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 struct cdev *zfsdev;
200 extern void zfs_init(void);
201 extern void zfs_fini(void);
203 uint_t zfs_fsyncer_key;
204 extern uint_t rrw_tsd_key;
205 static uint_t zfs_allow_log_key;
207 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *);
208 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *);
209 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *);
215 } zfs_ioc_namecheck_t;
218 POOL_CHECK_NONE = 1 << 0,
219 POOL_CHECK_SUSPENDED = 1 << 1,
220 POOL_CHECK_READONLY = 1 << 2,
221 } zfs_ioc_poolcheck_t;
223 typedef struct zfs_ioc_vec {
224 zfs_ioc_legacy_func_t *zvec_legacy_func;
225 zfs_ioc_func_t *zvec_func;
226 zfs_secpolicy_func_t *zvec_secpolicy;
227 zfs_ioc_namecheck_t zvec_namecheck;
228 boolean_t zvec_allow_log;
229 zfs_ioc_poolcheck_t zvec_pool_check;
230 boolean_t zvec_smush_outnvlist;
231 const char *zvec_name;
234 /* This array is indexed by zfs_userquota_prop_t */
235 static const char *userquota_perms[] = {
236 ZFS_DELEG_PERM_USERUSED,
237 ZFS_DELEG_PERM_USERQUOTA,
238 ZFS_DELEG_PERM_GROUPUSED,
239 ZFS_DELEG_PERM_GROUPQUOTA,
242 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
243 static int zfs_check_settable(const char *name, nvpair_t *property,
245 static int zfs_check_clearable(char *dataset, nvlist_t *props,
247 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
249 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
250 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
252 static void zfsdev_close(void *data);
254 static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature);
256 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
258 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
265 * Get rid of annoying "../common/" prefix to filename.
267 newfile = strrchr(file, '/');
268 if (newfile != NULL) {
269 newfile = newfile + 1; /* Get rid of leading / */
275 (void) vsnprintf(buf, sizeof (buf), fmt, adx);
279 * To get this data, use the zfs-dprintf probe as so:
280 * dtrace -q -n 'zfs-dprintf \
281 * /stringof(arg0) == "dbuf.c"/ \
282 * {printf("%s: %s", stringof(arg1), stringof(arg3))}'
284 * arg1 = function name
288 DTRACE_PROBE4(zfs__dprintf,
289 char *, newfile, char *, func, int, line, char *, buf);
293 history_str_free(char *buf)
295 kmem_free(buf, HIS_MAX_RECORD_LEN);
299 history_str_get(zfs_cmd_t *zc)
303 if (zc->zc_history == 0)
306 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
307 if (copyinstr((void *)(uintptr_t)zc->zc_history,
308 buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
309 history_str_free(buf);
313 buf[HIS_MAX_RECORD_LEN -1] = '\0';
319 * Check to see if the named dataset is currently defined as bootable
322 zfs_is_bootfs(const char *name)
326 if (dmu_objset_hold(name, FTAG, &os) == 0) {
328 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
329 dmu_objset_rele(os, FTAG);
336 * Return non-zero if the spa version is less than requested version.
339 zfs_earlier_version(const char *name, int version)
343 if (spa_open(name, &spa, FTAG) == 0) {
344 if (spa_version(spa) < version) {
345 spa_close(spa, FTAG);
348 spa_close(spa, FTAG);
354 * Return TRUE if the ZPL version is less than requested version.
357 zpl_earlier_version(const char *name, int version)
360 boolean_t rc = B_TRUE;
362 if (dmu_objset_hold(name, FTAG, &os) == 0) {
365 if (dmu_objset_type(os) != DMU_OST_ZFS) {
366 dmu_objset_rele(os, FTAG);
369 /* XXX reading from non-owned objset */
370 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
371 rc = zplversion < version;
372 dmu_objset_rele(os, FTAG);
378 zfs_log_history(zfs_cmd_t *zc)
383 if ((buf = history_str_get(zc)) == NULL)
386 if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
387 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
388 (void) spa_history_log(spa, buf);
389 spa_close(spa, FTAG);
391 history_str_free(buf);
395 * Policy for top-level read operations (list pools). Requires no privileges,
396 * and can be used in the local zone, as there is no associated dataset.
400 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
406 * Policy for dataset read operations (list children, get statistics). Requires
407 * no privileges, but must be visible in the local zone.
411 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
413 if (INGLOBALZONE(curthread) ||
414 zone_dataset_visible(zc->zc_name, NULL))
417 return (SET_ERROR(ENOENT));
421 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
426 * The dataset must be visible by this zone -- check this first
427 * so they don't see EPERM on something they shouldn't know about.
429 if (!INGLOBALZONE(curthread) &&
430 !zone_dataset_visible(dataset, &writable))
431 return (SET_ERROR(ENOENT));
433 if (INGLOBALZONE(curthread)) {
435 * If the fs is zoned, only root can access it from the
438 if (secpolicy_zfs(cr) && zoned)
439 return (SET_ERROR(EPERM));
442 * If we are in a local zone, the 'zoned' property must be set.
445 return (SET_ERROR(EPERM));
447 /* must be writable by this zone */
449 return (SET_ERROR(EPERM));
455 zfs_dozonecheck(const char *dataset, cred_t *cr)
459 if (dsl_prop_get_integer(dataset, "jailed", &zoned, NULL))
460 return (SET_ERROR(ENOENT));
462 return (zfs_dozonecheck_impl(dataset, zoned, cr));
466 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
470 if (dsl_prop_get_int_ds(ds, "jailed", &zoned))
471 return (SET_ERROR(ENOENT));
473 return (zfs_dozonecheck_impl(dataset, zoned, cr));
477 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
478 const char *perm, cred_t *cr)
482 error = zfs_dozonecheck_ds(name, ds, cr);
484 error = secpolicy_zfs(cr);
486 error = dsl_deleg_access_impl(ds, perm, cr);
492 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
498 error = dsl_pool_hold(name, FTAG, &dp);
502 error = dsl_dataset_hold(dp, name, FTAG, &ds);
504 dsl_pool_rele(dp, FTAG);
508 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
510 dsl_dataset_rele(ds, FTAG);
511 dsl_pool_rele(dp, FTAG);
517 * Policy for setting the security label property.
519 * Returns 0 for success, non-zero for access and other errors.
522 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
524 char ds_hexsl[MAXNAMELEN];
525 bslabel_t ds_sl, new_sl;
526 boolean_t new_default = FALSE;
528 int needed_priv = -1;
531 /* First get the existing dataset label. */
532 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
533 1, sizeof (ds_hexsl), &ds_hexsl, NULL);
535 return (SET_ERROR(EPERM));
537 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
540 /* The label must be translatable */
541 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
542 return (SET_ERROR(EINVAL));
545 * In a non-global zone, disallow attempts to set a label that
546 * doesn't match that of the zone; otherwise no other checks
549 if (!INGLOBALZONE(curproc)) {
550 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
551 return (SET_ERROR(EPERM));
556 * For global-zone datasets (i.e., those whose zoned property is
557 * "off", verify that the specified new label is valid for the
560 if (dsl_prop_get_integer(name,
561 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
562 return (SET_ERROR(EPERM));
564 if (zfs_check_global_label(name, strval) != 0)
565 return (SET_ERROR(EPERM));
569 * If the existing dataset label is nondefault, check if the
570 * dataset is mounted (label cannot be changed while mounted).
571 * Get the zfsvfs; if there isn't one, then the dataset isn't
572 * mounted (or isn't a dataset, doesn't exist, ...).
574 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
576 static char *setsl_tag = "setsl_tag";
579 * Try to own the dataset; abort if there is any error,
580 * (e.g., already mounted, in use, or other error).
582 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
585 return (SET_ERROR(EPERM));
587 dmu_objset_disown(os, setsl_tag);
590 needed_priv = PRIV_FILE_DOWNGRADE_SL;
594 if (hexstr_to_label(strval, &new_sl) != 0)
595 return (SET_ERROR(EPERM));
597 if (blstrictdom(&ds_sl, &new_sl))
598 needed_priv = PRIV_FILE_DOWNGRADE_SL;
599 else if (blstrictdom(&new_sl, &ds_sl))
600 needed_priv = PRIV_FILE_UPGRADE_SL;
602 /* dataset currently has a default label */
604 needed_priv = PRIV_FILE_UPGRADE_SL;
608 if (needed_priv != -1)
609 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
612 #endif /* SECLABEL */
615 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
621 * Check permissions for special properties.
626 * Disallow setting of 'zoned' from within a local zone.
628 if (!INGLOBALZONE(curthread))
629 return (SET_ERROR(EPERM));
633 case ZFS_PROP_FILESYSTEM_LIMIT:
634 case ZFS_PROP_SNAPSHOT_LIMIT:
635 if (!INGLOBALZONE(curthread)) {
637 char setpoint[MAXNAMELEN];
639 * Unprivileged users are allowed to modify the
640 * limit on things *under* (ie. contained by)
641 * the thing they own.
643 if (dsl_prop_get_integer(dsname, "jailed", &zoned,
645 return (SET_ERROR(EPERM));
646 if (!zoned || strlen(dsname) <= strlen(setpoint))
647 return (SET_ERROR(EPERM));
651 case ZFS_PROP_MLSLABEL:
653 if (!is_system_labeled())
654 return (SET_ERROR(EPERM));
656 if (nvpair_value_string(propval, &strval) == 0) {
659 err = zfs_set_slabel_policy(dsname, strval, CRED());
669 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
674 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
678 error = zfs_dozonecheck(zc->zc_name, cr);
683 * permission to set permissions will be evaluated later in
684 * dsl_deleg_can_allow()
691 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
693 return (zfs_secpolicy_write_perms(zc->zc_name,
694 ZFS_DELEG_PERM_ROLLBACK, cr));
699 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
707 * Generate the current snapshot name from the given objsetid, then
708 * use that name for the secpolicy/zone checks.
710 cp = strchr(zc->zc_name, '@');
712 return (SET_ERROR(EINVAL));
713 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
717 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
719 dsl_pool_rele(dp, FTAG);
723 dsl_dataset_name(ds, zc->zc_name);
725 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
726 ZFS_DELEG_PERM_SEND, cr);
727 dsl_dataset_rele(ds, FTAG);
728 dsl_pool_rele(dp, FTAG);
735 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
737 return (zfs_secpolicy_write_perms(zc->zc_name,
738 ZFS_DELEG_PERM_SEND, cr));
743 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
748 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
749 NO_FOLLOW, NULL, &vp)) != 0)
752 /* Now make sure mntpnt and dataset are ZFS */
754 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 ||
755 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
756 zc->zc_name) != 0)) {
758 return (SET_ERROR(EPERM));
762 return (dsl_deleg_access(zc->zc_name,
763 ZFS_DELEG_PERM_SHARE, cr));
767 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
769 if (!INGLOBALZONE(curthread))
770 return (SET_ERROR(EPERM));
772 if (secpolicy_nfs(cr) == 0) {
775 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
780 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
782 if (!INGLOBALZONE(curthread))
783 return (SET_ERROR(EPERM));
785 if (secpolicy_smb(cr) == 0) {
788 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
793 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
798 * Remove the @bla or /bla from the end of the name to get the parent.
800 (void) strncpy(parent, datasetname, parentsize);
801 cp = strrchr(parent, '@');
805 cp = strrchr(parent, '/');
807 return (SET_ERROR(ENOENT));
815 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
819 if ((error = zfs_secpolicy_write_perms(name,
820 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
823 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
828 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
830 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
834 * Destroying snapshots with delegated permissions requires
835 * descendant mount and destroy permissions.
839 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
842 nvpair_t *pair, *nextpair;
845 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
846 return (SET_ERROR(EINVAL));
847 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
849 nextpair = nvlist_next_nvpair(snaps, pair);
850 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr);
851 if (error == ENOENT) {
853 * Ignore any snapshots that don't exist (we consider
854 * them "already destroyed"). Remove the name from the
855 * nvl here in case the snapshot is created between
856 * now and when we try to destroy it (in which case
857 * we don't want to destroy it since we haven't
858 * checked for permission).
860 fnvlist_remove_nvpair(snaps, pair);
871 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
873 char parentname[MAXNAMELEN];
876 if ((error = zfs_secpolicy_write_perms(from,
877 ZFS_DELEG_PERM_RENAME, cr)) != 0)
880 if ((error = zfs_secpolicy_write_perms(from,
881 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
884 if ((error = zfs_get_parent(to, parentname,
885 sizeof (parentname))) != 0)
888 if ((error = zfs_secpolicy_write_perms(parentname,
889 ZFS_DELEG_PERM_CREATE, cr)) != 0)
892 if ((error = zfs_secpolicy_write_perms(parentname,
893 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
901 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
906 if ((zc->zc_cookie & 1) != 0) {
908 * This is recursive rename, so the starting snapshot might
909 * not exist. Check file system or volume permission instead.
911 at = strchr(zc->zc_name, '@');
917 error = zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr);
927 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
930 dsl_dataset_t *clone;
933 error = zfs_secpolicy_write_perms(zc->zc_name,
934 ZFS_DELEG_PERM_PROMOTE, cr);
938 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
942 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
945 char parentname[MAXNAMELEN];
946 dsl_dataset_t *origin = NULL;
950 error = dsl_dataset_hold_obj(dd->dd_pool,
951 dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin);
953 dsl_dataset_rele(clone, FTAG);
954 dsl_pool_rele(dp, FTAG);
958 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
959 ZFS_DELEG_PERM_MOUNT, cr);
961 dsl_dataset_name(origin, parentname);
963 error = zfs_secpolicy_write_perms_ds(parentname, origin,
964 ZFS_DELEG_PERM_PROMOTE, cr);
966 dsl_dataset_rele(clone, FTAG);
967 dsl_dataset_rele(origin, FTAG);
969 dsl_pool_rele(dp, FTAG);
975 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
979 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
980 ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
983 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
984 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
987 return (zfs_secpolicy_write_perms(zc->zc_name,
988 ZFS_DELEG_PERM_CREATE, cr));
992 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
994 return (zfs_secpolicy_write_perms(name,
995 ZFS_DELEG_PERM_SNAPSHOT, cr));
999 * Check for permission to create each snapshot in the nvlist.
1003 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1009 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
1010 return (SET_ERROR(EINVAL));
1011 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
1012 pair = nvlist_next_nvpair(snaps, pair)) {
1013 char *name = nvpair_name(pair);
1014 char *atp = strchr(name, '@');
1017 error = SET_ERROR(EINVAL);
1021 error = zfs_secpolicy_snapshot_perms(name, cr);
1030 * Check for permission to create each snapshot in the nvlist.
1034 zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1038 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
1039 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
1040 char *name = nvpair_name(pair);
1041 char *hashp = strchr(name, '#');
1043 if (hashp == NULL) {
1044 error = SET_ERROR(EINVAL);
1048 error = zfs_secpolicy_write_perms(name,
1049 ZFS_DELEG_PERM_BOOKMARK, cr);
1059 zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1061 nvpair_t *pair, *nextpair;
1064 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1066 char *name = nvpair_name(pair);
1067 char *hashp = strchr(name, '#');
1068 nextpair = nvlist_next_nvpair(innvl, pair);
1070 if (hashp == NULL) {
1071 error = SET_ERROR(EINVAL);
1076 error = zfs_secpolicy_write_perms(name,
1077 ZFS_DELEG_PERM_DESTROY, cr);
1079 if (error == ENOENT) {
1081 * Ignore any filesystems that don't exist (we consider
1082 * their bookmarks "already destroyed"). Remove
1083 * the name from the nvl here in case the filesystem
1084 * is created between now and when we try to destroy
1085 * the bookmark (in which case we don't want to
1086 * destroy it since we haven't checked for permission).
