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 Nexenta Systems, Inc. All rights reserved.
29 * Copyright (c) 2014, Joyent, Inc. All rights reserved.
30 * Copyright (c) 2011, 2014 by Delphix. All rights reserved.
31 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
32 * Copyright (c) 2013 Steven Hartland. All rights reserved.
38 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
39 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
41 * There are two ways that we handle ioctls: the legacy way where almost
42 * all of the logic is in the ioctl callback, and the new way where most
43 * of the marshalling is handled in the common entry point, zfsdev_ioctl().
45 * Non-legacy ioctls should be registered by calling
46 * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked
47 * from userland by lzc_ioctl().
49 * The registration arguments are as follows:
52 * The name of the ioctl. This is used for history logging. If the
53 * ioctl returns successfully (the callback returns 0), and allow_log
54 * is true, then a history log entry will be recorded with the input &
55 * output nvlists. The log entry can be printed with "zpool history -i".
58 * The ioctl request number, which userland will pass to ioctl(2).
59 * The ioctl numbers can change from release to release, because
60 * the caller (libzfs) must be matched to the kernel.
62 * zfs_secpolicy_func_t *secpolicy
63 * This function will be called before the zfs_ioc_func_t, to
64 * determine if this operation is permitted. It should return EPERM
65 * on failure, and 0 on success. Checks include determining if the
66 * dataset is visible in this zone, and if the user has either all
67 * zfs privileges in the zone (SYS_MOUNT), or has been granted permission
68 * to do this operation on this dataset with "zfs allow".
70 * zfs_ioc_namecheck_t namecheck
71 * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
72 * name, a dataset name, or nothing. If the name is not well-formed,
73 * the ioctl will fail and the callback will not be called.
74 * Therefore, the callback can assume that the name is well-formed
75 * (e.g. is null-terminated, doesn't have more than one '@' character,
76 * doesn't have invalid characters).
78 * zfs_ioc_poolcheck_t pool_check
79 * This specifies requirements on the pool state. If the pool does
80 * not meet them (is suspended or is readonly), the ioctl will fail
81 * and the callback will not be called. If any checks are specified
82 * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
83 * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED |
84 * POOL_CHECK_READONLY).
86 * boolean_t smush_outnvlist
87 * If smush_outnvlist is true, then the output is presumed to be a
88 * list of errors, and it will be "smushed" down to fit into the
89 * caller's buffer, by removing some entries and replacing them with a
90 * single "N_MORE_ERRORS" entry indicating how many were removed. See
91 * nvlist_smush() for details. If smush_outnvlist is false, and the
92 * outnvlist does not fit into the userland-provided buffer, then the
93 * ioctl will fail with ENOMEM.
95 * zfs_ioc_func_t *func
96 * The callback function that will perform the operation.
98 * The callback should return 0 on success, or an error number on
99 * failure. If the function fails, the userland ioctl will return -1,
100 * and errno will be set to the callback's return value. The callback
101 * will be called with the following arguments:
104 * The name of the pool or dataset to operate on, from
105 * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the
106 * expected type (pool, dataset, or none).
109 * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or
110 * NULL if no input nvlist was provided. Changes to this nvlist are
111 * ignored. If the input nvlist could not be deserialized, the
112 * ioctl will fail and the callback will not be called.
115 * The output nvlist, initially empty. The callback can fill it in,
116 * and it will be returned to userland by serializing it into
117 * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization
118 * fails (e.g. because the caller didn't supply a large enough
119 * buffer), then the overall ioctl will fail. See the
120 * 'smush_nvlist' argument above for additional behaviors.
122 * There are two typical uses of the output nvlist:
123 * - To return state, e.g. property values. In this case,
124 * smush_outnvlist should be false. If the buffer was not large
125 * enough, the caller will reallocate a larger buffer and try
128 * - To return multiple errors from an ioctl which makes on-disk
129 * changes. In this case, smush_outnvlist should be true.
130 * Ioctls which make on-disk modifications should generally not
131 * use the outnvl if they succeed, because the caller can not
132 * distinguish between the operation failing, and
133 * deserialization failing.
136 #include "opt_kstack_pages.h"
139 #include <sys/types.h>
140 #include <sys/param.h>
141 #include <sys/systm.h>
142 #include <sys/conf.h>
143 #include <sys/kernel.h>
144 #include <sys/lock.h>
145 #include <sys/malloc.h>
146 #include <sys/mutex.h>
147 #include <sys/proc.h>
148 #include <sys/errno.h>
151 #include <sys/file.h>
152 #include <sys/kmem.h>
153 #include <sys/conf.h>
154 #include <sys/cmn_err.h>
155 #include <sys/stat.h>
156 #include <sys/zfs_ioctl.h>
157 #include <sys/zfs_vfsops.h>
158 #include <sys/zfs_znode.h>
161 #include <sys/spa_impl.h>
162 #include <sys/vdev.h>
164 #include <sys/dsl_dir.h>
165 #include <sys/dsl_dataset.h>
166 #include <sys/dsl_prop.h>
167 #include <sys/dsl_deleg.h>
168 #include <sys/dmu_objset.h>
169 #include <sys/dmu_impl.h>
170 #include <sys/dmu_tx.h>
171 #include <sys/sunddi.h>
172 #include <sys/policy.h>
173 #include <sys/zone.h>
174 #include <sys/nvpair.h>
175 #include <sys/mount.h>
176 #include <sys/taskqueue.h>
178 #include <sys/varargs.h>
179 #include <sys/fs/zfs.h>
180 #include <sys/zfs_ctldir.h>
181 #include <sys/zfs_dir.h>
182 #include <sys/zfs_onexit.h>
183 #include <sys/zvol.h>
184 #include <sys/dsl_scan.h>
185 #include <sys/dmu_objset.h>
186 #include <sys/dmu_send.h>
187 #include <sys/dsl_destroy.h>
188 #include <sys/dsl_bookmark.h>
189 #include <sys/dsl_userhold.h>
190 #include <sys/zfeature.h>
192 #include "zfs_namecheck.h"
193 #include "zfs_prop.h"
194 #include "zfs_deleg.h"
195 #include "zfs_comutil.h"
196 #include "zfs_ioctl_compat.h"
198 CTASSERT(sizeof(zfs_cmd_t) < IOCPARM_MAX);
200 static struct cdev *zfsdev;
202 extern void zfs_init(void);
203 extern void zfs_fini(void);
205 uint_t zfs_fsyncer_key;
206 extern uint_t rrw_tsd_key;
207 static uint_t zfs_allow_log_key;
209 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *);
210 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *);
211 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *);
217 } zfs_ioc_namecheck_t;
220 POOL_CHECK_NONE = 1 << 0,
221 POOL_CHECK_SUSPENDED = 1 << 1,
222 POOL_CHECK_READONLY = 1 << 2,
223 } zfs_ioc_poolcheck_t;
225 typedef struct zfs_ioc_vec {
226 zfs_ioc_legacy_func_t *zvec_legacy_func;
227 zfs_ioc_func_t *zvec_func;
228 zfs_secpolicy_func_t *zvec_secpolicy;
229 zfs_ioc_namecheck_t zvec_namecheck;
230 boolean_t zvec_allow_log;
231 zfs_ioc_poolcheck_t zvec_pool_check;
232 boolean_t zvec_smush_outnvlist;
233 const char *zvec_name;
236 /* This array is indexed by zfs_userquota_prop_t */
237 static const char *userquota_perms[] = {
238 ZFS_DELEG_PERM_USERUSED,
239 ZFS_DELEG_PERM_USERQUOTA,
240 ZFS_DELEG_PERM_GROUPUSED,
241 ZFS_DELEG_PERM_GROUPQUOTA,
244 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
245 static int zfs_check_settable(const char *name, nvpair_t *property,
247 static int zfs_check_clearable(char *dataset, nvlist_t *props,
249 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
251 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
252 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
254 static void zfsdev_close(void *data);
256 static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature);
258 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
260 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
267 * Get rid of annoying "../common/" prefix to filename.
269 newfile = strrchr(file, '/');
270 if (newfile != NULL) {
271 newfile = newfile + 1; /* Get rid of leading / */
277 (void) vsnprintf(buf, sizeof (buf), fmt, adx);
281 * To get this data, use the zfs-dprintf probe as so:
282 * dtrace -q -n 'zfs-dprintf \
283 * /stringof(arg0) == "dbuf.c"/ \
284 * {printf("%s: %s", stringof(arg1), stringof(arg3))}'
286 * arg1 = function name
290 DTRACE_PROBE4(zfs__dprintf,
291 char *, newfile, char *, func, int, line, char *, buf);
295 history_str_free(char *buf)
297 kmem_free(buf, HIS_MAX_RECORD_LEN);
301 history_str_get(zfs_cmd_t *zc)
305 if (zc->zc_history == 0)
308 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
309 if (copyinstr((void *)(uintptr_t)zc->zc_history,
310 buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
311 history_str_free(buf);
315 buf[HIS_MAX_RECORD_LEN -1] = '\0';
321 * Check to see if the named dataset is currently defined as bootable
324 zfs_is_bootfs(const char *name)
328 if (dmu_objset_hold(name, FTAG, &os) == 0) {
330 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
331 dmu_objset_rele(os, FTAG);
338 * Return non-zero if the spa version is less than requested version.
341 zfs_earlier_version(const char *name, int version)
345 if (spa_open(name, &spa, FTAG) == 0) {
346 if (spa_version(spa) < version) {
347 spa_close(spa, FTAG);
350 spa_close(spa, FTAG);
356 * Return TRUE if the ZPL version is less than requested version.
359 zpl_earlier_version(const char *name, int version)
362 boolean_t rc = B_TRUE;
364 if (dmu_objset_hold(name, FTAG, &os) == 0) {
367 if (dmu_objset_type(os) != DMU_OST_ZFS) {
368 dmu_objset_rele(os, FTAG);
371 /* XXX reading from non-owned objset */
372 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
373 rc = zplversion < version;
374 dmu_objset_rele(os, FTAG);
380 zfs_log_history(zfs_cmd_t *zc)
385 if ((buf = history_str_get(zc)) == NULL)
388 if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
389 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
390 (void) spa_history_log(spa, buf);
391 spa_close(spa, FTAG);
393 history_str_free(buf);
397 * Policy for top-level read operations (list pools). Requires no privileges,
398 * and can be used in the local zone, as there is no associated dataset.
402 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
408 * Policy for dataset read operations (list children, get statistics). Requires
409 * no privileges, but must be visible in the local zone.
413 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
415 if (INGLOBALZONE(curthread) ||
416 zone_dataset_visible(zc->zc_name, NULL))
419 return (SET_ERROR(ENOENT));
423 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
428 * The dataset must be visible by this zone -- check this first
429 * so they don't see EPERM on something they shouldn't know about.
431 if (!INGLOBALZONE(curthread) &&
432 !zone_dataset_visible(dataset, &writable))
433 return (SET_ERROR(ENOENT));
435 if (INGLOBALZONE(curthread)) {
437 * If the fs is zoned, only root can access it from the
440 if (secpolicy_zfs(cr) && zoned)
441 return (SET_ERROR(EPERM));
444 * If we are in a local zone, the 'zoned' property must be set.
447 return (SET_ERROR(EPERM));
449 /* must be writable by this zone */
451 return (SET_ERROR(EPERM));
457 zfs_dozonecheck(const char *dataset, cred_t *cr)
461 if (dsl_prop_get_integer(dataset, "jailed", &zoned, NULL))
462 return (SET_ERROR(ENOENT));
464 return (zfs_dozonecheck_impl(dataset, zoned, cr));
468 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
472 if (dsl_prop_get_int_ds(ds, "jailed", &zoned))
473 return (SET_ERROR(ENOENT));
475 return (zfs_dozonecheck_impl(dataset, zoned, cr));
479 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
480 const char *perm, cred_t *cr)
484 error = zfs_dozonecheck_ds(name, ds, cr);
486 error = secpolicy_zfs(cr);
488 error = dsl_deleg_access_impl(ds, perm, cr);
494 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
500 error = dsl_pool_hold(name, FTAG, &dp);
504 error = dsl_dataset_hold(dp, name, FTAG, &ds);
506 dsl_pool_rele(dp, FTAG);
510 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
512 dsl_dataset_rele(ds, FTAG);
513 dsl_pool_rele(dp, FTAG);
519 * Policy for setting the security label property.
521 * Returns 0 for success, non-zero for access and other errors.
524 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
526 char ds_hexsl[MAXNAMELEN];
527 bslabel_t ds_sl, new_sl;
528 boolean_t new_default = FALSE;
530 int needed_priv = -1;
533 /* First get the existing dataset label. */
534 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
535 1, sizeof (ds_hexsl), &ds_hexsl, NULL);
537 return (SET_ERROR(EPERM));
539 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
542 /* The label must be translatable */
543 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
544 return (SET_ERROR(EINVAL));
547 * In a non-global zone, disallow attempts to set a label that
548 * doesn't match that of the zone; otherwise no other checks
551 if (!INGLOBALZONE(curproc)) {
552 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
553 return (SET_ERROR(EPERM));
558 * For global-zone datasets (i.e., those whose zoned property is
559 * "off", verify that the specified new label is valid for the
562 if (dsl_prop_get_integer(name,
563 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
564 return (SET_ERROR(EPERM));
566 if (zfs_check_global_label(name, strval) != 0)
567 return (SET_ERROR(EPERM));
571 * If the existing dataset label is nondefault, check if the
572 * dataset is mounted (label cannot be changed while mounted).
573 * Get the zfsvfs; if there isn't one, then the dataset isn't
574 * mounted (or isn't a dataset, doesn't exist, ...).
576 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
578 static char *setsl_tag = "setsl_tag";
581 * Try to own the dataset; abort if there is any error,
582 * (e.g., already mounted, in use, or other error).
584 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
587 return (SET_ERROR(EPERM));
589 dmu_objset_disown(os, setsl_tag);
592 needed_priv = PRIV_FILE_DOWNGRADE_SL;
596 if (hexstr_to_label(strval, &new_sl) != 0)
597 return (SET_ERROR(EPERM));
599 if (blstrictdom(&ds_sl, &new_sl))
600 needed_priv = PRIV_FILE_DOWNGRADE_SL;
601 else if (blstrictdom(&new_sl, &ds_sl))
602 needed_priv = PRIV_FILE_UPGRADE_SL;
604 /* dataset currently has a default label */
606 needed_priv = PRIV_FILE_UPGRADE_SL;
610 if (needed_priv != -1)
611 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
614 #endif /* SECLABEL */
617 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
623 * Check permissions for special properties.
628 * Disallow setting of 'zoned' from within a local zone.
630 if (!INGLOBALZONE(curthread))
631 return (SET_ERROR(EPERM));
635 case ZFS_PROP_FILESYSTEM_LIMIT:
636 case ZFS_PROP_SNAPSHOT_LIMIT:
637 if (!INGLOBALZONE(curthread)) {
639 char setpoint[MAXNAMELEN];
641 * Unprivileged users are allowed to modify the
642 * limit on things *under* (ie. contained by)
643 * the thing they own.
645 if (dsl_prop_get_integer(dsname, "jailed", &zoned,
647 return (SET_ERROR(EPERM));
648 if (!zoned || strlen(dsname) <= strlen(setpoint))
649 return (SET_ERROR(EPERM));
653 case ZFS_PROP_MLSLABEL:
655 if (!is_system_labeled())
656 return (SET_ERROR(EPERM));
658 if (nvpair_value_string(propval, &strval) == 0) {
661 err = zfs_set_slabel_policy(dsname, strval, CRED());
671 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
676 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
680 error = zfs_dozonecheck(zc->zc_name, cr);
685 * permission to set permissions will be evaluated later in
686 * dsl_deleg_can_allow()
693 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
695 return (zfs_secpolicy_write_perms(zc->zc_name,
696 ZFS_DELEG_PERM_ROLLBACK, cr));
701 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
709 * Generate the current snapshot name from the given objsetid, then
710 * use that name for the secpolicy/zone checks.
