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 2014 Xin Li <delphij@FreeBSD.org>. All rights reserved.
28 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
29 * Copyright (c) 2014, Joyent, Inc. All rights reserved.
30 * Copyright (c) 2011, 2014 by Delphix. All rights reserved.
31 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
32 * Copyright (c) 2013 Steven Hartland. All rights reserved.
33 * Copyright (c) 2014, Nexenta Systems, Inc. All rights reserved.
39 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
40 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
42 * There are two ways that we handle ioctls: the legacy way where almost
43 * all of the logic is in the ioctl callback, and the new way where most
44 * of the marshalling is handled in the common entry point, zfsdev_ioctl().
46 * Non-legacy ioctls should be registered by calling
47 * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked
48 * from userland by lzc_ioctl().
50 * The registration arguments are as follows:
53 * The name of the ioctl. This is used for history logging. If the
54 * ioctl returns successfully (the callback returns 0), and allow_log
55 * is true, then a history log entry will be recorded with the input &
56 * output nvlists. The log entry can be printed with "zpool history -i".
59 * The ioctl request number, which userland will pass to ioctl(2).
60 * The ioctl numbers can change from release to release, because
61 * the caller (libzfs) must be matched to the kernel.
63 * zfs_secpolicy_func_t *secpolicy
64 * This function will be called before the zfs_ioc_func_t, to
65 * determine if this operation is permitted. It should return EPERM
66 * on failure, and 0 on success. Checks include determining if the
67 * dataset is visible in this zone, and if the user has either all
68 * zfs privileges in the zone (SYS_MOUNT), or has been granted permission
69 * to do this operation on this dataset with "zfs allow".
71 * zfs_ioc_namecheck_t namecheck
72 * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
73 * name, a dataset name, or nothing. If the name is not well-formed,
74 * the ioctl will fail and the callback will not be called.
75 * Therefore, the callback can assume that the name is well-formed
76 * (e.g. is null-terminated, doesn't have more than one '@' character,
77 * doesn't have invalid characters).
79 * zfs_ioc_poolcheck_t pool_check
80 * This specifies requirements on the pool state. If the pool does
81 * not meet them (is suspended or is readonly), the ioctl will fail
82 * and the callback will not be called. If any checks are specified
83 * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
84 * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED |
85 * POOL_CHECK_READONLY).
87 * boolean_t smush_outnvlist
88 * If smush_outnvlist is true, then the output is presumed to be a
89 * list of errors, and it will be "smushed" down to fit into the
90 * caller's buffer, by removing some entries and replacing them with a
91 * single "N_MORE_ERRORS" entry indicating how many were removed. See
92 * nvlist_smush() for details. If smush_outnvlist is false, and the
93 * outnvlist does not fit into the userland-provided buffer, then the
94 * ioctl will fail with ENOMEM.
96 * zfs_ioc_func_t *func
97 * The callback function that will perform the operation.
99 * The callback should return 0 on success, or an error number on
100 * failure. If the function fails, the userland ioctl will return -1,
101 * and errno will be set to the callback's return value. The callback
102 * will be called with the following arguments:
105 * The name of the pool or dataset to operate on, from
106 * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the
107 * expected type (pool, dataset, or none).
110 * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or
111 * NULL if no input nvlist was provided. Changes to this nvlist are
112 * ignored. If the input nvlist could not be deserialized, the
113 * ioctl will fail and the callback will not be called.
116 * The output nvlist, initially empty. The callback can fill it in,
117 * and it will be returned to userland by serializing it into
118 * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization
119 * fails (e.g. because the caller didn't supply a large enough
120 * buffer), then the overall ioctl will fail. See the
121 * 'smush_nvlist' argument above for additional behaviors.
123 * There are two typical uses of the output nvlist:
124 * - To return state, e.g. property values. In this case,
125 * smush_outnvlist should be false. If the buffer was not large
126 * enough, the caller will reallocate a larger buffer and try
129 * - To return multiple errors from an ioctl which makes on-disk
130 * changes. In this case, smush_outnvlist should be true.
131 * Ioctls which make on-disk modifications should generally not
132 * use the outnvl if they succeed, because the caller can not
133 * distinguish between the operation failing, and
134 * deserialization failing.
137 #include <sys/types.h>
138 #include <sys/param.h>
139 #include <sys/systm.h>
140 #include <sys/conf.h>
141 #include <sys/kernel.h>
142 #include <sys/lock.h>
143 #include <sys/malloc.h>
144 #include <sys/mutex.h>
145 #include <sys/proc.h>
146 #include <sys/errno.h>
149 #include <sys/file.h>
150 #include <sys/kmem.h>
151 #include <sys/conf.h>
152 #include <sys/cmn_err.h>
153 #include <sys/stat.h>
154 #include <sys/zfs_ioctl.h>
155 #include <sys/zfs_vfsops.h>
156 #include <sys/zfs_znode.h>
159 #include <sys/spa_impl.h>
160 #include <sys/vdev.h>
162 #include <sys/dsl_dir.h>
163 #include <sys/dsl_dataset.h>
164 #include <sys/dsl_prop.h>
165 #include <sys/dsl_deleg.h>
166 #include <sys/dmu_objset.h>
167 #include <sys/dmu_impl.h>
168 #include <sys/dmu_tx.h>
169 #include <sys/sunddi.h>
170 #include <sys/policy.h>
171 #include <sys/zone.h>
172 #include <sys/nvpair.h>
173 #include <sys/mount.h>
174 #include <sys/taskqueue.h>
176 #include <sys/varargs.h>
177 #include <sys/fs/zfs.h>
178 #include <sys/zfs_ctldir.h>
179 #include <sys/zfs_dir.h>
180 #include <sys/zfs_onexit.h>
181 #include <sys/zvol.h>
182 #include <sys/dsl_scan.h>
183 #include <sys/dmu_objset.h>
184 #include <sys/dmu_send.h>
185 #include <sys/dsl_destroy.h>
186 #include <sys/dsl_bookmark.h>
187 #include <sys/dsl_userhold.h>
188 #include <sys/zfeature.h>
190 #include "zfs_namecheck.h"
191 #include "zfs_prop.h"
192 #include "zfs_deleg.h"
193 #include "zfs_comutil.h"
194 #include "zfs_ioctl_compat.h"
196 CTASSERT(sizeof(zfs_cmd_t) < IOCPARM_MAX);
198 static int snapshot_list_prefetch;
199 SYSCTL_DECL(_vfs_zfs);
200 SYSCTL_INT(_vfs_zfs, OID_AUTO, snapshot_list_prefetch, CTLFLAG_RWTUN,
201 &snapshot_list_prefetch, 0, "Prefetch data when listing snapshots");
203 static struct cdev *zfsdev;
205 extern void zfs_init(void);
206 extern void zfs_fini(void);
208 uint_t zfs_fsyncer_key;
209 extern uint_t rrw_tsd_key;
210 static uint_t zfs_allow_log_key;
212 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *);
213 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *);
214 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *);
220 } zfs_ioc_namecheck_t;
223 POOL_CHECK_NONE = 1 << 0,
224 POOL_CHECK_SUSPENDED = 1 << 1,
225 POOL_CHECK_READONLY = 1 << 2,
226 } zfs_ioc_poolcheck_t;
228 typedef struct zfs_ioc_vec {
229 zfs_ioc_legacy_func_t *zvec_legacy_func;
230 zfs_ioc_func_t *zvec_func;
231 zfs_secpolicy_func_t *zvec_secpolicy;
232 zfs_ioc_namecheck_t zvec_namecheck;
233 boolean_t zvec_allow_log;
234 zfs_ioc_poolcheck_t zvec_pool_check;
235 boolean_t zvec_smush_outnvlist;
236 const char *zvec_name;
239 /* This array is indexed by zfs_userquota_prop_t */
240 static const char *userquota_perms[] = {
241 ZFS_DELEG_PERM_USERUSED,
242 ZFS_DELEG_PERM_USERQUOTA,
243 ZFS_DELEG_PERM_GROUPUSED,
244 ZFS_DELEG_PERM_GROUPQUOTA,
247 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
248 static int zfs_check_settable(const char *name, nvpair_t *property,
250 static int zfs_check_clearable(char *dataset, nvlist_t *props,
252 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
254 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
255 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
257 static void zfsdev_close(void *data);
259 static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature);
261 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
263 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
270 * Get rid of annoying "../common/" prefix to filename.
272 newfile = strrchr(file, '/');
273 if (newfile != NULL) {
274 newfile = newfile + 1; /* Get rid of leading / */
280 (void) vsnprintf(buf, sizeof (buf), fmt, adx);
284 * To get this data, use the zfs-dprintf probe as so:
285 * dtrace -q -n 'zfs-dprintf \
286 * /stringof(arg0) == "dbuf.c"/ \
287 * {printf("%s: %s", stringof(arg1), stringof(arg3))}'
289 * arg1 = function name
293 DTRACE_PROBE4(zfs__dprintf,
294 char *, newfile, char *, func, int, line, char *, buf);
298 history_str_free(char *buf)
300 kmem_free(buf, HIS_MAX_RECORD_LEN);
304 history_str_get(zfs_cmd_t *zc)
308 if (zc->zc_history == 0)
311 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
312 if (copyinstr((void *)(uintptr_t)zc->zc_history,
313 buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
314 history_str_free(buf);
318 buf[HIS_MAX_RECORD_LEN -1] = '\0';
324 * Check to see if the named dataset is currently defined as bootable
327 zfs_is_bootfs(const char *name)
331 if (dmu_objset_hold(name, FTAG, &os) == 0) {
333 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
334 dmu_objset_rele(os, FTAG);
341 * Return non-zero if the spa version is less than requested version.
344 zfs_earlier_version(const char *name, int version)
348 if (spa_open(name, &spa, FTAG) == 0) {
349 if (spa_version(spa) < version) {
350 spa_close(spa, FTAG);
353 spa_close(spa, FTAG);
359 * Return TRUE if the ZPL version is less than requested version.
362 zpl_earlier_version(const char *name, int version)
365 boolean_t rc = B_TRUE;
367 if (dmu_objset_hold(name, FTAG, &os) == 0) {
370 if (dmu_objset_type(os) != DMU_OST_ZFS) {
371 dmu_objset_rele(os, FTAG);
374 /* XXX reading from non-owned objset */
375 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
376 rc = zplversion < version;
377 dmu_objset_rele(os, FTAG);
383 zfs_log_history(zfs_cmd_t *zc)
388 if ((buf = history_str_get(zc)) == NULL)
391 if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
392 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
393 (void) spa_history_log(spa, buf);
394 spa_close(spa, FTAG);
396 history_str_free(buf);
400 * Policy for top-level read operations (list pools). Requires no privileges,
401 * and can be used in the local zone, as there is no associated dataset.
405 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
411 * Policy for dataset read operations (list children, get statistics). Requires
412 * no privileges, but must be visible in the local zone.
416 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
418 if (INGLOBALZONE(curthread) ||
419 zone_dataset_visible(zc->zc_name, NULL))
422 return (SET_ERROR(ENOENT));
426 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
431 * The dataset must be visible by this zone -- check this first
432 * so they don't see EPERM on something they shouldn't know about.
434 if (!INGLOBALZONE(curthread) &&
435 !zone_dataset_visible(dataset, &writable))
436 return (SET_ERROR(ENOENT));
438 if (INGLOBALZONE(curthread)) {
440 * If the fs is zoned, only root can access it from the
443 if (secpolicy_zfs(cr) && zoned)
444 return (SET_ERROR(EPERM));
447 * If we are in a local zone, the 'zoned' property must be set.
450 return (SET_ERROR(EPERM));
452 /* must be writable by this zone */
454 return (SET_ERROR(EPERM));
460 zfs_dozonecheck(const char *dataset, cred_t *cr)
464 if (dsl_prop_get_integer(dataset, "jailed", &zoned, NULL))
465 return (SET_ERROR(ENOENT));
467 return (zfs_dozonecheck_impl(dataset, zoned, cr));
471 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
475 if (dsl_prop_get_int_ds(ds, "jailed", &zoned))
476 return (SET_ERROR(ENOENT));
478 return (zfs_dozonecheck_impl(dataset, zoned, cr));
482 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
483 const char *perm, cred_t *cr)
487 error = zfs_dozonecheck_ds(name, ds, cr);
489 error = secpolicy_zfs(cr);
491 error = dsl_deleg_access_impl(ds, perm, cr);
497 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
503 error = dsl_pool_hold(name, FTAG, &dp);
507 error = dsl_dataset_hold(dp, name, FTAG, &ds);
509 dsl_pool_rele(dp, FTAG);
513 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
515 dsl_dataset_rele(ds, FTAG);
516 dsl_pool_rele(dp, FTAG);
522 * Policy for setting the security label property.
524 * Returns 0 for success, non-zero for access and other errors.
527 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
529 char ds_hexsl[MAXNAMELEN];
530 bslabel_t ds_sl, new_sl;
531 boolean_t new_default = FALSE;
533 int needed_priv = -1;
536 /* First get the existing dataset label. */
537 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
538 1, sizeof (ds_hexsl), &ds_hexsl, NULL);
540 return (SET_ERROR(EPERM));
542 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
545 /* The label must be translatable */
546 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
547 return (SET_ERROR(EINVAL));
550 * In a non-global zone, disallow attempts to set a label that
551 * doesn't match that of the zone; otherwise no other checks
554 if (!INGLOBALZONE(curproc)) {
555 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
556 return (SET_ERROR(EPERM));
561 * For global-zone datasets (i.e., those whose zoned property is
562 * "off", verify that the specified new label is valid for the
565 if (dsl_prop_get_integer(name,
566 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
567 return (SET_ERROR(EPERM));
569 if (zfs_check_global_label(name, strval) != 0)
570 return (SET_ERROR(EPERM));
574 * If the existing dataset label is nondefault, check if the
575 * dataset is mounted (label cannot be changed while mounted).
576 * Get the zfsvfs; if there isn't one, then the dataset isn't
577 * mounted (or isn't a dataset, doesn't exist, ...).
579 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
581 static char *setsl_tag = "setsl_tag";
584 * Try to own the dataset; abort if there is any error,
585 * (e.g., already mounted, in use, or other error).
587 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
590 return (SET_ERROR(EPERM));
592 dmu_objset_disown(os, setsl_tag);
595 needed_priv = PRIV_FILE_DOWNGRADE_SL;
599 if (hexstr_to_label(strval, &new_sl) != 0)
600 return (SET_ERROR(EPERM));
602 if (blstrictdom(&ds_sl, &new_sl))
603 needed_priv = PRIV_FILE_DOWNGRADE_SL;
604 else if (blstrictdom(&new_sl, &ds_sl))
605 needed_priv = PRIV_FILE_UPGRADE_SL;
607 /* dataset currently has a default label */
609 needed_priv = PRIV_FILE_UPGRADE_SL;
613 if (needed_priv != -1)
614 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
617 #endif /* SECLABEL */
620 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
626 * Check permissions for special properties.
631 * Disallow setting of 'zoned' from within a local zone.
633 if (!INGLOBALZONE(curthread))
634 return (SET_ERROR(EPERM));
638 case ZFS_PROP_FILESYSTEM_LIMIT:
639 case ZFS_PROP_SNAPSHOT_LIMIT:
640 if (!INGLOBALZONE(curthread)) {
642 char setpoint[MAXNAMELEN];
644 * Unprivileged users are allowed to modify the
645 * limit on things *under* (ie. contained by)
646 * the thing they own.
648 if (dsl_prop_get_integer(dsname, "jailed", &zoned,
650 return (SET_ERROR(EPERM));
651 if (!zoned || strlen(dsname) <= strlen(setpoint))
652 return (SET_ERROR(EPERM));
656 case ZFS_PROP_MLSLABEL:
658 if (!is_system_labeled())
659 return (SET_ERROR(EPERM));
661 if (nvpair_value_string(propval, &strval) == 0) {
664 err = zfs_set_slabel_policy(dsname, strval, CRED());
674 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
679 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
683 error = zfs_dozonecheck(zc->zc_name, cr);
688 * permission to set permissions will be evaluated later in
689 * dsl_deleg_can_allow()
696 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
698 return (zfs_secpolicy_write_perms(zc->zc_name,
699 ZFS_DELEG_PERM_ROLLBACK, cr));
704 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
712 * Generate the current snapshot name from the given objsetid, then
713 * use that name for the secpolicy/zone checks.
