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 * Copyright (c) 2011-2012 Pawel Jakub Dawidek <pawel@dawidek.net>.
25 * All rights reserved.
26 * Copyright 2013 Martin Matuska <mm@FreeBSD.org>. All rights reserved.
27 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
28 * Copyright (c) 2014, Joyent, Inc. All rights reserved.
29 * Copyright (c) 2013 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/systm.h>
138 #include <sys/conf.h>
139 #include <sys/kernel.h>
140 #include <sys/lock.h>
141 #include <sys/malloc.h>
142 #include <sys/mutex.h>
143 #include <sys/proc.h>
144 #include <sys/errno.h>
147 #include <sys/file.h>
148 #include <sys/kmem.h>
149 #include <sys/conf.h>
150 #include <sys/cmn_err.h>
151 #include <sys/stat.h>
152 #include <sys/zfs_ioctl.h>
153 #include <sys/zfs_vfsops.h>
154 #include <sys/zfs_znode.h>
157 #include <sys/spa_impl.h>
158 #include <sys/vdev.h>
160 #include <sys/dsl_dir.h>
161 #include <sys/dsl_dataset.h>
162 #include <sys/dsl_prop.h>
163 #include <sys/dsl_deleg.h>
164 #include <sys/dmu_objset.h>
165 #include <sys/dmu_impl.h>
166 #include <sys/dmu_tx.h>
167 #include <sys/sunddi.h>
168 #include <sys/policy.h>
169 #include <sys/zone.h>
170 #include <sys/nvpair.h>
171 #include <sys/mount.h>
172 #include <sys/taskqueue.h>
174 #include <sys/varargs.h>
175 #include <sys/fs/zfs.h>
176 #include <sys/zfs_ctldir.h>
177 #include <sys/zfs_dir.h>
178 #include <sys/zfs_onexit.h>
179 #include <sys/zvol.h>
180 #include <sys/dsl_scan.h>
181 #include <sys/dmu_objset.h>
182 #include <sys/dmu_send.h>
183 #include <sys/dsl_destroy.h>
184 #include <sys/dsl_bookmark.h>
185 #include <sys/dsl_userhold.h>
186 #include <sys/zfeature.h>
188 #include "zfs_namecheck.h"
189 #include "zfs_prop.h"
190 #include "zfs_deleg.h"
191 #include "zfs_comutil.h"
192 #include "zfs_ioctl_compat.h"
194 CTASSERT(sizeof(zfs_cmd_t) < IOCPARM_MAX);
196 static int snapshot_list_prefetch;
197 SYSCTL_DECL(_vfs_zfs);
198 TUNABLE_INT("vfs.zfs.snapshot_list_prefetch", &snapshot_list_prefetch);
199 SYSCTL_INT(_vfs_zfs, OID_AUTO, snapshot_list_prefetch, CTLFLAG_RW,
200 &snapshot_list_prefetch, 0, "Prefetch data when listing snapshots");
202 static struct cdev *zfsdev;
204 extern void zfs_init(void);
205 extern void zfs_fini(void);
207 uint_t zfs_fsyncer_key;
208 extern uint_t rrw_tsd_key;
209 static uint_t zfs_allow_log_key;
211 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *);
212 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *);
213 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *);
219 } zfs_ioc_namecheck_t;
222 POOL_CHECK_NONE = 1 << 0,
223 POOL_CHECK_SUSPENDED = 1 << 1,
224 POOL_CHECK_READONLY = 1 << 2,
225 } zfs_ioc_poolcheck_t;
227 typedef struct zfs_ioc_vec {
228 zfs_ioc_legacy_func_t *zvec_legacy_func;
229 zfs_ioc_func_t *zvec_func;
230 zfs_secpolicy_func_t *zvec_secpolicy;
231 zfs_ioc_namecheck_t zvec_namecheck;
232 boolean_t zvec_allow_log;
233 zfs_ioc_poolcheck_t zvec_pool_check;
234 boolean_t zvec_smush_outnvlist;
235 const char *zvec_name;
238 /* This array is indexed by zfs_userquota_prop_t */
239 static const char *userquota_perms[] = {
240 ZFS_DELEG_PERM_USERUSED,
241 ZFS_DELEG_PERM_USERQUOTA,
242 ZFS_DELEG_PERM_GROUPUSED,
243 ZFS_DELEG_PERM_GROUPQUOTA,
246 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
247 static int zfs_check_settable(const char *name, nvpair_t *property,
249 static int zfs_check_clearable(char *dataset, nvlist_t *props,
251 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
253 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
254 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
256 static void zfsdev_close(void *data);
258 static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature);
260 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
262 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
269 * Get rid of annoying "../common/" prefix to filename.
271 newfile = strrchr(file, '/');
272 if (newfile != NULL) {
273 newfile = newfile + 1; /* Get rid of leading / */
279 (void) vsnprintf(buf, sizeof (buf), fmt, adx);
283 * To get this data, use the zfs-dprintf probe as so:
284 * dtrace -q -n 'zfs-dprintf \
285 * /stringof(arg0) == "dbuf.c"/ \
286 * {printf("%s: %s", stringof(arg1), stringof(arg3))}'
288 * arg1 = function name
292 DTRACE_PROBE4(zfs__dprintf,
293 char *, newfile, char *, func, int, line, char *, buf);
297 history_str_free(char *buf)
299 kmem_free(buf, HIS_MAX_RECORD_LEN);
303 history_str_get(zfs_cmd_t *zc)
307 if (zc->zc_history == 0)
310 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
311 if (copyinstr((void *)(uintptr_t)zc->zc_history,
312 buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
313 history_str_free(buf);
317 buf[HIS_MAX_RECORD_LEN -1] = '\0';
323 * Check to see if the named dataset is currently defined as bootable
326 zfs_is_bootfs(const char *name)
330 if (dmu_objset_hold(name, FTAG, &os) == 0) {
332 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
333 dmu_objset_rele(os, FTAG);
340 * Return non-zero if the spa version is less than requested version.
343 zfs_earlier_version(const char *name, int version)
347 if (spa_open(name, &spa, FTAG) == 0) {
348 if (spa_version(spa) < version) {
349 spa_close(spa, FTAG);
352 spa_close(spa, FTAG);
358 * Return TRUE if the ZPL version is less than requested version.
361 zpl_earlier_version(const char *name, int version)
364 boolean_t rc = B_TRUE;
366 if (dmu_objset_hold(name, FTAG, &os) == 0) {
369 if (dmu_objset_type(os) != DMU_OST_ZFS) {
370 dmu_objset_rele(os, FTAG);
373 /* XXX reading from non-owned objset */
374 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
375 rc = zplversion < version;
376 dmu_objset_rele(os, FTAG);
382 zfs_log_history(zfs_cmd_t *zc)
387 if ((buf = history_str_get(zc)) == NULL)
390 if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
391 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
392 (void) spa_history_log(spa, buf);
393 spa_close(spa, FTAG);
395 history_str_free(buf);
399 * Policy for top-level read operations (list pools). Requires no privileges,
400 * and can be used in the local zone, as there is no associated dataset.
404 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
410 * Policy for dataset read operations (list children, get statistics). Requires
411 * no privileges, but must be visible in the local zone.
415 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
417 if (INGLOBALZONE(curthread) ||
418 zone_dataset_visible(zc->zc_name, NULL))
421 return (SET_ERROR(ENOENT));
425 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
430 * The dataset must be visible by this zone -- check this first
431 * so they don't see EPERM on something they shouldn't know about.
433 if (!INGLOBALZONE(curthread) &&
434 !zone_dataset_visible(dataset, &writable))
435 return (SET_ERROR(ENOENT));
437 if (INGLOBALZONE(curthread)) {
439 * If the fs is zoned, only root can access it from the
442 if (secpolicy_zfs(cr) && zoned)
443 return (SET_ERROR(EPERM));
446 * If we are in a local zone, the 'zoned' property must be set.
449 return (SET_ERROR(EPERM));
451 /* must be writable by this zone */
453 return (SET_ERROR(EPERM));
459 zfs_dozonecheck(const char *dataset, cred_t *cr)
463 if (dsl_prop_get_integer(dataset, "jailed", &zoned, NULL))
464 return (SET_ERROR(ENOENT));
466 return (zfs_dozonecheck_impl(dataset, zoned, cr));
470 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
474 if (dsl_prop_get_int_ds(ds, "jailed", &zoned))
475 return (SET_ERROR(ENOENT));
477 return (zfs_dozonecheck_impl(dataset, zoned, cr));
481 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
482 const char *perm, cred_t *cr)
486 error = zfs_dozonecheck_ds(name, ds, cr);
488 error = secpolicy_zfs(cr);
490 error = dsl_deleg_access_impl(ds, perm, cr);
496 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
502 error = dsl_pool_hold(name, FTAG, &dp);
506 error = dsl_dataset_hold(dp, name, FTAG, &ds);
508 dsl_pool_rele(dp, FTAG);
512 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
514 dsl_dataset_rele(ds, FTAG);
515 dsl_pool_rele(dp, FTAG);
521 * Policy for setting the security label property.
523 * Returns 0 for success, non-zero for access and other errors.
526 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
528 char ds_hexsl[MAXNAMELEN];
529 bslabel_t ds_sl, new_sl;
530 boolean_t new_default = FALSE;
532 int needed_priv = -1;
535 /* First get the existing dataset label. */
536 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
537 1, sizeof (ds_hexsl), &ds_hexsl, NULL);
539 return (SET_ERROR(EPERM));
541 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
544 /* The label must be translatable */
545 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
546 return (SET_ERROR(EINVAL));
549 * In a non-global zone, disallow attempts to set a label that
550 * doesn't match that of the zone; otherwise no other checks
553 if (!INGLOBALZONE(curproc)) {
554 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
555 return (SET_ERROR(EPERM));
560 * For global-zone datasets (i.e., those whose zoned property is
561 * "off", verify that the specified new label is valid for the
564 if (dsl_prop_get_integer(name,
565 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
566 return (SET_ERROR(EPERM));
568 if (zfs_check_global_label(name, strval) != 0)
569 return (SET_ERROR(EPERM));
573 * If the existing dataset label is nondefault, check if the
574 * dataset is mounted (label cannot be changed while mounted).
575 * Get the zfsvfs; if there isn't one, then the dataset isn't
576 * mounted (or isn't a dataset, doesn't exist, ...).
578 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
580 static char *setsl_tag = "setsl_tag";
583 * Try to own the dataset; abort if there is any error,
584 * (e.g., already mounted, in use, or other error).
586 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
589 return (SET_ERROR(EPERM));
591 dmu_objset_disown(os, setsl_tag);
594 needed_priv = PRIV_FILE_DOWNGRADE_SL;
598 if (hexstr_to_label(strval, &new_sl) != 0)
599 return (SET_ERROR(EPERM));
601 if (blstrictdom(&ds_sl, &new_sl))
602 needed_priv = PRIV_FILE_DOWNGRADE_SL;
603 else if (blstrictdom(&new_sl, &ds_sl))
604 needed_priv = PRIV_FILE_UPGRADE_SL;
606 /* dataset currently has a default label */
608 needed_priv = PRIV_FILE_UPGRADE_SL;
612 if (needed_priv != -1)
613 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
616 #endif /* SECLABEL */
619 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
625 * Check permissions for special properties.
630 * Disallow setting of 'zoned' from within a local zone.
632 if (!INGLOBALZONE(curthread))
633 return (SET_ERROR(EPERM));
637 case ZFS_PROP_FILESYSTEM_LIMIT:
638 case ZFS_PROP_SNAPSHOT_LIMIT:
639 if (!INGLOBALZONE(curthread)) {
641 char setpoint[MAXNAMELEN];
643 * Unprivileged users are allowed to modify the
644 * limit on things *under* (ie. contained by)
645 * the thing they own.
647 if (dsl_prop_get_integer(dsname, "jailed", &zoned,
649 return (SET_ERROR(EPERM));
650 if (!zoned || strlen(dsname) <= strlen(setpoint))
651 return (SET_ERROR(EPERM));
655 case ZFS_PROP_MLSLABEL:
657 if (!is_system_labeled())
658 return (SET_ERROR(EPERM));
660 if (nvpair_value_string(propval, &strval) == 0) {
663 err = zfs_set_slabel_policy(dsname, strval, CRED());
673 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
678 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
682 error = zfs_dozonecheck(zc->zc_name, cr);
687 * permission to set permissions will be evaluated later in
688 * dsl_deleg_can_allow()
695 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
697 return (zfs_secpolicy_write_perms(zc->zc_name,
698 ZFS_DELEG_PERM_ROLLBACK, cr));
703 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
711 * Generate the current snapshot name from the given objsetid, then
712 * use that name for the secpolicy/zone checks.
714 cp = strchr(zc->zc_name, '@');
716 return (SET_ERROR(EINVAL));
717 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
721 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
723 dsl_pool_rele(dp, FTAG);
727 dsl_dataset_name(ds, zc->zc_name);
729 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
730 ZFS_DELEG_PERM_SEND, cr);
731 dsl_dataset_rele(ds, FTAG);
732 dsl_pool_rele(dp, FTAG);
739 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
741 return (zfs_secpolicy_write_perms(zc->zc_name,
742 ZFS_DELEG_PERM_SEND, cr));
747 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
752 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
753 NO_FOLLOW, NULL, &vp)) != 0)
756 /* Now make sure mntpnt and dataset are ZFS */
758 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 ||
759 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
760 zc->zc_name) != 0)) {
762 return (SET_ERROR(EPERM));
766 return (dsl_deleg_access(zc->zc_name,
767 ZFS_DELEG_PERM_SHARE, cr));
771 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
773 if (!INGLOBALZONE(curthread))
774 return (SET_ERROR(EPERM));
776 if (secpolicy_nfs(cr) == 0) {
779 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
784 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
786 if (!INGLOBALZONE(curthread))
787 return (SET_ERROR(EPERM));
789 if (secpolicy_smb(cr) == 0) {
792 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
797 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
802 * Remove the @bla or /bla from the end of the name to get the parent.
804 (void) strncpy(parent, datasetname, parentsize);
805 cp = strrchr(parent, '@');
809 cp = strrchr(parent, '/');
811 return (SET_ERROR(ENOENT));
819 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
823 if ((error = zfs_secpolicy_write_perms(name,
824 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
827 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
832 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
834 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
838 * Destroying snapshots with delegated permissions requires
839 * descendant mount and destroy permissions.
843 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
846 nvpair_t *pair, *nextpair;
849 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
850 return (SET_ERROR(EINVAL));
851 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
853 nextpair = nvlist_next_nvpair(snaps, pair);
854 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr);
855 if (error == ENOENT) {
857 * Ignore any snapshots that don't exist (we consider
858 * them "already destroyed"). Remove the name from the
859 * nvl here in case the snapshot is created between
860 * now and when we try to destroy it (in which case
861 * we don't want to destroy it since we haven't
862 * checked for permission).
864 fnvlist_remove_nvpair(snaps, pair);
875 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
877 char parentname[MAXNAMELEN];
880 if ((error = zfs_secpolicy_write_perms(from,
881 ZFS_DELEG_PERM_RENAME, cr)) != 0)
884 if ((error = zfs_secpolicy_write_perms(from,
885 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
888 if ((error = zfs_get_parent(to, parentname,
889 sizeof (parentname))) != 0)
892 if ((error = zfs_secpolicy_write_perms(parentname,
893 ZFS_DELEG_PERM_CREATE, cr)) != 0)
896 if ((error = zfs_secpolicy_write_perms(parentname,
897 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
905 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
910 if ((zc->zc_cookie & 1) != 0) {
912 * This is recursive rename, so the starting snapshot might
913 * not exist. Check file system or volume permission instead.
