
2754                         return (err); /* special property already handled */
2755         } else {
2756                 /*
2757                  * Only check this in the non-received case. We want to allow
2758                  * 'inherit -S' to revert non-inheritable properties like quota
2759                  * and reservation to the received or default values even though
2760                  * they are not considered inheritable.
2761                  */
2762                 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2763                         return (SET_ERROR(EINVAL));
2764         }
2765 
2766         /* property name has been validated by zfs_secpolicy_inherit_prop() */
2767         return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source));
2768 }
2769 
2770 static int
2771 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2772 {
2773         nvlist_t *props;
2774         spa_t *spa;
2775         int error;
2776         nvpair_t *pair;
2777 
2778         if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2779             zc->zc_iflags, &props))
2780                 return (error);
2781 
2782         /*
2783          * If the only property is the configfile, then just do a spa_lookup()
2784          * to handle the faulted case.
2785          */
2786         pair = nvlist_next_nvpair(props, NULL);
2787         if (pair != NULL && strcmp(nvpair_name(pair),
2788             zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2789             nvlist_next_nvpair(props, pair) == NULL) {
2790                 mutex_enter(&spa_namespace_lock);
2791                 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2792                         spa_configfile_set(spa, props, B_FALSE);
2793                         spa_config_sync(spa, B_FALSE, B_TRUE);
2794                 }
2795                 mutex_exit(&spa_namespace_lock);
2796                 if (spa != NULL) {
2797                         nvlist_free(props);
2798                         return (0);
2799                 }
2800         }
2801 
2802         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2803                 nvlist_free(props);
2804                 return (error);
2805         }
2806 
2807         error = spa_prop_set(spa, props);
2808 
2809         nvlist_free(props);
2810         spa_close(spa, FTAG);
2811 
2812         return (error);
2813 }
2814 
2815 static int
2816 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2817 {
2818         spa_t *spa;
2819         int error;
2820         nvlist_t *nvp = NULL;
2821 
2822         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2823                 /*
2824                  * If the pool is faulted, there may be properties we can still
2825                  * get (such as altroot and cachefile), so attempt to get them
2826                  * anyway.
2827                  */
2828                 mutex_enter(&spa_namespace_lock);
2829                 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2830                         error = spa_prop_get(spa, &nvp);
2831                 mutex_exit(&spa_namespace_lock);
2832         } else {
2833                 error = spa_prop_get(spa, &nvp);
2834                 spa_close(spa, FTAG);
2835         }
2836 
2837         if (error == 0 && zc->zc_nvlist_dst != NULL)
2838                 error = put_nvlist(zc, nvp);
2839         else
2840                 error = SET_ERROR(EFAULT);
2841 
2842         nvlist_free(nvp);
2843         return (error);
2844 }
2845 
2846 /*
2847  * inputs:
2848  * zc_name              name of filesystem
2849  * zc_nvlist_src{_size} nvlist of delegated permissions
2850  * zc_perm_action       allow/unallow flag
2851  *
2852  * outputs:             none
2853  */
2854 static int
2855 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2856 {
2857         int error;
2858         nvlist_t *fsaclnv = NULL;
2859 
2860         if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2861             zc->zc_iflags, &fsaclnv)) != 0)
2862                 return (error);
2863 
2864         /*
2865          * Verify nvlist is constructed correctly
2866          */
2867         if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2868                 nvlist_free(fsaclnv);
2869                 return (SET_ERROR(EINVAL));
2870         }
2871 
2872         /*
2873          * If we don't have PRIV_SYS_MOUNT, then validate
2874          * that user is allowed to hand out each permission in
2875          * the nvlist(s)
2876          */
2877 
2878         error = secpolicy_zfs(CRED());
2879         if (error != 0) {
2880                 if (zc->zc_perm_action == B_FALSE) {
2881                         error = dsl_deleg_can_allow(zc->zc_name,
2882                             fsaclnv, CRED());
2883                 } else {
2884                         error = dsl_deleg_can_unallow(zc->zc_name,
2885                             fsaclnv, CRED());
2886                 }
2887         }
2888 
2889         if (error == 0)
2890                 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2891 
2892         nvlist_free(fsaclnv);
2893         return (error);
2894 }
2895 
2896 /*
2897  * inputs:
2898  * zc_name              name of filesystem
2899  *
2900  * outputs:
2901  * zc_nvlist_src{_size} nvlist of delegated permissions
2902  */
2903 static int
2904 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
2905 {
2906         nvlist_t *nvp;
2907         int error;
2908 
2909         if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
2910                 error = put_nvlist(zc, nvp);
2911                 nvlist_free(nvp);
2912         }
2913 
2914         return (error);
2915 }
2916 
2917 /*
2918  * Search the vfs list for a specified resource.  Returns a pointer to it
2919  * or NULL if no suitable entry is found. The caller of this routine
2920  * is responsible for releasing the returned vfs pointer.
2921  */
2922 static vfs_t *
2923 zfs_get_vfs(const char *resource)
2924 {
2925         struct vfs *vfsp;
2926         struct vfs *vfs_found = NULL;
2927 
2928         vfs_list_read_lock();
2929         vfsp = rootvfs;
2930         do {
2931                 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
2932                         VFS_HOLD(vfsp);
2933                         vfs_found = vfsp;
2934                         break;
2935                 }
2936                 vfsp = vfsp->vfs_next;
2937         } while (vfsp != rootvfs);
2938         vfs_list_unlock();
2939         return (vfs_found);
2940 }
2941 
2942 /* ARGSUSED */
2943 static void
2944 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
2945 {
2946         zfs_creat_t *zct = arg;
2947 
2948         zfs_create_fs(os, cr, zct->zct_zplprops, tx);
2949 }
2950 
2951 #define ZFS_PROP_UNDEFINED      ((uint64_t)-1)
2952 
2953 /*
2954  * inputs:
2955  * os                   parent objset pointer (NULL if root fs)
2956  * fuids_ok             fuids allowed in this version of the spa?
2957  * sa_ok                SAs allowed in this version of the spa?
2958  * createprops          list of properties requested by creator
2959  *
2960  * outputs:
2961  * zplprops     values for the zplprops we attach to the master node object
2962  * is_ci        true if requested file system will be purely case-insensitive
2963  *
2964  * Determine the settings for utf8only, normalization and
2965  * casesensitivity.  Specific values may have been requested by the
2966  * creator and/or we can inherit values from the parent dataset.  If
2967  * the file system is of too early a vintage, a creator can not
2968  * request settings for these properties, even if the requested
2969  * setting is the default value.  We don't actually want to create dsl
2970  * properties for these, so remove them from the source nvlist after
2971  * processing.
2972  */
2973 static int
2974 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
2975     boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
2976     nvlist_t *zplprops, boolean_t *is_ci)
2977 {
2978         uint64_t sense = ZFS_PROP_UNDEFINED;
2979         uint64_t norm = ZFS_PROP_UNDEFINED;
2980         uint64_t u8 = ZFS_PROP_UNDEFINED;
2981 
2982         ASSERT(zplprops != NULL);
2983 
2984         /*
2985          * Pull out creator prop choices, if any.
2986          */
2987         if (createprops) {
2988                 (void) nvlist_lookup_uint64(createprops,
2989                     zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
2990                 (void) nvlist_lookup_uint64(createprops,
2991                     zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
2992                 (void) nvlist_remove_all(createprops,
2993                     zfs_prop_to_name(ZFS_PROP_NORMALIZE));
2994                 (void) nvlist_lookup_uint64(createprops,
2995                     zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
2996                 (void) nvlist_remove_all(createprops,
2997                     zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
2998                 (void) nvlist_lookup_uint64(createprops,
2999                     zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3000                 (void) nvlist_remove_all(createprops,
3001                     zfs_prop_to_name(ZFS_PROP_CASE));
3002         }
3003 
3004         /*
3005          * If the zpl version requested is whacky or the file system
3006          * or pool is version is too "young" to support normalization
3007          * and the creator tried to set a value for one of the props,
3008          * error out.
3009          */
3010         if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3011             (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3012             (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3013             (zplver < ZPL_VERSION_NORMALIZATION &&
3014             (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3015             sense != ZFS_PROP_UNDEFINED)))
3016                 return (SET_ERROR(ENOTSUP));
3017 
3018         /*
3019          * Put the version in the zplprops
3020          */
3021         VERIFY(nvlist_add_uint64(zplprops,
3022             zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3023 
3024         if (norm == ZFS_PROP_UNDEFINED)
3025                 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
3026         VERIFY(nvlist_add_uint64(zplprops,
3027             zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3028 
3029         /*
3030          * If we're normalizing, names must always be valid UTF-8 strings.
3031          */
3032         if (norm)
3033                 u8 = 1;
3034         if (u8 == ZFS_PROP_UNDEFINED)
3035                 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
3036         VERIFY(nvlist_add_uint64(zplprops,
3037             zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3038 
3039         if (sense == ZFS_PROP_UNDEFINED)
3040                 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
3041         VERIFY(nvlist_add_uint64(zplprops,
3042             zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3043 
3044         if (is_ci)
3045                 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
3046 
3047         return (0);
3048 }
3049 
3050 static int
3051 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3052     nvlist_t *zplprops, boolean_t *is_ci)
3053 {
3054         boolean_t fuids_ok, sa_ok;
3055         uint64_t zplver = ZPL_VERSION;
3056         objset_t *os = NULL;
3057         char parentname[MAXNAMELEN];
3058         char *cp;
3059         spa_t *spa;
3060         uint64_t spa_vers;
3061         int error;
3062 
3063         (void) strlcpy(parentname, dataset, sizeof (parentname));
3064         cp = strrchr(parentname, '/');
3065         ASSERT(cp != NULL);
3066         cp[0] = '\0';
3067 
3068         if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3069                 return (error);
3070 
3071         spa_vers = spa_version(spa);
3072         spa_close(spa, FTAG);
3073 
3074         zplver = zfs_zpl_version_map(spa_vers);
3075         fuids_ok = (zplver >= ZPL_VERSION_FUID);
3076         sa_ok = (zplver >= ZPL_VERSION_SA);
3077 
3078         /*
3079          * Open parent object set so we can inherit zplprop values.
3080          */
3081         if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3082                 return (error);
3083 
3084         error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3085             zplprops, is_ci);
3086         dmu_objset_rele(os, FTAG);
3087         return (error);
3088 }
3089 
3090 static int
3091 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3092     nvlist_t *zplprops, boolean_t *is_ci)
3093 {
3094         boolean_t fuids_ok;
3095         boolean_t sa_ok;
3096         uint64_t zplver = ZPL_VERSION;
3097         int error;
3098 
3099         zplver = zfs_zpl_version_map(spa_vers);
3100         fuids_ok = (zplver >= ZPL_VERSION_FUID);
3101         sa_ok = (zplver >= ZPL_VERSION_SA);
3102 
3103         error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3104             createprops, zplprops, is_ci);
3105         return (error);
3106 }
3107 
3108 /*
3109  * innvl: {
3110  *     "type" -> dmu_objset_type_t (int32)
3111  *     (optional) "props" -> { prop -> value }
3112  * }
3113  *
3114  * outnvl: propname -> error code (int32)
3115  */
3116 static int
3117 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3118 {
3119         int error = 0;
3120         zfs_creat_t zct = { 0 };
3121         nvlist_t *nvprops = NULL;
3122         void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3123         int32_t type32;
3124         dmu_objset_type_t type;
3125         boolean_t is_insensitive = B_FALSE;
3126 
3127         if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3128                 return (SET_ERROR(EINVAL));
3129         type = type32;
3130         (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3131 
3132         switch (type) {
3133         case DMU_OST_ZFS:
3134                 cbfunc = zfs_create_cb;
3135                 break;
3136 
3137         case DMU_OST_ZVOL:
3138                 cbfunc = zvol_create_cb;
3139                 break;
3140 
3141         default:
3142                 cbfunc = NULL;
3143                 break;
3144         }
3145         if (strchr(fsname, '@') ||
3146             strchr(fsname, '%'))
3147                 return (SET_ERROR(EINVAL));
3148 
3149         zct.zct_props = nvprops;
3150 
3151         if (cbfunc == NULL)
3152                 return (SET_ERROR(EINVAL));
3153 
3154         if (type == DMU_OST_ZVOL) {
3155                 uint64_t volsize, volblocksize;
3156 
3157                 if (nvprops == NULL)
3158                         return (SET_ERROR(EINVAL));
3159                 if (nvlist_lookup_uint64(nvprops,
3160                     zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3161                         return (SET_ERROR(EINVAL));
3162 
3163                 if ((error = nvlist_lookup_uint64(nvprops,
3164                     zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3165                     &volblocksize)) != 0 && error != ENOENT)
3166                         return (SET_ERROR(EINVAL));
3167 
3168                 if (error != 0)
3169                         volblocksize = zfs_prop_default_numeric(
3170                             ZFS_PROP_VOLBLOCKSIZE);
3171 
3172                 if ((error = zvol_check_volblocksize(
3173                     volblocksize)) != 0 ||
3174                     (error = zvol_check_volsize(volsize,
3175                     volblocksize)) != 0)
3176                         return (error);
3177         } else if (type == DMU_OST_ZFS) {
3178                 int error;
3179 
3180                 /*
3181                  * We have to have normalization and
3182                  * case-folding flags correct when we do the
3183                  * file system creation, so go figure them out
3184                  * now.