712 cp = strchr(zc->zc_name, '@');
714 return (SET_ERROR(EINVAL));
715 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
719 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
721 dsl_pool_rele(dp, FTAG);
725 dsl_dataset_name(ds, zc->zc_name);
727 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
728 ZFS_DELEG_PERM_SEND, cr);
729 dsl_dataset_rele(ds, FTAG);
730 dsl_pool_rele(dp, FTAG);
737 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
739 return (zfs_secpolicy_write_perms(zc->zc_name,
740 ZFS_DELEG_PERM_SEND, cr));
745 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
750 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
751 NO_FOLLOW, NULL, &vp)) != 0)
754 /* Now make sure mntpnt and dataset are ZFS */
756 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 ||
757 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
758 zc->zc_name) != 0)) {
760 return (SET_ERROR(EPERM));
764 return (dsl_deleg_access(zc->zc_name,
765 ZFS_DELEG_PERM_SHARE, cr));
769 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
771 if (!INGLOBALZONE(curthread))
772 return (SET_ERROR(EPERM));
774 if (secpolicy_nfs(cr) == 0) {
777 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
782 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
784 if (!INGLOBALZONE(curthread))
785 return (SET_ERROR(EPERM));
787 if (secpolicy_smb(cr) == 0) {
790 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
795 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
800 * Remove the @bla or /bla from the end of the name to get the parent.
802 (void) strncpy(parent, datasetname, parentsize);
803 cp = strrchr(parent, '@');
807 cp = strrchr(parent, '/');
809 return (SET_ERROR(ENOENT));
817 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
821 if ((error = zfs_secpolicy_write_perms(name,
822 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
825 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
830 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
832 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
836 * Destroying snapshots with delegated permissions requires
837 * descendant mount and destroy permissions.
841 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
844 nvpair_t *pair, *nextpair;
847 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
848 return (SET_ERROR(EINVAL));
849 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
851 nextpair = nvlist_next_nvpair(snaps, pair);
852 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr);
853 if (error == ENOENT) {
855 * Ignore any snapshots that don't exist (we consider
856 * them "already destroyed"). Remove the name from the
857 * nvl here in case the snapshot is created between
858 * now and when we try to destroy it (in which case
859 * we don't want to destroy it since we haven't
860 * checked for permission).
862 fnvlist_remove_nvpair(snaps, pair);
873 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
875 char parentname[MAXNAMELEN];
878 if ((error = zfs_secpolicy_write_perms(from,
879 ZFS_DELEG_PERM_RENAME, cr)) != 0)
882 if ((error = zfs_secpolicy_write_perms(from,
883 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
886 if ((error = zfs_get_parent(to, parentname,
887 sizeof (parentname))) != 0)
890 if ((error = zfs_secpolicy_write_perms(parentname,
891 ZFS_DELEG_PERM_CREATE, cr)) != 0)
894 if ((error = zfs_secpolicy_write_perms(parentname,
895 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
903 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
908 if ((zc->zc_cookie & 1) != 0) {
910 * This is recursive rename, so the starting snapshot might
911 * not exist. Check file system or volume permission instead.
913 at = strchr(zc->zc_name, '@');
919 error = zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr);
929 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
932 dsl_dataset_t *clone;
935 error = zfs_secpolicy_write_perms(zc->zc_name,
936 ZFS_DELEG_PERM_PROMOTE, cr);
940 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
944 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
947 char parentname[MAXNAMELEN];
948 dsl_dataset_t *origin = NULL;
952 error = dsl_dataset_hold_obj(dd->dd_pool,
953 dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin);
955 dsl_dataset_rele(clone, FTAG);
956 dsl_pool_rele(dp, FTAG);
960 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
961 ZFS_DELEG_PERM_MOUNT, cr);
963 dsl_dataset_name(origin, parentname);
965 error = zfs_secpolicy_write_perms_ds(parentname, origin,
966 ZFS_DELEG_PERM_PROMOTE, cr);
968 dsl_dataset_rele(clone, FTAG);
969 dsl_dataset_rele(origin, FTAG);
971 dsl_pool_rele(dp, FTAG);
977 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
981 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
982 ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
985 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
986 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
989 return (zfs_secpolicy_write_perms(zc->zc_name,
990 ZFS_DELEG_PERM_CREATE, cr));
994 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
996 return (zfs_secpolicy_write_perms(name,
997 ZFS_DELEG_PERM_SNAPSHOT, cr));
1001 * Check for permission to create each snapshot in the nvlist.
1005 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1011 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
1012 return (SET_ERROR(EINVAL));
1013 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
1014 pair = nvlist_next_nvpair(snaps, pair)) {
1015 char *name = nvpair_name(pair);
1016 char *atp = strchr(name, '@');
1019 error = SET_ERROR(EINVAL);
1023 error = zfs_secpolicy_snapshot_perms(name, cr);
1032 * Check for permission to create each snapshot in the nvlist.
1036 zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1040 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
1041 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
1042 char *name = nvpair_name(pair);
1043 char *hashp = strchr(name, '#');
1045 if (hashp == NULL) {
1046 error = SET_ERROR(EINVAL);
1050 error = zfs_secpolicy_write_perms(name,
1051 ZFS_DELEG_PERM_BOOKMARK, cr);
1061 zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1063 nvpair_t *pair, *nextpair;
1066 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1068 char *name = nvpair_name(pair);
1069 char *hashp = strchr(name, '#');
1070 nextpair = nvlist_next_nvpair(innvl, pair);
1072 if (hashp == NULL) {
1073 error = SET_ERROR(EINVAL);
1078 error = zfs_secpolicy_write_perms(name,
1079 ZFS_DELEG_PERM_DESTROY, cr);
1081 if (error == ENOENT) {
1083 * Ignore any filesystems that don't exist (we consider
1084 * their bookmarks "already destroyed"). Remove
1085 * the name from the nvl here in case the filesystem
1086 * is created between now and when we try to destroy
1087 * the bookmark (in which case we don't want to
1088 * destroy it since we haven't checked for permission).
1090 fnvlist_remove_nvpair(innvl, pair);
1102 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1105 * Even root must have a proper TSD so that we know what pool
1108 if (tsd_get(zfs_allow_log_key) == NULL)
1109 return (SET_ERROR(EPERM));
1114 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1116 char parentname[MAXNAMELEN];
1120 if ((error = zfs_get_parent(zc->zc_name, parentname,
1121 sizeof (parentname))) != 0)
1124 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1125 (error = zfs_secpolicy_write_perms(origin,
1126 ZFS_DELEG_PERM_CLONE, cr)) != 0)
1129 if ((error = zfs_secpolicy_write_perms(parentname,
1130 ZFS_DELEG_PERM_CREATE, cr)) != 0)
1133 return (zfs_secpolicy_write_perms(parentname,
1134 ZFS_DELEG_PERM_MOUNT, cr));
1138 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires
1139 * SYS_CONFIG privilege, which is not available in a local zone.
1143 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1145 if (secpolicy_sys_config(cr, B_FALSE) != 0)
1146 return (SET_ERROR(EPERM));
1152 * Policy for object to name lookups.
1156 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1160 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1163 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1168 * Policy for fault injection. Requires all privileges.
1172 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1174 return (secpolicy_zinject(cr));
1179 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1181 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1183 if (prop == ZPROP_INVAL) {
1184 if (!zfs_prop_user(zc->zc_value))
1185 return (SET_ERROR(EINVAL));
1186 return (zfs_secpolicy_write_perms(zc->zc_name,
1187 ZFS_DELEG_PERM_USERPROP, cr));
1189 return (zfs_secpolicy_setprop(zc->zc_name, prop,
1195 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1197 int err = zfs_secpolicy_read(zc, innvl, cr);
1201 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1202 return (SET_ERROR(EINVAL));
1204 if (zc->zc_value[0] == 0) {
1206 * They are asking about a posix uid/gid. If it's
1207 * themself, allow it.
1209 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1210 zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
1211 if (zc->zc_guid == crgetuid(cr))
1214 if (groupmember(zc->zc_guid, cr))
1219 return (zfs_secpolicy_write_perms(zc->zc_name,
1220 userquota_perms[zc->zc_objset_type], cr));
1224 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1226 int err = zfs_secpolicy_read(zc, innvl, cr);
1230 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1231 return (SET_ERROR(EINVAL));
1233 return (zfs_secpolicy_write_perms(zc->zc_name,
1234 userquota_perms[zc->zc_objset_type], cr));
1239 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1241 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1247 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1253 error = nvlist_lookup_nvlist(innvl, "holds", &holds);
1255 return (SET_ERROR(EINVAL));
1257 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
1258 pair = nvlist_next_nvpair(holds, pair)) {
1259 char fsname[MAXNAMELEN];
1260 error = dmu_fsname(nvpair_name(pair), fsname);
1263 error = zfs_secpolicy_write_perms(fsname,
1264 ZFS_DELEG_PERM_HOLD, cr);
1273 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1278 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1279 pair = nvlist_next_nvpair(innvl, pair)) {
1280 char fsname[MAXNAMELEN];
1281 error = dmu_fsname(nvpair_name(pair), fsname);
1284 error = zfs_secpolicy_write_perms(fsname,
1285 ZFS_DELEG_PERM_RELEASE, cr);
1293 * Policy for allowing temporary snapshots to be taken or released
1296 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1299 * A temporary snapshot is the same as a snapshot,
1300 * hold, destroy and release all rolled into one.
1301 * Delegated diff alone is sufficient that we allow this.
1305 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1306 ZFS_DELEG_PERM_DIFF, cr)) == 0)
1309 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1311 error = zfs_secpolicy_hold(zc, innvl, cr);
1313 error = zfs_secpolicy_release(zc, innvl, cr);
1315 error = zfs_secpolicy_destroy(zc, innvl, cr);
1320 * Returns the nvlist as specified by the user in the zfs_cmd_t.
1323 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1327 nvlist_t *list = NULL;
1330 * Read in and unpack the user-supplied nvlist.
1333 return (SET_ERROR(EINVAL));
1335 packed = kmem_alloc(size, KM_SLEEP);
1337 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1339 kmem_free(packed, size);
1343 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1344 kmem_free(packed, size);
1348 kmem_free(packed, size);
1355 * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1356 * Entries will be removed from the end of the nvlist, and one int32 entry
1357 * named "N_MORE_ERRORS" will be added indicating how many entries were
1361 nvlist_smush(nvlist_t *errors, size_t max)
1365 size = fnvlist_size(errors);
1368 nvpair_t *more_errors;
1372 return (SET_ERROR(ENOMEM));
1374 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1375 more_errors = nvlist_prev_nvpair(errors, NULL);
1378 nvpair_t *pair = nvlist_prev_nvpair(errors,
1380 fnvlist_remove_nvpair(errors, pair);
1382 size = fnvlist_size(errors);
1383 } while (size > max);
1385 fnvlist_remove_nvpair(errors, more_errors);
1386 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1387 ASSERT3U(fnvlist_size(errors), <=, max);
1394 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1396 char *packed = NULL;
1400 size = fnvlist_size(nvl);
1402 if (size > zc->zc_nvlist_dst_size) {
1404 * Solaris returns ENOMEM here, because even if an error is
1405 * returned from an ioctl(2), new zc_nvlist_dst_size will be
1406 * passed to the userland. This is not the case for FreeBSD.
1407 * We need to return 0, so the kernel will copy the
1408 * zc_nvlist_dst_size back and the userland can discover that a
1409 * bigger buffer is needed.
1413 packed = fnvlist_pack(nvl, &size);
1414 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1415 size, zc->zc_iflags) != 0)
1416 error = SET_ERROR(EFAULT);
1417 fnvlist_pack_free(packed, size);
1420 zc->zc_nvlist_dst_size = size;
1421 zc->zc_nvlist_dst_filled = B_TRUE;
1426 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1431 error = dmu_objset_hold(dsname, FTAG, &os);
1434 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1435 dmu_objset_rele(os, FTAG);
1436 return (SET_ERROR(EINVAL));
1439 mutex_enter(&os->os_user_ptr_lock);
1440 *zfvp = dmu_objset_get_user(os);
1442 VFS_HOLD((*zfvp)->z_vfs);
1444 error = SET_ERROR(ESRCH);
1446 mutex_exit(&os->os_user_ptr_lock);
1447 dmu_objset_rele(os, FTAG);
1452 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1453 * case its z_vfs will be NULL, and it will be opened as the owner.
1454 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1455 * which prevents all vnode ops from running.
1458 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1462 if (getzfsvfs(name, zfvp) != 0)
1463 error = zfsvfs_create(name, zfvp);
1465 rrm_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER :
1467 if ((*zfvp)->z_unmounted) {
1469 * XXX we could probably try again, since the unmounting
1470 * thread should be just about to disassociate the
1471 * objset from the zfsvfs.
1473 rrm_exit(&(*zfvp)->z_teardown_lock, tag);
1474 return (SET_ERROR(EBUSY));
1481 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1483 rrm_exit(&zfsvfs->z_teardown_lock, tag);
1485 if (zfsvfs->z_vfs) {
1486 VFS_RELE(zfsvfs->z_vfs);
1488 dmu_objset_disown(zfsvfs->z_os, zfsvfs);
1489 zfsvfs_free(zfsvfs);
1494 zfs_ioc_pool_create(zfs_cmd_t *zc)
1497 nvlist_t *config, *props = NULL;
1498 nvlist_t *rootprops = NULL;
1499 nvlist_t *zplprops = NULL;
1501 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1502 zc->zc_iflags, &config))
1505 if (zc->zc_nvlist_src_size != 0 && (error =
1506 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1507 zc->zc_iflags, &props))) {
1508 nvlist_free(config);
1513 nvlist_t *nvl = NULL;
1514 uint64_t version = SPA_VERSION;
1516 (void) nvlist_lookup_uint64(props,
1517 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1518 if (!SPA_VERSION_IS_SUPPORTED(version)) {
1519 error = SET_ERROR(EINVAL);
1520 goto pool_props_bad;
1522 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1524 error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1526 nvlist_free(config);
1530 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1532 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1533 error = zfs_fill_zplprops_root(version, rootprops,
1536 goto pool_props_bad;
1539 error = spa_create(zc->zc_name, config, props, zplprops);
1542 * Set the remaining root properties
1544 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1545 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1546 (void) spa_destroy(zc->zc_name);
1549 nvlist_free(rootprops);
1550 nvlist_free(zplprops);
1551 nvlist_free(config);
1558 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1561 zfs_log_history(zc);
1562 error = spa_destroy(zc->zc_name);
1564 zvol_remove_minors(zc->zc_name);
1569 zfs_ioc_pool_import(zfs_cmd_t *zc)
1571 nvlist_t *config, *props = NULL;
1575 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1576 zc->zc_iflags, &config)) != 0)
1579 if (zc->zc_nvlist_src_size != 0 && (error =
1580 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1581 zc->zc_iflags, &props))) {
1582 nvlist_free(config);
1586 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1587 guid != zc->zc_guid)
1588 error = SET_ERROR(EINVAL);
1590 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1592 if (zc->zc_nvlist_dst != 0) {
1595 if ((err = put_nvlist(zc, config)) != 0)
1599 nvlist_free(config);
1608 zfs_ioc_pool_export(zfs_cmd_t *zc)
1611 boolean_t force = (boolean_t)zc->zc_cookie;
1612 boolean_t hardforce = (boolean_t)zc->zc_guid;
1614 zfs_log_history(zc);
1615 error = spa_export(zc->zc_name, NULL, force, hardforce);
1617 zvol_remove_minors(zc->zc_name);
1622 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1627 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1628 return (SET_ERROR(EEXIST));
1630 error = put_nvlist(zc, configs);
1632 nvlist_free(configs);
1639 * zc_name name of the pool
1642 * zc_cookie real errno
1643 * zc_nvlist_dst config nvlist
1644 * zc_nvlist_dst_size size of config nvlist
1647 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1653 error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1654 sizeof (zc->zc_value));
1656 if (config != NULL) {
1657 ret = put_nvlist(zc, config);
1658 nvlist_free(config);
1661 * The config may be present even if 'error' is non-zero.
1662 * In this case we return success, and preserve the real errno
1665 zc->zc_cookie = error;
1674 * Try to import the given pool, returning pool stats as appropriate so that
1675 * user land knows which devices are available and overall pool health.
