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.
41 * Copyright (c) 2019, 2021, Klara Inc.
42 * Copyright (c) 2019, Allan Jude
48 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
49 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
51 * There are two ways that we handle ioctls: the legacy way where almost
52 * all of the logic is in the ioctl callback, and the new way where most
53 * of the marshalling is handled in the common entry point, zfsdev_ioctl().
55 * Non-legacy ioctls should be registered by calling
56 * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked
57 * from userland by lzc_ioctl().
59 * The registration arguments are as follows:
62 * The name of the ioctl. This is used for history logging. If the
63 * ioctl returns successfully (the callback returns 0), and allow_log
64 * is true, then a history log entry will be recorded with the input &
65 * output nvlists. The log entry can be printed with "zpool history -i".
68 * The ioctl request number, which userland will pass to ioctl(2).
69 * We want newer versions of libzfs and libzfs_core to run against
70 * existing zfs kernel modules (i.e. a deferred reboot after an update).
71 * Therefore the ioctl numbers cannot change from release to release.
73 * zfs_secpolicy_func_t *secpolicy
74 * This function will be called before the zfs_ioc_func_t, to
75 * determine if this operation is permitted. It should return EPERM
76 * on failure, and 0 on success. Checks include determining if the
77 * dataset is visible in this zone, and if the user has either all
78 * zfs privileges in the zone (SYS_MOUNT), or has been granted permission
79 * to do this operation on this dataset with "zfs allow".
81 * zfs_ioc_namecheck_t namecheck
82 * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
83 * name, a dataset name, or nothing. If the name is not well-formed,
84 * the ioctl will fail and the callback will not be called.
85 * Therefore, the callback can assume that the name is well-formed
86 * (e.g. is null-terminated, doesn't have more than one '@' character,
87 * doesn't have invalid characters).
89 * zfs_ioc_poolcheck_t pool_check
90 * This specifies requirements on the pool state. If the pool does
91 * not meet them (is suspended or is readonly), the ioctl will fail
92 * and the callback will not be called. If any checks are specified
93 * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
94 * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED |
95 * POOL_CHECK_READONLY).
97 * zfs_ioc_key_t *nvl_keys
98 * The list of expected/allowable innvl input keys. This list is used
99 * to validate the nvlist input to the ioctl.
101 * boolean_t smush_outnvlist
102 * If smush_outnvlist is true, then the output is presumed to be a
103 * list of errors, and it will be "smushed" down to fit into the
104 * caller's buffer, by removing some entries and replacing them with a
105 * single "N_MORE_ERRORS" entry indicating how many were removed. See
106 * nvlist_smush() for details. If smush_outnvlist is false, and the
107 * outnvlist does not fit into the userland-provided buffer, then the
108 * ioctl will fail with ENOMEM.
110 * zfs_ioc_func_t *func
111 * The callback function that will perform the operation.
113 * The callback should return 0 on success, or an error number on
114 * failure. If the function fails, the userland ioctl will return -1,
115 * and errno will be set to the callback's return value. The callback
116 * will be called with the following arguments:
119 * The name of the pool or dataset to operate on, from
120 * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the
121 * expected type (pool, dataset, or none).
124 * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or
125 * NULL if no input nvlist was provided. Changes to this nvlist are
126 * ignored. If the input nvlist could not be deserialized, the
127 * ioctl will fail and the callback will not be called.
130 * The output nvlist, initially empty. The callback can fill it in,
131 * and it will be returned to userland by serializing it into
132 * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization
133 * fails (e.g. because the caller didn't supply a large enough
134 * buffer), then the overall ioctl will fail. See the
135 * 'smush_nvlist' argument above for additional behaviors.
137 * There are two typical uses of the output nvlist:
138 * - To return state, e.g. property values. In this case,
139 * smush_outnvlist should be false. If the buffer was not large
140 * enough, the caller will reallocate a larger buffer and try
143 * - To return multiple errors from an ioctl which makes on-disk
144 * changes. In this case, smush_outnvlist should be true.
145 * Ioctls which make on-disk modifications should generally not
146 * use the outnvl if they succeed, because the caller can not
147 * distinguish between the operation failing, and
148 * deserialization failing.
150 * IOCTL Interface Errors
152 * The following ioctl input errors can be returned:
153 * ZFS_ERR_IOC_CMD_UNAVAIL the ioctl number is not supported by kernel
154 * ZFS_ERR_IOC_ARG_UNAVAIL an input argument is not supported by kernel
155 * ZFS_ERR_IOC_ARG_REQUIRED a required input argument is missing
156 * ZFS_ERR_IOC_ARG_BADTYPE an input argument has an invalid type
159 #include <sys/types.h>
160 #include <sys/param.h>
161 #include <sys/errno.h>
162 #include <sys/uio_impl.h>
163 #include <sys/file.h>
164 #include <sys/kmem.h>
165 #include <sys/cmn_err.h>
166 #include <sys/stat.h>
167 #include <sys/zfs_ioctl.h>
168 #include <sys/zfs_quota.h>
169 #include <sys/zfs_vfsops.h>
170 #include <sys/zfs_znode.h>
173 #include <sys/spa_impl.h>
174 #include <sys/vdev.h>
175 #include <sys/vdev_impl.h>
177 #include <sys/dsl_dir.h>
178 #include <sys/dsl_dataset.h>
179 #include <sys/dsl_prop.h>
180 #include <sys/dsl_deleg.h>
181 #include <sys/dmu_objset.h>
182 #include <sys/dmu_impl.h>
183 #include <sys/dmu_redact.h>
184 #include <sys/dmu_tx.h>
185 #include <sys/sunddi.h>
186 #include <sys/policy.h>
187 #include <sys/zone.h>
188 #include <sys/nvpair.h>
189 #include <sys/pathname.h>
190 #include <sys/fs/zfs.h>
191 #include <sys/zfs_ctldir.h>
192 #include <sys/zfs_dir.h>
193 #include <sys/zfs_onexit.h>
194 #include <sys/zvol.h>
195 #include <sys/dsl_scan.h>
196 #include <sys/fm/util.h>
197 #include <sys/dsl_crypt.h>
198 #include <sys/rrwlock.h>
199 #include <sys/zfs_file.h>
201 #include <sys/dmu_recv.h>
202 #include <sys/dmu_send.h>
203 #include <sys/dmu_recv.h>
204 #include <sys/dsl_destroy.h>
205 #include <sys/dsl_bookmark.h>
206 #include <sys/dsl_userhold.h>
207 #include <sys/zfeature.h>
209 #include <sys/zio_checksum.h>
210 #include <sys/vdev_removal.h>
211 #include <sys/vdev_impl.h>
212 #include <sys/vdev_initialize.h>
213 #include <sys/vdev_trim.h>
215 #include "zfs_namecheck.h"
216 #include "zfs_prop.h"
217 #include "zfs_deleg.h"
218 #include "zfs_comutil.h"
220 #include <sys/lua/lua.h>
221 #include <sys/lua/lauxlib.h>
222 #include <sys/zfs_ioctl_impl.h>
224 kmutex_t zfsdev_state_lock;
225 static zfsdev_state_t *zfsdev_state_list;
228 * Limit maximum nvlist size. We don't want users passing in insane values
229 * for zc->zc_nvlist_src_size, since we will need to allocate that much memory.
230 * Defaults to 0=auto which is handled by platform code.
232 unsigned long zfs_max_nvlist_src_size = 0;
235 * When logging the output nvlist of an ioctl in the on-disk history, limit
236 * the logged size to this many bytes. This must be less than DMU_MAX_ACCESS.
237 * This applies primarily to zfs_ioc_channel_program().
239 static unsigned long zfs_history_output_max = 1024 * 1024;
241 uint_t zfs_fsyncer_key;
242 uint_t zfs_allow_log_key;
244 /* DATA_TYPE_ANY is used when zkey_type can vary. */
245 #define DATA_TYPE_ANY DATA_TYPE_UNKNOWN
247 typedef struct zfs_ioc_vec {
248 zfs_ioc_legacy_func_t *zvec_legacy_func;
249 zfs_ioc_func_t *zvec_func;
250 zfs_secpolicy_func_t *zvec_secpolicy;
251 zfs_ioc_namecheck_t zvec_namecheck;
252 boolean_t zvec_allow_log;
253 zfs_ioc_poolcheck_t zvec_pool_check;
254 boolean_t zvec_smush_outnvlist;
255 const char *zvec_name;
256 const zfs_ioc_key_t *zvec_nvl_keys;
257 size_t zvec_nvl_key_count;
260 /* This array is indexed by zfs_userquota_prop_t */
261 static const char *userquota_perms[] = {
262 ZFS_DELEG_PERM_USERUSED,
263 ZFS_DELEG_PERM_USERQUOTA,
264 ZFS_DELEG_PERM_GROUPUSED,
265 ZFS_DELEG_PERM_GROUPQUOTA,
266 ZFS_DELEG_PERM_USEROBJUSED,
267 ZFS_DELEG_PERM_USEROBJQUOTA,
268 ZFS_DELEG_PERM_GROUPOBJUSED,
269 ZFS_DELEG_PERM_GROUPOBJQUOTA,
270 ZFS_DELEG_PERM_PROJECTUSED,
271 ZFS_DELEG_PERM_PROJECTQUOTA,
272 ZFS_DELEG_PERM_PROJECTOBJUSED,
273 ZFS_DELEG_PERM_PROJECTOBJQUOTA,
276 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
277 static int zfs_ioc_id_quota_upgrade(zfs_cmd_t *zc);
278 static int zfs_check_settable(const char *name, nvpair_t *property,
280 static int zfs_check_clearable(const char *dataset, nvlist_t *props,
282 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
284 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
285 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
288 history_str_free(char *buf)
290 kmem_free(buf, HIS_MAX_RECORD_LEN);
294 history_str_get(zfs_cmd_t *zc)
298 if (zc->zc_history == 0)
301 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
302 if (copyinstr((void *)(uintptr_t)zc->zc_history,
303 buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
304 history_str_free(buf);
308 buf[HIS_MAX_RECORD_LEN -1] = '\0';
314 * Return non-zero if the spa version is less than requested version.
317 zfs_earlier_version(const char *name, int version)
321 if (spa_open(name, &spa, FTAG) == 0) {
322 if (spa_version(spa) < version) {
323 spa_close(spa, FTAG);
326 spa_close(spa, FTAG);
332 * Return TRUE if the ZPL version is less than requested version.
335 zpl_earlier_version(const char *name, int version)
338 boolean_t rc = B_TRUE;
340 if (dmu_objset_hold(name, FTAG, &os) == 0) {
343 if (dmu_objset_type(os) != DMU_OST_ZFS) {
344 dmu_objset_rele(os, FTAG);
347 /* XXX reading from non-owned objset */
348 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
349 rc = zplversion < version;
350 dmu_objset_rele(os, FTAG);
356 zfs_log_history(zfs_cmd_t *zc)
361 if ((buf = history_str_get(zc)) == NULL)
364 if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
365 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
366 (void) spa_history_log(spa, buf);
367 spa_close(spa, FTAG);
369 history_str_free(buf);
373 * Policy for top-level read operations (list pools). Requires no privileges,
374 * and can be used in the local zone, as there is no associated dataset.
377 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
379 (void) zc, (void) innvl, (void) cr;
384 * Policy for dataset read operations (list children, get statistics). Requires
385 * no privileges, but must be visible in the local zone.
388 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
390 (void) innvl, (void) cr;
391 if (INGLOBALZONE(curproc) ||
392 zone_dataset_visible(zc->zc_name, NULL))
395 return (SET_ERROR(ENOENT));
399 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
404 * The dataset must be visible by this zone -- check this first
405 * so they don't see EPERM on something they shouldn't know about.
407 if (!INGLOBALZONE(curproc) &&
408 !zone_dataset_visible(dataset, &writable))
409 return (SET_ERROR(ENOENT));
411 if (INGLOBALZONE(curproc)) {
413 * If the fs is zoned, only root can access it from the
416 if (secpolicy_zfs(cr) && zoned)
417 return (SET_ERROR(EPERM));
420 * If we are in a local zone, the 'zoned' property must be set.
423 return (SET_ERROR(EPERM));
425 /* must be writable by this zone */
427 return (SET_ERROR(EPERM));
433 zfs_dozonecheck(const char *dataset, cred_t *cr)
437 if (dsl_prop_get_integer(dataset, zfs_prop_to_name(ZFS_PROP_ZONED),
439 return (SET_ERROR(ENOENT));
441 return (zfs_dozonecheck_impl(dataset, zoned, cr));
445 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
449 if (dsl_prop_get_int_ds(ds, zfs_prop_to_name(ZFS_PROP_ZONED), &zoned))
450 return (SET_ERROR(ENOENT));
452 return (zfs_dozonecheck_impl(dataset, zoned, cr));
456 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
457 const char *perm, cred_t *cr)
461 error = zfs_dozonecheck_ds(name, ds, cr);
463 error = secpolicy_zfs(cr);
465 error = dsl_deleg_access_impl(ds, perm, cr);
471 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
478 * First do a quick check for root in the global zone, which
479 * is allowed to do all write_perms. This ensures that zfs_ioc_*
480 * will get to handle nonexistent datasets.
482 if (INGLOBALZONE(curproc) && secpolicy_zfs(cr) == 0)
485 error = dsl_pool_hold(name, FTAG, &dp);
489 error = dsl_dataset_hold(dp, name, FTAG, &ds);
491 dsl_pool_rele(dp, FTAG);
495 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
497 dsl_dataset_rele(ds, FTAG);
498 dsl_pool_rele(dp, FTAG);
503 * Policy for setting the security label property.
505 * Returns 0 for success, non-zero for access and other errors.
508 zfs_set_slabel_policy(const char *name, const char *strval, cred_t *cr)
511 char ds_hexsl[MAXNAMELEN];
512 bslabel_t ds_sl, new_sl;
513 boolean_t new_default = FALSE;
515 int needed_priv = -1;
518 /* First get the existing dataset label. */
519 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
520 1, sizeof (ds_hexsl), &ds_hexsl, NULL);
522 return (SET_ERROR(EPERM));
524 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
527 /* The label must be translatable */
528 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
529 return (SET_ERROR(EINVAL));
532 * In a non-global zone, disallow attempts to set a label that
533 * doesn't match that of the zone; otherwise no other checks
536 if (!INGLOBALZONE(curproc)) {
537 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
538 return (SET_ERROR(EPERM));
543 * For global-zone datasets (i.e., those whose zoned property is
544 * "off", verify that the specified new label is valid for the
547 if (dsl_prop_get_integer(name,
548 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
549 return (SET_ERROR(EPERM));
551 if (zfs_check_global_label(name, strval) != 0)
552 return (SET_ERROR(EPERM));
556 * If the existing dataset label is nondefault, check if the
557 * dataset is mounted (label cannot be changed while mounted).
558 * Get the zfsvfs_t; if there isn't one, then the dataset isn't
559 * mounted (or isn't a dataset, doesn't exist, ...).
561 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
563 static const char *setsl_tag = "setsl_tag";
566 * Try to own the dataset; abort if there is any error,
567 * (e.g., already mounted, in use, or other error).
569 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE, B_TRUE,
572 return (SET_ERROR(EPERM));
574 dmu_objset_disown(os, B_TRUE, setsl_tag);
577 needed_priv = PRIV_FILE_DOWNGRADE_SL;
581 if (hexstr_to_label(strval, &new_sl) != 0)
582 return (SET_ERROR(EPERM));
584 if (blstrictdom(&ds_sl, &new_sl))
585 needed_priv = PRIV_FILE_DOWNGRADE_SL;
586 else if (blstrictdom(&new_sl, &ds_sl))
587 needed_priv = PRIV_FILE_UPGRADE_SL;
589 /* dataset currently has a default label */
591 needed_priv = PRIV_FILE_UPGRADE_SL;
595 if (needed_priv != -1)
596 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
599 return (SET_ERROR(ENOTSUP));
600 #endif /* HAVE_MLSLABEL */
604 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
610 * Check permissions for special properties.
617 * Disallow setting of 'zoned' from within a local zone.
619 if (!INGLOBALZONE(curproc))
620 return (SET_ERROR(EPERM));
624 case ZFS_PROP_FILESYSTEM_LIMIT:
625 case ZFS_PROP_SNAPSHOT_LIMIT:
626 if (!INGLOBALZONE(curproc)) {
628 char setpoint[ZFS_MAX_DATASET_NAME_LEN];
630 * Unprivileged users are allowed to modify the
631 * limit on things *under* (ie. contained by)
632 * the thing they own.
634 if (dsl_prop_get_integer(dsname,
635 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, setpoint))
636 return (SET_ERROR(EPERM));
637 if (!zoned || strlen(dsname) <= strlen(setpoint))
638 return (SET_ERROR(EPERM));
642 case ZFS_PROP_MLSLABEL:
643 if (!is_system_labeled())
644 return (SET_ERROR(EPERM));
646 if (nvpair_value_string(propval, &strval) == 0) {
649 err = zfs_set_slabel_policy(dsname, strval, CRED());
656 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
660 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
663 * permission to set permissions will be evaluated later in
664 * dsl_deleg_can_allow()
667 return (zfs_dozonecheck(zc->zc_name, cr));
671 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
674 return (zfs_secpolicy_write_perms(zc->zc_name,
675 ZFS_DELEG_PERM_ROLLBACK, cr));
679 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
688 * Generate the current snapshot name from the given objsetid, then
689 * use that name for the secpolicy/zone checks.
691 cp = strchr(zc->zc_name, '@');
693 return (SET_ERROR(EINVAL));
694 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
698 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
700 dsl_pool_rele(dp, FTAG);
704 dsl_dataset_name(ds, zc->zc_name);
706 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
707 ZFS_DELEG_PERM_SEND, cr);
708 dsl_dataset_rele(ds, FTAG);
709 dsl_pool_rele(dp, FTAG);
715 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
718 return (zfs_secpolicy_write_perms(zc->zc_name,
719 ZFS_DELEG_PERM_SEND, cr));
723 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
725 (void) zc, (void) innvl, (void) cr;
726 return (SET_ERROR(ENOTSUP));
730 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
732 (void) zc, (void) innvl, (void) cr;
733 return (SET_ERROR(ENOTSUP));
737 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
742 * Remove the @bla or /bla from the end of the name to get the parent.
744 (void) strncpy(parent, datasetname, parentsize);
745 cp = strrchr(parent, '@');
749 cp = strrchr(parent, '/');
751 return (SET_ERROR(ENOENT));
759 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
763 if ((error = zfs_secpolicy_write_perms(name,
764 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
767 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
771 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
774 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
778 * Destroying snapshots with delegated permissions requires
779 * descendant mount and destroy permissions.
782 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
786 nvpair_t *pair, *nextpair;
789 snaps = fnvlist_lookup_nvlist(innvl, "snaps");
791 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
793 nextpair = nvlist_next_nvpair(snaps, pair);
794 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr);
795 if (error == ENOENT) {
797 * Ignore any snapshots that don't exist (we consider
798 * them "already destroyed"). Remove the name from the
799 * nvl here in case the snapshot is created between
800 * now and when we try to destroy it (in which case
801 * we don't want to destroy it since we haven't
802 * checked for permission).
804 fnvlist_remove_nvpair(snaps, pair);
815 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
817 char parentname[ZFS_MAX_DATASET_NAME_LEN];
820 if ((error = zfs_secpolicy_write_perms(from,
821 ZFS_DELEG_PERM_RENAME, cr)) != 0)
824 if ((error = zfs_secpolicy_write_perms(from,
825 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
828 if ((error = zfs_get_parent(to, parentname,
829 sizeof (parentname))) != 0)
832 if ((error = zfs_secpolicy_write_perms(parentname,
833 ZFS_DELEG_PERM_CREATE, cr)) != 0)
836 if ((error = zfs_secpolicy_write_perms(parentname,
837 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
844 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
847 return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
851 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
855 dsl_dataset_t *clone;
858 error = zfs_secpolicy_write_perms(zc->zc_name,
859 ZFS_DELEG_PERM_PROMOTE, cr);
863 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
867 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
870 char parentname[ZFS_MAX_DATASET_NAME_LEN];
871 dsl_dataset_t *origin = NULL;
875 error = dsl_dataset_hold_obj(dd->dd_pool,
876 dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin);
878 dsl_dataset_rele(clone, FTAG);
879 dsl_pool_rele(dp, FTAG);
883 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
884 ZFS_DELEG_PERM_MOUNT, cr);
886 dsl_dataset_name(origin, parentname);
888 error = zfs_secpolicy_write_perms_ds(parentname, origin,
889 ZFS_DELEG_PERM_PROMOTE, cr);
891 dsl_dataset_rele(clone, FTAG);
892 dsl_dataset_rele(origin, FTAG);
894 dsl_pool_rele(dp, FTAG);
899 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
904 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
905 ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
908 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
909 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
912 return (zfs_secpolicy_write_perms(zc->zc_name,
913 ZFS_DELEG_PERM_CREATE, cr));
917 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
919 return (zfs_secpolicy_write_perms(name,
920 ZFS_DELEG_PERM_SNAPSHOT, cr));
924 * Check for permission to create each snapshot in the nvlist.
927 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
934 snaps = fnvlist_lookup_nvlist(innvl, "snaps");
936 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
937 pair = nvlist_next_nvpair(snaps, pair)) {
938 char *name = nvpair_name(pair);
939 char *atp = strchr(name, '@');
942 error = SET_ERROR(EINVAL);
946 error = zfs_secpolicy_snapshot_perms(name, cr);
955 * Check for permission to create each bookmark in the nvlist.
958 zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
963 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
964 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
965 char *name = nvpair_name(pair);
966 char *hashp = strchr(name, '#');
969 error = SET_ERROR(EINVAL);
973 error = zfs_secpolicy_write_perms(name,
974 ZFS_DELEG_PERM_BOOKMARK, cr);
983 zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
986 nvpair_t *pair, *nextpair;
989 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
991 char *name = nvpair_name(pair);
992 char *hashp = strchr(name, '#');
993 nextpair = nvlist_next_nvpair(innvl, pair);
996 error = SET_ERROR(EINVAL);
1001 error = zfs_secpolicy_write_perms(name,
1002 ZFS_DELEG_PERM_DESTROY, cr);
1004 if (error == ENOENT) {
1006 * Ignore any filesystems that don't exist (we consider
1007 * their bookmarks "already destroyed"). Remove
1008 * the name from the nvl here in case the filesystem
1009 * is created between now and when we try to destroy
1010 * the bookmark (in which case we don't want to
1011 * destroy it since we haven't checked for permission).
1013 fnvlist_remove_nvpair(innvl, pair);
1024 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1026 (void) zc, (void) innvl, (void) cr;
1028 * Even root must have a proper TSD so that we know what pool
1031 if (tsd_get(zfs_allow_log_key) == NULL)
1032 return (SET_ERROR(EPERM));
1037 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1039 char parentname[ZFS_MAX_DATASET_NAME_LEN];
1043 if ((error = zfs_get_parent(zc->zc_name, parentname,
1044 sizeof (parentname))) != 0)
1047 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1048 (error = zfs_secpolicy_write_perms(origin,
1049 ZFS_DELEG_PERM_CLONE, cr)) != 0)
1052 if ((error = zfs_secpolicy_write_perms(parentname,
1053 ZFS_DELEG_PERM_CREATE, cr)) != 0)
1056 return (zfs_secpolicy_write_perms(parentname,
1057 ZFS_DELEG_PERM_MOUNT, cr));
1061 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires
1062 * SYS_CONFIG privilege, which is not available in a local zone.
1065 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1067 (void) zc, (void) innvl;
1069 if (secpolicy_sys_config(cr, B_FALSE) != 0)
1070 return (SET_ERROR(EPERM));
1076 * Policy for object to name lookups.
1079 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1084 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1087 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1092 * Policy for fault injection. Requires all privileges.
1095 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1097 (void) zc, (void) innvl;
1098 return (secpolicy_zinject(cr));
1102 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1105 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1107 if (prop == ZPROP_USERPROP) {
1108 if (!zfs_prop_user(zc->zc_value))
1109 return (SET_ERROR(EINVAL));
1110 return (zfs_secpolicy_write_perms(zc->zc_name,
1111 ZFS_DELEG_PERM_USERPROP, cr));
1113 return (zfs_secpolicy_setprop(zc->zc_name, prop,
1119 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1121 int err = zfs_secpolicy_read(zc, innvl, cr);
1125 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1126 return (SET_ERROR(EINVAL));
1128 if (zc->zc_value[0] == 0) {
1130 * They are asking about a posix uid/gid. If it's
1131 * themself, allow it.