1088 fnvlist_remove_nvpair(innvl, pair);
1100 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1103 * Even root must have a proper TSD so that we know what pool
1106 if (tsd_get(zfs_allow_log_key) == NULL)
1107 return (SET_ERROR(EPERM));
1112 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1114 char parentname[MAXNAMELEN];
1118 if ((error = zfs_get_parent(zc->zc_name, parentname,
1119 sizeof (parentname))) != 0)
1122 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1123 (error = zfs_secpolicy_write_perms(origin,
1124 ZFS_DELEG_PERM_CLONE, cr)) != 0)
1127 if ((error = zfs_secpolicy_write_perms(parentname,
1128 ZFS_DELEG_PERM_CREATE, cr)) != 0)
1131 return (zfs_secpolicy_write_perms(parentname,
1132 ZFS_DELEG_PERM_MOUNT, cr));
1136 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires
1137 * SYS_CONFIG privilege, which is not available in a local zone.
1141 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1143 if (secpolicy_sys_config(cr, B_FALSE) != 0)
1144 return (SET_ERROR(EPERM));
1150 * Policy for object to name lookups.
1154 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1158 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1161 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1166 * Policy for fault injection. Requires all privileges.
1170 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1172 return (secpolicy_zinject(cr));
1177 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1179 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1181 if (prop == ZPROP_INVAL) {
1182 if (!zfs_prop_user(zc->zc_value))
1183 return (SET_ERROR(EINVAL));
1184 return (zfs_secpolicy_write_perms(zc->zc_name,
1185 ZFS_DELEG_PERM_USERPROP, cr));
1187 return (zfs_secpolicy_setprop(zc->zc_name, prop,
1193 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1195 int err = zfs_secpolicy_read(zc, innvl, cr);
1199 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1200 return (SET_ERROR(EINVAL));
1202 if (zc->zc_value[0] == 0) {
1204 * They are asking about a posix uid/gid. If it's
1205 * themself, allow it.
1207 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1208 zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
1209 if (zc->zc_guid == crgetuid(cr))
1212 if (groupmember(zc->zc_guid, cr))
1217 return (zfs_secpolicy_write_perms(zc->zc_name,
1218 userquota_perms[zc->zc_objset_type], cr));
1222 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1224 int err = zfs_secpolicy_read(zc, innvl, cr);
1228 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1229 return (SET_ERROR(EINVAL));
1231 return (zfs_secpolicy_write_perms(zc->zc_name,
1232 userquota_perms[zc->zc_objset_type], cr));
1237 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1239 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1245 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1251 error = nvlist_lookup_nvlist(innvl, "holds", &holds);
1253 return (SET_ERROR(EINVAL));
1255 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
1256 pair = nvlist_next_nvpair(holds, pair)) {
1257 char fsname[MAXNAMELEN];
1258 error = dmu_fsname(nvpair_name(pair), fsname);
1261 error = zfs_secpolicy_write_perms(fsname,
1262 ZFS_DELEG_PERM_HOLD, cr);
1271 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1276 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1277 pair = nvlist_next_nvpair(innvl, pair)) {
1278 char fsname[MAXNAMELEN];
1279 error = dmu_fsname(nvpair_name(pair), fsname);
1282 error = zfs_secpolicy_write_perms(fsname,
1283 ZFS_DELEG_PERM_RELEASE, cr);
1291 * Policy for allowing temporary snapshots to be taken or released
1294 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1297 * A temporary snapshot is the same as a snapshot,
1298 * hold, destroy and release all rolled into one.
1299 * Delegated diff alone is sufficient that we allow this.
1303 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1304 ZFS_DELEG_PERM_DIFF, cr)) == 0)
1307 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1309 error = zfs_secpolicy_hold(zc, innvl, cr);
1311 error = zfs_secpolicy_release(zc, innvl, cr);
1313 error = zfs_secpolicy_destroy(zc, innvl, cr);
1318 * Returns the nvlist as specified by the user in the zfs_cmd_t.
1321 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1325 nvlist_t *list = NULL;
1328 * Read in and unpack the user-supplied nvlist.
1331 return (SET_ERROR(EINVAL));
1333 packed = kmem_alloc(size, KM_SLEEP);
1335 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1337 kmem_free(packed, size);
1341 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1342 kmem_free(packed, size);
1346 kmem_free(packed, size);
1353 * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1354 * Entries will be removed from the end of the nvlist, and one int32 entry
1355 * named "N_MORE_ERRORS" will be added indicating how many entries were
1359 nvlist_smush(nvlist_t *errors, size_t max)
1363 size = fnvlist_size(errors);
1366 nvpair_t *more_errors;
1370 return (SET_ERROR(ENOMEM));
1372 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1373 more_errors = nvlist_prev_nvpair(errors, NULL);
1376 nvpair_t *pair = nvlist_prev_nvpair(errors,
1378 fnvlist_remove_nvpair(errors, pair);
1380 size = fnvlist_size(errors);
1381 } while (size > max);
1383 fnvlist_remove_nvpair(errors, more_errors);
1384 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1385 ASSERT3U(fnvlist_size(errors), <=, max);
1392 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1394 char *packed = NULL;
1398 size = fnvlist_size(nvl);
1400 if (size > zc->zc_nvlist_dst_size) {
1402 * Solaris returns ENOMEM here, because even if an error is
1403 * returned from an ioctl(2), new zc_nvlist_dst_size will be
1404 * passed to the userland. This is not the case for FreeBSD.
1405 * We need to return 0, so the kernel will copy the
1406 * zc_nvlist_dst_size back and the userland can discover that a
1407 * bigger buffer is needed.
1411 packed = fnvlist_pack(nvl, &size);
1412 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1413 size, zc->zc_iflags) != 0)
1414 error = SET_ERROR(EFAULT);
1415 fnvlist_pack_free(packed, size);
1418 zc->zc_nvlist_dst_size = size;
1419 zc->zc_nvlist_dst_filled = B_TRUE;
1424 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1429 error = dmu_objset_hold(dsname, FTAG, &os);
1432 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1433 dmu_objset_rele(os, FTAG);
1434 return (SET_ERROR(EINVAL));
1437 mutex_enter(&os->os_user_ptr_lock);
1438 *zfvp = dmu_objset_get_user(os);
1440 VFS_HOLD((*zfvp)->z_vfs);
1442 error = SET_ERROR(ESRCH);
1444 mutex_exit(&os->os_user_ptr_lock);
1445 dmu_objset_rele(os, FTAG);
1450 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1451 * case its z_vfs will be NULL, and it will be opened as the owner.
1452 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1453 * which prevents all vnode ops from running.
1456 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1460 if (getzfsvfs(name, zfvp) != 0)
1461 error = zfsvfs_create(name, zfvp);
1463 rrm_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER :
1465 if ((*zfvp)->z_unmounted) {
1467 * XXX we could probably try again, since the unmounting
1468 * thread should be just about to disassociate the
1469 * objset from the zfsvfs.
1471 rrm_exit(&(*zfvp)->z_teardown_lock, tag);
1472 return (SET_ERROR(EBUSY));
1479 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1481 rrm_exit(&zfsvfs->z_teardown_lock, tag);
1483 if (zfsvfs->z_vfs) {
1484 VFS_RELE(zfsvfs->z_vfs);
1486 dmu_objset_disown(zfsvfs->z_os, zfsvfs);
1487 zfsvfs_free(zfsvfs);
1492 zfs_ioc_pool_create(zfs_cmd_t *zc)
1495 nvlist_t *config, *props = NULL;
1496 nvlist_t *rootprops = NULL;
1497 nvlist_t *zplprops = NULL;
1499 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1500 zc->zc_iflags, &config))
1503 if (zc->zc_nvlist_src_size != 0 && (error =
1504 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1505 zc->zc_iflags, &props))) {
1506 nvlist_free(config);
1511 nvlist_t *nvl = NULL;
1512 uint64_t version = SPA_VERSION;
1514 (void) nvlist_lookup_uint64(props,
1515 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1516 if (!SPA_VERSION_IS_SUPPORTED(version)) {
1517 error = SET_ERROR(EINVAL);
1518 goto pool_props_bad;
1520 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1522 error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1524 nvlist_free(config);
1528 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1530 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1531 error = zfs_fill_zplprops_root(version, rootprops,
1534 goto pool_props_bad;
1537 error = spa_create(zc->zc_name, config, props, zplprops);
1540 * Set the remaining root properties
1542 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1543 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1544 (void) spa_destroy(zc->zc_name);
1547 nvlist_free(rootprops);
1548 nvlist_free(zplprops);
1549 nvlist_free(config);
1556 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1559 zfs_log_history(zc);
1560 error = spa_destroy(zc->zc_name);
1562 zvol_remove_minors(zc->zc_name);
1567 zfs_ioc_pool_import(zfs_cmd_t *zc)
1569 nvlist_t *config, *props = NULL;
1573 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1574 zc->zc_iflags, &config)) != 0)
1577 if (zc->zc_nvlist_src_size != 0 && (error =
1578 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1579 zc->zc_iflags, &props))) {
1580 nvlist_free(config);
1584 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1585 guid != zc->zc_guid)
1586 error = SET_ERROR(EINVAL);
1588 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1590 if (zc->zc_nvlist_dst != 0) {
1593 if ((err = put_nvlist(zc, config)) != 0)
1597 nvlist_free(config);
1606 zfs_ioc_pool_export(zfs_cmd_t *zc)
1609 boolean_t force = (boolean_t)zc->zc_cookie;
1610 boolean_t hardforce = (boolean_t)zc->zc_guid;
1612 zfs_log_history(zc);
1613 error = spa_export(zc->zc_name, NULL, force, hardforce);
1615 zvol_remove_minors(zc->zc_name);
1620 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1625 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1626 return (SET_ERROR(EEXIST));
1628 error = put_nvlist(zc, configs);
1630 nvlist_free(configs);
1637 * zc_name name of the pool
1640 * zc_cookie real errno
1641 * zc_nvlist_dst config nvlist
1642 * zc_nvlist_dst_size size of config nvlist
1645 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1651 error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1652 sizeof (zc->zc_value));
1654 if (config != NULL) {
1655 ret = put_nvlist(zc, config);
1656 nvlist_free(config);
1659 * The config may be present even if 'error' is non-zero.
1660 * In this case we return success, and preserve the real errno
1663 zc->zc_cookie = error;
1672 * Try to import the given pool, returning pool stats as appropriate so that
1673 * user land knows which devices are available and overall pool health.
1676 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1678 nvlist_t *tryconfig, *config;
1681 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1682 zc->zc_iflags, &tryconfig)) != 0)
1685 config = spa_tryimport(tryconfig);
1687 nvlist_free(tryconfig);
1690 return (SET_ERROR(EINVAL));
1692 error = put_nvlist(zc, config);
1693 nvlist_free(config);
1700 * zc_name name of the pool
1701 * zc_cookie scan func (pool_scan_func_t)
1704 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1709 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1712 if (zc->zc_cookie == POOL_SCAN_NONE)
1713 error = spa_scan_stop(spa);
1715 error = spa_scan(spa, zc->zc_cookie);
1717 spa_close(spa, FTAG);
1723 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1728 error = spa_open(zc->zc_name, &spa, FTAG);
1731 spa_close(spa, FTAG);
1737 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1742 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1745 if (zc->zc_cookie < spa_version(spa) ||
1746 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1747 spa_close(spa, FTAG);
1748 return (SET_ERROR(EINVAL));
1751 spa_upgrade(spa, zc->zc_cookie);
1752 spa_close(spa, FTAG);
1758 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1765 if ((size = zc->zc_history_len) == 0)
1766 return (SET_ERROR(EINVAL));
1768 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1771 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1772 spa_close(spa, FTAG);
1773 return (SET_ERROR(ENOTSUP));
1776 hist_buf = kmem_alloc(size, KM_SLEEP);
1777 if ((error = spa_history_get(spa, &zc->zc_history_offset,
1778 &zc->zc_history_len, hist_buf)) == 0) {
1779 error = ddi_copyout(hist_buf,
1780 (void *)(uintptr_t)zc->zc_history,
1781 zc->zc_history_len, zc->zc_iflags);
1784 spa_close(spa, FTAG);
1785 kmem_free(hist_buf, size);
1790 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1795 error = spa_open(zc->zc_name, &spa, FTAG);
1797 error = spa_change_guid(spa);
1798 spa_close(spa, FTAG);
1804 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1806 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1811 * zc_name name of filesystem
1812 * zc_obj object to find
1815 * zc_value name of object
1818 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1823 /* XXX reading from objset not owned */
1824 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1826 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1827 dmu_objset_rele(os, FTAG);
1828 return (SET_ERROR(EINVAL));
1830 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1831 sizeof (zc->zc_value));
1832 dmu_objset_rele(os, FTAG);
1839 * zc_name name of filesystem
1840 * zc_obj object to find
1843 * zc_stat stats on object
1844 * zc_value path to object
1847 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1852 /* XXX reading from objset not owned */
1853 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1855 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1856 dmu_objset_rele(os, FTAG);
1857 return (SET_ERROR(EINVAL));
1859 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1860 sizeof (zc->zc_value));
1861 dmu_objset_rele(os, FTAG);
1867 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1871 nvlist_t *config, **l2cache, **spares;
1872 uint_t nl2cache = 0, nspares = 0;
1874 error = spa_open(zc->zc_name, &spa, FTAG);
1878 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1879 zc->zc_iflags, &config);
1880 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1881 &l2cache, &nl2cache);
1883 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1888 * A root pool with concatenated devices is not supported.
1889 * Thus, can not add a device to a root pool.
1891 * Intent log device can not be added to a rootpool because
1892 * during mountroot, zil is replayed, a seperated log device
1893 * can not be accessed during the mountroot time.
1895 * l2cache and spare devices are ok to be added to a rootpool.
1897 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1898 nvlist_free(config);
1899 spa_close(spa, FTAG);
1900 return (SET_ERROR(EDOM));
1902 #endif /* illumos */
1905 error = spa_vdev_add(spa, config);
1906 nvlist_free(config);
1908 spa_close(spa, FTAG);
1914 * zc_name name of the pool
1915 * zc_nvlist_conf nvlist of devices to remove
1916 * zc_cookie to stop the remove?
1919 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1924 error = spa_open(zc->zc_name, &spa, FTAG);
1927 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1928 spa_close(spa, FTAG);
1933 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1937 vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1939 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1941 switch (zc->zc_cookie) {
1942 case VDEV_STATE_ONLINE:
1943 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1946 case VDEV_STATE_OFFLINE:
1947 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1950 case VDEV_STATE_FAULTED:
1951 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1952 zc->zc_obj != VDEV_AUX_EXTERNAL)
1953 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1955 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1958 case VDEV_STATE_DEGRADED:
1959 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1960 zc->zc_obj != VDEV_AUX_EXTERNAL)
1961 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1963 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1967 error = SET_ERROR(EINVAL);
1969 zc->zc_cookie = newstate;
1970 spa_close(spa, FTAG);
1975 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1978 int replacing = zc->zc_cookie;
1982 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1985 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1986 zc->zc_iflags, &config)) == 0) {
1987 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1988 nvlist_free(config);
1991 spa_close(spa, FTAG);
1996 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
2001 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2004 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
2006 spa_close(spa, FTAG);
2011 zfs_ioc_vdev_split(zfs_cmd_t *zc)
2014 nvlist_t *config, *props = NULL;
2016 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
2018 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2021 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
2022 zc->zc_iflags, &config)) {
2023 spa_close(spa, FTAG);
2027 if (zc->zc_nvlist_src_size != 0 && (error =
2028 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2029 zc->zc_iflags, &props))) {
2030 spa_close(spa, FTAG);
2031 nvlist_free(config);
2035 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
2037 spa_close(spa, FTAG);
2039 nvlist_free(config);
2046 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
2049 char *path = zc->zc_value;
2050 uint64_t guid = zc->zc_guid;
2053 error = spa_open(zc->zc_name, &spa, FTAG);
2057 error = spa_vdev_setpath(spa, guid, path);
2058 spa_close(spa, FTAG);
2063 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
2066 char *fru = zc->zc_value;
2067 uint64_t guid = zc->zc_guid;
2070 error = spa_open(zc->zc_name, &spa, FTAG);
2074 error = spa_vdev_setfru(spa, guid, fru);
2075 spa_close(spa, FTAG);
2080 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
2085 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2087 if (zc->zc_nvlist_dst != 0 &&
2088 (error = dsl_prop_get_all(os, &nv)) == 0) {
2089 dmu_objset_stats(os, nv);
2091 * NB: zvol_get_stats() will read the objset contents,
2092 * which we aren't supposed to do with a
2093 * DS_MODE_USER hold, because it could be
2094 * inconsistent. So this is a bit of a workaround...