1678 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1680 nvlist_t *tryconfig, *config;
1683 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1684 zc->zc_iflags, &tryconfig)) != 0)
1687 config = spa_tryimport(tryconfig);
1689 nvlist_free(tryconfig);
1692 return (SET_ERROR(EINVAL));
1694 error = put_nvlist(zc, config);
1695 nvlist_free(config);
1702 * zc_name name of the pool
1703 * zc_cookie scan func (pool_scan_func_t)
1706 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1711 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1714 if (zc->zc_cookie == POOL_SCAN_NONE)
1715 error = spa_scan_stop(spa);
1717 error = spa_scan(spa, zc->zc_cookie);
1719 spa_close(spa, FTAG);
1725 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1730 error = spa_open(zc->zc_name, &spa, FTAG);
1733 spa_close(spa, FTAG);
1739 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1744 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1747 if (zc->zc_cookie < spa_version(spa) ||
1748 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1749 spa_close(spa, FTAG);
1750 return (SET_ERROR(EINVAL));
1753 spa_upgrade(spa, zc->zc_cookie);
1754 spa_close(spa, FTAG);
1760 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1767 if ((size = zc->zc_history_len) == 0)
1768 return (SET_ERROR(EINVAL));
1770 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1773 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1774 spa_close(spa, FTAG);
1775 return (SET_ERROR(ENOTSUP));
1778 hist_buf = kmem_alloc(size, KM_SLEEP);
1779 if ((error = spa_history_get(spa, &zc->zc_history_offset,
1780 &zc->zc_history_len, hist_buf)) == 0) {
1781 error = ddi_copyout(hist_buf,
1782 (void *)(uintptr_t)zc->zc_history,
1783 zc->zc_history_len, zc->zc_iflags);
1786 spa_close(spa, FTAG);
1787 kmem_free(hist_buf, size);
1792 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1797 error = spa_open(zc->zc_name, &spa, FTAG);
1799 error = spa_change_guid(spa);
1800 spa_close(spa, FTAG);
1806 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1808 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1813 * zc_name name of filesystem
1814 * zc_obj object to find
1817 * zc_value name of object
1820 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1825 /* XXX reading from objset not owned */
1826 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1828 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1829 dmu_objset_rele(os, FTAG);
1830 return (SET_ERROR(EINVAL));
1832 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1833 sizeof (zc->zc_value));
1834 dmu_objset_rele(os, FTAG);
1841 * zc_name name of filesystem
1842 * zc_obj object to find
1845 * zc_stat stats on object
1846 * zc_value path to object
1849 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1854 /* XXX reading from objset not owned */
1855 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1857 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1858 dmu_objset_rele(os, FTAG);
1859 return (SET_ERROR(EINVAL));
1861 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1862 sizeof (zc->zc_value));
1863 dmu_objset_rele(os, FTAG);
1869 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1873 nvlist_t *config, **l2cache, **spares;
1874 uint_t nl2cache = 0, nspares = 0;
1876 error = spa_open(zc->zc_name, &spa, FTAG);
1880 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1881 zc->zc_iflags, &config);
1882 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1883 &l2cache, &nl2cache);
1885 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1890 * A root pool with concatenated devices is not supported.
1891 * Thus, can not add a device to a root pool.
1893 * Intent log device can not be added to a rootpool because
1894 * during mountroot, zil is replayed, a seperated log device
1895 * can not be accessed during the mountroot time.
1897 * l2cache and spare devices are ok to be added to a rootpool.
1899 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1900 nvlist_free(config);
1901 spa_close(spa, FTAG);
1902 return (SET_ERROR(EDOM));
1904 #endif /* illumos */
1907 error = spa_vdev_add(spa, config);
1908 nvlist_free(config);
1910 spa_close(spa, FTAG);
1916 * zc_name name of the pool
1917 * zc_nvlist_conf nvlist of devices to remove
1918 * zc_cookie to stop the remove?
1921 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1926 error = spa_open(zc->zc_name, &spa, FTAG);
1929 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1930 spa_close(spa, FTAG);
1935 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1939 vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1941 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1943 switch (zc->zc_cookie) {
1944 case VDEV_STATE_ONLINE:
1945 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1948 case VDEV_STATE_OFFLINE:
1949 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1952 case VDEV_STATE_FAULTED:
1953 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1954 zc->zc_obj != VDEV_AUX_EXTERNAL)
1955 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1957 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1960 case VDEV_STATE_DEGRADED:
1961 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1962 zc->zc_obj != VDEV_AUX_EXTERNAL)
1963 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1965 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1969 error = SET_ERROR(EINVAL);
1971 zc->zc_cookie = newstate;
1972 spa_close(spa, FTAG);
1977 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1980 int replacing = zc->zc_cookie;
1984 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1987 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1988 zc->zc_iflags, &config)) == 0) {
1989 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1990 nvlist_free(config);
1993 spa_close(spa, FTAG);
1998 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
2003 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2006 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
2008 spa_close(spa, FTAG);
2013 zfs_ioc_vdev_split(zfs_cmd_t *zc)
2016 nvlist_t *config, *props = NULL;
2018 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
2020 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2023 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
2024 zc->zc_iflags, &config)) {
2025 spa_close(spa, FTAG);
2029 if (zc->zc_nvlist_src_size != 0 && (error =
2030 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2031 zc->zc_iflags, &props))) {
2032 spa_close(spa, FTAG);
2033 nvlist_free(config);
2037 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
2039 spa_close(spa, FTAG);
2041 nvlist_free(config);
2048 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
2051 char *path = zc->zc_value;
2052 uint64_t guid = zc->zc_guid;
2055 error = spa_open(zc->zc_name, &spa, FTAG);
2059 error = spa_vdev_setpath(spa, guid, path);
2060 spa_close(spa, FTAG);
2065 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
2068 char *fru = zc->zc_value;
2069 uint64_t guid = zc->zc_guid;
2072 error = spa_open(zc->zc_name, &spa, FTAG);
2076 error = spa_vdev_setfru(spa, guid, fru);
2077 spa_close(spa, FTAG);
2082 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
2087 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2089 if (zc->zc_nvlist_dst != 0 &&
2090 (error = dsl_prop_get_all(os, &nv)) == 0) {
2091 dmu_objset_stats(os, nv);
2093 * NB: zvol_get_stats() will read the objset contents,
2094 * which we aren't supposed to do with a
2095 * DS_MODE_USER hold, because it could be
2096 * inconsistent. So this is a bit of a workaround...
2097 * XXX reading with out owning
2099 if (!zc->zc_objset_stats.dds_inconsistent &&
2100 dmu_objset_type(os) == DMU_OST_ZVOL) {
2101 error = zvol_get_stats(os, nv);
2106 error = put_nvlist(zc, nv);
2115 * zc_name name of filesystem
2116 * zc_nvlist_dst_size size of buffer for property nvlist
2119 * zc_objset_stats stats
2120 * zc_nvlist_dst property nvlist
2121 * zc_nvlist_dst_size size of property nvlist
2124 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2129 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2131 error = zfs_ioc_objset_stats_impl(zc, os);
2132 dmu_objset_rele(os, FTAG);
2135 if (error == ENOMEM)
2142 * zc_name name of filesystem
2143 * zc_nvlist_dst_size size of buffer for property nvlist
2146 * zc_nvlist_dst received property nvlist
2147 * zc_nvlist_dst_size size of received property nvlist
2149 * Gets received properties (distinct from local properties on or after
2150 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2151 * local property values.
2154 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2160 * Without this check, we would return local property values if the
2161 * caller has not already received properties on or after
2162 * SPA_VERSION_RECVD_PROPS.
2164 if (!dsl_prop_get_hasrecvd(zc->zc_name))
2165 return (SET_ERROR(ENOTSUP));
2167 if (zc->zc_nvlist_dst != 0 &&
2168 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2169 error = put_nvlist(zc, nv);
2177 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2183 * zfs_get_zplprop() will either find a value or give us
2184 * the default value (if there is one).
2186 if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2188 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2194 * zc_name name of filesystem
2195 * zc_nvlist_dst_size size of buffer for zpl property nvlist
2198 * zc_nvlist_dst zpl property nvlist
2199 * zc_nvlist_dst_size size of zpl property nvlist
2202 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2207 /* XXX reading without owning */
2208 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os))
2211 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2214 * NB: nvl_add_zplprop() will read the objset contents,
2215 * which we aren't supposed to do with a DS_MODE_USER
2216 * hold, because it could be inconsistent.
2218 if (zc->zc_nvlist_dst != 0 &&
2219 !zc->zc_objset_stats.dds_inconsistent &&
2220 dmu_objset_type(os) == DMU_OST_ZFS) {
2223 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2224 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2225 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2226 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2227 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2228 err = put_nvlist(zc, nv);
2231 err = SET_ERROR(ENOENT);
2233 dmu_objset_rele(os, FTAG);
2238 dataset_name_hidden(const char *name)
2241 * Skip over datasets that are not visible in this zone,
2242 * internal datasets (which have a $ in their name), and
2243 * temporary datasets (which have a % in their name).
2245 if (strchr(name, '$') != NULL)
2247 if (strchr(name, '%') != NULL)
2249 if (!INGLOBALZONE(curthread) && !zone_dataset_visible(name, NULL))
2256 * zc_name name of filesystem
2257 * zc_cookie zap cursor
2258 * zc_nvlist_dst_size size of buffer for property nvlist
2261 * zc_name name of next filesystem
2262 * zc_cookie zap cursor
2263 * zc_objset_stats stats
2264 * zc_nvlist_dst property nvlist
2265 * zc_nvlist_dst_size size of property nvlist
2268 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2273 size_t orig_len = strlen(zc->zc_name);
2276 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) {
2277 if (error == ENOENT)
2278 error = SET_ERROR(ESRCH);
2282 p = strrchr(zc->zc_name, '/');
2283 if (p == NULL || p[1] != '\0')
2284 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2285 p = zc->zc_name + strlen(zc->zc_name);
2288 error = dmu_dir_list_next(os,
2289 sizeof (zc->zc_name) - (p - zc->zc_name), p,
2290 NULL, &zc->zc_cookie);
2291 if (error == ENOENT)
2292 error = SET_ERROR(ESRCH);
2293 } while (error == 0 && dataset_name_hidden(zc->zc_name));
2294 dmu_objset_rele(os, FTAG);
2297 * If it's an internal dataset (ie. with a '$' in its name),
2298 * don't try to get stats for it, otherwise we'll return ENOENT.
2300 if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2301 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2302 if (error == ENOENT) {
2303 /* We lost a race with destroy, get the next one. */
2304 zc->zc_name[orig_len] = '\0';
2313 * zc_name name of filesystem
2314 * zc_cookie zap cursor
2315 * zc_nvlist_dst_size size of buffer for property nvlist
2316 * zc_simple when set, only name is requested
2319 * zc_name name of next snapshot
2320 * zc_objset_stats stats
2321 * zc_nvlist_dst property nvlist
2322 * zc_nvlist_dst_size size of property nvlist
2325 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2330 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2332 return (error == ENOENT ? ESRCH : error);
2336 * A dataset name of maximum length cannot have any snapshots,
2337 * so exit immediately.
2339 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
2340 dmu_objset_rele(os, FTAG);
2341 return (SET_ERROR(ESRCH));
2344 error = dmu_snapshot_list_next(os,
2345 sizeof (zc->zc_name) - strlen(zc->zc_name),
2346 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2349 if (error == 0 && !zc->zc_simple) {
2351 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2353 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2357 error = dmu_objset_from_ds(ds, &ossnap);
2359 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2360 dsl_dataset_rele(ds, FTAG);
2362 } else if (error == ENOENT) {
2363 error = SET_ERROR(ESRCH);
2366 dmu_objset_rele(os, FTAG);
2367 /* if we failed, undo the @ that we tacked on to zc_name */
2369 *strchr(zc->zc_name, '@') = '\0';
2374 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2376 const char *propname = nvpair_name(pair);
2378 unsigned int vallen;
2381 zfs_userquota_prop_t type;
2387 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2389 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2390 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2392 return (SET_ERROR(EINVAL));
2396 * A correctly constructed propname is encoded as
2397 * userquota@<rid>-<domain>.
2399 if ((dash = strchr(propname, '-')) == NULL ||
2400 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2402 return (SET_ERROR(EINVAL));
2409 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2411 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2412 zfsvfs_rele(zfsvfs, FTAG);
2419 * If the named property is one that has a special function to set its value,
2420 * return 0 on success and a positive error code on failure; otherwise if it is
2421 * not one of the special properties handled by this function, return -1.
2423 * XXX: It would be better for callers of the property interface if we handled
2424 * these special cases in dsl_prop.c (in the dsl layer).
2427 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2430 const char *propname = nvpair_name(pair);
2431 zfs_prop_t prop = zfs_name_to_prop(propname);
2435 if (prop == ZPROP_INVAL) {
2436 if (zfs_prop_userquota(propname))
2437 return (zfs_prop_set_userquota(dsname, pair));
2441 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2443 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2444 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2448 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2451 VERIFY(0 == nvpair_value_uint64(pair, &intval));
2454 case ZFS_PROP_QUOTA:
2455 err = dsl_dir_set_quota(dsname, source, intval);
2457 case ZFS_PROP_REFQUOTA:
2458 err = dsl_dataset_set_refquota(dsname, source, intval);
2460 case ZFS_PROP_FILESYSTEM_LIMIT:
2461 case ZFS_PROP_SNAPSHOT_LIMIT:
2462 if (intval == UINT64_MAX) {
2463 /* clearing the limit, just do it */
2466 err = dsl_dir_activate_fs_ss_limit(dsname);
2469 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2470 * default path to set the value in the nvlist.
2475 case ZFS_PROP_RESERVATION:
2476 err = dsl_dir_set_reservation(dsname, source, intval);
2478 case ZFS_PROP_REFRESERVATION:
2479 err = dsl_dataset_set_refreservation(dsname, source, intval);
2481 case ZFS_PROP_VOLSIZE:
2482 err = zvol_set_volsize(dsname, intval);
2484 case ZFS_PROP_VERSION:
2488 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2491 err = zfs_set_version(zfsvfs, intval);
2492 zfsvfs_rele(zfsvfs, FTAG);
2494 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2497 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2498 (void) strcpy(zc->zc_name, dsname);
2499 (void) zfs_ioc_userspace_upgrade(zc);
2500 kmem_free(zc, sizeof (zfs_cmd_t));
2512 * This function is best effort. If it fails to set any of the given properties,
2513 * it continues to set as many as it can and returns the last error
2514 * encountered. If the caller provides a non-NULL errlist, it will be filled in
2515 * with the list of names of all the properties that failed along with the
2516 * corresponding error numbers.
2518 * If every property is set successfully, zero is returned and errlist is not
2522 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2530 nvlist_t *genericnvl = fnvlist_alloc();
2531 nvlist_t *retrynvl = fnvlist_alloc();
2535 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2536 const char *propname = nvpair_name(pair);
2537 zfs_prop_t prop = zfs_name_to_prop(propname);
2540 /* decode the property value */
2542 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2544 attrs = fnvpair_value_nvlist(pair);
2545 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2547 err = SET_ERROR(EINVAL);
2550 /* Validate value type */
2551 if (err == 0 && prop == ZPROP_INVAL) {
2552 if (zfs_prop_user(propname)) {
2553 if (nvpair_type(propval) != DATA_TYPE_STRING)
2554 err = SET_ERROR(EINVAL);
2555 } else if (zfs_prop_userquota(propname)) {
2556 if (nvpair_type(propval) !=
2557 DATA_TYPE_UINT64_ARRAY)
2558 err = SET_ERROR(EINVAL);
2560 err = SET_ERROR(EINVAL);
2562 } else if (err == 0) {
2563 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2564 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2565 err = SET_ERROR(EINVAL);
2566 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2569 intval = fnvpair_value_uint64(propval);
2571 switch (zfs_prop_get_type(prop)) {
2572 case PROP_TYPE_NUMBER:
2574 case PROP_TYPE_STRING:
2575 err = SET_ERROR(EINVAL);
2577 case PROP_TYPE_INDEX:
2578 if (zfs_prop_index_to_string(prop,
2579 intval, &unused) != 0)
2580 err = SET_ERROR(EINVAL);
2584 "unknown property type");
2587 err = SET_ERROR(EINVAL);
2591 /* Validate permissions */
2593 err = zfs_check_settable(dsname, pair, CRED());
2596 err = zfs_prop_set_special(dsname, source, pair);
2599 * For better performance we build up a list of
2600 * properties to set in a single transaction.
2602 err = nvlist_add_nvpair(genericnvl, pair);
2603 } else if (err != 0 && nvl != retrynvl) {
2605 * This may be a spurious error caused by
2606 * receiving quota and reservation out of order.
2607 * Try again in a second pass.
2609 err = nvlist_add_nvpair(retrynvl, pair);
2614 if (errlist != NULL)
2615 fnvlist_add_int32(errlist, propname, err);
2620 if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2625 if (!nvlist_empty(genericnvl) &&
2626 dsl_props_set(dsname, source, genericnvl) != 0) {
2628 * If this fails, we still want to set as many properties as we
2629 * can, so try setting them individually.
2632 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2633 const char *propname = nvpair_name(pair);
2637 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2639 attrs = fnvpair_value_nvlist(pair);
2640 propval = fnvlist_lookup_nvpair(attrs,
2644 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2645 strval = fnvpair_value_string(propval);
2646 err = dsl_prop_set_string(dsname, propname,
2649 intval = fnvpair_value_uint64(propval);
2650 err = dsl_prop_set_int(dsname, propname, source,
2655 if (errlist != NULL) {
2656 fnvlist_add_int32(errlist, propname,
2663 nvlist_free(genericnvl);
2664 nvlist_free(retrynvl);
2670 * Check that all the properties are valid user properties.