915 at = strchr(zc->zc_name, '@');
921 error = zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr);
931 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
934 dsl_dataset_t *clone;
937 error = zfs_secpolicy_write_perms(zc->zc_name,
938 ZFS_DELEG_PERM_PROMOTE, cr);
942 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
946 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
949 char parentname[MAXNAMELEN];
950 dsl_dataset_t *origin = NULL;
954 error = dsl_dataset_hold_obj(dd->dd_pool,
955 dd->dd_phys->dd_origin_obj, FTAG, &origin);
957 dsl_dataset_rele(clone, FTAG);
958 dsl_pool_rele(dp, FTAG);
962 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
963 ZFS_DELEG_PERM_MOUNT, cr);
965 dsl_dataset_name(origin, parentname);
967 error = zfs_secpolicy_write_perms_ds(parentname, origin,
968 ZFS_DELEG_PERM_PROMOTE, cr);
970 dsl_dataset_rele(clone, FTAG);
971 dsl_dataset_rele(origin, FTAG);
973 dsl_pool_rele(dp, FTAG);
979 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
983 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
984 ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
987 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
988 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
991 return (zfs_secpolicy_write_perms(zc->zc_name,
992 ZFS_DELEG_PERM_CREATE, cr));
996 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
998 return (zfs_secpolicy_write_perms(name,
999 ZFS_DELEG_PERM_SNAPSHOT, cr));
1003 * Check for permission to create each snapshot in the nvlist.
1007 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1013 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
1014 return (SET_ERROR(EINVAL));
1015 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
1016 pair = nvlist_next_nvpair(snaps, pair)) {
1017 char *name = nvpair_name(pair);
1018 char *atp = strchr(name, '@');
1021 error = SET_ERROR(EINVAL);
1025 error = zfs_secpolicy_snapshot_perms(name, cr);
1034 * Check for permission to create each snapshot in the nvlist.
1038 zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1042 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
1043 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
1044 char *name = nvpair_name(pair);
1045 char *hashp = strchr(name, '#');
1047 if (hashp == NULL) {
1048 error = SET_ERROR(EINVAL);
1052 error = zfs_secpolicy_write_perms(name,
1053 ZFS_DELEG_PERM_BOOKMARK, cr);
1063 zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1065 nvpair_t *pair, *nextpair;
1068 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1070 char *name = nvpair_name(pair);
1071 char *hashp = strchr(name, '#');
1072 nextpair = nvlist_next_nvpair(innvl, pair);
1074 if (hashp == NULL) {
1075 error = SET_ERROR(EINVAL);
1080 error = zfs_secpolicy_write_perms(name,
1081 ZFS_DELEG_PERM_DESTROY, cr);
1083 if (error == ENOENT) {
1085 * Ignore any filesystems that don't exist (we consider
1086 * their bookmarks "already destroyed"). Remove
1087 * the name from the nvl here in case the filesystem
1088 * is created between now and when we try to destroy
1089 * the bookmark (in which case we don't want to
1090 * destroy it since we haven't checked for permission).
1092 fnvlist_remove_nvpair(innvl, pair);
1104 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1107 * Even root must have a proper TSD so that we know what pool
1110 if (tsd_get(zfs_allow_log_key) == NULL)
1111 return (SET_ERROR(EPERM));
1116 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1118 char parentname[MAXNAMELEN];
1122 if ((error = zfs_get_parent(zc->zc_name, parentname,
1123 sizeof (parentname))) != 0)
1126 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1127 (error = zfs_secpolicy_write_perms(origin,
1128 ZFS_DELEG_PERM_CLONE, cr)) != 0)
1131 if ((error = zfs_secpolicy_write_perms(parentname,
1132 ZFS_DELEG_PERM_CREATE, cr)) != 0)
1135 return (zfs_secpolicy_write_perms(parentname,
1136 ZFS_DELEG_PERM_MOUNT, cr));
1140 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires
1141 * SYS_CONFIG privilege, which is not available in a local zone.
1145 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1147 if (secpolicy_sys_config(cr, B_FALSE) != 0)
1148 return (SET_ERROR(EPERM));
1154 * Policy for object to name lookups.
1158 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1162 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1165 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1170 * Policy for fault injection. Requires all privileges.
1174 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1176 return (secpolicy_zinject(cr));
1181 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1183 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1185 if (prop == ZPROP_INVAL) {
1186 if (!zfs_prop_user(zc->zc_value))
1187 return (SET_ERROR(EINVAL));
1188 return (zfs_secpolicy_write_perms(zc->zc_name,
1189 ZFS_DELEG_PERM_USERPROP, cr));
1191 return (zfs_secpolicy_setprop(zc->zc_name, prop,
1197 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1199 int err = zfs_secpolicy_read(zc, innvl, cr);
1203 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1204 return (SET_ERROR(EINVAL));
1206 if (zc->zc_value[0] == 0) {
1208 * They are asking about a posix uid/gid. If it's
1209 * themself, allow it.
1211 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1212 zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
1213 if (zc->zc_guid == crgetuid(cr))
1216 if (groupmember(zc->zc_guid, cr))
1221 return (zfs_secpolicy_write_perms(zc->zc_name,
1222 userquota_perms[zc->zc_objset_type], cr));
1226 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1228 int err = zfs_secpolicy_read(zc, innvl, cr);
1232 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1233 return (SET_ERROR(EINVAL));
1235 return (zfs_secpolicy_write_perms(zc->zc_name,
1236 userquota_perms[zc->zc_objset_type], cr));
1241 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1243 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1249 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1255 error = nvlist_lookup_nvlist(innvl, "holds", &holds);
1257 return (SET_ERROR(EINVAL));
1259 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
1260 pair = nvlist_next_nvpair(holds, pair)) {
1261 char fsname[MAXNAMELEN];
1262 error = dmu_fsname(nvpair_name(pair), fsname);
1265 error = zfs_secpolicy_write_perms(fsname,
1266 ZFS_DELEG_PERM_HOLD, cr);
1275 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1280 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1281 pair = nvlist_next_nvpair(innvl, pair)) {
1282 char fsname[MAXNAMELEN];
1283 error = dmu_fsname(nvpair_name(pair), fsname);
1286 error = zfs_secpolicy_write_perms(fsname,
1287 ZFS_DELEG_PERM_RELEASE, cr);
1295 * Policy for allowing temporary snapshots to be taken or released
1298 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1301 * A temporary snapshot is the same as a snapshot,
1302 * hold, destroy and release all rolled into one.
1303 * Delegated diff alone is sufficient that we allow this.
1307 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1308 ZFS_DELEG_PERM_DIFF, cr)) == 0)
1311 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1313 error = zfs_secpolicy_hold(zc, innvl, cr);
1315 error = zfs_secpolicy_release(zc, innvl, cr);
1317 error = zfs_secpolicy_destroy(zc, innvl, cr);
1322 * Returns the nvlist as specified by the user in the zfs_cmd_t.
1325 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1329 nvlist_t *list = NULL;
1332 * Read in and unpack the user-supplied nvlist.
1335 return (SET_ERROR(EINVAL));
1337 packed = kmem_alloc(size, KM_SLEEP);
1339 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1341 kmem_free(packed, size);
1345 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1346 kmem_free(packed, size);
1350 kmem_free(packed, size);
1357 * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1358 * Entries will be removed from the end of the nvlist, and one int32 entry
1359 * named "N_MORE_ERRORS" will be added indicating how many entries were
1363 nvlist_smush(nvlist_t *errors, size_t max)
1367 size = fnvlist_size(errors);
1370 nvpair_t *more_errors;
1374 return (SET_ERROR(ENOMEM));
1376 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1377 more_errors = nvlist_prev_nvpair(errors, NULL);
1380 nvpair_t *pair = nvlist_prev_nvpair(errors,
1382 fnvlist_remove_nvpair(errors, pair);
1384 size = fnvlist_size(errors);
1385 } while (size > max);
1387 fnvlist_remove_nvpair(errors, more_errors);
1388 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1389 ASSERT3U(fnvlist_size(errors), <=, max);
1396 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1398 char *packed = NULL;
1402 size = fnvlist_size(nvl);
1404 if (size > zc->zc_nvlist_dst_size) {
1406 * Solaris returns ENOMEM here, because even if an error is
1407 * returned from an ioctl(2), new zc_nvlist_dst_size will be
1408 * passed to the userland. This is not the case for FreeBSD.
1409 * We need to return 0, so the kernel will copy the
1410 * zc_nvlist_dst_size back and the userland can discover that a
1411 * bigger buffer is needed.
1415 packed = fnvlist_pack(nvl, &size);
1416 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1417 size, zc->zc_iflags) != 0)
1418 error = SET_ERROR(EFAULT);
1419 fnvlist_pack_free(packed, size);
1422 zc->zc_nvlist_dst_size = size;
1423 zc->zc_nvlist_dst_filled = B_TRUE;
1428 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1433 error = dmu_objset_hold(dsname, FTAG, &os);
1436 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1437 dmu_objset_rele(os, FTAG);
1438 return (SET_ERROR(EINVAL));
1441 mutex_enter(&os->os_user_ptr_lock);
1442 *zfvp = dmu_objset_get_user(os);
1444 VFS_HOLD((*zfvp)->z_vfs);
1446 error = SET_ERROR(ESRCH);
1448 mutex_exit(&os->os_user_ptr_lock);
1449 dmu_objset_rele(os, FTAG);
1454 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1455 * case its z_vfs will be NULL, and it will be opened as the owner.
1456 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1457 * which prevents all vnode ops from running.
1460 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1464 if (getzfsvfs(name, zfvp) != 0)
1465 error = zfsvfs_create(name, zfvp);
1467 rrw_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER :
1469 if ((*zfvp)->z_unmounted) {
1471 * XXX we could probably try again, since the unmounting
1472 * thread should be just about to disassociate the
1473 * objset from the zfsvfs.
1475 rrw_exit(&(*zfvp)->z_teardown_lock, tag);
1476 return (SET_ERROR(EBUSY));
1483 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1485 rrw_exit(&zfsvfs->z_teardown_lock, tag);
1487 if (zfsvfs->z_vfs) {
1488 VFS_RELE(zfsvfs->z_vfs);
1490 dmu_objset_disown(zfsvfs->z_os, zfsvfs);
1491 zfsvfs_free(zfsvfs);
1496 zfs_ioc_pool_create(zfs_cmd_t *zc)
1499 nvlist_t *config, *props = NULL;
1500 nvlist_t *rootprops = NULL;
1501 nvlist_t *zplprops = NULL;
1503 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1504 zc->zc_iflags, &config))
1507 if (zc->zc_nvlist_src_size != 0 && (error =
1508 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1509 zc->zc_iflags, &props))) {
1510 nvlist_free(config);
1515 nvlist_t *nvl = NULL;
1516 uint64_t version = SPA_VERSION;
1518 (void) nvlist_lookup_uint64(props,
1519 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1520 if (!SPA_VERSION_IS_SUPPORTED(version)) {
1521 error = SET_ERROR(EINVAL);
1522 goto pool_props_bad;
1524 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1526 error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1528 nvlist_free(config);
1532 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1534 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1535 error = zfs_fill_zplprops_root(version, rootprops,
1538 goto pool_props_bad;
1541 error = spa_create(zc->zc_name, config, props, zplprops);
1544 * Set the remaining root properties
1546 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1547 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1548 (void) spa_destroy(zc->zc_name);
1551 nvlist_free(rootprops);
1552 nvlist_free(zplprops);
1553 nvlist_free(config);
1560 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1563 zfs_log_history(zc);
1564 error = spa_destroy(zc->zc_name);
1566 zvol_remove_minors(zc->zc_name);
1571 zfs_ioc_pool_import(zfs_cmd_t *zc)
1573 nvlist_t *config, *props = NULL;
1577 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1578 zc->zc_iflags, &config)) != 0)
1581 if (zc->zc_nvlist_src_size != 0 && (error =
1582 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1583 zc->zc_iflags, &props))) {
1584 nvlist_free(config);
1588 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1589 guid != zc->zc_guid)
1590 error = SET_ERROR(EINVAL);
1592 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1594 if (zc->zc_nvlist_dst != 0) {
1597 if ((err = put_nvlist(zc, config)) != 0)
1601 nvlist_free(config);
1610 zfs_ioc_pool_export(zfs_cmd_t *zc)
1613 boolean_t force = (boolean_t)zc->zc_cookie;
1614 boolean_t hardforce = (boolean_t)zc->zc_guid;
1616 zfs_log_history(zc);
1617 error = spa_export(zc->zc_name, NULL, force, hardforce);
1619 zvol_remove_minors(zc->zc_name);
1624 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1629 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1630 return (SET_ERROR(EEXIST));
1632 error = put_nvlist(zc, configs);
1634 nvlist_free(configs);
1641 * zc_name name of the pool
1644 * zc_cookie real errno
1645 * zc_nvlist_dst config nvlist
1646 * zc_nvlist_dst_size size of config nvlist
1649 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1655 error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1656 sizeof (zc->zc_value));
1658 if (config != NULL) {
1659 ret = put_nvlist(zc, config);
1660 nvlist_free(config);
1663 * The config may be present even if 'error' is non-zero.
1664 * In this case we return success, and preserve the real errno
1667 zc->zc_cookie = error;
1676 * Try to import the given pool, returning pool stats as appropriate so that
1677 * user land knows which devices are available and overall pool health.
1680 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1682 nvlist_t *tryconfig, *config;
1685 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1686 zc->zc_iflags, &tryconfig)) != 0)
1689 config = spa_tryimport(tryconfig);
1691 nvlist_free(tryconfig);
1694 return (SET_ERROR(EINVAL));
1696 error = put_nvlist(zc, config);
1697 nvlist_free(config);
1704 * zc_name name of the pool
1705 * zc_cookie scan func (pool_scan_func_t)
1708 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1713 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1716 if (zc->zc_cookie == POOL_SCAN_NONE)
1717 error = spa_scan_stop(spa);
1719 error = spa_scan(spa, zc->zc_cookie);
1721 spa_close(spa, FTAG);
1727 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1732 error = spa_open(zc->zc_name, &spa, FTAG);
1735 spa_close(spa, FTAG);
1741 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1746 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1749 if (zc->zc_cookie < spa_version(spa) ||
1750 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1751 spa_close(spa, FTAG);
1752 return (SET_ERROR(EINVAL));
1755 spa_upgrade(spa, zc->zc_cookie);
1756 spa_close(spa, FTAG);
1762 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1769 if ((size = zc->zc_history_len) == 0)
1770 return (SET_ERROR(EINVAL));
1772 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1775 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1776 spa_close(spa, FTAG);
1777 return (SET_ERROR(ENOTSUP));
1780 hist_buf = kmem_alloc(size, KM_SLEEP);
1781 if ((error = spa_history_get(spa, &zc->zc_history_offset,
1782 &zc->zc_history_len, hist_buf)) == 0) {
1783 error = ddi_copyout(hist_buf,
1784 (void *)(uintptr_t)zc->zc_history,
1785 zc->zc_history_len, zc->zc_iflags);
1788 spa_close(spa, FTAG);
1789 kmem_free(hist_buf, size);
1794 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1799 error = spa_open(zc->zc_name, &spa, FTAG);
1801 error = spa_change_guid(spa);
1802 spa_close(spa, FTAG);
1808 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1810 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1815 * zc_name name of filesystem
1816 * zc_obj object to find
1819 * zc_value name of object
1822 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1827 /* XXX reading from objset not owned */
1828 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1830 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1831 dmu_objset_rele(os, FTAG);
1832 return (SET_ERROR(EINVAL));
1834 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1835 sizeof (zc->zc_value));
1836 dmu_objset_rele(os, FTAG);
1843 * zc_name name of filesystem
1844 * zc_obj object to find
1847 * zc_stat stats on object
1848 * zc_value path to object
1851 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1856 /* XXX reading from objset not owned */
1857 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1859 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1860 dmu_objset_rele(os, FTAG);
1861 return (SET_ERROR(EINVAL));
1863 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1864 sizeof (zc->zc_value));
1865 dmu_objset_rele(os, FTAG);
1871 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1875 nvlist_t *config, **l2cache, **spares;
1876 uint_t nl2cache = 0, nspares = 0;
1878 error = spa_open(zc->zc_name, &spa, FTAG);
1882 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1883 zc->zc_iflags, &config);
1884 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1885 &l2cache, &nl2cache);
1887 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1892 * A root pool with concatenated devices is not supported.