2095 * XXX reading with out owning
2097 if (!zc->zc_objset_stats.dds_inconsistent &&
2098 dmu_objset_type(os) == DMU_OST_ZVOL) {
2099 error = zvol_get_stats(os, nv);
2104 error = put_nvlist(zc, nv);
2113 * zc_name name of filesystem
2114 * zc_nvlist_dst_size size of buffer for property nvlist
2117 * zc_objset_stats stats
2118 * zc_nvlist_dst property nvlist
2119 * zc_nvlist_dst_size size of property nvlist
2122 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2127 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2129 error = zfs_ioc_objset_stats_impl(zc, os);
2130 dmu_objset_rele(os, FTAG);
2133 if (error == ENOMEM)
2140 * zc_name name of filesystem
2141 * zc_nvlist_dst_size size of buffer for property nvlist
2144 * zc_nvlist_dst received property nvlist
2145 * zc_nvlist_dst_size size of received property nvlist
2147 * Gets received properties (distinct from local properties on or after
2148 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2149 * local property values.
2152 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2158 * Without this check, we would return local property values if the
2159 * caller has not already received properties on or after
2160 * SPA_VERSION_RECVD_PROPS.
2162 if (!dsl_prop_get_hasrecvd(zc->zc_name))
2163 return (SET_ERROR(ENOTSUP));
2165 if (zc->zc_nvlist_dst != 0 &&
2166 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2167 error = put_nvlist(zc, nv);
2175 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2181 * zfs_get_zplprop() will either find a value or give us
2182 * the default value (if there is one).
2184 if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2186 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2192 * zc_name name of filesystem
2193 * zc_nvlist_dst_size size of buffer for zpl property nvlist
2196 * zc_nvlist_dst zpl property nvlist
2197 * zc_nvlist_dst_size size of zpl property nvlist
2200 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2205 /* XXX reading without owning */
2206 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os))
2209 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2212 * NB: nvl_add_zplprop() will read the objset contents,
2213 * which we aren't supposed to do with a DS_MODE_USER
2214 * hold, because it could be inconsistent.
2216 if (zc->zc_nvlist_dst != 0 &&
2217 !zc->zc_objset_stats.dds_inconsistent &&
2218 dmu_objset_type(os) == DMU_OST_ZFS) {
2221 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2222 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2223 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2224 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2225 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2226 err = put_nvlist(zc, nv);
2229 err = SET_ERROR(ENOENT);
2231 dmu_objset_rele(os, FTAG);
2236 dataset_name_hidden(const char *name)
2239 * Skip over datasets that are not visible in this zone,
2240 * internal datasets (which have a $ in their name), and
2241 * temporary datasets (which have a % in their name).
2243 if (strchr(name, '$') != NULL)
2245 if (strchr(name, '%') != NULL)
2247 if (!INGLOBALZONE(curthread) && !zone_dataset_visible(name, NULL))
2254 * zc_name name of filesystem
2255 * zc_cookie zap cursor
2256 * zc_nvlist_dst_size size of buffer for property nvlist
2259 * zc_name name of next filesystem
2260 * zc_cookie zap cursor
2261 * zc_objset_stats stats
2262 * zc_nvlist_dst property nvlist
2263 * zc_nvlist_dst_size size of property nvlist
2266 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2271 size_t orig_len = strlen(zc->zc_name);
2274 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) {
2275 if (error == ENOENT)
2276 error = SET_ERROR(ESRCH);
2280 p = strrchr(zc->zc_name, '/');
2281 if (p == NULL || p[1] != '\0')
2282 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2283 p = zc->zc_name + strlen(zc->zc_name);
2286 error = dmu_dir_list_next(os,
2287 sizeof (zc->zc_name) - (p - zc->zc_name), p,
2288 NULL, &zc->zc_cookie);
2289 if (error == ENOENT)
2290 error = SET_ERROR(ESRCH);
2291 } while (error == 0 && dataset_name_hidden(zc->zc_name));
2292 dmu_objset_rele(os, FTAG);
2295 * If it's an internal dataset (ie. with a '$' in its name),
2296 * don't try to get stats for it, otherwise we'll return ENOENT.
2298 if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2299 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2300 if (error == ENOENT) {
2301 /* We lost a race with destroy, get the next one. */
2302 zc->zc_name[orig_len] = '\0';
2311 * zc_name name of filesystem
2312 * zc_cookie zap cursor
2313 * zc_nvlist_dst_size size of buffer for property nvlist
2314 * zc_simple when set, only name is requested
2317 * zc_name name of next snapshot
2318 * zc_objset_stats stats
2319 * zc_nvlist_dst property nvlist
2320 * zc_nvlist_dst_size size of property nvlist
2323 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2328 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2330 return (error == ENOENT ? ESRCH : error);
2334 * A dataset name of maximum length cannot have any snapshots,
2335 * so exit immediately.
2337 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
2338 dmu_objset_rele(os, FTAG);
2339 return (SET_ERROR(ESRCH));
2342 error = dmu_snapshot_list_next(os,
2343 sizeof (zc->zc_name) - strlen(zc->zc_name),
2344 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2347 if (error == 0 && !zc->zc_simple) {
2349 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2351 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2355 error = dmu_objset_from_ds(ds, &ossnap);
2357 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2358 dsl_dataset_rele(ds, FTAG);
2360 } else if (error == ENOENT) {
2361 error = SET_ERROR(ESRCH);
2364 dmu_objset_rele(os, FTAG);
2365 /* if we failed, undo the @ that we tacked on to zc_name */
2367 *strchr(zc->zc_name, '@') = '\0';
2372 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2374 const char *propname = nvpair_name(pair);
2376 unsigned int vallen;
2379 zfs_userquota_prop_t type;
2385 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2387 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2388 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2390 return (SET_ERROR(EINVAL));
2394 * A correctly constructed propname is encoded as
2395 * userquota@<rid>-<domain>.
2397 if ((dash = strchr(propname, '-')) == NULL ||
2398 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2400 return (SET_ERROR(EINVAL));
2407 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2409 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2410 zfsvfs_rele(zfsvfs, FTAG);
2417 * If the named property is one that has a special function to set its value,
2418 * return 0 on success and a positive error code on failure; otherwise if it is
2419 * not one of the special properties handled by this function, return -1.
2421 * XXX: It would be better for callers of the property interface if we handled
2422 * these special cases in dsl_prop.c (in the dsl layer).
2425 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2428 const char *propname = nvpair_name(pair);
2429 zfs_prop_t prop = zfs_name_to_prop(propname);
2433 if (prop == ZPROP_INVAL) {
2434 if (zfs_prop_userquota(propname))
2435 return (zfs_prop_set_userquota(dsname, pair));
2439 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2441 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2442 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2446 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2449 VERIFY(0 == nvpair_value_uint64(pair, &intval));
2452 case ZFS_PROP_QUOTA:
2453 err = dsl_dir_set_quota(dsname, source, intval);
2455 case ZFS_PROP_REFQUOTA:
2456 err = dsl_dataset_set_refquota(dsname, source, intval);
2458 case ZFS_PROP_FILESYSTEM_LIMIT:
2459 case ZFS_PROP_SNAPSHOT_LIMIT:
2460 if (intval == UINT64_MAX) {
2461 /* clearing the limit, just do it */
2464 err = dsl_dir_activate_fs_ss_limit(dsname);
2467 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2468 * default path to set the value in the nvlist.
2473 case ZFS_PROP_RESERVATION:
2474 err = dsl_dir_set_reservation(dsname, source, intval);
2476 case ZFS_PROP_REFRESERVATION:
2477 err = dsl_dataset_set_refreservation(dsname, source, intval);
2479 case ZFS_PROP_VOLSIZE:
2480 err = zvol_set_volsize(dsname, intval);
2482 case ZFS_PROP_VERSION:
2486 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2489 err = zfs_set_version(zfsvfs, intval);
2490 zfsvfs_rele(zfsvfs, FTAG);
2492 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2495 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2496 (void) strcpy(zc->zc_name, dsname);
2497 (void) zfs_ioc_userspace_upgrade(zc);
2498 kmem_free(zc, sizeof (zfs_cmd_t));
2510 * This function is best effort. If it fails to set any of the given properties,
2511 * it continues to set as many as it can and returns the last error
2512 * encountered. If the caller provides a non-NULL errlist, it will be filled in
2513 * with the list of names of all the properties that failed along with the
2514 * corresponding error numbers.
2516 * If every property is set successfully, zero is returned and errlist is not
2520 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2528 nvlist_t *genericnvl = fnvlist_alloc();
2529 nvlist_t *retrynvl = fnvlist_alloc();
2533 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2534 const char *propname = nvpair_name(pair);
2535 zfs_prop_t prop = zfs_name_to_prop(propname);
2538 /* decode the property value */
2540 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2542 attrs = fnvpair_value_nvlist(pair);
2543 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2545 err = SET_ERROR(EINVAL);
2548 /* Validate value type */
2549 if (err == 0 && prop == ZPROP_INVAL) {
2550 if (zfs_prop_user(propname)) {
2551 if (nvpair_type(propval) != DATA_TYPE_STRING)
2552 err = SET_ERROR(EINVAL);
2553 } else if (zfs_prop_userquota(propname)) {
2554 if (nvpair_type(propval) !=
2555 DATA_TYPE_UINT64_ARRAY)
2556 err = SET_ERROR(EINVAL);
2558 err = SET_ERROR(EINVAL);
2560 } else if (err == 0) {
2561 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2562 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2563 err = SET_ERROR(EINVAL);
2564 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2567 intval = fnvpair_value_uint64(propval);
2569 switch (zfs_prop_get_type(prop)) {
2570 case PROP_TYPE_NUMBER:
2572 case PROP_TYPE_STRING:
2573 err = SET_ERROR(EINVAL);
2575 case PROP_TYPE_INDEX:
2576 if (zfs_prop_index_to_string(prop,
2577 intval, &unused) != 0)
2578 err = SET_ERROR(EINVAL);
2582 "unknown property type");
2585 err = SET_ERROR(EINVAL);
2589 /* Validate permissions */
2591 err = zfs_check_settable(dsname, pair, CRED());
2594 err = zfs_prop_set_special(dsname, source, pair);
2597 * For better performance we build up a list of
2598 * properties to set in a single transaction.
2600 err = nvlist_add_nvpair(genericnvl, pair);
2601 } else if (err != 0 && nvl != retrynvl) {
2603 * This may be a spurious error caused by
2604 * receiving quota and reservation out of order.
2605 * Try again in a second pass.
2607 err = nvlist_add_nvpair(retrynvl, pair);
2612 if (errlist != NULL)
2613 fnvlist_add_int32(errlist, propname, err);
2618 if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2623 if (!nvlist_empty(genericnvl) &&
2624 dsl_props_set(dsname, source, genericnvl) != 0) {
2626 * If this fails, we still want to set as many properties as we
2627 * can, so try setting them individually.
2630 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2631 const char *propname = nvpair_name(pair);
2635 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2637 attrs = fnvpair_value_nvlist(pair);
2638 propval = fnvlist_lookup_nvpair(attrs,
2642 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2643 strval = fnvpair_value_string(propval);
2644 err = dsl_prop_set_string(dsname, propname,
2647 intval = fnvpair_value_uint64(propval);
2648 err = dsl_prop_set_int(dsname, propname, source,
2653 if (errlist != NULL) {
2654 fnvlist_add_int32(errlist, propname,
2661 nvlist_free(genericnvl);
2662 nvlist_free(retrynvl);
2668 * Check that all the properties are valid user properties.
2671 zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2673 nvpair_t *pair = NULL;
2676 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2677 const char *propname = nvpair_name(pair);
2679 if (!zfs_prop_user(propname) ||
2680 nvpair_type(pair) != DATA_TYPE_STRING)
2681 return (SET_ERROR(EINVAL));
2683 if (error = zfs_secpolicy_write_perms(fsname,
2684 ZFS_DELEG_PERM_USERPROP, CRED()))
2687 if (strlen(propname) >= ZAP_MAXNAMELEN)
2688 return (SET_ERROR(ENAMETOOLONG));
2690 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN)
2697 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2701 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2704 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2705 if (nvlist_exists(skipped, nvpair_name(pair)))
2708 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2713 clear_received_props(const char *dsname, nvlist_t *props,
2717 nvlist_t *cleared_props = NULL;
2718 props_skip(props, skipped, &cleared_props);
2719 if (!nvlist_empty(cleared_props)) {
2721 * Acts on local properties until the dataset has received
2722 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2724 zprop_source_t flags = (ZPROP_SRC_NONE |
2725 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2726 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2728 nvlist_free(cleared_props);
2734 * zc_name name of filesystem
2735 * zc_value name of property to set
2736 * zc_nvlist_src{_size} nvlist of properties to apply
2737 * zc_cookie received properties flag
2740 * zc_nvlist_dst{_size} error for each unapplied received property
2743 zfs_ioc_set_prop(zfs_cmd_t *zc)
2746 boolean_t received = zc->zc_cookie;
2747 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2752 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2753 zc->zc_iflags, &nvl)) != 0)
2757 nvlist_t *origprops;
2759 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2760 (void) clear_received_props(zc->zc_name,
2762 nvlist_free(origprops);
2765 error = dsl_prop_set_hasrecvd(zc->zc_name);
2768 errors = fnvlist_alloc();
2770 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2772 if (zc->zc_nvlist_dst != 0 && errors != NULL) {
2773 (void) put_nvlist(zc, errors);
2776 nvlist_free(errors);
2783 * zc_name name of filesystem
2784 * zc_value name of property to inherit
2785 * zc_cookie revert to received value if TRUE
2790 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2792 const char *propname = zc->zc_value;
2793 zfs_prop_t prop = zfs_name_to_prop(propname);
2794 boolean_t received = zc->zc_cookie;
2795 zprop_source_t source = (received
2796 ? ZPROP_SRC_NONE /* revert to received value, if any */
2797 : ZPROP_SRC_INHERITED); /* explicitly inherit */
2806 * zfs_prop_set_special() expects properties in the form of an
2807 * nvpair with type info.
