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>
148 #include <sys/modctl.h>
149 #include <sys/open.h>
150 #include <sys/file.h>
151 #include <sys/kmem.h>
152 #include <sys/conf.h>
153 #include <sys/cmn_err.h>
154 #include <sys/stat.h>
155 #include <sys/zfs_ioctl.h>
156 #include <sys/zfs_vfsops.h>
157 #include <sys/zfs_znode.h>
160 #include <sys/spa_impl.h>
161 #include <sys/vdev.h>
162 #include <sys/vdev_impl.h>
163 #include <sys/priv_impl.h>
165 #include <sys/dsl_dir.h>
166 #include <sys/dsl_dataset.h>
167 #include <sys/dsl_prop.h>
168 #include <sys/dsl_deleg.h>
169 #include <sys/dmu_objset.h>
170 #include <sys/dmu_impl.h>
171 #include <sys/dmu_tx.h>
173 #include <sys/sunddi.h>
174 #include <sys/sunldi.h>
175 #include <sys/policy.h>
176 #include <sys/zone.h>
177 #include <sys/nvpair.h>
178 #include <sys/pathname.h>
179 #include <sys/mount.h>
181 #include <sys/fs/zfs.h>
182 #include <sys/zfs_ctldir.h>
183 #include <sys/zfs_dir.h>
184 #include <sys/zfs_onexit.h>
185 #include <sys/zvol.h>
186 #include <sys/dsl_scan.h>
187 #include <sharefs/share.h>
188 #include <sys/fm/util.h>
189 #include <sys/dsl_crypt.h>
191 #include <sys/dmu_send.h>
192 #include <sys/dsl_destroy.h>
193 #include <sys/dsl_bookmark.h>
194 #include <sys/dsl_userhold.h>
195 #include <sys/zfeature.h>
197 #include <sys/zio_checksum.h>
199 #include <linux/miscdevice.h>
200 #include <linux/slab.h>
202 #include "zfs_namecheck.h"
203 #include "zfs_prop.h"
204 #include "zfs_deleg.h"
205 #include "zfs_comutil.h"
207 #include <sys/lua/lua.h>
208 #include <sys/lua/lauxlib.h>
211 * Limit maximum nvlist size. We don't want users passing in insane values
212 * for zc->zc_nvlist_src_size, since we will need to allocate that much memory.
214 #define MAX_NVLIST_SRC_SIZE KMALLOC_MAX_SIZE
216 kmutex_t zfsdev_state_lock;
217 zfsdev_state_t *zfsdev_state_list;
219 extern void zfs_init(void);
220 extern void zfs_fini(void);
222 uint_t zfs_fsyncer_key;
223 extern uint_t rrw_tsd_key;
224 static uint_t zfs_allow_log_key;
226 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *);
227 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *);
228 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *);
234 } zfs_ioc_namecheck_t;
237 POOL_CHECK_NONE = 1 << 0,
238 POOL_CHECK_SUSPENDED = 1 << 1,
239 POOL_CHECK_READONLY = 1 << 2,
240 } zfs_ioc_poolcheck_t;
242 typedef struct zfs_ioc_vec {
243 zfs_ioc_legacy_func_t *zvec_legacy_func;
244 zfs_ioc_func_t *zvec_func;
245 zfs_secpolicy_func_t *zvec_secpolicy;
246 zfs_ioc_namecheck_t zvec_namecheck;
247 boolean_t zvec_allow_log;
248 zfs_ioc_poolcheck_t zvec_pool_check;
249 boolean_t zvec_smush_outnvlist;
250 const char *zvec_name;
253 /* This array is indexed by zfs_userquota_prop_t */
254 static const char *userquota_perms[] = {
255 ZFS_DELEG_PERM_USERUSED,
256 ZFS_DELEG_PERM_USERQUOTA,
257 ZFS_DELEG_PERM_GROUPUSED,
258 ZFS_DELEG_PERM_GROUPQUOTA,
259 ZFS_DELEG_PERM_USEROBJUSED,
260 ZFS_DELEG_PERM_USEROBJQUOTA,
261 ZFS_DELEG_PERM_GROUPOBJUSED,
262 ZFS_DELEG_PERM_GROUPOBJQUOTA,
263 ZFS_DELEG_PERM_PROJECTUSED,
264 ZFS_DELEG_PERM_PROJECTQUOTA,
265 ZFS_DELEG_PERM_PROJECTOBJUSED,
266 ZFS_DELEG_PERM_PROJECTOBJQUOTA,
269 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
270 static int zfs_ioc_id_quota_upgrade(zfs_cmd_t *zc);
271 static int zfs_check_settable(const char *name, nvpair_t *property,
273 static int zfs_check_clearable(char *dataset, nvlist_t *props,
275 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
277 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
278 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
281 history_str_free(char *buf)
283 kmem_free(buf, HIS_MAX_RECORD_LEN);
287 history_str_get(zfs_cmd_t *zc)
291 if (zc->zc_history == 0)
294 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
295 if (copyinstr((void *)(uintptr_t)zc->zc_history,
296 buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
297 history_str_free(buf);
301 buf[HIS_MAX_RECORD_LEN -1] = '\0';
307 * Check to see if the named dataset is currently defined as bootable
310 zfs_is_bootfs(const char *name)
314 if (dmu_objset_hold(name, FTAG, &os) == 0) {
316 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
317 dmu_objset_rele(os, FTAG);
324 * Return non-zero if the spa version is less than requested version.
327 zfs_earlier_version(const char *name, int version)
331 if (spa_open(name, &spa, FTAG) == 0) {
332 if (spa_version(spa) < version) {
333 spa_close(spa, FTAG);
336 spa_close(spa, FTAG);
342 * Return TRUE if the ZPL version is less than requested version.
345 zpl_earlier_version(const char *name, int version)
348 boolean_t rc = B_TRUE;
350 if (dmu_objset_hold(name, FTAG, &os) == 0) {
353 if (dmu_objset_type(os) != DMU_OST_ZFS) {
354 dmu_objset_rele(os, FTAG);
357 /* XXX reading from non-owned objset */
358 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
359 rc = zplversion < version;
360 dmu_objset_rele(os, FTAG);
366 zfs_log_history(zfs_cmd_t *zc)
371 if ((buf = history_str_get(zc)) == NULL)
374 if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
375 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
376 (void) spa_history_log(spa, buf);
377 spa_close(spa, FTAG);
379 history_str_free(buf);
383 * Policy for top-level read operations (list pools). Requires no privileges,
384 * and can be used in the local zone, as there is no associated dataset.
388 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
394 * Policy for dataset read operations (list children, get statistics). Requires
395 * no privileges, but must be visible in the local zone.
399 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
401 if (INGLOBALZONE(curproc) ||
402 zone_dataset_visible(zc->zc_name, NULL))
405 return (SET_ERROR(ENOENT));
409 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
414 * The dataset must be visible by this zone -- check this first
415 * so they don't see EPERM on something they shouldn't know about.
417 if (!INGLOBALZONE(curproc) &&
418 !zone_dataset_visible(dataset, &writable))
419 return (SET_ERROR(ENOENT));
421 if (INGLOBALZONE(curproc)) {
423 * If the fs is zoned, only root can access it from the
426 if (secpolicy_zfs(cr) && zoned)
427 return (SET_ERROR(EPERM));
430 * If we are in a local zone, the 'zoned' property must be set.
433 return (SET_ERROR(EPERM));
435 /* must be writable by this zone */
437 return (SET_ERROR(EPERM));
443 zfs_dozonecheck(const char *dataset, cred_t *cr)
447 if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
448 return (SET_ERROR(ENOENT));
450 return (zfs_dozonecheck_impl(dataset, zoned, cr));
454 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
458 if (dsl_prop_get_int_ds(ds, "zoned", &zoned))
459 return (SET_ERROR(ENOENT));
461 return (zfs_dozonecheck_impl(dataset, zoned, cr));
465 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
466 const char *perm, cred_t *cr)
470 error = zfs_dozonecheck_ds(name, ds, cr);
472 error = secpolicy_zfs(cr);
474 error = dsl_deleg_access_impl(ds, perm, cr);
480 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
487 * First do a quick check for root in the global zone, which
488 * is allowed to do all write_perms. This ensures that zfs_ioc_*
489 * will get to handle nonexistent datasets.
491 if (INGLOBALZONE(curproc) && secpolicy_zfs(cr) == 0)
494 error = dsl_pool_hold(name, FTAG, &dp);
498 error = dsl_dataset_hold(dp, name, FTAG, &ds);
500 dsl_pool_rele(dp, FTAG);
504 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
506 dsl_dataset_rele(ds, FTAG);
507 dsl_pool_rele(dp, FTAG);
512 * Policy for setting the security label property.
514 * Returns 0 for success, non-zero for access and other errors.
517 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
520 char ds_hexsl[MAXNAMELEN];
521 bslabel_t ds_sl, new_sl;
522 boolean_t new_default = FALSE;
524 int needed_priv = -1;
527 /* First get the existing dataset label. */
528 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
529 1, sizeof (ds_hexsl), &ds_hexsl, NULL);
531 return (SET_ERROR(EPERM));
533 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
536 /* The label must be translatable */
537 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
538 return (SET_ERROR(EINVAL));
541 * In a non-global zone, disallow attempts to set a label that
542 * doesn't match that of the zone; otherwise no other checks
545 if (!INGLOBALZONE(curproc)) {
546 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
547 return (SET_ERROR(EPERM));
552 * For global-zone datasets (i.e., those whose zoned property is
553 * "off", verify that the specified new label is valid for the
556 if (dsl_prop_get_integer(name,
557 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
558 return (SET_ERROR(EPERM));
560 if (zfs_check_global_label(name, strval) != 0)
561 return (SET_ERROR(EPERM));
565 * If the existing dataset label is nondefault, check if the
566 * dataset is mounted (label cannot be changed while mounted).
567 * Get the zfsvfs_t; if there isn't one, then the dataset isn't
568 * mounted (or isn't a dataset, doesn't exist, ...).
570 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
572 static char *setsl_tag = "setsl_tag";
575 * Try to own the dataset; abort if there is any error,
576 * (e.g., already mounted, in use, or other error).
578 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE, B_TRUE,
581 return (SET_ERROR(EPERM));
583 dmu_objset_disown(os, B_TRUE, setsl_tag);
586 needed_priv = PRIV_FILE_DOWNGRADE_SL;
590 if (hexstr_to_label(strval, &new_sl) != 0)
591 return (SET_ERROR(EPERM));
593 if (blstrictdom(&ds_sl, &new_sl))
594 needed_priv = PRIV_FILE_DOWNGRADE_SL;
595 else if (blstrictdom(&new_sl, &ds_sl))
596 needed_priv = PRIV_FILE_UPGRADE_SL;
598 /* dataset currently has a default label */
600 needed_priv = PRIV_FILE_UPGRADE_SL;
604 if (needed_priv != -1)
605 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
608 return (SET_ERROR(ENOTSUP));
609 #endif /* HAVE_MLSLABEL */
613 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
619 * Check permissions for special properties.
626 * Disallow setting of 'zoned' from within a local zone.
628 if (!INGLOBALZONE(curproc))
629 return (SET_ERROR(EPERM));
633 case ZFS_PROP_FILESYSTEM_LIMIT:
634 case ZFS_PROP_SNAPSHOT_LIMIT:
635 if (!INGLOBALZONE(curproc)) {
637 char setpoint[ZFS_MAX_DATASET_NAME_LEN];
639 * Unprivileged users are allowed to modify the
640 * limit on things *under* (ie. contained by)
641 * the thing they own.
643 if (dsl_prop_get_integer(dsname, "zoned", &zoned,
645 return (SET_ERROR(EPERM));
646 if (!zoned || strlen(dsname) <= strlen(setpoint))
647 return (SET_ERROR(EPERM));
651 case ZFS_PROP_MLSLABEL:
652 if (!is_system_labeled())
653 return (SET_ERROR(EPERM));
655 if (nvpair_value_string(propval, &strval) == 0) {
658 err = zfs_set_slabel_policy(dsname, strval, CRED());
665 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
670 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
674 error = zfs_dozonecheck(zc->zc_name, cr);
679 * permission to set permissions will be evaluated later in
680 * dsl_deleg_can_allow()
687 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
689 return (zfs_secpolicy_write_perms(zc->zc_name,
690 ZFS_DELEG_PERM_ROLLBACK, cr));
695 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
703 * Generate the current snapshot name from the given objsetid, then
704 * use that name for the secpolicy/zone checks.
706 cp = strchr(zc->zc_name, '@');
708 return (SET_ERROR(EINVAL));
709 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
713 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
715 dsl_pool_rele(dp, FTAG);
719 dsl_dataset_name(ds, zc->zc_name);
721 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
722 ZFS_DELEG_PERM_SEND, cr);
723 dsl_dataset_rele(ds, FTAG);
724 dsl_pool_rele(dp, FTAG);
731 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
733 return (zfs_secpolicy_write_perms(zc->zc_name,
734 ZFS_DELEG_PERM_SEND, cr));
737 #ifdef HAVE_SMB_SHARE
740 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
745 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
746 NO_FOLLOW, NULL, &vp)) != 0)
749 /* Now make sure mntpnt and dataset are ZFS */
751 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
752 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
753 zc->zc_name) != 0)) {
755 return (SET_ERROR(EPERM));
759 return (dsl_deleg_access(zc->zc_name,
760 ZFS_DELEG_PERM_SHARE, cr));
762 #endif /* HAVE_SMB_SHARE */
765 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
767 #ifdef HAVE_SMB_SHARE
768 if (!INGLOBALZONE(curproc))
769 return (SET_ERROR(EPERM));
771 if (secpolicy_nfs(cr) == 0) {
774 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
777 return (SET_ERROR(ENOTSUP));
778 #endif /* HAVE_SMB_SHARE */
782 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
784 #ifdef HAVE_SMB_SHARE
785 if (!INGLOBALZONE(curproc))
786 return (SET_ERROR(EPERM));
788 if (secpolicy_smb(cr) == 0) {
791 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
794 return (SET_ERROR(ENOTSUP));
795 #endif /* HAVE_SMB_SHARE */
799 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
804 * Remove the @bla or /bla from the end of the name to get the parent.
806 (void) strncpy(parent, datasetname, parentsize);
807 cp = strrchr(parent, '@');
811 cp = strrchr(parent, '/');
813 return (SET_ERROR(ENOENT));
821 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
825 if ((error = zfs_secpolicy_write_perms(name,
826 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
829 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
834 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
836 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
840 * Destroying snapshots with delegated permissions requires
841 * descendant mount and destroy permissions.
845 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
848 nvpair_t *pair, *nextpair;
851 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
852 return (SET_ERROR(EINVAL));
853 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
855 nextpair = nvlist_next_nvpair(snaps, pair);
856 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr);
857 if (error == ENOENT) {
859 * Ignore any snapshots that don't exist (we consider
860 * them "already destroyed"). Remove the name from the
861 * nvl here in case the snapshot is created between
862 * now and when we try to destroy it (in which case
863 * we don't want to destroy it since we haven't
864 * checked for permission).
866 fnvlist_remove_nvpair(snaps, pair);
877 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
879 char parentname[ZFS_MAX_DATASET_NAME_LEN];
882 if ((error = zfs_secpolicy_write_perms(from,
883 ZFS_DELEG_PERM_RENAME, cr)) != 0)
886 if ((error = zfs_secpolicy_write_perms(from,
887 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
890 if ((error = zfs_get_parent(to, parentname,
891 sizeof (parentname))) != 0)
894 if ((error = zfs_secpolicy_write_perms(parentname,
895 ZFS_DELEG_PERM_CREATE, cr)) != 0)
898 if ((error = zfs_secpolicy_write_perms(parentname,
899 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
907 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
909 return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
914 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
917 dsl_dataset_t *clone;
920 error = zfs_secpolicy_write_perms(zc->zc_name,
921 ZFS_DELEG_PERM_PROMOTE, cr);
925 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
929 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
932 char parentname[ZFS_MAX_DATASET_NAME_LEN];
933 dsl_dataset_t *origin = NULL;
937 error = dsl_dataset_hold_obj(dd->dd_pool,
938 dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin);
940 dsl_dataset_rele(clone, FTAG);
941 dsl_pool_rele(dp, FTAG);
945 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
946 ZFS_DELEG_PERM_MOUNT, cr);
948 dsl_dataset_name(origin, parentname);
950 error = zfs_secpolicy_write_perms_ds(parentname, origin,
951 ZFS_DELEG_PERM_PROMOTE, cr);
953 dsl_dataset_rele(clone, FTAG);
954 dsl_dataset_rele(origin, FTAG);
956 dsl_pool_rele(dp, FTAG);
962 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
966 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
967 ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
970 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
971 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
974 return (zfs_secpolicy_write_perms(zc->zc_name,
975 ZFS_DELEG_PERM_CREATE, cr));
980 zfs_secpolicy_recv_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
982 return (zfs_secpolicy_recv(zc, innvl, cr));
986 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
988 return (zfs_secpolicy_write_perms(name,
989 ZFS_DELEG_PERM_SNAPSHOT, cr));
993 * Check for permission to create each snapshot in the nvlist.
997 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1003 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
1004 return (SET_ERROR(EINVAL));
1005 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
1006 pair = nvlist_next_nvpair(snaps, pair)) {
1007 char *name = nvpair_name(pair);
1008 char *atp = strchr(name, '@');
1011 error = SET_ERROR(EINVAL);
1015 error = zfs_secpolicy_snapshot_perms(name, cr);
1024 * Check for permission to create each snapshot in the nvlist.
