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 2015, OmniTI Computer Consulting, Inc. All rights reserved.
29 * Copyright 2015 Nexenta Systems, Inc. All rights reserved.
30 * Copyright (c) 2014, Joyent, Inc. All rights reserved.
31 * Copyright (c) 2011, 2015 by Delphix. All rights reserved.
32 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
33 * Copyright (c) 2013 Steven Hartland. All rights reserved.
34 * Copyright (c) 2014 Integros [integros.com]
40 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
41 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
43 * There are two ways that we handle ioctls: the legacy way where almost
44 * all of the logic is in the ioctl callback, and the new way where most
45 * of the marshalling is handled in the common entry point, zfsdev_ioctl().
47 * Non-legacy ioctls should be registered by calling
48 * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked
49 * from userland by lzc_ioctl().
51 * The registration arguments are as follows:
54 * The name of the ioctl. This is used for history logging. If the
55 * ioctl returns successfully (the callback returns 0), and allow_log
56 * is true, then a history log entry will be recorded with the input &
57 * output nvlists. The log entry can be printed with "zpool history -i".
60 * The ioctl request number, which userland will pass to ioctl(2).
61 * The ioctl numbers can change from release to release, because
62 * the caller (libzfs) must be matched to the kernel.
64 * zfs_secpolicy_func_t *secpolicy
65 * This function will be called before the zfs_ioc_func_t, to
66 * determine if this operation is permitted. It should return EPERM
67 * on failure, and 0 on success. Checks include determining if the
68 * dataset is visible in this zone, and if the user has either all
69 * zfs privileges in the zone (SYS_MOUNT), or has been granted permission
70 * to do this operation on this dataset with "zfs allow".
72 * zfs_ioc_namecheck_t namecheck
73 * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
74 * name, a dataset name, or nothing. If the name is not well-formed,
75 * the ioctl will fail and the callback will not be called.
76 * Therefore, the callback can assume that the name is well-formed
77 * (e.g. is null-terminated, doesn't have more than one '@' character,
78 * doesn't have invalid characters).
80 * zfs_ioc_poolcheck_t pool_check
81 * This specifies requirements on the pool state. If the pool does
82 * not meet them (is suspended or is readonly), the ioctl will fail
83 * and the callback will not be called. If any checks are specified
84 * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
85 * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED |
86 * POOL_CHECK_READONLY).
88 * boolean_t smush_outnvlist
89 * If smush_outnvlist is true, then the output is presumed to be a
90 * list of errors, and it will be "smushed" down to fit into the
91 * caller's buffer, by removing some entries and replacing them with a
92 * single "N_MORE_ERRORS" entry indicating how many were removed. See
93 * nvlist_smush() for details. If smush_outnvlist is false, and the
94 * outnvlist does not fit into the userland-provided buffer, then the
95 * ioctl will fail with ENOMEM.
97 * zfs_ioc_func_t *func
98 * The callback function that will perform the operation.
100 * The callback should return 0 on success, or an error number on
101 * failure. If the function fails, the userland ioctl will return -1,
102 * and errno will be set to the callback's return value. The callback
103 * will be called with the following arguments:
106 * The name of the pool or dataset to operate on, from
107 * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the
108 * expected type (pool, dataset, or none).
111 * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or
112 * NULL if no input nvlist was provided. Changes to this nvlist are
113 * ignored. If the input nvlist could not be deserialized, the
114 * ioctl will fail and the callback will not be called.
117 * The output nvlist, initially empty. The callback can fill it in,
118 * and it will be returned to userland by serializing it into
119 * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization
120 * fails (e.g. because the caller didn't supply a large enough
121 * buffer), then the overall ioctl will fail. See the
122 * 'smush_nvlist' argument above for additional behaviors.
124 * There are two typical uses of the output nvlist:
125 * - To return state, e.g. property values. In this case,
126 * smush_outnvlist should be false. If the buffer was not large
127 * enough, the caller will reallocate a larger buffer and try
130 * - To return multiple errors from an ioctl which makes on-disk
131 * changes. In this case, smush_outnvlist should be true.
132 * Ioctls which make on-disk modifications should generally not
133 * use the outnvl if they succeed, because the caller can not
134 * distinguish between the operation failing, and
135 * deserialization failing.
138 #include "opt_kstack_pages.h"
141 #include <sys/types.h>
142 #include <sys/param.h>
143 #include <sys/systm.h>
144 #include <sys/conf.h>
145 #include <sys/kernel.h>
146 #include <sys/lock.h>
147 #include <sys/malloc.h>
148 #include <sys/mutex.h>
149 #include <sys/proc.h>
150 #include <sys/errno.h>
153 #include <sys/file.h>
154 #include <sys/kmem.h>
155 #include <sys/conf.h>
156 #include <sys/cmn_err.h>
157 #include <sys/stat.h>
158 #include <sys/zfs_ioctl.h>
159 #include <sys/zfs_vfsops.h>
160 #include <sys/zfs_znode.h>
163 #include <sys/spa_impl.h>
164 #include <sys/vdev.h>
166 #include <sys/dsl_dir.h>
167 #include <sys/dsl_dataset.h>
168 #include <sys/dsl_prop.h>
169 #include <sys/dsl_deleg.h>
170 #include <sys/dmu_objset.h>
171 #include <sys/dmu_impl.h>
172 #include <sys/dmu_tx.h>
173 #include <sys/sunddi.h>
174 #include <sys/policy.h>
175 #include <sys/zone.h>
176 #include <sys/nvpair.h>
177 #include <sys/mount.h>
178 #include <sys/taskqueue.h>
180 #include <sys/varargs.h>
181 #include <sys/fs/zfs.h>
182 #include <sys/zfs_ctldir.h>
183 #include <sys/zfs_dir.h>
184 #include <sys/zfs_onexit.h>
185 #include <sys/zvol.h>
186 #include <sys/dsl_scan.h>
187 #include <sys/dmu_objset.h>
188 #include <sys/dmu_send.h>
189 #include <sys/dsl_destroy.h>
190 #include <sys/dsl_bookmark.h>
191 #include <sys/dsl_userhold.h>
192 #include <sys/zfeature.h>
193 #include <sys/zio_checksum.h>
195 #include "zfs_namecheck.h"
196 #include "zfs_prop.h"
197 #include "zfs_deleg.h"
198 #include "zfs_comutil.h"
199 #include "zfs_ioctl_compat.h"
201 CTASSERT(sizeof(zfs_cmd_t) < IOCPARM_MAX);
203 static struct cdev *zfsdev;
205 extern void zfs_init(void);
206 extern void zfs_fini(void);
208 uint_t zfs_fsyncer_key;
209 extern uint_t rrw_tsd_key;
210 static uint_t zfs_allow_log_key;
212 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *);
213 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *);
214 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *);
220 } zfs_ioc_namecheck_t;
223 POOL_CHECK_NONE = 1 << 0,
224 POOL_CHECK_SUSPENDED = 1 << 1,
225 POOL_CHECK_READONLY = 1 << 2,
226 } zfs_ioc_poolcheck_t;
228 typedef struct zfs_ioc_vec {
229 zfs_ioc_legacy_func_t *zvec_legacy_func;
230 zfs_ioc_func_t *zvec_func;
231 zfs_secpolicy_func_t *zvec_secpolicy;
232 zfs_ioc_namecheck_t zvec_namecheck;
233 boolean_t zvec_allow_log;
234 zfs_ioc_poolcheck_t zvec_pool_check;
235 boolean_t zvec_smush_outnvlist;
236 const char *zvec_name;
239 /* This array is indexed by zfs_userquota_prop_t */
240 static const char *userquota_perms[] = {
241 ZFS_DELEG_PERM_USERUSED,
242 ZFS_DELEG_PERM_USERQUOTA,
243 ZFS_DELEG_PERM_GROUPUSED,
244 ZFS_DELEG_PERM_GROUPQUOTA,
247 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
248 static int zfs_check_settable(const char *name, nvpair_t *property,
250 static int zfs_check_clearable(char *dataset, nvlist_t *props,
252 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
254 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
255 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
257 static void zfsdev_close(void *data);
259 static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature);
261 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
263 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
270 * Get rid of annoying "../common/" prefix to filename.
272 newfile = strrchr(file, '/');
273 if (newfile != NULL) {
274 newfile = newfile + 1; /* Get rid of leading / */
280 (void) vsnprintf(buf, sizeof (buf), fmt, adx);
284 * To get this data, use the zfs-dprintf probe as so:
285 * dtrace -q -n 'zfs-dprintf \
286 * /stringof(arg0) == "dbuf.c"/ \
287 * {printf("%s: %s", stringof(arg1), stringof(arg3))}'
289 * arg1 = function name
293 DTRACE_PROBE4(zfs__dprintf,
294 char *, newfile, char *, func, int, line, char *, buf);
298 history_str_free(char *buf)
300 kmem_free(buf, HIS_MAX_RECORD_LEN);
304 history_str_get(zfs_cmd_t *zc)
308 if (zc->zc_history == 0)
311 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
312 if (copyinstr((void *)(uintptr_t)zc->zc_history,
313 buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
314 history_str_free(buf);
318 buf[HIS_MAX_RECORD_LEN -1] = '\0';
324 * Check to see if the named dataset is currently defined as bootable
327 zfs_is_bootfs(const char *name)
331 if (dmu_objset_hold(name, FTAG, &os) == 0) {
333 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
334 dmu_objset_rele(os, FTAG);
341 * Return non-zero if the spa version is less than requested version.
344 zfs_earlier_version(const char *name, int version)
348 if (spa_open(name, &spa, FTAG) == 0) {
349 if (spa_version(spa) < version) {
350 spa_close(spa, FTAG);
353 spa_close(spa, FTAG);
359 * Return TRUE if the ZPL version is less than requested version.
362 zpl_earlier_version(const char *name, int version)
365 boolean_t rc = B_TRUE;
367 if (dmu_objset_hold(name, FTAG, &os) == 0) {
370 if (dmu_objset_type(os) != DMU_OST_ZFS) {
371 dmu_objset_rele(os, FTAG);
374 /* XXX reading from non-owned objset */
375 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
376 rc = zplversion < version;
377 dmu_objset_rele(os, FTAG);
383 zfs_log_history(zfs_cmd_t *zc)
388 if ((buf = history_str_get(zc)) == NULL)
391 if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
392 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
393 (void) spa_history_log(spa, buf);
394 spa_close(spa, FTAG);
396 history_str_free(buf);
400 * Policy for top-level read operations (list pools). Requires no privileges,
401 * and can be used in the local zone, as there is no associated dataset.
405 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
411 * Policy for dataset read operations (list children, get statistics). Requires
412 * no privileges, but must be visible in the local zone.
416 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
418 if (INGLOBALZONE(curthread) ||
419 zone_dataset_visible(zc->zc_name, NULL))
422 return (SET_ERROR(ENOENT));
426 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
431 * The dataset must be visible by this zone -- check this first
432 * so they don't see EPERM on something they shouldn't know about.
434 if (!INGLOBALZONE(curthread) &&
435 !zone_dataset_visible(dataset, &writable))
436 return (SET_ERROR(ENOENT));
438 if (INGLOBALZONE(curthread)) {
440 * If the fs is zoned, only root can access it from the
443 if (secpolicy_zfs(cr) && zoned)
444 return (SET_ERROR(EPERM));
447 * If we are in a local zone, the 'zoned' property must be set.
450 return (SET_ERROR(EPERM));
452 /* must be writable by this zone */
454 return (SET_ERROR(EPERM));
460 zfs_dozonecheck(const char *dataset, cred_t *cr)
464 if (dsl_prop_get_integer(dataset, "jailed", &zoned, NULL))
465 return (SET_ERROR(ENOENT));
467 return (zfs_dozonecheck_impl(dataset, zoned, cr));
471 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
475 if (dsl_prop_get_int_ds(ds, "jailed", &zoned))
476 return (SET_ERROR(ENOENT));
478 return (zfs_dozonecheck_impl(dataset, zoned, cr));
482 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
483 const char *perm, cred_t *cr)
487 error = zfs_dozonecheck_ds(name, ds, cr);
489 error = secpolicy_zfs(cr);
491 error = dsl_deleg_access_impl(ds, perm, cr);
497 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
503 error = dsl_pool_hold(name, FTAG, &dp);
507 error = dsl_dataset_hold(dp, name, FTAG, &ds);
509 dsl_pool_rele(dp, FTAG);
513 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
515 dsl_dataset_rele(ds, FTAG);
516 dsl_pool_rele(dp, FTAG);
522 * Policy for setting the security label property.
524 * Returns 0 for success, non-zero for access and other errors.
527 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
529 char ds_hexsl[MAXNAMELEN];
530 bslabel_t ds_sl, new_sl;
531 boolean_t new_default = FALSE;
533 int needed_priv = -1;
536 /* First get the existing dataset label. */
537 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
538 1, sizeof (ds_hexsl), &ds_hexsl, NULL);
540 return (SET_ERROR(EPERM));
542 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
545 /* The label must be translatable */
546 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
547 return (SET_ERROR(EINVAL));
550 * In a non-global zone, disallow attempts to set a label that
551 * doesn't match that of the zone; otherwise no other checks
554 if (!INGLOBALZONE(curproc)) {
555 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
556 return (SET_ERROR(EPERM));
561 * For global-zone datasets (i.e., those whose zoned property is
562 * "off", verify that the specified new label is valid for the
565 if (dsl_prop_get_integer(name,
566 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
567 return (SET_ERROR(EPERM));
569 if (zfs_check_global_label(name, strval) != 0)
570 return (SET_ERROR(EPERM));
574 * If the existing dataset label is nondefault, check if the
575 * dataset is mounted (label cannot be changed while mounted).
576 * Get the zfsvfs; if there isn't one, then the dataset isn't
577 * mounted (or isn't a dataset, doesn't exist, ...).
579 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
581 static char *setsl_tag = "setsl_tag";
584 * Try to own the dataset; abort if there is any error,
585 * (e.g., already mounted, in use, or other error).
587 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
590 return (SET_ERROR(EPERM));
592 dmu_objset_disown(os, setsl_tag);
595 needed_priv = PRIV_FILE_DOWNGRADE_SL;
599 if (hexstr_to_label(strval, &new_sl) != 0)
600 return (SET_ERROR(EPERM));
602 if (blstrictdom(&ds_sl, &new_sl))
603 needed_priv = PRIV_FILE_DOWNGRADE_SL;
604 else if (blstrictdom(&new_sl, &ds_sl))
605 needed_priv = PRIV_FILE_UPGRADE_SL;
607 /* dataset currently has a default label */
609 needed_priv = PRIV_FILE_UPGRADE_SL;
613 if (needed_priv != -1)
614 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
617 #endif /* SECLABEL */
620 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
626 * Check permissions for special properties.
631 * Disallow setting of 'zoned' from within a local zone.
633 if (!INGLOBALZONE(curthread))
634 return (SET_ERROR(EPERM));
638 case ZFS_PROP_FILESYSTEM_LIMIT:
639 case ZFS_PROP_SNAPSHOT_LIMIT:
640 if (!INGLOBALZONE(curthread)) {
642 char setpoint[MAXNAMELEN];
644 * Unprivileged users are allowed to modify the
645 * limit on things *under* (ie. contained by)
646 * the thing they own.
648 if (dsl_prop_get_integer(dsname, "jailed", &zoned,
650 return (SET_ERROR(EPERM));
651 if (!zoned || strlen(dsname) <= strlen(setpoint))
652 return (SET_ERROR(EPERM));
656 case ZFS_PROP_MLSLABEL:
658 if (!is_system_labeled())
659 return (SET_ERROR(EPERM));
661 if (nvpair_value_string(propval, &strval) == 0) {
664 err = zfs_set_slabel_policy(dsname, strval, CRED());
674 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
679 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
683 error = zfs_dozonecheck(zc->zc_name, cr);
688 * permission to set permissions will be evaluated later in
689 * dsl_deleg_can_allow()
696 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
698 return (zfs_secpolicy_write_perms(zc->zc_name,
699 ZFS_DELEG_PERM_ROLLBACK, cr));
704 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
712 * Generate the current snapshot name from the given objsetid, then
713 * use that name for the secpolicy/zone checks.
715 cp = strchr(zc->zc_name, '@');
717 return (SET_ERROR(EINVAL));
718 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
722 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
724 dsl_pool_rele(dp, FTAG);
728 dsl_dataset_name(ds, zc->zc_name);
730 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
731 ZFS_DELEG_PERM_SEND, cr);
732 dsl_dataset_rele(ds, FTAG);
733 dsl_pool_rele(dp, FTAG);
740 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
742 return (zfs_secpolicy_write_perms(zc->zc_name,
743 ZFS_DELEG_PERM_SEND, cr));
748 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
753 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
754 NO_FOLLOW, NULL, &vp)) != 0)
757 /* Now make sure mntpnt and dataset are ZFS */
759 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 ||
760 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
761 zc->zc_name) != 0)) {
763 return (SET_ERROR(EPERM));
767 return (dsl_deleg_access(zc->zc_name,
768 ZFS_DELEG_PERM_SHARE, cr));
772 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
774 if (!INGLOBALZONE(curthread))
775 return (SET_ERROR(EPERM));
777 if (secpolicy_nfs(cr) == 0) {
780 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
785 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
787 if (!INGLOBALZONE(curthread))
788 return (SET_ERROR(EPERM));
790 if (secpolicy_smb(cr) == 0) {
793 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
798 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
803 * Remove the @bla or /bla from the end of the name to get the parent.