2809 if (prop == ZPROP_INVAL) {
2810 if (!zfs_prop_user(propname))
2811 return (SET_ERROR(EINVAL));
2813 type = PROP_TYPE_STRING;
2814 } else if (prop == ZFS_PROP_VOLSIZE ||
2815 prop == ZFS_PROP_VERSION) {
2816 return (SET_ERROR(EINVAL));
2818 type = zfs_prop_get_type(prop);
2821 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2824 case PROP_TYPE_STRING:
2825 VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2827 case PROP_TYPE_NUMBER:
2828 case PROP_TYPE_INDEX:
2829 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2833 return (SET_ERROR(EINVAL));
2836 pair = nvlist_next_nvpair(dummy, NULL);
2837 err = zfs_prop_set_special(zc->zc_name, source, pair);
2840 return (err); /* special property already handled */
2843 * Only check this in the non-received case. We want to allow
2844 * 'inherit -S' to revert non-inheritable properties like quota
2845 * and reservation to the received or default values even though
2846 * they are not considered inheritable.
2848 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2849 return (SET_ERROR(EINVAL));
2852 /* property name has been validated by zfs_secpolicy_inherit_prop() */
2853 return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source));
2857 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2864 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2865 zc->zc_iflags, &props))
2869 * If the only property is the configfile, then just do a spa_lookup()
2870 * to handle the faulted case.
2872 pair = nvlist_next_nvpair(props, NULL);
2873 if (pair != NULL && strcmp(nvpair_name(pair),
2874 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2875 nvlist_next_nvpair(props, pair) == NULL) {
2876 mutex_enter(&spa_namespace_lock);
2877 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2878 spa_configfile_set(spa, props, B_FALSE);
2879 spa_config_sync(spa, B_FALSE, B_TRUE);
2881 mutex_exit(&spa_namespace_lock);
2888 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2893 error = spa_prop_set(spa, props);
2896 spa_close(spa, FTAG);
2902 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2906 nvlist_t *nvp = NULL;
2908 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2910 * If the pool is faulted, there may be properties we can still
2911 * get (such as altroot and cachefile), so attempt to get them
2914 mutex_enter(&spa_namespace_lock);
2915 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2916 error = spa_prop_get(spa, &nvp);
2917 mutex_exit(&spa_namespace_lock);
2919 error = spa_prop_get(spa, &nvp);
2920 spa_close(spa, FTAG);
2923 if (error == 0 && zc->zc_nvlist_dst != 0)
2924 error = put_nvlist(zc, nvp);
2926 error = SET_ERROR(EFAULT);
2934 * zc_name name of filesystem
2935 * zc_nvlist_src{_size} nvlist of delegated permissions
2936 * zc_perm_action allow/unallow flag
2941 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2944 nvlist_t *fsaclnv = NULL;
2946 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2947 zc->zc_iflags, &fsaclnv)) != 0)
2951 * Verify nvlist is constructed correctly
2953 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2954 nvlist_free(fsaclnv);
2955 return (SET_ERROR(EINVAL));
2959 * If we don't have PRIV_SYS_MOUNT, then validate
2960 * that user is allowed to hand out each permission in
2964 error = secpolicy_zfs(CRED());
2966 if (zc->zc_perm_action == B_FALSE) {
2967 error = dsl_deleg_can_allow(zc->zc_name,
2970 error = dsl_deleg_can_unallow(zc->zc_name,
2976 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2978 nvlist_free(fsaclnv);
2984 * zc_name name of filesystem
2987 * zc_nvlist_src{_size} nvlist of delegated permissions
2990 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
2995 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
2996 error = put_nvlist(zc, nvp);
3004 * Search the vfs list for a specified resource. Returns a pointer to it
3005 * or NULL if no suitable entry is found. The caller of this routine
3006 * is responsible for releasing the returned vfs pointer.
3009 zfs_get_vfs(const char *resource)
3013 mtx_lock(&mountlist_mtx);
3014 TAILQ_FOREACH(vfsp, &mountlist, mnt_list) {
3015 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
3020 mtx_unlock(&mountlist_mtx);
3026 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
3028 zfs_creat_t *zct = arg;
3030 zfs_create_fs(os, cr, zct->zct_zplprops, tx);
3033 #define ZFS_PROP_UNDEFINED ((uint64_t)-1)
3037 * os parent objset pointer (NULL if root fs)
3038 * fuids_ok fuids allowed in this version of the spa?
3039 * sa_ok SAs allowed in this version of the spa?
3040 * createprops list of properties requested by creator
3043 * zplprops values for the zplprops we attach to the master node object
3044 * is_ci true if requested file system will be purely case-insensitive
3046 * Determine the settings for utf8only, normalization and
3047 * casesensitivity. Specific values may have been requested by the
3048 * creator and/or we can inherit values from the parent dataset. If
3049 * the file system is of too early a vintage, a creator can not
3050 * request settings for these properties, even if the requested
3051 * setting is the default value. We don't actually want to create dsl
3052 * properties for these, so remove them from the source nvlist after
3056 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
3057 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
3058 nvlist_t *zplprops, boolean_t *is_ci)
3060 uint64_t sense = ZFS_PROP_UNDEFINED;
3061 uint64_t norm = ZFS_PROP_UNDEFINED;
3062 uint64_t u8 = ZFS_PROP_UNDEFINED;
3064 ASSERT(zplprops != NULL);
3067 * Pull out creator prop choices, if any.
3070 (void) nvlist_lookup_uint64(createprops,
3071 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
3072 (void) nvlist_lookup_uint64(createprops,
3073 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
3074 (void) nvlist_remove_all(createprops,
3075 zfs_prop_to_name(ZFS_PROP_NORMALIZE));
3076 (void) nvlist_lookup_uint64(createprops,
3077 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
3078 (void) nvlist_remove_all(createprops,
3079 zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
3080 (void) nvlist_lookup_uint64(createprops,
3081 zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3082 (void) nvlist_remove_all(createprops,
3083 zfs_prop_to_name(ZFS_PROP_CASE));
3087 * If the zpl version requested is whacky or the file system
3088 * or pool is version is too "young" to support normalization
3089 * and the creator tried to set a value for one of the props,
3092 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3093 (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3094 (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3095 (zplver < ZPL_VERSION_NORMALIZATION &&
3096 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3097 sense != ZFS_PROP_UNDEFINED)))
3098 return (SET_ERROR(ENOTSUP));
3101 * Put the version in the zplprops
3103 VERIFY(nvlist_add_uint64(zplprops,
3104 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3106 if (norm == ZFS_PROP_UNDEFINED)
3107 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
3108 VERIFY(nvlist_add_uint64(zplprops,
3109 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3112 * If we're normalizing, names must always be valid UTF-8 strings.
3116 if (u8 == ZFS_PROP_UNDEFINED)
3117 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
3118 VERIFY(nvlist_add_uint64(zplprops,
3119 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3121 if (sense == ZFS_PROP_UNDEFINED)
3122 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
3123 VERIFY(nvlist_add_uint64(zplprops,
3124 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3127 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
3133 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3134 nvlist_t *zplprops, boolean_t *is_ci)
3136 boolean_t fuids_ok, sa_ok;
3137 uint64_t zplver = ZPL_VERSION;
3138 objset_t *os = NULL;
3139 char parentname[MAXNAMELEN];
3145 (void) strlcpy(parentname, dataset, sizeof (parentname));
3146 cp = strrchr(parentname, '/');
3150 if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3153 spa_vers = spa_version(spa);
3154 spa_close(spa, FTAG);
3156 zplver = zfs_zpl_version_map(spa_vers);
3157 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3158 sa_ok = (zplver >= ZPL_VERSION_SA);
3161 * Open parent object set so we can inherit zplprop values.
3163 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3166 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3168 dmu_objset_rele(os, FTAG);
3173 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3174 nvlist_t *zplprops, boolean_t *is_ci)
3178 uint64_t zplver = ZPL_VERSION;
3181 zplver = zfs_zpl_version_map(spa_vers);
3182 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3183 sa_ok = (zplver >= ZPL_VERSION_SA);
3185 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3186 createprops, zplprops, is_ci);
3192 * "type" -> dmu_objset_type_t (int32)
3193 * (optional) "props" -> { prop -> value }
3196 * outnvl: propname -> error code (int32)
3199 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3202 zfs_creat_t zct = { 0 };
3203 nvlist_t *nvprops = NULL;
3204 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3206 dmu_objset_type_t type;
3207 boolean_t is_insensitive = B_FALSE;
3209 if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3210 return (SET_ERROR(EINVAL));
3212 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3216 cbfunc = zfs_create_cb;
3220 cbfunc = zvol_create_cb;
3227 if (strchr(fsname, '@') ||
3228 strchr(fsname, '%'))
3229 return (SET_ERROR(EINVAL));
3231 zct.zct_props = nvprops;
3234 return (SET_ERROR(EINVAL));
3236 if (type == DMU_OST_ZVOL) {
3237 uint64_t volsize, volblocksize;
3239 if (nvprops == NULL)
3240 return (SET_ERROR(EINVAL));
3241 if (nvlist_lookup_uint64(nvprops,
3242 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3243 return (SET_ERROR(EINVAL));
3245 if ((error = nvlist_lookup_uint64(nvprops,
3246 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3247 &volblocksize)) != 0 && error != ENOENT)
3248 return (SET_ERROR(EINVAL));
3251 volblocksize = zfs_prop_default_numeric(
3252 ZFS_PROP_VOLBLOCKSIZE);
3254 if ((error = zvol_check_volblocksize(
3255 volblocksize)) != 0 ||
3256 (error = zvol_check_volsize(volsize,
3257 volblocksize)) != 0)
3259 } else if (type == DMU_OST_ZFS) {
3263 * We have to have normalization and
3264 * case-folding flags correct when we do the
3265 * file system creation, so go figure them out
3268 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3269 NV_UNIQUE_NAME, KM_SLEEP) == 0);
3270 error = zfs_fill_zplprops(fsname, nvprops,
3271 zct.zct_zplprops, &is_insensitive);
3273 nvlist_free(zct.zct_zplprops);
3278 error = dmu_objset_create(fsname, type,
3279 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3280 nvlist_free(zct.zct_zplprops);
3283 * It would be nice to do this atomically.
3286 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3289 (void) dsl_destroy_head(fsname);
3292 if (error == 0 && type == DMU_OST_ZVOL)
3293 zvol_create_minors(fsname);
3300 * "origin" -> name of origin snapshot
3301 * (optional) "props" -> { prop -> value }
3304 * outnvl: propname -> error code (int32)
3307 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3310 nvlist_t *nvprops = NULL;
3313 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3314 return (SET_ERROR(EINVAL));
3315 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3317 if (strchr(fsname, '@') ||
3318 strchr(fsname, '%'))
3319 return (SET_ERROR(EINVAL));
3321 if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3322 return (SET_ERROR(EINVAL));
3323 error = dmu_objset_clone(fsname, origin_name);
3328 * It would be nice to do this atomically.
3331 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3334 (void) dsl_destroy_head(fsname);
3338 zvol_create_minors(fsname);
3345 * "snaps" -> { snapshot1, snapshot2 }
3346 * (optional) "props" -> { prop -> value (string) }
3349 * outnvl: snapshot -> error code (int32)
3352 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3355 nvlist_t *props = NULL;
3359 (void) nvlist_lookup_nvlist(innvl, "props", &props);
3360 if ((error = zfs_check_userprops(poolname, props)) != 0)
3363 if (!nvlist_empty(props) &&
3364 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3365 return (SET_ERROR(ENOTSUP));
3367 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3368 return (SET_ERROR(EINVAL));
3369 poollen = strlen(poolname);
3370 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3371 pair = nvlist_next_nvpair(snaps, pair)) {
3372 const char *name = nvpair_name(pair);
3373 const char *cp = strchr(name, '@');
3376 * The snap name must contain an @, and the part after it must
3377 * contain only valid characters.
3380 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3381 return (SET_ERROR(EINVAL));
3384 * The snap must be in the specified pool.
3386 if (strncmp(name, poolname, poollen) != 0 ||
3387 (name[poollen] != '/' && name[poollen] != '@'))
3388 return (SET_ERROR(EXDEV));
3390 /* This must be the only snap of this fs. */
3391 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3392 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3393 if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3395 return (SET_ERROR(EXDEV));
3400 error = dsl_dataset_snapshot(snaps, props, outnvl);
3405 * innvl: "message" -> string
3409 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3417 * The poolname in the ioctl is not set, we get it from the TSD,
3418 * which was set at the end of the last successful ioctl that allows
3419 * logging. The secpolicy func already checked that it is set.
3420 * Only one log ioctl is allowed after each successful ioctl, so
3421 * we clear the TSD here.
3423 poolname = tsd_get(zfs_allow_log_key);
3424 (void) tsd_set(zfs_allow_log_key, NULL);
3425 error = spa_open(poolname, &spa, FTAG);
3430 if (nvlist_lookup_string(innvl, "message", &message) != 0) {
3431 spa_close(spa, FTAG);
3432 return (SET_ERROR(EINVAL));
3435 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3436 spa_close(spa, FTAG);
3437 return (SET_ERROR(ENOTSUP));
3440 error = spa_history_log(spa, message);
3441 spa_close(spa, FTAG);
3446 * The dp_config_rwlock must not be held when calling this, because the
3447 * unmount may need to write out data.
3449 * This function is best-effort. Callers must deal gracefully if it
3450 * remains mounted (or is remounted after this call).
3452 * Returns 0 if the argument is not a snapshot, or it is not currently a
3453 * filesystem, or we were able to unmount it. Returns error code otherwise.
3456 zfs_unmount_snap(const char *snapname)
3462 if (strchr(snapname, '@') == NULL)
3465 vfsp = zfs_get_vfs(snapname);
3469 zfsvfs = vfsp->vfs_data;
3470 ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os)));
3472 err = vn_vfswlock(vfsp->vfs_vnodecovered);
3475 return (SET_ERROR(err));
3478 * Always force the unmount for snapshots.
3482 (void) dounmount(vfsp, MS_FORCE, kcred);
3485 (void) dounmount(vfsp, MS_FORCE, curthread);
3492 zfs_unmount_snap_cb(const char *snapname, void *arg)
3494 return (zfs_unmount_snap(snapname));
3498 * When a clone is destroyed, its origin may also need to be destroyed,
3499 * in which case it must be unmounted. This routine will do that unmount
3503 zfs_destroy_unmount_origin(const char *fsname)
3509 error = dmu_objset_hold(fsname, FTAG, &os);
3512 ds = dmu_objset_ds(os);
3513 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3514 char originname[MAXNAMELEN];
3515 dsl_dataset_name(ds->ds_prev, originname);
3516 dmu_objset_rele(os, FTAG);
3517 (void) zfs_unmount_snap(originname);
3519 dmu_objset_rele(os, FTAG);
3525 * "snaps" -> { snapshot1, snapshot2 }
3526 * (optional boolean) "defer"
3529 * outnvl: snapshot -> error code (int32)
3534 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3541 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3542 return (SET_ERROR(EINVAL));
3543 defer = nvlist_exists(innvl, "defer");
3545 poollen = strlen(poolname);
3546 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3547 pair = nvlist_next_nvpair(snaps, pair)) {
3548 const char *name = nvpair_name(pair);
3551 * The snap must be in the specified pool to prevent the
3552 * invalid removal of zvol minors below.
3554 if (strncmp(name, poolname, poollen) != 0 ||
3555 (name[poollen] != '/' && name[poollen] != '@'))
3556 return (SET_ERROR(EXDEV));
3558 error = zfs_unmount_snap(name);
3561 #if defined(__FreeBSD__)
3562 zvol_remove_minors(name);
3566 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3570 * Create bookmarks. Bookmark names are of the form <fs>#<bmark>.
3571 * All bookmarks must be in the same pool.
3574 * bookmark1 -> snapshot1, bookmark2 -> snapshot2
3577 * outnvl: bookmark -> error code (int32)
3582 zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3584 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3585 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3589 * Verify the snapshot argument.