3185                  */
3186                 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3187                     NV_UNIQUE_NAME, KM_SLEEP) == 0);
3188                 error = zfs_fill_zplprops(fsname, nvprops,
3189                     zct.zct_zplprops, &is_insensitive);
3190                 if (error != 0) {
3191                         nvlist_free(zct.zct_zplprops);
3192                         return (error);
3193                 }
3194         }
3195 
3196         error = dmu_objset_create(fsname, type,
3197             is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3198         nvlist_free(zct.zct_zplprops);
3199 
3200         /*
3201          * It would be nice to do this atomically.
3202          */
3203         if (error == 0) {
3204                 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3205                     nvprops, outnvl);
3206                 if (error != 0)
3207                         (void) dsl_destroy_head(fsname);
3208         }
3209         return (error);
3210 }
3211 
3212 /*
3213  * innvl: {
3214  *     "origin" -> name of origin snapshot
3215  *     (optional) "props" -> { prop -> value }
3216  * }
3217  *
3218  * outnvl: propname -> error code (int32)
3219  */
3220 static int
3221 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3222 {
3223         int error = 0;
3224         nvlist_t *nvprops = NULL;
3225         char *origin_name;
3226 
3227         if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3228                 return (SET_ERROR(EINVAL));
3229         (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3230 
3231         if (strchr(fsname, '@') ||
3232             strchr(fsname, '%'))
3233                 return (SET_ERROR(EINVAL));
3234 
3235         if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3236                 return (SET_ERROR(EINVAL));
3237         error = dmu_objset_clone(fsname, origin_name);
3238         if (error != 0)
3239                 return (error);
3240 
3241         /*
3242          * It would be nice to do this atomically.
3243          */
3244         if (error == 0) {
3245                 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3246                     nvprops, outnvl);
3247                 if (error != 0)
3248                         (void) dsl_destroy_head(fsname);
3249         }
3250         return (error);
3251 }
3252 
3253 /*
3254  * innvl: {
3255  *     "snaps" -> { snapshot1, snapshot2 }
3256  *     (optional) "props" -> { prop -> value (string) }
3257  * }
3258  *
3259  * outnvl: snapshot -> error code (int32)
3260  */
3261 static int
3262 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3263 {
3264         nvlist_t *snaps;
3265         nvlist_t *props = NULL;
3266         int error, poollen;
3267         nvpair_t *pair;
3268 
3269         (void) nvlist_lookup_nvlist(innvl, "props", &props);
3270         if ((error = zfs_check_userprops(poolname, props)) != 0)
3271                 return (error);
3272 
3273         if (!nvlist_empty(props) &&
3274             zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3275                 return (SET_ERROR(ENOTSUP));
3276 
3277         if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3278                 return (SET_ERROR(EINVAL));
3279         poollen = strlen(poolname);
3280         for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3281             pair = nvlist_next_nvpair(snaps, pair)) {
3282                 const char *name = nvpair_name(pair);
3283                 const char *cp = strchr(name, '@');
3284 
3285                 /*
3286                  * The snap name must contain an @, and the part after it must
3287                  * contain only valid characters.
3288                  */
3289                 if (cp == NULL || snapshot_namecheck(cp + 1, NULL, NULL) != 0)
3290                         return (SET_ERROR(EINVAL));
3291 
3292                 /*
3293                  * The snap must be in the specified pool.
3294                  */
3295                 if (strncmp(name, poolname, poollen) != 0 ||
3296                     (name[poollen] != '/' && name[poollen] != '@'))
3297                         return (SET_ERROR(EXDEV));
3298 
3299                 /* This must be the only snap of this fs. */
3300                 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3301                     pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3302                         if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3303                             == 0) {
3304                                 return (SET_ERROR(EXDEV));
3305                         }
3306                 }
3307         }
3308 
3309         error = dsl_dataset_snapshot(snaps, props, outnvl);
3310         return (error);
3311 }
3312 
3313 /*
3314  * innvl: "message" -> string
3315  */
3316 /* ARGSUSED */
3317 static int
3318 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3319 {
3320         char *message;
3321         spa_t *spa;
3322         int error;
3323         char *poolname;
3324 
3325         /*
3326          * The poolname in the ioctl is not set, we get it from the TSD,
3327          * which was set at the end of the last successful ioctl that allows
3328          * logging.  The secpolicy func already checked that it is set.
3329          * Only one log ioctl is allowed after each successful ioctl, so
3330          * we clear the TSD here.
3331          */
3332         poolname = tsd_get(zfs_allow_log_key);
3333         (void) tsd_set(zfs_allow_log_key, NULL);
3334         error = spa_open(poolname, &spa, FTAG);
3335         strfree(poolname);
3336         if (error != 0)
3337                 return (error);
3338 
3339         if (nvlist_lookup_string(innvl, "message", &message) != 0)  {
3340                 spa_close(spa, FTAG);
3341                 return (SET_ERROR(EINVAL));
3342         }
3343 
3344         if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3345                 spa_close(spa, FTAG);
3346                 return (SET_ERROR(ENOTSUP));
3347         }
3348 
3349         error = spa_history_log(spa, message);
3350         spa_close(spa, FTAG);
3351         return (error);
3352 }
3353 
3354 /*
3355  * The dp_config_rwlock must not be held when calling this, because the
3356  * unmount may need to write out data.
3357  *
3358  * This function is best-effort.  Callers must deal gracefully if it
3359  * remains mounted (or is remounted after this call).
3360  *
3361  * Returns 0 if the argument is not a snapshot, or it is not currently a
3362  * filesystem, or we were able to unmount it.  Returns error code otherwise.
3363  */
3364 int
3365 zfs_unmount_snap(const char *snapname)
3366 {
3367         vfs_t *vfsp;
3368         zfsvfs_t *zfsvfs;
3369         int err;
3370 
3371         if (strchr(snapname, '@') == NULL)
3372                 return (0);
3373 
3374         vfsp = zfs_get_vfs(snapname);
3375         if (vfsp == NULL)
3376                 return (0);
3377 
3378         zfsvfs = vfsp->vfs_data;
3379         ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os)));
3380 
3381         err = vn_vfswlock(vfsp->vfs_vnodecovered);
3382         VFS_RELE(vfsp);
3383         if (err != 0)
3384                 return (SET_ERROR(err));
3385 
3386         /*
3387          * Always force the unmount for snapshots.
3388          */
3389         (void) dounmount(vfsp, MS_FORCE, kcred);
3390         return (0);
3391 }
3392 
3393 /* ARGSUSED */
3394 static int
3395 zfs_unmount_snap_cb(const char *snapname, void *arg)
3396 {
3397         return (zfs_unmount_snap(snapname));
3398 }
3399 
3400 /*
3401  * When a clone is destroyed, its origin may also need to be destroyed,
3402  * in which case it must be unmounted.  This routine will do that unmount
3403  * if necessary.
3404  */
3405 void
3406 zfs_destroy_unmount_origin(const char *fsname)
3407 {
3408         int error;
3409         objset_t *os;
3410         dsl_dataset_t *ds;
3411 
3412         error = dmu_objset_hold(fsname, FTAG, &os);
3413         if (error != 0)
3414                 return;
3415         ds = dmu_objset_ds(os);
3416         if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3417                 char originname[MAXNAMELEN];
3418                 dsl_dataset_name(ds->ds_prev, originname);
3419                 dmu_objset_rele(os, FTAG);
3420                 (void) zfs_unmount_snap(originname);
3421         } else {
3422                 dmu_objset_rele(os, FTAG);
3423         }
3424 }
3425 
3426 /*
3427  * innvl: {
3428  *     "snaps" -> { snapshot1, snapshot2 }
3429  *     (optional boolean) "defer"
3430  * }
3431  *
3432  * outnvl: snapshot -> error code (int32)
3433  *
3434  */
3435 static int
3436 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3437 {
3438         int error, poollen;
3439         nvlist_t *snaps;
3440         nvpair_t *pair;
3441         boolean_t defer;
3442 
3443         if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3444                 return (SET_ERROR(EINVAL));
3445         defer = nvlist_exists(innvl, "defer");
3446 
3447         poollen = strlen(poolname);
3448         for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3449             pair = nvlist_next_nvpair(snaps, pair)) {
3450                 const char *name = nvpair_name(pair);
3451 
3452                 /*
3453                  * The snap must be in the specified pool.
3454                  */
3455                 if (strncmp(name, poolname, poollen) != 0 ||
3456                     (name[poollen] != '/' && name[poollen] != '@'))
3457                         return (SET_ERROR(EXDEV));
3458 
3459                 error = zfs_unmount_snap(name);
3460                 if (error != 0)
3461                         return (error);
3462         }
3463 
3464         return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3465 }
3466 
3467 /*
3468  * inputs:
3469  * zc_name              name of dataset to destroy
3470  * zc_objset_type       type of objset
3471  * zc_defer_destroy     mark for deferred destroy
3472  *
3473  * outputs:             none
3474  */
3475 static int
3476 zfs_ioc_destroy(zfs_cmd_t *zc)
3477 {
3478         int err;
3479 
3480         if (zc->zc_objset_type == DMU_OST_ZFS) {
3481                 err = zfs_unmount_snap(zc->zc_name);
3482                 if (err != 0)
3483                         return (err);
3484         }
3485 
3486         if (strchr(zc->zc_name, '@'))
3487                 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3488         else
3489                 err = dsl_destroy_head(zc->zc_name);
3490         if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3491                 (void) zvol_remove_minor(zc->zc_name);
3492         return (err);
3493 }
3494 
3495 /*
3496  * fsname is name of dataset to rollback (to most recent snapshot)
3497  *
3498  * innvl is not used.
