1 /*
   2  * CDDL HEADER START
   3  *
   4  * The contents of this file are subject to the terms of the
   5  * Common Development and Distribution License (the "License").
   6  * You may not use this file except in compliance with the License.
   7  *
   8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
   9  * or http://www.opensolaris.org/os/licensing.
  10  * See the License for the specific language governing permissions
  11  * and limitations under the License.
  12  *
  13  * When distributing Covered Code, include this CDDL HEADER in each
  14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
  15  * If applicable, add the following below this CDDL HEADER, with the
  16  * fields enclosed by brackets "[]" replaced with your own identifying
  17  * information: Portions Copyright [yyyy] [name of copyright owner]
  18  *
  19  * CDDL HEADER END
  20  */
  21 
  22 /*
  23  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
  24  * Portions Copyright 2011 Martin Matuska
  25  * Copyright 2011 Nexenta Systems, Inc.  All rights reserved.
  26  * Copyright (c) 2012, Joyent, Inc. All rights reserved.
  27  * Copyright (c) 2013 by Delphix. All rights reserved.
  28  * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
  29  * Copyright (c) 2013 Steven Hartland. All rights reserved.
  30  */
  31 
  32 /*
  33  * ZFS ioctls.
  34  *
  35  * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
  36  * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
  37  *
  38  * There are two ways that we handle ioctls: the legacy way where almost
  39  * all of the logic is in the ioctl callback, and the new way where most
  40  * of the marshalling is handled in the common entry point, zfsdev_ioctl().
  41  *
  42  * Non-legacy ioctls should be registered by calling
  43  * zfs_ioctl_register() from zfs_ioctl_init().  The ioctl is invoked
  44  * from userland by lzc_ioctl().
  45  *
  46  * The registration arguments are as follows:
  47  *
  48  * const char *name
  49  *   The name of the ioctl.  This is used for history logging.  If the
  50  *   ioctl returns successfully (the callback returns 0), and allow_log
  51  *   is true, then a history log entry will be recorded with the input &
  52  *   output nvlists.  The log entry can be printed with "zpool history -i".
  53  *
  54  * zfs_ioc_t ioc
  55  *   The ioctl request number, which userland will pass to ioctl(2).
  56  *   The ioctl numbers can change from release to release, because
  57  *   the caller (libzfs) must be matched to the kernel.
  58  *
  59  * zfs_secpolicy_func_t *secpolicy
  60  *   This function will be called before the zfs_ioc_func_t, to
  61  *   determine if this operation is permitted.  It should return EPERM
  62  *   on failure, and 0 on success.  Checks include determining if the
  63  *   dataset is visible in this zone, and if the user has either all
  64  *   zfs privileges in the zone (SYS_MOUNT), or has been granted permission
  65  *   to do this operation on this dataset with "zfs allow".
  66  *
  67  * zfs_ioc_namecheck_t namecheck
  68  *   This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
  69  *   name, a dataset name, or nothing.  If the name is not well-formed,
  70  *   the ioctl will fail and the callback will not be called.
  71  *   Therefore, the callback can assume that the name is well-formed
  72  *   (e.g. is null-terminated, doesn't have more than one '@' character,
  73  *   doesn't have invalid characters).
  74  *
  75  * zfs_ioc_poolcheck_t pool_check
  76  *   This specifies requirements on the pool state.  If the pool does
  77  *   not meet them (is suspended or is readonly), the ioctl will fail
  78  *   and the callback will not be called.  If any checks are specified
  79  *   (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
  80  *   Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED |
  81  *   POOL_CHECK_READONLY).
  82  *
  83  * boolean_t smush_outnvlist
  84  *   If smush_outnvlist is true, then the output is presumed to be a
  85  *   list of errors, and it will be "smushed" down to fit into the
  86  *   caller's buffer, by removing some entries and replacing them with a
  87  *   single "N_MORE_ERRORS" entry indicating how many were removed.  See
  88  *   nvlist_smush() for details.  If smush_outnvlist is false, and the
  89  *   outnvlist does not fit into the userland-provided buffer, then the
  90  *   ioctl will fail with ENOMEM.
  91  *
  92  * zfs_ioc_func_t *func
  93  *   The callback function that will perform the operation.
  94  *
  95  *   The callback should return 0 on success, or an error number on
  96  *   failure.  If the function fails, the userland ioctl will return -1,
  97  *   and errno will be set to the callback's return value.  The callback
  98  *   will be called with the following arguments:
  99  *
 100  *   const char *name
 101  *     The name of the pool or dataset to operate on, from
 102  *     zfs_cmd_t:zc_name.  The 'namecheck' argument specifies the
 103  *     expected type (pool, dataset, or none).
 104  *
 105  *   nvlist_t *innvl
 106  *     The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src.  Or
 107  *     NULL if no input nvlist was provided.  Changes to this nvlist are
 108  *     ignored.  If the input nvlist could not be deserialized, the
 109  *     ioctl will fail and the callback will not be called.
 110  *
 111  *   nvlist_t *outnvl
 112  *     The output nvlist, initially empty.  The callback can fill it in,
 113  *     and it will be returned to userland by serializing it into
 114  *     zfs_cmd_t:zc_nvlist_dst.  If it is non-empty, and serialization
 115  *     fails (e.g. because the caller didn't supply a large enough
 116  *     buffer), then the overall ioctl will fail.  See the
 117  *     'smush_nvlist' argument above for additional behaviors.
 118  *
 119  *     There are two typical uses of the output nvlist:
 120  *       - To return state, e.g. property values.  In this case,
 121  *         smush_outnvlist should be false.  If the buffer was not large
 122  *         enough, the caller will reallocate a larger buffer and try
 123  *         the ioctl again.
 124  *
 125  *       - To return multiple errors from an ioctl which makes on-disk
 126  *         changes.  In this case, smush_outnvlist should be true.
 127  *         Ioctls which make on-disk modifications should generally not
 128  *         use the outnvl if they succeed, because the caller can not
 129  *         distinguish between the operation failing, and
 130  *         deserialization failing.
 131  */
 132 
 133 #include <sys/types.h>
 134 #include <sys/param.h>
 135 #include <sys/errno.h>
 136 #include <sys/uio.h>
 137 #include <sys/buf.h>
 138 #include <sys/modctl.h>
 139 #include <sys/open.h>
 140 #include <sys/file.h>
 141 #include <sys/kmem.h>
 142 #include <sys/conf.h>
 143 #include <sys/cmn_err.h>
 144 #include <sys/stat.h>
 145 #include <sys/zfs_ioctl.h>
 146 #include <sys/zfs_vfsops.h>
 147 #include <sys/zfs_znode.h>
 148 #include <sys/zap.h>
 149 #include <sys/spa.h>
 150 #include <sys/spa_impl.h>
 151 #include <sys/vdev.h>
 152 #include <sys/priv_impl.h>
 153 #include <sys/dmu.h>
 154 #include <sys/dsl_dir.h>
 155 #include <sys/dsl_dataset.h>
 156 #include <sys/dsl_prop.h>
 157 #include <sys/dsl_deleg.h>
 158 #include <sys/dmu_objset.h>
 159 #include <sys/dmu_impl.h>
 160 #include <sys/dmu_tx.h>
 161 #include <sys/ddi.h>
 162 #include <sys/sunddi.h>
 163 #include <sys/sunldi.h>
 164 #include <sys/policy.h>
 165 #include <sys/zone.h>
 166 #include <sys/nvpair.h>
 167 #include <sys/pathname.h>
 168 #include <sys/mount.h>
 169 #include <sys/sdt.h>
 170 #include <sys/fs/zfs.h>
 171 #include <sys/zfs_ctldir.h>
 172 #include <sys/zfs_dir.h>
 173 #include <sys/zfs_onexit.h>
 174 #include <sys/zvol.h>
 175 #include <sys/dsl_scan.h>
 176 #include <sharefs/share.h>
 177 #include <sys/dmu_objset.h>
 178 #include <sys/dmu_send.h>
 179 #include <sys/dsl_destroy.h>
 180 #include <sys/dsl_userhold.h>
 181 #include <sys/zfeature.h>
 182 
 183 #include "zfs_namecheck.h"
 184 #include "zfs_prop.h"
 185 #include "zfs_deleg.h"
 186 #include "zfs_comutil.h"
 187 
 188 extern struct modlfs zfs_modlfs;
 189 
 190 extern void zfs_init(void);
 191 extern void zfs_fini(void);
 192 
 193 ldi_ident_t zfs_li = NULL;
 194 dev_info_t *zfs_dip;
 195 
 196 uint_t zfs_fsyncer_key;
 197 extern uint_t rrw_tsd_key;
 198 static uint_t zfs_allow_log_key;
 199 
 200 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *);
 201 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *);
 202 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *);
 203 
 204 typedef enum {
 205         NO_NAME,
 206         POOL_NAME,
 207         DATASET_NAME
 208 } zfs_ioc_namecheck_t;
 209 
 210 typedef enum {
 211         POOL_CHECK_NONE         = 1 << 0,
 212         POOL_CHECK_SUSPENDED    = 1 << 1,
 213         POOL_CHECK_READONLY     = 1 << 2,
 214 } zfs_ioc_poolcheck_t;
 215 
 216 typedef struct zfs_ioc_vec {
 217         zfs_ioc_legacy_func_t   *zvec_legacy_func;
 218         zfs_ioc_func_t          *zvec_func;
 219         zfs_secpolicy_func_t    *zvec_secpolicy;
 220         zfs_ioc_namecheck_t     zvec_namecheck;
 221         boolean_t               zvec_allow_log;
 222         zfs_ioc_poolcheck_t     zvec_pool_check;
 223         boolean_t               zvec_smush_outnvlist;
 224         const char              *zvec_name;
 225 } zfs_ioc_vec_t;
 226 
 227 /* This array is indexed by zfs_userquota_prop_t */
 228 static const char *userquota_perms[] = {
 229         ZFS_DELEG_PERM_USERUSED,
 230         ZFS_DELEG_PERM_USERQUOTA,
 231         ZFS_DELEG_PERM_GROUPUSED,
 232         ZFS_DELEG_PERM_GROUPQUOTA,
 233 };
 234 
 235 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
 236 static int zfs_check_settable(const char *name, nvpair_t *property,
 237     cred_t *cr);
 238 static int zfs_check_clearable(char *dataset, nvlist_t *props,
 239     nvlist_t **errors);
 240 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
 241     boolean_t *);
 242 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
 243 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
 244 
 245 static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature);
 246 
 247 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
 248 void
 249 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
 250 {
 251         const char *newfile;
 252         char buf[512];
 253         va_list adx;
 254 
 255         /*
 256          * Get rid of annoying "../common/" prefix to filename.
 257          */
 258         newfile = strrchr(file, '/');
 259         if (newfile != NULL) {
 260                 newfile = newfile + 1; /* Get rid of leading / */
 261         } else {
 262                 newfile = file;
 263         }
 264 
 265         va_start(adx, fmt);
 266         (void) vsnprintf(buf, sizeof (buf), fmt, adx);
 267         va_end(adx);
 268 
 269         /*
 270          * To get this data, use the zfs-dprintf probe as so:
 271          * dtrace -q -n 'zfs-dprintf \
 272          *      /stringof(arg0) == "dbuf.c"/ \
 273          *      {printf("%s: %s", stringof(arg1), stringof(arg3))}'
 274          * arg0 = file name
 275          * arg1 = function name
 276          * arg2 = line number
 277          * arg3 = message
 278          */
 279         DTRACE_PROBE4(zfs__dprintf,
 280             char *, newfile, char *, func, int, line, char *, buf);
 281 }
 282 
 283 static void
 284 history_str_free(char *buf)
 285 {
 286         kmem_free(buf, HIS_MAX_RECORD_LEN);
 287 }
 288 
 289 static char *
 290 history_str_get(zfs_cmd_t *zc)
 291 {
 292         char *buf;
 293 
 294         if (zc->zc_history == NULL)
 295                 return (NULL);
 296 
 297         buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
 298         if (copyinstr((void *)(uintptr_t)zc->zc_history,
 299             buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
 300                 history_str_free(buf);
 301                 return (NULL);
 302         }
 303 
 304         buf[HIS_MAX_RECORD_LEN -1] = '\0';
 305 
 306         return (buf);
 307 }
 308 
 309 /*
 310  * Check to see if the named dataset is currently defined as bootable
 311  */
 312 static boolean_t
 313 zfs_is_bootfs(const char *name)
 314 {
 315         objset_t *os;
 316 
 317         if (dmu_objset_hold(name, FTAG, &os) == 0) {
 318                 boolean_t ret;
 319                 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
 320                 dmu_objset_rele(os, FTAG);
 321                 return (ret);
 322         }
 323         return (B_FALSE);
 324 }
 325 
 326 /*
 327  * Return non-zero if the spa version is less than requested version.
 328  */
 329 static int
 330 zfs_earlier_version(const char *name, int version)
 331 {
 332         spa_t *spa;
 333 
 334         if (spa_open(name, &spa, FTAG) == 0) {
 335                 if (spa_version(spa) < version) {
 336                         spa_close(spa, FTAG);
 337                         return (1);
 338                 }
 339                 spa_close(spa, FTAG);
 340         }
 341         return (0);
 342 }
 343 
 344 /*
 345  * Return TRUE if the ZPL version is less than requested version.
 346  */
 347 static boolean_t
 348 zpl_earlier_version(const char *name, int version)
 349 {
 350         objset_t *os;
 351         boolean_t rc = B_TRUE;
 352 
 353         if (dmu_objset_hold(name, FTAG, &os) == 0) {
 354                 uint64_t zplversion;
 355 
 356                 if (dmu_objset_type(os) != DMU_OST_ZFS) {
 357                         dmu_objset_rele(os, FTAG);
 358                         return (B_TRUE);
 359                 }
 360                 /* XXX reading from non-owned objset */
 361                 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
 362                         rc = zplversion < version;
 363                 dmu_objset_rele(os, FTAG);
 364         }
 365         return (rc);
 366 }
 367 
 368 static void
 369 zfs_log_history(zfs_cmd_t *zc)
 370 {
 371         spa_t *spa;
 372         char *buf;
 373 
 374         if ((buf = history_str_get(zc)) == NULL)
 375                 return;
 376 
 377         if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
 378                 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
 379                         (void) spa_history_log(spa, buf);
 380                 spa_close(spa, FTAG);
 381         }
 382         history_str_free(buf);
 383 }
 384 
 385 /*
 386  * Policy for top-level read operations (list pools).  Requires no privileges,
 387  * and can be used in the local zone, as there is no associated dataset.
 388  */
 389 /* ARGSUSED */
 390 static int
 391 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 392 {
 393         return (0);
 394 }
 395 
 396 /*
 397  * Policy for dataset read operations (list children, get statistics).  Requires
 398  * no privileges, but must be visible in the local zone.
 399  */
 400 /* ARGSUSED */
 401 static int
 402 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 403 {
 404         if (INGLOBALZONE(curproc) ||
 405             zone_dataset_visible(zc->zc_name, NULL))
 406                 return (0);
 407 
 408         return (SET_ERROR(ENOENT));
 409 }
 410 
 411 static int
 412 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
 413 {
 414         int writable = 1;
 415 
 416         /*
 417          * The dataset must be visible by this zone -- check this first
 418          * so they don't see EPERM on something they shouldn't know about.
 419          */
 420         if (!INGLOBALZONE(curproc) &&
 421             !zone_dataset_visible(dataset, &writable))
 422                 return (SET_ERROR(ENOENT));
 423 
 424         if (INGLOBALZONE(curproc)) {
 425                 /*
 426                  * If the fs is zoned, only root can access it from the
 427                  * global zone.
 428                  */
 429                 if (secpolicy_zfs(cr) && zoned)
 430                         return (SET_ERROR(EPERM));
 431         } else {
 432                 /*
 433                  * If we are in a local zone, the 'zoned' property must be set.
 434                  */
 435                 if (!zoned)
 436                         return (SET_ERROR(EPERM));
 437 
 438                 /* must be writable by this zone */
 439                 if (!writable)
 440                         return (SET_ERROR(EPERM));
 441         }
 442         return (0);
 443 }
 444 
 445 static int
 446 zfs_dozonecheck(const char *dataset, cred_t *cr)
 447 {
 448         uint64_t zoned;
 449 
 450         if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
 451                 return (SET_ERROR(ENOENT));
 452 
 453         return (zfs_dozonecheck_impl(dataset, zoned, cr));
 454 }
 455 
 456 static int
 457 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
 458 {
 459         uint64_t zoned;
 460 
 461         if (dsl_prop_get_int_ds(ds, "zoned", &zoned))
 462                 return (SET_ERROR(ENOENT));
 463 
 464         return (zfs_dozonecheck_impl(dataset, zoned, cr));
 465 }
 466 
 467 static int
 468 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
 469     const char *perm, cred_t *cr)
 470 {
 471         int error;
 472 
 473         error = zfs_dozonecheck_ds(name, ds, cr);
 474         if (error == 0) {
 475                 error = secpolicy_zfs(cr);
 476                 if (error != 0)
 477                         error = dsl_deleg_access_impl(ds, perm, cr);
 478         }
 479         return (error);
 480 }
 481 
 482 static int
 483 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
 484 {
 485         int error;
 486         dsl_dataset_t *ds;
 487         dsl_pool_t *dp;
 488 
 489         error = dsl_pool_hold(name, FTAG, &dp);
 490         if (error != 0)
 491                 return (error);
 492 
 493         error = dsl_dataset_hold(dp, name, FTAG, &ds);
 494         if (error != 0) {
 495                 dsl_pool_rele(dp, FTAG);
 496                 return (error);
 497         }
 498 
 499         error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
 500 
 501         dsl_dataset_rele(ds, FTAG);
 502         dsl_pool_rele(dp, FTAG);
 503         return (error);
 504 }
 505 
 506 /*
 507  * Policy for setting the security label property.
 508  *
 509  * Returns 0 for success, non-zero for access and other errors.
 510  */
 511 static int
 512 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
 513 {
 514         char            ds_hexsl[MAXNAMELEN];
 515         bslabel_t       ds_sl, new_sl;
 516         boolean_t       new_default = FALSE;
 517         uint64_t        zoned;
 518         int             needed_priv = -1;
 519         int             error;
 520 
 521         /* First get the existing dataset label. */
 522         error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
 523             1, sizeof (ds_hexsl), &ds_hexsl, NULL);
 524         if (error != 0)
 525                 return (SET_ERROR(EPERM));
 526 
 527         if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
 528                 new_default = TRUE;
 529 
 530         /* The label must be translatable */
 531         if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
 532                 return (SET_ERROR(EINVAL));
 533 
 534         /*
 535          * In a non-global zone, disallow attempts to set a label that
 536          * doesn't match that of the zone; otherwise no other checks
 537          * are needed.
 538          */
 539         if (!INGLOBALZONE(curproc)) {
 540                 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
 541                         return (SET_ERROR(EPERM));
 542                 return (0);
 543         }
 544 
 545         /*
 546          * For global-zone datasets (i.e., those whose zoned property is
 547          * "off", verify that the specified new label is valid for the
 548          * global zone.
 549          */
 550         if (dsl_prop_get_integer(name,
 551             zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
 552                 return (SET_ERROR(EPERM));
 553         if (!zoned) {
 554                 if (zfs_check_global_label(name, strval) != 0)
 555                         return (SET_ERROR(EPERM));
 556         }
 557 
 558         /*
 559          * If the existing dataset label is nondefault, check if the
 560          * dataset is mounted (label cannot be changed while mounted).
 561          * Get the zfsvfs; if there isn't one, then the dataset isn't
 562          * mounted (or isn't a dataset, doesn't exist, ...).
 563          */
 564         if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
 565                 objset_t *os;
 566                 static char *setsl_tag = "setsl_tag";
 567 
 568                 /*
 569                  * Try to own the dataset; abort if there is any error,
 570                  * (e.g., already mounted, in use, or other error).
 571                  */
 572                 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
 573                     setsl_tag, &os);
 574                 if (error != 0)
 575                         return (SET_ERROR(EPERM));
 576 
 577                 dmu_objset_disown(os, setsl_tag);
 578 
 579                 if (new_default) {
 580                         needed_priv = PRIV_FILE_DOWNGRADE_SL;
 581                         goto out_check;
 582                 }
 583 
 584                 if (hexstr_to_label(strval, &new_sl) != 0)
 585                         return (SET_ERROR(EPERM));
 586 
 587                 if (blstrictdom(&ds_sl, &new_sl))
 588                         needed_priv = PRIV_FILE_DOWNGRADE_SL;
 589                 else if (blstrictdom(&new_sl, &ds_sl))
 590                         needed_priv = PRIV_FILE_UPGRADE_SL;
 591         } else {
 592                 /* dataset currently has a default label */
 593                 if (!new_default)
 594                         needed_priv = PRIV_FILE_UPGRADE_SL;
 595         }
 596 
 597 out_check:
 598         if (needed_priv != -1)
 599                 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
 600         return (0);
 601 }
 602 
 603 static int
 604 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
 605     cred_t *cr)
 606 {
 607         char *strval;
 608 
 609         /*
 610          * Check permissions for special properties.
 611          */
 612         switch (prop) {
 613         case ZFS_PROP_ZONED:
 614                 /*
 615                  * Disallow setting of 'zoned' from within a local zone.
 616                  */
 617                 if (!INGLOBALZONE(curproc))
 618                         return (SET_ERROR(EPERM));
 619                 break;
 620 
 621         case ZFS_PROP_QUOTA:
 622                 if (!INGLOBALZONE(curproc)) {
 623                         uint64_t zoned;
 624                         char setpoint[MAXNAMELEN];
 625                         /*
 626                          * Unprivileged users are allowed to modify the
 627                          * quota on things *under* (ie. contained by)
 628                          * the thing they own.