1893 * Thus, can not add a device to a root pool.
1895 * Intent log device can not be added to a rootpool because
1896 * during mountroot, zil is replayed, a seperated log device
1897 * can not be accessed during the mountroot time.
1899 * l2cache and spare devices are ok to be added to a rootpool.
1901 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1902 nvlist_free(config);
1903 spa_close(spa, FTAG);
1904 return (SET_ERROR(EDOM));
1906 #endif /* illumos */
1909 error = spa_vdev_add(spa, config);
1910 nvlist_free(config);
1912 spa_close(spa, FTAG);
1918 * zc_name name of the pool
1919 * zc_nvlist_conf nvlist of devices to remove
1920 * zc_cookie to stop the remove?
1923 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1928 error = spa_open(zc->zc_name, &spa, FTAG);
1931 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1932 spa_close(spa, FTAG);
1937 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1941 vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1943 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1945 switch (zc->zc_cookie) {
1946 case VDEV_STATE_ONLINE:
1947 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1950 case VDEV_STATE_OFFLINE:
1951 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1954 case VDEV_STATE_FAULTED:
1955 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1956 zc->zc_obj != VDEV_AUX_EXTERNAL)
1957 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1959 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1962 case VDEV_STATE_DEGRADED:
1963 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1964 zc->zc_obj != VDEV_AUX_EXTERNAL)
1965 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1967 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1971 error = SET_ERROR(EINVAL);
1973 zc->zc_cookie = newstate;
1974 spa_close(spa, FTAG);
1979 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1982 int replacing = zc->zc_cookie;
1986 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1989 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1990 zc->zc_iflags, &config)) == 0) {
1991 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1992 nvlist_free(config);
1995 spa_close(spa, FTAG);
2000 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
2005 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2008 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
2010 spa_close(spa, FTAG);
2015 zfs_ioc_vdev_split(zfs_cmd_t *zc)
2018 nvlist_t *config, *props = NULL;
2020 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
2022 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2025 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
2026 zc->zc_iflags, &config)) {
2027 spa_close(spa, FTAG);
2031 if (zc->zc_nvlist_src_size != 0 && (error =
2032 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2033 zc->zc_iflags, &props))) {
2034 spa_close(spa, FTAG);
2035 nvlist_free(config);
2039 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
2041 spa_close(spa, FTAG);
2043 nvlist_free(config);
2050 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
2053 char *path = zc->zc_value;
2054 uint64_t guid = zc->zc_guid;
2057 error = spa_open(zc->zc_name, &spa, FTAG);
2061 error = spa_vdev_setpath(spa, guid, path);
2062 spa_close(spa, FTAG);
2067 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
2070 char *fru = zc->zc_value;
2071 uint64_t guid = zc->zc_guid;
2074 error = spa_open(zc->zc_name, &spa, FTAG);
2078 error = spa_vdev_setfru(spa, guid, fru);
2079 spa_close(spa, FTAG);
2084 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
2089 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2091 if (zc->zc_nvlist_dst != 0 &&
2092 (error = dsl_prop_get_all(os, &nv)) == 0) {
2093 dmu_objset_stats(os, nv);
2095 * NB: zvol_get_stats() will read the objset contents,
2096 * which we aren't supposed to do with a
2097 * DS_MODE_USER hold, because it could be
2098 * inconsistent. So this is a bit of a workaround...
2099 * XXX reading with out owning
2101 if (!zc->zc_objset_stats.dds_inconsistent &&
2102 dmu_objset_type(os) == DMU_OST_ZVOL) {
2103 error = zvol_get_stats(os, nv);
2108 error = put_nvlist(zc, nv);
2117 * zc_name name of filesystem
2118 * zc_nvlist_dst_size size of buffer for property nvlist
2121 * zc_objset_stats stats
2122 * zc_nvlist_dst property nvlist
2123 * zc_nvlist_dst_size size of property nvlist
2126 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2131 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2133 error = zfs_ioc_objset_stats_impl(zc, os);
2134 dmu_objset_rele(os, FTAG);
2137 if (error == ENOMEM)
2144 * zc_name name of filesystem
2145 * zc_nvlist_dst_size size of buffer for property nvlist
2148 * zc_nvlist_dst received property nvlist
2149 * zc_nvlist_dst_size size of received property nvlist
2151 * Gets received properties (distinct from local properties on or after
2152 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2153 * local property values.
2156 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2162 * Without this check, we would return local property values if the
2163 * caller has not already received properties on or after
2164 * SPA_VERSION_RECVD_PROPS.
2166 if (!dsl_prop_get_hasrecvd(zc->zc_name))
2167 return (SET_ERROR(ENOTSUP));
2169 if (zc->zc_nvlist_dst != 0 &&
2170 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2171 error = put_nvlist(zc, nv);
2179 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2185 * zfs_get_zplprop() will either find a value or give us
2186 * the default value (if there is one).
2188 if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2190 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2196 * zc_name name of filesystem
2197 * zc_nvlist_dst_size size of buffer for zpl property nvlist
2200 * zc_nvlist_dst zpl property nvlist
2201 * zc_nvlist_dst_size size of zpl property nvlist
2204 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2209 /* XXX reading without owning */
2210 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os))
2213 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2216 * NB: nvl_add_zplprop() will read the objset contents,
2217 * which we aren't supposed to do with a DS_MODE_USER
2218 * hold, because it could be inconsistent.
2220 if (zc->zc_nvlist_dst != 0 &&
2221 !zc->zc_objset_stats.dds_inconsistent &&
2222 dmu_objset_type(os) == DMU_OST_ZFS) {
2225 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2226 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2227 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2228 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2229 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2230 err = put_nvlist(zc, nv);
2233 err = SET_ERROR(ENOENT);
2235 dmu_objset_rele(os, FTAG);
2240 dataset_name_hidden(const char *name)
2243 * Skip over datasets that are not visible in this zone,
2244 * internal datasets (which have a $ in their name), and
2245 * temporary datasets (which have a % in their name).
2247 if (strchr(name, '$') != NULL)
2249 if (strchr(name, '%') != NULL)
2251 if (!INGLOBALZONE(curthread) && !zone_dataset_visible(name, NULL))
2258 * zc_name name of filesystem
2259 * zc_cookie zap cursor
2260 * zc_nvlist_dst_size size of buffer for property nvlist
2263 * zc_name name of next filesystem
2264 * zc_cookie zap cursor
2265 * zc_objset_stats stats
2266 * zc_nvlist_dst property nvlist
2267 * zc_nvlist_dst_size size of property nvlist
2270 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2275 size_t orig_len = strlen(zc->zc_name);
2278 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) {
2279 if (error == ENOENT)
2280 error = SET_ERROR(ESRCH);
2284 p = strrchr(zc->zc_name, '/');
2285 if (p == NULL || p[1] != '\0')
2286 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2287 p = zc->zc_name + strlen(zc->zc_name);
2290 error = dmu_dir_list_next(os,
2291 sizeof (zc->zc_name) - (p - zc->zc_name), p,
2292 NULL, &zc->zc_cookie);
2293 if (error == ENOENT)
2294 error = SET_ERROR(ESRCH);
2295 } while (error == 0 && dataset_name_hidden(zc->zc_name));
2296 dmu_objset_rele(os, FTAG);
2299 * If it's an internal dataset (ie. with a '$' in its name),
2300 * don't try to get stats for it, otherwise we'll return ENOENT.
2302 if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2303 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2304 if (error == ENOENT) {
2305 /* We lost a race with destroy, get the next one. */
2306 zc->zc_name[orig_len] = '\0';
2315 * zc_name name of filesystem
2316 * zc_cookie zap cursor
2317 * zc_nvlist_dst_size size of buffer for property nvlist
2318 * zc_simple when set, only name is requested
2321 * zc_name name of next snapshot
2322 * zc_objset_stats stats
2323 * zc_nvlist_dst property nvlist
2324 * zc_nvlist_dst_size size of property nvlist
2327 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2332 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2334 return (error == ENOENT ? ESRCH : error);
2338 * A dataset name of maximum length cannot have any snapshots,
2339 * so exit immediately.
2341 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
2342 dmu_objset_rele(os, FTAG);
2343 return (SET_ERROR(ESRCH));
2346 error = dmu_snapshot_list_next(os,
2347 sizeof (zc->zc_name) - strlen(zc->zc_name),
2348 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2351 if (error == 0 && !zc->zc_simple) {
2353 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2355 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2359 error = dmu_objset_from_ds(ds, &ossnap);
2361 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2362 dsl_dataset_rele(ds, FTAG);
2364 } else if (error == ENOENT) {
2365 error = SET_ERROR(ESRCH);
2368 dmu_objset_rele(os, FTAG);
2369 /* if we failed, undo the @ that we tacked on to zc_name */
2371 *strchr(zc->zc_name, '@') = '\0';
2376 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2378 const char *propname = nvpair_name(pair);
2380 unsigned int vallen;
2383 zfs_userquota_prop_t type;
2389 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2391 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2392 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2394 return (SET_ERROR(EINVAL));
2398 * A correctly constructed propname is encoded as
2399 * userquota@<rid>-<domain>.
2401 if ((dash = strchr(propname, '-')) == NULL ||
2402 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2404 return (SET_ERROR(EINVAL));
2411 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2413 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2414 zfsvfs_rele(zfsvfs, FTAG);
2421 * If the named property is one that has a special function to set its value,
2422 * return 0 on success and a positive error code on failure; otherwise if it is
2423 * not one of the special properties handled by this function, return -1.
2425 * XXX: It would be better for callers of the property interface if we handled
2426 * these special cases in dsl_prop.c (in the dsl layer).
2429 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2432 const char *propname = nvpair_name(pair);
2433 zfs_prop_t prop = zfs_name_to_prop(propname);
2437 if (prop == ZPROP_INVAL) {
2438 if (zfs_prop_userquota(propname))
2439 return (zfs_prop_set_userquota(dsname, pair));
2443 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2445 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2446 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2450 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2453 VERIFY(0 == nvpair_value_uint64(pair, &intval));
2456 case ZFS_PROP_QUOTA:
2457 err = dsl_dir_set_quota(dsname, source, intval);
2459 case ZFS_PROP_REFQUOTA:
2460 err = dsl_dataset_set_refquota(dsname, source, intval);
2462 case ZFS_PROP_FILESYSTEM_LIMIT:
2463 case ZFS_PROP_SNAPSHOT_LIMIT:
2464 if (intval == UINT64_MAX) {
2465 /* clearing the limit, just do it */
2468 err = dsl_dir_activate_fs_ss_limit(dsname);
2471 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2472 * default path to set the value in the nvlist.
2477 case ZFS_PROP_RESERVATION:
2478 err = dsl_dir_set_reservation(dsname, source, intval);
2480 case ZFS_PROP_REFRESERVATION:
2481 err = dsl_dataset_set_refreservation(dsname, source, intval);
2483 case ZFS_PROP_VOLSIZE:
2484 err = zvol_set_volsize(dsname, ddi_driver_major(zfs_dip),
2487 case ZFS_PROP_VERSION:
2491 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2494 err = zfs_set_version(zfsvfs, intval);
2495 zfsvfs_rele(zfsvfs, FTAG);
2497 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2500 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2501 (void) strcpy(zc->zc_name, dsname);
2502 (void) zfs_ioc_userspace_upgrade(zc);
2503 kmem_free(zc, sizeof (zfs_cmd_t));
2507 case ZFS_PROP_COMPRESSION:
2509 if (intval == ZIO_COMPRESS_LZ4) {
2512 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
2516 * Setting the LZ4 compression algorithm activates
2519 if (!spa_feature_is_active(spa,
2520 SPA_FEATURE_LZ4_COMPRESS)) {
2521 if ((err = zfs_prop_activate_feature(spa,
2522 SPA_FEATURE_LZ4_COMPRESS)) != 0) {
2523 spa_close(spa, FTAG);
2528 spa_close(spa, FTAG);
2531 * We still want the default set action to be performed in the
2532 * caller, we only performed zfeature settings here.
2546 * This function is best effort. If it fails to set any of the given properties,
2547 * it continues to set as many as it can and returns the last error
2548 * encountered. If the caller provides a non-NULL errlist, it will be filled in
2549 * with the list of names of all the properties that failed along with the
2550 * corresponding error numbers.
2552 * If every property is set successfully, zero is returned and errlist is not
2556 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2564 nvlist_t *genericnvl = fnvlist_alloc();
2565 nvlist_t *retrynvl = fnvlist_alloc();
2569 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2570 const char *propname = nvpair_name(pair);
2571 zfs_prop_t prop = zfs_name_to_prop(propname);
2574 /* decode the property value */
2576 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2578 attrs = fnvpair_value_nvlist(pair);
2579 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2581 err = SET_ERROR(EINVAL);
2584 /* Validate value type */
2585 if (err == 0 && prop == ZPROP_INVAL) {
2586 if (zfs_prop_user(propname)) {
2587 if (nvpair_type(propval) != DATA_TYPE_STRING)
2588 err = SET_ERROR(EINVAL);
2589 } else if (zfs_prop_userquota(propname)) {
2590 if (nvpair_type(propval) !=
2591 DATA_TYPE_UINT64_ARRAY)
2592 err = SET_ERROR(EINVAL);
2594 err = SET_ERROR(EINVAL);
2596 } else if (err == 0) {
2597 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2598 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2599 err = SET_ERROR(EINVAL);
2600 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2603 intval = fnvpair_value_uint64(propval);
2605 switch (zfs_prop_get_type(prop)) {
2606 case PROP_TYPE_NUMBER:
2608 case PROP_TYPE_STRING:
2609 err = SET_ERROR(EINVAL);
2611 case PROP_TYPE_INDEX:
2612 if (zfs_prop_index_to_string(prop,
2613 intval, &unused) != 0)
2614 err = SET_ERROR(EINVAL);
2618 "unknown property type");
2621 err = SET_ERROR(EINVAL);
2625 /* Validate permissions */
2627 err = zfs_check_settable(dsname, pair, CRED());
2630 err = zfs_prop_set_special(dsname, source, pair);
2633 * For better performance we build up a list of
2634 * properties to set in a single transaction.
2636 err = nvlist_add_nvpair(genericnvl, pair);
2637 } else if (err != 0 && nvl != retrynvl) {
2639 * This may be a spurious error caused by
2640 * receiving quota and reservation out of order.
2641 * Try again in a second pass.
2643 err = nvlist_add_nvpair(retrynvl, pair);
2648 if (errlist != NULL)
2649 fnvlist_add_int32(errlist, propname, err);
2654 if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2659 if (!nvlist_empty(genericnvl) &&
2660 dsl_props_set(dsname, source, genericnvl) != 0) {
2662 * If this fails, we still want to set as many properties as we
2663 * can, so try setting them individually.