1133 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1134 zc->zc_objset_type == ZFS_PROP_USERQUOTA ||
1135 zc->zc_objset_type == ZFS_PROP_USEROBJUSED ||
1136 zc->zc_objset_type == ZFS_PROP_USEROBJQUOTA) {
1137 if (zc->zc_guid == crgetuid(cr))
1139 } else if (zc->zc_objset_type == ZFS_PROP_GROUPUSED ||
1140 zc->zc_objset_type == ZFS_PROP_GROUPQUOTA ||
1141 zc->zc_objset_type == ZFS_PROP_GROUPOBJUSED ||
1142 zc->zc_objset_type == ZFS_PROP_GROUPOBJQUOTA) {
1143 if (groupmember(zc->zc_guid, cr))
1146 /* else is for project quota/used */
1149 return (zfs_secpolicy_write_perms(zc->zc_name,
1150 userquota_perms[zc->zc_objset_type], cr));
1154 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1156 int err = zfs_secpolicy_read(zc, innvl, cr);
1160 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1161 return (SET_ERROR(EINVAL));
1163 return (zfs_secpolicy_write_perms(zc->zc_name,
1164 userquota_perms[zc->zc_objset_type], cr));
1168 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1171 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1176 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1183 holds = fnvlist_lookup_nvlist(innvl, "holds");
1185 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
1186 pair = nvlist_next_nvpair(holds, pair)) {
1187 char fsname[ZFS_MAX_DATASET_NAME_LEN];
1188 error = dmu_fsname(nvpair_name(pair), fsname);
1191 error = zfs_secpolicy_write_perms(fsname,
1192 ZFS_DELEG_PERM_HOLD, cr);
1200 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1206 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1207 pair = nvlist_next_nvpair(innvl, pair)) {
1208 char fsname[ZFS_MAX_DATASET_NAME_LEN];
1209 error = dmu_fsname(nvpair_name(pair), fsname);
1212 error = zfs_secpolicy_write_perms(fsname,
1213 ZFS_DELEG_PERM_RELEASE, cr);
1221 * Policy for allowing temporary snapshots to be taken or released
1224 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1227 * A temporary snapshot is the same as a snapshot,
1228 * hold, destroy and release all rolled into one.
1229 * Delegated diff alone is sufficient that we allow this.
1233 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1234 ZFS_DELEG_PERM_DIFF, cr)) == 0)
1237 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1239 if (innvl != NULL) {
1241 error = zfs_secpolicy_hold(zc, innvl, cr);
1243 error = zfs_secpolicy_release(zc, innvl, cr);
1245 error = zfs_secpolicy_destroy(zc, innvl, cr);
1251 zfs_secpolicy_load_key(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1253 return (zfs_secpolicy_write_perms(zc->zc_name,
1254 ZFS_DELEG_PERM_LOAD_KEY, cr));
1258 zfs_secpolicy_change_key(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1260 return (zfs_secpolicy_write_perms(zc->zc_name,
1261 ZFS_DELEG_PERM_CHANGE_KEY, cr));
1265 * Returns the nvlist as specified by the user in the zfs_cmd_t.
1268 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1272 nvlist_t *list = NULL;
1275 * Read in and unpack the user-supplied nvlist.
1278 return (SET_ERROR(EINVAL));
1280 packed = vmem_alloc(size, KM_SLEEP);
1282 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1284 vmem_free(packed, size);
1285 return (SET_ERROR(EFAULT));
1288 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1289 vmem_free(packed, size);
1293 vmem_free(packed, size);
1300 * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1301 * Entries will be removed from the end of the nvlist, and one int32 entry
1302 * named "N_MORE_ERRORS" will be added indicating how many entries were
1306 nvlist_smush(nvlist_t *errors, size_t max)
1310 size = fnvlist_size(errors);
1313 nvpair_t *more_errors;
1317 return (SET_ERROR(ENOMEM));
1319 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1320 more_errors = nvlist_prev_nvpair(errors, NULL);
1323 nvpair_t *pair = nvlist_prev_nvpair(errors,
1325 fnvlist_remove_nvpair(errors, pair);
1327 size = fnvlist_size(errors);
1328 } while (size > max);
1330 fnvlist_remove_nvpair(errors, more_errors);
1331 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1332 ASSERT3U(fnvlist_size(errors), <=, max);
1339 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1341 char *packed = NULL;
1345 size = fnvlist_size(nvl);
1347 if (size > zc->zc_nvlist_dst_size) {
1348 error = SET_ERROR(ENOMEM);
1350 packed = fnvlist_pack(nvl, &size);
1351 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1352 size, zc->zc_iflags) != 0)
1353 error = SET_ERROR(EFAULT);
1354 fnvlist_pack_free(packed, size);
1357 zc->zc_nvlist_dst_size = size;
1358 zc->zc_nvlist_dst_filled = B_TRUE;
1363 getzfsvfs_impl(objset_t *os, zfsvfs_t **zfvp)
1366 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1367 return (SET_ERROR(EINVAL));
1370 mutex_enter(&os->os_user_ptr_lock);
1371 *zfvp = dmu_objset_get_user(os);
1372 /* bump s_active only when non-zero to prevent umount race */
1373 error = zfs_vfs_ref(zfvp);
1374 mutex_exit(&os->os_user_ptr_lock);
1379 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1384 error = dmu_objset_hold(dsname, FTAG, &os);
1388 error = getzfsvfs_impl(os, zfvp);
1389 dmu_objset_rele(os, FTAG);
1394 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1395 * case its z_sb will be NULL, and it will be opened as the owner.
1396 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1397 * which prevents all inode ops from running.
1400 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1404 if (getzfsvfs(name, zfvp) != 0)
1405 error = zfsvfs_create(name, B_FALSE, zfvp);
1408 ZFS_TEARDOWN_ENTER_WRITE(*zfvp, tag);
1410 ZFS_TEARDOWN_ENTER_READ(*zfvp, tag);
1411 if ((*zfvp)->z_unmounted) {
1413 * XXX we could probably try again, since the unmounting
1414 * thread should be just about to disassociate the
1415 * objset from the zfsvfs.
1417 ZFS_TEARDOWN_EXIT(*zfvp, tag);
1418 return (SET_ERROR(EBUSY));
1425 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1427 ZFS_TEARDOWN_EXIT(zfsvfs, tag);
1429 if (zfs_vfs_held(zfsvfs)) {
1430 zfs_vfs_rele(zfsvfs);
1432 dmu_objset_disown(zfsvfs->z_os, B_TRUE, zfsvfs);
1433 zfsvfs_free(zfsvfs);
1438 zfs_ioc_pool_create(zfs_cmd_t *zc)
1441 nvlist_t *config, *props = NULL;
1442 nvlist_t *rootprops = NULL;
1443 nvlist_t *zplprops = NULL;
1444 dsl_crypto_params_t *dcp = NULL;
1445 const char *spa_name = zc->zc_name;
1446 boolean_t unload_wkey = B_TRUE;
1448 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1449 zc->zc_iflags, &config)))
1452 if (zc->zc_nvlist_src_size != 0 && (error =
1453 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1454 zc->zc_iflags, &props))) {
1455 nvlist_free(config);
1460 nvlist_t *nvl = NULL;
1461 nvlist_t *hidden_args = NULL;
1462 uint64_t version = SPA_VERSION;
1465 (void) nvlist_lookup_uint64(props,
1466 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1467 if (!SPA_VERSION_IS_SUPPORTED(version)) {
1468 error = SET_ERROR(EINVAL);
1469 goto pool_props_bad;
1471 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1473 error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1475 goto pool_props_bad;
1476 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1479 (void) nvlist_lookup_nvlist(props, ZPOOL_HIDDEN_ARGS,
1481 error = dsl_crypto_params_create_nvlist(DCP_CMD_NONE,
1482 rootprops, hidden_args, &dcp);
1484 goto pool_props_bad;
1485 (void) nvlist_remove_all(props, ZPOOL_HIDDEN_ARGS);
1487 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1488 error = zfs_fill_zplprops_root(version, rootprops,
1491 goto pool_props_bad;
1493 if (nvlist_lookup_string(props,
1494 zpool_prop_to_name(ZPOOL_PROP_TNAME), &tname) == 0)
1498 error = spa_create(zc->zc_name, config, props, zplprops, dcp);
1501 * Set the remaining root properties
1503 if (!error && (error = zfs_set_prop_nvlist(spa_name,
1504 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0) {
1505 (void) spa_destroy(spa_name);
1506 unload_wkey = B_FALSE; /* spa_destroy() unloads wrapping keys */
1510 nvlist_free(rootprops);
1511 nvlist_free(zplprops);
1512 nvlist_free(config);
1514 dsl_crypto_params_free(dcp, unload_wkey && !!error);
1520 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1523 zfs_log_history(zc);
1524 error = spa_destroy(zc->zc_name);
1530 zfs_ioc_pool_import(zfs_cmd_t *zc)
1532 nvlist_t *config, *props = NULL;
1536 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1537 zc->zc_iflags, &config)) != 0)
1540 if (zc->zc_nvlist_src_size != 0 && (error =
1541 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1542 zc->zc_iflags, &props))) {
1543 nvlist_free(config);
1547 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1548 guid != zc->zc_guid)
1549 error = SET_ERROR(EINVAL);
1551 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1553 if (zc->zc_nvlist_dst != 0) {
1556 if ((err = put_nvlist(zc, config)) != 0)
1560 nvlist_free(config);
1567 zfs_ioc_pool_export(zfs_cmd_t *zc)
1570 boolean_t force = (boolean_t)zc->zc_cookie;
1571 boolean_t hardforce = (boolean_t)zc->zc_guid;
1573 zfs_log_history(zc);
1574 error = spa_export(zc->zc_name, NULL, force, hardforce);
1580 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1585 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1586 return (SET_ERROR(EEXIST));
1588 error = put_nvlist(zc, configs);
1590 nvlist_free(configs);
1597 * zc_name name of the pool
1600 * zc_cookie real errno
1601 * zc_nvlist_dst config nvlist
1602 * zc_nvlist_dst_size size of config nvlist
1605 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1611 error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1612 sizeof (zc->zc_value));
1614 if (config != NULL) {
1615 ret = put_nvlist(zc, config);
1616 nvlist_free(config);
1619 * The config may be present even if 'error' is non-zero.
1620 * In this case we return success, and preserve the real errno
1623 zc->zc_cookie = error;
1632 * Try to import the given pool, returning pool stats as appropriate so that
1633 * user land knows which devices are available and overall pool health.
1636 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1638 nvlist_t *tryconfig, *config = NULL;
1641 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1642 zc->zc_iflags, &tryconfig)) != 0)
1645 config = spa_tryimport(tryconfig);
1647 nvlist_free(tryconfig);
1650 return (SET_ERROR(EINVAL));
1652 error = put_nvlist(zc, config);
1653 nvlist_free(config);
1660 * zc_name name of the pool
1661 * zc_cookie scan func (pool_scan_func_t)
1662 * zc_flags scrub pause/resume flag (pool_scrub_cmd_t)
1665 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1670 if (zc->zc_flags >= POOL_SCRUB_FLAGS_END)
1671 return (SET_ERROR(EINVAL));
1673 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1676 if (zc->zc_flags == POOL_SCRUB_PAUSE)
1677 error = spa_scrub_pause_resume(spa, POOL_SCRUB_PAUSE);
1678 else if (zc->zc_cookie == POOL_SCAN_NONE)
1679 error = spa_scan_stop(spa);
1681 error = spa_scan(spa, zc->zc_cookie);
1683 spa_close(spa, FTAG);
1689 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1694 error = spa_open(zc->zc_name, &spa, FTAG);
1697 spa_close(spa, FTAG);
1703 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1708 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1711 if (zc->zc_cookie < spa_version(spa) ||
1712 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1713 spa_close(spa, FTAG);
1714 return (SET_ERROR(EINVAL));
1717 spa_upgrade(spa, zc->zc_cookie);
1718 spa_close(spa, FTAG);
1724 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1731 if ((size = zc->zc_history_len) == 0)
1732 return (SET_ERROR(EINVAL));
1734 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1737 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1738 spa_close(spa, FTAG);
1739 return (SET_ERROR(ENOTSUP));
1742 hist_buf = vmem_alloc(size, KM_SLEEP);
1743 if ((error = spa_history_get(spa, &zc->zc_history_offset,
1744 &zc->zc_history_len, hist_buf)) == 0) {
1745 error = ddi_copyout(hist_buf,
1746 (void *)(uintptr_t)zc->zc_history,
1747 zc->zc_history_len, zc->zc_iflags);
1750 spa_close(spa, FTAG);
1751 vmem_free(hist_buf, size);
1756 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1761 error = spa_open(zc->zc_name, &spa, FTAG);
1763 error = spa_change_guid(spa);
1764 spa_close(spa, FTAG);
1770 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1772 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1777 * zc_name name of filesystem
1778 * zc_obj object to find
1781 * zc_value name of object
1784 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1789 /* XXX reading from objset not owned */
1790 if ((error = dmu_objset_hold_flags(zc->zc_name, B_TRUE,
1793 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1794 dmu_objset_rele_flags(os, B_TRUE, FTAG);
1795 return (SET_ERROR(EINVAL));
1797 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1798 sizeof (zc->zc_value));
1799 dmu_objset_rele_flags(os, B_TRUE, FTAG);
1806 * zc_name name of filesystem
1807 * zc_obj object to find
1810 * zc_stat stats on object
1811 * zc_value path to object
1814 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1819 /* XXX reading from objset not owned */
1820 if ((error = dmu_objset_hold_flags(zc->zc_name, B_TRUE,
1823 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1824 dmu_objset_rele_flags(os, B_TRUE, FTAG);
1825 return (SET_ERROR(EINVAL));
1827 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1828 sizeof (zc->zc_value));
1829 dmu_objset_rele_flags(os, B_TRUE, FTAG);
1835 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1841 error = spa_open(zc->zc_name, &spa, FTAG);
1845 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1846 zc->zc_iflags, &config);
1848 error = spa_vdev_add(spa, config);
1849 nvlist_free(config);
1851 spa_close(spa, FTAG);
1857 * zc_name name of the pool
1858 * zc_guid guid of vdev to remove
1859 * zc_cookie cancel removal
1862 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1867 error = spa_open(zc->zc_name, &spa, FTAG);
1870 if (zc->zc_cookie != 0) {
1871 error = spa_vdev_remove_cancel(spa);
1873 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1875 spa_close(spa, FTAG);
1880 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1884 vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1886 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1888 switch (zc->zc_cookie) {
1889 case VDEV_STATE_ONLINE:
1890 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1893 case VDEV_STATE_OFFLINE:
1894 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1897 case VDEV_STATE_FAULTED:
1898 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1899 zc->zc_obj != VDEV_AUX_EXTERNAL &&
1900 zc->zc_obj != VDEV_AUX_EXTERNAL_PERSIST)
1901 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1903 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1906 case VDEV_STATE_DEGRADED:
1907 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1908 zc->zc_obj != VDEV_AUX_EXTERNAL)
1909 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1911 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1915 error = SET_ERROR(EINVAL);
1917 zc->zc_cookie = newstate;
1918 spa_close(spa, FTAG);
1923 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1927 int replacing = zc->zc_cookie;
1928 int rebuild = zc->zc_simple;
1931 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1934 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1935 zc->zc_iflags, &config)) == 0) {
1936 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing,
1938 nvlist_free(config);
1941 spa_close(spa, FTAG);
1946 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1951 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1954 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
1956 spa_close(spa, FTAG);
1961 zfs_ioc_vdev_split(zfs_cmd_t *zc)
1964 nvlist_t *config, *props = NULL;
1966 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
1968 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1971 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1972 zc->zc_iflags, &config))) {
1973 spa_close(spa, FTAG);
1977 if (zc->zc_nvlist_src_size != 0 && (error =
1978 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1979 zc->zc_iflags, &props))) {
1980 spa_close(spa, FTAG);
1981 nvlist_free(config);
1985 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
1987 spa_close(spa, FTAG);
1989 nvlist_free(config);
1996 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
1999 const char *path = zc->zc_value;
2000 uint64_t guid = zc->zc_guid;
2003 error = spa_open(zc->zc_name, &spa, FTAG);
2007 error = spa_vdev_setpath(spa, guid, path);
2008 spa_close(spa, FTAG);
2013 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
2016 const char *fru = zc->zc_value;
2017 uint64_t guid = zc->zc_guid;
2020 error = spa_open(zc->zc_name, &spa, FTAG);
2024 error = spa_vdev_setfru(spa, guid, fru);
2025 spa_close(spa, FTAG);
2030 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
2035 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2037 if (zc->zc_nvlist_dst != 0 &&
2038 (error = dsl_prop_get_all(os, &nv)) == 0) {
2039 dmu_objset_stats(os, nv);
2041 * NB: zvol_get_stats() will read the objset contents,
2042 * which we aren't supposed to do with a
2043 * DS_MODE_USER hold, because it could be
2044 * inconsistent. So this is a bit of a workaround...
2045 * XXX reading without owning
2047 if (!zc->zc_objset_stats.dds_inconsistent &&
2048 dmu_objset_type(os) == DMU_OST_ZVOL) {
2049 error = zvol_get_stats(os, nv);
2057 error = put_nvlist(zc, nv);
2066 * zc_name name of filesystem
2067 * zc_nvlist_dst_size size of buffer for property nvlist
2070 * zc_objset_stats stats
2071 * zc_nvlist_dst property nvlist
2072 * zc_nvlist_dst_size size of property nvlist
2075 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2080 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2082 error = zfs_ioc_objset_stats_impl(zc, os);
2083 dmu_objset_rele(os, FTAG);
2091 * zc_name name of filesystem
2092 * zc_nvlist_dst_size size of buffer for property nvlist
2095 * zc_nvlist_dst received property nvlist
2096 * zc_nvlist_dst_size size of received property nvlist
2098 * Gets received properties (distinct from local properties on or after
2099 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2100 * local property values.
2103 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2109 * Without this check, we would return local property values if the
2110 * caller has not already received properties on or after
2111 * SPA_VERSION_RECVD_PROPS.
2113 if (!dsl_prop_get_hasrecvd(zc->zc_name))
2114 return (SET_ERROR(ENOTSUP));
2116 if (zc->zc_nvlist_dst != 0 &&
2117 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2118 error = put_nvlist(zc, nv);
2126 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2132 * zfs_get_zplprop() will either find a value or give us
2133 * the default value (if there is one).
2135 if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2137 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2143 * zc_name name of filesystem
2144 * zc_nvlist_dst_size size of buffer for zpl property nvlist
2147 * zc_nvlist_dst zpl property nvlist
2148 * zc_nvlist_dst_size size of zpl property nvlist
2151 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2156 /* XXX reading without owning */
2157 if ((err = dmu_objset_hold(zc->zc_name, FTAG, &os)))
2160 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2163 * NB: nvl_add_zplprop() will read the objset contents,
2164 * which we aren't supposed to do with a DS_MODE_USER
2165 * hold, because it could be inconsistent.
2167 if (zc->zc_nvlist_dst != 0 &&
2168 !zc->zc_objset_stats.dds_inconsistent &&
2169 dmu_objset_type(os) == DMU_OST_ZFS) {
2172 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2173 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2174 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2175 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2176 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2177 err = put_nvlist(zc, nv);
2180 err = SET_ERROR(ENOENT);
2182 dmu_objset_rele(os, FTAG);
2188 * zc_name name of filesystem
2189 * zc_cookie zap cursor
2190 * zc_nvlist_dst_size size of buffer for property nvlist
2193 * zc_name name of next filesystem
2194 * zc_cookie zap cursor
2195 * zc_objset_stats stats
2196 * zc_nvlist_dst property nvlist
2197 * zc_nvlist_dst_size size of property nvlist
2200 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2205 size_t orig_len = strlen(zc->zc_name);
2208 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os))) {
2209 if (error == ENOENT)
2210 error = SET_ERROR(ESRCH);
2214 p = strrchr(zc->zc_name, '/');
2215 if (p == NULL || p[1] != '\0')
2216 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2217 p = zc->zc_name + strlen(zc->zc_name);
2220 error = dmu_dir_list_next(os,
2221 sizeof (zc->zc_name) - (p - zc->zc_name), p,
2222 NULL, &zc->zc_cookie);
2223 if (error == ENOENT)
2224 error = SET_ERROR(ESRCH);
2225 } while (error == 0 && zfs_dataset_name_hidden(zc->zc_name));
2226 dmu_objset_rele(os, FTAG);
2229 * If it's an internal dataset (ie. with a '$' in its name),
2230 * don't try to get stats for it, otherwise we'll return ENOENT.
2232 if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2233 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2234 if (error == ENOENT) {
2235 /* We lost a race with destroy, get the next one. */
2236 zc->zc_name[orig_len] = '\0';
2245 * zc_name name of filesystem
2246 * zc_cookie zap cursor
2247 * zc_nvlist_src iteration range nvlist
2248 * zc_nvlist_src_size size of iteration range nvlist
2251 * zc_name name of next snapshot
2252 * zc_objset_stats stats
2253 * zc_nvlist_dst property nvlist
2254 * zc_nvlist_dst_size size of property nvlist
2257 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2260 objset_t *os, *ossnap;
2262 uint64_t min_txg = 0, max_txg = 0;
2264 if (zc->zc_nvlist_src_size != 0) {
2265 nvlist_t *props = NULL;
2266 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2267 zc->zc_iflags, &props);
2270 (void) nvlist_lookup_uint64(props, SNAP_ITER_MIN_TXG,
2272 (void) nvlist_lookup_uint64(props, SNAP_ITER_MAX_TXG,
2277 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2279 return (error == ENOENT ? SET_ERROR(ESRCH) : error);
2283 * A dataset name of maximum length cannot have any snapshots,
2284 * so exit immediately.
2286 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >=
2287 ZFS_MAX_DATASET_NAME_LEN) {
2288 dmu_objset_rele(os, FTAG);
2289 return (SET_ERROR(ESRCH));
2292 while (error == 0) {
2293 if (issig(JUSTLOOKING) && issig(FORREAL)) {
2294 error = SET_ERROR(EINTR);
2298 error = dmu_snapshot_list_next(os,
2299 sizeof (zc->zc_name) - strlen(zc->zc_name),
2300 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj,
2301 &zc->zc_cookie, NULL);
2302 if (error == ENOENT) {
2303 error = SET_ERROR(ESRCH);
2305 } else if (error != 0) {
2309 error = dsl_dataset_hold_obj(dmu_objset_pool(os), zc->zc_obj,
2314 if ((min_txg != 0 && dsl_get_creationtxg(ds) < min_txg) ||
2315 (max_txg != 0 && dsl_get_creationtxg(ds) > max_txg)) {
2316 dsl_dataset_rele(ds, FTAG);
2317 /* undo snapshot name append */
2318 *(strchr(zc->zc_name, '@') + 1) = '\0';
2323 if (zc->zc_simple) {
2324 dsl_dataset_rele(ds, FTAG);
2328 if ((error = dmu_objset_from_ds(ds, &ossnap)) != 0) {
2329 dsl_dataset_rele(ds, FTAG);
2332 if ((error = zfs_ioc_objset_stats_impl(zc, ossnap)) != 0) {
2333 dsl_dataset_rele(ds, FTAG);
2336 dsl_dataset_rele(ds, FTAG);
2340 dmu_objset_rele(os, FTAG);
2341 /* if we failed, undo the @ that we tacked on to zc_name */
2343 *strchr(zc->zc_name, '@') = '\0';
2348 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2350 const char *propname = nvpair_name(pair);
2352 unsigned int vallen;
2353 const char *dash, *domain;
2354 zfs_userquota_prop_t type;
2360 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2362 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2363 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2365 return (SET_ERROR(EINVAL));
2369 * A correctly constructed propname is encoded as
2370 * userquota@<rid>-<domain>.
2372 if ((dash = strchr(propname, '-')) == NULL ||
2373 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2375 return (SET_ERROR(EINVAL));
2382 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2384 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2385 zfsvfs_rele(zfsvfs, FTAG);
2392 * If the named property is one that has a special function to set its value,
2393 * return 0 on success and a positive error code on failure; otherwise if it is
2394 * not one of the special properties handled by this function, return -1.
2396 * XXX: It would be better for callers of the property interface if we handled
2397 * these special cases in dsl_prop.c (in the dsl layer).