 629                          */
 630                         if (dsl_prop_get_integer(dsname, "zoned", &zoned,
 631                             setpoint))
 632                                 return (SET_ERROR(EPERM));
 633                         if (!zoned || strlen(dsname) <= strlen(setpoint))
 634                                 return (SET_ERROR(EPERM));
 635                 }
 636                 break;
 637 
 638         case ZFS_PROP_MLSLABEL:
 639                 if (!is_system_labeled())
 640                         return (SET_ERROR(EPERM));
 641 
 642                 if (nvpair_value_string(propval, &strval) == 0) {
 643                         int err;
 644 
 645                         err = zfs_set_slabel_policy(dsname, strval, CRED());
 646                         if (err != 0)
 647                                 return (err);
 648                 }
 649                 break;
 650         }
 651 
 652         return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
 653 }
 654 
 655 /* ARGSUSED */
 656 static int
 657 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 658 {
 659         int error;
 660 
 661         error = zfs_dozonecheck(zc->zc_name, cr);
 662         if (error != 0)
 663                 return (error);
 664 
 665         /*
 666          * permission to set permissions will be evaluated later in
 667          * dsl_deleg_can_allow()
 668          */
 669         return (0);
 670 }
 671 
 672 /* ARGSUSED */
 673 static int
 674 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 675 {
 676         return (zfs_secpolicy_write_perms(zc->zc_name,
 677             ZFS_DELEG_PERM_ROLLBACK, cr));
 678 }
 679 
 680 /* ARGSUSED */
 681 static int
 682 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 683 {
 684         dsl_pool_t *dp;
 685         dsl_dataset_t *ds;
 686         char *cp;
 687         int error;
 688 
 689         /*
 690          * Generate the current snapshot name from the given objsetid, then
 691          * use that name for the secpolicy/zone checks.
 692          */
 693         cp = strchr(zc->zc_name, '@');
 694         if (cp == NULL)
 695                 return (SET_ERROR(EINVAL));
 696         error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
 697         if (error != 0)
 698                 return (error);
 699 
 700         error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
 701         if (error != 0) {
 702                 dsl_pool_rele(dp, FTAG);
 703                 return (error);
 704         }
 705 
 706         dsl_dataset_name(ds, zc->zc_name);
 707 
 708         error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
 709             ZFS_DELEG_PERM_SEND, cr);
 710         dsl_dataset_rele(ds, FTAG);
 711         dsl_pool_rele(dp, FTAG);
 712 
 713         return (error);
 714 }
 715 
 716 /* ARGSUSED */
 717 static int
 718 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 719 {
 720         return (zfs_secpolicy_write_perms(zc->zc_name,
 721             ZFS_DELEG_PERM_SEND, cr));
 722 }
 723 
 724 /* ARGSUSED */
 725 static int
 726 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 727 {
 728         vnode_t *vp;
 729         int error;
 730 
 731         if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
 732             NO_FOLLOW, NULL, &vp)) != 0)
 733                 return (error);
 734 
 735         /* Now make sure mntpnt and dataset are ZFS */
 736 
 737         if (vp->v_vfsp->vfs_fstype != zfsfstype ||
 738             (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
 739             zc->zc_name) != 0)) {
 740                 VN_RELE(vp);
 741                 return (SET_ERROR(EPERM));
 742         }
 743 
 744         VN_RELE(vp);
 745         return (dsl_deleg_access(zc->zc_name,
 746             ZFS_DELEG_PERM_SHARE, cr));
 747 }
 748 
 749 int
 750 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 751 {
 752         if (!INGLOBALZONE(curproc))
 753                 return (SET_ERROR(EPERM));
 754 
 755         if (secpolicy_nfs(cr) == 0) {
 756                 return (0);
 757         } else {
 758                 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
 759         }
 760 }
 761 
 762 int
 763 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 764 {
 765         if (!INGLOBALZONE(curproc))
 766                 return (SET_ERROR(EPERM));
 767 
 768         if (secpolicy_smb(cr) == 0) {
 769                 return (0);
 770         } else {
 771                 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
 772         }
 773 }
 774 
 775 static int
 776 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
 777 {
 778         char *cp;
 779 
 780         /*
 781          * Remove the @bla or /bla from the end of the name to get the parent.
 782          */
 783         (void) strncpy(parent, datasetname, parentsize);
 784         cp = strrchr(parent, '@');
 785         if (cp != NULL) {
 786                 cp[0] = '\0';
 787         } else {
 788                 cp = strrchr(parent, '/');
 789                 if (cp == NULL)
 790                         return (SET_ERROR(ENOENT));
 791                 cp[0] = '\0';
 792         }
 793 
 794         return (0);
 795 }
 796 
 797 int
 798 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
 799 {
 800         int error;
 801 
 802         if ((error = zfs_secpolicy_write_perms(name,
 803             ZFS_DELEG_PERM_MOUNT, cr)) != 0)
 804                 return (error);
 805 
 806         return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
 807 }
 808 
 809 /* ARGSUSED */
 810 static int
 811 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 812 {
 813         return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
 814 }
 815 
 816 /*
 817  * Destroying snapshots with delegated permissions requires
 818  * descendant mount and destroy permissions.
 819  */
 820 /* ARGSUSED */
 821 static int
 822 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 823 {
 824         nvlist_t *snaps;
 825         nvpair_t *pair, *nextpair;
 826         int error = 0;
 827 
 828         if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
 829                 return (SET_ERROR(EINVAL));
 830         for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
 831             pair = nextpair) {
 832                 dsl_pool_t *dp;
 833                 dsl_dataset_t *ds;
 834 
 835                 error = dsl_pool_hold(nvpair_name(pair), FTAG, &dp);
 836                 if (error != 0)
 837                         break;
 838                 nextpair = nvlist_next_nvpair(snaps, pair);
 839                 error = dsl_dataset_hold(dp, nvpair_name(pair), FTAG, &ds);
 840                 if (error == 0)
 841                         dsl_dataset_rele(ds, FTAG);
 842                 dsl_pool_rele(dp, FTAG);
 843 
 844                 if (error == 0) {
 845                         error = zfs_secpolicy_destroy_perms(nvpair_name(pair),
 846                             cr);
 847                 } else if (error == ENOENT) {
 848                         /*
 849                          * Ignore any snapshots that don't exist (we consider
 850                          * them "already destroyed").  Remove the name from the
 851                          * nvl here in case the snapshot is created between
 852                          * now and when we try to destroy it (in which case
 853                          * we don't want to destroy it since we haven't
 854                          * checked for permission).
 855                          */
 856                         fnvlist_remove_nvpair(snaps, pair);
 857                         error = 0;
 858                 }
 859                 if (error != 0)
 860                         break;
 861         }
 862 
 863         return (error);
 864 }
 865 
 866 int
 867 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
 868 {
 869         char    parentname[MAXNAMELEN];
 870         int     error;
 871 
 872         if ((error = zfs_secpolicy_write_perms(from,
 873             ZFS_DELEG_PERM_RENAME, cr)) != 0)
 874                 return (error);
 875 
 876         if ((error = zfs_secpolicy_write_perms(from,
 877             ZFS_DELEG_PERM_MOUNT, cr)) != 0)
 878                 return (error);
 879 
 880         if ((error = zfs_get_parent(to, parentname,
 881             sizeof (parentname))) != 0)
 882                 return (error);
 883 
 884         if ((error = zfs_secpolicy_write_perms(parentname,
 885             ZFS_DELEG_PERM_CREATE, cr)) != 0)
 886                 return (error);
 887 
 888         if ((error = zfs_secpolicy_write_perms(parentname,
 889             ZFS_DELEG_PERM_MOUNT, cr)) != 0)
 890                 return (error);
 891 
 892         return (error);
 893 }
 894 
 895 /* ARGSUSED */
 896 static int
 897 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 898 {
 899         return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
 900 }
 901 
 902 /* ARGSUSED */
 903 static int
 904 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 905 {
 906         dsl_pool_t *dp;
 907         dsl_dataset_t *clone;
 908         int error;
 909 
 910         error = zfs_secpolicy_write_perms(zc->zc_name,
 911             ZFS_DELEG_PERM_PROMOTE, cr);
 912         if (error != 0)
 913                 return (error);
 914 
 915         error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
 916         if (error != 0)
 917                 return (error);
 918 
 919         error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
 920 
 921         if (error == 0) {
 922                 char parentname[MAXNAMELEN];
 923                 dsl_dataset_t *origin = NULL;
 924                 dsl_dir_t *dd;
 925                 dd = clone->ds_dir;
 926 
 927                 error = dsl_dataset_hold_obj(dd->dd_pool,
 928                     dd->dd_phys->dd_origin_obj, FTAG, &origin);
 929                 if (error != 0) {
 930                         dsl_dataset_rele(clone, FTAG);
 931                         dsl_pool_rele(dp, FTAG);
 932                         return (error);
 933                 }
 934 
 935                 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
 936                     ZFS_DELEG_PERM_MOUNT, cr);
 937 
 938                 dsl_dataset_name(origin, parentname);
 939                 if (error == 0) {
 940                         error = zfs_secpolicy_write_perms_ds(parentname, origin,
 941                             ZFS_DELEG_PERM_PROMOTE, cr);
 942                 }
 943                 dsl_dataset_rele(clone, FTAG);
 944                 dsl_dataset_rele(origin, FTAG);
 945         }
 946         dsl_pool_rele(dp, FTAG);
 947         return (error);
 948 }
 949 
 950 /* ARGSUSED */
 951 static int
 952 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 953 {
 954         int error;
 955 
 956         if ((error = zfs_secpolicy_write_perms(zc->zc_name,
 957             ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
 958                 return (error);
 959 
 960         if ((error = zfs_secpolicy_write_perms(zc->zc_name,
 961             ZFS_DELEG_PERM_MOUNT, cr)) != 0)
 962                 return (error);
 963 
 964         return (zfs_secpolicy_write_perms(zc->zc_name,
 965             ZFS_DELEG_PERM_CREATE, cr));
 966 }
 967 
 968 int
 969 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
 970 {
 971         return (zfs_secpolicy_write_perms(name,
 972             ZFS_DELEG_PERM_SNAPSHOT, cr));
 973 }
 974 
 975 /*
 976  * Check for permission to create each snapshot in the nvlist.
 977  */
 978 /* ARGSUSED */
 979 static int
 980 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 981 {
 982         nvlist_t *snaps;
 983         int error = 0;
 984         nvpair_t *pair;
 985 
 986         if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
 987                 return (SET_ERROR(EINVAL));
 988         for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
 989             pair = nvlist_next_nvpair(snaps, pair)) {
 990                 char *name = nvpair_name(pair);
 991                 char *atp = strchr(name, '@');
 992 
 993                 if (atp == NULL) {
 994                         error = SET_ERROR(EINVAL);
 995                         break;
 996                 }
 997                 *atp = '\0';
 998                 error = zfs_secpolicy_snapshot_perms(name, cr);
 999                 *atp = '@';
1000                 if (error != 0)
1001                         break;
1002         }
1003         return (error);
1004 }
1005 
1006 /* ARGSUSED */
1007 static int
1008 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1009 {
1010         /*
1011          * Even root must have a proper TSD so that we know what pool
1012          * to log to.
1013          */
1014         if (tsd_get(zfs_allow_log_key) == NULL)
1015                 return (SET_ERROR(EPERM));
1016         return (0);
1017 }
1018 
1019 static int
1020 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1021 {
1022         char    parentname[MAXNAMELEN];
1023         int     error;
1024         char    *origin;
1025 
1026         if ((error = zfs_get_parent(zc->zc_name, parentname,
1027             sizeof (parentname))) != 0)
1028                 return (error);
1029 
1030         if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1031             (error = zfs_secpolicy_write_perms(origin,
1032             ZFS_DELEG_PERM_CLONE, cr)) != 0)
1033                 return (error);
1034 
1035         if ((error = zfs_secpolicy_write_perms(parentname,
1036             ZFS_DELEG_PERM_CREATE, cr)) != 0)
1037                 return (error);
1038 
1039         return (zfs_secpolicy_write_perms(parentname,
1040             ZFS_DELEG_PERM_MOUNT, cr));
1041 }
1042 
1043 /*
1044  * Policy for pool operations - create/destroy pools, add vdevs, etc.  Requires
1045  * SYS_CONFIG privilege, which is not available in a local zone.
1046  */
1047 /* ARGSUSED */
1048 static int
1049 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1050 {
1051         if (secpolicy_sys_config(cr, B_FALSE) != 0)
1052                 return (SET_ERROR(EPERM));
1053 
1054         return (0);
1055 }
1056 
1057 /*
1058  * Policy for object to name lookups.
1059  */
1060 /* ARGSUSED */
1061 static int
1062 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1063 {
1064         int error;
1065 
1066         if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1067                 return (0);
1068 
1069         error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1070         return (error);
1071 }
1072 
1073 /*
1074  * Policy for fault injection.  Requires all privileges.
1075  */
1076 /* ARGSUSED */
1077 static int
1078 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1079 {
1080         return (secpolicy_zinject(cr));
1081 }
1082 
1083 /* ARGSUSED */
1084 static int
1085 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1086 {
1087         zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1088 
1089         if (prop == ZPROP_INVAL) {
1090                 if (!zfs_prop_user(zc->zc_value))
1091                         return (SET_ERROR(EINVAL));
1092                 return (zfs_secpolicy_write_perms(zc->zc_name,
1093                     ZFS_DELEG_PERM_USERPROP, cr));
1094         } else {
1095                 return (zfs_secpolicy_setprop(zc->zc_name, prop,
1096                     NULL, cr));
1097         }
1098 }
1099 
1100 static int
1101 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1102 {
1103         int err = zfs_secpolicy_read(zc, innvl, cr);
1104         if (err)
1105                 return (err);
1106 
1107         if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1108                 return (SET_ERROR(EINVAL));
1109 
1110         if (zc->zc_value[0] == 0) {
1111                 /*
1112                  * They are asking about a posix uid/gid.  If it's
1113                  * themself, allow it.
1114                  */
1115                 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1116                     zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
1117                         if (zc->zc_guid == crgetuid(cr))
1118                                 return (0);
1119                 } else {
1120                         if (groupmember(zc->zc_guid, cr))
1121                                 return (0);
1122                 }
1123         }
1124 
1125         return (zfs_secpolicy_write_perms(zc->zc_name,
1126             userquota_perms[zc->zc_objset_type], cr));
1127 }
1128 
1129 static int
1130 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1131 {
1132         int err = zfs_secpolicy_read(zc, innvl, cr);
1133         if (err)
1134                 return (err);
1135 
1136         if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1137                 return (SET_ERROR(EINVAL));
1138 
1139         return (zfs_secpolicy_write_perms(zc->zc_name,
1140             userquota_perms[zc->zc_objset_type], cr));
1141 }
1142 
1143 /* ARGSUSED */
1144 static int
1145 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1146 {
1147         return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1148             NULL, cr));
1149 }
1150 
1151 /* ARGSUSED */
1152 static int
1153 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1154 {
1155         nvpair_t *pair;
1156         nvlist_t *holds;
1157         int error;
1158 
1159         error = nvlist_lookup_nvlist(innvl, "holds", &holds);
1160         if (error != 0)
1161                 return (SET_ERROR(EINVAL));
1162 
1163         for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
1164             pair = nvlist_next_nvpair(holds, pair)) {
1165                 char fsname[MAXNAMELEN];
1166                 error = dmu_fsname(nvpair_name(pair), fsname);
1167                 if (error != 0)
1168                         return (error);
1169                 error = zfs_secpolicy_write_perms(fsname,
1170                     ZFS_DELEG_PERM_HOLD, cr);
1171                 if (error != 0)
1172                         return (error);
1173         }
1174         return (0);
1175 }
1176 
1177 /* ARGSUSED */
1178 static int
1179 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1180 {
1181         nvpair_t *pair;
1182         int error;
1183 
1184         for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1185             pair = nvlist_next_nvpair(innvl, pair)) {
1186                 char fsname[MAXNAMELEN];
1187                 error = dmu_fsname(nvpair_name(pair), fsname);
1188                 if (error != 0)
1189                         return (error);
1190                 error = zfs_secpolicy_write_perms(fsname,
1191                     ZFS_DELEG_PERM_RELEASE, cr);
1192                 if (error != 0)
1193                         return (error);
1194         }
1195         return (0);
1196 }
1197 
1198 /*
1199  * Policy for allowing temporary snapshots to be taken or released
1200  */
1201 static int
1202 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1203 {
1204         /*
1205          * A temporary snapshot is the same as a snapshot,
1206          * hold, destroy and release all rolled into one.
1207          * Delegated diff alone is sufficient that we allow this.
1208          */
1209         int error;
1210 
1211         if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1212             ZFS_DELEG_PERM_DIFF, cr)) == 0)
1213                 return (0);
1214 
1215         error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1216         if (error == 0)
1217                 error = zfs_secpolicy_hold(zc, innvl, cr);
1218         if (error == 0)
1219                 error = zfs_secpolicy_release(zc, innvl, cr);
1220         if (error == 0)
1221                 error = zfs_secpolicy_destroy(zc, innvl, cr);
1222         return (error);
1223 }
1224 
1225 /*
1226  * Returns the nvlist as specified by the user in the zfs_cmd_t.
1227  */
1228 static int
1229 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1230 {
1231         char *packed;
1232         int error;
1233         nvlist_t *list = NULL;
1234 
1235         /*
1236          * Read in and unpack the user-supplied nvlist.
1237          */
1238         if (size == 0)
1239                 return (SET_ERROR(EINVAL));
1240 
1241         packed = kmem_alloc(size, KM_SLEEP);
1242 
1243         if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1244             iflag)) != 0) {
1245                 kmem_free(packed, size);
1246                 return (error);
1247         }
1248 
1249         if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1250                 kmem_free(packed, size);
1251                 return (error);
1252         }
1253 
1254         kmem_free(packed, size);
1255 
1256         *nvp = list;
1257         return (0);
1258 }
1259 
1260 /*
1261  * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1262  * Entries will be removed from the end of the nvlist, and one int32 entry
1263  * named "N_MORE_ERRORS" will be added indicating how many entries were
1264  * removed.
1265  */
1266 static int
1267 nvlist_smush(nvlist_t *errors, size_t max)
1268 {
1269         size_t size;
1270 
1271         size = fnvlist_size(errors);
1272 
1273         if (size > max) {
1274                 nvpair_t *more_errors;
1275                 int n = 0;
1276 
1277                 if (max < 1024)
1278                         return (SET_ERROR(ENOMEM));
1279 
1280                 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1281                 more_errors = nvlist_prev_nvpair(errors, NULL);
1282 
1283                 do {
1284                         nvpair_t *pair = nvlist_prev_nvpair(errors,
1285                             more_errors);
1286                         fnvlist_remove_nvpair(errors, pair);
1287                         n++;
1288                         size = fnvlist_size(errors);
1289                 } while (size > max);
1290 
1291                 fnvlist_remove_nvpair(errors, more_errors);
1292                 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1293                 ASSERT3U(fnvlist_size(errors), <=, max);
1294         }
1295 
1296         return (0);
1297 }
1298 
1299 static int
1300 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1301 {
1302         char *packed = NULL;
1303         int error = 0;
1304         size_t size;
1305 
1306         size = fnvlist_size(nvl);
1307 
1308         if (size > zc->zc_nvlist_dst_size) {
1309                 error = SET_ERROR(ENOMEM);
1310         } else {
1311                 packed = fnvlist_pack(nvl, &size);
1312                 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1313                     size, zc->zc_iflags) != 0)
1314                         error = SET_ERROR(EFAULT);
1315                 fnvlist_pack_free(packed, size);
1316         }
1317 
1318         zc->zc_nvlist_dst_size = size;
1319         zc->zc_nvlist_dst_filled = B_TRUE;
1320         return (error);
1321 }
1322 
1323 static int
1324 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1325 {
1326         objset_t *os;
1327         int error;
1328 
1329         error = dmu_objset_hold(dsname, FTAG, &os);
1330         if (error != 0)
1331                 return (error);
1332         if (dmu_objset_type(os) != DMU_OST_ZFS) {
1333                 dmu_objset_rele(os, FTAG);
1334                 return (SET_ERROR(EINVAL));
1335         }
1336 
1337         mutex_enter(&os->os_user_ptr_lock);
1338         *zfvp = dmu_objset_get_user(os);
1339         if (*zfvp) {
1340                 VFS_HOLD((*zfvp)->z_vfs);
1341         } else {
1342                 error = SET_ERROR(ESRCH);
1343         }
1344         mutex_exit(&os->os_user_ptr_lock);
1345         dmu_objset_rele(os, FTAG);
1346         return (error);
1347 }
1348 
1349 /*
1350  * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1351  * case its z_vfs will be NULL, and it will be opened as the owner.
1352  * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1353  * which prevents all vnode ops from running.
1354  */
1355 static int
1356 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1357 {
1358         int error = 0;
1359 
1360         if (getzfsvfs(name, zfvp) != 0)
1361                 error = zfsvfs_create(name, zfvp);
1362         if (error == 0) {
1363                 rrw_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER :
1364                     RW_READER, tag);
1365                 if ((*zfvp)->z_unmounted) {
1366                         /*
1367                          * XXX we could probably try again, since the unmounting
1368                          * thread should be just about to disassociate the
1369                          * objset from the zfsvfs.
