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 * Portions Copyright 2012 Pawel Jakub Dawidek <pawel@dawidek.net>
26 * Copyright (c) 2012, Joyent, Inc. All rights reserved.
27 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
28 * Copyright (c) 2012, Joyent, Inc. All rights reserved.
29 * Copyright (c) 201i3 by Delphix. All rights reserved.
30 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
31 * Copyright (c) 2013 Steven Hartland. All rights reserved.
37 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
38 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
40 * There are two ways that we handle ioctls: the legacy way where almost
41 * all of the logic is in the ioctl callback, and the new way where most
42 * of the marshalling is handled in the common entry point, zfsdev_ioctl().
44 * Non-legacy ioctls should be registered by calling
45 * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked
46 * from userland by lzc_ioctl().
48 * The registration arguments are as follows:
51 * The name of the ioctl. This is used for history logging. If the
52 * ioctl returns successfully (the callback returns 0), and allow_log
53 * is true, then a history log entry will be recorded with the input &
54 * output nvlists. The log entry can be printed with "zpool history -i".
57 * The ioctl request number, which userland will pass to ioctl(2).
58 * The ioctl numbers can change from release to release, because
59 * the caller (libzfs) must be matched to the kernel.
61 * zfs_secpolicy_func_t *secpolicy
62 * This function will be called before the zfs_ioc_func_t, to
63 * determine if this operation is permitted. It should return EPERM
64 * on failure, and 0 on success. Checks include determining if the
65 * dataset is visible in this zone, and if the user has either all
66 * zfs privileges in the zone (SYS_MOUNT), or has been granted permission
67 * to do this operation on this dataset with "zfs allow".
69 * zfs_ioc_namecheck_t namecheck
70 * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
71 * name, a dataset name, or nothing. If the name is not well-formed,
72 * the ioctl will fail and the callback will not be called.
73 * Therefore, the callback can assume that the name is well-formed
74 * (e.g. is null-terminated, doesn't have more than one '@' character,
75 * doesn't have invalid characters).
77 * zfs_ioc_poolcheck_t pool_check
78 * This specifies requirements on the pool state. If the pool does
79 * not meet them (is suspended or is readonly), the ioctl will fail
80 * and the callback will not be called. If any checks are specified
81 * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
82 * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED |
83 * POOL_CHECK_READONLY).
85 * boolean_t smush_outnvlist
86 * If smush_outnvlist is true, then the output is presumed to be a
87 * list of errors, and it will be "smushed" down to fit into the
88 * caller's buffer, by removing some entries and replacing them with a
89 * single "N_MORE_ERRORS" entry indicating how many were removed. See
90 * nvlist_smush() for details. If smush_outnvlist is false, and the
91 * outnvlist does not fit into the userland-provided buffer, then the
92 * ioctl will fail with ENOMEM.
94 * zfs_ioc_func_t *func
95 * The callback function that will perform the operation.
97 * The callback should return 0 on success, or an error number on
98 * failure. If the function fails, the userland ioctl will return -1,
99 * and errno will be set to the callback's return value. The callback
100 * will be called with the following arguments:
103 * The name of the pool or dataset to operate on, from
104 * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the
105 * expected type (pool, dataset, or none).
108 * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or
109 * NULL if no input nvlist was provided. Changes to this nvlist are
110 * ignored. If the input nvlist could not be deserialized, the
111 * ioctl will fail and the callback will not be called.
114 * The output nvlist, initially empty. The callback can fill it in,
115 * and it will be returned to userland by serializing it into
116 * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization
117 * fails (e.g. because the caller didn't supply a large enough
118 * buffer), then the overall ioctl will fail. See the
119 * 'smush_nvlist' argument above for additional behaviors.
121 * There are two typical uses of the output nvlist:
122 * - To return state, e.g. property values. In this case,
123 * smush_outnvlist should be false. If the buffer was not large
124 * enough, the caller will reallocate a larger buffer and try
127 * - To return multiple errors from an ioctl which makes on-disk
128 * changes. In this case, smush_outnvlist should be true.
129 * Ioctls which make on-disk modifications should generally not
130 * use the outnvl if they succeed, because the caller can not
131 * distinguish between the operation failing, and
132 * deserialization failing.
135 #include <sys/types.h>
136 #include <sys/param.h>
137 #include <sys/errno.h>
140 #include <sys/modctl.h>
141 #include <sys/open.h>
142 #include <sys/file.h>
143 #include <sys/kmem.h>
144 #include <sys/conf.h>
145 #include <sys/cmn_err.h>
146 #include <sys/stat.h>
147 #include <sys/zfs_ioctl.h>
148 #include <sys/zfs_vfsops.h>
149 #include <sys/zfs_znode.h>
152 #include <sys/spa_impl.h>
153 #include <sys/vdev.h>
154 #include <sys/priv_impl.h>
156 #include <sys/dsl_dir.h>
157 #include <sys/dsl_dataset.h>
158 #include <sys/dsl_prop.h>
159 #include <sys/dsl_deleg.h>
160 #include <sys/dmu_objset.h>
161 #include <sys/dmu_impl.h>
162 #include <sys/dmu_tx.h>
164 #include <sys/sunddi.h>
165 #include <sys/sunldi.h>
166 #include <sys/policy.h>
167 #include <sys/zone.h>
168 #include <sys/nvpair.h>
169 #include <sys/pathname.h>
170 #include <sys/mount.h>
172 #include <sys/fs/zfs.h>
173 #include <sys/zfs_ctldir.h>
174 #include <sys/zfs_dir.h>
175 #include <sys/zfs_onexit.h>
176 #include <sys/zvol.h>
177 #include <sys/dsl_scan.h>
178 #include <sharefs/share.h>
179 #include <sys/fm/util.h>
181 #include <sys/dmu_send.h>
182 #include <sys/dsl_destroy.h>
183 #include <sys/dsl_userhold.h>
184 #include <sys/zfeature.h>
186 #include <linux/miscdevice.h>
188 #include "zfs_namecheck.h"
189 #include "zfs_prop.h"
190 #include "zfs_deleg.h"
191 #include "zfs_comutil.h"
193 kmutex_t zfsdev_state_lock;
194 zfsdev_state_t *zfsdev_state_list;
196 extern void zfs_init(void);
197 extern void zfs_fini(void);
199 uint_t zfs_fsyncer_key;
200 extern uint_t rrw_tsd_key;
201 static uint_t zfs_allow_log_key;
203 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *);
204 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *);
205 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *);
211 } zfs_ioc_namecheck_t;
214 POOL_CHECK_NONE = 1 << 0,
215 POOL_CHECK_SUSPENDED = 1 << 1,
216 POOL_CHECK_READONLY = 1 << 2,
217 } zfs_ioc_poolcheck_t;
219 typedef struct zfs_ioc_vec {
220 zfs_ioc_legacy_func_t *zvec_legacy_func;
221 zfs_ioc_func_t *zvec_func;
222 zfs_secpolicy_func_t *zvec_secpolicy;
223 zfs_ioc_namecheck_t zvec_namecheck;
224 boolean_t zvec_allow_log;
225 zfs_ioc_poolcheck_t zvec_pool_check;
226 boolean_t zvec_smush_outnvlist;
227 const char *zvec_name;
230 /* This array is indexed by zfs_userquota_prop_t */
231 static const char *userquota_perms[] = {
232 ZFS_DELEG_PERM_USERUSED,
233 ZFS_DELEG_PERM_USERQUOTA,
234 ZFS_DELEG_PERM_GROUPUSED,
235 ZFS_DELEG_PERM_GROUPQUOTA,
238 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
239 static int zfs_check_settable(const char *name, nvpair_t *property,
241 static int zfs_check_clearable(char *dataset, nvlist_t *props,
243 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
245 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
246 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
248 static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature);
251 history_str_free(char *buf)
253 kmem_free(buf, HIS_MAX_RECORD_LEN);
257 history_str_get(zfs_cmd_t *zc)
261 if (zc->zc_history == 0)
264 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP | KM_NODEBUG);
265 if (copyinstr((void *)(uintptr_t)zc->zc_history,
266 buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
267 history_str_free(buf);
271 buf[HIS_MAX_RECORD_LEN -1] = '\0';
277 * Check to see if the named dataset is currently defined as bootable
280 zfs_is_bootfs(const char *name)
284 if (dmu_objset_hold(name, FTAG, &os) == 0) {
286 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
287 dmu_objset_rele(os, FTAG);
294 * Return non-zero if the spa version is less than requested version.
297 zfs_earlier_version(const char *name, int version)
301 if (spa_open(name, &spa, FTAG) == 0) {
302 if (spa_version(spa) < version) {
303 spa_close(spa, FTAG);
306 spa_close(spa, FTAG);
312 * Return TRUE if the ZPL version is less than requested version.
315 zpl_earlier_version(const char *name, int version)
318 boolean_t rc = B_TRUE;
320 if (dmu_objset_hold(name, FTAG, &os) == 0) {
323 if (dmu_objset_type(os) != DMU_OST_ZFS) {
324 dmu_objset_rele(os, FTAG);
327 /* XXX reading from non-owned objset */
328 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
329 rc = zplversion < version;
330 dmu_objset_rele(os, FTAG);
336 zfs_log_history(zfs_cmd_t *zc)
341 if ((buf = history_str_get(zc)) == NULL)
344 if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
345 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
346 (void) spa_history_log(spa, buf);
347 spa_close(spa, FTAG);
349 history_str_free(buf);
353 * Policy for top-level read operations (list pools). Requires no privileges,
354 * and can be used in the local zone, as there is no associated dataset.
358 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
364 * Policy for dataset read operations (list children, get statistics). Requires
365 * no privileges, but must be visible in the local zone.
369 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
371 if (INGLOBALZONE(curproc) ||
372 zone_dataset_visible(zc->zc_name, NULL))
375 return (SET_ERROR(ENOENT));
379 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
384 * The dataset must be visible by this zone -- check this first
385 * so they don't see EPERM on something they shouldn't know about.
387 if (!INGLOBALZONE(curproc) &&
388 !zone_dataset_visible(dataset, &writable))
389 return (SET_ERROR(ENOENT));
391 if (INGLOBALZONE(curproc)) {
393 * If the fs is zoned, only root can access it from the
396 if (secpolicy_zfs(cr) && zoned)
397 return (SET_ERROR(EPERM));
400 * If we are in a local zone, the 'zoned' property must be set.
403 return (SET_ERROR(EPERM));
405 /* must be writable by this zone */
407 return (SET_ERROR(EPERM));
413 zfs_dozonecheck(const char *dataset, cred_t *cr)
417 if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
418 return (SET_ERROR(ENOENT));
420 return (zfs_dozonecheck_impl(dataset, zoned, cr));
424 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
428 if (dsl_prop_get_int_ds(ds, "zoned", &zoned))
429 return (SET_ERROR(ENOENT));
431 return (zfs_dozonecheck_impl(dataset, zoned, cr));
435 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
436 const char *perm, cred_t *cr)
440 error = zfs_dozonecheck_ds(name, ds, cr);
442 error = secpolicy_zfs(cr);
444 error = dsl_deleg_access_impl(ds, perm, cr);
450 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
456 error = dsl_pool_hold(name, FTAG, &dp);
460 error = dsl_dataset_hold(dp, name, FTAG, &ds);
462 dsl_pool_rele(dp, FTAG);
466 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
468 dsl_dataset_rele(ds, FTAG);
469 dsl_pool_rele(dp, FTAG);
474 * Policy for setting the security label property.
476 * Returns 0 for success, non-zero for access and other errors.
479 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
482 char ds_hexsl[MAXNAMELEN];
483 bslabel_t ds_sl, new_sl;
484 boolean_t new_default = FALSE;
486 int needed_priv = -1;
489 /* First get the existing dataset label. */
490 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
491 1, sizeof (ds_hexsl), &ds_hexsl, NULL);
493 return (SET_ERROR(EPERM));
495 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
498 /* The label must be translatable */
499 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
500 return (SET_ERROR(EINVAL));
503 * In a non-global zone, disallow attempts to set a label that
504 * doesn't match that of the zone; otherwise no other checks
507 if (!INGLOBALZONE(curproc)) {
508 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
509 return (SET_ERROR(EPERM));
514 * For global-zone datasets (i.e., those whose zoned property is
515 * "off", verify that the specified new label is valid for the
518 if (dsl_prop_get_integer(name,
519 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
520 return (SET_ERROR(EPERM));
522 if (zfs_check_global_label(name, strval) != 0)
523 return (SET_ERROR(EPERM));
527 * If the existing dataset label is nondefault, check if the
528 * dataset is mounted (label cannot be changed while mounted).
529 * Get the zfs_sb_t; if there isn't one, then the dataset isn't
530 * mounted (or isn't a dataset, doesn't exist, ...).
532 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
534 static char *setsl_tag = "setsl_tag";
537 * Try to own the dataset; abort if there is any error,
538 * (e.g., already mounted, in use, or other error).
540 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
543 return (SET_ERROR(EPERM));
545 dmu_objset_disown(os, setsl_tag);
548 needed_priv = PRIV_FILE_DOWNGRADE_SL;
552 if (hexstr_to_label(strval, &new_sl) != 0)
553 return (SET_ERROR(EPERM));
555 if (blstrictdom(&ds_sl, &new_sl))
556 needed_priv = PRIV_FILE_DOWNGRADE_SL;
557 else if (blstrictdom(&new_sl, &ds_sl))
558 needed_priv = PRIV_FILE_UPGRADE_SL;
560 /* dataset currently has a default label */
562 needed_priv = PRIV_FILE_UPGRADE_SL;
566 if (needed_priv != -1)
567 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
571 #endif /* HAVE_MLSLABEL */
575 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
581 * Check permissions for special properties.
588 * Disallow setting of 'zoned' from within a local zone.
590 if (!INGLOBALZONE(curproc))
591 return (SET_ERROR(EPERM));
595 if (!INGLOBALZONE(curproc)) {
597 char setpoint[MAXNAMELEN];
599 * Unprivileged users are allowed to modify the
600 * quota on things *under* (ie. contained by)
601 * the thing they own.
603 if (dsl_prop_get_integer(dsname, "zoned", &zoned,
605 return (SET_ERROR(EPERM));
606 if (!zoned || strlen(dsname) <= strlen(setpoint))
607 return (SET_ERROR(EPERM));
611 case ZFS_PROP_MLSLABEL:
612 if (!is_system_labeled())
613 return (SET_ERROR(EPERM));
615 if (nvpair_value_string(propval, &strval) == 0) {
618 err = zfs_set_slabel_policy(dsname, strval, CRED());
625 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
630 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
634 error = zfs_dozonecheck(zc->zc_name, cr);
639 * permission to set permissions will be evaluated later in
640 * dsl_deleg_can_allow()
647 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
649 return (zfs_secpolicy_write_perms(zc->zc_name,
650 ZFS_DELEG_PERM_ROLLBACK, cr));
655 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
663 * Generate the current snapshot name from the given objsetid, then
664 * use that name for the secpolicy/zone checks.