805 (void) strncpy(parent, datasetname, parentsize);
806 cp = strrchr(parent, '@');
810 cp = strrchr(parent, '/');
812 return (SET_ERROR(ENOENT));
820 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
824 if ((error = zfs_secpolicy_write_perms(name,
825 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
828 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
833 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
835 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
839 * Destroying snapshots with delegated permissions requires
840 * descendant mount and destroy permissions.
844 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
847 nvpair_t *pair, *nextpair;
850 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
851 return (SET_ERROR(EINVAL));
852 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
854 nextpair = nvlist_next_nvpair(snaps, pair);
855 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr);
856 if (error == ENOENT) {
858 * Ignore any snapshots that don't exist (we consider
859 * them "already destroyed"). Remove the name from the
860 * nvl here in case the snapshot is created between
861 * now and when we try to destroy it (in which case
862 * we don't want to destroy it since we haven't
863 * checked for permission).
865 fnvlist_remove_nvpair(snaps, pair);
876 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
878 char parentname[MAXNAMELEN];
881 if ((error = zfs_secpolicy_write_perms(from,
882 ZFS_DELEG_PERM_RENAME, cr)) != 0)
885 if ((error = zfs_secpolicy_write_perms(from,
886 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
889 if ((error = zfs_get_parent(to, parentname,
890 sizeof (parentname))) != 0)
893 if ((error = zfs_secpolicy_write_perms(parentname,
894 ZFS_DELEG_PERM_CREATE, cr)) != 0)
897 if ((error = zfs_secpolicy_write_perms(parentname,
898 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
906 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
911 if ((zc->zc_cookie & 1) != 0) {
913 * This is recursive rename, so the starting snapshot might
914 * not exist. Check file system or volume permission instead.
916 at = strchr(zc->zc_name, '@');
922 error = zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr);
932 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
935 dsl_dataset_t *clone;
938 error = zfs_secpolicy_write_perms(zc->zc_name,
939 ZFS_DELEG_PERM_PROMOTE, cr);
943 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
947 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
950 char parentname[MAXNAMELEN];
951 dsl_dataset_t *origin = NULL;
955 error = dsl_dataset_hold_obj(dd->dd_pool,
956 dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin);
958 dsl_dataset_rele(clone, FTAG);
959 dsl_pool_rele(dp, FTAG);
963 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
964 ZFS_DELEG_PERM_MOUNT, cr);
966 dsl_dataset_name(origin, parentname);
968 error = zfs_secpolicy_write_perms_ds(parentname, origin,
969 ZFS_DELEG_PERM_PROMOTE, cr);
971 dsl_dataset_rele(clone, FTAG);
972 dsl_dataset_rele(origin, FTAG);
974 dsl_pool_rele(dp, FTAG);
980 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
984 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
985 ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
988 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
989 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
992 return (zfs_secpolicy_write_perms(zc->zc_name,
993 ZFS_DELEG_PERM_CREATE, cr));
997 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
999 return (zfs_secpolicy_write_perms(name,
1000 ZFS_DELEG_PERM_SNAPSHOT, cr));
1004 * Check for permission to create each snapshot in the nvlist.
1008 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1014 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
1015 return (SET_ERROR(EINVAL));
1016 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
1017 pair = nvlist_next_nvpair(snaps, pair)) {
1018 char *name = nvpair_name(pair);
1019 char *atp = strchr(name, '@');
1022 error = SET_ERROR(EINVAL);
1026 error = zfs_secpolicy_snapshot_perms(name, cr);
1035 * Check for permission to create each snapshot in the nvlist.
1039 zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1043 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
1044 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
1045 char *name = nvpair_name(pair);
1046 char *hashp = strchr(name, '#');
1048 if (hashp == NULL) {
1049 error = SET_ERROR(EINVAL);
1053 error = zfs_secpolicy_write_perms(name,
1054 ZFS_DELEG_PERM_BOOKMARK, cr);
1064 zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1066 nvpair_t *pair, *nextpair;
1069 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1071 char *name = nvpair_name(pair);
1072 char *hashp = strchr(name, '#');
1073 nextpair = nvlist_next_nvpair(innvl, pair);
1075 if (hashp == NULL) {
1076 error = SET_ERROR(EINVAL);
1081 error = zfs_secpolicy_write_perms(name,
1082 ZFS_DELEG_PERM_DESTROY, cr);
1084 if (error == ENOENT) {
1086 * Ignore any filesystems that don't exist (we consider
1087 * their bookmarks "already destroyed"). Remove
1088 * the name from the nvl here in case the filesystem
1089 * is created between now and when we try to destroy
1090 * the bookmark (in which case we don't want to
1091 * destroy it since we haven't checked for permission).
1093 fnvlist_remove_nvpair(innvl, pair);
1105 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1108 * Even root must have a proper TSD so that we know what pool
1111 if (tsd_get(zfs_allow_log_key) == NULL)
1112 return (SET_ERROR(EPERM));
1117 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1119 char parentname[MAXNAMELEN];
1123 if ((error = zfs_get_parent(zc->zc_name, parentname,
1124 sizeof (parentname))) != 0)
1127 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1128 (error = zfs_secpolicy_write_perms(origin,
1129 ZFS_DELEG_PERM_CLONE, cr)) != 0)
1132 if ((error = zfs_secpolicy_write_perms(parentname,
1133 ZFS_DELEG_PERM_CREATE, cr)) != 0)
1136 return (zfs_secpolicy_write_perms(parentname,
1137 ZFS_DELEG_PERM_MOUNT, cr));
1141 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires
1142 * SYS_CONFIG privilege, which is not available in a local zone.
1146 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1148 if (secpolicy_sys_config(cr, B_FALSE) != 0)
1149 return (SET_ERROR(EPERM));
1155 * Policy for object to name lookups.
1159 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1163 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1166 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1171 * Policy for fault injection. Requires all privileges.
1175 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1177 return (secpolicy_zinject(cr));
1182 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1184 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1186 if (prop == ZPROP_INVAL) {
1187 if (!zfs_prop_user(zc->zc_value))
1188 return (SET_ERROR(EINVAL));
1189 return (zfs_secpolicy_write_perms(zc->zc_name,
1190 ZFS_DELEG_PERM_USERPROP, cr));
1192 return (zfs_secpolicy_setprop(zc->zc_name, prop,
1198 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1200 int err = zfs_secpolicy_read(zc, innvl, cr);
1204 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1205 return (SET_ERROR(EINVAL));
1207 if (zc->zc_value[0] == 0) {
1209 * They are asking about a posix uid/gid. If it's
1210 * themself, allow it.
1212 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1213 zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
1214 if (zc->zc_guid == crgetuid(cr))
1217 if (groupmember(zc->zc_guid, cr))
1222 return (zfs_secpolicy_write_perms(zc->zc_name,
1223 userquota_perms[zc->zc_objset_type], cr));
1227 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1229 int err = zfs_secpolicy_read(zc, innvl, cr);
1233 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1234 return (SET_ERROR(EINVAL));
1236 return (zfs_secpolicy_write_perms(zc->zc_name,
1237 userquota_perms[zc->zc_objset_type], cr));
1242 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1244 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1250 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1256 error = nvlist_lookup_nvlist(innvl, "holds", &holds);
1258 return (SET_ERROR(EINVAL));
1260 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
1261 pair = nvlist_next_nvpair(holds, pair)) {
1262 char fsname[MAXNAMELEN];
1263 error = dmu_fsname(nvpair_name(pair), fsname);
1266 error = zfs_secpolicy_write_perms(fsname,
1267 ZFS_DELEG_PERM_HOLD, cr);
1276 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1281 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1282 pair = nvlist_next_nvpair(innvl, pair)) {
1283 char fsname[MAXNAMELEN];
1284 error = dmu_fsname(nvpair_name(pair), fsname);
1287 error = zfs_secpolicy_write_perms(fsname,
1288 ZFS_DELEG_PERM_RELEASE, cr);
1296 * Policy for allowing temporary snapshots to be taken or released
1299 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1302 * A temporary snapshot is the same as a snapshot,
1303 * hold, destroy and release all rolled into one.
1304 * Delegated diff alone is sufficient that we allow this.
1308 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1309 ZFS_DELEG_PERM_DIFF, cr)) == 0)
1312 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1314 error = zfs_secpolicy_hold(zc, innvl, cr);
1316 error = zfs_secpolicy_release(zc, innvl, cr);
1318 error = zfs_secpolicy_destroy(zc, innvl, cr);
1323 * Returns the nvlist as specified by the user in the zfs_cmd_t.
1326 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1330 nvlist_t *list = NULL;
1333 * Read in and unpack the user-supplied nvlist.
1336 return (SET_ERROR(EINVAL));
1338 packed = kmem_alloc(size, KM_SLEEP);
1340 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1342 kmem_free(packed, size);
1343 return (SET_ERROR(EFAULT));
1346 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1347 kmem_free(packed, size);
1351 kmem_free(packed, size);
1358 * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1359 * Entries will be removed from the end of the nvlist, and one int32 entry
1360 * named "N_MORE_ERRORS" will be added indicating how many entries were
1364 nvlist_smush(nvlist_t *errors, size_t max)
1368 size = fnvlist_size(errors);
1371 nvpair_t *more_errors;
1375 return (SET_ERROR(ENOMEM));
1377 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1378 more_errors = nvlist_prev_nvpair(errors, NULL);
1381 nvpair_t *pair = nvlist_prev_nvpair(errors,
1383 fnvlist_remove_nvpair(errors, pair);
1385 size = fnvlist_size(errors);
1386 } while (size > max);
1388 fnvlist_remove_nvpair(errors, more_errors);
1389 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1390 ASSERT3U(fnvlist_size(errors), <=, max);
1397 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1399 char *packed = NULL;
1403 size = fnvlist_size(nvl);
1405 if (size > zc->zc_nvlist_dst_size) {
1407 * Solaris returns ENOMEM here, because even if an error is
1408 * returned from an ioctl(2), new zc_nvlist_dst_size will be
1409 * passed to the userland. This is not the case for FreeBSD.
1410 * We need to return 0, so the kernel will copy the
1411 * zc_nvlist_dst_size back and the userland can discover that a
1412 * bigger buffer is needed.
1416 packed = fnvlist_pack(nvl, &size);
1417 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1418 size, zc->zc_iflags) != 0)
1419 error = SET_ERROR(EFAULT);
1420 fnvlist_pack_free(packed, size);
1423 zc->zc_nvlist_dst_size = size;
1424 zc->zc_nvlist_dst_filled = B_TRUE;
1429 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1434 error = dmu_objset_hold(dsname, FTAG, &os);
1437 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1438 dmu_objset_rele(os, FTAG);
1439 return (SET_ERROR(EINVAL));
1442 mutex_enter(&os->os_user_ptr_lock);
1443 *zfvp = dmu_objset_get_user(os);
1446 VFS_HOLD((*zfvp)->z_vfs);
1448 if (vfs_busy((*zfvp)->z_vfs, 0) != 0) {
1450 error = SET_ERROR(ESRCH);
1454 error = SET_ERROR(ESRCH);
1456 mutex_exit(&os->os_user_ptr_lock);
1457 dmu_objset_rele(os, FTAG);
1462 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1463 * case its z_vfs will be NULL, and it will be opened as the owner.
1464 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1465 * which prevents all vnode ops from running.
1468 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1472 if (getzfsvfs(name, zfvp) != 0)
1473 error = zfsvfs_create(name, zfvp);
1475 rrm_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER :
1477 if ((*zfvp)->z_unmounted) {
1479 * XXX we could probably try again, since the unmounting
1480 * thread should be just about to disassociate the
1481 * objset from the zfsvfs.
1483 rrm_exit(&(*zfvp)->z_teardown_lock, tag);
1484 return (SET_ERROR(EBUSY));
1491 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1493 rrm_exit(&zfsvfs->z_teardown_lock, tag);
1495 if (zfsvfs->z_vfs) {
1497 VFS_RELE(zfsvfs->z_vfs);
1499 vfs_unbusy(zfsvfs->z_vfs);
1502 dmu_objset_disown(zfsvfs->z_os, zfsvfs);
1503 zfsvfs_free(zfsvfs);
1508 zfs_ioc_pool_create(zfs_cmd_t *zc)
1511 nvlist_t *config, *props = NULL;
1512 nvlist_t *rootprops = NULL;
1513 nvlist_t *zplprops = NULL;
1515 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1516 zc->zc_iflags, &config))
1519 if (zc->zc_nvlist_src_size != 0 && (error =
1520 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1521 zc->zc_iflags, &props))) {
1522 nvlist_free(config);
1527 nvlist_t *nvl = NULL;
1528 uint64_t version = SPA_VERSION;
1530 (void) nvlist_lookup_uint64(props,
1531 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1532 if (!SPA_VERSION_IS_SUPPORTED(version)) {
1533 error = SET_ERROR(EINVAL);
1534 goto pool_props_bad;
1536 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1538 error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1540 nvlist_free(config);
1544 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1546 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1547 error = zfs_fill_zplprops_root(version, rootprops,
1550 goto pool_props_bad;
1553 error = spa_create(zc->zc_name, config, props, zplprops);
1556 * Set the remaining root properties
1558 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1559 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1560 (void) spa_destroy(zc->zc_name);
1563 nvlist_free(rootprops);
1564 nvlist_free(zplprops);
1565 nvlist_free(config);
1572 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1575 zfs_log_history(zc);
1576 error = spa_destroy(zc->zc_name);
1578 zvol_remove_minors(zc->zc_name);
1583 zfs_ioc_pool_import(zfs_cmd_t *zc)
1585 nvlist_t *config, *props = NULL;
1589 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1590 zc->zc_iflags, &config)) != 0)
1593 if (zc->zc_nvlist_src_size != 0 && (error =
1594 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1595 zc->zc_iflags, &props))) {
1596 nvlist_free(config);
1600 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1601 guid != zc->zc_guid)
1602 error = SET_ERROR(EINVAL);
1604 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1606 if (zc->zc_nvlist_dst != 0) {
1609 if ((err = put_nvlist(zc, config)) != 0)
1613 nvlist_free(config);
1621 zfs_ioc_pool_export(zfs_cmd_t *zc)
1624 boolean_t force = (boolean_t)zc->zc_cookie;
1625 boolean_t hardforce = (boolean_t)zc->zc_guid;
1627 zfs_log_history(zc);
1628 error = spa_export(zc->zc_name, NULL, force, hardforce);
1630 zvol_remove_minors(zc->zc_name);
1635 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1640 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1641 return (SET_ERROR(EEXIST));
1643 error = put_nvlist(zc, configs);
1645 nvlist_free(configs);
1652 * zc_name name of the pool
1655 * zc_cookie real errno
1656 * zc_nvlist_dst config nvlist
1657 * zc_nvlist_dst_size size of config nvlist
1660 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1666 error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1667 sizeof (zc->zc_value));
1669 if (config != NULL) {
1670 ret = put_nvlist(zc, config);
1671 nvlist_free(config);
1674 * The config may be present even if 'error' is non-zero.
1675 * In this case we return success, and preserve the real errno
1678 zc->zc_cookie = error;
1687 * Try to import the given pool, returning pool stats as appropriate so that
1688 * user land knows which devices are available and overall pool health.
1691 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1693 nvlist_t *tryconfig, *config;
1696 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1697 zc->zc_iflags, &tryconfig)) != 0)
1700 config = spa_tryimport(tryconfig);
1702 nvlist_free(tryconfig);
1705 return (SET_ERROR(EINVAL));
1707 error = put_nvlist(zc, config);
1708 nvlist_free(config);
1715 * zc_name name of the pool
1716 * zc_cookie scan func (pool_scan_func_t)
1719 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1724 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1727 if (zc->zc_cookie == POOL_SCAN_NONE)
1728 error = spa_scan_stop(spa);
1730 error = spa_scan(spa, zc->zc_cookie);
1732 spa_close(spa, FTAG);
1738 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1743 error = spa_open(zc->zc_name, &spa, FTAG);
1746 spa_close(spa, FTAG);
1752 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1757 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1760 if (zc->zc_cookie < spa_version(spa) ||
1761 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1762 spa_close(spa, FTAG);
1763 return (SET_ERROR(EINVAL));
1766 spa_upgrade(spa, zc->zc_cookie);
1767 spa_close(spa, FTAG);
1773 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1780 if ((size = zc->zc_history_len) == 0)
1781 return (SET_ERROR(EINVAL));
1783 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1786 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1787 spa_close(spa, FTAG);
1788 return (SET_ERROR(ENOTSUP));
1791 hist_buf = kmem_alloc(size, KM_SLEEP);
1792 if ((error = spa_history_get(spa, &zc->zc_history_offset,
1793 &zc->zc_history_len, hist_buf)) == 0) {
1794 error = ddi_copyout(hist_buf,
1795 (void *)(uintptr_t)zc->zc_history,
1796 zc->zc_history_len, zc->zc_iflags);
1799 spa_close(spa, FTAG);
1800 kmem_free(hist_buf, size);
1805 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1810 error = spa_open(zc->zc_name, &spa, FTAG);
1812 error = spa_change_guid(spa);
1813 spa_close(spa, FTAG);
1819 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1821 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1826 * zc_name name of filesystem
1827 * zc_obj object to find
1830 * zc_value name of object
1833 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1838 /* XXX reading from objset not owned */
1839 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1841 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1842 dmu_objset_rele(os, FTAG);
1843 return (SET_ERROR(EINVAL));
1845 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1846 sizeof (zc->zc_value));
1847 dmu_objset_rele(os, FTAG);
1854 * zc_name name of filesystem
1855 * zc_obj object to find
1858 * zc_stat stats on object
1859 * zc_value path to object
1862 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1867 /* XXX reading from objset not owned */
1868 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1870 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1871 dmu_objset_rele(os, FTAG);
1872 return (SET_ERROR(EINVAL));
1874 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1875 sizeof (zc->zc_value));
1876 dmu_objset_rele(os, FTAG);
1882 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1886 nvlist_t *config, **l2cache, **spares;
1887 uint_t nl2cache = 0, nspares = 0;
1889 error = spa_open(zc->zc_name, &spa, FTAG);
1893 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1894 zc->zc_iflags, &config);
1895 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1896 &l2cache, &nl2cache);
1898 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1903 * A root pool with concatenated devices is not supported.
