Use the LZ4 algorithm to compress metadata when the corresponding feature is enabled

   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) 2014, 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_bookmark.h>
 181 #include <sys/dsl_userhold.h>
 182 #include <sys/zfeature.h>
 183 
 184 #include "zfs_namecheck.h"
 185 #include "zfs_prop.h"
 186 #include "zfs_deleg.h"
 187 #include "zfs_comutil.h"
 188 
 189 extern struct modlfs zfs_modlfs;
 190 
 191 extern void zfs_init(void);
 192 extern void zfs_fini(void);
 193 
 194 ldi_ident_t zfs_li = NULL;
 195 dev_info_t *zfs_dip;
 196 
 197 uint_t zfs_fsyncer_key;
 198 extern uint_t rrw_tsd_key;
 199 static uint_t zfs_allow_log_key;
 200 
 201 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *);
 202 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *);
 203 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *);
 204 
 205 typedef enum {
 206         NO_NAME,
 207         POOL_NAME,
 208         DATASET_NAME
 209 } zfs_ioc_namecheck_t;
 210 
 211 typedef enum {
 212         POOL_CHECK_NONE         = 1 << 0,
 213         POOL_CHECK_SUSPENDED    = 1 << 1,
 214         POOL_CHECK_READONLY     = 1 << 2,
 215 } zfs_ioc_poolcheck_t;
 216 
 217 typedef struct zfs_ioc_vec {
 218         zfs_ioc_legacy_func_t   *zvec_legacy_func;
 219         zfs_ioc_func_t          *zvec_func;
 220         zfs_secpolicy_func_t    *zvec_secpolicy;
 221         zfs_ioc_namecheck_t     zvec_namecheck;
 222         boolean_t               zvec_allow_log;
 223         zfs_ioc_poolcheck_t     zvec_pool_check;
 224         boolean_t               zvec_smush_outnvlist;
 225         const char              *zvec_name;
 226 } zfs_ioc_vec_t;
 227 
 228 /* This array is indexed by zfs_userquota_prop_t */
 229 static const char *userquota_perms[] = {
 230         ZFS_DELEG_PERM_USERUSED,
 231         ZFS_DELEG_PERM_USERQUOTA,
 232         ZFS_DELEG_PERM_GROUPUSED,
 233         ZFS_DELEG_PERM_GROUPQUOTA,
 234 };
 235 
 236 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
 237 static int zfs_check_settable(const char *name, nvpair_t *property,
 238     cred_t *cr);
 239 static int zfs_check_clearable(char *dataset, nvlist_t *props,
 240     nvlist_t **errors);
 241 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
 242     boolean_t *);
 243 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
 244 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
 245 
 246 static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature);
 247 
 248 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
 249 void
 250 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
 251 {
 252         const char *newfile;
 253         char buf[512];
 254         va_list adx;
 255 
 256         /*
 257          * Get rid of annoying "../common/" prefix to filename.
 258          */
 259         newfile = strrchr(file, '/');
 260         if (newfile != NULL) {
 261                 newfile = newfile + 1; /* Get rid of leading / */
 262         } else {
 263                 newfile = file;
 264         }
 265 
 266         va_start(adx, fmt);
 267         (void) vsnprintf(buf, sizeof (buf), fmt, adx);
 268         va_end(adx);
 269 
 270         /*
 271          * To get this data, use the zfs-dprintf probe as so:
 272          * dtrace -q -n 'zfs-dprintf \
 273          *      /stringof(arg0) == "dbuf.c"/ \
 274          *      {printf("%s: %s", stringof(arg1), stringof(arg3))}'
 275          * arg0 = file name
 276          * arg1 = function name
 277          * arg2 = line number
 278          * arg3 = message
 279          */
 280         DTRACE_PROBE4(zfs__dprintf,
 281             char *, newfile, char *, func, int, line, char *, buf);
 282 }
 283 
 284 static void
 285 history_str_free(char *buf)
 286 {
 287         kmem_free(buf, HIS_MAX_RECORD_LEN);
 288 }
 289 
 290 static char *
 291 history_str_get(zfs_cmd_t *zc)
 292 {
 293         char *buf;
 294 
 295         if (zc->zc_history == NULL)
 296                 return (NULL);
 297 
 298         buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
 299         if (copyinstr((void *)(uintptr_t)zc->zc_history,
 300             buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
 301                 history_str_free(buf);
 302                 return (NULL);
 303         }
 304 
 305         buf[HIS_MAX_RECORD_LEN -1] = '\0';
 306 
 307         return (buf);
 308 }
 309 
 310 /*
 311  * Check to see if the named dataset is currently defined as bootable
 312  */
 313 static boolean_t
 314 zfs_is_bootfs(const char *name)
 315 {
 316         objset_t *os;
 317 
 318         if (dmu_objset_hold(name, FTAG, &os) == 0) {
 319                 boolean_t ret;
 320                 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
 321                 dmu_objset_rele(os, FTAG);
 322                 return (ret);
 323         }
 324         return (B_FALSE);
 325 }
 326 
 327 /*
 328  * Return non-zero if the spa version is less than requested version.
 329  */
 330 static int
 331 zfs_earlier_version(const char *name, int version)
 332 {
 333         spa_t *spa;
 334 
 335         if (spa_open(name, &spa, FTAG) == 0) {
 336                 if (spa_version(spa) < version) {
 337                         spa_close(spa, FTAG);
 338                         return (1);
 339                 }
 340                 spa_close(spa, FTAG);
 341         }
 342         return (0);
 343 }
 344 
 345 /*
 346  * Return TRUE if the ZPL version is less than requested version.
 347  */
 348 static boolean_t
 349 zpl_earlier_version(const char *name, int version)
 350 {
 351         objset_t *os;
 352         boolean_t rc = B_TRUE;
 353 
 354         if (dmu_objset_hold(name, FTAG, &os) == 0) {
 355                 uint64_t zplversion;
 356 
 357                 if (dmu_objset_type(os) != DMU_OST_ZFS) {
 358                         dmu_objset_rele(os, FTAG);
 359                         return (B_TRUE);
 360                 }
 361                 /* XXX reading from non-owned objset */
 362                 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
 363                         rc = zplversion < version;
 364                 dmu_objset_rele(os, FTAG);
 365         }
 366         return (rc);
 367 }
 368 
 369 static void
 370 zfs_log_history(zfs_cmd_t *zc)
 371 {
 372         spa_t *spa;
 373         char *buf;
 374 
 375         if ((buf = history_str_get(zc)) == NULL)
 376                 return;
 377 
 378         if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
 379                 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
 380                         (void) spa_history_log(spa, buf);
 381                 spa_close(spa, FTAG);
 382         }
 383         history_str_free(buf);
 384 }
 385 
 386 /*
 387  * Policy for top-level read operations (list pools).  Requires no privileges,
 388  * and can be used in the local zone, as there is no associated dataset.
 389  */
 390 /* ARGSUSED */
 391 static int
 392 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 393 {
 394         return (0);
 395 }
 396 
 397 /*
 398  * Policy for dataset read operations (list children, get statistics).  Requires
 399  * no privileges, but must be visible in the local zone.
 400  */
 401 /* ARGSUSED */
 402 static int
 403 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 404 {
 405         if (INGLOBALZONE(curproc) ||
 406             zone_dataset_visible(zc->zc_name, NULL))
 407                 return (0);
 408 
 409         return (SET_ERROR(ENOENT));
 410 }
 411 
 412 static int
 413 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
 414 {
 415         int writable = 1;
 416 
 417         /*
 418          * The dataset must be visible by this zone -- check this first
 419          * so they don't see EPERM on something they shouldn't know about.
 420          */
 421         if (!INGLOBALZONE(curproc) &&
 422             !zone_dataset_visible(dataset, &writable))
 423                 return (SET_ERROR(ENOENT));
 424 
 425         if (INGLOBALZONE(curproc)) {
 426                 /*
 427                  * If the fs is zoned, only root can access it from the
 428                  * global zone.
 429                  */
 430                 if (secpolicy_zfs(cr) && zoned)
 431                         return (SET_ERROR(EPERM));
 432         } else {
 433                 /*
 434                  * If we are in a local zone, the 'zoned' property must be set.
 435                  */
 436                 if (!zoned)
 437                         return (SET_ERROR(EPERM));
 438 
 439                 /* must be writable by this zone */
 440                 if (!writable)
 441                         return (SET_ERROR(EPERM));
 442         }
 443         return (0);
 444 }
 445 
 446 static int
 447 zfs_dozonecheck(const char *dataset, cred_t *cr)
 448 {
 449         uint64_t zoned;
 450 
 451         if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
 452                 return (SET_ERROR(ENOENT));
 453 
 454         return (zfs_dozonecheck_impl(dataset, zoned, cr));
 455 }
 456 
 457 static int
 458 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
 459 {
 460         uint64_t zoned;
 461 
 462         if (dsl_prop_get_int_ds(ds, "zoned", &zoned))
 463                 return (SET_ERROR(ENOENT));
 464 
 465         return (zfs_dozonecheck_impl(dataset, zoned, cr));
 466 }
 467 
 468 static int
 469 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
 470     const char *perm, cred_t *cr)
 471 {
 472         int error;
 473 
 474         error = zfs_dozonecheck_ds(name, ds, cr);
 475         if (error == 0) {
 476                 error = secpolicy_zfs(cr);
 477                 if (error != 0)
 478                         error = dsl_deleg_access_impl(ds, perm, cr);
 479         }
 480         return (error);
 481 }
 482 
 483 static int
 484 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
 485 {
 486         int error;
 487         dsl_dataset_t *ds;
 488         dsl_pool_t *dp;
 489 
 490         error = dsl_pool_hold(name, FTAG, &dp);
 491         if (error != 0)
 492                 return (error);
 493 
 494         error = dsl_dataset_hold(dp, name, FTAG, &ds);
 495         if (error != 0) {
 496                 dsl_pool_rele(dp, FTAG);
 497                 return (error);
 498         }
 499 
 500         error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
 501 
 502         dsl_dataset_rele(ds, FTAG);
 503         dsl_pool_rele(dp, FTAG);
 504         return (error);
 505 }
 506 
 507 /*
 508  * Policy for setting the security label property.
 509  *
 510  * Returns 0 for success, non-zero for access and other errors.
 511  */
 512 static int
 513 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
 514 {
 515         char            ds_hexsl[MAXNAMELEN];
 516         bslabel_t       ds_sl, new_sl;
 517         boolean_t       new_default = FALSE;
 518         uint64_t        zoned;
 519         int             needed_priv = -1;
 520         int             error;
 521 
 522         /* First get the existing dataset label. */
 523         error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
 524             1, sizeof (ds_hexsl), &ds_hexsl, NULL);
 525         if (error != 0)
 526                 return (SET_ERROR(EPERM));
 527 
 528         if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
 529                 new_default = TRUE;
 530 
 531         /* The label must be translatable */
 532         if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
 533                 return (SET_ERROR(EINVAL));
 534 
 535         /*
 536          * In a non-global zone, disallow attempts to set a label that
 537          * doesn't match that of the zone; otherwise no other checks
 538          * are needed.
 539          */
 540         if (!INGLOBALZONE(curproc)) {
 541                 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
 542                         return (SET_ERROR(EPERM));
 543                 return (0);
 544         }
 545 
 546         /*
 547          * For global-zone datasets (i.e., those whose zoned property is
 548          * "off", verify that the specified new label is valid for the
 549          * global zone.
 550          */
 551         if (dsl_prop_get_integer(name,
 552             zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
 553                 return (SET_ERROR(EPERM));
 554         if (!zoned) {
 555                 if (zfs_check_global_label(name, strval) != 0)
 556                         return (SET_ERROR(EPERM));
 557         }
 558 
 559         /*
 560          * If the existing dataset label is nondefault, check if the
 561          * dataset is mounted (label cannot be changed while mounted).
 562          * Get the zfsvfs; if there isn't one, then the dataset isn't
 563          * mounted (or isn't a dataset, doesn't exist, ...).
 564          */
 565         if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
 566                 objset_t *os;
 567                 static char *setsl_tag = "setsl_tag";
 568 
 569                 /*
 570                  * Try to own the dataset; abort if there is any error,
 571                  * (e.g., already mounted, in use, or other error).
 572                  */
 573                 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
 574                     setsl_tag, &os);
 575                 if (error != 0)
 576                         return (SET_ERROR(EPERM));
 577 
 578                 dmu_objset_disown(os, setsl_tag);
 579 
 580                 if (new_default) {
 581                         needed_priv = PRIV_FILE_DOWNGRADE_SL;
 582                         goto out_check;
 583                 }
 584 
 585                 if (hexstr_to_label(strval, &new_sl) != 0)
 586                         return (SET_ERROR(EPERM));
 587 
 588                 if (blstrictdom(&ds_sl, &new_sl))
 589                         needed_priv = PRIV_FILE_DOWNGRADE_SL;
 590                 else if (blstrictdom(&new_sl, &ds_sl))
 591                         needed_priv = PRIV_FILE_UPGRADE_SL;
 592         } else {
 593                 /* dataset currently has a default label */
 594                 if (!new_default)
 595                         needed_priv = PRIV_FILE_UPGRADE_SL;
 596         }
 597 
 598 out_check:
 599         if (needed_priv != -1)
 600                 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
 601         return (0);
 602 }
 603 
 604 static int
 605 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
 606     cred_t *cr)
 607 {
 608         char *strval;
 609 
 610         /*
 611          * Check permissions for special properties.
 612          */
 613         switch (prop) {
 614         case ZFS_PROP_ZONED:
 615                 /*
 616                  * Disallow setting of 'zoned' from within a local zone.
 617                  */
 618                 if (!INGLOBALZONE(curproc))
 619                         return (SET_ERROR(EPERM));
 620                 break;
 621 
 622         case ZFS_PROP_QUOTA:
 623         case ZFS_PROP_FILESYSTEM_LIMIT:
 624         case ZFS_PROP_SNAPSHOT_LIMIT:
 625                 if (!INGLOBALZONE(curproc)) {
 626                         uint64_t zoned;
 627                         char setpoint[MAXNAMELEN];
 628                         /*
 629                          * Unprivileged users are allowed to modify the
 630                          * limit on things *under* (ie. contained by)
 631                          * the thing they own.
 632                          */
 633                         if (dsl_prop_get_integer(dsname, "zoned", &zoned,
 634                             setpoint))
 635                                 return (SET_ERROR(EPERM));
 636                         if (!zoned || strlen(dsname) <= strlen(setpoint))
 637                                 return (SET_ERROR(EPERM));
 638                 }
 639                 break;
 640 
 641         case ZFS_PROP_MLSLABEL:
 642                 if (!is_system_labeled())
 643                         return (SET_ERROR(EPERM));
 644 
 645                 if (nvpair_value_string(propval, &strval) == 0) {
 646                         int err;
 647 
 648                         err = zfs_set_slabel_policy(dsname, strval, CRED());
 649                         if (err != 0)
 650                                 return (err);
 651                 }
 652                 break;
 653         }
 654 
 655         return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
 656 }
 657 
 658 /* ARGSUSED */
 659 static int
 660 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 661 {
 662         int error;
 663 
 664         error = zfs_dozonecheck(zc->zc_name, cr);
 665         if (error != 0)
 666                 return (error);
 667 
 668         /*
 669          * permission to set permissions will be evaluated later in
 670          * dsl_deleg_can_allow()
 671          */
 672         return (0);
 673 }
 674 
 675 /* ARGSUSED */
 676 static int
 677 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 678 {
 679         return (zfs_secpolicy_write_perms(zc->zc_name,
 680             ZFS_DELEG_PERM_ROLLBACK, cr));
 681 }
 682 
 683 /* ARGSUSED */
 684 static int
 685 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 686 {
 687         dsl_pool_t *dp;
 688         dsl_dataset_t *ds;
 689         char *cp;
 690         int error;
 691 
 692         /*
 693          * Generate the current snapshot name from the given objsetid, then
 694          * use that name for the secpolicy/zone checks.
 695          */
 696         cp = strchr(zc->zc_name, '@');
 697         if (cp == NULL)
 698                 return (SET_ERROR(EINVAL));
 699         error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
 700         if (error != 0)
 701                 return (error);
 702 
 703         error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
 704         if (error != 0) {
 705                 dsl_pool_rele(dp, FTAG);
 706                 return (error);
 707         }
 708 
 709         dsl_dataset_name(ds, zc->zc_name);
 710 
 711         error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
 712             ZFS_DELEG_PERM_SEND, cr);
 713         dsl_dataset_rele(ds, FTAG);
 714         dsl_pool_rele(dp, FTAG);
 715 
 716         return (error);
 717 }
 718 
 719 /* ARGSUSED */
 720 static int
 721 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 722 {
 723         return (zfs_secpolicy_write_perms(zc->zc_name,
 724             ZFS_DELEG_PERM_SEND, cr));
 725 }
 726 
 727 /* ARGSUSED */
 728 static int
 729 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 730 {
 731         vnode_t *vp;
 732         int error;
 733 
 734         if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
 735             NO_FOLLOW, NULL, &vp)) != 0)
 736                 return (error);
 737 
 738         /* Now make sure mntpnt and dataset are ZFS */
 739 
 740         if (vp->v_vfsp->vfs_fstype != zfsfstype ||
 741             (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
 742             zc->zc_name) != 0)) {
 743                 VN_RELE(vp);
 744                 return (SET_ERROR(EPERM));
 745         }
 746 
 747         VN_RELE(vp);
 748         return (dsl_deleg_access(zc->zc_name,
 749             ZFS_DELEG_PERM_SHARE, cr));
 750 }
 751 
 752 int
 753 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 754 {
 755         if (!INGLOBALZONE(curproc))
 756                 return (SET_ERROR(EPERM));
 757 
 758         if (secpolicy_nfs(cr) == 0) {
 759                 return (0);
 760         } else {
 761                 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
 762         }
 763 }
 764 
 765 int
 766 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 767 {
 768         if (!INGLOBALZONE(curproc))
 769                 return (SET_ERROR(EPERM));
 770 
 771         if (secpolicy_smb(cr) == 0) {
 772                 return (0);
 773         } else {
 774                 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
 775         }
 776 }
 777 
 778 static int
 779 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
 780 {
 781         char *cp;
 782 
 783         /*
 784          * Remove the @bla or /bla from the end of the name to get the parent.
 785          */
 786         (void) strncpy(parent, datasetname, parentsize);
 787         cp = strrchr(parent, '@');
 788         if (cp != NULL) {
 789                 cp[0] = '\0';
 790         } else {
 791                 cp = strrchr(parent, '/');
 792                 if (cp == NULL)
 793                         return (SET_ERROR(ENOENT));
 794                 cp[0] = '\0';
 795         }
 796 
 797         return (0);
 798 }
 799 
 800 int
 801 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
 802 {
 803         int error;
 804 
 805         if ((error = zfs_secpolicy_write_perms(name,
 806             ZFS_DELEG_PERM_MOUNT, cr)) != 0)
 807                 return (error);
 808 
 809         return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
 810 }
 811 
 812 /* ARGSUSED */
 813 static int
 814 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 815 {
 816         return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
 817 }
 818 
 819 /*
 820  * Destroying snapshots with delegated permissions requires
 821  * descendant mount and destroy permissions.
 822  */
 823 /* ARGSUSED */
 824 static int
 825 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 826 {
 827         nvlist_t *snaps;
 828         nvpair_t *pair, *nextpair;
 829         int error = 0;
 830 
 831         if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
 832                 return (SET_ERROR(EINVAL));
 833         for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
 834             pair = nextpair) {
 835                 nextpair = nvlist_next_nvpair(snaps, pair);
 836                 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr);
 837                 if (error == ENOENT) {
 838                         /*
 839                          * Ignore any snapshots that don't exist (we consider
 840                          * them "already destroyed").  Remove the name from the
 841                          * nvl here in case the snapshot is created between
 842                          * now and when we try to destroy it (in which case
 843                          * we don't want to destroy it since we haven't
 844                          * checked for permission).
 845                          */
 846                         fnvlist_remove_nvpair(snaps, pair);
 847                         error = 0;
 848                 }
 849                 if (error != 0)
 850                         break;
 851         }
 852 
 853         return (error);
 854 }
 855 
 856 int
 857 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
 858 {
 859         char    parentname[MAXNAMELEN];
 860         int     error;
 861 
 862         if ((error = zfs_secpolicy_write_perms(from,
 863             ZFS_DELEG_PERM_RENAME, cr)) != 0)
 864                 return (error);
 865 
 866         if ((error = zfs_secpolicy_write_perms(from,
 867             ZFS_DELEG_PERM_MOUNT, cr)) != 0)
 868                 return (error);
 869 
 870         if ((error = zfs_get_parent(to, parentname,
 871             sizeof (parentname))) != 0)
 872                 return (error);
 873 
 874         if ((error = zfs_secpolicy_write_perms(parentname,
 875             ZFS_DELEG_PERM_CREATE, cr)) != 0)
 876                 return (error);
 877 
 878         if ((error = zfs_secpolicy_write_perms(parentname,
 879             ZFS_DELEG_PERM_MOUNT, cr)) != 0)
 880                 return (error);
 881 
 882         return (error);
 883 }
 884 
 885 /* ARGSUSED */
 886 static int
 887 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 888 {
 889         return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
 890 }
 891 
 892 /* ARGSUSED */
 893 static int
 894 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 895 {
 896         dsl_pool_t *dp;
 897         dsl_dataset_t *clone;
 898         int error;
 899 
 900         error = zfs_secpolicy_write_perms(zc->zc_name,
 901             ZFS_DELEG_PERM_PROMOTE, cr);
 902         if (error != 0)
 903                 return (error);
 904 
 905         error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
 906         if (error != 0)
 907                 return (error);
 908 
 909         error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
 910 
 911         if (error == 0) {
 912                 char parentname[MAXNAMELEN];
 913                 dsl_dataset_t *origin = NULL;
 914                 dsl_dir_t *dd;
 915                 dd = clone->ds_dir;
 916 
 917                 error = dsl_dataset_hold_obj(dd->dd_pool,
 918                     dd->dd_phys->dd_origin_obj, FTAG, &origin);
 919                 if (error != 0) {
 920                         dsl_dataset_rele(clone, FTAG);
 921                         dsl_pool_rele(dp, FTAG);
 922                         return (error);
 923                 }
 924 
 925                 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
 926                     ZFS_DELEG_PERM_MOUNT, cr);
 927 
 928                 dsl_dataset_name(origin, parentname);
 929                 if (error == 0) {
 930                         error = zfs_secpolicy_write_perms_ds(parentname, origin,
 931                             ZFS_DELEG_PERM_PROMOTE, cr);
 932                 }
 933                 dsl_dataset_rele(clone, FTAG);
 934                 dsl_dataset_rele(origin, FTAG);
 935         }
 936         dsl_pool_rele(dp, FTAG);
 937         return (error);
 938 }
 939 
 940 /* ARGSUSED */
 941 static int
 942 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 943 {
 944         int error;
 945 
 946         if ((error = zfs_secpolicy_write_perms(zc->zc_name,
 947             ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
 948                 return (error);
 949 
 950         if ((error = zfs_secpolicy_write_perms(zc->zc_name,
 951             ZFS_DELEG_PERM_MOUNT, cr)) != 0)
 952                 return (error);
 953 
 954         return (zfs_secpolicy_write_perms(zc->zc_name,
 955             ZFS_DELEG_PERM_CREATE, cr));
 956 }
 957 
 958 int
 959 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
 960 {
 961         return (zfs_secpolicy_write_perms(name,
 962             ZFS_DELEG_PERM_SNAPSHOT, cr));
 963 }
 964 
 965 /*
 966  * Check for permission to create each snapshot in the nvlist.