3499  *
3500  * outnvl: "target" -> name of most recent snapshot
3501  * }
3502  */
3503 /* ARGSUSED */
3504 static int
3505 zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl)
3506 {
3507         zfsvfs_t *zfsvfs;
3508         int error;
3509 
3510         if (getzfsvfs(fsname, &zfsvfs) == 0) {
3511                 error = zfs_suspend_fs(zfsvfs);
3512                 if (error == 0) {
3513                         int resume_err;
3514 
3515                         error = dsl_dataset_rollback(fsname, zfsvfs, outnvl);
3516                         resume_err = zfs_resume_fs(zfsvfs, fsname);
3517                         error = error ? error : resume_err;
3518                 }
3519                 VFS_RELE(zfsvfs->z_vfs);
3520         } else {
3521                 error = dsl_dataset_rollback(fsname, NULL, outnvl);
3522         }
3523         return (error);
3524 }
3525 
3526 static int
3527 recursive_unmount(const char *fsname, void *arg)
3528 {
3529         const char *snapname = arg;
3530         char fullname[MAXNAMELEN];
3531 
3532         (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname);
3533         return (zfs_unmount_snap(fullname));
3534 }
3535 
3536 /*
3537  * inputs:
3538  * zc_name      old name of dataset
3539  * zc_value     new name of dataset
3540  * zc_cookie    recursive flag (only valid for snapshots)
3541  *
3542  * outputs:     none
3543  */
3544 static int
3545 zfs_ioc_rename(zfs_cmd_t *zc)
3546 {
3547         boolean_t recursive = zc->zc_cookie & 1;
3548         char *at;
3549 
3550         zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3551         if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3552             strchr(zc->zc_value, '%'))
3553                 return (SET_ERROR(EINVAL));
3554 
3555         at = strchr(zc->zc_name, '@');
3556         if (at != NULL) {
3557                 /* snaps must be in same fs */
3558                 int error;
3559 
3560                 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
3561                         return (SET_ERROR(EXDEV));
3562                 *at = '\0';
3563                 if (zc->zc_objset_type == DMU_OST_ZFS) {
3564                         error = dmu_objset_find(zc->zc_name,
3565                             recursive_unmount, at + 1,
3566                             recursive ? DS_FIND_CHILDREN : 0);
3567                         if (error != 0) {
3568                                 *at = '@';
3569                                 return (error);
3570                         }
3571                 }
3572                 error = dsl_dataset_rename_snapshot(zc->zc_name,
3573                     at + 1, strchr(zc->zc_value, '@') + 1, recursive);
3574                 *at = '@';
3575 
3576                 return (error);
3577         } else {
3578                 if (zc->zc_objset_type == DMU_OST_ZVOL)
3579                         (void) zvol_remove_minor(zc->zc_name);
3580                 return (dsl_dir_rename(zc->zc_name, zc->zc_value));
3581         }
3582 }
3583 
3584 static int
3585 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3586 {
3587         const char *propname = nvpair_name(pair);
3588         boolean_t issnap = (strchr(dsname, '@') != NULL);
3589         zfs_prop_t prop = zfs_name_to_prop(propname);
3590         uint64_t intval;
3591         int err;
3592 
3593         if (prop == ZPROP_INVAL) {
3594                 if (zfs_prop_user(propname)) {
3595                         if (err = zfs_secpolicy_write_perms(dsname,
3596                             ZFS_DELEG_PERM_USERPROP, cr))
3597                                 return (err);
3598                         return (0);
3599                 }
3600 
3601                 if (!issnap && zfs_prop_userquota(propname)) {
3602                         const char *perm = NULL;
3603                         const char *uq_prefix =
3604                             zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3605                         const char *gq_prefix =
3606                             zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3607 
3608                         if (strncmp(propname, uq_prefix,
3609                             strlen(uq_prefix)) == 0) {
3610                                 perm = ZFS_DELEG_PERM_USERQUOTA;
3611                         } else if (strncmp(propname, gq_prefix,
3612                             strlen(gq_prefix)) == 0) {
3613                                 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3614                         } else {
3615                                 /* USERUSED and GROUPUSED are read-only */
3616                                 return (SET_ERROR(EINVAL));
3617                         }
3618 
3619                         if (err = zfs_secpolicy_write_perms(dsname, perm, cr))
3620                                 return (err);
3621                         return (0);
3622                 }
3623 
3624                 return (SET_ERROR(EINVAL));
3625         }
3626 
3627         if (issnap)
3628                 return (SET_ERROR(EINVAL));
3629 
3630         if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3631                 /*
3632                  * dsl_prop_get_all_impl() returns properties in this
3633                  * format.
3634                  */
3635                 nvlist_t *attrs;
3636                 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3637                 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3638                     &pair) == 0);
3639         }
3640 
3641         /*
3642          * Check that this value is valid for this pool version
3643          */
3644         switch (prop) {
3645         case ZFS_PROP_COMPRESSION:
3646                 /*
3647                  * If the user specified gzip compression, make sure
3648                  * the SPA supports it. We ignore any errors here since
3649                  * we'll catch them later.
3650                  */
3651                 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3652                     nvpair_value_uint64(pair, &intval) == 0) {
3653                         if (intval >= ZIO_COMPRESS_GZIP_1 &&
3654                             intval <= ZIO_COMPRESS_GZIP_9 &&
3655                             zfs_earlier_version(dsname,
3656                             SPA_VERSION_GZIP_COMPRESSION)) {
3657                                 return (SET_ERROR(ENOTSUP));
3658                         }
3659 
3660                         if (intval == ZIO_COMPRESS_ZLE &&
3661                             zfs_earlier_version(dsname,
3662                             SPA_VERSION_ZLE_COMPRESSION))
3663                                 return (SET_ERROR(ENOTSUP));
3664 
3665                         if (intval == ZIO_COMPRESS_LZ4) {
3666                                 zfeature_info_t *feature =
3667                                     &spa_feature_table[
3668                                     SPA_FEATURE_LZ4_COMPRESS];
3669                                 spa_t *spa;
3670 
3671                                 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3672                                         return (err);
3673 
3674                                 if (!spa_feature_is_enabled(spa, feature)) {
3675                                         spa_close(spa, FTAG);
3676                                         return (SET_ERROR(ENOTSUP));
3677                                 }
3678                                 spa_close(spa, FTAG);
3679                         }
3680 
3681                         /*
3682                          * If this is a bootable dataset then
3683                          * verify that the compression algorithm
3684                          * is supported for booting. We must return
3685                          * something other than ENOTSUP since it
3686                          * implies a downrev pool version.
3687                          */
3688                         if (zfs_is_bootfs(dsname) &&
3689                             !BOOTFS_COMPRESS_VALID(intval)) {
3690                                 return (SET_ERROR(ERANGE));
3691                         }
3692                 }
3693                 break;
3694 
3695         case ZFS_PROP_COPIES:
3696                 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3697                         return (SET_ERROR(ENOTSUP));
3698                 break;
3699 
3700         case ZFS_PROP_DEDUP:
3701                 if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3702                         return (SET_ERROR(ENOTSUP));
3703                 break;
3704 
3705         case ZFS_PROP_SHARESMB:
3706                 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3707                         return (SET_ERROR(ENOTSUP));
3708                 break;
3709 
3710         case ZFS_PROP_ACLINHERIT:
3711                 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3712                     nvpair_value_uint64(pair, &intval) == 0) {
3713                         if (intval == ZFS_ACL_PASSTHROUGH_X &&
3714                             zfs_earlier_version(dsname,
3715                             SPA_VERSION_PASSTHROUGH_X))
3716                                 return (SET_ERROR(ENOTSUP));
3717                 }
3718                 break;
3719         }
3720 
3721         return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3722 }
3723 
3724 /*
3725  * Checks for a race condition to make sure we don't increment a feature flag
3726  * multiple times.
3727  */
3728 static int
3729 zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx)
3730 {
3731         spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3732         zfeature_info_t *feature = arg;
3733 
3734         if (!spa_feature_is_active(spa, feature))
3735                 return (0);
3736         else
3737                 return (SET_ERROR(EBUSY));
3738 }
3739 
3740 /*
3741  * The callback invoked on feature activation in the sync task caused by
3742  * zfs_prop_activate_feature.
3743  */
3744 static void
3745 zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx)
3746 {
3747         spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3748         zfeature_info_t *feature = arg;
3749 
3750         spa_feature_incr(spa, feature, tx);
3751 }
3752 
3753 /*
3754  * Activates a feature on a pool in response to a property setting. This
3755  * creates a new sync task which modifies the pool to reflect the feature
3756  * as being active.
3757  */
3758 static int
3759 zfs_prop_activate_feature(spa_t *spa, zfeature_info_t *feature)
3760 {
3761         int err;
3762 
3763         /* EBUSY here indicates that the feature is already active */
3764         err = dsl_sync_task(spa_name(spa),
3765             zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync,
3766             feature, 2);
3767 
3768         if (err != 0 && err != EBUSY)
3769                 return (err);
3770         else
3771                 return (0);
3772 }
3773 
3774 /*
3775  * Removes properties from the given props list that fail permission checks
3776  * needed to clear them and to restore them in case of a receive error. For each
3777  * property, make sure we have both set and inherit permissions.
3778  *
3779  * Returns the first error encountered if any permission checks fail. If the
3780  * caller provides a non-NULL errlist, it also gives the complete list of names
3781  * of all the properties that failed a permission check along with the
3782  * corresponding error numbers. The caller is responsible for freeing the
3783  * returned errlist.
3784  *
3785  * If every property checks out successfully, zero is returned and the list
3786  * pointed at by errlist is NULL.
3787  */
3788 static int
3789 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
3790 {
3791         zfs_cmd_t *zc;
3792         nvpair_t *pair, *next_pair;
3793         nvlist_t *errors;
3794         int err, rv = 0;
3795 
3796         if (props == NULL)
3797                 return (0);
3798 
3799         VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3800 
3801         zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
3802         (void) strcpy(zc->zc_name, dataset);
3803         pair = nvlist_next_nvpair(props, NULL);
3804         while (pair != NULL) {
3805                 next_pair = nvlist_next_nvpair(props, pair);
3806 
3807                 (void) strcpy(zc->zc_value, nvpair_name(pair));
3808                 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
3809                     (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
3810                         VERIFY(nvlist_remove_nvpair(props, pair) == 0);
3811                         VERIFY(nvlist_add_int32(errors,
3812                             zc->zc_value, err) == 0);
3813                 }
3814                 pair = next_pair;
3815         }
3816         kmem_free(zc, sizeof (zfs_cmd_t));
3817 
3818         if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
3819                 nvlist_free(errors);
3820                 errors = NULL;
3821         } else {
3822                 VERIFY(nvpair_value_int32(pair, &rv) == 0);
3823         }
3824 
3825         if (errlist == NULL)
3826                 nvlist_free(errors);
3827         else
3828                 *errlist = errors;
3829 
3830         return (rv);
3831 }
3832 
3833 static boolean_t
3834 propval_equals(nvpair_t *p1, nvpair_t *p2)
3835 {
3836         if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
3837                 /* dsl_prop_get_all_impl() format */
3838                 nvlist_t *attrs;
3839                 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
3840                 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3841                     &p1) == 0);
3842         }
3843 
3844         if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
3845                 nvlist_t *attrs;
3846                 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
3847                 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3848                     &p2) == 0);
3849         }
3850 
3851         if (nvpair_type(p1) != nvpair_type(p2))
3852                 return (B_FALSE);
3853 
3854         if (nvpair_type(p1) == DATA_TYPE_STRING) {
3855                 char *valstr1, *valstr2;
3856 
3857                 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
3858                 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
3859                 return (strcmp(valstr1, valstr2) == 0);
3860         } else {
3861                 uint64_t intval1, intval2;
3862 
3863                 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
3864                 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
3865                 return (intval1 == intval2);
3866         }
3867 }
3868 
3869 /*
3870  * Remove properties from props if they are not going to change (as determined
3871  * by comparison with origprops). Remove them from origprops as well, since we
3872  * do not need to clear or restore properties that won't change.