2400 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2403 const char *propname = nvpair_name(pair);
2404 zfs_prop_t prop = zfs_name_to_prop(propname);
2405 uint64_t intval = 0;
2406 const char *strval = NULL;
2409 if (prop == ZPROP_USERPROP) {
2410 if (zfs_prop_userquota(propname))
2411 return (zfs_prop_set_userquota(dsname, pair));
2415 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2417 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2418 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2422 /* all special properties are numeric except for keylocation */
2423 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING) {
2424 strval = fnvpair_value_string(pair);
2426 intval = fnvpair_value_uint64(pair);
2430 case ZFS_PROP_QUOTA:
2431 err = dsl_dir_set_quota(dsname, source, intval);
2433 case ZFS_PROP_REFQUOTA:
2434 err = dsl_dataset_set_refquota(dsname, source, intval);
2436 case ZFS_PROP_FILESYSTEM_LIMIT:
2437 case ZFS_PROP_SNAPSHOT_LIMIT:
2438 if (intval == UINT64_MAX) {
2439 /* clearing the limit, just do it */
2442 err = dsl_dir_activate_fs_ss_limit(dsname);
2445 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2446 * default path to set the value in the nvlist.
2451 case ZFS_PROP_KEYLOCATION:
2452 err = dsl_crypto_can_set_keylocation(dsname, strval);
2455 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2456 * default path to set the value in the nvlist.
2461 case ZFS_PROP_RESERVATION:
2462 err = dsl_dir_set_reservation(dsname, source, intval);
2464 case ZFS_PROP_REFRESERVATION:
2465 err = dsl_dataset_set_refreservation(dsname, source, intval);
2467 case ZFS_PROP_COMPRESSION:
2468 err = dsl_dataset_set_compression(dsname, source, intval);
2470 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2471 * default path to set the value in the nvlist.
2476 case ZFS_PROP_VOLSIZE:
2477 err = zvol_set_volsize(dsname, intval);
2479 case ZFS_PROP_SNAPDEV:
2480 err = zvol_set_snapdev(dsname, source, intval);
2482 case ZFS_PROP_VOLMODE:
2483 err = zvol_set_volmode(dsname, source, intval);
2485 case ZFS_PROP_VERSION:
2489 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2492 err = zfs_set_version(zfsvfs, intval);
2493 zfsvfs_rele(zfsvfs, FTAG);
2495 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2498 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2499 (void) strlcpy(zc->zc_name, dsname,
2500 sizeof (zc->zc_name));
2501 (void) zfs_ioc_userspace_upgrade(zc);
2502 (void) zfs_ioc_id_quota_upgrade(zc);
2503 kmem_free(zc, sizeof (zfs_cmd_t));
2515 zfs_is_namespace_prop(zfs_prop_t prop)
2519 case ZFS_PROP_ATIME:
2520 case ZFS_PROP_RELATIME:
2521 case ZFS_PROP_DEVICES:
2523 case ZFS_PROP_SETUID:
2524 case ZFS_PROP_READONLY:
2525 case ZFS_PROP_XATTR:
2526 case ZFS_PROP_NBMAND:
2535 * This function is best effort. If it fails to set any of the given properties,
2536 * it continues to set as many as it can and returns the last error
2537 * encountered. If the caller provides a non-NULL errlist, it will be filled in
2538 * with the list of names of all the properties that failed along with the
2539 * corresponding error numbers.
2541 * If every property is set successfully, zero is returned and errlist is not
2545 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2554 boolean_t should_update_mount_cache = B_FALSE;
2556 nvlist_t *genericnvl = fnvlist_alloc();
2557 nvlist_t *retrynvl = fnvlist_alloc();
2560 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2561 const char *propname = nvpair_name(pair);
2562 zfs_prop_t prop = zfs_name_to_prop(propname);
2565 /* decode the property value */
2567 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2569 attrs = fnvpair_value_nvlist(pair);
2570 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2572 err = SET_ERROR(EINVAL);
2575 /* Validate value type */
2576 if (err == 0 && source == ZPROP_SRC_INHERITED) {
2577 /* inherited properties are expected to be booleans */
2578 if (nvpair_type(propval) != DATA_TYPE_BOOLEAN)
2579 err = SET_ERROR(EINVAL);
2580 } else if (err == 0 && prop == ZPROP_USERPROP) {
2581 if (zfs_prop_user(propname)) {
2582 if (nvpair_type(propval) != DATA_TYPE_STRING)
2583 err = SET_ERROR(EINVAL);
2584 } else if (zfs_prop_userquota(propname)) {
2585 if (nvpair_type(propval) !=
2586 DATA_TYPE_UINT64_ARRAY)
2587 err = SET_ERROR(EINVAL);
2589 err = SET_ERROR(EINVAL);
2591 } else if (err == 0) {
2592 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2593 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2594 err = SET_ERROR(EINVAL);
2595 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2598 intval = fnvpair_value_uint64(propval);
2600 switch (zfs_prop_get_type(prop)) {
2601 case PROP_TYPE_NUMBER:
2603 case PROP_TYPE_STRING:
2604 err = SET_ERROR(EINVAL);
2606 case PROP_TYPE_INDEX:
2607 if (zfs_prop_index_to_string(prop,
2608 intval, &unused) != 0)
2610 SET_ERROR(ZFS_ERR_BADPROP);
2614 "unknown property type");
2617 err = SET_ERROR(EINVAL);
2621 /* Validate permissions */
2623 err = zfs_check_settable(dsname, pair, CRED());
2626 if (source == ZPROP_SRC_INHERITED)
2627 err = -1; /* does not need special handling */
2629 err = zfs_prop_set_special(dsname, source,
2633 * For better performance we build up a list of
2634 * properties to set in a single transaction.
2636 err = nvlist_add_nvpair(genericnvl, pair);
2637 } else if (err != 0 && nvl != retrynvl) {
2639 * This may be a spurious error caused by
2640 * receiving quota and reservation out of order.
2641 * Try again in a second pass.
2643 err = nvlist_add_nvpair(retrynvl, pair);
2648 if (errlist != NULL)
2649 fnvlist_add_int32(errlist, propname, err);
2653 if (zfs_is_namespace_prop(prop))
2654 should_update_mount_cache = B_TRUE;
2657 if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2662 if (nvlist_empty(genericnvl))
2666 * Try to set them all in one batch.
2668 err = dsl_props_set(dsname, source, genericnvl);
2673 * If batching fails, we still want to set as many properties as we
2674 * can, so try setting them individually.
2677 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2678 const char *propname = nvpair_name(pair);
2682 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2684 attrs = fnvpair_value_nvlist(pair);
2685 propval = fnvlist_lookup_nvpair(attrs, ZPROP_VALUE);
2688 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2689 strval = fnvpair_value_string(propval);
2690 err = dsl_prop_set_string(dsname, propname,
2692 } else if (nvpair_type(propval) == DATA_TYPE_BOOLEAN) {
2693 err = dsl_prop_inherit(dsname, propname, source);
2695 intval = fnvpair_value_uint64(propval);
2696 err = dsl_prop_set_int(dsname, propname, source,
2701 if (errlist != NULL) {
2702 fnvlist_add_int32(errlist, propname, err);
2709 if (should_update_mount_cache)
2710 zfs_ioctl_update_mount_cache(dsname);
2712 nvlist_free(genericnvl);
2713 nvlist_free(retrynvl);
2719 * Check that all the properties are valid user properties.
2722 zfs_check_userprops(nvlist_t *nvl)
2724 nvpair_t *pair = NULL;
2726 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2727 const char *propname = nvpair_name(pair);
2729 if (!zfs_prop_user(propname) ||
2730 nvpair_type(pair) != DATA_TYPE_STRING)
2731 return (SET_ERROR(EINVAL));
2733 if (strlen(propname) >= ZAP_MAXNAMELEN)
2734 return (SET_ERROR(ENAMETOOLONG));
2736 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN)
2737 return (SET_ERROR(E2BIG));
2743 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2747 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2750 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2751 if (nvlist_exists(skipped, nvpair_name(pair)))
2754 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2759 clear_received_props(const char *dsname, nvlist_t *props,
2763 nvlist_t *cleared_props = NULL;
2764 props_skip(props, skipped, &cleared_props);
2765 if (!nvlist_empty(cleared_props)) {
2767 * Acts on local properties until the dataset has received
2768 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2770 zprop_source_t flags = (ZPROP_SRC_NONE |
2771 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2772 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2774 nvlist_free(cleared_props);
2780 * zc_name name of filesystem
2781 * zc_value name of property to set
2782 * zc_nvlist_src{_size} nvlist of properties to apply
2783 * zc_cookie received properties flag
2786 * zc_nvlist_dst{_size} error for each unapplied received property
2789 zfs_ioc_set_prop(zfs_cmd_t *zc)
2792 boolean_t received = zc->zc_cookie;
2793 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2798 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2799 zc->zc_iflags, &nvl)) != 0)
2803 nvlist_t *origprops;
2805 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2806 (void) clear_received_props(zc->zc_name,
2808 nvlist_free(origprops);
2811 error = dsl_prop_set_hasrecvd(zc->zc_name);
2814 errors = fnvlist_alloc();
2816 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2818 if (zc->zc_nvlist_dst != 0 && errors != NULL) {
2819 (void) put_nvlist(zc, errors);
2822 nvlist_free(errors);
2829 * zc_name name of filesystem
2830 * zc_value name of property to inherit
2831 * zc_cookie revert to received value if TRUE
2836 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2838 const char *propname = zc->zc_value;
2839 zfs_prop_t prop = zfs_name_to_prop(propname);
2840 boolean_t received = zc->zc_cookie;
2841 zprop_source_t source = (received
2842 ? ZPROP_SRC_NONE /* revert to received value, if any */
2843 : ZPROP_SRC_INHERITED); /* explicitly inherit */
2851 * Only check this in the non-received case. We want to allow
2852 * 'inherit -S' to revert non-inheritable properties like quota
2853 * and reservation to the received or default values even though
2854 * they are not considered inheritable.
2856 if (prop != ZPROP_USERPROP && !zfs_prop_inheritable(prop))
2857 return (SET_ERROR(EINVAL));
2860 if (prop == ZPROP_USERPROP) {
2861 if (!zfs_prop_user(propname))
2862 return (SET_ERROR(EINVAL));
2864 type = PROP_TYPE_STRING;
2865 } else if (prop == ZFS_PROP_VOLSIZE || prop == ZFS_PROP_VERSION) {
2866 return (SET_ERROR(EINVAL));
2868 type = zfs_prop_get_type(prop);
2872 * zfs_prop_set_special() expects properties in the form of an
2873 * nvpair with type info.
2875 dummy = fnvlist_alloc();
2878 case PROP_TYPE_STRING:
2879 VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2881 case PROP_TYPE_NUMBER:
2882 case PROP_TYPE_INDEX:
2883 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2886 err = SET_ERROR(EINVAL);
2890 pair = nvlist_next_nvpair(dummy, NULL);
2892 err = SET_ERROR(EINVAL);
2894 err = zfs_prop_set_special(zc->zc_name, source, pair);
2895 if (err == -1) /* property is not "special", needs handling */
2896 err = dsl_prop_inherit(zc->zc_name, zc->zc_value,
2906 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2913 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2914 zc->zc_iflags, &props)))
2918 * If the only property is the configfile, then just do a spa_lookup()
2919 * to handle the faulted case.
2921 pair = nvlist_next_nvpair(props, NULL);
2922 if (pair != NULL && strcmp(nvpair_name(pair),
2923 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2924 nvlist_next_nvpair(props, pair) == NULL) {
2925 mutex_enter(&spa_namespace_lock);
2926 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2927 spa_configfile_set(spa, props, B_FALSE);
2928 spa_write_cachefile(spa, B_FALSE, B_TRUE);
2930 mutex_exit(&spa_namespace_lock);
2937 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2942 error = spa_prop_set(spa, props);
2945 spa_close(spa, FTAG);
2951 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2955 nvlist_t *nvp = NULL;
2957 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2959 * If the pool is faulted, there may be properties we can still
2960 * get (such as altroot and cachefile), so attempt to get them
2963 mutex_enter(&spa_namespace_lock);
2964 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2965 error = spa_prop_get(spa, &nvp);
2966 mutex_exit(&spa_namespace_lock);
2968 error = spa_prop_get(spa, &nvp);
2969 spa_close(spa, FTAG);
2972 if (error == 0 && zc->zc_nvlist_dst != 0)
2973 error = put_nvlist(zc, nvp);
2975 error = SET_ERROR(EFAULT);
2983 * "vdevprops_set_vdev" -> guid
2984 * "vdevprops_set_props" -> { prop -> value }
2987 * outnvl: propname -> error code (int32)
2989 static const zfs_ioc_key_t zfs_keys_vdev_set_props[] = {
2990 {ZPOOL_VDEV_PROPS_SET_VDEV, DATA_TYPE_UINT64, 0},
2991 {ZPOOL_VDEV_PROPS_SET_PROPS, DATA_TYPE_NVLIST, 0}
2995 zfs_ioc_vdev_set_props(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3002 /* Early validation */
3003 if (nvlist_lookup_uint64(innvl, ZPOOL_VDEV_PROPS_SET_VDEV,
3005 return (SET_ERROR(EINVAL));
3008 return (SET_ERROR(EINVAL));
3010 if ((error = spa_open(poolname, &spa, FTAG)) != 0)
3013 ASSERT(spa_writeable(spa));
3015 if ((vd = spa_lookup_by_guid(spa, vdev_guid, B_TRUE)) == NULL) {
3016 spa_close(spa, FTAG);
3017 return (SET_ERROR(ENOENT));
3020 error = vdev_prop_set(vd, innvl, outnvl);
3022 spa_close(spa, FTAG);
3029 * "vdevprops_get_vdev" -> guid
3030 * (optional) "vdevprops_get_props" -> { propname -> propid }
3033 * outnvl: propname -> value
3035 static const zfs_ioc_key_t zfs_keys_vdev_get_props[] = {
3036 {ZPOOL_VDEV_PROPS_GET_VDEV, DATA_TYPE_UINT64, 0},
3037 {ZPOOL_VDEV_PROPS_GET_PROPS, DATA_TYPE_NVLIST, ZK_OPTIONAL}
3041 zfs_ioc_vdev_get_props(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3048 /* Early validation */
3049 if (nvlist_lookup_uint64(innvl, ZPOOL_VDEV_PROPS_GET_VDEV,
3051 return (SET_ERROR(EINVAL));
3054 return (SET_ERROR(EINVAL));
3056 if ((error = spa_open(poolname, &spa, FTAG)) != 0)
3059 if ((vd = spa_lookup_by_guid(spa, vdev_guid, B_TRUE)) == NULL) {
3060 spa_close(spa, FTAG);
3061 return (SET_ERROR(ENOENT));
3064 error = vdev_prop_get(vd, innvl, outnvl);
3066 spa_close(spa, FTAG);
3073 * zc_name name of filesystem
3074 * zc_nvlist_src{_size} nvlist of delegated permissions
3075 * zc_perm_action allow/unallow flag
3080 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
3083 nvlist_t *fsaclnv = NULL;
3085 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
3086 zc->zc_iflags, &fsaclnv)) != 0)
3090 * Verify nvlist is constructed correctly
3092 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
3093 nvlist_free(fsaclnv);
3094 return (SET_ERROR(EINVAL));
3098 * If we don't have PRIV_SYS_MOUNT, then validate
3099 * that user is allowed to hand out each permission in
3103 error = secpolicy_zfs(CRED());
3105 if (zc->zc_perm_action == B_FALSE) {
3106 error = dsl_deleg_can_allow(zc->zc_name,
3109 error = dsl_deleg_can_unallow(zc->zc_name,
3115 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
3117 nvlist_free(fsaclnv);
3123 * zc_name name of filesystem
3126 * zc_nvlist_src{_size} nvlist of delegated permissions
3129 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
3134 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
3135 error = put_nvlist(zc, nvp);
3143 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
3145 zfs_creat_t *zct = arg;
3147 zfs_create_fs(os, cr, zct->zct_zplprops, tx);
3150 #define ZFS_PROP_UNDEFINED ((uint64_t)-1)
3154 * os parent objset pointer (NULL if root fs)
3155 * fuids_ok fuids allowed in this version of the spa?
3156 * sa_ok SAs allowed in this version of the spa?
3157 * createprops list of properties requested by creator
3160 * zplprops values for the zplprops we attach to the master node object
3161 * is_ci true if requested file system will be purely case-insensitive
3163 * Determine the settings for utf8only, normalization and
3164 * casesensitivity. Specific values may have been requested by the
3165 * creator and/or we can inherit values from the parent dataset. If
3166 * the file system is of too early a vintage, a creator can not
3167 * request settings for these properties, even if the requested
3168 * setting is the default value. We don't actually want to create dsl
3169 * properties for these, so remove them from the source nvlist after
3173 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
3174 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
3175 nvlist_t *zplprops, boolean_t *is_ci)
3177 uint64_t sense = ZFS_PROP_UNDEFINED;
3178 uint64_t norm = ZFS_PROP_UNDEFINED;
3179 uint64_t u8 = ZFS_PROP_UNDEFINED;
3182 ASSERT(zplprops != NULL);
3184 /* parent dataset must be a filesystem */
3185 if (os != NULL && os->os_phys->os_type != DMU_OST_ZFS)
3186 return (SET_ERROR(ZFS_ERR_WRONG_PARENT));
3189 * Pull out creator prop choices, if any.
3192 (void) nvlist_lookup_uint64(createprops,
3193 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
3194 (void) nvlist_lookup_uint64(createprops,
3195 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
3196 (void) nvlist_remove_all(createprops,
3197 zfs_prop_to_name(ZFS_PROP_NORMALIZE));
3198 (void) nvlist_lookup_uint64(createprops,
3199 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
3200 (void) nvlist_remove_all(createprops,
3201 zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
3202 (void) nvlist_lookup_uint64(createprops,
3203 zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3204 (void) nvlist_remove_all(createprops,
3205 zfs_prop_to_name(ZFS_PROP_CASE));
3209 * If the zpl version requested is whacky or the file system
3210 * or pool is version is too "young" to support normalization
3211 * and the creator tried to set a value for one of the props,
3214 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3215 (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3216 (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3217 (zplver < ZPL_VERSION_NORMALIZATION &&
3218 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3219 sense != ZFS_PROP_UNDEFINED)))
3220 return (SET_ERROR(ENOTSUP));
3223 * Put the version in the zplprops
3225 VERIFY(nvlist_add_uint64(zplprops,
3226 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3228 if (norm == ZFS_PROP_UNDEFINED &&
3229 (error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm)) != 0)
3231 VERIFY(nvlist_add_uint64(zplprops,
3232 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3235 * If we're normalizing, names must always be valid UTF-8 strings.
3239 if (u8 == ZFS_PROP_UNDEFINED &&
3240 (error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8)) != 0)
3242 VERIFY(nvlist_add_uint64(zplprops,
3243 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3245 if (sense == ZFS_PROP_UNDEFINED &&
3246 (error = zfs_get_zplprop(os, ZFS_PROP_CASE, &sense)) != 0)
3248 VERIFY(nvlist_add_uint64(zplprops,
3249 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3252 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
3258 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3259 nvlist_t *zplprops, boolean_t *is_ci)
3261 boolean_t fuids_ok, sa_ok;
3262 uint64_t zplver = ZPL_VERSION;
3263 objset_t *os = NULL;
3264 char parentname[ZFS_MAX_DATASET_NAME_LEN];
3269 zfs_get_parent(dataset, parentname, sizeof (parentname));
3271 if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3274 spa_vers = spa_version(spa);
3275 spa_close(spa, FTAG);
3277 zplver = zfs_zpl_version_map(spa_vers);
3278 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3279 sa_ok = (zplver >= ZPL_VERSION_SA);
3282 * Open parent object set so we can inherit zplprop values.
3284 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3287 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3289 dmu_objset_rele(os, FTAG);
3294 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3295 nvlist_t *zplprops, boolean_t *is_ci)
3299 uint64_t zplver = ZPL_VERSION;
3302 zplver = zfs_zpl_version_map(spa_vers);
3303 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3304 sa_ok = (zplver >= ZPL_VERSION_SA);
3306 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3307 createprops, zplprops, is_ci);
3313 * "type" -> dmu_objset_type_t (int32)
3314 * (optional) "props" -> { prop -> value }
3315 * (optional) "hidden_args" -> { "wkeydata" -> value }
3316 * raw uint8_t array of encryption wrapping key data (32 bytes)
3319 * outnvl: propname -> error code (int32)
3322 static const zfs_ioc_key_t zfs_keys_create[] = {
3323 {"type", DATA_TYPE_INT32, 0},
3324 {"props", DATA_TYPE_NVLIST, ZK_OPTIONAL},
3325 {"hidden_args", DATA_TYPE_NVLIST, ZK_OPTIONAL},
3329 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3332 zfs_creat_t zct = { 0 };
3333 nvlist_t *nvprops = NULL;
3334 nvlist_t *hidden_args = NULL;
3335 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3336 dmu_objset_type_t type;
3337 boolean_t is_insensitive = B_FALSE;
3338 dsl_crypto_params_t *dcp = NULL;
3340 type = (dmu_objset_type_t)fnvlist_lookup_int32(innvl, "type");
3341 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3342 (void) nvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS, &hidden_args);
3346 cbfunc = zfs_create_cb;
3350 cbfunc = zvol_create_cb;
3357 if (strchr(fsname, '@') ||
3358 strchr(fsname, '%'))
3359 return (SET_ERROR(EINVAL));
3361 zct.zct_props = nvprops;
3364 return (SET_ERROR(EINVAL));
3366 if (type == DMU_OST_ZVOL) {
3367 uint64_t volsize, volblocksize;
3369 if (nvprops == NULL)
3370 return (SET_ERROR(EINVAL));
3371 if (nvlist_lookup_uint64(nvprops,
3372 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3373 return (SET_ERROR(EINVAL));
3375 if ((error = nvlist_lookup_uint64(nvprops,
3376 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3377 &volblocksize)) != 0 && error != ENOENT)
3378 return (SET_ERROR(EINVAL));
3381 volblocksize = zfs_prop_default_numeric(
3382 ZFS_PROP_VOLBLOCKSIZE);
3384 if ((error = zvol_check_volblocksize(fsname,
3385 volblocksize)) != 0 ||
3386 (error = zvol_check_volsize(volsize,
3387 volblocksize)) != 0)
3389 } else if (type == DMU_OST_ZFS) {
3393 * We have to have normalization and
3394 * case-folding flags correct when we do the
3395 * file system creation, so go figure them out
3398 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3399 NV_UNIQUE_NAME, KM_SLEEP) == 0);
3400 error = zfs_fill_zplprops(fsname, nvprops,
3401 zct.zct_zplprops, &is_insensitive);
3403 nvlist_free(zct.zct_zplprops);
3408 error = dsl_crypto_params_create_nvlist(DCP_CMD_NONE, nvprops,
3411 nvlist_free(zct.zct_zplprops);
3415 error = dmu_objset_create(fsname, type,
3416 is_insensitive ? DS_FLAG_CI_DATASET : 0, dcp, cbfunc, &zct);
3418 nvlist_free(zct.zct_zplprops);
3419 dsl_crypto_params_free(dcp, !!error);
3422 * It would be nice to do this atomically.
3425 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3432 * Volumes will return EBUSY and cannot be destroyed
3433 * until all asynchronous minor handling (e.g. from
3434 * setting the volmode property) has completed. Wait for
3435 * the spa_zvol_taskq to drain then retry.
3437 error2 = dsl_destroy_head(fsname);
3438 while ((error2 == EBUSY) && (type == DMU_OST_ZVOL)) {
3439 error2 = spa_open(fsname, &spa, FTAG);
3441 taskq_wait(spa->spa_zvol_taskq);
3442 spa_close(spa, FTAG);
3444 error2 = dsl_destroy_head(fsname);
3453 * "origin" -> name of origin snapshot
3454 * (optional) "props" -> { prop -> value }
3455 * (optional) "hidden_args" -> { "wkeydata" -> value }
3456 * raw uint8_t array of encryption wrapping key data (32 bytes)
3460 * outnvl: propname -> error code (int32)
3462 static const zfs_ioc_key_t zfs_keys_clone[] = {
3463 {"origin", DATA_TYPE_STRING, 0},
3464 {"props", DATA_TYPE_NVLIST, ZK_OPTIONAL},
3465 {"hidden_args", DATA_TYPE_NVLIST, ZK_OPTIONAL},
3469 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3472 nvlist_t *nvprops = NULL;
3473 const char *origin_name;
3475 origin_name = fnvlist_lookup_string(innvl, "origin");
3476 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3478 if (strchr(fsname, '@') ||
3479 strchr(fsname, '%'))
3480 return (SET_ERROR(EINVAL));
3482 if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3483 return (SET_ERROR(EINVAL));
3485 error = dmu_objset_clone(fsname, origin_name);
3488 * It would be nice to do this atomically.