1028 zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1032 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
1033 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
1034 char *name = nvpair_name(pair);
1035 char *hashp = strchr(name, '#');
1037 if (hashp == NULL) {
1038 error = SET_ERROR(EINVAL);
1042 error = zfs_secpolicy_write_perms(name,
1043 ZFS_DELEG_PERM_BOOKMARK, 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;
1514 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1515 zc->zc_iflags, &config)))
1518 if (zc->zc_nvlist_src_size != 0 && (error =
1519 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1520 zc->zc_iflags, &props))) {
1521 nvlist_free(config);
1526 nvlist_t *nvl = NULL;
1527 nvlist_t *hidden_args = NULL;
1528 uint64_t version = SPA_VERSION;
1530 (void) nvlist_lookup_uint64(props,
1531 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1532 if (!SPA_VERSION_IS_SUPPORTED(version)) {
1533 error = SET_ERROR(EINVAL);
1534 goto pool_props_bad;
1536 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1538 error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1540 nvlist_free(config);
1544 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1547 (void) nvlist_lookup_nvlist(props, ZPOOL_HIDDEN_ARGS,
1549 error = dsl_crypto_params_create_nvlist(DCP_CMD_NONE,
1550 rootprops, hidden_args, &dcp);
1552 nvlist_free(config);
1556 (void) nvlist_remove_all(props, ZPOOL_HIDDEN_ARGS);
1558 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1559 error = zfs_fill_zplprops_root(version, rootprops,
1562 goto pool_props_bad;
1565 error = spa_create(zc->zc_name, config, props, zplprops, dcp);
1568 * Set the remaining root properties
1570 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1571 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1572 (void) spa_destroy(zc->zc_name);
1575 nvlist_free(rootprops);
1576 nvlist_free(zplprops);
1577 nvlist_free(config);
1579 dsl_crypto_params_free(dcp, !!error);
1585 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1588 zfs_log_history(zc);
1589 error = spa_destroy(zc->zc_name);
1595 zfs_ioc_pool_import(zfs_cmd_t *zc)
1597 nvlist_t *config, *props = NULL;
1601 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1602 zc->zc_iflags, &config)) != 0)
1605 if (zc->zc_nvlist_src_size != 0 && (error =
1606 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1607 zc->zc_iflags, &props))) {
1608 nvlist_free(config);
1612 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1613 guid != zc->zc_guid)
1614 error = SET_ERROR(EINVAL);
1616 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1618 if (zc->zc_nvlist_dst != 0) {
1621 if ((err = put_nvlist(zc, config)) != 0)
1625 nvlist_free(config);
1632 zfs_ioc_pool_export(zfs_cmd_t *zc)
1635 boolean_t force = (boolean_t)zc->zc_cookie;
1636 boolean_t hardforce = (boolean_t)zc->zc_guid;
1638 zfs_log_history(zc);
1639 error = spa_export(zc->zc_name, NULL, force, hardforce);
1645 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1650 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1651 return (SET_ERROR(EEXIST));
1653 error = put_nvlist(zc, configs);
1655 nvlist_free(configs);
1662 * zc_name name of the pool
1665 * zc_cookie real errno
1666 * zc_nvlist_dst config nvlist
1667 * zc_nvlist_dst_size size of config nvlist
1670 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1676 error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1677 sizeof (zc->zc_value));
1679 if (config != NULL) {
1680 ret = put_nvlist(zc, config);
1681 nvlist_free(config);
1684 * The config may be present even if 'error' is non-zero.
1685 * In this case we return success, and preserve the real errno
1688 zc->zc_cookie = error;
1697 * Try to import the given pool, returning pool stats as appropriate so that
1698 * user land knows which devices are available and overall pool health.
1701 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1703 nvlist_t *tryconfig, *config = NULL;
1706 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1707 zc->zc_iflags, &tryconfig)) != 0)
1710 config = spa_tryimport(tryconfig);
1712 nvlist_free(tryconfig);
1715 return (SET_ERROR(EINVAL));
1717 error = put_nvlist(zc, config);
1718 nvlist_free(config);
1725 * zc_name name of the pool
1726 * zc_cookie scan func (pool_scan_func_t)
1727 * zc_flags scrub pause/resume flag (pool_scrub_cmd_t)
1730 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1735 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1738 if (zc->zc_flags >= POOL_SCRUB_FLAGS_END)
1739 return (SET_ERROR(EINVAL));
1741 if (zc->zc_flags == POOL_SCRUB_PAUSE)
1742 error = spa_scrub_pause_resume(spa, POOL_SCRUB_PAUSE);
1743 else if (zc->zc_cookie == POOL_SCAN_NONE)
1744 error = spa_scan_stop(spa);
1746 error = spa_scan(spa, zc->zc_cookie);
1748 spa_close(spa, FTAG);
1754 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1759 error = spa_open(zc->zc_name, &spa, FTAG);
1762 spa_close(spa, FTAG);
1768 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1773 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1776 if (zc->zc_cookie < spa_version(spa) ||
1777 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1778 spa_close(spa, FTAG);
1779 return (SET_ERROR(EINVAL));
1782 spa_upgrade(spa, zc->zc_cookie);
1783 spa_close(spa, FTAG);
1789 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1796 if ((size = zc->zc_history_len) == 0)
1797 return (SET_ERROR(EINVAL));
1799 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1802 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1803 spa_close(spa, FTAG);
1804 return (SET_ERROR(ENOTSUP));
1807 hist_buf = vmem_alloc(size, KM_SLEEP);
1808 if ((error = spa_history_get(spa, &zc->zc_history_offset,
1809 &zc->zc_history_len, hist_buf)) == 0) {
1810 error = ddi_copyout(hist_buf,
1811 (void *)(uintptr_t)zc->zc_history,
1812 zc->zc_history_len, zc->zc_iflags);
1815 spa_close(spa, FTAG);
1816 vmem_free(hist_buf, size);
1821 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1826 error = spa_open(zc->zc_name, &spa, FTAG);
1828 error = spa_change_guid(spa);
1829 spa_close(spa, FTAG);
1835 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1837 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1842 * zc_name name of filesystem
1843 * zc_obj object to find
1846 * zc_value name of object
1849 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1854 /* XXX reading from objset not owned */
1855 if ((error = dmu_objset_hold_flags(zc->zc_name, B_TRUE,
1858 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1859 dmu_objset_rele_flags(os, B_TRUE, FTAG);
1860 return (SET_ERROR(EINVAL));
1862 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1863 sizeof (zc->zc_value));
1864 dmu_objset_rele_flags(os, B_TRUE, FTAG);
1871 * zc_name name of filesystem
1872 * zc_obj object to find
1875 * zc_stat stats on object
1876 * zc_value path to object
1879 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1884 /* XXX reading from objset not owned */
1885 if ((error = dmu_objset_hold_flags(zc->zc_name, B_TRUE,
1888 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1889 dmu_objset_rele_flags(os, B_TRUE, FTAG);
1890 return (SET_ERROR(EINVAL));
1892 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1893 sizeof (zc->zc_value));
1894 dmu_objset_rele_flags(os, B_TRUE, FTAG);
1900 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1906 error = spa_open(zc->zc_name, &spa, FTAG);
1910 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1911 zc->zc_iflags, &config);
1913 error = spa_vdev_add(spa, config);
1914 nvlist_free(config);
1916 spa_close(spa, FTAG);
1922 * zc_name name of the pool
1923 * zc_nvlist_conf nvlist of devices to remove
1924 * zc_cookie to stop the remove?
1927 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1932 error = spa_open(zc->zc_name, &spa, FTAG);
1935 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1936 spa_close(spa, FTAG);
1941 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1945 vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1947 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1949 switch (zc->zc_cookie) {
1950 case VDEV_STATE_ONLINE:
1951 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1954 case VDEV_STATE_OFFLINE:
1955 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1958 case VDEV_STATE_FAULTED:
1959 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1960 zc->zc_obj != VDEV_AUX_EXTERNAL &&
1961 zc->zc_obj != VDEV_AUX_EXTERNAL_PERSIST)
1962 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1964 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1967 case VDEV_STATE_DEGRADED:
1968 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1969 zc->zc_obj != VDEV_AUX_EXTERNAL)
1970 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1972 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1976 error = SET_ERROR(EINVAL);
1978 zc->zc_cookie = newstate;
1979 spa_close(spa, FTAG);
1984 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1987 int replacing = zc->zc_cookie;
1991 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1994 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1995 zc->zc_iflags, &config)) == 0) {
1996 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1997 nvlist_free(config);
2000 spa_close(spa, FTAG);
2005 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
2010 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2013 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
2015 spa_close(spa, FTAG);
2020 zfs_ioc_vdev_split(zfs_cmd_t *zc)
2023 nvlist_t *config, *props = NULL;
2025 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
2027 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2030 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
2031 zc->zc_iflags, &config))) {
2032 spa_close(spa, FTAG);
2036 if (zc->zc_nvlist_src_size != 0 && (error =
2037 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2038 zc->zc_iflags, &props))) {
2039 spa_close(spa, FTAG);
2040 nvlist_free(config);
2044 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
2046 spa_close(spa, FTAG);
2048 nvlist_free(config);
2055 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
2058 char *path = zc->zc_value;
2059 uint64_t guid = zc->zc_guid;
2062 error = spa_open(zc->zc_name, &spa, FTAG);
2066 error = spa_vdev_setpath(spa, guid, path);
2067 spa_close(spa, FTAG);
2072 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
2075 char *fru = zc->zc_value;
2076 uint64_t guid = zc->zc_guid;
2079 error = spa_open(zc->zc_name, &spa, FTAG);
2083 error = spa_vdev_setfru(spa, guid, fru);
2084 spa_close(spa, FTAG);
2089 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
2094 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2096 if (zc->zc_nvlist_dst != 0 &&
2097 (error = dsl_prop_get_all(os, &nv)) == 0) {
2098 dmu_objset_stats(os, nv);
2100 * NB: zvol_get_stats() will read the objset contents,
2101 * which we aren't supposed to do with a
2102 * DS_MODE_USER hold, because it could be
2103 * inconsistent. So this is a bit of a workaround...
2104 * XXX reading with out owning
2106 if (!zc->zc_objset_stats.dds_inconsistent &&
2107 dmu_objset_type(os) == DMU_OST_ZVOL) {
2108 error = zvol_get_stats(os, nv);
2116 error = put_nvlist(zc, nv);
2125 * zc_name name of filesystem
2126 * zc_nvlist_dst_size size of buffer for property nvlist
2129 * zc_objset_stats stats
2130 * zc_nvlist_dst property nvlist
2131 * zc_nvlist_dst_size size of property nvlist
2134 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2139 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2141 error = zfs_ioc_objset_stats_impl(zc, os);
2142 dmu_objset_rele(os, FTAG);
2150 * zc_name name of filesystem
2151 * zc_nvlist_dst_size size of buffer for property nvlist
2154 * zc_nvlist_dst received property nvlist
2155 * zc_nvlist_dst_size size of received property nvlist
2157 * Gets received properties (distinct from local properties on or after
2158 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2159 * local property values.
2162 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2168 * Without this check, we would return local property values if the
2169 * caller has not already received properties on or after
2170 * SPA_VERSION_RECVD_PROPS.
2172 if (!dsl_prop_get_hasrecvd(zc->zc_name))
2173 return (SET_ERROR(ENOTSUP));
2175 if (zc->zc_nvlist_dst != 0 &&
2176 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2177 error = put_nvlist(zc, nv);
2185 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2191 * zfs_get_zplprop() will either find a value or give us
2192 * the default value (if there is one).
2194 if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2196 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2202 * zc_name name of filesystem
2203 * zc_nvlist_dst_size size of buffer for zpl property nvlist
2206 * zc_nvlist_dst zpl property nvlist
2207 * zc_nvlist_dst_size size of zpl property nvlist
2210 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2215 /* XXX reading without owning */
2216 if ((err = dmu_objset_hold(zc->zc_name, FTAG, &os)))
2219 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2222 * NB: nvl_add_zplprop() will read the objset contents,
2223 * which we aren't supposed to do with a DS_MODE_USER
2224 * hold, because it could be inconsistent.
2226 if (zc->zc_nvlist_dst != 0 &&
2227 !zc->zc_objset_stats.dds_inconsistent &&
2228 dmu_objset_type(os) == DMU_OST_ZFS) {
2231 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2232 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2233 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2234 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2235 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2236 err = put_nvlist(zc, nv);
2239 err = SET_ERROR(ENOENT);
2241 dmu_objset_rele(os, FTAG);
2246 dataset_name_hidden(const char *name)
2249 * Skip over datasets that are not visible in this zone,
2250 * internal datasets (which have a $ in their name), and
2251 * temporary datasets (which have a % in their name).
2253 if (strchr(name, '$') != NULL)
2255 if (strchr(name, '%') != NULL)
2257 if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
2264 * zc_name name of filesystem
2265 * zc_cookie zap cursor
2266 * zc_nvlist_dst_size size of buffer for property nvlist
2269 * zc_name name of next filesystem
2270 * zc_cookie zap cursor
2271 * zc_objset_stats stats
2272 * zc_nvlist_dst property nvlist
2273 * zc_nvlist_dst_size size of property nvlist
2276 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2281 size_t orig_len = strlen(zc->zc_name);
2284 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os))) {
2285 if (error == ENOENT)
2286 error = SET_ERROR(ESRCH);
2290 p = strrchr(zc->zc_name, '/');
2291 if (p == NULL || p[1] != '\0')
2292 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2293 p = zc->zc_name + strlen(zc->zc_name);
2296 error = dmu_dir_list_next(os,
2297 sizeof (zc->zc_name) - (p - zc->zc_name), p,
2298 NULL, &zc->zc_cookie);
2299 if (error == ENOENT)
2300 error = SET_ERROR(ESRCH);
2301 } while (error == 0 && dataset_name_hidden(zc->zc_name));
2302 dmu_objset_rele(os, FTAG);
2305 * If it's an internal dataset (ie. with a '$' in its name),
2306 * don't try to get stats for it, otherwise we'll return ENOENT.
2308 if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2309 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2310 if (error == ENOENT) {
2311 /* We lost a race with destroy, get the next one. */
2312 zc->zc_name[orig_len] = '\0';
2321 * zc_name name of filesystem
2322 * zc_cookie zap cursor
2323 * zc_nvlist_dst_size size of buffer for property nvlist
2326 * zc_name name of next snapshot
2327 * zc_objset_stats stats
2328 * zc_nvlist_dst property nvlist
2329 * zc_nvlist_dst_size size of property nvlist
2332 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2337 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2339 return (error == ENOENT ? ESRCH : error);
2343 * A dataset name of maximum length cannot have any snapshots,
2344 * so exit immediately.
2346 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >=
2347 ZFS_MAX_DATASET_NAME_LEN) {
2348 dmu_objset_rele(os, FTAG);
2349 return (SET_ERROR(ESRCH));
2352 error = dmu_snapshot_list_next(os,
2353 sizeof (zc->zc_name) - strlen(zc->zc_name),
2354 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2357 if (error == 0 && !zc->zc_simple) {
2359 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2361 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2365 error = dmu_objset_from_ds(ds, &ossnap);
2367 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2368 dsl_dataset_rele(ds, FTAG);
2370 } else if (error == ENOENT) {
2371 error = SET_ERROR(ESRCH);
2374 dmu_objset_rele(os, FTAG);
2375 /* if we failed, undo the @ that we tacked on to zc_name */
2377 *strchr(zc->zc_name, '@') = '\0';
2382 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2384 const char *propname = nvpair_name(pair);
2386 unsigned int vallen;
2389 zfs_userquota_prop_t type;
2395 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2397 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2398 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2400 return (SET_ERROR(EINVAL));
2404 * A correctly constructed propname is encoded as
2405 * userquota@<rid>-<domain>.
2407 if ((dash = strchr(propname, '-')) == NULL ||
2408 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2410 return (SET_ERROR(EINVAL));
2417 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2419 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2420 zfsvfs_rele(zfsvfs, FTAG);
2427 * If the named property is one that has a special function to set its value,
2428 * return 0 on success and a positive error code on failure; otherwise if it is
2429 * not one of the special properties handled by this function, return -1.
2431 * XXX: It would be better for callers of the property interface if we handled
2432 * these special cases in dsl_prop.c (in the dsl layer).
2435 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2438 const char *propname = nvpair_name(pair);
2439 zfs_prop_t prop = zfs_name_to_prop(propname);
2440 uint64_t intval = 0;
2441 char *strval = NULL;
2444 if (prop == ZPROP_INVAL) {
2445 if (zfs_prop_userquota(propname))
2446 return (zfs_prop_set_userquota(dsname, pair));
2450 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2452 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2453 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2457 /* all special properties are numeric except for keylocation */
2458 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING) {
2459 strval = fnvpair_value_string(pair);
2461 intval = fnvpair_value_uint64(pair);
2465 case ZFS_PROP_QUOTA:
2466 err = dsl_dir_set_quota(dsname, source, intval);
2468 case ZFS_PROP_REFQUOTA:
2469 err = dsl_dataset_set_refquota(dsname, source, intval);
2471 case ZFS_PROP_FILESYSTEM_LIMIT:
2472 case ZFS_PROP_SNAPSHOT_LIMIT:
2473 if (intval == UINT64_MAX) {
2474 /* clearing the limit, just do it */
2477 err = dsl_dir_activate_fs_ss_limit(dsname);
2480 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2481 * default path to set the value in the nvlist.
2486 case ZFS_PROP_KEYLOCATION:
2487 err = dsl_crypto_can_set_keylocation(dsname, strval);
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_RESERVATION:
2497 err = dsl_dir_set_reservation(dsname, source, intval);
2499 case ZFS_PROP_REFRESERVATION:
2500 err = dsl_dataset_set_refreservation(dsname, source, intval);
2502 case ZFS_PROP_VOLSIZE:
2503 err = zvol_set_volsize(dsname, intval);
2505 case ZFS_PROP_SNAPDEV:
2506 err = zvol_set_snapdev(dsname, source, intval);
2508 case ZFS_PROP_VOLMODE:
2509 err = zvol_set_volmode(dsname, source, intval);
2511 case ZFS_PROP_VERSION:
2515 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2518 err = zfs_set_version(zfsvfs, intval);
2519 zfsvfs_rele(zfsvfs, FTAG);
2521 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2524 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2525 (void) strcpy(zc->zc_name, dsname);
2526 (void) zfs_ioc_userspace_upgrade(zc);
2527 (void) zfs_ioc_id_quota_upgrade(zc);
2528 kmem_free(zc, sizeof (zfs_cmd_t));
2540 * This function is best effort. If it fails to set any of the given properties,
2541 * it continues to set as many as it can and returns the last error
2542 * encountered. If the caller provides a non-NULL errlist, it will be filled in
2543 * with the list of names of all the properties that failed along with the
2544 * corresponding error numbers.