715 cp = strchr(zc->zc_name, '@');
717 return (SET_ERROR(EINVAL));
718 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
722 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
724 dsl_pool_rele(dp, FTAG);
728 dsl_dataset_name(ds, zc->zc_name);
730 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
731 ZFS_DELEG_PERM_SEND, cr);
732 dsl_dataset_rele(ds, FTAG);
733 dsl_pool_rele(dp, FTAG);
740 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
742 return (zfs_secpolicy_write_perms(zc->zc_name,
743 ZFS_DELEG_PERM_SEND, cr));
748 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
753 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
754 NO_FOLLOW, NULL, &vp)) != 0)
757 /* Now make sure mntpnt and dataset are ZFS */
759 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 ||
760 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
761 zc->zc_name) != 0)) {
763 return (SET_ERROR(EPERM));
767 return (dsl_deleg_access(zc->zc_name,
768 ZFS_DELEG_PERM_SHARE, cr));
772 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
774 if (!INGLOBALZONE(curthread))
775 return (SET_ERROR(EPERM));
777 if (secpolicy_nfs(cr) == 0) {
780 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
785 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
787 if (!INGLOBALZONE(curthread))
788 return (SET_ERROR(EPERM));
790 if (secpolicy_smb(cr) == 0) {
793 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
798 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
803 * Remove the @bla or /bla from the end of the name to get the parent.
805 (void) strncpy(parent, datasetname, parentsize);
806 cp = strrchr(parent, '@');
810 cp = strrchr(parent, '/');
812 return (SET_ERROR(ENOENT));
820 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
824 if ((error = zfs_secpolicy_write_perms(name,
825 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
828 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
833 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
835 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
839 * Destroying snapshots with delegated permissions requires
840 * descendant mount and destroy permissions.
844 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
847 nvpair_t *pair, *nextpair;
850 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
851 return (SET_ERROR(EINVAL));
852 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
854 nextpair = nvlist_next_nvpair(snaps, pair);
855 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr);
856 if (error == ENOENT) {
858 * Ignore any snapshots that don't exist (we consider
859 * them "already destroyed"). Remove the name from the
860 * nvl here in case the snapshot is created between
861 * now and when we try to destroy it (in which case
862 * we don't want to destroy it since we haven't
863 * checked for permission).
865 fnvlist_remove_nvpair(snaps, pair);
876 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
878 char parentname[MAXNAMELEN];
881 if ((error = zfs_secpolicy_write_perms(from,
882 ZFS_DELEG_PERM_RENAME, cr)) != 0)
885 if ((error = zfs_secpolicy_write_perms(from,
886 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
889 if ((error = zfs_get_parent(to, parentname,
890 sizeof (parentname))) != 0)
893 if ((error = zfs_secpolicy_write_perms(parentname,
894 ZFS_DELEG_PERM_CREATE, cr)) != 0)
897 if ((error = zfs_secpolicy_write_perms(parentname,
898 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
906 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
911 if ((zc->zc_cookie & 1) != 0) {
913 * This is recursive rename, so the starting snapshot might
914 * not exist. Check file system or volume permission instead.
916 at = strchr(zc->zc_name, '@');
922 error = zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr);
932 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
935 dsl_dataset_t *clone;
938 error = zfs_secpolicy_write_perms(zc->zc_name,
939 ZFS_DELEG_PERM_PROMOTE, cr);
943 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
947 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
950 char parentname[MAXNAMELEN];
951 dsl_dataset_t *origin = NULL;
955 error = dsl_dataset_hold_obj(dd->dd_pool,
956 dd->dd_phys->dd_origin_obj, FTAG, &origin);
958 dsl_dataset_rele(clone, FTAG);
959 dsl_pool_rele(dp, FTAG);
963 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
964 ZFS_DELEG_PERM_MOUNT, cr);
966 dsl_dataset_name(origin, parentname);
968 error = zfs_secpolicy_write_perms_ds(parentname, origin,
969 ZFS_DELEG_PERM_PROMOTE, cr);
971 dsl_dataset_rele(clone, FTAG);
972 dsl_dataset_rele(origin, FTAG);
974 dsl_pool_rele(dp, FTAG);
980 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
984 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
985 ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
988 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
989 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
992 return (zfs_secpolicy_write_perms(zc->zc_name,
993 ZFS_DELEG_PERM_CREATE, cr));
997 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
999 return (zfs_secpolicy_write_perms(name,
1000 ZFS_DELEG_PERM_SNAPSHOT, cr));
1004 * Check for permission to create each snapshot in the nvlist.
1008 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1014 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
1015 return (SET_ERROR(EINVAL));
1016 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
1017 pair = nvlist_next_nvpair(snaps, pair)) {
1018 char *name = nvpair_name(pair);
1019 char *atp = strchr(name, '@');
1022 error = SET_ERROR(EINVAL);
1026 error = zfs_secpolicy_snapshot_perms(name, cr);
1035 * Check for permission to create each snapshot in the nvlist.
1039 zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1043 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
1044 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
1045 char *name = nvpair_name(pair);
1046 char *hashp = strchr(name, '#');
1048 if (hashp == NULL) {
1049 error = SET_ERROR(EINVAL);
1053 error = zfs_secpolicy_write_perms(name,
1054 ZFS_DELEG_PERM_BOOKMARK, cr);
1064 zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1066 nvpair_t *pair, *nextpair;
1069 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1071 char *name = nvpair_name(pair);
1072 char *hashp = strchr(name, '#');
1073 nextpair = nvlist_next_nvpair(innvl, pair);
1075 if (hashp == NULL) {
1076 error = SET_ERROR(EINVAL);
1081 error = zfs_secpolicy_write_perms(name,
1082 ZFS_DELEG_PERM_DESTROY, cr);
1084 if (error == ENOENT) {
1086 * Ignore any filesystems that don't exist (we consider
1087 * their bookmarks "already destroyed"). Remove
1088 * the name from the nvl here in case the filesystem
1089 * is created between now and when we try to destroy
1090 * the bookmark (in which case we don't want to
1091 * destroy it since we haven't checked for permission).
1093 fnvlist_remove_nvpair(innvl, pair);
1105 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1108 * Even root must have a proper TSD so that we know what pool
1111 if (tsd_get(zfs_allow_log_key) == NULL)
1112 return (SET_ERROR(EPERM));
1117 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1119 char parentname[MAXNAMELEN];
1123 if ((error = zfs_get_parent(zc->zc_name, parentname,
1124 sizeof (parentname))) != 0)
1127 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1128 (error = zfs_secpolicy_write_perms(origin,
1129 ZFS_DELEG_PERM_CLONE, cr)) != 0)
1132 if ((error = zfs_secpolicy_write_perms(parentname,
1133 ZFS_DELEG_PERM_CREATE, cr)) != 0)
1136 return (zfs_secpolicy_write_perms(parentname,
1137 ZFS_DELEG_PERM_MOUNT, cr));
1141 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires
1142 * SYS_CONFIG privilege, which is not available in a local zone.
1146 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1148 if (secpolicy_sys_config(cr, B_FALSE) != 0)
1149 return (SET_ERROR(EPERM));
1155 * Policy for object to name lookups.
1159 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1163 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1166 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1171 * Policy for fault injection. Requires all privileges.
1175 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1177 return (secpolicy_zinject(cr));
1182 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1184 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1186 if (prop == ZPROP_INVAL) {
1187 if (!zfs_prop_user(zc->zc_value))
1188 return (SET_ERROR(EINVAL));
1189 return (zfs_secpolicy_write_perms(zc->zc_name,
1190 ZFS_DELEG_PERM_USERPROP, cr));
1192 return (zfs_secpolicy_setprop(zc->zc_name, prop,
1198 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1200 int err = zfs_secpolicy_read(zc, innvl, cr);
1204 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1205 return (SET_ERROR(EINVAL));
1207 if (zc->zc_value[0] == 0) {
1209 * They are asking about a posix uid/gid. If it's
1210 * themself, allow it.
1212 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1213 zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
1214 if (zc->zc_guid == crgetuid(cr))
1217 if (groupmember(zc->zc_guid, cr))
1222 return (zfs_secpolicy_write_perms(zc->zc_name,
1223 userquota_perms[zc->zc_objset_type], cr));
1227 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1229 int err = zfs_secpolicy_read(zc, innvl, cr);
1233 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1234 return (SET_ERROR(EINVAL));
1236 return (zfs_secpolicy_write_perms(zc->zc_name,
1237 userquota_perms[zc->zc_objset_type], cr));
1242 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1244 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1250 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1256 error = nvlist_lookup_nvlist(innvl, "holds", &holds);
1258 return (SET_ERROR(EINVAL));
1260 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
1261 pair = nvlist_next_nvpair(holds, pair)) {
1262 char fsname[MAXNAMELEN];
1263 error = dmu_fsname(nvpair_name(pair), fsname);
1266 error = zfs_secpolicy_write_perms(fsname,
1267 ZFS_DELEG_PERM_HOLD, cr);
1276 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1281 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1282 pair = nvlist_next_nvpair(innvl, pair)) {
1283 char fsname[MAXNAMELEN];
1284 error = dmu_fsname(nvpair_name(pair), fsname);
1287 error = zfs_secpolicy_write_perms(fsname,
1288 ZFS_DELEG_PERM_RELEASE, cr);
1296 * Policy for allowing temporary snapshots to be taken or released
1299 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1302 * A temporary snapshot is the same as a snapshot,
1303 * hold, destroy and release all rolled into one.
1304 * Delegated diff alone is sufficient that we allow this.
1308 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1309 ZFS_DELEG_PERM_DIFF, cr)) == 0)
1312 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1314 error = zfs_secpolicy_hold(zc, innvl, cr);
1316 error = zfs_secpolicy_release(zc, innvl, cr);
1318 error = zfs_secpolicy_destroy(zc, innvl, cr);
1323 * Returns the nvlist as specified by the user in the zfs_cmd_t.
1326 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1330 nvlist_t *list = NULL;
1333 * Read in and unpack the user-supplied nvlist.
1336 return (SET_ERROR(EINVAL));
1338 packed = kmem_alloc(size, KM_SLEEP);
1340 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1342 kmem_free(packed, size);
1346 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1347 kmem_free(packed, size);
1351 kmem_free(packed, size);
1358 * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1359 * Entries will be removed from the end of the nvlist, and one int32 entry
1360 * named "N_MORE_ERRORS" will be added indicating how many entries were
1364 nvlist_smush(nvlist_t *errors, size_t max)
1368 size = fnvlist_size(errors);
1371 nvpair_t *more_errors;
1375 return (SET_ERROR(ENOMEM));
1377 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1378 more_errors = nvlist_prev_nvpair(errors, NULL);
1381 nvpair_t *pair = nvlist_prev_nvpair(errors,
1383 fnvlist_remove_nvpair(errors, pair);
1385 size = fnvlist_size(errors);
1386 } while (size > max);
1388 fnvlist_remove_nvpair(errors, more_errors);
1389 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1390 ASSERT3U(fnvlist_size(errors), <=, max);
1397 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1399 char *packed = NULL;
1403 size = fnvlist_size(nvl);
1405 if (size > zc->zc_nvlist_dst_size) {
1407 * Solaris returns ENOMEM here, because even if an error is
1408 * returned from an ioctl(2), new zc_nvlist_dst_size will be
1409 * passed to the userland. This is not the case for FreeBSD.
1410 * We need to return 0, so the kernel will copy the
1411 * zc_nvlist_dst_size back and the userland can discover that a
1412 * bigger buffer is needed.
1416 packed = fnvlist_pack(nvl, &size);
1417 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1418 size, zc->zc_iflags) != 0)
1419 error = SET_ERROR(EFAULT);
1420 fnvlist_pack_free(packed, size);
1423 zc->zc_nvlist_dst_size = size;
1424 zc->zc_nvlist_dst_filled = B_TRUE;
1429 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1434 error = dmu_objset_hold(dsname, FTAG, &os);
1437 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1438 dmu_objset_rele(os, FTAG);
1439 return (SET_ERROR(EINVAL));
1442 mutex_enter(&os->os_user_ptr_lock);
1443 *zfvp = dmu_objset_get_user(os);
1445 VFS_HOLD((*zfvp)->z_vfs);
1447 error = SET_ERROR(ESRCH);
1449 mutex_exit(&os->os_user_ptr_lock);
1450 dmu_objset_rele(os, FTAG);
1455 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1456 * case its z_vfs will be NULL, and it will be opened as the owner.
1457 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1458 * which prevents all vnode ops from running.
1461 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1465 if (getzfsvfs(name, zfvp) != 0)
1466 error = zfsvfs_create(name, zfvp);
1468 rrm_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER :
1470 if ((*zfvp)->z_unmounted) {
1472 * XXX we could probably try again, since the unmounting
1473 * thread should be just about to disassociate the
1474 * objset from the zfsvfs.
1476 rrm_exit(&(*zfvp)->z_teardown_lock, tag);
1477 return (SET_ERROR(EBUSY));
1484 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1486 rrm_exit(&zfsvfs->z_teardown_lock, tag);
1488 if (zfsvfs->z_vfs) {
1489 VFS_RELE(zfsvfs->z_vfs);
1491 dmu_objset_disown(zfsvfs->z_os, zfsvfs);
1492 zfsvfs_free(zfsvfs);
1497 zfs_ioc_pool_create(zfs_cmd_t *zc)
1500 nvlist_t *config, *props = NULL;
1501 nvlist_t *rootprops = NULL;
1502 nvlist_t *zplprops = NULL;
1504 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1505 zc->zc_iflags, &config))
1508 if (zc->zc_nvlist_src_size != 0 && (error =
1509 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1510 zc->zc_iflags, &props))) {
1511 nvlist_free(config);
1516 nvlist_t *nvl = NULL;
1517 uint64_t version = SPA_VERSION;
1519 (void) nvlist_lookup_uint64(props,
1520 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1521 if (!SPA_VERSION_IS_SUPPORTED(version)) {
1522 error = SET_ERROR(EINVAL);
1523 goto pool_props_bad;
1525 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1527 error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1529 nvlist_free(config);
1533 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1535 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1536 error = zfs_fill_zplprops_root(version, rootprops,
1539 goto pool_props_bad;
1542 error = spa_create(zc->zc_name, config, props, zplprops);
1545 * Set the remaining root properties
1547 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1548 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1549 (void) spa_destroy(zc->zc_name);
1552 nvlist_free(rootprops);
1553 nvlist_free(zplprops);
1554 nvlist_free(config);
1561 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1564 zfs_log_history(zc);
1565 error = spa_destroy(zc->zc_name);
1567 zvol_remove_minors(zc->zc_name);
1572 zfs_ioc_pool_import(zfs_cmd_t *zc)
1574 nvlist_t *config, *props = NULL;
1578 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1579 zc->zc_iflags, &config)) != 0)
1582 if (zc->zc_nvlist_src_size != 0 && (error =
1583 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1584 zc->zc_iflags, &props))) {
1585 nvlist_free(config);
1589 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1590 guid != zc->zc_guid)
1591 error = SET_ERROR(EINVAL);
1593 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1595 if (zc->zc_nvlist_dst != 0) {
1598 if ((err = put_nvlist(zc, config)) != 0)
1602 nvlist_free(config);
1611 zfs_ioc_pool_export(zfs_cmd_t *zc)
1614 boolean_t force = (boolean_t)zc->zc_cookie;
1615 boolean_t hardforce = (boolean_t)zc->zc_guid;
1617 zfs_log_history(zc);
1618 error = spa_export(zc->zc_name, NULL, force, hardforce);
1620 zvol_remove_minors(zc->zc_name);
1625 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1630 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1631 return (SET_ERROR(EEXIST));
1633 error = put_nvlist(zc, configs);
1635 nvlist_free(configs);
1642 * zc_name name of the pool
1645 * zc_cookie real errno
1646 * zc_nvlist_dst config nvlist
1647 * zc_nvlist_dst_size size of config nvlist
1650 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1656 error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1657 sizeof (zc->zc_value));
1659 if (config != NULL) {
1660 ret = put_nvlist(zc, config);
1661 nvlist_free(config);
1664 * The config may be present even if 'error' is non-zero.
1665 * In this case we return success, and preserve the real errno
1668 zc->zc_cookie = error;
1677 * Try to import the given pool, returning pool stats as appropriate so that
1678 * user land knows which devices are available and overall pool health.
