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 * Portions Copyright 2011 Martin Matuska
25 * Copyright 2015, OmniTI Computer Consulting, Inc. All rights reserved.
26 * Portions Copyright 2012 Pawel Jakub Dawidek <pawel@dawidek.net>
27 * Copyright (c) 2014, 2016 Joyent, Inc. All rights reserved.
28 * Copyright 2016 Nexenta Systems, Inc. All rights reserved.
29 * Copyright (c) 2014, Joyent, Inc. All rights reserved.
30 * Copyright (c) 2011, 2020 by Delphix. All rights reserved.
31 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
32 * Copyright (c) 2013 Steven Hartland. All rights reserved.
33 * Copyright (c) 2014 Integros [integros.com]
34 * Copyright 2016 Toomas Soome <tsoome@me.com>
35 * Copyright (c) 2016 Actifio, Inc. All rights reserved.
36 * Copyright (c) 2018, loli10K <ezomori.nozomu@gmail.com>. All rights reserved.
37 * Copyright 2017 RackTop Systems.
38 * Copyright (c) 2017 Open-E, Inc. All Rights Reserved.
39 * Copyright (c) 2019 Datto Inc.
40 * Copyright (c) 2019, 2020 by Christian Schwarz. All rights reserved.
46 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
47 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
49 * There are two ways that we handle ioctls: the legacy way where almost
50 * all of the logic is in the ioctl callback, and the new way where most
51 * of the marshalling is handled in the common entry point, zfsdev_ioctl().
53 * Non-legacy ioctls should be registered by calling
54 * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked
55 * from userland by lzc_ioctl().
57 * The registration arguments are as follows:
60 * The name of the ioctl. This is used for history logging. If the
61 * ioctl returns successfully (the callback returns 0), and allow_log
62 * is true, then a history log entry will be recorded with the input &
63 * output nvlists. The log entry can be printed with "zpool history -i".
66 * The ioctl request number, which userland will pass to ioctl(2).
67 * We want newer versions of libzfs and libzfs_core to run against
68 * existing zfs kernel modules (i.e. a deferred reboot after an update).
69 * Therefore the ioctl numbers cannot change from release to release.
71 * zfs_secpolicy_func_t *secpolicy
72 * This function will be called before the zfs_ioc_func_t, to
73 * determine if this operation is permitted. It should return EPERM
74 * on failure, and 0 on success. Checks include determining if the
75 * dataset is visible in this zone, and if the user has either all
76 * zfs privileges in the zone (SYS_MOUNT), or has been granted permission
77 * to do this operation on this dataset with "zfs allow".
79 * zfs_ioc_namecheck_t namecheck
80 * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
81 * name, a dataset name, or nothing. If the name is not well-formed,
82 * the ioctl will fail and the callback will not be called.
83 * Therefore, the callback can assume that the name is well-formed
84 * (e.g. is null-terminated, doesn't have more than one '@' character,
85 * doesn't have invalid characters).
87 * zfs_ioc_poolcheck_t pool_check
88 * This specifies requirements on the pool state. If the pool does
89 * not meet them (is suspended or is readonly), the ioctl will fail
90 * and the callback will not be called. If any checks are specified
91 * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
92 * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED |
93 * POOL_CHECK_READONLY).
95 * zfs_ioc_key_t *nvl_keys
96 * The list of expected/allowable innvl input keys. This list is used
97 * to validate the nvlist input to the ioctl.
99 * boolean_t smush_outnvlist
100 * If smush_outnvlist is true, then the output is presumed to be a
101 * list of errors, and it will be "smushed" down to fit into the
102 * caller's buffer, by removing some entries and replacing them with a
103 * single "N_MORE_ERRORS" entry indicating how many were removed. See
104 * nvlist_smush() for details. If smush_outnvlist is false, and the
105 * outnvlist does not fit into the userland-provided buffer, then the
106 * ioctl will fail with ENOMEM.
108 * zfs_ioc_func_t *func
109 * The callback function that will perform the operation.
111 * The callback should return 0 on success, or an error number on
112 * failure. If the function fails, the userland ioctl will return -1,
113 * and errno will be set to the callback's return value. The callback
114 * will be called with the following arguments:
117 * The name of the pool or dataset to operate on, from
118 * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the
119 * expected type (pool, dataset, or none).
122 * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or
123 * NULL if no input nvlist was provided. Changes to this nvlist are
124 * ignored. If the input nvlist could not be deserialized, the
125 * ioctl will fail and the callback will not be called.
128 * The output nvlist, initially empty. The callback can fill it in,
129 * and it will be returned to userland by serializing it into
130 * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization
131 * fails (e.g. because the caller didn't supply a large enough
132 * buffer), then the overall ioctl will fail. See the
133 * 'smush_nvlist' argument above for additional behaviors.
135 * There are two typical uses of the output nvlist:
136 * - To return state, e.g. property values. In this case,
137 * smush_outnvlist should be false. If the buffer was not large
138 * enough, the caller will reallocate a larger buffer and try
141 * - To return multiple errors from an ioctl which makes on-disk
142 * changes. In this case, smush_outnvlist should be true.
143 * Ioctls which make on-disk modifications should generally not
144 * use the outnvl if they succeed, because the caller can not
145 * distinguish between the operation failing, and
146 * deserialization failing.
148 * IOCTL Interface Errors
150 * The following ioctl input errors can be returned:
151 * ZFS_ERR_IOC_CMD_UNAVAIL the ioctl number is not supported by kernel
152 * ZFS_ERR_IOC_ARG_UNAVAIL an input argument is not supported by kernel
153 * ZFS_ERR_IOC_ARG_REQUIRED a required input argument is missing
154 * ZFS_ERR_IOC_ARG_BADTYPE an input argument has an invalid type
157 #include <sys/types.h>
158 #include <sys/param.h>
159 #include <sys/errno.h>
161 #include <sys/file.h>
162 #include <sys/kmem.h>
163 #include <sys/cmn_err.h>
164 #include <sys/stat.h>
165 #include <sys/zfs_ioctl.h>
166 #include <sys/zfs_quota.h>
167 #include <sys/zfs_vfsops.h>
168 #include <sys/zfs_znode.h>
171 #include <sys/spa_impl.h>
172 #include <sys/vdev.h>
173 #include <sys/vdev_impl.h>
175 #include <sys/dsl_dir.h>
176 #include <sys/dsl_dataset.h>
177 #include <sys/dsl_prop.h>
178 #include <sys/dsl_deleg.h>
179 #include <sys/dmu_objset.h>
180 #include <sys/dmu_impl.h>
181 #include <sys/dmu_redact.h>
182 #include <sys/dmu_tx.h>
183 #include <sys/sunddi.h>
184 #include <sys/policy.h>
185 #include <sys/zone.h>
186 #include <sys/nvpair.h>
187 #include <sys/pathname.h>
188 #include <sys/fs/zfs.h>
189 #include <sys/zfs_ctldir.h>
190 #include <sys/zfs_dir.h>
191 #include <sys/zfs_onexit.h>
192 #include <sys/zvol.h>
193 #include <sys/dsl_scan.h>
194 #include <sys/fm/util.h>
195 #include <sys/dsl_crypt.h>
196 #include <sys/rrwlock.h>
197 #include <sys/zfs_file.h>
199 #include <sys/dmu_recv.h>
200 #include <sys/dmu_send.h>
201 #include <sys/dmu_recv.h>
202 #include <sys/dsl_destroy.h>
203 #include <sys/dsl_bookmark.h>
204 #include <sys/dsl_userhold.h>
205 #include <sys/zfeature.h>
207 #include <sys/zio_checksum.h>
208 #include <sys/vdev_removal.h>
209 #include <sys/vdev_impl.h>
210 #include <sys/vdev_initialize.h>
211 #include <sys/vdev_trim.h>
213 #include "zfs_namecheck.h"
214 #include "zfs_prop.h"
215 #include "zfs_deleg.h"
216 #include "zfs_comutil.h"
218 #include <sys/lua/lua.h>
219 #include <sys/lua/lauxlib.h>
220 #include <sys/zfs_ioctl_impl.h>
222 kmutex_t zfsdev_state_lock;
223 zfsdev_state_t *zfsdev_state_list;
226 * Limit maximum nvlist size. We don't want users passing in insane values
227 * for zc->zc_nvlist_src_size, since we will need to allocate that much memory.
229 #define MAX_NVLIST_SRC_SIZE KMALLOC_MAX_SIZE
231 uint_t zfs_fsyncer_key;
232 uint_t zfs_allow_log_key;
234 /* DATA_TYPE_ANY is used when zkey_type can vary. */
235 #define DATA_TYPE_ANY DATA_TYPE_UNKNOWN
237 typedef struct zfs_ioc_vec {
238 zfs_ioc_legacy_func_t *zvec_legacy_func;
239 zfs_ioc_func_t *zvec_func;
240 zfs_secpolicy_func_t *zvec_secpolicy;
241 zfs_ioc_namecheck_t zvec_namecheck;
242 boolean_t zvec_allow_log;
243 zfs_ioc_poolcheck_t zvec_pool_check;
244 boolean_t zvec_smush_outnvlist;
245 const char *zvec_name;
246 const zfs_ioc_key_t *zvec_nvl_keys;
247 size_t zvec_nvl_key_count;
250 /* This array is indexed by zfs_userquota_prop_t */
251 static const char *userquota_perms[] = {
252 ZFS_DELEG_PERM_USERUSED,
253 ZFS_DELEG_PERM_USERQUOTA,
254 ZFS_DELEG_PERM_GROUPUSED,
255 ZFS_DELEG_PERM_GROUPQUOTA,
256 ZFS_DELEG_PERM_USEROBJUSED,
257 ZFS_DELEG_PERM_USEROBJQUOTA,
258 ZFS_DELEG_PERM_GROUPOBJUSED,
259 ZFS_DELEG_PERM_GROUPOBJQUOTA,
260 ZFS_DELEG_PERM_PROJECTUSED,
261 ZFS_DELEG_PERM_PROJECTQUOTA,
262 ZFS_DELEG_PERM_PROJECTOBJUSED,
263 ZFS_DELEG_PERM_PROJECTOBJQUOTA,
266 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
267 static int zfs_ioc_id_quota_upgrade(zfs_cmd_t *zc);
268 static int zfs_check_settable(const char *name, nvpair_t *property,
270 static int zfs_check_clearable(char *dataset, nvlist_t *props,
272 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
274 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
275 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
278 history_str_free(char *buf)
280 kmem_free(buf, HIS_MAX_RECORD_LEN);
284 history_str_get(zfs_cmd_t *zc)
288 if (zc->zc_history == 0)
291 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
292 if (copyinstr((void *)(uintptr_t)zc->zc_history,
293 buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
294 history_str_free(buf);
298 buf[HIS_MAX_RECORD_LEN -1] = '\0';
304 * Check to see if the named dataset is currently defined as bootable
307 zfs_is_bootfs(const char *name)
311 if (dmu_objset_hold(name, FTAG, &os) == 0) {
313 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
314 dmu_objset_rele(os, FTAG);
321 * Return non-zero if the spa version is less than requested version.
324 zfs_earlier_version(const char *name, int version)
328 if (spa_open(name, &spa, FTAG) == 0) {
329 if (spa_version(spa) < version) {
330 spa_close(spa, FTAG);
333 spa_close(spa, FTAG);
339 * Return TRUE if the ZPL version is less than requested version.
342 zpl_earlier_version(const char *name, int version)
345 boolean_t rc = B_TRUE;
347 if (dmu_objset_hold(name, FTAG, &os) == 0) {
350 if (dmu_objset_type(os) != DMU_OST_ZFS) {
351 dmu_objset_rele(os, FTAG);
354 /* XXX reading from non-owned objset */
355 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
356 rc = zplversion < version;
357 dmu_objset_rele(os, FTAG);
363 zfs_log_history(zfs_cmd_t *zc)
368 if ((buf = history_str_get(zc)) == NULL)
371 if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
372 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
373 (void) spa_history_log(spa, buf);
374 spa_close(spa, FTAG);
376 history_str_free(buf);
380 * Policy for top-level read operations (list pools). Requires no privileges,
381 * and can be used in the local zone, as there is no associated dataset.
385 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
391 * Policy for dataset read operations (list children, get statistics). Requires
392 * no privileges, but must be visible in the local zone.
396 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
398 if (INGLOBALZONE(curproc) ||
399 zone_dataset_visible(zc->zc_name, NULL))
402 return (SET_ERROR(ENOENT));
406 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
411 * The dataset must be visible by this zone -- check this first
412 * so they don't see EPERM on something they shouldn't know about.
414 if (!INGLOBALZONE(curproc) &&
415 !zone_dataset_visible(dataset, &writable))
416 return (SET_ERROR(ENOENT));
418 if (INGLOBALZONE(curproc)) {
420 * If the fs is zoned, only root can access it from the
423 if (secpolicy_zfs(cr) && zoned)
424 return (SET_ERROR(EPERM));
427 * If we are in a local zone, the 'zoned' property must be set.
430 return (SET_ERROR(EPERM));
432 /* must be writable by this zone */
434 return (SET_ERROR(EPERM));
440 zfs_dozonecheck(const char *dataset, cred_t *cr)
444 if (dsl_prop_get_integer(dataset, zfs_prop_to_name(ZFS_PROP_ZONED),
446 return (SET_ERROR(ENOENT));
448 return (zfs_dozonecheck_impl(dataset, zoned, cr));
452 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
456 if (dsl_prop_get_int_ds(ds, zfs_prop_to_name(ZFS_PROP_ZONED), &zoned))
457 return (SET_ERROR(ENOENT));
459 return (zfs_dozonecheck_impl(dataset, zoned, cr));
463 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
464 const char *perm, cred_t *cr)
468 error = zfs_dozonecheck_ds(name, ds, cr);
470 error = secpolicy_zfs(cr);
472 error = dsl_deleg_access_impl(ds, perm, cr);
478 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
485 * First do a quick check for root in the global zone, which
486 * is allowed to do all write_perms. This ensures that zfs_ioc_*
487 * will get to handle nonexistent datasets.
489 if (INGLOBALZONE(curproc) && secpolicy_zfs(cr) == 0)
492 error = dsl_pool_hold(name, FTAG, &dp);
496 error = dsl_dataset_hold(dp, name, FTAG, &ds);
498 dsl_pool_rele(dp, FTAG);
502 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
504 dsl_dataset_rele(ds, FTAG);
505 dsl_pool_rele(dp, FTAG);
510 * Policy for setting the security label property.
512 * Returns 0 for success, non-zero for access and other errors.
515 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
518 char ds_hexsl[MAXNAMELEN];
519 bslabel_t ds_sl, new_sl;
520 boolean_t new_default = FALSE;
522 int needed_priv = -1;
525 /* First get the existing dataset label. */
526 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
527 1, sizeof (ds_hexsl), &ds_hexsl, NULL);
529 return (SET_ERROR(EPERM));
531 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
534 /* The label must be translatable */
535 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
536 return (SET_ERROR(EINVAL));
539 * In a non-global zone, disallow attempts to set a label that
540 * doesn't match that of the zone; otherwise no other checks
543 if (!INGLOBALZONE(curproc)) {
544 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
545 return (SET_ERROR(EPERM));
550 * For global-zone datasets (i.e., those whose zoned property is
551 * "off", verify that the specified new label is valid for the
554 if (dsl_prop_get_integer(name,
555 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
556 return (SET_ERROR(EPERM));
558 if (zfs_check_global_label(name, strval) != 0)
559 return (SET_ERROR(EPERM));
563 * If the existing dataset label is nondefault, check if the
564 * dataset is mounted (label cannot be changed while mounted).
565 * Get the zfsvfs_t; if there isn't one, then the dataset isn't
566 * mounted (or isn't a dataset, doesn't exist, ...).
568 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
570 static char *setsl_tag = "setsl_tag";
573 * Try to own the dataset; abort if there is any error,
574 * (e.g., already mounted, in use, or other error).
576 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE, B_TRUE,
579 return (SET_ERROR(EPERM));
581 dmu_objset_disown(os, B_TRUE, setsl_tag);
584 needed_priv = PRIV_FILE_DOWNGRADE_SL;
588 if (hexstr_to_label(strval, &new_sl) != 0)
589 return (SET_ERROR(EPERM));
591 if (blstrictdom(&ds_sl, &new_sl))
592 needed_priv = PRIV_FILE_DOWNGRADE_SL;
593 else if (blstrictdom(&new_sl, &ds_sl))
594 needed_priv = PRIV_FILE_UPGRADE_SL;
596 /* dataset currently has a default label */
598 needed_priv = PRIV_FILE_UPGRADE_SL;
602 if (needed_priv != -1)
603 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
606 return (SET_ERROR(ENOTSUP));
607 #endif /* HAVE_MLSLABEL */
611 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
617 * Check permissions for special properties.
624 * Disallow setting of 'zoned' from within a local zone.
626 if (!INGLOBALZONE(curproc))
627 return (SET_ERROR(EPERM));
631 case ZFS_PROP_FILESYSTEM_LIMIT:
632 case ZFS_PROP_SNAPSHOT_LIMIT:
633 if (!INGLOBALZONE(curproc)) {
635 char setpoint[ZFS_MAX_DATASET_NAME_LEN];
637 * Unprivileged users are allowed to modify the
638 * limit on things *under* (ie. contained by)
639 * the thing they own.
641 if (dsl_prop_get_integer(dsname,
642 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, setpoint))
643 return (SET_ERROR(EPERM));
644 if (!zoned || strlen(dsname) <= strlen(setpoint))
645 return (SET_ERROR(EPERM));
649 case ZFS_PROP_MLSLABEL:
650 if (!is_system_labeled())
651 return (SET_ERROR(EPERM));
653 if (nvpair_value_string(propval, &strval) == 0) {
656 err = zfs_set_slabel_policy(dsname, strval, CRED());
663 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
668 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
672 error = zfs_dozonecheck(zc->zc_name, cr);
677 * permission to set permissions will be evaluated later in
678 * dsl_deleg_can_allow()
685 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
687 return (zfs_secpolicy_write_perms(zc->zc_name,
688 ZFS_DELEG_PERM_ROLLBACK, cr));
693 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
701 * Generate the current snapshot name from the given objsetid, then
702 * use that name for the secpolicy/zone checks.
704 cp = strchr(zc->zc_name, '@');
706 return (SET_ERROR(EINVAL));
707 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
711 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
713 dsl_pool_rele(dp, FTAG);
717 dsl_dataset_name(ds, zc->zc_name);
719 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
720 ZFS_DELEG_PERM_SEND, cr);
721 dsl_dataset_rele(ds, FTAG);
722 dsl_pool_rele(dp, FTAG);
729 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
731 return (zfs_secpolicy_write_perms(zc->zc_name,
732 ZFS_DELEG_PERM_SEND, cr));
736 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
738 return (SET_ERROR(ENOTSUP));
742 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
744 return (SET_ERROR(ENOTSUP));
748 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
753 * Remove the @bla or /bla from the end of the name to get the parent.
755 (void) strncpy(parent, datasetname, parentsize);
756 cp = strrchr(parent, '@');
760 cp = strrchr(parent, '/');
762 return (SET_ERROR(ENOENT));
770 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
774 if ((error = zfs_secpolicy_write_perms(name,
775 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
778 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
783 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
785 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
789 * Destroying snapshots with delegated permissions requires
790 * descendant mount and destroy permissions.
794 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
797 nvpair_t *pair, *nextpair;
800 snaps = fnvlist_lookup_nvlist(innvl, "snaps");
802 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
804 nextpair = nvlist_next_nvpair(snaps, pair);
805 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr);
806 if (error == ENOENT) {
808 * Ignore any snapshots that don't exist (we consider
809 * them "already destroyed"). Remove the name from the
810 * nvl here in case the snapshot is created between
811 * now and when we try to destroy it (in which case
812 * we don't want to destroy it since we haven't
813 * checked for permission).
815 fnvlist_remove_nvpair(snaps, pair);
826 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
828 char parentname[ZFS_MAX_DATASET_NAME_LEN];
831 if ((error = zfs_secpolicy_write_perms(from,
832 ZFS_DELEG_PERM_RENAME, cr)) != 0)
835 if ((error = zfs_secpolicy_write_perms(from,
836 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
839 if ((error = zfs_get_parent(to, parentname,
840 sizeof (parentname))) != 0)
843 if ((error = zfs_secpolicy_write_perms(parentname,
844 ZFS_DELEG_PERM_CREATE, cr)) != 0)
847 if ((error = zfs_secpolicy_write_perms(parentname,
848 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
856 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
858 return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
863 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
866 dsl_dataset_t *clone;
869 error = zfs_secpolicy_write_perms(zc->zc_name,
870 ZFS_DELEG_PERM_PROMOTE, cr);
874 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
878 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
881 char parentname[ZFS_MAX_DATASET_NAME_LEN];
882 dsl_dataset_t *origin = NULL;
886 error = dsl_dataset_hold_obj(dd->dd_pool,
887 dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin);
889 dsl_dataset_rele(clone, FTAG);
890 dsl_pool_rele(dp, FTAG);
894 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
895 ZFS_DELEG_PERM_MOUNT, cr);
897 dsl_dataset_name(origin, parentname);
899 error = zfs_secpolicy_write_perms_ds(parentname, origin,
900 ZFS_DELEG_PERM_PROMOTE, cr);
902 dsl_dataset_rele(clone, FTAG);
903 dsl_dataset_rele(origin, FTAG);
905 dsl_pool_rele(dp, FTAG);
911 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
915 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
916 ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
919 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
920 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
923 return (zfs_secpolicy_write_perms(zc->zc_name,
924 ZFS_DELEG_PERM_CREATE, cr));
929 zfs_secpolicy_recv_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
931 return (zfs_secpolicy_recv(zc, innvl, cr));
935 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
937 return (zfs_secpolicy_write_perms(name,
938 ZFS_DELEG_PERM_SNAPSHOT, cr));
942 * Check for permission to create each snapshot in the nvlist.
946 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
952 snaps = fnvlist_lookup_nvlist(innvl, "snaps");
954 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
955 pair = nvlist_next_nvpair(snaps, pair)) {
956 char *name = nvpair_name(pair);
957 char *atp = strchr(name, '@');
960 error = SET_ERROR(EINVAL);
964 error = zfs_secpolicy_snapshot_perms(name, cr);
973 * Check for permission to create each bookmark in the nvlist.
977 zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
981 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
982 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
983 char *name = nvpair_name(pair);
984 char *hashp = strchr(name, '#');
987 error = SET_ERROR(EINVAL);
991 error = zfs_secpolicy_write_perms(name,
992 ZFS_DELEG_PERM_BOOKMARK, cr);
1002 zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1004 nvpair_t *pair, *nextpair;
1007 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1009 char *name = nvpair_name(pair);
1010 char *hashp = strchr(name, '#');
1011 nextpair = nvlist_next_nvpair(innvl, pair);
1013 if (hashp == NULL) {
1014 error = SET_ERROR(EINVAL);
1019 error = zfs_secpolicy_write_perms(name,
1020 ZFS_DELEG_PERM_DESTROY, cr);
1022 if (error == ENOENT) {
1024 * Ignore any filesystems that don't exist (we consider
1025 * their bookmarks "already destroyed"). Remove
1026 * the name from the nvl here in case the filesystem
1027 * is created between now and when we try to destroy
1028 * the bookmark (in which case we don't want to
1029 * destroy it since we haven't checked for permission).