2666 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2667 const char *propname = nvpair_name(pair);
2671 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2673 attrs = fnvpair_value_nvlist(pair);
2674 propval = fnvlist_lookup_nvpair(attrs,
2678 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2679 strval = fnvpair_value_string(propval);
2680 err = dsl_prop_set_string(dsname, propname,
2683 intval = fnvpair_value_uint64(propval);
2684 err = dsl_prop_set_int(dsname, propname, source,
2689 if (errlist != NULL) {
2690 fnvlist_add_int32(errlist, propname,
2697 nvlist_free(genericnvl);
2698 nvlist_free(retrynvl);
2704 * Check that all the properties are valid user properties.
2707 zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2709 nvpair_t *pair = NULL;
2712 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2713 const char *propname = nvpair_name(pair);
2715 if (!zfs_prop_user(propname) ||
2716 nvpair_type(pair) != DATA_TYPE_STRING)
2717 return (SET_ERROR(EINVAL));
2719 if (error = zfs_secpolicy_write_perms(fsname,
2720 ZFS_DELEG_PERM_USERPROP, CRED()))
2723 if (strlen(propname) >= ZAP_MAXNAMELEN)
2724 return (SET_ERROR(ENAMETOOLONG));
2726 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN)
2733 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2737 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2740 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2741 if (nvlist_exists(skipped, nvpair_name(pair)))
2744 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2749 clear_received_props(const char *dsname, nvlist_t *props,
2753 nvlist_t *cleared_props = NULL;
2754 props_skip(props, skipped, &cleared_props);
2755 if (!nvlist_empty(cleared_props)) {
2757 * Acts on local properties until the dataset has received
2758 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2760 zprop_source_t flags = (ZPROP_SRC_NONE |
2761 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2762 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2764 nvlist_free(cleared_props);
2770 * zc_name name of filesystem
2771 * zc_value name of property to set
2772 * zc_nvlist_src{_size} nvlist of properties to apply
2773 * zc_cookie received properties flag
2776 * zc_nvlist_dst{_size} error for each unapplied received property
2779 zfs_ioc_set_prop(zfs_cmd_t *zc)
2782 boolean_t received = zc->zc_cookie;
2783 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2788 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2789 zc->zc_iflags, &nvl)) != 0)
2793 nvlist_t *origprops;
2795 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2796 (void) clear_received_props(zc->zc_name,
2798 nvlist_free(origprops);
2801 error = dsl_prop_set_hasrecvd(zc->zc_name);
2804 errors = fnvlist_alloc();
2806 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2808 if (zc->zc_nvlist_dst != 0 && errors != NULL) {
2809 (void) put_nvlist(zc, errors);
2812 nvlist_free(errors);
2819 * zc_name name of filesystem
2820 * zc_value name of property to inherit
2821 * zc_cookie revert to received value if TRUE
2826 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2828 const char *propname = zc->zc_value;
2829 zfs_prop_t prop = zfs_name_to_prop(propname);
2830 boolean_t received = zc->zc_cookie;
2831 zprop_source_t source = (received
2832 ? ZPROP_SRC_NONE /* revert to received value, if any */
2833 : ZPROP_SRC_INHERITED); /* explicitly inherit */
2842 * zfs_prop_set_special() expects properties in the form of an
2843 * nvpair with type info.
2845 if (prop == ZPROP_INVAL) {
2846 if (!zfs_prop_user(propname))
2847 return (SET_ERROR(EINVAL));
2849 type = PROP_TYPE_STRING;
2850 } else if (prop == ZFS_PROP_VOLSIZE ||
2851 prop == ZFS_PROP_VERSION) {
2852 return (SET_ERROR(EINVAL));
2854 type = zfs_prop_get_type(prop);
2857 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2860 case PROP_TYPE_STRING:
2861 VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2863 case PROP_TYPE_NUMBER:
2864 case PROP_TYPE_INDEX:
2865 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2869 return (SET_ERROR(EINVAL));
2872 pair = nvlist_next_nvpair(dummy, NULL);
2873 err = zfs_prop_set_special(zc->zc_name, source, pair);
2876 return (err); /* special property already handled */
2879 * Only check this in the non-received case. We want to allow
2880 * 'inherit -S' to revert non-inheritable properties like quota
2881 * and reservation to the received or default values even though
2882 * they are not considered inheritable.
2884 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2885 return (SET_ERROR(EINVAL));
2888 /* property name has been validated by zfs_secpolicy_inherit_prop() */
2889 return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source));
2893 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2900 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2901 zc->zc_iflags, &props))
2905 * If the only property is the configfile, then just do a spa_lookup()
2906 * to handle the faulted case.
2908 pair = nvlist_next_nvpair(props, NULL);
2909 if (pair != NULL && strcmp(nvpair_name(pair),
2910 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2911 nvlist_next_nvpair(props, pair) == NULL) {
2912 mutex_enter(&spa_namespace_lock);
2913 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2914 spa_configfile_set(spa, props, B_FALSE);
2915 spa_config_sync(spa, B_FALSE, B_TRUE);
2917 mutex_exit(&spa_namespace_lock);
2924 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2929 error = spa_prop_set(spa, props);
2932 spa_close(spa, FTAG);
2938 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2942 nvlist_t *nvp = NULL;
2944 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2946 * If the pool is faulted, there may be properties we can still
2947 * get (such as altroot and cachefile), so attempt to get them
2950 mutex_enter(&spa_namespace_lock);
2951 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2952 error = spa_prop_get(spa, &nvp);
2953 mutex_exit(&spa_namespace_lock);
2955 error = spa_prop_get(spa, &nvp);
2956 spa_close(spa, FTAG);
2959 if (error == 0 && zc->zc_nvlist_dst != 0)
2960 error = put_nvlist(zc, nvp);
2962 error = SET_ERROR(EFAULT);
2970 * zc_name name of filesystem
2971 * zc_nvlist_src{_size} nvlist of delegated permissions
2972 * zc_perm_action allow/unallow flag
2977 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2980 nvlist_t *fsaclnv = NULL;
2982 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2983 zc->zc_iflags, &fsaclnv)) != 0)
2987 * Verify nvlist is constructed correctly
2989 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2990 nvlist_free(fsaclnv);
2991 return (SET_ERROR(EINVAL));
2995 * If we don't have PRIV_SYS_MOUNT, then validate
2996 * that user is allowed to hand out each permission in
3000 error = secpolicy_zfs(CRED());
3002 if (zc->zc_perm_action == B_FALSE) {
3003 error = dsl_deleg_can_allow(zc->zc_name,
3006 error = dsl_deleg_can_unallow(zc->zc_name,
3012 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
3014 nvlist_free(fsaclnv);
3020 * zc_name name of filesystem
3023 * zc_nvlist_src{_size} nvlist of delegated permissions
3026 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
3031 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
3032 error = put_nvlist(zc, nvp);
3040 * Search the vfs list for a specified resource. Returns a pointer to it
3041 * or NULL if no suitable entry is found. The caller of this routine
3042 * is responsible for releasing the returned vfs pointer.
3045 zfs_get_vfs(const char *resource)
3049 mtx_lock(&mountlist_mtx);
3050 TAILQ_FOREACH(vfsp, &mountlist, mnt_list) {
3051 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
3056 mtx_unlock(&mountlist_mtx);
3062 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
3064 zfs_creat_t *zct = arg;
3066 zfs_create_fs(os, cr, zct->zct_zplprops, tx);
3069 #define ZFS_PROP_UNDEFINED ((uint64_t)-1)
3073 * os parent objset pointer (NULL if root fs)
3074 * fuids_ok fuids allowed in this version of the spa?
3075 * sa_ok SAs allowed in this version of the spa?
3076 * createprops list of properties requested by creator
3079 * zplprops values for the zplprops we attach to the master node object
3080 * is_ci true if requested file system will be purely case-insensitive
3082 * Determine the settings for utf8only, normalization and
3083 * casesensitivity. Specific values may have been requested by the
3084 * creator and/or we can inherit values from the parent dataset. If
3085 * the file system is of too early a vintage, a creator can not
3086 * request settings for these properties, even if the requested
3087 * setting is the default value. We don't actually want to create dsl
3088 * properties for these, so remove them from the source nvlist after
3092 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
3093 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
3094 nvlist_t *zplprops, boolean_t *is_ci)
3096 uint64_t sense = ZFS_PROP_UNDEFINED;
3097 uint64_t norm = ZFS_PROP_UNDEFINED;
3098 uint64_t u8 = ZFS_PROP_UNDEFINED;
3100 ASSERT(zplprops != NULL);
3103 * Pull out creator prop choices, if any.
3106 (void) nvlist_lookup_uint64(createprops,
3107 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
3108 (void) nvlist_lookup_uint64(createprops,
3109 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
3110 (void) nvlist_remove_all(createprops,
3111 zfs_prop_to_name(ZFS_PROP_NORMALIZE));
3112 (void) nvlist_lookup_uint64(createprops,
3113 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
3114 (void) nvlist_remove_all(createprops,
3115 zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
3116 (void) nvlist_lookup_uint64(createprops,
3117 zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3118 (void) nvlist_remove_all(createprops,
3119 zfs_prop_to_name(ZFS_PROP_CASE));
3123 * If the zpl version requested is whacky or the file system
3124 * or pool is version is too "young" to support normalization
3125 * and the creator tried to set a value for one of the props,
3128 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3129 (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3130 (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3131 (zplver < ZPL_VERSION_NORMALIZATION &&
3132 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3133 sense != ZFS_PROP_UNDEFINED)))
3134 return (SET_ERROR(ENOTSUP));
3137 * Put the version in the zplprops
3139 VERIFY(nvlist_add_uint64(zplprops,
3140 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3142 if (norm == ZFS_PROP_UNDEFINED)
3143 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
3144 VERIFY(nvlist_add_uint64(zplprops,
3145 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3148 * If we're normalizing, names must always be valid UTF-8 strings.
3152 if (u8 == ZFS_PROP_UNDEFINED)
3153 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
3154 VERIFY(nvlist_add_uint64(zplprops,
3155 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3157 if (sense == ZFS_PROP_UNDEFINED)
3158 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
3159 VERIFY(nvlist_add_uint64(zplprops,
3160 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3163 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
3169 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3170 nvlist_t *zplprops, boolean_t *is_ci)
3172 boolean_t fuids_ok, sa_ok;
3173 uint64_t zplver = ZPL_VERSION;
3174 objset_t *os = NULL;
3175 char parentname[MAXNAMELEN];
3181 (void) strlcpy(parentname, dataset, sizeof (parentname));
3182 cp = strrchr(parentname, '/');
3186 if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3189 spa_vers = spa_version(spa);
3190 spa_close(spa, FTAG);
3192 zplver = zfs_zpl_version_map(spa_vers);
3193 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3194 sa_ok = (zplver >= ZPL_VERSION_SA);
3197 * Open parent object set so we can inherit zplprop values.
3199 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3202 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3204 dmu_objset_rele(os, FTAG);
3209 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3210 nvlist_t *zplprops, boolean_t *is_ci)
3214 uint64_t zplver = ZPL_VERSION;
3217 zplver = zfs_zpl_version_map(spa_vers);
3218 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3219 sa_ok = (zplver >= ZPL_VERSION_SA);
3221 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3222 createprops, zplprops, is_ci);
3228 * "type" -> dmu_objset_type_t (int32)
3229 * (optional) "props" -> { prop -> value }
3232 * outnvl: propname -> error code (int32)
3235 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3238 zfs_creat_t zct = { 0 };
3239 nvlist_t *nvprops = NULL;
3240 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3242 dmu_objset_type_t type;
3243 boolean_t is_insensitive = B_FALSE;
3245 if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3246 return (SET_ERROR(EINVAL));
3248 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3252 cbfunc = zfs_create_cb;
3256 cbfunc = zvol_create_cb;
3263 if (strchr(fsname, '@') ||
3264 strchr(fsname, '%'))
3265 return (SET_ERROR(EINVAL));
3267 zct.zct_props = nvprops;
3270 return (SET_ERROR(EINVAL));
3272 if (type == DMU_OST_ZVOL) {
3273 uint64_t volsize, volblocksize;
3275 if (nvprops == NULL)
3276 return (SET_ERROR(EINVAL));
3277 if (nvlist_lookup_uint64(nvprops,
3278 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3279 return (SET_ERROR(EINVAL));
3281 if ((error = nvlist_lookup_uint64(nvprops,
3282 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3283 &volblocksize)) != 0 && error != ENOENT)
3284 return (SET_ERROR(EINVAL));
3287 volblocksize = zfs_prop_default_numeric(
3288 ZFS_PROP_VOLBLOCKSIZE);
3290 if ((error = zvol_check_volblocksize(
3291 volblocksize)) != 0 ||
3292 (error = zvol_check_volsize(volsize,
3293 volblocksize)) != 0)
3295 } else if (type == DMU_OST_ZFS) {
3299 * We have to have normalization and
3300 * case-folding flags correct when we do the
3301 * file system creation, so go figure them out
3304 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3305 NV_UNIQUE_NAME, KM_SLEEP) == 0);
3306 error = zfs_fill_zplprops(fsname, nvprops,
3307 zct.zct_zplprops, &is_insensitive);
3309 nvlist_free(zct.zct_zplprops);
3314 error = dmu_objset_create(fsname, type,
3315 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3316 nvlist_free(zct.zct_zplprops);
3319 * It would be nice to do this atomically.
3322 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3325 (void) dsl_destroy_head(fsname);
3328 if (error == 0 && type == DMU_OST_ZVOL)
3329 zvol_create_minors(fsname);
3336 * "origin" -> name of origin snapshot
3337 * (optional) "props" -> { prop -> value }
3340 * outnvl: propname -> error code (int32)
3343 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3346 nvlist_t *nvprops = NULL;
3349 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3350 return (SET_ERROR(EINVAL));
3351 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3353 if (strchr(fsname, '@') ||
3354 strchr(fsname, '%'))
3355 return (SET_ERROR(EINVAL));
3357 if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3358 return (SET_ERROR(EINVAL));
3359 error = dmu_objset_clone(fsname, origin_name);
3364 * It would be nice to do this atomically.
3367 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3370 (void) dsl_destroy_head(fsname);
3374 zvol_create_minors(fsname);
3381 * "snaps" -> { snapshot1, snapshot2 }
3382 * (optional) "props" -> { prop -> value (string) }
3385 * outnvl: snapshot -> error code (int32)
3388 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3391 nvlist_t *props = NULL;
3395 (void) nvlist_lookup_nvlist(innvl, "props", &props);
3396 if ((error = zfs_check_userprops(poolname, props)) != 0)
3399 if (!nvlist_empty(props) &&
3400 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3401 return (SET_ERROR(ENOTSUP));
3403 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3404 return (SET_ERROR(EINVAL));
3405 poollen = strlen(poolname);
3406 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3407 pair = nvlist_next_nvpair(snaps, pair)) {
3408 const char *name = nvpair_name(pair);
3409 const char *cp = strchr(name, '@');
3412 * The snap name must contain an @, and the part after it must
3413 * contain only valid characters.
3416 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3417 return (SET_ERROR(EINVAL));
3420 * The snap must be in the specified pool.
3422 if (strncmp(name, poolname, poollen) != 0 ||
3423 (name[poollen] != '/' && name[poollen] != '@'))
3424 return (SET_ERROR(EXDEV));
3426 /* This must be the only snap of this fs. */
3427 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3428 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3429 if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3431 return (SET_ERROR(EXDEV));
3436 error = dsl_dataset_snapshot(snaps, props, outnvl);
3441 * innvl: "message" -> string
3445 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3453 * The poolname in the ioctl is not set, we get it from the TSD,
3454 * which was set at the end of the last successful ioctl that allows
3455 * logging. The secpolicy func already checked that it is set.