3591 if (nvpair_value_string(pair, &snap_name) != 0)
3592 return (SET_ERROR(EINVAL));
3595 /* Verify that the keys (bookmarks) are unique */
3596 for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair);
3597 pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) {
3598 if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0)
3599 return (SET_ERROR(EINVAL));
3603 return (dsl_bookmark_create(innvl, outnvl));
3608 * property 1, property 2, ...
3612 * bookmark name 1 -> { property 1, property 2, ... },
3613 * bookmark name 2 -> { property 1, property 2, ... }
3618 zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3620 return (dsl_get_bookmarks(fsname, innvl, outnvl));
3625 * bookmark name 1, bookmark name 2
3628 * outnvl: bookmark -> error code (int32)
3632 zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl,
3637 poollen = strlen(poolname);
3638 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3639 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3640 const char *name = nvpair_name(pair);
3641 const char *cp = strchr(name, '#');
3644 * The bookmark name must contain an #, and the part after it
3645 * must contain only valid characters.
3648 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3649 return (SET_ERROR(EINVAL));
3652 * The bookmark must be in the specified pool.
3654 if (strncmp(name, poolname, poollen) != 0 ||
3655 (name[poollen] != '/' && name[poollen] != '#'))
3656 return (SET_ERROR(EXDEV));
3659 error = dsl_bookmark_destroy(innvl, outnvl);
3665 * zc_name name of dataset to destroy
3666 * zc_objset_type type of objset
3667 * zc_defer_destroy mark for deferred destroy
3672 zfs_ioc_destroy(zfs_cmd_t *zc)
3676 if (zc->zc_objset_type == DMU_OST_ZFS) {
3677 err = zfs_unmount_snap(zc->zc_name);
3682 if (strchr(zc->zc_name, '@'))
3683 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3685 err = dsl_destroy_head(zc->zc_name);
3686 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3688 zvol_remove_minors(zc->zc_name);
3690 (void) zvol_remove_minor(zc->zc_name);
3696 * fsname is name of dataset to rollback (to most recent snapshot)
3698 * innvl is not used.
3700 * outnvl: "target" -> name of most recent snapshot
3705 zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl)
3710 if (getzfsvfs(fsname, &zfsvfs) == 0) {
3711 error = zfs_suspend_fs(zfsvfs);
3715 error = dsl_dataset_rollback(fsname, zfsvfs, outnvl);
3716 resume_err = zfs_resume_fs(zfsvfs, fsname);
3717 error = error ? error : resume_err;
3719 VFS_RELE(zfsvfs->z_vfs);
3721 error = dsl_dataset_rollback(fsname, NULL, outnvl);
3727 recursive_unmount(const char *fsname, void *arg)
3729 const char *snapname = arg;
3730 char fullname[MAXNAMELEN];
3732 (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname);
3733 return (zfs_unmount_snap(fullname));
3738 * zc_name old name of dataset
3739 * zc_value new name of dataset
3740 * zc_cookie recursive flag (only valid for snapshots)
3745 zfs_ioc_rename(zfs_cmd_t *zc)
3747 boolean_t recursive = zc->zc_cookie & 1;
3749 boolean_t allow_mounted = zc->zc_cookie & 2;
3753 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3754 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3755 strchr(zc->zc_value, '%'))
3756 return (SET_ERROR(EINVAL));
3758 at = strchr(zc->zc_name, '@');
3760 /* snaps must be in same fs */
3763 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
3764 return (SET_ERROR(EXDEV));
3767 if (zc->zc_objset_type == DMU_OST_ZFS) {
3769 if (zc->zc_objset_type == DMU_OST_ZFS && allow_mounted) {
3771 error = dmu_objset_find(zc->zc_name,
3772 recursive_unmount, at + 1,
3773 recursive ? DS_FIND_CHILDREN : 0);
3779 error = dsl_dataset_rename_snapshot(zc->zc_name,
3780 at + 1, strchr(zc->zc_value, '@') + 1, recursive);
3786 if (zc->zc_objset_type == DMU_OST_ZVOL)
3787 (void) zvol_remove_minor(zc->zc_name);
3789 return (dsl_dir_rename(zc->zc_name, zc->zc_value));
3794 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3796 const char *propname = nvpair_name(pair);
3797 boolean_t issnap = (strchr(dsname, '@') != NULL);
3798 zfs_prop_t prop = zfs_name_to_prop(propname);
3802 if (prop == ZPROP_INVAL) {
3803 if (zfs_prop_user(propname)) {
3804 if (err = zfs_secpolicy_write_perms(dsname,
3805 ZFS_DELEG_PERM_USERPROP, cr))
3810 if (!issnap && zfs_prop_userquota(propname)) {
3811 const char *perm = NULL;
3812 const char *uq_prefix =
3813 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3814 const char *gq_prefix =
3815 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3817 if (strncmp(propname, uq_prefix,
3818 strlen(uq_prefix)) == 0) {
3819 perm = ZFS_DELEG_PERM_USERQUOTA;
3820 } else if (strncmp(propname, gq_prefix,
3821 strlen(gq_prefix)) == 0) {
3822 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3824 /* USERUSED and GROUPUSED are read-only */
3825 return (SET_ERROR(EINVAL));
3828 if (err = zfs_secpolicy_write_perms(dsname, perm, cr))
3833 return (SET_ERROR(EINVAL));
3837 return (SET_ERROR(EINVAL));
3839 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3841 * dsl_prop_get_all_impl() returns properties in this
3845 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3846 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3851 * Check that this value is valid for this pool version
3854 case ZFS_PROP_COMPRESSION:
3856 * If the user specified gzip compression, make sure
3857 * the SPA supports it. We ignore any errors here since
3858 * we'll catch them later.
3860 if (nvpair_value_uint64(pair, &intval) == 0) {
3861 if (intval >= ZIO_COMPRESS_GZIP_1 &&
3862 intval <= ZIO_COMPRESS_GZIP_9 &&
3863 zfs_earlier_version(dsname,
3864 SPA_VERSION_GZIP_COMPRESSION)) {
3865 return (SET_ERROR(ENOTSUP));
3868 if (intval == ZIO_COMPRESS_ZLE &&
3869 zfs_earlier_version(dsname,
3870 SPA_VERSION_ZLE_COMPRESSION))
3871 return (SET_ERROR(ENOTSUP));
3873 if (intval == ZIO_COMPRESS_LZ4) {
3876 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3879 if (!spa_feature_is_enabled(spa,
3880 SPA_FEATURE_LZ4_COMPRESS)) {
3881 spa_close(spa, FTAG);
3882 return (SET_ERROR(ENOTSUP));
3884 spa_close(spa, FTAG);
3888 * If this is a bootable dataset then
3889 * verify that the compression algorithm
3890 * is supported for booting. We must return
3891 * something other than ENOTSUP since it
3892 * implies a downrev pool version.
3894 if (zfs_is_bootfs(dsname) &&
3895 !BOOTFS_COMPRESS_VALID(intval)) {
3896 return (SET_ERROR(ERANGE));
3901 case ZFS_PROP_COPIES:
3902 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3903 return (SET_ERROR(ENOTSUP));
3906 case ZFS_PROP_DEDUP:
3907 if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3908 return (SET_ERROR(ENOTSUP));
3911 case ZFS_PROP_RECORDSIZE:
3912 /* Record sizes above 128k need the feature to be enabled */
3913 if (nvpair_value_uint64(pair, &intval) == 0 &&
3914 intval > SPA_OLD_MAXBLOCKSIZE) {
3918 * If this is a bootable dataset then
3919 * the we don't allow large (>128K) blocks,
3920 * because GRUB doesn't support them.
3922 if (zfs_is_bootfs(dsname) &&
3923 intval > SPA_OLD_MAXBLOCKSIZE) {
3924 return (SET_ERROR(EDOM));
3928 * We don't allow setting the property above 1MB,
3929 * unless the tunable has been changed.
3931 if (intval > zfs_max_recordsize ||
3932 intval > SPA_MAXBLOCKSIZE)
3933 return (SET_ERROR(EDOM));
3935 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3938 if (!spa_feature_is_enabled(spa,
3939 SPA_FEATURE_LARGE_BLOCKS)) {
3940 spa_close(spa, FTAG);
3941 return (SET_ERROR(ENOTSUP));
3943 spa_close(spa, FTAG);
3947 case ZFS_PROP_SHARESMB:
3948 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3949 return (SET_ERROR(ENOTSUP));
3952 case ZFS_PROP_ACLINHERIT:
3953 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3954 nvpair_value_uint64(pair, &intval) == 0) {
3955 if (intval == ZFS_ACL_PASSTHROUGH_X &&
3956 zfs_earlier_version(dsname,
3957 SPA_VERSION_PASSTHROUGH_X))
3958 return (SET_ERROR(ENOTSUP));
3963 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3967 * Checks for a race condition to make sure we don't increment a feature flag
3971 zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx)
3973 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3974 spa_feature_t *featurep = arg;
3976 if (!spa_feature_is_active(spa, *featurep))
3979 return (SET_ERROR(EBUSY));
3983 * The callback invoked on feature activation in the sync task caused by
3984 * zfs_prop_activate_feature.
3987 zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx)
3989 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3990 spa_feature_t *featurep = arg;
3992 spa_feature_incr(spa, *featurep, tx);
3996 * Activates a feature on a pool in response to a property setting. This
3997 * creates a new sync task which modifies the pool to reflect the feature
4001 zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature)
4005 /* EBUSY here indicates that the feature is already active */
4006 err = dsl_sync_task(spa_name(spa),
4007 zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync,
4008 &feature, 2, ZFS_SPACE_CHECK_RESERVED);
4010 if (err != 0 && err != EBUSY)
4017 * Removes properties from the given props list that fail permission checks
4018 * needed to clear them and to restore them in case of a receive error. For each
4019 * property, make sure we have both set and inherit permissions.
4021 * Returns the first error encountered if any permission checks fail. If the
4022 * caller provides a non-NULL errlist, it also gives the complete list of names
4023 * of all the properties that failed a permission check along with the
4024 * corresponding error numbers. The caller is responsible for freeing the
4027 * If every property checks out successfully, zero is returned and the list
4028 * pointed at by errlist is NULL.
4031 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
4034 nvpair_t *pair, *next_pair;
4041 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4043 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
4044 (void) strcpy(zc->zc_name, dataset);
4045 pair = nvlist_next_nvpair(props, NULL);
4046 while (pair != NULL) {
4047 next_pair = nvlist_next_nvpair(props, pair);
4049 (void) strcpy(zc->zc_value, nvpair_name(pair));
4050 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
4051 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
4052 VERIFY(nvlist_remove_nvpair(props, pair) == 0);
4053 VERIFY(nvlist_add_int32(errors,
4054 zc->zc_value, err) == 0);
4058 kmem_free(zc, sizeof (zfs_cmd_t));
4060 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
4061 nvlist_free(errors);
4064 VERIFY(nvpair_value_int32(pair, &rv) == 0);
4067 if (errlist == NULL)
4068 nvlist_free(errors);
4076 propval_equals(nvpair_t *p1, nvpair_t *p2)
4078 if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
4079 /* dsl_prop_get_all_impl() format */
4081 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
4082 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4086 if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
4088 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
4089 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4093 if (nvpair_type(p1) != nvpair_type(p2))
4096 if (nvpair_type(p1) == DATA_TYPE_STRING) {
4097 char *valstr1, *valstr2;
4099 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
4100 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
4101 return (strcmp(valstr1, valstr2) == 0);
4103 uint64_t intval1, intval2;
4105 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
4106 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
4107 return (intval1 == intval2);
4112 * Remove properties from props if they are not going to change (as determined
4113 * by comparison with origprops). Remove them from origprops as well, since we
4114 * do not need to clear or restore properties that won't change.
4117 props_reduce(nvlist_t *props, nvlist_t *origprops)
4119 nvpair_t *pair, *next_pair;
4121 if (origprops == NULL)
4122 return; /* all props need to be received */
4124 pair = nvlist_next_nvpair(props, NULL);
4125 while (pair != NULL) {
4126 const char *propname = nvpair_name(pair);
4129 next_pair = nvlist_next_nvpair(props, pair);
4131 if ((nvlist_lookup_nvpair(origprops, propname,
4132 &match) != 0) || !propval_equals(pair, match))
4133 goto next; /* need to set received value */
4135 /* don't clear the existing received value */
4136 (void) nvlist_remove_nvpair(origprops, match);
4137 /* don't bother receiving the property */
4138 (void) nvlist_remove_nvpair(props, pair);
4145 static boolean_t zfs_ioc_recv_inject_err;
4150 * zc_name name of containing filesystem
4151 * zc_nvlist_src{_size} nvlist of properties to apply
4152 * zc_value name of snapshot to create
4153 * zc_string name of clone origin (if DRR_FLAG_CLONE)
4154 * zc_cookie file descriptor to recv from
4155 * zc_begin_record the BEGIN record of the stream (not byteswapped)
4156 * zc_guid force flag
4157 * zc_cleanup_fd cleanup-on-exit file descriptor
4158 * zc_action_handle handle for this guid/ds mapping (or zero on first call)
4161 * zc_cookie number of bytes read
4162 * zc_nvlist_dst{_size} error for each unapplied received property
4163 * zc_obj zprop_errflags_t
4164 * zc_action_handle handle for this guid/ds mapping
4167 zfs_ioc_recv(zfs_cmd_t *zc)
4170 dmu_recv_cookie_t drc;
4171 boolean_t force = (boolean_t)zc->zc_guid;
4174 int props_error = 0;
4177 nvlist_t *props = NULL; /* sent properties */
4178 nvlist_t *origprops = NULL; /* existing properties */
4179 char *origin = NULL;
4181 char tofs[ZFS_MAXNAMELEN];
4182 cap_rights_t rights;
4183 boolean_t first_recvd_props = B_FALSE;
4185 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
4186 strchr(zc->zc_value, '@') == NULL ||
4187 strchr(zc->zc_value, '%'))
4188 return (SET_ERROR(EINVAL));
4190 (void) strcpy(tofs, zc->zc_value);
4191 tosnap = strchr(tofs, '@');
4194 if (zc->zc_nvlist_src != 0 &&
4195 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
4196 zc->zc_iflags, &props)) != 0)
4203 fget_read(curthread, fd, cap_rights_init(&rights, CAP_PREAD), &fp);
4207 return (SET_ERROR(EBADF));
4210 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4212 if (zc->zc_string[0])
4213 origin = zc->zc_string;
4215 error = dmu_recv_begin(tofs, tosnap,
4216 &zc->zc_begin_record, force, origin, &drc);
4221 * Set properties before we receive the stream so that they are applied
4222 * to the new data. Note that we must call dmu_recv_stream() if
4223 * dmu_recv_begin() succeeds.
4225 if (props != NULL && !drc.drc_newfs) {
4226 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
4227 SPA_VERSION_RECVD_PROPS &&
4228 !dsl_prop_get_hasrecvd(tofs))
4229 first_recvd_props = B_TRUE;
4232 * If new received properties are supplied, they are to
4233 * completely replace the existing received properties, so stash
4234 * away the existing ones.
4236 if (dsl_prop_get_received(tofs, &origprops) == 0) {
4237 nvlist_t *errlist = NULL;
4239 * Don't bother writing a property if its value won't
4240 * change (and avoid the unnecessary security checks).
4242 * The first receive after SPA_VERSION_RECVD_PROPS is a
4243 * special case where we blow away all local properties
4246 if (!first_recvd_props)
4247 props_reduce(props, origprops);
4248 if (zfs_check_clearable(tofs, origprops, &errlist) != 0)
4249 (void) nvlist_merge(errors, errlist, 0);
4250 nvlist_free(errlist);
4252 if (clear_received_props(tofs, origprops,
4253 first_recvd_props ? NULL : props) != 0)
4254 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4256 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4260 if (props != NULL) {
4261 props_error = dsl_prop_set_hasrecvd(tofs);
4263 if (props_error == 0) {
4264 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4269 if (zc->zc_nvlist_dst_size != 0 &&
4270 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4271 put_nvlist(zc, errors) != 0)) {
4273 * Caller made zc->zc_nvlist_dst less than the minimum expected
4274 * size or supplied an invalid address.