2546 * If every property is set successfully, zero is returned and errlist is not
2550 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2559 nvlist_t *genericnvl = fnvlist_alloc();
2560 nvlist_t *retrynvl = fnvlist_alloc();
2563 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2564 const char *propname = nvpair_name(pair);
2565 zfs_prop_t prop = zfs_name_to_prop(propname);
2568 /* decode the property value */
2570 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2572 attrs = fnvpair_value_nvlist(pair);
2573 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2575 err = SET_ERROR(EINVAL);
2578 /* Validate value type */
2579 if (err == 0 && source == ZPROP_SRC_INHERITED) {
2580 /* inherited properties are expected to be booleans */
2581 if (nvpair_type(propval) != DATA_TYPE_BOOLEAN)
2582 err = SET_ERROR(EINVAL);
2583 } else if (err == 0 && prop == ZPROP_INVAL) {
2584 if (zfs_prop_user(propname)) {
2585 if (nvpair_type(propval) != DATA_TYPE_STRING)
2586 err = SET_ERROR(EINVAL);
2587 } else if (zfs_prop_userquota(propname)) {
2588 if (nvpair_type(propval) !=
2589 DATA_TYPE_UINT64_ARRAY)
2590 err = SET_ERROR(EINVAL);
2592 err = SET_ERROR(EINVAL);
2594 } else if (err == 0) {
2595 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2596 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2597 err = SET_ERROR(EINVAL);
2598 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2601 intval = fnvpair_value_uint64(propval);
2603 switch (zfs_prop_get_type(prop)) {
2604 case PROP_TYPE_NUMBER:
2606 case PROP_TYPE_STRING:
2607 err = SET_ERROR(EINVAL);
2609 case PROP_TYPE_INDEX:
2610 if (zfs_prop_index_to_string(prop,
2611 intval, &unused) != 0)
2612 err = SET_ERROR(EINVAL);
2616 "unknown property type");
2619 err = SET_ERROR(EINVAL);
2623 /* Validate permissions */
2625 err = zfs_check_settable(dsname, pair, CRED());
2628 if (source == ZPROP_SRC_INHERITED)
2629 err = -1; /* does not need special handling */
2631 err = zfs_prop_set_special(dsname, source,
2635 * For better performance we build up a list of
2636 * properties to set in a single transaction.
2638 err = nvlist_add_nvpair(genericnvl, pair);
2639 } else if (err != 0 && nvl != retrynvl) {
2641 * This may be a spurious error caused by
2642 * receiving quota and reservation out of order.
2643 * Try again in a second pass.
2645 err = nvlist_add_nvpair(retrynvl, pair);
2650 if (errlist != NULL)
2651 fnvlist_add_int32(errlist, propname, err);
2656 if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2661 if (!nvlist_empty(genericnvl) &&
2662 dsl_props_set(dsname, source, genericnvl) != 0) {
2664 * If this fails, we still want to set as many properties as we
2665 * can, so try setting them individually.
2668 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2669 const char *propname = nvpair_name(pair);
2673 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2675 attrs = fnvpair_value_nvlist(pair);
2676 propval = fnvlist_lookup_nvpair(attrs,
2680 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2681 strval = fnvpair_value_string(propval);
2682 err = dsl_prop_set_string(dsname, propname,
2684 } else if (nvpair_type(propval) == DATA_TYPE_BOOLEAN) {
2685 err = dsl_prop_inherit(dsname, propname,
2688 intval = fnvpair_value_uint64(propval);
2689 err = dsl_prop_set_int(dsname, propname, source,
2694 if (errlist != NULL) {
2695 fnvlist_add_int32(errlist, propname,
2702 nvlist_free(genericnvl);
2703 nvlist_free(retrynvl);
2709 * Check that all the properties are valid user properties.
2712 zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2714 nvpair_t *pair = NULL;
2717 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2718 const char *propname = nvpair_name(pair);
2720 if (!zfs_prop_user(propname) ||
2721 nvpair_type(pair) != DATA_TYPE_STRING)
2722 return (SET_ERROR(EINVAL));
2724 if ((error = zfs_secpolicy_write_perms(fsname,
2725 ZFS_DELEG_PERM_USERPROP, CRED())))
2728 if (strlen(propname) >= ZAP_MAXNAMELEN)
2729 return (SET_ERROR(ENAMETOOLONG));
2731 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN)
2732 return (SET_ERROR(E2BIG));
2738 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2742 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2745 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2746 if (nvlist_exists(skipped, nvpair_name(pair)))
2749 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2754 clear_received_props(const char *dsname, nvlist_t *props,
2758 nvlist_t *cleared_props = NULL;
2759 props_skip(props, skipped, &cleared_props);
2760 if (!nvlist_empty(cleared_props)) {
2762 * Acts on local properties until the dataset has received
2763 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2765 zprop_source_t flags = (ZPROP_SRC_NONE |
2766 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2767 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2769 nvlist_free(cleared_props);
2775 * zc_name name of filesystem
2776 * zc_value name of property to set
2777 * zc_nvlist_src{_size} nvlist of properties to apply
2778 * zc_cookie received properties flag
2781 * zc_nvlist_dst{_size} error for each unapplied received property
2784 zfs_ioc_set_prop(zfs_cmd_t *zc)
2787 boolean_t received = zc->zc_cookie;
2788 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2793 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2794 zc->zc_iflags, &nvl)) != 0)
2798 nvlist_t *origprops;
2800 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2801 (void) clear_received_props(zc->zc_name,
2803 nvlist_free(origprops);
2806 error = dsl_prop_set_hasrecvd(zc->zc_name);
2809 errors = fnvlist_alloc();
2811 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2813 if (zc->zc_nvlist_dst != 0 && errors != NULL) {
2814 (void) put_nvlist(zc, errors);
2817 nvlist_free(errors);
2824 * zc_name name of filesystem
2825 * zc_value name of property to inherit
2826 * zc_cookie revert to received value if TRUE
2831 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2833 const char *propname = zc->zc_value;
2834 zfs_prop_t prop = zfs_name_to_prop(propname);
2835 boolean_t received = zc->zc_cookie;
2836 zprop_source_t source = (received
2837 ? ZPROP_SRC_NONE /* revert to received value, if any */
2838 : ZPROP_SRC_INHERITED); /* explicitly inherit */
2846 * Only check this in the non-received case. We want to allow
2847 * 'inherit -S' to revert non-inheritable properties like quota
2848 * and reservation to the received or default values even though
2849 * they are not considered inheritable.
2851 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2852 return (SET_ERROR(EINVAL));
2855 if (prop == ZPROP_INVAL) {
2856 if (!zfs_prop_user(propname))
2857 return (SET_ERROR(EINVAL));
2859 type = PROP_TYPE_STRING;
2860 } else if (prop == ZFS_PROP_VOLSIZE || prop == ZFS_PROP_VERSION) {
2861 return (SET_ERROR(EINVAL));
2863 type = zfs_prop_get_type(prop);
2867 * zfs_prop_set_special() expects properties in the form of an
2868 * nvpair with type info.
2870 dummy = fnvlist_alloc();
2873 case PROP_TYPE_STRING:
2874 VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2876 case PROP_TYPE_NUMBER:
2877 case PROP_TYPE_INDEX:
2878 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2881 err = SET_ERROR(EINVAL);
2885 pair = nvlist_next_nvpair(dummy, NULL);
2887 err = SET_ERROR(EINVAL);
2889 err = zfs_prop_set_special(zc->zc_name, source, pair);
2890 if (err == -1) /* property is not "special", needs handling */
2891 err = dsl_prop_inherit(zc->zc_name, zc->zc_value,
2901 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2908 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2909 zc->zc_iflags, &props)))
2913 * If the only property is the configfile, then just do a spa_lookup()
2914 * to handle the faulted case.
2916 pair = nvlist_next_nvpair(props, NULL);
2917 if (pair != NULL && strcmp(nvpair_name(pair),
2918 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2919 nvlist_next_nvpair(props, pair) == NULL) {
2920 mutex_enter(&spa_namespace_lock);
2921 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2922 spa_configfile_set(spa, props, B_FALSE);
2923 spa_config_sync(spa, B_FALSE, B_TRUE);
2925 mutex_exit(&spa_namespace_lock);
2932 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2937 error = spa_prop_set(spa, props);
2940 spa_close(spa, FTAG);
2946 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2950 nvlist_t *nvp = NULL;
2952 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2954 * If the pool is faulted, there may be properties we can still
2955 * get (such as altroot and cachefile), so attempt to get them
2958 mutex_enter(&spa_namespace_lock);
2959 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2960 error = spa_prop_get(spa, &nvp);
2961 mutex_exit(&spa_namespace_lock);
2963 error = spa_prop_get(spa, &nvp);
2964 spa_close(spa, FTAG);
2967 if (error == 0 && zc->zc_nvlist_dst != 0)
2968 error = put_nvlist(zc, nvp);
2970 error = SET_ERROR(EFAULT);
2978 * zc_name name of filesystem
2979 * zc_nvlist_src{_size} nvlist of delegated permissions
2980 * zc_perm_action allow/unallow flag
2985 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2988 nvlist_t *fsaclnv = NULL;
2990 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2991 zc->zc_iflags, &fsaclnv)) != 0)
2995 * Verify nvlist is constructed correctly
2997 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2998 nvlist_free(fsaclnv);
2999 return (SET_ERROR(EINVAL));
3003 * If we don't have PRIV_SYS_MOUNT, then validate
3004 * that user is allowed to hand out each permission in
3008 error = secpolicy_zfs(CRED());
3010 if (zc->zc_perm_action == B_FALSE) {
3011 error = dsl_deleg_can_allow(zc->zc_name,
3014 error = dsl_deleg_can_unallow(zc->zc_name,
3020 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
3022 nvlist_free(fsaclnv);
3028 * zc_name name of filesystem
3031 * zc_nvlist_src{_size} nvlist of delegated permissions
3034 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
3039 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
3040 error = put_nvlist(zc, nvp);
3049 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
3051 zfs_creat_t *zct = arg;
3053 zfs_create_fs(os, cr, zct->zct_zplprops, tx);
3056 #define ZFS_PROP_UNDEFINED ((uint64_t)-1)
3060 * os parent objset pointer (NULL if root fs)
3061 * fuids_ok fuids allowed in this version of the spa?
3062 * sa_ok SAs allowed in this version of the spa?
3063 * createprops list of properties requested by creator
3066 * zplprops values for the zplprops we attach to the master node object
3067 * is_ci true if requested file system will be purely case-insensitive
3069 * Determine the settings for utf8only, normalization and
3070 * casesensitivity. Specific values may have been requested by the
3071 * creator and/or we can inherit values from the parent dataset. If
3072 * the file system is of too early a vintage, a creator can not
3073 * request settings for these properties, even if the requested
3074 * setting is the default value. We don't actually want to create dsl
3075 * properties for these, so remove them from the source nvlist after
3079 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
3080 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
3081 nvlist_t *zplprops, boolean_t *is_ci)
3083 uint64_t sense = ZFS_PROP_UNDEFINED;
3084 uint64_t norm = ZFS_PROP_UNDEFINED;
3085 uint64_t u8 = ZFS_PROP_UNDEFINED;
3088 ASSERT(zplprops != NULL);
3090 if (os != NULL && os->os_phys->os_type != DMU_OST_ZFS)
3091 return (SET_ERROR(EINVAL));
3094 * Pull out creator prop choices, if any.
3097 (void) nvlist_lookup_uint64(createprops,
3098 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
3099 (void) nvlist_lookup_uint64(createprops,
3100 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
3101 (void) nvlist_remove_all(createprops,
3102 zfs_prop_to_name(ZFS_PROP_NORMALIZE));
3103 (void) nvlist_lookup_uint64(createprops,
3104 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
3105 (void) nvlist_remove_all(createprops,
3106 zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
3107 (void) nvlist_lookup_uint64(createprops,
3108 zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3109 (void) nvlist_remove_all(createprops,
3110 zfs_prop_to_name(ZFS_PROP_CASE));
3114 * If the zpl version requested is whacky or the file system
3115 * or pool is version is too "young" to support normalization
3116 * and the creator tried to set a value for one of the props,
3119 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3120 (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3121 (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3122 (zplver < ZPL_VERSION_NORMALIZATION &&
3123 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3124 sense != ZFS_PROP_UNDEFINED)))
3125 return (SET_ERROR(ENOTSUP));
3128 * Put the version in the zplprops
3130 VERIFY(nvlist_add_uint64(zplprops,
3131 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3133 if (norm == ZFS_PROP_UNDEFINED &&
3134 (error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm)) != 0)
3136 VERIFY(nvlist_add_uint64(zplprops,
3137 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3140 * If we're normalizing, names must always be valid UTF-8 strings.
3144 if (u8 == ZFS_PROP_UNDEFINED &&
3145 (error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8)) != 0)
3147 VERIFY(nvlist_add_uint64(zplprops,
3148 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3150 if (sense == ZFS_PROP_UNDEFINED &&
3151 (error = zfs_get_zplprop(os, ZFS_PROP_CASE, &sense)) != 0)
3153 VERIFY(nvlist_add_uint64(zplprops,
3154 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3157 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
3163 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3164 nvlist_t *zplprops, boolean_t *is_ci)
3166 boolean_t fuids_ok, sa_ok;
3167 uint64_t zplver = ZPL_VERSION;
3168 objset_t *os = NULL;
3169 char parentname[ZFS_MAX_DATASET_NAME_LEN];
3175 (void) strlcpy(parentname, dataset, sizeof (parentname));
3176 cp = strrchr(parentname, '/');
3180 if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3183 spa_vers = spa_version(spa);
3184 spa_close(spa, FTAG);
3186 zplver = zfs_zpl_version_map(spa_vers);
3187 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3188 sa_ok = (zplver >= ZPL_VERSION_SA);
3191 * Open parent object set so we can inherit zplprop values.
3193 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3196 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3198 dmu_objset_rele(os, FTAG);
3203 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3204 nvlist_t *zplprops, boolean_t *is_ci)
3208 uint64_t zplver = ZPL_VERSION;
3211 zplver = zfs_zpl_version_map(spa_vers);
3212 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3213 sa_ok = (zplver >= ZPL_VERSION_SA);
3215 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3216 createprops, zplprops, is_ci);
3222 * "type" -> dmu_objset_type_t (int32)
3223 * (optional) "props" -> { prop -> value }
3224 * (optional) "hidden_args" -> { "wkeydata" -> value }
3225 * raw uint8_t array of encryption wrapping key data (32 bytes)
3228 * outnvl: propname -> error code (int32)
3231 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3234 zfs_creat_t zct = { 0 };
3235 nvlist_t *nvprops = NULL;
3236 nvlist_t *hidden_args = NULL;
3237 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3239 dmu_objset_type_t type;
3240 boolean_t is_insensitive = B_FALSE;
3241 dsl_crypto_params_t *dcp = NULL;
3243 if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3244 return (SET_ERROR(EINVAL));
3246 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3247 (void) nvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS, &hidden_args);
3251 cbfunc = zfs_create_cb;
3255 cbfunc = zvol_create_cb;
3262 if (strchr(fsname, '@') ||
3263 strchr(fsname, '%'))
3264 return (SET_ERROR(EINVAL));
3266 zct.zct_props = nvprops;
3269 return (SET_ERROR(EINVAL));
3271 if (type == DMU_OST_ZVOL) {
3272 uint64_t volsize, volblocksize;
3274 if (nvprops == NULL)
3275 return (SET_ERROR(EINVAL));
3276 if (nvlist_lookup_uint64(nvprops,
3277 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3278 return (SET_ERROR(EINVAL));
3280 if ((error = nvlist_lookup_uint64(nvprops,
3281 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3282 &volblocksize)) != 0 && error != ENOENT)
3283 return (SET_ERROR(EINVAL));
3286 volblocksize = zfs_prop_default_numeric(
3287 ZFS_PROP_VOLBLOCKSIZE);
3289 if ((error = zvol_check_volblocksize(fsname,
3290 volblocksize)) != 0 ||
3291 (error = zvol_check_volsize(volsize,
3292 volblocksize)) != 0)
3294 } else if (type == DMU_OST_ZFS) {
3298 * We have to have normalization and
3299 * case-folding flags correct when we do the
3300 * file system creation, so go figure them out
3303 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3304 NV_UNIQUE_NAME, KM_SLEEP) == 0);
3305 error = zfs_fill_zplprops(fsname, nvprops,
3306 zct.zct_zplprops, &is_insensitive);
3308 nvlist_free(zct.zct_zplprops);
3313 error = dsl_crypto_params_create_nvlist(DCP_CMD_NONE, nvprops,
3316 nvlist_free(zct.zct_zplprops);
3320 error = dmu_objset_create(fsname, type,
3321 is_insensitive ? DS_FLAG_CI_DATASET : 0, dcp, cbfunc, &zct);
3323 nvlist_free(zct.zct_zplprops);
3324 dsl_crypto_params_free(dcp, !!error);
3327 * It would be nice to do this atomically.
3330 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3337 * Volumes will return EBUSY and cannot be destroyed
3338 * until all asynchronous minor handling has completed.
3339 * Wait for the spa_zvol_taskq to drain then retry.