3456 * Only one log ioctl is allowed after each successful ioctl, so
3457 * we clear the TSD here.
3459 poolname = tsd_get(zfs_allow_log_key);
3460 (void) tsd_set(zfs_allow_log_key, NULL);
3461 error = spa_open(poolname, &spa, FTAG);
3466 if (nvlist_lookup_string(innvl, "message", &message) != 0) {
3467 spa_close(spa, FTAG);
3468 return (SET_ERROR(EINVAL));
3471 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3472 spa_close(spa, FTAG);
3473 return (SET_ERROR(ENOTSUP));
3476 error = spa_history_log(spa, message);
3477 spa_close(spa, FTAG);
3482 * The dp_config_rwlock must not be held when calling this, because the
3483 * unmount may need to write out data.
3485 * This function is best-effort. Callers must deal gracefully if it
3486 * remains mounted (or is remounted after this call).
3488 * Returns 0 if the argument is not a snapshot, or it is not currently a
3489 * filesystem, or we were able to unmount it. Returns error code otherwise.
3492 zfs_unmount_snap(const char *snapname)
3498 if (strchr(snapname, '@') == NULL)
3501 vfsp = zfs_get_vfs(snapname);
3505 zfsvfs = vfsp->vfs_data;
3506 ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os)));
3508 err = vn_vfswlock(vfsp->vfs_vnodecovered);
3511 return (SET_ERROR(err));
3514 * Always force the unmount for snapshots.
3518 (void) dounmount(vfsp, MS_FORCE, kcred);
3520 mtx_lock(&Giant); /* dounmount() */
3521 (void) dounmount(vfsp, MS_FORCE, curthread);
3522 mtx_unlock(&Giant); /* dounmount() */
3529 zfs_unmount_snap_cb(const char *snapname, void *arg)
3531 return (zfs_unmount_snap(snapname));
3535 * When a clone is destroyed, its origin may also need to be destroyed,
3536 * in which case it must be unmounted. This routine will do that unmount
3540 zfs_destroy_unmount_origin(const char *fsname)
3546 error = dmu_objset_hold(fsname, FTAG, &os);
3549 ds = dmu_objset_ds(os);
3550 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3551 char originname[MAXNAMELEN];
3552 dsl_dataset_name(ds->ds_prev, originname);
3553 dmu_objset_rele(os, FTAG);
3554 (void) zfs_unmount_snap(originname);
3556 dmu_objset_rele(os, FTAG);
3562 * "snaps" -> { snapshot1, snapshot2 }
3563 * (optional boolean) "defer"
3566 * outnvl: snapshot -> error code (int32)
3571 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3577 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3578 return (SET_ERROR(EINVAL));
3579 defer = nvlist_exists(innvl, "defer");
3581 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3582 pair = nvlist_next_nvpair(snaps, pair)) {
3583 (void) zfs_unmount_snap(nvpair_name(pair));
3586 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3590 * Create bookmarks. Bookmark names are of the form <fs>#<bmark>.
3591 * All bookmarks must be in the same pool.
3594 * bookmark1 -> snapshot1, bookmark2 -> snapshot2
3597 * outnvl: bookmark -> error code (int32)
3602 zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3604 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3605 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3609 * Verify the snapshot argument.
3611 if (nvpair_value_string(pair, &snap_name) != 0)
3612 return (SET_ERROR(EINVAL));
3615 /* Verify that the keys (bookmarks) are unique */
3616 for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair);
3617 pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) {
3618 if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0)
3619 return (SET_ERROR(EINVAL));
3623 return (dsl_bookmark_create(innvl, outnvl));
3628 * property 1, property 2, ...
3632 * bookmark name 1 -> { property 1, property 2, ... },
3633 * bookmark name 2 -> { property 1, property 2, ... }
3638 zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3640 return (dsl_get_bookmarks(fsname, innvl, outnvl));
3645 * bookmark name 1, bookmark name 2
3648 * outnvl: bookmark -> error code (int32)
3652 zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl,
3657 poollen = strlen(poolname);
3658 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3659 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3660 const char *name = nvpair_name(pair);
3661 const char *cp = strchr(name, '#');
3664 * The bookmark name must contain an #, and the part after it
3665 * must contain only valid characters.
3668 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3669 return (SET_ERROR(EINVAL));
3672 * The bookmark must be in the specified pool.
3674 if (strncmp(name, poolname, poollen) != 0 ||
3675 (name[poollen] != '/' && name[poollen] != '#'))
3676 return (SET_ERROR(EXDEV));
3677 (void) zvol_remove_minor(name);
3680 error = dsl_bookmark_destroy(innvl, outnvl);
3686 * zc_name name of dataset to destroy
3687 * zc_objset_type type of objset
3688 * zc_defer_destroy mark for deferred destroy
3693 zfs_ioc_destroy(zfs_cmd_t *zc)
3697 if (zc->zc_objset_type == DMU_OST_ZFS) {
3698 err = zfs_unmount_snap(zc->zc_name);
3703 if (strchr(zc->zc_name, '@'))
3704 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3706 err = dsl_destroy_head(zc->zc_name);
3707 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3708 (void) zvol_remove_minor(zc->zc_name);
3713 * fsname is name of dataset to rollback (to most recent snapshot)
3715 * innvl is not used.
3717 * outnvl: "target" -> name of most recent snapshot
3722 zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl)
3727 if (getzfsvfs(fsname, &zfsvfs) == 0) {
3728 error = zfs_suspend_fs(zfsvfs);
3732 error = dsl_dataset_rollback(fsname, zfsvfs, outnvl);
3733 resume_err = zfs_resume_fs(zfsvfs, fsname);
3734 error = error ? error : resume_err;
3736 VFS_RELE(zfsvfs->z_vfs);
3738 error = dsl_dataset_rollback(fsname, NULL, outnvl);
3744 recursive_unmount(const char *fsname, void *arg)
3746 const char *snapname = arg;
3747 char fullname[MAXNAMELEN];
3749 (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname);
3750 return (zfs_unmount_snap(fullname));
3755 * zc_name old name of dataset
3756 * zc_value new name of dataset
3757 * zc_cookie recursive flag (only valid for snapshots)
3762 zfs_ioc_rename(zfs_cmd_t *zc)
3764 boolean_t recursive = zc->zc_cookie & 1;
3766 boolean_t allow_mounted = zc->zc_cookie & 2;
3770 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3771 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3772 strchr(zc->zc_value, '%'))
3773 return (SET_ERROR(EINVAL));
3775 at = strchr(zc->zc_name, '@');
3777 /* snaps must be in same fs */
3780 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
3781 return (SET_ERROR(EXDEV));
3784 if (zc->zc_objset_type == DMU_OST_ZFS) {
3786 if (zc->zc_objset_type == DMU_OST_ZFS && allow_mounted) {
3788 error = dmu_objset_find(zc->zc_name,
3789 recursive_unmount, at + 1,
3790 recursive ? DS_FIND_CHILDREN : 0);
3796 error = dsl_dataset_rename_snapshot(zc->zc_name,
3797 at + 1, strchr(zc->zc_value, '@') + 1, recursive);
3803 if (zc->zc_objset_type == DMU_OST_ZVOL)
3804 (void) zvol_remove_minor(zc->zc_name);
3806 return (dsl_dir_rename(zc->zc_name, zc->zc_value));
3811 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3813 const char *propname = nvpair_name(pair);
3814 boolean_t issnap = (strchr(dsname, '@') != NULL);
3815 zfs_prop_t prop = zfs_name_to_prop(propname);
3819 if (prop == ZPROP_INVAL) {
3820 if (zfs_prop_user(propname)) {
3821 if (err = zfs_secpolicy_write_perms(dsname,
3822 ZFS_DELEG_PERM_USERPROP, cr))
3827 if (!issnap && zfs_prop_userquota(propname)) {
3828 const char *perm = NULL;
3829 const char *uq_prefix =
3830 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3831 const char *gq_prefix =
3832 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3834 if (strncmp(propname, uq_prefix,
3835 strlen(uq_prefix)) == 0) {
3836 perm = ZFS_DELEG_PERM_USERQUOTA;
3837 } else if (strncmp(propname, gq_prefix,
3838 strlen(gq_prefix)) == 0) {
3839 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3841 /* USERUSED and GROUPUSED are read-only */
3842 return (SET_ERROR(EINVAL));
3845 if (err = zfs_secpolicy_write_perms(dsname, perm, cr))
3850 return (SET_ERROR(EINVAL));
3854 return (SET_ERROR(EINVAL));
3856 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3858 * dsl_prop_get_all_impl() returns properties in this
3862 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3863 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3868 * Check that this value is valid for this pool version
3871 case ZFS_PROP_COMPRESSION:
3873 * If the user specified gzip compression, make sure
3874 * the SPA supports it. We ignore any errors here since
3875 * we'll catch them later.
3877 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3878 nvpair_value_uint64(pair, &intval) == 0) {
3879 if (intval >= ZIO_COMPRESS_GZIP_1 &&
3880 intval <= ZIO_COMPRESS_GZIP_9 &&
3881 zfs_earlier_version(dsname,
3882 SPA_VERSION_GZIP_COMPRESSION)) {
3883 return (SET_ERROR(ENOTSUP));
3886 if (intval == ZIO_COMPRESS_ZLE &&
3887 zfs_earlier_version(dsname,
3888 SPA_VERSION_ZLE_COMPRESSION))
3889 return (SET_ERROR(ENOTSUP));
3891 if (intval == ZIO_COMPRESS_LZ4) {
3894 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3897 if (!spa_feature_is_enabled(spa,
3898 SPA_FEATURE_LZ4_COMPRESS)) {
3899 spa_close(spa, FTAG);
3900 return (SET_ERROR(ENOTSUP));
3902 spa_close(spa, FTAG);
3906 * If this is a bootable dataset then
3907 * verify that the compression algorithm
3908 * is supported for booting. We must return
3909 * something other than ENOTSUP since it
3910 * implies a downrev pool version.
3912 if (zfs_is_bootfs(dsname) &&
3913 !BOOTFS_COMPRESS_VALID(intval)) {
3914 return (SET_ERROR(ERANGE));
3919 case ZFS_PROP_COPIES:
3920 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3921 return (SET_ERROR(ENOTSUP));
3924 case ZFS_PROP_DEDUP:
3925 if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3926 return (SET_ERROR(ENOTSUP));
3929 case ZFS_PROP_SHARESMB:
3930 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3931 return (SET_ERROR(ENOTSUP));
3934 case ZFS_PROP_ACLINHERIT:
3935 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3936 nvpair_value_uint64(pair, &intval) == 0) {
3937 if (intval == ZFS_ACL_PASSTHROUGH_X &&
3938 zfs_earlier_version(dsname,
3939 SPA_VERSION_PASSTHROUGH_X))
3940 return (SET_ERROR(ENOTSUP));
3945 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3949 * Checks for a race condition to make sure we don't increment a feature flag
3953 zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx)
3955 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3956 spa_feature_t *featurep = arg;
3958 if (!spa_feature_is_active(spa, *featurep))
3961 return (SET_ERROR(EBUSY));
3965 * The callback invoked on feature activation in the sync task caused by
3966 * zfs_prop_activate_feature.
3969 zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx)
3971 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3972 spa_feature_t *featurep = arg;
3974 spa_feature_incr(spa, *featurep, tx);
3978 * Activates a feature on a pool in response to a property setting. This
3979 * creates a new sync task which modifies the pool to reflect the feature
3983 zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature)
3987 /* EBUSY here indicates that the feature is already active */
3988 err = dsl_sync_task(spa_name(spa),
3989 zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync,
3992 if (err != 0 && err != EBUSY)
3999 * Removes properties from the given props list that fail permission checks
4000 * needed to clear them and to restore them in case of a receive error. For each
4001 * property, make sure we have both set and inherit permissions.
4003 * Returns the first error encountered if any permission checks fail. If the
4004 * caller provides a non-NULL errlist, it also gives the complete list of names
4005 * of all the properties that failed a permission check along with the
4006 * corresponding error numbers. The caller is responsible for freeing the
4009 * If every property checks out successfully, zero is returned and the list
4010 * pointed at by errlist is NULL.
4013 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
4016 nvpair_t *pair, *next_pair;
4023 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4025 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
4026 (void) strcpy(zc->zc_name, dataset);
4027 pair = nvlist_next_nvpair(props, NULL);
4028 while (pair != NULL) {
4029 next_pair = nvlist_next_nvpair(props, pair);
4031 (void) strcpy(zc->zc_value, nvpair_name(pair));
4032 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
4033 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
4034 VERIFY(nvlist_remove_nvpair(props, pair) == 0);
4035 VERIFY(nvlist_add_int32(errors,
4036 zc->zc_value, err) == 0);
4040 kmem_free(zc, sizeof (zfs_cmd_t));
4042 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
4043 nvlist_free(errors);
4046 VERIFY(nvpair_value_int32(pair, &rv) == 0);
4049 if (errlist == NULL)
4050 nvlist_free(errors);
4058 propval_equals(nvpair_t *p1, nvpair_t *p2)
4060 if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
4061 /* dsl_prop_get_all_impl() format */
4063 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
4064 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4068 if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
4070 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
4071 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4075 if (nvpair_type(p1) != nvpair_type(p2))
4078 if (nvpair_type(p1) == DATA_TYPE_STRING) {
4079 char *valstr1, *valstr2;
4081 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
4082 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
4083 return (strcmp(valstr1, valstr2) == 0);
4085 uint64_t intval1, intval2;
4087 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
4088 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
4089 return (intval1 == intval2);
4094 * Remove properties from props if they are not going to change (as determined
4095 * by comparison with origprops). Remove them from origprops as well, since we
4096 * do not need to clear or restore properties that won't change.
4099 props_reduce(nvlist_t *props, nvlist_t *origprops)
4101 nvpair_t *pair, *next_pair;
4103 if (origprops == NULL)
4104 return; /* all props need to be received */
4106 pair = nvlist_next_nvpair(props, NULL);
4107 while (pair != NULL) {
4108 const char *propname = nvpair_name(pair);
4111 next_pair = nvlist_next_nvpair(props, pair);
4113 if ((nvlist_lookup_nvpair(origprops, propname,
4114 &match) != 0) || !propval_equals(pair, match))
4115 goto next; /* need to set received value */
4117 /* don't clear the existing received value */
4118 (void) nvlist_remove_nvpair(origprops, match);
4119 /* don't bother receiving the property */
4120 (void) nvlist_remove_nvpair(props, pair);
4127 static boolean_t zfs_ioc_recv_inject_err;
4132 * zc_name name of containing filesystem
4133 * zc_nvlist_src{_size} nvlist of properties to apply
4134 * zc_value name of snapshot to create
4135 * zc_string name of clone origin (if DRR_FLAG_CLONE)
4136 * zc_cookie file descriptor to recv from
4137 * zc_begin_record the BEGIN record of the stream (not byteswapped)
4138 * zc_guid force flag
4139 * zc_cleanup_fd cleanup-on-exit file descriptor
4140 * zc_action_handle handle for this guid/ds mapping (or zero on first call)
4143 * zc_cookie number of bytes read
4144 * zc_nvlist_dst{_size} error for each unapplied received property
4145 * zc_obj zprop_errflags_t
4146 * zc_action_handle handle for this guid/ds mapping
4149 zfs_ioc_recv(zfs_cmd_t *zc)
4152 dmu_recv_cookie_t drc;
4153 boolean_t force = (boolean_t)zc->zc_guid;
4156 int props_error = 0;
4159 nvlist_t *props = NULL; /* sent properties */
4160 nvlist_t *origprops = NULL; /* existing properties */
4161 char *origin = NULL;
4163 char tofs[ZFS_MAXNAMELEN];
4164 boolean_t first_recvd_props = B_FALSE;
4166 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
4167 strchr(zc->zc_value, '@') == NULL ||
4168 strchr(zc->zc_value, '%'))
4169 return (SET_ERROR(EINVAL));
4171 (void) strcpy(tofs, zc->zc_value);
4172 tosnap = strchr(tofs, '@');
4175 if (zc->zc_nvlist_src != 0 &&
4176 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
4177 zc->zc_iflags, &props)) != 0)
4184 return (SET_ERROR(EBADF));
4187 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4189 if (zc->zc_string[0])
4190 origin = zc->zc_string;
4192 error = dmu_recv_begin(tofs, tosnap,
4193 &zc->zc_begin_record, force, origin, &drc);
4198 * Set properties before we receive the stream so that they are applied
4199 * to the new data. Note that we must call dmu_recv_stream() if
4200 * dmu_recv_begin() succeeds.