1031 fnvlist_remove_nvpair(innvl, pair);
1043 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1046 * Even root must have a proper TSD so that we know what pool
1049 if (tsd_get(zfs_allow_log_key) == NULL)
1050 return (SET_ERROR(EPERM));
1055 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1057 char parentname[ZFS_MAX_DATASET_NAME_LEN];
1061 if ((error = zfs_get_parent(zc->zc_name, parentname,
1062 sizeof (parentname))) != 0)
1065 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1066 (error = zfs_secpolicy_write_perms(origin,
1067 ZFS_DELEG_PERM_CLONE, cr)) != 0)
1070 if ((error = zfs_secpolicy_write_perms(parentname,
1071 ZFS_DELEG_PERM_CREATE, cr)) != 0)
1074 return (zfs_secpolicy_write_perms(parentname,
1075 ZFS_DELEG_PERM_MOUNT, cr));
1079 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires
1080 * SYS_CONFIG privilege, which is not available in a local zone.
1084 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1086 if (secpolicy_sys_config(cr, B_FALSE) != 0)
1087 return (SET_ERROR(EPERM));
1093 * Policy for object to name lookups.
1097 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1101 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1104 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1109 * Policy for fault injection. Requires all privileges.
1113 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1115 return (secpolicy_zinject(cr));
1120 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1122 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1124 if (prop == ZPROP_INVAL) {
1125 if (!zfs_prop_user(zc->zc_value))
1126 return (SET_ERROR(EINVAL));
1127 return (zfs_secpolicy_write_perms(zc->zc_name,
1128 ZFS_DELEG_PERM_USERPROP, cr));
1130 return (zfs_secpolicy_setprop(zc->zc_name, prop,
1136 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1138 int err = zfs_secpolicy_read(zc, innvl, cr);
1142 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1143 return (SET_ERROR(EINVAL));
1145 if (zc->zc_value[0] == 0) {
1147 * They are asking about a posix uid/gid. If it's
1148 * themself, allow it.
1150 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1151 zc->zc_objset_type == ZFS_PROP_USERQUOTA ||
1152 zc->zc_objset_type == ZFS_PROP_USEROBJUSED ||
1153 zc->zc_objset_type == ZFS_PROP_USEROBJQUOTA) {
1154 if (zc->zc_guid == crgetuid(cr))
1156 } else if (zc->zc_objset_type == ZFS_PROP_GROUPUSED ||
1157 zc->zc_objset_type == ZFS_PROP_GROUPQUOTA ||
1158 zc->zc_objset_type == ZFS_PROP_GROUPOBJUSED ||
1159 zc->zc_objset_type == ZFS_PROP_GROUPOBJQUOTA) {
1160 if (groupmember(zc->zc_guid, cr))
1163 /* else is for project quota/used */
1166 return (zfs_secpolicy_write_perms(zc->zc_name,
1167 userquota_perms[zc->zc_objset_type], cr));
1171 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1173 int err = zfs_secpolicy_read(zc, innvl, cr);
1177 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1178 return (SET_ERROR(EINVAL));
1180 return (zfs_secpolicy_write_perms(zc->zc_name,
1181 userquota_perms[zc->zc_objset_type], cr));
1186 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1188 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1194 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1200 holds = fnvlist_lookup_nvlist(innvl, "holds");
1202 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
1203 pair = nvlist_next_nvpair(holds, pair)) {
1204 char fsname[ZFS_MAX_DATASET_NAME_LEN];
1205 error = dmu_fsname(nvpair_name(pair), fsname);
1208 error = zfs_secpolicy_write_perms(fsname,
1209 ZFS_DELEG_PERM_HOLD, cr);
1218 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1223 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1224 pair = nvlist_next_nvpair(innvl, pair)) {
1225 char fsname[ZFS_MAX_DATASET_NAME_LEN];
1226 error = dmu_fsname(nvpair_name(pair), fsname);
1229 error = zfs_secpolicy_write_perms(fsname,
1230 ZFS_DELEG_PERM_RELEASE, cr);
1238 * Policy for allowing temporary snapshots to be taken or released
1241 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1244 * A temporary snapshot is the same as a snapshot,
1245 * hold, destroy and release all rolled into one.
1246 * Delegated diff alone is sufficient that we allow this.
1250 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1251 ZFS_DELEG_PERM_DIFF, cr)) == 0)
1254 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1256 if (innvl != NULL) {
1258 error = zfs_secpolicy_hold(zc, innvl, cr);
1260 error = zfs_secpolicy_release(zc, innvl, cr);
1262 error = zfs_secpolicy_destroy(zc, innvl, cr);
1268 zfs_secpolicy_load_key(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1270 return (zfs_secpolicy_write_perms(zc->zc_name,
1271 ZFS_DELEG_PERM_LOAD_KEY, cr));
1275 zfs_secpolicy_change_key(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1277 return (zfs_secpolicy_write_perms(zc->zc_name,
1278 ZFS_DELEG_PERM_CHANGE_KEY, cr));
1282 * Returns the nvlist as specified by the user in the zfs_cmd_t.
1285 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1289 nvlist_t *list = NULL;
1292 * Read in and unpack the user-supplied nvlist.
1295 return (SET_ERROR(EINVAL));
1297 packed = vmem_alloc(size, KM_SLEEP);
1299 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1301 vmem_free(packed, size);
1302 return (SET_ERROR(EFAULT));
1305 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1306 vmem_free(packed, size);
1310 vmem_free(packed, size);
1317 * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1318 * Entries will be removed from the end of the nvlist, and one int32 entry
1319 * named "N_MORE_ERRORS" will be added indicating how many entries were
1323 nvlist_smush(nvlist_t *errors, size_t max)
1327 size = fnvlist_size(errors);
1330 nvpair_t *more_errors;
1334 return (SET_ERROR(ENOMEM));
1336 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1337 more_errors = nvlist_prev_nvpair(errors, NULL);
1340 nvpair_t *pair = nvlist_prev_nvpair(errors,
1342 fnvlist_remove_nvpair(errors, pair);
1344 size = fnvlist_size(errors);
1345 } while (size > max);
1347 fnvlist_remove_nvpair(errors, more_errors);
1348 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1349 ASSERT3U(fnvlist_size(errors), <=, max);
1356 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1358 char *packed = NULL;
1362 size = fnvlist_size(nvl);
1364 if (size > zc->zc_nvlist_dst_size) {
1365 error = SET_ERROR(ENOMEM);
1367 packed = fnvlist_pack(nvl, &size);
1368 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1369 size, zc->zc_iflags) != 0)
1370 error = SET_ERROR(EFAULT);
1371 fnvlist_pack_free(packed, size);
1374 zc->zc_nvlist_dst_size = size;
1375 zc->zc_nvlist_dst_filled = B_TRUE;
1380 getzfsvfs_impl(objset_t *os, zfsvfs_t **zfvp)
1383 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1384 return (SET_ERROR(EINVAL));
1387 mutex_enter(&os->os_user_ptr_lock);
1388 *zfvp = dmu_objset_get_user(os);
1389 /* bump s_active only when non-zero to prevent umount race */
1390 error = zfs_vfs_ref(zfvp);
1391 mutex_exit(&os->os_user_ptr_lock);
1396 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1401 error = dmu_objset_hold(dsname, FTAG, &os);
1405 error = getzfsvfs_impl(os, zfvp);
1406 dmu_objset_rele(os, FTAG);
1411 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1412 * case its z_sb will be NULL, and it will be opened as the owner.
1413 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1414 * which prevents all inode ops from running.
1417 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1421 if (getzfsvfs(name, zfvp) != 0)
1422 error = zfsvfs_create(name, B_FALSE, zfvp);
1424 rrm_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER :
1426 if ((*zfvp)->z_unmounted) {
1428 * XXX we could probably try again, since the unmounting
1429 * thread should be just about to disassociate the
1430 * objset from the zfsvfs.
1432 rrm_exit(&(*zfvp)->z_teardown_lock, tag);
1433 return (SET_ERROR(EBUSY));
1440 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1442 rrm_exit(&zfsvfs->z_teardown_lock, tag);
1444 if (zfs_vfs_held(zfsvfs)) {
1445 zfs_vfs_rele(zfsvfs);
1447 dmu_objset_disown(zfsvfs->z_os, B_TRUE, zfsvfs);
1448 zfsvfs_free(zfsvfs);
1453 zfs_ioc_pool_create(zfs_cmd_t *zc)
1456 nvlist_t *config, *props = NULL;
1457 nvlist_t *rootprops = NULL;
1458 nvlist_t *zplprops = NULL;
1459 dsl_crypto_params_t *dcp = NULL;
1460 char *spa_name = zc->zc_name;
1461 boolean_t unload_wkey = B_TRUE;
1463 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1464 zc->zc_iflags, &config)))
1467 if (zc->zc_nvlist_src_size != 0 && (error =
1468 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1469 zc->zc_iflags, &props))) {
1470 nvlist_free(config);
1475 nvlist_t *nvl = NULL;
1476 nvlist_t *hidden_args = NULL;
1477 uint64_t version = SPA_VERSION;
1480 (void) nvlist_lookup_uint64(props,
1481 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1482 if (!SPA_VERSION_IS_SUPPORTED(version)) {
1483 error = SET_ERROR(EINVAL);
1484 goto pool_props_bad;
1486 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1488 error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1490 goto pool_props_bad;
1491 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1494 (void) nvlist_lookup_nvlist(props, ZPOOL_HIDDEN_ARGS,
1496 error = dsl_crypto_params_create_nvlist(DCP_CMD_NONE,
1497 rootprops, hidden_args, &dcp);
1499 goto pool_props_bad;
1500 (void) nvlist_remove_all(props, ZPOOL_HIDDEN_ARGS);
1502 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1503 error = zfs_fill_zplprops_root(version, rootprops,
1506 goto pool_props_bad;
1508 if (nvlist_lookup_string(props,
1509 zpool_prop_to_name(ZPOOL_PROP_TNAME), &tname) == 0)
1513 error = spa_create(zc->zc_name, config, props, zplprops, dcp);
1516 * Set the remaining root properties
1518 if (!error && (error = zfs_set_prop_nvlist(spa_name,
1519 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0) {
1520 (void) spa_destroy(spa_name);
1521 unload_wkey = B_FALSE; /* spa_destroy() unloads wrapping keys */
1525 nvlist_free(rootprops);
1526 nvlist_free(zplprops);
1527 nvlist_free(config);
1529 dsl_crypto_params_free(dcp, unload_wkey && !!error);
1535 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1538 zfs_log_history(zc);
1539 error = spa_destroy(zc->zc_name);
1545 zfs_ioc_pool_import(zfs_cmd_t *zc)
1547 nvlist_t *config, *props = NULL;
1551 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1552 zc->zc_iflags, &config)) != 0)
1555 if (zc->zc_nvlist_src_size != 0 && (error =
1556 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1557 zc->zc_iflags, &props))) {
1558 nvlist_free(config);
1562 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1563 guid != zc->zc_guid)
1564 error = SET_ERROR(EINVAL);
1566 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1568 if (zc->zc_nvlist_dst != 0) {
1571 if ((err = put_nvlist(zc, config)) != 0)
1575 nvlist_free(config);
1582 zfs_ioc_pool_export(zfs_cmd_t *zc)
1585 boolean_t force = (boolean_t)zc->zc_cookie;
1586 boolean_t hardforce = (boolean_t)zc->zc_guid;
1588 zfs_log_history(zc);
1589 error = spa_export(zc->zc_name, NULL, force, hardforce);
1595 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1600 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1601 return (SET_ERROR(EEXIST));
1603 error = put_nvlist(zc, configs);
1605 nvlist_free(configs);
1612 * zc_name name of the pool
1615 * zc_cookie real errno
1616 * zc_nvlist_dst config nvlist
1617 * zc_nvlist_dst_size size of config nvlist
1620 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1626 error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1627 sizeof (zc->zc_value));
1629 if (config != NULL) {
1630 ret = put_nvlist(zc, config);
1631 nvlist_free(config);
1634 * The config may be present even if 'error' is non-zero.
1635 * In this case we return success, and preserve the real errno
1638 zc->zc_cookie = error;
1647 * Try to import the given pool, returning pool stats as appropriate so that
1648 * user land knows which devices are available and overall pool health.
1651 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1653 nvlist_t *tryconfig, *config = NULL;
1656 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1657 zc->zc_iflags, &tryconfig)) != 0)
1660 config = spa_tryimport(tryconfig);
1662 nvlist_free(tryconfig);
1665 return (SET_ERROR(EINVAL));
1667 error = put_nvlist(zc, config);
1668 nvlist_free(config);
1675 * zc_name name of the pool
1676 * zc_cookie scan func (pool_scan_func_t)
1677 * zc_flags scrub pause/resume flag (pool_scrub_cmd_t)
1680 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1685 if (zc->zc_flags >= POOL_SCRUB_FLAGS_END)
1686 return (SET_ERROR(EINVAL));
1688 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1691 if (zc->zc_flags == POOL_SCRUB_PAUSE)
1692 error = spa_scrub_pause_resume(spa, POOL_SCRUB_PAUSE);
1693 else if (zc->zc_cookie == POOL_SCAN_NONE)
1694 error = spa_scan_stop(spa);
1696 error = spa_scan(spa, zc->zc_cookie);
1698 spa_close(spa, FTAG);
1704 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1709 error = spa_open(zc->zc_name, &spa, FTAG);
1712 spa_close(spa, FTAG);
1718 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1723 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1726 if (zc->zc_cookie < spa_version(spa) ||
1727 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1728 spa_close(spa, FTAG);
1729 return (SET_ERROR(EINVAL));
1732 spa_upgrade(spa, zc->zc_cookie);
1733 spa_close(spa, FTAG);
1739 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1746 if ((size = zc->zc_history_len) == 0)
1747 return (SET_ERROR(EINVAL));
1749 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1752 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1753 spa_close(spa, FTAG);
1754 return (SET_ERROR(ENOTSUP));
1757 hist_buf = vmem_alloc(size, KM_SLEEP);
1758 if ((error = spa_history_get(spa, &zc->zc_history_offset,
1759 &zc->zc_history_len, hist_buf)) == 0) {
1760 error = ddi_copyout(hist_buf,
1761 (void *)(uintptr_t)zc->zc_history,
1762 zc->zc_history_len, zc->zc_iflags);
1765 spa_close(spa, FTAG);
1766 vmem_free(hist_buf, size);
1771 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1776 error = spa_open(zc->zc_name, &spa, FTAG);
1778 error = spa_change_guid(spa);
1779 spa_close(spa, FTAG);
1785 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1787 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1792 * zc_name name of filesystem
1793 * zc_obj object to find
1796 * zc_value name of object
1799 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1804 /* XXX reading from objset not owned */
1805 if ((error = dmu_objset_hold_flags(zc->zc_name, B_TRUE,
1808 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1809 dmu_objset_rele_flags(os, B_TRUE, FTAG);
1810 return (SET_ERROR(EINVAL));
1812 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1813 sizeof (zc->zc_value));
1814 dmu_objset_rele_flags(os, B_TRUE, FTAG);
1821 * zc_name name of filesystem
1822 * zc_obj object to find
1825 * zc_stat stats on object
1826 * zc_value path to object
1829 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1834 /* XXX reading from objset not owned */
1835 if ((error = dmu_objset_hold_flags(zc->zc_name, B_TRUE,
1838 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1839 dmu_objset_rele_flags(os, B_TRUE, FTAG);
1840 return (SET_ERROR(EINVAL));
1842 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1843 sizeof (zc->zc_value));
1844 dmu_objset_rele_flags(os, B_TRUE, FTAG);
1850 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1856 error = spa_open(zc->zc_name, &spa, FTAG);
1860 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1861 zc->zc_iflags, &config);
1863 error = spa_vdev_add(spa, config);
1864 nvlist_free(config);
1866 spa_close(spa, FTAG);
1872 * zc_name name of the pool
1873 * zc_guid guid of vdev to remove
1874 * zc_cookie cancel removal
1877 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1882 error = spa_open(zc->zc_name, &spa, FTAG);
1885 if (zc->zc_cookie != 0) {
1886 error = spa_vdev_remove_cancel(spa);
1888 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1890 spa_close(spa, FTAG);
1895 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1899 vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1901 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1903 switch (zc->zc_cookie) {
1904 case VDEV_STATE_ONLINE:
1905 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1908 case VDEV_STATE_OFFLINE:
1909 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1912 case VDEV_STATE_FAULTED:
1913 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1914 zc->zc_obj != VDEV_AUX_EXTERNAL &&
1915 zc->zc_obj != VDEV_AUX_EXTERNAL_PERSIST)
1916 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1918 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1921 case VDEV_STATE_DEGRADED:
1922 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1923 zc->zc_obj != VDEV_AUX_EXTERNAL)
1924 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1926 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1930 error = SET_ERROR(EINVAL);
1932 zc->zc_cookie = newstate;
1933 spa_close(spa, FTAG);
1938 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1941 int replacing = zc->zc_cookie;
1945 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1948 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1949 zc->zc_iflags, &config)) == 0) {
1950 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1951 nvlist_free(config);
1954 spa_close(spa, FTAG);
1959 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1964 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1967 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
1969 spa_close(spa, FTAG);
1974 zfs_ioc_vdev_split(zfs_cmd_t *zc)
1977 nvlist_t *config, *props = NULL;
1979 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
1981 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1984 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1985 zc->zc_iflags, &config))) {
1986 spa_close(spa, FTAG);
1990 if (zc->zc_nvlist_src_size != 0 && (error =
1991 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1992 zc->zc_iflags, &props))) {
1993 spa_close(spa, FTAG);
1994 nvlist_free(config);
1998 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
2000 spa_close(spa, FTAG);
2002 nvlist_free(config);
2009 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
2012 char *path = zc->zc_value;
2013 uint64_t guid = zc->zc_guid;
2016 error = spa_open(zc->zc_name, &spa, FTAG);
2020 error = spa_vdev_setpath(spa, guid, path);
2021 spa_close(spa, FTAG);
2026 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
2029 char *fru = zc->zc_value;
2030 uint64_t guid = zc->zc_guid;
2033 error = spa_open(zc->zc_name, &spa, FTAG);
2037 error = spa_vdev_setfru(spa, guid, fru);
2038 spa_close(spa, FTAG);
2043 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
2048 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2050 if (zc->zc_nvlist_dst != 0 &&
2051 (error = dsl_prop_get_all(os, &nv)) == 0) {
2052 dmu_objset_stats(os, nv);
2054 * NB: zvol_get_stats() will read the objset contents,
2055 * which we aren't supposed to do with a
2056 * DS_MODE_USER hold, because it could be
2057 * inconsistent. So this is a bit of a workaround...
2058 * XXX reading without owning
2060 if (!zc->zc_objset_stats.dds_inconsistent &&
2061 dmu_objset_type(os) == DMU_OST_ZVOL) {
2062 error = zvol_get_stats(os, nv);
2070 error = put_nvlist(zc, nv);
2079 * zc_name name of filesystem
2080 * zc_nvlist_dst_size size of buffer for property nvlist
2083 * zc_objset_stats stats
2084 * zc_nvlist_dst property nvlist
2085 * zc_nvlist_dst_size size of property nvlist
2088 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2093 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2095 error = zfs_ioc_objset_stats_impl(zc, os);
2096 dmu_objset_rele(os, FTAG);
2104 * zc_name name of filesystem
2105 * zc_nvlist_dst_size size of buffer for property nvlist
2108 * zc_nvlist_dst received property nvlist
2109 * zc_nvlist_dst_size size of received property nvlist
2111 * Gets received properties (distinct from local properties on or after
2112 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2113 * local property values.
2116 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2122 * Without this check, we would return local property values if the
2123 * caller has not already received properties on or after
2124 * SPA_VERSION_RECVD_PROPS.
2126 if (!dsl_prop_get_hasrecvd(zc->zc_name))
2127 return (SET_ERROR(ENOTSUP));
2129 if (zc->zc_nvlist_dst != 0 &&
2130 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2131 error = put_nvlist(zc, nv);
2139 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2145 * zfs_get_zplprop() will either find a value or give us
2146 * the default value (if there is one).
2148 if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2150 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2156 * zc_name name of filesystem
2157 * zc_nvlist_dst_size size of buffer for zpl property nvlist
2160 * zc_nvlist_dst zpl property nvlist
2161 * zc_nvlist_dst_size size of zpl property nvlist
2164 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2169 /* XXX reading without owning */
2170 if ((err = dmu_objset_hold(zc->zc_name, FTAG, &os)))
2173 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2176 * NB: nvl_add_zplprop() will read the objset contents,
2177 * which we aren't supposed to do with a DS_MODE_USER
2178 * hold, because it could be inconsistent.
2180 if (zc->zc_nvlist_dst != 0 &&
2181 !zc->zc_objset_stats.dds_inconsistent &&
2182 dmu_objset_type(os) == DMU_OST_ZFS) {
2185 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2186 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2187 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2188 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2189 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2190 err = put_nvlist(zc, nv);
2193 err = SET_ERROR(ENOENT);
2195 dmu_objset_rele(os, FTAG);
2201 * zc_name name of filesystem
2202 * zc_cookie zap cursor
2203 * zc_nvlist_dst_size size of buffer for property nvlist
2206 * zc_name name of next filesystem
2207 * zc_cookie zap cursor
2208 * zc_objset_stats stats
2209 * zc_nvlist_dst property nvlist
2210 * zc_nvlist_dst_size size of property nvlist
2213 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2218 size_t orig_len = strlen(zc->zc_name);
2221 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os))) {
2222 if (error == ENOENT)
2223 error = SET_ERROR(ESRCH);
2227 p = strrchr(zc->zc_name, '/');
2228 if (p == NULL || p[1] != '\0')
2229 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2230 p = zc->zc_name + strlen(zc->zc_name);
2233 error = dmu_dir_list_next(os,
2234 sizeof (zc->zc_name) - (p - zc->zc_name), p,
2235 NULL, &zc->zc_cookie);
2236 if (error == ENOENT)
2237 error = SET_ERROR(ESRCH);
2238 } while (error == 0 && zfs_dataset_name_hidden(zc->zc_name));
2239 dmu_objset_rele(os, FTAG);
2242 * If it's an internal dataset (ie. with a '$' in its name),
2243 * don't try to get stats for it, otherwise we'll return ENOENT.
2245 if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2246 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2247 if (error == ENOENT) {
2248 /* We lost a race with destroy, get the next one. */
2249 zc->zc_name[orig_len] = '\0';
2258 * zc_name name of filesystem
2259 * zc_cookie zap cursor
2260 * zc_nvlist_src iteration range nvlist
2261 * zc_nvlist_src_size size of iteration range nvlist
2264 * zc_name name of next snapshot
2265 * zc_objset_stats stats
2266 * zc_nvlist_dst property nvlist
2267 * zc_nvlist_dst_size size of property nvlist
2270 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2273 objset_t *os, *ossnap;
2275 uint64_t min_txg = 0, max_txg = 0;
2277 if (zc->zc_nvlist_src_size != 0) {
2278 nvlist_t *props = NULL;
2279 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2280 zc->zc_iflags, &props);
2283 (void) nvlist_lookup_uint64(props, SNAP_ITER_MIN_TXG,
2285 (void) nvlist_lookup_uint64(props, SNAP_ITER_MAX_TXG,
2290 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2292 return (error == ENOENT ? SET_ERROR(ESRCH) : error);
2296 * A dataset name of maximum length cannot have any snapshots,
2297 * so exit immediately.