2673 zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2675 nvpair_t *pair = NULL;
2678 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2679 const char *propname = nvpair_name(pair);
2681 if (!zfs_prop_user(propname) ||
2682 nvpair_type(pair) != DATA_TYPE_STRING)
2683 return (SET_ERROR(EINVAL));
2685 if (error = zfs_secpolicy_write_perms(fsname,
2686 ZFS_DELEG_PERM_USERPROP, CRED()))
2689 if (strlen(propname) >= ZAP_MAXNAMELEN)
2690 return (SET_ERROR(ENAMETOOLONG));
2692 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN)
2699 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2703 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2706 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2707 if (nvlist_exists(skipped, nvpair_name(pair)))
2710 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2715 clear_received_props(const char *dsname, nvlist_t *props,
2719 nvlist_t *cleared_props = NULL;
2720 props_skip(props, skipped, &cleared_props);
2721 if (!nvlist_empty(cleared_props)) {
2723 * Acts on local properties until the dataset has received
2724 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2726 zprop_source_t flags = (ZPROP_SRC_NONE |
2727 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2728 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2730 nvlist_free(cleared_props);
2736 * zc_name name of filesystem
2737 * zc_value name of property to set
2738 * zc_nvlist_src{_size} nvlist of properties to apply
2739 * zc_cookie received properties flag
2742 * zc_nvlist_dst{_size} error for each unapplied received property
2745 zfs_ioc_set_prop(zfs_cmd_t *zc)
2748 boolean_t received = zc->zc_cookie;
2749 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2754 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2755 zc->zc_iflags, &nvl)) != 0)
2759 nvlist_t *origprops;
2761 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2762 (void) clear_received_props(zc->zc_name,
2764 nvlist_free(origprops);
2767 error = dsl_prop_set_hasrecvd(zc->zc_name);
2770 errors = fnvlist_alloc();
2772 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2774 if (zc->zc_nvlist_dst != 0 && errors != NULL) {
2775 (void) put_nvlist(zc, errors);
2778 nvlist_free(errors);
2785 * zc_name name of filesystem
2786 * zc_value name of property to inherit
2787 * zc_cookie revert to received value if TRUE
2792 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2794 const char *propname = zc->zc_value;
2795 zfs_prop_t prop = zfs_name_to_prop(propname);
2796 boolean_t received = zc->zc_cookie;
2797 zprop_source_t source = (received
2798 ? ZPROP_SRC_NONE /* revert to received value, if any */
2799 : ZPROP_SRC_INHERITED); /* explicitly inherit */
2808 * zfs_prop_set_special() expects properties in the form of an
2809 * nvpair with type info.
2811 if (prop == ZPROP_INVAL) {
2812 if (!zfs_prop_user(propname))
2813 return (SET_ERROR(EINVAL));
2815 type = PROP_TYPE_STRING;
2816 } else if (prop == ZFS_PROP_VOLSIZE ||
2817 prop == ZFS_PROP_VERSION) {
2818 return (SET_ERROR(EINVAL));
2820 type = zfs_prop_get_type(prop);
2823 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2826 case PROP_TYPE_STRING:
2827 VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2829 case PROP_TYPE_NUMBER:
2830 case PROP_TYPE_INDEX:
2831 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2835 return (SET_ERROR(EINVAL));
2838 pair = nvlist_next_nvpair(dummy, NULL);
2839 err = zfs_prop_set_special(zc->zc_name, source, pair);
2842 return (err); /* special property already handled */
2845 * Only check this in the non-received case. We want to allow
2846 * 'inherit -S' to revert non-inheritable properties like quota
2847 * and reservation to the received or default values even though
2848 * they are not considered inheritable.
2850 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2851 return (SET_ERROR(EINVAL));
2854 /* property name has been validated by zfs_secpolicy_inherit_prop() */
2855 return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source));
2859 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2866 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2867 zc->zc_iflags, &props))
2871 * If the only property is the configfile, then just do a spa_lookup()
2872 * to handle the faulted case.
2874 pair = nvlist_next_nvpair(props, NULL);
2875 if (pair != NULL && strcmp(nvpair_name(pair),
2876 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2877 nvlist_next_nvpair(props, pair) == NULL) {
2878 mutex_enter(&spa_namespace_lock);
2879 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2880 spa_configfile_set(spa, props, B_FALSE);
2881 spa_config_sync(spa, B_FALSE, B_TRUE);
2883 mutex_exit(&spa_namespace_lock);
2890 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2895 error = spa_prop_set(spa, props);
2898 spa_close(spa, FTAG);
2904 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2908 nvlist_t *nvp = NULL;
2910 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2912 * If the pool is faulted, there may be properties we can still
2913 * get (such as altroot and cachefile), so attempt to get them
2916 mutex_enter(&spa_namespace_lock);
2917 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2918 error = spa_prop_get(spa, &nvp);
2919 mutex_exit(&spa_namespace_lock);
2921 error = spa_prop_get(spa, &nvp);
2922 spa_close(spa, FTAG);
2925 if (error == 0 && zc->zc_nvlist_dst != 0)
2926 error = put_nvlist(zc, nvp);
2928 error = SET_ERROR(EFAULT);
2936 * zc_name name of filesystem
2937 * zc_nvlist_src{_size} nvlist of delegated permissions
2938 * zc_perm_action allow/unallow flag
2943 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2946 nvlist_t *fsaclnv = NULL;
2948 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2949 zc->zc_iflags, &fsaclnv)) != 0)
2953 * Verify nvlist is constructed correctly
2955 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2956 nvlist_free(fsaclnv);
2957 return (SET_ERROR(EINVAL));
2961 * If we don't have PRIV_SYS_MOUNT, then validate
2962 * that user is allowed to hand out each permission in
2966 error = secpolicy_zfs(CRED());
2968 if (zc->zc_perm_action == B_FALSE) {
2969 error = dsl_deleg_can_allow(zc->zc_name,
2972 error = dsl_deleg_can_unallow(zc->zc_name,
2978 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2980 nvlist_free(fsaclnv);
2986 * zc_name name of filesystem
2989 * zc_nvlist_src{_size} nvlist of delegated permissions
2992 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
2997 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
2998 error = put_nvlist(zc, nvp);
3006 * Search the vfs list for a specified resource. Returns a pointer to it
3007 * or NULL if no suitable entry is found. The caller of this routine
3008 * is responsible for releasing the returned vfs pointer.
3011 zfs_get_vfs(const char *resource)
3015 mtx_lock(&mountlist_mtx);
3016 TAILQ_FOREACH(vfsp, &mountlist, mnt_list) {
3017 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
3022 mtx_unlock(&mountlist_mtx);
3028 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
3030 zfs_creat_t *zct = arg;
3032 zfs_create_fs(os, cr, zct->zct_zplprops, tx);
3035 #define ZFS_PROP_UNDEFINED ((uint64_t)-1)
3039 * os parent objset pointer (NULL if root fs)
3040 * fuids_ok fuids allowed in this version of the spa?
3041 * sa_ok SAs allowed in this version of the spa?
3042 * createprops list of properties requested by creator
3045 * zplprops values for the zplprops we attach to the master node object
3046 * is_ci true if requested file system will be purely case-insensitive
3048 * Determine the settings for utf8only, normalization and
3049 * casesensitivity. Specific values may have been requested by the
3050 * creator and/or we can inherit values from the parent dataset. If
3051 * the file system is of too early a vintage, a creator can not
3052 * request settings for these properties, even if the requested
3053 * setting is the default value. We don't actually want to create dsl
3054 * properties for these, so remove them from the source nvlist after
3058 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
3059 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
3060 nvlist_t *zplprops, boolean_t *is_ci)
3062 uint64_t sense = ZFS_PROP_UNDEFINED;
3063 uint64_t norm = ZFS_PROP_UNDEFINED;
3064 uint64_t u8 = ZFS_PROP_UNDEFINED;
3066 ASSERT(zplprops != NULL);
3069 * Pull out creator prop choices, if any.
3072 (void) nvlist_lookup_uint64(createprops,
3073 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
3074 (void) nvlist_lookup_uint64(createprops,
3075 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
3076 (void) nvlist_remove_all(createprops,
3077 zfs_prop_to_name(ZFS_PROP_NORMALIZE));
3078 (void) nvlist_lookup_uint64(createprops,
3079 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
3080 (void) nvlist_remove_all(createprops,
3081 zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
3082 (void) nvlist_lookup_uint64(createprops,
3083 zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3084 (void) nvlist_remove_all(createprops,
3085 zfs_prop_to_name(ZFS_PROP_CASE));
3089 * If the zpl version requested is whacky or the file system
3090 * or pool is version is too "young" to support normalization
3091 * and the creator tried to set a value for one of the props,
3094 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3095 (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3096 (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3097 (zplver < ZPL_VERSION_NORMALIZATION &&
3098 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3099 sense != ZFS_PROP_UNDEFINED)))
3100 return (SET_ERROR(ENOTSUP));
3103 * Put the version in the zplprops
3105 VERIFY(nvlist_add_uint64(zplprops,
3106 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3108 if (norm == ZFS_PROP_UNDEFINED)
3109 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
3110 VERIFY(nvlist_add_uint64(zplprops,
3111 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3114 * If we're normalizing, names must always be valid UTF-8 strings.
3118 if (u8 == ZFS_PROP_UNDEFINED)
3119 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
3120 VERIFY(nvlist_add_uint64(zplprops,
3121 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3123 if (sense == ZFS_PROP_UNDEFINED)
3124 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
3125 VERIFY(nvlist_add_uint64(zplprops,
3126 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3129 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
3135 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3136 nvlist_t *zplprops, boolean_t *is_ci)
3138 boolean_t fuids_ok, sa_ok;
3139 uint64_t zplver = ZPL_VERSION;
3140 objset_t *os = NULL;
3141 char parentname[MAXNAMELEN];
3147 (void) strlcpy(parentname, dataset, sizeof (parentname));
3148 cp = strrchr(parentname, '/');
3152 if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3155 spa_vers = spa_version(spa);
3156 spa_close(spa, FTAG);
3158 zplver = zfs_zpl_version_map(spa_vers);
3159 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3160 sa_ok = (zplver >= ZPL_VERSION_SA);
3163 * Open parent object set so we can inherit zplprop values.
3165 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3168 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3170 dmu_objset_rele(os, FTAG);
3175 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3176 nvlist_t *zplprops, boolean_t *is_ci)
3180 uint64_t zplver = ZPL_VERSION;
3183 zplver = zfs_zpl_version_map(spa_vers);
3184 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3185 sa_ok = (zplver >= ZPL_VERSION_SA);
3187 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3188 createprops, zplprops, is_ci);
3194 * "type" -> dmu_objset_type_t (int32)
3195 * (optional) "props" -> { prop -> value }
3198 * outnvl: propname -> error code (int32)
3201 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3204 zfs_creat_t zct = { 0 };
3205 nvlist_t *nvprops = NULL;
3206 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3208 dmu_objset_type_t type;
3209 boolean_t is_insensitive = B_FALSE;
3211 if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3212 return (SET_ERROR(EINVAL));
3214 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3218 cbfunc = zfs_create_cb;
3222 cbfunc = zvol_create_cb;
3229 if (strchr(fsname, '@') ||
3230 strchr(fsname, '%'))
3231 return (SET_ERROR(EINVAL));
3233 zct.zct_props = nvprops;
3236 return (SET_ERROR(EINVAL));
3238 if (type == DMU_OST_ZVOL) {
3239 uint64_t volsize, volblocksize;
3241 if (nvprops == NULL)
3242 return (SET_ERROR(EINVAL));
3243 if (nvlist_lookup_uint64(nvprops,
3244 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3245 return (SET_ERROR(EINVAL));
3247 if ((error = nvlist_lookup_uint64(nvprops,
3248 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3249 &volblocksize)) != 0 && error != ENOENT)
3250 return (SET_ERROR(EINVAL));
3253 volblocksize = zfs_prop_default_numeric(
3254 ZFS_PROP_VOLBLOCKSIZE);
3256 if ((error = zvol_check_volblocksize(
3257 volblocksize)) != 0 ||
3258 (error = zvol_check_volsize(volsize,
3259 volblocksize)) != 0)
3261 } else if (type == DMU_OST_ZFS) {
3265 * We have to have normalization and
3266 * case-folding flags correct when we do the
3267 * file system creation, so go figure them out
3270 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3271 NV_UNIQUE_NAME, KM_SLEEP) == 0);
3272 error = zfs_fill_zplprops(fsname, nvprops,
3273 zct.zct_zplprops, &is_insensitive);
3275 nvlist_free(zct.zct_zplprops);
3280 error = dmu_objset_create(fsname, type,
3281 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3282 nvlist_free(zct.zct_zplprops);
3285 * It would be nice to do this atomically.
3288 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3291 (void) dsl_destroy_head(fsname);
3294 if (error == 0 && type == DMU_OST_ZVOL)
3295 zvol_create_minors(fsname);
3302 * "origin" -> name of origin snapshot
3303 * (optional) "props" -> { prop -> value }
3306 * outnvl: propname -> error code (int32)
3309 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3312 nvlist_t *nvprops = NULL;
3315 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3316 return (SET_ERROR(EINVAL));
3317 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3319 if (strchr(fsname, '@') ||
3320 strchr(fsname, '%'))
3321 return (SET_ERROR(EINVAL));
3323 if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3324 return (SET_ERROR(EINVAL));
3325 error = dmu_objset_clone(fsname, origin_name);
3330 * It would be nice to do this atomically.
3333 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3336 (void) dsl_destroy_head(fsname);
3340 zvol_create_minors(fsname);
3347 * "snaps" -> { snapshot1, snapshot2 }
3348 * (optional) "props" -> { prop -> value (string) }
3351 * outnvl: snapshot -> error code (int32)
3354 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3357 nvlist_t *props = NULL;
3361 (void) nvlist_lookup_nvlist(innvl, "props", &props);
3362 if ((error = zfs_check_userprops(poolname, props)) != 0)
3365 if (!nvlist_empty(props) &&
3366 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3367 return (SET_ERROR(ENOTSUP));
3369 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3370 return (SET_ERROR(EINVAL));
3371 poollen = strlen(poolname);
3372 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3373 pair = nvlist_next_nvpair(snaps, pair)) {
3374 const char *name = nvpair_name(pair);
3375 const char *cp = strchr(name, '@');
3378 * The snap name must contain an @, and the part after it must
3379 * contain only valid characters.
3382 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3383 return (SET_ERROR(EINVAL));
3386 * The snap must be in the specified pool.
3388 if (strncmp(name, poolname, poollen) != 0 ||
3389 (name[poollen] != '/' && name[poollen] != '@'))
3390 return (SET_ERROR(EXDEV));
3392 /* This must be the only snap of this fs. */
3393 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3394 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3395 if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3397 return (SET_ERROR(EXDEV));
3402 error = dsl_dataset_snapshot(snaps, props, outnvl);
3407 * innvl: "message" -> string
3411 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3419 * The poolname in the ioctl is not set, we get it from the TSD,
3420 * which was set at the end of the last successful ioctl that allows
3421 * logging. The secpolicy func already checked that it is set.
3422 * Only one log ioctl is allowed after each successful ioctl, so
3423 * we clear the TSD here.
3425 poolname = tsd_get(zfs_allow_log_key);
3426 (void) tsd_set(zfs_allow_log_key, NULL);
3427 error = spa_open(poolname, &spa, FTAG);
3432 if (nvlist_lookup_string(innvl, "message", &message) != 0) {
3433 spa_close(spa, FTAG);
3434 return (SET_ERROR(EINVAL));
3437 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3438 spa_close(spa, FTAG);
3439 return (SET_ERROR(ENOTSUP));
3442 error = spa_history_log(spa, message);
3443 spa_close(spa, FTAG);
3448 * The dp_config_rwlock must not be held when calling this, because the
3449 * unmount may need to write out data.
3451 * This function is best-effort. Callers must deal gracefully if it
3452 * remains mounted (or is remounted after this call).
3454 * Returns 0 if the argument is not a snapshot, or it is not currently a
3455 * filesystem, or we were able to unmount it. Returns error code otherwise.
3458 zfs_unmount_snap(const char *snapname)
3464 if (strchr(snapname, '@') == NULL)
3467 vfsp = zfs_get_vfs(snapname);
3471 zfsvfs = vfsp->vfs_data;
3472 ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os)));
3474 err = vn_vfswlock(vfsp->vfs_vnodecovered);
3477 return (SET_ERROR(err));
3480 * Always force the unmount for snapshots.