3873  */
3874 static void
3875 props_reduce(nvlist_t *props, nvlist_t *origprops)
3876 {
3877         nvpair_t *pair, *next_pair;
3878 
3879         if (origprops == NULL)
3880                 return; /* all props need to be received */
3881 
3882         pair = nvlist_next_nvpair(props, NULL);
3883         while (pair != NULL) {
3884                 const char *propname = nvpair_name(pair);
3885                 nvpair_t *match;
3886 
3887                 next_pair = nvlist_next_nvpair(props, pair);
3888 
3889                 if ((nvlist_lookup_nvpair(origprops, propname,
3890                     &match) != 0) || !propval_equals(pair, match))
3891                         goto next; /* need to set received value */
3892 
3893                 /* don't clear the existing received value */
3894                 (void) nvlist_remove_nvpair(origprops, match);
3895                 /* don't bother receiving the property */
3896                 (void) nvlist_remove_nvpair(props, pair);
3897 next:
3898                 pair = next_pair;
3899         }
3900 }
3901 
3902 #ifdef  DEBUG
3903 static boolean_t zfs_ioc_recv_inject_err;
3904 #endif
3905 
3906 /*
3907  * inputs:
3908  * zc_name              name of containing filesystem
3909  * zc_nvlist_src{_size} nvlist of properties to apply
3910  * zc_value             name of snapshot to create
3911  * zc_string            name of clone origin (if DRR_FLAG_CLONE)
3912  * zc_cookie            file descriptor to recv from
3913  * zc_begin_record      the BEGIN record of the stream (not byteswapped)
3914  * zc_guid              force flag
3915  * zc_cleanup_fd        cleanup-on-exit file descriptor
3916  * zc_action_handle     handle for this guid/ds mapping (or zero on first call)
3917  *
3918  * outputs:
3919  * zc_cookie            number of bytes read
3920  * zc_nvlist_dst{_size} error for each unapplied received property
3921  * zc_obj               zprop_errflags_t
3922  * zc_action_handle     handle for this guid/ds mapping
3923  */
3924 static int
3925 zfs_ioc_recv(zfs_cmd_t *zc)
3926 {
3927         file_t *fp;
3928         dmu_recv_cookie_t drc;
3929         boolean_t force = (boolean_t)zc->zc_guid;
3930         int fd;
3931         int error = 0;
3932         int props_error = 0;
3933         nvlist_t *errors;
3934         offset_t off;
3935         nvlist_t *props = NULL; /* sent properties */
3936         nvlist_t *origprops = NULL; /* existing properties */
3937         char *origin = NULL;
3938         char *tosnap;
3939         char tofs[ZFS_MAXNAMELEN];
3940         boolean_t first_recvd_props = B_FALSE;
3941 
3942         if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3943             strchr(zc->zc_value, '@') == NULL ||
3944             strchr(zc->zc_value, '%'))
3945                 return (SET_ERROR(EINVAL));
3946 
3947         (void) strcpy(tofs, zc->zc_value);
3948         tosnap = strchr(tofs, '@');
3949         *tosnap++ = '\0';
3950 
3951         if (zc->zc_nvlist_src != NULL &&
3952             (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
3953             zc->zc_iflags, &props)) != 0)
3954                 return (error);
3955 
3956         fd = zc->zc_cookie;
3957         fp = getf(fd);
3958         if (fp == NULL) {
3959                 nvlist_free(props);
3960                 return (SET_ERROR(EBADF));
3961         }
3962 
3963         VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3964 
3965         if (zc->zc_string[0])
3966                 origin = zc->zc_string;
3967 
3968         error = dmu_recv_begin(tofs, tosnap,
3969             &zc->zc_begin_record, force, origin, &drc);
3970         if (error != 0)
3971                 goto out;
3972 
3973         /*
3974          * Set properties before we receive the stream so that they are applied
3975          * to the new data. Note that we must call dmu_recv_stream() if
3976          * dmu_recv_begin() succeeds.
3977          */
3978         if (props != NULL && !drc.drc_newfs) {
3979                 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
3980                     SPA_VERSION_RECVD_PROPS &&
3981                     !dsl_prop_get_hasrecvd(tofs))
3982                         first_recvd_props = B_TRUE;
3983 
3984                 /*
3985                  * If new received properties are supplied, they are to
3986                  * completely replace the existing received properties, so stash
3987                  * away the existing ones.
3988                  */
3989                 if (dsl_prop_get_received(tofs, &origprops) == 0) {
3990                         nvlist_t *errlist = NULL;
3991                         /*
3992                          * Don't bother writing a property if its value won't
3993                          * change (and avoid the unnecessary security checks).
3994                          *
3995                          * The first receive after SPA_VERSION_RECVD_PROPS is a
3996                          * special case where we blow away all local properties
3997                          * regardless.
3998                          */
3999                         if (!first_recvd_props)
4000                                 props_reduce(props, origprops);
4001                         if (zfs_check_clearable(tofs, origprops, &errlist) != 0)
4002                                 (void) nvlist_merge(errors, errlist, 0);
4003                         nvlist_free(errlist);
4004 
4005                         if (clear_received_props(tofs, origprops,
4006                             first_recvd_props ? NULL : props) != 0)
4007                                 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4008                 } else {
4009                         zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4010                 }
4011         }
4012 
4013         if (props != NULL) {
4014                 props_error = dsl_prop_set_hasrecvd(tofs);
4015 
4016                 if (props_error == 0) {
4017                         (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4018                             props, errors);
4019                 }
4020         }
4021 
4022         if (zc->zc_nvlist_dst_size != 0 &&
4023             (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4024             put_nvlist(zc, errors) != 0)) {
4025                 /*
4026                  * Caller made zc->zc_nvlist_dst less than the minimum expected
4027                  * size or supplied an invalid address.
4028                  */
4029                 props_error = SET_ERROR(EINVAL);
4030         }
4031 
4032         off = fp->f_offset;
4033         error = dmu_recv_stream(&drc, fp->f_vnode, &off, zc->zc_cleanup_fd,
4034             &zc->zc_action_handle);
4035 
4036         if (error == 0) {
4037                 zfsvfs_t *zfsvfs = NULL;
4038 
4039                 if (getzfsvfs(tofs, &zfsvfs) == 0) {
4040                         /* online recv */
4041                         int end_err;
4042 
4043                         error = zfs_suspend_fs(zfsvfs);
4044                         /*
4045                          * If the suspend fails, then the recv_end will
4046                          * likely also fail, and clean up after itself.
4047                          */
4048                         end_err = dmu_recv_end(&drc, zfsvfs);
4049                         if (error == 0)
4050                                 error = zfs_resume_fs(zfsvfs, tofs);
4051                         error = error ? error : end_err;
4052                         VFS_RELE(zfsvfs->z_vfs);
4053                 } else {
4054                         error = dmu_recv_end(&drc, NULL);
4055                 }
4056         }
4057 
4058         zc->zc_cookie = off - fp->f_offset;
4059         if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4060                 fp->f_offset = off;
4061 
4062 #ifdef  DEBUG
4063         if (zfs_ioc_recv_inject_err) {
4064                 zfs_ioc_recv_inject_err = B_FALSE;
4065                 error = 1;
4066         }
4067 #endif
4068         /*
4069          * On error, restore the original props.
4070          */
4071         if (error != 0 && props != NULL && !drc.drc_newfs) {
4072                 if (clear_received_props(tofs, props, NULL) != 0) {
4073                         /*
4074                          * We failed to clear the received properties.
4075                          * Since we may have left a $recvd value on the
4076                          * system, we can't clear the $hasrecvd flag.
4077                          */
4078                         zc->zc_obj |= ZPROP_ERR_NORESTORE;
4079                 } else if (first_recvd_props) {
4080                         dsl_prop_unset_hasrecvd(tofs);
4081                 }
4082 
4083                 if (origprops == NULL && !drc.drc_newfs) {
4084                         /* We failed to stash the original properties. */
4085                         zc->zc_obj |= ZPROP_ERR_NORESTORE;
4086                 }
4087 
4088                 /*
4089                  * dsl_props_set() will not convert RECEIVED to LOCAL on or
4090                  * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4091                  * explictly if we're restoring local properties cleared in the
4092                  * first new-style receive.
4093                  */
4094                 if (origprops != NULL &&
4095                     zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4096                     ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4097                     origprops, NULL) != 0) {
4098                         /*
4099                          * We stashed the original properties but failed to
4100                          * restore them.
4101                          */
4102                         zc->zc_obj |= ZPROP_ERR_NORESTORE;
4103                 }
4104         }
4105 out:
4106         nvlist_free(props);
4107         nvlist_free(origprops);
4108         nvlist_free(errors);
4109         releasef(fd);
4110 
4111         if (error == 0)
4112                 error = props_error;
4113 
4114         return (error);
4115 }
4116 
4117 /*
4118  * inputs:
4119  * zc_name      name of snapshot to send
4120  * zc_cookie    file descriptor to send stream to
4121  * zc_obj       fromorigin flag (mutually exclusive with zc_fromobj)
4122  * zc_sendobj   objsetid of snapshot to send
4123  * zc_fromobj   objsetid of incremental fromsnap (may be zero)
4124  * zc_guid      if set, estimate size of stream only.  zc_cookie is ignored.
4125  *              output size in zc_objset_type.
4126  *
4127  * outputs: none
4128  */
4129 static int
4130 zfs_ioc_send(zfs_cmd_t *zc)
4131 {
4132         int error;
4133         offset_t off;
4134         boolean_t estimate = (zc->zc_guid != 0);
4135 
4136         if (zc->zc_obj != 0) {
4137                 dsl_pool_t *dp;
4138                 dsl_dataset_t *tosnap;
4139 
4140                 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4141                 if (error != 0)
4142                         return (error);
4143 
4144                 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4145                 if (error != 0) {
4146                         dsl_pool_rele(dp, FTAG);
4147                         return (error);
4148                 }
4149 
4150                 if (dsl_dir_is_clone(tosnap->ds_dir))
4151                         zc->zc_fromobj = tosnap->ds_dir->dd_phys->dd_origin_obj;
4152                 dsl_dataset_rele(tosnap, FTAG);
4153                 dsl_pool_rele(dp, FTAG);
4154         }
4155 
4156         if (estimate) {
4157                 dsl_pool_t *dp;
4158                 dsl_dataset_t *tosnap;
4159                 dsl_dataset_t *fromsnap = NULL;
4160 
4161                 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4162                 if (error != 0)
4163                         return (error);
4164 
4165                 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4166                 if (error != 0) {
4167                         dsl_pool_rele(dp, FTAG);
4168                         return (error);
4169                 }
4170 
4171                 if (zc->zc_fromobj != 0) {
4172                         error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
4173                             FTAG, &fromsnap);
4174                         if (error != 0) {
4175                                 dsl_dataset_rele(tosnap, FTAG);
4176                                 dsl_pool_rele(dp, FTAG);
4177                                 return (error);
4178                         }
4179                 }
4180 
4181                 error = dmu_send_estimate(tosnap, fromsnap,
4182                     &zc->zc_objset_type);
4183 
4184                 if (fromsnap != NULL)
4185                         dsl_dataset_rele(fromsnap, FTAG);
4186                 dsl_dataset_rele(tosnap, FTAG);
4187                 dsl_pool_rele(dp, FTAG);
4188         } else {
4189                 file_t *fp = getf(zc->zc_cookie);
4190                 if (fp == NULL)
4191                         return (SET_ERROR(EBADF));
4192 
4193                 off = fp->f_offset;
4194                 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
4195                     zc->zc_fromobj, zc->zc_cookie, fp->f_vnode, &off);
4196 
4197                 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4198                         fp->f_offset = off;
4199                 releasef(zc->zc_cookie);
4200         }
4201         return (error);
4202 }
4203 
4204 /*
4205  * inputs:
4206  * zc_name      name of snapshot on which to report progress
4207  * zc_cookie    file descriptor of send stream
4208  *
4209  * outputs:
4210  * zc_cookie    number of bytes written in send stream thus far
4211  */
4212 static int
4213 zfs_ioc_send_progress(zfs_cmd_t *zc)
4214 {
4215         dsl_pool_t *dp;
4216         dsl_dataset_t *ds;
4217         dmu_sendarg_t *dsp = NULL;
4218         int error;
4219 
4220         error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4221         if (error != 0)
4222                 return (error);
4223 
4224         error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4225         if (error != 0) {
4226                 dsl_pool_rele(dp, FTAG);
4227                 return (error);
4228         }
4229 
4230         mutex_enter(&ds->ds_sendstream_lock);
4231 
4232         /*
4233          * Iterate over all the send streams currently active on this dataset.
4234          * If there's one which matches the specified file descriptor _and_ the
4235          * stream was started by the current process, return the progress of
4236          * that stream.