2299 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >=
2300 ZFS_MAX_DATASET_NAME_LEN) {
2301 dmu_objset_rele(os, FTAG);
2302 return (SET_ERROR(ESRCH));
2305 while (error == 0) {
2306 if (issig(JUSTLOOKING) && issig(FORREAL)) {
2307 error = SET_ERROR(EINTR);
2311 error = dmu_snapshot_list_next(os,
2312 sizeof (zc->zc_name) - strlen(zc->zc_name),
2313 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj,
2314 &zc->zc_cookie, NULL);
2315 if (error == ENOENT) {
2316 error = SET_ERROR(ESRCH);
2318 } else if (error != 0) {
2322 error = dsl_dataset_hold_obj(dmu_objset_pool(os), zc->zc_obj,
2327 if ((min_txg != 0 && dsl_get_creationtxg(ds) < min_txg) ||
2328 (max_txg != 0 && dsl_get_creationtxg(ds) > max_txg)) {
2329 dsl_dataset_rele(ds, FTAG);
2330 /* undo snapshot name append */
2331 *(strchr(zc->zc_name, '@') + 1) = '\0';
2336 if (zc->zc_simple) {
2337 dsl_dataset_rele(ds, FTAG);
2341 if ((error = dmu_objset_from_ds(ds, &ossnap)) != 0) {
2342 dsl_dataset_rele(ds, FTAG);
2345 if ((error = zfs_ioc_objset_stats_impl(zc, ossnap)) != 0) {
2346 dsl_dataset_rele(ds, FTAG);
2349 dsl_dataset_rele(ds, FTAG);
2353 dmu_objset_rele(os, FTAG);
2354 /* if we failed, undo the @ that we tacked on to zc_name */
2356 *strchr(zc->zc_name, '@') = '\0';
2361 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2363 const char *propname = nvpair_name(pair);
2365 unsigned int vallen;
2368 zfs_userquota_prop_t type;
2374 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2376 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2377 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2379 return (SET_ERROR(EINVAL));
2383 * A correctly constructed propname is encoded as
2384 * userquota@<rid>-<domain>.
2386 if ((dash = strchr(propname, '-')) == NULL ||
2387 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2389 return (SET_ERROR(EINVAL));
2396 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2398 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2399 zfsvfs_rele(zfsvfs, FTAG);
2406 * If the named property is one that has a special function to set its value,
2407 * return 0 on success and a positive error code on failure; otherwise if it is
2408 * not one of the special properties handled by this function, return -1.
2410 * XXX: It would be better for callers of the property interface if we handled
2411 * these special cases in dsl_prop.c (in the dsl layer).
2414 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2417 const char *propname = nvpair_name(pair);
2418 zfs_prop_t prop = zfs_name_to_prop(propname);
2419 uint64_t intval = 0;
2420 char *strval = NULL;
2423 if (prop == ZPROP_INVAL) {
2424 if (zfs_prop_userquota(propname))
2425 return (zfs_prop_set_userquota(dsname, pair));
2429 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2431 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2432 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2436 /* all special properties are numeric except for keylocation */
2437 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING) {
2438 strval = fnvpair_value_string(pair);
2440 intval = fnvpair_value_uint64(pair);
2444 case ZFS_PROP_QUOTA:
2445 err = dsl_dir_set_quota(dsname, source, intval);
2447 case ZFS_PROP_REFQUOTA:
2448 err = dsl_dataset_set_refquota(dsname, source, intval);
2450 case ZFS_PROP_FILESYSTEM_LIMIT:
2451 case ZFS_PROP_SNAPSHOT_LIMIT:
2452 if (intval == UINT64_MAX) {
2453 /* clearing the limit, just do it */
2456 err = dsl_dir_activate_fs_ss_limit(dsname);
2459 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2460 * default path to set the value in the nvlist.
2465 case ZFS_PROP_KEYLOCATION:
2466 err = dsl_crypto_can_set_keylocation(dsname, strval);
2469 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2470 * default path to set the value in the nvlist.
2475 case ZFS_PROP_RESERVATION:
2476 err = dsl_dir_set_reservation(dsname, source, intval);
2478 case ZFS_PROP_REFRESERVATION:
2479 err = dsl_dataset_set_refreservation(dsname, source, intval);
2481 case ZFS_PROP_VOLSIZE:
2482 err = zvol_set_volsize(dsname, intval);
2484 case ZFS_PROP_SNAPDEV:
2485 err = zvol_set_snapdev(dsname, source, intval);
2487 case ZFS_PROP_VOLMODE:
2488 err = zvol_set_volmode(dsname, source, intval);
2490 case ZFS_PROP_VERSION:
2494 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2497 err = zfs_set_version(zfsvfs, intval);
2498 zfsvfs_rele(zfsvfs, FTAG);
2500 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2503 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2504 (void) strlcpy(zc->zc_name, dsname,
2505 sizeof (zc->zc_name));
2506 (void) zfs_ioc_userspace_upgrade(zc);
2507 (void) zfs_ioc_id_quota_upgrade(zc);
2508 kmem_free(zc, sizeof (zfs_cmd_t));
2520 * This function is best effort. If it fails to set any of the given properties,
2521 * it continues to set as many as it can and returns the last error
2522 * encountered. If the caller provides a non-NULL errlist, it will be filled in
2523 * with the list of names of all the properties that failed along with the
2524 * corresponding error numbers.
2526 * If every property is set successfully, zero is returned and errlist is not
2530 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2539 nvlist_t *genericnvl = fnvlist_alloc();
2540 nvlist_t *retrynvl = fnvlist_alloc();
2543 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2544 const char *propname = nvpair_name(pair);
2545 zfs_prop_t prop = zfs_name_to_prop(propname);
2548 /* decode the property value */
2550 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2552 attrs = fnvpair_value_nvlist(pair);
2553 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2555 err = SET_ERROR(EINVAL);
2558 /* Validate value type */
2559 if (err == 0 && source == ZPROP_SRC_INHERITED) {
2560 /* inherited properties are expected to be booleans */
2561 if (nvpair_type(propval) != DATA_TYPE_BOOLEAN)
2562 err = SET_ERROR(EINVAL);
2563 } else if (err == 0 && prop == ZPROP_INVAL) {
2564 if (zfs_prop_user(propname)) {
2565 if (nvpair_type(propval) != DATA_TYPE_STRING)
2566 err = SET_ERROR(EINVAL);
2567 } else if (zfs_prop_userquota(propname)) {
2568 if (nvpair_type(propval) !=
2569 DATA_TYPE_UINT64_ARRAY)
2570 err = SET_ERROR(EINVAL);
2572 err = SET_ERROR(EINVAL);
2574 } else if (err == 0) {
2575 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2576 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2577 err = SET_ERROR(EINVAL);
2578 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2581 intval = fnvpair_value_uint64(propval);
2583 switch (zfs_prop_get_type(prop)) {
2584 case PROP_TYPE_NUMBER:
2586 case PROP_TYPE_STRING:
2587 err = SET_ERROR(EINVAL);
2589 case PROP_TYPE_INDEX:
2590 if (zfs_prop_index_to_string(prop,
2591 intval, &unused) != 0)
2592 err = SET_ERROR(EINVAL);
2596 "unknown property type");
2599 err = SET_ERROR(EINVAL);
2603 /* Validate permissions */
2605 err = zfs_check_settable(dsname, pair, CRED());
2608 if (source == ZPROP_SRC_INHERITED)
2609 err = -1; /* does not need special handling */
2611 err = zfs_prop_set_special(dsname, source,
2615 * For better performance we build up a list of
2616 * properties to set in a single transaction.
2618 err = nvlist_add_nvpair(genericnvl, pair);
2619 } else if (err != 0 && nvl != retrynvl) {
2621 * This may be a spurious error caused by
2622 * receiving quota and reservation out of order.
2623 * Try again in a second pass.
2625 err = nvlist_add_nvpair(retrynvl, pair);
2630 if (errlist != NULL)
2631 fnvlist_add_int32(errlist, propname, err);
2636 if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2641 if (!nvlist_empty(genericnvl) &&
2642 dsl_props_set(dsname, source, genericnvl) != 0) {
2644 * If this fails, we still want to set as many properties as we
2645 * can, so try setting them individually.
2648 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2649 const char *propname = nvpair_name(pair);
2653 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2655 attrs = fnvpair_value_nvlist(pair);
2656 propval = fnvlist_lookup_nvpair(attrs,
2660 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2661 strval = fnvpair_value_string(propval);
2662 err = dsl_prop_set_string(dsname, propname,
2664 } else if (nvpair_type(propval) == DATA_TYPE_BOOLEAN) {
2665 err = dsl_prop_inherit(dsname, propname,
2668 intval = fnvpair_value_uint64(propval);
2669 err = dsl_prop_set_int(dsname, propname, source,
2674 if (errlist != NULL) {
2675 fnvlist_add_int32(errlist, propname,
2682 nvlist_free(genericnvl);
2683 nvlist_free(retrynvl);
2689 * Check that all the properties are valid user properties.
2692 zfs_check_userprops(nvlist_t *nvl)
2694 nvpair_t *pair = NULL;
2696 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2697 const char *propname = nvpair_name(pair);
2699 if (!zfs_prop_user(propname) ||
2700 nvpair_type(pair) != DATA_TYPE_STRING)
2701 return (SET_ERROR(EINVAL));
2703 if (strlen(propname) >= ZAP_MAXNAMELEN)
2704 return (SET_ERROR(ENAMETOOLONG));
2706 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN)
2707 return (SET_ERROR(E2BIG));
2713 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2717 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2720 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2721 if (nvlist_exists(skipped, nvpair_name(pair)))
2724 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2729 clear_received_props(const char *dsname, nvlist_t *props,
2733 nvlist_t *cleared_props = NULL;
2734 props_skip(props, skipped, &cleared_props);
2735 if (!nvlist_empty(cleared_props)) {
2737 * Acts on local properties until the dataset has received
2738 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2740 zprop_source_t flags = (ZPROP_SRC_NONE |
2741 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2742 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2744 nvlist_free(cleared_props);
2750 * zc_name name of filesystem
2751 * zc_value name of property to set
2752 * zc_nvlist_src{_size} nvlist of properties to apply
2753 * zc_cookie received properties flag
2756 * zc_nvlist_dst{_size} error for each unapplied received property
2759 zfs_ioc_set_prop(zfs_cmd_t *zc)
2762 boolean_t received = zc->zc_cookie;
2763 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2768 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2769 zc->zc_iflags, &nvl)) != 0)
2773 nvlist_t *origprops;
2775 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2776 (void) clear_received_props(zc->zc_name,
2778 nvlist_free(origprops);
2781 error = dsl_prop_set_hasrecvd(zc->zc_name);
2784 errors = fnvlist_alloc();
2786 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2788 if (zc->zc_nvlist_dst != 0 && errors != NULL) {
2789 (void) put_nvlist(zc, errors);
2792 nvlist_free(errors);
2799 * zc_name name of filesystem
2800 * zc_value name of property to inherit
2801 * zc_cookie revert to received value if TRUE
2806 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2808 const char *propname = zc->zc_value;
2809 zfs_prop_t prop = zfs_name_to_prop(propname);
2810 boolean_t received = zc->zc_cookie;
2811 zprop_source_t source = (received
2812 ? ZPROP_SRC_NONE /* revert to received value, if any */
2813 : ZPROP_SRC_INHERITED); /* explicitly inherit */
2821 * Only check this in the non-received case. We want to allow
2822 * 'inherit -S' to revert non-inheritable properties like quota
2823 * and reservation to the received or default values even though
2824 * they are not considered inheritable.
2826 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2827 return (SET_ERROR(EINVAL));
2830 if (prop == ZPROP_INVAL) {
2831 if (!zfs_prop_user(propname))
2832 return (SET_ERROR(EINVAL));
2834 type = PROP_TYPE_STRING;
2835 } else if (prop == ZFS_PROP_VOLSIZE || prop == ZFS_PROP_VERSION) {
2836 return (SET_ERROR(EINVAL));
2838 type = zfs_prop_get_type(prop);
2842 * zfs_prop_set_special() expects properties in the form of an
2843 * nvpair with type info.
2845 dummy = fnvlist_alloc();
2848 case PROP_TYPE_STRING:
2849 VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2851 case PROP_TYPE_NUMBER:
2852 case PROP_TYPE_INDEX:
2853 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2856 err = SET_ERROR(EINVAL);
2860 pair = nvlist_next_nvpair(dummy, NULL);
2862 err = SET_ERROR(EINVAL);
2864 err = zfs_prop_set_special(zc->zc_name, source, pair);
2865 if (err == -1) /* property is not "special", needs handling */
2866 err = dsl_prop_inherit(zc->zc_name, zc->zc_value,
2876 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2883 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2884 zc->zc_iflags, &props)))
2888 * If the only property is the configfile, then just do a spa_lookup()
2889 * to handle the faulted case.
2891 pair = nvlist_next_nvpair(props, NULL);
2892 if (pair != NULL && strcmp(nvpair_name(pair),
2893 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2894 nvlist_next_nvpair(props, pair) == NULL) {
2895 mutex_enter(&spa_namespace_lock);
2896 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2897 spa_configfile_set(spa, props, B_FALSE);
2898 spa_write_cachefile(spa, B_FALSE, B_TRUE);
2900 mutex_exit(&spa_namespace_lock);
2907 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2912 error = spa_prop_set(spa, props);
2915 spa_close(spa, FTAG);
2921 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2925 nvlist_t *nvp = NULL;
2927 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2929 * If the pool is faulted, there may be properties we can still
2930 * get (such as altroot and cachefile), so attempt to get them
2933 mutex_enter(&spa_namespace_lock);
2934 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2935 error = spa_prop_get(spa, &nvp);
2936 mutex_exit(&spa_namespace_lock);
2938 error = spa_prop_get(spa, &nvp);
2939 spa_close(spa, FTAG);
2942 if (error == 0 && zc->zc_nvlist_dst != 0)
2943 error = put_nvlist(zc, nvp);
2945 error = SET_ERROR(EFAULT);
2953 * zc_name name of filesystem
2954 * zc_nvlist_src{_size} nvlist of delegated permissions
2955 * zc_perm_action allow/unallow flag
2960 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2963 nvlist_t *fsaclnv = NULL;
2965 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2966 zc->zc_iflags, &fsaclnv)) != 0)
2970 * Verify nvlist is constructed correctly
2972 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2973 nvlist_free(fsaclnv);
2974 return (SET_ERROR(EINVAL));
2978 * If we don't have PRIV_SYS_MOUNT, then validate
2979 * that user is allowed to hand out each permission in
2983 error = secpolicy_zfs(CRED());
2985 if (zc->zc_perm_action == B_FALSE) {
2986 error = dsl_deleg_can_allow(zc->zc_name,
2989 error = dsl_deleg_can_unallow(zc->zc_name,
2995 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2997 nvlist_free(fsaclnv);
3003 * zc_name name of filesystem
3006 * zc_nvlist_src{_size} nvlist of delegated permissions
3009 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
3014 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
3015 error = put_nvlist(zc, nvp);
3024 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
3026 zfs_creat_t *zct = arg;
3028 zfs_create_fs(os, cr, zct->zct_zplprops, tx);
3031 #define ZFS_PROP_UNDEFINED ((uint64_t)-1)
3035 * os parent objset pointer (NULL if root fs)
3036 * fuids_ok fuids allowed in this version of the spa?
3037 * sa_ok SAs allowed in this version of the spa?
3038 * createprops list of properties requested by creator
3041 * zplprops values for the zplprops we attach to the master node object
3042 * is_ci true if requested file system will be purely case-insensitive
3044 * Determine the settings for utf8only, normalization and
3045 * casesensitivity. Specific values may have been requested by the
3046 * creator and/or we can inherit values from the parent dataset. If
3047 * the file system is of too early a vintage, a creator can not
3048 * request settings for these properties, even if the requested
3049 * setting is the default value. We don't actually want to create dsl
3050 * properties for these, so remove them from the source nvlist after
3054 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
3055 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
3056 nvlist_t *zplprops, boolean_t *is_ci)
3058 uint64_t sense = ZFS_PROP_UNDEFINED;
3059 uint64_t norm = ZFS_PROP_UNDEFINED;
3060 uint64_t u8 = ZFS_PROP_UNDEFINED;
3063 ASSERT(zplprops != NULL);
3065 /* parent dataset must be a filesystem */
3066 if (os != NULL && os->os_phys->os_type != DMU_OST_ZFS)
3067 return (SET_ERROR(ZFS_ERR_WRONG_PARENT));
3070 * Pull out creator prop choices, if any.
3073 (void) nvlist_lookup_uint64(createprops,
3074 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
3075 (void) nvlist_lookup_uint64(createprops,
3076 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
3077 (void) nvlist_remove_all(createprops,
3078 zfs_prop_to_name(ZFS_PROP_NORMALIZE));
3079 (void) nvlist_lookup_uint64(createprops,
3080 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
3081 (void) nvlist_remove_all(createprops,
3082 zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
3083 (void) nvlist_lookup_uint64(createprops,
3084 zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3085 (void) nvlist_remove_all(createprops,
3086 zfs_prop_to_name(ZFS_PROP_CASE));
3090 * If the zpl version requested is whacky or the file system
3091 * or pool is version is too "young" to support normalization
3092 * and the creator tried to set a value for one of the props,
3095 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3096 (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3097 (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3098 (zplver < ZPL_VERSION_NORMALIZATION &&
3099 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3100 sense != ZFS_PROP_UNDEFINED)))
3101 return (SET_ERROR(ENOTSUP));
3104 * Put the version in the zplprops
3106 VERIFY(nvlist_add_uint64(zplprops,
3107 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3109 if (norm == ZFS_PROP_UNDEFINED &&
3110 (error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm)) != 0)
3112 VERIFY(nvlist_add_uint64(zplprops,
3113 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3116 * If we're normalizing, names must always be valid UTF-8 strings.
3120 if (u8 == ZFS_PROP_UNDEFINED &&
3121 (error = 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 (error = zfs_get_zplprop(os, ZFS_PROP_CASE, &sense)) != 0)
3129 VERIFY(nvlist_add_uint64(zplprops,
3130 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3133 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
3139 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3140 nvlist_t *zplprops, boolean_t *is_ci)
3142 boolean_t fuids_ok, sa_ok;
3143 uint64_t zplver = ZPL_VERSION;
3144 objset_t *os = NULL;
3145 char parentname[ZFS_MAX_DATASET_NAME_LEN];
3150 zfs_get_parent(dataset, parentname, sizeof (parentname));
3152 if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3155 spa_vers = spa_version(spa);
3156 spa_close(spa, FTAG);
3158 zplver = zfs_zpl_version_map(spa_vers);
3159 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3160 sa_ok = (zplver >= ZPL_VERSION_SA);
3163 * Open parent object set so we can inherit zplprop values.
3165 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3168 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3170 dmu_objset_rele(os, FTAG);
3175 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3176 nvlist_t *zplprops, boolean_t *is_ci)
3180 uint64_t zplver = ZPL_VERSION;
3183 zplver = zfs_zpl_version_map(spa_vers);
3184 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3185 sa_ok = (zplver >= ZPL_VERSION_SA);
3187 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3188 createprops, zplprops, is_ci);
3194 * "type" -> dmu_objset_type_t (int32)
3195 * (optional) "props" -> { prop -> value }
3196 * (optional) "hidden_args" -> { "wkeydata" -> value }
3197 * raw uint8_t array of encryption wrapping key data (32 bytes)
3200 * outnvl: propname -> error code (int32)
3203 static const zfs_ioc_key_t zfs_keys_create[] = {
3204 {"type", DATA_TYPE_INT32, 0},
3205 {"props", DATA_TYPE_NVLIST, ZK_OPTIONAL},
3206 {"hidden_args", DATA_TYPE_NVLIST, ZK_OPTIONAL},
3210 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3213 zfs_creat_t zct = { 0 };
3214 nvlist_t *nvprops = NULL;
3215 nvlist_t *hidden_args = NULL;
3216 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3217 dmu_objset_type_t type;
3218 boolean_t is_insensitive = B_FALSE;
3219 dsl_crypto_params_t *dcp = NULL;
3221 type = (dmu_objset_type_t)fnvlist_lookup_int32(innvl, "type");
3222 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3223 (void) nvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS, &hidden_args);
3227 cbfunc = zfs_create_cb;
3231 cbfunc = zvol_create_cb;
3238 if (strchr(fsname, '@') ||
3239 strchr(fsname, '%'))
3240 return (SET_ERROR(EINVAL));
3242 zct.zct_props = nvprops;
3245 return (SET_ERROR(EINVAL));
3247 if (type == DMU_OST_ZVOL) {
3248 uint64_t volsize, volblocksize;
3250 if (nvprops == NULL)
3251 return (SET_ERROR(EINVAL));
3252 if (nvlist_lookup_uint64(nvprops,
3253 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3254 return (SET_ERROR(EINVAL));
3256 if ((error = nvlist_lookup_uint64(nvprops,
3257 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3258 &volblocksize)) != 0 && error != ENOENT)
3259 return (SET_ERROR(EINVAL));
3262 volblocksize = zfs_prop_default_numeric(
3263 ZFS_PROP_VOLBLOCKSIZE);
3265 if ((error = zvol_check_volblocksize(fsname,
3266 volblocksize)) != 0 ||
3267 (error = zvol_check_volsize(volsize,
3268 volblocksize)) != 0)
3270 } else if (type == DMU_OST_ZFS) {
3274 * We have to have normalization and
3275 * case-folding flags correct when we do the
3276 * file system creation, so go figure them out
3279 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3280 NV_UNIQUE_NAME, KM_SLEEP) == 0);
3281 error = zfs_fill_zplprops(fsname, nvprops,
3282 zct.zct_zplprops, &is_insensitive);
3284 nvlist_free(zct.zct_zplprops);
3289 error = dsl_crypto_params_create_nvlist(DCP_CMD_NONE, nvprops,
3292 nvlist_free(zct.zct_zplprops);
3296 error = dmu_objset_create(fsname, type,
3297 is_insensitive ? DS_FLAG_CI_DATASET : 0, dcp, cbfunc, &zct);
3299 nvlist_free(zct.zct_zplprops);
3300 dsl_crypto_params_free(dcp, !!error);
3303 * It would be nice to do this atomically.
3306 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3313 * Volumes will return EBUSY and cannot be destroyed
3314 * until all asynchronous minor handling (e.g. from
3315 * setting the volmode property) has completed. Wait for
3316 * the spa_zvol_taskq to drain then retry.
3318 error2 = dsl_destroy_head(fsname);
3319 while ((error2 == EBUSY) && (type == DMU_OST_ZVOL)) {
3320 error2 = spa_open(fsname, &spa, FTAG);
3322 taskq_wait(spa->spa_zvol_taskq);
3323 spa_close(spa, FTAG);
3325 error2 = dsl_destroy_head(fsname);
3334 * "origin" -> name of origin snapshot
3335 * (optional) "props" -> { prop -> value }
3336 * (optional) "hidden_args" -> { "wkeydata" -> value }
3337 * raw uint8_t array of encryption wrapping key data (32 bytes)
3341 * outnvl: propname -> error code (int32)
3343 static const zfs_ioc_key_t zfs_keys_clone[] = {
3344 {"origin", DATA_TYPE_STRING, 0},
3345 {"props", DATA_TYPE_NVLIST, ZK_OPTIONAL},
3346 {"hidden_args", DATA_TYPE_NVLIST, ZK_OPTIONAL},
3350 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3353 nvlist_t *nvprops = NULL;
3356 origin_name = fnvlist_lookup_string(innvl, "origin");
3357 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3359 if (strchr(fsname, '@') ||
3360 strchr(fsname, '%'))
3361 return (SET_ERROR(EINVAL));
3363 if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3364 return (SET_ERROR(EINVAL));
3366 error = dmu_objset_clone(fsname, origin_name);
3369 * It would be nice to do this atomically.
