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, 2017 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) 2017, loli10K <ezomori.nozomu@gmail.com>. All rights reserved.
37 * Copyright (c) 2017 Datto Inc. All rights reserved.
38 * Copyright 2017 RackTop Systems.
39 * Copyright (c) 2017 Open-E, Inc. All Rights Reserved.
45 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
46 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
48 * There are two ways that we handle ioctls: the legacy way where almost
49 * all of the logic is in the ioctl callback, and the new way where most
50 * of the marshalling is handled in the common entry point, zfsdev_ioctl().
52 * Non-legacy ioctls should be registered by calling
53 * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked
54 * from userland by lzc_ioctl().
56 * The registration arguments are as follows:
59 * The name of the ioctl. This is used for history logging. If the
60 * ioctl returns successfully (the callback returns 0), and allow_log
61 * is true, then a history log entry will be recorded with the input &
62 * output nvlists. The log entry can be printed with "zpool history -i".
65 * The ioctl request number, which userland will pass to ioctl(2).
66 * The ioctl numbers can change from release to release, because
67 * the caller (libzfs) must be matched to the kernel.
69 * zfs_secpolicy_func_t *secpolicy
70 * This function will be called before the zfs_ioc_func_t, to
71 * determine if this operation is permitted. It should return EPERM
72 * on failure, and 0 on success. Checks include determining if the
73 * dataset is visible in this zone, and if the user has either all
74 * zfs privileges in the zone (SYS_MOUNT), or has been granted permission
75 * to do this operation on this dataset with "zfs allow".
77 * zfs_ioc_namecheck_t namecheck
78 * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
79 * name, a dataset name, or nothing. If the name is not well-formed,
80 * the ioctl will fail and the callback will not be called.
81 * Therefore, the callback can assume that the name is well-formed
82 * (e.g. is null-terminated, doesn't have more than one '@' character,
83 * doesn't have invalid characters).
85 * zfs_ioc_poolcheck_t pool_check
86 * This specifies requirements on the pool state. If the pool does
87 * not meet them (is suspended or is readonly), the ioctl will fail
88 * and the callback will not be called. If any checks are specified
89 * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
90 * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED |
91 * POOL_CHECK_READONLY).
93 * boolean_t smush_outnvlist
94 * If smush_outnvlist is true, then the output is presumed to be a
95 * list of errors, and it will be "smushed" down to fit into the
96 * caller's buffer, by removing some entries and replacing them with a
97 * single "N_MORE_ERRORS" entry indicating how many were removed. See
98 * nvlist_smush() for details. If smush_outnvlist is false, and the
99 * outnvlist does not fit into the userland-provided buffer, then the
100 * ioctl will fail with ENOMEM.
102 * zfs_ioc_func_t *func
103 * The callback function that will perform the operation.
105 * The callback should return 0 on success, or an error number on
106 * failure. If the function fails, the userland ioctl will return -1,
107 * and errno will be set to the callback's return value. The callback
108 * will be called with the following arguments:
111 * The name of the pool or dataset to operate on, from
112 * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the
113 * expected type (pool, dataset, or none).
116 * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or
117 * NULL if no input nvlist was provided. Changes to this nvlist are
118 * ignored. If the input nvlist could not be deserialized, the
119 * ioctl will fail and the callback will not be called.
122 * The output nvlist, initially empty. The callback can fill it in,
123 * and it will be returned to userland by serializing it into
124 * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization
125 * fails (e.g. because the caller didn't supply a large enough
126 * buffer), then the overall ioctl will fail. See the
127 * 'smush_nvlist' argument above for additional behaviors.
129 * There are two typical uses of the output nvlist:
130 * - To return state, e.g. property values. In this case,
131 * smush_outnvlist should be false. If the buffer was not large
132 * enough, the caller will reallocate a larger buffer and try
135 * - To return multiple errors from an ioctl which makes on-disk
136 * changes. In this case, smush_outnvlist should be true.
137 * Ioctls which make on-disk modifications should generally not
138 * use the outnvl if they succeed, because the caller can not
139 * distinguish between the operation failing, and
140 * deserialization failing.
143 #include <sys/types.h>
144 #include <sys/param.h>
145 #include <sys/errno.h>
147 #include <sys/file.h>
148 #include <sys/kmem.h>
149 #include <sys/cmn_err.h>
150 #include <sys/stat.h>
151 #include <sys/zfs_ioctl.h>
152 #include <sys/zfs_vfsops.h>
153 #include <sys/zfs_znode.h>
156 #include <sys/spa_impl.h>
157 #include <sys/vdev.h>
158 #include <sys/vdev_impl.h>
160 #include <sys/dsl_dir.h>
161 #include <sys/dsl_dataset.h>
162 #include <sys/dsl_prop.h>
163 #include <sys/dsl_deleg.h>
164 #include <sys/dmu_objset.h>
165 #include <sys/dmu_impl.h>
166 #include <sys/dmu_tx.h>
167 #include <sys/sunddi.h>
168 #include <sys/policy.h>
169 #include <sys/zone.h>
170 #include <sys/nvpair.h>
171 #include <sys/pathname.h>
173 #include <sys/fs/zfs.h>
174 #include <sys/zfs_ctldir.h>
175 #include <sys/zfs_dir.h>
176 #include <sys/zfs_onexit.h>
177 #include <sys/zvol.h>
178 #include <sys/dsl_scan.h>
179 #include <sys/fm/util.h>
180 #include <sys/dsl_crypt.h>
182 #include <sys/dmu_send.h>
183 #include <sys/dsl_destroy.h>
184 #include <sys/dsl_bookmark.h>
185 #include <sys/dsl_userhold.h>
186 #include <sys/zfeature.h>
188 #include <sys/zio_checksum.h>
189 #include <sys/vdev_removal.h>
191 #include <linux/miscdevice.h>
192 #include <linux/slab.h>
194 #include "zfs_namecheck.h"
195 #include "zfs_prop.h"
196 #include "zfs_deleg.h"
197 #include "zfs_comutil.h"
199 #include <sys/lua/lua.h>
200 #include <sys/lua/lauxlib.h>
203 * Limit maximum nvlist size. We don't want users passing in insane values
204 * for zc->zc_nvlist_src_size, since we will need to allocate that much memory.
206 #define MAX_NVLIST_SRC_SIZE KMALLOC_MAX_SIZE
208 kmutex_t zfsdev_state_lock;
209 zfsdev_state_t *zfsdev_state_list;
211 extern void zfs_init(void);
212 extern void zfs_fini(void);
214 uint_t zfs_fsyncer_key;
215 extern uint_t rrw_tsd_key;
216 static uint_t zfs_allow_log_key;
218 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *);
219 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *);
220 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *);
226 } zfs_ioc_namecheck_t;
229 POOL_CHECK_NONE = 1 << 0,
230 POOL_CHECK_SUSPENDED = 1 << 1,
231 POOL_CHECK_READONLY = 1 << 2,
232 } zfs_ioc_poolcheck_t;
234 typedef struct zfs_ioc_vec {
235 zfs_ioc_legacy_func_t *zvec_legacy_func;
236 zfs_ioc_func_t *zvec_func;
237 zfs_secpolicy_func_t *zvec_secpolicy;
238 zfs_ioc_namecheck_t zvec_namecheck;
239 boolean_t zvec_allow_log;
240 zfs_ioc_poolcheck_t zvec_pool_check;
241 boolean_t zvec_smush_outnvlist;
242 const char *zvec_name;
245 /* This array is indexed by zfs_userquota_prop_t */
246 static const char *userquota_perms[] = {
247 ZFS_DELEG_PERM_USERUSED,
248 ZFS_DELEG_PERM_USERQUOTA,
249 ZFS_DELEG_PERM_GROUPUSED,
250 ZFS_DELEG_PERM_GROUPQUOTA,
251 ZFS_DELEG_PERM_USEROBJUSED,
252 ZFS_DELEG_PERM_USEROBJQUOTA,
253 ZFS_DELEG_PERM_GROUPOBJUSED,
254 ZFS_DELEG_PERM_GROUPOBJQUOTA,
255 ZFS_DELEG_PERM_PROJECTUSED,
256 ZFS_DELEG_PERM_PROJECTQUOTA,
257 ZFS_DELEG_PERM_PROJECTOBJUSED,
258 ZFS_DELEG_PERM_PROJECTOBJQUOTA,
261 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
262 static int zfs_ioc_id_quota_upgrade(zfs_cmd_t *zc);
263 static int zfs_check_settable(const char *name, nvpair_t *property,
265 static int zfs_check_clearable(char *dataset, nvlist_t *props,
267 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
269 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
270 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
273 history_str_free(char *buf)
275 kmem_free(buf, HIS_MAX_RECORD_LEN);
279 history_str_get(zfs_cmd_t *zc)
283 if (zc->zc_history == 0)
286 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
287 if (copyinstr((void *)(uintptr_t)zc->zc_history,
288 buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
289 history_str_free(buf);
293 buf[HIS_MAX_RECORD_LEN -1] = '\0';
299 * Check to see if the named dataset is currently defined as bootable
302 zfs_is_bootfs(const char *name)
306 if (dmu_objset_hold(name, FTAG, &os) == 0) {
308 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
309 dmu_objset_rele(os, FTAG);
316 * Return non-zero if the spa version is less than requested version.
319 zfs_earlier_version(const char *name, int version)
323 if (spa_open(name, &spa, FTAG) == 0) {
324 if (spa_version(spa) < version) {
325 spa_close(spa, FTAG);
328 spa_close(spa, FTAG);
334 * Return TRUE if the ZPL version is less than requested version.
337 zpl_earlier_version(const char *name, int version)
340 boolean_t rc = B_TRUE;
342 if (dmu_objset_hold(name, FTAG, &os) == 0) {
345 if (dmu_objset_type(os) != DMU_OST_ZFS) {
346 dmu_objset_rele(os, FTAG);
349 /* XXX reading from non-owned objset */
350 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
351 rc = zplversion < version;
352 dmu_objset_rele(os, FTAG);
358 zfs_log_history(zfs_cmd_t *zc)
363 if ((buf = history_str_get(zc)) == NULL)
366 if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
367 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
368 (void) spa_history_log(spa, buf);
369 spa_close(spa, FTAG);
371 history_str_free(buf);
375 * Policy for top-level read operations (list pools). Requires no privileges,
376 * and can be used in the local zone, as there is no associated dataset.
380 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
386 * Policy for dataset read operations (list children, get statistics). Requires
387 * no privileges, but must be visible in the local zone.
391 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
393 if (INGLOBALZONE(curproc) ||
394 zone_dataset_visible(zc->zc_name, NULL))
397 return (SET_ERROR(ENOENT));
401 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
406 * The dataset must be visible by this zone -- check this first
407 * so they don't see EPERM on something they shouldn't know about.
409 if (!INGLOBALZONE(curproc) &&
410 !zone_dataset_visible(dataset, &writable))
411 return (SET_ERROR(ENOENT));
413 if (INGLOBALZONE(curproc)) {
415 * If the fs is zoned, only root can access it from the
418 if (secpolicy_zfs(cr) && zoned)
419 return (SET_ERROR(EPERM));
422 * If we are in a local zone, the 'zoned' property must be set.
425 return (SET_ERROR(EPERM));
427 /* must be writable by this zone */
429 return (SET_ERROR(EPERM));
435 zfs_dozonecheck(const char *dataset, cred_t *cr)
439 if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
440 return (SET_ERROR(ENOENT));
442 return (zfs_dozonecheck_impl(dataset, zoned, cr));
446 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
450 if (dsl_prop_get_int_ds(ds, "zoned", &zoned))
451 return (SET_ERROR(ENOENT));
453 return (zfs_dozonecheck_impl(dataset, zoned, cr));
457 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
458 const char *perm, cred_t *cr)
462 error = zfs_dozonecheck_ds(name, ds, cr);
464 error = secpolicy_zfs(cr);
466 error = dsl_deleg_access_impl(ds, perm, cr);
472 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
479 * First do a quick check for root in the global zone, which
480 * is allowed to do all write_perms. This ensures that zfs_ioc_*
481 * will get to handle nonexistent datasets.
483 if (INGLOBALZONE(curproc) && secpolicy_zfs(cr) == 0)
486 error = dsl_pool_hold(name, FTAG, &dp);
490 error = dsl_dataset_hold(dp, name, FTAG, &ds);
492 dsl_pool_rele(dp, FTAG);
496 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
498 dsl_dataset_rele(ds, FTAG);
499 dsl_pool_rele(dp, FTAG);
504 * Policy for setting the security label property.
506 * Returns 0 for success, non-zero for access and other errors.
509 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
512 char ds_hexsl[MAXNAMELEN];
513 bslabel_t ds_sl, new_sl;
514 boolean_t new_default = FALSE;
516 int needed_priv = -1;
519 /* First get the existing dataset label. */
520 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
521 1, sizeof (ds_hexsl), &ds_hexsl, NULL);
523 return (SET_ERROR(EPERM));
525 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
528 /* The label must be translatable */
529 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
530 return (SET_ERROR(EINVAL));
533 * In a non-global zone, disallow attempts to set a label that
534 * doesn't match that of the zone; otherwise no other checks
537 if (!INGLOBALZONE(curproc)) {
538 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
539 return (SET_ERROR(EPERM));
544 * For global-zone datasets (i.e., those whose zoned property is
545 * "off", verify that the specified new label is valid for the
548 if (dsl_prop_get_integer(name,
549 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
550 return (SET_ERROR(EPERM));
552 if (zfs_check_global_label(name, strval) != 0)
553 return (SET_ERROR(EPERM));
557 * If the existing dataset label is nondefault, check if the
558 * dataset is mounted (label cannot be changed while mounted).
559 * Get the zfsvfs_t; if there isn't one, then the dataset isn't
560 * mounted (or isn't a dataset, doesn't exist, ...).
562 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
564 static char *setsl_tag = "setsl_tag";
567 * Try to own the dataset; abort if there is any error,
568 * (e.g., already mounted, in use, or other error).
570 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE, B_TRUE,
573 return (SET_ERROR(EPERM));
575 dmu_objset_disown(os, B_TRUE, setsl_tag);
578 needed_priv = PRIV_FILE_DOWNGRADE_SL;
582 if (hexstr_to_label(strval, &new_sl) != 0)
583 return (SET_ERROR(EPERM));
585 if (blstrictdom(&ds_sl, &new_sl))
586 needed_priv = PRIV_FILE_DOWNGRADE_SL;
587 else if (blstrictdom(&new_sl, &ds_sl))
588 needed_priv = PRIV_FILE_UPGRADE_SL;
590 /* dataset currently has a default label */
592 needed_priv = PRIV_FILE_UPGRADE_SL;
596 if (needed_priv != -1)
597 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
600 return (SET_ERROR(ENOTSUP));
601 #endif /* HAVE_MLSLABEL */
605 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
611 * Check permissions for special properties.
618 * Disallow setting of 'zoned' from within a local zone.
620 if (!INGLOBALZONE(curproc))
621 return (SET_ERROR(EPERM));
625 case ZFS_PROP_FILESYSTEM_LIMIT:
626 case ZFS_PROP_SNAPSHOT_LIMIT:
627 if (!INGLOBALZONE(curproc)) {
629 char setpoint[ZFS_MAX_DATASET_NAME_LEN];
631 * Unprivileged users are allowed to modify the
632 * limit on things *under* (ie. contained by)
633 * the thing they own.
635 if (dsl_prop_get_integer(dsname, "zoned", &zoned,
637 return (SET_ERROR(EPERM));
638 if (!zoned || strlen(dsname) <= strlen(setpoint))
639 return (SET_ERROR(EPERM));
643 case ZFS_PROP_MLSLABEL:
644 if (!is_system_labeled())
645 return (SET_ERROR(EPERM));
647 if (nvpair_value_string(propval, &strval) == 0) {
650 err = zfs_set_slabel_policy(dsname, strval, CRED());
657 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
662 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
666 error = zfs_dozonecheck(zc->zc_name, cr);
671 * permission to set permissions will be evaluated later in
672 * dsl_deleg_can_allow()
679 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
681 return (zfs_secpolicy_write_perms(zc->zc_name,
682 ZFS_DELEG_PERM_ROLLBACK, cr));
687 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
695 * Generate the current snapshot name from the given objsetid, then
696 * use that name for the secpolicy/zone checks.
698 cp = strchr(zc->zc_name, '@');
700 return (SET_ERROR(EINVAL));
701 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
705 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
707 dsl_pool_rele(dp, FTAG);
711 dsl_dataset_name(ds, zc->zc_name);
713 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
714 ZFS_DELEG_PERM_SEND, cr);
715 dsl_dataset_rele(ds, FTAG);
716 dsl_pool_rele(dp, FTAG);
723 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
725 return (zfs_secpolicy_write_perms(zc->zc_name,
726 ZFS_DELEG_PERM_SEND, cr));
729 #ifdef HAVE_SMB_SHARE
732 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
737 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
738 NO_FOLLOW, NULL, &vp)) != 0)
741 /* Now make sure mntpnt and dataset are ZFS */
743 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
744 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
745 zc->zc_name) != 0)) {
747 return (SET_ERROR(EPERM));
751 return (dsl_deleg_access(zc->zc_name,
752 ZFS_DELEG_PERM_SHARE, cr));
754 #endif /* HAVE_SMB_SHARE */
757 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
759 #ifdef HAVE_SMB_SHARE
760 if (!INGLOBALZONE(curproc))
761 return (SET_ERROR(EPERM));
763 if (secpolicy_nfs(cr) == 0) {
766 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
769 return (SET_ERROR(ENOTSUP));
770 #endif /* HAVE_SMB_SHARE */
774 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
776 #ifdef HAVE_SMB_SHARE
777 if (!INGLOBALZONE(curproc))
778 return (SET_ERROR(EPERM));
780 if (secpolicy_smb(cr) == 0) {
783 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
786 return (SET_ERROR(ENOTSUP));
787 #endif /* HAVE_SMB_SHARE */
791 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
796 * Remove the @bla or /bla from the end of the name to get the parent.
798 (void) strncpy(parent, datasetname, parentsize);
799 cp = strrchr(parent, '@');
803 cp = strrchr(parent, '/');
805 return (SET_ERROR(ENOENT));
813 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
817 if ((error = zfs_secpolicy_write_perms(name,
818 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
821 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
826 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
828 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
832 * Destroying snapshots with delegated permissions requires
833 * descendant mount and destroy permissions.
837 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
840 nvpair_t *pair, *nextpair;
843 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
844 return (SET_ERROR(EINVAL));
845 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
847 nextpair = nvlist_next_nvpair(snaps, pair);
848 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr);
849 if (error == ENOENT) {
851 * Ignore any snapshots that don't exist (we consider
852 * them "already destroyed"). Remove the name from the
853 * nvl here in case the snapshot is created between
854 * now and when we try to destroy it (in which case
855 * we don't want to destroy it since we haven't
856 * checked for permission).