1904 * Thus, can not add a device to a root pool.
1906 * Intent log device can not be added to a rootpool because
1907 * during mountroot, zil is replayed, a seperated log device
1908 * can not be accessed during the mountroot time.
1910 * l2cache and spare devices are ok to be added to a rootpool.
1912 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1913 nvlist_free(config);
1914 spa_close(spa, FTAG);
1915 return (SET_ERROR(EDOM));
1917 #endif /* illumos */
1920 error = spa_vdev_add(spa, config);
1921 nvlist_free(config);
1923 spa_close(spa, FTAG);
1929 * zc_name name of the pool
1930 * zc_nvlist_conf nvlist of devices to remove
1931 * zc_cookie to stop the remove?
1934 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1939 error = spa_open(zc->zc_name, &spa, FTAG);
1942 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1943 spa_close(spa, FTAG);
1948 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1952 vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1954 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1956 switch (zc->zc_cookie) {
1957 case VDEV_STATE_ONLINE:
1958 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1961 case VDEV_STATE_OFFLINE:
1962 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1965 case VDEV_STATE_FAULTED:
1966 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1967 zc->zc_obj != VDEV_AUX_EXTERNAL)
1968 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1970 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1973 case VDEV_STATE_DEGRADED:
1974 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1975 zc->zc_obj != VDEV_AUX_EXTERNAL)
1976 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1978 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1982 error = SET_ERROR(EINVAL);
1984 zc->zc_cookie = newstate;
1985 spa_close(spa, FTAG);
1990 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1993 int replacing = zc->zc_cookie;
1997 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2000 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
2001 zc->zc_iflags, &config)) == 0) {
2002 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
2003 nvlist_free(config);
2006 spa_close(spa, FTAG);
2011 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
2016 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2019 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
2021 spa_close(spa, FTAG);
2026 zfs_ioc_vdev_split(zfs_cmd_t *zc)
2029 nvlist_t *config, *props = NULL;
2031 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
2033 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2036 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
2037 zc->zc_iflags, &config)) {
2038 spa_close(spa, FTAG);
2042 if (zc->zc_nvlist_src_size != 0 && (error =
2043 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2044 zc->zc_iflags, &props))) {
2045 spa_close(spa, FTAG);
2046 nvlist_free(config);
2050 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
2052 spa_close(spa, FTAG);
2054 nvlist_free(config);
2061 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
2064 char *path = zc->zc_value;
2065 uint64_t guid = zc->zc_guid;
2068 error = spa_open(zc->zc_name, &spa, FTAG);
2072 error = spa_vdev_setpath(spa, guid, path);
2073 spa_close(spa, FTAG);
2078 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
2081 char *fru = zc->zc_value;
2082 uint64_t guid = zc->zc_guid;
2085 error = spa_open(zc->zc_name, &spa, FTAG);
2089 error = spa_vdev_setfru(spa, guid, fru);
2090 spa_close(spa, FTAG);
2095 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
2100 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2102 if (zc->zc_nvlist_dst != 0 &&
2103 (error = dsl_prop_get_all(os, &nv)) == 0) {
2104 dmu_objset_stats(os, nv);
2106 * NB: zvol_get_stats() will read the objset contents,
2107 * which we aren't supposed to do with a
2108 * DS_MODE_USER hold, because it could be
2109 * inconsistent. So this is a bit of a workaround...
2110 * XXX reading with out owning
2112 if (!zc->zc_objset_stats.dds_inconsistent &&
2113 dmu_objset_type(os) == DMU_OST_ZVOL) {
2114 error = zvol_get_stats(os, nv);
2119 error = put_nvlist(zc, nv);
2128 * zc_name name of filesystem
2129 * zc_nvlist_dst_size size of buffer for property nvlist
2132 * zc_objset_stats stats
2133 * zc_nvlist_dst property nvlist
2134 * zc_nvlist_dst_size size of property nvlist
2137 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2142 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2144 error = zfs_ioc_objset_stats_impl(zc, os);
2145 dmu_objset_rele(os, FTAG);
2148 if (error == ENOMEM)
2155 * zc_name name of filesystem
2156 * zc_nvlist_dst_size size of buffer for property nvlist
2159 * zc_nvlist_dst received property nvlist
2160 * zc_nvlist_dst_size size of received property nvlist
2162 * Gets received properties (distinct from local properties on or after
2163 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2164 * local property values.
2167 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2173 * Without this check, we would return local property values if the
2174 * caller has not already received properties on or after
2175 * SPA_VERSION_RECVD_PROPS.
2177 if (!dsl_prop_get_hasrecvd(zc->zc_name))
2178 return (SET_ERROR(ENOTSUP));
2180 if (zc->zc_nvlist_dst != 0 &&
2181 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2182 error = put_nvlist(zc, nv);
2190 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2196 * zfs_get_zplprop() will either find a value or give us
2197 * the default value (if there is one).
2199 if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2201 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2207 * zc_name name of filesystem
2208 * zc_nvlist_dst_size size of buffer for zpl property nvlist
2211 * zc_nvlist_dst zpl property nvlist
2212 * zc_nvlist_dst_size size of zpl property nvlist
2215 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2220 /* XXX reading without owning */
2221 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os))
2224 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2227 * NB: nvl_add_zplprop() will read the objset contents,
2228 * which we aren't supposed to do with a DS_MODE_USER
2229 * hold, because it could be inconsistent.
2231 if (zc->zc_nvlist_dst != 0 &&
2232 !zc->zc_objset_stats.dds_inconsistent &&
2233 dmu_objset_type(os) == DMU_OST_ZFS) {
2236 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2237 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2238 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2239 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2240 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2241 err = put_nvlist(zc, nv);
2244 err = SET_ERROR(ENOENT);
2246 dmu_objset_rele(os, FTAG);
2251 dataset_name_hidden(const char *name)
2254 * Skip over datasets that are not visible in this zone,
2255 * internal datasets (which have a $ in their name), and
2256 * temporary datasets (which have a % in their name).
2258 if (strchr(name, '$') != NULL)
2260 if (strchr(name, '%') != NULL)
2262 if (!INGLOBALZONE(curthread) && !zone_dataset_visible(name, NULL))
2269 * zc_name name of filesystem
2270 * zc_cookie zap cursor
2271 * zc_nvlist_dst_size size of buffer for property nvlist
2274 * zc_name name of next filesystem
2275 * zc_cookie zap cursor
2276 * zc_objset_stats stats
2277 * zc_nvlist_dst property nvlist
2278 * zc_nvlist_dst_size size of property nvlist
2281 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2286 size_t orig_len = strlen(zc->zc_name);
2289 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) {
2290 if (error == ENOENT)
2291 error = SET_ERROR(ESRCH);
2295 p = strrchr(zc->zc_name, '/');
2296 if (p == NULL || p[1] != '\0')
2297 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2298 p = zc->zc_name + strlen(zc->zc_name);
2301 error = dmu_dir_list_next(os,
2302 sizeof (zc->zc_name) - (p - zc->zc_name), p,
2303 NULL, &zc->zc_cookie);
2304 if (error == ENOENT)
2305 error = SET_ERROR(ESRCH);
2306 } while (error == 0 && dataset_name_hidden(zc->zc_name));
2307 dmu_objset_rele(os, FTAG);
2310 * If it's an internal dataset (ie. with a '$' in its name),
2311 * don't try to get stats for it, otherwise we'll return ENOENT.
2313 if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2314 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2315 if (error == ENOENT) {
2316 /* We lost a race with destroy, get the next one. */
2317 zc->zc_name[orig_len] = '\0';
2326 * zc_name name of filesystem
2327 * zc_cookie zap cursor
2328 * zc_nvlist_dst_size size of buffer for property nvlist
2329 * zc_simple when set, only name is requested
2332 * zc_name name of next snapshot
2333 * zc_objset_stats stats
2334 * zc_nvlist_dst property nvlist
2335 * zc_nvlist_dst_size size of property nvlist
2338 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2343 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2345 return (error == ENOENT ? ESRCH : error);
2349 * A dataset name of maximum length cannot have any snapshots,
2350 * so exit immediately.
2352 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
2353 dmu_objset_rele(os, FTAG);
2354 return (SET_ERROR(ESRCH));
2357 error = dmu_snapshot_list_next(os,
2358 sizeof (zc->zc_name) - strlen(zc->zc_name),
2359 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2362 if (error == 0 && !zc->zc_simple) {
2364 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2366 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2370 error = dmu_objset_from_ds(ds, &ossnap);
2372 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2373 dsl_dataset_rele(ds, FTAG);
2375 } else if (error == ENOENT) {
2376 error = SET_ERROR(ESRCH);
2379 dmu_objset_rele(os, FTAG);
2380 /* if we failed, undo the @ that we tacked on to zc_name */
2382 *strchr(zc->zc_name, '@') = '\0';
2387 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2389 const char *propname = nvpair_name(pair);
2391 unsigned int vallen;
2394 zfs_userquota_prop_t type;
2400 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2402 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2403 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2405 return (SET_ERROR(EINVAL));
2409 * A correctly constructed propname is encoded as
2410 * userquota@<rid>-<domain>.
2412 if ((dash = strchr(propname, '-')) == NULL ||
2413 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2415 return (SET_ERROR(EINVAL));
2422 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2424 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2425 zfsvfs_rele(zfsvfs, FTAG);
2432 * If the named property is one that has a special function to set its value,
2433 * return 0 on success and a positive error code on failure; otherwise if it is
2434 * not one of the special properties handled by this function, return -1.
2436 * XXX: It would be better for callers of the property interface if we handled
2437 * these special cases in dsl_prop.c (in the dsl layer).
2440 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2443 const char *propname = nvpair_name(pair);
2444 zfs_prop_t prop = zfs_name_to_prop(propname);
2448 if (prop == ZPROP_INVAL) {
2449 if (zfs_prop_userquota(propname))
2450 return (zfs_prop_set_userquota(dsname, pair));
2454 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2456 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2457 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2461 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2464 VERIFY(0 == nvpair_value_uint64(pair, &intval));
2467 case ZFS_PROP_QUOTA:
2468 err = dsl_dir_set_quota(dsname, source, intval);
2470 case ZFS_PROP_REFQUOTA:
2471 err = dsl_dataset_set_refquota(dsname, source, intval);
2473 case ZFS_PROP_FILESYSTEM_LIMIT:
2474 case ZFS_PROP_SNAPSHOT_LIMIT:
2475 if (intval == UINT64_MAX) {
2476 /* clearing the limit, just do it */
2479 err = dsl_dir_activate_fs_ss_limit(dsname);
2482 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2483 * default path to set the value in the nvlist.
2488 case ZFS_PROP_RESERVATION:
2489 err = dsl_dir_set_reservation(dsname, source, intval);
2491 case ZFS_PROP_REFRESERVATION:
2492 err = dsl_dataset_set_refreservation(dsname, source, intval);
2494 case ZFS_PROP_VOLSIZE:
2495 err = zvol_set_volsize(dsname, intval);
2497 case ZFS_PROP_VERSION:
2501 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2504 err = zfs_set_version(zfsvfs, intval);
2505 zfsvfs_rele(zfsvfs, FTAG);
2507 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2510 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2511 (void) strcpy(zc->zc_name, dsname);
2512 (void) zfs_ioc_userspace_upgrade(zc);
2513 kmem_free(zc, sizeof (zfs_cmd_t));
2525 * This function is best effort. If it fails to set any of the given properties,
2526 * it continues to set as many as it can and returns the last error
2527 * encountered. If the caller provides a non-NULL errlist, it will be filled in
2528 * with the list of names of all the properties that failed along with the
2529 * corresponding error numbers.
2531 * If every property is set successfully, zero is returned and errlist is not
2535 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2543 nvlist_t *genericnvl = fnvlist_alloc();
2544 nvlist_t *retrynvl = fnvlist_alloc();
2548 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2549 const char *propname = nvpair_name(pair);
2550 zfs_prop_t prop = zfs_name_to_prop(propname);
2553 /* decode the property value */
2555 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2557 attrs = fnvpair_value_nvlist(pair);
2558 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2560 err = SET_ERROR(EINVAL);
2563 /* Validate value type */
2564 if (err == 0 && prop == ZPROP_INVAL) {
2565 if (zfs_prop_user(propname)) {
2566 if (nvpair_type(propval) != DATA_TYPE_STRING)
2567 err = SET_ERROR(EINVAL);
2568 } else if (zfs_prop_userquota(propname)) {
2569 if (nvpair_type(propval) !=
2570 DATA_TYPE_UINT64_ARRAY)
2571 err = SET_ERROR(EINVAL);
2573 err = SET_ERROR(EINVAL);
2575 } else if (err == 0) {
2576 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2577 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2578 err = SET_ERROR(EINVAL);
2579 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2582 intval = fnvpair_value_uint64(propval);
2584 switch (zfs_prop_get_type(prop)) {
2585 case PROP_TYPE_NUMBER:
2587 case PROP_TYPE_STRING:
2588 err = SET_ERROR(EINVAL);
2590 case PROP_TYPE_INDEX:
2591 if (zfs_prop_index_to_string(prop,
2592 intval, &unused) != 0)
2593 err = SET_ERROR(EINVAL);
2597 "unknown property type");
2600 err = SET_ERROR(EINVAL);
2604 /* Validate permissions */
2606 err = zfs_check_settable(dsname, pair, CRED());
2609 err = zfs_prop_set_special(dsname, source, pair);
2612 * For better performance we build up a list of
2613 * properties to set in a single transaction.
2615 err = nvlist_add_nvpair(genericnvl, pair);
2616 } else if (err != 0 && nvl != retrynvl) {
2618 * This may be a spurious error caused by
2619 * receiving quota and reservation out of order.
2620 * Try again in a second pass.
2622 err = nvlist_add_nvpair(retrynvl, pair);
2627 if (errlist != NULL)
2628 fnvlist_add_int32(errlist, propname, err);
2633 if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2638 if (!nvlist_empty(genericnvl) &&
2639 dsl_props_set(dsname, source, genericnvl) != 0) {
2641 * If this fails, we still want to set as many properties as we
2642 * can, so try setting them individually.
2645 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2646 const char *propname = nvpair_name(pair);
2650 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2652 attrs = fnvpair_value_nvlist(pair);
2653 propval = fnvlist_lookup_nvpair(attrs,
2657 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2658 strval = fnvpair_value_string(propval);
2659 err = dsl_prop_set_string(dsname, propname,
2662 intval = fnvpair_value_uint64(propval);
2663 err = dsl_prop_set_int(dsname, propname, source,
2668 if (errlist != NULL) {
2669 fnvlist_add_int32(errlist, propname,
2676 nvlist_free(genericnvl);
2677 nvlist_free(retrynvl);
2683 * Check that all the properties are valid user properties.
2686 zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2688 nvpair_t *pair = NULL;
2691 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2692 const char *propname = nvpair_name(pair);
2694 if (!zfs_prop_user(propname) ||
2695 nvpair_type(pair) != DATA_TYPE_STRING)
2696 return (SET_ERROR(EINVAL));
2698 if (error = zfs_secpolicy_write_perms(fsname,
2699 ZFS_DELEG_PERM_USERPROP, CRED()))
2702 if (strlen(propname) >= ZAP_MAXNAMELEN)
2703 return (SET_ERROR(ENAMETOOLONG));
2705 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN)
2712 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2716 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2719 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2720 if (nvlist_exists(skipped, nvpair_name(pair)))
2723 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2728 clear_received_props(const char *dsname, nvlist_t *props,
2732 nvlist_t *cleared_props = NULL;
2733 props_skip(props, skipped, &cleared_props);
2734 if (!nvlist_empty(cleared_props)) {
2736 * Acts on local properties until the dataset has received
2737 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2739 zprop_source_t flags = (ZPROP_SRC_NONE |
2740 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2741 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2743 nvlist_free(cleared_props);
2749 * zc_name name of filesystem
2750 * zc_value name of property to set
2751 * zc_nvlist_src{_size} nvlist of properties to apply
2752 * zc_cookie received properties flag
2755 * zc_nvlist_dst{_size} error for each unapplied received property
2758 zfs_ioc_set_prop(zfs_cmd_t *zc)
2761 boolean_t received = zc->zc_cookie;
2762 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2767 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2768 zc->zc_iflags, &nvl)) != 0)
2772 nvlist_t *origprops;
2774 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2775 (void) clear_received_props(zc->zc_name,
2777 nvlist_free(origprops);
2780 error = dsl_prop_set_hasrecvd(zc->zc_name);
2783 errors = fnvlist_alloc();
2785 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2787 if (zc->zc_nvlist_dst != 0 && errors != NULL) {
2788 (void) put_nvlist(zc, errors);
2791 nvlist_free(errors);
2798 * zc_name name of filesystem
2799 * zc_value name of property to inherit
2800 * zc_cookie revert to received value if TRUE
2805 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2807 const char *propname = zc->zc_value;
2808 zfs_prop_t prop = zfs_name_to_prop(propname);
2809 boolean_t received = zc->zc_cookie;
2810 zprop_source_t source = (received
2811 ? ZPROP_SRC_NONE /* revert to received value, if any */
2812 : ZPROP_SRC_INHERITED); /* explicitly inherit */
2821 * zfs_prop_set_special() expects properties in the form of an
2822 * nvpair with type info.