1681 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1683 nvlist_t *tryconfig, *config;
1686 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1687 zc->zc_iflags, &tryconfig)) != 0)
1690 config = spa_tryimport(tryconfig);
1692 nvlist_free(tryconfig);
1695 return (SET_ERROR(EINVAL));
1697 error = put_nvlist(zc, config);
1698 nvlist_free(config);
1705 * zc_name name of the pool
1706 * zc_cookie scan func (pool_scan_func_t)
1709 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1714 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1717 if (zc->zc_cookie == POOL_SCAN_NONE)
1718 error = spa_scan_stop(spa);
1720 error = spa_scan(spa, zc->zc_cookie);
1722 spa_close(spa, FTAG);
1728 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1733 error = spa_open(zc->zc_name, &spa, FTAG);
1736 spa_close(spa, FTAG);
1742 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1747 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1750 if (zc->zc_cookie < spa_version(spa) ||
1751 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1752 spa_close(spa, FTAG);
1753 return (SET_ERROR(EINVAL));
1756 spa_upgrade(spa, zc->zc_cookie);
1757 spa_close(spa, FTAG);
1763 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1770 if ((size = zc->zc_history_len) == 0)
1771 return (SET_ERROR(EINVAL));
1773 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1776 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1777 spa_close(spa, FTAG);
1778 return (SET_ERROR(ENOTSUP));
1781 hist_buf = kmem_alloc(size, KM_SLEEP);
1782 if ((error = spa_history_get(spa, &zc->zc_history_offset,
1783 &zc->zc_history_len, hist_buf)) == 0) {
1784 error = ddi_copyout(hist_buf,
1785 (void *)(uintptr_t)zc->zc_history,
1786 zc->zc_history_len, zc->zc_iflags);
1789 spa_close(spa, FTAG);
1790 kmem_free(hist_buf, size);
1795 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1800 error = spa_open(zc->zc_name, &spa, FTAG);
1802 error = spa_change_guid(spa);
1803 spa_close(spa, FTAG);
1809 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1811 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1816 * zc_name name of filesystem
1817 * zc_obj object to find
1820 * zc_value name of object
1823 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1828 /* XXX reading from objset not owned */
1829 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1831 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1832 dmu_objset_rele(os, FTAG);
1833 return (SET_ERROR(EINVAL));
1835 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1836 sizeof (zc->zc_value));
1837 dmu_objset_rele(os, FTAG);
1844 * zc_name name of filesystem
1845 * zc_obj object to find
1848 * zc_stat stats on object
1849 * zc_value path to object
1852 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1857 /* XXX reading from objset not owned */
1858 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1860 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1861 dmu_objset_rele(os, FTAG);
1862 return (SET_ERROR(EINVAL));
1864 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1865 sizeof (zc->zc_value));
1866 dmu_objset_rele(os, FTAG);
1872 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1876 nvlist_t *config, **l2cache, **spares;
1877 uint_t nl2cache = 0, nspares = 0;
1879 error = spa_open(zc->zc_name, &spa, FTAG);
1883 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1884 zc->zc_iflags, &config);
1885 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1886 &l2cache, &nl2cache);
1888 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1893 * A root pool with concatenated devices is not supported.
1894 * Thus, can not add a device to a root pool.
1896 * Intent log device can not be added to a rootpool because
1897 * during mountroot, zil is replayed, a seperated log device
1898 * can not be accessed during the mountroot time.
1900 * l2cache and spare devices are ok to be added to a rootpool.
1902 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1903 nvlist_free(config);
1904 spa_close(spa, FTAG);
1905 return (SET_ERROR(EDOM));
1907 #endif /* illumos */
1910 error = spa_vdev_add(spa, config);
1911 nvlist_free(config);
1913 spa_close(spa, FTAG);
1919 * zc_name name of the pool
1920 * zc_nvlist_conf nvlist of devices to remove
1921 * zc_cookie to stop the remove?
1924 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1929 error = spa_open(zc->zc_name, &spa, FTAG);
1932 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1933 spa_close(spa, FTAG);
1938 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1942 vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1944 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1946 switch (zc->zc_cookie) {
1947 case VDEV_STATE_ONLINE:
1948 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1951 case VDEV_STATE_OFFLINE:
1952 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1955 case VDEV_STATE_FAULTED:
1956 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1957 zc->zc_obj != VDEV_AUX_EXTERNAL)
1958 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1960 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1963 case VDEV_STATE_DEGRADED:
1964 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1965 zc->zc_obj != VDEV_AUX_EXTERNAL)
1966 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1968 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1972 error = SET_ERROR(EINVAL);
1974 zc->zc_cookie = newstate;
1975 spa_close(spa, FTAG);
1980 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1983 int replacing = zc->zc_cookie;
1987 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1990 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1991 zc->zc_iflags, &config)) == 0) {
1992 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1993 nvlist_free(config);
1996 spa_close(spa, FTAG);
2001 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
2006 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2009 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
2011 spa_close(spa, FTAG);
2016 zfs_ioc_vdev_split(zfs_cmd_t *zc)
2019 nvlist_t *config, *props = NULL;
2021 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
2023 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2026 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
2027 zc->zc_iflags, &config)) {
2028 spa_close(spa, FTAG);
2032 if (zc->zc_nvlist_src_size != 0 && (error =
2033 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2034 zc->zc_iflags, &props))) {
2035 spa_close(spa, FTAG);
2036 nvlist_free(config);
2040 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
2042 spa_close(spa, FTAG);
2044 nvlist_free(config);
2051 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
2054 char *path = zc->zc_value;
2055 uint64_t guid = zc->zc_guid;
2058 error = spa_open(zc->zc_name, &spa, FTAG);
2062 error = spa_vdev_setpath(spa, guid, path);
2063 spa_close(spa, FTAG);
2068 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
2071 char *fru = zc->zc_value;
2072 uint64_t guid = zc->zc_guid;
2075 error = spa_open(zc->zc_name, &spa, FTAG);
2079 error = spa_vdev_setfru(spa, guid, fru);
2080 spa_close(spa, FTAG);
2085 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
2090 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2092 if (zc->zc_nvlist_dst != 0 &&
2093 (error = dsl_prop_get_all(os, &nv)) == 0) {
2094 dmu_objset_stats(os, nv);
2096 * NB: zvol_get_stats() will read the objset contents,
2097 * which we aren't supposed to do with a
2098 * DS_MODE_USER hold, because it could be
2099 * inconsistent. So this is a bit of a workaround...
2100 * XXX reading with out owning
2102 if (!zc->zc_objset_stats.dds_inconsistent &&
2103 dmu_objset_type(os) == DMU_OST_ZVOL) {
2104 error = zvol_get_stats(os, nv);
2109 error = put_nvlist(zc, nv);
2118 * zc_name name of filesystem
2119 * zc_nvlist_dst_size size of buffer for property nvlist
2122 * zc_objset_stats stats
2123 * zc_nvlist_dst property nvlist
2124 * zc_nvlist_dst_size size of property nvlist
2127 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2132 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2134 error = zfs_ioc_objset_stats_impl(zc, os);
2135 dmu_objset_rele(os, FTAG);
2138 if (error == ENOMEM)
2145 * zc_name name of filesystem
2146 * zc_nvlist_dst_size size of buffer for property nvlist
2149 * zc_nvlist_dst received property nvlist
2150 * zc_nvlist_dst_size size of received property nvlist
2152 * Gets received properties (distinct from local properties on or after
2153 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2154 * local property values.
2157 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2163 * Without this check, we would return local property values if the
2164 * caller has not already received properties on or after
2165 * SPA_VERSION_RECVD_PROPS.
2167 if (!dsl_prop_get_hasrecvd(zc->zc_name))
2168 return (SET_ERROR(ENOTSUP));
2170 if (zc->zc_nvlist_dst != 0 &&
2171 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2172 error = put_nvlist(zc, nv);
2180 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2186 * zfs_get_zplprop() will either find a value or give us
2187 * the default value (if there is one).
2189 if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2191 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2197 * zc_name name of filesystem
2198 * zc_nvlist_dst_size size of buffer for zpl property nvlist
2201 * zc_nvlist_dst zpl property nvlist
2202 * zc_nvlist_dst_size size of zpl property nvlist
2205 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2210 /* XXX reading without owning */
2211 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os))
2214 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2217 * NB: nvl_add_zplprop() will read the objset contents,
2218 * which we aren't supposed to do with a DS_MODE_USER
2219 * hold, because it could be inconsistent.
2221 if (zc->zc_nvlist_dst != 0 &&
2222 !zc->zc_objset_stats.dds_inconsistent &&
2223 dmu_objset_type(os) == DMU_OST_ZFS) {
2226 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2227 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2228 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2229 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2230 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2231 err = put_nvlist(zc, nv);
2234 err = SET_ERROR(ENOENT);
2236 dmu_objset_rele(os, FTAG);
2241 dataset_name_hidden(const char *name)
2244 * Skip over datasets that are not visible in this zone,
2245 * internal datasets (which have a $ in their name), and
2246 * temporary datasets (which have a % in their name).
2248 if (strchr(name, '$') != NULL)
2250 if (strchr(name, '%') != NULL)
2252 if (!INGLOBALZONE(curthread) && !zone_dataset_visible(name, NULL))
2259 * zc_name name of filesystem
2260 * zc_cookie zap cursor
2261 * zc_nvlist_dst_size size of buffer for property nvlist
2264 * zc_name name of next filesystem
2265 * zc_cookie zap cursor
2266 * zc_objset_stats stats
2267 * zc_nvlist_dst property nvlist
2268 * zc_nvlist_dst_size size of property nvlist
2271 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2276 size_t orig_len = strlen(zc->zc_name);
2279 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) {
2280 if (error == ENOENT)
2281 error = SET_ERROR(ESRCH);
2285 p = strrchr(zc->zc_name, '/');
2286 if (p == NULL || p[1] != '\0')
2287 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2288 p = zc->zc_name + strlen(zc->zc_name);
2291 error = dmu_dir_list_next(os,
2292 sizeof (zc->zc_name) - (p - zc->zc_name), p,
2293 NULL, &zc->zc_cookie);
2294 if (error == ENOENT)
2295 error = SET_ERROR(ESRCH);
2296 } while (error == 0 && dataset_name_hidden(zc->zc_name));
2297 dmu_objset_rele(os, FTAG);
2300 * If it's an internal dataset (ie. with a '$' in its name),
2301 * don't try to get stats for it, otherwise we'll return ENOENT.
2303 if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2304 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2305 if (error == ENOENT) {
2306 /* We lost a race with destroy, get the next one. */
2307 zc->zc_name[orig_len] = '\0';
2316 * zc_name name of filesystem
2317 * zc_cookie zap cursor
2318 * zc_nvlist_dst_size size of buffer for property nvlist
2319 * zc_simple when set, only name is requested
2322 * zc_name name of next snapshot
2323 * zc_objset_stats stats
2324 * zc_nvlist_dst property nvlist
2325 * zc_nvlist_dst_size size of property nvlist
2328 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2333 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2335 return (error == ENOENT ? ESRCH : error);
2339 * A dataset name of maximum length cannot have any snapshots,
2340 * so exit immediately.
2342 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
2343 dmu_objset_rele(os, FTAG);
2344 return (SET_ERROR(ESRCH));
2347 error = dmu_snapshot_list_next(os,
2348 sizeof (zc->zc_name) - strlen(zc->zc_name),
2349 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2352 if (error == 0 && !zc->zc_simple) {
2354 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2356 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2360 error = dmu_objset_from_ds(ds, &ossnap);
2362 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2363 dsl_dataset_rele(ds, FTAG);
2365 } else if (error == ENOENT) {
2366 error = SET_ERROR(ESRCH);
2369 dmu_objset_rele(os, FTAG);
2370 /* if we failed, undo the @ that we tacked on to zc_name */
2372 *strchr(zc->zc_name, '@') = '\0';
2377 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2379 const char *propname = nvpair_name(pair);
2381 unsigned int vallen;
2384 zfs_userquota_prop_t type;
2390 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2392 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2393 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2395 return (SET_ERROR(EINVAL));
2399 * A correctly constructed propname is encoded as
2400 * userquota@<rid>-<domain>.
2402 if ((dash = strchr(propname, '-')) == NULL ||
2403 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2405 return (SET_ERROR(EINVAL));
2412 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2414 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2415 zfsvfs_rele(zfsvfs, FTAG);
2422 * If the named property is one that has a special function to set its value,
2423 * return 0 on success and a positive error code on failure; otherwise if it is
2424 * not one of the special properties handled by this function, return -1.
2426 * XXX: It would be better for callers of the property interface if we handled
2427 * these special cases in dsl_prop.c (in the dsl layer).
2430 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2433 const char *propname = nvpair_name(pair);
2434 zfs_prop_t prop = zfs_name_to_prop(propname);
2438 if (prop == ZPROP_INVAL) {
2439 if (zfs_prop_userquota(propname))
2440 return (zfs_prop_set_userquota(dsname, pair));
2444 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2446 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2447 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2451 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2454 VERIFY(0 == nvpair_value_uint64(pair, &intval));
2457 case ZFS_PROP_QUOTA:
2458 err = dsl_dir_set_quota(dsname, source, intval);
2460 case ZFS_PROP_REFQUOTA:
2461 err = dsl_dataset_set_refquota(dsname, source, intval);
2463 case ZFS_PROP_FILESYSTEM_LIMIT:
2464 case ZFS_PROP_SNAPSHOT_LIMIT:
2465 if (intval == UINT64_MAX) {
2466 /* clearing the limit, just do it */
2469 err = dsl_dir_activate_fs_ss_limit(dsname);
2472 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2473 * default path to set the value in the nvlist.
2478 case ZFS_PROP_RESERVATION:
2479 err = dsl_dir_set_reservation(dsname, source, intval);
2481 case ZFS_PROP_REFRESERVATION:
2482 err = dsl_dataset_set_refreservation(dsname, source, intval);
2484 case ZFS_PROP_VOLSIZE:
2485 err = zvol_set_volsize(dsname, ddi_driver_major(zfs_dip),
2488 case ZFS_PROP_VERSION:
2492 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2495 err = zfs_set_version(zfsvfs, intval);
2496 zfsvfs_rele(zfsvfs, FTAG);
2498 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2501 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2502 (void) strcpy(zc->zc_name, dsname);
2503 (void) zfs_ioc_userspace_upgrade(zc);
2504 kmem_free(zc, sizeof (zfs_cmd_t));
2516 * This function is best effort. If it fails to set any of the given properties,
2517 * it continues to set as many as it can and returns the last error
2518 * encountered. If the caller provides a non-NULL errlist, it will be filled in
2519 * with the list of names of all the properties that failed along with the
2520 * corresponding error numbers.
2522 * If every property is set successfully, zero is returned and errlist is not
2526 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2534 nvlist_t *genericnvl = fnvlist_alloc();
2535 nvlist_t *retrynvl = fnvlist_alloc();
2539 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2540 const char *propname = nvpair_name(pair);
2541 zfs_prop_t prop = zfs_name_to_prop(propname);
2544 /* decode the property value */
2546 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2548 attrs = fnvpair_value_nvlist(pair);
2549 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2551 err = SET_ERROR(EINVAL);
2554 /* Validate value type */
2555 if (err == 0 && prop == ZPROP_INVAL) {
2556 if (zfs_prop_user(propname)) {
2557 if (nvpair_type(propval) != DATA_TYPE_STRING)
2558 err = SET_ERROR(EINVAL);
2559 } else if (zfs_prop_userquota(propname)) {
2560 if (nvpair_type(propval) !=
2561 DATA_TYPE_UINT64_ARRAY)
2562 err = SET_ERROR(EINVAL);
2564 err = SET_ERROR(EINVAL);
2566 } else if (err == 0) {
2567 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2568 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2569 err = SET_ERROR(EINVAL);
2570 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2573 intval = fnvpair_value_uint64(propval);
2575 switch (zfs_prop_get_type(prop)) {
2576 case PROP_TYPE_NUMBER:
2578 case PROP_TYPE_STRING:
2579 err = SET_ERROR(EINVAL);
2581 case PROP_TYPE_INDEX:
2582 if (zfs_prop_index_to_string(prop,
2583 intval, &unused) != 0)
2584 err = SET_ERROR(EINVAL);
2588 "unknown property type");
2591 err = SET_ERROR(EINVAL);
2595 /* Validate permissions */
2597 err = zfs_check_settable(dsname, pair, CRED());
2600 err = zfs_prop_set_special(dsname, source, pair);
2603 * For better performance we build up a list of
2604 * properties to set in a single transaction.
