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