2824 if (prop == ZPROP_INVAL) {
2825 if (!zfs_prop_user(propname))
2826 return (SET_ERROR(EINVAL));
2828 type = PROP_TYPE_STRING;
2829 } else if (prop == ZFS_PROP_VOLSIZE ||
2830 prop == ZFS_PROP_VERSION) {
2831 return (SET_ERROR(EINVAL));
2833 type = zfs_prop_get_type(prop);
2836 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2839 case PROP_TYPE_STRING:
2840 VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2842 case PROP_TYPE_NUMBER:
2843 case PROP_TYPE_INDEX:
2844 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2848 return (SET_ERROR(EINVAL));
2851 pair = nvlist_next_nvpair(dummy, NULL);
2852 err = zfs_prop_set_special(zc->zc_name, source, pair);
2855 return (err); /* special property already handled */
2858 * Only check this in the non-received case. We want to allow
2859 * 'inherit -S' to revert non-inheritable properties like quota
2860 * and reservation to the received or default values even though
2861 * they are not considered inheritable.
2863 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2864 return (SET_ERROR(EINVAL));
2867 /* property name has been validated by zfs_secpolicy_inherit_prop() */
2868 return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source));
2872 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2879 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2880 zc->zc_iflags, &props))
2884 * If the only property is the configfile, then just do a spa_lookup()
2885 * to handle the faulted case.
2887 pair = nvlist_next_nvpair(props, NULL);
2888 if (pair != NULL && strcmp(nvpair_name(pair),
2889 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2890 nvlist_next_nvpair(props, pair) == NULL) {
2891 mutex_enter(&spa_namespace_lock);
2892 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2893 spa_configfile_set(spa, props, B_FALSE);
2894 spa_config_sync(spa, B_FALSE, B_TRUE);
2896 mutex_exit(&spa_namespace_lock);
2903 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2908 error = spa_prop_set(spa, props);
2911 spa_close(spa, FTAG);
2917 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2921 nvlist_t *nvp = NULL;
2923 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2925 * If the pool is faulted, there may be properties we can still
2926 * get (such as altroot and cachefile), so attempt to get them
2929 mutex_enter(&spa_namespace_lock);
2930 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2931 error = spa_prop_get(spa, &nvp);
2932 mutex_exit(&spa_namespace_lock);
2934 error = spa_prop_get(spa, &nvp);
2935 spa_close(spa, FTAG);
2938 if (error == 0 && zc->zc_nvlist_dst != 0)
2939 error = put_nvlist(zc, nvp);
2941 error = SET_ERROR(EFAULT);
2949 * zc_name name of filesystem
2950 * zc_nvlist_src{_size} nvlist of delegated permissions
2951 * zc_perm_action allow/unallow flag
2956 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2959 nvlist_t *fsaclnv = NULL;
2961 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2962 zc->zc_iflags, &fsaclnv)) != 0)
2966 * Verify nvlist is constructed correctly
2968 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2969 nvlist_free(fsaclnv);
2970 return (SET_ERROR(EINVAL));
2974 * If we don't have PRIV_SYS_MOUNT, then validate
2975 * that user is allowed to hand out each permission in
2979 error = secpolicy_zfs(CRED());
2981 if (zc->zc_perm_action == B_FALSE) {
2982 error = dsl_deleg_can_allow(zc->zc_name,
2985 error = dsl_deleg_can_unallow(zc->zc_name,
2991 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2993 nvlist_free(fsaclnv);
2999 * zc_name name of filesystem
3002 * zc_nvlist_src{_size} nvlist of delegated permissions
3005 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
3010 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
3011 error = put_nvlist(zc, nvp);
3019 * Search the vfs list for a specified resource. Returns a pointer to it
3020 * or NULL if no suitable entry is found. The caller of this routine
3021 * is responsible for releasing the returned vfs pointer.
3024 zfs_get_vfs(const char *resource)
3028 mtx_lock(&mountlist_mtx);
3029 TAILQ_FOREACH(vfsp, &mountlist, mnt_list) {
3030 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
3031 if (vfs_busy(vfsp, MBF_MNTLSTLOCK) != 0)
3037 mtx_unlock(&mountlist_mtx);
3043 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
3045 zfs_creat_t *zct = arg;
3047 zfs_create_fs(os, cr, zct->zct_zplprops, tx);
3050 #define ZFS_PROP_UNDEFINED ((uint64_t)-1)
3054 * os parent objset pointer (NULL if root fs)
3055 * fuids_ok fuids allowed in this version of the spa?
3056 * sa_ok SAs allowed in this version of the spa?
3057 * createprops list of properties requested by creator
3060 * zplprops values for the zplprops we attach to the master node object
3061 * is_ci true if requested file system will be purely case-insensitive
3063 * Determine the settings for utf8only, normalization and
3064 * casesensitivity. Specific values may have been requested by the
3065 * creator and/or we can inherit values from the parent dataset. If
3066 * the file system is of too early a vintage, a creator can not
3067 * request settings for these properties, even if the requested
3068 * setting is the default value. We don't actually want to create dsl
3069 * properties for these, so remove them from the source nvlist after
3073 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
3074 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
3075 nvlist_t *zplprops, boolean_t *is_ci)
3077 uint64_t sense = ZFS_PROP_UNDEFINED;
3078 uint64_t norm = ZFS_PROP_UNDEFINED;
3079 uint64_t u8 = ZFS_PROP_UNDEFINED;
3081 ASSERT(zplprops != NULL);
3084 * Pull out creator prop choices, if any.
3087 (void) nvlist_lookup_uint64(createprops,
3088 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
3089 (void) nvlist_lookup_uint64(createprops,
3090 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
3091 (void) nvlist_remove_all(createprops,
3092 zfs_prop_to_name(ZFS_PROP_NORMALIZE));
3093 (void) nvlist_lookup_uint64(createprops,
3094 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
3095 (void) nvlist_remove_all(createprops,
3096 zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
3097 (void) nvlist_lookup_uint64(createprops,
3098 zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3099 (void) nvlist_remove_all(createprops,
3100 zfs_prop_to_name(ZFS_PROP_CASE));
3104 * If the zpl version requested is whacky or the file system
3105 * or pool is version is too "young" to support normalization
3106 * and the creator tried to set a value for one of the props,
3109 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3110 (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3111 (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3112 (zplver < ZPL_VERSION_NORMALIZATION &&
3113 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3114 sense != ZFS_PROP_UNDEFINED)))
3115 return (SET_ERROR(ENOTSUP));
3118 * Put the version in the zplprops
3120 VERIFY(nvlist_add_uint64(zplprops,
3121 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3123 if (norm == ZFS_PROP_UNDEFINED)
3124 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
3125 VERIFY(nvlist_add_uint64(zplprops,
3126 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3129 * If we're normalizing, names must always be valid UTF-8 strings.
3133 if (u8 == ZFS_PROP_UNDEFINED)
3134 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
3135 VERIFY(nvlist_add_uint64(zplprops,
3136 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3138 if (sense == ZFS_PROP_UNDEFINED)
3139 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
3140 VERIFY(nvlist_add_uint64(zplprops,
3141 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3144 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
3150 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3151 nvlist_t *zplprops, boolean_t *is_ci)
3153 boolean_t fuids_ok, sa_ok;
3154 uint64_t zplver = ZPL_VERSION;
3155 objset_t *os = NULL;
3156 char parentname[MAXNAMELEN];
3162 (void) strlcpy(parentname, dataset, sizeof (parentname));
3163 cp = strrchr(parentname, '/');
3167 if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3170 spa_vers = spa_version(spa);
3171 spa_close(spa, FTAG);
3173 zplver = zfs_zpl_version_map(spa_vers);
3174 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3175 sa_ok = (zplver >= ZPL_VERSION_SA);
3178 * Open parent object set so we can inherit zplprop values.
3180 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3183 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3185 dmu_objset_rele(os, FTAG);
3190 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3191 nvlist_t *zplprops, boolean_t *is_ci)
3195 uint64_t zplver = ZPL_VERSION;
3198 zplver = zfs_zpl_version_map(spa_vers);
3199 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3200 sa_ok = (zplver >= ZPL_VERSION_SA);
3202 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3203 createprops, zplprops, is_ci);
3209 * "type" -> dmu_objset_type_t (int32)
3210 * (optional) "props" -> { prop -> value }
3213 * outnvl: propname -> error code (int32)
3216 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3219 zfs_creat_t zct = { 0 };
3220 nvlist_t *nvprops = NULL;
3221 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3223 dmu_objset_type_t type;
3224 boolean_t is_insensitive = B_FALSE;
3226 if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3227 return (SET_ERROR(EINVAL));
3229 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3233 cbfunc = zfs_create_cb;
3237 cbfunc = zvol_create_cb;
3244 if (strchr(fsname, '@') ||
3245 strchr(fsname, '%'))
3246 return (SET_ERROR(EINVAL));
3248 zct.zct_props = nvprops;
3251 return (SET_ERROR(EINVAL));
3253 if (type == DMU_OST_ZVOL) {
3254 uint64_t volsize, volblocksize;
3256 if (nvprops == NULL)
3257 return (SET_ERROR(EINVAL));
3258 if (nvlist_lookup_uint64(nvprops,
3259 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3260 return (SET_ERROR(EINVAL));
3262 if ((error = nvlist_lookup_uint64(nvprops,
3263 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3264 &volblocksize)) != 0 && error != ENOENT)
3265 return (SET_ERROR(EINVAL));
3268 volblocksize = zfs_prop_default_numeric(
3269 ZFS_PROP_VOLBLOCKSIZE);
3271 if ((error = zvol_check_volblocksize(
3272 volblocksize)) != 0 ||
3273 (error = zvol_check_volsize(volsize,
3274 volblocksize)) != 0)
3276 } else if (type == DMU_OST_ZFS) {
3280 * We have to have normalization and
3281 * case-folding flags correct when we do the
3282 * file system creation, so go figure them out
3285 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3286 NV_UNIQUE_NAME, KM_SLEEP) == 0);
3287 error = zfs_fill_zplprops(fsname, nvprops,
3288 zct.zct_zplprops, &is_insensitive);
3290 nvlist_free(zct.zct_zplprops);
3295 error = dmu_objset_create(fsname, type,
3296 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3297 nvlist_free(zct.zct_zplprops);
3300 * It would be nice to do this atomically.
3303 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3306 (void) dsl_destroy_head(fsname);
3309 if (error == 0 && type == DMU_OST_ZVOL)
3310 zvol_create_minors(fsname);
3317 * "origin" -> name of origin snapshot
3318 * (optional) "props" -> { prop -> value }
3321 * outnvl: propname -> error code (int32)
3324 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3327 nvlist_t *nvprops = NULL;
3330 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3331 return (SET_ERROR(EINVAL));
3332 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3334 if (strchr(fsname, '@') ||
3335 strchr(fsname, '%'))
3336 return (SET_ERROR(EINVAL));
3338 if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3339 return (SET_ERROR(EINVAL));
3340 error = dmu_objset_clone(fsname, origin_name);
3345 * It would be nice to do this atomically.
3348 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3351 (void) dsl_destroy_head(fsname);
3355 zvol_create_minors(fsname);
3362 * "snaps" -> { snapshot1, snapshot2 }
3363 * (optional) "props" -> { prop -> value (string) }
3366 * outnvl: snapshot -> error code (int32)
3369 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3372 nvlist_t *props = NULL;
3376 (void) nvlist_lookup_nvlist(innvl, "props", &props);
3377 if ((error = zfs_check_userprops(poolname, props)) != 0)
3380 if (!nvlist_empty(props) &&
3381 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3382 return (SET_ERROR(ENOTSUP));
3384 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3385 return (SET_ERROR(EINVAL));
3386 poollen = strlen(poolname);
3387 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3388 pair = nvlist_next_nvpair(snaps, pair)) {
3389 const char *name = nvpair_name(pair);
3390 const char *cp = strchr(name, '@');
3393 * The snap name must contain an @, and the part after it must
3394 * contain only valid characters.
3397 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3398 return (SET_ERROR(EINVAL));
3401 * The snap must be in the specified pool.
3403 if (strncmp(name, poolname, poollen) != 0 ||
3404 (name[poollen] != '/' && name[poollen] != '@'))
3405 return (SET_ERROR(EXDEV));
3407 /* This must be the only snap of this fs. */
3408 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3409 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3410 if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3412 return (SET_ERROR(EXDEV));
3417 error = dsl_dataset_snapshot(snaps, props, outnvl);
3422 * innvl: "message" -> string
3426 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3434 * The poolname in the ioctl is not set, we get it from the TSD,
3435 * which was set at the end of the last successful ioctl that allows
3436 * logging. The secpolicy func already checked that it is set.
3437 * Only one log ioctl is allowed after each successful ioctl, so
3438 * we clear the TSD here.
3440 poolname = tsd_get(zfs_allow_log_key);
3441 (void) tsd_set(zfs_allow_log_key, NULL);
3442 error = spa_open(poolname, &spa, FTAG);
3447 if (nvlist_lookup_string(innvl, "message", &message) != 0) {
3448 spa_close(spa, FTAG);
3449 return (SET_ERROR(EINVAL));
3452 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3453 spa_close(spa, FTAG);
3454 return (SET_ERROR(ENOTSUP));
3457 error = spa_history_log(spa, message);
3458 spa_close(spa, FTAG);
3463 * The dp_config_rwlock must not be held when calling this, because the
3464 * unmount may need to write out data.
3466 * This function is best-effort. Callers must deal gracefully if it
3467 * remains mounted (or is remounted after this call).
3469 * Returns 0 if the argument is not a snapshot, or it is not currently a
3470 * filesystem, or we were able to unmount it. Returns error code otherwise.
3473 zfs_unmount_snap(const char *snapname)
3479 if (strchr(snapname, '@') == NULL)
3482 vfsp = zfs_get_vfs(snapname);
3486 zfsvfs = vfsp->vfs_data;
3487 ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os)));
3489 err = vn_vfswlock(vfsp->vfs_vnodecovered);
3496 return (SET_ERROR(err));
3499 * Always force the unmount for snapshots.
3503 (void) dounmount(vfsp, MS_FORCE, kcred);
3506 (void) dounmount(vfsp, MS_FORCE, curthread);
3513 zfs_unmount_snap_cb(const char *snapname, void *arg)
3515 return (zfs_unmount_snap(snapname));
3519 * When a clone is destroyed, its origin may also need to be destroyed,
3520 * in which case it must be unmounted. This routine will do that unmount
3524 zfs_destroy_unmount_origin(const char *fsname)
3530 error = dmu_objset_hold(fsname, FTAG, &os);
3533 ds = dmu_objset_ds(os);
3534 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3535 char originname[MAXNAMELEN];
3536 dsl_dataset_name(ds->ds_prev, originname);
3537 dmu_objset_rele(os, FTAG);
3538 (void) zfs_unmount_snap(originname);
3540 dmu_objset_rele(os, FTAG);
3546 * "snaps" -> { snapshot1, snapshot2 }
3547 * (optional boolean) "defer"
3550 * outnvl: snapshot -> error code (int32)
3555 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3562 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3563 return (SET_ERROR(EINVAL));
3564 defer = nvlist_exists(innvl, "defer");
3566 poollen = strlen(poolname);
3567 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3568 pair = nvlist_next_nvpair(snaps, pair)) {
3569 const char *name = nvpair_name(pair);
3572 * The snap must be in the specified pool to prevent the
3573 * invalid removal of zvol minors below.
3575 if (strncmp(name, poolname, poollen) != 0 ||
3576 (name[poollen] != '/' && name[poollen] != '@'))
3577 return (SET_ERROR(EXDEV));
3579 error = zfs_unmount_snap(name);
3582 #if defined(__FreeBSD__)
3583 zvol_remove_minors(name);
3587 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3591 * Create bookmarks. Bookmark names are of the form <fs>#<bmark>.
3592 * All bookmarks must be in the same pool.
3595 * bookmark1 -> snapshot1, bookmark2 -> snapshot2
3598 * outnvl: bookmark -> error code (int32)
3603 zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3605 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3606 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3610 * Verify the snapshot argument.