4237          */
4238         for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4239             dsp = list_next(&ds->ds_sendstreams, dsp)) {
4240                 if (dsp->dsa_outfd == zc->zc_cookie &&
4241                     dsp->dsa_proc == curproc)
4242                         break;
4243         }
4244 
4245         if (dsp != NULL)
4246                 zc->zc_cookie = *(dsp->dsa_off);
4247         else
4248                 error = SET_ERROR(ENOENT);
4249 
4250         mutex_exit(&ds->ds_sendstream_lock);
4251         dsl_dataset_rele(ds, FTAG);
4252         dsl_pool_rele(dp, FTAG);
4253         return (error);
4254 }
4255 
4256 static int
4257 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4258 {
4259         int id, error;
4260 
4261         error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4262             &zc->zc_inject_record);
4263 
4264         if (error == 0)
4265                 zc->zc_guid = (uint64_t)id;
4266 
4267         return (error);
4268 }
4269 
4270 static int
4271 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4272 {
4273         return (zio_clear_fault((int)zc->zc_guid));
4274 }
4275 
4276 static int
4277 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4278 {
4279         int id = (int)zc->zc_guid;
4280         int error;
4281 
4282         error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4283             &zc->zc_inject_record);
4284 
4285         zc->zc_guid = id;
4286 
4287         return (error);
4288 }
4289 
4290 static int
4291 zfs_ioc_error_log(zfs_cmd_t *zc)
4292 {
4293         spa_t *spa;
4294         int error;
4295         size_t count = (size_t)zc->zc_nvlist_dst_size;
4296 
4297         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4298                 return (error);
4299 
4300         error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4301             &count);
4302         if (error == 0)
4303                 zc->zc_nvlist_dst_size = count;
4304         else
4305                 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4306 
4307         spa_close(spa, FTAG);
4308 
4309         return (error);
4310 }
4311 
4312 static int
4313 zfs_ioc_clear(zfs_cmd_t *zc)
4314 {
4315         spa_t *spa;
4316         vdev_t *vd;
4317         int error;
4318 
4319         /*
4320          * On zpool clear we also fix up missing slogs
4321          */
4322         mutex_enter(&spa_namespace_lock);
4323         spa = spa_lookup(zc->zc_name);
4324         if (spa == NULL) {
4325                 mutex_exit(&spa_namespace_lock);
4326                 return (SET_ERROR(EIO));
4327         }
4328         if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4329                 /* we need to let spa_open/spa_load clear the chains */
4330                 spa_set_log_state(spa, SPA_LOG_CLEAR);
4331         }
4332         spa->spa_last_open_failed = 0;
4333         mutex_exit(&spa_namespace_lock);
4334 
4335         if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4336                 error = spa_open(zc->zc_name, &spa, FTAG);
4337         } else {
4338                 nvlist_t *policy;
4339                 nvlist_t *config = NULL;
4340 
4341                 if (zc->zc_nvlist_src == NULL)
4342                         return (SET_ERROR(EINVAL));
4343 
4344                 if ((error = get_nvlist(zc->zc_nvlist_src,
4345                     zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4346                         error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4347                             policy, &config);
4348                         if (config != NULL) {
4349                                 int err;
4350 
4351                                 if ((err = put_nvlist(zc, config)) != 0)
4352                                         error = err;
4353                                 nvlist_free(config);
4354                         }
4355                         nvlist_free(policy);
4356                 }
4357         }
4358 
4359         if (error != 0)
4360                 return (error);
4361 
4362         spa_vdev_state_enter(spa, SCL_NONE);
4363 
4364         if (zc->zc_guid == 0) {
4365                 vd = NULL;
4366         } else {
4367                 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4368                 if (vd == NULL) {
4369                         (void) spa_vdev_state_exit(spa, NULL, ENODEV);
4370                         spa_close(spa, FTAG);
4371                         return (SET_ERROR(ENODEV));
4372                 }
4373         }
4374 
4375         vdev_clear(spa, vd);
4376 
4377         (void) spa_vdev_state_exit(spa, NULL, 0);
4378 
4379         /*
4380          * Resume any suspended I/Os.
4381          */
4382         if (zio_resume(spa) != 0)
4383                 error = SET_ERROR(EIO);
4384 
4385         spa_close(spa, FTAG);
4386 
4387         return (error);
4388 }
4389 
4390 static int
4391 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4392 {
4393         spa_t *spa;
4394         int error;
4395 
4396         error = spa_open(zc->zc_name, &spa, FTAG);
4397         if (error != 0)
4398                 return (error);
4399 
4400         spa_vdev_state_enter(spa, SCL_NONE);
4401 
4402         /*
4403          * If a resilver is already in progress then set the
4404          * spa_scrub_reopen flag to B_TRUE so that we don't restart
4405          * the scan as a side effect of the reopen. Otherwise, let
4406          * vdev_open() decided if a resilver is required.
4407          */
4408         spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4409         vdev_reopen(spa->spa_root_vdev);
4410         spa->spa_scrub_reopen = B_FALSE;
4411 
4412         (void) spa_vdev_state_exit(spa, NULL, 0);
4413         spa_close(spa, FTAG);
4414         return (0);
4415 }
4416 /*
4417  * inputs:
4418  * zc_name      name of filesystem
4419  * zc_value     name of origin snapshot
4420  *
4421  * outputs:
4422  * zc_string    name of conflicting snapshot, if there is one
4423  */
4424 static int
4425 zfs_ioc_promote(zfs_cmd_t *zc)
4426 {
4427         char *cp;
4428 
4429         /*
4430          * We don't need to unmount *all* the origin fs's snapshots, but
4431          * it's easier.
4432          */
4433         cp = strchr(zc->zc_value, '@');
4434         if (cp)
4435                 *cp = '\0';
4436         (void) dmu_objset_find(zc->zc_value,
4437             zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
4438         return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4439 }
4440 
4441 /*
4442  * Retrieve a single {user|group}{used|quota}@... property.
4443  *
4444  * inputs:
4445  * zc_name      name of filesystem
4446  * zc_objset_type zfs_userquota_prop_t
4447  * zc_value     domain name (eg. "S-1-234-567-89")
4448  * zc_guid      RID/UID/GID
4449  *
4450  * outputs:
4451  * zc_cookie    property value
4452  */
4453 static int
4454 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4455 {
4456         zfsvfs_t *zfsvfs;
4457         int error;
4458 
4459         if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4460                 return (SET_ERROR(EINVAL));
4461 
4462         error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4463         if (error != 0)
4464                 return (error);
4465 
4466         error = zfs_userspace_one(zfsvfs,
4467             zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4468         zfsvfs_rele(zfsvfs, FTAG);
4469 
4470         return (error);
4471 }
4472 
4473 /*
4474  * inputs:
4475  * zc_name              name of filesystem
4476  * zc_cookie            zap cursor
4477  * zc_objset_type       zfs_userquota_prop_t
4478  * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4479  *
4480  * outputs:
4481  * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
4482  * zc_cookie    zap cursor
4483  */
4484 static int
4485 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4486 {
4487         zfsvfs_t *zfsvfs;
4488         int bufsize = zc->zc_nvlist_dst_size;
4489 
4490         if (bufsize <= 0)
4491                 return (SET_ERROR(ENOMEM));
4492 
4493         int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4494         if (error != 0)
4495                 return (error);
4496 
4497         void *buf = kmem_alloc(bufsize, KM_SLEEP);
4498 
4499         error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
4500             buf, &zc->zc_nvlist_dst_size);
4501 
4502         if (error == 0) {
4503                 error = xcopyout(buf,
4504                     (void *)(uintptr_t)zc->zc_nvlist_dst,
4505                     zc->zc_nvlist_dst_size);
4506         }
4507         kmem_free(buf, bufsize);
4508         zfsvfs_rele(zfsvfs, FTAG);
4509 
4510         return (error);
4511 }
4512 
4513 /*
4514  * inputs:
4515  * zc_name              name of filesystem
4516  *
4517  * outputs:
4518  * none
4519  */
4520 static int
4521 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4522 {
4523         objset_t *os;
4524         int error = 0;
4525         zfsvfs_t *zfsvfs;
4526 
4527         if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
4528                 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
4529                         /*
4530                          * If userused is not enabled, it may be because the
4531                          * objset needs to be closed & reopened (to grow the
4532                          * objset_phys_t).  Suspend/resume the fs will do that.
4533                          */
4534                         error = zfs_suspend_fs(zfsvfs);
4535                         if (error == 0) {
4536                                 dmu_objset_refresh_ownership(zfsvfs->z_os,
4537                                     zfsvfs);
4538                                 error = zfs_resume_fs(zfsvfs, zc->zc_name);
4539                         }
4540                 }
4541                 if (error == 0)
4542                         error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
4543                 VFS_RELE(zfsvfs->z_vfs);
4544         } else {
4545                 /* XXX kind of reading contents without owning */
4546                 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4547                 if (error != 0)
4548                         return (error);
4549 
4550                 error = dmu_objset_userspace_upgrade(os);
4551                 dmu_objset_rele(os, FTAG);
4552         }
4553 
4554         return (error);
4555 }
4556 
4557 /*
4558  * We don't want to have a hard dependency
4559  * against some special symbols in sharefs
4560  * nfs, and smbsrv.  Determine them if needed when
4561  * the first file system is shared.
4562  * Neither sharefs, nfs or smbsrv are unloadable modules.
4563  */
4564 int (*znfsexport_fs)(void *arg);
4565 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
4566 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
4567 
4568 int zfs_nfsshare_inited;
4569 int zfs_smbshare_inited;
4570 
4571 ddi_modhandle_t nfs_mod;
4572 ddi_modhandle_t sharefs_mod;
4573 ddi_modhandle_t smbsrv_mod;
4574 kmutex_t zfs_share_lock;
4575 
4576 static int
4577 zfs_init_sharefs()
4578 {
4579         int error;
4580 
4581         ASSERT(MUTEX_HELD(&zfs_share_lock));
4582         /* Both NFS and SMB shares also require sharetab support. */
4583         if (sharefs_mod == NULL && ((sharefs_mod =
4584             ddi_modopen("fs/sharefs",
4585             KRTLD_MODE_FIRST, &error)) == NULL)) {
4586                 return (SET_ERROR(ENOSYS));
4587         }
4588         if (zshare_fs == NULL && ((zshare_fs =
4589             (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
4590             ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
4591                 return (SET_ERROR(ENOSYS));
4592         }
4593         return (0);
4594 }
4595 
4596 static int
4597 zfs_ioc_share(zfs_cmd_t *zc)
4598 {
4599         int error;
4600         int opcode;
4601 
4602         switch (zc->zc_share.z_sharetype) {
4603         case ZFS_SHARE_NFS:
4604         case ZFS_UNSHARE_NFS:
4605                 if (zfs_nfsshare_inited == 0) {
4606                         mutex_enter(&zfs_share_lock);
4607                         if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
4608                             KRTLD_MODE_FIRST, &error)) == NULL)) {
4609                                 mutex_exit(&zfs_share_lock);
4610                                 return (SET_ERROR(ENOSYS));
4611                         }
4612                         if (znfsexport_fs == NULL &&
4613                             ((znfsexport_fs = (int (*)(void *))
4614                             ddi_modsym(nfs_mod,
4615                             "nfs_export", &error)) == NULL)) {
4616                                 mutex_exit(&zfs_share_lock);
4617                                 return (SET_ERROR(ENOSYS));
4618                         }
4619                         error = zfs_init_sharefs();