666 cp = strchr(zc->zc_name, '@');
668 return (SET_ERROR(EINVAL));
669 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
673 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
675 dsl_pool_rele(dp, FTAG);
679 dsl_dataset_name(ds, zc->zc_name);
681 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
682 ZFS_DELEG_PERM_SEND, cr);
683 dsl_dataset_rele(ds, FTAG);
684 dsl_pool_rele(dp, FTAG);
691 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
693 return (zfs_secpolicy_write_perms(zc->zc_name,
694 ZFS_DELEG_PERM_SEND, cr));
697 #ifdef HAVE_SMB_SHARE
700 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
705 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
706 NO_FOLLOW, NULL, &vp)) != 0)
709 /* Now make sure mntpnt and dataset are ZFS */
711 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
712 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
713 zc->zc_name) != 0)) {
715 return (SET_ERROR(EPERM));
719 return (dsl_deleg_access(zc->zc_name,
720 ZFS_DELEG_PERM_SHARE, cr));
722 #endif /* HAVE_SMB_SHARE */
725 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
727 #ifdef HAVE_SMB_SHARE
728 if (!INGLOBALZONE(curproc))
729 return (SET_ERROR(EPERM));
731 if (secpolicy_nfs(cr) == 0) {
734 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
737 return (SET_ERROR(ENOTSUP));
738 #endif /* HAVE_SMB_SHARE */
742 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
744 #ifdef HAVE_SMB_SHARE
745 if (!INGLOBALZONE(curproc))
746 return (SET_ERROR(EPERM));
748 if (secpolicy_smb(cr) == 0) {
751 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
754 return (SET_ERROR(ENOTSUP));
755 #endif /* HAVE_SMB_SHARE */
759 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
764 * Remove the @bla or /bla from the end of the name to get the parent.
766 (void) strncpy(parent, datasetname, parentsize);
767 cp = strrchr(parent, '@');
771 cp = strrchr(parent, '/');
773 return (SET_ERROR(ENOENT));
781 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
785 if ((error = zfs_secpolicy_write_perms(name,
786 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
789 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
794 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
796 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
800 * Destroying snapshots with delegated permissions requires
801 * descendant mount and destroy permissions.
805 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
808 nvpair_t *pair, *nextpair;
811 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
812 return (SET_ERROR(EINVAL));
813 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
818 error = dsl_pool_hold(nvpair_name(pair), FTAG, &dp);
821 nextpair = nvlist_next_nvpair(snaps, pair);
822 error = dsl_dataset_hold(dp, nvpair_name(pair), FTAG, &ds);
824 dsl_dataset_rele(ds, FTAG);
825 dsl_pool_rele(dp, FTAG);
828 error = zfs_secpolicy_destroy_perms(nvpair_name(pair),
830 } else if (error == ENOENT) {
832 * Ignore any snapshots that don't exist (we consider
833 * them "already destroyed"). Remove the name from the
834 * nvl here in case the snapshot is created between
835 * now and when we try to destroy it (in which case
836 * we don't want to destroy it since we haven't
837 * checked for permission).
839 fnvlist_remove_nvpair(snaps, pair);
850 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
852 char parentname[MAXNAMELEN];
855 if ((error = zfs_secpolicy_write_perms(from,
856 ZFS_DELEG_PERM_RENAME, cr)) != 0)
859 if ((error = zfs_secpolicy_write_perms(from,
860 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
863 if ((error = zfs_get_parent(to, parentname,
864 sizeof (parentname))) != 0)
867 if ((error = zfs_secpolicy_write_perms(parentname,
868 ZFS_DELEG_PERM_CREATE, cr)) != 0)
871 if ((error = zfs_secpolicy_write_perms(parentname,
872 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
880 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
882 return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
887 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
890 dsl_dataset_t *clone;
893 error = zfs_secpolicy_write_perms(zc->zc_name,
894 ZFS_DELEG_PERM_PROMOTE, cr);
898 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
902 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
905 char parentname[MAXNAMELEN];
906 dsl_dataset_t *origin = NULL;
910 error = dsl_dataset_hold_obj(dd->dd_pool,
911 dd->dd_phys->dd_origin_obj, FTAG, &origin);
913 dsl_dataset_rele(clone, FTAG);
914 dsl_pool_rele(dp, FTAG);
918 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
919 ZFS_DELEG_PERM_MOUNT, cr);
921 dsl_dataset_name(origin, parentname);
923 error = zfs_secpolicy_write_perms_ds(parentname, origin,
924 ZFS_DELEG_PERM_PROMOTE, cr);
926 dsl_dataset_rele(clone, FTAG);
927 dsl_dataset_rele(origin, FTAG);
929 dsl_pool_rele(dp, FTAG);
935 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
939 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
940 ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
943 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
944 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
947 return (zfs_secpolicy_write_perms(zc->zc_name,
948 ZFS_DELEG_PERM_CREATE, cr));
952 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
954 return (zfs_secpolicy_write_perms(name,
955 ZFS_DELEG_PERM_SNAPSHOT, cr));
959 * Check for permission to create each snapshot in the nvlist.
963 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
969 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
970 return (SET_ERROR(EINVAL));
971 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
972 pair = nvlist_next_nvpair(snaps, pair)) {
973 char *name = nvpair_name(pair);
974 char *atp = strchr(name, '@');
977 error = SET_ERROR(EINVAL);
981 error = zfs_secpolicy_snapshot_perms(name, cr);
991 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
994 * Even root must have a proper TSD so that we know what pool
997 if (tsd_get(zfs_allow_log_key) == NULL)
998 return (SET_ERROR(EPERM));
1003 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1005 char parentname[MAXNAMELEN];
1009 if ((error = zfs_get_parent(zc->zc_name, parentname,
1010 sizeof (parentname))) != 0)
1013 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1014 (error = zfs_secpolicy_write_perms(origin,
1015 ZFS_DELEG_PERM_CLONE, cr)) != 0)
1018 if ((error = zfs_secpolicy_write_perms(parentname,
1019 ZFS_DELEG_PERM_CREATE, cr)) != 0)
1022 return (zfs_secpolicy_write_perms(parentname,
1023 ZFS_DELEG_PERM_MOUNT, cr));
1027 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires
1028 * SYS_CONFIG privilege, which is not available in a local zone.
1032 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1034 if (secpolicy_sys_config(cr, B_FALSE) != 0)
1035 return (SET_ERROR(EPERM));
1041 * Policy for object to name lookups.
1045 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1049 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1052 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1057 * Policy for fault injection. Requires all privileges.
1061 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1063 return (secpolicy_zinject(cr));
1068 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1070 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1072 if (prop == ZPROP_INVAL) {
1073 if (!zfs_prop_user(zc->zc_value))
1074 return (SET_ERROR(EINVAL));
1075 return (zfs_secpolicy_write_perms(zc->zc_name,
1076 ZFS_DELEG_PERM_USERPROP, cr));
1078 return (zfs_secpolicy_setprop(zc->zc_name, prop,
1084 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1086 int err = zfs_secpolicy_read(zc, innvl, cr);
1090 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1091 return (SET_ERROR(EINVAL));
1093 if (zc->zc_value[0] == 0) {
1095 * They are asking about a posix uid/gid. If it's
1096 * themself, allow it.
1098 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1099 zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
1100 if (zc->zc_guid == crgetuid(cr))
1103 if (groupmember(zc->zc_guid, cr))
1108 return (zfs_secpolicy_write_perms(zc->zc_name,
1109 userquota_perms[zc->zc_objset_type], cr));
1113 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1115 int err = zfs_secpolicy_read(zc, innvl, cr);
1119 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1120 return (SET_ERROR(EINVAL));
1122 return (zfs_secpolicy_write_perms(zc->zc_name,
1123 userquota_perms[zc->zc_objset_type], cr));
1128 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1130 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1136 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1142 error = nvlist_lookup_nvlist(innvl, "holds", &holds);
1144 return (SET_ERROR(EINVAL));
1146 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
1147 pair = nvlist_next_nvpair(holds, pair)) {
1148 char fsname[MAXNAMELEN];
1149 error = dmu_fsname(nvpair_name(pair), fsname);
1152 error = zfs_secpolicy_write_perms(fsname,
1153 ZFS_DELEG_PERM_HOLD, cr);
1162 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1167 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1168 pair = nvlist_next_nvpair(innvl, pair)) {
1169 char fsname[MAXNAMELEN];
1170 error = dmu_fsname(nvpair_name(pair), fsname);
1173 error = zfs_secpolicy_write_perms(fsname,
1174 ZFS_DELEG_PERM_RELEASE, cr);
1182 * Policy for allowing temporary snapshots to be taken or released
1185 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1188 * A temporary snapshot is the same as a snapshot,
1189 * hold, destroy and release all rolled into one.
1190 * Delegated diff alone is sufficient that we allow this.
1194 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1195 ZFS_DELEG_PERM_DIFF, cr)) == 0)
1198 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1200 error = zfs_secpolicy_hold(zc, innvl, cr);
1202 error = zfs_secpolicy_release(zc, innvl, cr);
1204 error = zfs_secpolicy_destroy(zc, innvl, cr);
1209 * Returns the nvlist as specified by the user in the zfs_cmd_t.
1212 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1216 nvlist_t *list = NULL;
1219 * Read in and unpack the user-supplied nvlist.
1222 return (SET_ERROR(EINVAL));
1224 packed = kmem_alloc(size, KM_SLEEP | KM_NODEBUG);
1226 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1228 kmem_free(packed, size);
1232 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1233 kmem_free(packed, size);
1237 kmem_free(packed, size);
1244 * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1245 * Entries will be removed from the end of the nvlist, and one int32 entry
1246 * named "N_MORE_ERRORS" will be added indicating how many entries were
1250 nvlist_smush(nvlist_t *errors, size_t max)
1254 size = fnvlist_size(errors);
1257 nvpair_t *more_errors;
1261 return (SET_ERROR(ENOMEM));
1263 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1264 more_errors = nvlist_prev_nvpair(errors, NULL);
1267 nvpair_t *pair = nvlist_prev_nvpair(errors,
1269 fnvlist_remove_nvpair(errors, pair);
1271 size = fnvlist_size(errors);
1272 } while (size > max);
1274 fnvlist_remove_nvpair(errors, more_errors);
1275 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1276 ASSERT3U(fnvlist_size(errors), <=, max);
1283 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1285 char *packed = NULL;
1289 size = fnvlist_size(nvl);
1291 if (size > zc->zc_nvlist_dst_size) {
1292 error = SET_ERROR(ENOMEM);
1294 packed = fnvlist_pack(nvl, &size);
1295 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1296 size, zc->zc_iflags) != 0)
1297 error = SET_ERROR(EFAULT);
1298 fnvlist_pack_free(packed, size);
1301 zc->zc_nvlist_dst_size = size;
1302 zc->zc_nvlist_dst_filled = B_TRUE;
1307 get_zfs_sb(const char *dsname, zfs_sb_t **zsbp)
1312 error = dmu_objset_hold(dsname, FTAG, &os);
1315 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1316 dmu_objset_rele(os, FTAG);
1317 return (SET_ERROR(EINVAL));
1320 mutex_enter(&os->os_user_ptr_lock);
1321 *zsbp = dmu_objset_get_user(os);
1322 if (*zsbp && (*zsbp)->z_sb) {
1323 atomic_inc(&((*zsbp)->z_sb->s_active));
1325 error = SET_ERROR(ESRCH);
1327 mutex_exit(&os->os_user_ptr_lock);
1328 dmu_objset_rele(os, FTAG);
1333 * Find a zfs_sb_t for a mounted filesystem, or create our own, in which
1334 * case its z_sb will be NULL, and it will be opened as the owner.
1335 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1336 * which prevents all inode ops from running.
1339 zfs_sb_hold(const char *name, void *tag, zfs_sb_t **zsbp, boolean_t writer)
1343 if (get_zfs_sb(name, zsbp) != 0)
1344 error = zfs_sb_create(name, zsbp);
1346 rrw_enter(&(*zsbp)->z_teardown_lock, (writer) ? RW_WRITER :
1348 if ((*zsbp)->z_unmounted) {
1350 * XXX we could probably try again, since the unmounting
1351 * thread should be just about to disassociate the
1352 * objset from the zsb.
1354 rrw_exit(&(*zsbp)->z_teardown_lock, tag);
1355 return (SET_ERROR(EBUSY));
1362 zfs_sb_rele(zfs_sb_t *zsb, void *tag)
1364 rrw_exit(&zsb->z_teardown_lock, tag);
1367 deactivate_super(zsb->z_sb);
1369 dmu_objset_disown(zsb->z_os, zsb);
1375 zfs_ioc_pool_create(zfs_cmd_t *zc)
1378 nvlist_t *config, *props = NULL;
1379 nvlist_t *rootprops = NULL;
1380 nvlist_t *zplprops = NULL;
1382 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1383 zc->zc_iflags, &config)))
1386 if (zc->zc_nvlist_src_size != 0 && (error =
1387 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1388 zc->zc_iflags, &props))) {
1389 nvlist_free(config);
1394 nvlist_t *nvl = NULL;
1395 uint64_t version = SPA_VERSION;
1397 (void) nvlist_lookup_uint64(props,
1398 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1399 if (!SPA_VERSION_IS_SUPPORTED(version)) {
1400 error = SET_ERROR(EINVAL);
1401 goto pool_props_bad;
1403 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1405 error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1407 nvlist_free(config);
1411 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1413 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1414 error = zfs_fill_zplprops_root(version, rootprops,
1417 goto pool_props_bad;
1420 error = spa_create(zc->zc_name, config, props, zplprops);
1423 * Set the remaining root properties
1425 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1426 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1427 (void) spa_destroy(zc->zc_name);
1430 nvlist_free(rootprops);
1431 nvlist_free(zplprops);
1432 nvlist_free(config);
1439 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1442 zfs_log_history(zc);
1443 error = spa_destroy(zc->zc_name);
1445 zvol_remove_minors(zc->zc_name);
1450 zfs_ioc_pool_import(zfs_cmd_t *zc)
1452 nvlist_t *config, *props = NULL;
1456 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1457 zc->zc_iflags, &config)) != 0)
1460 if (zc->zc_nvlist_src_size != 0 && (error =
1461 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1462 zc->zc_iflags, &props))) {
1463 nvlist_free(config);
1467 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1468 guid != zc->zc_guid)
1469 error = SET_ERROR(EINVAL);
1471 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1473 if (zc->zc_nvlist_dst != 0) {
1476 if ((err = put_nvlist(zc, config)) != 0)
1480 nvlist_free(config);
1489 zfs_ioc_pool_export(zfs_cmd_t *zc)
1492 boolean_t force = (boolean_t)zc->zc_cookie;
1493 boolean_t hardforce = (boolean_t)zc->zc_guid;
1495 zfs_log_history(zc);
1496 error = spa_export(zc->zc_name, NULL, force, hardforce);
1498 zvol_remove_minors(zc->zc_name);
1503 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1508 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1509 return (SET_ERROR(EEXIST));
1511 error = put_nvlist(zc, configs);
1513 nvlist_free(configs);
1520 * zc_name name of the pool
1523 * zc_cookie real errno
1524 * zc_nvlist_dst config nvlist
1525 * zc_nvlist_dst_size size of config nvlist
1528 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1534 error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1535 sizeof (zc->zc_value));
1537 if (config != NULL) {
1538 ret = put_nvlist(zc, config);
1539 nvlist_free(config);
1542 * The config may be present even if 'error' is non-zero.