858 fnvlist_remove_nvpair(snaps, pair);
869 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
871 char parentname[ZFS_MAX_DATASET_NAME_LEN];
874 if ((error = zfs_secpolicy_write_perms(from,
875 ZFS_DELEG_PERM_RENAME, cr)) != 0)
878 if ((error = zfs_secpolicy_write_perms(from,
879 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
882 if ((error = zfs_get_parent(to, parentname,
883 sizeof (parentname))) != 0)
886 if ((error = zfs_secpolicy_write_perms(parentname,
887 ZFS_DELEG_PERM_CREATE, cr)) != 0)
890 if ((error = zfs_secpolicy_write_perms(parentname,
891 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
899 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
901 return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
906 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
909 dsl_dataset_t *clone;
912 error = zfs_secpolicy_write_perms(zc->zc_name,
913 ZFS_DELEG_PERM_PROMOTE, cr);
917 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
921 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
924 char parentname[ZFS_MAX_DATASET_NAME_LEN];
925 dsl_dataset_t *origin = NULL;
929 error = dsl_dataset_hold_obj(dd->dd_pool,
930 dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin);
932 dsl_dataset_rele(clone, FTAG);
933 dsl_pool_rele(dp, FTAG);
937 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
938 ZFS_DELEG_PERM_MOUNT, cr);
940 dsl_dataset_name(origin, parentname);
942 error = zfs_secpolicy_write_perms_ds(parentname, origin,
943 ZFS_DELEG_PERM_PROMOTE, cr);
945 dsl_dataset_rele(clone, FTAG);
946 dsl_dataset_rele(origin, FTAG);
948 dsl_pool_rele(dp, FTAG);
954 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
958 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
959 ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
962 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
963 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
966 return (zfs_secpolicy_write_perms(zc->zc_name,
967 ZFS_DELEG_PERM_CREATE, cr));
972 zfs_secpolicy_recv_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
974 return (zfs_secpolicy_recv(zc, innvl, cr));
978 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
980 return (zfs_secpolicy_write_perms(name,
981 ZFS_DELEG_PERM_SNAPSHOT, cr));
985 * Check for permission to create each snapshot in the nvlist.
989 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
995 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
996 return (SET_ERROR(EINVAL));
997 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
998 pair = nvlist_next_nvpair(snaps, pair)) {
999 char *name = nvpair_name(pair);
1000 char *atp = strchr(name, '@');
1003 error = SET_ERROR(EINVAL);
1007 error = zfs_secpolicy_snapshot_perms(name, cr);
1016 * Check for permission to create each snapshot in the nvlist.
1020 zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1024 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
1025 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
1026 char *name = nvpair_name(pair);
1027 char *hashp = strchr(name, '#');
1029 if (hashp == NULL) {
1030 error = SET_ERROR(EINVAL);
1034 error = zfs_secpolicy_write_perms(name,
1035 ZFS_DELEG_PERM_BOOKMARK, cr);
1045 zfs_secpolicy_remap(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1047 return (zfs_secpolicy_write_perms(zc->zc_name,
1048 ZFS_DELEG_PERM_REMAP, cr));
1053 zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1055 nvpair_t *pair, *nextpair;
1058 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1060 char *name = nvpair_name(pair);
1061 char *hashp = strchr(name, '#');
1062 nextpair = nvlist_next_nvpair(innvl, pair);
1064 if (hashp == NULL) {
1065 error = SET_ERROR(EINVAL);
1070 error = zfs_secpolicy_write_perms(name,
1071 ZFS_DELEG_PERM_DESTROY, cr);
1073 if (error == ENOENT) {
1075 * Ignore any filesystems that don't exist (we consider
1076 * their bookmarks "already destroyed"). Remove
1077 * the name from the nvl here in case the filesystem
1078 * is created between now and when we try to destroy
1079 * the bookmark (in which case we don't want to
1080 * destroy it since we haven't checked for permission).
1082 fnvlist_remove_nvpair(innvl, pair);
1094 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1097 * Even root must have a proper TSD so that we know what pool
1100 if (tsd_get(zfs_allow_log_key) == NULL)
1101 return (SET_ERROR(EPERM));
1106 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1108 char parentname[ZFS_MAX_DATASET_NAME_LEN];
1112 if ((error = zfs_get_parent(zc->zc_name, parentname,
1113 sizeof (parentname))) != 0)
1116 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1117 (error = zfs_secpolicy_write_perms(origin,
1118 ZFS_DELEG_PERM_CLONE, cr)) != 0)
1121 if ((error = zfs_secpolicy_write_perms(parentname,
1122 ZFS_DELEG_PERM_CREATE, cr)) != 0)
1125 return (zfs_secpolicy_write_perms(parentname,
1126 ZFS_DELEG_PERM_MOUNT, cr));
1130 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires
1131 * SYS_CONFIG privilege, which is not available in a local zone.
1135 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1137 if (secpolicy_sys_config(cr, B_FALSE) != 0)
1138 return (SET_ERROR(EPERM));
1144 * Policy for object to name lookups.
1148 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1152 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1155 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1160 * Policy for fault injection. Requires all privileges.
1164 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1166 return (secpolicy_zinject(cr));
1171 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1173 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1175 if (prop == ZPROP_INVAL) {
1176 if (!zfs_prop_user(zc->zc_value))
1177 return (SET_ERROR(EINVAL));
1178 return (zfs_secpolicy_write_perms(zc->zc_name,
1179 ZFS_DELEG_PERM_USERPROP, cr));
1181 return (zfs_secpolicy_setprop(zc->zc_name, prop,
1187 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1189 int err = zfs_secpolicy_read(zc, innvl, cr);
1193 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1194 return (SET_ERROR(EINVAL));
1196 if (zc->zc_value[0] == 0) {
1198 * They are asking about a posix uid/gid. If it's
1199 * themself, allow it.
1201 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1202 zc->zc_objset_type == ZFS_PROP_USERQUOTA ||
1203 zc->zc_objset_type == ZFS_PROP_USEROBJUSED ||
1204 zc->zc_objset_type == ZFS_PROP_USEROBJQUOTA) {
1205 if (zc->zc_guid == crgetuid(cr))
1207 } else if (zc->zc_objset_type == ZFS_PROP_GROUPUSED ||
1208 zc->zc_objset_type == ZFS_PROP_GROUPQUOTA ||
1209 zc->zc_objset_type == ZFS_PROP_GROUPOBJUSED ||
1210 zc->zc_objset_type == ZFS_PROP_GROUPOBJQUOTA) {
1211 if (groupmember(zc->zc_guid, cr))
1214 /* else is for project quota/used */
1217 return (zfs_secpolicy_write_perms(zc->zc_name,
1218 userquota_perms[zc->zc_objset_type], cr));
1222 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1224 int err = zfs_secpolicy_read(zc, innvl, cr);
1228 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1229 return (SET_ERROR(EINVAL));
1231 return (zfs_secpolicy_write_perms(zc->zc_name,
1232 userquota_perms[zc->zc_objset_type], cr));
1237 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1239 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1245 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1251 error = nvlist_lookup_nvlist(innvl, "holds", &holds);
1253 return (SET_ERROR(EINVAL));
1255 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
1256 pair = nvlist_next_nvpair(holds, pair)) {
1257 char fsname[ZFS_MAX_DATASET_NAME_LEN];
1258 error = dmu_fsname(nvpair_name(pair), fsname);
1261 error = zfs_secpolicy_write_perms(fsname,
1262 ZFS_DELEG_PERM_HOLD, cr);
1271 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1276 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1277 pair = nvlist_next_nvpair(innvl, pair)) {
1278 char fsname[ZFS_MAX_DATASET_NAME_LEN];
1279 error = dmu_fsname(nvpair_name(pair), fsname);
1282 error = zfs_secpolicy_write_perms(fsname,
1283 ZFS_DELEG_PERM_RELEASE, cr);
1291 * Policy for allowing temporary snapshots to be taken or released
1294 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1297 * A temporary snapshot is the same as a snapshot,
1298 * hold, destroy and release all rolled into one.
1299 * Delegated diff alone is sufficient that we allow this.
1303 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1304 ZFS_DELEG_PERM_DIFF, cr)) == 0)
1307 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1309 error = zfs_secpolicy_hold(zc, innvl, cr);
1311 error = zfs_secpolicy_release(zc, innvl, cr);
1313 error = zfs_secpolicy_destroy(zc, innvl, cr);
1318 zfs_secpolicy_load_key(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1320 return (zfs_secpolicy_write_perms(zc->zc_name,
1321 ZFS_DELEG_PERM_LOAD_KEY, cr));
1325 zfs_secpolicy_change_key(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1327 return (zfs_secpolicy_write_perms(zc->zc_name,
1328 ZFS_DELEG_PERM_CHANGE_KEY, cr));
1332 * Returns the nvlist as specified by the user in the zfs_cmd_t.
1335 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1339 nvlist_t *list = NULL;
1342 * Read in and unpack the user-supplied nvlist.
1345 return (SET_ERROR(EINVAL));
1347 packed = vmem_alloc(size, KM_SLEEP);
1349 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1351 vmem_free(packed, size);
1352 return (SET_ERROR(EFAULT));
1355 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1356 vmem_free(packed, size);
1360 vmem_free(packed, size);
1367 * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1368 * Entries will be removed from the end of the nvlist, and one int32 entry
1369 * named "N_MORE_ERRORS" will be added indicating how many entries were
1373 nvlist_smush(nvlist_t *errors, size_t max)
1377 size = fnvlist_size(errors);
1380 nvpair_t *more_errors;
1384 return (SET_ERROR(ENOMEM));
1386 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1387 more_errors = nvlist_prev_nvpair(errors, NULL);
1390 nvpair_t *pair = nvlist_prev_nvpair(errors,
1392 fnvlist_remove_nvpair(errors, pair);
1394 size = fnvlist_size(errors);
1395 } while (size > max);
1397 fnvlist_remove_nvpair(errors, more_errors);
1398 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1399 ASSERT3U(fnvlist_size(errors), <=, max);
1406 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1408 char *packed = NULL;
1412 size = fnvlist_size(nvl);
1414 if (size > zc->zc_nvlist_dst_size) {
1415 error = SET_ERROR(ENOMEM);
1417 packed = fnvlist_pack(nvl, &size);
1418 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1419 size, zc->zc_iflags) != 0)
1420 error = SET_ERROR(EFAULT);
1421 fnvlist_pack_free(packed, size);
1424 zc->zc_nvlist_dst_size = size;
1425 zc->zc_nvlist_dst_filled = B_TRUE;
1430 getzfsvfs_impl(objset_t *os, zfsvfs_t **zfvp)
1433 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1434 return (SET_ERROR(EINVAL));
1437 mutex_enter(&os->os_user_ptr_lock);
1438 *zfvp = dmu_objset_get_user(os);
1439 /* bump s_active only when non-zero to prevent umount race */
1440 if (*zfvp == NULL || (*zfvp)->z_sb == NULL ||
1441 !atomic_inc_not_zero(&((*zfvp)->z_sb->s_active))) {
1442 error = SET_ERROR(ESRCH);
1444 mutex_exit(&os->os_user_ptr_lock);
1449 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1454 error = dmu_objset_hold(dsname, FTAG, &os);
1458 error = getzfsvfs_impl(os, zfvp);
1459 dmu_objset_rele(os, FTAG);
1464 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1465 * case its z_sb will be NULL, and it will be opened as the owner.
1466 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1467 * which prevents all inode ops from running.
1470 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1474 if (getzfsvfs(name, zfvp) != 0)
1475 error = zfsvfs_create(name, B_FALSE, zfvp);
1477 rrm_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER :
1479 if ((*zfvp)->z_unmounted) {
1481 * XXX we could probably try again, since the unmounting
1482 * thread should be just about to disassociate the
1483 * objset from the zfsvfs.
1485 rrm_exit(&(*zfvp)->z_teardown_lock, tag);
1486 return (SET_ERROR(EBUSY));
1493 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1495 rrm_exit(&zfsvfs->z_teardown_lock, tag);
1498 deactivate_super(zfsvfs->z_sb);
1500 dmu_objset_disown(zfsvfs->z_os, B_TRUE, zfsvfs);
1501 zfsvfs_free(zfsvfs);
1506 zfs_ioc_pool_create(zfs_cmd_t *zc)
1509 nvlist_t *config, *props = NULL;
1510 nvlist_t *rootprops = NULL;
1511 nvlist_t *zplprops = NULL;
1512 dsl_crypto_params_t *dcp = NULL;
1513 char *spa_name = zc->zc_name;
1515 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1516 zc->zc_iflags, &config)))
1519 if (zc->zc_nvlist_src_size != 0 && (error =
1520 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1521 zc->zc_iflags, &props))) {
1522 nvlist_free(config);
1527 nvlist_t *nvl = NULL;
1528 nvlist_t *hidden_args = NULL;
1529 uint64_t version = SPA_VERSION;
1532 (void) nvlist_lookup_uint64(props,
1533 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1534 if (!SPA_VERSION_IS_SUPPORTED(version)) {
1535 error = SET_ERROR(EINVAL);
1536 goto pool_props_bad;
1538 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1540 error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1542 nvlist_free(config);
1546 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1549 (void) nvlist_lookup_nvlist(props, ZPOOL_HIDDEN_ARGS,
1551 error = dsl_crypto_params_create_nvlist(DCP_CMD_NONE,
1552 rootprops, hidden_args, &dcp);
1554 nvlist_free(config);
1558 (void) nvlist_remove_all(props, ZPOOL_HIDDEN_ARGS);
1560 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1561 error = zfs_fill_zplprops_root(version, rootprops,
1564 goto pool_props_bad;
1566 if (nvlist_lookup_string(props,
1567 zpool_prop_to_name(ZPOOL_PROP_TNAME), &tname) == 0)
1571 error = spa_create(zc->zc_name, config, props, zplprops, dcp);
1574 * Set the remaining root properties
1576 if (!error && (error = zfs_set_prop_nvlist(spa_name,
1577 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1578 (void) spa_destroy(spa_name);
1581 nvlist_free(rootprops);
1582 nvlist_free(zplprops);
1583 nvlist_free(config);
1585 dsl_crypto_params_free(dcp, !!error);
1591 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1594 zfs_log_history(zc);
1595 error = spa_destroy(zc->zc_name);
1601 zfs_ioc_pool_import(zfs_cmd_t *zc)
1603 nvlist_t *config, *props = NULL;
1607 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1608 zc->zc_iflags, &config)) != 0)
1611 if (zc->zc_nvlist_src_size != 0 && (error =
1612 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1613 zc->zc_iflags, &props))) {
1614 nvlist_free(config);
1618 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1619 guid != zc->zc_guid)
1620 error = SET_ERROR(EINVAL);
1622 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1624 if (zc->zc_nvlist_dst != 0) {
1627 if ((err = put_nvlist(zc, config)) != 0)
1631 nvlist_free(config);
1638 zfs_ioc_pool_export(zfs_cmd_t *zc)
1641 boolean_t force = (boolean_t)zc->zc_cookie;
1642 boolean_t hardforce = (boolean_t)zc->zc_guid;
1644 zfs_log_history(zc);
1645 error = spa_export(zc->zc_name, NULL, force, hardforce);
1651 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1656 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1657 return (SET_ERROR(EEXIST));
1659 error = put_nvlist(zc, configs);
1661 nvlist_free(configs);
1668 * zc_name name of the pool
1671 * zc_cookie real errno
1672 * zc_nvlist_dst config nvlist
1673 * zc_nvlist_dst_size size of config nvlist
1676 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1682 error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1683 sizeof (zc->zc_value));
1685 if (config != NULL) {
1686 ret = put_nvlist(zc, config);
1687 nvlist_free(config);
1690 * The config may be present even if 'error' is non-zero.
1691 * In this case we return success, and preserve the real errno
1694 zc->zc_cookie = error;
1703 * Try to import the given pool, returning pool stats as appropriate so that
1704 * user land knows which devices are available and overall pool health.
1707 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1709 nvlist_t *tryconfig, *config = NULL;
1712 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1713 zc->zc_iflags, &tryconfig)) != 0)
1716 config = spa_tryimport(tryconfig);
1718 nvlist_free(tryconfig);
1721 return (SET_ERROR(EINVAL));
1723 error = put_nvlist(zc, config);
1724 nvlist_free(config);
1731 * zc_name name of the pool
1732 * zc_cookie scan func (pool_scan_func_t)
1733 * zc_flags scrub pause/resume flag (pool_scrub_cmd_t)
1736 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1741 if (zc->zc_flags >= POOL_SCRUB_FLAGS_END)
1742 return (SET_ERROR(EINVAL));
1744 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1747 if (zc->zc_flags == POOL_SCRUB_PAUSE)
1748 error = spa_scrub_pause_resume(spa, POOL_SCRUB_PAUSE);
1749 else if (zc->zc_cookie == POOL_SCAN_NONE)
1750 error = spa_scan_stop(spa);
1752 error = spa_scan(spa, zc->zc_cookie);
1754 spa_close(spa, FTAG);
1760 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1765 error = spa_open(zc->zc_name, &spa, FTAG);
1768 spa_close(spa, FTAG);
1774 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1779 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1782 if (zc->zc_cookie < spa_version(spa) ||
1783 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1784 spa_close(spa, FTAG);
1785 return (SET_ERROR(EINVAL));
1788 spa_upgrade(spa, zc->zc_cookie);
1789 spa_close(spa, FTAG);
1795 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1802 if ((size = zc->zc_history_len) == 0)
1803 return (SET_ERROR(EINVAL));
1805 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1808 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1809 spa_close(spa, FTAG);
1810 return (SET_ERROR(ENOTSUP));
1813 hist_buf = vmem_alloc(size, KM_SLEEP);
1814 if ((error = spa_history_get(spa, &zc->zc_history_offset,
1815 &zc->zc_history_len, hist_buf)) == 0) {
1816 error = ddi_copyout(hist_buf,
1817 (void *)(uintptr_t)zc->zc_history,
1818 zc->zc_history_len, zc->zc_iflags);
1821 spa_close(spa, FTAG);
1822 vmem_free(hist_buf, size);
1827 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1832 error = spa_open(zc->zc_name, &spa, FTAG);
1834 error = spa_change_guid(spa);
1835 spa_close(spa, FTAG);
1841 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1843 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1848 * zc_name name of filesystem
1849 * zc_obj object to find
1852 * zc_value name of object
1855 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1860 /* XXX reading from objset not owned */
1861 if ((error = dmu_objset_hold_flags(zc->zc_name, B_TRUE,
1864 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1865 dmu_objset_rele_flags(os, B_TRUE, FTAG);
1866 return (SET_ERROR(EINVAL));
1868 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1869 sizeof (zc->zc_value));
1870 dmu_objset_rele_flags(os, B_TRUE, FTAG);
1877 * zc_name name of filesystem
1878 * zc_obj object to find
1881 * zc_stat stats on object
1882 * zc_value path to object
1885 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1890 /* XXX reading from objset not owned */
1891 if ((error = dmu_objset_hold_flags(zc->zc_name, B_TRUE,
1894 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1895 dmu_objset_rele_flags(os, B_TRUE, FTAG);
1896 return (SET_ERROR(EINVAL));
1898 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1899 sizeof (zc->zc_value));
1900 dmu_objset_rele_flags(os, B_TRUE, FTAG);
1906 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1912 error = spa_open(zc->zc_name, &spa, FTAG);
1916 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1917 zc->zc_iflags, &config);
1919 error = spa_vdev_add(spa, config);
1920 nvlist_free(config);
1922 spa_close(spa, FTAG);
1928 * zc_name name of the pool
1929 * zc_guid guid of vdev to remove
1930 * zc_cookie cancel removal
1933 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1938 error = spa_open(zc->zc_name, &spa, FTAG);
1941 if (zc->zc_cookie != 0) {
1942 error = spa_vdev_remove_cancel(spa);
1944 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1946 spa_close(spa, FTAG);
1951 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1955 vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1957 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1959 switch (zc->zc_cookie) {
1960 case VDEV_STATE_ONLINE:
1961 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1964 case VDEV_STATE_OFFLINE:
1965 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1968 case VDEV_STATE_FAULTED:
1969 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1970 zc->zc_obj != VDEV_AUX_EXTERNAL &&
1971 zc->zc_obj != VDEV_AUX_EXTERNAL_PERSIST)
1972 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1974 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1977 case VDEV_STATE_DEGRADED:
1978 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1979 zc->zc_obj != VDEV_AUX_EXTERNAL)
1980 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1982 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1986 error = SET_ERROR(EINVAL);
1988 zc->zc_cookie = newstate;
1989 spa_close(spa, FTAG);
1994 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1997 int replacing = zc->zc_cookie;
2001 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2004 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
2005 zc->zc_iflags, &config)) == 0) {
2006 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
2007 nvlist_free(config);
2010 spa_close(spa, FTAG);
2015 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
2020 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2023 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
2025 spa_close(spa, FTAG);
2030 zfs_ioc_vdev_split(zfs_cmd_t *zc)
2033 nvlist_t *config, *props = NULL;
2035 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
2037 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2040 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
2041 zc->zc_iflags, &config))) {
2042 spa_close(spa, FTAG);
2046 if (zc->zc_nvlist_src_size != 0 && (error =
2047 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2048 zc->zc_iflags, &props))) {
2049 spa_close(spa, FTAG);
2050 nvlist_free(config);
2054 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
2056 spa_close(spa, FTAG);
2058 nvlist_free(config);
2065 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
2068 char *path = zc->zc_value;
2069 uint64_t guid = zc->zc_guid;
2072 error = spa_open(zc->zc_name, &spa, FTAG);
2076 error = spa_vdev_setpath(spa, guid, path);
2077 spa_close(spa, FTAG);
2082 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
2085 char *fru = zc->zc_value;
2086 uint64_t guid = zc->zc_guid;
2089 error = spa_open(zc->zc_name, &spa, FTAG);
2093 error = spa_vdev_setfru(spa, guid, fru);
2094 spa_close(spa, FTAG);
2099 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
2104 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2106 if (zc->zc_nvlist_dst != 0 &&
2107 (error = dsl_prop_get_all(os, &nv)) == 0) {
2108 dmu_objset_stats(os, nv);
2110 * NB: zvol_get_stats() will read the objset contents,
2111 * which we aren't supposed to do with a
2112 * DS_MODE_USER hold, because it could be
2113 * inconsistent. So this is a bit of a workaround...