2606 err = nvlist_add_nvpair(genericnvl, pair);
2607 } else if (err != 0 && nvl != retrynvl) {
2609 * This may be a spurious error caused by
2610 * receiving quota and reservation out of order.
2611 * Try again in a second pass.
2613 err = nvlist_add_nvpair(retrynvl, pair);
2618 if (errlist != NULL)
2619 fnvlist_add_int32(errlist, propname, err);
2624 if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2629 if (!nvlist_empty(genericnvl) &&
2630 dsl_props_set(dsname, source, genericnvl) != 0) {
2632 * If this fails, we still want to set as many properties as we
2633 * can, so try setting them individually.
2636 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2637 const char *propname = nvpair_name(pair);
2641 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2643 attrs = fnvpair_value_nvlist(pair);
2644 propval = fnvlist_lookup_nvpair(attrs,
2648 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2649 strval = fnvpair_value_string(propval);
2650 err = dsl_prop_set_string(dsname, propname,
2653 intval = fnvpair_value_uint64(propval);
2654 err = dsl_prop_set_int(dsname, propname, source,
2659 if (errlist != NULL) {
2660 fnvlist_add_int32(errlist, propname,
2667 nvlist_free(genericnvl);
2668 nvlist_free(retrynvl);
2674 * Check that all the properties are valid user properties.
2677 zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2679 nvpair_t *pair = NULL;
2682 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2683 const char *propname = nvpair_name(pair);
2685 if (!zfs_prop_user(propname) ||
2686 nvpair_type(pair) != DATA_TYPE_STRING)
2687 return (SET_ERROR(EINVAL));
2689 if (error = zfs_secpolicy_write_perms(fsname,
2690 ZFS_DELEG_PERM_USERPROP, CRED()))
2693 if (strlen(propname) >= ZAP_MAXNAMELEN)
2694 return (SET_ERROR(ENAMETOOLONG));
2696 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN)
2703 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2707 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2710 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2711 if (nvlist_exists(skipped, nvpair_name(pair)))
2714 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2719 clear_received_props(const char *dsname, nvlist_t *props,
2723 nvlist_t *cleared_props = NULL;
2724 props_skip(props, skipped, &cleared_props);
2725 if (!nvlist_empty(cleared_props)) {
2727 * Acts on local properties until the dataset has received
2728 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2730 zprop_source_t flags = (ZPROP_SRC_NONE |
2731 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2732 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2734 nvlist_free(cleared_props);
2740 * zc_name name of filesystem
2741 * zc_value name of property to set
2742 * zc_nvlist_src{_size} nvlist of properties to apply
2743 * zc_cookie received properties flag
2746 * zc_nvlist_dst{_size} error for each unapplied received property
2749 zfs_ioc_set_prop(zfs_cmd_t *zc)
2752 boolean_t received = zc->zc_cookie;
2753 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2758 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2759 zc->zc_iflags, &nvl)) != 0)
2763 nvlist_t *origprops;
2765 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2766 (void) clear_received_props(zc->zc_name,
2768 nvlist_free(origprops);
2771 error = dsl_prop_set_hasrecvd(zc->zc_name);
2774 errors = fnvlist_alloc();
2776 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2778 if (zc->zc_nvlist_dst != 0 && errors != NULL) {
2779 (void) put_nvlist(zc, errors);
2782 nvlist_free(errors);
2789 * zc_name name of filesystem
2790 * zc_value name of property to inherit
2791 * zc_cookie revert to received value if TRUE
2796 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2798 const char *propname = zc->zc_value;
2799 zfs_prop_t prop = zfs_name_to_prop(propname);
2800 boolean_t received = zc->zc_cookie;
2801 zprop_source_t source = (received
2802 ? ZPROP_SRC_NONE /* revert to received value, if any */
2803 : ZPROP_SRC_INHERITED); /* explicitly inherit */
2812 * zfs_prop_set_special() expects properties in the form of an
2813 * nvpair with type info.
2815 if (prop == ZPROP_INVAL) {
2816 if (!zfs_prop_user(propname))
2817 return (SET_ERROR(EINVAL));
2819 type = PROP_TYPE_STRING;
2820 } else if (prop == ZFS_PROP_VOLSIZE ||
2821 prop == ZFS_PROP_VERSION) {
2822 return (SET_ERROR(EINVAL));
2824 type = zfs_prop_get_type(prop);
2827 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2830 case PROP_TYPE_STRING:
2831 VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2833 case PROP_TYPE_NUMBER:
2834 case PROP_TYPE_INDEX:
2835 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2839 return (SET_ERROR(EINVAL));
2842 pair = nvlist_next_nvpair(dummy, NULL);
2843 err = zfs_prop_set_special(zc->zc_name, source, pair);
2846 return (err); /* special property already handled */
2849 * Only check this in the non-received case. We want to allow
2850 * 'inherit -S' to revert non-inheritable properties like quota
2851 * and reservation to the received or default values even though
2852 * they are not considered inheritable.
2854 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2855 return (SET_ERROR(EINVAL));
2858 /* property name has been validated by zfs_secpolicy_inherit_prop() */
2859 return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source));
2863 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2870 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2871 zc->zc_iflags, &props))
2875 * If the only property is the configfile, then just do a spa_lookup()
2876 * to handle the faulted case.
2878 pair = nvlist_next_nvpair(props, NULL);
2879 if (pair != NULL && strcmp(nvpair_name(pair),
2880 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2881 nvlist_next_nvpair(props, pair) == NULL) {
2882 mutex_enter(&spa_namespace_lock);
2883 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2884 spa_configfile_set(spa, props, B_FALSE);
2885 spa_config_sync(spa, B_FALSE, B_TRUE);
2887 mutex_exit(&spa_namespace_lock);
2894 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2899 error = spa_prop_set(spa, props);
2902 spa_close(spa, FTAG);
2908 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2912 nvlist_t *nvp = NULL;
2914 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2916 * If the pool is faulted, there may be properties we can still
2917 * get (such as altroot and cachefile), so attempt to get them
2920 mutex_enter(&spa_namespace_lock);
2921 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2922 error = spa_prop_get(spa, &nvp);
2923 mutex_exit(&spa_namespace_lock);
2925 error = spa_prop_get(spa, &nvp);
2926 spa_close(spa, FTAG);
2929 if (error == 0 && zc->zc_nvlist_dst != 0)
2930 error = put_nvlist(zc, nvp);
2932 error = SET_ERROR(EFAULT);
2940 * zc_name name of filesystem
2941 * zc_nvlist_src{_size} nvlist of delegated permissions
2942 * zc_perm_action allow/unallow flag
2947 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2950 nvlist_t *fsaclnv = NULL;
2952 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2953 zc->zc_iflags, &fsaclnv)) != 0)
2957 * Verify nvlist is constructed correctly
2959 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2960 nvlist_free(fsaclnv);
2961 return (SET_ERROR(EINVAL));
2965 * If we don't have PRIV_SYS_MOUNT, then validate
2966 * that user is allowed to hand out each permission in
2970 error = secpolicy_zfs(CRED());
2972 if (zc->zc_perm_action == B_FALSE) {
2973 error = dsl_deleg_can_allow(zc->zc_name,
2976 error = dsl_deleg_can_unallow(zc->zc_name,
2982 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2984 nvlist_free(fsaclnv);
2990 * zc_name name of filesystem
2993 * zc_nvlist_src{_size} nvlist of delegated permissions
2996 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
3001 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
3002 error = put_nvlist(zc, nvp);
3010 * Search the vfs list for a specified resource. Returns a pointer to it
3011 * or NULL if no suitable entry is found. The caller of this routine
3012 * is responsible for releasing the returned vfs pointer.
3015 zfs_get_vfs(const char *resource)
3019 mtx_lock(&mountlist_mtx);
3020 TAILQ_FOREACH(vfsp, &mountlist, mnt_list) {
3021 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
3026 mtx_unlock(&mountlist_mtx);
3032 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
3034 zfs_creat_t *zct = arg;
3036 zfs_create_fs(os, cr, zct->zct_zplprops, tx);
3039 #define ZFS_PROP_UNDEFINED ((uint64_t)-1)
3043 * os parent objset pointer (NULL if root fs)
3044 * fuids_ok fuids allowed in this version of the spa?
3045 * sa_ok SAs allowed in this version of the spa?
3046 * createprops list of properties requested by creator
3049 * zplprops values for the zplprops we attach to the master node object
3050 * is_ci true if requested file system will be purely case-insensitive
3052 * Determine the settings for utf8only, normalization and
3053 * casesensitivity. Specific values may have been requested by the
3054 * creator and/or we can inherit values from the parent dataset. If
3055 * the file system is of too early a vintage, a creator can not
3056 * request settings for these properties, even if the requested
3057 * setting is the default value. We don't actually want to create dsl
3058 * properties for these, so remove them from the source nvlist after
3062 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
3063 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
3064 nvlist_t *zplprops, boolean_t *is_ci)
3066 uint64_t sense = ZFS_PROP_UNDEFINED;
3067 uint64_t norm = ZFS_PROP_UNDEFINED;
3068 uint64_t u8 = ZFS_PROP_UNDEFINED;
3070 ASSERT(zplprops != NULL);
3073 * Pull out creator prop choices, if any.
3076 (void) nvlist_lookup_uint64(createprops,
3077 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
3078 (void) nvlist_lookup_uint64(createprops,
3079 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
3080 (void) nvlist_remove_all(createprops,
3081 zfs_prop_to_name(ZFS_PROP_NORMALIZE));
3082 (void) nvlist_lookup_uint64(createprops,
3083 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
3084 (void) nvlist_remove_all(createprops,
3085 zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
3086 (void) nvlist_lookup_uint64(createprops,
3087 zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3088 (void) nvlist_remove_all(createprops,
3089 zfs_prop_to_name(ZFS_PROP_CASE));
3093 * If the zpl version requested is whacky or the file system
3094 * or pool is version is too "young" to support normalization
3095 * and the creator tried to set a value for one of the props,
3098 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3099 (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3100 (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3101 (zplver < ZPL_VERSION_NORMALIZATION &&
3102 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3103 sense != ZFS_PROP_UNDEFINED)))
3104 return (SET_ERROR(ENOTSUP));
3107 * Put the version in the zplprops
3109 VERIFY(nvlist_add_uint64(zplprops,
3110 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3112 if (norm == ZFS_PROP_UNDEFINED)
3113 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
3114 VERIFY(nvlist_add_uint64(zplprops,
3115 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3118 * If we're normalizing, names must always be valid UTF-8 strings.
3122 if (u8 == ZFS_PROP_UNDEFINED)
3123 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
3124 VERIFY(nvlist_add_uint64(zplprops,
3125 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3127 if (sense == ZFS_PROP_UNDEFINED)
3128 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
3129 VERIFY(nvlist_add_uint64(zplprops,
3130 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3133 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
3139 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3140 nvlist_t *zplprops, boolean_t *is_ci)
3142 boolean_t fuids_ok, sa_ok;
3143 uint64_t zplver = ZPL_VERSION;
3144 objset_t *os = NULL;
3145 char parentname[MAXNAMELEN];
3151 (void) strlcpy(parentname, dataset, sizeof (parentname));
3152 cp = strrchr(parentname, '/');
3156 if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3159 spa_vers = spa_version(spa);
3160 spa_close(spa, FTAG);
3162 zplver = zfs_zpl_version_map(spa_vers);
3163 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3164 sa_ok = (zplver >= ZPL_VERSION_SA);
3167 * Open parent object set so we can inherit zplprop values.
3169 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3172 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3174 dmu_objset_rele(os, FTAG);
3179 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3180 nvlist_t *zplprops, boolean_t *is_ci)
3184 uint64_t zplver = ZPL_VERSION;
3187 zplver = zfs_zpl_version_map(spa_vers);
3188 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3189 sa_ok = (zplver >= ZPL_VERSION_SA);
3191 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3192 createprops, zplprops, is_ci);
3198 * "type" -> dmu_objset_type_t (int32)
3199 * (optional) "props" -> { prop -> value }
3202 * outnvl: propname -> error code (int32)
3205 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3208 zfs_creat_t zct = { 0 };
3209 nvlist_t *nvprops = NULL;
3210 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3212 dmu_objset_type_t type;
3213 boolean_t is_insensitive = B_FALSE;
3215 if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3216 return (SET_ERROR(EINVAL));
3218 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3222 cbfunc = zfs_create_cb;
3226 cbfunc = zvol_create_cb;
3233 if (strchr(fsname, '@') ||
3234 strchr(fsname, '%'))
3235 return (SET_ERROR(EINVAL));
3237 zct.zct_props = nvprops;
3240 return (SET_ERROR(EINVAL));
3242 if (type == DMU_OST_ZVOL) {
3243 uint64_t volsize, volblocksize;
3245 if (nvprops == NULL)
3246 return (SET_ERROR(EINVAL));
3247 if (nvlist_lookup_uint64(nvprops,
3248 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3249 return (SET_ERROR(EINVAL));
3251 if ((error = nvlist_lookup_uint64(nvprops,
3252 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3253 &volblocksize)) != 0 && error != ENOENT)
3254 return (SET_ERROR(EINVAL));
3257 volblocksize = zfs_prop_default_numeric(
3258 ZFS_PROP_VOLBLOCKSIZE);
3260 if ((error = zvol_check_volblocksize(
3261 volblocksize)) != 0 ||
3262 (error = zvol_check_volsize(volsize,
3263 volblocksize)) != 0)
3265 } else if (type == DMU_OST_ZFS) {
3269 * We have to have normalization and
3270 * case-folding flags correct when we do the
3271 * file system creation, so go figure them out
3274 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3275 NV_UNIQUE_NAME, KM_SLEEP) == 0);
3276 error = zfs_fill_zplprops(fsname, nvprops,
3277 zct.zct_zplprops, &is_insensitive);
3279 nvlist_free(zct.zct_zplprops);
3284 error = dmu_objset_create(fsname, type,
3285 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3286 nvlist_free(zct.zct_zplprops);
3289 * It would be nice to do this atomically.
3292 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3295 (void) dsl_destroy_head(fsname);
3298 if (error == 0 && type == DMU_OST_ZVOL)
3299 zvol_create_minors(fsname);
3306 * "origin" -> name of origin snapshot
3307 * (optional) "props" -> { prop -> value }
3310 * outnvl: propname -> error code (int32)
3313 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3316 nvlist_t *nvprops = NULL;
3319 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3320 return (SET_ERROR(EINVAL));
3321 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3323 if (strchr(fsname, '@') ||
3324 strchr(fsname, '%'))
3325 return (SET_ERROR(EINVAL));
3327 if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3328 return (SET_ERROR(EINVAL));
3329 error = dmu_objset_clone(fsname, origin_name);
3334 * It would be nice to do this atomically.
3337 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3340 (void) dsl_destroy_head(fsname);
3344 zvol_create_minors(fsname);
3351 * "snaps" -> { snapshot1, snapshot2 }
3352 * (optional) "props" -> { prop -> value (string) }
3355 * outnvl: snapshot -> error code (int32)
3358 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3361 nvlist_t *props = NULL;
3365 (void) nvlist_lookup_nvlist(innvl, "props", &props);
3366 if ((error = zfs_check_userprops(poolname, props)) != 0)
3369 if (!nvlist_empty(props) &&
3370 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3371 return (SET_ERROR(ENOTSUP));
3373 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3374 return (SET_ERROR(EINVAL));
3375 poollen = strlen(poolname);
3376 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3377 pair = nvlist_next_nvpair(snaps, pair)) {
3378 const char *name = nvpair_name(pair);
3379 const char *cp = strchr(name, '@');
3382 * The snap name must contain an @, and the part after it must
3383 * contain only valid characters.
3386 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3387 return (SET_ERROR(EINVAL));
3390 * The snap must be in the specified pool.
3392 if (strncmp(name, poolname, poollen) != 0 ||
3393 (name[poollen] != '/' && name[poollen] != '@'))
3394 return (SET_ERROR(EXDEV));
3396 /* This must be the only snap of this fs. */
3397 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3398 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3399 if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3401 return (SET_ERROR(EXDEV));
3406 error = dsl_dataset_snapshot(snaps, props, outnvl);
3411 * innvl: "message" -> string
3415 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3423 * The poolname in the ioctl is not set, we get it from the TSD,
3424 * which was set at the end of the last successful ioctl that allows
3425 * logging. The secpolicy func already checked that it is set.