3612 if (nvpair_value_string(pair, &snap_name) != 0)
3613 return (SET_ERROR(EINVAL));
3616 /* Verify that the keys (bookmarks) are unique */
3617 for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair);
3618 pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) {
3619 if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0)
3620 return (SET_ERROR(EINVAL));
3624 return (dsl_bookmark_create(innvl, outnvl));
3629 * property 1, property 2, ...
3633 * bookmark name 1 -> { property 1, property 2, ... },
3634 * bookmark name 2 -> { property 1, property 2, ... }
3639 zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3641 return (dsl_get_bookmarks(fsname, innvl, outnvl));
3646 * bookmark name 1, bookmark name 2
3649 * outnvl: bookmark -> error code (int32)
3653 zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl,
3658 poollen = strlen(poolname);
3659 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3660 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3661 const char *name = nvpair_name(pair);
3662 const char *cp = strchr(name, '#');
3665 * The bookmark name must contain an #, and the part after it
3666 * must contain only valid characters.
3669 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3670 return (SET_ERROR(EINVAL));
3673 * The bookmark must be in the specified pool.
3675 if (strncmp(name, poolname, poollen) != 0 ||
3676 (name[poollen] != '/' && name[poollen] != '#'))
3677 return (SET_ERROR(EXDEV));
3680 error = dsl_bookmark_destroy(innvl, outnvl);
3686 * zc_name name of dataset to destroy
3687 * zc_objset_type type of objset
3688 * zc_defer_destroy mark for deferred destroy
3693 zfs_ioc_destroy(zfs_cmd_t *zc)
3697 if (zc->zc_objset_type == DMU_OST_ZFS) {
3698 err = zfs_unmount_snap(zc->zc_name);
3703 if (strchr(zc->zc_name, '@'))
3704 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3706 err = dsl_destroy_head(zc->zc_name);
3707 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3709 zvol_remove_minors(zc->zc_name);
3711 (void) zvol_remove_minor(zc->zc_name);
3717 * fsname is name of dataset to rollback (to most recent snapshot)
3719 * innvl is not used.
3721 * outnvl: "target" -> name of most recent snapshot
3726 zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl)
3731 if (getzfsvfs(fsname, &zfsvfs) == 0) {
3732 error = zfs_suspend_fs(zfsvfs);
3736 error = dsl_dataset_rollback(fsname, zfsvfs, outnvl);
3737 resume_err = zfs_resume_fs(zfsvfs, fsname);
3738 error = error ? error : resume_err;
3741 VFS_RELE(zfsvfs->z_vfs);
3743 vfs_unbusy(zfsvfs->z_vfs);
3746 error = dsl_dataset_rollback(fsname, NULL, outnvl);
3752 recursive_unmount(const char *fsname, void *arg)
3754 const char *snapname = arg;
3755 char fullname[MAXNAMELEN];
3757 (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname);
3758 return (zfs_unmount_snap(fullname));
3763 * zc_name old name of dataset
3764 * zc_value new name of dataset
3765 * zc_cookie recursive flag (only valid for snapshots)
3770 zfs_ioc_rename(zfs_cmd_t *zc)
3772 boolean_t recursive = zc->zc_cookie & 1;
3774 boolean_t allow_mounted = B_TRUE;
3777 allow_mounted = (zc->zc_cookie & 2) != 0;
3780 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3781 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3782 strchr(zc->zc_value, '%'))
3783 return (SET_ERROR(EINVAL));
3785 at = strchr(zc->zc_name, '@');
3787 /* snaps must be in same fs */
3790 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
3791 return (SET_ERROR(EXDEV));
3793 if (zc->zc_objset_type == DMU_OST_ZFS && allow_mounted) {
3794 error = dmu_objset_find(zc->zc_name,
3795 recursive_unmount, at + 1,
3796 recursive ? DS_FIND_CHILDREN : 0);
3802 error = dsl_dataset_rename_snapshot(zc->zc_name,
3803 at + 1, strchr(zc->zc_value, '@') + 1, recursive);
3809 if (zc->zc_objset_type == DMU_OST_ZVOL)
3810 (void) zvol_remove_minor(zc->zc_name);
3812 return (dsl_dir_rename(zc->zc_name, zc->zc_value));
3817 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3819 const char *propname = nvpair_name(pair);
3820 boolean_t issnap = (strchr(dsname, '@') != NULL);
3821 zfs_prop_t prop = zfs_name_to_prop(propname);
3825 if (prop == ZPROP_INVAL) {
3826 if (zfs_prop_user(propname)) {
3827 if (err = zfs_secpolicy_write_perms(dsname,
3828 ZFS_DELEG_PERM_USERPROP, cr))
3833 if (!issnap && zfs_prop_userquota(propname)) {
3834 const char *perm = NULL;
3835 const char *uq_prefix =
3836 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3837 const char *gq_prefix =
3838 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3840 if (strncmp(propname, uq_prefix,
3841 strlen(uq_prefix)) == 0) {
3842 perm = ZFS_DELEG_PERM_USERQUOTA;
3843 } else if (strncmp(propname, gq_prefix,
3844 strlen(gq_prefix)) == 0) {
3845 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3847 /* USERUSED and GROUPUSED are read-only */
3848 return (SET_ERROR(EINVAL));
3851 if (err = zfs_secpolicy_write_perms(dsname, perm, cr))
3856 return (SET_ERROR(EINVAL));
3860 return (SET_ERROR(EINVAL));
3862 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3864 * dsl_prop_get_all_impl() returns properties in this
3868 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3869 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3874 * Check that this value is valid for this pool version
3877 case ZFS_PROP_COMPRESSION:
3879 * If the user specified gzip compression, make sure
3880 * the SPA supports it. We ignore any errors here since
3881 * we'll catch them later.
3883 if (nvpair_value_uint64(pair, &intval) == 0) {
3884 if (intval >= ZIO_COMPRESS_GZIP_1 &&
3885 intval <= ZIO_COMPRESS_GZIP_9 &&
3886 zfs_earlier_version(dsname,
3887 SPA_VERSION_GZIP_COMPRESSION)) {
3888 return (SET_ERROR(ENOTSUP));
3891 if (intval == ZIO_COMPRESS_ZLE &&
3892 zfs_earlier_version(dsname,
3893 SPA_VERSION_ZLE_COMPRESSION))
3894 return (SET_ERROR(ENOTSUP));
3896 if (intval == ZIO_COMPRESS_LZ4) {
3899 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3902 if (!spa_feature_is_enabled(spa,
3903 SPA_FEATURE_LZ4_COMPRESS)) {
3904 spa_close(spa, FTAG);
3905 return (SET_ERROR(ENOTSUP));
3907 spa_close(spa, FTAG);
3911 * If this is a bootable dataset then
3912 * verify that the compression algorithm
3913 * is supported for booting. We must return
3914 * something other than ENOTSUP since it
3915 * implies a downrev pool version.
3917 if (zfs_is_bootfs(dsname) &&
3918 !BOOTFS_COMPRESS_VALID(intval)) {
3919 return (SET_ERROR(ERANGE));
3924 case ZFS_PROP_COPIES:
3925 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3926 return (SET_ERROR(ENOTSUP));
3929 case ZFS_PROP_RECORDSIZE:
3930 /* Record sizes above 128k need the feature to be enabled */
3931 if (nvpair_value_uint64(pair, &intval) == 0 &&
3932 intval > SPA_OLD_MAXBLOCKSIZE) {
3936 * If this is a bootable dataset then
3937 * the we don't allow large (>128K) blocks,
3938 * because GRUB doesn't support them.
3940 if (zfs_is_bootfs(dsname) &&
3941 intval > SPA_OLD_MAXBLOCKSIZE) {
3942 return (SET_ERROR(ERANGE));
3946 * We don't allow setting the property above 1MB,
3947 * unless the tunable has been changed.
3949 if (intval > zfs_max_recordsize ||
3950 intval > SPA_MAXBLOCKSIZE)
3951 return (SET_ERROR(ERANGE));
3953 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3956 if (!spa_feature_is_enabled(spa,
3957 SPA_FEATURE_LARGE_BLOCKS)) {
3958 spa_close(spa, FTAG);
3959 return (SET_ERROR(ENOTSUP));
3961 spa_close(spa, FTAG);
3965 case ZFS_PROP_SHARESMB:
3966 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3967 return (SET_ERROR(ENOTSUP));
3970 case ZFS_PROP_ACLINHERIT:
3971 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3972 nvpair_value_uint64(pair, &intval) == 0) {
3973 if (intval == ZFS_ACL_PASSTHROUGH_X &&
3974 zfs_earlier_version(dsname,
3975 SPA_VERSION_PASSTHROUGH_X))
3976 return (SET_ERROR(ENOTSUP));
3980 case ZFS_PROP_CHECKSUM:
3981 case ZFS_PROP_DEDUP:
3983 spa_feature_t feature;
3986 /* dedup feature version checks */
3987 if (prop == ZFS_PROP_DEDUP &&
3988 zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3989 return (SET_ERROR(ENOTSUP));
3991 if (nvpair_value_uint64(pair, &intval) != 0)
3992 return (SET_ERROR(EINVAL));
3994 /* check prop value is enabled in features */
3995 feature = zio_checksum_to_feature(intval & ZIO_CHECKSUM_MASK);
3996 if (feature == SPA_FEATURE_NONE)
3999 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4002 * Salted checksums are not supported on root pools.
4004 if (spa_bootfs(spa) != 0 &&
4005 intval < ZIO_CHECKSUM_FUNCTIONS &&
4006 (zio_checksum_table[intval].ci_flags &
4007 ZCHECKSUM_FLAG_SALTED)) {
4008 spa_close(spa, FTAG);
4009 return (SET_ERROR(ERANGE));
4011 if (!spa_feature_is_enabled(spa, feature)) {
4012 spa_close(spa, FTAG);
4013 return (SET_ERROR(ENOTSUP));
4015 spa_close(spa, FTAG);
4020 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
4024 * Checks for a race condition to make sure we don't increment a feature flag
4028 zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx)
4030 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
4031 spa_feature_t *featurep = arg;
4033 if (!spa_feature_is_active(spa, *featurep))
4036 return (SET_ERROR(EBUSY));
4040 * The callback invoked on feature activation in the sync task caused by
4041 * zfs_prop_activate_feature.
4044 zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx)
4046 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
4047 spa_feature_t *featurep = arg;
4049 spa_feature_incr(spa, *featurep, tx);
4053 * Activates a feature on a pool in response to a property setting. This
4054 * creates a new sync task which modifies the pool to reflect the feature
4058 zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature)
4062 /* EBUSY here indicates that the feature is already active */
4063 err = dsl_sync_task(spa_name(spa),
4064 zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync,
4065 &feature, 2, ZFS_SPACE_CHECK_RESERVED);
4067 if (err != 0 && err != EBUSY)
4074 * Removes properties from the given props list that fail permission checks
4075 * needed to clear them and to restore them in case of a receive error. For each
4076 * property, make sure we have both set and inherit permissions.
4078 * Returns the first error encountered if any permission checks fail. If the
4079 * caller provides a non-NULL errlist, it also gives the complete list of names
4080 * of all the properties that failed a permission check along with the
4081 * corresponding error numbers. The caller is responsible for freeing the
4084 * If every property checks out successfully, zero is returned and the list
4085 * pointed at by errlist is NULL.
4088 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
4091 nvpair_t *pair, *next_pair;
4098 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4100 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
4101 (void) strcpy(zc->zc_name, dataset);
4102 pair = nvlist_next_nvpair(props, NULL);
4103 while (pair != NULL) {
4104 next_pair = nvlist_next_nvpair(props, pair);
4106 (void) strcpy(zc->zc_value, nvpair_name(pair));
4107 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
4108 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
4109 VERIFY(nvlist_remove_nvpair(props, pair) == 0);
4110 VERIFY(nvlist_add_int32(errors,
4111 zc->zc_value, err) == 0);
4115 kmem_free(zc, sizeof (zfs_cmd_t));
4117 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
4118 nvlist_free(errors);
4121 VERIFY(nvpair_value_int32(pair, &rv) == 0);
4124 if (errlist == NULL)
4125 nvlist_free(errors);
4133 propval_equals(nvpair_t *p1, nvpair_t *p2)
4135 if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
4136 /* dsl_prop_get_all_impl() format */
4138 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
4139 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4143 if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
4145 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
4146 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4150 if (nvpair_type(p1) != nvpair_type(p2))
4153 if (nvpair_type(p1) == DATA_TYPE_STRING) {
4154 char *valstr1, *valstr2;
4156 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
4157 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
4158 return (strcmp(valstr1, valstr2) == 0);
4160 uint64_t intval1, intval2;
4162 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
4163 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
4164 return (intval1 == intval2);
4169 * Remove properties from props if they are not going to change (as determined
4170 * by comparison with origprops). Remove them from origprops as well, since we
4171 * do not need to clear or restore properties that won't change.
4174 props_reduce(nvlist_t *props, nvlist_t *origprops)
4176 nvpair_t *pair, *next_pair;
4178 if (origprops == NULL)
4179 return; /* all props need to be received */
4181 pair = nvlist_next_nvpair(props, NULL);
4182 while (pair != NULL) {
4183 const char *propname = nvpair_name(pair);
4186 next_pair = nvlist_next_nvpair(props, pair);
4188 if ((nvlist_lookup_nvpair(origprops, propname,
4189 &match) != 0) || !propval_equals(pair, match))
4190 goto next; /* need to set received value */
4192 /* don't clear the existing received value */
4193 (void) nvlist_remove_nvpair(origprops, match);
4194 /* don't bother receiving the property */
4195 (void) nvlist_remove_nvpair(props, pair);
4202 * Extract properties that cannot be set PRIOR to the receipt of a dataset.
4203 * For example, refquota cannot be set until after the receipt of a dataset,
4204 * because in replication streams, an older/earlier snapshot may exceed the
4205 * refquota. We want to receive the older/earlier snapshot, but setting
4206 * refquota pre-receipt will set the dsl's ACTUAL quota, which will prevent
4207 * the older/earlier snapshot from being received (with EDQUOT).
4209 * The ZFS test "zfs_receive_011_pos" demonstrates such a scenario.
4211 * libzfs will need to be judicious handling errors encountered by props
4212 * extracted by this function.
4215 extract_delay_props(nvlist_t *props)
4217 nvlist_t *delayprops;
4218 nvpair_t *nvp, *tmp;
4219 static const zfs_prop_t delayable[] = { ZFS_PROP_REFQUOTA, 0 };
4222 VERIFY(nvlist_alloc(&delayprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4224 for (nvp = nvlist_next_nvpair(props, NULL); nvp != NULL;
4225 nvp = nvlist_next_nvpair(props, nvp)) {
4227 * strcmp() is safe because zfs_prop_to_name() always returns
4230 for (i = 0; delayable[i] != 0; i++) {
4231 if (strcmp(zfs_prop_to_name(delayable[i]),
4232 nvpair_name(nvp)) == 0) {
4236 if (delayable[i] != 0) {
4237 tmp = nvlist_prev_nvpair(props, nvp);
4238 VERIFY(nvlist_add_nvpair(delayprops, nvp) == 0);
4239 VERIFY(nvlist_remove_nvpair(props, nvp) == 0);
4244 if (nvlist_empty(delayprops)) {
4245 nvlist_free(delayprops);
4248 return (delayprops);
4252 static boolean_t zfs_ioc_recv_inject_err;
4257 * zc_name name of containing filesystem
4258 * zc_nvlist_src{_size} nvlist of properties to apply
4259 * zc_value name of snapshot to create
4260 * zc_string name of clone origin (if DRR_FLAG_CLONE)
4261 * zc_cookie file descriptor to recv from
4262 * zc_begin_record the BEGIN record of the stream (not byteswapped)
4263 * zc_guid force flag
4264 * zc_cleanup_fd cleanup-on-exit file descriptor
4265 * zc_action_handle handle for this guid/ds mapping (or zero on first call)
4266 * zc_resumable if data is incomplete assume sender will resume
4269 * zc_cookie number of bytes read
4270 * zc_nvlist_dst{_size} error for each unapplied received property
4271 * zc_obj zprop_errflags_t
4272 * zc_action_handle handle for this guid/ds mapping
4275 zfs_ioc_recv(zfs_cmd_t *zc)
4278 dmu_recv_cookie_t drc;
4279 boolean_t force = (boolean_t)zc->zc_guid;
4282 int props_error = 0;
4285 nvlist_t *props = NULL; /* sent properties */
4286 nvlist_t *origprops = NULL; /* existing properties */
4287 nvlist_t *delayprops = NULL; /* sent properties applied post-receive */
4288 char *origin = NULL;
4290 char tofs[ZFS_MAXNAMELEN];
4291 cap_rights_t rights;
4292 boolean_t first_recvd_props = B_FALSE;
4294 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
4295 strchr(zc->zc_value, '@') == NULL ||
4296 strchr(zc->zc_value, '%'))
4297 return (SET_ERROR(EINVAL));
4299 (void) strcpy(tofs, zc->zc_value);
4300 tosnap = strchr(tofs, '@');
4303 if (zc->zc_nvlist_src != 0 &&
4304 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
4305 zc->zc_iflags, &props)) != 0)
4312 fget_read(curthread, fd, cap_rights_init(&rights, CAP_PREAD), &fp);
4316 return (SET_ERROR(EBADF));
4319 errors = fnvlist_alloc();
4321 if (zc->zc_string[0])
4322 origin = zc->zc_string;
4324 error = dmu_recv_begin(tofs, tosnap,
4325 &zc->zc_begin_record, force, zc->zc_resumable, origin, &drc);
4330 * Set properties before we receive the stream so that they are applied
4331 * to the new data. Note that we must call dmu_recv_stream() if
4332 * dmu_recv_begin() succeeds.