3491 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3494 (void) dsl_destroy_head(fsname);
3499 static const zfs_ioc_key_t zfs_keys_remap[] = {
3504 zfs_ioc_remap(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3506 /* This IOCTL is no longer supported. */
3507 (void) fsname, (void) innvl, (void) outnvl;
3513 * "snaps" -> { snapshot1, snapshot2 }
3514 * (optional) "props" -> { prop -> value (string) }
3517 * outnvl: snapshot -> error code (int32)
3519 static const zfs_ioc_key_t zfs_keys_snapshot[] = {
3520 {"snaps", DATA_TYPE_NVLIST, 0},
3521 {"props", DATA_TYPE_NVLIST, ZK_OPTIONAL},
3525 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3528 nvlist_t *props = NULL;
3532 (void) nvlist_lookup_nvlist(innvl, "props", &props);
3533 if (!nvlist_empty(props) &&
3534 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3535 return (SET_ERROR(ENOTSUP));
3536 if ((error = zfs_check_userprops(props)) != 0)
3539 snaps = fnvlist_lookup_nvlist(innvl, "snaps");
3540 poollen = strlen(poolname);
3541 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3542 pair = nvlist_next_nvpair(snaps, pair)) {
3543 const char *name = nvpair_name(pair);
3544 char *cp = strchr(name, '@');
3547 * The snap name must contain an @, and the part after it must
3548 * contain only valid characters.
3551 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3552 return (SET_ERROR(EINVAL));
3555 * The snap must be in the specified pool.
3557 if (strncmp(name, poolname, poollen) != 0 ||
3558 (name[poollen] != '/' && name[poollen] != '@'))
3559 return (SET_ERROR(EXDEV));
3562 * Check for permission to set the properties on the fs.
3564 if (!nvlist_empty(props)) {
3566 error = zfs_secpolicy_write_perms(name,
3567 ZFS_DELEG_PERM_USERPROP, CRED());
3573 /* This must be the only snap of this fs. */
3574 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3575 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3576 if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3578 return (SET_ERROR(EXDEV));
3583 error = dsl_dataset_snapshot(snaps, props, outnvl);
3589 * innvl: "message" -> string
3591 static const zfs_ioc_key_t zfs_keys_log_history[] = {
3592 {"message", DATA_TYPE_STRING, 0},
3596 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3598 (void) unused, (void) outnvl;
3599 const char *message;
3605 * The poolname in the ioctl is not set, we get it from the TSD,
3606 * which was set at the end of the last successful ioctl that allows
3607 * logging. The secpolicy func already checked that it is set.
3608 * Only one log ioctl is allowed after each successful ioctl, so
3609 * we clear the TSD here.
3611 poolname = tsd_get(zfs_allow_log_key);
3612 if (poolname == NULL)
3613 return (SET_ERROR(EINVAL));
3614 (void) tsd_set(zfs_allow_log_key, NULL);
3615 error = spa_open(poolname, &spa, FTAG);
3616 kmem_strfree(poolname);
3620 message = fnvlist_lookup_string(innvl, "message");
3622 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3623 spa_close(spa, FTAG);
3624 return (SET_ERROR(ENOTSUP));
3627 error = spa_history_log(spa, message);
3628 spa_close(spa, FTAG);
3633 * This ioctl is used to set the bootenv configuration on the current
3634 * pool. This configuration is stored in the second padding area of the label,
3635 * and it is used by the bootloader(s) to store the bootloader and/or system
3637 * The data is stored as nvlist data stream, and is protected by
3638 * an embedded checksum.
3639 * The version can have two possible values:
3640 * VB_RAW: nvlist should have key GRUB_ENVMAP, value DATA_TYPE_STRING.
3641 * VB_NVLIST: nvlist with arbitrary <key, value> pairs.
3643 static const zfs_ioc_key_t zfs_keys_set_bootenv[] = {
3644 {"version", DATA_TYPE_UINT64, 0},
3645 {"<keys>", DATA_TYPE_ANY, ZK_OPTIONAL | ZK_WILDCARDLIST},
3649 zfs_ioc_set_bootenv(const char *name, nvlist_t *innvl, nvlist_t *outnvl)
3654 if ((error = spa_open(name, &spa, FTAG)) != 0)
3656 spa_vdev_state_enter(spa, SCL_ALL);
3657 error = vdev_label_write_bootenv(spa->spa_root_vdev, innvl);
3658 (void) spa_vdev_state_exit(spa, NULL, 0);
3659 spa_close(spa, FTAG);
3663 static const zfs_ioc_key_t zfs_keys_get_bootenv[] = {
3668 zfs_ioc_get_bootenv(const char *name, nvlist_t *innvl, nvlist_t *outnvl)
3673 if ((error = spa_open(name, &spa, FTAG)) != 0)
3675 spa_vdev_state_enter(spa, SCL_ALL);
3676 error = vdev_label_read_bootenv(spa->spa_root_vdev, outnvl);
3677 (void) spa_vdev_state_exit(spa, NULL, 0);
3678 spa_close(spa, FTAG);
3683 * The dp_config_rwlock must not be held when calling this, because the
3684 * unmount may need to write out data.
3686 * This function is best-effort. Callers must deal gracefully if it
3687 * remains mounted (or is remounted after this call).
3689 * Returns 0 if the argument is not a snapshot, or it is not currently a
3690 * filesystem, or we were able to unmount it. Returns error code otherwise.
3693 zfs_unmount_snap(const char *snapname)
3695 if (strchr(snapname, '@') == NULL)
3698 (void) zfsctl_snapshot_unmount(snapname, MNT_FORCE);
3702 zfs_unmount_snap_cb(const char *snapname, void *arg)
3705 zfs_unmount_snap(snapname);
3710 * When a clone is destroyed, its origin may also need to be destroyed,
3711 * in which case it must be unmounted. This routine will do that unmount
3715 zfs_destroy_unmount_origin(const char *fsname)
3721 error = dmu_objset_hold(fsname, FTAG, &os);
3724 ds = dmu_objset_ds(os);
3725 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3726 char originname[ZFS_MAX_DATASET_NAME_LEN];
3727 dsl_dataset_name(ds->ds_prev, originname);
3728 dmu_objset_rele(os, FTAG);
3729 zfs_unmount_snap(originname);
3731 dmu_objset_rele(os, FTAG);
3737 * "snaps" -> { snapshot1, snapshot2 }
3738 * (optional boolean) "defer"
3741 * outnvl: snapshot -> error code (int32)
3743 static const zfs_ioc_key_t zfs_keys_destroy_snaps[] = {
3744 {"snaps", DATA_TYPE_NVLIST, 0},
3745 {"defer", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
3749 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3757 snaps = fnvlist_lookup_nvlist(innvl, "snaps");
3758 defer = nvlist_exists(innvl, "defer");
3760 poollen = strlen(poolname);
3761 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3762 pair = nvlist_next_nvpair(snaps, pair)) {
3763 const char *name = nvpair_name(pair);
3766 * The snap must be in the specified pool to prevent the
3767 * invalid removal of zvol minors below.
3769 if (strncmp(name, poolname, poollen) != 0 ||
3770 (name[poollen] != '/' && name[poollen] != '@'))
3771 return (SET_ERROR(EXDEV));
3773 zfs_unmount_snap(nvpair_name(pair));
3774 if (spa_open(name, &spa, FTAG) == 0) {
3775 zvol_remove_minors(spa, name, B_TRUE);
3776 spa_close(spa, FTAG);
3780 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3784 * Create bookmarks. The bookmark names are of the form <fs>#<bmark>.
3785 * All bookmarks and snapshots must be in the same pool.
3786 * dsl_bookmark_create_nvl_validate describes the nvlist schema in more detail.
3789 * new_bookmark1 -> existing_snapshot,
3790 * new_bookmark2 -> existing_bookmark,
3793 * outnvl: bookmark -> error code (int32)
3796 static const zfs_ioc_key_t zfs_keys_bookmark[] = {
3797 {"<bookmark>...", DATA_TYPE_STRING, ZK_WILDCARDLIST},
3801 zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3804 return (dsl_bookmark_create(innvl, outnvl));
3809 * property 1, property 2, ...
3813 * bookmark name 1 -> { property 1, property 2, ... },
3814 * bookmark name 2 -> { property 1, property 2, ... }
3818 static const zfs_ioc_key_t zfs_keys_get_bookmarks[] = {
3819 {"<property>...", DATA_TYPE_BOOLEAN, ZK_WILDCARDLIST | ZK_OPTIONAL},
3823 zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3825 return (dsl_get_bookmarks(fsname, innvl, outnvl));
3829 * innvl is not used.
3832 * property 1, property 2, ...
3836 static const zfs_ioc_key_t zfs_keys_get_bookmark_props[] = {
3841 zfs_ioc_get_bookmark_props(const char *bookmark, nvlist_t *innvl,
3845 char fsname[ZFS_MAX_DATASET_NAME_LEN];
3848 bmname = strchr(bookmark, '#');
3850 return (SET_ERROR(EINVAL));
3853 (void) strlcpy(fsname, bookmark, sizeof (fsname));
3854 *(strchr(fsname, '#')) = '\0';
3856 return (dsl_get_bookmark_props(fsname, bmname, outnvl));
3861 * bookmark name 1, bookmark name 2
3864 * outnvl: bookmark -> error code (int32)
3867 static const zfs_ioc_key_t zfs_keys_destroy_bookmarks[] = {
3868 {"<bookmark>...", DATA_TYPE_BOOLEAN, ZK_WILDCARDLIST},
3872 zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl,
3877 poollen = strlen(poolname);
3878 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3879 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3880 const char *name = nvpair_name(pair);
3881 const char *cp = strchr(name, '#');
3884 * The bookmark name must contain an #, and the part after it
3885 * must contain only valid characters.
3888 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3889 return (SET_ERROR(EINVAL));
3892 * The bookmark must be in the specified pool.
3894 if (strncmp(name, poolname, poollen) != 0 ||
3895 (name[poollen] != '/' && name[poollen] != '#'))
3896 return (SET_ERROR(EXDEV));
3899 error = dsl_bookmark_destroy(innvl, outnvl);
3903 static const zfs_ioc_key_t zfs_keys_channel_program[] = {
3904 {"program", DATA_TYPE_STRING, 0},
3905 {"arg", DATA_TYPE_ANY, 0},
3906 {"sync", DATA_TYPE_BOOLEAN_VALUE, ZK_OPTIONAL},
3907 {"instrlimit", DATA_TYPE_UINT64, ZK_OPTIONAL},
3908 {"memlimit", DATA_TYPE_UINT64, ZK_OPTIONAL},
3912 zfs_ioc_channel_program(const char *poolname, nvlist_t *innvl,
3916 uint64_t instrlimit, memlimit;
3917 boolean_t sync_flag;
3918 nvpair_t *nvarg = NULL;
3920 program = fnvlist_lookup_string(innvl, ZCP_ARG_PROGRAM);
3921 if (0 != nvlist_lookup_boolean_value(innvl, ZCP_ARG_SYNC, &sync_flag)) {
3924 if (0 != nvlist_lookup_uint64(innvl, ZCP_ARG_INSTRLIMIT, &instrlimit)) {
3925 instrlimit = ZCP_DEFAULT_INSTRLIMIT;
3927 if (0 != nvlist_lookup_uint64(innvl, ZCP_ARG_MEMLIMIT, &memlimit)) {
3928 memlimit = ZCP_DEFAULT_MEMLIMIT;
3930 nvarg = fnvlist_lookup_nvpair(innvl, ZCP_ARG_ARGLIST);
3932 if (instrlimit == 0 || instrlimit > zfs_lua_max_instrlimit)
3933 return (SET_ERROR(EINVAL));
3934 if (memlimit == 0 || memlimit > zfs_lua_max_memlimit)
3935 return (SET_ERROR(EINVAL));
3937 return (zcp_eval(poolname, program, sync_flag, instrlimit, memlimit,
3945 static const zfs_ioc_key_t zfs_keys_pool_checkpoint[] = {
3950 zfs_ioc_pool_checkpoint(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3952 (void) innvl, (void) outnvl;
3953 return (spa_checkpoint(poolname));
3960 static const zfs_ioc_key_t zfs_keys_pool_discard_checkpoint[] = {
3965 zfs_ioc_pool_discard_checkpoint(const char *poolname, nvlist_t *innvl,
3968 (void) innvl, (void) outnvl;
3969 return (spa_checkpoint_discard(poolname));
3974 * zc_name name of dataset to destroy
3975 * zc_defer_destroy mark for deferred destroy
3980 zfs_ioc_destroy(zfs_cmd_t *zc)
3983 dmu_objset_type_t ost;
3986 err = dmu_objset_hold(zc->zc_name, FTAG, &os);
3989 ost = dmu_objset_type(os);
3990 dmu_objset_rele(os, FTAG);
3992 if (ost == DMU_OST_ZFS)
3993 zfs_unmount_snap(zc->zc_name);
3995 if (strchr(zc->zc_name, '@')) {
3996 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3998 err = dsl_destroy_head(zc->zc_name);
3999 if (err == EEXIST) {
4001 * It is possible that the given DS may have
4002 * hidden child (%recv) datasets - "leftovers"
4003 * resulting from the previously interrupted
4006 * 6 extra bytes for /%recv
4008 char namebuf[ZFS_MAX_DATASET_NAME_LEN + 6];
4010 if (snprintf(namebuf, sizeof (namebuf), "%s/%s",
4011 zc->zc_name, recv_clone_name) >=
4013 return (SET_ERROR(EINVAL));
4016 * Try to remove the hidden child (%recv) and after
4017 * that try to remove the target dataset.
4018 * If the hidden child (%recv) does not exist
4019 * the original error (EEXIST) will be returned
4021 err = dsl_destroy_head(namebuf);
4023 err = dsl_destroy_head(zc->zc_name);
4024 else if (err == ENOENT)
4025 err = SET_ERROR(EEXIST);
4034 * "initialize_command" -> POOL_INITIALIZE_{CANCEL|START|SUSPEND} (uint64)
4035 * "initialize_vdevs": { -> guids to initialize (nvlist)
4036 * "vdev_path_1": vdev_guid_1, (uint64),
4037 * "vdev_path_2": vdev_guid_2, (uint64),
4043 * "initialize_vdevs": { -> initialization errors (nvlist)
4044 * "vdev_path_1": errno, see function body for possible errnos (uint64)
4045 * "vdev_path_2": errno, ... (uint64)
4050 * EINVAL is returned for an unknown commands or if any of the provided vdev
4051 * guids have be specified with a type other than uint64.
4053 static const zfs_ioc_key_t zfs_keys_pool_initialize[] = {
4054 {ZPOOL_INITIALIZE_COMMAND, DATA_TYPE_UINT64, 0},
4055 {ZPOOL_INITIALIZE_VDEVS, DATA_TYPE_NVLIST, 0}
4059 zfs_ioc_pool_initialize(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
4062 if (nvlist_lookup_uint64(innvl, ZPOOL_INITIALIZE_COMMAND,
4064 return (SET_ERROR(EINVAL));
4067 if (!(cmd_type == POOL_INITIALIZE_CANCEL ||
4068 cmd_type == POOL_INITIALIZE_START ||
4069 cmd_type == POOL_INITIALIZE_SUSPEND)) {
4070 return (SET_ERROR(EINVAL));
4073 nvlist_t *vdev_guids;
4074 if (nvlist_lookup_nvlist(innvl, ZPOOL_INITIALIZE_VDEVS,
4075 &vdev_guids) != 0) {
4076 return (SET_ERROR(EINVAL));
4079 for (nvpair_t *pair = nvlist_next_nvpair(vdev_guids, NULL);
4080 pair != NULL; pair = nvlist_next_nvpair(vdev_guids, pair)) {
4082 if (nvpair_value_uint64(pair, &vdev_guid) != 0) {
4083 return (SET_ERROR(EINVAL));
4088 int error = spa_open(poolname, &spa, FTAG);
4092 nvlist_t *vdev_errlist = fnvlist_alloc();
4093 int total_errors = spa_vdev_initialize(spa, vdev_guids, cmd_type,
4096 if (fnvlist_size(vdev_errlist) > 0) {
4097 fnvlist_add_nvlist(outnvl, ZPOOL_INITIALIZE_VDEVS,
4100 fnvlist_free(vdev_errlist);
4102 spa_close(spa, FTAG);
4103 return (total_errors > 0 ? SET_ERROR(EINVAL) : 0);
4108 * "trim_command" -> POOL_TRIM_{CANCEL|START|SUSPEND} (uint64)
4109 * "trim_vdevs": { -> guids to TRIM (nvlist)
4110 * "vdev_path_1": vdev_guid_1, (uint64),
4111 * "vdev_path_2": vdev_guid_2, (uint64),
4114 * "trim_rate" -> Target TRIM rate in bytes/sec.
4115 * "trim_secure" -> Set to request a secure TRIM.
4119 * "trim_vdevs": { -> TRIM errors (nvlist)
4120 * "vdev_path_1": errno, see function body for possible errnos (uint64)
4121 * "vdev_path_2": errno, ... (uint64)
4126 * EINVAL is returned for an unknown commands or if any of the provided vdev
4127 * guids have be specified with a type other than uint64.
4129 static const zfs_ioc_key_t zfs_keys_pool_trim[] = {
4130 {ZPOOL_TRIM_COMMAND, DATA_TYPE_UINT64, 0},
4131 {ZPOOL_TRIM_VDEVS, DATA_TYPE_NVLIST, 0},
4132 {ZPOOL_TRIM_RATE, DATA_TYPE_UINT64, ZK_OPTIONAL},
4133 {ZPOOL_TRIM_SECURE, DATA_TYPE_BOOLEAN_VALUE, ZK_OPTIONAL},
4137 zfs_ioc_pool_trim(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
4140 if (nvlist_lookup_uint64(innvl, ZPOOL_TRIM_COMMAND, &cmd_type) != 0)
4141 return (SET_ERROR(EINVAL));
4143 if (!(cmd_type == POOL_TRIM_CANCEL ||
4144 cmd_type == POOL_TRIM_START ||
4145 cmd_type == POOL_TRIM_SUSPEND)) {
4146 return (SET_ERROR(EINVAL));
4149 nvlist_t *vdev_guids;
4150 if (nvlist_lookup_nvlist(innvl, ZPOOL_TRIM_VDEVS, &vdev_guids) != 0)
4151 return (SET_ERROR(EINVAL));
4153 for (nvpair_t *pair = nvlist_next_nvpair(vdev_guids, NULL);
4154 pair != NULL; pair = nvlist_next_nvpair(vdev_guids, pair)) {
4156 if (nvpair_value_uint64(pair, &vdev_guid) != 0) {
4157 return (SET_ERROR(EINVAL));
4161 /* Optional, defaults to maximum rate when not provided */
4163 if (nvlist_lookup_uint64(innvl, ZPOOL_TRIM_RATE, &rate) != 0)
4166 /* Optional, defaults to standard TRIM when not provided */
4168 if (nvlist_lookup_boolean_value(innvl, ZPOOL_TRIM_SECURE,
4174 int error = spa_open(poolname, &spa, FTAG);
4178 nvlist_t *vdev_errlist = fnvlist_alloc();
4179 int total_errors = spa_vdev_trim(spa, vdev_guids, cmd_type,
4180 rate, !!zfs_trim_metaslab_skip, secure, vdev_errlist);
4182 if (fnvlist_size(vdev_errlist) > 0)
4183 fnvlist_add_nvlist(outnvl, ZPOOL_TRIM_VDEVS, vdev_errlist);
4185 fnvlist_free(vdev_errlist);
4187 spa_close(spa, FTAG);
4188 return (total_errors > 0 ? SET_ERROR(EINVAL) : 0);
4192 * This ioctl waits for activity of a particular type to complete. If there is
4193 * no activity of that type in progress, it returns immediately, and the
4194 * returned value "waited" is false. If there is activity in progress, and no
4195 * tag is passed in, the ioctl blocks until all activity of that type is
4196 * complete, and then returns with "waited" set to true.
4198 * If a tag is provided, it identifies a particular instance of an activity to
4199 * wait for. Currently, this is only valid for use with 'initialize', because
4200 * that is the only activity for which there can be multiple instances running
4201 * concurrently. In the case of 'initialize', the tag corresponds to the guid of
4202 * the vdev on which to wait.
4204 * If a thread waiting in the ioctl receives a signal, the call will return
4205 * immediately, and the return value will be EINTR.
4208 * "wait_activity" -> int32_t
4209 * (optional) "wait_tag" -> uint64_t
4212 * outnvl: "waited" -> boolean_t
4214 static const zfs_ioc_key_t zfs_keys_pool_wait[] = {
4215 {ZPOOL_WAIT_ACTIVITY, DATA_TYPE_INT32, 0},
4216 {ZPOOL_WAIT_TAG, DATA_TYPE_UINT64, ZK_OPTIONAL},
4220 zfs_ioc_wait(const char *name, nvlist_t *innvl, nvlist_t *outnvl)
4227 if (nvlist_lookup_int32(innvl, ZPOOL_WAIT_ACTIVITY, &activity) != 0)
4230 if (nvlist_lookup_uint64(innvl, ZPOOL_WAIT_TAG, &tag) == 0)
4231 error = spa_wait_tag(name, activity, tag, &waited);
4233 error = spa_wait(name, activity, &waited);
4236 fnvlist_add_boolean_value(outnvl, ZPOOL_WAIT_WAITED, waited);
4242 * This ioctl waits for activity of a particular type to complete. If there is
4243 * no activity of that type in progress, it returns immediately, and the
4244 * returned value "waited" is false. If there is activity in progress, and no
4245 * tag is passed in, the ioctl blocks until all activity of that type is
4246 * complete, and then returns with "waited" set to true.
4248 * If a thread waiting in the ioctl receives a signal, the call will return
4249 * immediately, and the return value will be EINTR.
4252 * "wait_activity" -> int32_t
4255 * outnvl: "waited" -> boolean_t
4257 static const zfs_ioc_key_t zfs_keys_fs_wait[] = {
4258 {ZFS_WAIT_ACTIVITY, DATA_TYPE_INT32, 0},
4262 zfs_ioc_wait_fs(const char *name, nvlist_t *innvl, nvlist_t *outnvl)
4265 boolean_t waited = B_FALSE;
4271 if (nvlist_lookup_int32(innvl, ZFS_WAIT_ACTIVITY, &activity) != 0)
4272 return (SET_ERROR(EINVAL));
4274 if (activity >= ZFS_WAIT_NUM_ACTIVITIES || activity < 0)
4275 return (SET_ERROR(EINVAL));
4277 if ((error = dsl_pool_hold(name, FTAG, &dp)) != 0)
4280 if ((error = dsl_dataset_hold(dp, name, FTAG, &ds)) != 0) {
4281 dsl_pool_rele(dp, FTAG);
4286 mutex_enter(&dd->dd_activity_lock);
4287 dd->dd_activity_waiters++;
4290 * We get a long-hold here so that the dsl_dataset_t and dsl_dir_t
4291 * aren't evicted while we're waiting. Normally this is prevented by
4292 * holding the pool, but we can't do that while we're waiting since
4293 * that would prevent TXGs from syncing out. Some of the functionality
4294 * of long-holds (e.g. preventing deletion) is unnecessary for this
4295 * case, since we would cancel the waiters before proceeding with a
4296 * deletion. An alternative mechanism for keeping the dataset around
4297 * could be developed but this is simpler.
4299 dsl_dataset_long_hold(ds, FTAG);
4300 dsl_pool_rele(dp, FTAG);
4302 error = dsl_dir_wait(dd, ds, activity, &waited);
4304 dsl_dataset_long_rele(ds, FTAG);
4305 dd->dd_activity_waiters--;
4306 if (dd->dd_activity_waiters == 0)
4307 cv_signal(&dd->dd_activity_cv);
4308 mutex_exit(&dd->dd_activity_lock);
4310 dsl_dataset_rele(ds, FTAG);
4313 fnvlist_add_boolean_value(outnvl, ZFS_WAIT_WAITED, waited);
4319 * fsname is name of dataset to rollback (to most recent snapshot)
4321 * innvl may contain name of expected target snapshot
4323 * outnvl: "target" -> name of most recent snapshot
4326 static const zfs_ioc_key_t zfs_keys_rollback[] = {
4327 {"target", DATA_TYPE_STRING, ZK_OPTIONAL},
4331 zfs_ioc_rollback(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
4334 zvol_state_handle_t *zv;
4335 char *target = NULL;
4338 (void) nvlist_lookup_string(innvl, "target", &target);
4339 if (target != NULL) {
4340 const char *cp = strchr(target, '@');
4343 * The snap name must contain an @, and the part after it must
4344 * contain only valid characters.
