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.
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.
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]
23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Copyright (c) 2011-2012 Pawel Jakub Dawidek <pawel@dawidek.net>.
25 * All rights reserved.
26 * Copyright 2013 Martin Matuska <mm@FreeBSD.org>. All rights reserved.
27 * Copyright 2014 Xin Li <delphij@FreeBSD.org>. All rights reserved.
28 * Copyright 2015, OmniTI Computer Consulting, Inc. All rights reserved.
29 * Copyright 2015 Nexenta Systems, Inc. All rights reserved.
30 * Copyright (c) 2014, Joyent, Inc. All rights reserved.
31 * Copyright (c) 2011, 2015 by Delphix. All rights reserved.
32 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
33 * Copyright (c) 2013 Steven Hartland. All rights reserved.
34 * Copyright (c) 2014 Integros [integros.com]
40 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
41 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
43 * There are two ways that we handle ioctls: the legacy way where almost
44 * all of the logic is in the ioctl callback, and the new way where most
45 * of the marshalling is handled in the common entry point, zfsdev_ioctl().
47 * Non-legacy ioctls should be registered by calling
48 * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked
49 * from userland by lzc_ioctl().
51 * The registration arguments are as follows:
54 * The name of the ioctl. This is used for history logging. If the
55 * ioctl returns successfully (the callback returns 0), and allow_log
56 * is true, then a history log entry will be recorded with the input &
57 * output nvlists. The log entry can be printed with "zpool history -i".
60 * The ioctl request number, which userland will pass to ioctl(2).
61 * The ioctl numbers can change from release to release, because
62 * the caller (libzfs) must be matched to the kernel.
64 * zfs_secpolicy_func_t *secpolicy
65 * This function will be called before the zfs_ioc_func_t, to
66 * determine if this operation is permitted. It should return EPERM
67 * on failure, and 0 on success. Checks include determining if the
68 * dataset is visible in this zone, and if the user has either all
69 * zfs privileges in the zone (SYS_MOUNT), or has been granted permission
70 * to do this operation on this dataset with "zfs allow".
72 * zfs_ioc_namecheck_t namecheck
73 * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
74 * name, a dataset name, or nothing. If the name is not well-formed,
75 * the ioctl will fail and the callback will not be called.
76 * Therefore, the callback can assume that the name is well-formed
77 * (e.g. is null-terminated, doesn't have more than one '@' character,
78 * doesn't have invalid characters).
80 * zfs_ioc_poolcheck_t pool_check
81 * This specifies requirements on the pool state. If the pool does
82 * not meet them (is suspended or is readonly), the ioctl will fail
83 * and the callback will not be called. If any checks are specified
84 * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
85 * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED |
86 * POOL_CHECK_READONLY).
88 * boolean_t smush_outnvlist
89 * If smush_outnvlist is true, then the output is presumed to be a
90 * list of errors, and it will be "smushed" down to fit into the
91 * caller's buffer, by removing some entries and replacing them with a
92 * single "N_MORE_ERRORS" entry indicating how many were removed. See
93 * nvlist_smush() for details. If smush_outnvlist is false, and the
94 * outnvlist does not fit into the userland-provided buffer, then the
95 * ioctl will fail with ENOMEM.
97 * zfs_ioc_func_t *func
98 * The callback function that will perform the operation.
100 * The callback should return 0 on success, or an error number on
101 * failure. If the function fails, the userland ioctl will return -1,
102 * and errno will be set to the callback's return value. The callback
103 * will be called with the following arguments:
106 * The name of the pool or dataset to operate on, from
107 * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the
108 * expected type (pool, dataset, or none).
111 * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or
112 * NULL if no input nvlist was provided. Changes to this nvlist are
113 * ignored. If the input nvlist could not be deserialized, the
114 * ioctl will fail and the callback will not be called.
117 * The output nvlist, initially empty. The callback can fill it in,
118 * and it will be returned to userland by serializing it into
119 * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization
120 * fails (e.g. because the caller didn't supply a large enough
121 * buffer), then the overall ioctl will fail. See the
122 * 'smush_nvlist' argument above for additional behaviors.
124 * There are two typical uses of the output nvlist:
125 * - To return state, e.g. property values. In this case,
126 * smush_outnvlist should be false. If the buffer was not large
127 * enough, the caller will reallocate a larger buffer and try
130 * - To return multiple errors from an ioctl which makes on-disk
131 * changes. In this case, smush_outnvlist should be true.
132 * Ioctls which make on-disk modifications should generally not
133 * use the outnvl if they succeed, because the caller can not
134 * distinguish between the operation failing, and
135 * deserialization failing.
138 #include "opt_kstack_pages.h"
141 #include <sys/types.h>
142 #include <sys/param.h>
143 #include <sys/systm.h>
144 #include <sys/conf.h>
145 #include <sys/kernel.h>
146 #include <sys/lock.h>
147 #include <sys/malloc.h>
148 #include <sys/mutex.h>
149 #include <sys/proc.h>
150 #include <sys/errno.h>
153 #include <sys/file.h>
154 #include <sys/kmem.h>
155 #include <sys/conf.h>
156 #include <sys/cmn_err.h>
157 #include <sys/stat.h>
158 #include <sys/zfs_ioctl.h>
159 #include <sys/zfs_vfsops.h>
160 #include <sys/zfs_znode.h>
163 #include <sys/spa_impl.h>
164 #include <sys/vdev.h>
166 #include <sys/dsl_dir.h>
167 #include <sys/dsl_dataset.h>
168 #include <sys/dsl_prop.h>
169 #include <sys/dsl_deleg.h>
170 #include <sys/dmu_objset.h>
171 #include <sys/dmu_impl.h>
172 #include <sys/dmu_tx.h>
173 #include <sys/sunddi.h>
174 #include <sys/policy.h>
175 #include <sys/zone.h>
176 #include <sys/nvpair.h>
177 #include <sys/mount.h>
178 #include <sys/taskqueue.h>
180 #include <sys/varargs.h>
181 #include <sys/fs/zfs.h>
182 #include <sys/zfs_ctldir.h>
183 #include <sys/zfs_dir.h>
184 #include <sys/zfs_onexit.h>
185 #include <sys/zvol.h>
186 #include <sys/dsl_scan.h>
187 #include <sys/dmu_objset.h>
188 #include <sys/dmu_send.h>
189 #include <sys/dsl_destroy.h>
190 #include <sys/dsl_bookmark.h>
191 #include <sys/dsl_userhold.h>
192 #include <sys/zfeature.h>
193 #include <sys/zio_checksum.h>
195 #include "zfs_namecheck.h"
196 #include "zfs_prop.h"
197 #include "zfs_deleg.h"
198 #include "zfs_comutil.h"
199 #include "zfs_ioctl_compat.h"
201 CTASSERT(sizeof(zfs_cmd_t) < IOCPARM_MAX);
203 static struct cdev *zfsdev;
205 extern void zfs_init(void);
206 extern void zfs_fini(void);
208 uint_t zfs_fsyncer_key;
209 extern uint_t rrw_tsd_key;
210 static uint_t zfs_allow_log_key;
212 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *);
213 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *);
214 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *);
220 } zfs_ioc_namecheck_t;
223 POOL_CHECK_NONE = 1 << 0,
224 POOL_CHECK_SUSPENDED = 1 << 1,
225 POOL_CHECK_READONLY = 1 << 2,
226 } zfs_ioc_poolcheck_t;
228 typedef struct zfs_ioc_vec {
229 zfs_ioc_legacy_func_t *zvec_legacy_func;
230 zfs_ioc_func_t *zvec_func;
231 zfs_secpolicy_func_t *zvec_secpolicy;
232 zfs_ioc_namecheck_t zvec_namecheck;
233 boolean_t zvec_allow_log;
234 zfs_ioc_poolcheck_t zvec_pool_check;
235 boolean_t zvec_smush_outnvlist;
236 const char *zvec_name;
239 /* This array is indexed by zfs_userquota_prop_t */
240 static const char *userquota_perms[] = {
241 ZFS_DELEG_PERM_USERUSED,
242 ZFS_DELEG_PERM_USERQUOTA,
243 ZFS_DELEG_PERM_GROUPUSED,
244 ZFS_DELEG_PERM_GROUPQUOTA,
247 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
248 static int zfs_check_settable(const char *name, nvpair_t *property,
250 static int zfs_check_clearable(char *dataset, nvlist_t *props,
252 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
254 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
255 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
257 static void zfsdev_close(void *data);
259 static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature);
261 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
263 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
270 * Get rid of annoying "../common/" prefix to filename.
272 newfile = strrchr(file, '/');
273 if (newfile != NULL) {
274 newfile = newfile + 1; /* Get rid of leading / */
280 (void) vsnprintf(buf, sizeof (buf), fmt, adx);
284 * To get this data, use the zfs-dprintf probe as so:
285 * dtrace -q -n 'zfs-dprintf \
286 * /stringof(arg0) == "dbuf.c"/ \
287 * {printf("%s: %s", stringof(arg1), stringof(arg3))}'
289 * arg1 = function name
293 DTRACE_PROBE4(zfs__dprintf,
294 char *, newfile, char *, func, int, line, char *, buf);
298 history_str_free(char *buf)
300 kmem_free(buf, HIS_MAX_RECORD_LEN);
304 history_str_get(zfs_cmd_t *zc)
308 if (zc->zc_history == 0)
311 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
312 if (copyinstr((void *)(uintptr_t)zc->zc_history,
313 buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
314 history_str_free(buf);
318 buf[HIS_MAX_RECORD_LEN -1] = '\0';
324 * Check to see if the named dataset is currently defined as bootable
327 zfs_is_bootfs(const char *name)
331 if (dmu_objset_hold(name, FTAG, &os) == 0) {
333 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
334 dmu_objset_rele(os, FTAG);
341 * Return non-zero if the spa version is less than requested version.
344 zfs_earlier_version(const char *name, int version)
348 if (spa_open(name, &spa, FTAG) == 0) {
349 if (spa_version(spa) < version) {
350 spa_close(spa, FTAG);
353 spa_close(spa, FTAG);
359 * Return TRUE if the ZPL version is less than requested version.
362 zpl_earlier_version(const char *name, int version)
365 boolean_t rc = B_TRUE;
367 if (dmu_objset_hold(name, FTAG, &os) == 0) {
370 if (dmu_objset_type(os) != DMU_OST_ZFS) {
371 dmu_objset_rele(os, FTAG);
374 /* XXX reading from non-owned objset */
375 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
376 rc = zplversion < version;
377 dmu_objset_rele(os, FTAG);
383 zfs_log_history(zfs_cmd_t *zc)
388 if ((buf = history_str_get(zc)) == NULL)
391 if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
392 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
393 (void) spa_history_log(spa, buf);
394 spa_close(spa, FTAG);
396 history_str_free(buf);
400 * Policy for top-level read operations (list pools). Requires no privileges,
401 * and can be used in the local zone, as there is no associated dataset.
405 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
411 * Policy for dataset read operations (list children, get statistics). Requires
412 * no privileges, but must be visible in the local zone.
416 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
418 if (INGLOBALZONE(curthread) ||
419 zone_dataset_visible(zc->zc_name, NULL))
422 return (SET_ERROR(ENOENT));
426 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
431 * The dataset must be visible by this zone -- check this first
432 * so they don't see EPERM on something they shouldn't know about.
434 if (!INGLOBALZONE(curthread) &&
435 !zone_dataset_visible(dataset, &writable))
436 return (SET_ERROR(ENOENT));
438 if (INGLOBALZONE(curthread)) {
440 * If the fs is zoned, only root can access it from the
443 if (secpolicy_zfs(cr) && zoned)
444 return (SET_ERROR(EPERM));
447 * If we are in a local zone, the 'zoned' property must be set.
450 return (SET_ERROR(EPERM));
452 /* must be writable by this zone */
454 return (SET_ERROR(EPERM));
460 zfs_dozonecheck(const char *dataset, cred_t *cr)
464 if (dsl_prop_get_integer(dataset, "jailed", &zoned, NULL))
465 return (SET_ERROR(ENOENT));
467 return (zfs_dozonecheck_impl(dataset, zoned, cr));
471 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
475 if (dsl_prop_get_int_ds(ds, "jailed", &zoned))
476 return (SET_ERROR(ENOENT));
478 return (zfs_dozonecheck_impl(dataset, zoned, cr));
482 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
483 const char *perm, cred_t *cr)
487 error = zfs_dozonecheck_ds(name, ds, cr);
489 error = secpolicy_zfs(cr);
491 error = dsl_deleg_access_impl(ds, perm, cr);
497 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
503 error = dsl_pool_hold(name, FTAG, &dp);
507 error = dsl_dataset_hold(dp, name, FTAG, &ds);
509 dsl_pool_rele(dp, FTAG);
513 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
515 dsl_dataset_rele(ds, FTAG);
516 dsl_pool_rele(dp, FTAG);
522 * Policy for setting the security label property.
524 * Returns 0 for success, non-zero for access and other errors.
527 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
529 char ds_hexsl[MAXNAMELEN];
530 bslabel_t ds_sl, new_sl;
531 boolean_t new_default = FALSE;
533 int needed_priv = -1;
536 /* First get the existing dataset label. */
537 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
538 1, sizeof (ds_hexsl), &ds_hexsl, NULL);
540 return (SET_ERROR(EPERM));
542 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
545 /* The label must be translatable */
546 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
547 return (SET_ERROR(EINVAL));
550 * In a non-global zone, disallow attempts to set a label that
551 * doesn't match that of the zone; otherwise no other checks
554 if (!INGLOBALZONE(curproc)) {
555 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
556 return (SET_ERROR(EPERM));
561 * For global-zone datasets (i.e., those whose zoned property is
562 * "off", verify that the specified new label is valid for the
565 if (dsl_prop_get_integer(name,
566 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
567 return (SET_ERROR(EPERM));
569 if (zfs_check_global_label(name, strval) != 0)
570 return (SET_ERROR(EPERM));
574 * If the existing dataset label is nondefault, check if the
575 * dataset is mounted (label cannot be changed while mounted).
576 * Get the zfsvfs; if there isn't one, then the dataset isn't
577 * mounted (or isn't a dataset, doesn't exist, ...).
579 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
581 static char *setsl_tag = "setsl_tag";
584 * Try to own the dataset; abort if there is any error,
585 * (e.g., already mounted, in use, or other error).
587 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
590 return (SET_ERROR(EPERM));
592 dmu_objset_disown(os, setsl_tag);
595 needed_priv = PRIV_FILE_DOWNGRADE_SL;
599 if (hexstr_to_label(strval, &new_sl) != 0)
600 return (SET_ERROR(EPERM));
602 if (blstrictdom(&ds_sl, &new_sl))
603 needed_priv = PRIV_FILE_DOWNGRADE_SL;
604 else if (blstrictdom(&new_sl, &ds_sl))
605 needed_priv = PRIV_FILE_UPGRADE_SL;
607 /* dataset currently has a default label */
609 needed_priv = PRIV_FILE_UPGRADE_SL;
613 if (needed_priv != -1)
614 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
617 #endif /* SECLABEL */
620 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
626 * Check permissions for special properties.
631 * Disallow setting of 'zoned' from within a local zone.
633 if (!INGLOBALZONE(curthread))
634 return (SET_ERROR(EPERM));
638 case ZFS_PROP_FILESYSTEM_LIMIT:
639 case ZFS_PROP_SNAPSHOT_LIMIT:
640 if (!INGLOBALZONE(curthread)) {
642 char setpoint[MAXNAMELEN];
644 * Unprivileged users are allowed to modify the
645 * limit on things *under* (ie. contained by)
646 * the thing they own.
648 if (dsl_prop_get_integer(dsname, "jailed", &zoned,
650 return (SET_ERROR(EPERM));
651 if (!zoned || strlen(dsname) <= strlen(setpoint))
652 return (SET_ERROR(EPERM));
656 case ZFS_PROP_MLSLABEL:
658 if (!is_system_labeled())
659 return (SET_ERROR(EPERM));
661 if (nvpair_value_string(propval, &strval) == 0) {
664 err = zfs_set_slabel_policy(dsname, strval, CRED());
674 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
679 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
683 error = zfs_dozonecheck(zc->zc_name, cr);
688 * permission to set permissions will be evaluated later in
689 * dsl_deleg_can_allow()
696 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
698 return (zfs_secpolicy_write_perms(zc->zc_name,
699 ZFS_DELEG_PERM_ROLLBACK, cr));
704 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
712 * Generate the current snapshot name from the given objsetid, then
713 * use that name for the secpolicy/zone checks.
715 cp = strchr(zc->zc_name, '@');
717 return (SET_ERROR(EINVAL));
718 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
722 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
724 dsl_pool_rele(dp, FTAG);
728 dsl_dataset_name(ds, zc->zc_name);
730 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
731 ZFS_DELEG_PERM_SEND, cr);
732 dsl_dataset_rele(ds, FTAG);
733 dsl_pool_rele(dp, FTAG);
740 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
742 return (zfs_secpolicy_write_perms(zc->zc_name,
743 ZFS_DELEG_PERM_SEND, cr));
748 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
753 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
754 NO_FOLLOW, NULL, &vp)) != 0)
757 /* Now make sure mntpnt and dataset are ZFS */
759 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 ||
760 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
761 zc->zc_name) != 0)) {
763 return (SET_ERROR(EPERM));
767 return (dsl_deleg_access(zc->zc_name,
768 ZFS_DELEG_PERM_SHARE, cr));
772 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
774 if (!INGLOBALZONE(curthread))
775 return (SET_ERROR(EPERM));
777 if (secpolicy_nfs(cr) == 0) {
780 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
785 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
787 if (!INGLOBALZONE(curthread))
788 return (SET_ERROR(EPERM));
790 if (secpolicy_smb(cr) == 0) {
793 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
798 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
803 * Remove the @bla or /bla from the end of the name to get the parent.