4620                         if (error != 0) {
4621                                 mutex_exit(&zfs_share_lock);
4622                                 return (SET_ERROR(ENOSYS));
4623                         }
4624                         zfs_nfsshare_inited = 1;
4625                         mutex_exit(&zfs_share_lock);
4626                 }
4627                 break;
4628         case ZFS_SHARE_SMB:
4629         case ZFS_UNSHARE_SMB:
4630                 if (zfs_smbshare_inited == 0) {
4631                         mutex_enter(&zfs_share_lock);
4632                         if (smbsrv_mod == NULL && ((smbsrv_mod =
4633                             ddi_modopen("drv/smbsrv",
4634                             KRTLD_MODE_FIRST, &error)) == NULL)) {
4635                                 mutex_exit(&zfs_share_lock);
4636                                 return (SET_ERROR(ENOSYS));
4637                         }
4638                         if (zsmbexport_fs == NULL && ((zsmbexport_fs =
4639                             (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
4640                             "smb_server_share", &error)) == NULL)) {
4641                                 mutex_exit(&zfs_share_lock);
4642                                 return (SET_ERROR(ENOSYS));
4643                         }
4644                         error = zfs_init_sharefs();
4645                         if (error != 0) {
4646                                 mutex_exit(&zfs_share_lock);
4647                                 return (SET_ERROR(ENOSYS));
4648                         }
4649                         zfs_smbshare_inited = 1;
4650                         mutex_exit(&zfs_share_lock);
4651                 }
4652                 break;
4653         default:
4654                 return (SET_ERROR(EINVAL));
4655         }
4656 
4657         switch (zc->zc_share.z_sharetype) {
4658         case ZFS_SHARE_NFS:
4659         case ZFS_UNSHARE_NFS:
4660                 if (error =
4661                     znfsexport_fs((void *)
4662                     (uintptr_t)zc->zc_share.z_exportdata))
4663                         return (error);
4664                 break;
4665         case ZFS_SHARE_SMB:
4666         case ZFS_UNSHARE_SMB:
4667                 if (error = zsmbexport_fs((void *)
4668                     (uintptr_t)zc->zc_share.z_exportdata,
4669                     zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
4670                     B_TRUE: B_FALSE)) {
4671                         return (error);
4672                 }
4673                 break;
4674         }
4675 
4676         opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
4677             zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
4678             SHAREFS_ADD : SHAREFS_REMOVE;
4679 
4680         /*
4681          * Add or remove share from sharetab
4682          */
4683         error = zshare_fs(opcode,
4684             (void *)(uintptr_t)zc->zc_share.z_sharedata,
4685             zc->zc_share.z_sharemax);
4686 
4687         return (error);
4688 
4689 }
4690 
4691 ace_t full_access[] = {
4692         {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4693 };
4694 
4695 /*
4696  * inputs:
4697  * zc_name              name of containing filesystem
4698  * zc_obj               object # beyond which we want next in-use object #
4699  *
4700  * outputs:
4701  * zc_obj               next in-use object #
4702  */
4703 static int
4704 zfs_ioc_next_obj(zfs_cmd_t *zc)
4705 {
4706         objset_t *os = NULL;
4707         int error;
4708 
4709         error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4710         if (error != 0)
4711                 return (error);
4712 
4713         error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
4714             os->os_dsl_dataset->ds_phys->ds_prev_snap_txg);
4715 
4716         dmu_objset_rele(os, FTAG);
4717         return (error);
4718 }
4719 
4720 /*
4721  * inputs:
4722  * zc_name              name of filesystem
4723  * zc_value             prefix name for snapshot
4724  * zc_cleanup_fd        cleanup-on-exit file descriptor for calling process
4725  *
4726  * outputs:
4727  * zc_value             short name of new snapshot
4728  */
4729 static int
4730 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
4731 {
4732         char *snap_name;
4733         char *hold_name;
4734         int error;
4735         minor_t minor;
4736 
4737         error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
4738         if (error != 0)
4739                 return (error);
4740 
4741         snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
4742             (u_longlong_t)ddi_get_lbolt64());
4743         hold_name = kmem_asprintf("%%%s", zc->zc_value);
4744 
4745         error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
4746             hold_name);
4747         if (error == 0)
4748                 (void) strcpy(zc->zc_value, snap_name);
4749         strfree(snap_name);
4750         strfree(hold_name);
4751         zfs_onexit_fd_rele(zc->zc_cleanup_fd);
4752         return (error);
4753 }
4754 
4755 /*
4756  * inputs:
4757  * zc_name              name of "to" snapshot
4758  * zc_value             name of "from" snapshot
4759  * zc_cookie            file descriptor to write diff data on
4760  *
4761  * outputs:
4762  * dmu_diff_record_t's to the file descriptor
4763  */
4764 static int
4765 zfs_ioc_diff(zfs_cmd_t *zc)
4766 {
4767         file_t *fp;
4768         offset_t off;
4769         int error;
4770 
4771         fp = getf(zc->zc_cookie);
4772         if (fp == NULL)
4773                 return (SET_ERROR(EBADF));
4774 
4775         off = fp->f_offset;
4776 
4777         error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off);
4778 
4779         if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4780                 fp->f_offset = off;
4781         releasef(zc->zc_cookie);
4782 
4783         return (error);
4784 }
4785 
4786 /*
4787  * Remove all ACL files in shares dir
4788  */
4789 static int
4790 zfs_smb_acl_purge(znode_t *dzp)
4791 {
4792         zap_cursor_t    zc;
4793         zap_attribute_t zap;
4794         zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
4795         int error;
4796 
4797         for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
4798             (error = zap_cursor_retrieve(&zc, &zap)) == 0;
4799             zap_cursor_advance(&zc)) {
4800                 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
4801                     NULL, 0)) != 0)
4802                         break;
4803         }
4804         zap_cursor_fini(&zc);
4805         return (error);
4806 }
4807 
4808 static int
4809 zfs_ioc_smb_acl(zfs_cmd_t *zc)
4810 {
4811         vnode_t *vp;
4812         znode_t *dzp;
4813         vnode_t *resourcevp = NULL;
4814         znode_t *sharedir;
4815         zfsvfs_t *zfsvfs;
4816         nvlist_t *nvlist;
4817         char *src, *target;
4818         vattr_t vattr;
4819         vsecattr_t vsec;
4820         int error = 0;
4821 
4822         if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
4823             NO_FOLLOW, NULL, &vp)) != 0)
4824                 return (error);
4825 
4826         /* Now make sure mntpnt and dataset are ZFS */
4827 
4828         if (vp->v_vfsp->vfs_fstype != zfsfstype ||
4829             (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
4830             zc->zc_name) != 0)) {
4831                 VN_RELE(vp);
4832                 return (SET_ERROR(EINVAL));
4833         }
4834 
4835         dzp = VTOZ(vp);
4836         zfsvfs = dzp->z_zfsvfs;
4837         ZFS_ENTER(zfsvfs);
4838 
4839         /*
4840          * Create share dir if its missing.
4841          */
4842         mutex_enter(&zfsvfs->z_lock);
4843         if (zfsvfs->z_shares_dir == 0) {
4844                 dmu_tx_t *tx;
4845 
4846                 tx = dmu_tx_create(zfsvfs->z_os);
4847                 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
4848                     ZFS_SHARES_DIR);
4849                 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
4850                 error = dmu_tx_assign(tx, TXG_WAIT);
4851                 if (error != 0) {
4852                         dmu_tx_abort(tx);
4853                 } else {
4854                         error = zfs_create_share_dir(zfsvfs, tx);
4855                         dmu_tx_commit(tx);
4856                 }
4857                 if (error != 0) {
4858                         mutex_exit(&zfsvfs->z_lock);
4859                         VN_RELE(vp);
4860                         ZFS_EXIT(zfsvfs);
4861                         return (error);
4862                 }
4863         }
4864         mutex_exit(&zfsvfs->z_lock);
4865 
4866         ASSERT(zfsvfs->z_shares_dir);
4867         if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
4868                 VN_RELE(vp);
4869                 ZFS_EXIT(zfsvfs);
4870                 return (error);
4871         }
4872 
4873         switch (zc->zc_cookie) {
4874         case ZFS_SMB_ACL_ADD:
4875                 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
4876                 vattr.va_type = VREG;
4877                 vattr.va_mode = S_IFREG|0777;
4878                 vattr.va_uid = 0;
4879                 vattr.va_gid = 0;
4880 
4881                 vsec.vsa_mask = VSA_ACE;
4882                 vsec.vsa_aclentp = &full_access;
4883                 vsec.vsa_aclentsz = sizeof (full_access);
4884                 vsec.vsa_aclcnt = 1;
4885 
4886                 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
4887                     &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
4888                 if (resourcevp)
4889                         VN_RELE(resourcevp);
4890                 break;
4891 
4892         case ZFS_SMB_ACL_REMOVE:
4893                 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
4894                     NULL, 0);
4895                 break;
4896 
4897         case ZFS_SMB_ACL_RENAME:
4898                 if ((error = get_nvlist(zc->zc_nvlist_src,
4899                     zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
4900                         VN_RELE(vp);
4901                         ZFS_EXIT(zfsvfs);
4902                         return (error);
4903                 }
4904                 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
4905                     nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
4906                     &target)) {
4907                         VN_RELE(vp);
4908                         VN_RELE(ZTOV(sharedir));
4909                         ZFS_EXIT(zfsvfs);
4910                         nvlist_free(nvlist);
4911                         return (error);
4912                 }
4913                 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
4914                     kcred, NULL, 0);
4915                 nvlist_free(nvlist);
4916                 break;
4917 
4918         case ZFS_SMB_ACL_PURGE:
4919                 error = zfs_smb_acl_purge(sharedir);
4920                 break;
4921 
4922         default:
4923                 error = SET_ERROR(EINVAL);
4924                 break;
4925         }
4926 
4927         VN_RELE(vp);
4928         VN_RELE(ZTOV(sharedir));
4929 
4930         ZFS_EXIT(zfsvfs);
4931 
4932         return (error);
4933 }
4934 
4935 /*
4936  * innvl: {
4937  *     "holds" -> { snapname -> holdname (string), ... }
4938  *     (optional) "cleanup_fd" -> fd (int32)
4939  * }
4940  *
4941  * outnvl: {
4942  *     snapname -> error value (int32)
4943  *     ...
4944  * }
4945  */
4946 /* ARGSUSED */
4947 static int
4948 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
4949 {
4950         nvlist_t *holds;
4951         int cleanup_fd = -1;
4952         int error;
4953         minor_t minor = 0;
4954 
4955         error = nvlist_lookup_nvlist(args, "holds", &holds);
4956         if (error != 0)
4957                 return (SET_ERROR(EINVAL));
4958 
4959         if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
4960                 error = zfs_onexit_fd_hold(cleanup_fd, &minor);
4961                 if (error != 0)
4962                         return (error);
4963         }
4964 
4965         error = dsl_dataset_user_hold(holds, minor, errlist);
4966         if (minor != 0)
4967                 zfs_onexit_fd_rele(cleanup_fd);
4968         return (error);
4969 }
4970 
4971 /*
4972  * innvl is not used.
4973  *
4974  * outnvl: {
4975  *    holdname -> time added (uint64 seconds since epoch)
4976  *    ...
4977  * }
4978  */
4979 /* ARGSUSED */
4980 static int
4981 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
4982 {
4983         return (dsl_dataset_get_holds(snapname, outnvl));
4984 }
4985 
4986 /*
4987  * innvl: {
4988  *     snapname -> { holdname, ... }
4989  *     ...
4990  * }
4991  *
4992  * outnvl: {
4993  *     snapname -> error value (int32)
4994  *     ...