4347 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
4348 return (SET_ERROR(EINVAL));
4351 if (getzfsvfs(fsname, &zfsvfs) == 0) {
4354 ds = dmu_objset_ds(zfsvfs->z_os);
4355 error = zfs_suspend_fs(zfsvfs);
4359 error = dsl_dataset_rollback(fsname, target, zfsvfs,
4361 resume_err = zfs_resume_fs(zfsvfs, ds);
4362 error = error ? error : resume_err;
4364 zfs_vfs_rele(zfsvfs);
4365 } else if ((zv = zvol_suspend(fsname)) != NULL) {
4366 error = dsl_dataset_rollback(fsname, target, zvol_tag(zv),
4370 error = dsl_dataset_rollback(fsname, target, NULL, outnvl);
4376 recursive_unmount(const char *fsname, void *arg)
4378 const char *snapname = arg;
4381 fullname = kmem_asprintf("%s@%s", fsname, snapname);
4382 zfs_unmount_snap(fullname);
4383 kmem_strfree(fullname);
4390 * snapname is the snapshot to redact.
4392 * "bookname" -> (string)
4393 * shortname of the redaction bookmark to generate
4394 * "snapnv" -> (nvlist, values ignored)
4395 * snapshots to redact snapname with respect to
4401 static const zfs_ioc_key_t zfs_keys_redact[] = {
4402 {"bookname", DATA_TYPE_STRING, 0},
4403 {"snapnv", DATA_TYPE_NVLIST, 0},
4407 zfs_ioc_redact(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
4410 nvlist_t *redactnvl = NULL;
4411 char *redactbook = NULL;
4413 if (nvlist_lookup_nvlist(innvl, "snapnv", &redactnvl) != 0)
4414 return (SET_ERROR(EINVAL));
4415 if (fnvlist_num_pairs(redactnvl) == 0)
4416 return (SET_ERROR(ENXIO));
4417 if (nvlist_lookup_string(innvl, "bookname", &redactbook) != 0)
4418 return (SET_ERROR(EINVAL));
4420 return (dmu_redact_snap(snapname, redactnvl, redactbook));
4425 * zc_name old name of dataset
4426 * zc_value new name of dataset
4427 * zc_cookie recursive flag (only valid for snapshots)
4432 zfs_ioc_rename(zfs_cmd_t *zc)
4435 dmu_objset_type_t ost;
4436 boolean_t recursive = zc->zc_cookie & 1;
4437 boolean_t nounmount = !!(zc->zc_cookie & 2);
4441 /* "zfs rename" from and to ...%recv datasets should both fail */
4442 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
4443 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
4444 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0 ||
4445 dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
4446 strchr(zc->zc_name, '%') || strchr(zc->zc_value, '%'))
4447 return (SET_ERROR(EINVAL));
4449 err = dmu_objset_hold(zc->zc_name, FTAG, &os);
4452 ost = dmu_objset_type(os);
4453 dmu_objset_rele(os, FTAG);
4455 at = strchr(zc->zc_name, '@');
4457 /* snaps must be in same fs */
4460 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
4461 return (SET_ERROR(EXDEV));
4463 if (ost == DMU_OST_ZFS && !nounmount) {
4464 error = dmu_objset_find(zc->zc_name,
4465 recursive_unmount, at + 1,
4466 recursive ? DS_FIND_CHILDREN : 0);
4472 error = dsl_dataset_rename_snapshot(zc->zc_name,
4473 at + 1, strchr(zc->zc_value, '@') + 1, recursive);
4478 return (dsl_dir_rename(zc->zc_name, zc->zc_value));
4483 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
4485 const char *propname = nvpair_name(pair);
4486 boolean_t issnap = (strchr(dsname, '@') != NULL);
4487 zfs_prop_t prop = zfs_name_to_prop(propname);
4488 uint64_t intval, compval;
4491 if (prop == ZPROP_USERPROP) {
4492 if (zfs_prop_user(propname)) {
4493 if ((err = zfs_secpolicy_write_perms(dsname,
4494 ZFS_DELEG_PERM_USERPROP, cr)))
4499 if (!issnap && zfs_prop_userquota(propname)) {
4500 const char *perm = NULL;
4501 const char *uq_prefix =
4502 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
4503 const char *gq_prefix =
4504 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
4505 const char *uiq_prefix =
4506 zfs_userquota_prop_prefixes[ZFS_PROP_USEROBJQUOTA];
4507 const char *giq_prefix =
4508 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPOBJQUOTA];
4509 const char *pq_prefix =
4510 zfs_userquota_prop_prefixes[ZFS_PROP_PROJECTQUOTA];
4511 const char *piq_prefix = zfs_userquota_prop_prefixes[\
4512 ZFS_PROP_PROJECTOBJQUOTA];
4514 if (strncmp(propname, uq_prefix,
4515 strlen(uq_prefix)) == 0) {
4516 perm = ZFS_DELEG_PERM_USERQUOTA;
4517 } else if (strncmp(propname, uiq_prefix,
4518 strlen(uiq_prefix)) == 0) {
4519 perm = ZFS_DELEG_PERM_USEROBJQUOTA;
4520 } else if (strncmp(propname, gq_prefix,
4521 strlen(gq_prefix)) == 0) {
4522 perm = ZFS_DELEG_PERM_GROUPQUOTA;
4523 } else if (strncmp(propname, giq_prefix,
4524 strlen(giq_prefix)) == 0) {
4525 perm = ZFS_DELEG_PERM_GROUPOBJQUOTA;
4526 } else if (strncmp(propname, pq_prefix,
4527 strlen(pq_prefix)) == 0) {
4528 perm = ZFS_DELEG_PERM_PROJECTQUOTA;
4529 } else if (strncmp(propname, piq_prefix,
4530 strlen(piq_prefix)) == 0) {
4531 perm = ZFS_DELEG_PERM_PROJECTOBJQUOTA;
4533 /* {USER|GROUP|PROJECT}USED are read-only */
4534 return (SET_ERROR(EINVAL));
4537 if ((err = zfs_secpolicy_write_perms(dsname, perm, cr)))
4542 return (SET_ERROR(EINVAL));
4546 return (SET_ERROR(EINVAL));
4548 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
4550 * dsl_prop_get_all_impl() returns properties in this
4554 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
4555 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4560 * Check that this value is valid for this pool version
4563 case ZFS_PROP_COMPRESSION:
4565 * If the user specified gzip compression, make sure
4566 * the SPA supports it. We ignore any errors here since
4567 * we'll catch them later.
4569 if (nvpair_value_uint64(pair, &intval) == 0) {
4570 compval = ZIO_COMPRESS_ALGO(intval);
4571 if (compval >= ZIO_COMPRESS_GZIP_1 &&
4572 compval <= ZIO_COMPRESS_GZIP_9 &&
4573 zfs_earlier_version(dsname,
4574 SPA_VERSION_GZIP_COMPRESSION)) {
4575 return (SET_ERROR(ENOTSUP));
4578 if (compval == ZIO_COMPRESS_ZLE &&
4579 zfs_earlier_version(dsname,
4580 SPA_VERSION_ZLE_COMPRESSION))
4581 return (SET_ERROR(ENOTSUP));
4583 if (compval == ZIO_COMPRESS_LZ4) {
4586 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4589 if (!spa_feature_is_enabled(spa,
4590 SPA_FEATURE_LZ4_COMPRESS)) {
4591 spa_close(spa, FTAG);
4592 return (SET_ERROR(ENOTSUP));
4594 spa_close(spa, FTAG);
4597 if (compval == ZIO_COMPRESS_ZSTD) {
4600 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4603 if (!spa_feature_is_enabled(spa,
4604 SPA_FEATURE_ZSTD_COMPRESS)) {
4605 spa_close(spa, FTAG);
4606 return (SET_ERROR(ENOTSUP));
4608 spa_close(spa, FTAG);
4613 case ZFS_PROP_COPIES:
4614 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
4615 return (SET_ERROR(ENOTSUP));
4618 case ZFS_PROP_VOLBLOCKSIZE:
4619 case ZFS_PROP_RECORDSIZE:
4620 /* Record sizes above 128k need the feature to be enabled */
4621 if (nvpair_value_uint64(pair, &intval) == 0 &&
4622 intval > SPA_OLD_MAXBLOCKSIZE) {
4626 * We don't allow setting the property above 1MB,
4627 * unless the tunable has been changed.
4629 if (intval > zfs_max_recordsize ||
4630 intval > SPA_MAXBLOCKSIZE)
4631 return (SET_ERROR(ERANGE));
4633 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4636 if (!spa_feature_is_enabled(spa,
4637 SPA_FEATURE_LARGE_BLOCKS)) {
4638 spa_close(spa, FTAG);
4639 return (SET_ERROR(ENOTSUP));
4641 spa_close(spa, FTAG);
4645 case ZFS_PROP_DNODESIZE:
4646 /* Dnode sizes above 512 need the feature to be enabled */
4647 if (nvpair_value_uint64(pair, &intval) == 0 &&
4648 intval != ZFS_DNSIZE_LEGACY) {
4651 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4654 if (!spa_feature_is_enabled(spa,
4655 SPA_FEATURE_LARGE_DNODE)) {
4656 spa_close(spa, FTAG);
4657 return (SET_ERROR(ENOTSUP));
4659 spa_close(spa, FTAG);
4663 case ZFS_PROP_SPECIAL_SMALL_BLOCKS:
4665 * This property could require the allocation classes
4666 * feature to be active for setting, however we allow
4667 * it so that tests of settable properties succeed.
4668 * The CLI will issue a warning in this case.
4672 case ZFS_PROP_SHARESMB:
4673 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
4674 return (SET_ERROR(ENOTSUP));
4677 case ZFS_PROP_ACLINHERIT:
4678 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
4679 nvpair_value_uint64(pair, &intval) == 0) {
4680 if (intval == ZFS_ACL_PASSTHROUGH_X &&
4681 zfs_earlier_version(dsname,
4682 SPA_VERSION_PASSTHROUGH_X))
4683 return (SET_ERROR(ENOTSUP));
4686 case ZFS_PROP_CHECKSUM:
4687 case ZFS_PROP_DEDUP:
4689 spa_feature_t feature;
4693 /* dedup feature version checks */
4694 if (prop == ZFS_PROP_DEDUP &&
4695 zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
4696 return (SET_ERROR(ENOTSUP));
4698 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
4699 nvpair_value_uint64(pair, &intval) == 0) {
4700 /* check prop value is enabled in features */
4701 feature = zio_checksum_to_feature(
4702 intval & ZIO_CHECKSUM_MASK);
4703 if (feature == SPA_FEATURE_NONE)
4706 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4709 if (!spa_feature_is_enabled(spa, feature)) {
4710 spa_close(spa, FTAG);
4711 return (SET_ERROR(ENOTSUP));
4713 spa_close(spa, FTAG);
4722 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
4726 * Removes properties from the given props list that fail permission checks
4727 * needed to clear them and to restore them in case of a receive error. For each
4728 * property, make sure we have both set and inherit permissions.
4730 * Returns the first error encountered if any permission checks fail. If the
4731 * caller provides a non-NULL errlist, it also gives the complete list of names
4732 * of all the properties that failed a permission check along with the
4733 * corresponding error numbers. The caller is responsible for freeing the
4736 * If every property checks out successfully, zero is returned and the list
4737 * pointed at by errlist is NULL.
4740 zfs_check_clearable(const char *dataset, nvlist_t *props, nvlist_t **errlist)
4743 nvpair_t *pair, *next_pair;
4750 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4752 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
4753 (void) strlcpy(zc->zc_name, dataset, sizeof (zc->zc_name));
4754 pair = nvlist_next_nvpair(props, NULL);
4755 while (pair != NULL) {
4756 next_pair = nvlist_next_nvpair(props, pair);
4758 (void) strlcpy(zc->zc_value, nvpair_name(pair),
4759 sizeof (zc->zc_value));
4760 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
4761 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
4762 VERIFY(nvlist_remove_nvpair(props, pair) == 0);
4763 VERIFY(nvlist_add_int32(errors,
4764 zc->zc_value, err) == 0);
4768 kmem_free(zc, sizeof (zfs_cmd_t));
4770 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
4771 nvlist_free(errors);
4774 VERIFY(nvpair_value_int32(pair, &rv) == 0);
4777 if (errlist == NULL)
4778 nvlist_free(errors);
4786 propval_equals(nvpair_t *p1, nvpair_t *p2)
4788 if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
4789 /* dsl_prop_get_all_impl() format */
4791 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
4792 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4796 if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
4798 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
4799 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4803 if (nvpair_type(p1) != nvpair_type(p2))
4806 if (nvpair_type(p1) == DATA_TYPE_STRING) {
4807 char *valstr1, *valstr2;
4809 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
4810 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
4811 return (strcmp(valstr1, valstr2) == 0);
4813 uint64_t intval1, intval2;
4815 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
4816 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
4817 return (intval1 == intval2);
4822 * Remove properties from props if they are not going to change (as determined
4823 * by comparison with origprops). Remove them from origprops as well, since we
4824 * do not need to clear or restore properties that won't change.
4827 props_reduce(nvlist_t *props, nvlist_t *origprops)
4829 nvpair_t *pair, *next_pair;
4831 if (origprops == NULL)
4832 return; /* all props need to be received */
4834 pair = nvlist_next_nvpair(props, NULL);
4835 while (pair != NULL) {
4836 const char *propname = nvpair_name(pair);
4839 next_pair = nvlist_next_nvpair(props, pair);
4841 if ((nvlist_lookup_nvpair(origprops, propname,
4842 &match) != 0) || !propval_equals(pair, match))
4843 goto next; /* need to set received value */
4845 /* don't clear the existing received value */
4846 (void) nvlist_remove_nvpair(origprops, match);
4847 /* don't bother receiving the property */
4848 (void) nvlist_remove_nvpair(props, pair);
4855 * Extract properties that cannot be set PRIOR to the receipt of a dataset.
4856 * For example, refquota cannot be set until after the receipt of a dataset,
4857 * because in replication streams, an older/earlier snapshot may exceed the
4858 * refquota. We want to receive the older/earlier snapshot, but setting
4859 * refquota pre-receipt will set the dsl's ACTUAL quota, which will prevent
4860 * the older/earlier snapshot from being received (with EDQUOT).
4862 * The ZFS test "zfs_receive_011_pos" demonstrates such a scenario.
4864 * libzfs will need to be judicious handling errors encountered by props
4865 * extracted by this function.
4868 extract_delay_props(nvlist_t *props)
4870 nvlist_t *delayprops;
4871 nvpair_t *nvp, *tmp;
4872 static const zfs_prop_t delayable[] = {
4874 ZFS_PROP_KEYLOCATION,
4879 VERIFY(nvlist_alloc(&delayprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4881 for (nvp = nvlist_next_nvpair(props, NULL); nvp != NULL;
4882 nvp = nvlist_next_nvpair(props, nvp)) {
4884 * strcmp() is safe because zfs_prop_to_name() always returns
4887 for (i = 0; delayable[i] != 0; i++) {
4888 if (strcmp(zfs_prop_to_name(delayable[i]),
4889 nvpair_name(nvp)) == 0) {
4893 if (delayable[i] != 0) {
4894 tmp = nvlist_prev_nvpair(props, nvp);
4895 VERIFY(nvlist_add_nvpair(delayprops, nvp) == 0);
4896 VERIFY(nvlist_remove_nvpair(props, nvp) == 0);
4901 if (nvlist_empty(delayprops)) {
4902 nvlist_free(delayprops);
4905 return (delayprops);
4909 zfs_allow_log_destroy(void *arg)
4911 char *poolname = arg;
4913 if (poolname != NULL)
4914 kmem_strfree(poolname);
4918 static boolean_t zfs_ioc_recv_inject_err;
4922 * nvlist 'errors' is always allocated. It will contain descriptions of
4923 * encountered errors, if any. It's the callers responsibility to free.
4926 zfs_ioc_recv_impl(char *tofs, char *tosnap, char *origin, nvlist_t *recvprops,
4927 nvlist_t *localprops, nvlist_t *hidden_args, boolean_t force,
4928 boolean_t resumable, int input_fd,
4929 dmu_replay_record_t *begin_record, uint64_t *read_bytes,
4930 uint64_t *errflags, nvlist_t **errors)
4932 dmu_recv_cookie_t drc;
4934 int props_error = 0;
4936 nvlist_t *local_delayprops = NULL;
4937 nvlist_t *recv_delayprops = NULL;
4938 nvlist_t *origprops = NULL; /* existing properties */
4939 nvlist_t *origrecvd = NULL; /* existing received properties */
4940 boolean_t first_recvd_props = B_FALSE;
4941 boolean_t tofs_was_redacted;
4942 zfs_file_t *input_fp;
4946 *errors = fnvlist_alloc();
4949 if ((input_fp = zfs_file_get(input_fd)) == NULL)
4950 return (SET_ERROR(EBADF));
4952 noff = off = zfs_file_off(input_fp);
4953 error = dmu_recv_begin(tofs, tosnap, begin_record, force,
4954 resumable, localprops, hidden_args, origin, &drc, input_fp,
4958 tofs_was_redacted = dsl_get_redacted(drc.drc_ds);
4961 * Set properties before we receive the stream so that they are applied
4962 * to the new data. Note that we must call dmu_recv_stream() if
4963 * dmu_recv_begin() succeeds.
4965 if (recvprops != NULL && !drc.drc_newfs) {
4966 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
4967 SPA_VERSION_RECVD_PROPS &&
4968 !dsl_prop_get_hasrecvd(tofs))
4969 first_recvd_props = B_TRUE;
4972 * If new received properties are supplied, they are to
4973 * completely replace the existing received properties,
4974 * so stash away the existing ones.
4976 if (dsl_prop_get_received(tofs, &origrecvd) == 0) {
4977 nvlist_t *errlist = NULL;
4979 * Don't bother writing a property if its value won't
4980 * change (and avoid the unnecessary security checks).
4982 * The first receive after SPA_VERSION_RECVD_PROPS is a
4983 * special case where we blow away all local properties
4986 if (!first_recvd_props)
4987 props_reduce(recvprops, origrecvd);
4988 if (zfs_check_clearable(tofs, origrecvd, &errlist) != 0)
4989 (void) nvlist_merge(*errors, errlist, 0);
4990 nvlist_free(errlist);
4992 if (clear_received_props(tofs, origrecvd,
4993 first_recvd_props ? NULL : recvprops) != 0)
4994 *errflags |= ZPROP_ERR_NOCLEAR;
4996 *errflags |= ZPROP_ERR_NOCLEAR;
5001 * Stash away existing properties so we can restore them on error unless
5002 * we're doing the first receive after SPA_VERSION_RECVD_PROPS, in which
5003 * case "origrecvd" will take care of that.
5005 if (localprops != NULL && !drc.drc_newfs && !first_recvd_props) {
5007 if (dmu_objset_hold(tofs, FTAG, &os) == 0) {
5008 if (dsl_prop_get_all(os, &origprops) != 0) {
5009 *errflags |= ZPROP_ERR_NOCLEAR;
5011 dmu_objset_rele(os, FTAG);
5013 *errflags |= ZPROP_ERR_NOCLEAR;
5017 if (recvprops != NULL) {
5018 props_error = dsl_prop_set_hasrecvd(tofs);
5020 if (props_error == 0) {
5021 recv_delayprops = extract_delay_props(recvprops);
5022 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
5023 recvprops, *errors);
5027 if (localprops != NULL) {
5028 nvlist_t *oprops = fnvlist_alloc();
5029 nvlist_t *xprops = fnvlist_alloc();
5030 nvpair_t *nvp = NULL;
5032 while ((nvp = nvlist_next_nvpair(localprops, nvp)) != NULL) {
5033 if (nvpair_type(nvp) == DATA_TYPE_BOOLEAN) {
5035 const char *name = nvpair_name(nvp);
5036 zfs_prop_t prop = zfs_name_to_prop(name);
5037 if (prop != ZPROP_USERPROP) {
5038 if (!zfs_prop_inheritable(prop))
5040 } else if (!zfs_prop_user(name))
5042 fnvlist_add_boolean(xprops, name);
5044 /* -o property=value */
5045 fnvlist_add_nvpair(oprops, nvp);
5049 local_delayprops = extract_delay_props(oprops);
5050 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_LOCAL,
5052 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_INHERITED,
5055 nvlist_free(oprops);
5056 nvlist_free(xprops);
5059 error = dmu_recv_stream(&drc, &off);
5062 zfsvfs_t *zfsvfs = NULL;
5063 zvol_state_handle_t *zv = NULL;
5065 if (getzfsvfs(tofs, &zfsvfs) == 0) {
5069 boolean_t stream_is_redacted = DMU_GET_FEATUREFLAGS(
5070 begin_record->drr_u.drr_begin.
5071 drr_versioninfo) & DMU_BACKUP_FEATURE_REDACTED;
5073 ds = dmu_objset_ds(zfsvfs->z_os);
5074 error = zfs_suspend_fs(zfsvfs);
5076 * If the suspend fails, then the recv_end will
5077 * likely also fail, and clean up after itself.
5079 end_err = dmu_recv_end(&drc, zfsvfs);
5081 * If the dataset was not redacted, but we received a
5082 * redacted stream onto it, we need to unmount the
5083 * dataset. Otherwise, resume the filesystem.
5085 if (error == 0 && !drc.drc_newfs &&
5086 stream_is_redacted && !tofs_was_redacted) {
5087 error = zfs_end_fs(zfsvfs, ds);
5088 } else if (error == 0) {
5089 error = zfs_resume_fs(zfsvfs, ds);
5091 error = error ? error : end_err;
5092 zfs_vfs_rele(zfsvfs);
5093 } else if ((zv = zvol_suspend(tofs)) != NULL) {
5094 error = dmu_recv_end(&drc, zvol_tag(zv));
5097 error = dmu_recv_end(&drc, NULL);
5100 /* Set delayed properties now, after we're done receiving. */
5101 if (recv_delayprops != NULL && error == 0) {
5102 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
5103 recv_delayprops, *errors);
5105 if (local_delayprops != NULL && error == 0) {
5106 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_LOCAL,
5107 local_delayprops, *errors);
5112 * Merge delayed props back in with initial props, in case
5113 * we're DEBUG and zfs_ioc_recv_inject_err is set (which means
5114 * we have to make sure clear_received_props() includes
5115 * the delayed properties).
5117 * Since zfs_ioc_recv_inject_err is only in DEBUG kernels,
5118 * using ASSERT() will be just like a VERIFY.
5120 if (recv_delayprops != NULL) {
5121 ASSERT(nvlist_merge(recvprops, recv_delayprops, 0) == 0);
5122 nvlist_free(recv_delayprops);
5124 if (local_delayprops != NULL) {
5125 ASSERT(nvlist_merge(localprops, local_delayprops, 0) == 0);
5126 nvlist_free(local_delayprops);
5128 *read_bytes = off - noff;
5131 if (zfs_ioc_recv_inject_err) {
5132 zfs_ioc_recv_inject_err = B_FALSE;
5138 * On error, restore the original props.
5140 if (error != 0 && recvprops != NULL && !drc.drc_newfs) {
5141 if (clear_received_props(tofs, recvprops, NULL) != 0) {
5143 * We failed to clear the received properties.
5144 * Since we may have left a $recvd value on the
5145 * system, we can't clear the $hasrecvd flag.
5147 *errflags |= ZPROP_ERR_NORESTORE;
5148 } else if (first_recvd_props) {
5149 dsl_prop_unset_hasrecvd(tofs);
5152 if (origrecvd == NULL && !drc.drc_newfs) {
5153 /* We failed to stash the original properties. */
5154 *errflags |= ZPROP_ERR_NORESTORE;
5158 * dsl_props_set() will not convert RECEIVED to LOCAL on or
5159 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
5160 * explicitly if we're restoring local properties cleared in the
5161 * first new-style receive.
5163 if (origrecvd != NULL &&
5164 zfs_set_prop_nvlist(tofs, (first_recvd_props ?
5165 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
5166 origrecvd, NULL) != 0) {
5168 * We stashed the original properties but failed to
5171 *errflags |= ZPROP_ERR_NORESTORE;
5174 if (error != 0 && localprops != NULL && !drc.drc_newfs &&
5175 !first_recvd_props) {
5177 nvlist_t *inheritprops;
5180 if (origprops == NULL) {
5181 /* We failed to stash the original properties. */
5182 *errflags |= ZPROP_ERR_NORESTORE;
5186 /* Restore original props */
5187 setprops = fnvlist_alloc();
5188 inheritprops = fnvlist_alloc();
5190 while ((nvp = nvlist_next_nvpair(localprops, nvp)) != NULL) {
5191 const char *name = nvpair_name(nvp);
5195 if (!nvlist_exists(origprops, name)) {
5197 * Property was not present or was explicitly
5198 * inherited before the receive, restore this.