2114 * XXX reading with out owning
2116 if (!zc->zc_objset_stats.dds_inconsistent &&
2117 dmu_objset_type(os) == DMU_OST_ZVOL) {
2118 error = zvol_get_stats(os, nv);
2126 error = put_nvlist(zc, nv);
2135 * zc_name name of filesystem
2136 * zc_nvlist_dst_size size of buffer for property nvlist
2139 * zc_objset_stats stats
2140 * zc_nvlist_dst property nvlist
2141 * zc_nvlist_dst_size size of property nvlist
2144 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2149 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2151 error = zfs_ioc_objset_stats_impl(zc, os);
2152 dmu_objset_rele(os, FTAG);
2160 * zc_name name of filesystem
2161 * zc_nvlist_dst_size size of buffer for property nvlist
2164 * zc_nvlist_dst received property nvlist
2165 * zc_nvlist_dst_size size of received property nvlist
2167 * Gets received properties (distinct from local properties on or after
2168 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2169 * local property values.
2172 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2178 * Without this check, we would return local property values if the
2179 * caller has not already received properties on or after
2180 * SPA_VERSION_RECVD_PROPS.
2182 if (!dsl_prop_get_hasrecvd(zc->zc_name))
2183 return (SET_ERROR(ENOTSUP));
2185 if (zc->zc_nvlist_dst != 0 &&
2186 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2187 error = put_nvlist(zc, nv);
2195 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2201 * zfs_get_zplprop() will either find a value or give us
2202 * the default value (if there is one).
2204 if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2206 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2212 * zc_name name of filesystem
2213 * zc_nvlist_dst_size size of buffer for zpl property nvlist
2216 * zc_nvlist_dst zpl property nvlist
2217 * zc_nvlist_dst_size size of zpl property nvlist
2220 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2225 /* XXX reading without owning */
2226 if ((err = dmu_objset_hold(zc->zc_name, FTAG, &os)))
2229 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2232 * NB: nvl_add_zplprop() will read the objset contents,
2233 * which we aren't supposed to do with a DS_MODE_USER
2234 * hold, because it could be inconsistent.
2236 if (zc->zc_nvlist_dst != 0 &&
2237 !zc->zc_objset_stats.dds_inconsistent &&
2238 dmu_objset_type(os) == DMU_OST_ZFS) {
2241 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2242 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2243 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2244 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2245 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2246 err = put_nvlist(zc, nv);
2249 err = SET_ERROR(ENOENT);
2251 dmu_objset_rele(os, FTAG);
2256 dataset_name_hidden(const char *name)
2259 * Skip over datasets that are not visible in this zone,
2260 * internal datasets (which have a $ in their name), and
2261 * temporary datasets (which have a % in their name).
2263 if (strchr(name, '$') != NULL)
2265 if (strchr(name, '%') != NULL)
2267 if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
2274 * zc_name name of filesystem
2275 * zc_cookie zap cursor
2276 * zc_nvlist_dst_size size of buffer for property nvlist
2279 * zc_name name of next filesystem
2280 * zc_cookie zap cursor
2281 * zc_objset_stats stats
2282 * zc_nvlist_dst property nvlist
2283 * zc_nvlist_dst_size size of property nvlist
2286 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2291 size_t orig_len = strlen(zc->zc_name);
2294 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os))) {
2295 if (error == ENOENT)
2296 error = SET_ERROR(ESRCH);
2300 p = strrchr(zc->zc_name, '/');
2301 if (p == NULL || p[1] != '\0')
2302 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2303 p = zc->zc_name + strlen(zc->zc_name);
2306 error = dmu_dir_list_next(os,
2307 sizeof (zc->zc_name) - (p - zc->zc_name), p,
2308 NULL, &zc->zc_cookie);
2309 if (error == ENOENT)
2310 error = SET_ERROR(ESRCH);
2311 } while (error == 0 && dataset_name_hidden(zc->zc_name));
2312 dmu_objset_rele(os, FTAG);
2315 * If it's an internal dataset (ie. with a '$' in its name),
2316 * don't try to get stats for it, otherwise we'll return ENOENT.
2318 if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2319 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2320 if (error == ENOENT) {
2321 /* We lost a race with destroy, get the next one. */
2322 zc->zc_name[orig_len] = '\0';
2331 * zc_name name of filesystem
2332 * zc_cookie zap cursor
2333 * zc_nvlist_dst_size size of buffer for property nvlist
2336 * zc_name name of next snapshot
2337 * zc_objset_stats stats
2338 * zc_nvlist_dst property nvlist
2339 * zc_nvlist_dst_size size of property nvlist
2342 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2347 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2349 return (error == ENOENT ? ESRCH : error);
2353 * A dataset name of maximum length cannot have any snapshots,
2354 * so exit immediately.
2356 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >=
2357 ZFS_MAX_DATASET_NAME_LEN) {
2358 dmu_objset_rele(os, FTAG);
2359 return (SET_ERROR(ESRCH));
2362 error = dmu_snapshot_list_next(os,
2363 sizeof (zc->zc_name) - strlen(zc->zc_name),
2364 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2367 if (error == 0 && !zc->zc_simple) {
2369 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2371 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2375 error = dmu_objset_from_ds(ds, &ossnap);
2377 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2378 dsl_dataset_rele(ds, FTAG);
2380 } else if (error == ENOENT) {
2381 error = SET_ERROR(ESRCH);
2384 dmu_objset_rele(os, FTAG);
2385 /* if we failed, undo the @ that we tacked on to zc_name */
2387 *strchr(zc->zc_name, '@') = '\0';
2392 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2394 const char *propname = nvpair_name(pair);
2396 unsigned int vallen;
2399 zfs_userquota_prop_t type;
2405 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2407 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2408 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2410 return (SET_ERROR(EINVAL));
2414 * A correctly constructed propname is encoded as
2415 * userquota@<rid>-<domain>.
2417 if ((dash = strchr(propname, '-')) == NULL ||
2418 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2420 return (SET_ERROR(EINVAL));
2427 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2429 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2430 zfsvfs_rele(zfsvfs, FTAG);
2437 * If the named property is one that has a special function to set its value,
2438 * return 0 on success and a positive error code on failure; otherwise if it is
2439 * not one of the special properties handled by this function, return -1.
2441 * XXX: It would be better for callers of the property interface if we handled
2442 * these special cases in dsl_prop.c (in the dsl layer).
2445 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2448 const char *propname = nvpair_name(pair);
2449 zfs_prop_t prop = zfs_name_to_prop(propname);
2450 uint64_t intval = 0;
2451 char *strval = NULL;
2454 if (prop == ZPROP_INVAL) {
2455 if (zfs_prop_userquota(propname))
2456 return (zfs_prop_set_userquota(dsname, pair));
2460 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2462 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2463 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2467 /* all special properties are numeric except for keylocation */
2468 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING) {
2469 strval = fnvpair_value_string(pair);
2471 intval = fnvpair_value_uint64(pair);
2475 case ZFS_PROP_QUOTA:
2476 err = dsl_dir_set_quota(dsname, source, intval);
2478 case ZFS_PROP_REFQUOTA:
2479 err = dsl_dataset_set_refquota(dsname, source, intval);
2481 case ZFS_PROP_FILESYSTEM_LIMIT:
2482 case ZFS_PROP_SNAPSHOT_LIMIT:
2483 if (intval == UINT64_MAX) {
2484 /* clearing the limit, just do it */
2487 err = dsl_dir_activate_fs_ss_limit(dsname);
2490 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2491 * default path to set the value in the nvlist.
2496 case ZFS_PROP_KEYLOCATION:
2497 err = dsl_crypto_can_set_keylocation(dsname, strval);
2500 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2501 * default path to set the value in the nvlist.
2506 case ZFS_PROP_RESERVATION:
2507 err = dsl_dir_set_reservation(dsname, source, intval);
2509 case ZFS_PROP_REFRESERVATION:
2510 err = dsl_dataset_set_refreservation(dsname, source, intval);
2512 case ZFS_PROP_VOLSIZE:
2513 err = zvol_set_volsize(dsname, intval);
2515 case ZFS_PROP_SNAPDEV:
2516 err = zvol_set_snapdev(dsname, source, intval);
2518 case ZFS_PROP_VOLMODE:
2519 err = zvol_set_volmode(dsname, source, intval);
2521 case ZFS_PROP_VERSION:
2525 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2528 err = zfs_set_version(zfsvfs, intval);
2529 zfsvfs_rele(zfsvfs, FTAG);
2531 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2534 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2535 (void) strcpy(zc->zc_name, dsname);
2536 (void) zfs_ioc_userspace_upgrade(zc);
2537 (void) zfs_ioc_id_quota_upgrade(zc);
2538 kmem_free(zc, sizeof (zfs_cmd_t));
2550 * This function is best effort. If it fails to set any of the given properties,
2551 * it continues to set as many as it can and returns the last error
2552 * encountered. If the caller provides a non-NULL errlist, it will be filled in
2553 * with the list of names of all the properties that failed along with the
2554 * corresponding error numbers.
2556 * If every property is set successfully, zero is returned and errlist is not
2560 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2569 nvlist_t *genericnvl = fnvlist_alloc();
2570 nvlist_t *retrynvl = fnvlist_alloc();
2573 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2574 const char *propname = nvpair_name(pair);
2575 zfs_prop_t prop = zfs_name_to_prop(propname);
2578 /* decode the property value */
2580 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2582 attrs = fnvpair_value_nvlist(pair);
2583 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2585 err = SET_ERROR(EINVAL);
2588 /* Validate value type */
2589 if (err == 0 && source == ZPROP_SRC_INHERITED) {
2590 /* inherited properties are expected to be booleans */
2591 if (nvpair_type(propval) != DATA_TYPE_BOOLEAN)
2592 err = SET_ERROR(EINVAL);
2593 } else if (err == 0 && prop == ZPROP_INVAL) {
2594 if (zfs_prop_user(propname)) {
2595 if (nvpair_type(propval) != DATA_TYPE_STRING)
2596 err = SET_ERROR(EINVAL);
2597 } else if (zfs_prop_userquota(propname)) {
2598 if (nvpair_type(propval) !=
2599 DATA_TYPE_UINT64_ARRAY)
2600 err = SET_ERROR(EINVAL);
2602 err = SET_ERROR(EINVAL);
2604 } else if (err == 0) {
2605 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2606 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2607 err = SET_ERROR(EINVAL);
2608 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2611 intval = fnvpair_value_uint64(propval);
2613 switch (zfs_prop_get_type(prop)) {
2614 case PROP_TYPE_NUMBER:
2616 case PROP_TYPE_STRING:
2617 err = SET_ERROR(EINVAL);
2619 case PROP_TYPE_INDEX:
2620 if (zfs_prop_index_to_string(prop,
2621 intval, &unused) != 0)
2622 err = SET_ERROR(EINVAL);
2626 "unknown property type");
2629 err = SET_ERROR(EINVAL);
2633 /* Validate permissions */
2635 err = zfs_check_settable(dsname, pair, CRED());
2638 if (source == ZPROP_SRC_INHERITED)
2639 err = -1; /* does not need special handling */
2641 err = zfs_prop_set_special(dsname, source,
2645 * For better performance we build up a list of
2646 * properties to set in a single transaction.
2648 err = nvlist_add_nvpair(genericnvl, pair);
2649 } else if (err != 0 && nvl != retrynvl) {
2651 * This may be a spurious error caused by
2652 * receiving quota and reservation out of order.
2653 * Try again in a second pass.
2655 err = nvlist_add_nvpair(retrynvl, pair);
2660 if (errlist != NULL)
2661 fnvlist_add_int32(errlist, propname, err);
2666 if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2671 if (!nvlist_empty(genericnvl) &&
2672 dsl_props_set(dsname, source, genericnvl) != 0) {
2674 * If this fails, we still want to set as many properties as we
2675 * can, so try setting them individually.
2678 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2679 const char *propname = nvpair_name(pair);
2683 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2685 attrs = fnvpair_value_nvlist(pair);
2686 propval = fnvlist_lookup_nvpair(attrs,
2690 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2691 strval = fnvpair_value_string(propval);
2692 err = dsl_prop_set_string(dsname, propname,
2694 } else if (nvpair_type(propval) == DATA_TYPE_BOOLEAN) {
2695 err = dsl_prop_inherit(dsname, propname,
2698 intval = fnvpair_value_uint64(propval);
2699 err = dsl_prop_set_int(dsname, propname, source,
2704 if (errlist != NULL) {
2705 fnvlist_add_int32(errlist, propname,
2712 nvlist_free(genericnvl);
2713 nvlist_free(retrynvl);
2719 * Check that all the properties are valid user properties.
2722 zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2724 nvpair_t *pair = NULL;
2727 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2728 const char *propname = nvpair_name(pair);
2730 if (!zfs_prop_user(propname) ||
2731 nvpair_type(pair) != DATA_TYPE_STRING)
2732 return (SET_ERROR(EINVAL));
2734 if ((error = zfs_secpolicy_write_perms(fsname,
2735 ZFS_DELEG_PERM_USERPROP, CRED())))
2738 if (strlen(propname) >= ZAP_MAXNAMELEN)
2739 return (SET_ERROR(ENAMETOOLONG));
2741 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN)
2742 return (SET_ERROR(E2BIG));
2748 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2752 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2755 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2756 if (nvlist_exists(skipped, nvpair_name(pair)))
2759 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2764 clear_received_props(const char *dsname, nvlist_t *props,
2768 nvlist_t *cleared_props = NULL;
2769 props_skip(props, skipped, &cleared_props);
2770 if (!nvlist_empty(cleared_props)) {
2772 * Acts on local properties until the dataset has received
2773 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2775 zprop_source_t flags = (ZPROP_SRC_NONE |
2776 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2777 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2779 nvlist_free(cleared_props);
2785 * zc_name name of filesystem
2786 * zc_value name of property to set
2787 * zc_nvlist_src{_size} nvlist of properties to apply
2788 * zc_cookie received properties flag
2791 * zc_nvlist_dst{_size} error for each unapplied received property
2794 zfs_ioc_set_prop(zfs_cmd_t *zc)
2797 boolean_t received = zc->zc_cookie;
2798 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2803 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2804 zc->zc_iflags, &nvl)) != 0)
2808 nvlist_t *origprops;
2810 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2811 (void) clear_received_props(zc->zc_name,
2813 nvlist_free(origprops);
2816 error = dsl_prop_set_hasrecvd(zc->zc_name);
2819 errors = fnvlist_alloc();
2821 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2823 if (zc->zc_nvlist_dst != 0 && errors != NULL) {
2824 (void) put_nvlist(zc, errors);
2827 nvlist_free(errors);
2834 * zc_name name of filesystem
2835 * zc_value name of property to inherit
2836 * zc_cookie revert to received value if TRUE
2841 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2843 const char *propname = zc->zc_value;
2844 zfs_prop_t prop = zfs_name_to_prop(propname);
2845 boolean_t received = zc->zc_cookie;
2846 zprop_source_t source = (received
2847 ? ZPROP_SRC_NONE /* revert to received value, if any */
2848 : ZPROP_SRC_INHERITED); /* explicitly inherit */
2856 * Only check this in the non-received case. We want to allow
2857 * 'inherit -S' to revert non-inheritable properties like quota
2858 * and reservation to the received or default values even though
2859 * they are not considered inheritable.
2861 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2862 return (SET_ERROR(EINVAL));
2865 if (prop == ZPROP_INVAL) {
2866 if (!zfs_prop_user(propname))
2867 return (SET_ERROR(EINVAL));
2869 type = PROP_TYPE_STRING;
2870 } else if (prop == ZFS_PROP_VOLSIZE || prop == ZFS_PROP_VERSION) {
2871 return (SET_ERROR(EINVAL));
2873 type = zfs_prop_get_type(prop);
2877 * zfs_prop_set_special() expects properties in the form of an
2878 * nvpair with type info.
2880 dummy = fnvlist_alloc();
2883 case PROP_TYPE_STRING:
2884 VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2886 case PROP_TYPE_NUMBER:
2887 case PROP_TYPE_INDEX:
2888 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2891 err = SET_ERROR(EINVAL);
2895 pair = nvlist_next_nvpair(dummy, NULL);
2897 err = SET_ERROR(EINVAL);
2899 err = zfs_prop_set_special(zc->zc_name, source, pair);
2900 if (err == -1) /* property is not "special", needs handling */
2901 err = dsl_prop_inherit(zc->zc_name, zc->zc_value,
2911 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2918 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2919 zc->zc_iflags, &props)))
2923 * If the only property is the configfile, then just do a spa_lookup()
2924 * to handle the faulted case.
2926 pair = nvlist_next_nvpair(props, NULL);
2927 if (pair != NULL && strcmp(nvpair_name(pair),
2928 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2929 nvlist_next_nvpair(props, pair) == NULL) {
2930 mutex_enter(&spa_namespace_lock);
2931 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2932 spa_configfile_set(spa, props, B_FALSE);
2933 spa_write_cachefile(spa, B_FALSE, B_TRUE);
2935 mutex_exit(&spa_namespace_lock);
2942 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2947 error = spa_prop_set(spa, props);
2950 spa_close(spa, FTAG);
2956 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2960 nvlist_t *nvp = NULL;
2962 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2964 * If the pool is faulted, there may be properties we can still
2965 * get (such as altroot and cachefile), so attempt to get them
2968 mutex_enter(&spa_namespace_lock);
2969 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2970 error = spa_prop_get(spa, &nvp);
2971 mutex_exit(&spa_namespace_lock);
2973 error = spa_prop_get(spa, &nvp);
2974 spa_close(spa, FTAG);
2977 if (error == 0 && zc->zc_nvlist_dst != 0)
2978 error = put_nvlist(zc, nvp);
2980 error = SET_ERROR(EFAULT);
2988 * zc_name name of filesystem
2989 * zc_nvlist_src{_size} nvlist of delegated permissions
2990 * zc_perm_action allow/unallow flag
2995 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2998 nvlist_t *fsaclnv = NULL;
3000 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
3001 zc->zc_iflags, &fsaclnv)) != 0)
3005 * Verify nvlist is constructed correctly
3007 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
3008 nvlist_free(fsaclnv);
3009 return (SET_ERROR(EINVAL));
3013 * If we don't have PRIV_SYS_MOUNT, then validate
3014 * that user is allowed to hand out each permission in
3018 error = secpolicy_zfs(CRED());
3020 if (zc->zc_perm_action == B_FALSE) {
3021 error = dsl_deleg_can_allow(zc->zc_name,
3024 error = dsl_deleg_can_unallow(zc->zc_name,
3030 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
3032 nvlist_free(fsaclnv);
3038 * zc_name name of filesystem
3041 * zc_nvlist_src{_size} nvlist of delegated permissions
3044 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
3049 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
3050 error = put_nvlist(zc, nvp);
3059 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
3061 zfs_creat_t *zct = arg;
3063 zfs_create_fs(os, cr, zct->zct_zplprops, tx);
3066 #define ZFS_PROP_UNDEFINED ((uint64_t)-1)
3070 * os parent objset pointer (NULL if root fs)
3071 * fuids_ok fuids allowed in this version of the spa?