3372 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3375 (void) dsl_destroy_head(fsname);
3380 static const zfs_ioc_key_t zfs_keys_remap[] = {
3386 zfs_ioc_remap(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3388 /* This IOCTL is no longer supported. */
3394 * "snaps" -> { snapshot1, snapshot2 }
3395 * (optional) "props" -> { prop -> value (string) }
3398 * outnvl: snapshot -> error code (int32)
3400 static const zfs_ioc_key_t zfs_keys_snapshot[] = {
3401 {"snaps", DATA_TYPE_NVLIST, 0},
3402 {"props", DATA_TYPE_NVLIST, ZK_OPTIONAL},
3406 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3409 nvlist_t *props = NULL;
3413 (void) nvlist_lookup_nvlist(innvl, "props", &props);
3414 if (!nvlist_empty(props) &&
3415 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3416 return (SET_ERROR(ENOTSUP));
3417 if ((error = zfs_check_userprops(props)) != 0)
3420 snaps = fnvlist_lookup_nvlist(innvl, "snaps");
3421 poollen = strlen(poolname);
3422 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3423 pair = nvlist_next_nvpair(snaps, pair)) {
3424 const char *name = nvpair_name(pair);
3425 char *cp = strchr(name, '@');
3428 * The snap name must contain an @, and the part after it must
3429 * contain only valid characters.
3432 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3433 return (SET_ERROR(EINVAL));
3436 * The snap must be in the specified pool.
3438 if (strncmp(name, poolname, poollen) != 0 ||
3439 (name[poollen] != '/' && name[poollen] != '@'))
3440 return (SET_ERROR(EXDEV));
3443 * Check for permission to set the properties on the fs.
3445 if (!nvlist_empty(props)) {
3447 error = zfs_secpolicy_write_perms(name,
3448 ZFS_DELEG_PERM_USERPROP, CRED());
3454 /* This must be the only snap of this fs. */
3455 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3456 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3457 if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3459 return (SET_ERROR(EXDEV));
3464 error = dsl_dataset_snapshot(snaps, props, outnvl);
3470 * innvl: "message" -> string
3472 static const zfs_ioc_key_t zfs_keys_log_history[] = {
3473 {"message", DATA_TYPE_STRING, 0},
3478 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3486 * The poolname in the ioctl is not set, we get it from the TSD,
3487 * which was set at the end of the last successful ioctl that allows
3488 * logging. The secpolicy func already checked that it is set.
3489 * Only one log ioctl is allowed after each successful ioctl, so
3490 * we clear the TSD here.
3492 poolname = tsd_get(zfs_allow_log_key);
3493 if (poolname == NULL)
3494 return (SET_ERROR(EINVAL));
3495 (void) tsd_set(zfs_allow_log_key, NULL);
3496 error = spa_open(poolname, &spa, FTAG);
3497 kmem_strfree(poolname);
3501 message = fnvlist_lookup_string(innvl, "message");
3503 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3504 spa_close(spa, FTAG);
3505 return (SET_ERROR(ENOTSUP));
3508 error = spa_history_log(spa, message);
3509 spa_close(spa, FTAG);
3514 * This ioctl is used to set the bootenv configuration on the current
3515 * pool. This configuration is stored in the second padding area of the label,
3516 * and it is used by the GRUB bootloader used on Linux to store the contents
3517 * of the grubenv file. The file is stored as raw ASCII, and is protected by
3518 * an embedded checksum. By default, GRUB will check if the boot filesystem
3519 * supports storing the environment data in a special location, and if so,
3520 * will invoke filesystem specific logic to retrieve it. This can be overridden
3521 * by a variable, should the user so desire.
3524 static const zfs_ioc_key_t zfs_keys_set_bootenv[] = {
3525 {"envmap", DATA_TYPE_STRING, 0},
3529 zfs_ioc_set_bootenv(const char *name, nvlist_t *innvl, nvlist_t *outnvl)
3535 envmap = fnvlist_lookup_string(innvl, "envmap");
3536 if ((error = spa_open(name, &spa, FTAG)) != 0)
3538 spa_vdev_state_enter(spa, SCL_ALL);
3539 error = vdev_label_write_bootenv(spa->spa_root_vdev, envmap);
3540 (void) spa_vdev_state_exit(spa, NULL, 0);
3541 spa_close(spa, FTAG);
3545 static const zfs_ioc_key_t zfs_keys_get_bootenv[] = {
3551 zfs_ioc_get_bootenv(const char *name, nvlist_t *innvl, nvlist_t *outnvl)
3556 if ((error = spa_open(name, &spa, FTAG)) != 0)
3558 spa_vdev_state_enter(spa, SCL_ALL);
3559 error = vdev_label_read_bootenv(spa->spa_root_vdev, outnvl);
3560 (void) spa_vdev_state_exit(spa, NULL, 0);
3561 spa_close(spa, FTAG);
3566 * The dp_config_rwlock must not be held when calling this, because the
3567 * unmount may need to write out data.
3569 * This function is best-effort. Callers must deal gracefully if it
3570 * remains mounted (or is remounted after this call).
3572 * Returns 0 if the argument is not a snapshot, or it is not currently a
3573 * filesystem, or we were able to unmount it. Returns error code otherwise.
3576 zfs_unmount_snap(const char *snapname)
3578 if (strchr(snapname, '@') == NULL)
3581 (void) zfsctl_snapshot_unmount((char *)snapname, MNT_FORCE);
3586 zfs_unmount_snap_cb(const char *snapname, void *arg)
3588 zfs_unmount_snap(snapname);
3593 * When a clone is destroyed, its origin may also need to be destroyed,
3594 * in which case it must be unmounted. This routine will do that unmount
3598 zfs_destroy_unmount_origin(const char *fsname)
3604 error = dmu_objset_hold(fsname, FTAG, &os);
3607 ds = dmu_objset_ds(os);
3608 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3609 char originname[ZFS_MAX_DATASET_NAME_LEN];
3610 dsl_dataset_name(ds->ds_prev, originname);
3611 dmu_objset_rele(os, FTAG);
3612 zfs_unmount_snap(originname);
3614 dmu_objset_rele(os, FTAG);
3620 * "snaps" -> { snapshot1, snapshot2 }
3621 * (optional boolean) "defer"
3624 * outnvl: snapshot -> error code (int32)
3626 static const zfs_ioc_key_t zfs_keys_destroy_snaps[] = {
3627 {"snaps", DATA_TYPE_NVLIST, 0},
3628 {"defer", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
3633 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3641 snaps = fnvlist_lookup_nvlist(innvl, "snaps");
3642 defer = nvlist_exists(innvl, "defer");
3644 poollen = strlen(poolname);
3645 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3646 pair = nvlist_next_nvpair(snaps, pair)) {
3647 const char *name = nvpair_name(pair);
3650 * The snap must be in the specified pool to prevent the
3651 * invalid removal of zvol minors below.
3653 if (strncmp(name, poolname, poollen) != 0 ||
3654 (name[poollen] != '/' && name[poollen] != '@'))
3655 return (SET_ERROR(EXDEV));
3657 zfs_unmount_snap(nvpair_name(pair));
3658 if (spa_open(name, &spa, FTAG) == 0) {
3659 zvol_remove_minors(spa, name, B_TRUE);
3660 spa_close(spa, FTAG);
3664 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3668 * Create bookmarks. The bookmark names are of the form <fs>#<bmark>.
3669 * All bookmarks and snapshots must be in the same pool.
3670 * dsl_bookmark_create_nvl_validate describes the nvlist schema in more detail.
3673 * new_bookmark1 -> existing_snapshot,
3674 * new_bookmark2 -> existing_bookmark,
3677 * outnvl: bookmark -> error code (int32)
3680 static const zfs_ioc_key_t zfs_keys_bookmark[] = {
3681 {"<bookmark>...", DATA_TYPE_STRING, ZK_WILDCARDLIST},
3686 zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3688 return (dsl_bookmark_create(innvl, outnvl));
3693 * property 1, property 2, ...
3697 * bookmark name 1 -> { property 1, property 2, ... },
3698 * bookmark name 2 -> { property 1, property 2, ... }
3702 static const zfs_ioc_key_t zfs_keys_get_bookmarks[] = {
3703 {"<property>...", DATA_TYPE_BOOLEAN, ZK_WILDCARDLIST | ZK_OPTIONAL},
3707 zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3709 return (dsl_get_bookmarks(fsname, innvl, outnvl));
3713 * innvl is not used.
3716 * property 1, property 2, ...
3720 static const zfs_ioc_key_t zfs_keys_get_bookmark_props[] = {
3726 zfs_ioc_get_bookmark_props(const char *bookmark, nvlist_t *innvl,
3729 char fsname[ZFS_MAX_DATASET_NAME_LEN];
3732 bmname = strchr(bookmark, '#');
3734 return (SET_ERROR(EINVAL));
3737 (void) strlcpy(fsname, bookmark, sizeof (fsname));
3738 *(strchr(fsname, '#')) = '\0';
3740 return (dsl_get_bookmark_props(fsname, bmname, outnvl));
3745 * bookmark name 1, bookmark name 2
3748 * outnvl: bookmark -> error code (int32)
3751 static const zfs_ioc_key_t zfs_keys_destroy_bookmarks[] = {
3752 {"<bookmark>...", DATA_TYPE_BOOLEAN, ZK_WILDCARDLIST},
3756 zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl,
3761 poollen = strlen(poolname);
3762 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3763 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3764 const char *name = nvpair_name(pair);
3765 const char *cp = strchr(name, '#');
3768 * The bookmark name must contain an #, and the part after it
3769 * must contain only valid characters.
3772 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3773 return (SET_ERROR(EINVAL));
3776 * The bookmark must be in the specified pool.
3778 if (strncmp(name, poolname, poollen) != 0 ||
3779 (name[poollen] != '/' && name[poollen] != '#'))
3780 return (SET_ERROR(EXDEV));
3783 error = dsl_bookmark_destroy(innvl, outnvl);
3787 static const zfs_ioc_key_t zfs_keys_channel_program[] = {
3788 {"program", DATA_TYPE_STRING, 0},
3789 {"arg", DATA_TYPE_ANY, 0},
3790 {"sync", DATA_TYPE_BOOLEAN_VALUE, ZK_OPTIONAL},
3791 {"instrlimit", DATA_TYPE_UINT64, ZK_OPTIONAL},
3792 {"memlimit", DATA_TYPE_UINT64, ZK_OPTIONAL},
3796 zfs_ioc_channel_program(const char *poolname, nvlist_t *innvl,
3800 uint64_t instrlimit, memlimit;
3801 boolean_t sync_flag;
3802 nvpair_t *nvarg = NULL;
3804 program = fnvlist_lookup_string(innvl, ZCP_ARG_PROGRAM);
3805 if (0 != nvlist_lookup_boolean_value(innvl, ZCP_ARG_SYNC, &sync_flag)) {
3808 if (0 != nvlist_lookup_uint64(innvl, ZCP_ARG_INSTRLIMIT, &instrlimit)) {
3809 instrlimit = ZCP_DEFAULT_INSTRLIMIT;
3811 if (0 != nvlist_lookup_uint64(innvl, ZCP_ARG_MEMLIMIT, &memlimit)) {
3812 memlimit = ZCP_DEFAULT_MEMLIMIT;
3814 nvarg = fnvlist_lookup_nvpair(innvl, ZCP_ARG_ARGLIST);
3816 if (instrlimit == 0 || instrlimit > zfs_lua_max_instrlimit)
3817 return (SET_ERROR(EINVAL));
3818 if (memlimit == 0 || memlimit > zfs_lua_max_memlimit)
3819 return (SET_ERROR(EINVAL));
3821 return (zcp_eval(poolname, program, sync_flag, instrlimit, memlimit,
3829 static const zfs_ioc_key_t zfs_keys_pool_checkpoint[] = {
3835 zfs_ioc_pool_checkpoint(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3837 return (spa_checkpoint(poolname));
3844 static const zfs_ioc_key_t zfs_keys_pool_discard_checkpoint[] = {
3850 zfs_ioc_pool_discard_checkpoint(const char *poolname, nvlist_t *innvl,
3853 return (spa_checkpoint_discard(poolname));
3858 * zc_name name of dataset to destroy
3859 * zc_defer_destroy mark for deferred destroy
3864 zfs_ioc_destroy(zfs_cmd_t *zc)
3867 dmu_objset_type_t ost;
3870 err = dmu_objset_hold(zc->zc_name, FTAG, &os);
3873 ost = dmu_objset_type(os);
3874 dmu_objset_rele(os, FTAG);
3876 if (ost == DMU_OST_ZFS)
3877 zfs_unmount_snap(zc->zc_name);
3879 if (strchr(zc->zc_name, '@')) {
3880 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3882 err = dsl_destroy_head(zc->zc_name);
3883 if (err == EEXIST) {
3885 * It is possible that the given DS may have
3886 * hidden child (%recv) datasets - "leftovers"
3887 * resulting from the previously interrupted
3890 * 6 extra bytes for /%recv
3892 char namebuf[ZFS_MAX_DATASET_NAME_LEN + 6];
3894 if (snprintf(namebuf, sizeof (namebuf), "%s/%s",
3895 zc->zc_name, recv_clone_name) >=
3897 return (SET_ERROR(EINVAL));
3900 * Try to remove the hidden child (%recv) and after
3901 * that try to remove the target dataset.
3902 * If the hidden child (%recv) does not exist
3903 * the original error (EEXIST) will be returned
3905 err = dsl_destroy_head(namebuf);
3907 err = dsl_destroy_head(zc->zc_name);
3908 else if (err == ENOENT)
3909 err = SET_ERROR(EEXIST);
3918 * "initialize_command" -> POOL_INITIALIZE_{CANCEL|START|SUSPEND} (uint64)
3919 * "initialize_vdevs": { -> guids to initialize (nvlist)
3920 * "vdev_path_1": vdev_guid_1, (uint64),
3921 * "vdev_path_2": vdev_guid_2, (uint64),
3927 * "initialize_vdevs": { -> initialization errors (nvlist)
3928 * "vdev_path_1": errno, see function body for possible errnos (uint64)
3929 * "vdev_path_2": errno, ... (uint64)
3934 * EINVAL is returned for an unknown commands or if any of the provided vdev
3935 * guids have be specified with a type other than uint64.
3937 static const zfs_ioc_key_t zfs_keys_pool_initialize[] = {
3938 {ZPOOL_INITIALIZE_COMMAND, DATA_TYPE_UINT64, 0},
3939 {ZPOOL_INITIALIZE_VDEVS, DATA_TYPE_NVLIST, 0}
3943 zfs_ioc_pool_initialize(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3946 if (nvlist_lookup_uint64(innvl, ZPOOL_INITIALIZE_COMMAND,
3948 return (SET_ERROR(EINVAL));
3951 if (!(cmd_type == POOL_INITIALIZE_CANCEL ||
3952 cmd_type == POOL_INITIALIZE_START ||
3953 cmd_type == POOL_INITIALIZE_SUSPEND)) {
3954 return (SET_ERROR(EINVAL));
3957 nvlist_t *vdev_guids;
3958 if (nvlist_lookup_nvlist(innvl, ZPOOL_INITIALIZE_VDEVS,
3959 &vdev_guids) != 0) {
3960 return (SET_ERROR(EINVAL));
3963 for (nvpair_t *pair = nvlist_next_nvpair(vdev_guids, NULL);
3964 pair != NULL; pair = nvlist_next_nvpair(vdev_guids, pair)) {
3966 if (nvpair_value_uint64(pair, &vdev_guid) != 0) {
3967 return (SET_ERROR(EINVAL));
3972 int error = spa_open(poolname, &spa, FTAG);
3976 nvlist_t *vdev_errlist = fnvlist_alloc();
3977 int total_errors = spa_vdev_initialize(spa, vdev_guids, cmd_type,
3980 if (fnvlist_size(vdev_errlist) > 0) {
3981 fnvlist_add_nvlist(outnvl, ZPOOL_INITIALIZE_VDEVS,
3984 fnvlist_free(vdev_errlist);
3986 spa_close(spa, FTAG);
3987 return (total_errors > 0 ? EINVAL : 0);
3992 * "trim_command" -> POOL_TRIM_{CANCEL|START|SUSPEND} (uint64)
3993 * "trim_vdevs": { -> guids to TRIM (nvlist)
3994 * "vdev_path_1": vdev_guid_1, (uint64),
3995 * "vdev_path_2": vdev_guid_2, (uint64),
3998 * "trim_rate" -> Target TRIM rate in bytes/sec.
3999 * "trim_secure" -> Set to request a secure TRIM.
4003 * "trim_vdevs": { -> TRIM errors (nvlist)
4004 * "vdev_path_1": errno, see function body for possible errnos (uint64)
4005 * "vdev_path_2": errno, ... (uint64)
4010 * EINVAL is returned for an unknown commands or if any of the provided vdev
4011 * guids have be specified with a type other than uint64.
4013 static const zfs_ioc_key_t zfs_keys_pool_trim[] = {
4014 {ZPOOL_TRIM_COMMAND, DATA_TYPE_UINT64, 0},
4015 {ZPOOL_TRIM_VDEVS, DATA_TYPE_NVLIST, 0},
4016 {ZPOOL_TRIM_RATE, DATA_TYPE_UINT64, ZK_OPTIONAL},
4017 {ZPOOL_TRIM_SECURE, DATA_TYPE_BOOLEAN_VALUE, ZK_OPTIONAL},
4021 zfs_ioc_pool_trim(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
4024 if (nvlist_lookup_uint64(innvl, ZPOOL_TRIM_COMMAND, &cmd_type) != 0)
4025 return (SET_ERROR(EINVAL));
4027 if (!(cmd_type == POOL_TRIM_CANCEL ||
4028 cmd_type == POOL_TRIM_START ||
4029 cmd_type == POOL_TRIM_SUSPEND)) {
4030 return (SET_ERROR(EINVAL));
4033 nvlist_t *vdev_guids;
4034 if (nvlist_lookup_nvlist(innvl, ZPOOL_TRIM_VDEVS, &vdev_guids) != 0)
4035 return (SET_ERROR(EINVAL));
4037 for (nvpair_t *pair = nvlist_next_nvpair(vdev_guids, NULL);
4038 pair != NULL; pair = nvlist_next_nvpair(vdev_guids, pair)) {
4040 if (nvpair_value_uint64(pair, &vdev_guid) != 0) {
4041 return (SET_ERROR(EINVAL));
4045 /* Optional, defaults to maximum rate when not provided */
4047 if (nvlist_lookup_uint64(innvl, ZPOOL_TRIM_RATE, &rate) != 0)
4050 /* Optional, defaults to standard TRIM when not provided */
4052 if (nvlist_lookup_boolean_value(innvl, ZPOOL_TRIM_SECURE,
4058 int error = spa_open(poolname, &spa, FTAG);
4062 nvlist_t *vdev_errlist = fnvlist_alloc();
4063 int total_errors = spa_vdev_trim(spa, vdev_guids, cmd_type,
4064 rate, !!zfs_trim_metaslab_skip, secure, vdev_errlist);
4066 if (fnvlist_size(vdev_errlist) > 0)
4067 fnvlist_add_nvlist(outnvl, ZPOOL_TRIM_VDEVS, vdev_errlist);
4069 fnvlist_free(vdev_errlist);
4071 spa_close(spa, FTAG);
4072 return (total_errors > 0 ? EINVAL : 0);
4076 * This ioctl waits for activity of a particular type to complete. If there is
4077 * no activity of that type in progress, it returns immediately, and the
4078 * returned value "waited" is false. If there is activity in progress, and no
4079 * tag is passed in, the ioctl blocks until all activity of that type is
4080 * complete, and then returns with "waited" set to true.
4082 * If a tag is provided, it identifies a particular instance of an activity to
4083 * wait for. Currently, this is only valid for use with 'initialize', because
4084 * that is the only activity for which there can be multiple instances running
4085 * concurrently. In the case of 'initialize', the tag corresponds to the guid of
4086 * the vdev on which to wait.
4088 * If a thread waiting in the ioctl receives a signal, the call will return
4089 * immediately, and the return value will be EINTR.
4092 * "wait_activity" -> int32_t
4093 * (optional) "wait_tag" -> uint64_t
4096 * outnvl: "waited" -> boolean_t
4098 static const zfs_ioc_key_t zfs_keys_pool_wait[] = {
4099 {ZPOOL_WAIT_ACTIVITY, DATA_TYPE_INT32, 0},
4100 {ZPOOL_WAIT_TAG, DATA_TYPE_UINT64, ZK_OPTIONAL},
4104 zfs_ioc_wait(const char *name, nvlist_t *innvl, nvlist_t *outnvl)
4111 if (nvlist_lookup_int32(innvl, ZPOOL_WAIT_ACTIVITY, &activity) != 0)
4114 if (nvlist_lookup_uint64(innvl, ZPOOL_WAIT_TAG, &tag) == 0)
4115 error = spa_wait_tag(name, activity, tag, &waited);
4117 error = spa_wait(name, activity, &waited);
4120 fnvlist_add_boolean_value(outnvl, ZPOOL_WAIT_WAITED, waited);
4126 * This ioctl waits for activity of a particular type to complete. If there is
4127 * no activity of that type in progress, it returns immediately, and the
4128 * returned value "waited" is false. If there is activity in progress, and no
4129 * tag is passed in, the ioctl blocks until all activity of that type is
4130 * complete, and then returns with "waited" set to true.
4132 * If a thread waiting in the ioctl receives a signal, the call will return
4133 * immediately, and the return value will be EINTR.
4136 * "wait_activity" -> int32_t
4139 * outnvl: "waited" -> boolean_t
4141 static const zfs_ioc_key_t zfs_keys_fs_wait[] = {
4142 {ZFS_WAIT_ACTIVITY, DATA_TYPE_INT32, 0},
4146 zfs_ioc_wait_fs(const char *name, nvlist_t *innvl, nvlist_t *outnvl)
4149 boolean_t waited = B_FALSE;
4155 if (nvlist_lookup_int32(innvl, ZFS_WAIT_ACTIVITY, &activity) != 0)
4156 return (SET_ERROR(EINVAL));
4158 if (activity >= ZFS_WAIT_NUM_ACTIVITIES || activity < 0)
4159 return (SET_ERROR(EINVAL));
4161 if ((error = dsl_pool_hold(name, FTAG, &dp)) != 0)
4164 if ((error = dsl_dataset_hold(dp, name, FTAG, &ds)) != 0) {
4165 dsl_pool_rele(dp, FTAG);
4170 mutex_enter(&dd->dd_activity_lock);
4171 dd->dd_activity_waiters++;
4174 * We get a long-hold here so that the dsl_dataset_t and dsl_dir_t
4175 * aren't evicted while we're waiting. Normally this is prevented by
4176 * holding the pool, but we can't do that while we're waiting since
4177 * that would prevent TXGs from syncing out. Some of the functionality
4178 * of long-holds (e.g. preventing deletion) is unnecessary for this
4179 * case, since we would cancel the waiters before proceeding with a
4180 * deletion. An alternative mechanism for keeping the dataset around
4181 * could be developed but this is simpler.