3341 error2 = dsl_destroy_head(fsname);
3342 while ((error2 == EBUSY) && (type == DMU_OST_ZVOL)) {
3343 error2 = spa_open(fsname, &spa, FTAG);
3345 taskq_wait(spa->spa_zvol_taskq);
3346 spa_close(spa, FTAG);
3348 error2 = dsl_destroy_head(fsname);
3357 * "origin" -> name of origin snapshot
3358 * (optional) "props" -> { prop -> value }
3359 * (optional) "hidden_args" -> { "wkeydata" -> value }
3360 * raw uint8_t array of encryption wrapping key data (32 bytes)
3364 * outnvl: propname -> error code (int32)
3367 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3370 nvlist_t *nvprops = NULL;
3373 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3374 return (SET_ERROR(EINVAL));
3375 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3377 if (strchr(fsname, '@') ||
3378 strchr(fsname, '%'))
3379 return (SET_ERROR(EINVAL));
3381 if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3382 return (SET_ERROR(EINVAL));
3384 error = dmu_objset_clone(fsname, origin_name);
3387 * It would be nice to do this atomically.
3390 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3393 (void) dsl_destroy_head(fsname);
3400 * "snaps" -> { snapshot1, snapshot2 }
3401 * (optional) "props" -> { prop -> value (string) }
3404 * outnvl: snapshot -> error code (int32)
3407 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3410 nvlist_t *props = NULL;
3414 (void) nvlist_lookup_nvlist(innvl, "props", &props);
3415 if ((error = zfs_check_userprops(poolname, props)) != 0)
3418 if (!nvlist_empty(props) &&
3419 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3420 return (SET_ERROR(ENOTSUP));
3422 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3423 return (SET_ERROR(EINVAL));
3424 poollen = strlen(poolname);
3425 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3426 pair = nvlist_next_nvpair(snaps, pair)) {
3427 const char *name = nvpair_name(pair);
3428 const char *cp = strchr(name, '@');
3431 * The snap name must contain an @, and the part after it must
3432 * contain only valid characters.
3435 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3436 return (SET_ERROR(EINVAL));
3439 * The snap must be in the specified pool.
3441 if (strncmp(name, poolname, poollen) != 0 ||
3442 (name[poollen] != '/' && name[poollen] != '@'))
3443 return (SET_ERROR(EXDEV));
3445 /* This must be the only snap of this fs. */
3446 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3447 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3448 if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3450 return (SET_ERROR(EXDEV));
3455 error = dsl_dataset_snapshot(snaps, props, outnvl);
3461 * innvl: "message" -> string
3465 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3473 * The poolname in the ioctl is not set, we get it from the TSD,
3474 * which was set at the end of the last successful ioctl that allows
3475 * logging. The secpolicy func already checked that it is set.
3476 * Only one log ioctl is allowed after each successful ioctl, so
3477 * we clear the TSD here.
3479 poolname = tsd_get(zfs_allow_log_key);
3480 if (poolname == NULL)
3481 return (SET_ERROR(EINVAL));
3482 (void) tsd_set(zfs_allow_log_key, NULL);
3483 error = spa_open(poolname, &spa, FTAG);
3488 if (nvlist_lookup_string(innvl, "message", &message) != 0) {
3489 spa_close(spa, FTAG);
3490 return (SET_ERROR(EINVAL));
3493 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3494 spa_close(spa, FTAG);
3495 return (SET_ERROR(ENOTSUP));
3498 error = spa_history_log(spa, message);
3499 spa_close(spa, FTAG);
3504 * The dp_config_rwlock must not be held when calling this, because the
3505 * unmount may need to write out data.
3507 * This function is best-effort. Callers must deal gracefully if it
3508 * remains mounted (or is remounted after this call).
3510 * Returns 0 if the argument is not a snapshot, or it is not currently a
3511 * filesystem, or we were able to unmount it. Returns error code otherwise.
3514 zfs_unmount_snap(const char *snapname)
3516 if (strchr(snapname, '@') == NULL)
3519 (void) zfsctl_snapshot_unmount((char *)snapname, MNT_FORCE);
3524 zfs_unmount_snap_cb(const char *snapname, void *arg)
3526 zfs_unmount_snap(snapname);
3531 * When a clone is destroyed, its origin may also need to be destroyed,
3532 * in which case it must be unmounted. This routine will do that unmount
3536 zfs_destroy_unmount_origin(const char *fsname)
3542 error = dmu_objset_hold(fsname, FTAG, &os);
3545 ds = dmu_objset_ds(os);
3546 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3547 char originname[ZFS_MAX_DATASET_NAME_LEN];
3548 dsl_dataset_name(ds->ds_prev, originname);
3549 dmu_objset_rele(os, FTAG);
3550 zfs_unmount_snap(originname);
3552 dmu_objset_rele(os, FTAG);
3558 * "snaps" -> { snapshot1, snapshot2 }
3559 * (optional boolean) "defer"
3562 * outnvl: snapshot -> error code (int32)
3566 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3572 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3573 return (SET_ERROR(EINVAL));
3574 defer = nvlist_exists(innvl, "defer");
3576 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3577 pair = nvlist_next_nvpair(snaps, pair)) {
3578 zfs_unmount_snap(nvpair_name(pair));
3581 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3585 * Create bookmarks. Bookmark names are of the form <fs>#<bmark>.
3586 * All bookmarks must be in the same pool.
3589 * bookmark1 -> snapshot1, bookmark2 -> snapshot2
3592 * outnvl: bookmark -> error code (int32)
3597 zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3599 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3600 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3604 * Verify the snapshot argument.
3606 if (nvpair_value_string(pair, &snap_name) != 0)
3607 return (SET_ERROR(EINVAL));
3610 /* Verify that the keys (bookmarks) are unique */
3611 for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair);
3612 pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) {
3613 if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0)
3614 return (SET_ERROR(EINVAL));
3618 return (dsl_bookmark_create(innvl, outnvl));
3623 * property 1, property 2, ...
3627 * bookmark name 1 -> { property 1, property 2, ... },
3628 * bookmark name 2 -> { property 1, property 2, ... }
3633 zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3635 return (dsl_get_bookmarks(fsname, innvl, outnvl));
3640 * bookmark name 1, bookmark name 2
3643 * outnvl: bookmark -> error code (int32)
3647 zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl,
3652 poollen = strlen(poolname);
3653 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3654 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3655 const char *name = nvpair_name(pair);
3656 const char *cp = strchr(name, '#');
3659 * The bookmark name must contain an #, and the part after it
3660 * must contain only valid characters.
3663 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3664 return (SET_ERROR(EINVAL));
3667 * The bookmark must be in the specified pool.
3669 if (strncmp(name, poolname, poollen) != 0 ||
3670 (name[poollen] != '/' && name[poollen] != '#'))
3671 return (SET_ERROR(EXDEV));
3674 error = dsl_bookmark_destroy(innvl, outnvl);
3679 zfs_ioc_channel_program(const char *poolname, nvlist_t *innvl,
3683 uint64_t instrlimit, memlimit;
3684 boolean_t sync_flag;
3685 nvpair_t *nvarg = NULL;
3687 if (0 != nvlist_lookup_string(innvl, ZCP_ARG_PROGRAM, &program)) {
3690 if (0 != nvlist_lookup_boolean_value(innvl, ZCP_ARG_SYNC, &sync_flag)) {
3693 if (0 != nvlist_lookup_uint64(innvl, ZCP_ARG_INSTRLIMIT, &instrlimit)) {
3694 instrlimit = ZCP_DEFAULT_INSTRLIMIT;
3696 if (0 != nvlist_lookup_uint64(innvl, ZCP_ARG_MEMLIMIT, &memlimit)) {
3697 memlimit = ZCP_DEFAULT_MEMLIMIT;
3699 if (0 != nvlist_lookup_nvpair(innvl, ZCP_ARG_ARGLIST, &nvarg)) {
3703 if (instrlimit == 0 || instrlimit > zfs_lua_max_instrlimit)
3705 if (memlimit == 0 || memlimit > zfs_lua_max_memlimit)
3708 return (zcp_eval(poolname, program, sync_flag, instrlimit, memlimit,
3714 * zc_name name of dataset to destroy
3715 * zc_objset_type type of objset
3716 * zc_defer_destroy mark for deferred destroy
3721 zfs_ioc_destroy(zfs_cmd_t *zc)
3725 if (zc->zc_objset_type == DMU_OST_ZFS)
3726 zfs_unmount_snap(zc->zc_name);
3728 if (strchr(zc->zc_name, '@')) {
3729 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3731 err = dsl_destroy_head(zc->zc_name);
3732 if (err == EEXIST) {
3734 * It is possible that the given DS may have
3735 * hidden child (%recv) datasets - "leftovers"
3736 * resulting from the previously interrupted
3739 * 6 extra bytes for /%recv
3741 char namebuf[ZFS_MAX_DATASET_NAME_LEN + 6];
3743 if (snprintf(namebuf, sizeof (namebuf), "%s/%s",
3744 zc->zc_name, recv_clone_name) >=
3746 return (SET_ERROR(EINVAL));
3749 * Try to remove the hidden child (%recv) and after
3750 * that try to remove the target dataset.
3751 * If the hidden child (%recv) does not exist
3752 * the original error (EEXIST) will be returned
3754 err = dsl_destroy_head(namebuf);
3756 err = dsl_destroy_head(zc->zc_name);
3757 else if (err == ENOENT)
3758 err = SET_ERROR(EEXIST);
3766 * fsname is name of dataset to rollback (to most recent snapshot)
3768 * innvl may contain name of expected target snapshot
3770 * outnvl: "target" -> name of most recent snapshot
3775 zfs_ioc_rollback(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3779 char *target = NULL;
3782 (void) nvlist_lookup_string(innvl, "target", &target);
3783 if (target != NULL) {
3784 const char *cp = strchr(target, '@');
3787 * The snap name must contain an @, and the part after it must
3788 * contain only valid characters.
3791 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3792 return (SET_ERROR(EINVAL));
3795 if (getzfsvfs(fsname, &zfsvfs) == 0) {
3798 ds = dmu_objset_ds(zfsvfs->z_os);
3799 error = zfs_suspend_fs(zfsvfs);
3803 error = dsl_dataset_rollback(fsname, target, zfsvfs,
3805 resume_err = zfs_resume_fs(zfsvfs, ds);
3806 error = error ? error : resume_err;
3808 deactivate_super(zfsvfs->z_sb);
3809 } else if ((zv = zvol_suspend(fsname)) != NULL) {
3810 error = dsl_dataset_rollback(fsname, target, zvol_tag(zv),
3814 error = dsl_dataset_rollback(fsname, target, NULL, outnvl);
3820 recursive_unmount(const char *fsname, void *arg)
3822 const char *snapname = arg;
3825 fullname = kmem_asprintf("%s@%s", fsname, snapname);
3826 zfs_unmount_snap(fullname);
3834 * zc_name old name of dataset
3835 * zc_value new name of dataset
3836 * zc_cookie recursive flag (only valid for snapshots)
3841 zfs_ioc_rename(zfs_cmd_t *zc)
3843 boolean_t recursive = zc->zc_cookie & 1;
3846 /* "zfs rename" from and to ...%recv datasets should both fail */
3847 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
3848 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3849 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0 ||
3850 dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3851 strchr(zc->zc_name, '%') || strchr(zc->zc_value, '%'))
3852 return (SET_ERROR(EINVAL));
3854 at = strchr(zc->zc_name, '@');
3856 /* snaps must be in same fs */
3859 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
3860 return (SET_ERROR(EXDEV));
3862 if (zc->zc_objset_type == DMU_OST_ZFS) {
3863 error = dmu_objset_find(zc->zc_name,
3864 recursive_unmount, at + 1,
3865 recursive ? DS_FIND_CHILDREN : 0);
3871 error = dsl_dataset_rename_snapshot(zc->zc_name,
3872 at + 1, strchr(zc->zc_value, '@') + 1, recursive);
3877 return (dsl_dir_rename(zc->zc_name, zc->zc_value));
3882 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3884 const char *propname = nvpair_name(pair);
3885 boolean_t issnap = (strchr(dsname, '@') != NULL);
3886 zfs_prop_t prop = zfs_name_to_prop(propname);
3890 if (prop == ZPROP_INVAL) {
3891 if (zfs_prop_user(propname)) {
3892 if ((err = zfs_secpolicy_write_perms(dsname,
3893 ZFS_DELEG_PERM_USERPROP, cr)))
3898 if (!issnap && zfs_prop_userquota(propname)) {
3899 const char *perm = NULL;
3900 const char *uq_prefix =
3901 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3902 const char *gq_prefix =
3903 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3904 const char *uiq_prefix =
3905 zfs_userquota_prop_prefixes[ZFS_PROP_USEROBJQUOTA];
3906 const char *giq_prefix =
3907 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPOBJQUOTA];
3908 const char *pq_prefix =
3909 zfs_userquota_prop_prefixes[ZFS_PROP_PROJECTQUOTA];
3910 const char *piq_prefix = zfs_userquota_prop_prefixes[\
3911 ZFS_PROP_PROJECTOBJQUOTA];
3913 if (strncmp(propname, uq_prefix,
3914 strlen(uq_prefix)) == 0) {
3915 perm = ZFS_DELEG_PERM_USERQUOTA;
3916 } else if (strncmp(propname, uiq_prefix,
3917 strlen(uiq_prefix)) == 0) {
3918 perm = ZFS_DELEG_PERM_USEROBJQUOTA;
3919 } else if (strncmp(propname, gq_prefix,
3920 strlen(gq_prefix)) == 0) {
3921 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3922 } else if (strncmp(propname, giq_prefix,
3923 strlen(giq_prefix)) == 0) {
3924 perm = ZFS_DELEG_PERM_GROUPOBJQUOTA;
3925 } else if (strncmp(propname, pq_prefix,
3926 strlen(pq_prefix)) == 0) {
3927 perm = ZFS_DELEG_PERM_PROJECTQUOTA;
3928 } else if (strncmp(propname, piq_prefix,
3929 strlen(piq_prefix)) == 0) {
3930 perm = ZFS_DELEG_PERM_PROJECTOBJQUOTA;
3932 /* {USER|GROUP|PROJECT}USED are read-only */
3933 return (SET_ERROR(EINVAL));
3936 if ((err = zfs_secpolicy_write_perms(dsname, perm, cr)))
3941 return (SET_ERROR(EINVAL));
3945 return (SET_ERROR(EINVAL));
3947 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3949 * dsl_prop_get_all_impl() returns properties in this
3953 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3954 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3959 * Check that this value is valid for this pool version
3962 case ZFS_PROP_COMPRESSION:
3964 * If the user specified gzip compression, make sure
3965 * the SPA supports it. We ignore any errors here since
3966 * we'll catch them later.
3968 if (nvpair_value_uint64(pair, &intval) == 0) {
3969 if (intval >= ZIO_COMPRESS_GZIP_1 &&
3970 intval <= ZIO_COMPRESS_GZIP_9 &&
3971 zfs_earlier_version(dsname,
3972 SPA_VERSION_GZIP_COMPRESSION)) {
3973 return (SET_ERROR(ENOTSUP));
3976 if (intval == ZIO_COMPRESS_ZLE &&
3977 zfs_earlier_version(dsname,
3978 SPA_VERSION_ZLE_COMPRESSION))
3979 return (SET_ERROR(ENOTSUP));
3981 if (intval == ZIO_COMPRESS_LZ4) {
3984 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3987 if (!spa_feature_is_enabled(spa,
3988 SPA_FEATURE_LZ4_COMPRESS)) {
3989 spa_close(spa, FTAG);
3990 return (SET_ERROR(ENOTSUP));
3992 spa_close(spa, FTAG);
3996 * If this is a bootable dataset then
3997 * verify that the compression algorithm
3998 * is supported for booting. We must return
3999 * something other than ENOTSUP since it
4000 * implies a downrev pool version.
4002 if (zfs_is_bootfs(dsname) &&
4003 !BOOTFS_COMPRESS_VALID(intval)) {
4004 return (SET_ERROR(ERANGE));
4009 case ZFS_PROP_COPIES:
4010 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
4011 return (SET_ERROR(ENOTSUP));
4014 case ZFS_PROP_VOLBLOCKSIZE:
4015 case ZFS_PROP_RECORDSIZE:
4016 /* Record sizes above 128k need the feature to be enabled */
4017 if (nvpair_value_uint64(pair, &intval) == 0 &&
4018 intval > SPA_OLD_MAXBLOCKSIZE) {
4022 * We don't allow setting the property above 1MB,
4023 * unless the tunable has been changed.
4025 if (intval > zfs_max_recordsize ||
4026 intval > SPA_MAXBLOCKSIZE)
4027 return (SET_ERROR(ERANGE));
4029 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4032 if (!spa_feature_is_enabled(spa,
4033 SPA_FEATURE_LARGE_BLOCKS)) {
4034 spa_close(spa, FTAG);
4035 return (SET_ERROR(ENOTSUP));
4037 spa_close(spa, FTAG);
4041 case ZFS_PROP_DNODESIZE:
4042 /* Dnode sizes above 512 need the feature to be enabled */
4043 if (nvpair_value_uint64(pair, &intval) == 0 &&
4044 intval != ZFS_DNSIZE_LEGACY) {
4048 * If this is a bootable dataset then
4049 * we don't allow large (>512B) dnodes,
4050 * because GRUB doesn't support them.
4052 if (zfs_is_bootfs(dsname) &&
4053 intval != ZFS_DNSIZE_LEGACY) {
4054 return (SET_ERROR(EDOM));
4057 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4060 if (!spa_feature_is_enabled(spa,
4061 SPA_FEATURE_LARGE_DNODE)) {
4062 spa_close(spa, FTAG);
4063 return (SET_ERROR(ENOTSUP));
4065 spa_close(spa, FTAG);
4069 case ZFS_PROP_SHARESMB:
4070 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
4071 return (SET_ERROR(ENOTSUP));
4074 case ZFS_PROP_ACLINHERIT:
4075 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
4076 nvpair_value_uint64(pair, &intval) == 0) {
4077 if (intval == ZFS_ACL_PASSTHROUGH_X &&
4078 zfs_earlier_version(dsname,
4079 SPA_VERSION_PASSTHROUGH_X))
4080 return (SET_ERROR(ENOTSUP));
4083 case ZFS_PROP_CHECKSUM:
4084 case ZFS_PROP_DEDUP:
4086 spa_feature_t feature;
4091 /* dedup feature version checks */
4092 if (prop == ZFS_PROP_DEDUP &&
4093 zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
4094 return (SET_ERROR(ENOTSUP));
4096 if (nvpair_value_uint64(pair, &intval) != 0)
4097 return (SET_ERROR(EINVAL));
4099 /* check prop value is enabled in features */
4100 feature = zio_checksum_to_feature(intval & ZIO_CHECKSUM_MASK);
4101 if (feature == SPA_FEATURE_NONE)
4104 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4107 * Salted checksums are not supported on root pools.