5200 fnvlist_add_boolean(inheritprops, name);
5203 attrs = fnvlist_lookup_nvlist(origprops, name);
5204 source = fnvlist_lookup_string(attrs, ZPROP_SOURCE);
5206 /* Skip received properties */
5207 if (strcmp(source, ZPROP_SOURCE_VAL_RECVD) == 0)
5210 if (strcmp(source, tofs) == 0) {
5211 /* Property was locally set */
5212 fnvlist_add_nvlist(setprops, name, attrs);
5214 /* Property was implicitly inherited */
5215 fnvlist_add_boolean(inheritprops, name);
5219 if (zfs_set_prop_nvlist(tofs, ZPROP_SRC_LOCAL, setprops,
5221 *errflags |= ZPROP_ERR_NORESTORE;
5222 if (zfs_set_prop_nvlist(tofs, ZPROP_SRC_INHERITED, inheritprops,
5224 *errflags |= ZPROP_ERR_NORESTORE;
5226 nvlist_free(setprops);
5227 nvlist_free(inheritprops);
5230 zfs_file_put(input_fp);
5231 nvlist_free(origrecvd);
5232 nvlist_free(origprops);
5235 error = props_error;
5242 * zc_name name of containing filesystem (unused)
5243 * zc_nvlist_src{_size} nvlist of properties to apply
5244 * zc_nvlist_conf{_size} nvlist of properties to exclude
5245 * (DATA_TYPE_BOOLEAN) and override (everything else)
5246 * zc_value name of snapshot to create
5247 * zc_string name of clone origin (if DRR_FLAG_CLONE)
5248 * zc_cookie file descriptor to recv from
5249 * zc_begin_record the BEGIN record of the stream (not byteswapped)
5250 * zc_guid force flag
5253 * zc_cookie number of bytes read
5254 * zc_obj zprop_errflags_t
5255 * zc_nvlist_dst{_size} error for each unapplied received property
5258 zfs_ioc_recv(zfs_cmd_t *zc)
5260 dmu_replay_record_t begin_record;
5261 nvlist_t *errors = NULL;
5262 nvlist_t *recvdprops = NULL;
5263 nvlist_t *localprops = NULL;
5264 char *origin = NULL;
5266 char tofs[ZFS_MAX_DATASET_NAME_LEN];
5269 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
5270 strchr(zc->zc_value, '@') == NULL ||
5271 strchr(zc->zc_value, '%'))
5272 return (SET_ERROR(EINVAL));
5274 (void) strlcpy(tofs, zc->zc_value, sizeof (tofs));
5275 tosnap = strchr(tofs, '@');
5278 if (zc->zc_nvlist_src != 0 &&
5279 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
5280 zc->zc_iflags, &recvdprops)) != 0)
5283 if (zc->zc_nvlist_conf != 0 &&
5284 (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
5285 zc->zc_iflags, &localprops)) != 0)
5288 if (zc->zc_string[0])
5289 origin = zc->zc_string;
5291 begin_record.drr_type = DRR_BEGIN;
5292 begin_record.drr_payloadlen = 0;
5293 begin_record.drr_u.drr_begin = zc->zc_begin_record;
5295 error = zfs_ioc_recv_impl(tofs, tosnap, origin, recvdprops, localprops,
5296 NULL, zc->zc_guid, B_FALSE, zc->zc_cookie, &begin_record,
5297 &zc->zc_cookie, &zc->zc_obj, &errors);
5298 nvlist_free(recvdprops);
5299 nvlist_free(localprops);
5302 * Now that all props, initial and delayed, are set, report the prop
5303 * errors to the caller.
5305 if (zc->zc_nvlist_dst_size != 0 && errors != NULL &&
5306 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
5307 put_nvlist(zc, errors) != 0)) {
5309 * Caller made zc->zc_nvlist_dst less than the minimum expected
5310 * size or supplied an invalid address.
5312 error = SET_ERROR(EINVAL);
5315 nvlist_free(errors);
5322 * "snapname" -> full name of the snapshot to create
5323 * (optional) "props" -> received properties to set (nvlist)
5324 * (optional) "localprops" -> override and exclude properties (nvlist)
5325 * (optional) "origin" -> name of clone origin (DRR_FLAG_CLONE)
5326 * "begin_record" -> non-byteswapped dmu_replay_record_t
5327 * "input_fd" -> file descriptor to read stream from (int32)
5328 * (optional) "force" -> force flag (value ignored)
5329 * (optional) "resumable" -> resumable flag (value ignored)
5330 * (optional) "cleanup_fd" -> unused
5331 * (optional) "action_handle" -> unused
5332 * (optional) "hidden_args" -> { "wkeydata" -> value }
5336 * "read_bytes" -> number of bytes read
5337 * "error_flags" -> zprop_errflags_t
5338 * "errors" -> error for each unapplied received property (nvlist)
5341 static const zfs_ioc_key_t zfs_keys_recv_new[] = {
5342 {"snapname", DATA_TYPE_STRING, 0},
5343 {"props", DATA_TYPE_NVLIST, ZK_OPTIONAL},
5344 {"localprops", DATA_TYPE_NVLIST, ZK_OPTIONAL},
5345 {"origin", DATA_TYPE_STRING, ZK_OPTIONAL},
5346 {"begin_record", DATA_TYPE_BYTE_ARRAY, 0},
5347 {"input_fd", DATA_TYPE_INT32, 0},
5348 {"force", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
5349 {"resumable", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
5350 {"cleanup_fd", DATA_TYPE_INT32, ZK_OPTIONAL},
5351 {"action_handle", DATA_TYPE_UINT64, ZK_OPTIONAL},
5352 {"hidden_args", DATA_TYPE_NVLIST, ZK_OPTIONAL},
5356 zfs_ioc_recv_new(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
5358 dmu_replay_record_t *begin_record;
5359 uint_t begin_record_size;
5360 nvlist_t *errors = NULL;
5361 nvlist_t *recvprops = NULL;
5362 nvlist_t *localprops = NULL;
5363 nvlist_t *hidden_args = NULL;
5365 char *origin = NULL;
5367 char tofs[ZFS_MAX_DATASET_NAME_LEN];
5369 boolean_t resumable;
5370 uint64_t read_bytes = 0;
5371 uint64_t errflags = 0;
5375 snapname = fnvlist_lookup_string(innvl, "snapname");
5377 if (dataset_namecheck(snapname, NULL, NULL) != 0 ||
5378 strchr(snapname, '@') == NULL ||
5379 strchr(snapname, '%'))
5380 return (SET_ERROR(EINVAL));
5382 (void) strlcpy(tofs, snapname, sizeof (tofs));
5383 tosnap = strchr(tofs, '@');
5386 error = nvlist_lookup_string(innvl, "origin", &origin);
5387 if (error && error != ENOENT)
5390 error = nvlist_lookup_byte_array(innvl, "begin_record",
5391 (uchar_t **)&begin_record, &begin_record_size);
5392 if (error != 0 || begin_record_size != sizeof (*begin_record))
5393 return (SET_ERROR(EINVAL));
5395 input_fd = fnvlist_lookup_int32(innvl, "input_fd");
5397 force = nvlist_exists(innvl, "force");
5398 resumable = nvlist_exists(innvl, "resumable");
5400 /* we still use "props" here for backwards compatibility */
5401 error = nvlist_lookup_nvlist(innvl, "props", &recvprops);
5402 if (error && error != ENOENT)
5405 error = nvlist_lookup_nvlist(innvl, "localprops", &localprops);
5406 if (error && error != ENOENT)
5409 error = nvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS, &hidden_args);
5410 if (error && error != ENOENT)
5413 error = zfs_ioc_recv_impl(tofs, tosnap, origin, recvprops, localprops,
5414 hidden_args, force, resumable, input_fd, begin_record,
5415 &read_bytes, &errflags, &errors);
5417 fnvlist_add_uint64(outnvl, "read_bytes", read_bytes);
5418 fnvlist_add_uint64(outnvl, "error_flags", errflags);
5419 fnvlist_add_nvlist(outnvl, "errors", errors);
5421 nvlist_free(errors);
5422 nvlist_free(recvprops);
5423 nvlist_free(localprops);
5428 typedef struct dump_bytes_io {
5436 dump_bytes_cb(void *arg)
5438 dump_bytes_io_t *dbi = (dump_bytes_io_t *)arg;
5445 dbi->dbi_err = zfs_file_write(fp, buf, dbi->dbi_len, NULL);
5449 dump_bytes(objset_t *os, void *buf, int len, void *arg)
5451 dump_bytes_io_t dbi;
5457 #if defined(HAVE_LARGE_STACKS)
5458 dump_bytes_cb(&dbi);
5461 * The vn_rdwr() call is performed in a taskq to ensure that there is
5462 * always enough stack space to write safely to the target filesystem.
5463 * The ZIO_TYPE_FREE threads are used because there can be a lot of
5464 * them and they are used in vdev_file.c for a similar purpose.
5466 spa_taskq_dispatch_sync(dmu_objset_spa(os), ZIO_TYPE_FREE,
5467 ZIO_TASKQ_ISSUE, dump_bytes_cb, &dbi, TQ_SLEEP);
5468 #endif /* HAVE_LARGE_STACKS */
5470 return (dbi.dbi_err);
5475 * zc_name name of snapshot to send
5476 * zc_cookie file descriptor to send stream to
5477 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj)
5478 * zc_sendobj objsetid of snapshot to send
5479 * zc_fromobj objsetid of incremental fromsnap (may be zero)
5480 * zc_guid if set, estimate size of stream only. zc_cookie is ignored.
5481 * output size in zc_objset_type.
5482 * zc_flags lzc_send_flags
5485 * zc_objset_type estimated size, if zc_guid is set
5487 * NOTE: This is no longer the preferred interface, any new functionality
5488 * should be added to zfs_ioc_send_new() instead.
5491 zfs_ioc_send(zfs_cmd_t *zc)
5495 boolean_t estimate = (zc->zc_guid != 0);
5496 boolean_t embedok = (zc->zc_flags & 0x1);
5497 boolean_t large_block_ok = (zc->zc_flags & 0x2);
5498 boolean_t compressok = (zc->zc_flags & 0x4);
5499 boolean_t rawok = (zc->zc_flags & 0x8);
5500 boolean_t savedok = (zc->zc_flags & 0x10);
5502 if (zc->zc_obj != 0) {
5504 dsl_dataset_t *tosnap;
5506 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5510 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
5512 dsl_pool_rele(dp, FTAG);
5516 if (dsl_dir_is_clone(tosnap->ds_dir))
5518 dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj;
5519 dsl_dataset_rele(tosnap, FTAG);
5520 dsl_pool_rele(dp, FTAG);
5525 dsl_dataset_t *tosnap;
5526 dsl_dataset_t *fromsnap = NULL;
5528 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5532 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj,
5535 dsl_pool_rele(dp, FTAG);
5539 if (zc->zc_fromobj != 0) {
5540 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
5543 dsl_dataset_rele(tosnap, FTAG);
5544 dsl_pool_rele(dp, FTAG);
5549 error = dmu_send_estimate_fast(tosnap, fromsnap, NULL,
5550 compressok || rawok, savedok, &zc->zc_objset_type);
5552 if (fromsnap != NULL)
5553 dsl_dataset_rele(fromsnap, FTAG);
5554 dsl_dataset_rele(tosnap, FTAG);
5555 dsl_pool_rele(dp, FTAG);
5558 dmu_send_outparams_t out = {0};
5560 if ((fp = zfs_file_get(zc->zc_cookie)) == NULL)
5561 return (SET_ERROR(EBADF));
5563 off = zfs_file_off(fp);
5564 out.dso_outfunc = dump_bytes;
5566 out.dso_dryrun = B_FALSE;
5567 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
5568 zc->zc_fromobj, embedok, large_block_ok, compressok,
5569 rawok, savedok, zc->zc_cookie, &off, &out);
5578 * zc_name name of snapshot on which to report progress
5579 * zc_cookie file descriptor of send stream
5582 * zc_cookie number of bytes written in send stream thus far
5583 * zc_objset_type logical size of data traversed by send thus far
5586 zfs_ioc_send_progress(zfs_cmd_t *zc)
5590 dmu_sendstatus_t *dsp = NULL;
5593 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5597 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
5599 dsl_pool_rele(dp, FTAG);
5603 mutex_enter(&ds->ds_sendstream_lock);
5606 * Iterate over all the send streams currently active on this dataset.
5607 * If there's one which matches the specified file descriptor _and_ the
5608 * stream was started by the current process, return the progress of
5612 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
5613 dsp = list_next(&ds->ds_sendstreams, dsp)) {
5614 if (dsp->dss_outfd == zc->zc_cookie &&
5615 zfs_proc_is_caller(dsp->dss_proc))
5620 zc->zc_cookie = atomic_cas_64((volatile uint64_t *)dsp->dss_off,
5622 /* This is the closest thing we have to atomic_read_64. */
5623 zc->zc_objset_type = atomic_cas_64(&dsp->dss_blocks, 0, 0);
5625 error = SET_ERROR(ENOENT);
5628 mutex_exit(&ds->ds_sendstream_lock);
5629 dsl_dataset_rele(ds, FTAG);
5630 dsl_pool_rele(dp, FTAG);
5635 zfs_ioc_inject_fault(zfs_cmd_t *zc)
5639 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
5640 &zc->zc_inject_record);
5643 zc->zc_guid = (uint64_t)id;
5649 zfs_ioc_clear_fault(zfs_cmd_t *zc)
5651 return (zio_clear_fault((int)zc->zc_guid));
5655 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
5657 int id = (int)zc->zc_guid;
5660 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
5661 &zc->zc_inject_record);
5669 zfs_ioc_error_log(zfs_cmd_t *zc)
5673 uint64_t count = zc->zc_nvlist_dst_size;
5675 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
5678 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
5681 zc->zc_nvlist_dst_size = count;
5683 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
5685 spa_close(spa, FTAG);
5691 zfs_ioc_clear(zfs_cmd_t *zc)
5698 * On zpool clear we also fix up missing slogs
5700 mutex_enter(&spa_namespace_lock);
5701 spa = spa_lookup(zc->zc_name);
5703 mutex_exit(&spa_namespace_lock);
5704 return (SET_ERROR(EIO));
5706 if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
5707 /* we need to let spa_open/spa_load clear the chains */
5708 spa_set_log_state(spa, SPA_LOG_CLEAR);
5710 spa->spa_last_open_failed = 0;
5711 mutex_exit(&spa_namespace_lock);
5713 if (zc->zc_cookie & ZPOOL_NO_REWIND) {
5714 error = spa_open(zc->zc_name, &spa, FTAG);
5717 nvlist_t *config = NULL;
5719 if (zc->zc_nvlist_src == 0)
5720 return (SET_ERROR(EINVAL));
5722 if ((error = get_nvlist(zc->zc_nvlist_src,
5723 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
5724 error = spa_open_rewind(zc->zc_name, &spa, FTAG,
5726 if (config != NULL) {
5729 if ((err = put_nvlist(zc, config)) != 0)
5731 nvlist_free(config);
5733 nvlist_free(policy);
5741 * If multihost is enabled, resuming I/O is unsafe as another
5742 * host may have imported the pool.
5744 if (spa_multihost(spa) && spa_suspended(spa))
5745 return (SET_ERROR(EINVAL));
5747 spa_vdev_state_enter(spa, SCL_NONE);
5749 if (zc->zc_guid == 0) {
5752 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
5754 error = SET_ERROR(ENODEV);
5755 (void) spa_vdev_state_exit(spa, NULL, error);
5756 spa_close(spa, FTAG);
5761 vdev_clear(spa, vd);
5763 (void) spa_vdev_state_exit(spa, spa_suspended(spa) ?
5764 NULL : spa->spa_root_vdev, 0);
5767 * Resume any suspended I/Os.
5769 if (zio_resume(spa) != 0)
5770 error = SET_ERROR(EIO);
5772 spa_close(spa, FTAG);
5778 * Reopen all the vdevs associated with the pool.
5781 * "scrub_restart" -> when true and scrub is running, allow to restart
5782 * scrub as the side effect of the reopen (boolean).
5787 static const zfs_ioc_key_t zfs_keys_pool_reopen[] = {
5788 {"scrub_restart", DATA_TYPE_BOOLEAN_VALUE, ZK_OPTIONAL},
5792 zfs_ioc_pool_reopen(const char *pool, nvlist_t *innvl, nvlist_t *outnvl)
5797 boolean_t rc, scrub_restart = B_TRUE;
5800 error = nvlist_lookup_boolean_value(innvl,
5801 "scrub_restart", &rc);
5806 error = spa_open(pool, &spa, FTAG);
5810 spa_vdev_state_enter(spa, SCL_NONE);
5813 * If the scrub_restart flag is B_FALSE and a scrub is already
5814 * in progress then set spa_scrub_reopen flag to B_TRUE so that
5815 * we don't restart the scrub as a side effect of the reopen.
5816 * Otherwise, let vdev_open() decided if a resilver is required.
5819 spa->spa_scrub_reopen = (!scrub_restart &&
5820 dsl_scan_scrubbing(spa->spa_dsl_pool));
5821 vdev_reopen(spa->spa_root_vdev);
5822 spa->spa_scrub_reopen = B_FALSE;
5824 (void) spa_vdev_state_exit(spa, NULL, 0);
5825 spa_close(spa, FTAG);
5831 * zc_name name of filesystem
5834 * zc_string name of conflicting snapshot, if there is one
5837 zfs_ioc_promote(zfs_cmd_t *zc)
5840 dsl_dataset_t *ds, *ods;
5841 char origin[ZFS_MAX_DATASET_NAME_LEN];
5845 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
5846 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0 ||
5847 strchr(zc->zc_name, '%'))
5848 return (SET_ERROR(EINVAL));
5850 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5854 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
5856 dsl_pool_rele(dp, FTAG);
5860 if (!dsl_dir_is_clone(ds->ds_dir)) {
5861 dsl_dataset_rele(ds, FTAG);
5862 dsl_pool_rele(dp, FTAG);
5863 return (SET_ERROR(EINVAL));
5866 error = dsl_dataset_hold_obj(dp,
5867 dsl_dir_phys(ds->ds_dir)->dd_origin_obj, FTAG, &ods);
5869 dsl_dataset_rele(ds, FTAG);
5870 dsl_pool_rele(dp, FTAG);
5874 dsl_dataset_name(ods, origin);
5875 dsl_dataset_rele(ods, FTAG);
5876 dsl_dataset_rele(ds, FTAG);
5877 dsl_pool_rele(dp, FTAG);
5880 * We don't need to unmount *all* the origin fs's snapshots, but
5883 cp = strchr(origin, '@');
5886 (void) dmu_objset_find(origin,
5887 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
5888 return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
5892 * Retrieve a single {user|group|project}{used|quota}@... property.
5895 * zc_name name of filesystem
5896 * zc_objset_type zfs_userquota_prop_t
5897 * zc_value domain name (eg. "S-1-234-567-89")
5898 * zc_guid RID/UID/GID
5901 * zc_cookie property value
5904 zfs_ioc_userspace_one(zfs_cmd_t *zc)
5909 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
5910 return (SET_ERROR(EINVAL));
5912 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
5916 error = zfs_userspace_one(zfsvfs,
5917 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
5918 zfsvfs_rele(zfsvfs, FTAG);
5925 * zc_name name of filesystem
5926 * zc_cookie zap cursor
5927 * zc_objset_type zfs_userquota_prop_t
5928 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
5931 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
5932 * zc_cookie zap cursor
5935 zfs_ioc_userspace_many(zfs_cmd_t *zc)
5938 int bufsize = zc->zc_nvlist_dst_size;
5941 return (SET_ERROR(ENOMEM));
5943 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
5947 void *buf = vmem_alloc(bufsize, KM_SLEEP);
5949 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
5950 buf, &zc->zc_nvlist_dst_size);
5953 error = xcopyout(buf,
5954 (void *)(uintptr_t)zc->zc_nvlist_dst,
5955 zc->zc_nvlist_dst_size);
5957 vmem_free(buf, bufsize);
5958 zfsvfs_rele(zfsvfs, FTAG);
5965 * zc_name name of filesystem
5971 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
5976 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
5977 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
5979 * If userused is not enabled, it may be because the
5980 * objset needs to be closed & reopened (to grow the
5981 * objset_phys_t). Suspend/resume the fs will do that.
5983 dsl_dataset_t *ds, *newds;
5985 ds = dmu_objset_ds(zfsvfs->z_os);
5986 error = zfs_suspend_fs(zfsvfs);
5988 dmu_objset_refresh_ownership(ds, &newds,
5990 error = zfs_resume_fs(zfsvfs, newds);
5994 mutex_enter(&zfsvfs->z_os->os_upgrade_lock);
5995 if (zfsvfs->z_os->os_upgrade_id == 0) {
5996 /* clear potential error code and retry */
5997 zfsvfs->z_os->os_upgrade_status = 0;
5998 mutex_exit(&zfsvfs->z_os->os_upgrade_lock);
6000 dsl_pool_config_enter(
6001 dmu_objset_pool(zfsvfs->z_os), FTAG);
6002 dmu_objset_userspace_upgrade(zfsvfs->z_os);
6003 dsl_pool_config_exit(
6004 dmu_objset_pool(zfsvfs->z_os), FTAG);
6006 mutex_exit(&zfsvfs->z_os->os_upgrade_lock);
6009 taskq_wait_id(zfsvfs->z_os->os_spa->spa_upgrade_taskq,
6010 zfsvfs->z_os->os_upgrade_id);
6011 error = zfsvfs->z_os->os_upgrade_status;
6013 zfs_vfs_rele(zfsvfs);
6017 /* XXX kind of reading contents without owning */
6018 error = dmu_objset_hold_flags(zc->zc_name, B_TRUE, FTAG, &os);
6022 mutex_enter(&os->os_upgrade_lock);
6023 if (os->os_upgrade_id == 0) {
6024 /* clear potential error code and retry */
6025 os->os_upgrade_status = 0;
6026 mutex_exit(&os->os_upgrade_lock);
6028 dmu_objset_userspace_upgrade(os);
6030 mutex_exit(&os->os_upgrade_lock);
6033 dsl_pool_rele(dmu_objset_pool(os), FTAG);
6035 taskq_wait_id(os->os_spa->spa_upgrade_taskq, os->os_upgrade_id);
6036 error = os->os_upgrade_status;
6038 dsl_dataset_rele_flags(dmu_objset_ds(os), DS_HOLD_FLAG_DECRYPT,
6046 * zc_name name of filesystem
6052 zfs_ioc_id_quota_upgrade(zfs_cmd_t *zc)
6057 error = dmu_objset_hold_flags(zc->zc_name, B_TRUE, FTAG, &os);
6061 if (dmu_objset_userobjspace_upgradable(os) ||
6062 dmu_objset_projectquota_upgradable(os)) {
6063 mutex_enter(&os->os_upgrade_lock);
6064 if (os->os_upgrade_id == 0) {
6065 /* clear potential error code and retry */
6066 os->os_upgrade_status = 0;
6067 mutex_exit(&os->os_upgrade_lock);
6069 dmu_objset_id_quota_upgrade(os);
6071 mutex_exit(&os->os_upgrade_lock);
6074 dsl_pool_rele(dmu_objset_pool(os), FTAG);
6076 taskq_wait_id(os->os_spa->spa_upgrade_taskq, os->os_upgrade_id);
6077 error = os->os_upgrade_status;
6079 dsl_pool_rele(dmu_objset_pool(os), FTAG);
6082 dsl_dataset_rele_flags(dmu_objset_ds(os), DS_HOLD_FLAG_DECRYPT, FTAG);
6088 zfs_ioc_share(zfs_cmd_t *zc)
6090 return (SET_ERROR(ENOSYS));
6095 * zc_name name of containing filesystem
6096 * zc_obj object # beyond which we want next in-use object #
6099 * zc_obj next in-use object #
6102 zfs_ioc_next_obj(zfs_cmd_t *zc)
6104 objset_t *os = NULL;
6107 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
6111 error = dmu_object_next(os, &zc->zc_obj, B_FALSE, 0);
6113 dmu_objset_rele(os, FTAG);
6119 * zc_name name of filesystem
6120 * zc_value prefix name for snapshot
6121 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
6124 * zc_value short name of new snapshot
6127 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
6133 zfs_file_t *fp = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
6135 return (SET_ERROR(EBADF));
6137 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
6138 (u_longlong_t)ddi_get_lbolt64());
6139 hold_name = kmem_asprintf("%%%s", zc->zc_value);
6141 int error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
6144 (void) strlcpy(zc->zc_value, snap_name,
6145 sizeof (zc->zc_value));
6146 kmem_strfree(snap_name);
6147 kmem_strfree(hold_name);
6148 zfs_onexit_fd_rele(fp);
6154 * zc_name name of "to" snapshot
6155 * zc_value name of "from" snapshot
6156 * zc_cookie file descriptor to write diff data on
6159 * dmu_diff_record_t's to the file descriptor
6162 zfs_ioc_diff(zfs_cmd_t *zc)
6168 if ((fp = zfs_file_get(zc->zc_cookie)) == NULL)
6169 return (SET_ERROR(EBADF));
6171 off = zfs_file_off(fp);
6172 error = dmu_diff(zc->zc_name, zc->zc_value, fp, &off);
6180 zfs_ioc_smb_acl(zfs_cmd_t *zc)
6182 return (SET_ERROR(ENOTSUP));
6187 * "holds" -> { snapname -> holdname (string), ... }
6188 * (optional) "cleanup_fd" -> fd (int32)
6192 * snapname -> error value (int32)
6196 static const zfs_ioc_key_t zfs_keys_hold[] = {
6197 {"holds", DATA_TYPE_NVLIST, 0},
6198 {"cleanup_fd", DATA_TYPE_INT32, ZK_OPTIONAL},
6202 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
6207 int cleanup_fd = -1;
6210 zfs_file_t *fp = NULL;
6212 holds = fnvlist_lookup_nvlist(args, "holds");
6214 /* make sure the user didn't pass us any invalid (empty) tags */
6215 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
6216 pair = nvlist_next_nvpair(holds, pair)) {
6219 error = nvpair_value_string(pair, &htag);
6221 return (SET_ERROR(error));
6223 if (strlen(htag) == 0)
6224 return (SET_ERROR(EINVAL));
6227 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
6228 fp = zfs_onexit_fd_hold(cleanup_fd, &minor);
6230 return (SET_ERROR(EBADF));
6233 error = dsl_dataset_user_hold(holds, minor, errlist);
6235 ASSERT3U(minor, !=, 0);
6236 zfs_onexit_fd_rele(fp);
6238 return (SET_ERROR(error));
6242 * innvl is not used.