1543 * In this case we return success, and preserve the real errno
1546 zc->zc_cookie = error;
1555 * Try to import the given pool, returning pool stats as appropriate so that
1556 * user land knows which devices are available and overall pool health.
1559 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1561 nvlist_t *tryconfig, *config;
1564 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1565 zc->zc_iflags, &tryconfig)) != 0)
1568 config = spa_tryimport(tryconfig);
1570 nvlist_free(tryconfig);
1573 return (SET_ERROR(EINVAL));
1575 error = put_nvlist(zc, config);
1576 nvlist_free(config);
1583 * zc_name name of the pool
1584 * zc_cookie scan func (pool_scan_func_t)
1587 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1592 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1595 if (zc->zc_cookie == POOL_SCAN_NONE)
1596 error = spa_scan_stop(spa);
1598 error = spa_scan(spa, zc->zc_cookie);
1600 spa_close(spa, FTAG);
1606 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1611 error = spa_open(zc->zc_name, &spa, FTAG);
1614 spa_close(spa, FTAG);
1620 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1625 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1628 if (zc->zc_cookie < spa_version(spa) ||
1629 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1630 spa_close(spa, FTAG);
1631 return (SET_ERROR(EINVAL));
1634 spa_upgrade(spa, zc->zc_cookie);
1635 spa_close(spa, FTAG);
1641 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1648 if ((size = zc->zc_history_len) == 0)
1649 return (SET_ERROR(EINVAL));
1651 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1654 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1655 spa_close(spa, FTAG);
1656 return (SET_ERROR(ENOTSUP));
1659 hist_buf = vmem_alloc(size, KM_SLEEP);
1660 if ((error = spa_history_get(spa, &zc->zc_history_offset,
1661 &zc->zc_history_len, hist_buf)) == 0) {
1662 error = ddi_copyout(hist_buf,
1663 (void *)(uintptr_t)zc->zc_history,
1664 zc->zc_history_len, zc->zc_iflags);
1667 spa_close(spa, FTAG);
1668 vmem_free(hist_buf, size);
1673 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1678 error = spa_open(zc->zc_name, &spa, FTAG);
1680 error = spa_change_guid(spa);
1681 spa_close(spa, FTAG);
1687 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1689 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1694 * zc_name name of filesystem
1695 * zc_obj object to find
1698 * zc_value name of object
1701 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1706 /* XXX reading from objset not owned */
1707 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1709 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1710 dmu_objset_rele(os, FTAG);
1711 return (SET_ERROR(EINVAL));
1713 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1714 sizeof (zc->zc_value));
1715 dmu_objset_rele(os, FTAG);
1722 * zc_name name of filesystem
1723 * zc_obj object to find
1726 * zc_stat stats on object
1727 * zc_value path to object
1730 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1735 /* XXX reading from objset not owned */
1736 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1738 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1739 dmu_objset_rele(os, FTAG);
1740 return (SET_ERROR(EINVAL));
1742 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1743 sizeof (zc->zc_value));
1744 dmu_objset_rele(os, FTAG);
1750 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1756 error = spa_open(zc->zc_name, &spa, FTAG);
1760 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1761 zc->zc_iflags, &config);
1763 error = spa_vdev_add(spa, config);
1764 nvlist_free(config);
1766 spa_close(spa, FTAG);
1772 * zc_name name of the pool
1773 * zc_nvlist_conf nvlist of devices to remove
1774 * zc_cookie to stop the remove?
1777 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1782 error = spa_open(zc->zc_name, &spa, FTAG);
1785 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1786 spa_close(spa, FTAG);
1791 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1795 vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1797 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1799 switch (zc->zc_cookie) {
1800 case VDEV_STATE_ONLINE:
1801 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1804 case VDEV_STATE_OFFLINE:
1805 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1808 case VDEV_STATE_FAULTED:
1809 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1810 zc->zc_obj != VDEV_AUX_EXTERNAL)
1811 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1813 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1816 case VDEV_STATE_DEGRADED:
1817 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1818 zc->zc_obj != VDEV_AUX_EXTERNAL)
1819 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1821 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1825 error = SET_ERROR(EINVAL);
1827 zc->zc_cookie = newstate;
1828 spa_close(spa, FTAG);
1833 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1836 int replacing = zc->zc_cookie;
1840 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1843 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1844 zc->zc_iflags, &config)) == 0) {
1845 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1846 nvlist_free(config);
1849 spa_close(spa, FTAG);
1854 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1859 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1862 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
1864 spa_close(spa, FTAG);
1869 zfs_ioc_vdev_split(zfs_cmd_t *zc)
1872 nvlist_t *config, *props = NULL;
1874 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
1876 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1879 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1880 zc->zc_iflags, &config))) {
1881 spa_close(spa, FTAG);
1885 if (zc->zc_nvlist_src_size != 0 && (error =
1886 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1887 zc->zc_iflags, &props))) {
1888 spa_close(spa, FTAG);
1889 nvlist_free(config);
1893 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
1895 spa_close(spa, FTAG);
1897 nvlist_free(config);
1904 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
1907 char *path = zc->zc_value;
1908 uint64_t guid = zc->zc_guid;
1911 error = spa_open(zc->zc_name, &spa, FTAG);
1915 error = spa_vdev_setpath(spa, guid, path);
1916 spa_close(spa, FTAG);
1921 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
1924 char *fru = zc->zc_value;
1925 uint64_t guid = zc->zc_guid;
1928 error = spa_open(zc->zc_name, &spa, FTAG);
1932 error = spa_vdev_setfru(spa, guid, fru);
1933 spa_close(spa, FTAG);
1938 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
1943 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1945 if (zc->zc_nvlist_dst != 0 &&
1946 (error = dsl_prop_get_all(os, &nv)) == 0) {
1947 dmu_objset_stats(os, nv);
1949 * NB: zvol_get_stats() will read the objset contents,
1950 * which we aren't supposed to do with a
1951 * DS_MODE_USER hold, because it could be
1952 * inconsistent. So this is a bit of a workaround...
1953 * XXX reading with out owning
1955 if (!zc->zc_objset_stats.dds_inconsistent &&
1956 dmu_objset_type(os) == DMU_OST_ZVOL) {
1957 error = zvol_get_stats(os, nv);
1963 error = put_nvlist(zc, nv);
1972 * zc_name name of filesystem
1973 * zc_nvlist_dst_size size of buffer for property nvlist
1976 * zc_objset_stats stats
1977 * zc_nvlist_dst property nvlist
1978 * zc_nvlist_dst_size size of property nvlist
1981 zfs_ioc_objset_stats(zfs_cmd_t *zc)
1986 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
1988 error = zfs_ioc_objset_stats_impl(zc, os);
1989 dmu_objset_rele(os, FTAG);
1997 * zc_name name of filesystem
1998 * zc_nvlist_dst_size size of buffer for property nvlist
2001 * zc_nvlist_dst received property nvlist
2002 * zc_nvlist_dst_size size of received property nvlist
2004 * Gets received properties (distinct from local properties on or after
2005 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2006 * local property values.
2009 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2015 * Without this check, we would return local property values if the
2016 * caller has not already received properties on or after
2017 * SPA_VERSION_RECVD_PROPS.
2019 if (!dsl_prop_get_hasrecvd(zc->zc_name))
2020 return (SET_ERROR(ENOTSUP));
2022 if (zc->zc_nvlist_dst != 0 &&
2023 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2024 error = put_nvlist(zc, nv);
2032 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2038 * zfs_get_zplprop() will either find a value or give us
2039 * the default value (if there is one).
2041 if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2043 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2049 * zc_name name of filesystem
2050 * zc_nvlist_dst_size size of buffer for zpl property nvlist
2053 * zc_nvlist_dst zpl property nvlist
2054 * zc_nvlist_dst_size size of zpl property nvlist
2057 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2062 /* XXX reading without owning */
2063 if ((err = dmu_objset_hold(zc->zc_name, FTAG, &os)))
2066 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2069 * NB: nvl_add_zplprop() will read the objset contents,
2070 * which we aren't supposed to do with a DS_MODE_USER
2071 * hold, because it could be inconsistent.
2073 if (zc->zc_nvlist_dst != 0 &&
2074 !zc->zc_objset_stats.dds_inconsistent &&
2075 dmu_objset_type(os) == DMU_OST_ZFS) {
2078 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2079 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2080 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2081 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2082 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2083 err = put_nvlist(zc, nv);
2086 err = SET_ERROR(ENOENT);
2088 dmu_objset_rele(os, FTAG);
2093 dataset_name_hidden(const char *name)
2096 * Skip over datasets that are not visible in this zone,
2097 * internal datasets (which have a $ in their name), and
2098 * temporary datasets (which have a % in their name).
2100 if (strchr(name, '$') != NULL)
2102 if (strchr(name, '%') != NULL)
2104 if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
2111 * zc_name name of filesystem
2112 * zc_cookie zap cursor
2113 * zc_nvlist_dst_size size of buffer for property nvlist
2116 * zc_name name of next filesystem
2117 * zc_cookie zap cursor
2118 * zc_objset_stats stats
2119 * zc_nvlist_dst property nvlist
2120 * zc_nvlist_dst_size size of property nvlist
2123 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2128 size_t orig_len = strlen(zc->zc_name);
2131 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os))) {
2132 if (error == ENOENT)
2133 error = SET_ERROR(ESRCH);
2137 p = strrchr(zc->zc_name, '/');
2138 if (p == NULL || p[1] != '\0')
2139 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2140 p = zc->zc_name + strlen(zc->zc_name);
2143 error = dmu_dir_list_next(os,
2144 sizeof (zc->zc_name) - (p - zc->zc_name), p,
2145 NULL, &zc->zc_cookie);
2146 if (error == ENOENT)
2147 error = SET_ERROR(ESRCH);
2148 } while (error == 0 && dataset_name_hidden(zc->zc_name));
2149 dmu_objset_rele(os, FTAG);
2152 * If it's an internal dataset (ie. with a '$' in its name),
2153 * don't try to get stats for it, otherwise we'll return ENOENT.
2155 if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2156 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2157 if (error == ENOENT) {
2158 /* We lost a race with destroy, get the next one. */
2159 zc->zc_name[orig_len] = '\0';
2168 * zc_name name of filesystem
2169 * zc_cookie zap cursor
2170 * zc_nvlist_dst_size size of buffer for property nvlist
2173 * zc_name name of next snapshot
2174 * zc_objset_stats stats
2175 * zc_nvlist_dst property nvlist
2176 * zc_nvlist_dst_size size of property nvlist
2179 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2184 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2186 return (error == ENOENT ? ESRCH : error);
2190 * A dataset name of maximum length cannot have any snapshots,
2191 * so exit immediately.
2193 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
2194 dmu_objset_rele(os, FTAG);
2195 return (SET_ERROR(ESRCH));
2198 error = dmu_snapshot_list_next(os,
2199 sizeof (zc->zc_name) - strlen(zc->zc_name),
2200 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2203 if (error == 0 && !zc->zc_simple) {
2205 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2207 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2211 error = dmu_objset_from_ds(ds, &ossnap);
2213 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2214 dsl_dataset_rele(ds, FTAG);
2216 } else if (error == ENOENT) {
2217 error = SET_ERROR(ESRCH);
2220 dmu_objset_rele(os, FTAG);
2221 /* if we failed, undo the @ that we tacked on to zc_name */
2223 *strchr(zc->zc_name, '@') = '\0';
2228 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2230 const char *propname = nvpair_name(pair);
2232 unsigned int vallen;
2235 zfs_userquota_prop_t type;
2241 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2243 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2244 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2246 return (SET_ERROR(EINVAL));
2250 * A correctly constructed propname is encoded as
2251 * userquota@<rid>-<domain>.
2253 if ((dash = strchr(propname, '-')) == NULL ||
2254 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2256 return (SET_ERROR(EINVAL));
2263 err = zfs_sb_hold(dsname, FTAG, &zsb, B_FALSE);
2265 err = zfs_set_userquota(zsb, type, domain, rid, quota);
2266 zfs_sb_rele(zsb, FTAG);
2273 * If the named property is one that has a special function to set its value,
2274 * return 0 on success and a positive error code on failure; otherwise if it is
2275 * not one of the special properties handled by this function, return -1.
2277 * XXX: It would be better for callers of the property interface if we handled
2278 * these special cases in dsl_prop.c (in the dsl layer).
2281 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2284 const char *propname = nvpair_name(pair);
2285 zfs_prop_t prop = zfs_name_to_prop(propname);
2289 if (prop == ZPROP_INVAL) {
2290 if (zfs_prop_userquota(propname))
2291 return (zfs_prop_set_userquota(dsname, pair));
2295 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2297 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2298 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2302 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2305 VERIFY(0 == nvpair_value_uint64(pair, &intval));
2308 case ZFS_PROP_QUOTA:
2309 err = dsl_dir_set_quota(dsname, source, intval);
2311 case ZFS_PROP_REFQUOTA:
2312 err = dsl_dataset_set_refquota(dsname, source, intval);
2314 case ZFS_PROP_RESERVATION:
2315 err = dsl_dir_set_reservation(dsname, source, intval);
2317 case ZFS_PROP_REFRESERVATION:
2318 err = dsl_dataset_set_refreservation(dsname, source, intval);
2320 case ZFS_PROP_VOLSIZE:
2321 err = zvol_set_volsize(dsname, intval);
2323 case ZFS_PROP_SNAPDEV:
2324 err = zvol_set_snapdev(dsname, intval);
2326 case ZFS_PROP_VERSION:
2330 if ((err = zfs_sb_hold(dsname, FTAG, &zsb, B_TRUE)) != 0)
2333 err = zfs_set_version(zsb, intval);
2334 zfs_sb_rele(zsb, FTAG);
2336 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2339 zc = kmem_zalloc(sizeof (zfs_cmd_t),
2340 KM_SLEEP | KM_NODEBUG);
2341 (void) strcpy(zc->zc_name, dsname);
2342 (void) zfs_ioc_userspace_upgrade(zc);
2343 kmem_free(zc, sizeof (zfs_cmd_t));
2347 case ZFS_PROP_COMPRESSION:
2349 if (intval == ZIO_COMPRESS_LZ4) {
2352 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
2356 * Setting the LZ4 compression algorithm activates
2359 if (!spa_feature_is_active(spa,
2360 SPA_FEATURE_LZ4_COMPRESS)) {
2361 if ((err = zfs_prop_activate_feature(spa,
2362 SPA_FEATURE_LZ4_COMPRESS)) != 0) {
2363 spa_close(spa, FTAG);
2368 spa_close(spa, FTAG);
2371 * We still want the default set action to be performed in the
2372 * caller, we only performed zfeature settings here.