1370                          */
1371                         rrw_exit(&(*zfvp)->z_teardown_lock, tag);
1372                         return (SET_ERROR(EBUSY));
1373                 }
1374         }
1375         return (error);
1376 }
1377 
1378 static void
1379 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1380 {
1381         rrw_exit(&zfsvfs->z_teardown_lock, tag);
1382 
1383         if (zfsvfs->z_vfs) {
1384                 VFS_RELE(zfsvfs->z_vfs);
1385         } else {
1386                 dmu_objset_disown(zfsvfs->z_os, zfsvfs);
1387                 zfsvfs_free(zfsvfs);
1388         }
1389 }
1390 
1391 static int
1392 zfs_ioc_pool_create(zfs_cmd_t *zc)
1393 {
1394         int error;
1395         nvlist_t *config, *props = NULL;
1396         nvlist_t *rootprops = NULL;
1397         nvlist_t *zplprops = NULL;
1398 
1399         if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1400             zc->zc_iflags, &config))
1401                 return (error);
1402 
1403         if (zc->zc_nvlist_src_size != 0 && (error =
1404             get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1405             zc->zc_iflags, &props))) {
1406                 nvlist_free(config);
1407                 return (error);
1408         }
1409 
1410         if (props) {
1411                 nvlist_t *nvl = NULL;
1412                 uint64_t version = SPA_VERSION;
1413 
1414                 (void) nvlist_lookup_uint64(props,
1415                     zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1416                 if (!SPA_VERSION_IS_SUPPORTED(version)) {
1417                         error = SET_ERROR(EINVAL);
1418                         goto pool_props_bad;
1419                 }
1420                 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1421                 if (nvl) {
1422                         error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1423                         if (error != 0) {
1424                                 nvlist_free(config);
1425                                 nvlist_free(props);
1426                                 return (error);
1427                         }
1428                         (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1429                 }
1430                 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1431                 error = zfs_fill_zplprops_root(version, rootprops,
1432                     zplprops, NULL);
1433                 if (error != 0)
1434                         goto pool_props_bad;
1435         }
1436 
1437         error = spa_create(zc->zc_name, config, props, zplprops);
1438 
1439         /*
1440          * Set the remaining root properties
1441          */
1442         if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1443             ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1444                 (void) spa_destroy(zc->zc_name);
1445 
1446 pool_props_bad:
1447         nvlist_free(rootprops);
1448         nvlist_free(zplprops);
1449         nvlist_free(config);
1450         nvlist_free(props);
1451 
1452         return (error);
1453 }
1454 
1455 static int
1456 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1457 {
1458         int error;
1459         zfs_log_history(zc);
1460         error = spa_destroy(zc->zc_name);
1461         if (error == 0)
1462                 zvol_remove_minors(zc->zc_name);
1463         return (error);
1464 }
1465 
1466 static int
1467 zfs_ioc_pool_import(zfs_cmd_t *zc)
1468 {
1469         nvlist_t *config, *props = NULL;
1470         uint64_t guid;
1471         int error;
1472 
1473         if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1474             zc->zc_iflags, &config)) != 0)
1475                 return (error);
1476 
1477         if (zc->zc_nvlist_src_size != 0 && (error =
1478             get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1479             zc->zc_iflags, &props))) {
1480                 nvlist_free(config);
1481                 return (error);
1482         }
1483 
1484         if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1485             guid != zc->zc_guid)
1486                 error = SET_ERROR(EINVAL);
1487         else
1488                 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1489 
1490         if (zc->zc_nvlist_dst != 0) {
1491                 int err;
1492 
1493                 if ((err = put_nvlist(zc, config)) != 0)
1494                         error = err;
1495         }
1496 
1497         nvlist_free(config);
1498 
1499         if (props)
1500                 nvlist_free(props);
1501 
1502         return (error);
1503 }
1504 
1505 static int
1506 zfs_ioc_pool_export(zfs_cmd_t *zc)
1507 {
1508         int error;
1509         boolean_t force = (boolean_t)zc->zc_cookie;
1510         boolean_t hardforce = (boolean_t)zc->zc_guid;
1511 
1512         zfs_log_history(zc);
1513         error = spa_export(zc->zc_name, NULL, force, hardforce);
1514         if (error == 0)
1515                 zvol_remove_minors(zc->zc_name);
1516         return (error);
1517 }
1518 
1519 static int
1520 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1521 {
1522         nvlist_t *configs;
1523         int error;
1524 
1525         if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1526                 return (SET_ERROR(EEXIST));
1527 
1528         error = put_nvlist(zc, configs);
1529 
1530         nvlist_free(configs);
1531 
1532         return (error);
1533 }
1534 
1535 /*
1536  * inputs:
1537  * zc_name              name of the pool
1538  *
1539  * outputs:
1540  * zc_cookie            real errno
1541  * zc_nvlist_dst        config nvlist
1542  * zc_nvlist_dst_size   size of config nvlist
1543  */
1544 static int
1545 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1546 {
1547         nvlist_t *config;
1548         int error;
1549         int ret = 0;
1550 
1551         error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1552             sizeof (zc->zc_value));
1553 
1554         if (config != NULL) {
1555                 ret = put_nvlist(zc, config);
1556                 nvlist_free(config);
1557 
1558                 /*
1559                  * The config may be present even if 'error' is non-zero.
1560                  * In this case we return success, and preserve the real errno
1561                  * in 'zc_cookie'.
1562                  */
1563                 zc->zc_cookie = error;
1564         } else {
1565                 ret = error;
1566         }
1567 
1568         return (ret);
1569 }
1570 
1571 /*
1572  * Try to import the given pool, returning pool stats as appropriate so that
1573  * user land knows which devices are available and overall pool health.
1574  */
1575 static int
1576 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1577 {
1578         nvlist_t *tryconfig, *config;
1579         int error;
1580 
1581         if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1582             zc->zc_iflags, &tryconfig)) != 0)
1583                 return (error);
1584 
1585         config = spa_tryimport(tryconfig);
1586 
1587         nvlist_free(tryconfig);
1588 
1589         if (config == NULL)
1590                 return (SET_ERROR(EINVAL));
1591 
1592         error = put_nvlist(zc, config);
1593         nvlist_free(config);
1594 
1595         return (error);
1596 }
1597 
1598 /*
1599  * inputs:
1600  * zc_name              name of the pool
1601  * zc_cookie            scan func (pool_scan_func_t)
1602  */
1603 static int
1604 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1605 {
1606         spa_t *spa;
1607         int error;
1608 
1609         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1610                 return (error);
1611 
1612         if (zc->zc_cookie == POOL_SCAN_NONE)
1613                 error = spa_scan_stop(spa);
1614         else
1615                 error = spa_scan(spa, zc->zc_cookie);
1616 
1617         spa_close(spa, FTAG);
1618 
1619         return (error);
1620 }
1621 
1622 static int
1623 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1624 {
1625         spa_t *spa;
1626         int error;
1627 
1628         error = spa_open(zc->zc_name, &spa, FTAG);
1629         if (error == 0) {
1630                 spa_freeze(spa);
1631                 spa_close(spa, FTAG);
1632         }
1633         return (error);
1634 }
1635 
1636 static int
1637 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1638 {
1639         spa_t *spa;
1640         int error;
1641 
1642         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1643                 return (error);
1644 
1645         if (zc->zc_cookie < spa_version(spa) ||
1646             !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1647                 spa_close(spa, FTAG);
1648                 return (SET_ERROR(EINVAL));
1649         }
1650 
1651         spa_upgrade(spa, zc->zc_cookie);
1652         spa_close(spa, FTAG);
1653 
1654         return (error);
1655 }
1656 
1657 static int
1658 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1659 {
1660         spa_t *spa;
1661         char *hist_buf;
1662         uint64_t size;
1663         int error;
1664 
1665         if ((size = zc->zc_history_len) == 0)
1666                 return (SET_ERROR(EINVAL));
1667 
1668         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1669                 return (error);
1670 
1671         if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1672                 spa_close(spa, FTAG);
1673                 return (SET_ERROR(ENOTSUP));
1674         }
1675 
1676         hist_buf = kmem_alloc(size, KM_SLEEP);
1677         if ((error = spa_history_get(spa, &zc->zc_history_offset,
1678             &zc->zc_history_len, hist_buf)) == 0) {
1679                 error = ddi_copyout(hist_buf,
1680                     (void *)(uintptr_t)zc->zc_history,
1681                     zc->zc_history_len, zc->zc_iflags);
1682         }
1683 
1684         spa_close(spa, FTAG);
1685         kmem_free(hist_buf, size);
1686         return (error);
1687 }
1688 
1689 static int
1690 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1691 {
1692         spa_t *spa;
1693         int error;
1694 
1695         error = spa_open(zc->zc_name, &spa, FTAG);
1696         if (error == 0) {
1697                 error = spa_change_guid(spa);
1698                 spa_close(spa, FTAG);
1699         }
1700         return (error);
1701 }
1702 
1703 static int
1704 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1705 {
1706         return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1707 }
1708 
1709 /*
1710  * inputs:
1711  * zc_name              name of filesystem
1712  * zc_obj               object to find
1713  *
1714  * outputs:
1715  * zc_value             name of object
1716  */
1717 static int
1718 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1719 {
1720         objset_t *os;
1721         int error;
1722 
1723         /* XXX reading from objset not owned */
1724         if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1725                 return (error);
1726         if (dmu_objset_type(os) != DMU_OST_ZFS) {
1727                 dmu_objset_rele(os, FTAG);
1728                 return (SET_ERROR(EINVAL));
1729         }
1730         error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1731             sizeof (zc->zc_value));
1732         dmu_objset_rele(os, FTAG);
1733 
1734         return (error);
1735 }
1736 
1737 /*
1738  * inputs:
1739  * zc_name              name of filesystem
1740  * zc_obj               object to find
1741  *
1742  * outputs:
1743  * zc_stat              stats on object
1744  * zc_value             path to object
1745  */
1746 static int
1747 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1748 {
1749         objset_t *os;
1750         int error;
1751 
1752         /* XXX reading from objset not owned */
1753         if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1754                 return (error);
1755         if (dmu_objset_type(os) != DMU_OST_ZFS) {
1756                 dmu_objset_rele(os, FTAG);
1757                 return (SET_ERROR(EINVAL));
1758         }
1759         error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1760             sizeof (zc->zc_value));
1761         dmu_objset_rele(os, FTAG);
1762 
1763         return (error);
1764 }
1765 
1766 static int
1767 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1768 {
1769         spa_t *spa;
1770         int error;
1771         nvlist_t *config, **l2cache, **spares;
1772         uint_t nl2cache = 0, nspares = 0;
1773 
1774         error = spa_open(zc->zc_name, &spa, FTAG);
1775         if (error != 0)
1776                 return (error);
1777 
1778         error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1779             zc->zc_iflags, &config);
1780         (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1781             &l2cache, &nl2cache);
1782 
1783         (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1784             &spares, &nspares);
1785 
1786         /*
1787          * A root pool with concatenated devices is not supported.
1788          * Thus, can not add a device to a root pool.
1789          *
1790          * Intent log device can not be added to a rootpool because
1791          * during mountroot, zil is replayed, a seperated log device
1792          * can not be accessed during the mountroot time.
1793          *
1794          * l2cache and spare devices are ok to be added to a rootpool.
1795          */
1796         if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1797                 nvlist_free(config);
1798                 spa_close(spa, FTAG);
1799                 return (SET_ERROR(EDOM));
1800         }
1801 
1802         if (error == 0) {
1803                 error = spa_vdev_add(spa, config);
1804                 nvlist_free(config);
1805         }
1806         spa_close(spa, FTAG);
1807         return (error);
1808 }
1809 
1810 /*
1811  * inputs:
1812  * zc_name              name of the pool
1813  * zc_nvlist_conf       nvlist of devices to remove
1814  * zc_cookie            to stop the remove?
1815  */
1816 static int
1817 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1818 {
1819         spa_t *spa;
1820         int error;
1821 
1822         error = spa_open(zc->zc_name, &spa, FTAG);
1823         if (error != 0)
1824                 return (error);
1825         error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1826         spa_close(spa, FTAG);
1827         return (error);
1828 }
1829 
1830 static int
1831 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1832 {
1833         spa_t *spa;
1834         int error;
1835         vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1836 
1837         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1838                 return (error);
1839         switch (zc->zc_cookie) {
1840         case VDEV_STATE_ONLINE:
1841                 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1842                 break;
1843 
1844         case VDEV_STATE_OFFLINE:
1845                 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1846                 break;
1847 
1848         case VDEV_STATE_FAULTED:
1849                 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1850                     zc->zc_obj != VDEV_AUX_EXTERNAL)
1851                         zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1852 
1853                 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1854                 break;
1855 
1856         case VDEV_STATE_DEGRADED:
1857                 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1858                     zc->zc_obj != VDEV_AUX_EXTERNAL)
1859                         zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1860 
1861                 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1862                 break;
1863 
1864         default:
1865                 error = SET_ERROR(EINVAL);
1866         }
1867         zc->zc_cookie = newstate;
1868         spa_close(spa, FTAG);
1869         return (error);
1870 }
1871 
1872 static int
1873 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1874 {
1875         spa_t *spa;
1876         int replacing = zc->zc_cookie;
1877         nvlist_t *config;
1878         int error;
1879 
1880         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1881                 return (error);
1882 
1883         if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1884             zc->zc_iflags, &config)) == 0) {
1885                 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1886                 nvlist_free(config);
1887         }
1888 
1889         spa_close(spa, FTAG);
1890         return (error);
1891 }
1892 
1893 static int
1894 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1895 {
1896         spa_t *spa;
1897         int error;
1898 
1899         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1900                 return (error);
1901 
1902         error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
1903 
1904         spa_close(spa, FTAG);
1905         return (error);
1906 }
1907 
1908 static int
1909 zfs_ioc_vdev_split(zfs_cmd_t *zc)
1910 {
1911         spa_t *spa;
1912         nvlist_t *config, *props = NULL;
1913         int error;
1914         boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
1915 
1916         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1917                 return (error);
1918 
1919         if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1920             zc->zc_iflags, &config)) {
1921                 spa_close(spa, FTAG);
1922                 return (error);
1923         }
1924 
1925         if (zc->zc_nvlist_src_size != 0 && (error =
1926             get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1927             zc->zc_iflags, &props))) {
1928                 spa_close(spa, FTAG);
1929                 nvlist_free(config);
1930                 return (error);
1931         }
1932 
1933         error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
1934 
1935         spa_close(spa, FTAG);
1936 
1937         nvlist_free(config);
1938         nvlist_free(props);
1939 
1940         return (error);
1941 }
1942 
1943 static int
1944 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
1945 {
1946         spa_t *spa;
1947         char *path = zc->zc_value;
1948         uint64_t guid = zc->zc_guid;
1949         int error;
1950 
1951         error = spa_open(zc->zc_name, &spa, FTAG);
1952         if (error != 0)
1953                 return (error);
1954 
1955         error = spa_vdev_setpath(spa, guid, path);
1956         spa_close(spa, FTAG);
1957         return (error);
1958 }
1959 
1960 static int
1961 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
1962 {
1963         spa_t *spa;
1964         char *fru = zc->zc_value;
1965         uint64_t guid = zc->zc_guid;
1966         int error;
1967 
1968         error = spa_open(zc->zc_name, &spa, FTAG);
1969         if (error != 0)
1970                 return (error);
1971 
1972         error = spa_vdev_setfru(spa, guid, fru);
1973         spa_close(spa, FTAG);
1974         return (error);
1975 }
1976 
1977 static int
1978 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
1979 {
1980         int error = 0;
1981         nvlist_t *nv;
1982 
1983         dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1984 
1985         if (zc->zc_nvlist_dst != 0 &&
1986             (error = dsl_prop_get_all(os, &nv)) == 0) {
1987                 dmu_objset_stats(os, nv);
1988                 /*
1989                  * NB: zvol_get_stats() will read the objset contents,
1990                  * which we aren't supposed to do with a
1991                  * DS_MODE_USER hold, because it could be
1992                  * inconsistent.  So this is a bit of a workaround...
1993                  * XXX reading with out owning
1994                  */
1995                 if (!zc->zc_objset_stats.dds_inconsistent &&
1996                     dmu_objset_type(os) == DMU_OST_ZVOL) {
1997                         error = zvol_get_stats(os, nv);
1998                         if (error == EIO)
1999                                 return (error);
2000                         VERIFY0(error);
2001                 }
2002                 error = put_nvlist(zc, nv);
2003                 nvlist_free(nv);
2004         }
2005 
2006         return (error);
2007 }
2008 
2009 /*
2010  * inputs:
2011  * zc_name              name of filesystem
2012  * zc_nvlist_dst_size   size of buffer for property nvlist
2013  *
2014  * outputs:
2015  * zc_objset_stats      stats
2016  * zc_nvlist_dst        property nvlist
2017  * zc_nvlist_dst_size   size of property nvlist
2018  */
2019 static int
2020 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2021 {
2022         objset_t *os;
2023         int error;
2024 
2025         error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2026         if (error == 0) {
2027                 error = zfs_ioc_objset_stats_impl(zc, os);
2028                 dmu_objset_rele(os, FTAG);
2029         }
2030 
2031         return (error);
2032 }
2033 
2034 /*
2035  * inputs:
2036  * zc_name              name of filesystem
2037  * zc_nvlist_dst_size   size of buffer for property nvlist
2038  *
2039  * outputs:
2040  * zc_nvlist_dst        received property nvlist
2041  * zc_nvlist_dst_size   size of received property nvlist
2042  *
2043  * Gets received properties (distinct from local properties on or after
2044  * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2045  * local property values.
2046  */
2047 static int
2048 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2049 {
2050         int error = 0;
2051         nvlist_t *nv;
2052 
2053         /*
2054          * Without this check, we would return local property values if the
2055          * caller has not already received properties on or after
2056          * SPA_VERSION_RECVD_PROPS.
2057          */
2058         if (!dsl_prop_get_hasrecvd(zc->zc_name))
2059                 return (SET_ERROR(ENOTSUP));
2060 
2061         if (zc->zc_nvlist_dst != 0 &&
2062             (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2063                 error = put_nvlist(zc, nv);
2064                 nvlist_free(nv);
2065         }
2066 
2067         return (error);
2068 }
2069 
2070 static int
2071 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2072 {
2073         uint64_t value;
2074         int error;
2075 
2076         /*
2077          * zfs_get_zplprop() will either find a value or give us
2078          * the default value (if there is one).
2079          */
2080         if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2081                 return (error);
2082         VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2083         return (0);
2084 }
2085 
2086 /*
2087  * inputs:
2088  * zc_name              name of filesystem
2089  * zc_nvlist_dst_size   size of buffer for zpl property nvlist
2090  *
2091  * outputs:
2092  * zc_nvlist_dst        zpl property nvlist
2093  * zc_nvlist_dst_size   size of zpl property nvlist
2094  */
2095 static int
2096 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2097 {
2098         objset_t *os;
2099         int err;
2100 
2101         /* XXX reading without owning */
2102         if (err = dmu_objset_hold(zc->zc_name, FTAG, &os))
2103                 return (err);
2104 
2105         dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2106 
2107         /*
2108          * NB: nvl_add_zplprop() will read the objset contents,
2109          * which we aren't supposed to do with a DS_MODE_USER
2110          * hold, because it could be inconsistent.
2111          */
2112         if (zc->zc_nvlist_dst != NULL &&
2113             !zc->zc_objset_stats.dds_inconsistent &&
2114             dmu_objset_type(os) == DMU_OST_ZFS) {
2115                 nvlist_t *nv;
2116 
2117                 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2118                 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2119                     (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2120                     (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2121                     (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2122                         err = put_nvlist(zc, nv);
2123                 nvlist_free(nv);
2124         } else {
2125                 err = SET_ERROR(ENOENT);
2126         }
2127         dmu_objset_rele(os, FTAG);
2128         return (err);
2129 }
2130 
2131 static boolean_t
2132 dataset_name_hidden(const char *name)
2133 {
2134         /*
2135          * Skip over datasets that are not visible in this zone,
2136          * internal datasets (which have a $ in their name), and
2137          * temporary datasets (which have a % in their name).
2138          */
2139         if (strchr(name, '$') != NULL)
2140                 return (B_TRUE);
2141         if (strchr(name, '%') != NULL)
2142                 return (B_TRUE);
2143         if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
2144                 return (B_TRUE);
2145         return (B_FALSE);
2146 }
2147 
2148 /*
2149  * inputs:
2150  * zc_name              name of filesystem
2151  * zc_cookie            zap cursor
2152  * zc_nvlist_dst_size   size of buffer for property nvlist
2153  *
2154  * outputs:
2155  * zc_name              name of next filesystem
2156  * zc_cookie            zap cursor
2157  * zc_objset_stats      stats
2158  * zc_nvlist_dst        property nvlist
2159  * zc_nvlist_dst_size   size of property nvlist
2160  */
2161 static int
2162 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2163 {
2164         objset_t *os;
2165         int error;
2166         char *p;
2167         size_t orig_len = strlen(zc->zc_name);
2168 
2169 top:
2170         if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) {
2171                 if (error == ENOENT)
2172                         error = SET_ERROR(ESRCH);
2173                 return (error);
2174         }
2175 
2176         p = strrchr(zc->zc_name, '/');
2177         if (p == NULL || p[1] != '\0')
2178                 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2179         p = zc->zc_name + strlen(zc->zc_name);
2180 
2181         do {
2182                 error = dmu_dir_list_next(os,
2183                     sizeof (zc->zc_name) - (p - zc->zc_name), p,
2184                     NULL, &zc->zc_cookie);
2185                 if (error == ENOENT)
2186                         error = SET_ERROR(ESRCH);
2187         } while (error == 0 && dataset_name_hidden(zc->zc_name));
2188         dmu_objset_rele(os, FTAG);
2189 
2190         /*
2191          * If it's an internal dataset (ie. with a '$' in its name),
2192          * don't try to get stats for it, otherwise we'll return ENOENT.
2193          */
2194         if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2195                 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2196                 if (error == ENOENT) {
2197                         /* We lost a race with destroy, get the next one. */
2198                         zc->zc_name[orig_len] = '\0';
2199                         goto top;
2200                 }
2201         }
2202         return (error);
2203 }
2204 
2205 /*
2206  * inputs:
2207  * zc_name              name of filesystem
2208  * zc_cookie            zap cursor
2209  * zc_nvlist_dst_size   size of buffer for property nvlist
2210  *
2211  * outputs:
2212  * zc_name              name of next snapshot
2213  * zc_objset_stats      stats
2214  * zc_nvlist_dst        property nvlist
2215  * zc_nvlist_dst_size   size of property nvlist
2216  */
2217 static int
2218 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2219 {
2220         objset_t *os;
2221         int error;
2222 
2223         error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2224         if (error != 0) {
2225                 return (error == ENOENT ? ESRCH : error);
2226         }
2227 
2228         /*
2229          * A dataset name of maximum length cannot have any snapshots,
2230          * so exit immediately.
2231          */
2232         if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
2233                 dmu_objset_rele(os, FTAG);
2234                 return (SET_ERROR(ESRCH));
2235         }
2236 
2237         error = dmu_snapshot_list_next(os,
2238             sizeof (zc->zc_name) - strlen(zc->zc_name),
2239             zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2240             NULL);
2241 
2242         if (error == 0) {
2243                 dsl_dataset_t *ds;
2244                 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2245 
2246                 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2247                 if (error == 0) {
2248                         objset_t *ossnap;
2249 
2250                         error = dmu_objset_from_ds(ds, &ossnap);
2251                         if (error == 0)
2252                                 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2253                         dsl_dataset_rele(ds, FTAG);
2254                 }
2255         } else if (error == ENOENT) {
2256                 error = SET_ERROR(ESRCH);
2257         }
2258 
2259         dmu_objset_rele(os, FTAG);
2260         /* if we failed, undo the @ that we tacked on to zc_name */
2261         if (error != 0)
2262                 *strchr(zc->zc_name, '@') = '\0';
2263         return (error);
2264 }
2265 
2266 static int
2267 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2268 {
2269         const char *propname = nvpair_name(pair);
2270         uint64_t *valary;
2271         unsigned int vallen;
2272         const char *domain;
2273         char *dash;
2274         zfs_userquota_prop_t type;
2275         uint64_t rid;
2276         uint64_t quota;
2277         zfsvfs_t *zfsvfs;
2278         int err;
2279 
2280         if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2281                 nvlist_t *attrs;
2282                 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2283                 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2284                     &pair) != 0)
2285                         return (SET_ERROR(EINVAL));
2286         }
2287 
2288         /*
2289          * A correctly constructed propname is encoded as
2290          * userquota@<rid>-<domain>.