6245 * holdname -> time added (uint64 seconds since epoch)
6249 static const zfs_ioc_key_t zfs_keys_get_holds[] = {
6254 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
6257 return (dsl_dataset_get_holds(snapname, outnvl));
6262 * snapname -> { holdname, ... }
6267 * snapname -> error value (int32)
6271 static const zfs_ioc_key_t zfs_keys_release[] = {
6272 {"<snapname>...", DATA_TYPE_NVLIST, ZK_WILDCARDLIST},
6276 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
6279 return (dsl_dataset_user_release(holds, errlist));
6284 * zc_guid flags (ZEVENT_NONBLOCK)
6285 * zc_cleanup_fd zevent file descriptor
6288 * zc_nvlist_dst next nvlist event
6289 * zc_cookie dropped events since last get
6292 zfs_ioc_events_next(zfs_cmd_t *zc)
6295 nvlist_t *event = NULL;
6297 uint64_t dropped = 0;
6300 zfs_file_t *fp = zfs_zevent_fd_hold(zc->zc_cleanup_fd, &minor, &ze);
6302 return (SET_ERROR(EBADF));
6305 error = zfs_zevent_next(ze, &event,
6306 &zc->zc_nvlist_dst_size, &dropped);
6307 if (event != NULL) {
6308 zc->zc_cookie = dropped;
6309 error = put_nvlist(zc, event);
6313 if (zc->zc_guid & ZEVENT_NONBLOCK)
6316 if ((error == 0) || (error != ENOENT))
6319 error = zfs_zevent_wait(ze);
6324 zfs_zevent_fd_rele(fp);
6331 * zc_cookie cleared events count
6334 zfs_ioc_events_clear(zfs_cmd_t *zc)
6338 zfs_zevent_drain_all(&count);
6339 zc->zc_cookie = count;
6346 * zc_guid eid | ZEVENT_SEEK_START | ZEVENT_SEEK_END
6347 * zc_cleanup zevent file descriptor
6350 zfs_ioc_events_seek(zfs_cmd_t *zc)
6356 zfs_file_t *fp = zfs_zevent_fd_hold(zc->zc_cleanup_fd, &minor, &ze);
6358 return (SET_ERROR(EBADF));
6360 error = zfs_zevent_seek(ze, zc->zc_guid);
6361 zfs_zevent_fd_rele(fp);
6368 * zc_name name of later filesystem or snapshot
6369 * zc_value full name of old snapshot or bookmark
6372 * zc_cookie space in bytes
6373 * zc_objset_type compressed space in bytes
6374 * zc_perm_action uncompressed space in bytes
6377 zfs_ioc_space_written(zfs_cmd_t *zc)
6383 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
6386 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
6388 dsl_pool_rele(dp, FTAG);
6391 if (strchr(zc->zc_value, '#') != NULL) {
6392 zfs_bookmark_phys_t bmp;
6393 error = dsl_bookmark_lookup(dp, zc->zc_value,
6396 error = dsl_dataset_space_written_bookmark(&bmp, new,
6398 &zc->zc_objset_type, &zc->zc_perm_action);
6402 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
6405 error = dsl_dataset_space_written(old, new,
6407 &zc->zc_objset_type, &zc->zc_perm_action);
6408 dsl_dataset_rele(old, FTAG);
6411 dsl_dataset_rele(new, FTAG);
6412 dsl_pool_rele(dp, FTAG);
6418 * "firstsnap" -> snapshot name
6422 * "used" -> space in bytes
6423 * "compressed" -> compressed space in bytes
6424 * "uncompressed" -> uncompressed space in bytes
6427 static const zfs_ioc_key_t zfs_keys_space_snaps[] = {
6428 {"firstsnap", DATA_TYPE_STRING, 0},
6432 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
6436 dsl_dataset_t *new, *old;
6438 uint64_t used, comp, uncomp;
6440 firstsnap = fnvlist_lookup_string(innvl, "firstsnap");
6442 error = dsl_pool_hold(lastsnap, FTAG, &dp);
6446 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
6447 if (error == 0 && !new->ds_is_snapshot) {
6448 dsl_dataset_rele(new, FTAG);
6449 error = SET_ERROR(EINVAL);
6452 dsl_pool_rele(dp, FTAG);
6455 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
6456 if (error == 0 && !old->ds_is_snapshot) {
6457 dsl_dataset_rele(old, FTAG);
6458 error = SET_ERROR(EINVAL);
6461 dsl_dataset_rele(new, FTAG);
6462 dsl_pool_rele(dp, FTAG);
6466 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
6467 dsl_dataset_rele(old, FTAG);
6468 dsl_dataset_rele(new, FTAG);
6469 dsl_pool_rele(dp, FTAG);
6470 fnvlist_add_uint64(outnvl, "used", used);
6471 fnvlist_add_uint64(outnvl, "compressed", comp);
6472 fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
6478 * "fd" -> file descriptor to write stream to (int32)
6479 * (optional) "fromsnap" -> full snap name to send an incremental from
6480 * (optional) "largeblockok" -> (value ignored)
6481 * indicates that blocks > 128KB are permitted
6482 * (optional) "embedok" -> (value ignored)
6483 * presence indicates DRR_WRITE_EMBEDDED records are permitted
6484 * (optional) "compressok" -> (value ignored)
6485 * presence indicates compressed DRR_WRITE records are permitted
6486 * (optional) "rawok" -> (value ignored)
6487 * presence indicates raw encrypted records should be used.
6488 * (optional) "savedok" -> (value ignored)
6489 * presence indicates we should send a partially received snapshot
6490 * (optional) "resume_object" and "resume_offset" -> (uint64)
6491 * if present, resume send stream from specified object and offset.
6492 * (optional) "redactbook" -> (string)
6493 * if present, use this bookmark's redaction list to generate a redacted
6499 static const zfs_ioc_key_t zfs_keys_send_new[] = {
6500 {"fd", DATA_TYPE_INT32, 0},
6501 {"fromsnap", DATA_TYPE_STRING, ZK_OPTIONAL},
6502 {"largeblockok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
6503 {"embedok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
6504 {"compressok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
6505 {"rawok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
6506 {"savedok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
6507 {"resume_object", DATA_TYPE_UINT64, ZK_OPTIONAL},
6508 {"resume_offset", DATA_TYPE_UINT64, ZK_OPTIONAL},
6509 {"redactbook", DATA_TYPE_STRING, ZK_OPTIONAL},
6513 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
6518 char *fromname = NULL;
6521 boolean_t largeblockok;
6523 boolean_t compressok;
6526 uint64_t resumeobj = 0;
6527 uint64_t resumeoff = 0;
6528 char *redactbook = NULL;
6530 fd = fnvlist_lookup_int32(innvl, "fd");
6532 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
6534 largeblockok = nvlist_exists(innvl, "largeblockok");
6535 embedok = nvlist_exists(innvl, "embedok");
6536 compressok = nvlist_exists(innvl, "compressok");
6537 rawok = nvlist_exists(innvl, "rawok");
6538 savedok = nvlist_exists(innvl, "savedok");
6540 (void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj);
6541 (void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff);
6543 (void) nvlist_lookup_string(innvl, "redactbook", &redactbook);
6545 if ((fp = zfs_file_get(fd)) == NULL)
6546 return (SET_ERROR(EBADF));
6548 off = zfs_file_off(fp);
6550 dmu_send_outparams_t out = {0};
6551 out.dso_outfunc = dump_bytes;
6553 out.dso_dryrun = B_FALSE;
6554 error = dmu_send(snapname, fromname, embedok, largeblockok,
6555 compressok, rawok, savedok, resumeobj, resumeoff,
6556 redactbook, fd, &off, &out);
6563 send_space_sum(objset_t *os, void *buf, int len, void *arg)
6565 (void) os, (void) buf;
6566 uint64_t *size = arg;
6573 * Determine approximately how large a zfs send stream will be -- the number
6574 * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
6577 * (optional) "from" -> full snap or bookmark name to send an incremental
6579 * (optional) "largeblockok" -> (value ignored)
6580 * indicates that blocks > 128KB are permitted
6581 * (optional) "embedok" -> (value ignored)
6582 * presence indicates DRR_WRITE_EMBEDDED records are permitted
6583 * (optional) "compressok" -> (value ignored)
6584 * presence indicates compressed DRR_WRITE records are permitted
6585 * (optional) "rawok" -> (value ignored)
6586 * presence indicates raw encrypted records should be used.
6587 * (optional) "resume_object" and "resume_offset" -> (uint64)
6588 * if present, resume send stream from specified object and offset.
6589 * (optional) "fd" -> file descriptor to use as a cookie for progress
6594 * "space" -> bytes of space (uint64)
6597 static const zfs_ioc_key_t zfs_keys_send_space[] = {
6598 {"from", DATA_TYPE_STRING, ZK_OPTIONAL},
6599 {"fromsnap", DATA_TYPE_STRING, ZK_OPTIONAL},
6600 {"largeblockok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
6601 {"embedok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
6602 {"compressok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
6603 {"rawok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
6604 {"fd", DATA_TYPE_INT32, ZK_OPTIONAL},
6605 {"redactbook", DATA_TYPE_STRING, ZK_OPTIONAL},
6606 {"resume_object", DATA_TYPE_UINT64, ZK_OPTIONAL},
6607 {"resume_offset", DATA_TYPE_UINT64, ZK_OPTIONAL},
6608 {"bytes", DATA_TYPE_UINT64, ZK_OPTIONAL},
6612 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
6615 dsl_dataset_t *tosnap;
6616 dsl_dataset_t *fromsnap = NULL;
6618 char *fromname = NULL;
6619 char *redactlist_book = NULL;
6620 boolean_t largeblockok;
6622 boolean_t compressok;
6626 boolean_t full_estimate = B_FALSE;
6627 uint64_t resumeobj = 0;
6628 uint64_t resumeoff = 0;
6629 uint64_t resume_bytes = 0;
6631 zfs_bookmark_phys_t zbm = {0};
6633 error = dsl_pool_hold(snapname, FTAG, &dp);
6637 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
6639 dsl_pool_rele(dp, FTAG);
6642 (void) nvlist_lookup_int32(innvl, "fd", &fd);
6644 largeblockok = nvlist_exists(innvl, "largeblockok");
6645 embedok = nvlist_exists(innvl, "embedok");
6646 compressok = nvlist_exists(innvl, "compressok");
6647 rawok = nvlist_exists(innvl, "rawok");
6648 savedok = nvlist_exists(innvl, "savedok");
6649 boolean_t from = (nvlist_lookup_string(innvl, "from", &fromname) == 0);
6650 boolean_t altbook = (nvlist_lookup_string(innvl, "redactbook",
6651 &redactlist_book) == 0);
6653 (void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj);
6654 (void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff);
6655 (void) nvlist_lookup_uint64(innvl, "bytes", &resume_bytes);
6658 full_estimate = B_TRUE;
6660 if (strchr(fromname, '#')) {
6661 error = dsl_bookmark_lookup(dp, fromname, tosnap, &zbm);
6664 * dsl_bookmark_lookup() will fail with EXDEV if
6665 * the from-bookmark and tosnap are at the same txg.
6666 * However, it's valid to do a send (and therefore,
6667 * a send estimate) from and to the same time point,
6668 * if the bookmark is redacted (the incremental send
6669 * can change what's redacted on the target). In
6670 * this case, dsl_bookmark_lookup() fills in zbm
6671 * but returns EXDEV. Ignore this error.
6673 if (error == EXDEV && zbm.zbm_redaction_obj != 0 &&
6675 dsl_dataset_phys(tosnap)->ds_guid)
6679 dsl_dataset_rele(tosnap, FTAG);
6680 dsl_pool_rele(dp, FTAG);
6683 if (zbm.zbm_redaction_obj != 0 || !(zbm.zbm_flags &
6684 ZBM_FLAG_HAS_FBN)) {
6685 full_estimate = B_TRUE;
6687 } else if (strchr(fromname, '@')) {
6688 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
6690 dsl_dataset_rele(tosnap, FTAG);
6691 dsl_pool_rele(dp, FTAG);
6695 if (!dsl_dataset_is_before(tosnap, fromsnap, 0)) {
6696 full_estimate = B_TRUE;
6697 dsl_dataset_rele(fromsnap, FTAG);
6701 * from is not properly formatted as a snapshot or
6704 dsl_dataset_rele(tosnap, FTAG);
6705 dsl_pool_rele(dp, FTAG);
6706 return (SET_ERROR(EINVAL));
6710 if (full_estimate) {
6711 dmu_send_outparams_t out = {0};
6713 out.dso_outfunc = send_space_sum;
6714 out.dso_arg = &space;
6715 out.dso_dryrun = B_TRUE;
6717 * We have to release these holds so dmu_send can take them. It
6718 * will do all the error checking we need.
6720 dsl_dataset_rele(tosnap, FTAG);
6721 dsl_pool_rele(dp, FTAG);
6722 error = dmu_send(snapname, fromname, embedok, largeblockok,
6723 compressok, rawok, savedok, resumeobj, resumeoff,
6724 redactlist_book, fd, &off, &out);
6726 error = dmu_send_estimate_fast(tosnap, fromsnap,
6727 (from && strchr(fromname, '#') != NULL ? &zbm : NULL),
6728 compressok || rawok, savedok, &space);
6729 space -= resume_bytes;
6730 if (fromsnap != NULL)
6731 dsl_dataset_rele(fromsnap, FTAG);
6732 dsl_dataset_rele(tosnap, FTAG);
6733 dsl_pool_rele(dp, FTAG);
6736 fnvlist_add_uint64(outnvl, "space", space);
6742 * Sync the currently open TXG to disk for the specified pool.
6743 * This is somewhat similar to 'zfs_sync()'.
6744 * For cases that do not result in error this ioctl will wait for
6745 * the currently open TXG to commit before returning back to the caller.
6748 * "force" -> when true, force uberblock update even if there is no dirty data.
6749 * In addition this will cause the vdev configuration to be written
6750 * out including updating the zpool cache file. (boolean_t)
6755 static const zfs_ioc_key_t zfs_keys_pool_sync[] = {
6756 {"force", DATA_TYPE_BOOLEAN_VALUE, 0},
6760 zfs_ioc_pool_sync(const char *pool, nvlist_t *innvl, nvlist_t *onvl)
6764 boolean_t rc, force = B_FALSE;
6767 if ((err = spa_open(pool, &spa, FTAG)) != 0)
6771 err = nvlist_lookup_boolean_value(innvl, "force", &rc);
6777 spa_config_enter(spa, SCL_CONFIG, FTAG, RW_WRITER);
6778 vdev_config_dirty(spa->spa_root_vdev);
6779 spa_config_exit(spa, SCL_CONFIG, FTAG);
6781 txg_wait_synced(spa_get_dsl(spa), 0);
6783 spa_close(spa, FTAG);
6789 * Load a user's wrapping key into the kernel.
6791 * "hidden_args" -> { "wkeydata" -> value }
6792 * raw uint8_t array of encryption wrapping key data (32 bytes)
6793 * (optional) "noop" -> (value ignored)
6794 * presence indicated key should only be verified, not loaded
6797 static const zfs_ioc_key_t zfs_keys_load_key[] = {
6798 {"hidden_args", DATA_TYPE_NVLIST, 0},
6799 {"noop", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
6803 zfs_ioc_load_key(const char *dsname, nvlist_t *innvl, nvlist_t *outnvl)
6807 dsl_crypto_params_t *dcp = NULL;
6808 nvlist_t *hidden_args;
6809 boolean_t noop = nvlist_exists(innvl, "noop");
6811 if (strchr(dsname, '@') != NULL || strchr(dsname, '%') != NULL) {
6812 ret = SET_ERROR(EINVAL);
6816 hidden_args = fnvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS);
6818 ret = dsl_crypto_params_create_nvlist(DCP_CMD_NONE, NULL,
6823 ret = spa_keystore_load_wkey(dsname, dcp, noop);
6827 dsl_crypto_params_free(dcp, noop);
6832 dsl_crypto_params_free(dcp, B_TRUE);
6837 * Unload a user's wrapping key from the kernel.
6838 * Both innvl and outnvl are unused.
6840 static const zfs_ioc_key_t zfs_keys_unload_key[] = {
6845 zfs_ioc_unload_key(const char *dsname, nvlist_t *innvl, nvlist_t *outnvl)
6847 (void) innvl, (void) outnvl;
6850 if (strchr(dsname, '@') != NULL || strchr(dsname, '%') != NULL) {
6851 ret = (SET_ERROR(EINVAL));
6855 ret = spa_keystore_unload_wkey(dsname);
6864 * Changes a user's wrapping key used to decrypt a dataset. The keyformat,
6865 * keylocation, pbkdf2salt, and pbkdf2iters properties can also be specified
6866 * here to change how the key is derived in userspace.
6869 * "hidden_args" (optional) -> { "wkeydata" -> value }
6870 * raw uint8_t array of new encryption wrapping key data (32 bytes)
6871 * "props" (optional) -> { prop -> value }
6876 static const zfs_ioc_key_t zfs_keys_change_key[] = {
6877 {"crypt_cmd", DATA_TYPE_UINT64, ZK_OPTIONAL},
6878 {"hidden_args", DATA_TYPE_NVLIST, ZK_OPTIONAL},
6879 {"props", DATA_TYPE_NVLIST, ZK_OPTIONAL},
6883 zfs_ioc_change_key(const char *dsname, nvlist_t *innvl, nvlist_t *outnvl)
6887 uint64_t cmd = DCP_CMD_NONE;
6888 dsl_crypto_params_t *dcp = NULL;
6889 nvlist_t *args = NULL, *hidden_args = NULL;
6891 if (strchr(dsname, '@') != NULL || strchr(dsname, '%') != NULL) {
6892 ret = (SET_ERROR(EINVAL));
6896 (void) nvlist_lookup_uint64(innvl, "crypt_cmd", &cmd);
6897 (void) nvlist_lookup_nvlist(innvl, "props", &args);
6898 (void) nvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS, &hidden_args);
6900 ret = dsl_crypto_params_create_nvlist(cmd, args, hidden_args, &dcp);
6904 ret = spa_keystore_change_key(dsname, dcp);
6908 dsl_crypto_params_free(dcp, B_FALSE);
6913 dsl_crypto_params_free(dcp, B_TRUE);
6917 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
6920 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
6921 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
6922 boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
6924 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
6926 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
6927 ASSERT3U(ioc, <, ZFS_IOC_LAST);
6928 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
6929 ASSERT3P(vec->zvec_func, ==, NULL);
6931 vec->zvec_legacy_func = func;
6932 vec->zvec_secpolicy = secpolicy;
6933 vec->zvec_namecheck = namecheck;
6934 vec->zvec_allow_log = log_history;
6935 vec->zvec_pool_check = pool_check;
6939 * See the block comment at the beginning of this file for details on
6940 * each argument to this function.