4995  * }
4996  */
4997 /* ARGSUSED */
4998 static int
4999 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
5000 {
5001         return (dsl_dataset_user_release(holds, errlist));
5002 }
5003 
5004 /*
5005  * inputs:
5006  * zc_name              name of new filesystem or snapshot
5007  * zc_value             full name of old snapshot
5008  *
5009  * outputs:
5010  * zc_cookie            space in bytes
5011  * zc_objset_type       compressed space in bytes
5012  * zc_perm_action       uncompressed space in bytes
5013  */
5014 static int
5015 zfs_ioc_space_written(zfs_cmd_t *zc)
5016 {
5017         int error;
5018         dsl_pool_t *dp;
5019         dsl_dataset_t *new, *old;
5020 
5021         error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5022         if (error != 0)
5023                 return (error);
5024         error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
5025         if (error != 0) {
5026                 dsl_pool_rele(dp, FTAG);
5027                 return (error);
5028         }
5029         error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
5030         if (error != 0) {
5031                 dsl_dataset_rele(new, FTAG);
5032                 dsl_pool_rele(dp, FTAG);
5033                 return (error);
5034         }
5035 
5036         error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
5037             &zc->zc_objset_type, &zc->zc_perm_action);
5038         dsl_dataset_rele(old, FTAG);
5039         dsl_dataset_rele(new, FTAG);
5040         dsl_pool_rele(dp, FTAG);
5041         return (error);
5042 }
5043 
5044 /*
5045  * innvl: {
5046  *     "firstsnap" -> snapshot name
5047  * }
5048  *
5049  * outnvl: {
5050  *     "used" -> space in bytes
5051  *     "compressed" -> compressed space in bytes
5052  *     "uncompressed" -> uncompressed space in bytes
5053  * }
5054  */
5055 static int
5056 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
5057 {
5058         int error;
5059         dsl_pool_t *dp;
5060         dsl_dataset_t *new, *old;
5061         char *firstsnap;
5062         uint64_t used, comp, uncomp;
5063 
5064         if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5065                 return (SET_ERROR(EINVAL));
5066 
5067         error = dsl_pool_hold(lastsnap, FTAG, &dp);
5068         if (error != 0)
5069                 return (error);
5070 
5071         error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
5072         if (error != 0) {
5073                 dsl_pool_rele(dp, FTAG);
5074                 return (error);
5075         }
5076         error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
5077         if (error != 0) {
5078                 dsl_dataset_rele(new, FTAG);
5079                 dsl_pool_rele(dp, FTAG);
5080                 return (error);
5081         }
5082 
5083         error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5084         dsl_dataset_rele(old, FTAG);
5085         dsl_dataset_rele(new, FTAG);
5086         dsl_pool_rele(dp, FTAG);
5087         fnvlist_add_uint64(outnvl, "used", used);
5088         fnvlist_add_uint64(outnvl, "compressed", comp);
5089         fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5090         return (error);
5091 }
5092 
5093 /*
5094  * innvl: {
5095  *     "fd" -> file descriptor to write stream to (int32)
5096  *     (optional) "fromsnap" -> full snap name to send an incremental from
5097  * }
5098  *
5099  * outnvl is unused
5100  */
5101 /* ARGSUSED */
5102 static int
5103 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5104 {
5105         int error;
5106         offset_t off;
5107         char *fromname = NULL;
5108         int fd;
5109 
5110         error = nvlist_lookup_int32(innvl, "fd", &fd);
5111         if (error != 0)
5112                 return (SET_ERROR(EINVAL));
5113 
5114         (void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
5115 
5116         file_t *fp = getf(fd);
5117         if (fp == NULL)
5118                 return (SET_ERROR(EBADF));
5119 
5120         off = fp->f_offset;
5121         error = dmu_send(snapname, fromname, fd, fp->f_vnode, &off);
5122 
5123         if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5124                 fp->f_offset = off;
5125         releasef(fd);
5126         return (error);
5127 }
5128 
5129 /*
5130  * Determine approximately how large a zfs send stream will be -- the number
5131  * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5132  *
5133  * innvl: {
5134  *     (optional) "fromsnap" -> full snap name to send an incremental from
5135  * }
5136  *
5137  * outnvl: {
5138  *     "space" -> bytes of space (uint64)
5139  * }
5140  */
5141 static int
5142 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5143 {
5144         dsl_pool_t *dp;
5145         dsl_dataset_t *fromsnap = NULL;
5146         dsl_dataset_t *tosnap;
5147         int error;
5148         char *fromname;
5149         uint64_t space;
5150 
5151         error = dsl_pool_hold(snapname, FTAG, &dp);
5152         if (error != 0)
5153                 return (error);
5154 
5155         error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
5156         if (error != 0) {
5157                 dsl_pool_rele(dp, FTAG);
5158                 return (error);
5159         }
5160 
5161         error = nvlist_lookup_string(innvl, "fromsnap", &fromname);
5162         if (error == 0) {
5163                 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
5164                 if (error != 0) {
5165                         dsl_dataset_rele(tosnap, FTAG);
5166                         dsl_pool_rele(dp, FTAG);
5167                         return (error);
5168                 }
5169         }
5170 
5171         error = dmu_send_estimate(tosnap, fromsnap, &space);
5172         fnvlist_add_uint64(outnvl, "space", space);
5173 
5174         if (fromsnap != NULL)
5175                 dsl_dataset_rele(fromsnap, FTAG);
5176         dsl_dataset_rele(tosnap, FTAG);
5177         dsl_pool_rele(dp, FTAG);
5178         return (error);
5179 }
5180 
5181 
5182 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
5183 
5184 static void
5185 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5186     zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5187     boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
5188 {
5189         zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5190 
5191         ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5192         ASSERT3U(ioc, <, ZFS_IOC_LAST);
5193         ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5194         ASSERT3P(vec->zvec_func, ==, NULL);
5195 
5196         vec->zvec_legacy_func = func;
5197         vec->zvec_secpolicy = secpolicy;
5198         vec->zvec_namecheck = namecheck;
5199         vec->zvec_allow_log = log_history;
5200         vec->zvec_pool_check = pool_check;
5201 }
5202 
5203 /*
5204  * See the block comment at the beginning of this file for details on
5205  * each argument to this function.
5206  */
5207 static void
5208 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
5209     zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5210     zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
5211     boolean_t allow_log)
5212 {
5213         zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5214 
5215         ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5216         ASSERT3U(ioc, <, ZFS_IOC_LAST);
5217         ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5218         ASSERT3P(vec->zvec_func, ==, NULL);
5219 
5220         /* if we are logging, the name must be valid */
5221         ASSERT(!allow_log || namecheck != NO_NAME);
5222 
5223         vec->zvec_name = name;
5224         vec->zvec_func = func;
5225         vec->zvec_secpolicy = secpolicy;
5226         vec->zvec_namecheck = namecheck;
5227         vec->zvec_pool_check = pool_check;
5228         vec->zvec_smush_outnvlist = smush_outnvlist;
5229         vec->zvec_allow_log = allow_log;
5230 }
5231 
5232 static void
5233 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5234     zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
5235     zfs_ioc_poolcheck_t pool_check)
5236 {
5237         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5238             POOL_NAME, log_history, pool_check);
5239 }
5240 
5241 static void
5242 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5243     zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
5244 {
5245         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5246             DATASET_NAME, B_FALSE, pool_check);
5247 }
5248 
5249 static void
5250 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5251 {
5252         zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
5253             POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5254 }
5255 
5256 static void
5257 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5258     zfs_secpolicy_func_t *secpolicy)
5259 {
5260         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5261             NO_NAME, B_FALSE, POOL_CHECK_NONE);
5262 }
5263 
5264 static void
5265 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
5266     zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
5267 {
5268         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5269             DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
5270 }
5271 
5272 static void
5273 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5274 {
5275         zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
5276             zfs_secpolicy_read);
5277 }
5278 
5279 static void
5280 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5281         zfs_secpolicy_func_t *secpolicy)
5282 {
5283         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5284             DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5285 }
5286 
5287 static void
5288 zfs_ioctl_init(void)
5289 {
5290         zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
5291             zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
5292             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5293 
5294         zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
5295             zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
5296             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
5297 
5298         zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
5299             zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
5300             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5301 
5302         zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
5303             zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
5304             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5305 
5306         zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
5307             zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
5308             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5309 
5310         zfs_ioctl_register("create", ZFS_IOC_CREATE,
5311             zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
5312             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5313 
5314         zfs_ioctl_register("clone", ZFS_IOC_CLONE,
5315             zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
5316             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5317 
5318         zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
5319             zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
5320             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5321 
5322         zfs_ioctl_register("hold", ZFS_IOC_HOLD,
5323             zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
5324             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5325         zfs_ioctl_register("release", ZFS_IOC_RELEASE,
5326             zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
5327             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5328 
5329         zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
5330             zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
5331             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5332 
5333         zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
5334             zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
5335             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE);
5336 
5337         /* IOCTLS that use the legacy function signature */
5338 
5339         zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
5340             zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
5341 
5342         zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
5343             zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5344         zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
5345             zfs_ioc_pool_scan);
5346         zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
5347             zfs_ioc_pool_upgrade);
5348         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
5349             zfs_ioc_vdev_add);
5350         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
5351             zfs_ioc_vdev_remove);
5352         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
5353             zfs_ioc_vdev_set_state);
5354         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
5355             zfs_ioc_vdev_attach);
5356         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
5357             zfs_ioc_vdev_detach);
5358         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
5359             zfs_ioc_vdev_setpath);
5360         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
5361             zfs_ioc_vdev_setfru);
5362         zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
5363             zfs_ioc_pool_set_props);
5364         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
5365             zfs_ioc_vdev_split);
5366         zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
5367             zfs_ioc_pool_reguid);
5368 
5369         zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
5370             zfs_ioc_pool_configs, zfs_secpolicy_none);
5371         zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
5372             zfs_ioc_pool_tryimport, zfs_secpolicy_config);
5373         zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
5374             zfs_ioc_inject_fault, zfs_secpolicy_inject);
5375         zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
5376             zfs_ioc_clear_fault, zfs_secpolicy_inject);
5377         zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
5378             zfs_ioc_inject_list_next, zfs_secpolicy_inject);
5379 
5380         /*
5381          * pool destroy, and export don't log the history as part of
5382          * zfsdev_ioctl, but rather zfs_ioc_pool_export
5383          * does the logging of those commands.
5384          */
5385         zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
5386             zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5387         zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
5388             zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5389 
5390         zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
5391             zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5392         zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
5393             zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5394 
5395         zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
5396             zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED);
5397         zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
5398             zfs_ioc_dsobj_to_dsname,
5399             zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED);
5400         zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
5401             zfs_ioc_pool_get_history,
5402             zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
5403 
5404         zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
5405             zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5406 
5407         zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
5408             zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5409         zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
5410             zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5411 
5412         zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
5413             zfs_ioc_space_written);
5414         zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
5415             zfs_ioc_objset_recvd_props);
5416         zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
5417             zfs_ioc_next_obj);
5418         zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
5419             zfs_ioc_get_fsacl);
5420         zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
5421             zfs_ioc_objset_stats);
5422         zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
5423             zfs_ioc_objset_zplprops);
5424         zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
5425             zfs_ioc_dataset_list_next);
5426         zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
5427             zfs_ioc_snapshot_list_next);
5428         zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
5429             zfs_ioc_send_progress);
5430 
5431         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
5432             zfs_ioc_diff, zfs_secpolicy_diff);
5433         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
5434             zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
5435         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
5436             zfs_ioc_obj_to_path, zfs_secpolicy_diff);
5437         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
5438             zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
5439         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
5440             zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
5441         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
5442             zfs_ioc_send, zfs_secpolicy_send);
5443 
5444         zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
5445             zfs_secpolicy_none);
5446         zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
5447             zfs_secpolicy_destroy);
5448         zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
5449             zfs_secpolicy_rename);
5450         zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
5451             zfs_secpolicy_recv);
5452         zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
5453             zfs_secpolicy_promote);
5454         zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
5455             zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
5456         zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
5457             zfs_secpolicy_set_fsacl);
5458 
5459         zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
5460             zfs_secpolicy_share, POOL_CHECK_NONE);
5461         zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
5462             zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
5463         zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
5464             zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
5465             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5466         zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
5467             zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
5468             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5469 }
5470 
5471 int
5472 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
5473     zfs_ioc_poolcheck_t check)
5474 {
5475         spa_t *spa;
5476         int error;
5477 
5478         ASSERT(type == POOL_NAME || type == DATASET_NAME);
5479 
5480         if (check & POOL_CHECK_NONE)
5481                 return (0);
5482 
5483         error = spa_open(name, &spa, FTAG);
5484         if (error == 0) {
5485                 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
5486                         error = SET_ERROR(EAGAIN);
5487                 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
5488                         error = SET_ERROR(EROFS);
5489                 spa_close(spa, FTAG);
5490         }
5491         return (error);
5492 }
5493 
5494 /*
5495  * Find a free minor number.