4183 dsl_dataset_long_hold(ds, FTAG);
4184 dsl_pool_rele(dp, FTAG);
4186 error = dsl_dir_wait(dd, ds, activity, &waited);
4188 dsl_dataset_long_rele(ds, FTAG);
4189 dd->dd_activity_waiters--;
4190 if (dd->dd_activity_waiters == 0)
4191 cv_signal(&dd->dd_activity_cv);
4192 mutex_exit(&dd->dd_activity_lock);
4194 dsl_dataset_rele(ds, FTAG);
4197 fnvlist_add_boolean_value(outnvl, ZFS_WAIT_WAITED, waited);
4203 * fsname is name of dataset to rollback (to most recent snapshot)
4205 * innvl may contain name of expected target snapshot
4207 * outnvl: "target" -> name of most recent snapshot
4210 static const zfs_ioc_key_t zfs_keys_rollback[] = {
4211 {"target", DATA_TYPE_STRING, ZK_OPTIONAL},
4216 zfs_ioc_rollback(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
4219 zvol_state_handle_t *zv;
4220 char *target = NULL;
4223 (void) nvlist_lookup_string(innvl, "target", &target);
4224 if (target != NULL) {
4225 const char *cp = strchr(target, '@');
4228 * The snap name must contain an @, and the part after it must
4229 * contain only valid characters.
4232 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
4233 return (SET_ERROR(EINVAL));
4236 if (getzfsvfs(fsname, &zfsvfs) == 0) {
4239 ds = dmu_objset_ds(zfsvfs->z_os);
4240 error = zfs_suspend_fs(zfsvfs);
4244 error = dsl_dataset_rollback(fsname, target, zfsvfs,
4246 resume_err = zfs_resume_fs(zfsvfs, ds);
4247 error = error ? error : resume_err;
4249 zfs_vfs_rele(zfsvfs);
4250 } else if ((zv = zvol_suspend(fsname)) != NULL) {
4251 error = dsl_dataset_rollback(fsname, target, zvol_tag(zv),
4255 error = dsl_dataset_rollback(fsname, target, NULL, outnvl);
4261 recursive_unmount(const char *fsname, void *arg)
4263 const char *snapname = arg;
4266 fullname = kmem_asprintf("%s@%s", fsname, snapname);
4267 zfs_unmount_snap(fullname);
4268 kmem_strfree(fullname);
4275 * snapname is the snapshot to redact.
4277 * "bookname" -> (string)
4278 * shortname of the redaction bookmark to generate
4279 * "snapnv" -> (nvlist, values ignored)
4280 * snapshots to redact snapname with respect to
4287 static const zfs_ioc_key_t zfs_keys_redact[] = {
4288 {"bookname", DATA_TYPE_STRING, 0},
4289 {"snapnv", DATA_TYPE_NVLIST, 0},
4292 zfs_ioc_redact(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
4294 nvlist_t *redactnvl = NULL;
4295 char *redactbook = NULL;
4297 if (nvlist_lookup_nvlist(innvl, "snapnv", &redactnvl) != 0)
4298 return (SET_ERROR(EINVAL));
4299 if (fnvlist_num_pairs(redactnvl) == 0)
4300 return (SET_ERROR(ENXIO));
4301 if (nvlist_lookup_string(innvl, "bookname", &redactbook) != 0)
4302 return (SET_ERROR(EINVAL));
4304 return (dmu_redact_snap(snapname, redactnvl, redactbook));
4309 * zc_name old name of dataset
4310 * zc_value new name of dataset
4311 * zc_cookie recursive flag (only valid for snapshots)
4316 zfs_ioc_rename(zfs_cmd_t *zc)
4319 dmu_objset_type_t ost;
4320 boolean_t recursive = zc->zc_cookie & 1;
4324 /* "zfs rename" from and to ...%recv datasets should both fail */
4325 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
4326 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
4327 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0 ||
4328 dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
4329 strchr(zc->zc_name, '%') || strchr(zc->zc_value, '%'))
4330 return (SET_ERROR(EINVAL));
4332 err = dmu_objset_hold(zc->zc_name, FTAG, &os);
4335 ost = dmu_objset_type(os);
4336 dmu_objset_rele(os, FTAG);
4338 at = strchr(zc->zc_name, '@');
4340 /* snaps must be in same fs */
4343 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
4344 return (SET_ERROR(EXDEV));
4346 if (ost == DMU_OST_ZFS) {
4347 error = dmu_objset_find(zc->zc_name,
4348 recursive_unmount, at + 1,
4349 recursive ? DS_FIND_CHILDREN : 0);
4355 error = dsl_dataset_rename_snapshot(zc->zc_name,
4356 at + 1, strchr(zc->zc_value, '@') + 1, recursive);
4361 return (dsl_dir_rename(zc->zc_name, zc->zc_value));
4366 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
4368 const char *propname = nvpair_name(pair);
4369 boolean_t issnap = (strchr(dsname, '@') != NULL);
4370 zfs_prop_t prop = zfs_name_to_prop(propname);
4374 if (prop == ZPROP_INVAL) {
4375 if (zfs_prop_user(propname)) {
4376 if ((err = zfs_secpolicy_write_perms(dsname,
4377 ZFS_DELEG_PERM_USERPROP, cr)))
4382 if (!issnap && zfs_prop_userquota(propname)) {
4383 const char *perm = NULL;
4384 const char *uq_prefix =
4385 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
4386 const char *gq_prefix =
4387 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
4388 const char *uiq_prefix =
4389 zfs_userquota_prop_prefixes[ZFS_PROP_USEROBJQUOTA];
4390 const char *giq_prefix =
4391 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPOBJQUOTA];
4392 const char *pq_prefix =
4393 zfs_userquota_prop_prefixes[ZFS_PROP_PROJECTQUOTA];
4394 const char *piq_prefix = zfs_userquota_prop_prefixes[\
4395 ZFS_PROP_PROJECTOBJQUOTA];
4397 if (strncmp(propname, uq_prefix,
4398 strlen(uq_prefix)) == 0) {
4399 perm = ZFS_DELEG_PERM_USERQUOTA;
4400 } else if (strncmp(propname, uiq_prefix,
4401 strlen(uiq_prefix)) == 0) {
4402 perm = ZFS_DELEG_PERM_USEROBJQUOTA;
4403 } else if (strncmp(propname, gq_prefix,
4404 strlen(gq_prefix)) == 0) {
4405 perm = ZFS_DELEG_PERM_GROUPQUOTA;
4406 } else if (strncmp(propname, giq_prefix,
4407 strlen(giq_prefix)) == 0) {
4408 perm = ZFS_DELEG_PERM_GROUPOBJQUOTA;
4409 } else if (strncmp(propname, pq_prefix,
4410 strlen(pq_prefix)) == 0) {
4411 perm = ZFS_DELEG_PERM_PROJECTQUOTA;
4412 } else if (strncmp(propname, piq_prefix,
4413 strlen(piq_prefix)) == 0) {
4414 perm = ZFS_DELEG_PERM_PROJECTOBJQUOTA;
4416 /* {USER|GROUP|PROJECT}USED are read-only */
4417 return (SET_ERROR(EINVAL));
4420 if ((err = zfs_secpolicy_write_perms(dsname, perm, cr)))
4425 return (SET_ERROR(EINVAL));
4429 return (SET_ERROR(EINVAL));
4431 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
4433 * dsl_prop_get_all_impl() returns properties in this
4437 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
4438 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4443 * Check that this value is valid for this pool version
4446 case ZFS_PROP_COMPRESSION:
4448 * If the user specified gzip compression, make sure
4449 * the SPA supports it. We ignore any errors here since
4450 * we'll catch them later.
4452 if (nvpair_value_uint64(pair, &intval) == 0) {
4453 if (intval >= ZIO_COMPRESS_GZIP_1 &&
4454 intval <= ZIO_COMPRESS_GZIP_9 &&
4455 zfs_earlier_version(dsname,
4456 SPA_VERSION_GZIP_COMPRESSION)) {
4457 return (SET_ERROR(ENOTSUP));
4460 if (intval == ZIO_COMPRESS_ZLE &&
4461 zfs_earlier_version(dsname,
4462 SPA_VERSION_ZLE_COMPRESSION))
4463 return (SET_ERROR(ENOTSUP));
4465 if (intval == ZIO_COMPRESS_LZ4) {
4468 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4471 if (!spa_feature_is_enabled(spa,
4472 SPA_FEATURE_LZ4_COMPRESS)) {
4473 spa_close(spa, FTAG);
4474 return (SET_ERROR(ENOTSUP));
4476 spa_close(spa, FTAG);
4480 * If this is a bootable dataset then
4481 * verify that the compression algorithm
4482 * is supported for booting. We must return
4483 * something other than ENOTSUP since it
4484 * implies a downrev pool version.
4486 if (zfs_is_bootfs(dsname) &&
4487 !BOOTFS_COMPRESS_VALID(intval)) {
4488 return (SET_ERROR(ERANGE));
4493 case ZFS_PROP_COPIES:
4494 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
4495 return (SET_ERROR(ENOTSUP));
4498 case ZFS_PROP_VOLBLOCKSIZE:
4499 case ZFS_PROP_RECORDSIZE:
4500 /* Record sizes above 128k need the feature to be enabled */
4501 if (nvpair_value_uint64(pair, &intval) == 0 &&
4502 intval > SPA_OLD_MAXBLOCKSIZE) {
4506 * We don't allow setting the property above 1MB,
4507 * unless the tunable has been changed.
4509 if (intval > zfs_max_recordsize ||
4510 intval > SPA_MAXBLOCKSIZE)
4511 return (SET_ERROR(ERANGE));
4513 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4516 if (!spa_feature_is_enabled(spa,
4517 SPA_FEATURE_LARGE_BLOCKS)) {
4518 spa_close(spa, FTAG);
4519 return (SET_ERROR(ENOTSUP));
4521 spa_close(spa, FTAG);
4525 case ZFS_PROP_DNODESIZE:
4526 /* Dnode sizes above 512 need the feature to be enabled */
4527 if (nvpair_value_uint64(pair, &intval) == 0 &&
4528 intval != ZFS_DNSIZE_LEGACY) {
4532 * If this is a bootable dataset then
4533 * we don't allow large (>512B) dnodes,
4534 * because GRUB doesn't support them.
4536 if (zfs_is_bootfs(dsname) &&
4537 intval != ZFS_DNSIZE_LEGACY) {
4538 return (SET_ERROR(EDOM));
4541 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4544 if (!spa_feature_is_enabled(spa,
4545 SPA_FEATURE_LARGE_DNODE)) {
4546 spa_close(spa, FTAG);
4547 return (SET_ERROR(ENOTSUP));
4549 spa_close(spa, FTAG);
4553 case ZFS_PROP_SPECIAL_SMALL_BLOCKS:
4555 * This property could require the allocation classes
4556 * feature to be active for setting, however we allow
4557 * it so that tests of settable properties succeed.
4558 * The CLI will issue a warning in this case.
4562 case ZFS_PROP_SHARESMB:
4563 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
4564 return (SET_ERROR(ENOTSUP));
4567 case ZFS_PROP_ACLINHERIT:
4568 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
4569 nvpair_value_uint64(pair, &intval) == 0) {
4570 if (intval == ZFS_ACL_PASSTHROUGH_X &&
4571 zfs_earlier_version(dsname,
4572 SPA_VERSION_PASSTHROUGH_X))
4573 return (SET_ERROR(ENOTSUP));
4576 case ZFS_PROP_CHECKSUM:
4577 case ZFS_PROP_DEDUP:
4579 spa_feature_t feature;
4583 /* dedup feature version checks */
4584 if (prop == ZFS_PROP_DEDUP &&
4585 zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
4586 return (SET_ERROR(ENOTSUP));
4588 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
4589 nvpair_value_uint64(pair, &intval) == 0) {
4590 /* check prop value is enabled in features */
4591 feature = zio_checksum_to_feature(
4592 intval & ZIO_CHECKSUM_MASK);
4593 if (feature == SPA_FEATURE_NONE)
4596 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4599 if (!spa_feature_is_enabled(spa, feature)) {
4600 spa_close(spa, FTAG);
4601 return (SET_ERROR(ENOTSUP));
4603 spa_close(spa, FTAG);
4612 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
4616 * Removes properties from the given props list that fail permission checks
4617 * needed to clear them and to restore them in case of a receive error. For each
4618 * property, make sure we have both set and inherit permissions.
4620 * Returns the first error encountered if any permission checks fail. If the
4621 * caller provides a non-NULL errlist, it also gives the complete list of names
4622 * of all the properties that failed a permission check along with the
4623 * corresponding error numbers. The caller is responsible for freeing the
4626 * If every property checks out successfully, zero is returned and the list
4627 * pointed at by errlist is NULL.
4630 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
4633 nvpair_t *pair, *next_pair;
4640 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4642 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
4643 (void) strlcpy(zc->zc_name, dataset, sizeof (zc->zc_name));
4644 pair = nvlist_next_nvpair(props, NULL);
4645 while (pair != NULL) {
4646 next_pair = nvlist_next_nvpair(props, pair);
4648 (void) strlcpy(zc->zc_value, nvpair_name(pair),
4649 sizeof (zc->zc_value));
4650 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
4651 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
4652 VERIFY(nvlist_remove_nvpair(props, pair) == 0);
4653 VERIFY(nvlist_add_int32(errors,
4654 zc->zc_value, err) == 0);
4658 kmem_free(zc, sizeof (zfs_cmd_t));
4660 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
4661 nvlist_free(errors);
4664 VERIFY(nvpair_value_int32(pair, &rv) == 0);
4667 if (errlist == NULL)
4668 nvlist_free(errors);
4676 propval_equals(nvpair_t *p1, nvpair_t *p2)
4678 if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
4679 /* dsl_prop_get_all_impl() format */
4681 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
4682 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4686 if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
4688 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
4689 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4693 if (nvpair_type(p1) != nvpair_type(p2))
4696 if (nvpair_type(p1) == DATA_TYPE_STRING) {
4697 char *valstr1, *valstr2;
4699 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
4700 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
4701 return (strcmp(valstr1, valstr2) == 0);
4703 uint64_t intval1, intval2;
4705 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
4706 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
4707 return (intval1 == intval2);
4712 * Remove properties from props if they are not going to change (as determined
4713 * by comparison with origprops). Remove them from origprops as well, since we
4714 * do not need to clear or restore properties that won't change.
4717 props_reduce(nvlist_t *props, nvlist_t *origprops)
4719 nvpair_t *pair, *next_pair;
4721 if (origprops == NULL)
4722 return; /* all props need to be received */
4724 pair = nvlist_next_nvpair(props, NULL);
4725 while (pair != NULL) {
4726 const char *propname = nvpair_name(pair);
4729 next_pair = nvlist_next_nvpair(props, pair);
4731 if ((nvlist_lookup_nvpair(origprops, propname,
4732 &match) != 0) || !propval_equals(pair, match))
4733 goto next; /* need to set received value */
4735 /* don't clear the existing received value */
4736 (void) nvlist_remove_nvpair(origprops, match);
4737 /* don't bother receiving the property */
4738 (void) nvlist_remove_nvpair(props, pair);
4745 * Extract properties that cannot be set PRIOR to the receipt of a dataset.
4746 * For example, refquota cannot be set until after the receipt of a dataset,
4747 * because in replication streams, an older/earlier snapshot may exceed the
4748 * refquota. We want to receive the older/earlier snapshot, but setting
4749 * refquota pre-receipt will set the dsl's ACTUAL quota, which will prevent
4750 * the older/earlier snapshot from being received (with EDQUOT).
4752 * The ZFS test "zfs_receive_011_pos" demonstrates such a scenario.
4754 * libzfs will need to be judicious handling errors encountered by props
4755 * extracted by this function.
4758 extract_delay_props(nvlist_t *props)
4760 nvlist_t *delayprops;
4761 nvpair_t *nvp, *tmp;
4762 static const zfs_prop_t delayable[] = {
4764 ZFS_PROP_KEYLOCATION,
4769 VERIFY(nvlist_alloc(&delayprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4771 for (nvp = nvlist_next_nvpair(props, NULL); nvp != NULL;
4772 nvp = nvlist_next_nvpair(props, nvp)) {
4774 * strcmp() is safe because zfs_prop_to_name() always returns
4777 for (i = 0; delayable[i] != 0; i++) {
4778 if (strcmp(zfs_prop_to_name(delayable[i]),
4779 nvpair_name(nvp)) == 0) {
4783 if (delayable[i] != 0) {
4784 tmp = nvlist_prev_nvpair(props, nvp);
4785 VERIFY(nvlist_add_nvpair(delayprops, nvp) == 0);
4786 VERIFY(nvlist_remove_nvpair(props, nvp) == 0);
4791 if (nvlist_empty(delayprops)) {
4792 nvlist_free(delayprops);
4795 return (delayprops);
4799 zfs_allow_log_destroy(void *arg)
4801 char *poolname = arg;
4803 if (poolname != NULL)
4804 kmem_strfree(poolname);
4808 static boolean_t zfs_ioc_recv_inject_err;
4812 * nvlist 'errors' is always allocated. It will contain descriptions of
4813 * encountered errors, if any. It's the callers responsibility to free.
4816 zfs_ioc_recv_impl(char *tofs, char *tosnap, char *origin, nvlist_t *recvprops,
4817 nvlist_t *localprops, nvlist_t *hidden_args, boolean_t force,
4818 boolean_t resumable, int input_fd,
4819 dmu_replay_record_t *begin_record, uint64_t *read_bytes,
4820 uint64_t *errflags, nvlist_t **errors)
4822 dmu_recv_cookie_t drc;
4824 int props_error = 0;
4826 nvlist_t *local_delayprops = NULL;
4827 nvlist_t *recv_delayprops = NULL;
4828 nvlist_t *origprops = NULL; /* existing properties */
4829 nvlist_t *origrecvd = NULL; /* existing received properties */
4830 boolean_t first_recvd_props = B_FALSE;
4831 boolean_t tofs_was_redacted;
4832 zfs_file_t *input_fp;
4836 *errors = fnvlist_alloc();
4839 if ((error = zfs_file_get(input_fd, &input_fp)))
4842 noff = off = zfs_file_off(input_fp);
4843 error = dmu_recv_begin(tofs, tosnap, begin_record, force,
4844 resumable, localprops, hidden_args, origin, &drc, input_fp,
4848 tofs_was_redacted = dsl_get_redacted(drc.drc_ds);
4851 * Set properties before we receive the stream so that they are applied
4852 * to the new data. Note that we must call dmu_recv_stream() if
4853 * dmu_recv_begin() succeeds.
4855 if (recvprops != NULL && !drc.drc_newfs) {
4856 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
4857 SPA_VERSION_RECVD_PROPS &&
4858 !dsl_prop_get_hasrecvd(tofs))
4859 first_recvd_props = B_TRUE;
4862 * If new received properties are supplied, they are to
4863 * completely replace the existing received properties,
4864 * so stash away the existing ones.
4866 if (dsl_prop_get_received(tofs, &origrecvd) == 0) {
4867 nvlist_t *errlist = NULL;
4869 * Don't bother writing a property if its value won't
4870 * change (and avoid the unnecessary security checks).
4872 * The first receive after SPA_VERSION_RECVD_PROPS is a
4873 * special case where we blow away all local properties
4876 if (!first_recvd_props)
4877 props_reduce(recvprops, origrecvd);
4878 if (zfs_check_clearable(tofs, origrecvd, &errlist) != 0)
4879 (void) nvlist_merge(*errors, errlist, 0);
4880 nvlist_free(errlist);
4882 if (clear_received_props(tofs, origrecvd,
4883 first_recvd_props ? NULL : recvprops) != 0)
4884 *errflags |= ZPROP_ERR_NOCLEAR;
4886 *errflags |= ZPROP_ERR_NOCLEAR;
4891 * Stash away existing properties so we can restore them on error unless
4892 * we're doing the first receive after SPA_VERSION_RECVD_PROPS, in which
4893 * case "origrecvd" will take care of that.
4895 if (localprops != NULL && !drc.drc_newfs && !first_recvd_props) {
4897 if (dmu_objset_hold(tofs, FTAG, &os) == 0) {
4898 if (dsl_prop_get_all(os, &origprops) != 0) {
4899 *errflags |= ZPROP_ERR_NOCLEAR;
4901 dmu_objset_rele(os, FTAG);
4903 *errflags |= ZPROP_ERR_NOCLEAR;
4907 if (recvprops != NULL) {
4908 props_error = dsl_prop_set_hasrecvd(tofs);
4910 if (props_error == 0) {
4911 recv_delayprops = extract_delay_props(recvprops);
4912 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4913 recvprops, *errors);
4917 if (localprops != NULL) {
4918 nvlist_t *oprops = fnvlist_alloc();
4919 nvlist_t *xprops = fnvlist_alloc();
4920 nvpair_t *nvp = NULL;
4922 while ((nvp = nvlist_next_nvpair(localprops, nvp)) != NULL) {
4923 if (nvpair_type(nvp) == DATA_TYPE_BOOLEAN) {
4925 const char *name = nvpair_name(nvp);
4926 zfs_prop_t prop = zfs_name_to_prop(name);
4927 if (prop != ZPROP_INVAL) {
4928 if (!zfs_prop_inheritable(prop))
4930 } else if (!zfs_prop_user(name))
4932 fnvlist_add_boolean(xprops, name);
4934 /* -o property=value */
4935 fnvlist_add_nvpair(oprops, nvp);
4939 local_delayprops = extract_delay_props(oprops);
4940 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_LOCAL,
4942 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_INHERITED,
4945 nvlist_free(oprops);
4946 nvlist_free(xprops);
4949 error = dmu_recv_stream(&drc, &off);
4952 zfsvfs_t *zfsvfs = NULL;
4953 zvol_state_handle_t *zv = NULL;
4955 if (getzfsvfs(tofs, &zfsvfs) == 0) {
4959 boolean_t stream_is_redacted = DMU_GET_FEATUREFLAGS(
4960 begin_record->drr_u.drr_begin.
4961 drr_versioninfo) & DMU_BACKUP_FEATURE_REDACTED;
4963 ds = dmu_objset_ds(zfsvfs->z_os);
4964 error = zfs_suspend_fs(zfsvfs);
4966 * If the suspend fails, then the recv_end will
4967 * likely also fail, and clean up after itself.
4969 end_err = dmu_recv_end(&drc, zfsvfs);
4971 * If the dataset was not redacted, but we received a
4972 * redacted stream onto it, we need to unmount the
4973 * dataset. Otherwise, resume the filesystem.
4975 if (error == 0 && !drc.drc_newfs &&
4976 stream_is_redacted && !tofs_was_redacted) {
4977 error = zfs_end_fs(zfsvfs, ds);
4978 } else if (error == 0) {
4979 error = zfs_resume_fs(zfsvfs, ds);
4981 error = error ? error : end_err;
4982 zfs_vfs_rele(zfsvfs);
4983 } else if ((zv = zvol_suspend(tofs)) != NULL) {
4984 error = dmu_recv_end(&drc, zvol_tag(zv));
4987 error = dmu_recv_end(&drc, NULL);
4990 /* Set delayed properties now, after we're done receiving. */
4991 if (recv_delayprops != NULL && error == 0) {
4992 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4993 recv_delayprops, *errors);
4995 if (local_delayprops != NULL && error == 0) {
4996 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_LOCAL,
4997 local_delayprops, *errors);
5002 * Merge delayed props back in with initial props, in case
5003 * we're DEBUG and zfs_ioc_recv_inject_err is set (which means
5004 * we have to make sure clear_received_props() includes
5005 * the delayed properties).
5007 * Since zfs_ioc_recv_inject_err is only in DEBUG kernels,
5008 * using ASSERT() will be just like a VERIFY.