6943 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
6944 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
6945 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
6946 boolean_t allow_log, const zfs_ioc_key_t *nvl_keys, size_t num_keys)
6948 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
6950 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
6951 ASSERT3U(ioc, <, ZFS_IOC_LAST);
6952 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
6953 ASSERT3P(vec->zvec_func, ==, NULL);
6955 /* if we are logging, the name must be valid */
6956 ASSERT(!allow_log || namecheck != NO_NAME);
6958 vec->zvec_name = name;
6959 vec->zvec_func = func;
6960 vec->zvec_secpolicy = secpolicy;
6961 vec->zvec_namecheck = namecheck;
6962 vec->zvec_pool_check = pool_check;
6963 vec->zvec_smush_outnvlist = smush_outnvlist;
6964 vec->zvec_allow_log = allow_log;
6965 vec->zvec_nvl_keys = nvl_keys;
6966 vec->zvec_nvl_key_count = num_keys;
6970 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
6971 zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
6972 zfs_ioc_poolcheck_t pool_check)
6974 zfs_ioctl_register_legacy(ioc, func, secpolicy,
6975 POOL_NAME, log_history, pool_check);
6979 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
6980 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
6982 zfs_ioctl_register_legacy(ioc, func, secpolicy,
6983 DATASET_NAME, B_FALSE, pool_check);
6987 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
6989 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
6990 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
6994 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
6995 zfs_secpolicy_func_t *secpolicy)
6997 zfs_ioctl_register_legacy(ioc, func, secpolicy,
6998 NO_NAME, B_FALSE, POOL_CHECK_NONE);
7002 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
7003 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
7005 zfs_ioctl_register_legacy(ioc, func, secpolicy,
7006 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
7010 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
7012 zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
7013 zfs_secpolicy_read);
7017 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
7018 zfs_secpolicy_func_t *secpolicy)
7020 zfs_ioctl_register_legacy(ioc, func, secpolicy,
7021 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
7025 zfs_ioctl_init(void)
7027 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
7028 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
7029 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7030 zfs_keys_snapshot, ARRAY_SIZE(zfs_keys_snapshot));
7032 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
7033 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
7034 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE,
7035 zfs_keys_log_history, ARRAY_SIZE(zfs_keys_log_history));
7037 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
7038 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
7039 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
7040 zfs_keys_space_snaps, ARRAY_SIZE(zfs_keys_space_snaps));
7042 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
7043 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
7044 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
7045 zfs_keys_send_new, ARRAY_SIZE(zfs_keys_send_new));
7047 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
7048 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
7049 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
7050 zfs_keys_send_space, ARRAY_SIZE(zfs_keys_send_space));
7052 zfs_ioctl_register("create", ZFS_IOC_CREATE,
7053 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
7054 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7055 zfs_keys_create, ARRAY_SIZE(zfs_keys_create));
7057 zfs_ioctl_register("clone", ZFS_IOC_CLONE,
7058 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
7059 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7060 zfs_keys_clone, ARRAY_SIZE(zfs_keys_clone));
7062 zfs_ioctl_register("remap", ZFS_IOC_REMAP,
7063 zfs_ioc_remap, zfs_secpolicy_none, DATASET_NAME,
7064 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE,
7065 zfs_keys_remap, ARRAY_SIZE(zfs_keys_remap));
7067 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
7068 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
7069 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7070 zfs_keys_destroy_snaps, ARRAY_SIZE(zfs_keys_destroy_snaps));
7072 zfs_ioctl_register("hold", ZFS_IOC_HOLD,
7073 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
7074 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7075 zfs_keys_hold, ARRAY_SIZE(zfs_keys_hold));
7076 zfs_ioctl_register("release", ZFS_IOC_RELEASE,
7077 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
7078 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7079 zfs_keys_release, ARRAY_SIZE(zfs_keys_release));
7081 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
7082 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
7083 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
7084 zfs_keys_get_holds, ARRAY_SIZE(zfs_keys_get_holds));
7086 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
7087 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
7088 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE,
7089 zfs_keys_rollback, ARRAY_SIZE(zfs_keys_rollback));
7091 zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK,
7092 zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME,
7093 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7094 zfs_keys_bookmark, ARRAY_SIZE(zfs_keys_bookmark));
7096 zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS,
7097 zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME,
7098 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
7099 zfs_keys_get_bookmarks, ARRAY_SIZE(zfs_keys_get_bookmarks));
7101 zfs_ioctl_register("get_bookmark_props", ZFS_IOC_GET_BOOKMARK_PROPS,
7102 zfs_ioc_get_bookmark_props, zfs_secpolicy_read, ENTITY_NAME,
7103 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE, zfs_keys_get_bookmark_props,
7104 ARRAY_SIZE(zfs_keys_get_bookmark_props));
7106 zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS,
7107 zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks,
7109 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7110 zfs_keys_destroy_bookmarks,
7111 ARRAY_SIZE(zfs_keys_destroy_bookmarks));
7113 zfs_ioctl_register("receive", ZFS_IOC_RECV_NEW,
7114 zfs_ioc_recv_new, zfs_secpolicy_recv, DATASET_NAME,
7115 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7116 zfs_keys_recv_new, ARRAY_SIZE(zfs_keys_recv_new));
7117 zfs_ioctl_register("load-key", ZFS_IOC_LOAD_KEY,
7118 zfs_ioc_load_key, zfs_secpolicy_load_key,
7119 DATASET_NAME, POOL_CHECK_SUSPENDED, B_TRUE, B_TRUE,
7120 zfs_keys_load_key, ARRAY_SIZE(zfs_keys_load_key));
7121 zfs_ioctl_register("unload-key", ZFS_IOC_UNLOAD_KEY,
7122 zfs_ioc_unload_key, zfs_secpolicy_load_key,
7123 DATASET_NAME, POOL_CHECK_SUSPENDED, B_TRUE, B_TRUE,
7124 zfs_keys_unload_key, ARRAY_SIZE(zfs_keys_unload_key));
7125 zfs_ioctl_register("change-key", ZFS_IOC_CHANGE_KEY,
7126 zfs_ioc_change_key, zfs_secpolicy_change_key,
7127 DATASET_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY,
7128 B_TRUE, B_TRUE, zfs_keys_change_key,
7129 ARRAY_SIZE(zfs_keys_change_key));
7131 zfs_ioctl_register("sync", ZFS_IOC_POOL_SYNC,
7132 zfs_ioc_pool_sync, zfs_secpolicy_none, POOL_NAME,
7133 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE,
7134 zfs_keys_pool_sync, ARRAY_SIZE(zfs_keys_pool_sync));
7135 zfs_ioctl_register("reopen", ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
7136 zfs_secpolicy_config, POOL_NAME, POOL_CHECK_SUSPENDED, B_TRUE,
7137 B_TRUE, zfs_keys_pool_reopen, ARRAY_SIZE(zfs_keys_pool_reopen));
7139 zfs_ioctl_register("channel_program", ZFS_IOC_CHANNEL_PROGRAM,
7140 zfs_ioc_channel_program, zfs_secpolicy_config,
7141 POOL_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE,
7142 B_TRUE, zfs_keys_channel_program,
7143 ARRAY_SIZE(zfs_keys_channel_program));
7145 zfs_ioctl_register("redact", ZFS_IOC_REDACT,
7146 zfs_ioc_redact, zfs_secpolicy_config, DATASET_NAME,
7147 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7148 zfs_keys_redact, ARRAY_SIZE(zfs_keys_redact));
7150 zfs_ioctl_register("zpool_checkpoint", ZFS_IOC_POOL_CHECKPOINT,
7151 zfs_ioc_pool_checkpoint, zfs_secpolicy_config, POOL_NAME,
7152 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7153 zfs_keys_pool_checkpoint, ARRAY_SIZE(zfs_keys_pool_checkpoint));
7155 zfs_ioctl_register("zpool_discard_checkpoint",
7156 ZFS_IOC_POOL_DISCARD_CHECKPOINT, zfs_ioc_pool_discard_checkpoint,
7157 zfs_secpolicy_config, POOL_NAME,
7158 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7159 zfs_keys_pool_discard_checkpoint,
7160 ARRAY_SIZE(zfs_keys_pool_discard_checkpoint));
7162 zfs_ioctl_register("initialize", ZFS_IOC_POOL_INITIALIZE,
7163 zfs_ioc_pool_initialize, zfs_secpolicy_config, POOL_NAME,
7164 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7165 zfs_keys_pool_initialize, ARRAY_SIZE(zfs_keys_pool_initialize));
7167 zfs_ioctl_register("trim", ZFS_IOC_POOL_TRIM,
7168 zfs_ioc_pool_trim, zfs_secpolicy_config, POOL_NAME,
7169 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7170 zfs_keys_pool_trim, ARRAY_SIZE(zfs_keys_pool_trim));
7172 zfs_ioctl_register("wait", ZFS_IOC_WAIT,
7173 zfs_ioc_wait, zfs_secpolicy_none, POOL_NAME,
7174 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE,
7175 zfs_keys_pool_wait, ARRAY_SIZE(zfs_keys_pool_wait));
7177 zfs_ioctl_register("wait_fs", ZFS_IOC_WAIT_FS,
7178 zfs_ioc_wait_fs, zfs_secpolicy_none, DATASET_NAME,
7179 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE,
7180 zfs_keys_fs_wait, ARRAY_SIZE(zfs_keys_fs_wait));
7182 zfs_ioctl_register("set_bootenv", ZFS_IOC_SET_BOOTENV,
7183 zfs_ioc_set_bootenv, zfs_secpolicy_config, POOL_NAME,
7184 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE,
7185 zfs_keys_set_bootenv, ARRAY_SIZE(zfs_keys_set_bootenv));
7187 zfs_ioctl_register("get_bootenv", ZFS_IOC_GET_BOOTENV,
7188 zfs_ioc_get_bootenv, zfs_secpolicy_none, POOL_NAME,
7189 POOL_CHECK_SUSPENDED, B_FALSE, B_TRUE,
7190 zfs_keys_get_bootenv, ARRAY_SIZE(zfs_keys_get_bootenv));
7192 zfs_ioctl_register("zpool_vdev_get_props", ZFS_IOC_VDEV_GET_PROPS,
7193 zfs_ioc_vdev_get_props, zfs_secpolicy_read, POOL_NAME,
7194 POOL_CHECK_NONE, B_FALSE, B_FALSE, zfs_keys_vdev_get_props,
7195 ARRAY_SIZE(zfs_keys_vdev_get_props));
7197 zfs_ioctl_register("zpool_vdev_set_props", ZFS_IOC_VDEV_SET_PROPS,
7198 zfs_ioc_vdev_set_props, zfs_secpolicy_config, POOL_NAME,
7199 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE,
7200 zfs_keys_vdev_set_props, ARRAY_SIZE(zfs_keys_vdev_set_props));
7202 /* IOCTLS that use the legacy function signature */
7204 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
7205 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
7207 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
7208 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
7209 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
7211 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
7212 zfs_ioc_pool_upgrade);
7213 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
7215 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
7216 zfs_ioc_vdev_remove);
7217 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
7218 zfs_ioc_vdev_set_state);
7219 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
7220 zfs_ioc_vdev_attach);
7221 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
7222 zfs_ioc_vdev_detach);
7223 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
7224 zfs_ioc_vdev_setpath);
7225 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
7226 zfs_ioc_vdev_setfru);
7227 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
7228 zfs_ioc_pool_set_props);
7229 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
7230 zfs_ioc_vdev_split);
7231 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
7232 zfs_ioc_pool_reguid);
7234 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
7235 zfs_ioc_pool_configs, zfs_secpolicy_none);
7236 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
7237 zfs_ioc_pool_tryimport, zfs_secpolicy_config);
7238 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
7239 zfs_ioc_inject_fault, zfs_secpolicy_inject);
7240 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
7241 zfs_ioc_clear_fault, zfs_secpolicy_inject);
7242 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
7243 zfs_ioc_inject_list_next, zfs_secpolicy_inject);
7246 * pool destroy, and export don't log the history as part of
7247 * zfsdev_ioctl, but rather zfs_ioc_pool_export
7248 * does the logging of those commands.
7250 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
7251 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
7252 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
7253 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
7255 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
7256 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
7257 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
7258 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
7260 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
7261 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED);
7262 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
7263 zfs_ioc_dsobj_to_dsname,
7264 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED);
7265 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
7266 zfs_ioc_pool_get_history,
7267 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
7269 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
7270 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
7272 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
7273 zfs_secpolicy_config, B_TRUE, POOL_CHECK_READONLY);
7275 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
7276 zfs_ioc_space_written);
7277 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
7278 zfs_ioc_objset_recvd_props);
7279 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
7281 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
7283 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
7284 zfs_ioc_objset_stats);
7285 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
7286 zfs_ioc_objset_zplprops);
7287 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
7288 zfs_ioc_dataset_list_next);
7289 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
7290 zfs_ioc_snapshot_list_next);
7291 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
7292 zfs_ioc_send_progress);
7294 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
7295 zfs_ioc_diff, zfs_secpolicy_diff);
7296 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
7297 zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
7298 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
7299 zfs_ioc_obj_to_path, zfs_secpolicy_diff);
7300 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
7301 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
7302 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
7303 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
7304 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
7305 zfs_ioc_send, zfs_secpolicy_send);
7307 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
7308 zfs_secpolicy_none);
7309 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
7310 zfs_secpolicy_destroy);
7311 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
7312 zfs_secpolicy_rename);
7313 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
7314 zfs_secpolicy_recv);
7315 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
7316 zfs_secpolicy_promote);
7317 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
7318 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
7319 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
7320 zfs_secpolicy_set_fsacl);
7322 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
7323 zfs_secpolicy_share, POOL_CHECK_NONE);
7324 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
7325 zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
7326 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
7327 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
7328 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
7329 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
7330 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
7331 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
7333 zfs_ioctl_register_legacy(ZFS_IOC_EVENTS_NEXT, zfs_ioc_events_next,
7334 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);
7335 zfs_ioctl_register_legacy(ZFS_IOC_EVENTS_CLEAR, zfs_ioc_events_clear,
7336 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);
7337 zfs_ioctl_register_legacy(ZFS_IOC_EVENTS_SEEK, zfs_ioc_events_seek,
7338 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);
7340 zfs_ioctl_init_os();
7344 * Verify that for non-legacy ioctls the input nvlist
7345 * pairs match against the expected input.
7347 * Possible errors are:
7348 * ZFS_ERR_IOC_ARG_UNAVAIL An unrecognized nvpair was encountered
7349 * ZFS_ERR_IOC_ARG_REQUIRED A required nvpair is missing
7350 * ZFS_ERR_IOC_ARG_BADTYPE Invalid type for nvpair
7353 zfs_check_input_nvpairs(nvlist_t *innvl, const zfs_ioc_vec_t *vec)
7355 const zfs_ioc_key_t *nvl_keys = vec->zvec_nvl_keys;
7356 boolean_t required_keys_found = B_FALSE;
7359 * examine each input pair
7361 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
7362 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
7363 char *name = nvpair_name(pair);
7364 data_type_t type = nvpair_type(pair);
7365 boolean_t identified = B_FALSE;
7368 * check pair against the documented names and type
7370 for (int k = 0; k < vec->zvec_nvl_key_count; k++) {
7371 /* if not a wild card name, check for an exact match */
7372 if ((nvl_keys[k].zkey_flags & ZK_WILDCARDLIST) == 0 &&
7373 strcmp(nvl_keys[k].zkey_name, name) != 0)
7376 identified = B_TRUE;
7378 if (nvl_keys[k].zkey_type != DATA_TYPE_ANY &&
7379 nvl_keys[k].zkey_type != type) {
7380 return (SET_ERROR(ZFS_ERR_IOC_ARG_BADTYPE));
7383 if (nvl_keys[k].zkey_flags & ZK_OPTIONAL)
7386 required_keys_found = B_TRUE;
7390 /* allow an 'optional' key, everything else is invalid */
7392 (strcmp(name, "optional") != 0 ||
7393 type != DATA_TYPE_NVLIST)) {
7394 return (SET_ERROR(ZFS_ERR_IOC_ARG_UNAVAIL));
7398 /* verify that all required keys were found */
7399 for (int k = 0; k < vec->zvec_nvl_key_count; k++) {
7400 if (nvl_keys[k].zkey_flags & ZK_OPTIONAL)
7403 if (nvl_keys[k].zkey_flags & ZK_WILDCARDLIST) {
7404 /* at least one non-optional key is expected here */
7405 if (!required_keys_found)
7406 return (SET_ERROR(ZFS_ERR_IOC_ARG_REQUIRED));
7410 if (!nvlist_exists(innvl, nvl_keys[k].zkey_name))
7411 return (SET_ERROR(ZFS_ERR_IOC_ARG_REQUIRED));
7418 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
7419 zfs_ioc_poolcheck_t check)
7424 ASSERT(type == POOL_NAME || type == DATASET_NAME ||
7425 type == ENTITY_NAME);
7427 if (check & POOL_CHECK_NONE)
7430 error = spa_open(name, &spa, FTAG);
7432 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
7433 error = SET_ERROR(EAGAIN);
7434 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
7435 error = SET_ERROR(EROFS);
7436 spa_close(spa, FTAG);
7442 zfsdev_getminor(zfs_file_t *fp, minor_t *minorp)
7444 zfsdev_state_t *zs, *fpd;
7446 ASSERT(!MUTEX_HELD(&zfsdev_state_lock));
7448 fpd = zfs_file_private(fp);
7450 return (SET_ERROR(EBADF));
7452 mutex_enter(&zfsdev_state_lock);
7454 for (zs = zfsdev_state_list; zs != NULL; zs = zs->zs_next) {
7456 if (zs->zs_minor == -1)
7460 *minorp = fpd->zs_minor;
7461 mutex_exit(&zfsdev_state_lock);
7466 mutex_exit(&zfsdev_state_lock);
7468 return (SET_ERROR(EBADF));
7472 zfsdev_get_state(minor_t minor, enum zfsdev_state_type which)
7476 for (zs = zfsdev_state_list; zs != NULL; zs = zs->zs_next) {
7477 if (zs->zs_minor == minor) {
7481 return (zs->zs_onexit);
7483 return (zs->zs_zevent);
7494 * Find a free minor number. The zfsdev_state_list is expected to
7495 * be short since it is only a list of currently open file handles.
7498 zfsdev_minor_alloc(void)
7500 static minor_t last_minor = 0;
7503 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
7505 for (m = last_minor + 1; m != last_minor; m++) {
7506 if (m > ZFSDEV_MAX_MINOR)
7508 if (zfsdev_get_state(m, ZST_ALL) == NULL) {
7518 zfsdev_state_init(void *priv)
7520 zfsdev_state_t *zs, *zsprev = NULL;
7522 boolean_t newzs = B_FALSE;
7524 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
7526 minor = zfsdev_minor_alloc();
7528 return (SET_ERROR(ENXIO));
7530 for (zs = zfsdev_state_list; zs != NULL; zs = zs->zs_next) {
7531 if (zs->zs_minor == -1)
7537 zs = kmem_zalloc(sizeof (zfsdev_state_t), KM_SLEEP);
7541 zfsdev_private_set_state(priv, zs);
7543 zfs_onexit_init((zfs_onexit_t **)&zs->zs_onexit);
7544 zfs_zevent_init((zfs_zevent_t **)&zs->zs_zevent);
7547 * In order to provide for lock-free concurrent read access
7548 * to the minor list in zfsdev_get_state(), new entries
7549 * must be completely written before linking them into the
7550 * list whereas existing entries are already linked; the last
7551 * operation must be updating zs_minor (from -1 to the new
7555 zs->zs_minor = minor;
7557 zsprev->zs_next = zs;
7560 zs->zs_minor = minor;
7567 zfsdev_state_destroy(void *priv)
7569 zfsdev_state_t *zs = zfsdev_private_get_state(priv);
7572 ASSERT3S(zs->zs_minor, >, 0);
7575 * The last reference to this zfsdev file descriptor is being dropped.
7576 * We don't have to worry about lookup grabbing this state object, and
7577 * zfsdev_state_init() will not try to reuse this object until it is
7578 * invalidated by setting zs_minor to -1. Invalidation must be done
7579 * last, with a memory barrier to ensure ordering. This lets us avoid
7580 * taking the global zfsdev state lock around destruction.
7582 zfs_onexit_destroy(zs->zs_onexit);
7583 zfs_zevent_destroy(zs->zs_zevent);
7584 zs->zs_onexit = NULL;
7585 zs->zs_zevent = NULL;
7591 zfsdev_ioctl_common(uint_t vecnum, zfs_cmd_t *zc, int flag)
7594 const zfs_ioc_vec_t *vec;
7595 char *saved_poolname = NULL;
7596 uint64_t max_nvlist_src_size;
7597 size_t saved_poolname_len = 0;
7598 nvlist_t *innvl = NULL;
7599 fstrans_cookie_t cookie;
7600 hrtime_t start_time = gethrtime();
7604 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
7605 return (SET_ERROR(ZFS_ERR_IOC_CMD_UNAVAIL));
7607 vec = &zfs_ioc_vec[vecnum];
7610 * The registered ioctl list may be sparse, verify that either
7611 * a normal or legacy handler are registered.
7613 if (vec->zvec_func == NULL && vec->zvec_legacy_func == NULL)
7614 return (SET_ERROR(ZFS_ERR_IOC_CMD_UNAVAIL));
7616 zc->zc_iflags = flag & FKIOCTL;
7617 max_nvlist_src_size = zfs_max_nvlist_src_size_os();
7618 if (zc->zc_nvlist_src_size > max_nvlist_src_size) {
7620 * Make sure the user doesn't pass in an insane value for
7621 * zc_nvlist_src_size. We have to check, since we will end
7622 * up allocating that much memory inside of get_nvlist(). This
7623 * prevents a nefarious user from allocating tons of kernel
7626 * Also, we return EINVAL instead of ENOMEM here. The reason
7627 * being that returning ENOMEM from an ioctl() has a special
7628 * connotation; that the user's size value is too small and
7629 * needs to be expanded to hold the nvlist. See
7630 * zcmd_expand_dst_nvlist() for details.
7632 error = SET_ERROR(EINVAL); /* User's size too big */
7634 } else if (zc->zc_nvlist_src_size != 0) {
7635 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
7636 zc->zc_iflags, &innvl);
7642 * Ensure that all pool/dataset names are valid before we pass down to
7645 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
7646 switch (vec->zvec_namecheck) {
7648 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
7649 error = SET_ERROR(EINVAL);
7651 error = pool_status_check(zc->zc_name,
7652 vec->zvec_namecheck, vec->zvec_pool_check);
7656 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
7657 error = SET_ERROR(EINVAL);
7659 error = pool_status_check(zc->zc_name,
7660 vec->zvec_namecheck, vec->zvec_pool_check);
7664 if (entity_namecheck(zc->zc_name, NULL, NULL) != 0) {
7665 error = SET_ERROR(EINVAL);
7667 error = pool_status_check(zc->zc_name,
7668 vec->zvec_namecheck, vec->zvec_pool_check);
7676 * Ensure that all input pairs are valid before we pass them down
7677 * to the lower layers.
7679 * The vectored functions can use fnvlist_lookup_{type} for any
7680 * required pairs since zfs_check_input_nvpairs() confirmed that
7681 * they exist and are of the correct type.
7683 if (error == 0 && vec->zvec_func != NULL) {
7684 error = zfs_check_input_nvpairs(innvl, vec);
7690 cookie = spl_fstrans_mark();
7691 error = vec->zvec_secpolicy(zc, innvl, CRED());
7692 spl_fstrans_unmark(cookie);
7698 /* legacy ioctls can modify zc_name */
7700 * Can't use kmem_strdup() as we might truncate the string and
7701 * kmem_strfree() would then free with incorrect size.
7703 saved_poolname_len = strlen(zc->zc_name) + 1;
7704 saved_poolname = kmem_alloc(saved_poolname_len, KM_SLEEP);
7706 strlcpy(saved_poolname, zc->zc_name, saved_poolname_len);
7707 saved_poolname[strcspn(saved_poolname, "/@#")] = '\0';
7709 if (vec->zvec_func != NULL) {
7713 nvlist_t *lognv = NULL;
7715 ASSERT(vec->zvec_legacy_func == NULL);
7718 * Add the innvl to the lognv before calling the func,
7719 * in case the func changes the innvl.
7721 if (vec->zvec_allow_log) {
7722 lognv = fnvlist_alloc();
7723 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
7725 if (!nvlist_empty(innvl)) {
7726 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
7731 outnvl = fnvlist_alloc();
7732 cookie = spl_fstrans_mark();
7733 error = vec->zvec_func(zc->zc_name, innvl, outnvl);
7734 spl_fstrans_unmark(cookie);
7737 * Some commands can partially execute, modify state, and still
7738 * return an error. In these cases, attempt to record what
7742 (cmd == ZFS_IOC_CHANNEL_PROGRAM && error != EINVAL)) &&
7743 vec->zvec_allow_log &&
7744 spa_open(zc->zc_name, &spa, FTAG) == 0) {
7745 if (!nvlist_empty(outnvl)) {
7746 size_t out_size = fnvlist_size(outnvl);
7747 if (out_size > zfs_history_output_max) {
7748 fnvlist_add_int64(lognv,
7749 ZPOOL_HIST_OUTPUT_SIZE, out_size);
7751 fnvlist_add_nvlist(lognv,
7752 ZPOOL_HIST_OUTPUT_NVL, outnvl);
7756 fnvlist_add_int64(lognv, ZPOOL_HIST_ERRNO,
7759 fnvlist_add_int64(lognv, ZPOOL_HIST_ELAPSED_NS,
7760 gethrtime() - start_time);
7761 (void) spa_history_log_nvl(spa, lognv);
7762 spa_close(spa, FTAG);
7764 fnvlist_free(lognv);
7766 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
7768 if (vec->zvec_smush_outnvlist) {
7769 smusherror = nvlist_smush(outnvl,
7770 zc->zc_nvlist_dst_size);
7772 if (smusherror == 0)
7773 puterror = put_nvlist(zc, outnvl);
7779 nvlist_free(outnvl);
7781 cookie = spl_fstrans_mark();
7782 error = vec->zvec_legacy_func(zc);
7783 spl_fstrans_unmark(cookie);
7788 if (error == 0 && vec->zvec_allow_log) {
7789 char *s = tsd_get(zfs_allow_log_key);
7792 (void) tsd_set(zfs_allow_log_key, kmem_strdup(saved_poolname));
7794 if (saved_poolname != NULL)
7795 kmem_free(saved_poolname, saved_poolname_len);
7805 if ((error = zvol_init()) != 0)
7808 spa_init(SPA_MODE_READ | SPA_MODE_WRITE);
7813 mutex_init(&zfsdev_state_lock, NULL, MUTEX_DEFAULT, NULL);
7814 zfsdev_state_list = kmem_zalloc(sizeof (zfsdev_state_t), KM_SLEEP);
7815 zfsdev_state_list->zs_minor = -1;
7817 if ((error = zfsdev_attach()) != 0)
7820 tsd_create(&zfs_fsyncer_key, NULL);
7821 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
7822 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
7836 zfsdev_state_t *zs, *zsnext = NULL;
7840 mutex_destroy(&zfsdev_state_lock);
7842 for (zs = zfsdev_state_list; zs != NULL; zs = zsnext) {
7843 zsnext = zs->zs_next;
7845 zfs_onexit_destroy(zs->zs_onexit);
7847 zfs_zevent_destroy(zs->zs_zevent);
7848 kmem_free(zs, sizeof (zfsdev_state_t));
7851 zfs_ereport_taskq_fini(); /* run before zfs_fini() on Linux */
7856 tsd_destroy(&zfs_fsyncer_key);
7857 tsd_destroy(&rrw_tsd_key);
7858 tsd_destroy(&zfs_allow_log_key);
7861 ZFS_MODULE_PARAM(zfs, zfs_, max_nvlist_src_size, ULONG, ZMOD_RW,
7862 "Maximum size in bytes allowed for src nvlist passed with ZFS ioctls");
7864 ZFS_MODULE_PARAM(zfs, zfs_, history_output_max, ULONG, ZMOD_RW,
7865 "Maximum size in bytes of ZFS ioctl output that will be logged");