4276 props_error = SET_ERROR(EINVAL);
4280 error = dmu_recv_stream(&drc, fp, &off, zc->zc_cleanup_fd,
4281 &zc->zc_action_handle);
4284 zfsvfs_t *zfsvfs = NULL;
4286 if (getzfsvfs(tofs, &zfsvfs) == 0) {
4290 error = zfs_suspend_fs(zfsvfs);
4292 * If the suspend fails, then the recv_end will
4293 * likely also fail, and clean up after itself.
4295 end_err = dmu_recv_end(&drc, zfsvfs);
4297 error = zfs_resume_fs(zfsvfs, tofs);
4298 error = error ? error : end_err;
4299 VFS_RELE(zfsvfs->z_vfs);
4301 error = dmu_recv_end(&drc, NULL);
4305 zc->zc_cookie = off - fp->f_offset;
4306 if (off >= 0 && off <= MAXOFFSET_T)
4310 if (zfs_ioc_recv_inject_err) {
4311 zfs_ioc_recv_inject_err = B_FALSE;
4318 zvol_create_minors(tofs);
4322 * On error, restore the original props.
4324 if (error != 0 && props != NULL && !drc.drc_newfs) {
4325 if (clear_received_props(tofs, props, NULL) != 0) {
4327 * We failed to clear the received properties.
4328 * Since we may have left a $recvd value on the
4329 * system, we can't clear the $hasrecvd flag.
4331 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4332 } else if (first_recvd_props) {
4333 dsl_prop_unset_hasrecvd(tofs);
4336 if (origprops == NULL && !drc.drc_newfs) {
4337 /* We failed to stash the original properties. */
4338 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4342 * dsl_props_set() will not convert RECEIVED to LOCAL on or
4343 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4344 * explictly if we're restoring local properties cleared in the
4345 * first new-style receive.
4347 if (origprops != NULL &&
4348 zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4349 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4350 origprops, NULL) != 0) {
4352 * We stashed the original properties but failed to
4355 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4360 nvlist_free(origprops);
4361 nvlist_free(errors);
4365 error = props_error;
4372 * zc_name name of snapshot to send
4373 * zc_cookie file descriptor to send stream to
4374 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj)
4375 * zc_sendobj objsetid of snapshot to send
4376 * zc_fromobj objsetid of incremental fromsnap (may be zero)
4377 * zc_guid if set, estimate size of stream only. zc_cookie is ignored.
4378 * output size in zc_objset_type.
4379 * zc_flags lzc_send_flags
4382 * zc_objset_type estimated size, if zc_guid is set
4385 zfs_ioc_send(zfs_cmd_t *zc)
4389 boolean_t estimate = (zc->zc_guid != 0);
4390 boolean_t embedok = (zc->zc_flags & 0x1);
4391 boolean_t large_block_ok = (zc->zc_flags & 0x2);
4393 if (zc->zc_obj != 0) {
4395 dsl_dataset_t *tosnap;
4397 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4401 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4403 dsl_pool_rele(dp, FTAG);
4407 if (dsl_dir_is_clone(tosnap->ds_dir))
4409 dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj;
4410 dsl_dataset_rele(tosnap, FTAG);
4411 dsl_pool_rele(dp, FTAG);
4416 dsl_dataset_t *tosnap;
4417 dsl_dataset_t *fromsnap = NULL;
4419 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4423 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4425 dsl_pool_rele(dp, FTAG);
4429 if (zc->zc_fromobj != 0) {
4430 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
4433 dsl_dataset_rele(tosnap, FTAG);
4434 dsl_pool_rele(dp, FTAG);
4439 error = dmu_send_estimate(tosnap, fromsnap,
4440 &zc->zc_objset_type);
4442 if (fromsnap != NULL)
4443 dsl_dataset_rele(fromsnap, FTAG);
4444 dsl_dataset_rele(tosnap, FTAG);
4445 dsl_pool_rele(dp, FTAG);
4448 cap_rights_t rights;
4451 fp = getf(zc->zc_cookie);
4453 fget_write(curthread, zc->zc_cookie,
4454 cap_rights_init(&rights, CAP_WRITE), &fp);
4457 return (SET_ERROR(EBADF));
4460 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
4461 zc->zc_fromobj, embedok, large_block_ok,
4463 zc->zc_cookie, fp->f_vnode, &off);
4465 zc->zc_cookie, fp, &off);
4468 if (off >= 0 && off <= MAXOFFSET_T)
4470 releasef(zc->zc_cookie);
4477 * zc_name name of snapshot on which to report progress
4478 * zc_cookie file descriptor of send stream
4481 * zc_cookie number of bytes written in send stream thus far
4484 zfs_ioc_send_progress(zfs_cmd_t *zc)
4488 dmu_sendarg_t *dsp = NULL;
4491 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4495 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4497 dsl_pool_rele(dp, FTAG);
4501 mutex_enter(&ds->ds_sendstream_lock);
4504 * Iterate over all the send streams currently active on this dataset.
4505 * If there's one which matches the specified file descriptor _and_ the
4506 * stream was started by the current process, return the progress of
4509 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4510 dsp = list_next(&ds->ds_sendstreams, dsp)) {
4511 if (dsp->dsa_outfd == zc->zc_cookie &&
4512 dsp->dsa_proc == curproc)
4517 zc->zc_cookie = *(dsp->dsa_off);
4519 error = SET_ERROR(ENOENT);
4521 mutex_exit(&ds->ds_sendstream_lock);
4522 dsl_dataset_rele(ds, FTAG);
4523 dsl_pool_rele(dp, FTAG);
4528 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4532 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4533 &zc->zc_inject_record);
4536 zc->zc_guid = (uint64_t)id;
4542 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4544 return (zio_clear_fault((int)zc->zc_guid));
4548 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4550 int id = (int)zc->zc_guid;
4553 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4554 &zc->zc_inject_record);
4562 zfs_ioc_error_log(zfs_cmd_t *zc)
4566 size_t count = (size_t)zc->zc_nvlist_dst_size;
4568 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4571 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4574 zc->zc_nvlist_dst_size = count;
4576 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4578 spa_close(spa, FTAG);
4584 zfs_ioc_clear(zfs_cmd_t *zc)
4591 * On zpool clear we also fix up missing slogs
4593 mutex_enter(&spa_namespace_lock);
4594 spa = spa_lookup(zc->zc_name);
4596 mutex_exit(&spa_namespace_lock);
4597 return (SET_ERROR(EIO));
4599 if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4600 /* we need to let spa_open/spa_load clear the chains */
4601 spa_set_log_state(spa, SPA_LOG_CLEAR);
4603 spa->spa_last_open_failed = 0;
4604 mutex_exit(&spa_namespace_lock);
4606 if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4607 error = spa_open(zc->zc_name, &spa, FTAG);
4610 nvlist_t *config = NULL;
4612 if (zc->zc_nvlist_src == 0)
4613 return (SET_ERROR(EINVAL));
4615 if ((error = get_nvlist(zc->zc_nvlist_src,
4616 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4617 error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4619 if (config != NULL) {
4622 if ((err = put_nvlist(zc, config)) != 0)
4624 nvlist_free(config);
4626 nvlist_free(policy);
4633 spa_vdev_state_enter(spa, SCL_NONE);
4635 if (zc->zc_guid == 0) {
4638 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4640 (void) spa_vdev_state_exit(spa, NULL, ENODEV);
4641 spa_close(spa, FTAG);
4642 return (SET_ERROR(ENODEV));
4646 vdev_clear(spa, vd);
4648 (void) spa_vdev_state_exit(spa, NULL, 0);
4651 * Resume any suspended I/Os.
4653 if (zio_resume(spa) != 0)
4654 error = SET_ERROR(EIO);
4656 spa_close(spa, FTAG);
4662 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4667 error = spa_open(zc->zc_name, &spa, FTAG);
4671 spa_vdev_state_enter(spa, SCL_NONE);
4674 * If a resilver is already in progress then set the
4675 * spa_scrub_reopen flag to B_TRUE so that we don't restart
4676 * the scan as a side effect of the reopen. Otherwise, let
4677 * vdev_open() decided if a resilver is required.
4679 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4680 vdev_reopen(spa->spa_root_vdev);
4681 spa->spa_scrub_reopen = B_FALSE;
4683 (void) spa_vdev_state_exit(spa, NULL, 0);
4684 spa_close(spa, FTAG);
4689 * zc_name name of filesystem
4690 * zc_value name of origin snapshot
4693 * zc_string name of conflicting snapshot, if there is one
4696 zfs_ioc_promote(zfs_cmd_t *zc)
4701 * We don't need to unmount *all* the origin fs's snapshots, but
4704 cp = strchr(zc->zc_value, '@');
4707 (void) dmu_objset_find(zc->zc_value,
4708 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
4709 return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4713 * Retrieve a single {user|group}{used|quota}@... property.
4716 * zc_name name of filesystem
4717 * zc_objset_type zfs_userquota_prop_t
4718 * zc_value domain name (eg. "S-1-234-567-89")
4719 * zc_guid RID/UID/GID
4722 * zc_cookie property value
4725 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4730 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4731 return (SET_ERROR(EINVAL));
4733 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4737 error = zfs_userspace_one(zfsvfs,
4738 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4739 zfsvfs_rele(zfsvfs, FTAG);
4746 * zc_name name of filesystem
4747 * zc_cookie zap cursor
4748 * zc_objset_type zfs_userquota_prop_t
4749 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4752 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
4753 * zc_cookie zap cursor
4756 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4759 int bufsize = zc->zc_nvlist_dst_size;
4762 return (SET_ERROR(ENOMEM));
4764 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4768 void *buf = kmem_alloc(bufsize, KM_SLEEP);
4770 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
4771 buf, &zc->zc_nvlist_dst_size);
4774 error = ddi_copyout(buf,
4775 (void *)(uintptr_t)zc->zc_nvlist_dst,
4776 zc->zc_nvlist_dst_size, zc->zc_iflags);
4778 kmem_free(buf, bufsize);
4779 zfsvfs_rele(zfsvfs, FTAG);
4786 * zc_name name of filesystem
4792 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4798 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
4799 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
4801 * If userused is not enabled, it may be because the
4802 * objset needs to be closed & reopened (to grow the
4803 * objset_phys_t). Suspend/resume the fs will do that.
4805 error = zfs_suspend_fs(zfsvfs);
4807 dmu_objset_refresh_ownership(zfsvfs->z_os,
4809 error = zfs_resume_fs(zfsvfs, zc->zc_name);
4813 error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
4814 VFS_RELE(zfsvfs->z_vfs);
4816 /* XXX kind of reading contents without owning */
4817 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4821 error = dmu_objset_userspace_upgrade(os);
4822 dmu_objset_rele(os, FTAG);
4830 * We don't want to have a hard dependency
4831 * against some special symbols in sharefs
4832 * nfs, and smbsrv. Determine them if needed when
4833 * the first file system is shared.
4834 * Neither sharefs, nfs or smbsrv are unloadable modules.
4836 int (*znfsexport_fs)(void *arg);
4837 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
4838 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
4840 int zfs_nfsshare_inited;
4841 int zfs_smbshare_inited;
4843 ddi_modhandle_t nfs_mod;
4844 ddi_modhandle_t sharefs_mod;
4845 ddi_modhandle_t smbsrv_mod;
4847 kmutex_t zfs_share_lock;
4855 ASSERT(MUTEX_HELD(&zfs_share_lock));
4856 /* Both NFS and SMB shares also require sharetab support. */
4857 if (sharefs_mod == NULL && ((sharefs_mod =
4858 ddi_modopen("fs/sharefs",
4859 KRTLD_MODE_FIRST, &error)) == NULL)) {
4860 return (SET_ERROR(ENOSYS));
4862 if (zshare_fs == NULL && ((zshare_fs =
4863 (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
4864 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
4865 return (SET_ERROR(ENOSYS));
4872 zfs_ioc_share(zfs_cmd_t *zc)
4878 switch (zc->zc_share.z_sharetype) {
4880 case ZFS_UNSHARE_NFS:
4881 if (zfs_nfsshare_inited == 0) {
4882 mutex_enter(&zfs_share_lock);
4883 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
4884 KRTLD_MODE_FIRST, &error)) == NULL)) {
4885 mutex_exit(&zfs_share_lock);
4886 return (SET_ERROR(ENOSYS));
4888 if (znfsexport_fs == NULL &&
4889 ((znfsexport_fs = (int (*)(void *))
4891 "nfs_export", &error)) == NULL)) {
4892 mutex_exit(&zfs_share_lock);
4893 return (SET_ERROR(ENOSYS));
4895 error = zfs_init_sharefs();
4897 mutex_exit(&zfs_share_lock);
4898 return (SET_ERROR(ENOSYS));
4900 zfs_nfsshare_inited = 1;
4901 mutex_exit(&zfs_share_lock);
4905 case ZFS_UNSHARE_SMB:
4906 if (zfs_smbshare_inited == 0) {
4907 mutex_enter(&zfs_share_lock);
4908 if (smbsrv_mod == NULL && ((smbsrv_mod =
4909 ddi_modopen("drv/smbsrv",
4910 KRTLD_MODE_FIRST, &error)) == NULL)) {
4911 mutex_exit(&zfs_share_lock);
4912 return (SET_ERROR(ENOSYS));
4914 if (zsmbexport_fs == NULL && ((zsmbexport_fs =
4915 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
4916 "smb_server_share", &error)) == NULL)) {
4917 mutex_exit(&zfs_share_lock);
4918 return (SET_ERROR(ENOSYS));
4920 error = zfs_init_sharefs();
4922 mutex_exit(&zfs_share_lock);
4923 return (SET_ERROR(ENOSYS));
4925 zfs_smbshare_inited = 1;
4926 mutex_exit(&zfs_share_lock);
4930 return (SET_ERROR(EINVAL));
4933 switch (zc->zc_share.z_sharetype) {
4935 case ZFS_UNSHARE_NFS:
4937 znfsexport_fs((void *)
4938 (uintptr_t)zc->zc_share.z_exportdata))
4942 case ZFS_UNSHARE_SMB:
4943 if (error = zsmbexport_fs((void *)
4944 (uintptr_t)zc->zc_share.z_exportdata,
4945 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
4952 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
4953 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
4954 SHAREFS_ADD : SHAREFS_REMOVE;
4957 * Add or remove share from sharetab
4959 error = zshare_fs(opcode,
4960 (void *)(uintptr_t)zc->zc_share.z_sharedata,
4961 zc->zc_share.z_sharemax);
4970 ace_t full_access[] = {
4971 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4976 * zc_name name of containing filesystem
4977 * zc_obj object # beyond which we want next in-use object #
4980 * zc_obj next in-use object #
4983 zfs_ioc_next_obj(zfs_cmd_t *zc)
4985 objset_t *os = NULL;
4988 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4992 error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
4993 dsl_dataset_phys(os->os_dsl_dataset)->ds_prev_snap_txg);
4995 dmu_objset_rele(os, FTAG);
5001 * zc_name name of filesystem
5002 * zc_value prefix name for snapshot
5003 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
5006 * zc_value short name of new snapshot
5009 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
5016 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
5020 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
5021 (u_longlong_t)ddi_get_lbolt64());
5022 hold_name = kmem_asprintf("%%%s", zc->zc_value);
5024 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
5027 (void) strcpy(zc->zc_value, snap_name);
5030 zfs_onexit_fd_rele(zc->zc_cleanup_fd);
5036 * zc_name name of "to" snapshot
5037 * zc_value name of "from" snapshot
5038 * zc_cookie file descriptor to write diff data on
5041 * dmu_diff_record_t's to the file descriptor
5044 zfs_ioc_diff(zfs_cmd_t *zc)
5047 cap_rights_t rights;
5052 fp = getf(zc->zc_cookie);
5054 fget_write(curthread, zc->zc_cookie,
5055 cap_rights_init(&rights, CAP_WRITE), &fp);
5058 return (SET_ERROR(EBADF));
5063 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off);
5065 error = dmu_diff(zc->zc_name, zc->zc_value, fp, &off);
5068 if (off >= 0 && off <= MAXOFFSET_T)
5070 releasef(zc->zc_cookie);
5077 * Remove all ACL files in shares dir
5080 zfs_smb_acl_purge(znode_t *dzp)
5083 zap_attribute_t zap;
5084 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
5087 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
5088 (error = zap_cursor_retrieve(&zc, &zap)) == 0;
5089 zap_cursor_advance(&zc)) {
5090 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
5094 zap_cursor_fini(&zc);
5100 zfs_ioc_smb_acl(zfs_cmd_t *zc)
5105 vnode_t *resourcevp = NULL;
5114 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
5115 NO_FOLLOW, NULL, &vp)) != 0)
5118 /* Now make sure mntpnt and dataset are ZFS */
5120 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 ||
5121 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
5122 zc->zc_name) != 0)) {
5124 return (SET_ERROR(EINVAL));
5128 zfsvfs = dzp->z_zfsvfs;
5132 * Create share dir if its missing.