3484 (void) dounmount(vfsp, MS_FORCE, kcred);
3487 (void) dounmount(vfsp, MS_FORCE, curthread);
3494 zfs_unmount_snap_cb(const char *snapname, void *arg)
3496 return (zfs_unmount_snap(snapname));
3500 * When a clone is destroyed, its origin may also need to be destroyed,
3501 * in which case it must be unmounted. This routine will do that unmount
3505 zfs_destroy_unmount_origin(const char *fsname)
3511 error = dmu_objset_hold(fsname, FTAG, &os);
3514 ds = dmu_objset_ds(os);
3515 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3516 char originname[MAXNAMELEN];
3517 dsl_dataset_name(ds->ds_prev, originname);
3518 dmu_objset_rele(os, FTAG);
3519 (void) zfs_unmount_snap(originname);
3521 dmu_objset_rele(os, FTAG);
3527 * "snaps" -> { snapshot1, snapshot2 }
3528 * (optional boolean) "defer"
3531 * outnvl: snapshot -> error code (int32)
3536 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3543 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3544 return (SET_ERROR(EINVAL));
3545 defer = nvlist_exists(innvl, "defer");
3547 poollen = strlen(poolname);
3548 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3549 pair = nvlist_next_nvpair(snaps, pair)) {
3550 const char *name = nvpair_name(pair);
3553 * The snap must be in the specified pool to prevent the
3554 * invalid removal of zvol minors below.
3556 if (strncmp(name, poolname, poollen) != 0 ||
3557 (name[poollen] != '/' && name[poollen] != '@'))
3558 return (SET_ERROR(EXDEV));
3560 error = zfs_unmount_snap(name);
3563 #if defined(__FreeBSD__)
3564 zvol_remove_minors(name);
3568 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3572 * Create bookmarks. Bookmark names are of the form <fs>#<bmark>.
3573 * All bookmarks must be in the same pool.
3576 * bookmark1 -> snapshot1, bookmark2 -> snapshot2
3579 * outnvl: bookmark -> error code (int32)
3584 zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3586 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3587 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3591 * Verify the snapshot argument.
3593 if (nvpair_value_string(pair, &snap_name) != 0)
3594 return (SET_ERROR(EINVAL));
3597 /* Verify that the keys (bookmarks) are unique */
3598 for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair);
3599 pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) {
3600 if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0)
3601 return (SET_ERROR(EINVAL));
3605 return (dsl_bookmark_create(innvl, outnvl));
3610 * property 1, property 2, ...
3614 * bookmark name 1 -> { property 1, property 2, ... },
3615 * bookmark name 2 -> { property 1, property 2, ... }
3620 zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3622 return (dsl_get_bookmarks(fsname, innvl, outnvl));
3627 * bookmark name 1, bookmark name 2
3630 * outnvl: bookmark -> error code (int32)
3634 zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl,
3639 poollen = strlen(poolname);
3640 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3641 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3642 const char *name = nvpair_name(pair);
3643 const char *cp = strchr(name, '#');
3646 * The bookmark name must contain an #, and the part after it
3647 * must contain only valid characters.
3650 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3651 return (SET_ERROR(EINVAL));
3654 * The bookmark must be in the specified pool.
3656 if (strncmp(name, poolname, poollen) != 0 ||
3657 (name[poollen] != '/' && name[poollen] != '#'))
3658 return (SET_ERROR(EXDEV));
3661 error = dsl_bookmark_destroy(innvl, outnvl);
3667 * zc_name name of dataset to destroy
3668 * zc_objset_type type of objset
3669 * zc_defer_destroy mark for deferred destroy
3674 zfs_ioc_destroy(zfs_cmd_t *zc)
3678 if (zc->zc_objset_type == DMU_OST_ZFS) {
3679 err = zfs_unmount_snap(zc->zc_name);
3684 if (strchr(zc->zc_name, '@'))
3685 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3687 err = dsl_destroy_head(zc->zc_name);
3688 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3690 zvol_remove_minors(zc->zc_name);
3692 (void) zvol_remove_minor(zc->zc_name);
3698 * fsname is name of dataset to rollback (to most recent snapshot)
3700 * innvl is not used.
3702 * outnvl: "target" -> name of most recent snapshot
3707 zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl)
3712 if (getzfsvfs(fsname, &zfsvfs) == 0) {
3713 error = zfs_suspend_fs(zfsvfs);
3717 error = dsl_dataset_rollback(fsname, zfsvfs, outnvl);
3718 resume_err = zfs_resume_fs(zfsvfs, fsname);
3719 error = error ? error : resume_err;
3721 VFS_RELE(zfsvfs->z_vfs);
3723 error = dsl_dataset_rollback(fsname, NULL, outnvl);
3729 recursive_unmount(const char *fsname, void *arg)
3731 const char *snapname = arg;
3732 char fullname[MAXNAMELEN];
3734 (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname);
3735 return (zfs_unmount_snap(fullname));
3740 * zc_name old name of dataset
3741 * zc_value new name of dataset
3742 * zc_cookie recursive flag (only valid for snapshots)
3747 zfs_ioc_rename(zfs_cmd_t *zc)
3749 boolean_t recursive = zc->zc_cookie & 1;
3751 boolean_t allow_mounted = B_TRUE;
3754 allow_mounted = (zc->zc_cookie & 2) != 0;
3757 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3758 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3759 strchr(zc->zc_value, '%'))
3760 return (SET_ERROR(EINVAL));
3762 at = strchr(zc->zc_name, '@');
3764 /* snaps must be in same fs */
3767 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
3768 return (SET_ERROR(EXDEV));
3770 if (zc->zc_objset_type == DMU_OST_ZFS && allow_mounted) {
3771 error = dmu_objset_find(zc->zc_name,
3772 recursive_unmount, at + 1,
3773 recursive ? DS_FIND_CHILDREN : 0);
3779 error = dsl_dataset_rename_snapshot(zc->zc_name,
3780 at + 1, strchr(zc->zc_value, '@') + 1, recursive);
3786 if (zc->zc_objset_type == DMU_OST_ZVOL)
3787 (void) zvol_remove_minor(zc->zc_name);
3789 return (dsl_dir_rename(zc->zc_name, zc->zc_value));
3794 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3796 const char *propname = nvpair_name(pair);
3797 boolean_t issnap = (strchr(dsname, '@') != NULL);
3798 zfs_prop_t prop = zfs_name_to_prop(propname);
3802 if (prop == ZPROP_INVAL) {
3803 if (zfs_prop_user(propname)) {
3804 if (err = zfs_secpolicy_write_perms(dsname,
3805 ZFS_DELEG_PERM_USERPROP, cr))
3810 if (!issnap && zfs_prop_userquota(propname)) {
3811 const char *perm = NULL;
3812 const char *uq_prefix =
3813 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3814 const char *gq_prefix =
3815 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3817 if (strncmp(propname, uq_prefix,
3818 strlen(uq_prefix)) == 0) {
3819 perm = ZFS_DELEG_PERM_USERQUOTA;
3820 } else if (strncmp(propname, gq_prefix,
3821 strlen(gq_prefix)) == 0) {
3822 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3824 /* USERUSED and GROUPUSED are read-only */
3825 return (SET_ERROR(EINVAL));
3828 if (err = zfs_secpolicy_write_perms(dsname, perm, cr))
3833 return (SET_ERROR(EINVAL));
3837 return (SET_ERROR(EINVAL));
3839 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3841 * dsl_prop_get_all_impl() returns properties in this
3845 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3846 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3851 * Check that this value is valid for this pool version
3854 case ZFS_PROP_COMPRESSION:
3856 * If the user specified gzip compression, make sure
3857 * the SPA supports it. We ignore any errors here since
3858 * we'll catch them later.
3860 if (nvpair_value_uint64(pair, &intval) == 0) {
3861 if (intval >= ZIO_COMPRESS_GZIP_1 &&
3862 intval <= ZIO_COMPRESS_GZIP_9 &&
3863 zfs_earlier_version(dsname,
3864 SPA_VERSION_GZIP_COMPRESSION)) {
3865 return (SET_ERROR(ENOTSUP));
3868 if (intval == ZIO_COMPRESS_ZLE &&
3869 zfs_earlier_version(dsname,
3870 SPA_VERSION_ZLE_COMPRESSION))
3871 return (SET_ERROR(ENOTSUP));
3873 if (intval == ZIO_COMPRESS_LZ4) {
3876 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3879 if (!spa_feature_is_enabled(spa,
3880 SPA_FEATURE_LZ4_COMPRESS)) {
3881 spa_close(spa, FTAG);
3882 return (SET_ERROR(ENOTSUP));
3884 spa_close(spa, FTAG);
3888 * If this is a bootable dataset then
3889 * verify that the compression algorithm
3890 * is supported for booting. We must return
3891 * something other than ENOTSUP since it
3892 * implies a downrev pool version.
3894 if (zfs_is_bootfs(dsname) &&
3895 !BOOTFS_COMPRESS_VALID(intval)) {
3896 return (SET_ERROR(ERANGE));
3901 case ZFS_PROP_COPIES:
3902 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3903 return (SET_ERROR(ENOTSUP));
3906 case ZFS_PROP_DEDUP:
3907 if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3908 return (SET_ERROR(ENOTSUP));
3911 case ZFS_PROP_RECORDSIZE:
3912 /* Record sizes above 128k need the feature to be enabled */
3913 if (nvpair_value_uint64(pair, &intval) == 0 &&
3914 intval > SPA_OLD_MAXBLOCKSIZE) {
3918 * If this is a bootable dataset then
3919 * the we don't allow large (>128K) blocks,
3920 * because GRUB doesn't support them.
3922 if (zfs_is_bootfs(dsname) &&
3923 intval > SPA_OLD_MAXBLOCKSIZE) {
3924 return (SET_ERROR(EDOM));
3928 * We don't allow setting the property above 1MB,
3929 * unless the tunable has been changed.
3931 if (intval > zfs_max_recordsize ||
3932 intval > SPA_MAXBLOCKSIZE)
3933 return (SET_ERROR(EDOM));
3935 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3938 if (!spa_feature_is_enabled(spa,
3939 SPA_FEATURE_LARGE_BLOCKS)) {
3940 spa_close(spa, FTAG);
3941 return (SET_ERROR(ENOTSUP));
3943 spa_close(spa, FTAG);
3947 case ZFS_PROP_SHARESMB:
3948 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3949 return (SET_ERROR(ENOTSUP));
3952 case ZFS_PROP_ACLINHERIT:
3953 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3954 nvpair_value_uint64(pair, &intval) == 0) {
3955 if (intval == ZFS_ACL_PASSTHROUGH_X &&
3956 zfs_earlier_version(dsname,
3957 SPA_VERSION_PASSTHROUGH_X))
3958 return (SET_ERROR(ENOTSUP));
3963 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3967 * Checks for a race condition to make sure we don't increment a feature flag
3971 zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx)
3973 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3974 spa_feature_t *featurep = arg;
3976 if (!spa_feature_is_active(spa, *featurep))
3979 return (SET_ERROR(EBUSY));
3983 * The callback invoked on feature activation in the sync task caused by
3984 * zfs_prop_activate_feature.
3987 zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx)
3989 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3990 spa_feature_t *featurep = arg;
3992 spa_feature_incr(spa, *featurep, tx);
3996 * Activates a feature on a pool in response to a property setting. This
3997 * creates a new sync task which modifies the pool to reflect the feature
4001 zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature)
4005 /* EBUSY here indicates that the feature is already active */
4006 err = dsl_sync_task(spa_name(spa),
4007 zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync,
4008 &feature, 2, ZFS_SPACE_CHECK_RESERVED);
4010 if (err != 0 && err != EBUSY)
4017 * Removes properties from the given props list that fail permission checks
4018 * needed to clear them and to restore them in case of a receive error. For each
4019 * property, make sure we have both set and inherit permissions.
4021 * Returns the first error encountered if any permission checks fail. If the
4022 * caller provides a non-NULL errlist, it also gives the complete list of names
4023 * of all the properties that failed a permission check along with the
4024 * corresponding error numbers. The caller is responsible for freeing the
4027 * If every property checks out successfully, zero is returned and the list
4028 * pointed at by errlist is NULL.
4031 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
4034 nvpair_t *pair, *next_pair;
4041 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4043 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
4044 (void) strcpy(zc->zc_name, dataset);
4045 pair = nvlist_next_nvpair(props, NULL);
4046 while (pair != NULL) {
4047 next_pair = nvlist_next_nvpair(props, pair);
4049 (void) strcpy(zc->zc_value, nvpair_name(pair));
4050 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
4051 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
4052 VERIFY(nvlist_remove_nvpair(props, pair) == 0);
4053 VERIFY(nvlist_add_int32(errors,
4054 zc->zc_value, err) == 0);
4058 kmem_free(zc, sizeof (zfs_cmd_t));
4060 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
4061 nvlist_free(errors);
4064 VERIFY(nvpair_value_int32(pair, &rv) == 0);
4067 if (errlist == NULL)
4068 nvlist_free(errors);
4076 propval_equals(nvpair_t *p1, nvpair_t *p2)
4078 if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
4079 /* dsl_prop_get_all_impl() format */
4081 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
4082 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4086 if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
4088 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
4089 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4093 if (nvpair_type(p1) != nvpair_type(p2))
4096 if (nvpair_type(p1) == DATA_TYPE_STRING) {
4097 char *valstr1, *valstr2;
4099 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
4100 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
4101 return (strcmp(valstr1, valstr2) == 0);
4103 uint64_t intval1, intval2;
4105 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
4106 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
4107 return (intval1 == intval2);
4112 * Remove properties from props if they are not going to change (as determined
4113 * by comparison with origprops). Remove them from origprops as well, since we
4114 * do not need to clear or restore properties that won't change.
4117 props_reduce(nvlist_t *props, nvlist_t *origprops)
4119 nvpair_t *pair, *next_pair;
4121 if (origprops == NULL)
4122 return; /* all props need to be received */
4124 pair = nvlist_next_nvpair(props, NULL);
4125 while (pair != NULL) {
4126 const char *propname = nvpair_name(pair);
4129 next_pair = nvlist_next_nvpair(props, pair);
4131 if ((nvlist_lookup_nvpair(origprops, propname,
4132 &match) != 0) || !propval_equals(pair, match))
4133 goto next; /* need to set received value */
4135 /* don't clear the existing received value */
4136 (void) nvlist_remove_nvpair(origprops, match);
4137 /* don't bother receiving the property */
4138 (void) nvlist_remove_nvpair(props, pair);
4145 static boolean_t zfs_ioc_recv_inject_err;
4150 * zc_name name of containing filesystem
4151 * zc_nvlist_src{_size} nvlist of properties to apply
4152 * zc_value name of snapshot to create
4153 * zc_string name of clone origin (if DRR_FLAG_CLONE)
4154 * zc_cookie file descriptor to recv from
4155 * zc_begin_record the BEGIN record of the stream (not byteswapped)
4156 * zc_guid force flag
4157 * zc_cleanup_fd cleanup-on-exit file descriptor
4158 * zc_action_handle handle for this guid/ds mapping (or zero on first call)
4161 * zc_cookie number of bytes read
4162 * zc_nvlist_dst{_size} error for each unapplied received property
4163 * zc_obj zprop_errflags_t
4164 * zc_action_handle handle for this guid/ds mapping
4167 zfs_ioc_recv(zfs_cmd_t *zc)
4170 dmu_recv_cookie_t drc;
4171 boolean_t force = (boolean_t)zc->zc_guid;
4174 int props_error = 0;
4177 nvlist_t *props = NULL; /* sent properties */
4178 nvlist_t *origprops = NULL; /* existing properties */
4179 char *origin = NULL;
4181 char tofs[ZFS_MAXNAMELEN];
4182 cap_rights_t rights;
4183 boolean_t first_recvd_props = B_FALSE;
4185 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
4186 strchr(zc->zc_value, '@') == NULL ||
4187 strchr(zc->zc_value, '%'))
4188 return (SET_ERROR(EINVAL));
4190 (void) strcpy(tofs, zc->zc_value);
4191 tosnap = strchr(tofs, '@');
4194 if (zc->zc_nvlist_src != 0 &&
4195 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
4196 zc->zc_iflags, &props)) != 0)
4203 fget_read(curthread, fd, cap_rights_init(&rights, CAP_PREAD), &fp);
4207 return (SET_ERROR(EBADF));
4210 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4212 if (zc->zc_string[0])
4213 origin = zc->zc_string;
4215 error = dmu_recv_begin(tofs, tosnap,
4216 &zc->zc_begin_record, force, origin, &drc);
4221 * Set properties before we receive the stream so that they are applied
4222 * to the new data. Note that we must call dmu_recv_stream() if
4223 * dmu_recv_begin() succeeds.