2291          */
2292         if ((dash = strchr(propname, '-')) == NULL ||
2293             nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2294             vallen != 3)
2295                 return (SET_ERROR(EINVAL));
2296 
2297         domain = dash + 1;
2298         type = valary[0];
2299         rid = valary[1];
2300         quota = valary[2];
2301 
2302         err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2303         if (err == 0) {
2304                 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2305                 zfsvfs_rele(zfsvfs, FTAG);
2306         }
2307 
2308         return (err);
2309 }
2310 
2311 /*
2312  * If the named property is one that has a special function to set its value,
2313  * return 0 on success and a positive error code on failure; otherwise if it is
2314  * not one of the special properties handled by this function, return -1.
2315  *
2316  * XXX: It would be better for callers of the property interface if we handled
2317  * these special cases in dsl_prop.c (in the dsl layer).
2318  */
2319 static int
2320 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2321     nvpair_t *pair)
2322 {
2323         const char *propname = nvpair_name(pair);
2324         zfs_prop_t prop = zfs_name_to_prop(propname);
2325         uint64_t intval;
2326         int err;
2327 
2328         if (prop == ZPROP_INVAL) {
2329                 if (zfs_prop_userquota(propname))
2330                         return (zfs_prop_set_userquota(dsname, pair));
2331                 return (-1);
2332         }
2333 
2334         if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2335                 nvlist_t *attrs;
2336                 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2337                 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2338                     &pair) == 0);
2339         }
2340 
2341         if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2342                 return (-1);
2343 
2344         VERIFY(0 == nvpair_value_uint64(pair, &intval));
2345 
2346         switch (prop) {
2347         case ZFS_PROP_QUOTA:
2348                 err = dsl_dir_set_quota(dsname, source, intval);
2349                 break;
2350         case ZFS_PROP_REFQUOTA:
2351                 err = dsl_dataset_set_refquota(dsname, source, intval);
2352                 break;
2353         case ZFS_PROP_RESERVATION:
2354                 err = dsl_dir_set_reservation(dsname, source, intval);
2355                 break;
2356         case ZFS_PROP_REFRESERVATION:
2357                 err = dsl_dataset_set_refreservation(dsname, source, intval);
2358                 break;
2359         case ZFS_PROP_VOLSIZE:
2360                 err = zvol_set_volsize(dsname, intval);
2361                 break;
2362         case ZFS_PROP_VERSION:
2363         {
2364                 zfsvfs_t *zfsvfs;
2365 
2366                 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2367                         break;
2368 
2369                 err = zfs_set_version(zfsvfs, intval);
2370                 zfsvfs_rele(zfsvfs, FTAG);
2371 
2372                 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2373                         zfs_cmd_t *zc;
2374 
2375                         zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2376                         (void) strcpy(zc->zc_name, dsname);
2377                         (void) zfs_ioc_userspace_upgrade(zc);
2378                         kmem_free(zc, sizeof (zfs_cmd_t));
2379                 }
2380                 break;
2381         }
2382         case ZFS_PROP_COMPRESSION:
2383         {
2384                 if (intval == ZIO_COMPRESS_LZ4) {
2385                         spa_t *spa;
2386 
2387                         if ((err = spa_open(dsname, &spa, FTAG)) != 0)
2388                                 return (err);
2389 
2390                         /*
2391                          * Setting the LZ4 compression algorithm activates
2392                          * the feature.
2393                          */
2394                         if (!spa_feature_is_active(spa,
2395                             SPA_FEATURE_LZ4_COMPRESS)) {
2396                                 if ((err = zfs_prop_activate_feature(spa,
2397                                     SPA_FEATURE_LZ4_COMPRESS)) != 0) {
2398                                         spa_close(spa, FTAG);
2399                                         return (err);
2400                                 }
2401                         }
2402 
2403                         spa_close(spa, FTAG);
2404                 }
2405                 /*
2406                  * We still want the default set action to be performed in the
2407                  * caller, we only performed zfeature settings here.
2408                  */
2409                 err = -1;
2410                 break;
2411         }
2412 
2413         default:
2414                 err = -1;
2415         }
2416 
2417         return (err);
2418 }
2419 
2420 /*
2421  * This function is best effort. If it fails to set any of the given properties,
2422  * it continues to set as many as it can and returns the last error
2423  * encountered. If the caller provides a non-NULL errlist, it will be filled in
2424  * with the list of names of all the properties that failed along with the
2425  * corresponding error numbers.
2426  *
2427  * If every property is set successfully, zero is returned and errlist is not
2428  * modified.
2429  */
2430 int
2431 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2432     nvlist_t *errlist)
2433 {
2434         nvpair_t *pair;
2435         nvpair_t *propval;
2436         int rv = 0;
2437         uint64_t intval;
2438         char *strval;
2439         nvlist_t *genericnvl = fnvlist_alloc();
2440         nvlist_t *retrynvl = fnvlist_alloc();
2441 
2442 retry:
2443         pair = NULL;
2444         while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2445                 const char *propname = nvpair_name(pair);
2446                 zfs_prop_t prop = zfs_name_to_prop(propname);
2447                 int err = 0;
2448 
2449                 /* decode the property value */
2450                 propval = pair;
2451                 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2452                         nvlist_t *attrs;
2453                         attrs = fnvpair_value_nvlist(pair);
2454                         if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2455                             &propval) != 0)
2456                                 err = SET_ERROR(EINVAL);
2457                 }
2458 
2459                 /* Validate value type */
2460                 if (err == 0 && prop == ZPROP_INVAL) {
2461                         if (zfs_prop_user(propname)) {
2462                                 if (nvpair_type(propval) != DATA_TYPE_STRING)
2463                                         err = SET_ERROR(EINVAL);
2464                         } else if (zfs_prop_userquota(propname)) {
2465                                 if (nvpair_type(propval) !=
2466                                     DATA_TYPE_UINT64_ARRAY)
2467                                         err = SET_ERROR(EINVAL);
2468                         } else {
2469                                 err = SET_ERROR(EINVAL);
2470                         }
2471                 } else if (err == 0) {
2472                         if (nvpair_type(propval) == DATA_TYPE_STRING) {
2473                                 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2474                                         err = SET_ERROR(EINVAL);
2475                         } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2476                                 const char *unused;
2477 
2478                                 intval = fnvpair_value_uint64(propval);
2479 
2480                                 switch (zfs_prop_get_type(prop)) {
2481                                 case PROP_TYPE_NUMBER:
2482                                         break;
2483                                 case PROP_TYPE_STRING:
2484                                         err = SET_ERROR(EINVAL);
2485                                         break;
2486                                 case PROP_TYPE_INDEX:
2487                                         if (zfs_prop_index_to_string(prop,
2488                                             intval, &unused) != 0)
2489                                                 err = SET_ERROR(EINVAL);
2490                                         break;
2491                                 default:
2492                                         cmn_err(CE_PANIC,
2493                                             "unknown property type");
2494                                 }
2495                         } else {
2496                                 err = SET_ERROR(EINVAL);
2497                         }
2498                 }
2499 
2500                 /* Validate permissions */
2501                 if (err == 0)
2502                         err = zfs_check_settable(dsname, pair, CRED());
2503 
2504                 if (err == 0) {
2505                         err = zfs_prop_set_special(dsname, source, pair);
2506                         if (err == -1) {
2507                                 /*
2508                                  * For better performance we build up a list of
2509                                  * properties to set in a single transaction.
2510                                  */
2511                                 err = nvlist_add_nvpair(genericnvl, pair);
2512                         } else if (err != 0 && nvl != retrynvl) {
2513                                 /*
2514                                  * This may be a spurious error caused by
2515                                  * receiving quota and reservation out of order.
2516                                  * Try again in a second pass.
2517                                  */
2518                                 err = nvlist_add_nvpair(retrynvl, pair);
2519                         }
2520                 }
2521 
2522                 if (err != 0) {
2523                         if (errlist != NULL)
2524                                 fnvlist_add_int32(errlist, propname, err);
2525                         rv = err;
2526                 }
2527         }
2528 
2529         if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2530                 nvl = retrynvl;
2531                 goto retry;
2532         }
2533 
2534         if (!nvlist_empty(genericnvl) &&
2535             dsl_props_set(dsname, source, genericnvl) != 0) {
2536                 /*
2537                  * If this fails, we still want to set as many properties as we
2538                  * can, so try setting them individually.
2539                  */
2540                 pair = NULL;
2541                 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2542                         const char *propname = nvpair_name(pair);
2543                         int err = 0;
2544 
2545                         propval = pair;
2546                         if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2547                                 nvlist_t *attrs;
2548                                 attrs = fnvpair_value_nvlist(pair);
2549                                 propval = fnvlist_lookup_nvpair(attrs,
2550                                     ZPROP_VALUE);
2551                         }
2552 
2553                         if (nvpair_type(propval) == DATA_TYPE_STRING) {
2554                                 strval = fnvpair_value_string(propval);
2555                                 err = dsl_prop_set_string(dsname, propname,
2556                                     source, strval);
2557                         } else {
2558                                 intval = fnvpair_value_uint64(propval);
2559                                 err = dsl_prop_set_int(dsname, propname, source,
2560                                     intval);
2561                         }
2562 
2563                         if (err != 0) {
2564                                 if (errlist != NULL) {
2565                                         fnvlist_add_int32(errlist, propname,
2566                                             err);
2567                                 }
2568                                 rv = err;
2569                         }
2570                 }
2571         }
2572         nvlist_free(genericnvl);
2573         nvlist_free(retrynvl);
2574 
2575         return (rv);
2576 }
2577 
2578 /*
2579  * Check that all the properties are valid user properties.
2580  */
2581 static int
2582 zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2583 {
2584         nvpair_t *pair = NULL;
2585         int error = 0;
2586 
2587         while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2588                 const char *propname = nvpair_name(pair);
2589                 char *valstr;
2590 
2591                 if (!zfs_prop_user(propname) ||
2592                     nvpair_type(pair) != DATA_TYPE_STRING)
2593                         return (SET_ERROR(EINVAL));
2594 
2595                 if (error = zfs_secpolicy_write_perms(fsname,
2596                     ZFS_DELEG_PERM_USERPROP, CRED()))
2597                         return (error);
2598 
2599                 if (strlen(propname) >= ZAP_MAXNAMELEN)
2600                         return (SET_ERROR(ENAMETOOLONG));
2601 
2602                 VERIFY(nvpair_value_string(pair, &valstr) == 0);
2603                 if (strlen(valstr) >= ZAP_MAXVALUELEN)
2604                         return (E2BIG);
2605         }
2606         return (0);
2607 }
2608 
2609 static void
2610 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2611 {
2612         nvpair_t *pair;
2613 
2614         VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2615 
2616         pair = NULL;
2617         while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2618                 if (nvlist_exists(skipped, nvpair_name(pair)))
2619                         continue;
2620 
2621                 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2622         }
2623 }
2624 
2625 static int
2626 clear_received_props(const char *dsname, nvlist_t *props,
2627     nvlist_t *skipped)
2628 {
2629         int err = 0;
2630         nvlist_t *cleared_props = NULL;
2631         props_skip(props, skipped, &cleared_props);
2632         if (!nvlist_empty(cleared_props)) {
2633                 /*
2634                  * Acts on local properties until the dataset has received
2635                  * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2636                  */
2637                 zprop_source_t flags = (ZPROP_SRC_NONE |
2638                     (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2639                 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2640         }
2641         nvlist_free(cleared_props);
2642         return (err);
2643 }
2644 
2645 /*
2646  * inputs:
2647  * zc_name              name of filesystem
2648  * zc_value             name of property to set
2649  * zc_nvlist_src{_size} nvlist of properties to apply
2650  * zc_cookie            received properties flag
2651  *
2652  * outputs:
2653  * zc_nvlist_dst{_size} error for each unapplied received property
2654  */
2655 static int
2656 zfs_ioc_set_prop(zfs_cmd_t *zc)
2657 {
2658         nvlist_t *nvl;
2659         boolean_t received = zc->zc_cookie;
2660         zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2661             ZPROP_SRC_LOCAL);
2662         nvlist_t *errors;
2663         int error;
2664 
2665         if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2666             zc->zc_iflags, &nvl)) != 0)
2667                 return (error);
2668 
2669         if (received) {
2670                 nvlist_t *origprops;
2671 
2672                 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2673                         (void) clear_received_props(zc->zc_name,
2674                             origprops, nvl);
2675                         nvlist_free(origprops);
2676                 }
2677 
2678                 error = dsl_prop_set_hasrecvd(zc->zc_name);
2679         }
2680 
2681         errors = fnvlist_alloc();
2682         if (error == 0)
2683                 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2684 
2685         if (zc->zc_nvlist_dst != NULL && errors != NULL) {
2686                 (void) put_nvlist(zc, errors);
2687         }
2688 
2689         nvlist_free(errors);
2690         nvlist_free(nvl);
2691         return (error);
2692 }
2693 
2694 /*
2695  * inputs:
2696  * zc_name              name of filesystem
2697  * zc_value             name of property to inherit
2698  * zc_cookie            revert to received value if TRUE
2699  *
2700  * outputs:             none
2701  */
2702 static int
2703 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2704 {
2705         const char *propname = zc->zc_value;
2706         zfs_prop_t prop = zfs_name_to_prop(propname);
2707         boolean_t received = zc->zc_cookie;
2708         zprop_source_t source = (received
2709             ? ZPROP_SRC_NONE            /* revert to received value, if any */
2710             : ZPROP_SRC_INHERITED);     /* explicitly inherit */
2711 
2712         if (received) {
2713                 nvlist_t *dummy;
2714                 nvpair_t *pair;
2715                 zprop_type_t type;
2716                 int err;
2717 
2718                 /*
2719                  * zfs_prop_set_special() expects properties in the form of an
2720                  * nvpair with type info.
2721                  */
2722                 if (prop == ZPROP_INVAL) {
2723                         if (!zfs_prop_user(propname))
2724                                 return (SET_ERROR(EINVAL));
2725 
2726                         type = PROP_TYPE_STRING;
2727                 } else if (prop == ZFS_PROP_VOLSIZE ||
2728                     prop == ZFS_PROP_VERSION) {
2729                         return (SET_ERROR(EINVAL));
2730                 } else {
2731                         type = zfs_prop_get_type(prop);
2732                 }
2733 
2734                 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2735 
2736                 switch (type) {
2737                 case PROP_TYPE_STRING:
2738                         VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2739                         break;
2740                 case PROP_TYPE_NUMBER:
2741                 case PROP_TYPE_INDEX:
2742                         VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2743                         break;
2744                 default:
2745                         nvlist_free(dummy);
2746                         return (SET_ERROR(EINVAL));
2747                 }
2748 
2749                 pair = nvlist_next_nvpair(dummy, NULL);
2750                 err = zfs_prop_set_special(zc->zc_name, source, pair);
2751                 nvlist_free(dummy);
2752                 if (err != -1)
2753                         return (err); /* special property already handled */
2754         } else {
2755                 /*
2756                  * Only check this in the non-received case. We want to allow
2757                  * 'inherit -S' to revert non-inheritable properties like quota
2758                  * and reservation to the received or default values even though
2759                  * they are not considered inheritable.
2760                  */
2761                 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2762                         return (SET_ERROR(EINVAL));
2763         }
2764 
2765         /* property name has been validated by zfs_secpolicy_inherit_prop() */
2766         return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source));
2767 }
2768 
2769 static int
2770 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2771 {
2772         nvlist_t *props;
2773         spa_t *spa;
2774         int error;
2775         nvpair_t *pair;
2776 
2777         if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2778             zc->zc_iflags, &props))
2779                 return (error);
2780 
2781         /*
2782          * If the only property is the configfile, then just do a spa_lookup()
2783          * to handle the faulted case.
2784          */
2785         pair = nvlist_next_nvpair(props, NULL);
2786         if (pair != NULL && strcmp(nvpair_name(pair),
2787             zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2788             nvlist_next_nvpair(props, pair) == NULL) {
2789                 mutex_enter(&spa_namespace_lock);
2790                 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2791                         spa_configfile_set(spa, props, B_FALSE);
2792                         spa_config_sync(spa, B_FALSE, B_TRUE);
2793                 }
2794                 mutex_exit(&spa_namespace_lock);
2795                 if (spa != NULL) {
2796                         nvlist_free(props);
2797                         return (0);
2798                 }
2799         }
2800 
2801         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2802                 nvlist_free(props);
2803                 return (error);
2804         }
2805 
2806         error = spa_prop_set(spa, props);
2807 
2808         nvlist_free(props);
2809         spa_close(spa, FTAG);
2810 
2811         return (error);
2812 }
2813 
2814 static int
2815 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2816 {
2817         spa_t *spa;
2818         int error;
2819         nvlist_t *nvp = NULL;
2820 
2821         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2822                 /*
2823                  * If the pool is faulted, there may be properties we can still
2824                  * get (such as altroot and cachefile), so attempt to get them
2825                  * anyway.
2826                  */
2827                 mutex_enter(&spa_namespace_lock);
2828                 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2829                         error = spa_prop_get(spa, &nvp);
2830                 mutex_exit(&spa_namespace_lock);
2831         } else {
2832                 error = spa_prop_get(spa, &nvp);
2833                 spa_close(spa, FTAG);
2834         }
2835 
2836         if (error == 0 && zc->zc_nvlist_dst != NULL)
2837                 error = put_nvlist(zc, nvp);
2838         else
2839                 error = SET_ERROR(EFAULT);
2840 
2841         nvlist_free(nvp);
2842         return (error);
2843 }
2844 
2845 /*
2846  * inputs:
2847  * zc_name              name of filesystem
2848  * zc_nvlist_src{_size} nvlist of delegated permissions
2849  * zc_perm_action       allow/unallow flag
2850  *
2851  * outputs:             none
2852  */
2853 static int
2854 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2855 {
2856         int error;
2857         nvlist_t *fsaclnv = NULL;
2858 
2859         if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2860             zc->zc_iflags, &fsaclnv)) != 0)
2861                 return (error);
2862 
2863         /*
2864          * Verify nvlist is constructed correctly
2865          */
2866         if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2867                 nvlist_free(fsaclnv);
2868                 return (SET_ERROR(EINVAL));
2869         }
2870 
2871         /*
2872          * If we don't have PRIV_SYS_MOUNT, then validate
2873          * that user is allowed to hand out each permission in
2874          * the nvlist(s)
2875          */
2876 
2877         error = secpolicy_zfs(CRED());
2878         if (error != 0) {
2879                 if (zc->zc_perm_action == B_FALSE) {
2880                         error = dsl_deleg_can_allow(zc->zc_name,
2881                             fsaclnv, CRED());
2882                 } else {
2883                         error = dsl_deleg_can_unallow(zc->zc_name,
2884                             fsaclnv, CRED());
2885                 }
2886         }
2887 
2888         if (error == 0)
2889                 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2890 
2891         nvlist_free(fsaclnv);
2892         return (error);
2893 }
2894 
2895 /*
2896  * inputs:
2897  * zc_name              name of filesystem
2898  *
2899  * outputs:
2900  * zc_nvlist_src{_size} nvlist of delegated permissions
2901  */
2902 static int
2903 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
2904 {
2905         nvlist_t *nvp;
2906         int error;
2907 
2908         if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
2909                 error = put_nvlist(zc, nvp);
2910                 nvlist_free(nvp);
2911         }
2912 
2913         return (error);
2914 }
2915 
2916 /*
2917  * Search the vfs list for a specified resource.  Returns a pointer to it
2918  * or NULL if no suitable entry is found. The caller of this routine
2919  * is responsible for releasing the returned vfs pointer.
2920  */
2921 static vfs_t *
2922 zfs_get_vfs(const char *resource)
2923 {
2924         struct vfs *vfsp;
2925         struct vfs *vfs_found = NULL;
2926 
2927         vfs_list_read_lock();
2928         vfsp = rootvfs;
2929         do {
2930                 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
2931                         VFS_HOLD(vfsp);
2932                         vfs_found = vfsp;
2933                         break;
2934                 }
2935                 vfsp = vfsp->vfs_next;
2936         } while (vfsp != rootvfs);
2937         vfs_list_unlock();
2938         return (vfs_found);
2939 }
2940 
2941 /* ARGSUSED */
2942 static void
2943 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
2944 {
2945         zfs_creat_t *zct = arg;
2946 
2947         zfs_create_fs(os, cr, zct->zct_zplprops, tx);
2948 }
2949 
2950 #define ZFS_PROP_UNDEFINED      ((uint64_t)-1)
2951 
2952 /*
2953  * inputs:
2954  * os                   parent objset pointer (NULL if root fs)
2955  * fuids_ok             fuids allowed in this version of the spa?
2956  * sa_ok                SAs allowed in this version of the spa?
2957  * createprops          list of properties requested by creator
2958  *
2959  * outputs:
2960  * zplprops     values for the zplprops we attach to the master node object
2961  * is_ci        true if requested file system will be purely case-insensitive
2962  *
2963  * Determine the settings for utf8only, normalization and
2964  * casesensitivity.  Specific values may have been requested by the
2965  * creator and/or we can inherit values from the parent dataset.  If
2966  * the file system is of too early a vintage, a creator can not
2967  * request settings for these properties, even if the requested
2968  * setting is the default value.  We don't actually want to create dsl
2969  * properties for these, so remove them from the source nvlist after
2970  * processing.
2971  */
2972 static int
2973 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
2974     boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
2975     nvlist_t *zplprops, boolean_t *is_ci)
2976 {
2977         uint64_t sense = ZFS_PROP_UNDEFINED;
2978         uint64_t norm = ZFS_PROP_UNDEFINED;
2979         uint64_t u8 = ZFS_PROP_UNDEFINED;
2980 
2981         ASSERT(zplprops != NULL);
2982 
2983         /*
2984          * Pull out creator prop choices, if any.