3426 * Only one log ioctl is allowed after each successful ioctl, so
3427 * we clear the TSD here.
3429 poolname = tsd_get(zfs_allow_log_key);
3430 (void) tsd_set(zfs_allow_log_key, NULL);
3431 error = spa_open(poolname, &spa, FTAG);
3436 if (nvlist_lookup_string(innvl, "message", &message) != 0) {
3437 spa_close(spa, FTAG);
3438 return (SET_ERROR(EINVAL));
3441 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3442 spa_close(spa, FTAG);
3443 return (SET_ERROR(ENOTSUP));
3446 error = spa_history_log(spa, message);
3447 spa_close(spa, FTAG);
3452 * The dp_config_rwlock must not be held when calling this, because the
3453 * unmount may need to write out data.
3455 * This function is best-effort. Callers must deal gracefully if it
3456 * remains mounted (or is remounted after this call).
3458 * Returns 0 if the argument is not a snapshot, or it is not currently a
3459 * filesystem, or we were able to unmount it. Returns error code otherwise.
3462 zfs_unmount_snap(const char *snapname)
3468 if (strchr(snapname, '@') == NULL)
3471 vfsp = zfs_get_vfs(snapname);
3475 zfsvfs = vfsp->vfs_data;
3476 ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os)));
3478 err = vn_vfswlock(vfsp->vfs_vnodecovered);
3481 return (SET_ERROR(err));
3484 * Always force the unmount for snapshots.
3488 (void) dounmount(vfsp, MS_FORCE, kcred);
3490 mtx_lock(&Giant); /* dounmount() */
3491 (void) dounmount(vfsp, MS_FORCE, curthread);
3492 mtx_unlock(&Giant); /* dounmount() */
3499 zfs_unmount_snap_cb(const char *snapname, void *arg)
3501 return (zfs_unmount_snap(snapname));
3505 * When a clone is destroyed, its origin may also need to be destroyed,
3506 * in which case it must be unmounted. This routine will do that unmount
3510 zfs_destroy_unmount_origin(const char *fsname)
3516 error = dmu_objset_hold(fsname, FTAG, &os);
3519 ds = dmu_objset_ds(os);
3520 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3521 char originname[MAXNAMELEN];
3522 dsl_dataset_name(ds->ds_prev, originname);
3523 dmu_objset_rele(os, FTAG);
3524 (void) zfs_unmount_snap(originname);
3526 dmu_objset_rele(os, FTAG);
3532 * "snaps" -> { snapshot1, snapshot2 }
3533 * (optional boolean) "defer"
3536 * outnvl: snapshot -> error code (int32)
3541 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3548 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3549 return (SET_ERROR(EINVAL));
3550 defer = nvlist_exists(innvl, "defer");
3552 poollen = strlen(poolname);
3553 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3554 pair = nvlist_next_nvpair(snaps, pair)) {
3555 const char *name = nvpair_name(pair);
3558 * The snap must be in the specified pool to prevent the
3559 * invalid removal of zvol minors below.
3561 if (strncmp(name, poolname, poollen) != 0 ||
3562 (name[poollen] != '/' && name[poollen] != '@'))
3563 return (SET_ERROR(EXDEV));
3565 error = zfs_unmount_snap(name);
3568 #if defined(__FreeBSD__)
3569 zvol_remove_minors(name);
3573 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3577 * Create bookmarks. Bookmark names are of the form <fs>#<bmark>.
3578 * All bookmarks must be in the same pool.
3581 * bookmark1 -> snapshot1, bookmark2 -> snapshot2
3584 * outnvl: bookmark -> error code (int32)
3589 zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3591 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3592 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3596 * Verify the snapshot argument.
3598 if (nvpair_value_string(pair, &snap_name) != 0)
3599 return (SET_ERROR(EINVAL));
3602 /* Verify that the keys (bookmarks) are unique */
3603 for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair);
3604 pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) {
3605 if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0)
3606 return (SET_ERROR(EINVAL));
3610 return (dsl_bookmark_create(innvl, outnvl));
3615 * property 1, property 2, ...
3619 * bookmark name 1 -> { property 1, property 2, ... },
3620 * bookmark name 2 -> { property 1, property 2, ... }
3625 zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3627 return (dsl_get_bookmarks(fsname, innvl, outnvl));
3632 * bookmark name 1, bookmark name 2
3635 * outnvl: bookmark -> error code (int32)
3639 zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl,
3644 poollen = strlen(poolname);
3645 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3646 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3647 const char *name = nvpair_name(pair);
3648 const char *cp = strchr(name, '#');
3651 * The bookmark name must contain an #, and the part after it
3652 * must contain only valid characters.
3655 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3656 return (SET_ERROR(EINVAL));
3659 * The bookmark must be in the specified pool.
3661 if (strncmp(name, poolname, poollen) != 0 ||
3662 (name[poollen] != '/' && name[poollen] != '#'))
3663 return (SET_ERROR(EXDEV));
3666 error = dsl_bookmark_destroy(innvl, outnvl);
3672 * zc_name name of dataset to destroy
3673 * zc_objset_type type of objset
3674 * zc_defer_destroy mark for deferred destroy
3679 zfs_ioc_destroy(zfs_cmd_t *zc)
3683 if (zc->zc_objset_type == DMU_OST_ZFS) {
3684 err = zfs_unmount_snap(zc->zc_name);
3689 if (strchr(zc->zc_name, '@'))
3690 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3692 err = dsl_destroy_head(zc->zc_name);
3693 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3695 zvol_remove_minors(zc->zc_name);
3697 (void) zvol_remove_minor(zc->zc_name);
3703 * fsname is name of dataset to rollback (to most recent snapshot)
3705 * innvl is not used.
3707 * outnvl: "target" -> name of most recent snapshot
3712 zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl)
3717 if (getzfsvfs(fsname, &zfsvfs) == 0) {
3718 error = zfs_suspend_fs(zfsvfs);
3722 error = dsl_dataset_rollback(fsname, zfsvfs, outnvl);
3723 resume_err = zfs_resume_fs(zfsvfs, fsname);
3724 error = error ? error : resume_err;
3726 VFS_RELE(zfsvfs->z_vfs);
3728 error = dsl_dataset_rollback(fsname, NULL, outnvl);
3734 recursive_unmount(const char *fsname, void *arg)
3736 const char *snapname = arg;
3737 char fullname[MAXNAMELEN];
3739 (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname);
3740 return (zfs_unmount_snap(fullname));
3745 * zc_name old name of dataset
3746 * zc_value new name of dataset
3747 * zc_cookie recursive flag (only valid for snapshots)
3752 zfs_ioc_rename(zfs_cmd_t *zc)
3754 boolean_t recursive = zc->zc_cookie & 1;
3756 boolean_t allow_mounted = zc->zc_cookie & 2;
3760 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3761 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3762 strchr(zc->zc_value, '%'))
3763 return (SET_ERROR(EINVAL));
3765 at = strchr(zc->zc_name, '@');
3767 /* snaps must be in same fs */
3770 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
3771 return (SET_ERROR(EXDEV));
3774 if (zc->zc_objset_type == DMU_OST_ZFS) {
3776 if (zc->zc_objset_type == DMU_OST_ZFS && allow_mounted) {
3778 error = dmu_objset_find(zc->zc_name,
3779 recursive_unmount, at + 1,
3780 recursive ? DS_FIND_CHILDREN : 0);
3786 error = dsl_dataset_rename_snapshot(zc->zc_name,
3787 at + 1, strchr(zc->zc_value, '@') + 1, recursive);
3793 if (zc->zc_objset_type == DMU_OST_ZVOL)
3794 (void) zvol_remove_minor(zc->zc_name);
3796 return (dsl_dir_rename(zc->zc_name, zc->zc_value));
3801 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3803 const char *propname = nvpair_name(pair);
3804 boolean_t issnap = (strchr(dsname, '@') != NULL);
3805 zfs_prop_t prop = zfs_name_to_prop(propname);
3809 if (prop == ZPROP_INVAL) {
3810 if (zfs_prop_user(propname)) {
3811 if (err = zfs_secpolicy_write_perms(dsname,
3812 ZFS_DELEG_PERM_USERPROP, cr))
3817 if (!issnap && zfs_prop_userquota(propname)) {
3818 const char *perm = NULL;
3819 const char *uq_prefix =
3820 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3821 const char *gq_prefix =
3822 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3824 if (strncmp(propname, uq_prefix,
3825 strlen(uq_prefix)) == 0) {
3826 perm = ZFS_DELEG_PERM_USERQUOTA;
3827 } else if (strncmp(propname, gq_prefix,
3828 strlen(gq_prefix)) == 0) {
3829 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3831 /* USERUSED and GROUPUSED are read-only */
3832 return (SET_ERROR(EINVAL));
3835 if (err = zfs_secpolicy_write_perms(dsname, perm, cr))
3840 return (SET_ERROR(EINVAL));
3844 return (SET_ERROR(EINVAL));
3846 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3848 * dsl_prop_get_all_impl() returns properties in this
3852 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3853 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3858 * Check that this value is valid for this pool version
3861 case ZFS_PROP_COMPRESSION:
3863 * If the user specified gzip compression, make sure
3864 * the SPA supports it. We ignore any errors here since
3865 * we'll catch them later.
3867 if (nvpair_value_uint64(pair, &intval) == 0) {
3868 if (intval >= ZIO_COMPRESS_GZIP_1 &&
3869 intval <= ZIO_COMPRESS_GZIP_9 &&
3870 zfs_earlier_version(dsname,
3871 SPA_VERSION_GZIP_COMPRESSION)) {
3872 return (SET_ERROR(ENOTSUP));
3875 if (intval == ZIO_COMPRESS_ZLE &&
3876 zfs_earlier_version(dsname,
3877 SPA_VERSION_ZLE_COMPRESSION))
3878 return (SET_ERROR(ENOTSUP));
3880 if (intval == ZIO_COMPRESS_LZ4) {
3883 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3886 if (!spa_feature_is_enabled(spa,
3887 SPA_FEATURE_LZ4_COMPRESS)) {
3888 spa_close(spa, FTAG);
3889 return (SET_ERROR(ENOTSUP));
3891 spa_close(spa, FTAG);
3895 * If this is a bootable dataset then
3896 * verify that the compression algorithm
3897 * is supported for booting. We must return
3898 * something other than ENOTSUP since it
3899 * implies a downrev pool version.
3901 if (zfs_is_bootfs(dsname) &&
3902 !BOOTFS_COMPRESS_VALID(intval)) {
3903 return (SET_ERROR(ERANGE));
3908 case ZFS_PROP_COPIES:
3909 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3910 return (SET_ERROR(ENOTSUP));
3913 case ZFS_PROP_DEDUP:
3914 if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3915 return (SET_ERROR(ENOTSUP));
3918 case ZFS_PROP_RECORDSIZE:
3919 /* Record sizes above 128k need the feature to be enabled */
3920 if (nvpair_value_uint64(pair, &intval) == 0 &&
3921 intval > SPA_OLD_MAXBLOCKSIZE) {
3925 * If this is a bootable dataset then
3926 * the we don't allow large (>128K) blocks,
3927 * because GRUB doesn't support them.
3929 if (zfs_is_bootfs(dsname) &&
3930 intval > SPA_OLD_MAXBLOCKSIZE) {
3931 return (SET_ERROR(EDOM));
3935 * We don't allow setting the property above 1MB,
3936 * unless the tunable has been changed.
3938 if (intval > zfs_max_recordsize ||
3939 intval > SPA_MAXBLOCKSIZE)
3940 return (SET_ERROR(EDOM));
3942 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3945 if (!spa_feature_is_enabled(spa,
3946 SPA_FEATURE_LARGE_BLOCKS)) {
3947 spa_close(spa, FTAG);
3948 return (SET_ERROR(ENOTSUP));
3950 spa_close(spa, FTAG);
3954 case ZFS_PROP_SHARESMB:
3955 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3956 return (SET_ERROR(ENOTSUP));
3959 case ZFS_PROP_ACLINHERIT:
3960 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3961 nvpair_value_uint64(pair, &intval) == 0) {
3962 if (intval == ZFS_ACL_PASSTHROUGH_X &&
3963 zfs_earlier_version(dsname,
3964 SPA_VERSION_PASSTHROUGH_X))
3965 return (SET_ERROR(ENOTSUP));
3970 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3974 * Checks for a race condition to make sure we don't increment a feature flag
3978 zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx)
3980 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3981 spa_feature_t *featurep = arg;
3983 if (!spa_feature_is_active(spa, *featurep))
3986 return (SET_ERROR(EBUSY));
3990 * The callback invoked on feature activation in the sync task caused by
3991 * zfs_prop_activate_feature.
3994 zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx)
3996 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3997 spa_feature_t *featurep = arg;
3999 spa_feature_incr(spa, *featurep, tx);
4003 * Activates a feature on a pool in response to a property setting. This
4004 * creates a new sync task which modifies the pool to reflect the feature
4008 zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature)
4012 /* EBUSY here indicates that the feature is already active */
4013 err = dsl_sync_task(spa_name(spa),
4014 zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync,
4015 &feature, 2, ZFS_SPACE_CHECK_RESERVED);
4017 if (err != 0 && err != EBUSY)
4024 * Removes properties from the given props list that fail permission checks
4025 * needed to clear them and to restore them in case of a receive error. For each
4026 * property, make sure we have both set and inherit permissions.
4028 * Returns the first error encountered if any permission checks fail. If the
4029 * caller provides a non-NULL errlist, it also gives the complete list of names
4030 * of all the properties that failed a permission check along with the
4031 * corresponding error numbers. The caller is responsible for freeing the
4034 * If every property checks out successfully, zero is returned and the list
4035 * pointed at by errlist is NULL.
4038 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
4041 nvpair_t *pair, *next_pair;
4048 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4050 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
4051 (void) strcpy(zc->zc_name, dataset);
4052 pair = nvlist_next_nvpair(props, NULL);
4053 while (pair != NULL) {
4054 next_pair = nvlist_next_nvpair(props, pair);
4056 (void) strcpy(zc->zc_value, nvpair_name(pair));
4057 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
4058 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
4059 VERIFY(nvlist_remove_nvpair(props, pair) == 0);
4060 VERIFY(nvlist_add_int32(errors,
4061 zc->zc_value, err) == 0);
4065 kmem_free(zc, sizeof (zfs_cmd_t));
4067 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
4068 nvlist_free(errors);
4071 VERIFY(nvpair_value_int32(pair, &rv) == 0);
4074 if (errlist == NULL)
4075 nvlist_free(errors);
4083 propval_equals(nvpair_t *p1, nvpair_t *p2)
4085 if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
4086 /* dsl_prop_get_all_impl() format */
4088 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
4089 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4093 if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
4095 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
4096 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4100 if (nvpair_type(p1) != nvpair_type(p2))
4103 if (nvpair_type(p1) == DATA_TYPE_STRING) {
4104 char *valstr1, *valstr2;
4106 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
4107 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
4108 return (strcmp(valstr1, valstr2) == 0);
4110 uint64_t intval1, intval2;
4112 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
4113 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
4114 return (intval1 == intval2);
4119 * Remove properties from props if they are not going to change (as determined
4120 * by comparison with origprops). Remove them from origprops as well, since we
4121 * do not need to clear or restore properties that won't change.