805 (void) strncpy(parent, datasetname, parentsize);
806 cp = strrchr(parent, '@');
810 cp = strrchr(parent, '/');
812 return (SET_ERROR(ENOENT));
820 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
824 if ((error = zfs_secpolicy_write_perms(name,
825 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
828 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
833 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
835 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
839 * Destroying snapshots with delegated permissions requires
840 * descendant mount and destroy permissions.
844 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
847 nvpair_t *pair, *nextpair;
850 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
851 return (SET_ERROR(EINVAL));
852 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
854 nextpair = nvlist_next_nvpair(snaps, pair);
855 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr);
856 if (error == ENOENT) {
858 * Ignore any snapshots that don't exist (we consider
859 * them "already destroyed"). Remove the name from the
860 * nvl here in case the snapshot is created between
861 * now and when we try to destroy it (in which case
862 * we don't want to destroy it since we haven't
863 * checked for permission).
865 fnvlist_remove_nvpair(snaps, pair);
876 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
878 char parentname[MAXNAMELEN];
881 if ((error = zfs_secpolicy_write_perms(from,
882 ZFS_DELEG_PERM_RENAME, cr)) != 0)
885 if ((error = zfs_secpolicy_write_perms(from,
886 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
889 if ((error = zfs_get_parent(to, parentname,
890 sizeof (parentname))) != 0)
893 if ((error = zfs_secpolicy_write_perms(parentname,
894 ZFS_DELEG_PERM_CREATE, cr)) != 0)
897 if ((error = zfs_secpolicy_write_perms(parentname,
898 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
906 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
911 if ((zc->zc_cookie & 1) != 0) {
913 * This is recursive rename, so the starting snapshot might
914 * not exist. Check file system or volume permission instead.
916 at = strchr(zc->zc_name, '@');
922 error = zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr);
932 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
935 dsl_dataset_t *clone;
938 error = zfs_secpolicy_write_perms(zc->zc_name,
939 ZFS_DELEG_PERM_PROMOTE, cr);
943 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
947 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
950 char parentname[MAXNAMELEN];
951 dsl_dataset_t *origin = NULL;
955 error = dsl_dataset_hold_obj(dd->dd_pool,
956 dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin);
958 dsl_dataset_rele(clone, FTAG);
959 dsl_pool_rele(dp, FTAG);
963 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
964 ZFS_DELEG_PERM_MOUNT, cr);
966 dsl_dataset_name(origin, parentname);
968 error = zfs_secpolicy_write_perms_ds(parentname, origin,
969 ZFS_DELEG_PERM_PROMOTE, cr);
971 dsl_dataset_rele(clone, FTAG);
972 dsl_dataset_rele(origin, FTAG);
974 dsl_pool_rele(dp, FTAG);
980 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
984 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
985 ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
988 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
989 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
992 return (zfs_secpolicy_write_perms(zc->zc_name,
993 ZFS_DELEG_PERM_CREATE, cr));
997 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
999 return (zfs_secpolicy_write_perms(name,
1000 ZFS_DELEG_PERM_SNAPSHOT, cr));
1004 * Check for permission to create each snapshot in the nvlist.
1008 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1014 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
1015 return (SET_ERROR(EINVAL));
1016 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
1017 pair = nvlist_next_nvpair(snaps, pair)) {
1018 char *name = nvpair_name(pair);
1019 char *atp = strchr(name, '@');
1022 error = SET_ERROR(EINVAL);
1026 error = zfs_secpolicy_snapshot_perms(name, cr);
1035 * Check for permission to create each snapshot in the nvlist.
1039 zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1043 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
1044 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
1045 char *name = nvpair_name(pair);
1046 char *hashp = strchr(name, '#');
1048 if (hashp == NULL) {
1049 error = SET_ERROR(EINVAL);
1053 error = zfs_secpolicy_write_perms(name,
1054 ZFS_DELEG_PERM_BOOKMARK, cr);
1064 zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1066 nvpair_t *pair, *nextpair;
1069 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1071 char *name = nvpair_name(pair);
1072 char *hashp = strchr(name, '#');
1073 nextpair = nvlist_next_nvpair(innvl, pair);
1075 if (hashp == NULL) {
1076 error = SET_ERROR(EINVAL);
1081 error = zfs_secpolicy_write_perms(name,
1082 ZFS_DELEG_PERM_DESTROY, cr);
1084 if (error == ENOENT) {
1086 * Ignore any filesystems that don't exist (we consider
1087 * their bookmarks "already destroyed"). Remove
1088 * the name from the nvl here in case the filesystem
1089 * is created between now and when we try to destroy
1090 * the bookmark (in which case we don't want to
1091 * destroy it since we haven't checked for permission).
1093 fnvlist_remove_nvpair(innvl, pair);
1105 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1108 * Even root must have a proper TSD so that we know what pool
1111 if (tsd_get(zfs_allow_log_key) == NULL)
1112 return (SET_ERROR(EPERM));
1117 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1119 char parentname[MAXNAMELEN];
1123 if ((error = zfs_get_parent(zc->zc_name, parentname,
1124 sizeof (parentname))) != 0)
1127 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1128 (error = zfs_secpolicy_write_perms(origin,
1129 ZFS_DELEG_PERM_CLONE, cr)) != 0)
1132 if ((error = zfs_secpolicy_write_perms(parentname,
1133 ZFS_DELEG_PERM_CREATE, cr)) != 0)
1136 return (zfs_secpolicy_write_perms(parentname,
1137 ZFS_DELEG_PERM_MOUNT, cr));
1141 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires
1142 * SYS_CONFIG privilege, which is not available in a local zone.
1146 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1148 if (secpolicy_sys_config(cr, B_FALSE) != 0)
1149 return (SET_ERROR(EPERM));
1155 * Policy for object to name lookups.
1159 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1163 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1166 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1171 * Policy for fault injection. Requires all privileges.
1175 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1177 return (secpolicy_zinject(cr));
1182 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1184 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1186 if (prop == ZPROP_INVAL) {
1187 if (!zfs_prop_user(zc->zc_value))
1188 return (SET_ERROR(EINVAL));
1189 return (zfs_secpolicy_write_perms(zc->zc_name,
1190 ZFS_DELEG_PERM_USERPROP, cr));
1192 return (zfs_secpolicy_setprop(zc->zc_name, prop,
1198 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1200 int err = zfs_secpolicy_read(zc, innvl, cr);
1204 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1205 return (SET_ERROR(EINVAL));
1207 if (zc->zc_value[0] == 0) {
1209 * They are asking about a posix uid/gid. If it's
1210 * themself, allow it.
1212 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1213 zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
1214 if (zc->zc_guid == crgetuid(cr))
1217 if (groupmember(zc->zc_guid, cr))
1222 return (zfs_secpolicy_write_perms(zc->zc_name,
1223 userquota_perms[zc->zc_objset_type], cr));
1227 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1229 int err = zfs_secpolicy_read(zc, innvl, cr);
1233 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1234 return (SET_ERROR(EINVAL));
1236 return (zfs_secpolicy_write_perms(zc->zc_name,
1237 userquota_perms[zc->zc_objset_type], cr));
1242 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1244 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1250 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1256 error = nvlist_lookup_nvlist(innvl, "holds", &holds);
1258 return (SET_ERROR(EINVAL));
1260 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
1261 pair = nvlist_next_nvpair(holds, pair)) {
1262 char fsname[MAXNAMELEN];
1263 error = dmu_fsname(nvpair_name(pair), fsname);
1266 error = zfs_secpolicy_write_perms(fsname,
1267 ZFS_DELEG_PERM_HOLD, cr);
1276 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1281 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1282 pair = nvlist_next_nvpair(innvl, pair)) {
1283 char fsname[MAXNAMELEN];
1284 error = dmu_fsname(nvpair_name(pair), fsname);
1287 error = zfs_secpolicy_write_perms(fsname,
1288 ZFS_DELEG_PERM_RELEASE, cr);
1296 * Policy for allowing temporary snapshots to be taken or released
1299 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1302 * A temporary snapshot is the same as a snapshot,
1303 * hold, destroy and release all rolled into one.
1304 * Delegated diff alone is sufficient that we allow this.
1308 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1309 ZFS_DELEG_PERM_DIFF, cr)) == 0)
1312 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1314 error = zfs_secpolicy_hold(zc, innvl, cr);
1316 error = zfs_secpolicy_release(zc, innvl, cr);
1318 error = zfs_secpolicy_destroy(zc, innvl, cr);
1323 * Returns the nvlist as specified by the user in the zfs_cmd_t.
1326 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1330 nvlist_t *list = NULL;
1333 * Read in and unpack the user-supplied nvlist.
1336 return (SET_ERROR(EINVAL));
1338 packed = kmem_alloc(size, KM_SLEEP);
1340 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1342 kmem_free(packed, size);
1343 return (SET_ERROR(EFAULT));
1346 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1347 kmem_free(packed, size);
1351 kmem_free(packed, size);
1358 * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1359 * Entries will be removed from the end of the nvlist, and one int32 entry
1360 * named "N_MORE_ERRORS" will be added indicating how many entries were
1364 nvlist_smush(nvlist_t *errors, size_t max)
1368 size = fnvlist_size(errors);
1371 nvpair_t *more_errors;
1375 return (SET_ERROR(ENOMEM));
1377 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1378 more_errors = nvlist_prev_nvpair(errors, NULL);
1381 nvpair_t *pair = nvlist_prev_nvpair(errors,
1383 fnvlist_remove_nvpair(errors, pair);
1385 size = fnvlist_size(errors);
1386 } while (size > max);
1388 fnvlist_remove_nvpair(errors, more_errors);
1389 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1390 ASSERT3U(fnvlist_size(errors), <=, max);
1397 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1399 char *packed = NULL;
1403 size = fnvlist_size(nvl);
1405 if (size > zc->zc_nvlist_dst_size) {
1407 * Solaris returns ENOMEM here, because even if an error is
1408 * returned from an ioctl(2), new zc_nvlist_dst_size will be
1409 * passed to the userland. This is not the case for FreeBSD.
1410 * We need to return 0, so the kernel will copy the
1411 * zc_nvlist_dst_size back and the userland can discover that a
1412 * bigger buffer is needed.
1416 packed = fnvlist_pack(nvl, &size);
1417 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1418 size, zc->zc_iflags) != 0)
1419 error = SET_ERROR(EFAULT);
1420 fnvlist_pack_free(packed, size);
1423 zc->zc_nvlist_dst_size = size;
1424 zc->zc_nvlist_dst_filled = B_TRUE;
1429 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1434 error = dmu_objset_hold(dsname, FTAG, &os);
1437 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1438 dmu_objset_rele(os, FTAG);
1439 return (SET_ERROR(EINVAL));
1442 mutex_enter(&os->os_user_ptr_lock);
1443 *zfvp = dmu_objset_get_user(os);
1445 VFS_HOLD((*zfvp)->z_vfs);
1447 error = SET_ERROR(ESRCH);
1449 mutex_exit(&os->os_user_ptr_lock);
1450 dmu_objset_rele(os, FTAG);
1455 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1456 * case its z_vfs will be NULL, and it will be opened as the owner.
1457 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1458 * which prevents all vnode ops from running.
1461 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1465 if (getzfsvfs(name, zfvp) != 0)
1466 error = zfsvfs_create(name, zfvp);
1468 rrm_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER :
1470 if ((*zfvp)->z_unmounted) {
1472 * XXX we could probably try again, since the unmounting
1473 * thread should be just about to disassociate the
1474 * objset from the zfsvfs.
1476 rrm_exit(&(*zfvp)->z_teardown_lock, tag);
1477 return (SET_ERROR(EBUSY));
1484 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1486 rrm_exit(&zfsvfs->z_teardown_lock, tag);
1488 if (zfsvfs->z_vfs) {
1489 VFS_RELE(zfsvfs->z_vfs);
1491 dmu_objset_disown(zfsvfs->z_os, zfsvfs);
1492 zfsvfs_free(zfsvfs);
1497 zfs_ioc_pool_create(zfs_cmd_t *zc)
1500 nvlist_t *config, *props = NULL;
1501 nvlist_t *rootprops = NULL;
1502 nvlist_t *zplprops = NULL;
1504 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1505 zc->zc_iflags, &config))
1508 if (zc->zc_nvlist_src_size != 0 && (error =
1509 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1510 zc->zc_iflags, &props))) {
1511 nvlist_free(config);
1516 nvlist_t *nvl = NULL;
1517 uint64_t version = SPA_VERSION;
1519 (void) nvlist_lookup_uint64(props,
1520 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1521 if (!SPA_VERSION_IS_SUPPORTED(version)) {
1522 error = SET_ERROR(EINVAL);
1523 goto pool_props_bad;
1525 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1527 error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1529 nvlist_free(config);
1533 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1535 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1536 error = zfs_fill_zplprops_root(version, rootprops,
1539 goto pool_props_bad;
1542 error = spa_create(zc->zc_name, config, props, zplprops);
1545 * Set the remaining root properties
1547 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1548 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1549 (void) spa_destroy(zc->zc_name);
1552 nvlist_free(rootprops);
1553 nvlist_free(zplprops);
1554 nvlist_free(config);
1561 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1564 zfs_log_history(zc);
1565 error = spa_destroy(zc->zc_name);
1567 zvol_remove_minors(zc->zc_name);
1572 zfs_ioc_pool_import(zfs_cmd_t *zc)
1574 nvlist_t *config, *props = NULL;
1578 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1579 zc->zc_iflags, &config)) != 0)
1582 if (zc->zc_nvlist_src_size != 0 && (error =
1583 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1584 zc->zc_iflags, &props))) {
1585 nvlist_free(config);
1589 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1590 guid != zc->zc_guid)
1591 error = SET_ERROR(EINVAL);
1593 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1595 if (zc->zc_nvlist_dst != 0) {
1598 if ((err = put_nvlist(zc, config)) != 0)
1602 nvlist_free(config);
1611 zfs_ioc_pool_export(zfs_cmd_t *zc)
1614 boolean_t force = (boolean_t)zc->zc_cookie;
1615 boolean_t hardforce = (boolean_t)zc->zc_guid;
1617 zfs_log_history(zc);
1618 error = spa_export(zc->zc_name, NULL, force, hardforce);
1620 zvol_remove_minors(zc->zc_name);
1625 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1630 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1631 return (SET_ERROR(EEXIST));
1633 error = put_nvlist(zc, configs);
1635 nvlist_free(configs);
1642 * zc_name name of the pool
1645 * zc_cookie real errno
1646 * zc_nvlist_dst config nvlist
1647 * zc_nvlist_dst_size size of config nvlist
1650 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1656 error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1657 sizeof (zc->zc_value));
1659 if (config != NULL) {
1660 ret = put_nvlist(zc, config);
1661 nvlist_free(config);
1664 * The config may be present even if 'error' is non-zero.
1665 * In this case we return success, and preserve the real errno
1668 zc->zc_cookie = error;
1677 * Try to import the given pool, returning pool stats as appropriate so that
1678 * user land knows which devices are available and overall pool health.
1681 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1683 nvlist_t *tryconfig, *config;
1686 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1687 zc->zc_iflags, &tryconfig)) != 0)
1690 config = spa_tryimport(tryconfig);
1692 nvlist_free(tryconfig);
1695 return (SET_ERROR(EINVAL));
1697 error = put_nvlist(zc, config);
1698 nvlist_free(config);
1705 * zc_name name of the pool
1706 * zc_cookie scan func (pool_scan_func_t)
1709 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1714 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1717 if (zc->zc_cookie == POOL_SCAN_NONE)
1718 error = spa_scan_stop(spa);
1720 error = spa_scan(spa, zc->zc_cookie);
1722 spa_close(spa, FTAG);
1728 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1733 error = spa_open(zc->zc_name, &spa, FTAG);
1736 spa_close(spa, FTAG);
1742 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1747 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1750 if (zc->zc_cookie < spa_version(spa) ||
1751 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1752 spa_close(spa, FTAG);
1753 return (SET_ERROR(EINVAL));
1756 spa_upgrade(spa, zc->zc_cookie);
1757 spa_close(spa, FTAG);
1763 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1770 if ((size = zc->zc_history_len) == 0)
1771 return (SET_ERROR(EINVAL));
1773 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1776 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1777 spa_close(spa, FTAG);
1778 return (SET_ERROR(ENOTSUP));
1781 hist_buf = kmem_alloc(size, KM_SLEEP);
1782 if ((error = spa_history_get(spa, &zc->zc_history_offset,
1783 &zc->zc_history_len, hist_buf)) == 0) {
1784 error = ddi_copyout(hist_buf,
1785 (void *)(uintptr_t)zc->zc_history,
1786 zc->zc_history_len, zc->zc_iflags);
1789 spa_close(spa, FTAG);
1790 kmem_free(hist_buf, size);
1795 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1800 error = spa_open(zc->zc_name, &spa, FTAG);
1802 error = spa_change_guid(spa);
1803 spa_close(spa, FTAG);
1809 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1811 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1816 * zc_name name of filesystem
1817 * zc_obj object to find
1820 * zc_value name of object
1823 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1828 /* XXX reading from objset not owned */
1829 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1831 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1832 dmu_objset_rele(os, FTAG);
1833 return (SET_ERROR(EINVAL));
1835 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1836 sizeof (zc->zc_value));
1837 dmu_objset_rele(os, FTAG);
1844 * zc_name name of filesystem
1845 * zc_obj object to find
1848 * zc_stat stats on object
1849 * zc_value path to object
1852 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1857 /* XXX reading from objset not owned */
1858 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1860 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1861 dmu_objset_rele(os, FTAG);
1862 return (SET_ERROR(EINVAL));
1864 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1865 sizeof (zc->zc_value));
1866 dmu_objset_rele(os, FTAG);
1872 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1876 nvlist_t *config, **l2cache, **spares;
1877 uint_t nl2cache = 0, nspares = 0;
1879 error = spa_open(zc->zc_name, &spa, FTAG);
1883 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1884 zc->zc_iflags, &config);
1885 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1886 &l2cache, &nl2cache);
1888 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1893 * A root pool with concatenated devices is not supported.