2985          */
2986         if (createprops) {
2987                 (void) nvlist_lookup_uint64(createprops,
2988                     zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
2989                 (void) nvlist_lookup_uint64(createprops,
2990                     zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
2991                 (void) nvlist_remove_all(createprops,
2992                     zfs_prop_to_name(ZFS_PROP_NORMALIZE));
2993                 (void) nvlist_lookup_uint64(createprops,
2994                     zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
2995                 (void) nvlist_remove_all(createprops,
2996                     zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
2997                 (void) nvlist_lookup_uint64(createprops,
2998                     zfs_prop_to_name(ZFS_PROP_CASE), &sense);
2999                 (void) nvlist_remove_all(createprops,
3000                     zfs_prop_to_name(ZFS_PROP_CASE));
3001         }
3002 
3003         /*
3004          * If the zpl version requested is whacky or the file system
3005          * or pool is version is too "young" to support normalization
3006          * and the creator tried to set a value for one of the props,
3007          * error out.
3008          */
3009         if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3010             (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3011             (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3012             (zplver < ZPL_VERSION_NORMALIZATION &&
3013             (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3014             sense != ZFS_PROP_UNDEFINED)))
3015                 return (SET_ERROR(ENOTSUP));
3016 
3017         /*
3018          * Put the version in the zplprops
3019          */
3020         VERIFY(nvlist_add_uint64(zplprops,
3021             zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3022 
3023         if (norm == ZFS_PROP_UNDEFINED)
3024                 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
3025         VERIFY(nvlist_add_uint64(zplprops,
3026             zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3027 
3028         /*
3029          * If we're normalizing, names must always be valid UTF-8 strings.
3030          */
3031         if (norm)
3032                 u8 = 1;
3033         if (u8 == ZFS_PROP_UNDEFINED)
3034                 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
3035         VERIFY(nvlist_add_uint64(zplprops,
3036             zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3037 
3038         if (sense == ZFS_PROP_UNDEFINED)
3039                 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
3040         VERIFY(nvlist_add_uint64(zplprops,
3041             zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3042 
3043         if (is_ci)
3044                 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
3045 
3046         return (0);
3047 }
3048 
3049 static int
3050 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3051     nvlist_t *zplprops, boolean_t *is_ci)
3052 {
3053         boolean_t fuids_ok, sa_ok;
3054         uint64_t zplver = ZPL_VERSION;
3055         objset_t *os = NULL;
3056         char parentname[MAXNAMELEN];
3057         char *cp;
3058         spa_t *spa;
3059         uint64_t spa_vers;
3060         int error;
3061 
3062         (void) strlcpy(parentname, dataset, sizeof (parentname));
3063         cp = strrchr(parentname, '/');
3064         ASSERT(cp != NULL);
3065         cp[0] = '\0';
3066 
3067         if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3068                 return (error);
3069 
3070         spa_vers = spa_version(spa);
3071         spa_close(spa, FTAG);
3072 
3073         zplver = zfs_zpl_version_map(spa_vers);
3074         fuids_ok = (zplver >= ZPL_VERSION_FUID);
3075         sa_ok = (zplver >= ZPL_VERSION_SA);
3076 
3077         /*
3078          * Open parent object set so we can inherit zplprop values.
3079          */
3080         if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3081                 return (error);
3082 
3083         error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3084             zplprops, is_ci);
3085         dmu_objset_rele(os, FTAG);
3086         return (error);
3087 }
3088 
3089 static int
3090 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3091     nvlist_t *zplprops, boolean_t *is_ci)
3092 {
3093         boolean_t fuids_ok;
3094         boolean_t sa_ok;
3095         uint64_t zplver = ZPL_VERSION;
3096         int error;
3097 
3098         zplver = zfs_zpl_version_map(spa_vers);
3099         fuids_ok = (zplver >= ZPL_VERSION_FUID);
3100         sa_ok = (zplver >= ZPL_VERSION_SA);
3101 
3102         error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3103             createprops, zplprops, is_ci);
3104         return (error);
3105 }
3106 
3107 /*
3108  * innvl: {
3109  *     "type" -> dmu_objset_type_t (int32)
3110  *     (optional) "props" -> { prop -> value }
3111  * }
3112  *
3113  * outnvl: propname -> error code (int32)
3114  */
3115 static int
3116 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3117 {
3118         int error = 0;
3119         zfs_creat_t zct = { 0 };
3120         nvlist_t *nvprops = NULL;
3121         void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3122         int32_t type32;
3123         dmu_objset_type_t type;
3124         boolean_t is_insensitive = B_FALSE;
3125 
3126         if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3127                 return (SET_ERROR(EINVAL));
3128         type = type32;
3129         (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3130 
3131         switch (type) {
3132         case DMU_OST_ZFS:
3133                 cbfunc = zfs_create_cb;
3134                 break;
3135 
3136         case DMU_OST_ZVOL:
3137                 cbfunc = zvol_create_cb;
3138                 break;
3139 
3140         default:
3141                 cbfunc = NULL;
3142                 break;
3143         }
3144         if (strchr(fsname, '@') ||
3145             strchr(fsname, '%'))
3146                 return (SET_ERROR(EINVAL));
3147 
3148         zct.zct_props = nvprops;
3149 
3150         if (cbfunc == NULL)
3151                 return (SET_ERROR(EINVAL));
3152 
3153         if (type == DMU_OST_ZVOL) {
3154                 uint64_t volsize, volblocksize;
3155 
3156                 if (nvprops == NULL)
3157                         return (SET_ERROR(EINVAL));
3158                 if (nvlist_lookup_uint64(nvprops,
3159                     zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3160                         return (SET_ERROR(EINVAL));
3161 
3162                 if ((error = nvlist_lookup_uint64(nvprops,
3163                     zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3164                     &volblocksize)) != 0 && error != ENOENT)
3165                         return (SET_ERROR(EINVAL));
3166 
3167                 if (error != 0)
3168                         volblocksize = zfs_prop_default_numeric(
3169                             ZFS_PROP_VOLBLOCKSIZE);
3170 
3171                 if ((error = zvol_check_volblocksize(
3172                     volblocksize)) != 0 ||
3173                     (error = zvol_check_volsize(volsize,
3174                     volblocksize)) != 0)
3175                         return (error);
3176         } else if (type == DMU_OST_ZFS) {
3177                 int error;
3178 
3179                 /*
3180                  * We have to have normalization and
3181                  * case-folding flags correct when we do the
3182                  * file system creation, so go figure them out
3183                  * now.
3184                  */
3185                 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3186                     NV_UNIQUE_NAME, KM_SLEEP) == 0);
3187                 error = zfs_fill_zplprops(fsname, nvprops,
3188                     zct.zct_zplprops, &is_insensitive);
3189                 if (error != 0) {
3190                         nvlist_free(zct.zct_zplprops);
3191                         return (error);
3192                 }
3193         }
3194 
3195         error = dmu_objset_create(fsname, type,
3196             is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3197         nvlist_free(zct.zct_zplprops);
3198 
3199         /*
3200          * It would be nice to do this atomically.
3201          */
3202         if (error == 0) {
3203                 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3204                     nvprops, outnvl);
3205                 if (error != 0)
3206                         (void) dsl_destroy_head(fsname);
3207         }
3208         return (error);
3209 }
3210 
3211 /*
3212  * innvl: {
3213  *     "origin" -> name of origin snapshot
3214  *     (optional) "props" -> { prop -> value }
3215  * }
3216  *
3217  * outnvl: propname -> error code (int32)
3218  */
3219 static int
3220 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3221 {
3222         int error = 0;
3223         nvlist_t *nvprops = NULL;
3224         char *origin_name;
3225 
3226         if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3227                 return (SET_ERROR(EINVAL));
3228         (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3229 
3230         if (strchr(fsname, '@') ||
3231             strchr(fsname, '%'))
3232                 return (SET_ERROR(EINVAL));
3233 
3234         if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3235                 return (SET_ERROR(EINVAL));
3236         error = dmu_objset_clone(fsname, origin_name);
3237         if (error != 0)
3238                 return (error);
3239 
3240         /*
3241          * It would be nice to do this atomically.
3242          */
3243         if (error == 0) {
3244                 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3245                     nvprops, outnvl);
3246                 if (error != 0)
3247                         (void) dsl_destroy_head(fsname);
3248         }
3249         return (error);
3250 }
3251 
3252 /*
3253  * innvl: {
3254  *     "snaps" -> { snapshot1, snapshot2 }
3255  *     (optional) "props" -> { prop -> value (string) }
3256  * }
3257  *
3258  * outnvl: snapshot -> error code (int32)
3259  */
3260 static int
3261 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3262 {
3263         nvlist_t *snaps;
3264         nvlist_t *props = NULL;
3265         int error, poollen;
3266         nvpair_t *pair;
3267 
3268         (void) nvlist_lookup_nvlist(innvl, "props", &props);
3269         if ((error = zfs_check_userprops(poolname, props)) != 0)
3270                 return (error);
3271 
3272         if (!nvlist_empty(props) &&
3273             zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3274                 return (SET_ERROR(ENOTSUP));
3275 
3276         if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3277                 return (SET_ERROR(EINVAL));
3278         poollen = strlen(poolname);
3279         for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3280             pair = nvlist_next_nvpair(snaps, pair)) {
3281                 const char *name = nvpair_name(pair);
3282                 const char *cp = strchr(name, '@');
3283 
3284                 /*
3285                  * The snap name must contain an @, and the part after it must
3286                  * contain only valid characters.
3287                  */
3288                 if (cp == NULL || snapshot_namecheck(cp + 1, NULL, NULL) != 0)
3289                         return (SET_ERROR(EINVAL));
3290 
3291                 /*
3292                  * The snap must be in the specified pool.
3293                  */
3294                 if (strncmp(name, poolname, poollen) != 0 ||
3295                     (name[poollen] != '/' && name[poollen] != '@'))
3296                         return (SET_ERROR(EXDEV));
3297 
3298                 /* This must be the only snap of this fs. */
3299                 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3300                     pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3301                         if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3302                             == 0) {
3303                                 return (SET_ERROR(EXDEV));
3304                         }
3305                 }
3306         }
3307 
3308         error = dsl_dataset_snapshot(snaps, props, outnvl);
3309         return (error);
3310 }
3311 
3312 /*
3313  * innvl: "message" -> string
3314  */
3315 /* ARGSUSED */
3316 static int
3317 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3318 {
3319         char *message;
3320         spa_t *spa;
3321         int error;
3322         char *poolname;
3323 
3324         /*
3325          * The poolname in the ioctl is not set, we get it from the TSD,
3326          * which was set at the end of the last successful ioctl that allows
3327          * logging.  The secpolicy func already checked that it is set.
3328          * Only one log ioctl is allowed after each successful ioctl, so
3329          * we clear the TSD here.
3330          */
3331         poolname = tsd_get(zfs_allow_log_key);
3332         (void) tsd_set(zfs_allow_log_key, NULL);
3333         error = spa_open(poolname, &spa, FTAG);
3334         strfree(poolname);
3335         if (error != 0)
3336                 return (error);
3337 
3338         if (nvlist_lookup_string(innvl, "message", &message) != 0)  {
3339                 spa_close(spa, FTAG);
3340                 return (SET_ERROR(EINVAL));
3341         }
3342 
3343         if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3344                 spa_close(spa, FTAG);
3345                 return (SET_ERROR(ENOTSUP));
3346         }
3347 
3348         error = spa_history_log(spa, message);
3349         spa_close(spa, FTAG);
3350         return (error);
3351 }
3352 
3353 /*
3354  * The dp_config_rwlock must not be held when calling this, because the
3355  * unmount may need to write out data.
3356  *
3357  * This function is best-effort.  Callers must deal gracefully if it
3358  * remains mounted (or is remounted after this call).
3359  *
3360  * Returns 0 if the argument is not a snapshot, or it is not currently a
3361  * filesystem, or we were able to unmount it.  Returns error code otherwise.
3362  */
3363 int
3364 zfs_unmount_snap(const char *snapname)
3365 {
3366         vfs_t *vfsp;
3367         zfsvfs_t *zfsvfs;
3368         int err;
3369 
3370         if (strchr(snapname, '@') == NULL)
3371                 return (0);
3372 
3373         vfsp = zfs_get_vfs(snapname);
3374         if (vfsp == NULL)
3375                 return (0);
3376 
3377         zfsvfs = vfsp->vfs_data;
3378         ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os)));
3379 
3380         err = vn_vfswlock(vfsp->vfs_vnodecovered);
3381         VFS_RELE(vfsp);
3382         if (err != 0)
3383                 return (SET_ERROR(err));
3384 
3385         /*
3386          * Always force the unmount for snapshots.
3387          */
3388         (void) dounmount(vfsp, MS_FORCE, kcred);
3389         return (0);
3390 }
3391 
3392 /* ARGSUSED */
3393 static int
3394 zfs_unmount_snap_cb(const char *snapname, void *arg)
3395 {
3396         return (zfs_unmount_snap(snapname));
3397 }
3398 
3399 /*
3400  * When a clone is destroyed, its origin may also need to be destroyed,
3401  * in which case it must be unmounted.  This routine will do that unmount
3402  * if necessary.
3403  */
3404 void
3405 zfs_destroy_unmount_origin(const char *fsname)
3406 {
3407         int error;
3408         objset_t *os;
3409         dsl_dataset_t *ds;
3410 
3411         error = dmu_objset_hold(fsname, FTAG, &os);
3412         if (error != 0)
3413                 return;
3414         ds = dmu_objset_ds(os);
3415         if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3416                 char originname[MAXNAMELEN];
3417                 dsl_dataset_name(ds->ds_prev, originname);
3418                 dmu_objset_rele(os, FTAG);
3419                 (void) zfs_unmount_snap(originname);
3420         } else {
3421                 dmu_objset_rele(os, FTAG);
3422         }
3423 }
3424 
3425 /*
3426  * innvl: {
3427  *     "snaps" -> { snapshot1, snapshot2 }
3428  *     (optional boolean) "defer"
3429  * }
3430  *
3431  * outnvl: snapshot -> error code (int32)
3432  *
3433  */
3434 static int
3435 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3436 {
3437         int error, poollen;
3438         nvlist_t *snaps;
3439         nvpair_t *pair;
3440         boolean_t defer;
3441 
3442         if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3443                 return (SET_ERROR(EINVAL));
3444         defer = nvlist_exists(innvl, "defer");
3445 
3446         poollen = strlen(poolname);
3447         for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3448             pair = nvlist_next_nvpair(snaps, pair)) {
3449                 const char *name = nvpair_name(pair);
3450 
3451                 /*
3452                  * The snap must be in the specified pool.
3453                  */
3454                 if (strncmp(name, poolname, poollen) != 0 ||
3455                     (name[poollen] != '/' && name[poollen] != '@'))
3456                         return (SET_ERROR(EXDEV));
3457 
3458                 error = zfs_unmount_snap(name);
3459                 if (error != 0)
3460                         return (error);
3461         }
3462 
3463         return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3464 }
3465 
3466 /*
3467  * inputs:
3468  * zc_name              name of dataset to destroy
3469  * zc_objset_type       type of objset
3470  * zc_defer_destroy     mark for deferred destroy
3471  *
3472  * outputs:             none
3473  */
3474 static int
3475 zfs_ioc_destroy(zfs_cmd_t *zc)
3476 {
3477         int err;
3478 
3479         if (zc->zc_objset_type == DMU_OST_ZFS) {
3480                 err = zfs_unmount_snap(zc->zc_name);
3481                 if (err != 0)
3482                         return (err);
3483         }
3484 
3485         if (strchr(zc->zc_name, '@'))
3486                 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3487         else
3488                 err = dsl_destroy_head(zc->zc_name);
3489         if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3490                 (void) zvol_remove_minor(zc->zc_name);
3491         return (err);
3492 }
3493 
3494 /*
3495  * fsname is name of dataset to rollback (to most recent snapshot)
3496  *
3497  * innvl is not used.
3498  *
3499  * outnvl: "target" -> name of most recent snapshot
3500  * }
3501  */
3502 /* ARGSUSED */
3503 static int
3504 zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl)
3505 {
3506         zfsvfs_t *zfsvfs;
3507         int error;
3508 
3509         if (getzfsvfs(fsname, &zfsvfs) == 0) {
3510                 error = zfs_suspend_fs(zfsvfs);
3511                 if (error == 0) {
3512                         int resume_err;
3513 
3514                         error = dsl_dataset_rollback(fsname, zfsvfs, outnvl);
3515                         resume_err = zfs_resume_fs(zfsvfs, fsname);
3516                         error = error ? error : resume_err;
3517                 }
3518                 VFS_RELE(zfsvfs->z_vfs);
3519         } else {
3520                 error = dsl_dataset_rollback(fsname, NULL, outnvl);
3521         }
3522         return (error);
3523 }
3524 
3525 static int
3526 recursive_unmount(const char *fsname, void *arg)
3527 {
3528         const char *snapname = arg;
3529         char fullname[MAXNAMELEN];
3530 
3531         (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname);
3532         return (zfs_unmount_snap(fullname));
3533 }
3534 
3535 /*
3536  * inputs:
3537  * zc_name      old name of dataset
3538  * zc_value     new name of dataset
3539  * zc_cookie    recursive flag (only valid for snapshots)
3540  *
3541  * outputs:     none
3542  */
3543 static int
3544 zfs_ioc_rename(zfs_cmd_t *zc)
3545 {
3546         boolean_t recursive = zc->zc_cookie & 1;
3547         char *at;
3548 
3549         zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3550         if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3551             strchr(zc->zc_value, '%'))
3552                 return (SET_ERROR(EINVAL));
3553 
3554         at = strchr(zc->zc_name, '@');
3555         if (at != NULL) {
3556                 /* snaps must be in same fs */
3557                 int error;
3558 
3559                 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
3560                         return (SET_ERROR(EXDEV));
3561                 *at = '\0';
3562                 if (zc->zc_objset_type == DMU_OST_ZFS) {
3563                         error = dmu_objset_find(zc->zc_name,
3564                             recursive_unmount, at + 1,
3565                             recursive ? DS_FIND_CHILDREN : 0);
3566                         if (error != 0) {
3567                                 *at = '@';
3568                                 return (error);
3569                         }
3570                 }
3571                 error = dsl_dataset_rename_snapshot(zc->zc_name,
3572                     at + 1, strchr(zc->zc_value, '@') + 1, recursive);
3573                 *at = '@';
3574 
3575                 return (error);
3576         } else {
3577                 if (zc->zc_objset_type == DMU_OST_ZVOL)
3578                         (void) zvol_remove_minor(zc->zc_name);
3579                 return (dsl_dir_rename(zc->zc_name, zc->zc_value));
3580         }
3581 }
3582 
3583 static int
3584 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3585 {
3586         const char *propname = nvpair_name(pair);
3587         boolean_t issnap = (strchr(dsname, '@') != NULL);
3588         zfs_prop_t prop = zfs_name_to_prop(propname);
3589         uint64_t intval;
3590         int err;
3591 
3592         if (prop == ZPROP_INVAL) {
3593                 if (zfs_prop_user(propname)) {
3594                         if (err = zfs_secpolicy_write_perms(dsname,
3595                             ZFS_DELEG_PERM_USERPROP, cr))
3596                                 return (err);
3597                         return (0);
3598                 }
3599 
3600                 if (!issnap && zfs_prop_userquota(propname)) {
3601                         const char *perm = NULL;
3602                         const char *uq_prefix =
3603                             zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3604                         const char *gq_prefix =
3605                             zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3606 
3607                         if (strncmp(propname, uq_prefix,
3608                             strlen(uq_prefix)) == 0) {
3609                                 perm = ZFS_DELEG_PERM_USERQUOTA;
3610                         } else if (strncmp(propname, gq_prefix,
3611                             strlen(gq_prefix)) == 0) {
3612                                 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3613                         } else {
3614                                 /* USERUSED and GROUPUSED are read-only */
3615                                 return (SET_ERROR(EINVAL));
3616                         }
3617 
3618                         if (err = zfs_secpolicy_write_perms(dsname, perm, cr))
3619                                 return (err);
3620                         return (0);
3621                 }
3622 
3623                 return (SET_ERROR(EINVAL));
3624         }
3625 
3626         if (issnap)
3627                 return (SET_ERROR(EINVAL));
3628 
3629         if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3630                 /*
3631                  * dsl_prop_get_all_impl() returns properties in this
3632                  * format.
3633                  */
3634                 nvlist_t *attrs;
3635                 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3636                 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3637                     &pair) == 0);
3638         }
3639 
3640         /*
3641          * Check that this value is valid for this pool version
3642          */
3643         switch (prop) {
3644         case ZFS_PROP_COMPRESSION:
3645                 /*
3646                  * If the user specified gzip compression, make sure
3647                  * the SPA supports it. We ignore any errors here since
3648                  * we'll catch them later.
3649                  */
3650                 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3651                     nvpair_value_uint64(pair, &intval) == 0) {
3652                         if (intval >= ZIO_COMPRESS_GZIP_1 &&
3653                             intval <= ZIO_COMPRESS_GZIP_9 &&
3654                             zfs_earlier_version(dsname,
3655                             SPA_VERSION_GZIP_COMPRESSION)) {
3656                                 return (SET_ERROR(ENOTSUP));
3657                         }
3658 
3659                         if (intval == ZIO_COMPRESS_ZLE &&
3660                             zfs_earlier_version(dsname,
3661                             SPA_VERSION_ZLE_COMPRESSION))
3662                                 return (SET_ERROR(ENOTSUP));
3663 
3664                         if (intval == ZIO_COMPRESS_LZ4) {
3665                                 spa_t *spa;
3666 
3667                                 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3668                                         return (err);
3669 
3670                                 if (!spa_feature_is_enabled(spa,
3671                                     SPA_FEATURE_LZ4_COMPRESS)) {
3672                                         spa_close(spa, FTAG);
3673                                         return (SET_ERROR(ENOTSUP));
3674                                 }
3675                                 spa_close(spa, FTAG);
3676                         }
3677 
3678                         /*
3679                          * If this is a bootable dataset then
3680                          * verify that the compression algorithm
3681                          * is supported for booting. We must return
3682                          * something other than ENOTSUP since it
3683                          * implies a downrev pool version.