3072 * sa_ok SAs allowed in this version of the spa?
3073 * createprops list of properties requested by creator
3076 * zplprops values for the zplprops we attach to the master node object
3077 * is_ci true if requested file system will be purely case-insensitive
3079 * Determine the settings for utf8only, normalization and
3080 * casesensitivity. Specific values may have been requested by the
3081 * creator and/or we can inherit values from the parent dataset. If
3082 * the file system is of too early a vintage, a creator can not
3083 * request settings for these properties, even if the requested
3084 * setting is the default value. We don't actually want to create dsl
3085 * properties for these, so remove them from the source nvlist after
3089 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
3090 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
3091 nvlist_t *zplprops, boolean_t *is_ci)
3093 uint64_t sense = ZFS_PROP_UNDEFINED;
3094 uint64_t norm = ZFS_PROP_UNDEFINED;
3095 uint64_t u8 = ZFS_PROP_UNDEFINED;
3098 ASSERT(zplprops != NULL);
3100 if (os != NULL && os->os_phys->os_type != DMU_OST_ZFS)
3101 return (SET_ERROR(EINVAL));
3104 * Pull out creator prop choices, if any.
3107 (void) nvlist_lookup_uint64(createprops,
3108 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
3109 (void) nvlist_lookup_uint64(createprops,
3110 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
3111 (void) nvlist_remove_all(createprops,
3112 zfs_prop_to_name(ZFS_PROP_NORMALIZE));
3113 (void) nvlist_lookup_uint64(createprops,
3114 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
3115 (void) nvlist_remove_all(createprops,
3116 zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
3117 (void) nvlist_lookup_uint64(createprops,
3118 zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3119 (void) nvlist_remove_all(createprops,
3120 zfs_prop_to_name(ZFS_PROP_CASE));
3124 * If the zpl version requested is whacky or the file system
3125 * or pool is version is too "young" to support normalization
3126 * and the creator tried to set a value for one of the props,
3129 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3130 (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3131 (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3132 (zplver < ZPL_VERSION_NORMALIZATION &&
3133 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3134 sense != ZFS_PROP_UNDEFINED)))
3135 return (SET_ERROR(ENOTSUP));
3138 * Put the version in the zplprops
3140 VERIFY(nvlist_add_uint64(zplprops,
3141 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3143 if (norm == ZFS_PROP_UNDEFINED &&
3144 (error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm)) != 0)
3146 VERIFY(nvlist_add_uint64(zplprops,
3147 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3150 * If we're normalizing, names must always be valid UTF-8 strings.
3154 if (u8 == ZFS_PROP_UNDEFINED &&
3155 (error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8)) != 0)
3157 VERIFY(nvlist_add_uint64(zplprops,
3158 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3160 if (sense == ZFS_PROP_UNDEFINED &&
3161 (error = zfs_get_zplprop(os, ZFS_PROP_CASE, &sense)) != 0)
3163 VERIFY(nvlist_add_uint64(zplprops,
3164 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3167 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
3173 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3174 nvlist_t *zplprops, boolean_t *is_ci)
3176 boolean_t fuids_ok, sa_ok;
3177 uint64_t zplver = ZPL_VERSION;
3178 objset_t *os = NULL;
3179 char parentname[ZFS_MAX_DATASET_NAME_LEN];
3185 (void) strlcpy(parentname, dataset, sizeof (parentname));
3186 cp = strrchr(parentname, '/');
3190 if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3193 spa_vers = spa_version(spa);
3194 spa_close(spa, FTAG);
3196 zplver = zfs_zpl_version_map(spa_vers);
3197 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3198 sa_ok = (zplver >= ZPL_VERSION_SA);
3201 * Open parent object set so we can inherit zplprop values.
3203 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3206 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3208 dmu_objset_rele(os, FTAG);
3213 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3214 nvlist_t *zplprops, boolean_t *is_ci)
3218 uint64_t zplver = ZPL_VERSION;
3221 zplver = zfs_zpl_version_map(spa_vers);
3222 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3223 sa_ok = (zplver >= ZPL_VERSION_SA);
3225 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3226 createprops, zplprops, is_ci);
3232 * "type" -> dmu_objset_type_t (int32)
3233 * (optional) "props" -> { prop -> value }
3234 * (optional) "hidden_args" -> { "wkeydata" -> value }
3235 * raw uint8_t array of encryption wrapping key data (32 bytes)
3238 * outnvl: propname -> error code (int32)
3241 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3244 zfs_creat_t zct = { 0 };
3245 nvlist_t *nvprops = NULL;
3246 nvlist_t *hidden_args = NULL;
3247 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3249 dmu_objset_type_t type;
3250 boolean_t is_insensitive = B_FALSE;
3251 dsl_crypto_params_t *dcp = NULL;
3253 if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3254 return (SET_ERROR(EINVAL));
3256 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3257 (void) nvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS, &hidden_args);
3261 cbfunc = zfs_create_cb;
3265 cbfunc = zvol_create_cb;
3272 if (strchr(fsname, '@') ||
3273 strchr(fsname, '%'))
3274 return (SET_ERROR(EINVAL));
3276 zct.zct_props = nvprops;
3279 return (SET_ERROR(EINVAL));
3281 if (type == DMU_OST_ZVOL) {
3282 uint64_t volsize, volblocksize;
3284 if (nvprops == NULL)
3285 return (SET_ERROR(EINVAL));
3286 if (nvlist_lookup_uint64(nvprops,
3287 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3288 return (SET_ERROR(EINVAL));
3290 if ((error = nvlist_lookup_uint64(nvprops,
3291 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3292 &volblocksize)) != 0 && error != ENOENT)
3293 return (SET_ERROR(EINVAL));
3296 volblocksize = zfs_prop_default_numeric(
3297 ZFS_PROP_VOLBLOCKSIZE);
3299 if ((error = zvol_check_volblocksize(fsname,
3300 volblocksize)) != 0 ||
3301 (error = zvol_check_volsize(volsize,
3302 volblocksize)) != 0)
3304 } else if (type == DMU_OST_ZFS) {
3308 * We have to have normalization and
3309 * case-folding flags correct when we do the
3310 * file system creation, so go figure them out
3313 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3314 NV_UNIQUE_NAME, KM_SLEEP) == 0);
3315 error = zfs_fill_zplprops(fsname, nvprops,
3316 zct.zct_zplprops, &is_insensitive);
3318 nvlist_free(zct.zct_zplprops);
3323 error = dsl_crypto_params_create_nvlist(DCP_CMD_NONE, nvprops,
3326 nvlist_free(zct.zct_zplprops);
3330 error = dmu_objset_create(fsname, type,
3331 is_insensitive ? DS_FLAG_CI_DATASET : 0, dcp, cbfunc, &zct);
3333 nvlist_free(zct.zct_zplprops);
3334 dsl_crypto_params_free(dcp, !!error);
3337 * It would be nice to do this atomically.
3340 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3347 * Volumes will return EBUSY and cannot be destroyed
3348 * until all asynchronous minor handling has completed.
3349 * Wait for the spa_zvol_taskq to drain then retry.
3351 error2 = dsl_destroy_head(fsname);
3352 while ((error2 == EBUSY) && (type == DMU_OST_ZVOL)) {
3353 error2 = spa_open(fsname, &spa, FTAG);
3355 taskq_wait(spa->spa_zvol_taskq);
3356 spa_close(spa, FTAG);
3358 error2 = dsl_destroy_head(fsname);
3367 * "origin" -> name of origin snapshot
3368 * (optional) "props" -> { prop -> value }
3369 * (optional) "hidden_args" -> { "wkeydata" -> value }
3370 * raw uint8_t array of encryption wrapping key data (32 bytes)
3374 * outnvl: propname -> error code (int32)
3377 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3380 nvlist_t *nvprops = NULL;
3383 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3384 return (SET_ERROR(EINVAL));
3385 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3387 if (strchr(fsname, '@') ||
3388 strchr(fsname, '%'))
3389 return (SET_ERROR(EINVAL));
3391 if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3392 return (SET_ERROR(EINVAL));
3394 error = dmu_objset_clone(fsname, origin_name);
3397 * It would be nice to do this atomically.
3400 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3403 (void) dsl_destroy_head(fsname);
3410 zfs_ioc_remap(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3412 if (strchr(fsname, '@') ||
3413 strchr(fsname, '%'))
3414 return (SET_ERROR(EINVAL));
3416 return (dmu_objset_remap_indirects(fsname));
3421 * "snaps" -> { snapshot1, snapshot2 }
3422 * (optional) "props" -> { prop -> value (string) }
3425 * outnvl: snapshot -> error code (int32)
3428 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3431 nvlist_t *props = NULL;
3435 (void) nvlist_lookup_nvlist(innvl, "props", &props);
3436 if ((error = zfs_check_userprops(poolname, props)) != 0)
3439 if (!nvlist_empty(props) &&
3440 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3441 return (SET_ERROR(ENOTSUP));
3443 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3444 return (SET_ERROR(EINVAL));
3445 poollen = strlen(poolname);
3446 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3447 pair = nvlist_next_nvpair(snaps, pair)) {
3448 const char *name = nvpair_name(pair);
3449 const char *cp = strchr(name, '@');
3452 * The snap name must contain an @, and the part after it must
3453 * contain only valid characters.
3456 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3457 return (SET_ERROR(EINVAL));
3460 * The snap must be in the specified pool.
3462 if (strncmp(name, poolname, poollen) != 0 ||
3463 (name[poollen] != '/' && name[poollen] != '@'))
3464 return (SET_ERROR(EXDEV));
3466 /* This must be the only snap of this fs. */
3467 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3468 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3469 if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3471 return (SET_ERROR(EXDEV));
3476 error = dsl_dataset_snapshot(snaps, props, outnvl);
3482 * innvl: "message" -> string
3486 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3494 * The poolname in the ioctl is not set, we get it from the TSD,
3495 * which was set at the end of the last successful ioctl that allows
3496 * logging. The secpolicy func already checked that it is set.
3497 * Only one log ioctl is allowed after each successful ioctl, so
3498 * we clear the TSD here.
3500 poolname = tsd_get(zfs_allow_log_key);
3501 if (poolname == NULL)
3502 return (SET_ERROR(EINVAL));
3503 (void) tsd_set(zfs_allow_log_key, NULL);
3504 error = spa_open(poolname, &spa, FTAG);
3509 if (nvlist_lookup_string(innvl, "message", &message) != 0) {
3510 spa_close(spa, FTAG);
3511 return (SET_ERROR(EINVAL));
3514 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3515 spa_close(spa, FTAG);
3516 return (SET_ERROR(ENOTSUP));
3519 error = spa_history_log(spa, message);
3520 spa_close(spa, FTAG);
3525 * The dp_config_rwlock must not be held when calling this, because the
3526 * unmount may need to write out data.
3528 * This function is best-effort. Callers must deal gracefully if it
3529 * remains mounted (or is remounted after this call).
3531 * Returns 0 if the argument is not a snapshot, or it is not currently a
3532 * filesystem, or we were able to unmount it. Returns error code otherwise.
3535 zfs_unmount_snap(const char *snapname)
3537 if (strchr(snapname, '@') == NULL)
3540 (void) zfsctl_snapshot_unmount((char *)snapname, MNT_FORCE);
3545 zfs_unmount_snap_cb(const char *snapname, void *arg)
3547 zfs_unmount_snap(snapname);
3552 * When a clone is destroyed, its origin may also need to be destroyed,
3553 * in which case it must be unmounted. This routine will do that unmount
3557 zfs_destroy_unmount_origin(const char *fsname)
3563 error = dmu_objset_hold(fsname, FTAG, &os);
3566 ds = dmu_objset_ds(os);
3567 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3568 char originname[ZFS_MAX_DATASET_NAME_LEN];
3569 dsl_dataset_name(ds->ds_prev, originname);
3570 dmu_objset_rele(os, FTAG);
3571 zfs_unmount_snap(originname);
3573 dmu_objset_rele(os, FTAG);
3579 * "snaps" -> { snapshot1, snapshot2 }
3580 * (optional boolean) "defer"
3583 * outnvl: snapshot -> error code (int32)
3587 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3593 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3594 return (SET_ERROR(EINVAL));
3595 defer = nvlist_exists(innvl, "defer");
3597 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3598 pair = nvlist_next_nvpair(snaps, pair)) {
3599 zfs_unmount_snap(nvpair_name(pair));
3602 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3606 * Create bookmarks. Bookmark names are of the form <fs>#<bmark>.
3607 * All bookmarks must be in the same pool.
3610 * bookmark1 -> snapshot1, bookmark2 -> snapshot2
3613 * outnvl: bookmark -> error code (int32)
3618 zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3620 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3621 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3625 * Verify the snapshot argument.
3627 if (nvpair_value_string(pair, &snap_name) != 0)
3628 return (SET_ERROR(EINVAL));
3631 /* Verify that the keys (bookmarks) are unique */
3632 for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair);
3633 pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) {
3634 if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0)
3635 return (SET_ERROR(EINVAL));
3639 return (dsl_bookmark_create(innvl, outnvl));
3644 * property 1, property 2, ...
3648 * bookmark name 1 -> { property 1, property 2, ... },
3649 * bookmark name 2 -> { property 1, property 2, ... }
3654 zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3656 return (dsl_get_bookmarks(fsname, innvl, outnvl));
3661 * bookmark name 1, bookmark name 2
3664 * outnvl: bookmark -> error code (int32)
3668 zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl,
3673 poollen = strlen(poolname);
3674 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3675 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3676 const char *name = nvpair_name(pair);
3677 const char *cp = strchr(name, '#');
3680 * The bookmark name must contain an #, and the part after it
3681 * must contain only valid characters.
3684 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3685 return (SET_ERROR(EINVAL));
3688 * The bookmark must be in the specified pool.
3690 if (strncmp(name, poolname, poollen) != 0 ||
3691 (name[poollen] != '/' && name[poollen] != '#'))
3692 return (SET_ERROR(EXDEV));
3695 error = dsl_bookmark_destroy(innvl, outnvl);
3700 zfs_ioc_channel_program(const char *poolname, nvlist_t *innvl,
3704 uint64_t instrlimit, memlimit;
3705 boolean_t sync_flag;
3706 nvpair_t *nvarg = NULL;
3708 if (0 != nvlist_lookup_string(innvl, ZCP_ARG_PROGRAM, &program)) {
3711 if (0 != nvlist_lookup_boolean_value(innvl, ZCP_ARG_SYNC, &sync_flag)) {
3714 if (0 != nvlist_lookup_uint64(innvl, ZCP_ARG_INSTRLIMIT, &instrlimit)) {
3715 instrlimit = ZCP_DEFAULT_INSTRLIMIT;
3717 if (0 != nvlist_lookup_uint64(innvl, ZCP_ARG_MEMLIMIT, &memlimit)) {
3718 memlimit = ZCP_DEFAULT_MEMLIMIT;
3720 if (0 != nvlist_lookup_nvpair(innvl, ZCP_ARG_ARGLIST, &nvarg)) {
3724 if (instrlimit == 0 || instrlimit > zfs_lua_max_instrlimit)
3726 if (memlimit == 0 || memlimit > zfs_lua_max_memlimit)
3729 return (zcp_eval(poolname, program, sync_flag, instrlimit, memlimit,
3735 * zc_name name of dataset to destroy
3736 * zc_objset_type type of objset
3737 * zc_defer_destroy mark for deferred destroy
3742 zfs_ioc_destroy(zfs_cmd_t *zc)
3746 if (zc->zc_objset_type == DMU_OST_ZFS)
3747 zfs_unmount_snap(zc->zc_name);
3749 if (strchr(zc->zc_name, '@')) {
3750 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3752 err = dsl_destroy_head(zc->zc_name);
3753 if (err == EEXIST) {
3755 * It is possible that the given DS may have
3756 * hidden child (%recv) datasets - "leftovers"
3757 * resulting from the previously interrupted
3760 * 6 extra bytes for /%recv
3762 char namebuf[ZFS_MAX_DATASET_NAME_LEN + 6];
3764 if (snprintf(namebuf, sizeof (namebuf), "%s/%s",
3765 zc->zc_name, recv_clone_name) >=
3767 return (SET_ERROR(EINVAL));
3770 * Try to remove the hidden child (%recv) and after
3771 * that try to remove the target dataset.
3772 * If the hidden child (%recv) does not exist
3773 * the original error (EEXIST) will be returned
3775 err = dsl_destroy_head(namebuf);
3777 err = dsl_destroy_head(zc->zc_name);
3778 else if (err == ENOENT)
3779 err = SET_ERROR(EEXIST);
3787 * fsname is name of dataset to rollback (to most recent snapshot)
3789 * innvl may contain name of expected target snapshot
3791 * outnvl: "target" -> name of most recent snapshot
3796 zfs_ioc_rollback(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3800 char *target = NULL;
3803 (void) nvlist_lookup_string(innvl, "target", &target);
3804 if (target != NULL) {
3805 const char *cp = strchr(target, '@');
3808 * The snap name must contain an @, and the part after it must
3809 * contain only valid characters.