1894 * Thus, can not add a device to a root pool.
1896 * Intent log device can not be added to a rootpool because
1897 * during mountroot, zil is replayed, a seperated log device
1898 * can not be accessed during the mountroot time.
1900 * l2cache and spare devices are ok to be added to a rootpool.
1902 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1903 nvlist_free(config);
1904 spa_close(spa, FTAG);
1905 return (SET_ERROR(EDOM));
1907 #endif /* illumos */
1910 error = spa_vdev_add(spa, config);
1911 nvlist_free(config);
1913 spa_close(spa, FTAG);
1919 * zc_name name of the pool
1920 * zc_nvlist_conf nvlist of devices to remove
1921 * zc_cookie to stop the remove?
1924 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1929 error = spa_open(zc->zc_name, &spa, FTAG);
1932 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1933 spa_close(spa, FTAG);
1938 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1942 vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1944 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1946 switch (zc->zc_cookie) {
1947 case VDEV_STATE_ONLINE:
1948 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1951 case VDEV_STATE_OFFLINE:
1952 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1955 case VDEV_STATE_FAULTED:
1956 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1957 zc->zc_obj != VDEV_AUX_EXTERNAL)
1958 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1960 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1963 case VDEV_STATE_DEGRADED:
1964 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1965 zc->zc_obj != VDEV_AUX_EXTERNAL)
1966 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1968 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1972 error = SET_ERROR(EINVAL);
1974 zc->zc_cookie = newstate;
1975 spa_close(spa, FTAG);
1980 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1983 int replacing = zc->zc_cookie;
1987 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1990 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1991 zc->zc_iflags, &config)) == 0) {
1992 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1993 nvlist_free(config);
1996 spa_close(spa, FTAG);
2001 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
2006 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2009 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
2011 spa_close(spa, FTAG);
2016 zfs_ioc_vdev_split(zfs_cmd_t *zc)
2019 nvlist_t *config, *props = NULL;
2021 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
2023 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2026 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
2027 zc->zc_iflags, &config)) {
2028 spa_close(spa, FTAG);
2032 if (zc->zc_nvlist_src_size != 0 && (error =
2033 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2034 zc->zc_iflags, &props))) {
2035 spa_close(spa, FTAG);
2036 nvlist_free(config);
2040 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
2042 spa_close(spa, FTAG);
2044 nvlist_free(config);
2051 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
2054 char *path = zc->zc_value;
2055 uint64_t guid = zc->zc_guid;
2058 error = spa_open(zc->zc_name, &spa, FTAG);
2062 error = spa_vdev_setpath(spa, guid, path);
2063 spa_close(spa, FTAG);
2068 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
2071 char *fru = zc->zc_value;
2072 uint64_t guid = zc->zc_guid;
2075 error = spa_open(zc->zc_name, &spa, FTAG);
2079 error = spa_vdev_setfru(spa, guid, fru);
2080 spa_close(spa, FTAG);
2085 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
2090 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2092 if (zc->zc_nvlist_dst != 0 &&
2093 (error = dsl_prop_get_all(os, &nv)) == 0) {
2094 dmu_objset_stats(os, nv);
2096 * NB: zvol_get_stats() will read the objset contents,
2097 * which we aren't supposed to do with a
2098 * DS_MODE_USER hold, because it could be
2099 * inconsistent. So this is a bit of a workaround...
2100 * XXX reading with out owning
2102 if (!zc->zc_objset_stats.dds_inconsistent &&
2103 dmu_objset_type(os) == DMU_OST_ZVOL) {
2104 error = zvol_get_stats(os, nv);
2109 error = put_nvlist(zc, nv);
2118 * zc_name name of filesystem
2119 * zc_nvlist_dst_size size of buffer for property nvlist
2122 * zc_objset_stats stats
2123 * zc_nvlist_dst property nvlist
2124 * zc_nvlist_dst_size size of property nvlist
2127 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2132 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2134 error = zfs_ioc_objset_stats_impl(zc, os);
2135 dmu_objset_rele(os, FTAG);
2138 if (error == ENOMEM)
2145 * zc_name name of filesystem
2146 * zc_nvlist_dst_size size of buffer for property nvlist
2149 * zc_nvlist_dst received property nvlist
2150 * zc_nvlist_dst_size size of received property nvlist
2152 * Gets received properties (distinct from local properties on or after
2153 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2154 * local property values.
2157 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2163 * Without this check, we would return local property values if the
2164 * caller has not already received properties on or after
2165 * SPA_VERSION_RECVD_PROPS.
2167 if (!dsl_prop_get_hasrecvd(zc->zc_name))
2168 return (SET_ERROR(ENOTSUP));
2170 if (zc->zc_nvlist_dst != 0 &&
2171 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2172 error = put_nvlist(zc, nv);
2180 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2186 * zfs_get_zplprop() will either find a value or give us
2187 * the default value (if there is one).
2189 if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2191 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2197 * zc_name name of filesystem
2198 * zc_nvlist_dst_size size of buffer for zpl property nvlist
2201 * zc_nvlist_dst zpl property nvlist
2202 * zc_nvlist_dst_size size of zpl property nvlist
2205 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2210 /* XXX reading without owning */
2211 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os))
2214 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2217 * NB: nvl_add_zplprop() will read the objset contents,
2218 * which we aren't supposed to do with a DS_MODE_USER
2219 * hold, because it could be inconsistent.
2221 if (zc->zc_nvlist_dst != 0 &&
2222 !zc->zc_objset_stats.dds_inconsistent &&
2223 dmu_objset_type(os) == DMU_OST_ZFS) {
2226 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2227 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2228 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2229 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2230 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2231 err = put_nvlist(zc, nv);
2234 err = SET_ERROR(ENOENT);
2236 dmu_objset_rele(os, FTAG);
2241 dataset_name_hidden(const char *name)
2244 * Skip over datasets that are not visible in this zone,
2245 * internal datasets (which have a $ in their name), and
2246 * temporary datasets (which have a % in their name).
2248 if (strchr(name, '$') != NULL)
2250 if (strchr(name, '%') != NULL)
2252 if (!INGLOBALZONE(curthread) && !zone_dataset_visible(name, NULL))
2259 * zc_name name of filesystem
2260 * zc_cookie zap cursor
2261 * zc_nvlist_dst_size size of buffer for property nvlist
2264 * zc_name name of next filesystem
2265 * zc_cookie zap cursor
2266 * zc_objset_stats stats
2267 * zc_nvlist_dst property nvlist
2268 * zc_nvlist_dst_size size of property nvlist
2271 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2276 size_t orig_len = strlen(zc->zc_name);
2279 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) {
2280 if (error == ENOENT)
2281 error = SET_ERROR(ESRCH);
2285 p = strrchr(zc->zc_name, '/');
2286 if (p == NULL || p[1] != '\0')
2287 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2288 p = zc->zc_name + strlen(zc->zc_name);
2291 error = dmu_dir_list_next(os,
2292 sizeof (zc->zc_name) - (p - zc->zc_name), p,
2293 NULL, &zc->zc_cookie);
2294 if (error == ENOENT)
2295 error = SET_ERROR(ESRCH);
2296 } while (error == 0 && dataset_name_hidden(zc->zc_name));
2297 dmu_objset_rele(os, FTAG);
2300 * If it's an internal dataset (ie. with a '$' in its name),
2301 * don't try to get stats for it, otherwise we'll return ENOENT.
2303 if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2304 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2305 if (error == ENOENT) {
2306 /* We lost a race with destroy, get the next one. */
2307 zc->zc_name[orig_len] = '\0';
2316 * zc_name name of filesystem
2317 * zc_cookie zap cursor
2318 * zc_nvlist_dst_size size of buffer for property nvlist
2319 * zc_simple when set, only name is requested
2322 * zc_name name of next snapshot
2323 * zc_objset_stats stats
2324 * zc_nvlist_dst property nvlist
2325 * zc_nvlist_dst_size size of property nvlist
2328 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2333 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2335 return (error == ENOENT ? ESRCH : error);
2339 * A dataset name of maximum length cannot have any snapshots,
2340 * so exit immediately.
2342 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
2343 dmu_objset_rele(os, FTAG);
2344 return (SET_ERROR(ESRCH));
2347 error = dmu_snapshot_list_next(os,
2348 sizeof (zc->zc_name) - strlen(zc->zc_name),
2349 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2352 if (error == 0 && !zc->zc_simple) {
2354 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2356 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2360 error = dmu_objset_from_ds(ds, &ossnap);
2362 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2363 dsl_dataset_rele(ds, FTAG);
2365 } else if (error == ENOENT) {
2366 error = SET_ERROR(ESRCH);
2369 dmu_objset_rele(os, FTAG);
2370 /* if we failed, undo the @ that we tacked on to zc_name */
2372 *strchr(zc->zc_name, '@') = '\0';
2377 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2379 const char *propname = nvpair_name(pair);
2381 unsigned int vallen;
2384 zfs_userquota_prop_t type;
2390 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2392 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2393 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2395 return (SET_ERROR(EINVAL));
2399 * A correctly constructed propname is encoded as
2400 * userquota@<rid>-<domain>.
2402 if ((dash = strchr(propname, '-')) == NULL ||
2403 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2405 return (SET_ERROR(EINVAL));
2412 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2414 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2415 zfsvfs_rele(zfsvfs, FTAG);
2422 * If the named property is one that has a special function to set its value,
2423 * return 0 on success and a positive error code on failure; otherwise if it is
2424 * not one of the special properties handled by this function, return -1.
2426 * XXX: It would be better for callers of the property interface if we handled
2427 * these special cases in dsl_prop.c (in the dsl layer).
2430 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2433 const char *propname = nvpair_name(pair);
2434 zfs_prop_t prop = zfs_name_to_prop(propname);
2438 if (prop == ZPROP_INVAL) {
2439 if (zfs_prop_userquota(propname))
2440 return (zfs_prop_set_userquota(dsname, pair));
2444 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2446 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2447 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2451 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2454 VERIFY(0 == nvpair_value_uint64(pair, &intval));
2457 case ZFS_PROP_QUOTA:
2458 err = dsl_dir_set_quota(dsname, source, intval);
2460 case ZFS_PROP_REFQUOTA:
2461 err = dsl_dataset_set_refquota(dsname, source, intval);
2463 case ZFS_PROP_FILESYSTEM_LIMIT:
2464 case ZFS_PROP_SNAPSHOT_LIMIT:
2465 if (intval == UINT64_MAX) {
2466 /* clearing the limit, just do it */
2469 err = dsl_dir_activate_fs_ss_limit(dsname);
2472 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2473 * default path to set the value in the nvlist.
2478 case ZFS_PROP_RESERVATION:
2479 err = dsl_dir_set_reservation(dsname, source, intval);
2481 case ZFS_PROP_REFRESERVATION:
2482 err = dsl_dataset_set_refreservation(dsname, source, intval);
2484 case ZFS_PROP_VOLSIZE:
2485 err = zvol_set_volsize(dsname, intval);
2487 case ZFS_PROP_VERSION:
2491 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2494 err = zfs_set_version(zfsvfs, intval);
2495 zfsvfs_rele(zfsvfs, FTAG);
2497 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2500 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2501 (void) strcpy(zc->zc_name, dsname);
2502 (void) zfs_ioc_userspace_upgrade(zc);
2503 kmem_free(zc, sizeof (zfs_cmd_t));
2515 * This function is best effort. If it fails to set any of the given properties,
2516 * it continues to set as many as it can and returns the last error
2517 * encountered. If the caller provides a non-NULL errlist, it will be filled in
2518 * with the list of names of all the properties that failed along with the
2519 * corresponding error numbers.
2521 * If every property is set successfully, zero is returned and errlist is not
2525 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2533 nvlist_t *genericnvl = fnvlist_alloc();
2534 nvlist_t *retrynvl = fnvlist_alloc();
2538 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2539 const char *propname = nvpair_name(pair);
2540 zfs_prop_t prop = zfs_name_to_prop(propname);
2543 /* decode the property value */
2545 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2547 attrs = fnvpair_value_nvlist(pair);
2548 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2550 err = SET_ERROR(EINVAL);
2553 /* Validate value type */
2554 if (err == 0 && prop == ZPROP_INVAL) {
2555 if (zfs_prop_user(propname)) {
2556 if (nvpair_type(propval) != DATA_TYPE_STRING)
2557 err = SET_ERROR(EINVAL);
2558 } else if (zfs_prop_userquota(propname)) {
2559 if (nvpair_type(propval) !=
2560 DATA_TYPE_UINT64_ARRAY)
2561 err = SET_ERROR(EINVAL);
2563 err = SET_ERROR(EINVAL);
2565 } else if (err == 0) {
2566 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2567 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2568 err = SET_ERROR(EINVAL);
2569 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2572 intval = fnvpair_value_uint64(propval);
2574 switch (zfs_prop_get_type(prop)) {
2575 case PROP_TYPE_NUMBER:
2577 case PROP_TYPE_STRING:
2578 err = SET_ERROR(EINVAL);
2580 case PROP_TYPE_INDEX:
2581 if (zfs_prop_index_to_string(prop,
2582 intval, &unused) != 0)
2583 err = SET_ERROR(EINVAL);
2587 "unknown property type");
2590 err = SET_ERROR(EINVAL);
2594 /* Validate permissions */
2596 err = zfs_check_settable(dsname, pair, CRED());
2599 err = zfs_prop_set_special(dsname, source, pair);
2602 * For better performance we build up a list of
2603 * properties to set in a single transaction.
2605 err = nvlist_add_nvpair(genericnvl, pair);
2606 } else if (err != 0 && nvl != retrynvl) {
2608 * This may be a spurious error caused by
2609 * receiving quota and reservation out of order.
2610 * Try again in a second pass.
2612 err = nvlist_add_nvpair(retrynvl, pair);
2617 if (errlist != NULL)
2618 fnvlist_add_int32(errlist, propname, err);
2623 if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2628 if (!nvlist_empty(genericnvl) &&
2629 dsl_props_set(dsname, source, genericnvl) != 0) {
2631 * If this fails, we still want to set as many properties as we
2632 * can, so try setting them individually.
2635 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2636 const char *propname = nvpair_name(pair);
2640 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2642 attrs = fnvpair_value_nvlist(pair);
2643 propval = fnvlist_lookup_nvpair(attrs,
2647 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2648 strval = fnvpair_value_string(propval);
2649 err = dsl_prop_set_string(dsname, propname,
2652 intval = fnvpair_value_uint64(propval);
2653 err = dsl_prop_set_int(dsname, propname, source,
2658 if (errlist != NULL) {
2659 fnvlist_add_int32(errlist, propname,
2666 nvlist_free(genericnvl);
2667 nvlist_free(retrynvl);
2673 * Check that all the properties are valid user properties.