4109 if (spa_bootfs(spa) != 0 &&
4110 intval < ZIO_CHECKSUM_FUNCTIONS &&
4111 (zio_checksum_table[intval].ci_flags &
4112 ZCHECKSUM_FLAG_SALTED)) {
4113 spa_close(spa, FTAG);
4114 return (SET_ERROR(ERANGE));
4116 if (!spa_feature_is_enabled(spa, feature)) {
4117 spa_close(spa, FTAG);
4118 return (SET_ERROR(ENOTSUP));
4120 spa_close(spa, FTAG);
4128 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
4132 * Removes properties from the given props list that fail permission checks
4133 * needed to clear them and to restore them in case of a receive error. For each
4134 * property, make sure we have both set and inherit permissions.
4136 * Returns the first error encountered if any permission checks fail. If the
4137 * caller provides a non-NULL errlist, it also gives the complete list of names
4138 * of all the properties that failed a permission check along with the
4139 * corresponding error numbers. The caller is responsible for freeing the
4142 * If every property checks out successfully, zero is returned and the list
4143 * pointed at by errlist is NULL.
4146 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
4149 nvpair_t *pair, *next_pair;
4156 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4158 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
4159 (void) strlcpy(zc->zc_name, dataset, sizeof (zc->zc_name));
4160 pair = nvlist_next_nvpair(props, NULL);
4161 while (pair != NULL) {
4162 next_pair = nvlist_next_nvpair(props, pair);
4164 (void) strlcpy(zc->zc_value, nvpair_name(pair),
4165 sizeof (zc->zc_value));
4166 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
4167 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
4168 VERIFY(nvlist_remove_nvpair(props, pair) == 0);
4169 VERIFY(nvlist_add_int32(errors,
4170 zc->zc_value, err) == 0);
4174 kmem_free(zc, sizeof (zfs_cmd_t));
4176 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
4177 nvlist_free(errors);
4180 VERIFY(nvpair_value_int32(pair, &rv) == 0);
4183 if (errlist == NULL)
4184 nvlist_free(errors);
4192 propval_equals(nvpair_t *p1, nvpair_t *p2)
4194 if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
4195 /* dsl_prop_get_all_impl() format */
4197 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
4198 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4202 if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
4204 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
4205 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4209 if (nvpair_type(p1) != nvpair_type(p2))
4212 if (nvpair_type(p1) == DATA_TYPE_STRING) {
4213 char *valstr1, *valstr2;
4215 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
4216 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
4217 return (strcmp(valstr1, valstr2) == 0);
4219 uint64_t intval1, intval2;
4221 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
4222 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
4223 return (intval1 == intval2);
4228 * Remove properties from props if they are not going to change (as determined
4229 * by comparison with origprops). Remove them from origprops as well, since we
4230 * do not need to clear or restore properties that won't change.
4233 props_reduce(nvlist_t *props, nvlist_t *origprops)
4235 nvpair_t *pair, *next_pair;
4237 if (origprops == NULL)
4238 return; /* all props need to be received */
4240 pair = nvlist_next_nvpair(props, NULL);
4241 while (pair != NULL) {
4242 const char *propname = nvpair_name(pair);
4245 next_pair = nvlist_next_nvpair(props, pair);
4247 if ((nvlist_lookup_nvpair(origprops, propname,
4248 &match) != 0) || !propval_equals(pair, match))
4249 goto next; /* need to set received value */
4251 /* don't clear the existing received value */
4252 (void) nvlist_remove_nvpair(origprops, match);
4253 /* don't bother receiving the property */
4254 (void) nvlist_remove_nvpair(props, pair);
4261 * Extract properties that cannot be set PRIOR to the receipt of a dataset.
4262 * For example, refquota cannot be set until after the receipt of a dataset,
4263 * because in replication streams, an older/earlier snapshot may exceed the
4264 * refquota. We want to receive the older/earlier snapshot, but setting
4265 * refquota pre-receipt will set the dsl's ACTUAL quota, which will prevent
4266 * the older/earlier snapshot from being received (with EDQUOT).
4268 * The ZFS test "zfs_receive_011_pos" demonstrates such a scenario.
4270 * libzfs will need to be judicious handling errors encountered by props
4271 * extracted by this function.
4274 extract_delay_props(nvlist_t *props)
4276 nvlist_t *delayprops;
4277 nvpair_t *nvp, *tmp;
4278 static const zfs_prop_t delayable[] = {
4280 ZFS_PROP_KEYLOCATION,
4285 VERIFY(nvlist_alloc(&delayprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4287 for (nvp = nvlist_next_nvpair(props, NULL); nvp != NULL;
4288 nvp = nvlist_next_nvpair(props, nvp)) {
4290 * strcmp() is safe because zfs_prop_to_name() always returns
4293 for (i = 0; delayable[i] != 0; i++) {
4294 if (strcmp(zfs_prop_to_name(delayable[i]),
4295 nvpair_name(nvp)) == 0) {
4299 if (delayable[i] != 0) {
4300 tmp = nvlist_prev_nvpair(props, nvp);
4301 VERIFY(nvlist_add_nvpair(delayprops, nvp) == 0);
4302 VERIFY(nvlist_remove_nvpair(props, nvp) == 0);
4307 if (nvlist_empty(delayprops)) {
4308 nvlist_free(delayprops);
4311 return (delayprops);
4315 static boolean_t zfs_ioc_recv_inject_err;
4319 * nvlist 'errors' is always allocated. It will contain descriptions of
4320 * encountered errors, if any. It's the callers responsibility to free.
4323 zfs_ioc_recv_impl(char *tofs, char *tosnap, char *origin, nvlist_t *recvprops,
4324 nvlist_t *localprops, boolean_t force, boolean_t resumable, int input_fd,
4325 dmu_replay_record_t *begin_record, int cleanup_fd, uint64_t *read_bytes,
4326 uint64_t *errflags, uint64_t *action_handle, nvlist_t **errors)
4328 dmu_recv_cookie_t drc;
4330 int props_error = 0;
4332 nvlist_t *delayprops = NULL; /* sent properties applied post-receive */
4333 nvlist_t *origprops = NULL; /* existing properties */
4334 nvlist_t *origrecvd = NULL; /* existing received properties */
4335 boolean_t first_recvd_props = B_FALSE;
4340 *errors = fnvlist_alloc();
4342 input_fp = getf(input_fd);
4343 if (input_fp == NULL)
4344 return (SET_ERROR(EBADF));
4346 error = dmu_recv_begin(tofs, tosnap,
4347 begin_record, force, resumable, origin, &drc);
4352 * Set properties before we receive the stream so that they are applied
4353 * to the new data. Note that we must call dmu_recv_stream() if
4354 * dmu_recv_begin() succeeds.
4356 if (recvprops != NULL && !drc.drc_newfs) {
4357 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
4358 SPA_VERSION_RECVD_PROPS &&
4359 !dsl_prop_get_hasrecvd(tofs))
4360 first_recvd_props = B_TRUE;
4363 * If new received properties are supplied, they are to
4364 * completely replace the existing received properties, so stash
4365 * away the existing ones.
4367 if (dsl_prop_get_received(tofs, &origrecvd) == 0) {
4368 nvlist_t *errlist = NULL;
4370 * Don't bother writing a property if its value won't
4371 * change (and avoid the unnecessary security checks).
4373 * The first receive after SPA_VERSION_RECVD_PROPS is a
4374 * special case where we blow away all local properties
4377 if (!first_recvd_props)
4378 props_reduce(recvprops, origrecvd);
4379 if (zfs_check_clearable(tofs, origrecvd, &errlist) != 0)
4380 (void) nvlist_merge(*errors, errlist, 0);
4381 nvlist_free(errlist);
4383 if (clear_received_props(tofs, origrecvd,
4384 first_recvd_props ? NULL : recvprops) != 0)
4385 *errflags |= ZPROP_ERR_NOCLEAR;
4387 *errflags |= ZPROP_ERR_NOCLEAR;
4392 * Stash away existing properties so we can restore them on error unless
4393 * we're doing the first receive after SPA_VERSION_RECVD_PROPS, in which
4394 * case "origrecvd" will take care of that.
4396 if (localprops != NULL && !drc.drc_newfs && !first_recvd_props) {
4398 if (dmu_objset_hold(tofs, FTAG, &os) == 0) {
4399 if (dsl_prop_get_all(os, &origprops) != 0) {
4400 *errflags |= ZPROP_ERR_NOCLEAR;
4402 dmu_objset_rele(os, FTAG);
4404 *errflags |= ZPROP_ERR_NOCLEAR;
4408 if (recvprops != NULL) {
4409 props_error = dsl_prop_set_hasrecvd(tofs);
4411 if (props_error == 0) {
4412 delayprops = extract_delay_props(recvprops);
4413 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4414 recvprops, *errors);
4418 if (localprops != NULL) {
4419 nvlist_t *oprops = fnvlist_alloc();
4420 nvlist_t *xprops = fnvlist_alloc();
4421 nvpair_t *nvp = NULL;
4423 while ((nvp = nvlist_next_nvpair(localprops, nvp)) != NULL) {
4424 if (nvpair_type(nvp) == DATA_TYPE_BOOLEAN) {
4426 const char *name = nvpair_name(nvp);
4427 zfs_prop_t prop = zfs_name_to_prop(name);
4428 if (prop != ZPROP_INVAL) {
4429 if (!zfs_prop_inheritable(prop))
4431 } else if (!zfs_prop_user(name))
4433 fnvlist_add_boolean(xprops, name);
4435 /* -o property=value */
4436 fnvlist_add_nvpair(oprops, nvp);
4439 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_LOCAL,
4441 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_INHERITED,
4444 nvlist_free(oprops);
4445 nvlist_free(xprops);
4448 off = input_fp->f_offset;
4449 error = dmu_recv_stream(&drc, input_fp->f_vnode, &off, cleanup_fd,
4453 zfsvfs_t *zfsvfs = NULL;
4454 zvol_state_t *zv = NULL;
4456 if (getzfsvfs(tofs, &zfsvfs) == 0) {
4461 ds = dmu_objset_ds(zfsvfs->z_os);
4462 error = zfs_suspend_fs(zfsvfs);
4464 * If the suspend fails, then the recv_end will
4465 * likely also fail, and clean up after itself.
4467 end_err = dmu_recv_end(&drc, zfsvfs);
4469 error = zfs_resume_fs(zfsvfs, ds);
4470 error = error ? error : end_err;
4471 deactivate_super(zfsvfs->z_sb);
4472 } else if ((zv = zvol_suspend(tofs)) != NULL) {
4473 error = dmu_recv_end(&drc, zvol_tag(zv));
4476 error = dmu_recv_end(&drc, NULL);
4479 /* Set delayed properties now, after we're done receiving. */
4480 if (delayprops != NULL && error == 0) {
4481 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4482 delayprops, *errors);
4486 if (delayprops != NULL) {
4488 * Merge delayed props back in with initial props, in case
4489 * we're DEBUG and zfs_ioc_recv_inject_err is set (which means
4490 * we have to make sure clear_received_props() includes
4491 * the delayed properties).
4493 * Since zfs_ioc_recv_inject_err is only in DEBUG kernels,
4494 * using ASSERT() will be just like a VERIFY.
4496 ASSERT(nvlist_merge(recvprops, delayprops, 0) == 0);
4497 nvlist_free(delayprops);
4501 *read_bytes = off - input_fp->f_offset;
4502 if (VOP_SEEK(input_fp->f_vnode, input_fp->f_offset, &off, NULL) == 0)
4503 input_fp->f_offset = off;
4506 if (zfs_ioc_recv_inject_err) {
4507 zfs_ioc_recv_inject_err = B_FALSE;
4513 * On error, restore the original props.
4515 if (error != 0 && recvprops != NULL && !drc.drc_newfs) {
4516 if (clear_received_props(tofs, recvprops, NULL) != 0) {
4518 * We failed to clear the received properties.
4519 * Since we may have left a $recvd value on the
4520 * system, we can't clear the $hasrecvd flag.
4522 *errflags |= ZPROP_ERR_NORESTORE;
4523 } else if (first_recvd_props) {
4524 dsl_prop_unset_hasrecvd(tofs);
4527 if (origrecvd == NULL && !drc.drc_newfs) {
4528 /* We failed to stash the original properties. */
4529 *errflags |= ZPROP_ERR_NORESTORE;
4533 * dsl_props_set() will not convert RECEIVED to LOCAL on or
4534 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4535 * explicitly if we're restoring local properties cleared in the
4536 * first new-style receive.
4538 if (origrecvd != NULL &&
4539 zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4540 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4541 origrecvd, NULL) != 0) {
4543 * We stashed the original properties but failed to
4546 *errflags |= ZPROP_ERR_NORESTORE;
4549 if (error != 0 && localprops != NULL && !drc.drc_newfs &&
4550 !first_recvd_props) {
4552 nvlist_t *inheritprops;
4555 if (origprops == NULL) {
4556 /* We failed to stash the original properties. */
4557 *errflags |= ZPROP_ERR_NORESTORE;
4561 /* Restore original props */
4562 setprops = fnvlist_alloc();
4563 inheritprops = fnvlist_alloc();
4565 while ((nvp = nvlist_next_nvpair(localprops, nvp)) != NULL) {
4566 const char *name = nvpair_name(nvp);
4570 if (!nvlist_exists(origprops, name)) {
4572 * Property was not present or was explicitly
4573 * inherited before the receive, restore this.
4575 fnvlist_add_boolean(inheritprops, name);
4578 attrs = fnvlist_lookup_nvlist(origprops, name);
4579 source = fnvlist_lookup_string(attrs, ZPROP_SOURCE);
4581 /* Skip received properties */
4582 if (strcmp(source, ZPROP_SOURCE_VAL_RECVD) == 0)
4585 if (strcmp(source, tofs) == 0) {
4586 /* Property was locally set */
4587 fnvlist_add_nvlist(setprops, name, attrs);
4589 /* Property was implicitly inherited */
4590 fnvlist_add_boolean(inheritprops, name);
4594 if (zfs_set_prop_nvlist(tofs, ZPROP_SRC_LOCAL, setprops,
4596 *errflags |= ZPROP_ERR_NORESTORE;
4597 if (zfs_set_prop_nvlist(tofs, ZPROP_SRC_INHERITED, inheritprops,
4599 *errflags |= ZPROP_ERR_NORESTORE;
4601 nvlist_free(setprops);
4602 nvlist_free(inheritprops);
4606 nvlist_free(origrecvd);
4607 nvlist_free(origprops);
4610 error = props_error;
4617 * zc_name name of containing filesystem (unused)
4618 * zc_nvlist_src{_size} nvlist of properties to apply
4619 * zc_nvlist_conf{_size} nvlist of properties to exclude
4620 * (DATA_TYPE_BOOLEAN) and override (everything else)
4621 * zc_value name of snapshot to create
4622 * zc_string name of clone origin (if DRR_FLAG_CLONE)
4623 * zc_cookie file descriptor to recv from
4624 * zc_begin_record the BEGIN record of the stream (not byteswapped)
4625 * zc_guid force flag
4626 * zc_cleanup_fd cleanup-on-exit file descriptor
4627 * zc_action_handle handle for this guid/ds mapping (or zero on first call)
4630 * zc_cookie number of bytes read
4631 * zc_obj zprop_errflags_t
4632 * zc_action_handle handle for this guid/ds mapping
4633 * zc_nvlist_dst{_size} error for each unapplied received property
4636 zfs_ioc_recv(zfs_cmd_t *zc)
4638 dmu_replay_record_t begin_record;
4639 nvlist_t *errors = NULL;
4640 nvlist_t *recvdprops = NULL;
4641 nvlist_t *localprops = NULL;
4642 char *origin = NULL;
4644 char tofs[ZFS_MAX_DATASET_NAME_LEN];
4647 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
4648 strchr(zc->zc_value, '@') == NULL ||
4649 strchr(zc->zc_value, '%'))
4650 return (SET_ERROR(EINVAL));
4652 (void) strlcpy(tofs, zc->zc_value, sizeof (tofs));
4653 tosnap = strchr(tofs, '@');
4656 if (zc->zc_nvlist_src != 0 &&
4657 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
4658 zc->zc_iflags, &recvdprops)) != 0)
4661 if (zc->zc_nvlist_conf != 0 &&
4662 (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
4663 zc->zc_iflags, &localprops)) != 0)
4666 if (zc->zc_string[0])
4667 origin = zc->zc_string;
4669 begin_record.drr_type = DRR_BEGIN;
4670 begin_record.drr_payloadlen = 0;
4671 begin_record.drr_u.drr_begin = zc->zc_begin_record;
4673 error = zfs_ioc_recv_impl(tofs, tosnap, origin, recvdprops, localprops,
4674 zc->zc_guid, B_FALSE, zc->zc_cookie, &begin_record,
4675 zc->zc_cleanup_fd, &zc->zc_cookie, &zc->zc_obj,
4676 &zc->zc_action_handle, &errors);
4677 nvlist_free(recvdprops);
4678 nvlist_free(localprops);
4681 * Now that all props, initial and delayed, are set, report the prop
4682 * errors to the caller.
4684 if (zc->zc_nvlist_dst_size != 0 && errors != NULL &&
4685 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4686 put_nvlist(zc, errors) != 0)) {
4688 * Caller made zc->zc_nvlist_dst less than the minimum expected
4689 * size or supplied an invalid address.