4202 if (props != NULL && !drc.drc_newfs) {
4203 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
4204 SPA_VERSION_RECVD_PROPS &&
4205 !dsl_prop_get_hasrecvd(tofs))
4206 first_recvd_props = B_TRUE;
4209 * If new received properties are supplied, they are to
4210 * completely replace the existing received properties, so stash
4211 * away the existing ones.
4213 if (dsl_prop_get_received(tofs, &origprops) == 0) {
4214 nvlist_t *errlist = NULL;
4216 * Don't bother writing a property if its value won't
4217 * change (and avoid the unnecessary security checks).
4219 * The first receive after SPA_VERSION_RECVD_PROPS is a
4220 * special case where we blow away all local properties
4223 if (!first_recvd_props)
4224 props_reduce(props, origprops);
4225 if (zfs_check_clearable(tofs, origprops, &errlist) != 0)
4226 (void) nvlist_merge(errors, errlist, 0);
4227 nvlist_free(errlist);
4229 if (clear_received_props(tofs, origprops,
4230 first_recvd_props ? NULL : props) != 0)
4231 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4233 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4237 if (props != NULL) {
4238 props_error = dsl_prop_set_hasrecvd(tofs);
4240 if (props_error == 0) {
4241 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4246 if (zc->zc_nvlist_dst_size != 0 &&
4247 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4248 put_nvlist(zc, errors) != 0)) {
4250 * Caller made zc->zc_nvlist_dst less than the minimum expected
4251 * size or supplied an invalid address.
4253 props_error = SET_ERROR(EINVAL);
4257 error = dmu_recv_stream(&drc, fp, &off, zc->zc_cleanup_fd,
4258 &zc->zc_action_handle);
4261 zfsvfs_t *zfsvfs = NULL;
4263 if (getzfsvfs(tofs, &zfsvfs) == 0) {
4267 error = zfs_suspend_fs(zfsvfs);
4269 * If the suspend fails, then the recv_end will
4270 * likely also fail, and clean up after itself.
4272 end_err = dmu_recv_end(&drc, zfsvfs);
4274 error = zfs_resume_fs(zfsvfs, tofs);
4275 error = error ? error : end_err;
4276 VFS_RELE(zfsvfs->z_vfs);
4278 error = dmu_recv_end(&drc, NULL);
4282 zc->zc_cookie = off - fp->f_offset;
4283 if (off >= 0 && off <= MAXOFFSET_T)
4287 if (zfs_ioc_recv_inject_err) {
4288 zfs_ioc_recv_inject_err = B_FALSE;
4295 zvol_create_minors(tofs);
4299 * On error, restore the original props.
4301 if (error != 0 && props != NULL && !drc.drc_newfs) {
4302 if (clear_received_props(tofs, props, NULL) != 0) {
4304 * We failed to clear the received properties.
4305 * Since we may have left a $recvd value on the
4306 * system, we can't clear the $hasrecvd flag.
4308 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4309 } else if (first_recvd_props) {
4310 dsl_prop_unset_hasrecvd(tofs);
4313 if (origprops == NULL && !drc.drc_newfs) {
4314 /* We failed to stash the original properties. */
4315 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4319 * dsl_props_set() will not convert RECEIVED to LOCAL on or
4320 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4321 * explictly if we're restoring local properties cleared in the
4322 * first new-style receive.
4324 if (origprops != NULL &&
4325 zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4326 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4327 origprops, NULL) != 0) {
4329 * We stashed the original properties but failed to
4332 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4337 nvlist_free(origprops);
4338 nvlist_free(errors);
4342 error = props_error;
4349 * zc_name name of snapshot to send
4350 * zc_cookie file descriptor to send stream to
4351 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj)
4352 * zc_sendobj objsetid of snapshot to send
4353 * zc_fromobj objsetid of incremental fromsnap (may be zero)
4354 * zc_guid if set, estimate size of stream only. zc_cookie is ignored.
4355 * output size in zc_objset_type.
4358 * zc_objset_type estimated size, if zc_guid is set
4361 zfs_ioc_send(zfs_cmd_t *zc)
4365 boolean_t estimate = (zc->zc_guid != 0);
4367 if (zc->zc_obj != 0) {
4369 dsl_dataset_t *tosnap;
4371 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4375 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4377 dsl_pool_rele(dp, FTAG);
4381 if (dsl_dir_is_clone(tosnap->ds_dir))
4382 zc->zc_fromobj = tosnap->ds_dir->dd_phys->dd_origin_obj;
4383 dsl_dataset_rele(tosnap, FTAG);
4384 dsl_pool_rele(dp, FTAG);
4389 dsl_dataset_t *tosnap;
4390 dsl_dataset_t *fromsnap = NULL;
4392 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4396 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4398 dsl_pool_rele(dp, FTAG);
4402 if (zc->zc_fromobj != 0) {
4403 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
4406 dsl_dataset_rele(tosnap, FTAG);
4407 dsl_pool_rele(dp, FTAG);
4412 error = dmu_send_estimate(tosnap, fromsnap,
4413 &zc->zc_objset_type);
4415 if (fromsnap != NULL)
4416 dsl_dataset_rele(fromsnap, FTAG);
4417 dsl_dataset_rele(tosnap, FTAG);
4418 dsl_pool_rele(dp, FTAG);
4420 file_t *fp = getf(zc->zc_cookie);
4422 return (SET_ERROR(EBADF));
4425 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
4427 zc->zc_fromobj, zc->zc_cookie, fp->f_vnode, &off);
4429 zc->zc_fromobj, zc->zc_cookie, fp, &off);
4432 if (off >= 0 && off <= MAXOFFSET_T)
4434 releasef(zc->zc_cookie);
4441 * zc_name name of snapshot on which to report progress
4442 * zc_cookie file descriptor of send stream
4445 * zc_cookie number of bytes written in send stream thus far
4448 zfs_ioc_send_progress(zfs_cmd_t *zc)
4452 dmu_sendarg_t *dsp = NULL;
4455 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4459 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4461 dsl_pool_rele(dp, FTAG);
4465 mutex_enter(&ds->ds_sendstream_lock);
4468 * Iterate over all the send streams currently active on this dataset.
4469 * If there's one which matches the specified file descriptor _and_ the
4470 * stream was started by the current process, return the progress of
4473 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4474 dsp = list_next(&ds->ds_sendstreams, dsp)) {
4475 if (dsp->dsa_outfd == zc->zc_cookie &&
4476 dsp->dsa_proc == curproc)
4481 zc->zc_cookie = *(dsp->dsa_off);
4483 error = SET_ERROR(ENOENT);
4485 mutex_exit(&ds->ds_sendstream_lock);
4486 dsl_dataset_rele(ds, FTAG);
4487 dsl_pool_rele(dp, FTAG);
4492 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4496 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4497 &zc->zc_inject_record);
4500 zc->zc_guid = (uint64_t)id;
4506 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4508 return (zio_clear_fault((int)zc->zc_guid));
4512 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4514 int id = (int)zc->zc_guid;
4517 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4518 &zc->zc_inject_record);
4526 zfs_ioc_error_log(zfs_cmd_t *zc)
4530 size_t count = (size_t)zc->zc_nvlist_dst_size;
4532 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4535 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4538 zc->zc_nvlist_dst_size = count;
4540 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4542 spa_close(spa, FTAG);
4548 zfs_ioc_clear(zfs_cmd_t *zc)
4555 * On zpool clear we also fix up missing slogs
4557 mutex_enter(&spa_namespace_lock);
4558 spa = spa_lookup(zc->zc_name);
4560 mutex_exit(&spa_namespace_lock);
4561 return (SET_ERROR(EIO));
4563 if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4564 /* we need to let spa_open/spa_load clear the chains */
4565 spa_set_log_state(spa, SPA_LOG_CLEAR);
4567 spa->spa_last_open_failed = 0;
4568 mutex_exit(&spa_namespace_lock);
4570 if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4571 error = spa_open(zc->zc_name, &spa, FTAG);
4574 nvlist_t *config = NULL;
4576 if (zc->zc_nvlist_src == 0)
4577 return (SET_ERROR(EINVAL));
4579 if ((error = get_nvlist(zc->zc_nvlist_src,
4580 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4581 error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4583 if (config != NULL) {
4586 if ((err = put_nvlist(zc, config)) != 0)
4588 nvlist_free(config);
4590 nvlist_free(policy);
4597 spa_vdev_state_enter(spa, SCL_NONE);
4599 if (zc->zc_guid == 0) {
4602 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4604 (void) spa_vdev_state_exit(spa, NULL, ENODEV);
4605 spa_close(spa, FTAG);
4606 return (SET_ERROR(ENODEV));
4610 vdev_clear(spa, vd);
4612 (void) spa_vdev_state_exit(spa, NULL, 0);
4615 * Resume any suspended I/Os.
4617 if (zio_resume(spa) != 0)
4618 error = SET_ERROR(EIO);
4620 spa_close(spa, FTAG);
4626 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4631 error = spa_open(zc->zc_name, &spa, FTAG);
4635 spa_vdev_state_enter(spa, SCL_NONE);
4638 * If a resilver is already in progress then set the
4639 * spa_scrub_reopen flag to B_TRUE so that we don't restart
4640 * the scan as a side effect of the reopen. Otherwise, let
4641 * vdev_open() decided if a resilver is required.
4643 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4644 vdev_reopen(spa->spa_root_vdev);
4645 spa->spa_scrub_reopen = B_FALSE;
4647 (void) spa_vdev_state_exit(spa, NULL, 0);
4648 spa_close(spa, FTAG);
4653 * zc_name name of filesystem
4654 * zc_value name of origin snapshot
4657 * zc_string name of conflicting snapshot, if there is one
4660 zfs_ioc_promote(zfs_cmd_t *zc)
4665 * We don't need to unmount *all* the origin fs's snapshots, but
4668 cp = strchr(zc->zc_value, '@');
4671 (void) dmu_objset_find(zc->zc_value,
4672 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
4673 return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4677 * Retrieve a single {user|group}{used|quota}@... property.
4680 * zc_name name of filesystem
4681 * zc_objset_type zfs_userquota_prop_t
4682 * zc_value domain name (eg. "S-1-234-567-89")
4683 * zc_guid RID/UID/GID
4686 * zc_cookie property value
4689 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4694 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4695 return (SET_ERROR(EINVAL));
4697 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4701 error = zfs_userspace_one(zfsvfs,
4702 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4703 zfsvfs_rele(zfsvfs, FTAG);
4710 * zc_name name of filesystem
4711 * zc_cookie zap cursor
4712 * zc_objset_type zfs_userquota_prop_t
4713 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4716 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
4717 * zc_cookie zap cursor
4720 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4723 int bufsize = zc->zc_nvlist_dst_size;
4726 return (SET_ERROR(ENOMEM));
4728 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4732 void *buf = kmem_alloc(bufsize, KM_SLEEP);
4734 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
4735 buf, &zc->zc_nvlist_dst_size);
4738 error = ddi_copyout(buf,
4739 (void *)(uintptr_t)zc->zc_nvlist_dst,
4740 zc->zc_nvlist_dst_size, zc->zc_iflags);
4742 kmem_free(buf, bufsize);
4743 zfsvfs_rele(zfsvfs, FTAG);
4750 * zc_name name of filesystem
4756 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4762 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
4763 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
4765 * If userused is not enabled, it may be because the
4766 * objset needs to be closed & reopened (to grow the
4767 * objset_phys_t). Suspend/resume the fs will do that.
4769 error = zfs_suspend_fs(zfsvfs);
4771 dmu_objset_refresh_ownership(zfsvfs->z_os,
4773 error = zfs_resume_fs(zfsvfs, zc->zc_name);
4777 error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
4778 VFS_RELE(zfsvfs->z_vfs);
4780 /* XXX kind of reading contents without owning */
4781 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4785 error = dmu_objset_userspace_upgrade(os);
4786 dmu_objset_rele(os, FTAG);
4794 * We don't want to have a hard dependency
4795 * against some special symbols in sharefs
4796 * nfs, and smbsrv. Determine them if needed when
4797 * the first file system is shared.
4798 * Neither sharefs, nfs or smbsrv are unloadable modules.