2676 zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2678 nvpair_t *pair = NULL;
2681 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2682 const char *propname = nvpair_name(pair);
2684 if (!zfs_prop_user(propname) ||
2685 nvpair_type(pair) != DATA_TYPE_STRING)
2686 return (SET_ERROR(EINVAL));
2688 if (error = zfs_secpolicy_write_perms(fsname,
2689 ZFS_DELEG_PERM_USERPROP, CRED()))
2692 if (strlen(propname) >= ZAP_MAXNAMELEN)
2693 return (SET_ERROR(ENAMETOOLONG));
2695 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN)
2702 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2706 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2709 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2710 if (nvlist_exists(skipped, nvpair_name(pair)))
2713 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2718 clear_received_props(const char *dsname, nvlist_t *props,
2722 nvlist_t *cleared_props = NULL;
2723 props_skip(props, skipped, &cleared_props);
2724 if (!nvlist_empty(cleared_props)) {
2726 * Acts on local properties until the dataset has received
2727 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2729 zprop_source_t flags = (ZPROP_SRC_NONE |
2730 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2731 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2733 nvlist_free(cleared_props);
2739 * zc_name name of filesystem
2740 * zc_value name of property to set
2741 * zc_nvlist_src{_size} nvlist of properties to apply
2742 * zc_cookie received properties flag
2745 * zc_nvlist_dst{_size} error for each unapplied received property
2748 zfs_ioc_set_prop(zfs_cmd_t *zc)
2751 boolean_t received = zc->zc_cookie;
2752 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2757 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2758 zc->zc_iflags, &nvl)) != 0)
2762 nvlist_t *origprops;
2764 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2765 (void) clear_received_props(zc->zc_name,
2767 nvlist_free(origprops);
2770 error = dsl_prop_set_hasrecvd(zc->zc_name);
2773 errors = fnvlist_alloc();
2775 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2777 if (zc->zc_nvlist_dst != 0 && errors != NULL) {
2778 (void) put_nvlist(zc, errors);
2781 nvlist_free(errors);
2788 * zc_name name of filesystem
2789 * zc_value name of property to inherit
2790 * zc_cookie revert to received value if TRUE
2795 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2797 const char *propname = zc->zc_value;
2798 zfs_prop_t prop = zfs_name_to_prop(propname);
2799 boolean_t received = zc->zc_cookie;
2800 zprop_source_t source = (received
2801 ? ZPROP_SRC_NONE /* revert to received value, if any */
2802 : ZPROP_SRC_INHERITED); /* explicitly inherit */
2811 * zfs_prop_set_special() expects properties in the form of an
2812 * nvpair with type info.
2814 if (prop == ZPROP_INVAL) {
2815 if (!zfs_prop_user(propname))
2816 return (SET_ERROR(EINVAL));
2818 type = PROP_TYPE_STRING;
2819 } else if (prop == ZFS_PROP_VOLSIZE ||
2820 prop == ZFS_PROP_VERSION) {
2821 return (SET_ERROR(EINVAL));
2823 type = zfs_prop_get_type(prop);
2826 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2829 case PROP_TYPE_STRING:
2830 VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2832 case PROP_TYPE_NUMBER:
2833 case PROP_TYPE_INDEX:
2834 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2838 return (SET_ERROR(EINVAL));
2841 pair = nvlist_next_nvpair(dummy, NULL);
2842 err = zfs_prop_set_special(zc->zc_name, source, pair);
2845 return (err); /* special property already handled */
2848 * Only check this in the non-received case. We want to allow
2849 * 'inherit -S' to revert non-inheritable properties like quota
2850 * and reservation to the received or default values even though
2851 * they are not considered inheritable.
2853 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2854 return (SET_ERROR(EINVAL));
2857 /* property name has been validated by zfs_secpolicy_inherit_prop() */
2858 return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source));
2862 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2869 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2870 zc->zc_iflags, &props))
2874 * If the only property is the configfile, then just do a spa_lookup()
2875 * to handle the faulted case.
2877 pair = nvlist_next_nvpair(props, NULL);
2878 if (pair != NULL && strcmp(nvpair_name(pair),
2879 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2880 nvlist_next_nvpair(props, pair) == NULL) {
2881 mutex_enter(&spa_namespace_lock);
2882 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2883 spa_configfile_set(spa, props, B_FALSE);
2884 spa_config_sync(spa, B_FALSE, B_TRUE);
2886 mutex_exit(&spa_namespace_lock);
2893 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2898 error = spa_prop_set(spa, props);
2901 spa_close(spa, FTAG);
2907 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2911 nvlist_t *nvp = NULL;
2913 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2915 * If the pool is faulted, there may be properties we can still
2916 * get (such as altroot and cachefile), so attempt to get them
2919 mutex_enter(&spa_namespace_lock);
2920 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2921 error = spa_prop_get(spa, &nvp);
2922 mutex_exit(&spa_namespace_lock);
2924 error = spa_prop_get(spa, &nvp);
2925 spa_close(spa, FTAG);
2928 if (error == 0 && zc->zc_nvlist_dst != 0)
2929 error = put_nvlist(zc, nvp);
2931 error = SET_ERROR(EFAULT);
2939 * zc_name name of filesystem
2940 * zc_nvlist_src{_size} nvlist of delegated permissions
2941 * zc_perm_action allow/unallow flag
2946 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2949 nvlist_t *fsaclnv = NULL;
2951 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2952 zc->zc_iflags, &fsaclnv)) != 0)
2956 * Verify nvlist is constructed correctly
2958 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2959 nvlist_free(fsaclnv);
2960 return (SET_ERROR(EINVAL));
2964 * If we don't have PRIV_SYS_MOUNT, then validate
2965 * that user is allowed to hand out each permission in
2969 error = secpolicy_zfs(CRED());
2971 if (zc->zc_perm_action == B_FALSE) {
2972 error = dsl_deleg_can_allow(zc->zc_name,
2975 error = dsl_deleg_can_unallow(zc->zc_name,
2981 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2983 nvlist_free(fsaclnv);
2989 * zc_name name of filesystem
2992 * zc_nvlist_src{_size} nvlist of delegated permissions
2995 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
3000 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
3001 error = put_nvlist(zc, nvp);
3009 * Search the vfs list for a specified resource. Returns a pointer to it
3010 * or NULL if no suitable entry is found. The caller of this routine
3011 * is responsible for releasing the returned vfs pointer.
3014 zfs_get_vfs(const char *resource)
3018 mtx_lock(&mountlist_mtx);
3019 TAILQ_FOREACH(vfsp, &mountlist, mnt_list) {
3020 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
3025 mtx_unlock(&mountlist_mtx);
3031 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
3033 zfs_creat_t *zct = arg;
3035 zfs_create_fs(os, cr, zct->zct_zplprops, tx);
3038 #define ZFS_PROP_UNDEFINED ((uint64_t)-1)
3042 * os parent objset pointer (NULL if root fs)
3043 * fuids_ok fuids allowed in this version of the spa?
3044 * sa_ok SAs allowed in this version of the spa?
3045 * createprops list of properties requested by creator
3048 * zplprops values for the zplprops we attach to the master node object
3049 * is_ci true if requested file system will be purely case-insensitive
3051 * Determine the settings for utf8only, normalization and
3052 * casesensitivity. Specific values may have been requested by the
3053 * creator and/or we can inherit values from the parent dataset. If
3054 * the file system is of too early a vintage, a creator can not
3055 * request settings for these properties, even if the requested
3056 * setting is the default value. We don't actually want to create dsl
3057 * properties for these, so remove them from the source nvlist after
3061 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
3062 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
3063 nvlist_t *zplprops, boolean_t *is_ci)
3065 uint64_t sense = ZFS_PROP_UNDEFINED;
3066 uint64_t norm = ZFS_PROP_UNDEFINED;
3067 uint64_t u8 = ZFS_PROP_UNDEFINED;
3069 ASSERT(zplprops != NULL);
3072 * Pull out creator prop choices, if any.
3075 (void) nvlist_lookup_uint64(createprops,
3076 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
3077 (void) nvlist_lookup_uint64(createprops,
3078 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
3079 (void) nvlist_remove_all(createprops,
3080 zfs_prop_to_name(ZFS_PROP_NORMALIZE));
3081 (void) nvlist_lookup_uint64(createprops,
3082 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
3083 (void) nvlist_remove_all(createprops,
3084 zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
3085 (void) nvlist_lookup_uint64(createprops,
3086 zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3087 (void) nvlist_remove_all(createprops,
3088 zfs_prop_to_name(ZFS_PROP_CASE));
3092 * If the zpl version requested is whacky or the file system
3093 * or pool is version is too "young" to support normalization
3094 * and the creator tried to set a value for one of the props,
3097 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3098 (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3099 (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3100 (zplver < ZPL_VERSION_NORMALIZATION &&
3101 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3102 sense != ZFS_PROP_UNDEFINED)))
3103 return (SET_ERROR(ENOTSUP));
3106 * Put the version in the zplprops
3108 VERIFY(nvlist_add_uint64(zplprops,
3109 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3111 if (norm == ZFS_PROP_UNDEFINED)
3112 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
3113 VERIFY(nvlist_add_uint64(zplprops,
3114 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3117 * If we're normalizing, names must always be valid UTF-8 strings.
3121 if (u8 == ZFS_PROP_UNDEFINED)
3122 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
3123 VERIFY(nvlist_add_uint64(zplprops,
3124 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3126 if (sense == ZFS_PROP_UNDEFINED)
3127 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
3128 VERIFY(nvlist_add_uint64(zplprops,
3129 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3132 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
3138 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3139 nvlist_t *zplprops, boolean_t *is_ci)
3141 boolean_t fuids_ok, sa_ok;
3142 uint64_t zplver = ZPL_VERSION;
3143 objset_t *os = NULL;
3144 char parentname[MAXNAMELEN];
3150 (void) strlcpy(parentname, dataset, sizeof (parentname));
3151 cp = strrchr(parentname, '/');
3155 if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3158 spa_vers = spa_version(spa);
3159 spa_close(spa, FTAG);
3161 zplver = zfs_zpl_version_map(spa_vers);
3162 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3163 sa_ok = (zplver >= ZPL_VERSION_SA);
3166 * Open parent object set so we can inherit zplprop values.
3168 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3171 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3173 dmu_objset_rele(os, FTAG);
3178 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3179 nvlist_t *zplprops, boolean_t *is_ci)
3183 uint64_t zplver = ZPL_VERSION;
3186 zplver = zfs_zpl_version_map(spa_vers);
3187 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3188 sa_ok = (zplver >= ZPL_VERSION_SA);
3190 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3191 createprops, zplprops, is_ci);
3197 * "type" -> dmu_objset_type_t (int32)
3198 * (optional) "props" -> { prop -> value }
3201 * outnvl: propname -> error code (int32)
3204 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3207 zfs_creat_t zct = { 0 };
3208 nvlist_t *nvprops = NULL;
3209 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3211 dmu_objset_type_t type;
3212 boolean_t is_insensitive = B_FALSE;
3214 if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3215 return (SET_ERROR(EINVAL));
3217 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3221 cbfunc = zfs_create_cb;
3225 cbfunc = zvol_create_cb;
3232 if (strchr(fsname, '@') ||
3233 strchr(fsname, '%'))
3234 return (SET_ERROR(EINVAL));
3236 zct.zct_props = nvprops;
3239 return (SET_ERROR(EINVAL));
3241 if (type == DMU_OST_ZVOL) {
3242 uint64_t volsize, volblocksize;
3244 if (nvprops == NULL)
3245 return (SET_ERROR(EINVAL));
3246 if (nvlist_lookup_uint64(nvprops,
3247 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3248 return (SET_ERROR(EINVAL));
3250 if ((error = nvlist_lookup_uint64(nvprops,
3251 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3252 &volblocksize)) != 0 && error != ENOENT)
3253 return (SET_ERROR(EINVAL));
3256 volblocksize = zfs_prop_default_numeric(
3257 ZFS_PROP_VOLBLOCKSIZE);
3259 if ((error = zvol_check_volblocksize(
3260 volblocksize)) != 0 ||
3261 (error = zvol_check_volsize(volsize,
3262 volblocksize)) != 0)
3264 } else if (type == DMU_OST_ZFS) {
3268 * We have to have normalization and
3269 * case-folding flags correct when we do the
3270 * file system creation, so go figure them out
3273 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3274 NV_UNIQUE_NAME, KM_SLEEP) == 0);
3275 error = zfs_fill_zplprops(fsname, nvprops,
3276 zct.zct_zplprops, &is_insensitive);
3278 nvlist_free(zct.zct_zplprops);
3283 error = dmu_objset_create(fsname, type,
3284 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3285 nvlist_free(zct.zct_zplprops);
3288 * It would be nice to do this atomically.
3291 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3294 (void) dsl_destroy_head(fsname);
3297 if (error == 0 && type == DMU_OST_ZVOL)
3298 zvol_create_minors(fsname);
3305 * "origin" -> name of origin snapshot
3306 * (optional) "props" -> { prop -> value }
3309 * outnvl: propname -> error code (int32)
3312 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3315 nvlist_t *nvprops = NULL;
3318 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3319 return (SET_ERROR(EINVAL));
3320 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3322 if (strchr(fsname, '@') ||
3323 strchr(fsname, '%'))
3324 return (SET_ERROR(EINVAL));
3326 if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3327 return (SET_ERROR(EINVAL));
3328 error = dmu_objset_clone(fsname, origin_name);
3333 * It would be nice to do this atomically.
3336 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3339 (void) dsl_destroy_head(fsname);
3343 zvol_create_minors(fsname);
3350 * "snaps" -> { snapshot1, snapshot2 }
3351 * (optional) "props" -> { prop -> value (string) }
3354 * outnvl: snapshot -> error code (int32)
3357 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3360 nvlist_t *props = NULL;
3364 (void) nvlist_lookup_nvlist(innvl, "props", &props);
3365 if ((error = zfs_check_userprops(poolname, props)) != 0)
3368 if (!nvlist_empty(props) &&
3369 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3370 return (SET_ERROR(ENOTSUP));
3372 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3373 return (SET_ERROR(EINVAL));
3374 poollen = strlen(poolname);
3375 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3376 pair = nvlist_next_nvpair(snaps, pair)) {
3377 const char *name = nvpair_name(pair);
3378 const char *cp = strchr(name, '@');
3381 * The snap name must contain an @, and the part after it must
3382 * contain only valid characters.
3385 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3386 return (SET_ERROR(EINVAL));
3389 * The snap must be in the specified pool.
3391 if (strncmp(name, poolname, poollen) != 0 ||
3392 (name[poollen] != '/' && name[poollen] != '@'))
3393 return (SET_ERROR(EXDEV));
3395 /* This must be the only snap of this fs. */
3396 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3397 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3398 if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3400 return (SET_ERROR(EXDEV));
3405 error = dsl_dataset_snapshot(snaps, props, outnvl);
3410 * innvl: "message" -> string
3414 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3422 * The poolname in the ioctl is not set, we get it from the TSD,
3423 * which was set at the end of the last successful ioctl that allows
3424 * logging. The secpolicy func already checked that it is set.
3425 * Only one log ioctl is allowed after each successful ioctl, so
3426 * we clear the TSD here.
3428 poolname = tsd_get(zfs_allow_log_key);
3429 (void) tsd_set(zfs_allow_log_key, NULL);
3430 error = spa_open(poolname, &spa, FTAG);
3435 if (nvlist_lookup_string(innvl, "message", &message) != 0) {
3436 spa_close(spa, FTAG);
3437 return (SET_ERROR(EINVAL));
3440 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3441 spa_close(spa, FTAG);
3442 return (SET_ERROR(ENOTSUP));
3445 error = spa_history_log(spa, message);
3446 spa_close(spa, FTAG);
3451 * The dp_config_rwlock must not be held when calling this, because the
3452 * unmount may need to write out data.
3454 * This function is best-effort. Callers must deal gracefully if it
3455 * remains mounted (or is remounted after this call).
3457 * Returns 0 if the argument is not a snapshot, or it is not currently a
3458 * filesystem, or we were able to unmount it. Returns error code otherwise.
3461 zfs_unmount_snap(const char *snapname)
3467 if (strchr(snapname, '@') == NULL)
3470 vfsp = zfs_get_vfs(snapname);
3474 zfsvfs = vfsp->vfs_data;
3475 ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os)));
3477 err = vn_vfswlock(vfsp->vfs_vnodecovered);
3480 return (SET_ERROR(err));
3483 * Always force the unmount for snapshots.
3487 (void) dounmount(vfsp, MS_FORCE, kcred);
3490 (void) dounmount(vfsp, MS_FORCE, curthread);
3497 zfs_unmount_snap_cb(const char *snapname, void *arg)
3499 return (zfs_unmount_snap(snapname));
3503 * When a clone is destroyed, its origin may also need to be destroyed,
3504 * in which case it must be unmounted. This routine will do that unmount
3508 zfs_destroy_unmount_origin(const char *fsname)
3514 error = dmu_objset_hold(fsname, FTAG, &os);
3517 ds = dmu_objset_ds(os);
3518 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3519 char originname[MAXNAMELEN];
3520 dsl_dataset_name(ds->ds_prev, originname);
3521 dmu_objset_rele(os, FTAG);
3522 (void) zfs_unmount_snap(originname);
3524 dmu_objset_rele(os, FTAG);
3530 * "snaps" -> { snapshot1, snapshot2 }
3531 * (optional boolean) "defer"
3534 * outnvl: snapshot -> error code (int32)
3539 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3546 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3547 return (SET_ERROR(EINVAL));
3548 defer = nvlist_exists(innvl, "defer");
3550 poollen = strlen(poolname);
3551 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3552 pair = nvlist_next_nvpair(snaps, pair)) {
3553 const char *name = nvpair_name(pair);
3556 * The snap must be in the specified pool to prevent the
3557 * invalid removal of zvol minors below.
3559 if (strncmp(name, poolname, poollen) != 0 ||
3560 (name[poollen] != '/' && name[poollen] != '@'))
3561 return (SET_ERROR(EXDEV));
3563 error = zfs_unmount_snap(name);
3566 #if defined(__FreeBSD__)
3567 zvol_remove_minors(name);
3571 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3575 * Create bookmarks. Bookmark names are of the form <fs>#<bmark>.
3576 * All bookmarks must be in the same pool.
3579 * bookmark1 -> snapshot1, bookmark2 -> snapshot2
3582 * outnvl: bookmark -> error code (int32)
3587 zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3589 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3590 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3594 * Verify the snapshot argument.