 967  */
 968 /* ARGSUSED */
 969 static int
 970 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 971 {
 972         nvlist_t *snaps;
 973         int error = 0;
 974         nvpair_t *pair;
 975 
 976         if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
 977                 return (SET_ERROR(EINVAL));
 978         for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
 979             pair = nvlist_next_nvpair(snaps, pair)) {
 980                 char *name = nvpair_name(pair);
 981                 char *atp = strchr(name, '@');
 982 
 983                 if (atp == NULL) {
 984                         error = SET_ERROR(EINVAL);
 985                         break;
 986                 }
 987                 *atp = '\0';
 988                 error = zfs_secpolicy_snapshot_perms(name, cr);
 989                 *atp = '@';
 990                 if (error != 0)
 991                         break;
 992         }
 993         return (error);
 994 }
 995 
 996 /*
 997  * Check for permission to create each snapshot in the nvlist.
 998  */
 999 /* ARGSUSED */
1000 static int
1001 zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1002 {
1003         int error = 0;
1004 
1005         for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
1006             pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
1007                 char *name = nvpair_name(pair);
1008                 char *hashp = strchr(name, '#');
1009 
1010                 if (hashp == NULL) {
1011                         error = SET_ERROR(EINVAL);
1012                         break;
1013                 }
1014                 *hashp = '\0';
1015                 error = zfs_secpolicy_write_perms(name,
1016                     ZFS_DELEG_PERM_BOOKMARK, cr);
1017                 *hashp = '#';
1018                 if (error != 0)
1019                         break;
1020         }
1021         return (error);
1022 }
1023 
1024 /* ARGSUSED */
1025 static int
1026 zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1027 {
1028         nvpair_t *pair, *nextpair;
1029         int error = 0;
1030 
1031         for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1032             pair = nextpair) {
1033                 char *name = nvpair_name(pair);
1034                 char *hashp = strchr(name, '#');
1035                 nextpair = nvlist_next_nvpair(innvl, pair);
1036 
1037                 if (hashp == NULL) {
1038                         error = SET_ERROR(EINVAL);
1039                         break;
1040                 }
1041 
1042                 *hashp = '\0';
1043                 error = zfs_secpolicy_write_perms(name,
1044                     ZFS_DELEG_PERM_DESTROY, cr);
1045                 *hashp = '#';
1046                 if (error == ENOENT) {
1047                         /*
1048                          * Ignore any filesystems that don't exist (we consider
1049                          * their bookmarks "already destroyed").  Remove
1050                          * the name from the nvl here in case the filesystem
1051                          * is created between now and when we try to destroy
1052                          * the bookmark (in which case we don't want to
1053                          * destroy it since we haven't checked for permission).
1054                          */
1055                         fnvlist_remove_nvpair(innvl, pair);
1056                         error = 0;
1057                 }
1058                 if (error != 0)
1059                         break;
1060         }
1061 
1062         return (error);
1063 }
1064 
1065 /* ARGSUSED */
1066 static int
1067 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1068 {
1069         /*
1070          * Even root must have a proper TSD so that we know what pool
1071          * to log to.
1072          */
1073         if (tsd_get(zfs_allow_log_key) == NULL)
1074                 return (SET_ERROR(EPERM));
1075         return (0);
1076 }
1077 
1078 static int
1079 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1080 {
1081         char    parentname[MAXNAMELEN];
1082         int     error;
1083         char    *origin;
1084 
1085         if ((error = zfs_get_parent(zc->zc_name, parentname,
1086             sizeof (parentname))) != 0)
1087                 return (error);
1088 
1089         if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1090             (error = zfs_secpolicy_write_perms(origin,
1091             ZFS_DELEG_PERM_CLONE, cr)) != 0)
1092                 return (error);
1093 
1094         if ((error = zfs_secpolicy_write_perms(parentname,
1095             ZFS_DELEG_PERM_CREATE, cr)) != 0)
1096                 return (error);
1097 
1098         return (zfs_secpolicy_write_perms(parentname,
1099             ZFS_DELEG_PERM_MOUNT, cr));
1100 }
1101 
1102 /*
1103  * Policy for pool operations - create/destroy pools, add vdevs, etc.  Requires
1104  * SYS_CONFIG privilege, which is not available in a local zone.
1105  */
1106 /* ARGSUSED */
1107 static int
1108 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1109 {
1110         if (secpolicy_sys_config(cr, B_FALSE) != 0)
1111                 return (SET_ERROR(EPERM));
1112 
1113         return (0);
1114 }
1115 
1116 /*
1117  * Policy for object to name lookups.
1118  */
1119 /* ARGSUSED */
1120 static int
1121 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1122 {
1123         int error;
1124 
1125         if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1126                 return (0);
1127 
1128         error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1129         return (error);
1130 }
1131 
1132 /*
1133  * Policy for fault injection.  Requires all privileges.
1134  */
1135 /* ARGSUSED */
1136 static int
1137 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1138 {
1139         return (secpolicy_zinject(cr));
1140 }
1141 
1142 /* ARGSUSED */
1143 static int
1144 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1145 {
1146         zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1147 
1148         if (prop == ZPROP_INVAL) {
1149                 if (!zfs_prop_user(zc->zc_value))
1150                         return (SET_ERROR(EINVAL));
1151                 return (zfs_secpolicy_write_perms(zc->zc_name,
1152                     ZFS_DELEG_PERM_USERPROP, cr));
1153         } else {
1154                 return (zfs_secpolicy_setprop(zc->zc_name, prop,
1155                     NULL, cr));
1156         }
1157 }
1158 
1159 static int
1160 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1161 {
1162         int err = zfs_secpolicy_read(zc, innvl, cr);
1163         if (err)
1164                 return (err);
1165 
1166         if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1167                 return (SET_ERROR(EINVAL));
1168 
1169         if (zc->zc_value[0] == 0) {
1170                 /*
1171                  * They are asking about a posix uid/gid.  If it's
1172                  * themself, allow it.
1173                  */
1174                 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1175                     zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
1176                         if (zc->zc_guid == crgetuid(cr))
1177                                 return (0);
1178                 } else {
1179                         if (groupmember(zc->zc_guid, cr))
1180                                 return (0);
1181                 }
1182         }
1183 
1184         return (zfs_secpolicy_write_perms(zc->zc_name,
1185             userquota_perms[zc->zc_objset_type], cr));
1186 }
1187 
1188 static int
1189 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1190 {
1191         int err = zfs_secpolicy_read(zc, innvl, cr);
1192         if (err)
1193                 return (err);
1194 
1195         if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1196                 return (SET_ERROR(EINVAL));
1197 
1198         return (zfs_secpolicy_write_perms(zc->zc_name,
1199             userquota_perms[zc->zc_objset_type], cr));
1200 }
1201 
1202 /* ARGSUSED */
1203 static int
1204 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1205 {
1206         return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1207             NULL, cr));
1208 }
1209 
1210 /* ARGSUSED */
1211 static int
1212 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1213 {
1214         nvpair_t *pair;
1215         nvlist_t *holds;
1216         int error;
1217 
1218         error = nvlist_lookup_nvlist(innvl, "holds", &holds);
1219         if (error != 0)
1220                 return (SET_ERROR(EINVAL));
1221 
1222         for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
1223             pair = nvlist_next_nvpair(holds, pair)) {
1224                 char fsname[MAXNAMELEN];
1225                 error = dmu_fsname(nvpair_name(pair), fsname);
1226                 if (error != 0)
1227                         return (error);
1228                 error = zfs_secpolicy_write_perms(fsname,
1229                     ZFS_DELEG_PERM_HOLD, cr);
1230                 if (error != 0)
1231                         return (error);
1232         }
1233         return (0);
1234 }
1235 
1236 /* ARGSUSED */
1237 static int
1238 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1239 {
1240         nvpair_t *pair;
1241         int error;
1242 
1243         for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1244             pair = nvlist_next_nvpair(innvl, pair)) {
1245                 char fsname[MAXNAMELEN];
1246                 error = dmu_fsname(nvpair_name(pair), fsname);
1247                 if (error != 0)
1248                         return (error);
1249                 error = zfs_secpolicy_write_perms(fsname,
1250                     ZFS_DELEG_PERM_RELEASE, cr);
1251                 if (error != 0)
1252                         return (error);
1253         }
1254         return (0);
1255 }
1256 
1257 /*
1258  * Policy for allowing temporary snapshots to be taken or released
1259  */
1260 static int
1261 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1262 {
1263         /*
1264          * A temporary snapshot is the same as a snapshot,
1265          * hold, destroy and release all rolled into one.
1266          * Delegated diff alone is sufficient that we allow this.
1267          */
1268         int error;
1269 
1270         if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1271             ZFS_DELEG_PERM_DIFF, cr)) == 0)
1272                 return (0);
1273 
1274         error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1275         if (error == 0)
1276                 error = zfs_secpolicy_hold(zc, innvl, cr);
1277         if (error == 0)
1278                 error = zfs_secpolicy_release(zc, innvl, cr);
1279         if (error == 0)
1280                 error = zfs_secpolicy_destroy(zc, innvl, cr);
1281         return (error);
1282 }
1283 
1284 /*
1285  * Returns the nvlist as specified by the user in the zfs_cmd_t.
1286  */
1287 static int
1288 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1289 {
1290         char *packed;
1291         int error;
1292         nvlist_t *list = NULL;
1293 
1294         /*
1295          * Read in and unpack the user-supplied nvlist.
1296          */
1297         if (size == 0)
1298                 return (SET_ERROR(EINVAL));
1299 
1300         packed = kmem_alloc(size, KM_SLEEP);
1301 
1302         if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1303             iflag)) != 0) {
1304                 kmem_free(packed, size);
1305                 return (error);
1306         }
1307 
1308         if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1309                 kmem_free(packed, size);
1310                 return (error);
1311         }
1312 
1313         kmem_free(packed, size);
1314 
1315         *nvp = list;
1316         return (0);
1317 }
1318 
1319 /*
1320  * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1321  * Entries will be removed from the end of the nvlist, and one int32 entry
1322  * named "N_MORE_ERRORS" will be added indicating how many entries were
1323  * removed.
1324  */
1325 static int
1326 nvlist_smush(nvlist_t *errors, size_t max)
1327 {
1328         size_t size;
1329 
1330         size = fnvlist_size(errors);
1331 
1332         if (size > max) {
1333                 nvpair_t *more_errors;
1334                 int n = 0;
1335 
1336                 if (max < 1024)
1337                         return (SET_ERROR(ENOMEM));
1338 
1339                 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1340                 more_errors = nvlist_prev_nvpair(errors, NULL);
1341 
1342                 do {
1343                         nvpair_t *pair = nvlist_prev_nvpair(errors,
1344                             more_errors);
1345                         fnvlist_remove_nvpair(errors, pair);
1346                         n++;
1347                         size = fnvlist_size(errors);
1348                 } while (size > max);
1349 
1350                 fnvlist_remove_nvpair(errors, more_errors);
1351                 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1352                 ASSERT3U(fnvlist_size(errors), <=, max);
1353         }
1354 
1355         return (0);
1356 }
1357 
1358 static int
1359 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1360 {
1361         char *packed = NULL;
1362         int error = 0;
1363         size_t size;
1364 
1365         size = fnvlist_size(nvl);
1366 
1367         if (size > zc->zc_nvlist_dst_size) {
1368                 error = SET_ERROR(ENOMEM);
1369         } else {
1370                 packed = fnvlist_pack(nvl, &size);
1371                 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1372                     size, zc->zc_iflags) != 0)
1373                         error = SET_ERROR(EFAULT);
1374                 fnvlist_pack_free(packed, size);
1375         }
1376 
1377         zc->zc_nvlist_dst_size = size;
1378         zc->zc_nvlist_dst_filled = B_TRUE;
1379         return (error);
1380 }
1381 
1382 static int
1383 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1384 {
1385         objset_t *os;
1386         int error;
1387 
1388         error = dmu_objset_hold(dsname, FTAG, &os);
1389         if (error != 0)
1390                 return (error);
1391         if (dmu_objset_type(os) != DMU_OST_ZFS) {
1392                 dmu_objset_rele(os, FTAG);
1393                 return (SET_ERROR(EINVAL));
1394         }
1395 
1396         mutex_enter(&os->os_user_ptr_lock);
1397         *zfvp = dmu_objset_get_user(os);
1398         if (*zfvp) {
1399                 VFS_HOLD((*zfvp)->z_vfs);
1400         } else {
1401                 error = SET_ERROR(ESRCH);
1402         }
1403         mutex_exit(&os->os_user_ptr_lock);
1404         dmu_objset_rele(os, FTAG);
1405         return (error);
1406 }
1407 
1408 /*
1409  * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1410  * case its z_vfs will be NULL, and it will be opened as the owner.
1411  * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1412  * which prevents all vnode ops from running.
1413  */
1414 static int
1415 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1416 {
1417         int error = 0;
1418 
1419         if (getzfsvfs(name, zfvp) != 0)
1420                 error = zfsvfs_create(name, zfvp);
1421         if (error == 0) {
1422                 rrw_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER :
1423                     RW_READER, tag);
1424                 if ((*zfvp)->z_unmounted) {
1425                         /*
1426                          * XXX we could probably try again, since the unmounting
1427                          * thread should be just about to disassociate the
1428                          * objset from the zfsvfs.
1429                          */
1430                         rrw_exit(&(*zfvp)->z_teardown_lock, tag);
1431                         return (SET_ERROR(EBUSY));
1432                 }
1433         }
1434         return (error);
1435 }
1436 
1437 static void
1438 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1439 {
1440         rrw_exit(&zfsvfs->z_teardown_lock, tag);
1441 
1442         if (zfsvfs->z_vfs) {
1443                 VFS_RELE(zfsvfs->z_vfs);
1444         } else {
1445                 dmu_objset_disown(zfsvfs->z_os, zfsvfs);
1446                 zfsvfs_free(zfsvfs);
1447         }
1448 }
1449 
1450 static int
1451 zfs_ioc_pool_create(zfs_cmd_t *zc)
1452 {
1453         int error;
1454         nvlist_t *config, *props = NULL;
1455         nvlist_t *rootprops = NULL;
1456         nvlist_t *zplprops = NULL;
1457 
1458         if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1459             zc->zc_iflags, &config))
1460                 return (error);
1461 
1462         if (zc->zc_nvlist_src_size != 0 && (error =
1463             get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1464             zc->zc_iflags, &props))) {
1465                 nvlist_free(config);
1466                 return (error);
1467         }
1468 
1469         if (props) {
1470                 nvlist_t *nvl = NULL;
1471                 uint64_t version = SPA_VERSION;
1472 
1473                 (void) nvlist_lookup_uint64(props,
1474                     zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1475                 if (!SPA_VERSION_IS_SUPPORTED(version)) {
1476                         error = SET_ERROR(EINVAL);
1477                         goto pool_props_bad;
1478                 }
1479                 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1480                 if (nvl) {
1481                         error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1482                         if (error != 0) {
1483                                 nvlist_free(config);
1484                                 nvlist_free(props);
1485                                 return (error);
1486                         }
1487                         (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1488                 }
1489                 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1490                 error = zfs_fill_zplprops_root(version, rootprops,
1491                     zplprops, NULL);
1492                 if (error != 0)
1493                         goto pool_props_bad;
1494         }
1495 
1496         error = spa_create(zc->zc_name, config, props, zplprops);
1497 
1498         /*
1499          * Set the remaining root properties
1500          */
1501         if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1502             ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1503                 (void) spa_destroy(zc->zc_name);
1504 
1505 pool_props_bad:
1506         nvlist_free(rootprops);
1507         nvlist_free(zplprops);
1508         nvlist_free(config);
1509         nvlist_free(props);
1510 
1511         return (error);
1512 }
1513 
1514 static int
1515 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1516 {
1517         int error;
1518         zfs_log_history(zc);
1519         error = spa_destroy(zc->zc_name);
1520         if (error == 0)
1521                 zvol_remove_minors(zc->zc_name);
1522         return (error);
1523 }
1524 
1525 static int
1526 zfs_ioc_pool_import(zfs_cmd_t *zc)
1527 {
1528         nvlist_t *config, *props = NULL;
1529         uint64_t guid;
1530         int error;
1531 
1532         if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1533             zc->zc_iflags, &config)) != 0)
1534                 return (error);
1535 
1536         if (zc->zc_nvlist_src_size != 0 && (error =
1537             get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1538             zc->zc_iflags, &props))) {
1539                 nvlist_free(config);
1540                 return (error);
1541         }
1542 
1543         if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1544             guid != zc->zc_guid)
1545                 error = SET_ERROR(EINVAL);
1546         else
1547                 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1548 
1549         if (zc->zc_nvlist_dst != 0) {
1550                 int err;
1551 
1552                 if ((err = put_nvlist(zc, config)) != 0)
1553                         error = err;
1554         }
1555 
1556         nvlist_free(config);
1557 
1558         if (props)
1559                 nvlist_free(props);
1560 
1561         return (error);
1562 }
1563 
1564 static int
1565 zfs_ioc_pool_export(zfs_cmd_t *zc)
1566 {
1567         int error;
1568         boolean_t force = (boolean_t)zc->zc_cookie;
1569         boolean_t hardforce = (boolean_t)zc->zc_guid;
1570 
1571         zfs_log_history(zc);
1572         error = spa_export(zc->zc_name, NULL, force, hardforce);
1573         if (error == 0)
1574                 zvol_remove_minors(zc->zc_name);
1575         return (error);
1576 }
1577 
1578 static int
1579 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1580 {
1581         nvlist_t *configs;
1582         int error;
1583 
1584         if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1585                 return (SET_ERROR(EEXIST));
1586 
1587         error = put_nvlist(zc, configs);
1588 
1589         nvlist_free(configs);
1590 
1591         return (error);
1592 }
1593 
1594 /*
1595  * inputs:
1596  * zc_name              name of the pool
1597  *
1598  * outputs:
1599  * zc_cookie            real errno
1600  * zc_nvlist_dst        config nvlist
1601  * zc_nvlist_dst_size   size of config nvlist
1602  */
1603 static int
1604 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1605 {
1606         nvlist_t *config;
1607         int error;
1608         int ret = 0;
1609 
1610         error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1611             sizeof (zc->zc_value));
1612 
1613         if (config != NULL) {
1614                 ret = put_nvlist(zc, config);
1615                 nvlist_free(config);
1616 
1617                 /*
1618                  * The config may be present even if 'error' is non-zero.
1619                  * In this case we return success, and preserve the real errno
1620                  * in 'zc_cookie'.
1621                  */
1622                 zc->zc_cookie = error;
1623         } else {
1624                 ret = error;
1625         }
1626 
1627         return (ret);
1628 }
1629 
1630 /*
1631  * Try to import the given pool, returning pool stats as appropriate so that
1632  * user land knows which devices are available and overall pool health.