4691 error = SET_ERROR(EINVAL);
4694 nvlist_free(errors);
4701 * "snapname" -> full name of the snapshot to create
4702 * (optional) "props" -> received properties to set (nvlist)
4703 * (optional) "localprops" -> override and exclude properties (nvlist)
4704 * (optional) "origin" -> name of clone origin (DRR_FLAG_CLONE)
4705 * "begin_record" -> non-byteswapped dmu_replay_record_t
4706 * "input_fd" -> file descriptor to read stream from (int32)
4707 * (optional) "force" -> force flag (value ignored)
4708 * (optional) "resumable" -> resumable flag (value ignored)
4709 * (optional) "cleanup_fd" -> cleanup-on-exit file descriptor
4710 * (optional) "action_handle" -> handle for this guid/ds mapping
4714 * "read_bytes" -> number of bytes read
4715 * "error_flags" -> zprop_errflags_t
4716 * "action_handle" -> handle for this guid/ds mapping
4717 * "errors" -> error for each unapplied received property (nvlist)
4721 zfs_ioc_recv_new(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
4723 dmu_replay_record_t *begin_record;
4724 uint_t begin_record_size;
4725 nvlist_t *errors = NULL;
4726 nvlist_t *recvprops = NULL;
4727 nvlist_t *localprops = NULL;
4728 char *snapname = NULL;
4729 char *origin = NULL;
4731 char tofs[ZFS_MAX_DATASET_NAME_LEN];
4733 boolean_t resumable;
4734 uint64_t action_handle = 0;
4735 uint64_t read_bytes = 0;
4736 uint64_t errflags = 0;
4738 int cleanup_fd = -1;
4741 error = nvlist_lookup_string(innvl, "snapname", &snapname);
4743 return (SET_ERROR(EINVAL));
4745 if (dataset_namecheck(snapname, NULL, NULL) != 0 ||
4746 strchr(snapname, '@') == NULL ||
4747 strchr(snapname, '%'))
4748 return (SET_ERROR(EINVAL));
4750 (void) strcpy(tofs, snapname);
4751 tosnap = strchr(tofs, '@');
4754 error = nvlist_lookup_string(innvl, "origin", &origin);
4755 if (error && error != ENOENT)
4758 error = nvlist_lookup_byte_array(innvl, "begin_record",
4759 (uchar_t **)&begin_record, &begin_record_size);
4760 if (error != 0 || begin_record_size != sizeof (*begin_record))
4761 return (SET_ERROR(EINVAL));
4763 error = nvlist_lookup_int32(innvl, "input_fd", &input_fd);
4765 return (SET_ERROR(EINVAL));
4767 force = nvlist_exists(innvl, "force");
4768 resumable = nvlist_exists(innvl, "resumable");
4770 error = nvlist_lookup_int32(innvl, "cleanup_fd", &cleanup_fd);
4771 if (error && error != ENOENT)
4774 error = nvlist_lookup_uint64(innvl, "action_handle", &action_handle);
4775 if (error && error != ENOENT)
4778 /* we still use "props" here for backwards compatibility */
4779 error = nvlist_lookup_nvlist(innvl, "props", &recvprops);
4780 if (error && error != ENOENT)
4783 error = nvlist_lookup_nvlist(innvl, "localprops", &localprops);
4784 if (error && error != ENOENT)
4787 error = zfs_ioc_recv_impl(tofs, tosnap, origin, recvprops, localprops,
4788 force, resumable, input_fd, begin_record, cleanup_fd, &read_bytes,
4789 &errflags, &action_handle, &errors);
4791 fnvlist_add_uint64(outnvl, "read_bytes", read_bytes);
4792 fnvlist_add_uint64(outnvl, "error_flags", errflags);
4793 fnvlist_add_uint64(outnvl, "action_handle", action_handle);
4794 fnvlist_add_nvlist(outnvl, "errors", errors);
4796 nvlist_free(errors);
4797 nvlist_free(recvprops);
4798 nvlist_free(localprops);
4805 * zc_name name of snapshot to send
4806 * zc_cookie file descriptor to send stream to
4807 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj)
4808 * zc_sendobj objsetid of snapshot to send
4809 * zc_fromobj objsetid of incremental fromsnap (may be zero)
4810 * zc_guid if set, estimate size of stream only. zc_cookie is ignored.
4811 * output size in zc_objset_type.
4812 * zc_flags lzc_send_flags
4815 * zc_objset_type estimated size, if zc_guid is set
4817 * NOTE: This is no longer the preferred interface, any new functionality
4818 * should be added to zfs_ioc_send_new() instead.
4821 zfs_ioc_send(zfs_cmd_t *zc)
4825 boolean_t estimate = (zc->zc_guid != 0);
4826 boolean_t embedok = (zc->zc_flags & 0x1);
4827 boolean_t large_block_ok = (zc->zc_flags & 0x2);
4828 boolean_t compressok = (zc->zc_flags & 0x4);
4829 boolean_t rawok = (zc->zc_flags & 0x8);
4831 if (zc->zc_obj != 0) {
4833 dsl_dataset_t *tosnap;
4835 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4839 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4841 dsl_pool_rele(dp, FTAG);
4845 if (dsl_dir_is_clone(tosnap->ds_dir))
4847 dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj;
4848 dsl_dataset_rele(tosnap, FTAG);
4849 dsl_pool_rele(dp, FTAG);
4854 dsl_dataset_t *tosnap;
4855 dsl_dataset_t *fromsnap = NULL;
4857 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4861 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj,
4864 dsl_pool_rele(dp, FTAG);
4868 if (zc->zc_fromobj != 0) {
4869 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
4872 dsl_dataset_rele(tosnap, FTAG);
4873 dsl_pool_rele(dp, FTAG);
4878 error = dmu_send_estimate(tosnap, fromsnap, compressok || rawok,
4879 &zc->zc_objset_type);
4881 if (fromsnap != NULL)
4882 dsl_dataset_rele(fromsnap, FTAG);
4883 dsl_dataset_rele(tosnap, FTAG);
4884 dsl_pool_rele(dp, FTAG);
4886 file_t *fp = getf(zc->zc_cookie);
4888 return (SET_ERROR(EBADF));
4891 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
4892 zc->zc_fromobj, embedok, large_block_ok, compressok, rawok,
4893 zc->zc_cookie, fp->f_vnode, &off);
4895 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4897 releasef(zc->zc_cookie);
4904 * zc_name name of snapshot on which to report progress
4905 * zc_cookie file descriptor of send stream
4908 * zc_cookie number of bytes written in send stream thus far
4911 zfs_ioc_send_progress(zfs_cmd_t *zc)
4915 dmu_sendarg_t *dsp = NULL;
4918 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4922 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4924 dsl_pool_rele(dp, FTAG);
4928 mutex_enter(&ds->ds_sendstream_lock);
4931 * Iterate over all the send streams currently active on this dataset.
4932 * If there's one which matches the specified file descriptor _and_ the
4933 * stream was started by the current process, return the progress of
4937 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4938 dsp = list_next(&ds->ds_sendstreams, dsp)) {
4939 if (dsp->dsa_outfd == zc->zc_cookie &&
4940 dsp->dsa_proc->group_leader == curproc->group_leader)
4945 zc->zc_cookie = *(dsp->dsa_off);
4947 error = SET_ERROR(ENOENT);
4949 mutex_exit(&ds->ds_sendstream_lock);
4950 dsl_dataset_rele(ds, FTAG);
4951 dsl_pool_rele(dp, FTAG);
4956 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4960 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4961 &zc->zc_inject_record);
4964 zc->zc_guid = (uint64_t)id;
4970 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4972 return (zio_clear_fault((int)zc->zc_guid));
4976 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4978 int id = (int)zc->zc_guid;
4981 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4982 &zc->zc_inject_record);
4990 zfs_ioc_error_log(zfs_cmd_t *zc)
4994 size_t count = (size_t)zc->zc_nvlist_dst_size;
4996 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4999 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
5002 zc->zc_nvlist_dst_size = count;
5004 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
5006 spa_close(spa, FTAG);
5012 zfs_ioc_clear(zfs_cmd_t *zc)
5019 * On zpool clear we also fix up missing slogs
5021 mutex_enter(&spa_namespace_lock);
5022 spa = spa_lookup(zc->zc_name);
5024 mutex_exit(&spa_namespace_lock);
5025 return (SET_ERROR(EIO));
5027 if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
5028 /* we need to let spa_open/spa_load clear the chains */
5029 spa_set_log_state(spa, SPA_LOG_CLEAR);
5031 spa->spa_last_open_failed = 0;
5032 mutex_exit(&spa_namespace_lock);
5034 if (zc->zc_cookie & ZPOOL_NO_REWIND) {
5035 error = spa_open(zc->zc_name, &spa, FTAG);
5038 nvlist_t *config = NULL;
5040 if (zc->zc_nvlist_src == 0)
5041 return (SET_ERROR(EINVAL));
5043 if ((error = get_nvlist(zc->zc_nvlist_src,
5044 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
5045 error = spa_open_rewind(zc->zc_name, &spa, FTAG,
5047 if (config != NULL) {
5050 if ((err = put_nvlist(zc, config)) != 0)
5052 nvlist_free(config);
5054 nvlist_free(policy);
5061 spa_vdev_state_enter(spa, SCL_NONE);
5063 if (zc->zc_guid == 0) {
5066 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
5068 (void) spa_vdev_state_exit(spa, NULL, ENODEV);
5069 spa_close(spa, FTAG);
5070 return (SET_ERROR(ENODEV));
5074 vdev_clear(spa, vd);
5076 (void) spa_vdev_state_exit(spa, spa_suspended(spa) ?
5077 NULL : spa->spa_root_vdev, 0);
5080 * Resume any suspended I/Os.
5082 if (zio_resume(spa) != 0)
5083 error = SET_ERROR(EIO);
5085 spa_close(spa, FTAG);
5091 * Reopen all the vdevs associated with the pool.
5094 * "scrub_restart" -> when true and scrub is running, allow to restart
5095 * scrub as the side effect of the reopen (boolean).
5102 zfs_ioc_pool_reopen(const char *pool, nvlist_t *innvl, nvlist_t *outnvl)
5106 boolean_t scrub_restart = B_TRUE;
5109 if (nvlist_lookup_boolean_value(innvl, "scrub_restart",
5110 &scrub_restart) != 0) {
5111 return (SET_ERROR(EINVAL));
5115 error = spa_open(pool, &spa, FTAG);
5119 spa_vdev_state_enter(spa, SCL_NONE);
5122 * If the scrub_restart flag is B_FALSE and a scrub is already
5123 * in progress then set spa_scrub_reopen flag to B_TRUE so that
5124 * we don't restart the scrub as a side effect of the reopen.
5125 * Otherwise, let vdev_open() decided if a resilver is required.
5128 spa->spa_scrub_reopen = (!scrub_restart &&
5129 dsl_scan_scrubbing(spa->spa_dsl_pool));
5130 vdev_reopen(spa->spa_root_vdev);
5131 spa->spa_scrub_reopen = B_FALSE;
5133 (void) spa_vdev_state_exit(spa, NULL, 0);
5134 spa_close(spa, FTAG);
5140 * zc_name name of filesystem
5143 * zc_string name of conflicting snapshot, if there is one
5146 zfs_ioc_promote(zfs_cmd_t *zc)
5149 dsl_dataset_t *ds, *ods;
5150 char origin[ZFS_MAX_DATASET_NAME_LEN];
5154 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
5155 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0 ||
5156 strchr(zc->zc_name, '%'))
5157 return (SET_ERROR(EINVAL));
5159 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5163 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
5165 dsl_pool_rele(dp, FTAG);
5169 if (!dsl_dir_is_clone(ds->ds_dir)) {
5170 dsl_dataset_rele(ds, FTAG);
5171 dsl_pool_rele(dp, FTAG);
5172 return (SET_ERROR(EINVAL));
5175 error = dsl_dataset_hold_obj(dp,
5176 dsl_dir_phys(ds->ds_dir)->dd_origin_obj, FTAG, &ods);
5178 dsl_dataset_rele(ds, FTAG);
5179 dsl_pool_rele(dp, FTAG);
5183 dsl_dataset_name(ods, origin);
5184 dsl_dataset_rele(ods, FTAG);
5185 dsl_dataset_rele(ds, FTAG);
5186 dsl_pool_rele(dp, FTAG);
5189 * We don't need to unmount *all* the origin fs's snapshots, but
5192 cp = strchr(origin, '@');
5195 (void) dmu_objset_find(origin,
5196 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
5197 return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
5201 * Retrieve a single {user|group|project}{used|quota}@... property.
5204 * zc_name name of filesystem
5205 * zc_objset_type zfs_userquota_prop_t
5206 * zc_value domain name (eg. "S-1-234-567-89")
5207 * zc_guid RID/UID/GID
5210 * zc_cookie property value
5213 zfs_ioc_userspace_one(zfs_cmd_t *zc)
5218 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
5219 return (SET_ERROR(EINVAL));
5221 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
5225 error = zfs_userspace_one(zfsvfs,
5226 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
5227 zfsvfs_rele(zfsvfs, FTAG);
5234 * zc_name name of filesystem
5235 * zc_cookie zap cursor
5236 * zc_objset_type zfs_userquota_prop_t
5237 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
5240 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
5241 * zc_cookie zap cursor
5244 zfs_ioc_userspace_many(zfs_cmd_t *zc)
5247 int bufsize = zc->zc_nvlist_dst_size;
5250 return (SET_ERROR(ENOMEM));
5252 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
5256 void *buf = vmem_alloc(bufsize, KM_SLEEP);
5258 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
5259 buf, &zc->zc_nvlist_dst_size);
5262 error = xcopyout(buf,
5263 (void *)(uintptr_t)zc->zc_nvlist_dst,
5264 zc->zc_nvlist_dst_size);
5266 vmem_free(buf, bufsize);
5267 zfsvfs_rele(zfsvfs, FTAG);
5274 * zc_name name of filesystem
5280 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
5286 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
5287 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
5289 * If userused is not enabled, it may be because the
5290 * objset needs to be closed & reopened (to grow the
5291 * objset_phys_t). Suspend/resume the fs will do that.
5293 dsl_dataset_t *ds, *newds;
5295 ds = dmu_objset_ds(zfsvfs->z_os);
5296 error = zfs_suspend_fs(zfsvfs);
5298 dmu_objset_refresh_ownership(ds, &newds,
5300 error = zfs_resume_fs(zfsvfs, newds);
5304 error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
5305 deactivate_super(zfsvfs->z_sb);
5307 /* XXX kind of reading contents without owning */
5308 error = dmu_objset_hold_flags(zc->zc_name, B_TRUE, FTAG, &os);
5312 error = dmu_objset_userspace_upgrade(os);
5313 dmu_objset_rele_flags(os, B_TRUE, FTAG);
5321 * zc_name name of filesystem
5327 zfs_ioc_id_quota_upgrade(zfs_cmd_t *zc)
5332 error = dmu_objset_hold_flags(zc->zc_name, B_TRUE, FTAG, &os);
5336 if (dmu_objset_userobjspace_upgradable(os) ||
5337 dmu_objset_projectquota_upgradable(os)) {
5338 mutex_enter(&os->os_upgrade_lock);
5339 if (os->os_upgrade_id == 0) {
5340 /* clear potential error code and retry */
5341 os->os_upgrade_status = 0;
5342 mutex_exit(&os->os_upgrade_lock);
5344 dmu_objset_id_quota_upgrade(os);
5346 mutex_exit(&os->os_upgrade_lock);
5349 dsl_pool_rele(dmu_objset_pool(os), FTAG);
5351 taskq_wait_id(os->os_spa->spa_upgrade_taskq, os->os_upgrade_id);
5352 error = os->os_upgrade_status;
5354 dsl_pool_rele(dmu_objset_pool(os), FTAG);
5357 dsl_dataset_rele_flags(dmu_objset_ds(os), DS_HOLD_FLAG_DECRYPT, FTAG);
5363 zfs_ioc_share(zfs_cmd_t *zc)
5365 return (SET_ERROR(ENOSYS));
5368 ace_t full_access[] = {
5369 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
5374 * zc_name name of containing filesystem
5375 * zc_obj object # beyond which we want next in-use object #
5378 * zc_obj next in-use object #
5381 zfs_ioc_next_obj(zfs_cmd_t *zc)
5383 objset_t *os = NULL;
5386 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
5390 error = dmu_object_next(os, &zc->zc_obj, B_FALSE, 0);
5392 dmu_objset_rele(os, FTAG);
5398 * zc_name name of filesystem
5399 * zc_value prefix name for snapshot
5400 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
5403 * zc_value short name of new snapshot
5406 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
5413 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
5417 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
5418 (u_longlong_t)ddi_get_lbolt64());
5419 hold_name = kmem_asprintf("%%%s", zc->zc_value);
5421 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
5424 (void) strlcpy(zc->zc_value, snap_name,
5425 sizeof (zc->zc_value));
5428 zfs_onexit_fd_rele(zc->zc_cleanup_fd);
5434 * zc_name name of "to" snapshot
5435 * zc_value name of "from" snapshot
5436 * zc_cookie file descriptor to write diff data on
5439 * dmu_diff_record_t's to the file descriptor
5442 zfs_ioc_diff(zfs_cmd_t *zc)
5448 fp = getf(zc->zc_cookie);
5450 return (SET_ERROR(EBADF));
5454 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off);
5456 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5458 releasef(zc->zc_cookie);
5464 * Remove all ACL files in shares dir
5466 #ifdef HAVE_SMB_SHARE
5468 zfs_smb_acl_purge(znode_t *dzp)
5471 zap_attribute_t zap;
5472 zfsvfs_t *zfsvfs = ZTOZSB(dzp);
5475 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
5476 (error = zap_cursor_retrieve(&zc, &zap)) == 0;
5477 zap_cursor_advance(&zc)) {
5478 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
5482 zap_cursor_fini(&zc);
5485 #endif /* HAVE_SMB_SHARE */
5488 zfs_ioc_smb_acl(zfs_cmd_t *zc)
5490 #ifdef HAVE_SMB_SHARE
5493 vnode_t *resourcevp = NULL;
5502 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
5503 NO_FOLLOW, NULL, &vp)) != 0)
5506 /* Now make sure mntpnt and dataset are ZFS */
5508 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
5509 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
5510 zc->zc_name) != 0)) {
5512 return (SET_ERROR(EINVAL));
5516 zfsvfs = ZTOZSB(dzp);
5520 * Create share dir if its missing.