3596 if (nvpair_value_string(pair, &snap_name) != 0)
3597 return (SET_ERROR(EINVAL));
3600 /* Verify that the keys (bookmarks) are unique */
3601 for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair);
3602 pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) {
3603 if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0)
3604 return (SET_ERROR(EINVAL));
3608 return (dsl_bookmark_create(innvl, outnvl));
3613 * property 1, property 2, ...
3617 * bookmark name 1 -> { property 1, property 2, ... },
3618 * bookmark name 2 -> { property 1, property 2, ... }
3623 zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3625 return (dsl_get_bookmarks(fsname, innvl, outnvl));
3630 * bookmark name 1, bookmark name 2
3633 * outnvl: bookmark -> error code (int32)
3637 zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl,
3642 poollen = strlen(poolname);
3643 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3644 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3645 const char *name = nvpair_name(pair);
3646 const char *cp = strchr(name, '#');
3649 * The bookmark name must contain an #, and the part after it
3650 * must contain only valid characters.
3653 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3654 return (SET_ERROR(EINVAL));
3657 * The bookmark must be in the specified pool.
3659 if (strncmp(name, poolname, poollen) != 0 ||
3660 (name[poollen] != '/' && name[poollen] != '#'))
3661 return (SET_ERROR(EXDEV));
3664 error = dsl_bookmark_destroy(innvl, outnvl);
3670 * zc_name name of dataset to destroy
3671 * zc_objset_type type of objset
3672 * zc_defer_destroy mark for deferred destroy
3677 zfs_ioc_destroy(zfs_cmd_t *zc)
3681 if (zc->zc_objset_type == DMU_OST_ZFS) {
3682 err = zfs_unmount_snap(zc->zc_name);
3687 if (strchr(zc->zc_name, '@'))
3688 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3690 err = dsl_destroy_head(zc->zc_name);
3691 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3693 zvol_remove_minors(zc->zc_name);
3695 (void) zvol_remove_minor(zc->zc_name);
3701 * fsname is name of dataset to rollback (to most recent snapshot)
3703 * innvl is not used.
3705 * outnvl: "target" -> name of most recent snapshot
3710 zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl)
3715 if (getzfsvfs(fsname, &zfsvfs) == 0) {
3716 error = zfs_suspend_fs(zfsvfs);
3720 error = dsl_dataset_rollback(fsname, zfsvfs, outnvl);
3721 resume_err = zfs_resume_fs(zfsvfs, fsname);
3722 error = error ? error : resume_err;
3724 VFS_RELE(zfsvfs->z_vfs);
3726 error = dsl_dataset_rollback(fsname, NULL, outnvl);
3732 recursive_unmount(const char *fsname, void *arg)
3734 const char *snapname = arg;
3735 char fullname[MAXNAMELEN];
3737 (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname);
3738 return (zfs_unmount_snap(fullname));
3743 * zc_name old name of dataset
3744 * zc_value new name of dataset
3745 * zc_cookie recursive flag (only valid for snapshots)
3750 zfs_ioc_rename(zfs_cmd_t *zc)
3752 boolean_t recursive = zc->zc_cookie & 1;
3754 boolean_t allow_mounted = B_TRUE;
3757 allow_mounted = (zc->zc_cookie & 2) != 0;
3760 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3761 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3762 strchr(zc->zc_value, '%'))
3763 return (SET_ERROR(EINVAL));
3765 at = strchr(zc->zc_name, '@');
3767 /* snaps must be in same fs */
3770 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
3771 return (SET_ERROR(EXDEV));
3773 if (zc->zc_objset_type == DMU_OST_ZFS && allow_mounted) {
3774 error = dmu_objset_find(zc->zc_name,
3775 recursive_unmount, at + 1,
3776 recursive ? DS_FIND_CHILDREN : 0);
3782 error = dsl_dataset_rename_snapshot(zc->zc_name,
3783 at + 1, strchr(zc->zc_value, '@') + 1, recursive);
3789 if (zc->zc_objset_type == DMU_OST_ZVOL)
3790 (void) zvol_remove_minor(zc->zc_name);
3792 return (dsl_dir_rename(zc->zc_name, zc->zc_value));
3797 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3799 const char *propname = nvpair_name(pair);
3800 boolean_t issnap = (strchr(dsname, '@') != NULL);
3801 zfs_prop_t prop = zfs_name_to_prop(propname);
3805 if (prop == ZPROP_INVAL) {
3806 if (zfs_prop_user(propname)) {
3807 if (err = zfs_secpolicy_write_perms(dsname,
3808 ZFS_DELEG_PERM_USERPROP, cr))
3813 if (!issnap && zfs_prop_userquota(propname)) {
3814 const char *perm = NULL;
3815 const char *uq_prefix =
3816 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3817 const char *gq_prefix =
3818 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3820 if (strncmp(propname, uq_prefix,
3821 strlen(uq_prefix)) == 0) {
3822 perm = ZFS_DELEG_PERM_USERQUOTA;
3823 } else if (strncmp(propname, gq_prefix,
3824 strlen(gq_prefix)) == 0) {
3825 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3827 /* USERUSED and GROUPUSED are read-only */
3828 return (SET_ERROR(EINVAL));
3831 if (err = zfs_secpolicy_write_perms(dsname, perm, cr))
3836 return (SET_ERROR(EINVAL));
3840 return (SET_ERROR(EINVAL));
3842 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3844 * dsl_prop_get_all_impl() returns properties in this
3848 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3849 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3854 * Check that this value is valid for this pool version
3857 case ZFS_PROP_COMPRESSION:
3859 * If the user specified gzip compression, make sure
3860 * the SPA supports it. We ignore any errors here since
3861 * we'll catch them later.
3863 if (nvpair_value_uint64(pair, &intval) == 0) {
3864 if (intval >= ZIO_COMPRESS_GZIP_1 &&
3865 intval <= ZIO_COMPRESS_GZIP_9 &&
3866 zfs_earlier_version(dsname,
3867 SPA_VERSION_GZIP_COMPRESSION)) {
3868 return (SET_ERROR(ENOTSUP));
3871 if (intval == ZIO_COMPRESS_ZLE &&
3872 zfs_earlier_version(dsname,
3873 SPA_VERSION_ZLE_COMPRESSION))
3874 return (SET_ERROR(ENOTSUP));
3876 if (intval == ZIO_COMPRESS_LZ4) {
3879 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3882 if (!spa_feature_is_enabled(spa,
3883 SPA_FEATURE_LZ4_COMPRESS)) {
3884 spa_close(spa, FTAG);
3885 return (SET_ERROR(ENOTSUP));
3887 spa_close(spa, FTAG);
3891 * If this is a bootable dataset then
3892 * verify that the compression algorithm
3893 * is supported for booting. We must return
3894 * something other than ENOTSUP since it
3895 * implies a downrev pool version.
3897 if (zfs_is_bootfs(dsname) &&
3898 !BOOTFS_COMPRESS_VALID(intval)) {
3899 return (SET_ERROR(ERANGE));
3904 case ZFS_PROP_COPIES:
3905 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3906 return (SET_ERROR(ENOTSUP));
3909 case ZFS_PROP_RECORDSIZE:
3910 /* Record sizes above 128k need the feature to be enabled */
3911 if (nvpair_value_uint64(pair, &intval) == 0 &&
3912 intval > SPA_OLD_MAXBLOCKSIZE) {
3916 * If this is a bootable dataset then
3917 * the we don't allow large (>128K) blocks,
3918 * because GRUB doesn't support them.
3920 if (zfs_is_bootfs(dsname) &&
3921 intval > SPA_OLD_MAXBLOCKSIZE) {
3922 return (SET_ERROR(ERANGE));
3926 * We don't allow setting the property above 1MB,
3927 * unless the tunable has been changed.
3929 if (intval > zfs_max_recordsize ||
3930 intval > SPA_MAXBLOCKSIZE)
3931 return (SET_ERROR(ERANGE));
3933 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3936 if (!spa_feature_is_enabled(spa,
3937 SPA_FEATURE_LARGE_BLOCKS)) {
3938 spa_close(spa, FTAG);
3939 return (SET_ERROR(ENOTSUP));
3941 spa_close(spa, FTAG);
3945 case ZFS_PROP_SHARESMB:
3946 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3947 return (SET_ERROR(ENOTSUP));
3950 case ZFS_PROP_ACLINHERIT:
3951 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3952 nvpair_value_uint64(pair, &intval) == 0) {
3953 if (intval == ZFS_ACL_PASSTHROUGH_X &&
3954 zfs_earlier_version(dsname,
3955 SPA_VERSION_PASSTHROUGH_X))
3956 return (SET_ERROR(ENOTSUP));
3960 case ZFS_PROP_CHECKSUM:
3961 case ZFS_PROP_DEDUP:
3963 spa_feature_t feature;
3966 /* dedup feature version checks */
3967 if (prop == ZFS_PROP_DEDUP &&
3968 zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3969 return (SET_ERROR(ENOTSUP));
3971 if (nvpair_value_uint64(pair, &intval) != 0)
3972 return (SET_ERROR(EINVAL));
3974 /* check prop value is enabled in features */
3975 feature = zio_checksum_to_feature(intval);
3976 if (feature == SPA_FEATURE_NONE)
3979 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3982 * Salted checksums are not supported on root pools.
3984 if (spa_bootfs(spa) != 0 &&
3985 intval < ZIO_CHECKSUM_FUNCTIONS &&
3986 (zio_checksum_table[intval].ci_flags &
3987 ZCHECKSUM_FLAG_SALTED)) {
3988 spa_close(spa, FTAG);
3989 return (SET_ERROR(ERANGE));
3991 if (!spa_feature_is_enabled(spa, feature)) {
3992 spa_close(spa, FTAG);
3993 return (SET_ERROR(ENOTSUP));
3995 spa_close(spa, FTAG);
4000 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
4004 * Checks for a race condition to make sure we don't increment a feature flag
4008 zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx)
4010 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
4011 spa_feature_t *featurep = arg;
4013 if (!spa_feature_is_active(spa, *featurep))
4016 return (SET_ERROR(EBUSY));
4020 * The callback invoked on feature activation in the sync task caused by
4021 * zfs_prop_activate_feature.
4024 zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx)
4026 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
4027 spa_feature_t *featurep = arg;
4029 spa_feature_incr(spa, *featurep, tx);
4033 * Activates a feature on a pool in response to a property setting. This
4034 * creates a new sync task which modifies the pool to reflect the feature
4038 zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature)
4042 /* EBUSY here indicates that the feature is already active */
4043 err = dsl_sync_task(spa_name(spa),
4044 zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync,
4045 &feature, 2, ZFS_SPACE_CHECK_RESERVED);
4047 if (err != 0 && err != EBUSY)
4054 * Removes properties from the given props list that fail permission checks
4055 * needed to clear them and to restore them in case of a receive error. For each
4056 * property, make sure we have both set and inherit permissions.
4058 * Returns the first error encountered if any permission checks fail. If the
4059 * caller provides a non-NULL errlist, it also gives the complete list of names
4060 * of all the properties that failed a permission check along with the
4061 * corresponding error numbers. The caller is responsible for freeing the
4064 * If every property checks out successfully, zero is returned and the list
4065 * pointed at by errlist is NULL.
4068 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
4071 nvpair_t *pair, *next_pair;
4078 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4080 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
4081 (void) strcpy(zc->zc_name, dataset);
4082 pair = nvlist_next_nvpair(props, NULL);
4083 while (pair != NULL) {
4084 next_pair = nvlist_next_nvpair(props, pair);
4086 (void) strcpy(zc->zc_value, nvpair_name(pair));
4087 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
4088 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
4089 VERIFY(nvlist_remove_nvpair(props, pair) == 0);
4090 VERIFY(nvlist_add_int32(errors,
4091 zc->zc_value, err) == 0);
4095 kmem_free(zc, sizeof (zfs_cmd_t));
4097 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
4098 nvlist_free(errors);
4101 VERIFY(nvpair_value_int32(pair, &rv) == 0);
4104 if (errlist == NULL)
4105 nvlist_free(errors);
4113 propval_equals(nvpair_t *p1, nvpair_t *p2)
4115 if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
4116 /* dsl_prop_get_all_impl() format */
4118 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
4119 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4123 if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
4125 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
4126 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4130 if (nvpair_type(p1) != nvpair_type(p2))
4133 if (nvpair_type(p1) == DATA_TYPE_STRING) {
4134 char *valstr1, *valstr2;
4136 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
4137 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
4138 return (strcmp(valstr1, valstr2) == 0);
4140 uint64_t intval1, intval2;
4142 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
4143 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
4144 return (intval1 == intval2);
4149 * Remove properties from props if they are not going to change (as determined
4150 * by comparison with origprops). Remove them from origprops as well, since we
4151 * do not need to clear or restore properties that won't change.
4154 props_reduce(nvlist_t *props, nvlist_t *origprops)
4156 nvpair_t *pair, *next_pair;
4158 if (origprops == NULL)
4159 return; /* all props need to be received */
4161 pair = nvlist_next_nvpair(props, NULL);
4162 while (pair != NULL) {
4163 const char *propname = nvpair_name(pair);
4166 next_pair = nvlist_next_nvpair(props, pair);
4168 if ((nvlist_lookup_nvpair(origprops, propname,
4169 &match) != 0) || !propval_equals(pair, match))
4170 goto next; /* need to set received value */
4172 /* don't clear the existing received value */
4173 (void) nvlist_remove_nvpair(origprops, match);
4174 /* don't bother receiving the property */
4175 (void) nvlist_remove_nvpair(props, pair);
4182 * Extract properties that cannot be set PRIOR to the receipt of a dataset.
4183 * For example, refquota cannot be set until after the receipt of a dataset,
4184 * because in replication streams, an older/earlier snapshot may exceed the
4185 * refquota. We want to receive the older/earlier snapshot, but setting
4186 * refquota pre-receipt will set the dsl's ACTUAL quota, which will prevent
4187 * the older/earlier snapshot from being received (with EDQUOT).
4189 * The ZFS test "zfs_receive_011_pos" demonstrates such a scenario.
4191 * libzfs will need to be judicious handling errors encountered by props
4192 * extracted by this function.
4195 extract_delay_props(nvlist_t *props)
4197 nvlist_t *delayprops;
4198 nvpair_t *nvp, *tmp;
4199 static const zfs_prop_t delayable[] = { ZFS_PROP_REFQUOTA, 0 };
4202 VERIFY(nvlist_alloc(&delayprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4204 for (nvp = nvlist_next_nvpair(props, NULL); nvp != NULL;
4205 nvp = nvlist_next_nvpair(props, nvp)) {
4207 * strcmp() is safe because zfs_prop_to_name() always returns
4210 for (i = 0; delayable[i] != 0; i++) {
4211 if (strcmp(zfs_prop_to_name(delayable[i]),
4212 nvpair_name(nvp)) == 0) {
4216 if (delayable[i] != 0) {
4217 tmp = nvlist_prev_nvpair(props, nvp);
4218 VERIFY(nvlist_add_nvpair(delayprops, nvp) == 0);
4219 VERIFY(nvlist_remove_nvpair(props, nvp) == 0);
4224 if (nvlist_empty(delayprops)) {
4225 nvlist_free(delayprops);
4228 return (delayprops);
4232 static boolean_t zfs_ioc_recv_inject_err;
4237 * zc_name name of containing filesystem
4238 * zc_nvlist_src{_size} nvlist of properties to apply
4239 * zc_value name of snapshot to create
4240 * zc_string name of clone origin (if DRR_FLAG_CLONE)
4241 * zc_cookie file descriptor to recv from
4242 * zc_begin_record the BEGIN record of the stream (not byteswapped)
4243 * zc_guid force flag
4244 * zc_cleanup_fd cleanup-on-exit file descriptor
4245 * zc_action_handle handle for this guid/ds mapping (or zero on first call)
4246 * zc_resumable if data is incomplete assume sender will resume
4249 * zc_cookie number of bytes read
4250 * zc_nvlist_dst{_size} error for each unapplied received property
4251 * zc_obj zprop_errflags_t
4252 * zc_action_handle handle for this guid/ds mapping
4255 zfs_ioc_recv(zfs_cmd_t *zc)
4258 dmu_recv_cookie_t drc;
4259 boolean_t force = (boolean_t)zc->zc_guid;
4262 int props_error = 0;
4265 nvlist_t *props = NULL; /* sent properties */
4266 nvlist_t *origprops = NULL; /* existing properties */
4267 nvlist_t *delayprops = NULL; /* sent properties applied post-receive */
4268 char *origin = NULL;
4270 char tofs[ZFS_MAXNAMELEN];
4271 cap_rights_t rights;
4272 boolean_t first_recvd_props = B_FALSE;
4274 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
4275 strchr(zc->zc_value, '@') == NULL ||
4276 strchr(zc->zc_value, '%'))
4277 return (SET_ERROR(EINVAL));
4279 (void) strcpy(tofs, zc->zc_value);
4280 tosnap = strchr(tofs, '@');
4283 if (zc->zc_nvlist_src != 0 &&
4284 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
4285 zc->zc_iflags, &props)) != 0)
4292 fget_read(curthread, fd, cap_rights_init(&rights, CAP_PREAD), &fp);
4296 return (SET_ERROR(EBADF));
4299 errors = fnvlist_alloc();
4301 if (zc->zc_string[0])
4302 origin = zc->zc_string;
4304 error = dmu_recv_begin(tofs, tosnap,
4305 &zc->zc_begin_record, force, zc->zc_resumable, origin, &drc);
4310 * Set properties before we receive the stream so that they are applied
4311 * to the new data. Note that we must call dmu_recv_stream() if
4312 * dmu_recv_begin() succeeds.