1633  */
1634 static int
1635 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1636 {
1637         nvlist_t *tryconfig, *config;
1638         int error;
1639 
1640         if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1641             zc->zc_iflags, &tryconfig)) != 0)
1642                 return (error);
1643 
1644         config = spa_tryimport(tryconfig);
1645 
1646         nvlist_free(tryconfig);
1647 
1648         if (config == NULL)
1649                 return (SET_ERROR(EINVAL));
1650 
1651         error = put_nvlist(zc, config);
1652         nvlist_free(config);
1653 
1654         return (error);
1655 }
1656 
1657 /*
1658  * inputs:
1659  * zc_name              name of the pool
1660  * zc_cookie            scan func (pool_scan_func_t)
1661  */
1662 static int
1663 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1664 {
1665         spa_t *spa;
1666         int error;
1667 
1668         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1669                 return (error);
1670 
1671         if (zc->zc_cookie == POOL_SCAN_NONE)
1672                 error = spa_scan_stop(spa);
1673         else
1674                 error = spa_scan(spa, zc->zc_cookie);
1675 
1676         spa_close(spa, FTAG);
1677 
1678         return (error);
1679 }
1680 
1681 static int
1682 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1683 {
1684         spa_t *spa;
1685         int error;
1686 
1687         error = spa_open(zc->zc_name, &spa, FTAG);
1688         if (error == 0) {
1689                 spa_freeze(spa);
1690                 spa_close(spa, FTAG);
1691         }
1692         return (error);
1693 }
1694 
1695 static int
1696 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1697 {
1698         spa_t *spa;
1699         int error;
1700 
1701         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1702                 return (error);
1703 
1704         if (zc->zc_cookie < spa_version(spa) ||
1705             !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1706                 spa_close(spa, FTAG);
1707                 return (SET_ERROR(EINVAL));
1708         }
1709 
1710         spa_upgrade(spa, zc->zc_cookie);
1711         spa_close(spa, FTAG);
1712 
1713         return (error);
1714 }
1715 
1716 static int
1717 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1718 {
1719         spa_t *spa;
1720         char *hist_buf;
1721         uint64_t size;
1722         int error;
1723 
1724         if ((size = zc->zc_history_len) == 0)
1725                 return (SET_ERROR(EINVAL));
1726 
1727         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1728                 return (error);
1729 
1730         if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1731                 spa_close(spa, FTAG);
1732                 return (SET_ERROR(ENOTSUP));
1733         }
1734 
1735         hist_buf = kmem_alloc(size, KM_SLEEP);
1736         if ((error = spa_history_get(spa, &zc->zc_history_offset,
1737             &zc->zc_history_len, hist_buf)) == 0) {
1738                 error = ddi_copyout(hist_buf,
1739                     (void *)(uintptr_t)zc->zc_history,
1740                     zc->zc_history_len, zc->zc_iflags);
1741         }
1742 
1743         spa_close(spa, FTAG);
1744         kmem_free(hist_buf, size);
1745         return (error);
1746 }
1747 
1748 static int
1749 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1750 {
1751         spa_t *spa;
1752         int error;
1753 
1754         error = spa_open(zc->zc_name, &spa, FTAG);
1755         if (error == 0) {
1756                 error = spa_change_guid(spa);
1757                 spa_close(spa, FTAG);
1758         }
1759         return (error);
1760 }
1761 
1762 static int
1763 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1764 {
1765         return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1766 }
1767 
1768 /*
1769  * inputs:
1770  * zc_name              name of filesystem
1771  * zc_obj               object to find
1772  *
1773  * outputs:
1774  * zc_value             name of object
1775  */
1776 static int
1777 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1778 {
1779         objset_t *os;
1780         int error;
1781 
1782         /* XXX reading from objset not owned */
1783         if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1784                 return (error);
1785         if (dmu_objset_type(os) != DMU_OST_ZFS) {
1786                 dmu_objset_rele(os, FTAG);
1787                 return (SET_ERROR(EINVAL));
1788         }
1789         error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1790             sizeof (zc->zc_value));
1791         dmu_objset_rele(os, FTAG);
1792 
1793         return (error);
1794 }
1795 
1796 /*
1797  * inputs:
1798  * zc_name              name of filesystem
1799  * zc_obj               object to find
1800  *
1801  * outputs:
1802  * zc_stat              stats on object
1803  * zc_value             path to object
1804  */
1805 static int
1806 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1807 {
1808         objset_t *os;
1809         int error;
1810 
1811         /* XXX reading from objset not owned */
1812         if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1813                 return (error);
1814         if (dmu_objset_type(os) != DMU_OST_ZFS) {
1815                 dmu_objset_rele(os, FTAG);
1816                 return (SET_ERROR(EINVAL));
1817         }
1818         error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1819             sizeof (zc->zc_value));
1820         dmu_objset_rele(os, FTAG);
1821 
1822         return (error);
1823 }
1824 
1825 static int
1826 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1827 {
1828         spa_t *spa;
1829         int error;
1830         nvlist_t *config, **l2cache, **spares;
1831         uint_t nl2cache = 0, nspares = 0;
1832 
1833         error = spa_open(zc->zc_name, &spa, FTAG);
1834         if (error != 0)
1835                 return (error);
1836 
1837         error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1838             zc->zc_iflags, &config);
1839         (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1840             &l2cache, &nl2cache);
1841 
1842         (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1843             &spares, &nspares);
1844 
1845         /*
1846          * A root pool with concatenated devices is not supported.
1847          * Thus, can not add a device to a root pool.
1848          *
1849          * Intent log device can not be added to a rootpool because
1850          * during mountroot, zil is replayed, a seperated log device
1851          * can not be accessed during the mountroot time.
1852          *
1853          * l2cache and spare devices are ok to be added to a rootpool.
1854          */
1855         if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1856                 nvlist_free(config);
1857                 spa_close(spa, FTAG);
1858                 return (SET_ERROR(EDOM));
1859         }
1860 
1861         if (error == 0) {
1862                 error = spa_vdev_add(spa, config);
1863                 nvlist_free(config);
1864         }
1865         spa_close(spa, FTAG);
1866         return (error);
1867 }
1868 
1869 /*
1870  * inputs:
1871  * zc_name              name of the pool
1872  * zc_nvlist_conf       nvlist of devices to remove
1873  * zc_cookie            to stop the remove?
1874  */
1875 static int
1876 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1877 {
1878         spa_t *spa;
1879         int error;
1880 
1881         error = spa_open(zc->zc_name, &spa, FTAG);
1882         if (error != 0)
1883                 return (error);
1884         error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1885         spa_close(spa, FTAG);
1886         return (error);
1887 }
1888 
1889 static int
1890 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1891 {
1892         spa_t *spa;
1893         int error;
1894         vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1895 
1896         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1897                 return (error);
1898         switch (zc->zc_cookie) {
1899         case VDEV_STATE_ONLINE:
1900                 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1901                 break;
1902 
1903         case VDEV_STATE_OFFLINE:
1904                 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1905                 break;
1906 
1907         case VDEV_STATE_FAULTED:
1908                 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1909                     zc->zc_obj != VDEV_AUX_EXTERNAL)
1910                         zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1911 
1912                 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1913                 break;
1914 
1915         case VDEV_STATE_DEGRADED:
1916                 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1917                     zc->zc_obj != VDEV_AUX_EXTERNAL)
1918                         zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1919 
1920                 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1921                 break;
1922 
1923         default:
1924                 error = SET_ERROR(EINVAL);
1925         }
1926         zc->zc_cookie = newstate;
1927         spa_close(spa, FTAG);
1928         return (error);
1929 }
1930 
1931 static int
1932 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1933 {
1934         spa_t *spa;
1935         int replacing = zc->zc_cookie;
1936         nvlist_t *config;
1937         int error;
1938 
1939         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1940                 return (error);
1941 
1942         if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1943             zc->zc_iflags, &config)) == 0) {
1944                 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1945                 nvlist_free(config);
1946         }
1947 
1948         spa_close(spa, FTAG);
1949         return (error);
1950 }
1951 
1952 static int
1953 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1954 {
1955         spa_t *spa;
1956         int error;
1957 
1958         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1959                 return (error);
1960 
1961         error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
1962 
1963         spa_close(spa, FTAG);
1964         return (error);
1965 }
1966 
1967 static int
1968 zfs_ioc_vdev_split(zfs_cmd_t *zc)
1969 {
1970         spa_t *spa;
1971         nvlist_t *config, *props = NULL;
1972         int error;
1973         boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
1974 
1975         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1976                 return (error);
1977 
1978         if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1979             zc->zc_iflags, &config)) {
1980                 spa_close(spa, FTAG);
1981                 return (error);
1982         }
1983 
1984         if (zc->zc_nvlist_src_size != 0 && (error =
1985             get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1986             zc->zc_iflags, &props))) {
1987                 spa_close(spa, FTAG);
1988                 nvlist_free(config);
1989                 return (error);
1990         }
1991 
1992         error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
1993 
1994         spa_close(spa, FTAG);
1995 
1996         nvlist_free(config);
1997         nvlist_free(props);
1998 
1999         return (error);
2000 }
2001 
2002 static int
2003 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
2004 {
2005         spa_t *spa;
2006         char *path = zc->zc_value;
2007         uint64_t guid = zc->zc_guid;
2008         int error;
2009 
2010         error = spa_open(zc->zc_name, &spa, FTAG);
2011         if (error != 0)
2012                 return (error);
2013 
2014         error = spa_vdev_setpath(spa, guid, path);
2015         spa_close(spa, FTAG);
2016         return (error);
2017 }
2018 
2019 static int
2020 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
2021 {
2022         spa_t *spa;
2023         char *fru = zc->zc_value;
2024         uint64_t guid = zc->zc_guid;
2025         int error;
2026 
2027         error = spa_open(zc->zc_name, &spa, FTAG);
2028         if (error != 0)
2029                 return (error);
2030 
2031         error = spa_vdev_setfru(spa, guid, fru);
2032         spa_close(spa, FTAG);
2033         return (error);
2034 }
2035 
2036 static int
2037 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
2038 {
2039         int error = 0;
2040         nvlist_t *nv;
2041 
2042         dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2043 
2044         if (zc->zc_nvlist_dst != 0 &&
2045             (error = dsl_prop_get_all(os, &nv)) == 0) {
2046                 dmu_objset_stats(os, nv);
2047                 /*
2048                  * NB: zvol_get_stats() will read the objset contents,
2049                  * which we aren't supposed to do with a
2050                  * DS_MODE_USER hold, because it could be
2051                  * inconsistent.  So this is a bit of a workaround...
2052                  * XXX reading with out owning
2053                  */
2054                 if (!zc->zc_objset_stats.dds_inconsistent &&
2055                     dmu_objset_type(os) == DMU_OST_ZVOL) {
2056                         error = zvol_get_stats(os, nv);
2057                         if (error == EIO)
2058                                 return (error);
2059                         VERIFY0(error);
2060                 }
2061                 error = put_nvlist(zc, nv);
2062                 nvlist_free(nv);
2063         }
2064 
2065         return (error);
2066 }
2067 
2068 /*
2069  * inputs:
2070  * zc_name              name of filesystem
2071  * zc_nvlist_dst_size   size of buffer for property nvlist
2072  *
2073  * outputs:
2074  * zc_objset_stats      stats
2075  * zc_nvlist_dst        property nvlist
2076  * zc_nvlist_dst_size   size of property nvlist
2077  */
2078 static int
2079 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2080 {
2081         objset_t *os;
2082         int error;
2083 
2084         error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2085         if (error == 0) {
2086                 error = zfs_ioc_objset_stats_impl(zc, os);
2087                 dmu_objset_rele(os, FTAG);
2088         }
2089 
2090         return (error);
2091 }
2092 
2093 /*
2094  * inputs:
2095  * zc_name              name of filesystem
2096  * zc_nvlist_dst_size   size of buffer for property nvlist
2097  *
2098  * outputs:
2099  * zc_nvlist_dst        received property nvlist
2100  * zc_nvlist_dst_size   size of received property nvlist
2101  *
2102  * Gets received properties (distinct from local properties on or after
2103  * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2104  * local property values.
2105  */
2106 static int
2107 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2108 {
2109         int error = 0;
2110         nvlist_t *nv;
2111 
2112         /*
2113          * Without this check, we would return local property values if the
2114          * caller has not already received properties on or after
2115          * SPA_VERSION_RECVD_PROPS.
2116          */
2117         if (!dsl_prop_get_hasrecvd(zc->zc_name))
2118                 return (SET_ERROR(ENOTSUP));
2119 
2120         if (zc->zc_nvlist_dst != 0 &&
2121             (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2122                 error = put_nvlist(zc, nv);
2123                 nvlist_free(nv);
2124         }
2125 
2126         return (error);
2127 }
2128 
2129 static int
2130 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2131 {
2132         uint64_t value;
2133         int error;
2134 
2135         /*
2136          * zfs_get_zplprop() will either find a value or give us
2137          * the default value (if there is one).
2138          */
2139         if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2140                 return (error);
2141         VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2142         return (0);
2143 }
2144 
2145 /*
2146  * inputs:
2147  * zc_name              name of filesystem
2148  * zc_nvlist_dst_size   size of buffer for zpl property nvlist
2149  *
2150  * outputs:
2151  * zc_nvlist_dst        zpl property nvlist
2152  * zc_nvlist_dst_size   size of zpl property nvlist
2153  */
2154 static int
2155 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2156 {
2157         objset_t *os;
2158         int err;
2159 
2160         /* XXX reading without owning */
2161         if (err = dmu_objset_hold(zc->zc_name, FTAG, &os))
2162                 return (err);
2163 
2164         dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2165 
2166         /*
2167          * NB: nvl_add_zplprop() will read the objset contents,
2168          * which we aren't supposed to do with a DS_MODE_USER
2169          * hold, because it could be inconsistent.
2170          */
2171         if (zc->zc_nvlist_dst != NULL &&
2172             !zc->zc_objset_stats.dds_inconsistent &&
2173             dmu_objset_type(os) == DMU_OST_ZFS) {
2174                 nvlist_t *nv;
2175 
2176                 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2177                 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2178                     (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2179                     (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2180                     (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2181                         err = put_nvlist(zc, nv);
2182                 nvlist_free(nv);
2183         } else {
2184                 err = SET_ERROR(ENOENT);
2185         }
2186         dmu_objset_rele(os, FTAG);
2187         return (err);
2188 }
2189 
2190 static boolean_t
2191 dataset_name_hidden(const char *name)
2192 {
2193         /*
2194          * Skip over datasets that are not visible in this zone,
2195          * internal datasets (which have a $ in their name), and
2196          * temporary datasets (which have a % in their name).
2197          */
2198         if (strchr(name, '$') != NULL)
2199                 return (B_TRUE);
2200         if (strchr(name, '%') != NULL)
2201                 return (B_TRUE);
2202         if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
2203                 return (B_TRUE);
2204         return (B_FALSE);
2205 }
2206 
2207 /*
2208  * inputs:
2209  * zc_name              name of filesystem
2210  * zc_cookie            zap cursor
2211  * zc_nvlist_dst_size   size of buffer for property nvlist
2212  *
2213  * outputs:
2214  * zc_name              name of next filesystem
2215  * zc_cookie            zap cursor
2216  * zc_objset_stats      stats
2217  * zc_nvlist_dst        property nvlist
2218  * zc_nvlist_dst_size   size of property nvlist
2219  */
2220 static int
2221 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2222 {
2223         objset_t *os;
2224         int error;
2225         char *p;
2226         size_t orig_len = strlen(zc->zc_name);
2227 
2228 top:
2229         if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) {
2230                 if (error == ENOENT)
2231                         error = SET_ERROR(ESRCH);
2232                 return (error);
2233         }
2234 
2235         p = strrchr(zc->zc_name, '/');
2236         if (p == NULL || p[1] != '\0')
2237                 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2238         p = zc->zc_name + strlen(zc->zc_name);
2239 
2240         do {
2241                 error = dmu_dir_list_next(os,
2242                     sizeof (zc->zc_name) - (p - zc->zc_name), p,
2243                     NULL, &zc->zc_cookie);
2244                 if (error == ENOENT)
2245                         error = SET_ERROR(ESRCH);
2246         } while (error == 0 && dataset_name_hidden(zc->zc_name));
2247         dmu_objset_rele(os, FTAG);
2248 
2249         /*
2250          * If it's an internal dataset (ie. with a '$' in its name),
2251          * don't try to get stats for it, otherwise we'll return ENOENT.
2252          */
2253         if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2254                 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2255                 if (error == ENOENT) {
2256                         /* We lost a race with destroy, get the next one. */
2257                         zc->zc_name[orig_len] = '\0';
2258                         goto top;
2259                 }
2260         }
2261         return (error);
2262 }
2263 
2264 /*
2265  * inputs:
2266  * zc_name              name of filesystem
2267  * zc_cookie            zap cursor
2268  * zc_nvlist_dst_size   size of buffer for property nvlist
2269  *
2270  * outputs:
2271  * zc_name              name of next snapshot
2272  * zc_objset_stats      stats
2273  * zc_nvlist_dst        property nvlist
2274  * zc_nvlist_dst_size   size of property nvlist
2275  */
2276 static int
2277 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2278 {
2279         objset_t *os;
2280         int error;
2281 
2282         error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2283         if (error != 0) {
2284                 return (error == ENOENT ? ESRCH : error);
2285         }
2286 
2287         /*
2288          * A dataset name of maximum length cannot have any snapshots,
2289          * so exit immediately.
2290          */
2291         if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
2292                 dmu_objset_rele(os, FTAG);
2293                 return (SET_ERROR(ESRCH));
2294         }
2295 
2296         error = dmu_snapshot_list_next(os,
2297             sizeof (zc->zc_name) - strlen(zc->zc_name),
2298             zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2299             NULL);
2300 
2301         if (error == 0) {
2302                 dsl_dataset_t *ds;
2303                 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2304 
2305                 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2306                 if (error == 0) {
2307                         objset_t *ossnap;
2308 
2309                         error = dmu_objset_from_ds(ds, &ossnap);
2310                         if (error == 0)
2311                                 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2312                         dsl_dataset_rele(ds, FTAG);
2313                 }
2314         } else if (error == ENOENT) {
2315                 error = SET_ERROR(ESRCH);
2316         }
2317 
2318         dmu_objset_rele(os, FTAG);
2319         /* if we failed, undo the @ that we tacked on to zc_name */
2320         if (error != 0)
2321                 *strchr(zc->zc_name, '@') = '\0';
2322         return (error);
2323 }
2324 
2325 static int
2326 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2327 {
2328         const char *propname = nvpair_name(pair);
2329         uint64_t *valary;
2330         unsigned int vallen;
2331         const char *domain;
2332         char *dash;
2333         zfs_userquota_prop_t type;
2334         uint64_t rid;
2335         uint64_t quota;
2336         zfsvfs_t *zfsvfs;
2337         int err;
2338 
2339         if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2340                 nvlist_t *attrs;
2341                 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2342                 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2343                     &pair) != 0)
2344                         return (SET_ERROR(EINVAL));
2345         }
2346 
2347         /*
2348          * A correctly constructed propname is encoded as
2349          * userquota@<rid>-<domain>.
2350          */
2351         if ((dash = strchr(propname, '-')) == NULL ||
2352             nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2353             vallen != 3)
2354                 return (SET_ERROR(EINVAL));
2355 
2356         domain = dash + 1;
2357         type = valary[0];
2358         rid = valary[1];
2359         quota = valary[2];
2360 
2361         err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2362         if (err == 0) {
2363                 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2364                 zfsvfs_rele(zfsvfs, FTAG);
2365         }
2366 
2367         return (err);
2368 }
2369 
2370 /*
2371  * If the named property is one that has a special function to set its value,
2372  * return 0 on success and a positive error code on failure; otherwise if it is
2373  * not one of the special properties handled by this function, return -1.
2374  *
2375  * XXX: It would be better for callers of the property interface if we handled
2376  * these special cases in dsl_prop.c (in the dsl layer).
2377  */
2378 static int
2379 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2380     nvpair_t *pair)
2381 {
2382         const char *propname = nvpair_name(pair);
2383         zfs_prop_t prop = zfs_name_to_prop(propname);
2384         uint64_t intval;
2385         int err;
2386 
2387         if (prop == ZPROP_INVAL) {
2388                 if (zfs_prop_userquota(propname))
2389                         return (zfs_prop_set_userquota(dsname, pair));
2390                 return (-1);
2391         }
2392 
2393         if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2394                 nvlist_t *attrs;
2395                 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2396                 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2397                     &pair) == 0);
2398         }
2399 
2400         if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2401                 return (-1);
2402 
2403         VERIFY(0 == nvpair_value_uint64(pair, &intval));
2404 
2405         switch (prop) {
2406         case ZFS_PROP_QUOTA:
2407                 err = dsl_dir_set_quota(dsname, source, intval);
2408                 break;
2409         case ZFS_PROP_REFQUOTA:
2410                 err = dsl_dataset_set_refquota(dsname, source, intval);
2411                 break;
2412         case ZFS_PROP_FILESYSTEM_LIMIT:
2413         case ZFS_PROP_SNAPSHOT_LIMIT:
2414                 if (intval == UINT64_MAX) {
2415                         /* clearing the limit, just do it */
2416                         err = 0;
2417                 } else {
2418                         err = dsl_dir_activate_fs_ss_limit(dsname);
2419                 }
2420                 /*
2421                  * Set err to -1 to force the zfs_set_prop_nvlist code down the
2422                  * default path to set the value in the nvlist.
2423                  */
2424                 if (err == 0)
2425                         err = -1;
2426                 break;
2427         case ZFS_PROP_RESERVATION:
2428                 err = dsl_dir_set_reservation(dsname, source, intval);
2429                 break;
2430         case ZFS_PROP_REFRESERVATION:
2431                 err = dsl_dataset_set_refreservation(dsname, source, intval);
2432                 break;
2433         case ZFS_PROP_VOLSIZE:
2434                 err = zvol_set_volsize(dsname, intval);
2435                 break;
2436         case ZFS_PROP_VERSION:
2437         {
2438                 zfsvfs_t *zfsvfs;
2439 
2440                 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2441                         break;
2442 
2443                 err = zfs_set_version(zfsvfs, intval);
2444                 zfsvfs_rele(zfsvfs, FTAG);
2445 
2446                 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2447                         zfs_cmd_t *zc;
2448 
2449                         zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2450                         (void) strcpy(zc->zc_name, dsname);
2451                         (void) zfs_ioc_userspace_upgrade(zc);
2452                         kmem_free(zc, sizeof (zfs_cmd_t));
2453                 }
2454                 break;
2455         }
2456         case ZFS_PROP_COMPRESSION:
2457         {
2458                 if (intval == ZIO_COMPRESS_LZ4) {
2459                         spa_t *spa;
2460 
2461                         if ((err = spa_open(dsname, &spa, FTAG)) != 0)
2462                                 return (err);
2463 
2464                         /*
2465                          * Setting the LZ4 compression algorithm activates
2466                          * the feature.
2467                          */
2468                         if (!spa_feature_is_active(spa,
2469                             SPA_FEATURE_LZ4_COMPRESS)) {
2470                                 if ((err = zfs_prop_activate_feature(spa,
2471                                     SPA_FEATURE_LZ4_COMPRESS)) != 0) {
2472                                         spa_close(spa, FTAG);
2473                                         return (err);
2474                                 }
2475                         }
2476 
2477                         spa_close(spa, FTAG);
2478                 }
2479                 /*
2480                  * We still want the default set action to be performed in the
2481                  * caller, we only performed zfeature settings here.