5134 mutex_enter(&zfsvfs->z_lock);
5135 if (zfsvfs->z_shares_dir == 0) {
5138 tx = dmu_tx_create(zfsvfs->z_os);
5139 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
5141 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
5142 error = dmu_tx_assign(tx, TXG_WAIT);
5146 error = zfs_create_share_dir(zfsvfs, tx);
5150 mutex_exit(&zfsvfs->z_lock);
5156 mutex_exit(&zfsvfs->z_lock);
5158 ASSERT(zfsvfs->z_shares_dir);
5159 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
5165 switch (zc->zc_cookie) {
5166 case ZFS_SMB_ACL_ADD:
5167 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
5168 vattr.va_type = VREG;
5169 vattr.va_mode = S_IFREG|0777;
5173 vsec.vsa_mask = VSA_ACE;
5174 vsec.vsa_aclentp = &full_access;
5175 vsec.vsa_aclentsz = sizeof (full_access);
5176 vsec.vsa_aclcnt = 1;
5178 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
5179 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
5181 VN_RELE(resourcevp);
5184 case ZFS_SMB_ACL_REMOVE:
5185 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
5189 case ZFS_SMB_ACL_RENAME:
5190 if ((error = get_nvlist(zc->zc_nvlist_src,
5191 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
5196 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
5197 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
5200 VN_RELE(ZTOV(sharedir));
5202 nvlist_free(nvlist);
5205 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
5207 nvlist_free(nvlist);
5210 case ZFS_SMB_ACL_PURGE:
5211 error = zfs_smb_acl_purge(sharedir);
5215 error = SET_ERROR(EINVAL);
5220 VN_RELE(ZTOV(sharedir));
5226 return (EOPNOTSUPP);
5232 * "holds" -> { snapname -> holdname (string), ... }
5233 * (optional) "cleanup_fd" -> fd (int32)
5237 * snapname -> error value (int32)
5243 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
5246 int cleanup_fd = -1;
5250 error = nvlist_lookup_nvlist(args, "holds", &holds);
5252 return (SET_ERROR(EINVAL));
5254 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
5255 error = zfs_onexit_fd_hold(cleanup_fd, &minor);
5260 error = dsl_dataset_user_hold(holds, minor, errlist);
5262 zfs_onexit_fd_rele(cleanup_fd);
5267 * innvl is not used.
5270 * holdname -> time added (uint64 seconds since epoch)
5276 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
5278 return (dsl_dataset_get_holds(snapname, outnvl));
5283 * snapname -> { holdname, ... }
5288 * snapname -> error value (int32)
5294 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
5296 return (dsl_dataset_user_release(holds, errlist));
5301 * zc_name name of new filesystem or snapshot
5302 * zc_value full name of old snapshot
5305 * zc_cookie space in bytes
5306 * zc_objset_type compressed space in bytes
5307 * zc_perm_action uncompressed space in bytes
5310 zfs_ioc_space_written(zfs_cmd_t *zc)
5314 dsl_dataset_t *new, *old;
5316 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5319 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
5321 dsl_pool_rele(dp, FTAG);
5324 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
5326 dsl_dataset_rele(new, FTAG);
5327 dsl_pool_rele(dp, FTAG);
5331 error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
5332 &zc->zc_objset_type, &zc->zc_perm_action);
5333 dsl_dataset_rele(old, FTAG);
5334 dsl_dataset_rele(new, FTAG);
5335 dsl_pool_rele(dp, FTAG);
5341 * "firstsnap" -> snapshot name
5345 * "used" -> space in bytes
5346 * "compressed" -> compressed space in bytes
5347 * "uncompressed" -> uncompressed space in bytes
5351 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
5355 dsl_dataset_t *new, *old;
5357 uint64_t used, comp, uncomp;
5359 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5360 return (SET_ERROR(EINVAL));
5362 error = dsl_pool_hold(lastsnap, FTAG, &dp);
5366 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
5367 if (error == 0 && !dsl_dataset_is_snapshot(new)) {
5368 dsl_dataset_rele(new, FTAG);
5369 error = SET_ERROR(EINVAL);
5372 dsl_pool_rele(dp, FTAG);
5375 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
5376 if (error == 0 && !dsl_dataset_is_snapshot(old)) {
5377 dsl_dataset_rele(old, FTAG);
5378 error = SET_ERROR(EINVAL);
5381 dsl_dataset_rele(new, FTAG);
5382 dsl_pool_rele(dp, FTAG);
5386 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5387 dsl_dataset_rele(old, FTAG);
5388 dsl_dataset_rele(new, FTAG);
5389 dsl_pool_rele(dp, FTAG);
5390 fnvlist_add_uint64(outnvl, "used", used);
5391 fnvlist_add_uint64(outnvl, "compressed", comp);
5392 fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5397 zfs_ioc_jail(zfs_cmd_t *zc)
5400 return (zone_dataset_attach(curthread->td_ucred, zc->zc_name,
5401 (int)zc->zc_jailid));
5405 zfs_ioc_unjail(zfs_cmd_t *zc)
5408 return (zone_dataset_detach(curthread->td_ucred, zc->zc_name,
5409 (int)zc->zc_jailid));
5414 * "fd" -> file descriptor to write stream to (int32)
5415 * (optional) "fromsnap" -> full snap name to send an incremental from
5416 * (optional) "largeblockok" -> (value ignored)
5417 * indicates that blocks > 128KB are permitted
5418 * (optional) "embedok" -> (value ignored)
5419 * presence indicates DRR_WRITE_EMBEDDED records are permitted
5426 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5428 cap_rights_t rights;
5432 char *fromname = NULL;
5434 boolean_t largeblockok;
5437 error = nvlist_lookup_int32(innvl, "fd", &fd);
5439 return (SET_ERROR(EINVAL));
5441 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
5443 largeblockok = nvlist_exists(innvl, "largeblockok");
5444 embedok = nvlist_exists(innvl, "embedok");
5447 file_t *fp = getf(fd);
5449 fget_write(curthread, fd, cap_rights_init(&rights, CAP_WRITE), &fp);
5452 return (SET_ERROR(EBADF));
5455 error = dmu_send(snapname, fromname, embedok, largeblockok,
5457 fd, fp->f_vnode, &off);
5463 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5474 * Determine approximately how large a zfs send stream will be -- the number
5475 * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5478 * (optional) "fromsnap" -> full snap name to send an incremental from
5482 * "space" -> bytes of space (uint64)
5486 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5489 dsl_dataset_t *fromsnap = NULL;
5490 dsl_dataset_t *tosnap;
5495 error = dsl_pool_hold(snapname, FTAG, &dp);
5499 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
5501 dsl_pool_rele(dp, FTAG);
5505 error = nvlist_lookup_string(innvl, "fromsnap", &fromname);
5507 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
5509 dsl_dataset_rele(tosnap, FTAG);
5510 dsl_pool_rele(dp, FTAG);
5515 error = dmu_send_estimate(tosnap, fromsnap, &space);
5516 fnvlist_add_uint64(outnvl, "space", space);
5518 if (fromsnap != NULL)
5519 dsl_dataset_rele(fromsnap, FTAG);
5520 dsl_dataset_rele(tosnap, FTAG);
5521 dsl_pool_rele(dp, FTAG);
5526 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
5529 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5530 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5531 boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
5533 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5535 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5536 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5537 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5538 ASSERT3P(vec->zvec_func, ==, NULL);
5540 vec->zvec_legacy_func = func;
5541 vec->zvec_secpolicy = secpolicy;
5542 vec->zvec_namecheck = namecheck;
5543 vec->zvec_allow_log = log_history;
5544 vec->zvec_pool_check = pool_check;
5548 * See the block comment at the beginning of this file for details on
5549 * each argument to this function.
5552 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
5553 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5554 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
5555 boolean_t allow_log)
5557 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5559 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5560 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5561 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5562 ASSERT3P(vec->zvec_func, ==, NULL);
5564 /* if we are logging, the name must be valid */
5565 ASSERT(!allow_log || namecheck != NO_NAME);
5567 vec->zvec_name = name;
5568 vec->zvec_func = func;
5569 vec->zvec_secpolicy = secpolicy;
5570 vec->zvec_namecheck = namecheck;
5571 vec->zvec_pool_check = pool_check;
5572 vec->zvec_smush_outnvlist = smush_outnvlist;
5573 vec->zvec_allow_log = allow_log;
5577 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5578 zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
5579 zfs_ioc_poolcheck_t pool_check)
5581 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5582 POOL_NAME, log_history, pool_check);
5586 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5587 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
5589 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5590 DATASET_NAME, B_FALSE, pool_check);
5594 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5596 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
5597 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5601 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5602 zfs_secpolicy_func_t *secpolicy)
5604 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5605 NO_NAME, B_FALSE, POOL_CHECK_NONE);
5609 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
5610 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
5612 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5613 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
5617 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5619 zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
5620 zfs_secpolicy_read);
5624 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5625 zfs_secpolicy_func_t *secpolicy)
5627 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5628 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5632 zfs_ioctl_init(void)
5634 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
5635 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
5636 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5638 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
5639 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
5640 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
5642 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
5643 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
5644 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5646 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
5647 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
5648 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5650 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
5651 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
5652 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5654 zfs_ioctl_register("create", ZFS_IOC_CREATE,
5655 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
5656 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5658 zfs_ioctl_register("clone", ZFS_IOC_CLONE,
5659 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
5660 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5662 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
5663 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
5664 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5666 zfs_ioctl_register("hold", ZFS_IOC_HOLD,
5667 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
5668 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5669 zfs_ioctl_register("release", ZFS_IOC_RELEASE,
5670 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
5671 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5673 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
5674 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
5675 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5677 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
5678 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
5679 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE);
5681 zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK,
5682 zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME,
5683 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5685 zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS,
5686 zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME,
5687 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5689 zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS,
5690 zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks,
5692 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5694 /* IOCTLS that use the legacy function signature */
5696 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
5697 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
5699 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
5700 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5701 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
5703 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
5704 zfs_ioc_pool_upgrade);
5705 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
5707 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
5708 zfs_ioc_vdev_remove);
5709 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
5710 zfs_ioc_vdev_set_state);
5711 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
5712 zfs_ioc_vdev_attach);
5713 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
5714 zfs_ioc_vdev_detach);
5715 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
5716 zfs_ioc_vdev_setpath);
5717 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
5718 zfs_ioc_vdev_setfru);
5719 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
5720 zfs_ioc_pool_set_props);
5721 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
5722 zfs_ioc_vdev_split);
5723 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
5724 zfs_ioc_pool_reguid);
5726 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
5727 zfs_ioc_pool_configs, zfs_secpolicy_none);
5728 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
5729 zfs_ioc_pool_tryimport, zfs_secpolicy_config);
5730 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
5731 zfs_ioc_inject_fault, zfs_secpolicy_inject);
5732 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
5733 zfs_ioc_clear_fault, zfs_secpolicy_inject);
5734 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
5735 zfs_ioc_inject_list_next, zfs_secpolicy_inject);
5738 * pool destroy, and export don't log the history as part of
5739 * zfsdev_ioctl, but rather zfs_ioc_pool_export
5740 * does the logging of those commands.
5742 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
5743 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5744 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
5745 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5747 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
5748 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5749 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
5750 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5752 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
5753 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_NONE);
5754 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
5755 zfs_ioc_dsobj_to_dsname,
5756 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_NONE);
5757 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
5758 zfs_ioc_pool_get_history,
5759 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
5761 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
5762 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5764 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
5765 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5766 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
5767 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5769 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
5770 zfs_ioc_space_written);
5771 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
5772 zfs_ioc_objset_recvd_props);
5773 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
5775 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
5777 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
5778 zfs_ioc_objset_stats);
5779 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
5780 zfs_ioc_objset_zplprops);
5781 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
5782 zfs_ioc_dataset_list_next);
5783 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
5784 zfs_ioc_snapshot_list_next);
5785 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
5786 zfs_ioc_send_progress);
5788 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
5789 zfs_ioc_diff, zfs_secpolicy_diff);
5790 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
5791 zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
5792 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
5793 zfs_ioc_obj_to_path, zfs_secpolicy_diff);
5794 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
5795 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
5796 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
5797 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
5798 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
5799 zfs_ioc_send, zfs_secpolicy_send);
5801 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
5802 zfs_secpolicy_none);
5803 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
5804 zfs_secpolicy_destroy);
5805 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
5806 zfs_secpolicy_rename);
5807 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
5808 zfs_secpolicy_recv);
5809 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
5810 zfs_secpolicy_promote);
5811 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
5812 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
5813 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
5814 zfs_secpolicy_set_fsacl);
5816 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
5817 zfs_secpolicy_share, POOL_CHECK_NONE);
5818 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
5819 zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
5820 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
5821 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
5822 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5823 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
5824 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
5825 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5828 zfs_ioctl_register_dataset_nolog(ZFS_IOC_JAIL, zfs_ioc_jail,
5829 zfs_secpolicy_config, POOL_CHECK_NONE);
5830 zfs_ioctl_register_dataset_nolog(ZFS_IOC_UNJAIL, zfs_ioc_unjail,
5831 zfs_secpolicy_config, POOL_CHECK_NONE);
5836 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
5837 zfs_ioc_poolcheck_t check)
5842 ASSERT(type == POOL_NAME || type == DATASET_NAME);
5844 if (check & POOL_CHECK_NONE)
5847 error = spa_open(name, &spa, FTAG);
5849 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
5850 error = SET_ERROR(EAGAIN);
5851 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
5852 error = SET_ERROR(EROFS);
5853 spa_close(spa, FTAG);
5859 * Find a free minor number.