5010 if (recv_delayprops != NULL) {
5011 ASSERT(nvlist_merge(recvprops, recv_delayprops, 0) == 0);
5012 nvlist_free(recv_delayprops);
5014 if (local_delayprops != NULL) {
5015 ASSERT(nvlist_merge(localprops, local_delayprops, 0) == 0);
5016 nvlist_free(local_delayprops);
5018 *read_bytes = off - noff;
5021 if (zfs_ioc_recv_inject_err) {
5022 zfs_ioc_recv_inject_err = B_FALSE;
5028 * On error, restore the original props.
5030 if (error != 0 && recvprops != NULL && !drc.drc_newfs) {
5031 if (clear_received_props(tofs, recvprops, NULL) != 0) {
5033 * We failed to clear the received properties.
5034 * Since we may have left a $recvd value on the
5035 * system, we can't clear the $hasrecvd flag.
5037 *errflags |= ZPROP_ERR_NORESTORE;
5038 } else if (first_recvd_props) {
5039 dsl_prop_unset_hasrecvd(tofs);
5042 if (origrecvd == NULL && !drc.drc_newfs) {
5043 /* We failed to stash the original properties. */
5044 *errflags |= ZPROP_ERR_NORESTORE;
5048 * dsl_props_set() will not convert RECEIVED to LOCAL on or
5049 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
5050 * explicitly if we're restoring local properties cleared in the
5051 * first new-style receive.
5053 if (origrecvd != NULL &&
5054 zfs_set_prop_nvlist(tofs, (first_recvd_props ?
5055 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
5056 origrecvd, NULL) != 0) {
5058 * We stashed the original properties but failed to
5061 *errflags |= ZPROP_ERR_NORESTORE;
5064 if (error != 0 && localprops != NULL && !drc.drc_newfs &&
5065 !first_recvd_props) {
5067 nvlist_t *inheritprops;
5070 if (origprops == NULL) {
5071 /* We failed to stash the original properties. */
5072 *errflags |= ZPROP_ERR_NORESTORE;
5076 /* Restore original props */
5077 setprops = fnvlist_alloc();
5078 inheritprops = fnvlist_alloc();
5080 while ((nvp = nvlist_next_nvpair(localprops, nvp)) != NULL) {
5081 const char *name = nvpair_name(nvp);
5085 if (!nvlist_exists(origprops, name)) {
5087 * Property was not present or was explicitly
5088 * inherited before the receive, restore this.
5090 fnvlist_add_boolean(inheritprops, name);
5093 attrs = fnvlist_lookup_nvlist(origprops, name);
5094 source = fnvlist_lookup_string(attrs, ZPROP_SOURCE);
5096 /* Skip received properties */
5097 if (strcmp(source, ZPROP_SOURCE_VAL_RECVD) == 0)
5100 if (strcmp(source, tofs) == 0) {
5101 /* Property was locally set */
5102 fnvlist_add_nvlist(setprops, name, attrs);
5104 /* Property was implicitly inherited */
5105 fnvlist_add_boolean(inheritprops, name);
5109 if (zfs_set_prop_nvlist(tofs, ZPROP_SRC_LOCAL, setprops,
5111 *errflags |= ZPROP_ERR_NORESTORE;
5112 if (zfs_set_prop_nvlist(tofs, ZPROP_SRC_INHERITED, inheritprops,
5114 *errflags |= ZPROP_ERR_NORESTORE;
5116 nvlist_free(setprops);
5117 nvlist_free(inheritprops);
5120 zfs_file_put(input_fd);
5121 nvlist_free(origrecvd);
5122 nvlist_free(origprops);
5125 error = props_error;
5132 * zc_name name of containing filesystem (unused)
5133 * zc_nvlist_src{_size} nvlist of properties to apply
5134 * zc_nvlist_conf{_size} nvlist of properties to exclude
5135 * (DATA_TYPE_BOOLEAN) and override (everything else)
5136 * zc_value name of snapshot to create
5137 * zc_string name of clone origin (if DRR_FLAG_CLONE)
5138 * zc_cookie file descriptor to recv from
5139 * zc_begin_record the BEGIN record of the stream (not byteswapped)
5140 * zc_guid force flag
5143 * zc_cookie number of bytes read
5144 * zc_obj zprop_errflags_t
5145 * zc_nvlist_dst{_size} error for each unapplied received property
5148 zfs_ioc_recv(zfs_cmd_t *zc)
5150 dmu_replay_record_t begin_record;
5151 nvlist_t *errors = NULL;
5152 nvlist_t *recvdprops = NULL;
5153 nvlist_t *localprops = NULL;
5154 char *origin = NULL;
5156 char tofs[ZFS_MAX_DATASET_NAME_LEN];
5159 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
5160 strchr(zc->zc_value, '@') == NULL ||
5161 strchr(zc->zc_value, '%'))
5162 return (SET_ERROR(EINVAL));
5164 (void) strlcpy(tofs, zc->zc_value, sizeof (tofs));
5165 tosnap = strchr(tofs, '@');
5168 if (zc->zc_nvlist_src != 0 &&
5169 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
5170 zc->zc_iflags, &recvdprops)) != 0)
5173 if (zc->zc_nvlist_conf != 0 &&
5174 (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
5175 zc->zc_iflags, &localprops)) != 0)
5178 if (zc->zc_string[0])
5179 origin = zc->zc_string;
5181 begin_record.drr_type = DRR_BEGIN;
5182 begin_record.drr_payloadlen = 0;
5183 begin_record.drr_u.drr_begin = zc->zc_begin_record;
5185 error = zfs_ioc_recv_impl(tofs, tosnap, origin, recvdprops, localprops,
5186 NULL, zc->zc_guid, B_FALSE, zc->zc_cookie, &begin_record,
5187 &zc->zc_cookie, &zc->zc_obj, &errors);
5188 nvlist_free(recvdprops);
5189 nvlist_free(localprops);
5192 * Now that all props, initial and delayed, are set, report the prop
5193 * errors to the caller.
5195 if (zc->zc_nvlist_dst_size != 0 && errors != NULL &&
5196 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
5197 put_nvlist(zc, errors) != 0)) {
5199 * Caller made zc->zc_nvlist_dst less than the minimum expected
5200 * size or supplied an invalid address.
5202 error = SET_ERROR(EINVAL);
5205 nvlist_free(errors);
5212 * "snapname" -> full name of the snapshot to create
5213 * (optional) "props" -> received properties to set (nvlist)
5214 * (optional) "localprops" -> override and exclude properties (nvlist)
5215 * (optional) "origin" -> name of clone origin (DRR_FLAG_CLONE)
5216 * "begin_record" -> non-byteswapped dmu_replay_record_t
5217 * "input_fd" -> file descriptor to read stream from (int32)
5218 * (optional) "force" -> force flag (value ignored)
5219 * (optional) "resumable" -> resumable flag (value ignored)
5220 * (optional) "cleanup_fd" -> unused
5221 * (optional) "action_handle" -> unused
5222 * (optional) "hidden_args" -> { "wkeydata" -> value }
5226 * "read_bytes" -> number of bytes read
5227 * "error_flags" -> zprop_errflags_t
5228 * "errors" -> error for each unapplied received property (nvlist)
5231 static const zfs_ioc_key_t zfs_keys_recv_new[] = {
5232 {"snapname", DATA_TYPE_STRING, 0},
5233 {"props", DATA_TYPE_NVLIST, ZK_OPTIONAL},
5234 {"localprops", DATA_TYPE_NVLIST, ZK_OPTIONAL},
5235 {"origin", DATA_TYPE_STRING, ZK_OPTIONAL},
5236 {"begin_record", DATA_TYPE_BYTE_ARRAY, 0},
5237 {"input_fd", DATA_TYPE_INT32, 0},
5238 {"force", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
5239 {"resumable", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
5240 {"cleanup_fd", DATA_TYPE_INT32, ZK_OPTIONAL},
5241 {"action_handle", DATA_TYPE_UINT64, ZK_OPTIONAL},
5242 {"hidden_args", DATA_TYPE_NVLIST, ZK_OPTIONAL},
5246 zfs_ioc_recv_new(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
5248 dmu_replay_record_t *begin_record;
5249 uint_t begin_record_size;
5250 nvlist_t *errors = NULL;
5251 nvlist_t *recvprops = NULL;
5252 nvlist_t *localprops = NULL;
5253 nvlist_t *hidden_args = NULL;
5255 char *origin = NULL;
5257 char tofs[ZFS_MAX_DATASET_NAME_LEN];
5259 boolean_t resumable;
5260 uint64_t read_bytes = 0;
5261 uint64_t errflags = 0;
5265 snapname = fnvlist_lookup_string(innvl, "snapname");
5267 if (dataset_namecheck(snapname, NULL, NULL) != 0 ||
5268 strchr(snapname, '@') == NULL ||
5269 strchr(snapname, '%'))
5270 return (SET_ERROR(EINVAL));
5272 (void) strlcpy(tofs, snapname, sizeof (tofs));
5273 tosnap = strchr(tofs, '@');
5276 error = nvlist_lookup_string(innvl, "origin", &origin);
5277 if (error && error != ENOENT)
5280 error = nvlist_lookup_byte_array(innvl, "begin_record",
5281 (uchar_t **)&begin_record, &begin_record_size);
5282 if (error != 0 || begin_record_size != sizeof (*begin_record))
5283 return (SET_ERROR(EINVAL));
5285 input_fd = fnvlist_lookup_int32(innvl, "input_fd");
5287 force = nvlist_exists(innvl, "force");
5288 resumable = nvlist_exists(innvl, "resumable");
5290 /* we still use "props" here for backwards compatibility */
5291 error = nvlist_lookup_nvlist(innvl, "props", &recvprops);
5292 if (error && error != ENOENT)
5295 error = nvlist_lookup_nvlist(innvl, "localprops", &localprops);
5296 if (error && error != ENOENT)
5299 error = nvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS, &hidden_args);
5300 if (error && error != ENOENT)
5303 error = zfs_ioc_recv_impl(tofs, tosnap, origin, recvprops, localprops,
5304 hidden_args, force, resumable, input_fd, begin_record,
5305 &read_bytes, &errflags, &errors);
5307 fnvlist_add_uint64(outnvl, "read_bytes", read_bytes);
5308 fnvlist_add_uint64(outnvl, "error_flags", errflags);
5309 fnvlist_add_nvlist(outnvl, "errors", errors);
5311 nvlist_free(errors);
5312 nvlist_free(recvprops);
5313 nvlist_free(localprops);
5318 typedef struct dump_bytes_io {
5326 dump_bytes_cb(void *arg)
5328 dump_bytes_io_t *dbi = (dump_bytes_io_t *)arg;
5335 dbi->dbi_err = zfs_file_write(fp, buf, dbi->dbi_len, NULL);
5339 dump_bytes(objset_t *os, void *buf, int len, void *arg)
5341 dump_bytes_io_t dbi;
5347 #if defined(HAVE_LARGE_STACKS)
5348 dump_bytes_cb(&dbi);
5351 * The vn_rdwr() call is performed in a taskq to ensure that there is
5352 * always enough stack space to write safely to the target filesystem.
5353 * The ZIO_TYPE_FREE threads are used because there can be a lot of
5354 * them and they are used in vdev_file.c for a similar purpose.
5356 spa_taskq_dispatch_sync(dmu_objset_spa(os), ZIO_TYPE_FREE,
5357 ZIO_TASKQ_ISSUE, dump_bytes_cb, &dbi, TQ_SLEEP);
5358 #endif /* HAVE_LARGE_STACKS */
5360 return (dbi.dbi_err);
5365 * zc_name name of snapshot to send
5366 * zc_cookie file descriptor to send stream to
5367 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj)
5368 * zc_sendobj objsetid of snapshot to send
5369 * zc_fromobj objsetid of incremental fromsnap (may be zero)
5370 * zc_guid if set, estimate size of stream only. zc_cookie is ignored.
5371 * output size in zc_objset_type.
5372 * zc_flags lzc_send_flags
5375 * zc_objset_type estimated size, if zc_guid is set
5377 * NOTE: This is no longer the preferred interface, any new functionality
5378 * should be added to zfs_ioc_send_new() instead.
5381 zfs_ioc_send(zfs_cmd_t *zc)
5385 boolean_t estimate = (zc->zc_guid != 0);
5386 boolean_t embedok = (zc->zc_flags & 0x1);
5387 boolean_t large_block_ok = (zc->zc_flags & 0x2);
5388 boolean_t compressok = (zc->zc_flags & 0x4);
5389 boolean_t rawok = (zc->zc_flags & 0x8);
5390 boolean_t savedok = (zc->zc_flags & 0x10);
5392 if (zc->zc_obj != 0) {
5394 dsl_dataset_t *tosnap;
5396 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5400 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
5402 dsl_pool_rele(dp, FTAG);
5406 if (dsl_dir_is_clone(tosnap->ds_dir))
5408 dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj;
5409 dsl_dataset_rele(tosnap, FTAG);
5410 dsl_pool_rele(dp, FTAG);
5415 dsl_dataset_t *tosnap;
5416 dsl_dataset_t *fromsnap = NULL;
5418 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5422 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj,
5425 dsl_pool_rele(dp, FTAG);
5429 if (zc->zc_fromobj != 0) {
5430 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
5433 dsl_dataset_rele(tosnap, FTAG);
5434 dsl_pool_rele(dp, FTAG);
5439 error = dmu_send_estimate_fast(tosnap, fromsnap, NULL,
5440 compressok || rawok, savedok, &zc->zc_objset_type);
5442 if (fromsnap != NULL)
5443 dsl_dataset_rele(fromsnap, FTAG);
5444 dsl_dataset_rele(tosnap, FTAG);
5445 dsl_pool_rele(dp, FTAG);
5448 dmu_send_outparams_t out = {0};
5450 if ((error = zfs_file_get(zc->zc_cookie, &fp)))
5453 off = zfs_file_off(fp);
5454 out.dso_outfunc = dump_bytes;
5456 out.dso_dryrun = B_FALSE;
5457 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
5458 zc->zc_fromobj, embedok, large_block_ok, compressok,
5459 rawok, savedok, zc->zc_cookie, &off, &out);
5461 zfs_file_put(zc->zc_cookie);
5468 * zc_name name of snapshot on which to report progress
5469 * zc_cookie file descriptor of send stream
5472 * zc_cookie number of bytes written in send stream thus far
5473 * zc_objset_type logical size of data traversed by send thus far
5476 zfs_ioc_send_progress(zfs_cmd_t *zc)
5480 dmu_sendstatus_t *dsp = NULL;
5483 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5487 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
5489 dsl_pool_rele(dp, FTAG);
5493 mutex_enter(&ds->ds_sendstream_lock);
5496 * Iterate over all the send streams currently active on this dataset.
5497 * If there's one which matches the specified file descriptor _and_ the
5498 * stream was started by the current process, return the progress of
5502 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
5503 dsp = list_next(&ds->ds_sendstreams, dsp)) {
5504 if (dsp->dss_outfd == zc->zc_cookie &&
5505 zfs_proc_is_caller(dsp->dss_proc))
5510 zc->zc_cookie = atomic_cas_64((volatile uint64_t *)dsp->dss_off,
5512 /* This is the closest thing we have to atomic_read_64. */
5513 zc->zc_objset_type = atomic_cas_64(&dsp->dss_blocks, 0, 0);
5515 error = SET_ERROR(ENOENT);
5518 mutex_exit(&ds->ds_sendstream_lock);
5519 dsl_dataset_rele(ds, FTAG);
5520 dsl_pool_rele(dp, FTAG);
5525 zfs_ioc_inject_fault(zfs_cmd_t *zc)
5529 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
5530 &zc->zc_inject_record);
5533 zc->zc_guid = (uint64_t)id;
5539 zfs_ioc_clear_fault(zfs_cmd_t *zc)
5541 return (zio_clear_fault((int)zc->zc_guid));
5545 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
5547 int id = (int)zc->zc_guid;
5550 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
5551 &zc->zc_inject_record);
5559 zfs_ioc_error_log(zfs_cmd_t *zc)
5563 size_t count = (size_t)zc->zc_nvlist_dst_size;
5565 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
5568 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
5571 zc->zc_nvlist_dst_size = count;
5573 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
5575 spa_close(spa, FTAG);
5581 zfs_ioc_clear(zfs_cmd_t *zc)
5588 * On zpool clear we also fix up missing slogs
5590 mutex_enter(&spa_namespace_lock);
5591 spa = spa_lookup(zc->zc_name);
5593 mutex_exit(&spa_namespace_lock);
5594 return (SET_ERROR(EIO));
5596 if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
5597 /* we need to let spa_open/spa_load clear the chains */
5598 spa_set_log_state(spa, SPA_LOG_CLEAR);
5600 spa->spa_last_open_failed = 0;
5601 mutex_exit(&spa_namespace_lock);
5603 if (zc->zc_cookie & ZPOOL_NO_REWIND) {
5604 error = spa_open(zc->zc_name, &spa, FTAG);
5607 nvlist_t *config = NULL;
5609 if (zc->zc_nvlist_src == 0)
5610 return (SET_ERROR(EINVAL));
5612 if ((error = get_nvlist(zc->zc_nvlist_src,
5613 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
5614 error = spa_open_rewind(zc->zc_name, &spa, FTAG,
5616 if (config != NULL) {
5619 if ((err = put_nvlist(zc, config)) != 0)
5621 nvlist_free(config);
5623 nvlist_free(policy);
5631 * If multihost is enabled, resuming I/O is unsafe as another
5632 * host may have imported the pool.
5634 if (spa_multihost(spa) && spa_suspended(spa))
5635 return (SET_ERROR(EINVAL));
5637 spa_vdev_state_enter(spa, SCL_NONE);
5639 if (zc->zc_guid == 0) {
5642 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
5644 error = SET_ERROR(ENODEV);
5645 (void) spa_vdev_state_exit(spa, NULL, error);
5646 spa_close(spa, FTAG);
5651 vdev_clear(spa, vd);
5653 (void) spa_vdev_state_exit(spa, spa_suspended(spa) ?
5654 NULL : spa->spa_root_vdev, 0);
5657 * Resume any suspended I/Os.
5659 if (zio_resume(spa) != 0)
5660 error = SET_ERROR(EIO);
5662 spa_close(spa, FTAG);
5668 * Reopen all the vdevs associated with the pool.
5671 * "scrub_restart" -> when true and scrub is running, allow to restart
5672 * scrub as the side effect of the reopen (boolean).
5677 static const zfs_ioc_key_t zfs_keys_pool_reopen[] = {
5678 {"scrub_restart", DATA_TYPE_BOOLEAN_VALUE, ZK_OPTIONAL},
5683 zfs_ioc_pool_reopen(const char *pool, nvlist_t *innvl, nvlist_t *outnvl)
5687 boolean_t rc, scrub_restart = B_TRUE;
5690 error = nvlist_lookup_boolean_value(innvl,
5691 "scrub_restart", &rc);
5696 error = spa_open(pool, &spa, FTAG);
5700 spa_vdev_state_enter(spa, SCL_NONE);
5703 * If the scrub_restart flag is B_FALSE and a scrub is already
5704 * in progress then set spa_scrub_reopen flag to B_TRUE so that
5705 * we don't restart the scrub as a side effect of the reopen.
5706 * Otherwise, let vdev_open() decided if a resilver is required.
5709 spa->spa_scrub_reopen = (!scrub_restart &&
5710 dsl_scan_scrubbing(spa->spa_dsl_pool));
5711 vdev_reopen(spa->spa_root_vdev);
5712 spa->spa_scrub_reopen = B_FALSE;
5714 (void) spa_vdev_state_exit(spa, NULL, 0);
5715 spa_close(spa, FTAG);
5721 * zc_name name of filesystem
5724 * zc_string name of conflicting snapshot, if there is one
5727 zfs_ioc_promote(zfs_cmd_t *zc)
5730 dsl_dataset_t *ds, *ods;
5731 char origin[ZFS_MAX_DATASET_NAME_LEN];
5735 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
5736 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0 ||
5737 strchr(zc->zc_name, '%'))
5738 return (SET_ERROR(EINVAL));
5740 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5744 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
5746 dsl_pool_rele(dp, FTAG);
5750 if (!dsl_dir_is_clone(ds->ds_dir)) {
5751 dsl_dataset_rele(ds, FTAG);
5752 dsl_pool_rele(dp, FTAG);
5753 return (SET_ERROR(EINVAL));
5756 error = dsl_dataset_hold_obj(dp,
5757 dsl_dir_phys(ds->ds_dir)->dd_origin_obj, FTAG, &ods);
5759 dsl_dataset_rele(ds, FTAG);
5760 dsl_pool_rele(dp, FTAG);
5764 dsl_dataset_name(ods, origin);
5765 dsl_dataset_rele(ods, FTAG);
5766 dsl_dataset_rele(ds, FTAG);
5767 dsl_pool_rele(dp, FTAG);
5770 * We don't need to unmount *all* the origin fs's snapshots, but
5773 cp = strchr(origin, '@');
5776 (void) dmu_objset_find(origin,
5777 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
5778 return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
5782 * Retrieve a single {user|group|project}{used|quota}@... property.
5785 * zc_name name of filesystem
5786 * zc_objset_type zfs_userquota_prop_t
5787 * zc_value domain name (eg. "S-1-234-567-89")
5788 * zc_guid RID/UID/GID
5791 * zc_cookie property value
5794 zfs_ioc_userspace_one(zfs_cmd_t *zc)
5799 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
5800 return (SET_ERROR(EINVAL));
5802 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
5806 error = zfs_userspace_one(zfsvfs,
5807 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
5808 zfsvfs_rele(zfsvfs, FTAG);
5815 * zc_name name of filesystem
5816 * zc_cookie zap cursor
5817 * zc_objset_type zfs_userquota_prop_t
5818 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
5821 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
5822 * zc_cookie zap cursor
5825 zfs_ioc_userspace_many(zfs_cmd_t *zc)
5828 int bufsize = zc->zc_nvlist_dst_size;
5831 return (SET_ERROR(ENOMEM));
5833 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
5837 void *buf = vmem_alloc(bufsize, KM_SLEEP);
5839 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
5840 buf, &zc->zc_nvlist_dst_size);
5843 error = xcopyout(buf,
5844 (void *)(uintptr_t)zc->zc_nvlist_dst,
5845 zc->zc_nvlist_dst_size);
5847 vmem_free(buf, bufsize);
5848 zfsvfs_rele(zfsvfs, FTAG);
5855 * zc_name name of filesystem
5861 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
5867 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
5868 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
5870 * If userused is not enabled, it may be because the
5871 * objset needs to be closed & reopened (to grow the
5872 * objset_phys_t). Suspend/resume the fs will do that.