2482                  */
2483                 err = -1;
2484                 break;
2485         }
2486 
2487         default:
2488                 err = -1;
2489         }
2490 
2491         return (err);
2492 }
2493 
2494 /*
2495  * This function is best effort. If it fails to set any of the given properties,
2496  * it continues to set as many as it can and returns the last error
2497  * encountered. If the caller provides a non-NULL errlist, it will be filled in
2498  * with the list of names of all the properties that failed along with the
2499  * corresponding error numbers.
2500  *
2501  * If every property is set successfully, zero is returned and errlist is not
2502  * modified.
2503  */
2504 int
2505 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2506     nvlist_t *errlist)
2507 {
2508         nvpair_t *pair;
2509         nvpair_t *propval;
2510         int rv = 0;
2511         uint64_t intval;
2512         char *strval;
2513         nvlist_t *genericnvl = fnvlist_alloc();
2514         nvlist_t *retrynvl = fnvlist_alloc();
2515 
2516 retry:
2517         pair = NULL;
2518         while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2519                 const char *propname = nvpair_name(pair);
2520                 zfs_prop_t prop = zfs_name_to_prop(propname);
2521                 int err = 0;
2522 
2523                 /* decode the property value */
2524                 propval = pair;
2525                 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2526                         nvlist_t *attrs;
2527                         attrs = fnvpair_value_nvlist(pair);
2528                         if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2529                             &propval) != 0)
2530                                 err = SET_ERROR(EINVAL);
2531                 }
2532 
2533                 /* Validate value type */
2534                 if (err == 0 && prop == ZPROP_INVAL) {
2535                         if (zfs_prop_user(propname)) {
2536                                 if (nvpair_type(propval) != DATA_TYPE_STRING)
2537                                         err = SET_ERROR(EINVAL);
2538                         } else if (zfs_prop_userquota(propname)) {
2539                                 if (nvpair_type(propval) !=
2540                                     DATA_TYPE_UINT64_ARRAY)
2541                                         err = SET_ERROR(EINVAL);
2542                         } else {
2543                                 err = SET_ERROR(EINVAL);
2544                         }
2545                 } else if (err == 0) {
2546                         if (nvpair_type(propval) == DATA_TYPE_STRING) {
2547                                 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2548                                         err = SET_ERROR(EINVAL);
2549                         } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2550                                 const char *unused;
2551 
2552                                 intval = fnvpair_value_uint64(propval);
2553 
2554                                 switch (zfs_prop_get_type(prop)) {
2555                                 case PROP_TYPE_NUMBER:
2556                                         break;
2557                                 case PROP_TYPE_STRING:
2558                                         err = SET_ERROR(EINVAL);
2559                                         break;
2560                                 case PROP_TYPE_INDEX:
2561                                         if (zfs_prop_index_to_string(prop,
2562                                             intval, &unused) != 0)
2563                                                 err = SET_ERROR(EINVAL);
2564                                         break;
2565                                 default:
2566                                         cmn_err(CE_PANIC,
2567                                             "unknown property type");
2568                                 }
2569                         } else {
2570                                 err = SET_ERROR(EINVAL);
2571                         }
2572                 }
2573 
2574                 /* Validate permissions */
2575                 if (err == 0)
2576                         err = zfs_check_settable(dsname, pair, CRED());
2577 
2578                 if (err == 0) {
2579                         err = zfs_prop_set_special(dsname, source, pair);
2580                         if (err == -1) {
2581                                 /*
2582                                  * For better performance we build up a list of
2583                                  * properties to set in a single transaction.
2584                                  */
2585                                 err = nvlist_add_nvpair(genericnvl, pair);
2586                         } else if (err != 0 && nvl != retrynvl) {
2587                                 /*
2588                                  * This may be a spurious error caused by
2589                                  * receiving quota and reservation out of order.
2590                                  * Try again in a second pass.
2591                                  */
2592                                 err = nvlist_add_nvpair(retrynvl, pair);
2593                         }
2594                 }
2595 
2596                 if (err != 0) {
2597                         if (errlist != NULL)
2598                                 fnvlist_add_int32(errlist, propname, err);
2599                         rv = err;
2600                 }
2601         }
2602 
2603         if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2604                 nvl = retrynvl;
2605                 goto retry;
2606         }
2607 
2608         if (!nvlist_empty(genericnvl) &&
2609             dsl_props_set(dsname, source, genericnvl) != 0) {
2610                 /*
2611                  * If this fails, we still want to set as many properties as we
2612                  * can, so try setting them individually.
2613                  */
2614                 pair = NULL;
2615                 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2616                         const char *propname = nvpair_name(pair);
2617                         int err = 0;
2618 
2619                         propval = pair;
2620                         if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2621                                 nvlist_t *attrs;
2622                                 attrs = fnvpair_value_nvlist(pair);
2623                                 propval = fnvlist_lookup_nvpair(attrs,
2624                                     ZPROP_VALUE);
2625                         }
2626 
2627                         if (nvpair_type(propval) == DATA_TYPE_STRING) {
2628                                 strval = fnvpair_value_string(propval);
2629                                 err = dsl_prop_set_string(dsname, propname,
2630                                     source, strval);
2631                         } else {
2632                                 intval = fnvpair_value_uint64(propval);
2633                                 err = dsl_prop_set_int(dsname, propname, source,
2634                                     intval);
2635                         }
2636 
2637                         if (err != 0) {
2638                                 if (errlist != NULL) {
2639                                         fnvlist_add_int32(errlist, propname,
2640                                             err);
2641                                 }
2642                                 rv = err;
2643                         }
2644                 }
2645         }
2646         nvlist_free(genericnvl);
2647         nvlist_free(retrynvl);
2648 
2649         return (rv);
2650 }
2651 
2652 /*
2653  * Check that all the properties are valid user properties.
2654  */
2655 static int
2656 zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2657 {
2658         nvpair_t *pair = NULL;
2659         int error = 0;
2660 
2661         while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2662                 const char *propname = nvpair_name(pair);
2663 
2664                 if (!zfs_prop_user(propname) ||
2665                     nvpair_type(pair) != DATA_TYPE_STRING)
2666                         return (SET_ERROR(EINVAL));
2667 
2668                 if (error = zfs_secpolicy_write_perms(fsname,
2669                     ZFS_DELEG_PERM_USERPROP, CRED()))
2670                         return (error);
2671 
2672                 if (strlen(propname) >= ZAP_MAXNAMELEN)
2673                         return (SET_ERROR(ENAMETOOLONG));
2674 
2675                 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN)
2676                         return (E2BIG);
2677         }
2678         return (0);
2679 }
2680 
2681 static void
2682 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2683 {
2684         nvpair_t *pair;
2685 
2686         VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2687 
2688         pair = NULL;
2689         while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2690                 if (nvlist_exists(skipped, nvpair_name(pair)))
2691                         continue;
2692 
2693                 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2694         }
2695 }
2696 
2697 static int
2698 clear_received_props(const char *dsname, nvlist_t *props,
2699     nvlist_t *skipped)
2700 {
2701         int err = 0;
2702         nvlist_t *cleared_props = NULL;
2703         props_skip(props, skipped, &cleared_props);
2704         if (!nvlist_empty(cleared_props)) {
2705                 /*
2706                  * Acts on local properties until the dataset has received
2707                  * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2708                  */
2709                 zprop_source_t flags = (ZPROP_SRC_NONE |
2710                     (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2711                 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2712         }
2713         nvlist_free(cleared_props);
2714         return (err);
2715 }
2716 
2717 /*
2718  * inputs:
2719  * zc_name              name of filesystem
2720  * zc_value             name of property to set
2721  * zc_nvlist_src{_size} nvlist of properties to apply
2722  * zc_cookie            received properties flag
2723  *
2724  * outputs:
2725  * zc_nvlist_dst{_size} error for each unapplied received property
2726  */
2727 static int
2728 zfs_ioc_set_prop(zfs_cmd_t *zc)
2729 {
2730         nvlist_t *nvl;
2731         boolean_t received = zc->zc_cookie;
2732         zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2733             ZPROP_SRC_LOCAL);
2734         nvlist_t *errors;
2735         int error;
2736 
2737         if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2738             zc->zc_iflags, &nvl)) != 0)
2739                 return (error);
2740 
2741         if (received) {
2742                 nvlist_t *origprops;
2743 
2744                 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2745                         (void) clear_received_props(zc->zc_name,
2746                             origprops, nvl);
2747                         nvlist_free(origprops);
2748                 }
2749 
2750                 error = dsl_prop_set_hasrecvd(zc->zc_name);
2751         }
2752 
2753         errors = fnvlist_alloc();
2754         if (error == 0)
2755                 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2756 
2757         if (zc->zc_nvlist_dst != NULL && errors != NULL) {
2758                 (void) put_nvlist(zc, errors);
2759         }
2760 
2761         nvlist_free(errors);
2762         nvlist_free(nvl);
2763         return (error);
2764 }
2765 
2766 /*
2767  * inputs:
2768  * zc_name              name of filesystem
2769  * zc_value             name of property to inherit
2770  * zc_cookie            revert to received value if TRUE
2771  *
2772  * outputs:             none
2773  */
2774 static int
2775 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2776 {
2777         const char *propname = zc->zc_value;
2778         zfs_prop_t prop = zfs_name_to_prop(propname);
2779         boolean_t received = zc->zc_cookie;
2780         zprop_source_t source = (received
2781             ? ZPROP_SRC_NONE            /* revert to received value, if any */
2782             : ZPROP_SRC_INHERITED);     /* explicitly inherit */
2783 
2784         if (received) {
2785                 nvlist_t *dummy;
2786                 nvpair_t *pair;
2787                 zprop_type_t type;
2788                 int err;
2789 
2790                 /*
2791                  * zfs_prop_set_special() expects properties in the form of an
2792                  * nvpair with type info.
2793                  */
2794                 if (prop == ZPROP_INVAL) {
2795                         if (!zfs_prop_user(propname))
2796                                 return (SET_ERROR(EINVAL));
2797 
2798                         type = PROP_TYPE_STRING;
2799                 } else if (prop == ZFS_PROP_VOLSIZE ||
2800                     prop == ZFS_PROP_VERSION) {
2801                         return (SET_ERROR(EINVAL));
2802                 } else {
2803                         type = zfs_prop_get_type(prop);
2804                 }
2805 
2806                 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2807 
2808                 switch (type) {
2809                 case PROP_TYPE_STRING:
2810                         VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2811                         break;
2812                 case PROP_TYPE_NUMBER:
2813                 case PROP_TYPE_INDEX:
2814                         VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2815                         break;
2816                 default:
2817                         nvlist_free(dummy);
2818                         return (SET_ERROR(EINVAL));
2819                 }
2820 
2821                 pair = nvlist_next_nvpair(dummy, NULL);
2822                 err = zfs_prop_set_special(zc->zc_name, source, pair);
2823                 nvlist_free(dummy);
2824                 if (err != -1)
2825                         return (err); /* special property already handled */
2826         } else {
2827                 /*
2828                  * Only check this in the non-received case. We want to allow
2829                  * 'inherit -S' to revert non-inheritable properties like quota
2830                  * and reservation to the received or default values even though
2831                  * they are not considered inheritable.
2832                  */
2833                 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2834                         return (SET_ERROR(EINVAL));
2835         }
2836 
2837         /* property name has been validated by zfs_secpolicy_inherit_prop() */
2838         return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source));
2839 }
2840 
2841 static int
2842 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2843 {
2844         nvlist_t *props;
2845         spa_t *spa;
2846         int error;
2847         nvpair_t *pair;
2848 
2849         if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2850             zc->zc_iflags, &props))
2851                 return (error);
2852 
2853         /*
2854          * If the only property is the configfile, then just do a spa_lookup()
2855          * to handle the faulted case.
2856          */
2857         pair = nvlist_next_nvpair(props, NULL);
2858         if (pair != NULL && strcmp(nvpair_name(pair),
2859             zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2860             nvlist_next_nvpair(props, pair) == NULL) {
2861                 mutex_enter(&spa_namespace_lock);
2862                 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2863                         spa_configfile_set(spa, props, B_FALSE);
2864                         spa_config_sync(spa, B_FALSE, B_TRUE);
2865                 }
2866                 mutex_exit(&spa_namespace_lock);
2867                 if (spa != NULL) {
2868                         nvlist_free(props);
2869                         return (0);
2870                 }
2871         }
2872 
2873         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2874                 nvlist_free(props);
2875                 return (error);
2876         }
2877 
2878         error = spa_prop_set(spa, props);
2879 
2880         nvlist_free(props);
2881         spa_close(spa, FTAG);
2882 
2883         return (error);
2884 }
2885 
2886 static int
2887 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2888 {
2889         spa_t *spa;
2890         int error;
2891         nvlist_t *nvp = NULL;
2892 
2893         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2894                 /*
2895                  * If the pool is faulted, there may be properties we can still
2896                  * get (such as altroot and cachefile), so attempt to get them
2897                  * anyway.
2898                  */
2899                 mutex_enter(&spa_namespace_lock);
2900                 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2901                         error = spa_prop_get(spa, &nvp);
2902                 mutex_exit(&spa_namespace_lock);
2903         } else {
2904                 error = spa_prop_get(spa, &nvp);
2905                 spa_close(spa, FTAG);
2906         }
2907 
2908         if (error == 0 && zc->zc_nvlist_dst != NULL)
2909                 error = put_nvlist(zc, nvp);
2910         else
2911                 error = SET_ERROR(EFAULT);
2912 
2913         nvlist_free(nvp);
2914         return (error);
2915 }
2916 
2917 /*
2918  * inputs:
2919  * zc_name              name of filesystem
2920  * zc_nvlist_src{_size} nvlist of delegated permissions
2921  * zc_perm_action       allow/unallow flag
2922  *
2923  * outputs:             none
2924  */
2925 static int
2926 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2927 {
2928         int error;
2929         nvlist_t *fsaclnv = NULL;
2930 
2931         if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2932             zc->zc_iflags, &fsaclnv)) != 0)
2933                 return (error);
2934 
2935         /*
2936          * Verify nvlist is constructed correctly
2937          */
2938         if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2939                 nvlist_free(fsaclnv);
2940                 return (SET_ERROR(EINVAL));
2941         }
2942 
2943         /*
2944          * If we don't have PRIV_SYS_MOUNT, then validate
2945          * that user is allowed to hand out each permission in
2946          * the nvlist(s)
2947          */
2948 
2949         error = secpolicy_zfs(CRED());
2950         if (error != 0) {
2951                 if (zc->zc_perm_action == B_FALSE) {
2952                         error = dsl_deleg_can_allow(zc->zc_name,
2953                             fsaclnv, CRED());
2954                 } else {
2955                         error = dsl_deleg_can_unallow(zc->zc_name,
2956                             fsaclnv, CRED());
2957                 }
2958         }
2959 
2960         if (error == 0)
2961                 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2962 
2963         nvlist_free(fsaclnv);
2964         return (error);
2965 }
2966 
2967 /*
2968  * inputs:
2969  * zc_name              name of filesystem
2970  *
2971  * outputs:
2972  * zc_nvlist_src{_size} nvlist of delegated permissions
2973  */
2974 static int
2975 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
2976 {
2977         nvlist_t *nvp;
2978         int error;
2979 
2980         if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
2981                 error = put_nvlist(zc, nvp);
2982                 nvlist_free(nvp);
2983         }
2984 
2985         return (error);
2986 }
2987 
2988 /*
2989  * Search the vfs list for a specified resource.  Returns a pointer to it
2990  * or NULL if no suitable entry is found. The caller of this routine
2991  * is responsible for releasing the returned vfs pointer.
2992  */
2993 static vfs_t *
2994 zfs_get_vfs(const char *resource)
2995 {
2996         struct vfs *vfsp;
2997         struct vfs *vfs_found = NULL;
2998 
2999         vfs_list_read_lock();
3000         vfsp = rootvfs;
3001         do {
3002                 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
3003                         VFS_HOLD(vfsp);
3004                         vfs_found = vfsp;
3005                         break;
3006                 }
3007                 vfsp = vfsp->vfs_next;
3008         } while (vfsp != rootvfs);
3009         vfs_list_unlock();
3010         return (vfs_found);
3011 }
3012 
3013 /* ARGSUSED */
3014 static void
3015 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
3016 {
3017         zfs_creat_t *zct = arg;
3018 
3019         zfs_create_fs(os, cr, zct->zct_zplprops, tx);
3020 }
3021 
3022 #define ZFS_PROP_UNDEFINED      ((uint64_t)-1)
3023 
3024 /*
3025  * inputs:
3026  * os                   parent objset pointer (NULL if root fs)
3027  * fuids_ok             fuids allowed in this version of the spa?
3028  * sa_ok                SAs allowed in this version of the spa?
3029  * createprops          list of properties requested by creator
3030  *
3031  * outputs:
3032  * zplprops     values for the zplprops we attach to the master node object
3033  * is_ci        true if requested file system will be purely case-insensitive
3034  *
3035  * Determine the settings for utf8only, normalization and
3036  * casesensitivity.  Specific values may have been requested by the
3037  * creator and/or we can inherit values from the parent dataset.  If
3038  * the file system is of too early a vintage, a creator can not
3039  * request settings for these properties, even if the requested
3040  * setting is the default value.  We don't actually want to create dsl
3041  * properties for these, so remove them from the source nvlist after
3042  * processing.
3043  */
3044 static int
3045 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
3046     boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
3047     nvlist_t *zplprops, boolean_t *is_ci)
3048 {
3049         uint64_t sense = ZFS_PROP_UNDEFINED;
3050         uint64_t norm = ZFS_PROP_UNDEFINED;
3051         uint64_t u8 = ZFS_PROP_UNDEFINED;
3052 
3053         ASSERT(zplprops != NULL);
3054 
3055         /*
3056          * Pull out creator prop choices, if any.
3057          */
3058         if (createprops) {
3059                 (void) nvlist_lookup_uint64(createprops,
3060                     zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
3061                 (void) nvlist_lookup_uint64(createprops,
3062                     zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
3063                 (void) nvlist_remove_all(createprops,
3064                     zfs_prop_to_name(ZFS_PROP_NORMALIZE));
3065                 (void) nvlist_lookup_uint64(createprops,
3066                     zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
3067                 (void) nvlist_remove_all(createprops,
3068                     zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
3069                 (void) nvlist_lookup_uint64(createprops,
3070                     zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3071                 (void) nvlist_remove_all(createprops,
3072                     zfs_prop_to_name(ZFS_PROP_CASE));
3073         }
3074 
3075         /*
3076          * If the zpl version requested is whacky or the file system
3077          * or pool is version is too "young" to support normalization
3078          * and the creator tried to set a value for one of the props,
3079          * error out.
3080          */
3081         if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3082             (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3083             (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3084             (zplver < ZPL_VERSION_NORMALIZATION &&
3085             (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3086             sense != ZFS_PROP_UNDEFINED)))
3087                 return (SET_ERROR(ENOTSUP));
3088 
3089         /*
3090          * Put the version in the zplprops
3091          */
3092         VERIFY(nvlist_add_uint64(zplprops,
3093             zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3094 
3095         if (norm == ZFS_PROP_UNDEFINED)
3096                 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
3097         VERIFY(nvlist_add_uint64(zplprops,
3098             zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3099 
3100         /*
3101          * If we're normalizing, names must always be valid UTF-8 strings.
3102          */
3103         if (norm)
3104                 u8 = 1;
3105         if (u8 == ZFS_PROP_UNDEFINED)
3106                 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
3107         VERIFY(nvlist_add_uint64(zplprops,
3108             zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3109 
3110         if (sense == ZFS_PROP_UNDEFINED)
3111                 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
3112         VERIFY(nvlist_add_uint64(zplprops,
3113             zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3114 
3115         if (is_ci)
3116                 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
3117 
3118         return (0);
3119 }
3120 
3121 static int
3122 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3123     nvlist_t *zplprops, boolean_t *is_ci)
3124 {
3125         boolean_t fuids_ok, sa_ok;
3126         uint64_t zplver = ZPL_VERSION;
3127         objset_t *os = NULL;
3128         char parentname[MAXNAMELEN];
3129         char *cp;
3130         spa_t *spa;
3131         uint64_t spa_vers;
3132         int error;
3133 
3134         (void) strlcpy(parentname, dataset, sizeof (parentname));
3135         cp = strrchr(parentname, '/');
3136         ASSERT(cp != NULL);
3137         cp[0] = '\0';
3138 
3139         if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3140                 return (error);
3141 
3142         spa_vers = spa_version(spa);
3143         spa_close(spa, FTAG);
3144 
3145         zplver = zfs_zpl_version_map(spa_vers);
3146         fuids_ok = (zplver >= ZPL_VERSION_FUID);
3147         sa_ok = (zplver >= ZPL_VERSION_SA);
3148 
3149         /*
3150          * Open parent object set so we can inherit zplprop values.