4225 if (props != NULL && !drc.drc_newfs) {
4226 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
4227 SPA_VERSION_RECVD_PROPS &&
4228 !dsl_prop_get_hasrecvd(tofs))
4229 first_recvd_props = B_TRUE;
4232 * If new received properties are supplied, they are to
4233 * completely replace the existing received properties, so stash
4234 * away the existing ones.
4236 if (dsl_prop_get_received(tofs, &origprops) == 0) {
4237 nvlist_t *errlist = NULL;
4239 * Don't bother writing a property if its value won't
4240 * change (and avoid the unnecessary security checks).
4242 * The first receive after SPA_VERSION_RECVD_PROPS is a
4243 * special case where we blow away all local properties
4246 if (!first_recvd_props)
4247 props_reduce(props, origprops);
4248 if (zfs_check_clearable(tofs, origprops, &errlist) != 0)
4249 (void) nvlist_merge(errors, errlist, 0);
4250 nvlist_free(errlist);
4252 if (clear_received_props(tofs, origprops,
4253 first_recvd_props ? NULL : props) != 0)
4254 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4256 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4260 if (props != NULL) {
4261 props_error = dsl_prop_set_hasrecvd(tofs);
4263 if (props_error == 0) {
4264 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4269 if (zc->zc_nvlist_dst_size != 0 &&
4270 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4271 put_nvlist(zc, errors) != 0)) {
4273 * Caller made zc->zc_nvlist_dst less than the minimum expected
4274 * size or supplied an invalid address.
4276 props_error = SET_ERROR(EINVAL);
4280 error = dmu_recv_stream(&drc, fp, &off, zc->zc_cleanup_fd,
4281 &zc->zc_action_handle);
4284 zfsvfs_t *zfsvfs = NULL;
4286 if (getzfsvfs(tofs, &zfsvfs) == 0) {
4290 error = zfs_suspend_fs(zfsvfs);
4292 * If the suspend fails, then the recv_end will
4293 * likely also fail, and clean up after itself.
4295 end_err = dmu_recv_end(&drc, zfsvfs);
4297 error = zfs_resume_fs(zfsvfs, tofs);
4298 error = error ? error : end_err;
4299 VFS_RELE(zfsvfs->z_vfs);
4301 error = dmu_recv_end(&drc, NULL);
4305 zc->zc_cookie = off - fp->f_offset;
4306 if (off >= 0 && off <= MAXOFFSET_T)
4310 if (zfs_ioc_recv_inject_err) {
4311 zfs_ioc_recv_inject_err = B_FALSE;
4318 zvol_create_minors(tofs);
4322 * On error, restore the original props.
4324 if (error != 0 && props != NULL && !drc.drc_newfs) {
4325 if (clear_received_props(tofs, props, NULL) != 0) {
4327 * We failed to clear the received properties.
4328 * Since we may have left a $recvd value on the
4329 * system, we can't clear the $hasrecvd flag.
4331 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4332 } else if (first_recvd_props) {
4333 dsl_prop_unset_hasrecvd(tofs);
4336 if (origprops == NULL && !drc.drc_newfs) {
4337 /* We failed to stash the original properties. */
4338 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4342 * dsl_props_set() will not convert RECEIVED to LOCAL on or
4343 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4344 * explictly if we're restoring local properties cleared in the
4345 * first new-style receive.
4347 if (origprops != NULL &&
4348 zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4349 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4350 origprops, NULL) != 0) {
4352 * We stashed the original properties but failed to
4355 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4360 nvlist_free(origprops);
4361 nvlist_free(errors);
4365 error = props_error;
4372 * zc_name name of snapshot to send
4373 * zc_cookie file descriptor to send stream to
4374 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj)
4375 * zc_sendobj objsetid of snapshot to send
4376 * zc_fromobj objsetid of incremental fromsnap (may be zero)
4377 * zc_guid if set, estimate size of stream only. zc_cookie is ignored.
4378 * output size in zc_objset_type.
4379 * zc_flags lzc_send_flags
4382 * zc_objset_type estimated size, if zc_guid is set
4385 zfs_ioc_send(zfs_cmd_t *zc)
4389 boolean_t estimate = (zc->zc_guid != 0);
4390 boolean_t embedok = (zc->zc_flags & 0x1);
4391 boolean_t large_block_ok = (zc->zc_flags & 0x2);
4393 if (zc->zc_obj != 0) {
4395 dsl_dataset_t *tosnap;
4397 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4401 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4403 dsl_pool_rele(dp, FTAG);
4407 if (dsl_dir_is_clone(tosnap->ds_dir))
4409 dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj;
4410 dsl_dataset_rele(tosnap, FTAG);
4411 dsl_pool_rele(dp, FTAG);
4416 dsl_dataset_t *tosnap;
4417 dsl_dataset_t *fromsnap = NULL;
4419 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4423 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4425 dsl_pool_rele(dp, FTAG);
4429 if (zc->zc_fromobj != 0) {
4430 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
4433 dsl_dataset_rele(tosnap, FTAG);
4434 dsl_pool_rele(dp, FTAG);
4439 error = dmu_send_estimate(tosnap, fromsnap,
4440 &zc->zc_objset_type);
4442 if (fromsnap != NULL)
4443 dsl_dataset_rele(fromsnap, FTAG);
4444 dsl_dataset_rele(tosnap, FTAG);
4445 dsl_pool_rele(dp, FTAG);
4448 cap_rights_t rights;
4451 fp = getf(zc->zc_cookie);
4453 fget_write(curthread, zc->zc_cookie,
4454 cap_rights_init(&rights, CAP_WRITE), &fp);
4457 return (SET_ERROR(EBADF));
4460 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
4461 zc->zc_fromobj, embedok, large_block_ok,
4463 zc->zc_cookie, fp->f_vnode, &off);
4465 zc->zc_cookie, fp, &off);
4468 if (off >= 0 && off <= MAXOFFSET_T)
4470 releasef(zc->zc_cookie);
4477 * zc_name name of snapshot on which to report progress
4478 * zc_cookie file descriptor of send stream
4481 * zc_cookie number of bytes written in send stream thus far
4484 zfs_ioc_send_progress(zfs_cmd_t *zc)
4488 dmu_sendarg_t *dsp = NULL;
4491 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4495 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4497 dsl_pool_rele(dp, FTAG);
4501 mutex_enter(&ds->ds_sendstream_lock);
4504 * Iterate over all the send streams currently active on this dataset.
4505 * If there's one which matches the specified file descriptor _and_ the
4506 * stream was started by the current process, return the progress of
4509 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4510 dsp = list_next(&ds->ds_sendstreams, dsp)) {
4511 if (dsp->dsa_outfd == zc->zc_cookie &&
4512 dsp->dsa_proc == curproc)
4517 zc->zc_cookie = *(dsp->dsa_off);
4519 error = SET_ERROR(ENOENT);
4521 mutex_exit(&ds->ds_sendstream_lock);
4522 dsl_dataset_rele(ds, FTAG);
4523 dsl_pool_rele(dp, FTAG);
4528 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4532 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4533 &zc->zc_inject_record);
4536 zc->zc_guid = (uint64_t)id;
4542 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4544 return (zio_clear_fault((int)zc->zc_guid));
4548 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4550 int id = (int)zc->zc_guid;
4553 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4554 &zc->zc_inject_record);
4562 zfs_ioc_error_log(zfs_cmd_t *zc)
4566 size_t count = (size_t)zc->zc_nvlist_dst_size;
4568 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4571 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4574 zc->zc_nvlist_dst_size = count;
4576 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4578 spa_close(spa, FTAG);
4584 zfs_ioc_clear(zfs_cmd_t *zc)
4591 * On zpool clear we also fix up missing slogs
4593 mutex_enter(&spa_namespace_lock);
4594 spa = spa_lookup(zc->zc_name);
4596 mutex_exit(&spa_namespace_lock);
4597 return (SET_ERROR(EIO));
4599 if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4600 /* we need to let spa_open/spa_load clear the chains */
4601 spa_set_log_state(spa, SPA_LOG_CLEAR);
4603 spa->spa_last_open_failed = 0;
4604 mutex_exit(&spa_namespace_lock);
4606 if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4607 error = spa_open(zc->zc_name, &spa, FTAG);
4610 nvlist_t *config = NULL;
4612 if (zc->zc_nvlist_src == 0)
4613 return (SET_ERROR(EINVAL));
4615 if ((error = get_nvlist(zc->zc_nvlist_src,
4616 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4617 error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4619 if (config != NULL) {
4622 if ((err = put_nvlist(zc, config)) != 0)
4624 nvlist_free(config);
4626 nvlist_free(policy);
4633 spa_vdev_state_enter(spa, SCL_NONE);
4635 if (zc->zc_guid == 0) {
4638 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4640 (void) spa_vdev_state_exit(spa, NULL, ENODEV);
4641 spa_close(spa, FTAG);
4642 return (SET_ERROR(ENODEV));
4646 vdev_clear(spa, vd);
4648 (void) spa_vdev_state_exit(spa, NULL, 0);
4651 * Resume any suspended I/Os.
4653 if (zio_resume(spa) != 0)
4654 error = SET_ERROR(EIO);
4656 spa_close(spa, FTAG);
4662 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4667 error = spa_open(zc->zc_name, &spa, FTAG);
4671 spa_vdev_state_enter(spa, SCL_NONE);
4674 * If a resilver is already in progress then set the
4675 * spa_scrub_reopen flag to B_TRUE so that we don't restart
4676 * the scan as a side effect of the reopen. Otherwise, let
4677 * vdev_open() decided if a resilver is required.
4679 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4680 vdev_reopen(spa->spa_root_vdev);
4681 spa->spa_scrub_reopen = B_FALSE;
4683 (void) spa_vdev_state_exit(spa, NULL, 0);
4684 spa_close(spa, FTAG);
4689 * zc_name name of filesystem
4690 * zc_value name of origin snapshot
4693 * zc_string name of conflicting snapshot, if there is one
4696 zfs_ioc_promote(zfs_cmd_t *zc)
4701 * We don't need to unmount *all* the origin fs's snapshots, but
4704 cp = strchr(zc->zc_value, '@');
4707 (void) dmu_objset_find(zc->zc_value,
4708 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
4709 return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4713 * Retrieve a single {user|group}{used|quota}@... property.
4716 * zc_name name of filesystem
4717 * zc_objset_type zfs_userquota_prop_t
4718 * zc_value domain name (eg. "S-1-234-567-89")
4719 * zc_guid RID/UID/GID
4722 * zc_cookie property value
4725 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4730 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4731 return (SET_ERROR(EINVAL));
4733 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4737 error = zfs_userspace_one(zfsvfs,
4738 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4739 zfsvfs_rele(zfsvfs, FTAG);
4746 * zc_name name of filesystem
4747 * zc_cookie zap cursor
4748 * zc_objset_type zfs_userquota_prop_t
4749 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4752 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
4753 * zc_cookie zap cursor
4756 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4759 int bufsize = zc->zc_nvlist_dst_size;
4762 return (SET_ERROR(ENOMEM));
4764 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4768 void *buf = kmem_alloc(bufsize, KM_SLEEP);
4770 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
4771 buf, &zc->zc_nvlist_dst_size);
4774 error = ddi_copyout(buf,
4775 (void *)(uintptr_t)zc->zc_nvlist_dst,
4776 zc->zc_nvlist_dst_size, zc->zc_iflags);
4778 kmem_free(buf, bufsize);
4779 zfsvfs_rele(zfsvfs, FTAG);
4786 * zc_name name of filesystem
4792 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4798 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
4799 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
4801 * If userused is not enabled, it may be because the
4802 * objset needs to be closed & reopened (to grow the
4803 * objset_phys_t). Suspend/resume the fs will do that.
4805 error = zfs_suspend_fs(zfsvfs);
4807 dmu_objset_refresh_ownership(zfsvfs->z_os,
4809 error = zfs_resume_fs(zfsvfs, zc->zc_name);
4813 error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
4814 VFS_RELE(zfsvfs->z_vfs);
4816 /* XXX kind of reading contents without owning */
4817 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4821 error = dmu_objset_userspace_upgrade(os);
4822 dmu_objset_rele(os, FTAG);
4830 * We don't want to have a hard dependency
4831 * against some special symbols in sharefs
4832 * nfs, and smbsrv. Determine them if needed when
4833 * the first file system is shared.
4834 * Neither sharefs, nfs or smbsrv are unloadable modules.
4836 int (*znfsexport_fs)(void *arg);
4837 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
4838 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
4840 int zfs_nfsshare_inited;
4841 int zfs_smbshare_inited;
4843 ddi_modhandle_t nfs_mod;
4844 ddi_modhandle_t sharefs_mod;
4845 ddi_modhandle_t smbsrv_mod;
4846 #endif /* illumos */
4847 kmutex_t zfs_share_lock;
4855 ASSERT(MUTEX_HELD(&zfs_share_lock));
4856 /* Both NFS and SMB shares also require sharetab support. */
4857 if (sharefs_mod == NULL && ((sharefs_mod =
4858 ddi_modopen("fs/sharefs",
4859 KRTLD_MODE_FIRST, &error)) == NULL)) {
4860 return (SET_ERROR(ENOSYS));
4862 if (zshare_fs == NULL && ((zshare_fs =
4863 (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
4864 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
4865 return (SET_ERROR(ENOSYS));
4869 #endif /* illumos */
4872 zfs_ioc_share(zfs_cmd_t *zc)
4878 switch (zc->zc_share.z_sharetype) {
4880 case ZFS_UNSHARE_NFS:
4881 if (zfs_nfsshare_inited == 0) {
4882 mutex_enter(&zfs_share_lock);
4883 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
4884 KRTLD_MODE_FIRST, &error)) == NULL)) {
4885 mutex_exit(&zfs_share_lock);
4886 return (SET_ERROR(ENOSYS));
4888 if (znfsexport_fs == NULL &&
4889 ((znfsexport_fs = (int (*)(void *))
4891 "nfs_export", &error)) == NULL)) {
4892 mutex_exit(&zfs_share_lock);
4893 return (SET_ERROR(ENOSYS));
4895 error = zfs_init_sharefs();
4897 mutex_exit(&zfs_share_lock);
4898 return (SET_ERROR(ENOSYS));
4900 zfs_nfsshare_inited = 1;
4901 mutex_exit(&zfs_share_lock);
4905 case ZFS_UNSHARE_SMB:
4906 if (zfs_smbshare_inited == 0) {
4907 mutex_enter(&zfs_share_lock);
4908 if (smbsrv_mod == NULL && ((smbsrv_mod =
4909 ddi_modopen("drv/smbsrv",
4910 KRTLD_MODE_FIRST, &error)) == NULL)) {
4911 mutex_exit(&zfs_share_lock);
4912 return (SET_ERROR(ENOSYS));
4914 if (zsmbexport_fs == NULL && ((zsmbexport_fs =
4915 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
4916 "smb_server_share", &error)) == NULL)) {
4917 mutex_exit(&zfs_share_lock);
4918 return (SET_ERROR(ENOSYS));
4920 error = zfs_init_sharefs();
4922 mutex_exit(&zfs_share_lock);
4923 return (SET_ERROR(ENOSYS));
4925 zfs_smbshare_inited = 1;
4926 mutex_exit(&zfs_share_lock);
4930 return (SET_ERROR(EINVAL));
4933 switch (zc->zc_share.z_sharetype) {
4935 case ZFS_UNSHARE_NFS:
4937 znfsexport_fs((void *)
4938 (uintptr_t)zc->zc_share.z_exportdata))
4942 case ZFS_UNSHARE_SMB:
4943 if (error = zsmbexport_fs((void *)
4944 (uintptr_t)zc->zc_share.z_exportdata,
4945 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
4952 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
4953 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
4954 SHAREFS_ADD : SHAREFS_REMOVE;
4957 * Add or remove share from sharetab
4959 error = zshare_fs(opcode,
4960 (void *)(uintptr_t)zc->zc_share.z_sharedata,
4961 zc->zc_share.z_sharemax);
4965 #else /* !illumos */
4967 #endif /* illumos */
4970 ace_t full_access[] = {
4971 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4976 * zc_name name of containing filesystem
4977 * zc_obj object # beyond which we want next in-use object #
4980 * zc_obj next in-use object #
4983 zfs_ioc_next_obj(zfs_cmd_t *zc)
4985 objset_t *os = NULL;
4988 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4992 error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
4993 dsl_dataset_phys(os->os_dsl_dataset)->ds_prev_snap_txg);
4995 dmu_objset_rele(os, FTAG);
5001 * zc_name name of filesystem
5002 * zc_value prefix name for snapshot
5003 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
5006 * zc_value short name of new snapshot
5009 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
5016 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
5020 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
5021 (u_longlong_t)ddi_get_lbolt64());
5022 hold_name = kmem_asprintf("%%%s", zc->zc_value);
5024 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
5027 (void) strcpy(zc->zc_value, snap_name);
5030 zfs_onexit_fd_rele(zc->zc_cleanup_fd);
5036 * zc_name name of "to" snapshot
5037 * zc_value name of "from" snapshot
5038 * zc_cookie file descriptor to write diff data on
5041 * dmu_diff_record_t's to the file descriptor
5044 zfs_ioc_diff(zfs_cmd_t *zc)
5047 cap_rights_t rights;
5052 fp = getf(zc->zc_cookie);
5054 fget_write(curthread, zc->zc_cookie,
5055 cap_rights_init(&rights, CAP_WRITE), &fp);
5058 return (SET_ERROR(EBADF));
5063 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off);
5065 error = dmu_diff(zc->zc_name, zc->zc_value, fp, &off);
5068 if (off >= 0 && off <= MAXOFFSET_T)
5070 releasef(zc->zc_cookie);
5077 * Remove all ACL files in shares dir
5080 zfs_smb_acl_purge(znode_t *dzp)
5083 zap_attribute_t zap;
5084 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
5087 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
5088 (error = zap_cursor_retrieve(&zc, &zap)) == 0;
5089 zap_cursor_advance(&zc)) {
5090 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
5094 zap_cursor_fini(&zc);
5097 #endif /* illumos */
5100 zfs_ioc_smb_acl(zfs_cmd_t *zc)
5105 vnode_t *resourcevp = NULL;
5114 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
5115 NO_FOLLOW, NULL, &vp)) != 0)
5118 /* Now make sure mntpnt and dataset are ZFS */
5120 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 ||
5121 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
5122 zc->zc_name) != 0)) {
5124 return (SET_ERROR(EINVAL));
5128 zfsvfs = dzp->z_zfsvfs;
5132 * Create share dir if its missing.