3812 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3813 return (SET_ERROR(EINVAL));
3816 if (getzfsvfs(fsname, &zfsvfs) == 0) {
3819 ds = dmu_objset_ds(zfsvfs->z_os);
3820 error = zfs_suspend_fs(zfsvfs);
3824 error = dsl_dataset_rollback(fsname, target, zfsvfs,
3826 resume_err = zfs_resume_fs(zfsvfs, ds);
3827 error = error ? error : resume_err;
3829 deactivate_super(zfsvfs->z_sb);
3830 } else if ((zv = zvol_suspend(fsname)) != NULL) {
3831 error = dsl_dataset_rollback(fsname, target, zvol_tag(zv),
3835 error = dsl_dataset_rollback(fsname, target, NULL, outnvl);
3841 recursive_unmount(const char *fsname, void *arg)
3843 const char *snapname = arg;
3846 fullname = kmem_asprintf("%s@%s", fsname, snapname);
3847 zfs_unmount_snap(fullname);
3855 * zc_name old name of dataset
3856 * zc_value new name of dataset
3857 * zc_cookie recursive flag (only valid for snapshots)
3862 zfs_ioc_rename(zfs_cmd_t *zc)
3864 boolean_t recursive = zc->zc_cookie & 1;
3867 /* "zfs rename" from and to ...%recv datasets should both fail */
3868 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
3869 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3870 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0 ||
3871 dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3872 strchr(zc->zc_name, '%') || strchr(zc->zc_value, '%'))
3873 return (SET_ERROR(EINVAL));
3875 at = strchr(zc->zc_name, '@');
3877 /* snaps must be in same fs */
3880 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
3881 return (SET_ERROR(EXDEV));
3883 if (zc->zc_objset_type == DMU_OST_ZFS) {
3884 error = dmu_objset_find(zc->zc_name,
3885 recursive_unmount, at + 1,
3886 recursive ? DS_FIND_CHILDREN : 0);
3892 error = dsl_dataset_rename_snapshot(zc->zc_name,
3893 at + 1, strchr(zc->zc_value, '@') + 1, recursive);
3898 return (dsl_dir_rename(zc->zc_name, zc->zc_value));
3903 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3905 const char *propname = nvpair_name(pair);
3906 boolean_t issnap = (strchr(dsname, '@') != NULL);
3907 zfs_prop_t prop = zfs_name_to_prop(propname);
3911 if (prop == ZPROP_INVAL) {
3912 if (zfs_prop_user(propname)) {
3913 if ((err = zfs_secpolicy_write_perms(dsname,
3914 ZFS_DELEG_PERM_USERPROP, cr)))
3919 if (!issnap && zfs_prop_userquota(propname)) {
3920 const char *perm = NULL;
3921 const char *uq_prefix =
3922 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3923 const char *gq_prefix =
3924 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3925 const char *uiq_prefix =
3926 zfs_userquota_prop_prefixes[ZFS_PROP_USEROBJQUOTA];
3927 const char *giq_prefix =
3928 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPOBJQUOTA];
3929 const char *pq_prefix =
3930 zfs_userquota_prop_prefixes[ZFS_PROP_PROJECTQUOTA];
3931 const char *piq_prefix = zfs_userquota_prop_prefixes[\
3932 ZFS_PROP_PROJECTOBJQUOTA];
3934 if (strncmp(propname, uq_prefix,
3935 strlen(uq_prefix)) == 0) {
3936 perm = ZFS_DELEG_PERM_USERQUOTA;
3937 } else if (strncmp(propname, uiq_prefix,
3938 strlen(uiq_prefix)) == 0) {
3939 perm = ZFS_DELEG_PERM_USEROBJQUOTA;
3940 } else if (strncmp(propname, gq_prefix,
3941 strlen(gq_prefix)) == 0) {
3942 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3943 } else if (strncmp(propname, giq_prefix,
3944 strlen(giq_prefix)) == 0) {
3945 perm = ZFS_DELEG_PERM_GROUPOBJQUOTA;
3946 } else if (strncmp(propname, pq_prefix,
3947 strlen(pq_prefix)) == 0) {
3948 perm = ZFS_DELEG_PERM_PROJECTQUOTA;
3949 } else if (strncmp(propname, piq_prefix,
3950 strlen(piq_prefix)) == 0) {
3951 perm = ZFS_DELEG_PERM_PROJECTOBJQUOTA;
3953 /* {USER|GROUP|PROJECT}USED are read-only */
3954 return (SET_ERROR(EINVAL));
3957 if ((err = zfs_secpolicy_write_perms(dsname, perm, cr)))
3962 return (SET_ERROR(EINVAL));
3966 return (SET_ERROR(EINVAL));
3968 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3970 * dsl_prop_get_all_impl() returns properties in this
3974 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3975 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3980 * Check that this value is valid for this pool version
3983 case ZFS_PROP_COMPRESSION:
3985 * If the user specified gzip compression, make sure
3986 * the SPA supports it. We ignore any errors here since
3987 * we'll catch them later.
3989 if (nvpair_value_uint64(pair, &intval) == 0) {
3990 if (intval >= ZIO_COMPRESS_GZIP_1 &&
3991 intval <= ZIO_COMPRESS_GZIP_9 &&
3992 zfs_earlier_version(dsname,
3993 SPA_VERSION_GZIP_COMPRESSION)) {
3994 return (SET_ERROR(ENOTSUP));
3997 if (intval == ZIO_COMPRESS_ZLE &&
3998 zfs_earlier_version(dsname,
3999 SPA_VERSION_ZLE_COMPRESSION))
4000 return (SET_ERROR(ENOTSUP));
4002 if (intval == ZIO_COMPRESS_LZ4) {
4005 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4008 if (!spa_feature_is_enabled(spa,
4009 SPA_FEATURE_LZ4_COMPRESS)) {
4010 spa_close(spa, FTAG);
4011 return (SET_ERROR(ENOTSUP));
4013 spa_close(spa, FTAG);
4017 * If this is a bootable dataset then
4018 * verify that the compression algorithm
4019 * is supported for booting. We must return
4020 * something other than ENOTSUP since it
4021 * implies a downrev pool version.
4023 if (zfs_is_bootfs(dsname) &&
4024 !BOOTFS_COMPRESS_VALID(intval)) {
4025 return (SET_ERROR(ERANGE));
4030 case ZFS_PROP_COPIES:
4031 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
4032 return (SET_ERROR(ENOTSUP));
4035 case ZFS_PROP_VOLBLOCKSIZE:
4036 case ZFS_PROP_RECORDSIZE:
4037 /* Record sizes above 128k need the feature to be enabled */
4038 if (nvpair_value_uint64(pair, &intval) == 0 &&
4039 intval > SPA_OLD_MAXBLOCKSIZE) {
4043 * We don't allow setting the property above 1MB,
4044 * unless the tunable has been changed.
4046 if (intval > zfs_max_recordsize ||
4047 intval > SPA_MAXBLOCKSIZE)
4048 return (SET_ERROR(ERANGE));
4050 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4053 if (!spa_feature_is_enabled(spa,
4054 SPA_FEATURE_LARGE_BLOCKS)) {
4055 spa_close(spa, FTAG);
4056 return (SET_ERROR(ENOTSUP));
4058 spa_close(spa, FTAG);
4062 case ZFS_PROP_DNODESIZE:
4063 /* Dnode sizes above 512 need the feature to be enabled */
4064 if (nvpair_value_uint64(pair, &intval) == 0 &&
4065 intval != ZFS_DNSIZE_LEGACY) {
4069 * If this is a bootable dataset then
4070 * we don't allow large (>512B) dnodes,
4071 * because GRUB doesn't support them.
4073 if (zfs_is_bootfs(dsname) &&
4074 intval != ZFS_DNSIZE_LEGACY) {
4075 return (SET_ERROR(EDOM));
4078 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4081 if (!spa_feature_is_enabled(spa,
4082 SPA_FEATURE_LARGE_DNODE)) {
4083 spa_close(spa, FTAG);
4084 return (SET_ERROR(ENOTSUP));
4086 spa_close(spa, FTAG);
4090 case ZFS_PROP_SHARESMB:
4091 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
4092 return (SET_ERROR(ENOTSUP));
4095 case ZFS_PROP_ACLINHERIT:
4096 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
4097 nvpair_value_uint64(pair, &intval) == 0) {
4098 if (intval == ZFS_ACL_PASSTHROUGH_X &&
4099 zfs_earlier_version(dsname,
4100 SPA_VERSION_PASSTHROUGH_X))
4101 return (SET_ERROR(ENOTSUP));
4104 case ZFS_PROP_CHECKSUM:
4105 case ZFS_PROP_DEDUP:
4107 spa_feature_t feature;
4112 /* dedup feature version checks */
4113 if (prop == ZFS_PROP_DEDUP &&
4114 zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
4115 return (SET_ERROR(ENOTSUP));
4117 if (nvpair_value_uint64(pair, &intval) != 0)
4118 return (SET_ERROR(EINVAL));
4120 /* check prop value is enabled in features */
4121 feature = zio_checksum_to_feature(intval & ZIO_CHECKSUM_MASK);
4122 if (feature == SPA_FEATURE_NONE)
4125 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4128 * Salted checksums are not supported on root pools.
4130 if (spa_bootfs(spa) != 0 &&
4131 intval < ZIO_CHECKSUM_FUNCTIONS &&
4132 (zio_checksum_table[intval].ci_flags &
4133 ZCHECKSUM_FLAG_SALTED)) {
4134 spa_close(spa, FTAG);
4135 return (SET_ERROR(ERANGE));
4137 if (!spa_feature_is_enabled(spa, feature)) {
4138 spa_close(spa, FTAG);
4139 return (SET_ERROR(ENOTSUP));
4141 spa_close(spa, FTAG);
4149 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
4153 * Removes properties from the given props list that fail permission checks
4154 * needed to clear them and to restore them in case of a receive error. For each
4155 * property, make sure we have both set and inherit permissions.
4157 * Returns the first error encountered if any permission checks fail. If the
4158 * caller provides a non-NULL errlist, it also gives the complete list of names
4159 * of all the properties that failed a permission check along with the
4160 * corresponding error numbers. The caller is responsible for freeing the
4163 * If every property checks out successfully, zero is returned and the list
4164 * pointed at by errlist is NULL.
4167 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
4170 nvpair_t *pair, *next_pair;
4177 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4179 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
4180 (void) strlcpy(zc->zc_name, dataset, sizeof (zc->zc_name));
4181 pair = nvlist_next_nvpair(props, NULL);
4182 while (pair != NULL) {
4183 next_pair = nvlist_next_nvpair(props, pair);
4185 (void) strlcpy(zc->zc_value, nvpair_name(pair),
4186 sizeof (zc->zc_value));
4187 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
4188 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
4189 VERIFY(nvlist_remove_nvpair(props, pair) == 0);
4190 VERIFY(nvlist_add_int32(errors,
4191 zc->zc_value, err) == 0);
4195 kmem_free(zc, sizeof (zfs_cmd_t));
4197 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
4198 nvlist_free(errors);
4201 VERIFY(nvpair_value_int32(pair, &rv) == 0);
4204 if (errlist == NULL)
4205 nvlist_free(errors);
4213 propval_equals(nvpair_t *p1, nvpair_t *p2)
4215 if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
4216 /* dsl_prop_get_all_impl() format */
4218 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
4219 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4223 if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
4225 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
4226 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4230 if (nvpair_type(p1) != nvpair_type(p2))
4233 if (nvpair_type(p1) == DATA_TYPE_STRING) {
4234 char *valstr1, *valstr2;
4236 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
4237 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
4238 return (strcmp(valstr1, valstr2) == 0);
4240 uint64_t intval1, intval2;
4242 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
4243 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
4244 return (intval1 == intval2);
4249 * Remove properties from props if they are not going to change (as determined
4250 * by comparison with origprops). Remove them from origprops as well, since we
4251 * do not need to clear or restore properties that won't change.
4254 props_reduce(nvlist_t *props, nvlist_t *origprops)
4256 nvpair_t *pair, *next_pair;
4258 if (origprops == NULL)
4259 return; /* all props need to be received */
4261 pair = nvlist_next_nvpair(props, NULL);
4262 while (pair != NULL) {
4263 const char *propname = nvpair_name(pair);
4266 next_pair = nvlist_next_nvpair(props, pair);
4268 if ((nvlist_lookup_nvpair(origprops, propname,
4269 &match) != 0) || !propval_equals(pair, match))
4270 goto next; /* need to set received value */
4272 /* don't clear the existing received value */
4273 (void) nvlist_remove_nvpair(origprops, match);
4274 /* don't bother receiving the property */
4275 (void) nvlist_remove_nvpair(props, pair);
4282 * Extract properties that cannot be set PRIOR to the receipt of a dataset.
4283 * For example, refquota cannot be set until after the receipt of a dataset,
4284 * because in replication streams, an older/earlier snapshot may exceed the
4285 * refquota. We want to receive the older/earlier snapshot, but setting
4286 * refquota pre-receipt will set the dsl's ACTUAL quota, which will prevent
4287 * the older/earlier snapshot from being received (with EDQUOT).
4289 * The ZFS test "zfs_receive_011_pos" demonstrates such a scenario.
4291 * libzfs will need to be judicious handling errors encountered by props
4292 * extracted by this function.
4295 extract_delay_props(nvlist_t *props)
4297 nvlist_t *delayprops;
4298 nvpair_t *nvp, *tmp;
4299 static const zfs_prop_t delayable[] = {
4301 ZFS_PROP_KEYLOCATION,
4306 VERIFY(nvlist_alloc(&delayprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4308 for (nvp = nvlist_next_nvpair(props, NULL); nvp != NULL;
4309 nvp = nvlist_next_nvpair(props, nvp)) {
4311 * strcmp() is safe because zfs_prop_to_name() always returns
4314 for (i = 0; delayable[i] != 0; i++) {
4315 if (strcmp(zfs_prop_to_name(delayable[i]),
4316 nvpair_name(nvp)) == 0) {
4320 if (delayable[i] != 0) {
4321 tmp = nvlist_prev_nvpair(props, nvp);
4322 VERIFY(nvlist_add_nvpair(delayprops, nvp) == 0);
4323 VERIFY(nvlist_remove_nvpair(props, nvp) == 0);
4328 if (nvlist_empty(delayprops)) {
4329 nvlist_free(delayprops);
4332 return (delayprops);
4336 static boolean_t zfs_ioc_recv_inject_err;
4340 * nvlist 'errors' is always allocated. It will contain descriptions of
4341 * encountered errors, if any. It's the callers responsibility to free.
4344 zfs_ioc_recv_impl(char *tofs, char *tosnap, char *origin, nvlist_t *recvprops,
4345 nvlist_t *localprops, boolean_t force, boolean_t resumable, int input_fd,
4346 dmu_replay_record_t *begin_record, int cleanup_fd, uint64_t *read_bytes,
4347 uint64_t *errflags, uint64_t *action_handle, nvlist_t **errors)
4349 dmu_recv_cookie_t drc;
4351 int props_error = 0;
4353 nvlist_t *delayprops = NULL; /* sent properties applied post-receive */
4354 nvlist_t *origprops = NULL; /* existing properties */
4355 nvlist_t *origrecvd = NULL; /* existing received properties */
4356 boolean_t first_recvd_props = B_FALSE;
4361 *errors = fnvlist_alloc();
4363 input_fp = getf(input_fd);
4364 if (input_fp == NULL)
4365 return (SET_ERROR(EBADF));
4367 error = dmu_recv_begin(tofs, tosnap,
4368 begin_record, force, resumable, origin, &drc);
4373 * Set properties before we receive the stream so that they are applied
4374 * to the new data. Note that we must call dmu_recv_stream() if
4375 * dmu_recv_begin() succeeds.
4377 if (recvprops != NULL && !drc.drc_newfs) {
4378 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
4379 SPA_VERSION_RECVD_PROPS &&
4380 !dsl_prop_get_hasrecvd(tofs))
4381 first_recvd_props = B_TRUE;
4384 * If new received properties are supplied, they are to
4385 * completely replace the existing received properties, so stash
4386 * away the existing ones.
4388 if (dsl_prop_get_received(tofs, &origrecvd) == 0) {
4389 nvlist_t *errlist = NULL;
4391 * Don't bother writing a property if its value won't
4392 * change (and avoid the unnecessary security checks).
4394 * The first receive after SPA_VERSION_RECVD_PROPS is a
4395 * special case where we blow away all local properties
4398 if (!first_recvd_props)
4399 props_reduce(recvprops, origrecvd);
4400 if (zfs_check_clearable(tofs, origrecvd, &errlist) != 0)
4401 (void) nvlist_merge(*errors, errlist, 0);
4402 nvlist_free(errlist);
4404 if (clear_received_props(tofs, origrecvd,
4405 first_recvd_props ? NULL : recvprops) != 0)
4406 *errflags |= ZPROP_ERR_NOCLEAR;
4408 *errflags |= ZPROP_ERR_NOCLEAR;
4413 * Stash away existing properties so we can restore them on error unless
4414 * we're doing the first receive after SPA_VERSION_RECVD_PROPS, in which
4415 * case "origrecvd" will take care of that.
4417 if (localprops != NULL && !drc.drc_newfs && !first_recvd_props) {
4419 if (dmu_objset_hold(tofs, FTAG, &os) == 0) {
4420 if (dsl_prop_get_all(os, &origprops) != 0) {
4421 *errflags |= ZPROP_ERR_NOCLEAR;
4423 dmu_objset_rele(os, FTAG);
4425 *errflags |= ZPROP_ERR_NOCLEAR;
4429 if (recvprops != NULL) {
4430 props_error = dsl_prop_set_hasrecvd(tofs);
4432 if (props_error == 0) {
4433 delayprops = extract_delay_props(recvprops);
4434 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4435 recvprops, *errors);
4439 if (localprops != NULL) {
4440 nvlist_t *oprops = fnvlist_alloc();
4441 nvlist_t *xprops = fnvlist_alloc();
4442 nvpair_t *nvp = NULL;
4444 while ((nvp = nvlist_next_nvpair(localprops, nvp)) != NULL) {
4445 if (nvpair_type(nvp) == DATA_TYPE_BOOLEAN) {
4447 const char *name = nvpair_name(nvp);
4448 zfs_prop_t prop = zfs_name_to_prop(name);
4449 if (prop != ZPROP_INVAL) {
4450 if (!zfs_prop_inheritable(prop))
4452 } else if (!zfs_prop_user(name))
4454 fnvlist_add_boolean(xprops, name);
4456 /* -o property=value */
4457 fnvlist_add_nvpair(oprops, nvp);
4460 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_LOCAL,
4462 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_INHERITED,
4465 nvlist_free(oprops);
4466 nvlist_free(xprops);
4469 off = input_fp->f_offset;
4470 error = dmu_recv_stream(&drc, input_fp->f_vnode, &off, cleanup_fd,
4474 zfsvfs_t *zfsvfs = NULL;
4475 zvol_state_t *zv = NULL;
4477 if (getzfsvfs(tofs, &zfsvfs) == 0) {
4482 ds = dmu_objset_ds(zfsvfs->z_os);
4483 error = zfs_suspend_fs(zfsvfs);
4485 * If the suspend fails, then the recv_end will
4486 * likely also fail, and clean up after itself.
4488 end_err = dmu_recv_end(&drc, zfsvfs);
4490 error = zfs_resume_fs(zfsvfs, ds);
4491 error = error ? error : end_err;
4492 deactivate_super(zfsvfs->z_sb);
4493 } else if ((zv = zvol_suspend(tofs)) != NULL) {
4494 error = dmu_recv_end(&drc, zvol_tag(zv));
4497 error = dmu_recv_end(&drc, NULL);
4500 /* Set delayed properties now, after we're done receiving. */
4501 if (delayprops != NULL && error == 0) {
4502 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4503 delayprops, *errors);
4507 if (delayprops != NULL) {
4509 * Merge delayed props back in with initial props, in case
4510 * we're DEBUG and zfs_ioc_recv_inject_err is set (which means
4511 * we have to make sure clear_received_props() includes
4512 * the delayed properties).
4514 * Since zfs_ioc_recv_inject_err is only in DEBUG kernels,
4515 * using ASSERT() will be just like a VERIFY.
4517 ASSERT(nvlist_merge(recvprops, delayprops, 0) == 0);
4518 nvlist_free(delayprops);
4522 *read_bytes = off - input_fp->f_offset;
4523 if (VOP_SEEK(input_fp->f_vnode, input_fp->f_offset, &off, NULL) == 0)
4524 input_fp->f_offset = off;
4527 if (zfs_ioc_recv_inject_err) {
4528 zfs_ioc_recv_inject_err = B_FALSE;
4534 * On error, restore the original props.
4536 if (error != 0 && recvprops != NULL && !drc.drc_newfs) {
4537 if (clear_received_props(tofs, recvprops, NULL) != 0) {
4539 * We failed to clear the received properties.
4540 * Since we may have left a $recvd value on the
4541 * system, we can't clear the $hasrecvd flag.
4543 *errflags |= ZPROP_ERR_NORESTORE;
4544 } else if (first_recvd_props) {
4545 dsl_prop_unset_hasrecvd(tofs);
4548 if (origrecvd == NULL && !drc.drc_newfs) {
4549 /* We failed to stash the original properties. */
4550 *errflags |= ZPROP_ERR_NORESTORE;
4554 * dsl_props_set() will not convert RECEIVED to LOCAL on or
4555 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4556 * explicitly if we're restoring local properties cleared in the
4557 * first new-style receive.