3151          */
3152         if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3153                 return (error);
3154 
3155         error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3156             zplprops, is_ci);
3157         dmu_objset_rele(os, FTAG);
3158         return (error);
3159 }
3160 
3161 static int
3162 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3163     nvlist_t *zplprops, boolean_t *is_ci)
3164 {
3165         boolean_t fuids_ok;
3166         boolean_t sa_ok;
3167         uint64_t zplver = ZPL_VERSION;
3168         int error;
3169 
3170         zplver = zfs_zpl_version_map(spa_vers);
3171         fuids_ok = (zplver >= ZPL_VERSION_FUID);
3172         sa_ok = (zplver >= ZPL_VERSION_SA);
3173 
3174         error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3175             createprops, zplprops, is_ci);
3176         return (error);
3177 }
3178 
3179 /*
3180  * innvl: {
3181  *     "type" -> dmu_objset_type_t (int32)
3182  *     (optional) "props" -> { prop -> value }
3183  * }
3184  *
3185  * outnvl: propname -> error code (int32)
3186  */
3187 static int
3188 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3189 {
3190         int error = 0;
3191         zfs_creat_t zct = { 0 };
3192         nvlist_t *nvprops = NULL;
3193         void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3194         int32_t type32;
3195         dmu_objset_type_t type;
3196         boolean_t is_insensitive = B_FALSE;
3197 
3198         if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3199                 return (SET_ERROR(EINVAL));
3200         type = type32;
3201         (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3202 
3203         switch (type) {
3204         case DMU_OST_ZFS:
3205                 cbfunc = zfs_create_cb;
3206                 break;
3207 
3208         case DMU_OST_ZVOL:
3209                 cbfunc = zvol_create_cb;
3210                 break;
3211 
3212         default:
3213                 cbfunc = NULL;
3214                 break;
3215         }
3216         if (strchr(fsname, '@') ||
3217             strchr(fsname, '%'))
3218                 return (SET_ERROR(EINVAL));
3219 
3220         zct.zct_props = nvprops;
3221 
3222         if (cbfunc == NULL)
3223                 return (SET_ERROR(EINVAL));
3224 
3225         if (type == DMU_OST_ZVOL) {
3226                 uint64_t volsize, volblocksize;
3227 
3228                 if (nvprops == NULL)
3229                         return (SET_ERROR(EINVAL));
3230                 if (nvlist_lookup_uint64(nvprops,
3231                     zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3232                         return (SET_ERROR(EINVAL));
3233 
3234                 if ((error = nvlist_lookup_uint64(nvprops,
3235                     zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3236                     &volblocksize)) != 0 && error != ENOENT)
3237                         return (SET_ERROR(EINVAL));
3238 
3239                 if (error != 0)
3240                         volblocksize = zfs_prop_default_numeric(
3241                             ZFS_PROP_VOLBLOCKSIZE);
3242 
3243                 if ((error = zvol_check_volblocksize(
3244                     volblocksize)) != 0 ||
3245                     (error = zvol_check_volsize(volsize,
3246                     volblocksize)) != 0)
3247                         return (error);
3248         } else if (type == DMU_OST_ZFS) {
3249                 int error;
3250 
3251                 /*
3252                  * We have to have normalization and
3253                  * case-folding flags correct when we do the
3254                  * file system creation, so go figure them out
3255                  * now.
3256                  */
3257                 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3258                     NV_UNIQUE_NAME, KM_SLEEP) == 0);
3259                 error = zfs_fill_zplprops(fsname, nvprops,
3260                     zct.zct_zplprops, &is_insensitive);
3261                 if (error != 0) {
3262                         nvlist_free(zct.zct_zplprops);
3263                         return (error);
3264                 }
3265         }
3266 
3267         error = dmu_objset_create(fsname, type,
3268             is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3269         nvlist_free(zct.zct_zplprops);
3270 
3271         /*
3272          * It would be nice to do this atomically.
3273          */
3274         if (error == 0) {
3275                 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3276                     nvprops, outnvl);
3277                 if (error != 0)
3278                         (void) dsl_destroy_head(fsname);
3279         }
3280         return (error);
3281 }
3282 
3283 /*
3284  * innvl: {
3285  *     "origin" -> name of origin snapshot
3286  *     (optional) "props" -> { prop -> value }
3287  * }
3288  *
3289  * outnvl: propname -> error code (int32)
3290  */
3291 static int
3292 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3293 {
3294         int error = 0;
3295         nvlist_t *nvprops = NULL;
3296         char *origin_name;
3297 
3298         if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3299                 return (SET_ERROR(EINVAL));
3300         (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3301 
3302         if (strchr(fsname, '@') ||
3303             strchr(fsname, '%'))
3304                 return (SET_ERROR(EINVAL));
3305 
3306         if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3307                 return (SET_ERROR(EINVAL));
3308         error = dmu_objset_clone(fsname, origin_name);
3309         if (error != 0)
3310                 return (error);
3311 
3312         /*
3313          * It would be nice to do this atomically.
3314          */
3315         if (error == 0) {
3316                 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3317                     nvprops, outnvl);
3318                 if (error != 0)
3319                         (void) dsl_destroy_head(fsname);
3320         }
3321         return (error);
3322 }
3323 
3324 /*
3325  * innvl: {
3326  *     "snaps" -> { snapshot1, snapshot2 }
3327  *     (optional) "props" -> { prop -> value (string) }
3328  * }
3329  *
3330  * outnvl: snapshot -> error code (int32)
3331  */
3332 static int
3333 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3334 {
3335         nvlist_t *snaps;
3336         nvlist_t *props = NULL;
3337         int error, poollen;
3338         nvpair_t *pair;
3339 
3340         (void) nvlist_lookup_nvlist(innvl, "props", &props);
3341         if ((error = zfs_check_userprops(poolname, props)) != 0)
3342                 return (error);
3343 
3344         if (!nvlist_empty(props) &&
3345             zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3346                 return (SET_ERROR(ENOTSUP));
3347 
3348         if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3349                 return (SET_ERROR(EINVAL));
3350         poollen = strlen(poolname);
3351         for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3352             pair = nvlist_next_nvpair(snaps, pair)) {
3353                 const char *name = nvpair_name(pair);
3354                 const char *cp = strchr(name, '@');
3355 
3356                 /*
3357                  * The snap name must contain an @, and the part after it must
3358                  * contain only valid characters.
3359                  */
3360                 if (cp == NULL ||
3361                     zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3362                         return (SET_ERROR(EINVAL));
3363 
3364                 /*
3365                  * The snap must be in the specified pool.
3366                  */
3367                 if (strncmp(name, poolname, poollen) != 0 ||
3368                     (name[poollen] != '/' && name[poollen] != '@'))
3369                         return (SET_ERROR(EXDEV));
3370 
3371                 /* This must be the only snap of this fs. */
3372                 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3373                     pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3374                         if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3375                             == 0) {
3376                                 return (SET_ERROR(EXDEV));
3377                         }
3378                 }
3379         }
3380 
3381         error = dsl_dataset_snapshot(snaps, props, outnvl);
3382         return (error);
3383 }
3384 
3385 /*
3386  * innvl: "message" -> string
3387  */
3388 /* ARGSUSED */
3389 static int
3390 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3391 {
3392         char *message;
3393         spa_t *spa;
3394         int error;
3395         char *poolname;
3396 
3397         /*
3398          * The poolname in the ioctl is not set, we get it from the TSD,
3399          * which was set at the end of the last successful ioctl that allows
3400          * logging.  The secpolicy func already checked that it is set.
3401          * Only one log ioctl is allowed after each successful ioctl, so
3402          * we clear the TSD here.
3403          */
3404         poolname = tsd_get(zfs_allow_log_key);
3405         (void) tsd_set(zfs_allow_log_key, NULL);
3406         error = spa_open(poolname, &spa, FTAG);
3407         strfree(poolname);
3408         if (error != 0)
3409                 return (error);
3410 
3411         if (nvlist_lookup_string(innvl, "message", &message) != 0)  {
3412                 spa_close(spa, FTAG);
3413                 return (SET_ERROR(EINVAL));
3414         }
3415 
3416         if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3417                 spa_close(spa, FTAG);
3418                 return (SET_ERROR(ENOTSUP));
3419         }
3420 
3421         error = spa_history_log(spa, message);
3422         spa_close(spa, FTAG);
3423         return (error);
3424 }
3425 
3426 /*
3427  * The dp_config_rwlock must not be held when calling this, because the
3428  * unmount may need to write out data.
3429  *
3430  * This function is best-effort.  Callers must deal gracefully if it
3431  * remains mounted (or is remounted after this call).
3432  *
3433  * Returns 0 if the argument is not a snapshot, or it is not currently a
3434  * filesystem, or we were able to unmount it.  Returns error code otherwise.
3435  */
3436 int
3437 zfs_unmount_snap(const char *snapname)
3438 {
3439         vfs_t *vfsp;
3440         zfsvfs_t *zfsvfs;
3441         int err;
3442 
3443         if (strchr(snapname, '@') == NULL)
3444                 return (0);
3445 
3446         vfsp = zfs_get_vfs(snapname);
3447         if (vfsp == NULL)
3448                 return (0);
3449 
3450         zfsvfs = vfsp->vfs_data;
3451         ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os)));
3452 
3453         err = vn_vfswlock(vfsp->vfs_vnodecovered);
3454         VFS_RELE(vfsp);
3455         if (err != 0)
3456                 return (SET_ERROR(err));
3457 
3458         /*
3459          * Always force the unmount for snapshots.
3460          */
3461         (void) dounmount(vfsp, MS_FORCE, kcred);
3462         return (0);
3463 }
3464 
3465 /* ARGSUSED */
3466 static int
3467 zfs_unmount_snap_cb(const char *snapname, void *arg)
3468 {
3469         return (zfs_unmount_snap(snapname));
3470 }
3471 
3472 /*
3473  * When a clone is destroyed, its origin may also need to be destroyed,
3474  * in which case it must be unmounted.  This routine will do that unmount
3475  * if necessary.
3476  */
3477 void
3478 zfs_destroy_unmount_origin(const char *fsname)
3479 {
3480         int error;
3481         objset_t *os;
3482         dsl_dataset_t *ds;
3483 
3484         error = dmu_objset_hold(fsname, FTAG, &os);
3485         if (error != 0)
3486                 return;
3487         ds = dmu_objset_ds(os);
3488         if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3489                 char originname[MAXNAMELEN];
3490                 dsl_dataset_name(ds->ds_prev, originname);
3491                 dmu_objset_rele(os, FTAG);
3492                 (void) zfs_unmount_snap(originname);
3493         } else {
3494                 dmu_objset_rele(os, FTAG);
3495         }
3496 }
3497 
3498 /*
3499  * innvl: {
3500  *     "snaps" -> { snapshot1, snapshot2 }
3501  *     (optional boolean) "defer"
3502  * }
3503  *
3504  * outnvl: snapshot -> error code (int32)
3505  *
3506  */
3507 /* ARGSUSED */
3508 static int
3509 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3510 {
3511         nvlist_t *snaps;
3512         nvpair_t *pair;
3513         boolean_t defer;
3514 
3515         if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3516                 return (SET_ERROR(EINVAL));
3517         defer = nvlist_exists(innvl, "defer");
3518 
3519         for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3520             pair = nvlist_next_nvpair(snaps, pair)) {
3521                 (void) zfs_unmount_snap(nvpair_name(pair));
3522         }
3523 
3524         return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3525 }
3526 
3527 /*
3528  * Create bookmarks.  Bookmark names are of the form <fs>#<bmark>.
3529  * All bookmarks must be in the same pool.
3530  *
3531  * innvl: {
3532  *     bookmark1 -> snapshot1, bookmark2 -> snapshot2
3533  * }
3534  *
3535  * outnvl: bookmark -> error code (int32)
3536  *
3537  */
3538 /* ARGSUSED */
3539 static int
3540 zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3541 {
3542         for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3543             pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3544                 char *snap_name;
3545 
3546                 /*
3547                  * Verify the snapshot argument.
3548                  */
3549                 if (nvpair_value_string(pair, &snap_name) != 0)
3550                         return (SET_ERROR(EINVAL));
3551 
3552 
3553                 /* Verify that the keys (bookmarks) are unique */
3554                 for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair);
3555                     pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) {
3556                         if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0)
3557                                 return (SET_ERROR(EINVAL));
3558                 }
3559         }
3560 
3561         return (dsl_bookmark_create(innvl, outnvl));
3562 }
3563 
3564 /*
3565  * innvl: {
3566  *     property 1, property 2, ...
3567  * }
3568  *
3569  * outnvl: {
3570  *     bookmark name 1 -> { property 1, property 2, ... },
3571  *     bookmark name 2 -> { property 1, property 2, ... }
3572  * }
3573  *
3574  */
3575 static int
3576 zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3577 {
3578         return (dsl_get_bookmarks(fsname, innvl, outnvl));
3579 }
3580 
3581 /*
3582  * innvl: {
3583  *     bookmark name 1, bookmark name 2
3584  * }
3585  *
3586  * outnvl: bookmark -> error code (int32)
3587  *
3588  */
3589 static int
3590 zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl,
3591     nvlist_t *outnvl)
3592 {
3593         int error, poollen;
3594 
3595         poollen = strlen(poolname);
3596         for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3597             pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3598                 const char *name = nvpair_name(pair);
3599                 const char *cp = strchr(name, '#');
3600 
3601                 /*
3602                  * The bookmark name must contain an #, and the part after it
3603                  * must contain only valid characters.
3604                  */
3605                 if (cp == NULL ||
3606                     zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3607                         return (SET_ERROR(EINVAL));
3608 
3609                 /*
3610                  * The bookmark must be in the specified pool.
3611                  */
3612                 if (strncmp(name, poolname, poollen) != 0 ||
3613                     (name[poollen] != '/' && name[poollen] != '#'))
3614                         return (SET_ERROR(EXDEV));
3615         }
3616 
3617         error = dsl_bookmark_destroy(innvl, outnvl);
3618         return (error);
3619 }
3620 
3621 /*
3622  * inputs:
3623  * zc_name              name of dataset to destroy
3624  * zc_objset_type       type of objset
3625  * zc_defer_destroy     mark for deferred destroy
3626  *
3627  * outputs:             none
3628  */
3629 static int
3630 zfs_ioc_destroy(zfs_cmd_t *zc)
3631 {
3632         int err;
3633 
3634         if (zc->zc_objset_type == DMU_OST_ZFS) {
3635                 err = zfs_unmount_snap(zc->zc_name);
3636                 if (err != 0)
3637                         return (err);
3638         }
3639 
3640         if (strchr(zc->zc_name, '@'))
3641                 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3642         else
3643                 err = dsl_destroy_head(zc->zc_name);
3644         if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3645                 (void) zvol_remove_minor(zc->zc_name);
3646         return (err);
3647 }
3648 
3649 /*
3650  * fsname is name of dataset to rollback (to most recent snapshot)
3651  *
3652  * innvl is not used.
3653  *
3654  * outnvl: "target" -> name of most recent snapshot
3655  * }
3656  */
3657 /* ARGSUSED */
3658 static int
3659 zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl)
3660 {
3661         zfsvfs_t *zfsvfs;
3662         int error;
3663 
3664         if (getzfsvfs(fsname, &zfsvfs) == 0) {
3665                 error = zfs_suspend_fs(zfsvfs);
3666                 if (error == 0) {
3667                         int resume_err;
3668 
3669                         error = dsl_dataset_rollback(fsname, zfsvfs, outnvl);
3670                         resume_err = zfs_resume_fs(zfsvfs, fsname);
3671                         error = error ? error : resume_err;
3672                 }
3673                 VFS_RELE(zfsvfs->z_vfs);
3674         } else {
3675                 error = dsl_dataset_rollback(fsname, NULL, outnvl);
3676         }
3677         return (error);
3678 }
3679 
3680 static int
3681 recursive_unmount(const char *fsname, void *arg)
3682 {
3683         const char *snapname = arg;
3684         char fullname[MAXNAMELEN];
3685 
3686         (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname);
3687         return (zfs_unmount_snap(fullname));
3688 }
3689 
3690 /*
3691  * inputs:
3692  * zc_name      old name of dataset
3693  * zc_value     new name of dataset
3694  * zc_cookie    recursive flag (only valid for snapshots)
3695  *
3696  * outputs:     none
3697  */
3698 static int
3699 zfs_ioc_rename(zfs_cmd_t *zc)
3700 {
3701         boolean_t recursive = zc->zc_cookie & 1;
3702         char *at;
3703 
3704         zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3705         if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3706             strchr(zc->zc_value, '%'))
3707                 return (SET_ERROR(EINVAL));
3708 
3709         at = strchr(zc->zc_name, '@');
3710         if (at != NULL) {
3711                 /* snaps must be in same fs */
3712                 int error;
3713 
3714                 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
3715                         return (SET_ERROR(EXDEV));
3716                 *at = '\0';
3717                 if (zc->zc_objset_type == DMU_OST_ZFS) {
3718                         error = dmu_objset_find(zc->zc_name,
3719                             recursive_unmount, at + 1,
3720                             recursive ? DS_FIND_CHILDREN : 0);
3721                         if (error != 0) {
3722                                 *at = '@';
3723                                 return (error);
3724                         }
3725                 }
3726                 error = dsl_dataset_rename_snapshot(zc->zc_name,
3727                     at + 1, strchr(zc->zc_value, '@') + 1, recursive);
3728                 *at = '@';
3729 
3730                 return (error);
3731         } else {
3732                 if (zc->zc_objset_type == DMU_OST_ZVOL)
3733                         (void) zvol_remove_minor(zc->zc_name);
3734                 return (dsl_dir_rename(zc->zc_name, zc->zc_value));
3735         }
3736 }
3737 
3738 static int
3739 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3740 {
3741         const char *propname = nvpair_name(pair);
3742         boolean_t issnap = (strchr(dsname, '@') != NULL);
3743         zfs_prop_t prop = zfs_name_to_prop(propname);
3744         uint64_t intval;
3745         int err;
3746 
3747         if (prop == ZPROP_INVAL) {
3748                 if (zfs_prop_user(propname)) {
3749                         if (err = zfs_secpolicy_write_perms(dsname,
3750                             ZFS_DELEG_PERM_USERPROP, cr))
3751                                 return (err);
3752                         return (0);
3753                 }
3754 
3755                 if (!issnap && zfs_prop_userquota(propname)) {
3756                         const char *perm = NULL;
3757                         const char *uq_prefix =
3758                             zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3759                         const char *gq_prefix =
3760                             zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3761 
3762                         if (strncmp(propname, uq_prefix,
3763                             strlen(uq_prefix)) == 0) {
3764                                 perm = ZFS_DELEG_PERM_USERQUOTA;
3765                         } else if (strncmp(propname, gq_prefix,
3766                             strlen(gq_prefix)) == 0) {
3767                                 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3768                         } else {
3769                                 /* USERUSED and GROUPUSED are read-only */
3770                                 return (SET_ERROR(EINVAL));
3771                         }
3772 
3773                         if (err = zfs_secpolicy_write_perms(dsname, perm, cr))
3774                                 return (err);
3775                         return (0);
3776                 }
3777 
3778                 return (SET_ERROR(EINVAL));
3779         }
3780 
3781         if (issnap)
3782                 return (SET_ERROR(EINVAL));
3783 
3784         if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3785                 /*
3786                  * dsl_prop_get_all_impl() returns properties in this
3787                  * format.
3788                  */
3789                 nvlist_t *attrs;
3790                 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3791                 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3792                     &pair) == 0);
3793         }
3794 
3795         /*
3796          * Check that this value is valid for this pool version
3797          */
3798         switch (prop) {
3799         case ZFS_PROP_COMPRESSION:
3800                 /*
3801                  * If the user specified gzip compression, make sure
3802                  * the SPA supports it. We ignore any errors here since
3803                  * we'll catch them later.
3804                  */
3805                 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3806                     nvpair_value_uint64(pair, &intval) == 0) {
3807                         if (intval >= ZIO_COMPRESS_GZIP_1 &&
3808                             intval <= ZIO_COMPRESS_GZIP_9 &&
3809                             zfs_earlier_version(dsname,
3810                             SPA_VERSION_GZIP_COMPRESSION)) {
3811                                 return (SET_ERROR(ENOTSUP));
3812                         }
3813 
3814                         if (intval == ZIO_COMPRESS_ZLE &&
3815                             zfs_earlier_version(dsname,
3816                             SPA_VERSION_ZLE_COMPRESSION))
3817                                 return (SET_ERROR(ENOTSUP));
3818 
3819                         if (intval == ZIO_COMPRESS_LZ4) {
3820                                 spa_t *spa;
3821 
3822                                 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3823                                         return (err);
3824 
3825                                 if (!spa_feature_is_enabled(spa,
3826                                     SPA_FEATURE_LZ4_COMPRESS)) {
3827                                         spa_close(spa, FTAG);
3828                                         return (SET_ERROR(ENOTSUP));
3829                                 }
3830                                 spa_close(spa, FTAG);
3831                         }
3832 
3833                         /*
3834                          * If this is a bootable dataset then
3835                          * verify that the compression algorithm
3836                          * is supported for booting. We must return
3837                          * something other than ENOTSUP since it
3838                          * implies a downrev pool version.
3839                          */
3840                         if (zfs_is_bootfs(dsname) &&
3841                             !BOOTFS_COMPRESS_VALID(intval)) {
3842                                 return (SET_ERROR(ERANGE));
3843                         }
3844                 }
3845                 break;
3846 
3847         case ZFS_PROP_COPIES:
3848                 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3849                         return (SET_ERROR(ENOTSUP));
3850                 break;
3851 
3852         case ZFS_PROP_DEDUP:
3853                 if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3854                         return (SET_ERROR(ENOTSUP));
3855                 break;
3856 
3857         case ZFS_PROP_SHARESMB:
3858                 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3859                         return (SET_ERROR(ENOTSUP));
3860                 break;
3861 
3862         case ZFS_PROP_ACLINHERIT:
3863                 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3864                     nvpair_value_uint64(pair, &intval) == 0) {
3865                         if (intval == ZFS_ACL_PASSTHROUGH_X &&
3866                             zfs_earlier_version(dsname,
3867                             SPA_VERSION_PASSTHROUGH_X))
3868                                 return (SET_ERROR(ENOTSUP));
3869                 }
3870                 break;
3871         }
3872 
3873         return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3874 }
3875 
3876 /*
3877  * Checks for a race condition to make sure we don't increment a feature flag
3878  * multiple times.
3879  */
3880 static int
3881 zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx)
3882 {
3883         spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3884         spa_feature_t *featurep = arg;
3885 
3886         if (!spa_feature_is_active(spa, *featurep))
3887                 return (0);
3888         else
3889                 return (SET_ERROR(EBUSY));
3890 }
3891 
3892 /*
3893  * The callback invoked on feature activation in the sync task caused by
3894  * zfs_prop_activate_feature.