5522 mutex_enter(&zfsvfs->z_lock);
5523 if (zfsvfs->z_shares_dir == 0) {
5526 tx = dmu_tx_create(zfsvfs->z_os);
5527 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
5529 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
5530 error = dmu_tx_assign(tx, TXG_WAIT);
5534 error = zfs_create_share_dir(zfsvfs, tx);
5538 mutex_exit(&zfsvfs->z_lock);
5544 mutex_exit(&zfsvfs->z_lock);
5546 ASSERT(zfsvfs->z_shares_dir);
5547 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
5553 switch (zc->zc_cookie) {
5554 case ZFS_SMB_ACL_ADD:
5555 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
5556 vattr.va_mode = S_IFREG|0777;
5560 vsec.vsa_mask = VSA_ACE;
5561 vsec.vsa_aclentp = &full_access;
5562 vsec.vsa_aclentsz = sizeof (full_access);
5563 vsec.vsa_aclcnt = 1;
5565 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
5566 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
5568 VN_RELE(resourcevp);
5571 case ZFS_SMB_ACL_REMOVE:
5572 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
5576 case ZFS_SMB_ACL_RENAME:
5577 if ((error = get_nvlist(zc->zc_nvlist_src,
5578 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
5580 VN_RELE(ZTOV(sharedir));
5584 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
5585 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
5588 VN_RELE(ZTOV(sharedir));
5590 nvlist_free(nvlist);
5593 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
5595 nvlist_free(nvlist);
5598 case ZFS_SMB_ACL_PURGE:
5599 error = zfs_smb_acl_purge(sharedir);
5603 error = SET_ERROR(EINVAL);
5608 VN_RELE(ZTOV(sharedir));
5614 return (SET_ERROR(ENOTSUP));
5615 #endif /* HAVE_SMB_SHARE */
5620 * "holds" -> { snapname -> holdname (string), ... }
5621 * (optional) "cleanup_fd" -> fd (int32)
5625 * snapname -> error value (int32)
5631 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
5635 int cleanup_fd = -1;
5639 error = nvlist_lookup_nvlist(args, "holds", &holds);
5641 return (SET_ERROR(EINVAL));
5643 /* make sure the user didn't pass us any invalid (empty) tags */
5644 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
5645 pair = nvlist_next_nvpair(holds, pair)) {
5648 error = nvpair_value_string(pair, &htag);
5650 return (SET_ERROR(error));
5652 if (strlen(htag) == 0)
5653 return (SET_ERROR(EINVAL));
5656 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
5657 error = zfs_onexit_fd_hold(cleanup_fd, &minor);
5662 error = dsl_dataset_user_hold(holds, minor, errlist);
5664 zfs_onexit_fd_rele(cleanup_fd);
5669 * innvl is not used.
5672 * holdname -> time added (uint64 seconds since epoch)
5678 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
5680 ASSERT3P(args, ==, NULL);
5681 return (dsl_dataset_get_holds(snapname, outnvl));
5686 * snapname -> { holdname, ... }
5691 * snapname -> error value (int32)
5697 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
5699 return (dsl_dataset_user_release(holds, errlist));
5704 * zc_guid flags (ZEVENT_NONBLOCK)
5705 * zc_cleanup_fd zevent file descriptor
5708 * zc_nvlist_dst next nvlist event
5709 * zc_cookie dropped events since last get
5712 zfs_ioc_events_next(zfs_cmd_t *zc)
5715 nvlist_t *event = NULL;
5717 uint64_t dropped = 0;
5720 error = zfs_zevent_fd_hold(zc->zc_cleanup_fd, &minor, &ze);
5725 error = zfs_zevent_next(ze, &event,
5726 &zc->zc_nvlist_dst_size, &dropped);
5727 if (event != NULL) {
5728 zc->zc_cookie = dropped;
5729 error = put_nvlist(zc, event);
5733 if (zc->zc_guid & ZEVENT_NONBLOCK)
5736 if ((error == 0) || (error != ENOENT))
5739 error = zfs_zevent_wait(ze);
5744 zfs_zevent_fd_rele(zc->zc_cleanup_fd);
5751 * zc_cookie cleared events count
5754 zfs_ioc_events_clear(zfs_cmd_t *zc)
5758 zfs_zevent_drain_all(&count);
5759 zc->zc_cookie = count;
5766 * zc_guid eid | ZEVENT_SEEK_START | ZEVENT_SEEK_END
5767 * zc_cleanup zevent file descriptor
5770 zfs_ioc_events_seek(zfs_cmd_t *zc)
5776 error = zfs_zevent_fd_hold(zc->zc_cleanup_fd, &minor, &ze);
5780 error = zfs_zevent_seek(ze, zc->zc_guid);
5781 zfs_zevent_fd_rele(zc->zc_cleanup_fd);
5788 * zc_name name of new filesystem or snapshot
5789 * zc_value full name of old snapshot
5792 * zc_cookie space in bytes
5793 * zc_objset_type compressed space in bytes
5794 * zc_perm_action uncompressed space in bytes
5797 zfs_ioc_space_written(zfs_cmd_t *zc)
5801 dsl_dataset_t *new, *old;
5803 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5806 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
5808 dsl_pool_rele(dp, FTAG);
5811 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
5813 dsl_dataset_rele(new, FTAG);
5814 dsl_pool_rele(dp, FTAG);
5818 error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
5819 &zc->zc_objset_type, &zc->zc_perm_action);
5820 dsl_dataset_rele(old, FTAG);
5821 dsl_dataset_rele(new, FTAG);
5822 dsl_pool_rele(dp, FTAG);
5828 * "firstsnap" -> snapshot name
5832 * "used" -> space in bytes
5833 * "compressed" -> compressed space in bytes
5834 * "uncompressed" -> uncompressed space in bytes
5838 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
5842 dsl_dataset_t *new, *old;
5844 uint64_t used, comp, uncomp;
5846 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5847 return (SET_ERROR(EINVAL));
5849 error = dsl_pool_hold(lastsnap, FTAG, &dp);
5853 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
5854 if (error == 0 && !new->ds_is_snapshot) {
5855 dsl_dataset_rele(new, FTAG);
5856 error = SET_ERROR(EINVAL);
5859 dsl_pool_rele(dp, FTAG);
5862 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
5863 if (error == 0 && !old->ds_is_snapshot) {
5864 dsl_dataset_rele(old, FTAG);
5865 error = SET_ERROR(EINVAL);
5868 dsl_dataset_rele(new, FTAG);
5869 dsl_pool_rele(dp, FTAG);
5873 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5874 dsl_dataset_rele(old, FTAG);
5875 dsl_dataset_rele(new, FTAG);
5876 dsl_pool_rele(dp, FTAG);
5877 fnvlist_add_uint64(outnvl, "used", used);
5878 fnvlist_add_uint64(outnvl, "compressed", comp);
5879 fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5885 * "fd" -> file descriptor to write stream to (int32)
5886 * (optional) "fromsnap" -> full snap name to send an incremental from
5887 * (optional) "largeblockok" -> (value ignored)
5888 * indicates that blocks > 128KB are permitted
5889 * (optional) "embedok" -> (value ignored)
5890 * presence indicates DRR_WRITE_EMBEDDED records are permitted
5891 * (optional) "compressok" -> (value ignored)
5892 * presence indicates compressed DRR_WRITE records are permitted
5893 * (optional) "rawok" -> (value ignored)
5894 * presence indicates raw encrypted records should be used.
5895 * (optional) "resume_object" and "resume_offset" -> (uint64)
5896 * if present, resume send stream from specified object and offset.
5903 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5907 char *fromname = NULL;
5910 boolean_t largeblockok;
5912 boolean_t compressok;
5914 uint64_t resumeobj = 0;
5915 uint64_t resumeoff = 0;
5917 error = nvlist_lookup_int32(innvl, "fd", &fd);
5919 return (SET_ERROR(EINVAL));
5921 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
5923 largeblockok = nvlist_exists(innvl, "largeblockok");
5924 embedok = nvlist_exists(innvl, "embedok");
5925 compressok = nvlist_exists(innvl, "compressok");
5926 rawok = nvlist_exists(innvl, "rawok");
5928 (void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj);
5929 (void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff);
5931 if ((fp = getf(fd)) == NULL)
5932 return (SET_ERROR(EBADF));
5935 error = dmu_send(snapname, fromname, embedok, largeblockok, compressok,
5936 rawok, fd, resumeobj, resumeoff, fp->f_vnode, &off);
5938 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5946 * Determine approximately how large a zfs send stream will be -- the number
5947 * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5950 * (optional) "from" -> full snap or bookmark name to send an incremental
5952 * (optional) "largeblockok" -> (value ignored)
5953 * indicates that blocks > 128KB are permitted
5954 * (optional) "embedok" -> (value ignored)
5955 * presence indicates DRR_WRITE_EMBEDDED records are permitted
5956 * (optional) "compressok" -> (value ignored)
5957 * presence indicates compressed DRR_WRITE records are permitted
5958 * (optional) "rawok" -> (value ignored)
5959 * presence indicates raw encrypted records should be used.
5963 * "space" -> bytes of space (uint64)
5967 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5970 dsl_dataset_t *tosnap;
5973 boolean_t compressok;
5977 error = dsl_pool_hold(snapname, FTAG, &dp);
5981 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
5983 dsl_pool_rele(dp, FTAG);
5987 compressok = nvlist_exists(innvl, "compressok");
5988 rawok = nvlist_exists(innvl, "rawok");
5990 error = nvlist_lookup_string(innvl, "from", &fromname);
5992 if (strchr(fromname, '@') != NULL) {
5994 * If from is a snapshot, hold it and use the more
5995 * efficient dmu_send_estimate to estimate send space
5996 * size using deadlists.
5998 dsl_dataset_t *fromsnap;
5999 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
6002 error = dmu_send_estimate(tosnap, fromsnap,
6003 compressok || rawok, &space);
6004 dsl_dataset_rele(fromsnap, FTAG);
6005 } else if (strchr(fromname, '#') != NULL) {
6007 * If from is a bookmark, fetch the creation TXG of the
6008 * snapshot it was created from and use that to find
6009 * blocks that were born after it.
6011 zfs_bookmark_phys_t frombm;
6013 error = dsl_bookmark_lookup(dp, fromname, tosnap,
6017 error = dmu_send_estimate_from_txg(tosnap,
6018 frombm.zbm_creation_txg, compressok || rawok,
6022 * from is not properly formatted as a snapshot or
6025 error = SET_ERROR(EINVAL);
6030 * If estimating the size of a full send, use dmu_send_estimate.
6032 error = dmu_send_estimate(tosnap, NULL, compressok || rawok,
6036 fnvlist_add_uint64(outnvl, "space", space);
6039 dsl_dataset_rele(tosnap, FTAG);
6040 dsl_pool_rele(dp, FTAG);
6045 * Sync the currently open TXG to disk for the specified pool.
6046 * This is somewhat similar to 'zfs_sync()'.
6047 * For cases that do not result in error this ioctl will wait for
6048 * the currently open TXG to commit before returning back to the caller.
6051 * "force" -> when true, force uberblock update even if there is no dirty data.
6052 * In addition this will cause the vdev configuration to be written
6053 * out including updating the zpool cache file. (boolean_t)
6060 zfs_ioc_pool_sync(const char *pool, nvlist_t *innvl, nvlist_t *onvl)
6063 boolean_t force = B_FALSE;
6066 if ((err = spa_open(pool, &spa, FTAG)) != 0)
6070 if (nvlist_lookup_boolean_value(innvl, "force", &force) != 0) {
6071 err = SET_ERROR(EINVAL);
6077 spa_config_enter(spa, SCL_CONFIG, FTAG, RW_WRITER);
6078 vdev_config_dirty(spa->spa_root_vdev);
6079 spa_config_exit(spa, SCL_CONFIG, FTAG);
6081 txg_wait_synced(spa_get_dsl(spa), 0);
6083 spa_close(spa, FTAG);
6089 * Load a user's wrapping key into the kernel.
6091 * "hidden_args" -> { "wkeydata" -> value }
6092 * raw uint8_t array of encryption wrapping key data (32 bytes)
6093 * (optional) "noop" -> (value ignored)
6094 * presence indicated key should only be verified, not loaded
6099 zfs_ioc_load_key(const char *dsname, nvlist_t *innvl, nvlist_t *outnvl)
6102 dsl_crypto_params_t *dcp = NULL;
6103 nvlist_t *hidden_args;
6104 boolean_t noop = nvlist_exists(innvl, "noop");
6106 if (strchr(dsname, '@') != NULL || strchr(dsname, '%') != NULL) {
6107 ret = SET_ERROR(EINVAL);
6111 ret = nvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS, &hidden_args);
6113 ret = SET_ERROR(EINVAL);
6117 ret = dsl_crypto_params_create_nvlist(DCP_CMD_NONE, NULL,
6122 ret = spa_keystore_load_wkey(dsname, dcp, noop);
6126 dsl_crypto_params_free(dcp, noop);
6131 dsl_crypto_params_free(dcp, B_TRUE);
6136 * Unload a user's wrapping key from the kernel.
6137 * Both innvl and outnvl are unused.
6141 zfs_ioc_unload_key(const char *dsname, nvlist_t *innvl, nvlist_t *outnvl)
6145 if (strchr(dsname, '@') != NULL || strchr(dsname, '%') != NULL) {
6146 ret = (SET_ERROR(EINVAL));
6150 ret = spa_keystore_unload_wkey(dsname);
6159 * Changes a user's wrapping key used to decrypt a dataset. The keyformat,
6160 * keylocation, pbkdf2salt, and pbkdf2iters properties can also be specified
6161 * here to change how the key is derived in userspace.
6164 * "hidden_args" (optional) -> { "wkeydata" -> value }
6165 * raw uint8_t array of new encryption wrapping key data (32 bytes)
6166 * "props" (optional) -> { prop -> value }
6173 zfs_ioc_change_key(const char *dsname, nvlist_t *innvl, nvlist_t *outnvl)
6176 uint64_t cmd = DCP_CMD_NONE;
6177 dsl_crypto_params_t *dcp = NULL;
6178 nvlist_t *args = NULL, *hidden_args = NULL;
6180 if (strchr(dsname, '@') != NULL || strchr(dsname, '%') != NULL) {
6181 ret = (SET_ERROR(EINVAL));
6185 (void) nvlist_lookup_uint64(innvl, "crypt_cmd", &cmd);
6186 (void) nvlist_lookup_nvlist(innvl, "props", &args);
6187 (void) nvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS, &hidden_args);
6189 ret = dsl_crypto_params_create_nvlist(cmd, args, hidden_args, &dcp);
6193 ret = spa_keystore_change_key(dsname, dcp);
6197 dsl_crypto_params_free(dcp, B_FALSE);
6202 dsl_crypto_params_free(dcp, B_TRUE);
6206 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
6209 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
6210 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
6211 boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
6213 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
6215 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
6216 ASSERT3U(ioc, <, ZFS_IOC_LAST);
6217 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
6218 ASSERT3P(vec->zvec_func, ==, NULL);
6220 vec->zvec_legacy_func = func;
6221 vec->zvec_secpolicy = secpolicy;
6222 vec->zvec_namecheck = namecheck;
6223 vec->zvec_allow_log = log_history;
6224 vec->zvec_pool_check = pool_check;
6228 * See the block comment at the beginning of this file for details on
6229 * each argument to this function.