5862 zfsdev_minor_alloc(void)
5864 static minor_t last_minor;
5867 ASSERT(MUTEX_HELD(&spa_namespace_lock));
5869 for (m = last_minor + 1; m != last_minor; m++) {
5870 if (m > ZFSDEV_MAX_MINOR)
5872 if (ddi_get_soft_state(zfsdev_state, m) == NULL) {
5882 zfs_ctldev_init(struct cdev *devp)
5885 zfs_soft_state_t *zs;
5887 ASSERT(MUTEX_HELD(&spa_namespace_lock));
5889 minor = zfsdev_minor_alloc();
5891 return (SET_ERROR(ENXIO));
5893 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS)
5894 return (SET_ERROR(EAGAIN));
5896 devfs_set_cdevpriv((void *)(uintptr_t)minor, zfsdev_close);
5898 zs = ddi_get_soft_state(zfsdev_state, minor);
5899 zs->zss_type = ZSST_CTLDEV;
5900 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data);
5906 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor)
5908 ASSERT(MUTEX_HELD(&spa_namespace_lock));
5910 zfs_onexit_destroy(zo);
5911 ddi_soft_state_free(zfsdev_state, minor);
5915 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which)
5917 zfs_soft_state_t *zp;
5919 zp = ddi_get_soft_state(zfsdev_state, minor);
5920 if (zp == NULL || zp->zss_type != which)
5923 return (zp->zss_data);
5927 zfsdev_open(struct cdev *devp, int flag, int mode, struct thread *td)
5932 if (getminor(*devp) != 0)
5933 return (zvol_open(devp, flag, otyp, cr));
5936 /* This is the control device. Allocate a new minor if requested. */
5938 mutex_enter(&spa_namespace_lock);
5939 error = zfs_ctldev_init(devp);
5940 mutex_exit(&spa_namespace_lock);
5947 zfsdev_close(void *data)
5950 minor_t minor = (minor_t)(uintptr_t)data;
5955 mutex_enter(&spa_namespace_lock);
5956 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV);
5958 mutex_exit(&spa_namespace_lock);
5961 zfs_ctldev_destroy(zo, minor);
5962 mutex_exit(&spa_namespace_lock);
5966 zfsdev_ioctl(struct cdev *dev, u_long zcmd, caddr_t arg, int flag,
5973 minor_t minor = getminor(dev);
5975 zfs_iocparm_t *zc_iocparm;
5976 int cflag, cmd, oldvecnum;
5977 boolean_t newioc, compat;
5978 void *compat_zc = NULL;
5979 cred_t *cr = td->td_ucred;
5981 const zfs_ioc_vec_t *vec;
5982 char *saved_poolname = NULL;
5983 nvlist_t *innvl = NULL;
5985 cflag = ZFS_CMD_COMPAT_NONE;
5987 newioc = B_TRUE; /* "new" style (zfs_iocparm_t) ioctl */
5989 len = IOCPARM_LEN(zcmd);
5990 vecnum = cmd = zcmd & 0xff;
5993 * Check if we are talking to supported older binaries
5994 * and translate zfs_cmd if necessary
5996 if (len != sizeof(zfs_iocparm_t)) {
6003 case sizeof(zfs_cmd_zcmd_t):
6004 cflag = ZFS_CMD_COMPAT_LZC;
6006 case sizeof(zfs_cmd_deadman_t):
6007 cflag = ZFS_CMD_COMPAT_DEADMAN;
6009 case sizeof(zfs_cmd_v28_t):
6010 cflag = ZFS_CMD_COMPAT_V28;
6012 case sizeof(zfs_cmd_v15_t):
6013 cflag = ZFS_CMD_COMPAT_V15;
6014 vecnum = zfs_ioctl_v15_to_v28[cmd];
6017 * Return without further handling
6018 * if the command is blacklisted.
6020 if (vecnum == ZFS_IOC_COMPAT_PASS)
6022 else if (vecnum == ZFS_IOC_COMPAT_FAIL)
6031 vecnum = cmd - ZFS_IOC_FIRST;
6032 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
6035 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
6036 return (SET_ERROR(EINVAL));
6037 vec = &zfs_ioc_vec[vecnum];
6039 zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP);
6042 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
6044 error = SET_ERROR(EFAULT);
6047 #else /* !illumos */
6048 bzero(zc, sizeof(zfs_cmd_t));
6051 zc_iocparm = (void *)arg;
6053 switch (zc_iocparm->zfs_ioctl_version) {
6054 case ZFS_IOCVER_CURRENT:
6055 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_t)) {
6056 error = SET_ERROR(EINVAL);
6060 case ZFS_IOCVER_ZCMD:
6061 if (zc_iocparm->zfs_cmd_size > sizeof(zfs_cmd_t) ||
6062 zc_iocparm->zfs_cmd_size < sizeof(zfs_cmd_zcmd_t)) {
6063 error = SET_ERROR(EFAULT);
6067 cflag = ZFS_CMD_COMPAT_ZCMD;
6070 error = SET_ERROR(EINVAL);
6076 ASSERT(sizeof(zfs_cmd_t) >= zc_iocparm->zfs_cmd_size);
6077 compat_zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP);
6078 bzero(compat_zc, sizeof(zfs_cmd_t));
6080 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd,
6081 compat_zc, zc_iocparm->zfs_cmd_size, flag);
6083 error = SET_ERROR(EFAULT);
6087 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd,
6088 zc, zc_iocparm->zfs_cmd_size, flag);
6090 error = SET_ERROR(EFAULT);
6098 ASSERT(compat_zc != NULL);
6099 zfs_cmd_compat_get(zc, compat_zc, cflag);
6101 ASSERT(compat_zc == NULL);
6102 zfs_cmd_compat_get(zc, arg, cflag);
6105 error = zfs_ioctl_compat_pre(zc, &vecnum, cflag);
6108 if (oldvecnum != vecnum)
6109 vec = &zfs_ioc_vec[vecnum];
6111 #endif /* !illumos */
6113 zc->zc_iflags = flag & FKIOCTL;
6114 if (zc->zc_nvlist_src_size != 0) {
6115 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
6116 zc->zc_iflags, &innvl);
6121 /* rewrite innvl for backwards compatibility */
6123 innvl = zfs_ioctl_compat_innvl(zc, innvl, vecnum, cflag);
6126 * Ensure that all pool/dataset names are valid before we pass down to
6129 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
6130 switch (vec->zvec_namecheck) {
6132 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
6133 error = SET_ERROR(EINVAL);
6135 error = pool_status_check(zc->zc_name,
6136 vec->zvec_namecheck, vec->zvec_pool_check);
6140 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
6141 error = SET_ERROR(EINVAL);
6143 error = pool_status_check(zc->zc_name,
6144 vec->zvec_namecheck, vec->zvec_pool_check);
6151 if (error == 0 && !(flag & FKIOCTL))
6152 error = vec->zvec_secpolicy(zc, innvl, cr);
6157 /* legacy ioctls can modify zc_name */
6158 len = strcspn(zc->zc_name, "/@#") + 1;
6159 saved_poolname = kmem_alloc(len, KM_SLEEP);
6160 (void) strlcpy(saved_poolname, zc->zc_name, len);
6162 if (vec->zvec_func != NULL) {
6166 nvlist_t *lognv = NULL;
6168 ASSERT(vec->zvec_legacy_func == NULL);
6171 * Add the innvl to the lognv before calling the func,
6172 * in case the func changes the innvl.
6174 if (vec->zvec_allow_log) {
6175 lognv = fnvlist_alloc();
6176 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
6178 if (!nvlist_empty(innvl)) {
6179 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
6184 outnvl = fnvlist_alloc();
6185 error = vec->zvec_func(zc->zc_name, innvl, outnvl);
6187 if (error == 0 && vec->zvec_allow_log &&
6188 spa_open(zc->zc_name, &spa, FTAG) == 0) {
6189 if (!nvlist_empty(outnvl)) {
6190 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
6193 (void) spa_history_log_nvl(spa, lognv);
6194 spa_close(spa, FTAG);
6196 fnvlist_free(lognv);
6198 /* rewrite outnvl for backwards compatibility */
6200 outnvl = zfs_ioctl_compat_outnvl(zc, outnvl, vecnum,
6203 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
6205 if (vec->zvec_smush_outnvlist) {
6206 smusherror = nvlist_smush(outnvl,
6207 zc->zc_nvlist_dst_size);
6209 if (smusherror == 0)
6210 puterror = put_nvlist(zc, outnvl);
6216 nvlist_free(outnvl);
6218 error = vec->zvec_legacy_func(zc);
6225 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
6226 if (error == 0 && rc != 0)
6227 error = SET_ERROR(EFAULT);
6230 zfs_ioctl_compat_post(zc, cmd, cflag);
6232 ASSERT(compat_zc != NULL);
6233 ASSERT(sizeof(zfs_cmd_t) >= zc_iocparm->zfs_cmd_size);
6235 zfs_cmd_compat_put(zc, compat_zc, vecnum, cflag);
6236 rc = ddi_copyout(compat_zc,
6237 (void *)(uintptr_t)zc_iocparm->zfs_cmd,
6238 zc_iocparm->zfs_cmd_size, flag);
6239 if (error == 0 && rc != 0)
6240 error = SET_ERROR(EFAULT);
6241 kmem_free(compat_zc, sizeof (zfs_cmd_t));
6243 zfs_cmd_compat_put(zc, arg, vecnum, cflag);
6248 rc = ddi_copyout(zc, (void *)(uintptr_t)zc_iocparm->zfs_cmd,
6249 sizeof (zfs_cmd_t), flag);
6250 if (error == 0 && rc != 0)
6251 error = SET_ERROR(EFAULT);
6254 if (error == 0 && vec->zvec_allow_log) {
6255 char *s = tsd_get(zfs_allow_log_key);
6258 (void) tsd_set(zfs_allow_log_key, saved_poolname);
6260 if (saved_poolname != NULL)
6261 strfree(saved_poolname);
6264 kmem_free(zc, sizeof (zfs_cmd_t));
6270 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
6272 if (cmd != DDI_ATTACH)
6273 return (DDI_FAILURE);
6275 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
6276 DDI_PSEUDO, 0) == DDI_FAILURE)
6277 return (DDI_FAILURE);
6281 ddi_report_dev(dip);
6283 return (DDI_SUCCESS);
6287 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
6289 if (spa_busy() || zfs_busy() || zvol_busy())
6290 return (DDI_FAILURE);
6292 if (cmd != DDI_DETACH)
6293 return (DDI_FAILURE);
6297 ddi_prop_remove_all(dip);
6298 ddi_remove_minor_node(dip, NULL);
6300 return (DDI_SUCCESS);
6305 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
6308 case DDI_INFO_DEVT2DEVINFO:
6310 return (DDI_SUCCESS);
6312 case DDI_INFO_DEVT2INSTANCE:
6313 *result = (void *)0;
6314 return (DDI_SUCCESS);
6317 return (DDI_FAILURE);
6322 * OK, so this is a little weird.
6324 * /dev/zfs is the control node, i.e. minor 0.
6325 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
6327 * /dev/zfs has basically nothing to do except serve up ioctls,
6328 * so most of the standard driver entry points are in zvol.c.
6331 static struct cb_ops zfs_cb_ops = {
6332 zfsdev_open, /* open */
6333 zfsdev_close, /* close */
6334 zvol_strategy, /* strategy */
6336 zvol_dump, /* dump */
6337 zvol_read, /* read */
6338 zvol_write, /* write */
6339 zfsdev_ioctl, /* ioctl */
6343 nochpoll, /* poll */
6344 ddi_prop_op, /* prop_op */
6345 NULL, /* streamtab */
6346 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */
6347 CB_REV, /* version */
6348 nodev, /* async read */
6349 nodev, /* async write */
6352 static struct dev_ops zfs_dev_ops = {
6353 DEVO_REV, /* version */
6355 zfs_info, /* info */
6356 nulldev, /* identify */
6357 nulldev, /* probe */
6358 zfs_attach, /* attach */
6359 zfs_detach, /* detach */
6361 &zfs_cb_ops, /* driver operations */
6362 NULL, /* no bus operations */
6364 ddi_quiesce_not_needed, /* quiesce */
6367 static struct modldrv zfs_modldrv = {
6373 static struct modlinkage modlinkage = {
6375 (void *)&zfs_modlfs,
6376 (void *)&zfs_modldrv,
6381 static struct cdevsw zfs_cdevsw = {
6382 .d_version = D_VERSION,
6383 .d_open = zfsdev_open,
6384 .d_ioctl = zfsdev_ioctl,
6385 .d_name = ZFS_DEV_NAME
6389 zfs_allow_log_destroy(void *arg)
6391 char *poolname = arg;
6398 zfsdev = make_dev(&zfs_cdevsw, 0x0, UID_ROOT, GID_OPERATOR, 0666,
6406 destroy_dev(zfsdev);
6409 static struct root_hold_token *zfs_root_token;
6410 struct proc *zfsproc;
6418 spa_init(FREAD | FWRITE);
6423 if ((error = mod_install(&modlinkage)) != 0) {
6430 tsd_create(&zfs_fsyncer_key, NULL);
6431 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
6432 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
6434 error = ldi_ident_from_mod(&modlinkage, &zfs_li);
6436 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
6446 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
6447 return (SET_ERROR(EBUSY));
6449 if ((error = mod_remove(&modlinkage)) != 0)
6455 if (zfs_nfsshare_inited)
6456 (void) ddi_modclose(nfs_mod);
6457 if (zfs_smbshare_inited)
6458 (void) ddi_modclose(smbsrv_mod);
6459 if (zfs_nfsshare_inited || zfs_smbshare_inited)
6460 (void) ddi_modclose(sharefs_mod);
6462 tsd_destroy(&zfs_fsyncer_key);
6463 ldi_ident_release(zfs_li);
6465 mutex_destroy(&zfs_share_lock);
6471 _info(struct modinfo *modinfop)
6473 return (mod_info(&modlinkage, modinfop));
6477 static int zfs__init(void);
6478 static int zfs__fini(void);
6479 static void zfs_shutdown(void *, int);
6481 static eventhandler_tag zfs_shutdown_event_tag;
6487 zfs_root_token = root_mount_hold("ZFS");
6489 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
6491 spa_init(FREAD | FWRITE);
6496 tsd_create(&zfs_fsyncer_key, NULL);
6497 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
6498 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
6500 printf("ZFS storage pool version: features support (" SPA_VERSION_STRING ")\n");
6501 root_mount_rel(zfs_root_token);
6511 if (spa_busy() || zfs_busy() || zvol_busy() ||
6512 zio_injection_enabled) {
6521 tsd_destroy(&zfs_fsyncer_key);
6522 tsd_destroy(&rrw_tsd_key);
6523 tsd_destroy(&zfs_allow_log_key);
6525 mutex_destroy(&zfs_share_lock);
6531 zfs_shutdown(void *arg __unused, int howto __unused)
6535 * ZFS fini routines can not properly work in a panic-ed system.
6537 if (panicstr == NULL)
6543 zfs_modevent(module_t mod, int type, void *unused __unused)
6551 zfs_shutdown_event_tag = EVENTHANDLER_REGISTER(
6552 shutdown_post_sync, zfs_shutdown, NULL,
6553 SHUTDOWN_PRI_FIRST);
6557 if (err == 0 && zfs_shutdown_event_tag != NULL)
6558 EVENTHANDLER_DEREGISTER(shutdown_post_sync,
6559 zfs_shutdown_event_tag);
6566 return (EOPNOTSUPP);
6569 static moduledata_t zfs_mod = {
6574 DECLARE_MODULE(zfsctrl, zfs_mod, SI_SUB_VFS, SI_ORDER_ANY);
6575 MODULE_VERSION(zfsctrl, 1);
6576 MODULE_DEPEND(zfsctrl, opensolaris, 1, 1, 1);
6577 MODULE_DEPEND(zfsctrl, krpc, 1, 1, 1);
6578 MODULE_DEPEND(zfsctrl, acl_nfs4, 1, 1, 1);