5134 mutex_enter(&zfsvfs->z_lock);
5135 if (zfsvfs->z_shares_dir == 0) {
5138 tx = dmu_tx_create(zfsvfs->z_os);
5139 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
5141 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
5142 error = dmu_tx_assign(tx, TXG_WAIT);
5146 error = zfs_create_share_dir(zfsvfs, tx);
5150 mutex_exit(&zfsvfs->z_lock);
5156 mutex_exit(&zfsvfs->z_lock);
5158 ASSERT(zfsvfs->z_shares_dir);
5159 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
5165 switch (zc->zc_cookie) {
5166 case ZFS_SMB_ACL_ADD:
5167 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
5168 vattr.va_type = VREG;
5169 vattr.va_mode = S_IFREG|0777;
5173 vsec.vsa_mask = VSA_ACE;
5174 vsec.vsa_aclentp = &full_access;
5175 vsec.vsa_aclentsz = sizeof (full_access);
5176 vsec.vsa_aclcnt = 1;
5178 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
5179 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
5181 VN_RELE(resourcevp);
5184 case ZFS_SMB_ACL_REMOVE:
5185 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
5189 case ZFS_SMB_ACL_RENAME:
5190 if ((error = get_nvlist(zc->zc_nvlist_src,
5191 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
5196 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
5197 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
5200 VN_RELE(ZTOV(sharedir));
5202 nvlist_free(nvlist);
5205 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
5207 nvlist_free(nvlist);
5210 case ZFS_SMB_ACL_PURGE:
5211 error = zfs_smb_acl_purge(sharedir);
5215 error = SET_ERROR(EINVAL);
5220 VN_RELE(ZTOV(sharedir));
5225 #else /* !illumos */
5226 return (EOPNOTSUPP);
5227 #endif /* illumos */
5232 * "holds" -> { snapname -> holdname (string), ... }
5233 * (optional) "cleanup_fd" -> fd (int32)
5237 * snapname -> error value (int32)
5243 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
5247 int cleanup_fd = -1;
5251 error = nvlist_lookup_nvlist(args, "holds", &holds);
5253 return (SET_ERROR(EINVAL));
5255 /* make sure the user didn't pass us any invalid (empty) tags */
5256 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
5257 pair = nvlist_next_nvpair(holds, pair)) {
5260 error = nvpair_value_string(pair, &htag);
5262 return (SET_ERROR(error));
5264 if (strlen(htag) == 0)
5265 return (SET_ERROR(EINVAL));
5268 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
5269 error = zfs_onexit_fd_hold(cleanup_fd, &minor);
5274 error = dsl_dataset_user_hold(holds, minor, errlist);
5276 zfs_onexit_fd_rele(cleanup_fd);
5281 * innvl is not used.
5284 * holdname -> time added (uint64 seconds since epoch)
5290 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
5292 return (dsl_dataset_get_holds(snapname, outnvl));
5297 * snapname -> { holdname, ... }
5302 * snapname -> error value (int32)
5308 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
5310 return (dsl_dataset_user_release(holds, errlist));
5315 * zc_name name of new filesystem or snapshot
5316 * zc_value full name of old snapshot
5319 * zc_cookie space in bytes
5320 * zc_objset_type compressed space in bytes
5321 * zc_perm_action uncompressed space in bytes
5324 zfs_ioc_space_written(zfs_cmd_t *zc)
5328 dsl_dataset_t *new, *old;
5330 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5333 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
5335 dsl_pool_rele(dp, FTAG);
5338 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
5340 dsl_dataset_rele(new, FTAG);
5341 dsl_pool_rele(dp, FTAG);
5345 error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
5346 &zc->zc_objset_type, &zc->zc_perm_action);
5347 dsl_dataset_rele(old, FTAG);
5348 dsl_dataset_rele(new, FTAG);
5349 dsl_pool_rele(dp, FTAG);
5355 * "firstsnap" -> snapshot name
5359 * "used" -> space in bytes
5360 * "compressed" -> compressed space in bytes
5361 * "uncompressed" -> uncompressed space in bytes
5365 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
5369 dsl_dataset_t *new, *old;
5371 uint64_t used, comp, uncomp;
5373 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5374 return (SET_ERROR(EINVAL));
5376 error = dsl_pool_hold(lastsnap, FTAG, &dp);
5380 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
5381 if (error == 0 && !new->ds_is_snapshot) {
5382 dsl_dataset_rele(new, FTAG);
5383 error = SET_ERROR(EINVAL);
5386 dsl_pool_rele(dp, FTAG);
5389 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
5390 if (error == 0 && !old->ds_is_snapshot) {
5391 dsl_dataset_rele(old, FTAG);
5392 error = SET_ERROR(EINVAL);
5395 dsl_dataset_rele(new, FTAG);
5396 dsl_pool_rele(dp, FTAG);
5400 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5401 dsl_dataset_rele(old, FTAG);
5402 dsl_dataset_rele(new, FTAG);
5403 dsl_pool_rele(dp, FTAG);
5404 fnvlist_add_uint64(outnvl, "used", used);
5405 fnvlist_add_uint64(outnvl, "compressed", comp);
5406 fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5411 zfs_ioc_jail(zfs_cmd_t *zc)
5414 return (zone_dataset_attach(curthread->td_ucred, zc->zc_name,
5415 (int)zc->zc_jailid));
5419 zfs_ioc_unjail(zfs_cmd_t *zc)
5422 return (zone_dataset_detach(curthread->td_ucred, zc->zc_name,
5423 (int)zc->zc_jailid));
5428 * "fd" -> file descriptor to write stream to (int32)
5429 * (optional) "fromsnap" -> full snap name to send an incremental from
5430 * (optional) "largeblockok" -> (value ignored)
5431 * indicates that blocks > 128KB are permitted
5432 * (optional) "embedok" -> (value ignored)
5433 * presence indicates DRR_WRITE_EMBEDDED records are permitted
5440 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5442 cap_rights_t rights;
5446 char *fromname = NULL;
5448 boolean_t largeblockok;
5451 error = nvlist_lookup_int32(innvl, "fd", &fd);
5453 return (SET_ERROR(EINVAL));
5455 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
5457 largeblockok = nvlist_exists(innvl, "largeblockok");
5458 embedok = nvlist_exists(innvl, "embedok");
5461 file_t *fp = getf(fd);
5463 fget_write(curthread, fd, cap_rights_init(&rights, CAP_WRITE), &fp);
5466 return (SET_ERROR(EBADF));
5469 error = dmu_send(snapname, fromname, embedok, largeblockok,
5471 fd, fp->f_vnode, &off);
5477 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5488 * Determine approximately how large a zfs send stream will be -- the number
5489 * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5492 * (optional) "from" -> full snap or bookmark name to send an incremental
5497 * "space" -> bytes of space (uint64)
5501 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5504 dsl_dataset_t *tosnap;
5509 error = dsl_pool_hold(snapname, FTAG, &dp);
5513 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
5515 dsl_pool_rele(dp, FTAG);
5519 error = nvlist_lookup_string(innvl, "from", &fromname);
5521 if (strchr(fromname, '@') != NULL) {
5523 * If from is a snapshot, hold it and use the more
5524 * efficient dmu_send_estimate to estimate send space
5525 * size using deadlists.
5527 dsl_dataset_t *fromsnap;
5528 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
5531 error = dmu_send_estimate(tosnap, fromsnap, &space);
5532 dsl_dataset_rele(fromsnap, FTAG);
5533 } else if (strchr(fromname, '#') != NULL) {
5535 * If from is a bookmark, fetch the creation TXG of the
5536 * snapshot it was created from and use that to find
5537 * blocks that were born after it.
5539 zfs_bookmark_phys_t frombm;
5541 error = dsl_bookmark_lookup(dp, fromname, tosnap,
5545 error = dmu_send_estimate_from_txg(tosnap,
5546 frombm.zbm_creation_txg, &space);
5549 * from is not properly formatted as a snapshot or
5552 error = SET_ERROR(EINVAL);
5556 // If estimating the size of a full send, use dmu_send_estimate
5557 error = dmu_send_estimate(tosnap, NULL, &space);
5560 fnvlist_add_uint64(outnvl, "space", space);
5563 dsl_dataset_rele(tosnap, FTAG);
5564 dsl_pool_rele(dp, FTAG);
5568 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
5571 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5572 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5573 boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
5575 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5577 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5578 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5579 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5580 ASSERT3P(vec->zvec_func, ==, NULL);
5582 vec->zvec_legacy_func = func;
5583 vec->zvec_secpolicy = secpolicy;
5584 vec->zvec_namecheck = namecheck;
5585 vec->zvec_allow_log = log_history;
5586 vec->zvec_pool_check = pool_check;
5590 * See the block comment at the beginning of this file for details on
5591 * each argument to this function.
5594 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
5595 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5596 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
5597 boolean_t allow_log)
5599 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5601 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5602 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5603 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5604 ASSERT3P(vec->zvec_func, ==, NULL);
5606 /* if we are logging, the name must be valid */
5607 ASSERT(!allow_log || namecheck != NO_NAME);
5609 vec->zvec_name = name;
5610 vec->zvec_func = func;
5611 vec->zvec_secpolicy = secpolicy;
5612 vec->zvec_namecheck = namecheck;
5613 vec->zvec_pool_check = pool_check;
5614 vec->zvec_smush_outnvlist = smush_outnvlist;
5615 vec->zvec_allow_log = allow_log;
5619 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5620 zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
5621 zfs_ioc_poolcheck_t pool_check)
5623 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5624 POOL_NAME, log_history, pool_check);
5628 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5629 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
5631 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5632 DATASET_NAME, B_FALSE, pool_check);
5636 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5638 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
5639 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5643 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5644 zfs_secpolicy_func_t *secpolicy)
5646 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5647 NO_NAME, B_FALSE, POOL_CHECK_NONE);
5651 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
5652 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
5654 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5655 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
5659 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5661 zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
5662 zfs_secpolicy_read);
5666 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5667 zfs_secpolicy_func_t *secpolicy)
5669 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5670 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5674 zfs_ioctl_init(void)
5676 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
5677 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
5678 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5680 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
5681 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
5682 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
5684 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
5685 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
5686 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5688 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
5689 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
5690 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5692 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
5693 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
5694 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5696 zfs_ioctl_register("create", ZFS_IOC_CREATE,
5697 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
5698 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5700 zfs_ioctl_register("clone", ZFS_IOC_CLONE,
5701 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
5702 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5704 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
5705 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
5706 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5708 zfs_ioctl_register("hold", ZFS_IOC_HOLD,
5709 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
5710 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5711 zfs_ioctl_register("release", ZFS_IOC_RELEASE,
5712 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
5713 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5715 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
5716 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
5717 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5719 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
5720 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
5721 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE);
5723 zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK,
5724 zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME,
5725 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5727 zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS,
5728 zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME,
5729 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5731 zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS,
5732 zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks,
5734 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5736 /* IOCTLS that use the legacy function signature */
5738 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
5739 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
5741 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
5742 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5743 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
5745 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
5746 zfs_ioc_pool_upgrade);
5747 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
5749 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
5750 zfs_ioc_vdev_remove);
5751 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
5752 zfs_ioc_vdev_set_state);
5753 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
5754 zfs_ioc_vdev_attach);
5755 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
5756 zfs_ioc_vdev_detach);
5757 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
5758 zfs_ioc_vdev_setpath);
5759 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
5760 zfs_ioc_vdev_setfru);
5761 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
5762 zfs_ioc_pool_set_props);
5763 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
5764 zfs_ioc_vdev_split);
5765 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
5766 zfs_ioc_pool_reguid);
5768 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
5769 zfs_ioc_pool_configs, zfs_secpolicy_none);
5770 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
5771 zfs_ioc_pool_tryimport, zfs_secpolicy_config);
5772 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
5773 zfs_ioc_inject_fault, zfs_secpolicy_inject);
5774 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
5775 zfs_ioc_clear_fault, zfs_secpolicy_inject);
5776 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
5777 zfs_ioc_inject_list_next, zfs_secpolicy_inject);
5780 * pool destroy, and export don't log the history as part of
5781 * zfsdev_ioctl, but rather zfs_ioc_pool_export
5782 * does the logging of those commands.
5784 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
5785 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5786 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
5787 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5789 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
5790 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5791 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
5792 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5794 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
5795 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_NONE);
5796 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
5797 zfs_ioc_dsobj_to_dsname,
5798 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_NONE);
5799 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
5800 zfs_ioc_pool_get_history,
5801 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
5803 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
5804 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5806 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
5807 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5808 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
5809 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5811 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
5812 zfs_ioc_space_written);
5813 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
5814 zfs_ioc_objset_recvd_props);
5815 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
5817 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
5819 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
5820 zfs_ioc_objset_stats);
5821 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
5822 zfs_ioc_objset_zplprops);
5823 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
5824 zfs_ioc_dataset_list_next);
5825 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
5826 zfs_ioc_snapshot_list_next);
5827 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
5828 zfs_ioc_send_progress);
5830 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
5831 zfs_ioc_diff, zfs_secpolicy_diff);
5832 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
5833 zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
5834 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
5835 zfs_ioc_obj_to_path, zfs_secpolicy_diff);
5836 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
5837 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
5838 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
5839 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
5840 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
5841 zfs_ioc_send, zfs_secpolicy_send);
5843 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
5844 zfs_secpolicy_none);
5845 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
5846 zfs_secpolicy_destroy);
5847 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
5848 zfs_secpolicy_rename);
5849 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
5850 zfs_secpolicy_recv);
5851 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
5852 zfs_secpolicy_promote);
5853 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
5854 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
5855 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
5856 zfs_secpolicy_set_fsacl);
5858 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
5859 zfs_secpolicy_share, POOL_CHECK_NONE);
5860 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
5861 zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
5862 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
5863 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
5864 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5865 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
5866 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
5867 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5870 zfs_ioctl_register_dataset_nolog(ZFS_IOC_JAIL, zfs_ioc_jail,
5871 zfs_secpolicy_config, POOL_CHECK_NONE);
5872 zfs_ioctl_register_dataset_nolog(ZFS_IOC_UNJAIL, zfs_ioc_unjail,
5873 zfs_secpolicy_config, POOL_CHECK_NONE);
5878 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
5879 zfs_ioc_poolcheck_t check)
5884 ASSERT(type == POOL_NAME || type == DATASET_NAME);
5886 if (check & POOL_CHECK_NONE)
5889 error = spa_open(name, &spa, FTAG);
5891 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
5892 error = SET_ERROR(EAGAIN);
5893 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
5894 error = SET_ERROR(EROFS);
5895 spa_close(spa, FTAG);
5901 * Find a free minor number.