2386 * This function is best effort. If it fails to set any of the given properties,
2387 * it continues to set as many as it can and returns the last error
2388 * encountered. If the caller provides a non-NULL errlist, it will be filled in
2389 * with the list of names of all the properties that failed along with the
2390 * corresponding error numbers.
2392 * If every property is set successfully, zero is returned and errlist is not
2396 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2405 nvlist_t *genericnvl = fnvlist_alloc();
2406 nvlist_t *retrynvl = fnvlist_alloc();
2409 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2410 const char *propname = nvpair_name(pair);
2411 zfs_prop_t prop = zfs_name_to_prop(propname);
2414 /* decode the property value */
2416 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2418 attrs = fnvpair_value_nvlist(pair);
2419 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2421 err = SET_ERROR(EINVAL);
2424 /* Validate value type */
2425 if (err == 0 && prop == ZPROP_INVAL) {
2426 if (zfs_prop_user(propname)) {
2427 if (nvpair_type(propval) != DATA_TYPE_STRING)
2428 err = SET_ERROR(EINVAL);
2429 } else if (zfs_prop_userquota(propname)) {
2430 if (nvpair_type(propval) !=
2431 DATA_TYPE_UINT64_ARRAY)
2432 err = SET_ERROR(EINVAL);
2434 err = SET_ERROR(EINVAL);
2436 } else if (err == 0) {
2437 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2438 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2439 err = SET_ERROR(EINVAL);
2440 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2443 intval = fnvpair_value_uint64(propval);
2445 switch (zfs_prop_get_type(prop)) {
2446 case PROP_TYPE_NUMBER:
2448 case PROP_TYPE_STRING:
2449 err = SET_ERROR(EINVAL);
2451 case PROP_TYPE_INDEX:
2452 if (zfs_prop_index_to_string(prop,
2453 intval, &unused) != 0)
2454 err = SET_ERROR(EINVAL);
2458 "unknown property type");
2461 err = SET_ERROR(EINVAL);
2465 /* Validate permissions */
2467 err = zfs_check_settable(dsname, pair, CRED());
2470 err = zfs_prop_set_special(dsname, source, pair);
2473 * For better performance we build up a list of
2474 * properties to set in a single transaction.
2476 err = nvlist_add_nvpair(genericnvl, pair);
2477 } else if (err != 0 && nvl != retrynvl) {
2479 * This may be a spurious error caused by
2480 * receiving quota and reservation out of order.
2481 * Try again in a second pass.
2483 err = nvlist_add_nvpair(retrynvl, pair);
2488 if (errlist != NULL)
2489 fnvlist_add_int32(errlist, propname, err);
2494 if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2499 if (!nvlist_empty(genericnvl) &&
2500 dsl_props_set(dsname, source, genericnvl) != 0) {
2502 * If this fails, we still want to set as many properties as we
2503 * can, so try setting them individually.
2506 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2507 const char *propname = nvpair_name(pair);
2511 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2513 attrs = fnvpair_value_nvlist(pair);
2514 propval = fnvlist_lookup_nvpair(attrs,
2518 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2519 strval = fnvpair_value_string(propval);
2520 err = dsl_prop_set_string(dsname, propname,
2523 intval = fnvpair_value_uint64(propval);
2524 err = dsl_prop_set_int(dsname, propname, source,
2529 if (errlist != NULL) {
2530 fnvlist_add_int32(errlist, propname,
2537 nvlist_free(genericnvl);
2538 nvlist_free(retrynvl);
2544 * Check that all the properties are valid user properties.
2547 zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2549 nvpair_t *pair = NULL;
2552 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2553 const char *propname = nvpair_name(pair);
2556 if (!zfs_prop_user(propname) ||
2557 nvpair_type(pair) != DATA_TYPE_STRING)
2558 return (SET_ERROR(EINVAL));
2560 if ((error = zfs_secpolicy_write_perms(fsname,
2561 ZFS_DELEG_PERM_USERPROP, CRED())))
2564 if (strlen(propname) >= ZAP_MAXNAMELEN)
2565 return (SET_ERROR(ENAMETOOLONG));
2567 VERIFY(nvpair_value_string(pair, &valstr) == 0);
2568 if (strlen(valstr) >= ZAP_MAXVALUELEN)
2569 return (SET_ERROR(E2BIG));
2575 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2579 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2582 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2583 if (nvlist_exists(skipped, nvpair_name(pair)))
2586 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2591 clear_received_props(const char *dsname, nvlist_t *props,
2595 nvlist_t *cleared_props = NULL;
2596 props_skip(props, skipped, &cleared_props);
2597 if (!nvlist_empty(cleared_props)) {
2599 * Acts on local properties until the dataset has received
2600 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2602 zprop_source_t flags = (ZPROP_SRC_NONE |
2603 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2604 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2606 nvlist_free(cleared_props);
2612 * zc_name name of filesystem
2613 * zc_value name of property to set
2614 * zc_nvlist_src{_size} nvlist of properties to apply
2615 * zc_cookie received properties flag
2618 * zc_nvlist_dst{_size} error for each unapplied received property
2621 zfs_ioc_set_prop(zfs_cmd_t *zc)
2624 boolean_t received = zc->zc_cookie;
2625 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2630 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2631 zc->zc_iflags, &nvl)) != 0)
2635 nvlist_t *origprops;
2637 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2638 (void) clear_received_props(zc->zc_name,
2640 nvlist_free(origprops);
2643 error = dsl_prop_set_hasrecvd(zc->zc_name);
2646 errors = fnvlist_alloc();
2648 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2650 if (zc->zc_nvlist_dst != 0 && errors != NULL) {
2651 (void) put_nvlist(zc, errors);
2654 nvlist_free(errors);
2661 * zc_name name of filesystem
2662 * zc_value name of property to inherit
2663 * zc_cookie revert to received value if TRUE
2668 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2670 const char *propname = zc->zc_value;
2671 zfs_prop_t prop = zfs_name_to_prop(propname);
2672 boolean_t received = zc->zc_cookie;
2673 zprop_source_t source = (received
2674 ? ZPROP_SRC_NONE /* revert to received value, if any */
2675 : ZPROP_SRC_INHERITED); /* explicitly inherit */
2684 * zfs_prop_set_special() expects properties in the form of an
2685 * nvpair with type info.
2687 if (prop == ZPROP_INVAL) {
2688 if (!zfs_prop_user(propname))
2689 return (SET_ERROR(EINVAL));
2691 type = PROP_TYPE_STRING;
2692 } else if (prop == ZFS_PROP_VOLSIZE ||
2693 prop == ZFS_PROP_VERSION) {
2694 return (SET_ERROR(EINVAL));
2696 type = zfs_prop_get_type(prop);
2699 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2702 case PROP_TYPE_STRING:
2703 VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2705 case PROP_TYPE_NUMBER:
2706 case PROP_TYPE_INDEX:
2707 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2711 return (SET_ERROR(EINVAL));
2714 pair = nvlist_next_nvpair(dummy, NULL);
2715 err = zfs_prop_set_special(zc->zc_name, source, pair);
2718 return (err); /* special property already handled */
2721 * Only check this in the non-received case. We want to allow
2722 * 'inherit -S' to revert non-inheritable properties like quota
2723 * and reservation to the received or default values even though
2724 * they are not considered inheritable.
2726 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2727 return (SET_ERROR(EINVAL));
2730 /* property name has been validated by zfs_secpolicy_inherit_prop() */
2731 return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source));
2735 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2742 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2743 zc->zc_iflags, &props)))
2747 * If the only property is the configfile, then just do a spa_lookup()
2748 * to handle the faulted case.
2750 pair = nvlist_next_nvpair(props, NULL);
2751 if (pair != NULL && strcmp(nvpair_name(pair),
2752 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2753 nvlist_next_nvpair(props, pair) == NULL) {
2754 mutex_enter(&spa_namespace_lock);
2755 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2756 spa_configfile_set(spa, props, B_FALSE);
2757 spa_config_sync(spa, B_FALSE, B_TRUE);
2759 mutex_exit(&spa_namespace_lock);
2766 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2771 error = spa_prop_set(spa, props);
2774 spa_close(spa, FTAG);
2780 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2784 nvlist_t *nvp = NULL;
2786 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2788 * If the pool is faulted, there may be properties we can still
2789 * get (such as altroot and cachefile), so attempt to get them
2792 mutex_enter(&spa_namespace_lock);
2793 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2794 error = spa_prop_get(spa, &nvp);
2795 mutex_exit(&spa_namespace_lock);
2797 error = spa_prop_get(spa, &nvp);
2798 spa_close(spa, FTAG);
2801 if (error == 0 && zc->zc_nvlist_dst != 0)
2802 error = put_nvlist(zc, nvp);
2804 error = SET_ERROR(EFAULT);
2812 * zc_name name of filesystem
2813 * zc_nvlist_src{_size} nvlist of delegated permissions
2814 * zc_perm_action allow/unallow flag
2819 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2822 nvlist_t *fsaclnv = NULL;
2824 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2825 zc->zc_iflags, &fsaclnv)) != 0)
2829 * Verify nvlist is constructed correctly
2831 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2832 nvlist_free(fsaclnv);
2833 return (SET_ERROR(EINVAL));
2837 * If we don't have PRIV_SYS_MOUNT, then validate
2838 * that user is allowed to hand out each permission in
2842 error = secpolicy_zfs(CRED());
2844 if (zc->zc_perm_action == B_FALSE) {
2845 error = dsl_deleg_can_allow(zc->zc_name,
2848 error = dsl_deleg_can_unallow(zc->zc_name,
2854 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2856 nvlist_free(fsaclnv);
2862 * zc_name name of filesystem
2865 * zc_nvlist_src{_size} nvlist of delegated permissions
2868 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
2873 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
2874 error = put_nvlist(zc, nvp);
2883 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
2885 zfs_creat_t *zct = arg;
2887 zfs_create_fs(os, cr, zct->zct_zplprops, tx);
2890 #define ZFS_PROP_UNDEFINED ((uint64_t)-1)
2894 * os parent objset pointer (NULL if root fs)
2895 * fuids_ok fuids allowed in this version of the spa?
2896 * sa_ok SAs allowed in this version of the spa?
2897 * createprops list of properties requested by creator
2900 * zplprops values for the zplprops we attach to the master node object
2901 * is_ci true if requested file system will be purely case-insensitive
2903 * Determine the settings for utf8only, normalization and
2904 * casesensitivity. Specific values may have been requested by the
2905 * creator and/or we can inherit values from the parent dataset. If
2906 * the file system is of too early a vintage, a creator can not
2907 * request settings for these properties, even if the requested
2908 * setting is the default value. We don't actually want to create dsl
2909 * properties for these, so remove them from the source nvlist after
2913 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
2914 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
2915 nvlist_t *zplprops, boolean_t *is_ci)
2917 uint64_t sense = ZFS_PROP_UNDEFINED;
2918 uint64_t norm = ZFS_PROP_UNDEFINED;
2919 uint64_t u8 = ZFS_PROP_UNDEFINED;
2922 ASSERT(zplprops != NULL);
2925 * Pull out creator prop choices, if any.
2928 (void) nvlist_lookup_uint64(createprops,
2929 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
2930 (void) nvlist_lookup_uint64(createprops,
2931 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
2932 (void) nvlist_remove_all(createprops,
2933 zfs_prop_to_name(ZFS_PROP_NORMALIZE));
2934 (void) nvlist_lookup_uint64(createprops,
2935 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
2936 (void) nvlist_remove_all(createprops,
2937 zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
2938 (void) nvlist_lookup_uint64(createprops,
2939 zfs_prop_to_name(ZFS_PROP_CASE), &sense);
2940 (void) nvlist_remove_all(createprops,
2941 zfs_prop_to_name(ZFS_PROP_CASE));
2945 * If the zpl version requested is whacky or the file system
2946 * or pool is version is too "young" to support normalization
2947 * and the creator tried to set a value for one of the props,
2950 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
2951 (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
2952 (zplver >= ZPL_VERSION_SA && !sa_ok) ||
2953 (zplver < ZPL_VERSION_NORMALIZATION &&
2954 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
2955 sense != ZFS_PROP_UNDEFINED)))
2956 return (SET_ERROR(ENOTSUP));
2959 * Put the version in the zplprops
2961 VERIFY(nvlist_add_uint64(zplprops,
2962 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
2964 if (norm == ZFS_PROP_UNDEFINED &&
2965 (error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm)) != 0)
2967 VERIFY(nvlist_add_uint64(zplprops,
2968 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
2971 * If we're normalizing, names must always be valid UTF-8 strings.
2975 if (u8 == ZFS_PROP_UNDEFINED &&
2976 (error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8)) != 0)
2978 VERIFY(nvlist_add_uint64(zplprops,
2979 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
2981 if (sense == ZFS_PROP_UNDEFINED &&
2982 (error = zfs_get_zplprop(os, ZFS_PROP_CASE, &sense)) != 0)
2984 VERIFY(nvlist_add_uint64(zplprops,
2985 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
2988 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
2994 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
2995 nvlist_t *zplprops, boolean_t *is_ci)
2997 boolean_t fuids_ok, sa_ok;
2998 uint64_t zplver = ZPL_VERSION;
2999 objset_t *os = NULL;
3000 char parentname[MAXNAMELEN];
3006 (void) strlcpy(parentname, dataset, sizeof (parentname));
3007 cp = strrchr(parentname, '/');
3011 if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3014 spa_vers = spa_version(spa);
3015 spa_close(spa, FTAG);
3017 zplver = zfs_zpl_version_map(spa_vers);
3018 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3019 sa_ok = (zplver >= ZPL_VERSION_SA);
3022 * Open parent object set so we can inherit zplprop values.