3684                          */
3685                         if (zfs_is_bootfs(dsname) &&
3686                             !BOOTFS_COMPRESS_VALID(intval)) {
3687                                 return (SET_ERROR(ERANGE));
3688                         }
3689                 }
3690                 break;
3691 
3692         case ZFS_PROP_COPIES:
3693                 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3694                         return (SET_ERROR(ENOTSUP));
3695                 break;
3696 
3697         case ZFS_PROP_DEDUP:
3698                 if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3699                         return (SET_ERROR(ENOTSUP));
3700                 break;
3701 
3702         case ZFS_PROP_SHARESMB:
3703                 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3704                         return (SET_ERROR(ENOTSUP));
3705                 break;
3706 
3707         case ZFS_PROP_ACLINHERIT:
3708                 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3709                     nvpair_value_uint64(pair, &intval) == 0) {
3710                         if (intval == ZFS_ACL_PASSTHROUGH_X &&
3711                             zfs_earlier_version(dsname,
3712                             SPA_VERSION_PASSTHROUGH_X))
3713                                 return (SET_ERROR(ENOTSUP));
3714                 }
3715                 break;
3716         }
3717 
3718         return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3719 }
3720 
3721 /*
3722  * Checks for a race condition to make sure we don't increment a feature flag
3723  * multiple times.
3724  */
3725 static int
3726 zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx)
3727 {
3728         spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3729         spa_feature_t *featurep = arg;
3730 
3731         if (!spa_feature_is_active(spa, *featurep))
3732                 return (0);
3733         else
3734                 return (SET_ERROR(EBUSY));
3735 }
3736 
3737 /*
3738  * The callback invoked on feature activation in the sync task caused by
3739  * zfs_prop_activate_feature.
3740  */
3741 static void
3742 zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx)
3743 {
3744         spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3745         spa_feature_t *featurep = arg;
3746 
3747         spa_feature_incr(spa, *featurep, tx);
3748 }
3749 
3750 /*
3751  * Activates a feature on a pool in response to a property setting. This
3752  * creates a new sync task which modifies the pool to reflect the feature
3753  * as being active.
3754  */
3755 static int
3756 zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature)
3757 {
3758         int err;
3759 
3760         /* EBUSY here indicates that the feature is already active */
3761         err = dsl_sync_task(spa_name(spa),
3762             zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync,
3763             &feature, 2);
3764 
3765         if (err != 0 && err != EBUSY)
3766                 return (err);
3767         else
3768                 return (0);
3769 }
3770 
3771 /*
3772  * Removes properties from the given props list that fail permission checks
3773  * needed to clear them and to restore them in case of a receive error. For each
3774  * property, make sure we have both set and inherit permissions.
3775  *
3776  * Returns the first error encountered if any permission checks fail. If the
3777  * caller provides a non-NULL errlist, it also gives the complete list of names
3778  * of all the properties that failed a permission check along with the
3779  * corresponding error numbers. The caller is responsible for freeing the
3780  * returned errlist.
3781  *
3782  * If every property checks out successfully, zero is returned and the list
3783  * pointed at by errlist is NULL.
3784  */
3785 static int
3786 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
3787 {
3788         zfs_cmd_t *zc;
3789         nvpair_t *pair, *next_pair;
3790         nvlist_t *errors;
3791         int err, rv = 0;
3792 
3793         if (props == NULL)
3794                 return (0);
3795 
3796         VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3797 
3798         zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
3799         (void) strcpy(zc->zc_name, dataset);
3800         pair = nvlist_next_nvpair(props, NULL);
3801         while (pair != NULL) {
3802                 next_pair = nvlist_next_nvpair(props, pair);
3803 
3804                 (void) strcpy(zc->zc_value, nvpair_name(pair));
3805                 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
3806                     (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
3807                         VERIFY(nvlist_remove_nvpair(props, pair) == 0);
3808                         VERIFY(nvlist_add_int32(errors,
3809                             zc->zc_value, err) == 0);
3810                 }
3811                 pair = next_pair;
3812         }
3813         kmem_free(zc, sizeof (zfs_cmd_t));
3814 
3815         if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
3816                 nvlist_free(errors);
3817                 errors = NULL;
3818         } else {
3819                 VERIFY(nvpair_value_int32(pair, &rv) == 0);
3820         }
3821 
3822         if (errlist == NULL)
3823                 nvlist_free(errors);
3824         else
3825                 *errlist = errors;
3826 
3827         return (rv);
3828 }
3829 
3830 static boolean_t
3831 propval_equals(nvpair_t *p1, nvpair_t *p2)
3832 {
3833         if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
3834                 /* dsl_prop_get_all_impl() format */
3835                 nvlist_t *attrs;
3836                 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
3837                 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3838                     &p1) == 0);
3839         }
3840 
3841         if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
3842                 nvlist_t *attrs;
3843                 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
3844                 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3845                     &p2) == 0);
3846         }
3847 
3848         if (nvpair_type(p1) != nvpair_type(p2))
3849                 return (B_FALSE);
3850 
3851         if (nvpair_type(p1) == DATA_TYPE_STRING) {
3852                 char *valstr1, *valstr2;
3853 
3854                 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
3855                 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
3856                 return (strcmp(valstr1, valstr2) == 0);
3857         } else {
3858                 uint64_t intval1, intval2;
3859 
3860                 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
3861                 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
3862                 return (intval1 == intval2);
3863         }
3864 }
3865 
3866 /*
3867  * Remove properties from props if they are not going to change (as determined
3868  * by comparison with origprops). Remove them from origprops as well, since we
3869  * do not need to clear or restore properties that won't change.
3870  */
3871 static void
3872 props_reduce(nvlist_t *props, nvlist_t *origprops)
3873 {
3874         nvpair_t *pair, *next_pair;
3875 
3876         if (origprops == NULL)
3877                 return; /* all props need to be received */
3878 
3879         pair = nvlist_next_nvpair(props, NULL);
3880         while (pair != NULL) {
3881                 const char *propname = nvpair_name(pair);
3882                 nvpair_t *match;
3883 
3884                 next_pair = nvlist_next_nvpair(props, pair);
3885 
3886                 if ((nvlist_lookup_nvpair(origprops, propname,
3887                     &match) != 0) || !propval_equals(pair, match))
3888                         goto next; /* need to set received value */
3889 
3890                 /* don't clear the existing received value */
3891                 (void) nvlist_remove_nvpair(origprops, match);
3892                 /* don't bother receiving the property */
3893                 (void) nvlist_remove_nvpair(props, pair);
3894 next:
3895                 pair = next_pair;
3896         }
3897 }
3898 
3899 #ifdef  DEBUG
3900 static boolean_t zfs_ioc_recv_inject_err;
3901 #endif
3902 
3903 /*
3904  * inputs:
3905  * zc_name              name of containing filesystem
3906  * zc_nvlist_src{_size} nvlist of properties to apply
3907  * zc_value             name of snapshot to create
3908  * zc_string            name of clone origin (if DRR_FLAG_CLONE)
3909  * zc_cookie            file descriptor to recv from
3910  * zc_begin_record      the BEGIN record of the stream (not byteswapped)
3911  * zc_guid              force flag
3912  * zc_cleanup_fd        cleanup-on-exit file descriptor
3913  * zc_action_handle     handle for this guid/ds mapping (or zero on first call)
3914  *
3915  * outputs:
3916  * zc_cookie            number of bytes read
3917  * zc_nvlist_dst{_size} error for each unapplied received property
3918  * zc_obj               zprop_errflags_t
3919  * zc_action_handle     handle for this guid/ds mapping
3920  */
3921 static int
3922 zfs_ioc_recv(zfs_cmd_t *zc)
3923 {
3924         file_t *fp;
3925         dmu_recv_cookie_t drc;
3926         boolean_t force = (boolean_t)zc->zc_guid;
3927         int fd;
3928         int error = 0;
3929         int props_error = 0;
3930         nvlist_t *errors;
3931         offset_t off;
3932         nvlist_t *props = NULL; /* sent properties */
3933         nvlist_t *origprops = NULL; /* existing properties */
3934         char *origin = NULL;
3935         char *tosnap;
3936         char tofs[ZFS_MAXNAMELEN];
3937         boolean_t first_recvd_props = B_FALSE;
3938 
3939         if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3940             strchr(zc->zc_value, '@') == NULL ||
3941             strchr(zc->zc_value, '%'))
3942                 return (SET_ERROR(EINVAL));
3943 
3944         (void) strcpy(tofs, zc->zc_value);
3945         tosnap = strchr(tofs, '@');
3946         *tosnap++ = '\0';
3947 
3948         if (zc->zc_nvlist_src != NULL &&
3949             (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
3950             zc->zc_iflags, &props)) != 0)
3951                 return (error);
3952 
3953         fd = zc->zc_cookie;
3954         fp = getf(fd);
3955         if (fp == NULL) {
3956                 nvlist_free(props);
3957                 return (SET_ERROR(EBADF));
3958         }
3959 
3960         VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3961 
3962         if (zc->zc_string[0])
3963                 origin = zc->zc_string;
3964 
3965         error = dmu_recv_begin(tofs, tosnap,
3966             &zc->zc_begin_record, force, origin, &drc);
3967         if (error != 0)
3968                 goto out;
3969 
3970         /*
3971          * Set properties before we receive the stream so that they are applied
3972          * to the new data. Note that we must call dmu_recv_stream() if
3973          * dmu_recv_begin() succeeds.
3974          */
3975         if (props != NULL && !drc.drc_newfs) {
3976                 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
3977                     SPA_VERSION_RECVD_PROPS &&
3978                     !dsl_prop_get_hasrecvd(tofs))
3979                         first_recvd_props = B_TRUE;
3980 
3981                 /*
3982                  * If new received properties are supplied, they are to
3983                  * completely replace the existing received properties, so stash
3984                  * away the existing ones.
3985                  */
3986                 if (dsl_prop_get_received(tofs, &origprops) == 0) {
3987                         nvlist_t *errlist = NULL;
3988                         /*
3989                          * Don't bother writing a property if its value won't
3990                          * change (and avoid the unnecessary security checks).
3991                          *
3992                          * The first receive after SPA_VERSION_RECVD_PROPS is a
3993                          * special case where we blow away all local properties
3994                          * regardless.
3995                          */
3996                         if (!first_recvd_props)
3997                                 props_reduce(props, origprops);
3998                         if (zfs_check_clearable(tofs, origprops, &errlist) != 0)
3999                                 (void) nvlist_merge(errors, errlist, 0);
4000                         nvlist_free(errlist);
4001 
4002                         if (clear_received_props(tofs, origprops,
4003                             first_recvd_props ? NULL : props) != 0)
4004                                 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4005                 } else {
4006                         zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4007                 }
4008         }
4009 
4010         if (props != NULL) {
4011                 props_error = dsl_prop_set_hasrecvd(tofs);
4012 
4013                 if (props_error == 0) {
4014                         (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4015                             props, errors);
4016                 }
4017         }
4018 
4019         if (zc->zc_nvlist_dst_size != 0 &&
4020             (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4021             put_nvlist(zc, errors) != 0)) {
4022                 /*
4023                  * Caller made zc->zc_nvlist_dst less than the minimum expected
4024                  * size or supplied an invalid address.
4025                  */
4026                 props_error = SET_ERROR(EINVAL);
4027         }
4028 
4029         off = fp->f_offset;
4030         error = dmu_recv_stream(&drc, fp->f_vnode, &off, zc->zc_cleanup_fd,
4031             &zc->zc_action_handle);
4032 
4033         if (error == 0) {
4034                 zfsvfs_t *zfsvfs = NULL;
4035 
4036                 if (getzfsvfs(tofs, &zfsvfs) == 0) {
4037                         /* online recv */
4038                         int end_err;
4039 
4040                         error = zfs_suspend_fs(zfsvfs);
4041                         /*
4042                          * If the suspend fails, then the recv_end will
4043                          * likely also fail, and clean up after itself.
4044                          */
4045                         end_err = dmu_recv_end(&drc, zfsvfs);
4046                         if (error == 0)
4047                                 error = zfs_resume_fs(zfsvfs, tofs);
4048                         error = error ? error : end_err;
4049                         VFS_RELE(zfsvfs->z_vfs);
4050                 } else {
4051                         error = dmu_recv_end(&drc, NULL);
4052                 }
4053         }
4054 
4055         zc->zc_cookie = off - fp->f_offset;
4056         if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4057                 fp->f_offset = off;
4058 
4059 #ifdef  DEBUG
4060         if (zfs_ioc_recv_inject_err) {
4061                 zfs_ioc_recv_inject_err = B_FALSE;
4062                 error = 1;
4063         }
4064 #endif
4065         /*
4066          * On error, restore the original props.
4067          */
4068         if (error != 0 && props != NULL && !drc.drc_newfs) {
4069                 if (clear_received_props(tofs, props, NULL) != 0) {
4070                         /*
4071                          * We failed to clear the received properties.
4072                          * Since we may have left a $recvd value on the
4073                          * system, we can't clear the $hasrecvd flag.
4074                          */
4075                         zc->zc_obj |= ZPROP_ERR_NORESTORE;
4076                 } else if (first_recvd_props) {
4077                         dsl_prop_unset_hasrecvd(tofs);
4078                 }
4079 
4080                 if (origprops == NULL && !drc.drc_newfs) {
4081                         /* We failed to stash the original properties. */
4082                         zc->zc_obj |= ZPROP_ERR_NORESTORE;
4083                 }
4084 
4085                 /*
4086                  * dsl_props_set() will not convert RECEIVED to LOCAL on or
4087                  * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4088                  * explictly if we're restoring local properties cleared in the
4089                  * first new-style receive.
4090                  */
4091                 if (origprops != NULL &&
4092                     zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4093                     ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4094                     origprops, NULL) != 0) {
4095                         /*
4096                          * We stashed the original properties but failed to
4097                          * restore them.
4098                          */
4099                         zc->zc_obj |= ZPROP_ERR_NORESTORE;
4100                 }
4101         }
4102 out:
4103         nvlist_free(props);
4104         nvlist_free(origprops);
4105         nvlist_free(errors);
4106         releasef(fd);
4107 
4108         if (error == 0)
4109                 error = props_error;
4110 
4111         return (error);
4112 }
4113 
4114 /*
4115  * inputs:
4116  * zc_name      name of snapshot to send
4117  * zc_cookie    file descriptor to send stream to
4118  * zc_obj       fromorigin flag (mutually exclusive with zc_fromobj)
4119  * zc_sendobj   objsetid of snapshot to send
4120  * zc_fromobj   objsetid of incremental fromsnap (may be zero)
4121  * zc_guid      if set, estimate size of stream only.  zc_cookie is ignored.
4122  *              output size in zc_objset_type.
4123  *
4124  * outputs: none
4125  */
4126 static int
4127 zfs_ioc_send(zfs_cmd_t *zc)
4128 {
4129         int error;
4130         offset_t off;
4131         boolean_t estimate = (zc->zc_guid != 0);
4132 
4133         if (zc->zc_obj != 0) {
4134                 dsl_pool_t *dp;
4135                 dsl_dataset_t *tosnap;
4136 
4137                 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4138                 if (error != 0)
4139                         return (error);
4140 
4141                 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4142                 if (error != 0) {
4143                         dsl_pool_rele(dp, FTAG);
4144                         return (error);
4145                 }
4146 
4147                 if (dsl_dir_is_clone(tosnap->ds_dir))
4148                         zc->zc_fromobj = tosnap->ds_dir->dd_phys->dd_origin_obj;
4149                 dsl_dataset_rele(tosnap, FTAG);
4150                 dsl_pool_rele(dp, FTAG);
4151         }
4152 
4153         if (estimate) {
4154                 dsl_pool_t *dp;
4155                 dsl_dataset_t *tosnap;
4156                 dsl_dataset_t *fromsnap = NULL;
4157 
4158                 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4159                 if (error != 0)
4160                         return (error);
4161 
4162                 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4163                 if (error != 0) {
4164                         dsl_pool_rele(dp, FTAG);
4165                         return (error);
4166                 }
4167 
4168                 if (zc->zc_fromobj != 0) {
4169                         error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
4170                             FTAG, &fromsnap);
4171                         if (error != 0) {
4172                                 dsl_dataset_rele(tosnap, FTAG);
4173                                 dsl_pool_rele(dp, FTAG);
4174                                 return (error);
4175                         }
4176                 }
4177 
4178                 error = dmu_send_estimate(tosnap, fromsnap,
4179                     &zc->zc_objset_type);
4180 
4181                 if (fromsnap != NULL)
4182                         dsl_dataset_rele(fromsnap, FTAG);
4183                 dsl_dataset_rele(tosnap, FTAG);
4184                 dsl_pool_rele(dp, FTAG);
4185         } else {
4186                 file_t *fp = getf(zc->zc_cookie);
4187                 if (fp == NULL)
4188                         return (SET_ERROR(EBADF));
4189 
4190                 off = fp->f_offset;
4191                 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
4192                     zc->zc_fromobj, zc->zc_cookie, fp->f_vnode, &off);
4193 
4194                 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4195                         fp->f_offset = off;
4196                 releasef(zc->zc_cookie);
4197         }
4198         return (error);
4199 }
4200 
4201 /*
4202  * inputs:
4203  * zc_name      name of snapshot on which to report progress
4204  * zc_cookie    file descriptor of send stream
4205  *
4206  * outputs:
4207  * zc_cookie    number of bytes written in send stream thus far
4208  */
4209 static int
4210 zfs_ioc_send_progress(zfs_cmd_t *zc)
4211 {
4212         dsl_pool_t *dp;
4213         dsl_dataset_t *ds;
4214         dmu_sendarg_t *dsp = NULL;
4215         int error;
4216 
4217         error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4218         if (error != 0)
4219                 return (error);
4220 
4221         error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4222         if (error != 0) {
4223                 dsl_pool_rele(dp, FTAG);
4224                 return (error);
4225         }
4226 
4227         mutex_enter(&ds->ds_sendstream_lock);
4228 
4229         /*
4230          * Iterate over all the send streams currently active on this dataset.
4231          * If there's one which matches the specified file descriptor _and_ the
4232          * stream was started by the current process, return the progress of
4233          * that stream.
4234          */
4235         for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4236             dsp = list_next(&ds->ds_sendstreams, dsp)) {
4237                 if (dsp->dsa_outfd == zc->zc_cookie &&
4238                     dsp->dsa_proc == curproc)
4239                         break;
4240         }
4241 
4242         if (dsp != NULL)
4243                 zc->zc_cookie = *(dsp->dsa_off);
4244         else
4245                 error = SET_ERROR(ENOENT);
4246 
4247         mutex_exit(&ds->ds_sendstream_lock);
4248         dsl_dataset_rele(ds, FTAG);
4249         dsl_pool_rele(dp, FTAG);
4250         return (error);
4251 }
4252 
4253 static int
4254 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4255 {
4256         int id, error;
4257 
4258         error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4259             &zc->zc_inject_record);
4260 
4261         if (error == 0)
4262                 zc->zc_guid = (uint64_t)id;
4263 
4264         return (error);
4265 }
4266 
4267 static int
4268 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4269 {
4270         return (zio_clear_fault((int)zc->zc_guid));
4271 }
4272 
4273 static int
4274 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4275 {
4276         int id = (int)zc->zc_guid;
4277         int error;
4278 
4279         error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4280             &zc->zc_inject_record);
4281 
4282         zc->zc_guid = id;
4283 
4284         return (error);
4285 }
4286 
4287 static int
4288 zfs_ioc_error_log(zfs_cmd_t *zc)
4289 {
4290         spa_t *spa;
4291         int error;
4292         size_t count = (size_t)zc->zc_nvlist_dst_size;
4293 
4294         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4295                 return (error);
4296 
4297         error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4298             &count);
4299         if (error == 0)
4300                 zc->zc_nvlist_dst_size = count;
4301         else
4302                 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4303 
4304         spa_close(spa, FTAG);
4305 
4306         return (error);
4307 }
4308 
4309 static int
4310 zfs_ioc_clear(zfs_cmd_t *zc)
4311 {
4312         spa_t *spa;
4313         vdev_t *vd;
4314         int error;
4315 
4316         /*
4317          * On zpool clear we also fix up missing slogs
4318          */
4319         mutex_enter(&spa_namespace_lock);
4320         spa = spa_lookup(zc->zc_name);
4321         if (spa == NULL) {
4322                 mutex_exit(&spa_namespace_lock);
4323                 return (SET_ERROR(EIO));
4324         }
4325         if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4326                 /* we need to let spa_open/spa_load clear the chains */
4327                 spa_set_log_state(spa, SPA_LOG_CLEAR);
4328         }
4329         spa->spa_last_open_failed = 0;
4330         mutex_exit(&spa_namespace_lock);
4331 
4332         if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4333                 error = spa_open(zc->zc_name, &spa, FTAG);
4334         } else {
4335                 nvlist_t *policy;
4336                 nvlist_t *config = NULL;
4337 
4338                 if (zc->zc_nvlist_src == NULL)
4339                         return (SET_ERROR(EINVAL));
4340 
4341                 if ((error = get_nvlist(zc->zc_nvlist_src,
4342                     zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4343                         error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4344                             policy, &config);
4345                         if (config != NULL) {
4346                                 int err;
4347 
4348                                 if ((err = put_nvlist(zc, config)) != 0)
4349                                         error = err;
4350                                 nvlist_free(config);
4351                         }
4352                         nvlist_free(policy);
4353                 }
4354         }
4355 
4356         if (error != 0)
4357                 return (error);
4358 
4359         spa_vdev_state_enter(spa, SCL_NONE);
4360 
4361         if (zc->zc_guid == 0) {
4362                 vd = NULL;
4363         } else {
4364                 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4365                 if (vd == NULL) {
4366                         (void) spa_vdev_state_exit(spa, NULL, ENODEV);
4367                         spa_close(spa, FTAG);
4368                         return (SET_ERROR(ENODEV));
4369                 }
4370         }
4371 
4372         vdev_clear(spa, vd);
4373 
4374         (void) spa_vdev_state_exit(spa, NULL, 0);
4375 
4376         /*
4377          * Resume any suspended I/Os.
4378          */
4379         if (zio_resume(spa) != 0)
4380                 error = SET_ERROR(EIO);
4381 
4382         spa_close(spa, FTAG);
4383 
4384         return (error);
4385 }
4386 
4387 static int
4388 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4389 {
4390         spa_t *spa;
4391         int error;
4392 
4393         error = spa_open(zc->zc_name, &spa, FTAG);
4394         if (error != 0)
4395                 return (error);
4396 
4397         spa_vdev_state_enter(spa, SCL_NONE);
4398 
4399         /*
4400          * If a resilver is already in progress then set the
4401          * spa_scrub_reopen flag to B_TRUE so that we don't restart
4402          * the scan as a side effect of the reopen. Otherwise, let
4403          * vdev_open() decided if a resilver is required.