5874 dsl_dataset_t *ds, *newds;
5876 ds = dmu_objset_ds(zfsvfs->z_os);
5877 error = zfs_suspend_fs(zfsvfs);
5879 dmu_objset_refresh_ownership(ds, &newds,
5881 error = zfs_resume_fs(zfsvfs, newds);
5885 error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
5886 zfs_vfs_rele(zfsvfs);
5888 /* XXX kind of reading contents without owning */
5889 error = dmu_objset_hold_flags(zc->zc_name, B_TRUE, FTAG, &os);
5893 error = dmu_objset_userspace_upgrade(os);
5894 dmu_objset_rele_flags(os, B_TRUE, FTAG);
5902 * zc_name name of filesystem
5908 zfs_ioc_id_quota_upgrade(zfs_cmd_t *zc)
5913 error = dmu_objset_hold_flags(zc->zc_name, B_TRUE, FTAG, &os);
5917 if (dmu_objset_userobjspace_upgradable(os) ||
5918 dmu_objset_projectquota_upgradable(os)) {
5919 mutex_enter(&os->os_upgrade_lock);
5920 if (os->os_upgrade_id == 0) {
5921 /* clear potential error code and retry */
5922 os->os_upgrade_status = 0;
5923 mutex_exit(&os->os_upgrade_lock);
5925 dmu_objset_id_quota_upgrade(os);
5927 mutex_exit(&os->os_upgrade_lock);
5930 dsl_pool_rele(dmu_objset_pool(os), FTAG);
5932 taskq_wait_id(os->os_spa->spa_upgrade_taskq, os->os_upgrade_id);
5933 error = os->os_upgrade_status;
5935 dsl_pool_rele(dmu_objset_pool(os), FTAG);
5938 dsl_dataset_rele_flags(dmu_objset_ds(os), DS_HOLD_FLAG_DECRYPT, FTAG);
5944 zfs_ioc_share(zfs_cmd_t *zc)
5946 return (SET_ERROR(ENOSYS));
5949 ace_t full_access[] = {
5950 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
5955 * zc_name name of containing filesystem
5956 * zc_obj object # beyond which we want next in-use object #
5959 * zc_obj next in-use object #
5962 zfs_ioc_next_obj(zfs_cmd_t *zc)
5964 objset_t *os = NULL;
5967 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
5971 error = dmu_object_next(os, &zc->zc_obj, B_FALSE, 0);
5973 dmu_objset_rele(os, FTAG);
5979 * zc_name name of filesystem
5980 * zc_value prefix name for snapshot
5981 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
5984 * zc_value short name of new snapshot
5987 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
5994 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
5998 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
5999 (u_longlong_t)ddi_get_lbolt64());
6000 hold_name = kmem_asprintf("%%%s", zc->zc_value);
6002 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
6005 (void) strlcpy(zc->zc_value, snap_name,
6006 sizeof (zc->zc_value));
6007 kmem_strfree(snap_name);
6008 kmem_strfree(hold_name);
6009 zfs_onexit_fd_rele(zc->zc_cleanup_fd);
6015 * zc_name name of "to" snapshot
6016 * zc_value name of "from" snapshot
6017 * zc_cookie file descriptor to write diff data on
6020 * dmu_diff_record_t's to the file descriptor
6023 zfs_ioc_diff(zfs_cmd_t *zc)
6029 if ((error = zfs_file_get(zc->zc_cookie, &fp)))
6032 off = zfs_file_off(fp);
6033 error = dmu_diff(zc->zc_name, zc->zc_value, fp, &off);
6035 zfs_file_put(zc->zc_cookie);
6041 zfs_ioc_smb_acl(zfs_cmd_t *zc)
6043 return (SET_ERROR(ENOTSUP));
6048 * "holds" -> { snapname -> holdname (string), ... }
6049 * (optional) "cleanup_fd" -> fd (int32)
6053 * snapname -> error value (int32)
6057 static const zfs_ioc_key_t zfs_keys_hold[] = {
6058 {"holds", DATA_TYPE_NVLIST, 0},
6059 {"cleanup_fd", DATA_TYPE_INT32, ZK_OPTIONAL},
6064 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
6068 int cleanup_fd = -1;
6072 holds = fnvlist_lookup_nvlist(args, "holds");
6074 /* make sure the user didn't pass us any invalid (empty) tags */
6075 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
6076 pair = nvlist_next_nvpair(holds, pair)) {
6079 error = nvpair_value_string(pair, &htag);
6081 return (SET_ERROR(error));
6083 if (strlen(htag) == 0)
6084 return (SET_ERROR(EINVAL));
6087 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
6088 error = zfs_onexit_fd_hold(cleanup_fd, &minor);
6090 return (SET_ERROR(error));
6093 error = dsl_dataset_user_hold(holds, minor, errlist);
6095 zfs_onexit_fd_rele(cleanup_fd);
6096 return (SET_ERROR(error));
6100 * innvl is not used.
6103 * holdname -> time added (uint64 seconds since epoch)
6107 static const zfs_ioc_key_t zfs_keys_get_holds[] = {
6113 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
6115 return (dsl_dataset_get_holds(snapname, outnvl));
6120 * snapname -> { holdname, ... }
6125 * snapname -> error value (int32)
6129 static const zfs_ioc_key_t zfs_keys_release[] = {
6130 {"<snapname>...", DATA_TYPE_NVLIST, ZK_WILDCARDLIST},
6135 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
6137 return (dsl_dataset_user_release(holds, errlist));
6142 * zc_guid flags (ZEVENT_NONBLOCK)
6143 * zc_cleanup_fd zevent file descriptor
6146 * zc_nvlist_dst next nvlist event
6147 * zc_cookie dropped events since last get
6150 zfs_ioc_events_next(zfs_cmd_t *zc)
6153 nvlist_t *event = NULL;
6155 uint64_t dropped = 0;
6158 error = zfs_zevent_fd_hold(zc->zc_cleanup_fd, &minor, &ze);
6163 error = zfs_zevent_next(ze, &event,
6164 &zc->zc_nvlist_dst_size, &dropped);
6165 if (event != NULL) {
6166 zc->zc_cookie = dropped;
6167 error = put_nvlist(zc, event);
6171 if (zc->zc_guid & ZEVENT_NONBLOCK)
6174 if ((error == 0) || (error != ENOENT))
6177 error = zfs_zevent_wait(ze);
6182 zfs_zevent_fd_rele(zc->zc_cleanup_fd);
6189 * zc_cookie cleared events count
6192 zfs_ioc_events_clear(zfs_cmd_t *zc)
6196 zfs_zevent_drain_all(&count);
6197 zc->zc_cookie = count;
6204 * zc_guid eid | ZEVENT_SEEK_START | ZEVENT_SEEK_END
6205 * zc_cleanup zevent file descriptor
6208 zfs_ioc_events_seek(zfs_cmd_t *zc)
6214 error = zfs_zevent_fd_hold(zc->zc_cleanup_fd, &minor, &ze);
6218 error = zfs_zevent_seek(ze, zc->zc_guid);
6219 zfs_zevent_fd_rele(zc->zc_cleanup_fd);
6226 * zc_name name of later filesystem or snapshot
6227 * zc_value full name of old snapshot or bookmark
6230 * zc_cookie space in bytes
6231 * zc_objset_type compressed space in bytes
6232 * zc_perm_action uncompressed space in bytes
6235 zfs_ioc_space_written(zfs_cmd_t *zc)
6241 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
6244 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
6246 dsl_pool_rele(dp, FTAG);
6249 if (strchr(zc->zc_value, '#') != NULL) {
6250 zfs_bookmark_phys_t bmp;
6251 error = dsl_bookmark_lookup(dp, zc->zc_value,
6254 error = dsl_dataset_space_written_bookmark(&bmp, new,
6256 &zc->zc_objset_type, &zc->zc_perm_action);
6260 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
6263 error = dsl_dataset_space_written(old, new,
6265 &zc->zc_objset_type, &zc->zc_perm_action);
6266 dsl_dataset_rele(old, FTAG);
6269 dsl_dataset_rele(new, FTAG);
6270 dsl_pool_rele(dp, FTAG);
6276 * "firstsnap" -> snapshot name
6280 * "used" -> space in bytes
6281 * "compressed" -> compressed space in bytes
6282 * "uncompressed" -> uncompressed space in bytes
6285 static const zfs_ioc_key_t zfs_keys_space_snaps[] = {
6286 {"firstsnap", DATA_TYPE_STRING, 0},
6290 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
6294 dsl_dataset_t *new, *old;
6296 uint64_t used, comp, uncomp;
6298 firstsnap = fnvlist_lookup_string(innvl, "firstsnap");
6300 error = dsl_pool_hold(lastsnap, FTAG, &dp);
6304 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
6305 if (error == 0 && !new->ds_is_snapshot) {
6306 dsl_dataset_rele(new, FTAG);
6307 error = SET_ERROR(EINVAL);
6310 dsl_pool_rele(dp, FTAG);
6313 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
6314 if (error == 0 && !old->ds_is_snapshot) {
6315 dsl_dataset_rele(old, FTAG);
6316 error = SET_ERROR(EINVAL);
6319 dsl_dataset_rele(new, FTAG);
6320 dsl_pool_rele(dp, FTAG);
6324 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
6325 dsl_dataset_rele(old, FTAG);
6326 dsl_dataset_rele(new, FTAG);
6327 dsl_pool_rele(dp, FTAG);
6328 fnvlist_add_uint64(outnvl, "used", used);
6329 fnvlist_add_uint64(outnvl, "compressed", comp);
6330 fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
6336 * "fd" -> file descriptor to write stream to (int32)
6337 * (optional) "fromsnap" -> full snap name to send an incremental from
6338 * (optional) "largeblockok" -> (value ignored)
6339 * indicates that blocks > 128KB are permitted
6340 * (optional) "embedok" -> (value ignored)
6341 * presence indicates DRR_WRITE_EMBEDDED records are permitted
6342 * (optional) "compressok" -> (value ignored)
6343 * presence indicates compressed DRR_WRITE records are permitted
6344 * (optional) "rawok" -> (value ignored)
6345 * presence indicates raw encrypted records should be used.
6346 * (optional) "savedok" -> (value ignored)
6347 * presence indicates we should send a partially received snapshot
6348 * (optional) "resume_object" and "resume_offset" -> (uint64)
6349 * if present, resume send stream from specified object and offset.
6350 * (optional) "redactbook" -> (string)
6351 * if present, use this bookmark's redaction list to generate a redacted
6357 static const zfs_ioc_key_t zfs_keys_send_new[] = {
6358 {"fd", DATA_TYPE_INT32, 0},
6359 {"fromsnap", DATA_TYPE_STRING, ZK_OPTIONAL},
6360 {"largeblockok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
6361 {"embedok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
6362 {"compressok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
6363 {"rawok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
6364 {"savedok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
6365 {"resume_object", DATA_TYPE_UINT64, ZK_OPTIONAL},
6366 {"resume_offset", DATA_TYPE_UINT64, ZK_OPTIONAL},
6367 {"redactbook", DATA_TYPE_STRING, ZK_OPTIONAL},
6372 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
6376 char *fromname = NULL;
6379 boolean_t largeblockok;
6381 boolean_t compressok;
6384 uint64_t resumeobj = 0;
6385 uint64_t resumeoff = 0;
6386 char *redactbook = NULL;
6388 fd = fnvlist_lookup_int32(innvl, "fd");
6390 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
6392 largeblockok = nvlist_exists(innvl, "largeblockok");
6393 embedok = nvlist_exists(innvl, "embedok");
6394 compressok = nvlist_exists(innvl, "compressok");
6395 rawok = nvlist_exists(innvl, "rawok");
6396 savedok = nvlist_exists(innvl, "savedok");
6398 (void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj);
6399 (void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff);
6401 (void) nvlist_lookup_string(innvl, "redactbook", &redactbook);
6403 if ((error = zfs_file_get(fd, &fp)))
6406 off = zfs_file_off(fp);
6408 dmu_send_outparams_t out = {0};
6409 out.dso_outfunc = dump_bytes;
6411 out.dso_dryrun = B_FALSE;
6412 error = dmu_send(snapname, fromname, embedok, largeblockok,
6413 compressok, rawok, savedok, resumeobj, resumeoff,
6414 redactbook, fd, &off, &out);
6422 send_space_sum(objset_t *os, void *buf, int len, void *arg)
6424 uint64_t *size = arg;
6430 * Determine approximately how large a zfs send stream will be -- the number
6431 * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
6434 * (optional) "from" -> full snap or bookmark name to send an incremental
6436 * (optional) "largeblockok" -> (value ignored)
6437 * indicates that blocks > 128KB are permitted
6438 * (optional) "embedok" -> (value ignored)
6439 * presence indicates DRR_WRITE_EMBEDDED records are permitted
6440 * (optional) "compressok" -> (value ignored)
6441 * presence indicates compressed DRR_WRITE records are permitted
6442 * (optional) "rawok" -> (value ignored)
6443 * presence indicates raw encrypted records should be used.
6444 * (optional) "fd" -> file descriptor to use as a cookie for progress
6449 * "space" -> bytes of space (uint64)
6452 static const zfs_ioc_key_t zfs_keys_send_space[] = {
6453 {"from", DATA_TYPE_STRING, ZK_OPTIONAL},
6454 {"fromsnap", DATA_TYPE_STRING, ZK_OPTIONAL},
6455 {"largeblockok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
6456 {"embedok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
6457 {"compressok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
6458 {"rawok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
6459 {"fd", DATA_TYPE_INT32, ZK_OPTIONAL},
6460 {"redactbook", DATA_TYPE_STRING, ZK_OPTIONAL},
6461 {"resumeobj", DATA_TYPE_UINT64, ZK_OPTIONAL},
6462 {"resumeoff", DATA_TYPE_UINT64, ZK_OPTIONAL},
6463 {"bytes", DATA_TYPE_UINT64, ZK_OPTIONAL},
6467 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
6470 dsl_dataset_t *tosnap;
6471 dsl_dataset_t *fromsnap = NULL;
6473 char *fromname = NULL;
6474 char *redactlist_book = NULL;
6475 boolean_t largeblockok;
6477 boolean_t compressok;
6481 boolean_t full_estimate = B_FALSE;
6482 uint64_t resumeobj = 0;
6483 uint64_t resumeoff = 0;
6484 uint64_t resume_bytes = 0;
6486 zfs_bookmark_phys_t zbm = {0};
6488 error = dsl_pool_hold(snapname, FTAG, &dp);
6492 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
6494 dsl_pool_rele(dp, FTAG);
6497 (void) nvlist_lookup_int32(innvl, "fd", &fd);
6499 largeblockok = nvlist_exists(innvl, "largeblockok");
6500 embedok = nvlist_exists(innvl, "embedok");
6501 compressok = nvlist_exists(innvl, "compressok");
6502 rawok = nvlist_exists(innvl, "rawok");
6503 savedok = nvlist_exists(innvl, "savedok");
6504 boolean_t from = (nvlist_lookup_string(innvl, "from", &fromname) == 0);
6505 boolean_t altbook = (nvlist_lookup_string(innvl, "redactbook",
6506 &redactlist_book) == 0);
6508 (void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj);
6509 (void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff);
6510 (void) nvlist_lookup_uint64(innvl, "bytes", &resume_bytes);
6513 full_estimate = B_TRUE;
6515 if (strchr(fromname, '#')) {
6516 error = dsl_bookmark_lookup(dp, fromname, tosnap, &zbm);
6519 * dsl_bookmark_lookup() will fail with EXDEV if
6520 * the from-bookmark and tosnap are at the same txg.
6521 * However, it's valid to do a send (and therefore,
6522 * a send estimate) from and to the same time point,
6523 * if the bookmark is redacted (the incremental send
6524 * can change what's redacted on the target). In
6525 * this case, dsl_bookmark_lookup() fills in zbm
6526 * but returns EXDEV. Ignore this error.
6528 if (error == EXDEV && zbm.zbm_redaction_obj != 0 &&
6530 dsl_dataset_phys(tosnap)->ds_guid)
6534 dsl_dataset_rele(tosnap, FTAG);
6535 dsl_pool_rele(dp, FTAG);
6538 if (zbm.zbm_redaction_obj != 0 || !(zbm.zbm_flags &
6539 ZBM_FLAG_HAS_FBN)) {
6540 full_estimate = B_TRUE;
6542 } else if (strchr(fromname, '@')) {
6543 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
6545 dsl_dataset_rele(tosnap, FTAG);
6546 dsl_pool_rele(dp, FTAG);
6550 if (!dsl_dataset_is_before(tosnap, fromsnap, 0)) {
6551 full_estimate = B_TRUE;
6552 dsl_dataset_rele(fromsnap, FTAG);
6556 * from is not properly formatted as a snapshot or
6559 dsl_dataset_rele(tosnap, FTAG);
6560 dsl_pool_rele(dp, FTAG);
6561 return (SET_ERROR(EINVAL));
6565 if (full_estimate) {
6566 dmu_send_outparams_t out = {0};
6568 out.dso_outfunc = send_space_sum;
6569 out.dso_arg = &space;
6570 out.dso_dryrun = B_TRUE;
6572 * We have to release these holds so dmu_send can take them. It
6573 * will do all the error checking we need.
6575 dsl_dataset_rele(tosnap, FTAG);
6576 dsl_pool_rele(dp, FTAG);
6577 error = dmu_send(snapname, fromname, embedok, largeblockok,
6578 compressok, rawok, savedok, resumeobj, resumeoff,
6579 redactlist_book, fd, &off, &out);
6581 error = dmu_send_estimate_fast(tosnap, fromsnap,
6582 (from && strchr(fromname, '#') != NULL ? &zbm : NULL),
6583 compressok || rawok, savedok, &space);
6584 space -= resume_bytes;
6585 if (fromsnap != NULL)
6586 dsl_dataset_rele(fromsnap, FTAG);
6587 dsl_dataset_rele(tosnap, FTAG);
6588 dsl_pool_rele(dp, FTAG);
6591 fnvlist_add_uint64(outnvl, "space", space);
6597 * Sync the currently open TXG to disk for the specified pool.
6598 * This is somewhat similar to 'zfs_sync()'.
6599 * For cases that do not result in error this ioctl will wait for
6600 * the currently open TXG to commit before returning back to the caller.
6603 * "force" -> when true, force uberblock update even if there is no dirty data.
6604 * In addition this will cause the vdev configuration to be written
6605 * out including updating the zpool cache file. (boolean_t)
6610 static const zfs_ioc_key_t zfs_keys_pool_sync[] = {
6611 {"force", DATA_TYPE_BOOLEAN_VALUE, 0},
6616 zfs_ioc_pool_sync(const char *pool, nvlist_t *innvl, nvlist_t *onvl)
6619 boolean_t force = B_FALSE;
6622 if ((err = spa_open(pool, &spa, FTAG)) != 0)
6626 force = fnvlist_lookup_boolean_value(innvl, "force");
6629 spa_config_enter(spa, SCL_CONFIG, FTAG, RW_WRITER);
6630 vdev_config_dirty(spa->spa_root_vdev);
6631 spa_config_exit(spa, SCL_CONFIG, FTAG);
6633 txg_wait_synced(spa_get_dsl(spa), 0);
6635 spa_close(spa, FTAG);
6641 * Load a user's wrapping key into the kernel.
6643 * "hidden_args" -> { "wkeydata" -> value }
6644 * raw uint8_t array of encryption wrapping key data (32 bytes)
6645 * (optional) "noop" -> (value ignored)
6646 * presence indicated key should only be verified, not loaded
6649 static const zfs_ioc_key_t zfs_keys_load_key[] = {
6650 {"hidden_args", DATA_TYPE_NVLIST, 0},
6651 {"noop", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
6656 zfs_ioc_load_key(const char *dsname, nvlist_t *innvl, nvlist_t *outnvl)
6659 dsl_crypto_params_t *dcp = NULL;
6660 nvlist_t *hidden_args;
6661 boolean_t noop = nvlist_exists(innvl, "noop");
6663 if (strchr(dsname, '@') != NULL || strchr(dsname, '%') != NULL) {
6664 ret = SET_ERROR(EINVAL);
6668 hidden_args = fnvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS);
6670 ret = dsl_crypto_params_create_nvlist(DCP_CMD_NONE, NULL,
6675 ret = spa_keystore_load_wkey(dsname, dcp, noop);
6679 dsl_crypto_params_free(dcp, noop);
6684 dsl_crypto_params_free(dcp, B_TRUE);
6689 * Unload a user's wrapping key from the kernel.
6690 * Both innvl and outnvl are unused.
6692 static const zfs_ioc_key_t zfs_keys_unload_key[] = {
6698 zfs_ioc_unload_key(const char *dsname, nvlist_t *innvl, nvlist_t *outnvl)
6702 if (strchr(dsname, '@') != NULL || strchr(dsname, '%') != NULL) {
6703 ret = (SET_ERROR(EINVAL));
6707 ret = spa_keystore_unload_wkey(dsname);
6716 * Changes a user's wrapping key used to decrypt a dataset. The keyformat,
6717 * keylocation, pbkdf2salt, and pbkdf2iters properties can also be specified
6718 * here to change how the key is derived in userspace.
6721 * "hidden_args" (optional) -> { "wkeydata" -> value }
6722 * raw uint8_t array of new encryption wrapping key data (32 bytes)
6723 * "props" (optional) -> { prop -> value }
6728 static const zfs_ioc_key_t zfs_keys_change_key[] = {
6729 {"crypt_cmd", DATA_TYPE_UINT64, ZK_OPTIONAL},
6730 {"hidden_args", DATA_TYPE_NVLIST, ZK_OPTIONAL},
6731 {"props", DATA_TYPE_NVLIST, ZK_OPTIONAL},
6736 zfs_ioc_change_key(const char *dsname, nvlist_t *innvl, nvlist_t *outnvl)
6739 uint64_t cmd = DCP_CMD_NONE;
6740 dsl_crypto_params_t *dcp = NULL;
6741 nvlist_t *args = NULL, *hidden_args = NULL;
6743 if (strchr(dsname, '@') != NULL || strchr(dsname, '%') != NULL) {
6744 ret = (SET_ERROR(EINVAL));
6748 (void) nvlist_lookup_uint64(innvl, "crypt_cmd", &cmd);
6749 (void) nvlist_lookup_nvlist(innvl, "props", &args);
6750 (void) nvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS, &hidden_args);
6752 ret = dsl_crypto_params_create_nvlist(cmd, args, hidden_args, &dcp);
6756 ret = spa_keystore_change_key(dsname, dcp);
6760 dsl_crypto_params_free(dcp, B_FALSE);
6765 dsl_crypto_params_free(dcp, B_TRUE);
6769 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
6772 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
6773 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
6774 boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
6776 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
6778 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
6779 ASSERT3U(ioc, <, ZFS_IOC_LAST);
6780 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
6781 ASSERT3P(vec->zvec_func, ==, NULL);
6783 vec->zvec_legacy_func = func;
6784 vec->zvec_secpolicy = secpolicy;
6785 vec->zvec_namecheck = namecheck;
6786 vec->zvec_allow_log = log_history;
6787 vec->zvec_pool_check = pool_check;
6791 * See the block comment at the beginning of this file for details on
6792 * each argument to this function.