3024 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3027 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3029 dmu_objset_rele(os, FTAG);
3034 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3035 nvlist_t *zplprops, boolean_t *is_ci)
3039 uint64_t zplver = ZPL_VERSION;
3042 zplver = zfs_zpl_version_map(spa_vers);
3043 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3044 sa_ok = (zplver >= ZPL_VERSION_SA);
3046 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3047 createprops, zplprops, is_ci);
3053 * "type" -> dmu_objset_type_t (int32)
3054 * (optional) "props" -> { prop -> value }
3057 * outnvl: propname -> error code (int32)
3060 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3063 zfs_creat_t zct = { 0 };
3064 nvlist_t *nvprops = NULL;
3065 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3067 dmu_objset_type_t type;
3068 boolean_t is_insensitive = B_FALSE;
3070 if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3071 return (SET_ERROR(EINVAL));
3073 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3077 cbfunc = zfs_create_cb;
3081 cbfunc = zvol_create_cb;
3088 if (strchr(fsname, '@') ||
3089 strchr(fsname, '%'))
3090 return (SET_ERROR(EINVAL));
3092 zct.zct_props = nvprops;
3095 return (SET_ERROR(EINVAL));
3097 if (type == DMU_OST_ZVOL) {
3098 uint64_t volsize, volblocksize;
3100 if (nvprops == NULL)
3101 return (SET_ERROR(EINVAL));
3102 if (nvlist_lookup_uint64(nvprops,
3103 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3104 return (SET_ERROR(EINVAL));
3106 if ((error = nvlist_lookup_uint64(nvprops,
3107 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3108 &volblocksize)) != 0 && error != ENOENT)
3109 return (SET_ERROR(EINVAL));
3112 volblocksize = zfs_prop_default_numeric(
3113 ZFS_PROP_VOLBLOCKSIZE);
3115 if ((error = zvol_check_volblocksize(
3116 volblocksize)) != 0 ||
3117 (error = zvol_check_volsize(volsize,
3118 volblocksize)) != 0)
3120 } else if (type == DMU_OST_ZFS) {
3124 * We have to have normalization and
3125 * case-folding flags correct when we do the
3126 * file system creation, so go figure them out
3129 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3130 NV_UNIQUE_NAME, KM_SLEEP) == 0);
3131 error = zfs_fill_zplprops(fsname, nvprops,
3132 zct.zct_zplprops, &is_insensitive);
3134 nvlist_free(zct.zct_zplprops);
3139 error = dmu_objset_create(fsname, type,
3140 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3141 nvlist_free(zct.zct_zplprops);
3144 * It would be nice to do this atomically.
3147 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3150 (void) dsl_destroy_head(fsname);
3154 if (error == 0 && type == DMU_OST_ZVOL)
3155 zvol_create_minors(fsname);
3163 * "origin" -> name of origin snapshot
3164 * (optional) "props" -> { prop -> value }
3168 * outnvl: propname -> error code (int32)
3171 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3174 nvlist_t *nvprops = NULL;
3177 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3178 return (SET_ERROR(EINVAL));
3179 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3181 if (strchr(fsname, '@') ||
3182 strchr(fsname, '%'))
3183 return (SET_ERROR(EINVAL));
3185 if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3186 return (SET_ERROR(EINVAL));
3187 error = dmu_objset_clone(fsname, origin_name);
3192 * It would be nice to do this atomically.
3195 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3198 (void) dsl_destroy_head(fsname);
3203 zvol_create_minors(fsname);
3211 * "snaps" -> { snapshot1, snapshot2 }
3212 * (optional) "props" -> { prop -> value (string) }
3215 * outnvl: snapshot -> error code (int32)
3218 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3221 nvlist_t *props = NULL;
3223 nvpair_t *pair, *pair2;
3225 (void) nvlist_lookup_nvlist(innvl, "props", &props);
3226 if ((error = zfs_check_userprops(poolname, props)) != 0)
3229 if (!nvlist_empty(props) &&
3230 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3231 return (SET_ERROR(ENOTSUP));
3233 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3234 return (SET_ERROR(EINVAL));
3235 poollen = strlen(poolname);
3236 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3237 pair = nvlist_next_nvpair(snaps, pair)) {
3238 const char *name = nvpair_name(pair);
3239 const char *cp = strchr(name, '@');
3242 * The snap name must contain an @, and the part after it must
3243 * contain only valid characters.
3245 if (cp == NULL || snapshot_namecheck(cp + 1, NULL, NULL) != 0)
3246 return (SET_ERROR(EINVAL));
3249 * The snap must be in the specified pool.
3251 if (strncmp(name, poolname, poollen) != 0 ||
3252 (name[poollen] != '/' && name[poollen] != '@'))
3253 return (SET_ERROR(EXDEV));
3255 /* This must be the only snap of this fs. */
3256 for (pair2 = nvlist_next_nvpair(snaps, pair);
3257 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3258 if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3260 return (SET_ERROR(EXDEV));
3265 error = dsl_dataset_snapshot(snaps, props, outnvl);
3269 zvol_create_minors(poolname);
3276 * innvl: "message" -> string
3280 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3288 * The poolname in the ioctl is not set, we get it from the TSD,
3289 * which was set at the end of the last successful ioctl that allows
3290 * logging. The secpolicy func already checked that it is set.
3291 * Only one log ioctl is allowed after each successful ioctl, so
3292 * we clear the TSD here.
3294 poolname = tsd_get(zfs_allow_log_key);
3295 (void) tsd_set(zfs_allow_log_key, NULL);
3296 error = spa_open(poolname, &spa, FTAG);
3301 if (nvlist_lookup_string(innvl, "message", &message) != 0) {
3302 spa_close(spa, FTAG);
3303 return (SET_ERROR(EINVAL));
3306 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3307 spa_close(spa, FTAG);
3308 return (SET_ERROR(ENOTSUP));
3311 error = spa_history_log(spa, message);
3312 spa_close(spa, FTAG);
3317 * The dp_config_rwlock must not be held when calling this, because the
3318 * unmount may need to write out data.
3320 * This function is best-effort. Callers must deal gracefully if it
3321 * remains mounted (or is remounted after this call).
3323 * XXX: This function should detect a failure to unmount a snapdir of a dataset
3324 * and return the appropriate error code when it is mounted. Its Illumos and
3325 * FreeBSD counterparts do this. We do not do this on Linux because there is no
3326 * clear way to access the mount information that FreeBSD and Illumos use to
3327 * distinguish between things with mounted snapshot directories, and things
3328 * without mounted snapshot directories, which include zvols. Returning a
3329 * failure for the latter causes `zfs destroy` to fail on zvol snapshots.
3332 zfs_unmount_snap(const char *snapname)
3334 zfs_sb_t *zsb = NULL;
3339 if ((ptr = strchr(snapname, '@')) == NULL)
3342 dsname = kmem_alloc(ptr - snapname + 1, KM_SLEEP);
3343 strlcpy(dsname, snapname, ptr - snapname + 1);
3344 fullname = strdup(snapname);
3346 if (zfs_sb_hold(dsname, FTAG, &zsb, B_FALSE) == 0) {
3347 ASSERT(!dsl_pool_config_held(dmu_objset_pool(zsb->z_os)));
3348 (void) zfsctl_unmount_snapshot(zsb, fullname, MNT_FORCE);
3349 zfs_sb_rele(zsb, FTAG);
3352 kmem_free(dsname, ptr - snapname + 1);
3360 zfs_unmount_snap_cb(const char *snapname, void *arg)
3362 return (zfs_unmount_snap(snapname));
3366 * When a clone is destroyed, its origin may also need to be destroyed,
3367 * in which case it must be unmounted. This routine will do that unmount
3371 zfs_destroy_unmount_origin(const char *fsname)
3377 error = dmu_objset_hold(fsname, FTAG, &os);
3380 ds = dmu_objset_ds(os);
3381 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3382 char originname[MAXNAMELEN];
3383 dsl_dataset_name(ds->ds_prev, originname);
3384 dmu_objset_rele(os, FTAG);
3385 (void) zfs_unmount_snap(originname);
3387 dmu_objset_rele(os, FTAG);
3393 * "snaps" -> { snapshot1, snapshot2 }
3394 * (optional boolean) "defer"
3397 * outnvl: snapshot -> error code (int32)
3400 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3407 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3408 return (SET_ERROR(EINVAL));
3409 defer = nvlist_exists(innvl, "defer");
3411 poollen = strlen(poolname);
3412 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3413 pair = nvlist_next_nvpair(snaps, pair)) {
3414 const char *name = nvpair_name(pair);
3417 * The snap must be in the specified pool.
3419 if (strncmp(name, poolname, poollen) != 0 ||
3420 (name[poollen] != '/' && name[poollen] != '@'))
3421 return (SET_ERROR(EXDEV));
3423 error = zfs_unmount_snap(name);
3426 (void) zvol_remove_minor(name);
3429 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3434 * zc_name name of dataset to destroy
3435 * zc_objset_type type of objset
3436 * zc_defer_destroy mark for deferred destroy
3441 zfs_ioc_destroy(zfs_cmd_t *zc)
3445 if (zc->zc_objset_type == DMU_OST_ZFS) {
3446 err = zfs_unmount_snap(zc->zc_name);
3451 if (strchr(zc->zc_name, '@'))
3452 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3454 err = dsl_destroy_head(zc->zc_name);
3455 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3456 (void) zvol_remove_minor(zc->zc_name);
3461 * fsname is name of dataset to rollback (to most recent snapshot)
3463 * innvl is not used.
3465 * outnvl: "target" -> name of most recent snapshot
3470 zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl)
3475 if (get_zfs_sb(fsname, &zsb) == 0) {
3476 error = zfs_suspend_fs(zsb);
3480 error = dsl_dataset_rollback(fsname, zsb, outnvl);
3481 resume_err = zfs_resume_fs(zsb, fsname);
3482 error = error ? error : resume_err;
3484 deactivate_super(zsb->z_sb);
3486 error = dsl_dataset_rollback(fsname, NULL, outnvl);
3492 recursive_unmount(const char *fsname, void *arg)
3494 const char *snapname = arg;
3498 fullname = kmem_asprintf("%s@%s", fsname, snapname);
3499 error = zfs_unmount_snap(fullname);
3507 * zc_name old name of dataset
3508 * zc_value new name of dataset
3509 * zc_cookie recursive flag (only valid for snapshots)
3514 zfs_ioc_rename(zfs_cmd_t *zc)
3516 boolean_t recursive = zc->zc_cookie & 1;
3519 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3520 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3521 strchr(zc->zc_value, '%'))
3522 return (SET_ERROR(EINVAL));
3524 at = strchr(zc->zc_name, '@');
3526 /* snaps must be in same fs */
3529 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
3530 return (SET_ERROR(EXDEV));
3532 if (zc->zc_objset_type == DMU_OST_ZFS) {
3533 error = dmu_objset_find(zc->zc_name,
3534 recursive_unmount, at + 1,
3535 recursive ? DS_FIND_CHILDREN : 0);
3541 error = dsl_dataset_rename_snapshot(zc->zc_name,
3542 at + 1, strchr(zc->zc_value, '@') + 1, recursive);
3547 return (dsl_dir_rename(zc->zc_name, zc->zc_value));
3552 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3554 const char *propname = nvpair_name(pair);
3555 boolean_t issnap = (strchr(dsname, '@') != NULL);
3556 zfs_prop_t prop = zfs_name_to_prop(propname);
3560 if (prop == ZPROP_INVAL) {
3561 if (zfs_prop_user(propname)) {
3562 if ((err = zfs_secpolicy_write_perms(dsname,
3563 ZFS_DELEG_PERM_USERPROP, cr)))
3568 if (!issnap && zfs_prop_userquota(propname)) {
3569 const char *perm = NULL;
3570 const char *uq_prefix =
3571 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3572 const char *gq_prefix =
3573 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3575 if (strncmp(propname, uq_prefix,
3576 strlen(uq_prefix)) == 0) {
3577 perm = ZFS_DELEG_PERM_USERQUOTA;
3578 } else if (strncmp(propname, gq_prefix,
3579 strlen(gq_prefix)) == 0) {
3580 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3582 /* USERUSED and GROUPUSED are read-only */
3583 return (SET_ERROR(EINVAL));
3586 if ((err = zfs_secpolicy_write_perms(dsname, perm, cr)))
3591 return (SET_ERROR(EINVAL));
3595 return (SET_ERROR(EINVAL));
3597 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3599 * dsl_prop_get_all_impl() returns properties in this
3603 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3604 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3609 * Check that this value is valid for this pool version
3612 case ZFS_PROP_COMPRESSION:
3614 * If the user specified gzip compression, make sure
3615 * the SPA supports it. We ignore any errors here since
3616 * we'll catch them later.
3618 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3619 nvpair_value_uint64(pair, &intval) == 0) {
3620 if (intval >= ZIO_COMPRESS_GZIP_1 &&
3621 intval <= ZIO_COMPRESS_GZIP_9 &&
3622 zfs_earlier_version(dsname,
3623 SPA_VERSION_GZIP_COMPRESSION)) {
3624 return (SET_ERROR(ENOTSUP));
3627 if (intval == ZIO_COMPRESS_ZLE &&
3628 zfs_earlier_version(dsname,
3629 SPA_VERSION_ZLE_COMPRESSION))
3630 return (SET_ERROR(ENOTSUP));
3632 if (intval == ZIO_COMPRESS_LZ4) {
3635 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3638 if (!spa_feature_is_enabled(spa,
3639 SPA_FEATURE_LZ4_COMPRESS)) {
3640 spa_close(spa, FTAG);
3641 return (SET_ERROR(ENOTSUP));
3643 spa_close(spa, FTAG);
3647 * If this is a bootable dataset then
3648 * verify that the compression algorithm
3649 * is supported for booting. We must return
3650 * something other than ENOTSUP since it
3651 * implies a downrev pool version.
3653 if (zfs_is_bootfs(dsname) &&
3654 !BOOTFS_COMPRESS_VALID(intval)) {
3655 return (SET_ERROR(ERANGE));
3660 case ZFS_PROP_COPIES:
3661 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3662 return (SET_ERROR(ENOTSUP));
3665 case ZFS_PROP_DEDUP:
3666 if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3667 return (SET_ERROR(ENOTSUP));
3670 case ZFS_PROP_SHARESMB:
3671 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3672 return (SET_ERROR(ENOTSUP));
3675 case ZFS_PROP_ACLINHERIT:
3676 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3677 nvpair_value_uint64(pair, &intval) == 0) {
3678 if (intval == ZFS_ACL_PASSTHROUGH_X &&
3679 zfs_earlier_version(dsname,
3680 SPA_VERSION_PASSTHROUGH_X))
3681 return (SET_ERROR(ENOTSUP));
3688 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3692 * Checks for a race condition to make sure we don't increment a feature flag
3696 zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx)
3698 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3699 spa_feature_t *featurep = arg;
3701 if (!spa_feature_is_active(spa, *featurep))
3704 return (SET_ERROR(EBUSY));
3708 * The callback invoked on feature activation in the sync task caused by
3709 * zfs_prop_activate_feature.