5496  */
5497 minor_t
5498 zfsdev_minor_alloc(void)
5499 {
5500         static minor_t last_minor;
5501         minor_t m;
5502 
5503         ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5504 
5505         for (m = last_minor + 1; m != last_minor; m++) {
5506                 if (m > ZFSDEV_MAX_MINOR)
5507                         m = 1;
5508                 if (ddi_get_soft_state(zfsdev_state, m) == NULL) {
5509                         last_minor = m;
5510                         return (m);
5511                 }
5512         }
5513 
5514         return (0);
5515 }
5516 
5517 static int
5518 zfs_ctldev_init(dev_t *devp)
5519 {
5520         minor_t minor;
5521         zfs_soft_state_t *zs;
5522 
5523         ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5524         ASSERT(getminor(*devp) == 0);
5525 
5526         minor = zfsdev_minor_alloc();
5527         if (minor == 0)
5528                 return (SET_ERROR(ENXIO));
5529 
5530         if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS)
5531                 return (SET_ERROR(EAGAIN));
5532 
5533         *devp = makedevice(getemajor(*devp), minor);
5534 
5535         zs = ddi_get_soft_state(zfsdev_state, minor);
5536         zs->zss_type = ZSST_CTLDEV;
5537         zfs_onexit_init((zfs_onexit_t **)&zs->zss_data);
5538 
5539         return (0);
5540 }
5541 
5542 static void
5543 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor)
5544 {
5545         ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5546 
5547         zfs_onexit_destroy(zo);
5548         ddi_soft_state_free(zfsdev_state, minor);
5549 }
5550 
5551 void *
5552 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which)
5553 {
5554         zfs_soft_state_t *zp;
5555 
5556         zp = ddi_get_soft_state(zfsdev_state, minor);
5557         if (zp == NULL || zp->zss_type != which)
5558                 return (NULL);
5559 
5560         return (zp->zss_data);
5561 }
5562 
5563 static int
5564 zfsdev_open(dev_t *devp, int flag, int otyp, cred_t *cr)
5565 {
5566         int error = 0;
5567 
5568         if (getminor(*devp) != 0)
5569                 return (zvol_open(devp, flag, otyp, cr));
5570 
5571         /* This is the control device. Allocate a new minor if requested. */
5572         if (flag & FEXCL) {
5573                 mutex_enter(&zfsdev_state_lock);
5574                 error = zfs_ctldev_init(devp);
5575                 mutex_exit(&zfsdev_state_lock);
5576         }
5577 
5578         return (error);
5579 }
5580 
5581 static int
5582 zfsdev_close(dev_t dev, int flag, int otyp, cred_t *cr)
5583 {
5584         zfs_onexit_t *zo;
5585         minor_t minor = getminor(dev);
5586 
5587         if (minor == 0)
5588                 return (0);
5589 
5590         mutex_enter(&zfsdev_state_lock);
5591         zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV);
5592         if (zo == NULL) {
5593                 mutex_exit(&zfsdev_state_lock);
5594                 return (zvol_close(dev, flag, otyp, cr));
5595         }
5596         zfs_ctldev_destroy(zo, minor);
5597         mutex_exit(&zfsdev_state_lock);
5598 
5599         return (0);
5600 }
5601 
5602 static int
5603 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
5604 {
5605         zfs_cmd_t *zc;
5606         uint_t vecnum;
5607         int error, rc, len;
5608         minor_t minor = getminor(dev);
5609         const zfs_ioc_vec_t *vec;
5610         char *saved_poolname = NULL;
5611         nvlist_t *innvl = NULL;
5612 
5613         if (minor != 0 &&
5614             zfsdev_get_soft_state(minor, ZSST_CTLDEV) == NULL)
5615                 return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
5616 
5617         vecnum = cmd - ZFS_IOC_FIRST;
5618         ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
5619 
5620         if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
5621                 return (SET_ERROR(EINVAL));
5622         vec = &zfs_ioc_vec[vecnum];
5623 
5624         zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
5625 
5626         error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
5627         if (error != 0) {
5628                 error = SET_ERROR(EFAULT);
5629                 goto out;
5630         }
5631 
5632         zc->zc_iflags = flag & FKIOCTL;
5633         if (zc->zc_nvlist_src_size != 0) {
5634                 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
5635                     zc->zc_iflags, &innvl);
5636                 if (error != 0)
5637                         goto out;
5638         }
5639 
5640         /*
5641          * Ensure that all pool/dataset names are valid before we pass down to
5642          * the lower layers.
5643          */
5644         zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
5645         switch (vec->zvec_namecheck) {
5646         case POOL_NAME:
5647                 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
5648                         error = SET_ERROR(EINVAL);
5649                 else
5650                         error = pool_status_check(zc->zc_name,
5651                             vec->zvec_namecheck, vec->zvec_pool_check);
5652                 break;
5653 
5654         case DATASET_NAME:
5655                 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
5656                         error = SET_ERROR(EINVAL);
5657                 else
5658                         error = pool_status_check(zc->zc_name,
5659                             vec->zvec_namecheck, vec->zvec_pool_check);
5660                 break;
5661 
5662         case NO_NAME:
5663                 break;
5664         }
5665 
5666 
5667         if (error == 0 && !(flag & FKIOCTL))
5668                 error = vec->zvec_secpolicy(zc, innvl, cr);
5669 
5670         if (error != 0)
5671                 goto out;
5672 
5673         /* legacy ioctls can modify zc_name */
5674         len = strcspn(zc->zc_name, "/@") + 1;
5675         saved_poolname = kmem_alloc(len, KM_SLEEP);
5676         (void) strlcpy(saved_poolname, zc->zc_name, len);
5677 
5678         if (vec->zvec_func != NULL) {
5679                 nvlist_t *outnvl;
5680                 int puterror = 0;
5681                 spa_t *spa;
5682                 nvlist_t *lognv = NULL;
5683 
5684                 ASSERT(vec->zvec_legacy_func == NULL);
5685 
5686                 /*
5687                  * Add the innvl to the lognv before calling the func,
5688                  * in case the func changes the innvl.
5689                  */
5690                 if (vec->zvec_allow_log) {
5691                         lognv = fnvlist_alloc();
5692                         fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
5693                             vec->zvec_name);
5694                         if (!nvlist_empty(innvl)) {
5695                                 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
5696                                     innvl);
5697                         }
5698                 }
5699 
5700                 outnvl = fnvlist_alloc();
5701                 error = vec->zvec_func(zc->zc_name, innvl, outnvl);
5702 
5703                 if (error == 0 && vec->zvec_allow_log &&
5704                     spa_open(zc->zc_name, &spa, FTAG) == 0) {
5705                         if (!nvlist_empty(outnvl)) {
5706                                 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
5707                                     outnvl);
5708                         }
5709                         (void) spa_history_log_nvl(spa, lognv);
5710                         spa_close(spa, FTAG);
5711                 }
5712                 fnvlist_free(lognv);
5713 
5714                 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
5715                         int smusherror = 0;
5716                         if (vec->zvec_smush_outnvlist) {
5717                                 smusherror = nvlist_smush(outnvl,
5718                                     zc->zc_nvlist_dst_size);
5719                         }
5720                         if (smusherror == 0)
5721                                 puterror = put_nvlist(zc, outnvl);
5722                 }
5723 
5724                 if (puterror != 0)
5725                         error = puterror;
5726 
5727                 nvlist_free(outnvl);
5728         } else {
5729                 error = vec->zvec_legacy_func(zc);
5730         }
5731 
5732 out:
5733         nvlist_free(innvl);
5734         rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
5735         if (error == 0 && rc != 0)
5736                 error = SET_ERROR(EFAULT);
5737         if (error == 0 && vec->zvec_allow_log) {
5738                 char *s = tsd_get(zfs_allow_log_key);
5739                 if (s != NULL)
5740                         strfree(s);
5741                 (void) tsd_set(zfs_allow_log_key, saved_poolname);
5742         } else {
5743                 if (saved_poolname != NULL)
5744                         strfree(saved_poolname);
5745         }
5746 
5747         kmem_free(zc, sizeof (zfs_cmd_t));
5748         return (error);
5749 }
5750 
5751 static int
5752 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
5753 {
5754         if (cmd != DDI_ATTACH)
5755                 return (DDI_FAILURE);
5756 
5757         if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
5758             DDI_PSEUDO, 0) == DDI_FAILURE)
5759                 return (DDI_FAILURE);
5760 
5761         zfs_dip = dip;
5762 
5763         ddi_report_dev(dip);
5764 
5765         return (DDI_SUCCESS);
5766 }
5767 
5768 static int
5769 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
5770 {
5771         if (spa_busy() || zfs_busy() || zvol_busy())
5772                 return (DDI_FAILURE);
5773 
5774         if (cmd != DDI_DETACH)
5775                 return (DDI_FAILURE);
5776 
5777         zfs_dip = NULL;
5778 
5779         ddi_prop_remove_all(dip);
5780         ddi_remove_minor_node(dip, NULL);
5781 
5782         return (DDI_SUCCESS);
5783 }
5784 
5785 /*ARGSUSED*/
5786 static int
5787 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
5788 {
5789         switch (infocmd) {
5790         case DDI_INFO_DEVT2DEVINFO:
5791                 *result = zfs_dip;
5792                 return (DDI_SUCCESS);
5793 
5794         case DDI_INFO_DEVT2INSTANCE:
5795                 *result = (void *)0;
5796                 return (DDI_SUCCESS);
5797         }
5798 
5799         return (DDI_FAILURE);
5800 }
5801 
5802 /*
5803  * OK, so this is a little weird.
5804  *
5805  * /dev/zfs is the control node, i.e. minor 0.
5806  * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
5807  *
5808  * /dev/zfs has basically nothing to do except serve up ioctls,
5809  * so most of the standard driver entry points are in zvol.c.
5810  */
5811 static struct cb_ops zfs_cb_ops = {
5812         zfsdev_open,    /* open */
5813         zfsdev_close,   /* close */
5814         zvol_strategy,  /* strategy */
5815         nodev,          /* print */
5816         zvol_dump,      /* dump */
5817         zvol_read,      /* read */
5818         zvol_write,     /* write */
5819         zfsdev_ioctl,   /* ioctl */
5820         nodev,          /* devmap */
5821         nodev,          /* mmap */
5822         nodev,          /* segmap */
5823         nochpoll,       /* poll */
5824         ddi_prop_op,    /* prop_op */
5825         NULL,           /* streamtab */
5826         D_NEW | D_MP | D_64BIT,         /* Driver compatibility flag */
5827         CB_REV,         /* version */
5828         nodev,          /* async read */
5829         nodev,          /* async write */
5830 };
5831 
5832 static struct dev_ops zfs_dev_ops = {
5833         DEVO_REV,       /* version */
5834         0,              /* refcnt */
5835         zfs_info,       /* info */
5836         nulldev,        /* identify */
5837         nulldev,        /* probe */
5838         zfs_attach,     /* attach */
5839         zfs_detach,     /* detach */
5840         nodev,          /* reset */
5841         &zfs_cb_ops,        /* driver operations */
5842         NULL,           /* no bus operations */
5843         NULL,           /* power */
5844         ddi_quiesce_not_needed, /* quiesce */
5845 };
5846 
5847 static struct modldrv zfs_modldrv = {
5848         &mod_driverops,
5849         "ZFS storage pool",
5850         &zfs_dev_ops
5851 };
5852 
5853 static struct modlinkage modlinkage = {
5854         MODREV_1,
5855         (void *)&zfs_modlfs,
5856         (void *)&zfs_modldrv,
5857         NULL
5858 };
5859 
5860 static void
5861 zfs_allow_log_destroy(void *arg)
5862 {
5863         char *poolname = arg;
5864         strfree(poolname);
5865 }
5866 
5867 int
5868 _init(void)
5869 {
5870         int error;
5871 
5872         spa_init(FREAD | FWRITE);
5873         zfs_init();
5874         zvol_init();
5875         zfs_ioctl_init();
5876 
5877         if ((error = mod_install(&modlinkage)) != 0) {
5878                 zvol_fini();
5879                 zfs_fini();
5880                 spa_fini();
5881                 return (error);
5882         }
5883 
5884         tsd_create(&zfs_fsyncer_key, NULL);
5885         tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
5886         tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
5887 
5888         error = ldi_ident_from_mod(&modlinkage, &zfs_li);
5889         ASSERT(error == 0);
5890         mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
5891 
5892         return (0);
5893 }
5894 
5895 int
5896 _fini(void)
5897 {
5898         int error;
5899 
5900         if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
5901                 return (SET_ERROR(EBUSY));
5902 
5903         if ((error = mod_remove(&modlinkage)) != 0)
5904                 return (error);
5905 
5906         zvol_fini();
5907         zfs_fini();
5908         spa_fini();
5909         if (zfs_nfsshare_inited)
5910                 (void) ddi_modclose(nfs_mod);
5911         if (zfs_smbshare_inited)
5912                 (void) ddi_modclose(smbsrv_mod);
5913         if (zfs_nfsshare_inited || zfs_smbshare_inited)
5914                 (void) ddi_modclose(sharefs_mod);
5915 
5916         tsd_destroy(&zfs_fsyncer_key);
5917         ldi_ident_release(zfs_li);
5918         zfs_li = NULL;
5919         mutex_destroy(&zfs_share_lock);
5920 
5921         return (error);
5922 }
5923 
5924 int
5925 _info(struct modinfo *modinfop)
5926 {
5927         return (mod_info(&modlinkage, modinfop));
5928 }