4559 if (origrecvd != NULL &&
4560 zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4561 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4562 origrecvd, NULL) != 0) {
4564 * We stashed the original properties but failed to
4567 *errflags |= ZPROP_ERR_NORESTORE;
4570 if (error != 0 && localprops != NULL && !drc.drc_newfs &&
4571 !first_recvd_props) {
4573 nvlist_t *inheritprops;
4576 if (origprops == NULL) {
4577 /* We failed to stash the original properties. */
4578 *errflags |= ZPROP_ERR_NORESTORE;
4582 /* Restore original props */
4583 setprops = fnvlist_alloc();
4584 inheritprops = fnvlist_alloc();
4586 while ((nvp = nvlist_next_nvpair(localprops, nvp)) != NULL) {
4587 const char *name = nvpair_name(nvp);
4591 if (!nvlist_exists(origprops, name)) {
4593 * Property was not present or was explicitly
4594 * inherited before the receive, restore this.
4596 fnvlist_add_boolean(inheritprops, name);
4599 attrs = fnvlist_lookup_nvlist(origprops, name);
4600 source = fnvlist_lookup_string(attrs, ZPROP_SOURCE);
4602 /* Skip received properties */
4603 if (strcmp(source, ZPROP_SOURCE_VAL_RECVD) == 0)
4606 if (strcmp(source, tofs) == 0) {
4607 /* Property was locally set */
4608 fnvlist_add_nvlist(setprops, name, attrs);
4610 /* Property was implicitly inherited */
4611 fnvlist_add_boolean(inheritprops, name);
4615 if (zfs_set_prop_nvlist(tofs, ZPROP_SRC_LOCAL, setprops,
4617 *errflags |= ZPROP_ERR_NORESTORE;
4618 if (zfs_set_prop_nvlist(tofs, ZPROP_SRC_INHERITED, inheritprops,
4620 *errflags |= ZPROP_ERR_NORESTORE;
4622 nvlist_free(setprops);
4623 nvlist_free(inheritprops);
4627 nvlist_free(origrecvd);
4628 nvlist_free(origprops);
4631 error = props_error;
4638 * zc_name name of containing filesystem (unused)
4639 * zc_nvlist_src{_size} nvlist of properties to apply
4640 * zc_nvlist_conf{_size} nvlist of properties to exclude
4641 * (DATA_TYPE_BOOLEAN) and override (everything else)
4642 * zc_value name of snapshot to create
4643 * zc_string name of clone origin (if DRR_FLAG_CLONE)
4644 * zc_cookie file descriptor to recv from
4645 * zc_begin_record the BEGIN record of the stream (not byteswapped)
4646 * zc_guid force flag
4647 * zc_cleanup_fd cleanup-on-exit file descriptor
4648 * zc_action_handle handle for this guid/ds mapping (or zero on first call)
4651 * zc_cookie number of bytes read
4652 * zc_obj zprop_errflags_t
4653 * zc_action_handle handle for this guid/ds mapping
4654 * zc_nvlist_dst{_size} error for each unapplied received property
4657 zfs_ioc_recv(zfs_cmd_t *zc)
4659 dmu_replay_record_t begin_record;
4660 nvlist_t *errors = NULL;
4661 nvlist_t *recvdprops = NULL;
4662 nvlist_t *localprops = NULL;
4663 char *origin = NULL;
4665 char tofs[ZFS_MAX_DATASET_NAME_LEN];
4668 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
4669 strchr(zc->zc_value, '@') == NULL ||
4670 strchr(zc->zc_value, '%'))
4671 return (SET_ERROR(EINVAL));
4673 (void) strlcpy(tofs, zc->zc_value, sizeof (tofs));
4674 tosnap = strchr(tofs, '@');
4677 if (zc->zc_nvlist_src != 0 &&
4678 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
4679 zc->zc_iflags, &recvdprops)) != 0)
4682 if (zc->zc_nvlist_conf != 0 &&
4683 (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
4684 zc->zc_iflags, &localprops)) != 0)
4687 if (zc->zc_string[0])
4688 origin = zc->zc_string;
4690 begin_record.drr_type = DRR_BEGIN;
4691 begin_record.drr_payloadlen = 0;
4692 begin_record.drr_u.drr_begin = zc->zc_begin_record;
4694 error = zfs_ioc_recv_impl(tofs, tosnap, origin, recvdprops, localprops,
4695 zc->zc_guid, B_FALSE, zc->zc_cookie, &begin_record,
4696 zc->zc_cleanup_fd, &zc->zc_cookie, &zc->zc_obj,
4697 &zc->zc_action_handle, &errors);
4698 nvlist_free(recvdprops);
4699 nvlist_free(localprops);
4702 * Now that all props, initial and delayed, are set, report the prop
4703 * errors to the caller.
4705 if (zc->zc_nvlist_dst_size != 0 && errors != NULL &&
4706 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4707 put_nvlist(zc, errors) != 0)) {
4709 * Caller made zc->zc_nvlist_dst less than the minimum expected
4710 * size or supplied an invalid address.
4712 error = SET_ERROR(EINVAL);
4715 nvlist_free(errors);
4722 * "snapname" -> full name of the snapshot to create
4723 * (optional) "props" -> received properties to set (nvlist)
4724 * (optional) "localprops" -> override and exclude properties (nvlist)
4725 * (optional) "origin" -> name of clone origin (DRR_FLAG_CLONE)
4726 * "begin_record" -> non-byteswapped dmu_replay_record_t
4727 * "input_fd" -> file descriptor to read stream from (int32)
4728 * (optional) "force" -> force flag (value ignored)
4729 * (optional) "resumable" -> resumable flag (value ignored)
4730 * (optional) "cleanup_fd" -> cleanup-on-exit file descriptor
4731 * (optional) "action_handle" -> handle for this guid/ds mapping
4735 * "read_bytes" -> number of bytes read
4736 * "error_flags" -> zprop_errflags_t
4737 * "action_handle" -> handle for this guid/ds mapping
4738 * "errors" -> error for each unapplied received property (nvlist)
4742 zfs_ioc_recv_new(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
4744 dmu_replay_record_t *begin_record;
4745 uint_t begin_record_size;
4746 nvlist_t *errors = NULL;
4747 nvlist_t *recvprops = NULL;
4748 nvlist_t *localprops = NULL;
4749 char *snapname = NULL;
4750 char *origin = NULL;
4752 char tofs[ZFS_MAX_DATASET_NAME_LEN];
4754 boolean_t resumable;
4755 uint64_t action_handle = 0;
4756 uint64_t read_bytes = 0;
4757 uint64_t errflags = 0;
4759 int cleanup_fd = -1;
4762 error = nvlist_lookup_string(innvl, "snapname", &snapname);
4764 return (SET_ERROR(EINVAL));
4766 if (dataset_namecheck(snapname, NULL, NULL) != 0 ||
4767 strchr(snapname, '@') == NULL ||
4768 strchr(snapname, '%'))
4769 return (SET_ERROR(EINVAL));
4771 (void) strcpy(tofs, snapname);
4772 tosnap = strchr(tofs, '@');
4775 error = nvlist_lookup_string(innvl, "origin", &origin);
4776 if (error && error != ENOENT)
4779 error = nvlist_lookup_byte_array(innvl, "begin_record",
4780 (uchar_t **)&begin_record, &begin_record_size);
4781 if (error != 0 || begin_record_size != sizeof (*begin_record))
4782 return (SET_ERROR(EINVAL));
4784 error = nvlist_lookup_int32(innvl, "input_fd", &input_fd);
4786 return (SET_ERROR(EINVAL));
4788 force = nvlist_exists(innvl, "force");
4789 resumable = nvlist_exists(innvl, "resumable");
4791 error = nvlist_lookup_int32(innvl, "cleanup_fd", &cleanup_fd);
4792 if (error && error != ENOENT)
4795 error = nvlist_lookup_uint64(innvl, "action_handle", &action_handle);
4796 if (error && error != ENOENT)
4799 /* we still use "props" here for backwards compatibility */
4800 error = nvlist_lookup_nvlist(innvl, "props", &recvprops);
4801 if (error && error != ENOENT)
4804 error = nvlist_lookup_nvlist(innvl, "localprops", &localprops);
4805 if (error && error != ENOENT)
4808 error = zfs_ioc_recv_impl(tofs, tosnap, origin, recvprops, localprops,
4809 force, resumable, input_fd, begin_record, cleanup_fd, &read_bytes,
4810 &errflags, &action_handle, &errors);
4812 fnvlist_add_uint64(outnvl, "read_bytes", read_bytes);
4813 fnvlist_add_uint64(outnvl, "error_flags", errflags);
4814 fnvlist_add_uint64(outnvl, "action_handle", action_handle);
4815 fnvlist_add_nvlist(outnvl, "errors", errors);
4817 nvlist_free(errors);
4818 nvlist_free(recvprops);
4819 nvlist_free(localprops);
4826 * zc_name name of snapshot to send
4827 * zc_cookie file descriptor to send stream to
4828 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj)
4829 * zc_sendobj objsetid of snapshot to send
4830 * zc_fromobj objsetid of incremental fromsnap (may be zero)
4831 * zc_guid if set, estimate size of stream only. zc_cookie is ignored.
4832 * output size in zc_objset_type.
4833 * zc_flags lzc_send_flags
4836 * zc_objset_type estimated size, if zc_guid is set
4838 * NOTE: This is no longer the preferred interface, any new functionality
4839 * should be added to zfs_ioc_send_new() instead.
4842 zfs_ioc_send(zfs_cmd_t *zc)
4846 boolean_t estimate = (zc->zc_guid != 0);
4847 boolean_t embedok = (zc->zc_flags & 0x1);
4848 boolean_t large_block_ok = (zc->zc_flags & 0x2);
4849 boolean_t compressok = (zc->zc_flags & 0x4);
4850 boolean_t rawok = (zc->zc_flags & 0x8);
4852 if (zc->zc_obj != 0) {
4854 dsl_dataset_t *tosnap;
4856 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4860 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4862 dsl_pool_rele(dp, FTAG);
4866 if (dsl_dir_is_clone(tosnap->ds_dir))
4868 dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj;
4869 dsl_dataset_rele(tosnap, FTAG);
4870 dsl_pool_rele(dp, FTAG);
4875 dsl_dataset_t *tosnap;
4876 dsl_dataset_t *fromsnap = NULL;
4878 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4882 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj,
4885 dsl_pool_rele(dp, FTAG);
4889 if (zc->zc_fromobj != 0) {
4890 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
4893 dsl_dataset_rele(tosnap, FTAG);
4894 dsl_pool_rele(dp, FTAG);
4899 error = dmu_send_estimate(tosnap, fromsnap, compressok || rawok,
4900 &zc->zc_objset_type);
4902 if (fromsnap != NULL)
4903 dsl_dataset_rele(fromsnap, FTAG);
4904 dsl_dataset_rele(tosnap, FTAG);
4905 dsl_pool_rele(dp, FTAG);
4907 file_t *fp = getf(zc->zc_cookie);
4909 return (SET_ERROR(EBADF));
4912 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
4913 zc->zc_fromobj, embedok, large_block_ok, compressok, rawok,
4914 zc->zc_cookie, fp->f_vnode, &off);
4916 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4918 releasef(zc->zc_cookie);
4925 * zc_name name of snapshot on which to report progress
4926 * zc_cookie file descriptor of send stream
4929 * zc_cookie number of bytes written in send stream thus far
4932 zfs_ioc_send_progress(zfs_cmd_t *zc)
4936 dmu_sendarg_t *dsp = NULL;
4939 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4943 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4945 dsl_pool_rele(dp, FTAG);
4949 mutex_enter(&ds->ds_sendstream_lock);
4952 * Iterate over all the send streams currently active on this dataset.
4953 * If there's one which matches the specified file descriptor _and_ the
4954 * stream was started by the current process, return the progress of
4958 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4959 dsp = list_next(&ds->ds_sendstreams, dsp)) {
4960 if (dsp->dsa_outfd == zc->zc_cookie &&
4961 dsp->dsa_proc->group_leader == curproc->group_leader)
4966 zc->zc_cookie = *(dsp->dsa_off);
4968 error = SET_ERROR(ENOENT);
4970 mutex_exit(&ds->ds_sendstream_lock);
4971 dsl_dataset_rele(ds, FTAG);
4972 dsl_pool_rele(dp, FTAG);
4977 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4981 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4982 &zc->zc_inject_record);
4985 zc->zc_guid = (uint64_t)id;
4991 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4993 return (zio_clear_fault((int)zc->zc_guid));
4997 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4999 int id = (int)zc->zc_guid;
5002 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
5003 &zc->zc_inject_record);
5011 zfs_ioc_error_log(zfs_cmd_t *zc)
5015 size_t count = (size_t)zc->zc_nvlist_dst_size;
5017 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
5020 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
5023 zc->zc_nvlist_dst_size = count;
5025 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
5027 spa_close(spa, FTAG);
5033 zfs_ioc_clear(zfs_cmd_t *zc)
5040 * On zpool clear we also fix up missing slogs
5042 mutex_enter(&spa_namespace_lock);
5043 spa = spa_lookup(zc->zc_name);
5045 mutex_exit(&spa_namespace_lock);
5046 return (SET_ERROR(EIO));
5048 if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
5049 /* we need to let spa_open/spa_load clear the chains */
5050 spa_set_log_state(spa, SPA_LOG_CLEAR);
5052 spa->spa_last_open_failed = 0;
5053 mutex_exit(&spa_namespace_lock);
5055 if (zc->zc_cookie & ZPOOL_NO_REWIND) {
5056 error = spa_open(zc->zc_name, &spa, FTAG);
5059 nvlist_t *config = NULL;
5061 if (zc->zc_nvlist_src == 0)
5062 return (SET_ERROR(EINVAL));
5064 if ((error = get_nvlist(zc->zc_nvlist_src,
5065 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
5066 error = spa_open_rewind(zc->zc_name, &spa, FTAG,
5068 if (config != NULL) {
5071 if ((err = put_nvlist(zc, config)) != 0)
5073 nvlist_free(config);
5075 nvlist_free(policy);
5082 spa_vdev_state_enter(spa, SCL_NONE);
5084 if (zc->zc_guid == 0) {
5087 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
5089 (void) spa_vdev_state_exit(spa, NULL, ENODEV);
5090 spa_close(spa, FTAG);
5091 return (SET_ERROR(ENODEV));
5095 vdev_clear(spa, vd);
5097 (void) spa_vdev_state_exit(spa, spa_suspended(spa) ?
5098 NULL : spa->spa_root_vdev, 0);
5101 * Resume any suspended I/Os.
5103 if (zio_resume(spa) != 0)
5104 error = SET_ERROR(EIO);
5106 spa_close(spa, FTAG);
5112 * Reopen all the vdevs associated with the pool.
5115 * "scrub_restart" -> when true and scrub is running, allow to restart
5116 * scrub as the side effect of the reopen (boolean).
5123 zfs_ioc_pool_reopen(const char *pool, nvlist_t *innvl, nvlist_t *outnvl)
5127 boolean_t scrub_restart = B_TRUE;
5130 if (nvlist_lookup_boolean_value(innvl, "scrub_restart",
5131 &scrub_restart) != 0) {
5132 return (SET_ERROR(EINVAL));
5136 error = spa_open(pool, &spa, FTAG);
5140 spa_vdev_state_enter(spa, SCL_NONE);
5143 * If the scrub_restart flag is B_FALSE and a scrub is already
5144 * in progress then set spa_scrub_reopen flag to B_TRUE so that
5145 * we don't restart the scrub as a side effect of the reopen.
5146 * Otherwise, let vdev_open() decided if a resilver is required.
5149 spa->spa_scrub_reopen = (!scrub_restart &&
5150 dsl_scan_scrubbing(spa->spa_dsl_pool));
5151 vdev_reopen(spa->spa_root_vdev);
5152 spa->spa_scrub_reopen = B_FALSE;
5154 (void) spa_vdev_state_exit(spa, NULL, 0);
5155 spa_close(spa, FTAG);
5161 * zc_name name of filesystem
5164 * zc_string name of conflicting snapshot, if there is one
5167 zfs_ioc_promote(zfs_cmd_t *zc)
5170 dsl_dataset_t *ds, *ods;
5171 char origin[ZFS_MAX_DATASET_NAME_LEN];
5175 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
5176 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0 ||
5177 strchr(zc->zc_name, '%'))
5178 return (SET_ERROR(EINVAL));
5180 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5184 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
5186 dsl_pool_rele(dp, FTAG);
5190 if (!dsl_dir_is_clone(ds->ds_dir)) {
5191 dsl_dataset_rele(ds, FTAG);
5192 dsl_pool_rele(dp, FTAG);
5193 return (SET_ERROR(EINVAL));
5196 error = dsl_dataset_hold_obj(dp,
5197 dsl_dir_phys(ds->ds_dir)->dd_origin_obj, FTAG, &ods);
5199 dsl_dataset_rele(ds, FTAG);
5200 dsl_pool_rele(dp, FTAG);
5204 dsl_dataset_name(ods, origin);
5205 dsl_dataset_rele(ods, FTAG);
5206 dsl_dataset_rele(ds, FTAG);
5207 dsl_pool_rele(dp, FTAG);
5210 * We don't need to unmount *all* the origin fs's snapshots, but
5213 cp = strchr(origin, '@');
5216 (void) dmu_objset_find(origin,
5217 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
5218 return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
5222 * Retrieve a single {user|group|project}{used|quota}@... property.
5225 * zc_name name of filesystem
5226 * zc_objset_type zfs_userquota_prop_t
5227 * zc_value domain name (eg. "S-1-234-567-89")
5228 * zc_guid RID/UID/GID
5231 * zc_cookie property value
5234 zfs_ioc_userspace_one(zfs_cmd_t *zc)
5239 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
5240 return (SET_ERROR(EINVAL));
5242 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
5246 error = zfs_userspace_one(zfsvfs,
5247 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
5248 zfsvfs_rele(zfsvfs, FTAG);
5255 * zc_name name of filesystem
5256 * zc_cookie zap cursor
5257 * zc_objset_type zfs_userquota_prop_t
5258 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
5261 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
5262 * zc_cookie zap cursor
5265 zfs_ioc_userspace_many(zfs_cmd_t *zc)
5268 int bufsize = zc->zc_nvlist_dst_size;
5271 return (SET_ERROR(ENOMEM));
5273 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
5277 void *buf = vmem_alloc(bufsize, KM_SLEEP);
5279 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
5280 buf, &zc->zc_nvlist_dst_size);
5283 error = xcopyout(buf,
5284 (void *)(uintptr_t)zc->zc_nvlist_dst,
5285 zc->zc_nvlist_dst_size);
5287 vmem_free(buf, bufsize);
5288 zfsvfs_rele(zfsvfs, FTAG);
5295 * zc_name name of filesystem
5301 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
5307 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
5308 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
5310 * If userused is not enabled, it may be because the
5311 * objset needs to be closed & reopened (to grow the
5312 * objset_phys_t). Suspend/resume the fs will do that.