3712 zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx)
3714 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3715 spa_feature_t *featurep = arg;
3717 spa_feature_incr(spa, *featurep, tx);
3721 * Activates a feature on a pool in response to a property setting. This
3722 * creates a new sync task which modifies the pool to reflect the feature
3726 zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature)
3730 /* EBUSY here indicates that the feature is already active */
3731 err = dsl_sync_task(spa_name(spa),
3732 zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync,
3735 if (err != 0 && err != EBUSY)
3742 * Removes properties from the given props list that fail permission checks
3743 * needed to clear them and to restore them in case of a receive error. For each
3744 * property, make sure we have both set and inherit permissions.
3746 * Returns the first error encountered if any permission checks fail. If the
3747 * caller provides a non-NULL errlist, it also gives the complete list of names
3748 * of all the properties that failed a permission check along with the
3749 * corresponding error numbers. The caller is responsible for freeing the
3752 * If every property checks out successfully, zero is returned and the list
3753 * pointed at by errlist is NULL.
3756 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
3759 nvpair_t *pair, *next_pair;
3766 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3768 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP | KM_NODEBUG);
3769 (void) strcpy(zc->zc_name, dataset);
3770 pair = nvlist_next_nvpair(props, NULL);
3771 while (pair != NULL) {
3772 next_pair = nvlist_next_nvpair(props, pair);
3774 (void) strcpy(zc->zc_value, nvpair_name(pair));
3775 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
3776 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
3777 VERIFY(nvlist_remove_nvpair(props, pair) == 0);
3778 VERIFY(nvlist_add_int32(errors,
3779 zc->zc_value, err) == 0);
3783 kmem_free(zc, sizeof (zfs_cmd_t));
3785 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
3786 nvlist_free(errors);
3789 VERIFY(nvpair_value_int32(pair, &rv) == 0);
3792 if (errlist == NULL)
3793 nvlist_free(errors);
3801 propval_equals(nvpair_t *p1, nvpair_t *p2)
3803 if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
3804 /* dsl_prop_get_all_impl() format */
3806 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
3807 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3811 if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
3813 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
3814 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3818 if (nvpair_type(p1) != nvpair_type(p2))
3821 if (nvpair_type(p1) == DATA_TYPE_STRING) {
3822 char *valstr1, *valstr2;
3824 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
3825 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
3826 return (strcmp(valstr1, valstr2) == 0);
3828 uint64_t intval1, intval2;
3830 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
3831 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
3832 return (intval1 == intval2);
3837 * Remove properties from props if they are not going to change (as determined
3838 * by comparison with origprops). Remove them from origprops as well, since we
3839 * do not need to clear or restore properties that won't change.
3842 props_reduce(nvlist_t *props, nvlist_t *origprops)
3844 nvpair_t *pair, *next_pair;
3846 if (origprops == NULL)
3847 return; /* all props need to be received */
3849 pair = nvlist_next_nvpair(props, NULL);
3850 while (pair != NULL) {
3851 const char *propname = nvpair_name(pair);
3854 next_pair = nvlist_next_nvpair(props, pair);
3856 if ((nvlist_lookup_nvpair(origprops, propname,
3857 &match) != 0) || !propval_equals(pair, match))
3858 goto next; /* need to set received value */
3860 /* don't clear the existing received value */
3861 (void) nvlist_remove_nvpair(origprops, match);
3862 /* don't bother receiving the property */
3863 (void) nvlist_remove_nvpair(props, pair);
3870 static boolean_t zfs_ioc_recv_inject_err;
3875 * zc_name name of containing filesystem
3876 * zc_nvlist_src{_size} nvlist of properties to apply
3877 * zc_value name of snapshot to create
3878 * zc_string name of clone origin (if DRR_FLAG_CLONE)
3879 * zc_cookie file descriptor to recv from
3880 * zc_begin_record the BEGIN record of the stream (not byteswapped)
3881 * zc_guid force flag
3882 * zc_cleanup_fd cleanup-on-exit file descriptor
3883 * zc_action_handle handle for this guid/ds mapping (or zero on first call)
3886 * zc_cookie number of bytes read
3887 * zc_nvlist_dst{_size} error for each unapplied received property
3888 * zc_obj zprop_errflags_t
3889 * zc_action_handle handle for this guid/ds mapping
3892 zfs_ioc_recv(zfs_cmd_t *zc)
3895 dmu_recv_cookie_t drc;
3896 boolean_t force = (boolean_t)zc->zc_guid;
3899 int props_error = 0;
3902 nvlist_t *props = NULL; /* sent properties */
3903 nvlist_t *origprops = NULL; /* existing properties */
3904 char *origin = NULL;
3906 char tofs[ZFS_MAXNAMELEN];
3907 boolean_t first_recvd_props = B_FALSE;
3909 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3910 strchr(zc->zc_value, '@') == NULL ||
3911 strchr(zc->zc_value, '%'))
3912 return (SET_ERROR(EINVAL));
3914 (void) strcpy(tofs, zc->zc_value);
3915 tosnap = strchr(tofs, '@');
3918 if (zc->zc_nvlist_src != 0 &&
3919 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
3920 zc->zc_iflags, &props)) != 0)
3927 return (SET_ERROR(EBADF));
3930 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3932 if (zc->zc_string[0])
3933 origin = zc->zc_string;
3935 error = dmu_recv_begin(tofs, tosnap,
3936 &zc->zc_begin_record, force, origin, &drc);
3941 * Set properties before we receive the stream so that they are applied
3942 * to the new data. Note that we must call dmu_recv_stream() if
3943 * dmu_recv_begin() succeeds.
3945 if (props != NULL && !drc.drc_newfs) {
3946 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
3947 SPA_VERSION_RECVD_PROPS &&
3948 !dsl_prop_get_hasrecvd(tofs))
3949 first_recvd_props = B_TRUE;
3952 * If new received properties are supplied, they are to
3953 * completely replace the existing received properties, so stash
3954 * away the existing ones.
3956 if (dsl_prop_get_received(tofs, &origprops) == 0) {
3957 nvlist_t *errlist = NULL;
3959 * Don't bother writing a property if its value won't
3960 * change (and avoid the unnecessary security checks).
3962 * The first receive after SPA_VERSION_RECVD_PROPS is a
3963 * special case where we blow away all local properties
3966 if (!first_recvd_props)
3967 props_reduce(props, origprops);
3968 if (zfs_check_clearable(tofs, origprops, &errlist) != 0)
3969 (void) nvlist_merge(errors, errlist, 0);
3970 nvlist_free(errlist);
3972 if (clear_received_props(tofs, origprops,
3973 first_recvd_props ? NULL : props) != 0)
3974 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
3976 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
3980 if (props != NULL) {
3981 props_error = dsl_prop_set_hasrecvd(tofs);
3983 if (props_error == 0) {
3984 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
3989 if (zc->zc_nvlist_dst_size != 0 &&
3990 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
3991 put_nvlist(zc, errors) != 0)) {
3993 * Caller made zc->zc_nvlist_dst less than the minimum expected
3994 * size or supplied an invalid address.
3996 props_error = SET_ERROR(EINVAL);
4000 error = dmu_recv_stream(&drc, fp->f_vnode, &off, zc->zc_cleanup_fd,
4001 &zc->zc_action_handle);
4004 zfs_sb_t *zsb = NULL;
4006 if (get_zfs_sb(tofs, &zsb) == 0) {
4010 error = zfs_suspend_fs(zsb);
4012 * If the suspend fails, then the recv_end will
4013 * likely also fail, and clean up after itself.
4015 end_err = dmu_recv_end(&drc, zsb);
4017 error = zfs_resume_fs(zsb, tofs);
4018 error = error ? error : end_err;
4019 deactivate_super(zsb->z_sb);
4021 error = dmu_recv_end(&drc, NULL);
4025 zc->zc_cookie = off - fp->f_offset;
4026 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4030 if (zfs_ioc_recv_inject_err) {
4031 zfs_ioc_recv_inject_err = B_FALSE;
4038 zvol_create_minors(tofs);
4042 * On error, restore the original props.
4044 if (error != 0 && props != NULL && !drc.drc_newfs) {
4045 if (clear_received_props(tofs, props, NULL) != 0) {
4047 * We failed to clear the received properties.
4048 * Since we may have left a $recvd value on the
4049 * system, we can't clear the $hasrecvd flag.
4051 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4052 } else if (first_recvd_props) {
4053 dsl_prop_unset_hasrecvd(tofs);
4056 if (origprops == NULL && !drc.drc_newfs) {
4057 /* We failed to stash the original properties. */
4058 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4062 * dsl_props_set() will not convert RECEIVED to LOCAL on or
4063 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4064 * explictly if we're restoring local properties cleared in the
4065 * first new-style receive.
4067 if (origprops != NULL &&
4068 zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4069 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4070 origprops, NULL) != 0) {
4072 * We stashed the original properties but failed to
4075 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4080 nvlist_free(origprops);
4081 nvlist_free(errors);
4085 error = props_error;
4092 * zc_name name of snapshot to send
4093 * zc_cookie file descriptor to send stream to
4094 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj)
4095 * zc_sendobj objsetid of snapshot to send
4096 * zc_fromobj objsetid of incremental fromsnap (may be zero)
4097 * zc_guid if set, estimate size of stream only. zc_cookie is ignored.
4098 * output size in zc_objset_type.
4103 zfs_ioc_send(zfs_cmd_t *zc)
4107 boolean_t estimate = (zc->zc_guid != 0);
4109 if (zc->zc_obj != 0) {
4111 dsl_dataset_t *tosnap;
4113 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4117 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4119 dsl_pool_rele(dp, FTAG);
4123 if (dsl_dir_is_clone(tosnap->ds_dir))
4124 zc->zc_fromobj = tosnap->ds_dir->dd_phys->dd_origin_obj;
4125 dsl_dataset_rele(tosnap, FTAG);
4126 dsl_pool_rele(dp, FTAG);
4131 dsl_dataset_t *tosnap;
4132 dsl_dataset_t *fromsnap = NULL;
4134 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4138 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4140 dsl_pool_rele(dp, FTAG);
4144 if (zc->zc_fromobj != 0) {
4145 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
4148 dsl_dataset_rele(tosnap, FTAG);
4149 dsl_pool_rele(dp, FTAG);
4154 error = dmu_send_estimate(tosnap, fromsnap,
4155 &zc->zc_objset_type);
4157 if (fromsnap != NULL)
4158 dsl_dataset_rele(fromsnap, FTAG);
4159 dsl_dataset_rele(tosnap, FTAG);
4160 dsl_pool_rele(dp, FTAG);
4162 file_t *fp = getf(zc->zc_cookie);
4164 return (SET_ERROR(EBADF));
4167 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
4168 zc->zc_fromobj, zc->zc_cookie, fp->f_vnode, &off);
4170 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4172 releasef(zc->zc_cookie);
4179 * zc_name name of snapshot on which to report progress
4180 * zc_cookie file descriptor of send stream
4183 * zc_cookie number of bytes written in send stream thus far
4186 zfs_ioc_send_progress(zfs_cmd_t *zc)
4190 dmu_sendarg_t *dsp = NULL;
4193 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4197 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4199 dsl_pool_rele(dp, FTAG);
4203 mutex_enter(&ds->ds_sendstream_lock);
4206 * Iterate over all the send streams currently active on this dataset.
4207 * If there's one which matches the specified file descriptor _and_ the
4208 * stream was started by the current process, return the progress of
4212 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4213 dsp = list_next(&ds->ds_sendstreams, dsp)) {
4214 if (dsp->dsa_outfd == zc->zc_cookie &&
4215 dsp->dsa_proc->group_leader == curproc->group_leader)
4220 zc->zc_cookie = *(dsp->dsa_off);
4222 error = SET_ERROR(ENOENT);
4224 mutex_exit(&ds->ds_sendstream_lock);
4225 dsl_dataset_rele(ds, FTAG);
4226 dsl_pool_rele(dp, FTAG);
4231 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4235 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4236 &zc->zc_inject_record);
4239 zc->zc_guid = (uint64_t)id;
4245 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4247 return (zio_clear_fault((int)zc->zc_guid));
4251 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4253 int id = (int)zc->zc_guid;
4256 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4257 &zc->zc_inject_record);
4265 zfs_ioc_error_log(zfs_cmd_t *zc)
4269 size_t count = (size_t)zc->zc_nvlist_dst_size;
4271 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4274 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4277 zc->zc_nvlist_dst_size = count;
4279 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4281 spa_close(spa, FTAG);
4287 zfs_ioc_clear(zfs_cmd_t *zc)
4294 * On zpool clear we also fix up missing slogs
4296 mutex_enter(&spa_namespace_lock);
4297 spa = spa_lookup(zc->zc_name);
4299 mutex_exit(&spa_namespace_lock);
4300 return (SET_ERROR(EIO));
4302 if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4303 /* we need to let spa_open/spa_load clear the chains */
4304 spa_set_log_state(spa, SPA_LOG_CLEAR);
4306 spa->spa_last_open_failed = 0;
4307 mutex_exit(&spa_namespace_lock);
4309 if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4310 error = spa_open(zc->zc_name, &spa, FTAG);
4313 nvlist_t *config = NULL;
4315 if (zc->zc_nvlist_src == 0)
4316 return (SET_ERROR(EINVAL));
4318 if ((error = get_nvlist(zc->zc_nvlist_src,
4319 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4320 error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4322 if (config != NULL) {
4325 if ((err = put_nvlist(zc, config)) != 0)
4327 nvlist_free(config);
4329 nvlist_free(policy);
4336 spa_vdev_state_enter(spa, SCL_NONE);
4338 if (zc->zc_guid == 0) {
4341 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4343 (void) spa_vdev_state_exit(spa, NULL, ENODEV);
4344 spa_close(spa, FTAG);
4345 return (SET_ERROR(ENODEV));
4349 vdev_clear(spa, vd);
4351 (void) spa_vdev_state_exit(spa, NULL, 0);
4354 * Resume any suspended I/Os.
4356 if (zio_resume(spa) != 0)
4357 error = SET_ERROR(EIO);
4359 spa_close(spa, FTAG);
4365 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4370 error = spa_open(zc->zc_name, &spa, FTAG);
4374 spa_vdev_state_enter(spa, SCL_NONE);
4377 * If a resilver is already in progress then set the
4378 * spa_scrub_reopen flag to B_TRUE so that we don't restart
4379 * the scan as a side effect of the reopen. Otherwise, let
4380 * vdev_open() decided if a resilver is required.