4404          */
4405         spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4406         vdev_reopen(spa->spa_root_vdev);
4407         spa->spa_scrub_reopen = B_FALSE;
4408 
4409         (void) spa_vdev_state_exit(spa, NULL, 0);
4410         spa_close(spa, FTAG);
4411         return (0);
4412 }
4413 /*
4414  * inputs:
4415  * zc_name      name of filesystem
4416  * zc_value     name of origin snapshot
4417  *
4418  * outputs:
4419  * zc_string    name of conflicting snapshot, if there is one
4420  */
4421 static int
4422 zfs_ioc_promote(zfs_cmd_t *zc)
4423 {
4424         char *cp;
4425 
4426         /*
4427          * We don't need to unmount *all* the origin fs's snapshots, but
4428          * it's easier.
4429          */
4430         cp = strchr(zc->zc_value, '@');
4431         if (cp)
4432                 *cp = '\0';
4433         (void) dmu_objset_find(zc->zc_value,
4434             zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
4435         return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4436 }
4437 
4438 /*
4439  * Retrieve a single {user|group}{used|quota}@... property.
4440  *
4441  * inputs:
4442  * zc_name      name of filesystem
4443  * zc_objset_type zfs_userquota_prop_t
4444  * zc_value     domain name (eg. "S-1-234-567-89")
4445  * zc_guid      RID/UID/GID
4446  *
4447  * outputs:
4448  * zc_cookie    property value
4449  */
4450 static int
4451 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4452 {
4453         zfsvfs_t *zfsvfs;
4454         int error;
4455 
4456         if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4457                 return (SET_ERROR(EINVAL));
4458 
4459         error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4460         if (error != 0)
4461                 return (error);
4462 
4463         error = zfs_userspace_one(zfsvfs,
4464             zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4465         zfsvfs_rele(zfsvfs, FTAG);
4466 
4467         return (error);
4468 }
4469 
4470 /*
4471  * inputs:
4472  * zc_name              name of filesystem
4473  * zc_cookie            zap cursor
4474  * zc_objset_type       zfs_userquota_prop_t
4475  * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4476  *
4477  * outputs:
4478  * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
4479  * zc_cookie    zap cursor
4480  */
4481 static int
4482 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4483 {
4484         zfsvfs_t *zfsvfs;
4485         int bufsize = zc->zc_nvlist_dst_size;
4486 
4487         if (bufsize <= 0)
4488                 return (SET_ERROR(ENOMEM));
4489 
4490         int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4491         if (error != 0)
4492                 return (error);
4493 
4494         void *buf = kmem_alloc(bufsize, KM_SLEEP);
4495 
4496         error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
4497             buf, &zc->zc_nvlist_dst_size);
4498 
4499         if (error == 0) {
4500                 error = xcopyout(buf,
4501                     (void *)(uintptr_t)zc->zc_nvlist_dst,
4502                     zc->zc_nvlist_dst_size);
4503         }
4504         kmem_free(buf, bufsize);
4505         zfsvfs_rele(zfsvfs, FTAG);
4506 
4507         return (error);
4508 }
4509 
4510 /*
4511  * inputs:
4512  * zc_name              name of filesystem
4513  *
4514  * outputs:
4515  * none
4516  */
4517 static int
4518 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4519 {
4520         objset_t *os;
4521         int error = 0;
4522         zfsvfs_t *zfsvfs;
4523 
4524         if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
4525                 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
4526                         /*
4527                          * If userused is not enabled, it may be because the
4528                          * objset needs to be closed & reopened (to grow the
4529                          * objset_phys_t).  Suspend/resume the fs will do that.
4530                          */
4531                         error = zfs_suspend_fs(zfsvfs);
4532                         if (error == 0) {
4533                                 dmu_objset_refresh_ownership(zfsvfs->z_os,
4534                                     zfsvfs);
4535                                 error = zfs_resume_fs(zfsvfs, zc->zc_name);
4536                         }
4537                 }
4538                 if (error == 0)
4539                         error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
4540                 VFS_RELE(zfsvfs->z_vfs);
4541         } else {
4542                 /* XXX kind of reading contents without owning */
4543                 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4544                 if (error != 0)
4545                         return (error);
4546 
4547                 error = dmu_objset_userspace_upgrade(os);
4548                 dmu_objset_rele(os, FTAG);
4549         }
4550 
4551         return (error);
4552 }
4553 
4554 /*
4555  * We don't want to have a hard dependency
4556  * against some special symbols in sharefs
4557  * nfs, and smbsrv.  Determine them if needed when
4558  * the first file system is shared.
4559  * Neither sharefs, nfs or smbsrv are unloadable modules.
4560  */
4561 int (*znfsexport_fs)(void *arg);
4562 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
4563 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
4564 
4565 int zfs_nfsshare_inited;
4566 int zfs_smbshare_inited;
4567 
4568 ddi_modhandle_t nfs_mod;
4569 ddi_modhandle_t sharefs_mod;
4570 ddi_modhandle_t smbsrv_mod;
4571 kmutex_t zfs_share_lock;
4572 
4573 static int
4574 zfs_init_sharefs()
4575 {
4576         int error;
4577 
4578         ASSERT(MUTEX_HELD(&zfs_share_lock));
4579         /* Both NFS and SMB shares also require sharetab support. */
4580         if (sharefs_mod == NULL && ((sharefs_mod =
4581             ddi_modopen("fs/sharefs",
4582             KRTLD_MODE_FIRST, &error)) == NULL)) {
4583                 return (SET_ERROR(ENOSYS));
4584         }
4585         if (zshare_fs == NULL && ((zshare_fs =
4586             (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
4587             ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
4588                 return (SET_ERROR(ENOSYS));
4589         }
4590         return (0);
4591 }
4592 
4593 static int
4594 zfs_ioc_share(zfs_cmd_t *zc)
4595 {
4596         int error;
4597         int opcode;
4598 
4599         switch (zc->zc_share.z_sharetype) {
4600         case ZFS_SHARE_NFS:
4601         case ZFS_UNSHARE_NFS:
4602                 if (zfs_nfsshare_inited == 0) {
4603                         mutex_enter(&zfs_share_lock);
4604                         if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
4605                             KRTLD_MODE_FIRST, &error)) == NULL)) {
4606                                 mutex_exit(&zfs_share_lock);
4607                                 return (SET_ERROR(ENOSYS));
4608                         }
4609                         if (znfsexport_fs == NULL &&
4610                             ((znfsexport_fs = (int (*)(void *))
4611                             ddi_modsym(nfs_mod,
4612                             "nfs_export", &error)) == NULL)) {
4613                                 mutex_exit(&zfs_share_lock);
4614                                 return (SET_ERROR(ENOSYS));
4615                         }
4616                         error = zfs_init_sharefs();
4617                         if (error != 0) {
4618                                 mutex_exit(&zfs_share_lock);
4619                                 return (SET_ERROR(ENOSYS));
4620                         }
4621                         zfs_nfsshare_inited = 1;
4622                         mutex_exit(&zfs_share_lock);
4623                 }
4624                 break;
4625         case ZFS_SHARE_SMB:
4626         case ZFS_UNSHARE_SMB:
4627                 if (zfs_smbshare_inited == 0) {
4628                         mutex_enter(&zfs_share_lock);
4629                         if (smbsrv_mod == NULL && ((smbsrv_mod =
4630                             ddi_modopen("drv/smbsrv",
4631                             KRTLD_MODE_FIRST, &error)) == NULL)) {
4632                                 mutex_exit(&zfs_share_lock);
4633                                 return (SET_ERROR(ENOSYS));
4634                         }
4635                         if (zsmbexport_fs == NULL && ((zsmbexport_fs =
4636                             (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
4637                             "smb_server_share", &error)) == NULL)) {
4638                                 mutex_exit(&zfs_share_lock);
4639                                 return (SET_ERROR(ENOSYS));
4640                         }
4641                         error = zfs_init_sharefs();
4642                         if (error != 0) {
4643                                 mutex_exit(&zfs_share_lock);
4644                                 return (SET_ERROR(ENOSYS));
4645                         }
4646                         zfs_smbshare_inited = 1;
4647                         mutex_exit(&zfs_share_lock);
4648                 }
4649                 break;
4650         default:
4651                 return (SET_ERROR(EINVAL));
4652         }
4653 
4654         switch (zc->zc_share.z_sharetype) {
4655         case ZFS_SHARE_NFS:
4656         case ZFS_UNSHARE_NFS:
4657                 if (error =
4658                     znfsexport_fs((void *)
4659                     (uintptr_t)zc->zc_share.z_exportdata))
4660                         return (error);
4661                 break;
4662         case ZFS_SHARE_SMB:
4663         case ZFS_UNSHARE_SMB:
4664                 if (error = zsmbexport_fs((void *)
4665                     (uintptr_t)zc->zc_share.z_exportdata,
4666                     zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
4667                     B_TRUE: B_FALSE)) {
4668                         return (error);
4669                 }
4670                 break;
4671         }
4672 
4673         opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
4674             zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
4675             SHAREFS_ADD : SHAREFS_REMOVE;
4676 
4677         /*
4678          * Add or remove share from sharetab
4679          */
4680         error = zshare_fs(opcode,
4681             (void *)(uintptr_t)zc->zc_share.z_sharedata,
4682             zc->zc_share.z_sharemax);
4683 
4684         return (error);
4685 
4686 }
4687 
4688 ace_t full_access[] = {
4689         {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4690 };
4691 
4692 /*
4693  * inputs:
4694  * zc_name              name of containing filesystem
4695  * zc_obj               object # beyond which we want next in-use object #
4696  *
4697  * outputs:
4698  * zc_obj               next in-use object #
4699  */
4700 static int
4701 zfs_ioc_next_obj(zfs_cmd_t *zc)
4702 {
4703         objset_t *os = NULL;
4704         int error;
4705 
4706         error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4707         if (error != 0)
4708                 return (error);
4709 
4710         error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
4711             os->os_dsl_dataset->ds_phys->ds_prev_snap_txg);
4712 
4713         dmu_objset_rele(os, FTAG);
4714         return (error);
4715 }
4716 
4717 /*
4718  * inputs:
4719  * zc_name              name of filesystem
4720  * zc_value             prefix name for snapshot
4721  * zc_cleanup_fd        cleanup-on-exit file descriptor for calling process
4722  *
4723  * outputs:
4724  * zc_value             short name of new snapshot
4725  */
4726 static int
4727 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
4728 {
4729         char *snap_name;
4730         char *hold_name;
4731         int error;
4732         minor_t minor;
4733 
4734         error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
4735         if (error != 0)
4736                 return (error);
4737 
4738         snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
4739             (u_longlong_t)ddi_get_lbolt64());
4740         hold_name = kmem_asprintf("%%%s", zc->zc_value);
4741 
4742         error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
4743             hold_name);
4744         if (error == 0)
4745                 (void) strcpy(zc->zc_value, snap_name);
4746         strfree(snap_name);
4747         strfree(hold_name);
4748         zfs_onexit_fd_rele(zc->zc_cleanup_fd);
4749         return (error);
4750 }
4751 
4752 /*
4753  * inputs:
4754  * zc_name              name of "to" snapshot
4755  * zc_value             name of "from" snapshot
4756  * zc_cookie            file descriptor to write diff data on
4757  *
4758  * outputs:
4759  * dmu_diff_record_t's to the file descriptor
4760  */
4761 static int
4762 zfs_ioc_diff(zfs_cmd_t *zc)
4763 {
4764         file_t *fp;
4765         offset_t off;
4766         int error;
4767 
4768         fp = getf(zc->zc_cookie);
4769         if (fp == NULL)
4770                 return (SET_ERROR(EBADF));
4771 
4772         off = fp->f_offset;
4773 
4774         error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off);
4775 
4776         if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4777                 fp->f_offset = off;
4778         releasef(zc->zc_cookie);
4779 
4780         return (error);
4781 }
4782 
4783 /*
4784  * Remove all ACL files in shares dir
4785  */
4786 static int
4787 zfs_smb_acl_purge(znode_t *dzp)
4788 {
4789         zap_cursor_t    zc;
4790         zap_attribute_t zap;
4791         zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
4792         int error;
4793 
4794         for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
4795             (error = zap_cursor_retrieve(&zc, &zap)) == 0;
4796             zap_cursor_advance(&zc)) {
4797                 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
4798                     NULL, 0)) != 0)
4799                         break;
4800         }
4801         zap_cursor_fini(&zc);
4802         return (error);
4803 }
4804 
4805 static int
4806 zfs_ioc_smb_acl(zfs_cmd_t *zc)
4807 {
4808         vnode_t *vp;
4809         znode_t *dzp;
4810         vnode_t *resourcevp = NULL;
4811         znode_t *sharedir;
4812         zfsvfs_t *zfsvfs;
4813         nvlist_t *nvlist;
4814         char *src, *target;
4815         vattr_t vattr;
4816         vsecattr_t vsec;
4817         int error = 0;
4818 
4819         if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
4820             NO_FOLLOW, NULL, &vp)) != 0)
4821                 return (error);
4822 
4823         /* Now make sure mntpnt and dataset are ZFS */
4824 
4825         if (vp->v_vfsp->vfs_fstype != zfsfstype ||
4826             (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
4827             zc->zc_name) != 0)) {
4828                 VN_RELE(vp);
4829                 return (SET_ERROR(EINVAL));
4830         }
4831 
4832         dzp = VTOZ(vp);
4833         zfsvfs = dzp->z_zfsvfs;
4834         ZFS_ENTER(zfsvfs);
4835 
4836         /*
4837          * Create share dir if its missing.
4838          */
4839         mutex_enter(&zfsvfs->z_lock);
4840         if (zfsvfs->z_shares_dir == 0) {
4841                 dmu_tx_t *tx;
4842 
4843                 tx = dmu_tx_create(zfsvfs->z_os);
4844                 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
4845                     ZFS_SHARES_DIR);
4846                 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
4847                 error = dmu_tx_assign(tx, TXG_WAIT);
4848                 if (error != 0) {
4849                         dmu_tx_abort(tx);
4850                 } else {
4851                         error = zfs_create_share_dir(zfsvfs, tx);
4852                         dmu_tx_commit(tx);
4853                 }
4854                 if (error != 0) {
4855                         mutex_exit(&zfsvfs->z_lock);
4856                         VN_RELE(vp);
4857                         ZFS_EXIT(zfsvfs);
4858                         return (error);
4859                 }
4860         }
4861         mutex_exit(&zfsvfs->z_lock);
4862 
4863         ASSERT(zfsvfs->z_shares_dir);
4864         if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
4865                 VN_RELE(vp);
4866                 ZFS_EXIT(zfsvfs);
4867                 return (error);
4868         }
4869 
4870         switch (zc->zc_cookie) {
4871         case ZFS_SMB_ACL_ADD:
4872                 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
4873                 vattr.va_type = VREG;
4874                 vattr.va_mode = S_IFREG|0777;
4875                 vattr.va_uid = 0;
4876                 vattr.va_gid = 0;
4877 
4878                 vsec.vsa_mask = VSA_ACE;
4879                 vsec.vsa_aclentp = &full_access;
4880                 vsec.vsa_aclentsz = sizeof (full_access);
4881                 vsec.vsa_aclcnt = 1;
4882 
4883                 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
4884                     &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
4885                 if (resourcevp)
4886                         VN_RELE(resourcevp);
4887                 break;
4888 
4889         case ZFS_SMB_ACL_REMOVE:
4890                 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
4891                     NULL, 0);
4892                 break;
4893 
4894         case ZFS_SMB_ACL_RENAME:
4895                 if ((error = get_nvlist(zc->zc_nvlist_src,
4896                     zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
4897                         VN_RELE(vp);
4898                         ZFS_EXIT(zfsvfs);
4899                         return (error);
4900                 }
4901                 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
4902                     nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
4903                     &target)) {
4904                         VN_RELE(vp);
4905                         VN_RELE(ZTOV(sharedir));
4906                         ZFS_EXIT(zfsvfs);
4907                         nvlist_free(nvlist);
4908                         return (error);
4909                 }
4910                 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
4911                     kcred, NULL, 0);
4912                 nvlist_free(nvlist);
4913                 break;
4914 
4915         case ZFS_SMB_ACL_PURGE:
4916                 error = zfs_smb_acl_purge(sharedir);
4917                 break;
4918 
4919         default:
4920                 error = SET_ERROR(EINVAL);
4921                 break;
4922         }
4923 
4924         VN_RELE(vp);
4925         VN_RELE(ZTOV(sharedir));
4926 
4927         ZFS_EXIT(zfsvfs);
4928 
4929         return (error);
4930 }
4931 
4932 /*
4933  * innvl: {
4934  *     "holds" -> { snapname -> holdname (string), ... }
4935  *     (optional) "cleanup_fd" -> fd (int32)
4936  * }
4937  *
4938  * outnvl: {
4939  *     snapname -> error value (int32)
4940  *     ...
4941  * }
4942  */
4943 /* ARGSUSED */
4944 static int
4945 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
4946 {
4947         nvlist_t *holds;
4948         int cleanup_fd = -1;
4949         int error;
4950         minor_t minor = 0;
4951 
4952         error = nvlist_lookup_nvlist(args, "holds", &holds);
4953         if (error != 0)
4954                 return (SET_ERROR(EINVAL));
4955 
4956         if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
4957                 error = zfs_onexit_fd_hold(cleanup_fd, &minor);
4958                 if (error != 0)
4959                         return (error);
4960         }
4961 
4962         error = dsl_dataset_user_hold(holds, minor, errlist);
4963         if (minor != 0)
4964                 zfs_onexit_fd_rele(cleanup_fd);
4965         return (error);
4966 }
4967 
4968 /*
4969  * innvl is not used.
4970  *
4971  * outnvl: {
4972  *    holdname -> time added (uint64 seconds since epoch)
4973  *    ...
4974  * }
4975  */
4976 /* ARGSUSED */
4977 static int
4978 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
4979 {
4980         return (dsl_dataset_get_holds(snapname, outnvl));
4981 }
4982 
4983 /*
4984  * innvl: {
4985  *     snapname -> { holdname, ... }
4986  *     ...
4987  * }
4988  *
4989  * outnvl: {
4990  *     snapname -> error value (int32)
4991  *     ...
4992  * }
4993  */
4994 /* ARGSUSED */
4995 static int
4996 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
4997 {
4998         return (dsl_dataset_user_release(holds, errlist));
4999 }
5000 
5001 /*
5002  * inputs:
5003  * zc_name              name of new filesystem or snapshot
5004  * zc_value             full name of old snapshot
5005  *
5006  * outputs:
5007  * zc_cookie            space in bytes
5008  * zc_objset_type       compressed space in bytes
5009  * zc_perm_action       uncompressed space in bytes
5010  */
5011 static int
5012 zfs_ioc_space_written(zfs_cmd_t *zc)
5013 {
5014         int error;
5015         dsl_pool_t *dp;
5016         dsl_dataset_t *new, *old;
5017 
5018         error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5019         if (error != 0)
5020                 return (error);
5021         error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
5022         if (error != 0) {
5023                 dsl_pool_rele(dp, FTAG);
5024                 return (error);
5025         }
5026         error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
5027         if (error != 0) {
5028                 dsl_dataset_rele(new, FTAG);
5029                 dsl_pool_rele(dp, FTAG);
5030                 return (error);
5031         }
5032 
5033         error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
5034             &zc->zc_objset_type, &zc->zc_perm_action);
5035         dsl_dataset_rele(old, FTAG);
5036         dsl_dataset_rele(new, FTAG);
5037         dsl_pool_rele(dp, FTAG);
5038         return (error);
5039 }
5040 
5041 /*
5042  * innvl: {
5043  *     "firstsnap" -> snapshot name
5044  * }
5045  *
5046  * outnvl: {
5047  *     "used" -> space in bytes
5048  *     "compressed" -> compressed space in bytes
5049  *     "uncompressed" -> uncompressed space in bytes
5050  * }
5051  */
5052 static int
5053 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
5054 {
5055         int error;
5056         dsl_pool_t *dp;
5057         dsl_dataset_t *new, *old;
5058         char *firstsnap;
5059         uint64_t used, comp, uncomp;
5060 
5061         if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5062                 return (SET_ERROR(EINVAL));
5063 
5064         error = dsl_pool_hold(lastsnap, FTAG, &dp);
5065         if (error != 0)
5066                 return (error);
5067 
5068         error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
5069         if (error != 0) {
5070                 dsl_pool_rele(dp, FTAG);
5071                 return (error);
5072         }
5073         error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
5074         if (error != 0) {
5075                 dsl_dataset_rele(new, FTAG);
5076                 dsl_pool_rele(dp, FTAG);
5077                 return (error);
5078         }
5079 
5080         error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5081         dsl_dataset_rele(old, FTAG);
5082         dsl_dataset_rele(new, FTAG);
5083         dsl_pool_rele(dp, FTAG);
5084         fnvlist_add_uint64(outnvl, "used", used);
5085         fnvlist_add_uint64(outnvl, "compressed", comp);
5086         fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5087         return (error);
5088 }
5089 
5090 /*
5091  * innvl: {
5092  *     "fd" -> file descriptor to write stream to (int32)
5093  *     (optional) "fromsnap" -> full snap name to send an incremental from
5094  * }
5095  *
5096  * outnvl is unused
5097  */
5098 /* ARGSUSED */
5099 static int
5100 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5101 {
5102         int error;
5103         offset_t off;
5104         char *fromname = NULL;
5105         int fd;
5106 
5107         error = nvlist_lookup_int32(innvl, "fd", &fd);
5108         if (error != 0)
5109                 return (SET_ERROR(EINVAL));
5110 
5111         (void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
5112 
5113         file_t *fp = getf(fd);
5114         if (fp == NULL)
5115                 return (SET_ERROR(EBADF));
5116 
5117         off = fp->f_offset;
5118         error = dmu_send(snapname, fromname, fd, fp->f_vnode, &off);
5119 
5120         if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5121                 fp->f_offset = off;
5122         releasef(fd);
5123         return (error);
5124 }
5125 
5126 /*
5127  * Determine approximately how large a zfs send stream will be -- the number
5128  * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5129  *
5130  * innvl: {
5131  *     (optional) "fromsnap" -> full snap name to send an incremental from
5132  * }
5133  *
5134  * outnvl: {
5135  *     "space" -> bytes of space (uint64)
5136  * }
5137  */
5138 static int
5139 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5140 {
5141         dsl_pool_t *dp;
5142         dsl_dataset_t *fromsnap = NULL;
5143         dsl_dataset_t *tosnap;
5144         int error;
5145         char *fromname;
5146         uint64_t space;
5147 
5148         error = dsl_pool_hold(snapname, FTAG, &dp);
5149         if (error != 0)
5150                 return (error);
5151 
5152         error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
5153         if (error != 0) {
5154                 dsl_pool_rele(dp, FTAG);
5155                 return (error);
5156         }
5157 
5158         error = nvlist_lookup_string(innvl, "fromsnap", &fromname);
5159         if (error == 0) {
5160                 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
5161                 if (error != 0) {
5162                         dsl_dataset_rele(tosnap, FTAG);
5163                         dsl_pool_rele(dp, FTAG);
5164                         return (error);
5165                 }
5166         }
5167 
5168         error = dmu_send_estimate(tosnap, fromsnap, &space);
5169         fnvlist_add_uint64(outnvl, "space", space);
5170 
5171         if (fromsnap != NULL)
5172                 dsl_dataset_rele(fromsnap, FTAG);
5173         dsl_dataset_rele(tosnap, FTAG);
5174         dsl_pool_rele(dp, FTAG);
5175         return (error);
5176 }
5177 
5178 
5179 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
5180 
5181 static void
5182 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5183     zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5184     boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
5185 {
5186         zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5187 
5188         ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5189         ASSERT3U(ioc, <, ZFS_IOC_LAST);
5190         ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5191         ASSERT3P(vec->zvec_func, ==, NULL);
5192 
5193         vec->zvec_legacy_func = func;
5194         vec->zvec_secpolicy = secpolicy;
5195         vec->zvec_namecheck = namecheck;
5196         vec->zvec_allow_log = log_history;
5197         vec->zvec_pool_check = pool_check;
5198 }
5199 
5200 /*
5201  * See the block comment at the beginning of this file for details on
5202  * each argument to this function.