5314 dsl_dataset_t *ds, *newds;
5316 ds = dmu_objset_ds(zfsvfs->z_os);
5317 error = zfs_suspend_fs(zfsvfs);
5319 dmu_objset_refresh_ownership(ds, &newds,
5321 error = zfs_resume_fs(zfsvfs, newds);
5325 error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
5326 deactivate_super(zfsvfs->z_sb);
5328 /* XXX kind of reading contents without owning */
5329 error = dmu_objset_hold_flags(zc->zc_name, B_TRUE, FTAG, &os);
5333 error = dmu_objset_userspace_upgrade(os);
5334 dmu_objset_rele_flags(os, B_TRUE, FTAG);
5342 * zc_name name of filesystem
5348 zfs_ioc_id_quota_upgrade(zfs_cmd_t *zc)
5353 error = dmu_objset_hold_flags(zc->zc_name, B_TRUE, FTAG, &os);
5357 if (dmu_objset_userobjspace_upgradable(os) ||
5358 dmu_objset_projectquota_upgradable(os)) {
5359 mutex_enter(&os->os_upgrade_lock);
5360 if (os->os_upgrade_id == 0) {
5361 /* clear potential error code and retry */
5362 os->os_upgrade_status = 0;
5363 mutex_exit(&os->os_upgrade_lock);
5365 dmu_objset_id_quota_upgrade(os);
5367 mutex_exit(&os->os_upgrade_lock);
5370 dsl_pool_rele(dmu_objset_pool(os), FTAG);
5372 taskq_wait_id(os->os_spa->spa_upgrade_taskq, os->os_upgrade_id);
5373 error = os->os_upgrade_status;
5375 dsl_pool_rele(dmu_objset_pool(os), FTAG);
5378 dsl_dataset_rele_flags(dmu_objset_ds(os), DS_HOLD_FLAG_DECRYPT, FTAG);
5384 zfs_ioc_share(zfs_cmd_t *zc)
5386 return (SET_ERROR(ENOSYS));
5389 ace_t full_access[] = {
5390 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
5395 * zc_name name of containing filesystem
5396 * zc_obj object # beyond which we want next in-use object #
5399 * zc_obj next in-use object #
5402 zfs_ioc_next_obj(zfs_cmd_t *zc)
5404 objset_t *os = NULL;
5407 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
5411 error = dmu_object_next(os, &zc->zc_obj, B_FALSE, 0);
5413 dmu_objset_rele(os, FTAG);
5419 * zc_name name of filesystem
5420 * zc_value prefix name for snapshot
5421 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
5424 * zc_value short name of new snapshot
5427 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
5434 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
5438 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
5439 (u_longlong_t)ddi_get_lbolt64());
5440 hold_name = kmem_asprintf("%%%s", zc->zc_value);
5442 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
5445 (void) strlcpy(zc->zc_value, snap_name,
5446 sizeof (zc->zc_value));
5449 zfs_onexit_fd_rele(zc->zc_cleanup_fd);
5455 * zc_name name of "to" snapshot
5456 * zc_value name of "from" snapshot
5457 * zc_cookie file descriptor to write diff data on
5460 * dmu_diff_record_t's to the file descriptor
5463 zfs_ioc_diff(zfs_cmd_t *zc)
5469 fp = getf(zc->zc_cookie);
5471 return (SET_ERROR(EBADF));
5475 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off);
5477 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5479 releasef(zc->zc_cookie);
5485 * Remove all ACL files in shares dir
5487 #ifdef HAVE_SMB_SHARE
5489 zfs_smb_acl_purge(znode_t *dzp)
5492 zap_attribute_t zap;
5493 zfsvfs_t *zfsvfs = ZTOZSB(dzp);
5496 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
5497 (error = zap_cursor_retrieve(&zc, &zap)) == 0;
5498 zap_cursor_advance(&zc)) {
5499 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
5503 zap_cursor_fini(&zc);
5506 #endif /* HAVE_SMB_SHARE */
5509 zfs_ioc_smb_acl(zfs_cmd_t *zc)
5511 #ifdef HAVE_SMB_SHARE
5514 vnode_t *resourcevp = NULL;
5523 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
5524 NO_FOLLOW, NULL, &vp)) != 0)
5527 /* Now make sure mntpnt and dataset are ZFS */
5529 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
5530 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
5531 zc->zc_name) != 0)) {
5533 return (SET_ERROR(EINVAL));
5537 zfsvfs = ZTOZSB(dzp);
5541 * Create share dir if its missing.
5543 mutex_enter(&zfsvfs->z_lock);
5544 if (zfsvfs->z_shares_dir == 0) {
5547 tx = dmu_tx_create(zfsvfs->z_os);
5548 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
5550 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
5551 error = dmu_tx_assign(tx, TXG_WAIT);
5555 error = zfs_create_share_dir(zfsvfs, tx);
5559 mutex_exit(&zfsvfs->z_lock);
5565 mutex_exit(&zfsvfs->z_lock);
5567 ASSERT(zfsvfs->z_shares_dir);
5568 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
5574 switch (zc->zc_cookie) {
5575 case ZFS_SMB_ACL_ADD:
5576 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
5577 vattr.va_mode = S_IFREG|0777;
5581 vsec.vsa_mask = VSA_ACE;
5582 vsec.vsa_aclentp = &full_access;
5583 vsec.vsa_aclentsz = sizeof (full_access);
5584 vsec.vsa_aclcnt = 1;
5586 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
5587 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
5589 VN_RELE(resourcevp);
5592 case ZFS_SMB_ACL_REMOVE:
5593 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
5597 case ZFS_SMB_ACL_RENAME:
5598 if ((error = get_nvlist(zc->zc_nvlist_src,
5599 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
5601 VN_RELE(ZTOV(sharedir));
5605 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
5606 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
5609 VN_RELE(ZTOV(sharedir));
5611 nvlist_free(nvlist);
5614 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
5616 nvlist_free(nvlist);
5619 case ZFS_SMB_ACL_PURGE:
5620 error = zfs_smb_acl_purge(sharedir);
5624 error = SET_ERROR(EINVAL);
5629 VN_RELE(ZTOV(sharedir));
5635 return (SET_ERROR(ENOTSUP));
5636 #endif /* HAVE_SMB_SHARE */
5641 * "holds" -> { snapname -> holdname (string), ... }
5642 * (optional) "cleanup_fd" -> fd (int32)
5646 * snapname -> error value (int32)
5652 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
5656 int cleanup_fd = -1;
5660 error = nvlist_lookup_nvlist(args, "holds", &holds);
5662 return (SET_ERROR(EINVAL));
5664 /* make sure the user didn't pass us any invalid (empty) tags */
5665 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
5666 pair = nvlist_next_nvpair(holds, pair)) {
5669 error = nvpair_value_string(pair, &htag);
5671 return (SET_ERROR(error));
5673 if (strlen(htag) == 0)
5674 return (SET_ERROR(EINVAL));
5677 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
5678 error = zfs_onexit_fd_hold(cleanup_fd, &minor);
5683 error = dsl_dataset_user_hold(holds, minor, errlist);
5685 zfs_onexit_fd_rele(cleanup_fd);
5690 * innvl is not used.
5693 * holdname -> time added (uint64 seconds since epoch)
5699 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
5701 ASSERT3P(args, ==, NULL);
5702 return (dsl_dataset_get_holds(snapname, outnvl));
5707 * snapname -> { holdname, ... }
5712 * snapname -> error value (int32)
5718 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
5720 return (dsl_dataset_user_release(holds, errlist));
5725 * zc_guid flags (ZEVENT_NONBLOCK)
5726 * zc_cleanup_fd zevent file descriptor
5729 * zc_nvlist_dst next nvlist event
5730 * zc_cookie dropped events since last get
5733 zfs_ioc_events_next(zfs_cmd_t *zc)
5736 nvlist_t *event = NULL;
5738 uint64_t dropped = 0;
5741 error = zfs_zevent_fd_hold(zc->zc_cleanup_fd, &minor, &ze);
5746 error = zfs_zevent_next(ze, &event,
5747 &zc->zc_nvlist_dst_size, &dropped);
5748 if (event != NULL) {
5749 zc->zc_cookie = dropped;
5750 error = put_nvlist(zc, event);
5754 if (zc->zc_guid & ZEVENT_NONBLOCK)
5757 if ((error == 0) || (error != ENOENT))
5760 error = zfs_zevent_wait(ze);
5765 zfs_zevent_fd_rele(zc->zc_cleanup_fd);
5772 * zc_cookie cleared events count
5775 zfs_ioc_events_clear(zfs_cmd_t *zc)
5779 zfs_zevent_drain_all(&count);
5780 zc->zc_cookie = count;
5787 * zc_guid eid | ZEVENT_SEEK_START | ZEVENT_SEEK_END
5788 * zc_cleanup zevent file descriptor
5791 zfs_ioc_events_seek(zfs_cmd_t *zc)
5797 error = zfs_zevent_fd_hold(zc->zc_cleanup_fd, &minor, &ze);
5801 error = zfs_zevent_seek(ze, zc->zc_guid);
5802 zfs_zevent_fd_rele(zc->zc_cleanup_fd);
5809 * zc_name name of new filesystem or snapshot
5810 * zc_value full name of old snapshot
5813 * zc_cookie space in bytes
5814 * zc_objset_type compressed space in bytes
5815 * zc_perm_action uncompressed space in bytes
5818 zfs_ioc_space_written(zfs_cmd_t *zc)
5822 dsl_dataset_t *new, *old;
5824 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5827 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
5829 dsl_pool_rele(dp, FTAG);
5832 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
5834 dsl_dataset_rele(new, FTAG);
5835 dsl_pool_rele(dp, FTAG);
5839 error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
5840 &zc->zc_objset_type, &zc->zc_perm_action);
5841 dsl_dataset_rele(old, FTAG);
5842 dsl_dataset_rele(new, FTAG);
5843 dsl_pool_rele(dp, FTAG);
5849 * "firstsnap" -> snapshot name
5853 * "used" -> space in bytes
5854 * "compressed" -> compressed space in bytes
5855 * "uncompressed" -> uncompressed space in bytes
5859 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
5863 dsl_dataset_t *new, *old;
5865 uint64_t used, comp, uncomp;
5867 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5868 return (SET_ERROR(EINVAL));
5870 error = dsl_pool_hold(lastsnap, FTAG, &dp);
5874 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
5875 if (error == 0 && !new->ds_is_snapshot) {
5876 dsl_dataset_rele(new, FTAG);
5877 error = SET_ERROR(EINVAL);
5880 dsl_pool_rele(dp, FTAG);
5883 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
5884 if (error == 0 && !old->ds_is_snapshot) {
5885 dsl_dataset_rele(old, FTAG);
5886 error = SET_ERROR(EINVAL);
5889 dsl_dataset_rele(new, FTAG);
5890 dsl_pool_rele(dp, FTAG);
5894 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5895 dsl_dataset_rele(old, FTAG);
5896 dsl_dataset_rele(new, FTAG);
5897 dsl_pool_rele(dp, FTAG);
5898 fnvlist_add_uint64(outnvl, "used", used);
5899 fnvlist_add_uint64(outnvl, "compressed", comp);
5900 fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5906 * "fd" -> file descriptor to write stream to (int32)
5907 * (optional) "fromsnap" -> full snap name to send an incremental from
5908 * (optional) "largeblockok" -> (value ignored)
5909 * indicates that blocks > 128KB are permitted
5910 * (optional) "embedok" -> (value ignored)
5911 * presence indicates DRR_WRITE_EMBEDDED records are permitted
5912 * (optional) "compressok" -> (value ignored)
5913 * presence indicates compressed DRR_WRITE records are permitted
5914 * (optional) "rawok" -> (value ignored)
5915 * presence indicates raw encrypted records should be used.
5916 * (optional) "resume_object" and "resume_offset" -> (uint64)
5917 * if present, resume send stream from specified object and offset.
5924 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5928 char *fromname = NULL;
5931 boolean_t largeblockok;
5933 boolean_t compressok;
5935 uint64_t resumeobj = 0;
5936 uint64_t resumeoff = 0;
5938 error = nvlist_lookup_int32(innvl, "fd", &fd);
5940 return (SET_ERROR(EINVAL));
5942 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
5944 largeblockok = nvlist_exists(innvl, "largeblockok");
5945 embedok = nvlist_exists(innvl, "embedok");
5946 compressok = nvlist_exists(innvl, "compressok");
5947 rawok = nvlist_exists(innvl, "rawok");
5949 (void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj);
5950 (void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff);
5952 if ((fp = getf(fd)) == NULL)
5953 return (SET_ERROR(EBADF));
5956 error = dmu_send(snapname, fromname, embedok, largeblockok, compressok,
5957 rawok, fd, resumeobj, resumeoff, fp->f_vnode, &off);
5959 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5967 * Determine approximately how large a zfs send stream will be -- the number
5968 * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5971 * (optional) "from" -> full snap or bookmark name to send an incremental
5973 * (optional) "largeblockok" -> (value ignored)
5974 * indicates that blocks > 128KB are permitted
5975 * (optional) "embedok" -> (value ignored)
5976 * presence indicates DRR_WRITE_EMBEDDED records are permitted
5977 * (optional) "compressok" -> (value ignored)
5978 * presence indicates compressed DRR_WRITE records are permitted
5979 * (optional) "rawok" -> (value ignored)
5980 * presence indicates raw encrypted records should be used.
5984 * "space" -> bytes of space (uint64)
5988 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5991 dsl_dataset_t *tosnap;
5994 boolean_t compressok;
5998 error = dsl_pool_hold(snapname, FTAG, &dp);
6002 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
6004 dsl_pool_rele(dp, FTAG);
6008 compressok = nvlist_exists(innvl, "compressok");
6009 rawok = nvlist_exists(innvl, "rawok");
6011 error = nvlist_lookup_string(innvl, "from", &fromname);
6013 if (strchr(fromname, '@') != NULL) {
6015 * If from is a snapshot, hold it and use the more
6016 * efficient dmu_send_estimate to estimate send space
6017 * size using deadlists.
6019 dsl_dataset_t *fromsnap;
6020 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
6023 error = dmu_send_estimate(tosnap, fromsnap,
6024 compressok || rawok, &space);
6025 dsl_dataset_rele(fromsnap, FTAG);
6026 } else if (strchr(fromname, '#') != NULL) {
6028 * If from is a bookmark, fetch the creation TXG of the
6029 * snapshot it was created from and use that to find
6030 * blocks that were born after it.
6032 zfs_bookmark_phys_t frombm;
6034 error = dsl_bookmark_lookup(dp, fromname, tosnap,
6038 error = dmu_send_estimate_from_txg(tosnap,
6039 frombm.zbm_creation_txg, compressok || rawok,
6043 * from is not properly formatted as a snapshot or
6046 error = SET_ERROR(EINVAL);
6051 * If estimating the size of a full send, use dmu_send_estimate.
6053 error = dmu_send_estimate(tosnap, NULL, compressok || rawok,
6057 fnvlist_add_uint64(outnvl, "space", space);
6060 dsl_dataset_rele(tosnap, FTAG);
6061 dsl_pool_rele(dp, FTAG);
6066 * Sync the currently open TXG to disk for the specified pool.
6067 * This is somewhat similar to 'zfs_sync()'.
6068 * For cases that do not result in error this ioctl will wait for
6069 * the currently open TXG to commit before returning back to the caller.
6072 * "force" -> when true, force uberblock update even if there is no dirty data.
6073 * In addition this will cause the vdev configuration to be written
6074 * out including updating the zpool cache file. (boolean_t)
6081 zfs_ioc_pool_sync(const char *pool, nvlist_t *innvl, nvlist_t *onvl)
6084 boolean_t force = B_FALSE;
6087 if ((err = spa_open(pool, &spa, FTAG)) != 0)
6091 if (nvlist_lookup_boolean_value(innvl, "force", &force) != 0) {
6092 err = SET_ERROR(EINVAL);
6098 spa_config_enter(spa, SCL_CONFIG, FTAG, RW_WRITER);
6099 vdev_config_dirty(spa->spa_root_vdev);
6100 spa_config_exit(spa, SCL_CONFIG, FTAG);
6102 txg_wait_synced(spa_get_dsl(spa), 0);
6104 spa_close(spa, FTAG);
6110 * Load a user's wrapping key into the kernel.
6112 * "hidden_args" -> { "wkeydata" -> value }
6113 * raw uint8_t array of encryption wrapping key data (32 bytes)
6114 * (optional) "noop" -> (value ignored)
6115 * presence indicated key should only be verified, not loaded
6120 zfs_ioc_load_key(const char *dsname, nvlist_t *innvl, nvlist_t *outnvl)
6123 dsl_crypto_params_t *dcp = NULL;
6124 nvlist_t *hidden_args;
6125 boolean_t noop = nvlist_exists(innvl, "noop");
6127 if (strchr(dsname, '@') != NULL || strchr(dsname, '%') != NULL) {
6128 ret = SET_ERROR(EINVAL);
6132 ret = nvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS, &hidden_args);
6134 ret = SET_ERROR(EINVAL);
6138 ret = dsl_crypto_params_create_nvlist(DCP_CMD_NONE, NULL,
6143 ret = spa_keystore_load_wkey(dsname, dcp, noop);
6147 dsl_crypto_params_free(dcp, noop);
6152 dsl_crypto_params_free(dcp, B_TRUE);
6157 * Unload a user's wrapping key from the kernel.
6158 * Both innvl and outnvl are unused.
6162 zfs_ioc_unload_key(const char *dsname, nvlist_t *innvl, nvlist_t *outnvl)
6166 if (strchr(dsname, '@') != NULL || strchr(dsname, '%') != NULL) {
6167 ret = (SET_ERROR(EINVAL));
6171 ret = spa_keystore_unload_wkey(dsname);
6180 * Changes a user's wrapping key used to decrypt a dataset. The keyformat,
6181 * keylocation, pbkdf2salt, and pbkdf2iters properties can also be specified
6182 * here to change how the key is derived in userspace.
6185 * "hidden_args" (optional) -> { "wkeydata" -> value }
6186 * raw uint8_t array of new encryption wrapping key data (32 bytes)
6187 * "props" (optional) -> { prop -> value }
6194 zfs_ioc_change_key(const char *dsname, nvlist_t *innvl, nvlist_t *outnvl)
6197 uint64_t cmd = DCP_CMD_NONE;
6198 dsl_crypto_params_t *dcp = NULL;
6199 nvlist_t *args = NULL, *hidden_args = NULL;
6201 if (strchr(dsname, '@') != NULL || strchr(dsname, '%') != NULL) {
6202 ret = (SET_ERROR(EINVAL));
6206 (void) nvlist_lookup_uint64(innvl, "crypt_cmd", &cmd);
6207 (void) nvlist_lookup_nvlist(innvl, "props", &args);
6208 (void) nvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS, &hidden_args);
6210 ret = dsl_crypto_params_create_nvlist(cmd, args, hidden_args, &dcp);
6214 ret = spa_keystore_change_key(dsname, dcp);
6218 dsl_crypto_params_free(dcp, B_FALSE);
6223 dsl_crypto_params_free(dcp, B_TRUE);
6227 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
6230 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
6231 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
6232 boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
6234 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
6236 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
6237 ASSERT3U(ioc, <, ZFS_IOC_LAST);
6238 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
6239 ASSERT3P(vec->zvec_func, ==, NULL);
6241 vec->zvec_legacy_func = func;
6242 vec->zvec_secpolicy = secpolicy;
6243 vec->zvec_namecheck = namecheck;
6244 vec->zvec_allow_log = log_history;
6245 vec->zvec_pool_check = pool_check;
6249 * See the block comment at the beginning of this file for details on
6250 * each argument to this function.