4314 if (props != NULL && !drc.drc_newfs) {
4315 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
4316 SPA_VERSION_RECVD_PROPS &&
4317 !dsl_prop_get_hasrecvd(tofs))
4318 first_recvd_props = B_TRUE;
4321 * If new received properties are supplied, they are to
4322 * completely replace the existing received properties, so stash
4323 * away the existing ones.
4325 if (dsl_prop_get_received(tofs, &origprops) == 0) {
4326 nvlist_t *errlist = NULL;
4328 * Don't bother writing a property if its value won't
4329 * change (and avoid the unnecessary security checks).
4331 * The first receive after SPA_VERSION_RECVD_PROPS is a
4332 * special case where we blow away all local properties
4335 if (!first_recvd_props)
4336 props_reduce(props, origprops);
4337 if (zfs_check_clearable(tofs, origprops, &errlist) != 0)
4338 (void) nvlist_merge(errors, errlist, 0);
4339 nvlist_free(errlist);
4341 if (clear_received_props(tofs, origprops,
4342 first_recvd_props ? NULL : props) != 0)
4343 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4345 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4349 if (props != NULL) {
4350 props_error = dsl_prop_set_hasrecvd(tofs);
4352 if (props_error == 0) {
4353 delayprops = extract_delay_props(props);
4354 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4360 error = dmu_recv_stream(&drc, fp, &off, zc->zc_cleanup_fd,
4361 &zc->zc_action_handle);
4364 zfsvfs_t *zfsvfs = NULL;
4366 if (getzfsvfs(tofs, &zfsvfs) == 0) {
4370 error = zfs_suspend_fs(zfsvfs);
4372 * If the suspend fails, then the recv_end will
4373 * likely also fail, and clean up after itself.
4375 end_err = dmu_recv_end(&drc, zfsvfs);
4377 error = zfs_resume_fs(zfsvfs, tofs);
4378 error = error ? error : end_err;
4379 VFS_RELE(zfsvfs->z_vfs);
4381 error = dmu_recv_end(&drc, NULL);
4384 /* Set delayed properties now, after we're done receiving. */
4385 if (delayprops != NULL && error == 0) {
4386 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4387 delayprops, errors);
4391 if (delayprops != NULL) {
4393 * Merge delayed props back in with initial props, in case
4394 * we're DEBUG and zfs_ioc_recv_inject_err is set (which means
4395 * we have to make sure clear_received_props() includes
4396 * the delayed properties).
4398 * Since zfs_ioc_recv_inject_err is only in DEBUG kernels,
4399 * using ASSERT() will be just like a VERIFY.
4401 ASSERT(nvlist_merge(props, delayprops, 0) == 0);
4402 nvlist_free(delayprops);
4406 * Now that all props, initial and delayed, are set, report the prop
4407 * errors to the caller.
4409 if (zc->zc_nvlist_dst_size != 0 &&
4410 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4411 put_nvlist(zc, errors) != 0)) {
4413 * Caller made zc->zc_nvlist_dst less than the minimum expected
4414 * size or supplied an invalid address.
4416 props_error = SET_ERROR(EINVAL);
4419 zc->zc_cookie = off - fp->f_offset;
4420 if (off >= 0 && off <= MAXOFFSET_T)
4424 if (zfs_ioc_recv_inject_err) {
4425 zfs_ioc_recv_inject_err = B_FALSE;
4432 zvol_create_minors(tofs);
4436 * On error, restore the original props.
4438 if (error != 0 && props != NULL && !drc.drc_newfs) {
4439 if (clear_received_props(tofs, props, NULL) != 0) {
4441 * We failed to clear the received properties.
4442 * Since we may have left a $recvd value on the
4443 * system, we can't clear the $hasrecvd flag.
4445 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4446 } else if (first_recvd_props) {
4447 dsl_prop_unset_hasrecvd(tofs);
4450 if (origprops == NULL && !drc.drc_newfs) {
4451 /* We failed to stash the original properties. */
4452 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4456 * dsl_props_set() will not convert RECEIVED to LOCAL on or
4457 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4458 * explictly if we're restoring local properties cleared in the
4459 * first new-style receive.
4461 if (origprops != NULL &&
4462 zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4463 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4464 origprops, NULL) != 0) {
4466 * We stashed the original properties but failed to
4469 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4474 nvlist_free(origprops);
4475 nvlist_free(errors);
4479 error = props_error;
4486 * zc_name name of snapshot to send
4487 * zc_cookie file descriptor to send stream to
4488 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj)
4489 * zc_sendobj objsetid of snapshot to send
4490 * zc_fromobj objsetid of incremental fromsnap (may be zero)
4491 * zc_guid if set, estimate size of stream only. zc_cookie is ignored.
4492 * output size in zc_objset_type.
4493 * zc_flags lzc_send_flags
4496 * zc_objset_type estimated size, if zc_guid is set
4499 zfs_ioc_send(zfs_cmd_t *zc)
4503 boolean_t estimate = (zc->zc_guid != 0);
4504 boolean_t embedok = (zc->zc_flags & 0x1);
4505 boolean_t large_block_ok = (zc->zc_flags & 0x2);
4507 if (zc->zc_obj != 0) {
4509 dsl_dataset_t *tosnap;
4511 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4515 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4517 dsl_pool_rele(dp, FTAG);
4521 if (dsl_dir_is_clone(tosnap->ds_dir))
4523 dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj;
4524 dsl_dataset_rele(tosnap, FTAG);
4525 dsl_pool_rele(dp, FTAG);
4530 dsl_dataset_t *tosnap;
4531 dsl_dataset_t *fromsnap = NULL;
4533 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4537 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4539 dsl_pool_rele(dp, FTAG);
4543 if (zc->zc_fromobj != 0) {
4544 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
4547 dsl_dataset_rele(tosnap, FTAG);
4548 dsl_pool_rele(dp, FTAG);
4553 error = dmu_send_estimate(tosnap, fromsnap,
4554 &zc->zc_objset_type);
4556 if (fromsnap != NULL)
4557 dsl_dataset_rele(fromsnap, FTAG);
4558 dsl_dataset_rele(tosnap, FTAG);
4559 dsl_pool_rele(dp, FTAG);
4562 cap_rights_t rights;
4565 fp = getf(zc->zc_cookie);
4567 fget_write(curthread, zc->zc_cookie,
4568 cap_rights_init(&rights, CAP_WRITE), &fp);
4571 return (SET_ERROR(EBADF));
4574 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
4575 zc->zc_fromobj, embedok, large_block_ok,
4577 zc->zc_cookie, fp->f_vnode, &off);
4579 zc->zc_cookie, fp, &off);
4582 if (off >= 0 && off <= MAXOFFSET_T)
4584 releasef(zc->zc_cookie);
4591 * zc_name name of snapshot on which to report progress
4592 * zc_cookie file descriptor of send stream
4595 * zc_cookie number of bytes written in send stream thus far
4598 zfs_ioc_send_progress(zfs_cmd_t *zc)
4602 dmu_sendarg_t *dsp = NULL;
4605 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4609 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4611 dsl_pool_rele(dp, FTAG);
4615 mutex_enter(&ds->ds_sendstream_lock);
4618 * Iterate over all the send streams currently active on this dataset.
4619 * If there's one which matches the specified file descriptor _and_ the
4620 * stream was started by the current process, return the progress of
4623 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4624 dsp = list_next(&ds->ds_sendstreams, dsp)) {
4625 if (dsp->dsa_outfd == zc->zc_cookie &&
4626 dsp->dsa_proc == curproc)
4631 zc->zc_cookie = *(dsp->dsa_off);
4633 error = SET_ERROR(ENOENT);
4635 mutex_exit(&ds->ds_sendstream_lock);
4636 dsl_dataset_rele(ds, FTAG);
4637 dsl_pool_rele(dp, FTAG);
4642 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4646 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4647 &zc->zc_inject_record);
4650 zc->zc_guid = (uint64_t)id;
4656 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4658 return (zio_clear_fault((int)zc->zc_guid));
4662 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4664 int id = (int)zc->zc_guid;
4667 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4668 &zc->zc_inject_record);
4676 zfs_ioc_error_log(zfs_cmd_t *zc)
4680 size_t count = (size_t)zc->zc_nvlist_dst_size;
4682 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4685 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4688 zc->zc_nvlist_dst_size = count;
4690 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4692 spa_close(spa, FTAG);
4698 zfs_ioc_clear(zfs_cmd_t *zc)
4705 * On zpool clear we also fix up missing slogs
4707 mutex_enter(&spa_namespace_lock);
4708 spa = spa_lookup(zc->zc_name);
4710 mutex_exit(&spa_namespace_lock);
4711 return (SET_ERROR(EIO));
4713 if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4714 /* we need to let spa_open/spa_load clear the chains */
4715 spa_set_log_state(spa, SPA_LOG_CLEAR);
4717 spa->spa_last_open_failed = 0;
4718 mutex_exit(&spa_namespace_lock);
4720 if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4721 error = spa_open(zc->zc_name, &spa, FTAG);
4724 nvlist_t *config = NULL;
4726 if (zc->zc_nvlist_src == 0)
4727 return (SET_ERROR(EINVAL));
4729 if ((error = get_nvlist(zc->zc_nvlist_src,
4730 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4731 error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4733 if (config != NULL) {
4736 if ((err = put_nvlist(zc, config)) != 0)
4738 nvlist_free(config);
4740 nvlist_free(policy);
4747 spa_vdev_state_enter(spa, SCL_NONE);
4749 if (zc->zc_guid == 0) {
4752 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4754 (void) spa_vdev_state_exit(spa, NULL, ENODEV);
4755 spa_close(spa, FTAG);
4756 return (SET_ERROR(ENODEV));
4760 vdev_clear(spa, vd);
4762 (void) spa_vdev_state_exit(spa, NULL, 0);
4765 * Resume any suspended I/Os.
4767 if (zio_resume(spa) != 0)
4768 error = SET_ERROR(EIO);
4770 spa_close(spa, FTAG);
4776 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4781 error = spa_open(zc->zc_name, &spa, FTAG);
4785 spa_vdev_state_enter(spa, SCL_NONE);
4788 * If a resilver is already in progress then set the
4789 * spa_scrub_reopen flag to B_TRUE so that we don't restart
4790 * the scan as a side effect of the reopen. Otherwise, let
4791 * vdev_open() decided if a resilver is required.
4793 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4794 vdev_reopen(spa->spa_root_vdev);
4795 spa->spa_scrub_reopen = B_FALSE;
4797 (void) spa_vdev_state_exit(spa, NULL, 0);
4798 spa_close(spa, FTAG);
4803 * zc_name name of filesystem
4804 * zc_value name of origin snapshot
4807 * zc_string name of conflicting snapshot, if there is one
4810 zfs_ioc_promote(zfs_cmd_t *zc)
4815 * We don't need to unmount *all* the origin fs's snapshots, but
4818 cp = strchr(zc->zc_value, '@');
4821 (void) dmu_objset_find(zc->zc_value,
4822 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
4823 return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4827 * Retrieve a single {user|group}{used|quota}@... property.
4830 * zc_name name of filesystem
4831 * zc_objset_type zfs_userquota_prop_t
4832 * zc_value domain name (eg. "S-1-234-567-89")
4833 * zc_guid RID/UID/GID
4836 * zc_cookie property value
4839 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4844 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4845 return (SET_ERROR(EINVAL));
4847 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4851 error = zfs_userspace_one(zfsvfs,
4852 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4853 zfsvfs_rele(zfsvfs, FTAG);
4860 * zc_name name of filesystem
4861 * zc_cookie zap cursor
4862 * zc_objset_type zfs_userquota_prop_t
4863 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4866 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
4867 * zc_cookie zap cursor
4870 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4873 int bufsize = zc->zc_nvlist_dst_size;
4876 return (SET_ERROR(ENOMEM));
4878 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4882 void *buf = kmem_alloc(bufsize, KM_SLEEP);
4884 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
4885 buf, &zc->zc_nvlist_dst_size);
4888 error = ddi_copyout(buf,
4889 (void *)(uintptr_t)zc->zc_nvlist_dst,
4890 zc->zc_nvlist_dst_size, zc->zc_iflags);
4892 kmem_free(buf, bufsize);
4893 zfsvfs_rele(zfsvfs, FTAG);
4900 * zc_name name of filesystem
4906 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4912 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
4913 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
4915 * If userused is not enabled, it may be because the
4916 * objset needs to be closed & reopened (to grow the
4917 * objset_phys_t). Suspend/resume the fs will do that.
4919 error = zfs_suspend_fs(zfsvfs);
4921 dmu_objset_refresh_ownership(zfsvfs->z_os,
4923 error = zfs_resume_fs(zfsvfs, zc->zc_name);
4927 error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
4928 VFS_RELE(zfsvfs->z_vfs);
4930 /* XXX kind of reading contents without owning */
4931 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4935 error = dmu_objset_userspace_upgrade(os);
4936 dmu_objset_rele(os, FTAG);
4944 * We don't want to have a hard dependency
4945 * against some special symbols in sharefs
4946 * nfs, and smbsrv. Determine them if needed when
4947 * the first file system is shared.
4948 * Neither sharefs, nfs or smbsrv are unloadable modules.
4950 int (*znfsexport_fs)(void *arg);
4951 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
4952 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
4954 int zfs_nfsshare_inited;
4955 int zfs_smbshare_inited;
4957 ddi_modhandle_t nfs_mod;
4958 ddi_modhandle_t sharefs_mod;
4959 ddi_modhandle_t smbsrv_mod;
4960 #endif /* illumos */
4961 kmutex_t zfs_share_lock;
4969 ASSERT(MUTEX_HELD(&zfs_share_lock));
4970 /* Both NFS and SMB shares also require sharetab support. */
4971 if (sharefs_mod == NULL && ((sharefs_mod =
4972 ddi_modopen("fs/sharefs",
4973 KRTLD_MODE_FIRST, &error)) == NULL)) {
4974 return (SET_ERROR(ENOSYS));
4976 if (zshare_fs == NULL && ((zshare_fs =
4977 (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
4978 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
4979 return (SET_ERROR(ENOSYS));
4983 #endif /* illumos */
4986 zfs_ioc_share(zfs_cmd_t *zc)
4992 switch (zc->zc_share.z_sharetype) {
4994 case ZFS_UNSHARE_NFS:
4995 if (zfs_nfsshare_inited == 0) {
4996 mutex_enter(&zfs_share_lock);
4997 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
4998 KRTLD_MODE_FIRST, &error)) == NULL)) {
4999 mutex_exit(&zfs_share_lock);
5000 return (SET_ERROR(ENOSYS));
5002 if (znfsexport_fs == NULL &&
5003 ((znfsexport_fs = (int (*)(void *))
5005 "nfs_export", &error)) == NULL)) {
5006 mutex_exit(&zfs_share_lock);
5007 return (SET_ERROR(ENOSYS));
5009 error = zfs_init_sharefs();
5011 mutex_exit(&zfs_share_lock);
5012 return (SET_ERROR(ENOSYS));
5014 zfs_nfsshare_inited = 1;
5015 mutex_exit(&zfs_share_lock);
5019 case ZFS_UNSHARE_SMB:
5020 if (zfs_smbshare_inited == 0) {
5021 mutex_enter(&zfs_share_lock);
5022 if (smbsrv_mod == NULL && ((smbsrv_mod =
5023 ddi_modopen("drv/smbsrv",
5024 KRTLD_MODE_FIRST, &error)) == NULL)) {
5025 mutex_exit(&zfs_share_lock);
5026 return (SET_ERROR(ENOSYS));
5028 if (zsmbexport_fs == NULL && ((zsmbexport_fs =
5029 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
5030 "smb_server_share", &error)) == NULL)) {
5031 mutex_exit(&zfs_share_lock);
5032 return (SET_ERROR(ENOSYS));
5034 error = zfs_init_sharefs();
5036 mutex_exit(&zfs_share_lock);
5037 return (SET_ERROR(ENOSYS));
5039 zfs_smbshare_inited = 1;
5040 mutex_exit(&zfs_share_lock);
5044 return (SET_ERROR(EINVAL));
5047 switch (zc->zc_share.z_sharetype) {
5049 case ZFS_UNSHARE_NFS:
5051 znfsexport_fs((void *)
5052 (uintptr_t)zc->zc_share.z_exportdata))
5056 case ZFS_UNSHARE_SMB:
5057 if (error = zsmbexport_fs((void *)
5058 (uintptr_t)zc->zc_share.z_exportdata,
5059 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
5066 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
5067 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
5068 SHAREFS_ADD : SHAREFS_REMOVE;
5071 * Add or remove share from sharetab
5073 error = zshare_fs(opcode,
5074 (void *)(uintptr_t)zc->zc_share.z_sharedata,
5075 zc->zc_share.z_sharemax);
5079 #else /* !illumos */
5081 #endif /* illumos */
5084 ace_t full_access[] = {
5085 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
5090 * zc_name name of containing filesystem
5091 * zc_obj object # beyond which we want next in-use object #
5094 * zc_obj next in-use object #
5097 zfs_ioc_next_obj(zfs_cmd_t *zc)
5099 objset_t *os = NULL;
5102 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
5106 error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
5107 dsl_dataset_phys(os->os_dsl_dataset)->ds_prev_snap_txg);
5109 dmu_objset_rele(os, FTAG);
5115 * zc_name name of filesystem
5116 * zc_value prefix name for snapshot
5117 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
5120 * zc_value short name of new snapshot
5123 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
5130 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
5134 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
5135 (u_longlong_t)ddi_get_lbolt64());
5136 hold_name = kmem_asprintf("%%%s", zc->zc_value);
5138 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
5141 (void) strcpy(zc->zc_value, snap_name);
5144 zfs_onexit_fd_rele(zc->zc_cleanup_fd);
5150 * zc_name name of "to" snapshot
5151 * zc_value name of "from" snapshot
5152 * zc_cookie file descriptor to write diff data on
5155 * dmu_diff_record_t's to the file descriptor
5158 zfs_ioc_diff(zfs_cmd_t *zc)
5161 cap_rights_t rights;
5166 fp = getf(zc->zc_cookie);
5168 fget_write(curthread, zc->zc_cookie,
5169 cap_rights_init(&rights, CAP_WRITE), &fp);
5172 return (SET_ERROR(EBADF));
5177 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off);
5179 error = dmu_diff(zc->zc_name, zc->zc_value, fp, &off);
5182 if (off >= 0 && off <= MAXOFFSET_T)
5184 releasef(zc->zc_cookie);
5191 * Remove all ACL files in shares dir
5194 zfs_smb_acl_purge(znode_t *dzp)
5197 zap_attribute_t zap;
5198 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
5201 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
5202 (error = zap_cursor_retrieve(&zc, &zap)) == 0;
5203 zap_cursor_advance(&zc)) {
5204 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
5208 zap_cursor_fini(&zc);
5211 #endif /* illumos */
5214 zfs_ioc_smb_acl(zfs_cmd_t *zc)
5219 vnode_t *resourcevp = NULL;
5228 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
5229 NO_FOLLOW, NULL, &vp)) != 0)
5232 /* Now make sure mntpnt and dataset are ZFS */
5234 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 ||
5235 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
5236 zc->zc_name) != 0)) {
5238 return (SET_ERROR(EINVAL));
5242 zfsvfs = dzp->z_zfsvfs;
5246 * Create share dir if its missing.