5203  */
5204 static void
5205 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
5206     zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5207     zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
5208     boolean_t allow_log)
5209 {
5210         zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5211 
5212         ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5213         ASSERT3U(ioc, <, ZFS_IOC_LAST);
5214         ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5215         ASSERT3P(vec->zvec_func, ==, NULL);
5216 
5217         /* if we are logging, the name must be valid */
5218         ASSERT(!allow_log || namecheck != NO_NAME);
5219 
5220         vec->zvec_name = name;
5221         vec->zvec_func = func;
5222         vec->zvec_secpolicy = secpolicy;
5223         vec->zvec_namecheck = namecheck;
5224         vec->zvec_pool_check = pool_check;
5225         vec->zvec_smush_outnvlist = smush_outnvlist;
5226         vec->zvec_allow_log = allow_log;
5227 }
5228 
5229 static void
5230 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5231     zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
5232     zfs_ioc_poolcheck_t pool_check)
5233 {
5234         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5235             POOL_NAME, log_history, pool_check);
5236 }
5237 
5238 static void
5239 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5240     zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
5241 {
5242         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5243             DATASET_NAME, B_FALSE, pool_check);
5244 }
5245 
5246 static void
5247 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5248 {
5249         zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
5250             POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5251 }
5252 
5253 static void
5254 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5255     zfs_secpolicy_func_t *secpolicy)
5256 {
5257         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5258             NO_NAME, B_FALSE, POOL_CHECK_NONE);
5259 }
5260 
5261 static void
5262 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
5263     zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
5264 {
5265         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5266             DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
5267 }
5268 
5269 static void
5270 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5271 {
5272         zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
5273             zfs_secpolicy_read);
5274 }
5275 
5276 static void
5277 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5278         zfs_secpolicy_func_t *secpolicy)
5279 {
5280         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5281             DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5282 }
5283 
5284 static void
5285 zfs_ioctl_init(void)
5286 {
5287         zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
5288             zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
5289             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5290 
5291         zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
5292             zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
5293             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
5294 
5295         zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
5296             zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
5297             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5298 
5299         zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
5300             zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
5301             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5302 
5303         zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
5304             zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
5305             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5306 
5307         zfs_ioctl_register("create", ZFS_IOC_CREATE,
5308             zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
5309             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5310 
5311         zfs_ioctl_register("clone", ZFS_IOC_CLONE,
5312             zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
5313             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5314 
5315         zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
5316             zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
5317             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5318 
5319         zfs_ioctl_register("hold", ZFS_IOC_HOLD,
5320             zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
5321             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5322         zfs_ioctl_register("release", ZFS_IOC_RELEASE,
5323             zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
5324             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5325 
5326         zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
5327             zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
5328             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5329 
5330         zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
5331             zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
5332             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE);
5333 
5334         /* IOCTLS that use the legacy function signature */
5335 
5336         zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
5337             zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
5338 
5339         zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
5340             zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5341         zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
5342             zfs_ioc_pool_scan);
5343         zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
5344             zfs_ioc_pool_upgrade);
5345         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
5346             zfs_ioc_vdev_add);
5347         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
5348             zfs_ioc_vdev_remove);
5349         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
5350             zfs_ioc_vdev_set_state);
5351         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
5352             zfs_ioc_vdev_attach);
5353         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
5354             zfs_ioc_vdev_detach);
5355         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
5356             zfs_ioc_vdev_setpath);
5357         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
5358             zfs_ioc_vdev_setfru);
5359         zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
5360             zfs_ioc_pool_set_props);
5361         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
5362             zfs_ioc_vdev_split);
5363         zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
5364             zfs_ioc_pool_reguid);
5365 
5366         zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
5367             zfs_ioc_pool_configs, zfs_secpolicy_none);
5368         zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
5369             zfs_ioc_pool_tryimport, zfs_secpolicy_config);
5370         zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
5371             zfs_ioc_inject_fault, zfs_secpolicy_inject);
5372         zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
5373             zfs_ioc_clear_fault, zfs_secpolicy_inject);
5374         zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
5375             zfs_ioc_inject_list_next, zfs_secpolicy_inject);
5376 
5377         /*
5378          * pool destroy, and export don't log the history as part of
5379          * zfsdev_ioctl, but rather zfs_ioc_pool_export
5380          * does the logging of those commands.
5381          */
5382         zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
5383             zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5384         zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
5385             zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5386 
5387         zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
5388             zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5389         zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
5390             zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5391 
5392         zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
5393             zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED);
5394         zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
5395             zfs_ioc_dsobj_to_dsname,
5396             zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED);
5397         zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
5398             zfs_ioc_pool_get_history,
5399             zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
5400 
5401         zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
5402             zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5403 
5404         zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
5405             zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5406         zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
5407             zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5408 
5409         zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
5410             zfs_ioc_space_written);
5411         zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
5412             zfs_ioc_objset_recvd_props);
5413         zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
5414             zfs_ioc_next_obj);
5415         zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
5416             zfs_ioc_get_fsacl);
5417         zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
5418             zfs_ioc_objset_stats);
5419         zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
5420             zfs_ioc_objset_zplprops);
5421         zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
5422             zfs_ioc_dataset_list_next);
5423         zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
5424             zfs_ioc_snapshot_list_next);
5425         zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
5426             zfs_ioc_send_progress);
5427 
5428         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
5429             zfs_ioc_diff, zfs_secpolicy_diff);
5430         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
5431             zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
5432         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
5433             zfs_ioc_obj_to_path, zfs_secpolicy_diff);
5434         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
5435             zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
5436         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
5437             zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
5438         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
5439             zfs_ioc_send, zfs_secpolicy_send);
5440 
5441         zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
5442             zfs_secpolicy_none);
5443         zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
5444             zfs_secpolicy_destroy);
5445         zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
5446             zfs_secpolicy_rename);
5447         zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
5448             zfs_secpolicy_recv);
5449         zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
5450             zfs_secpolicy_promote);
5451         zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
5452             zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
5453         zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
5454             zfs_secpolicy_set_fsacl);
5455 
5456         zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
5457             zfs_secpolicy_share, POOL_CHECK_NONE);
5458         zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
5459             zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
5460         zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
5461             zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
5462             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5463         zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
5464             zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
5465             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5466 }
5467 
5468 int
5469 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
5470     zfs_ioc_poolcheck_t check)
5471 {
5472         spa_t *spa;
5473         int error;
5474 
5475         ASSERT(type == POOL_NAME || type == DATASET_NAME);
5476 
5477         if (check & POOL_CHECK_NONE)
5478                 return (0);
5479 
5480         error = spa_open(name, &spa, FTAG);
5481         if (error == 0) {
5482                 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
5483                         error = SET_ERROR(EAGAIN);
5484                 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
5485                         error = SET_ERROR(EROFS);
5486                 spa_close(spa, FTAG);
5487         }
5488         return (error);
5489 }
5490 
5491 /*
5492  * Find a free minor number.
5493  */
5494 minor_t
5495 zfsdev_minor_alloc(void)
5496 {
5497         static minor_t last_minor;
5498         minor_t m;
5499 
5500         ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5501 
5502         for (m = last_minor + 1; m != last_minor; m++) {
5503                 if (m > ZFSDEV_MAX_MINOR)
5504                         m = 1;
5505                 if (ddi_get_soft_state(zfsdev_state, m) == NULL) {
5506                         last_minor = m;
5507                         return (m);
5508                 }
5509         }
5510 
5511         return (0);
5512 }
5513 
5514 static int
5515 zfs_ctldev_init(dev_t *devp)
5516 {
5517         minor_t minor;
5518         zfs_soft_state_t *zs;
5519 
5520         ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5521         ASSERT(getminor(*devp) == 0);
5522 
5523         minor = zfsdev_minor_alloc();
5524         if (minor == 0)
5525                 return (SET_ERROR(ENXIO));
5526 
5527         if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS)
5528                 return (SET_ERROR(EAGAIN));
5529 
5530         *devp = makedevice(getemajor(*devp), minor);
5531 
5532         zs = ddi_get_soft_state(zfsdev_state, minor);
5533         zs->zss_type = ZSST_CTLDEV;
5534         zfs_onexit_init((zfs_onexit_t **)&zs->zss_data);
5535 
5536         return (0);
5537 }
5538 
5539 static void
5540 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor)
5541 {
5542         ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5543 
5544         zfs_onexit_destroy(zo);
5545         ddi_soft_state_free(zfsdev_state, minor);
5546 }
5547 
5548 void *
5549 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which)
5550 {
5551         zfs_soft_state_t *zp;
5552 
5553         zp = ddi_get_soft_state(zfsdev_state, minor);
5554         if (zp == NULL || zp->zss_type != which)
5555                 return (NULL);
5556 
5557         return (zp->zss_data);
5558 }
5559 
5560 static int
5561 zfsdev_open(dev_t *devp, int flag, int otyp, cred_t *cr)
5562 {
5563         int error = 0;
5564 
5565         if (getminor(*devp) != 0)
5566                 return (zvol_open(devp, flag, otyp, cr));
5567 
5568         /* This is the control device. Allocate a new minor if requested. */
5569         if (flag & FEXCL) {
5570                 mutex_enter(&zfsdev_state_lock);
5571                 error = zfs_ctldev_init(devp);
5572                 mutex_exit(&zfsdev_state_lock);
5573         }
5574 
5575         return (error);
5576 }
5577 
5578 static int
5579 zfsdev_close(dev_t dev, int flag, int otyp, cred_t *cr)
5580 {
5581         zfs_onexit_t *zo;
5582         minor_t minor = getminor(dev);
5583 
5584         if (minor == 0)
5585                 return (0);
5586 
5587         mutex_enter(&zfsdev_state_lock);
5588         zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV);
5589         if (zo == NULL) {
5590                 mutex_exit(&zfsdev_state_lock);
5591                 return (zvol_close(dev, flag, otyp, cr));
5592         }
5593         zfs_ctldev_destroy(zo, minor);
5594         mutex_exit(&zfsdev_state_lock);
5595 
5596         return (0);
5597 }
5598 
5599 static int
5600 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
5601 {
5602         zfs_cmd_t *zc;
5603         uint_t vecnum;
5604         int error, rc, len;
5605         minor_t minor = getminor(dev);
5606         const zfs_ioc_vec_t *vec;
5607         char *saved_poolname = NULL;
5608         nvlist_t *innvl = NULL;
5609 
5610         if (minor != 0 &&
5611             zfsdev_get_soft_state(minor, ZSST_CTLDEV) == NULL)
5612                 return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
5613 
5614         vecnum = cmd - ZFS_IOC_FIRST;
5615         ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
5616 
5617         if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
5618                 return (SET_ERROR(EINVAL));
5619         vec = &zfs_ioc_vec[vecnum];
5620 
5621         zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
5622 
5623         error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
5624         if (error != 0) {
5625                 error = SET_ERROR(EFAULT);
5626                 goto out;
5627         }
5628 
5629         zc->zc_iflags = flag & FKIOCTL;
5630         if (zc->zc_nvlist_src_size != 0) {
5631                 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
5632                     zc->zc_iflags, &innvl);
5633                 if (error != 0)
5634                         goto out;
5635         }
5636 
5637         /*
5638          * Ensure that all pool/dataset names are valid before we pass down to
5639          * the lower layers.
5640          */
5641         zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
5642         switch (vec->zvec_namecheck) {
5643         case POOL_NAME:
5644                 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
5645                         error = SET_ERROR(EINVAL);
5646                 else
5647                         error = pool_status_check(zc->zc_name,
5648                             vec->zvec_namecheck, vec->zvec_pool_check);
5649                 break;
5650 
5651         case DATASET_NAME:
5652                 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
5653                         error = SET_ERROR(EINVAL);
5654                 else
5655                         error = pool_status_check(zc->zc_name,
5656                             vec->zvec_namecheck, vec->zvec_pool_check);
5657                 break;
5658 
5659         case NO_NAME:
5660                 break;
5661         }
5662 
5663 
5664         if (error == 0 && !(flag & FKIOCTL))
5665                 error = vec->zvec_secpolicy(zc, innvl, cr);
5666 
5667         if (error != 0)
5668                 goto out;
5669 
5670         /* legacy ioctls can modify zc_name */
5671         len = strcspn(zc->zc_name, "/@") + 1;
5672         saved_poolname = kmem_alloc(len, KM_SLEEP);
5673         (void) strlcpy(saved_poolname, zc->zc_name, len);
5674 
5675         if (vec->zvec_func != NULL) {
5676                 nvlist_t *outnvl;
5677                 int puterror = 0;
5678                 spa_t *spa;
5679                 nvlist_t *lognv = NULL;
5680 
5681                 ASSERT(vec->zvec_legacy_func == NULL);
5682 
5683                 /*
5684                  * Add the innvl to the lognv before calling the func,
5685                  * in case the func changes the innvl.
5686                  */
5687                 if (vec->zvec_allow_log) {
5688                         lognv = fnvlist_alloc();
5689                         fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
5690                             vec->zvec_name);
5691                         if (!nvlist_empty(innvl)) {
5692                                 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
5693                                     innvl);
5694                         }
5695                 }
5696 
5697                 outnvl = fnvlist_alloc();
5698                 error = vec->zvec_func(zc->zc_name, innvl, outnvl);
5699 
5700                 if (error == 0 && vec->zvec_allow_log &&
5701                     spa_open(zc->zc_name, &spa, FTAG) == 0) {
5702                         if (!nvlist_empty(outnvl)) {
5703                                 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
5704                                     outnvl);
5705                         }
5706                         (void) spa_history_log_nvl(spa, lognv);
5707                         spa_close(spa, FTAG);
5708                 }
5709                 fnvlist_free(lognv);
5710 
5711                 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
5712                         int smusherror = 0;
5713                         if (vec->zvec_smush_outnvlist) {
5714                                 smusherror = nvlist_smush(outnvl,
5715                                     zc->zc_nvlist_dst_size);
5716                         }
5717                         if (smusherror == 0)
5718                                 puterror = put_nvlist(zc, outnvl);
5719                 }
5720 
5721                 if (puterror != 0)
5722                         error = puterror;
5723 
5724                 nvlist_free(outnvl);
5725         } else {
5726                 error = vec->zvec_legacy_func(zc);
5727         }
5728 
5729 out:
5730         nvlist_free(innvl);
5731         rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
5732         if (error == 0 && rc != 0)
5733                 error = SET_ERROR(EFAULT);
5734         if (error == 0 && vec->zvec_allow_log) {
5735                 char *s = tsd_get(zfs_allow_log_key);
5736                 if (s != NULL)
5737                         strfree(s);
5738                 (void) tsd_set(zfs_allow_log_key, saved_poolname);
5739         } else {
5740                 if (saved_poolname != NULL)
5741                         strfree(saved_poolname);
5742         }
5743 
5744         kmem_free(zc, sizeof (zfs_cmd_t));
5745         return (error);
5746 }
5747 
5748 static int
5749 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
5750 {
5751         if (cmd != DDI_ATTACH)
5752                 return (DDI_FAILURE);
5753 
5754         if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
5755             DDI_PSEUDO, 0) == DDI_FAILURE)
5756                 return (DDI_FAILURE);
5757 
5758         zfs_dip = dip;
5759 
5760         ddi_report_dev(dip);
5761 
5762         return (DDI_SUCCESS);
5763 }
5764 
5765 static int
5766 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
5767 {
5768         if (spa_busy() || zfs_busy() || zvol_busy())
5769                 return (DDI_FAILURE);
5770 
5771         if (cmd != DDI_DETACH)
5772                 return (DDI_FAILURE);
5773 
5774         zfs_dip = NULL;
5775 
5776         ddi_prop_remove_all(dip);
5777         ddi_remove_minor_node(dip, NULL);
5778 
5779         return (DDI_SUCCESS);
5780 }
5781 
5782 /*ARGSUSED*/
5783 static int
5784 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
5785 {
5786         switch (infocmd) {
5787         case DDI_INFO_DEVT2DEVINFO:
5788                 *result = zfs_dip;
5789                 return (DDI_SUCCESS);
5790 
5791         case DDI_INFO_DEVT2INSTANCE:
5792                 *result = (void *)0;
5793                 return (DDI_SUCCESS);
5794         }
5795 
5796         return (DDI_FAILURE);
5797 }
5798 
5799 /*
5800  * OK, so this is a little weird.
5801  *
5802  * /dev/zfs is the control node, i.e. minor 0.
5803  * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
5804  *
5805  * /dev/zfs has basically nothing to do except serve up ioctls,
5806  * so most of the standard driver entry points are in zvol.c.
5807  */
5808 static struct cb_ops zfs_cb_ops = {
5809         zfsdev_open,    /* open */
5810         zfsdev_close,   /* close */
5811         zvol_strategy,  /* strategy */
5812         nodev,          /* print */
5813         zvol_dump,      /* dump */
5814         zvol_read,      /* read */
5815         zvol_write,     /* write */
5816         zfsdev_ioctl,   /* ioctl */
5817         nodev,          /* devmap */
5818         nodev,          /* mmap */
5819         nodev,          /* segmap */
5820         nochpoll,       /* poll */
5821         ddi_prop_op,    /* prop_op */
5822         NULL,           /* streamtab */
5823         D_NEW | D_MP | D_64BIT,         /* Driver compatibility flag */
5824         CB_REV,         /* version */
5825         nodev,          /* async read */
5826         nodev,          /* async write */
5827 };
5828 
5829 static struct dev_ops zfs_dev_ops = {
5830         DEVO_REV,       /* version */
5831         0,              /* refcnt */
5832         zfs_info,       /* info */
5833         nulldev,        /* identify */
5834         nulldev,        /* probe */
5835         zfs_attach,     /* attach */
5836         zfs_detach,     /* detach */
5837         nodev,          /* reset */
5838         &zfs_cb_ops,        /* driver operations */
5839         NULL,           /* no bus operations */
5840         NULL,           /* power */
5841         ddi_quiesce_not_needed, /* quiesce */
5842 };
5843 
5844 static struct modldrv zfs_modldrv = {
5845         &mod_driverops,
5846         "ZFS storage pool",
5847         &zfs_dev_ops
5848 };
5849 
5850 static struct modlinkage modlinkage = {
5851         MODREV_1,
5852         (void *)&zfs_modlfs,
5853         (void *)&zfs_modldrv,
5854         NULL
5855 };
5856 
5857 static void
5858 zfs_allow_log_destroy(void *arg)
5859 {
5860         char *poolname = arg;
5861         strfree(poolname);
5862 }
5863 
5864 int
5865 _init(void)
5866 {
5867         int error;
5868 
5869         spa_init(FREAD | FWRITE);
5870         zfs_init();
5871         zvol_init();
5872         zfs_ioctl_init();
5873 
5874         if ((error = mod_install(&modlinkage)) != 0) {
5875                 zvol_fini();
5876                 zfs_fini();
5877                 spa_fini();
5878                 return (error);
5879         }
5880 
5881         tsd_create(&zfs_fsyncer_key, NULL);
5882         tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
5883         tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
5884 
5885         error = ldi_ident_from_mod(&modlinkage, &zfs_li);
5886         ASSERT(error == 0);
5887         mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
5888 
5889         return (0);
5890 }
5891 
5892 int
5893 _fini(void)
5894 {
5895         int error;
5896 
5897         if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
5898                 return (SET_ERROR(EBUSY));
5899 
5900         if ((error = mod_remove(&modlinkage)) != 0)
5901                 return (error);
5902 
5903         zvol_fini();
5904         zfs_fini();
5905         spa_fini();
5906         if (zfs_nfsshare_inited)
5907                 (void) ddi_modclose(nfs_mod);
5908         if (zfs_smbshare_inited)
5909                 (void) ddi_modclose(smbsrv_mod);
5910         if (zfs_nfsshare_inited || zfs_smbshare_inited)
5911                 (void) ddi_modclose(sharefs_mod);
5912 
5913         tsd_destroy(&zfs_fsyncer_key);
5914         ldi_ident_release(zfs_li);
5915         zfs_li = NULL;
5916         mutex_destroy(&zfs_share_lock);
5917 
5918         return (error);
5919 }
5920 
5921 int
5922 _info(struct modinfo *modinfop)
5923 {
5924         return (mod_info(&modlinkage, modinfop));
5925 }