6232 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
6233 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
6234 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
6235 boolean_t allow_log)
6237 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
6239 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
6240 ASSERT3U(ioc, <, ZFS_IOC_LAST);
6241 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
6242 ASSERT3P(vec->zvec_func, ==, NULL);
6244 /* if we are logging, the name must be valid */
6245 ASSERT(!allow_log || namecheck != NO_NAME);
6247 vec->zvec_name = name;
6248 vec->zvec_func = func;
6249 vec->zvec_secpolicy = secpolicy;
6250 vec->zvec_namecheck = namecheck;
6251 vec->zvec_pool_check = pool_check;
6252 vec->zvec_smush_outnvlist = smush_outnvlist;
6253 vec->zvec_allow_log = allow_log;
6257 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
6258 zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
6259 zfs_ioc_poolcheck_t pool_check)
6261 zfs_ioctl_register_legacy(ioc, func, secpolicy,
6262 POOL_NAME, log_history, pool_check);
6266 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
6267 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
6269 zfs_ioctl_register_legacy(ioc, func, secpolicy,
6270 DATASET_NAME, B_FALSE, pool_check);
6274 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
6276 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
6277 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
6281 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
6282 zfs_secpolicy_func_t *secpolicy)
6284 zfs_ioctl_register_legacy(ioc, func, secpolicy,
6285 NO_NAME, B_FALSE, POOL_CHECK_NONE);
6289 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
6290 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
6292 zfs_ioctl_register_legacy(ioc, func, secpolicy,
6293 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
6297 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
6299 zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
6300 zfs_secpolicy_read);
6304 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
6305 zfs_secpolicy_func_t *secpolicy)
6307 zfs_ioctl_register_legacy(ioc, func, secpolicy,
6308 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
6312 zfs_ioctl_init(void)
6314 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
6315 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
6316 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6318 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
6319 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
6320 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
6322 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
6323 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
6324 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
6326 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
6327 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
6328 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
6330 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
6331 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
6332 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
6334 zfs_ioctl_register("create", ZFS_IOC_CREATE,
6335 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
6336 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6338 zfs_ioctl_register("clone", ZFS_IOC_CLONE,
6339 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
6340 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6342 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
6343 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
6344 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6346 zfs_ioctl_register("hold", ZFS_IOC_HOLD,
6347 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
6348 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6349 zfs_ioctl_register("release", ZFS_IOC_RELEASE,
6350 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
6351 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6353 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
6354 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
6355 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
6357 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
6358 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
6359 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE);
6361 zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK,
6362 zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME,
6363 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6365 zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS,
6366 zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME,
6367 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
6369 zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS,
6370 zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks,
6372 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6374 zfs_ioctl_register("receive", ZFS_IOC_RECV_NEW,
6375 zfs_ioc_recv_new, zfs_secpolicy_recv_new, DATASET_NAME,
6376 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6377 zfs_ioctl_register("load-key", ZFS_IOC_LOAD_KEY,
6378 zfs_ioc_load_key, zfs_secpolicy_load_key,
6379 DATASET_NAME, POOL_CHECK_SUSPENDED, B_TRUE, B_TRUE);
6380 zfs_ioctl_register("unload-key", ZFS_IOC_UNLOAD_KEY,
6381 zfs_ioc_unload_key, zfs_secpolicy_load_key,
6382 DATASET_NAME, POOL_CHECK_SUSPENDED, B_TRUE, B_TRUE);
6383 zfs_ioctl_register("change-key", ZFS_IOC_CHANGE_KEY,
6384 zfs_ioc_change_key, zfs_secpolicy_change_key,
6385 DATASET_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY,
6388 zfs_ioctl_register("sync", ZFS_IOC_POOL_SYNC,
6389 zfs_ioc_pool_sync, zfs_secpolicy_none, POOL_NAME,
6390 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
6391 zfs_ioctl_register("reopen", ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
6392 zfs_secpolicy_config, POOL_NAME, POOL_CHECK_SUSPENDED, B_TRUE,
6395 zfs_ioctl_register("channel_program", ZFS_IOC_CHANNEL_PROGRAM,
6396 zfs_ioc_channel_program, zfs_secpolicy_config,
6397 POOL_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE,
6400 /* IOCTLS that use the legacy function signature */
6402 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
6403 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
6405 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
6406 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
6407 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
6409 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
6410 zfs_ioc_pool_upgrade);
6411 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
6413 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
6414 zfs_ioc_vdev_remove);
6415 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
6416 zfs_ioc_vdev_set_state);
6417 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
6418 zfs_ioc_vdev_attach);
6419 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
6420 zfs_ioc_vdev_detach);
6421 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
6422 zfs_ioc_vdev_setpath);
6423 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
6424 zfs_ioc_vdev_setfru);
6425 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
6426 zfs_ioc_pool_set_props);
6427 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
6428 zfs_ioc_vdev_split);
6429 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
6430 zfs_ioc_pool_reguid);
6432 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
6433 zfs_ioc_pool_configs, zfs_secpolicy_none);
6434 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
6435 zfs_ioc_pool_tryimport, zfs_secpolicy_config);
6436 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
6437 zfs_ioc_inject_fault, zfs_secpolicy_inject);
6438 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
6439 zfs_ioc_clear_fault, zfs_secpolicy_inject);
6440 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
6441 zfs_ioc_inject_list_next, zfs_secpolicy_inject);
6444 * pool destroy, and export don't log the history as part of
6445 * zfsdev_ioctl, but rather zfs_ioc_pool_export
6446 * does the logging of those commands.
6448 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
6449 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
6450 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
6451 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
6453 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
6454 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
6455 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
6456 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
6458 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
6459 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED);
6460 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
6461 zfs_ioc_dsobj_to_dsname,
6462 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED);
6463 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
6464 zfs_ioc_pool_get_history,
6465 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
6467 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
6468 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
6470 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
6471 zfs_secpolicy_config, B_TRUE, POOL_CHECK_READONLY);
6473 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
6474 zfs_ioc_space_written);
6475 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
6476 zfs_ioc_objset_recvd_props);
6477 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
6479 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
6481 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
6482 zfs_ioc_objset_stats);
6483 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
6484 zfs_ioc_objset_zplprops);
6485 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
6486 zfs_ioc_dataset_list_next);
6487 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
6488 zfs_ioc_snapshot_list_next);
6489 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
6490 zfs_ioc_send_progress);
6492 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
6493 zfs_ioc_diff, zfs_secpolicy_diff);
6494 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
6495 zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
6496 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
6497 zfs_ioc_obj_to_path, zfs_secpolicy_diff);
6498 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
6499 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
6500 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
6501 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
6502 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
6503 zfs_ioc_send, zfs_secpolicy_send);
6505 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
6506 zfs_secpolicy_none);
6507 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
6508 zfs_secpolicy_destroy);
6509 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
6510 zfs_secpolicy_rename);
6511 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
6512 zfs_secpolicy_recv);
6513 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
6514 zfs_secpolicy_promote);
6515 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
6516 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
6517 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
6518 zfs_secpolicy_set_fsacl);
6520 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
6521 zfs_secpolicy_share, POOL_CHECK_NONE);
6522 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
6523 zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
6524 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
6525 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
6526 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
6527 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
6528 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
6529 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
6534 zfs_ioctl_register_legacy(ZFS_IOC_EVENTS_NEXT, zfs_ioc_events_next,
6535 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);
6536 zfs_ioctl_register_legacy(ZFS_IOC_EVENTS_CLEAR, zfs_ioc_events_clear,
6537 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);
6538 zfs_ioctl_register_legacy(ZFS_IOC_EVENTS_SEEK, zfs_ioc_events_seek,
6539 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);
6543 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
6544 zfs_ioc_poolcheck_t check)
6549 ASSERT(type == POOL_NAME || type == DATASET_NAME);
6551 if (check & POOL_CHECK_NONE)
6554 error = spa_open(name, &spa, FTAG);
6556 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
6557 error = SET_ERROR(EAGAIN);
6558 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
6559 error = SET_ERROR(EROFS);
6560 spa_close(spa, FTAG);
6566 zfsdev_get_state_impl(minor_t minor, enum zfsdev_state_type which)
6570 for (zs = zfsdev_state_list; zs != NULL; zs = zs->zs_next) {
6571 if (zs->zs_minor == minor) {
6575 return (zs->zs_onexit);
6577 return (zs->zs_zevent);
6588 zfsdev_get_state(minor_t minor, enum zfsdev_state_type which)
6592 ptr = zfsdev_get_state_impl(minor, which);
6598 zfsdev_getminor(struct file *filp, minor_t *minorp)
6600 zfsdev_state_t *zs, *fpd;
6602 ASSERT(filp != NULL);
6603 ASSERT(!MUTEX_HELD(&zfsdev_state_lock));
6605 fpd = filp->private_data;
6607 return (SET_ERROR(EBADF));
6609 mutex_enter(&zfsdev_state_lock);
6611 for (zs = zfsdev_state_list; zs != NULL; zs = zs->zs_next) {
6613 if (zs->zs_minor == -1)
6617 *minorp = fpd->zs_minor;
6618 mutex_exit(&zfsdev_state_lock);
6623 mutex_exit(&zfsdev_state_lock);
6625 return (SET_ERROR(EBADF));
6629 * Find a free minor number. The zfsdev_state_list is expected to
6630 * be short since it is only a list of currently open file handles.
6633 zfsdev_minor_alloc(void)
6635 static minor_t last_minor = 0;
6638 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
6640 for (m = last_minor + 1; m != last_minor; m++) {
6641 if (m > ZFSDEV_MAX_MINOR)
6643 if (zfsdev_get_state_impl(m, ZST_ALL) == NULL) {
6653 zfsdev_state_init(struct file *filp)
6655 zfsdev_state_t *zs, *zsprev = NULL;
6657 boolean_t newzs = B_FALSE;
6659 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
6661 minor = zfsdev_minor_alloc();
6663 return (SET_ERROR(ENXIO));
6665 for (zs = zfsdev_state_list; zs != NULL; zs = zs->zs_next) {
6666 if (zs->zs_minor == -1)
6672 zs = kmem_zalloc(sizeof (zfsdev_state_t), KM_SLEEP);
6677 filp->private_data = zs;
6679 zfs_onexit_init((zfs_onexit_t **)&zs->zs_onexit);
6680 zfs_zevent_init((zfs_zevent_t **)&zs->zs_zevent);
6684 * In order to provide for lock-free concurrent read access
6685 * to the minor list in zfsdev_get_state_impl(), new entries
6686 * must be completely written before linking them into the
6687 * list whereas existing entries are already linked; the last
6688 * operation must be updating zs_minor (from -1 to the new
6692 zs->zs_minor = minor;
6694 zsprev->zs_next = zs;
6697 zs->zs_minor = minor;
6704 zfsdev_state_destroy(struct file *filp)
6708 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
6709 ASSERT(filp->private_data != NULL);
6711 zs = filp->private_data;
6713 zfs_onexit_destroy(zs->zs_onexit);
6714 zfs_zevent_destroy(zs->zs_zevent);
6720 zfsdev_open(struct inode *ino, struct file *filp)
6724 mutex_enter(&zfsdev_state_lock);
6725 error = zfsdev_state_init(filp);
6726 mutex_exit(&zfsdev_state_lock);
6732 zfsdev_release(struct inode *ino, struct file *filp)
6736 mutex_enter(&zfsdev_state_lock);
6737 error = zfsdev_state_destroy(filp);
6738 mutex_exit(&zfsdev_state_lock);
6744 zfsdev_ioctl(struct file *filp, unsigned cmd, unsigned long arg)
6748 int error, rc, flag = 0;
6749 const zfs_ioc_vec_t *vec;
6750 char *saved_poolname = NULL;
6751 nvlist_t *innvl = NULL;
6752 fstrans_cookie_t cookie;
6754 vecnum = cmd - ZFS_IOC_FIRST;
6755 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
6756 return (-SET_ERROR(EINVAL));
6757 vec = &zfs_ioc_vec[vecnum];
6760 * The registered ioctl list may be sparse, verify that either
6761 * a normal or legacy handler are registered.
6763 if (vec->zvec_func == NULL && vec->zvec_legacy_func == NULL)
6764 return (-SET_ERROR(EINVAL));
6766 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
6768 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
6770 error = SET_ERROR(EFAULT);
6774 zc->zc_iflags = flag & FKIOCTL;
6775 if (zc->zc_nvlist_src_size > MAX_NVLIST_SRC_SIZE) {
6777 * Make sure the user doesn't pass in an insane value for
6778 * zc_nvlist_src_size. We have to check, since we will end
6779 * up allocating that much memory inside of get_nvlist(). This
6780 * prevents a nefarious user from allocating tons of kernel
6783 * Also, we return EINVAL instead of ENOMEM here. The reason
6784 * being that returning ENOMEM from an ioctl() has a special
6785 * connotation; that the user's size value is too small and
6786 * needs to be expanded to hold the nvlist. See
6787 * zcmd_expand_dst_nvlist() for details.
6789 error = SET_ERROR(EINVAL); /* User's size too big */
6791 } else if (zc->zc_nvlist_src_size != 0) {
6792 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
6793 zc->zc_iflags, &innvl);
6799 * Ensure that all pool/dataset names are valid before we pass down to
6802 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
6803 switch (vec->zvec_namecheck) {
6805 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
6806 error = SET_ERROR(EINVAL);
6808 error = pool_status_check(zc->zc_name,
6809 vec->zvec_namecheck, vec->zvec_pool_check);
6813 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
6814 error = SET_ERROR(EINVAL);
6816 error = pool_status_check(zc->zc_name,
6817 vec->zvec_namecheck, vec->zvec_pool_check);
6826 cookie = spl_fstrans_mark();
6827 error = vec->zvec_secpolicy(zc, innvl, CRED());
6828 spl_fstrans_unmark(cookie);
6834 /* legacy ioctls can modify zc_name */
6835 saved_poolname = strdup(zc->zc_name);
6836 if (saved_poolname == NULL) {
6837 error = SET_ERROR(ENOMEM);
6840 saved_poolname[strcspn(saved_poolname, "/@#")] = '\0';
6843 if (vec->zvec_func != NULL) {
6847 nvlist_t *lognv = NULL;
6849 ASSERT(vec->zvec_legacy_func == NULL);
6852 * Add the innvl to the lognv before calling the func,
6853 * in case the func changes the innvl.
6855 if (vec->zvec_allow_log) {
6856 lognv = fnvlist_alloc();
6857 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
6859 if (!nvlist_empty(innvl)) {
6860 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
6865 outnvl = fnvlist_alloc();
6866 cookie = spl_fstrans_mark();
6867 error = vec->zvec_func(zc->zc_name, innvl, outnvl);
6868 spl_fstrans_unmark(cookie);
6871 * Some commands can partially execute, modify state, and still
6872 * return an error. In these cases, attempt to record what
6876 (cmd == ZFS_IOC_CHANNEL_PROGRAM && error != EINVAL)) &&
6877 vec->zvec_allow_log &&
6878 spa_open(zc->zc_name, &spa, FTAG) == 0) {
6879 if (!nvlist_empty(outnvl)) {
6880 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
6884 fnvlist_add_int64(lognv, ZPOOL_HIST_ERRNO,
6887 (void) spa_history_log_nvl(spa, lognv);
6888 spa_close(spa, FTAG);
6890 fnvlist_free(lognv);
6892 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
6894 if (vec->zvec_smush_outnvlist) {
6895 smusherror = nvlist_smush(outnvl,
6896 zc->zc_nvlist_dst_size);
6898 if (smusherror == 0)
6899 puterror = put_nvlist(zc, outnvl);
6905 nvlist_free(outnvl);
6907 cookie = spl_fstrans_mark();
6908 error = vec->zvec_legacy_func(zc);
6909 spl_fstrans_unmark(cookie);
6914 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
6915 if (error == 0 && rc != 0)
6916 error = SET_ERROR(EFAULT);
6917 if (error == 0 && vec->zvec_allow_log) {
6918 char *s = tsd_get(zfs_allow_log_key);
6921 (void) tsd_set(zfs_allow_log_key, saved_poolname);
6923 if (saved_poolname != NULL)
6924 strfree(saved_poolname);
6927 kmem_free(zc, sizeof (zfs_cmd_t));
6931 #ifdef CONFIG_COMPAT
6933 zfsdev_compat_ioctl(struct file *filp, unsigned cmd, unsigned long arg)
6935 return (zfsdev_ioctl(filp, cmd, arg));
6938 #define zfsdev_compat_ioctl NULL
6941 static const struct file_operations zfsdev_fops = {
6942 .open = zfsdev_open,
6943 .release = zfsdev_release,
6944 .unlocked_ioctl = zfsdev_ioctl,
6945 .compat_ioctl = zfsdev_compat_ioctl,
6946 .owner = THIS_MODULE,
6949 static struct miscdevice zfs_misc = {
6952 .fops = &zfsdev_fops,
6955 MODULE_ALIAS_MISCDEV(ZFS_MINOR);
6956 MODULE_ALIAS("devname:zfs");
6963 mutex_init(&zfsdev_state_lock, NULL, MUTEX_DEFAULT, NULL);
6964 zfsdev_state_list = kmem_zalloc(sizeof (zfsdev_state_t), KM_SLEEP);
6965 zfsdev_state_list->zs_minor = -1;
6967 error = misc_register(&zfs_misc);
6968 if (error == -EBUSY) {
6970 * Fallback to dynamic minor allocation in the event of a
6971 * collision with a reserved minor in linux/miscdevice.h.
6972 * In this case the kernel modules must be manually loaded.
6974 printk(KERN_INFO "ZFS: misc_register() with static minor %d "
6975 "failed %d, retrying with MISC_DYNAMIC_MINOR\n",
6978 zfs_misc.minor = MISC_DYNAMIC_MINOR;
6979 error = misc_register(&zfs_misc);
6983 printk(KERN_INFO "ZFS: misc_register() failed %d\n", error);
6991 zfsdev_state_t *zs, *zsprev = NULL;
6993 misc_deregister(&zfs_misc);
6994 mutex_destroy(&zfsdev_state_lock);
6996 for (zs = zfsdev_state_list; zs != NULL; zs = zs->zs_next) {
6998 kmem_free(zsprev, sizeof (zfsdev_state_t));
7002 kmem_free(zsprev, sizeof (zfsdev_state_t));
7006 zfs_allow_log_destroy(void *arg)
7008 char *poolname = arg;
7010 if (poolname != NULL)
7015 #define ZFS_DEBUG_STR " (DEBUG mode)"
7017 #define ZFS_DEBUG_STR ""
7025 error = -vn_set_pwd("/");
7028 "ZFS: Warning unable to set pwd to '/': %d\n", error);
7032 if ((error = -zvol_init()) != 0)
7035 spa_init(FREAD | FWRITE);
7040 if ((error = zfs_attach()) != 0)
7043 tsd_create(&zfs_fsyncer_key, NULL);
7044 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
7045 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
7047 printk(KERN_NOTICE "ZFS: Loaded module v%s-%s%s, "
7048 "ZFS pool version %s, ZFS filesystem version %s\n",
7049 ZFS_META_VERSION, ZFS_META_RELEASE, ZFS_DEBUG_STR,
7050 SPA_VERSION_STRING, ZPL_VERSION_STRING);
7051 #ifndef CONFIG_FS_POSIX_ACL
7052 printk(KERN_NOTICE "ZFS: Posix ACLs disabled by kernel\n");
7053 #endif /* CONFIG_FS_POSIX_ACL */
7061 printk(KERN_NOTICE "ZFS: Failed to Load ZFS Filesystem v%s-%s%s"
7062 ", rc = %d\n", ZFS_META_VERSION, ZFS_META_RELEASE,
7063 ZFS_DEBUG_STR, error);
7076 tsd_destroy(&zfs_fsyncer_key);
7077 tsd_destroy(&rrw_tsd_key);
7078 tsd_destroy(&zfs_allow_log_key);
7080 printk(KERN_NOTICE "ZFS: Unloaded module v%s-%s%s\n",
7081 ZFS_META_VERSION, ZFS_META_RELEASE, ZFS_DEBUG_STR);
7088 MODULE_DESCRIPTION("ZFS");
7089 MODULE_AUTHOR(ZFS_META_AUTHOR);
7090 MODULE_LICENSE(ZFS_META_LICENSE);
7091 MODULE_VERSION(ZFS_META_VERSION "-" ZFS_META_RELEASE);
7092 #endif /* HAVE_SPL */