4124 props_reduce(nvlist_t *props, nvlist_t *origprops)
4126 nvpair_t *pair, *next_pair;
4128 if (origprops == NULL)
4129 return; /* all props need to be received */
4131 pair = nvlist_next_nvpair(props, NULL);
4132 while (pair != NULL) {
4133 const char *propname = nvpair_name(pair);
4136 next_pair = nvlist_next_nvpair(props, pair);
4138 if ((nvlist_lookup_nvpair(origprops, propname,
4139 &match) != 0) || !propval_equals(pair, match))
4140 goto next; /* need to set received value */
4142 /* don't clear the existing received value */
4143 (void) nvlist_remove_nvpair(origprops, match);
4144 /* don't bother receiving the property */
4145 (void) nvlist_remove_nvpair(props, pair);
4152 static boolean_t zfs_ioc_recv_inject_err;
4157 * zc_name name of containing filesystem
4158 * zc_nvlist_src{_size} nvlist of properties to apply
4159 * zc_value name of snapshot to create
4160 * zc_string name of clone origin (if DRR_FLAG_CLONE)
4161 * zc_cookie file descriptor to recv from
4162 * zc_begin_record the BEGIN record of the stream (not byteswapped)
4163 * zc_guid force flag
4164 * zc_cleanup_fd cleanup-on-exit file descriptor
4165 * zc_action_handle handle for this guid/ds mapping (or zero on first call)
4168 * zc_cookie number of bytes read
4169 * zc_nvlist_dst{_size} error for each unapplied received property
4170 * zc_obj zprop_errflags_t
4171 * zc_action_handle handle for this guid/ds mapping
4174 zfs_ioc_recv(zfs_cmd_t *zc)
4177 dmu_recv_cookie_t drc;
4178 boolean_t force = (boolean_t)zc->zc_guid;
4181 int props_error = 0;
4184 nvlist_t *props = NULL; /* sent properties */
4185 nvlist_t *origprops = NULL; /* existing properties */
4186 char *origin = NULL;
4188 char tofs[ZFS_MAXNAMELEN];
4189 cap_rights_t rights;
4190 boolean_t first_recvd_props = B_FALSE;
4192 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
4193 strchr(zc->zc_value, '@') == NULL ||
4194 strchr(zc->zc_value, '%'))
4195 return (SET_ERROR(EINVAL));
4197 (void) strcpy(tofs, zc->zc_value);
4198 tosnap = strchr(tofs, '@');
4201 if (zc->zc_nvlist_src != 0 &&
4202 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
4203 zc->zc_iflags, &props)) != 0)
4207 fp = getf(fd, cap_rights_init(&rights, CAP_PREAD));
4210 return (SET_ERROR(EBADF));
4213 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4215 if (zc->zc_string[0])
4216 origin = zc->zc_string;
4218 error = dmu_recv_begin(tofs, tosnap,
4219 &zc->zc_begin_record, force, origin, &drc);
4224 * Set properties before we receive the stream so that they are applied
4225 * to the new data. Note that we must call dmu_recv_stream() if
4226 * dmu_recv_begin() succeeds.
4228 if (props != NULL && !drc.drc_newfs) {
4229 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
4230 SPA_VERSION_RECVD_PROPS &&
4231 !dsl_prop_get_hasrecvd(tofs))
4232 first_recvd_props = B_TRUE;
4235 * If new received properties are supplied, they are to
4236 * completely replace the existing received properties, so stash
4237 * away the existing ones.
4239 if (dsl_prop_get_received(tofs, &origprops) == 0) {
4240 nvlist_t *errlist = NULL;
4242 * Don't bother writing a property if its value won't
4243 * change (and avoid the unnecessary security checks).
4245 * The first receive after SPA_VERSION_RECVD_PROPS is a
4246 * special case where we blow away all local properties
4249 if (!first_recvd_props)
4250 props_reduce(props, origprops);
4251 if (zfs_check_clearable(tofs, origprops, &errlist) != 0)
4252 (void) nvlist_merge(errors, errlist, 0);
4253 nvlist_free(errlist);
4255 if (clear_received_props(tofs, origprops,
4256 first_recvd_props ? NULL : props) != 0)
4257 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4259 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4263 if (props != NULL) {
4264 props_error = dsl_prop_set_hasrecvd(tofs);
4266 if (props_error == 0) {
4267 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4272 if (zc->zc_nvlist_dst_size != 0 &&
4273 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4274 put_nvlist(zc, errors) != 0)) {
4276 * Caller made zc->zc_nvlist_dst less than the minimum expected
4277 * size or supplied an invalid address.
4279 props_error = SET_ERROR(EINVAL);
4283 error = dmu_recv_stream(&drc, fp, &off, zc->zc_cleanup_fd,
4284 &zc->zc_action_handle);
4287 zfsvfs_t *zfsvfs = NULL;
4289 if (getzfsvfs(tofs, &zfsvfs) == 0) {
4293 error = zfs_suspend_fs(zfsvfs);
4295 * If the suspend fails, then the recv_end will
4296 * likely also fail, and clean up after itself.
4298 end_err = dmu_recv_end(&drc, zfsvfs);
4300 error = zfs_resume_fs(zfsvfs, tofs);
4301 error = error ? error : end_err;
4302 VFS_RELE(zfsvfs->z_vfs);
4304 error = dmu_recv_end(&drc, NULL);
4308 zc->zc_cookie = off - fp->f_offset;
4309 if (off >= 0 && off <= MAXOFFSET_T)
4313 if (zfs_ioc_recv_inject_err) {
4314 zfs_ioc_recv_inject_err = B_FALSE;
4321 zvol_create_minors(tofs);
4325 * On error, restore the original props.
4327 if (error != 0 && props != NULL && !drc.drc_newfs) {
4328 if (clear_received_props(tofs, props, NULL) != 0) {
4330 * We failed to clear the received properties.
4331 * Since we may have left a $recvd value on the
4332 * system, we can't clear the $hasrecvd flag.
4334 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4335 } else if (first_recvd_props) {
4336 dsl_prop_unset_hasrecvd(tofs);
4339 if (origprops == NULL && !drc.drc_newfs) {
4340 /* We failed to stash the original properties. */
4341 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4345 * dsl_props_set() will not convert RECEIVED to LOCAL on or
4346 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4347 * explictly if we're restoring local properties cleared in the
4348 * first new-style receive.
4350 if (origprops != NULL &&
4351 zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4352 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4353 origprops, NULL) != 0) {
4355 * We stashed the original properties but failed to
4358 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4363 nvlist_free(origprops);
4364 nvlist_free(errors);
4368 error = props_error;
4375 * zc_name name of snapshot to send
4376 * zc_cookie file descriptor to send stream to
4377 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj)
4378 * zc_sendobj objsetid of snapshot to send
4379 * zc_fromobj objsetid of incremental fromsnap (may be zero)
4380 * zc_guid if set, estimate size of stream only. zc_cookie is ignored.
4381 * output size in zc_objset_type.
4382 * zc_flags lzc_send_flags
4385 * zc_objset_type estimated size, if zc_guid is set
4388 zfs_ioc_send(zfs_cmd_t *zc)
4392 boolean_t estimate = (zc->zc_guid != 0);
4393 boolean_t embedok = (zc->zc_flags & 0x1);
4394 boolean_t large_block_ok = (zc->zc_flags & 0x2);
4396 if (zc->zc_obj != 0) {
4398 dsl_dataset_t *tosnap;
4400 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4404 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4406 dsl_pool_rele(dp, FTAG);
4410 if (dsl_dir_is_clone(tosnap->ds_dir))
4411 zc->zc_fromobj = tosnap->ds_dir->dd_phys->dd_origin_obj;
4412 dsl_dataset_rele(tosnap, FTAG);
4413 dsl_pool_rele(dp, FTAG);
4418 dsl_dataset_t *tosnap;
4419 dsl_dataset_t *fromsnap = NULL;
4421 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4425 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4427 dsl_pool_rele(dp, FTAG);
4431 if (zc->zc_fromobj != 0) {
4432 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
4435 dsl_dataset_rele(tosnap, FTAG);
4436 dsl_pool_rele(dp, FTAG);
4441 error = dmu_send_estimate(tosnap, fromsnap,
4442 &zc->zc_objset_type);
4444 if (fromsnap != NULL)
4445 dsl_dataset_rele(fromsnap, FTAG);
4446 dsl_dataset_rele(tosnap, FTAG);
4447 dsl_pool_rele(dp, FTAG);
4450 cap_rights_t rights;
4452 fp = getf(zc->zc_cookie,
4453 cap_rights_init(&rights, CAP_WRITE));
4455 return (SET_ERROR(EBADF));
4458 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
4459 zc->zc_fromobj, embedok, large_block_ok,
4461 zc->zc_cookie, fp->f_vnode, &off);
4463 zc->zc_cookie, fp, &off);
4466 if (off >= 0 && off <= MAXOFFSET_T)
4468 releasef(zc->zc_cookie);
4475 * zc_name name of snapshot on which to report progress
4476 * zc_cookie file descriptor of send stream
4479 * zc_cookie number of bytes written in send stream thus far
4482 zfs_ioc_send_progress(zfs_cmd_t *zc)
4486 dmu_sendarg_t *dsp = NULL;
4489 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4493 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4495 dsl_pool_rele(dp, FTAG);
4499 mutex_enter(&ds->ds_sendstream_lock);
4502 * Iterate over all the send streams currently active on this dataset.
4503 * If there's one which matches the specified file descriptor _and_ the
4504 * stream was started by the current process, return the progress of
4507 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4508 dsp = list_next(&ds->ds_sendstreams, dsp)) {
4509 if (dsp->dsa_outfd == zc->zc_cookie &&
4510 dsp->dsa_proc == curproc)
4515 zc->zc_cookie = *(dsp->dsa_off);
4517 error = SET_ERROR(ENOENT);
4519 mutex_exit(&ds->ds_sendstream_lock);
4520 dsl_dataset_rele(ds, FTAG);
4521 dsl_pool_rele(dp, FTAG);
4526 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4530 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4531 &zc->zc_inject_record);
4534 zc->zc_guid = (uint64_t)id;
4540 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4542 return (zio_clear_fault((int)zc->zc_guid));
4546 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4548 int id = (int)zc->zc_guid;
4551 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4552 &zc->zc_inject_record);
4560 zfs_ioc_error_log(zfs_cmd_t *zc)
4564 size_t count = (size_t)zc->zc_nvlist_dst_size;
4566 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4569 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4572 zc->zc_nvlist_dst_size = count;
4574 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4576 spa_close(spa, FTAG);
4582 zfs_ioc_clear(zfs_cmd_t *zc)
4589 * On zpool clear we also fix up missing slogs
4591 mutex_enter(&spa_namespace_lock);
4592 spa = spa_lookup(zc->zc_name);
4594 mutex_exit(&spa_namespace_lock);
4595 return (SET_ERROR(EIO));
4597 if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4598 /* we need to let spa_open/spa_load clear the chains */
4599 spa_set_log_state(spa, SPA_LOG_CLEAR);
4601 spa->spa_last_open_failed = 0;
4602 mutex_exit(&spa_namespace_lock);
4604 if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4605 error = spa_open(zc->zc_name, &spa, FTAG);
4608 nvlist_t *config = NULL;
4610 if (zc->zc_nvlist_src == 0)
4611 return (SET_ERROR(EINVAL));
4613 if ((error = get_nvlist(zc->zc_nvlist_src,
4614 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4615 error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4617 if (config != NULL) {
4620 if ((err = put_nvlist(zc, config)) != 0)
4622 nvlist_free(config);
4624 nvlist_free(policy);
4631 spa_vdev_state_enter(spa, SCL_NONE);
4633 if (zc->zc_guid == 0) {
4636 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4638 (void) spa_vdev_state_exit(spa, NULL, ENODEV);
4639 spa_close(spa, FTAG);
4640 return (SET_ERROR(ENODEV));
4644 vdev_clear(spa, vd);
4646 (void) spa_vdev_state_exit(spa, NULL, 0);
4649 * Resume any suspended I/Os.
4651 if (zio_resume(spa) != 0)
4652 error = SET_ERROR(EIO);
4654 spa_close(spa, FTAG);
4660 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4665 error = spa_open(zc->zc_name, &spa, FTAG);
4669 spa_vdev_state_enter(spa, SCL_NONE);
4672 * If a resilver is already in progress then set the
4673 * spa_scrub_reopen flag to B_TRUE so that we don't restart
4674 * the scan as a side effect of the reopen. Otherwise, let
4675 * vdev_open() decided if a resilver is required.
4677 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4678 vdev_reopen(spa->spa_root_vdev);
4679 spa->spa_scrub_reopen = B_FALSE;
4681 (void) spa_vdev_state_exit(spa, NULL, 0);
4682 spa_close(spa, FTAG);
4687 * zc_name name of filesystem
4688 * zc_value name of origin snapshot
4691 * zc_string name of conflicting snapshot, if there is one
4694 zfs_ioc_promote(zfs_cmd_t *zc)
4699 * We don't need to unmount *all* the origin fs's snapshots, but
4702 cp = strchr(zc->zc_value, '@');
4705 (void) dmu_objset_find(zc->zc_value,
4706 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
4707 return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4711 * Retrieve a single {user|group}{used|quota}@... property.
4714 * zc_name name of filesystem
4715 * zc_objset_type zfs_userquota_prop_t
4716 * zc_value domain name (eg. "S-1-234-567-89")
4717 * zc_guid RID/UID/GID
4720 * zc_cookie property value
4723 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4728 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4729 return (SET_ERROR(EINVAL));
4731 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4735 error = zfs_userspace_one(zfsvfs,
4736 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4737 zfsvfs_rele(zfsvfs, FTAG);
4744 * zc_name name of filesystem
4745 * zc_cookie zap cursor
4746 * zc_objset_type zfs_userquota_prop_t
4747 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4750 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
4751 * zc_cookie zap cursor
4754 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4757 int bufsize = zc->zc_nvlist_dst_size;
4760 return (SET_ERROR(ENOMEM));
4762 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4766 void *buf = kmem_alloc(bufsize, KM_SLEEP);
4768 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
4769 buf, &zc->zc_nvlist_dst_size);
4772 error = ddi_copyout(buf,
4773 (void *)(uintptr_t)zc->zc_nvlist_dst,
4774 zc->zc_nvlist_dst_size, zc->zc_iflags);
4776 kmem_free(buf, bufsize);
4777 zfsvfs_rele(zfsvfs, FTAG);
4784 * zc_name name of filesystem
4790 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4796 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
4797 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
4799 * If userused is not enabled, it may be because the
4800 * objset needs to be closed & reopened (to grow the
4801 * objset_phys_t). Suspend/resume the fs will do that.
4803 error = zfs_suspend_fs(zfsvfs);
4805 dmu_objset_refresh_ownership(zfsvfs->z_os,
4807 error = zfs_resume_fs(zfsvfs, zc->zc_name);
4811 error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
4812 VFS_RELE(zfsvfs->z_vfs);
4814 /* XXX kind of reading contents without owning */
4815 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4819 error = dmu_objset_userspace_upgrade(os);
4820 dmu_objset_rele(os, FTAG);
4828 * We don't want to have a hard dependency
4829 * against some special symbols in sharefs
4830 * nfs, and smbsrv. Determine them if needed when
4831 * the first file system is shared.
4832 * Neither sharefs, nfs or smbsrv are unloadable modules.