3895  */
3896 static void
3897 zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx)
3898 {
3899         spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3900         spa_feature_t *featurep = arg;
3901 
3902         spa_feature_incr(spa, *featurep, tx);
3903 }
3904 
3905 /*
3906  * Activates a feature on a pool in response to a property setting. This
3907  * creates a new sync task which modifies the pool to reflect the feature
3908  * as being active.
3909  */
3910 static int
3911 zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature)
3912 {
3913         int err;
3914 
3915         /* EBUSY here indicates that the feature is already active */
3916         err = dsl_sync_task(spa_name(spa),
3917             zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync,
3918             &feature, 2);
3919 
3920         if (err != 0 && err != EBUSY)
3921                 return (err);
3922         else
3923                 return (0);
3924 }
3925 
3926 /*
3927  * Removes properties from the given props list that fail permission checks
3928  * needed to clear them and to restore them in case of a receive error. For each
3929  * property, make sure we have both set and inherit permissions.
3930  *
3931  * Returns the first error encountered if any permission checks fail. If the
3932  * caller provides a non-NULL errlist, it also gives the complete list of names
3933  * of all the properties that failed a permission check along with the
3934  * corresponding error numbers. The caller is responsible for freeing the
3935  * returned errlist.
3936  *
3937  * If every property checks out successfully, zero is returned and the list
3938  * pointed at by errlist is NULL.
3939  */
3940 static int
3941 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
3942 {
3943         zfs_cmd_t *zc;
3944         nvpair_t *pair, *next_pair;
3945         nvlist_t *errors;
3946         int err, rv = 0;
3947 
3948         if (props == NULL)
3949                 return (0);
3950 
3951         VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3952 
3953         zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
3954         (void) strcpy(zc->zc_name, dataset);
3955         pair = nvlist_next_nvpair(props, NULL);
3956         while (pair != NULL) {
3957                 next_pair = nvlist_next_nvpair(props, pair);
3958 
3959                 (void) strcpy(zc->zc_value, nvpair_name(pair));
3960                 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
3961                     (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
3962                         VERIFY(nvlist_remove_nvpair(props, pair) == 0);
3963                         VERIFY(nvlist_add_int32(errors,
3964                             zc->zc_value, err) == 0);
3965                 }
3966                 pair = next_pair;
3967         }
3968         kmem_free(zc, sizeof (zfs_cmd_t));
3969 
3970         if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
3971                 nvlist_free(errors);
3972                 errors = NULL;
3973         } else {
3974                 VERIFY(nvpair_value_int32(pair, &rv) == 0);
3975         }
3976 
3977         if (errlist == NULL)
3978                 nvlist_free(errors);
3979         else
3980                 *errlist = errors;
3981 
3982         return (rv);
3983 }
3984 
3985 static boolean_t
3986 propval_equals(nvpair_t *p1, nvpair_t *p2)
3987 {
3988         if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
3989                 /* dsl_prop_get_all_impl() format */
3990                 nvlist_t *attrs;
3991                 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
3992                 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3993                     &p1) == 0);
3994         }
3995 
3996         if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
3997                 nvlist_t *attrs;
3998                 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
3999                 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4000                     &p2) == 0);
4001         }
4002 
4003         if (nvpair_type(p1) != nvpair_type(p2))
4004                 return (B_FALSE);
4005 
4006         if (nvpair_type(p1) == DATA_TYPE_STRING) {
4007                 char *valstr1, *valstr2;
4008 
4009                 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
4010                 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
4011                 return (strcmp(valstr1, valstr2) == 0);
4012         } else {
4013                 uint64_t intval1, intval2;
4014 
4015                 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
4016                 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
4017                 return (intval1 == intval2);
4018         }
4019 }
4020 
4021 /*
4022  * Remove properties from props if they are not going to change (as determined
4023  * by comparison with origprops). Remove them from origprops as well, since we
4024  * do not need to clear or restore properties that won't change.
4025  */
4026 static void
4027 props_reduce(nvlist_t *props, nvlist_t *origprops)
4028 {
4029         nvpair_t *pair, *next_pair;
4030 
4031         if (origprops == NULL)
4032                 return; /* all props need to be received */
4033 
4034         pair = nvlist_next_nvpair(props, NULL);
4035         while (pair != NULL) {
4036                 const char *propname = nvpair_name(pair);
4037                 nvpair_t *match;
4038 
4039                 next_pair = nvlist_next_nvpair(props, pair);
4040 
4041                 if ((nvlist_lookup_nvpair(origprops, propname,
4042                     &match) != 0) || !propval_equals(pair, match))
4043                         goto next; /* need to set received value */
4044 
4045                 /* don't clear the existing received value */
4046                 (void) nvlist_remove_nvpair(origprops, match);
4047                 /* don't bother receiving the property */
4048                 (void) nvlist_remove_nvpair(props, pair);
4049 next:
4050                 pair = next_pair;
4051         }
4052 }
4053 
4054 #ifdef  DEBUG
4055 static boolean_t zfs_ioc_recv_inject_err;
4056 #endif
4057 
4058 /*
4059  * inputs:
4060  * zc_name              name of containing filesystem
4061  * zc_nvlist_src{_size} nvlist of properties to apply
4062  * zc_value             name of snapshot to create
4063  * zc_string            name of clone origin (if DRR_FLAG_CLONE)
4064  * zc_cookie            file descriptor to recv from
4065  * zc_begin_record      the BEGIN record of the stream (not byteswapped)
4066  * zc_guid              force flag
4067  * zc_cleanup_fd        cleanup-on-exit file descriptor
4068  * zc_action_handle     handle for this guid/ds mapping (or zero on first call)
4069  *
4070  * outputs:
4071  * zc_cookie            number of bytes read
4072  * zc_nvlist_dst{_size} error for each unapplied received property
4073  * zc_obj               zprop_errflags_t
4074  * zc_action_handle     handle for this guid/ds mapping
4075  */
4076 static int
4077 zfs_ioc_recv(zfs_cmd_t *zc)
4078 {
4079         file_t *fp;
4080         dmu_recv_cookie_t drc;
4081         boolean_t force = (boolean_t)zc->zc_guid;
4082         int fd;
4083         int error = 0;
4084         int props_error = 0;
4085         nvlist_t *errors;
4086         offset_t off;
4087         nvlist_t *props = NULL; /* sent properties */
4088         nvlist_t *origprops = NULL; /* existing properties */
4089         char *origin = NULL;
4090         char *tosnap;
4091         char tofs[ZFS_MAXNAMELEN];
4092         boolean_t first_recvd_props = B_FALSE;
4093 
4094         if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
4095             strchr(zc->zc_value, '@') == NULL ||
4096             strchr(zc->zc_value, '%'))
4097                 return (SET_ERROR(EINVAL));
4098 
4099         (void) strcpy(tofs, zc->zc_value);
4100         tosnap = strchr(tofs, '@');
4101         *tosnap++ = '\0';
4102 
4103         if (zc->zc_nvlist_src != NULL &&
4104             (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
4105             zc->zc_iflags, &props)) != 0)
4106                 return (error);
4107 
4108         fd = zc->zc_cookie;
4109         fp = getf(fd);
4110         if (fp == NULL) {
4111                 nvlist_free(props);
4112                 return (SET_ERROR(EBADF));
4113         }
4114 
4115         VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4116 
4117         if (zc->zc_string[0])
4118                 origin = zc->zc_string;
4119 
4120         error = dmu_recv_begin(tofs, tosnap,
4121             &zc->zc_begin_record, force, origin, &drc);
4122         if (error != 0)
4123                 goto out;
4124 
4125         /*
4126          * Set properties before we receive the stream so that they are applied
4127          * to the new data. Note that we must call dmu_recv_stream() if
4128          * dmu_recv_begin() succeeds.
4129          */
4130         if (props != NULL && !drc.drc_newfs) {
4131                 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
4132                     SPA_VERSION_RECVD_PROPS &&
4133                     !dsl_prop_get_hasrecvd(tofs))
4134                         first_recvd_props = B_TRUE;
4135 
4136                 /*
4137                  * If new received properties are supplied, they are to
4138                  * completely replace the existing received properties, so stash
4139                  * away the existing ones.
4140                  */
4141                 if (dsl_prop_get_received(tofs, &origprops) == 0) {
4142                         nvlist_t *errlist = NULL;
4143                         /*
4144                          * Don't bother writing a property if its value won't
4145                          * change (and avoid the unnecessary security checks).
4146                          *
4147                          * The first receive after SPA_VERSION_RECVD_PROPS is a
4148                          * special case where we blow away all local properties
4149                          * regardless.
4150                          */
4151                         if (!first_recvd_props)
4152                                 props_reduce(props, origprops);
4153                         if (zfs_check_clearable(tofs, origprops, &errlist) != 0)
4154                                 (void) nvlist_merge(errors, errlist, 0);
4155                         nvlist_free(errlist);
4156 
4157                         if (clear_received_props(tofs, origprops,
4158                             first_recvd_props ? NULL : props) != 0)
4159                                 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4160                 } else {
4161                         zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4162                 }
4163         }
4164 
4165         if (props != NULL) {
4166                 props_error = dsl_prop_set_hasrecvd(tofs);
4167 
4168                 if (props_error == 0) {
4169                         (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4170                             props, errors);
4171                 }
4172         }
4173 
4174         if (zc->zc_nvlist_dst_size != 0 &&
4175             (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4176             put_nvlist(zc, errors) != 0)) {
4177                 /*
4178                  * Caller made zc->zc_nvlist_dst less than the minimum expected
4179                  * size or supplied an invalid address.
4180                  */
4181                 props_error = SET_ERROR(EINVAL);
4182         }
4183 
4184         off = fp->f_offset;
4185         error = dmu_recv_stream(&drc, fp->f_vnode, &off, zc->zc_cleanup_fd,
4186             &zc->zc_action_handle);
4187 
4188         if (error == 0) {
4189                 zfsvfs_t *zfsvfs = NULL;
4190 
4191                 if (getzfsvfs(tofs, &zfsvfs) == 0) {
4192                         /* online recv */
4193                         int end_err;
4194 
4195                         error = zfs_suspend_fs(zfsvfs);
4196                         /*
4197                          * If the suspend fails, then the recv_end will
4198                          * likely also fail, and clean up after itself.
4199                          */
4200                         end_err = dmu_recv_end(&drc, zfsvfs);
4201                         if (error == 0)
4202                                 error = zfs_resume_fs(zfsvfs, tofs);
4203                         error = error ? error : end_err;
4204                         VFS_RELE(zfsvfs->z_vfs);
4205                 } else {
4206                         error = dmu_recv_end(&drc, NULL);
4207                 }
4208         }
4209 
4210         zc->zc_cookie = off - fp->f_offset;
4211         if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4212                 fp->f_offset = off;
4213 
4214 #ifdef  DEBUG
4215         if (zfs_ioc_recv_inject_err) {
4216                 zfs_ioc_recv_inject_err = B_FALSE;
4217                 error = 1;
4218         }
4219 #endif
4220         /*
4221          * On error, restore the original props.
4222          */
4223         if (error != 0 && props != NULL && !drc.drc_newfs) {
4224                 if (clear_received_props(tofs, props, NULL) != 0) {
4225                         /*
4226                          * We failed to clear the received properties.
4227                          * Since we may have left a $recvd value on the
4228                          * system, we can't clear the $hasrecvd flag.
4229                          */
4230                         zc->zc_obj |= ZPROP_ERR_NORESTORE;
4231                 } else if (first_recvd_props) {
4232                         dsl_prop_unset_hasrecvd(tofs);
4233                 }
4234 
4235                 if (origprops == NULL && !drc.drc_newfs) {
4236                         /* We failed to stash the original properties. */
4237                         zc->zc_obj |= ZPROP_ERR_NORESTORE;
4238                 }
4239 
4240                 /*
4241                  * dsl_props_set() will not convert RECEIVED to LOCAL on or
4242                  * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4243                  * explictly if we're restoring local properties cleared in the
4244                  * first new-style receive.
4245                  */
4246                 if (origprops != NULL &&
4247                     zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4248                     ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4249                     origprops, NULL) != 0) {
4250                         /*
4251                          * We stashed the original properties but failed to
4252                          * restore them.
4253                          */
4254                         zc->zc_obj |= ZPROP_ERR_NORESTORE;
4255                 }
4256         }
4257 out:
4258         nvlist_free(props);
4259         nvlist_free(origprops);
4260         nvlist_free(errors);
4261         releasef(fd);
4262 
4263         if (error == 0)
4264                 error = props_error;
4265 
4266         return (error);
4267 }
4268 
4269 /*
4270  * inputs:
4271  * zc_name      name of snapshot to send
4272  * zc_cookie    file descriptor to send stream to
4273  * zc_obj       fromorigin flag (mutually exclusive with zc_fromobj)
4274  * zc_sendobj   objsetid of snapshot to send
4275  * zc_fromobj   objsetid of incremental fromsnap (may be zero)
4276  * zc_guid      if set, estimate size of stream only.  zc_cookie is ignored.
4277  *              output size in zc_objset_type.
4278  *
4279  * outputs:
4280  * zc_objset_type       estimated size, if zc_guid is set
4281  */
4282 static int
4283 zfs_ioc_send(zfs_cmd_t *zc)
4284 {
4285         int error;
4286         offset_t off;
4287         boolean_t estimate = (zc->zc_guid != 0);
4288 
4289         if (zc->zc_obj != 0) {
4290                 dsl_pool_t *dp;
4291                 dsl_dataset_t *tosnap;
4292 
4293                 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4294                 if (error != 0)
4295                         return (error);
4296 
4297                 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4298                 if (error != 0) {
4299                         dsl_pool_rele(dp, FTAG);
4300                         return (error);
4301                 }
4302 
4303                 if (dsl_dir_is_clone(tosnap->ds_dir))
4304                         zc->zc_fromobj = tosnap->ds_dir->dd_phys->dd_origin_obj;
4305                 dsl_dataset_rele(tosnap, FTAG);
4306                 dsl_pool_rele(dp, FTAG);
4307         }
4308 
4309         if (estimate) {
4310                 dsl_pool_t *dp;
4311                 dsl_dataset_t *tosnap;
4312                 dsl_dataset_t *fromsnap = NULL;
4313 
4314                 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4315                 if (error != 0)
4316                         return (error);
4317 
4318                 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4319                 if (error != 0) {
4320                         dsl_pool_rele(dp, FTAG);
4321                         return (error);
4322                 }
4323 
4324                 if (zc->zc_fromobj != 0) {
4325                         error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
4326                             FTAG, &fromsnap);
4327                         if (error != 0) {
4328                                 dsl_dataset_rele(tosnap, FTAG);
4329                                 dsl_pool_rele(dp, FTAG);
4330                                 return (error);
4331                         }
4332                 }
4333 
4334                 error = dmu_send_estimate(tosnap, fromsnap,
4335                     &zc->zc_objset_type);
4336 
4337                 if (fromsnap != NULL)
4338                         dsl_dataset_rele(fromsnap, FTAG);
4339                 dsl_dataset_rele(tosnap, FTAG);
4340                 dsl_pool_rele(dp, FTAG);
4341         } else {
4342                 file_t *fp = getf(zc->zc_cookie);
4343                 if (fp == NULL)
4344                         return (SET_ERROR(EBADF));
4345 
4346                 off = fp->f_offset;
4347                 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
4348                     zc->zc_fromobj, zc->zc_cookie, fp->f_vnode, &off);
4349 
4350                 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4351                         fp->f_offset = off;
4352                 releasef(zc->zc_cookie);
4353         }
4354         return (error);
4355 }
4356 
4357 /*
4358  * inputs:
4359  * zc_name      name of snapshot on which to report progress
4360  * zc_cookie    file descriptor of send stream
4361  *
4362  * outputs:
4363  * zc_cookie    number of bytes written in send stream thus far
4364  */
4365 static int
4366 zfs_ioc_send_progress(zfs_cmd_t *zc)
4367 {
4368         dsl_pool_t *dp;
4369         dsl_dataset_t *ds;
4370         dmu_sendarg_t *dsp = NULL;
4371         int error;
4372 
4373         error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4374         if (error != 0)
4375                 return (error);
4376 
4377         error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4378         if (error != 0) {
4379                 dsl_pool_rele(dp, FTAG);
4380                 return (error);
4381         }
4382 
4383         mutex_enter(&ds->ds_sendstream_lock);
4384 
4385         /*
4386          * Iterate over all the send streams currently active on this dataset.
4387          * If there's one which matches the specified file descriptor _and_ the
4388          * stream was started by the current process, return the progress of
4389          * that stream.
4390          */
4391         for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4392             dsp = list_next(&ds->ds_sendstreams, dsp)) {
4393                 if (dsp->dsa_outfd == zc->zc_cookie &&
4394                     dsp->dsa_proc == curproc)
4395                         break;
4396         }
4397 
4398         if (dsp != NULL)
4399                 zc->zc_cookie = *(dsp->dsa_off);
4400         else
4401                 error = SET_ERROR(ENOENT);
4402 
4403         mutex_exit(&ds->ds_sendstream_lock);
4404         dsl_dataset_rele(ds, FTAG);
4405         dsl_pool_rele(dp, FTAG);
4406         return (error);
4407 }
4408 
4409 static int
4410 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4411 {
4412         int id, error;
4413 
4414         error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4415             &zc->zc_inject_record);
4416 
4417         if (error == 0)
4418                 zc->zc_guid = (uint64_t)id;
4419 
4420         return (error);
4421 }
4422 
4423 static int
4424 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4425 {
4426         return (zio_clear_fault((int)zc->zc_guid));
4427 }
4428 
4429 static int
4430 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4431 {
4432         int id = (int)zc->zc_guid;
4433         int error;
4434 
4435         error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4436             &zc->zc_inject_record);
4437 
4438         zc->zc_guid = id;
4439 
4440         return (error);
4441 }
4442 
4443 static int
4444 zfs_ioc_error_log(zfs_cmd_t *zc)
4445 {
4446         spa_t *spa;
4447         int error;
4448         size_t count = (size_t)zc->zc_nvlist_dst_size;
4449 
4450         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4451                 return (error);
4452 
4453         error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4454             &count);
4455         if (error == 0)
4456                 zc->zc_nvlist_dst_size = count;
4457         else
4458                 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4459 
4460         spa_close(spa, FTAG);
4461 
4462         return (error);
4463 }
4464 
4465 static int
4466 zfs_ioc_clear(zfs_cmd_t *zc)
4467 {
4468         spa_t *spa;
4469         vdev_t *vd;
4470         int error;
4471 
4472         /*
4473          * On zpool clear we also fix up missing slogs
4474          */
4475         mutex_enter(&spa_namespace_lock);
4476         spa = spa_lookup(zc->zc_name);
4477         if (spa == NULL) {
4478                 mutex_exit(&spa_namespace_lock);
4479                 return (SET_ERROR(EIO));
4480         }
4481         if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4482                 /* we need to let spa_open/spa_load clear the chains */
4483                 spa_set_log_state(spa, SPA_LOG_CLEAR);
4484         }
4485         spa->spa_last_open_failed = 0;
4486         mutex_exit(&spa_namespace_lock);
4487 
4488         if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4489                 error = spa_open(zc->zc_name, &spa, FTAG);
4490         } else {
4491                 nvlist_t *policy;
4492                 nvlist_t *config = NULL;
4493 
4494                 if (zc->zc_nvlist_src == NULL)
4495                         return (SET_ERROR(EINVAL));
4496 
4497                 if ((error = get_nvlist(zc->zc_nvlist_src,
4498                     zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4499                         error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4500                             policy, &config);
4501                         if (config != NULL) {
4502                                 int err;
4503 
4504                                 if ((err = put_nvlist(zc, config)) != 0)
4505                                         error = err;
4506                                 nvlist_free(config);
4507                         }
4508                         nvlist_free(policy);
4509                 }
4510         }
4511 
4512         if (error != 0)
4513                 return (error);
4514 
4515         spa_vdev_state_enter(spa, SCL_NONE);
4516 
4517         if (zc->zc_guid == 0) {
4518                 vd = NULL;
4519         } else {
4520                 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4521                 if (vd == NULL) {
4522                         (void) spa_vdev_state_exit(spa, NULL, ENODEV);
4523                         spa_close(spa, FTAG);
4524                         return (SET_ERROR(ENODEV));
4525                 }
4526         }
4527 
4528         vdev_clear(spa, vd);
4529 
4530         (void) spa_vdev_state_exit(spa, NULL, 0);
4531 
4532         /*
4533          * Resume any suspended I/Os.
4534          */
4535         if (zio_resume(spa) != 0)
4536                 error = SET_ERROR(EIO);
4537 
4538         spa_close(spa, FTAG);
4539 
4540         return (error);
4541 }
4542 
4543 static int
4544 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4545 {
4546         spa_t *spa;
4547         int error;
4548 
4549         error = spa_open(zc->zc_name, &spa, FTAG);
4550         if (error != 0)
4551                 return (error);
4552 
4553         spa_vdev_state_enter(spa, SCL_NONE);
4554 
4555         /*
4556          * If a resilver is already in progress then set the
4557          * spa_scrub_reopen flag to B_TRUE so that we don't restart
4558          * the scan as a side effect of the reopen. Otherwise, let
4559          * vdev_open() decided if a resilver is required.
4560          */
4561         spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4562         vdev_reopen(spa->spa_root_vdev);
4563         spa->spa_scrub_reopen = B_FALSE;
4564 
4565         (void) spa_vdev_state_exit(spa, NULL, 0);
4566         spa_close(spa, FTAG);
4567         return (0);
4568 }
4569 /*
4570  * inputs:
4571  * zc_name      name of filesystem
4572  * zc_value     name of origin snapshot
4573  *
4574  * outputs:
4575  * zc_string    name of conflicting snapshot, if there is one
4576  */
4577 static int
4578 zfs_ioc_promote(zfs_cmd_t *zc)
4579 {
4580         char *cp;
4581 
4582         /*
4583          * We don't need to unmount *all* the origin fs's snapshots, but
4584          * it's easier.