4800 int (*znfsexport_fs)(void *arg);
4801 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
4802 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
4804 int zfs_nfsshare_inited;
4805 int zfs_smbshare_inited;
4807 ddi_modhandle_t nfs_mod;
4808 ddi_modhandle_t sharefs_mod;
4809 ddi_modhandle_t smbsrv_mod;
4811 kmutex_t zfs_share_lock;
4819 ASSERT(MUTEX_HELD(&zfs_share_lock));
4820 /* Both NFS and SMB shares also require sharetab support. */
4821 if (sharefs_mod == NULL && ((sharefs_mod =
4822 ddi_modopen("fs/sharefs",
4823 KRTLD_MODE_FIRST, &error)) == NULL)) {
4824 return (SET_ERROR(ENOSYS));
4826 if (zshare_fs == NULL && ((zshare_fs =
4827 (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
4828 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
4829 return (SET_ERROR(ENOSYS));
4836 zfs_ioc_share(zfs_cmd_t *zc)
4842 switch (zc->zc_share.z_sharetype) {
4844 case ZFS_UNSHARE_NFS:
4845 if (zfs_nfsshare_inited == 0) {
4846 mutex_enter(&zfs_share_lock);
4847 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
4848 KRTLD_MODE_FIRST, &error)) == NULL)) {
4849 mutex_exit(&zfs_share_lock);
4850 return (SET_ERROR(ENOSYS));
4852 if (znfsexport_fs == NULL &&
4853 ((znfsexport_fs = (int (*)(void *))
4855 "nfs_export", &error)) == NULL)) {
4856 mutex_exit(&zfs_share_lock);
4857 return (SET_ERROR(ENOSYS));
4859 error = zfs_init_sharefs();
4861 mutex_exit(&zfs_share_lock);
4862 return (SET_ERROR(ENOSYS));
4864 zfs_nfsshare_inited = 1;
4865 mutex_exit(&zfs_share_lock);
4869 case ZFS_UNSHARE_SMB:
4870 if (zfs_smbshare_inited == 0) {
4871 mutex_enter(&zfs_share_lock);
4872 if (smbsrv_mod == NULL && ((smbsrv_mod =
4873 ddi_modopen("drv/smbsrv",
4874 KRTLD_MODE_FIRST, &error)) == NULL)) {
4875 mutex_exit(&zfs_share_lock);
4876 return (SET_ERROR(ENOSYS));
4878 if (zsmbexport_fs == NULL && ((zsmbexport_fs =
4879 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
4880 "smb_server_share", &error)) == NULL)) {
4881 mutex_exit(&zfs_share_lock);
4882 return (SET_ERROR(ENOSYS));
4884 error = zfs_init_sharefs();
4886 mutex_exit(&zfs_share_lock);
4887 return (SET_ERROR(ENOSYS));
4889 zfs_smbshare_inited = 1;
4890 mutex_exit(&zfs_share_lock);
4894 return (SET_ERROR(EINVAL));
4897 switch (zc->zc_share.z_sharetype) {
4899 case ZFS_UNSHARE_NFS:
4901 znfsexport_fs((void *)
4902 (uintptr_t)zc->zc_share.z_exportdata))
4906 case ZFS_UNSHARE_SMB:
4907 if (error = zsmbexport_fs((void *)
4908 (uintptr_t)zc->zc_share.z_exportdata,
4909 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
4916 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
4917 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
4918 SHAREFS_ADD : SHAREFS_REMOVE;
4921 * Add or remove share from sharetab
4923 error = zshare_fs(opcode,
4924 (void *)(uintptr_t)zc->zc_share.z_sharedata,
4925 zc->zc_share.z_sharemax);
4934 ace_t full_access[] = {
4935 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4940 * zc_name name of containing filesystem
4941 * zc_obj object # beyond which we want next in-use object #
4944 * zc_obj next in-use object #
4947 zfs_ioc_next_obj(zfs_cmd_t *zc)
4949 objset_t *os = NULL;
4952 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4956 error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
4957 os->os_dsl_dataset->ds_phys->ds_prev_snap_txg);
4959 dmu_objset_rele(os, FTAG);
4965 * zc_name name of filesystem
4966 * zc_value prefix name for snapshot
4967 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
4970 * zc_value short name of new snapshot
4973 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
4980 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
4984 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
4985 (u_longlong_t)ddi_get_lbolt64());
4986 hold_name = kmem_asprintf("%%%s", zc->zc_value);
4988 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
4991 (void) strcpy(zc->zc_value, snap_name);
4994 zfs_onexit_fd_rele(zc->zc_cleanup_fd);
5000 * zc_name name of "to" snapshot
5001 * zc_value name of "from" snapshot
5002 * zc_cookie file descriptor to write diff data on
5005 * dmu_diff_record_t's to the file descriptor
5008 zfs_ioc_diff(zfs_cmd_t *zc)
5014 fp = getf(zc->zc_cookie);
5016 return (SET_ERROR(EBADF));
5021 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off);
5023 error = dmu_diff(zc->zc_name, zc->zc_value, fp, &off);
5026 if (off >= 0 && off <= MAXOFFSET_T)
5028 releasef(zc->zc_cookie);
5035 * Remove all ACL files in shares dir
5038 zfs_smb_acl_purge(znode_t *dzp)
5041 zap_attribute_t zap;
5042 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
5045 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
5046 (error = zap_cursor_retrieve(&zc, &zap)) == 0;
5047 zap_cursor_advance(&zc)) {
5048 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
5052 zap_cursor_fini(&zc);
5058 zfs_ioc_smb_acl(zfs_cmd_t *zc)
5063 vnode_t *resourcevp = NULL;
5072 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
5073 NO_FOLLOW, NULL, &vp)) != 0)
5076 /* Now make sure mntpnt and dataset are ZFS */
5078 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 ||
5079 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
5080 zc->zc_name) != 0)) {
5082 return (SET_ERROR(EINVAL));
5086 zfsvfs = dzp->z_zfsvfs;
5090 * Create share dir if its missing.
5092 mutex_enter(&zfsvfs->z_lock);
5093 if (zfsvfs->z_shares_dir == 0) {
5096 tx = dmu_tx_create(zfsvfs->z_os);
5097 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
5099 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
5100 error = dmu_tx_assign(tx, TXG_WAIT);
5104 error = zfs_create_share_dir(zfsvfs, tx);
5108 mutex_exit(&zfsvfs->z_lock);
5114 mutex_exit(&zfsvfs->z_lock);
5116 ASSERT(zfsvfs->z_shares_dir);
5117 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
5123 switch (zc->zc_cookie) {
5124 case ZFS_SMB_ACL_ADD:
5125 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
5126 vattr.va_type = VREG;
5127 vattr.va_mode = S_IFREG|0777;
5131 vsec.vsa_mask = VSA_ACE;
5132 vsec.vsa_aclentp = &full_access;
5133 vsec.vsa_aclentsz = sizeof (full_access);
5134 vsec.vsa_aclcnt = 1;
5136 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
5137 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
5139 VN_RELE(resourcevp);
5142 case ZFS_SMB_ACL_REMOVE:
5143 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
5147 case ZFS_SMB_ACL_RENAME:
5148 if ((error = get_nvlist(zc->zc_nvlist_src,
5149 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
5154 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
5155 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
5158 VN_RELE(ZTOV(sharedir));
5160 nvlist_free(nvlist);
5163 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
5165 nvlist_free(nvlist);
5168 case ZFS_SMB_ACL_PURGE:
5169 error = zfs_smb_acl_purge(sharedir);
5173 error = SET_ERROR(EINVAL);
5178 VN_RELE(ZTOV(sharedir));
5184 return (EOPNOTSUPP);
5190 * "holds" -> { snapname -> holdname (string), ... }
5191 * (optional) "cleanup_fd" -> fd (int32)
5195 * snapname -> error value (int32)
5201 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
5204 int cleanup_fd = -1;
5208 error = nvlist_lookup_nvlist(args, "holds", &holds);
5210 return (SET_ERROR(EINVAL));
5212 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
5213 error = zfs_onexit_fd_hold(cleanup_fd, &minor);
5218 error = dsl_dataset_user_hold(holds, minor, errlist);
5220 zfs_onexit_fd_rele(cleanup_fd);
5225 * innvl is not used.
5228 * holdname -> time added (uint64 seconds since epoch)
5234 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
5236 return (dsl_dataset_get_holds(snapname, outnvl));
5241 * snapname -> { holdname, ... }
5246 * snapname -> error value (int32)
5252 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
5254 return (dsl_dataset_user_release(holds, errlist));
5259 * zc_name name of new filesystem or snapshot
5260 * zc_value full name of old snapshot
5263 * zc_cookie space in bytes
5264 * zc_objset_type compressed space in bytes
5265 * zc_perm_action uncompressed space in bytes
5268 zfs_ioc_space_written(zfs_cmd_t *zc)
5272 dsl_dataset_t *new, *old;
5274 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5277 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
5279 dsl_pool_rele(dp, FTAG);
5282 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
5284 dsl_dataset_rele(new, FTAG);
5285 dsl_pool_rele(dp, FTAG);
5289 error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
5290 &zc->zc_objset_type, &zc->zc_perm_action);
5291 dsl_dataset_rele(old, FTAG);
5292 dsl_dataset_rele(new, FTAG);
5293 dsl_pool_rele(dp, FTAG);
5299 * "firstsnap" -> snapshot name
5303 * "used" -> space in bytes
5304 * "compressed" -> compressed space in bytes
5305 * "uncompressed" -> uncompressed space in bytes
5309 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
5313 dsl_dataset_t *new, *old;
5315 uint64_t used, comp, uncomp;
5317 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5318 return (SET_ERROR(EINVAL));
5320 error = dsl_pool_hold(lastsnap, FTAG, &dp);
5324 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
5326 dsl_pool_rele(dp, FTAG);
5329 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
5331 dsl_dataset_rele(new, FTAG);
5332 dsl_pool_rele(dp, FTAG);
5336 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5337 dsl_dataset_rele(old, FTAG);
5338 dsl_dataset_rele(new, FTAG);
5339 dsl_pool_rele(dp, FTAG);
5340 fnvlist_add_uint64(outnvl, "used", used);
5341 fnvlist_add_uint64(outnvl, "compressed", comp);
5342 fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5347 zfs_ioc_jail(zfs_cmd_t *zc)
5350 return (zone_dataset_attach(curthread->td_ucred, zc->zc_name,
5351 (int)zc->zc_jailid));
5355 zfs_ioc_unjail(zfs_cmd_t *zc)
5358 return (zone_dataset_detach(curthread->td_ucred, zc->zc_name,
5359 (int)zc->zc_jailid));
5364 * "fd" -> file descriptor to write stream to (int32)
5365 * (optional) "fromsnap" -> full snap name to send an incremental from
5372 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5376 char *fromname = NULL;
5379 error = nvlist_lookup_int32(innvl, "fd", &fd);
5381 return (SET_ERROR(EINVAL));
5383 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
5385 file_t *fp = getf(fd);
5387 return (SET_ERROR(EBADF));
5391 error = dmu_send(snapname, fromname, fd, fp->f_vnode, &off);
5393 error = dmu_send(snapname, fromname, fd, fp, &off);
5397 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5408 * Determine approximately how large a zfs send stream will be -- the number
5409 * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5412 * (optional) "fromsnap" -> full snap name to send an incremental from
5416 * "space" -> bytes of space (uint64)
5420 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5423 dsl_dataset_t *fromsnap = NULL;
5424 dsl_dataset_t *tosnap;
5429 error = dsl_pool_hold(snapname, FTAG, &dp);
5433 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
5435 dsl_pool_rele(dp, FTAG);
5439 error = nvlist_lookup_string(innvl, "fromsnap", &fromname);
5441 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
5443 dsl_dataset_rele(tosnap, FTAG);
5444 dsl_pool_rele(dp, FTAG);
5449 error = dmu_send_estimate(tosnap, fromsnap, &space);
5450 fnvlist_add_uint64(outnvl, "space", space);
5452 if (fromsnap != NULL)
5453 dsl_dataset_rele(fromsnap, FTAG);
5454 dsl_dataset_rele(tosnap, FTAG);
5455 dsl_pool_rele(dp, FTAG);
5460 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
5463 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5464 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5465 boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
5467 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5469 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5470 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5471 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5472 ASSERT3P(vec->zvec_func, ==, NULL);
5474 vec->zvec_legacy_func = func;
5475 vec->zvec_secpolicy = secpolicy;
5476 vec->zvec_namecheck = namecheck;
5477 vec->zvec_allow_log = log_history;
5478 vec->zvec_pool_check = pool_check;
5482 * See the block comment at the beginning of this file for details on
5483 * each argument to this function.
5486 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
5487 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5488 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
5489 boolean_t allow_log)
5491 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5493 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5494 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5495 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5496 ASSERT3P(vec->zvec_func, ==, NULL);
5498 /* if we are logging, the name must be valid */
5499 ASSERT(!allow_log || namecheck != NO_NAME);
5501 vec->zvec_name = name;
5502 vec->zvec_func = func;
5503 vec->zvec_secpolicy = secpolicy;
5504 vec->zvec_namecheck = namecheck;
5505 vec->zvec_pool_check = pool_check;
5506 vec->zvec_smush_outnvlist = smush_outnvlist;
5507 vec->zvec_allow_log = allow_log;
5511 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5512 zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
5513 zfs_ioc_poolcheck_t pool_check)
5515 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5516 POOL_NAME, log_history, pool_check);
5520 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5521 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
5523 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5524 DATASET_NAME, B_FALSE, pool_check);
5528 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5530 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
5531 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5535 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5536 zfs_secpolicy_func_t *secpolicy)
5538 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5539 NO_NAME, B_FALSE, POOL_CHECK_NONE);
5543 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
5544 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
5546 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5547 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
5551 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5553 zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
5554 zfs_secpolicy_read);
5558 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5559 zfs_secpolicy_func_t *secpolicy)
5561 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5562 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5566 zfs_ioctl_init(void)
5568 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
5569 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
5570 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5572 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
5573 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
5574 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
5576 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
5577 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
5578 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5580 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
5581 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
5582 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5584 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
5585 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
5586 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5588 zfs_ioctl_register("create", ZFS_IOC_CREATE,
5589 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
5590 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5592 zfs_ioctl_register("clone", ZFS_IOC_CLONE,
5593 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
5594 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5596 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
5597 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
5598 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5600 zfs_ioctl_register("hold", ZFS_IOC_HOLD,
5601 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
5602 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5603 zfs_ioctl_register("release", ZFS_IOC_RELEASE,
5604 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
5605 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5607 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
5608 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
5609 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5611 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
5612 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
5613 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE);
5615 zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK,
5616 zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME,
5617 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5619 zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS,
5620 zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME,
5621 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5623 zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS,
5624 zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks,
5626 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5628 /* IOCTLS that use the legacy function signature */
5630 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
5631 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
5633 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
5634 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5635 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
5637 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
5638 zfs_ioc_pool_upgrade);
5639 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
5641 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
5642 zfs_ioc_vdev_remove);
5643 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
5644 zfs_ioc_vdev_set_state);
5645 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
5646 zfs_ioc_vdev_attach);
5647 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
5648 zfs_ioc_vdev_detach);
5649 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
5650 zfs_ioc_vdev_setpath);
5651 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
5652 zfs_ioc_vdev_setfru);
5653 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
5654 zfs_ioc_pool_set_props);
5655 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
5656 zfs_ioc_vdev_split);
5657 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
5658 zfs_ioc_pool_reguid);
5660 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
5661 zfs_ioc_pool_configs, zfs_secpolicy_none);
5662 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
5663 zfs_ioc_pool_tryimport, zfs_secpolicy_config);
5664 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
5665 zfs_ioc_inject_fault, zfs_secpolicy_inject);
5666 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
5667 zfs_ioc_clear_fault, zfs_secpolicy_inject);
5668 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
5669 zfs_ioc_inject_list_next, zfs_secpolicy_inject);
5672 * pool destroy, and export don't log the history as part of
5673 * zfsdev_ioctl, but rather zfs_ioc_pool_export
5674 * does the logging of those commands.
5676 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
5677 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5678 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
5679 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5681 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
5682 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5683 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
5684 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5686 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
5687 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_NONE);
5688 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
5689 zfs_ioc_dsobj_to_dsname,
5690 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_NONE);
5691 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
5692 zfs_ioc_pool_get_history,
5693 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
5695 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
5696 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5698 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
5699 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5700 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
5701 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5703 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
5704 zfs_ioc_space_written);
5705 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
5706 zfs_ioc_objset_recvd_props);
5707 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
5709 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
5711 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
5712 zfs_ioc_objset_stats);
5713 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
5714 zfs_ioc_objset_zplprops);
5715 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
5716 zfs_ioc_dataset_list_next);
5717 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
5718 zfs_ioc_snapshot_list_next);
5719 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
5720 zfs_ioc_send_progress);
5722 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
5723 zfs_ioc_diff, zfs_secpolicy_diff);
5724 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
5725 zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
5726 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
5727 zfs_ioc_obj_to_path, zfs_secpolicy_diff);
5728 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
5729 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
5730 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
5731 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
5732 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
5733 zfs_ioc_send, zfs_secpolicy_send);
5735 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
5736 zfs_secpolicy_none);
5737 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
5738 zfs_secpolicy_destroy);
5739 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
5740 zfs_secpolicy_rename);
5741 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
5742 zfs_secpolicy_recv);
5743 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
5744 zfs_secpolicy_promote);
5745 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
5746 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
5747 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
5748 zfs_secpolicy_set_fsacl);
5750 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
5751 zfs_secpolicy_share, POOL_CHECK_NONE);
5752 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
5753 zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
5754 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
5755 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
5756 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5757 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
5758 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
5759 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5762 zfs_ioctl_register_dataset_nolog(ZFS_IOC_JAIL, zfs_ioc_jail,
5763 zfs_secpolicy_config, POOL_CHECK_NONE);
5764 zfs_ioctl_register_dataset_nolog(ZFS_IOC_UNJAIL, zfs_ioc_unjail,
5765 zfs_secpolicy_config, POOL_CHECK_NONE);
5770 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
5771 zfs_ioc_poolcheck_t check)
5776 ASSERT(type == POOL_NAME || type == DATASET_NAME);
5778 if (check & POOL_CHECK_NONE)
5781 error = spa_open(name, &spa, FTAG);
5783 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
5784 error = SET_ERROR(EAGAIN);
5785 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
5786 error = SET_ERROR(EROFS);
5787 spa_close(spa, FTAG);
5793 * Find a free minor number.