4382 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4383 vdev_reopen(spa->spa_root_vdev);
4384 spa->spa_scrub_reopen = B_FALSE;
4386 (void) spa_vdev_state_exit(spa, NULL, 0);
4387 spa_close(spa, FTAG);
4392 * zc_name name of filesystem
4393 * zc_value name of origin snapshot
4396 * zc_string name of conflicting snapshot, if there is one
4399 zfs_ioc_promote(zfs_cmd_t *zc)
4404 * We don't need to unmount *all* the origin fs's snapshots, but
4407 cp = strchr(zc->zc_value, '@');
4410 (void) dmu_objset_find(zc->zc_value,
4411 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
4412 return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4416 * Retrieve a single {user|group}{used|quota}@... property.
4419 * zc_name name of filesystem
4420 * zc_objset_type zfs_userquota_prop_t
4421 * zc_value domain name (eg. "S-1-234-567-89")
4422 * zc_guid RID/UID/GID
4425 * zc_cookie property value
4428 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4433 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4434 return (SET_ERROR(EINVAL));
4436 error = zfs_sb_hold(zc->zc_name, FTAG, &zsb, B_FALSE);
4440 error = zfs_userspace_one(zsb,
4441 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4442 zfs_sb_rele(zsb, FTAG);
4449 * zc_name name of filesystem
4450 * zc_cookie zap cursor
4451 * zc_objset_type zfs_userquota_prop_t
4452 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4455 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
4456 * zc_cookie zap cursor
4459 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4462 int bufsize = zc->zc_nvlist_dst_size;
4467 return (SET_ERROR(ENOMEM));
4469 error = zfs_sb_hold(zc->zc_name, FTAG, &zsb, B_FALSE);
4473 buf = vmem_alloc(bufsize, KM_SLEEP);
4475 error = zfs_userspace_many(zsb, zc->zc_objset_type, &zc->zc_cookie,
4476 buf, &zc->zc_nvlist_dst_size);
4479 error = xcopyout(buf,
4480 (void *)(uintptr_t)zc->zc_nvlist_dst,
4481 zc->zc_nvlist_dst_size);
4483 vmem_free(buf, bufsize);
4484 zfs_sb_rele(zsb, FTAG);
4491 * zc_name name of filesystem
4497 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4503 if (get_zfs_sb(zc->zc_name, &zsb) == 0) {
4504 if (!dmu_objset_userused_enabled(zsb->z_os)) {
4506 * If userused is not enabled, it may be because the
4507 * objset needs to be closed & reopened (to grow the
4508 * objset_phys_t). Suspend/resume the fs will do that.
4510 error = zfs_suspend_fs(zsb);
4512 dmu_objset_refresh_ownership(zsb->z_os,
4514 error = zfs_resume_fs(zsb, zc->zc_name);
4518 error = dmu_objset_userspace_upgrade(zsb->z_os);
4519 deactivate_super(zsb->z_sb);
4521 /* XXX kind of reading contents without owning */
4522 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4526 error = dmu_objset_userspace_upgrade(os);
4527 dmu_objset_rele(os, FTAG);
4534 zfs_ioc_share(zfs_cmd_t *zc)
4536 return (SET_ERROR(ENOSYS));
4539 ace_t full_access[] = {
4540 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4545 * zc_name name of containing filesystem
4546 * zc_obj object # beyond which we want next in-use object #
4549 * zc_obj next in-use object #
4552 zfs_ioc_next_obj(zfs_cmd_t *zc)
4554 objset_t *os = NULL;
4557 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4561 error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
4562 os->os_dsl_dataset->ds_phys->ds_prev_snap_txg);
4564 dmu_objset_rele(os, FTAG);
4570 * zc_name name of filesystem
4571 * zc_value prefix name for snapshot
4572 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
4575 * zc_value short name of new snapshot
4578 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
4585 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
4589 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
4590 (u_longlong_t)ddi_get_lbolt64());
4591 hold_name = kmem_asprintf("%%%s", zc->zc_value);
4593 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
4596 (void) strcpy(zc->zc_value, snap_name);
4599 zfs_onexit_fd_rele(zc->zc_cleanup_fd);
4605 * zc_name name of "to" snapshot
4606 * zc_value name of "from" snapshot
4607 * zc_cookie file descriptor to write diff data on
4610 * dmu_diff_record_t's to the file descriptor
4613 zfs_ioc_diff(zfs_cmd_t *zc)
4619 fp = getf(zc->zc_cookie);
4621 return (SET_ERROR(EBADF));
4625 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off);
4627 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4629 releasef(zc->zc_cookie);
4635 * Remove all ACL files in shares dir
4637 #ifdef HAVE_SMB_SHARE
4639 zfs_smb_acl_purge(znode_t *dzp)
4642 zap_attribute_t zap;
4643 zfs_sb_t *zsb = ZTOZSB(dzp);
4646 for (zap_cursor_init(&zc, zsb->z_os, dzp->z_id);
4647 (error = zap_cursor_retrieve(&zc, &zap)) == 0;
4648 zap_cursor_advance(&zc)) {
4649 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
4653 zap_cursor_fini(&zc);
4656 #endif /* HAVE_SMB_SHARE */
4659 zfs_ioc_smb_acl(zfs_cmd_t *zc)
4661 #ifdef HAVE_SMB_SHARE
4664 vnode_t *resourcevp = NULL;
4673 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
4674 NO_FOLLOW, NULL, &vp)) != 0)
4677 /* Now make sure mntpnt and dataset are ZFS */
4679 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
4680 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
4681 zc->zc_name) != 0)) {
4683 return (SET_ERROR(EINVAL));
4691 * Create share dir if its missing.
4693 mutex_enter(&zsb->z_lock);
4694 if (zsb->z_shares_dir == 0) {
4697 tx = dmu_tx_create(zsb->z_os);
4698 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
4700 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
4701 error = dmu_tx_assign(tx, TXG_WAIT);
4705 error = zfs_create_share_dir(zsb, tx);
4709 mutex_exit(&zsb->z_lock);
4715 mutex_exit(&zsb->z_lock);
4717 ASSERT(zsb->z_shares_dir);
4718 if ((error = zfs_zget(zsb, zsb->z_shares_dir, &sharedir)) != 0) {
4724 switch (zc->zc_cookie) {
4725 case ZFS_SMB_ACL_ADD:
4726 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
4727 vattr.va_mode = S_IFREG|0777;
4731 vsec.vsa_mask = VSA_ACE;
4732 vsec.vsa_aclentp = &full_access;
4733 vsec.vsa_aclentsz = sizeof (full_access);
4734 vsec.vsa_aclcnt = 1;
4736 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
4737 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
4739 VN_RELE(resourcevp);
4742 case ZFS_SMB_ACL_REMOVE:
4743 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
4747 case ZFS_SMB_ACL_RENAME:
4748 if ((error = get_nvlist(zc->zc_nvlist_src,
4749 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
4754 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
4755 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
4758 VN_RELE(ZTOV(sharedir));
4760 nvlist_free(nvlist);
4763 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
4765 nvlist_free(nvlist);
4768 case ZFS_SMB_ACL_PURGE:
4769 error = zfs_smb_acl_purge(sharedir);
4773 error = SET_ERROR(EINVAL);
4778 VN_RELE(ZTOV(sharedir));
4784 return (SET_ERROR(ENOTSUP));
4785 #endif /* HAVE_SMB_SHARE */
4790 * "holds" -> { snapname -> holdname (string), ... }
4791 * (optional) "cleanup_fd" -> fd (int32)
4795 * snapname -> error value (int32)
4801 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
4804 int cleanup_fd = -1;
4808 error = nvlist_lookup_nvlist(args, "holds", &holds);
4810 return (SET_ERROR(EINVAL));
4812 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
4813 error = zfs_onexit_fd_hold(cleanup_fd, &minor);
4818 error = dsl_dataset_user_hold(holds, minor, errlist);
4820 zfs_onexit_fd_rele(cleanup_fd);
4825 * innvl is not used.
4828 * holdname -> time added (uint64 seconds since epoch)
4834 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
4836 return (dsl_dataset_get_holds(snapname, outnvl));
4841 * snapname -> { holdname, ... }
4846 * snapname -> error value (int32)
4852 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
4854 return (dsl_dataset_user_release(holds, errlist));
4859 * zc_guid flags (ZEVENT_NONBLOCK)
4860 * zc_cleanup_fd zevent file descriptor
4863 * zc_nvlist_dst next nvlist event
4864 * zc_cookie dropped events since last get
4867 zfs_ioc_events_next(zfs_cmd_t *zc)
4870 nvlist_t *event = NULL;
4872 uint64_t dropped = 0;
4875 error = zfs_zevent_fd_hold(zc->zc_cleanup_fd, &minor, &ze);
4880 error = zfs_zevent_next(ze, &event,
4881 &zc->zc_nvlist_dst_size, &dropped);
4882 if (event != NULL) {
4883 zc->zc_cookie = dropped;
4884 error = put_nvlist(zc, event);
4888 if (zc->zc_guid & ZEVENT_NONBLOCK)
4891 if ((error == 0) || (error != ENOENT))
4894 error = zfs_zevent_wait(ze);
4899 zfs_zevent_fd_rele(zc->zc_cleanup_fd);
4906 * zc_cookie cleared events count
4909 zfs_ioc_events_clear(zfs_cmd_t *zc)
4913 zfs_zevent_drain_all(&count);
4914 zc->zc_cookie = count;
4921 * zc_guid eid | ZEVENT_SEEK_START | ZEVENT_SEEK_END
4922 * zc_cleanup zevent file descriptor
4925 zfs_ioc_events_seek(zfs_cmd_t *zc)
4931 error = zfs_zevent_fd_hold(zc->zc_cleanup_fd, &minor, &ze);
4935 error = zfs_zevent_seek(ze, zc->zc_guid);
4936 zfs_zevent_fd_rele(zc->zc_cleanup_fd);
4943 * zc_name name of new filesystem or snapshot
4944 * zc_value full name of old snapshot
4947 * zc_cookie space in bytes
4948 * zc_objset_type compressed space in bytes
4949 * zc_perm_action uncompressed space in bytes
4952 zfs_ioc_space_written(zfs_cmd_t *zc)
4956 dsl_dataset_t *new, *old;
4958 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4961 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
4963 dsl_pool_rele(dp, FTAG);
4966 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
4968 dsl_dataset_rele(new, FTAG);
4969 dsl_pool_rele(dp, FTAG);
4973 error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
4974 &zc->zc_objset_type, &zc->zc_perm_action);
4975 dsl_dataset_rele(old, FTAG);
4976 dsl_dataset_rele(new, FTAG);
4977 dsl_pool_rele(dp, FTAG);
4983 * "firstsnap" -> snapshot name
4987 * "used" -> space in bytes
4988 * "compressed" -> compressed space in bytes
4989 * "uncompressed" -> uncompressed space in bytes
4993 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
4997 dsl_dataset_t *new, *old;
4999 uint64_t used, comp, uncomp;
5001 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5002 return (SET_ERROR(EINVAL));
5004 error = dsl_pool_hold(lastsnap, FTAG, &dp);
5008 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
5010 dsl_pool_rele(dp, FTAG);
5013 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
5015 dsl_dataset_rele(new, FTAG);
5016 dsl_pool_rele(dp, FTAG);
5020 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5021 dsl_dataset_rele(old, FTAG);
5022 dsl_dataset_rele(new, FTAG);
5023 dsl_pool_rele(dp, FTAG);
5024 fnvlist_add_uint64(outnvl, "used", used);
5025 fnvlist_add_uint64(outnvl, "compressed", comp);
5026 fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5032 * "fd" -> file descriptor to write stream to (int32)
5033 * (optional) "fromsnap" -> full snap name to send an incremental from
5040 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5044 char *fromname = NULL;
5048 error = nvlist_lookup_int32(innvl, "fd", &fd);
5050 return (SET_ERROR(EINVAL));
5052 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
5054 if ((fp = getf(fd)) == NULL)
5055 return (SET_ERROR(EBADF));
5058 error = dmu_send(snapname, fromname, fd, fp->f_vnode, &off);
5060 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5068 * Determine approximately how large a zfs send stream will be -- the number
5069 * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5072 * (optional) "fromsnap" -> full snap name to send an incremental from
5076 * "space" -> bytes of space (uint64)
5080 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5083 dsl_dataset_t *fromsnap = NULL;
5084 dsl_dataset_t *tosnap;
5089 error = dsl_pool_hold(snapname, FTAG, &dp);
5093 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
5095 dsl_pool_rele(dp, FTAG);
5099 error = nvlist_lookup_string(innvl, "fromsnap", &fromname);
5101 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
5103 dsl_dataset_rele(tosnap, FTAG);
5104 dsl_pool_rele(dp, FTAG);
5109 error = dmu_send_estimate(tosnap, fromsnap, &space);
5110 fnvlist_add_uint64(outnvl, "space", space);
5112 if (fromsnap != NULL)
5113 dsl_dataset_rele(fromsnap, FTAG);
5114 dsl_dataset_rele(tosnap, FTAG);
5115 dsl_pool_rele(dp, FTAG);
5120 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
5123 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5124 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5125 boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
5127 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5129 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5130 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5131 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5132 ASSERT3P(vec->zvec_func, ==, NULL);
5134 vec->zvec_legacy_func = func;
5135 vec->zvec_secpolicy = secpolicy;
5136 vec->zvec_namecheck = namecheck;
5137 vec->zvec_allow_log = log_history;
5138 vec->zvec_pool_check = pool_check;
5142 * See the block comment at the beginning of this file for details on
5143 * each argument to this function.