4585          */
4586         cp = strchr(zc->zc_value, '@');
4587         if (cp)
4588                 *cp = '\0';
4589         (void) dmu_objset_find(zc->zc_value,
4590             zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
4591         return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4592 }
4593 
4594 /*
4595  * Retrieve a single {user|group}{used|quota}@... property.
4596  *
4597  * inputs:
4598  * zc_name      name of filesystem
4599  * zc_objset_type zfs_userquota_prop_t
4600  * zc_value     domain name (eg. "S-1-234-567-89")
4601  * zc_guid      RID/UID/GID
4602  *
4603  * outputs:
4604  * zc_cookie    property value
4605  */
4606 static int
4607 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4608 {
4609         zfsvfs_t *zfsvfs;
4610         int error;
4611 
4612         if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4613                 return (SET_ERROR(EINVAL));
4614 
4615         error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4616         if (error != 0)
4617                 return (error);
4618 
4619         error = zfs_userspace_one(zfsvfs,
4620             zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4621         zfsvfs_rele(zfsvfs, FTAG);
4622 
4623         return (error);
4624 }
4625 
4626 /*
4627  * inputs:
4628  * zc_name              name of filesystem
4629  * zc_cookie            zap cursor
4630  * zc_objset_type       zfs_userquota_prop_t
4631  * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4632  *
4633  * outputs:
4634  * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
4635  * zc_cookie    zap cursor
4636  */
4637 static int
4638 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4639 {
4640         zfsvfs_t *zfsvfs;
4641         int bufsize = zc->zc_nvlist_dst_size;
4642 
4643         if (bufsize <= 0)
4644                 return (SET_ERROR(ENOMEM));
4645 
4646         int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4647         if (error != 0)
4648                 return (error);
4649 
4650         void *buf = kmem_alloc(bufsize, KM_SLEEP);
4651 
4652         error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
4653             buf, &zc->zc_nvlist_dst_size);
4654 
4655         if (error == 0) {
4656                 error = xcopyout(buf,
4657                     (void *)(uintptr_t)zc->zc_nvlist_dst,
4658                     zc->zc_nvlist_dst_size);
4659         }
4660         kmem_free(buf, bufsize);
4661         zfsvfs_rele(zfsvfs, FTAG);
4662 
4663         return (error);
4664 }
4665 
4666 /*
4667  * inputs:
4668  * zc_name              name of filesystem
4669  *
4670  * outputs:
4671  * none
4672  */
4673 static int
4674 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4675 {
4676         objset_t *os;
4677         int error = 0;
4678         zfsvfs_t *zfsvfs;
4679 
4680         if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
4681                 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
4682                         /*
4683                          * If userused is not enabled, it may be because the
4684                          * objset needs to be closed & reopened (to grow the
4685                          * objset_phys_t).  Suspend/resume the fs will do that.
4686                          */
4687                         error = zfs_suspend_fs(zfsvfs);
4688                         if (error == 0) {
4689                                 dmu_objset_refresh_ownership(zfsvfs->z_os,
4690                                     zfsvfs);
4691                                 error = zfs_resume_fs(zfsvfs, zc->zc_name);
4692                         }
4693                 }
4694                 if (error == 0)
4695                         error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
4696                 VFS_RELE(zfsvfs->z_vfs);
4697         } else {
4698                 /* XXX kind of reading contents without owning */
4699                 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4700                 if (error != 0)
4701                         return (error);
4702 
4703                 error = dmu_objset_userspace_upgrade(os);
4704                 dmu_objset_rele(os, FTAG);
4705         }
4706 
4707         return (error);
4708 }
4709 
4710 /*
4711  * We don't want to have a hard dependency
4712  * against some special symbols in sharefs
4713  * nfs, and smbsrv.  Determine them if needed when
4714  * the first file system is shared.
4715  * Neither sharefs, nfs or smbsrv are unloadable modules.
4716  */
4717 int (*znfsexport_fs)(void *arg);
4718 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
4719 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
4720 
4721 int zfs_nfsshare_inited;
4722 int zfs_smbshare_inited;
4723 
4724 ddi_modhandle_t nfs_mod;
4725 ddi_modhandle_t sharefs_mod;
4726 ddi_modhandle_t smbsrv_mod;
4727 kmutex_t zfs_share_lock;
4728 
4729 static int
4730 zfs_init_sharefs()
4731 {
4732         int error;
4733 
4734         ASSERT(MUTEX_HELD(&zfs_share_lock));
4735         /* Both NFS and SMB shares also require sharetab support. */
4736         if (sharefs_mod == NULL && ((sharefs_mod =
4737             ddi_modopen("fs/sharefs",
4738             KRTLD_MODE_FIRST, &error)) == NULL)) {
4739                 return (SET_ERROR(ENOSYS));
4740         }
4741         if (zshare_fs == NULL && ((zshare_fs =
4742             (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
4743             ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
4744                 return (SET_ERROR(ENOSYS));
4745         }
4746         return (0);
4747 }
4748 
4749 static int
4750 zfs_ioc_share(zfs_cmd_t *zc)
4751 {
4752         int error;
4753         int opcode;
4754 
4755         switch (zc->zc_share.z_sharetype) {
4756         case ZFS_SHARE_NFS:
4757         case ZFS_UNSHARE_NFS:
4758                 if (zfs_nfsshare_inited == 0) {
4759                         mutex_enter(&zfs_share_lock);
4760                         if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
4761                             KRTLD_MODE_FIRST, &error)) == NULL)) {
4762                                 mutex_exit(&zfs_share_lock);
4763                                 return (SET_ERROR(ENOSYS));
4764                         }
4765                         if (znfsexport_fs == NULL &&
4766                             ((znfsexport_fs = (int (*)(void *))
4767                             ddi_modsym(nfs_mod,
4768                             "nfs_export", &error)) == NULL)) {
4769                                 mutex_exit(&zfs_share_lock);
4770                                 return (SET_ERROR(ENOSYS));
4771                         }
4772                         error = zfs_init_sharefs();
4773                         if (error != 0) {
4774                                 mutex_exit(&zfs_share_lock);
4775                                 return (SET_ERROR(ENOSYS));
4776                         }
4777                         zfs_nfsshare_inited = 1;
4778                         mutex_exit(&zfs_share_lock);
4779                 }
4780                 break;
4781         case ZFS_SHARE_SMB:
4782         case ZFS_UNSHARE_SMB:
4783                 if (zfs_smbshare_inited == 0) {
4784                         mutex_enter(&zfs_share_lock);
4785                         if (smbsrv_mod == NULL && ((smbsrv_mod =
4786                             ddi_modopen("drv/smbsrv",
4787                             KRTLD_MODE_FIRST, &error)) == NULL)) {
4788                                 mutex_exit(&zfs_share_lock);
4789                                 return (SET_ERROR(ENOSYS));
4790                         }
4791                         if (zsmbexport_fs == NULL && ((zsmbexport_fs =
4792                             (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
4793                             "smb_server_share", &error)) == NULL)) {
4794                                 mutex_exit(&zfs_share_lock);
4795                                 return (SET_ERROR(ENOSYS));
4796                         }
4797                         error = zfs_init_sharefs();
4798                         if (error != 0) {
4799                                 mutex_exit(&zfs_share_lock);
4800                                 return (SET_ERROR(ENOSYS));
4801                         }
4802                         zfs_smbshare_inited = 1;
4803                         mutex_exit(&zfs_share_lock);
4804                 }
4805                 break;
4806         default:
4807                 return (SET_ERROR(EINVAL));
4808         }
4809 
4810         switch (zc->zc_share.z_sharetype) {
4811         case ZFS_SHARE_NFS:
4812         case ZFS_UNSHARE_NFS:
4813                 if (error =
4814                     znfsexport_fs((void *)
4815                     (uintptr_t)zc->zc_share.z_exportdata))
4816                         return (error);
4817                 break;
4818         case ZFS_SHARE_SMB:
4819         case ZFS_UNSHARE_SMB:
4820                 if (error = zsmbexport_fs((void *)
4821                     (uintptr_t)zc->zc_share.z_exportdata,
4822                     zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
4823                     B_TRUE: B_FALSE)) {
4824                         return (error);
4825                 }
4826                 break;
4827         }
4828 
4829         opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
4830             zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
4831             SHAREFS_ADD : SHAREFS_REMOVE;
4832 
4833         /*
4834          * Add or remove share from sharetab
4835          */
4836         error = zshare_fs(opcode,
4837             (void *)(uintptr_t)zc->zc_share.z_sharedata,
4838             zc->zc_share.z_sharemax);
4839 
4840         return (error);
4841 
4842 }
4843 
4844 ace_t full_access[] = {
4845         {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4846 };
4847 
4848 /*
4849  * inputs:
4850  * zc_name              name of containing filesystem
4851  * zc_obj               object # beyond which we want next in-use object #
4852  *
4853  * outputs:
4854  * zc_obj               next in-use object #
4855  */
4856 static int
4857 zfs_ioc_next_obj(zfs_cmd_t *zc)
4858 {
4859         objset_t *os = NULL;
4860         int error;
4861 
4862         error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4863         if (error != 0)
4864                 return (error);
4865 
4866         error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
4867             os->os_dsl_dataset->ds_phys->ds_prev_snap_txg);
4868 
4869         dmu_objset_rele(os, FTAG);
4870         return (error);
4871 }
4872 
4873 /*
4874  * inputs:
4875  * zc_name              name of filesystem
4876  * zc_value             prefix name for snapshot
4877  * zc_cleanup_fd        cleanup-on-exit file descriptor for calling process
4878  *
4879  * outputs:
4880  * zc_value             short name of new snapshot
4881  */
4882 static int
4883 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
4884 {
4885         char *snap_name;
4886         char *hold_name;
4887         int error;
4888         minor_t minor;
4889 
4890         error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
4891         if (error != 0)
4892                 return (error);
4893 
4894         snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
4895             (u_longlong_t)ddi_get_lbolt64());
4896         hold_name = kmem_asprintf("%%%s", zc->zc_value);
4897 
4898         error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
4899             hold_name);
4900         if (error == 0)
4901                 (void) strcpy(zc->zc_value, snap_name);
4902         strfree(snap_name);
4903         strfree(hold_name);
4904         zfs_onexit_fd_rele(zc->zc_cleanup_fd);
4905         return (error);
4906 }
4907 
4908 /*
4909  * inputs:
4910  * zc_name              name of "to" snapshot
4911  * zc_value             name of "from" snapshot
4912  * zc_cookie            file descriptor to write diff data on
4913  *
4914  * outputs:
4915  * dmu_diff_record_t's to the file descriptor
4916  */
4917 static int
4918 zfs_ioc_diff(zfs_cmd_t *zc)
4919 {
4920         file_t *fp;
4921         offset_t off;
4922         int error;
4923 
4924         fp = getf(zc->zc_cookie);
4925         if (fp == NULL)
4926                 return (SET_ERROR(EBADF));
4927 
4928         off = fp->f_offset;
4929 
4930         error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off);
4931 
4932         if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4933                 fp->f_offset = off;
4934         releasef(zc->zc_cookie);
4935 
4936         return (error);
4937 }
4938 
4939 /*
4940  * Remove all ACL files in shares dir
4941  */
4942 static int
4943 zfs_smb_acl_purge(znode_t *dzp)
4944 {
4945         zap_cursor_t    zc;
4946         zap_attribute_t zap;
4947         zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
4948         int error;
4949 
4950         for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
4951             (error = zap_cursor_retrieve(&zc, &zap)) == 0;
4952             zap_cursor_advance(&zc)) {
4953                 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
4954                     NULL, 0)) != 0)
4955                         break;
4956         }
4957         zap_cursor_fini(&zc);
4958         return (error);
4959 }
4960 
4961 static int
4962 zfs_ioc_smb_acl(zfs_cmd_t *zc)
4963 {
4964         vnode_t *vp;
4965         znode_t *dzp;
4966         vnode_t *resourcevp = NULL;
4967         znode_t *sharedir;
4968         zfsvfs_t *zfsvfs;
4969         nvlist_t *nvlist;
4970         char *src, *target;
4971         vattr_t vattr;
4972         vsecattr_t vsec;
4973         int error = 0;
4974 
4975         if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
4976             NO_FOLLOW, NULL, &vp)) != 0)
4977                 return (error);
4978 
4979         /* Now make sure mntpnt and dataset are ZFS */
4980 
4981         if (vp->v_vfsp->vfs_fstype != zfsfstype ||
4982             (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
4983             zc->zc_name) != 0)) {
4984                 VN_RELE(vp);
4985                 return (SET_ERROR(EINVAL));
4986         }
4987 
4988         dzp = VTOZ(vp);
4989         zfsvfs = dzp->z_zfsvfs;
4990         ZFS_ENTER(zfsvfs);
4991 
4992         /*
4993          * Create share dir if its missing.
4994          */
4995         mutex_enter(&zfsvfs->z_lock);
4996         if (zfsvfs->z_shares_dir == 0) {
4997                 dmu_tx_t *tx;
4998 
4999                 tx = dmu_tx_create(zfsvfs->z_os);
5000                 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
5001                     ZFS_SHARES_DIR);
5002                 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
5003                 error = dmu_tx_assign(tx, TXG_WAIT);
5004                 if (error != 0) {
5005                         dmu_tx_abort(tx);
5006                 } else {
5007                         error = zfs_create_share_dir(zfsvfs, tx);
5008                         dmu_tx_commit(tx);
5009                 }
5010                 if (error != 0) {
5011                         mutex_exit(&zfsvfs->z_lock);
5012                         VN_RELE(vp);
5013                         ZFS_EXIT(zfsvfs);
5014                         return (error);
5015                 }
5016         }
5017         mutex_exit(&zfsvfs->z_lock);
5018 
5019         ASSERT(zfsvfs->z_shares_dir);
5020         if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
5021                 VN_RELE(vp);
5022                 ZFS_EXIT(zfsvfs);
5023                 return (error);
5024         }
5025 
5026         switch (zc->zc_cookie) {
5027         case ZFS_SMB_ACL_ADD:
5028                 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
5029                 vattr.va_type = VREG;
5030                 vattr.va_mode = S_IFREG|0777;
5031                 vattr.va_uid = 0;
5032                 vattr.va_gid = 0;
5033 
5034                 vsec.vsa_mask = VSA_ACE;
5035                 vsec.vsa_aclentp = &full_access;
5036                 vsec.vsa_aclentsz = sizeof (full_access);
5037                 vsec.vsa_aclcnt = 1;
5038 
5039                 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
5040                     &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
5041                 if (resourcevp)
5042                         VN_RELE(resourcevp);
5043                 break;
5044 
5045         case ZFS_SMB_ACL_REMOVE:
5046                 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
5047                     NULL, 0);
5048                 break;
5049 
5050         case ZFS_SMB_ACL_RENAME:
5051                 if ((error = get_nvlist(zc->zc_nvlist_src,
5052                     zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
5053                         VN_RELE(vp);
5054                         ZFS_EXIT(zfsvfs);
5055                         return (error);
5056                 }
5057                 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
5058                     nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
5059                     &target)) {
5060                         VN_RELE(vp);
5061                         VN_RELE(ZTOV(sharedir));
5062                         ZFS_EXIT(zfsvfs);
5063                         nvlist_free(nvlist);
5064                         return (error);
5065                 }
5066                 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
5067                     kcred, NULL, 0);
5068                 nvlist_free(nvlist);
5069                 break;
5070 
5071         case ZFS_SMB_ACL_PURGE:
5072                 error = zfs_smb_acl_purge(sharedir);
5073                 break;
5074 
5075         default:
5076                 error = SET_ERROR(EINVAL);
5077                 break;
5078         }
5079 
5080         VN_RELE(vp);
5081         VN_RELE(ZTOV(sharedir));
5082 
5083         ZFS_EXIT(zfsvfs);
5084 
5085         return (error);
5086 }
5087 
5088 /*
5089  * innvl: {
5090  *     "holds" -> { snapname -> holdname (string), ... }
5091  *     (optional) "cleanup_fd" -> fd (int32)
5092  * }
5093  *
5094  * outnvl: {
5095  *     snapname -> error value (int32)
5096  *     ...
5097  * }
5098  */
5099 /* ARGSUSED */
5100 static int
5101 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
5102 {
5103         nvlist_t *holds;
5104         int cleanup_fd = -1;
5105         int error;
5106         minor_t minor = 0;
5107 
5108         error = nvlist_lookup_nvlist(args, "holds", &holds);
5109         if (error != 0)
5110                 return (SET_ERROR(EINVAL));
5111 
5112         if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
5113                 error = zfs_onexit_fd_hold(cleanup_fd, &minor);
5114                 if (error != 0)
5115                         return (error);
5116         }
5117 
5118         error = dsl_dataset_user_hold(holds, minor, errlist);
5119         if (minor != 0)
5120                 zfs_onexit_fd_rele(cleanup_fd);
5121         return (error);
5122 }
5123 
5124 /*
5125  * innvl is not used.
5126  *
5127  * outnvl: {
5128  *    holdname -> time added (uint64 seconds since epoch)
5129  *    ...
5130  * }
5131  */
5132 /* ARGSUSED */
5133 static int
5134 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
5135 {
5136         return (dsl_dataset_get_holds(snapname, outnvl));
5137 }
5138 
5139 /*
5140  * innvl: {
5141  *     snapname -> { holdname, ... }
5142  *     ...
5143  * }
5144  *
5145  * outnvl: {
5146  *     snapname -> error value (int32)
5147  *     ...
5148  * }
5149  */
5150 /* ARGSUSED */
5151 static int
5152 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
5153 {
5154         return (dsl_dataset_user_release(holds, errlist));
5155 }
5156 
5157 /*
5158  * inputs:
5159  * zc_name              name of new filesystem or snapshot
5160  * zc_value             full name of old snapshot
5161  *
5162  * outputs:
5163  * zc_cookie            space in bytes
5164  * zc_objset_type       compressed space in bytes
5165  * zc_perm_action       uncompressed space in bytes
5166  */
5167 static int
5168 zfs_ioc_space_written(zfs_cmd_t *zc)
5169 {
5170         int error;
5171         dsl_pool_t *dp;
5172         dsl_dataset_t *new, *old;
5173 
5174         error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5175         if (error != 0)
5176                 return (error);
5177         error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
5178         if (error != 0) {
5179                 dsl_pool_rele(dp, FTAG);
5180                 return (error);
5181         }
5182         error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
5183         if (error != 0) {
5184                 dsl_dataset_rele(new, FTAG);
5185                 dsl_pool_rele(dp, FTAG);
5186                 return (error);
5187         }
5188 
5189         error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
5190             &zc->zc_objset_type, &zc->zc_perm_action);
5191         dsl_dataset_rele(old, FTAG);
5192         dsl_dataset_rele(new, FTAG);
5193         dsl_pool_rele(dp, FTAG);
5194         return (error);
5195 }
5196 
5197 /*
5198  * innvl: {
5199  *     "firstsnap" -> snapshot name
5200  * }
5201  *
5202  * outnvl: {
5203  *     "used" -> space in bytes
5204  *     "compressed" -> compressed space in bytes
5205  *     "uncompressed" -> uncompressed space in bytes
5206  * }
5207  */
5208 static int
5209 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
5210 {
5211         int error;
5212         dsl_pool_t *dp;
5213         dsl_dataset_t *new, *old;
5214         char *firstsnap;
5215         uint64_t used, comp, uncomp;
5216 
5217         if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5218                 return (SET_ERROR(EINVAL));
5219 
5220         error = dsl_pool_hold(lastsnap, FTAG, &dp);
5221         if (error != 0)
5222                 return (error);
5223 
5224         error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
5225         if (error != 0) {
5226                 dsl_pool_rele(dp, FTAG);
5227                 return (error);
5228         }
5229         error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
5230         if (error != 0) {
5231                 dsl_dataset_rele(new, FTAG);
5232                 dsl_pool_rele(dp, FTAG);
5233                 return (error);
5234         }
5235 
5236         error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5237         dsl_dataset_rele(old, FTAG);
5238         dsl_dataset_rele(new, FTAG);
5239         dsl_pool_rele(dp, FTAG);
5240         fnvlist_add_uint64(outnvl, "used", used);
5241         fnvlist_add_uint64(outnvl, "compressed", comp);
5242         fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5243         return (error);
5244 }
5245 
5246 /*
5247  * innvl: {
5248  *     "fd" -> file descriptor to write stream to (int32)
5249  *     (optional) "fromsnap" -> full snap name to send an incremental from
5250  * }
5251  *
5252  * outnvl is unused
5253  */
5254 /* ARGSUSED */
5255 static int
5256 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5257 {
5258         int error;
5259         offset_t off;
5260         char *fromname = NULL;
5261         int fd;
5262 
5263         error = nvlist_lookup_int32(innvl, "fd", &fd);
5264         if (error != 0)
5265                 return (SET_ERROR(EINVAL));
5266 
5267         (void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
5268 
5269         file_t *fp = getf(fd);
5270         if (fp == NULL)
5271                 return (SET_ERROR(EBADF));
5272 
5273         off = fp->f_offset;
5274         error = dmu_send(snapname, fromname, fd, fp->f_vnode, &off);
5275 
5276         if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5277                 fp->f_offset = off;
5278         releasef(fd);
5279         return (error);
5280 }
5281 
5282 /*
5283  * Determine approximately how large a zfs send stream will be -- the number
5284  * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5285  *
5286  * innvl: {
5287  *     (optional) "fromsnap" -> full snap name to send an incremental from
5288  * }
5289  *
5290  * outnvl: {
5291  *     "space" -> bytes of space (uint64)
5292  * }
5293  */
5294 static int
5295 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5296 {
5297         dsl_pool_t *dp;
5298         dsl_dataset_t *fromsnap = NULL;
5299         dsl_dataset_t *tosnap;
5300         int error;
5301         char *fromname;
5302         uint64_t space;
5303 
5304         error = dsl_pool_hold(snapname, FTAG, &dp);
5305         if (error != 0)
5306                 return (error);
5307 
5308         error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
5309         if (error != 0) {
5310                 dsl_pool_rele(dp, FTAG);
5311                 return (error);
5312         }
5313 
5314         error = nvlist_lookup_string(innvl, "fromsnap", &fromname);
5315         if (error == 0) {
5316                 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
5317                 if (error != 0) {
5318                         dsl_dataset_rele(tosnap, FTAG);
5319                         dsl_pool_rele(dp, FTAG);
5320                         return (error);
5321                 }
5322         }
5323 
5324         error = dmu_send_estimate(tosnap, fromsnap, &space);
5325         fnvlist_add_uint64(outnvl, "space", space);
5326 
5327         if (fromsnap != NULL)
5328                 dsl_dataset_rele(fromsnap, FTAG);
5329         dsl_dataset_rele(tosnap, FTAG);
5330         dsl_pool_rele(dp, FTAG);
5331         return (error);
5332 }
5333 
5334 
5335 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
5336 
5337 static void
5338 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5339     zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5340     boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
5341 {
5342         zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5343 
5344         ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5345         ASSERT3U(ioc, <, ZFS_IOC_LAST);
5346         ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5347         ASSERT3P(vec->zvec_func, ==, NULL);
5348 
5349         vec->zvec_legacy_func = func;
5350         vec->zvec_secpolicy = secpolicy;
5351         vec->zvec_namecheck = namecheck;
5352         vec->zvec_allow_log = log_history;
5353         vec->zvec_pool_check = pool_check;
5354 }
5355 
5356 /*
5357  * See the block comment at the beginning of this file for details on
5358  * each argument to this function.