6253 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
6254 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
6255 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
6256 boolean_t allow_log)
6258 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
6260 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
6261 ASSERT3U(ioc, <, ZFS_IOC_LAST);
6262 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
6263 ASSERT3P(vec->zvec_func, ==, NULL);
6265 /* if we are logging, the name must be valid */
6266 ASSERT(!allow_log || namecheck != NO_NAME);
6268 vec->zvec_name = name;
6269 vec->zvec_func = func;
6270 vec->zvec_secpolicy = secpolicy;
6271 vec->zvec_namecheck = namecheck;
6272 vec->zvec_pool_check = pool_check;
6273 vec->zvec_smush_outnvlist = smush_outnvlist;
6274 vec->zvec_allow_log = allow_log;
6278 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
6279 zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
6280 zfs_ioc_poolcheck_t pool_check)
6282 zfs_ioctl_register_legacy(ioc, func, secpolicy,
6283 POOL_NAME, log_history, pool_check);
6287 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
6288 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
6290 zfs_ioctl_register_legacy(ioc, func, secpolicy,
6291 DATASET_NAME, B_FALSE, pool_check);
6295 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
6297 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
6298 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
6302 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
6303 zfs_secpolicy_func_t *secpolicy)
6305 zfs_ioctl_register_legacy(ioc, func, secpolicy,
6306 NO_NAME, B_FALSE, POOL_CHECK_NONE);
6310 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
6311 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
6313 zfs_ioctl_register_legacy(ioc, func, secpolicy,
6314 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
6318 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
6320 zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
6321 zfs_secpolicy_read);
6325 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
6326 zfs_secpolicy_func_t *secpolicy)
6328 zfs_ioctl_register_legacy(ioc, func, secpolicy,
6329 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
6333 zfs_ioctl_init(void)
6335 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
6336 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
6337 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6339 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
6340 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
6341 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
6343 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
6344 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
6345 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
6347 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
6348 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
6349 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
6351 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
6352 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
6353 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
6355 zfs_ioctl_register("create", ZFS_IOC_CREATE,
6356 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
6357 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6359 zfs_ioctl_register("clone", ZFS_IOC_CLONE,
6360 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
6361 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6363 zfs_ioctl_register("remap", ZFS_IOC_REMAP,
6364 zfs_ioc_remap, zfs_secpolicy_remap, DATASET_NAME,
6365 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE);
6367 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
6368 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
6369 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6371 zfs_ioctl_register("hold", ZFS_IOC_HOLD,
6372 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
6373 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6374 zfs_ioctl_register("release", ZFS_IOC_RELEASE,
6375 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
6376 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6378 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
6379 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
6380 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
6382 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
6383 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
6384 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE);
6386 zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK,
6387 zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME,
6388 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6390 zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS,
6391 zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME,
6392 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
6394 zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS,
6395 zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks,
6397 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6399 zfs_ioctl_register("receive", ZFS_IOC_RECV_NEW,
6400 zfs_ioc_recv_new, zfs_secpolicy_recv_new, DATASET_NAME,
6401 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6402 zfs_ioctl_register("load-key", ZFS_IOC_LOAD_KEY,
6403 zfs_ioc_load_key, zfs_secpolicy_load_key,
6404 DATASET_NAME, POOL_CHECK_SUSPENDED, B_TRUE, B_TRUE);
6405 zfs_ioctl_register("unload-key", ZFS_IOC_UNLOAD_KEY,
6406 zfs_ioc_unload_key, zfs_secpolicy_load_key,
6407 DATASET_NAME, POOL_CHECK_SUSPENDED, B_TRUE, B_TRUE);
6408 zfs_ioctl_register("change-key", ZFS_IOC_CHANGE_KEY,
6409 zfs_ioc_change_key, zfs_secpolicy_change_key,
6410 DATASET_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY,
6413 zfs_ioctl_register("sync", ZFS_IOC_POOL_SYNC,
6414 zfs_ioc_pool_sync, zfs_secpolicy_none, POOL_NAME,
6415 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
6416 zfs_ioctl_register("reopen", ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
6417 zfs_secpolicy_config, POOL_NAME, POOL_CHECK_SUSPENDED, B_TRUE,
6420 zfs_ioctl_register("channel_program", ZFS_IOC_CHANNEL_PROGRAM,
6421 zfs_ioc_channel_program, zfs_secpolicy_config,
6422 POOL_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE,
6425 /* IOCTLS that use the legacy function signature */
6427 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
6428 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
6430 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
6431 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
6432 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
6434 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
6435 zfs_ioc_pool_upgrade);
6436 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
6438 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
6439 zfs_ioc_vdev_remove);
6440 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
6441 zfs_ioc_vdev_set_state);
6442 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
6443 zfs_ioc_vdev_attach);
6444 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
6445 zfs_ioc_vdev_detach);
6446 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
6447 zfs_ioc_vdev_setpath);
6448 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
6449 zfs_ioc_vdev_setfru);
6450 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
6451 zfs_ioc_pool_set_props);
6452 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
6453 zfs_ioc_vdev_split);
6454 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
6455 zfs_ioc_pool_reguid);
6457 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
6458 zfs_ioc_pool_configs, zfs_secpolicy_none);
6459 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
6460 zfs_ioc_pool_tryimport, zfs_secpolicy_config);
6461 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
6462 zfs_ioc_inject_fault, zfs_secpolicy_inject);
6463 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
6464 zfs_ioc_clear_fault, zfs_secpolicy_inject);
6465 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
6466 zfs_ioc_inject_list_next, zfs_secpolicy_inject);
6469 * pool destroy, and export don't log the history as part of
6470 * zfsdev_ioctl, but rather zfs_ioc_pool_export
6471 * does the logging of those commands.
6473 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
6474 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
6475 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
6476 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
6478 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
6479 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
6480 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
6481 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
6483 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
6484 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED);
6485 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
6486 zfs_ioc_dsobj_to_dsname,
6487 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED);
6488 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
6489 zfs_ioc_pool_get_history,
6490 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
6492 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
6493 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
6495 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
6496 zfs_secpolicy_config, B_TRUE, POOL_CHECK_READONLY);
6498 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
6499 zfs_ioc_space_written);
6500 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
6501 zfs_ioc_objset_recvd_props);
6502 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
6504 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
6506 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
6507 zfs_ioc_objset_stats);
6508 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
6509 zfs_ioc_objset_zplprops);
6510 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
6511 zfs_ioc_dataset_list_next);
6512 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
6513 zfs_ioc_snapshot_list_next);
6514 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
6515 zfs_ioc_send_progress);
6517 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
6518 zfs_ioc_diff, zfs_secpolicy_diff);
6519 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
6520 zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
6521 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
6522 zfs_ioc_obj_to_path, zfs_secpolicy_diff);
6523 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
6524 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
6525 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
6526 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
6527 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
6528 zfs_ioc_send, zfs_secpolicy_send);
6530 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
6531 zfs_secpolicy_none);
6532 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
6533 zfs_secpolicy_destroy);
6534 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
6535 zfs_secpolicy_rename);
6536 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
6537 zfs_secpolicy_recv);
6538 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
6539 zfs_secpolicy_promote);
6540 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
6541 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
6542 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
6543 zfs_secpolicy_set_fsacl);
6545 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
6546 zfs_secpolicy_share, POOL_CHECK_NONE);
6547 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
6548 zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
6549 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
6550 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
6551 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
6552 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
6553 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
6554 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
6559 zfs_ioctl_register_legacy(ZFS_IOC_EVENTS_NEXT, zfs_ioc_events_next,
6560 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);
6561 zfs_ioctl_register_legacy(ZFS_IOC_EVENTS_CLEAR, zfs_ioc_events_clear,
6562 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);
6563 zfs_ioctl_register_legacy(ZFS_IOC_EVENTS_SEEK, zfs_ioc_events_seek,
6564 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);
6568 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
6569 zfs_ioc_poolcheck_t check)
6574 ASSERT(type == POOL_NAME || type == DATASET_NAME);
6576 if (check & POOL_CHECK_NONE)
6579 error = spa_open(name, &spa, FTAG);
6581 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
6582 error = SET_ERROR(EAGAIN);
6583 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
6584 error = SET_ERROR(EROFS);
6585 spa_close(spa, FTAG);
6591 zfsdev_get_state_impl(minor_t minor, enum zfsdev_state_type which)
6595 for (zs = zfsdev_state_list; zs != NULL; zs = zs->zs_next) {
6596 if (zs->zs_minor == minor) {
6600 return (zs->zs_onexit);
6602 return (zs->zs_zevent);
6613 zfsdev_get_state(minor_t minor, enum zfsdev_state_type which)
6617 ptr = zfsdev_get_state_impl(minor, which);
6623 zfsdev_getminor(struct file *filp, minor_t *minorp)
6625 zfsdev_state_t *zs, *fpd;
6627 ASSERT(filp != NULL);
6628 ASSERT(!MUTEX_HELD(&zfsdev_state_lock));
6630 fpd = filp->private_data;
6632 return (SET_ERROR(EBADF));
6634 mutex_enter(&zfsdev_state_lock);
6636 for (zs = zfsdev_state_list; zs != NULL; zs = zs->zs_next) {
6638 if (zs->zs_minor == -1)
6642 *minorp = fpd->zs_minor;
6643 mutex_exit(&zfsdev_state_lock);
6648 mutex_exit(&zfsdev_state_lock);
6650 return (SET_ERROR(EBADF));
6654 * Find a free minor number. The zfsdev_state_list is expected to
6655 * be short since it is only a list of currently open file handles.
6658 zfsdev_minor_alloc(void)
6660 static minor_t last_minor = 0;
6663 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
6665 for (m = last_minor + 1; m != last_minor; m++) {
6666 if (m > ZFSDEV_MAX_MINOR)
6668 if (zfsdev_get_state_impl(m, ZST_ALL) == NULL) {
6678 zfsdev_state_init(struct file *filp)
6680 zfsdev_state_t *zs, *zsprev = NULL;
6682 boolean_t newzs = B_FALSE;
6684 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
6686 minor = zfsdev_minor_alloc();
6688 return (SET_ERROR(ENXIO));
6690 for (zs = zfsdev_state_list; zs != NULL; zs = zs->zs_next) {
6691 if (zs->zs_minor == -1)
6697 zs = kmem_zalloc(sizeof (zfsdev_state_t), KM_SLEEP);
6702 filp->private_data = zs;
6704 zfs_onexit_init((zfs_onexit_t **)&zs->zs_onexit);
6705 zfs_zevent_init((zfs_zevent_t **)&zs->zs_zevent);
6709 * In order to provide for lock-free concurrent read access
6710 * to the minor list in zfsdev_get_state_impl(), new entries
6711 * must be completely written before linking them into the
6712 * list whereas existing entries are already linked; the last
6713 * operation must be updating zs_minor (from -1 to the new
6717 zs->zs_minor = minor;
6719 zsprev->zs_next = zs;
6722 zs->zs_minor = minor;
6729 zfsdev_state_destroy(struct file *filp)
6733 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
6734 ASSERT(filp->private_data != NULL);
6736 zs = filp->private_data;
6738 zfs_onexit_destroy(zs->zs_onexit);
6739 zfs_zevent_destroy(zs->zs_zevent);
6745 zfsdev_open(struct inode *ino, struct file *filp)
6749 mutex_enter(&zfsdev_state_lock);
6750 error = zfsdev_state_init(filp);
6751 mutex_exit(&zfsdev_state_lock);
6757 zfsdev_release(struct inode *ino, struct file *filp)
6761 mutex_enter(&zfsdev_state_lock);
6762 error = zfsdev_state_destroy(filp);
6763 mutex_exit(&zfsdev_state_lock);
6769 zfsdev_ioctl(struct file *filp, unsigned cmd, unsigned long arg)
6773 int error, rc, flag = 0;
6774 const zfs_ioc_vec_t *vec;
6775 char *saved_poolname = NULL;
6776 nvlist_t *innvl = NULL;
6777 fstrans_cookie_t cookie;
6779 vecnum = cmd - ZFS_IOC_FIRST;
6780 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
6781 return (-SET_ERROR(EINVAL));
6782 vec = &zfs_ioc_vec[vecnum];
6785 * The registered ioctl list may be sparse, verify that either
6786 * a normal or legacy handler are registered.
6788 if (vec->zvec_func == NULL && vec->zvec_legacy_func == NULL)
6789 return (-SET_ERROR(EINVAL));
6791 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
6793 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
6795 error = SET_ERROR(EFAULT);
6799 zc->zc_iflags = flag & FKIOCTL;
6800 if (zc->zc_nvlist_src_size > MAX_NVLIST_SRC_SIZE) {
6802 * Make sure the user doesn't pass in an insane value for
6803 * zc_nvlist_src_size. We have to check, since we will end
6804 * up allocating that much memory inside of get_nvlist(). This
6805 * prevents a nefarious user from allocating tons of kernel
6808 * Also, we return EINVAL instead of ENOMEM here. The reason
6809 * being that returning ENOMEM from an ioctl() has a special
6810 * connotation; that the user's size value is too small and
6811 * needs to be expanded to hold the nvlist. See
6812 * zcmd_expand_dst_nvlist() for details.
6814 error = SET_ERROR(EINVAL); /* User's size too big */
6816 } else if (zc->zc_nvlist_src_size != 0) {
6817 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
6818 zc->zc_iflags, &innvl);
6824 * Ensure that all pool/dataset names are valid before we pass down to
6827 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
6828 switch (vec->zvec_namecheck) {
6830 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
6831 error = SET_ERROR(EINVAL);
6833 error = pool_status_check(zc->zc_name,
6834 vec->zvec_namecheck, vec->zvec_pool_check);
6838 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
6839 error = SET_ERROR(EINVAL);
6841 error = pool_status_check(zc->zc_name,
6842 vec->zvec_namecheck, vec->zvec_pool_check);
6851 cookie = spl_fstrans_mark();
6852 error = vec->zvec_secpolicy(zc, innvl, CRED());
6853 spl_fstrans_unmark(cookie);
6859 /* legacy ioctls can modify zc_name */
6860 saved_poolname = strdup(zc->zc_name);
6861 if (saved_poolname == NULL) {
6862 error = SET_ERROR(ENOMEM);
6865 saved_poolname[strcspn(saved_poolname, "/@#")] = '\0';
6868 if (vec->zvec_func != NULL) {
6872 nvlist_t *lognv = NULL;
6874 ASSERT(vec->zvec_legacy_func == NULL);
6877 * Add the innvl to the lognv before calling the func,
6878 * in case the func changes the innvl.
6880 if (vec->zvec_allow_log) {
6881 lognv = fnvlist_alloc();
6882 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
6884 if (!nvlist_empty(innvl)) {
6885 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
6890 outnvl = fnvlist_alloc();
6891 cookie = spl_fstrans_mark();
6892 error = vec->zvec_func(zc->zc_name, innvl, outnvl);
6893 spl_fstrans_unmark(cookie);
6896 * Some commands can partially execute, modify state, and still
6897 * return an error. In these cases, attempt to record what
6901 (cmd == ZFS_IOC_CHANNEL_PROGRAM && error != EINVAL)) &&
6902 vec->zvec_allow_log &&
6903 spa_open(zc->zc_name, &spa, FTAG) == 0) {
6904 if (!nvlist_empty(outnvl)) {
6905 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
6909 fnvlist_add_int64(lognv, ZPOOL_HIST_ERRNO,
6912 (void) spa_history_log_nvl(spa, lognv);
6913 spa_close(spa, FTAG);
6915 fnvlist_free(lognv);
6917 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
6919 if (vec->zvec_smush_outnvlist) {
6920 smusherror = nvlist_smush(outnvl,
6921 zc->zc_nvlist_dst_size);
6923 if (smusherror == 0)
6924 puterror = put_nvlist(zc, outnvl);
6930 nvlist_free(outnvl);
6932 cookie = spl_fstrans_mark();
6933 error = vec->zvec_legacy_func(zc);
6934 spl_fstrans_unmark(cookie);
6939 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
6940 if (error == 0 && rc != 0)
6941 error = SET_ERROR(EFAULT);
6942 if (error == 0 && vec->zvec_allow_log) {
6943 char *s = tsd_get(zfs_allow_log_key);
6946 (void) tsd_set(zfs_allow_log_key, saved_poolname);
6948 if (saved_poolname != NULL)
6949 strfree(saved_poolname);
6952 kmem_free(zc, sizeof (zfs_cmd_t));
6956 #ifdef CONFIG_COMPAT
6958 zfsdev_compat_ioctl(struct file *filp, unsigned cmd, unsigned long arg)
6960 return (zfsdev_ioctl(filp, cmd, arg));
6963 #define zfsdev_compat_ioctl NULL
6966 static const struct file_operations zfsdev_fops = {
6967 .open = zfsdev_open,
6968 .release = zfsdev_release,
6969 .unlocked_ioctl = zfsdev_ioctl,
6970 .compat_ioctl = zfsdev_compat_ioctl,
6971 .owner = THIS_MODULE,
6974 static struct miscdevice zfs_misc = {
6977 .fops = &zfsdev_fops,
6980 MODULE_ALIAS_MISCDEV(ZFS_MINOR);
6981 MODULE_ALIAS("devname:zfs");
6988 mutex_init(&zfsdev_state_lock, NULL, MUTEX_DEFAULT, NULL);
6989 zfsdev_state_list = kmem_zalloc(sizeof (zfsdev_state_t), KM_SLEEP);
6990 zfsdev_state_list->zs_minor = -1;
6992 error = misc_register(&zfs_misc);
6993 if (error == -EBUSY) {
6995 * Fallback to dynamic minor allocation in the event of a
6996 * collision with a reserved minor in linux/miscdevice.h.
6997 * In this case the kernel modules must be manually loaded.
6999 printk(KERN_INFO "ZFS: misc_register() with static minor %d "
7000 "failed %d, retrying with MISC_DYNAMIC_MINOR\n",
7003 zfs_misc.minor = MISC_DYNAMIC_MINOR;
7004 error = misc_register(&zfs_misc);
7008 printk(KERN_INFO "ZFS: misc_register() failed %d\n", error);
7016 zfsdev_state_t *zs, *zsprev = NULL;
7018 misc_deregister(&zfs_misc);
7019 mutex_destroy(&zfsdev_state_lock);
7021 for (zs = zfsdev_state_list; zs != NULL; zs = zs->zs_next) {
7023 kmem_free(zsprev, sizeof (zfsdev_state_t));
7027 kmem_free(zsprev, sizeof (zfsdev_state_t));
7031 zfs_allow_log_destroy(void *arg)
7033 char *poolname = arg;
7035 if (poolname != NULL)
7040 #define ZFS_DEBUG_STR " (DEBUG mode)"
7042 #define ZFS_DEBUG_STR ""
7050 error = -vn_set_pwd("/");
7053 "ZFS: Warning unable to set pwd to '/': %d\n", error);
7057 if ((error = -zvol_init()) != 0)
7060 spa_init(FREAD | FWRITE);
7065 if ((error = zfs_attach()) != 0)
7068 tsd_create(&zfs_fsyncer_key, NULL);
7069 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
7070 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
7072 printk(KERN_NOTICE "ZFS: Loaded module v%s-%s%s, "
7073 "ZFS pool version %s, ZFS filesystem version %s\n",
7074 ZFS_META_VERSION, ZFS_META_RELEASE, ZFS_DEBUG_STR,
7075 SPA_VERSION_STRING, ZPL_VERSION_STRING);
7076 #ifndef CONFIG_FS_POSIX_ACL
7077 printk(KERN_NOTICE "ZFS: Posix ACLs disabled by kernel\n");
7078 #endif /* CONFIG_FS_POSIX_ACL */
7086 printk(KERN_NOTICE "ZFS: Failed to Load ZFS Filesystem v%s-%s%s"
7087 ", rc = %d\n", ZFS_META_VERSION, ZFS_META_RELEASE,
7088 ZFS_DEBUG_STR, error);
7101 tsd_destroy(&zfs_fsyncer_key);
7102 tsd_destroy(&rrw_tsd_key);
7103 tsd_destroy(&zfs_allow_log_key);
7105 printk(KERN_NOTICE "ZFS: Unloaded module v%s-%s%s\n",
7106 ZFS_META_VERSION, ZFS_META_RELEASE, ZFS_DEBUG_STR);
7109 #if defined(_KERNEL)
7113 MODULE_DESCRIPTION("ZFS");
7114 MODULE_AUTHOR(ZFS_META_AUTHOR);
7115 MODULE_LICENSE(ZFS_META_LICENSE);
7116 MODULE_VERSION(ZFS_META_VERSION "-" ZFS_META_RELEASE);