4334 if (props != NULL && !drc.drc_newfs) {
4335 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
4336 SPA_VERSION_RECVD_PROPS &&
4337 !dsl_prop_get_hasrecvd(tofs))
4338 first_recvd_props = B_TRUE;
4341 * If new received properties are supplied, they are to
4342 * completely replace the existing received properties, so stash
4343 * away the existing ones.
4345 if (dsl_prop_get_received(tofs, &origprops) == 0) {
4346 nvlist_t *errlist = NULL;
4348 * Don't bother writing a property if its value won't
4349 * change (and avoid the unnecessary security checks).
4351 * The first receive after SPA_VERSION_RECVD_PROPS is a
4352 * special case where we blow away all local properties
4355 if (!first_recvd_props)
4356 props_reduce(props, origprops);
4357 if (zfs_check_clearable(tofs, origprops, &errlist) != 0)
4358 (void) nvlist_merge(errors, errlist, 0);
4359 nvlist_free(errlist);
4361 if (clear_received_props(tofs, origprops,
4362 first_recvd_props ? NULL : props) != 0)
4363 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4365 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4369 if (props != NULL) {
4370 props_error = dsl_prop_set_hasrecvd(tofs);
4372 if (props_error == 0) {
4373 delayprops = extract_delay_props(props);
4374 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4380 error = dmu_recv_stream(&drc, fp, &off, zc->zc_cleanup_fd,
4381 &zc->zc_action_handle);
4384 zfsvfs_t *zfsvfs = NULL;
4386 if (getzfsvfs(tofs, &zfsvfs) == 0) {
4390 error = zfs_suspend_fs(zfsvfs);
4392 * If the suspend fails, then the recv_end will
4393 * likely also fail, and clean up after itself.
4395 end_err = dmu_recv_end(&drc, zfsvfs);
4397 error = zfs_resume_fs(zfsvfs, tofs);
4398 error = error ? error : end_err;
4400 VFS_RELE(zfsvfs->z_vfs);
4402 vfs_unbusy(zfsvfs->z_vfs);
4405 error = dmu_recv_end(&drc, NULL);
4408 /* Set delayed properties now, after we're done receiving. */
4409 if (delayprops != NULL && error == 0) {
4410 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4411 delayprops, errors);
4415 if (delayprops != NULL) {
4417 * Merge delayed props back in with initial props, in case
4418 * we're DEBUG and zfs_ioc_recv_inject_err is set (which means
4419 * we have to make sure clear_received_props() includes
4420 * the delayed properties).
4422 * Since zfs_ioc_recv_inject_err is only in DEBUG kernels,
4423 * using ASSERT() will be just like a VERIFY.
4425 ASSERT(nvlist_merge(props, delayprops, 0) == 0);
4426 nvlist_free(delayprops);
4430 * Now that all props, initial and delayed, are set, report the prop
4431 * errors to the caller.
4433 if (zc->zc_nvlist_dst_size != 0 &&
4434 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4435 put_nvlist(zc, errors) != 0)) {
4437 * Caller made zc->zc_nvlist_dst less than the minimum expected
4438 * size or supplied an invalid address.
4440 props_error = SET_ERROR(EINVAL);
4443 zc->zc_cookie = off - fp->f_offset;
4444 if (off >= 0 && off <= MAXOFFSET_T)
4448 if (zfs_ioc_recv_inject_err) {
4449 zfs_ioc_recv_inject_err = B_FALSE;
4456 zvol_create_minors(tofs);
4460 * On error, restore the original props.
4462 if (error != 0 && props != NULL && !drc.drc_newfs) {
4463 if (clear_received_props(tofs, props, NULL) != 0) {
4465 * We failed to clear the received properties.
4466 * Since we may have left a $recvd value on the
4467 * system, we can't clear the $hasrecvd flag.
4469 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4470 } else if (first_recvd_props) {
4471 dsl_prop_unset_hasrecvd(tofs);
4474 if (origprops == NULL && !drc.drc_newfs) {
4475 /* We failed to stash the original properties. */
4476 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4480 * dsl_props_set() will not convert RECEIVED to LOCAL on or
4481 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4482 * explictly if we're restoring local properties cleared in the
4483 * first new-style receive.
4485 if (origprops != NULL &&
4486 zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4487 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4488 origprops, NULL) != 0) {
4490 * We stashed the original properties but failed to
4493 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4498 nvlist_free(origprops);
4499 nvlist_free(errors);
4503 error = props_error;
4510 * zc_name name of snapshot to send
4511 * zc_cookie file descriptor to send stream to
4512 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj)
4513 * zc_sendobj objsetid of snapshot to send
4514 * zc_fromobj objsetid of incremental fromsnap (may be zero)
4515 * zc_guid if set, estimate size of stream only. zc_cookie is ignored.
4516 * output size in zc_objset_type.
4517 * zc_flags lzc_send_flags
4520 * zc_objset_type estimated size, if zc_guid is set
4523 zfs_ioc_send(zfs_cmd_t *zc)
4527 boolean_t estimate = (zc->zc_guid != 0);
4528 boolean_t embedok = (zc->zc_flags & 0x1);
4529 boolean_t large_block_ok = (zc->zc_flags & 0x2);
4531 if (zc->zc_obj != 0) {
4533 dsl_dataset_t *tosnap;
4535 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4539 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4541 dsl_pool_rele(dp, FTAG);
4545 if (dsl_dir_is_clone(tosnap->ds_dir))
4547 dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj;
4548 dsl_dataset_rele(tosnap, FTAG);
4549 dsl_pool_rele(dp, FTAG);
4554 dsl_dataset_t *tosnap;
4555 dsl_dataset_t *fromsnap = NULL;
4557 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4561 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4563 dsl_pool_rele(dp, FTAG);
4567 if (zc->zc_fromobj != 0) {
4568 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
4571 dsl_dataset_rele(tosnap, FTAG);
4572 dsl_pool_rele(dp, FTAG);
4577 error = dmu_send_estimate(tosnap, fromsnap,
4578 &zc->zc_objset_type);
4580 if (fromsnap != NULL)
4581 dsl_dataset_rele(fromsnap, FTAG);
4582 dsl_dataset_rele(tosnap, FTAG);
4583 dsl_pool_rele(dp, FTAG);
4586 cap_rights_t rights;
4589 fp = getf(zc->zc_cookie);
4591 fget_write(curthread, zc->zc_cookie,
4592 cap_rights_init(&rights, CAP_WRITE), &fp);
4595 return (SET_ERROR(EBADF));
4598 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
4599 zc->zc_fromobj, embedok, large_block_ok,
4601 zc->zc_cookie, fp->f_vnode, &off);
4603 zc->zc_cookie, fp, &off);
4606 if (off >= 0 && off <= MAXOFFSET_T)
4608 releasef(zc->zc_cookie);
4615 * zc_name name of snapshot on which to report progress
4616 * zc_cookie file descriptor of send stream
4619 * zc_cookie number of bytes written in send stream thus far
4622 zfs_ioc_send_progress(zfs_cmd_t *zc)
4626 dmu_sendarg_t *dsp = NULL;
4629 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4633 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4635 dsl_pool_rele(dp, FTAG);
4639 mutex_enter(&ds->ds_sendstream_lock);
4642 * Iterate over all the send streams currently active on this dataset.
4643 * If there's one which matches the specified file descriptor _and_ the
4644 * stream was started by the current process, return the progress of
4647 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4648 dsp = list_next(&ds->ds_sendstreams, dsp)) {
4649 if (dsp->dsa_outfd == zc->zc_cookie &&
4650 dsp->dsa_proc == curproc)
4655 zc->zc_cookie = *(dsp->dsa_off);
4657 error = SET_ERROR(ENOENT);
4659 mutex_exit(&ds->ds_sendstream_lock);
4660 dsl_dataset_rele(ds, FTAG);
4661 dsl_pool_rele(dp, FTAG);
4666 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4670 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4671 &zc->zc_inject_record);
4674 zc->zc_guid = (uint64_t)id;
4680 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4682 return (zio_clear_fault((int)zc->zc_guid));
4686 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4688 int id = (int)zc->zc_guid;
4691 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4692 &zc->zc_inject_record);
4700 zfs_ioc_error_log(zfs_cmd_t *zc)
4704 size_t count = (size_t)zc->zc_nvlist_dst_size;
4706 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4709 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4712 zc->zc_nvlist_dst_size = count;
4714 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4716 spa_close(spa, FTAG);
4722 zfs_ioc_clear(zfs_cmd_t *zc)
4729 * On zpool clear we also fix up missing slogs
4731 mutex_enter(&spa_namespace_lock);
4732 spa = spa_lookup(zc->zc_name);
4734 mutex_exit(&spa_namespace_lock);
4735 return (SET_ERROR(EIO));
4737 if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4738 /* we need to let spa_open/spa_load clear the chains */
4739 spa_set_log_state(spa, SPA_LOG_CLEAR);
4741 spa->spa_last_open_failed = 0;
4742 mutex_exit(&spa_namespace_lock);
4744 if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4745 error = spa_open(zc->zc_name, &spa, FTAG);
4748 nvlist_t *config = NULL;
4750 if (zc->zc_nvlist_src == 0)
4751 return (SET_ERROR(EINVAL));
4753 if ((error = get_nvlist(zc->zc_nvlist_src,
4754 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4755 error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4757 if (config != NULL) {
4760 if ((err = put_nvlist(zc, config)) != 0)
4762 nvlist_free(config);
4764 nvlist_free(policy);
4771 spa_vdev_state_enter(spa, SCL_NONE);
4773 if (zc->zc_guid == 0) {
4776 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4778 (void) spa_vdev_state_exit(spa, NULL, ENODEV);
4779 spa_close(spa, FTAG);
4780 return (SET_ERROR(ENODEV));
4784 vdev_clear(spa, vd);
4786 (void) spa_vdev_state_exit(spa, NULL, 0);
4789 * Resume any suspended I/Os.
4791 if (zio_resume(spa) != 0)
4792 error = SET_ERROR(EIO);
4794 spa_close(spa, FTAG);
4800 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4805 error = spa_open(zc->zc_name, &spa, FTAG);
4809 spa_vdev_state_enter(spa, SCL_NONE);
4812 * If a resilver is already in progress then set the
4813 * spa_scrub_reopen flag to B_TRUE so that we don't restart
4814 * the scan as a side effect of the reopen. Otherwise, let
4815 * vdev_open() decided if a resilver is required.
4817 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4818 vdev_reopen(spa->spa_root_vdev);
4819 spa->spa_scrub_reopen = B_FALSE;
4821 (void) spa_vdev_state_exit(spa, NULL, 0);
4822 spa_close(spa, FTAG);
4827 * zc_name name of filesystem
4828 * zc_value name of origin snapshot
4831 * zc_string name of conflicting snapshot, if there is one
4834 zfs_ioc_promote(zfs_cmd_t *zc)
4839 * We don't need to unmount *all* the origin fs's snapshots, but
4842 cp = strchr(zc->zc_value, '@');
4845 (void) dmu_objset_find(zc->zc_value,
4846 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
4847 return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4851 * Retrieve a single {user|group}{used|quota}@... property.
4854 * zc_name name of filesystem
4855 * zc_objset_type zfs_userquota_prop_t
4856 * zc_value domain name (eg. "S-1-234-567-89")
4857 * zc_guid RID/UID/GID
4860 * zc_cookie property value
4863 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4868 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4869 return (SET_ERROR(EINVAL));
4871 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4875 error = zfs_userspace_one(zfsvfs,
4876 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4877 zfsvfs_rele(zfsvfs, FTAG);
4884 * zc_name name of filesystem
4885 * zc_cookie zap cursor
4886 * zc_objset_type zfs_userquota_prop_t
4887 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4890 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
4891 * zc_cookie zap cursor
4894 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4897 int bufsize = zc->zc_nvlist_dst_size;
4900 return (SET_ERROR(ENOMEM));
4902 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4906 void *buf = kmem_alloc(bufsize, KM_SLEEP);
4908 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
4909 buf, &zc->zc_nvlist_dst_size);
4912 error = ddi_copyout(buf,
4913 (void *)(uintptr_t)zc->zc_nvlist_dst,
4914 zc->zc_nvlist_dst_size, zc->zc_iflags);
4916 kmem_free(buf, bufsize);
4917 zfsvfs_rele(zfsvfs, FTAG);
4924 * zc_name name of filesystem
4930 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4936 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
4937 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
4939 * If userused is not enabled, it may be because the
4940 * objset needs to be closed & reopened (to grow the
4941 * objset_phys_t). Suspend/resume the fs will do that.
4943 error = zfs_suspend_fs(zfsvfs);
4945 dmu_objset_refresh_ownership(zfsvfs->z_os,
4947 error = zfs_resume_fs(zfsvfs, zc->zc_name);
4951 error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
4953 VFS_RELE(zfsvfs->z_vfs);
4955 vfs_unbusy(zfsvfs->z_vfs);
4958 /* XXX kind of reading contents without owning */
4959 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4963 error = dmu_objset_userspace_upgrade(os);
4964 dmu_objset_rele(os, FTAG);
4972 * We don't want to have a hard dependency
4973 * against some special symbols in sharefs
4974 * nfs, and smbsrv. Determine them if needed when
4975 * the first file system is shared.
4976 * Neither sharefs, nfs or smbsrv are unloadable modules.
4978 int (*znfsexport_fs)(void *arg);
4979 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
4980 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
4982 int zfs_nfsshare_inited;
4983 int zfs_smbshare_inited;
4985 ddi_modhandle_t nfs_mod;
4986 ddi_modhandle_t sharefs_mod;
4987 ddi_modhandle_t smbsrv_mod;
4988 #endif /* illumos */
4989 kmutex_t zfs_share_lock;
4997 ASSERT(MUTEX_HELD(&zfs_share_lock));
4998 /* Both NFS and SMB shares also require sharetab support. */
4999 if (sharefs_mod == NULL && ((sharefs_mod =
5000 ddi_modopen("fs/sharefs",
5001 KRTLD_MODE_FIRST, &error)) == NULL)) {
5002 return (SET_ERROR(ENOSYS));
5004 if (zshare_fs == NULL && ((zshare_fs =
5005 (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
5006 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
5007 return (SET_ERROR(ENOSYS));
5011 #endif /* illumos */
5014 zfs_ioc_share(zfs_cmd_t *zc)
5020 switch (zc->zc_share.z_sharetype) {
5022 case ZFS_UNSHARE_NFS:
5023 if (zfs_nfsshare_inited == 0) {
5024 mutex_enter(&zfs_share_lock);
5025 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
5026 KRTLD_MODE_FIRST, &error)) == NULL)) {
5027 mutex_exit(&zfs_share_lock);
5028 return (SET_ERROR(ENOSYS));
5030 if (znfsexport_fs == NULL &&
5031 ((znfsexport_fs = (int (*)(void *))
5033 "nfs_export", &error)) == NULL)) {
5034 mutex_exit(&zfs_share_lock);
5035 return (SET_ERROR(ENOSYS));
5037 error = zfs_init_sharefs();
5039 mutex_exit(&zfs_share_lock);
5040 return (SET_ERROR(ENOSYS));
5042 zfs_nfsshare_inited = 1;
5043 mutex_exit(&zfs_share_lock);
5047 case ZFS_UNSHARE_SMB:
5048 if (zfs_smbshare_inited == 0) {
5049 mutex_enter(&zfs_share_lock);
5050 if (smbsrv_mod == NULL && ((smbsrv_mod =
5051 ddi_modopen("drv/smbsrv",
5052 KRTLD_MODE_FIRST, &error)) == NULL)) {
5053 mutex_exit(&zfs_share_lock);
5054 return (SET_ERROR(ENOSYS));
5056 if (zsmbexport_fs == NULL && ((zsmbexport_fs =
5057 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
5058 "smb_server_share", &error)) == NULL)) {
5059 mutex_exit(&zfs_share_lock);
5060 return (SET_ERROR(ENOSYS));
5062 error = zfs_init_sharefs();
5064 mutex_exit(&zfs_share_lock);
5065 return (SET_ERROR(ENOSYS));
5067 zfs_smbshare_inited = 1;
5068 mutex_exit(&zfs_share_lock);
5072 return (SET_ERROR(EINVAL));
5075 switch (zc->zc_share.z_sharetype) {
5077 case ZFS_UNSHARE_NFS:
5079 znfsexport_fs((void *)
5080 (uintptr_t)zc->zc_share.z_exportdata))
5084 case ZFS_UNSHARE_SMB:
5085 if (error = zsmbexport_fs((void *)
5086 (uintptr_t)zc->zc_share.z_exportdata,
5087 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
5094 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
5095 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
5096 SHAREFS_ADD : SHAREFS_REMOVE;
5099 * Add or remove share from sharetab
5101 error = zshare_fs(opcode,
5102 (void *)(uintptr_t)zc->zc_share.z_sharedata,
5103 zc->zc_share.z_sharemax);
5107 #else /* !illumos */
5109 #endif /* illumos */
5112 ace_t full_access[] = {
5113 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
5118 * zc_name name of containing filesystem
5119 * zc_obj object # beyond which we want next in-use object #
5122 * zc_obj next in-use object #
5125 zfs_ioc_next_obj(zfs_cmd_t *zc)
5127 objset_t *os = NULL;
5130 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
5134 error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
5135 dsl_dataset_phys(os->os_dsl_dataset)->ds_prev_snap_txg);
5137 dmu_objset_rele(os, FTAG);
5143 * zc_name name of filesystem
5144 * zc_value prefix name for snapshot
5145 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
5148 * zc_value short name of new snapshot
5151 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
5158 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
5162 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
5163 (u_longlong_t)ddi_get_lbolt64());
5164 hold_name = kmem_asprintf("%%%s", zc->zc_value);
5166 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
5169 (void) strcpy(zc->zc_value, snap_name);
5172 zfs_onexit_fd_rele(zc->zc_cleanup_fd);
5178 * zc_name name of "to" snapshot
5179 * zc_value name of "from" snapshot
5180 * zc_cookie file descriptor to write diff data on
5183 * dmu_diff_record_t's to the file descriptor
5186 zfs_ioc_diff(zfs_cmd_t *zc)
5189 cap_rights_t rights;
5194 fp = getf(zc->zc_cookie);
5196 fget_write(curthread, zc->zc_cookie,
5197 cap_rights_init(&rights, CAP_WRITE), &fp);
5200 return (SET_ERROR(EBADF));
5205 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off);
5207 error = dmu_diff(zc->zc_name, zc->zc_value, fp, &off);
5210 if (off >= 0 && off <= MAXOFFSET_T)
5212 releasef(zc->zc_cookie);
5219 * Remove all ACL files in shares dir
5222 zfs_smb_acl_purge(znode_t *dzp)
5225 zap_attribute_t zap;
5226 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
5229 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
5230 (error = zap_cursor_retrieve(&zc, &zap)) == 0;
5231 zap_cursor_advance(&zc)) {
5232 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
5236 zap_cursor_fini(&zc);
5239 #endif /* illumos */
5242 zfs_ioc_smb_acl(zfs_cmd_t *zc)
5247 vnode_t *resourcevp = NULL;
5256 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
5257 NO_FOLLOW, NULL, &vp)) != 0)
5260 /* Now make sure mntpnt and dataset are ZFS */
5262 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 ||
5263 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
5264 zc->zc_name) != 0)) {
5266 return (SET_ERROR(EINVAL));
5270 zfsvfs = dzp->z_zfsvfs;
5274 * Create share dir if its missing.