4834 int (*znfsexport_fs)(void *arg);
4835 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
4836 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
4838 int zfs_nfsshare_inited;
4839 int zfs_smbshare_inited;
4841 ddi_modhandle_t nfs_mod;
4842 ddi_modhandle_t sharefs_mod;
4843 ddi_modhandle_t smbsrv_mod;
4845 kmutex_t zfs_share_lock;
4853 ASSERT(MUTEX_HELD(&zfs_share_lock));
4854 /* Both NFS and SMB shares also require sharetab support. */
4855 if (sharefs_mod == NULL && ((sharefs_mod =
4856 ddi_modopen("fs/sharefs",
4857 KRTLD_MODE_FIRST, &error)) == NULL)) {
4858 return (SET_ERROR(ENOSYS));
4860 if (zshare_fs == NULL && ((zshare_fs =
4861 (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
4862 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
4863 return (SET_ERROR(ENOSYS));
4870 zfs_ioc_share(zfs_cmd_t *zc)
4876 switch (zc->zc_share.z_sharetype) {
4878 case ZFS_UNSHARE_NFS:
4879 if (zfs_nfsshare_inited == 0) {
4880 mutex_enter(&zfs_share_lock);
4881 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
4882 KRTLD_MODE_FIRST, &error)) == NULL)) {
4883 mutex_exit(&zfs_share_lock);
4884 return (SET_ERROR(ENOSYS));
4886 if (znfsexport_fs == NULL &&
4887 ((znfsexport_fs = (int (*)(void *))
4889 "nfs_export", &error)) == NULL)) {
4890 mutex_exit(&zfs_share_lock);
4891 return (SET_ERROR(ENOSYS));
4893 error = zfs_init_sharefs();
4895 mutex_exit(&zfs_share_lock);
4896 return (SET_ERROR(ENOSYS));
4898 zfs_nfsshare_inited = 1;
4899 mutex_exit(&zfs_share_lock);
4903 case ZFS_UNSHARE_SMB:
4904 if (zfs_smbshare_inited == 0) {
4905 mutex_enter(&zfs_share_lock);
4906 if (smbsrv_mod == NULL && ((smbsrv_mod =
4907 ddi_modopen("drv/smbsrv",
4908 KRTLD_MODE_FIRST, &error)) == NULL)) {
4909 mutex_exit(&zfs_share_lock);
4910 return (SET_ERROR(ENOSYS));
4912 if (zsmbexport_fs == NULL && ((zsmbexport_fs =
4913 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
4914 "smb_server_share", &error)) == NULL)) {
4915 mutex_exit(&zfs_share_lock);
4916 return (SET_ERROR(ENOSYS));
4918 error = zfs_init_sharefs();
4920 mutex_exit(&zfs_share_lock);
4921 return (SET_ERROR(ENOSYS));
4923 zfs_smbshare_inited = 1;
4924 mutex_exit(&zfs_share_lock);
4928 return (SET_ERROR(EINVAL));
4931 switch (zc->zc_share.z_sharetype) {
4933 case ZFS_UNSHARE_NFS:
4935 znfsexport_fs((void *)
4936 (uintptr_t)zc->zc_share.z_exportdata))
4940 case ZFS_UNSHARE_SMB:
4941 if (error = zsmbexport_fs((void *)
4942 (uintptr_t)zc->zc_share.z_exportdata,
4943 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
4950 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
4951 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
4952 SHAREFS_ADD : SHAREFS_REMOVE;
4955 * Add or remove share from sharetab
4957 error = zshare_fs(opcode,
4958 (void *)(uintptr_t)zc->zc_share.z_sharedata,
4959 zc->zc_share.z_sharemax);
4968 ace_t full_access[] = {
4969 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4974 * zc_name name of containing filesystem
4975 * zc_obj object # beyond which we want next in-use object #
4978 * zc_obj next in-use object #
4981 zfs_ioc_next_obj(zfs_cmd_t *zc)
4983 objset_t *os = NULL;
4986 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4990 error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
4991 os->os_dsl_dataset->ds_phys->ds_prev_snap_txg);
4993 dmu_objset_rele(os, FTAG);
4999 * zc_name name of filesystem
5000 * zc_value prefix name for snapshot
5001 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
5004 * zc_value short name of new snapshot
5007 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
5014 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
5018 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
5019 (u_longlong_t)ddi_get_lbolt64());
5020 hold_name = kmem_asprintf("%%%s", zc->zc_value);
5022 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
5025 (void) strcpy(zc->zc_value, snap_name);
5028 zfs_onexit_fd_rele(zc->zc_cleanup_fd);
5034 * zc_name name of "to" snapshot
5035 * zc_value name of "from" snapshot
5036 * zc_cookie file descriptor to write diff data on
5039 * dmu_diff_record_t's to the file descriptor
5042 zfs_ioc_diff(zfs_cmd_t *zc)
5045 cap_rights_t rights;
5049 fp = getf(zc->zc_cookie, cap_rights_init(&rights, CAP_WRITE));
5051 return (SET_ERROR(EBADF));
5056 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off);
5058 error = dmu_diff(zc->zc_name, zc->zc_value, fp, &off);
5061 if (off >= 0 && off <= MAXOFFSET_T)
5063 releasef(zc->zc_cookie);
5070 * Remove all ACL files in shares dir
5073 zfs_smb_acl_purge(znode_t *dzp)
5076 zap_attribute_t zap;
5077 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
5080 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
5081 (error = zap_cursor_retrieve(&zc, &zap)) == 0;
5082 zap_cursor_advance(&zc)) {
5083 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
5087 zap_cursor_fini(&zc);
5093 zfs_ioc_smb_acl(zfs_cmd_t *zc)
5098 vnode_t *resourcevp = NULL;
5107 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
5108 NO_FOLLOW, NULL, &vp)) != 0)
5111 /* Now make sure mntpnt and dataset are ZFS */
5113 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 ||
5114 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
5115 zc->zc_name) != 0)) {
5117 return (SET_ERROR(EINVAL));
5121 zfsvfs = dzp->z_zfsvfs;
5125 * Create share dir if its missing.
5127 mutex_enter(&zfsvfs->z_lock);
5128 if (zfsvfs->z_shares_dir == 0) {
5131 tx = dmu_tx_create(zfsvfs->z_os);
5132 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
5134 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
5135 error = dmu_tx_assign(tx, TXG_WAIT);
5139 error = zfs_create_share_dir(zfsvfs, tx);
5143 mutex_exit(&zfsvfs->z_lock);
5149 mutex_exit(&zfsvfs->z_lock);
5151 ASSERT(zfsvfs->z_shares_dir);
5152 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
5158 switch (zc->zc_cookie) {
5159 case ZFS_SMB_ACL_ADD:
5160 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
5161 vattr.va_type = VREG;
5162 vattr.va_mode = S_IFREG|0777;
5166 vsec.vsa_mask = VSA_ACE;
5167 vsec.vsa_aclentp = &full_access;
5168 vsec.vsa_aclentsz = sizeof (full_access);
5169 vsec.vsa_aclcnt = 1;
5171 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
5172 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
5174 VN_RELE(resourcevp);
5177 case ZFS_SMB_ACL_REMOVE:
5178 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
5182 case ZFS_SMB_ACL_RENAME:
5183 if ((error = get_nvlist(zc->zc_nvlist_src,
5184 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
5189 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
5190 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
5193 VN_RELE(ZTOV(sharedir));
5195 nvlist_free(nvlist);
5198 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
5200 nvlist_free(nvlist);
5203 case ZFS_SMB_ACL_PURGE:
5204 error = zfs_smb_acl_purge(sharedir);
5208 error = SET_ERROR(EINVAL);
5213 VN_RELE(ZTOV(sharedir));
5219 return (EOPNOTSUPP);
5225 * "holds" -> { snapname -> holdname (string), ... }
5226 * (optional) "cleanup_fd" -> fd (int32)
5230 * snapname -> error value (int32)
5236 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
5239 int cleanup_fd = -1;
5243 error = nvlist_lookup_nvlist(args, "holds", &holds);
5245 return (SET_ERROR(EINVAL));
5247 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
5248 error = zfs_onexit_fd_hold(cleanup_fd, &minor);
5253 error = dsl_dataset_user_hold(holds, minor, errlist);
5255 zfs_onexit_fd_rele(cleanup_fd);
5260 * innvl is not used.
5263 * holdname -> time added (uint64 seconds since epoch)
5269 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
5271 return (dsl_dataset_get_holds(snapname, outnvl));
5276 * snapname -> { holdname, ... }
5281 * snapname -> error value (int32)
5287 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
5289 return (dsl_dataset_user_release(holds, errlist));
5294 * zc_name name of new filesystem or snapshot
5295 * zc_value full name of old snapshot
5298 * zc_cookie space in bytes
5299 * zc_objset_type compressed space in bytes
5300 * zc_perm_action uncompressed space in bytes
5303 zfs_ioc_space_written(zfs_cmd_t *zc)
5307 dsl_dataset_t *new, *old;
5309 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5312 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
5314 dsl_pool_rele(dp, FTAG);
5317 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
5319 dsl_dataset_rele(new, FTAG);
5320 dsl_pool_rele(dp, FTAG);
5324 error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
5325 &zc->zc_objset_type, &zc->zc_perm_action);
5326 dsl_dataset_rele(old, FTAG);
5327 dsl_dataset_rele(new, FTAG);
5328 dsl_pool_rele(dp, FTAG);
5334 * "firstsnap" -> snapshot name
5338 * "used" -> space in bytes
5339 * "compressed" -> compressed space in bytes
5340 * "uncompressed" -> uncompressed space in bytes
5344 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
5348 dsl_dataset_t *new, *old;
5350 uint64_t used, comp, uncomp;
5352 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5353 return (SET_ERROR(EINVAL));
5355 error = dsl_pool_hold(lastsnap, FTAG, &dp);
5359 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
5361 dsl_pool_rele(dp, FTAG);
5364 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
5366 dsl_dataset_rele(new, FTAG);
5367 dsl_pool_rele(dp, FTAG);
5371 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5372 dsl_dataset_rele(old, FTAG);
5373 dsl_dataset_rele(new, FTAG);
5374 dsl_pool_rele(dp, FTAG);
5375 fnvlist_add_uint64(outnvl, "used", used);
5376 fnvlist_add_uint64(outnvl, "compressed", comp);
5377 fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5382 zfs_ioc_jail(zfs_cmd_t *zc)
5385 return (zone_dataset_attach(curthread->td_ucred, zc->zc_name,
5386 (int)zc->zc_jailid));
5390 zfs_ioc_unjail(zfs_cmd_t *zc)
5393 return (zone_dataset_detach(curthread->td_ucred, zc->zc_name,
5394 (int)zc->zc_jailid));
5399 * "fd" -> file descriptor to write stream to (int32)
5400 * (optional) "fromsnap" -> full snap name to send an incremental from
5401 * (optional) "largeblockok" -> (value ignored)
5402 * indicates that blocks > 128KB are permitted
5403 * (optional) "embedok" -> (value ignored)
5404 * presence indicates DRR_WRITE_EMBEDDED records are permitted
5411 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5413 cap_rights_t rights;
5416 char *fromname = NULL;
5418 boolean_t largeblockok;
5421 error = nvlist_lookup_int32(innvl, "fd", &fd);
5423 return (SET_ERROR(EINVAL));
5425 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
5427 largeblockok = nvlist_exists(innvl, "largeblockok");
5428 embedok = nvlist_exists(innvl, "embedok");
5430 file_t *fp = getf(fd, cap_rights_init(&rights, CAP_READ));
5432 return (SET_ERROR(EBADF));
5435 error = dmu_send(snapname, fromname, embedok, largeblockok,
5437 fd, fp->f_vnode, &off);
5443 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5454 * Determine approximately how large a zfs send stream will be -- the number
5455 * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5458 * (optional) "fromsnap" -> full snap name to send an incremental from
5462 * "space" -> bytes of space (uint64)
5466 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5469 dsl_dataset_t *fromsnap = NULL;
5470 dsl_dataset_t *tosnap;
5475 error = dsl_pool_hold(snapname, FTAG, &dp);
5479 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
5481 dsl_pool_rele(dp, FTAG);
5485 error = nvlist_lookup_string(innvl, "fromsnap", &fromname);
5487 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
5489 dsl_dataset_rele(tosnap, FTAG);
5490 dsl_pool_rele(dp, FTAG);
5495 error = dmu_send_estimate(tosnap, fromsnap, &space);
5496 fnvlist_add_uint64(outnvl, "space", space);
5498 if (fromsnap != NULL)
5499 dsl_dataset_rele(fromsnap, FTAG);
5500 dsl_dataset_rele(tosnap, FTAG);
5501 dsl_pool_rele(dp, FTAG);
5506 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
5509 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5510 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5511 boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
5513 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5515 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5516 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5517 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5518 ASSERT3P(vec->zvec_func, ==, NULL);
5520 vec->zvec_legacy_func = func;
5521 vec->zvec_secpolicy = secpolicy;
5522 vec->zvec_namecheck = namecheck;
5523 vec->zvec_allow_log = log_history;
5524 vec->zvec_pool_check = pool_check;
5528 * See the block comment at the beginning of this file for details on
5529 * each argument to this function.
5532 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
5533 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5534 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
5535 boolean_t allow_log)
5537 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5539 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5540 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5541 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5542 ASSERT3P(vec->zvec_func, ==, NULL);
5544 /* if we are logging, the name must be valid */
5545 ASSERT(!allow_log || namecheck != NO_NAME);
5547 vec->zvec_name = name;
5548 vec->zvec_func = func;
5549 vec->zvec_secpolicy = secpolicy;
5550 vec->zvec_namecheck = namecheck;
5551 vec->zvec_pool_check = pool_check;
5552 vec->zvec_smush_outnvlist = smush_outnvlist;
5553 vec->zvec_allow_log = allow_log;
5557 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5558 zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
5559 zfs_ioc_poolcheck_t pool_check)
5561 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5562 POOL_NAME, log_history, pool_check);
5566 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5567 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
5569 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5570 DATASET_NAME, B_FALSE, pool_check);
5574 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5576 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
5577 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5581 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5582 zfs_secpolicy_func_t *secpolicy)
5584 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5585 NO_NAME, B_FALSE, POOL_CHECK_NONE);
5589 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
5590 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
5592 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5593 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
5597 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5599 zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
5600 zfs_secpolicy_read);
5604 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5605 zfs_secpolicy_func_t *secpolicy)
5607 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5608 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5612 zfs_ioctl_init(void)
5614 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
5615 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
5616 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5618 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
5619 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
5620 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
5622 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
5623 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
5624 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5626 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
5627 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
5628 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5630 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
5631 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
5632 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5634 zfs_ioctl_register("create", ZFS_IOC_CREATE,
5635 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
5636 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5638 zfs_ioctl_register("clone", ZFS_IOC_CLONE,
5639 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
5640 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5642 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
5643 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
5644 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5646 zfs_ioctl_register("hold", ZFS_IOC_HOLD,
5647 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
5648 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5649 zfs_ioctl_register("release", ZFS_IOC_RELEASE,
5650 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
5651 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5653 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
5654 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
5655 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5657 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
5658 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
5659 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE);
5661 zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK,
5662 zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME,
5663 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5665 zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS,
5666 zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME,
5667 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5669 zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS,
5670 zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks,
5672 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5674 /* IOCTLS that use the legacy function signature */
5676 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
5677 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
5679 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
5680 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5681 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
5683 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
5684 zfs_ioc_pool_upgrade);
5685 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
5687 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
5688 zfs_ioc_vdev_remove);
5689 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
5690 zfs_ioc_vdev_set_state);
5691 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
5692 zfs_ioc_vdev_attach);
5693 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
5694 zfs_ioc_vdev_detach);
5695 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
5696 zfs_ioc_vdev_setpath);
5697 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
5698 zfs_ioc_vdev_setfru);
5699 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
5700 zfs_ioc_pool_set_props);
5701 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
5702 zfs_ioc_vdev_split);
5703 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
5704 zfs_ioc_pool_reguid);
5706 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
5707 zfs_ioc_pool_configs, zfs_secpolicy_none);
5708 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
5709 zfs_ioc_pool_tryimport, zfs_secpolicy_config);
5710 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
5711 zfs_ioc_inject_fault, zfs_secpolicy_inject);
5712 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
5713 zfs_ioc_clear_fault, zfs_secpolicy_inject);
5714 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
5715 zfs_ioc_inject_list_next, zfs_secpolicy_inject);
5718 * pool destroy, and export don't log the history as part of
5719 * zfsdev_ioctl, but rather zfs_ioc_pool_export
5720 * does the logging of those commands.
5722 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
5723 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5724 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
5725 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5727 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
5728 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5729 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
5730 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5732 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
5733 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_NONE);
5734 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
5735 zfs_ioc_dsobj_to_dsname,
5736 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_NONE);
5737 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
5738 zfs_ioc_pool_get_history,
5739 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
5741 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
5742 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5744 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
5745 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5746 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
5747 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5749 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
5750 zfs_ioc_space_written);
5751 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
5752 zfs_ioc_objset_recvd_props);
5753 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
5755 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
5757 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
5758 zfs_ioc_objset_stats);
5759 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
5760 zfs_ioc_objset_zplprops);
5761 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
5762 zfs_ioc_dataset_list_next);
5763 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
5764 zfs_ioc_snapshot_list_next);
5765 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
5766 zfs_ioc_send_progress);
5768 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
5769 zfs_ioc_diff, zfs_secpolicy_diff);
5770 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
5771 zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
5772 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
5773 zfs_ioc_obj_to_path, zfs_secpolicy_diff);
5774 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
5775 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
5776 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
5777 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
5778 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
5779 zfs_ioc_send, zfs_secpolicy_send);
5781 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
5782 zfs_secpolicy_none);
5783 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
5784 zfs_secpolicy_destroy);
5785 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
5786 zfs_secpolicy_rename);
5787 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
5788 zfs_secpolicy_recv);
5789 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
5790 zfs_secpolicy_promote);
5791 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
5792 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
5793 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
5794 zfs_secpolicy_set_fsacl);
5796 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
5797 zfs_secpolicy_share, POOL_CHECK_NONE);
5798 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
5799 zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
5800 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
5801 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
5802 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5803 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
5804 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
5805 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5808 zfs_ioctl_register_dataset_nolog(ZFS_IOC_JAIL, zfs_ioc_jail,
5809 zfs_secpolicy_config, POOL_CHECK_NONE);
5810 zfs_ioctl_register_dataset_nolog(ZFS_IOC_UNJAIL, zfs_ioc_unjail,
5811 zfs_secpolicy_config, POOL_CHECK_NONE);
5816 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
5817 zfs_ioc_poolcheck_t check)
5822 ASSERT(type == POOL_NAME || type == DATASET_NAME);
5824 if (check & POOL_CHECK_NONE)
5827 error = spa_open(name, &spa, FTAG);
5829 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
5830 error = SET_ERROR(EAGAIN);
5831 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
5832 error = SET_ERROR(EROFS);
5833 spa_close(spa, FTAG);
5839 * Find a free minor number.