5796 zfsdev_minor_alloc(void)
5798 static minor_t last_minor;
5801 ASSERT(MUTEX_HELD(&spa_namespace_lock));
5803 for (m = last_minor + 1; m != last_minor; m++) {
5804 if (m > ZFSDEV_MAX_MINOR)
5806 if (ddi_get_soft_state(zfsdev_state, m) == NULL) {
5816 zfs_ctldev_init(struct cdev *devp)
5819 zfs_soft_state_t *zs;
5821 ASSERT(MUTEX_HELD(&spa_namespace_lock));
5823 minor = zfsdev_minor_alloc();
5825 return (SET_ERROR(ENXIO));
5827 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS)
5828 return (SET_ERROR(EAGAIN));
5830 devfs_set_cdevpriv((void *)(uintptr_t)minor, zfsdev_close);
5832 zs = ddi_get_soft_state(zfsdev_state, minor);
5833 zs->zss_type = ZSST_CTLDEV;
5834 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data);
5840 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor)
5842 ASSERT(MUTEX_HELD(&spa_namespace_lock));
5844 zfs_onexit_destroy(zo);
5845 ddi_soft_state_free(zfsdev_state, minor);
5849 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which)
5851 zfs_soft_state_t *zp;
5853 zp = ddi_get_soft_state(zfsdev_state, minor);
5854 if (zp == NULL || zp->zss_type != which)
5857 return (zp->zss_data);
5861 zfsdev_open(struct cdev *devp, int flag, int mode, struct thread *td)
5866 if (getminor(*devp) != 0)
5867 return (zvol_open(devp, flag, otyp, cr));
5870 /* This is the control device. Allocate a new minor if requested. */
5872 mutex_enter(&spa_namespace_lock);
5873 error = zfs_ctldev_init(devp);
5874 mutex_exit(&spa_namespace_lock);
5881 zfsdev_close(void *data)
5884 minor_t minor = (minor_t)(uintptr_t)data;
5889 mutex_enter(&spa_namespace_lock);
5890 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV);
5892 mutex_exit(&spa_namespace_lock);
5895 zfs_ctldev_destroy(zo, minor);
5896 mutex_exit(&spa_namespace_lock);
5900 zfsdev_ioctl(struct cdev *dev, u_long zcmd, caddr_t arg, int flag,
5907 minor_t minor = getminor(dev);
5909 zfs_iocparm_t *zc_iocparm;
5910 int cflag, cmd, oldvecnum;
5911 boolean_t newioc, compat;
5912 cred_t *cr = td->td_ucred;
5914 const zfs_ioc_vec_t *vec;
5915 char *saved_poolname = NULL;
5916 nvlist_t *innvl = NULL;
5918 cflag = ZFS_CMD_COMPAT_NONE;
5922 len = IOCPARM_LEN(zcmd);
5926 * Check if we are talking to supported older binaries
5927 * and translate zfs_cmd if necessary
5929 if (len != sizeof(zfs_iocparm_t)) {
5931 if (len == sizeof(zfs_cmd_t)) {
5932 cflag = ZFS_CMD_COMPAT_LZC;
5934 } else if (len == sizeof(zfs_cmd_deadman_t)) {
5935 cflag = ZFS_CMD_COMPAT_DEADMAN;
5938 } else if (len == sizeof(zfs_cmd_v28_t)) {
5939 cflag = ZFS_CMD_COMPAT_V28;
5942 } else if (len == sizeof(zfs_cmd_v15_t)) {
5943 cflag = ZFS_CMD_COMPAT_V15;
5945 vecnum = zfs_ioctl_v15_to_v28[cmd];
5952 vecnum = cmd - ZFS_IOC_FIRST;
5953 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
5957 if (vecnum == ZFS_IOC_COMPAT_PASS)
5959 else if (vecnum == ZFS_IOC_COMPAT_FAIL)
5964 * Check if we have sufficient kernel memory allocated
5965 * for the zfs_cmd_t request. Bail out if not so we
5966 * will not access undefined memory region.
5968 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
5969 return (SET_ERROR(EINVAL));
5970 vec = &zfs_ioc_vec[vecnum];
5973 zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP);
5974 bzero(zc, sizeof(zfs_cmd_t));
5976 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
5978 error = SET_ERROR(EFAULT);
5981 #else /* !illumos */
5983 * We don't alloc/free zc only if talking to library ioctl version 2
5985 if (cflag != ZFS_CMD_COMPAT_LZC) {
5986 zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP);
5987 bzero(zc, sizeof(zfs_cmd_t));
5994 zc_iocparm = (void *)arg;
5995 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_t)) {
5996 error = SET_ERROR(EFAULT);
5999 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd, zc,
6000 sizeof(zfs_cmd_t), flag);
6002 error = SET_ERROR(EFAULT);
6008 zfs_cmd_compat_get(zc, arg, cflag);
6010 error = zfs_ioctl_compat_pre(zc, &vecnum, cflag);
6013 if (oldvecnum != vecnum)
6014 vec = &zfs_ioc_vec[vecnum];
6016 #endif /* !illumos */
6018 zc->zc_iflags = flag & FKIOCTL;
6019 if (zc->zc_nvlist_src_size != 0) {
6020 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
6021 zc->zc_iflags, &innvl);
6026 /* rewrite innvl for backwards compatibility */
6028 innvl = zfs_ioctl_compat_innvl(zc, innvl, vecnum, cflag);
6031 * Ensure that all pool/dataset names are valid before we pass down to
6034 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
6035 switch (vec->zvec_namecheck) {
6037 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
6038 error = SET_ERROR(EINVAL);
6040 error = pool_status_check(zc->zc_name,
6041 vec->zvec_namecheck, vec->zvec_pool_check);
6045 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
6046 error = SET_ERROR(EINVAL);
6048 error = pool_status_check(zc->zc_name,
6049 vec->zvec_namecheck, vec->zvec_pool_check);
6056 if (error == 0 && !(flag & FKIOCTL))
6057 error = vec->zvec_secpolicy(zc, innvl, cr);
6062 /* legacy ioctls can modify zc_name */
6063 len = strcspn(zc->zc_name, "/@#") + 1;
6064 saved_poolname = kmem_alloc(len, KM_SLEEP);
6065 (void) strlcpy(saved_poolname, zc->zc_name, len);
6067 if (vec->zvec_func != NULL) {
6071 nvlist_t *lognv = NULL;
6073 ASSERT(vec->zvec_legacy_func == NULL);
6076 * Add the innvl to the lognv before calling the func,
6077 * in case the func changes the innvl.
6079 if (vec->zvec_allow_log) {
6080 lognv = fnvlist_alloc();
6081 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
6083 if (!nvlist_empty(innvl)) {
6084 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
6089 outnvl = fnvlist_alloc();
6090 error = vec->zvec_func(zc->zc_name, innvl, outnvl);
6092 if (error == 0 && vec->zvec_allow_log &&
6093 spa_open(zc->zc_name, &spa, FTAG) == 0) {
6094 if (!nvlist_empty(outnvl)) {
6095 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
6098 (void) spa_history_log_nvl(spa, lognv);
6099 spa_close(spa, FTAG);
6101 fnvlist_free(lognv);
6103 /* rewrite outnvl for backwards compatibility */
6104 if (cflag != ZFS_CMD_COMPAT_NONE && cflag != ZFS_CMD_COMPAT_LZC)
6105 outnvl = zfs_ioctl_compat_outnvl(zc, outnvl, vecnum,
6108 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
6110 if (vec->zvec_smush_outnvlist) {
6111 smusherror = nvlist_smush(outnvl,
6112 zc->zc_nvlist_dst_size);
6114 if (smusherror == 0)
6115 puterror = put_nvlist(zc, outnvl);
6121 nvlist_free(outnvl);
6123 error = vec->zvec_legacy_func(zc);
6130 zfs_ioctl_compat_post(zc, cmd, cflag);
6131 zfs_cmd_compat_put(zc, arg, vecnum, cflag);
6135 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
6136 if (error == 0 && rc != 0)
6137 error = SET_ERROR(EFAULT);
6140 rc = ddi_copyout(zc, (void *)(uintptr_t)zc_iocparm->zfs_cmd,
6141 sizeof (zfs_cmd_t), flag);
6142 if (error == 0 && rc != 0)
6143 error = SET_ERROR(EFAULT);
6146 if (error == 0 && vec->zvec_allow_log) {
6147 char *s = tsd_get(zfs_allow_log_key);
6150 (void) tsd_set(zfs_allow_log_key, saved_poolname);
6152 if (saved_poolname != NULL)
6153 strfree(saved_poolname);
6157 kmem_free(zc, sizeof (zfs_cmd_t));
6160 * We don't alloc/free zc only if talking to library ioctl version 2
6162 if (cflag != ZFS_CMD_COMPAT_LZC)
6163 kmem_free(zc, sizeof (zfs_cmd_t));
6170 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
6172 if (cmd != DDI_ATTACH)
6173 return (DDI_FAILURE);
6175 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
6176 DDI_PSEUDO, 0) == DDI_FAILURE)
6177 return (DDI_FAILURE);
6181 ddi_report_dev(dip);
6183 return (DDI_SUCCESS);
6187 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
6189 if (spa_busy() || zfs_busy() || zvol_busy())
6190 return (DDI_FAILURE);
6192 if (cmd != DDI_DETACH)
6193 return (DDI_FAILURE);
6197 ddi_prop_remove_all(dip);
6198 ddi_remove_minor_node(dip, NULL);
6200 return (DDI_SUCCESS);
6205 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
6208 case DDI_INFO_DEVT2DEVINFO:
6210 return (DDI_SUCCESS);
6212 case DDI_INFO_DEVT2INSTANCE:
6213 *result = (void *)0;
6214 return (DDI_SUCCESS);
6217 return (DDI_FAILURE);
6222 * OK, so this is a little weird.
6224 * /dev/zfs is the control node, i.e. minor 0.
6225 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
6227 * /dev/zfs has basically nothing to do except serve up ioctls,
6228 * so most of the standard driver entry points are in zvol.c.
6231 static struct cb_ops zfs_cb_ops = {
6232 zfsdev_open, /* open */
6233 zfsdev_close, /* close */
6234 zvol_strategy, /* strategy */
6236 zvol_dump, /* dump */
6237 zvol_read, /* read */
6238 zvol_write, /* write */
6239 zfsdev_ioctl, /* ioctl */
6243 nochpoll, /* poll */
6244 ddi_prop_op, /* prop_op */
6245 NULL, /* streamtab */
6246 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */
6247 CB_REV, /* version */
6248 nodev, /* async read */
6249 nodev, /* async write */
6252 static struct dev_ops zfs_dev_ops = {
6253 DEVO_REV, /* version */
6255 zfs_info, /* info */
6256 nulldev, /* identify */
6257 nulldev, /* probe */
6258 zfs_attach, /* attach */
6259 zfs_detach, /* detach */
6261 &zfs_cb_ops, /* driver operations */
6262 NULL, /* no bus operations */
6264 ddi_quiesce_not_needed, /* quiesce */
6267 static struct modldrv zfs_modldrv = {
6273 static struct modlinkage modlinkage = {
6275 (void *)&zfs_modlfs,
6276 (void *)&zfs_modldrv,
6281 static struct cdevsw zfs_cdevsw = {
6282 .d_version = D_VERSION,
6283 .d_open = zfsdev_open,
6284 .d_ioctl = zfsdev_ioctl,
6285 .d_name = ZFS_DEV_NAME
6289 zfs_allow_log_destroy(void *arg)
6291 char *poolname = arg;
6298 zfsdev = make_dev(&zfs_cdevsw, 0x0, UID_ROOT, GID_OPERATOR, 0666,
6306 destroy_dev(zfsdev);
6309 static struct root_hold_token *zfs_root_token;
6310 struct proc *zfsproc;
6318 spa_init(FREAD | FWRITE);
6323 if ((error = mod_install(&modlinkage)) != 0) {
6330 tsd_create(&zfs_fsyncer_key, NULL);
6331 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
6332 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
6334 error = ldi_ident_from_mod(&modlinkage, &zfs_li);
6336 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
6346 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
6347 return (SET_ERROR(EBUSY));
6349 if ((error = mod_remove(&modlinkage)) != 0)
6355 if (zfs_nfsshare_inited)
6356 (void) ddi_modclose(nfs_mod);
6357 if (zfs_smbshare_inited)
6358 (void) ddi_modclose(smbsrv_mod);
6359 if (zfs_nfsshare_inited || zfs_smbshare_inited)
6360 (void) ddi_modclose(sharefs_mod);
6362 tsd_destroy(&zfs_fsyncer_key);
6363 ldi_ident_release(zfs_li);
6365 mutex_destroy(&zfs_share_lock);
6371 _info(struct modinfo *modinfop)
6373 return (mod_info(&modlinkage, modinfop));
6377 static int zfs__init(void);
6378 static int zfs__fini(void);
6379 static void zfs_shutdown(void *, int);
6381 static eventhandler_tag zfs_shutdown_event_tag;
6387 zfs_root_token = root_mount_hold("ZFS");
6389 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
6391 spa_init(FREAD | FWRITE);
6396 tsd_create(&zfs_fsyncer_key, NULL);
6397 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
6398 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
6400 printf("ZFS storage pool version: features support (" SPA_VERSION_STRING ")\n");
6401 root_mount_rel(zfs_root_token);
6411 if (spa_busy() || zfs_busy() || zvol_busy() ||
6412 zio_injection_enabled) {
6421 tsd_destroy(&zfs_fsyncer_key);
6422 tsd_destroy(&rrw_tsd_key);
6423 tsd_destroy(&zfs_allow_log_key);
6425 mutex_destroy(&zfs_share_lock);
6431 zfs_shutdown(void *arg __unused, int howto __unused)
6435 * ZFS fini routines can not properly work in a panic-ed system.
6437 if (panicstr == NULL)
6443 zfs_modevent(module_t mod, int type, void *unused __unused)
6451 zfs_shutdown_event_tag = EVENTHANDLER_REGISTER(
6452 shutdown_post_sync, zfs_shutdown, NULL,
6453 SHUTDOWN_PRI_FIRST);
6457 if (err == 0 && zfs_shutdown_event_tag != NULL)
6458 EVENTHANDLER_DEREGISTER(shutdown_post_sync,
6459 zfs_shutdown_event_tag);
6466 return (EOPNOTSUPP);
6469 static moduledata_t zfs_mod = {
6474 DECLARE_MODULE(zfsctrl, zfs_mod, SI_SUB_VFS, SI_ORDER_ANY);
6475 MODULE_VERSION(zfsctrl, 1);
6476 MODULE_DEPEND(zfsctrl, opensolaris, 1, 1, 1);
6477 MODULE_DEPEND(zfsctrl, krpc, 1, 1, 1);