5276 mutex_enter(&zfsvfs->z_lock);
5277 if (zfsvfs->z_shares_dir == 0) {
5280 tx = dmu_tx_create(zfsvfs->z_os);
5281 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
5283 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
5284 error = dmu_tx_assign(tx, TXG_WAIT);
5288 error = zfs_create_share_dir(zfsvfs, tx);
5292 mutex_exit(&zfsvfs->z_lock);
5298 mutex_exit(&zfsvfs->z_lock);
5300 ASSERT(zfsvfs->z_shares_dir);
5301 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
5307 switch (zc->zc_cookie) {
5308 case ZFS_SMB_ACL_ADD:
5309 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
5310 vattr.va_type = VREG;
5311 vattr.va_mode = S_IFREG|0777;
5315 vsec.vsa_mask = VSA_ACE;
5316 vsec.vsa_aclentp = &full_access;
5317 vsec.vsa_aclentsz = sizeof (full_access);
5318 vsec.vsa_aclcnt = 1;
5320 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
5321 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
5323 VN_RELE(resourcevp);
5326 case ZFS_SMB_ACL_REMOVE:
5327 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
5331 case ZFS_SMB_ACL_RENAME:
5332 if ((error = get_nvlist(zc->zc_nvlist_src,
5333 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
5335 VN_RELE(ZTOV(sharedir));
5339 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
5340 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
5343 VN_RELE(ZTOV(sharedir));
5345 nvlist_free(nvlist);
5348 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
5350 nvlist_free(nvlist);
5353 case ZFS_SMB_ACL_PURGE:
5354 error = zfs_smb_acl_purge(sharedir);
5358 error = SET_ERROR(EINVAL);
5363 VN_RELE(ZTOV(sharedir));
5368 #else /* !illumos */
5369 return (EOPNOTSUPP);
5370 #endif /* illumos */
5375 * "holds" -> { snapname -> holdname (string), ... }
5376 * (optional) "cleanup_fd" -> fd (int32)
5380 * snapname -> error value (int32)
5386 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
5390 int cleanup_fd = -1;
5394 error = nvlist_lookup_nvlist(args, "holds", &holds);
5396 return (SET_ERROR(EINVAL));
5398 /* make sure the user didn't pass us any invalid (empty) tags */
5399 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
5400 pair = nvlist_next_nvpair(holds, pair)) {
5403 error = nvpair_value_string(pair, &htag);
5405 return (SET_ERROR(error));
5407 if (strlen(htag) == 0)
5408 return (SET_ERROR(EINVAL));
5411 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
5412 error = zfs_onexit_fd_hold(cleanup_fd, &minor);
5417 error = dsl_dataset_user_hold(holds, minor, errlist);
5419 zfs_onexit_fd_rele(cleanup_fd);
5424 * innvl is not used.
5427 * holdname -> time added (uint64 seconds since epoch)
5433 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
5435 return (dsl_dataset_get_holds(snapname, outnvl));
5440 * snapname -> { holdname, ... }
5445 * snapname -> error value (int32)
5451 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
5453 return (dsl_dataset_user_release(holds, errlist));
5458 * zc_name name of new filesystem or snapshot
5459 * zc_value full name of old snapshot
5462 * zc_cookie space in bytes
5463 * zc_objset_type compressed space in bytes
5464 * zc_perm_action uncompressed space in bytes
5467 zfs_ioc_space_written(zfs_cmd_t *zc)
5471 dsl_dataset_t *new, *old;
5473 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5476 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
5478 dsl_pool_rele(dp, FTAG);
5481 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
5483 dsl_dataset_rele(new, FTAG);
5484 dsl_pool_rele(dp, FTAG);
5488 error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
5489 &zc->zc_objset_type, &zc->zc_perm_action);
5490 dsl_dataset_rele(old, FTAG);
5491 dsl_dataset_rele(new, FTAG);
5492 dsl_pool_rele(dp, FTAG);
5498 * "firstsnap" -> snapshot name
5502 * "used" -> space in bytes
5503 * "compressed" -> compressed space in bytes
5504 * "uncompressed" -> uncompressed space in bytes
5508 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
5512 dsl_dataset_t *new, *old;
5514 uint64_t used, comp, uncomp;
5516 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5517 return (SET_ERROR(EINVAL));
5519 error = dsl_pool_hold(lastsnap, FTAG, &dp);
5523 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
5524 if (error == 0 && !new->ds_is_snapshot) {
5525 dsl_dataset_rele(new, FTAG);
5526 error = SET_ERROR(EINVAL);
5529 dsl_pool_rele(dp, FTAG);
5532 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
5533 if (error == 0 && !old->ds_is_snapshot) {
5534 dsl_dataset_rele(old, FTAG);
5535 error = SET_ERROR(EINVAL);
5538 dsl_dataset_rele(new, FTAG);
5539 dsl_pool_rele(dp, FTAG);
5543 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5544 dsl_dataset_rele(old, FTAG);
5545 dsl_dataset_rele(new, FTAG);
5546 dsl_pool_rele(dp, FTAG);
5547 fnvlist_add_uint64(outnvl, "used", used);
5548 fnvlist_add_uint64(outnvl, "compressed", comp);
5549 fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5554 zfs_ioc_jail(zfs_cmd_t *zc)
5557 return (zone_dataset_attach(curthread->td_ucred, zc->zc_name,
5558 (int)zc->zc_jailid));
5562 zfs_ioc_unjail(zfs_cmd_t *zc)
5565 return (zone_dataset_detach(curthread->td_ucred, zc->zc_name,
5566 (int)zc->zc_jailid));
5571 * "fd" -> file descriptor to write stream to (int32)
5572 * (optional) "fromsnap" -> full snap name to send an incremental from
5573 * (optional) "largeblockok" -> (value ignored)
5574 * indicates that blocks > 128KB are permitted
5575 * (optional) "embedok" -> (value ignored)
5576 * presence indicates DRR_WRITE_EMBEDDED records are permitted
5577 * (optional) "resume_object" and "resume_offset" -> (uint64)
5578 * if present, resume send stream from specified object and offset.
5585 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5587 cap_rights_t rights;
5591 char *fromname = NULL;
5593 boolean_t largeblockok;
5595 uint64_t resumeobj = 0;
5596 uint64_t resumeoff = 0;
5598 error = nvlist_lookup_int32(innvl, "fd", &fd);
5600 return (SET_ERROR(EINVAL));
5602 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
5604 largeblockok = nvlist_exists(innvl, "largeblockok");
5605 embedok = nvlist_exists(innvl, "embedok");
5607 (void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj);
5608 (void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff);
5611 file_t *fp = getf(fd);
5613 fget_write(curthread, fd, cap_rights_init(&rights, CAP_WRITE), &fp);
5616 return (SET_ERROR(EBADF));
5619 error = dmu_send(snapname, fromname, embedok, largeblockok, fd,
5621 resumeobj, resumeoff, fp->f_vnode, &off);
5623 resumeobj, resumeoff, fp, &off);
5627 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5638 * Determine approximately how large a zfs send stream will be -- the number
5639 * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5642 * (optional) "from" -> full snap or bookmark name to send an incremental
5647 * "space" -> bytes of space (uint64)
5651 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5654 dsl_dataset_t *tosnap;
5659 error = dsl_pool_hold(snapname, FTAG, &dp);
5663 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
5665 dsl_pool_rele(dp, FTAG);
5669 error = nvlist_lookup_string(innvl, "from", &fromname);
5671 if (strchr(fromname, '@') != NULL) {
5673 * If from is a snapshot, hold it and use the more
5674 * efficient dmu_send_estimate to estimate send space
5675 * size using deadlists.
5677 dsl_dataset_t *fromsnap;
5678 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
5681 error = dmu_send_estimate(tosnap, fromsnap, &space);
5682 dsl_dataset_rele(fromsnap, FTAG);
5683 } else if (strchr(fromname, '#') != NULL) {
5685 * If from is a bookmark, fetch the creation TXG of the
5686 * snapshot it was created from and use that to find
5687 * blocks that were born after it.
5689 zfs_bookmark_phys_t frombm;
5691 error = dsl_bookmark_lookup(dp, fromname, tosnap,
5695 error = dmu_send_estimate_from_txg(tosnap,
5696 frombm.zbm_creation_txg, &space);
5699 * from is not properly formatted as a snapshot or
5702 error = SET_ERROR(EINVAL);
5706 // If estimating the size of a full send, use dmu_send_estimate
5707 error = dmu_send_estimate(tosnap, NULL, &space);
5710 fnvlist_add_uint64(outnvl, "space", space);
5713 dsl_dataset_rele(tosnap, FTAG);
5714 dsl_pool_rele(dp, FTAG);
5718 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
5721 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5722 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5723 boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
5725 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5727 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5728 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5729 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5730 ASSERT3P(vec->zvec_func, ==, NULL);
5732 vec->zvec_legacy_func = func;
5733 vec->zvec_secpolicy = secpolicy;
5734 vec->zvec_namecheck = namecheck;
5735 vec->zvec_allow_log = log_history;
5736 vec->zvec_pool_check = pool_check;
5740 * See the block comment at the beginning of this file for details on
5741 * each argument to this function.
5744 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
5745 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5746 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
5747 boolean_t allow_log)
5749 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5751 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5752 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5753 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5754 ASSERT3P(vec->zvec_func, ==, NULL);
5756 /* if we are logging, the name must be valid */
5757 ASSERT(!allow_log || namecheck != NO_NAME);
5759 vec->zvec_name = name;
5760 vec->zvec_func = func;
5761 vec->zvec_secpolicy = secpolicy;
5762 vec->zvec_namecheck = namecheck;
5763 vec->zvec_pool_check = pool_check;
5764 vec->zvec_smush_outnvlist = smush_outnvlist;
5765 vec->zvec_allow_log = allow_log;
5769 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5770 zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
5771 zfs_ioc_poolcheck_t pool_check)
5773 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5774 POOL_NAME, log_history, pool_check);
5778 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5779 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
5781 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5782 DATASET_NAME, B_FALSE, pool_check);
5786 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5788 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
5789 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5793 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5794 zfs_secpolicy_func_t *secpolicy)
5796 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5797 NO_NAME, B_FALSE, POOL_CHECK_NONE);
5801 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
5802 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
5804 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5805 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
5809 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5811 zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
5812 zfs_secpolicy_read);
5816 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5817 zfs_secpolicy_func_t *secpolicy)
5819 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5820 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5824 zfs_ioctl_init(void)
5826 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
5827 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
5828 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5830 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
5831 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
5832 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
5834 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
5835 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
5836 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5838 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
5839 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
5840 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5842 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
5843 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
5844 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5846 zfs_ioctl_register("create", ZFS_IOC_CREATE,
5847 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
5848 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5850 zfs_ioctl_register("clone", ZFS_IOC_CLONE,
5851 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
5852 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5854 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
5855 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
5856 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5858 zfs_ioctl_register("hold", ZFS_IOC_HOLD,
5859 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
5860 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5861 zfs_ioctl_register("release", ZFS_IOC_RELEASE,
5862 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
5863 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5865 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
5866 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
5867 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5869 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
5870 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
5871 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE);
5873 zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK,
5874 zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME,
5875 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5877 zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS,
5878 zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME,
5879 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5881 zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS,
5882 zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks,
5884 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5886 /* IOCTLS that use the legacy function signature */
5888 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
5889 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
5891 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
5892 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5893 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
5895 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
5896 zfs_ioc_pool_upgrade);
5897 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
5899 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
5900 zfs_ioc_vdev_remove);
5901 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
5902 zfs_ioc_vdev_set_state);
5903 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
5904 zfs_ioc_vdev_attach);
5905 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
5906 zfs_ioc_vdev_detach);
5907 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
5908 zfs_ioc_vdev_setpath);
5909 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
5910 zfs_ioc_vdev_setfru);
5911 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
5912 zfs_ioc_pool_set_props);
5913 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
5914 zfs_ioc_vdev_split);
5915 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
5916 zfs_ioc_pool_reguid);
5918 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
5919 zfs_ioc_pool_configs, zfs_secpolicy_none);
5920 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
5921 zfs_ioc_pool_tryimport, zfs_secpolicy_config);
5922 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
5923 zfs_ioc_inject_fault, zfs_secpolicy_inject);
5924 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
5925 zfs_ioc_clear_fault, zfs_secpolicy_inject);
5926 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
5927 zfs_ioc_inject_list_next, zfs_secpolicy_inject);
5930 * pool destroy, and export don't log the history as part of
5931 * zfsdev_ioctl, but rather zfs_ioc_pool_export
5932 * does the logging of those commands.
5934 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
5935 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5936 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
5937 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5939 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
5940 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5941 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
5942 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5944 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
5945 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_NONE);
5946 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
5947 zfs_ioc_dsobj_to_dsname,
5948 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_NONE);
5949 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
5950 zfs_ioc_pool_get_history,
5951 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
5953 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
5954 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5956 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
5957 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5958 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
5959 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5961 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
5962 zfs_ioc_space_written);
5963 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
5964 zfs_ioc_objset_recvd_props);
5965 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
5967 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
5969 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
5970 zfs_ioc_objset_stats);
5971 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
5972 zfs_ioc_objset_zplprops);
5973 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
5974 zfs_ioc_dataset_list_next);
5975 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
5976 zfs_ioc_snapshot_list_next);
5977 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
5978 zfs_ioc_send_progress);
5980 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
5981 zfs_ioc_diff, zfs_secpolicy_diff);
5982 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
5983 zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
5984 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
5985 zfs_ioc_obj_to_path, zfs_secpolicy_diff);
5986 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
5987 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
5988 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
5989 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
5990 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
5991 zfs_ioc_send, zfs_secpolicy_send);
5993 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
5994 zfs_secpolicy_none);
5995 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
5996 zfs_secpolicy_destroy);
5997 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
5998 zfs_secpolicy_rename);
5999 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
6000 zfs_secpolicy_recv);
6001 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
6002 zfs_secpolicy_promote);
6003 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
6004 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
6005 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
6006 zfs_secpolicy_set_fsacl);
6008 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
6009 zfs_secpolicy_share, POOL_CHECK_NONE);
6010 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
6011 zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
6012 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
6013 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
6014 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
6015 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
6016 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
6017 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
6020 zfs_ioctl_register_dataset_nolog(ZFS_IOC_JAIL, zfs_ioc_jail,
6021 zfs_secpolicy_config, POOL_CHECK_NONE);
6022 zfs_ioctl_register_dataset_nolog(ZFS_IOC_UNJAIL, zfs_ioc_unjail,
6023 zfs_secpolicy_config, POOL_CHECK_NONE);
6028 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
6029 zfs_ioc_poolcheck_t check)
6034 ASSERT(type == POOL_NAME || type == DATASET_NAME);
6036 if (check & POOL_CHECK_NONE)
6039 error = spa_open(name, &spa, FTAG);
6041 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
6042 error = SET_ERROR(EAGAIN);
6043 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
6044 error = SET_ERROR(EROFS);
6045 spa_close(spa, FTAG);
6051 * Find a free minor number.