5248 mutex_enter(&zfsvfs->z_lock);
5249 if (zfsvfs->z_shares_dir == 0) {
5252 tx = dmu_tx_create(zfsvfs->z_os);
5253 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
5255 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
5256 error = dmu_tx_assign(tx, TXG_WAIT);
5260 error = zfs_create_share_dir(zfsvfs, tx);
5264 mutex_exit(&zfsvfs->z_lock);
5270 mutex_exit(&zfsvfs->z_lock);
5272 ASSERT(zfsvfs->z_shares_dir);
5273 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
5279 switch (zc->zc_cookie) {
5280 case ZFS_SMB_ACL_ADD:
5281 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
5282 vattr.va_type = VREG;
5283 vattr.va_mode = S_IFREG|0777;
5287 vsec.vsa_mask = VSA_ACE;
5288 vsec.vsa_aclentp = &full_access;
5289 vsec.vsa_aclentsz = sizeof (full_access);
5290 vsec.vsa_aclcnt = 1;
5292 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
5293 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
5295 VN_RELE(resourcevp);
5298 case ZFS_SMB_ACL_REMOVE:
5299 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
5303 case ZFS_SMB_ACL_RENAME:
5304 if ((error = get_nvlist(zc->zc_nvlist_src,
5305 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
5307 VN_RELE(ZTOV(sharedir));
5311 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
5312 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
5315 VN_RELE(ZTOV(sharedir));
5317 nvlist_free(nvlist);
5320 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
5322 nvlist_free(nvlist);
5325 case ZFS_SMB_ACL_PURGE:
5326 error = zfs_smb_acl_purge(sharedir);
5330 error = SET_ERROR(EINVAL);
5335 VN_RELE(ZTOV(sharedir));
5340 #else /* !illumos */
5341 return (EOPNOTSUPP);
5342 #endif /* illumos */
5347 * "holds" -> { snapname -> holdname (string), ... }
5348 * (optional) "cleanup_fd" -> fd (int32)
5352 * snapname -> error value (int32)
5358 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
5362 int cleanup_fd = -1;
5366 error = nvlist_lookup_nvlist(args, "holds", &holds);
5368 return (SET_ERROR(EINVAL));
5370 /* make sure the user didn't pass us any invalid (empty) tags */
5371 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
5372 pair = nvlist_next_nvpair(holds, pair)) {
5375 error = nvpair_value_string(pair, &htag);
5377 return (SET_ERROR(error));
5379 if (strlen(htag) == 0)
5380 return (SET_ERROR(EINVAL));
5383 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
5384 error = zfs_onexit_fd_hold(cleanup_fd, &minor);
5389 error = dsl_dataset_user_hold(holds, minor, errlist);
5391 zfs_onexit_fd_rele(cleanup_fd);
5396 * innvl is not used.
5399 * holdname -> time added (uint64 seconds since epoch)
5405 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
5407 return (dsl_dataset_get_holds(snapname, outnvl));
5412 * snapname -> { holdname, ... }
5417 * snapname -> error value (int32)
5423 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
5425 return (dsl_dataset_user_release(holds, errlist));
5430 * zc_name name of new filesystem or snapshot
5431 * zc_value full name of old snapshot
5434 * zc_cookie space in bytes
5435 * zc_objset_type compressed space in bytes
5436 * zc_perm_action uncompressed space in bytes
5439 zfs_ioc_space_written(zfs_cmd_t *zc)
5443 dsl_dataset_t *new, *old;
5445 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5448 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
5450 dsl_pool_rele(dp, FTAG);
5453 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
5455 dsl_dataset_rele(new, FTAG);
5456 dsl_pool_rele(dp, FTAG);
5460 error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
5461 &zc->zc_objset_type, &zc->zc_perm_action);
5462 dsl_dataset_rele(old, FTAG);
5463 dsl_dataset_rele(new, FTAG);
5464 dsl_pool_rele(dp, FTAG);
5470 * "firstsnap" -> snapshot name
5474 * "used" -> space in bytes
5475 * "compressed" -> compressed space in bytes
5476 * "uncompressed" -> uncompressed space in bytes
5480 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
5484 dsl_dataset_t *new, *old;
5486 uint64_t used, comp, uncomp;
5488 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5489 return (SET_ERROR(EINVAL));
5491 error = dsl_pool_hold(lastsnap, FTAG, &dp);
5495 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
5496 if (error == 0 && !new->ds_is_snapshot) {
5497 dsl_dataset_rele(new, FTAG);
5498 error = SET_ERROR(EINVAL);
5501 dsl_pool_rele(dp, FTAG);
5504 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
5505 if (error == 0 && !old->ds_is_snapshot) {
5506 dsl_dataset_rele(old, FTAG);
5507 error = SET_ERROR(EINVAL);
5510 dsl_dataset_rele(new, FTAG);
5511 dsl_pool_rele(dp, FTAG);
5515 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5516 dsl_dataset_rele(old, FTAG);
5517 dsl_dataset_rele(new, FTAG);
5518 dsl_pool_rele(dp, FTAG);
5519 fnvlist_add_uint64(outnvl, "used", used);
5520 fnvlist_add_uint64(outnvl, "compressed", comp);
5521 fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5526 zfs_ioc_jail(zfs_cmd_t *zc)
5529 return (zone_dataset_attach(curthread->td_ucred, zc->zc_name,
5530 (int)zc->zc_jailid));
5534 zfs_ioc_unjail(zfs_cmd_t *zc)
5537 return (zone_dataset_detach(curthread->td_ucred, zc->zc_name,
5538 (int)zc->zc_jailid));
5543 * "fd" -> file descriptor to write stream to (int32)
5544 * (optional) "fromsnap" -> full snap name to send an incremental from
5545 * (optional) "largeblockok" -> (value ignored)
5546 * indicates that blocks > 128KB are permitted
5547 * (optional) "embedok" -> (value ignored)
5548 * presence indicates DRR_WRITE_EMBEDDED records are permitted
5549 * (optional) "resume_object" and "resume_offset" -> (uint64)
5550 * if present, resume send stream from specified object and offset.
5557 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5559 cap_rights_t rights;
5563 char *fromname = NULL;
5565 boolean_t largeblockok;
5567 uint64_t resumeobj = 0;
5568 uint64_t resumeoff = 0;
5570 error = nvlist_lookup_int32(innvl, "fd", &fd);
5572 return (SET_ERROR(EINVAL));
5574 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
5576 largeblockok = nvlist_exists(innvl, "largeblockok");
5577 embedok = nvlist_exists(innvl, "embedok");
5579 (void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj);
5580 (void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff);
5583 file_t *fp = getf(fd);
5585 fget_write(curthread, fd, cap_rights_init(&rights, CAP_WRITE), &fp);
5588 return (SET_ERROR(EBADF));
5591 error = dmu_send(snapname, fromname, embedok, largeblockok, fd,
5593 resumeobj, resumeoff, fp->f_vnode, &off);
5595 resumeobj, resumeoff, fp, &off);
5599 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5610 * Determine approximately how large a zfs send stream will be -- the number
5611 * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5614 * (optional) "from" -> full snap or bookmark name to send an incremental
5619 * "space" -> bytes of space (uint64)
5623 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5626 dsl_dataset_t *tosnap;
5631 error = dsl_pool_hold(snapname, FTAG, &dp);
5635 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
5637 dsl_pool_rele(dp, FTAG);
5641 error = nvlist_lookup_string(innvl, "from", &fromname);
5643 if (strchr(fromname, '@') != NULL) {
5645 * If from is a snapshot, hold it and use the more
5646 * efficient dmu_send_estimate to estimate send space
5647 * size using deadlists.
5649 dsl_dataset_t *fromsnap;
5650 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
5653 error = dmu_send_estimate(tosnap, fromsnap, &space);
5654 dsl_dataset_rele(fromsnap, FTAG);
5655 } else if (strchr(fromname, '#') != NULL) {
5657 * If from is a bookmark, fetch the creation TXG of the
5658 * snapshot it was created from and use that to find
5659 * blocks that were born after it.
5661 zfs_bookmark_phys_t frombm;
5663 error = dsl_bookmark_lookup(dp, fromname, tosnap,
5667 error = dmu_send_estimate_from_txg(tosnap,
5668 frombm.zbm_creation_txg, &space);
5671 * from is not properly formatted as a snapshot or
5674 error = SET_ERROR(EINVAL);
5678 // If estimating the size of a full send, use dmu_send_estimate
5679 error = dmu_send_estimate(tosnap, NULL, &space);
5682 fnvlist_add_uint64(outnvl, "space", space);
5685 dsl_dataset_rele(tosnap, FTAG);
5686 dsl_pool_rele(dp, FTAG);
5690 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
5693 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5694 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5695 boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
5697 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5699 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5700 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5701 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5702 ASSERT3P(vec->zvec_func, ==, NULL);
5704 vec->zvec_legacy_func = func;
5705 vec->zvec_secpolicy = secpolicy;
5706 vec->zvec_namecheck = namecheck;
5707 vec->zvec_allow_log = log_history;
5708 vec->zvec_pool_check = pool_check;
5712 * See the block comment at the beginning of this file for details on
5713 * each argument to this function.
5716 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
5717 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5718 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
5719 boolean_t allow_log)
5721 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5723 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5724 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5725 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5726 ASSERT3P(vec->zvec_func, ==, NULL);
5728 /* if we are logging, the name must be valid */
5729 ASSERT(!allow_log || namecheck != NO_NAME);
5731 vec->zvec_name = name;
5732 vec->zvec_func = func;
5733 vec->zvec_secpolicy = secpolicy;
5734 vec->zvec_namecheck = namecheck;
5735 vec->zvec_pool_check = pool_check;
5736 vec->zvec_smush_outnvlist = smush_outnvlist;
5737 vec->zvec_allow_log = allow_log;
5741 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5742 zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
5743 zfs_ioc_poolcheck_t pool_check)
5745 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5746 POOL_NAME, log_history, pool_check);
5750 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5751 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
5753 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5754 DATASET_NAME, B_FALSE, pool_check);
5758 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5760 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
5761 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5765 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5766 zfs_secpolicy_func_t *secpolicy)
5768 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5769 NO_NAME, B_FALSE, POOL_CHECK_NONE);
5773 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
5774 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
5776 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5777 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
5781 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5783 zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
5784 zfs_secpolicy_read);
5788 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5789 zfs_secpolicy_func_t *secpolicy)
5791 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5792 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5796 zfs_ioctl_init(void)
5798 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
5799 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
5800 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5802 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
5803 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
5804 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
5806 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
5807 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
5808 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5810 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
5811 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
5812 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5814 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
5815 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
5816 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5818 zfs_ioctl_register("create", ZFS_IOC_CREATE,
5819 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
5820 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5822 zfs_ioctl_register("clone", ZFS_IOC_CLONE,
5823 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
5824 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5826 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
5827 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
5828 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5830 zfs_ioctl_register("hold", ZFS_IOC_HOLD,
5831 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
5832 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5833 zfs_ioctl_register("release", ZFS_IOC_RELEASE,
5834 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
5835 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5837 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
5838 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
5839 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5841 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
5842 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
5843 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE);
5845 zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK,
5846 zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME,
5847 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5849 zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS,
5850 zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME,
5851 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5853 zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS,
5854 zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks,
5856 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5858 /* IOCTLS that use the legacy function signature */
5860 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
5861 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
5863 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
5864 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5865 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
5867 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
5868 zfs_ioc_pool_upgrade);
5869 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
5871 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
5872 zfs_ioc_vdev_remove);
5873 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
5874 zfs_ioc_vdev_set_state);
5875 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
5876 zfs_ioc_vdev_attach);
5877 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
5878 zfs_ioc_vdev_detach);
5879 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
5880 zfs_ioc_vdev_setpath);
5881 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
5882 zfs_ioc_vdev_setfru);
5883 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
5884 zfs_ioc_pool_set_props);
5885 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
5886 zfs_ioc_vdev_split);
5887 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
5888 zfs_ioc_pool_reguid);
5890 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
5891 zfs_ioc_pool_configs, zfs_secpolicy_none);
5892 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
5893 zfs_ioc_pool_tryimport, zfs_secpolicy_config);
5894 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
5895 zfs_ioc_inject_fault, zfs_secpolicy_inject);
5896 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
5897 zfs_ioc_clear_fault, zfs_secpolicy_inject);
5898 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
5899 zfs_ioc_inject_list_next, zfs_secpolicy_inject);
5902 * pool destroy, and export don't log the history as part of
5903 * zfsdev_ioctl, but rather zfs_ioc_pool_export
5904 * does the logging of those commands.
5906 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
5907 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5908 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
5909 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5911 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
5912 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5913 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
5914 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5916 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
5917 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_NONE);
5918 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
5919 zfs_ioc_dsobj_to_dsname,
5920 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_NONE);
5921 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
5922 zfs_ioc_pool_get_history,
5923 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
5925 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
5926 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5928 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
5929 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5930 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
5931 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5933 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
5934 zfs_ioc_space_written);
5935 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
5936 zfs_ioc_objset_recvd_props);
5937 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
5939 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
5941 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
5942 zfs_ioc_objset_stats);
5943 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
5944 zfs_ioc_objset_zplprops);
5945 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
5946 zfs_ioc_dataset_list_next);
5947 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
5948 zfs_ioc_snapshot_list_next);
5949 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
5950 zfs_ioc_send_progress);
5952 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
5953 zfs_ioc_diff, zfs_secpolicy_diff);
5954 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
5955 zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
5956 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
5957 zfs_ioc_obj_to_path, zfs_secpolicy_diff);
5958 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
5959 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
5960 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
5961 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
5962 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
5963 zfs_ioc_send, zfs_secpolicy_send);
5965 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
5966 zfs_secpolicy_none);
5967 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
5968 zfs_secpolicy_destroy);
5969 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
5970 zfs_secpolicy_rename);
5971 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
5972 zfs_secpolicy_recv);
5973 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
5974 zfs_secpolicy_promote);
5975 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
5976 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
5977 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
5978 zfs_secpolicy_set_fsacl);
5980 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
5981 zfs_secpolicy_share, POOL_CHECK_NONE);
5982 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
5983 zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
5984 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
5985 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
5986 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5987 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
5988 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
5989 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5992 zfs_ioctl_register_dataset_nolog(ZFS_IOC_JAIL, zfs_ioc_jail,
5993 zfs_secpolicy_config, POOL_CHECK_NONE);
5994 zfs_ioctl_register_dataset_nolog(ZFS_IOC_UNJAIL, zfs_ioc_unjail,
5995 zfs_secpolicy_config, POOL_CHECK_NONE);
6000 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
6001 zfs_ioc_poolcheck_t check)
6006 ASSERT(type == POOL_NAME || type == DATASET_NAME);
6008 if (check & POOL_CHECK_NONE)
6011 error = spa_open(name, &spa, FTAG);
6013 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
6014 error = SET_ERROR(EAGAIN);
6015 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
6016 error = SET_ERROR(EROFS);
6017 spa_close(spa, FTAG);
6023 * Find a free minor number.