5146 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
5147 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5148 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
5149 boolean_t allow_log)
5151 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5153 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5154 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5155 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5156 ASSERT3P(vec->zvec_func, ==, NULL);
5158 /* if we are logging, the name must be valid */
5159 ASSERT(!allow_log || namecheck != NO_NAME);
5161 vec->zvec_name = name;
5162 vec->zvec_func = func;
5163 vec->zvec_secpolicy = secpolicy;
5164 vec->zvec_namecheck = namecheck;
5165 vec->zvec_pool_check = pool_check;
5166 vec->zvec_smush_outnvlist = smush_outnvlist;
5167 vec->zvec_allow_log = allow_log;
5171 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5172 zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
5173 zfs_ioc_poolcheck_t pool_check)
5175 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5176 POOL_NAME, log_history, pool_check);
5180 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5181 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
5183 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5184 DATASET_NAME, B_FALSE, pool_check);
5188 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5190 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
5191 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5195 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5196 zfs_secpolicy_func_t *secpolicy)
5198 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5199 NO_NAME, B_FALSE, POOL_CHECK_NONE);
5203 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
5204 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
5206 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5207 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
5211 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5213 zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
5214 zfs_secpolicy_read);
5218 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5219 zfs_secpolicy_func_t *secpolicy)
5221 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5222 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5226 zfs_ioctl_init(void)
5228 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
5229 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
5230 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5232 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
5233 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
5234 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
5236 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
5237 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
5238 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5240 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
5241 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
5242 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5244 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
5245 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
5246 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5248 zfs_ioctl_register("create", ZFS_IOC_CREATE,
5249 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
5250 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5252 zfs_ioctl_register("clone", ZFS_IOC_CLONE,
5253 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
5254 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5256 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
5257 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
5258 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5260 zfs_ioctl_register("hold", ZFS_IOC_HOLD,
5261 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
5262 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5263 zfs_ioctl_register("release", ZFS_IOC_RELEASE,
5264 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
5265 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5267 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
5268 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
5269 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5271 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
5272 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
5273 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE);
5275 /* IOCTLS that use the legacy function signature */
5277 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
5278 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
5280 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
5281 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5282 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
5284 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
5285 zfs_ioc_pool_upgrade);
5286 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
5288 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
5289 zfs_ioc_vdev_remove);
5290 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
5291 zfs_ioc_vdev_set_state);
5292 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
5293 zfs_ioc_vdev_attach);
5294 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
5295 zfs_ioc_vdev_detach);
5296 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
5297 zfs_ioc_vdev_setpath);
5298 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
5299 zfs_ioc_vdev_setfru);
5300 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
5301 zfs_ioc_pool_set_props);
5302 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
5303 zfs_ioc_vdev_split);
5304 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
5305 zfs_ioc_pool_reguid);
5307 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
5308 zfs_ioc_pool_configs, zfs_secpolicy_none);
5309 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
5310 zfs_ioc_pool_tryimport, zfs_secpolicy_config);
5311 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
5312 zfs_ioc_inject_fault, zfs_secpolicy_inject);
5313 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
5314 zfs_ioc_clear_fault, zfs_secpolicy_inject);
5315 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
5316 zfs_ioc_inject_list_next, zfs_secpolicy_inject);
5319 * pool destroy, and export don't log the history as part of
5320 * zfsdev_ioctl, but rather zfs_ioc_pool_export
5321 * does the logging of those commands.
5323 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
5324 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5325 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
5326 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5328 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
5329 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5330 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
5331 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5333 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
5334 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED);
5335 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
5336 zfs_ioc_dsobj_to_dsname,
5337 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED);
5338 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
5339 zfs_ioc_pool_get_history,
5340 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
5342 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
5343 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5345 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
5346 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5347 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
5348 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5350 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
5351 zfs_ioc_space_written);
5352 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
5353 zfs_ioc_objset_recvd_props);
5354 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
5356 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
5358 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
5359 zfs_ioc_objset_stats);
5360 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
5361 zfs_ioc_objset_zplprops);
5362 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
5363 zfs_ioc_dataset_list_next);
5364 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
5365 zfs_ioc_snapshot_list_next);
5366 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
5367 zfs_ioc_send_progress);
5369 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
5370 zfs_ioc_diff, zfs_secpolicy_diff);
5371 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
5372 zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
5373 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
5374 zfs_ioc_obj_to_path, zfs_secpolicy_diff);
5375 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
5376 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
5377 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
5378 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
5379 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
5380 zfs_ioc_send, zfs_secpolicy_send);
5382 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
5383 zfs_secpolicy_none);
5384 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
5385 zfs_secpolicy_destroy);
5386 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
5387 zfs_secpolicy_rename);
5388 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
5389 zfs_secpolicy_recv);
5390 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
5391 zfs_secpolicy_promote);
5392 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
5393 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
5394 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
5395 zfs_secpolicy_set_fsacl);
5397 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
5398 zfs_secpolicy_share, POOL_CHECK_NONE);
5399 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
5400 zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
5401 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
5402 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
5403 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5404 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
5405 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
5406 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5411 zfs_ioctl_register_legacy(ZFS_IOC_EVENTS_NEXT, zfs_ioc_events_next,
5412 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);
5413 zfs_ioctl_register_legacy(ZFS_IOC_EVENTS_CLEAR, zfs_ioc_events_clear,
5414 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);
5415 zfs_ioctl_register_legacy(ZFS_IOC_EVENTS_SEEK, zfs_ioc_events_seek,
5416 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);
5420 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
5421 zfs_ioc_poolcheck_t check)
5426 ASSERT(type == POOL_NAME || type == DATASET_NAME);
5428 if (check & POOL_CHECK_NONE)
5431 error = spa_open(name, &spa, FTAG);
5433 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
5434 error = SET_ERROR(EAGAIN);
5435 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
5436 error = SET_ERROR(EROFS);
5437 spa_close(spa, FTAG);
5443 zfsdev_get_state_impl(minor_t minor, enum zfsdev_state_type which)
5447 for (zs = zfsdev_state_list; zs != NULL; zs = zs->zs_next) {
5448 if (zs->zs_minor == minor) {
5452 return (zs->zs_onexit);
5454 return (zs->zs_zevent);
5465 zfsdev_get_state(minor_t minor, enum zfsdev_state_type which)
5469 ptr = zfsdev_get_state_impl(minor, which);
5475 zfsdev_getminor(struct file *filp)
5477 ASSERT(filp != NULL);
5478 ASSERT(filp->private_data != NULL);
5480 return (((zfsdev_state_t *)filp->private_data)->zs_minor);
5484 * Find a free minor number. The zfsdev_state_list is expected to
5485 * be short since it is only a list of currently open file handles.
5488 zfsdev_minor_alloc(void)
5490 static minor_t last_minor = 0;
5493 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5495 for (m = last_minor + 1; m != last_minor; m++) {
5496 if (m > ZFSDEV_MAX_MINOR)
5498 if (zfsdev_get_state_impl(m, ZST_ALL) == NULL) {
5508 zfsdev_state_init(struct file *filp)
5510 zfsdev_state_t *zs, *zsprev = NULL;
5512 boolean_t newzs = B_FALSE;
5514 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5516 minor = zfsdev_minor_alloc();
5518 return (SET_ERROR(ENXIO));
5520 for (zs = zfsdev_state_list; zs != NULL; zs = zs->zs_next) {
5521 if (zs->zs_minor == -1)
5527 zs = kmem_zalloc(sizeof (zfsdev_state_t), KM_SLEEP);
5532 filp->private_data = zs;
5534 zfs_onexit_init((zfs_onexit_t **)&zs->zs_onexit);
5535 zfs_zevent_init((zfs_zevent_t **)&zs->zs_zevent);
5539 * In order to provide for lock-free concurrent read access
5540 * to the minor list in zfsdev_get_state_impl(), new entries
5541 * must be completely written before linking them into the
5542 * list whereas existing entries are already linked; the last
5543 * operation must be updating zs_minor (from -1 to the new
5547 zs->zs_minor = minor;
5549 zsprev->zs_next = zs;
5552 zs->zs_minor = minor;
5559 zfsdev_state_destroy(struct file *filp)
5563 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5564 ASSERT(filp->private_data != NULL);
5566 zs = filp->private_data;
5568 zfs_onexit_destroy(zs->zs_onexit);
5569 zfs_zevent_destroy(zs->zs_zevent);
5575 zfsdev_open(struct inode *ino, struct file *filp)
5579 mutex_enter(&zfsdev_state_lock);
5580 error = zfsdev_state_init(filp);
5581 mutex_exit(&zfsdev_state_lock);
5587 zfsdev_release(struct inode *ino, struct file *filp)
5591 mutex_enter(&zfsdev_state_lock);
5592 error = zfsdev_state_destroy(filp);
5593 mutex_exit(&zfsdev_state_lock);
5599 zfsdev_ioctl(struct file *filp, unsigned cmd, unsigned long arg)
5603 int error, rc, flag = 0;
5604 const zfs_ioc_vec_t *vec;
5605 char *saved_poolname = NULL;
5606 nvlist_t *innvl = NULL;
5608 vecnum = cmd - ZFS_IOC_FIRST;
5609 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
5610 return (-SET_ERROR(EINVAL));
5611 vec = &zfs_ioc_vec[vecnum];
5614 * The registered ioctl list may be sparse, verify that either
5615 * a normal or legacy handler are registered.
5617 if (vec->zvec_func == NULL && vec->zvec_legacy_func == NULL)
5618 return (-SET_ERROR(EINVAL));
5620 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP | KM_NODEBUG);
5622 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
5624 error = SET_ERROR(EFAULT);
5628 zc->zc_iflags = flag & FKIOCTL;
5629 if (zc->zc_nvlist_src_size != 0) {
5630 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
5631 zc->zc_iflags, &innvl);
5637 * Ensure that all pool/dataset names are valid before we pass down to
5640 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
5641 switch (vec->zvec_namecheck) {
5643 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
5644 error = SET_ERROR(EINVAL);
5646 error = pool_status_check(zc->zc_name,
5647 vec->zvec_namecheck, vec->zvec_pool_check);
5651 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
5652 error = SET_ERROR(EINVAL);
5654 error = pool_status_check(zc->zc_name,
5655 vec->zvec_namecheck, vec->zvec_pool_check);
5663 if (error == 0 && !(flag & FKIOCTL))
5664 error = vec->zvec_secpolicy(zc, innvl, CRED());
5669 /* legacy ioctls can modify zc_name */
5670 saved_poolname = strdup(zc->zc_name);
5671 if (saved_poolname == NULL) {
5672 error = SET_ERROR(ENOMEM);
5675 saved_poolname[strcspn(saved_poolname, "/@#")] = '\0';
5678 if (vec->zvec_func != NULL) {
5682 nvlist_t *lognv = NULL;
5684 ASSERT(vec->zvec_legacy_func == NULL);
5687 * Add the innvl to the lognv before calling the func,
5688 * in case the func changes the innvl.
5690 if (vec->zvec_allow_log) {
5691 lognv = fnvlist_alloc();
5692 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
5694 if (!nvlist_empty(innvl)) {
5695 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
5700 VERIFY0(nvlist_alloc(&outnvl, NV_UNIQUE_NAME, KM_PUSHPAGE));
5701 error = vec->zvec_func(zc->zc_name, innvl, outnvl);
5703 if (error == 0 && vec->zvec_allow_log &&
5704 spa_open(zc->zc_name, &spa, FTAG) == 0) {
5705 if (!nvlist_empty(outnvl)) {
5706 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
5709 (void) spa_history_log_nvl(spa, lognv);
5710 spa_close(spa, FTAG);
5712 fnvlist_free(lognv);
5714 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
5716 if (vec->zvec_smush_outnvlist) {
5717 smusherror = nvlist_smush(outnvl,
5718 zc->zc_nvlist_dst_size);
5720 if (smusherror == 0)
5721 puterror = put_nvlist(zc, outnvl);
5727 nvlist_free(outnvl);
5729 error = vec->zvec_legacy_func(zc);
5734 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
5735 if (error == 0 && rc != 0)
5736 error = SET_ERROR(EFAULT);
5737 if (error == 0 && vec->zvec_allow_log) {
5738 char *s = tsd_get(zfs_allow_log_key);
5741 (void) tsd_set(zfs_allow_log_key, saved_poolname);
5743 if (saved_poolname != NULL)
5744 strfree(saved_poolname);
5747 kmem_free(zc, sizeof (zfs_cmd_t));
5751 #ifdef CONFIG_COMPAT
5753 zfsdev_compat_ioctl(struct file *filp, unsigned cmd, unsigned long arg)
5755 return (zfsdev_ioctl(filp, cmd, arg));
5758 #define zfsdev_compat_ioctl NULL
5761 static const struct file_operations zfsdev_fops = {
5762 .open = zfsdev_open,
5763 .release = zfsdev_release,
5764 .unlocked_ioctl = zfsdev_ioctl,
5765 .compat_ioctl = zfsdev_compat_ioctl,
5766 .owner = THIS_MODULE,
5769 static struct miscdevice zfs_misc = {
5770 .minor = MISC_DYNAMIC_MINOR,
5772 .fops = &zfsdev_fops,
5780 mutex_init(&zfsdev_state_lock, NULL, MUTEX_DEFAULT, NULL);
5781 zfsdev_state_list = kmem_zalloc(sizeof (zfsdev_state_t), KM_SLEEP);
5782 zfsdev_state_list->zs_minor = -1;
5784 error = misc_register(&zfs_misc);
5786 printk(KERN_INFO "ZFS: misc_register() failed %d\n", error);
5797 zfsdev_state_t *zs, *zsprev = NULL;
5799 error = misc_deregister(&zfs_misc);
5801 printk(KERN_INFO "ZFS: misc_deregister() failed %d\n", error);
5803 mutex_destroy(&zfsdev_state_lock);
5805 for (zs = zfsdev_state_list; zs != NULL; zs = zs->zs_next) {
5807 kmem_free(zsprev, sizeof (zfsdev_state_t));
5811 kmem_free(zsprev, sizeof (zfsdev_state_t));
5815 zfs_allow_log_destroy(void *arg)
5817 char *poolname = arg;
5822 #define ZFS_DEBUG_STR " (DEBUG mode)"
5824 #define ZFS_DEBUG_STR ""
5832 spa_init(FREAD | FWRITE);
5835 if ((error = zvol_init()) != 0)
5840 if ((error = zfs_attach()) != 0)
5843 tsd_create(&zfs_fsyncer_key, NULL);
5844 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
5845 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
5847 printk(KERN_NOTICE "ZFS: Loaded module v%s-%s%s, "
5848 "ZFS pool version %s, ZFS filesystem version %s\n",
5849 ZFS_META_VERSION, ZFS_META_RELEASE, ZFS_DEBUG_STR,
5850 SPA_VERSION_STRING, ZPL_VERSION_STRING);
5851 #ifndef CONFIG_FS_POSIX_ACL
5852 printk(KERN_NOTICE "ZFS: Posix ACLs disabled by kernel\n");
5853 #endif /* CONFIG_FS_POSIX_ACL */
5862 printk(KERN_NOTICE "ZFS: Failed to Load ZFS Filesystem v%s-%s%s"
5863 ", rc = %d\n", ZFS_META_VERSION, ZFS_META_RELEASE,
5864 ZFS_DEBUG_STR, error);
5877 tsd_destroy(&zfs_fsyncer_key);
5878 tsd_destroy(&rrw_tsd_key);
5879 tsd_destroy(&zfs_allow_log_key);
5881 printk(KERN_NOTICE "ZFS: Unloaded module v%s-%s%s\n",
5882 ZFS_META_VERSION, ZFS_META_RELEASE, ZFS_DEBUG_STR);
5888 spl_module_init(_init);
5889 spl_module_exit(_fini);
5891 MODULE_DESCRIPTION("ZFS");
5892 MODULE_AUTHOR(ZFS_META_AUTHOR);
5893 MODULE_LICENSE(ZFS_META_LICENSE);
5894 MODULE_VERSION(ZFS_META_VERSION "-" ZFS_META_RELEASE);
5895 #endif /* HAVE_SPL */