6795 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
6796 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
6797 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
6798 boolean_t allow_log, const zfs_ioc_key_t *nvl_keys, size_t num_keys)
6800 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
6802 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
6803 ASSERT3U(ioc, <, ZFS_IOC_LAST);
6804 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
6805 ASSERT3P(vec->zvec_func, ==, NULL);
6807 /* if we are logging, the name must be valid */
6808 ASSERT(!allow_log || namecheck != NO_NAME);
6810 vec->zvec_name = name;
6811 vec->zvec_func = func;
6812 vec->zvec_secpolicy = secpolicy;
6813 vec->zvec_namecheck = namecheck;
6814 vec->zvec_pool_check = pool_check;
6815 vec->zvec_smush_outnvlist = smush_outnvlist;
6816 vec->zvec_allow_log = allow_log;
6817 vec->zvec_nvl_keys = nvl_keys;
6818 vec->zvec_nvl_key_count = num_keys;
6822 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
6823 zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
6824 zfs_ioc_poolcheck_t pool_check)
6826 zfs_ioctl_register_legacy(ioc, func, secpolicy,
6827 POOL_NAME, log_history, pool_check);
6831 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
6832 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
6834 zfs_ioctl_register_legacy(ioc, func, secpolicy,
6835 DATASET_NAME, B_FALSE, pool_check);
6839 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
6841 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
6842 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
6846 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
6847 zfs_secpolicy_func_t *secpolicy)
6849 zfs_ioctl_register_legacy(ioc, func, secpolicy,
6850 NO_NAME, B_FALSE, POOL_CHECK_NONE);
6854 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
6855 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
6857 zfs_ioctl_register_legacy(ioc, func, secpolicy,
6858 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
6862 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
6864 zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
6865 zfs_secpolicy_read);
6869 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
6870 zfs_secpolicy_func_t *secpolicy)
6872 zfs_ioctl_register_legacy(ioc, func, secpolicy,
6873 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
6877 zfs_ioctl_init(void)
6879 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
6880 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
6881 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
6882 zfs_keys_snapshot, ARRAY_SIZE(zfs_keys_snapshot));
6884 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
6885 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
6886 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE,
6887 zfs_keys_log_history, ARRAY_SIZE(zfs_keys_log_history));
6889 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
6890 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
6891 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
6892 zfs_keys_space_snaps, ARRAY_SIZE(zfs_keys_space_snaps));
6894 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
6895 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
6896 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
6897 zfs_keys_send_new, ARRAY_SIZE(zfs_keys_send_new));
6899 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
6900 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
6901 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
6902 zfs_keys_send_space, ARRAY_SIZE(zfs_keys_send_space));
6904 zfs_ioctl_register("create", ZFS_IOC_CREATE,
6905 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
6906 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
6907 zfs_keys_create, ARRAY_SIZE(zfs_keys_create));
6909 zfs_ioctl_register("clone", ZFS_IOC_CLONE,
6910 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
6911 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
6912 zfs_keys_clone, ARRAY_SIZE(zfs_keys_clone));
6914 zfs_ioctl_register("remap", ZFS_IOC_REMAP,
6915 zfs_ioc_remap, zfs_secpolicy_none, DATASET_NAME,
6916 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE,
6917 zfs_keys_remap, ARRAY_SIZE(zfs_keys_remap));
6919 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
6920 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
6921 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
6922 zfs_keys_destroy_snaps, ARRAY_SIZE(zfs_keys_destroy_snaps));
6924 zfs_ioctl_register("hold", ZFS_IOC_HOLD,
6925 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
6926 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
6927 zfs_keys_hold, ARRAY_SIZE(zfs_keys_hold));
6928 zfs_ioctl_register("release", ZFS_IOC_RELEASE,
6929 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
6930 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
6931 zfs_keys_release, ARRAY_SIZE(zfs_keys_release));
6933 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
6934 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
6935 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
6936 zfs_keys_get_holds, ARRAY_SIZE(zfs_keys_get_holds));
6938 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
6939 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
6940 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE,
6941 zfs_keys_rollback, ARRAY_SIZE(zfs_keys_rollback));
6943 zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK,
6944 zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME,
6945 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
6946 zfs_keys_bookmark, ARRAY_SIZE(zfs_keys_bookmark));
6948 zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS,
6949 zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME,
6950 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
6951 zfs_keys_get_bookmarks, ARRAY_SIZE(zfs_keys_get_bookmarks));
6953 zfs_ioctl_register("get_bookmark_props", ZFS_IOC_GET_BOOKMARK_PROPS,
6954 zfs_ioc_get_bookmark_props, zfs_secpolicy_read, ENTITY_NAME,
6955 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE, zfs_keys_get_bookmark_props,
6956 ARRAY_SIZE(zfs_keys_get_bookmark_props));
6958 zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS,
6959 zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks,
6961 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
6962 zfs_keys_destroy_bookmarks,
6963 ARRAY_SIZE(zfs_keys_destroy_bookmarks));
6965 zfs_ioctl_register("receive", ZFS_IOC_RECV_NEW,
6966 zfs_ioc_recv_new, zfs_secpolicy_recv_new, DATASET_NAME,
6967 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
6968 zfs_keys_recv_new, ARRAY_SIZE(zfs_keys_recv_new));
6969 zfs_ioctl_register("load-key", ZFS_IOC_LOAD_KEY,
6970 zfs_ioc_load_key, zfs_secpolicy_load_key,
6971 DATASET_NAME, POOL_CHECK_SUSPENDED, B_TRUE, B_TRUE,
6972 zfs_keys_load_key, ARRAY_SIZE(zfs_keys_load_key));
6973 zfs_ioctl_register("unload-key", ZFS_IOC_UNLOAD_KEY,
6974 zfs_ioc_unload_key, zfs_secpolicy_load_key,
6975 DATASET_NAME, POOL_CHECK_SUSPENDED, B_TRUE, B_TRUE,
6976 zfs_keys_unload_key, ARRAY_SIZE(zfs_keys_unload_key));
6977 zfs_ioctl_register("change-key", ZFS_IOC_CHANGE_KEY,
6978 zfs_ioc_change_key, zfs_secpolicy_change_key,
6979 DATASET_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY,
6980 B_TRUE, B_TRUE, zfs_keys_change_key,
6981 ARRAY_SIZE(zfs_keys_change_key));
6983 zfs_ioctl_register("sync", ZFS_IOC_POOL_SYNC,
6984 zfs_ioc_pool_sync, zfs_secpolicy_none, POOL_NAME,
6985 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE,
6986 zfs_keys_pool_sync, ARRAY_SIZE(zfs_keys_pool_sync));
6987 zfs_ioctl_register("reopen", ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
6988 zfs_secpolicy_config, POOL_NAME, POOL_CHECK_SUSPENDED, B_TRUE,
6989 B_TRUE, zfs_keys_pool_reopen, ARRAY_SIZE(zfs_keys_pool_reopen));
6991 zfs_ioctl_register("channel_program", ZFS_IOC_CHANNEL_PROGRAM,
6992 zfs_ioc_channel_program, zfs_secpolicy_config,
6993 POOL_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE,
6994 B_TRUE, zfs_keys_channel_program,
6995 ARRAY_SIZE(zfs_keys_channel_program));
6997 zfs_ioctl_register("redact", ZFS_IOC_REDACT,
6998 zfs_ioc_redact, zfs_secpolicy_config, DATASET_NAME,
6999 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7000 zfs_keys_redact, ARRAY_SIZE(zfs_keys_redact));
7002 zfs_ioctl_register("zpool_checkpoint", ZFS_IOC_POOL_CHECKPOINT,
7003 zfs_ioc_pool_checkpoint, zfs_secpolicy_config, POOL_NAME,
7004 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7005 zfs_keys_pool_checkpoint, ARRAY_SIZE(zfs_keys_pool_checkpoint));
7007 zfs_ioctl_register("zpool_discard_checkpoint",
7008 ZFS_IOC_POOL_DISCARD_CHECKPOINT, zfs_ioc_pool_discard_checkpoint,
7009 zfs_secpolicy_config, POOL_NAME,
7010 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7011 zfs_keys_pool_discard_checkpoint,
7012 ARRAY_SIZE(zfs_keys_pool_discard_checkpoint));
7014 zfs_ioctl_register("initialize", ZFS_IOC_POOL_INITIALIZE,
7015 zfs_ioc_pool_initialize, zfs_secpolicy_config, POOL_NAME,
7016 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7017 zfs_keys_pool_initialize, ARRAY_SIZE(zfs_keys_pool_initialize));
7019 zfs_ioctl_register("trim", ZFS_IOC_POOL_TRIM,
7020 zfs_ioc_pool_trim, zfs_secpolicy_config, POOL_NAME,
7021 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7022 zfs_keys_pool_trim, ARRAY_SIZE(zfs_keys_pool_trim));
7024 zfs_ioctl_register("wait", ZFS_IOC_WAIT,
7025 zfs_ioc_wait, zfs_secpolicy_none, POOL_NAME,
7026 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE,
7027 zfs_keys_pool_wait, ARRAY_SIZE(zfs_keys_pool_wait));
7029 zfs_ioctl_register("wait_fs", ZFS_IOC_WAIT_FS,
7030 zfs_ioc_wait_fs, zfs_secpolicy_none, DATASET_NAME,
7031 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE,
7032 zfs_keys_fs_wait, ARRAY_SIZE(zfs_keys_fs_wait));
7034 zfs_ioctl_register("set_bootenv", ZFS_IOC_SET_BOOTENV,
7035 zfs_ioc_set_bootenv, zfs_secpolicy_config, POOL_NAME,
7036 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE,
7037 zfs_keys_set_bootenv, ARRAY_SIZE(zfs_keys_set_bootenv));
7039 zfs_ioctl_register("get_bootenv", ZFS_IOC_GET_BOOTENV,
7040 zfs_ioc_get_bootenv, zfs_secpolicy_none, POOL_NAME,
7041 POOL_CHECK_SUSPENDED, B_FALSE, B_TRUE,
7042 zfs_keys_get_bootenv, ARRAY_SIZE(zfs_keys_get_bootenv));
7044 /* IOCTLS that use the legacy function signature */
7046 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
7047 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
7049 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
7050 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
7051 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
7053 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
7054 zfs_ioc_pool_upgrade);
7055 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
7057 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
7058 zfs_ioc_vdev_remove);
7059 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
7060 zfs_ioc_vdev_set_state);
7061 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
7062 zfs_ioc_vdev_attach);
7063 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
7064 zfs_ioc_vdev_detach);
7065 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
7066 zfs_ioc_vdev_setpath);
7067 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
7068 zfs_ioc_vdev_setfru);
7069 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
7070 zfs_ioc_pool_set_props);
7071 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
7072 zfs_ioc_vdev_split);
7073 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
7074 zfs_ioc_pool_reguid);
7076 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
7077 zfs_ioc_pool_configs, zfs_secpolicy_none);
7078 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
7079 zfs_ioc_pool_tryimport, zfs_secpolicy_config);
7080 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
7081 zfs_ioc_inject_fault, zfs_secpolicy_inject);
7082 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
7083 zfs_ioc_clear_fault, zfs_secpolicy_inject);
7084 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
7085 zfs_ioc_inject_list_next, zfs_secpolicy_inject);
7088 * pool destroy, and export don't log the history as part of
7089 * zfsdev_ioctl, but rather zfs_ioc_pool_export
7090 * does the logging of those commands.
7092 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
7093 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
7094 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
7095 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
7097 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
7098 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
7099 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
7100 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
7102 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
7103 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED);
7104 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
7105 zfs_ioc_dsobj_to_dsname,
7106 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED);
7107 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
7108 zfs_ioc_pool_get_history,
7109 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
7111 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
7112 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
7114 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
7115 zfs_secpolicy_config, B_TRUE, POOL_CHECK_READONLY);
7117 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
7118 zfs_ioc_space_written);
7119 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
7120 zfs_ioc_objset_recvd_props);
7121 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
7123 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
7125 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
7126 zfs_ioc_objset_stats);
7127 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
7128 zfs_ioc_objset_zplprops);
7129 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
7130 zfs_ioc_dataset_list_next);
7131 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
7132 zfs_ioc_snapshot_list_next);
7133 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
7134 zfs_ioc_send_progress);
7136 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
7137 zfs_ioc_diff, zfs_secpolicy_diff);
7138 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
7139 zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
7140 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
7141 zfs_ioc_obj_to_path, zfs_secpolicy_diff);
7142 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
7143 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
7144 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
7145 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
7146 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
7147 zfs_ioc_send, zfs_secpolicy_send);
7149 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
7150 zfs_secpolicy_none);
7151 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
7152 zfs_secpolicy_destroy);
7153 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
7154 zfs_secpolicy_rename);
7155 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
7156 zfs_secpolicy_recv);
7157 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
7158 zfs_secpolicy_promote);
7159 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
7160 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
7161 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
7162 zfs_secpolicy_set_fsacl);
7164 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
7165 zfs_secpolicy_share, POOL_CHECK_NONE);
7166 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
7167 zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
7168 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
7169 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
7170 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
7171 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
7172 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
7173 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
7175 zfs_ioctl_register_legacy(ZFS_IOC_EVENTS_NEXT, zfs_ioc_events_next,
7176 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);
7177 zfs_ioctl_register_legacy(ZFS_IOC_EVENTS_CLEAR, zfs_ioc_events_clear,
7178 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);
7179 zfs_ioctl_register_legacy(ZFS_IOC_EVENTS_SEEK, zfs_ioc_events_seek,
7180 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);
7182 zfs_ioctl_init_os();
7186 * Verify that for non-legacy ioctls the input nvlist
7187 * pairs match against the expected input.
7189 * Possible errors are:
7190 * ZFS_ERR_IOC_ARG_UNAVAIL An unrecognized nvpair was encountered
7191 * ZFS_ERR_IOC_ARG_REQUIRED A required nvpair is missing
7192 * ZFS_ERR_IOC_ARG_BADTYPE Invalid type for nvpair
7195 zfs_check_input_nvpairs(nvlist_t *innvl, const zfs_ioc_vec_t *vec)
7197 const zfs_ioc_key_t *nvl_keys = vec->zvec_nvl_keys;
7198 boolean_t required_keys_found = B_FALSE;
7201 * examine each input pair
7203 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
7204 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
7205 char *name = nvpair_name(pair);
7206 data_type_t type = nvpair_type(pair);
7207 boolean_t identified = B_FALSE;
7210 * check pair against the documented names and type
7212 for (int k = 0; k < vec->zvec_nvl_key_count; k++) {
7213 /* if not a wild card name, check for an exact match */
7214 if ((nvl_keys[k].zkey_flags & ZK_WILDCARDLIST) == 0 &&
7215 strcmp(nvl_keys[k].zkey_name, name) != 0)
7218 identified = B_TRUE;
7220 if (nvl_keys[k].zkey_type != DATA_TYPE_ANY &&
7221 nvl_keys[k].zkey_type != type) {
7222 return (SET_ERROR(ZFS_ERR_IOC_ARG_BADTYPE));
7225 if (nvl_keys[k].zkey_flags & ZK_OPTIONAL)
7228 required_keys_found = B_TRUE;
7232 /* allow an 'optional' key, everything else is invalid */
7234 (strcmp(name, "optional") != 0 ||
7235 type != DATA_TYPE_NVLIST)) {
7236 return (SET_ERROR(ZFS_ERR_IOC_ARG_UNAVAIL));
7240 /* verify that all required keys were found */
7241 for (int k = 0; k < vec->zvec_nvl_key_count; k++) {
7242 if (nvl_keys[k].zkey_flags & ZK_OPTIONAL)
7245 if (nvl_keys[k].zkey_flags & ZK_WILDCARDLIST) {
7246 /* at least one non-optional key is expected here */
7247 if (!required_keys_found)
7248 return (SET_ERROR(ZFS_ERR_IOC_ARG_REQUIRED));
7252 if (!nvlist_exists(innvl, nvl_keys[k].zkey_name))
7253 return (SET_ERROR(ZFS_ERR_IOC_ARG_REQUIRED));
7260 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
7261 zfs_ioc_poolcheck_t check)
7266 ASSERT(type == POOL_NAME || type == DATASET_NAME ||
7267 type == ENTITY_NAME);
7269 if (check & POOL_CHECK_NONE)
7272 error = spa_open(name, &spa, FTAG);
7274 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
7275 error = SET_ERROR(EAGAIN);
7276 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
7277 error = SET_ERROR(EROFS);
7278 spa_close(spa, FTAG);
7284 zfsdev_getminor(int fd, minor_t *minorp)
7286 zfsdev_state_t *zs, *fpd;
7290 ASSERT(!MUTEX_HELD(&zfsdev_state_lock));
7292 if ((rc = zfs_file_get(fd, &fp)))
7295 fpd = zfs_file_private(fp);
7297 return (SET_ERROR(EBADF));
7299 mutex_enter(&zfsdev_state_lock);
7301 for (zs = zfsdev_state_list; zs != NULL; zs = zs->zs_next) {
7303 if (zs->zs_minor == -1)
7307 *minorp = fpd->zs_minor;
7308 mutex_exit(&zfsdev_state_lock);
7313 mutex_exit(&zfsdev_state_lock);
7315 return (SET_ERROR(EBADF));
7319 zfsdev_get_state_impl(minor_t minor, enum zfsdev_state_type which)
7323 for (zs = zfsdev_state_list; zs != NULL; zs = zs->zs_next) {
7324 if (zs->zs_minor == minor) {
7328 return (zs->zs_onexit);
7330 return (zs->zs_zevent);
7341 zfsdev_get_state(minor_t minor, enum zfsdev_state_type which)
7345 ptr = zfsdev_get_state_impl(minor, which);
7351 * Find a free minor number. The zfsdev_state_list is expected to
7352 * be short since it is only a list of currently open file handles.
7355 zfsdev_minor_alloc(void)
7357 static minor_t last_minor = 0;
7360 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
7362 for (m = last_minor + 1; m != last_minor; m++) {
7363 if (m > ZFSDEV_MAX_MINOR)
7365 if (zfsdev_get_state_impl(m, ZST_ALL) == NULL) {
7375 zfsdev_ioctl_common(uint_t vecnum, zfs_cmd_t *zc, int flag)
7378 const zfs_ioc_vec_t *vec;
7379 char *saved_poolname = NULL;
7380 size_t saved_poolname_len = 0;
7381 nvlist_t *innvl = NULL;
7382 fstrans_cookie_t cookie;
7386 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
7387 return (SET_ERROR(ZFS_ERR_IOC_CMD_UNAVAIL));
7389 vec = &zfs_ioc_vec[vecnum];
7392 * The registered ioctl list may be sparse, verify that either
7393 * a normal or legacy handler are registered.
7395 if (vec->zvec_func == NULL && vec->zvec_legacy_func == NULL)
7396 return (SET_ERROR(ZFS_ERR_IOC_CMD_UNAVAIL));
7398 zc->zc_iflags = flag & FKIOCTL;
7399 if (zc->zc_nvlist_src_size > MAX_NVLIST_SRC_SIZE) {
7401 * Make sure the user doesn't pass in an insane value for
7402 * zc_nvlist_src_size. We have to check, since we will end
7403 * up allocating that much memory inside of get_nvlist(). This
7404 * prevents a nefarious user from allocating tons of kernel
7407 * Also, we return EINVAL instead of ENOMEM here. The reason
7408 * being that returning ENOMEM from an ioctl() has a special
7409 * connotation; that the user's size value is too small and
7410 * needs to be expanded to hold the nvlist. See
7411 * zcmd_expand_dst_nvlist() for details.
7413 error = SET_ERROR(EINVAL); /* User's size too big */
7415 } else if (zc->zc_nvlist_src_size != 0) {
7416 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
7417 zc->zc_iflags, &innvl);
7423 * Ensure that all pool/dataset names are valid before we pass down to
7426 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
7427 switch (vec->zvec_namecheck) {
7429 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
7430 error = SET_ERROR(EINVAL);
7432 error = pool_status_check(zc->zc_name,
7433 vec->zvec_namecheck, vec->zvec_pool_check);
7437 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
7438 error = SET_ERROR(EINVAL);
7440 error = pool_status_check(zc->zc_name,
7441 vec->zvec_namecheck, vec->zvec_pool_check);
7445 if (entity_namecheck(zc->zc_name, NULL, NULL) != 0) {
7446 error = SET_ERROR(EINVAL);
7448 error = pool_status_check(zc->zc_name,
7449 vec->zvec_namecheck, vec->zvec_pool_check);
7457 * Ensure that all input pairs are valid before we pass them down
7458 * to the lower layers.
7460 * The vectored functions can use fnvlist_lookup_{type} for any
7461 * required pairs since zfs_check_input_nvpairs() confirmed that
7462 * they exist and are of the correct type.
7464 if (error == 0 && vec->zvec_func != NULL) {
7465 error = zfs_check_input_nvpairs(innvl, vec);
7471 cookie = spl_fstrans_mark();
7472 error = vec->zvec_secpolicy(zc, innvl, CRED());
7473 spl_fstrans_unmark(cookie);
7479 /* legacy ioctls can modify zc_name */
7481 * Can't use kmem_strdup() as we might truncate the string and
7482 * kmem_strfree() would then free with incorrect size.
7484 saved_poolname_len = strlen(zc->zc_name) + 1;
7485 saved_poolname = kmem_alloc(saved_poolname_len, KM_SLEEP);
7487 strlcpy(saved_poolname, zc->zc_name, saved_poolname_len);
7488 saved_poolname[strcspn(saved_poolname, "/@#")] = '\0';
7490 if (vec->zvec_func != NULL) {
7494 nvlist_t *lognv = NULL;
7496 ASSERT(vec->zvec_legacy_func == NULL);
7499 * Add the innvl to the lognv before calling the func,
7500 * in case the func changes the innvl.
7502 if (vec->zvec_allow_log) {
7503 lognv = fnvlist_alloc();
7504 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
7506 if (!nvlist_empty(innvl)) {
7507 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
7512 outnvl = fnvlist_alloc();
7513 cookie = spl_fstrans_mark();
7514 error = vec->zvec_func(zc->zc_name, innvl, outnvl);
7515 spl_fstrans_unmark(cookie);
7518 * Some commands can partially execute, modify state, and still
7519 * return an error. In these cases, attempt to record what
7523 (cmd == ZFS_IOC_CHANNEL_PROGRAM && error != EINVAL)) &&
7524 vec->zvec_allow_log &&
7525 spa_open(zc->zc_name, &spa, FTAG) == 0) {
7526 if (!nvlist_empty(outnvl)) {
7527 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
7531 fnvlist_add_int64(lognv, ZPOOL_HIST_ERRNO,
7534 (void) spa_history_log_nvl(spa, lognv);
7535 spa_close(spa, FTAG);
7537 fnvlist_free(lognv);
7539 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
7541 if (vec->zvec_smush_outnvlist) {
7542 smusherror = nvlist_smush(outnvl,
7543 zc->zc_nvlist_dst_size);
7545 if (smusherror == 0)
7546 puterror = put_nvlist(zc, outnvl);
7552 nvlist_free(outnvl);
7554 cookie = spl_fstrans_mark();
7555 error = vec->zvec_legacy_func(zc);
7556 spl_fstrans_unmark(cookie);
7561 if (error == 0 && vec->zvec_allow_log) {
7562 char *s = tsd_get(zfs_allow_log_key);
7565 (void) tsd_set(zfs_allow_log_key, kmem_strdup(saved_poolname));
7567 if (saved_poolname != NULL)
7568 kmem_free(saved_poolname, saved_poolname_len);
7578 if ((error = zvol_init()) != 0)
7581 spa_init(SPA_MODE_READ | SPA_MODE_WRITE);
7586 mutex_init(&zfsdev_state_lock, NULL, MUTEX_DEFAULT, NULL);
7587 zfsdev_state_list = kmem_zalloc(sizeof (zfsdev_state_t), KM_SLEEP);
7588 zfsdev_state_list->zs_minor = -1;
7590 if ((error = zfsdev_attach()) != 0)
7593 tsd_create(&zfs_fsyncer_key, NULL);
7594 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
7595 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
7609 zfsdev_state_t *zs, *zsprev = NULL;
7613 mutex_destroy(&zfsdev_state_lock);
7615 for (zs = zfsdev_state_list; zs != NULL; zs = zs->zs_next) {
7617 kmem_free(zsprev, sizeof (zfsdev_state_t));
7621 kmem_free(zsprev, sizeof (zfsdev_state_t));
7627 tsd_destroy(&zfs_fsyncer_key);
7628 tsd_destroy(&rrw_tsd_key);
7629 tsd_destroy(&zfs_allow_log_key);