5842 zfsdev_minor_alloc(void)
5844 static minor_t last_minor;
5847 ASSERT(MUTEX_HELD(&spa_namespace_lock));
5849 for (m = last_minor + 1; m != last_minor; m++) {
5850 if (m > ZFSDEV_MAX_MINOR)
5852 if (ddi_get_soft_state(zfsdev_state, m) == NULL) {
5862 zfs_ctldev_init(struct cdev *devp)
5865 zfs_soft_state_t *zs;
5867 ASSERT(MUTEX_HELD(&spa_namespace_lock));
5869 minor = zfsdev_minor_alloc();
5871 return (SET_ERROR(ENXIO));
5873 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS)
5874 return (SET_ERROR(EAGAIN));
5876 devfs_set_cdevpriv((void *)(uintptr_t)minor, zfsdev_close);
5878 zs = ddi_get_soft_state(zfsdev_state, minor);
5879 zs->zss_type = ZSST_CTLDEV;
5880 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data);
5886 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor)
5888 ASSERT(MUTEX_HELD(&spa_namespace_lock));
5890 zfs_onexit_destroy(zo);
5891 ddi_soft_state_free(zfsdev_state, minor);
5895 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which)
5897 zfs_soft_state_t *zp;
5899 zp = ddi_get_soft_state(zfsdev_state, minor);
5900 if (zp == NULL || zp->zss_type != which)
5903 return (zp->zss_data);
5907 zfsdev_open(struct cdev *devp, int flag, int mode, struct thread *td)
5912 if (getminor(*devp) != 0)
5913 return (zvol_open(devp, flag, otyp, cr));
5916 /* This is the control device. Allocate a new minor if requested. */
5918 mutex_enter(&spa_namespace_lock);
5919 error = zfs_ctldev_init(devp);
5920 mutex_exit(&spa_namespace_lock);
5927 zfsdev_close(void *data)
5930 minor_t minor = (minor_t)(uintptr_t)data;
5935 mutex_enter(&spa_namespace_lock);
5936 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV);
5938 mutex_exit(&spa_namespace_lock);
5941 zfs_ctldev_destroy(zo, minor);
5942 mutex_exit(&spa_namespace_lock);
5946 zfsdev_ioctl(struct cdev *dev, u_long zcmd, caddr_t arg, int flag,
5953 minor_t minor = getminor(dev);
5955 zfs_iocparm_t *zc_iocparm;
5956 int cflag, cmd, oldvecnum;
5957 boolean_t newioc, compat;
5958 void *compat_zc = NULL;
5959 cred_t *cr = td->td_ucred;
5961 const zfs_ioc_vec_t *vec;
5962 char *saved_poolname = NULL;
5963 nvlist_t *innvl = NULL;
5965 cflag = ZFS_CMD_COMPAT_NONE;
5967 newioc = B_TRUE; /* "new" style (zfs_iocparm_t) ioctl */
5969 len = IOCPARM_LEN(zcmd);
5970 vecnum = cmd = zcmd & 0xff;
5973 * Check if we are talking to supported older binaries
5974 * and translate zfs_cmd if necessary
5976 if (len != sizeof(zfs_iocparm_t)) {
5983 case sizeof(zfs_cmd_zcmd_t):
5984 cflag = ZFS_CMD_COMPAT_LZC;
5986 case sizeof(zfs_cmd_deadman_t):
5987 cflag = ZFS_CMD_COMPAT_DEADMAN;
5989 case sizeof(zfs_cmd_v28_t):
5990 cflag = ZFS_CMD_COMPAT_V28;
5992 case sizeof(zfs_cmd_v15_t):
5993 cflag = ZFS_CMD_COMPAT_V15;
5994 vecnum = zfs_ioctl_v15_to_v28[cmd];
5997 * Return without further handling
5998 * if the command is blacklisted.
6000 if (vecnum == ZFS_IOC_COMPAT_PASS)
6002 else if (vecnum == ZFS_IOC_COMPAT_FAIL)
6011 vecnum = cmd - ZFS_IOC_FIRST;
6012 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
6015 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
6016 return (SET_ERROR(EINVAL));
6017 vec = &zfs_ioc_vec[vecnum];
6019 zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP);
6022 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
6024 error = SET_ERROR(EFAULT);
6027 #else /* !illumos */
6028 bzero(zc, sizeof(zfs_cmd_t));
6031 zc_iocparm = (void *)arg;
6033 switch (zc_iocparm->zfs_ioctl_version) {
6034 case ZFS_IOCVER_CURRENT:
6035 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_t)) {
6036 error = SET_ERROR(EINVAL);
6040 case ZFS_IOCVER_ZCMD:
6041 if (zc_iocparm->zfs_cmd_size > sizeof(zfs_cmd_t) ||
6042 zc_iocparm->zfs_cmd_size < sizeof(zfs_cmd_zcmd_t)) {
6043 error = SET_ERROR(EFAULT);
6047 cflag = ZFS_CMD_COMPAT_ZCMD;
6050 error = SET_ERROR(EINVAL);
6056 ASSERT(sizeof(zfs_cmd_t) >= zc_iocparm->zfs_cmd_size);
6057 compat_zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP);
6058 bzero(compat_zc, sizeof(zfs_cmd_t));
6060 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd,
6061 compat_zc, zc_iocparm->zfs_cmd_size, flag);
6063 error = SET_ERROR(EFAULT);
6067 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd,
6068 zc, zc_iocparm->zfs_cmd_size, flag);
6070 error = SET_ERROR(EFAULT);
6078 ASSERT(compat_zc != NULL);
6079 zfs_cmd_compat_get(zc, compat_zc, cflag);
6081 ASSERT(compat_zc == NULL);
6082 zfs_cmd_compat_get(zc, arg, cflag);
6085 error = zfs_ioctl_compat_pre(zc, &vecnum, cflag);
6088 if (oldvecnum != vecnum)
6089 vec = &zfs_ioc_vec[vecnum];
6091 #endif /* !illumos */
6093 zc->zc_iflags = flag & FKIOCTL;
6094 if (zc->zc_nvlist_src_size != 0) {
6095 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
6096 zc->zc_iflags, &innvl);
6101 /* rewrite innvl for backwards compatibility */
6103 innvl = zfs_ioctl_compat_innvl(zc, innvl, vecnum, cflag);
6106 * Ensure that all pool/dataset names are valid before we pass down to
6109 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
6110 switch (vec->zvec_namecheck) {
6112 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
6113 error = SET_ERROR(EINVAL);
6115 error = pool_status_check(zc->zc_name,
6116 vec->zvec_namecheck, vec->zvec_pool_check);
6120 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
6121 error = SET_ERROR(EINVAL);
6123 error = pool_status_check(zc->zc_name,
6124 vec->zvec_namecheck, vec->zvec_pool_check);
6131 if (error == 0 && !(flag & FKIOCTL))
6132 error = vec->zvec_secpolicy(zc, innvl, cr);
6137 /* legacy ioctls can modify zc_name */
6138 len = strcspn(zc->zc_name, "/@#") + 1;
6139 saved_poolname = kmem_alloc(len, KM_SLEEP);
6140 (void) strlcpy(saved_poolname, zc->zc_name, len);
6142 if (vec->zvec_func != NULL) {
6146 nvlist_t *lognv = NULL;
6148 ASSERT(vec->zvec_legacy_func == NULL);
6151 * Add the innvl to the lognv before calling the func,
6152 * in case the func changes the innvl.
6154 if (vec->zvec_allow_log) {
6155 lognv = fnvlist_alloc();
6156 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
6158 if (!nvlist_empty(innvl)) {
6159 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
6164 outnvl = fnvlist_alloc();
6165 error = vec->zvec_func(zc->zc_name, innvl, outnvl);
6167 if (error == 0 && vec->zvec_allow_log &&
6168 spa_open(zc->zc_name, &spa, FTAG) == 0) {
6169 if (!nvlist_empty(outnvl)) {
6170 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
6173 (void) spa_history_log_nvl(spa, lognv);
6174 spa_close(spa, FTAG);
6176 fnvlist_free(lognv);
6178 /* rewrite outnvl for backwards compatibility */
6180 outnvl = zfs_ioctl_compat_outnvl(zc, outnvl, vecnum,
6183 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
6185 if (vec->zvec_smush_outnvlist) {
6186 smusherror = nvlist_smush(outnvl,
6187 zc->zc_nvlist_dst_size);
6189 if (smusherror == 0)
6190 puterror = put_nvlist(zc, outnvl);
6196 nvlist_free(outnvl);
6198 error = vec->zvec_legacy_func(zc);
6205 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
6206 if (error == 0 && rc != 0)
6207 error = SET_ERROR(EFAULT);
6210 zfs_ioctl_compat_post(zc, cmd, cflag);
6212 ASSERT(compat_zc != NULL);
6213 ASSERT(sizeof(zfs_cmd_t) >= zc_iocparm->zfs_cmd_size);
6215 zfs_cmd_compat_put(zc, compat_zc, vecnum, cflag);
6216 rc = ddi_copyout(compat_zc,
6217 (void *)(uintptr_t)zc_iocparm->zfs_cmd,
6218 zc_iocparm->zfs_cmd_size, flag);
6219 if (error == 0 && rc != 0)
6220 error = SET_ERROR(EFAULT);
6221 kmem_free(compat_zc, sizeof (zfs_cmd_t));
6223 zfs_cmd_compat_put(zc, arg, vecnum, cflag);
6228 rc = ddi_copyout(zc, (void *)(uintptr_t)zc_iocparm->zfs_cmd,
6229 sizeof (zfs_cmd_t), flag);
6230 if (error == 0 && rc != 0)
6231 error = SET_ERROR(EFAULT);
6234 if (error == 0 && vec->zvec_allow_log) {
6235 char *s = tsd_get(zfs_allow_log_key);
6238 (void) tsd_set(zfs_allow_log_key, saved_poolname);
6240 if (saved_poolname != NULL)
6241 strfree(saved_poolname);
6244 kmem_free(zc, sizeof (zfs_cmd_t));
6250 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
6252 if (cmd != DDI_ATTACH)
6253 return (DDI_FAILURE);
6255 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
6256 DDI_PSEUDO, 0) == DDI_FAILURE)
6257 return (DDI_FAILURE);
6261 ddi_report_dev(dip);
6263 return (DDI_SUCCESS);
6267 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
6269 if (spa_busy() || zfs_busy() || zvol_busy())
6270 return (DDI_FAILURE);
6272 if (cmd != DDI_DETACH)
6273 return (DDI_FAILURE);
6277 ddi_prop_remove_all(dip);
6278 ddi_remove_minor_node(dip, NULL);
6280 return (DDI_SUCCESS);
6285 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
6288 case DDI_INFO_DEVT2DEVINFO:
6290 return (DDI_SUCCESS);
6292 case DDI_INFO_DEVT2INSTANCE:
6293 *result = (void *)0;
6294 return (DDI_SUCCESS);
6297 return (DDI_FAILURE);
6302 * OK, so this is a little weird.
6304 * /dev/zfs is the control node, i.e. minor 0.
6305 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
6307 * /dev/zfs has basically nothing to do except serve up ioctls,
6308 * so most of the standard driver entry points are in zvol.c.
6311 static struct cb_ops zfs_cb_ops = {
6312 zfsdev_open, /* open */
6313 zfsdev_close, /* close */
6314 zvol_strategy, /* strategy */
6316 zvol_dump, /* dump */
6317 zvol_read, /* read */
6318 zvol_write, /* write */
6319 zfsdev_ioctl, /* ioctl */
6323 nochpoll, /* poll */
6324 ddi_prop_op, /* prop_op */
6325 NULL, /* streamtab */
6326 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */
6327 CB_REV, /* version */
6328 nodev, /* async read */
6329 nodev, /* async write */
6332 static struct dev_ops zfs_dev_ops = {
6333 DEVO_REV, /* version */
6335 zfs_info, /* info */
6336 nulldev, /* identify */
6337 nulldev, /* probe */
6338 zfs_attach, /* attach */
6339 zfs_detach, /* detach */
6341 &zfs_cb_ops, /* driver operations */
6342 NULL, /* no bus operations */
6344 ddi_quiesce_not_needed, /* quiesce */
6347 static struct modldrv zfs_modldrv = {
6353 static struct modlinkage modlinkage = {
6355 (void *)&zfs_modlfs,
6356 (void *)&zfs_modldrv,
6361 static struct cdevsw zfs_cdevsw = {
6362 .d_version = D_VERSION,
6363 .d_open = zfsdev_open,
6364 .d_ioctl = zfsdev_ioctl,
6365 .d_name = ZFS_DEV_NAME
6369 zfs_allow_log_destroy(void *arg)
6371 char *poolname = arg;
6378 zfsdev = make_dev(&zfs_cdevsw, 0x0, UID_ROOT, GID_OPERATOR, 0666,
6386 destroy_dev(zfsdev);
6389 static struct root_hold_token *zfs_root_token;
6390 struct proc *zfsproc;
6398 spa_init(FREAD | FWRITE);
6403 if ((error = mod_install(&modlinkage)) != 0) {
6410 tsd_create(&zfs_fsyncer_key, NULL);
6411 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
6412 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
6414 error = ldi_ident_from_mod(&modlinkage, &zfs_li);
6416 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
6426 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
6427 return (SET_ERROR(EBUSY));
6429 if ((error = mod_remove(&modlinkage)) != 0)
6435 if (zfs_nfsshare_inited)
6436 (void) ddi_modclose(nfs_mod);
6437 if (zfs_smbshare_inited)
6438 (void) ddi_modclose(smbsrv_mod);
6439 if (zfs_nfsshare_inited || zfs_smbshare_inited)
6440 (void) ddi_modclose(sharefs_mod);
6442 tsd_destroy(&zfs_fsyncer_key);
6443 ldi_ident_release(zfs_li);
6445 mutex_destroy(&zfs_share_lock);
6451 _info(struct modinfo *modinfop)
6453 return (mod_info(&modlinkage, modinfop));
6457 static int zfs__init(void);
6458 static int zfs__fini(void);
6459 static void zfs_shutdown(void *, int);
6461 static eventhandler_tag zfs_shutdown_event_tag;
6467 zfs_root_token = root_mount_hold("ZFS");
6469 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
6471 spa_init(FREAD | FWRITE);
6476 tsd_create(&zfs_fsyncer_key, NULL);
6477 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
6478 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
6480 printf("ZFS storage pool version: features support (" SPA_VERSION_STRING ")\n");
6481 root_mount_rel(zfs_root_token);
6491 if (spa_busy() || zfs_busy() || zvol_busy() ||
6492 zio_injection_enabled) {
6501 tsd_destroy(&zfs_fsyncer_key);
6502 tsd_destroy(&rrw_tsd_key);
6503 tsd_destroy(&zfs_allow_log_key);
6505 mutex_destroy(&zfs_share_lock);
6511 zfs_shutdown(void *arg __unused, int howto __unused)
6515 * ZFS fini routines can not properly work in a panic-ed system.
6517 if (panicstr == NULL)
6523 zfs_modevent(module_t mod, int type, void *unused __unused)
6531 zfs_shutdown_event_tag = EVENTHANDLER_REGISTER(
6532 shutdown_post_sync, zfs_shutdown, NULL,
6533 SHUTDOWN_PRI_FIRST);
6537 if (err == 0 && zfs_shutdown_event_tag != NULL)
6538 EVENTHANDLER_DEREGISTER(shutdown_post_sync,
6539 zfs_shutdown_event_tag);
6546 return (EOPNOTSUPP);
6549 static moduledata_t zfs_mod = {
6554 DECLARE_MODULE(zfsctrl, zfs_mod, SI_SUB_VFS, SI_ORDER_ANY);
6555 MODULE_VERSION(zfsctrl, 1);
6556 MODULE_DEPEND(zfsctrl, opensolaris, 1, 1, 1);
6557 MODULE_DEPEND(zfsctrl, krpc, 1, 1, 1);
6558 MODULE_DEPEND(zfsctrl, acl_nfs4, 1, 1, 1);