6054 zfsdev_minor_alloc(void)
6056 static minor_t last_minor;
6059 ASSERT(MUTEX_HELD(&spa_namespace_lock));
6061 for (m = last_minor + 1; m != last_minor; m++) {
6062 if (m > ZFSDEV_MAX_MINOR)
6064 if (ddi_get_soft_state(zfsdev_state, m) == NULL) {
6074 zfs_ctldev_init(struct cdev *devp)
6077 zfs_soft_state_t *zs;
6079 ASSERT(MUTEX_HELD(&spa_namespace_lock));
6081 minor = zfsdev_minor_alloc();
6083 return (SET_ERROR(ENXIO));
6085 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS)
6086 return (SET_ERROR(EAGAIN));
6088 devfs_set_cdevpriv((void *)(uintptr_t)minor, zfsdev_close);
6090 zs = ddi_get_soft_state(zfsdev_state, minor);
6091 zs->zss_type = ZSST_CTLDEV;
6092 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data);
6098 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor)
6100 ASSERT(MUTEX_HELD(&spa_namespace_lock));
6102 zfs_onexit_destroy(zo);
6103 ddi_soft_state_free(zfsdev_state, minor);
6107 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which)
6109 zfs_soft_state_t *zp;
6111 zp = ddi_get_soft_state(zfsdev_state, minor);
6112 if (zp == NULL || zp->zss_type != which)
6115 return (zp->zss_data);
6119 zfsdev_open(struct cdev *devp, int flag, int mode, struct thread *td)
6124 if (getminor(*devp) != 0)
6125 return (zvol_open(devp, flag, otyp, cr));
6128 /* This is the control device. Allocate a new minor if requested. */
6130 mutex_enter(&spa_namespace_lock);
6131 error = zfs_ctldev_init(devp);
6132 mutex_exit(&spa_namespace_lock);
6139 zfsdev_close(void *data)
6142 minor_t minor = (minor_t)(uintptr_t)data;
6147 mutex_enter(&spa_namespace_lock);
6148 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV);
6150 mutex_exit(&spa_namespace_lock);
6153 zfs_ctldev_destroy(zo, minor);
6154 mutex_exit(&spa_namespace_lock);
6158 zfsdev_ioctl(struct cdev *dev, u_long zcmd, caddr_t arg, int flag,
6165 minor_t minor = getminor(dev);
6167 zfs_iocparm_t *zc_iocparm;
6168 int cflag, cmd, oldvecnum;
6169 boolean_t newioc, compat;
6170 void *compat_zc = NULL;
6171 cred_t *cr = td->td_ucred;
6173 const zfs_ioc_vec_t *vec;
6174 char *saved_poolname = NULL;
6175 nvlist_t *innvl = NULL;
6177 cflag = ZFS_CMD_COMPAT_NONE;
6179 newioc = B_TRUE; /* "new" style (zfs_iocparm_t) ioctl */
6181 len = IOCPARM_LEN(zcmd);
6182 vecnum = cmd = zcmd & 0xff;
6185 * Check if we are talking to supported older binaries
6186 * and translate zfs_cmd if necessary
6188 if (len != sizeof(zfs_iocparm_t)) {
6195 case sizeof(zfs_cmd_zcmd_t):
6196 cflag = ZFS_CMD_COMPAT_LZC;
6198 case sizeof(zfs_cmd_deadman_t):
6199 cflag = ZFS_CMD_COMPAT_DEADMAN;
6201 case sizeof(zfs_cmd_v28_t):
6202 cflag = ZFS_CMD_COMPAT_V28;
6204 case sizeof(zfs_cmd_v15_t):
6205 cflag = ZFS_CMD_COMPAT_V15;
6206 vecnum = zfs_ioctl_v15_to_v28[cmd];
6209 * Return without further handling
6210 * if the command is blacklisted.
6212 if (vecnum == ZFS_IOC_COMPAT_PASS)
6214 else if (vecnum == ZFS_IOC_COMPAT_FAIL)
6223 vecnum = cmd - ZFS_IOC_FIRST;
6224 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
6227 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
6228 return (SET_ERROR(EINVAL));
6229 vec = &zfs_ioc_vec[vecnum];
6231 zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP);
6234 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
6236 error = SET_ERROR(EFAULT);
6239 #else /* !illumos */
6240 bzero(zc, sizeof(zfs_cmd_t));
6243 zc_iocparm = (void *)arg;
6245 switch (zc_iocparm->zfs_ioctl_version) {
6246 case ZFS_IOCVER_CURRENT:
6247 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_t)) {
6248 error = SET_ERROR(EINVAL);
6252 case ZFS_IOCVER_RESUME:
6253 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_resume_t)) {
6254 error = SET_ERROR(EFAULT);
6258 cflag = ZFS_CMD_COMPAT_RESUME;
6260 case ZFS_IOCVER_EDBP:
6261 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_edbp_t)) {
6262 error = SET_ERROR(EFAULT);
6266 cflag = ZFS_CMD_COMPAT_EDBP;
6268 case ZFS_IOCVER_ZCMD:
6269 if (zc_iocparm->zfs_cmd_size > sizeof(zfs_cmd_t) ||
6270 zc_iocparm->zfs_cmd_size < sizeof(zfs_cmd_zcmd_t)) {
6271 error = SET_ERROR(EFAULT);
6275 cflag = ZFS_CMD_COMPAT_ZCMD;
6278 error = SET_ERROR(EINVAL);
6284 ASSERT(sizeof(zfs_cmd_t) >= zc_iocparm->zfs_cmd_size);
6285 compat_zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP);
6286 bzero(compat_zc, sizeof(zfs_cmd_t));
6288 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd,
6289 compat_zc, zc_iocparm->zfs_cmd_size, flag);
6291 error = SET_ERROR(EFAULT);
6295 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd,
6296 zc, zc_iocparm->zfs_cmd_size, flag);
6298 error = SET_ERROR(EFAULT);
6306 ASSERT(compat_zc != NULL);
6307 zfs_cmd_compat_get(zc, compat_zc, cflag);
6309 ASSERT(compat_zc == NULL);
6310 zfs_cmd_compat_get(zc, arg, cflag);
6313 error = zfs_ioctl_compat_pre(zc, &vecnum, cflag);
6316 if (oldvecnum != vecnum)
6317 vec = &zfs_ioc_vec[vecnum];
6319 #endif /* !illumos */
6321 zc->zc_iflags = flag & FKIOCTL;
6322 if (zc->zc_nvlist_src_size != 0) {
6323 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
6324 zc->zc_iflags, &innvl);
6329 /* rewrite innvl for backwards compatibility */
6331 innvl = zfs_ioctl_compat_innvl(zc, innvl, vecnum, cflag);
6334 * Ensure that all pool/dataset names are valid before we pass down to
6337 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
6338 switch (vec->zvec_namecheck) {
6340 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
6341 error = SET_ERROR(EINVAL);
6343 error = pool_status_check(zc->zc_name,
6344 vec->zvec_namecheck, vec->zvec_pool_check);
6348 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
6349 error = SET_ERROR(EINVAL);
6351 error = pool_status_check(zc->zc_name,
6352 vec->zvec_namecheck, vec->zvec_pool_check);
6359 if (error == 0 && !(flag & FKIOCTL))
6360 error = vec->zvec_secpolicy(zc, innvl, cr);
6365 /* legacy ioctls can modify zc_name */
6366 len = strcspn(zc->zc_name, "/@#") + 1;
6367 saved_poolname = kmem_alloc(len, KM_SLEEP);
6368 (void) strlcpy(saved_poolname, zc->zc_name, len);
6370 if (vec->zvec_func != NULL) {
6374 nvlist_t *lognv = NULL;
6376 ASSERT(vec->zvec_legacy_func == NULL);
6379 * Add the innvl to the lognv before calling the func,
6380 * in case the func changes the innvl.
6382 if (vec->zvec_allow_log) {
6383 lognv = fnvlist_alloc();
6384 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
6386 if (!nvlist_empty(innvl)) {
6387 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
6392 outnvl = fnvlist_alloc();
6393 error = vec->zvec_func(zc->zc_name, innvl, outnvl);
6395 if (error == 0 && vec->zvec_allow_log &&
6396 spa_open(zc->zc_name, &spa, FTAG) == 0) {
6397 if (!nvlist_empty(outnvl)) {
6398 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
6401 (void) spa_history_log_nvl(spa, lognv);
6402 spa_close(spa, FTAG);
6404 fnvlist_free(lognv);
6406 /* rewrite outnvl for backwards compatibility */
6408 outnvl = zfs_ioctl_compat_outnvl(zc, outnvl, vecnum,
6411 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
6413 if (vec->zvec_smush_outnvlist) {
6414 smusherror = nvlist_smush(outnvl,
6415 zc->zc_nvlist_dst_size);
6417 if (smusherror == 0)
6418 puterror = put_nvlist(zc, outnvl);
6424 nvlist_free(outnvl);
6426 error = vec->zvec_legacy_func(zc);
6433 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
6434 if (error == 0 && rc != 0)
6435 error = SET_ERROR(EFAULT);
6438 zfs_ioctl_compat_post(zc, cmd, cflag);
6440 ASSERT(compat_zc != NULL);
6441 ASSERT(sizeof(zfs_cmd_t) >= zc_iocparm->zfs_cmd_size);
6443 zfs_cmd_compat_put(zc, compat_zc, vecnum, cflag);
6444 rc = ddi_copyout(compat_zc,
6445 (void *)(uintptr_t)zc_iocparm->zfs_cmd,
6446 zc_iocparm->zfs_cmd_size, flag);
6447 if (error == 0 && rc != 0)
6448 error = SET_ERROR(EFAULT);
6449 kmem_free(compat_zc, sizeof (zfs_cmd_t));
6451 zfs_cmd_compat_put(zc, arg, vecnum, cflag);
6456 rc = ddi_copyout(zc, (void *)(uintptr_t)zc_iocparm->zfs_cmd,
6457 sizeof (zfs_cmd_t), flag);
6458 if (error == 0 && rc != 0)
6459 error = SET_ERROR(EFAULT);
6462 if (error == 0 && vec->zvec_allow_log) {
6463 char *s = tsd_get(zfs_allow_log_key);
6466 (void) tsd_set(zfs_allow_log_key, saved_poolname);
6468 if (saved_poolname != NULL)
6469 strfree(saved_poolname);
6472 kmem_free(zc, sizeof (zfs_cmd_t));
6478 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
6480 if (cmd != DDI_ATTACH)
6481 return (DDI_FAILURE);
6483 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
6484 DDI_PSEUDO, 0) == DDI_FAILURE)
6485 return (DDI_FAILURE);
6489 ddi_report_dev(dip);
6491 return (DDI_SUCCESS);
6495 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
6497 if (spa_busy() || zfs_busy() || zvol_busy())
6498 return (DDI_FAILURE);
6500 if (cmd != DDI_DETACH)
6501 return (DDI_FAILURE);
6505 ddi_prop_remove_all(dip);
6506 ddi_remove_minor_node(dip, NULL);
6508 return (DDI_SUCCESS);
6513 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
6516 case DDI_INFO_DEVT2DEVINFO:
6518 return (DDI_SUCCESS);
6520 case DDI_INFO_DEVT2INSTANCE:
6521 *result = (void *)0;
6522 return (DDI_SUCCESS);
6525 return (DDI_FAILURE);
6527 #endif /* illumos */
6530 * OK, so this is a little weird.
6532 * /dev/zfs is the control node, i.e. minor 0.
6533 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
6535 * /dev/zfs has basically nothing to do except serve up ioctls,
6536 * so most of the standard driver entry points are in zvol.c.
6539 static struct cb_ops zfs_cb_ops = {
6540 zfsdev_open, /* open */
6541 zfsdev_close, /* close */
6542 zvol_strategy, /* strategy */
6544 zvol_dump, /* dump */
6545 zvol_read, /* read */
6546 zvol_write, /* write */
6547 zfsdev_ioctl, /* ioctl */
6551 nochpoll, /* poll */
6552 ddi_prop_op, /* prop_op */
6553 NULL, /* streamtab */
6554 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */
6555 CB_REV, /* version */
6556 nodev, /* async read */
6557 nodev, /* async write */
6560 static struct dev_ops zfs_dev_ops = {
6561 DEVO_REV, /* version */
6563 zfs_info, /* info */
6564 nulldev, /* identify */
6565 nulldev, /* probe */
6566 zfs_attach, /* attach */
6567 zfs_detach, /* detach */
6569 &zfs_cb_ops, /* driver operations */
6570 NULL, /* no bus operations */
6572 ddi_quiesce_not_needed, /* quiesce */
6575 static struct modldrv zfs_modldrv = {
6581 static struct modlinkage modlinkage = {
6583 (void *)&zfs_modlfs,
6584 (void *)&zfs_modldrv,
6587 #endif /* illumos */
6589 static struct cdevsw zfs_cdevsw = {
6590 .d_version = D_VERSION,
6591 .d_open = zfsdev_open,
6592 .d_ioctl = zfsdev_ioctl,
6593 .d_name = ZFS_DEV_NAME
6597 zfs_allow_log_destroy(void *arg)
6599 char *poolname = arg;
6606 zfsdev = make_dev(&zfs_cdevsw, 0x0, UID_ROOT, GID_OPERATOR, 0666,
6614 destroy_dev(zfsdev);
6617 static struct root_hold_token *zfs_root_token;
6618 struct proc *zfsproc;
6626 spa_init(FREAD | FWRITE);
6631 if ((error = mod_install(&modlinkage)) != 0) {
6638 tsd_create(&zfs_fsyncer_key, NULL);
6639 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
6640 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
6642 error = ldi_ident_from_mod(&modlinkage, &zfs_li);
6644 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
6654 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
6655 return (SET_ERROR(EBUSY));
6657 if ((error = mod_remove(&modlinkage)) != 0)
6663 if (zfs_nfsshare_inited)
6664 (void) ddi_modclose(nfs_mod);
6665 if (zfs_smbshare_inited)
6666 (void) ddi_modclose(smbsrv_mod);
6667 if (zfs_nfsshare_inited || zfs_smbshare_inited)
6668 (void) ddi_modclose(sharefs_mod);
6670 tsd_destroy(&zfs_fsyncer_key);
6671 ldi_ident_release(zfs_li);
6673 mutex_destroy(&zfs_share_lock);
6679 _info(struct modinfo *modinfop)
6681 return (mod_info(&modlinkage, modinfop));
6683 #endif /* illumos */
6685 static int zfs__init(void);
6686 static int zfs__fini(void);
6687 static void zfs_shutdown(void *, int);
6689 static eventhandler_tag zfs_shutdown_event_tag;
6692 #define ZFS_MIN_KSTACK_PAGES 4
6700 #if KSTACK_PAGES < ZFS_MIN_KSTACK_PAGES
6701 printf("ZFS NOTICE: KSTACK_PAGES is %d which could result in stack "
6702 "overflow panic!\nPlease consider adding "
6703 "'options KSTACK_PAGES=%d' to your kernel config\n", KSTACK_PAGES,
6704 ZFS_MIN_KSTACK_PAGES);
6707 zfs_root_token = root_mount_hold("ZFS");
6709 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
6711 spa_init(FREAD | FWRITE);
6716 tsd_create(&zfs_fsyncer_key, NULL);
6717 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
6718 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
6720 printf("ZFS storage pool version: features support (" SPA_VERSION_STRING ")\n");
6721 root_mount_rel(zfs_root_token);
6731 if (spa_busy() || zfs_busy() || zvol_busy() ||
6732 zio_injection_enabled) {
6741 tsd_destroy(&zfs_fsyncer_key);
6742 tsd_destroy(&rrw_tsd_key);
6743 tsd_destroy(&zfs_allow_log_key);
6745 mutex_destroy(&zfs_share_lock);
6751 zfs_shutdown(void *arg __unused, int howto __unused)
6755 * ZFS fini routines can not properly work in a panic-ed system.
6757 if (panicstr == NULL)
6763 zfs_modevent(module_t mod, int type, void *unused __unused)
6771 zfs_shutdown_event_tag = EVENTHANDLER_REGISTER(
6772 shutdown_post_sync, zfs_shutdown, NULL,
6773 SHUTDOWN_PRI_FIRST);
6777 if (err == 0 && zfs_shutdown_event_tag != NULL)
6778 EVENTHANDLER_DEREGISTER(shutdown_post_sync,
6779 zfs_shutdown_event_tag);
6786 return (EOPNOTSUPP);
6789 static moduledata_t zfs_mod = {
6794 DECLARE_MODULE(zfsctrl, zfs_mod, SI_SUB_VFS, SI_ORDER_ANY);
6795 MODULE_VERSION(zfsctrl, 1);
6796 MODULE_DEPEND(zfsctrl, opensolaris, 1, 1, 1);
6797 MODULE_DEPEND(zfsctrl, krpc, 1, 1, 1);
6798 MODULE_DEPEND(zfsctrl, acl_nfs4, 1, 1, 1);