5359  */
5360 static void
5361 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
5362     zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5363     zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
5364     boolean_t allow_log)
5365 {
5366         zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5367 
5368         ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5369         ASSERT3U(ioc, <, ZFS_IOC_LAST);
5370         ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5371         ASSERT3P(vec->zvec_func, ==, NULL);
5372 
5373         /* if we are logging, the name must be valid */
5374         ASSERT(!allow_log || namecheck != NO_NAME);
5375 
5376         vec->zvec_name = name;
5377         vec->zvec_func = func;
5378         vec->zvec_secpolicy = secpolicy;
5379         vec->zvec_namecheck = namecheck;
5380         vec->zvec_pool_check = pool_check;
5381         vec->zvec_smush_outnvlist = smush_outnvlist;
5382         vec->zvec_allow_log = allow_log;
5383 }
5384 
5385 static void
5386 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5387     zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
5388     zfs_ioc_poolcheck_t pool_check)
5389 {
5390         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5391             POOL_NAME, log_history, pool_check);
5392 }
5393 
5394 static void
5395 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5396     zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
5397 {
5398         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5399             DATASET_NAME, B_FALSE, pool_check);
5400 }
5401 
5402 static void
5403 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5404 {
5405         zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
5406             POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5407 }
5408 
5409 static void
5410 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5411     zfs_secpolicy_func_t *secpolicy)
5412 {
5413         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5414             NO_NAME, B_FALSE, POOL_CHECK_NONE);
5415 }
5416 
5417 static void
5418 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
5419     zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
5420 {
5421         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5422             DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
5423 }
5424 
5425 static void
5426 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5427 {
5428         zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
5429             zfs_secpolicy_read);
5430 }
5431 
5432 static void
5433 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5434         zfs_secpolicy_func_t *secpolicy)
5435 {
5436         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5437             DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5438 }
5439 
5440 static void
5441 zfs_ioctl_init(void)
5442 {
5443         zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
5444             zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
5445             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5446 
5447         zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
5448             zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
5449             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
5450 
5451         zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
5452             zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
5453             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5454 
5455         zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
5456             zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
5457             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5458 
5459         zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
5460             zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
5461             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5462 
5463         zfs_ioctl_register("create", ZFS_IOC_CREATE,
5464             zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
5465             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5466 
5467         zfs_ioctl_register("clone", ZFS_IOC_CLONE,
5468             zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
5469             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5470 
5471         zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
5472             zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
5473             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5474 
5475         zfs_ioctl_register("hold", ZFS_IOC_HOLD,
5476             zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
5477             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5478         zfs_ioctl_register("release", ZFS_IOC_RELEASE,
5479             zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
5480             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5481 
5482         zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
5483             zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
5484             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5485 
5486         zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
5487             zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
5488             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE);
5489 
5490         zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK,
5491             zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME,
5492             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5493 
5494         zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS,
5495             zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME,
5496             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5497 
5498         zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS,
5499             zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks,
5500             POOL_NAME,
5501             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5502 
5503         /* IOCTLS that use the legacy function signature */
5504 
5505         zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
5506             zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
5507 
5508         zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
5509             zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5510         zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
5511             zfs_ioc_pool_scan);
5512         zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
5513             zfs_ioc_pool_upgrade);
5514         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
5515             zfs_ioc_vdev_add);
5516         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
5517             zfs_ioc_vdev_remove);
5518         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
5519             zfs_ioc_vdev_set_state);
5520         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
5521             zfs_ioc_vdev_attach);
5522         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
5523             zfs_ioc_vdev_detach);
5524         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
5525             zfs_ioc_vdev_setpath);
5526         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
5527             zfs_ioc_vdev_setfru);
5528         zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
5529             zfs_ioc_pool_set_props);
5530         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
5531             zfs_ioc_vdev_split);
5532         zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
5533             zfs_ioc_pool_reguid);
5534 
5535         zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
5536             zfs_ioc_pool_configs, zfs_secpolicy_none);
5537         zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
5538             zfs_ioc_pool_tryimport, zfs_secpolicy_config);
5539         zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
5540             zfs_ioc_inject_fault, zfs_secpolicy_inject);
5541         zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
5542             zfs_ioc_clear_fault, zfs_secpolicy_inject);
5543         zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
5544             zfs_ioc_inject_list_next, zfs_secpolicy_inject);
5545 
5546         /*
5547          * pool destroy, and export don't log the history as part of
5548          * zfsdev_ioctl, but rather zfs_ioc_pool_export
5549          * does the logging of those commands.
5550          */
5551         zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
5552             zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5553         zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
5554             zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5555 
5556         zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
5557             zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5558         zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
5559             zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5560 
5561         zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
5562             zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED);
5563         zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
5564             zfs_ioc_dsobj_to_dsname,
5565             zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED);
5566         zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
5567             zfs_ioc_pool_get_history,
5568             zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
5569 
5570         zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
5571             zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5572 
5573         zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
5574             zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5575         zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
5576             zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5577 
5578         zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
5579             zfs_ioc_space_written);
5580         zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
5581             zfs_ioc_objset_recvd_props);
5582         zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
5583             zfs_ioc_next_obj);
5584         zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
5585             zfs_ioc_get_fsacl);
5586         zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
5587             zfs_ioc_objset_stats);
5588         zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
5589             zfs_ioc_objset_zplprops);
5590         zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
5591             zfs_ioc_dataset_list_next);
5592         zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
5593             zfs_ioc_snapshot_list_next);
5594         zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
5595             zfs_ioc_send_progress);
5596 
5597         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
5598             zfs_ioc_diff, zfs_secpolicy_diff);
5599         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
5600             zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
5601         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
5602             zfs_ioc_obj_to_path, zfs_secpolicy_diff);
5603         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
5604             zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
5605         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
5606             zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
5607         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
5608             zfs_ioc_send, zfs_secpolicy_send);
5609 
5610         zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
5611             zfs_secpolicy_none);
5612         zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
5613             zfs_secpolicy_destroy);
5614         zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
5615             zfs_secpolicy_rename);
5616         zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
5617             zfs_secpolicy_recv);
5618         zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
5619             zfs_secpolicy_promote);
5620         zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
5621             zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
5622         zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
5623             zfs_secpolicy_set_fsacl);
5624 
5625         zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
5626             zfs_secpolicy_share, POOL_CHECK_NONE);
5627         zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
5628             zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
5629         zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
5630             zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
5631             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5632         zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
5633             zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
5634             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5635 }
5636 
5637 int
5638 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
5639     zfs_ioc_poolcheck_t check)
5640 {
5641         spa_t *spa;
5642         int error;
5643 
5644         ASSERT(type == POOL_NAME || type == DATASET_NAME);
5645 
5646         if (check & POOL_CHECK_NONE)
5647                 return (0);
5648 
5649         error = spa_open(name, &spa, FTAG);
5650         if (error == 0) {
5651                 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
5652                         error = SET_ERROR(EAGAIN);
5653                 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
5654                         error = SET_ERROR(EROFS);
5655                 spa_close(spa, FTAG);
5656         }
5657         return (error);
5658 }
5659 
5660 /*
5661  * Find a free minor number.
5662  */
5663 minor_t
5664 zfsdev_minor_alloc(void)
5665 {
5666         static minor_t last_minor;
5667         minor_t m;
5668 
5669         ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5670 
5671         for (m = last_minor + 1; m != last_minor; m++) {
5672                 if (m > ZFSDEV_MAX_MINOR)
5673                         m = 1;
5674                 if (ddi_get_soft_state(zfsdev_state, m) == NULL) {
5675                         last_minor = m;
5676                         return (m);
5677                 }
5678         }
5679 
5680         return (0);
5681 }
5682 
5683 static int
5684 zfs_ctldev_init(dev_t *devp)
5685 {
5686         minor_t minor;
5687         zfs_soft_state_t *zs;
5688 
5689         ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5690         ASSERT(getminor(*devp) == 0);
5691 
5692         minor = zfsdev_minor_alloc();
5693         if (minor == 0)
5694                 return (SET_ERROR(ENXIO));
5695 
5696         if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS)
5697                 return (SET_ERROR(EAGAIN));
5698 
5699         *devp = makedevice(getemajor(*devp), minor);
5700 
5701         zs = ddi_get_soft_state(zfsdev_state, minor);
5702         zs->zss_type = ZSST_CTLDEV;
5703         zfs_onexit_init((zfs_onexit_t **)&zs->zss_data);
5704 
5705         return (0);
5706 }
5707 
5708 static void
5709 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor)
5710 {
5711         ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5712 
5713         zfs_onexit_destroy(zo);
5714         ddi_soft_state_free(zfsdev_state, minor);
5715 }
5716 
5717 void *
5718 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which)
5719 {
5720         zfs_soft_state_t *zp;
5721 
5722         zp = ddi_get_soft_state(zfsdev_state, minor);
5723         if (zp == NULL || zp->zss_type != which)
5724                 return (NULL);
5725 
5726         return (zp->zss_data);
5727 }
5728 
5729 static int
5730 zfsdev_open(dev_t *devp, int flag, int otyp, cred_t *cr)
5731 {
5732         int error = 0;
5733 
5734         if (getminor(*devp) != 0)
5735                 return (zvol_open(devp, flag, otyp, cr));
5736 
5737         /* This is the control device. Allocate a new minor if requested. */
5738         if (flag & FEXCL) {
5739                 mutex_enter(&zfsdev_state_lock);
5740                 error = zfs_ctldev_init(devp);
5741                 mutex_exit(&zfsdev_state_lock);
5742         }
5743 
5744         return (error);
5745 }
5746 
5747 static int
5748 zfsdev_close(dev_t dev, int flag, int otyp, cred_t *cr)
5749 {
5750         zfs_onexit_t *zo;
5751         minor_t minor = getminor(dev);
5752 
5753         if (minor == 0)
5754                 return (0);
5755 
5756         mutex_enter(&zfsdev_state_lock);
5757         zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV);
5758         if (zo == NULL) {
5759                 mutex_exit(&zfsdev_state_lock);
5760                 return (zvol_close(dev, flag, otyp, cr));
5761         }
5762         zfs_ctldev_destroy(zo, minor);
5763         mutex_exit(&zfsdev_state_lock);
5764 
5765         return (0);
5766 }
5767 
5768 static int
5769 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
5770 {
5771         zfs_cmd_t *zc;
5772         uint_t vecnum;
5773         int error, rc, len;
5774         minor_t minor = getminor(dev);
5775         const zfs_ioc_vec_t *vec;
5776         char *saved_poolname = NULL;
5777         nvlist_t *innvl = NULL;
5778 
5779         if (minor != 0 &&
5780             zfsdev_get_soft_state(minor, ZSST_CTLDEV) == NULL)
5781                 return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
5782 
5783         vecnum = cmd - ZFS_IOC_FIRST;
5784         ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
5785 
5786         if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
5787                 return (SET_ERROR(EINVAL));
5788         vec = &zfs_ioc_vec[vecnum];
5789 
5790         zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
5791 
5792         error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
5793         if (error != 0) {
5794                 error = SET_ERROR(EFAULT);
5795                 goto out;
5796         }
5797 
5798         zc->zc_iflags = flag & FKIOCTL;
5799         if (zc->zc_nvlist_src_size != 0) {
5800                 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
5801                     zc->zc_iflags, &innvl);
5802                 if (error != 0)
5803                         goto out;
5804         }
5805 
5806         /*
5807          * Ensure that all pool/dataset names are valid before we pass down to
5808          * the lower layers.
5809          */
5810         zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
5811         switch (vec->zvec_namecheck) {
5812         case POOL_NAME:
5813                 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
5814                         error = SET_ERROR(EINVAL);
5815                 else
5816                         error = pool_status_check(zc->zc_name,
5817                             vec->zvec_namecheck, vec->zvec_pool_check);
5818                 break;
5819 
5820         case DATASET_NAME:
5821                 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
5822                         error = SET_ERROR(EINVAL);
5823                 else
5824                         error = pool_status_check(zc->zc_name,
5825                             vec->zvec_namecheck, vec->zvec_pool_check);
5826                 break;
5827 
5828         case NO_NAME:
5829                 break;
5830         }
5831 
5832 
5833         if (error == 0 && !(flag & FKIOCTL))
5834                 error = vec->zvec_secpolicy(zc, innvl, cr);
5835 
5836         if (error != 0)
5837                 goto out;
5838 
5839         /* legacy ioctls can modify zc_name */
5840         len = strcspn(zc->zc_name, "/@#") + 1;
5841         saved_poolname = kmem_alloc(len, KM_SLEEP);
5842         (void) strlcpy(saved_poolname, zc->zc_name, len);
5843 
5844         if (vec->zvec_func != NULL) {
5845                 nvlist_t *outnvl;
5846                 int puterror = 0;
5847                 spa_t *spa;
5848                 nvlist_t *lognv = NULL;
5849 
5850                 ASSERT(vec->zvec_legacy_func == NULL);
5851 
5852                 /*
5853                  * Add the innvl to the lognv before calling the func,
5854                  * in case the func changes the innvl.
5855                  */
5856                 if (vec->zvec_allow_log) {
5857                         lognv = fnvlist_alloc();
5858                         fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
5859                             vec->zvec_name);
5860                         if (!nvlist_empty(innvl)) {
5861                                 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
5862                                     innvl);
5863                         }
5864                 }
5865 
5866                 outnvl = fnvlist_alloc();
5867                 error = vec->zvec_func(zc->zc_name, innvl, outnvl);
5868 
5869                 if (error == 0 && vec->zvec_allow_log &&
5870                     spa_open(zc->zc_name, &spa, FTAG) == 0) {
5871                         if (!nvlist_empty(outnvl)) {
5872                                 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
5873                                     outnvl);
5874                         }
5875                         (void) spa_history_log_nvl(spa, lognv);
5876                         spa_close(spa, FTAG);
5877                 }
5878                 fnvlist_free(lognv);
5879 
5880                 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
5881                         int smusherror = 0;
5882                         if (vec->zvec_smush_outnvlist) {
5883                                 smusherror = nvlist_smush(outnvl,
5884                                     zc->zc_nvlist_dst_size);
5885                         }
5886                         if (smusherror == 0)
5887                                 puterror = put_nvlist(zc, outnvl);
5888                 }
5889 
5890                 if (puterror != 0)
5891                         error = puterror;
5892 
5893                 nvlist_free(outnvl);
5894         } else {
5895                 error = vec->zvec_legacy_func(zc);
5896         }
5897 
5898 out:
5899         nvlist_free(innvl);
5900         rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
5901         if (error == 0 && rc != 0)
5902                 error = SET_ERROR(EFAULT);
5903         if (error == 0 && vec->zvec_allow_log) {
5904                 char *s = tsd_get(zfs_allow_log_key);
5905                 if (s != NULL)
5906                         strfree(s);
5907                 (void) tsd_set(zfs_allow_log_key, saved_poolname);
5908         } else {
5909                 if (saved_poolname != NULL)
5910                         strfree(saved_poolname);
5911         }
5912 
5913         kmem_free(zc, sizeof (zfs_cmd_t));
5914         return (error);
5915 }
5916 
5917 static int
5918 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
5919 {
5920         if (cmd != DDI_ATTACH)
5921                 return (DDI_FAILURE);
5922 
5923         if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
5924             DDI_PSEUDO, 0) == DDI_FAILURE)
5925                 return (DDI_FAILURE);
5926 
5927         zfs_dip = dip;
5928 
5929         ddi_report_dev(dip);
5930 
5931         return (DDI_SUCCESS);
5932 }
5933 
5934 static int
5935 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
5936 {
5937         if (spa_busy() || zfs_busy() || zvol_busy())
5938                 return (DDI_FAILURE);
5939 
5940         if (cmd != DDI_DETACH)
5941                 return (DDI_FAILURE);
5942 
5943         zfs_dip = NULL;
5944 
5945         ddi_prop_remove_all(dip);
5946         ddi_remove_minor_node(dip, NULL);
5947 
5948         return (DDI_SUCCESS);
5949 }
5950 
5951 /*ARGSUSED*/
5952 static int
5953 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
5954 {
5955         switch (infocmd) {
5956         case DDI_INFO_DEVT2DEVINFO:
5957                 *result = zfs_dip;
5958                 return (DDI_SUCCESS);
5959 
5960         case DDI_INFO_DEVT2INSTANCE:
5961                 *result = (void *)0;
5962                 return (DDI_SUCCESS);
5963         }
5964 
5965         return (DDI_FAILURE);
5966 }
5967 
5968 /*
5969  * OK, so this is a little weird.
5970  *
5971  * /dev/zfs is the control node, i.e. minor 0.
5972  * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
5973  *
5974  * /dev/zfs has basically nothing to do except serve up ioctls,
5975  * so most of the standard driver entry points are in zvol.c.
5976  */
5977 static struct cb_ops zfs_cb_ops = {
5978         zfsdev_open,    /* open */
5979         zfsdev_close,   /* close */
5980         zvol_strategy,  /* strategy */
5981         nodev,          /* print */
5982         zvol_dump,      /* dump */
5983         zvol_read,      /* read */
5984         zvol_write,     /* write */
5985         zfsdev_ioctl,   /* ioctl */
5986         nodev,          /* devmap */
5987         nodev,          /* mmap */
5988         nodev,          /* segmap */
5989         nochpoll,       /* poll */
5990         ddi_prop_op,    /* prop_op */
5991         NULL,           /* streamtab */
5992         D_NEW | D_MP | D_64BIT,         /* Driver compatibility flag */
5993         CB_REV,         /* version */
5994         nodev,          /* async read */
5995         nodev,          /* async write */
5996 };
5997 
5998 static struct dev_ops zfs_dev_ops = {
5999         DEVO_REV,       /* version */
6000         0,              /* refcnt */
6001         zfs_info,       /* info */
6002         nulldev,        /* identify */
6003         nulldev,        /* probe */
6004         zfs_attach,     /* attach */
6005         zfs_detach,     /* detach */
6006         nodev,          /* reset */
6007         &zfs_cb_ops,        /* driver operations */
6008         NULL,           /* no bus operations */
6009         NULL,           /* power */
6010         ddi_quiesce_not_needed, /* quiesce */
6011 };
6012 
6013 static struct modldrv zfs_modldrv = {
6014         &mod_driverops,
6015         "ZFS storage pool",
6016         &zfs_dev_ops
6017 };
6018 
6019 static struct modlinkage modlinkage = {
6020         MODREV_1,
6021         (void *)&zfs_modlfs,
6022         (void *)&zfs_modldrv,
6023         NULL
6024 };
6025 
6026 static void
6027 zfs_allow_log_destroy(void *arg)
6028 {
6029         char *poolname = arg;
6030         strfree(poolname);
6031 }
6032 
6033 int
6034 _init(void)
6035 {
6036         int error;
6037 
6038         spa_init(FREAD | FWRITE);
6039         zfs_init();
6040         zvol_init();
6041         zfs_ioctl_init();
6042 
6043         if ((error = mod_install(&modlinkage)) != 0) {
6044                 zvol_fini();
6045                 zfs_fini();
6046                 spa_fini();
6047                 return (error);
6048         }
6049 
6050         tsd_create(&zfs_fsyncer_key, NULL);
6051         tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
6052         tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
6053 
6054         error = ldi_ident_from_mod(&modlinkage, &zfs_li);
6055         ASSERT(error == 0);
6056         mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
6057 
6058         return (0);
6059 }
6060 
6061 int
6062 _fini(void)
6063 {
6064         int error;
6065 
6066         if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
6067                 return (SET_ERROR(EBUSY));
6068 
6069         if ((error = mod_remove(&modlinkage)) != 0)
6070                 return (error);
6071 
6072         zvol_fini();
6073         zfs_fini();
6074         spa_fini();
6075         if (zfs_nfsshare_inited)
6076                 (void) ddi_modclose(nfs_mod);
6077         if (zfs_smbshare_inited)
6078                 (void) ddi_modclose(smbsrv_mod);
6079         if (zfs_nfsshare_inited || zfs_smbshare_inited)
6080                 (void) ddi_modclose(sharefs_mod);
6081 
6082         tsd_destroy(&zfs_fsyncer_key);
6083         ldi_ident_release(zfs_li);
6084         zfs_li = NULL;
6085         mutex_destroy(&zfs_share_lock);
6086 
6087         return (error);
6088 }
6089 
6090 int
6091 _info(struct modinfo *modinfop)
6092 {
6093         return (mod_info(&modlinkage, modinfop));
6094 }
--- EOF ---