5904 zfsdev_minor_alloc(void)
5906 static minor_t last_minor;
5909 ASSERT(MUTEX_HELD(&spa_namespace_lock));
5911 for (m = last_minor + 1; m != last_minor; m++) {
5912 if (m > ZFSDEV_MAX_MINOR)
5914 if (ddi_get_soft_state(zfsdev_state, m) == NULL) {
5924 zfs_ctldev_init(struct cdev *devp)
5927 zfs_soft_state_t *zs;
5929 ASSERT(MUTEX_HELD(&spa_namespace_lock));
5931 minor = zfsdev_minor_alloc();
5933 return (SET_ERROR(ENXIO));
5935 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS)
5936 return (SET_ERROR(EAGAIN));
5938 devfs_set_cdevpriv((void *)(uintptr_t)minor, zfsdev_close);
5940 zs = ddi_get_soft_state(zfsdev_state, minor);
5941 zs->zss_type = ZSST_CTLDEV;
5942 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data);
5948 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor)
5950 ASSERT(MUTEX_HELD(&spa_namespace_lock));
5952 zfs_onexit_destroy(zo);
5953 ddi_soft_state_free(zfsdev_state, minor);
5957 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which)
5959 zfs_soft_state_t *zp;
5961 zp = ddi_get_soft_state(zfsdev_state, minor);
5962 if (zp == NULL || zp->zss_type != which)
5965 return (zp->zss_data);
5969 zfsdev_open(struct cdev *devp, int flag, int mode, struct thread *td)
5974 if (getminor(*devp) != 0)
5975 return (zvol_open(devp, flag, otyp, cr));
5978 /* This is the control device. Allocate a new minor if requested. */
5980 mutex_enter(&spa_namespace_lock);
5981 error = zfs_ctldev_init(devp);
5982 mutex_exit(&spa_namespace_lock);
5989 zfsdev_close(void *data)
5992 minor_t minor = (minor_t)(uintptr_t)data;
5997 mutex_enter(&spa_namespace_lock);
5998 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV);
6000 mutex_exit(&spa_namespace_lock);
6003 zfs_ctldev_destroy(zo, minor);
6004 mutex_exit(&spa_namespace_lock);
6008 zfsdev_ioctl(struct cdev *dev, u_long zcmd, caddr_t arg, int flag,
6015 minor_t minor = getminor(dev);
6017 zfs_iocparm_t *zc_iocparm;
6018 int cflag, cmd, oldvecnum;
6019 boolean_t newioc, compat;
6020 void *compat_zc = NULL;
6021 cred_t *cr = td->td_ucred;
6023 const zfs_ioc_vec_t *vec;
6024 char *saved_poolname = NULL;
6025 nvlist_t *innvl = NULL;
6027 cflag = ZFS_CMD_COMPAT_NONE;
6029 newioc = B_TRUE; /* "new" style (zfs_iocparm_t) ioctl */
6031 len = IOCPARM_LEN(zcmd);
6032 vecnum = cmd = zcmd & 0xff;
6035 * Check if we are talking to supported older binaries
6036 * and translate zfs_cmd if necessary
6038 if (len != sizeof(zfs_iocparm_t)) {
6045 case sizeof(zfs_cmd_zcmd_t):
6046 cflag = ZFS_CMD_COMPAT_LZC;
6048 case sizeof(zfs_cmd_deadman_t):
6049 cflag = ZFS_CMD_COMPAT_DEADMAN;
6051 case sizeof(zfs_cmd_v28_t):
6052 cflag = ZFS_CMD_COMPAT_V28;
6054 case sizeof(zfs_cmd_v15_t):
6055 cflag = ZFS_CMD_COMPAT_V15;
6056 vecnum = zfs_ioctl_v15_to_v28[cmd];
6059 * Return without further handling
6060 * if the command is blacklisted.
6062 if (vecnum == ZFS_IOC_COMPAT_PASS)
6064 else if (vecnum == ZFS_IOC_COMPAT_FAIL)
6073 vecnum = cmd - ZFS_IOC_FIRST;
6074 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
6077 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
6078 return (SET_ERROR(EINVAL));
6079 vec = &zfs_ioc_vec[vecnum];
6081 zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP);
6084 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
6086 error = SET_ERROR(EFAULT);
6089 #else /* !illumos */
6090 bzero(zc, sizeof(zfs_cmd_t));
6093 zc_iocparm = (void *)arg;
6095 switch (zc_iocparm->zfs_ioctl_version) {
6096 case ZFS_IOCVER_CURRENT:
6097 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_t)) {
6098 error = SET_ERROR(EINVAL);
6102 case ZFS_IOCVER_ZCMD:
6103 if (zc_iocparm->zfs_cmd_size > sizeof(zfs_cmd_t) ||
6104 zc_iocparm->zfs_cmd_size < sizeof(zfs_cmd_zcmd_t)) {
6105 error = SET_ERROR(EFAULT);
6109 cflag = ZFS_CMD_COMPAT_ZCMD;
6112 error = SET_ERROR(EINVAL);
6118 ASSERT(sizeof(zfs_cmd_t) >= zc_iocparm->zfs_cmd_size);
6119 compat_zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP);
6120 bzero(compat_zc, sizeof(zfs_cmd_t));
6122 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd,
6123 compat_zc, zc_iocparm->zfs_cmd_size, flag);
6125 error = SET_ERROR(EFAULT);
6129 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd,
6130 zc, zc_iocparm->zfs_cmd_size, flag);
6132 error = SET_ERROR(EFAULT);
6140 ASSERT(compat_zc != NULL);
6141 zfs_cmd_compat_get(zc, compat_zc, cflag);
6143 ASSERT(compat_zc == NULL);
6144 zfs_cmd_compat_get(zc, arg, cflag);
6147 error = zfs_ioctl_compat_pre(zc, &vecnum, cflag);
6150 if (oldvecnum != vecnum)
6151 vec = &zfs_ioc_vec[vecnum];
6153 #endif /* !illumos */
6155 zc->zc_iflags = flag & FKIOCTL;
6156 if (zc->zc_nvlist_src_size != 0) {
6157 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
6158 zc->zc_iflags, &innvl);
6163 /* rewrite innvl for backwards compatibility */
6165 innvl = zfs_ioctl_compat_innvl(zc, innvl, vecnum, cflag);
6168 * Ensure that all pool/dataset names are valid before we pass down to
6171 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
6172 switch (vec->zvec_namecheck) {
6174 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
6175 error = SET_ERROR(EINVAL);
6177 error = pool_status_check(zc->zc_name,
6178 vec->zvec_namecheck, vec->zvec_pool_check);
6182 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
6183 error = SET_ERROR(EINVAL);
6185 error = pool_status_check(zc->zc_name,
6186 vec->zvec_namecheck, vec->zvec_pool_check);
6193 if (error == 0 && !(flag & FKIOCTL))
6194 error = vec->zvec_secpolicy(zc, innvl, cr);
6199 /* legacy ioctls can modify zc_name */
6200 len = strcspn(zc->zc_name, "/@#") + 1;
6201 saved_poolname = kmem_alloc(len, KM_SLEEP);
6202 (void) strlcpy(saved_poolname, zc->zc_name, len);
6204 if (vec->zvec_func != NULL) {
6208 nvlist_t *lognv = NULL;
6210 ASSERT(vec->zvec_legacy_func == NULL);
6213 * Add the innvl to the lognv before calling the func,
6214 * in case the func changes the innvl.
6216 if (vec->zvec_allow_log) {
6217 lognv = fnvlist_alloc();
6218 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
6220 if (!nvlist_empty(innvl)) {
6221 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
6226 outnvl = fnvlist_alloc();
6227 error = vec->zvec_func(zc->zc_name, innvl, outnvl);
6229 if (error == 0 && vec->zvec_allow_log &&
6230 spa_open(zc->zc_name, &spa, FTAG) == 0) {
6231 if (!nvlist_empty(outnvl)) {
6232 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
6235 (void) spa_history_log_nvl(spa, lognv);
6236 spa_close(spa, FTAG);
6238 fnvlist_free(lognv);
6240 /* rewrite outnvl for backwards compatibility */
6242 outnvl = zfs_ioctl_compat_outnvl(zc, outnvl, vecnum,
6245 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
6247 if (vec->zvec_smush_outnvlist) {
6248 smusherror = nvlist_smush(outnvl,
6249 zc->zc_nvlist_dst_size);
6251 if (smusherror == 0)
6252 puterror = put_nvlist(zc, outnvl);
6258 nvlist_free(outnvl);
6260 error = vec->zvec_legacy_func(zc);
6267 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
6268 if (error == 0 && rc != 0)
6269 error = SET_ERROR(EFAULT);
6272 zfs_ioctl_compat_post(zc, cmd, cflag);
6274 ASSERT(compat_zc != NULL);
6275 ASSERT(sizeof(zfs_cmd_t) >= zc_iocparm->zfs_cmd_size);
6277 zfs_cmd_compat_put(zc, compat_zc, vecnum, cflag);
6278 rc = ddi_copyout(compat_zc,
6279 (void *)(uintptr_t)zc_iocparm->zfs_cmd,
6280 zc_iocparm->zfs_cmd_size, flag);
6281 if (error == 0 && rc != 0)
6282 error = SET_ERROR(EFAULT);
6283 kmem_free(compat_zc, sizeof (zfs_cmd_t));
6285 zfs_cmd_compat_put(zc, arg, vecnum, cflag);
6290 rc = ddi_copyout(zc, (void *)(uintptr_t)zc_iocparm->zfs_cmd,
6291 sizeof (zfs_cmd_t), flag);
6292 if (error == 0 && rc != 0)
6293 error = SET_ERROR(EFAULT);
6296 if (error == 0 && vec->zvec_allow_log) {
6297 char *s = tsd_get(zfs_allow_log_key);
6300 (void) tsd_set(zfs_allow_log_key, saved_poolname);
6302 if (saved_poolname != NULL)
6303 strfree(saved_poolname);
6306 kmem_free(zc, sizeof (zfs_cmd_t));
6312 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
6314 if (cmd != DDI_ATTACH)
6315 return (DDI_FAILURE);
6317 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
6318 DDI_PSEUDO, 0) == DDI_FAILURE)
6319 return (DDI_FAILURE);
6323 ddi_report_dev(dip);
6325 return (DDI_SUCCESS);
6329 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
6331 if (spa_busy() || zfs_busy() || zvol_busy())
6332 return (DDI_FAILURE);
6334 if (cmd != DDI_DETACH)
6335 return (DDI_FAILURE);
6339 ddi_prop_remove_all(dip);
6340 ddi_remove_minor_node(dip, NULL);
6342 return (DDI_SUCCESS);
6347 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
6350 case DDI_INFO_DEVT2DEVINFO:
6352 return (DDI_SUCCESS);
6354 case DDI_INFO_DEVT2INSTANCE:
6355 *result = (void *)0;
6356 return (DDI_SUCCESS);
6359 return (DDI_FAILURE);
6361 #endif /* illumos */
6364 * OK, so this is a little weird.
6366 * /dev/zfs is the control node, i.e. minor 0.
6367 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
6369 * /dev/zfs has basically nothing to do except serve up ioctls,
6370 * so most of the standard driver entry points are in zvol.c.
6373 static struct cb_ops zfs_cb_ops = {
6374 zfsdev_open, /* open */
6375 zfsdev_close, /* close */
6376 zvol_strategy, /* strategy */
6378 zvol_dump, /* dump */
6379 zvol_read, /* read */
6380 zvol_write, /* write */
6381 zfsdev_ioctl, /* ioctl */
6385 nochpoll, /* poll */
6386 ddi_prop_op, /* prop_op */
6387 NULL, /* streamtab */
6388 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */
6389 CB_REV, /* version */
6390 nodev, /* async read */
6391 nodev, /* async write */
6394 static struct dev_ops zfs_dev_ops = {
6395 DEVO_REV, /* version */
6397 zfs_info, /* info */
6398 nulldev, /* identify */
6399 nulldev, /* probe */
6400 zfs_attach, /* attach */
6401 zfs_detach, /* detach */
6403 &zfs_cb_ops, /* driver operations */
6404 NULL, /* no bus operations */
6406 ddi_quiesce_not_needed, /* quiesce */
6409 static struct modldrv zfs_modldrv = {
6415 static struct modlinkage modlinkage = {
6417 (void *)&zfs_modlfs,
6418 (void *)&zfs_modldrv,
6421 #endif /* illumos */
6423 static struct cdevsw zfs_cdevsw = {
6424 .d_version = D_VERSION,
6425 .d_open = zfsdev_open,
6426 .d_ioctl = zfsdev_ioctl,
6427 .d_name = ZFS_DEV_NAME
6431 zfs_allow_log_destroy(void *arg)
6433 char *poolname = arg;
6440 zfsdev = make_dev(&zfs_cdevsw, 0x0, UID_ROOT, GID_OPERATOR, 0666,
6448 destroy_dev(zfsdev);
6451 static struct root_hold_token *zfs_root_token;
6452 struct proc *zfsproc;
6460 spa_init(FREAD | FWRITE);
6465 if ((error = mod_install(&modlinkage)) != 0) {
6472 tsd_create(&zfs_fsyncer_key, NULL);
6473 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
6474 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
6476 error = ldi_ident_from_mod(&modlinkage, &zfs_li);
6478 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
6488 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
6489 return (SET_ERROR(EBUSY));
6491 if ((error = mod_remove(&modlinkage)) != 0)
6497 if (zfs_nfsshare_inited)
6498 (void) ddi_modclose(nfs_mod);
6499 if (zfs_smbshare_inited)
6500 (void) ddi_modclose(smbsrv_mod);
6501 if (zfs_nfsshare_inited || zfs_smbshare_inited)
6502 (void) ddi_modclose(sharefs_mod);
6504 tsd_destroy(&zfs_fsyncer_key);
6505 ldi_ident_release(zfs_li);
6507 mutex_destroy(&zfs_share_lock);
6513 _info(struct modinfo *modinfop)
6515 return (mod_info(&modlinkage, modinfop));
6517 #endif /* illumos */
6519 static int zfs__init(void);
6520 static int zfs__fini(void);
6521 static void zfs_shutdown(void *, int);
6523 static eventhandler_tag zfs_shutdown_event_tag;
6526 #define ZFS_MIN_KSTACK_PAGES 4
6534 #if KSTACK_PAGES < ZFS_MIN_KSTACK_PAGES
6535 printf("ZFS NOTICE: KSTACK_PAGES is %d which could result in stack "
6536 "overflow panic!\nPlease consider adding "
6537 "'options KSTACK_PAGES=%d' to your kernel config\n", KSTACK_PAGES,
6538 ZFS_MIN_KSTACK_PAGES);
6541 zfs_root_token = root_mount_hold("ZFS");
6543 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
6545 spa_init(FREAD | FWRITE);
6550 tsd_create(&zfs_fsyncer_key, NULL);
6551 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
6552 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
6554 printf("ZFS storage pool version: features support (" SPA_VERSION_STRING ")\n");
6555 root_mount_rel(zfs_root_token);
6565 if (spa_busy() || zfs_busy() || zvol_busy() ||
6566 zio_injection_enabled) {
6575 tsd_destroy(&zfs_fsyncer_key);
6576 tsd_destroy(&rrw_tsd_key);
6577 tsd_destroy(&zfs_allow_log_key);
6579 mutex_destroy(&zfs_share_lock);
6585 zfs_shutdown(void *arg __unused, int howto __unused)
6589 * ZFS fini routines can not properly work in a panic-ed system.
6591 if (panicstr == NULL)
6597 zfs_modevent(module_t mod, int type, void *unused __unused)
6605 zfs_shutdown_event_tag = EVENTHANDLER_REGISTER(
6606 shutdown_post_sync, zfs_shutdown, NULL,
6607 SHUTDOWN_PRI_FIRST);
6611 if (err == 0 && zfs_shutdown_event_tag != NULL)
6612 EVENTHANDLER_DEREGISTER(shutdown_post_sync,
6613 zfs_shutdown_event_tag);
6620 return (EOPNOTSUPP);
6623 static moduledata_t zfs_mod = {
6628 DECLARE_MODULE(zfsctrl, zfs_mod, SI_SUB_VFS, SI_ORDER_ANY);
6629 MODULE_VERSION(zfsctrl, 1);
6630 MODULE_DEPEND(zfsctrl, opensolaris, 1, 1, 1);
6631 MODULE_DEPEND(zfsctrl, krpc, 1, 1, 1);
6632 MODULE_DEPEND(zfsctrl, acl_nfs4, 1, 1, 1);