6026 zfsdev_minor_alloc(void)
6028 static minor_t last_minor;
6031 ASSERT(MUTEX_HELD(&spa_namespace_lock));
6033 for (m = last_minor + 1; m != last_minor; m++) {
6034 if (m > ZFSDEV_MAX_MINOR)
6036 if (ddi_get_soft_state(zfsdev_state, m) == NULL) {
6046 zfs_ctldev_init(struct cdev *devp)
6049 zfs_soft_state_t *zs;
6051 ASSERT(MUTEX_HELD(&spa_namespace_lock));
6053 minor = zfsdev_minor_alloc();
6055 return (SET_ERROR(ENXIO));
6057 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS)
6058 return (SET_ERROR(EAGAIN));
6060 devfs_set_cdevpriv((void *)(uintptr_t)minor, zfsdev_close);
6062 zs = ddi_get_soft_state(zfsdev_state, minor);
6063 zs->zss_type = ZSST_CTLDEV;
6064 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data);
6070 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor)
6072 ASSERT(MUTEX_HELD(&spa_namespace_lock));
6074 zfs_onexit_destroy(zo);
6075 ddi_soft_state_free(zfsdev_state, minor);
6079 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which)
6081 zfs_soft_state_t *zp;
6083 zp = ddi_get_soft_state(zfsdev_state, minor);
6084 if (zp == NULL || zp->zss_type != which)
6087 return (zp->zss_data);
6091 zfsdev_open(struct cdev *devp, int flag, int mode, struct thread *td)
6096 if (getminor(*devp) != 0)
6097 return (zvol_open(devp, flag, otyp, cr));
6100 /* This is the control device. Allocate a new minor if requested. */
6102 mutex_enter(&spa_namespace_lock);
6103 error = zfs_ctldev_init(devp);
6104 mutex_exit(&spa_namespace_lock);
6111 zfsdev_close(void *data)
6114 minor_t minor = (minor_t)(uintptr_t)data;
6119 mutex_enter(&spa_namespace_lock);
6120 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV);
6122 mutex_exit(&spa_namespace_lock);
6125 zfs_ctldev_destroy(zo, minor);
6126 mutex_exit(&spa_namespace_lock);
6130 zfsdev_ioctl(struct cdev *dev, u_long zcmd, caddr_t arg, int flag,
6137 minor_t minor = getminor(dev);
6139 zfs_iocparm_t *zc_iocparm;
6140 int cflag, cmd, oldvecnum;
6141 boolean_t newioc, compat;
6142 void *compat_zc = NULL;
6143 cred_t *cr = td->td_ucred;
6145 const zfs_ioc_vec_t *vec;
6146 char *saved_poolname = NULL;
6147 nvlist_t *innvl = NULL;
6149 cflag = ZFS_CMD_COMPAT_NONE;
6151 newioc = B_TRUE; /* "new" style (zfs_iocparm_t) ioctl */
6153 len = IOCPARM_LEN(zcmd);
6154 vecnum = cmd = zcmd & 0xff;
6157 * Check if we are talking to supported older binaries
6158 * and translate zfs_cmd if necessary
6160 if (len != sizeof(zfs_iocparm_t)) {
6167 case sizeof(zfs_cmd_zcmd_t):
6168 cflag = ZFS_CMD_COMPAT_LZC;
6170 case sizeof(zfs_cmd_deadman_t):
6171 cflag = ZFS_CMD_COMPAT_DEADMAN;
6173 case sizeof(zfs_cmd_v28_t):
6174 cflag = ZFS_CMD_COMPAT_V28;
6176 case sizeof(zfs_cmd_v15_t):
6177 cflag = ZFS_CMD_COMPAT_V15;
6178 vecnum = zfs_ioctl_v15_to_v28[cmd];
6181 * Return without further handling
6182 * if the command is blacklisted.
6184 if (vecnum == ZFS_IOC_COMPAT_PASS)
6186 else if (vecnum == ZFS_IOC_COMPAT_FAIL)
6195 vecnum = cmd - ZFS_IOC_FIRST;
6196 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
6199 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
6200 return (SET_ERROR(EINVAL));
6201 vec = &zfs_ioc_vec[vecnum];
6203 zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP);
6206 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
6208 error = SET_ERROR(EFAULT);
6211 #else /* !illumos */
6212 bzero(zc, sizeof(zfs_cmd_t));
6215 zc_iocparm = (void *)arg;
6217 switch (zc_iocparm->zfs_ioctl_version) {
6218 case ZFS_IOCVER_CURRENT:
6219 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_t)) {
6220 error = SET_ERROR(EINVAL);
6224 case ZFS_IOCVER_RESUME:
6225 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_resume_t)) {
6226 error = SET_ERROR(EFAULT);
6230 cflag = ZFS_CMD_COMPAT_RESUME;
6232 case ZFS_IOCVER_EDBP:
6233 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_edbp_t)) {
6234 error = SET_ERROR(EFAULT);
6238 cflag = ZFS_CMD_COMPAT_EDBP;
6240 case ZFS_IOCVER_ZCMD:
6241 if (zc_iocparm->zfs_cmd_size > sizeof(zfs_cmd_t) ||
6242 zc_iocparm->zfs_cmd_size < sizeof(zfs_cmd_zcmd_t)) {
6243 error = SET_ERROR(EFAULT);
6247 cflag = ZFS_CMD_COMPAT_ZCMD;
6250 error = SET_ERROR(EINVAL);
6256 ASSERT(sizeof(zfs_cmd_t) >= zc_iocparm->zfs_cmd_size);
6257 compat_zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP);
6258 bzero(compat_zc, sizeof(zfs_cmd_t));
6260 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd,
6261 compat_zc, zc_iocparm->zfs_cmd_size, flag);
6263 error = SET_ERROR(EFAULT);
6267 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd,
6268 zc, zc_iocparm->zfs_cmd_size, flag);
6270 error = SET_ERROR(EFAULT);
6278 ASSERT(compat_zc != NULL);
6279 zfs_cmd_compat_get(zc, compat_zc, cflag);
6281 ASSERT(compat_zc == NULL);
6282 zfs_cmd_compat_get(zc, arg, cflag);
6285 error = zfs_ioctl_compat_pre(zc, &vecnum, cflag);
6288 if (oldvecnum != vecnum)
6289 vec = &zfs_ioc_vec[vecnum];
6291 #endif /* !illumos */
6293 zc->zc_iflags = flag & FKIOCTL;
6294 if (zc->zc_nvlist_src_size != 0) {
6295 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
6296 zc->zc_iflags, &innvl);
6301 /* rewrite innvl for backwards compatibility */
6303 innvl = zfs_ioctl_compat_innvl(zc, innvl, vecnum, cflag);
6306 * Ensure that all pool/dataset names are valid before we pass down to
6309 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
6310 switch (vec->zvec_namecheck) {
6312 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
6313 error = SET_ERROR(EINVAL);
6315 error = pool_status_check(zc->zc_name,
6316 vec->zvec_namecheck, vec->zvec_pool_check);
6320 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
6321 error = SET_ERROR(EINVAL);
6323 error = pool_status_check(zc->zc_name,
6324 vec->zvec_namecheck, vec->zvec_pool_check);
6331 if (error == 0 && !(flag & FKIOCTL))
6332 error = vec->zvec_secpolicy(zc, innvl, cr);
6337 /* legacy ioctls can modify zc_name */
6338 len = strcspn(zc->zc_name, "/@#") + 1;
6339 saved_poolname = kmem_alloc(len, KM_SLEEP);
6340 (void) strlcpy(saved_poolname, zc->zc_name, len);
6342 if (vec->zvec_func != NULL) {
6346 nvlist_t *lognv = NULL;
6348 ASSERT(vec->zvec_legacy_func == NULL);
6351 * Add the innvl to the lognv before calling the func,
6352 * in case the func changes the innvl.
6354 if (vec->zvec_allow_log) {
6355 lognv = fnvlist_alloc();
6356 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
6358 if (!nvlist_empty(innvl)) {
6359 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
6364 outnvl = fnvlist_alloc();
6365 error = vec->zvec_func(zc->zc_name, innvl, outnvl);
6367 if (error == 0 && vec->zvec_allow_log &&
6368 spa_open(zc->zc_name, &spa, FTAG) == 0) {
6369 if (!nvlist_empty(outnvl)) {
6370 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
6373 (void) spa_history_log_nvl(spa, lognv);
6374 spa_close(spa, FTAG);
6376 fnvlist_free(lognv);
6378 /* rewrite outnvl for backwards compatibility */
6380 outnvl = zfs_ioctl_compat_outnvl(zc, outnvl, vecnum,
6383 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
6385 if (vec->zvec_smush_outnvlist) {
6386 smusherror = nvlist_smush(outnvl,
6387 zc->zc_nvlist_dst_size);
6389 if (smusherror == 0)
6390 puterror = put_nvlist(zc, outnvl);
6396 nvlist_free(outnvl);
6398 error = vec->zvec_legacy_func(zc);
6405 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
6406 if (error == 0 && rc != 0)
6407 error = SET_ERROR(EFAULT);
6410 zfs_ioctl_compat_post(zc, cmd, cflag);
6412 ASSERT(compat_zc != NULL);
6413 ASSERT(sizeof(zfs_cmd_t) >= zc_iocparm->zfs_cmd_size);
6415 zfs_cmd_compat_put(zc, compat_zc, vecnum, cflag);
6416 rc = ddi_copyout(compat_zc,
6417 (void *)(uintptr_t)zc_iocparm->zfs_cmd,
6418 zc_iocparm->zfs_cmd_size, flag);
6419 if (error == 0 && rc != 0)
6420 error = SET_ERROR(EFAULT);
6421 kmem_free(compat_zc, sizeof (zfs_cmd_t));
6423 zfs_cmd_compat_put(zc, arg, vecnum, cflag);
6428 rc = ddi_copyout(zc, (void *)(uintptr_t)zc_iocparm->zfs_cmd,
6429 sizeof (zfs_cmd_t), flag);
6430 if (error == 0 && rc != 0)
6431 error = SET_ERROR(EFAULT);
6434 if (error == 0 && vec->zvec_allow_log) {
6435 char *s = tsd_get(zfs_allow_log_key);
6438 (void) tsd_set(zfs_allow_log_key, saved_poolname);
6440 if (saved_poolname != NULL)
6441 strfree(saved_poolname);
6444 kmem_free(zc, sizeof (zfs_cmd_t));
6450 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
6452 if (cmd != DDI_ATTACH)
6453 return (DDI_FAILURE);
6455 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
6456 DDI_PSEUDO, 0) == DDI_FAILURE)
6457 return (DDI_FAILURE);
6461 ddi_report_dev(dip);
6463 return (DDI_SUCCESS);
6467 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
6469 if (spa_busy() || zfs_busy() || zvol_busy())
6470 return (DDI_FAILURE);
6472 if (cmd != DDI_DETACH)
6473 return (DDI_FAILURE);
6477 ddi_prop_remove_all(dip);
6478 ddi_remove_minor_node(dip, NULL);
6480 return (DDI_SUCCESS);
6485 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
6488 case DDI_INFO_DEVT2DEVINFO:
6490 return (DDI_SUCCESS);
6492 case DDI_INFO_DEVT2INSTANCE:
6493 *result = (void *)0;
6494 return (DDI_SUCCESS);
6497 return (DDI_FAILURE);
6499 #endif /* illumos */
6502 * OK, so this is a little weird.
6504 * /dev/zfs is the control node, i.e. minor 0.
6505 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
6507 * /dev/zfs has basically nothing to do except serve up ioctls,
6508 * so most of the standard driver entry points are in zvol.c.
6511 static struct cb_ops zfs_cb_ops = {
6512 zfsdev_open, /* open */
6513 zfsdev_close, /* close */
6514 zvol_strategy, /* strategy */
6516 zvol_dump, /* dump */
6517 zvol_read, /* read */
6518 zvol_write, /* write */
6519 zfsdev_ioctl, /* ioctl */
6523 nochpoll, /* poll */
6524 ddi_prop_op, /* prop_op */
6525 NULL, /* streamtab */
6526 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */
6527 CB_REV, /* version */
6528 nodev, /* async read */
6529 nodev, /* async write */
6532 static struct dev_ops zfs_dev_ops = {
6533 DEVO_REV, /* version */
6535 zfs_info, /* info */
6536 nulldev, /* identify */
6537 nulldev, /* probe */
6538 zfs_attach, /* attach */
6539 zfs_detach, /* detach */
6541 &zfs_cb_ops, /* driver operations */
6542 NULL, /* no bus operations */
6544 ddi_quiesce_not_needed, /* quiesce */
6547 static struct modldrv zfs_modldrv = {
6553 static struct modlinkage modlinkage = {
6555 (void *)&zfs_modlfs,
6556 (void *)&zfs_modldrv,
6559 #endif /* illumos */
6561 static struct cdevsw zfs_cdevsw = {
6562 .d_version = D_VERSION,
6563 .d_open = zfsdev_open,
6564 .d_ioctl = zfsdev_ioctl,
6565 .d_name = ZFS_DEV_NAME
6569 zfs_allow_log_destroy(void *arg)
6571 char *poolname = arg;
6578 zfsdev = make_dev(&zfs_cdevsw, 0x0, UID_ROOT, GID_OPERATOR, 0666,
6586 destroy_dev(zfsdev);
6589 static struct root_hold_token *zfs_root_token;
6590 struct proc *zfsproc;
6598 spa_init(FREAD | FWRITE);
6603 if ((error = mod_install(&modlinkage)) != 0) {
6610 tsd_create(&zfs_fsyncer_key, NULL);
6611 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
6612 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
6614 error = ldi_ident_from_mod(&modlinkage, &zfs_li);
6616 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
6626 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
6627 return (SET_ERROR(EBUSY));
6629 if ((error = mod_remove(&modlinkage)) != 0)
6635 if (zfs_nfsshare_inited)
6636 (void) ddi_modclose(nfs_mod);
6637 if (zfs_smbshare_inited)
6638 (void) ddi_modclose(smbsrv_mod);
6639 if (zfs_nfsshare_inited || zfs_smbshare_inited)
6640 (void) ddi_modclose(sharefs_mod);
6642 tsd_destroy(&zfs_fsyncer_key);
6643 ldi_ident_release(zfs_li);
6645 mutex_destroy(&zfs_share_lock);
6651 _info(struct modinfo *modinfop)
6653 return (mod_info(&modlinkage, modinfop));
6655 #endif /* illumos */
6657 static int zfs__init(void);
6658 static int zfs__fini(void);
6659 static void zfs_shutdown(void *, int);
6661 static eventhandler_tag zfs_shutdown_event_tag;
6664 #define ZFS_MIN_KSTACK_PAGES 4
6672 #if KSTACK_PAGES < ZFS_MIN_KSTACK_PAGES
6673 printf("ZFS NOTICE: KSTACK_PAGES is %d which could result in stack "
6674 "overflow panic!\nPlease consider adding "
6675 "'options KSTACK_PAGES=%d' to your kernel config\n", KSTACK_PAGES,
6676 ZFS_MIN_KSTACK_PAGES);
6679 zfs_root_token = root_mount_hold("ZFS");
6681 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
6683 spa_init(FREAD | FWRITE);
6688 tsd_create(&zfs_fsyncer_key, NULL);
6689 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
6690 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
6692 printf("ZFS storage pool version: features support (" SPA_VERSION_STRING ")\n");
6693 root_mount_rel(zfs_root_token);
6703 if (spa_busy() || zfs_busy() || zvol_busy() ||
6704 zio_injection_enabled) {
6713 tsd_destroy(&zfs_fsyncer_key);
6714 tsd_destroy(&rrw_tsd_key);
6715 tsd_destroy(&zfs_allow_log_key);
6717 mutex_destroy(&zfs_share_lock);
6723 zfs_shutdown(void *arg __unused, int howto __unused)
6727 * ZFS fini routines can not properly work in a panic-ed system.
6729 if (panicstr == NULL)
6735 zfs_modevent(module_t mod, int type, void *unused __unused)
6743 zfs_shutdown_event_tag = EVENTHANDLER_REGISTER(
6744 shutdown_post_sync, zfs_shutdown, NULL,
6745 SHUTDOWN_PRI_FIRST);
6749 if (err == 0 && zfs_shutdown_event_tag != NULL)
6750 EVENTHANDLER_DEREGISTER(shutdown_post_sync,
6751 zfs_shutdown_event_tag);
6758 return (EOPNOTSUPP);
6761 static moduledata_t zfs_mod = {
6766 DECLARE_MODULE(zfsctrl, zfs_mod, SI_SUB_VFS, SI_ORDER_ANY);
6767 MODULE_VERSION(zfsctrl, 1);
6768 MODULE_DEPEND(zfsctrl, opensolaris, 1, 1, 1);
6769 MODULE_DEPEND(zfsctrl, krpc, 1, 1, 1);
6770 MODULE_DEPEND(zfsctrl, acl_nfs4, 1, 1, 1);